blob: eac26b73bcc3f717c02d93d9ed1e2b6ce7e8fb1c [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>
Glenn Griffinc6aefaf2008-02-07 21:49:26 -080032#include <linux/cryptohash.h>
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070033
34#include <net/inet_connection_sock.h>
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -070035#include <net/inet_timewait_sock.h>
Arnaldo Carvalho de Melo77d8bf92005-08-09 20:00:51 -070036#include <net/inet_hashtables.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <net/checksum.h>
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070038#include <net/request_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#include <net/sock.h>
40#include <net/snmp.h>
41#include <net/ip.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070042#include <net/tcp_states.h>
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -070043#include <net/inet_ecn.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070044
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#include <linux/seq_file.h>
46
Arnaldo Carvalho de Melo6e04e022005-08-09 20:07:35 -070047extern struct inet_hashinfo tcp_hashinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
Linus Torvalds1da177e2005-04-16 15:20:36 -070049extern atomic_t tcp_orphan_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -070050extern void tcp_time_wait(struct sock *sk, int state, int timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -070051
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#define MAX_TCP_HEADER (128 + MAX_HEADER)
53
54/*
55 * Never offer a window over 32767 without using window scaling. Some
56 * poor stacks do signed 16bit maths!
57 */
58#define MAX_TCP_WINDOW 32767U
59
60/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
61#define TCP_MIN_MSS 88U
62
63/* Minimal RCV_MSS. */
64#define TCP_MIN_RCVMSS 536U
65
John Heffner5d424d52006-03-20 17:53:41 -080066/* The least MTU to use for probing */
67#define TCP_BASE_MSS 512
68
Linus Torvalds1da177e2005-04-16 15:20:36 -070069/* After receiving this amount of duplicate ACKs fast retransmit starts. */
70#define TCP_FASTRETRANS_THRESH 3
71
72/* Maximal reordering. */
73#define TCP_MAX_REORDERING 127
74
75/* Maximal number of ACKs sent quickly to accelerate slow-start. */
76#define TCP_MAX_QUICKACKS 16U
77
78/* urg_data states */
79#define TCP_URG_VALID 0x0100
80#define TCP_URG_NOTYET 0x0200
81#define TCP_URG_READ 0x0400
82
83#define TCP_RETR1 3 /*
84 * This is how many retries it does before it
85 * tries to figure out if the gateway is
86 * down. Minimal RFC value is 3; it corresponds
87 * to ~3sec-8min depending on RTO.
88 */
89
90#define TCP_RETR2 15 /*
91 * This should take at least
92 * 90 minutes to time out.
93 * RFC1122 says that the limit is 100 sec.
94 * 15 is ~13-30min depending on RTO.
95 */
96
97#define TCP_SYN_RETRIES 5 /* number of times to retry active opening a
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -080098 * connection: ~180sec is RFC minimum */
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100#define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800101 * connection: ~180sec is RFC minimum */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102
103
104#define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned
105 * socket. 7 is ~50sec-16min.
106 */
107
108
109#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
110 * state, about 60 seconds */
111#define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
112 /* BSD style FIN_WAIT2 deadlock breaker.
113 * It used to be 3min, new value is 60sec,
114 * to combine FIN-WAIT-2 timeout with
115 * TIME-WAIT timer.
116 */
117
118#define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
119#if HZ >= 100
120#define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
121#define TCP_ATO_MIN ((unsigned)(HZ/25))
122#else
123#define TCP_DELACK_MIN 4U
124#define TCP_ATO_MIN 4U
125#endif
126#define TCP_RTO_MAX ((unsigned)(120*HZ))
127#define TCP_RTO_MIN ((unsigned)(HZ/5))
128#define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */
129
130#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
131 * for local resources.
132 */
133
134#define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
135#define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
136#define TCP_KEEPALIVE_INTVL (75*HZ)
137
138#define MAX_TCP_KEEPIDLE 32767
139#define MAX_TCP_KEEPINTVL 32767
140#define MAX_TCP_KEEPCNT 127
141#define MAX_TCP_SYNCNT 127
Patrick McManusec3c0982008-03-21 16:33:01 -0700142#define MAX_TCP_ACCEPT_DEFERRED 65535
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
144#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145
146#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
147#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
148 * after this time. It should be equal
149 * (or greater than) TCP_TIMEWAIT_LEN
150 * to provide reliability equal to one
151 * provided by timewait state.
152 */
153#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
154 * timestamps. It must be less than
155 * minimal timewait lifetime.
156 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157/*
158 * TCP option
159 */
160
161#define TCPOPT_NOP 1 /* Padding */
162#define TCPOPT_EOL 0 /* End of options */
163#define TCPOPT_MSS 2 /* Segment size negotiating */
164#define TCPOPT_WINDOW 3 /* Window scaling */
165#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
166#define TCPOPT_SACK 5 /* SACK Block */
167#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800168#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169
170/*
171 * TCP option lengths
172 */
173
174#define TCPOLEN_MSS 4
175#define TCPOLEN_WINDOW 3
176#define TCPOLEN_SACK_PERM 2
177#define TCPOLEN_TIMESTAMP 10
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800178#define TCPOLEN_MD5SIG 18
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179
180/* But this is what stacks really send out. */
181#define TCPOLEN_TSTAMP_ALIGNED 12
182#define TCPOLEN_WSCALE_ALIGNED 4
183#define TCPOLEN_SACKPERM_ALIGNED 4
184#define TCPOLEN_SACK_BASE 2
185#define TCPOLEN_SACK_BASE_ALIGNED 4
186#define TCPOLEN_SACK_PERBLOCK 8
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800187#define TCPOLEN_MD5SIG_ALIGNED 20
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189/* Flags in tp->nonagle */
190#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
191#define TCP_NAGLE_CORK 2 /* Socket is corked */
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800192#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -0700194extern struct inet_timewait_death_row tcp_death_row;
195
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196/* sysctl variables for tcp */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197extern int sysctl_tcp_timestamps;
198extern int sysctl_tcp_window_scaling;
199extern int sysctl_tcp_sack;
200extern int sysctl_tcp_fin_timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201extern int sysctl_tcp_keepalive_time;
202extern int sysctl_tcp_keepalive_probes;
203extern int sysctl_tcp_keepalive_intvl;
204extern int sysctl_tcp_syn_retries;
205extern int sysctl_tcp_synack_retries;
206extern int sysctl_tcp_retries1;
207extern int sysctl_tcp_retries2;
208extern int sysctl_tcp_orphan_retries;
209extern int sysctl_tcp_syncookies;
210extern int sysctl_tcp_retrans_collapse;
211extern int sysctl_tcp_stdurg;
212extern int sysctl_tcp_rfc1337;
213extern int sysctl_tcp_abort_on_overflow;
214extern int sysctl_tcp_max_orphans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215extern int sysctl_tcp_fack;
216extern int sysctl_tcp_reordering;
217extern int sysctl_tcp_ecn;
218extern int sysctl_tcp_dsack;
219extern int sysctl_tcp_mem[3];
220extern int sysctl_tcp_wmem[3];
221extern int sysctl_tcp_rmem[3];
222extern int sysctl_tcp_app_win;
223extern int sysctl_tcp_adv_win_scale;
224extern int sysctl_tcp_tw_reuse;
225extern int sysctl_tcp_frto;
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800226extern int sysctl_tcp_frto_response;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227extern int sysctl_tcp_low_latency;
Chris Leech95937822006-05-23 18:02:55 -0700228extern int sysctl_tcp_dma_copybreak;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229extern int sysctl_tcp_nometrics_save;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230extern int sysctl_tcp_moderate_rcvbuf;
231extern int sysctl_tcp_tso_win_divisor;
Stephen Hemminger9772efb2005-11-10 17:09:53 -0800232extern int sysctl_tcp_abc;
John Heffner5d424d52006-03-20 17:53:41 -0800233extern int sysctl_tcp_mtu_probing;
234extern int sysctl_tcp_base_mss;
Rick Jones15d99e02006-03-20 22:40:29 -0800235extern int sysctl_tcp_workaround_signed_windows;
David S. Miller35089bb2006-06-13 22:33:04 -0700236extern int sysctl_tcp_slow_start_after_idle;
John Heffner886236c2007-03-25 19:21:45 -0700237extern int sysctl_tcp_max_ssthresh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
239extern atomic_t tcp_memory_allocated;
240extern atomic_t tcp_sockets_allocated;
241extern int tcp_memory_pressure;
242
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 * The next routines deal with comparing 32 bit unsigned ints
245 * and worry about wraparound (automatic with unsigned arithmetic).
246 */
247
248static inline int before(__u32 seq1, __u32 seq2)
249{
Gerrit Renker0d630cc2007-01-04 12:25:16 -0800250 return (__s32)(seq1-seq2) < 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251}
Gerrit Renker9a036b9c2006-12-20 10:25:55 -0800252#define after(seq2, seq1) before(seq1, seq2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253
254/* is s2<=s1<=s3 ? */
255static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
256{
257 return seq3 - seq2 >= seq1 - seq2;
258}
259
Pavel Emelianove4fd5da2007-05-29 13:19:18 -0700260static inline int tcp_too_many_orphans(struct sock *sk, int num)
261{
262 return (num > sysctl_tcp_max_orphans) ||
263 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
264 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
265}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266
267extern struct proto tcp_prot;
268
269DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
270#define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field)
271#define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field)
272#define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field)
273#define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field)
274#define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val)
275#define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val)
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277extern void tcp_v4_err(struct sk_buff *skb, u32);
278
279extern void tcp_shutdown (struct sock *sk, int how);
280
281extern int tcp_v4_rcv(struct sk_buff *skb);
282
283extern int tcp_v4_remember_stamp(struct sock *sk);
284
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700285extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
David S. Miller3516ffb2007-08-02 19:23:56 -0700287extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 struct msghdr *msg, size_t size);
289extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
290
291extern int tcp_ioctl(struct sock *sk,
292 int cmd,
293 unsigned long arg);
294
295extern int tcp_rcv_state_process(struct sock *sk,
296 struct sk_buff *skb,
297 struct tcphdr *th,
298 unsigned len);
299
300extern int tcp_rcv_established(struct sock *sk,
301 struct sk_buff *skb,
302 struct tcphdr *th,
303 unsigned len);
304
305extern void tcp_rcv_space_adjust(struct sock *sk);
306
Chris Leech0e4b4992006-05-23 18:00:16 -0700307extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
308
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800309extern int tcp_twsk_unique(struct sock *sk,
310 struct sock *sktw, void *twp);
311
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800312extern void tcp_twsk_destructor(struct sock *sk);
313
Jens Axboe9c55e012007-11-06 23:30:13 -0800314extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
315 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
316
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700317static inline void tcp_dec_quickack_mode(struct sock *sk,
318 const unsigned int pkts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700320 struct inet_connection_sock *icsk = inet_csk(sk);
David S. Millerfc6415bc2005-07-05 15:17:45 -0700321
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700322 if (icsk->icsk_ack.quick) {
323 if (pkts >= icsk->icsk_ack.quick) {
324 icsk->icsk_ack.quick = 0;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700325 /* Leaving quickack mode we deflate ATO. */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700326 icsk->icsk_ack.ato = TCP_ATO_MIN;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700327 } else
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700328 icsk->icsk_ack.quick -= pkts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 }
330}
331
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700332extern void tcp_enter_quickack_mode(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
335{
336 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
337}
338
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -0700339#define TCP_ECN_OK 1
340#define TCP_ECN_QUEUE_CWR 2
341#define TCP_ECN_DEMAND_CWR 4
342
343static __inline__ void
344TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
345{
346 if (sysctl_tcp_ecn && th->ece && th->cwr)
347 inet_rsk(req)->ecn_ok = 1;
348}
349
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350enum tcp_tw_status
351{
352 TCP_TW_SUCCESS = 0,
353 TCP_TW_RST = 1,
354 TCP_TW_ACK = 2,
355 TCP_TW_SYN = 3
356};
357
358
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700359extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 struct sk_buff *skb,
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700361 const struct tcphdr *th);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362
363extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700364 struct request_sock *req,
365 struct request_sock **prev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366extern int tcp_child_process(struct sock *parent,
367 struct sock *child,
368 struct sk_buff *skb);
Ilpo Järvinen46d0de42007-02-21 23:10:39 -0800369extern int tcp_use_frto(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370extern void tcp_enter_frto(struct sock *sk);
371extern void tcp_enter_loss(struct sock *sk, int how);
372extern void tcp_clear_retrans(struct tcp_sock *tp);
373extern void tcp_update_metrics(struct sock *sk);
374
375extern void tcp_close(struct sock *sk,
376 long timeout);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
378
379extern int tcp_getsockopt(struct sock *sk, int level,
380 int optname,
381 char __user *optval,
382 int __user *optlen);
383extern int tcp_setsockopt(struct sock *sk, int level,
384 int optname, char __user *optval,
385 int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800386extern int compat_tcp_getsockopt(struct sock *sk,
387 int level, int optname,
388 char __user *optval, int __user *optlen);
389extern int compat_tcp_setsockopt(struct sock *sk,
390 int level, int optname,
391 char __user *optval, int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392extern void tcp_set_keepalive(struct sock *sk, int val);
393extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
394 struct msghdr *msg,
395 size_t len, int nonblock,
396 int flags, int *addr_len);
397
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398extern void tcp_parse_options(struct sk_buff *skb,
399 struct tcp_options_received *opt_rx,
400 int estab);
401
YOSHIFUJI Hideaki7d5d5522008-04-17 12:29:53 +0900402extern u8 *tcp_parse_md5sig_option(struct tcphdr *th);
403
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404/*
405 * TCP v4 functions exported for the inet6 API
406 */
407
Arnaldo Carvalho de Melo8292a172005-12-13 23:15:52 -0800408extern void tcp_v4_send_check(struct sock *sk, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 struct sk_buff *skb);
410
411extern int tcp_v4_conn_request(struct sock *sk,
412 struct sk_buff *skb);
413
414extern struct sock * tcp_create_openreq_child(struct sock *sk,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700415 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 struct sk_buff *skb);
417
418extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
419 struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700420 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 struct dst_entry *dst);
422
423extern int tcp_v4_do_rcv(struct sock *sk,
424 struct sk_buff *skb);
425
426extern int tcp_v4_connect(struct sock *sk,
427 struct sockaddr *uaddr,
428 int addr_len);
429
430extern int tcp_connect(struct sock *sk);
431
432extern struct sk_buff * tcp_make_synack(struct sock *sk,
433 struct dst_entry *dst,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700434 struct request_sock *req);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435
436extern int tcp_disconnect(struct sock *sk, int flags);
437
438extern void tcp_unhash(struct sock *sk);
439
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440/* From syncookies.c */
Florian Westphal2051f112008-03-23 22:21:28 -0700441extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
443 struct ip_options *opt);
444extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
445 __u16 *mss);
446
Florian Westphal4dfc2812008-04-10 03:12:40 -0700447extern __u32 cookie_init_timestamp(struct request_sock *req);
448extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt);
449
Glenn Griffinc6aefaf2008-02-07 21:49:26 -0800450/* From net/ipv6/syncookies.c */
451extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
452extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb,
453 __u16 *mss);
454
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455/* tcp_output.c */
456
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700457extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
458 int nonagle);
459extern int tcp_may_send_now(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
461extern void tcp_xmit_retransmit_queue(struct sock *);
462extern void tcp_simple_retransmit(struct sock *);
463extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
David S. Miller6475be12005-09-01 22:47:01 -0700464extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
466extern void tcp_send_probe0(struct sock *);
467extern void tcp_send_partial(struct sock *);
468extern int tcp_write_wakeup(struct sock *);
469extern void tcp_send_fin(struct sock *sk);
Al Virodd0fc662005-10-07 07:46:04 +0100470extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471extern int tcp_send_synack(struct sock *);
David S. Millerc1b4a7e2005-07-05 15:24:38 -0700472extern void tcp_push_one(struct sock *, unsigned int mss_now);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473extern void tcp_send_ack(struct sock *sk);
474extern void tcp_send_delayed_ack(struct sock *sk);
475
David S. Millera762a982005-07-05 15:18:51 -0700476/* tcp_input.c */
477extern void tcp_cwnd_application_limited(struct sock *sk);
478
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479/* tcp_timer.c */
480extern void tcp_init_xmit_timers(struct sock *);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700481static inline void tcp_clear_xmit_timers(struct sock *sk)
482{
483 inet_csk_clear_xmit_timers(sk);
484}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
487extern unsigned int tcp_current_mss(struct sock *sk, int large);
488
Arnaldo Carvalho de Melo17b085e2005-08-12 12:59:17 -0300489/* tcp.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490extern void tcp_get_info(struct sock *, struct tcp_info *);
491
492/* Read 'sendfile()'-style from a TCP socket */
493typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
494 unsigned int, size_t);
495extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
496 sk_read_actor_t recv_actor);
497
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800498extern void tcp_initialize_rcv_mss(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
John Heffner5d424d52006-03-20 17:53:41 -0800500extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
501extern int tcp_mss_to_mtu(struct sock *sk, int mss);
502extern void tcp_mtup_init(struct sock *sk);
503
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800504static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505{
506 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
507 ntohl(TCP_FLAG_ACK) |
508 snd_wnd);
509}
510
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800511static inline void tcp_fast_path_on(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512{
513 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
514}
515
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700516static inline void tcp_fast_path_check(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700518 struct tcp_sock *tp = tcp_sk(sk);
519
David S. Millerb03efcf2005-07-08 14:57:23 -0700520 if (skb_queue_empty(&tp->out_of_order_queue) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 tp->rcv_wnd &&
522 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
523 !tp->urg_data)
524 tcp_fast_path_on(tp);
525}
526
527/* Compute the actual receive window we are currently advertising.
528 * Rcv_nxt can be after the window if our peer push more data
529 * than the offered window.
530 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800531static inline u32 tcp_receive_window(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532{
533 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
534
535 if (win < 0)
536 win = 0;
537 return (u32) win;
538}
539
540/* Choose a new window, without checks for shrinking, and without
541 * scaling applied to the result. The caller does these things
542 * if necessary. This is a "raw" window selection.
543 */
544extern u32 __tcp_select_window(struct sock *sk);
545
546/* TCP timestamps are only 32-bits, this causes a slight
547 * complication on 64-bit systems since we store a snapshot
Stephen Hemminger31f34262005-11-15 15:17:10 -0800548 * of jiffies in the buffer control blocks below. We decided
549 * to use only the low 32-bits of jiffies and hide the ugly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 * casts with the following macro.
551 */
552#define tcp_time_stamp ((__u32)(jiffies))
553
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800554/* This is what the send packet queuing engine uses to pass
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 * TCP per-packet control information to the transmission
556 * code. We also store the host-order sequence numbers in
557 * here too. This is 36 bytes on 32-bit architectures,
558 * 40 bytes on 64-bit machines, if this grows please adjust
559 * skbuff.h:skbuff->cb[xxx] size appropriately.
560 */
561struct tcp_skb_cb {
562 union {
563 struct inet_skb_parm h4;
564#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
565 struct inet6_skb_parm h6;
566#endif
567 } header; /* For incoming frames */
568 __u32 seq; /* Starting sequence number */
569 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
570 __u32 when; /* used to compute rtt's */
571 __u8 flags; /* TCP header flags. */
572
573 /* NOTE: These must match up to the flags byte in a
574 * real TCP header.
575 */
576#define TCPCB_FLAG_FIN 0x01
577#define TCPCB_FLAG_SYN 0x02
578#define TCPCB_FLAG_RST 0x04
579#define TCPCB_FLAG_PSH 0x08
580#define TCPCB_FLAG_ACK 0x10
581#define TCPCB_FLAG_URG 0x20
582#define TCPCB_FLAG_ECE 0x40
583#define TCPCB_FLAG_CWR 0x80
584
585 __u8 sacked; /* State flags for SACK/FACK. */
586#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
587#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
588#define TCPCB_LOST 0x04 /* SKB is lost */
589#define TCPCB_TAGBITS 0x07 /* All tag bits */
590
591#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
592#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
593
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 __u16 urg_ptr; /* Valid w/URG flags is set. */
595 __u32 ack_seq; /* Sequence number ACK'd */
596};
597
598#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
599
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600/* Due to TSO, an SKB can be composed of multiple actual
601 * packets. To keep these tracked properly, we use this.
602 */
603static inline int tcp_skb_pcount(const struct sk_buff *skb)
604{
Herbert Xu79671682006-06-22 02:40:14 -0700605 return skb_shinfo(skb)->gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606}
607
608/* This is valid iff tcp_skb_pcount() > 1. */
609static inline int tcp_skb_mss(const struct sk_buff *skb)
610{
Herbert Xu79671682006-06-22 02:40:14 -0700611 return skb_shinfo(skb)->gso_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612}
613
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300614static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615{
616 if (*count) {
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300617 *count -= decr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 if ((int)*count < 0)
619 *count = 0;
620 }
621}
622
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300623static inline void tcp_dec_pcount_approx(__u32 *count,
624 const struct sk_buff *skb)
625{
626 tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb));
627}
628
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700629/* Events passed to congestion control interface */
630enum tcp_ca_event {
631 CA_EVENT_TX_START, /* first transmit when no packets in flight */
632 CA_EVENT_CWND_RESTART, /* congestion window restart */
633 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
634 CA_EVENT_FRTO, /* fast recovery timeout */
635 CA_EVENT_LOSS, /* loss timeout */
636 CA_EVENT_FAST_ACK, /* in sequence ack */
637 CA_EVENT_SLOW_ACK, /* other ack */
638};
639
640/*
641 * Interface for adding new TCP congestion control handlers
642 */
643#define TCP_CA_NAME_MAX 16
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800644#define TCP_CA_MAX 128
645#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
646
Stephen Hemminger164891a2007-04-23 22:26:16 -0700647#define TCP_CONG_NON_RESTRICTED 0x1
648#define TCP_CONG_RTT_STAMP 0x2
649
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700650struct tcp_congestion_ops {
651 struct list_head list;
Stephen Hemminger164891a2007-04-23 22:26:16 -0700652 unsigned long flags;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700653
654 /* initialize private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300655 void (*init)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700656 /* cleanup private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300657 void (*release)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700658
659 /* return slow start threshold (required) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300660 u32 (*ssthresh)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700661 /* lower bound for congestion window (optional) */
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700662 u32 (*min_cwnd)(const struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700663 /* do new cwnd calculation (required) */
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200664 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700665 /* call before changing ca_state (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300666 void (*set_state)(struct sock *sk, u8 new_state);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700667 /* call when cwnd event occurs (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300668 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700669 /* new value of cwnd after loss (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300670 u32 (*undo_cwnd)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700671 /* hook for packet ack accounting (optional) */
Stephen Hemminger30cfd0b2007-07-25 23:49:34 -0700672 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
Arnaldo Carvalho de Melo73c1f4a2005-08-12 12:51:49 -0300673 /* get info for inet_diag (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300674 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700675
676 char name[TCP_CA_NAME_MAX];
677 struct module *owner;
678};
679
680extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
681extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
682
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300683extern void tcp_init_congestion_control(struct sock *sk);
684extern void tcp_cleanup_congestion_control(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700685extern int tcp_set_default_congestion_control(const char *name);
686extern void tcp_get_default_congestion_control(char *name);
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800687extern void tcp_get_available_congestion_control(char *buf, size_t len);
Stephen Hemmingerce7bc3b2006-11-09 16:35:15 -0800688extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
689extern int tcp_set_allowed_congestion_control(char *allowed);
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300690extern int tcp_set_congestion_control(struct sock *sk, const char *name);
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800691extern void tcp_slow_start(struct tcp_sock *tp);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700692
Stephen Hemminger5f8ef482005-06-23 20:37:36 -0700693extern struct tcp_congestion_ops tcp_init_congestion_ops;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300694extern u32 tcp_reno_ssthresh(struct sock *sk);
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200695extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700696extern u32 tcp_reno_min_cwnd(const struct sock *sk);
David S. Millera8acfba2005-06-23 23:45:02 -0700697extern struct tcp_congestion_ops tcp_reno;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700698
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300699static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700700{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300701 struct inet_connection_sock *icsk = inet_csk(sk);
702
703 if (icsk->icsk_ca_ops->set_state)
704 icsk->icsk_ca_ops->set_state(sk, ca_state);
705 icsk->icsk_ca_state = ca_state;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700706}
707
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300708static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700709{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300710 const struct inet_connection_sock *icsk = inet_csk(sk);
711
712 if (icsk->icsk_ca_ops->cwnd_event)
713 icsk->icsk_ca_ops->cwnd_event(sk, event);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700714}
715
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300716/* These functions determine how the current flow behaves in respect of SACK
717 * handling. SACK is negotiated with the peer, and therefore it can vary
718 * between different flows.
719 *
720 * tcp_is_sack - SACK enabled
721 * tcp_is_reno - No SACK
722 * tcp_is_fack - FACK enabled, implies SACK enabled
723 */
724static inline int tcp_is_sack(const struct tcp_sock *tp)
725{
726 return tp->rx_opt.sack_ok;
727}
728
729static inline int tcp_is_reno(const struct tcp_sock *tp)
730{
731 return !tcp_is_sack(tp);
732}
733
734static inline int tcp_is_fack(const struct tcp_sock *tp)
735{
736 return tp->rx_opt.sack_ok & 2;
737}
738
739static inline void tcp_enable_fack(struct tcp_sock *tp)
740{
741 tp->rx_opt.sack_ok |= 2;
742}
743
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300744static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
745{
746 return tp->sacked_out + tp->lost_out;
747}
748
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749/* This determines how many packets are "in the network" to the best
750 * of our knowledge. In many cases it is conservative, but where
751 * detailed information is available from the receiver (via SACK
752 * blocks etc.) we can make more aggressive calculations.
753 *
754 * Use this for decisions involving congestion control, use just
755 * tp->packets_out to determine if the send queue is empty or not.
756 *
757 * Read this equation as:
758 *
759 * "Packets sent once on transmission queue" MINUS
760 * "Packets left network, but not honestly ACKed yet" PLUS
761 * "Packets fast retransmitted"
762 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800763static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764{
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300765 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766}
767
Ilpo Järvinen882beba2008-04-07 22:33:07 -0700768extern int tcp_limit_reno_sacked(struct tcp_sock *tp);
769
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
771 * The exception is rate halving phase, when cwnd is decreasing towards
772 * ssthresh.
773 */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300774static inline __u32 tcp_current_ssthresh(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300776 const struct tcp_sock *tp = tcp_sk(sk);
777 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 return tp->snd_ssthresh;
779 else
780 return max(tp->snd_ssthresh,
781 ((tp->snd_cwnd >> 1) +
782 (tp->snd_cwnd >> 2)));
783}
784
Ilpo Järvinenb9c45952007-07-27 16:36:17 +0300785/* Use define here intentionally to get WARN_ON location shown at the caller */
786#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800788extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
790
791/* Slow start with delack produces 3 packets of burst, so that
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700792 * it is safe "de facto". This will be the default - same as
793 * the default reordering threshold - but if reordering increases,
794 * we must be able to allow cwnd to burst at least this much in order
795 * to not pull it back when holes are filled.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 */
797static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
798{
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700799 return tp->reordering;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800}
801
Ilpo Järvinen90840de2007-12-31 04:48:41 -0800802/* Returns end sequence number of the receiver's advertised window */
803static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
804{
805 return tp->snd_una + tp->snd_wnd;
806}
Ilpo Järvinencea14e02008-01-12 03:19:12 -0800807extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800808
Chuck Leverc1bd24b2007-10-23 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
Ilpo Järvinen490d5042008-01-12 03:17:20 -0800927extern void tcp_set_state(struct sock *sk, int state);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928
Andi Kleen4ac02ba2007-04-20 17:11:46 -0700929extern void tcp_done(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800931static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932{
933 rx_opt->dsack = 0;
934 rx_opt->eff_sacks = 0;
935 rx_opt->num_sacks = 0;
936}
937
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938/* Determine a window scaling and initial window to offer. */
939extern void tcp_select_initial_window(int __space, __u32 mss,
940 __u32 *rcv_wnd, __u32 *window_clamp,
941 int wscale_ok, __u8 *rcv_wscale);
942
943static inline int tcp_win_from_space(int space)
944{
945 return sysctl_tcp_adv_win_scale<=0 ?
946 (space>>(-sysctl_tcp_adv_win_scale)) :
947 space - (space>>sysctl_tcp_adv_win_scale);
948}
949
950/* Note: caller must be prepared to deal with negative returns */
951static inline int tcp_space(const struct sock *sk)
952{
953 return tcp_win_from_space(sk->sk_rcvbuf -
954 atomic_read(&sk->sk_rmem_alloc));
955}
956
957static inline int tcp_full_space(const struct sock *sk)
958{
959 return tcp_win_from_space(sk->sk_rcvbuf);
960}
961
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800962static inline void tcp_openreq_init(struct request_sock *req,
963 struct tcp_options_received *rx_opt,
964 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700966 struct inet_request_sock *ireq = inet_rsk(req);
967
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
Florian Westphal4dfc2812008-04-10 03:12:40 -0700969 req->cookie_ts = 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700970 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 req->mss = rx_opt->mss_clamp;
972 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700973 ireq->tstamp_ok = rx_opt->tstamp_ok;
974 ireq->sack_ok = rx_opt->sack_ok;
975 ireq->snd_wscale = rx_opt->snd_wscale;
976 ireq->wscale_ok = rx_opt->wscale_ok;
977 ireq->acked = 0;
978 ireq->ecn_ok = 0;
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -0700979 ireq->rmt_port = tcp_hdr(skb)->source;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980}
981
982extern void tcp_enter_memory_pressure(void);
983
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984static inline int keepalive_intvl_when(const struct tcp_sock *tp)
985{
986 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
987}
988
989static inline int keepalive_time_when(const struct tcp_sock *tp)
990{
991 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
992}
993
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700994static inline int tcp_fin_time(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700996 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
997 const int rto = inet_csk(sk)->icsk_rto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700999 if (fin_timeout < (rto << 2) - (rto >> 1))
1000 fin_timeout = (rto << 2) - (rto >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001
1002 return fin_timeout;
1003}
1004
1005static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1006{
1007 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1008 return 0;
James Morris9d729f72007-03-04 16:12:44 -08001009 if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 return 0;
1011
1012 /* RST segments are not recommended to carry timestamp,
1013 and, if they do, it is recommended to ignore PAWS because
1014 "their cleanup function should take precedence over timestamps."
1015 Certainly, it is mistake. It is necessary to understand the reasons
1016 of this constraint to relax it: if peer reboots, clock may go
1017 out-of-sync and half-open connections will not be reset.
1018 Actually, the problem would be not existing if all
1019 the implementations followed draft about maintaining clock
1020 via reboots. Linux-2.2 DOES NOT!
1021
1022 However, we can relax time bounds for RST segments to MSL.
1023 */
James Morris9d729f72007-03-04 16:12:44 -08001024 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 return 0;
1026 return 1;
1027}
1028
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029#define TCP_CHECK_TIMER(sk) do { } while (0)
1030
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031static inline void tcp_mib_init(void)
1032{
1033 /* See RFC 2012 */
1034 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1035 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1036 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1037 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1038}
1039
Ilpo Järvinen5af4ec22007-09-20 11:30:48 -07001040/* from STCP */
David S. Miller0800f172007-09-20 11:40:37 -07001041static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1042{
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001043 tp->lost_skb_hint = NULL;
1044 tp->scoreboard_skb_hint = NULL;
1045 tp->retransmit_skb_hint = NULL;
1046 tp->forward_skb_hint = NULL;
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001047}
1048
David S. Miller0800f172007-09-20 11:40:37 -07001049static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1050{
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001051 tcp_clear_retrans_hints_partial(tp);
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001052}
1053
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001054/* MD5 Signature */
1055struct crypto_hash;
1056
1057/* - key database */
1058struct tcp_md5sig_key {
1059 u8 *key;
1060 u8 keylen;
1061};
1062
1063struct tcp4_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001064 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001065 __be32 addr;
1066};
1067
1068struct tcp6_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001069 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001070#if 0
1071 u32 scope_id; /* XXX */
1072#endif
1073 struct in6_addr addr;
1074};
1075
1076/* - sock block */
1077struct tcp_md5sig_info {
1078 struct tcp4_md5sig_key *keys4;
1079#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1080 struct tcp6_md5sig_key *keys6;
1081 u32 entries6;
1082 u32 alloced6;
1083#endif
1084 u32 entries4;
1085 u32 alloced4;
1086};
1087
1088/* - pseudo header */
1089struct tcp4_pseudohdr {
1090 __be32 saddr;
1091 __be32 daddr;
1092 __u8 pad;
1093 __u8 protocol;
1094 __be16 len;
1095};
1096
1097struct tcp6_pseudohdr {
1098 struct in6_addr saddr;
1099 struct in6_addr daddr;
1100 __be32 len;
1101 __be32 protocol; /* including padding */
1102};
1103
1104union tcp_md5sum_block {
1105 struct tcp4_pseudohdr ip4;
1106#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1107 struct tcp6_pseudohdr ip6;
1108#endif
1109};
1110
1111/* - pool: digest algorithm, hash description and scratch buffer */
1112struct tcp_md5sig_pool {
1113 struct hash_desc md5_desc;
1114 union tcp_md5sum_block md5_blk;
1115};
1116
1117#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1118
1119/* - functions */
YOSHIFUJI Hideaki8d26d762008-04-17 13:19:16 +09001120extern int tcp_calc_md5_hash(char *md5_hash,
1121 struct tcp_md5sig_key *key,
1122 int bplen,
1123 struct tcphdr *th,
1124 unsigned int tcplen,
1125 struct tcp_md5sig_pool *hp);
1126
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001127extern int tcp_v4_calc_md5_hash(char *md5_hash,
1128 struct tcp_md5sig_key *key,
1129 struct sock *sk,
1130 struct dst_entry *dst,
1131 struct request_sock *req,
1132 struct tcphdr *th,
YOSHIFUJI Hideaki9cb57342008-01-12 02:16:03 -08001133 unsigned int tcplen);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001134extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1135 struct sock *addr_sk);
1136
1137extern int tcp_v4_md5_do_add(struct sock *sk,
1138 __be32 addr,
1139 u8 *newkey,
1140 u8 newkeylen);
1141
1142extern int tcp_v4_md5_do_del(struct sock *sk,
Al Viro8e5200f2006-11-20 18:06:37 -08001143 __be32 addr);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001144
1145extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1146extern void tcp_free_md5sig_pool(void);
1147
1148extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1149extern void __tcp_put_md5sig_pool(void);
1150
1151static inline
1152struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1153{
1154 int cpu = get_cpu();
1155 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1156 if (!ret)
1157 put_cpu();
1158 return ret;
1159}
1160
1161static inline void tcp_put_md5sig_pool(void)
1162{
1163 __tcp_put_md5sig_pool();
1164 put_cpu();
1165}
1166
David S. Millerfe067e82007-03-07 12:12:44 -08001167/* write queue abstraction */
1168static inline void tcp_write_queue_purge(struct sock *sk)
1169{
1170 struct sk_buff *skb;
1171
1172 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001173 sk_wmem_free_skb(sk, skb);
1174 sk_mem_reclaim(sk);
David S. Millerfe067e82007-03-07 12:12:44 -08001175}
1176
1177static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1178{
1179 struct sk_buff *skb = sk->sk_write_queue.next;
1180 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1181 return NULL;
1182 return skb;
1183}
1184
1185static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1186{
1187 struct sk_buff *skb = sk->sk_write_queue.prev;
1188 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1189 return NULL;
1190 return skb;
1191}
1192
1193static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1194{
1195 return skb->next;
1196}
1197
1198#define tcp_for_write_queue(skb, sk) \
1199 for (skb = (sk)->sk_write_queue.next; \
1200 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1201 skb = skb->next)
1202
1203#define tcp_for_write_queue_from(skb, sk) \
1204 for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\
1205 skb = skb->next)
1206
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001207#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
1208 for (tmp = skb->next; \
1209 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1210 skb = tmp, tmp = skb->next)
1211
David S. Millerfe067e82007-03-07 12:12:44 -08001212static inline struct sk_buff *tcp_send_head(struct sock *sk)
1213{
1214 return sk->sk_send_head;
1215}
1216
1217static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1218{
1219 sk->sk_send_head = skb->next;
1220 if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue)
1221 sk->sk_send_head = NULL;
1222}
1223
1224static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1225{
1226 if (sk->sk_send_head == skb_unlinked)
1227 sk->sk_send_head = NULL;
1228}
1229
1230static inline void tcp_init_send_head(struct sock *sk)
1231{
1232 sk->sk_send_head = NULL;
1233}
1234
1235static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1236{
1237 __skb_queue_tail(&sk->sk_write_queue, skb);
1238}
1239
1240static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1241{
1242 __tcp_add_write_queue_tail(sk, skb);
1243
1244 /* Queue it, remembering where we must start sending. */
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001245 if (sk->sk_send_head == NULL) {
David S. Millerfe067e82007-03-07 12:12:44 -08001246 sk->sk_send_head = skb;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001247
1248 if (tcp_sk(sk)->highest_sack == NULL)
1249 tcp_sk(sk)->highest_sack = skb;
1250 }
David S. Millerfe067e82007-03-07 12:12:44 -08001251}
1252
1253static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1254{
1255 __skb_queue_head(&sk->sk_write_queue, skb);
1256}
1257
1258/* Insert buff after skb on the write queue of sk. */
1259static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1260 struct sk_buff *buff,
1261 struct sock *sk)
1262{
Gerrit Renker7de6c032008-04-14 00:05:09 -07001263 __skb_queue_after(&sk->sk_write_queue, skb, buff);
David S. Millerfe067e82007-03-07 12:12:44 -08001264}
1265
1266/* Insert skb between prev and next on the write queue of sk. */
1267static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1268 struct sk_buff *skb,
1269 struct sock *sk)
1270{
1271 __skb_insert(new, skb->prev, skb, &sk->sk_write_queue);
Ilpo Järvinen6e421412007-11-19 23:24:09 -08001272
1273 if (sk->sk_send_head == skb)
1274 sk->sk_send_head = new;
David S. Millerfe067e82007-03-07 12:12:44 -08001275}
1276
1277static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1278{
1279 __skb_unlink(skb, &sk->sk_write_queue);
1280}
1281
1282static inline int tcp_skb_is_last(const struct sock *sk,
1283 const struct sk_buff *skb)
1284{
1285 return skb->next == (struct sk_buff *)&sk->sk_write_queue;
1286}
1287
1288static inline int tcp_write_queue_empty(struct sock *sk)
1289{
1290 return skb_queue_empty(&sk->sk_write_queue);
1291}
1292
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001293/* Start sequence of the highest skb with SACKed bit, valid only if
1294 * sacked > 0 or when the caller has ensured validity by itself.
1295 */
1296static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1297{
1298 if (!tp->sacked_out)
1299 return tp->snd_una;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001300
1301 if (tp->highest_sack == NULL)
1302 return tp->snd_nxt;
1303
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001304 return TCP_SKB_CB(tp->highest_sack)->seq;
1305}
1306
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001307static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1308{
1309 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1310 tcp_write_queue_next(sk, skb);
1311}
1312
1313static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1314{
1315 return tcp_sk(sk)->highest_sack;
1316}
1317
1318static inline void tcp_highest_sack_reset(struct sock *sk)
1319{
1320 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1321}
1322
1323/* Called when old skb is about to be deleted (to be combined with new skb) */
1324static inline void tcp_highest_sack_combine(struct sock *sk,
1325 struct sk_buff *old,
1326 struct sk_buff *new)
1327{
1328 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1329 tcp_sk(sk)->highest_sack = new;
1330}
1331
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332/* /proc */
1333enum tcp_seq_states {
1334 TCP_SEQ_STATE_LISTENING,
1335 TCP_SEQ_STATE_OPENREQ,
1336 TCP_SEQ_STATE_ESTABLISHED,
1337 TCP_SEQ_STATE_TIME_WAIT,
1338};
1339
1340struct tcp_seq_afinfo {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 char *name;
1342 sa_family_t family;
Denis V. Lunev68fcadd2008-04-13 22:13:30 -07001343 struct file_operations seq_fops;
Denis V. Lunev9427c4b2008-04-13 22:12:13 -07001344 struct seq_operations seq_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345};
1346
1347struct tcp_iter_state {
Denis V. Luneva4146b12008-04-13 22:11:14 -07001348 struct seq_net_private p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 sa_family_t family;
1350 enum tcp_seq_states state;
1351 struct sock *syn_wait_sk;
1352 int bucket, sbucket, num, uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353};
1354
Daniel Lezcano6f8b13b2008-03-21 04:14:45 -07001355extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1356extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001358extern struct request_sock_ops tcp_request_sock_ops;
Glenn Griffinc6aefaf2008-02-07 21:49:26 -08001359extern struct request_sock_ops tcp6_request_sock_ops;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001360
1361extern int tcp_v4_destroy_sock(struct sock *sk);
1362
Herbert Xua430a432006-07-08 13:34:56 -07001363extern int tcp_v4_gso_send_check(struct sk_buff *skb);
Herbert Xu576a30e2006-06-27 13:22:38 -07001364extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001365
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001366#ifdef CONFIG_PROC_FS
1367extern int tcp4_proc_init(void);
1368extern void tcp4_proc_exit(void);
1369#endif
1370
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001371/* TCP af-specific functions */
1372struct tcp_sock_af_ops {
1373#ifdef CONFIG_TCP_MD5SIG
1374 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1375 struct sock *addr_sk);
1376 int (*calc_md5_hash) (char *location,
1377 struct tcp_md5sig_key *md5,
1378 struct sock *sk,
1379 struct dst_entry *dst,
1380 struct request_sock *req,
1381 struct tcphdr *th,
YOSHIFUJI Hideaki9cb57342008-01-12 02:16:03 -08001382 unsigned int len);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001383 int (*md5_add) (struct sock *sk,
1384 struct sock *addr_sk,
1385 u8 *newkey,
1386 u8 len);
1387 int (*md5_parse) (struct sock *sk,
1388 char __user *optval,
1389 int optlen);
1390#endif
1391};
1392
1393struct tcp_request_sock_ops {
1394#ifdef CONFIG_TCP_MD5SIG
1395 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1396 struct request_sock *req);
1397#endif
1398};
1399
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001400extern void tcp_v4_init(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001401extern void tcp_init(void);
1402
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403#endif /* _TCP_H */