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 #define OPTION_MPTCP_CSUMREQD BIT(13) 33 34 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \ 35 OPTION_MPTCP_MPC_ACK) 36 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \ 37 OPTION_MPTCP_MPJ_ACK) 38 39 /* MPTCP option subtypes */ 40 #define MPTCPOPT_MP_CAPABLE 0 41 #define MPTCPOPT_MP_JOIN 1 42 #define MPTCPOPT_DSS 2 43 #define MPTCPOPT_ADD_ADDR 3 44 #define MPTCPOPT_RM_ADDR 4 45 #define MPTCPOPT_MP_PRIO 5 46 #define MPTCPOPT_MP_FAIL 6 47 #define MPTCPOPT_MP_FASTCLOSE 7 48 #define MPTCPOPT_RST 8 49 50 /* MPTCP suboption lengths */ 51 #define TCPOLEN_MPTCP_MPC_SYN 4 52 #define TCPOLEN_MPTCP_MPC_SYNACK 12 53 #define TCPOLEN_MPTCP_MPC_ACK 20 54 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 55 #define TCPOLEN_MPTCP_MPJ_SYN 12 56 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 57 #define TCPOLEN_MPTCP_MPJ_ACK 24 58 #define TCPOLEN_MPTCP_DSS_BASE 4 59 #define TCPOLEN_MPTCP_DSS_ACK32 4 60 #define TCPOLEN_MPTCP_DSS_ACK64 8 61 #define TCPOLEN_MPTCP_DSS_MAP32 10 62 #define TCPOLEN_MPTCP_DSS_MAP64 14 63 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 64 #define TCPOLEN_MPTCP_ADD_ADDR 16 65 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 66 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 68 #define TCPOLEN_MPTCP_ADD_ADDR6 28 69 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 70 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 72 #define TCPOLEN_MPTCP_PORT_LEN 2 73 #define TCPOLEN_MPTCP_PORT_ALIGN 2 74 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 75 #define TCPOLEN_MPTCP_PRIO 3 76 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 77 #define TCPOLEN_MPTCP_FASTCLOSE 12 78 #define TCPOLEN_MPTCP_RST 4 79 #define TCPOLEN_MPTCP_FAIL 12 80 81 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 82 83 /* MPTCP MP_JOIN flags */ 84 #define MPTCPOPT_BACKUP BIT(0) 85 #define MPTCPOPT_HMAC_LEN 20 86 #define MPTCPOPT_THMAC_LEN 8 87 88 /* MPTCP MP_CAPABLE flags */ 89 #define MPTCP_VERSION_MASK (0x0F) 90 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 91 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 92 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 93 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 94 #define MPTCP_CAP_FLAG_MASK (0x1F) 95 96 /* MPTCP DSS flags */ 97 #define MPTCP_DSS_DATA_FIN BIT(4) 98 #define MPTCP_DSS_DSN64 BIT(3) 99 #define MPTCP_DSS_HAS_MAP BIT(2) 100 #define MPTCP_DSS_ACK64 BIT(1) 101 #define MPTCP_DSS_HAS_ACK BIT(0) 102 #define MPTCP_DSS_FLAG_MASK (0x1F) 103 104 /* MPTCP ADD_ADDR flags */ 105 #define MPTCP_ADDR_ECHO BIT(0) 106 107 /* MPTCP MP_PRIO flags */ 108 #define MPTCP_PRIO_BKUP BIT(0) 109 110 /* MPTCP TCPRST flags */ 111 #define MPTCP_RST_TRANSIENT BIT(0) 112 113 /* MPTCP socket flags */ 114 #define MPTCP_DATA_READY 0 115 #define MPTCP_NOSPACE 1 116 #define MPTCP_WORK_RTX 2 117 #define MPTCP_WORK_EOF 3 118 #define MPTCP_FALLBACK_DONE 4 119 #define MPTCP_WORK_CLOSE_SUBFLOW 5 120 #define MPTCP_PUSH_PENDING 6 121 #define MPTCP_CLEAN_UNA 7 122 #define MPTCP_ERROR_REPORT 8 123 #define MPTCP_RETRANSMIT 9 124 #define MPTCP_WORK_SYNC_SETSOCKOPT 10 125 #define MPTCP_CONNECTED 11 126 127 static inline bool before64(__u64 seq1, __u64 seq2) 128 { 129 return (__s64)(seq1 - seq2) < 0; 130 } 131 132 #define after64(seq2, seq1) before64(seq1, seq2) 133 134 struct mptcp_options_received { 135 u64 sndr_key; 136 u64 rcvr_key; 137 u64 data_ack; 138 u64 data_seq; 139 u32 subflow_seq; 140 u16 data_len; 141 __sum16 csum; 142 u16 suboptions; 143 u32 token; 144 u32 nonce; 145 u16 use_map:1, 146 dsn64:1, 147 data_fin:1, 148 use_ack:1, 149 ack64:1, 150 mpc_map:1, 151 reset_reason:4, 152 reset_transient:1, 153 echo:1, 154 backup:1, 155 deny_join_id0:1, 156 __unused:2; 157 u8 join_id; 158 u64 thmac; 159 u8 hmac[MPTCPOPT_HMAC_LEN]; 160 struct mptcp_addr_info addr; 161 struct mptcp_rm_list rm_list; 162 u64 ahmac; 163 u64 fail_seq; 164 }; 165 166 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 167 { 168 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 169 ((nib & 0xF) << 8) | field); 170 } 171 172 enum mptcp_pm_status { 173 MPTCP_PM_ADD_ADDR_RECEIVED, 174 MPTCP_PM_ADD_ADDR_SEND_ACK, 175 MPTCP_PM_RM_ADDR_RECEIVED, 176 MPTCP_PM_ESTABLISHED, 177 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 178 MPTCP_PM_SUBFLOW_ESTABLISHED, 179 }; 180 181 enum mptcp_addr_signal_status { 182 MPTCP_ADD_ADDR_SIGNAL, 183 MPTCP_ADD_ADDR_ECHO, 184 MPTCP_RM_ADDR_SIGNAL, 185 }; 186 187 struct mptcp_pm_data { 188 struct mptcp_addr_info local; 189 struct mptcp_addr_info remote; 190 struct list_head anno_list; 191 192 spinlock_t lock; /*protects the whole PM data */ 193 194 u8 addr_signal; 195 bool server_side; 196 bool work_pending; 197 bool accept_addr; 198 bool accept_subflow; 199 bool remote_deny_join_id0; 200 u8 add_addr_signaled; 201 u8 add_addr_accepted; 202 u8 local_addr_used; 203 u8 subflows; 204 u8 status; 205 struct mptcp_rm_list rm_list_tx; 206 struct mptcp_rm_list rm_list_rx; 207 }; 208 209 struct mptcp_data_frag { 210 struct list_head list; 211 u64 data_seq; 212 u16 data_len; 213 u16 offset; 214 u16 overhead; 215 u16 already_sent; 216 struct page *page; 217 }; 218 219 /* MPTCP connection sock */ 220 struct mptcp_sock { 221 /* inet_connection_sock must be the first member */ 222 struct inet_connection_sock sk; 223 u64 local_key; 224 u64 remote_key; 225 u64 write_seq; 226 u64 snd_nxt; 227 u64 ack_seq; 228 u64 rcv_wnd_sent; 229 u64 rcv_data_fin_seq; 230 int rmem_fwd_alloc; 231 struct sock *last_snd; 232 int snd_burst; 233 int old_wspace; 234 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 235 * recovery related fields are under data_lock 236 * protection 237 */ 238 u64 snd_una; 239 u64 wnd_end; 240 unsigned long timer_ival; 241 u32 token; 242 int rmem_released; 243 unsigned long flags; 244 bool recovery; /* closing subflow write queue reinjected */ 245 bool can_ack; 246 bool fully_established; 247 bool rcv_data_fin; 248 bool snd_data_fin_enable; 249 bool rcv_fastclose; 250 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 251 bool csum_enabled; 252 spinlock_t join_list_lock; 253 struct work_struct work; 254 struct sk_buff *ooo_last_skb; 255 struct rb_root out_of_order_queue; 256 struct sk_buff_head receive_queue; 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_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 276 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 277 278 #define mptcp_for_each_subflow(__msk, __subflow) \ 279 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 280 281 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 282 { 283 sock_owned_by_me((const struct sock *)msk); 284 } 285 286 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 287 { 288 return (struct mptcp_sock *)sk; 289 } 290 291 /* the msk socket don't use the backlog, also account for the bulk 292 * free memory 293 */ 294 static inline int __mptcp_rmem(const struct sock *sk) 295 { 296 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 297 } 298 299 static inline int __mptcp_space(const struct sock *sk) 300 { 301 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 302 } 303 304 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 305 { 306 const struct mptcp_sock *msk = mptcp_sk(sk); 307 308 return READ_ONCE(msk->first_pending); 309 } 310 311 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 312 { 313 struct mptcp_sock *msk = mptcp_sk(sk); 314 struct mptcp_data_frag *cur; 315 316 cur = msk->first_pending; 317 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 318 list_next_entry(cur, list); 319 } 320 321 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 322 { 323 struct mptcp_sock *msk = mptcp_sk(sk); 324 325 if (!msk->first_pending) 326 return NULL; 327 328 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 329 return NULL; 330 331 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 332 } 333 334 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 335 { 336 struct mptcp_sock *msk = mptcp_sk(sk); 337 338 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 339 return NULL; 340 341 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 342 } 343 344 struct csum_pseudo_header { 345 __be64 data_seq; 346 __be32 subflow_seq; 347 __be16 data_len; 348 __sum16 csum; 349 }; 350 351 struct mptcp_subflow_request_sock { 352 struct tcp_request_sock sk; 353 u16 mp_capable : 1, 354 mp_join : 1, 355 backup : 1, 356 csum_reqd : 1, 357 allow_join_id0 : 1; 358 u8 local_id; 359 u8 remote_id; 360 u64 local_key; 361 u64 idsn; 362 u32 token; 363 u32 ssn_offset; 364 u64 thmac; 365 u32 local_nonce; 366 u32 remote_nonce; 367 struct mptcp_sock *msk; 368 struct hlist_nulls_node token_node; 369 }; 370 371 static inline struct mptcp_subflow_request_sock * 372 mptcp_subflow_rsk(const struct request_sock *rsk) 373 { 374 return (struct mptcp_subflow_request_sock *)rsk; 375 } 376 377 enum mptcp_data_avail { 378 MPTCP_SUBFLOW_NODATA, 379 MPTCP_SUBFLOW_DATA_AVAIL, 380 }; 381 382 struct mptcp_delegated_action { 383 struct napi_struct napi; 384 struct list_head head; 385 }; 386 387 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 388 389 #define MPTCP_DELEGATE_SEND 0 390 #define MPTCP_DELEGATE_ACK 1 391 392 /* MPTCP subflow context */ 393 struct mptcp_subflow_context { 394 struct list_head node;/* conn_list of subflows */ 395 u64 local_key; 396 u64 remote_key; 397 u64 idsn; 398 u64 map_seq; 399 u32 snd_isn; 400 u32 token; 401 u32 rel_write_seq; 402 u32 map_subflow_seq; 403 u32 ssn_offset; 404 u32 map_data_len; 405 __wsum map_data_csum; 406 u32 map_csum_len; 407 u32 request_mptcp : 1, /* send MP_CAPABLE */ 408 request_join : 1, /* send MP_JOIN */ 409 request_bkup : 1, 410 mp_capable : 1, /* remote is MPTCP capable */ 411 mp_join : 1, /* remote is JOINing */ 412 fully_established : 1, /* path validated */ 413 pm_notified : 1, /* PM hook called for established status */ 414 conn_finished : 1, 415 map_valid : 1, 416 map_csum_reqd : 1, 417 map_data_fin : 1, 418 mpc_map : 1, 419 backup : 1, 420 send_mp_prio : 1, 421 send_mp_fail : 1, 422 rx_eof : 1, 423 can_ack : 1, /* only after processing the remote a key */ 424 disposable : 1, /* ctx can be free at ulp release time */ 425 stale : 1; /* unable to snd/rcv data, do not use for xmit */ 426 enum mptcp_data_avail data_avail; 427 u32 remote_nonce; 428 u64 thmac; 429 u32 local_nonce; 430 u32 remote_token; 431 u8 hmac[MPTCPOPT_HMAC_LEN]; 432 u8 local_id; 433 u8 remote_id; 434 u8 reset_seen:1; 435 u8 reset_transient:1; 436 u8 reset_reason:4; 437 u8 stale_count; 438 439 long delegated_status; 440 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 441 442 u32 setsockopt_seq; 443 u32 stale_rcv_tstamp; 444 445 struct sock *tcp_sock; /* tcp sk backpointer */ 446 struct sock *conn; /* parent mptcp_sock */ 447 const struct inet_connection_sock_af_ops *icsk_af_ops; 448 void (*tcp_data_ready)(struct sock *sk); 449 void (*tcp_state_change)(struct sock *sk); 450 void (*tcp_write_space)(struct sock *sk); 451 void (*tcp_error_report)(struct sock *sk); 452 453 struct rcu_head rcu; 454 }; 455 456 static inline struct mptcp_subflow_context * 457 mptcp_subflow_ctx(const struct sock *sk) 458 { 459 struct inet_connection_sock *icsk = inet_csk(sk); 460 461 /* Use RCU on icsk_ulp_data only for sock diag code */ 462 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 463 } 464 465 static inline struct sock * 466 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 467 { 468 return subflow->tcp_sock; 469 } 470 471 static inline u64 472 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 473 { 474 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 475 subflow->ssn_offset - 476 subflow->map_subflow_seq; 477 } 478 479 static inline u64 480 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 481 { 482 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 483 } 484 485 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk, 486 struct mptcp_subflow_context *subflow) 487 { 488 sock_hold(mptcp_subflow_tcp_sock(subflow)); 489 spin_lock_bh(&msk->join_list_lock); 490 list_add_tail(&subflow->node, &msk->join_list); 491 spin_unlock_bh(&msk->join_list_lock); 492 } 493 494 void mptcp_subflow_process_delegated(struct sock *ssk); 495 496 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 497 { 498 struct mptcp_delegated_action *delegated; 499 bool schedule; 500 501 /* the caller held the subflow bh socket lock */ 502 lockdep_assert_in_softirq(); 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(action, &subflow->delegated_status)) { 509 /* still on delegated list from previous scheduling */ 510 if (!list_empty(&subflow->delegated_node)) 511 return; 512 513 delegated = this_cpu_ptr(&mptcp_delegated_actions); 514 schedule = list_empty(&delegated->head); 515 list_add_tail(&subflow->delegated_node, &delegated->head); 516 sock_hold(mptcp_subflow_tcp_sock(subflow)); 517 if (schedule) 518 napi_schedule(&delegated->napi); 519 } 520 } 521 522 static inline struct mptcp_subflow_context * 523 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 524 { 525 struct mptcp_subflow_context *ret; 526 527 if (list_empty(&delegated->head)) 528 return NULL; 529 530 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 531 list_del_init(&ret->delegated_node); 532 return ret; 533 } 534 535 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 536 { 537 return !!READ_ONCE(subflow->delegated_status); 538 } 539 540 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action) 541 { 542 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 543 * touching the status bit 544 */ 545 smp_wmb(); 546 clear_bit(action, &subflow->delegated_status); 547 } 548 549 int mptcp_is_enabled(const struct net *net); 550 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 551 int mptcp_is_checksum_enabled(const struct net *net); 552 int mptcp_allow_join_id0(const struct net *net); 553 unsigned int mptcp_stale_loss_cnt(const struct net *net); 554 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 555 struct mptcp_options_received *mp_opt); 556 bool __mptcp_retransmit_pending_data(struct sock *sk); 557 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 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_send_ack(struct sock *ssk); 564 void mptcp_subflow_reset(struct sock *ssk); 565 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 566 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); 567 568 /* called with sk socket lock held */ 569 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 570 const struct mptcp_addr_info *remote); 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 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 589 590 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 591 592 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 593 struct mptcp_subflow_context *ctx) 594 { 595 sk->sk_data_ready = ctx->tcp_data_ready; 596 sk->sk_state_change = ctx->tcp_state_change; 597 sk->sk_write_space = ctx->tcp_write_space; 598 sk->sk_error_report = ctx->tcp_error_report; 599 600 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 601 } 602 603 static inline bool mptcp_has_another_subflow(struct sock *ssk) 604 { 605 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk), *tmp; 606 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 607 608 mptcp_for_each_subflow(msk, tmp) { 609 if (tmp != subflow) 610 return true; 611 } 612 613 return false; 614 } 615 616 void __init mptcp_proto_init(void); 617 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 618 int __init mptcp_proto_v6_init(void); 619 #endif 620 621 struct sock *mptcp_sk_clone(const struct sock *sk, 622 const struct mptcp_options_received *mp_opt, 623 struct request_sock *req); 624 void mptcp_get_options(const struct sock *sk, 625 const struct sk_buff *skb, 626 struct mptcp_options_received *mp_opt); 627 628 void mptcp_finish_connect(struct sock *sk); 629 void __mptcp_set_connected(struct sock *sk); 630 static inline bool mptcp_is_fully_established(struct sock *sk) 631 { 632 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 633 READ_ONCE(mptcp_sk(sk)->fully_established); 634 } 635 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 636 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 637 bool mptcp_finish_join(struct sock *sk); 638 bool mptcp_schedule_work(struct sock *sk); 639 int mptcp_setsockopt(struct sock *sk, int level, int optname, 640 sockptr_t optval, unsigned int optlen); 641 int mptcp_getsockopt(struct sock *sk, int level, int optname, 642 char __user *optval, int __user *option); 643 644 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 645 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 646 { 647 if (use_64bit) 648 return cur_seq; 649 650 return __mptcp_expand_seq(old_seq, cur_seq); 651 } 652 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 653 void __mptcp_data_acked(struct sock *sk); 654 void __mptcp_error_report(struct sock *sk); 655 void mptcp_subflow_eof(struct sock *sk); 656 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 657 void __mptcp_flush_join_list(struct mptcp_sock *msk); 658 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 659 { 660 return READ_ONCE(msk->snd_data_fin_enable) && 661 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 662 } 663 664 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 665 { 666 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 667 return false; 668 669 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 670 return true; 671 } 672 673 static inline void mptcp_write_space(struct sock *sk) 674 { 675 if (sk_stream_is_writeable(sk)) { 676 /* pairs with memory barrier in mptcp_poll */ 677 smp_mb(); 678 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 679 sk_stream_write_space(sk); 680 } 681 } 682 683 void mptcp_destroy_common(struct mptcp_sock *msk); 684 685 #define MPTCP_TOKEN_MAX_RETRIES 4 686 687 void __init mptcp_token_init(void); 688 static inline void mptcp_token_init_request(struct request_sock *req) 689 { 690 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 691 } 692 693 int mptcp_token_new_request(struct request_sock *req); 694 void mptcp_token_destroy_request(struct request_sock *req); 695 int mptcp_token_new_connect(struct sock *sk); 696 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 697 struct mptcp_sock *msk); 698 bool mptcp_token_exists(u32 token); 699 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 700 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 701 long *s_num); 702 void mptcp_token_destroy(struct mptcp_sock *msk); 703 704 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 705 706 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 707 708 void __init mptcp_pm_init(void); 709 void mptcp_pm_data_init(struct mptcp_sock *msk); 710 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 711 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 712 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 713 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp); 714 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 715 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 716 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 717 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id); 718 void mptcp_pm_add_addr_received(struct mptcp_sock *msk, 719 const struct mptcp_addr_info *addr); 720 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 721 struct mptcp_addr_info *addr); 722 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 723 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 724 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 725 const struct mptcp_rm_list *rm_list); 726 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 727 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 728 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 729 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 730 struct mptcp_pm_add_entry * 731 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 732 struct mptcp_addr_info *addr, bool check_id); 733 struct mptcp_pm_add_entry * 734 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk, 735 struct mptcp_addr_info *addr); 736 int mptcp_pm_get_flags_and_ifindex_by_id(struct net *net, unsigned int id, 737 u8 *flags, int *ifindex); 738 739 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 740 const struct mptcp_addr_info *addr, 741 bool echo); 742 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 743 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 744 745 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 746 const struct sock *ssk, gfp_t gfp); 747 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info); 748 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 749 750 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 751 { 752 return READ_ONCE(msk->pm.addr_signal) & 753 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 754 } 755 756 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 757 { 758 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 759 } 760 761 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 762 { 763 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 764 } 765 766 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 767 { 768 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 769 } 770 771 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 772 { 773 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 774 775 if (family == AF_INET6) 776 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 777 if (!echo) 778 len += MPTCPOPT_THMAC_LEN; 779 /* account for 2 trailing 'nop' options */ 780 if (port) 781 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 782 783 return len; 784 } 785 786 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 787 { 788 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 789 return -EINVAL; 790 791 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 792 } 793 794 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, struct sk_buff *skb, 795 unsigned int opt_size, unsigned int remaining, 796 struct mptcp_addr_info *addr, bool *echo, 797 bool *port, bool *drop_other_suboptions); 798 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 799 struct mptcp_rm_list *rm_list); 800 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 801 802 void __init mptcp_pm_nl_init(void); 803 void mptcp_pm_nl_data_init(struct mptcp_sock *msk); 804 void mptcp_pm_nl_work(struct mptcp_sock *msk); 805 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 806 const struct mptcp_rm_list *rm_list); 807 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 808 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk); 809 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk); 810 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk); 811 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk); 812 813 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 814 void mptcp_sockopt_sync_all(struct mptcp_sock *msk); 815 816 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 817 { 818 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 819 } 820 821 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 822 823 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 824 { 825 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 826 } 827 828 static inline bool mptcp_check_fallback(const 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 return __mptcp_check_fallback(msk); 834 } 835 836 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 837 { 838 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 839 pr_debug("TCP fallback already done (msk=%p)", msk); 840 return; 841 } 842 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 843 } 844 845 static inline void mptcp_do_fallback(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 __mptcp_do_fallback(msk); 851 } 852 853 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 854 855 static inline bool subflow_simultaneous_connect(struct sock *sk) 856 { 857 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 858 struct sock *parent = subflow->conn; 859 860 return sk->sk_state == TCP_ESTABLISHED && 861 !mptcp_sk(parent)->pm.server_side && 862 !subflow->conn_finished; 863 } 864 865 #ifdef CONFIG_SYN_COOKIES 866 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 867 struct sk_buff *skb); 868 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 869 struct sk_buff *skb); 870 void __init mptcp_join_cookie_init(void); 871 #else 872 static inline void 873 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 874 struct sk_buff *skb) {} 875 static inline bool 876 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 877 struct sk_buff *skb) 878 { 879 return false; 880 } 881 882 static inline void mptcp_join_cookie_init(void) {} 883 #endif 884 885 #endif /* __MPTCP_PROTOCOL_H */ 886