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 #define OPTION_MPTCP_DSS BIT(11) 30 #define OPTION_MPTCP_FAIL BIT(12) 31 32 /* MPTCP option subtypes */ 33 #define MPTCPOPT_MP_CAPABLE 0 34 #define MPTCPOPT_MP_JOIN 1 35 #define MPTCPOPT_DSS 2 36 #define MPTCPOPT_ADD_ADDR 3 37 #define MPTCPOPT_RM_ADDR 4 38 #define MPTCPOPT_MP_PRIO 5 39 #define MPTCPOPT_MP_FAIL 6 40 #define MPTCPOPT_MP_FASTCLOSE 7 41 #define MPTCPOPT_RST 8 42 43 /* MPTCP suboption lengths */ 44 #define TCPOLEN_MPTCP_MPC_SYN 4 45 #define TCPOLEN_MPTCP_MPC_SYNACK 12 46 #define TCPOLEN_MPTCP_MPC_ACK 20 47 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 48 #define TCPOLEN_MPTCP_MPJ_SYN 12 49 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 50 #define TCPOLEN_MPTCP_MPJ_ACK 24 51 #define TCPOLEN_MPTCP_DSS_BASE 4 52 #define TCPOLEN_MPTCP_DSS_ACK32 4 53 #define TCPOLEN_MPTCP_DSS_ACK64 8 54 #define TCPOLEN_MPTCP_DSS_MAP32 10 55 #define TCPOLEN_MPTCP_DSS_MAP64 14 56 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 57 #define TCPOLEN_MPTCP_ADD_ADDR 16 58 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 59 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 60 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 61 #define TCPOLEN_MPTCP_ADD_ADDR6 28 62 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 63 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 64 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 65 #define TCPOLEN_MPTCP_PORT_LEN 2 66 #define TCPOLEN_MPTCP_PORT_ALIGN 2 67 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 68 #define TCPOLEN_MPTCP_PRIO 3 69 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 70 #define TCPOLEN_MPTCP_FASTCLOSE 12 71 #define TCPOLEN_MPTCP_RST 4 72 #define TCPOLEN_MPTCP_FAIL 12 73 74 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 75 76 /* MPTCP MP_JOIN flags */ 77 #define MPTCPOPT_BACKUP BIT(0) 78 #define MPTCPOPT_HMAC_LEN 20 79 #define MPTCPOPT_THMAC_LEN 8 80 81 /* MPTCP MP_CAPABLE flags */ 82 #define MPTCP_VERSION_MASK (0x0F) 83 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 84 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 85 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 86 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 87 #define MPTCP_CAP_FLAG_MASK (0x1F) 88 89 /* MPTCP DSS flags */ 90 #define MPTCP_DSS_DATA_FIN BIT(4) 91 #define MPTCP_DSS_DSN64 BIT(3) 92 #define MPTCP_DSS_HAS_MAP BIT(2) 93 #define MPTCP_DSS_ACK64 BIT(1) 94 #define MPTCP_DSS_HAS_ACK BIT(0) 95 #define MPTCP_DSS_FLAG_MASK (0x1F) 96 97 /* MPTCP ADD_ADDR flags */ 98 #define MPTCP_ADDR_ECHO BIT(0) 99 100 /* MPTCP MP_PRIO flags */ 101 #define MPTCP_PRIO_BKUP BIT(0) 102 103 /* MPTCP TCPRST flags */ 104 #define MPTCP_RST_TRANSIENT BIT(0) 105 106 /* MPTCP socket flags */ 107 #define MPTCP_DATA_READY 0 108 #define MPTCP_NOSPACE 1 109 #define MPTCP_WORK_RTX 2 110 #define MPTCP_WORK_EOF 3 111 #define MPTCP_FALLBACK_DONE 4 112 #define MPTCP_WORK_CLOSE_SUBFLOW 5 113 #define MPTCP_PUSH_PENDING 6 114 #define MPTCP_CLEAN_UNA 7 115 #define MPTCP_ERROR_REPORT 8 116 #define MPTCP_RETRANSMIT 9 117 #define MPTCP_WORK_SYNC_SETSOCKOPT 10 118 #define MPTCP_CONNECTED 11 119 120 static inline bool before64(__u64 seq1, __u64 seq2) 121 { 122 return (__s64)(seq1 - seq2) < 0; 123 } 124 125 #define after64(seq2, seq1) before64(seq1, seq2) 126 127 struct mptcp_options_received { 128 u64 sndr_key; 129 u64 rcvr_key; 130 u64 data_ack; 131 u64 data_seq; 132 u32 subflow_seq; 133 u16 data_len; 134 __sum16 csum; 135 u16 mp_capable : 1, 136 mp_join : 1, 137 fastclose : 1, 138 reset : 1, 139 dss : 1, 140 add_addr : 1, 141 rm_addr : 1, 142 mp_prio : 1, 143 mp_fail : 1, 144 echo : 1, 145 csum_reqd : 1, 146 backup : 1, 147 deny_join_id0 : 1; 148 u32 token; 149 u32 nonce; 150 u64 thmac; 151 u8 hmac[MPTCPOPT_HMAC_LEN]; 152 u8 join_id; 153 u8 use_map:1, 154 dsn64:1, 155 data_fin:1, 156 use_ack:1, 157 ack64:1, 158 mpc_map:1, 159 __unused:2; 160 struct mptcp_addr_info addr; 161 struct mptcp_rm_list rm_list; 162 u64 ahmac; 163 u8 reset_reason:4; 164 u8 reset_transient:1; 165 u64 fail_seq; 166 }; 167 168 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 169 { 170 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 171 ((nib & 0xF) << 8) | field); 172 } 173 174 enum mptcp_pm_status { 175 MPTCP_PM_ADD_ADDR_RECEIVED, 176 MPTCP_PM_ADD_ADDR_SEND_ACK, 177 MPTCP_PM_RM_ADDR_RECEIVED, 178 MPTCP_PM_ESTABLISHED, 179 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 180 MPTCP_PM_SUBFLOW_ESTABLISHED, 181 }; 182 183 enum mptcp_addr_signal_status { 184 MPTCP_ADD_ADDR_SIGNAL, 185 MPTCP_ADD_ADDR_ECHO, 186 MPTCP_RM_ADDR_SIGNAL, 187 }; 188 189 struct mptcp_pm_data { 190 struct mptcp_addr_info local; 191 struct mptcp_addr_info remote; 192 struct list_head anno_list; 193 194 spinlock_t lock; /*protects the whole PM data */ 195 196 u8 addr_signal; 197 bool server_side; 198 bool work_pending; 199 bool accept_addr; 200 bool accept_subflow; 201 bool remote_deny_join_id0; 202 u8 add_addr_signaled; 203 u8 add_addr_accepted; 204 u8 local_addr_used; 205 u8 subflows; 206 u8 status; 207 struct mptcp_rm_list rm_list_tx; 208 struct mptcp_rm_list rm_list_rx; 209 }; 210 211 struct mptcp_data_frag { 212 struct list_head list; 213 u64 data_seq; 214 u16 data_len; 215 u16 offset; 216 u16 overhead; 217 u16 already_sent; 218 struct page *page; 219 }; 220 221 /* MPTCP connection sock */ 222 struct mptcp_sock { 223 /* inet_connection_sock must be the first member */ 224 struct inet_connection_sock sk; 225 u64 local_key; 226 u64 remote_key; 227 u64 write_seq; 228 u64 snd_nxt; 229 u64 ack_seq; 230 u64 rcv_wnd_sent; 231 u64 rcv_data_fin_seq; 232 int wmem_reserved; 233 struct sock *last_snd; 234 int snd_burst; 235 int old_wspace; 236 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 237 * recovery related fields are under data_lock 238 * protection 239 */ 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 recovery; /* closing subflow write queue reinjected */ 247 bool can_ack; 248 bool fully_established; 249 bool rcv_data_fin; 250 bool snd_data_fin_enable; 251 bool rcv_fastclose; 252 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 253 bool csum_enabled; 254 spinlock_t join_list_lock; 255 struct work_struct work; 256 struct sk_buff *ooo_last_skb; 257 struct rb_root out_of_order_queue; 258 struct sk_buff_head receive_queue; 259 int tx_pending_data; 260 struct list_head conn_list; 261 struct list_head rtx_queue; 262 struct mptcp_data_frag *first_pending; 263 struct list_head join_list; 264 struct socket *subflow; /* outgoing connect/listener/!mp_capable */ 265 struct sock *first; 266 struct mptcp_pm_data pm; 267 struct { 268 u32 space; /* bytes copied in last measurement window */ 269 u32 copied; /* bytes copied in this measurement window */ 270 u64 time; /* start time of measurement window */ 271 u64 rtt_us; /* last maximum rtt of subflows */ 272 } rcvq_space; 273 274 u32 setsockopt_seq; 275 char ca_name[TCP_CA_NAME_MAX]; 276 }; 277 278 #define mptcp_lock_sock(___sk, cb) do { \ 279 struct sock *__sk = (___sk); /* silence macro reuse warning */ \ 280 might_sleep(); \ 281 spin_lock_bh(&__sk->sk_lock.slock); \ 282 if (__sk->sk_lock.owned) \ 283 __lock_sock(__sk); \ 284 cb; \ 285 __sk->sk_lock.owned = 1; \ 286 spin_unlock(&__sk->sk_lock.slock); \ 287 mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_); \ 288 local_bh_enable(); \ 289 } while (0) 290 291 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 292 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 293 294 #define mptcp_for_each_subflow(__msk, __subflow) \ 295 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 296 297 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 298 { 299 sock_owned_by_me((const struct sock *)msk); 300 } 301 302 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 303 { 304 return (struct mptcp_sock *)sk; 305 } 306 307 /* the msk socket don't use the backlog, also account for the bulk 308 * free memory 309 */ 310 static inline int __mptcp_rmem(const struct sock *sk) 311 { 312 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 313 } 314 315 static inline int __mptcp_space(const struct sock *sk) 316 { 317 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 318 } 319 320 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 321 { 322 const struct mptcp_sock *msk = mptcp_sk(sk); 323 324 return READ_ONCE(msk->first_pending); 325 } 326 327 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 328 { 329 struct mptcp_sock *msk = mptcp_sk(sk); 330 struct mptcp_data_frag *cur; 331 332 cur = msk->first_pending; 333 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 334 list_next_entry(cur, list); 335 } 336 337 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 338 { 339 struct mptcp_sock *msk = mptcp_sk(sk); 340 341 if (!msk->first_pending) 342 return NULL; 343 344 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 345 return NULL; 346 347 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 348 } 349 350 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 351 { 352 struct mptcp_sock *msk = mptcp_sk(sk); 353 354 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 355 return NULL; 356 357 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 358 } 359 360 struct csum_pseudo_header { 361 __be64 data_seq; 362 __be32 subflow_seq; 363 __be16 data_len; 364 __sum16 csum; 365 }; 366 367 struct mptcp_subflow_request_sock { 368 struct tcp_request_sock sk; 369 u16 mp_capable : 1, 370 mp_join : 1, 371 backup : 1, 372 csum_reqd : 1, 373 allow_join_id0 : 1; 374 u8 local_id; 375 u8 remote_id; 376 u64 local_key; 377 u64 idsn; 378 u32 token; 379 u32 ssn_offset; 380 u64 thmac; 381 u32 local_nonce; 382 u32 remote_nonce; 383 struct mptcp_sock *msk; 384 struct hlist_nulls_node token_node; 385 }; 386 387 static inline struct mptcp_subflow_request_sock * 388 mptcp_subflow_rsk(const struct request_sock *rsk) 389 { 390 return (struct mptcp_subflow_request_sock *)rsk; 391 } 392 393 enum mptcp_data_avail { 394 MPTCP_SUBFLOW_NODATA, 395 MPTCP_SUBFLOW_DATA_AVAIL, 396 }; 397 398 struct mptcp_delegated_action { 399 struct napi_struct napi; 400 struct list_head head; 401 }; 402 403 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 404 405 #define MPTCP_DELEGATE_SEND 0 406 407 /* MPTCP subflow context */ 408 struct mptcp_subflow_context { 409 struct list_head node;/* conn_list of subflows */ 410 u64 local_key; 411 u64 remote_key; 412 u64 idsn; 413 u64 map_seq; 414 u32 snd_isn; 415 u32 token; 416 u32 rel_write_seq; 417 u32 map_subflow_seq; 418 u32 ssn_offset; 419 u32 map_data_len; 420 __wsum map_data_csum; 421 u32 map_csum_len; 422 u32 request_mptcp : 1, /* send MP_CAPABLE */ 423 request_join : 1, /* send MP_JOIN */ 424 request_bkup : 1, 425 mp_capable : 1, /* remote is MPTCP capable */ 426 mp_join : 1, /* remote is JOINing */ 427 fully_established : 1, /* path validated */ 428 pm_notified : 1, /* PM hook called for established status */ 429 conn_finished : 1, 430 map_valid : 1, 431 map_csum_reqd : 1, 432 map_data_fin : 1, 433 mpc_map : 1, 434 backup : 1, 435 send_mp_prio : 1, 436 send_mp_fail : 1, 437 rx_eof : 1, 438 can_ack : 1, /* only after processing the remote a key */ 439 disposable : 1, /* ctx can be free at ulp release time */ 440 stale : 1; /* unable to snd/rcv data, do not use for xmit */ 441 enum mptcp_data_avail data_avail; 442 u32 remote_nonce; 443 u64 thmac; 444 u32 local_nonce; 445 u32 remote_token; 446 u8 hmac[MPTCPOPT_HMAC_LEN]; 447 u8 local_id; 448 u8 remote_id; 449 u8 reset_seen:1; 450 u8 reset_transient:1; 451 u8 reset_reason:4; 452 u8 stale_count; 453 454 long delegated_status; 455 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 456 457 u32 setsockopt_seq; 458 u32 stale_rcv_tstamp; 459 460 struct sock *tcp_sock; /* tcp sk backpointer */ 461 struct sock *conn; /* parent mptcp_sock */ 462 const struct inet_connection_sock_af_ops *icsk_af_ops; 463 void (*tcp_data_ready)(struct sock *sk); 464 void (*tcp_state_change)(struct sock *sk); 465 void (*tcp_write_space)(struct sock *sk); 466 void (*tcp_error_report)(struct sock *sk); 467 468 struct rcu_head rcu; 469 }; 470 471 static inline struct mptcp_subflow_context * 472 mptcp_subflow_ctx(const struct sock *sk) 473 { 474 struct inet_connection_sock *icsk = inet_csk(sk); 475 476 /* Use RCU on icsk_ulp_data only for sock diag code */ 477 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 478 } 479 480 static inline struct sock * 481 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 482 { 483 return subflow->tcp_sock; 484 } 485 486 static inline u64 487 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 488 { 489 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 490 subflow->ssn_offset - 491 subflow->map_subflow_seq; 492 } 493 494 static inline u64 495 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 496 { 497 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 498 } 499 500 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk, 501 struct mptcp_subflow_context *subflow) 502 { 503 sock_hold(mptcp_subflow_tcp_sock(subflow)); 504 spin_lock_bh(&msk->join_list_lock); 505 list_add_tail(&subflow->node, &msk->join_list); 506 spin_unlock_bh(&msk->join_list_lock); 507 } 508 509 void mptcp_subflow_process_delegated(struct sock *ssk); 510 511 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow) 512 { 513 struct mptcp_delegated_action *delegated; 514 bool schedule; 515 516 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 517 * ensures the below list check sees list updates done prior to status 518 * bit changes 519 */ 520 if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) { 521 /* still on delegated list from previous scheduling */ 522 if (!list_empty(&subflow->delegated_node)) 523 return; 524 525 /* the caller held the subflow bh socket lock */ 526 lockdep_assert_in_softirq(); 527 528 delegated = this_cpu_ptr(&mptcp_delegated_actions); 529 schedule = list_empty(&delegated->head); 530 list_add_tail(&subflow->delegated_node, &delegated->head); 531 sock_hold(mptcp_subflow_tcp_sock(subflow)); 532 if (schedule) 533 napi_schedule(&delegated->napi); 534 } 535 } 536 537 static inline struct mptcp_subflow_context * 538 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 539 { 540 struct mptcp_subflow_context *ret; 541 542 if (list_empty(&delegated->head)) 543 return NULL; 544 545 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 546 list_del_init(&ret->delegated_node); 547 return ret; 548 } 549 550 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 551 { 552 return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 553 } 554 555 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow) 556 { 557 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 558 * touching the status bit 559 */ 560 smp_wmb(); 561 clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 562 } 563 564 int mptcp_is_enabled(const struct net *net); 565 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 566 int mptcp_is_checksum_enabled(const struct net *net); 567 int mptcp_allow_join_id0(const struct net *net); 568 unsigned int mptcp_stale_loss_cnt(const struct net *net); 569 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 570 struct mptcp_options_received *mp_opt); 571 bool __mptcp_retransmit_pending_data(struct sock *sk); 572 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 573 bool mptcp_subflow_data_available(struct sock *sk); 574 void __init mptcp_subflow_init(void); 575 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 576 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 577 struct mptcp_subflow_context *subflow); 578 void mptcp_subflow_reset(struct sock *ssk); 579 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 580 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); 581 582 /* called with sk socket lock held */ 583 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 584 const struct mptcp_addr_info *remote); 585 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); 586 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 587 struct sockaddr_storage *addr, 588 unsigned short family); 589 590 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 591 { 592 struct sock *ssk = mptcp_subflow_tcp_sock(subflow); 593 594 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 595 if (subflow->request_join && !subflow->fully_established) 596 return false; 597 598 /* only send if our side has not closed yet */ 599 return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 600 } 601 602 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 603 604 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 605 606 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 607 struct mptcp_subflow_context *ctx) 608 { 609 sk->sk_data_ready = ctx->tcp_data_ready; 610 sk->sk_state_change = ctx->tcp_state_change; 611 sk->sk_write_space = ctx->tcp_write_space; 612 sk->sk_error_report = ctx->tcp_error_report; 613 614 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 615 } 616 617 static inline bool mptcp_has_another_subflow(struct sock *ssk) 618 { 619 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk), *tmp; 620 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 621 622 mptcp_for_each_subflow(msk, tmp) { 623 if (tmp != subflow) 624 return true; 625 } 626 627 return false; 628 } 629 630 void __init mptcp_proto_init(void); 631 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 632 int __init mptcp_proto_v6_init(void); 633 #endif 634 635 struct sock *mptcp_sk_clone(const struct sock *sk, 636 const struct mptcp_options_received *mp_opt, 637 struct request_sock *req); 638 void mptcp_get_options(const struct sock *sk, 639 const struct sk_buff *skb, 640 struct mptcp_options_received *mp_opt); 641 642 void mptcp_finish_connect(struct sock *sk); 643 void __mptcp_set_connected(struct sock *sk); 644 static inline bool mptcp_is_fully_established(struct sock *sk) 645 { 646 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 647 READ_ONCE(mptcp_sk(sk)->fully_established); 648 } 649 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 650 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 651 bool mptcp_finish_join(struct sock *sk); 652 bool mptcp_schedule_work(struct sock *sk); 653 int mptcp_setsockopt(struct sock *sk, int level, int optname, 654 sockptr_t optval, unsigned int optlen); 655 int mptcp_getsockopt(struct sock *sk, int level, int optname, 656 char __user *optval, int __user *option); 657 658 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 659 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 660 { 661 if (use_64bit) 662 return cur_seq; 663 664 return __mptcp_expand_seq(old_seq, cur_seq); 665 } 666 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 667 void __mptcp_data_acked(struct sock *sk); 668 void __mptcp_error_report(struct sock *sk); 669 void mptcp_subflow_eof(struct sock *sk); 670 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 671 void __mptcp_flush_join_list(struct mptcp_sock *msk); 672 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 673 { 674 return READ_ONCE(msk->snd_data_fin_enable) && 675 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 676 } 677 678 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 679 { 680 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 681 return false; 682 683 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 684 return true; 685 } 686 687 static inline void mptcp_write_space(struct sock *sk) 688 { 689 if (sk_stream_is_writeable(sk)) { 690 /* pairs with memory barrier in mptcp_poll */ 691 smp_mb(); 692 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 693 sk_stream_write_space(sk); 694 } 695 } 696 697 void mptcp_destroy_common(struct mptcp_sock *msk); 698 699 #define MPTCP_TOKEN_MAX_RETRIES 4 700 701 void __init mptcp_token_init(void); 702 static inline void mptcp_token_init_request(struct request_sock *req) 703 { 704 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 705 } 706 707 int mptcp_token_new_request(struct request_sock *req); 708 void mptcp_token_destroy_request(struct request_sock *req); 709 int mptcp_token_new_connect(struct sock *sk); 710 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 711 struct mptcp_sock *msk); 712 bool mptcp_token_exists(u32 token); 713 struct mptcp_sock *mptcp_token_get_sock(u32 token); 714 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 715 long *s_num); 716 void mptcp_token_destroy(struct mptcp_sock *msk); 717 718 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 719 720 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 721 722 void __init mptcp_pm_init(void); 723 void mptcp_pm_data_init(struct mptcp_sock *msk); 724 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 725 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 726 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 727 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp); 728 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 729 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 730 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 731 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id); 732 void mptcp_pm_add_addr_received(struct mptcp_sock *msk, 733 const struct mptcp_addr_info *addr); 734 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 735 struct mptcp_addr_info *addr); 736 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 737 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 738 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 739 const struct mptcp_rm_list *rm_list); 740 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 741 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 742 struct mptcp_addr_info *addr, 743 u8 bkup); 744 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 745 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 746 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 747 struct mptcp_pm_add_entry * 748 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 749 struct mptcp_addr_info *addr, bool check_id); 750 struct mptcp_pm_add_entry * 751 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk, 752 struct mptcp_addr_info *addr); 753 int mptcp_pm_get_flags_and_ifindex_by_id(struct net *net, unsigned int id, 754 u8 *flags, int *ifindex); 755 756 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 757 const struct mptcp_addr_info *addr, 758 bool echo); 759 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 760 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 761 762 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 763 const struct sock *ssk, gfp_t gfp); 764 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info); 765 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 766 767 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 768 { 769 return READ_ONCE(msk->pm.addr_signal) & 770 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 771 } 772 773 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 774 { 775 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 776 } 777 778 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 779 { 780 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 781 } 782 783 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 784 { 785 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 786 } 787 788 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 789 { 790 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 791 792 if (family == AF_INET6) 793 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 794 if (!echo) 795 len += MPTCPOPT_THMAC_LEN; 796 /* account for 2 trailing 'nop' options */ 797 if (port) 798 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 799 800 return len; 801 } 802 803 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 804 { 805 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 806 return -EINVAL; 807 808 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 809 } 810 811 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, struct sk_buff *skb, 812 unsigned int opt_size, unsigned int remaining, 813 struct mptcp_addr_info *addr, bool *echo, 814 bool *port, bool *drop_other_suboptions); 815 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 816 struct mptcp_rm_list *rm_list); 817 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 818 819 void __init mptcp_pm_nl_init(void); 820 void mptcp_pm_nl_data_init(struct mptcp_sock *msk); 821 void mptcp_pm_nl_work(struct mptcp_sock *msk); 822 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 823 const struct mptcp_rm_list *rm_list); 824 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 825 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk); 826 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk); 827 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk); 828 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk); 829 830 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 831 void mptcp_sockopt_sync_all(struct mptcp_sock *msk); 832 833 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 834 { 835 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 836 } 837 838 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 839 840 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 841 { 842 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 843 } 844 845 static inline bool mptcp_check_fallback(const struct sock *sk) 846 { 847 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 848 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 849 850 return __mptcp_check_fallback(msk); 851 } 852 853 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 854 { 855 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 856 pr_debug("TCP fallback already done (msk=%p)", msk); 857 return; 858 } 859 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 860 } 861 862 static inline void mptcp_do_fallback(struct sock *sk) 863 { 864 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 865 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 866 867 __mptcp_do_fallback(msk); 868 } 869 870 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 871 872 static inline bool subflow_simultaneous_connect(struct sock *sk) 873 { 874 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 875 struct sock *parent = subflow->conn; 876 877 return sk->sk_state == TCP_ESTABLISHED && 878 !mptcp_sk(parent)->pm.server_side && 879 !subflow->conn_finished; 880 } 881 882 #ifdef CONFIG_SYN_COOKIES 883 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 884 struct sk_buff *skb); 885 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 886 struct sk_buff *skb); 887 void __init mptcp_join_cookie_init(void); 888 #else 889 static inline void 890 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 891 struct sk_buff *skb) {} 892 static inline bool 893 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 894 struct sk_buff *skb) 895 { 896 return false; 897 } 898 899 static inline void mptcp_join_cookie_init(void) {} 900 #endif 901 902 #endif /* __MPTCP_PROTOCOL_H */ 903