blob: 456983d1fcec80f327ec7994eccedecba668c6dd [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 module.
7 *
8 * Version: @(#)tcp.h 1.0.5 05/23/93
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * 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
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/list.h>
25#include <linux/tcp.h>
26#include <linux/slab.h>
27#include <linux/cache.h>
28#include <linux/percpu.h>
Herbert Xufb286bb2005-11-10 13:01:24 -080029#include <linux/skbuff.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070030#include <linux/dmaengine.h>
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -080031#include <linux/crypto.h>
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070032
33#include <net/inet_connection_sock.h>
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -070034#include <net/inet_timewait_sock.h>
Arnaldo Carvalho de Melo77d8bf92005-08-09 20:00:51 -070035#include <net/inet_hashtables.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <net/checksum.h>
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070037#include <net/request_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#include <net/sock.h>
39#include <net/snmp.h>
40#include <net/ip.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070041#include <net/tcp_states.h>
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -070042#include <net/inet_ecn.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070043
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/seq_file.h>
45
Arnaldo Carvalho de Melo6e04e022005-08-09 20:07:35 -070046extern struct inet_hashinfo tcp_hashinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -070047
Linus Torvalds1da177e2005-04-16 15:20:36 -070048extern atomic_t tcp_orphan_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -070049extern void tcp_time_wait(struct sock *sk, int state, int timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
Linus Torvalds1da177e2005-04-16 15:20:36 -070051#define MAX_TCP_HEADER (128 + MAX_HEADER)
52
53/*
54 * Never offer a window over 32767 without using window scaling. Some
55 * poor stacks do signed 16bit maths!
56 */
57#define MAX_TCP_WINDOW 32767U
58
59/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
60#define TCP_MIN_MSS 88U
61
62/* Minimal RCV_MSS. */
63#define TCP_MIN_RCVMSS 536U
64
John Heffner5d424d52006-03-20 17:53:41 -080065/* The least MTU to use for probing */
66#define TCP_BASE_MSS 512
67
Linus Torvalds1da177e2005-04-16 15:20:36 -070068/* After receiving this amount of duplicate ACKs fast retransmit starts. */
69#define TCP_FASTRETRANS_THRESH 3
70
71/* Maximal reordering. */
72#define TCP_MAX_REORDERING 127
73
74/* Maximal number of ACKs sent quickly to accelerate slow-start. */
75#define TCP_MAX_QUICKACKS 16U
76
77/* urg_data states */
78#define TCP_URG_VALID 0x0100
79#define TCP_URG_NOTYET 0x0200
80#define TCP_URG_READ 0x0400
81
82#define TCP_RETR1 3 /*
83 * This is how many retries it does before it
84 * tries to figure out if the gateway is
85 * down. Minimal RFC value is 3; it corresponds
86 * to ~3sec-8min depending on RTO.
87 */
88
89#define TCP_RETR2 15 /*
90 * This should take at least
91 * 90 minutes to time out.
92 * RFC1122 says that the limit is 100 sec.
93 * 15 is ~13-30min depending on RTO.
94 */
95
96#define TCP_SYN_RETRIES 5 /* number of times to retry active opening a
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -080097 * connection: ~180sec is RFC minimum */
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
99#define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800100 * connection: ~180sec is RFC minimum */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
102
103#define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned
104 * socket. 7 is ~50sec-16min.
105 */
106
107
108#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
109 * state, about 60 seconds */
110#define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
111 /* BSD style FIN_WAIT2 deadlock breaker.
112 * It used to be 3min, new value is 60sec,
113 * to combine FIN-WAIT-2 timeout with
114 * TIME-WAIT timer.
115 */
116
117#define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
118#if HZ >= 100
119#define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
120#define TCP_ATO_MIN ((unsigned)(HZ/25))
121#else
122#define TCP_DELACK_MIN 4U
123#define TCP_ATO_MIN 4U
124#endif
125#define TCP_RTO_MAX ((unsigned)(120*HZ))
126#define TCP_RTO_MIN ((unsigned)(HZ/5))
127#define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */
128
129#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
130 * for local resources.
131 */
132
133#define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
134#define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
135#define TCP_KEEPALIVE_INTVL (75*HZ)
136
137#define MAX_TCP_KEEPIDLE 32767
138#define MAX_TCP_KEEPINTVL 32767
139#define MAX_TCP_KEEPCNT 127
140#define MAX_TCP_SYNCNT 127
141
142#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
144#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
145#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
146 * after this time. It should be equal
147 * (or greater than) TCP_TIMEWAIT_LEN
148 * to provide reliability equal to one
149 * provided by timewait state.
150 */
151#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
152 * timestamps. It must be less than
153 * minimal timewait lifetime.
154 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155/*
156 * TCP option
157 */
158
159#define TCPOPT_NOP 1 /* Padding */
160#define TCPOPT_EOL 0 /* End of options */
161#define TCPOPT_MSS 2 /* Segment size negotiating */
162#define TCPOPT_WINDOW 3 /* Window scaling */
163#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
164#define TCPOPT_SACK 5 /* SACK Block */
165#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800166#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167
168/*
169 * TCP option lengths
170 */
171
172#define TCPOLEN_MSS 4
173#define TCPOLEN_WINDOW 3
174#define TCPOLEN_SACK_PERM 2
175#define TCPOLEN_TIMESTAMP 10
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800176#define TCPOLEN_MD5SIG 18
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177
178/* But this is what stacks really send out. */
179#define TCPOLEN_TSTAMP_ALIGNED 12
180#define TCPOLEN_WSCALE_ALIGNED 4
181#define TCPOLEN_SACKPERM_ALIGNED 4
182#define TCPOLEN_SACK_BASE 2
183#define TCPOLEN_SACK_BASE_ALIGNED 4
184#define TCPOLEN_SACK_PERBLOCK 8
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800185#define TCPOLEN_MD5SIG_ALIGNED 20
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187/* Flags in tp->nonagle */
188#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
189#define TCP_NAGLE_CORK 2 /* Socket is corked */
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800190#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -0700192extern struct inet_timewait_death_row tcp_death_row;
193
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194/* sysctl variables for tcp */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195extern int sysctl_tcp_timestamps;
196extern int sysctl_tcp_window_scaling;
197extern int sysctl_tcp_sack;
198extern int sysctl_tcp_fin_timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199extern int sysctl_tcp_keepalive_time;
200extern int sysctl_tcp_keepalive_probes;
201extern int sysctl_tcp_keepalive_intvl;
202extern int sysctl_tcp_syn_retries;
203extern int sysctl_tcp_synack_retries;
204extern int sysctl_tcp_retries1;
205extern int sysctl_tcp_retries2;
206extern int sysctl_tcp_orphan_retries;
207extern int sysctl_tcp_syncookies;
208extern int sysctl_tcp_retrans_collapse;
209extern int sysctl_tcp_stdurg;
210extern int sysctl_tcp_rfc1337;
211extern int sysctl_tcp_abort_on_overflow;
212extern int sysctl_tcp_max_orphans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213extern int sysctl_tcp_fack;
214extern int sysctl_tcp_reordering;
215extern int sysctl_tcp_ecn;
216extern int sysctl_tcp_dsack;
217extern int sysctl_tcp_mem[3];
218extern int sysctl_tcp_wmem[3];
219extern int sysctl_tcp_rmem[3];
220extern int sysctl_tcp_app_win;
221extern int sysctl_tcp_adv_win_scale;
222extern int sysctl_tcp_tw_reuse;
223extern int sysctl_tcp_frto;
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800224extern int sysctl_tcp_frto_response;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225extern int sysctl_tcp_low_latency;
Chris Leech95937822006-05-23 18:02:55 -0700226extern int sysctl_tcp_dma_copybreak;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227extern int sysctl_tcp_nometrics_save;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228extern int sysctl_tcp_moderate_rcvbuf;
229extern int sysctl_tcp_tso_win_divisor;
Stephen Hemminger9772efb2005-11-10 17:09:53 -0800230extern int sysctl_tcp_abc;
John Heffner5d424d52006-03-20 17:53:41 -0800231extern int sysctl_tcp_mtu_probing;
232extern int sysctl_tcp_base_mss;
Rick Jones15d99e02006-03-20 22:40:29 -0800233extern int sysctl_tcp_workaround_signed_windows;
David S. Miller35089bb2006-06-13 22:33:04 -0700234extern int sysctl_tcp_slow_start_after_idle;
John Heffner886236c2007-03-25 19:21:45 -0700235extern int sysctl_tcp_max_ssthresh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
237extern atomic_t tcp_memory_allocated;
238extern atomic_t tcp_sockets_allocated;
239extern int tcp_memory_pressure;
240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 * The next routines deal with comparing 32 bit unsigned ints
243 * and worry about wraparound (automatic with unsigned arithmetic).
244 */
245
246static inline int before(__u32 seq1, __u32 seq2)
247{
Gerrit Renker0d630cc2007-01-04 12:25:16 -0800248 return (__s32)(seq1-seq2) < 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249}
Gerrit Renker9a036b9c2006-12-20 10:25:55 -0800250#define after(seq2, seq1) before(seq1, seq2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251
252/* is s2<=s1<=s3 ? */
253static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
254{
255 return seq3 - seq2 >= seq1 - seq2;
256}
257
Pavel Emelianove4fd5da2007-05-29 13:19:18 -0700258static inline int tcp_too_many_orphans(struct sock *sk, int num)
259{
260 return (num > sysctl_tcp_max_orphans) ||
261 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
262 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
263}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
265extern struct proto tcp_prot;
266
267DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
268#define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field)
269#define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field)
270#define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field)
271#define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field)
272#define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val)
273#define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val)
274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275extern void tcp_v4_err(struct sk_buff *skb, u32);
276
277extern void tcp_shutdown (struct sock *sk, int how);
278
279extern int tcp_v4_rcv(struct sk_buff *skb);
280
281extern int tcp_v4_remember_stamp(struct sock *sk);
282
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700283extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284
David S. Miller3516ffb2007-08-02 19:23:56 -0700285extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 struct msghdr *msg, size_t size);
287extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
288
289extern int tcp_ioctl(struct sock *sk,
290 int cmd,
291 unsigned long arg);
292
293extern int tcp_rcv_state_process(struct sock *sk,
294 struct sk_buff *skb,
295 struct tcphdr *th,
296 unsigned len);
297
298extern int tcp_rcv_established(struct sock *sk,
299 struct sk_buff *skb,
300 struct tcphdr *th,
301 unsigned len);
302
303extern void tcp_rcv_space_adjust(struct sock *sk);
304
Chris Leech0e4b4992006-05-23 18:00:16 -0700305extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
306
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800307extern int tcp_twsk_unique(struct sock *sk,
308 struct sock *sktw, void *twp);
309
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800310extern void tcp_twsk_destructor(struct sock *sk);
311
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700312static inline void tcp_dec_quickack_mode(struct sock *sk,
313 const unsigned int pkts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700315 struct inet_connection_sock *icsk = inet_csk(sk);
David S. Millerfc6415bc2005-07-05 15:17:45 -0700316
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700317 if (icsk->icsk_ack.quick) {
318 if (pkts >= icsk->icsk_ack.quick) {
319 icsk->icsk_ack.quick = 0;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700320 /* Leaving quickack mode we deflate ATO. */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700321 icsk->icsk_ack.ato = TCP_ATO_MIN;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700322 } else
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700323 icsk->icsk_ack.quick -= pkts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 }
325}
326
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700327extern void tcp_enter_quickack_mode(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
330{
331 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
332}
333
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -0700334#define TCP_ECN_OK 1
335#define TCP_ECN_QUEUE_CWR 2
336#define TCP_ECN_DEMAND_CWR 4
337
338static __inline__ void
339TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
340{
341 if (sysctl_tcp_ecn && th->ece && th->cwr)
342 inet_rsk(req)->ecn_ok = 1;
343}
344
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345enum tcp_tw_status
346{
347 TCP_TW_SUCCESS = 0,
348 TCP_TW_RST = 1,
349 TCP_TW_ACK = 2,
350 TCP_TW_SYN = 3
351};
352
353
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700354extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 struct sk_buff *skb,
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700356 const struct tcphdr *th);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357
358extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700359 struct request_sock *req,
360 struct request_sock **prev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361extern int tcp_child_process(struct sock *parent,
362 struct sock *child,
363 struct sk_buff *skb);
Ilpo Järvinen46d0de42007-02-21 23:10:39 -0800364extern int tcp_use_frto(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365extern void tcp_enter_frto(struct sock *sk);
366extern void tcp_enter_loss(struct sock *sk, int how);
367extern void tcp_clear_retrans(struct tcp_sock *tp);
368extern void tcp_update_metrics(struct sock *sk);
369
370extern void tcp_close(struct sock *sk,
371 long timeout);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
373
374extern int tcp_getsockopt(struct sock *sk, int level,
375 int optname,
376 char __user *optval,
377 int __user *optlen);
378extern int tcp_setsockopt(struct sock *sk, int level,
379 int optname, char __user *optval,
380 int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800381extern int compat_tcp_getsockopt(struct sock *sk,
382 int level, int optname,
383 char __user *optval, int __user *optlen);
384extern int compat_tcp_setsockopt(struct sock *sk,
385 int level, int optname,
386 char __user *optval, int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387extern void tcp_set_keepalive(struct sock *sk, int val);
388extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
389 struct msghdr *msg,
390 size_t len, int nonblock,
391 int flags, int *addr_len);
392
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393extern void tcp_parse_options(struct sk_buff *skb,
394 struct tcp_options_received *opt_rx,
395 int estab);
396
397/*
398 * TCP v4 functions exported for the inet6 API
399 */
400
Arnaldo Carvalho de Melo8292a172005-12-13 23:15:52 -0800401extern void tcp_v4_send_check(struct sock *sk, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 struct sk_buff *skb);
403
404extern int tcp_v4_conn_request(struct sock *sk,
405 struct sk_buff *skb);
406
407extern struct sock * tcp_create_openreq_child(struct sock *sk,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700408 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 struct sk_buff *skb);
410
411extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
412 struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700413 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 struct dst_entry *dst);
415
416extern int tcp_v4_do_rcv(struct sock *sk,
417 struct sk_buff *skb);
418
419extern int tcp_v4_connect(struct sock *sk,
420 struct sockaddr *uaddr,
421 int addr_len);
422
423extern int tcp_connect(struct sock *sk);
424
425extern struct sk_buff * tcp_make_synack(struct sock *sk,
426 struct dst_entry *dst,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700427 struct request_sock *req);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428
429extern int tcp_disconnect(struct sock *sk, int flags);
430
431extern void tcp_unhash(struct sock *sk);
432
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433/* From syncookies.c */
434extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
435 struct ip_options *opt);
436extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
437 __u16 *mss);
438
439/* tcp_output.c */
440
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700441extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
442 int nonagle);
443extern int tcp_may_send_now(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
445extern void tcp_xmit_retransmit_queue(struct sock *);
446extern void tcp_simple_retransmit(struct sock *);
447extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
David S. Miller6475be12005-09-01 22:47:01 -0700448extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
450extern void tcp_send_probe0(struct sock *);
451extern void tcp_send_partial(struct sock *);
452extern int tcp_write_wakeup(struct sock *);
453extern void tcp_send_fin(struct sock *sk);
Al Virodd0fc662005-10-07 07:46:04 +0100454extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455extern int tcp_send_synack(struct sock *);
David S. Millerc1b4a7e2005-07-05 15:24:38 -0700456extern void tcp_push_one(struct sock *, unsigned int mss_now);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457extern void tcp_send_ack(struct sock *sk);
458extern void tcp_send_delayed_ack(struct sock *sk);
459
David S. Millera762a982005-07-05 15:18:51 -0700460/* tcp_input.c */
461extern void tcp_cwnd_application_limited(struct sock *sk);
462
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463/* tcp_timer.c */
464extern void tcp_init_xmit_timers(struct sock *);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700465static inline void tcp_clear_xmit_timers(struct sock *sk)
466{
467 inet_csk_clear_xmit_timers(sk);
468}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
471extern unsigned int tcp_current_mss(struct sock *sk, int large);
472
Arnaldo Carvalho de Melo17b085e2005-08-12 12:59:17 -0300473/* tcp.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474extern void tcp_get_info(struct sock *, struct tcp_info *);
475
476/* Read 'sendfile()'-style from a TCP socket */
477typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
478 unsigned int, size_t);
479extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
480 sk_read_actor_t recv_actor);
481
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800482extern void tcp_initialize_rcv_mss(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
John Heffner5d424d52006-03-20 17:53:41 -0800484extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
485extern int tcp_mss_to_mtu(struct sock *sk, int mss);
486extern void tcp_mtup_init(struct sock *sk);
487
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800488static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
490 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
491 ntohl(TCP_FLAG_ACK) |
492 snd_wnd);
493}
494
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800495static inline void tcp_fast_path_on(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496{
497 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
498}
499
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700500static inline void tcp_fast_path_check(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700502 struct tcp_sock *tp = tcp_sk(sk);
503
David S. Millerb03efcf2005-07-08 14:57:23 -0700504 if (skb_queue_empty(&tp->out_of_order_queue) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 tp->rcv_wnd &&
506 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
507 !tp->urg_data)
508 tcp_fast_path_on(tp);
509}
510
511/* Compute the actual receive window we are currently advertising.
512 * Rcv_nxt can be after the window if our peer push more data
513 * than the offered window.
514 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800515static inline u32 tcp_receive_window(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516{
517 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
518
519 if (win < 0)
520 win = 0;
521 return (u32) win;
522}
523
524/* Choose a new window, without checks for shrinking, and without
525 * scaling applied to the result. The caller does these things
526 * if necessary. This is a "raw" window selection.
527 */
528extern u32 __tcp_select_window(struct sock *sk);
529
530/* TCP timestamps are only 32-bits, this causes a slight
531 * complication on 64-bit systems since we store a snapshot
Stephen Hemminger31f34262005-11-15 15:17:10 -0800532 * of jiffies in the buffer control blocks below. We decided
533 * to use only the low 32-bits of jiffies and hide the ugly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 * casts with the following macro.
535 */
536#define tcp_time_stamp ((__u32)(jiffies))
537
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800538/* This is what the send packet queuing engine uses to pass
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 * TCP per-packet control information to the transmission
540 * code. We also store the host-order sequence numbers in
541 * here too. This is 36 bytes on 32-bit architectures,
542 * 40 bytes on 64-bit machines, if this grows please adjust
543 * skbuff.h:skbuff->cb[xxx] size appropriately.
544 */
545struct tcp_skb_cb {
546 union {
547 struct inet_skb_parm h4;
548#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
549 struct inet6_skb_parm h6;
550#endif
551 } header; /* For incoming frames */
552 __u32 seq; /* Starting sequence number */
553 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
554 __u32 when; /* used to compute rtt's */
555 __u8 flags; /* TCP header flags. */
556
557 /* NOTE: These must match up to the flags byte in a
558 * real TCP header.
559 */
560#define TCPCB_FLAG_FIN 0x01
561#define TCPCB_FLAG_SYN 0x02
562#define TCPCB_FLAG_RST 0x04
563#define TCPCB_FLAG_PSH 0x08
564#define TCPCB_FLAG_ACK 0x10
565#define TCPCB_FLAG_URG 0x20
566#define TCPCB_FLAG_ECE 0x40
567#define TCPCB_FLAG_CWR 0x80
568
569 __u8 sacked; /* State flags for SACK/FACK. */
570#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
571#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
572#define TCPCB_LOST 0x04 /* SKB is lost */
573#define TCPCB_TAGBITS 0x07 /* All tag bits */
574
575#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
576#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
577
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800578#define TCPCB_URG 0x20 /* Urgent pointer advanced here */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579
580#define TCPCB_AT_TAIL (TCPCB_URG)
581
582 __u16 urg_ptr; /* Valid w/URG flags is set. */
583 __u32 ack_seq; /* Sequence number ACK'd */
584};
585
586#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
587
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588/* Due to TSO, an SKB can be composed of multiple actual
589 * packets. To keep these tracked properly, we use this.
590 */
591static inline int tcp_skb_pcount(const struct sk_buff *skb)
592{
Herbert Xu79671682006-06-22 02:40:14 -0700593 return skb_shinfo(skb)->gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594}
595
596/* This is valid iff tcp_skb_pcount() > 1. */
597static inline int tcp_skb_mss(const struct sk_buff *skb)
598{
Herbert Xu79671682006-06-22 02:40:14 -0700599 return skb_shinfo(skb)->gso_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600}
601
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300602static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603{
604 if (*count) {
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300605 *count -= decr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 if ((int)*count < 0)
607 *count = 0;
608 }
609}
610
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300611static inline void tcp_dec_pcount_approx(__u32 *count,
612 const struct sk_buff *skb)
613{
614 tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb));
615}
616
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700617/* Events passed to congestion control interface */
618enum tcp_ca_event {
619 CA_EVENT_TX_START, /* first transmit when no packets in flight */
620 CA_EVENT_CWND_RESTART, /* congestion window restart */
621 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
622 CA_EVENT_FRTO, /* fast recovery timeout */
623 CA_EVENT_LOSS, /* loss timeout */
624 CA_EVENT_FAST_ACK, /* in sequence ack */
625 CA_EVENT_SLOW_ACK, /* other ack */
626};
627
628/*
629 * Interface for adding new TCP congestion control handlers
630 */
631#define TCP_CA_NAME_MAX 16
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800632#define TCP_CA_MAX 128
633#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
634
Stephen Hemminger164891a2007-04-23 22:26:16 -0700635#define TCP_CONG_NON_RESTRICTED 0x1
636#define TCP_CONG_RTT_STAMP 0x2
637
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700638struct tcp_congestion_ops {
639 struct list_head list;
Stephen Hemminger164891a2007-04-23 22:26:16 -0700640 unsigned long flags;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700641
642 /* initialize private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300643 void (*init)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700644 /* cleanup private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300645 void (*release)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700646
647 /* return slow start threshold (required) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300648 u32 (*ssthresh)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700649 /* lower bound for congestion window (optional) */
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700650 u32 (*min_cwnd)(const struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700651 /* do new cwnd calculation (required) */
Stephen Hemminger16751342007-07-16 18:35:52 -0700652 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight, int good_ack);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700653 /* call before changing ca_state (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300654 void (*set_state)(struct sock *sk, u8 new_state);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700655 /* call when cwnd event occurs (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300656 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700657 /* new value of cwnd after loss (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300658 u32 (*undo_cwnd)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700659 /* hook for packet ack accounting (optional) */
Stephen Hemminger30cfd0b2007-07-25 23:49:34 -0700660 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
Arnaldo Carvalho de Melo73c1f4a2005-08-12 12:51:49 -0300661 /* get info for inet_diag (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300662 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700663
664 char name[TCP_CA_NAME_MAX];
665 struct module *owner;
666};
667
668extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
669extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
670
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300671extern void tcp_init_congestion_control(struct sock *sk);
672extern void tcp_cleanup_congestion_control(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700673extern int tcp_set_default_congestion_control(const char *name);
674extern void tcp_get_default_congestion_control(char *name);
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800675extern void tcp_get_available_congestion_control(char *buf, size_t len);
Stephen Hemmingerce7bc3b2006-11-09 16:35:15 -0800676extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
677extern int tcp_set_allowed_congestion_control(char *allowed);
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300678extern int tcp_set_congestion_control(struct sock *sk, const char *name);
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800679extern void tcp_slow_start(struct tcp_sock *tp);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700680
Stephen Hemminger5f8ef482005-06-23 20:37:36 -0700681extern struct tcp_congestion_ops tcp_init_congestion_ops;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300682extern u32 tcp_reno_ssthresh(struct sock *sk);
Stephen Hemminger16751342007-07-16 18:35:52 -0700683extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight, int flag);
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700684extern u32 tcp_reno_min_cwnd(const struct sock *sk);
David S. Millera8acfba2005-06-23 23:45:02 -0700685extern struct tcp_congestion_ops tcp_reno;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700686
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300687static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700688{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300689 struct inet_connection_sock *icsk = inet_csk(sk);
690
691 if (icsk->icsk_ca_ops->set_state)
692 icsk->icsk_ca_ops->set_state(sk, ca_state);
693 icsk->icsk_ca_state = ca_state;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700694}
695
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300696static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700697{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300698 const struct inet_connection_sock *icsk = inet_csk(sk);
699
700 if (icsk->icsk_ca_ops->cwnd_event)
701 icsk->icsk_ca_ops->cwnd_event(sk, event);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700702}
703
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300704/* These functions determine how the current flow behaves in respect of SACK
705 * handling. SACK is negotiated with the peer, and therefore it can vary
706 * between different flows.
707 *
708 * tcp_is_sack - SACK enabled
709 * tcp_is_reno - No SACK
710 * tcp_is_fack - FACK enabled, implies SACK enabled
711 */
712static inline int tcp_is_sack(const struct tcp_sock *tp)
713{
714 return tp->rx_opt.sack_ok;
715}
716
717static inline int tcp_is_reno(const struct tcp_sock *tp)
718{
719 return !tcp_is_sack(tp);
720}
721
722static inline int tcp_is_fack(const struct tcp_sock *tp)
723{
724 return tp->rx_opt.sack_ok & 2;
725}
726
727static inline void tcp_enable_fack(struct tcp_sock *tp)
728{
729 tp->rx_opt.sack_ok |= 2;
730}
731
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300732static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
733{
734 return tp->sacked_out + tp->lost_out;
735}
736
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737/* This determines how many packets are "in the network" to the best
738 * of our knowledge. In many cases it is conservative, but where
739 * detailed information is available from the receiver (via SACK
740 * blocks etc.) we can make more aggressive calculations.
741 *
742 * Use this for decisions involving congestion control, use just
743 * tp->packets_out to determine if the send queue is empty or not.
744 *
745 * Read this equation as:
746 *
747 * "Packets sent once on transmission queue" MINUS
748 * "Packets left network, but not honestly ACKed yet" PLUS
749 * "Packets fast retransmitted"
750 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800751static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752{
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300753 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754}
755
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
757 * The exception is rate halving phase, when cwnd is decreasing towards
758 * ssthresh.
759 */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300760static inline __u32 tcp_current_ssthresh(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300762 const struct tcp_sock *tp = tcp_sk(sk);
763 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 return tp->snd_ssthresh;
765 else
766 return max(tp->snd_ssthresh,
767 ((tp->snd_cwnd >> 1) +
768 (tp->snd_cwnd >> 2)));
769}
770
Ilpo Järvinenb9c45952007-07-27 16:36:17 +0300771/* Use define here intentionally to get WARN_ON location shown at the caller */
772#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800774extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
776
777/* Slow start with delack produces 3 packets of burst, so that
778 * it is safe "de facto".
779 */
780static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
781{
782 return 3;
783}
784
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800785/* RFC2861 Check whether we are limited by application or congestion window
786 * This is the inverse of cwnd check in tcp_tso_should_defer
787 */
788static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
789{
790 const struct tcp_sock *tp = tcp_sk(sk);
791 u32 left;
792
793 if (in_flight >= tp->snd_cwnd)
794 return 1;
795
Herbert Xubcd76112006-06-30 13:36:35 -0700796 if (!sk_can_gso(sk))
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800797 return 0;
798
799 left = tp->snd_cwnd - in_flight;
800 if (sysctl_tcp_tso_win_divisor)
801 return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd;
802 else
803 return left <= tcp_max_burst(tp);
804}
805
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800806static inline void tcp_minshall_update(struct tcp_sock *tp, int mss,
807 const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
809 if (skb->len < mss)
810 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
811}
812
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700813static inline void tcp_check_probe_timer(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700815 struct tcp_sock *tp = tcp_sk(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700816 const struct inet_connection_sock *icsk = inet_csk(sk);
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700817
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700818 if (!tp->packets_out && !icsk->icsk_pending)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700819 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
820 icsk->icsk_rto, TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821}
822
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700823static inline void tcp_push_pending_frames(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700825 struct tcp_sock *tp = tcp_sk(sk);
826
827 __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828}
829
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800830static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831{
832 tp->snd_wl1 = seq;
833}
834
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800835static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836{
837 tp->snd_wl1 = seq;
838}
839
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840/*
841 * Calculate(/check) TCP checksum
842 */
Frederik Deweerdtba7808e2007-02-04 20:15:27 -0800843static inline __sum16 tcp_v4_check(int len, __be32 saddr,
844 __be32 daddr, __wsum base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845{
846 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
847}
848
Al Virob51655b2006-11-14 21:40:42 -0800849static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850{
Herbert Xufb286bb2005-11-10 13:01:24 -0800851 return __skb_checksum_complete(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852}
853
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800854static inline int tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855{
Herbert Xu60476372007-04-09 11:59:39 -0700856 return !skb_csum_unnecessary(skb) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 __tcp_checksum_complete(skb);
858}
859
860/* Prequeue for VJ style copy to user, combined with checksumming. */
861
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800862static inline void tcp_prequeue_init(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863{
864 tp->ucopy.task = NULL;
865 tp->ucopy.len = 0;
866 tp->ucopy.memory = 0;
867 skb_queue_head_init(&tp->ucopy.prequeue);
Chris Leech97fc2f02006-05-23 17:55:33 -0700868#ifdef CONFIG_NET_DMA
869 tp->ucopy.dma_chan = NULL;
870 tp->ucopy.wakeup = 0;
871 tp->ucopy.pinned_list = NULL;
872 tp->ucopy.dma_cookie = 0;
873#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874}
875
876/* Packet is added to VJ-style prequeue for processing in process
877 * context, if a reader task is waiting. Apparently, this exciting
878 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
879 * failed somewhere. Latency? Burstiness? Well, at least now we will
880 * see, why it failed. 8)8) --ANK
881 *
882 * NOTE: is this not too big to inline?
883 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800884static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885{
886 struct tcp_sock *tp = tcp_sk(sk);
887
888 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
889 __skb_queue_tail(&tp->ucopy.prequeue, skb);
890 tp->ucopy.memory += skb->truesize;
891 if (tp->ucopy.memory > sk->sk_rcvbuf) {
892 struct sk_buff *skb1;
893
894 BUG_ON(sock_owned_by_user(sk));
895
896 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
897 sk->sk_backlog_rcv(sk, skb1);
898 NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED);
899 }
900
901 tp->ucopy.memory = 0;
902 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
903 wake_up_interruptible(sk->sk_sleep);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700904 if (!inet_csk_ack_scheduled(sk))
905 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700906 (3 * TCP_RTO_MIN) / 4,
907 TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 }
909 return 1;
910 }
911 return 0;
912}
913
914
915#undef STATE_TRACE
916
917#ifdef STATE_TRACE
918static const char *statename[]={
919 "Unused","Established","Syn Sent","Syn Recv",
920 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
921 "Close Wait","Last ACK","Listen","Closing"
922};
923#endif
924
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800925static inline void tcp_set_state(struct sock *sk, int state)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926{
927 int oldstate = sk->sk_state;
928
929 switch (state) {
930 case TCP_ESTABLISHED:
931 if (oldstate != TCP_ESTABLISHED)
932 TCP_INC_STATS(TCP_MIB_CURRESTAB);
933 break;
934
935 case TCP_CLOSE:
936 if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED)
937 TCP_INC_STATS(TCP_MIB_ESTABRESETS);
938
939 sk->sk_prot->unhash(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700940 if (inet_csk(sk)->icsk_bind_hash &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 !(sk->sk_userlocks & SOCK_BINDPORT_LOCK))
Arnaldo Carvalho de Melo2d8c4ce2005-08-09 20:07:13 -0700942 inet_put_port(&tcp_hashinfo, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 /* fall through */
944 default:
945 if (oldstate==TCP_ESTABLISHED)
946 TCP_DEC_STATS(TCP_MIB_CURRESTAB);
947 }
948
949 /* Change state AFTER socket is unhashed to avoid closed
950 * socket sitting in hash tables.
951 */
952 sk->sk_state = state;
953
954#ifdef STATE_TRACE
955 SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n",sk, statename[oldstate],statename[state]);
956#endif
957}
958
Andi Kleen4ac02ba2007-04-20 17:11:46 -0700959extern void tcp_done(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800961static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962{
963 rx_opt->dsack = 0;
964 rx_opt->eff_sacks = 0;
965 rx_opt->num_sacks = 0;
966}
967
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968/* Determine a window scaling and initial window to offer. */
969extern void tcp_select_initial_window(int __space, __u32 mss,
970 __u32 *rcv_wnd, __u32 *window_clamp,
971 int wscale_ok, __u8 *rcv_wscale);
972
973static inline int tcp_win_from_space(int space)
974{
975 return sysctl_tcp_adv_win_scale<=0 ?
976 (space>>(-sysctl_tcp_adv_win_scale)) :
977 space - (space>>sysctl_tcp_adv_win_scale);
978}
979
980/* Note: caller must be prepared to deal with negative returns */
981static inline int tcp_space(const struct sock *sk)
982{
983 return tcp_win_from_space(sk->sk_rcvbuf -
984 atomic_read(&sk->sk_rmem_alloc));
985}
986
987static inline int tcp_full_space(const struct sock *sk)
988{
989 return tcp_win_from_space(sk->sk_rcvbuf);
990}
991
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800992static inline void tcp_openreq_init(struct request_sock *req,
993 struct tcp_options_received *rx_opt,
994 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700996 struct inet_request_sock *ireq = inet_rsk(req);
997
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700999 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 req->mss = rx_opt->mss_clamp;
1001 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001002 ireq->tstamp_ok = rx_opt->tstamp_ok;
1003 ireq->sack_ok = rx_opt->sack_ok;
1004 ireq->snd_wscale = rx_opt->snd_wscale;
1005 ireq->wscale_ok = rx_opt->wscale_ok;
1006 ireq->acked = 0;
1007 ireq->ecn_ok = 0;
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001008 ireq->rmt_port = tcp_hdr(skb)->source;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009}
1010
1011extern void tcp_enter_memory_pressure(void);
1012
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1014{
1015 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1016}
1017
1018static inline int keepalive_time_when(const struct tcp_sock *tp)
1019{
1020 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1021}
1022
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001023static inline int tcp_fin_time(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001025 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1026 const int rto = inet_csk(sk)->icsk_rto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001028 if (fin_timeout < (rto << 2) - (rto >> 1))
1029 fin_timeout = (rto << 2) - (rto >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030
1031 return fin_timeout;
1032}
1033
1034static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1035{
1036 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1037 return 0;
James Morris9d729f72007-03-04 16:12:44 -08001038 if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 return 0;
1040
1041 /* RST segments are not recommended to carry timestamp,
1042 and, if they do, it is recommended to ignore PAWS because
1043 "their cleanup function should take precedence over timestamps."
1044 Certainly, it is mistake. It is necessary to understand the reasons
1045 of this constraint to relax it: if peer reboots, clock may go
1046 out-of-sync and half-open connections will not be reset.
1047 Actually, the problem would be not existing if all
1048 the implementations followed draft about maintaining clock
1049 via reboots. Linux-2.2 DOES NOT!
1050
1051 However, we can relax time bounds for RST segments to MSL.
1052 */
James Morris9d729f72007-03-04 16:12:44 -08001053 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054 return 0;
1055 return 1;
1056}
1057
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058#define TCP_CHECK_TIMER(sk) do { } while (0)
1059
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060static inline void tcp_mib_init(void)
1061{
1062 /* See RFC 2012 */
1063 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1064 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1065 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1066 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1067}
1068
Ilpo Järvinen5af4ec22007-09-20 11:30:48 -07001069/* from STCP */
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001070static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp) {
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001071 tp->lost_skb_hint = NULL;
1072 tp->scoreboard_skb_hint = NULL;
1073 tp->retransmit_skb_hint = NULL;
1074 tp->forward_skb_hint = NULL;
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001075}
1076
1077static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp) {
1078 tcp_clear_retrans_hints_partial(tp);
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001079 tp->fastpath_skb_hint = NULL;
1080}
1081
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001082/* MD5 Signature */
1083struct crypto_hash;
1084
1085/* - key database */
1086struct tcp_md5sig_key {
1087 u8 *key;
1088 u8 keylen;
1089};
1090
1091struct tcp4_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001092 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001093 __be32 addr;
1094};
1095
1096struct tcp6_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001097 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001098#if 0
1099 u32 scope_id; /* XXX */
1100#endif
1101 struct in6_addr addr;
1102};
1103
1104/* - sock block */
1105struct tcp_md5sig_info {
1106 struct tcp4_md5sig_key *keys4;
1107#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1108 struct tcp6_md5sig_key *keys6;
1109 u32 entries6;
1110 u32 alloced6;
1111#endif
1112 u32 entries4;
1113 u32 alloced4;
1114};
1115
1116/* - pseudo header */
1117struct tcp4_pseudohdr {
1118 __be32 saddr;
1119 __be32 daddr;
1120 __u8 pad;
1121 __u8 protocol;
1122 __be16 len;
1123};
1124
1125struct tcp6_pseudohdr {
1126 struct in6_addr saddr;
1127 struct in6_addr daddr;
1128 __be32 len;
1129 __be32 protocol; /* including padding */
1130};
1131
1132union tcp_md5sum_block {
1133 struct tcp4_pseudohdr ip4;
1134#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1135 struct tcp6_pseudohdr ip6;
1136#endif
1137};
1138
1139/* - pool: digest algorithm, hash description and scratch buffer */
1140struct tcp_md5sig_pool {
1141 struct hash_desc md5_desc;
1142 union tcp_md5sum_block md5_blk;
1143};
1144
1145#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1146
1147/* - functions */
1148extern int tcp_v4_calc_md5_hash(char *md5_hash,
1149 struct tcp_md5sig_key *key,
1150 struct sock *sk,
1151 struct dst_entry *dst,
1152 struct request_sock *req,
1153 struct tcphdr *th,
1154 int protocol, int tcplen);
1155extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1156 struct sock *addr_sk);
1157
1158extern int tcp_v4_md5_do_add(struct sock *sk,
1159 __be32 addr,
1160 u8 *newkey,
1161 u8 newkeylen);
1162
1163extern int tcp_v4_md5_do_del(struct sock *sk,
Al Viro8e5200f2006-11-20 18:06:37 -08001164 __be32 addr);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001165
1166extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1167extern void tcp_free_md5sig_pool(void);
1168
1169extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1170extern void __tcp_put_md5sig_pool(void);
1171
1172static inline
1173struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1174{
1175 int cpu = get_cpu();
1176 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1177 if (!ret)
1178 put_cpu();
1179 return ret;
1180}
1181
1182static inline void tcp_put_md5sig_pool(void)
1183{
1184 __tcp_put_md5sig_pool();
1185 put_cpu();
1186}
1187
David S. Millerfe067e82007-03-07 12:12:44 -08001188/* write queue abstraction */
1189static inline void tcp_write_queue_purge(struct sock *sk)
1190{
1191 struct sk_buff *skb;
1192
1193 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
1194 sk_stream_free_skb(sk, skb);
1195 sk_stream_mem_reclaim(sk);
1196}
1197
1198static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1199{
1200 struct sk_buff *skb = sk->sk_write_queue.next;
1201 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1202 return NULL;
1203 return skb;
1204}
1205
1206static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1207{
1208 struct sk_buff *skb = sk->sk_write_queue.prev;
1209 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1210 return NULL;
1211 return skb;
1212}
1213
1214static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1215{
1216 return skb->next;
1217}
1218
1219#define tcp_for_write_queue(skb, sk) \
1220 for (skb = (sk)->sk_write_queue.next; \
1221 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1222 skb = skb->next)
1223
1224#define tcp_for_write_queue_from(skb, sk) \
1225 for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\
1226 skb = skb->next)
1227
1228static inline struct sk_buff *tcp_send_head(struct sock *sk)
1229{
1230 return sk->sk_send_head;
1231}
1232
1233static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1234{
Ilpo Järvinend551e452007-04-30 00:42:20 -07001235 struct tcp_sock *tp = tcp_sk(sk);
1236
David S. Millerfe067e82007-03-07 12:12:44 -08001237 sk->sk_send_head = skb->next;
1238 if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue)
1239 sk->sk_send_head = NULL;
Ilpo Järvinend551e452007-04-30 00:42:20 -07001240 /* Don't override Nagle indefinately with F-RTO */
1241 if (tp->frto_counter == 2)
1242 tp->frto_counter = 3;
David S. Millerfe067e82007-03-07 12:12:44 -08001243}
1244
1245static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1246{
1247 if (sk->sk_send_head == skb_unlinked)
1248 sk->sk_send_head = NULL;
1249}
1250
1251static inline void tcp_init_send_head(struct sock *sk)
1252{
1253 sk->sk_send_head = NULL;
1254}
1255
1256static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1257{
1258 __skb_queue_tail(&sk->sk_write_queue, skb);
1259}
1260
1261static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1262{
1263 __tcp_add_write_queue_tail(sk, skb);
1264
1265 /* Queue it, remembering where we must start sending. */
1266 if (sk->sk_send_head == NULL)
1267 sk->sk_send_head = skb;
1268}
1269
1270static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1271{
1272 __skb_queue_head(&sk->sk_write_queue, skb);
1273}
1274
1275/* Insert buff after skb on the write queue of sk. */
1276static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1277 struct sk_buff *buff,
1278 struct sock *sk)
1279{
1280 __skb_append(skb, buff, &sk->sk_write_queue);
1281}
1282
1283/* Insert skb between prev and next on the write queue of sk. */
1284static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1285 struct sk_buff *skb,
1286 struct sock *sk)
1287{
1288 __skb_insert(new, skb->prev, skb, &sk->sk_write_queue);
1289}
1290
1291static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1292{
1293 __skb_unlink(skb, &sk->sk_write_queue);
1294}
1295
1296static inline int tcp_skb_is_last(const struct sock *sk,
1297 const struct sk_buff *skb)
1298{
1299 return skb->next == (struct sk_buff *)&sk->sk_write_queue;
1300}
1301
1302static inline int tcp_write_queue_empty(struct sock *sk)
1303{
1304 return skb_queue_empty(&sk->sk_write_queue);
1305}
1306
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307/* /proc */
1308enum tcp_seq_states {
1309 TCP_SEQ_STATE_LISTENING,
1310 TCP_SEQ_STATE_OPENREQ,
1311 TCP_SEQ_STATE_ESTABLISHED,
1312 TCP_SEQ_STATE_TIME_WAIT,
1313};
1314
1315struct tcp_seq_afinfo {
1316 struct module *owner;
1317 char *name;
1318 sa_family_t family;
1319 int (*seq_show) (struct seq_file *m, void *v);
1320 struct file_operations *seq_fops;
1321};
1322
1323struct tcp_iter_state {
1324 sa_family_t family;
1325 enum tcp_seq_states state;
1326 struct sock *syn_wait_sk;
1327 int bucket, sbucket, num, uid;
1328 struct seq_operations seq_ops;
1329};
1330
1331extern int tcp_proc_register(struct tcp_seq_afinfo *afinfo);
1332extern void tcp_proc_unregister(struct tcp_seq_afinfo *afinfo);
1333
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001334extern struct request_sock_ops tcp_request_sock_ops;
1335
1336extern int tcp_v4_destroy_sock(struct sock *sk);
1337
Herbert Xua430a432006-07-08 13:34:56 -07001338extern int tcp_v4_gso_send_check(struct sk_buff *skb);
Herbert Xu576a30e2006-06-27 13:22:38 -07001339extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001340
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001341#ifdef CONFIG_PROC_FS
1342extern int tcp4_proc_init(void);
1343extern void tcp4_proc_exit(void);
1344#endif
1345
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001346/* TCP af-specific functions */
1347struct tcp_sock_af_ops {
1348#ifdef CONFIG_TCP_MD5SIG
1349 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1350 struct sock *addr_sk);
1351 int (*calc_md5_hash) (char *location,
1352 struct tcp_md5sig_key *md5,
1353 struct sock *sk,
1354 struct dst_entry *dst,
1355 struct request_sock *req,
1356 struct tcphdr *th,
1357 int protocol, int len);
1358 int (*md5_add) (struct sock *sk,
1359 struct sock *addr_sk,
1360 u8 *newkey,
1361 u8 len);
1362 int (*md5_parse) (struct sock *sk,
1363 char __user *optval,
1364 int optlen);
1365#endif
1366};
1367
1368struct tcp_request_sock_ops {
1369#ifdef CONFIG_TCP_MD5SIG
1370 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1371 struct request_sock *req);
1372#endif
1373};
1374
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001375extern void tcp_v4_init(struct net_proto_family *ops);
1376extern void tcp_init(void);
1377
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378#endif /* _TCP_H */