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.
6 * Definitions for the TCP protocol.
8 * Version: @(#)tcp.h 1.0.2 04/28/93
10 * Author: Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
20 #include <linux/types.h>
21 #include <asm/byteorder.h>
28 #if defined(__LITTLE_ENDIAN_BITFIELD)
39 #elif defined(__BIG_ENDIAN_BITFIELD)
51 #error "Adjust your <asm/byteorder.h> defines"
70 TCP_CLOSING, /* now a valid state */
72 TCP_MAX_STATES /* Leave at the end! */
75 #define TCP_STATE_MASK 0xF
76 #define TCP_ACTION_FIN (1 << 7)
79 TCPF_ESTABLISHED = (1 << 1),
80 TCPF_SYN_SENT = (1 << 2),
81 TCPF_SYN_RECV = (1 << 3),
82 TCPF_FIN_WAIT1 = (1 << 4),
83 TCPF_FIN_WAIT2 = (1 << 5),
84 TCPF_TIME_WAIT = (1 << 6),
85 TCPF_CLOSE = (1 << 7),
86 TCPF_CLOSE_WAIT = (1 << 8),
87 TCPF_LAST_ACK = (1 << 9),
88 TCPF_LISTEN = (1 << 10),
89 TCPF_CLOSING = (1 << 11)
93 * The union cast uses a gcc extension to avoid aliasing problems
94 * (union is compatible to any of its members)
95 * This means this part of the code is -fstrict-aliasing safe now.
102 #define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3])
105 TCP_FLAG_CWR = __constant_htonl(0x00800000),
106 TCP_FLAG_ECE = __constant_htonl(0x00400000),
107 TCP_FLAG_URG = __constant_htonl(0x00200000),
108 TCP_FLAG_ACK = __constant_htonl(0x00100000),
109 TCP_FLAG_PSH = __constant_htonl(0x00080000),
110 TCP_FLAG_RST = __constant_htonl(0x00040000),
111 TCP_FLAG_SYN = __constant_htonl(0x00020000),
112 TCP_FLAG_FIN = __constant_htonl(0x00010000),
113 TCP_RESERVED_BITS = __constant_htonl(0x0F000000),
114 TCP_DATA_OFFSET = __constant_htonl(0xF0000000)
117 /* TCP socket options */
118 #define TCP_NODELAY 1 /* Turn off Nagle's algorithm. */
119 #define TCP_MAXSEG 2 /* Limit MSS */
120 #define TCP_CORK 3 /* Never send partially complete segments */
121 #define TCP_KEEPIDLE 4 /* Start keeplives after this period */
122 #define TCP_KEEPINTVL 5 /* Interval between keepalives */
123 #define TCP_KEEPCNT 6 /* Number of keepalives before death */
124 #define TCP_SYNCNT 7 /* Number of SYN retransmits */
125 #define TCP_LINGER2 8 /* Life time of orphaned FIN-WAIT-2 state */
126 #define TCP_DEFER_ACCEPT 9 /* Wake up listener only when data arrive */
127 #define TCP_WINDOW_CLAMP 10 /* Bound advertised window */
128 #define TCP_INFO 11 /* Information about this connection. */
129 #define TCP_QUICKACK 12 /* Block/reenable quick acks */
131 #define TCPI_OPT_TIMESTAMPS 1
132 #define TCPI_OPT_SACK 2
133 #define TCPI_OPT_WSCALE 4
134 #define TCPI_OPT_ECN 8
139 #define TCPF_CA_Open (1<<TCP_CA_Open)
141 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
143 #define TCPF_CA_CWR (1<<TCP_CA_CWR)
145 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
147 #define TCPF_CA_Loss (1<<TCP_CA_Loss)
154 __u8 tcpi_retransmits;
158 __u8 tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
172 __u32 tcpi_last_data_sent;
173 __u32 tcpi_last_ack_sent; /* Not remembered, sorry. */
174 __u32 tcpi_last_data_recv;
175 __u32 tcpi_last_ack_recv;
179 __u32 tcpi_rcv_ssthresh;
182 __u32 tcpi_snd_ssthresh;
185 __u32 tcpi_reordering;
188 __u32 tcpi_rcv_space;
190 __u32 tcpi_total_retrans;
195 #include <linux/config.h>
196 #include <linux/skbuff.h>
197 #include <linux/ip.h>
198 #include <net/sock.h>
200 /* This defines a selective acknowledgement block. */
201 struct tcp_sack_block {
206 enum tcp_congestion_algo {
213 struct tcp_options_received {
215 long ts_recent_stamp;/* Time we stored ts_recent (for aging) */
216 __u32 ts_recent; /* Time stamp to echo next */
217 __u32 rcv_tsval; /* Time stamp value */
218 __u32 rcv_tsecr; /* Time stamp echo reply */
219 __u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */
220 tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */
221 dsack : 1, /* D-SACK is scheduled */
222 wscale_ok : 1, /* Wscale seen on SYN packet */
223 sack_ok : 4, /* SACK seen on SYN packet */
224 snd_wscale : 4, /* Window scaling received from sender */
225 rcv_wscale : 4; /* Window scaling to send to receiver */
227 __u8 eff_sacks; /* Size of SACK array to send with next packet */
228 __u8 num_sacks; /* Number of SACK blocks */
229 __u16 user_mss; /* mss requested by user in ioctl */
230 __u16 mss_clamp; /* Maximal mss, negotiated at connection setup */
234 /* inet_sock has to be the first member of tcp_sock */
235 struct inet_sock inet;
236 int tcp_header_len; /* Bytes of tcp header to send */
239 * Header prediction flags
240 * 0x5?10 << 16 + snd_wnd in net byte order
245 * RFC793 variables by their proper names. This means you can
246 * read the code and the spec side by side (and laugh ...)
247 * See RFC793 and RFC1122. The RFC writes these in capitals.
249 __u32 rcv_nxt; /* What we want to receive next */
250 __u32 snd_nxt; /* Next sequence we send */
252 __u32 snd_una; /* First byte we want an ack for */
253 __u32 snd_sml; /* Last byte of the most recently transmitted small packet */
254 __u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */
255 __u32 lsndtime; /* timestamp of last sent data packet (for restart window) */
256 struct tcp_bind_bucket *bind_hash;
257 /* Delayed ACK control data */
259 __u8 pending; /* ACK is pending */
260 __u8 quick; /* Scheduled number of quick acks */
261 __u8 pingpong; /* The session is interactive */
262 __u8 blocked; /* Delayed ACK was blocked by socket lock*/
263 __u32 ato; /* Predicted tick of soft clock */
264 unsigned long timeout; /* Currently scheduled timeout */
265 __u32 lrcvtime; /* timestamp of last received data packet*/
266 __u16 last_seg_size; /* Size of last incoming segment */
267 __u16 rcv_mss; /* MSS used for delayed ACK decisions */
270 /* Data for direct copy to user */
272 struct sk_buff_head prequeue;
273 struct task_struct *task;
279 __u32 snd_wl1; /* Sequence for window update */
280 __u32 snd_wnd; /* The window we expect to receive */
281 __u32 max_window; /* Maximal window ever seen from peer */
282 __u32 pmtu_cookie; /* Last pmtu seen by socket */
283 __u32 mss_cache; /* Cached effective mss, not including SACKS */
284 __u16 mss_cache_std; /* Like mss_cache, but without TSO */
285 __u16 ext_header_len; /* Network protocol overhead (IP/IPv6 options) */
286 __u8 ca_state; /* State of fast-retransmit machine */
287 __u8 retransmits; /* Number of unrecovered RTO timeouts. */
289 __u16 advmss; /* Advertised MSS */
290 __u32 window_clamp; /* Maximal window to advertise */
291 __u32 rcv_ssthresh; /* Current window clamp */
293 __u32 frto_highmark; /* snd_nxt when RTO occurred */
294 __u8 reordering; /* Packet reordering metric. */
295 __u8 frto_counter; /* Number of new acks after RTO */
297 __u8 adv_cong; /* Using Vegas, Westwood, or BIC */
298 __u8 defer_accept; /* User waits for some data after accept() */
300 /* RTT measurement */
301 __u32 srtt; /* smoothed round trip time << 3 */
302 __u32 mdev; /* medium deviation */
303 __u32 mdev_max; /* maximal mdev for the last rtt period */
304 __u32 rttvar; /* smoothed mdev_max */
305 __u32 rtt_seq; /* sequence number to update rttvar */
306 __u32 rto; /* retransmit timeout */
308 __u32 packets_out; /* Packets which are "in flight" */
309 __u32 left_out; /* Packets which leaved network */
310 __u32 retrans_out; /* Retransmitted packets out */
311 __u8 backoff; /* backoff */
313 * Options received (usually on last packet, some only on SYN packets).
315 __u8 nonagle; /* Disable Nagle algorithm? */
316 __u8 keepalive_probes; /* num of allowed keep alive probes */
318 __u8 probes_out; /* unanswered 0 window probes */
319 struct tcp_options_received rx_opt;
322 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
324 __u32 snd_ssthresh; /* Slow start size threshold */
325 __u32 snd_cwnd; /* Sending congestion window */
326 __u16 snd_cwnd_cnt; /* Linear increase counter */
327 __u16 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
329 __u32 snd_cwnd_stamp;
331 /* Two commonly used timers in both sender and receiver paths. */
332 unsigned long timeout;
333 struct timer_list retransmit_timer; /* Resend (no ack) */
334 struct timer_list delack_timer; /* Ack delay */
336 struct sk_buff_head out_of_order_queue; /* Out of order segments go here */
338 struct tcp_func *af_specific; /* Operations which are AF_INET{4,6} specific */
340 __u32 rcv_wnd; /* Current receiver window */
341 __u32 rcv_wup; /* rcv_nxt on last window update sent */
342 __u32 write_seq; /* Tail(+1) of data held in tcp send buffer */
343 __u32 pushed_seq; /* Last pushed seq, required to talk to windows */
344 __u32 copied_seq; /* Head of yet unread data */
347 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
348 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
350 __u8 syn_retries; /* num of allowed syn retries */
351 __u8 ecn_flags; /* ECN status bits. */
352 __u16 prior_ssthresh; /* ssthresh saved at recovery start */
353 __u32 lost_out; /* Lost packets */
354 __u32 sacked_out; /* SACK'd packets */
355 __u32 fackets_out; /* FACK'd packets */
356 __u32 high_seq; /* snd_nxt at onset of congestion */
358 __u32 retrans_stamp; /* Timestamp of the last retransmit,
359 * also used in SYN-SENT to remember stamp of
361 __u32 undo_marker; /* tracking retrans started here. */
362 int undo_retrans; /* number of undoable retransmissions. */
363 __u32 urg_seq; /* Seq of received urgent pointer */
364 __u16 urg_data; /* Saved octet of OOB data and control flags */
365 __u8 pending; /* Scheduled timer event */
366 __u8 urg_mode; /* In urgent mode */
367 __u32 snd_up; /* Urgent pointer */
369 __u32 total_retrans; /* Total retransmits for entire connection */
371 /* The syn_wait_lock is necessary only to avoid proc interface having
372 * to grab the main lock sock while browsing the listening hash
373 * (otherwise it's deadlock prone).
374 * This lock is acquired in read mode only from listening_get_next()
375 * and it's acquired in write mode _only_ from code that is actively
376 * changing the syn_wait_queue. All readers that are holding
377 * the master sock lock don't need to grab this lock in read mode
378 * too as the syn_wait_queue writes are always protected from
379 * the main sock lock.
381 rwlock_t syn_wait_lock;
382 struct tcp_listen_opt *listen_opt;
384 /* FIFO of established children */
385 struct open_request *accept_queue;
386 struct open_request *accept_queue_tail;
388 unsigned int keepalive_time; /* time before keep alive takes place */
389 unsigned int keepalive_intvl; /* time interval between keep alive probes */
392 unsigned long last_synq_overflow;
394 /* Receiver side RTT estimation */
401 /* Receiver queue space */
408 /* TCP Westwood structure */
410 __u32 bw_ns_est; /* first bandwidth estimation..not too smoothed 8) */
411 __u32 bw_est; /* bandwidth estimate */
412 __u32 rtt_win_sx; /* here starts a new evaluation... */
414 __u32 snd_una; /* used for evaluating the number of acked bytes */
418 __u32 rtt_min; /* minimum observed RTT */
421 /* Vegas variables */
423 __u32 beg_snd_nxt; /* right edge during last RTT */
424 __u32 beg_snd_una; /* left edge during last RTT */
425 __u32 beg_snd_cwnd; /* saves the size of the cwnd */
426 __u8 doing_vegas_now;/* if true, do vegas for this RTT */
427 __u16 cntRTT; /* # of RTTs measured within last RTT */
428 __u32 minRTT; /* min of RTTs measured within last RTT (in usec) */
429 __u32 baseRTT; /* the min of all Vegas RTT measurements seen (in usec) */
432 /* BI TCP Parameters */
434 __u32 cnt; /* increase cwnd by 1 after this number of ACKs */
435 __u32 last_max_cwnd; /* last maximium snd_cwnd */
436 __u32 last_cwnd; /* the last snd_cwnd */
437 __u32 last_stamp; /* time when updated last_cwnd */
441 static inline struct tcp_sock *tcp_sk(const struct sock *sk)
443 return (struct tcp_sock *)sk;
448 #endif /* _LINUX_TCP_H */