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 snd_una; 234 u64 wnd_end; 235 unsigned long timer_ival; 236 u32 token; 237 int rmem_released; 238 unsigned long flags; 239 bool can_ack; 240 bool fully_established; 241 bool rcv_data_fin; 242 bool snd_data_fin_enable; 243 bool rcv_fastclose; 244 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 245 bool csum_enabled; 246 spinlock_t join_list_lock; 247 struct work_struct work; 248 struct sk_buff *ooo_last_skb; 249 struct rb_root out_of_order_queue; 250 struct sk_buff_head receive_queue; 251 int tx_pending_data; 252 struct list_head conn_list; 253 struct list_head rtx_queue; 254 struct mptcp_data_frag *first_pending; 255 struct list_head join_list; 256 struct socket *subflow; /* outgoing connect/listener/!mp_capable */ 257 struct sock *first; 258 struct mptcp_pm_data pm; 259 struct { 260 u32 space; /* bytes copied in last measurement window */ 261 u32 copied; /* bytes copied in this measurement window */ 262 u64 time; /* start time of measurement window */ 263 u64 rtt_us; /* last maximum rtt of subflows */ 264 } rcvq_space; 265 266 u32 setsockopt_seq; 267 char ca_name[TCP_CA_NAME_MAX]; 268 }; 269 270 #define mptcp_lock_sock(___sk, cb) do { \ 271 struct sock *__sk = (___sk); /* silence macro reuse warning */ \ 272 might_sleep(); \ 273 spin_lock_bh(&__sk->sk_lock.slock); \ 274 if (__sk->sk_lock.owned) \ 275 __lock_sock(__sk); \ 276 cb; \ 277 __sk->sk_lock.owned = 1; \ 278 spin_unlock(&__sk->sk_lock.slock); \ 279 mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_); \ 280 local_bh_enable(); \ 281 } while (0) 282 283 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 284 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 285 286 #define mptcp_for_each_subflow(__msk, __subflow) \ 287 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 288 289 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 290 { 291 sock_owned_by_me((const struct sock *)msk); 292 } 293 294 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 295 { 296 return (struct mptcp_sock *)sk; 297 } 298 299 /* the msk socket don't use the backlog, also account for the bulk 300 * free memory 301 */ 302 static inline int __mptcp_rmem(const struct sock *sk) 303 { 304 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 305 } 306 307 static inline int __mptcp_space(const struct sock *sk) 308 { 309 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 310 } 311 312 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 313 { 314 const struct mptcp_sock *msk = mptcp_sk(sk); 315 316 return READ_ONCE(msk->first_pending); 317 } 318 319 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 320 { 321 struct mptcp_sock *msk = mptcp_sk(sk); 322 struct mptcp_data_frag *cur; 323 324 cur = msk->first_pending; 325 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 326 list_next_entry(cur, list); 327 } 328 329 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 330 { 331 struct mptcp_sock *msk = mptcp_sk(sk); 332 333 if (!msk->first_pending) 334 return NULL; 335 336 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 337 return NULL; 338 339 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 340 } 341 342 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 343 { 344 struct mptcp_sock *msk = mptcp_sk(sk); 345 346 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 347 return NULL; 348 349 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 350 } 351 352 struct csum_pseudo_header { 353 __be64 data_seq; 354 __be32 subflow_seq; 355 __be16 data_len; 356 __sum16 csum; 357 }; 358 359 struct mptcp_subflow_request_sock { 360 struct tcp_request_sock sk; 361 u16 mp_capable : 1, 362 mp_join : 1, 363 backup : 1, 364 csum_reqd : 1, 365 allow_join_id0 : 1; 366 u8 local_id; 367 u8 remote_id; 368 u64 local_key; 369 u64 idsn; 370 u32 token; 371 u32 ssn_offset; 372 u64 thmac; 373 u32 local_nonce; 374 u32 remote_nonce; 375 struct mptcp_sock *msk; 376 struct hlist_nulls_node token_node; 377 }; 378 379 static inline struct mptcp_subflow_request_sock * 380 mptcp_subflow_rsk(const struct request_sock *rsk) 381 { 382 return (struct mptcp_subflow_request_sock *)rsk; 383 } 384 385 enum mptcp_data_avail { 386 MPTCP_SUBFLOW_NODATA, 387 MPTCP_SUBFLOW_DATA_AVAIL, 388 }; 389 390 struct mptcp_delegated_action { 391 struct napi_struct napi; 392 struct list_head head; 393 }; 394 395 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 396 397 #define MPTCP_DELEGATE_SEND 0 398 399 /* MPTCP subflow context */ 400 struct mptcp_subflow_context { 401 struct list_head node;/* conn_list of subflows */ 402 u64 local_key; 403 u64 remote_key; 404 u64 idsn; 405 u64 map_seq; 406 u32 snd_isn; 407 u32 token; 408 u32 rel_write_seq; 409 u32 map_subflow_seq; 410 u32 ssn_offset; 411 u32 map_data_len; 412 __wsum map_data_csum; 413 u32 map_csum_len; 414 u32 request_mptcp : 1, /* send MP_CAPABLE */ 415 request_join : 1, /* send MP_JOIN */ 416 request_bkup : 1, 417 mp_capable : 1, /* remote is MPTCP capable */ 418 mp_join : 1, /* remote is JOINing */ 419 fully_established : 1, /* path validated */ 420 pm_notified : 1, /* PM hook called for established status */ 421 conn_finished : 1, 422 map_valid : 1, 423 map_csum_reqd : 1, 424 map_data_fin : 1, 425 mpc_map : 1, 426 backup : 1, 427 send_mp_prio : 1, 428 rx_eof : 1, 429 can_ack : 1, /* only after processing the remote a key */ 430 disposable : 1; /* ctx can be free at ulp release time */ 431 enum mptcp_data_avail data_avail; 432 u32 remote_nonce; 433 u64 thmac; 434 u32 local_nonce; 435 u32 remote_token; 436 u8 hmac[MPTCPOPT_HMAC_LEN]; 437 u8 local_id; 438 u8 remote_id; 439 u8 reset_seen:1; 440 u8 reset_transient:1; 441 u8 reset_reason:4; 442 443 long delegated_status; 444 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 445 446 u32 setsockopt_seq; 447 448 struct sock *tcp_sock; /* tcp sk backpointer */ 449 struct sock *conn; /* parent mptcp_sock */ 450 const struct inet_connection_sock_af_ops *icsk_af_ops; 451 void (*tcp_data_ready)(struct sock *sk); 452 void (*tcp_state_change)(struct sock *sk); 453 void (*tcp_write_space)(struct sock *sk); 454 void (*tcp_error_report)(struct sock *sk); 455 456 struct rcu_head rcu; 457 }; 458 459 static inline struct mptcp_subflow_context * 460 mptcp_subflow_ctx(const struct sock *sk) 461 { 462 struct inet_connection_sock *icsk = inet_csk(sk); 463 464 /* Use RCU on icsk_ulp_data only for sock diag code */ 465 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 466 } 467 468 static inline struct sock * 469 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 470 { 471 return subflow->tcp_sock; 472 } 473 474 static inline u64 475 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 476 { 477 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 478 subflow->ssn_offset - 479 subflow->map_subflow_seq; 480 } 481 482 static inline u64 483 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 484 { 485 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 486 } 487 488 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk, 489 struct mptcp_subflow_context *subflow) 490 { 491 sock_hold(mptcp_subflow_tcp_sock(subflow)); 492 spin_lock_bh(&msk->join_list_lock); 493 list_add_tail(&subflow->node, &msk->join_list); 494 spin_unlock_bh(&msk->join_list_lock); 495 } 496 497 void mptcp_subflow_process_delegated(struct sock *ssk); 498 499 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow) 500 { 501 struct mptcp_delegated_action *delegated; 502 bool schedule; 503 504 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 505 * ensures the below list check sees list updates done prior to status 506 * bit changes 507 */ 508 if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) { 509 /* still on delegated list from previous scheduling */ 510 if (!list_empty(&subflow->delegated_node)) 511 return; 512 513 /* the caller held the subflow bh socket lock */ 514 lockdep_assert_in_softirq(); 515 516 delegated = this_cpu_ptr(&mptcp_delegated_actions); 517 schedule = list_empty(&delegated->head); 518 list_add_tail(&subflow->delegated_node, &delegated->head); 519 sock_hold(mptcp_subflow_tcp_sock(subflow)); 520 if (schedule) 521 napi_schedule(&delegated->napi); 522 } 523 } 524 525 static inline struct mptcp_subflow_context * 526 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 527 { 528 struct mptcp_subflow_context *ret; 529 530 if (list_empty(&delegated->head)) 531 return NULL; 532 533 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 534 list_del_init(&ret->delegated_node); 535 return ret; 536 } 537 538 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 539 { 540 return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 541 } 542 543 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow) 544 { 545 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 546 * touching the status bit 547 */ 548 smp_wmb(); 549 clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 550 } 551 552 int mptcp_is_enabled(struct net *net); 553 unsigned int mptcp_get_add_addr_timeout(struct net *net); 554 int mptcp_is_checksum_enabled(struct net *net); 555 int mptcp_allow_join_id0(struct net *net); 556 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 557 struct mptcp_options_received *mp_opt); 558 bool mptcp_subflow_data_available(struct sock *sk); 559 void __init mptcp_subflow_init(void); 560 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 561 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 562 struct mptcp_subflow_context *subflow); 563 void mptcp_subflow_reset(struct sock *ssk); 564 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 565 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); 566 567 /* called with sk socket lock held */ 568 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 569 const struct mptcp_addr_info *remote, 570 u8 flags, int ifindex); 571 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); 572 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 573 struct sockaddr_storage *addr, 574 unsigned short family); 575 576 static inline bool mptcp_subflow_active(struct mptcp_subflow_context *subflow) 577 { 578 struct sock *ssk = mptcp_subflow_tcp_sock(subflow); 579 580 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 581 if (subflow->request_join && !subflow->fully_established) 582 return false; 583 584 /* only send if our side has not closed yet */ 585 return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 586 } 587 588 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 589 struct mptcp_subflow_context *ctx) 590 { 591 sk->sk_data_ready = ctx->tcp_data_ready; 592 sk->sk_state_change = ctx->tcp_state_change; 593 sk->sk_write_space = ctx->tcp_write_space; 594 sk->sk_error_report = ctx->tcp_error_report; 595 596 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 597 } 598 599 void __init mptcp_proto_init(void); 600 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 601 int __init mptcp_proto_v6_init(void); 602 #endif 603 604 struct sock *mptcp_sk_clone(const struct sock *sk, 605 const struct mptcp_options_received *mp_opt, 606 struct request_sock *req); 607 void mptcp_get_options(const struct sock *sk, 608 const struct sk_buff *skb, 609 struct mptcp_options_received *mp_opt); 610 611 void mptcp_finish_connect(struct sock *sk); 612 void __mptcp_set_connected(struct sock *sk); 613 static inline bool mptcp_is_fully_established(struct sock *sk) 614 { 615 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 616 READ_ONCE(mptcp_sk(sk)->fully_established); 617 } 618 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 619 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 620 bool mptcp_finish_join(struct sock *sk); 621 bool mptcp_schedule_work(struct sock *sk); 622 int mptcp_setsockopt(struct sock *sk, int level, int optname, 623 sockptr_t optval, unsigned int optlen); 624 int mptcp_getsockopt(struct sock *sk, int level, int optname, 625 char __user *optval, int __user *option); 626 627 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 628 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 629 { 630 if (use_64bit) 631 return cur_seq; 632 633 return __mptcp_expand_seq(old_seq, cur_seq); 634 } 635 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 636 void __mptcp_data_acked(struct sock *sk); 637 void __mptcp_error_report(struct sock *sk); 638 void mptcp_subflow_eof(struct sock *sk); 639 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 640 void __mptcp_flush_join_list(struct mptcp_sock *msk); 641 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 642 { 643 return READ_ONCE(msk->snd_data_fin_enable) && 644 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 645 } 646 647 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 648 { 649 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 650 return false; 651 652 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 653 return true; 654 } 655 656 static inline void mptcp_write_space(struct sock *sk) 657 { 658 if (sk_stream_is_writeable(sk)) { 659 /* pairs with memory barrier in mptcp_poll */ 660 smp_mb(); 661 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 662 sk_stream_write_space(sk); 663 } 664 } 665 666 void mptcp_destroy_common(struct mptcp_sock *msk); 667 668 #define MPTCP_TOKEN_MAX_RETRIES 4 669 670 void __init mptcp_token_init(void); 671 static inline void mptcp_token_init_request(struct request_sock *req) 672 { 673 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 674 } 675 676 int mptcp_token_new_request(struct request_sock *req); 677 void mptcp_token_destroy_request(struct request_sock *req); 678 int mptcp_token_new_connect(struct sock *sk); 679 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 680 struct mptcp_sock *msk); 681 bool mptcp_token_exists(u32 token); 682 struct mptcp_sock *mptcp_token_get_sock(u32 token); 683 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 684 long *s_num); 685 void mptcp_token_destroy(struct mptcp_sock *msk); 686 687 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 688 689 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 690 691 void __init mptcp_pm_init(void); 692 void mptcp_pm_data_init(struct mptcp_sock *msk); 693 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 694 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp); 695 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 696 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 697 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 698 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id); 699 void mptcp_pm_add_addr_received(struct mptcp_sock *msk, 700 const struct mptcp_addr_info *addr); 701 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 702 struct mptcp_addr_info *addr); 703 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 704 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 705 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 706 const struct mptcp_rm_list *rm_list); 707 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 708 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 709 struct mptcp_addr_info *addr, 710 u8 bkup); 711 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 712 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 713 struct mptcp_pm_add_entry * 714 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 715 struct mptcp_addr_info *addr, bool check_id); 716 struct mptcp_pm_add_entry * 717 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk, 718 struct mptcp_addr_info *addr); 719 720 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 721 const struct mptcp_addr_info *addr, 722 bool echo); 723 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 724 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 725 726 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 727 const struct sock *ssk, gfp_t gfp); 728 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info); 729 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 730 731 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 732 { 733 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 734 } 735 736 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 737 { 738 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 739 } 740 741 static inline bool mptcp_pm_should_add_signal_ipv6(struct mptcp_sock *msk) 742 { 743 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_IPV6); 744 } 745 746 static inline bool mptcp_pm_should_add_signal_port(struct mptcp_sock *msk) 747 { 748 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_PORT); 749 } 750 751 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 752 { 753 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 754 } 755 756 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 757 { 758 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 759 760 if (family == AF_INET6) 761 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 762 if (!echo) 763 len += MPTCPOPT_THMAC_LEN; 764 /* account for 2 trailing 'nop' options */ 765 if (port) 766 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 767 768 return len; 769 } 770 771 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 772 { 773 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 774 return -EINVAL; 775 776 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 777 } 778 779 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 780 struct mptcp_addr_info *saddr, bool *echo, bool *port); 781 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 782 struct mptcp_rm_list *rm_list); 783 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 784 785 void __init mptcp_pm_nl_init(void); 786 void mptcp_pm_nl_data_init(struct mptcp_sock *msk); 787 void mptcp_pm_nl_work(struct mptcp_sock *msk); 788 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 789 const struct mptcp_rm_list *rm_list); 790 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 791 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk); 792 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk); 793 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk); 794 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk); 795 796 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 797 void mptcp_sockopt_sync_all(struct mptcp_sock *msk); 798 799 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 800 { 801 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 802 } 803 804 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 805 806 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 807 { 808 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 809 } 810 811 static inline bool mptcp_check_fallback(const struct sock *sk) 812 { 813 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 814 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 815 816 return __mptcp_check_fallback(msk); 817 } 818 819 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 820 { 821 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 822 pr_debug("TCP fallback already done (msk=%p)", msk); 823 return; 824 } 825 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 826 } 827 828 static inline void mptcp_do_fallback(struct sock *sk) 829 { 830 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 831 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 832 833 __mptcp_do_fallback(msk); 834 } 835 836 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 837 838 static inline bool subflow_simultaneous_connect(struct sock *sk) 839 { 840 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 841 struct sock *parent = subflow->conn; 842 843 return sk->sk_state == TCP_ESTABLISHED && 844 !mptcp_sk(parent)->pm.server_side && 845 !subflow->conn_finished; 846 } 847 848 #ifdef CONFIG_SYN_COOKIES 849 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 850 struct sk_buff *skb); 851 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 852 struct sk_buff *skb); 853 void __init mptcp_join_cookie_init(void); 854 #else 855 static inline void 856 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 857 struct sk_buff *skb) {} 858 static inline bool 859 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 860 struct sk_buff *skb) 861 { 862 return false; 863 } 864 865 static inline void mptcp_join_cookie_init(void) {} 866 #endif 867 868 #endif /* __MPTCP_PROTOCOL_H */ 869