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 16 #include "mptcp_pm_gen.h" 17 18 #define MPTCP_SUPPORTED_VERSION 1 19 20 /* MPTCP option bits */ 21 #define OPTION_MPTCP_MPC_SYN BIT(0) 22 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 23 #define OPTION_MPTCP_MPC_ACK BIT(2) 24 #define OPTION_MPTCP_MPJ_SYN BIT(3) 25 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 26 #define OPTION_MPTCP_MPJ_ACK BIT(5) 27 #define OPTION_MPTCP_ADD_ADDR BIT(6) 28 #define OPTION_MPTCP_RM_ADDR BIT(7) 29 #define OPTION_MPTCP_FASTCLOSE BIT(8) 30 #define OPTION_MPTCP_PRIO BIT(9) 31 #define OPTION_MPTCP_RST BIT(10) 32 #define OPTION_MPTCP_DSS BIT(11) 33 #define OPTION_MPTCP_FAIL BIT(12) 34 35 #define OPTION_MPTCP_CSUMREQD BIT(13) 36 37 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \ 38 OPTION_MPTCP_MPC_ACK) 39 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \ 40 OPTION_MPTCP_MPJ_ACK) 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_NOSPACE 1 117 #define MPTCP_WORK_RTX 2 118 #define MPTCP_FALLBACK_DONE 4 119 #define MPTCP_WORK_CLOSE_SUBFLOW 5 120 121 /* 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:1; 135 }; 136 137 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0])) 138 139 static inline bool before64(__u64 seq1, __u64 seq2) 140 { 141 return (__s64)(seq1 - seq2) < 0; 142 } 143 144 #define after64(seq2, seq1) before64(seq1, seq2) 145 146 struct mptcp_options_received { 147 u64 sndr_key; 148 u64 rcvr_key; 149 u64 data_ack; 150 u64 data_seq; 151 u32 subflow_seq; 152 u16 data_len; 153 __sum16 csum; 154 u16 suboptions; 155 u32 token; 156 u32 nonce; 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 u8 join_id; 170 u64 thmac; 171 u8 hmac[MPTCPOPT_HMAC_LEN]; 172 struct mptcp_addr_info addr; 173 struct mptcp_rm_list rm_list; 174 u64 ahmac; 175 u64 fail_seq; 176 }; 177 178 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 179 { 180 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 181 ((nib & 0xF) << 8) | field); 182 } 183 184 enum mptcp_pm_status { 185 MPTCP_PM_ADD_ADDR_RECEIVED, 186 MPTCP_PM_ADD_ADDR_SEND_ACK, 187 MPTCP_PM_RM_ADDR_RECEIVED, 188 MPTCP_PM_ESTABLISHED, 189 MPTCP_PM_SUBFLOW_ESTABLISHED, 190 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 191 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 192 * accounted int id_avail_bitmap 193 */ 194 }; 195 196 enum mptcp_pm_type { 197 MPTCP_PM_TYPE_KERNEL = 0, 198 MPTCP_PM_TYPE_USERSPACE, 199 200 __MPTCP_PM_TYPE_NR, 201 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 202 }; 203 204 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 205 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 206 207 enum mptcp_addr_signal_status { 208 MPTCP_ADD_ADDR_SIGNAL, 209 MPTCP_ADD_ADDR_ECHO, 210 MPTCP_RM_ADDR_SIGNAL, 211 }; 212 213 /* max value of mptcp_addr_info.id */ 214 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 215 216 struct mptcp_pm_data { 217 struct mptcp_addr_info local; 218 struct mptcp_addr_info remote; 219 struct list_head anno_list; 220 struct list_head userspace_pm_local_addr_list; 221 222 spinlock_t lock; /*protects the whole PM data */ 223 224 u8 addr_signal; 225 bool server_side; 226 bool work_pending; 227 bool accept_addr; 228 bool accept_subflow; 229 bool remote_deny_join_id0; 230 u8 add_addr_signaled; 231 u8 add_addr_accepted; 232 u8 local_addr_used; 233 u8 pm_type; 234 u8 subflows; 235 u8 status; 236 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 237 struct mptcp_rm_list rm_list_tx; 238 struct mptcp_rm_list rm_list_rx; 239 }; 240 241 struct mptcp_pm_addr_entry { 242 struct list_head list; 243 struct mptcp_addr_info addr; 244 u8 flags; 245 int ifindex; 246 struct socket *lsk; 247 }; 248 249 struct mptcp_data_frag { 250 struct list_head list; 251 u64 data_seq; 252 u16 data_len; 253 u16 offset; 254 u16 overhead; 255 u16 already_sent; 256 struct page *page; 257 }; 258 259 /* MPTCP connection sock */ 260 struct mptcp_sock { 261 /* inet_connection_sock must be the first member */ 262 struct inet_connection_sock sk; 263 u64 local_key; /* protected by the first subflow socket lock 264 * lockless access read 265 */ 266 u64 remote_key; /* same as above */ 267 u64 write_seq; 268 u64 bytes_sent; 269 u64 snd_nxt; 270 u64 bytes_received; 271 u64 ack_seq; 272 atomic64_t rcv_wnd_sent; 273 u64 rcv_data_fin_seq; 274 u64 bytes_retrans; 275 u64 bytes_consumed; 276 int rmem_fwd_alloc; 277 int snd_burst; 278 int old_wspace; 279 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 280 * recovery related fields are under data_lock 281 * protection 282 */ 283 u64 bytes_acked; 284 u64 snd_una; 285 u64 wnd_end; 286 unsigned long timer_ival; 287 u32 token; 288 int rmem_released; 289 unsigned long flags; 290 unsigned long cb_flags; 291 bool recovery; /* closing subflow write queue reinjected */ 292 bool can_ack; 293 bool fully_established; 294 bool rcv_data_fin; 295 bool snd_data_fin_enable; 296 bool rcv_fastclose; 297 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 298 bool csum_enabled; 299 bool allow_infinite_fallback; 300 u8 pending_state; /* A subflow asked to set this sk_state, 301 * protected by the msk data lock 302 */ 303 u8 mpc_endpoint_id; 304 u8 recvmsg_inq:1, 305 cork:1, 306 nodelay:1, 307 fastopening:1, 308 in_accept_queue:1, 309 free_first:1, 310 rcvspace_init:1; 311 struct work_struct work; 312 struct sk_buff *ooo_last_skb; 313 struct rb_root out_of_order_queue; 314 struct sk_buff_head receive_queue; 315 struct list_head conn_list; 316 struct list_head rtx_queue; 317 struct mptcp_data_frag *first_pending; 318 struct list_head join_list; 319 struct sock *first; /* The mptcp ops can safely dereference, using suitable 320 * ONCE annotation, the subflow outside the socket 321 * lock as such sock is freed after close(). 322 */ 323 struct mptcp_pm_data pm; 324 struct mptcp_sched_ops *sched; 325 struct { 326 u32 space; /* bytes copied in last measurement window */ 327 u32 copied; /* bytes copied in this measurement window */ 328 u64 time; /* start time of measurement window */ 329 u64 rtt_us; /* last maximum rtt of subflows */ 330 } rcvq_space; 331 u8 scaling_ratio; 332 333 u32 subflow_id; 334 u32 setsockopt_seq; 335 char ca_name[TCP_CA_NAME_MAX]; 336 }; 337 338 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 339 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 340 341 #define mptcp_for_each_subflow(__msk, __subflow) \ 342 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 343 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 344 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 345 346 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 347 { 348 sock_owned_by_me((const struct sock *)msk); 349 } 350 351 #ifdef CONFIG_DEBUG_NET 352 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 353 #undef tcp_sk 354 #define tcp_sk(ptr) ({ \ 355 typeof(ptr) _ptr = (ptr); \ 356 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 357 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 358 }) 359 #define mptcp_sk(ptr) ({ \ 360 typeof(ptr) _ptr = (ptr); \ 361 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 362 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 363 }) 364 365 #else /* !CONFIG_DEBUG_NET */ 366 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 367 #endif 368 369 /* the msk socket don't use the backlog, also account for the bulk 370 * free memory 371 */ 372 static inline int __mptcp_rmem(const struct sock *sk) 373 { 374 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 375 } 376 377 static inline int mptcp_win_from_space(const struct sock *sk, int space) 378 { 379 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 380 } 381 382 static inline int __mptcp_space(const struct sock *sk) 383 { 384 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 385 } 386 387 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 388 { 389 const struct mptcp_sock *msk = mptcp_sk(sk); 390 391 return READ_ONCE(msk->first_pending); 392 } 393 394 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 395 { 396 struct mptcp_sock *msk = mptcp_sk(sk); 397 struct mptcp_data_frag *cur; 398 399 cur = msk->first_pending; 400 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 401 list_next_entry(cur, list); 402 } 403 404 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 405 { 406 const struct mptcp_sock *msk = mptcp_sk(sk); 407 408 if (!msk->first_pending) 409 return NULL; 410 411 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 412 return NULL; 413 414 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 415 } 416 417 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 418 { 419 struct mptcp_sock *msk = mptcp_sk(sk); 420 421 if (msk->snd_una == msk->snd_nxt) 422 return NULL; 423 424 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 425 } 426 427 struct csum_pseudo_header { 428 __be64 data_seq; 429 __be32 subflow_seq; 430 __be16 data_len; 431 __sum16 csum; 432 }; 433 434 struct mptcp_subflow_request_sock { 435 struct tcp_request_sock sk; 436 u16 mp_capable : 1, 437 mp_join : 1, 438 backup : 1, 439 csum_reqd : 1, 440 allow_join_id0 : 1; 441 u8 local_id; 442 u8 remote_id; 443 u64 local_key; 444 u64 idsn; 445 u32 token; 446 u32 ssn_offset; 447 u64 thmac; 448 u32 local_nonce; 449 u32 remote_nonce; 450 struct mptcp_sock *msk; 451 struct hlist_nulls_node token_node; 452 }; 453 454 static inline struct mptcp_subflow_request_sock * 455 mptcp_subflow_rsk(const struct request_sock *rsk) 456 { 457 return (struct mptcp_subflow_request_sock *)rsk; 458 } 459 460 struct mptcp_delegated_action { 461 struct napi_struct napi; 462 struct list_head head; 463 }; 464 465 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 466 467 #define MPTCP_DELEGATE_SCHEDULED 0 468 #define MPTCP_DELEGATE_SEND 1 469 #define MPTCP_DELEGATE_ACK 2 470 #define MPTCP_DELEGATE_SNDBUF 3 471 472 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 473 /* MPTCP subflow context */ 474 struct mptcp_subflow_context { 475 struct list_head node;/* conn_list of subflows */ 476 477 struct_group(reset, 478 479 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 480 u64 local_key; 481 u64 remote_key; 482 u64 idsn; 483 u64 map_seq; 484 u32 snd_isn; 485 u32 token; 486 u32 rel_write_seq; 487 u32 map_subflow_seq; 488 u32 ssn_offset; 489 u32 map_data_len; 490 __wsum map_data_csum; 491 u32 map_csum_len; 492 u32 request_mptcp : 1, /* send MP_CAPABLE */ 493 request_join : 1, /* send MP_JOIN */ 494 request_bkup : 1, 495 mp_capable : 1, /* remote is MPTCP capable */ 496 mp_join : 1, /* remote is JOINing */ 497 fully_established : 1, /* path validated */ 498 pm_notified : 1, /* PM hook called for established status */ 499 conn_finished : 1, 500 map_valid : 1, 501 map_csum_reqd : 1, 502 map_data_fin : 1, 503 mpc_map : 1, 504 backup : 1, 505 send_mp_prio : 1, 506 send_mp_fail : 1, 507 send_fastclose : 1, 508 send_infinite_map : 1, 509 remote_key_valid : 1, /* received the peer key from */ 510 disposable : 1, /* ctx can be free at ulp release time */ 511 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 512 valid_csum_seen : 1, /* at least one csum validated */ 513 is_mptfo : 1, /* subflow is doing TFO */ 514 __unused : 10; 515 bool data_avail; 516 bool scheduled; 517 u32 remote_nonce; 518 u64 thmac; 519 u32 local_nonce; 520 u32 remote_token; 521 union { 522 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 523 u64 iasn; /* initial ack sequence number, MPC subflows only */ 524 }; 525 s16 local_id; /* if negative not initialized yet */ 526 u8 remote_id; 527 u8 reset_seen:1; 528 u8 reset_transient:1; 529 u8 reset_reason:4; 530 u8 stale_count; 531 532 u32 subflow_id; 533 534 long delegated_status; 535 unsigned long fail_tout; 536 537 ); 538 539 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 540 541 u32 setsockopt_seq; 542 u32 stale_rcv_tstamp; 543 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 544 * protected by the msk socket lock 545 */ 546 547 struct sock *tcp_sock; /* tcp sk backpointer */ 548 struct sock *conn; /* parent mptcp_sock */ 549 const struct inet_connection_sock_af_ops *icsk_af_ops; 550 void (*tcp_state_change)(struct sock *sk); 551 void (*tcp_error_report)(struct sock *sk); 552 553 struct rcu_head rcu; 554 }; 555 556 static inline struct mptcp_subflow_context * 557 mptcp_subflow_ctx(const struct sock *sk) 558 { 559 struct inet_connection_sock *icsk = inet_csk(sk); 560 561 /* Use RCU on icsk_ulp_data only for sock diag code */ 562 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 563 } 564 565 static inline struct sock * 566 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 567 { 568 return subflow->tcp_sock; 569 } 570 571 static inline void 572 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 573 { 574 memset(&subflow->reset, 0, sizeof(subflow->reset)); 575 subflow->request_mptcp = 1; 576 WRITE_ONCE(subflow->local_id, -1); 577 } 578 579 static inline u64 580 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 581 { 582 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 583 subflow->ssn_offset - 584 subflow->map_subflow_seq; 585 } 586 587 static inline u64 588 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 589 { 590 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 591 } 592 593 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 594 595 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 596 { 597 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 598 struct mptcp_delegated_action *delegated; 599 bool schedule; 600 601 /* the caller held the subflow bh socket lock */ 602 lockdep_assert_in_softirq(); 603 604 /* The implied barrier pairs with tcp_release_cb_override() 605 * mptcp_napi_poll(), and ensures the below list check sees list 606 * updates done prior to delegated status bits changes 607 */ 608 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 609 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 610 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 611 return; 612 613 delegated = this_cpu_ptr(&mptcp_delegated_actions); 614 schedule = list_empty(&delegated->head); 615 list_add_tail(&subflow->delegated_node, &delegated->head); 616 sock_hold(mptcp_subflow_tcp_sock(subflow)); 617 if (schedule) 618 napi_schedule(&delegated->napi); 619 } 620 } 621 622 static inline struct mptcp_subflow_context * 623 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 624 { 625 struct mptcp_subflow_context *ret; 626 627 if (list_empty(&delegated->head)) 628 return NULL; 629 630 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 631 list_del_init(&ret->delegated_node); 632 return ret; 633 } 634 635 int mptcp_is_enabled(const struct net *net); 636 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 637 int mptcp_is_checksum_enabled(const struct net *net); 638 int mptcp_allow_join_id0(const struct net *net); 639 unsigned int mptcp_stale_loss_cnt(const struct net *net); 640 unsigned int mptcp_close_timeout(const struct sock *sk); 641 int mptcp_get_pm_type(const struct net *net); 642 const char *mptcp_get_scheduler(const struct net *net); 643 void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 644 struct mptcp_subflow_context *subflow, 645 const struct mptcp_options_received *mp_opt); 646 bool __mptcp_retransmit_pending_data(struct sock *sk); 647 void mptcp_check_and_set_pending(struct sock *sk); 648 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 649 bool mptcp_subflow_data_available(struct sock *sk); 650 void __init mptcp_subflow_init(void); 651 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 652 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 653 struct mptcp_subflow_context *subflow); 654 void __mptcp_subflow_send_ack(struct sock *ssk); 655 void mptcp_subflow_reset(struct sock *ssk); 656 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 657 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 658 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 659 bool __mptcp_close(struct sock *sk, long timeout); 660 void mptcp_cancel_work(struct sock *sk); 661 void __mptcp_unaccepted_force_close(struct sock *sk); 662 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk); 663 void mptcp_set_state(struct sock *sk, int state); 664 665 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 666 const struct mptcp_addr_info *b, bool use_port); 667 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr); 668 669 /* called with sk socket lock held */ 670 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 671 const struct mptcp_addr_info *remote); 672 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 673 struct socket **new_sock); 674 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 675 struct sockaddr_storage *addr, 676 unsigned short family); 677 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 678 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 679 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 680 void mptcp_sched_init(void); 681 int mptcp_init_sched(struct mptcp_sock *msk, 682 struct mptcp_sched_ops *sched); 683 void mptcp_release_sched(struct mptcp_sock *msk); 684 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 685 bool scheduled); 686 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 687 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 688 int mptcp_sched_get_send(struct mptcp_sock *msk); 689 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 690 691 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 692 { 693 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 694 } 695 696 static inline bool mptcp_epollin_ready(const struct sock *sk) 697 { 698 /* mptcp doesn't have to deal with small skbs in the receive queue, 699 * at it can always coalesce them 700 */ 701 return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) || 702 (mem_cgroup_sockets_enabled && sk->sk_memcg && 703 mem_cgroup_under_socket_pressure(sk->sk_memcg)) || 704 READ_ONCE(tcp_memory_pressure); 705 } 706 707 int mptcp_set_rcvlowat(struct sock *sk, int val); 708 709 static inline bool __tcp_can_send(const struct sock *ssk) 710 { 711 /* only send if our side has not closed yet */ 712 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 713 } 714 715 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 716 { 717 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 718 if (subflow->request_join && !subflow->fully_established) 719 return false; 720 721 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 722 } 723 724 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 725 726 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 727 728 void mptcp_subflow_drop_ctx(struct sock *ssk); 729 730 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 731 struct mptcp_subflow_context *ctx) 732 { 733 sk->sk_data_ready = sock_def_readable; 734 sk->sk_state_change = ctx->tcp_state_change; 735 sk->sk_write_space = sk_stream_write_space; 736 sk->sk_error_report = ctx->tcp_error_report; 737 738 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 739 } 740 741 void __init mptcp_proto_init(void); 742 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 743 int __init mptcp_proto_v6_init(void); 744 #endif 745 746 struct sock *mptcp_sk_clone_init(const struct sock *sk, 747 const struct mptcp_options_received *mp_opt, 748 struct sock *ssk, 749 struct request_sock *req); 750 void mptcp_get_options(const struct sk_buff *skb, 751 struct mptcp_options_received *mp_opt); 752 753 void mptcp_finish_connect(struct sock *sk); 754 void __mptcp_sync_state(struct sock *sk, int state); 755 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 756 757 static inline void mptcp_stop_tout_timer(struct sock *sk) 758 { 759 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 760 return; 761 762 sk_stop_timer(sk, &sk->sk_timer); 763 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 764 } 765 766 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 767 { 768 /* avoid 0 timestamp, as that means no close timeout */ 769 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 770 } 771 772 static inline void mptcp_start_tout_timer(struct sock *sk) 773 { 774 mptcp_set_close_tout(sk, tcp_jiffies32); 775 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 776 } 777 778 static inline bool mptcp_is_fully_established(struct sock *sk) 779 { 780 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 781 READ_ONCE(mptcp_sk(sk)->fully_established); 782 } 783 784 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 785 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 786 bool mptcp_finish_join(struct sock *sk); 787 bool mptcp_schedule_work(struct sock *sk); 788 int mptcp_setsockopt(struct sock *sk, int level, int optname, 789 sockptr_t optval, unsigned int optlen); 790 int mptcp_getsockopt(struct sock *sk, int level, int optname, 791 char __user *optval, int __user *option); 792 793 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 794 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 795 { 796 if (use_64bit) 797 return cur_seq; 798 799 return __mptcp_expand_seq(old_seq, cur_seq); 800 } 801 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 802 void __mptcp_data_acked(struct sock *sk); 803 void __mptcp_error_report(struct sock *sk); 804 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 805 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 806 { 807 return READ_ONCE(msk->snd_data_fin_enable) && 808 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 809 } 810 811 static inline void mptcp_write_space(struct sock *sk) 812 { 813 if (sk_stream_is_writeable(sk)) { 814 /* pairs with memory barrier in mptcp_poll */ 815 smp_mb(); 816 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 817 sk_stream_write_space(sk); 818 } 819 } 820 821 static inline void __mptcp_sync_sndbuf(struct sock *sk) 822 { 823 struct mptcp_subflow_context *subflow; 824 int ssk_sndbuf, new_sndbuf; 825 826 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 827 return; 828 829 new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0]; 830 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 831 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 832 833 subflow->cached_sndbuf = ssk_sndbuf; 834 new_sndbuf += ssk_sndbuf; 835 } 836 837 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 838 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 839 mptcp_write_space(sk); 840 } 841 842 /* The called held both the msk socket and the subflow socket locks, 843 * possibly under BH 844 */ 845 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 846 { 847 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 848 849 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 850 __mptcp_sync_sndbuf(sk); 851 } 852 853 /* the caller held only the subflow socket lock, either in process or 854 * BH context. Additionally this can be called under the msk data lock, 855 * so we can't acquire such lock here: let the delegate action acquires 856 * the needed locks in suitable order. 857 */ 858 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 859 { 860 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 861 862 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 863 return; 864 865 local_bh_disable(); 866 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 867 local_bh_enable(); 868 } 869 870 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 871 872 #define MPTCP_TOKEN_MAX_RETRIES 4 873 874 void __init mptcp_token_init(void); 875 static inline void mptcp_token_init_request(struct request_sock *req) 876 { 877 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 878 } 879 880 int mptcp_token_new_request(struct request_sock *req); 881 void mptcp_token_destroy_request(struct request_sock *req); 882 int mptcp_token_new_connect(struct sock *ssk); 883 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 884 struct mptcp_sock *msk); 885 bool mptcp_token_exists(u32 token); 886 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 887 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 888 long *s_num); 889 void mptcp_token_destroy(struct mptcp_sock *msk); 890 891 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 892 893 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 894 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 895 896 void __init mptcp_pm_init(void); 897 void mptcp_pm_data_init(struct mptcp_sock *msk); 898 void mptcp_pm_data_reset(struct mptcp_sock *msk); 899 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 900 struct mptcp_addr_info *addr); 901 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 902 bool require_family, 903 struct mptcp_pm_addr_entry *entry); 904 bool mptcp_pm_addr_families_match(const struct sock *sk, 905 const struct mptcp_addr_info *loc, 906 const struct mptcp_addr_info *rem); 907 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 908 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 909 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 910 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 911 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 912 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 913 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 914 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 915 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 916 const struct mptcp_subflow_context *subflow); 917 void mptcp_pm_add_addr_received(const struct sock *ssk, 918 const struct mptcp_addr_info *addr); 919 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 920 const struct mptcp_addr_info *addr); 921 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 922 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 923 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 924 const struct mptcp_rm_list *rm_list); 925 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 926 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 927 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 928 struct mptcp_addr_info *addr, 929 struct mptcp_addr_info *rem, 930 u8 bkup); 931 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 932 const struct mptcp_addr_info *addr); 933 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 934 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 935 struct mptcp_pm_add_entry * 936 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 937 const struct mptcp_addr_info *addr, bool check_id); 938 struct mptcp_pm_add_entry * 939 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 940 const struct mptcp_addr_info *addr); 941 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 942 unsigned int id, 943 u8 *flags, int *ifindex); 944 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id, 945 u8 *flags, int *ifindex); 946 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 947 unsigned int id, 948 u8 *flags, int *ifindex); 949 int mptcp_pm_set_flags(struct net *net, struct nlattr *token, 950 struct mptcp_pm_addr_entry *loc, 951 struct mptcp_pm_addr_entry *rem, u8 bkup); 952 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup); 953 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token, 954 struct mptcp_pm_addr_entry *loc, 955 struct mptcp_pm_addr_entry *rem, u8 bkup); 956 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 957 const struct mptcp_addr_info *addr, 958 bool echo); 959 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 960 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 961 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list); 962 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk, 963 struct list_head *rm_list); 964 965 void mptcp_free_local_addr_list(struct mptcp_sock *msk); 966 967 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 968 const struct sock *ssk, gfp_t gfp); 969 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 970 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 971 void mptcp_event_pm_listener(const struct sock *ssk, 972 enum mptcp_event_type event); 973 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 974 975 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow, 976 const struct mptcp_options_received *mp_opt); 977 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 978 struct request_sock *req); 979 980 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 981 { 982 return READ_ONCE(msk->pm.addr_signal) & 983 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 984 } 985 986 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 987 { 988 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 989 } 990 991 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 992 { 993 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 994 } 995 996 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 997 { 998 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 999 } 1000 1001 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1002 { 1003 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1004 } 1005 1006 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1007 { 1008 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1009 } 1010 1011 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 1012 { 1013 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 1014 1015 if (family == AF_INET6) 1016 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 1017 if (!echo) 1018 len += MPTCPOPT_THMAC_LEN; 1019 /* account for 2 trailing 'nop' options */ 1020 if (port) 1021 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 1022 1023 return len; 1024 } 1025 1026 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 1027 { 1028 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 1029 return -EINVAL; 1030 1031 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 1032 } 1033 1034 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 1035 unsigned int opt_size, unsigned int remaining, 1036 struct mptcp_addr_info *addr, bool *echo, 1037 bool *drop_other_suboptions); 1038 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1039 struct mptcp_rm_list *rm_list); 1040 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1041 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1042 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1043 1044 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1045 { 1046 int local_id = READ_ONCE(subflow->local_id); 1047 1048 if (local_id < 0) 1049 return 0; 1050 return local_id; 1051 } 1052 1053 void __init mptcp_pm_nl_init(void); 1054 void mptcp_pm_nl_work(struct mptcp_sock *msk); 1055 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 1056 const struct mptcp_rm_list *rm_list); 1057 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 1058 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 1059 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 1060 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 1061 1062 /* called under PM lock */ 1063 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1064 { 1065 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 1066 WRITE_ONCE(msk->pm.accept_subflow, true); 1067 } 1068 1069 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1070 { 1071 spin_lock_bh(&msk->pm.lock); 1072 __mptcp_pm_close_subflow(msk); 1073 spin_unlock_bh(&msk->pm.lock); 1074 } 1075 1076 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 1077 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1078 1079 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1080 { 1081 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1082 } 1083 1084 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1085 1086 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1087 { 1088 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1089 } 1090 1091 static inline bool mptcp_check_fallback(const struct sock *sk) 1092 { 1093 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1094 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1095 1096 return __mptcp_check_fallback(msk); 1097 } 1098 1099 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 1100 { 1101 if (__mptcp_check_fallback(msk)) { 1102 pr_debug("TCP fallback already done (msk=%p)", msk); 1103 return; 1104 } 1105 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1106 } 1107 1108 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1109 { 1110 struct sock *ssk = READ_ONCE(msk->first); 1111 1112 return ssk && ((1 << inet_sk_state_load(ssk)) & 1113 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1114 TCPF_SYN_RECV | TCPF_LISTEN)); 1115 } 1116 1117 static inline void mptcp_do_fallback(struct sock *ssk) 1118 { 1119 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1120 struct sock *sk = subflow->conn; 1121 struct mptcp_sock *msk; 1122 1123 msk = mptcp_sk(sk); 1124 __mptcp_do_fallback(msk); 1125 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1126 gfp_t saved_allocation = ssk->sk_allocation; 1127 1128 /* we are in a atomic (BH) scope, override ssk default for data 1129 * fin allocation 1130 */ 1131 ssk->sk_allocation = GFP_ATOMIC; 1132 ssk->sk_shutdown |= SEND_SHUTDOWN; 1133 tcp_shutdown(ssk, SEND_SHUTDOWN); 1134 ssk->sk_allocation = saved_allocation; 1135 } 1136 } 1137 1138 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 1139 1140 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1141 { 1142 struct mptcp_ext *mpext; 1143 1144 mpext = skb ? mptcp_get_ext(skb) : NULL; 1145 if (mpext && mpext->infinite_map) 1146 return true; 1147 1148 return false; 1149 } 1150 1151 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1152 { 1153 return (subflow->request_mptcp || subflow->request_join); 1154 } 1155 1156 static inline bool subflow_simultaneous_connect(struct sock *sk) 1157 { 1158 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1159 1160 return (1 << sk->sk_state) & 1161 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) && 1162 is_active_ssk(subflow) && 1163 !subflow->conn_finished; 1164 } 1165 1166 #ifdef CONFIG_SYN_COOKIES 1167 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1168 struct sk_buff *skb); 1169 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1170 struct sk_buff *skb); 1171 void __init mptcp_join_cookie_init(void); 1172 #else 1173 static inline void 1174 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1175 struct sk_buff *skb) {} 1176 static inline bool 1177 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1178 struct sk_buff *skb) 1179 { 1180 return false; 1181 } 1182 1183 static inline void mptcp_join_cookie_init(void) {} 1184 #endif 1185 1186 #endif /* __MPTCP_PROTOCOL_H */ 1187