1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* Multipath TCP 3 * 4 * Copyright (c) 2017 - 2019, Intel Corporation. 5 */ 6 7 #ifndef __MPTCP_PROTOCOL_H 8 #define __MPTCP_PROTOCOL_H 9 10 #include <linux/random.h> 11 #include <net/tcp.h> 12 #include <net/inet_connection_sock.h> 13 #include <uapi/linux/mptcp.h> 14 #include <net/genetlink.h> 15 #include <net/rstreason.h> 16 17 #define MPTCP_SUPPORTED_VERSION 1 18 19 /* MPTCP option bits */ 20 #define OPTION_MPTCP_MPC_SYN BIT(0) 21 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 22 #define OPTION_MPTCP_MPC_ACK BIT(2) 23 #define OPTION_MPTCP_MPJ_SYN BIT(3) 24 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 25 #define OPTION_MPTCP_MPJ_ACK BIT(5) 26 #define OPTION_MPTCP_ADD_ADDR BIT(6) 27 #define OPTION_MPTCP_RM_ADDR BIT(7) 28 #define OPTION_MPTCP_FASTCLOSE BIT(8) 29 #define OPTION_MPTCP_PRIO BIT(9) 30 #define OPTION_MPTCP_RST BIT(10) 31 #define OPTION_MPTCP_DSS BIT(11) 32 #define OPTION_MPTCP_FAIL BIT(12) 33 34 #define OPTION_MPTCP_CSUMREQD BIT(13) 35 36 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \ 37 OPTION_MPTCP_MPC_ACK) 38 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \ 39 OPTION_MPTCP_MPJ_ACK) 40 41 /* MPTCP option subtypes */ 42 #define MPTCPOPT_MP_CAPABLE 0 43 #define MPTCPOPT_MP_JOIN 1 44 #define MPTCPOPT_DSS 2 45 #define MPTCPOPT_ADD_ADDR 3 46 #define MPTCPOPT_RM_ADDR 4 47 #define MPTCPOPT_MP_PRIO 5 48 #define MPTCPOPT_MP_FAIL 6 49 #define MPTCPOPT_MP_FASTCLOSE 7 50 #define MPTCPOPT_RST 8 51 52 /* MPTCP suboption lengths */ 53 #define TCPOLEN_MPTCP_MPC_SYN 4 54 #define TCPOLEN_MPTCP_MPC_SYNACK 12 55 #define TCPOLEN_MPTCP_MPC_ACK 20 56 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 57 #define TCPOLEN_MPTCP_MPJ_SYN 12 58 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 59 #define TCPOLEN_MPTCP_MPJ_ACK 24 60 #define TCPOLEN_MPTCP_DSS_BASE 4 61 #define TCPOLEN_MPTCP_DSS_ACK32 4 62 #define TCPOLEN_MPTCP_DSS_ACK64 8 63 #define TCPOLEN_MPTCP_DSS_MAP32 10 64 #define TCPOLEN_MPTCP_DSS_MAP64 14 65 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 66 #define TCPOLEN_MPTCP_ADD_ADDR 16 67 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 70 #define TCPOLEN_MPTCP_ADD_ADDR6 28 71 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 74 #define TCPOLEN_MPTCP_PORT_LEN 2 75 #define TCPOLEN_MPTCP_PORT_ALIGN 2 76 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 77 #define TCPOLEN_MPTCP_PRIO 3 78 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 79 #define TCPOLEN_MPTCP_FASTCLOSE 12 80 #define TCPOLEN_MPTCP_RST 4 81 #define TCPOLEN_MPTCP_FAIL 12 82 83 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 84 85 /* MPTCP MP_JOIN flags */ 86 #define MPTCPOPT_BACKUP BIT(0) 87 #define MPTCPOPT_THMAC_LEN 8 88 89 /* MPTCP MP_CAPABLE flags */ 90 #define MPTCP_VERSION_MASK (0x0F) 91 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 92 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 93 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 94 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 95 #define MPTCP_CAP_FLAG_MASK (0x1F) 96 97 /* MPTCP DSS flags */ 98 #define MPTCP_DSS_DATA_FIN BIT(4) 99 #define MPTCP_DSS_DSN64 BIT(3) 100 #define MPTCP_DSS_HAS_MAP BIT(2) 101 #define MPTCP_DSS_ACK64 BIT(1) 102 #define MPTCP_DSS_HAS_ACK BIT(0) 103 #define MPTCP_DSS_FLAG_MASK (0x1F) 104 105 /* MPTCP ADD_ADDR flags */ 106 #define MPTCP_ADDR_ECHO BIT(0) 107 108 /* MPTCP MP_PRIO flags */ 109 #define MPTCP_PRIO_BKUP BIT(0) 110 111 /* MPTCP TCPRST flags */ 112 #define MPTCP_RST_TRANSIENT BIT(0) 113 114 /* MPTCP socket atomic flags */ 115 #define MPTCP_WORK_RTX 1 116 #define MPTCP_FALLBACK_DONE 2 117 #define MPTCP_WORK_CLOSE_SUBFLOW 3 118 119 /* MPTCP socket release cb flags */ 120 #define MPTCP_PUSH_PENDING 1 121 #define MPTCP_CLEAN_UNA 2 122 #define MPTCP_ERROR_REPORT 3 123 #define MPTCP_RETRANSMIT 4 124 #define MPTCP_FLUSH_JOIN_LIST 5 125 #define MPTCP_SYNC_STATE 6 126 #define MPTCP_SYNC_SNDBUF 7 127 128 struct mptcp_skb_cb { 129 u64 map_seq; 130 u64 end_seq; 131 u32 offset; 132 u8 has_rxtstamp; 133 u8 cant_coalesce; 134 }; 135 136 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0])) 137 138 static inline bool before64(__u64 seq1, __u64 seq2) 139 { 140 return (__s64)(seq1 - seq2) < 0; 141 } 142 143 #define after64(seq2, seq1) before64(seq1, seq2) 144 145 struct mptcp_options_received { 146 u64 sndr_key; 147 u64 rcvr_key; 148 u64 data_ack; 149 u64 data_seq; 150 u32 subflow_seq; 151 u16 data_len; 152 __sum16 csum; 153 struct_group(status, 154 u16 suboptions; 155 u16 use_map:1, 156 dsn64:1, 157 data_fin:1, 158 use_ack:1, 159 ack64:1, 160 mpc_map:1, 161 reset_reason:4, 162 reset_transient:1, 163 echo:1, 164 backup:1, 165 deny_join_id0:1, 166 __unused:2; 167 ); 168 u8 join_id; 169 u32 token; 170 u32 nonce; 171 u64 thmac; 172 u8 hmac[MPTCPOPT_HMAC_LEN]; 173 struct mptcp_addr_info addr; 174 struct mptcp_rm_list rm_list; 175 u64 ahmac; 176 u64 fail_seq; 177 }; 178 179 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 180 { 181 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 182 ((nib & 0xF) << 8) | field); 183 } 184 185 enum mptcp_pm_status { 186 MPTCP_PM_ADD_ADDR_RECEIVED, 187 MPTCP_PM_ADD_ADDR_SEND_ACK, 188 MPTCP_PM_RM_ADDR_RECEIVED, 189 MPTCP_PM_ESTABLISHED, 190 MPTCP_PM_SUBFLOW_ESTABLISHED, 191 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 192 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 193 * accounted int id_avail_bitmap 194 */ 195 }; 196 197 enum mptcp_pm_type { 198 MPTCP_PM_TYPE_KERNEL = 0, 199 MPTCP_PM_TYPE_USERSPACE, 200 201 __MPTCP_PM_TYPE_NR, 202 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 203 }; 204 205 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 206 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 207 208 enum mptcp_addr_signal_status { 209 MPTCP_ADD_ADDR_SIGNAL, 210 MPTCP_ADD_ADDR_ECHO, 211 MPTCP_RM_ADDR_SIGNAL, 212 }; 213 214 /* max value of mptcp_addr_info.id */ 215 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 216 217 struct mptcp_pm_data { 218 struct mptcp_addr_info local; 219 struct mptcp_addr_info remote; 220 struct list_head anno_list; 221 struct list_head userspace_pm_local_addr_list; 222 223 spinlock_t lock; /*protects the whole PM data */ 224 225 struct_group(reset, 226 227 u8 addr_signal; 228 bool server_side; 229 bool work_pending; 230 bool accept_addr; 231 bool accept_subflow; 232 bool remote_deny_join_id0; 233 u8 add_addr_signaled; 234 u8 add_addr_accepted; 235 u8 local_addr_used; 236 u8 pm_type; 237 u8 extra_subflows; 238 u8 status; 239 240 ); 241 242 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 243 struct mptcp_rm_list rm_list_tx; 244 struct mptcp_rm_list rm_list_rx; 245 }; 246 247 struct mptcp_pm_local { 248 struct mptcp_addr_info addr; 249 u32 flags; 250 int ifindex; 251 }; 252 253 struct mptcp_pm_addr_entry { 254 struct list_head list; 255 struct mptcp_addr_info addr; 256 u32 flags; 257 int ifindex; 258 struct socket *lsk; 259 }; 260 261 struct mptcp_data_frag { 262 struct list_head list; 263 u64 data_seq; 264 u16 data_len; 265 u16 offset; 266 u8 overhead; 267 u8 eor; /* currently using 1 bit */ 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 fastclosing:1; 326 u32 notsent_lowat; 327 int keepalive_cnt; 328 int keepalive_idle; 329 int keepalive_intvl; 330 int maxseg; 331 struct work_struct work; 332 struct sk_buff *ooo_last_skb; 333 struct rb_root out_of_order_queue; 334 struct list_head conn_list; 335 struct list_head rtx_queue; 336 struct mptcp_data_frag *first_pending; 337 struct list_head join_list; 338 struct sock *first; /* The mptcp ops can safely dereference, using suitable 339 * ONCE annotation, the subflow outside the socket 340 * lock as such sock is freed after close(). 341 */ 342 struct mptcp_pm_data pm; 343 struct mptcp_sched_ops *sched; 344 struct { 345 int space; /* bytes copied in last measurement window */ 346 int copied; /* bytes copied in this measurement window */ 347 u64 time; /* start time of measurement window */ 348 u64 rtt_us; /* last maximum rtt of subflows */ 349 } rcvq_space; 350 u8 scaling_ratio; 351 bool allow_subflows; 352 353 u32 subflow_id; 354 u32 setsockopt_seq; 355 char ca_name[TCP_CA_NAME_MAX]; 356 357 spinlock_t fallback_lock; /* protects fallback, 358 * allow_infinite_fallback and 359 * allow_join 360 */ 361 362 struct list_head backlog_list; /* protected by the data lock */ 363 u32 backlog_len; 364 u32 backlog_unaccounted; 365 }; 366 367 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 368 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 369 370 #define mptcp_for_each_subflow(__msk, __subflow) \ 371 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 372 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 373 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 374 #define mptcp_next_subflow(__msk, __subflow) \ 375 list_next_entry_circular(__subflow, &((__msk)->conn_list), node) 376 377 extern struct genl_family mptcp_genl_family; 378 379 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 380 { 381 sock_owned_by_me((const struct sock *)msk); 382 } 383 384 #ifdef CONFIG_DEBUG_NET 385 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 386 #undef tcp_sk 387 #define tcp_sk(ptr) ({ \ 388 typeof(ptr) _ptr = (ptr); \ 389 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 390 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 391 }) 392 #define mptcp_sk(ptr) ({ \ 393 typeof(ptr) _ptr = (ptr); \ 394 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 395 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 396 }) 397 398 #else /* !CONFIG_DEBUG_NET */ 399 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 400 #endif 401 402 static inline int mptcp_win_from_space(const struct sock *sk, int space) 403 { 404 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 405 } 406 407 static inline int mptcp_space_from_win(const struct sock *sk, int win) 408 { 409 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win); 410 } 411 412 static inline int __mptcp_space(const struct sock *sk) 413 { 414 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - 415 READ_ONCE(mptcp_sk(sk)->backlog_len) - 416 sk_rmem_alloc_get(sk)); 417 } 418 419 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 420 { 421 const struct mptcp_sock *msk = mptcp_sk(sk); 422 423 return msk->first_pending; 424 } 425 426 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 427 { 428 struct mptcp_sock *msk = mptcp_sk(sk); 429 struct mptcp_data_frag *cur; 430 431 cur = msk->first_pending; 432 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 433 list_next_entry(cur, list); 434 } 435 436 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 437 { 438 const struct mptcp_sock *msk = mptcp_sk(sk); 439 440 if (!msk->first_pending) 441 return NULL; 442 443 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 444 return NULL; 445 446 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 447 } 448 449 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 450 { 451 struct mptcp_sock *msk = mptcp_sk(sk); 452 453 if (msk->snd_una == msk->snd_nxt) 454 return NULL; 455 456 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 457 } 458 459 struct csum_pseudo_header { 460 __be64 data_seq; 461 __be32 subflow_seq; 462 __be16 data_len; 463 __sum16 csum; 464 }; 465 466 struct mptcp_subflow_request_sock { 467 struct tcp_request_sock sk; 468 u16 mp_capable : 1, 469 mp_join : 1, 470 backup : 1, 471 request_bkup : 1, 472 csum_reqd : 1, 473 allow_join_id0 : 1; 474 u8 local_id; 475 u8 remote_id; 476 u64 local_key; 477 u64 idsn; 478 u32 token; 479 u32 ssn_offset; 480 u64 thmac; 481 u32 local_nonce; 482 u32 remote_nonce; 483 struct mptcp_sock *msk; 484 struct hlist_nulls_node token_node; 485 }; 486 487 static inline struct mptcp_subflow_request_sock * 488 mptcp_subflow_rsk(const struct request_sock *rsk) 489 { 490 return (struct mptcp_subflow_request_sock *)rsk; 491 } 492 493 struct mptcp_delegated_action { 494 struct napi_struct napi; 495 local_lock_t bh_lock; 496 struct list_head head; 497 }; 498 499 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 500 501 #define MPTCP_DELEGATE_SCHEDULED 0 502 #define MPTCP_DELEGATE_SEND 1 503 #define MPTCP_DELEGATE_ACK 2 504 #define MPTCP_DELEGATE_SNDBUF 3 505 506 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 507 /* MPTCP subflow context */ 508 struct mptcp_subflow_context { 509 struct list_head node;/* conn_list of subflows */ 510 511 struct_group(reset, 512 513 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 514 u64 local_key; 515 u64 remote_key; 516 u64 idsn; 517 u64 map_seq; 518 u64 rcv_wnd_sent; 519 u32 snd_isn; 520 u32 token; 521 u32 rel_write_seq; 522 u32 map_subflow_seq; 523 u32 ssn_offset; 524 u32 map_data_len; 525 __wsum map_data_csum; 526 u32 map_csum_len; 527 u32 request_mptcp : 1, /* send MP_CAPABLE */ 528 request_join : 1, /* send MP_JOIN */ 529 request_bkup : 1, 530 mp_capable : 1, /* remote is MPTCP capable */ 531 mp_join : 1, /* remote is JOINing */ 532 pm_notified : 1, /* PM hook called for established status */ 533 conn_finished : 1, 534 map_valid : 1, 535 map_csum_reqd : 1, 536 map_data_fin : 1, 537 mpc_map : 1, 538 backup : 1, 539 send_mp_prio : 1, 540 send_mp_fail : 1, 541 send_fastclose : 1, 542 send_infinite_map : 1, 543 remote_key_valid : 1, /* received the peer key from */ 544 disposable : 1, /* ctx can be free at ulp release time */ 545 closing : 1, /* must not pass rx data to msk anymore */ 546 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 547 valid_csum_seen : 1, /* at least one csum validated */ 548 is_mptfo : 1, /* subflow is doing TFO */ 549 close_event_done : 1, /* has done the post-closed part */ 550 mpc_drop : 1, /* the MPC option has been dropped in a rtx */ 551 __unused : 8; 552 bool data_avail; 553 bool scheduled; 554 bool pm_listener; /* a listener managed by the kernel PM? */ 555 bool fully_established; /* path validated */ 556 u32 lent_mem_frag; 557 u32 remote_nonce; 558 u64 thmac; 559 u32 local_nonce; 560 u32 remote_token; 561 union { 562 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 563 u64 iasn; /* initial ack sequence number, MPC subflows only */ 564 }; 565 s16 local_id; /* if negative not initialized yet */ 566 u8 remote_id; 567 u8 reset_seen:1; 568 u8 reset_transient:1; 569 u8 reset_reason:4; 570 u8 stale_count; 571 572 u32 subflow_id; 573 574 long delegated_status; 575 unsigned long fail_tout; 576 577 ); 578 579 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 580 581 u32 setsockopt_seq; 582 u32 stale_rcv_tstamp; 583 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 584 * protected by the msk socket lock 585 */ 586 587 struct sock *tcp_sock; /* tcp sk backpointer */ 588 struct sock *conn; /* parent mptcp_sock */ 589 const struct inet_connection_sock_af_ops *icsk_af_ops; 590 void (*tcp_state_change)(struct sock *sk); 591 void (*tcp_error_report)(struct sock *sk); 592 593 struct rcu_head rcu; 594 }; 595 596 static inline struct mptcp_subflow_context * 597 mptcp_subflow_ctx(const struct sock *sk) 598 { 599 const struct inet_connection_sock *icsk = inet_csk(sk); 600 601 /* Use RCU on icsk_ulp_data only for sock diag code */ 602 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 603 } 604 605 static inline struct sock * 606 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 607 { 608 return subflow->tcp_sock; 609 } 610 611 static inline void 612 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 613 { 614 memset(&subflow->reset, 0, sizeof(subflow->reset)); 615 subflow->request_mptcp = 1; 616 WRITE_ONCE(subflow->local_id, -1); 617 } 618 619 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */ 620 static inline enum sk_rst_reason 621 sk_rst_convert_mptcp_reason(u32 reason) 622 { 623 switch (reason) { 624 case MPTCP_RST_EUNSPEC: 625 return SK_RST_REASON_MPTCP_RST_EUNSPEC; 626 case MPTCP_RST_EMPTCP: 627 return SK_RST_REASON_MPTCP_RST_EMPTCP; 628 case MPTCP_RST_ERESOURCE: 629 return SK_RST_REASON_MPTCP_RST_ERESOURCE; 630 case MPTCP_RST_EPROHIBIT: 631 return SK_RST_REASON_MPTCP_RST_EPROHIBIT; 632 case MPTCP_RST_EWQ2BIG: 633 return SK_RST_REASON_MPTCP_RST_EWQ2BIG; 634 case MPTCP_RST_EBADPERF: 635 return SK_RST_REASON_MPTCP_RST_EBADPERF; 636 case MPTCP_RST_EMIDDLEBOX: 637 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX; 638 default: 639 /* It should not happen, or else errors may occur 640 * in MPTCP layer 641 */ 642 return SK_RST_REASON_ERROR; 643 } 644 } 645 646 static inline void 647 mptcp_send_active_reset_reason(struct sock *sk) 648 { 649 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 650 enum sk_rst_reason reason; 651 652 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason); 653 tcp_send_active_reset(sk, GFP_ATOMIC, reason); 654 } 655 656 /* Made the fwd mem carried by the given skb available to the msk, 657 * To be paired with a previous mptcp_subflow_lend_fwdmem() before freeing 658 * the skb or setting the skb ownership. 659 */ 660 static inline void mptcp_borrow_fwdmem(struct sock *sk, struct sk_buff *skb) 661 { 662 struct sock *ssk = skb->sk; 663 664 /* The subflow just lend the skb fwd memory; if the subflow meanwhile 665 * closed, mptcp_close_ssk() already released the ssk rcv memory. 666 */ 667 DEBUG_NET_WARN_ON_ONCE(skb->destructor); 668 sk_forward_alloc_add(sk, skb->truesize); 669 if (!ssk) 670 return; 671 672 atomic_sub(skb->truesize, &ssk->sk_rmem_alloc); 673 skb->sk = NULL; 674 } 675 676 static inline void 677 __mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow, int size) 678 { 679 int frag = (subflow->lent_mem_frag + size) & (PAGE_SIZE - 1); 680 681 subflow->lent_mem_frag = frag; 682 } 683 684 static inline void 685 mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow, 686 struct sk_buff *skb) 687 { 688 __mptcp_subflow_lend_fwdmem(subflow, skb->truesize); 689 skb->destructor = NULL; 690 } 691 692 static inline u64 693 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 694 { 695 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 696 subflow->ssn_offset - 697 subflow->map_subflow_seq; 698 } 699 700 static inline u64 701 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 702 { 703 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 704 } 705 706 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 707 708 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 709 { 710 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 711 struct mptcp_delegated_action *delegated; 712 bool schedule; 713 714 /* the caller held the subflow bh socket lock */ 715 lockdep_assert_in_softirq(); 716 717 /* The implied barrier pairs with tcp_release_cb_override() 718 * mptcp_napi_poll(), and ensures the below list check sees list 719 * updates done prior to delegated status bits changes 720 */ 721 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 722 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 723 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 724 return; 725 726 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 727 delegated = this_cpu_ptr(&mptcp_delegated_actions); 728 schedule = list_empty(&delegated->head); 729 list_add_tail(&subflow->delegated_node, &delegated->head); 730 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 731 sock_hold(mptcp_subflow_tcp_sock(subflow)); 732 if (schedule) 733 napi_schedule(&delegated->napi); 734 } 735 } 736 737 static inline struct mptcp_subflow_context * 738 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 739 { 740 struct mptcp_subflow_context *ret; 741 742 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 743 if (list_empty(&delegated->head)) { 744 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 745 return NULL; 746 } 747 748 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 749 list_del_init(&ret->delegated_node); 750 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 751 return ret; 752 } 753 754 void __mptcp_inherit_memcg(struct sock *sk, struct sock *ssk, gfp_t gfp); 755 void __mptcp_inherit_cgrp_data(struct sock *sk, struct sock *ssk); 756 757 int mptcp_is_enabled(const struct net *net); 758 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 759 int mptcp_is_checksum_enabled(const struct net *net); 760 int mptcp_allow_join_id0(const struct net *net); 761 unsigned int mptcp_stale_loss_cnt(const struct net *net); 762 unsigned int mptcp_close_timeout(const struct sock *sk); 763 int mptcp_get_pm_type(const struct net *net); 764 const char *mptcp_get_path_manager(const struct net *net); 765 const char *mptcp_get_scheduler(const struct net *net); 766 767 void mptcp_active_disable(struct sock *sk); 768 bool mptcp_active_should_disable(struct sock *ssk); 769 void mptcp_active_enable(struct sock *sk); 770 771 void mptcp_get_available_schedulers(char *buf, size_t maxlen); 772 void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 773 struct mptcp_subflow_context *subflow, 774 const struct mptcp_options_received *mp_opt); 775 bool __mptcp_retransmit_pending_data(struct sock *sk); 776 void mptcp_check_and_set_pending(struct sock *sk); 777 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 778 bool mptcp_subflow_data_available(struct sock *sk); 779 void __init mptcp_subflow_init(void); 780 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 781 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 782 struct mptcp_subflow_context *subflow); 783 void __mptcp_subflow_send_ack(struct sock *ssk); 784 void mptcp_subflow_reset(struct sock *ssk); 785 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 786 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 787 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 788 bool __mptcp_close(struct sock *sk, long timeout); 789 void mptcp_cancel_work(struct sock *sk); 790 void __mptcp_unaccepted_force_close(struct sock *sk); 791 void mptcp_set_state(struct sock *sk, int state); 792 793 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 794 const struct mptcp_addr_info *b, bool use_port); 795 void mptcp_local_address(const struct sock_common *skc, 796 struct mptcp_addr_info *addr); 797 void mptcp_remote_address(const struct sock_common *skc, 798 struct mptcp_addr_info *addr); 799 800 /* called with sk socket lock held */ 801 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local, 802 const struct mptcp_addr_info *remote); 803 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 804 struct socket **new_sock); 805 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 806 struct sockaddr_storage *addr, 807 unsigned short family); 808 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 809 int mptcp_validate_scheduler(struct mptcp_sched_ops *sched); 810 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 811 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 812 void mptcp_sched_init(void); 813 int mptcp_init_sched(struct mptcp_sock *msk, 814 struct mptcp_sched_ops *sched); 815 void mptcp_release_sched(struct mptcp_sock *msk); 816 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 817 bool scheduled); 818 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 819 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 820 int mptcp_sched_get_send(struct mptcp_sock *msk); 821 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 822 823 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 824 { 825 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 826 } 827 828 static inline bool mptcp_epollin_ready(const struct sock *sk) 829 { 830 u64 data_avail = mptcp_data_avail(mptcp_sk(sk)); 831 832 if (!data_avail) 833 return false; 834 835 /* mptcp doesn't have to deal with small skbs in the receive queue, 836 * as it can always coalesce them 837 */ 838 return (data_avail >= sk->sk_rcvlowat) || 839 tcp_under_memory_pressure(sk); 840 } 841 842 int mptcp_set_rcvlowat(struct sock *sk, int val); 843 844 static inline bool __tcp_can_send(const struct sock *ssk) 845 { 846 /* only send if our side has not closed yet */ 847 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 848 } 849 850 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 851 { 852 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 853 if (subflow->request_join && !READ_ONCE(subflow->fully_established)) 854 return false; 855 856 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 857 } 858 859 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 860 861 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 862 863 void mptcp_subflow_drop_ctx(struct sock *ssk); 864 865 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 866 struct mptcp_subflow_context *ctx) 867 { 868 sk->sk_data_ready = sock_def_readable; 869 sk->sk_state_change = ctx->tcp_state_change; 870 sk->sk_write_space = sk_stream_write_space; 871 sk->sk_error_report = ctx->tcp_error_report; 872 873 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 874 } 875 876 void __init mptcp_proto_init(void); 877 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 878 int __init mptcp_proto_v6_init(void); 879 #endif 880 881 struct sock *mptcp_sk_clone_init(const struct sock *sk, 882 const struct mptcp_options_received *mp_opt, 883 struct sock *ssk, 884 struct request_sock *req); 885 void mptcp_get_options(const struct sk_buff *skb, 886 struct mptcp_options_received *mp_opt); 887 888 void mptcp_finish_connect(struct sock *sk); 889 void __mptcp_sync_state(struct sock *sk, int state); 890 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 891 892 static inline void mptcp_stop_tout_timer(struct sock *sk) 893 { 894 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 895 return; 896 897 sk_stop_timer(sk, &inet_csk(sk)->mptcp_tout_timer); 898 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 899 } 900 901 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 902 { 903 /* avoid 0 timestamp, as that means no close timeout */ 904 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 905 } 906 907 static inline void mptcp_start_tout_timer(struct sock *sk) 908 { 909 mptcp_set_close_tout(sk, tcp_jiffies32); 910 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 911 } 912 913 static inline bool mptcp_is_fully_established(struct sock *sk) 914 { 915 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 916 READ_ONCE(mptcp_sk(sk)->fully_established); 917 } 918 919 static inline u64 mptcp_stamp(void) 920 { 921 return div_u64(tcp_clock_ns(), NSEC_PER_USEC); 922 } 923 924 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 925 bool mptcp_finish_join(struct sock *sk); 926 bool mptcp_schedule_work(struct sock *sk); 927 int mptcp_setsockopt(struct sock *sk, int level, int optname, 928 sockptr_t optval, unsigned int optlen); 929 int mptcp_getsockopt(struct sock *sk, int level, int optname, 930 char __user *optval, int __user *option); 931 932 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 933 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 934 { 935 if (use_64bit) 936 return cur_seq; 937 938 return __mptcp_expand_seq(old_seq, cur_seq); 939 } 940 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 941 void __mptcp_data_acked(struct sock *sk); 942 void __mptcp_error_report(struct sock *sk); 943 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 944 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 945 { 946 return READ_ONCE(msk->snd_data_fin_enable) && 947 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 948 } 949 950 static inline u32 mptcp_notsent_lowat(const struct sock *sk) 951 { 952 struct net *net = sock_net(sk); 953 u32 val; 954 955 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat); 956 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat); 957 } 958 959 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake) 960 { 961 const struct mptcp_sock *msk = mptcp_sk(sk); 962 u32 notsent_bytes; 963 964 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt); 965 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk); 966 } 967 968 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake) 969 { 970 return mptcp_stream_memory_free(sk, wake) && 971 __sk_stream_is_writeable(sk, wake); 972 } 973 974 static inline void mptcp_write_space(struct sock *sk) 975 { 976 /* pairs with memory barrier in mptcp_poll */ 977 smp_mb(); 978 if (mptcp_stream_memory_free(sk, 1)) 979 INDIRECT_CALL_1(sk->sk_write_space, sk_stream_write_space, sk); 980 } 981 982 static inline void __mptcp_sync_sndbuf(struct sock *sk) 983 { 984 struct mptcp_subflow_context *subflow; 985 int ssk_sndbuf, new_sndbuf; 986 987 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 988 return; 989 990 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]); 991 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 992 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 993 994 subflow->cached_sndbuf = ssk_sndbuf; 995 new_sndbuf += ssk_sndbuf; 996 } 997 998 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 999 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 1000 mptcp_write_space(sk); 1001 } 1002 1003 /* The called held both the msk socket and the subflow socket locks, 1004 * possibly under BH 1005 */ 1006 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 1007 { 1008 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1009 1010 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 1011 __mptcp_sync_sndbuf(sk); 1012 } 1013 1014 /* the caller held only the subflow socket lock, either in process or 1015 * BH context. Additionally this can be called under the msk data lock, 1016 * so we can't acquire such lock here: let the delegate action acquires 1017 * the needed locks in suitable order. 1018 */ 1019 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 1020 { 1021 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1022 1023 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 1024 return; 1025 1026 local_bh_disable(); 1027 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 1028 local_bh_enable(); 1029 } 1030 1031 #define MPTCP_TOKEN_MAX_RETRIES 4 1032 1033 void __init mptcp_token_init(void); 1034 static inline void mptcp_token_init_request(struct request_sock *req) 1035 { 1036 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 1037 } 1038 1039 int mptcp_token_new_request(struct request_sock *req); 1040 void mptcp_token_destroy_request(struct request_sock *req); 1041 int mptcp_token_new_connect(struct sock *ssk); 1042 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 1043 struct mptcp_sock *msk); 1044 bool mptcp_token_exists(u32 token); 1045 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 1046 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 1047 long *s_num); 1048 void mptcp_token_destroy(struct mptcp_sock *msk); 1049 1050 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 1051 1052 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 1053 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 1054 1055 void __init mptcp_pm_init(void); 1056 void mptcp_pm_data_init(struct mptcp_sock *msk); 1057 void mptcp_pm_data_reset(struct mptcp_sock *msk); 1058 void mptcp_pm_destroy(struct mptcp_sock *msk); 1059 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 1060 struct mptcp_addr_info *addr); 1061 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 1062 bool require_family, 1063 struct mptcp_pm_addr_entry *entry); 1064 bool mptcp_pm_addr_families_match(const struct sock *sk, 1065 const struct mptcp_addr_info *loc, 1066 const struct mptcp_addr_info *rem); 1067 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 1068 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 1069 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 1070 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 1071 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 1072 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 1073 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 1074 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 1075 const struct mptcp_subflow_context *subflow); 1076 void mptcp_pm_add_addr_received(const struct sock *ssk, 1077 const struct mptcp_addr_info *addr); 1078 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 1079 const struct mptcp_addr_info *addr); 1080 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 1081 void mptcp_pm_send_ack(struct mptcp_sock *msk, 1082 struct mptcp_subflow_context *subflow, 1083 bool prio, bool backup); 1084 void mptcp_pm_addr_send_ack(struct mptcp_sock *msk); 1085 void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id); 1086 void mptcp_pm_rm_subflow(struct mptcp_sock *msk, 1087 const struct mptcp_rm_list *rm_list); 1088 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 1089 const struct mptcp_rm_list *rm_list); 1090 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 1091 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 1092 int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk, 1093 struct mptcp_addr_info *addr, 1094 struct mptcp_addr_info *rem, 1095 u8 bkup); 1096 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 1097 const struct mptcp_addr_info *addr); 1098 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 1099 struct mptcp_pm_add_entry * 1100 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 1101 const struct mptcp_addr_info *addr, bool check_id); 1102 bool mptcp_lookup_subflow_by_saddr(const struct list_head *list, 1103 const struct mptcp_addr_info *saddr); 1104 bool mptcp_remove_anno_list_by_saddr(struct mptcp_sock *msk, 1105 const struct mptcp_addr_info *addr); 1106 int mptcp_pm_nl_set_flags(struct mptcp_pm_addr_entry *local, 1107 struct genl_info *info); 1108 int mptcp_userspace_pm_set_flags(struct mptcp_pm_addr_entry *local, 1109 struct genl_info *info); 1110 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 1111 const struct mptcp_addr_info *addr, 1112 bool echo); 1113 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 1114 void mptcp_pm_remove_addr_entry(struct mptcp_sock *msk, 1115 struct mptcp_pm_addr_entry *entry); 1116 1117 /* the default path manager, used in mptcp_pm_unregister */ 1118 extern struct mptcp_pm_ops mptcp_pm_kernel; 1119 1120 struct mptcp_pm_ops *mptcp_pm_find(const char *name); 1121 int mptcp_pm_register(struct mptcp_pm_ops *pm_ops); 1122 void mptcp_pm_unregister(struct mptcp_pm_ops *pm_ops); 1123 int mptcp_pm_validate(struct mptcp_pm_ops *pm_ops); 1124 void mptcp_pm_get_available(char *buf, size_t maxlen); 1125 1126 void mptcp_userspace_pm_free_local_addr_list(struct mptcp_sock *msk); 1127 1128 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 1129 const struct sock *ssk, gfp_t gfp); 1130 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 1131 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 1132 void mptcp_event_pm_listener(const struct sock *ssk, 1133 enum mptcp_event_type event); 1134 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 1135 1136 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 1137 struct request_sock *req); 1138 int mptcp_pm_genl_fill_addr(struct sk_buff *msg, 1139 struct netlink_callback *cb, 1140 struct mptcp_pm_addr_entry *entry); 1141 1142 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 1143 { 1144 return READ_ONCE(msk->pm.addr_signal) & 1145 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 1146 } 1147 1148 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 1149 { 1150 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 1151 } 1152 1153 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 1154 { 1155 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 1156 } 1157 1158 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 1159 { 1160 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 1161 } 1162 1163 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1164 { 1165 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1166 } 1167 1168 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1169 { 1170 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1171 } 1172 1173 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 1174 { 1175 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 1176 1177 if (family == AF_INET6) 1178 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 1179 if (!echo) 1180 len += MPTCPOPT_THMAC_LEN; 1181 /* account for 2 trailing 'nop' options */ 1182 if (port) 1183 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 1184 1185 return len; 1186 } 1187 1188 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 1189 { 1190 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 1191 return -EINVAL; 1192 1193 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 1194 } 1195 1196 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 1197 unsigned int opt_size, unsigned int remaining, 1198 struct mptcp_addr_info *addr, bool *echo, 1199 bool *drop_other_suboptions); 1200 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1201 struct mptcp_rm_list *rm_list); 1202 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1203 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, 1204 struct mptcp_pm_addr_entry *skc); 1205 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, 1206 struct mptcp_pm_addr_entry *skc); 1207 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc); 1208 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1209 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1210 int mptcp_pm_nl_dump_addr(struct sk_buff *msg, 1211 struct netlink_callback *cb); 1212 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg, 1213 struct netlink_callback *cb); 1214 int mptcp_pm_nl_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1215 struct genl_info *info); 1216 int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1217 struct genl_info *info); 1218 1219 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1220 { 1221 int local_id = READ_ONCE(subflow->local_id); 1222 1223 if (local_id < 0) 1224 return 0; 1225 return local_id; 1226 } 1227 1228 void __init mptcp_pm_kernel_register(void); 1229 void __init mptcp_pm_userspace_register(void); 1230 void __init mptcp_pm_nl_init(void); 1231 void mptcp_pm_worker(struct mptcp_sock *msk); 1232 void __mptcp_pm_kernel_worker(struct mptcp_sock *msk); 1233 u8 mptcp_pm_get_endp_signal_max(const struct mptcp_sock *msk); 1234 u8 mptcp_pm_get_endp_subflow_max(const struct mptcp_sock *msk); 1235 u8 mptcp_pm_get_endp_laminar_max(const struct mptcp_sock *msk); 1236 u8 mptcp_pm_get_endp_fullmesh_max(const struct mptcp_sock *msk); 1237 u8 mptcp_pm_get_limit_add_addr_accepted(const struct mptcp_sock *msk); 1238 u8 mptcp_pm_get_limit_extra_subflows(const struct mptcp_sock *msk); 1239 1240 /* called under PM lock */ 1241 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1242 { 1243 if (--msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk)) 1244 WRITE_ONCE(msk->pm.accept_subflow, true); 1245 } 1246 1247 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1248 { 1249 spin_lock_bh(&msk->pm.lock); 1250 __mptcp_pm_close_subflow(msk); 1251 spin_unlock_bh(&msk->pm.lock); 1252 } 1253 1254 static inline bool mptcp_pm_add_addr_c_flag_case(struct mptcp_sock *msk) 1255 { 1256 return READ_ONCE(msk->pm.remote_deny_join_id0) && 1257 msk->pm.local_addr_used == 0 && 1258 mptcp_pm_get_limit_add_addr_accepted(msk) == 0 && 1259 msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk); 1260 } 1261 1262 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1263 1264 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1265 { 1266 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1267 } 1268 1269 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1270 1271 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1272 { 1273 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1274 } 1275 1276 static inline bool mptcp_check_fallback(const struct sock *sk) 1277 { 1278 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1279 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1280 1281 return __mptcp_check_fallback(msk); 1282 } 1283 1284 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1285 { 1286 struct sock *ssk = READ_ONCE(msk->first); 1287 1288 return ssk && ((1 << inet_sk_state_load(ssk)) & 1289 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1290 TCPF_SYN_RECV | TCPF_LISTEN)); 1291 } 1292 1293 bool __mptcp_try_fallback(struct mptcp_sock *msk, int fb_mib); 1294 1295 static inline bool mptcp_try_fallback(struct sock *ssk, int fb_mib) 1296 { 1297 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1298 struct sock *sk = subflow->conn; 1299 struct mptcp_sock *msk; 1300 1301 msk = mptcp_sk(sk); 1302 if (!__mptcp_try_fallback(msk, fb_mib)) 1303 return false; 1304 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1305 gfp_t saved_allocation = ssk->sk_allocation; 1306 1307 /* we are in a atomic (BH) scope, override ssk default for data 1308 * fin allocation 1309 */ 1310 ssk->sk_allocation = GFP_ATOMIC; 1311 ssk->sk_shutdown |= SEND_SHUTDOWN; 1312 tcp_shutdown(ssk, SEND_SHUTDOWN); 1313 ssk->sk_allocation = saved_allocation; 1314 } 1315 return true; 1316 } 1317 1318 static inline void mptcp_early_fallback(struct mptcp_sock *msk, 1319 struct mptcp_subflow_context *subflow, 1320 int fb_mib) 1321 { 1322 subflow->request_mptcp = 0; 1323 WARN_ON_ONCE(!__mptcp_try_fallback(msk, fb_mib)); 1324 } 1325 1326 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1327 { 1328 struct mptcp_ext *mpext; 1329 1330 mpext = skb ? mptcp_get_ext(skb) : NULL; 1331 if (mpext && mpext->infinite_map) 1332 return true; 1333 1334 return false; 1335 } 1336 1337 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1338 { 1339 return (subflow->request_mptcp || subflow->request_join); 1340 } 1341 1342 static inline bool subflow_simultaneous_connect(struct sock *sk) 1343 { 1344 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1345 1346 /* Note that the sk state implies !subflow->conn_finished. */ 1347 return sk->sk_state == TCP_SYN_RECV && is_active_ssk(subflow); 1348 } 1349 1350 #ifdef CONFIG_SYN_COOKIES 1351 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1352 struct sk_buff *skb); 1353 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1354 struct sk_buff *skb); 1355 void __init mptcp_join_cookie_init(void); 1356 #else 1357 static inline void 1358 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1359 struct sk_buff *skb) {} 1360 static inline bool 1361 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1362 struct sk_buff *skb) 1363 { 1364 return false; 1365 } 1366 1367 static inline void mptcp_join_cookie_init(void) {} 1368 #endif 1369 1370 #endif /* __MPTCP_PROTOCOL_H */ 1371