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 #define OPTIONS_MPTCP_DSS (OPTION_MPTCP_DSS | OPTION_MPTCP_CSUMREQD) 41 42 /* MPTCP option subtypes */ 43 #define MPTCPOPT_MP_CAPABLE 0 44 #define MPTCPOPT_MP_JOIN 1 45 #define MPTCPOPT_DSS 2 46 #define MPTCPOPT_ADD_ADDR 3 47 #define MPTCPOPT_RM_ADDR 4 48 #define MPTCPOPT_MP_PRIO 5 49 #define MPTCPOPT_MP_FAIL 6 50 #define MPTCPOPT_MP_FASTCLOSE 7 51 #define MPTCPOPT_RST 8 52 53 /* MPTCP suboption lengths */ 54 #define TCPOLEN_MPTCP_MPC_SYN 4 55 #define TCPOLEN_MPTCP_MPC_SYNACK 12 56 #define TCPOLEN_MPTCP_MPC_ACK 20 57 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 58 #define TCPOLEN_MPTCP_MPJ_SYN 12 59 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 60 #define TCPOLEN_MPTCP_MPJ_ACK 24 61 #define TCPOLEN_MPTCP_DSS_BASE 4 62 #define TCPOLEN_MPTCP_DSS_ACK32 4 63 #define TCPOLEN_MPTCP_DSS_ACK64 8 64 #define TCPOLEN_MPTCP_DSS_MAP32 10 65 #define TCPOLEN_MPTCP_DSS_MAP64 14 66 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 67 #define TCPOLEN_MPTCP_ADD_ADDR 16 68 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 70 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 71 #define TCPOLEN_MPTCP_ADD_ADDR6 28 72 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 74 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 75 #define TCPOLEN_MPTCP_PORT_LEN 2 76 #define TCPOLEN_MPTCP_PORT_ALIGN 2 77 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 78 #define TCPOLEN_MPTCP_PRIO 3 79 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 80 #define TCPOLEN_MPTCP_FASTCLOSE 12 81 #define TCPOLEN_MPTCP_RST 4 82 #define TCPOLEN_MPTCP_FAIL 12 83 84 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 85 86 /* MPTCP MP_JOIN flags */ 87 #define MPTCPOPT_BACKUP BIT(0) 88 #define MPTCPOPT_THMAC_LEN 8 89 90 /* MPTCP MP_CAPABLE flags */ 91 #define MPTCP_VERSION_MASK (0x0F) 92 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 93 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 94 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 95 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 96 #define MPTCP_CAP_FLAG_MASK (0x1F) 97 98 /* MPTCP DSS flags */ 99 #define MPTCP_DSS_DATA_FIN BIT(4) 100 #define MPTCP_DSS_DSN64 BIT(3) 101 #define MPTCP_DSS_HAS_MAP BIT(2) 102 #define MPTCP_DSS_ACK64 BIT(1) 103 #define MPTCP_DSS_HAS_ACK BIT(0) 104 #define MPTCP_DSS_FLAG_MASK (0x1F) 105 106 /* MPTCP ADD_ADDR flags */ 107 #define MPTCP_ADDR_ECHO BIT(0) 108 109 /* MPTCP MP_PRIO flags */ 110 #define MPTCP_PRIO_BKUP BIT(0) 111 112 /* MPTCP TCPRST flags */ 113 #define MPTCP_RST_TRANSIENT BIT(0) 114 115 /* MPTCP socket atomic flags */ 116 #define MPTCP_WORK_RTX 1 117 #define MPTCP_FALLBACK_DONE 2 118 #define MPTCP_WORK_CLOSE_SUBFLOW 3 119 #define MPTCP_RTX_ENABLED 4 120 121 /* MPTCP socket release cb flags */ 122 #define MPTCP_PUSH_PENDING 1 123 #define MPTCP_CLEAN_UNA 2 124 #define MPTCP_ERROR_REPORT 3 125 #define MPTCP_RETRANSMIT 4 126 #define MPTCP_FLUSH_JOIN_LIST 5 127 #define MPTCP_SYNC_STATE 6 128 #define MPTCP_SYNC_SNDBUF 7 129 130 struct mptcp_skb_cb { 131 u64 map_seq; 132 u64 end_seq; 133 u32 offset; 134 u8 has_rxtstamp; 135 u8 cant_coalesce; 136 }; 137 138 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0])) 139 140 static inline bool before64(__u64 seq1, __u64 seq2) 141 { 142 return (__s64)(seq1 - seq2) < 0; 143 } 144 145 #define after64(seq2, seq1) before64(seq1, seq2) 146 147 struct mptcp_options_received { 148 struct { /* DSS, also used by MP_CAPABLE with data */ 149 u64 data_ack; 150 u64 data_seq; 151 u32 subflow_seq; 152 u16 data_len; 153 __sum16 csum; 154 }; 155 struct_group(status, 156 u16 suboptions; 157 u16 use_map:1, 158 dsn64:1, 159 data_fin:1, 160 use_ack:1, 161 ack64:1, 162 mpc_map:1, 163 reset_reason:4, 164 reset_transient:1, 165 echo:1, 166 backup:1, 167 deny_join_id0:1, 168 __unused:2; 169 ); 170 struct mptcp_rm_list rm_list; 171 /* Options below are mutually exclusive, see mptcp_parse_option() */ 172 union { 173 struct { /* MP_CAPABLE */ 174 u64 sndr_key; 175 u64 rcvr_key; 176 }; 177 struct { /* MP_JOIN */ 178 u32 nonce; 179 u8 join_id; 180 union { 181 u32 token; /* SYN */ 182 u64 thmac; /* SYN + ACK */ 183 u8 hmac[MPTCPOPT_HMAC_LEN]; /* ACK */ 184 }; 185 }; 186 struct { /* ADD_ADDR */ 187 struct mptcp_addr_info addr; 188 u64 ahmac; 189 }; 190 u64 fail_seq; /* MP_FAIL */ 191 u64 fc_recv_key; /* MP_FASTCLOSE */ 192 }; 193 }; 194 195 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 196 { 197 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 198 ((nib & 0xF) << 8) | field); 199 } 200 201 enum mptcp_pm_status { 202 MPTCP_PM_ADD_ADDR_RECEIVED, 203 MPTCP_PM_ADD_ADDR_SEND_ACK, 204 MPTCP_PM_RM_ADDR_RECEIVED, 205 MPTCP_PM_ESTABLISHED, 206 MPTCP_PM_SUBFLOW_ESTABLISHED, 207 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 208 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED, /* persistent status, set after MPC local address is 209 * accounted int id_avail_bitmap 210 */ 211 MPTCP_PM_DESTROYING, /* To fence out PM list allocs */ 212 }; 213 214 enum mptcp_pm_type { 215 MPTCP_PM_TYPE_KERNEL = 0, 216 MPTCP_PM_TYPE_USERSPACE, 217 218 __MPTCP_PM_TYPE_NR, 219 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 220 }; 221 222 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 223 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 224 225 enum mptcp_addr_signal_status { 226 MPTCP_ADD_ADDR_SIGNAL, 227 MPTCP_ADD_ADDR_ECHO, 228 MPTCP_RM_ADDR_SIGNAL, 229 }; 230 231 /* max value of mptcp_addr_info.id */ 232 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 233 234 struct mptcp_pm_data { 235 struct mptcp_addr_info local; 236 struct mptcp_addr_info remote; 237 struct list_head anno_list; 238 struct list_head userspace_pm_local_addr_list; 239 240 spinlock_t lock; /*protects the whole PM data */ 241 242 struct_group(reset, 243 244 u8 addr_signal; 245 bool server_side; 246 bool work_pending; 247 bool accept_addr; 248 bool accept_subflow; 249 bool remote_deny_join_id0; 250 u8 add_addr_signaled; 251 u8 add_addr_accepted; 252 u8 local_addr_used; 253 u8 pm_type; 254 u8 extra_subflows; 255 u8 status; 256 257 ); 258 259 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 260 struct mptcp_rm_list rm_list_tx; 261 struct mptcp_rm_list rm_list_rx; 262 }; 263 264 struct mptcp_pm_local { 265 struct mptcp_addr_info addr; 266 u32 flags; 267 int ifindex; 268 }; 269 270 struct mptcp_pm_addr_entry { 271 struct list_head list; 272 struct mptcp_addr_info addr; 273 u32 flags; 274 int ifindex; 275 struct socket *lsk; 276 }; 277 278 struct mptcp_data_frag { 279 struct list_head list; 280 u64 data_seq; 281 u16 data_len; 282 u16 offset; 283 u8 overhead; 284 u8 eor; /* currently using 1 bit */ 285 u16 already_sent; 286 struct page *page; 287 }; 288 289 /* Arbitrary compromise between as low as possible to react timely to subflow 290 * close event and as big as possible to avoid being fouled by biased large 291 * samples due to peer sending data on a different subflow WRT to the incoming 292 * ack. 293 */ 294 #define MPTCP_RTT_SAMPLES 5 295 296 /* MPTCP connection sock */ 297 struct mptcp_sock { 298 /* inet_connection_sock must be the first member */ 299 struct inet_connection_sock sk; 300 u64 local_key; /* protected by the first subflow socket lock 301 * lockless access read 302 */ 303 u64 remote_key; /* same as above */ 304 u64 write_seq; 305 u64 bytes_sent; 306 u64 snd_nxt; 307 u64 bytes_received; 308 u64 ack_seq; 309 atomic64_t rcv_wnd_sent; 310 u64 rcv_data_fin_seq; 311 u64 bytes_retrans; 312 u64 bytes_consumed; 313 int snd_burst; 314 int old_wspace; 315 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 316 * recovery related fields are under data_lock 317 * protection 318 */ 319 u64 bytes_acked; 320 u64 snd_una; 321 u64 wnd_end; 322 u32 last_data_sent; 323 u32 last_data_recv; 324 u32 last_ack_recv; 325 unsigned long timer_ival; 326 u32 token; 327 unsigned long flags; 328 unsigned long cb_flags; 329 bool recovery; /* closing subflow write queue reinjected */ 330 bool can_ack; 331 bool fully_established; 332 bool rcv_data_fin; 333 bool snd_data_fin_enable; 334 bool rcv_fastclose; 335 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 336 bool csum_enabled; 337 bool allow_infinite_fallback; 338 u8 pending_state; /* A subflow asked to set this sk_state, 339 * protected by the msk data lock 340 */ 341 u8 mpc_endpoint_id; 342 u8 recvmsg_inq:1, 343 cork:1, 344 nodelay:1, 345 fastopening:1, 346 in_accept_queue:1, 347 free_first:1, 348 rcvspace_init:1, 349 fastclosing:1; 350 u32 notsent_lowat; 351 int keepalive_cnt; 352 int keepalive_idle; 353 int keepalive_intvl; 354 int maxseg; 355 struct work_struct work; 356 struct sk_buff *ooo_last_skb; 357 struct rb_root out_of_order_queue; 358 struct list_head conn_list; 359 struct list_head rtx_queue; 360 struct mptcp_data_frag *first_pending; 361 struct list_head join_list; 362 struct sock *first; /* The mptcp ops can safely dereference, using suitable 363 * ONCE annotation, the subflow outside the socket 364 * lock as such sock is freed after close(). 365 */ 366 struct mptcp_pm_data pm; 367 struct mptcp_sched_ops *sched; 368 369 /* Most recent rtt_us observed by in use incoming subflows. */ 370 struct { 371 u32 samples[MPTCP_RTT_SAMPLES]; 372 u32 next_sample; 373 } rcv_rtt_est; 374 375 struct { 376 int space; /* bytes copied in last measurement window */ 377 int copied; /* bytes copied in this measurement window */ 378 u64 time; /* start time of measurement window */ 379 } rcvq_space; 380 u8 scaling_ratio; 381 bool allow_subflows; 382 383 u32 subflow_id; 384 u32 setsockopt_seq; 385 char ca_name[TCP_CA_NAME_MAX]; 386 387 spinlock_t fallback_lock; /* protects fallback, 388 * allow_infinite_fallback and 389 * allow_join 390 */ 391 392 struct list_head backlog_list; /* protected by the data lock */ 393 u32 backlog_len; 394 u32 backlog_unaccounted; 395 }; 396 397 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 398 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 399 400 #define mptcp_for_each_subflow(__msk, __subflow) \ 401 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 402 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 403 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 404 #define mptcp_next_subflow(__msk, __subflow) \ 405 list_next_entry_circular(__subflow, &((__msk)->conn_list), node) 406 407 extern struct genl_family mptcp_genl_family; 408 409 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 410 { 411 sock_owned_by_me((const struct sock *)msk); 412 } 413 414 #ifdef CONFIG_DEBUG_NET 415 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 416 #undef tcp_sk 417 #define tcp_sk(ptr) ({ \ 418 typeof(ptr) _ptr = (ptr); \ 419 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 420 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 421 }) 422 #define mptcp_sk(ptr) ({ \ 423 typeof(ptr) _ptr = (ptr); \ 424 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 425 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 426 }) 427 428 #else /* !CONFIG_DEBUG_NET */ 429 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 430 #endif 431 432 static inline int mptcp_win_from_space(const struct sock *sk, int space) 433 { 434 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 435 } 436 437 static inline int mptcp_space_from_win(const struct sock *sk, int win) 438 { 439 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win); 440 } 441 442 static inline int __mptcp_space(const struct sock *sk) 443 { 444 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - 445 READ_ONCE(mptcp_sk(sk)->backlog_len) - 446 sk_rmem_alloc_get(sk)); 447 } 448 449 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 450 { 451 const struct mptcp_sock *msk = mptcp_sk(sk); 452 453 return msk->first_pending; 454 } 455 456 static inline void mptcp_init_rtt_est(struct mptcp_sock *msk) 457 { 458 int i; 459 460 for (i = 0; i < MPTCP_RTT_SAMPLES; ++i) 461 msk->rcv_rtt_est.samples[i] = U32_MAX; 462 msk->rcv_rtt_est.next_sample = 0; 463 msk->scaling_ratio = TCP_DEFAULT_SCALING_RATIO; 464 } 465 466 static inline u32 mptcp_rtt_us_est(const struct mptcp_sock *msk) 467 { 468 u32 rtt_us = READ_ONCE(msk->rcv_rtt_est.samples[0]); 469 int i; 470 471 /* Lockless access of collected samples. */ 472 for (i = 1; i < MPTCP_RTT_SAMPLES; ++i) 473 rtt_us = min(rtt_us, READ_ONCE(msk->rcv_rtt_est.samples[i])); 474 return rtt_us; 475 } 476 477 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 478 { 479 struct mptcp_sock *msk = mptcp_sk(sk); 480 struct mptcp_data_frag *cur; 481 482 cur = msk->first_pending; 483 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 484 list_next_entry(cur, list); 485 } 486 487 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 488 { 489 const struct mptcp_sock *msk = mptcp_sk(sk); 490 491 if (!msk->first_pending) 492 return NULL; 493 494 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 495 return NULL; 496 497 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 498 } 499 500 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 501 { 502 struct mptcp_sock *msk = mptcp_sk(sk); 503 504 if (msk->snd_una == msk->snd_nxt) 505 return NULL; 506 507 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 508 } 509 510 struct csum_pseudo_header { 511 __be64 data_seq; 512 __be32 subflow_seq; 513 __be16 data_len; 514 __sum16 csum; 515 }; 516 517 struct mptcp_subflow_request_sock { 518 struct tcp_request_sock sk; 519 u16 mp_capable : 1, 520 mp_join : 1, 521 backup : 1, 522 request_bkup : 1, 523 csum_reqd : 1, 524 allow_join_id0 : 1; 525 u8 local_id; 526 u8 remote_id; 527 u64 local_key; 528 u64 idsn; 529 u32 token; 530 u32 ssn_offset; 531 u64 thmac; 532 u32 local_nonce; 533 u32 remote_nonce; 534 struct mptcp_sock *msk; 535 struct hlist_nulls_node token_node; 536 }; 537 538 static inline struct mptcp_subflow_request_sock * 539 mptcp_subflow_rsk(const struct request_sock *rsk) 540 { 541 return (struct mptcp_subflow_request_sock *)rsk; 542 } 543 544 struct mptcp_delegated_action { 545 struct napi_struct napi; 546 local_lock_t bh_lock; 547 struct list_head head; 548 }; 549 550 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 551 552 #define MPTCP_DELEGATE_SCHEDULED 0 553 #define MPTCP_DELEGATE_SEND 1 554 #define MPTCP_DELEGATE_ACK 2 555 #define MPTCP_DELEGATE_SNDBUF 3 556 557 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 558 /* MPTCP subflow context */ 559 struct mptcp_subflow_context { 560 struct list_head node;/* conn_list of subflows */ 561 562 struct_group(reset, 563 564 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 565 u64 local_key; 566 u64 remote_key; 567 u64 idsn; 568 u64 map_seq; 569 u64 rcv_wnd_sent; 570 u32 snd_isn; 571 u32 token; 572 u32 rel_write_seq; 573 u32 map_subflow_seq; 574 u32 ssn_offset; 575 u32 map_data_len; 576 __wsum map_data_csum; 577 u32 map_csum_len; 578 u32 prev_rtt_seq; 579 u32 request_mptcp : 1, /* send MP_CAPABLE */ 580 request_join : 1, /* send MP_JOIN */ 581 request_bkup : 1, 582 mp_capable : 1, /* remote is MPTCP capable */ 583 mp_join : 1, /* remote is JOINing */ 584 pm_notified : 1, /* PM hook called for established status */ 585 conn_finished : 1, 586 map_valid : 1, 587 map_csum_reqd : 1, 588 map_data_fin : 1, 589 mpc_map : 1, 590 backup : 1, 591 send_mp_prio : 1, 592 send_mp_fail : 1, 593 send_fastclose : 1, 594 send_infinite_map : 1, 595 remote_key_valid : 1, /* received the peer key from */ 596 disposable : 1, /* ctx can be free at ulp release time */ 597 closing : 1, /* must not pass rx data to msk anymore */ 598 valid_csum_seen : 1, /* at least one csum validated */ 599 is_mptfo : 1, /* subflow is doing TFO */ 600 close_event_done : 1, /* has done the post-closed part */ 601 mpc_drop : 1, /* the MPC option has been dropped in a rtx */ 602 resetting : 1, /* subflow is resetting */ 603 __unused : 8; 604 bool data_avail; 605 bool scheduled; 606 bool pm_listener; /* a listener managed by the kernel PM? */ 607 bool fully_established; /* path validated */ 608 u32 lent_mem_frag; 609 u32 remote_nonce; 610 u32 local_nonce; 611 u32 remote_token; 612 union { 613 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 614 u64 iasn; /* initial ack sequence number, MPC subflows only */ 615 }; 616 s16 local_id; /* if negative not initialized yet */ 617 u8 remote_id; 618 u8 reset_seen:1; 619 u8 reset_transient:1; 620 u8 reset_reason:4; 621 u8 stale_count; /* Protected by the msk socket lock */ 622 u8 stale; /* Protected by the msk socket lock, 623 * if set the subflow is unable to snd/rcv 624 * data, the schedule should skip it 625 */ 626 627 u32 subflow_id; 628 629 long delegated_status; 630 unsigned long fail_tout; 631 632 ); 633 634 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 635 636 u32 setsockopt_seq; 637 u32 stale_rcv_tstamp; 638 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 639 * protected by the msk socket lock 640 */ 641 642 struct sock *tcp_sock; /* tcp sk backpointer */ 643 struct sock *conn; /* parent mptcp_sock */ 644 const struct inet_connection_sock_af_ops *icsk_af_ops; 645 void (*tcp_state_change)(struct sock *sk); 646 void (*tcp_error_report)(struct sock *sk); 647 648 struct rcu_head rcu; 649 }; 650 651 static inline struct mptcp_subflow_context * 652 mptcp_subflow_ctx(const struct sock *sk) 653 { 654 const struct inet_connection_sock *icsk = inet_csk(sk); 655 656 /* Use RCU on icsk_ulp_data only for sock diag code */ 657 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 658 } 659 660 static inline struct sock * 661 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 662 { 663 return subflow->tcp_sock; 664 } 665 666 static inline void 667 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 668 { 669 memset(&subflow->reset, 0, sizeof(subflow->reset)); 670 subflow->request_mptcp = 1; 671 WRITE_ONCE(subflow->local_id, -1); 672 } 673 674 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */ 675 static inline enum sk_rst_reason 676 sk_rst_convert_mptcp_reason(u32 reason) 677 { 678 switch (reason) { 679 case MPTCP_RST_EUNSPEC: 680 return SK_RST_REASON_MPTCP_RST_EUNSPEC; 681 case MPTCP_RST_EMPTCP: 682 return SK_RST_REASON_MPTCP_RST_EMPTCP; 683 case MPTCP_RST_ERESOURCE: 684 return SK_RST_REASON_MPTCP_RST_ERESOURCE; 685 case MPTCP_RST_EPROHIBIT: 686 return SK_RST_REASON_MPTCP_RST_EPROHIBIT; 687 case MPTCP_RST_EWQ2BIG: 688 return SK_RST_REASON_MPTCP_RST_EWQ2BIG; 689 case MPTCP_RST_EBADPERF: 690 return SK_RST_REASON_MPTCP_RST_EBADPERF; 691 case MPTCP_RST_EMIDDLEBOX: 692 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX; 693 default: 694 /* It should not happen, or else errors may occur 695 * in MPTCP layer 696 */ 697 return SK_RST_REASON_ERROR; 698 } 699 } 700 701 static inline void 702 mptcp_send_active_reset_reason(struct sock *sk) 703 { 704 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 705 enum sk_rst_reason reason; 706 707 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason); 708 tcp_send_active_reset(sk, reason); 709 } 710 711 /* Made the fwd mem carried by the given skb available to the msk, 712 * To be paired with a previous mptcp_subflow_lend_fwdmem() before freeing 713 * the skb or setting the skb ownership. 714 */ 715 static inline void mptcp_borrow_fwdmem(struct sock *sk, struct sk_buff *skb) 716 { 717 struct sock *ssk = skb->sk; 718 719 /* The subflow just lend the skb fwd memory; if the subflow meanwhile 720 * closed, mptcp_close_ssk() already released the ssk rcv memory. 721 */ 722 DEBUG_NET_WARN_ON_ONCE(skb->destructor); 723 sk_forward_alloc_add(sk, skb->truesize); 724 if (!ssk) 725 return; 726 727 atomic_sub(skb->truesize, &ssk->sk_rmem_alloc); 728 skb->sk = NULL; 729 } 730 731 static inline void 732 __mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow, int size) 733 { 734 int frag = (subflow->lent_mem_frag + size) & (PAGE_SIZE - 1); 735 736 subflow->lent_mem_frag = frag; 737 } 738 739 static inline void 740 mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow, 741 struct sk_buff *skb) 742 { 743 __mptcp_subflow_lend_fwdmem(subflow, skb->truesize); 744 skb->destructor = NULL; 745 } 746 747 static inline u64 748 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 749 { 750 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 751 subflow->ssn_offset - 752 subflow->map_subflow_seq; 753 } 754 755 static inline u64 756 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 757 { 758 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 759 } 760 761 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 762 763 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 764 { 765 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 766 struct mptcp_delegated_action *delegated; 767 bool schedule; 768 769 /* the caller held the subflow bh socket lock */ 770 lockdep_assert_in_softirq(); 771 772 /* The implied barrier pairs with tcp_release_cb_override() 773 * mptcp_napi_poll(), and ensures the below list check sees list 774 * updates done prior to delegated status bits changes 775 */ 776 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 777 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 778 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 779 return; 780 781 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 782 delegated = this_cpu_ptr(&mptcp_delegated_actions); 783 schedule = list_empty(&delegated->head); 784 list_add_tail(&subflow->delegated_node, &delegated->head); 785 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 786 sock_hold(mptcp_subflow_tcp_sock(subflow)); 787 if (schedule) 788 napi_schedule(&delegated->napi); 789 } 790 } 791 792 static inline struct mptcp_subflow_context * 793 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 794 { 795 struct mptcp_subflow_context *ret; 796 797 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 798 if (list_empty(&delegated->head)) { 799 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 800 return NULL; 801 } 802 803 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 804 list_del_init(&ret->delegated_node); 805 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 806 return ret; 807 } 808 809 void __mptcp_inherit_memcg(struct sock *sk, struct sock *ssk, gfp_t gfp); 810 void __mptcp_inherit_cgrp_data(struct sock *sk, struct sock *ssk); 811 812 int mptcp_is_enabled(const struct net *net); 813 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 814 int mptcp_is_checksum_enabled(const struct net *net); 815 int mptcp_allow_join_id0(const struct net *net); 816 unsigned int mptcp_stale_loss_cnt(const struct net *net); 817 unsigned int mptcp_close_timeout(const struct sock *sk); 818 int mptcp_get_pm_type(const struct net *net); 819 const char *mptcp_get_path_manager(const struct net *net); 820 const char *mptcp_get_scheduler(const struct net *net); 821 unsigned int mptcp_add_addr_v6_port_drop_ts(const struct net *net); 822 823 void mptcp_active_disable(struct sock *sk); 824 bool mptcp_active_should_disable(struct sock *ssk); 825 void mptcp_active_enable(struct sock *sk); 826 827 void mptcp_get_available_schedulers(char *buf, size_t maxlen); 828 void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 829 struct mptcp_subflow_context *subflow, 830 const struct mptcp_options_received *mp_opt); 831 bool __mptcp_retransmit_pending_data(struct sock *sk); 832 void mptcp_check_and_set_pending(struct sock *sk); 833 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 834 bool mptcp_subflow_data_available(struct sock *sk); 835 void __init mptcp_subflow_init(void); 836 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 837 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 838 struct mptcp_subflow_context *subflow); 839 void __mptcp_subflow_send_ack(struct sock *ssk); 840 void mptcp_subflow_reset(struct sock *ssk); 841 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 842 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 843 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 844 bool __mptcp_close(struct sock *sk, long timeout); 845 void mptcp_cancel_work(struct sock *sk); 846 void __mptcp_unaccepted_force_close(struct sock *sk); 847 void mptcp_set_state(struct sock *sk, int state); 848 849 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 850 const struct mptcp_addr_info *b, bool use_port); 851 void mptcp_local_address(const struct sock_common *skc, 852 struct mptcp_addr_info *addr); 853 void mptcp_remote_address(const struct sock_common *skc, 854 struct mptcp_addr_info *addr); 855 856 /* called with sk socket lock held */ 857 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local, 858 const struct mptcp_addr_info *remote); 859 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 860 struct socket **new_sock); 861 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 862 struct sockaddr_storage *addr, 863 unsigned short family); 864 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 865 int mptcp_validate_scheduler(struct mptcp_sched_ops *sched); 866 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 867 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 868 void mptcp_sched_init(void); 869 int mptcp_init_sched(struct mptcp_sock *msk, 870 struct mptcp_sched_ops *sched); 871 void mptcp_release_sched(struct mptcp_sock *msk); 872 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 873 bool scheduled); 874 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 875 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 876 int mptcp_sched_get_send(struct mptcp_sock *msk); 877 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 878 879 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 880 { 881 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 882 } 883 884 static inline bool mptcp_epollin_ready(const struct sock *sk) 885 { 886 u64 data_avail = mptcp_data_avail(mptcp_sk(sk)); 887 888 if (!data_avail) 889 return false; 890 891 /* mptcp doesn't have to deal with small skbs in the receive queue, 892 * as it can always coalesce them 893 */ 894 return (data_avail >= sk->sk_rcvlowat) || 895 tcp_under_memory_pressure(sk); 896 } 897 898 int mptcp_set_rcvlowat(struct sock *sk, int val); 899 900 static inline bool __tcp_can_send(const struct sock *ssk) 901 { 902 /* only send if our side has not closed yet */ 903 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 904 } 905 906 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 907 { 908 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 909 if (subflow->request_join && !READ_ONCE(subflow->fully_established)) 910 return false; 911 912 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 913 } 914 915 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 916 917 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 918 919 void mptcp_subflow_drop_ctx(struct sock *ssk); 920 921 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 922 struct mptcp_subflow_context *ctx) 923 { 924 sk->sk_data_ready = sock_def_readable; 925 sk->sk_state_change = ctx->tcp_state_change; 926 sk->sk_write_space = sk_stream_write_space; 927 sk->sk_error_report = ctx->tcp_error_report; 928 929 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 930 } 931 932 void __init mptcp_proto_init(void); 933 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 934 int __init mptcp_proto_v6_init(void); 935 void __init mptcp_subflow_v6_init(void); 936 #endif 937 938 struct sock *mptcp_sk_clone_init(const struct sock *sk, 939 const struct mptcp_options_received *mp_opt, 940 struct sock *ssk, 941 struct request_sock *req); 942 void mptcp_get_options(const struct sk_buff *skb, 943 struct mptcp_options_received *mp_opt); 944 945 void mptcp_finish_connect(struct sock *sk); 946 void __mptcp_sync_state(struct sock *sk, int state); 947 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 948 949 static inline void mptcp_stop_tout_timer(struct sock *sk) 950 { 951 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 952 return; 953 954 sk_stop_timer(sk, &inet_csk(sk)->mptcp_tout_timer); 955 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 956 } 957 958 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 959 { 960 /* avoid 0 timestamp, as that means no close timeout */ 961 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 962 } 963 964 static inline void mptcp_start_tout_timer(struct sock *sk) 965 { 966 mptcp_set_close_tout(sk, tcp_jiffies32); 967 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 968 } 969 970 static inline bool mptcp_is_fully_established(struct sock *sk) 971 { 972 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 973 READ_ONCE(mptcp_sk(sk)->fully_established); 974 } 975 976 static inline u64 mptcp_stamp(void) 977 { 978 return div_u64(tcp_clock_ns(), NSEC_PER_USEC); 979 } 980 981 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 982 bool mptcp_finish_join(struct sock *sk); 983 bool mptcp_schedule_work(struct sock *sk); 984 int mptcp_setsockopt(struct sock *sk, int level, int optname, 985 sockptr_t optval, unsigned int optlen); 986 int mptcp_getsockopt(struct sock *sk, int level, int optname, 987 char __user *optval, int __user *option); 988 989 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 990 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 991 { 992 if (use_64bit) 993 return cur_seq; 994 995 return __mptcp_expand_seq(old_seq, cur_seq); 996 } 997 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 998 void __mptcp_data_acked(struct sock *sk); 999 void __mptcp_error_report(struct sock *sk); 1000 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 1001 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 1002 { 1003 return READ_ONCE(msk->snd_data_fin_enable) && 1004 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 1005 } 1006 1007 static inline u32 mptcp_notsent_lowat(const struct sock *sk) 1008 { 1009 struct net *net = sock_net(sk); 1010 u32 val; 1011 1012 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat); 1013 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat); 1014 } 1015 1016 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake) 1017 { 1018 const struct mptcp_sock *msk = mptcp_sk(sk); 1019 u32 notsent_bytes; 1020 1021 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt); 1022 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk); 1023 } 1024 1025 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake) 1026 { 1027 return mptcp_stream_memory_free(sk, wake) && 1028 __sk_stream_is_writeable(sk, wake); 1029 } 1030 1031 static inline void mptcp_write_space(struct sock *sk) 1032 { 1033 /* pairs with memory barrier in mptcp_poll */ 1034 smp_mb(); 1035 if (mptcp_stream_memory_free(sk, 1)) 1036 INDIRECT_CALL_1(sk->sk_write_space, sk_stream_write_space, sk); 1037 } 1038 1039 static inline void __mptcp_sync_sndbuf(struct sock *sk) 1040 { 1041 struct mptcp_subflow_context *subflow; 1042 int ssk_sndbuf, new_sndbuf; 1043 1044 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 1045 return; 1046 1047 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]); 1048 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 1049 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 1050 1051 subflow->cached_sndbuf = ssk_sndbuf; 1052 new_sndbuf += ssk_sndbuf; 1053 } 1054 1055 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 1056 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 1057 mptcp_write_space(sk); 1058 } 1059 1060 /* The called held both the msk socket and the subflow socket locks, 1061 * possibly under BH 1062 */ 1063 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 1064 { 1065 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1066 1067 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 1068 __mptcp_sync_sndbuf(sk); 1069 } 1070 1071 /* the caller held only the subflow socket lock, either in process or 1072 * BH context. Additionally this can be called under the msk data lock, 1073 * so we can't acquire such lock here: let the delegate action acquires 1074 * the needed locks in suitable order. 1075 */ 1076 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 1077 { 1078 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1079 1080 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 1081 return; 1082 1083 local_bh_disable(); 1084 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 1085 local_bh_enable(); 1086 } 1087 1088 #define MPTCP_TOKEN_MAX_RETRIES 4 1089 1090 void __init mptcp_token_init(void); 1091 static inline void mptcp_token_init_request(struct request_sock *req) 1092 { 1093 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 1094 } 1095 1096 int mptcp_token_new_request(struct request_sock *req); 1097 void mptcp_token_destroy_request(struct request_sock *req); 1098 int mptcp_token_new_connect(struct sock *ssk); 1099 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 1100 struct mptcp_sock *msk); 1101 bool mptcp_token_exists(u32 token); 1102 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 1103 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 1104 long *s_num); 1105 void mptcp_token_destroy(struct mptcp_sock *msk); 1106 1107 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 1108 1109 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 1110 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 1111 1112 void __init mptcp_pm_init(void); 1113 void mptcp_pm_data_init(struct mptcp_sock *msk); 1114 void mptcp_pm_data_reset(struct mptcp_sock *msk); 1115 void mptcp_pm_destroy(struct mptcp_sock *msk); 1116 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 1117 struct mptcp_addr_info *addr); 1118 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 1119 bool require_family, 1120 struct mptcp_pm_addr_entry *entry); 1121 bool mptcp_pm_addr_families_match(const struct sock *sk, 1122 const struct mptcp_addr_info *loc, 1123 const struct mptcp_addr_info *rem); 1124 void mptcp_pm_chk_stale(const struct mptcp_sock *msk); 1125 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 1126 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 1127 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 1128 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 1129 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 1130 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 1131 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 1132 const struct mptcp_subflow_context *subflow); 1133 void mptcp_pm_add_addr_received(const struct sock *ssk, 1134 const struct mptcp_addr_info *addr); 1135 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 1136 const struct mptcp_addr_info *addr); 1137 void mptcp_pm_send_ack(struct mptcp_sock *msk, 1138 struct mptcp_subflow_context *subflow, 1139 bool prio, bool backup); 1140 void mptcp_pm_addr_send_ack(struct mptcp_sock *msk); 1141 void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id); 1142 void mptcp_pm_rm_subflow(struct mptcp_sock *msk, 1143 const struct mptcp_rm_list *rm_list); 1144 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 1145 const struct mptcp_rm_list *rm_list); 1146 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 1147 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 1148 int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk, 1149 struct mptcp_addr_info *addr, 1150 struct mptcp_addr_info *rem, 1151 u8 bkup); 1152 bool mptcp_pm_announced_alloc(struct mptcp_sock *msk, 1153 const struct mptcp_addr_info *addr); 1154 bool mptcp_pm_announced_remove(struct mptcp_sock *msk, 1155 const struct mptcp_addr_info *addr); 1156 bool mptcp_pm_announced_has_ssk(struct mptcp_sock *msk, const struct sock *ssk); 1157 bool mptcp_pm_has_subflow_saddr(const struct mptcp_sock *msk, 1158 const struct mptcp_addr_info *saddr); 1159 int mptcp_pm_nl_set_flags(struct mptcp_pm_addr_entry *local, 1160 struct genl_info *info); 1161 int mptcp_userspace_pm_set_flags(struct mptcp_pm_addr_entry *local, 1162 struct genl_info *info); 1163 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 1164 const struct mptcp_addr_info *addr, 1165 bool echo); 1166 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 1167 1168 /* the default path manager, used in mptcp_pm_unregister */ 1169 extern struct mptcp_pm_ops mptcp_pm_kernel; 1170 1171 struct mptcp_pm_ops *mptcp_pm_find(const char *name); 1172 int mptcp_pm_register(struct mptcp_pm_ops *pm_ops); 1173 void mptcp_pm_unregister(struct mptcp_pm_ops *pm_ops); 1174 int mptcp_pm_validate(struct mptcp_pm_ops *pm_ops); 1175 void mptcp_pm_get_available(char *buf, size_t maxlen); 1176 1177 void mptcp_userspace_pm_free_local_addr_list(struct mptcp_sock *msk); 1178 1179 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 1180 const struct sock *ssk, gfp_t gfp); 1181 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 1182 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 1183 void mptcp_event_pm_listener(const struct sock *ssk, 1184 enum mptcp_event_type event); 1185 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 1186 1187 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 1188 struct request_sock *req); 1189 int mptcp_pm_genl_fill_addr(struct sk_buff *msg, 1190 struct netlink_callback *cb, 1191 struct mptcp_pm_addr_entry *entry); 1192 1193 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 1194 { 1195 return READ_ONCE(msk->pm.addr_signal) & 1196 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 1197 } 1198 1199 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 1200 { 1201 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 1202 } 1203 1204 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 1205 { 1206 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 1207 } 1208 1209 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 1210 { 1211 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 1212 } 1213 1214 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1215 { 1216 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1217 } 1218 1219 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1220 { 1221 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1222 } 1223 1224 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, int *size, int remaining, 1225 struct mptcp_addr_info *addr, bool *echo, 1226 bool *drop_ts); 1227 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1228 struct mptcp_rm_list *rm_list, int *len); 1229 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1230 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, 1231 struct mptcp_pm_addr_entry *skc); 1232 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, 1233 struct mptcp_pm_addr_entry *skc); 1234 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc); 1235 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1236 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1237 int mptcp_pm_nl_dump_addr(struct sk_buff *msg, 1238 struct netlink_callback *cb); 1239 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg, 1240 struct netlink_callback *cb); 1241 int mptcp_pm_nl_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1242 struct genl_info *info); 1243 int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1244 struct genl_info *info); 1245 1246 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1247 { 1248 int local_id = READ_ONCE(subflow->local_id); 1249 1250 if (local_id < 0) 1251 return 0; 1252 return local_id; 1253 } 1254 1255 void __init mptcp_pm_kernel_register(void); 1256 void __init mptcp_pm_userspace_register(void); 1257 void __init mptcp_pm_nl_init(void); 1258 void mptcp_pm_worker(struct mptcp_sock *msk); 1259 void __mptcp_pm_kernel_worker(struct mptcp_sock *msk); 1260 u8 mptcp_pm_get_endp_signal_max(const struct mptcp_sock *msk); 1261 u8 mptcp_pm_get_endp_subflow_max(const struct mptcp_sock *msk); 1262 u8 mptcp_pm_get_endp_laminar_max(const struct mptcp_sock *msk); 1263 u8 mptcp_pm_get_endp_fullmesh_max(const struct mptcp_sock *msk); 1264 u8 mptcp_pm_get_limit_add_addr_accepted(const struct mptcp_sock *msk); 1265 u8 mptcp_pm_get_limit_extra_subflows(const struct mptcp_sock *msk); 1266 1267 /* called under PM lock */ 1268 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1269 { 1270 if (!WARN_ON_ONCE(msk->pm.extra_subflows == 0) && 1271 --msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk)) 1272 WRITE_ONCE(msk->pm.accept_subflow, true); 1273 } 1274 1275 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1276 { 1277 spin_lock_bh(&msk->pm.lock); 1278 __mptcp_pm_close_subflow(msk); 1279 spin_unlock_bh(&msk->pm.lock); 1280 } 1281 1282 static inline bool mptcp_pm_add_addr_c_flag_case(struct mptcp_sock *msk) 1283 { 1284 return READ_ONCE(msk->pm.remote_deny_join_id0) && 1285 msk->pm.local_addr_used == 0 && 1286 mptcp_pm_get_limit_add_addr_accepted(msk) == 0 && 1287 msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk); 1288 } 1289 1290 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1291 1292 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1293 { 1294 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1295 } 1296 1297 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1298 1299 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1300 { 1301 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1302 } 1303 1304 static inline bool mptcp_check_fallback(const struct sock *sk) 1305 { 1306 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1307 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1308 1309 return __mptcp_check_fallback(msk); 1310 } 1311 1312 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1313 { 1314 struct sock *ssk = READ_ONCE(msk->first); 1315 1316 return ssk && ((1 << inet_sk_state_load(ssk)) & 1317 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1318 TCPF_SYN_RECV | TCPF_LISTEN)); 1319 } 1320 1321 bool __mptcp_try_fallback(struct mptcp_sock *msk, int fb_mib); 1322 1323 static inline bool mptcp_try_fallback(struct sock *ssk, int fb_mib) 1324 { 1325 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1326 struct sock *sk = subflow->conn; 1327 struct mptcp_sock *msk; 1328 1329 msk = mptcp_sk(sk); 1330 if (!__mptcp_try_fallback(msk, fb_mib)) 1331 return false; 1332 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1333 gfp_t saved_allocation = ssk->sk_allocation; 1334 1335 /* we are in a atomic (BH) scope, override ssk default for data 1336 * fin allocation 1337 */ 1338 ssk->sk_allocation = GFP_ATOMIC; 1339 ssk->sk_shutdown |= SEND_SHUTDOWN; 1340 tcp_shutdown(ssk, SEND_SHUTDOWN); 1341 ssk->sk_allocation = saved_allocation; 1342 } 1343 return true; 1344 } 1345 1346 static inline void mptcp_early_fallback(struct mptcp_sock *msk, 1347 struct mptcp_subflow_context *subflow, 1348 int fb_mib) 1349 { 1350 subflow->request_mptcp = 0; 1351 WARN_ON_ONCE(!__mptcp_try_fallback(msk, fb_mib)); 1352 } 1353 1354 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1355 { 1356 struct mptcp_ext *mpext; 1357 1358 mpext = skb ? mptcp_get_ext(skb) : NULL; 1359 if (mpext && mpext->infinite_map) 1360 return true; 1361 1362 return false; 1363 } 1364 1365 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1366 { 1367 return (subflow->request_mptcp || subflow->request_join); 1368 } 1369 1370 static inline bool subflow_simultaneous_connect(struct sock *sk) 1371 { 1372 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1373 1374 /* Note that the sk state implies !subflow->conn_finished. */ 1375 return sk->sk_state == TCP_SYN_RECV && is_active_ssk(subflow); 1376 } 1377 1378 #ifdef CONFIG_SYN_COOKIES 1379 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1380 struct sk_buff *skb); 1381 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1382 struct sk_buff *skb); 1383 void __init mptcp_join_cookie_init(void); 1384 #else 1385 static inline void 1386 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1387 struct sk_buff *skb) {} 1388 static inline bool 1389 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1390 struct sk_buff *skb) 1391 { 1392 return false; 1393 } 1394 1395 static inline void mptcp_join_cookie_init(void) {} 1396 #endif 1397 1398 #endif /* __MPTCP_PROTOCOL_H */ 1399