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 14 #define MPTCP_SUPPORTED_VERSION 1 15 16 /* MPTCP option bits */ 17 #define OPTION_MPTCP_MPC_SYN BIT(0) 18 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 19 #define OPTION_MPTCP_MPC_ACK BIT(2) 20 #define OPTION_MPTCP_MPJ_SYN BIT(3) 21 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 22 #define OPTION_MPTCP_MPJ_ACK BIT(5) 23 #define OPTION_MPTCP_ADD_ADDR BIT(6) 24 #define OPTION_MPTCP_ADD_ADDR6 BIT(7) 25 #define OPTION_MPTCP_RM_ADDR BIT(8) 26 27 /* MPTCP option subtypes */ 28 #define MPTCPOPT_MP_CAPABLE 0 29 #define MPTCPOPT_MP_JOIN 1 30 #define MPTCPOPT_DSS 2 31 #define MPTCPOPT_ADD_ADDR 3 32 #define MPTCPOPT_RM_ADDR 4 33 #define MPTCPOPT_MP_PRIO 5 34 #define MPTCPOPT_MP_FAIL 6 35 #define MPTCPOPT_MP_FASTCLOSE 7 36 37 /* MPTCP suboption lengths */ 38 #define TCPOLEN_MPTCP_MPC_SYN 4 39 #define TCPOLEN_MPTCP_MPC_SYNACK 12 40 #define TCPOLEN_MPTCP_MPC_ACK 20 41 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 42 #define TCPOLEN_MPTCP_MPJ_SYN 12 43 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 44 #define TCPOLEN_MPTCP_MPJ_ACK 24 45 #define TCPOLEN_MPTCP_DSS_BASE 4 46 #define TCPOLEN_MPTCP_DSS_ACK32 4 47 #define TCPOLEN_MPTCP_DSS_ACK64 8 48 #define TCPOLEN_MPTCP_DSS_MAP32 10 49 #define TCPOLEN_MPTCP_DSS_MAP64 14 50 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 51 #define TCPOLEN_MPTCP_ADD_ADDR 16 52 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 53 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 54 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 55 #define TCPOLEN_MPTCP_ADD_ADDR6 28 56 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 57 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 58 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 59 #define TCPOLEN_MPTCP_PORT_LEN 2 60 #define TCPOLEN_MPTCP_RM_ADDR_BASE 4 61 62 /* MPTCP MP_JOIN flags */ 63 #define MPTCPOPT_BACKUP BIT(0) 64 #define MPTCPOPT_HMAC_LEN 20 65 #define MPTCPOPT_THMAC_LEN 8 66 67 /* MPTCP MP_CAPABLE flags */ 68 #define MPTCP_VERSION_MASK (0x0F) 69 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 70 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 71 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 72 #define MPTCP_CAP_FLAG_MASK (0x3F) 73 74 /* MPTCP DSS flags */ 75 #define MPTCP_DSS_DATA_FIN BIT(4) 76 #define MPTCP_DSS_DSN64 BIT(3) 77 #define MPTCP_DSS_HAS_MAP BIT(2) 78 #define MPTCP_DSS_ACK64 BIT(1) 79 #define MPTCP_DSS_HAS_ACK BIT(0) 80 #define MPTCP_DSS_FLAG_MASK (0x1F) 81 82 /* MPTCP ADD_ADDR flags */ 83 #define MPTCP_ADDR_ECHO BIT(0) 84 #define MPTCP_ADDR_IPVERSION_4 4 85 #define MPTCP_ADDR_IPVERSION_6 6 86 87 /* MPTCP socket flags */ 88 #define MPTCP_DATA_READY 0 89 #define MPTCP_SEND_SPACE 1 90 #define MPTCP_WORK_RTX 2 91 #define MPTCP_WORK_EOF 3 92 93 struct mptcp_options_received { 94 u64 sndr_key; 95 u64 rcvr_key; 96 u64 data_ack; 97 u64 data_seq; 98 u32 subflow_seq; 99 u16 data_len; 100 u16 mp_capable : 1, 101 mp_join : 1, 102 dss : 1, 103 add_addr : 1, 104 rm_addr : 1, 105 family : 4, 106 echo : 1, 107 backup : 1; 108 u32 token; 109 u32 nonce; 110 u64 thmac; 111 u8 hmac[20]; 112 u8 join_id; 113 u8 use_map:1, 114 dsn64:1, 115 data_fin:1, 116 use_ack:1, 117 ack64:1, 118 mpc_map:1, 119 __unused:2; 120 u8 addr_id; 121 u8 rm_id; 122 union { 123 struct in_addr addr; 124 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 125 struct in6_addr addr6; 126 #endif 127 }; 128 u64 ahmac; 129 u16 port; 130 }; 131 132 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 133 { 134 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 135 ((nib & 0xF) << 8) | field); 136 } 137 138 #define MPTCP_PM_MAX_ADDR 4 139 140 struct mptcp_addr_info { 141 sa_family_t family; 142 __be16 port; 143 u8 id; 144 union { 145 struct in_addr addr; 146 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 147 struct in6_addr addr6; 148 #endif 149 }; 150 }; 151 152 enum mptcp_pm_status { 153 MPTCP_PM_ADD_ADDR_RECEIVED, 154 MPTCP_PM_ESTABLISHED, 155 MPTCP_PM_SUBFLOW_ESTABLISHED, 156 }; 157 158 struct mptcp_pm_data { 159 struct mptcp_addr_info local; 160 struct mptcp_addr_info remote; 161 162 spinlock_t lock; /*protects the whole PM data */ 163 164 bool addr_signal; 165 bool server_side; 166 bool work_pending; 167 bool accept_addr; 168 bool accept_subflow; 169 u8 add_addr_signaled; 170 u8 add_addr_accepted; 171 u8 local_addr_used; 172 u8 subflows; 173 u8 add_addr_signal_max; 174 u8 add_addr_accept_max; 175 u8 local_addr_max; 176 u8 subflows_max; 177 u8 status; 178 179 struct work_struct work; 180 }; 181 182 struct mptcp_data_frag { 183 struct list_head list; 184 u64 data_seq; 185 int data_len; 186 int offset; 187 int overhead; 188 struct page *page; 189 }; 190 191 /* MPTCP connection sock */ 192 struct mptcp_sock { 193 /* inet_connection_sock must be the first member */ 194 struct inet_connection_sock sk; 195 u64 local_key; 196 u64 remote_key; 197 u64 write_seq; 198 u64 ack_seq; 199 atomic64_t snd_una; 200 unsigned long timer_ival; 201 u32 token; 202 unsigned long flags; 203 bool can_ack; 204 spinlock_t join_list_lock; 205 struct work_struct work; 206 struct list_head conn_list; 207 struct list_head rtx_queue; 208 struct list_head join_list; 209 struct skb_ext *cached_ext; /* for the next sendmsg */ 210 struct socket *subflow; /* outgoing connect/listener/!mp_capable */ 211 struct sock *first; 212 struct mptcp_pm_data pm; 213 }; 214 215 #define mptcp_for_each_subflow(__msk, __subflow) \ 216 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 217 218 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 219 { 220 return (struct mptcp_sock *)sk; 221 } 222 223 static inline struct mptcp_data_frag *mptcp_rtx_tail(const struct sock *sk) 224 { 225 struct mptcp_sock *msk = mptcp_sk(sk); 226 227 if (list_empty(&msk->rtx_queue)) 228 return NULL; 229 230 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 231 } 232 233 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 234 { 235 struct mptcp_sock *msk = mptcp_sk(sk); 236 237 if (list_empty(&msk->rtx_queue)) 238 return NULL; 239 240 return list_first_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 241 } 242 243 struct mptcp_subflow_request_sock { 244 struct tcp_request_sock sk; 245 u16 mp_capable : 1, 246 mp_join : 1, 247 backup : 1; 248 u8 local_id; 249 u8 remote_id; 250 u64 local_key; 251 u64 idsn; 252 u32 token; 253 u32 ssn_offset; 254 u64 thmac; 255 u32 local_nonce; 256 u32 remote_nonce; 257 }; 258 259 static inline struct mptcp_subflow_request_sock * 260 mptcp_subflow_rsk(const struct request_sock *rsk) 261 { 262 return (struct mptcp_subflow_request_sock *)rsk; 263 } 264 265 /* MPTCP subflow context */ 266 struct mptcp_subflow_context { 267 struct list_head node;/* conn_list of subflows */ 268 u64 local_key; 269 u64 remote_key; 270 u64 idsn; 271 u64 map_seq; 272 u32 snd_isn; 273 u32 token; 274 u32 rel_write_seq; 275 u32 map_subflow_seq; 276 u32 ssn_offset; 277 u32 map_data_len; 278 u32 request_mptcp : 1, /* send MP_CAPABLE */ 279 request_join : 1, /* send MP_JOIN */ 280 request_bkup : 1, 281 mp_capable : 1, /* remote is MPTCP capable */ 282 mp_join : 1, /* remote is JOINing */ 283 fully_established : 1, /* path validated */ 284 pm_notified : 1, /* PM hook called for established status */ 285 conn_finished : 1, 286 map_valid : 1, 287 mpc_map : 1, 288 backup : 1, 289 data_avail : 1, 290 rx_eof : 1, 291 data_fin_tx_enable : 1, 292 can_ack : 1; /* only after processing the remote a key */ 293 u64 data_fin_tx_seq; 294 u32 remote_nonce; 295 u64 thmac; 296 u32 local_nonce; 297 u32 remote_token; 298 u8 hmac[MPTCPOPT_HMAC_LEN]; 299 u8 local_id; 300 u8 remote_id; 301 302 struct sock *tcp_sock; /* tcp sk backpointer */ 303 struct sock *conn; /* parent mptcp_sock */ 304 const struct inet_connection_sock_af_ops *icsk_af_ops; 305 void (*tcp_data_ready)(struct sock *sk); 306 void (*tcp_state_change)(struct sock *sk); 307 void (*tcp_write_space)(struct sock *sk); 308 309 struct rcu_head rcu; 310 }; 311 312 static inline struct mptcp_subflow_context * 313 mptcp_subflow_ctx(const struct sock *sk) 314 { 315 struct inet_connection_sock *icsk = inet_csk(sk); 316 317 /* Use RCU on icsk_ulp_data only for sock diag code */ 318 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 319 } 320 321 static inline struct sock * 322 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 323 { 324 return subflow->tcp_sock; 325 } 326 327 static inline u64 328 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 329 { 330 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 331 subflow->ssn_offset - 332 subflow->map_subflow_seq; 333 } 334 335 static inline u64 336 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 337 { 338 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 339 } 340 341 int mptcp_is_enabled(struct net *net); 342 bool mptcp_subflow_data_available(struct sock *sk); 343 void mptcp_subflow_init(void); 344 345 /* called with sk socket lock held */ 346 int __mptcp_subflow_connect(struct sock *sk, int ifindex, 347 const struct mptcp_addr_info *loc, 348 const struct mptcp_addr_info *remote); 349 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); 350 351 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 352 struct mptcp_subflow_context *ctx) 353 { 354 sk->sk_data_ready = ctx->tcp_data_ready; 355 sk->sk_state_change = ctx->tcp_state_change; 356 sk->sk_write_space = ctx->tcp_write_space; 357 358 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 359 } 360 361 extern const struct inet_connection_sock_af_ops ipv4_specific; 362 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 363 extern const struct inet_connection_sock_af_ops ipv6_specific; 364 #endif 365 366 void mptcp_proto_init(void); 367 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 368 int mptcp_proto_v6_init(void); 369 #endif 370 371 struct sock *mptcp_sk_clone(const struct sock *sk, 372 const struct mptcp_options_received *mp_opt, 373 struct request_sock *req); 374 void mptcp_get_options(const struct sk_buff *skb, 375 struct mptcp_options_received *mp_opt); 376 377 void mptcp_finish_connect(struct sock *sk); 378 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 379 bool mptcp_finish_join(struct sock *sk); 380 void mptcp_data_acked(struct sock *sk); 381 void mptcp_subflow_eof(struct sock *sk); 382 383 int mptcp_token_new_request(struct request_sock *req); 384 void mptcp_token_destroy_request(u32 token); 385 int mptcp_token_new_connect(struct sock *sk); 386 int mptcp_token_new_accept(u32 token, struct sock *conn); 387 struct mptcp_sock *mptcp_token_get_sock(u32 token); 388 void mptcp_token_destroy(u32 token); 389 390 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 391 static inline void mptcp_crypto_key_gen_sha(u64 *key, u32 *token, u64 *idsn) 392 { 393 /* we might consider a faster version that computes the key as a 394 * hash of some information available in the MPTCP socket. Use 395 * random data at the moment, as it's probably the safest option 396 * in case multiple sockets are opened in different namespaces at 397 * the same time. 398 */ 399 get_random_bytes(key, sizeof(u64)); 400 mptcp_crypto_key_sha(*key, token, idsn); 401 } 402 403 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 404 405 void mptcp_pm_init(void); 406 void mptcp_pm_data_init(struct mptcp_sock *msk); 407 void mptcp_pm_close(struct mptcp_sock *msk); 408 void mptcp_pm_new_connection(struct mptcp_sock *msk, int server_side); 409 void mptcp_pm_fully_established(struct mptcp_sock *msk); 410 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 411 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 412 void mptcp_pm_subflow_established(struct mptcp_sock *msk, 413 struct mptcp_subflow_context *subflow); 414 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id); 415 void mptcp_pm_add_addr_received(struct mptcp_sock *msk, 416 const struct mptcp_addr_info *addr); 417 418 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 419 const struct mptcp_addr_info *addr); 420 int mptcp_pm_remove_addr(struct mptcp_sock *msk, u8 local_id); 421 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, u8 remote_id); 422 423 static inline bool mptcp_pm_should_signal(struct mptcp_sock *msk) 424 { 425 return READ_ONCE(msk->pm.addr_signal); 426 } 427 428 static inline unsigned int mptcp_add_addr_len(int family) 429 { 430 if (family == AF_INET) 431 return TCPOLEN_MPTCP_ADD_ADDR; 432 return TCPOLEN_MPTCP_ADD_ADDR6; 433 } 434 435 bool mptcp_pm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 436 struct mptcp_addr_info *saddr); 437 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 438 439 void mptcp_pm_nl_init(void); 440 void mptcp_pm_nl_data_init(struct mptcp_sock *msk); 441 void mptcp_pm_nl_fully_established(struct mptcp_sock *msk); 442 void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk); 443 void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk); 444 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 445 446 static inline struct mptcp_ext *mptcp_get_ext(struct sk_buff *skb) 447 { 448 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 449 } 450 451 static inline bool before64(__u64 seq1, __u64 seq2) 452 { 453 return (__s64)(seq1 - seq2) < 0; 454 } 455 456 #define after64(seq2, seq1) before64(seq1, seq2) 457 458 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 459 460 #endif /* __MPTCP_PROTOCOL_H */ 461