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 #include <net/rstreason.h> 16 17 #include "mptcp_pm_gen.h" 18 19 #define MPTCP_SUPPORTED_VERSION 1 20 21 /* MPTCP option bits */ 22 #define OPTION_MPTCP_MPC_SYN BIT(0) 23 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 24 #define OPTION_MPTCP_MPC_ACK BIT(2) 25 #define OPTION_MPTCP_MPJ_SYN BIT(3) 26 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 27 #define OPTION_MPTCP_MPJ_ACK BIT(5) 28 #define OPTION_MPTCP_ADD_ADDR BIT(6) 29 #define OPTION_MPTCP_RM_ADDR BIT(7) 30 #define OPTION_MPTCP_FASTCLOSE BIT(8) 31 #define OPTION_MPTCP_PRIO BIT(9) 32 #define OPTION_MPTCP_RST BIT(10) 33 #define OPTION_MPTCP_DSS BIT(11) 34 #define OPTION_MPTCP_FAIL BIT(12) 35 36 #define OPTION_MPTCP_CSUMREQD BIT(13) 37 38 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \ 39 OPTION_MPTCP_MPC_ACK) 40 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \ 41 OPTION_MPTCP_MPJ_ACK) 42 43 /* MPTCP option subtypes */ 44 #define MPTCPOPT_MP_CAPABLE 0 45 #define MPTCPOPT_MP_JOIN 1 46 #define MPTCPOPT_DSS 2 47 #define MPTCPOPT_ADD_ADDR 3 48 #define MPTCPOPT_RM_ADDR 4 49 #define MPTCPOPT_MP_PRIO 5 50 #define MPTCPOPT_MP_FAIL 6 51 #define MPTCPOPT_MP_FASTCLOSE 7 52 #define MPTCPOPT_RST 8 53 54 /* MPTCP suboption lengths */ 55 #define TCPOLEN_MPTCP_MPC_SYN 4 56 #define TCPOLEN_MPTCP_MPC_SYNACK 12 57 #define TCPOLEN_MPTCP_MPC_ACK 20 58 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 59 #define TCPOLEN_MPTCP_MPJ_SYN 12 60 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 61 #define TCPOLEN_MPTCP_MPJ_ACK 24 62 #define TCPOLEN_MPTCP_DSS_BASE 4 63 #define TCPOLEN_MPTCP_DSS_ACK32 4 64 #define TCPOLEN_MPTCP_DSS_ACK64 8 65 #define TCPOLEN_MPTCP_DSS_MAP32 10 66 #define TCPOLEN_MPTCP_DSS_MAP64 14 67 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 68 #define TCPOLEN_MPTCP_ADD_ADDR 16 69 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 70 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 71 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 72 #define TCPOLEN_MPTCP_ADD_ADDR6 28 73 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 74 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 75 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 76 #define TCPOLEN_MPTCP_PORT_LEN 2 77 #define TCPOLEN_MPTCP_PORT_ALIGN 2 78 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 79 #define TCPOLEN_MPTCP_PRIO 3 80 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 81 #define TCPOLEN_MPTCP_FASTCLOSE 12 82 #define TCPOLEN_MPTCP_RST 4 83 #define TCPOLEN_MPTCP_FAIL 12 84 85 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 86 87 /* MPTCP MP_JOIN flags */ 88 #define MPTCPOPT_BACKUP BIT(0) 89 #define MPTCPOPT_THMAC_LEN 8 90 91 /* MPTCP MP_CAPABLE flags */ 92 #define MPTCP_VERSION_MASK (0x0F) 93 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 94 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 95 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 96 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 97 #define MPTCP_CAP_FLAG_MASK (0x1F) 98 99 /* MPTCP DSS flags */ 100 #define MPTCP_DSS_DATA_FIN BIT(4) 101 #define MPTCP_DSS_DSN64 BIT(3) 102 #define MPTCP_DSS_HAS_MAP BIT(2) 103 #define MPTCP_DSS_ACK64 BIT(1) 104 #define MPTCP_DSS_HAS_ACK BIT(0) 105 #define MPTCP_DSS_FLAG_MASK (0x1F) 106 107 /* MPTCP ADD_ADDR flags */ 108 #define MPTCP_ADDR_ECHO BIT(0) 109 110 /* MPTCP MP_PRIO flags */ 111 #define MPTCP_PRIO_BKUP BIT(0) 112 113 /* MPTCP TCPRST flags */ 114 #define MPTCP_RST_TRANSIENT BIT(0) 115 116 /* MPTCP socket atomic flags */ 117 #define MPTCP_WORK_RTX 1 118 #define MPTCP_FALLBACK_DONE 2 119 #define MPTCP_WORK_CLOSE_SUBFLOW 3 120 121 /* 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_SYNC_STATE 6 128 #define MPTCP_SYNC_SNDBUF 7 129 130 struct mptcp_skb_cb { 131 u64 map_seq; 132 u64 end_seq; 133 u32 offset; 134 u8 has_rxtstamp:1; 135 }; 136 137 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0])) 138 139 static inline bool before64(__u64 seq1, __u64 seq2) 140 { 141 return (__s64)(seq1 - seq2) < 0; 142 } 143 144 #define after64(seq2, seq1) before64(seq1, seq2) 145 146 struct mptcp_options_received { 147 u64 sndr_key; 148 u64 rcvr_key; 149 u64 data_ack; 150 u64 data_seq; 151 u32 subflow_seq; 152 u16 data_len; 153 __sum16 csum; 154 u16 suboptions; 155 u32 token; 156 u32 nonce; 157 u16 use_map:1, 158 dsn64:1, 159 data_fin:1, 160 use_ack:1, 161 ack64:1, 162 mpc_map:1, 163 reset_reason:4, 164 reset_transient:1, 165 echo:1, 166 backup:1, 167 deny_join_id0:1, 168 __unused:2; 169 u8 join_id; 170 u64 thmac; 171 u8 hmac[MPTCPOPT_HMAC_LEN]; 172 struct mptcp_addr_info addr; 173 struct mptcp_rm_list rm_list; 174 u64 ahmac; 175 u64 fail_seq; 176 }; 177 178 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 179 { 180 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 181 ((nib & 0xF) << 8) | field); 182 } 183 184 enum mptcp_pm_status { 185 MPTCP_PM_ADD_ADDR_RECEIVED, 186 MPTCP_PM_ADD_ADDR_SEND_ACK, 187 MPTCP_PM_RM_ADDR_RECEIVED, 188 MPTCP_PM_ESTABLISHED, 189 MPTCP_PM_SUBFLOW_ESTABLISHED, 190 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 191 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 192 * accounted int id_avail_bitmap 193 */ 194 }; 195 196 enum mptcp_pm_type { 197 MPTCP_PM_TYPE_KERNEL = 0, 198 MPTCP_PM_TYPE_USERSPACE, 199 200 __MPTCP_PM_TYPE_NR, 201 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 202 }; 203 204 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 205 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 206 207 enum mptcp_addr_signal_status { 208 MPTCP_ADD_ADDR_SIGNAL, 209 MPTCP_ADD_ADDR_ECHO, 210 MPTCP_RM_ADDR_SIGNAL, 211 }; 212 213 /* max value of mptcp_addr_info.id */ 214 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 215 216 struct mptcp_pm_data { 217 struct mptcp_addr_info local; 218 struct mptcp_addr_info remote; 219 struct list_head anno_list; 220 struct list_head userspace_pm_local_addr_list; 221 222 spinlock_t lock; /*protects the whole PM data */ 223 224 u8 addr_signal; 225 bool server_side; 226 bool work_pending; 227 bool accept_addr; 228 bool accept_subflow; 229 bool remote_deny_join_id0; 230 u8 add_addr_signaled; 231 u8 add_addr_accepted; 232 u8 local_addr_used; 233 u8 pm_type; 234 u8 subflows; 235 u8 status; 236 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 237 struct mptcp_rm_list rm_list_tx; 238 struct mptcp_rm_list rm_list_rx; 239 }; 240 241 struct mptcp_pm_addr_entry { 242 struct list_head list; 243 struct mptcp_addr_info addr; 244 u8 flags; 245 int ifindex; 246 struct socket *lsk; 247 }; 248 249 struct mptcp_data_frag { 250 struct list_head list; 251 u64 data_seq; 252 u16 data_len; 253 u16 offset; 254 u16 overhead; 255 u16 already_sent; 256 struct page *page; 257 }; 258 259 /* MPTCP connection sock */ 260 struct mptcp_sock { 261 /* inet_connection_sock must be the first member */ 262 struct inet_connection_sock sk; 263 u64 local_key; /* protected by the first subflow socket lock 264 * lockless access read 265 */ 266 u64 remote_key; /* same as above */ 267 u64 write_seq; 268 u64 bytes_sent; 269 u64 snd_nxt; 270 u64 bytes_received; 271 u64 ack_seq; 272 atomic64_t rcv_wnd_sent; 273 u64 rcv_data_fin_seq; 274 u64 bytes_retrans; 275 u64 bytes_consumed; 276 int rmem_fwd_alloc; 277 int snd_burst; 278 int old_wspace; 279 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 280 * recovery related fields are under data_lock 281 * protection 282 */ 283 u64 bytes_acked; 284 u64 snd_una; 285 u64 wnd_end; 286 u32 last_data_sent; 287 u32 last_data_recv; 288 u32 last_ack_recv; 289 unsigned long timer_ival; 290 u32 token; 291 int rmem_released; 292 unsigned long flags; 293 unsigned long cb_flags; 294 bool recovery; /* closing subflow write queue reinjected */ 295 bool can_ack; 296 bool fully_established; 297 bool rcv_data_fin; 298 bool snd_data_fin_enable; 299 bool rcv_fastclose; 300 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 301 bool csum_enabled; 302 bool allow_infinite_fallback; 303 u8 pending_state; /* A subflow asked to set this sk_state, 304 * protected by the msk data lock 305 */ 306 u8 mpc_endpoint_id; 307 u8 recvmsg_inq:1, 308 cork:1, 309 nodelay:1, 310 fastopening:1, 311 in_accept_queue:1, 312 free_first:1, 313 rcvspace_init:1; 314 u32 notsent_lowat; 315 struct work_struct work; 316 struct sk_buff *ooo_last_skb; 317 struct rb_root out_of_order_queue; 318 struct sk_buff_head receive_queue; 319 struct list_head conn_list; 320 struct list_head rtx_queue; 321 struct mptcp_data_frag *first_pending; 322 struct list_head join_list; 323 struct sock *first; /* The mptcp ops can safely dereference, using suitable 324 * ONCE annotation, the subflow outside the socket 325 * lock as such sock is freed after close(). 326 */ 327 struct mptcp_pm_data pm; 328 struct mptcp_sched_ops *sched; 329 struct { 330 u32 space; /* bytes copied in last measurement window */ 331 u32 copied; /* bytes copied in this measurement window */ 332 u64 time; /* start time of measurement window */ 333 u64 rtt_us; /* last maximum rtt of subflows */ 334 } rcvq_space; 335 u8 scaling_ratio; 336 337 u32 subflow_id; 338 u32 setsockopt_seq; 339 char ca_name[TCP_CA_NAME_MAX]; 340 }; 341 342 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 343 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 344 345 #define mptcp_for_each_subflow(__msk, __subflow) \ 346 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 347 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 348 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 349 350 extern struct genl_family mptcp_genl_family; 351 352 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 353 { 354 sock_owned_by_me((const struct sock *)msk); 355 } 356 357 #ifdef CONFIG_DEBUG_NET 358 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 359 #undef tcp_sk 360 #define tcp_sk(ptr) ({ \ 361 typeof(ptr) _ptr = (ptr); \ 362 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 363 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 364 }) 365 #define mptcp_sk(ptr) ({ \ 366 typeof(ptr) _ptr = (ptr); \ 367 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 368 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 369 }) 370 371 #else /* !CONFIG_DEBUG_NET */ 372 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 373 #endif 374 375 /* the msk socket don't use the backlog, also account for the bulk 376 * free memory 377 */ 378 static inline int __mptcp_rmem(const struct sock *sk) 379 { 380 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 381 } 382 383 static inline int mptcp_win_from_space(const struct sock *sk, int space) 384 { 385 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 386 } 387 388 static inline int __mptcp_space(const struct sock *sk) 389 { 390 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 391 } 392 393 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 394 { 395 const struct mptcp_sock *msk = mptcp_sk(sk); 396 397 return READ_ONCE(msk->first_pending); 398 } 399 400 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 401 { 402 struct mptcp_sock *msk = mptcp_sk(sk); 403 struct mptcp_data_frag *cur; 404 405 cur = msk->first_pending; 406 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 407 list_next_entry(cur, list); 408 } 409 410 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 411 { 412 const struct mptcp_sock *msk = mptcp_sk(sk); 413 414 if (!msk->first_pending) 415 return NULL; 416 417 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 418 return NULL; 419 420 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 421 } 422 423 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 424 { 425 struct mptcp_sock *msk = mptcp_sk(sk); 426 427 if (msk->snd_una == msk->snd_nxt) 428 return NULL; 429 430 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 431 } 432 433 struct csum_pseudo_header { 434 __be64 data_seq; 435 __be32 subflow_seq; 436 __be16 data_len; 437 __sum16 csum; 438 }; 439 440 struct mptcp_subflow_request_sock { 441 struct tcp_request_sock sk; 442 u16 mp_capable : 1, 443 mp_join : 1, 444 backup : 1, 445 csum_reqd : 1, 446 allow_join_id0 : 1; 447 u8 local_id; 448 u8 remote_id; 449 u64 local_key; 450 u64 idsn; 451 u32 token; 452 u32 ssn_offset; 453 u64 thmac; 454 u32 local_nonce; 455 u32 remote_nonce; 456 struct mptcp_sock *msk; 457 struct hlist_nulls_node token_node; 458 }; 459 460 static inline struct mptcp_subflow_request_sock * 461 mptcp_subflow_rsk(const struct request_sock *rsk) 462 { 463 return (struct mptcp_subflow_request_sock *)rsk; 464 } 465 466 struct mptcp_delegated_action { 467 struct napi_struct napi; 468 struct list_head head; 469 }; 470 471 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 472 473 #define MPTCP_DELEGATE_SCHEDULED 0 474 #define MPTCP_DELEGATE_SEND 1 475 #define MPTCP_DELEGATE_ACK 2 476 #define MPTCP_DELEGATE_SNDBUF 3 477 478 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 479 /* MPTCP subflow context */ 480 struct mptcp_subflow_context { 481 struct list_head node;/* conn_list of subflows */ 482 483 struct_group(reset, 484 485 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 486 u64 local_key; 487 u64 remote_key; 488 u64 idsn; 489 u64 map_seq; 490 u32 snd_isn; 491 u32 token; 492 u32 rel_write_seq; 493 u32 map_subflow_seq; 494 u32 ssn_offset; 495 u32 map_data_len; 496 __wsum map_data_csum; 497 u32 map_csum_len; 498 u32 request_mptcp : 1, /* send MP_CAPABLE */ 499 request_join : 1, /* send MP_JOIN */ 500 request_bkup : 1, 501 mp_capable : 1, /* remote is MPTCP capable */ 502 mp_join : 1, /* remote is JOINing */ 503 fully_established : 1, /* path validated */ 504 pm_notified : 1, /* PM hook called for established status */ 505 conn_finished : 1, 506 map_valid : 1, 507 map_csum_reqd : 1, 508 map_data_fin : 1, 509 mpc_map : 1, 510 backup : 1, 511 send_mp_prio : 1, 512 send_mp_fail : 1, 513 send_fastclose : 1, 514 send_infinite_map : 1, 515 remote_key_valid : 1, /* received the peer key from */ 516 disposable : 1, /* ctx can be free at ulp release time */ 517 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 518 valid_csum_seen : 1, /* at least one csum validated */ 519 is_mptfo : 1, /* subflow is doing TFO */ 520 __unused : 10; 521 bool data_avail; 522 bool scheduled; 523 u32 remote_nonce; 524 u64 thmac; 525 u32 local_nonce; 526 u32 remote_token; 527 union { 528 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 529 u64 iasn; /* initial ack sequence number, MPC subflows only */ 530 }; 531 s16 local_id; /* if negative not initialized yet */ 532 u8 remote_id; 533 u8 reset_seen:1; 534 u8 reset_transient:1; 535 u8 reset_reason:4; 536 u8 stale_count; 537 538 u32 subflow_id; 539 540 long delegated_status; 541 unsigned long fail_tout; 542 543 ); 544 545 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 546 547 u32 setsockopt_seq; 548 u32 stale_rcv_tstamp; 549 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 550 * protected by the msk socket lock 551 */ 552 553 struct sock *tcp_sock; /* tcp sk backpointer */ 554 struct sock *conn; /* parent mptcp_sock */ 555 const struct inet_connection_sock_af_ops *icsk_af_ops; 556 void (*tcp_state_change)(struct sock *sk); 557 void (*tcp_error_report)(struct sock *sk); 558 559 struct rcu_head rcu; 560 }; 561 562 static inline struct mptcp_subflow_context * 563 mptcp_subflow_ctx(const struct sock *sk) 564 { 565 const struct inet_connection_sock *icsk = inet_csk(sk); 566 567 /* Use RCU on icsk_ulp_data only for sock diag code */ 568 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 569 } 570 571 static inline struct sock * 572 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 573 { 574 return subflow->tcp_sock; 575 } 576 577 static inline void 578 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 579 { 580 memset(&subflow->reset, 0, sizeof(subflow->reset)); 581 subflow->request_mptcp = 1; 582 WRITE_ONCE(subflow->local_id, -1); 583 } 584 585 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */ 586 static inline enum sk_rst_reason 587 sk_rst_convert_mptcp_reason(u32 reason) 588 { 589 switch (reason) { 590 case MPTCP_RST_EUNSPEC: 591 return SK_RST_REASON_MPTCP_RST_EUNSPEC; 592 case MPTCP_RST_EMPTCP: 593 return SK_RST_REASON_MPTCP_RST_EMPTCP; 594 case MPTCP_RST_ERESOURCE: 595 return SK_RST_REASON_MPTCP_RST_ERESOURCE; 596 case MPTCP_RST_EPROHIBIT: 597 return SK_RST_REASON_MPTCP_RST_EPROHIBIT; 598 case MPTCP_RST_EWQ2BIG: 599 return SK_RST_REASON_MPTCP_RST_EWQ2BIG; 600 case MPTCP_RST_EBADPERF: 601 return SK_RST_REASON_MPTCP_RST_EBADPERF; 602 case MPTCP_RST_EMIDDLEBOX: 603 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX; 604 default: 605 /* It should not happen, or else errors may occur 606 * in MPTCP layer 607 */ 608 return SK_RST_REASON_ERROR; 609 } 610 } 611 612 static inline void 613 mptcp_send_active_reset_reason(struct sock *sk) 614 { 615 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 616 enum sk_rst_reason reason; 617 618 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason); 619 tcp_send_active_reset(sk, GFP_ATOMIC, reason); 620 } 621 622 static inline u64 623 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 624 { 625 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 626 subflow->ssn_offset - 627 subflow->map_subflow_seq; 628 } 629 630 static inline u64 631 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 632 { 633 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 634 } 635 636 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 637 638 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 639 { 640 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 641 struct mptcp_delegated_action *delegated; 642 bool schedule; 643 644 /* the caller held the subflow bh socket lock */ 645 lockdep_assert_in_softirq(); 646 647 /* The implied barrier pairs with tcp_release_cb_override() 648 * mptcp_napi_poll(), and ensures the below list check sees list 649 * updates done prior to delegated status bits changes 650 */ 651 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 652 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 653 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 654 return; 655 656 delegated = this_cpu_ptr(&mptcp_delegated_actions); 657 schedule = list_empty(&delegated->head); 658 list_add_tail(&subflow->delegated_node, &delegated->head); 659 sock_hold(mptcp_subflow_tcp_sock(subflow)); 660 if (schedule) 661 napi_schedule(&delegated->napi); 662 } 663 } 664 665 static inline struct mptcp_subflow_context * 666 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 667 { 668 struct mptcp_subflow_context *ret; 669 670 if (list_empty(&delegated->head)) 671 return NULL; 672 673 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 674 list_del_init(&ret->delegated_node); 675 return ret; 676 } 677 678 int mptcp_is_enabled(const struct net *net); 679 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 680 int mptcp_is_checksum_enabled(const struct net *net); 681 int mptcp_allow_join_id0(const struct net *net); 682 unsigned int mptcp_stale_loss_cnt(const struct net *net); 683 unsigned int mptcp_close_timeout(const struct sock *sk); 684 int mptcp_get_pm_type(const struct net *net); 685 const char *mptcp_get_scheduler(const struct net *net); 686 void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 687 struct mptcp_subflow_context *subflow, 688 const struct mptcp_options_received *mp_opt); 689 bool __mptcp_retransmit_pending_data(struct sock *sk); 690 void mptcp_check_and_set_pending(struct sock *sk); 691 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 692 bool mptcp_subflow_data_available(struct sock *sk); 693 void __init mptcp_subflow_init(void); 694 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 695 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 696 struct mptcp_subflow_context *subflow); 697 void __mptcp_subflow_send_ack(struct sock *ssk); 698 void mptcp_subflow_reset(struct sock *ssk); 699 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 700 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 701 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 702 bool __mptcp_close(struct sock *sk, long timeout); 703 void mptcp_cancel_work(struct sock *sk); 704 void __mptcp_unaccepted_force_close(struct sock *sk); 705 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk); 706 void mptcp_set_state(struct sock *sk, int state); 707 708 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 709 const struct mptcp_addr_info *b, bool use_port); 710 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr); 711 712 /* called with sk socket lock held */ 713 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 714 const struct mptcp_addr_info *remote); 715 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 716 struct socket **new_sock); 717 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 718 struct sockaddr_storage *addr, 719 unsigned short family); 720 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 721 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 722 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 723 void mptcp_sched_init(void); 724 int mptcp_init_sched(struct mptcp_sock *msk, 725 struct mptcp_sched_ops *sched); 726 void mptcp_release_sched(struct mptcp_sock *msk); 727 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 728 bool scheduled); 729 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 730 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 731 int mptcp_sched_get_send(struct mptcp_sock *msk); 732 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 733 734 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 735 { 736 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 737 } 738 739 static inline bool mptcp_epollin_ready(const struct sock *sk) 740 { 741 /* mptcp doesn't have to deal with small skbs in the receive queue, 742 * at it can always coalesce them 743 */ 744 return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) || 745 (mem_cgroup_sockets_enabled && sk->sk_memcg && 746 mem_cgroup_under_socket_pressure(sk->sk_memcg)) || 747 READ_ONCE(tcp_memory_pressure); 748 } 749 750 int mptcp_set_rcvlowat(struct sock *sk, int val); 751 752 static inline bool __tcp_can_send(const struct sock *ssk) 753 { 754 /* only send if our side has not closed yet */ 755 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 756 } 757 758 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 759 { 760 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 761 if (subflow->request_join && !subflow->fully_established) 762 return false; 763 764 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 765 } 766 767 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 768 769 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 770 771 void mptcp_subflow_drop_ctx(struct sock *ssk); 772 773 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 774 struct mptcp_subflow_context *ctx) 775 { 776 sk->sk_data_ready = sock_def_readable; 777 sk->sk_state_change = ctx->tcp_state_change; 778 sk->sk_write_space = sk_stream_write_space; 779 sk->sk_error_report = ctx->tcp_error_report; 780 781 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 782 } 783 784 void __init mptcp_proto_init(void); 785 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 786 int __init mptcp_proto_v6_init(void); 787 #endif 788 789 struct sock *mptcp_sk_clone_init(const struct sock *sk, 790 const struct mptcp_options_received *mp_opt, 791 struct sock *ssk, 792 struct request_sock *req); 793 void mptcp_get_options(const struct sk_buff *skb, 794 struct mptcp_options_received *mp_opt); 795 796 void mptcp_finish_connect(struct sock *sk); 797 void __mptcp_sync_state(struct sock *sk, int state); 798 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 799 800 static inline void mptcp_stop_tout_timer(struct sock *sk) 801 { 802 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 803 return; 804 805 sk_stop_timer(sk, &sk->sk_timer); 806 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 807 } 808 809 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 810 { 811 /* avoid 0 timestamp, as that means no close timeout */ 812 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 813 } 814 815 static inline void mptcp_start_tout_timer(struct sock *sk) 816 { 817 mptcp_set_close_tout(sk, tcp_jiffies32); 818 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 819 } 820 821 static inline bool mptcp_is_fully_established(struct sock *sk) 822 { 823 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 824 READ_ONCE(mptcp_sk(sk)->fully_established); 825 } 826 827 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 828 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 829 bool mptcp_finish_join(struct sock *sk); 830 bool mptcp_schedule_work(struct sock *sk); 831 int mptcp_setsockopt(struct sock *sk, int level, int optname, 832 sockptr_t optval, unsigned int optlen); 833 int mptcp_getsockopt(struct sock *sk, int level, int optname, 834 char __user *optval, int __user *option); 835 836 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 837 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 838 { 839 if (use_64bit) 840 return cur_seq; 841 842 return __mptcp_expand_seq(old_seq, cur_seq); 843 } 844 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 845 void __mptcp_data_acked(struct sock *sk); 846 void __mptcp_error_report(struct sock *sk); 847 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 848 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 849 { 850 return READ_ONCE(msk->snd_data_fin_enable) && 851 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 852 } 853 854 static inline u32 mptcp_notsent_lowat(const struct sock *sk) 855 { 856 struct net *net = sock_net(sk); 857 u32 val; 858 859 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat); 860 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat); 861 } 862 863 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake) 864 { 865 const struct mptcp_sock *msk = mptcp_sk(sk); 866 u32 notsent_bytes; 867 868 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt); 869 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk); 870 } 871 872 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake) 873 { 874 return mptcp_stream_memory_free(sk, wake) && 875 __sk_stream_is_writeable(sk, wake); 876 } 877 878 static inline void mptcp_write_space(struct sock *sk) 879 { 880 /* pairs with memory barrier in mptcp_poll */ 881 smp_mb(); 882 if (mptcp_stream_memory_free(sk, 1)) 883 sk_stream_write_space(sk); 884 } 885 886 static inline void __mptcp_sync_sndbuf(struct sock *sk) 887 { 888 struct mptcp_subflow_context *subflow; 889 int ssk_sndbuf, new_sndbuf; 890 891 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 892 return; 893 894 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]); 895 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 896 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 897 898 subflow->cached_sndbuf = ssk_sndbuf; 899 new_sndbuf += ssk_sndbuf; 900 } 901 902 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 903 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 904 mptcp_write_space(sk); 905 } 906 907 /* The called held both the msk socket and the subflow socket locks, 908 * possibly under BH 909 */ 910 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 911 { 912 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 913 914 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 915 __mptcp_sync_sndbuf(sk); 916 } 917 918 /* the caller held only the subflow socket lock, either in process or 919 * BH context. Additionally this can be called under the msk data lock, 920 * so we can't acquire such lock here: let the delegate action acquires 921 * the needed locks in suitable order. 922 */ 923 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 924 { 925 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 926 927 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 928 return; 929 930 local_bh_disable(); 931 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 932 local_bh_enable(); 933 } 934 935 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 936 937 #define MPTCP_TOKEN_MAX_RETRIES 4 938 939 void __init mptcp_token_init(void); 940 static inline void mptcp_token_init_request(struct request_sock *req) 941 { 942 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 943 } 944 945 int mptcp_token_new_request(struct request_sock *req); 946 void mptcp_token_destroy_request(struct request_sock *req); 947 int mptcp_token_new_connect(struct sock *ssk); 948 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 949 struct mptcp_sock *msk); 950 bool mptcp_token_exists(u32 token); 951 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 952 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 953 long *s_num); 954 void mptcp_token_destroy(struct mptcp_sock *msk); 955 956 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 957 958 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 959 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 960 961 void __init mptcp_pm_init(void); 962 void mptcp_pm_data_init(struct mptcp_sock *msk); 963 void mptcp_pm_data_reset(struct mptcp_sock *msk); 964 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 965 struct mptcp_addr_info *addr); 966 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 967 bool require_family, 968 struct mptcp_pm_addr_entry *entry); 969 bool mptcp_pm_addr_families_match(const struct sock *sk, 970 const struct mptcp_addr_info *loc, 971 const struct mptcp_addr_info *rem); 972 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 973 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 974 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 975 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 976 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 977 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 978 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 979 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 980 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 981 const struct mptcp_subflow_context *subflow); 982 void mptcp_pm_add_addr_received(const struct sock *ssk, 983 const struct mptcp_addr_info *addr); 984 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 985 const struct mptcp_addr_info *addr); 986 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 987 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 988 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 989 const struct mptcp_rm_list *rm_list); 990 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 991 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 992 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 993 struct mptcp_addr_info *addr, 994 struct mptcp_addr_info *rem, 995 u8 bkup); 996 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 997 const struct mptcp_addr_info *addr); 998 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 999 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 1000 struct mptcp_pm_add_entry * 1001 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 1002 const struct mptcp_addr_info *addr, bool check_id); 1003 struct mptcp_pm_add_entry * 1004 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 1005 const struct mptcp_addr_info *addr); 1006 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 1007 unsigned int id, 1008 u8 *flags, int *ifindex); 1009 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id, 1010 u8 *flags, int *ifindex); 1011 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 1012 unsigned int id, 1013 u8 *flags, int *ifindex); 1014 int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info); 1015 int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info); 1016 int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info); 1017 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 1018 const struct mptcp_addr_info *addr, 1019 bool echo); 1020 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 1021 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 1022 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list); 1023 1024 void mptcp_free_local_addr_list(struct mptcp_sock *msk); 1025 1026 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 1027 const struct sock *ssk, gfp_t gfp); 1028 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 1029 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 1030 void mptcp_event_pm_listener(const struct sock *ssk, 1031 enum mptcp_event_type event); 1032 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 1033 1034 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow, 1035 const struct mptcp_options_received *mp_opt); 1036 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 1037 struct request_sock *req); 1038 int mptcp_nl_fill_addr(struct sk_buff *skb, 1039 struct mptcp_pm_addr_entry *entry); 1040 1041 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 1042 { 1043 return READ_ONCE(msk->pm.addr_signal) & 1044 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 1045 } 1046 1047 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 1048 { 1049 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 1050 } 1051 1052 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 1053 { 1054 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 1055 } 1056 1057 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 1058 { 1059 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 1060 } 1061 1062 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1063 { 1064 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1065 } 1066 1067 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1068 { 1069 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1070 } 1071 1072 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 1073 { 1074 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 1075 1076 if (family == AF_INET6) 1077 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 1078 if (!echo) 1079 len += MPTCPOPT_THMAC_LEN; 1080 /* account for 2 trailing 'nop' options */ 1081 if (port) 1082 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 1083 1084 return len; 1085 } 1086 1087 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 1088 { 1089 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 1090 return -EINVAL; 1091 1092 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 1093 } 1094 1095 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 1096 unsigned int opt_size, unsigned int remaining, 1097 struct mptcp_addr_info *addr, bool *echo, 1098 bool *drop_other_suboptions); 1099 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1100 struct mptcp_rm_list *rm_list); 1101 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1102 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1103 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1104 int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb); 1105 int mptcp_pm_nl_dump_addr(struct sk_buff *msg, 1106 struct netlink_callback *cb); 1107 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg, 1108 struct netlink_callback *cb); 1109 int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info); 1110 int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info); 1111 int mptcp_userspace_pm_get_addr(struct sk_buff *skb, 1112 struct genl_info *info); 1113 1114 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1115 { 1116 int local_id = READ_ONCE(subflow->local_id); 1117 1118 if (local_id < 0) 1119 return 0; 1120 return local_id; 1121 } 1122 1123 void __init mptcp_pm_nl_init(void); 1124 void mptcp_pm_nl_work(struct mptcp_sock *msk); 1125 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 1126 const struct mptcp_rm_list *rm_list); 1127 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 1128 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 1129 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 1130 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 1131 1132 /* called under PM lock */ 1133 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1134 { 1135 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 1136 WRITE_ONCE(msk->pm.accept_subflow, true); 1137 } 1138 1139 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1140 { 1141 spin_lock_bh(&msk->pm.lock); 1142 __mptcp_pm_close_subflow(msk); 1143 spin_unlock_bh(&msk->pm.lock); 1144 } 1145 1146 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 1147 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1148 1149 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1150 { 1151 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1152 } 1153 1154 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1155 1156 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1157 { 1158 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1159 } 1160 1161 static inline bool mptcp_check_fallback(const struct sock *sk) 1162 { 1163 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1164 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1165 1166 return __mptcp_check_fallback(msk); 1167 } 1168 1169 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 1170 { 1171 if (__mptcp_check_fallback(msk)) { 1172 pr_debug("TCP fallback already done (msk=%p)", msk); 1173 return; 1174 } 1175 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1176 } 1177 1178 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1179 { 1180 struct sock *ssk = READ_ONCE(msk->first); 1181 1182 return ssk && ((1 << inet_sk_state_load(ssk)) & 1183 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1184 TCPF_SYN_RECV | TCPF_LISTEN)); 1185 } 1186 1187 static inline void mptcp_do_fallback(struct sock *ssk) 1188 { 1189 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1190 struct sock *sk = subflow->conn; 1191 struct mptcp_sock *msk; 1192 1193 msk = mptcp_sk(sk); 1194 __mptcp_do_fallback(msk); 1195 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1196 gfp_t saved_allocation = ssk->sk_allocation; 1197 1198 /* we are in a atomic (BH) scope, override ssk default for data 1199 * fin allocation 1200 */ 1201 ssk->sk_allocation = GFP_ATOMIC; 1202 ssk->sk_shutdown |= SEND_SHUTDOWN; 1203 tcp_shutdown(ssk, SEND_SHUTDOWN); 1204 ssk->sk_allocation = saved_allocation; 1205 } 1206 } 1207 1208 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 1209 1210 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1211 { 1212 struct mptcp_ext *mpext; 1213 1214 mpext = skb ? mptcp_get_ext(skb) : NULL; 1215 if (mpext && mpext->infinite_map) 1216 return true; 1217 1218 return false; 1219 } 1220 1221 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1222 { 1223 return (subflow->request_mptcp || subflow->request_join); 1224 } 1225 1226 static inline bool subflow_simultaneous_connect(struct sock *sk) 1227 { 1228 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1229 1230 return (1 << sk->sk_state) & 1231 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) && 1232 is_active_ssk(subflow) && 1233 !subflow->conn_finished; 1234 } 1235 1236 #ifdef CONFIG_SYN_COOKIES 1237 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1238 struct sk_buff *skb); 1239 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1240 struct sk_buff *skb); 1241 void __init mptcp_join_cookie_init(void); 1242 #else 1243 static inline void 1244 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1245 struct sk_buff *skb) {} 1246 static inline bool 1247 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1248 struct sk_buff *skb) 1249 { 1250 return false; 1251 } 1252 1253 static inline void mptcp_join_cookie_init(void) {} 1254 #endif 1255 1256 #endif /* __MPTCP_PROTOCOL_H */ 1257