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 atomic flags */ 114 #define MPTCP_NOSPACE 1 115 #define MPTCP_WORK_RTX 2 116 #define MPTCP_WORK_EOF 3 117 #define MPTCP_FALLBACK_DONE 4 118 #define MPTCP_WORK_CLOSE_SUBFLOW 5 119 120 /* MPTCP socket release cb flags */ 121 #define MPTCP_PUSH_PENDING 1 122 #define MPTCP_CLEAN_UNA 2 123 #define MPTCP_ERROR_REPORT 3 124 #define MPTCP_RETRANSMIT 4 125 #define MPTCP_FLUSH_JOIN_LIST 5 126 #define MPTCP_CONNECTED 6 127 #define MPTCP_RESET_SCHEDULER 7 128 129 static inline bool before64(__u64 seq1, __u64 seq2) 130 { 131 return (__s64)(seq1 - seq2) < 0; 132 } 133 134 #define after64(seq2, seq1) before64(seq1, seq2) 135 136 struct mptcp_options_received { 137 u64 sndr_key; 138 u64 rcvr_key; 139 u64 data_ack; 140 u64 data_seq; 141 u32 subflow_seq; 142 u16 data_len; 143 __sum16 csum; 144 u16 suboptions; 145 u32 token; 146 u32 nonce; 147 u16 use_map:1, 148 dsn64:1, 149 data_fin:1, 150 use_ack:1, 151 ack64:1, 152 mpc_map:1, 153 reset_reason:4, 154 reset_transient:1, 155 echo:1, 156 backup:1, 157 deny_join_id0:1, 158 __unused:2; 159 u8 join_id; 160 u64 thmac; 161 u8 hmac[MPTCPOPT_HMAC_LEN]; 162 struct mptcp_addr_info addr; 163 struct mptcp_rm_list rm_list; 164 u64 ahmac; 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_SUBFLOW_ESTABLISHED, 180 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 181 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 182 * accounted int id_avail_bitmap 183 */ 184 }; 185 186 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 187 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 188 189 enum mptcp_addr_signal_status { 190 MPTCP_ADD_ADDR_SIGNAL, 191 MPTCP_ADD_ADDR_ECHO, 192 MPTCP_RM_ADDR_SIGNAL, 193 }; 194 195 /* max value of mptcp_addr_info.id */ 196 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 197 198 struct mptcp_pm_data { 199 struct mptcp_addr_info local; 200 struct mptcp_addr_info remote; 201 struct list_head anno_list; 202 203 spinlock_t lock; /*protects the whole PM data */ 204 205 u8 addr_signal; 206 bool server_side; 207 bool work_pending; 208 bool accept_addr; 209 bool accept_subflow; 210 bool remote_deny_join_id0; 211 u8 add_addr_signaled; 212 u8 add_addr_accepted; 213 u8 local_addr_used; 214 u8 subflows; 215 u8 status; 216 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 217 struct mptcp_rm_list rm_list_tx; 218 struct mptcp_rm_list rm_list_rx; 219 }; 220 221 struct mptcp_data_frag { 222 struct list_head list; 223 u64 data_seq; 224 u16 data_len; 225 u16 offset; 226 u16 overhead; 227 u16 already_sent; 228 struct page *page; 229 }; 230 231 /* MPTCP connection sock */ 232 struct mptcp_sock { 233 /* inet_connection_sock must be the first member */ 234 struct inet_connection_sock sk; 235 u64 local_key; 236 u64 remote_key; 237 u64 write_seq; 238 u64 snd_nxt; 239 u64 ack_seq; 240 u64 rcv_wnd_sent; 241 u64 rcv_data_fin_seq; 242 int rmem_fwd_alloc; 243 struct sock *last_snd; 244 int snd_burst; 245 int old_wspace; 246 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 247 * recovery related fields are under data_lock 248 * protection 249 */ 250 u64 snd_una; 251 u64 wnd_end; 252 unsigned long timer_ival; 253 u32 token; 254 int rmem_released; 255 unsigned long flags; 256 unsigned long cb_flags; 257 unsigned long push_pending; 258 bool recovery; /* closing subflow write queue reinjected */ 259 bool can_ack; 260 bool fully_established; 261 bool rcv_data_fin; 262 bool snd_data_fin_enable; 263 bool rcv_fastclose; 264 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 265 bool csum_enabled; 266 bool allow_infinite_fallback; 267 u8 recvmsg_inq:1, 268 cork:1, 269 nodelay:1; 270 struct work_struct work; 271 struct sk_buff *ooo_last_skb; 272 struct rb_root out_of_order_queue; 273 struct sk_buff_head receive_queue; 274 struct list_head conn_list; 275 struct list_head rtx_queue; 276 struct mptcp_data_frag *first_pending; 277 struct list_head join_list; 278 struct socket *subflow; /* outgoing connect/listener/!mp_capable */ 279 struct sock *first; 280 struct mptcp_pm_data pm; 281 struct { 282 u32 space; /* bytes copied in last measurement window */ 283 u32 copied; /* bytes copied in this measurement window */ 284 u64 time; /* start time of measurement window */ 285 u64 rtt_us; /* last maximum rtt of subflows */ 286 } rcvq_space; 287 288 u32 setsockopt_seq; 289 char ca_name[TCP_CA_NAME_MAX]; 290 }; 291 292 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 293 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 294 295 #define mptcp_for_each_subflow(__msk, __subflow) \ 296 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 297 298 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 299 { 300 sock_owned_by_me((const struct sock *)msk); 301 } 302 303 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 304 { 305 return (struct mptcp_sock *)sk; 306 } 307 308 /* the msk socket don't use the backlog, also account for the bulk 309 * free memory 310 */ 311 static inline int __mptcp_rmem(const struct sock *sk) 312 { 313 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 314 } 315 316 static inline int __mptcp_space(const struct sock *sk) 317 { 318 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 319 } 320 321 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 322 { 323 const struct mptcp_sock *msk = mptcp_sk(sk); 324 325 return READ_ONCE(msk->first_pending); 326 } 327 328 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 329 { 330 struct mptcp_sock *msk = mptcp_sk(sk); 331 struct mptcp_data_frag *cur; 332 333 cur = msk->first_pending; 334 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 335 list_next_entry(cur, list); 336 } 337 338 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 339 { 340 struct mptcp_sock *msk = mptcp_sk(sk); 341 342 if (!msk->first_pending) 343 return NULL; 344 345 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 346 return NULL; 347 348 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 349 } 350 351 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 352 { 353 struct mptcp_sock *msk = mptcp_sk(sk); 354 355 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 356 return NULL; 357 358 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 359 } 360 361 struct csum_pseudo_header { 362 __be64 data_seq; 363 __be32 subflow_seq; 364 __be16 data_len; 365 __sum16 csum; 366 }; 367 368 struct mptcp_subflow_request_sock { 369 struct tcp_request_sock sk; 370 u16 mp_capable : 1, 371 mp_join : 1, 372 backup : 1, 373 csum_reqd : 1, 374 allow_join_id0 : 1; 375 u8 local_id; 376 u8 remote_id; 377 u64 local_key; 378 u64 idsn; 379 u32 token; 380 u32 ssn_offset; 381 u64 thmac; 382 u32 local_nonce; 383 u32 remote_nonce; 384 struct mptcp_sock *msk; 385 struct hlist_nulls_node token_node; 386 }; 387 388 static inline struct mptcp_subflow_request_sock * 389 mptcp_subflow_rsk(const struct request_sock *rsk) 390 { 391 return (struct mptcp_subflow_request_sock *)rsk; 392 } 393 394 enum mptcp_data_avail { 395 MPTCP_SUBFLOW_NODATA, 396 MPTCP_SUBFLOW_DATA_AVAIL, 397 }; 398 399 struct mptcp_delegated_action { 400 struct napi_struct napi; 401 struct list_head head; 402 }; 403 404 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 405 406 #define MPTCP_DELEGATE_SEND 0 407 #define MPTCP_DELEGATE_ACK 1 408 409 /* MPTCP subflow context */ 410 struct mptcp_subflow_context { 411 struct list_head node;/* conn_list of subflows */ 412 413 struct_group(reset, 414 415 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 416 u64 local_key; 417 u64 remote_key; 418 u64 idsn; 419 u64 map_seq; 420 u32 snd_isn; 421 u32 token; 422 u32 rel_write_seq; 423 u32 map_subflow_seq; 424 u32 ssn_offset; 425 u32 map_data_len; 426 __wsum map_data_csum; 427 u32 map_csum_len; 428 u32 request_mptcp : 1, /* send MP_CAPABLE */ 429 request_join : 1, /* send MP_JOIN */ 430 request_bkup : 1, 431 mp_capable : 1, /* remote is MPTCP capable */ 432 mp_join : 1, /* remote is JOINing */ 433 fully_established : 1, /* path validated */ 434 pm_notified : 1, /* PM hook called for established status */ 435 conn_finished : 1, 436 map_valid : 1, 437 map_csum_reqd : 1, 438 map_data_fin : 1, 439 mpc_map : 1, 440 backup : 1, 441 send_mp_prio : 1, 442 send_mp_fail : 1, 443 send_fastclose : 1, 444 send_infinite_map : 1, 445 rx_eof : 1, 446 can_ack : 1, /* only after processing the remote a key */ 447 disposable : 1, /* ctx can be free at ulp release time */ 448 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 449 local_id_valid : 1; /* local_id is correctly initialized */ 450 enum mptcp_data_avail data_avail; 451 u32 remote_nonce; 452 u64 thmac; 453 u32 local_nonce; 454 u32 remote_token; 455 u8 hmac[MPTCPOPT_HMAC_LEN]; 456 u8 local_id; 457 u8 remote_id; 458 u8 reset_seen:1; 459 u8 reset_transient:1; 460 u8 reset_reason:4; 461 u8 stale_count; 462 463 long delegated_status; 464 465 ); 466 467 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 468 469 u32 setsockopt_seq; 470 u32 stale_rcv_tstamp; 471 472 struct sock *tcp_sock; /* tcp sk backpointer */ 473 struct sock *conn; /* parent mptcp_sock */ 474 const struct inet_connection_sock_af_ops *icsk_af_ops; 475 void (*tcp_state_change)(struct sock *sk); 476 void (*tcp_error_report)(struct sock *sk); 477 478 struct rcu_head rcu; 479 }; 480 481 static inline struct mptcp_subflow_context * 482 mptcp_subflow_ctx(const struct sock *sk) 483 { 484 struct inet_connection_sock *icsk = inet_csk(sk); 485 486 /* Use RCU on icsk_ulp_data only for sock diag code */ 487 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 488 } 489 490 static inline struct sock * 491 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 492 { 493 return subflow->tcp_sock; 494 } 495 496 static inline void 497 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 498 { 499 memset(&subflow->reset, 0, sizeof(subflow->reset)); 500 subflow->request_mptcp = 1; 501 } 502 503 static inline u64 504 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 505 { 506 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 507 subflow->ssn_offset - 508 subflow->map_subflow_seq; 509 } 510 511 static inline u64 512 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 513 { 514 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 515 } 516 517 void mptcp_subflow_process_delegated(struct sock *ssk); 518 519 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 520 { 521 struct mptcp_delegated_action *delegated; 522 bool schedule; 523 524 /* the caller held the subflow bh socket lock */ 525 lockdep_assert_in_softirq(); 526 527 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 528 * ensures the below list check sees list updates done prior to status 529 * bit changes 530 */ 531 if (!test_and_set_bit(action, &subflow->delegated_status)) { 532 /* still on delegated list from previous scheduling */ 533 if (!list_empty(&subflow->delegated_node)) 534 return; 535 536 delegated = this_cpu_ptr(&mptcp_delegated_actions); 537 schedule = list_empty(&delegated->head); 538 list_add_tail(&subflow->delegated_node, &delegated->head); 539 sock_hold(mptcp_subflow_tcp_sock(subflow)); 540 if (schedule) 541 napi_schedule(&delegated->napi); 542 } 543 } 544 545 static inline struct mptcp_subflow_context * 546 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 547 { 548 struct mptcp_subflow_context *ret; 549 550 if (list_empty(&delegated->head)) 551 return NULL; 552 553 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 554 list_del_init(&ret->delegated_node); 555 return ret; 556 } 557 558 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 559 { 560 return !!READ_ONCE(subflow->delegated_status); 561 } 562 563 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action) 564 { 565 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 566 * touching the status bit 567 */ 568 smp_wmb(); 569 clear_bit(action, &subflow->delegated_status); 570 } 571 572 int mptcp_is_enabled(const struct net *net); 573 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 574 int mptcp_is_checksum_enabled(const struct net *net); 575 int mptcp_allow_join_id0(const struct net *net); 576 unsigned int mptcp_stale_loss_cnt(const struct net *net); 577 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 578 struct mptcp_options_received *mp_opt); 579 bool __mptcp_retransmit_pending_data(struct sock *sk); 580 void mptcp_check_and_set_pending(struct sock *sk); 581 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 582 bool mptcp_subflow_data_available(struct sock *sk); 583 void __init mptcp_subflow_init(void); 584 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 585 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 586 struct mptcp_subflow_context *subflow); 587 void mptcp_subflow_send_ack(struct sock *ssk); 588 void mptcp_subflow_reset(struct sock *ssk); 589 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 590 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); 591 592 /* called with sk socket lock held */ 593 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 594 const struct mptcp_addr_info *remote); 595 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); 596 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 597 struct sockaddr_storage *addr, 598 unsigned short family); 599 600 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 601 { 602 struct sock *ssk = mptcp_subflow_tcp_sock(subflow); 603 604 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 605 if (subflow->request_join && !subflow->fully_established) 606 return false; 607 608 /* only send if our side has not closed yet */ 609 return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 610 } 611 612 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 613 614 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 615 616 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 617 struct mptcp_subflow_context *ctx) 618 { 619 sk->sk_data_ready = sock_def_readable; 620 sk->sk_state_change = ctx->tcp_state_change; 621 sk->sk_write_space = sk_stream_write_space; 622 sk->sk_error_report = ctx->tcp_error_report; 623 624 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 625 } 626 627 static inline bool mptcp_has_another_subflow(struct sock *ssk) 628 { 629 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk), *tmp; 630 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 631 632 mptcp_for_each_subflow(msk, tmp) { 633 if (tmp != subflow) 634 return true; 635 } 636 637 return false; 638 } 639 640 void __init mptcp_proto_init(void); 641 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 642 int __init mptcp_proto_v6_init(void); 643 #endif 644 645 struct sock *mptcp_sk_clone(const struct sock *sk, 646 const struct mptcp_options_received *mp_opt, 647 struct request_sock *req); 648 void mptcp_get_options(const struct sk_buff *skb, 649 struct mptcp_options_received *mp_opt); 650 651 void mptcp_finish_connect(struct sock *sk); 652 void __mptcp_set_connected(struct sock *sk); 653 static inline bool mptcp_is_fully_established(struct sock *sk) 654 { 655 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 656 READ_ONCE(mptcp_sk(sk)->fully_established); 657 } 658 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 659 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 660 bool mptcp_finish_join(struct sock *sk); 661 bool mptcp_schedule_work(struct sock *sk); 662 int mptcp_setsockopt(struct sock *sk, int level, int optname, 663 sockptr_t optval, unsigned int optlen); 664 int mptcp_getsockopt(struct sock *sk, int level, int optname, 665 char __user *optval, int __user *option); 666 667 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 668 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 669 { 670 if (use_64bit) 671 return cur_seq; 672 673 return __mptcp_expand_seq(old_seq, cur_seq); 674 } 675 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 676 void __mptcp_data_acked(struct sock *sk); 677 void __mptcp_error_report(struct sock *sk); 678 void mptcp_subflow_eof(struct sock *sk); 679 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 680 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 681 { 682 return READ_ONCE(msk->snd_data_fin_enable) && 683 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 684 } 685 686 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 687 { 688 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 689 return false; 690 691 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 692 return true; 693 } 694 695 static inline void mptcp_write_space(struct sock *sk) 696 { 697 if (sk_stream_is_writeable(sk)) { 698 /* pairs with memory barrier in mptcp_poll */ 699 smp_mb(); 700 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 701 sk_stream_write_space(sk); 702 } 703 } 704 705 void mptcp_destroy_common(struct mptcp_sock *msk); 706 707 #define MPTCP_TOKEN_MAX_RETRIES 4 708 709 void __init mptcp_token_init(void); 710 static inline void mptcp_token_init_request(struct request_sock *req) 711 { 712 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 713 } 714 715 int mptcp_token_new_request(struct request_sock *req); 716 void mptcp_token_destroy_request(struct request_sock *req); 717 int mptcp_token_new_connect(struct sock *sk); 718 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 719 struct mptcp_sock *msk); 720 bool mptcp_token_exists(u32 token); 721 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 722 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 723 long *s_num); 724 void mptcp_token_destroy(struct mptcp_sock *msk); 725 726 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 727 728 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 729 u16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 730 731 void __init mptcp_pm_init(void); 732 void mptcp_pm_data_init(struct mptcp_sock *msk); 733 void mptcp_pm_data_reset(struct mptcp_sock *msk); 734 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 735 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 736 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 737 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp); 738 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 739 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 740 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 741 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 742 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk, 743 const struct mptcp_subflow_context *subflow); 744 void mptcp_pm_add_addr_received(struct mptcp_sock *msk, 745 const struct mptcp_addr_info *addr); 746 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 747 const struct mptcp_addr_info *addr); 748 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 749 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 750 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 751 const struct mptcp_rm_list *rm_list); 752 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 753 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 754 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 755 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 756 struct mptcp_pm_add_entry * 757 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 758 const struct mptcp_addr_info *addr, bool check_id); 759 struct mptcp_pm_add_entry * 760 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 761 const struct mptcp_addr_info *addr); 762 int mptcp_pm_get_flags_and_ifindex_by_id(struct net *net, unsigned int id, 763 u8 *flags, int *ifindex); 764 765 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 766 const struct mptcp_addr_info *addr, 767 bool echo); 768 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 769 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 770 771 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 772 const struct sock *ssk, gfp_t gfp); 773 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info); 774 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 775 776 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 777 { 778 return READ_ONCE(msk->pm.addr_signal) & 779 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 780 } 781 782 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 783 { 784 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 785 } 786 787 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 788 { 789 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 790 } 791 792 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 793 { 794 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 795 } 796 797 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 798 { 799 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 800 801 if (family == AF_INET6) 802 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 803 if (!echo) 804 len += MPTCPOPT_THMAC_LEN; 805 /* account for 2 trailing 'nop' options */ 806 if (port) 807 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 808 809 return len; 810 } 811 812 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 813 { 814 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 815 return -EINVAL; 816 817 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 818 } 819 820 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 821 unsigned int opt_size, unsigned int remaining, 822 struct mptcp_addr_info *addr, bool *echo, 823 bool *drop_other_suboptions); 824 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 825 struct mptcp_rm_list *rm_list); 826 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 827 828 void __init mptcp_pm_nl_init(void); 829 void mptcp_pm_nl_data_init(struct mptcp_sock *msk); 830 void mptcp_pm_nl_work(struct mptcp_sock *msk); 831 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 832 const struct mptcp_rm_list *rm_list); 833 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 834 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 835 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 836 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 837 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 838 839 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 840 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 841 842 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 843 { 844 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 845 } 846 847 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 848 849 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 850 { 851 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 852 } 853 854 static inline bool mptcp_check_fallback(const struct sock *sk) 855 { 856 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 857 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 858 859 return __mptcp_check_fallback(msk); 860 } 861 862 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 863 { 864 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 865 pr_debug("TCP fallback already done (msk=%p)", msk); 866 return; 867 } 868 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 869 } 870 871 static inline void mptcp_do_fallback(struct sock *sk) 872 { 873 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 874 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 875 876 __mptcp_do_fallback(msk); 877 } 878 879 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 880 881 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 882 { 883 struct mptcp_ext *mpext; 884 885 mpext = skb ? mptcp_get_ext(skb) : NULL; 886 if (mpext && mpext->infinite_map) 887 return true; 888 889 return false; 890 } 891 892 static inline bool subflow_simultaneous_connect(struct sock *sk) 893 { 894 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 895 struct sock *parent = subflow->conn; 896 897 return sk->sk_state == TCP_ESTABLISHED && 898 !mptcp_sk(parent)->pm.server_side && 899 !subflow->conn_finished; 900 } 901 902 #ifdef CONFIG_SYN_COOKIES 903 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 904 struct sk_buff *skb); 905 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 906 struct sk_buff *skb); 907 void __init mptcp_join_cookie_init(void); 908 #else 909 static inline void 910 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 911 struct sk_buff *skb) {} 912 static inline bool 913 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 914 struct sk_buff *skb) 915 { 916 return false; 917 } 918 919 static inline void mptcp_join_cookie_init(void) {} 920 #endif 921 922 #endif /* __MPTCP_PROTOCOL_H */ 923