| 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 TCP_DEBUG 1 |
| 22 | #define FASTRETRANS_DEBUG 1 |
| 23 | |
| 24 | #include <linux/list.h> |
| 25 | #include <linux/tcp.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 | |
| 34 | #include <net/inet_connection_sock.h> |
| 35 | #include <net/inet_timewait_sock.h> |
| 36 | #include <net/inet_hashtables.h> |
| 37 | #include <net/checksum.h> |
| 38 | #include <net/request_sock.h> |
| 39 | #include <net/sock.h> |
| 40 | #include <net/snmp.h> |
| 41 | #include <net/ip.h> |
| 42 | #include <net/tcp_states.h> |
| 43 | #include <net/inet_ecn.h> |
| 44 | |
| 45 | #include <linux/seq_file.h> |
| 46 | |
| 47 | extern struct inet_hashinfo tcp_hashinfo; |
| 48 | |
| 49 | extern atomic_t tcp_orphan_count; |
| 50 | extern void tcp_time_wait(struct sock *sk, int state, int timeo); |
| 51 | |
| 52 | #define MAX_TCP_HEADER (128 + MAX_HEADER) |
| 53 | #define MAX_TCP_OPTION_SPACE 40 |
| 54 | |
| 55 | /* |
| 56 | * Never offer a window over 32767 without using window scaling. Some |
| 57 | * poor stacks do signed 16bit maths! |
| 58 | */ |
| 59 | #define MAX_TCP_WINDOW 32767U |
| 60 | |
| 61 | /* Minimal accepted MSS. It is (60+60+8) - (20+20). */ |
| 62 | #define TCP_MIN_MSS 88U |
| 63 | |
| 64 | /* Minimal RCV_MSS. */ |
| 65 | #define TCP_MIN_RCVMSS 536U |
| 66 | |
| 67 | /* The least MTU to use for probing */ |
| 68 | #define TCP_BASE_MSS 512 |
| 69 | |
| 70 | /* After receiving this amount of duplicate ACKs fast retransmit starts. */ |
| 71 | #define TCP_FASTRETRANS_THRESH 3 |
| 72 | |
| 73 | /* Maximal reordering. */ |
| 74 | #define TCP_MAX_REORDERING 127 |
| 75 | |
| 76 | /* Maximal number of ACKs sent quickly to accelerate slow-start. */ |
| 77 | #define TCP_MAX_QUICKACKS 16U |
| 78 | |
| 79 | /* urg_data states */ |
| 80 | #define TCP_URG_VALID 0x0100 |
| 81 | #define TCP_URG_NOTYET 0x0200 |
| 82 | #define TCP_URG_READ 0x0400 |
| 83 | |
| 84 | #define TCP_RETR1 3 /* |
| 85 | * This is how many retries it does before it |
| 86 | * tries to figure out if the gateway is |
| 87 | * down. Minimal RFC value is 3; it corresponds |
| 88 | * to ~3sec-8min depending on RTO. |
| 89 | */ |
| 90 | |
| 91 | #define TCP_RETR2 15 /* |
| 92 | * This should take at least |
| 93 | * 90 minutes to time out. |
| 94 | * RFC1122 says that the limit is 100 sec. |
| 95 | * 15 is ~13-30min depending on RTO. |
| 96 | */ |
| 97 | |
| 98 | #define TCP_SYN_RETRIES 5 /* number of times to retry active opening a |
| 99 | * connection: ~180sec is RFC minimum */ |
| 100 | |
| 101 | #define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a |
| 102 | * connection: ~180sec is RFC minimum */ |
| 103 | |
| 104 | |
| 105 | #define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned |
| 106 | * socket. 7 is ~50sec-16min. |
| 107 | */ |
| 108 | |
| 109 | |
| 110 | #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT |
| 111 | * state, about 60 seconds */ |
| 112 | #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN |
| 113 | /* BSD style FIN_WAIT2 deadlock breaker. |
| 114 | * It used to be 3min, new value is 60sec, |
| 115 | * to combine FIN-WAIT-2 timeout with |
| 116 | * TIME-WAIT timer. |
| 117 | */ |
| 118 | |
| 119 | #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */ |
| 120 | #if HZ >= 100 |
| 121 | #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */ |
| 122 | #define TCP_ATO_MIN ((unsigned)(HZ/25)) |
| 123 | #else |
| 124 | #define TCP_DELACK_MIN 4U |
| 125 | #define TCP_ATO_MIN 4U |
| 126 | #endif |
| 127 | #define TCP_RTO_MAX ((unsigned)(120*HZ)) |
| 128 | #define TCP_RTO_MIN ((unsigned)(HZ/5)) |
| 129 | #define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */ |
| 130 | |
| 131 | #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes |
| 132 | * for local resources. |
| 133 | */ |
| 134 | |
| 135 | #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */ |
| 136 | #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */ |
| 137 | #define TCP_KEEPALIVE_INTVL (75*HZ) |
| 138 | |
| 139 | #define MAX_TCP_KEEPIDLE 32767 |
| 140 | #define MAX_TCP_KEEPINTVL 32767 |
| 141 | #define MAX_TCP_KEEPCNT 127 |
| 142 | #define MAX_TCP_SYNCNT 127 |
| 143 | |
| 144 | #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */ |
| 145 | |
| 146 | #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24) |
| 147 | #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated |
| 148 | * after this time. It should be equal |
| 149 | * (or greater than) TCP_TIMEWAIT_LEN |
| 150 | * to provide reliability equal to one |
| 151 | * provided by timewait state. |
| 152 | */ |
| 153 | #define TCP_PAWS_WINDOW 1 /* Replay window for per-host |
| 154 | * timestamps. It must be less than |
| 155 | * minimal timewait lifetime. |
| 156 | */ |
| 157 | /* |
| 158 | * TCP option |
| 159 | */ |
| 160 | |
| 161 | #define TCPOPT_NOP 1 /* Padding */ |
| 162 | #define TCPOPT_EOL 0 /* End of options */ |
| 163 | #define TCPOPT_MSS 2 /* Segment size negotiating */ |
| 164 | #define TCPOPT_WINDOW 3 /* Window scaling */ |
| 165 | #define TCPOPT_SACK_PERM 4 /* SACK Permitted */ |
| 166 | #define TCPOPT_SACK 5 /* SACK Block */ |
| 167 | #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */ |
| 168 | #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */ |
| 169 | |
| 170 | /* |
| 171 | * TCP option lengths |
| 172 | */ |
| 173 | |
| 174 | #define TCPOLEN_MSS 4 |
| 175 | #define TCPOLEN_WINDOW 3 |
| 176 | #define TCPOLEN_SACK_PERM 2 |
| 177 | #define TCPOLEN_TIMESTAMP 10 |
| 178 | #define TCPOLEN_MD5SIG 18 |
| 179 | |
| 180 | /* But this is what stacks really send out. */ |
| 181 | #define TCPOLEN_TSTAMP_ALIGNED 12 |
| 182 | #define TCPOLEN_WSCALE_ALIGNED 4 |
| 183 | #define TCPOLEN_SACKPERM_ALIGNED 4 |
| 184 | #define TCPOLEN_SACK_BASE 2 |
| 185 | #define TCPOLEN_SACK_BASE_ALIGNED 4 |
| 186 | #define TCPOLEN_SACK_PERBLOCK 8 |
| 187 | #define TCPOLEN_MD5SIG_ALIGNED 20 |
| 188 | #define TCPOLEN_MSS_ALIGNED 4 |
| 189 | |
| 190 | /* Flags in tp->nonagle */ |
| 191 | #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */ |
| 192 | #define TCP_NAGLE_CORK 2 /* Socket is corked */ |
| 193 | #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */ |
| 194 | |
| 195 | extern struct inet_timewait_death_row tcp_death_row; |
| 196 | |
| 197 | /* sysctl variables for tcp */ |
| 198 | extern int sysctl_tcp_timestamps; |
| 199 | extern int sysctl_tcp_window_scaling; |
| 200 | extern int sysctl_tcp_sack; |
| 201 | extern int sysctl_tcp_fin_timeout; |
| 202 | extern int sysctl_tcp_keepalive_time; |
| 203 | extern int sysctl_tcp_keepalive_probes; |
| 204 | extern int sysctl_tcp_keepalive_intvl; |
| 205 | extern int sysctl_tcp_syn_retries; |
| 206 | extern int sysctl_tcp_synack_retries; |
| 207 | extern int sysctl_tcp_retries1; |
| 208 | extern int sysctl_tcp_retries2; |
| 209 | extern int sysctl_tcp_orphan_retries; |
| 210 | extern int sysctl_tcp_syncookies; |
| 211 | extern int sysctl_tcp_retrans_collapse; |
| 212 | extern int sysctl_tcp_stdurg; |
| 213 | extern int sysctl_tcp_rfc1337; |
| 214 | extern int sysctl_tcp_abort_on_overflow; |
| 215 | extern int sysctl_tcp_max_orphans; |
| 216 | extern int sysctl_tcp_fack; |
| 217 | extern int sysctl_tcp_reordering; |
| 218 | extern int sysctl_tcp_ecn; |
| 219 | extern int sysctl_tcp_dsack; |
| 220 | extern int sysctl_tcp_mem[3]; |
| 221 | extern int sysctl_tcp_wmem[3]; |
| 222 | extern int sysctl_tcp_rmem[3]; |
| 223 | extern int sysctl_tcp_app_win; |
| 224 | extern int sysctl_tcp_adv_win_scale; |
| 225 | extern int sysctl_tcp_tw_reuse; |
| 226 | extern int sysctl_tcp_frto; |
| 227 | extern int sysctl_tcp_frto_response; |
| 228 | extern int sysctl_tcp_low_latency; |
| 229 | extern int sysctl_tcp_dma_copybreak; |
| 230 | extern int sysctl_tcp_nometrics_save; |
| 231 | extern int sysctl_tcp_moderate_rcvbuf; |
| 232 | extern int sysctl_tcp_tso_win_divisor; |
| 233 | extern int sysctl_tcp_abc; |
| 234 | extern int sysctl_tcp_mtu_probing; |
| 235 | extern int sysctl_tcp_base_mss; |
| 236 | extern int sysctl_tcp_workaround_signed_windows; |
| 237 | extern int sysctl_tcp_slow_start_after_idle; |
| 238 | extern int sysctl_tcp_max_ssthresh; |
| 239 | |
| 240 | extern atomic_t tcp_memory_allocated; |
| 241 | extern atomic_t tcp_sockets_allocated; |
| 242 | extern int tcp_memory_pressure; |
| 243 | |
| 244 | /* |
| 245 | * The next routines deal with comparing 32 bit unsigned ints |
| 246 | * and worry about wraparound (automatic with unsigned arithmetic). |
| 247 | */ |
| 248 | |
| 249 | static inline int before(__u32 seq1, __u32 seq2) |
| 250 | { |
| 251 | return (__s32)(seq1-seq2) < 0; |
| 252 | } |
| 253 | #define after(seq2, seq1) before(seq1, seq2) |
| 254 | |
| 255 | /* is s2<=s1<=s3 ? */ |
| 256 | static inline int between(__u32 seq1, __u32 seq2, __u32 seq3) |
| 257 | { |
| 258 | return seq3 - seq2 >= seq1 - seq2; |
| 259 | } |
| 260 | |
| 261 | static inline int tcp_too_many_orphans(struct sock *sk, int num) |
| 262 | { |
| 263 | return (num > sysctl_tcp_max_orphans) || |
| 264 | (sk->sk_wmem_queued > SOCK_MIN_SNDBUF && |
| 265 | atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]); |
| 266 | } |
| 267 | |
| 268 | extern struct proto tcp_prot; |
| 269 | |
| 270 | #define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field) |
| 271 | #define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field) |
| 272 | #define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field) |
| 273 | #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val) |
| 274 | |
| 275 | extern void tcp_v4_err(struct sk_buff *skb, u32); |
| 276 | |
| 277 | extern void tcp_shutdown (struct sock *sk, int how); |
| 278 | |
| 279 | extern int tcp_v4_rcv(struct sk_buff *skb); |
| 280 | |
| 281 | extern int tcp_v4_remember_stamp(struct sock *sk); |
| 282 | |
| 283 | extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw); |
| 284 | |
| 285 | extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock, |
| 286 | struct msghdr *msg, size_t size); |
| 287 | extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags); |
| 288 | |
| 289 | extern int tcp_ioctl(struct sock *sk, |
| 290 | int cmd, |
| 291 | unsigned long arg); |
| 292 | |
| 293 | extern int tcp_rcv_state_process(struct sock *sk, |
| 294 | struct sk_buff *skb, |
| 295 | struct tcphdr *th, |
| 296 | unsigned len); |
| 297 | |
| 298 | extern int tcp_rcv_established(struct sock *sk, |
| 299 | struct sk_buff *skb, |
| 300 | struct tcphdr *th, |
| 301 | unsigned len); |
| 302 | |
| 303 | extern void tcp_rcv_space_adjust(struct sock *sk); |
| 304 | |
| 305 | extern void tcp_cleanup_rbuf(struct sock *sk, int copied); |
| 306 | |
| 307 | extern int tcp_twsk_unique(struct sock *sk, |
| 308 | struct sock *sktw, void *twp); |
| 309 | |
| 310 | extern void tcp_twsk_destructor(struct sock *sk); |
| 311 | |
| 312 | extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos, |
| 313 | struct pipe_inode_info *pipe, size_t len, unsigned int flags); |
| 314 | |
| 315 | static inline void tcp_dec_quickack_mode(struct sock *sk, |
| 316 | const unsigned int pkts) |
| 317 | { |
| 318 | struct inet_connection_sock *icsk = inet_csk(sk); |
| 319 | |
| 320 | if (icsk->icsk_ack.quick) { |
| 321 | if (pkts >= icsk->icsk_ack.quick) { |
| 322 | icsk->icsk_ack.quick = 0; |
| 323 | /* Leaving quickack mode we deflate ATO. */ |
| 324 | icsk->icsk_ack.ato = TCP_ATO_MIN; |
| 325 | } else |
| 326 | icsk->icsk_ack.quick -= pkts; |
| 327 | } |
| 328 | } |
| 329 | |
| 330 | extern void tcp_enter_quickack_mode(struct sock *sk); |
| 331 | |
| 332 | static inline void tcp_clear_options(struct tcp_options_received *rx_opt) |
| 333 | { |
| 334 | rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0; |
| 335 | } |
| 336 | |
| 337 | #define TCP_ECN_OK 1 |
| 338 | #define TCP_ECN_QUEUE_CWR 2 |
| 339 | #define TCP_ECN_DEMAND_CWR 4 |
| 340 | |
| 341 | static __inline__ void |
| 342 | TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th) |
| 343 | { |
| 344 | if (sysctl_tcp_ecn && th->ece && th->cwr) |
| 345 | inet_rsk(req)->ecn_ok = 1; |
| 346 | } |
| 347 | |
| 348 | enum tcp_tw_status |
| 349 | { |
| 350 | TCP_TW_SUCCESS = 0, |
| 351 | TCP_TW_RST = 1, |
| 352 | TCP_TW_ACK = 2, |
| 353 | TCP_TW_SYN = 3 |
| 354 | }; |
| 355 | |
| 356 | |
| 357 | extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw, |
| 358 | struct sk_buff *skb, |
| 359 | const struct tcphdr *th); |
| 360 | |
| 361 | extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb, |
| 362 | struct request_sock *req, |
| 363 | struct request_sock **prev); |
| 364 | extern int tcp_child_process(struct sock *parent, |
| 365 | struct sock *child, |
| 366 | struct sk_buff *skb); |
| 367 | extern int tcp_use_frto(struct sock *sk); |
| 368 | extern void tcp_enter_frto(struct sock *sk); |
| 369 | extern void tcp_enter_loss(struct sock *sk, int how); |
| 370 | extern void tcp_clear_retrans(struct tcp_sock *tp); |
| 371 | extern void tcp_update_metrics(struct sock *sk); |
| 372 | |
| 373 | extern void tcp_close(struct sock *sk, |
| 374 | long timeout); |
| 375 | extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait); |
| 376 | |
| 377 | extern int tcp_getsockopt(struct sock *sk, int level, |
| 378 | int optname, |
| 379 | char __user *optval, |
| 380 | int __user *optlen); |
| 381 | extern int tcp_setsockopt(struct sock *sk, int level, |
| 382 | int optname, char __user *optval, |
| 383 | int optlen); |
| 384 | extern int compat_tcp_getsockopt(struct sock *sk, |
| 385 | int level, int optname, |
| 386 | char __user *optval, int __user *optlen); |
| 387 | extern int compat_tcp_setsockopt(struct sock *sk, |
| 388 | int level, int optname, |
| 389 | char __user *optval, int optlen); |
| 390 | extern void tcp_set_keepalive(struct sock *sk, int val); |
| 391 | extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, |
| 392 | struct msghdr *msg, |
| 393 | size_t len, int nonblock, |
| 394 | int flags, int *addr_len); |
| 395 | |
| 396 | extern void tcp_parse_options(struct sk_buff *skb, |
| 397 | struct tcp_options_received *opt_rx, |
| 398 | int estab); |
| 399 | |
| 400 | extern u8 *tcp_parse_md5sig_option(struct tcphdr *th); |
| 401 | |
| 402 | /* |
| 403 | * TCP v4 functions exported for the inet6 API |
| 404 | */ |
| 405 | |
| 406 | extern void tcp_v4_send_check(struct sock *sk, int len, |
| 407 | struct sk_buff *skb); |
| 408 | |
| 409 | extern int tcp_v4_conn_request(struct sock *sk, |
| 410 | struct sk_buff *skb); |
| 411 | |
| 412 | extern struct sock * tcp_create_openreq_child(struct sock *sk, |
| 413 | struct request_sock *req, |
| 414 | struct sk_buff *skb); |
| 415 | |
| 416 | extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, |
| 417 | struct sk_buff *skb, |
| 418 | struct request_sock *req, |
| 419 | struct dst_entry *dst); |
| 420 | |
| 421 | extern int tcp_v4_do_rcv(struct sock *sk, |
| 422 | struct sk_buff *skb); |
| 423 | |
| 424 | extern int tcp_v4_connect(struct sock *sk, |
| 425 | struct sockaddr *uaddr, |
| 426 | int addr_len); |
| 427 | |
| 428 | extern int tcp_connect(struct sock *sk); |
| 429 | |
| 430 | extern struct sk_buff * tcp_make_synack(struct sock *sk, |
| 431 | struct dst_entry *dst, |
| 432 | struct request_sock *req); |
| 433 | |
| 434 | extern int tcp_disconnect(struct sock *sk, int flags); |
| 435 | |
| 436 | |
| 437 | /* From syncookies.c */ |
| 438 | extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS]; |
| 439 | extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb, |
| 440 | struct ip_options *opt); |
| 441 | extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, |
| 442 | __u16 *mss); |
| 443 | |
| 444 | extern __u32 cookie_init_timestamp(struct request_sock *req); |
| 445 | extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt); |
| 446 | |
| 447 | /* From net/ipv6/syncookies.c */ |
| 448 | extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb); |
| 449 | extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb, |
| 450 | __u16 *mss); |
| 451 | |
| 452 | /* tcp_output.c */ |
| 453 | |
| 454 | extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss, |
| 455 | int nonagle); |
| 456 | extern int tcp_may_send_now(struct sock *sk); |
| 457 | extern int tcp_retransmit_skb(struct sock *, struct sk_buff *); |
| 458 | extern void tcp_xmit_retransmit_queue(struct sock *); |
| 459 | extern void tcp_simple_retransmit(struct sock *); |
| 460 | extern int tcp_trim_head(struct sock *, struct sk_buff *, u32); |
| 461 | extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int); |
| 462 | |
| 463 | extern void tcp_send_probe0(struct sock *); |
| 464 | extern void tcp_send_partial(struct sock *); |
| 465 | extern int tcp_write_wakeup(struct sock *); |
| 466 | extern void tcp_send_fin(struct sock *sk); |
| 467 | extern void tcp_send_active_reset(struct sock *sk, gfp_t priority); |
| 468 | extern int tcp_send_synack(struct sock *); |
| 469 | extern void tcp_push_one(struct sock *, unsigned int mss_now); |
| 470 | extern void tcp_send_ack(struct sock *sk); |
| 471 | extern void tcp_send_delayed_ack(struct sock *sk); |
| 472 | |
| 473 | /* tcp_input.c */ |
| 474 | extern void tcp_cwnd_application_limited(struct sock *sk); |
| 475 | extern void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp, |
| 476 | struct sk_buff *skb); |
| 477 | |
| 478 | /* tcp_timer.c */ |
| 479 | extern void tcp_init_xmit_timers(struct sock *); |
| 480 | static inline void tcp_clear_xmit_timers(struct sock *sk) |
| 481 | { |
| 482 | inet_csk_clear_xmit_timers(sk); |
| 483 | } |
| 484 | |
| 485 | extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu); |
| 486 | extern unsigned int tcp_current_mss(struct sock *sk, int large); |
| 487 | |
| 488 | /* tcp.c */ |
| 489 | extern void tcp_get_info(struct sock *, struct tcp_info *); |
| 490 | |
| 491 | /* Read 'sendfile()'-style from a TCP socket */ |
| 492 | typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, |
| 493 | unsigned int, size_t); |
| 494 | extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc, |
| 495 | sk_read_actor_t recv_actor); |
| 496 | |
| 497 | extern void tcp_initialize_rcv_mss(struct sock *sk); |
| 498 | |
| 499 | extern int tcp_mtu_to_mss(struct sock *sk, int pmtu); |
| 500 | extern int tcp_mss_to_mtu(struct sock *sk, int mss); |
| 501 | extern void tcp_mtup_init(struct sock *sk); |
| 502 | |
| 503 | static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd) |
| 504 | { |
| 505 | tp->pred_flags = htonl((tp->tcp_header_len << 26) | |
| 506 | ntohl(TCP_FLAG_ACK) | |
| 507 | snd_wnd); |
| 508 | } |
| 509 | |
| 510 | static inline void tcp_fast_path_on(struct tcp_sock *tp) |
| 511 | { |
| 512 | __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale); |
| 513 | } |
| 514 | |
| 515 | static inline void tcp_fast_path_check(struct sock *sk) |
| 516 | { |
| 517 | struct tcp_sock *tp = tcp_sk(sk); |
| 518 | |
| 519 | if (skb_queue_empty(&tp->out_of_order_queue) && |
| 520 | tp->rcv_wnd && |
| 521 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf && |
| 522 | !tp->urg_data) |
| 523 | tcp_fast_path_on(tp); |
| 524 | } |
| 525 | |
| 526 | /* Compute the actual receive window we are currently advertising. |
| 527 | * Rcv_nxt can be after the window if our peer push more data |
| 528 | * than the offered window. |
| 529 | */ |
| 530 | static inline u32 tcp_receive_window(const struct tcp_sock *tp) |
| 531 | { |
| 532 | s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt; |
| 533 | |
| 534 | if (win < 0) |
| 535 | win = 0; |
| 536 | return (u32) win; |
| 537 | } |
| 538 | |
| 539 | /* Choose a new window, without checks for shrinking, and without |
| 540 | * scaling applied to the result. The caller does these things |
| 541 | * if necessary. This is a "raw" window selection. |
| 542 | */ |
| 543 | extern u32 __tcp_select_window(struct sock *sk); |
| 544 | |
| 545 | /* TCP timestamps are only 32-bits, this causes a slight |
| 546 | * complication on 64-bit systems since we store a snapshot |
| 547 | * of jiffies in the buffer control blocks below. We decided |
| 548 | * to use only the low 32-bits of jiffies and hide the ugly |
| 549 | * casts with the following macro. |
| 550 | */ |
| 551 | #define tcp_time_stamp ((__u32)(jiffies)) |
| 552 | |
| 553 | /* This is what the send packet queuing engine uses to pass |
| 554 | * TCP per-packet control information to the transmission |
| 555 | * code. We also store the host-order sequence numbers in |
| 556 | * here too. This is 36 bytes on 32-bit architectures, |
| 557 | * 40 bytes on 64-bit machines, if this grows please adjust |
| 558 | * skbuff.h:skbuff->cb[xxx] size appropriately. |
| 559 | */ |
| 560 | struct tcp_skb_cb { |
| 561 | union { |
| 562 | struct inet_skb_parm h4; |
| 563 | #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE) |
| 564 | struct inet6_skb_parm h6; |
| 565 | #endif |
| 566 | } header; /* For incoming frames */ |
| 567 | __u32 seq; /* Starting sequence number */ |
| 568 | __u32 end_seq; /* SEQ + FIN + SYN + datalen */ |
| 569 | __u32 when; /* used to compute rtt's */ |
| 570 | __u8 flags; /* TCP header flags. */ |
| 571 | |
| 572 | /* NOTE: These must match up to the flags byte in a |
| 573 | * real TCP header. |
| 574 | */ |
| 575 | #define TCPCB_FLAG_FIN 0x01 |
| 576 | #define TCPCB_FLAG_SYN 0x02 |
| 577 | #define TCPCB_FLAG_RST 0x04 |
| 578 | #define TCPCB_FLAG_PSH 0x08 |
| 579 | #define TCPCB_FLAG_ACK 0x10 |
| 580 | #define TCPCB_FLAG_URG 0x20 |
| 581 | #define TCPCB_FLAG_ECE 0x40 |
| 582 | #define TCPCB_FLAG_CWR 0x80 |
| 583 | |
| 584 | __u8 sacked; /* State flags for SACK/FACK. */ |
| 585 | #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */ |
| 586 | #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */ |
| 587 | #define TCPCB_LOST 0x04 /* SKB is lost */ |
| 588 | #define TCPCB_TAGBITS 0x07 /* All tag bits */ |
| 589 | |
| 590 | #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */ |
| 591 | #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS) |
| 592 | |
| 593 | __u16 urg_ptr; /* Valid w/URG flags is set. */ |
| 594 | __u32 ack_seq; /* Sequence number ACK'd */ |
| 595 | }; |
| 596 | |
| 597 | #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0])) |
| 598 | |
| 599 | /* Due to TSO, an SKB can be composed of multiple actual |
| 600 | * packets. To keep these tracked properly, we use this. |
| 601 | */ |
| 602 | static inline int tcp_skb_pcount(const struct sk_buff *skb) |
| 603 | { |
| 604 | return skb_shinfo(skb)->gso_segs; |
| 605 | } |
| 606 | |
| 607 | /* This is valid iff tcp_skb_pcount() > 1. */ |
| 608 | static inline int tcp_skb_mss(const struct sk_buff *skb) |
| 609 | { |
| 610 | return skb_shinfo(skb)->gso_size; |
| 611 | } |
| 612 | |
| 613 | static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr) |
| 614 | { |
| 615 | if (*count) { |
| 616 | *count -= decr; |
| 617 | if ((int)*count < 0) |
| 618 | *count = 0; |
| 619 | } |
| 620 | } |
| 621 | |
| 622 | static inline void tcp_dec_pcount_approx(__u32 *count, |
| 623 | const struct sk_buff *skb) |
| 624 | { |
| 625 | tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb)); |
| 626 | } |
| 627 | |
| 628 | /* Events passed to congestion control interface */ |
| 629 | enum tcp_ca_event { |
| 630 | CA_EVENT_TX_START, /* first transmit when no packets in flight */ |
| 631 | CA_EVENT_CWND_RESTART, /* congestion window restart */ |
| 632 | CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */ |
| 633 | CA_EVENT_FRTO, /* fast recovery timeout */ |
| 634 | CA_EVENT_LOSS, /* loss timeout */ |
| 635 | CA_EVENT_FAST_ACK, /* in sequence ack */ |
| 636 | CA_EVENT_SLOW_ACK, /* other ack */ |
| 637 | }; |
| 638 | |
| 639 | /* |
| 640 | * Interface for adding new TCP congestion control handlers |
| 641 | */ |
| 642 | #define TCP_CA_NAME_MAX 16 |
| 643 | #define TCP_CA_MAX 128 |
| 644 | #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX) |
| 645 | |
| 646 | #define TCP_CONG_NON_RESTRICTED 0x1 |
| 647 | #define TCP_CONG_RTT_STAMP 0x2 |
| 648 | |
| 649 | struct tcp_congestion_ops { |
| 650 | struct list_head list; |
| 651 | unsigned long flags; |
| 652 | |
| 653 | /* initialize private data (optional) */ |
| 654 | void (*init)(struct sock *sk); |
| 655 | /* cleanup private data (optional) */ |
| 656 | void (*release)(struct sock *sk); |
| 657 | |
| 658 | /* return slow start threshold (required) */ |
| 659 | u32 (*ssthresh)(struct sock *sk); |
| 660 | /* lower bound for congestion window (optional) */ |
| 661 | u32 (*min_cwnd)(const struct sock *sk); |
| 662 | /* do new cwnd calculation (required) */ |
| 663 | void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight); |
| 664 | /* call before changing ca_state (optional) */ |
| 665 | void (*set_state)(struct sock *sk, u8 new_state); |
| 666 | /* call when cwnd event occurs (optional) */ |
| 667 | void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev); |
| 668 | /* new value of cwnd after loss (optional) */ |
| 669 | u32 (*undo_cwnd)(struct sock *sk); |
| 670 | /* hook for packet ack accounting (optional) */ |
| 671 | void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us); |
| 672 | /* get info for inet_diag (optional) */ |
| 673 | void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb); |
| 674 | |
| 675 | char name[TCP_CA_NAME_MAX]; |
| 676 | struct module *owner; |
| 677 | }; |
| 678 | |
| 679 | extern int tcp_register_congestion_control(struct tcp_congestion_ops *type); |
| 680 | extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type); |
| 681 | |
| 682 | extern void tcp_init_congestion_control(struct sock *sk); |
| 683 | extern void tcp_cleanup_congestion_control(struct sock *sk); |
| 684 | extern int tcp_set_default_congestion_control(const char *name); |
| 685 | extern void tcp_get_default_congestion_control(char *name); |
| 686 | extern void tcp_get_available_congestion_control(char *buf, size_t len); |
| 687 | extern void tcp_get_allowed_congestion_control(char *buf, size_t len); |
| 688 | extern int tcp_set_allowed_congestion_control(char *allowed); |
| 689 | extern int tcp_set_congestion_control(struct sock *sk, const char *name); |
| 690 | extern void tcp_slow_start(struct tcp_sock *tp); |
| 691 | |
| 692 | extern struct tcp_congestion_ops tcp_init_congestion_ops; |
| 693 | extern u32 tcp_reno_ssthresh(struct sock *sk); |
| 694 | extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight); |
| 695 | extern u32 tcp_reno_min_cwnd(const struct sock *sk); |
| 696 | extern struct tcp_congestion_ops tcp_reno; |
| 697 | |
| 698 | static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state) |
| 699 | { |
| 700 | struct inet_connection_sock *icsk = inet_csk(sk); |
| 701 | |
| 702 | if (icsk->icsk_ca_ops->set_state) |
| 703 | icsk->icsk_ca_ops->set_state(sk, ca_state); |
| 704 | icsk->icsk_ca_state = ca_state; |
| 705 | } |
| 706 | |
| 707 | static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event) |
| 708 | { |
| 709 | const struct inet_connection_sock *icsk = inet_csk(sk); |
| 710 | |
| 711 | if (icsk->icsk_ca_ops->cwnd_event) |
| 712 | icsk->icsk_ca_ops->cwnd_event(sk, event); |
| 713 | } |
| 714 | |
| 715 | /* These functions determine how the current flow behaves in respect of SACK |
| 716 | * handling. SACK is negotiated with the peer, and therefore it can vary |
| 717 | * between different flows. |
| 718 | * |
| 719 | * tcp_is_sack - SACK enabled |
| 720 | * tcp_is_reno - No SACK |
| 721 | * tcp_is_fack - FACK enabled, implies SACK enabled |
| 722 | */ |
| 723 | static inline int tcp_is_sack(const struct tcp_sock *tp) |
| 724 | { |
| 725 | return tp->rx_opt.sack_ok; |
| 726 | } |
| 727 | |
| 728 | static inline int tcp_is_reno(const struct tcp_sock *tp) |
| 729 | { |
| 730 | return !tcp_is_sack(tp); |
| 731 | } |
| 732 | |
| 733 | static inline int tcp_is_fack(const struct tcp_sock *tp) |
| 734 | { |
| 735 | return tp->rx_opt.sack_ok & 2; |
| 736 | } |
| 737 | |
| 738 | static inline void tcp_enable_fack(struct tcp_sock *tp) |
| 739 | { |
| 740 | tp->rx_opt.sack_ok |= 2; |
| 741 | } |
| 742 | |
| 743 | static inline unsigned int tcp_left_out(const struct tcp_sock *tp) |
| 744 | { |
| 745 | return tp->sacked_out + tp->lost_out; |
| 746 | } |
| 747 | |
| 748 | /* This determines how many packets are "in the network" to the best |
| 749 | * of our knowledge. In many cases it is conservative, but where |
| 750 | * detailed information is available from the receiver (via SACK |
| 751 | * blocks etc.) we can make more aggressive calculations. |
| 752 | * |
| 753 | * Use this for decisions involving congestion control, use just |
| 754 | * tp->packets_out to determine if the send queue is empty or not. |
| 755 | * |
| 756 | * Read this equation as: |
| 757 | * |
| 758 | * "Packets sent once on transmission queue" MINUS |
| 759 | * "Packets left network, but not honestly ACKed yet" PLUS |
| 760 | * "Packets fast retransmitted" |
| 761 | */ |
| 762 | static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp) |
| 763 | { |
| 764 | return tp->packets_out - tcp_left_out(tp) + tp->retrans_out; |
| 765 | } |
| 766 | |
| 767 | extern int tcp_limit_reno_sacked(struct tcp_sock *tp); |
| 768 | |
| 769 | /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd. |
| 770 | * The exception is rate halving phase, when cwnd is decreasing towards |
| 771 | * ssthresh. |
| 772 | */ |
| 773 | static inline __u32 tcp_current_ssthresh(const struct sock *sk) |
| 774 | { |
| 775 | const struct tcp_sock *tp = tcp_sk(sk); |
| 776 | if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery)) |
| 777 | return tp->snd_ssthresh; |
| 778 | else |
| 779 | return max(tp->snd_ssthresh, |
| 780 | ((tp->snd_cwnd >> 1) + |
| 781 | (tp->snd_cwnd >> 2))); |
| 782 | } |
| 783 | |
| 784 | /* Use define here intentionally to get WARN_ON location shown at the caller */ |
| 785 | #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out) |
| 786 | |
| 787 | extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh); |
| 788 | extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst); |
| 789 | |
| 790 | /* Slow start with delack produces 3 packets of burst, so that |
| 791 | * it is safe "de facto". This will be the default - same as |
| 792 | * the default reordering threshold - but if reordering increases, |
| 793 | * we must be able to allow cwnd to burst at least this much in order |
| 794 | * to not pull it back when holes are filled. |
| 795 | */ |
| 796 | static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp) |
| 797 | { |
| 798 | return tp->reordering; |
| 799 | } |
| 800 | |
| 801 | /* Returns end sequence number of the receiver's advertised window */ |
| 802 | static inline u32 tcp_wnd_end(const struct tcp_sock *tp) |
| 803 | { |
| 804 | return tp->snd_una + tp->snd_wnd; |
| 805 | } |
| 806 | extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight); |
| 807 | |
| 808 | static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss, |
| 809 | const struct sk_buff *skb) |
| 810 | { |
| 811 | if (skb->len < mss) |
| 812 | tp->snd_sml = TCP_SKB_CB(skb)->end_seq; |
| 813 | } |
| 814 | |
| 815 | static inline void tcp_check_probe_timer(struct sock *sk) |
| 816 | { |
| 817 | struct tcp_sock *tp = tcp_sk(sk); |
| 818 | const struct inet_connection_sock *icsk = inet_csk(sk); |
| 819 | |
| 820 | if (!tp->packets_out && !icsk->icsk_pending) |
| 821 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0, |
| 822 | icsk->icsk_rto, TCP_RTO_MAX); |
| 823 | } |
| 824 | |
| 825 | static inline void tcp_push_pending_frames(struct sock *sk) |
| 826 | { |
| 827 | struct tcp_sock *tp = tcp_sk(sk); |
| 828 | |
| 829 | __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle); |
| 830 | } |
| 831 | |
| 832 | static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq) |
| 833 | { |
| 834 | tp->snd_wl1 = seq; |
| 835 | } |
| 836 | |
| 837 | static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq) |
| 838 | { |
| 839 | tp->snd_wl1 = seq; |
| 840 | } |
| 841 | |
| 842 | /* |
| 843 | * Calculate(/check) TCP checksum |
| 844 | */ |
| 845 | static inline __sum16 tcp_v4_check(int len, __be32 saddr, |
| 846 | __be32 daddr, __wsum base) |
| 847 | { |
| 848 | return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base); |
| 849 | } |
| 850 | |
| 851 | static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb) |
| 852 | { |
| 853 | return __skb_checksum_complete(skb); |
| 854 | } |
| 855 | |
| 856 | static inline int tcp_checksum_complete(struct sk_buff *skb) |
| 857 | { |
| 858 | return !skb_csum_unnecessary(skb) && |
| 859 | __tcp_checksum_complete(skb); |
| 860 | } |
| 861 | |
| 862 | /* Prequeue for VJ style copy to user, combined with checksumming. */ |
| 863 | |
| 864 | static inline void tcp_prequeue_init(struct tcp_sock *tp) |
| 865 | { |
| 866 | tp->ucopy.task = NULL; |
| 867 | tp->ucopy.len = 0; |
| 868 | tp->ucopy.memory = 0; |
| 869 | skb_queue_head_init(&tp->ucopy.prequeue); |
| 870 | #ifdef CONFIG_NET_DMA |
| 871 | tp->ucopy.dma_chan = NULL; |
| 872 | tp->ucopy.wakeup = 0; |
| 873 | tp->ucopy.pinned_list = NULL; |
| 874 | tp->ucopy.dma_cookie = 0; |
| 875 | #endif |
| 876 | } |
| 877 | |
| 878 | /* Packet is added to VJ-style prequeue for processing in process |
| 879 | * context, if a reader task is waiting. Apparently, this exciting |
| 880 | * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93) |
| 881 | * failed somewhere. Latency? Burstiness? Well, at least now we will |
| 882 | * see, why it failed. 8)8) --ANK |
| 883 | * |
| 884 | * NOTE: is this not too big to inline? |
| 885 | */ |
| 886 | static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb) |
| 887 | { |
| 888 | struct tcp_sock *tp = tcp_sk(sk); |
| 889 | |
| 890 | if (!sysctl_tcp_low_latency && tp->ucopy.task) { |
| 891 | __skb_queue_tail(&tp->ucopy.prequeue, skb); |
| 892 | tp->ucopy.memory += skb->truesize; |
| 893 | if (tp->ucopy.memory > sk->sk_rcvbuf) { |
| 894 | struct sk_buff *skb1; |
| 895 | |
| 896 | BUG_ON(sock_owned_by_user(sk)); |
| 897 | |
| 898 | while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) { |
| 899 | sk->sk_backlog_rcv(sk, skb1); |
| 900 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPPREQUEUEDROPPED); |
| 901 | } |
| 902 | |
| 903 | tp->ucopy.memory = 0; |
| 904 | } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) { |
| 905 | wake_up_interruptible(sk->sk_sleep); |
| 906 | if (!inet_csk_ack_scheduled(sk)) |
| 907 | inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK, |
| 908 | (3 * TCP_RTO_MIN) / 4, |
| 909 | TCP_RTO_MAX); |
| 910 | } |
| 911 | return 1; |
| 912 | } |
| 913 | return 0; |
| 914 | } |
| 915 | |
| 916 | |
| 917 | #undef STATE_TRACE |
| 918 | |
| 919 | #ifdef STATE_TRACE |
| 920 | static const char *statename[]={ |
| 921 | "Unused","Established","Syn Sent","Syn Recv", |
| 922 | "Fin Wait 1","Fin Wait 2","Time Wait", "Close", |
| 923 | "Close Wait","Last ACK","Listen","Closing" |
| 924 | }; |
| 925 | #endif |
| 926 | extern void tcp_set_state(struct sock *sk, int state); |
| 927 | |
| 928 | extern void tcp_done(struct sock *sk); |
| 929 | |
| 930 | static inline void tcp_sack_reset(struct tcp_options_received *rx_opt) |
| 931 | { |
| 932 | rx_opt->dsack = 0; |
| 933 | rx_opt->eff_sacks = 0; |
| 934 | rx_opt->num_sacks = 0; |
| 935 | } |
| 936 | |
| 937 | /* Determine a window scaling and initial window to offer. */ |
| 938 | extern void tcp_select_initial_window(int __space, __u32 mss, |
| 939 | __u32 *rcv_wnd, __u32 *window_clamp, |
| 940 | int wscale_ok, __u8 *rcv_wscale); |
| 941 | |
| 942 | static inline int tcp_win_from_space(int space) |
| 943 | { |
| 944 | return sysctl_tcp_adv_win_scale<=0 ? |
| 945 | (space>>(-sysctl_tcp_adv_win_scale)) : |
| 946 | space - (space>>sysctl_tcp_adv_win_scale); |
| 947 | } |
| 948 | |
| 949 | /* Note: caller must be prepared to deal with negative returns */ |
| 950 | static inline int tcp_space(const struct sock *sk) |
| 951 | { |
| 952 | return tcp_win_from_space(sk->sk_rcvbuf - |
| 953 | atomic_read(&sk->sk_rmem_alloc)); |
| 954 | } |
| 955 | |
| 956 | static inline int tcp_full_space(const struct sock *sk) |
| 957 | { |
| 958 | return tcp_win_from_space(sk->sk_rcvbuf); |
| 959 | } |
| 960 | |
| 961 | static inline void tcp_openreq_init(struct request_sock *req, |
| 962 | struct tcp_options_received *rx_opt, |
| 963 | struct sk_buff *skb) |
| 964 | { |
| 965 | struct inet_request_sock *ireq = inet_rsk(req); |
| 966 | |
| 967 | req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */ |
| 968 | req->cookie_ts = 0; |
| 969 | tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq; |
| 970 | req->mss = rx_opt->mss_clamp; |
| 971 | req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0; |
| 972 | ireq->tstamp_ok = rx_opt->tstamp_ok; |
| 973 | ireq->sack_ok = rx_opt->sack_ok; |
| 974 | ireq->snd_wscale = rx_opt->snd_wscale; |
| 975 | ireq->wscale_ok = rx_opt->wscale_ok; |
| 976 | ireq->acked = 0; |
| 977 | ireq->ecn_ok = 0; |
| 978 | ireq->rmt_port = tcp_hdr(skb)->source; |
| 979 | } |
| 980 | |
| 981 | extern void tcp_enter_memory_pressure(struct sock *sk); |
| 982 | |
| 983 | static inline int keepalive_intvl_when(const struct tcp_sock *tp) |
| 984 | { |
| 985 | return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl; |
| 986 | } |
| 987 | |
| 988 | static inline int keepalive_time_when(const struct tcp_sock *tp) |
| 989 | { |
| 990 | return tp->keepalive_time ? : sysctl_tcp_keepalive_time; |
| 991 | } |
| 992 | |
| 993 | static inline int tcp_fin_time(const struct sock *sk) |
| 994 | { |
| 995 | int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout; |
| 996 | const int rto = inet_csk(sk)->icsk_rto; |
| 997 | |
| 998 | if (fin_timeout < (rto << 2) - (rto >> 1)) |
| 999 | fin_timeout = (rto << 2) - (rto >> 1); |
| 1000 | |
| 1001 | return fin_timeout; |
| 1002 | } |
| 1003 | |
| 1004 | static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst) |
| 1005 | { |
| 1006 | if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0) |
| 1007 | return 0; |
| 1008 | if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS) |
| 1009 | return 0; |
| 1010 | |
| 1011 | /* RST segments are not recommended to carry timestamp, |
| 1012 | and, if they do, it is recommended to ignore PAWS because |
| 1013 | "their cleanup function should take precedence over timestamps." |
| 1014 | Certainly, it is mistake. It is necessary to understand the reasons |
| 1015 | of this constraint to relax it: if peer reboots, clock may go |
| 1016 | out-of-sync and half-open connections will not be reset. |
| 1017 | Actually, the problem would be not existing if all |
| 1018 | the implementations followed draft about maintaining clock |
| 1019 | via reboots. Linux-2.2 DOES NOT! |
| 1020 | |
| 1021 | However, we can relax time bounds for RST segments to MSL. |
| 1022 | */ |
| 1023 | if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL) |
| 1024 | return 0; |
| 1025 | return 1; |
| 1026 | } |
| 1027 | |
| 1028 | #define TCP_CHECK_TIMER(sk) do { } while (0) |
| 1029 | |
| 1030 | static inline void tcp_mib_init(struct net *net) |
| 1031 | { |
| 1032 | /* See RFC 2012 */ |
| 1033 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1); |
| 1034 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ); |
| 1035 | TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ); |
| 1036 | TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1); |
| 1037 | } |
| 1038 | |
| 1039 | /* from STCP */ |
| 1040 | static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp) |
| 1041 | { |
| 1042 | tp->lost_skb_hint = NULL; |
| 1043 | tp->scoreboard_skb_hint = NULL; |
| 1044 | tp->retransmit_skb_hint = NULL; |
| 1045 | } |
| 1046 | |
| 1047 | /* MD5 Signature */ |
| 1048 | struct crypto_hash; |
| 1049 | |
| 1050 | /* - key database */ |
| 1051 | struct tcp_md5sig_key { |
| 1052 | u8 *key; |
| 1053 | u8 keylen; |
| 1054 | }; |
| 1055 | |
| 1056 | struct tcp4_md5sig_key { |
| 1057 | struct tcp_md5sig_key base; |
| 1058 | __be32 addr; |
| 1059 | }; |
| 1060 | |
| 1061 | struct tcp6_md5sig_key { |
| 1062 | struct tcp_md5sig_key base; |
| 1063 | #if 0 |
| 1064 | u32 scope_id; /* XXX */ |
| 1065 | #endif |
| 1066 | struct in6_addr addr; |
| 1067 | }; |
| 1068 | |
| 1069 | /* - sock block */ |
| 1070 | struct tcp_md5sig_info { |
| 1071 | struct tcp4_md5sig_key *keys4; |
| 1072 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) |
| 1073 | struct tcp6_md5sig_key *keys6; |
| 1074 | u32 entries6; |
| 1075 | u32 alloced6; |
| 1076 | #endif |
| 1077 | u32 entries4; |
| 1078 | u32 alloced4; |
| 1079 | }; |
| 1080 | |
| 1081 | /* - pseudo header */ |
| 1082 | struct tcp4_pseudohdr { |
| 1083 | __be32 saddr; |
| 1084 | __be32 daddr; |
| 1085 | __u8 pad; |
| 1086 | __u8 protocol; |
| 1087 | __be16 len; |
| 1088 | }; |
| 1089 | |
| 1090 | struct tcp6_pseudohdr { |
| 1091 | struct in6_addr saddr; |
| 1092 | struct in6_addr daddr; |
| 1093 | __be32 len; |
| 1094 | __be32 protocol; /* including padding */ |
| 1095 | }; |
| 1096 | |
| 1097 | union tcp_md5sum_block { |
| 1098 | struct tcp4_pseudohdr ip4; |
| 1099 | #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) |
| 1100 | struct tcp6_pseudohdr ip6; |
| 1101 | #endif |
| 1102 | }; |
| 1103 | |
| 1104 | /* - pool: digest algorithm, hash description and scratch buffer */ |
| 1105 | struct tcp_md5sig_pool { |
| 1106 | struct hash_desc md5_desc; |
| 1107 | union tcp_md5sum_block md5_blk; |
| 1108 | }; |
| 1109 | |
| 1110 | #define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */ |
| 1111 | |
| 1112 | /* - functions */ |
| 1113 | extern int tcp_v4_md5_hash_skb(char *md5_hash, |
| 1114 | struct tcp_md5sig_key *key, |
| 1115 | struct sock *sk, |
| 1116 | struct request_sock *req, |
| 1117 | struct sk_buff *skb); |
| 1118 | |
| 1119 | extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk, |
| 1120 | struct sock *addr_sk); |
| 1121 | |
| 1122 | extern int tcp_v4_md5_do_add(struct sock *sk, |
| 1123 | __be32 addr, |
| 1124 | u8 *newkey, |
| 1125 | u8 newkeylen); |
| 1126 | |
| 1127 | extern int tcp_v4_md5_do_del(struct sock *sk, |
| 1128 | __be32 addr); |
| 1129 | |
| 1130 | #ifdef CONFIG_TCP_MD5SIG |
| 1131 | #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_keylen ? \ |
| 1132 | &(struct tcp_md5sig_key) { \ |
| 1133 | .key = (twsk)->tw_md5_key, \ |
| 1134 | .keylen = (twsk)->tw_md5_keylen, \ |
| 1135 | } : NULL) |
| 1136 | #else |
| 1137 | #define tcp_twsk_md5_key(twsk) NULL |
| 1138 | #endif |
| 1139 | |
| 1140 | extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void); |
| 1141 | extern void tcp_free_md5sig_pool(void); |
| 1142 | |
| 1143 | extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu); |
| 1144 | extern void __tcp_put_md5sig_pool(void); |
| 1145 | extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, struct tcphdr *); |
| 1146 | extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, struct sk_buff *, |
| 1147 | unsigned header_len); |
| 1148 | extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp, |
| 1149 | struct tcp_md5sig_key *key); |
| 1150 | |
| 1151 | static inline |
| 1152 | struct tcp_md5sig_pool *tcp_get_md5sig_pool(void) |
| 1153 | { |
| 1154 | int cpu = get_cpu(); |
| 1155 | struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu); |
| 1156 | if (!ret) |
| 1157 | put_cpu(); |
| 1158 | return ret; |
| 1159 | } |
| 1160 | |
| 1161 | static inline void tcp_put_md5sig_pool(void) |
| 1162 | { |
| 1163 | __tcp_put_md5sig_pool(); |
| 1164 | put_cpu(); |
| 1165 | } |
| 1166 | |
| 1167 | /* write queue abstraction */ |
| 1168 | static inline void tcp_write_queue_purge(struct sock *sk) |
| 1169 | { |
| 1170 | struct sk_buff *skb; |
| 1171 | |
| 1172 | while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) |
| 1173 | sk_wmem_free_skb(sk, skb); |
| 1174 | sk_mem_reclaim(sk); |
| 1175 | } |
| 1176 | |
| 1177 | static inline struct sk_buff *tcp_write_queue_head(struct sock *sk) |
| 1178 | { |
| 1179 | struct sk_buff *skb = sk->sk_write_queue.next; |
| 1180 | if (skb == (struct sk_buff *) &sk->sk_write_queue) |
| 1181 | return NULL; |
| 1182 | return skb; |
| 1183 | } |
| 1184 | |
| 1185 | static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk) |
| 1186 | { |
| 1187 | struct sk_buff *skb = sk->sk_write_queue.prev; |
| 1188 | if (skb == (struct sk_buff *) &sk->sk_write_queue) |
| 1189 | return NULL; |
| 1190 | return skb; |
| 1191 | } |
| 1192 | |
| 1193 | static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb) |
| 1194 | { |
| 1195 | return skb->next; |
| 1196 | } |
| 1197 | |
| 1198 | #define tcp_for_write_queue(skb, sk) \ |
| 1199 | for (skb = (sk)->sk_write_queue.next; \ |
| 1200 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ |
| 1201 | skb = skb->next) |
| 1202 | |
| 1203 | #define tcp_for_write_queue_from(skb, sk) \ |
| 1204 | for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\ |
| 1205 | skb = skb->next) |
| 1206 | |
| 1207 | #define tcp_for_write_queue_from_safe(skb, tmp, sk) \ |
| 1208 | for (tmp = skb->next; \ |
| 1209 | (skb != (struct sk_buff *)&(sk)->sk_write_queue); \ |
| 1210 | skb = tmp, tmp = skb->next) |
| 1211 | |
| 1212 | static inline struct sk_buff *tcp_send_head(struct sock *sk) |
| 1213 | { |
| 1214 | return sk->sk_send_head; |
| 1215 | } |
| 1216 | |
| 1217 | static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb) |
| 1218 | { |
| 1219 | sk->sk_send_head = skb->next; |
| 1220 | if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue) |
| 1221 | sk->sk_send_head = NULL; |
| 1222 | } |
| 1223 | |
| 1224 | static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked) |
| 1225 | { |
| 1226 | if (sk->sk_send_head == skb_unlinked) |
| 1227 | sk->sk_send_head = NULL; |
| 1228 | } |
| 1229 | |
| 1230 | static inline void tcp_init_send_head(struct sock *sk) |
| 1231 | { |
| 1232 | sk->sk_send_head = NULL; |
| 1233 | } |
| 1234 | |
| 1235 | static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) |
| 1236 | { |
| 1237 | __skb_queue_tail(&sk->sk_write_queue, skb); |
| 1238 | } |
| 1239 | |
| 1240 | static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb) |
| 1241 | { |
| 1242 | __tcp_add_write_queue_tail(sk, skb); |
| 1243 | |
| 1244 | /* Queue it, remembering where we must start sending. */ |
| 1245 | if (sk->sk_send_head == NULL) { |
| 1246 | sk->sk_send_head = skb; |
| 1247 | |
| 1248 | if (tcp_sk(sk)->highest_sack == NULL) |
| 1249 | tcp_sk(sk)->highest_sack = skb; |
| 1250 | } |
| 1251 | } |
| 1252 | |
| 1253 | static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb) |
| 1254 | { |
| 1255 | __skb_queue_head(&sk->sk_write_queue, skb); |
| 1256 | } |
| 1257 | |
| 1258 | /* Insert buff after skb on the write queue of sk. */ |
| 1259 | static inline void tcp_insert_write_queue_after(struct sk_buff *skb, |
| 1260 | struct sk_buff *buff, |
| 1261 | struct sock *sk) |
| 1262 | { |
| 1263 | __skb_queue_after(&sk->sk_write_queue, skb, buff); |
| 1264 | } |
| 1265 | |
| 1266 | /* Insert skb between prev and next on the write queue of sk. */ |
| 1267 | static inline void tcp_insert_write_queue_before(struct sk_buff *new, |
| 1268 | struct sk_buff *skb, |
| 1269 | struct sock *sk) |
| 1270 | { |
| 1271 | __skb_insert(new, skb->prev, skb, &sk->sk_write_queue); |
| 1272 | |
| 1273 | if (sk->sk_send_head == skb) |
| 1274 | sk->sk_send_head = new; |
| 1275 | } |
| 1276 | |
| 1277 | static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk) |
| 1278 | { |
| 1279 | __skb_unlink(skb, &sk->sk_write_queue); |
| 1280 | } |
| 1281 | |
| 1282 | static inline int tcp_skb_is_last(const struct sock *sk, |
| 1283 | const struct sk_buff *skb) |
| 1284 | { |
| 1285 | return skb->next == (struct sk_buff *)&sk->sk_write_queue; |
| 1286 | } |
| 1287 | |
| 1288 | static inline int tcp_write_queue_empty(struct sock *sk) |
| 1289 | { |
| 1290 | return skb_queue_empty(&sk->sk_write_queue); |
| 1291 | } |
| 1292 | |
| 1293 | /* Start sequence of the highest skb with SACKed bit, valid only if |
| 1294 | * sacked > 0 or when the caller has ensured validity by itself. |
| 1295 | */ |
| 1296 | static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp) |
| 1297 | { |
| 1298 | if (!tp->sacked_out) |
| 1299 | return tp->snd_una; |
| 1300 | |
| 1301 | if (tp->highest_sack == NULL) |
| 1302 | return tp->snd_nxt; |
| 1303 | |
| 1304 | return TCP_SKB_CB(tp->highest_sack)->seq; |
| 1305 | } |
| 1306 | |
| 1307 | static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb) |
| 1308 | { |
| 1309 | tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL : |
| 1310 | tcp_write_queue_next(sk, skb); |
| 1311 | } |
| 1312 | |
| 1313 | static inline struct sk_buff *tcp_highest_sack(struct sock *sk) |
| 1314 | { |
| 1315 | return tcp_sk(sk)->highest_sack; |
| 1316 | } |
| 1317 | |
| 1318 | static inline void tcp_highest_sack_reset(struct sock *sk) |
| 1319 | { |
| 1320 | tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk); |
| 1321 | } |
| 1322 | |
| 1323 | /* Called when old skb is about to be deleted (to be combined with new skb) */ |
| 1324 | static inline void tcp_highest_sack_combine(struct sock *sk, |
| 1325 | struct sk_buff *old, |
| 1326 | struct sk_buff *new) |
| 1327 | { |
| 1328 | if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack)) |
| 1329 | tcp_sk(sk)->highest_sack = new; |
| 1330 | } |
| 1331 | |
| 1332 | /* /proc */ |
| 1333 | enum tcp_seq_states { |
| 1334 | TCP_SEQ_STATE_LISTENING, |
| 1335 | TCP_SEQ_STATE_OPENREQ, |
| 1336 | TCP_SEQ_STATE_ESTABLISHED, |
| 1337 | TCP_SEQ_STATE_TIME_WAIT, |
| 1338 | }; |
| 1339 | |
| 1340 | struct tcp_seq_afinfo { |
| 1341 | char *name; |
| 1342 | sa_family_t family; |
| 1343 | struct file_operations seq_fops; |
| 1344 | struct seq_operations seq_ops; |
| 1345 | }; |
| 1346 | |
| 1347 | struct tcp_iter_state { |
| 1348 | struct seq_net_private p; |
| 1349 | sa_family_t family; |
| 1350 | enum tcp_seq_states state; |
| 1351 | struct sock *syn_wait_sk; |
| 1352 | int bucket, sbucket, num, uid; |
| 1353 | }; |
| 1354 | |
| 1355 | extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo); |
| 1356 | extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo); |
| 1357 | |
| 1358 | extern struct request_sock_ops tcp_request_sock_ops; |
| 1359 | extern struct request_sock_ops tcp6_request_sock_ops; |
| 1360 | |
| 1361 | extern void tcp_v4_destroy_sock(struct sock *sk); |
| 1362 | |
| 1363 | extern int tcp_v4_gso_send_check(struct sk_buff *skb); |
| 1364 | extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features); |
| 1365 | |
| 1366 | #ifdef CONFIG_PROC_FS |
| 1367 | extern int tcp4_proc_init(void); |
| 1368 | extern void tcp4_proc_exit(void); |
| 1369 | #endif |
| 1370 | |
| 1371 | /* TCP af-specific functions */ |
| 1372 | struct tcp_sock_af_ops { |
| 1373 | #ifdef CONFIG_TCP_MD5SIG |
| 1374 | struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk, |
| 1375 | struct sock *addr_sk); |
| 1376 | int (*calc_md5_hash) (char *location, |
| 1377 | struct tcp_md5sig_key *md5, |
| 1378 | struct sock *sk, |
| 1379 | struct request_sock *req, |
| 1380 | struct sk_buff *skb); |
| 1381 | int (*md5_add) (struct sock *sk, |
| 1382 | struct sock *addr_sk, |
| 1383 | u8 *newkey, |
| 1384 | u8 len); |
| 1385 | int (*md5_parse) (struct sock *sk, |
| 1386 | char __user *optval, |
| 1387 | int optlen); |
| 1388 | #endif |
| 1389 | }; |
| 1390 | |
| 1391 | struct tcp_request_sock_ops { |
| 1392 | #ifdef CONFIG_TCP_MD5SIG |
| 1393 | struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk, |
| 1394 | struct request_sock *req); |
| 1395 | #endif |
| 1396 | }; |
| 1397 | |
| 1398 | extern void tcp_v4_init(void); |
| 1399 | extern void tcp_init(void); |
| 1400 | |
| 1401 | #endif /* _TCP_H */ |