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_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 enum mptcp_pm_type { 187 MPTCP_PM_TYPE_KERNEL = 0, 188 MPTCP_PM_TYPE_USERSPACE, 189 190 __MPTCP_PM_TYPE_NR, 191 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 192 }; 193 194 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 195 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 196 197 enum mptcp_addr_signal_status { 198 MPTCP_ADD_ADDR_SIGNAL, 199 MPTCP_ADD_ADDR_ECHO, 200 MPTCP_RM_ADDR_SIGNAL, 201 }; 202 203 /* max value of mptcp_addr_info.id */ 204 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 205 206 struct mptcp_pm_data { 207 struct mptcp_addr_info local; 208 struct mptcp_addr_info remote; 209 struct list_head anno_list; 210 struct list_head userspace_pm_local_addr_list; 211 212 spinlock_t lock; /*protects the whole PM data */ 213 214 u8 addr_signal; 215 bool server_side; 216 bool work_pending; 217 bool accept_addr; 218 bool accept_subflow; 219 bool remote_deny_join_id0; 220 u8 add_addr_signaled; 221 u8 add_addr_accepted; 222 u8 local_addr_used; 223 u8 pm_type; 224 u8 subflows; 225 u8 status; 226 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 227 struct mptcp_rm_list rm_list_tx; 228 struct mptcp_rm_list rm_list_rx; 229 }; 230 231 struct mptcp_pm_addr_entry { 232 struct list_head list; 233 struct mptcp_addr_info addr; 234 u8 flags; 235 int ifindex; 236 struct socket *lsk; 237 }; 238 239 struct mptcp_data_frag { 240 struct list_head list; 241 u64 data_seq; 242 u16 data_len; 243 u16 offset; 244 u16 overhead; 245 u16 already_sent; 246 struct page *page; 247 }; 248 249 /* MPTCP connection sock */ 250 struct mptcp_sock { 251 /* inet_connection_sock must be the first member */ 252 struct inet_connection_sock sk; 253 u64 local_key; 254 u64 remote_key; 255 u64 write_seq; 256 u64 snd_nxt; 257 u64 ack_seq; 258 atomic64_t rcv_wnd_sent; 259 u64 rcv_data_fin_seq; 260 int rmem_fwd_alloc; 261 struct sock *last_snd; 262 int snd_burst; 263 int old_wspace; 264 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 265 * recovery related fields are under data_lock 266 * protection 267 */ 268 u64 snd_una; 269 u64 wnd_end; 270 unsigned long timer_ival; 271 u32 token; 272 int rmem_released; 273 unsigned long flags; 274 unsigned long cb_flags; 275 unsigned long push_pending; 276 bool recovery; /* closing subflow write queue reinjected */ 277 bool can_ack; 278 bool fully_established; 279 bool rcv_data_fin; 280 bool snd_data_fin_enable; 281 bool rcv_fastclose; 282 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 283 bool csum_enabled; 284 bool allow_infinite_fallback; 285 u8 mpc_endpoint_id; 286 u8 recvmsg_inq:1, 287 cork:1, 288 nodelay:1, 289 is_sendmsg:1; 290 int connect_flags; 291 struct work_struct work; 292 struct sk_buff *ooo_last_skb; 293 struct rb_root out_of_order_queue; 294 struct sk_buff_head receive_queue; 295 struct list_head conn_list; 296 struct list_head rtx_queue; 297 struct mptcp_data_frag *first_pending; 298 struct list_head join_list; 299 struct socket *subflow; /* outgoing connect/listener/!mp_capable */ 300 struct sock *first; 301 struct mptcp_pm_data pm; 302 struct { 303 u32 space; /* bytes copied in last measurement window */ 304 u32 copied; /* bytes copied in this measurement window */ 305 u64 time; /* start time of measurement window */ 306 u64 rtt_us; /* last maximum rtt of subflows */ 307 } rcvq_space; 308 309 u32 setsockopt_seq; 310 char ca_name[TCP_CA_NAME_MAX]; 311 struct mptcp_sock *dl_next; 312 }; 313 314 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 315 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 316 317 #define mptcp_for_each_subflow(__msk, __subflow) \ 318 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 319 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 320 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 321 322 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 323 { 324 sock_owned_by_me((const struct sock *)msk); 325 } 326 327 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 328 { 329 return (struct mptcp_sock *)sk; 330 } 331 332 /* the msk socket don't use the backlog, also account for the bulk 333 * free memory 334 */ 335 static inline int __mptcp_rmem(const struct sock *sk) 336 { 337 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 338 } 339 340 static inline int __mptcp_space(const struct sock *sk) 341 { 342 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 343 } 344 345 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 346 { 347 const struct mptcp_sock *msk = mptcp_sk(sk); 348 349 return READ_ONCE(msk->first_pending); 350 } 351 352 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 353 { 354 struct mptcp_sock *msk = mptcp_sk(sk); 355 struct mptcp_data_frag *cur; 356 357 cur = msk->first_pending; 358 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 359 list_next_entry(cur, list); 360 } 361 362 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 363 { 364 struct mptcp_sock *msk = mptcp_sk(sk); 365 366 if (!msk->first_pending) 367 return NULL; 368 369 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 370 return NULL; 371 372 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 373 } 374 375 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 376 { 377 struct mptcp_sock *msk = mptcp_sk(sk); 378 379 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 380 return NULL; 381 382 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 383 } 384 385 struct csum_pseudo_header { 386 __be64 data_seq; 387 __be32 subflow_seq; 388 __be16 data_len; 389 __sum16 csum; 390 }; 391 392 struct mptcp_subflow_request_sock { 393 struct tcp_request_sock sk; 394 u16 mp_capable : 1, 395 mp_join : 1, 396 backup : 1, 397 csum_reqd : 1, 398 allow_join_id0 : 1; 399 u8 local_id; 400 u8 remote_id; 401 u64 local_key; 402 u64 idsn; 403 u32 token; 404 u32 ssn_offset; 405 u64 thmac; 406 u32 local_nonce; 407 u32 remote_nonce; 408 struct mptcp_sock *msk; 409 struct hlist_nulls_node token_node; 410 }; 411 412 static inline struct mptcp_subflow_request_sock * 413 mptcp_subflow_rsk(const struct request_sock *rsk) 414 { 415 return (struct mptcp_subflow_request_sock *)rsk; 416 } 417 418 enum mptcp_data_avail { 419 MPTCP_SUBFLOW_NODATA, 420 MPTCP_SUBFLOW_DATA_AVAIL, 421 }; 422 423 struct mptcp_delegated_action { 424 struct napi_struct napi; 425 struct list_head head; 426 }; 427 428 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 429 430 #define MPTCP_DELEGATE_SEND 0 431 #define MPTCP_DELEGATE_ACK 1 432 433 /* MPTCP subflow context */ 434 struct mptcp_subflow_context { 435 struct list_head node;/* conn_list of subflows */ 436 437 struct_group(reset, 438 439 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 440 u64 local_key; 441 u64 remote_key; 442 u64 idsn; 443 u64 map_seq; 444 u32 snd_isn; 445 u32 token; 446 u32 rel_write_seq; 447 u32 map_subflow_seq; 448 u32 ssn_offset; 449 u32 map_data_len; 450 __wsum map_data_csum; 451 u32 map_csum_len; 452 u32 request_mptcp : 1, /* send MP_CAPABLE */ 453 request_join : 1, /* send MP_JOIN */ 454 request_bkup : 1, 455 mp_capable : 1, /* remote is MPTCP capable */ 456 mp_join : 1, /* remote is JOINing */ 457 fully_established : 1, /* path validated */ 458 pm_notified : 1, /* PM hook called for established status */ 459 conn_finished : 1, 460 map_valid : 1, 461 map_csum_reqd : 1, 462 map_data_fin : 1, 463 mpc_map : 1, 464 backup : 1, 465 send_mp_prio : 1, 466 send_mp_fail : 1, 467 send_fastclose : 1, 468 send_infinite_map : 1, 469 rx_eof : 1, 470 can_ack : 1, /* only after processing the remote a key */ 471 disposable : 1, /* ctx can be free at ulp release time */ 472 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 473 local_id_valid : 1, /* local_id is correctly initialized */ 474 valid_csum_seen : 1; /* at least one csum validated */ 475 enum mptcp_data_avail data_avail; 476 u32 remote_nonce; 477 u64 thmac; 478 u32 local_nonce; 479 u32 remote_token; 480 u8 hmac[MPTCPOPT_HMAC_LEN]; 481 u8 local_id; 482 u8 remote_id; 483 u8 reset_seen:1; 484 u8 reset_transient:1; 485 u8 reset_reason:4; 486 u8 stale_count; 487 488 long delegated_status; 489 unsigned long fail_tout; 490 491 ); 492 493 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 494 495 u32 setsockopt_seq; 496 u32 stale_rcv_tstamp; 497 498 struct sock *tcp_sock; /* tcp sk backpointer */ 499 struct sock *conn; /* parent mptcp_sock */ 500 const struct inet_connection_sock_af_ops *icsk_af_ops; 501 void (*tcp_state_change)(struct sock *sk); 502 void (*tcp_error_report)(struct sock *sk); 503 504 struct rcu_head rcu; 505 }; 506 507 static inline struct mptcp_subflow_context * 508 mptcp_subflow_ctx(const struct sock *sk) 509 { 510 struct inet_connection_sock *icsk = inet_csk(sk); 511 512 /* Use RCU on icsk_ulp_data only for sock diag code */ 513 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 514 } 515 516 static inline struct sock * 517 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 518 { 519 return subflow->tcp_sock; 520 } 521 522 static inline void 523 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 524 { 525 memset(&subflow->reset, 0, sizeof(subflow->reset)); 526 subflow->request_mptcp = 1; 527 } 528 529 static inline u64 530 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 531 { 532 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 533 subflow->ssn_offset - 534 subflow->map_subflow_seq; 535 } 536 537 static inline u64 538 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 539 { 540 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 541 } 542 543 void mptcp_subflow_process_delegated(struct sock *ssk); 544 545 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 546 { 547 struct mptcp_delegated_action *delegated; 548 bool schedule; 549 550 /* the caller held the subflow bh socket lock */ 551 lockdep_assert_in_softirq(); 552 553 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 554 * ensures the below list check sees list updates done prior to status 555 * bit changes 556 */ 557 if (!test_and_set_bit(action, &subflow->delegated_status)) { 558 /* still on delegated list from previous scheduling */ 559 if (!list_empty(&subflow->delegated_node)) 560 return; 561 562 delegated = this_cpu_ptr(&mptcp_delegated_actions); 563 schedule = list_empty(&delegated->head); 564 list_add_tail(&subflow->delegated_node, &delegated->head); 565 sock_hold(mptcp_subflow_tcp_sock(subflow)); 566 if (schedule) 567 napi_schedule(&delegated->napi); 568 } 569 } 570 571 static inline struct mptcp_subflow_context * 572 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 573 { 574 struct mptcp_subflow_context *ret; 575 576 if (list_empty(&delegated->head)) 577 return NULL; 578 579 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 580 list_del_init(&ret->delegated_node); 581 return ret; 582 } 583 584 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 585 { 586 return !!READ_ONCE(subflow->delegated_status); 587 } 588 589 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action) 590 { 591 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 592 * touching the status bit 593 */ 594 smp_wmb(); 595 clear_bit(action, &subflow->delegated_status); 596 } 597 598 int mptcp_is_enabled(const struct net *net); 599 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 600 int mptcp_is_checksum_enabled(const struct net *net); 601 int mptcp_allow_join_id0(const struct net *net); 602 unsigned int mptcp_stale_loss_cnt(const struct net *net); 603 int mptcp_get_pm_type(const struct net *net); 604 void mptcp_copy_inaddrs(struct sock *msk, const struct sock *ssk); 605 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 606 struct mptcp_options_received *mp_opt); 607 bool __mptcp_retransmit_pending_data(struct sock *sk); 608 void mptcp_check_and_set_pending(struct sock *sk); 609 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 610 bool mptcp_subflow_data_available(struct sock *sk); 611 void __init mptcp_subflow_init(void); 612 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 613 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 614 struct mptcp_subflow_context *subflow); 615 void __mptcp_subflow_send_ack(struct sock *ssk); 616 void mptcp_subflow_reset(struct sock *ssk); 617 void mptcp_subflow_queue_clean(struct sock *ssk); 618 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 619 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); 620 bool __mptcp_close(struct sock *sk, long timeout); 621 void mptcp_cancel_work(struct sock *sk); 622 623 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 624 const struct mptcp_addr_info *b, bool use_port); 625 626 /* called with sk socket lock held */ 627 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 628 const struct mptcp_addr_info *remote); 629 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); 630 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 631 struct sockaddr_storage *addr, 632 unsigned short family); 633 634 static inline bool __tcp_can_send(const struct sock *ssk) 635 { 636 /* only send if our side has not closed yet */ 637 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 638 } 639 640 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 641 { 642 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 643 if (subflow->request_join && !subflow->fully_established) 644 return false; 645 646 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 647 } 648 649 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 650 651 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 652 653 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 654 struct mptcp_subflow_context *ctx) 655 { 656 sk->sk_data_ready = sock_def_readable; 657 sk->sk_state_change = ctx->tcp_state_change; 658 sk->sk_write_space = sk_stream_write_space; 659 sk->sk_error_report = ctx->tcp_error_report; 660 661 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 662 } 663 664 void __init mptcp_proto_init(void); 665 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 666 int __init mptcp_proto_v6_init(void); 667 #endif 668 669 struct sock *mptcp_sk_clone(const struct sock *sk, 670 const struct mptcp_options_received *mp_opt, 671 struct request_sock *req); 672 void mptcp_get_options(const struct sk_buff *skb, 673 struct mptcp_options_received *mp_opt); 674 675 void mptcp_finish_connect(struct sock *sk); 676 void __mptcp_set_connected(struct sock *sk); 677 void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout); 678 static inline bool mptcp_is_fully_established(struct sock *sk) 679 { 680 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 681 READ_ONCE(mptcp_sk(sk)->fully_established); 682 } 683 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 684 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 685 bool mptcp_finish_join(struct sock *sk); 686 bool mptcp_schedule_work(struct sock *sk); 687 int mptcp_setsockopt(struct sock *sk, int level, int optname, 688 sockptr_t optval, unsigned int optlen); 689 int mptcp_getsockopt(struct sock *sk, int level, int optname, 690 char __user *optval, int __user *option); 691 692 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 693 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 694 { 695 if (use_64bit) 696 return cur_seq; 697 698 return __mptcp_expand_seq(old_seq, cur_seq); 699 } 700 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 701 void __mptcp_data_acked(struct sock *sk); 702 void __mptcp_error_report(struct sock *sk); 703 void mptcp_subflow_eof(struct sock *sk); 704 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 705 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 706 { 707 return READ_ONCE(msk->snd_data_fin_enable) && 708 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 709 } 710 711 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 712 { 713 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 714 return false; 715 716 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 717 return true; 718 } 719 720 static inline void mptcp_write_space(struct sock *sk) 721 { 722 if (sk_stream_is_writeable(sk)) { 723 /* pairs with memory barrier in mptcp_poll */ 724 smp_mb(); 725 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 726 sk_stream_write_space(sk); 727 } 728 } 729 730 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 731 732 #define MPTCP_TOKEN_MAX_RETRIES 4 733 734 void __init mptcp_token_init(void); 735 static inline void mptcp_token_init_request(struct request_sock *req) 736 { 737 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 738 } 739 740 int mptcp_token_new_request(struct request_sock *req); 741 void mptcp_token_destroy_request(struct request_sock *req); 742 int mptcp_token_new_connect(struct sock *sk); 743 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 744 struct mptcp_sock *msk); 745 bool mptcp_token_exists(u32 token); 746 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 747 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 748 long *s_num); 749 void mptcp_token_destroy(struct mptcp_sock *msk); 750 751 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 752 753 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 754 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 755 756 void __init mptcp_pm_init(void); 757 void mptcp_pm_data_init(struct mptcp_sock *msk); 758 void mptcp_pm_data_reset(struct mptcp_sock *msk); 759 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 760 struct mptcp_addr_info *addr); 761 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 762 bool require_family, 763 struct mptcp_pm_addr_entry *entry); 764 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 765 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 766 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 767 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp); 768 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 769 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 770 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 771 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 772 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk, 773 const struct mptcp_subflow_context *subflow); 774 void mptcp_pm_add_addr_received(const struct sock *ssk, 775 const struct mptcp_addr_info *addr); 776 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 777 const struct mptcp_addr_info *addr); 778 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 779 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 780 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 781 const struct mptcp_rm_list *rm_list); 782 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 783 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 784 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 785 struct mptcp_addr_info *addr, 786 struct mptcp_addr_info *rem, 787 u8 bkup); 788 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 789 const struct mptcp_pm_addr_entry *entry); 790 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 791 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 792 struct mptcp_pm_add_entry * 793 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 794 const struct mptcp_addr_info *addr, bool check_id); 795 struct mptcp_pm_add_entry * 796 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 797 const struct mptcp_addr_info *addr); 798 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 799 unsigned int id, 800 u8 *flags, int *ifindex); 801 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 802 unsigned int id, 803 u8 *flags, int *ifindex); 804 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token, 805 struct mptcp_pm_addr_entry *loc, 806 struct mptcp_pm_addr_entry *rem, u8 bkup); 807 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 808 const struct mptcp_addr_info *addr, 809 bool echo); 810 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 811 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 812 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk, 813 struct list_head *rm_list); 814 815 int mptcp_userspace_pm_append_new_local_addr(struct mptcp_sock *msk, 816 struct mptcp_pm_addr_entry *entry); 817 void mptcp_free_local_addr_list(struct mptcp_sock *msk); 818 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info); 819 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info); 820 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info); 821 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info); 822 823 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 824 const struct sock *ssk, gfp_t gfp); 825 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 826 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 827 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 828 829 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 830 { 831 return READ_ONCE(msk->pm.addr_signal) & 832 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 833 } 834 835 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 836 { 837 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 838 } 839 840 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 841 { 842 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 843 } 844 845 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 846 { 847 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 848 } 849 850 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 851 { 852 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 853 } 854 855 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 856 { 857 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 858 } 859 860 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 861 { 862 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 863 864 if (family == AF_INET6) 865 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 866 if (!echo) 867 len += MPTCPOPT_THMAC_LEN; 868 /* account for 2 trailing 'nop' options */ 869 if (port) 870 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 871 872 return len; 873 } 874 875 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 876 { 877 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 878 return -EINVAL; 879 880 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 881 } 882 883 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 884 unsigned int opt_size, unsigned int remaining, 885 struct mptcp_addr_info *addr, bool *echo, 886 bool *drop_other_suboptions); 887 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 888 struct mptcp_rm_list *rm_list); 889 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 890 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 891 892 void __init mptcp_pm_nl_init(void); 893 void mptcp_pm_nl_work(struct mptcp_sock *msk); 894 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 895 const struct mptcp_rm_list *rm_list); 896 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 897 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 898 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 899 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 900 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 901 902 /* called under PM lock */ 903 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 904 { 905 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 906 WRITE_ONCE(msk->pm.accept_subflow, true); 907 } 908 909 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 910 { 911 spin_lock_bh(&msk->pm.lock); 912 __mptcp_pm_close_subflow(msk); 913 spin_unlock_bh(&msk->pm.lock); 914 } 915 916 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 917 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 918 919 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 920 { 921 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 922 } 923 924 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 925 926 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 927 { 928 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 929 } 930 931 static inline bool mptcp_check_fallback(const struct sock *sk) 932 { 933 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 934 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 935 936 return __mptcp_check_fallback(msk); 937 } 938 939 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 940 { 941 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 942 pr_debug("TCP fallback already done (msk=%p)", msk); 943 return; 944 } 945 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 946 } 947 948 static inline void mptcp_do_fallback(struct sock *ssk) 949 { 950 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 951 struct sock *sk = subflow->conn; 952 struct mptcp_sock *msk; 953 954 msk = mptcp_sk(sk); 955 __mptcp_do_fallback(msk); 956 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 957 gfp_t saved_allocation = ssk->sk_allocation; 958 959 /* we are in a atomic (BH) scope, override ssk default for data 960 * fin allocation 961 */ 962 ssk->sk_allocation = GFP_ATOMIC; 963 ssk->sk_shutdown |= SEND_SHUTDOWN; 964 tcp_shutdown(ssk, SEND_SHUTDOWN); 965 ssk->sk_allocation = saved_allocation; 966 } 967 } 968 969 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 970 971 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 972 { 973 struct mptcp_ext *mpext; 974 975 mpext = skb ? mptcp_get_ext(skb) : NULL; 976 if (mpext && mpext->infinite_map) 977 return true; 978 979 return false; 980 } 981 982 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 983 { 984 return (subflow->request_mptcp || subflow->request_join); 985 } 986 987 static inline bool subflow_simultaneous_connect(struct sock *sk) 988 { 989 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 990 991 return sk->sk_state == TCP_ESTABLISHED && 992 is_active_ssk(subflow) && 993 !subflow->conn_finished; 994 } 995 996 #ifdef CONFIG_SYN_COOKIES 997 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 998 struct sk_buff *skb); 999 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1000 struct sk_buff *skb); 1001 void __init mptcp_join_cookie_init(void); 1002 #else 1003 static inline void 1004 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1005 struct sk_buff *skb) {} 1006 static inline bool 1007 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1008 struct sk_buff *skb) 1009 { 1010 return false; 1011 } 1012 1013 static inline void mptcp_join_cookie_init(void) {} 1014 #endif 1015 1016 #endif /* __MPTCP_PROTOCOL_H */ 1017