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