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 struct_group(reset, 227 228 u8 addr_signal; 229 bool server_side; 230 bool work_pending; 231 bool accept_addr; 232 bool accept_subflow; 233 bool remote_deny_join_id0; 234 u8 add_addr_signaled; 235 u8 add_addr_accepted; 236 u8 local_addr_used; 237 u8 pm_type; 238 u8 subflows; 239 u8 status; 240 241 ); 242 243 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 244 struct mptcp_rm_list rm_list_tx; 245 struct mptcp_rm_list rm_list_rx; 246 }; 247 248 struct mptcp_pm_local { 249 struct mptcp_addr_info addr; 250 u8 flags; 251 int ifindex; 252 }; 253 254 struct mptcp_pm_addr_entry { 255 struct list_head list; 256 struct mptcp_addr_info addr; 257 u8 flags; 258 int ifindex; 259 struct socket *lsk; 260 }; 261 262 struct mptcp_data_frag { 263 struct list_head list; 264 u64 data_seq; 265 u16 data_len; 266 u16 offset; 267 u16 overhead; 268 u16 already_sent; 269 struct page *page; 270 }; 271 272 /* MPTCP connection sock */ 273 struct mptcp_sock { 274 /* inet_connection_sock must be the first member */ 275 struct inet_connection_sock sk; 276 u64 local_key; /* protected by the first subflow socket lock 277 * lockless access read 278 */ 279 u64 remote_key; /* same as above */ 280 u64 write_seq; 281 u64 bytes_sent; 282 u64 snd_nxt; 283 u64 bytes_received; 284 u64 ack_seq; 285 atomic64_t rcv_wnd_sent; 286 u64 rcv_data_fin_seq; 287 u64 bytes_retrans; 288 u64 bytes_consumed; 289 int snd_burst; 290 int old_wspace; 291 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 292 * recovery related fields are under data_lock 293 * protection 294 */ 295 u64 bytes_acked; 296 u64 snd_una; 297 u64 wnd_end; 298 u32 last_data_sent; 299 u32 last_data_recv; 300 u32 last_ack_recv; 301 unsigned long timer_ival; 302 u32 token; 303 unsigned long flags; 304 unsigned long cb_flags; 305 bool recovery; /* closing subflow write queue reinjected */ 306 bool can_ack; 307 bool fully_established; 308 bool rcv_data_fin; 309 bool snd_data_fin_enable; 310 bool rcv_fastclose; 311 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 312 bool csum_enabled; 313 bool allow_infinite_fallback; 314 u8 pending_state; /* A subflow asked to set this sk_state, 315 * protected by the msk data lock 316 */ 317 u8 mpc_endpoint_id; 318 u8 recvmsg_inq:1, 319 cork:1, 320 nodelay:1, 321 fastopening:1, 322 in_accept_queue:1, 323 free_first:1, 324 rcvspace_init:1; 325 u32 notsent_lowat; 326 int keepalive_cnt; 327 int keepalive_idle; 328 int keepalive_intvl; 329 struct work_struct work; 330 struct sk_buff *ooo_last_skb; 331 struct rb_root out_of_order_queue; 332 struct list_head conn_list; 333 struct list_head rtx_queue; 334 struct mptcp_data_frag *first_pending; 335 struct list_head join_list; 336 struct sock *first; /* The mptcp ops can safely dereference, using suitable 337 * ONCE annotation, the subflow outside the socket 338 * lock as such sock is freed after close(). 339 */ 340 struct mptcp_pm_data pm; 341 struct mptcp_sched_ops *sched; 342 struct { 343 u32 space; /* bytes copied in last measurement window */ 344 u32 copied; /* bytes copied in this measurement window */ 345 u64 time; /* start time of measurement window */ 346 u64 rtt_us; /* last maximum rtt of subflows */ 347 } rcvq_space; 348 u8 scaling_ratio; 349 350 u32 subflow_id; 351 u32 setsockopt_seq; 352 char ca_name[TCP_CA_NAME_MAX]; 353 }; 354 355 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 356 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 357 358 #define mptcp_for_each_subflow(__msk, __subflow) \ 359 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 360 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 361 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 362 #define mptcp_next_subflow(__msk, __subflow) \ 363 list_next_entry_circular(__subflow, &((__msk)->conn_list), node) 364 365 extern struct genl_family mptcp_genl_family; 366 367 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 368 { 369 sock_owned_by_me((const struct sock *)msk); 370 } 371 372 #ifdef CONFIG_DEBUG_NET 373 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 374 #undef tcp_sk 375 #define tcp_sk(ptr) ({ \ 376 typeof(ptr) _ptr = (ptr); \ 377 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 378 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 379 }) 380 #define mptcp_sk(ptr) ({ \ 381 typeof(ptr) _ptr = (ptr); \ 382 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 383 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 384 }) 385 386 #else /* !CONFIG_DEBUG_NET */ 387 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 388 #endif 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) - 403 sk_rmem_alloc_get(sk)); 404 } 405 406 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 407 { 408 const struct mptcp_sock *msk = mptcp_sk(sk); 409 410 return READ_ONCE(msk->first_pending); 411 } 412 413 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 414 { 415 struct mptcp_sock *msk = mptcp_sk(sk); 416 struct mptcp_data_frag *cur; 417 418 cur = msk->first_pending; 419 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 420 list_next_entry(cur, list); 421 } 422 423 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 424 { 425 const struct mptcp_sock *msk = mptcp_sk(sk); 426 427 if (!msk->first_pending) 428 return NULL; 429 430 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 431 return NULL; 432 433 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 434 } 435 436 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 437 { 438 struct mptcp_sock *msk = mptcp_sk(sk); 439 440 if (msk->snd_una == msk->snd_nxt) 441 return NULL; 442 443 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 444 } 445 446 struct csum_pseudo_header { 447 __be64 data_seq; 448 __be32 subflow_seq; 449 __be16 data_len; 450 __sum16 csum; 451 }; 452 453 struct mptcp_subflow_request_sock { 454 struct tcp_request_sock sk; 455 u16 mp_capable : 1, 456 mp_join : 1, 457 backup : 1, 458 request_bkup : 1, 459 csum_reqd : 1, 460 allow_join_id0 : 1; 461 u8 local_id; 462 u8 remote_id; 463 u64 local_key; 464 u64 idsn; 465 u32 token; 466 u32 ssn_offset; 467 u64 thmac; 468 u32 local_nonce; 469 u32 remote_nonce; 470 struct mptcp_sock *msk; 471 struct hlist_nulls_node token_node; 472 }; 473 474 static inline struct mptcp_subflow_request_sock * 475 mptcp_subflow_rsk(const struct request_sock *rsk) 476 { 477 return (struct mptcp_subflow_request_sock *)rsk; 478 } 479 480 struct mptcp_delegated_action { 481 struct napi_struct napi; 482 local_lock_t bh_lock; 483 struct list_head head; 484 }; 485 486 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 487 488 #define MPTCP_DELEGATE_SCHEDULED 0 489 #define MPTCP_DELEGATE_SEND 1 490 #define MPTCP_DELEGATE_ACK 2 491 #define MPTCP_DELEGATE_SNDBUF 3 492 493 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 494 /* MPTCP subflow context */ 495 struct mptcp_subflow_context { 496 struct list_head node;/* conn_list of subflows */ 497 498 struct_group(reset, 499 500 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 501 u64 local_key; 502 u64 remote_key; 503 u64 idsn; 504 u64 map_seq; 505 u32 snd_isn; 506 u32 token; 507 u32 rel_write_seq; 508 u32 map_subflow_seq; 509 u32 ssn_offset; 510 u32 map_data_len; 511 __wsum map_data_csum; 512 u32 map_csum_len; 513 u32 request_mptcp : 1, /* send MP_CAPABLE */ 514 request_join : 1, /* send MP_JOIN */ 515 request_bkup : 1, 516 mp_capable : 1, /* remote is MPTCP capable */ 517 mp_join : 1, /* remote is JOINing */ 518 pm_notified : 1, /* PM hook called for established status */ 519 conn_finished : 1, 520 map_valid : 1, 521 map_csum_reqd : 1, 522 map_data_fin : 1, 523 mpc_map : 1, 524 backup : 1, 525 send_mp_prio : 1, 526 send_mp_fail : 1, 527 send_fastclose : 1, 528 send_infinite_map : 1, 529 remote_key_valid : 1, /* received the peer key from */ 530 disposable : 1, /* ctx can be free at ulp release time */ 531 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 532 valid_csum_seen : 1, /* at least one csum validated */ 533 is_mptfo : 1, /* subflow is doing TFO */ 534 close_event_done : 1, /* has done the post-closed part */ 535 mpc_drop : 1, /* the MPC option has been dropped in a rtx */ 536 __unused : 9; 537 bool data_avail; 538 bool scheduled; 539 bool pm_listener; /* a listener managed by the kernel PM? */ 540 bool fully_established; /* path validated */ 541 u32 remote_nonce; 542 u64 thmac; 543 u32 local_nonce; 544 u32 remote_token; 545 union { 546 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 547 u64 iasn; /* initial ack sequence number, MPC subflows only */ 548 }; 549 s16 local_id; /* if negative not initialized yet */ 550 u8 remote_id; 551 u8 reset_seen:1; 552 u8 reset_transient:1; 553 u8 reset_reason:4; 554 u8 stale_count; 555 556 u32 subflow_id; 557 558 long delegated_status; 559 unsigned long fail_tout; 560 561 ); 562 563 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 564 565 u32 setsockopt_seq; 566 u32 stale_rcv_tstamp; 567 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 568 * protected by the msk socket lock 569 */ 570 571 struct sock *tcp_sock; /* tcp sk backpointer */ 572 struct sock *conn; /* parent mptcp_sock */ 573 const struct inet_connection_sock_af_ops *icsk_af_ops; 574 void (*tcp_state_change)(struct sock *sk); 575 void (*tcp_error_report)(struct sock *sk); 576 577 struct rcu_head rcu; 578 }; 579 580 static inline struct mptcp_subflow_context * 581 mptcp_subflow_ctx(const struct sock *sk) 582 { 583 const struct inet_connection_sock *icsk = inet_csk(sk); 584 585 /* Use RCU on icsk_ulp_data only for sock diag code */ 586 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 587 } 588 589 static inline struct sock * 590 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 591 { 592 return subflow->tcp_sock; 593 } 594 595 static inline void 596 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 597 { 598 memset(&subflow->reset, 0, sizeof(subflow->reset)); 599 subflow->request_mptcp = 1; 600 WRITE_ONCE(subflow->local_id, -1); 601 } 602 603 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */ 604 static inline enum sk_rst_reason 605 sk_rst_convert_mptcp_reason(u32 reason) 606 { 607 switch (reason) { 608 case MPTCP_RST_EUNSPEC: 609 return SK_RST_REASON_MPTCP_RST_EUNSPEC; 610 case MPTCP_RST_EMPTCP: 611 return SK_RST_REASON_MPTCP_RST_EMPTCP; 612 case MPTCP_RST_ERESOURCE: 613 return SK_RST_REASON_MPTCP_RST_ERESOURCE; 614 case MPTCP_RST_EPROHIBIT: 615 return SK_RST_REASON_MPTCP_RST_EPROHIBIT; 616 case MPTCP_RST_EWQ2BIG: 617 return SK_RST_REASON_MPTCP_RST_EWQ2BIG; 618 case MPTCP_RST_EBADPERF: 619 return SK_RST_REASON_MPTCP_RST_EBADPERF; 620 case MPTCP_RST_EMIDDLEBOX: 621 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX; 622 default: 623 /* It should not happen, or else errors may occur 624 * in MPTCP layer 625 */ 626 return SK_RST_REASON_ERROR; 627 } 628 } 629 630 static inline void 631 mptcp_send_active_reset_reason(struct sock *sk) 632 { 633 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 634 enum sk_rst_reason reason; 635 636 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason); 637 tcp_send_active_reset(sk, GFP_ATOMIC, reason); 638 } 639 640 static inline u64 641 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 642 { 643 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 644 subflow->ssn_offset - 645 subflow->map_subflow_seq; 646 } 647 648 static inline u64 649 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 650 { 651 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 652 } 653 654 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 655 656 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 657 { 658 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 659 struct mptcp_delegated_action *delegated; 660 bool schedule; 661 662 /* the caller held the subflow bh socket lock */ 663 lockdep_assert_in_softirq(); 664 665 /* The implied barrier pairs with tcp_release_cb_override() 666 * mptcp_napi_poll(), and ensures the below list check sees list 667 * updates done prior to delegated status bits changes 668 */ 669 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 670 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 671 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 672 return; 673 674 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 675 delegated = this_cpu_ptr(&mptcp_delegated_actions); 676 schedule = list_empty(&delegated->head); 677 list_add_tail(&subflow->delegated_node, &delegated->head); 678 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 679 sock_hold(mptcp_subflow_tcp_sock(subflow)); 680 if (schedule) 681 napi_schedule(&delegated->napi); 682 } 683 } 684 685 static inline struct mptcp_subflow_context * 686 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 687 { 688 struct mptcp_subflow_context *ret; 689 690 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 691 if (list_empty(&delegated->head)) { 692 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 693 return NULL; 694 } 695 696 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 697 list_del_init(&ret->delegated_node); 698 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 699 return ret; 700 } 701 702 int mptcp_is_enabled(const struct net *net); 703 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 704 int mptcp_is_checksum_enabled(const struct net *net); 705 int mptcp_allow_join_id0(const struct net *net); 706 unsigned int mptcp_stale_loss_cnt(const struct net *net); 707 unsigned int mptcp_close_timeout(const struct sock *sk); 708 int mptcp_get_pm_type(const struct net *net); 709 const char *mptcp_get_path_manager(const struct net *net); 710 const char *mptcp_get_scheduler(const struct net *net); 711 712 void mptcp_active_disable(struct sock *sk); 713 bool mptcp_active_should_disable(struct sock *ssk); 714 void mptcp_active_enable(struct sock *sk); 715 716 void mptcp_get_available_schedulers(char *buf, size_t maxlen); 717 void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 718 struct mptcp_subflow_context *subflow, 719 const struct mptcp_options_received *mp_opt); 720 bool __mptcp_retransmit_pending_data(struct sock *sk); 721 void mptcp_check_and_set_pending(struct sock *sk); 722 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 723 bool mptcp_subflow_data_available(struct sock *sk); 724 void __init mptcp_subflow_init(void); 725 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 726 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 727 struct mptcp_subflow_context *subflow); 728 void __mptcp_subflow_send_ack(struct sock *ssk); 729 void mptcp_subflow_reset(struct sock *ssk); 730 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 731 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 732 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 733 bool __mptcp_close(struct sock *sk, long timeout); 734 void mptcp_cancel_work(struct sock *sk); 735 void __mptcp_unaccepted_force_close(struct sock *sk); 736 void mptcp_set_state(struct sock *sk, int state); 737 738 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 739 const struct mptcp_addr_info *b, bool use_port); 740 void mptcp_local_address(const struct sock_common *skc, 741 struct mptcp_addr_info *addr); 742 void mptcp_remote_address(const struct sock_common *skc, 743 struct mptcp_addr_info *addr); 744 745 /* called with sk socket lock held */ 746 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local, 747 const struct mptcp_addr_info *remote); 748 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 749 struct socket **new_sock); 750 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 751 struct sockaddr_storage *addr, 752 unsigned short family); 753 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 754 int mptcp_validate_scheduler(struct mptcp_sched_ops *sched); 755 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 756 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 757 void mptcp_sched_init(void); 758 int mptcp_init_sched(struct mptcp_sock *msk, 759 struct mptcp_sched_ops *sched); 760 void mptcp_release_sched(struct mptcp_sock *msk); 761 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 762 bool scheduled); 763 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 764 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 765 int mptcp_sched_get_send(struct mptcp_sock *msk); 766 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 767 768 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 769 { 770 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 771 } 772 773 static inline bool mptcp_epollin_ready(const struct sock *sk) 774 { 775 u64 data_avail = mptcp_data_avail(mptcp_sk(sk)); 776 777 if (!data_avail) 778 return false; 779 780 /* mptcp doesn't have to deal with small skbs in the receive queue, 781 * as it can always coalesce them 782 */ 783 return (data_avail >= sk->sk_rcvlowat) || 784 (mem_cgroup_sockets_enabled && sk->sk_memcg && 785 mem_cgroup_under_socket_pressure(sk->sk_memcg)) || 786 READ_ONCE(tcp_memory_pressure); 787 } 788 789 int mptcp_set_rcvlowat(struct sock *sk, int val); 790 791 static inline bool __tcp_can_send(const struct sock *ssk) 792 { 793 /* only send if our side has not closed yet */ 794 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 795 } 796 797 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 798 { 799 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 800 if (subflow->request_join && !READ_ONCE(subflow->fully_established)) 801 return false; 802 803 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 804 } 805 806 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 807 808 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 809 810 void mptcp_subflow_drop_ctx(struct sock *ssk); 811 812 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 813 struct mptcp_subflow_context *ctx) 814 { 815 sk->sk_data_ready = sock_def_readable; 816 sk->sk_state_change = ctx->tcp_state_change; 817 sk->sk_write_space = sk_stream_write_space; 818 sk->sk_error_report = ctx->tcp_error_report; 819 820 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 821 } 822 823 void __init mptcp_proto_init(void); 824 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 825 int __init mptcp_proto_v6_init(void); 826 #endif 827 828 struct sock *mptcp_sk_clone_init(const struct sock *sk, 829 const struct mptcp_options_received *mp_opt, 830 struct sock *ssk, 831 struct request_sock *req); 832 void mptcp_get_options(const struct sk_buff *skb, 833 struct mptcp_options_received *mp_opt); 834 835 void mptcp_finish_connect(struct sock *sk); 836 void __mptcp_sync_state(struct sock *sk, int state); 837 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 838 839 static inline void mptcp_stop_tout_timer(struct sock *sk) 840 { 841 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 842 return; 843 844 sk_stop_timer(sk, &sk->sk_timer); 845 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 846 } 847 848 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 849 { 850 /* avoid 0 timestamp, as that means no close timeout */ 851 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 852 } 853 854 static inline void mptcp_start_tout_timer(struct sock *sk) 855 { 856 mptcp_set_close_tout(sk, tcp_jiffies32); 857 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 858 } 859 860 static inline bool mptcp_is_fully_established(struct sock *sk) 861 { 862 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 863 READ_ONCE(mptcp_sk(sk)->fully_established); 864 } 865 866 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 867 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 868 bool mptcp_finish_join(struct sock *sk); 869 bool mptcp_schedule_work(struct sock *sk); 870 int mptcp_setsockopt(struct sock *sk, int level, int optname, 871 sockptr_t optval, unsigned int optlen); 872 int mptcp_getsockopt(struct sock *sk, int level, int optname, 873 char __user *optval, int __user *option); 874 875 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 876 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 877 { 878 if (use_64bit) 879 return cur_seq; 880 881 return __mptcp_expand_seq(old_seq, cur_seq); 882 } 883 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 884 void __mptcp_data_acked(struct sock *sk); 885 void __mptcp_error_report(struct sock *sk); 886 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 887 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 888 { 889 return READ_ONCE(msk->snd_data_fin_enable) && 890 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 891 } 892 893 static inline u32 mptcp_notsent_lowat(const struct sock *sk) 894 { 895 struct net *net = sock_net(sk); 896 u32 val; 897 898 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat); 899 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat); 900 } 901 902 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake) 903 { 904 const struct mptcp_sock *msk = mptcp_sk(sk); 905 u32 notsent_bytes; 906 907 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt); 908 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk); 909 } 910 911 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake) 912 { 913 return mptcp_stream_memory_free(sk, wake) && 914 __sk_stream_is_writeable(sk, wake); 915 } 916 917 static inline void mptcp_write_space(struct sock *sk) 918 { 919 /* pairs with memory barrier in mptcp_poll */ 920 smp_mb(); 921 if (mptcp_stream_memory_free(sk, 1)) 922 sk_stream_write_space(sk); 923 } 924 925 static inline void __mptcp_sync_sndbuf(struct sock *sk) 926 { 927 struct mptcp_subflow_context *subflow; 928 int ssk_sndbuf, new_sndbuf; 929 930 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 931 return; 932 933 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]); 934 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 935 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 936 937 subflow->cached_sndbuf = ssk_sndbuf; 938 new_sndbuf += ssk_sndbuf; 939 } 940 941 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 942 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 943 mptcp_write_space(sk); 944 } 945 946 /* The called held both the msk socket and the subflow socket locks, 947 * possibly under BH 948 */ 949 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 950 { 951 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 952 953 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 954 __mptcp_sync_sndbuf(sk); 955 } 956 957 /* the caller held only the subflow socket lock, either in process or 958 * BH context. Additionally this can be called under the msk data lock, 959 * so we can't acquire such lock here: let the delegate action acquires 960 * the needed locks in suitable order. 961 */ 962 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 963 { 964 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 965 966 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 967 return; 968 969 local_bh_disable(); 970 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 971 local_bh_enable(); 972 } 973 974 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 975 976 #define MPTCP_TOKEN_MAX_RETRIES 4 977 978 void __init mptcp_token_init(void); 979 static inline void mptcp_token_init_request(struct request_sock *req) 980 { 981 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 982 } 983 984 int mptcp_token_new_request(struct request_sock *req); 985 void mptcp_token_destroy_request(struct request_sock *req); 986 int mptcp_token_new_connect(struct sock *ssk); 987 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 988 struct mptcp_sock *msk); 989 bool mptcp_token_exists(u32 token); 990 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 991 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 992 long *s_num); 993 void mptcp_token_destroy(struct mptcp_sock *msk); 994 995 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 996 997 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 998 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 999 1000 void __init mptcp_pm_init(void); 1001 void mptcp_pm_data_init(struct mptcp_sock *msk); 1002 void mptcp_pm_data_reset(struct mptcp_sock *msk); 1003 void mptcp_pm_destroy(struct mptcp_sock *msk); 1004 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 1005 struct mptcp_addr_info *addr); 1006 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 1007 bool require_family, 1008 struct mptcp_pm_addr_entry *entry); 1009 bool mptcp_pm_addr_families_match(const struct sock *sk, 1010 const struct mptcp_addr_info *loc, 1011 const struct mptcp_addr_info *rem); 1012 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 1013 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 1014 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 1015 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 1016 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 1017 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 1018 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 1019 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 1020 const struct mptcp_subflow_context *subflow); 1021 void mptcp_pm_add_addr_received(const struct sock *ssk, 1022 const struct mptcp_addr_info *addr); 1023 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 1024 const struct mptcp_addr_info *addr); 1025 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 1026 void mptcp_pm_send_ack(struct mptcp_sock *msk, 1027 struct mptcp_subflow_context *subflow, 1028 bool prio, bool backup); 1029 void mptcp_pm_addr_send_ack(struct mptcp_sock *msk); 1030 void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id); 1031 void mptcp_pm_rm_subflow(struct mptcp_sock *msk, 1032 const struct mptcp_rm_list *rm_list); 1033 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 1034 const struct mptcp_rm_list *rm_list); 1035 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 1036 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 1037 int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk, 1038 struct mptcp_addr_info *addr, 1039 struct mptcp_addr_info *rem, 1040 u8 bkup); 1041 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 1042 const struct mptcp_addr_info *addr); 1043 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 1044 struct mptcp_pm_add_entry * 1045 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 1046 const struct mptcp_addr_info *addr, bool check_id); 1047 bool mptcp_lookup_subflow_by_saddr(const struct list_head *list, 1048 const struct mptcp_addr_info *saddr); 1049 bool mptcp_remove_anno_list_by_saddr(struct mptcp_sock *msk, 1050 const struct mptcp_addr_info *addr); 1051 int mptcp_pm_nl_set_flags(struct mptcp_pm_addr_entry *local, 1052 struct genl_info *info); 1053 int mptcp_userspace_pm_set_flags(struct mptcp_pm_addr_entry *local, 1054 struct genl_info *info); 1055 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 1056 const struct mptcp_addr_info *addr, 1057 bool echo); 1058 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 1059 void mptcp_pm_remove_addr_entry(struct mptcp_sock *msk, 1060 struct mptcp_pm_addr_entry *entry); 1061 1062 /* the default path manager, used in mptcp_pm_unregister */ 1063 extern struct mptcp_pm_ops mptcp_pm_kernel; 1064 1065 struct mptcp_pm_ops *mptcp_pm_find(const char *name); 1066 int mptcp_pm_register(struct mptcp_pm_ops *pm_ops); 1067 void mptcp_pm_unregister(struct mptcp_pm_ops *pm_ops); 1068 int mptcp_pm_validate(struct mptcp_pm_ops *pm_ops); 1069 void mptcp_pm_get_available(char *buf, size_t maxlen); 1070 1071 void mptcp_userspace_pm_free_local_addr_list(struct mptcp_sock *msk); 1072 1073 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 1074 const struct sock *ssk, gfp_t gfp); 1075 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 1076 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 1077 void mptcp_event_pm_listener(const struct sock *ssk, 1078 enum mptcp_event_type event); 1079 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 1080 1081 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 1082 struct request_sock *req); 1083 int mptcp_pm_genl_fill_addr(struct sk_buff *msg, 1084 struct netlink_callback *cb, 1085 struct mptcp_pm_addr_entry *entry); 1086 1087 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 1088 { 1089 return READ_ONCE(msk->pm.addr_signal) & 1090 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 1091 } 1092 1093 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 1094 { 1095 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 1096 } 1097 1098 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 1099 { 1100 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 1101 } 1102 1103 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 1104 { 1105 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 1106 } 1107 1108 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1109 { 1110 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1111 } 1112 1113 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1114 { 1115 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1116 } 1117 1118 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 1119 { 1120 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 1121 1122 if (family == AF_INET6) 1123 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 1124 if (!echo) 1125 len += MPTCPOPT_THMAC_LEN; 1126 /* account for 2 trailing 'nop' options */ 1127 if (port) 1128 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 1129 1130 return len; 1131 } 1132 1133 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 1134 { 1135 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 1136 return -EINVAL; 1137 1138 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 1139 } 1140 1141 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 1142 unsigned int opt_size, unsigned int remaining, 1143 struct mptcp_addr_info *addr, bool *echo, 1144 bool *drop_other_suboptions); 1145 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1146 struct mptcp_rm_list *rm_list); 1147 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1148 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, 1149 struct mptcp_pm_addr_entry *skc); 1150 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, 1151 struct mptcp_pm_addr_entry *skc); 1152 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc); 1153 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1154 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1155 int mptcp_pm_nl_dump_addr(struct sk_buff *msg, 1156 struct netlink_callback *cb); 1157 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg, 1158 struct netlink_callback *cb); 1159 int mptcp_pm_nl_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1160 struct genl_info *info); 1161 int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1162 struct genl_info *info); 1163 1164 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1165 { 1166 int local_id = READ_ONCE(subflow->local_id); 1167 1168 if (local_id < 0) 1169 return 0; 1170 return local_id; 1171 } 1172 1173 void __init mptcp_pm_kernel_register(void); 1174 void __init mptcp_pm_userspace_register(void); 1175 void __init mptcp_pm_nl_init(void); 1176 void mptcp_pm_worker(struct mptcp_sock *msk); 1177 void __mptcp_pm_kernel_worker(struct mptcp_sock *msk); 1178 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 1179 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 1180 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 1181 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 1182 1183 /* called under PM lock */ 1184 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1185 { 1186 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 1187 WRITE_ONCE(msk->pm.accept_subflow, true); 1188 } 1189 1190 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1191 { 1192 spin_lock_bh(&msk->pm.lock); 1193 __mptcp_pm_close_subflow(msk); 1194 spin_unlock_bh(&msk->pm.lock); 1195 } 1196 1197 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1198 1199 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1200 { 1201 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1202 } 1203 1204 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1205 1206 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1207 { 1208 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1209 } 1210 1211 static inline bool mptcp_check_fallback(const struct sock *sk) 1212 { 1213 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1214 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1215 1216 return __mptcp_check_fallback(msk); 1217 } 1218 1219 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 1220 { 1221 if (__mptcp_check_fallback(msk)) { 1222 pr_debug("TCP fallback already done (msk=%p)\n", msk); 1223 return; 1224 } 1225 if (WARN_ON_ONCE(!READ_ONCE(msk->allow_infinite_fallback))) 1226 return; 1227 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1228 } 1229 1230 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1231 { 1232 struct sock *ssk = READ_ONCE(msk->first); 1233 1234 return ssk && ((1 << inet_sk_state_load(ssk)) & 1235 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1236 TCPF_SYN_RECV | TCPF_LISTEN)); 1237 } 1238 1239 static inline void mptcp_do_fallback(struct sock *ssk) 1240 { 1241 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1242 struct sock *sk = subflow->conn; 1243 struct mptcp_sock *msk; 1244 1245 msk = mptcp_sk(sk); 1246 __mptcp_do_fallback(msk); 1247 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1248 gfp_t saved_allocation = ssk->sk_allocation; 1249 1250 /* we are in a atomic (BH) scope, override ssk default for data 1251 * fin allocation 1252 */ 1253 ssk->sk_allocation = GFP_ATOMIC; 1254 ssk->sk_shutdown |= SEND_SHUTDOWN; 1255 tcp_shutdown(ssk, SEND_SHUTDOWN); 1256 ssk->sk_allocation = saved_allocation; 1257 } 1258 } 1259 1260 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)\n", __func__, a) 1261 1262 static inline void mptcp_subflow_early_fallback(struct mptcp_sock *msk, 1263 struct mptcp_subflow_context *subflow) 1264 { 1265 pr_fallback(msk); 1266 subflow->request_mptcp = 0; 1267 __mptcp_do_fallback(msk); 1268 } 1269 1270 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1271 { 1272 struct mptcp_ext *mpext; 1273 1274 mpext = skb ? mptcp_get_ext(skb) : NULL; 1275 if (mpext && mpext->infinite_map) 1276 return true; 1277 1278 return false; 1279 } 1280 1281 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1282 { 1283 return (subflow->request_mptcp || subflow->request_join); 1284 } 1285 1286 static inline bool subflow_simultaneous_connect(struct sock *sk) 1287 { 1288 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1289 1290 return (1 << sk->sk_state) & 1291 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) && 1292 is_active_ssk(subflow) && 1293 !subflow->conn_finished; 1294 } 1295 1296 #ifdef CONFIG_SYN_COOKIES 1297 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1298 struct sk_buff *skb); 1299 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1300 struct sk_buff *skb); 1301 void __init mptcp_join_cookie_init(void); 1302 #else 1303 static inline void 1304 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1305 struct sk_buff *skb) {} 1306 static inline bool 1307 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1308 struct sk_buff *skb) 1309 { 1310 return false; 1311 } 1312 1313 static inline void mptcp_join_cookie_init(void) {} 1314 #endif 1315 1316 #endif /* __MPTCP_PROTOCOL_H */ 1317