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 #ifdef CONFIG_ACCEPT_QUEUES
132 #define TCP_ACCEPTQ_SHARE 13 /* Set accept queue share */
135 #define TCPI_OPT_TIMESTAMPS 1
136 #define TCPI_OPT_SACK 2
137 #define TCPI_OPT_WSCALE 4
138 #define TCPI_OPT_ECN 8
143 #define TCPF_CA_Open (1<<TCP_CA_Open)
145 #define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
147 #define TCPF_CA_CWR (1<<TCP_CA_CWR)
149 #define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
151 #define TCPF_CA_Loss (1<<TCP_CA_Loss)
158 __u8 tcpi_retransmits;
162 __u8 tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
176 __u32 tcpi_last_data_sent;
177 __u32 tcpi_last_ack_sent; /* Not remembered, sorry. */
178 __u32 tcpi_last_data_recv;
179 __u32 tcpi_last_ack_recv;
183 __u32 tcpi_rcv_ssthresh;
186 __u32 tcpi_snd_ssthresh;
189 __u32 tcpi_reordering;
192 __u32 tcpi_rcv_space;
195 #ifdef CONFIG_ACCEPT_QUEUES
197 #define NUM_ACCEPT_QUEUES 8 /* Must be power of 2 */
199 struct tcp_acceptq_info {
200 unsigned char acceptq_shares;
201 unsigned long acceptq_wait_time;
202 unsigned int acceptq_qcount;
203 unsigned int acceptq_count;
209 #include <linux/config.h>
210 #include <linux/skbuff.h>
211 #include <linux/ip.h>
212 #include <net/sock.h>
214 /* This defines a selective acknowledgement block. */
215 struct tcp_sack_block {
221 int tcp_header_len; /* Bytes of tcp header to send */
224 * Header prediction flags
225 * 0x5?10 << 16 + snd_wnd in net byte order
230 * RFC793 variables by their proper names. This means you can
231 * read the code and the spec side by side (and laugh ...)
232 * See RFC793 and RFC1122. The RFC writes these in capitals.
234 __u32 rcv_nxt; /* What we want to receive next */
235 __u32 snd_nxt; /* Next sequence we send */
237 __u32 snd_una; /* First byte we want an ack for */
238 __u32 snd_sml; /* Last byte of the most recently transmitted small packet */
239 __u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */
240 __u32 lsndtime; /* timestamp of last sent data packet (for restart window) */
241 struct tcp_bind_bucket *bind_hash;
242 /* Delayed ACK control data */
244 __u8 pending; /* ACK is pending */
245 __u8 quick; /* Scheduled number of quick acks */
246 __u8 pingpong; /* The session is interactive */
247 __u8 blocked; /* Delayed ACK was blocked by socket lock*/
248 __u32 ato; /* Predicted tick of soft clock */
249 unsigned long timeout; /* Currently scheduled timeout */
250 __u32 lrcvtime; /* timestamp of last received data packet*/
251 __u16 last_seg_size; /* Size of last incoming segment */
252 __u16 rcv_mss; /* MSS used for delayed ACK decisions */
255 /* Data for direct copy to user */
257 struct sk_buff_head prequeue;
258 struct task_struct *task;
264 __u32 snd_wl1; /* Sequence for window update */
265 __u32 snd_wnd; /* The window we expect to receive */
266 __u32 max_window; /* Maximal window ever seen from peer */
267 __u32 pmtu_cookie; /* Last pmtu seen by socket */
268 __u32 mss_cache; /* Cached effective mss, not including SACKS */
269 __u16 mss_cache_std; /* Like mss_cache, but without TSO */
270 __u16 mss_clamp; /* Maximal mss, negotiated at connection setup */
271 __u16 ext_header_len; /* Network protocol overhead (IP/IPv6 options) */
272 __u16 ext2_header_len;/* Options depending on route */
273 __u8 ca_state; /* State of fast-retransmit machine */
274 __u8 retransmits; /* Number of unrecovered RTO timeouts. */
276 __u8 reordering; /* Packet reordering metric. */
277 __u8 frto_counter; /* Number of new acks after RTO */
278 __u32 frto_highmark; /* snd_nxt when RTO occurred */
281 __u8 defer_accept; /* User waits for some data after accept() */
282 /* one byte hole, try to pack */
284 /* RTT measurement */
285 __u8 backoff; /* backoff */
286 __u32 srtt; /* smothed round trip time << 3 */
287 __u32 mdev; /* medium deviation */
288 __u32 mdev_max; /* maximal mdev for the last rtt period */
289 __u32 rttvar; /* smoothed mdev_max */
290 __u32 rtt_seq; /* sequence number to update rttvar */
291 __u32 rto; /* retransmit timeout */
293 __u32 packets_out; /* Packets which are "in flight" */
294 __u32 left_out; /* Packets which leaved network */
295 __u32 retrans_out; /* Retransmitted packets out */
299 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
301 __u32 snd_ssthresh; /* Slow start size threshold */
302 __u32 snd_cwnd; /* Sending congestion window */
303 __u16 snd_cwnd_cnt; /* Linear increase counter */
304 __u16 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
306 __u32 snd_cwnd_stamp;
308 /* Two commonly used timers in both sender and receiver paths. */
309 unsigned long timeout;
310 struct timer_list retransmit_timer; /* Resend (no ack) */
311 struct timer_list delack_timer; /* Ack delay */
313 struct sk_buff_head out_of_order_queue; /* Out of order segments go here */
315 struct tcp_func *af_specific; /* Operations which are AF_INET{4,6} specific */
317 __u32 rcv_wnd; /* Current receiver window */
318 __u32 rcv_wup; /* rcv_nxt on last window update sent */
319 __u32 write_seq; /* Tail(+1) of data held in tcp send buffer */
320 __u32 pushed_seq; /* Last pushed seq, required to talk to windows */
321 __u32 copied_seq; /* Head of yet unread data */
323 * Options received (usually on last packet, some only on SYN packets).
325 char tstamp_ok, /* TIMESTAMP seen on SYN packet */
326 wscale_ok, /* Wscale seen on SYN packet */
327 sack_ok; /* SACK seen on SYN packet */
328 char saw_tstamp; /* Saw TIMESTAMP on last packet */
329 __u8 snd_wscale; /* Window scaling received from sender */
330 __u8 rcv_wscale; /* Window scaling to send to receiver */
331 __u8 nonagle; /* Disable Nagle algorithm? */
332 __u8 keepalive_probes; /* num of allowed keep alive probes */
335 __u32 rcv_tsval; /* Time stamp value */
336 __u32 rcv_tsecr; /* Time stamp echo reply */
337 __u32 ts_recent; /* Time stamp to echo next */
338 long ts_recent_stamp;/* Time we stored ts_recent (for aging) */
341 __u16 user_mss; /* mss requested by user in ioctl */
342 __u8 dsack; /* D-SACK is scheduled */
343 __u8 eff_sacks; /* Size of SACK array to send with next packet */
344 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
345 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
347 __u32 window_clamp; /* Maximal window to advertise */
348 __u32 rcv_ssthresh; /* Current window clamp */
349 __u8 probes_out; /* unanswered 0 window probes */
350 __u8 num_sacks; /* Number of SACK blocks */
351 __u16 advmss; /* Advertised MSS */
353 __u8 syn_retries; /* num of allowed syn retries */
354 __u8 ecn_flags; /* ECN status bits. */
355 __u16 prior_ssthresh; /* ssthresh saved at recovery start */
356 __u32 lost_out; /* Lost packets */
357 __u32 sacked_out; /* SACK'd packets */
358 __u32 fackets_out; /* FACK'd packets */
359 __u32 high_seq; /* snd_nxt at onset of congestion */
361 __u32 retrans_stamp; /* Timestamp of the last retransmit,
362 * also used in SYN-SENT to remember stamp of
364 __u32 undo_marker; /* tracking retrans started here. */
365 int undo_retrans; /* number of undoable retransmissions. */
366 __u32 urg_seq; /* Seq of received urgent pointer */
367 __u16 urg_data; /* Saved octet of OOB data and control flags */
368 __u8 pending; /* Scheduled timer event */
369 __u8 urg_mode; /* In urgent mode */
370 __u32 snd_up; /* Urgent pointer */
372 /* The syn_wait_lock is necessary only to avoid proc interface having
373 * to grab the main lock sock while browsing the listening hash
374 * (otherwise it's deadlock prone).
375 * This lock is acquired in read mode only from listening_get_next()
376 * and it's acquired in write mode _only_ from code that is actively
377 * changing the syn_wait_queue. All readers that are holding
378 * the master sock lock don't need to grab this lock in read mode
379 * too as the syn_wait_queue writes are always protected from
380 * the main sock lock.
382 rwlock_t syn_wait_lock;
383 struct tcp_listen_opt *listen_opt;
385 /* FIFO of established children */
386 struct open_request *accept_queue;
387 #ifndef CONFIG_ACCEPT_QUEUES
388 struct open_request *accept_queue_tail;
391 unsigned int keepalive_time; /* time before keep alive takes place */
392 unsigned int keepalive_intvl; /* time interval between keep alive probes */
395 unsigned long last_synq_overflow;
397 /* Receiver side RTT estimation */
404 /* Receiver queue space */
411 /* TCP Westwood structure */
413 __u32 bw_ns_est; /* first bandwidth estimation..not too smoothed 8) */
414 __u32 bw_est; /* bandwidth estimate */
415 __u32 rtt_win_sx; /* here starts a new evaluation... */
417 __u32 snd_una; /* used for evaluating the number of acked bytes */
421 __u32 rtt_min; /* minimum observed RTT */
424 /* Vegas variables */
426 __u32 beg_snd_nxt; /* right edge during last RTT */
427 __u32 beg_snd_una; /* left edge during last RTT */
428 __u32 beg_snd_cwnd; /* saves the size of the cwnd */
429 __u8 do_vegas; /* do vegas for this connection */
430 __u8 doing_vegas_now;/* if true, do vegas for this RTT */
431 __u16 cntRTT; /* # of RTTs measured within last RTT */
432 __u32 minRTT; /* min of RTTs measured within last RTT (in usec) */
433 __u32 baseRTT; /* the min of all Vegas RTT measurements seen (in usec) */
436 /* BI TCP Parameters */
438 __u32 cnt; /* increase cwnd by 1 after this number of ACKs */
439 __u32 last_max_cwnd; /* last maximium snd_cwnd */
440 __u32 last_cwnd; /* the last snd_cwnd */
443 #ifdef CONFIG_ACCEPT_QUEUES
444 /* move to listen opt... */
447 struct open_request *aq_head;
448 struct open_request *aq_tail;
450 unsigned int aq_ratio;
451 unsigned int aq_count;
452 unsigned int aq_qcount;
453 unsigned int aq_backlog;
454 unsigned int aq_wait_time;
455 } acceptq[NUM_ACCEPT_QUEUES];
459 /* WARNING: don't change the layout of the members in tcp_sock! */
462 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
463 struct ipv6_pinfo *pinet6;
465 struct inet_opt inet;
469 static inline struct tcp_opt * tcp_sk(const struct sock *__sk)
471 return &((struct tcp_sock *)__sk)->tcp;
476 #endif /* _LINUX_TCP_H */