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 #include <net/genetlink.h> 15 16 #define MPTCP_SUPPORTED_VERSION 1 17 18 /* MPTCP option bits */ 19 #define OPTION_MPTCP_MPC_SYN BIT(0) 20 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 21 #define OPTION_MPTCP_MPC_ACK BIT(2) 22 #define OPTION_MPTCP_MPJ_SYN BIT(3) 23 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 24 #define OPTION_MPTCP_MPJ_ACK BIT(5) 25 #define OPTION_MPTCP_ADD_ADDR BIT(6) 26 #define OPTION_MPTCP_RM_ADDR BIT(7) 27 #define OPTION_MPTCP_FASTCLOSE BIT(8) 28 #define OPTION_MPTCP_PRIO BIT(9) 29 #define OPTION_MPTCP_RST BIT(10) 30 #define OPTION_MPTCP_DSS BIT(11) 31 #define OPTION_MPTCP_FAIL BIT(12) 32 33 #define OPTION_MPTCP_CSUMREQD BIT(13) 34 35 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \ 36 OPTION_MPTCP_MPC_ACK) 37 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \ 38 OPTION_MPTCP_MPJ_ACK) 39 40 /* MPTCP option subtypes */ 41 #define MPTCPOPT_MP_CAPABLE 0 42 #define MPTCPOPT_MP_JOIN 1 43 #define MPTCPOPT_DSS 2 44 #define MPTCPOPT_ADD_ADDR 3 45 #define MPTCPOPT_RM_ADDR 4 46 #define MPTCPOPT_MP_PRIO 5 47 #define MPTCPOPT_MP_FAIL 6 48 #define MPTCPOPT_MP_FASTCLOSE 7 49 #define MPTCPOPT_RST 8 50 51 /* MPTCP suboption lengths */ 52 #define TCPOLEN_MPTCP_MPC_SYN 4 53 #define TCPOLEN_MPTCP_MPC_SYNACK 12 54 #define TCPOLEN_MPTCP_MPC_ACK 20 55 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 56 #define TCPOLEN_MPTCP_MPJ_SYN 12 57 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 58 #define TCPOLEN_MPTCP_MPJ_ACK 24 59 #define TCPOLEN_MPTCP_DSS_BASE 4 60 #define TCPOLEN_MPTCP_DSS_ACK32 4 61 #define TCPOLEN_MPTCP_DSS_ACK64 8 62 #define TCPOLEN_MPTCP_DSS_MAP32 10 63 #define TCPOLEN_MPTCP_DSS_MAP64 14 64 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 65 #define TCPOLEN_MPTCP_ADD_ADDR 16 66 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 69 #define TCPOLEN_MPTCP_ADD_ADDR6 28 70 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 73 #define TCPOLEN_MPTCP_PORT_LEN 2 74 #define TCPOLEN_MPTCP_PORT_ALIGN 2 75 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 76 #define TCPOLEN_MPTCP_PRIO 3 77 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 78 #define TCPOLEN_MPTCP_FASTCLOSE 12 79 #define TCPOLEN_MPTCP_RST 4 80 #define TCPOLEN_MPTCP_FAIL 12 81 82 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 83 84 /* MPTCP MP_JOIN flags */ 85 #define MPTCPOPT_BACKUP BIT(0) 86 #define MPTCPOPT_HMAC_LEN 20 87 #define MPTCPOPT_THMAC_LEN 8 88 89 /* MPTCP MP_CAPABLE flags */ 90 #define MPTCP_VERSION_MASK (0x0F) 91 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 92 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 93 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 94 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 95 #define MPTCP_CAP_FLAG_MASK (0x1F) 96 97 /* MPTCP DSS flags */ 98 #define MPTCP_DSS_DATA_FIN BIT(4) 99 #define MPTCP_DSS_DSN64 BIT(3) 100 #define MPTCP_DSS_HAS_MAP BIT(2) 101 #define MPTCP_DSS_ACK64 BIT(1) 102 #define MPTCP_DSS_HAS_ACK BIT(0) 103 #define MPTCP_DSS_FLAG_MASK (0x1F) 104 105 /* MPTCP ADD_ADDR flags */ 106 #define MPTCP_ADDR_ECHO BIT(0) 107 108 /* MPTCP MP_PRIO flags */ 109 #define MPTCP_PRIO_BKUP BIT(0) 110 111 /* MPTCP TCPRST flags */ 112 #define MPTCP_RST_TRANSIENT BIT(0) 113 114 /* MPTCP socket atomic flags */ 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 5 120 121 /* MPTCP socket release cb flags */ 122 #define MPTCP_PUSH_PENDING 1 123 #define MPTCP_CLEAN_UNA 2 124 #define MPTCP_ERROR_REPORT 3 125 #define MPTCP_RETRANSMIT 4 126 #define MPTCP_FLUSH_JOIN_LIST 5 127 #define MPTCP_CONNECTED 6 128 #define MPTCP_RESET_SCHEDULER 7 129 130 static inline bool before64(__u64 seq1, __u64 seq2) 131 { 132 return (__s64)(seq1 - seq2) < 0; 133 } 134 135 #define after64(seq2, seq1) before64(seq1, seq2) 136 137 struct mptcp_options_received { 138 u64 sndr_key; 139 u64 rcvr_key; 140 u64 data_ack; 141 u64 data_seq; 142 u32 subflow_seq; 143 u16 data_len; 144 __sum16 csum; 145 u16 suboptions; 146 u32 token; 147 u32 nonce; 148 u16 use_map:1, 149 dsn64:1, 150 data_fin:1, 151 use_ack:1, 152 ack64:1, 153 mpc_map:1, 154 reset_reason:4, 155 reset_transient:1, 156 echo:1, 157 backup:1, 158 deny_join_id0:1, 159 __unused:2; 160 u8 join_id; 161 u64 thmac; 162 u8 hmac[MPTCPOPT_HMAC_LEN]; 163 struct mptcp_addr_info addr; 164 struct mptcp_rm_list rm_list; 165 u64 ahmac; 166 u64 fail_seq; 167 }; 168 169 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 170 { 171 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 172 ((nib & 0xF) << 8) | field); 173 } 174 175 enum mptcp_pm_status { 176 MPTCP_PM_ADD_ADDR_RECEIVED, 177 MPTCP_PM_ADD_ADDR_SEND_ACK, 178 MPTCP_PM_RM_ADDR_RECEIVED, 179 MPTCP_PM_ESTABLISHED, 180 MPTCP_PM_SUBFLOW_ESTABLISHED, 181 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 182 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 183 * accounted int id_avail_bitmap 184 */ 185 }; 186 187 enum mptcp_pm_type { 188 MPTCP_PM_TYPE_KERNEL = 0, 189 MPTCP_PM_TYPE_USERSPACE, 190 191 __MPTCP_PM_TYPE_NR, 192 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 193 }; 194 195 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 196 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 197 198 enum mptcp_addr_signal_status { 199 MPTCP_ADD_ADDR_SIGNAL, 200 MPTCP_ADD_ADDR_ECHO, 201 MPTCP_RM_ADDR_SIGNAL, 202 }; 203 204 /* max value of mptcp_addr_info.id */ 205 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 206 207 struct mptcp_pm_data { 208 struct mptcp_addr_info local; 209 struct mptcp_addr_info remote; 210 struct list_head anno_list; 211 struct list_head userspace_pm_local_addr_list; 212 213 spinlock_t lock; /*protects the whole PM data */ 214 215 u8 addr_signal; 216 bool server_side; 217 bool work_pending; 218 bool accept_addr; 219 bool accept_subflow; 220 bool remote_deny_join_id0; 221 u8 add_addr_signaled; 222 u8 add_addr_accepted; 223 u8 local_addr_used; 224 u8 pm_type; 225 u8 subflows; 226 u8 status; 227 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 228 struct mptcp_rm_list rm_list_tx; 229 struct mptcp_rm_list rm_list_rx; 230 }; 231 232 struct mptcp_pm_addr_entry { 233 struct list_head list; 234 struct mptcp_addr_info addr; 235 u8 flags; 236 int ifindex; 237 struct socket *lsk; 238 }; 239 240 struct mptcp_data_frag { 241 struct list_head list; 242 u64 data_seq; 243 u16 data_len; 244 u16 offset; 245 u16 overhead; 246 u16 already_sent; 247 struct page *page; 248 }; 249 250 /* MPTCP connection sock */ 251 struct mptcp_sock { 252 /* inet_connection_sock must be the first member */ 253 struct inet_connection_sock sk; 254 u64 local_key; 255 u64 remote_key; 256 u64 write_seq; 257 u64 snd_nxt; 258 u64 ack_seq; 259 atomic64_t rcv_wnd_sent; 260 u64 rcv_data_fin_seq; 261 int rmem_fwd_alloc; 262 struct sock *last_snd; 263 int snd_burst; 264 int old_wspace; 265 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 266 * recovery related fields are under data_lock 267 * protection 268 */ 269 u64 snd_una; 270 u64 wnd_end; 271 unsigned long timer_ival; 272 u32 token; 273 int rmem_released; 274 unsigned long flags; 275 unsigned long cb_flags; 276 unsigned long push_pending; 277 bool recovery; /* closing subflow write queue reinjected */ 278 bool can_ack; 279 bool fully_established; 280 bool rcv_data_fin; 281 bool snd_data_fin_enable; 282 bool rcv_fastclose; 283 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 284 bool csum_enabled; 285 bool allow_infinite_fallback; 286 u8 recvmsg_inq:1, 287 cork:1, 288 nodelay:1; 289 struct work_struct work; 290 struct sk_buff *ooo_last_skb; 291 struct rb_root out_of_order_queue; 292 struct sk_buff_head receive_queue; 293 struct list_head conn_list; 294 struct list_head rtx_queue; 295 struct mptcp_data_frag *first_pending; 296 struct list_head join_list; 297 struct socket *subflow; /* outgoing connect/listener/!mp_capable */ 298 struct sock *first; 299 struct mptcp_pm_data pm; 300 struct { 301 u32 space; /* bytes copied in last measurement window */ 302 u32 copied; /* bytes copied in this measurement window */ 303 u64 time; /* start time of measurement window */ 304 u64 rtt_us; /* last maximum rtt of subflows */ 305 } rcvq_space; 306 307 u32 setsockopt_seq; 308 char ca_name[TCP_CA_NAME_MAX]; 309 }; 310 311 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 312 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 313 314 #define mptcp_for_each_subflow(__msk, __subflow) \ 315 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 316 317 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 318 { 319 sock_owned_by_me((const struct sock *)msk); 320 } 321 322 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 323 { 324 return (struct mptcp_sock *)sk; 325 } 326 327 /* the msk socket don't use the backlog, also account for the bulk 328 * free memory 329 */ 330 static inline int __mptcp_rmem(const struct sock *sk) 331 { 332 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 333 } 334 335 static inline int __mptcp_space(const struct sock *sk) 336 { 337 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 338 } 339 340 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 341 { 342 const struct mptcp_sock *msk = mptcp_sk(sk); 343 344 return READ_ONCE(msk->first_pending); 345 } 346 347 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 348 { 349 struct mptcp_sock *msk = mptcp_sk(sk); 350 struct mptcp_data_frag *cur; 351 352 cur = msk->first_pending; 353 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 354 list_next_entry(cur, list); 355 } 356 357 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 358 { 359 struct mptcp_sock *msk = mptcp_sk(sk); 360 361 if (!msk->first_pending) 362 return NULL; 363 364 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 365 return NULL; 366 367 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 368 } 369 370 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 371 { 372 struct mptcp_sock *msk = mptcp_sk(sk); 373 374 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 375 return NULL; 376 377 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 378 } 379 380 struct csum_pseudo_header { 381 __be64 data_seq; 382 __be32 subflow_seq; 383 __be16 data_len; 384 __sum16 csum; 385 }; 386 387 struct mptcp_subflow_request_sock { 388 struct tcp_request_sock sk; 389 u16 mp_capable : 1, 390 mp_join : 1, 391 backup : 1, 392 csum_reqd : 1, 393 allow_join_id0 : 1; 394 u8 local_id; 395 u8 remote_id; 396 u64 local_key; 397 u64 idsn; 398 u32 token; 399 u32 ssn_offset; 400 u64 thmac; 401 u32 local_nonce; 402 u32 remote_nonce; 403 struct mptcp_sock *msk; 404 struct hlist_nulls_node token_node; 405 }; 406 407 static inline struct mptcp_subflow_request_sock * 408 mptcp_subflow_rsk(const struct request_sock *rsk) 409 { 410 return (struct mptcp_subflow_request_sock *)rsk; 411 } 412 413 enum mptcp_data_avail { 414 MPTCP_SUBFLOW_NODATA, 415 MPTCP_SUBFLOW_DATA_AVAIL, 416 }; 417 418 struct mptcp_delegated_action { 419 struct napi_struct napi; 420 struct list_head head; 421 }; 422 423 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 424 425 #define MPTCP_DELEGATE_SEND 0 426 #define MPTCP_DELEGATE_ACK 1 427 428 /* MPTCP subflow context */ 429 struct mptcp_subflow_context { 430 struct list_head node;/* conn_list of subflows */ 431 432 struct_group(reset, 433 434 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 435 u64 local_key; 436 u64 remote_key; 437 u64 idsn; 438 u64 map_seq; 439 u32 snd_isn; 440 u32 token; 441 u32 rel_write_seq; 442 u32 map_subflow_seq; 443 u32 ssn_offset; 444 u32 map_data_len; 445 __wsum map_data_csum; 446 u32 map_csum_len; 447 u32 request_mptcp : 1, /* send MP_CAPABLE */ 448 request_join : 1, /* send MP_JOIN */ 449 request_bkup : 1, 450 mp_capable : 1, /* remote is MPTCP capable */ 451 mp_join : 1, /* remote is JOINing */ 452 fully_established : 1, /* path validated */ 453 pm_notified : 1, /* PM hook called for established status */ 454 conn_finished : 1, 455 map_valid : 1, 456 map_csum_reqd : 1, 457 map_data_fin : 1, 458 mpc_map : 1, 459 backup : 1, 460 send_mp_prio : 1, 461 send_mp_fail : 1, 462 send_fastclose : 1, 463 send_infinite_map : 1, 464 rx_eof : 1, 465 can_ack : 1, /* only after processing the remote a key */ 466 disposable : 1, /* ctx can be free at ulp release time */ 467 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 468 local_id_valid : 1, /* local_id is correctly initialized */ 469 valid_csum_seen : 1; /* at least one csum validated */ 470 enum mptcp_data_avail data_avail; 471 bool mp_fail_response_expect; 472 u32 remote_nonce; 473 u64 thmac; 474 u32 local_nonce; 475 u32 remote_token; 476 u8 hmac[MPTCPOPT_HMAC_LEN]; 477 u8 local_id; 478 u8 remote_id; 479 u8 reset_seen:1; 480 u8 reset_transient:1; 481 u8 reset_reason:4; 482 u8 stale_count; 483 484 long delegated_status; 485 486 ); 487 488 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 489 490 u32 setsockopt_seq; 491 u32 stale_rcv_tstamp; 492 493 struct sock *tcp_sock; /* tcp sk backpointer */ 494 struct sock *conn; /* parent mptcp_sock */ 495 const struct inet_connection_sock_af_ops *icsk_af_ops; 496 void (*tcp_state_change)(struct sock *sk); 497 void (*tcp_error_report)(struct sock *sk); 498 499 struct rcu_head rcu; 500 }; 501 502 static inline struct mptcp_subflow_context * 503 mptcp_subflow_ctx(const struct sock *sk) 504 { 505 struct inet_connection_sock *icsk = inet_csk(sk); 506 507 /* Use RCU on icsk_ulp_data only for sock diag code */ 508 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 509 } 510 511 static inline struct sock * 512 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 513 { 514 return subflow->tcp_sock; 515 } 516 517 static inline void 518 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 519 { 520 memset(&subflow->reset, 0, sizeof(subflow->reset)); 521 subflow->request_mptcp = 1; 522 } 523 524 static inline u64 525 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 526 { 527 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 528 subflow->ssn_offset - 529 subflow->map_subflow_seq; 530 } 531 532 static inline u64 533 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 534 { 535 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 536 } 537 538 void mptcp_subflow_process_delegated(struct sock *ssk); 539 540 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 541 { 542 struct mptcp_delegated_action *delegated; 543 bool schedule; 544 545 /* the caller held the subflow bh socket lock */ 546 lockdep_assert_in_softirq(); 547 548 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 549 * ensures the below list check sees list updates done prior to status 550 * bit changes 551 */ 552 if (!test_and_set_bit(action, &subflow->delegated_status)) { 553 /* still on delegated list from previous scheduling */ 554 if (!list_empty(&subflow->delegated_node)) 555 return; 556 557 delegated = this_cpu_ptr(&mptcp_delegated_actions); 558 schedule = list_empty(&delegated->head); 559 list_add_tail(&subflow->delegated_node, &delegated->head); 560 sock_hold(mptcp_subflow_tcp_sock(subflow)); 561 if (schedule) 562 napi_schedule(&delegated->napi); 563 } 564 } 565 566 static inline struct mptcp_subflow_context * 567 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 568 { 569 struct mptcp_subflow_context *ret; 570 571 if (list_empty(&delegated->head)) 572 return NULL; 573 574 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 575 list_del_init(&ret->delegated_node); 576 return ret; 577 } 578 579 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 580 { 581 return !!READ_ONCE(subflow->delegated_status); 582 } 583 584 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action) 585 { 586 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 587 * touching the status bit 588 */ 589 smp_wmb(); 590 clear_bit(action, &subflow->delegated_status); 591 } 592 593 int mptcp_is_enabled(const struct net *net); 594 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 595 int mptcp_is_checksum_enabled(const struct net *net); 596 int mptcp_allow_join_id0(const struct net *net); 597 unsigned int mptcp_stale_loss_cnt(const struct net *net); 598 int mptcp_get_pm_type(const struct net *net); 599 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 600 struct mptcp_options_received *mp_opt); 601 bool __mptcp_retransmit_pending_data(struct sock *sk); 602 void mptcp_check_and_set_pending(struct sock *sk); 603 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 604 bool mptcp_subflow_data_available(struct sock *sk); 605 void __init mptcp_subflow_init(void); 606 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 607 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 608 struct mptcp_subflow_context *subflow); 609 void mptcp_subflow_send_ack(struct sock *ssk); 610 void mptcp_subflow_reset(struct sock *ssk); 611 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 612 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); 613 614 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 615 const struct mptcp_addr_info *b, bool use_port); 616 617 /* called with sk socket lock held */ 618 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 619 const struct mptcp_addr_info *remote); 620 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); 621 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 622 struct sockaddr_storage *addr, 623 unsigned short family); 624 625 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 626 { 627 struct sock *ssk = mptcp_subflow_tcp_sock(subflow); 628 629 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 630 if (subflow->request_join && !subflow->fully_established) 631 return false; 632 633 /* only send if our side has not closed yet */ 634 return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 635 } 636 637 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 638 639 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 640 641 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 642 struct mptcp_subflow_context *ctx) 643 { 644 sk->sk_data_ready = sock_def_readable; 645 sk->sk_state_change = ctx->tcp_state_change; 646 sk->sk_write_space = sk_stream_write_space; 647 sk->sk_error_report = ctx->tcp_error_report; 648 649 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 650 } 651 652 void __init mptcp_proto_init(void); 653 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 654 int __init mptcp_proto_v6_init(void); 655 #endif 656 657 struct sock *mptcp_sk_clone(const struct sock *sk, 658 const struct mptcp_options_received *mp_opt, 659 struct request_sock *req); 660 void mptcp_get_options(const struct sk_buff *skb, 661 struct mptcp_options_received *mp_opt); 662 663 void mptcp_finish_connect(struct sock *sk); 664 void __mptcp_set_connected(struct sock *sk); 665 static inline bool mptcp_is_fully_established(struct sock *sk) 666 { 667 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 668 READ_ONCE(mptcp_sk(sk)->fully_established); 669 } 670 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 671 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 672 bool mptcp_finish_join(struct sock *sk); 673 bool mptcp_schedule_work(struct sock *sk); 674 int mptcp_setsockopt(struct sock *sk, int level, int optname, 675 sockptr_t optval, unsigned int optlen); 676 int mptcp_getsockopt(struct sock *sk, int level, int optname, 677 char __user *optval, int __user *option); 678 679 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 680 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 681 { 682 if (use_64bit) 683 return cur_seq; 684 685 return __mptcp_expand_seq(old_seq, cur_seq); 686 } 687 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 688 void __mptcp_data_acked(struct sock *sk); 689 void __mptcp_error_report(struct sock *sk); 690 void mptcp_subflow_eof(struct sock *sk); 691 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 692 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 693 { 694 return READ_ONCE(msk->snd_data_fin_enable) && 695 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 696 } 697 698 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 699 { 700 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 701 return false; 702 703 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 704 return true; 705 } 706 707 static inline void mptcp_write_space(struct sock *sk) 708 { 709 if (sk_stream_is_writeable(sk)) { 710 /* pairs with memory barrier in mptcp_poll */ 711 smp_mb(); 712 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 713 sk_stream_write_space(sk); 714 } 715 } 716 717 void mptcp_destroy_common(struct mptcp_sock *msk); 718 719 #define MPTCP_TOKEN_MAX_RETRIES 4 720 721 void __init mptcp_token_init(void); 722 static inline void mptcp_token_init_request(struct request_sock *req) 723 { 724 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 725 } 726 727 int mptcp_token_new_request(struct request_sock *req); 728 void mptcp_token_destroy_request(struct request_sock *req); 729 int mptcp_token_new_connect(struct sock *sk); 730 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 731 struct mptcp_sock *msk); 732 bool mptcp_token_exists(u32 token); 733 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 734 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 735 long *s_num); 736 void mptcp_token_destroy(struct mptcp_sock *msk); 737 738 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 739 740 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 741 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 742 743 void __init mptcp_pm_init(void); 744 void mptcp_pm_data_init(struct mptcp_sock *msk); 745 void mptcp_pm_data_reset(struct mptcp_sock *msk); 746 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 747 struct mptcp_addr_info *addr); 748 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 749 bool require_family, 750 struct mptcp_pm_addr_entry *entry); 751 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 752 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 753 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 754 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp); 755 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 756 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 757 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 758 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 759 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk, 760 const struct mptcp_subflow_context *subflow); 761 void mptcp_pm_add_addr_received(const struct sock *ssk, 762 const struct mptcp_addr_info *addr); 763 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 764 const struct mptcp_addr_info *addr); 765 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 766 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 767 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 768 const struct mptcp_rm_list *rm_list); 769 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 770 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 771 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 772 const struct mptcp_pm_addr_entry *entry); 773 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 774 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 775 struct mptcp_pm_add_entry * 776 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 777 const struct mptcp_addr_info *addr, bool check_id); 778 struct mptcp_pm_add_entry * 779 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 780 const struct mptcp_addr_info *addr); 781 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 782 unsigned int id, 783 u8 *flags, int *ifindex); 784 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 785 unsigned int id, 786 u8 *flags, int *ifindex); 787 788 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 789 const struct mptcp_addr_info *addr, 790 bool echo); 791 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 792 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 793 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk, 794 struct list_head *rm_list); 795 796 int mptcp_userspace_pm_append_new_local_addr(struct mptcp_sock *msk, 797 struct mptcp_pm_addr_entry *entry); 798 void mptcp_free_local_addr_list(struct mptcp_sock *msk); 799 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info); 800 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info); 801 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info); 802 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info); 803 804 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 805 const struct sock *ssk, gfp_t gfp); 806 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 807 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 808 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 809 810 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 811 { 812 return READ_ONCE(msk->pm.addr_signal) & 813 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 814 } 815 816 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 817 { 818 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 819 } 820 821 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 822 { 823 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 824 } 825 826 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 827 { 828 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 829 } 830 831 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 832 { 833 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 834 } 835 836 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 837 { 838 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 839 } 840 841 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 842 { 843 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 844 845 if (family == AF_INET6) 846 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 847 if (!echo) 848 len += MPTCPOPT_THMAC_LEN; 849 /* account for 2 trailing 'nop' options */ 850 if (port) 851 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 852 853 return len; 854 } 855 856 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 857 { 858 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 859 return -EINVAL; 860 861 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 862 } 863 864 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 865 unsigned int opt_size, unsigned int remaining, 866 struct mptcp_addr_info *addr, bool *echo, 867 bool *drop_other_suboptions); 868 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 869 struct mptcp_rm_list *rm_list); 870 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 871 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 872 873 void __init mptcp_pm_nl_init(void); 874 void mptcp_pm_nl_work(struct mptcp_sock *msk); 875 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 876 const struct mptcp_rm_list *rm_list); 877 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 878 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 879 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 880 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 881 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 882 883 /* called under PM lock */ 884 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 885 { 886 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 887 WRITE_ONCE(msk->pm.accept_subflow, true); 888 } 889 890 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 891 { 892 spin_lock_bh(&msk->pm.lock); 893 __mptcp_pm_close_subflow(msk); 894 spin_unlock_bh(&msk->pm.lock); 895 } 896 897 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 898 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 899 900 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 901 { 902 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 903 } 904 905 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 906 907 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 908 { 909 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 910 } 911 912 static inline bool mptcp_check_fallback(const struct sock *sk) 913 { 914 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 915 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 916 917 return __mptcp_check_fallback(msk); 918 } 919 920 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 921 { 922 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 923 pr_debug("TCP fallback already done (msk=%p)", msk); 924 return; 925 } 926 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 927 } 928 929 static inline void mptcp_do_fallback(struct sock *sk) 930 { 931 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 932 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 933 934 __mptcp_do_fallback(msk); 935 } 936 937 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 938 939 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 940 { 941 struct mptcp_ext *mpext; 942 943 mpext = skb ? mptcp_get_ext(skb) : NULL; 944 if (mpext && mpext->infinite_map) 945 return true; 946 947 return false; 948 } 949 950 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 951 { 952 return (subflow->request_mptcp || subflow->request_join); 953 } 954 955 static inline bool subflow_simultaneous_connect(struct sock *sk) 956 { 957 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 958 959 return sk->sk_state == TCP_ESTABLISHED && 960 is_active_ssk(subflow) && 961 !subflow->conn_finished; 962 } 963 964 #ifdef CONFIG_SYN_COOKIES 965 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 966 struct sk_buff *skb); 967 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 968 struct sk_buff *skb); 969 void __init mptcp_join_cookie_init(void); 970 #else 971 static inline void 972 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 973 struct sk_buff *skb) {} 974 static inline bool 975 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 976 struct sk_buff *skb) 977 { 978 return false; 979 } 980 981 static inline void mptcp_join_cookie_init(void) {} 982 #endif 983 984 #endif /* __MPTCP_PROTOCOL_H */ 985