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 #define MPTCP_DEQUEUE 8 128 129 struct mptcp_skb_cb { 130 u64 map_seq; 131 u64 end_seq; 132 u32 offset; 133 u8 has_rxtstamp; 134 u8 cant_coalesce; 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 struct_group(status, 155 u16 suboptions; 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 ); 169 u8 join_id; 170 u32 token; 171 u32 nonce; 172 u64 thmac; 173 u8 hmac[MPTCPOPT_HMAC_LEN]; 174 struct mptcp_addr_info addr; 175 struct mptcp_rm_list rm_list; 176 u64 ahmac; 177 u64 fail_seq; 178 }; 179 180 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 181 { 182 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 183 ((nib & 0xF) << 8) | field); 184 } 185 186 enum mptcp_pm_status { 187 MPTCP_PM_ADD_ADDR_RECEIVED, 188 MPTCP_PM_ADD_ADDR_SEND_ACK, 189 MPTCP_PM_RM_ADDR_RECEIVED, 190 MPTCP_PM_ESTABLISHED, 191 MPTCP_PM_SUBFLOW_ESTABLISHED, 192 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 193 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 194 * accounted int id_avail_bitmap 195 */ 196 }; 197 198 enum mptcp_pm_type { 199 MPTCP_PM_TYPE_KERNEL = 0, 200 MPTCP_PM_TYPE_USERSPACE, 201 202 __MPTCP_PM_TYPE_NR, 203 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 204 }; 205 206 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 207 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 208 209 enum mptcp_addr_signal_status { 210 MPTCP_ADD_ADDR_SIGNAL, 211 MPTCP_ADD_ADDR_ECHO, 212 MPTCP_RM_ADDR_SIGNAL, 213 }; 214 215 /* max value of mptcp_addr_info.id */ 216 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 217 218 struct mptcp_pm_data { 219 struct mptcp_addr_info local; 220 struct mptcp_addr_info remote; 221 struct list_head anno_list; 222 struct list_head userspace_pm_local_addr_list; 223 224 spinlock_t lock; /*protects the whole PM data */ 225 226 u8 addr_signal; 227 bool server_side; 228 bool work_pending; 229 bool accept_addr; 230 bool accept_subflow; 231 bool remote_deny_join_id0; 232 u8 add_addr_signaled; 233 u8 add_addr_accepted; 234 u8 local_addr_used; 235 u8 pm_type; 236 u8 subflows; 237 u8 status; 238 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 239 struct mptcp_rm_list rm_list_tx; 240 struct mptcp_rm_list rm_list_rx; 241 }; 242 243 struct mptcp_pm_local { 244 struct mptcp_addr_info addr; 245 u8 flags; 246 int ifindex; 247 }; 248 249 struct mptcp_pm_addr_entry { 250 struct list_head list; 251 struct mptcp_addr_info addr; 252 u8 flags; 253 int ifindex; 254 struct socket *lsk; 255 }; 256 257 struct mptcp_data_frag { 258 struct list_head list; 259 u64 data_seq; 260 u16 data_len; 261 u16 offset; 262 u16 overhead; 263 u16 already_sent; 264 struct page *page; 265 }; 266 267 /* MPTCP connection sock */ 268 struct mptcp_sock { 269 /* inet_connection_sock must be the first member */ 270 struct inet_connection_sock sk; 271 u64 local_key; /* protected by the first subflow socket lock 272 * lockless access read 273 */ 274 u64 remote_key; /* same as above */ 275 u64 write_seq; 276 u64 bytes_sent; 277 u64 snd_nxt; 278 u64 bytes_received; 279 u64 ack_seq; 280 atomic64_t rcv_wnd_sent; 281 u64 rcv_data_fin_seq; 282 u64 bytes_retrans; 283 u64 bytes_consumed; 284 int snd_burst; 285 int old_wspace; 286 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 287 * recovery related fields are under data_lock 288 * protection 289 */ 290 u64 bytes_acked; 291 u64 snd_una; 292 u64 wnd_end; 293 u32 last_data_sent; 294 u32 last_data_recv; 295 u32 last_ack_recv; 296 unsigned long timer_ival; 297 u32 token; 298 unsigned long flags; 299 unsigned long cb_flags; 300 bool recovery; /* closing subflow write queue reinjected */ 301 bool can_ack; 302 bool fully_established; 303 bool rcv_data_fin; 304 bool snd_data_fin_enable; 305 bool rcv_fastclose; 306 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 307 bool csum_enabled; 308 bool allow_infinite_fallback; 309 u8 pending_state; /* A subflow asked to set this sk_state, 310 * protected by the msk data lock 311 */ 312 u8 mpc_endpoint_id; 313 u8 recvmsg_inq:1, 314 cork:1, 315 nodelay:1, 316 fastopening:1, 317 in_accept_queue:1, 318 free_first:1, 319 rcvspace_init:1; 320 u32 notsent_lowat; 321 int keepalive_cnt; 322 int keepalive_idle; 323 int keepalive_intvl; 324 struct work_struct work; 325 struct sk_buff *ooo_last_skb; 326 struct rb_root out_of_order_queue; 327 struct list_head conn_list; 328 struct list_head rtx_queue; 329 struct mptcp_data_frag *first_pending; 330 struct list_head join_list; 331 struct sock *first; /* The mptcp ops can safely dereference, using suitable 332 * ONCE annotation, the subflow outside the socket 333 * lock as such sock is freed after close(). 334 */ 335 struct mptcp_pm_data pm; 336 struct mptcp_sched_ops *sched; 337 struct { 338 u32 space; /* bytes copied in last measurement window */ 339 u32 copied; /* bytes copied in this measurement window */ 340 u64 time; /* start time of measurement window */ 341 u64 rtt_us; /* last maximum rtt of subflows */ 342 } rcvq_space; 343 u8 scaling_ratio; 344 345 u32 subflow_id; 346 u32 setsockopt_seq; 347 char ca_name[TCP_CA_NAME_MAX]; 348 }; 349 350 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 351 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 352 353 #define mptcp_for_each_subflow(__msk, __subflow) \ 354 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 355 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 356 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 357 #define mptcp_next_subflow(__msk, __subflow) \ 358 list_next_entry_circular(__subflow, &((__msk)->conn_list), node) 359 360 extern struct genl_family mptcp_genl_family; 361 362 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 363 { 364 sock_owned_by_me((const struct sock *)msk); 365 } 366 367 #ifdef CONFIG_DEBUG_NET 368 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 369 #undef tcp_sk 370 #define tcp_sk(ptr) ({ \ 371 typeof(ptr) _ptr = (ptr); \ 372 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 373 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 374 }) 375 #define mptcp_sk(ptr) ({ \ 376 typeof(ptr) _ptr = (ptr); \ 377 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 378 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 379 }) 380 381 #else /* !CONFIG_DEBUG_NET */ 382 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 383 #endif 384 385 static inline int mptcp_win_from_space(const struct sock *sk, int space) 386 { 387 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 388 } 389 390 static inline int mptcp_space_from_win(const struct sock *sk, int win) 391 { 392 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win); 393 } 394 395 static inline int __mptcp_space(const struct sock *sk) 396 { 397 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - 398 sk_rmem_alloc_get(sk)); 399 } 400 401 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 402 { 403 const struct mptcp_sock *msk = mptcp_sk(sk); 404 405 return READ_ONCE(msk->first_pending); 406 } 407 408 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 409 { 410 struct mptcp_sock *msk = mptcp_sk(sk); 411 struct mptcp_data_frag *cur; 412 413 cur = msk->first_pending; 414 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 415 list_next_entry(cur, list); 416 } 417 418 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 419 { 420 const struct mptcp_sock *msk = mptcp_sk(sk); 421 422 if (!msk->first_pending) 423 return NULL; 424 425 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 426 return NULL; 427 428 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 429 } 430 431 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 432 { 433 struct mptcp_sock *msk = mptcp_sk(sk); 434 435 if (msk->snd_una == msk->snd_nxt) 436 return NULL; 437 438 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 439 } 440 441 struct csum_pseudo_header { 442 __be64 data_seq; 443 __be32 subflow_seq; 444 __be16 data_len; 445 __sum16 csum; 446 }; 447 448 struct mptcp_subflow_request_sock { 449 struct tcp_request_sock sk; 450 u16 mp_capable : 1, 451 mp_join : 1, 452 backup : 1, 453 request_bkup : 1, 454 csum_reqd : 1, 455 allow_join_id0 : 1; 456 u8 local_id; 457 u8 remote_id; 458 u64 local_key; 459 u64 idsn; 460 u32 token; 461 u32 ssn_offset; 462 u64 thmac; 463 u32 local_nonce; 464 u32 remote_nonce; 465 struct mptcp_sock *msk; 466 struct hlist_nulls_node token_node; 467 }; 468 469 static inline struct mptcp_subflow_request_sock * 470 mptcp_subflow_rsk(const struct request_sock *rsk) 471 { 472 return (struct mptcp_subflow_request_sock *)rsk; 473 } 474 475 struct mptcp_delegated_action { 476 struct napi_struct napi; 477 struct list_head head; 478 }; 479 480 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 481 482 #define MPTCP_DELEGATE_SCHEDULED 0 483 #define MPTCP_DELEGATE_SEND 1 484 #define MPTCP_DELEGATE_ACK 2 485 #define MPTCP_DELEGATE_SNDBUF 3 486 487 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 488 /* MPTCP subflow context */ 489 struct mptcp_subflow_context { 490 struct list_head node;/* conn_list of subflows */ 491 492 struct_group(reset, 493 494 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 495 u64 local_key; 496 u64 remote_key; 497 u64 idsn; 498 u64 map_seq; 499 u32 snd_isn; 500 u32 token; 501 u32 rel_write_seq; 502 u32 map_subflow_seq; 503 u32 ssn_offset; 504 u32 map_data_len; 505 __wsum map_data_csum; 506 u32 map_csum_len; 507 u32 request_mptcp : 1, /* send MP_CAPABLE */ 508 request_join : 1, /* send MP_JOIN */ 509 request_bkup : 1, 510 mp_capable : 1, /* remote is MPTCP capable */ 511 mp_join : 1, /* remote is JOINing */ 512 pm_notified : 1, /* PM hook called for established status */ 513 conn_finished : 1, 514 map_valid : 1, 515 map_csum_reqd : 1, 516 map_data_fin : 1, 517 mpc_map : 1, 518 backup : 1, 519 send_mp_prio : 1, 520 send_mp_fail : 1, 521 send_fastclose : 1, 522 send_infinite_map : 1, 523 remote_key_valid : 1, /* received the peer key from */ 524 disposable : 1, /* ctx can be free at ulp release time */ 525 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 526 valid_csum_seen : 1, /* at least one csum validated */ 527 is_mptfo : 1, /* subflow is doing TFO */ 528 close_event_done : 1, /* has done the post-closed part */ 529 mpc_drop : 1, /* the MPC option has been dropped in a rtx */ 530 __unused : 9; 531 bool data_avail; 532 bool scheduled; 533 bool pm_listener; /* a listener managed by the kernel PM? */ 534 bool fully_established; /* path validated */ 535 u32 remote_nonce; 536 u64 thmac; 537 u32 local_nonce; 538 u32 remote_token; 539 union { 540 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 541 u64 iasn; /* initial ack sequence number, MPC subflows only */ 542 }; 543 s16 local_id; /* if negative not initialized yet */ 544 u8 remote_id; 545 u8 reset_seen:1; 546 u8 reset_transient:1; 547 u8 reset_reason:4; 548 u8 stale_count; 549 550 u32 subflow_id; 551 552 long delegated_status; 553 unsigned long fail_tout; 554 555 ); 556 557 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 558 559 u32 setsockopt_seq; 560 u32 stale_rcv_tstamp; 561 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 562 * protected by the msk socket lock 563 */ 564 565 struct sock *tcp_sock; /* tcp sk backpointer */ 566 struct sock *conn; /* parent mptcp_sock */ 567 const struct inet_connection_sock_af_ops *icsk_af_ops; 568 void (*tcp_state_change)(struct sock *sk); 569 void (*tcp_error_report)(struct sock *sk); 570 571 struct rcu_head rcu; 572 }; 573 574 static inline struct mptcp_subflow_context * 575 mptcp_subflow_ctx(const struct sock *sk) 576 { 577 const struct inet_connection_sock *icsk = inet_csk(sk); 578 579 /* Use RCU on icsk_ulp_data only for sock diag code */ 580 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 581 } 582 583 static inline struct sock * 584 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 585 { 586 return subflow->tcp_sock; 587 } 588 589 static inline void 590 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 591 { 592 memset(&subflow->reset, 0, sizeof(subflow->reset)); 593 subflow->request_mptcp = 1; 594 WRITE_ONCE(subflow->local_id, -1); 595 } 596 597 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */ 598 static inline enum sk_rst_reason 599 sk_rst_convert_mptcp_reason(u32 reason) 600 { 601 switch (reason) { 602 case MPTCP_RST_EUNSPEC: 603 return SK_RST_REASON_MPTCP_RST_EUNSPEC; 604 case MPTCP_RST_EMPTCP: 605 return SK_RST_REASON_MPTCP_RST_EMPTCP; 606 case MPTCP_RST_ERESOURCE: 607 return SK_RST_REASON_MPTCP_RST_ERESOURCE; 608 case MPTCP_RST_EPROHIBIT: 609 return SK_RST_REASON_MPTCP_RST_EPROHIBIT; 610 case MPTCP_RST_EWQ2BIG: 611 return SK_RST_REASON_MPTCP_RST_EWQ2BIG; 612 case MPTCP_RST_EBADPERF: 613 return SK_RST_REASON_MPTCP_RST_EBADPERF; 614 case MPTCP_RST_EMIDDLEBOX: 615 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX; 616 default: 617 /* It should not happen, or else errors may occur 618 * in MPTCP layer 619 */ 620 return SK_RST_REASON_ERROR; 621 } 622 } 623 624 static inline void 625 mptcp_send_active_reset_reason(struct sock *sk) 626 { 627 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 628 enum sk_rst_reason reason; 629 630 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason); 631 tcp_send_active_reset(sk, GFP_ATOMIC, reason); 632 } 633 634 static inline u64 635 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 636 { 637 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 638 subflow->ssn_offset - 639 subflow->map_subflow_seq; 640 } 641 642 static inline u64 643 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 644 { 645 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 646 } 647 648 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 649 650 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 651 { 652 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 653 struct mptcp_delegated_action *delegated; 654 bool schedule; 655 656 /* the caller held the subflow bh socket lock */ 657 lockdep_assert_in_softirq(); 658 659 /* The implied barrier pairs with tcp_release_cb_override() 660 * mptcp_napi_poll(), and ensures the below list check sees list 661 * updates done prior to delegated status bits changes 662 */ 663 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 664 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 665 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 666 return; 667 668 delegated = this_cpu_ptr(&mptcp_delegated_actions); 669 schedule = list_empty(&delegated->head); 670 list_add_tail(&subflow->delegated_node, &delegated->head); 671 sock_hold(mptcp_subflow_tcp_sock(subflow)); 672 if (schedule) 673 napi_schedule(&delegated->napi); 674 } 675 } 676 677 static inline struct mptcp_subflow_context * 678 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 679 { 680 struct mptcp_subflow_context *ret; 681 682 if (list_empty(&delegated->head)) 683 return NULL; 684 685 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 686 list_del_init(&ret->delegated_node); 687 return ret; 688 } 689 690 int mptcp_is_enabled(const struct net *net); 691 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 692 int mptcp_is_checksum_enabled(const struct net *net); 693 int mptcp_allow_join_id0(const struct net *net); 694 unsigned int mptcp_stale_loss_cnt(const struct net *net); 695 unsigned int mptcp_close_timeout(const struct sock *sk); 696 int mptcp_get_pm_type(const struct net *net); 697 const char *mptcp_get_scheduler(const struct net *net); 698 699 void mptcp_active_disable(struct sock *sk); 700 bool mptcp_active_should_disable(struct sock *ssk); 701 void mptcp_active_enable(struct sock *sk); 702 703 void mptcp_get_available_schedulers(char *buf, size_t maxlen); 704 void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 705 struct mptcp_subflow_context *subflow, 706 const struct mptcp_options_received *mp_opt); 707 bool __mptcp_retransmit_pending_data(struct sock *sk); 708 void mptcp_check_and_set_pending(struct sock *sk); 709 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 710 bool mptcp_subflow_data_available(struct sock *sk); 711 void __init mptcp_subflow_init(void); 712 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 713 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 714 struct mptcp_subflow_context *subflow); 715 void __mptcp_subflow_send_ack(struct sock *ssk); 716 void mptcp_subflow_reset(struct sock *ssk); 717 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 718 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 719 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 720 bool __mptcp_close(struct sock *sk, long timeout); 721 void mptcp_cancel_work(struct sock *sk); 722 void __mptcp_unaccepted_force_close(struct sock *sk); 723 void mptcp_set_state(struct sock *sk, int state); 724 725 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 726 const struct mptcp_addr_info *b, bool use_port); 727 void mptcp_local_address(const struct sock_common *skc, 728 struct mptcp_addr_info *addr); 729 void mptcp_remote_address(const struct sock_common *skc, 730 struct mptcp_addr_info *addr); 731 732 /* called with sk socket lock held */ 733 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local, 734 const struct mptcp_addr_info *remote); 735 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 736 struct socket **new_sock); 737 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 738 struct sockaddr_storage *addr, 739 unsigned short family); 740 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 741 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 742 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 743 void mptcp_sched_init(void); 744 int mptcp_init_sched(struct mptcp_sock *msk, 745 struct mptcp_sched_ops *sched); 746 void mptcp_release_sched(struct mptcp_sock *msk); 747 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 748 bool scheduled); 749 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 750 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 751 int mptcp_sched_get_send(struct mptcp_sock *msk); 752 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 753 754 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 755 { 756 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 757 } 758 759 static inline bool mptcp_epollin_ready(const struct sock *sk) 760 { 761 u64 data_avail = mptcp_data_avail(mptcp_sk(sk)); 762 763 if (!data_avail) 764 return false; 765 766 /* mptcp doesn't have to deal with small skbs in the receive queue, 767 * as it can always coalesce them 768 */ 769 return (data_avail >= sk->sk_rcvlowat) || 770 (mem_cgroup_sockets_enabled && sk->sk_memcg && 771 mem_cgroup_under_socket_pressure(sk->sk_memcg)) || 772 READ_ONCE(tcp_memory_pressure); 773 } 774 775 int mptcp_set_rcvlowat(struct sock *sk, int val); 776 777 static inline bool __tcp_can_send(const struct sock *ssk) 778 { 779 /* only send if our side has not closed yet */ 780 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 781 } 782 783 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 784 { 785 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 786 if (subflow->request_join && !READ_ONCE(subflow->fully_established)) 787 return false; 788 789 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 790 } 791 792 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 793 794 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 795 796 void mptcp_subflow_drop_ctx(struct sock *ssk); 797 798 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 799 struct mptcp_subflow_context *ctx) 800 { 801 sk->sk_data_ready = sock_def_readable; 802 sk->sk_state_change = ctx->tcp_state_change; 803 sk->sk_write_space = sk_stream_write_space; 804 sk->sk_error_report = ctx->tcp_error_report; 805 806 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 807 } 808 809 void __init mptcp_proto_init(void); 810 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 811 int __init mptcp_proto_v6_init(void); 812 #endif 813 814 struct sock *mptcp_sk_clone_init(const struct sock *sk, 815 const struct mptcp_options_received *mp_opt, 816 struct sock *ssk, 817 struct request_sock *req); 818 void mptcp_get_options(const struct sk_buff *skb, 819 struct mptcp_options_received *mp_opt); 820 821 void mptcp_finish_connect(struct sock *sk); 822 void __mptcp_sync_state(struct sock *sk, int state); 823 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 824 825 static inline void mptcp_stop_tout_timer(struct sock *sk) 826 { 827 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 828 return; 829 830 sk_stop_timer(sk, &sk->sk_timer); 831 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 832 } 833 834 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 835 { 836 /* avoid 0 timestamp, as that means no close timeout */ 837 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 838 } 839 840 static inline void mptcp_start_tout_timer(struct sock *sk) 841 { 842 mptcp_set_close_tout(sk, tcp_jiffies32); 843 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 844 } 845 846 static inline bool mptcp_is_fully_established(struct sock *sk) 847 { 848 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 849 READ_ONCE(mptcp_sk(sk)->fully_established); 850 } 851 852 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 853 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 854 bool mptcp_finish_join(struct sock *sk); 855 bool mptcp_schedule_work(struct sock *sk); 856 int mptcp_setsockopt(struct sock *sk, int level, int optname, 857 sockptr_t optval, unsigned int optlen); 858 int mptcp_getsockopt(struct sock *sk, int level, int optname, 859 char __user *optval, int __user *option); 860 861 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 862 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 863 { 864 if (use_64bit) 865 return cur_seq; 866 867 return __mptcp_expand_seq(old_seq, cur_seq); 868 } 869 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 870 void __mptcp_data_acked(struct sock *sk); 871 void __mptcp_error_report(struct sock *sk); 872 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 873 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 874 { 875 return READ_ONCE(msk->snd_data_fin_enable) && 876 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 877 } 878 879 static inline u32 mptcp_notsent_lowat(const struct sock *sk) 880 { 881 struct net *net = sock_net(sk); 882 u32 val; 883 884 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat); 885 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat); 886 } 887 888 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake) 889 { 890 const struct mptcp_sock *msk = mptcp_sk(sk); 891 u32 notsent_bytes; 892 893 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt); 894 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk); 895 } 896 897 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake) 898 { 899 return mptcp_stream_memory_free(sk, wake) && 900 __sk_stream_is_writeable(sk, wake); 901 } 902 903 static inline void mptcp_write_space(struct sock *sk) 904 { 905 /* pairs with memory barrier in mptcp_poll */ 906 smp_mb(); 907 if (mptcp_stream_memory_free(sk, 1)) 908 sk_stream_write_space(sk); 909 } 910 911 static inline void __mptcp_sync_sndbuf(struct sock *sk) 912 { 913 struct mptcp_subflow_context *subflow; 914 int ssk_sndbuf, new_sndbuf; 915 916 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 917 return; 918 919 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]); 920 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 921 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 922 923 subflow->cached_sndbuf = ssk_sndbuf; 924 new_sndbuf += ssk_sndbuf; 925 } 926 927 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 928 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 929 mptcp_write_space(sk); 930 } 931 932 /* The called held both the msk socket and the subflow socket locks, 933 * possibly under BH 934 */ 935 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 936 { 937 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 938 939 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 940 __mptcp_sync_sndbuf(sk); 941 } 942 943 /* the caller held only the subflow socket lock, either in process or 944 * BH context. Additionally this can be called under the msk data lock, 945 * so we can't acquire such lock here: let the delegate action acquires 946 * the needed locks in suitable order. 947 */ 948 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 949 { 950 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 951 952 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 953 return; 954 955 local_bh_disable(); 956 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 957 local_bh_enable(); 958 } 959 960 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 961 962 #define MPTCP_TOKEN_MAX_RETRIES 4 963 964 void __init mptcp_token_init(void); 965 static inline void mptcp_token_init_request(struct request_sock *req) 966 { 967 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 968 } 969 970 int mptcp_token_new_request(struct request_sock *req); 971 void mptcp_token_destroy_request(struct request_sock *req); 972 int mptcp_token_new_connect(struct sock *ssk); 973 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 974 struct mptcp_sock *msk); 975 bool mptcp_token_exists(u32 token); 976 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 977 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 978 long *s_num); 979 void mptcp_token_destroy(struct mptcp_sock *msk); 980 981 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 982 983 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 984 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 985 986 void __init mptcp_pm_init(void); 987 void mptcp_pm_data_init(struct mptcp_sock *msk); 988 void mptcp_pm_data_reset(struct mptcp_sock *msk); 989 void mptcp_pm_destroy(struct mptcp_sock *msk); 990 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 991 struct mptcp_addr_info *addr); 992 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 993 bool require_family, 994 struct mptcp_pm_addr_entry *entry); 995 bool mptcp_pm_addr_families_match(const struct sock *sk, 996 const struct mptcp_addr_info *loc, 997 const struct mptcp_addr_info *rem); 998 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 999 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 1000 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 1001 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 1002 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 1003 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 1004 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 1005 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 1006 const struct mptcp_subflow_context *subflow); 1007 void mptcp_pm_add_addr_received(const struct sock *ssk, 1008 const struct mptcp_addr_info *addr); 1009 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 1010 const struct mptcp_addr_info *addr); 1011 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 1012 void mptcp_pm_send_ack(struct mptcp_sock *msk, 1013 struct mptcp_subflow_context *subflow, 1014 bool prio, bool backup); 1015 void mptcp_pm_addr_send_ack(struct mptcp_sock *msk); 1016 void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id); 1017 void mptcp_pm_rm_subflow(struct mptcp_sock *msk, 1018 const struct mptcp_rm_list *rm_list); 1019 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 1020 const struct mptcp_rm_list *rm_list); 1021 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 1022 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 1023 int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk, 1024 struct mptcp_addr_info *addr, 1025 struct mptcp_addr_info *rem, 1026 u8 bkup); 1027 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 1028 const struct mptcp_addr_info *addr); 1029 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 1030 struct mptcp_pm_add_entry * 1031 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 1032 const struct mptcp_addr_info *addr, bool check_id); 1033 bool mptcp_lookup_subflow_by_saddr(const struct list_head *list, 1034 const struct mptcp_addr_info *saddr); 1035 bool mptcp_remove_anno_list_by_saddr(struct mptcp_sock *msk, 1036 const struct mptcp_addr_info *addr); 1037 int mptcp_pm_nl_set_flags(struct mptcp_pm_addr_entry *local, 1038 struct genl_info *info); 1039 int mptcp_userspace_pm_set_flags(struct mptcp_pm_addr_entry *local, 1040 struct genl_info *info); 1041 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 1042 const struct mptcp_addr_info *addr, 1043 bool echo); 1044 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 1045 void mptcp_pm_remove_addr_entry(struct mptcp_sock *msk, 1046 struct mptcp_pm_addr_entry *entry); 1047 1048 void mptcp_userspace_pm_free_local_addr_list(struct mptcp_sock *msk); 1049 1050 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 1051 const struct sock *ssk, gfp_t gfp); 1052 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 1053 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 1054 void mptcp_event_pm_listener(const struct sock *ssk, 1055 enum mptcp_event_type event); 1056 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 1057 1058 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 1059 struct request_sock *req); 1060 int mptcp_pm_genl_fill_addr(struct sk_buff *msg, 1061 struct netlink_callback *cb, 1062 struct mptcp_pm_addr_entry *entry); 1063 1064 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 1065 { 1066 return READ_ONCE(msk->pm.addr_signal) & 1067 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 1068 } 1069 1070 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 1071 { 1072 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 1073 } 1074 1075 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 1076 { 1077 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 1078 } 1079 1080 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 1081 { 1082 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 1083 } 1084 1085 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1086 { 1087 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1088 } 1089 1090 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1091 { 1092 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1093 } 1094 1095 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 1096 { 1097 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 1098 1099 if (family == AF_INET6) 1100 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 1101 if (!echo) 1102 len += MPTCPOPT_THMAC_LEN; 1103 /* account for 2 trailing 'nop' options */ 1104 if (port) 1105 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 1106 1107 return len; 1108 } 1109 1110 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 1111 { 1112 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 1113 return -EINVAL; 1114 1115 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 1116 } 1117 1118 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 1119 unsigned int opt_size, unsigned int remaining, 1120 struct mptcp_addr_info *addr, bool *echo, 1121 bool *drop_other_suboptions); 1122 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1123 struct mptcp_rm_list *rm_list); 1124 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1125 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, 1126 struct mptcp_pm_addr_entry *skc); 1127 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, 1128 struct mptcp_pm_addr_entry *skc); 1129 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc); 1130 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1131 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1132 int mptcp_pm_nl_dump_addr(struct sk_buff *msg, 1133 struct netlink_callback *cb); 1134 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg, 1135 struct netlink_callback *cb); 1136 int mptcp_pm_nl_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1137 struct genl_info *info); 1138 int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1139 struct genl_info *info); 1140 1141 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1142 { 1143 int local_id = READ_ONCE(subflow->local_id); 1144 1145 if (local_id < 0) 1146 return 0; 1147 return local_id; 1148 } 1149 1150 void __init mptcp_pm_nl_init(void); 1151 void mptcp_pm_worker(struct mptcp_sock *msk); 1152 void __mptcp_pm_kernel_worker(struct mptcp_sock *msk); 1153 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 1154 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 1155 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 1156 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 1157 1158 /* called under PM lock */ 1159 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1160 { 1161 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 1162 WRITE_ONCE(msk->pm.accept_subflow, true); 1163 } 1164 1165 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1166 { 1167 spin_lock_bh(&msk->pm.lock); 1168 __mptcp_pm_close_subflow(msk); 1169 spin_unlock_bh(&msk->pm.lock); 1170 } 1171 1172 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1173 1174 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1175 { 1176 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1177 } 1178 1179 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1180 1181 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1182 { 1183 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1184 } 1185 1186 static inline bool mptcp_check_fallback(const struct sock *sk) 1187 { 1188 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1189 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1190 1191 return __mptcp_check_fallback(msk); 1192 } 1193 1194 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 1195 { 1196 if (__mptcp_check_fallback(msk)) { 1197 pr_debug("TCP fallback already done (msk=%p)\n", msk); 1198 return; 1199 } 1200 if (WARN_ON_ONCE(!READ_ONCE(msk->allow_infinite_fallback))) 1201 return; 1202 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1203 } 1204 1205 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1206 { 1207 struct sock *ssk = READ_ONCE(msk->first); 1208 1209 return ssk && ((1 << inet_sk_state_load(ssk)) & 1210 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1211 TCPF_SYN_RECV | TCPF_LISTEN)); 1212 } 1213 1214 static inline void mptcp_do_fallback(struct sock *ssk) 1215 { 1216 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1217 struct sock *sk = subflow->conn; 1218 struct mptcp_sock *msk; 1219 1220 msk = mptcp_sk(sk); 1221 __mptcp_do_fallback(msk); 1222 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1223 gfp_t saved_allocation = ssk->sk_allocation; 1224 1225 /* we are in a atomic (BH) scope, override ssk default for data 1226 * fin allocation 1227 */ 1228 ssk->sk_allocation = GFP_ATOMIC; 1229 ssk->sk_shutdown |= SEND_SHUTDOWN; 1230 tcp_shutdown(ssk, SEND_SHUTDOWN); 1231 ssk->sk_allocation = saved_allocation; 1232 } 1233 } 1234 1235 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)\n", __func__, a) 1236 1237 static inline void mptcp_subflow_early_fallback(struct mptcp_sock *msk, 1238 struct mptcp_subflow_context *subflow) 1239 { 1240 pr_fallback(msk); 1241 subflow->request_mptcp = 0; 1242 __mptcp_do_fallback(msk); 1243 } 1244 1245 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1246 { 1247 struct mptcp_ext *mpext; 1248 1249 mpext = skb ? mptcp_get_ext(skb) : NULL; 1250 if (mpext && mpext->infinite_map) 1251 return true; 1252 1253 return false; 1254 } 1255 1256 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1257 { 1258 return (subflow->request_mptcp || subflow->request_join); 1259 } 1260 1261 static inline bool subflow_simultaneous_connect(struct sock *sk) 1262 { 1263 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1264 1265 return (1 << sk->sk_state) & 1266 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) && 1267 is_active_ssk(subflow) && 1268 !subflow->conn_finished; 1269 } 1270 1271 #ifdef CONFIG_SYN_COOKIES 1272 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1273 struct sk_buff *skb); 1274 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1275 struct sk_buff *skb); 1276 void __init mptcp_join_cookie_init(void); 1277 #else 1278 static inline void 1279 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1280 struct sk_buff *skb) {} 1281 static inline bool 1282 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1283 struct sk_buff *skb) 1284 { 1285 return false; 1286 } 1287 1288 static inline void mptcp_join_cookie_init(void) {} 1289 #endif 1290 1291 #endif /* __MPTCP_PROTOCOL_H */ 1292