1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* Multipath TCP 3 * 4 * Copyright (c) 2017 - 2019, Intel Corporation. 5 */ 6 7 #ifndef __MPTCP_PROTOCOL_H 8 #define __MPTCP_PROTOCOL_H 9 10 #include <linux/random.h> 11 #include <net/tcp.h> 12 #include <net/inet_connection_sock.h> 13 #include <uapi/linux/mptcp.h> 14 #include <net/genetlink.h> 15 16 #define MPTCP_SUPPORTED_VERSION 1 17 18 /* MPTCP option bits */ 19 #define OPTION_MPTCP_MPC_SYN BIT(0) 20 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 21 #define OPTION_MPTCP_MPC_ACK BIT(2) 22 #define OPTION_MPTCP_MPJ_SYN BIT(3) 23 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 24 #define OPTION_MPTCP_MPJ_ACK BIT(5) 25 #define OPTION_MPTCP_ADD_ADDR BIT(6) 26 #define OPTION_MPTCP_RM_ADDR BIT(7) 27 #define OPTION_MPTCP_FASTCLOSE BIT(8) 28 #define OPTION_MPTCP_PRIO BIT(9) 29 #define OPTION_MPTCP_RST BIT(10) 30 #define OPTION_MPTCP_DSS BIT(11) 31 #define OPTION_MPTCP_FAIL BIT(12) 32 33 #define OPTION_MPTCP_CSUMREQD BIT(13) 34 35 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \ 36 OPTION_MPTCP_MPC_ACK) 37 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \ 38 OPTION_MPTCP_MPJ_ACK) 39 40 /* MPTCP option subtypes */ 41 #define MPTCPOPT_MP_CAPABLE 0 42 #define MPTCPOPT_MP_JOIN 1 43 #define MPTCPOPT_DSS 2 44 #define MPTCPOPT_ADD_ADDR 3 45 #define MPTCPOPT_RM_ADDR 4 46 #define MPTCPOPT_MP_PRIO 5 47 #define MPTCPOPT_MP_FAIL 6 48 #define MPTCPOPT_MP_FASTCLOSE 7 49 #define MPTCPOPT_RST 8 50 51 /* MPTCP suboption lengths */ 52 #define TCPOLEN_MPTCP_MPC_SYN 4 53 #define TCPOLEN_MPTCP_MPC_SYNACK 12 54 #define TCPOLEN_MPTCP_MPC_ACK 20 55 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 56 #define TCPOLEN_MPTCP_MPJ_SYN 12 57 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 58 #define TCPOLEN_MPTCP_MPJ_ACK 24 59 #define TCPOLEN_MPTCP_DSS_BASE 4 60 #define TCPOLEN_MPTCP_DSS_ACK32 4 61 #define TCPOLEN_MPTCP_DSS_ACK64 8 62 #define TCPOLEN_MPTCP_DSS_MAP32 10 63 #define TCPOLEN_MPTCP_DSS_MAP64 14 64 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 65 #define TCPOLEN_MPTCP_ADD_ADDR 16 66 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 69 #define TCPOLEN_MPTCP_ADD_ADDR6 28 70 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 73 #define TCPOLEN_MPTCP_PORT_LEN 2 74 #define TCPOLEN_MPTCP_PORT_ALIGN 2 75 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 76 #define TCPOLEN_MPTCP_PRIO 3 77 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 78 #define TCPOLEN_MPTCP_FASTCLOSE 12 79 #define TCPOLEN_MPTCP_RST 4 80 #define TCPOLEN_MPTCP_FAIL 12 81 82 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 83 84 /* MPTCP MP_JOIN flags */ 85 #define MPTCPOPT_BACKUP BIT(0) 86 #define MPTCPOPT_THMAC_LEN 8 87 88 /* MPTCP MP_CAPABLE flags */ 89 #define MPTCP_VERSION_MASK (0x0F) 90 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 91 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 92 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 93 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 94 #define MPTCP_CAP_FLAG_MASK (0x1F) 95 96 /* MPTCP DSS flags */ 97 #define MPTCP_DSS_DATA_FIN BIT(4) 98 #define MPTCP_DSS_DSN64 BIT(3) 99 #define MPTCP_DSS_HAS_MAP BIT(2) 100 #define MPTCP_DSS_ACK64 BIT(1) 101 #define MPTCP_DSS_HAS_ACK BIT(0) 102 #define MPTCP_DSS_FLAG_MASK (0x1F) 103 104 /* MPTCP ADD_ADDR flags */ 105 #define MPTCP_ADDR_ECHO BIT(0) 106 107 /* MPTCP MP_PRIO flags */ 108 #define MPTCP_PRIO_BKUP BIT(0) 109 110 /* MPTCP TCPRST flags */ 111 #define MPTCP_RST_TRANSIENT BIT(0) 112 113 /* MPTCP socket atomic flags */ 114 #define MPTCP_NOSPACE 1 115 #define MPTCP_WORK_RTX 2 116 #define MPTCP_WORK_EOF 3 117 #define MPTCP_FALLBACK_DONE 4 118 #define MPTCP_WORK_CLOSE_SUBFLOW 5 119 120 /* MPTCP socket release cb flags */ 121 #define MPTCP_PUSH_PENDING 1 122 #define MPTCP_CLEAN_UNA 2 123 #define MPTCP_ERROR_REPORT 3 124 #define MPTCP_RETRANSMIT 4 125 #define MPTCP_FLUSH_JOIN_LIST 5 126 #define MPTCP_CONNECTED 6 127 #define MPTCP_RESET_SCHEDULER 7 128 129 struct mptcp_skb_cb { 130 u64 map_seq; 131 u64 end_seq; 132 u32 offset; 133 u8 has_rxtstamp:1; 134 }; 135 136 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0])) 137 138 static inline bool before64(__u64 seq1, __u64 seq2) 139 { 140 return (__s64)(seq1 - seq2) < 0; 141 } 142 143 #define after64(seq2, seq1) before64(seq1, seq2) 144 145 struct mptcp_options_received { 146 u64 sndr_key; 147 u64 rcvr_key; 148 u64 data_ack; 149 u64 data_seq; 150 u32 subflow_seq; 151 u16 data_len; 152 __sum16 csum; 153 u16 suboptions; 154 u32 token; 155 u32 nonce; 156 u16 use_map:1, 157 dsn64:1, 158 data_fin:1, 159 use_ack:1, 160 ack64:1, 161 mpc_map:1, 162 reset_reason:4, 163 reset_transient:1, 164 echo:1, 165 backup:1, 166 deny_join_id0:1, 167 __unused:2; 168 u8 join_id; 169 u64 thmac; 170 u8 hmac[MPTCPOPT_HMAC_LEN]; 171 struct mptcp_addr_info addr; 172 struct mptcp_rm_list rm_list; 173 u64 ahmac; 174 u64 fail_seq; 175 }; 176 177 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 178 { 179 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 180 ((nib & 0xF) << 8) | field); 181 } 182 183 enum mptcp_pm_status { 184 MPTCP_PM_ADD_ADDR_RECEIVED, 185 MPTCP_PM_ADD_ADDR_SEND_ACK, 186 MPTCP_PM_RM_ADDR_RECEIVED, 187 MPTCP_PM_ESTABLISHED, 188 MPTCP_PM_SUBFLOW_ESTABLISHED, 189 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 190 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 191 * accounted int id_avail_bitmap 192 */ 193 }; 194 195 enum mptcp_pm_type { 196 MPTCP_PM_TYPE_KERNEL = 0, 197 MPTCP_PM_TYPE_USERSPACE, 198 199 __MPTCP_PM_TYPE_NR, 200 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 201 }; 202 203 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 204 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 205 206 enum mptcp_addr_signal_status { 207 MPTCP_ADD_ADDR_SIGNAL, 208 MPTCP_ADD_ADDR_ECHO, 209 MPTCP_RM_ADDR_SIGNAL, 210 }; 211 212 /* max value of mptcp_addr_info.id */ 213 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 214 215 struct mptcp_pm_data { 216 struct mptcp_addr_info local; 217 struct mptcp_addr_info remote; 218 struct list_head anno_list; 219 struct list_head userspace_pm_local_addr_list; 220 221 spinlock_t lock; /*protects the whole PM data */ 222 223 u8 addr_signal; 224 bool server_side; 225 bool work_pending; 226 bool accept_addr; 227 bool accept_subflow; 228 bool remote_deny_join_id0; 229 u8 add_addr_signaled; 230 u8 add_addr_accepted; 231 u8 local_addr_used; 232 u8 pm_type; 233 u8 subflows; 234 u8 status; 235 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 236 struct mptcp_rm_list rm_list_tx; 237 struct mptcp_rm_list rm_list_rx; 238 }; 239 240 struct mptcp_pm_addr_entry { 241 struct list_head list; 242 struct mptcp_addr_info addr; 243 u8 flags; 244 int ifindex; 245 struct socket *lsk; 246 }; 247 248 struct mptcp_data_frag { 249 struct list_head list; 250 u64 data_seq; 251 u16 data_len; 252 u16 offset; 253 u16 overhead; 254 u16 already_sent; 255 struct page *page; 256 }; 257 258 /* MPTCP connection sock */ 259 struct mptcp_sock { 260 /* inet_connection_sock must be the first member */ 261 struct inet_connection_sock sk; 262 u64 local_key; 263 u64 remote_key; 264 u64 write_seq; 265 u64 snd_nxt; 266 u64 ack_seq; 267 atomic64_t rcv_wnd_sent; 268 u64 rcv_data_fin_seq; 269 int rmem_fwd_alloc; 270 struct sock *last_snd; 271 int snd_burst; 272 int old_wspace; 273 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 274 * recovery related fields are under data_lock 275 * protection 276 */ 277 u64 snd_una; 278 u64 wnd_end; 279 unsigned long timer_ival; 280 u32 token; 281 int rmem_released; 282 unsigned long flags; 283 unsigned long cb_flags; 284 unsigned long push_pending; 285 bool recovery; /* closing subflow write queue reinjected */ 286 bool can_ack; 287 bool fully_established; 288 bool rcv_data_fin; 289 bool snd_data_fin_enable; 290 bool rcv_fastclose; 291 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 292 bool csum_enabled; 293 bool allow_infinite_fallback; 294 u8 mpc_endpoint_id; 295 u8 recvmsg_inq:1, 296 cork:1, 297 nodelay:1, 298 fastopening:1, 299 in_accept_queue:1; 300 struct work_struct work; 301 struct sk_buff *ooo_last_skb; 302 struct rb_root out_of_order_queue; 303 struct sk_buff_head receive_queue; 304 struct list_head conn_list; 305 struct list_head rtx_queue; 306 struct mptcp_data_frag *first_pending; 307 struct list_head join_list; 308 struct socket *subflow; /* outgoing connect/listener/!mp_capable 309 * The mptcp ops can safely dereference, using suitable 310 * ONCE annotation, the subflow outside the socket 311 * lock as such sock is freed after close(). 312 */ 313 struct sock *first; 314 struct mptcp_pm_data pm; 315 struct { 316 u32 space; /* bytes copied in last measurement window */ 317 u32 copied; /* bytes copied in this measurement window */ 318 u64 time; /* start time of measurement window */ 319 u64 rtt_us; /* last maximum rtt of subflows */ 320 } rcvq_space; 321 322 u32 setsockopt_seq; 323 char ca_name[TCP_CA_NAME_MAX]; 324 struct mptcp_sock *dl_next; 325 }; 326 327 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 328 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 329 330 #define mptcp_for_each_subflow(__msk, __subflow) \ 331 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 332 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 333 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 334 335 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 336 { 337 sock_owned_by_me((const struct sock *)msk); 338 } 339 340 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 341 342 /* the msk socket don't use the backlog, also account for the bulk 343 * free memory 344 */ 345 static inline int __mptcp_rmem(const struct sock *sk) 346 { 347 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 348 } 349 350 static inline int __mptcp_space(const struct sock *sk) 351 { 352 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 353 } 354 355 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 356 { 357 const struct mptcp_sock *msk = mptcp_sk(sk); 358 359 return READ_ONCE(msk->first_pending); 360 } 361 362 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 363 { 364 struct mptcp_sock *msk = mptcp_sk(sk); 365 struct mptcp_data_frag *cur; 366 367 cur = msk->first_pending; 368 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 369 list_next_entry(cur, list); 370 } 371 372 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 373 { 374 const struct mptcp_sock *msk = mptcp_sk(sk); 375 376 if (!msk->first_pending) 377 return NULL; 378 379 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 380 return NULL; 381 382 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 383 } 384 385 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 386 { 387 struct mptcp_sock *msk = mptcp_sk(sk); 388 389 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 390 return NULL; 391 392 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 393 } 394 395 struct csum_pseudo_header { 396 __be64 data_seq; 397 __be32 subflow_seq; 398 __be16 data_len; 399 __sum16 csum; 400 }; 401 402 struct mptcp_subflow_request_sock { 403 struct tcp_request_sock sk; 404 u16 mp_capable : 1, 405 mp_join : 1, 406 backup : 1, 407 csum_reqd : 1, 408 allow_join_id0 : 1; 409 u8 local_id; 410 u8 remote_id; 411 u64 local_key; 412 u64 idsn; 413 u32 token; 414 u32 ssn_offset; 415 u64 thmac; 416 u32 local_nonce; 417 u32 remote_nonce; 418 struct mptcp_sock *msk; 419 struct hlist_nulls_node token_node; 420 }; 421 422 static inline struct mptcp_subflow_request_sock * 423 mptcp_subflow_rsk(const struct request_sock *rsk) 424 { 425 return (struct mptcp_subflow_request_sock *)rsk; 426 } 427 428 enum mptcp_data_avail { 429 MPTCP_SUBFLOW_NODATA, 430 MPTCP_SUBFLOW_DATA_AVAIL, 431 }; 432 433 struct mptcp_delegated_action { 434 struct napi_struct napi; 435 struct list_head head; 436 }; 437 438 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 439 440 #define MPTCP_DELEGATE_SEND 0 441 #define MPTCP_DELEGATE_ACK 1 442 443 /* MPTCP subflow context */ 444 struct mptcp_subflow_context { 445 struct list_head node;/* conn_list of subflows */ 446 447 struct_group(reset, 448 449 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 450 u64 local_key; 451 u64 remote_key; 452 u64 idsn; 453 u64 map_seq; 454 u32 snd_isn; 455 u32 token; 456 u32 rel_write_seq; 457 u32 map_subflow_seq; 458 u32 ssn_offset; 459 u32 map_data_len; 460 __wsum map_data_csum; 461 u32 map_csum_len; 462 u32 request_mptcp : 1, /* send MP_CAPABLE */ 463 request_join : 1, /* send MP_JOIN */ 464 request_bkup : 1, 465 mp_capable : 1, /* remote is MPTCP capable */ 466 mp_join : 1, /* remote is JOINing */ 467 fully_established : 1, /* path validated */ 468 pm_notified : 1, /* PM hook called for established status */ 469 conn_finished : 1, 470 map_valid : 1, 471 map_csum_reqd : 1, 472 map_data_fin : 1, 473 mpc_map : 1, 474 backup : 1, 475 send_mp_prio : 1, 476 send_mp_fail : 1, 477 send_fastclose : 1, 478 send_infinite_map : 1, 479 rx_eof : 1, 480 remote_key_valid : 1, /* received the peer key from */ 481 disposable : 1, /* ctx can be free at ulp release time */ 482 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 483 local_id_valid : 1, /* local_id is correctly initialized */ 484 valid_csum_seen : 1, /* at least one csum validated */ 485 is_mptfo : 1, /* subflow is doing TFO */ 486 __unused : 8; 487 enum mptcp_data_avail data_avail; 488 u32 remote_nonce; 489 u64 thmac; 490 u32 local_nonce; 491 u32 remote_token; 492 union { 493 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 494 u64 iasn; /* initial ack sequence number, MPC subflows only */ 495 }; 496 u8 local_id; 497 u8 remote_id; 498 u8 reset_seen:1; 499 u8 reset_transient:1; 500 u8 reset_reason:4; 501 u8 stale_count; 502 503 long delegated_status; 504 unsigned long fail_tout; 505 506 ); 507 508 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 509 510 u32 setsockopt_seq; 511 u32 stale_rcv_tstamp; 512 513 struct sock *tcp_sock; /* tcp sk backpointer */ 514 struct sock *conn; /* parent mptcp_sock */ 515 const struct inet_connection_sock_af_ops *icsk_af_ops; 516 void (*tcp_state_change)(struct sock *sk); 517 void (*tcp_error_report)(struct sock *sk); 518 519 struct rcu_head rcu; 520 }; 521 522 static inline struct mptcp_subflow_context * 523 mptcp_subflow_ctx(const struct sock *sk) 524 { 525 struct inet_connection_sock *icsk = inet_csk(sk); 526 527 /* Use RCU on icsk_ulp_data only for sock diag code */ 528 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 529 } 530 531 static inline struct sock * 532 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 533 { 534 return subflow->tcp_sock; 535 } 536 537 static inline void 538 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 539 { 540 memset(&subflow->reset, 0, sizeof(subflow->reset)); 541 subflow->request_mptcp = 1; 542 } 543 544 static inline u64 545 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 546 { 547 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 548 subflow->ssn_offset - 549 subflow->map_subflow_seq; 550 } 551 552 static inline u64 553 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 554 { 555 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 556 } 557 558 void mptcp_subflow_process_delegated(struct sock *ssk); 559 560 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 561 { 562 struct mptcp_delegated_action *delegated; 563 bool schedule; 564 565 /* the caller held the subflow bh socket lock */ 566 lockdep_assert_in_softirq(); 567 568 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 569 * ensures the below list check sees list updates done prior to status 570 * bit changes 571 */ 572 if (!test_and_set_bit(action, &subflow->delegated_status)) { 573 /* still on delegated list from previous scheduling */ 574 if (!list_empty(&subflow->delegated_node)) 575 return; 576 577 delegated = this_cpu_ptr(&mptcp_delegated_actions); 578 schedule = list_empty(&delegated->head); 579 list_add_tail(&subflow->delegated_node, &delegated->head); 580 sock_hold(mptcp_subflow_tcp_sock(subflow)); 581 if (schedule) 582 napi_schedule(&delegated->napi); 583 } 584 } 585 586 static inline struct mptcp_subflow_context * 587 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 588 { 589 struct mptcp_subflow_context *ret; 590 591 if (list_empty(&delegated->head)) 592 return NULL; 593 594 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 595 list_del_init(&ret->delegated_node); 596 return ret; 597 } 598 599 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 600 { 601 return !!READ_ONCE(subflow->delegated_status); 602 } 603 604 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action) 605 { 606 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 607 * touching the status bit 608 */ 609 smp_wmb(); 610 clear_bit(action, &subflow->delegated_status); 611 } 612 613 int mptcp_is_enabled(const struct net *net); 614 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 615 int mptcp_is_checksum_enabled(const struct net *net); 616 int mptcp_allow_join_id0(const struct net *net); 617 unsigned int mptcp_stale_loss_cnt(const struct net *net); 618 int mptcp_get_pm_type(const struct net *net); 619 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 620 const struct mptcp_options_received *mp_opt); 621 bool __mptcp_retransmit_pending_data(struct sock *sk); 622 void mptcp_check_and_set_pending(struct sock *sk); 623 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 624 bool mptcp_subflow_data_available(struct sock *sk); 625 void __init mptcp_subflow_init(void); 626 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 627 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 628 struct mptcp_subflow_context *subflow); 629 void __mptcp_subflow_send_ack(struct sock *ssk); 630 void mptcp_subflow_reset(struct sock *ssk); 631 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 632 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 633 struct socket *__mptcp_nmpc_socket(struct mptcp_sock *msk); 634 bool __mptcp_close(struct sock *sk, long timeout); 635 void mptcp_cancel_work(struct sock *sk); 636 void __mptcp_unaccepted_force_close(struct sock *sk); 637 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk); 638 639 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 640 const struct mptcp_addr_info *b, bool use_port); 641 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr); 642 643 /* called with sk socket lock held */ 644 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 645 const struct mptcp_addr_info *remote); 646 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 647 struct socket **new_sock); 648 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 649 struct sockaddr_storage *addr, 650 unsigned short family); 651 652 static inline bool __tcp_can_send(const struct sock *ssk) 653 { 654 /* only send if our side has not closed yet */ 655 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 656 } 657 658 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 659 { 660 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 661 if (subflow->request_join && !subflow->fully_established) 662 return false; 663 664 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 665 } 666 667 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 668 669 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 670 671 void mptcp_subflow_drop_ctx(struct sock *ssk); 672 673 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 674 struct mptcp_subflow_context *ctx) 675 { 676 sk->sk_data_ready = sock_def_readable; 677 sk->sk_state_change = ctx->tcp_state_change; 678 sk->sk_write_space = sk_stream_write_space; 679 sk->sk_error_report = ctx->tcp_error_report; 680 681 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 682 } 683 684 void __init mptcp_proto_init(void); 685 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 686 int __init mptcp_proto_v6_init(void); 687 #endif 688 689 struct sock *mptcp_sk_clone_init(const struct sock *sk, 690 const struct mptcp_options_received *mp_opt, 691 struct sock *ssk, 692 struct request_sock *req); 693 void mptcp_get_options(const struct sk_buff *skb, 694 struct mptcp_options_received *mp_opt); 695 696 void mptcp_finish_connect(struct sock *sk); 697 void __mptcp_set_connected(struct sock *sk); 698 void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout); 699 static inline bool mptcp_is_fully_established(struct sock *sk) 700 { 701 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 702 READ_ONCE(mptcp_sk(sk)->fully_established); 703 } 704 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 705 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 706 bool mptcp_finish_join(struct sock *sk); 707 bool mptcp_schedule_work(struct sock *sk); 708 int mptcp_setsockopt(struct sock *sk, int level, int optname, 709 sockptr_t optval, unsigned int optlen); 710 int mptcp_getsockopt(struct sock *sk, int level, int optname, 711 char __user *optval, int __user *option); 712 713 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 714 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 715 { 716 if (use_64bit) 717 return cur_seq; 718 719 return __mptcp_expand_seq(old_seq, cur_seq); 720 } 721 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 722 void __mptcp_data_acked(struct sock *sk); 723 void __mptcp_error_report(struct sock *sk); 724 void mptcp_subflow_eof(struct sock *sk); 725 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 726 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 727 { 728 return READ_ONCE(msk->snd_data_fin_enable) && 729 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 730 } 731 732 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 733 { 734 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 735 return false; 736 737 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 738 return true; 739 } 740 741 static inline void mptcp_write_space(struct sock *sk) 742 { 743 if (sk_stream_is_writeable(sk)) { 744 /* pairs with memory barrier in mptcp_poll */ 745 smp_mb(); 746 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 747 sk_stream_write_space(sk); 748 } 749 } 750 751 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 752 753 #define MPTCP_TOKEN_MAX_RETRIES 4 754 755 void __init mptcp_token_init(void); 756 static inline void mptcp_token_init_request(struct request_sock *req) 757 { 758 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 759 } 760 761 int mptcp_token_new_request(struct request_sock *req); 762 void mptcp_token_destroy_request(struct request_sock *req); 763 int mptcp_token_new_connect(struct sock *ssk); 764 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 765 struct mptcp_sock *msk); 766 bool mptcp_token_exists(u32 token); 767 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 768 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 769 long *s_num); 770 void mptcp_token_destroy(struct mptcp_sock *msk); 771 772 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 773 774 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 775 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 776 777 void __init mptcp_pm_init(void); 778 void mptcp_pm_data_init(struct mptcp_sock *msk); 779 void mptcp_pm_data_reset(struct mptcp_sock *msk); 780 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 781 struct mptcp_addr_info *addr); 782 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 783 bool require_family, 784 struct mptcp_pm_addr_entry *entry); 785 bool mptcp_pm_addr_families_match(const struct sock *sk, 786 const struct mptcp_addr_info *loc, 787 const struct mptcp_addr_info *rem); 788 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 789 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 790 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 791 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 792 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 793 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 794 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 795 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 796 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk, 797 const struct mptcp_subflow_context *subflow); 798 void mptcp_pm_add_addr_received(const struct sock *ssk, 799 const struct mptcp_addr_info *addr); 800 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 801 const struct mptcp_addr_info *addr); 802 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 803 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 804 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 805 const struct mptcp_rm_list *rm_list); 806 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 807 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 808 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 809 struct mptcp_addr_info *addr, 810 struct mptcp_addr_info *rem, 811 u8 bkup); 812 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 813 const struct mptcp_pm_addr_entry *entry); 814 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 815 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 816 struct mptcp_pm_add_entry * 817 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 818 const struct mptcp_addr_info *addr, bool check_id); 819 struct mptcp_pm_add_entry * 820 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 821 const struct mptcp_addr_info *addr); 822 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 823 unsigned int id, 824 u8 *flags, int *ifindex); 825 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id, 826 u8 *flags, int *ifindex); 827 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 828 unsigned int id, 829 u8 *flags, int *ifindex); 830 int mptcp_pm_set_flags(struct net *net, struct nlattr *token, 831 struct mptcp_pm_addr_entry *loc, 832 struct mptcp_pm_addr_entry *rem, u8 bkup); 833 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup); 834 int mptcp_userspace_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_announce_addr(struct mptcp_sock *msk, 838 const struct mptcp_addr_info *addr, 839 bool echo); 840 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 841 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 842 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list); 843 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk, 844 struct list_head *rm_list); 845 846 void mptcp_free_local_addr_list(struct mptcp_sock *msk); 847 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info); 848 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info); 849 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info); 850 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info); 851 852 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 853 const struct sock *ssk, gfp_t gfp); 854 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 855 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 856 void mptcp_event_pm_listener(const struct sock *ssk, 857 enum mptcp_event_type event); 858 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 859 860 void mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow, 861 const struct mptcp_options_received *mp_opt); 862 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 863 struct request_sock *req); 864 865 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 866 { 867 return READ_ONCE(msk->pm.addr_signal) & 868 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 869 } 870 871 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 872 { 873 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 874 } 875 876 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 877 { 878 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 879 } 880 881 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 882 { 883 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 884 } 885 886 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 887 { 888 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 889 } 890 891 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 892 { 893 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 894 } 895 896 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 897 { 898 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 899 900 if (family == AF_INET6) 901 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 902 if (!echo) 903 len += MPTCPOPT_THMAC_LEN; 904 /* account for 2 trailing 'nop' options */ 905 if (port) 906 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 907 908 return len; 909 } 910 911 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 912 { 913 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 914 return -EINVAL; 915 916 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 917 } 918 919 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 920 unsigned int opt_size, unsigned int remaining, 921 struct mptcp_addr_info *addr, bool *echo, 922 bool *drop_other_suboptions); 923 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 924 struct mptcp_rm_list *rm_list); 925 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 926 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 927 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 928 929 void __init mptcp_pm_nl_init(void); 930 void mptcp_pm_nl_work(struct mptcp_sock *msk); 931 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 932 const struct mptcp_rm_list *rm_list); 933 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 934 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 935 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 936 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 937 938 /* called under PM lock */ 939 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 940 { 941 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 942 WRITE_ONCE(msk->pm.accept_subflow, true); 943 } 944 945 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 946 { 947 spin_lock_bh(&msk->pm.lock); 948 __mptcp_pm_close_subflow(msk); 949 spin_unlock_bh(&msk->pm.lock); 950 } 951 952 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 953 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 954 955 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 956 { 957 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 958 } 959 960 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 961 962 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 963 { 964 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 965 } 966 967 static inline bool mptcp_check_fallback(const struct sock *sk) 968 { 969 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 970 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 971 972 return __mptcp_check_fallback(msk); 973 } 974 975 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 976 { 977 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 978 pr_debug("TCP fallback already done (msk=%p)", msk); 979 return; 980 } 981 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 982 } 983 984 static inline void mptcp_do_fallback(struct sock *ssk) 985 { 986 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 987 struct sock *sk = subflow->conn; 988 struct mptcp_sock *msk; 989 990 msk = mptcp_sk(sk); 991 __mptcp_do_fallback(msk); 992 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 993 gfp_t saved_allocation = ssk->sk_allocation; 994 995 /* we are in a atomic (BH) scope, override ssk default for data 996 * fin allocation 997 */ 998 ssk->sk_allocation = GFP_ATOMIC; 999 ssk->sk_shutdown |= SEND_SHUTDOWN; 1000 tcp_shutdown(ssk, SEND_SHUTDOWN); 1001 ssk->sk_allocation = saved_allocation; 1002 } 1003 } 1004 1005 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 1006 1007 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1008 { 1009 struct mptcp_ext *mpext; 1010 1011 mpext = skb ? mptcp_get_ext(skb) : NULL; 1012 if (mpext && mpext->infinite_map) 1013 return true; 1014 1015 return false; 1016 } 1017 1018 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1019 { 1020 return (subflow->request_mptcp || subflow->request_join); 1021 } 1022 1023 static inline bool subflow_simultaneous_connect(struct sock *sk) 1024 { 1025 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1026 1027 return sk->sk_state == TCP_ESTABLISHED && 1028 is_active_ssk(subflow) && 1029 !subflow->conn_finished; 1030 } 1031 1032 #ifdef CONFIG_SYN_COOKIES 1033 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1034 struct sk_buff *skb); 1035 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1036 struct sk_buff *skb); 1037 void __init mptcp_join_cookie_init(void); 1038 #else 1039 static inline void 1040 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1041 struct sk_buff *skb) {} 1042 static inline bool 1043 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1044 struct sk_buff *skb) 1045 { 1046 return false; 1047 } 1048 1049 static inline void mptcp_join_cookie_init(void) {} 1050 #endif 1051 1052 #endif /* __MPTCP_PROTOCOL_H */ 1053