blob: 76286e80205afa0ff4c6029a9e43412e189b3e0e [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
Jens Axboe9c55e012007-11-06 23:30:13 -0800312extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
313 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
314
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700315static inline void tcp_dec_quickack_mode(struct sock *sk,
316 const unsigned int pkts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700318 struct inet_connection_sock *icsk = inet_csk(sk);
David S. Millerfc6415bc2005-07-05 15:17:45 -0700319
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700320 if (icsk->icsk_ack.quick) {
321 if (pkts >= icsk->icsk_ack.quick) {
322 icsk->icsk_ack.quick = 0;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700323 /* Leaving quickack mode we deflate ATO. */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700324 icsk->icsk_ack.ato = TCP_ATO_MIN;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700325 } else
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700326 icsk->icsk_ack.quick -= pkts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 }
328}
329
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700330extern void tcp_enter_quickack_mode(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
333{
334 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
335}
336
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -0700337#define TCP_ECN_OK 1
338#define TCP_ECN_QUEUE_CWR 2
339#define TCP_ECN_DEMAND_CWR 4
340
341static __inline__ void
342TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
343{
344 if (sysctl_tcp_ecn && th->ece && th->cwr)
345 inet_rsk(req)->ecn_ok = 1;
346}
347
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348enum tcp_tw_status
349{
350 TCP_TW_SUCCESS = 0,
351 TCP_TW_RST = 1,
352 TCP_TW_ACK = 2,
353 TCP_TW_SYN = 3
354};
355
356
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700357extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 struct sk_buff *skb,
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700359 const struct tcphdr *th);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360
361extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700362 struct request_sock *req,
363 struct request_sock **prev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364extern int tcp_child_process(struct sock *parent,
365 struct sock *child,
366 struct sk_buff *skb);
Ilpo Järvinen46d0de42007-02-21 23:10:39 -0800367extern int tcp_use_frto(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368extern void tcp_enter_frto(struct sock *sk);
369extern void tcp_enter_loss(struct sock *sk, int how);
370extern void tcp_clear_retrans(struct tcp_sock *tp);
371extern void tcp_update_metrics(struct sock *sk);
372
373extern void tcp_close(struct sock *sk,
374 long timeout);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
376
377extern int tcp_getsockopt(struct sock *sk, int level,
378 int optname,
379 char __user *optval,
380 int __user *optlen);
381extern int tcp_setsockopt(struct sock *sk, int level,
382 int optname, char __user *optval,
383 int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800384extern int compat_tcp_getsockopt(struct sock *sk,
385 int level, int optname,
386 char __user *optval, int __user *optlen);
387extern int compat_tcp_setsockopt(struct sock *sk,
388 int level, int optname,
389 char __user *optval, int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390extern void tcp_set_keepalive(struct sock *sk, int val);
391extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
392 struct msghdr *msg,
393 size_t len, int nonblock,
394 int flags, int *addr_len);
395
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396extern void tcp_parse_options(struct sk_buff *skb,
397 struct tcp_options_received *opt_rx,
398 int estab);
399
400/*
401 * TCP v4 functions exported for the inet6 API
402 */
403
Arnaldo Carvalho de Melo8292a172005-12-13 23:15:52 -0800404extern void tcp_v4_send_check(struct sock *sk, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 struct sk_buff *skb);
406
407extern int tcp_v4_conn_request(struct sock *sk,
408 struct sk_buff *skb);
409
410extern struct sock * tcp_create_openreq_child(struct sock *sk,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700411 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 struct sk_buff *skb);
413
414extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
415 struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700416 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 struct dst_entry *dst);
418
419extern int tcp_v4_do_rcv(struct sock *sk,
420 struct sk_buff *skb);
421
422extern int tcp_v4_connect(struct sock *sk,
423 struct sockaddr *uaddr,
424 int addr_len);
425
426extern int tcp_connect(struct sock *sk);
427
428extern struct sk_buff * tcp_make_synack(struct sock *sk,
429 struct dst_entry *dst,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700430 struct request_sock *req);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431
432extern int tcp_disconnect(struct sock *sk, int flags);
433
434extern void tcp_unhash(struct sock *sk);
435
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436/* From syncookies.c */
437extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
438 struct ip_options *opt);
439extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
440 __u16 *mss);
441
442/* tcp_output.c */
443
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700444extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
445 int nonagle);
446extern int tcp_may_send_now(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
448extern void tcp_xmit_retransmit_queue(struct sock *);
449extern void tcp_simple_retransmit(struct sock *);
450extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
David S. Miller6475be12005-09-01 22:47:01 -0700451extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452
453extern void tcp_send_probe0(struct sock *);
454extern void tcp_send_partial(struct sock *);
455extern int tcp_write_wakeup(struct sock *);
456extern void tcp_send_fin(struct sock *sk);
Al Virodd0fc662005-10-07 07:46:04 +0100457extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458extern int tcp_send_synack(struct sock *);
David S. Millerc1b4a7e2005-07-05 15:24:38 -0700459extern void tcp_push_one(struct sock *, unsigned int mss_now);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460extern void tcp_send_ack(struct sock *sk);
461extern void tcp_send_delayed_ack(struct sock *sk);
462
David S. Millera762a982005-07-05 15:18:51 -0700463/* tcp_input.c */
464extern void tcp_cwnd_application_limited(struct sock *sk);
465
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466/* tcp_timer.c */
467extern void tcp_init_xmit_timers(struct sock *);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700468static inline void tcp_clear_xmit_timers(struct sock *sk)
469{
470 inet_csk_clear_xmit_timers(sk);
471}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
474extern unsigned int tcp_current_mss(struct sock *sk, int large);
475
Arnaldo Carvalho de Melo17b085e2005-08-12 12:59:17 -0300476/* tcp.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477extern void tcp_get_info(struct sock *, struct tcp_info *);
478
479/* Read 'sendfile()'-style from a TCP socket */
480typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
481 unsigned int, size_t);
482extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
483 sk_read_actor_t recv_actor);
484
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800485extern void tcp_initialize_rcv_mss(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486
John Heffner5d424d52006-03-20 17:53:41 -0800487extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
488extern int tcp_mss_to_mtu(struct sock *sk, int mss);
489extern void tcp_mtup_init(struct sock *sk);
490
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800491static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
493 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
494 ntohl(TCP_FLAG_ACK) |
495 snd_wnd);
496}
497
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800498static inline void tcp_fast_path_on(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499{
500 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
501}
502
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700503static inline void tcp_fast_path_check(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700505 struct tcp_sock *tp = tcp_sk(sk);
506
David S. Millerb03efcf2005-07-08 14:57:23 -0700507 if (skb_queue_empty(&tp->out_of_order_queue) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 tp->rcv_wnd &&
509 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
510 !tp->urg_data)
511 tcp_fast_path_on(tp);
512}
513
514/* Compute the actual receive window we are currently advertising.
515 * Rcv_nxt can be after the window if our peer push more data
516 * than the offered window.
517 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800518static inline u32 tcp_receive_window(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519{
520 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
521
522 if (win < 0)
523 win = 0;
524 return (u32) win;
525}
526
527/* Choose a new window, without checks for shrinking, and without
528 * scaling applied to the result. The caller does these things
529 * if necessary. This is a "raw" window selection.
530 */
531extern u32 __tcp_select_window(struct sock *sk);
532
533/* TCP timestamps are only 32-bits, this causes a slight
534 * complication on 64-bit systems since we store a snapshot
Stephen Hemminger31f34262005-11-15 15:17:10 -0800535 * of jiffies in the buffer control blocks below. We decided
536 * to use only the low 32-bits of jiffies and hide the ugly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 * casts with the following macro.
538 */
539#define tcp_time_stamp ((__u32)(jiffies))
540
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800541/* This is what the send packet queuing engine uses to pass
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 * TCP per-packet control information to the transmission
543 * code. We also store the host-order sequence numbers in
544 * here too. This is 36 bytes on 32-bit architectures,
545 * 40 bytes on 64-bit machines, if this grows please adjust
546 * skbuff.h:skbuff->cb[xxx] size appropriately.
547 */
548struct tcp_skb_cb {
549 union {
550 struct inet_skb_parm h4;
551#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
552 struct inet6_skb_parm h6;
553#endif
554 } header; /* For incoming frames */
555 __u32 seq; /* Starting sequence number */
556 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
557 __u32 when; /* used to compute rtt's */
558 __u8 flags; /* TCP header flags. */
559
560 /* NOTE: These must match up to the flags byte in a
561 * real TCP header.
562 */
563#define TCPCB_FLAG_FIN 0x01
564#define TCPCB_FLAG_SYN 0x02
565#define TCPCB_FLAG_RST 0x04
566#define TCPCB_FLAG_PSH 0x08
567#define TCPCB_FLAG_ACK 0x10
568#define TCPCB_FLAG_URG 0x20
569#define TCPCB_FLAG_ECE 0x40
570#define TCPCB_FLAG_CWR 0x80
571
572 __u8 sacked; /* State flags for SACK/FACK. */
573#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
574#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
575#define TCPCB_LOST 0x04 /* SKB is lost */
576#define TCPCB_TAGBITS 0x07 /* All tag bits */
577
578#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
579#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
580
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800581#define TCPCB_URG 0x20 /* Urgent pointer advanced here */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
583#define TCPCB_AT_TAIL (TCPCB_URG)
584
585 __u16 urg_ptr; /* Valid w/URG flags is set. */
586 __u32 ack_seq; /* Sequence number ACK'd */
587};
588
589#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
590
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591/* Due to TSO, an SKB can be composed of multiple actual
592 * packets. To keep these tracked properly, we use this.
593 */
594static inline int tcp_skb_pcount(const struct sk_buff *skb)
595{
Herbert Xu79671682006-06-22 02:40:14 -0700596 return skb_shinfo(skb)->gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597}
598
599/* This is valid iff tcp_skb_pcount() > 1. */
600static inline int tcp_skb_mss(const struct sk_buff *skb)
601{
Herbert Xu79671682006-06-22 02:40:14 -0700602 return skb_shinfo(skb)->gso_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603}
604
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300605static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606{
607 if (*count) {
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300608 *count -= decr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 if ((int)*count < 0)
610 *count = 0;
611 }
612}
613
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300614static inline void tcp_dec_pcount_approx(__u32 *count,
615 const struct sk_buff *skb)
616{
617 tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb));
618}
619
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700620/* Events passed to congestion control interface */
621enum tcp_ca_event {
622 CA_EVENT_TX_START, /* first transmit when no packets in flight */
623 CA_EVENT_CWND_RESTART, /* congestion window restart */
624 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
625 CA_EVENT_FRTO, /* fast recovery timeout */
626 CA_EVENT_LOSS, /* loss timeout */
627 CA_EVENT_FAST_ACK, /* in sequence ack */
628 CA_EVENT_SLOW_ACK, /* other ack */
629};
630
631/*
632 * Interface for adding new TCP congestion control handlers
633 */
634#define TCP_CA_NAME_MAX 16
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800635#define TCP_CA_MAX 128
636#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
637
Stephen Hemminger164891a2007-04-23 22:26:16 -0700638#define TCP_CONG_NON_RESTRICTED 0x1
639#define TCP_CONG_RTT_STAMP 0x2
640
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700641struct tcp_congestion_ops {
642 struct list_head list;
Stephen Hemminger164891a2007-04-23 22:26:16 -0700643 unsigned long flags;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700644
645 /* initialize private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300646 void (*init)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700647 /* cleanup private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300648 void (*release)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700649
650 /* return slow start threshold (required) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300651 u32 (*ssthresh)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700652 /* lower bound for congestion window (optional) */
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700653 u32 (*min_cwnd)(const struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700654 /* do new cwnd calculation (required) */
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200655 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700656 /* call before changing ca_state (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300657 void (*set_state)(struct sock *sk, u8 new_state);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700658 /* call when cwnd event occurs (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300659 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700660 /* new value of cwnd after loss (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300661 u32 (*undo_cwnd)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700662 /* hook for packet ack accounting (optional) */
Stephen Hemminger30cfd0b2007-07-25 23:49:34 -0700663 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
Arnaldo Carvalho de Melo73c1f4a2005-08-12 12:51:49 -0300664 /* get info for inet_diag (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300665 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700666
667 char name[TCP_CA_NAME_MAX];
668 struct module *owner;
669};
670
671extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
672extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
673
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300674extern void tcp_init_congestion_control(struct sock *sk);
675extern void tcp_cleanup_congestion_control(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700676extern int tcp_set_default_congestion_control(const char *name);
677extern void tcp_get_default_congestion_control(char *name);
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800678extern void tcp_get_available_congestion_control(char *buf, size_t len);
Stephen Hemmingerce7bc3b2006-11-09 16:35:15 -0800679extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
680extern int tcp_set_allowed_congestion_control(char *allowed);
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300681extern int tcp_set_congestion_control(struct sock *sk, const char *name);
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800682extern void tcp_slow_start(struct tcp_sock *tp);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700683
Stephen Hemminger5f8ef482005-06-23 20:37:36 -0700684extern struct tcp_congestion_ops tcp_init_congestion_ops;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300685extern u32 tcp_reno_ssthresh(struct sock *sk);
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200686extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700687extern u32 tcp_reno_min_cwnd(const struct sock *sk);
David S. Millera8acfba2005-06-23 23:45:02 -0700688extern struct tcp_congestion_ops tcp_reno;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700689
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300690static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700691{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300692 struct inet_connection_sock *icsk = inet_csk(sk);
693
694 if (icsk->icsk_ca_ops->set_state)
695 icsk->icsk_ca_ops->set_state(sk, ca_state);
696 icsk->icsk_ca_state = ca_state;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700697}
698
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300699static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700700{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300701 const struct inet_connection_sock *icsk = inet_csk(sk);
702
703 if (icsk->icsk_ca_ops->cwnd_event)
704 icsk->icsk_ca_ops->cwnd_event(sk, event);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700705}
706
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300707/* These functions determine how the current flow behaves in respect of SACK
708 * handling. SACK is negotiated with the peer, and therefore it can vary
709 * between different flows.
710 *
711 * tcp_is_sack - SACK enabled
712 * tcp_is_reno - No SACK
713 * tcp_is_fack - FACK enabled, implies SACK enabled
714 */
715static inline int tcp_is_sack(const struct tcp_sock *tp)
716{
717 return tp->rx_opt.sack_ok;
718}
719
720static inline int tcp_is_reno(const struct tcp_sock *tp)
721{
722 return !tcp_is_sack(tp);
723}
724
725static inline int tcp_is_fack(const struct tcp_sock *tp)
726{
727 return tp->rx_opt.sack_ok & 2;
728}
729
730static inline void tcp_enable_fack(struct tcp_sock *tp)
731{
732 tp->rx_opt.sack_ok |= 2;
733}
734
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300735static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
736{
737 return tp->sacked_out + tp->lost_out;
738}
739
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740/* This determines how many packets are "in the network" to the best
741 * of our knowledge. In many cases it is conservative, but where
742 * detailed information is available from the receiver (via SACK
743 * blocks etc.) we can make more aggressive calculations.
744 *
745 * Use this for decisions involving congestion control, use just
746 * tp->packets_out to determine if the send queue is empty or not.
747 *
748 * Read this equation as:
749 *
750 * "Packets sent once on transmission queue" MINUS
751 * "Packets left network, but not honestly ACKed yet" PLUS
752 * "Packets fast retransmitted"
753 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800754static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755{
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300756 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757}
758
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
760 * The exception is rate halving phase, when cwnd is decreasing towards
761 * ssthresh.
762 */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300763static inline __u32 tcp_current_ssthresh(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300765 const struct tcp_sock *tp = tcp_sk(sk);
766 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 return tp->snd_ssthresh;
768 else
769 return max(tp->snd_ssthresh,
770 ((tp->snd_cwnd >> 1) +
771 (tp->snd_cwnd >> 2)));
772}
773
Ilpo Järvinenb9c45952007-07-27 16:36:17 +0300774/* Use define here intentionally to get WARN_ON location shown at the caller */
775#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800777extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
779
780/* Slow start with delack produces 3 packets of burst, so that
781 * it is safe "de facto".
782 */
783static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
784{
785 return 3;
786}
787
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800788/* RFC2861 Check whether we are limited by application or congestion window
789 * This is the inverse of cwnd check in tcp_tso_should_defer
790 */
791static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
792{
793 const struct tcp_sock *tp = tcp_sk(sk);
794 u32 left;
795
796 if (in_flight >= tp->snd_cwnd)
797 return 1;
798
Herbert Xubcd76112006-06-30 13:36:35 -0700799 if (!sk_can_gso(sk))
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800800 return 0;
801
802 left = tp->snd_cwnd - in_flight;
803 if (sysctl_tcp_tso_win_divisor)
804 return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd;
805 else
806 return left <= tcp_max_burst(tp);
807}
808
Chuck Leverc1bd24b72007-10-23 21:08:54 -0700809static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800810 const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811{
812 if (skb->len < mss)
813 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
814}
815
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700816static inline void tcp_check_probe_timer(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700818 struct tcp_sock *tp = tcp_sk(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700819 const struct inet_connection_sock *icsk = inet_csk(sk);
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700820
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700821 if (!tp->packets_out && !icsk->icsk_pending)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700822 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
823 icsk->icsk_rto, TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824}
825
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700826static inline void tcp_push_pending_frames(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700828 struct tcp_sock *tp = tcp_sk(sk);
829
830 __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831}
832
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800833static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834{
835 tp->snd_wl1 = seq;
836}
837
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800838static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839{
840 tp->snd_wl1 = seq;
841}
842
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843/*
844 * Calculate(/check) TCP checksum
845 */
Frederik Deweerdtba7808e2007-02-04 20:15:27 -0800846static inline __sum16 tcp_v4_check(int len, __be32 saddr,
847 __be32 daddr, __wsum base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848{
849 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
850}
851
Al Virob51655b2006-11-14 21:40:42 -0800852static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853{
Herbert Xufb286bb2005-11-10 13:01:24 -0800854 return __skb_checksum_complete(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855}
856
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800857static inline int tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858{
Herbert Xu60476372007-04-09 11:59:39 -0700859 return !skb_csum_unnecessary(skb) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 __tcp_checksum_complete(skb);
861}
862
863/* Prequeue for VJ style copy to user, combined with checksumming. */
864
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800865static inline void tcp_prequeue_init(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866{
867 tp->ucopy.task = NULL;
868 tp->ucopy.len = 0;
869 tp->ucopy.memory = 0;
870 skb_queue_head_init(&tp->ucopy.prequeue);
Chris Leech97fc2f02006-05-23 17:55:33 -0700871#ifdef CONFIG_NET_DMA
872 tp->ucopy.dma_chan = NULL;
873 tp->ucopy.wakeup = 0;
874 tp->ucopy.pinned_list = NULL;
875 tp->ucopy.dma_cookie = 0;
876#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877}
878
879/* Packet is added to VJ-style prequeue for processing in process
880 * context, if a reader task is waiting. Apparently, this exciting
881 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
882 * failed somewhere. Latency? Burstiness? Well, at least now we will
883 * see, why it failed. 8)8) --ANK
884 *
885 * NOTE: is this not too big to inline?
886 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800887static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888{
889 struct tcp_sock *tp = tcp_sk(sk);
890
891 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
892 __skb_queue_tail(&tp->ucopy.prequeue, skb);
893 tp->ucopy.memory += skb->truesize;
894 if (tp->ucopy.memory > sk->sk_rcvbuf) {
895 struct sk_buff *skb1;
896
897 BUG_ON(sock_owned_by_user(sk));
898
899 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
900 sk->sk_backlog_rcv(sk, skb1);
901 NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED);
902 }
903
904 tp->ucopy.memory = 0;
905 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
906 wake_up_interruptible(sk->sk_sleep);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700907 if (!inet_csk_ack_scheduled(sk))
908 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700909 (3 * TCP_RTO_MIN) / 4,
910 TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 }
912 return 1;
913 }
914 return 0;
915}
916
917
918#undef STATE_TRACE
919
920#ifdef STATE_TRACE
921static const char *statename[]={
922 "Unused","Established","Syn Sent","Syn Recv",
923 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
924 "Close Wait","Last ACK","Listen","Closing"
925};
926#endif
927
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800928static inline void tcp_set_state(struct sock *sk, int state)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929{
930 int oldstate = sk->sk_state;
931
932 switch (state) {
933 case TCP_ESTABLISHED:
934 if (oldstate != TCP_ESTABLISHED)
935 TCP_INC_STATS(TCP_MIB_CURRESTAB);
936 break;
937
938 case TCP_CLOSE:
939 if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED)
940 TCP_INC_STATS(TCP_MIB_ESTABRESETS);
941
942 sk->sk_prot->unhash(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700943 if (inet_csk(sk)->icsk_bind_hash &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 !(sk->sk_userlocks & SOCK_BINDPORT_LOCK))
Arnaldo Carvalho de Melo2d8c4ce2005-08-09 20:07:13 -0700945 inet_put_port(&tcp_hashinfo, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 /* fall through */
947 default:
948 if (oldstate==TCP_ESTABLISHED)
949 TCP_DEC_STATS(TCP_MIB_CURRESTAB);
950 }
951
952 /* Change state AFTER socket is unhashed to avoid closed
953 * socket sitting in hash tables.
954 */
955 sk->sk_state = state;
956
957#ifdef STATE_TRACE
958 SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n",sk, statename[oldstate],statename[state]);
959#endif
960}
961
Andi Kleen4ac02ba2007-04-20 17:11:46 -0700962extern void tcp_done(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800964static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965{
966 rx_opt->dsack = 0;
967 rx_opt->eff_sacks = 0;
968 rx_opt->num_sacks = 0;
969}
970
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971/* Determine a window scaling and initial window to offer. */
972extern void tcp_select_initial_window(int __space, __u32 mss,
973 __u32 *rcv_wnd, __u32 *window_clamp,
974 int wscale_ok, __u8 *rcv_wscale);
975
976static inline int tcp_win_from_space(int space)
977{
978 return sysctl_tcp_adv_win_scale<=0 ?
979 (space>>(-sysctl_tcp_adv_win_scale)) :
980 space - (space>>sysctl_tcp_adv_win_scale);
981}
982
983/* Note: caller must be prepared to deal with negative returns */
984static inline int tcp_space(const struct sock *sk)
985{
986 return tcp_win_from_space(sk->sk_rcvbuf -
987 atomic_read(&sk->sk_rmem_alloc));
988}
989
990static inline int tcp_full_space(const struct sock *sk)
991{
992 return tcp_win_from_space(sk->sk_rcvbuf);
993}
994
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800995static inline void tcp_openreq_init(struct request_sock *req,
996 struct tcp_options_received *rx_opt,
997 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700999 struct inet_request_sock *ireq = inet_rsk(req);
1000
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001002 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 req->mss = rx_opt->mss_clamp;
1004 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001005 ireq->tstamp_ok = rx_opt->tstamp_ok;
1006 ireq->sack_ok = rx_opt->sack_ok;
1007 ireq->snd_wscale = rx_opt->snd_wscale;
1008 ireq->wscale_ok = rx_opt->wscale_ok;
1009 ireq->acked = 0;
1010 ireq->ecn_ok = 0;
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001011 ireq->rmt_port = tcp_hdr(skb)->source;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012}
1013
1014extern void tcp_enter_memory_pressure(void);
1015
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1017{
1018 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1019}
1020
1021static inline int keepalive_time_when(const struct tcp_sock *tp)
1022{
1023 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1024}
1025
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001026static inline int tcp_fin_time(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001028 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1029 const int rto = inet_csk(sk)->icsk_rto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001031 if (fin_timeout < (rto << 2) - (rto >> 1))
1032 fin_timeout = (rto << 2) - (rto >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033
1034 return fin_timeout;
1035}
1036
1037static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1038{
1039 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1040 return 0;
James Morris9d729f72007-03-04 16:12:44 -08001041 if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 return 0;
1043
1044 /* RST segments are not recommended to carry timestamp,
1045 and, if they do, it is recommended to ignore PAWS because
1046 "their cleanup function should take precedence over timestamps."
1047 Certainly, it is mistake. It is necessary to understand the reasons
1048 of this constraint to relax it: if peer reboots, clock may go
1049 out-of-sync and half-open connections will not be reset.
1050 Actually, the problem would be not existing if all
1051 the implementations followed draft about maintaining clock
1052 via reboots. Linux-2.2 DOES NOT!
1053
1054 However, we can relax time bounds for RST segments to MSL.
1055 */
James Morris9d729f72007-03-04 16:12:44 -08001056 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 return 0;
1058 return 1;
1059}
1060
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061#define TCP_CHECK_TIMER(sk) do { } while (0)
1062
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063static inline void tcp_mib_init(void)
1064{
1065 /* See RFC 2012 */
1066 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1067 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1068 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1069 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1070}
1071
Ilpo Järvinen5af4ec22007-09-20 11:30:48 -07001072/* from STCP */
David S. Miller0800f172007-09-20 11:40:37 -07001073static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1074{
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001075 tp->lost_skb_hint = NULL;
1076 tp->scoreboard_skb_hint = NULL;
1077 tp->retransmit_skb_hint = NULL;
1078 tp->forward_skb_hint = NULL;
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001079}
1080
David S. Miller0800f172007-09-20 11:40:37 -07001081static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1082{
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001083 tcp_clear_retrans_hints_partial(tp);
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001084}
1085
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001086/* MD5 Signature */
1087struct crypto_hash;
1088
1089/* - key database */
1090struct tcp_md5sig_key {
1091 u8 *key;
1092 u8 keylen;
1093};
1094
1095struct tcp4_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001096 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001097 __be32 addr;
1098};
1099
1100struct tcp6_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001101 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001102#if 0
1103 u32 scope_id; /* XXX */
1104#endif
1105 struct in6_addr addr;
1106};
1107
1108/* - sock block */
1109struct tcp_md5sig_info {
1110 struct tcp4_md5sig_key *keys4;
1111#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1112 struct tcp6_md5sig_key *keys6;
1113 u32 entries6;
1114 u32 alloced6;
1115#endif
1116 u32 entries4;
1117 u32 alloced4;
1118};
1119
1120/* - pseudo header */
1121struct tcp4_pseudohdr {
1122 __be32 saddr;
1123 __be32 daddr;
1124 __u8 pad;
1125 __u8 protocol;
1126 __be16 len;
1127};
1128
1129struct tcp6_pseudohdr {
1130 struct in6_addr saddr;
1131 struct in6_addr daddr;
1132 __be32 len;
1133 __be32 protocol; /* including padding */
1134};
1135
1136union tcp_md5sum_block {
1137 struct tcp4_pseudohdr ip4;
1138#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1139 struct tcp6_pseudohdr ip6;
1140#endif
1141};
1142
1143/* - pool: digest algorithm, hash description and scratch buffer */
1144struct tcp_md5sig_pool {
1145 struct hash_desc md5_desc;
1146 union tcp_md5sum_block md5_blk;
1147};
1148
1149#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1150
1151/* - functions */
1152extern int tcp_v4_calc_md5_hash(char *md5_hash,
1153 struct tcp_md5sig_key *key,
1154 struct sock *sk,
1155 struct dst_entry *dst,
1156 struct request_sock *req,
1157 struct tcphdr *th,
YOSHIFUJI Hideaki9cb5734e2008-01-12 02:16:03 -08001158 int protocol,
1159 unsigned int tcplen);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001160extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1161 struct sock *addr_sk);
1162
1163extern int tcp_v4_md5_do_add(struct sock *sk,
1164 __be32 addr,
1165 u8 *newkey,
1166 u8 newkeylen);
1167
1168extern int tcp_v4_md5_do_del(struct sock *sk,
Al Viro8e5200f2006-11-20 18:06:37 -08001169 __be32 addr);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001170
1171extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1172extern void tcp_free_md5sig_pool(void);
1173
1174extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1175extern void __tcp_put_md5sig_pool(void);
1176
1177static inline
1178struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1179{
1180 int cpu = get_cpu();
1181 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1182 if (!ret)
1183 put_cpu();
1184 return ret;
1185}
1186
1187static inline void tcp_put_md5sig_pool(void)
1188{
1189 __tcp_put_md5sig_pool();
1190 put_cpu();
1191}
1192
David S. Millerfe067e82007-03-07 12:12:44 -08001193/* write queue abstraction */
1194static inline void tcp_write_queue_purge(struct sock *sk)
1195{
1196 struct sk_buff *skb;
1197
1198 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001199 sk_wmem_free_skb(sk, skb);
1200 sk_mem_reclaim(sk);
David S. Millerfe067e82007-03-07 12:12:44 -08001201}
1202
1203static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1204{
1205 struct sk_buff *skb = sk->sk_write_queue.next;
1206 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1207 return NULL;
1208 return skb;
1209}
1210
1211static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1212{
1213 struct sk_buff *skb = sk->sk_write_queue.prev;
1214 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1215 return NULL;
1216 return skb;
1217}
1218
1219static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1220{
1221 return skb->next;
1222}
1223
1224#define tcp_for_write_queue(skb, sk) \
1225 for (skb = (sk)->sk_write_queue.next; \
1226 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1227 skb = skb->next)
1228
1229#define tcp_for_write_queue_from(skb, sk) \
1230 for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\
1231 skb = skb->next)
1232
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001233#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
1234 for (tmp = skb->next; \
1235 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1236 skb = tmp, tmp = skb->next)
1237
David S. Millerfe067e82007-03-07 12:12:44 -08001238static inline struct sk_buff *tcp_send_head(struct sock *sk)
1239{
1240 return sk->sk_send_head;
1241}
1242
1243static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1244{
1245 sk->sk_send_head = skb->next;
1246 if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue)
1247 sk->sk_send_head = NULL;
1248}
1249
1250static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1251{
1252 if (sk->sk_send_head == skb_unlinked)
1253 sk->sk_send_head = NULL;
1254}
1255
1256static inline void tcp_init_send_head(struct sock *sk)
1257{
1258 sk->sk_send_head = NULL;
1259}
1260
1261static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1262{
1263 __skb_queue_tail(&sk->sk_write_queue, skb);
1264}
1265
1266static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1267{
1268 __tcp_add_write_queue_tail(sk, skb);
1269
1270 /* Queue it, remembering where we must start sending. */
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001271 if (sk->sk_send_head == NULL) {
David S. Millerfe067e82007-03-07 12:12:44 -08001272 sk->sk_send_head = skb;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001273
1274 if (tcp_sk(sk)->highest_sack == NULL)
1275 tcp_sk(sk)->highest_sack = skb;
1276 }
David S. Millerfe067e82007-03-07 12:12:44 -08001277}
1278
1279static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1280{
1281 __skb_queue_head(&sk->sk_write_queue, skb);
1282}
1283
1284/* Insert buff after skb on the write queue of sk. */
1285static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1286 struct sk_buff *buff,
1287 struct sock *sk)
1288{
1289 __skb_append(skb, buff, &sk->sk_write_queue);
1290}
1291
1292/* Insert skb between prev and next on the write queue of sk. */
1293static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1294 struct sk_buff *skb,
1295 struct sock *sk)
1296{
1297 __skb_insert(new, skb->prev, skb, &sk->sk_write_queue);
Ilpo Järvinen6e421412007-11-19 23:24:09 -08001298
1299 if (sk->sk_send_head == skb)
1300 sk->sk_send_head = new;
David S. Millerfe067e82007-03-07 12:12:44 -08001301}
1302
1303static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1304{
1305 __skb_unlink(skb, &sk->sk_write_queue);
1306}
1307
1308static inline int tcp_skb_is_last(const struct sock *sk,
1309 const struct sk_buff *skb)
1310{
1311 return skb->next == (struct sk_buff *)&sk->sk_write_queue;
1312}
1313
1314static inline int tcp_write_queue_empty(struct sock *sk)
1315{
1316 return skb_queue_empty(&sk->sk_write_queue);
1317}
1318
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001319/* Start sequence of the highest skb with SACKed bit, valid only if
1320 * sacked > 0 or when the caller has ensured validity by itself.
1321 */
1322static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1323{
1324 if (!tp->sacked_out)
1325 return tp->snd_una;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001326
1327 if (tp->highest_sack == NULL)
1328 return tp->snd_nxt;
1329
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001330 return TCP_SKB_CB(tp->highest_sack)->seq;
1331}
1332
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001333static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1334{
1335 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1336 tcp_write_queue_next(sk, skb);
1337}
1338
1339static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1340{
1341 return tcp_sk(sk)->highest_sack;
1342}
1343
1344static inline void tcp_highest_sack_reset(struct sock *sk)
1345{
1346 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1347}
1348
1349/* Called when old skb is about to be deleted (to be combined with new skb) */
1350static inline void tcp_highest_sack_combine(struct sock *sk,
1351 struct sk_buff *old,
1352 struct sk_buff *new)
1353{
1354 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1355 tcp_sk(sk)->highest_sack = new;
1356}
1357
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358/* /proc */
1359enum tcp_seq_states {
1360 TCP_SEQ_STATE_LISTENING,
1361 TCP_SEQ_STATE_OPENREQ,
1362 TCP_SEQ_STATE_ESTABLISHED,
1363 TCP_SEQ_STATE_TIME_WAIT,
1364};
1365
1366struct tcp_seq_afinfo {
1367 struct module *owner;
1368 char *name;
1369 sa_family_t family;
1370 int (*seq_show) (struct seq_file *m, void *v);
1371 struct file_operations *seq_fops;
1372};
1373
1374struct tcp_iter_state {
1375 sa_family_t family;
1376 enum tcp_seq_states state;
1377 struct sock *syn_wait_sk;
1378 int bucket, sbucket, num, uid;
1379 struct seq_operations seq_ops;
1380};
1381
1382extern int tcp_proc_register(struct tcp_seq_afinfo *afinfo);
1383extern void tcp_proc_unregister(struct tcp_seq_afinfo *afinfo);
1384
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001385extern struct request_sock_ops tcp_request_sock_ops;
1386
1387extern int tcp_v4_destroy_sock(struct sock *sk);
1388
Herbert Xua430a432006-07-08 13:34:56 -07001389extern int tcp_v4_gso_send_check(struct sk_buff *skb);
Herbert Xu576a30e2006-06-27 13:22:38 -07001390extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001391
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001392#ifdef CONFIG_PROC_FS
1393extern int tcp4_proc_init(void);
1394extern void tcp4_proc_exit(void);
1395#endif
1396
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001397/* TCP af-specific functions */
1398struct tcp_sock_af_ops {
1399#ifdef CONFIG_TCP_MD5SIG
1400 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1401 struct sock *addr_sk);
1402 int (*calc_md5_hash) (char *location,
1403 struct tcp_md5sig_key *md5,
1404 struct sock *sk,
1405 struct dst_entry *dst,
1406 struct request_sock *req,
1407 struct tcphdr *th,
YOSHIFUJI Hideaki9cb5734e2008-01-12 02:16:03 -08001408 int protocol,
1409 unsigned int len);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001410 int (*md5_add) (struct sock *sk,
1411 struct sock *addr_sk,
1412 u8 *newkey,
1413 u8 len);
1414 int (*md5_parse) (struct sock *sk,
1415 char __user *optval,
1416 int optlen);
1417#endif
1418};
1419
1420struct tcp_request_sock_ops {
1421#ifdef CONFIG_TCP_MD5SIG
1422 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1423 struct request_sock *req);
1424#endif
1425};
1426
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001427extern void tcp_v4_init(struct net_proto_family *ops);
1428extern void tcp_init(void);
1429
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430#endif /* _TCP_H */