1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* Multipath TCP 3 * 4 * Copyright (c) 2017 - 2019, Intel Corporation. 5 */ 6 7 #ifndef __MPTCP_PROTOCOL_H 8 #define __MPTCP_PROTOCOL_H 9 10 #include <linux/random.h> 11 #include <net/tcp.h> 12 #include <net/inet_connection_sock.h> 13 #include <uapi/linux/mptcp.h> 14 #include <net/genetlink.h> 15 #include <net/rstreason.h> 16 17 #define MPTCP_SUPPORTED_VERSION 1 18 19 /* MPTCP option bits */ 20 #define OPTION_MPTCP_MPC_SYN BIT(0) 21 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 22 #define OPTION_MPTCP_MPC_ACK BIT(2) 23 #define OPTION_MPTCP_MPJ_SYN BIT(3) 24 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 25 #define OPTION_MPTCP_MPJ_ACK BIT(5) 26 #define OPTION_MPTCP_ADD_ADDR BIT(6) 27 #define OPTION_MPTCP_RM_ADDR BIT(7) 28 #define OPTION_MPTCP_FASTCLOSE BIT(8) 29 #define OPTION_MPTCP_PRIO BIT(9) 30 #define OPTION_MPTCP_RST BIT(10) 31 #define OPTION_MPTCP_DSS BIT(11) 32 #define OPTION_MPTCP_FAIL BIT(12) 33 34 #define OPTION_MPTCP_CSUMREQD BIT(13) 35 36 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \ 37 OPTION_MPTCP_MPC_ACK) 38 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \ 39 OPTION_MPTCP_MPJ_ACK) 40 41 /* MPTCP option subtypes */ 42 #define MPTCPOPT_MP_CAPABLE 0 43 #define MPTCPOPT_MP_JOIN 1 44 #define MPTCPOPT_DSS 2 45 #define MPTCPOPT_ADD_ADDR 3 46 #define MPTCPOPT_RM_ADDR 4 47 #define MPTCPOPT_MP_PRIO 5 48 #define MPTCPOPT_MP_FAIL 6 49 #define MPTCPOPT_MP_FASTCLOSE 7 50 #define MPTCPOPT_RST 8 51 52 /* MPTCP suboption lengths */ 53 #define TCPOLEN_MPTCP_MPC_SYN 4 54 #define TCPOLEN_MPTCP_MPC_SYNACK 12 55 #define TCPOLEN_MPTCP_MPC_ACK 20 56 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 57 #define TCPOLEN_MPTCP_MPJ_SYN 12 58 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 59 #define TCPOLEN_MPTCP_MPJ_ACK 24 60 #define TCPOLEN_MPTCP_DSS_BASE 4 61 #define TCPOLEN_MPTCP_DSS_ACK32 4 62 #define TCPOLEN_MPTCP_DSS_ACK64 8 63 #define TCPOLEN_MPTCP_DSS_MAP32 10 64 #define TCPOLEN_MPTCP_DSS_MAP64 14 65 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 66 #define TCPOLEN_MPTCP_ADD_ADDR 16 67 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 70 #define TCPOLEN_MPTCP_ADD_ADDR6 28 71 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 74 #define TCPOLEN_MPTCP_PORT_LEN 2 75 #define TCPOLEN_MPTCP_PORT_ALIGN 2 76 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 77 #define TCPOLEN_MPTCP_PRIO 3 78 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 79 #define TCPOLEN_MPTCP_FASTCLOSE 12 80 #define TCPOLEN_MPTCP_RST 4 81 #define TCPOLEN_MPTCP_FAIL 12 82 83 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 84 85 /* MPTCP MP_JOIN flags */ 86 #define MPTCPOPT_BACKUP BIT(0) 87 #define MPTCPOPT_THMAC_LEN 8 88 89 /* MPTCP MP_CAPABLE flags */ 90 #define MPTCP_VERSION_MASK (0x0F) 91 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 92 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 93 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 94 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 95 #define MPTCP_CAP_FLAG_MASK (0x1F) 96 97 /* MPTCP DSS flags */ 98 #define MPTCP_DSS_DATA_FIN BIT(4) 99 #define MPTCP_DSS_DSN64 BIT(3) 100 #define MPTCP_DSS_HAS_MAP BIT(2) 101 #define MPTCP_DSS_ACK64 BIT(1) 102 #define MPTCP_DSS_HAS_ACK BIT(0) 103 #define MPTCP_DSS_FLAG_MASK (0x1F) 104 105 /* MPTCP ADD_ADDR flags */ 106 #define MPTCP_ADDR_ECHO BIT(0) 107 108 /* MPTCP MP_PRIO flags */ 109 #define MPTCP_PRIO_BKUP BIT(0) 110 111 /* MPTCP TCPRST flags */ 112 #define MPTCP_RST_TRANSIENT BIT(0) 113 114 /* MPTCP socket atomic flags */ 115 #define MPTCP_WORK_RTX 1 116 #define MPTCP_FALLBACK_DONE 2 117 #define MPTCP_WORK_CLOSE_SUBFLOW 3 118 119 /* MPTCP socket release cb flags */ 120 #define MPTCP_PUSH_PENDING 1 121 #define MPTCP_CLEAN_UNA 2 122 #define MPTCP_ERROR_REPORT 3 123 #define MPTCP_RETRANSMIT 4 124 #define MPTCP_FLUSH_JOIN_LIST 5 125 #define MPTCP_SYNC_STATE 6 126 #define MPTCP_SYNC_SNDBUF 7 127 128 struct mptcp_skb_cb { 129 u64 map_seq; 130 u64 end_seq; 131 u32 offset; 132 u8 has_rxtstamp; 133 u8 cant_coalesce; 134 }; 135 136 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0])) 137 138 static inline bool before64(__u64 seq1, __u64 seq2) 139 { 140 return (__s64)(seq1 - seq2) < 0; 141 } 142 143 #define after64(seq2, seq1) before64(seq1, seq2) 144 145 struct mptcp_options_received { 146 u64 sndr_key; 147 u64 rcvr_key; 148 u64 data_ack; 149 u64 data_seq; 150 u32 subflow_seq; 151 u16 data_len; 152 __sum16 csum; 153 struct_group(status, 154 u16 suboptions; 155 u16 use_map:1, 156 dsn64:1, 157 data_fin:1, 158 use_ack:1, 159 ack64:1, 160 mpc_map:1, 161 reset_reason:4, 162 reset_transient:1, 163 echo:1, 164 backup:1, 165 deny_join_id0:1, 166 __unused:2; 167 ); 168 u8 join_id; 169 u32 token; 170 u32 nonce; 171 u64 thmac; 172 u8 hmac[MPTCPOPT_HMAC_LEN]; 173 struct mptcp_addr_info addr; 174 struct mptcp_rm_list rm_list; 175 u64 ahmac; 176 u64 fail_seq; 177 }; 178 179 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 180 { 181 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 182 ((nib & 0xF) << 8) | field); 183 } 184 185 enum mptcp_pm_status { 186 MPTCP_PM_ADD_ADDR_RECEIVED, 187 MPTCP_PM_ADD_ADDR_SEND_ACK, 188 MPTCP_PM_RM_ADDR_RECEIVED, 189 MPTCP_PM_ESTABLISHED, 190 MPTCP_PM_SUBFLOW_ESTABLISHED, 191 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 192 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is 193 * accounted int id_avail_bitmap 194 */ 195 }; 196 197 enum mptcp_pm_type { 198 MPTCP_PM_TYPE_KERNEL = 0, 199 MPTCP_PM_TYPE_USERSPACE, 200 201 __MPTCP_PM_TYPE_NR, 202 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, 203 }; 204 205 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */ 206 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1) 207 208 enum mptcp_addr_signal_status { 209 MPTCP_ADD_ADDR_SIGNAL, 210 MPTCP_ADD_ADDR_ECHO, 211 MPTCP_RM_ADDR_SIGNAL, 212 }; 213 214 /* max value of mptcp_addr_info.id */ 215 #define MPTCP_PM_MAX_ADDR_ID U8_MAX 216 217 struct mptcp_pm_data { 218 struct mptcp_addr_info local; 219 struct mptcp_addr_info remote; 220 struct list_head anno_list; 221 struct list_head userspace_pm_local_addr_list; 222 223 spinlock_t lock; /*protects the whole PM data */ 224 225 struct_group(reset, 226 227 u8 addr_signal; 228 bool server_side; 229 bool work_pending; 230 bool accept_addr; 231 bool accept_subflow; 232 bool remote_deny_join_id0; 233 u8 add_addr_signaled; 234 u8 add_addr_accepted; 235 u8 local_addr_used; 236 u8 pm_type; 237 u8 extra_subflows; 238 u8 status; 239 240 ); 241 242 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1); 243 struct mptcp_rm_list rm_list_tx; 244 struct mptcp_rm_list rm_list_rx; 245 }; 246 247 struct mptcp_pm_local { 248 struct mptcp_addr_info addr; 249 u32 flags; 250 int ifindex; 251 }; 252 253 struct mptcp_pm_addr_entry { 254 struct list_head list; 255 struct mptcp_addr_info addr; 256 u32 flags; 257 int ifindex; 258 struct socket *lsk; 259 }; 260 261 struct mptcp_data_frag { 262 struct list_head list; 263 u64 data_seq; 264 u16 data_len; 265 u16 offset; 266 u8 overhead; 267 u8 eor; /* currently using 1 bit */ 268 u16 already_sent; 269 struct page *page; 270 }; 271 272 /* Arbitrary compromise between as low as possible to react timely to subflow 273 * close event and as big as possible to avoid being fouled by biased large 274 * samples due to peer sending data on a different subflow WRT to the incoming 275 * ack. 276 */ 277 #define MPTCP_RTT_SAMPLES 5 278 279 /* MPTCP connection sock */ 280 struct mptcp_sock { 281 /* inet_connection_sock must be the first member */ 282 struct inet_connection_sock sk; 283 u64 local_key; /* protected by the first subflow socket lock 284 * lockless access read 285 */ 286 u64 remote_key; /* same as above */ 287 u64 write_seq; 288 u64 bytes_sent; 289 u64 snd_nxt; 290 u64 bytes_received; 291 u64 ack_seq; 292 atomic64_t rcv_wnd_sent; 293 u64 rcv_data_fin_seq; 294 u64 bytes_retrans; 295 u64 bytes_consumed; 296 int snd_burst; 297 int old_wspace; 298 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 299 * recovery related fields are under data_lock 300 * protection 301 */ 302 u64 bytes_acked; 303 u64 snd_una; 304 u64 wnd_end; 305 u32 last_data_sent; 306 u32 last_data_recv; 307 u32 last_ack_recv; 308 unsigned long timer_ival; 309 u32 token; 310 unsigned long flags; 311 unsigned long cb_flags; 312 bool recovery; /* closing subflow write queue reinjected */ 313 bool can_ack; 314 bool fully_established; 315 bool rcv_data_fin; 316 bool snd_data_fin_enable; 317 bool rcv_fastclose; 318 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 319 bool csum_enabled; 320 bool allow_infinite_fallback; 321 u8 pending_state; /* A subflow asked to set this sk_state, 322 * protected by the msk data lock 323 */ 324 u8 mpc_endpoint_id; 325 u8 recvmsg_inq:1, 326 cork:1, 327 nodelay:1, 328 fastopening:1, 329 in_accept_queue:1, 330 free_first:1, 331 rcvspace_init:1, 332 fastclosing:1; 333 u32 notsent_lowat; 334 int keepalive_cnt; 335 int keepalive_idle; 336 int keepalive_intvl; 337 int maxseg; 338 struct work_struct work; 339 struct sk_buff *ooo_last_skb; 340 struct rb_root out_of_order_queue; 341 struct list_head conn_list; 342 struct list_head rtx_queue; 343 struct mptcp_data_frag *first_pending; 344 struct list_head join_list; 345 struct sock *first; /* The mptcp ops can safely dereference, using suitable 346 * ONCE annotation, the subflow outside the socket 347 * lock as such sock is freed after close(). 348 */ 349 struct mptcp_pm_data pm; 350 struct mptcp_sched_ops *sched; 351 352 /* Most recent rtt_us observed by in use incoming subflows. */ 353 struct { 354 u32 samples[MPTCP_RTT_SAMPLES]; 355 u32 next_sample; 356 } rcv_rtt_est; 357 358 struct { 359 int space; /* bytes copied in last measurement window */ 360 int copied; /* bytes copied in this measurement window */ 361 u64 time; /* start time of measurement window */ 362 } rcvq_space; 363 u8 scaling_ratio; 364 bool allow_subflows; 365 366 u32 subflow_id; 367 u32 setsockopt_seq; 368 char ca_name[TCP_CA_NAME_MAX]; 369 370 spinlock_t fallback_lock; /* protects fallback, 371 * allow_infinite_fallback and 372 * allow_join 373 */ 374 375 struct list_head backlog_list; /* protected by the data lock */ 376 u32 backlog_len; 377 u32 backlog_unaccounted; 378 }; 379 380 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 381 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 382 383 #define mptcp_for_each_subflow(__msk, __subflow) \ 384 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 385 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \ 386 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node) 387 #define mptcp_next_subflow(__msk, __subflow) \ 388 list_next_entry_circular(__subflow, &((__msk)->conn_list), node) 389 390 extern struct genl_family mptcp_genl_family; 391 392 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 393 { 394 sock_owned_by_me((const struct sock *)msk); 395 } 396 397 #ifdef CONFIG_DEBUG_NET 398 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */ 399 #undef tcp_sk 400 #define tcp_sk(ptr) ({ \ 401 typeof(ptr) _ptr = (ptr); \ 402 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \ 403 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \ 404 }) 405 #define mptcp_sk(ptr) ({ \ 406 typeof(ptr) _ptr = (ptr); \ 407 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \ 408 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \ 409 }) 410 411 #else /* !CONFIG_DEBUG_NET */ 412 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk) 413 #endif 414 415 static inline int mptcp_win_from_space(const struct sock *sk, int space) 416 { 417 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space); 418 } 419 420 static inline int mptcp_space_from_win(const struct sock *sk, int win) 421 { 422 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win); 423 } 424 425 static inline int __mptcp_space(const struct sock *sk) 426 { 427 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - 428 READ_ONCE(mptcp_sk(sk)->backlog_len) - 429 sk_rmem_alloc_get(sk)); 430 } 431 432 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 433 { 434 const struct mptcp_sock *msk = mptcp_sk(sk); 435 436 return msk->first_pending; 437 } 438 439 static inline void mptcp_init_rtt_est(struct mptcp_sock *msk) 440 { 441 int i; 442 443 for (i = 0; i < MPTCP_RTT_SAMPLES; ++i) 444 msk->rcv_rtt_est.samples[i] = U32_MAX; 445 msk->rcv_rtt_est.next_sample = 0; 446 msk->scaling_ratio = TCP_DEFAULT_SCALING_RATIO; 447 } 448 449 static inline u32 mptcp_rtt_us_est(const struct mptcp_sock *msk) 450 { 451 u32 rtt_us = READ_ONCE(msk->rcv_rtt_est.samples[0]); 452 int i; 453 454 /* Lockless access of collected samples. */ 455 for (i = 1; i < MPTCP_RTT_SAMPLES; ++i) 456 rtt_us = min(rtt_us, READ_ONCE(msk->rcv_rtt_est.samples[i])); 457 return rtt_us; 458 } 459 460 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 461 { 462 struct mptcp_sock *msk = mptcp_sk(sk); 463 struct mptcp_data_frag *cur; 464 465 cur = msk->first_pending; 466 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 467 list_next_entry(cur, list); 468 } 469 470 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 471 { 472 const struct mptcp_sock *msk = mptcp_sk(sk); 473 474 if (!msk->first_pending) 475 return NULL; 476 477 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 478 return NULL; 479 480 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 481 } 482 483 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk) 484 { 485 struct mptcp_sock *msk = mptcp_sk(sk); 486 487 if (msk->snd_una == msk->snd_nxt) 488 return NULL; 489 490 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 491 } 492 493 struct csum_pseudo_header { 494 __be64 data_seq; 495 __be32 subflow_seq; 496 __be16 data_len; 497 __sum16 csum; 498 }; 499 500 struct mptcp_subflow_request_sock { 501 struct tcp_request_sock sk; 502 u16 mp_capable : 1, 503 mp_join : 1, 504 backup : 1, 505 request_bkup : 1, 506 csum_reqd : 1, 507 allow_join_id0 : 1; 508 u8 local_id; 509 u8 remote_id; 510 u64 local_key; 511 u64 idsn; 512 u32 token; 513 u32 ssn_offset; 514 u64 thmac; 515 u32 local_nonce; 516 u32 remote_nonce; 517 struct mptcp_sock *msk; 518 struct hlist_nulls_node token_node; 519 }; 520 521 static inline struct mptcp_subflow_request_sock * 522 mptcp_subflow_rsk(const struct request_sock *rsk) 523 { 524 return (struct mptcp_subflow_request_sock *)rsk; 525 } 526 527 struct mptcp_delegated_action { 528 struct napi_struct napi; 529 local_lock_t bh_lock; 530 struct list_head head; 531 }; 532 533 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 534 535 #define MPTCP_DELEGATE_SCHEDULED 0 536 #define MPTCP_DELEGATE_SEND 1 537 #define MPTCP_DELEGATE_ACK 2 538 #define MPTCP_DELEGATE_SNDBUF 3 539 540 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED)) 541 /* MPTCP subflow context */ 542 struct mptcp_subflow_context { 543 struct list_head node;/* conn_list of subflows */ 544 545 struct_group(reset, 546 547 unsigned long avg_pacing_rate; /* protected by msk socket lock */ 548 u64 local_key; 549 u64 remote_key; 550 u64 idsn; 551 u64 map_seq; 552 u64 rcv_wnd_sent; 553 u32 snd_isn; 554 u32 token; 555 u32 rel_write_seq; 556 u32 map_subflow_seq; 557 u32 ssn_offset; 558 u32 map_data_len; 559 __wsum map_data_csum; 560 u32 map_csum_len; 561 u32 prev_rtt_seq; 562 u32 request_mptcp : 1, /* send MP_CAPABLE */ 563 request_join : 1, /* send MP_JOIN */ 564 request_bkup : 1, 565 mp_capable : 1, /* remote is MPTCP capable */ 566 mp_join : 1, /* remote is JOINing */ 567 pm_notified : 1, /* PM hook called for established status */ 568 conn_finished : 1, 569 map_valid : 1, 570 map_csum_reqd : 1, 571 map_data_fin : 1, 572 mpc_map : 1, 573 backup : 1, 574 send_mp_prio : 1, 575 send_mp_fail : 1, 576 send_fastclose : 1, 577 send_infinite_map : 1, 578 remote_key_valid : 1, /* received the peer key from */ 579 disposable : 1, /* ctx can be free at ulp release time */ 580 closing : 1, /* must not pass rx data to msk anymore */ 581 stale : 1, /* unable to snd/rcv data, do not use for xmit */ 582 valid_csum_seen : 1, /* at least one csum validated */ 583 is_mptfo : 1, /* subflow is doing TFO */ 584 close_event_done : 1, /* has done the post-closed part */ 585 mpc_drop : 1, /* the MPC option has been dropped in a rtx */ 586 __unused : 8; 587 bool data_avail; 588 bool scheduled; 589 bool pm_listener; /* a listener managed by the kernel PM? */ 590 bool fully_established; /* path validated */ 591 u32 lent_mem_frag; 592 u32 remote_nonce; 593 u64 thmac; 594 u32 local_nonce; 595 u32 remote_token; 596 union { 597 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */ 598 u64 iasn; /* initial ack sequence number, MPC subflows only */ 599 }; 600 s16 local_id; /* if negative not initialized yet */ 601 u8 remote_id; 602 u8 reset_seen:1; 603 u8 reset_transient:1; 604 u8 reset_reason:4; 605 u8 stale_count; 606 607 u32 subflow_id; 608 609 long delegated_status; 610 unsigned long fail_tout; 611 612 ); 613 614 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 615 616 u32 setsockopt_seq; 617 u32 stale_rcv_tstamp; 618 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf, 619 * protected by the msk socket lock 620 */ 621 622 struct sock *tcp_sock; /* tcp sk backpointer */ 623 struct sock *conn; /* parent mptcp_sock */ 624 const struct inet_connection_sock_af_ops *icsk_af_ops; 625 void (*tcp_state_change)(struct sock *sk); 626 void (*tcp_error_report)(struct sock *sk); 627 628 struct rcu_head rcu; 629 }; 630 631 static inline struct mptcp_subflow_context * 632 mptcp_subflow_ctx(const struct sock *sk) 633 { 634 const struct inet_connection_sock *icsk = inet_csk(sk); 635 636 /* Use RCU on icsk_ulp_data only for sock diag code */ 637 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 638 } 639 640 static inline struct sock * 641 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 642 { 643 return subflow->tcp_sock; 644 } 645 646 static inline void 647 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow) 648 { 649 memset(&subflow->reset, 0, sizeof(subflow->reset)); 650 subflow->request_mptcp = 1; 651 WRITE_ONCE(subflow->local_id, -1); 652 } 653 654 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */ 655 static inline enum sk_rst_reason 656 sk_rst_convert_mptcp_reason(u32 reason) 657 { 658 switch (reason) { 659 case MPTCP_RST_EUNSPEC: 660 return SK_RST_REASON_MPTCP_RST_EUNSPEC; 661 case MPTCP_RST_EMPTCP: 662 return SK_RST_REASON_MPTCP_RST_EMPTCP; 663 case MPTCP_RST_ERESOURCE: 664 return SK_RST_REASON_MPTCP_RST_ERESOURCE; 665 case MPTCP_RST_EPROHIBIT: 666 return SK_RST_REASON_MPTCP_RST_EPROHIBIT; 667 case MPTCP_RST_EWQ2BIG: 668 return SK_RST_REASON_MPTCP_RST_EWQ2BIG; 669 case MPTCP_RST_EBADPERF: 670 return SK_RST_REASON_MPTCP_RST_EBADPERF; 671 case MPTCP_RST_EMIDDLEBOX: 672 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX; 673 default: 674 /* It should not happen, or else errors may occur 675 * in MPTCP layer 676 */ 677 return SK_RST_REASON_ERROR; 678 } 679 } 680 681 static inline void 682 mptcp_send_active_reset_reason(struct sock *sk) 683 { 684 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 685 enum sk_rst_reason reason; 686 687 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason); 688 tcp_send_active_reset(sk, GFP_ATOMIC, reason); 689 } 690 691 /* Made the fwd mem carried by the given skb available to the msk, 692 * To be paired with a previous mptcp_subflow_lend_fwdmem() before freeing 693 * the skb or setting the skb ownership. 694 */ 695 static inline void mptcp_borrow_fwdmem(struct sock *sk, struct sk_buff *skb) 696 { 697 struct sock *ssk = skb->sk; 698 699 /* The subflow just lend the skb fwd memory; if the subflow meanwhile 700 * closed, mptcp_close_ssk() already released the ssk rcv memory. 701 */ 702 DEBUG_NET_WARN_ON_ONCE(skb->destructor); 703 sk_forward_alloc_add(sk, skb->truesize); 704 if (!ssk) 705 return; 706 707 atomic_sub(skb->truesize, &ssk->sk_rmem_alloc); 708 skb->sk = NULL; 709 } 710 711 static inline void 712 __mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow, int size) 713 { 714 int frag = (subflow->lent_mem_frag + size) & (PAGE_SIZE - 1); 715 716 subflow->lent_mem_frag = frag; 717 } 718 719 static inline void 720 mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow, 721 struct sk_buff *skb) 722 { 723 __mptcp_subflow_lend_fwdmem(subflow, skb->truesize); 724 skb->destructor = NULL; 725 } 726 727 static inline u64 728 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 729 { 730 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 731 subflow->ssn_offset - 732 subflow->map_subflow_seq; 733 } 734 735 static inline u64 736 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 737 { 738 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 739 } 740 741 void mptcp_subflow_process_delegated(struct sock *ssk, long actions); 742 743 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action) 744 { 745 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action); 746 struct mptcp_delegated_action *delegated; 747 bool schedule; 748 749 /* the caller held the subflow bh socket lock */ 750 lockdep_assert_in_softirq(); 751 752 /* The implied barrier pairs with tcp_release_cb_override() 753 * mptcp_napi_poll(), and ensures the below list check sees list 754 * updates done prior to delegated status bits changes 755 */ 756 old = set_mask_bits(&subflow->delegated_status, 0, set_bits); 757 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) { 758 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node))) 759 return; 760 761 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 762 delegated = this_cpu_ptr(&mptcp_delegated_actions); 763 schedule = list_empty(&delegated->head); 764 list_add_tail(&subflow->delegated_node, &delegated->head); 765 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 766 sock_hold(mptcp_subflow_tcp_sock(subflow)); 767 if (schedule) 768 napi_schedule(&delegated->napi); 769 } 770 } 771 772 static inline struct mptcp_subflow_context * 773 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 774 { 775 struct mptcp_subflow_context *ret; 776 777 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock); 778 if (list_empty(&delegated->head)) { 779 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 780 return NULL; 781 } 782 783 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 784 list_del_init(&ret->delegated_node); 785 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock); 786 return ret; 787 } 788 789 void __mptcp_inherit_memcg(struct sock *sk, struct sock *ssk, gfp_t gfp); 790 void __mptcp_inherit_cgrp_data(struct sock *sk, struct sock *ssk); 791 792 int mptcp_is_enabled(const struct net *net); 793 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 794 int mptcp_is_checksum_enabled(const struct net *net); 795 int mptcp_allow_join_id0(const struct net *net); 796 unsigned int mptcp_stale_loss_cnt(const struct net *net); 797 unsigned int mptcp_close_timeout(const struct sock *sk); 798 int mptcp_get_pm_type(const struct net *net); 799 const char *mptcp_get_path_manager(const struct net *net); 800 const char *mptcp_get_scheduler(const struct net *net); 801 802 void mptcp_active_disable(struct sock *sk); 803 bool mptcp_active_should_disable(struct sock *ssk); 804 void mptcp_active_enable(struct sock *sk); 805 806 void mptcp_get_available_schedulers(char *buf, size_t maxlen); 807 void __mptcp_subflow_fully_established(struct mptcp_sock *msk, 808 struct mptcp_subflow_context *subflow, 809 const struct mptcp_options_received *mp_opt); 810 bool __mptcp_retransmit_pending_data(struct sock *sk); 811 void mptcp_check_and_set_pending(struct sock *sk); 812 void __mptcp_push_pending(struct sock *sk, unsigned int flags); 813 bool mptcp_subflow_data_available(struct sock *sk); 814 void __init mptcp_subflow_init(void); 815 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 816 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 817 struct mptcp_subflow_context *subflow); 818 void __mptcp_subflow_send_ack(struct sock *ssk); 819 void mptcp_subflow_reset(struct sock *ssk); 820 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); 821 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 822 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk); 823 bool __mptcp_close(struct sock *sk, long timeout); 824 void mptcp_cancel_work(struct sock *sk); 825 void __mptcp_unaccepted_force_close(struct sock *sk); 826 void mptcp_set_state(struct sock *sk, int state); 827 828 bool mptcp_addresses_equal(const struct mptcp_addr_info *a, 829 const struct mptcp_addr_info *b, bool use_port); 830 void mptcp_local_address(const struct sock_common *skc, 831 struct mptcp_addr_info *addr); 832 void mptcp_remote_address(const struct sock_common *skc, 833 struct mptcp_addr_info *addr); 834 835 /* called with sk socket lock held */ 836 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local, 837 const struct mptcp_addr_info *remote); 838 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family, 839 struct socket **new_sock); 840 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 841 struct sockaddr_storage *addr, 842 unsigned short family); 843 struct mptcp_sched_ops *mptcp_sched_find(const char *name); 844 int mptcp_validate_scheduler(struct mptcp_sched_ops *sched); 845 int mptcp_register_scheduler(struct mptcp_sched_ops *sched); 846 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched); 847 void mptcp_sched_init(void); 848 int mptcp_init_sched(struct mptcp_sock *msk, 849 struct mptcp_sched_ops *sched); 850 void mptcp_release_sched(struct mptcp_sock *msk); 851 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, 852 bool scheduled); 853 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk); 854 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk); 855 int mptcp_sched_get_send(struct mptcp_sock *msk); 856 int mptcp_sched_get_retrans(struct mptcp_sock *msk); 857 858 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk) 859 { 860 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed); 861 } 862 863 static inline bool mptcp_epollin_ready(const struct sock *sk) 864 { 865 u64 data_avail = mptcp_data_avail(mptcp_sk(sk)); 866 867 if (!data_avail) 868 return false; 869 870 /* mptcp doesn't have to deal with small skbs in the receive queue, 871 * as it can always coalesce them 872 */ 873 return (data_avail >= sk->sk_rcvlowat) || 874 tcp_under_memory_pressure(sk); 875 } 876 877 int mptcp_set_rcvlowat(struct sock *sk, int val); 878 879 static inline bool __tcp_can_send(const struct sock *ssk) 880 { 881 /* only send if our side has not closed yet */ 882 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 883 } 884 885 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow) 886 { 887 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 888 if (subflow->request_join && !READ_ONCE(subflow->fully_established)) 889 return false; 890 891 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow)); 892 } 893 894 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow); 895 896 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow); 897 898 void mptcp_subflow_drop_ctx(struct sock *ssk); 899 900 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 901 struct mptcp_subflow_context *ctx) 902 { 903 sk->sk_data_ready = sock_def_readable; 904 sk->sk_state_change = ctx->tcp_state_change; 905 sk->sk_write_space = sk_stream_write_space; 906 sk->sk_error_report = ctx->tcp_error_report; 907 908 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 909 } 910 911 void __init mptcp_proto_init(void); 912 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 913 int __init mptcp_proto_v6_init(void); 914 void __init mptcp_subflow_v6_init(void); 915 #endif 916 917 struct sock *mptcp_sk_clone_init(const struct sock *sk, 918 const struct mptcp_options_received *mp_opt, 919 struct sock *ssk, 920 struct request_sock *req); 921 void mptcp_get_options(const struct sk_buff *skb, 922 struct mptcp_options_received *mp_opt); 923 924 void mptcp_finish_connect(struct sock *sk); 925 void __mptcp_sync_state(struct sock *sk, int state); 926 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout); 927 928 static inline void mptcp_stop_tout_timer(struct sock *sk) 929 { 930 if (!inet_csk(sk)->icsk_mtup.probe_timestamp) 931 return; 932 933 sk_stop_timer(sk, &inet_csk(sk)->mptcp_tout_timer); 934 inet_csk(sk)->icsk_mtup.probe_timestamp = 0; 935 } 936 937 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout) 938 { 939 /* avoid 0 timestamp, as that means no close timeout */ 940 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1; 941 } 942 943 static inline void mptcp_start_tout_timer(struct sock *sk) 944 { 945 mptcp_set_close_tout(sk, tcp_jiffies32); 946 mptcp_reset_tout_timer(mptcp_sk(sk), 0); 947 } 948 949 static inline bool mptcp_is_fully_established(struct sock *sk) 950 { 951 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 952 READ_ONCE(mptcp_sk(sk)->fully_established); 953 } 954 955 static inline u64 mptcp_stamp(void) 956 { 957 return div_u64(tcp_clock_ns(), NSEC_PER_USEC); 958 } 959 960 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 961 bool mptcp_finish_join(struct sock *sk); 962 bool mptcp_schedule_work(struct sock *sk); 963 int mptcp_setsockopt(struct sock *sk, int level, int optname, 964 sockptr_t optval, unsigned int optlen); 965 int mptcp_getsockopt(struct sock *sk, int level, int optname, 966 char __user *optval, int __user *option); 967 968 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 969 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 970 { 971 if (use_64bit) 972 return cur_seq; 973 974 return __mptcp_expand_seq(old_seq, cur_seq); 975 } 976 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 977 void __mptcp_data_acked(struct sock *sk); 978 void __mptcp_error_report(struct sock *sk); 979 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 980 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 981 { 982 return READ_ONCE(msk->snd_data_fin_enable) && 983 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 984 } 985 986 static inline u32 mptcp_notsent_lowat(const struct sock *sk) 987 { 988 struct net *net = sock_net(sk); 989 u32 val; 990 991 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat); 992 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat); 993 } 994 995 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake) 996 { 997 const struct mptcp_sock *msk = mptcp_sk(sk); 998 u32 notsent_bytes; 999 1000 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt); 1001 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk); 1002 } 1003 1004 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake) 1005 { 1006 return mptcp_stream_memory_free(sk, wake) && 1007 __sk_stream_is_writeable(sk, wake); 1008 } 1009 1010 static inline void mptcp_write_space(struct sock *sk) 1011 { 1012 /* pairs with memory barrier in mptcp_poll */ 1013 smp_mb(); 1014 if (mptcp_stream_memory_free(sk, 1)) 1015 INDIRECT_CALL_1(sk->sk_write_space, sk_stream_write_space, sk); 1016 } 1017 1018 static inline void __mptcp_sync_sndbuf(struct sock *sk) 1019 { 1020 struct mptcp_subflow_context *subflow; 1021 int ssk_sndbuf, new_sndbuf; 1022 1023 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) 1024 return; 1025 1026 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]); 1027 mptcp_for_each_subflow(mptcp_sk(sk), subflow) { 1028 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf); 1029 1030 subflow->cached_sndbuf = ssk_sndbuf; 1031 new_sndbuf += ssk_sndbuf; 1032 } 1033 1034 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */ 1035 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf); 1036 mptcp_write_space(sk); 1037 } 1038 1039 /* The called held both the msk socket and the subflow socket locks, 1040 * possibly under BH 1041 */ 1042 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 1043 { 1044 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1045 1046 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf) 1047 __mptcp_sync_sndbuf(sk); 1048 } 1049 1050 /* the caller held only the subflow socket lock, either in process or 1051 * BH context. Additionally this can be called under the msk data lock, 1052 * so we can't acquire such lock here: let the delegate action acquires 1053 * the needed locks in suitable order. 1054 */ 1055 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 1056 { 1057 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1058 1059 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf)) 1060 return; 1061 1062 local_bh_disable(); 1063 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF); 1064 local_bh_enable(); 1065 } 1066 1067 #define MPTCP_TOKEN_MAX_RETRIES 4 1068 1069 void __init mptcp_token_init(void); 1070 static inline void mptcp_token_init_request(struct request_sock *req) 1071 { 1072 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 1073 } 1074 1075 int mptcp_token_new_request(struct request_sock *req); 1076 void mptcp_token_destroy_request(struct request_sock *req); 1077 int mptcp_token_new_connect(struct sock *ssk); 1078 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 1079 struct mptcp_sock *msk); 1080 bool mptcp_token_exists(u32 token); 1081 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token); 1082 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 1083 long *s_num); 1084 void mptcp_token_destroy(struct mptcp_sock *msk); 1085 1086 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 1087 1088 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 1089 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum); 1090 1091 void __init mptcp_pm_init(void); 1092 void mptcp_pm_data_init(struct mptcp_sock *msk); 1093 void mptcp_pm_data_reset(struct mptcp_sock *msk); 1094 void mptcp_pm_destroy(struct mptcp_sock *msk); 1095 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info, 1096 struct mptcp_addr_info *addr); 1097 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info, 1098 bool require_family, 1099 struct mptcp_pm_addr_entry *entry); 1100 bool mptcp_pm_addr_families_match(const struct sock *sk, 1101 const struct mptcp_addr_info *loc, 1102 const struct mptcp_addr_info *rem); 1103 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 1104 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 1105 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk); 1106 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 1107 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 1108 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 1109 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk); 1110 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, 1111 const struct mptcp_subflow_context *subflow); 1112 void mptcp_pm_add_addr_received(const struct sock *ssk, 1113 const struct mptcp_addr_info *addr); 1114 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 1115 const struct mptcp_addr_info *addr); 1116 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 1117 void mptcp_pm_send_ack(struct mptcp_sock *msk, 1118 struct mptcp_subflow_context *subflow, 1119 bool prio, bool backup); 1120 void mptcp_pm_addr_send_ack(struct mptcp_sock *msk); 1121 void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id); 1122 void mptcp_pm_rm_subflow(struct mptcp_sock *msk, 1123 const struct mptcp_rm_list *rm_list); 1124 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 1125 const struct mptcp_rm_list *rm_list); 1126 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 1127 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq); 1128 int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk, 1129 struct mptcp_addr_info *addr, 1130 struct mptcp_addr_info *rem, 1131 u8 bkup); 1132 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk, 1133 const struct mptcp_addr_info *addr); 1134 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 1135 struct mptcp_pm_add_entry * 1136 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 1137 const struct mptcp_addr_info *addr, bool check_id); 1138 bool mptcp_lookup_subflow_by_saddr(const struct list_head *list, 1139 const struct mptcp_addr_info *saddr); 1140 bool mptcp_remove_anno_list_by_saddr(struct mptcp_sock *msk, 1141 const struct mptcp_addr_info *addr); 1142 int mptcp_pm_nl_set_flags(struct mptcp_pm_addr_entry *local, 1143 struct genl_info *info); 1144 int mptcp_userspace_pm_set_flags(struct mptcp_pm_addr_entry *local, 1145 struct genl_info *info); 1146 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 1147 const struct mptcp_addr_info *addr, 1148 bool echo); 1149 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 1150 void mptcp_pm_remove_addr_entry(struct mptcp_sock *msk, 1151 struct mptcp_pm_addr_entry *entry); 1152 1153 /* the default path manager, used in mptcp_pm_unregister */ 1154 extern struct mptcp_pm_ops mptcp_pm_kernel; 1155 1156 struct mptcp_pm_ops *mptcp_pm_find(const char *name); 1157 int mptcp_pm_register(struct mptcp_pm_ops *pm_ops); 1158 void mptcp_pm_unregister(struct mptcp_pm_ops *pm_ops); 1159 int mptcp_pm_validate(struct mptcp_pm_ops *pm_ops); 1160 void mptcp_pm_get_available(char *buf, size_t maxlen); 1161 1162 void mptcp_userspace_pm_free_local_addr_list(struct mptcp_sock *msk); 1163 1164 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 1165 const struct sock *ssk, gfp_t gfp); 1166 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info); 1167 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 1168 void mptcp_event_pm_listener(const struct sock *ssk, 1169 enum mptcp_event_type event); 1170 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk); 1171 1172 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow, 1173 struct request_sock *req); 1174 int mptcp_pm_genl_fill_addr(struct sk_buff *msg, 1175 struct netlink_callback *cb, 1176 struct mptcp_pm_addr_entry *entry); 1177 1178 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 1179 { 1180 return READ_ONCE(msk->pm.addr_signal) & 1181 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO)); 1182 } 1183 1184 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk) 1185 { 1186 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 1187 } 1188 1189 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 1190 { 1191 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 1192 } 1193 1194 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 1195 { 1196 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 1197 } 1198 1199 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk) 1200 { 1201 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE; 1202 } 1203 1204 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk) 1205 { 1206 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL; 1207 } 1208 1209 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 1210 { 1211 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 1212 1213 if (family == AF_INET6) 1214 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 1215 if (!echo) 1216 len += MPTCPOPT_THMAC_LEN; 1217 /* account for 2 trailing 'nop' options */ 1218 if (port) 1219 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 1220 1221 return len; 1222 } 1223 1224 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 1225 { 1226 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 1227 return -EINVAL; 1228 1229 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 1230 } 1231 1232 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb, 1233 unsigned int opt_size, unsigned int remaining, 1234 struct mptcp_addr_info *addr, bool *echo, 1235 bool *drop_other_suboptions); 1236 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 1237 struct mptcp_rm_list *rm_list); 1238 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 1239 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, 1240 struct mptcp_pm_addr_entry *skc); 1241 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, 1242 struct mptcp_pm_addr_entry *skc); 1243 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc); 1244 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1245 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc); 1246 int mptcp_pm_nl_dump_addr(struct sk_buff *msg, 1247 struct netlink_callback *cb); 1248 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg, 1249 struct netlink_callback *cb); 1250 int mptcp_pm_nl_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1251 struct genl_info *info); 1252 int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr, 1253 struct genl_info *info); 1254 1255 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow) 1256 { 1257 int local_id = READ_ONCE(subflow->local_id); 1258 1259 if (local_id < 0) 1260 return 0; 1261 return local_id; 1262 } 1263 1264 void __init mptcp_pm_kernel_register(void); 1265 void __init mptcp_pm_userspace_register(void); 1266 void __init mptcp_pm_nl_init(void); 1267 void mptcp_pm_worker(struct mptcp_sock *msk); 1268 void __mptcp_pm_kernel_worker(struct mptcp_sock *msk); 1269 u8 mptcp_pm_get_endp_signal_max(const struct mptcp_sock *msk); 1270 u8 mptcp_pm_get_endp_subflow_max(const struct mptcp_sock *msk); 1271 u8 mptcp_pm_get_endp_laminar_max(const struct mptcp_sock *msk); 1272 u8 mptcp_pm_get_endp_fullmesh_max(const struct mptcp_sock *msk); 1273 u8 mptcp_pm_get_limit_add_addr_accepted(const struct mptcp_sock *msk); 1274 u8 mptcp_pm_get_limit_extra_subflows(const struct mptcp_sock *msk); 1275 1276 /* called under PM lock */ 1277 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk) 1278 { 1279 if (--msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk)) 1280 WRITE_ONCE(msk->pm.accept_subflow, true); 1281 } 1282 1283 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk) 1284 { 1285 spin_lock_bh(&msk->pm.lock); 1286 __mptcp_pm_close_subflow(msk); 1287 spin_unlock_bh(&msk->pm.lock); 1288 } 1289 1290 static inline bool mptcp_pm_add_addr_c_flag_case(struct mptcp_sock *msk) 1291 { 1292 return READ_ONCE(msk->pm.remote_deny_join_id0) && 1293 msk->pm.local_addr_used == 0 && 1294 mptcp_pm_get_limit_add_addr_accepted(msk) == 0 && 1295 msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk); 1296 } 1297 1298 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk); 1299 1300 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 1301 { 1302 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 1303 } 1304 1305 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 1306 1307 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 1308 { 1309 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 1310 } 1311 1312 static inline bool mptcp_check_fallback(const struct sock *sk) 1313 { 1314 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1315 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 1316 1317 return __mptcp_check_fallback(msk); 1318 } 1319 1320 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk) 1321 { 1322 struct sock *ssk = READ_ONCE(msk->first); 1323 1324 return ssk && ((1 << inet_sk_state_load(ssk)) & 1325 (TCPF_ESTABLISHED | TCPF_SYN_SENT | 1326 TCPF_SYN_RECV | TCPF_LISTEN)); 1327 } 1328 1329 bool __mptcp_try_fallback(struct mptcp_sock *msk, int fb_mib); 1330 1331 static inline bool mptcp_try_fallback(struct sock *ssk, int fb_mib) 1332 { 1333 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); 1334 struct sock *sk = subflow->conn; 1335 struct mptcp_sock *msk; 1336 1337 msk = mptcp_sk(sk); 1338 if (!__mptcp_try_fallback(msk, fb_mib)) 1339 return false; 1340 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) { 1341 gfp_t saved_allocation = ssk->sk_allocation; 1342 1343 /* we are in a atomic (BH) scope, override ssk default for data 1344 * fin allocation 1345 */ 1346 ssk->sk_allocation = GFP_ATOMIC; 1347 ssk->sk_shutdown |= SEND_SHUTDOWN; 1348 tcp_shutdown(ssk, SEND_SHUTDOWN); 1349 ssk->sk_allocation = saved_allocation; 1350 } 1351 return true; 1352 } 1353 1354 static inline void mptcp_early_fallback(struct mptcp_sock *msk, 1355 struct mptcp_subflow_context *subflow, 1356 int fb_mib) 1357 { 1358 subflow->request_mptcp = 0; 1359 WARN_ON_ONCE(!__mptcp_try_fallback(msk, fb_mib)); 1360 } 1361 1362 static inline bool mptcp_check_infinite_map(struct sk_buff *skb) 1363 { 1364 struct mptcp_ext *mpext; 1365 1366 mpext = skb ? mptcp_get_ext(skb) : NULL; 1367 if (mpext && mpext->infinite_map) 1368 return true; 1369 1370 return false; 1371 } 1372 1373 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow) 1374 { 1375 return (subflow->request_mptcp || subflow->request_join); 1376 } 1377 1378 static inline bool subflow_simultaneous_connect(struct sock *sk) 1379 { 1380 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 1381 1382 /* Note that the sk state implies !subflow->conn_finished. */ 1383 return sk->sk_state == TCP_SYN_RECV && is_active_ssk(subflow); 1384 } 1385 1386 #ifdef CONFIG_SYN_COOKIES 1387 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1388 struct sk_buff *skb); 1389 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1390 struct sk_buff *skb); 1391 void __init mptcp_join_cookie_init(void); 1392 #else 1393 static inline void 1394 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 1395 struct sk_buff *skb) {} 1396 static inline bool 1397 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 1398 struct sk_buff *skb) 1399 { 1400 return false; 1401 } 1402 1403 static inline void mptcp_join_cookie_init(void) {} 1404 #endif 1405 1406 #endif /* __MPTCP_PROTOCOL_H */ 1407