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 #define MPTCP_SUPPORTED_VERSION 1 18 19 /* MPTCP option bits */ 20 #define OPTION_MPTCP_MPC_SYN BIT(0) 21 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 22 #define OPTION_MPTCP_MPC_ACK BIT(2) 23 #define OPTION_MPTCP_MPJ_SYN BIT(3) 24 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 25 #define OPTION_MPTCP_MPJ_ACK BIT(5) 26 #define OPTION_MPTCP_ADD_ADDR BIT(6) 27 #define OPTION_MPTCP_RM_ADDR BIT(7) 28 #define OPTION_MPTCP_FASTCLOSE BIT(8) 29 #define OPTION_MPTCP_PRIO BIT(9) 30 #define OPTION_MPTCP_RST BIT(10) 31 #define OPTION_MPTCP_DSS BIT(11) 32 #define OPTION_MPTCP_FAIL BIT(12) 33 34 #define OPTION_MPTCP_CSUMREQD BIT(13) 35 36 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \ 37 OPTION_MPTCP_MPC_ACK) 38 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \ 39 OPTION_MPTCP_MPJ_ACK) 40 41 /* MPTCP option subtypes */ 42 #define MPTCPOPT_MP_CAPABLE 0 43 #define MPTCPOPT_MP_JOIN 1 44 #define MPTCPOPT_DSS 2 45 #define MPTCPOPT_ADD_ADDR 3 46 #define MPTCPOPT_RM_ADDR 4 47 #define MPTCPOPT_MP_PRIO 5 48 #define MPTCPOPT_MP_FAIL 6 49 #define MPTCPOPT_MP_FASTCLOSE 7 50 #define MPTCPOPT_RST 8 51 52 /* MPTCP suboption lengths */ 53 #define TCPOLEN_MPTCP_MPC_SYN 4 54 #define TCPOLEN_MPTCP_MPC_SYNACK 12 55 #define TCPOLEN_MPTCP_MPC_ACK 20 56 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 57 #define TCPOLEN_MPTCP_MPJ_SYN 12 58 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 59 #define TCPOLEN_MPTCP_MPJ_ACK 24 60 #define TCPOLEN_MPTCP_DSS_BASE 4 61 #define TCPOLEN_MPTCP_DSS_ACK32 4 62 #define TCPOLEN_MPTCP_DSS_ACK64 8 63 #define TCPOLEN_MPTCP_DSS_MAP32 10 64 #define TCPOLEN_MPTCP_DSS_MAP64 14 65 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 66 #define TCPOLEN_MPTCP_ADD_ADDR 16 67 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 70 #define TCPOLEN_MPTCP_ADD_ADDR6 28 71 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 74 #define TCPOLEN_MPTCP_PORT_LEN 2 75 #define TCPOLEN_MPTCP_PORT_ALIGN 2 76 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 77 #define TCPOLEN_MPTCP_PRIO 3 78 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 79 #define TCPOLEN_MPTCP_FASTCLOSE 12 80 #define TCPOLEN_MPTCP_RST 4 81 #define TCPOLEN_MPTCP_FAIL 12 82 83 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 84 85 /* MPTCP MP_JOIN flags */ 86 #define MPTCPOPT_BACKUP BIT(0) 87 #define MPTCPOPT_THMAC_LEN 8 88 89 /* MPTCP MP_CAPABLE flags */ 90 #define MPTCP_VERSION_MASK (0x0F) 91 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 92 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 93 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 94 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 95 #define MPTCP_CAP_FLAG_MASK (0x1F) 96 97 /* MPTCP DSS flags */ 98 #define MPTCP_DSS_DATA_FIN BIT(4) 99 #define MPTCP_DSS_DSN64 BIT(3) 100 #define MPTCP_DSS_HAS_MAP BIT(2) 101 #define MPTCP_DSS_ACK64 BIT(1) 102 #define MPTCP_DSS_HAS_ACK BIT(0) 103 #define MPTCP_DSS_FLAG_MASK (0x1F) 104 105 /* MPTCP ADD_ADDR flags */ 106 #define MPTCP_ADDR_ECHO BIT(0) 107 108 /* MPTCP MP_PRIO flags */ 109 #define MPTCP_PRIO_BKUP BIT(0) 110 111 /* MPTCP TCPRST flags */ 112 #define MPTCP_RST_TRANSIENT BIT(0) 113 114 /* MPTCP socket atomic flags */ 115 #define MPTCP_WORK_RTX 1 116 #define MPTCP_FALLBACK_DONE 2 117 #define MPTCP_WORK_CLOSE_SUBFLOW 3 118 119 /* MPTCP socket release cb flags */ 120 #define MPTCP_PUSH_PENDING 1 121 #define MPTCP_CLEAN_UNA 2 122 #define MPTCP_ERROR_REPORT 3 123 #define MPTCP_RETRANSMIT 4 124 #define MPTCP_FLUSH_JOIN_LIST 5 125 #define MPTCP_SYNC_STATE 6 126 #define MPTCP_SYNC_SNDBUF 7 127 128 struct mptcp_skb_cb { 129 u64 map_seq; 130 u64 end_seq; 131 u32 offset; 132 u8 has_rxtstamp:1; 133 }; 134 135 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0])) 136 137 static inline bool before64(__u64 seq1, __u64 seq2) 138 { 139 return (__s64)(seq1 - seq2) < 0; 140 } 141 142 #define after64(seq2, seq1) before64(seq1, seq2) 143 144 struct mptcp_options_received { 145 u64 sndr_key; 146 u64 rcvr_key; 147 u64 data_ack; 148 u64 data_seq; 149 u32 subflow_seq; 150 u16 data_len; 151 __sum16 csum; 152 u16 suboptions; 153 u32 token; 154 u32 nonce; 155 u16 use_map:1, 156 dsn64:1, 157 data_fin:1, 158 use_ack:1, 159 ack64:1, 160 mpc_map:1, 161 reset_reason:4, 162 reset_transient:1, 163 echo:1, 164 backup:1, 165 deny_join_id0:1, 166 __unused:2; 167 u8 join_id; 168 u64 thmac; 169 u8 hmac[MPTCPOPT_HMAC_LEN]; 170 struct mptcp_addr_info addr; 171 struct mptcp_rm_list rm_list; 172 u64 ahmac; 173 u64 fail_seq; 174 }; 175 176 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 177 { 178 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 179 ((nib & 0xF) << 8) | field); 180 } 181 182 enum mptcp_pm_status { 183 MPTCP_PM_ADD_ADDR_RECEIVED, 184 MPTCP_PM_ADD_ADDR_SEND_ACK, 185 MPTCP_PM_RM_ADDR_RECEIVED, 186 MPTCP_PM_ESTABLISHED, 187 MPTCP_PM_SUBFLOW_ESTABLISHED, 188 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 189 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 190 * accounted int id_avail_bitmap 191 */ 192 }; 193 194 enum mptcp_pm_type { 195 MPTCP_PM_TYPE_KERNEL = 0, 196 MPTCP_PM_TYPE_USERSPACE, 197 198 __MPTCP_PM_TYPE_NR, 199 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 200 }; 201 202 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 203 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 204 205 enum mptcp_addr_signal_status { 206 MPTCP_ADD_ADDR_SIGNAL, 207 MPTCP_ADD_ADDR_ECHO, 208 MPTCP_RM_ADDR_SIGNAL, 209 }; 210 211 /* max value of mptcp_addr_info.id */ 212 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 213 214 struct mptcp_pm_data { 215 struct mptcp_addr_info local; 216 struct mptcp_addr_info remote; 217 struct list_head anno_list; 218 struct list_head userspace_pm_local_addr_list; 219 220 spinlock_t lock; /*protects the whole PM data */ 221 222 u8 addr_signal; 223 bool server_side; 224 bool work_pending; 225 bool accept_addr; 226 bool accept_subflow; 227 bool remote_deny_join_id0; 228 u8 add_addr_signaled; 229 u8 add_addr_accepted; 230 u8 local_addr_used; 231 u8 pm_type; 232 u8 subflows; 233 u8 status; 234 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 235 struct mptcp_rm_list rm_list_tx; 236 struct mptcp_rm_list rm_list_rx; 237 }; 238 239 struct mptcp_pm_local { 240 struct mptcp_addr_info addr; 241 u8 flags; 242 int ifindex; 243 }; 244 245 struct mptcp_pm_addr_entry { 246 struct list_head list; 247 struct mptcp_addr_info addr; 248 u8 flags; 249 int ifindex; 250 struct socket *lsk; 251 }; 252 253 struct mptcp_data_frag { 254 struct list_head list; 255 u64 data_seq; 256 u16 data_len; 257 u16 offset; 258 u16 overhead; 259 u16 already_sent; 260 struct page *page; 261 }; 262 263 /* MPTCP connection sock */ 264 struct mptcp_sock { 265 /* inet_connection_sock must be the first member */ 266 struct inet_connection_sock sk; 267 u64 local_key; /* protected by the first subflow socket lock 268 * lockless access read 269 */ 270 u64 remote_key; /* same as above */ 271 u64 write_seq; 272 u64 bytes_sent; 273 u64 snd_nxt; 274 u64 bytes_received; 275 u64 ack_seq; 276 atomic64_t rcv_wnd_sent; 277 u64 rcv_data_fin_seq; 278 u64 bytes_retrans; 279 u64 bytes_consumed; 280 int rmem_fwd_alloc; 281 int snd_burst; 282 int old_wspace; 283 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 284 * recovery related fields are under data_lock 285 * protection 286 */ 287 u64 bytes_acked; 288 u64 snd_una; 289 u64 wnd_end; 290 u32 last_data_sent; 291 u32 last_data_recv; 292 u32 last_ack_recv; 293 unsigned long timer_ival; 294 u32 token; 295 int rmem_released; 296 unsigned long flags; 297 unsigned long cb_flags; 298 bool recovery; /* closing subflow write queue reinjected */ 299 bool can_ack; 300 bool fully_established; 301 bool rcv_data_fin; 302 bool snd_data_fin_enable; 303 bool rcv_fastclose; 304 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 305 bool csum_enabled; 306 bool allow_infinite_fallback; 307 u8 pending_state; /* A subflow asked to set this sk_state, 308 * protected by the msk data lock 309 */ 310 u8 mpc_endpoint_id; 311 u8 recvmsg_inq:1, 312 cork:1, 313 nodelay:1, 314 fastopening:1, 315 in_accept_queue:1, 316 free_first:1, 317 rcvspace_init:1; 318 u32 notsent_lowat; 319 int keepalive_cnt; 320 int keepalive_idle; 321 int keepalive_intvl; 322 struct work_struct work; 323 struct sk_buff *ooo_last_skb; 324 struct rb_root out_of_order_queue; 325 struct sk_buff_head receive_queue; 326 struct list_head conn_list; 327 struct list_head rtx_queue; 328 struct mptcp_data_frag *first_pending; 329 struct list_head join_list; 330 struct sock *first; /* The mptcp ops can safely dereference, using suitable 331 * ONCE annotation, the subflow outside the socket 332 * lock as such sock is freed after close(). 333 */ 334 struct mptcp_pm_data pm; 335 struct mptcp_sched_ops *sched; 336 struct { 337 u32 space; /* bytes copied in last measurement window */ 338 u32 copied; /* bytes copied in this measurement window */ 339 u64 time; /* start time of measurement window */ 340 u64 rtt_us; /* last maximum rtt of subflows */ 341 } rcvq_space; 342 u8 scaling_ratio; 343 344 u32 subflow_id; 345 u32 setsockopt_seq; 346 char ca_name[TCP_CA_NAME_MAX]; 347 }; 348 349 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 350 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 351 352 #define mptcp_for_each_subflow(__msk, __subflow) \ 353 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 354 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 355 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 356 357 extern struct genl_family mptcp_genl_family; 358 359 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 360 { 361 sock_owned_by_me((const struct sock *)msk); 362 } 363 364 #ifdef CONFIG_DEBUG_NET 365 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 366 #undef tcp_sk 367 #define tcp_sk(ptr) ({ \ 368 typeof(ptr) _ptr = (ptr); \ 369 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 370 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 371 }) 372 #define mptcp_sk(ptr) ({ \ 373 typeof(ptr) _ptr = (ptr); \ 374 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 375 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 376 }) 377 378 #else /* !CONFIG_DEBUG_NET */ 379 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 380 #endif 381 382 /* the msk socket don't use the backlog, also account for the bulk 383 * free memory 384 */ 385 static inline int __mptcp_rmem(const struct sock *sk) 386 { 387 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 388 } 389 390 static inline int mptcp_win_from_space(const struct sock *sk, int space) 391 { 392 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 393 } 394 395 static inline int mptcp_space_from_win(const struct sock *sk, int win) 396 { 397 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win); 398 } 399 400 static inline int __mptcp_space(const struct sock *sk) 401 { 402 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 403 } 404 405 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 406 { 407 const struct mptcp_sock *msk = mptcp_sk(sk); 408 409 return READ_ONCE(msk->first_pending); 410 } 411 412 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 413 { 414 struct mptcp_sock *msk = mptcp_sk(sk); 415 struct mptcp_data_frag *cur; 416 417 cur = msk->first_pending; 418 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 419 list_next_entry(cur, list); 420 } 421 422 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 423 { 424 const struct mptcp_sock *msk = mptcp_sk(sk); 425 426 if (!msk->first_pending) 427 return NULL; 428 429 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 430 return NULL; 431 432 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 433 } 434 435 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 436 { 437 struct mptcp_sock *msk = mptcp_sk(sk); 438 439 if (msk->snd_una == msk->snd_nxt) 440 return NULL; 441 442 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 443 } 444 445 struct csum_pseudo_header { 446 __be64 data_seq; 447 __be32 subflow_seq; 448 __be16 data_len; 449 __sum16 csum; 450 }; 451 452 struct mptcp_subflow_request_sock { 453 struct tcp_request_sock sk; 454 u16 mp_capable : 1, 455 mp_join : 1, 456 backup : 1, 457 request_bkup : 1, 458 csum_reqd : 1, 459 allow_join_id0 : 1; 460 u8 local_id; 461 u8 remote_id; 462 u64 local_key; 463 u64 idsn; 464 u32 token; 465 u32 ssn_offset; 466 u64 thmac; 467 u32 local_nonce; 468 u32 remote_nonce; 469 struct mptcp_sock *msk; 470 struct hlist_nulls_node token_node; 471 }; 472 473 static inline struct mptcp_subflow_request_sock * 474 mptcp_subflow_rsk(const struct request_sock *rsk) 475 { 476 return (struct mptcp_subflow_request_sock *)rsk; 477 } 478 479 struct mptcp_delegated_action { 480 struct napi_struct napi; 481 struct list_head head; 482 }; 483 484 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 485 486 #define MPTCP_DELEGATE_SCHEDULED 0 487 #define MPTCP_DELEGATE_SEND 1 488 #define MPTCP_DELEGATE_ACK 2 489 #define MPTCP_DELEGATE_SNDBUF 3 490 491 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 492 /* MPTCP subflow context */ 493 struct mptcp_subflow_context { 494 struct list_head node;/* conn_list of subflows */ 495 496 struct_group(reset, 497 498 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 499 u64 local_key; 500 u64 remote_key; 501 u64 idsn; 502 u64 map_seq; 503 u32 snd_isn; 504 u32 token; 505 u32 rel_write_seq; 506 u32 map_subflow_seq; 507 u32 ssn_offset; 508 u32 map_data_len; 509 __wsum map_data_csum; 510 u32 map_csum_len; 511 u32 request_mptcp : 1, /* send MP_CAPABLE */ 512 request_join : 1, /* send MP_JOIN */ 513 request_bkup : 1, 514 mp_capable : 1, /* remote is MPTCP capable */ 515 mp_join : 1, /* remote is JOINing */ 516 fully_established : 1, /* path validated */ 517 pm_notified : 1, /* PM hook called for established status */ 518 conn_finished : 1, 519 map_valid : 1, 520 map_csum_reqd : 1, 521 map_data_fin : 1, 522 mpc_map : 1, 523 backup : 1, 524 send_mp_prio : 1, 525 send_mp_fail : 1, 526 send_fastclose : 1, 527 send_infinite_map : 1, 528 remote_key_valid : 1, /* received the peer key from */ 529 disposable : 1, /* ctx can be free at ulp release time */ 530 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 531 valid_csum_seen : 1, /* at least one csum validated */ 532 is_mptfo : 1, /* subflow is doing TFO */ 533 close_event_done : 1, /* has done the post-closed part */ 534 __unused : 9; 535 bool data_avail; 536 bool scheduled; 537 u32 remote_nonce; 538 u64 thmac; 539 u32 local_nonce; 540 u32 remote_token; 541 union { 542 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 543 u64 iasn; /* initial ack sequence number, MPC subflows only */ 544 }; 545 s16 local_id; /* if negative not initialized yet */ 546 u8 remote_id; 547 u8 reset_seen:1; 548 u8 reset_transient:1; 549 u8 reset_reason:4; 550 u8 stale_count; 551 552 u32 subflow_id; 553 554 long delegated_status; 555 unsigned long fail_tout; 556 557 ); 558 559 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 560 561 u32 setsockopt_seq; 562 u32 stale_rcv_tstamp; 563 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 564 * protected by the msk socket lock 565 */ 566 567 struct sock *tcp_sock; /* tcp sk backpointer */ 568 struct sock *conn; /* parent mptcp_sock */ 569 const struct inet_connection_sock_af_ops *icsk_af_ops; 570 void (*tcp_state_change)(struct sock *sk); 571 void (*tcp_error_report)(struct sock *sk); 572 573 struct rcu_head rcu; 574 }; 575 576 static inline struct mptcp_subflow_context * 577 mptcp_subflow_ctx(const struct sock *sk) 578 { 579 const struct inet_connection_sock *icsk = inet_csk(sk); 580 581 /* Use RCU on icsk_ulp_data only for sock diag code */ 582 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 583 } 584 585 static inline struct sock * 586 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 587 { 588 return subflow->tcp_sock; 589 } 590 591 static inline void 592 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 593 { 594 memset(&subflow->reset, 0, sizeof(subflow->reset)); 595 subflow->request_mptcp = 1; 596 WRITE_ONCE(subflow->local_id, -1); 597 } 598 599 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */ 600 static inline enum sk_rst_reason 601 sk_rst_convert_mptcp_reason(u32 reason) 602 { 603 switch (reason) { 604 case MPTCP_RST_EUNSPEC: 605 return SK_RST_REASON_MPTCP_RST_EUNSPEC; 606 case MPTCP_RST_EMPTCP: 607 return SK_RST_REASON_MPTCP_RST_EMPTCP; 608 case MPTCP_RST_ERESOURCE: 609 return SK_RST_REASON_MPTCP_RST_ERESOURCE; 610 case MPTCP_RST_EPROHIBIT: 611 return SK_RST_REASON_MPTCP_RST_EPROHIBIT; 612 case MPTCP_RST_EWQ2BIG: 613 return SK_RST_REASON_MPTCP_RST_EWQ2BIG; 614 case MPTCP_RST_EBADPERF: 615 return SK_RST_REASON_MPTCP_RST_EBADPERF; 616 case MPTCP_RST_EMIDDLEBOX: 617 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX; 618 default: 619 /* It should not happen, or else errors may occur 620 * in MPTCP layer 621 */ 622 return SK_RST_REASON_ERROR; 623 } 624 } 625 626 static inline void 627 mptcp_send_active_reset_reason(struct sock *sk) 628 { 629 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 630 enum sk_rst_reason reason; 631 632 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason); 633 tcp_send_active_reset(sk, GFP_ATOMIC, reason); 634 } 635 636 static inline u64 637 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 638 { 639 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 640 subflow->ssn_offset - 641 subflow->map_subflow_seq; 642 } 643 644 static inline u64 645 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 646 { 647 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 648 } 649 650 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 651 652 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 653 { 654 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 655 struct mptcp_delegated_action *delegated; 656 bool schedule; 657 658 /* the caller held the subflow bh socket lock */ 659 lockdep_assert_in_softirq(); 660 661 /* The implied barrier pairs with tcp_release_cb_override() 662 * mptcp_napi_poll(), and ensures the below list check sees list 663 * updates done prior to delegated status bits changes 664 */ 665 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 666 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 667 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 668 return; 669 670 delegated = this_cpu_ptr(&mptcp_delegated_actions); 671 schedule = list_empty(&delegated->head); 672 list_add_tail(&subflow->delegated_node, &delegated->head); 673 sock_hold(mptcp_subflow_tcp_sock(subflow)); 674 if (schedule) 675 napi_schedule(&delegated->napi); 676 } 677 } 678 679 static inline struct mptcp_subflow_context * 680 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 681 { 682 struct mptcp_subflow_context *ret; 683 684 if (list_empty(&delegated->head)) 685 return NULL; 686 687 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 688 list_del_init(&ret->delegated_node); 689 return ret; 690 } 691 692 int mptcp_is_enabled(const struct net *net); 693 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 694 int mptcp_is_checksum_enabled(const struct net *net); 695 int mptcp_allow_join_id0(const struct net *net); 696 unsigned int mptcp_stale_loss_cnt(const struct net *net); 697 unsigned int mptcp_close_timeout(const struct sock *sk); 698 int mptcp_get_pm_type(const struct net *net); 699 const char *mptcp_get_scheduler(const struct net *net); 700 void mptcp_get_available_schedulers(char *buf, size_t maxlen); 701 void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 702 struct mptcp_subflow_context *subflow, 703 const struct mptcp_options_received *mp_opt); 704 bool __mptcp_retransmit_pending_data(struct sock *sk); 705 void mptcp_check_and_set_pending(struct sock *sk); 706 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 707 bool mptcp_subflow_data_available(struct sock *sk); 708 void __init mptcp_subflow_init(void); 709 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 710 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 711 struct mptcp_subflow_context *subflow); 712 void __mptcp_subflow_send_ack(struct sock *ssk); 713 void mptcp_subflow_reset(struct sock *ssk); 714 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 715 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 716 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 717 bool __mptcp_close(struct sock *sk, long timeout); 718 void mptcp_cancel_work(struct sock *sk); 719 void __mptcp_unaccepted_force_close(struct sock *sk); 720 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk); 721 void mptcp_set_state(struct sock *sk, int state); 722 723 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 724 const struct mptcp_addr_info *b, bool use_port); 725 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr); 726 727 /* called with sk socket lock held */ 728 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local, 729 const struct mptcp_addr_info *remote); 730 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 731 struct socket **new_sock); 732 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 733 struct sockaddr_storage *addr, 734 unsigned short family); 735 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 736 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 737 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 738 void mptcp_sched_init(void); 739 int mptcp_init_sched(struct mptcp_sock *msk, 740 struct mptcp_sched_ops *sched); 741 void mptcp_release_sched(struct mptcp_sock *msk); 742 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 743 bool scheduled); 744 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 745 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 746 int mptcp_sched_get_send(struct mptcp_sock *msk); 747 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 748 749 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 750 { 751 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 752 } 753 754 static inline bool mptcp_epollin_ready(const struct sock *sk) 755 { 756 /* mptcp doesn't have to deal with small skbs in the receive queue, 757 * at it can always coalesce them 758 */ 759 return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) || 760 (mem_cgroup_sockets_enabled && sk->sk_memcg && 761 mem_cgroup_under_socket_pressure(sk->sk_memcg)) || 762 READ_ONCE(tcp_memory_pressure); 763 } 764 765 int mptcp_set_rcvlowat(struct sock *sk, int val); 766 767 static inline bool __tcp_can_send(const struct sock *ssk) 768 { 769 /* only send if our side has not closed yet */ 770 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 771 } 772 773 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 774 { 775 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 776 if (subflow->request_join && !subflow->fully_established) 777 return false; 778 779 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 780 } 781 782 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 783 784 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 785 786 void mptcp_subflow_drop_ctx(struct sock *ssk); 787 788 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 789 struct mptcp_subflow_context *ctx) 790 { 791 sk->sk_data_ready = sock_def_readable; 792 sk->sk_state_change = ctx->tcp_state_change; 793 sk->sk_write_space = sk_stream_write_space; 794 sk->sk_error_report = ctx->tcp_error_report; 795 796 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 797 } 798 799 void __init mptcp_proto_init(void); 800 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 801 int __init mptcp_proto_v6_init(void); 802 #endif 803 804 struct sock *mptcp_sk_clone_init(const struct sock *sk, 805 const struct mptcp_options_received *mp_opt, 806 struct sock *ssk, 807 struct request_sock *req); 808 void mptcp_get_options(const struct sk_buff *skb, 809 struct mptcp_options_received *mp_opt); 810 811 void mptcp_finish_connect(struct sock *sk); 812 void __mptcp_sync_state(struct sock *sk, int state); 813 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 814 815 static inline void mptcp_stop_tout_timer(struct sock *sk) 816 { 817 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 818 return; 819 820 sk_stop_timer(sk, &sk->sk_timer); 821 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 822 } 823 824 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 825 { 826 /* avoid 0 timestamp, as that means no close timeout */ 827 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 828 } 829 830 static inline void mptcp_start_tout_timer(struct sock *sk) 831 { 832 mptcp_set_close_tout(sk, tcp_jiffies32); 833 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 834 } 835 836 static inline bool mptcp_is_fully_established(struct sock *sk) 837 { 838 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 839 READ_ONCE(mptcp_sk(sk)->fully_established); 840 } 841 842 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 843 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 844 bool mptcp_finish_join(struct sock *sk); 845 bool mptcp_schedule_work(struct sock *sk); 846 int mptcp_setsockopt(struct sock *sk, int level, int optname, 847 sockptr_t optval, unsigned int optlen); 848 int mptcp_getsockopt(struct sock *sk, int level, int optname, 849 char __user *optval, int __user *option); 850 851 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 852 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 853 { 854 if (use_64bit) 855 return cur_seq; 856 857 return __mptcp_expand_seq(old_seq, cur_seq); 858 } 859 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 860 void __mptcp_data_acked(struct sock *sk); 861 void __mptcp_error_report(struct sock *sk); 862 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 863 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 864 { 865 return READ_ONCE(msk->snd_data_fin_enable) && 866 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 867 } 868 869 static inline u32 mptcp_notsent_lowat(const struct sock *sk) 870 { 871 struct net *net = sock_net(sk); 872 u32 val; 873 874 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat); 875 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat); 876 } 877 878 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake) 879 { 880 const struct mptcp_sock *msk = mptcp_sk(sk); 881 u32 notsent_bytes; 882 883 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt); 884 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk); 885 } 886 887 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake) 888 { 889 return mptcp_stream_memory_free(sk, wake) && 890 __sk_stream_is_writeable(sk, wake); 891 } 892 893 static inline void mptcp_write_space(struct sock *sk) 894 { 895 /* pairs with memory barrier in mptcp_poll */ 896 smp_mb(); 897 if (mptcp_stream_memory_free(sk, 1)) 898 sk_stream_write_space(sk); 899 } 900 901 static inline void __mptcp_sync_sndbuf(struct sock *sk) 902 { 903 struct mptcp_subflow_context *subflow; 904 int ssk_sndbuf, new_sndbuf; 905 906 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 907 return; 908 909 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]); 910 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 911 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 912 913 subflow->cached_sndbuf = ssk_sndbuf; 914 new_sndbuf += ssk_sndbuf; 915 } 916 917 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 918 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 919 mptcp_write_space(sk); 920 } 921 922 /* The called held both the msk socket and the subflow socket locks, 923 * possibly under BH 924 */ 925 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 926 { 927 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 928 929 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 930 __mptcp_sync_sndbuf(sk); 931 } 932 933 /* the caller held only the subflow socket lock, either in process or 934 * BH context. Additionally this can be called under the msk data lock, 935 * so we can't acquire such lock here: let the delegate action acquires 936 * the needed locks in suitable order. 937 */ 938 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 939 { 940 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 941 942 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 943 return; 944 945 local_bh_disable(); 946 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 947 local_bh_enable(); 948 } 949 950 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 951 952 #define MPTCP_TOKEN_MAX_RETRIES 4 953 954 void __init mptcp_token_init(void); 955 static inline void mptcp_token_init_request(struct request_sock *req) 956 { 957 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 958 } 959 960 int mptcp_token_new_request(struct request_sock *req); 961 void mptcp_token_destroy_request(struct request_sock *req); 962 int mptcp_token_new_connect(struct sock *ssk); 963 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 964 struct mptcp_sock *msk); 965 bool mptcp_token_exists(u32 token); 966 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 967 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 968 long *s_num); 969 void mptcp_token_destroy(struct mptcp_sock *msk); 970 971 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 972 973 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 974 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 975 976 void __init mptcp_pm_init(void); 977 void mptcp_pm_data_init(struct mptcp_sock *msk); 978 void mptcp_pm_data_reset(struct mptcp_sock *msk); 979 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 980 struct mptcp_addr_info *addr); 981 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 982 bool require_family, 983 struct mptcp_pm_addr_entry *entry); 984 bool mptcp_pm_addr_families_match(const struct sock *sk, 985 const struct mptcp_addr_info *loc, 986 const struct mptcp_addr_info *rem); 987 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 988 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 989 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 990 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 991 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 992 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 993 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 994 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 995 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 996 const struct mptcp_subflow_context *subflow); 997 void mptcp_pm_add_addr_received(const struct sock *ssk, 998 const struct mptcp_addr_info *addr); 999 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 1000 const struct mptcp_addr_info *addr); 1001 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 1002 bool mptcp_pm_nl_is_init_remote_addr(struct mptcp_sock *msk, 1003 const struct mptcp_addr_info *remote); 1004 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 1005 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 1006 const struct mptcp_rm_list *rm_list); 1007 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 1008 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 1009 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 1010 struct mptcp_addr_info *addr, 1011 struct mptcp_addr_info *rem, 1012 u8 bkup); 1013 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 1014 const struct mptcp_addr_info *addr); 1015 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 1016 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 1017 struct mptcp_pm_add_entry * 1018 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 1019 const struct mptcp_addr_info *addr, bool check_id); 1020 struct mptcp_pm_add_entry * 1021 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 1022 const struct mptcp_addr_info *addr); 1023 int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info); 1024 int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info); 1025 int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info); 1026 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 1027 const struct mptcp_addr_info *addr, 1028 bool echo); 1029 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 1030 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list); 1031 1032 void mptcp_free_local_addr_list(struct mptcp_sock *msk); 1033 1034 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 1035 const struct sock *ssk, gfp_t gfp); 1036 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 1037 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 1038 void mptcp_event_pm_listener(const struct sock *ssk, 1039 enum mptcp_event_type event); 1040 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 1041 1042 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow, 1043 const struct mptcp_options_received *mp_opt); 1044 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 1045 struct request_sock *req); 1046 int mptcp_nl_fill_addr(struct sk_buff *skb, 1047 struct mptcp_pm_addr_entry *entry); 1048 1049 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 1050 { 1051 return READ_ONCE(msk->pm.addr_signal) & 1052 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 1053 } 1054 1055 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 1056 { 1057 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 1058 } 1059 1060 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 1061 { 1062 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 1063 } 1064 1065 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 1066 { 1067 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 1068 } 1069 1070 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1071 { 1072 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1073 } 1074 1075 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1076 { 1077 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1078 } 1079 1080 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 1081 { 1082 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 1083 1084 if (family == AF_INET6) 1085 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 1086 if (!echo) 1087 len += MPTCPOPT_THMAC_LEN; 1088 /* account for 2 trailing 'nop' options */ 1089 if (port) 1090 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 1091 1092 return len; 1093 } 1094 1095 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 1096 { 1097 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 1098 return -EINVAL; 1099 1100 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 1101 } 1102 1103 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 1104 unsigned int opt_size, unsigned int remaining, 1105 struct mptcp_addr_info *addr, bool *echo, 1106 bool *drop_other_suboptions); 1107 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1108 struct mptcp_rm_list *rm_list); 1109 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1110 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1111 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1112 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc); 1113 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1114 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1115 int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb); 1116 int mptcp_pm_nl_dump_addr(struct sk_buff *msg, 1117 struct netlink_callback *cb); 1118 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg, 1119 struct netlink_callback *cb); 1120 int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info); 1121 int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info); 1122 int mptcp_userspace_pm_get_addr(struct sk_buff *skb, 1123 struct genl_info *info); 1124 1125 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1126 { 1127 int local_id = READ_ONCE(subflow->local_id); 1128 1129 if (local_id < 0) 1130 return 0; 1131 return local_id; 1132 } 1133 1134 void __init mptcp_pm_nl_init(void); 1135 void mptcp_pm_nl_work(struct mptcp_sock *msk); 1136 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 1137 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 1138 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 1139 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 1140 1141 /* called under PM lock */ 1142 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1143 { 1144 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 1145 WRITE_ONCE(msk->pm.accept_subflow, true); 1146 } 1147 1148 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1149 { 1150 spin_lock_bh(&msk->pm.lock); 1151 __mptcp_pm_close_subflow(msk); 1152 spin_unlock_bh(&msk->pm.lock); 1153 } 1154 1155 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1156 1157 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1158 { 1159 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1160 } 1161 1162 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1163 1164 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1165 { 1166 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1167 } 1168 1169 static inline bool mptcp_check_fallback(const struct sock *sk) 1170 { 1171 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1172 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1173 1174 return __mptcp_check_fallback(msk); 1175 } 1176 1177 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 1178 { 1179 if (__mptcp_check_fallback(msk)) { 1180 pr_debug("TCP fallback already done (msk=%p)\n", msk); 1181 return; 1182 } 1183 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1184 } 1185 1186 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1187 { 1188 struct sock *ssk = READ_ONCE(msk->first); 1189 1190 return ssk && ((1 << inet_sk_state_load(ssk)) & 1191 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1192 TCPF_SYN_RECV | TCPF_LISTEN)); 1193 } 1194 1195 static inline void mptcp_do_fallback(struct sock *ssk) 1196 { 1197 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1198 struct sock *sk = subflow->conn; 1199 struct mptcp_sock *msk; 1200 1201 msk = mptcp_sk(sk); 1202 __mptcp_do_fallback(msk); 1203 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1204 gfp_t saved_allocation = ssk->sk_allocation; 1205 1206 /* we are in a atomic (BH) scope, override ssk default for data 1207 * fin allocation 1208 */ 1209 ssk->sk_allocation = GFP_ATOMIC; 1210 ssk->sk_shutdown |= SEND_SHUTDOWN; 1211 tcp_shutdown(ssk, SEND_SHUTDOWN); 1212 ssk->sk_allocation = saved_allocation; 1213 } 1214 } 1215 1216 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)\n", __func__, a) 1217 1218 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1219 { 1220 struct mptcp_ext *mpext; 1221 1222 mpext = skb ? mptcp_get_ext(skb) : NULL; 1223 if (mpext && mpext->infinite_map) 1224 return true; 1225 1226 return false; 1227 } 1228 1229 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1230 { 1231 return (subflow->request_mptcp || subflow->request_join); 1232 } 1233 1234 static inline bool subflow_simultaneous_connect(struct sock *sk) 1235 { 1236 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1237 1238 return (1 << sk->sk_state) & 1239 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) && 1240 is_active_ssk(subflow) && 1241 !subflow->conn_finished; 1242 } 1243 1244 #ifdef CONFIG_SYN_COOKIES 1245 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1246 struct sk_buff *skb); 1247 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1248 struct sk_buff *skb); 1249 void __init mptcp_join_cookie_init(void); 1250 #else 1251 static inline void 1252 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1253 struct sk_buff *skb) {} 1254 static inline bool 1255 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1256 struct sk_buff *skb) 1257 { 1258 return false; 1259 } 1260 1261 static inline void mptcp_join_cookie_init(void) {} 1262 #endif 1263 1264 #endif /* __MPTCP_PROTOCOL_H */ 1265