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 #include "mptcp_pm_gen.h" 17 18 #define MPTCP_SUPPORTED_VERSION 1 19 20 /* MPTCP option bits */ 21 #define OPTION_MPTCP_MPC_SYN BIT(0) 22 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 23 #define OPTION_MPTCP_MPC_ACK BIT(2) 24 #define OPTION_MPTCP_MPJ_SYN BIT(3) 25 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 26 #define OPTION_MPTCP_MPJ_ACK BIT(5) 27 #define OPTION_MPTCP_ADD_ADDR BIT(6) 28 #define OPTION_MPTCP_RM_ADDR BIT(7) 29 #define OPTION_MPTCP_FASTCLOSE BIT(8) 30 #define OPTION_MPTCP_PRIO BIT(9) 31 #define OPTION_MPTCP_RST BIT(10) 32 #define OPTION_MPTCP_DSS BIT(11) 33 #define OPTION_MPTCP_FAIL BIT(12) 34 35 #define OPTION_MPTCP_CSUMREQD BIT(13) 36 37 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \ 38 OPTION_MPTCP_MPC_ACK) 39 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \ 40 OPTION_MPTCP_MPJ_ACK) 41 42 /* MPTCP option subtypes */ 43 #define MPTCPOPT_MP_CAPABLE 0 44 #define MPTCPOPT_MP_JOIN 1 45 #define MPTCPOPT_DSS 2 46 #define MPTCPOPT_ADD_ADDR 3 47 #define MPTCPOPT_RM_ADDR 4 48 #define MPTCPOPT_MP_PRIO 5 49 #define MPTCPOPT_MP_FAIL 6 50 #define MPTCPOPT_MP_FASTCLOSE 7 51 #define MPTCPOPT_RST 8 52 53 /* MPTCP suboption lengths */ 54 #define TCPOLEN_MPTCP_MPC_SYN 4 55 #define TCPOLEN_MPTCP_MPC_SYNACK 12 56 #define TCPOLEN_MPTCP_MPC_ACK 20 57 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 58 #define TCPOLEN_MPTCP_MPJ_SYN 12 59 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 60 #define TCPOLEN_MPTCP_MPJ_ACK 24 61 #define TCPOLEN_MPTCP_DSS_BASE 4 62 #define TCPOLEN_MPTCP_DSS_ACK32 4 63 #define TCPOLEN_MPTCP_DSS_ACK64 8 64 #define TCPOLEN_MPTCP_DSS_MAP32 10 65 #define TCPOLEN_MPTCP_DSS_MAP64 14 66 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 67 #define TCPOLEN_MPTCP_ADD_ADDR 16 68 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 70 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 71 #define TCPOLEN_MPTCP_ADD_ADDR6 28 72 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 74 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 75 #define TCPOLEN_MPTCP_PORT_LEN 2 76 #define TCPOLEN_MPTCP_PORT_ALIGN 2 77 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 78 #define TCPOLEN_MPTCP_PRIO 3 79 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 80 #define TCPOLEN_MPTCP_FASTCLOSE 12 81 #define TCPOLEN_MPTCP_RST 4 82 #define TCPOLEN_MPTCP_FAIL 12 83 84 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 85 86 /* MPTCP MP_JOIN flags */ 87 #define MPTCPOPT_BACKUP BIT(0) 88 #define MPTCPOPT_THMAC_LEN 8 89 90 /* MPTCP MP_CAPABLE flags */ 91 #define MPTCP_VERSION_MASK (0x0F) 92 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 93 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 94 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 95 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 96 #define MPTCP_CAP_FLAG_MASK (0x1F) 97 98 /* MPTCP DSS flags */ 99 #define MPTCP_DSS_DATA_FIN BIT(4) 100 #define MPTCP_DSS_DSN64 BIT(3) 101 #define MPTCP_DSS_HAS_MAP BIT(2) 102 #define MPTCP_DSS_ACK64 BIT(1) 103 #define MPTCP_DSS_HAS_ACK BIT(0) 104 #define MPTCP_DSS_FLAG_MASK (0x1F) 105 106 /* MPTCP ADD_ADDR flags */ 107 #define MPTCP_ADDR_ECHO BIT(0) 108 109 /* MPTCP MP_PRIO flags */ 110 #define MPTCP_PRIO_BKUP BIT(0) 111 112 /* MPTCP TCPRST flags */ 113 #define MPTCP_RST_TRANSIENT BIT(0) 114 115 /* MPTCP socket atomic flags */ 116 #define MPTCP_NOSPACE 1 117 #define MPTCP_WORK_RTX 2 118 #define MPTCP_FALLBACK_DONE 4 119 #define MPTCP_WORK_CLOSE_SUBFLOW 5 120 121 /* MPTCP socket release cb flags */ 122 #define MPTCP_PUSH_PENDING 1 123 #define MPTCP_CLEAN_UNA 2 124 #define MPTCP_ERROR_REPORT 3 125 #define MPTCP_RETRANSMIT 4 126 #define MPTCP_FLUSH_JOIN_LIST 5 127 #define MPTCP_CONNECTED 6 128 129 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 bytes_sent; 266 u64 snd_nxt; 267 u64 bytes_received; 268 u64 ack_seq; 269 atomic64_t rcv_wnd_sent; 270 u64 rcv_data_fin_seq; 271 u64 bytes_retrans; 272 int rmem_fwd_alloc; 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 free_first:1; 304 struct work_struct work; 305 struct sk_buff *ooo_last_skb; 306 struct rb_root out_of_order_queue; 307 struct sk_buff_head receive_queue; 308 struct list_head conn_list; 309 struct list_head rtx_queue; 310 struct mptcp_data_frag *first_pending; 311 struct list_head join_list; 312 struct sock *first; /* 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 mptcp_pm_data pm; 317 struct mptcp_sched_ops *sched; 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_SCHEDULED 0 450 #define MPTCP_DELEGATE_SEND 1 451 #define MPTCP_DELEGATE_ACK 2 452 453 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 454 /* MPTCP subflow context */ 455 struct mptcp_subflow_context { 456 struct list_head node;/* conn_list of subflows */ 457 458 struct_group(reset, 459 460 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 461 u64 local_key; 462 u64 remote_key; 463 u64 idsn; 464 u64 map_seq; 465 u32 snd_isn; 466 u32 token; 467 u32 rel_write_seq; 468 u32 map_subflow_seq; 469 u32 ssn_offset; 470 u32 map_data_len; 471 __wsum map_data_csum; 472 u32 map_csum_len; 473 u32 request_mptcp : 1, /* send MP_CAPABLE */ 474 request_join : 1, /* send MP_JOIN */ 475 request_bkup : 1, 476 mp_capable : 1, /* remote is MPTCP capable */ 477 mp_join : 1, /* remote is JOINing */ 478 fully_established : 1, /* path validated */ 479 pm_notified : 1, /* PM hook called for established status */ 480 conn_finished : 1, 481 map_valid : 1, 482 map_csum_reqd : 1, 483 map_data_fin : 1, 484 mpc_map : 1, 485 backup : 1, 486 send_mp_prio : 1, 487 send_mp_fail : 1, 488 send_fastclose : 1, 489 send_infinite_map : 1, 490 remote_key_valid : 1, /* received the peer key from */ 491 disposable : 1, /* ctx can be free at ulp release time */ 492 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 493 local_id_valid : 1, /* local_id is correctly initialized */ 494 valid_csum_seen : 1, /* at least one csum validated */ 495 is_mptfo : 1, /* subflow is doing TFO */ 496 __unused : 9; 497 enum mptcp_data_avail data_avail; 498 bool scheduled; 499 u32 remote_nonce; 500 u64 thmac; 501 u32 local_nonce; 502 u32 remote_token; 503 union { 504 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 505 u64 iasn; /* initial ack sequence number, MPC subflows only */ 506 }; 507 u8 local_id; 508 u8 remote_id; 509 u8 reset_seen:1; 510 u8 reset_transient:1; 511 u8 reset_reason:4; 512 u8 stale_count; 513 514 u32 subflow_id; 515 516 long delegated_status; 517 unsigned long fail_tout; 518 519 ); 520 521 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 522 523 u32 setsockopt_seq; 524 u32 stale_rcv_tstamp; 525 526 struct sock *tcp_sock; /* tcp sk backpointer */ 527 struct sock *conn; /* parent mptcp_sock */ 528 const struct inet_connection_sock_af_ops *icsk_af_ops; 529 void (*tcp_state_change)(struct sock *sk); 530 void (*tcp_error_report)(struct sock *sk); 531 532 struct rcu_head rcu; 533 }; 534 535 static inline struct mptcp_subflow_context * 536 mptcp_subflow_ctx(const struct sock *sk) 537 { 538 struct inet_connection_sock *icsk = inet_csk(sk); 539 540 /* Use RCU on icsk_ulp_data only for sock diag code */ 541 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 542 } 543 544 static inline struct sock * 545 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 546 { 547 return subflow->tcp_sock; 548 } 549 550 static inline void 551 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 552 { 553 memset(&subflow->reset, 0, sizeof(subflow->reset)); 554 subflow->request_mptcp = 1; 555 } 556 557 static inline u64 558 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 559 { 560 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 561 subflow->ssn_offset - 562 subflow->map_subflow_seq; 563 } 564 565 static inline u64 566 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 567 { 568 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 569 } 570 571 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 572 573 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 574 { 575 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 576 struct mptcp_delegated_action *delegated; 577 bool schedule; 578 579 /* the caller held the subflow bh socket lock */ 580 lockdep_assert_in_softirq(); 581 582 /* The implied barrier pairs with tcp_release_cb_override() 583 * mptcp_napi_poll(), and ensures the below list check sees list 584 * updates done prior to delegated status bits changes 585 */ 586 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 587 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 588 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 589 return; 590 591 delegated = this_cpu_ptr(&mptcp_delegated_actions); 592 schedule = list_empty(&delegated->head); 593 list_add_tail(&subflow->delegated_node, &delegated->head); 594 sock_hold(mptcp_subflow_tcp_sock(subflow)); 595 if (schedule) 596 napi_schedule(&delegated->napi); 597 } 598 } 599 600 static inline struct mptcp_subflow_context * 601 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 602 { 603 struct mptcp_subflow_context *ret; 604 605 if (list_empty(&delegated->head)) 606 return NULL; 607 608 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 609 list_del_init(&ret->delegated_node); 610 return ret; 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 const char *mptcp_get_scheduler(const struct net *net); 620 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 621 const struct mptcp_options_received *mp_opt); 622 bool __mptcp_retransmit_pending_data(struct sock *sk); 623 void mptcp_check_and_set_pending(struct sock *sk); 624 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 625 bool mptcp_subflow_data_available(struct sock *sk); 626 void __init mptcp_subflow_init(void); 627 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 628 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 629 struct mptcp_subflow_context *subflow); 630 void __mptcp_subflow_send_ack(struct sock *ssk); 631 void mptcp_subflow_reset(struct sock *ssk); 632 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 633 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 634 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 635 bool __mptcp_close(struct sock *sk, long timeout); 636 void mptcp_cancel_work(struct sock *sk); 637 void __mptcp_unaccepted_force_close(struct sock *sk); 638 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk); 639 640 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 641 const struct mptcp_addr_info *b, bool use_port); 642 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr); 643 644 /* called with sk socket lock held */ 645 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 646 const struct mptcp_addr_info *remote); 647 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 648 struct socket **new_sock); 649 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 650 struct sockaddr_storage *addr, 651 unsigned short family); 652 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 653 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 654 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 655 void mptcp_sched_init(void); 656 int mptcp_init_sched(struct mptcp_sock *msk, 657 struct mptcp_sched_ops *sched); 658 void mptcp_release_sched(struct mptcp_sock *msk); 659 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 660 bool scheduled); 661 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 662 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 663 int mptcp_sched_get_send(struct mptcp_sock *msk); 664 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 665 666 static inline bool __tcp_can_send(const struct sock *ssk) 667 { 668 /* only send if our side has not closed yet */ 669 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 670 } 671 672 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 673 { 674 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 675 if (subflow->request_join && !subflow->fully_established) 676 return false; 677 678 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 679 } 680 681 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 682 683 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 684 685 void mptcp_subflow_drop_ctx(struct sock *ssk); 686 687 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 688 struct mptcp_subflow_context *ctx) 689 { 690 sk->sk_data_ready = sock_def_readable; 691 sk->sk_state_change = ctx->tcp_state_change; 692 sk->sk_write_space = sk_stream_write_space; 693 sk->sk_error_report = ctx->tcp_error_report; 694 695 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 696 } 697 698 void __init mptcp_proto_init(void); 699 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 700 int __init mptcp_proto_v6_init(void); 701 #endif 702 703 struct sock *mptcp_sk_clone_init(const struct sock *sk, 704 const struct mptcp_options_received *mp_opt, 705 struct sock *ssk, 706 struct request_sock *req); 707 void mptcp_get_options(const struct sk_buff *skb, 708 struct mptcp_options_received *mp_opt); 709 710 void mptcp_finish_connect(struct sock *sk); 711 void __mptcp_set_connected(struct sock *sk); 712 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 713 714 static inline void mptcp_stop_tout_timer(struct sock *sk) 715 { 716 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 717 return; 718 719 sk_stop_timer(sk, &sk->sk_timer); 720 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 721 } 722 723 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 724 { 725 /* avoid 0 timestamp, as that means no close timeout */ 726 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 727 } 728 729 static inline void mptcp_start_tout_timer(struct sock *sk) 730 { 731 mptcp_set_close_tout(sk, tcp_jiffies32); 732 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 733 } 734 735 static inline bool mptcp_is_fully_established(struct sock *sk) 736 { 737 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 738 READ_ONCE(mptcp_sk(sk)->fully_established); 739 } 740 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 741 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 742 bool mptcp_finish_join(struct sock *sk); 743 bool mptcp_schedule_work(struct sock *sk); 744 int mptcp_setsockopt(struct sock *sk, int level, int optname, 745 sockptr_t optval, unsigned int optlen); 746 int mptcp_getsockopt(struct sock *sk, int level, int optname, 747 char __user *optval, int __user *option); 748 749 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 750 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 751 { 752 if (use_64bit) 753 return cur_seq; 754 755 return __mptcp_expand_seq(old_seq, cur_seq); 756 } 757 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 758 void __mptcp_data_acked(struct sock *sk); 759 void __mptcp_error_report(struct sock *sk); 760 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 761 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 762 { 763 return READ_ONCE(msk->snd_data_fin_enable) && 764 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 765 } 766 767 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 768 { 769 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 770 return false; 771 772 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 773 return true; 774 } 775 776 static inline void mptcp_write_space(struct sock *sk) 777 { 778 if (sk_stream_is_writeable(sk)) { 779 /* pairs with memory barrier in mptcp_poll */ 780 smp_mb(); 781 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 782 sk_stream_write_space(sk); 783 } 784 } 785 786 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 787 788 #define MPTCP_TOKEN_MAX_RETRIES 4 789 790 void __init mptcp_token_init(void); 791 static inline void mptcp_token_init_request(struct request_sock *req) 792 { 793 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 794 } 795 796 int mptcp_token_new_request(struct request_sock *req); 797 void mptcp_token_destroy_request(struct request_sock *req); 798 int mptcp_token_new_connect(struct sock *ssk); 799 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 800 struct mptcp_sock *msk); 801 bool mptcp_token_exists(u32 token); 802 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 803 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 804 long *s_num); 805 void mptcp_token_destroy(struct mptcp_sock *msk); 806 807 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 808 809 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 810 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 811 812 void __init mptcp_pm_init(void); 813 void mptcp_pm_data_init(struct mptcp_sock *msk); 814 void mptcp_pm_data_reset(struct mptcp_sock *msk); 815 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 816 struct mptcp_addr_info *addr); 817 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 818 bool require_family, 819 struct mptcp_pm_addr_entry *entry); 820 bool mptcp_pm_addr_families_match(const struct sock *sk, 821 const struct mptcp_addr_info *loc, 822 const struct mptcp_addr_info *rem); 823 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 824 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 825 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 826 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 827 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 828 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 829 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 830 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 831 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk, 832 const struct mptcp_subflow_context *subflow); 833 void mptcp_pm_add_addr_received(const struct sock *ssk, 834 const struct mptcp_addr_info *addr); 835 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 836 const struct mptcp_addr_info *addr); 837 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 838 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 839 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 840 const struct mptcp_rm_list *rm_list); 841 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 842 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 843 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 844 struct mptcp_addr_info *addr, 845 struct mptcp_addr_info *rem, 846 u8 bkup); 847 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 848 const struct mptcp_addr_info *addr); 849 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 850 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 851 struct mptcp_pm_add_entry * 852 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 853 const struct mptcp_addr_info *addr, bool check_id); 854 struct mptcp_pm_add_entry * 855 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 856 const struct mptcp_addr_info *addr); 857 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 858 unsigned int id, 859 u8 *flags, int *ifindex); 860 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id, 861 u8 *flags, int *ifindex); 862 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 863 unsigned int id, 864 u8 *flags, int *ifindex); 865 int mptcp_pm_set_flags(struct net *net, struct nlattr *token, 866 struct mptcp_pm_addr_entry *loc, 867 struct mptcp_pm_addr_entry *rem, u8 bkup); 868 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup); 869 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token, 870 struct mptcp_pm_addr_entry *loc, 871 struct mptcp_pm_addr_entry *rem, u8 bkup); 872 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 873 const struct mptcp_addr_info *addr, 874 bool echo); 875 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 876 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 877 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list); 878 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk, 879 struct list_head *rm_list); 880 881 void mptcp_free_local_addr_list(struct mptcp_sock *msk); 882 883 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 884 const struct sock *ssk, gfp_t gfp); 885 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 886 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 887 void mptcp_event_pm_listener(const struct sock *ssk, 888 enum mptcp_event_type event); 889 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 890 891 void mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow, 892 const struct mptcp_options_received *mp_opt); 893 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 894 struct request_sock *req); 895 896 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 897 { 898 return READ_ONCE(msk->pm.addr_signal) & 899 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 900 } 901 902 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 903 { 904 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 905 } 906 907 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 908 { 909 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 910 } 911 912 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 913 { 914 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 915 } 916 917 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 918 { 919 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 920 } 921 922 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 923 { 924 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 925 } 926 927 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 928 { 929 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 930 931 if (family == AF_INET6) 932 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 933 if (!echo) 934 len += MPTCPOPT_THMAC_LEN; 935 /* account for 2 trailing 'nop' options */ 936 if (port) 937 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 938 939 return len; 940 } 941 942 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 943 { 944 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 945 return -EINVAL; 946 947 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 948 } 949 950 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 951 unsigned int opt_size, unsigned int remaining, 952 struct mptcp_addr_info *addr, bool *echo, 953 bool *drop_other_suboptions); 954 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 955 struct mptcp_rm_list *rm_list); 956 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 957 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 958 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 959 960 void __init mptcp_pm_nl_init(void); 961 void mptcp_pm_nl_work(struct mptcp_sock *msk); 962 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 963 const struct mptcp_rm_list *rm_list); 964 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 965 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 966 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 967 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 968 969 /* called under PM lock */ 970 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 971 { 972 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 973 WRITE_ONCE(msk->pm.accept_subflow, true); 974 } 975 976 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 977 { 978 spin_lock_bh(&msk->pm.lock); 979 __mptcp_pm_close_subflow(msk); 980 spin_unlock_bh(&msk->pm.lock); 981 } 982 983 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 984 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 985 986 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 987 { 988 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 989 } 990 991 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 992 993 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 994 { 995 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 996 } 997 998 static inline bool mptcp_check_fallback(const struct sock *sk) 999 { 1000 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1001 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1002 1003 return __mptcp_check_fallback(msk); 1004 } 1005 1006 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 1007 { 1008 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 1009 pr_debug("TCP fallback already done (msk=%p)", msk); 1010 return; 1011 } 1012 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1013 } 1014 1015 static inline void mptcp_do_fallback(struct sock *ssk) 1016 { 1017 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1018 struct sock *sk = subflow->conn; 1019 struct mptcp_sock *msk; 1020 1021 msk = mptcp_sk(sk); 1022 __mptcp_do_fallback(msk); 1023 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1024 gfp_t saved_allocation = ssk->sk_allocation; 1025 1026 /* we are in a atomic (BH) scope, override ssk default for data 1027 * fin allocation 1028 */ 1029 ssk->sk_allocation = GFP_ATOMIC; 1030 ssk->sk_shutdown |= SEND_SHUTDOWN; 1031 tcp_shutdown(ssk, SEND_SHUTDOWN); 1032 ssk->sk_allocation = saved_allocation; 1033 } 1034 } 1035 1036 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 1037 1038 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1039 { 1040 struct mptcp_ext *mpext; 1041 1042 mpext = skb ? mptcp_get_ext(skb) : NULL; 1043 if (mpext && mpext->infinite_map) 1044 return true; 1045 1046 return false; 1047 } 1048 1049 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1050 { 1051 return (subflow->request_mptcp || subflow->request_join); 1052 } 1053 1054 static inline bool subflow_simultaneous_connect(struct sock *sk) 1055 { 1056 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1057 1058 return sk->sk_state == TCP_ESTABLISHED && 1059 is_active_ssk(subflow) && 1060 !subflow->conn_finished; 1061 } 1062 1063 #ifdef CONFIG_SYN_COOKIES 1064 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1065 struct sk_buff *skb); 1066 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1067 struct sk_buff *skb); 1068 void __init mptcp_join_cookie_init(void); 1069 #else 1070 static inline void 1071 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1072 struct sk_buff *skb) {} 1073 static inline bool 1074 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1075 struct sk_buff *skb) 1076 { 1077 return false; 1078 } 1079 1080 static inline void mptcp_join_cookie_init(void) {} 1081 #endif 1082 1083 #endif /* __MPTCP_PROTOCOL_H */ 1084