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_FALLBACK_DONE 4 117 #define MPTCP_WORK_CLOSE_SUBFLOW 5 118 119 /* MPTCP socket release cb flags */ 120 #define MPTCP_PUSH_PENDING 1 121 #define MPTCP_CLEAN_UNA 2 122 #define MPTCP_ERROR_REPORT 3 123 #define MPTCP_RETRANSMIT 4 124 #define MPTCP_FLUSH_JOIN_LIST 5 125 #define MPTCP_CONNECTED 6 126 #define MPTCP_RESET_SCHEDULER 7 127 128 struct mptcp_skb_cb { 129 u64 map_seq; 130 u64 end_seq; 131 u32 offset; 132 u8 has_rxtstamp:1; 133 }; 134 135 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0])) 136 137 static inline bool before64(__u64 seq1, __u64 seq2) 138 { 139 return (__s64)(seq1 - seq2) < 0; 140 } 141 142 #define after64(seq2, seq1) before64(seq1, seq2) 143 144 struct mptcp_options_received { 145 u64 sndr_key; 146 u64 rcvr_key; 147 u64 data_ack; 148 u64 data_seq; 149 u32 subflow_seq; 150 u16 data_len; 151 __sum16 csum; 152 u16 suboptions; 153 u32 token; 154 u32 nonce; 155 u16 use_map:1, 156 dsn64:1, 157 data_fin:1, 158 use_ack:1, 159 ack64:1, 160 mpc_map:1, 161 reset_reason:4, 162 reset_transient:1, 163 echo:1, 164 backup:1, 165 deny_join_id0:1, 166 __unused:2; 167 u8 join_id; 168 u64 thmac; 169 u8 hmac[MPTCPOPT_HMAC_LEN]; 170 struct mptcp_addr_info addr; 171 struct mptcp_rm_list rm_list; 172 u64 ahmac; 173 u64 fail_seq; 174 }; 175 176 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 177 { 178 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 179 ((nib & 0xF) << 8) | field); 180 } 181 182 enum mptcp_pm_status { 183 MPTCP_PM_ADD_ADDR_RECEIVED, 184 MPTCP_PM_ADD_ADDR_SEND_ACK, 185 MPTCP_PM_RM_ADDR_RECEIVED, 186 MPTCP_PM_ESTABLISHED, 187 MPTCP_PM_SUBFLOW_ESTABLISHED, 188 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 189 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 190 * accounted int id_avail_bitmap 191 */ 192 }; 193 194 enum mptcp_pm_type { 195 MPTCP_PM_TYPE_KERNEL = 0, 196 MPTCP_PM_TYPE_USERSPACE, 197 198 __MPTCP_PM_TYPE_NR, 199 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 200 }; 201 202 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 203 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 204 205 enum mptcp_addr_signal_status { 206 MPTCP_ADD_ADDR_SIGNAL, 207 MPTCP_ADD_ADDR_ECHO, 208 MPTCP_RM_ADDR_SIGNAL, 209 }; 210 211 /* max value of mptcp_addr_info.id */ 212 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 213 214 struct mptcp_pm_data { 215 struct mptcp_addr_info local; 216 struct mptcp_addr_info remote; 217 struct list_head anno_list; 218 struct list_head userspace_pm_local_addr_list; 219 220 spinlock_t lock; /*protects the whole PM data */ 221 222 u8 addr_signal; 223 bool server_side; 224 bool work_pending; 225 bool accept_addr; 226 bool accept_subflow; 227 bool remote_deny_join_id0; 228 u8 add_addr_signaled; 229 u8 add_addr_accepted; 230 u8 local_addr_used; 231 u8 pm_type; 232 u8 subflows; 233 u8 status; 234 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 235 struct mptcp_rm_list rm_list_tx; 236 struct mptcp_rm_list rm_list_rx; 237 }; 238 239 struct mptcp_pm_addr_entry { 240 struct list_head list; 241 struct mptcp_addr_info addr; 242 u8 flags; 243 int ifindex; 244 struct socket *lsk; 245 }; 246 247 struct mptcp_data_frag { 248 struct list_head list; 249 u64 data_seq; 250 u16 data_len; 251 u16 offset; 252 u16 overhead; 253 u16 already_sent; 254 struct page *page; 255 }; 256 257 /* MPTCP connection sock */ 258 struct mptcp_sock { 259 /* inet_connection_sock must be the first member */ 260 struct inet_connection_sock sk; 261 u64 local_key; 262 u64 remote_key; 263 u64 write_seq; 264 u64 bytes_sent; 265 u64 snd_nxt; 266 u64 bytes_received; 267 u64 ack_seq; 268 atomic64_t rcv_wnd_sent; 269 u64 rcv_data_fin_seq; 270 u64 bytes_retrans; 271 int rmem_fwd_alloc; 272 struct sock *last_snd; 273 int snd_burst; 274 int old_wspace; 275 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 276 * recovery related fields are under data_lock 277 * protection 278 */ 279 u64 bytes_acked; 280 u64 snd_una; 281 u64 wnd_end; 282 unsigned long timer_ival; 283 u32 token; 284 int rmem_released; 285 unsigned long flags; 286 unsigned long cb_flags; 287 unsigned long push_pending; 288 bool recovery; /* closing subflow write queue reinjected */ 289 bool can_ack; 290 bool fully_established; 291 bool rcv_data_fin; 292 bool snd_data_fin_enable; 293 bool rcv_fastclose; 294 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 295 bool csum_enabled; 296 bool allow_infinite_fallback; 297 u8 mpc_endpoint_id; 298 u8 recvmsg_inq:1, 299 cork:1, 300 nodelay:1, 301 fastopening:1, 302 in_accept_queue:1; 303 struct work_struct work; 304 struct sk_buff *ooo_last_skb; 305 struct rb_root out_of_order_queue; 306 struct sk_buff_head receive_queue; 307 struct list_head conn_list; 308 struct list_head rtx_queue; 309 struct mptcp_data_frag *first_pending; 310 struct list_head join_list; 311 struct socket *subflow; /* outgoing connect/listener/!mp_capable 312 * The mptcp ops can safely dereference, using suitable 313 * ONCE annotation, the subflow outside the socket 314 * lock as such sock is freed after close(). 315 */ 316 struct sock *first; 317 struct mptcp_pm_data pm; 318 struct { 319 u32 space; /* bytes copied in last measurement window */ 320 u32 copied; /* bytes copied in this measurement window */ 321 u64 time; /* start time of measurement window */ 322 u64 rtt_us; /* last maximum rtt of subflows */ 323 } rcvq_space; 324 325 u32 subflow_id; 326 u32 setsockopt_seq; 327 char ca_name[TCP_CA_NAME_MAX]; 328 struct mptcp_sock *dl_next; 329 }; 330 331 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 332 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 333 334 #define mptcp_for_each_subflow(__msk, __subflow) \ 335 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 336 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 337 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 338 339 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 340 { 341 sock_owned_by_me((const struct sock *)msk); 342 } 343 344 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 345 346 /* the msk socket don't use the backlog, also account for the bulk 347 * free memory 348 */ 349 static inline int __mptcp_rmem(const struct sock *sk) 350 { 351 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 352 } 353 354 static inline int __mptcp_space(const struct sock *sk) 355 { 356 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 357 } 358 359 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 360 { 361 const struct mptcp_sock *msk = mptcp_sk(sk); 362 363 return READ_ONCE(msk->first_pending); 364 } 365 366 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 367 { 368 struct mptcp_sock *msk = mptcp_sk(sk); 369 struct mptcp_data_frag *cur; 370 371 cur = msk->first_pending; 372 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 373 list_next_entry(cur, list); 374 } 375 376 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 377 { 378 const struct mptcp_sock *msk = mptcp_sk(sk); 379 380 if (!msk->first_pending) 381 return NULL; 382 383 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 384 return NULL; 385 386 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 387 } 388 389 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 390 { 391 struct mptcp_sock *msk = mptcp_sk(sk); 392 393 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 394 return NULL; 395 396 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 397 } 398 399 struct csum_pseudo_header { 400 __be64 data_seq; 401 __be32 subflow_seq; 402 __be16 data_len; 403 __sum16 csum; 404 }; 405 406 struct mptcp_subflow_request_sock { 407 struct tcp_request_sock sk; 408 u16 mp_capable : 1, 409 mp_join : 1, 410 backup : 1, 411 csum_reqd : 1, 412 allow_join_id0 : 1; 413 u8 local_id; 414 u8 remote_id; 415 u64 local_key; 416 u64 idsn; 417 u32 token; 418 u32 ssn_offset; 419 u64 thmac; 420 u32 local_nonce; 421 u32 remote_nonce; 422 struct mptcp_sock *msk; 423 struct hlist_nulls_node token_node; 424 }; 425 426 static inline struct mptcp_subflow_request_sock * 427 mptcp_subflow_rsk(const struct request_sock *rsk) 428 { 429 return (struct mptcp_subflow_request_sock *)rsk; 430 } 431 432 enum mptcp_data_avail { 433 MPTCP_SUBFLOW_NODATA, 434 MPTCP_SUBFLOW_DATA_AVAIL, 435 }; 436 437 struct mptcp_delegated_action { 438 struct napi_struct napi; 439 struct list_head head; 440 }; 441 442 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 443 444 #define MPTCP_DELEGATE_SEND 0 445 #define MPTCP_DELEGATE_ACK 1 446 447 /* MPTCP subflow context */ 448 struct mptcp_subflow_context { 449 struct list_head node;/* conn_list of subflows */ 450 451 struct_group(reset, 452 453 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 454 u64 local_key; 455 u64 remote_key; 456 u64 idsn; 457 u64 map_seq; 458 u32 snd_isn; 459 u32 token; 460 u32 rel_write_seq; 461 u32 map_subflow_seq; 462 u32 ssn_offset; 463 u32 map_data_len; 464 __wsum map_data_csum; 465 u32 map_csum_len; 466 u32 request_mptcp : 1, /* send MP_CAPABLE */ 467 request_join : 1, /* send MP_JOIN */ 468 request_bkup : 1, 469 mp_capable : 1, /* remote is MPTCP capable */ 470 mp_join : 1, /* remote is JOINing */ 471 fully_established : 1, /* path validated */ 472 pm_notified : 1, /* PM hook called for established status */ 473 conn_finished : 1, 474 map_valid : 1, 475 map_csum_reqd : 1, 476 map_data_fin : 1, 477 mpc_map : 1, 478 backup : 1, 479 send_mp_prio : 1, 480 send_mp_fail : 1, 481 send_fastclose : 1, 482 send_infinite_map : 1, 483 remote_key_valid : 1, /* received the peer key from */ 484 disposable : 1, /* ctx can be free at ulp release time */ 485 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 486 local_id_valid : 1, /* local_id is correctly initialized */ 487 valid_csum_seen : 1, /* at least one csum validated */ 488 is_mptfo : 1, /* subflow is doing TFO */ 489 __unused : 9; 490 enum mptcp_data_avail data_avail; 491 u32 remote_nonce; 492 u64 thmac; 493 u32 local_nonce; 494 u32 remote_token; 495 union { 496 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 497 u64 iasn; /* initial ack sequence number, MPC subflows only */ 498 }; 499 u8 local_id; 500 u8 remote_id; 501 u8 reset_seen:1; 502 u8 reset_transient:1; 503 u8 reset_reason:4; 504 u8 stale_count; 505 506 u32 subflow_id; 507 508 long delegated_status; 509 unsigned long fail_tout; 510 511 ); 512 513 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 514 515 u32 setsockopt_seq; 516 u32 stale_rcv_tstamp; 517 518 struct sock *tcp_sock; /* tcp sk backpointer */ 519 struct sock *conn; /* parent mptcp_sock */ 520 const struct inet_connection_sock_af_ops *icsk_af_ops; 521 void (*tcp_state_change)(struct sock *sk); 522 void (*tcp_error_report)(struct sock *sk); 523 524 struct rcu_head rcu; 525 }; 526 527 static inline struct mptcp_subflow_context * 528 mptcp_subflow_ctx(const struct sock *sk) 529 { 530 struct inet_connection_sock *icsk = inet_csk(sk); 531 532 /* Use RCU on icsk_ulp_data only for sock diag code */ 533 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 534 } 535 536 static inline struct sock * 537 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 538 { 539 return subflow->tcp_sock; 540 } 541 542 static inline void 543 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 544 { 545 memset(&subflow->reset, 0, sizeof(subflow->reset)); 546 subflow->request_mptcp = 1; 547 } 548 549 static inline u64 550 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 551 { 552 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 553 subflow->ssn_offset - 554 subflow->map_subflow_seq; 555 } 556 557 static inline u64 558 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 559 { 560 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 561 } 562 563 void mptcp_subflow_process_delegated(struct sock *ssk); 564 565 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 566 { 567 struct mptcp_delegated_action *delegated; 568 bool schedule; 569 570 /* the caller held the subflow bh socket lock */ 571 lockdep_assert_in_softirq(); 572 573 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 574 * ensures the below list check sees list updates done prior to status 575 * bit changes 576 */ 577 if (!test_and_set_bit(action, &subflow->delegated_status)) { 578 /* still on delegated list from previous scheduling */ 579 if (!list_empty(&subflow->delegated_node)) 580 return; 581 582 delegated = this_cpu_ptr(&mptcp_delegated_actions); 583 schedule = list_empty(&delegated->head); 584 list_add_tail(&subflow->delegated_node, &delegated->head); 585 sock_hold(mptcp_subflow_tcp_sock(subflow)); 586 if (schedule) 587 napi_schedule(&delegated->napi); 588 } 589 } 590 591 static inline struct mptcp_subflow_context * 592 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 593 { 594 struct mptcp_subflow_context *ret; 595 596 if (list_empty(&delegated->head)) 597 return NULL; 598 599 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 600 list_del_init(&ret->delegated_node); 601 return ret; 602 } 603 604 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 605 { 606 return !!READ_ONCE(subflow->delegated_status); 607 } 608 609 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action) 610 { 611 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 612 * touching the status bit 613 */ 614 smp_wmb(); 615 clear_bit(action, &subflow->delegated_status); 616 } 617 618 int mptcp_is_enabled(const struct net *net); 619 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 620 int mptcp_is_checksum_enabled(const struct net *net); 621 int mptcp_allow_join_id0(const struct net *net); 622 unsigned int mptcp_stale_loss_cnt(const struct net *net); 623 int mptcp_get_pm_type(const struct net *net); 624 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 625 const struct mptcp_options_received *mp_opt); 626 bool __mptcp_retransmit_pending_data(struct sock *sk); 627 void mptcp_check_and_set_pending(struct sock *sk); 628 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 629 bool mptcp_subflow_data_available(struct sock *sk); 630 void __init mptcp_subflow_init(void); 631 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 632 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 633 struct mptcp_subflow_context *subflow); 634 void __mptcp_subflow_send_ack(struct sock *ssk); 635 void mptcp_subflow_reset(struct sock *ssk); 636 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 637 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 638 struct socket *__mptcp_nmpc_socket(struct mptcp_sock *msk); 639 bool __mptcp_close(struct sock *sk, long timeout); 640 void mptcp_cancel_work(struct sock *sk); 641 void __mptcp_unaccepted_force_close(struct sock *sk); 642 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk); 643 644 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 645 const struct mptcp_addr_info *b, bool use_port); 646 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr); 647 648 /* called with sk socket lock held */ 649 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 650 const struct mptcp_addr_info *remote); 651 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 652 struct socket **new_sock); 653 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 654 struct sockaddr_storage *addr, 655 unsigned short family); 656 657 static inline bool __tcp_can_send(const struct sock *ssk) 658 { 659 /* only send if our side has not closed yet */ 660 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 661 } 662 663 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 664 { 665 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 666 if (subflow->request_join && !subflow->fully_established) 667 return false; 668 669 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 670 } 671 672 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 673 674 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 675 676 void mptcp_subflow_drop_ctx(struct sock *ssk); 677 678 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 679 struct mptcp_subflow_context *ctx) 680 { 681 sk->sk_data_ready = sock_def_readable; 682 sk->sk_state_change = ctx->tcp_state_change; 683 sk->sk_write_space = sk_stream_write_space; 684 sk->sk_error_report = ctx->tcp_error_report; 685 686 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 687 } 688 689 void __init mptcp_proto_init(void); 690 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 691 int __init mptcp_proto_v6_init(void); 692 #endif 693 694 struct sock *mptcp_sk_clone_init(const struct sock *sk, 695 const struct mptcp_options_received *mp_opt, 696 struct sock *ssk, 697 struct request_sock *req); 698 void mptcp_get_options(const struct sk_buff *skb, 699 struct mptcp_options_received *mp_opt); 700 701 void mptcp_finish_connect(struct sock *sk); 702 void __mptcp_set_connected(struct sock *sk); 703 void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout); 704 static inline bool mptcp_is_fully_established(struct sock *sk) 705 { 706 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 707 READ_ONCE(mptcp_sk(sk)->fully_established); 708 } 709 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 710 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 711 bool mptcp_finish_join(struct sock *sk); 712 bool mptcp_schedule_work(struct sock *sk); 713 int mptcp_setsockopt(struct sock *sk, int level, int optname, 714 sockptr_t optval, unsigned int optlen); 715 int mptcp_getsockopt(struct sock *sk, int level, int optname, 716 char __user *optval, int __user *option); 717 718 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 719 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 720 { 721 if (use_64bit) 722 return cur_seq; 723 724 return __mptcp_expand_seq(old_seq, cur_seq); 725 } 726 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 727 void __mptcp_data_acked(struct sock *sk); 728 void __mptcp_error_report(struct sock *sk); 729 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 730 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 731 { 732 return READ_ONCE(msk->snd_data_fin_enable) && 733 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 734 } 735 736 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 737 { 738 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 739 return false; 740 741 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 742 return true; 743 } 744 745 static inline void mptcp_write_space(struct sock *sk) 746 { 747 if (sk_stream_is_writeable(sk)) { 748 /* pairs with memory barrier in mptcp_poll */ 749 smp_mb(); 750 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 751 sk_stream_write_space(sk); 752 } 753 } 754 755 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 756 757 #define MPTCP_TOKEN_MAX_RETRIES 4 758 759 void __init mptcp_token_init(void); 760 static inline void mptcp_token_init_request(struct request_sock *req) 761 { 762 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 763 } 764 765 int mptcp_token_new_request(struct request_sock *req); 766 void mptcp_token_destroy_request(struct request_sock *req); 767 int mptcp_token_new_connect(struct sock *ssk); 768 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 769 struct mptcp_sock *msk); 770 bool mptcp_token_exists(u32 token); 771 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 772 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 773 long *s_num); 774 void mptcp_token_destroy(struct mptcp_sock *msk); 775 776 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 777 778 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 779 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 780 781 void __init mptcp_pm_init(void); 782 void mptcp_pm_data_init(struct mptcp_sock *msk); 783 void mptcp_pm_data_reset(struct mptcp_sock *msk); 784 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 785 struct mptcp_addr_info *addr); 786 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 787 bool require_family, 788 struct mptcp_pm_addr_entry *entry); 789 bool mptcp_pm_addr_families_match(const struct sock *sk, 790 const struct mptcp_addr_info *loc, 791 const struct mptcp_addr_info *rem); 792 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 793 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 794 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 795 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 796 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 797 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 798 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 799 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 800 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk, 801 const struct mptcp_subflow_context *subflow); 802 void mptcp_pm_add_addr_received(const struct sock *ssk, 803 const struct mptcp_addr_info *addr); 804 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 805 const struct mptcp_addr_info *addr); 806 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 807 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 808 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 809 const struct mptcp_rm_list *rm_list); 810 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 811 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 812 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 813 struct mptcp_addr_info *addr, 814 struct mptcp_addr_info *rem, 815 u8 bkup); 816 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 817 const struct mptcp_addr_info *addr); 818 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 819 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 820 struct mptcp_pm_add_entry * 821 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 822 const struct mptcp_addr_info *addr, bool check_id); 823 struct mptcp_pm_add_entry * 824 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 825 const struct mptcp_addr_info *addr); 826 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 827 unsigned int id, 828 u8 *flags, int *ifindex); 829 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id, 830 u8 *flags, int *ifindex); 831 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 832 unsigned int id, 833 u8 *flags, int *ifindex); 834 int mptcp_pm_set_flags(struct net *net, struct nlattr *token, 835 struct mptcp_pm_addr_entry *loc, 836 struct mptcp_pm_addr_entry *rem, u8 bkup); 837 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup); 838 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token, 839 struct mptcp_pm_addr_entry *loc, 840 struct mptcp_pm_addr_entry *rem, u8 bkup); 841 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 842 const struct mptcp_addr_info *addr, 843 bool echo); 844 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 845 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 846 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list); 847 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk, 848 struct list_head *rm_list); 849 850 void mptcp_free_local_addr_list(struct mptcp_sock *msk); 851 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info); 852 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info); 853 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info); 854 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info); 855 856 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 857 const struct sock *ssk, gfp_t gfp); 858 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 859 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 860 void mptcp_event_pm_listener(const struct sock *ssk, 861 enum mptcp_event_type event); 862 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 863 864 void mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow, 865 const struct mptcp_options_received *mp_opt); 866 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 867 struct request_sock *req); 868 869 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 870 { 871 return READ_ONCE(msk->pm.addr_signal) & 872 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 873 } 874 875 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 876 { 877 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 878 } 879 880 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 881 { 882 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 883 } 884 885 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 886 { 887 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 888 } 889 890 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 891 { 892 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 893 } 894 895 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 896 { 897 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 898 } 899 900 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 901 { 902 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 903 904 if (family == AF_INET6) 905 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 906 if (!echo) 907 len += MPTCPOPT_THMAC_LEN; 908 /* account for 2 trailing 'nop' options */ 909 if (port) 910 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 911 912 return len; 913 } 914 915 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 916 { 917 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 918 return -EINVAL; 919 920 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 921 } 922 923 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 924 unsigned int opt_size, unsigned int remaining, 925 struct mptcp_addr_info *addr, bool *echo, 926 bool *drop_other_suboptions); 927 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 928 struct mptcp_rm_list *rm_list); 929 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 930 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 931 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 932 933 void __init mptcp_pm_nl_init(void); 934 void mptcp_pm_nl_work(struct mptcp_sock *msk); 935 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 936 const struct mptcp_rm_list *rm_list); 937 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 938 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 939 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 940 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 941 942 /* called under PM lock */ 943 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 944 { 945 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 946 WRITE_ONCE(msk->pm.accept_subflow, true); 947 } 948 949 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 950 { 951 spin_lock_bh(&msk->pm.lock); 952 __mptcp_pm_close_subflow(msk); 953 spin_unlock_bh(&msk->pm.lock); 954 } 955 956 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 957 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 958 959 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 960 { 961 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 962 } 963 964 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 965 966 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 967 { 968 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 969 } 970 971 static inline bool mptcp_check_fallback(const struct sock *sk) 972 { 973 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 974 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 975 976 return __mptcp_check_fallback(msk); 977 } 978 979 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 980 { 981 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 982 pr_debug("TCP fallback already done (msk=%p)", msk); 983 return; 984 } 985 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 986 } 987 988 static inline void mptcp_do_fallback(struct sock *ssk) 989 { 990 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 991 struct sock *sk = subflow->conn; 992 struct mptcp_sock *msk; 993 994 msk = mptcp_sk(sk); 995 __mptcp_do_fallback(msk); 996 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 997 gfp_t saved_allocation = ssk->sk_allocation; 998 999 /* we are in a atomic (BH) scope, override ssk default for data 1000 * fin allocation 1001 */ 1002 ssk->sk_allocation = GFP_ATOMIC; 1003 ssk->sk_shutdown |= SEND_SHUTDOWN; 1004 tcp_shutdown(ssk, SEND_SHUTDOWN); 1005 ssk->sk_allocation = saved_allocation; 1006 } 1007 } 1008 1009 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 1010 1011 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1012 { 1013 struct mptcp_ext *mpext; 1014 1015 mpext = skb ? mptcp_get_ext(skb) : NULL; 1016 if (mpext && mpext->infinite_map) 1017 return true; 1018 1019 return false; 1020 } 1021 1022 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1023 { 1024 return (subflow->request_mptcp || subflow->request_join); 1025 } 1026 1027 static inline bool subflow_simultaneous_connect(struct sock *sk) 1028 { 1029 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1030 1031 return sk->sk_state == TCP_ESTABLISHED && 1032 is_active_ssk(subflow) && 1033 !subflow->conn_finished; 1034 } 1035 1036 #ifdef CONFIG_SYN_COOKIES 1037 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1038 struct sk_buff *skb); 1039 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1040 struct sk_buff *skb); 1041 void __init mptcp_join_cookie_init(void); 1042 #else 1043 static inline void 1044 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1045 struct sk_buff *skb) {} 1046 static inline bool 1047 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1048 struct sk_buff *skb) 1049 { 1050 return false; 1051 } 1052 1053 static inline void mptcp_join_cookie_init(void) {} 1054 #endif 1055 1056 #endif /* __MPTCP_PROTOCOL_H */ 1057