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_HMAC_SHA256 BIT(0) 83 #define MPTCP_CAP_FLAG_MASK (0x3F) 84 85 /* MPTCP DSS flags */ 86 #define MPTCP_DSS_DATA_FIN BIT(4) 87 #define MPTCP_DSS_DSN64 BIT(3) 88 #define MPTCP_DSS_HAS_MAP BIT(2) 89 #define MPTCP_DSS_ACK64 BIT(1) 90 #define MPTCP_DSS_HAS_ACK BIT(0) 91 #define MPTCP_DSS_FLAG_MASK (0x1F) 92 93 /* MPTCP ADD_ADDR flags */ 94 #define MPTCP_ADDR_ECHO BIT(0) 95 96 /* MPTCP MP_PRIO flags */ 97 #define MPTCP_PRIO_BKUP BIT(0) 98 99 /* MPTCP TCPRST flags */ 100 #define MPTCP_RST_TRANSIENT BIT(0) 101 102 /* MPTCP socket flags */ 103 #define MPTCP_DATA_READY 0 104 #define MPTCP_NOSPACE 1 105 #define MPTCP_WORK_RTX 2 106 #define MPTCP_WORK_EOF 3 107 #define MPTCP_FALLBACK_DONE 4 108 #define MPTCP_WORK_CLOSE_SUBFLOW 5 109 #define MPTCP_PUSH_PENDING 6 110 #define MPTCP_CLEAN_UNA 7 111 #define MPTCP_ERROR_REPORT 8 112 #define MPTCP_RETRANSMIT 9 113 #define MPTCP_WORK_SYNC_SETSOCKOPT 10 114 115 static inline bool before64(__u64 seq1, __u64 seq2) 116 { 117 return (__s64)(seq1 - seq2) < 0; 118 } 119 120 #define after64(seq2, seq1) before64(seq1, seq2) 121 122 struct mptcp_options_received { 123 u64 sndr_key; 124 u64 rcvr_key; 125 u64 data_ack; 126 u64 data_seq; 127 u32 subflow_seq; 128 u16 data_len; 129 __sum16 csum; 130 u16 mp_capable : 1, 131 mp_join : 1, 132 fastclose : 1, 133 reset : 1, 134 dss : 1, 135 add_addr : 1, 136 rm_addr : 1, 137 mp_prio : 1, 138 echo : 1, 139 csum_reqd : 1, 140 backup : 1; 141 u32 token; 142 u32 nonce; 143 u64 thmac; 144 u8 hmac[MPTCPOPT_HMAC_LEN]; 145 u8 join_id; 146 u8 use_map:1, 147 dsn64:1, 148 data_fin:1, 149 use_ack:1, 150 ack64:1, 151 mpc_map:1, 152 __unused:2; 153 struct mptcp_addr_info addr; 154 struct mptcp_rm_list rm_list; 155 u64 ahmac; 156 u8 reset_reason:4; 157 u8 reset_transient:1; 158 }; 159 160 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 161 { 162 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 163 ((nib & 0xF) << 8) | field); 164 } 165 166 enum mptcp_pm_status { 167 MPTCP_PM_ADD_ADDR_RECEIVED, 168 MPTCP_PM_ADD_ADDR_SEND_ACK, 169 MPTCP_PM_RM_ADDR_RECEIVED, 170 MPTCP_PM_ESTABLISHED, 171 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 172 MPTCP_PM_SUBFLOW_ESTABLISHED, 173 }; 174 175 enum mptcp_addr_signal_status { 176 MPTCP_ADD_ADDR_SIGNAL, 177 MPTCP_ADD_ADDR_ECHO, 178 MPTCP_ADD_ADDR_IPV6, 179 MPTCP_ADD_ADDR_PORT, 180 MPTCP_RM_ADDR_SIGNAL, 181 }; 182 183 struct mptcp_pm_data { 184 struct mptcp_addr_info local; 185 struct mptcp_addr_info remote; 186 struct list_head anno_list; 187 188 spinlock_t lock; /*protects the whole PM data */ 189 190 u8 addr_signal; 191 bool server_side; 192 bool work_pending; 193 bool accept_addr; 194 bool accept_subflow; 195 u8 add_addr_signaled; 196 u8 add_addr_accepted; 197 u8 local_addr_used; 198 u8 subflows; 199 u8 status; 200 struct mptcp_rm_list rm_list_tx; 201 struct mptcp_rm_list rm_list_rx; 202 }; 203 204 struct mptcp_data_frag { 205 struct list_head list; 206 u64 data_seq; 207 u16 data_len; 208 u16 offset; 209 u16 overhead; 210 u16 already_sent; 211 struct page *page; 212 }; 213 214 /* MPTCP connection sock */ 215 struct mptcp_sock { 216 /* inet_connection_sock must be the first member */ 217 struct inet_connection_sock sk; 218 u64 local_key; 219 u64 remote_key; 220 u64 write_seq; 221 u64 snd_nxt; 222 u64 ack_seq; 223 u64 rcv_wnd_sent; 224 u64 rcv_data_fin_seq; 225 int wmem_reserved; 226 struct sock *last_snd; 227 int snd_burst; 228 int old_wspace; 229 u64 snd_una; 230 u64 wnd_end; 231 unsigned long timer_ival; 232 u32 token; 233 int rmem_released; 234 unsigned long flags; 235 bool can_ack; 236 bool fully_established; 237 bool rcv_data_fin; 238 bool snd_data_fin_enable; 239 bool rcv_fastclose; 240 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 241 bool csum_enabled; 242 spinlock_t join_list_lock; 243 struct sock *ack_hint; 244 struct work_struct work; 245 struct sk_buff *ooo_last_skb; 246 struct rb_root out_of_order_queue; 247 struct sk_buff_head receive_queue; 248 struct sk_buff_head skb_tx_cache; /* this is wmem accounted */ 249 int tx_pending_data; 250 int size_goal_cache; 251 struct list_head conn_list; 252 struct list_head rtx_queue; 253 struct mptcp_data_frag *first_pending; 254 struct list_head join_list; 255 struct socket *subflow; /* outgoing connect/listener/!mp_capable */ 256 struct sock *first; 257 struct mptcp_pm_data pm; 258 struct { 259 u32 space; /* bytes copied in last measurement window */ 260 u32 copied; /* bytes copied in this measurement window */ 261 u64 time; /* start time of measurement window */ 262 u64 rtt_us; /* last maximum rtt of subflows */ 263 } rcvq_space; 264 265 u32 setsockopt_seq; 266 char ca_name[TCP_CA_NAME_MAX]; 267 }; 268 269 #define mptcp_lock_sock(___sk, cb) do { \ 270 struct sock *__sk = (___sk); /* silence macro reuse warning */ \ 271 might_sleep(); \ 272 spin_lock_bh(&__sk->sk_lock.slock); \ 273 if (__sk->sk_lock.owned) \ 274 __lock_sock(__sk); \ 275 cb; \ 276 __sk->sk_lock.owned = 1; \ 277 spin_unlock(&__sk->sk_lock.slock); \ 278 mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_); \ 279 local_bh_enable(); \ 280 } while (0) 281 282 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 283 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 284 285 #define mptcp_for_each_subflow(__msk, __subflow) \ 286 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 287 288 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 289 { 290 sock_owned_by_me((const struct sock *)msk); 291 } 292 293 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 294 { 295 return (struct mptcp_sock *)sk; 296 } 297 298 static inline int __mptcp_space(const struct sock *sk) 299 { 300 return tcp_space(sk) + READ_ONCE(mptcp_sk(sk)->rmem_released); 301 } 302 303 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 304 { 305 const struct mptcp_sock *msk = mptcp_sk(sk); 306 307 return READ_ONCE(msk->first_pending); 308 } 309 310 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 311 { 312 struct mptcp_sock *msk = mptcp_sk(sk); 313 struct mptcp_data_frag *cur; 314 315 cur = msk->first_pending; 316 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 317 list_next_entry(cur, list); 318 } 319 320 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 321 { 322 struct mptcp_sock *msk = mptcp_sk(sk); 323 324 if (!msk->first_pending) 325 return NULL; 326 327 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 328 return NULL; 329 330 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 331 } 332 333 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 334 { 335 struct mptcp_sock *msk = mptcp_sk(sk); 336 337 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 338 return NULL; 339 340 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 341 } 342 343 struct csum_pseudo_header { 344 __be64 data_seq; 345 __be32 subflow_seq; 346 __be16 data_len; 347 __sum16 csum; 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 csum_reqd : 1; 356 u8 local_id; 357 u8 remote_id; 358 u64 local_key; 359 u64 idsn; 360 u32 token; 361 u32 ssn_offset; 362 u64 thmac; 363 u32 local_nonce; 364 u32 remote_nonce; 365 struct mptcp_sock *msk; 366 struct hlist_nulls_node token_node; 367 }; 368 369 static inline struct mptcp_subflow_request_sock * 370 mptcp_subflow_rsk(const struct request_sock *rsk) 371 { 372 return (struct mptcp_subflow_request_sock *)rsk; 373 } 374 375 enum mptcp_data_avail { 376 MPTCP_SUBFLOW_NODATA, 377 MPTCP_SUBFLOW_DATA_AVAIL, 378 MPTCP_SUBFLOW_OOO_DATA 379 }; 380 381 struct mptcp_delegated_action { 382 struct napi_struct napi; 383 struct list_head head; 384 }; 385 386 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 387 388 #define MPTCP_DELEGATE_SEND 0 389 390 /* MPTCP subflow context */ 391 struct mptcp_subflow_context { 392 struct list_head node;/* conn_list of subflows */ 393 u64 local_key; 394 u64 remote_key; 395 u64 idsn; 396 u64 map_seq; 397 u32 snd_isn; 398 u32 token; 399 u32 rel_write_seq; 400 u32 map_subflow_seq; 401 u32 ssn_offset; 402 u32 map_data_len; 403 __wsum map_data_csum; 404 u32 map_csum_len; 405 u32 request_mptcp : 1, /* send MP_CAPABLE */ 406 request_join : 1, /* send MP_JOIN */ 407 request_bkup : 1, 408 mp_capable : 1, /* remote is MPTCP capable */ 409 mp_join : 1, /* remote is JOINing */ 410 fully_established : 1, /* path validated */ 411 pm_notified : 1, /* PM hook called for established status */ 412 conn_finished : 1, 413 map_valid : 1, 414 map_csum_reqd : 1, 415 map_data_fin : 1, 416 mpc_map : 1, 417 backup : 1, 418 send_mp_prio : 1, 419 rx_eof : 1, 420 can_ack : 1, /* only after processing the remote a key */ 421 disposable : 1; /* ctx can be free at ulp release time */ 422 enum mptcp_data_avail data_avail; 423 u32 remote_nonce; 424 u64 thmac; 425 u32 local_nonce; 426 u32 remote_token; 427 u8 hmac[MPTCPOPT_HMAC_LEN]; 428 u8 local_id; 429 u8 remote_id; 430 u8 reset_seen:1; 431 u8 reset_transient:1; 432 u8 reset_reason:4; 433 434 long delegated_status; 435 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 436 437 u32 setsockopt_seq; 438 439 struct sock *tcp_sock; /* tcp sk backpointer */ 440 struct sock *conn; /* parent mptcp_sock */ 441 const struct inet_connection_sock_af_ops *icsk_af_ops; 442 void (*tcp_data_ready)(struct sock *sk); 443 void (*tcp_state_change)(struct sock *sk); 444 void (*tcp_write_space)(struct sock *sk); 445 void (*tcp_error_report)(struct sock *sk); 446 447 struct rcu_head rcu; 448 }; 449 450 static inline struct mptcp_subflow_context * 451 mptcp_subflow_ctx(const struct sock *sk) 452 { 453 struct inet_connection_sock *icsk = inet_csk(sk); 454 455 /* Use RCU on icsk_ulp_data only for sock diag code */ 456 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 457 } 458 459 static inline struct sock * 460 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 461 { 462 return subflow->tcp_sock; 463 } 464 465 static inline u64 466 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 467 { 468 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 469 subflow->ssn_offset - 470 subflow->map_subflow_seq; 471 } 472 473 static inline u64 474 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 475 { 476 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 477 } 478 479 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk, 480 struct mptcp_subflow_context *subflow) 481 { 482 sock_hold(mptcp_subflow_tcp_sock(subflow)); 483 spin_lock_bh(&msk->join_list_lock); 484 list_add_tail(&subflow->node, &msk->join_list); 485 spin_unlock_bh(&msk->join_list_lock); 486 } 487 488 void mptcp_subflow_process_delegated(struct sock *ssk); 489 490 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow) 491 { 492 struct mptcp_delegated_action *delegated; 493 bool schedule; 494 495 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 496 * ensures the below list check sees list updates done prior to status 497 * bit changes 498 */ 499 if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) { 500 /* still on delegated list from previous scheduling */ 501 if (!list_empty(&subflow->delegated_node)) 502 return; 503 504 /* the caller held the subflow bh socket lock */ 505 lockdep_assert_in_softirq(); 506 507 delegated = this_cpu_ptr(&mptcp_delegated_actions); 508 schedule = list_empty(&delegated->head); 509 list_add_tail(&subflow->delegated_node, &delegated->head); 510 sock_hold(mptcp_subflow_tcp_sock(subflow)); 511 if (schedule) 512 napi_schedule(&delegated->napi); 513 } 514 } 515 516 static inline struct mptcp_subflow_context * 517 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 518 { 519 struct mptcp_subflow_context *ret; 520 521 if (list_empty(&delegated->head)) 522 return NULL; 523 524 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 525 list_del_init(&ret->delegated_node); 526 return ret; 527 } 528 529 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 530 { 531 return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 532 } 533 534 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow) 535 { 536 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 537 * touching the status bit 538 */ 539 smp_wmb(); 540 clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 541 } 542 543 int mptcp_is_enabled(struct net *net); 544 unsigned int mptcp_get_add_addr_timeout(struct net *net); 545 int mptcp_is_checksum_enabled(struct net *net); 546 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 547 struct mptcp_options_received *mp_opt); 548 bool mptcp_subflow_data_available(struct sock *sk); 549 void __init mptcp_subflow_init(void); 550 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 551 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 552 struct mptcp_subflow_context *subflow); 553 void mptcp_subflow_reset(struct sock *ssk); 554 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 555 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); 556 557 /* called with sk socket lock held */ 558 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 559 const struct mptcp_addr_info *remote, 560 u8 flags, int ifindex); 561 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); 562 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 563 struct sockaddr_storage *addr, 564 unsigned short family); 565 566 static inline bool mptcp_subflow_active(struct mptcp_subflow_context *subflow) 567 { 568 struct sock *ssk = mptcp_subflow_tcp_sock(subflow); 569 570 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 571 if (subflow->request_join && !subflow->fully_established) 572 return false; 573 574 /* only send if our side has not closed yet */ 575 return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 576 } 577 578 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 579 struct mptcp_subflow_context *ctx) 580 { 581 sk->sk_data_ready = ctx->tcp_data_ready; 582 sk->sk_state_change = ctx->tcp_state_change; 583 sk->sk_write_space = ctx->tcp_write_space; 584 sk->sk_error_report = ctx->tcp_error_report; 585 586 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 587 } 588 589 void __init mptcp_proto_init(void); 590 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 591 int __init mptcp_proto_v6_init(void); 592 #endif 593 594 struct sock *mptcp_sk_clone(const struct sock *sk, 595 const struct mptcp_options_received *mp_opt, 596 struct request_sock *req); 597 void mptcp_get_options(const struct sock *sk, 598 const struct sk_buff *skb, 599 struct mptcp_options_received *mp_opt); 600 601 void mptcp_finish_connect(struct sock *sk); 602 static inline bool mptcp_is_fully_established(struct sock *sk) 603 { 604 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 605 READ_ONCE(mptcp_sk(sk)->fully_established); 606 } 607 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 608 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 609 bool mptcp_finish_join(struct sock *sk); 610 bool mptcp_schedule_work(struct sock *sk); 611 int mptcp_setsockopt(struct sock *sk, int level, int optname, 612 sockptr_t optval, unsigned int optlen); 613 int mptcp_getsockopt(struct sock *sk, int level, int optname, 614 char __user *optval, int __user *option); 615 616 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 617 void __mptcp_data_acked(struct sock *sk); 618 void __mptcp_error_report(struct sock *sk); 619 void mptcp_subflow_eof(struct sock *sk); 620 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 621 void __mptcp_flush_join_list(struct mptcp_sock *msk); 622 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 623 { 624 return READ_ONCE(msk->snd_data_fin_enable) && 625 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 626 } 627 628 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 629 { 630 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 631 return false; 632 633 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 634 return true; 635 } 636 637 static inline void mptcp_write_space(struct sock *sk) 638 { 639 if (sk_stream_is_writeable(sk)) { 640 /* pairs with memory barrier in mptcp_poll */ 641 smp_mb(); 642 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 643 sk_stream_write_space(sk); 644 } 645 } 646 647 void mptcp_destroy_common(struct mptcp_sock *msk); 648 649 #define MPTCP_TOKEN_MAX_RETRIES 4 650 651 void __init mptcp_token_init(void); 652 static inline void mptcp_token_init_request(struct request_sock *req) 653 { 654 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 655 } 656 657 int mptcp_token_new_request(struct request_sock *req); 658 void mptcp_token_destroy_request(struct request_sock *req); 659 int mptcp_token_new_connect(struct sock *sk); 660 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 661 struct mptcp_sock *msk); 662 bool mptcp_token_exists(u32 token); 663 struct mptcp_sock *mptcp_token_get_sock(u32 token); 664 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 665 long *s_num); 666 void mptcp_token_destroy(struct mptcp_sock *msk); 667 668 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 669 670 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 671 672 void __init mptcp_pm_init(void); 673 void mptcp_pm_data_init(struct mptcp_sock *msk); 674 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 675 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp); 676 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 677 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 678 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 679 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id); 680 void mptcp_pm_add_addr_received(struct mptcp_sock *msk, 681 const struct mptcp_addr_info *addr); 682 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 683 struct mptcp_addr_info *addr); 684 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 685 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 686 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 687 const struct mptcp_rm_list *rm_list); 688 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 689 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 690 struct mptcp_addr_info *addr, 691 u8 bkup); 692 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 693 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 694 struct mptcp_pm_add_entry * 695 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 696 struct mptcp_addr_info *addr, bool check_id); 697 struct mptcp_pm_add_entry * 698 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk, 699 struct mptcp_addr_info *addr); 700 701 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 702 const struct mptcp_addr_info *addr, 703 bool echo); 704 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 705 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 706 707 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 708 const struct sock *ssk, gfp_t gfp); 709 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info); 710 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 711 712 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 713 { 714 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 715 } 716 717 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 718 { 719 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 720 } 721 722 static inline bool mptcp_pm_should_add_signal_ipv6(struct mptcp_sock *msk) 723 { 724 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_IPV6); 725 } 726 727 static inline bool mptcp_pm_should_add_signal_port(struct mptcp_sock *msk) 728 { 729 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_PORT); 730 } 731 732 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 733 { 734 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 735 } 736 737 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 738 { 739 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 740 741 if (family == AF_INET6) 742 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 743 if (!echo) 744 len += MPTCPOPT_THMAC_LEN; 745 /* account for 2 trailing 'nop' options */ 746 if (port) 747 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 748 749 return len; 750 } 751 752 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 753 { 754 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 755 return -EINVAL; 756 757 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 758 } 759 760 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 761 struct mptcp_addr_info *saddr, bool *echo, bool *port); 762 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 763 struct mptcp_rm_list *rm_list); 764 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 765 766 void __init mptcp_pm_nl_init(void); 767 void mptcp_pm_nl_data_init(struct mptcp_sock *msk); 768 void mptcp_pm_nl_work(struct mptcp_sock *msk); 769 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 770 const struct mptcp_rm_list *rm_list); 771 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 772 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk); 773 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk); 774 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk); 775 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk); 776 777 int mptcp_setsockopt(struct sock *sk, int level, int optname, 778 sockptr_t optval, unsigned int optlen); 779 780 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 781 void mptcp_sockopt_sync_all(struct mptcp_sock *msk); 782 783 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 784 { 785 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 786 } 787 788 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 789 790 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 791 { 792 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 793 } 794 795 static inline bool mptcp_check_fallback(const struct sock *sk) 796 { 797 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 798 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 799 800 return __mptcp_check_fallback(msk); 801 } 802 803 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 804 { 805 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 806 pr_debug("TCP fallback already done (msk=%p)", msk); 807 return; 808 } 809 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 810 } 811 812 static inline void mptcp_do_fallback(struct sock *sk) 813 { 814 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 815 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 816 817 __mptcp_do_fallback(msk); 818 } 819 820 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 821 822 static inline bool subflow_simultaneous_connect(struct sock *sk) 823 { 824 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 825 struct sock *parent = subflow->conn; 826 827 return sk->sk_state == TCP_ESTABLISHED && 828 !mptcp_sk(parent)->pm.server_side && 829 !subflow->conn_finished; 830 } 831 832 #ifdef CONFIG_SYN_COOKIES 833 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 834 struct sk_buff *skb); 835 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 836 struct sk_buff *skb); 837 void __init mptcp_join_cookie_init(void); 838 #else 839 static inline void 840 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 841 struct sk_buff *skb) {} 842 static inline bool 843 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 844 struct sk_buff *skb) 845 { 846 return false; 847 } 848 849 static inline void mptcp_join_cookie_init(void) {} 850 #endif 851 852 #endif /* __MPTCP_PROTOCOL_H */ 853