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 int maxseg; 330 struct work_struct work; 331 struct sk_buff *ooo_last_skb; 332 struct rb_root out_of_order_queue; 333 struct list_head conn_list; 334 struct list_head rtx_queue; 335 struct mptcp_data_frag *first_pending; 336 struct list_head join_list; 337 struct sock *first; /* The mptcp ops can safely dereference, using suitable 338 * ONCE annotation, the subflow outside the socket 339 * lock as such sock is freed after close(). 340 */ 341 struct mptcp_pm_data pm; 342 struct mptcp_sched_ops *sched; 343 struct { 344 u32 space; /* bytes copied in last measurement window */ 345 u32 copied; /* bytes copied in this measurement window */ 346 u64 time; /* start time of measurement window */ 347 u64 rtt_us; /* last maximum rtt of subflows */ 348 } rcvq_space; 349 u8 scaling_ratio; 350 bool allow_subflows; 351 352 u32 subflow_id; 353 u32 setsockopt_seq; 354 char ca_name[TCP_CA_NAME_MAX]; 355 356 spinlock_t fallback_lock; /* protects fallback, 357 * allow_infinite_fallback and 358 * allow_join 359 */ 360 }; 361 362 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 363 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 364 365 #define mptcp_for_each_subflow(__msk, __subflow) \ 366 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 367 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 368 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 369 #define mptcp_next_subflow(__msk, __subflow) \ 370 list_next_entry_circular(__subflow, &((__msk)->conn_list), node) 371 372 extern struct genl_family mptcp_genl_family; 373 374 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 375 { 376 sock_owned_by_me((const struct sock *)msk); 377 } 378 379 #ifdef CONFIG_DEBUG_NET 380 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 381 #undef tcp_sk 382 #define tcp_sk(ptr) ({ \ 383 typeof(ptr) _ptr = (ptr); \ 384 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 385 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 386 }) 387 #define mptcp_sk(ptr) ({ \ 388 typeof(ptr) _ptr = (ptr); \ 389 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 390 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 391 }) 392 393 #else /* !CONFIG_DEBUG_NET */ 394 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 395 #endif 396 397 static inline int mptcp_win_from_space(const struct sock *sk, int space) 398 { 399 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 400 } 401 402 static inline int mptcp_space_from_win(const struct sock *sk, int win) 403 { 404 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win); 405 } 406 407 static inline int __mptcp_space(const struct sock *sk) 408 { 409 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - 410 sk_rmem_alloc_get(sk)); 411 } 412 413 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 414 { 415 const struct mptcp_sock *msk = mptcp_sk(sk); 416 417 return READ_ONCE(msk->first_pending); 418 } 419 420 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 421 { 422 struct mptcp_sock *msk = mptcp_sk(sk); 423 struct mptcp_data_frag *cur; 424 425 cur = msk->first_pending; 426 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 427 list_next_entry(cur, list); 428 } 429 430 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 431 { 432 const struct mptcp_sock *msk = mptcp_sk(sk); 433 434 if (!msk->first_pending) 435 return NULL; 436 437 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 438 return NULL; 439 440 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 441 } 442 443 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 444 { 445 struct mptcp_sock *msk = mptcp_sk(sk); 446 447 if (msk->snd_una == msk->snd_nxt) 448 return NULL; 449 450 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 451 } 452 453 struct csum_pseudo_header { 454 __be64 data_seq; 455 __be32 subflow_seq; 456 __be16 data_len; 457 __sum16 csum; 458 }; 459 460 struct mptcp_subflow_request_sock { 461 struct tcp_request_sock sk; 462 u16 mp_capable : 1, 463 mp_join : 1, 464 backup : 1, 465 request_bkup : 1, 466 csum_reqd : 1, 467 allow_join_id0 : 1; 468 u8 local_id; 469 u8 remote_id; 470 u64 local_key; 471 u64 idsn; 472 u32 token; 473 u32 ssn_offset; 474 u64 thmac; 475 u32 local_nonce; 476 u32 remote_nonce; 477 struct mptcp_sock *msk; 478 struct hlist_nulls_node token_node; 479 }; 480 481 static inline struct mptcp_subflow_request_sock * 482 mptcp_subflow_rsk(const struct request_sock *rsk) 483 { 484 return (struct mptcp_subflow_request_sock *)rsk; 485 } 486 487 struct mptcp_delegated_action { 488 struct napi_struct napi; 489 local_lock_t bh_lock; 490 struct list_head head; 491 }; 492 493 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 494 495 #define MPTCP_DELEGATE_SCHEDULED 0 496 #define MPTCP_DELEGATE_SEND 1 497 #define MPTCP_DELEGATE_ACK 2 498 #define MPTCP_DELEGATE_SNDBUF 3 499 500 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 501 /* MPTCP subflow context */ 502 struct mptcp_subflow_context { 503 struct list_head node;/* conn_list of subflows */ 504 505 struct_group(reset, 506 507 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 508 u64 local_key; 509 u64 remote_key; 510 u64 idsn; 511 u64 map_seq; 512 u32 snd_isn; 513 u32 token; 514 u32 rel_write_seq; 515 u32 map_subflow_seq; 516 u32 ssn_offset; 517 u32 map_data_len; 518 __wsum map_data_csum; 519 u32 map_csum_len; 520 u32 request_mptcp : 1, /* send MP_CAPABLE */ 521 request_join : 1, /* send MP_JOIN */ 522 request_bkup : 1, 523 mp_capable : 1, /* remote is MPTCP capable */ 524 mp_join : 1, /* remote is JOINing */ 525 pm_notified : 1, /* PM hook called for established status */ 526 conn_finished : 1, 527 map_valid : 1, 528 map_csum_reqd : 1, 529 map_data_fin : 1, 530 mpc_map : 1, 531 backup : 1, 532 send_mp_prio : 1, 533 send_mp_fail : 1, 534 send_fastclose : 1, 535 send_infinite_map : 1, 536 remote_key_valid : 1, /* received the peer key from */ 537 disposable : 1, /* ctx can be free at ulp release time */ 538 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 539 valid_csum_seen : 1, /* at least one csum validated */ 540 is_mptfo : 1, /* subflow is doing TFO */ 541 close_event_done : 1, /* has done the post-closed part */ 542 mpc_drop : 1, /* the MPC option has been dropped in a rtx */ 543 __unused : 9; 544 bool data_avail; 545 bool scheduled; 546 bool pm_listener; /* a listener managed by the kernel PM? */ 547 bool fully_established; /* path validated */ 548 u32 remote_nonce; 549 u64 thmac; 550 u32 local_nonce; 551 u32 remote_token; 552 union { 553 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 554 u64 iasn; /* initial ack sequence number, MPC subflows only */ 555 }; 556 s16 local_id; /* if negative not initialized yet */ 557 u8 remote_id; 558 u8 reset_seen:1; 559 u8 reset_transient:1; 560 u8 reset_reason:4; 561 u8 stale_count; 562 563 u32 subflow_id; 564 565 long delegated_status; 566 unsigned long fail_tout; 567 568 ); 569 570 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 571 572 u32 setsockopt_seq; 573 u32 stale_rcv_tstamp; 574 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 575 * protected by the msk socket lock 576 */ 577 578 struct sock *tcp_sock; /* tcp sk backpointer */ 579 struct sock *conn; /* parent mptcp_sock */ 580 const struct inet_connection_sock_af_ops *icsk_af_ops; 581 void (*tcp_state_change)(struct sock *sk); 582 void (*tcp_error_report)(struct sock *sk); 583 584 struct rcu_head rcu; 585 }; 586 587 static inline struct mptcp_subflow_context * 588 mptcp_subflow_ctx(const struct sock *sk) 589 { 590 const struct inet_connection_sock *icsk = inet_csk(sk); 591 592 /* Use RCU on icsk_ulp_data only for sock diag code */ 593 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 594 } 595 596 static inline struct sock * 597 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 598 { 599 return subflow->tcp_sock; 600 } 601 602 static inline void 603 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 604 { 605 memset(&subflow->reset, 0, sizeof(subflow->reset)); 606 subflow->request_mptcp = 1; 607 WRITE_ONCE(subflow->local_id, -1); 608 } 609 610 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */ 611 static inline enum sk_rst_reason 612 sk_rst_convert_mptcp_reason(u32 reason) 613 { 614 switch (reason) { 615 case MPTCP_RST_EUNSPEC: 616 return SK_RST_REASON_MPTCP_RST_EUNSPEC; 617 case MPTCP_RST_EMPTCP: 618 return SK_RST_REASON_MPTCP_RST_EMPTCP; 619 case MPTCP_RST_ERESOURCE: 620 return SK_RST_REASON_MPTCP_RST_ERESOURCE; 621 case MPTCP_RST_EPROHIBIT: 622 return SK_RST_REASON_MPTCP_RST_EPROHIBIT; 623 case MPTCP_RST_EWQ2BIG: 624 return SK_RST_REASON_MPTCP_RST_EWQ2BIG; 625 case MPTCP_RST_EBADPERF: 626 return SK_RST_REASON_MPTCP_RST_EBADPERF; 627 case MPTCP_RST_EMIDDLEBOX: 628 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX; 629 default: 630 /* It should not happen, or else errors may occur 631 * in MPTCP layer 632 */ 633 return SK_RST_REASON_ERROR; 634 } 635 } 636 637 static inline void 638 mptcp_send_active_reset_reason(struct sock *sk) 639 { 640 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 641 enum sk_rst_reason reason; 642 643 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason); 644 tcp_send_active_reset(sk, GFP_ATOMIC, reason); 645 } 646 647 static inline u64 648 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 649 { 650 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 651 subflow->ssn_offset - 652 subflow->map_subflow_seq; 653 } 654 655 static inline u64 656 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 657 { 658 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 659 } 660 661 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 662 663 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 664 { 665 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 666 struct mptcp_delegated_action *delegated; 667 bool schedule; 668 669 /* the caller held the subflow bh socket lock */ 670 lockdep_assert_in_softirq(); 671 672 /* The implied barrier pairs with tcp_release_cb_override() 673 * mptcp_napi_poll(), and ensures the below list check sees list 674 * updates done prior to delegated status bits changes 675 */ 676 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 677 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 678 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 679 return; 680 681 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 682 delegated = this_cpu_ptr(&mptcp_delegated_actions); 683 schedule = list_empty(&delegated->head); 684 list_add_tail(&subflow->delegated_node, &delegated->head); 685 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 686 sock_hold(mptcp_subflow_tcp_sock(subflow)); 687 if (schedule) 688 napi_schedule(&delegated->napi); 689 } 690 } 691 692 static inline struct mptcp_subflow_context * 693 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 694 { 695 struct mptcp_subflow_context *ret; 696 697 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 698 if (list_empty(&delegated->head)) { 699 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 700 return NULL; 701 } 702 703 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 704 list_del_init(&ret->delegated_node); 705 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 706 return ret; 707 } 708 709 int mptcp_is_enabled(const struct net *net); 710 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 711 int mptcp_is_checksum_enabled(const struct net *net); 712 int mptcp_allow_join_id0(const struct net *net); 713 unsigned int mptcp_stale_loss_cnt(const struct net *net); 714 unsigned int mptcp_close_timeout(const struct sock *sk); 715 int mptcp_get_pm_type(const struct net *net); 716 const char *mptcp_get_path_manager(const struct net *net); 717 const char *mptcp_get_scheduler(const struct net *net); 718 719 void mptcp_active_disable(struct sock *sk); 720 bool mptcp_active_should_disable(struct sock *ssk); 721 void mptcp_active_enable(struct sock *sk); 722 723 void mptcp_get_available_schedulers(char *buf, size_t maxlen); 724 void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 725 struct mptcp_subflow_context *subflow, 726 const struct mptcp_options_received *mp_opt); 727 bool __mptcp_retransmit_pending_data(struct sock *sk); 728 void mptcp_check_and_set_pending(struct sock *sk); 729 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 730 bool mptcp_subflow_data_available(struct sock *sk); 731 void __init mptcp_subflow_init(void); 732 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 733 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 734 struct mptcp_subflow_context *subflow); 735 void __mptcp_subflow_send_ack(struct sock *ssk); 736 void mptcp_subflow_reset(struct sock *ssk); 737 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 738 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 739 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 740 bool __mptcp_close(struct sock *sk, long timeout); 741 void mptcp_cancel_work(struct sock *sk); 742 void __mptcp_unaccepted_force_close(struct sock *sk); 743 void mptcp_set_state(struct sock *sk, int state); 744 745 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 746 const struct mptcp_addr_info *b, bool use_port); 747 void mptcp_local_address(const struct sock_common *skc, 748 struct mptcp_addr_info *addr); 749 void mptcp_remote_address(const struct sock_common *skc, 750 struct mptcp_addr_info *addr); 751 752 /* called with sk socket lock held */ 753 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local, 754 const struct mptcp_addr_info *remote); 755 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 756 struct socket **new_sock); 757 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 758 struct sockaddr_storage *addr, 759 unsigned short family); 760 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 761 int mptcp_validate_scheduler(struct mptcp_sched_ops *sched); 762 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 763 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 764 void mptcp_sched_init(void); 765 int mptcp_init_sched(struct mptcp_sock *msk, 766 struct mptcp_sched_ops *sched); 767 void mptcp_release_sched(struct mptcp_sock *msk); 768 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 769 bool scheduled); 770 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 771 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 772 int mptcp_sched_get_send(struct mptcp_sock *msk); 773 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 774 775 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 776 { 777 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 778 } 779 780 static inline bool mptcp_epollin_ready(const struct sock *sk) 781 { 782 u64 data_avail = mptcp_data_avail(mptcp_sk(sk)); 783 784 if (!data_avail) 785 return false; 786 787 /* mptcp doesn't have to deal with small skbs in the receive queue, 788 * as it can always coalesce them 789 */ 790 return (data_avail >= sk->sk_rcvlowat) || 791 (mem_cgroup_sockets_enabled && sk->sk_memcg && 792 mem_cgroup_under_socket_pressure(sk->sk_memcg)) || 793 READ_ONCE(tcp_memory_pressure); 794 } 795 796 int mptcp_set_rcvlowat(struct sock *sk, int val); 797 798 static inline bool __tcp_can_send(const struct sock *ssk) 799 { 800 /* only send if our side has not closed yet */ 801 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 802 } 803 804 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 805 { 806 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 807 if (subflow->request_join && !READ_ONCE(subflow->fully_established)) 808 return false; 809 810 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 811 } 812 813 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 814 815 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 816 817 void mptcp_subflow_drop_ctx(struct sock *ssk); 818 819 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 820 struct mptcp_subflow_context *ctx) 821 { 822 sk->sk_data_ready = sock_def_readable; 823 sk->sk_state_change = ctx->tcp_state_change; 824 sk->sk_write_space = sk_stream_write_space; 825 sk->sk_error_report = ctx->tcp_error_report; 826 827 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 828 } 829 830 void __init mptcp_proto_init(void); 831 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 832 int __init mptcp_proto_v6_init(void); 833 #endif 834 835 struct sock *mptcp_sk_clone_init(const struct sock *sk, 836 const struct mptcp_options_received *mp_opt, 837 struct sock *ssk, 838 struct request_sock *req); 839 void mptcp_get_options(const struct sk_buff *skb, 840 struct mptcp_options_received *mp_opt); 841 842 void mptcp_finish_connect(struct sock *sk); 843 void __mptcp_sync_state(struct sock *sk, int state); 844 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 845 846 static inline void mptcp_stop_tout_timer(struct sock *sk) 847 { 848 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 849 return; 850 851 sk_stop_timer(sk, &sk->sk_timer); 852 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 853 } 854 855 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 856 { 857 /* avoid 0 timestamp, as that means no close timeout */ 858 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 859 } 860 861 static inline void mptcp_start_tout_timer(struct sock *sk) 862 { 863 mptcp_set_close_tout(sk, tcp_jiffies32); 864 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 865 } 866 867 static inline bool mptcp_is_fully_established(struct sock *sk) 868 { 869 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 870 READ_ONCE(mptcp_sk(sk)->fully_established); 871 } 872 873 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 874 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 875 bool mptcp_finish_join(struct sock *sk); 876 bool mptcp_schedule_work(struct sock *sk); 877 int mptcp_setsockopt(struct sock *sk, int level, int optname, 878 sockptr_t optval, unsigned int optlen); 879 int mptcp_getsockopt(struct sock *sk, int level, int optname, 880 char __user *optval, int __user *option); 881 882 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 883 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 884 { 885 if (use_64bit) 886 return cur_seq; 887 888 return __mptcp_expand_seq(old_seq, cur_seq); 889 } 890 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 891 void __mptcp_data_acked(struct sock *sk); 892 void __mptcp_error_report(struct sock *sk); 893 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 894 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 895 { 896 return READ_ONCE(msk->snd_data_fin_enable) && 897 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 898 } 899 900 static inline u32 mptcp_notsent_lowat(const struct sock *sk) 901 { 902 struct net *net = sock_net(sk); 903 u32 val; 904 905 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat); 906 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat); 907 } 908 909 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake) 910 { 911 const struct mptcp_sock *msk = mptcp_sk(sk); 912 u32 notsent_bytes; 913 914 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt); 915 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk); 916 } 917 918 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake) 919 { 920 return mptcp_stream_memory_free(sk, wake) && 921 __sk_stream_is_writeable(sk, wake); 922 } 923 924 static inline void mptcp_write_space(struct sock *sk) 925 { 926 /* pairs with memory barrier in mptcp_poll */ 927 smp_mb(); 928 if (mptcp_stream_memory_free(sk, 1)) 929 sk_stream_write_space(sk); 930 } 931 932 static inline void __mptcp_sync_sndbuf(struct sock *sk) 933 { 934 struct mptcp_subflow_context *subflow; 935 int ssk_sndbuf, new_sndbuf; 936 937 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 938 return; 939 940 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]); 941 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 942 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 943 944 subflow->cached_sndbuf = ssk_sndbuf; 945 new_sndbuf += ssk_sndbuf; 946 } 947 948 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 949 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 950 mptcp_write_space(sk); 951 } 952 953 /* The called held both the msk socket and the subflow socket locks, 954 * possibly under BH 955 */ 956 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 957 { 958 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 959 960 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 961 __mptcp_sync_sndbuf(sk); 962 } 963 964 /* the caller held only the subflow socket lock, either in process or 965 * BH context. Additionally this can be called under the msk data lock, 966 * so we can't acquire such lock here: let the delegate action acquires 967 * the needed locks in suitable order. 968 */ 969 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 970 { 971 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 972 973 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 974 return; 975 976 local_bh_disable(); 977 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 978 local_bh_enable(); 979 } 980 981 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 982 983 #define MPTCP_TOKEN_MAX_RETRIES 4 984 985 void __init mptcp_token_init(void); 986 static inline void mptcp_token_init_request(struct request_sock *req) 987 { 988 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 989 } 990 991 int mptcp_token_new_request(struct request_sock *req); 992 void mptcp_token_destroy_request(struct request_sock *req); 993 int mptcp_token_new_connect(struct sock *ssk); 994 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 995 struct mptcp_sock *msk); 996 bool mptcp_token_exists(u32 token); 997 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 998 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 999 long *s_num); 1000 void mptcp_token_destroy(struct mptcp_sock *msk); 1001 1002 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 1003 1004 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 1005 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 1006 1007 void __init mptcp_pm_init(void); 1008 void mptcp_pm_data_init(struct mptcp_sock *msk); 1009 void mptcp_pm_data_reset(struct mptcp_sock *msk); 1010 void mptcp_pm_destroy(struct mptcp_sock *msk); 1011 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 1012 struct mptcp_addr_info *addr); 1013 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 1014 bool require_family, 1015 struct mptcp_pm_addr_entry *entry); 1016 bool mptcp_pm_addr_families_match(const struct sock *sk, 1017 const struct mptcp_addr_info *loc, 1018 const struct mptcp_addr_info *rem); 1019 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 1020 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 1021 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 1022 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 1023 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 1024 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 1025 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 1026 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 1027 const struct mptcp_subflow_context *subflow); 1028 void mptcp_pm_add_addr_received(const struct sock *ssk, 1029 const struct mptcp_addr_info *addr); 1030 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 1031 const struct mptcp_addr_info *addr); 1032 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 1033 void mptcp_pm_send_ack(struct mptcp_sock *msk, 1034 struct mptcp_subflow_context *subflow, 1035 bool prio, bool backup); 1036 void mptcp_pm_addr_send_ack(struct mptcp_sock *msk); 1037 void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id); 1038 void mptcp_pm_rm_subflow(struct mptcp_sock *msk, 1039 const struct mptcp_rm_list *rm_list); 1040 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 1041 const struct mptcp_rm_list *rm_list); 1042 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 1043 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 1044 int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk, 1045 struct mptcp_addr_info *addr, 1046 struct mptcp_addr_info *rem, 1047 u8 bkup); 1048 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 1049 const struct mptcp_addr_info *addr); 1050 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 1051 struct mptcp_pm_add_entry * 1052 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 1053 const struct mptcp_addr_info *addr, bool check_id); 1054 bool mptcp_lookup_subflow_by_saddr(const struct list_head *list, 1055 const struct mptcp_addr_info *saddr); 1056 bool mptcp_remove_anno_list_by_saddr(struct mptcp_sock *msk, 1057 const struct mptcp_addr_info *addr); 1058 int mptcp_pm_nl_set_flags(struct mptcp_pm_addr_entry *local, 1059 struct genl_info *info); 1060 int mptcp_userspace_pm_set_flags(struct mptcp_pm_addr_entry *local, 1061 struct genl_info *info); 1062 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 1063 const struct mptcp_addr_info *addr, 1064 bool echo); 1065 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 1066 void mptcp_pm_remove_addr_entry(struct mptcp_sock *msk, 1067 struct mptcp_pm_addr_entry *entry); 1068 1069 /* the default path manager, used in mptcp_pm_unregister */ 1070 extern struct mptcp_pm_ops mptcp_pm_kernel; 1071 1072 struct mptcp_pm_ops *mptcp_pm_find(const char *name); 1073 int mptcp_pm_register(struct mptcp_pm_ops *pm_ops); 1074 void mptcp_pm_unregister(struct mptcp_pm_ops *pm_ops); 1075 int mptcp_pm_validate(struct mptcp_pm_ops *pm_ops); 1076 void mptcp_pm_get_available(char *buf, size_t maxlen); 1077 1078 void mptcp_userspace_pm_free_local_addr_list(struct mptcp_sock *msk); 1079 1080 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 1081 const struct sock *ssk, gfp_t gfp); 1082 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 1083 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 1084 void mptcp_event_pm_listener(const struct sock *ssk, 1085 enum mptcp_event_type event); 1086 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 1087 1088 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 1089 struct request_sock *req); 1090 int mptcp_pm_genl_fill_addr(struct sk_buff *msg, 1091 struct netlink_callback *cb, 1092 struct mptcp_pm_addr_entry *entry); 1093 1094 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 1095 { 1096 return READ_ONCE(msk->pm.addr_signal) & 1097 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 1098 } 1099 1100 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 1101 { 1102 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 1103 } 1104 1105 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 1106 { 1107 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 1108 } 1109 1110 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 1111 { 1112 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 1113 } 1114 1115 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1116 { 1117 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1118 } 1119 1120 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1121 { 1122 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1123 } 1124 1125 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 1126 { 1127 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 1128 1129 if (family == AF_INET6) 1130 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 1131 if (!echo) 1132 len += MPTCPOPT_THMAC_LEN; 1133 /* account for 2 trailing 'nop' options */ 1134 if (port) 1135 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 1136 1137 return len; 1138 } 1139 1140 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 1141 { 1142 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 1143 return -EINVAL; 1144 1145 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 1146 } 1147 1148 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 1149 unsigned int opt_size, unsigned int remaining, 1150 struct mptcp_addr_info *addr, bool *echo, 1151 bool *drop_other_suboptions); 1152 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1153 struct mptcp_rm_list *rm_list); 1154 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1155 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, 1156 struct mptcp_pm_addr_entry *skc); 1157 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, 1158 struct mptcp_pm_addr_entry *skc); 1159 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc); 1160 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1161 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1162 int mptcp_pm_nl_dump_addr(struct sk_buff *msg, 1163 struct netlink_callback *cb); 1164 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg, 1165 struct netlink_callback *cb); 1166 int mptcp_pm_nl_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1167 struct genl_info *info); 1168 int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1169 struct genl_info *info); 1170 1171 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1172 { 1173 int local_id = READ_ONCE(subflow->local_id); 1174 1175 if (local_id < 0) 1176 return 0; 1177 return local_id; 1178 } 1179 1180 void __init mptcp_pm_kernel_register(void); 1181 void __init mptcp_pm_userspace_register(void); 1182 void __init mptcp_pm_nl_init(void); 1183 void mptcp_pm_worker(struct mptcp_sock *msk); 1184 void __mptcp_pm_kernel_worker(struct mptcp_sock *msk); 1185 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 1186 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 1187 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 1188 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 1189 1190 /* called under PM lock */ 1191 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1192 { 1193 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 1194 WRITE_ONCE(msk->pm.accept_subflow, true); 1195 } 1196 1197 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1198 { 1199 spin_lock_bh(&msk->pm.lock); 1200 __mptcp_pm_close_subflow(msk); 1201 spin_unlock_bh(&msk->pm.lock); 1202 } 1203 1204 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1205 1206 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1207 { 1208 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1209 } 1210 1211 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1212 1213 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1214 { 1215 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1216 } 1217 1218 static inline bool mptcp_check_fallback(const struct sock *sk) 1219 { 1220 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1221 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1222 1223 return __mptcp_check_fallback(msk); 1224 } 1225 1226 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1227 { 1228 struct sock *ssk = READ_ONCE(msk->first); 1229 1230 return ssk && ((1 << inet_sk_state_load(ssk)) & 1231 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1232 TCPF_SYN_RECV | TCPF_LISTEN)); 1233 } 1234 1235 bool __mptcp_try_fallback(struct mptcp_sock *msk, int fb_mib); 1236 1237 static inline bool mptcp_try_fallback(struct sock *ssk, int fb_mib) 1238 { 1239 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1240 struct sock *sk = subflow->conn; 1241 struct mptcp_sock *msk; 1242 1243 msk = mptcp_sk(sk); 1244 if (!__mptcp_try_fallback(msk, fb_mib)) 1245 return false; 1246 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1247 gfp_t saved_allocation = ssk->sk_allocation; 1248 1249 /* we are in a atomic (BH) scope, override ssk default for data 1250 * fin allocation 1251 */ 1252 ssk->sk_allocation = GFP_ATOMIC; 1253 ssk->sk_shutdown |= SEND_SHUTDOWN; 1254 tcp_shutdown(ssk, SEND_SHUTDOWN); 1255 ssk->sk_allocation = saved_allocation; 1256 } 1257 return true; 1258 } 1259 1260 static inline void mptcp_early_fallback(struct mptcp_sock *msk, 1261 struct mptcp_subflow_context *subflow, 1262 int fb_mib) 1263 { 1264 subflow->request_mptcp = 0; 1265 WARN_ON_ONCE(!__mptcp_try_fallback(msk, fb_mib)); 1266 } 1267 1268 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1269 { 1270 struct mptcp_ext *mpext; 1271 1272 mpext = skb ? mptcp_get_ext(skb) : NULL; 1273 if (mpext && mpext->infinite_map) 1274 return true; 1275 1276 return false; 1277 } 1278 1279 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1280 { 1281 return (subflow->request_mptcp || subflow->request_join); 1282 } 1283 1284 static inline bool subflow_simultaneous_connect(struct sock *sk) 1285 { 1286 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1287 1288 return (1 << sk->sk_state) & 1289 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) && 1290 is_active_ssk(subflow) && 1291 !subflow->conn_finished; 1292 } 1293 1294 #ifdef CONFIG_SYN_COOKIES 1295 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1296 struct sk_buff *skb); 1297 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1298 struct sk_buff *skb); 1299 void __init mptcp_join_cookie_init(void); 1300 #else 1301 static inline void 1302 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1303 struct sk_buff *skb) {} 1304 static inline bool 1305 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1306 struct sk_buff *skb) 1307 { 1308 return false; 1309 } 1310 1311 static inline void mptcp_join_cookie_init(void) {} 1312 #endif 1313 1314 #endif /* __MPTCP_PROTOCOL_H */ 1315