blob: e4fd82e4210428458754a263c7e09930c2cb7e8a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 protocol.
7 *
8 * Version: @(#)tcp.h 1.0.2 04/28/93
9 *
10 * Author: Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 *
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.
16 */
17#ifndef _LINUX_TCP_H
18#define _LINUX_TCP_H
19
20#include <linux/types.h>
21#include <asm/byteorder.h>
22
23struct tcphdr {
24 __u16 source;
25 __u16 dest;
26 __u32 seq;
27 __u32 ack_seq;
28#if defined(__LITTLE_ENDIAN_BITFIELD)
29 __u16 res1:4,
30 doff:4,
31 fin:1,
32 syn:1,
33 rst:1,
34 psh:1,
35 ack:1,
36 urg:1,
37 ece:1,
38 cwr:1;
39#elif defined(__BIG_ENDIAN_BITFIELD)
40 __u16 doff:4,
41 res1:4,
42 cwr:1,
43 ece:1,
44 urg:1,
45 ack:1,
46 psh:1,
47 rst:1,
48 syn:1,
49 fin:1;
50#else
51#error "Adjust your <asm/byteorder.h> defines"
52#endif
53 __u16 window;
54 __u16 check;
55 __u16 urg_ptr;
56};
57
58
59enum {
60 TCP_ESTABLISHED = 1,
61 TCP_SYN_SENT,
62 TCP_SYN_RECV,
63 TCP_FIN_WAIT1,
64 TCP_FIN_WAIT2,
65 TCP_TIME_WAIT,
66 TCP_CLOSE,
67 TCP_CLOSE_WAIT,
68 TCP_LAST_ACK,
69 TCP_LISTEN,
70 TCP_CLOSING, /* now a valid state */
71
72 TCP_MAX_STATES /* Leave at the end! */
73};
74
75#define TCP_STATE_MASK 0xF
76#define TCP_ACTION_FIN (1 << 7)
77
78enum {
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)
90};
91
92/*
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.
96 */
97union tcp_word_hdr {
98 struct tcphdr hdr;
99 __u32 words[5];
100};
101
102#define tcp_flag_word(tp) ( ((union tcp_word_hdr *)(tp))->words [3])
103
104enum {
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)
115};
116
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 */
Stephen Hemminger5f8ef482005-06-23 20:37:36 -0700130#define TCP_CONGESTION 13 /* Congestion control algorithm */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131
132#define TCPI_OPT_TIMESTAMPS 1
133#define TCPI_OPT_SACK 2
134#define TCPI_OPT_WSCALE 4
135#define TCPI_OPT_ECN 8
136
137enum tcp_ca_state
138{
139 TCP_CA_Open = 0,
140#define TCPF_CA_Open (1<<TCP_CA_Open)
141 TCP_CA_Disorder = 1,
142#define TCPF_CA_Disorder (1<<TCP_CA_Disorder)
143 TCP_CA_CWR = 2,
144#define TCPF_CA_CWR (1<<TCP_CA_CWR)
145 TCP_CA_Recovery = 3,
146#define TCPF_CA_Recovery (1<<TCP_CA_Recovery)
147 TCP_CA_Loss = 4
148#define TCPF_CA_Loss (1<<TCP_CA_Loss)
149};
150
151struct tcp_info
152{
153 __u8 tcpi_state;
154 __u8 tcpi_ca_state;
155 __u8 tcpi_retransmits;
156 __u8 tcpi_probes;
157 __u8 tcpi_backoff;
158 __u8 tcpi_options;
159 __u8 tcpi_snd_wscale : 4, tcpi_rcv_wscale : 4;
160
161 __u32 tcpi_rto;
162 __u32 tcpi_ato;
163 __u32 tcpi_snd_mss;
164 __u32 tcpi_rcv_mss;
165
166 __u32 tcpi_unacked;
167 __u32 tcpi_sacked;
168 __u32 tcpi_lost;
169 __u32 tcpi_retrans;
170 __u32 tcpi_fackets;
171
172 /* Times. */
173 __u32 tcpi_last_data_sent;
174 __u32 tcpi_last_ack_sent; /* Not remembered, sorry. */
175 __u32 tcpi_last_data_recv;
176 __u32 tcpi_last_ack_recv;
177
178 /* Metrics. */
179 __u32 tcpi_pmtu;
180 __u32 tcpi_rcv_ssthresh;
181 __u32 tcpi_rtt;
182 __u32 tcpi_rttvar;
183 __u32 tcpi_snd_ssthresh;
184 __u32 tcpi_snd_cwnd;
185 __u32 tcpi_advmss;
186 __u32 tcpi_reordering;
187
188 __u32 tcpi_rcv_rtt;
189 __u32 tcpi_rcv_space;
190
191 __u32 tcpi_total_retrans;
192};
193
194#ifdef __KERNEL__
195
196#include <linux/config.h>
197#include <linux/skbuff.h>
198#include <linux/ip.h>
199#include <net/sock.h>
200
201/* This defines a selective acknowledgement block. */
202struct tcp_sack_block {
203 __u32 start_seq;
204 __u32 end_seq;
205};
206
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207struct tcp_options_received {
208/* PAWS/RTTM data */
209 long ts_recent_stamp;/* Time we stored ts_recent (for aging) */
210 __u32 ts_recent; /* Time stamp to echo next */
211 __u32 rcv_tsval; /* Time stamp value */
212 __u32 rcv_tsecr; /* Time stamp echo reply */
213 __u16 saw_tstamp : 1, /* Saw TIMESTAMP on last packet */
214 tstamp_ok : 1, /* TIMESTAMP seen on SYN packet */
215 dsack : 1, /* D-SACK is scheduled */
216 wscale_ok : 1, /* Wscale seen on SYN packet */
217 sack_ok : 4, /* SACK seen on SYN packet */
218 snd_wscale : 4, /* Window scaling received from sender */
219 rcv_wscale : 4; /* Window scaling to send to receiver */
220/* SACKs data */
221 __u8 eff_sacks; /* Size of SACK array to send with next packet */
222 __u8 num_sacks; /* Number of SACK blocks */
223 __u16 user_mss; /* mss requested by user in ioctl */
224 __u16 mss_clamp; /* Maximal mss, negotiated at connection setup */
225};
226
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700227struct tcp_request_sock {
228 struct inet_request_sock req;
229 __u32 rcv_isn;
230 __u32 snt_isn;
231};
232
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700233static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700234{
235 return (struct tcp_request_sock *)req;
236}
237
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238struct tcp_sock {
239 /* inet_sock has to be the first member of tcp_sock */
240 struct inet_sock inet;
241 int tcp_header_len; /* Bytes of tcp header to send */
242
243/*
244 * Header prediction flags
245 * 0x5?10 << 16 + snd_wnd in net byte order
246 */
247 __u32 pred_flags;
248
249/*
250 * RFC793 variables by their proper names. This means you can
251 * read the code and the spec side by side (and laugh ...)
252 * See RFC793 and RFC1122. The RFC writes these in capitals.
253 */
254 __u32 rcv_nxt; /* What we want to receive next */
255 __u32 snd_nxt; /* Next sequence we send */
256
257 __u32 snd_una; /* First byte we want an ack for */
258 __u32 snd_sml; /* Last byte of the most recently transmitted small packet */
259 __u32 rcv_tstamp; /* timestamp of last received ACK (for keepalives) */
260 __u32 lsndtime; /* timestamp of last sent data packet (for restart window) */
261 struct tcp_bind_bucket *bind_hash;
262 /* Delayed ACK control data */
263 struct {
264 __u8 pending; /* ACK is pending */
265 __u8 quick; /* Scheduled number of quick acks */
266 __u8 pingpong; /* The session is interactive */
267 __u8 blocked; /* Delayed ACK was blocked by socket lock*/
268 __u32 ato; /* Predicted tick of soft clock */
269 unsigned long timeout; /* Currently scheduled timeout */
270 __u32 lrcvtime; /* timestamp of last received data packet*/
271 __u16 last_seg_size; /* Size of last incoming segment */
272 __u16 rcv_mss; /* MSS used for delayed ACK decisions */
273 } ack;
274
275 /* Data for direct copy to user */
276 struct {
277 struct sk_buff_head prequeue;
278 struct task_struct *task;
279 struct iovec *iov;
280 int memory;
281 int len;
282 } ucopy;
283
284 __u32 snd_wl1; /* Sequence for window update */
285 __u32 snd_wnd; /* The window we expect to receive */
286 __u32 max_window; /* Maximal window ever seen from peer */
287 __u32 pmtu_cookie; /* Last pmtu seen by socket */
288 __u32 mss_cache; /* Cached effective mss, not including SACKS */
David S. Millerc1b4a7e2005-07-05 15:24:38 -0700289 __u16 xmit_size_goal; /* Goal for segmenting output packets */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 __u16 ext_header_len; /* Network protocol overhead (IP/IPv6 options) */
291 __u8 ca_state; /* State of fast-retransmit machine */
292 __u8 retransmits; /* Number of unrecovered RTO timeouts. */
293
294 __u16 advmss; /* Advertised MSS */
295 __u32 window_clamp; /* Maximal window to advertise */
296 __u32 rcv_ssthresh; /* Current window clamp */
297
298 __u32 frto_highmark; /* snd_nxt when RTO occurred */
299 __u8 reordering; /* Packet reordering metric. */
300 __u8 frto_counter; /* Number of new acks after RTO */
301
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700302 __u8 unused;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 __u8 defer_accept; /* User waits for some data after accept() */
304
305/* RTT measurement */
306 __u32 srtt; /* smoothed round trip time << 3 */
307 __u32 mdev; /* medium deviation */
308 __u32 mdev_max; /* maximal mdev for the last rtt period */
309 __u32 rttvar; /* smoothed mdev_max */
310 __u32 rtt_seq; /* sequence number to update rttvar */
311 __u32 rto; /* retransmit timeout */
312
313 __u32 packets_out; /* Packets which are "in flight" */
314 __u32 left_out; /* Packets which leaved network */
315 __u32 retrans_out; /* Retransmitted packets out */
316 __u8 backoff; /* backoff */
317/*
318 * Options received (usually on last packet, some only on SYN packets).
319 */
320 __u8 nonagle; /* Disable Nagle algorithm? */
321 __u8 keepalive_probes; /* num of allowed keep alive probes */
322
323 __u8 probes_out; /* unanswered 0 window probes */
324 struct tcp_options_received rx_opt;
325
326/*
327 * Slow start and congestion control (see also Nagle, and Karn & Partridge)
328 */
329 __u32 snd_ssthresh; /* Slow start size threshold */
330 __u32 snd_cwnd; /* Sending congestion window */
331 __u16 snd_cwnd_cnt; /* Linear increase counter */
332 __u16 snd_cwnd_clamp; /* Do not allow snd_cwnd to grow above this */
333 __u32 snd_cwnd_used;
334 __u32 snd_cwnd_stamp;
335
336 /* Two commonly used timers in both sender and receiver paths. */
337 unsigned long timeout;
338 struct timer_list retransmit_timer; /* Resend (no ack) */
339 struct timer_list delack_timer; /* Ack delay */
340
341 struct sk_buff_head out_of_order_queue; /* Out of order segments go here */
342
343 struct tcp_func *af_specific; /* Operations which are AF_INET{4,6} specific */
344
345 __u32 rcv_wnd; /* Current receiver window */
346 __u32 rcv_wup; /* rcv_nxt on last window update sent */
347 __u32 write_seq; /* Tail(+1) of data held in tcp send buffer */
348 __u32 pushed_seq; /* Last pushed seq, required to talk to windows */
349 __u32 copied_seq; /* Head of yet unread data */
350
351/* SACKs data */
352 struct tcp_sack_block duplicate_sack[1]; /* D-SACK block */
353 struct tcp_sack_block selective_acks[4]; /* The SACKS themselves*/
354
355 __u8 syn_retries; /* num of allowed syn retries */
356 __u8 ecn_flags; /* ECN status bits. */
357 __u16 prior_ssthresh; /* ssthresh saved at recovery start */
358 __u32 lost_out; /* Lost packets */
359 __u32 sacked_out; /* SACK'd packets */
360 __u32 fackets_out; /* FACK'd packets */
361 __u32 high_seq; /* snd_nxt at onset of congestion */
362
363 __u32 retrans_stamp; /* Timestamp of the last retransmit,
364 * also used in SYN-SENT to remember stamp of
365 * the first SYN. */
366 __u32 undo_marker; /* tracking retrans started here. */
367 int undo_retrans; /* number of undoable retransmissions. */
368 __u32 urg_seq; /* Seq of received urgent pointer */
369 __u16 urg_data; /* Saved octet of OOB data and control flags */
370 __u8 pending; /* Scheduled timer event */
371 __u8 urg_mode; /* In urgent mode */
372 __u32 snd_up; /* Urgent pointer */
373
374 __u32 total_retrans; /* Total retransmits for entire connection */
375
Arnaldo Carvalho de Melo0e875062005-06-18 22:47:59 -0700376 struct request_sock_queue accept_queue; /* FIFO of established children */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377
378 unsigned int keepalive_time; /* time before keep alive takes place */
379 unsigned int keepalive_intvl; /* time interval between keep alive probes */
380 int linger2;
381
382 unsigned long last_synq_overflow;
383
384/* Receiver side RTT estimation */
385 struct {
386 __u32 rtt;
387 __u32 seq;
388 __u32 time;
389 } rcv_rtt_est;
390
391/* Receiver queue space */
392 struct {
393 int space;
394 __u32 seq;
395 __u32 time;
396 } rcvq_space;
397
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700398 /* Pluggable TCP congestion control hook */
399 struct tcp_congestion_ops *ca_ops;
400 u32 ca_priv[16];
401#define TCP_CA_PRIV_SIZE (16*sizeof(u32))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402};
403
404static inline struct tcp_sock *tcp_sk(const struct sock *sk)
405{
406 return (struct tcp_sock *)sk;
407}
408
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700409static inline void *tcp_ca(const struct tcp_sock *tp)
410{
411 return (void *) tp->ca_priv;
412}
413
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414#endif
415
416#endif /* _LINUX_TCP_H */