1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * Definitions for the TCP module. 7 * 8 * Version: @(#)tcp.h 1.0.5 05/23/93 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * 13 * This program is free software; you can redistribute it and/or 14 * modify it under the terms of the GNU General Public License 15 * as published by the Free Software Foundation; either version 16 * 2 of the License, or (at your option) any later version. 17 */ 18 #ifndef _TCP_H 19 #define _TCP_H 20 21 #define FASTRETRANS_DEBUG 1 22 23 #include <linux/list.h> 24 #include <linux/tcp.h> 25 #include <linux/bug.h> 26 #include <linux/slab.h> 27 #include <linux/cache.h> 28 #include <linux/percpu.h> 29 #include <linux/skbuff.h> 30 #include <linux/dmaengine.h> 31 #include <linux/crypto.h> 32 #include <linux/cryptohash.h> 33 #include <linux/kref.h> 34 35 #include <net/inet_connection_sock.h> 36 #include <net/inet_timewait_sock.h> 37 #include <net/inet_hashtables.h> 38 #include <net/checksum.h> 39 #include <net/request_sock.h> 40 #include <net/sock.h> 41 #include <net/snmp.h> 42 #include <net/ip.h> 43 #include <net/tcp_states.h> 44 #include <net/inet_ecn.h> 45 #include <net/dst.h> 46 47 #include <linux/seq_file.h> 48 #include <linux/memcontrol.h> 49 50 extern struct inet_hashinfo tcp_hashinfo; 51 52 extern struct percpu_counter tcp_orphan_count; 53 extern void tcp_time_wait(struct sock *sk, int state, int timeo); 54 55 #define MAX_TCP_HEADER (128 + MAX_HEADER) 56 #define MAX_TCP_OPTION_SPACE 40 57 58 /* 59 * Never offer a window over 32767 without using window scaling. Some 60 * poor stacks do signed 16bit maths! 61 */ 62 #define MAX_TCP_WINDOW 32767U 63 64 /* Offer an initial receive window of 10 mss. */ 65 #define TCP_DEFAULT_INIT_RCVWND 10 66 67 /* Minimal accepted MSS. It is (60+60+8) - (20+20). */ 68 #define TCP_MIN_MSS 88U 69 70 /* The least MTU to use for probing */ 71 #define TCP_BASE_MSS 512 72 73 /* After receiving this amount of duplicate ACKs fast retransmit starts. */ 74 #define TCP_FASTRETRANS_THRESH 3 75 76 /* Maximal reordering. */ 77 #define TCP_MAX_REORDERING 127 78 79 /* Maximal number of ACKs sent quickly to accelerate slow-start. */ 80 #define TCP_MAX_QUICKACKS 16U 81 82 /* urg_data states */ 83 #define TCP_URG_VALID 0x0100 84 #define TCP_URG_NOTYET 0x0200 85 #define TCP_URG_READ 0x0400 86 87 #define TCP_RETR1 3 /* 88 * This is how many retries it does before it 89 * tries to figure out if the gateway is 90 * down. Minimal RFC value is 3; it corresponds 91 * to ~3sec-8min depending on RTO. 92 */ 93 94 #define TCP_RETR2 15 /* 95 * This should take at least 96 * 90 minutes to time out. 97 * RFC1122 says that the limit is 100 sec. 98 * 15 is ~13-30min depending on RTO. 99 */ 100 101 #define TCP_SYN_RETRIES 6 /* This is how many retries are done 102 * when active opening a connection. 103 * RFC1122 says the minimum retry MUST 104 * be at least 180secs. Nevertheless 105 * this value is corresponding to 106 * 63secs of retransmission with the 107 * current initial RTO. 108 */ 109 110 #define TCP_SYNACK_RETRIES 5 /* This is how may retries are done 111 * when passive opening a connection. 112 * This is corresponding to 31secs of 113 * retransmission with the current 114 * initial RTO. 115 */ 116 117 #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT 118 * state, about 60 seconds */ 119 #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN 120 /* BSD style FIN_WAIT2 deadlock breaker. 121 * It used to be 3min, new value is 60sec, 122 * to combine FIN-WAIT-2 timeout with 123 * TIME-WAIT timer. 124 */ 125 126 #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */ 127 #if HZ >= 100 128 #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */ 129 #define TCP_ATO_MIN ((unsigned)(HZ/25)) 130 #else 131 #define TCP_DELACK_MIN 4U 132 #define TCP_ATO_MIN 4U 133 #endif 134 #define TCP_RTO_MAX ((unsigned)(120*HZ)) 135 #define TCP_RTO_MIN ((unsigned)(HZ/5)) 136 #define TCP_TIMEOUT_INIT ((unsigned)(1*HZ)) /* RFC6298 2.1 initial RTO value */ 137 #define TCP_TIMEOUT_FALLBACK ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value, now 138 * used as a fallback RTO for the 139 * initial data transmission if no 140 * valid RTT sample has been acquired, 141 * most likely due to retrans in 3WHS. 142 */ 143 144 #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes 145 * for local resources. 146 */ 147 148 #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */ 149 #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */ 150 #define TCP_KEEPALIVE_INTVL (75*HZ) 151 152 #define MAX_TCP_KEEPIDLE 32767 153 #define MAX_TCP_KEEPINTVL 32767 154 #define MAX_TCP_KEEPCNT 127 155 #define MAX_TCP_SYNCNT 127 156 157 #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */ 158 159 #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24) 160 #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated 161 * after this time. It should be equal 162 * (or greater than) TCP_TIMEWAIT_LEN 163 * to provide reliability equal to one 164 * provided by timewait state. 165 */ 166 #define TCP_PAWS_WINDOW 1 /* Replay window for per-host 167 * timestamps. It must be less than 168 * minimal timewait lifetime. 169 */ 170 /* 171 * TCP option 172 */ 173 174 #define TCPOPT_NOP 1 /* Padding */ 175 #define TCPOPT_EOL 0 /* End of options */ 176 #define TCPOPT_MSS 2 /* Segment size negotiating */ 177 #define TCPOPT_WINDOW 3 /* Window scaling */ 178 #define TCPOPT_SACK_PERM 4 /* SACK Permitted */ 179 #define TCPOPT_SACK 5 /* SACK Block */ 180 #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */ 181 #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */ 182 #define TCPOPT_EXP 254 /* Experimental */ 183 /* Magic number to be after the option value for sharing TCP 184 * experimental options. See draft-ietf-tcpm-experimental-options-00.txt 185 */ 186 #define TCPOPT_FASTOPEN_MAGIC 0xF989 187 188 /* 189 * TCP option lengths 190 */ 191 192 #define TCPOLEN_MSS 4 193 #define TCPOLEN_WINDOW 3 194 #define TCPOLEN_SACK_PERM 2 195 #define TCPOLEN_TIMESTAMP 10 196 #define TCPOLEN_MD5SIG 18 197 #define TCPOLEN_EXP_FASTOPEN_BASE 4 198 #define TCPOLEN_COOKIE_BASE 2 /* Cookie-less header extension */ 199 #define TCPOLEN_COOKIE_PAIR 3 /* Cookie pair header extension */ 200 #define TCPOLEN_COOKIE_MIN (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MIN) 201 #define TCPOLEN_COOKIE_MAX (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MAX) 202 203 /* But this is what stacks really send out. */ 204 #define TCPOLEN_TSTAMP_ALIGNED 12 205 #define TCPOLEN_WSCALE_ALIGNED 4 206 #define TCPOLEN_SACKPERM_ALIGNED 4 207 #define TCPOLEN_SACK_BASE 2 208 #define TCPOLEN_SACK_BASE_ALIGNED 4 209 #define TCPOLEN_SACK_PERBLOCK 8 210 #define TCPOLEN_MD5SIG_ALIGNED 20 211 #define TCPOLEN_MSS_ALIGNED 4 212 213 /* Flags in tp->nonagle */ 214 #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */ 215 #define TCP_NAGLE_CORK 2 /* Socket is corked */ 216 #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */ 217 218 /* TCP thin-stream limits */ 219 #define TCP_THIN_LINEAR_RETRIES 6 /* After 6 linear retries, do exp. backoff */ 220 221 /* TCP initial congestion window as per draft-hkchu-tcpm-initcwnd-01 */ 222 #define TCP_INIT_CWND 10 223 224 /* Bit Flags for sysctl_tcp_fastopen */ 225 #define TFO_CLIENT_ENABLE 1 226 #define TFO_SERVER_ENABLE 2 227 #define TFO_CLIENT_NO_COOKIE 4 /* Data in SYN w/o cookie option */ 228 229 /* Process SYN data but skip cookie validation */ 230 #define TFO_SERVER_COOKIE_NOT_CHKED 0x100 231 /* Accept SYN data w/o any cookie option */ 232 #define TFO_SERVER_COOKIE_NOT_REQD 0x200 233 234 /* Force enable TFO on all listeners, i.e., not requiring the 235 * TCP_FASTOPEN socket option. SOCKOPT1/2 determine how to set max_qlen. 236 */ 237 #define TFO_SERVER_WO_SOCKOPT1 0x400 238 #define TFO_SERVER_WO_SOCKOPT2 0x800 239 /* Always create TFO child sockets on a TFO listener even when 240 * cookie/data not present. (For testing purpose!) 241 */ 242 #define TFO_SERVER_ALWAYS 0x1000 243 244 extern struct inet_timewait_death_row tcp_death_row; 245 246 /* sysctl variables for tcp */ 247 extern int sysctl_tcp_timestamps; 248 extern int sysctl_tcp_window_scaling; 249 extern int sysctl_tcp_sack; 250 extern int sysctl_tcp_fin_timeout; 251 extern int sysctl_tcp_keepalive_time; 252 extern int sysctl_tcp_keepalive_probes; 253 extern int sysctl_tcp_keepalive_intvl; 254 extern int sysctl_tcp_syn_retries; 255 extern int sysctl_tcp_synack_retries; 256 extern int sysctl_tcp_retries1; 257 extern int sysctl_tcp_retries2; 258 extern int sysctl_tcp_orphan_retries; 259 extern int sysctl_tcp_syncookies; 260 extern int sysctl_tcp_fastopen; 261 extern int sysctl_tcp_retrans_collapse; 262 extern int sysctl_tcp_stdurg; 263 extern int sysctl_tcp_rfc1337; 264 extern int sysctl_tcp_abort_on_overflow; 265 extern int sysctl_tcp_max_orphans; 266 extern int sysctl_tcp_fack; 267 extern int sysctl_tcp_reordering; 268 extern int sysctl_tcp_dsack; 269 extern int sysctl_tcp_wmem[3]; 270 extern int sysctl_tcp_rmem[3]; 271 extern int sysctl_tcp_app_win; 272 extern int sysctl_tcp_adv_win_scale; 273 extern int sysctl_tcp_tw_reuse; 274 extern int sysctl_tcp_frto; 275 extern int sysctl_tcp_low_latency; 276 extern int sysctl_tcp_dma_copybreak; 277 extern int sysctl_tcp_nometrics_save; 278 extern int sysctl_tcp_moderate_rcvbuf; 279 extern int sysctl_tcp_tso_win_divisor; 280 extern int sysctl_tcp_mtu_probing; 281 extern int sysctl_tcp_base_mss; 282 extern int sysctl_tcp_workaround_signed_windows; 283 extern int sysctl_tcp_slow_start_after_idle; 284 extern int sysctl_tcp_max_ssthresh; 285 extern int sysctl_tcp_thin_linear_timeouts; 286 extern int sysctl_tcp_thin_dupack; 287 extern int sysctl_tcp_early_retrans; 288 extern int sysctl_tcp_limit_output_bytes; 289 extern int sysctl_tcp_challenge_ack_limit; 290 291 extern atomic_long_t tcp_memory_allocated; 292 extern struct percpu_counter tcp_sockets_allocated; 293 extern int tcp_memory_pressure; 294 295 /* 296 * The next routines deal with comparing 32 bit unsigned ints 297 * and worry about wraparound (automatic with unsigned arithmetic). 298 */ 299 300 static inline bool before(__u32 seq1, __u32 seq2) 301 { 302 return (__s32)(seq1-seq2) < 0; 303 } 304 #define after(seq2, seq1) before(seq1, seq2) 305 306 /* is s2<=s1<=s3 ? */ 307 static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3) 308 { 309 return seq3 - seq2 >= seq1 - seq2; 310 } 311 312 static inline bool tcp_out_of_memory(struct sock *sk) 313 { 314 if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF && 315 sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2)) 316 return true; 317 return false; 318 } 319 320 static inline bool tcp_too_many_orphans(struct sock *sk, int shift) 321 { 322 struct percpu_counter *ocp = sk->sk_prot->orphan_count; 323 int orphans = percpu_counter_read_positive(ocp); 324 325 if (orphans << shift > sysctl_tcp_max_orphans) { 326 orphans = percpu_counter_sum_positive(ocp); 327 if (orphans << shift > sysctl_tcp_max_orphans) 328 return true; 329 } 330 return false; 331 } 332 333 extern bool tcp_check_oom(struct sock *sk, int shift); 334 335 /* syncookies: remember time of last synqueue overflow */ 336 static inline void tcp_synq_overflow(struct sock *sk) 337 { 338 tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies; 339 } 340 341 /* syncookies: no recent synqueue overflow on this listening socket? */ 342 static inline bool tcp_synq_no_recent_overflow(const struct sock *sk) 343 { 344 unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp; 345 return time_after(jiffies, last_overflow + TCP_TIMEOUT_FALLBACK); 346 } 347 348 extern struct proto tcp_prot; 349 350 #define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field) 351 #define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field) 352 #define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field) 353 #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val) 354 #define TCP_ADD_STATS(net, field, val) SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val) 355 356 extern void tcp_init_mem(struct net *net); 357 358 extern void tcp_tasklet_init(void); 359 360 extern void tcp_v4_err(struct sk_buff *skb, u32); 361 362 extern void tcp_shutdown (struct sock *sk, int how); 363 364 extern void tcp_v4_early_demux(struct sk_buff *skb); 365 extern int tcp_v4_rcv(struct sk_buff *skb); 366 367 extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw); 368 extern int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 369 size_t size); 370 extern int tcp_sendpage(struct sock *sk, struct page *page, int offset, 371 size_t size, int flags); 372 extern void tcp_release_cb(struct sock *sk); 373 extern void tcp_wfree(struct sk_buff *skb); 374 extern void tcp_write_timer_handler(struct sock *sk); 375 extern void tcp_delack_timer_handler(struct sock *sk); 376 extern int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg); 377 extern int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb, 378 const struct tcphdr *th, unsigned int len); 379 extern int tcp_rcv_established(struct sock *sk, struct sk_buff *skb, 380 const struct tcphdr *th, unsigned int len); 381 extern void tcp_rcv_space_adjust(struct sock *sk); 382 extern void tcp_cleanup_rbuf(struct sock *sk, int copied); 383 extern int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp); 384 extern void tcp_twsk_destructor(struct sock *sk); 385 extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos, 386 struct pipe_inode_info *pipe, size_t len, 387 unsigned int flags); 388 389 static inline void tcp_dec_quickack_mode(struct sock *sk, 390 const unsigned int pkts) 391 { 392 struct inet_connection_sock *icsk = inet_csk(sk); 393 394 if (icsk->icsk_ack.quick) { 395 if (pkts >= icsk->icsk_ack.quick) { 396 icsk->icsk_ack.quick = 0; 397 /* Leaving quickack mode we deflate ATO. */ 398 icsk->icsk_ack.ato = TCP_ATO_MIN; 399 } else 400 icsk->icsk_ack.quick -= pkts; 401 } 402 } 403 404 #define TCP_ECN_OK 1 405 #define TCP_ECN_QUEUE_CWR 2 406 #define TCP_ECN_DEMAND_CWR 4 407 #define TCP_ECN_SEEN 8 408 409 enum tcp_tw_status { 410 TCP_TW_SUCCESS = 0, 411 TCP_TW_RST = 1, 412 TCP_TW_ACK = 2, 413 TCP_TW_SYN = 3 414 }; 415 416 417 extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw, 418 struct sk_buff *skb, 419 const struct tcphdr *th); 420 extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb, 421 struct request_sock *req, 422 struct request_sock **prev, 423 bool fastopen); 424 extern int tcp_child_process(struct sock *parent, struct sock *child, 425 struct sk_buff *skb); 426 extern void tcp_enter_loss(struct sock *sk, int how); 427 extern void tcp_clear_retrans(struct tcp_sock *tp); 428 extern void tcp_update_metrics(struct sock *sk); 429 extern void tcp_init_metrics(struct sock *sk); 430 extern void tcp_metrics_init(void); 431 extern bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst, bool paws_check); 432 extern bool tcp_remember_stamp(struct sock *sk); 433 extern bool tcp_tw_remember_stamp(struct inet_timewait_sock *tw); 434 extern void tcp_fetch_timewait_stamp(struct sock *sk, struct dst_entry *dst); 435 extern void tcp_disable_fack(struct tcp_sock *tp); 436 extern void tcp_close(struct sock *sk, long timeout); 437 extern void tcp_init_sock(struct sock *sk); 438 extern unsigned int tcp_poll(struct file * file, struct socket *sock, 439 struct poll_table_struct *wait); 440 extern int tcp_getsockopt(struct sock *sk, int level, int optname, 441 char __user *optval, int __user *optlen); 442 extern int tcp_setsockopt(struct sock *sk, int level, int optname, 443 char __user *optval, unsigned int optlen); 444 extern int compat_tcp_getsockopt(struct sock *sk, int level, int optname, 445 char __user *optval, int __user *optlen); 446 extern int compat_tcp_setsockopt(struct sock *sk, int level, int optname, 447 char __user *optval, unsigned int optlen); 448 extern void tcp_set_keepalive(struct sock *sk, int val); 449 extern void tcp_syn_ack_timeout(struct sock *sk, struct request_sock *req); 450 extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 451 size_t len, int nonblock, int flags, int *addr_len); 452 extern void tcp_parse_options(const struct sk_buff *skb, 453 struct tcp_options_received *opt_rx, 454 int estab, struct tcp_fastopen_cookie *foc); 455 extern const u8 *tcp_parse_md5sig_option(const struct tcphdr *th); 456 457 /* 458 * TCP v4 functions exported for the inet6 API 459 */ 460 461 extern void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb); 462 extern int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb); 463 extern struct sock * tcp_create_openreq_child(struct sock *sk, 464 struct request_sock *req, 465 struct sk_buff *skb); 466 extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb, 467 struct request_sock *req, 468 struct dst_entry *dst); 469 extern int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb); 470 extern int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, 471 int addr_len); 472 extern int tcp_connect(struct sock *sk); 473 extern struct sk_buff * tcp_make_synack(struct sock *sk, struct dst_entry *dst, 474 struct request_sock *req, 475 struct tcp_fastopen_cookie *foc); 476 extern int tcp_disconnect(struct sock *sk, int flags); 477 478 void tcp_connect_init(struct sock *sk); 479 void tcp_finish_connect(struct sock *sk, struct sk_buff *skb); 480 int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size); 481 void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb); 482 483 /* From syncookies.c */ 484 extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS]; 485 extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb, 486 struct ip_options *opt); 487 #ifdef CONFIG_SYN_COOKIES 488 extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, 489 __u16 *mss); 490 #else 491 static inline __u32 cookie_v4_init_sequence(struct sock *sk, 492 struct sk_buff *skb, 493 __u16 *mss) 494 { 495 return 0; 496 } 497 #endif 498 499 extern __u32 cookie_init_timestamp(struct request_sock *req); 500 extern bool cookie_check_timestamp(struct tcp_options_received *opt, 501 struct net *net, bool *ecn_ok); 502 503 /* From net/ipv6/syncookies.c */ 504 extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb); 505 #ifdef CONFIG_SYN_COOKIES 506 extern __u32 cookie_v6_init_sequence(struct sock *sk, const struct sk_buff *skb, 507 __u16 *mss); 508 #else 509 static inline __u32 cookie_v6_init_sequence(struct sock *sk, 510 struct sk_buff *skb, 511 __u16 *mss) 512 { 513 return 0; 514 } 515 #endif 516 /* tcp_output.c */ 517 518 extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss, 519 int nonagle); 520 extern bool tcp_may_send_now(struct sock *sk); 521 extern int __tcp_retransmit_skb(struct sock *, struct sk_buff *); 522 extern int tcp_retransmit_skb(struct sock *, struct sk_buff *); 523 extern void tcp_retransmit_timer(struct sock *sk); 524 extern void tcp_xmit_retransmit_queue(struct sock *); 525 extern void tcp_simple_retransmit(struct sock *); 526 extern int tcp_trim_head(struct sock *, struct sk_buff *, u32); 527 extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int); 528 529 extern void tcp_send_probe0(struct sock *); 530 extern void tcp_send_partial(struct sock *); 531 extern int tcp_write_wakeup(struct sock *); 532 extern void tcp_send_fin(struct sock *sk); 533 extern void tcp_send_active_reset(struct sock *sk, gfp_t priority); 534 extern int tcp_send_synack(struct sock *); 535 extern bool tcp_syn_flood_action(struct sock *sk, 536 const struct sk_buff *skb, 537 const char *proto); 538 extern void tcp_push_one(struct sock *, unsigned int mss_now); 539 extern void tcp_send_ack(struct sock *sk); 540 extern void tcp_send_delayed_ack(struct sock *sk); 541 extern void tcp_send_loss_probe(struct sock *sk); 542 extern bool tcp_schedule_loss_probe(struct sock *sk); 543 544 /* tcp_input.c */ 545 extern void tcp_cwnd_application_limited(struct sock *sk); 546 extern void tcp_resume_early_retransmit(struct sock *sk); 547 extern void tcp_rearm_rto(struct sock *sk); 548 extern void tcp_reset(struct sock *sk); 549 550 /* tcp_timer.c */ 551 extern void tcp_init_xmit_timers(struct sock *); 552 static inline void tcp_clear_xmit_timers(struct sock *sk) 553 { 554 inet_csk_clear_xmit_timers(sk); 555 } 556 557 extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu); 558 extern unsigned int tcp_current_mss(struct sock *sk); 559 560 /* Bound MSS / TSO packet size with the half of the window */ 561 static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize) 562 { 563 int cutoff; 564 565 /* When peer uses tiny windows, there is no use in packetizing 566 * to sub-MSS pieces for the sake of SWS or making sure there 567 * are enough packets in the pipe for fast recovery. 568 * 569 * On the other hand, for extremely large MSS devices, handling 570 * smaller than MSS windows in this way does make sense. 571 */ 572 if (tp->max_window >= 512) 573 cutoff = (tp->max_window >> 1); 574 else 575 cutoff = tp->max_window; 576 577 if (cutoff && pktsize > cutoff) 578 return max_t(int, cutoff, 68U - tp->tcp_header_len); 579 else 580 return pktsize; 581 } 582 583 /* tcp.c */ 584 extern void tcp_get_info(const struct sock *, struct tcp_info *); 585 586 /* Read 'sendfile()'-style from a TCP socket */ 587 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, 588 unsigned int, size_t); 589 extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc, 590 sk_read_actor_t recv_actor); 591 592 extern void tcp_initialize_rcv_mss(struct sock *sk); 593 594 extern int tcp_mtu_to_mss(struct sock *sk, int pmtu); 595 extern int tcp_mss_to_mtu(struct sock *sk, int mss); 596 extern void tcp_mtup_init(struct sock *sk); 597 extern void tcp_valid_rtt_meas(struct sock *sk, u32 seq_rtt); 598 extern void tcp_init_buffer_space(struct sock *sk); 599 600 static inline void tcp_bound_rto(const struct sock *sk) 601 { 602 if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX) 603 inet_csk(sk)->icsk_rto = TCP_RTO_MAX; 604 } 605 606 static inline u32 __tcp_set_rto(const struct tcp_sock *tp) 607 { 608 return (tp->srtt >> 3) + tp->rttvar; 609 } 610 611 extern void tcp_set_rto(struct sock *sk); 612 613 static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd) 614 { 615 tp->pred_flags = htonl((tp->tcp_header_len << 26) | 616 ntohl(TCP_FLAG_ACK) | 617 snd_wnd); 618 } 619 620 static inline void tcp_fast_path_on(struct tcp_sock *tp) 621 { 622 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale); 623 } 624 625 static inline void tcp_fast_path_check(struct sock *sk) 626 { 627 struct tcp_sock *tp = tcp_sk(sk); 628 629 if (skb_queue_empty(&tp->out_of_order_queue) && 630 tp->rcv_wnd && 631 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && 632 !tp->urg_data) 633 tcp_fast_path_on(tp); 634 } 635 636 /* Compute the actual rto_min value */ 637 static inline u32 tcp_rto_min(struct sock *sk) 638 { 639 const struct dst_entry *dst = __sk_dst_get(sk); 640 u32 rto_min = TCP_RTO_MIN; 641 642 if (dst && dst_metric_locked(dst, RTAX_RTO_MIN)) 643 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN); 644 return rto_min; 645 } 646 647 /* Compute the actual receive window we are currently advertising. 648 * Rcv_nxt can be after the window if our peer push more data 649 * than the offered window. 650 */ 651 static inline u32 tcp_receive_window(const struct tcp_sock *tp) 652 { 653 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt; 654 655 if (win < 0) 656 win = 0; 657 return (u32) win; 658 } 659 660 /* Choose a new window, without checks for shrinking, and without 661 * scaling applied to the result. The caller does these things 662 * if necessary. This is a "raw" window selection. 663 */ 664 extern u32 __tcp_select_window(struct sock *sk); 665 666 void tcp_send_window_probe(struct sock *sk); 667 668 /* TCP timestamps are only 32-bits, this causes a slight 669 * complication on 64-bit systems since we store a snapshot 670 * of jiffies in the buffer control blocks below. We decided 671 * to use only the low 32-bits of jiffies and hide the ugly 672 * casts with the following macro. 673 */ 674 #define tcp_time_stamp ((__u32)(jiffies)) 675 676 #define tcp_flag_byte(th) (((u_int8_t *)th)[13]) 677 678 #define TCPHDR_FIN 0x01 679 #define TCPHDR_SYN 0x02 680 #define TCPHDR_RST 0x04 681 #define TCPHDR_PSH 0x08 682 #define TCPHDR_ACK 0x10 683 #define TCPHDR_URG 0x20 684 #define TCPHDR_ECE 0x40 685 #define TCPHDR_CWR 0x80 686 687 /* This is what the send packet queuing engine uses to pass 688 * TCP per-packet control information to the transmission code. 689 * We also store the host-order sequence numbers in here too. 690 * This is 44 bytes if IPV6 is enabled. 691 * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately. 692 */ 693 struct tcp_skb_cb { 694 union { 695 struct inet_skb_parm h4; 696 #if IS_ENABLED(CONFIG_IPV6) 697 struct inet6_skb_parm h6; 698 #endif 699 } header; /* For incoming frames */ 700 __u32 seq; /* Starting sequence number */ 701 __u32 end_seq; /* SEQ + FIN + SYN + datalen */ 702 __u32 when; /* used to compute rtt's */ 703 __u8 tcp_flags; /* TCP header flags. (tcp[13]) */ 704 705 __u8 sacked; /* State flags for SACK/FACK. */ 706 #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */ 707 #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */ 708 #define TCPCB_LOST 0x04 /* SKB is lost */ 709 #define TCPCB_TAGBITS 0x07 /* All tag bits */ 710 #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */ 711 #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS) 712 713 __u8 ip_dsfield; /* IPv4 tos or IPv6 dsfield */ 714 /* 1 byte hole */ 715 __u32 ack_seq; /* Sequence number ACK'd */ 716 }; 717 718 #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) 719 720 /* RFC3168 : 6.1.1 SYN packets must not have ECT/ECN bits set 721 * 722 * If we receive a SYN packet with these bits set, it means a network is 723 * playing bad games with TOS bits. In order to avoid possible false congestion 724 * notifications, we disable TCP ECN negociation. 725 */ 726 static inline void 727 TCP_ECN_create_request(struct request_sock *req, const struct sk_buff *skb, 728 struct net *net) 729 { 730 const struct tcphdr *th = tcp_hdr(skb); 731 732 if (net->ipv4.sysctl_tcp_ecn && th->ece && th->cwr && 733 INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield)) 734 inet_rsk(req)->ecn_ok = 1; 735 } 736 737 /* Due to TSO, an SKB can be composed of multiple actual 738 * packets. To keep these tracked properly, we use this. 739 */ 740 static inline int tcp_skb_pcount(const struct sk_buff *skb) 741 { 742 return skb_shinfo(skb)->gso_segs; 743 } 744 745 /* This is valid iff tcp_skb_pcount() > 1. */ 746 static inline int tcp_skb_mss(const struct sk_buff *skb) 747 { 748 return skb_shinfo(skb)->gso_size; 749 } 750 751 /* Events passed to congestion control interface */ 752 enum tcp_ca_event { 753 CA_EVENT_TX_START, /* first transmit when no packets in flight */ 754 CA_EVENT_CWND_RESTART, /* congestion window restart */ 755 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */ 756 CA_EVENT_LOSS, /* loss timeout */ 757 CA_EVENT_FAST_ACK, /* in sequence ack */ 758 CA_EVENT_SLOW_ACK, /* other ack */ 759 }; 760 761 /* 762 * Interface for adding new TCP congestion control handlers 763 */ 764 #define TCP_CA_NAME_MAX 16 765 #define TCP_CA_MAX 128 766 #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX) 767 768 #define TCP_CONG_NON_RESTRICTED 0x1 769 #define TCP_CONG_RTT_STAMP 0x2 770 771 struct tcp_congestion_ops { 772 struct list_head list; 773 unsigned long flags; 774 775 /* initialize private data (optional) */ 776 void (*init)(struct sock *sk); 777 /* cleanup private data (optional) */ 778 void (*release)(struct sock *sk); 779 780 /* return slow start threshold (required) */ 781 u32 (*ssthresh)(struct sock *sk); 782 /* lower bound for congestion window (optional) */ 783 u32 (*min_cwnd)(const struct sock *sk); 784 /* do new cwnd calculation (required) */ 785 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight); 786 /* call before changing ca_state (optional) */ 787 void (*set_state)(struct sock *sk, u8 new_state); 788 /* call when cwnd event occurs (optional) */ 789 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev); 790 /* new value of cwnd after loss (optional) */ 791 u32 (*undo_cwnd)(struct sock *sk); 792 /* hook for packet ack accounting (optional) */ 793 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us); 794 /* get info for inet_diag (optional) */ 795 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb); 796 797 char name[TCP_CA_NAME_MAX]; 798 struct module *owner; 799 }; 800 801 extern int tcp_register_congestion_control(struct tcp_congestion_ops *type); 802 extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type); 803 804 extern void tcp_init_congestion_control(struct sock *sk); 805 extern void tcp_cleanup_congestion_control(struct sock *sk); 806 extern int tcp_set_default_congestion_control(const char *name); 807 extern void tcp_get_default_congestion_control(char *name); 808 extern void tcp_get_available_congestion_control(char *buf, size_t len); 809 extern void tcp_get_allowed_congestion_control(char *buf, size_t len); 810 extern int tcp_set_allowed_congestion_control(char *allowed); 811 extern int tcp_set_congestion_control(struct sock *sk, const char *name); 812 extern void tcp_slow_start(struct tcp_sock *tp); 813 extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w); 814 815 extern struct tcp_congestion_ops tcp_init_congestion_ops; 816 extern u32 tcp_reno_ssthresh(struct sock *sk); 817 extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight); 818 extern u32 tcp_reno_min_cwnd(const struct sock *sk); 819 extern struct tcp_congestion_ops tcp_reno; 820 821 static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state) 822 { 823 struct inet_connection_sock *icsk = inet_csk(sk); 824 825 if (icsk->icsk_ca_ops->set_state) 826 icsk->icsk_ca_ops->set_state(sk, ca_state); 827 icsk->icsk_ca_state = ca_state; 828 } 829 830 static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) 831 { 832 const struct inet_connection_sock *icsk = inet_csk(sk); 833 834 if (icsk->icsk_ca_ops->cwnd_event) 835 icsk->icsk_ca_ops->cwnd_event(sk, event); 836 } 837 838 /* These functions determine how the current flow behaves in respect of SACK 839 * handling. SACK is negotiated with the peer, and therefore it can vary 840 * between different flows. 841 * 842 * tcp_is_sack - SACK enabled 843 * tcp_is_reno - No SACK 844 * tcp_is_fack - FACK enabled, implies SACK enabled 845 */ 846 static inline int tcp_is_sack(const struct tcp_sock *tp) 847 { 848 return tp->rx_opt.sack_ok; 849 } 850 851 static inline bool tcp_is_reno(const struct tcp_sock *tp) 852 { 853 return !tcp_is_sack(tp); 854 } 855 856 static inline bool tcp_is_fack(const struct tcp_sock *tp) 857 { 858 return tp->rx_opt.sack_ok & TCP_FACK_ENABLED; 859 } 860 861 static inline void tcp_enable_fack(struct tcp_sock *tp) 862 { 863 tp->rx_opt.sack_ok |= TCP_FACK_ENABLED; 864 } 865 866 /* TCP early-retransmit (ER) is similar to but more conservative than 867 * the thin-dupack feature. Enable ER only if thin-dupack is disabled. 868 */ 869 static inline void tcp_enable_early_retrans(struct tcp_sock *tp) 870 { 871 tp->do_early_retrans = sysctl_tcp_early_retrans && 872 sysctl_tcp_early_retrans < 4 && !sysctl_tcp_thin_dupack && 873 sysctl_tcp_reordering == 3; 874 } 875 876 static inline void tcp_disable_early_retrans(struct tcp_sock *tp) 877 { 878 tp->do_early_retrans = 0; 879 } 880 881 static inline unsigned int tcp_left_out(const struct tcp_sock *tp) 882 { 883 return tp->sacked_out + tp->lost_out; 884 } 885 886 /* This determines how many packets are "in the network" to the best 887 * of our knowledge. In many cases it is conservative, but where 888 * detailed information is available from the receiver (via SACK 889 * blocks etc.) we can make more aggressive calculations. 890 * 891 * Use this for decisions involving congestion control, use just 892 * tp->packets_out to determine if the send queue is empty or not. 893 * 894 * Read this equation as: 895 * 896 * "Packets sent once on transmission queue" MINUS 897 * "Packets left network, but not honestly ACKed yet" PLUS 898 * "Packets fast retransmitted" 899 */ 900 static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) 901 { 902 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out; 903 } 904 905 #define TCP_INFINITE_SSTHRESH 0x7fffffff 906 907 static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp) 908 { 909 return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH; 910 } 911 912 static inline bool tcp_in_cwnd_reduction(const struct sock *sk) 913 { 914 return (TCPF_CA_CWR | TCPF_CA_Recovery) & 915 (1 << inet_csk(sk)->icsk_ca_state); 916 } 917 918 /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. 919 * The exception is cwnd reduction phase, when cwnd is decreasing towards 920 * ssthresh. 921 */ 922 static inline __u32 tcp_current_ssthresh(const struct sock *sk) 923 { 924 const struct tcp_sock *tp = tcp_sk(sk); 925 926 if (tcp_in_cwnd_reduction(sk)) 927 return tp->snd_ssthresh; 928 else 929 return max(tp->snd_ssthresh, 930 ((tp->snd_cwnd >> 1) + 931 (tp->snd_cwnd >> 2))); 932 } 933 934 /* Use define here intentionally to get WARN_ON location shown at the caller */ 935 #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out) 936 937 extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh); 938 extern __u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst); 939 940 /* The maximum number of MSS of available cwnd for which TSO defers 941 * sending if not using sysctl_tcp_tso_win_divisor. 942 */ 943 static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp) 944 { 945 return 3; 946 } 947 948 /* Slow start with delack produces 3 packets of burst, so that 949 * it is safe "de facto". This will be the default - same as 950 * the default reordering threshold - but if reordering increases, 951 * we must be able to allow cwnd to burst at least this much in order 952 * to not pull it back when holes are filled. 953 */ 954 static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp) 955 { 956 return tp->reordering; 957 } 958 959 /* Returns end sequence number of the receiver's advertised window */ 960 static inline u32 tcp_wnd_end(const struct tcp_sock *tp) 961 { 962 return tp->snd_una + tp->snd_wnd; 963 } 964 extern bool tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight); 965 966 static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss, 967 const struct sk_buff *skb) 968 { 969 if (skb->len < mss) 970 tp->snd_sml = TCP_SKB_CB(skb)->end_seq; 971 } 972 973 static inline void tcp_check_probe_timer(struct sock *sk) 974 { 975 const struct tcp_sock *tp = tcp_sk(sk); 976 const struct inet_connection_sock *icsk = inet_csk(sk); 977 978 if (!tp->packets_out && !icsk->icsk_pending) 979 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, 980 icsk->icsk_rto, TCP_RTO_MAX); 981 } 982 983 static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq) 984 { 985 tp->snd_wl1 = seq; 986 } 987 988 static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq) 989 { 990 tp->snd_wl1 = seq; 991 } 992 993 /* 994 * Calculate(/check) TCP checksum 995 */ 996 static inline __sum16 tcp_v4_check(int len, __be32 saddr, 997 __be32 daddr, __wsum base) 998 { 999 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base); 1000 } 1001 1002 static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb) 1003 { 1004 return __skb_checksum_complete(skb); 1005 } 1006 1007 static inline bool tcp_checksum_complete(struct sk_buff *skb) 1008 { 1009 return !skb_csum_unnecessary(skb) && 1010 __tcp_checksum_complete(skb); 1011 } 1012 1013 /* Prequeue for VJ style copy to user, combined with checksumming. */ 1014 1015 static inline void tcp_prequeue_init(struct tcp_sock *tp) 1016 { 1017 tp->ucopy.task = NULL; 1018 tp->ucopy.len = 0; 1019 tp->ucopy.memory = 0; 1020 skb_queue_head_init(&tp->ucopy.prequeue); 1021 #ifdef CONFIG_NET_DMA 1022 tp->ucopy.dma_chan = NULL; 1023 tp->ucopy.wakeup = 0; 1024 tp->ucopy.pinned_list = NULL; 1025 tp->ucopy.dma_cookie = 0; 1026 #endif 1027 } 1028 1029 extern bool tcp_prequeue(struct sock *sk, struct sk_buff *skb); 1030 1031 #undef STATE_TRACE 1032 1033 #ifdef STATE_TRACE 1034 static const char *statename[]={ 1035 "Unused","Established","Syn Sent","Syn Recv", 1036 "Fin Wait 1","Fin Wait 2","Time Wait", "Close", 1037 "Close Wait","Last ACK","Listen","Closing" 1038 }; 1039 #endif 1040 extern void tcp_set_state(struct sock *sk, int state); 1041 1042 extern void tcp_done(struct sock *sk); 1043 1044 static inline void tcp_sack_reset(struct tcp_options_received *rx_opt) 1045 { 1046 rx_opt->dsack = 0; 1047 rx_opt->num_sacks = 0; 1048 } 1049 1050 /* Determine a window scaling and initial window to offer. */ 1051 extern void tcp_select_initial_window(int __space, __u32 mss, 1052 __u32 *rcv_wnd, __u32 *window_clamp, 1053 int wscale_ok, __u8 *rcv_wscale, 1054 __u32 init_rcv_wnd); 1055 1056 static inline int tcp_win_from_space(int space) 1057 { 1058 return sysctl_tcp_adv_win_scale<=0 ? 1059 (space>>(-sysctl_tcp_adv_win_scale)) : 1060 space - (space>>sysctl_tcp_adv_win_scale); 1061 } 1062 1063 /* Note: caller must be prepared to deal with negative returns */ 1064 static inline int tcp_space(const struct sock *sk) 1065 { 1066 return tcp_win_from_space(sk->sk_rcvbuf - 1067 atomic_read(&sk->sk_rmem_alloc)); 1068 } 1069 1070 static inline int tcp_full_space(const struct sock *sk) 1071 { 1072 return tcp_win_from_space(sk->sk_rcvbuf); 1073 } 1074 1075 static inline void tcp_openreq_init(struct request_sock *req, 1076 struct tcp_options_received *rx_opt, 1077 struct sk_buff *skb) 1078 { 1079 struct inet_request_sock *ireq = inet_rsk(req); 1080 1081 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */ 1082 req->cookie_ts = 0; 1083 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq; 1084 tcp_rsk(req)->rcv_nxt = TCP_SKB_CB(skb)->seq + 1; 1085 tcp_rsk(req)->snt_synack = 0; 1086 req->mss = rx_opt->mss_clamp; 1087 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0; 1088 ireq->tstamp_ok = rx_opt->tstamp_ok; 1089 ireq->sack_ok = rx_opt->sack_ok; 1090 ireq->snd_wscale = rx_opt->snd_wscale; 1091 ireq->wscale_ok = rx_opt->wscale_ok; 1092 ireq->acked = 0; 1093 ireq->ecn_ok = 0; 1094 ireq->rmt_port = tcp_hdr(skb)->source; 1095 ireq->loc_port = tcp_hdr(skb)->dest; 1096 } 1097 1098 /* Compute time elapsed between SYNACK and the ACK completing 3WHS */ 1099 static inline void tcp_synack_rtt_meas(struct sock *sk, 1100 struct request_sock *req) 1101 { 1102 if (tcp_rsk(req)->snt_synack) 1103 tcp_valid_rtt_meas(sk, 1104 tcp_time_stamp - tcp_rsk(req)->snt_synack); 1105 } 1106 1107 extern void tcp_enter_memory_pressure(struct sock *sk); 1108 1109 static inline int keepalive_intvl_when(const struct tcp_sock *tp) 1110 { 1111 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl; 1112 } 1113 1114 static inline int keepalive_time_when(const struct tcp_sock *tp) 1115 { 1116 return tp->keepalive_time ? : sysctl_tcp_keepalive_time; 1117 } 1118 1119 static inline int keepalive_probes(const struct tcp_sock *tp) 1120 { 1121 return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes; 1122 } 1123 1124 static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp) 1125 { 1126 const struct inet_connection_sock *icsk = &tp->inet_conn; 1127 1128 return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime, 1129 tcp_time_stamp - tp->rcv_tstamp); 1130 } 1131 1132 static inline int tcp_fin_time(const struct sock *sk) 1133 { 1134 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout; 1135 const int rto = inet_csk(sk)->icsk_rto; 1136 1137 if (fin_timeout < (rto << 2) - (rto >> 1)) 1138 fin_timeout = (rto << 2) - (rto >> 1); 1139 1140 return fin_timeout; 1141 } 1142 1143 static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt, 1144 int paws_win) 1145 { 1146 if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win) 1147 return true; 1148 if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)) 1149 return true; 1150 /* 1151 * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0, 1152 * then following tcp messages have valid values. Ignore 0 value, 1153 * or else 'negative' tsval might forbid us to accept their packets. 1154 */ 1155 if (!rx_opt->ts_recent) 1156 return true; 1157 return false; 1158 } 1159 1160 static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt, 1161 int rst) 1162 { 1163 if (tcp_paws_check(rx_opt, 0)) 1164 return false; 1165 1166 /* RST segments are not recommended to carry timestamp, 1167 and, if they do, it is recommended to ignore PAWS because 1168 "their cleanup function should take precedence over timestamps." 1169 Certainly, it is mistake. It is necessary to understand the reasons 1170 of this constraint to relax it: if peer reboots, clock may go 1171 out-of-sync and half-open connections will not be reset. 1172 Actually, the problem would be not existing if all 1173 the implementations followed draft about maintaining clock 1174 via reboots. Linux-2.2 DOES NOT! 1175 1176 However, we can relax time bounds for RST segments to MSL. 1177 */ 1178 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL) 1179 return false; 1180 return true; 1181 } 1182 1183 static inline void tcp_mib_init(struct net *net) 1184 { 1185 /* See RFC 2012 */ 1186 TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1); 1187 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ); 1188 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ); 1189 TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1); 1190 } 1191 1192 /* from STCP */ 1193 static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp) 1194 { 1195 tp->lost_skb_hint = NULL; 1196 tp->scoreboard_skb_hint = NULL; 1197 } 1198 1199 static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp) 1200 { 1201 tcp_clear_retrans_hints_partial(tp); 1202 tp->retransmit_skb_hint = NULL; 1203 } 1204 1205 /* MD5 Signature */ 1206 struct crypto_hash; 1207 1208 union tcp_md5_addr { 1209 struct in_addr a4; 1210 #if IS_ENABLED(CONFIG_IPV6) 1211 struct in6_addr a6; 1212 #endif 1213 }; 1214 1215 /* - key database */ 1216 struct tcp_md5sig_key { 1217 struct hlist_node node; 1218 u8 keylen; 1219 u8 family; /* AF_INET or AF_INET6 */ 1220 union tcp_md5_addr addr; 1221 u8 key[TCP_MD5SIG_MAXKEYLEN]; 1222 struct rcu_head rcu; 1223 }; 1224 1225 /* - sock block */ 1226 struct tcp_md5sig_info { 1227 struct hlist_head head; 1228 struct rcu_head rcu; 1229 }; 1230 1231 /* - pseudo header */ 1232 struct tcp4_pseudohdr { 1233 __be32 saddr; 1234 __be32 daddr; 1235 __u8 pad; 1236 __u8 protocol; 1237 __be16 len; 1238 }; 1239 1240 struct tcp6_pseudohdr { 1241 struct in6_addr saddr; 1242 struct in6_addr daddr; 1243 __be32 len; 1244 __be32 protocol; /* including padding */ 1245 }; 1246 1247 union tcp_md5sum_block { 1248 struct tcp4_pseudohdr ip4; 1249 #if IS_ENABLED(CONFIG_IPV6) 1250 struct tcp6_pseudohdr ip6; 1251 #endif 1252 }; 1253 1254 /* - pool: digest algorithm, hash description and scratch buffer */ 1255 struct tcp_md5sig_pool { 1256 struct hash_desc md5_desc; 1257 union tcp_md5sum_block md5_blk; 1258 }; 1259 1260 /* - functions */ 1261 extern int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key, 1262 const struct sock *sk, 1263 const struct request_sock *req, 1264 const struct sk_buff *skb); 1265 extern int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr, 1266 int family, const u8 *newkey, 1267 u8 newkeylen, gfp_t gfp); 1268 extern int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, 1269 int family); 1270 extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, 1271 struct sock *addr_sk); 1272 1273 #ifdef CONFIG_TCP_MD5SIG 1274 extern struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk, 1275 const union tcp_md5_addr *addr, int family); 1276 #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_key) 1277 #else 1278 static inline struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk, 1279 const union tcp_md5_addr *addr, 1280 int family) 1281 { 1282 return NULL; 1283 } 1284 #define tcp_twsk_md5_key(twsk) NULL 1285 #endif 1286 1287 extern struct tcp_md5sig_pool __percpu *tcp_alloc_md5sig_pool(struct sock *); 1288 extern void tcp_free_md5sig_pool(void); 1289 1290 extern struct tcp_md5sig_pool *tcp_get_md5sig_pool(void); 1291 extern void tcp_put_md5sig_pool(void); 1292 1293 extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, const struct tcphdr *); 1294 extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *, 1295 unsigned int header_len); 1296 extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp, 1297 const struct tcp_md5sig_key *key); 1298 1299 /* From tcp_fastopen.c */ 1300 extern void tcp_fastopen_cache_get(struct sock *sk, u16 *mss, 1301 struct tcp_fastopen_cookie *cookie, 1302 int *syn_loss, unsigned long *last_syn_loss); 1303 extern void tcp_fastopen_cache_set(struct sock *sk, u16 mss, 1304 struct tcp_fastopen_cookie *cookie, 1305 bool syn_lost); 1306 struct tcp_fastopen_request { 1307 /* Fast Open cookie. Size 0 means a cookie request */ 1308 struct tcp_fastopen_cookie cookie; 1309 struct msghdr *data; /* data in MSG_FASTOPEN */ 1310 u16 copied; /* queued in tcp_connect() */ 1311 }; 1312 void tcp_free_fastopen_req(struct tcp_sock *tp); 1313 1314 extern struct tcp_fastopen_context __rcu *tcp_fastopen_ctx; 1315 int tcp_fastopen_reset_cipher(void *key, unsigned int len); 1316 void tcp_fastopen_cookie_gen(__be32 addr, struct tcp_fastopen_cookie *foc); 1317 1318 #define TCP_FASTOPEN_KEY_LENGTH 16 1319 1320 /* Fastopen key context */ 1321 struct tcp_fastopen_context { 1322 struct crypto_cipher __rcu *tfm; 1323 __u8 key[TCP_FASTOPEN_KEY_LENGTH]; 1324 struct rcu_head rcu; 1325 }; 1326 1327 /* write queue abstraction */ 1328 static inline void tcp_write_queue_purge(struct sock *sk) 1329 { 1330 struct sk_buff *skb; 1331 1332 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) 1333 sk_wmem_free_skb(sk, skb); 1334 sk_mem_reclaim(sk); 1335 tcp_clear_all_retrans_hints(tcp_sk(sk)); 1336 } 1337 1338 static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk) 1339 { 1340 return skb_peek(&sk->sk_write_queue); 1341 } 1342 1343 static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk) 1344 { 1345 return skb_peek_tail(&sk->sk_write_queue); 1346 } 1347 1348 static inline struct sk_buff *tcp_write_queue_next(const struct sock *sk, 1349 const struct sk_buff *skb) 1350 { 1351 return skb_queue_next(&sk->sk_write_queue, skb); 1352 } 1353 1354 static inline struct sk_buff *tcp_write_queue_prev(const struct sock *sk, 1355 const struct sk_buff *skb) 1356 { 1357 return skb_queue_prev(&sk->sk_write_queue, skb); 1358 } 1359 1360 #define tcp_for_write_queue(skb, sk) \ 1361 skb_queue_walk(&(sk)->sk_write_queue, skb) 1362 1363 #define tcp_for_write_queue_from(skb, sk) \ 1364 skb_queue_walk_from(&(sk)->sk_write_queue, skb) 1365 1366 #define tcp_for_write_queue_from_safe(skb, tmp, sk) \ 1367 skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp) 1368 1369 static inline struct sk_buff *tcp_send_head(const struct sock *sk) 1370 { 1371 return sk->sk_send_head; 1372 } 1373 1374 static inline bool tcp_skb_is_last(const struct sock *sk, 1375 const struct sk_buff *skb) 1376 { 1377 return skb_queue_is_last(&sk->sk_write_queue, skb); 1378 } 1379 1380 static inline void tcp_advance_send_head(struct sock *sk, const struct sk_buff *skb) 1381 { 1382 if (tcp_skb_is_last(sk, skb)) 1383 sk->sk_send_head = NULL; 1384 else 1385 sk->sk_send_head = tcp_write_queue_next(sk, skb); 1386 } 1387 1388 static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked) 1389 { 1390 if (sk->sk_send_head == skb_unlinked) 1391 sk->sk_send_head = NULL; 1392 } 1393 1394 static inline void tcp_init_send_head(struct sock *sk) 1395 { 1396 sk->sk_send_head = NULL; 1397 } 1398 1399 static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) 1400 { 1401 __skb_queue_tail(&sk->sk_write_queue, skb); 1402 } 1403 1404 static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) 1405 { 1406 __tcp_add_write_queue_tail(sk, skb); 1407 1408 /* Queue it, remembering where we must start sending. */ 1409 if (sk->sk_send_head == NULL) { 1410 sk->sk_send_head = skb; 1411 1412 if (tcp_sk(sk)->highest_sack == NULL) 1413 tcp_sk(sk)->highest_sack = skb; 1414 } 1415 } 1416 1417 static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb) 1418 { 1419 __skb_queue_head(&sk->sk_write_queue, skb); 1420 } 1421 1422 /* Insert buff after skb on the write queue of sk. */ 1423 static inline void tcp_insert_write_queue_after(struct sk_buff *skb, 1424 struct sk_buff *buff, 1425 struct sock *sk) 1426 { 1427 __skb_queue_after(&sk->sk_write_queue, skb, buff); 1428 } 1429 1430 /* Insert new before skb on the write queue of sk. */ 1431 static inline void tcp_insert_write_queue_before(struct sk_buff *new, 1432 struct sk_buff *skb, 1433 struct sock *sk) 1434 { 1435 __skb_queue_before(&sk->sk_write_queue, skb, new); 1436 1437 if (sk->sk_send_head == skb) 1438 sk->sk_send_head = new; 1439 } 1440 1441 static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk) 1442 { 1443 __skb_unlink(skb, &sk->sk_write_queue); 1444 } 1445 1446 static inline bool tcp_write_queue_empty(struct sock *sk) 1447 { 1448 return skb_queue_empty(&sk->sk_write_queue); 1449 } 1450 1451 static inline void tcp_push_pending_frames(struct sock *sk) 1452 { 1453 if (tcp_send_head(sk)) { 1454 struct tcp_sock *tp = tcp_sk(sk); 1455 1456 __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle); 1457 } 1458 } 1459 1460 /* Start sequence of the skb just after the highest skb with SACKed 1461 * bit, valid only if sacked_out > 0 or when the caller has ensured 1462 * validity by itself. 1463 */ 1464 static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp) 1465 { 1466 if (!tp->sacked_out) 1467 return tp->snd_una; 1468 1469 if (tp->highest_sack == NULL) 1470 return tp->snd_nxt; 1471 1472 return TCP_SKB_CB(tp->highest_sack)->seq; 1473 } 1474 1475 static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb) 1476 { 1477 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL : 1478 tcp_write_queue_next(sk, skb); 1479 } 1480 1481 static inline struct sk_buff *tcp_highest_sack(struct sock *sk) 1482 { 1483 return tcp_sk(sk)->highest_sack; 1484 } 1485 1486 static inline void tcp_highest_sack_reset(struct sock *sk) 1487 { 1488 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk); 1489 } 1490 1491 /* Called when old skb is about to be deleted (to be combined with new skb) */ 1492 static inline void tcp_highest_sack_combine(struct sock *sk, 1493 struct sk_buff *old, 1494 struct sk_buff *new) 1495 { 1496 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack)) 1497 tcp_sk(sk)->highest_sack = new; 1498 } 1499 1500 /* Determines whether this is a thin stream (which may suffer from 1501 * increased latency). Used to trigger latency-reducing mechanisms. 1502 */ 1503 static inline bool tcp_stream_is_thin(struct tcp_sock *tp) 1504 { 1505 return tp->packets_out < 4 && !tcp_in_initial_slowstart(tp); 1506 } 1507 1508 /* /proc */ 1509 enum tcp_seq_states { 1510 TCP_SEQ_STATE_LISTENING, 1511 TCP_SEQ_STATE_OPENREQ, 1512 TCP_SEQ_STATE_ESTABLISHED, 1513 TCP_SEQ_STATE_TIME_WAIT, 1514 }; 1515 1516 int tcp_seq_open(struct inode *inode, struct file *file); 1517 1518 struct tcp_seq_afinfo { 1519 char *name; 1520 sa_family_t family; 1521 const struct file_operations *seq_fops; 1522 struct seq_operations seq_ops; 1523 }; 1524 1525 struct tcp_iter_state { 1526 struct seq_net_private p; 1527 sa_family_t family; 1528 enum tcp_seq_states state; 1529 struct sock *syn_wait_sk; 1530 int bucket, offset, sbucket, num; 1531 kuid_t uid; 1532 loff_t last_pos; 1533 }; 1534 1535 extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo); 1536 extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo); 1537 1538 extern struct request_sock_ops tcp_request_sock_ops; 1539 extern struct request_sock_ops tcp6_request_sock_ops; 1540 1541 extern void tcp_v4_destroy_sock(struct sock *sk); 1542 1543 extern int tcp_v4_gso_send_check(struct sk_buff *skb); 1544 extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, 1545 netdev_features_t features); 1546 extern struct sk_buff **tcp_gro_receive(struct sk_buff **head, 1547 struct sk_buff *skb); 1548 extern struct sk_buff **tcp4_gro_receive(struct sk_buff **head, 1549 struct sk_buff *skb); 1550 extern int tcp_gro_complete(struct sk_buff *skb); 1551 extern int tcp4_gro_complete(struct sk_buff *skb); 1552 1553 #ifdef CONFIG_PROC_FS 1554 extern int tcp4_proc_init(void); 1555 extern void tcp4_proc_exit(void); 1556 #endif 1557 1558 /* TCP af-specific functions */ 1559 struct tcp_sock_af_ops { 1560 #ifdef CONFIG_TCP_MD5SIG 1561 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk, 1562 struct sock *addr_sk); 1563 int (*calc_md5_hash) (char *location, 1564 struct tcp_md5sig_key *md5, 1565 const struct sock *sk, 1566 const struct request_sock *req, 1567 const struct sk_buff *skb); 1568 int (*md5_parse) (struct sock *sk, 1569 char __user *optval, 1570 int optlen); 1571 #endif 1572 }; 1573 1574 struct tcp_request_sock_ops { 1575 #ifdef CONFIG_TCP_MD5SIG 1576 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk, 1577 struct request_sock *req); 1578 int (*calc_md5_hash) (char *location, 1579 struct tcp_md5sig_key *md5, 1580 const struct sock *sk, 1581 const struct request_sock *req, 1582 const struct sk_buff *skb); 1583 #endif 1584 }; 1585 1586 extern void tcp_v4_init(void); 1587 extern void tcp_init(void); 1588 1589 #endif /* _TCP_H */ 1590