blob: 8983386356a59d228958718d04750a557da131b3 [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)
Adam Langley33ad7982008-07-19 00:04:31 -070053#define MAX_TCP_OPTION_SPACE 40
Linus Torvalds1da177e2005-04-16 15:20:36 -070054
55/*
56 * Never offer a window over 32767 without using window scaling. Some
57 * poor stacks do signed 16bit maths!
58 */
59#define MAX_TCP_WINDOW 32767U
60
61/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
62#define TCP_MIN_MSS 88U
63
64/* Minimal RCV_MSS. */
65#define TCP_MIN_RCVMSS 536U
66
John Heffner5d424d52006-03-20 17:53:41 -080067/* The least MTU to use for probing */
68#define TCP_BASE_MSS 512
69
Linus Torvalds1da177e2005-04-16 15:20:36 -070070/* After receiving this amount of duplicate ACKs fast retransmit starts. */
71#define TCP_FASTRETRANS_THRESH 3
72
73/* Maximal reordering. */
74#define TCP_MAX_REORDERING 127
75
76/* Maximal number of ACKs sent quickly to accelerate slow-start. */
77#define TCP_MAX_QUICKACKS 16U
78
79/* urg_data states */
80#define TCP_URG_VALID 0x0100
81#define TCP_URG_NOTYET 0x0200
82#define TCP_URG_READ 0x0400
83
84#define TCP_RETR1 3 /*
85 * This is how many retries it does before it
86 * tries to figure out if the gateway is
87 * down. Minimal RFC value is 3; it corresponds
88 * to ~3sec-8min depending on RTO.
89 */
90
91#define TCP_RETR2 15 /*
92 * This should take at least
93 * 90 minutes to time out.
94 * RFC1122 says that the limit is 100 sec.
95 * 15 is ~13-30min depending on RTO.
96 */
97
98#define TCP_SYN_RETRIES 5 /* number of times to retry active opening a
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -080099 * connection: ~180sec is RFC minimum */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
101#define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800102 * connection: ~180sec is RFC minimum */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
104
105#define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned
106 * socket. 7 is ~50sec-16min.
107 */
108
109
110#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
111 * state, about 60 seconds */
112#define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
113 /* BSD style FIN_WAIT2 deadlock breaker.
114 * It used to be 3min, new value is 60sec,
115 * to combine FIN-WAIT-2 timeout with
116 * TIME-WAIT timer.
117 */
118
119#define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
120#if HZ >= 100
121#define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
122#define TCP_ATO_MIN ((unsigned)(HZ/25))
123#else
124#define TCP_DELACK_MIN 4U
125#define TCP_ATO_MIN 4U
126#endif
127#define TCP_RTO_MAX ((unsigned)(120*HZ))
128#define TCP_RTO_MIN ((unsigned)(HZ/5))
129#define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */
130
131#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
132 * for local resources.
133 */
134
135#define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
136#define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
137#define TCP_KEEPALIVE_INTVL (75*HZ)
138
139#define MAX_TCP_KEEPIDLE 32767
140#define MAX_TCP_KEEPINTVL 32767
141#define MAX_TCP_KEEPCNT 127
142#define MAX_TCP_SYNCNT 127
143
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
Adam Langley33ad7982008-07-19 00:04:31 -0700188#define TCPOLEN_MSS_ALIGNED 4
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190/* Flags in tp->nonagle */
191#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
192#define TCP_NAGLE_CORK 2 /* Socket is corked */
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800193#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -0700195extern struct inet_timewait_death_row tcp_death_row;
196
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197/* sysctl variables for tcp */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198extern int sysctl_tcp_timestamps;
199extern int sysctl_tcp_window_scaling;
200extern int sysctl_tcp_sack;
201extern int sysctl_tcp_fin_timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202extern int sysctl_tcp_keepalive_time;
203extern int sysctl_tcp_keepalive_probes;
204extern int sysctl_tcp_keepalive_intvl;
205extern int sysctl_tcp_syn_retries;
206extern int sysctl_tcp_synack_retries;
207extern int sysctl_tcp_retries1;
208extern int sysctl_tcp_retries2;
209extern int sysctl_tcp_orphan_retries;
210extern int sysctl_tcp_syncookies;
211extern int sysctl_tcp_retrans_collapse;
212extern int sysctl_tcp_stdurg;
213extern int sysctl_tcp_rfc1337;
214extern int sysctl_tcp_abort_on_overflow;
215extern int sysctl_tcp_max_orphans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216extern int sysctl_tcp_fack;
217extern int sysctl_tcp_reordering;
218extern int sysctl_tcp_ecn;
219extern int sysctl_tcp_dsack;
220extern int sysctl_tcp_mem[3];
221extern int sysctl_tcp_wmem[3];
222extern int sysctl_tcp_rmem[3];
223extern int sysctl_tcp_app_win;
224extern int sysctl_tcp_adv_win_scale;
225extern int sysctl_tcp_tw_reuse;
226extern int sysctl_tcp_frto;
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800227extern int sysctl_tcp_frto_response;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228extern int sysctl_tcp_low_latency;
Chris Leech95937822006-05-23 18:02:55 -0700229extern int sysctl_tcp_dma_copybreak;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230extern int sysctl_tcp_nometrics_save;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231extern int sysctl_tcp_moderate_rcvbuf;
232extern int sysctl_tcp_tso_win_divisor;
Stephen Hemminger9772efb2005-11-10 17:09:53 -0800233extern int sysctl_tcp_abc;
John Heffner5d424d52006-03-20 17:53:41 -0800234extern int sysctl_tcp_mtu_probing;
235extern int sysctl_tcp_base_mss;
Rick Jones15d99e02006-03-20 22:40:29 -0800236extern int sysctl_tcp_workaround_signed_windows;
David S. Miller35089bb2006-06-13 22:33:04 -0700237extern int sysctl_tcp_slow_start_after_idle;
John Heffner886236c2007-03-25 19:21:45 -0700238extern int sysctl_tcp_max_ssthresh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240extern atomic_t tcp_memory_allocated;
241extern atomic_t tcp_sockets_allocated;
242extern int tcp_memory_pressure;
243
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 * The next routines deal with comparing 32 bit unsigned ints
246 * and worry about wraparound (automatic with unsigned arithmetic).
247 */
248
249static inline int before(__u32 seq1, __u32 seq2)
250{
Gerrit Renker0d630cc2007-01-04 12:25:16 -0800251 return (__s32)(seq1-seq2) < 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252}
Gerrit Renker9a036b92006-12-20 10:25:55 -0800253#define after(seq2, seq1) before(seq1, seq2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254
255/* is s2<=s1<=s3 ? */
256static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
257{
258 return seq3 - seq2 >= seq1 - seq2;
259}
260
Pavel Emelianove4fd5da2007-05-29 13:19:18 -0700261static inline int tcp_too_many_orphans(struct sock *sk, int num)
262{
263 return (num > sysctl_tcp_max_orphans) ||
264 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
265 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
266}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
268extern struct proto tcp_prot;
269
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -0700270#define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field)
271#define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
272#define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
273#define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274
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 Melo8feaf0c2005-08-09 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. Millerfc6415b2005-07-05 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. Millerfc6415b2005-07-05 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. Millerfc6415b2005-07-05 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 Melo8feaf0c2005-08-09 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 Melo8feaf0c2005-08-09 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
YOSHIFUJI Hideaki7d5d5522008-04-17 12:29:53 +0900400extern u8 *tcp_parse_md5sig_option(struct tcphdr *th);
401
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402/*
403 * TCP v4 functions exported for the inet6 API
404 */
405
Arnaldo Carvalho de Melo8292a172005-12-13 23:15:52 -0800406extern void tcp_v4_send_check(struct sock *sk, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 struct sk_buff *skb);
408
409extern int tcp_v4_conn_request(struct sock *sk,
410 struct sk_buff *skb);
411
412extern struct sock * tcp_create_openreq_child(struct sock *sk,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700413 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 struct sk_buff *skb);
415
416extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
417 struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700418 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 struct dst_entry *dst);
420
421extern int tcp_v4_do_rcv(struct sock *sk,
422 struct sk_buff *skb);
423
424extern int tcp_v4_connect(struct sock *sk,
425 struct sockaddr *uaddr,
426 int addr_len);
427
428extern int tcp_connect(struct sock *sk);
429
430extern struct sk_buff * tcp_make_synack(struct sock *sk,
431 struct dst_entry *dst,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700432 struct request_sock *req);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
434extern int tcp_disconnect(struct sock *sk, int flags);
435
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437/* From syncookies.c */
Florian Westphal2051f112008-03-23 22:21:28 -0700438extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
440 struct ip_options *opt);
441extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
442 __u16 *mss);
443
Florian Westphal4dfc2812008-04-10 03:12:40 -0700444extern __u32 cookie_init_timestamp(struct request_sock *req);
445extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt);
446
Glenn Griffinc6aefaf2008-02-07 21:49:26 -0800447/* From net/ipv6/syncookies.c */
448extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
449extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb,
450 __u16 *mss);
451
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452/* tcp_output.c */
453
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700454extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
455 int nonagle);
456extern int tcp_may_send_now(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
458extern void tcp_xmit_retransmit_queue(struct sock *);
459extern void tcp_simple_retransmit(struct sock *);
460extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
David S. Miller6475be12005-09-01 22:47:01 -0700461extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462
463extern void tcp_send_probe0(struct sock *);
464extern void tcp_send_partial(struct sock *);
465extern int tcp_write_wakeup(struct sock *);
466extern void tcp_send_fin(struct sock *sk);
Al Virodd0fc662005-10-07 07:46:04 +0100467extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468extern int tcp_send_synack(struct sock *);
David S. Millerc1b4a7e2005-07-05 15:24:38 -0700469extern void tcp_push_one(struct sock *, unsigned int mss_now);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470extern void tcp_send_ack(struct sock *sk);
471extern void tcp_send_delayed_ack(struct sock *sk);
472
David S. Millera762a982005-07-05 15:18:51 -0700473/* tcp_input.c */
474extern void tcp_cwnd_application_limited(struct sock *sk);
475
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476/* tcp_timer.c */
477extern void tcp_init_xmit_timers(struct sock *);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700478static inline void tcp_clear_xmit_timers(struct sock *sk)
479{
480 inet_csk_clear_xmit_timers(sk);
481}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
484extern unsigned int tcp_current_mss(struct sock *sk, int large);
485
Arnaldo Carvalho de Melo17b085e2005-08-12 12:59:17 -0300486/* tcp.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487extern void tcp_get_info(struct sock *, struct tcp_info *);
488
489/* Read 'sendfile()'-style from a TCP socket */
490typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
491 unsigned int, size_t);
492extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
493 sk_read_actor_t recv_actor);
494
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800495extern void tcp_initialize_rcv_mss(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496
John Heffner5d424d52006-03-20 17:53:41 -0800497extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
498extern int tcp_mss_to_mtu(struct sock *sk, int mss);
499extern void tcp_mtup_init(struct sock *sk);
500
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800501static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502{
503 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
504 ntohl(TCP_FLAG_ACK) |
505 snd_wnd);
506}
507
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800508static inline void tcp_fast_path_on(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509{
510 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
511}
512
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700513static inline void tcp_fast_path_check(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700515 struct tcp_sock *tp = tcp_sk(sk);
516
David S. Millerb03efcf2005-07-08 14:57:23 -0700517 if (skb_queue_empty(&tp->out_of_order_queue) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 tp->rcv_wnd &&
519 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
520 !tp->urg_data)
521 tcp_fast_path_on(tp);
522}
523
524/* Compute the actual receive window we are currently advertising.
525 * Rcv_nxt can be after the window if our peer push more data
526 * than the offered window.
527 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800528static inline u32 tcp_receive_window(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529{
530 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
531
532 if (win < 0)
533 win = 0;
534 return (u32) win;
535}
536
537/* Choose a new window, without checks for shrinking, and without
538 * scaling applied to the result. The caller does these things
539 * if necessary. This is a "raw" window selection.
540 */
541extern u32 __tcp_select_window(struct sock *sk);
542
543/* TCP timestamps are only 32-bits, this causes a slight
544 * complication on 64-bit systems since we store a snapshot
Stephen Hemminger31f34262005-11-15 15:17:10 -0800545 * of jiffies in the buffer control blocks below. We decided
546 * to use only the low 32-bits of jiffies and hide the ugly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 * casts with the following macro.
548 */
549#define tcp_time_stamp ((__u32)(jiffies))
550
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800551/* This is what the send packet queuing engine uses to pass
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 * TCP per-packet control information to the transmission
553 * code. We also store the host-order sequence numbers in
554 * here too. This is 36 bytes on 32-bit architectures,
555 * 40 bytes on 64-bit machines, if this grows please adjust
556 * skbuff.h:skbuff->cb[xxx] size appropriately.
557 */
558struct tcp_skb_cb {
559 union {
560 struct inet_skb_parm h4;
561#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
562 struct inet6_skb_parm h6;
563#endif
564 } header; /* For incoming frames */
565 __u32 seq; /* Starting sequence number */
566 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
567 __u32 when; /* used to compute rtt's */
568 __u8 flags; /* TCP header flags. */
569
570 /* NOTE: These must match up to the flags byte in a
571 * real TCP header.
572 */
573#define TCPCB_FLAG_FIN 0x01
574#define TCPCB_FLAG_SYN 0x02
575#define TCPCB_FLAG_RST 0x04
576#define TCPCB_FLAG_PSH 0x08
577#define TCPCB_FLAG_ACK 0x10
578#define TCPCB_FLAG_URG 0x20
579#define TCPCB_FLAG_ECE 0x40
580#define TCPCB_FLAG_CWR 0x80
581
582 __u8 sacked; /* State flags for SACK/FACK. */
583#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
584#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
585#define TCPCB_LOST 0x04 /* SKB is lost */
586#define TCPCB_TAGBITS 0x07 /* All tag bits */
587
588#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
589#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
590
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 __u16 urg_ptr; /* Valid w/URG flags is set. */
592 __u32 ack_seq; /* Sequence number ACK'd */
593};
594
595#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
596
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597/* Due to TSO, an SKB can be composed of multiple actual
598 * packets. To keep these tracked properly, we use this.
599 */
600static inline int tcp_skb_pcount(const struct sk_buff *skb)
601{
Herbert Xu79671682006-06-22 02:40:14 -0700602 return skb_shinfo(skb)->gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603}
604
605/* This is valid iff tcp_skb_pcount() > 1. */
606static inline int tcp_skb_mss(const struct sk_buff *skb)
607{
Herbert Xu79671682006-06-22 02:40:14 -0700608 return skb_shinfo(skb)->gso_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609}
610
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300611static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612{
613 if (*count) {
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300614 *count -= decr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 if ((int)*count < 0)
616 *count = 0;
617 }
618}
619
Ilpo Järvinenaf610b42007-06-15 12:58:38 +0300620static inline void tcp_dec_pcount_approx(__u32 *count,
621 const struct sk_buff *skb)
622{
623 tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb));
624}
625
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700626/* Events passed to congestion control interface */
627enum tcp_ca_event {
628 CA_EVENT_TX_START, /* first transmit when no packets in flight */
629 CA_EVENT_CWND_RESTART, /* congestion window restart */
630 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
631 CA_EVENT_FRTO, /* fast recovery timeout */
632 CA_EVENT_LOSS, /* loss timeout */
633 CA_EVENT_FAST_ACK, /* in sequence ack */
634 CA_EVENT_SLOW_ACK, /* other ack */
635};
636
637/*
638 * Interface for adding new TCP congestion control handlers
639 */
640#define TCP_CA_NAME_MAX 16
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800641#define TCP_CA_MAX 128
642#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
643
Stephen Hemminger164891a2007-04-23 22:26:16 -0700644#define TCP_CONG_NON_RESTRICTED 0x1
645#define TCP_CONG_RTT_STAMP 0x2
646
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700647struct tcp_congestion_ops {
648 struct list_head list;
Stephen Hemminger164891a2007-04-23 22:26:16 -0700649 unsigned long flags;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700650
651 /* initialize private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300652 void (*init)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700653 /* cleanup private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300654 void (*release)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700655
656 /* return slow start threshold (required) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300657 u32 (*ssthresh)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700658 /* lower bound for congestion window (optional) */
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700659 u32 (*min_cwnd)(const struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700660 /* do new cwnd calculation (required) */
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200661 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700662 /* call before changing ca_state (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300663 void (*set_state)(struct sock *sk, u8 new_state);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700664 /* call when cwnd event occurs (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300665 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700666 /* new value of cwnd after loss (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300667 u32 (*undo_cwnd)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700668 /* hook for packet ack accounting (optional) */
Stephen Hemminger30cfd0b2007-07-25 23:49:34 -0700669 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
Arnaldo Carvalho de Melo73c1f4a2005-08-12 12:51:49 -0300670 /* get info for inet_diag (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300671 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700672
673 char name[TCP_CA_NAME_MAX];
674 struct module *owner;
675};
676
677extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
678extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
679
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300680extern void tcp_init_congestion_control(struct sock *sk);
681extern void tcp_cleanup_congestion_control(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700682extern int tcp_set_default_congestion_control(const char *name);
683extern void tcp_get_default_congestion_control(char *name);
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800684extern void tcp_get_available_congestion_control(char *buf, size_t len);
Stephen Hemmingerce7bc3b2006-11-09 16:35:15 -0800685extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
686extern int tcp_set_allowed_congestion_control(char *allowed);
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300687extern int tcp_set_congestion_control(struct sock *sk, const char *name);
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800688extern void tcp_slow_start(struct tcp_sock *tp);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700689
Stephen Hemminger5f8ef482005-06-23 20:37:36 -0700690extern struct tcp_congestion_ops tcp_init_congestion_ops;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300691extern u32 tcp_reno_ssthresh(struct sock *sk);
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200692extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700693extern u32 tcp_reno_min_cwnd(const struct sock *sk);
David S. Millera8acfba2005-06-23 23:45:02 -0700694extern struct tcp_congestion_ops tcp_reno;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700695
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300696static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700697{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300698 struct inet_connection_sock *icsk = inet_csk(sk);
699
700 if (icsk->icsk_ca_ops->set_state)
701 icsk->icsk_ca_ops->set_state(sk, ca_state);
702 icsk->icsk_ca_state = ca_state;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700703}
704
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300705static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700706{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300707 const struct inet_connection_sock *icsk = inet_csk(sk);
708
709 if (icsk->icsk_ca_ops->cwnd_event)
710 icsk->icsk_ca_ops->cwnd_event(sk, event);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700711}
712
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300713/* These functions determine how the current flow behaves in respect of SACK
714 * handling. SACK is negotiated with the peer, and therefore it can vary
715 * between different flows.
716 *
717 * tcp_is_sack - SACK enabled
718 * tcp_is_reno - No SACK
719 * tcp_is_fack - FACK enabled, implies SACK enabled
720 */
721static inline int tcp_is_sack(const struct tcp_sock *tp)
722{
723 return tp->rx_opt.sack_ok;
724}
725
726static inline int tcp_is_reno(const struct tcp_sock *tp)
727{
728 return !tcp_is_sack(tp);
729}
730
731static inline int tcp_is_fack(const struct tcp_sock *tp)
732{
733 return tp->rx_opt.sack_ok & 2;
734}
735
736static inline void tcp_enable_fack(struct tcp_sock *tp)
737{
738 tp->rx_opt.sack_ok |= 2;
739}
740
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300741static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
742{
743 return tp->sacked_out + tp->lost_out;
744}
745
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746/* This determines how many packets are "in the network" to the best
747 * of our knowledge. In many cases it is conservative, but where
748 * detailed information is available from the receiver (via SACK
749 * blocks etc.) we can make more aggressive calculations.
750 *
751 * Use this for decisions involving congestion control, use just
752 * tp->packets_out to determine if the send queue is empty or not.
753 *
754 * Read this equation as:
755 *
756 * "Packets sent once on transmission queue" MINUS
757 * "Packets left network, but not honestly ACKed yet" PLUS
758 * "Packets fast retransmitted"
759 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800760static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761{
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300762 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763}
764
Ilpo Järvinen882beba2008-04-07 22:33:07 -0700765extern int tcp_limit_reno_sacked(struct tcp_sock *tp);
766
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
768 * The exception is rate halving phase, when cwnd is decreasing towards
769 * ssthresh.
770 */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300771static inline __u32 tcp_current_ssthresh(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300773 const struct tcp_sock *tp = tcp_sk(sk);
774 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 return tp->snd_ssthresh;
776 else
777 return max(tp->snd_ssthresh,
778 ((tp->snd_cwnd >> 1) +
779 (tp->snd_cwnd >> 2)));
780}
781
Ilpo Järvinenb9c45952007-07-27 16:36:17 +0300782/* Use define here intentionally to get WARN_ON location shown at the caller */
783#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800785extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
787
788/* Slow start with delack produces 3 packets of burst, so that
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700789 * it is safe "de facto". This will be the default - same as
790 * the default reordering threshold - but if reordering increases,
791 * we must be able to allow cwnd to burst at least this much in order
792 * to not pull it back when holes are filled.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 */
794static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
795{
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700796 return tp->reordering;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797}
798
Ilpo Järvinen90840de2007-12-31 04:48:41 -0800799/* Returns end sequence number of the receiver's advertised window */
800static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
801{
802 return tp->snd_una + tp->snd_wnd;
803}
Ilpo Järvinencea14e02008-01-12 03:19:12 -0800804extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800805
Chuck Leverc1bd24b2007-10-23 21:08:54 -0700806static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800807 const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
809 if (skb->len < mss)
810 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
811}
812
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700813static inline void tcp_check_probe_timer(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700815 struct tcp_sock *tp = tcp_sk(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700816 const struct inet_connection_sock *icsk = inet_csk(sk);
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700817
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700818 if (!tp->packets_out && !icsk->icsk_pending)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700819 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
820 icsk->icsk_rto, TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821}
822
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700823static inline void tcp_push_pending_frames(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700825 struct tcp_sock *tp = tcp_sk(sk);
826
827 __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828}
829
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800830static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831{
832 tp->snd_wl1 = seq;
833}
834
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800835static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836{
837 tp->snd_wl1 = seq;
838}
839
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840/*
841 * Calculate(/check) TCP checksum
842 */
Frederik Deweerdtba7808e2007-02-04 20:15:27 -0800843static inline __sum16 tcp_v4_check(int len, __be32 saddr,
844 __be32 daddr, __wsum base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845{
846 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
847}
848
Al Virob51655b2006-11-14 21:40:42 -0800849static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850{
Herbert Xufb286bb2005-11-10 13:01:24 -0800851 return __skb_checksum_complete(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852}
853
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800854static inline int tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855{
Herbert Xu60476372007-04-09 11:59:39 -0700856 return !skb_csum_unnecessary(skb) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 __tcp_checksum_complete(skb);
858}
859
860/* Prequeue for VJ style copy to user, combined with checksumming. */
861
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800862static inline void tcp_prequeue_init(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863{
864 tp->ucopy.task = NULL;
865 tp->ucopy.len = 0;
866 tp->ucopy.memory = 0;
867 skb_queue_head_init(&tp->ucopy.prequeue);
Chris Leech97fc2f02006-05-23 17:55:33 -0700868#ifdef CONFIG_NET_DMA
869 tp->ucopy.dma_chan = NULL;
870 tp->ucopy.wakeup = 0;
871 tp->ucopy.pinned_list = NULL;
872 tp->ucopy.dma_cookie = 0;
873#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874}
875
876/* Packet is added to VJ-style prequeue for processing in process
877 * context, if a reader task is waiting. Apparently, this exciting
878 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
879 * failed somewhere. Latency? Burstiness? Well, at least now we will
880 * see, why it failed. 8)8) --ANK
881 *
882 * NOTE: is this not too big to inline?
883 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800884static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885{
886 struct tcp_sock *tp = tcp_sk(sk);
887
888 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
889 __skb_queue_tail(&tp->ucopy.prequeue, skb);
890 tp->ucopy.memory += skb->truesize;
891 if (tp->ucopy.memory > sk->sk_rcvbuf) {
892 struct sk_buff *skb1;
893
894 BUG_ON(sock_owned_by_user(sk));
895
896 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
897 sk->sk_backlog_rcv(sk, skb1);
Pavel Emelyanovde0744a2008-07-16 20:31:16 -0700898 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPPREQUEUEDROPPED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 }
900
901 tp->ucopy.memory = 0;
902 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
903 wake_up_interruptible(sk->sk_sleep);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700904 if (!inet_csk_ack_scheduled(sk))
905 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700906 (3 * TCP_RTO_MIN) / 4,
907 TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 }
909 return 1;
910 }
911 return 0;
912}
913
914
915#undef STATE_TRACE
916
917#ifdef STATE_TRACE
918static const char *statename[]={
919 "Unused","Established","Syn Sent","Syn Recv",
920 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
921 "Close Wait","Last ACK","Listen","Closing"
922};
923#endif
Ilpo Järvinen490d5042008-01-12 03:17:20 -0800924extern void tcp_set_state(struct sock *sk, int state);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925
Andi Kleen4ac02ba2007-04-20 17:11:46 -0700926extern void tcp_done(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800928static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929{
930 rx_opt->dsack = 0;
931 rx_opt->eff_sacks = 0;
932 rx_opt->num_sacks = 0;
933}
934
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935/* Determine a window scaling and initial window to offer. */
936extern void tcp_select_initial_window(int __space, __u32 mss,
937 __u32 *rcv_wnd, __u32 *window_clamp,
938 int wscale_ok, __u8 *rcv_wscale);
939
940static inline int tcp_win_from_space(int space)
941{
942 return sysctl_tcp_adv_win_scale<=0 ?
943 (space>>(-sysctl_tcp_adv_win_scale)) :
944 space - (space>>sysctl_tcp_adv_win_scale);
945}
946
947/* Note: caller must be prepared to deal with negative returns */
948static inline int tcp_space(const struct sock *sk)
949{
950 return tcp_win_from_space(sk->sk_rcvbuf -
951 atomic_read(&sk->sk_rmem_alloc));
952}
953
954static inline int tcp_full_space(const struct sock *sk)
955{
956 return tcp_win_from_space(sk->sk_rcvbuf);
957}
958
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800959static inline void tcp_openreq_init(struct request_sock *req,
960 struct tcp_options_received *rx_opt,
961 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700963 struct inet_request_sock *ireq = inet_rsk(req);
964
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
Florian Westphal4dfc2812008-04-10 03:12:40 -0700966 req->cookie_ts = 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700967 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 req->mss = rx_opt->mss_clamp;
969 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700970 ireq->tstamp_ok = rx_opt->tstamp_ok;
971 ireq->sack_ok = rx_opt->sack_ok;
972 ireq->snd_wscale = rx_opt->snd_wscale;
973 ireq->wscale_ok = rx_opt->wscale_ok;
974 ireq->acked = 0;
975 ireq->ecn_ok = 0;
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -0700976 ireq->rmt_port = tcp_hdr(skb)->source;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977}
978
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700979extern void tcp_enter_memory_pressure(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981static inline int keepalive_intvl_when(const struct tcp_sock *tp)
982{
983 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
984}
985
986static inline int keepalive_time_when(const struct tcp_sock *tp)
987{
988 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
989}
990
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700991static inline int tcp_fin_time(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700993 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
994 const int rto = inet_csk(sk)->icsk_rto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700996 if (fin_timeout < (rto << 2) - (rto >> 1))
997 fin_timeout = (rto << 2) - (rto >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998
999 return fin_timeout;
1000}
1001
1002static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1003{
1004 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1005 return 0;
James Morris9d729f72007-03-04 16:12:44 -08001006 if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 return 0;
1008
1009 /* RST segments are not recommended to carry timestamp,
1010 and, if they do, it is recommended to ignore PAWS because
1011 "their cleanup function should take precedence over timestamps."
1012 Certainly, it is mistake. It is necessary to understand the reasons
1013 of this constraint to relax it: if peer reboots, clock may go
1014 out-of-sync and half-open connections will not be reset.
1015 Actually, the problem would be not existing if all
1016 the implementations followed draft about maintaining clock
1017 via reboots. Linux-2.2 DOES NOT!
1018
1019 However, we can relax time bounds for RST segments to MSL.
1020 */
James Morris9d729f72007-03-04 16:12:44 -08001021 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 return 0;
1023 return 1;
1024}
1025
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026#define TCP_CHECK_TIMER(sk) do { } while (0)
1027
Pavel Emelyanova9c19322008-07-16 20:21:42 -07001028static inline void tcp_mib_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029{
1030 /* See RFC 2012 */
Pavel Emelyanovcf1100a2008-07-16 20:27:38 -07001031 TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1032 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1033 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1034 TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035}
1036
Ilpo Järvinen5af4ec22007-09-20 11:30:48 -07001037/* from STCP */
David S. Miller0800f172007-09-20 11:40:37 -07001038static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1039{
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001040 tp->lost_skb_hint = NULL;
1041 tp->scoreboard_skb_hint = NULL;
1042 tp->retransmit_skb_hint = NULL;
1043 tp->forward_skb_hint = NULL;
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001044}
1045
David S. Miller0800f172007-09-20 11:40:37 -07001046static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1047{
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001048 tcp_clear_retrans_hints_partial(tp);
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001049}
1050
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001051/* MD5 Signature */
1052struct crypto_hash;
1053
1054/* - key database */
1055struct tcp_md5sig_key {
1056 u8 *key;
1057 u8 keylen;
1058};
1059
1060struct tcp4_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001061 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001062 __be32 addr;
1063};
1064
1065struct tcp6_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001066 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001067#if 0
1068 u32 scope_id; /* XXX */
1069#endif
1070 struct in6_addr addr;
1071};
1072
1073/* - sock block */
1074struct tcp_md5sig_info {
1075 struct tcp4_md5sig_key *keys4;
1076#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1077 struct tcp6_md5sig_key *keys6;
1078 u32 entries6;
1079 u32 alloced6;
1080#endif
1081 u32 entries4;
1082 u32 alloced4;
1083};
1084
1085/* - pseudo header */
1086struct tcp4_pseudohdr {
1087 __be32 saddr;
1088 __be32 daddr;
1089 __u8 pad;
1090 __u8 protocol;
1091 __be16 len;
1092};
1093
1094struct tcp6_pseudohdr {
1095 struct in6_addr saddr;
1096 struct in6_addr daddr;
1097 __be32 len;
1098 __be32 protocol; /* including padding */
1099};
1100
1101union tcp_md5sum_block {
1102 struct tcp4_pseudohdr ip4;
1103#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1104 struct tcp6_pseudohdr ip6;
1105#endif
1106};
1107
1108/* - pool: digest algorithm, hash description and scratch buffer */
1109struct tcp_md5sig_pool {
1110 struct hash_desc md5_desc;
1111 union tcp_md5sum_block md5_blk;
1112};
1113
1114#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1115
1116/* - functions */
Adam Langley49a72df2008-07-19 00:01:42 -07001117extern int tcp_v4_md5_hash_skb(char *md5_hash,
1118 struct tcp_md5sig_key *key,
1119 struct sock *sk,
1120 struct request_sock *req,
1121 struct sk_buff *skb);
YOSHIFUJI Hideaki8d26d762008-04-17 13:19:16 +09001122
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001123extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1124 struct sock *addr_sk);
1125
1126extern int tcp_v4_md5_do_add(struct sock *sk,
1127 __be32 addr,
1128 u8 *newkey,
1129 u8 newkeylen);
1130
1131extern int tcp_v4_md5_do_del(struct sock *sk,
Al Viro8e5200f2006-11-20 18:06:37 -08001132 __be32 addr);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001133
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +09001134#ifdef CONFIG_TCP_MD5SIG
1135#define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_keylen ? \
1136 &(struct tcp_md5sig_key) { \
1137 .key = (twsk)->tw_md5_key, \
1138 .keylen = (twsk)->tw_md5_keylen, \
1139 } : NULL)
1140#else
1141#define tcp_twsk_md5_key(twsk) NULL
1142#endif
1143
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001144extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1145extern void tcp_free_md5sig_pool(void);
1146
1147extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1148extern void __tcp_put_md5sig_pool(void);
Adam Langley49a72df2008-07-19 00:01:42 -07001149extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, struct tcphdr *);
1150extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, struct sk_buff *,
1151 unsigned header_len);
1152extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
1153 struct tcp_md5sig_key *key);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001154
1155static inline
1156struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1157{
1158 int cpu = get_cpu();
1159 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1160 if (!ret)
1161 put_cpu();
1162 return ret;
1163}
1164
1165static inline void tcp_put_md5sig_pool(void)
1166{
1167 __tcp_put_md5sig_pool();
1168 put_cpu();
1169}
1170
David S. Millerfe067e82007-03-07 12:12:44 -08001171/* write queue abstraction */
1172static inline void tcp_write_queue_purge(struct sock *sk)
1173{
1174 struct sk_buff *skb;
1175
1176 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001177 sk_wmem_free_skb(sk, skb);
1178 sk_mem_reclaim(sk);
David S. Millerfe067e82007-03-07 12:12:44 -08001179}
1180
1181static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1182{
1183 struct sk_buff *skb = sk->sk_write_queue.next;
1184 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1185 return NULL;
1186 return skb;
1187}
1188
1189static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1190{
1191 struct sk_buff *skb = sk->sk_write_queue.prev;
1192 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1193 return NULL;
1194 return skb;
1195}
1196
1197static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1198{
1199 return skb->next;
1200}
1201
1202#define tcp_for_write_queue(skb, sk) \
1203 for (skb = (sk)->sk_write_queue.next; \
1204 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1205 skb = skb->next)
1206
1207#define tcp_for_write_queue_from(skb, sk) \
1208 for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\
1209 skb = skb->next)
1210
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001211#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
1212 for (tmp = skb->next; \
1213 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1214 skb = tmp, tmp = skb->next)
1215
David S. Millerfe067e82007-03-07 12:12:44 -08001216static inline struct sk_buff *tcp_send_head(struct sock *sk)
1217{
1218 return sk->sk_send_head;
1219}
1220
1221static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1222{
1223 sk->sk_send_head = skb->next;
1224 if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue)
1225 sk->sk_send_head = NULL;
1226}
1227
1228static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1229{
1230 if (sk->sk_send_head == skb_unlinked)
1231 sk->sk_send_head = NULL;
1232}
1233
1234static inline void tcp_init_send_head(struct sock *sk)
1235{
1236 sk->sk_send_head = NULL;
1237}
1238
1239static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1240{
1241 __skb_queue_tail(&sk->sk_write_queue, skb);
1242}
1243
1244static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1245{
1246 __tcp_add_write_queue_tail(sk, skb);
1247
1248 /* Queue it, remembering where we must start sending. */
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001249 if (sk->sk_send_head == NULL) {
David S. Millerfe067e82007-03-07 12:12:44 -08001250 sk->sk_send_head = skb;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001251
1252 if (tcp_sk(sk)->highest_sack == NULL)
1253 tcp_sk(sk)->highest_sack = skb;
1254 }
David S. Millerfe067e82007-03-07 12:12:44 -08001255}
1256
1257static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1258{
1259 __skb_queue_head(&sk->sk_write_queue, skb);
1260}
1261
1262/* Insert buff after skb on the write queue of sk. */
1263static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1264 struct sk_buff *buff,
1265 struct sock *sk)
1266{
Gerrit Renker7de6c032008-04-14 00:05:09 -07001267 __skb_queue_after(&sk->sk_write_queue, skb, buff);
David S. Millerfe067e82007-03-07 12:12:44 -08001268}
1269
1270/* Insert skb between prev and next on the write queue of sk. */
1271static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1272 struct sk_buff *skb,
1273 struct sock *sk)
1274{
1275 __skb_insert(new, skb->prev, skb, &sk->sk_write_queue);
Ilpo Järvinen6e421412007-11-19 23:24:09 -08001276
1277 if (sk->sk_send_head == skb)
1278 sk->sk_send_head = new;
David S. Millerfe067e82007-03-07 12:12:44 -08001279}
1280
1281static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1282{
1283 __skb_unlink(skb, &sk->sk_write_queue);
1284}
1285
1286static inline int tcp_skb_is_last(const struct sock *sk,
1287 const struct sk_buff *skb)
1288{
1289 return skb->next == (struct sk_buff *)&sk->sk_write_queue;
1290}
1291
1292static inline int tcp_write_queue_empty(struct sock *sk)
1293{
1294 return skb_queue_empty(&sk->sk_write_queue);
1295}
1296
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001297/* Start sequence of the highest skb with SACKed bit, valid only if
1298 * sacked > 0 or when the caller has ensured validity by itself.
1299 */
1300static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1301{
1302 if (!tp->sacked_out)
1303 return tp->snd_una;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001304
1305 if (tp->highest_sack == NULL)
1306 return tp->snd_nxt;
1307
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001308 return TCP_SKB_CB(tp->highest_sack)->seq;
1309}
1310
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001311static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1312{
1313 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1314 tcp_write_queue_next(sk, skb);
1315}
1316
1317static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1318{
1319 return tcp_sk(sk)->highest_sack;
1320}
1321
1322static inline void tcp_highest_sack_reset(struct sock *sk)
1323{
1324 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1325}
1326
1327/* Called when old skb is about to be deleted (to be combined with new skb) */
1328static inline void tcp_highest_sack_combine(struct sock *sk,
1329 struct sk_buff *old,
1330 struct sk_buff *new)
1331{
1332 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1333 tcp_sk(sk)->highest_sack = new;
1334}
1335
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336/* /proc */
1337enum tcp_seq_states {
1338 TCP_SEQ_STATE_LISTENING,
1339 TCP_SEQ_STATE_OPENREQ,
1340 TCP_SEQ_STATE_ESTABLISHED,
1341 TCP_SEQ_STATE_TIME_WAIT,
1342};
1343
1344struct tcp_seq_afinfo {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 char *name;
1346 sa_family_t family;
Denis V. Lunev68fcadd2008-04-13 22:13:30 -07001347 struct file_operations seq_fops;
Denis V. Lunev9427c4b2008-04-13 22:12:13 -07001348 struct seq_operations seq_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349};
1350
1351struct tcp_iter_state {
Denis V. Luneva4146b12008-04-13 22:11:14 -07001352 struct seq_net_private p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 sa_family_t family;
1354 enum tcp_seq_states state;
1355 struct sock *syn_wait_sk;
1356 int bucket, sbucket, num, uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357};
1358
Daniel Lezcano6f8b13b2008-03-21 04:14:45 -07001359extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1360extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001362extern struct request_sock_ops tcp_request_sock_ops;
Glenn Griffinc6aefaf2008-02-07 21:49:26 -08001363extern struct request_sock_ops tcp6_request_sock_ops;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001364
Brian Haley7d06b2e2008-06-14 17:04:49 -07001365extern void tcp_v4_destroy_sock(struct sock *sk);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001366
Herbert Xua430a432006-07-08 13:34:56 -07001367extern int tcp_v4_gso_send_check(struct sk_buff *skb);
Herbert Xu576a30e2006-06-27 13:22:38 -07001368extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001369
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001370#ifdef CONFIG_PROC_FS
1371extern int tcp4_proc_init(void);
1372extern void tcp4_proc_exit(void);
1373#endif
1374
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001375/* TCP af-specific functions */
1376struct tcp_sock_af_ops {
1377#ifdef CONFIG_TCP_MD5SIG
1378 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1379 struct sock *addr_sk);
1380 int (*calc_md5_hash) (char *location,
1381 struct tcp_md5sig_key *md5,
1382 struct sock *sk,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001383 struct request_sock *req,
Adam Langley49a72df2008-07-19 00:01:42 -07001384 struct sk_buff *skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001385 int (*md5_add) (struct sock *sk,
1386 struct sock *addr_sk,
1387 u8 *newkey,
1388 u8 len);
1389 int (*md5_parse) (struct sock *sk,
1390 char __user *optval,
1391 int optlen);
1392#endif
1393};
1394
1395struct tcp_request_sock_ops {
1396#ifdef CONFIG_TCP_MD5SIG
1397 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1398 struct request_sock *req);
1399#endif
1400};
1401
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001402extern void tcp_v4_init(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001403extern void tcp_init(void);
1404
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405#endif /* _TCP_H */