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_WORK_RTX 1 117#define MPTCP_FALLBACK_DONE 2 118#define MPTCP_WORK_CLOSE_SUBFLOW 3 119 120/* MPTCP socket release cb flags */ 121#define MPTCP_PUSH_PENDING 1 122#define MPTCP_CLEAN_UNA 2 123#define MPTCP_ERROR_REPORT 3 124#define MPTCP_RETRANSMIT 4 125#define MPTCP_FLUSH_JOIN_LIST 5 126#define MPTCP_SYNC_STATE 6 127#define MPTCP_SYNC_SNDBUF 7 128 129struct mptcp_skb_cb { 130 u64 map_seq; 131 u64 end_seq; 132 u32 offset; 133 u8 has_rxtstamp:1; 134}; 135 136#define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0])) 137 138static 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 145struct 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 u16 suboptions; 154 u32 token; 155 u32 nonce; 156 u16 use_map:1, 157 dsn64:1, 158 data_fin:1, 159 use_ack:1, 160 ack64:1, 161 mpc_map:1, 162 reset_reason:4, 163 reset_transient:1, 164 echo:1, 165 backup:1, 166 deny_join_id0:1, 167 __unused:2; 168 u8 join_id; 169 u64 thmac; 170 u8 hmac[MPTCPOPT_HMAC_LEN]; 171 struct mptcp_addr_info addr; 172 struct mptcp_rm_list rm_list; 173 u64 ahmac; 174 u64 fail_seq; 175}; 176 177static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 178{ 179 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 180 ((nib & 0xF) << 8) | field); 181} 182 183enum mptcp_pm_status { 184 MPTCP_PM_ADD_ADDR_RECEIVED, 185 MPTCP_PM_ADD_ADDR_SEND_ACK, 186 MPTCP_PM_RM_ADDR_RECEIVED, 187 MPTCP_PM_ESTABLISHED, 188 MPTCP_PM_SUBFLOW_ESTABLISHED, 189 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 190 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 191 * accounted int id_avail_bitmap 192 */ 193}; 194 195enum mptcp_pm_type { 196 MPTCP_PM_TYPE_KERNEL = 0, 197 MPTCP_PM_TYPE_USERSPACE, 198 199 __MPTCP_PM_TYPE_NR, 200 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 201}; 202 203/* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 204#define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 205 206enum mptcp_addr_signal_status { 207 MPTCP_ADD_ADDR_SIGNAL, 208 MPTCP_ADD_ADDR_ECHO, 209 MPTCP_RM_ADDR_SIGNAL, 210}; 211 212/* max value of mptcp_addr_info.id */ 213#define MPTCP_PM_MAX_ADDR_ID U8_MAX 214 215struct mptcp_pm_data { 216 struct mptcp_addr_info local; 217 struct mptcp_addr_info remote; 218 struct list_head anno_list; 219 struct list_head userspace_pm_local_addr_list; 220 221 spinlock_t lock; /*protects the whole PM data */ 222 223 u8 addr_signal; 224 bool server_side; 225 bool work_pending; 226 bool accept_addr; 227 bool accept_subflow; 228 bool remote_deny_join_id0; 229 u8 add_addr_signaled; 230 u8 add_addr_accepted; 231 u8 local_addr_used; 232 u8 pm_type; 233 u8 subflows; 234 u8 status; 235 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 236 struct mptcp_rm_list rm_list_tx; 237 struct mptcp_rm_list rm_list_rx; 238}; 239 240struct mptcp_pm_addr_entry { 241 struct list_head list; 242 struct mptcp_addr_info addr; 243 u8 flags; 244 int ifindex; 245 struct socket *lsk; 246}; 247 248struct mptcp_data_frag { 249 struct list_head list; 250 u64 data_seq; 251 u16 data_len; 252 u16 offset; 253 u16 overhead; 254 u16 already_sent; 255 struct page *page; 256}; 257 258/* MPTCP connection sock */ 259struct mptcp_sock { 260 /* inet_connection_sock must be the first member */ 261 struct inet_connection_sock sk; 262 u64 local_key; /* protected by the first subflow socket lock 263 * lockless access read 264 */ 265 u64 remote_key; /* same as above */ 266 u64 write_seq; 267 u64 bytes_sent; 268 u64 snd_nxt; 269 u64 bytes_received; 270 u64 ack_seq; 271 atomic64_t rcv_wnd_sent; 272 u64 rcv_data_fin_seq; 273 u64 bytes_retrans; 274 u64 bytes_consumed; 275 int rmem_fwd_alloc; 276 int snd_burst; 277 int old_wspace; 278 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 279 * recovery related fields are under data_lock 280 * protection 281 */ 282 u64 bytes_acked; 283 u64 snd_una; 284 u64 wnd_end; 285 unsigned long timer_ival; 286 u32 token; 287 int rmem_released; 288 unsigned long flags; 289 unsigned long cb_flags; 290 bool recovery; /* closing subflow write queue reinjected */ 291 bool can_ack; 292 bool fully_established; 293 bool rcv_data_fin; 294 bool snd_data_fin_enable; 295 bool rcv_fastclose; 296 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 297 bool csum_enabled; 298 bool allow_infinite_fallback; 299 u8 pending_state; /* A subflow asked to set this sk_state, 300 * protected by the msk data lock 301 */ 302 u8 mpc_endpoint_id; 303 u8 recvmsg_inq:1, 304 cork:1, 305 nodelay:1, 306 fastopening:1, 307 in_accept_queue:1, 308 free_first:1, 309 rcvspace_init:1; 310 u32 notsent_lowat; 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 346extern struct genl_family mptcp_genl_family; 347 348static inline void msk_owned_by_me(const struct mptcp_sock *msk) 349{ 350 sock_owned_by_me((const struct sock *)msk); 351} 352 353#ifdef CONFIG_DEBUG_NET 354/* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 355#undef tcp_sk 356#define tcp_sk(ptr) ({ \ 357 typeof(ptr) _ptr = (ptr); \ 358 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 359 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 360}) 361#define mptcp_sk(ptr) ({ \ 362 typeof(ptr) _ptr = (ptr); \ 363 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 364 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 365}) 366 367#else /* !CONFIG_DEBUG_NET */ 368#define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 369#endif 370 371/* the msk socket don't use the backlog, also account for the bulk 372 * free memory 373 */ 374static inline int __mptcp_rmem(const struct sock *sk) 375{ 376 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 377} 378 379static inline int mptcp_win_from_space(const struct sock *sk, int space) 380{ 381 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 382} 383 384static inline int __mptcp_space(const struct sock *sk) 385{ 386 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 387} 388 389static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 390{ 391 const struct mptcp_sock *msk = mptcp_sk(sk); 392 393 return READ_ONCE(msk->first_pending); 394} 395 396static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 397{ 398 struct mptcp_sock *msk = mptcp_sk(sk); 399 struct mptcp_data_frag *cur; 400 401 cur = msk->first_pending; 402 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 403 list_next_entry(cur, list); 404} 405 406static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 407{ 408 const struct mptcp_sock *msk = mptcp_sk(sk); 409 410 if (!msk->first_pending) 411 return NULL; 412 413 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 414 return NULL; 415 416 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 417} 418 419static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 420{ 421 struct mptcp_sock *msk = mptcp_sk(sk); 422 423 if (msk->snd_una == msk->snd_nxt) 424 return NULL; 425 426 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 427} 428 429struct csum_pseudo_header { 430 __be64 data_seq; 431 __be32 subflow_seq; 432 __be16 data_len; 433 __sum16 csum; 434}; 435 436struct mptcp_subflow_request_sock { 437 struct tcp_request_sock sk; 438 u16 mp_capable : 1, 439 mp_join : 1, 440 backup : 1, 441 csum_reqd : 1, 442 allow_join_id0 : 1; 443 u8 local_id; 444 u8 remote_id; 445 u64 local_key; 446 u64 idsn; 447 u32 token; 448 u32 ssn_offset; 449 u64 thmac; 450 u32 local_nonce; 451 u32 remote_nonce; 452 struct mptcp_sock *msk; 453 struct hlist_nulls_node token_node; 454}; 455 456static inline struct mptcp_subflow_request_sock * 457mptcp_subflow_rsk(const struct request_sock *rsk) 458{ 459 return (struct mptcp_subflow_request_sock *)rsk; 460} 461 462struct mptcp_delegated_action { 463 struct napi_struct napi; 464 struct list_head head; 465}; 466 467DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 468 469#define MPTCP_DELEGATE_SCHEDULED 0 470#define MPTCP_DELEGATE_SEND 1 471#define MPTCP_DELEGATE_ACK 2 472#define MPTCP_DELEGATE_SNDBUF 3 473 474#define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 475/* MPTCP subflow context */ 476struct mptcp_subflow_context { 477 struct list_head node;/* conn_list of subflows */ 478 479 struct_group(reset, 480 481 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 482 u64 local_key; 483 u64 remote_key; 484 u64 idsn; 485 u64 map_seq; 486 u32 snd_isn; 487 u32 token; 488 u32 rel_write_seq; 489 u32 map_subflow_seq; 490 u32 ssn_offset; 491 u32 map_data_len; 492 __wsum map_data_csum; 493 u32 map_csum_len; 494 u32 request_mptcp : 1, /* send MP_CAPABLE */ 495 request_join : 1, /* send MP_JOIN */ 496 request_bkup : 1, 497 mp_capable : 1, /* remote is MPTCP capable */ 498 mp_join : 1, /* remote is JOINing */ 499 fully_established : 1, /* path validated */ 500 pm_notified : 1, /* PM hook called for established status */ 501 conn_finished : 1, 502 map_valid : 1, 503 map_csum_reqd : 1, 504 map_data_fin : 1, 505 mpc_map : 1, 506 backup : 1, 507 send_mp_prio : 1, 508 send_mp_fail : 1, 509 send_fastclose : 1, 510 send_infinite_map : 1, 511 remote_key_valid : 1, /* received the peer key from */ 512 disposable : 1, /* ctx can be free at ulp release time */ 513 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 514 valid_csum_seen : 1, /* at least one csum validated */ 515 is_mptfo : 1, /* subflow is doing TFO */ 516 __unused : 10; 517 bool data_avail; 518 bool scheduled; 519 u32 remote_nonce; 520 u64 thmac; 521 u32 local_nonce; 522 u32 remote_token; 523 union { 524 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 525 u64 iasn; /* initial ack sequence number, MPC subflows only */ 526 }; 527 s16 local_id; /* if negative not initialized yet */ 528 u8 remote_id; 529 u8 reset_seen:1; 530 u8 reset_transient:1; 531 u8 reset_reason:4; 532 u8 stale_count; 533 534 u32 subflow_id; 535 536 long delegated_status; 537 unsigned long fail_tout; 538 539 ); 540 541 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 542 543 u32 setsockopt_seq; 544 u32 stale_rcv_tstamp; 545 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 546 * protected by the msk socket lock 547 */ 548 549 struct sock *tcp_sock; /* tcp sk backpointer */ 550 struct sock *conn; /* parent mptcp_sock */ 551 const struct inet_connection_sock_af_ops *icsk_af_ops; 552 void (*tcp_state_change)(struct sock *sk); 553 void (*tcp_error_report)(struct sock *sk); 554 555 struct rcu_head rcu; 556}; 557 558static inline struct mptcp_subflow_context * 559mptcp_subflow_ctx(const struct sock *sk) 560{ 561 struct inet_connection_sock *icsk = inet_csk(sk); 562 563 /* Use RCU on icsk_ulp_data only for sock diag code */ 564 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 565} 566 567static inline struct sock * 568mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 569{ 570 return subflow->tcp_sock; 571} 572 573static inline void 574mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 575{ 576 memset(&subflow->reset, 0, sizeof(subflow->reset)); 577 subflow->request_mptcp = 1; 578 WRITE_ONCE(subflow->local_id, -1); 579} 580 581static inline u64 582mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 583{ 584 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 585 subflow->ssn_offset - 586 subflow->map_subflow_seq; 587} 588 589static inline u64 590mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 591{ 592 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 593} 594 595void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 596 597static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 598{ 599 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 600 struct mptcp_delegated_action *delegated; 601 bool schedule; 602 603 /* the caller held the subflow bh socket lock */ 604 lockdep_assert_in_softirq(); 605 606 /* The implied barrier pairs with tcp_release_cb_override() 607 * mptcp_napi_poll(), and ensures the below list check sees list 608 * updates done prior to delegated status bits changes 609 */ 610 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 611 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 612 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 613 return; 614 615 delegated = this_cpu_ptr(&mptcp_delegated_actions); 616 schedule = list_empty(&delegated->head); 617 list_add_tail(&subflow->delegated_node, &delegated->head); 618 sock_hold(mptcp_subflow_tcp_sock(subflow)); 619 if (schedule) 620 napi_schedule(&delegated->napi); 621 } 622} 623 624static inline struct mptcp_subflow_context * 625mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 626{ 627 struct mptcp_subflow_context *ret; 628 629 if (list_empty(&delegated->head)) 630 return NULL; 631 632 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 633 list_del_init(&ret->delegated_node); 634 return ret; 635} 636 637int mptcp_is_enabled(const struct net *net); 638unsigned int mptcp_get_add_addr_timeout(const struct net *net); 639int mptcp_is_checksum_enabled(const struct net *net); 640int mptcp_allow_join_id0(const struct net *net); 641unsigned int mptcp_stale_loss_cnt(const struct net *net); 642unsigned int mptcp_close_timeout(const struct sock *sk); 643int mptcp_get_pm_type(const struct net *net); 644const char *mptcp_get_scheduler(const struct net *net); 645void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 646 struct mptcp_subflow_context *subflow, 647 const struct mptcp_options_received *mp_opt); 648bool __mptcp_retransmit_pending_data(struct sock *sk); 649void mptcp_check_and_set_pending(struct sock *sk); 650void __mptcp_push_pending(struct sock *sk, unsigned int flags); 651bool mptcp_subflow_data_available(struct sock *sk); 652void __init mptcp_subflow_init(void); 653void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 654void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 655 struct mptcp_subflow_context *subflow); 656void __mptcp_subflow_send_ack(struct sock *ssk); 657void mptcp_subflow_reset(struct sock *ssk); 658void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 659void mptcp_sock_graft(struct sock *sk, struct socket *parent); 660struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 661bool __mptcp_close(struct sock *sk, long timeout); 662void mptcp_cancel_work(struct sock *sk); 663void __mptcp_unaccepted_force_close(struct sock *sk); 664void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk); 665void mptcp_set_state(struct sock *sk, int state); 666 667bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 668 const struct mptcp_addr_info *b, bool use_port); 669void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr); 670 671/* called with sk socket lock held */ 672int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 673 const struct mptcp_addr_info *remote); 674int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 675 struct socket **new_sock); 676void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 677 struct sockaddr_storage *addr, 678 unsigned short family); 679struct mptcp_sched_ops *mptcp_sched_find(const char *name); 680int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 681void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 682void mptcp_sched_init(void); 683int mptcp_init_sched(struct mptcp_sock *msk, 684 struct mptcp_sched_ops *sched); 685void mptcp_release_sched(struct mptcp_sock *msk); 686void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 687 bool scheduled); 688struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 689struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 690int mptcp_sched_get_send(struct mptcp_sock *msk); 691int mptcp_sched_get_retrans(struct mptcp_sock *msk); 692 693static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 694{ 695 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 696} 697 698static inline bool mptcp_epollin_ready(const struct sock *sk) 699{ 700 /* mptcp doesn't have to deal with small skbs in the receive queue, 701 * at it can always coalesce them 702 */ 703 return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) || 704 (mem_cgroup_sockets_enabled && sk->sk_memcg && 705 mem_cgroup_under_socket_pressure(sk->sk_memcg)) || 706 READ_ONCE(tcp_memory_pressure); 707} 708 709int mptcp_set_rcvlowat(struct sock *sk, int val); 710 711static inline bool __tcp_can_send(const struct sock *ssk) 712{ 713 /* only send if our side has not closed yet */ 714 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 715} 716 717static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 718{ 719 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 720 if (subflow->request_join && !subflow->fully_established) 721 return false; 722 723 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 724} 725 726void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 727 728bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 729 730void mptcp_subflow_drop_ctx(struct sock *ssk); 731 732static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 733 struct mptcp_subflow_context *ctx) 734{ 735 sk->sk_data_ready = sock_def_readable; 736 sk->sk_state_change = ctx->tcp_state_change; 737 sk->sk_write_space = sk_stream_write_space; 738 sk->sk_error_report = ctx->tcp_error_report; 739 740 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 741} 742 743void __init mptcp_proto_init(void); 744#if IS_ENABLED(CONFIG_MPTCP_IPV6) 745int __init mptcp_proto_v6_init(void); 746#endif 747 748struct sock *mptcp_sk_clone_init(const struct sock *sk, 749 const struct mptcp_options_received *mp_opt, 750 struct sock *ssk, 751 struct request_sock *req); 752void mptcp_get_options(const struct sk_buff *skb, 753 struct mptcp_options_received *mp_opt); 754 755void mptcp_finish_connect(struct sock *sk); 756void __mptcp_sync_state(struct sock *sk, int state); 757void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 758 759static inline void mptcp_stop_tout_timer(struct sock *sk) 760{ 761 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 762 return; 763 764 sk_stop_timer(sk, &sk->sk_timer); 765 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 766} 767 768static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 769{ 770 /* avoid 0 timestamp, as that means no close timeout */ 771 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 772} 773 774static inline void mptcp_start_tout_timer(struct sock *sk) 775{ 776 mptcp_set_close_tout(sk, tcp_jiffies32); 777 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 778} 779 780static inline bool mptcp_is_fully_established(struct sock *sk) 781{ 782 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 783 READ_ONCE(mptcp_sk(sk)->fully_established); 784} 785 786void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 787void mptcp_data_ready(struct sock *sk, struct sock *ssk); 788bool mptcp_finish_join(struct sock *sk); 789bool mptcp_schedule_work(struct sock *sk); 790int mptcp_setsockopt(struct sock *sk, int level, int optname, 791 sockptr_t optval, unsigned int optlen); 792int mptcp_getsockopt(struct sock *sk, int level, int optname, 793 char __user *optval, int __user *option); 794 795u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 796static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 797{ 798 if (use_64bit) 799 return cur_seq; 800 801 return __mptcp_expand_seq(old_seq, cur_seq); 802} 803void __mptcp_check_push(struct sock *sk, struct sock *ssk); 804void __mptcp_data_acked(struct sock *sk); 805void __mptcp_error_report(struct sock *sk); 806bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 807static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 808{ 809 return READ_ONCE(msk->snd_data_fin_enable) && 810 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 811} 812 813static inline u32 mptcp_notsent_lowat(const struct sock *sk) 814{ 815 struct net *net = sock_net(sk); 816 u32 val; 817 818 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat); 819 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat); 820} 821 822static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake) 823{ 824 const struct mptcp_sock *msk = mptcp_sk(sk); 825 u32 notsent_bytes; 826 827 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt); 828 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk); 829} 830 831static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake) 832{ 833 return mptcp_stream_memory_free(sk, wake) && 834 __sk_stream_is_writeable(sk, wake); 835} 836 837static inline void mptcp_write_space(struct sock *sk) 838{ 839 /* pairs with memory barrier in mptcp_poll */ 840 smp_mb(); 841 if (mptcp_stream_memory_free(sk, 1)) 842 sk_stream_write_space(sk); 843} 844 845static inline void __mptcp_sync_sndbuf(struct sock *sk) 846{ 847 struct mptcp_subflow_context *subflow; 848 int ssk_sndbuf, new_sndbuf; 849 850 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 851 return; 852 853 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]); 854 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 855 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 856 857 subflow->cached_sndbuf = ssk_sndbuf; 858 new_sndbuf += ssk_sndbuf; 859 } 860 861 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 862 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 863 mptcp_write_space(sk); 864} 865 866/* The called held both the msk socket and the subflow socket locks, 867 * possibly under BH 868 */ 869static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 870{ 871 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 872 873 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 874 __mptcp_sync_sndbuf(sk); 875} 876 877/* the caller held only the subflow socket lock, either in process or 878 * BH context. Additionally this can be called under the msk data lock, 879 * so we can't acquire such lock here: let the delegate action acquires 880 * the needed locks in suitable order. 881 */ 882static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 883{ 884 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 885 886 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 887 return; 888 889 local_bh_disable(); 890 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 891 local_bh_enable(); 892} 893 894void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags); 895 896#define MPTCP_TOKEN_MAX_RETRIES 4 897 898void __init mptcp_token_init(void); 899static inline void mptcp_token_init_request(struct request_sock *req) 900{ 901 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 902} 903 904int mptcp_token_new_request(struct request_sock *req); 905void mptcp_token_destroy_request(struct request_sock *req); 906int mptcp_token_new_connect(struct sock *ssk); 907void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 908 struct mptcp_sock *msk); 909bool mptcp_token_exists(u32 token); 910struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 911struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 912 long *s_num); 913void mptcp_token_destroy(struct mptcp_sock *msk); 914 915void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 916 917void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 918__sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 919 920void __init mptcp_pm_init(void); 921void mptcp_pm_data_init(struct mptcp_sock *msk); 922void mptcp_pm_data_reset(struct mptcp_sock *msk); 923int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 924 struct mptcp_addr_info *addr); 925int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 926 bool require_family, 927 struct mptcp_pm_addr_entry *entry); 928bool mptcp_pm_addr_families_match(const struct sock *sk, 929 const struct mptcp_addr_info *loc, 930 const struct mptcp_addr_info *rem); 931void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 932void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 933void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 934void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 935bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 936void mptcp_pm_connection_closed(struct mptcp_sock *msk); 937void mptcp_pm_subflow_established(struct mptcp_sock *msk); 938bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 939void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 940 const struct mptcp_subflow_context *subflow); 941void mptcp_pm_add_addr_received(const struct sock *ssk, 942 const struct mptcp_addr_info *addr); 943void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 944 const struct mptcp_addr_info *addr); 945void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 946void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 947void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 948 const struct mptcp_rm_list *rm_list); 949void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 950void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 951int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 952 struct mptcp_addr_info *addr, 953 struct mptcp_addr_info *rem, 954 u8 bkup); 955bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 956 const struct mptcp_addr_info *addr); 957void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 958bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 959struct mptcp_pm_add_entry * 960mptcp_pm_del_add_timer(struct mptcp_sock *msk, 961 const struct mptcp_addr_info *addr, bool check_id); 962struct mptcp_pm_add_entry * 963mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk, 964 const struct mptcp_addr_info *addr); 965int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 966 unsigned int id, 967 u8 *flags, int *ifindex); 968int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id, 969 u8 *flags, int *ifindex); 970int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, 971 unsigned int id, 972 u8 *flags, int *ifindex); 973int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info); 974int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info); 975int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info); 976int mptcp_pm_announce_addr(struct mptcp_sock *msk, 977 const struct mptcp_addr_info *addr, 978 bool echo); 979int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 980int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 981void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list); 982 983void mptcp_free_local_addr_list(struct mptcp_sock *msk); 984 985void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 986 const struct sock *ssk, gfp_t gfp); 987void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 988void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 989void mptcp_event_pm_listener(const struct sock *ssk, 990 enum mptcp_event_type event); 991bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 992 993void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow, 994 const struct mptcp_options_received *mp_opt); 995void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 996 struct request_sock *req); 997int mptcp_nl_fill_addr(struct sk_buff *skb, 998 struct mptcp_pm_addr_entry *entry); 999 1000static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 1001{ 1002 return READ_ONCE(msk->pm.addr_signal) & 1003 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 1004} 1005 1006static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 1007{ 1008 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 1009} 1010 1011static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 1012{ 1013 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 1014} 1015 1016static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 1017{ 1018 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 1019} 1020 1021static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1022{ 1023 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1024} 1025 1026static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1027{ 1028 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1029} 1030 1031static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 1032{ 1033 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 1034 1035 if (family == AF_INET6) 1036 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 1037 if (!echo) 1038 len += MPTCPOPT_THMAC_LEN; 1039 /* account for 2 trailing 'nop' options */ 1040 if (port) 1041 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 1042 1043 return len; 1044} 1045 1046static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 1047{ 1048 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 1049 return -EINVAL; 1050 1051 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 1052} 1053 1054bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 1055 unsigned int opt_size, unsigned int remaining, 1056 struct mptcp_addr_info *addr, bool *echo, 1057 bool *drop_other_suboptions); 1058bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1059 struct mptcp_rm_list *rm_list); 1060int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1061int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1062int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1063int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb); 1064int mptcp_pm_nl_dump_addr(struct sk_buff *msg, 1065 struct netlink_callback *cb); 1066int mptcp_userspace_pm_dump_addr(struct sk_buff *msg, 1067 struct netlink_callback *cb); 1068int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info); 1069int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info); 1070int mptcp_userspace_pm_get_addr(struct sk_buff *skb, 1071 struct genl_info *info); 1072 1073static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1074{ 1075 int local_id = READ_ONCE(subflow->local_id); 1076 1077 if (local_id < 0) 1078 return 0; 1079 return local_id; 1080} 1081 1082void __init mptcp_pm_nl_init(void); 1083void mptcp_pm_nl_work(struct mptcp_sock *msk); 1084void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 1085 const struct mptcp_rm_list *rm_list); 1086unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk); 1087unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk); 1088unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk); 1089unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk); 1090 1091/* called under PM lock */ 1092static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1093{ 1094 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk)) 1095 WRITE_ONCE(msk->pm.accept_subflow, true); 1096} 1097 1098static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1099{ 1100 spin_lock_bh(&msk->pm.lock); 1101 __mptcp_pm_close_subflow(msk); 1102 spin_unlock_bh(&msk->pm.lock); 1103} 1104 1105void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 1106void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1107 1108static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1109{ 1110 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1111} 1112 1113void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1114 1115static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1116{ 1117 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1118} 1119 1120static inline bool mptcp_check_fallback(const struct sock *sk) 1121{ 1122 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1123 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1124 1125 return __mptcp_check_fallback(msk); 1126} 1127 1128static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 1129{ 1130 if (__mptcp_check_fallback(msk)) { 1131 pr_debug("TCP fallback already done (msk=%p)", msk); 1132 return; 1133 } 1134 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1135} 1136 1137static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1138{ 1139 struct sock *ssk = READ_ONCE(msk->first); 1140 1141 return ssk && ((1 << inet_sk_state_load(ssk)) & 1142 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1143 TCPF_SYN_RECV | TCPF_LISTEN)); 1144} 1145 1146static inline void mptcp_do_fallback(struct sock *ssk) 1147{ 1148 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1149 struct sock *sk = subflow->conn; 1150 struct mptcp_sock *msk; 1151 1152 msk = mptcp_sk(sk); 1153 __mptcp_do_fallback(msk); 1154 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1155 gfp_t saved_allocation = ssk->sk_allocation; 1156 1157 /* we are in a atomic (BH) scope, override ssk default for data 1158 * fin allocation 1159 */ 1160 ssk->sk_allocation = GFP_ATOMIC; 1161 ssk->sk_shutdown |= SEND_SHUTDOWN; 1162 tcp_shutdown(ssk, SEND_SHUTDOWN); 1163 ssk->sk_allocation = saved_allocation; 1164 } 1165} 1166 1167#define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 1168 1169static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1170{ 1171 struct mptcp_ext *mpext; 1172 1173 mpext = skb ? mptcp_get_ext(skb) : NULL; 1174 if (mpext && mpext->infinite_map) 1175 return true; 1176 1177 return false; 1178} 1179 1180static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1181{ 1182 return (subflow->request_mptcp || subflow->request_join); 1183} 1184 1185static inline bool subflow_simultaneous_connect(struct sock *sk) 1186{ 1187 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1188 1189 return (1 << sk->sk_state) & 1190 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) && 1191 is_active_ssk(subflow) && 1192 !subflow->conn_finished; 1193} 1194 1195#ifdef CONFIG_SYN_COOKIES 1196void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1197 struct sk_buff *skb); 1198bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1199 struct sk_buff *skb); 1200void __init mptcp_join_cookie_init(void); 1201#else 1202static inline void 1203subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1204 struct sk_buff *skb) {} 1205static inline bool 1206mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1207 struct sk_buff *skb) 1208{ 1209 return false; 1210} 1211 1212static inline void mptcp_join_cookie_init(void) {} 1213#endif 1214 1215#endif /* __MPTCP_PROTOCOL_H */ 1216