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 u8 scaling_ratio; 325 326 u32 subflow_id; 327 u32 setsockopt_seq; 328 char ca_name[TCP_CA_NAME_MAX]; 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_win_from_space(const struct sock *sk, int space) 355 { 356 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 357 } 358 359 static inline int __mptcp_space(const struct sock *sk) 360 { 361 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 362 } 363 364 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 365 { 366 const struct mptcp_sock *msk = mptcp_sk(sk); 367 368 return READ_ONCE(msk->first_pending); 369 } 370 371 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 372 { 373 struct mptcp_sock *msk = mptcp_sk(sk); 374 struct mptcp_data_frag *cur; 375 376 cur = msk->first_pending; 377 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 378 list_next_entry(cur, list); 379 } 380 381 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 382 { 383 const struct mptcp_sock *msk = mptcp_sk(sk); 384 385 if (!msk->first_pending) 386 return NULL; 387 388 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 389 return NULL; 390 391 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 392 } 393 394 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 395 { 396 struct mptcp_sock *msk = mptcp_sk(sk); 397 398 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 399 return NULL; 400 401 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 402 } 403 404 struct csum_pseudo_header { 405 __be64 data_seq; 406 __be32 subflow_seq; 407 __be16 data_len; 408 __sum16 csum; 409 }; 410 411 struct mptcp_subflow_request_sock { 412 struct tcp_request_sock sk; 413 u16 mp_capable : 1, 414 mp_join : 1, 415 backup : 1, 416 csum_reqd : 1, 417 allow_join_id0 : 1; 418 u8 local_id; 419 u8 remote_id; 420 u64 local_key; 421 u64 idsn; 422 u32 token; 423 u32 ssn_offset; 424 u64 thmac; 425 u32 local_nonce; 426 u32 remote_nonce; 427 struct mptcp_sock *msk; 428 struct hlist_nulls_node token_node; 429 }; 430 431 static inline struct mptcp_subflow_request_sock * 432 mptcp_subflow_rsk(const struct request_sock *rsk) 433 { 434 return (struct mptcp_subflow_request_sock *)rsk; 435 } 436 437 enum mptcp_data_avail { 438 MPTCP_SUBFLOW_NODATA, 439 MPTCP_SUBFLOW_DATA_AVAIL, 440 }; 441 442 struct mptcp_delegated_action { 443 struct napi_struct napi; 444 struct list_head head; 445 }; 446 447 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 448 449 #define MPTCP_DELEGATE_SEND 0 450 #define MPTCP_DELEGATE_ACK 1 451 452 /* MPTCP subflow context */ 453 struct mptcp_subflow_context { 454 struct list_head node;/* conn_list of subflows */ 455 456 struct_group(reset, 457 458 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 459 u64 local_key; 460 u64 remote_key; 461 u64 idsn; 462 u64 map_seq; 463 u32 snd_isn; 464 u32 token; 465 u32 rel_write_seq; 466 u32 map_subflow_seq; 467 u32 ssn_offset; 468 u32 map_data_len; 469 __wsum map_data_csum; 470 u32 map_csum_len; 471 u32 request_mptcp : 1, /* send MP_CAPABLE */ 472 request_join : 1, /* send MP_JOIN */ 473 request_bkup : 1, 474 mp_capable : 1, /* remote is MPTCP capable */ 475 mp_join : 1, /* remote is JOINing */ 476 fully_established : 1, /* path validated */ 477 pm_notified : 1, /* PM hook called for established status */ 478 conn_finished : 1, 479 map_valid : 1, 480 map_csum_reqd : 1, 481 map_data_fin : 1, 482 mpc_map : 1, 483 backup : 1, 484 send_mp_prio : 1, 485 send_mp_fail : 1, 486 send_fastclose : 1, 487 send_infinite_map : 1, 488 remote_key_valid : 1, /* received the peer key from */ 489 disposable : 1, /* ctx can be free at ulp release time */ 490 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 491 local_id_valid : 1, /* local_id is correctly initialized */ 492 valid_csum_seen : 1, /* at least one csum validated */ 493 is_mptfo : 1, /* subflow is doing TFO */ 494 __unused : 9; 495 enum mptcp_data_avail data_avail; 496 u32 remote_nonce; 497 u64 thmac; 498 u32 local_nonce; 499 u32 remote_token; 500 union { 501 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 502 u64 iasn; /* initial ack sequence number, MPC subflows only */ 503 }; 504 u8 local_id; 505 u8 remote_id; 506 u8 reset_seen:1; 507 u8 reset_transient:1; 508 u8 reset_reason:4; 509 u8 stale_count; 510 511 u32 subflow_id; 512 513 long delegated_status; 514 unsigned long fail_tout; 515 516 ); 517 518 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 519 520 u32 setsockopt_seq; 521 u32 stale_rcv_tstamp; 522 523 struct sock *tcp_sock; /* tcp sk backpointer */ 524 struct sock *conn; /* parent mptcp_sock */ 525 const struct inet_connection_sock_af_ops *icsk_af_ops; 526 void (*tcp_state_change)(struct sock *sk); 527 void (*tcp_error_report)(struct sock *sk); 528 529 struct rcu_head rcu; 530 }; 531 532 static inline struct mptcp_subflow_context * 533 mptcp_subflow_ctx(const struct sock *sk) 534 { 535 struct inet_connection_sock *icsk = inet_csk(sk); 536 537 /* Use RCU on icsk_ulp_data only for sock diag code */ 538 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 539 } 540 541 static inline struct sock * 542 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 543 { 544 return subflow->tcp_sock; 545 } 546 547 static inline void 548 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 549 { 550 memset(&subflow->reset, 0, sizeof(subflow->reset)); 551 subflow->request_mptcp = 1; 552 } 553 554 static inline u64 555 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 556 { 557 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 558 subflow->ssn_offset - 559 subflow->map_subflow_seq; 560 } 561 562 static inline u64 563 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 564 { 565 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 566 } 567 568 void mptcp_subflow_process_delegated(struct sock *ssk); 569 570 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 571 { 572 struct mptcp_delegated_action *delegated; 573 bool schedule; 574 575 /* the caller held the subflow bh socket lock */ 576 lockdep_assert_in_softirq(); 577 578 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 579 * ensures the below list check sees list updates done prior to status 580 * bit changes 581 */ 582 if (!test_and_set_bit(action, &subflow->delegated_status)) { 583 /* still on delegated list from previous scheduling */ 584 if (!list_empty(&subflow->delegated_node)) 585 return; 586 587 delegated = this_cpu_ptr(&mptcp_delegated_actions); 588 schedule = list_empty(&delegated->head); 589 list_add_tail(&subflow->delegated_node, &delegated->head); 590 sock_hold(mptcp_subflow_tcp_sock(subflow)); 591 if (schedule) 592 napi_schedule(&delegated->napi); 593 } 594 } 595 596 static inline struct mptcp_subflow_context * 597 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 598 { 599 struct mptcp_subflow_context *ret; 600 601 if (list_empty(&delegated->head)) 602 return NULL; 603 604 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 605 list_del_init(&ret->delegated_node); 606 return ret; 607 } 608 609 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 610 { 611 return !!READ_ONCE(subflow->delegated_status); 612 } 613 614 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action) 615 { 616 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 617 * touching the status bit 618 */ 619 smp_wmb(); 620 clear_bit(action, &subflow->delegated_status); 621 } 622 623 int mptcp_is_enabled(const struct net *net); 624 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 625 int mptcp_is_checksum_enabled(const struct net *net); 626 int mptcp_allow_join_id0(const struct net *net); 627 unsigned int mptcp_stale_loss_cnt(const struct net *net); 628 int mptcp_get_pm_type(const struct net *net); 629 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 630 const struct mptcp_options_received *mp_opt); 631 bool __mptcp_retransmit_pending_data(struct sock *sk); 632 void mptcp_check_and_set_pending(struct sock *sk); 633 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 634 bool mptcp_subflow_data_available(struct sock *sk); 635 void __init mptcp_subflow_init(void); 636 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 637 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 638 struct mptcp_subflow_context *subflow); 639 void __mptcp_subflow_send_ack(struct sock *ssk); 640 void mptcp_subflow_reset(struct sock *ssk); 641 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 642 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 643 struct socket *__mptcp_nmpc_socket(struct mptcp_sock *msk); 644 bool __mptcp_close(struct sock *sk, long timeout); 645 void mptcp_cancel_work(struct sock *sk); 646 void __mptcp_unaccepted_force_close(struct sock *sk); 647 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk); 648 649 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 650 const struct mptcp_addr_info *b, bool use_port); 651 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr); 652 653 /* called with sk socket lock held */ 654 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 655 const struct mptcp_addr_info *remote); 656 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 657 struct socket **new_sock); 658 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 659 struct sockaddr_storage *addr, 660 unsigned short family); 661 662 static inline bool __tcp_can_send(const struct sock *ssk) 663 { 664 /* only send if our side has not closed yet */ 665 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 666 } 667 668 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 669 { 670 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 671 if (subflow->request_join && !subflow->fully_established) 672 return false; 673 674 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 675 } 676 677 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 678 679 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 680 681 void mptcp_subflow_drop_ctx(struct sock *ssk); 682 683 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 684 struct mptcp_subflow_context *ctx) 685 { 686 sk->sk_data_ready = sock_def_readable; 687 sk->sk_state_change = ctx->tcp_state_change; 688 sk->sk_write_space = sk_stream_write_space; 689 sk->sk_error_report = ctx->tcp_error_report; 690 691 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 692 } 693 694 void __init mptcp_proto_init(void); 695 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 696 int __init mptcp_proto_v6_init(void); 697 #endif 698 699 struct sock *mptcp_sk_clone_init(const struct sock *sk, 700 const struct mptcp_options_received *mp_opt, 701 struct sock *ssk, 702 struct request_sock *req); 703 void mptcp_get_options(const struct sk_buff *skb, 704 struct mptcp_options_received *mp_opt); 705 706 void mptcp_finish_connect(struct sock *sk); 707 void __mptcp_set_connected(struct sock *sk); 708 void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout); 709 static inline bool mptcp_is_fully_established(struct sock *sk) 710 { 711 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 712 READ_ONCE(mptcp_sk(sk)->fully_established); 713 } 714 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 715 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 716 bool mptcp_finish_join(struct sock *sk); 717 bool mptcp_schedule_work(struct sock *sk); 718 int mptcp_setsockopt(struct sock *sk, int level, int optname, 719 sockptr_t optval, unsigned int optlen); 720 int mptcp_getsockopt(struct sock *sk, int level, int optname, 721 char __user *optval, int __user *option); 722 723 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 724 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 725 { 726 if (use_64bit) 727 return cur_seq; 728 729 return __mptcp_expand_seq(old_seq, cur_seq); 730 } 731 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 732 void __mptcp_data_acked(struct sock *sk); 733 void __mptcp_error_report(struct sock *sk); 734 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 735 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 736 { 737 return READ_ONCE(msk->snd_data_fin_enable) && 738 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 739 } 740 741 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 742 { 743 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 744 return false; 745 746 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 747 return true; 748 } 749 750 static inline void mptcp_write_space(struct sock *sk) 751 { 752 if (sk_stream_is_writeable(sk)) { 753 /* pairs with memory barrier in mptcp_poll */ 754 smp_mb(); 755 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 756 sk_stream_write_space(sk); 757 } 758 } 759 760 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 761 762 #define MPTCP_TOKEN_MAX_RETRIES 4 763 764 void __init mptcp_token_init(void); 765 static inline void mptcp_token_init_request(struct request_sock *req) 766 { 767 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 768 } 769 770 int mptcp_token_new_request(struct request_sock *req); 771 void mptcp_token_destroy_request(struct request_sock *req); 772 int mptcp_token_new_connect(struct sock *ssk); 773 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 774 struct mptcp_sock *msk); 775 bool mptcp_token_exists(u32 token); 776 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 777 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 778 long *s_num); 779 void mptcp_token_destroy(struct mptcp_sock *msk); 780 781 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 782 783 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 784 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 785 786 void __init mptcp_pm_init(void); 787 void mptcp_pm_data_init(struct mptcp_sock *msk); 788 void mptcp_pm_data_reset(struct mptcp_sock *msk); 789 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 790 struct mptcp_addr_info *addr); 791 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 792 bool require_family, 793 struct mptcp_pm_addr_entry *entry); 794 bool mptcp_pm_addr_families_match(const struct sock *sk, 795 const struct mptcp_addr_info *loc, 796 const struct mptcp_addr_info *rem); 797 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 798 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 799 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 800 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 801 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 802 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 803 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 804 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 805 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk, 806 const struct mptcp_subflow_context *subflow); 807 void mptcp_pm_add_addr_received(const struct sock *ssk, 808 const struct mptcp_addr_info *addr); 809 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 810 const struct mptcp_addr_info *addr); 811 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 812 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 813 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 814 const struct mptcp_rm_list *rm_list); 815 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 816 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 817 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 818 struct mptcp_addr_info *addr, 819 struct mptcp_addr_info *rem, 820 u8 bkup); 821 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 822 const struct mptcp_addr_info *addr); 823 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 824 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 825 struct mptcp_pm_add_entry * 826 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 827 const struct mptcp_addr_info *addr, bool check_id); 828 struct mptcp_pm_add_entry * 829 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 830 const struct mptcp_addr_info *addr); 831 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 832 unsigned int id, 833 u8 *flags, int *ifindex); 834 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id, 835 u8 *flags, int *ifindex); 836 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 837 unsigned int id, 838 u8 *flags, int *ifindex); 839 int mptcp_pm_set_flags(struct net *net, struct nlattr *token, 840 struct mptcp_pm_addr_entry *loc, 841 struct mptcp_pm_addr_entry *rem, u8 bkup); 842 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup); 843 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token, 844 struct mptcp_pm_addr_entry *loc, 845 struct mptcp_pm_addr_entry *rem, u8 bkup); 846 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 847 const struct mptcp_addr_info *addr, 848 bool echo); 849 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 850 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 851 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list); 852 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk, 853 struct list_head *rm_list); 854 855 void mptcp_free_local_addr_list(struct mptcp_sock *msk); 856 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info); 857 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info); 858 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info); 859 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info); 860 861 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 862 const struct sock *ssk, gfp_t gfp); 863 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 864 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 865 void mptcp_event_pm_listener(const struct sock *ssk, 866 enum mptcp_event_type event); 867 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 868 869 void mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow, 870 const struct mptcp_options_received *mp_opt); 871 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 872 struct request_sock *req); 873 874 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 875 { 876 return READ_ONCE(msk->pm.addr_signal) & 877 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 878 } 879 880 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 881 { 882 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 883 } 884 885 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 886 { 887 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 888 } 889 890 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 891 { 892 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 893 } 894 895 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 896 { 897 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 898 } 899 900 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 901 { 902 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 903 } 904 905 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 906 { 907 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 908 909 if (family == AF_INET6) 910 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 911 if (!echo) 912 len += MPTCPOPT_THMAC_LEN; 913 /* account for 2 trailing 'nop' options */ 914 if (port) 915 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 916 917 return len; 918 } 919 920 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 921 { 922 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 923 return -EINVAL; 924 925 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 926 } 927 928 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 929 unsigned int opt_size, unsigned int remaining, 930 struct mptcp_addr_info *addr, bool *echo, 931 bool *drop_other_suboptions); 932 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 933 struct mptcp_rm_list *rm_list); 934 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 935 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 936 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 937 938 void __init mptcp_pm_nl_init(void); 939 void mptcp_pm_nl_work(struct mptcp_sock *msk); 940 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 941 const struct mptcp_rm_list *rm_list); 942 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 943 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 944 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 945 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 946 947 /* called under PM lock */ 948 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 949 { 950 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 951 WRITE_ONCE(msk->pm.accept_subflow, true); 952 } 953 954 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 955 { 956 spin_lock_bh(&msk->pm.lock); 957 __mptcp_pm_close_subflow(msk); 958 spin_unlock_bh(&msk->pm.lock); 959 } 960 961 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 962 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 963 964 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 965 { 966 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 967 } 968 969 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 970 971 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 972 { 973 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 974 } 975 976 static inline bool mptcp_check_fallback(const struct sock *sk) 977 { 978 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 979 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 980 981 return __mptcp_check_fallback(msk); 982 } 983 984 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 985 { 986 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 987 pr_debug("TCP fallback already done (msk=%p)", msk); 988 return; 989 } 990 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 991 } 992 993 static inline void mptcp_do_fallback(struct sock *ssk) 994 { 995 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 996 struct sock *sk = subflow->conn; 997 struct mptcp_sock *msk; 998 999 msk = mptcp_sk(sk); 1000 __mptcp_do_fallback(msk); 1001 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1002 gfp_t saved_allocation = ssk->sk_allocation; 1003 1004 /* we are in a atomic (BH) scope, override ssk default for data 1005 * fin allocation 1006 */ 1007 ssk->sk_allocation = GFP_ATOMIC; 1008 ssk->sk_shutdown |= SEND_SHUTDOWN; 1009 tcp_shutdown(ssk, SEND_SHUTDOWN); 1010 ssk->sk_allocation = saved_allocation; 1011 } 1012 } 1013 1014 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 1015 1016 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1017 { 1018 struct mptcp_ext *mpext; 1019 1020 mpext = skb ? mptcp_get_ext(skb) : NULL; 1021 if (mpext && mpext->infinite_map) 1022 return true; 1023 1024 return false; 1025 } 1026 1027 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1028 { 1029 return (subflow->request_mptcp || subflow->request_join); 1030 } 1031 1032 static inline bool subflow_simultaneous_connect(struct sock *sk) 1033 { 1034 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1035 1036 return sk->sk_state == TCP_ESTABLISHED && 1037 is_active_ssk(subflow) && 1038 !subflow->conn_finished; 1039 } 1040 1041 #ifdef CONFIG_SYN_COOKIES 1042 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1043 struct sk_buff *skb); 1044 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1045 struct sk_buff *skb); 1046 void __init mptcp_join_cookie_init(void); 1047 #else 1048 static inline void 1049 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1050 struct sk_buff *skb) {} 1051 static inline bool 1052 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1053 struct sk_buff *skb) 1054 { 1055 return false; 1056 } 1057 1058 static inline void mptcp_join_cookie_init(void) {} 1059 #endif 1060 1061 #endif /* __MPTCP_PROTOCOL_H */ 1062