blob: 5740b85bc5a0978349eb042caf17ec6255c98158 [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>
William Allen Simpson435cf552009-12-02 18:17:05 +000033#include <linux/kref.h>
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070034
35#include <net/inet_connection_sock.h>
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -070036#include <net/inet_timewait_sock.h>
Arnaldo Carvalho de Melo77d8bf92005-08-09 20:00:51 -070037#include <net/inet_hashtables.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#include <net/checksum.h>
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -070039#include <net/request_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070040#include <net/sock.h>
41#include <net/snmp.h>
42#include <net/ip.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070043#include <net/tcp_states.h>
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -070044#include <net/inet_ecn.h>
Satoru SATOH0c266892009-05-04 11:11:01 -070045#include <net/dst.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070046
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#include <linux/seq_file.h>
48
Arnaldo Carvalho de Melo6e04e022005-08-09 20:07:35 -070049extern struct inet_hashinfo tcp_hashinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
Eric Dumazetdd24c002008-11-25 21:17:14 -080051extern struct percpu_counter tcp_orphan_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -070052extern void tcp_time_wait(struct sock *sk, int state, int timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
Linus Torvalds1da177e2005-04-16 15:20:36 -070054#define MAX_TCP_HEADER (128 + MAX_HEADER)
Adam Langley33ad7982008-07-19 00:04:31 -070055#define MAX_TCP_OPTION_SPACE 40
Linus Torvalds1da177e2005-04-16 15:20:36 -070056
57/*
58 * Never offer a window over 32767 without using window scaling. Some
59 * poor stacks do signed 16bit maths!
60 */
61#define MAX_TCP_WINDOW 32767U
62
63/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
64#define TCP_MIN_MSS 88U
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
142
143#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144
145#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
146#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
147 * after this time. It should be equal
148 * (or greater than) TCP_TIMEWAIT_LEN
149 * to provide reliability equal to one
150 * provided by timewait state.
151 */
152#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
153 * timestamps. It must be less than
154 * minimal timewait lifetime.
155 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156/*
157 * TCP option
158 */
159
160#define TCPOPT_NOP 1 /* Padding */
161#define TCPOPT_EOL 0 /* End of options */
162#define TCPOPT_MSS 2 /* Segment size negotiating */
163#define TCPOPT_WINDOW 3 /* Window scaling */
164#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
165#define TCPOPT_SACK 5 /* SACK Block */
166#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800167#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
William Allen Simpson435cf552009-12-02 18:17:05 +0000168#define TCPOPT_COOKIE 253 /* Cookie extension (experimental) */
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
William Allen Simpson435cf552009-12-02 18:17:05 +0000179#define TCPOLEN_COOKIE_BASE 2 /* Cookie-less header extension */
180#define TCPOLEN_COOKIE_PAIR 3 /* Cookie pair header extension */
181#define TCPOLEN_COOKIE_MIN (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MIN)
182#define TCPOLEN_COOKIE_MAX (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183
184/* But this is what stacks really send out. */
185#define TCPOLEN_TSTAMP_ALIGNED 12
186#define TCPOLEN_WSCALE_ALIGNED 4
187#define TCPOLEN_SACKPERM_ALIGNED 4
188#define TCPOLEN_SACK_BASE 2
189#define TCPOLEN_SACK_BASE_ALIGNED 4
190#define TCPOLEN_SACK_PERBLOCK 8
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800191#define TCPOLEN_MD5SIG_ALIGNED 20
Adam Langley33ad7982008-07-19 00:04:31 -0700192#define TCPOLEN_MSS_ALIGNED 4
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194/* Flags in tp->nonagle */
195#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
196#define TCP_NAGLE_CORK 2 /* Socket is corked */
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800197#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -0700199extern struct inet_timewait_death_row tcp_death_row;
200
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201/* sysctl variables for tcp */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202extern int sysctl_tcp_timestamps;
203extern int sysctl_tcp_window_scaling;
204extern int sysctl_tcp_sack;
205extern int sysctl_tcp_fin_timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206extern int sysctl_tcp_keepalive_time;
207extern int sysctl_tcp_keepalive_probes;
208extern int sysctl_tcp_keepalive_intvl;
209extern int sysctl_tcp_syn_retries;
210extern int sysctl_tcp_synack_retries;
211extern int sysctl_tcp_retries1;
212extern int sysctl_tcp_retries2;
213extern int sysctl_tcp_orphan_retries;
214extern int sysctl_tcp_syncookies;
215extern int sysctl_tcp_retrans_collapse;
216extern int sysctl_tcp_stdurg;
217extern int sysctl_tcp_rfc1337;
218extern int sysctl_tcp_abort_on_overflow;
219extern int sysctl_tcp_max_orphans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220extern int sysctl_tcp_fack;
221extern int sysctl_tcp_reordering;
222extern int sysctl_tcp_ecn;
223extern int sysctl_tcp_dsack;
224extern int sysctl_tcp_mem[3];
225extern int sysctl_tcp_wmem[3];
226extern int sysctl_tcp_rmem[3];
227extern int sysctl_tcp_app_win;
228extern int sysctl_tcp_adv_win_scale;
229extern int sysctl_tcp_tw_reuse;
230extern int sysctl_tcp_frto;
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800231extern int sysctl_tcp_frto_response;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232extern int sysctl_tcp_low_latency;
Chris Leech95937822006-05-23 18:02:55 -0700233extern int sysctl_tcp_dma_copybreak;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234extern int sysctl_tcp_nometrics_save;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235extern int sysctl_tcp_moderate_rcvbuf;
236extern int sysctl_tcp_tso_win_divisor;
Stephen Hemminger9772efb2005-11-10 17:09:53 -0800237extern int sysctl_tcp_abc;
John Heffner5d424d52006-03-20 17:53:41 -0800238extern int sysctl_tcp_mtu_probing;
239extern int sysctl_tcp_base_mss;
Rick Jones15d99e02006-03-20 22:40:29 -0800240extern int sysctl_tcp_workaround_signed_windows;
David S. Miller35089bb2006-06-13 22:33:04 -0700241extern int sysctl_tcp_slow_start_after_idle;
John Heffner886236c2007-03-25 19:21:45 -0700242extern int sysctl_tcp_max_ssthresh;
William Allen Simpson519855c2009-12-02 18:14:19 +0000243extern int sysctl_tcp_cookie_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
245extern atomic_t tcp_memory_allocated;
Eric Dumazet17483762008-11-25 21:16:35 -0800246extern struct percpu_counter tcp_sockets_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247extern int tcp_memory_pressure;
248
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 * The next routines deal with comparing 32 bit unsigned ints
251 * and worry about wraparound (automatic with unsigned arithmetic).
252 */
253
254static inline int before(__u32 seq1, __u32 seq2)
255{
Gerrit Renker0d630cc2007-01-04 12:25:16 -0800256 return (__s32)(seq1-seq2) < 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257}
Gerrit Renker9a036b9c2006-12-20 10:25:55 -0800258#define after(seq2, seq1) before(seq1, seq2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259
260/* is s2<=s1<=s3 ? */
261static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
262{
263 return seq3 - seq2 >= seq1 - seq2;
264}
265
Pavel Emelianove4fd5da2007-05-29 13:19:18 -0700266static inline int tcp_too_many_orphans(struct sock *sk, int num)
267{
268 return (num > sysctl_tcp_max_orphans) ||
269 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
270 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
271}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
Florian Westphala0f82f62009-04-19 09:43:48 +0000273/* syncookies: remember time of last synqueue overflow */
274static inline void tcp_synq_overflow(struct sock *sk)
275{
276 tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies;
277}
278
279/* syncookies: no recent synqueue overflow on this listening socket? */
280static inline int tcp_synq_no_recent_overflow(const struct sock *sk)
281{
282 unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp;
283 return time_after(jiffies, last_overflow + TCP_TIMEOUT_INIT);
284}
285
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286extern struct proto tcp_prot;
287
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -0700288#define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field)
289#define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
290#define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
291#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 -0700292
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293extern void tcp_v4_err(struct sk_buff *skb, u32);
294
295extern void tcp_shutdown (struct sock *sk, int how);
296
297extern int tcp_v4_rcv(struct sk_buff *skb);
298
299extern int tcp_v4_remember_stamp(struct sock *sk);
300
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700301extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
David S. Miller3516ffb2007-08-02 19:23:56 -0700303extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 struct msghdr *msg, size_t size);
305extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
306
307extern int tcp_ioctl(struct sock *sk,
308 int cmd,
309 unsigned long arg);
310
311extern int tcp_rcv_state_process(struct sock *sk,
312 struct sk_buff *skb,
313 struct tcphdr *th,
314 unsigned len);
315
316extern int tcp_rcv_established(struct sock *sk,
317 struct sk_buff *skb,
318 struct tcphdr *th,
319 unsigned len);
320
321extern void tcp_rcv_space_adjust(struct sock *sk);
322
Chris Leech0e4b4992006-05-23 18:00:16 -0700323extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
324
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800325extern int tcp_twsk_unique(struct sock *sk,
326 struct sock *sktw, void *twp);
327
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800328extern void tcp_twsk_destructor(struct sock *sk);
329
Jens Axboe9c55e012007-11-06 23:30:13 -0800330extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
331 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
332
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700333static inline void tcp_dec_quickack_mode(struct sock *sk,
334 const unsigned int pkts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700336 struct inet_connection_sock *icsk = inet_csk(sk);
David S. Millerfc6415bc2005-07-05 15:17:45 -0700337
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700338 if (icsk->icsk_ack.quick) {
339 if (pkts >= icsk->icsk_ack.quick) {
340 icsk->icsk_ack.quick = 0;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700341 /* Leaving quickack mode we deflate ATO. */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700342 icsk->icsk_ack.ato = TCP_ATO_MIN;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700343 } else
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700344 icsk->icsk_ack.quick -= pkts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 }
346}
347
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700348extern void tcp_enter_quickack_mode(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -0700350#define TCP_ECN_OK 1
351#define TCP_ECN_QUEUE_CWR 2
352#define TCP_ECN_DEMAND_CWR 4
353
354static __inline__ void
355TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
356{
357 if (sysctl_tcp_ecn && th->ece && th->cwr)
358 inet_rsk(req)->ecn_ok = 1;
359}
360
Eric Dumazetfd2c3ef2009-11-03 03:26:03 +0000361enum tcp_tw_status {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 TCP_TW_SUCCESS = 0,
363 TCP_TW_RST = 1,
364 TCP_TW_ACK = 2,
365 TCP_TW_SYN = 3
366};
367
368
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700369extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 struct sk_buff *skb,
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700371 const struct tcphdr *th);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
373extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700374 struct request_sock *req,
375 struct request_sock **prev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376extern int tcp_child_process(struct sock *parent,
377 struct sock *child,
378 struct sk_buff *skb);
Ilpo Järvinen46d0de42007-02-21 23:10:39 -0800379extern int tcp_use_frto(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380extern void tcp_enter_frto(struct sock *sk);
381extern void tcp_enter_loss(struct sock *sk, int how);
382extern void tcp_clear_retrans(struct tcp_sock *tp);
383extern void tcp_update_metrics(struct sock *sk);
384
385extern void tcp_close(struct sock *sk,
386 long timeout);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
388
389extern int tcp_getsockopt(struct sock *sk, int level,
390 int optname,
391 char __user *optval,
392 int __user *optlen);
393extern int tcp_setsockopt(struct sock *sk, int level,
394 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700395 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800396extern int compat_tcp_getsockopt(struct sock *sk,
397 int level, int optname,
398 char __user *optval, int __user *optlen);
399extern int compat_tcp_setsockopt(struct sock *sk,
400 int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -0700401 char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402extern void tcp_set_keepalive(struct sock *sk, int val);
403extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
404 struct msghdr *msg,
405 size_t len, int nonblock,
406 int flags, int *addr_len);
407
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408extern void tcp_parse_options(struct sk_buff *skb,
409 struct tcp_options_received *opt_rx,
William Allen Simpson4957faade2009-12-02 18:25:27 +0000410 u8 **hvpp,
Gilad Ben-Yossef022c3f72009-10-28 04:15:22 +0000411 int estab,
412 struct dst_entry *dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413
YOSHIFUJI Hideaki7d5d5522008-04-17 12:29:53 +0900414extern u8 *tcp_parse_md5sig_option(struct tcphdr *th);
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416/*
417 * TCP v4 functions exported for the inet6 API
418 */
419
Arnaldo Carvalho de Melo8292a172005-12-13 23:15:52 -0800420extern void tcp_v4_send_check(struct sock *sk, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 struct sk_buff *skb);
422
423extern int tcp_v4_conn_request(struct sock *sk,
424 struct sk_buff *skb);
425
426extern struct sock * tcp_create_openreq_child(struct sock *sk,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700427 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 struct sk_buff *skb);
429
430extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
431 struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700432 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 struct dst_entry *dst);
434
435extern int tcp_v4_do_rcv(struct sock *sk,
436 struct sk_buff *skb);
437
438extern int tcp_v4_connect(struct sock *sk,
439 struct sockaddr *uaddr,
440 int addr_len);
441
442extern int tcp_connect(struct sock *sk);
443
444extern struct sk_buff * tcp_make_synack(struct sock *sk,
445 struct dst_entry *dst,
William Allen Simpsone6b4d112009-12-02 18:07:39 +0000446 struct request_sock *req,
447 struct request_values *rvp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448
449extern int tcp_disconnect(struct sock *sk, int flags);
450
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452/* From syncookies.c */
Florian Westphal2051f112008-03-23 22:21:28 -0700453extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
455 struct ip_options *opt);
456extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
457 __u16 *mss);
458
Florian Westphal4dfc2812008-04-10 03:12:40 -0700459extern __u32 cookie_init_timestamp(struct request_sock *req);
460extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt);
461
Glenn Griffinc6aefaf2008-02-07 21:49:26 -0800462/* From net/ipv6/syncookies.c */
463extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
464extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb,
465 __u16 *mss);
466
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467/* tcp_output.c */
468
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700469extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
470 int nonagle);
471extern int tcp_may_send_now(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000473extern void tcp_retransmit_timer(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474extern void tcp_xmit_retransmit_queue(struct sock *);
475extern void tcp_simple_retransmit(struct sock *);
476extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
David S. Miller6475be12005-09-01 22:47:01 -0700477extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478
479extern void tcp_send_probe0(struct sock *);
480extern void tcp_send_partial(struct sock *);
481extern int tcp_write_wakeup(struct sock *);
482extern void tcp_send_fin(struct sock *sk);
Al Virodd0fc662005-10-07 07:46:04 +0100483extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484extern int tcp_send_synack(struct sock *);
David S. Millerc1b4a7e2005-07-05 15:24:38 -0700485extern void tcp_push_one(struct sock *, unsigned int mss_now);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486extern void tcp_send_ack(struct sock *sk);
487extern void tcp_send_delayed_ack(struct sock *sk);
488
David S. Millera762a982005-07-05 15:18:51 -0700489/* tcp_input.c */
490extern void tcp_cwnd_application_limited(struct sock *sk);
491
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492/* tcp_timer.c */
493extern void tcp_init_xmit_timers(struct sock *);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700494static inline void tcp_clear_xmit_timers(struct sock *sk)
495{
496 inet_csk_clear_xmit_timers(sk);
497}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
Ilpo Järvinen0c54b852009-03-14 14:23:05 +0000500extern unsigned int tcp_current_mss(struct sock *sk);
501
502/* Bound MSS / TSO packet size with the half of the window */
503static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
504{
505 if (tp->max_window && pktsize > (tp->max_window >> 1))
506 return max(tp->max_window >> 1, 68U - tp->tcp_header_len);
507 else
508 return pktsize;
509}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510
Arnaldo Carvalho de Melo17b085e2005-08-12 12:59:17 -0300511/* tcp.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512extern void tcp_get_info(struct sock *, struct tcp_info *);
513
514/* Read 'sendfile()'-style from a TCP socket */
515typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
516 unsigned int, size_t);
517extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
518 sk_read_actor_t recv_actor);
519
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800520extern void tcp_initialize_rcv_mss(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
John Heffner5d424d52006-03-20 17:53:41 -0800522extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
523extern int tcp_mss_to_mtu(struct sock *sk, int mss);
524extern void tcp_mtup_init(struct sock *sk);
525
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000526static inline void tcp_bound_rto(const struct sock *sk)
527{
528 if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX)
529 inet_csk(sk)->icsk_rto = TCP_RTO_MAX;
530}
531
532static inline u32 __tcp_set_rto(const struct tcp_sock *tp)
533{
534 return (tp->srtt >> 3) + tp->rttvar;
535}
536
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800537static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538{
539 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
540 ntohl(TCP_FLAG_ACK) |
541 snd_wnd);
542}
543
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800544static inline void tcp_fast_path_on(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545{
546 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
547}
548
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700549static inline void tcp_fast_path_check(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700551 struct tcp_sock *tp = tcp_sk(sk);
552
David S. Millerb03efcf2005-07-08 14:57:23 -0700553 if (skb_queue_empty(&tp->out_of_order_queue) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 tp->rcv_wnd &&
555 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
556 !tp->urg_data)
557 tcp_fast_path_on(tp);
558}
559
Satoru SATOH0c266892009-05-04 11:11:01 -0700560/* Compute the actual rto_min value */
561static inline u32 tcp_rto_min(struct sock *sk)
562{
563 struct dst_entry *dst = __sk_dst_get(sk);
564 u32 rto_min = TCP_RTO_MIN;
565
566 if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
567 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
568 return rto_min;
569}
570
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571/* Compute the actual receive window we are currently advertising.
572 * Rcv_nxt can be after the window if our peer push more data
573 * than the offered window.
574 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800575static inline u32 tcp_receive_window(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576{
577 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
578
579 if (win < 0)
580 win = 0;
581 return (u32) win;
582}
583
584/* Choose a new window, without checks for shrinking, and without
585 * scaling applied to the result. The caller does these things
586 * if necessary. This is a "raw" window selection.
587 */
588extern u32 __tcp_select_window(struct sock *sk);
589
590/* TCP timestamps are only 32-bits, this causes a slight
591 * complication on 64-bit systems since we store a snapshot
Stephen Hemminger31f34262005-11-15 15:17:10 -0800592 * of jiffies in the buffer control blocks below. We decided
593 * to use only the low 32-bits of jiffies and hide the ugly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 * casts with the following macro.
595 */
596#define tcp_time_stamp ((__u32)(jiffies))
597
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800598/* This is what the send packet queuing engine uses to pass
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 * TCP per-packet control information to the transmission
600 * code. We also store the host-order sequence numbers in
601 * here too. This is 36 bytes on 32-bit architectures,
602 * 40 bytes on 64-bit machines, if this grows please adjust
603 * skbuff.h:skbuff->cb[xxx] size appropriately.
604 */
605struct tcp_skb_cb {
606 union {
607 struct inet_skb_parm h4;
608#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
609 struct inet6_skb_parm h6;
610#endif
611 } header; /* For incoming frames */
612 __u32 seq; /* Starting sequence number */
613 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
614 __u32 when; /* used to compute rtt's */
615 __u8 flags; /* TCP header flags. */
616
617 /* NOTE: These must match up to the flags byte in a
618 * real TCP header.
619 */
620#define TCPCB_FLAG_FIN 0x01
621#define TCPCB_FLAG_SYN 0x02
622#define TCPCB_FLAG_RST 0x04
623#define TCPCB_FLAG_PSH 0x08
624#define TCPCB_FLAG_ACK 0x10
625#define TCPCB_FLAG_URG 0x20
626#define TCPCB_FLAG_ECE 0x40
627#define TCPCB_FLAG_CWR 0x80
628
629 __u8 sacked; /* State flags for SACK/FACK. */
630#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
631#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
632#define TCPCB_LOST 0x04 /* SKB is lost */
633#define TCPCB_TAGBITS 0x07 /* All tag bits */
634
635#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
636#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
637
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 __u32 ack_seq; /* Sequence number ACK'd */
639};
640
641#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
642
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643/* Due to TSO, an SKB can be composed of multiple actual
644 * packets. To keep these tracked properly, we use this.
645 */
646static inline int tcp_skb_pcount(const struct sk_buff *skb)
647{
Herbert Xu79671682006-06-22 02:40:14 -0700648 return skb_shinfo(skb)->gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649}
650
651/* This is valid iff tcp_skb_pcount() > 1. */
652static inline int tcp_skb_mss(const struct sk_buff *skb)
653{
Herbert Xu79671682006-06-22 02:40:14 -0700654 return skb_shinfo(skb)->gso_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655}
656
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700657/* Events passed to congestion control interface */
658enum tcp_ca_event {
659 CA_EVENT_TX_START, /* first transmit when no packets in flight */
660 CA_EVENT_CWND_RESTART, /* congestion window restart */
661 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
662 CA_EVENT_FRTO, /* fast recovery timeout */
663 CA_EVENT_LOSS, /* loss timeout */
664 CA_EVENT_FAST_ACK, /* in sequence ack */
665 CA_EVENT_SLOW_ACK, /* other ack */
666};
667
668/*
669 * Interface for adding new TCP congestion control handlers
670 */
671#define TCP_CA_NAME_MAX 16
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800672#define TCP_CA_MAX 128
673#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
674
Stephen Hemminger164891a2007-04-23 22:26:16 -0700675#define TCP_CONG_NON_RESTRICTED 0x1
676#define TCP_CONG_RTT_STAMP 0x2
677
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700678struct tcp_congestion_ops {
679 struct list_head list;
Stephen Hemminger164891a2007-04-23 22:26:16 -0700680 unsigned long flags;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700681
682 /* initialize private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300683 void (*init)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700684 /* cleanup private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300685 void (*release)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700686
687 /* return slow start threshold (required) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300688 u32 (*ssthresh)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700689 /* lower bound for congestion window (optional) */
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700690 u32 (*min_cwnd)(const struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700691 /* do new cwnd calculation (required) */
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200692 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700693 /* call before changing ca_state (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300694 void (*set_state)(struct sock *sk, u8 new_state);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700695 /* call when cwnd event occurs (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300696 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700697 /* new value of cwnd after loss (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300698 u32 (*undo_cwnd)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700699 /* hook for packet ack accounting (optional) */
Stephen Hemminger30cfd0b2007-07-25 23:49:34 -0700700 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
Arnaldo Carvalho de Melo73c1f4a2005-08-12 12:51:49 -0300701 /* get info for inet_diag (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300702 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700703
704 char name[TCP_CA_NAME_MAX];
705 struct module *owner;
706};
707
708extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
709extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
710
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300711extern void tcp_init_congestion_control(struct sock *sk);
712extern void tcp_cleanup_congestion_control(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700713extern int tcp_set_default_congestion_control(const char *name);
714extern void tcp_get_default_congestion_control(char *name);
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800715extern void tcp_get_available_congestion_control(char *buf, size_t len);
Stephen Hemmingerce7bc3b2006-11-09 16:35:15 -0800716extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
717extern int tcp_set_allowed_congestion_control(char *allowed);
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300718extern int tcp_set_congestion_control(struct sock *sk, const char *name);
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800719extern void tcp_slow_start(struct tcp_sock *tp);
Ilpo Järvinen758ce5c2009-02-28 04:44:37 +0000720extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700721
Stephen Hemminger5f8ef482005-06-23 20:37:36 -0700722extern struct tcp_congestion_ops tcp_init_congestion_ops;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300723extern u32 tcp_reno_ssthresh(struct sock *sk);
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200724extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700725extern u32 tcp_reno_min_cwnd(const struct sock *sk);
David S. Millera8acfba2005-06-23 23:45:02 -0700726extern struct tcp_congestion_ops tcp_reno;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700727
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300728static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700729{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300730 struct inet_connection_sock *icsk = inet_csk(sk);
731
732 if (icsk->icsk_ca_ops->set_state)
733 icsk->icsk_ca_ops->set_state(sk, ca_state);
734 icsk->icsk_ca_state = ca_state;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700735}
736
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300737static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700738{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300739 const struct inet_connection_sock *icsk = inet_csk(sk);
740
741 if (icsk->icsk_ca_ops->cwnd_event)
742 icsk->icsk_ca_ops->cwnd_event(sk, event);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700743}
744
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300745/* These functions determine how the current flow behaves in respect of SACK
746 * handling. SACK is negotiated with the peer, and therefore it can vary
747 * between different flows.
748 *
749 * tcp_is_sack - SACK enabled
750 * tcp_is_reno - No SACK
751 * tcp_is_fack - FACK enabled, implies SACK enabled
752 */
753static inline int tcp_is_sack(const struct tcp_sock *tp)
754{
755 return tp->rx_opt.sack_ok;
756}
757
758static inline int tcp_is_reno(const struct tcp_sock *tp)
759{
760 return !tcp_is_sack(tp);
761}
762
763static inline int tcp_is_fack(const struct tcp_sock *tp)
764{
765 return tp->rx_opt.sack_ok & 2;
766}
767
768static inline void tcp_enable_fack(struct tcp_sock *tp)
769{
770 tp->rx_opt.sack_ok |= 2;
771}
772
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300773static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
774{
775 return tp->sacked_out + tp->lost_out;
776}
777
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778/* This determines how many packets are "in the network" to the best
779 * of our knowledge. In many cases it is conservative, but where
780 * detailed information is available from the receiver (via SACK
781 * blocks etc.) we can make more aggressive calculations.
782 *
783 * Use this for decisions involving congestion control, use just
784 * tp->packets_out to determine if the send queue is empty or not.
785 *
786 * Read this equation as:
787 *
788 * "Packets sent once on transmission queue" MINUS
789 * "Packets left network, but not honestly ACKed yet" PLUS
790 * "Packets fast retransmitted"
791 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800792static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793{
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300794 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795}
796
Ilpo Järvinen0b6a05c2009-09-15 01:30:10 -0700797#define TCP_INFINITE_SSTHRESH 0x7fffffff
798
799static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp)
800{
801 return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH;
802}
803
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
805 * The exception is rate halving phase, when cwnd is decreasing towards
806 * ssthresh.
807 */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300808static inline __u32 tcp_current_ssthresh(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300810 const struct tcp_sock *tp = tcp_sk(sk);
811 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 return tp->snd_ssthresh;
813 else
814 return max(tp->snd_ssthresh,
815 ((tp->snd_cwnd >> 1) +
816 (tp->snd_cwnd >> 2)));
817}
818
Ilpo Järvinenb9c45952007-07-27 16:36:17 +0300819/* Use define here intentionally to get WARN_ON location shown at the caller */
820#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800822extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
824
825/* Slow start with delack produces 3 packets of burst, so that
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700826 * it is safe "de facto". This will be the default - same as
827 * the default reordering threshold - but if reordering increases,
828 * we must be able to allow cwnd to burst at least this much in order
829 * to not pull it back when holes are filled.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 */
831static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
832{
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700833 return tp->reordering;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834}
835
Ilpo Järvinen90840de2007-12-31 04:48:41 -0800836/* Returns end sequence number of the receiver's advertised window */
837static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
838{
839 return tp->snd_una + tp->snd_wnd;
840}
Ilpo Järvinencea14e02008-01-12 03:19:12 -0800841extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800842
Chuck Leverc1bd24b2007-10-23 21:08:54 -0700843static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800844 const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845{
846 if (skb->len < mss)
847 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
848}
849
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700850static inline void tcp_check_probe_timer(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700852 struct tcp_sock *tp = tcp_sk(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700853 const struct inet_connection_sock *icsk = inet_csk(sk);
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700854
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700855 if (!tp->packets_out && !icsk->icsk_pending)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700856 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
857 icsk->icsk_rto, TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858}
859
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700860static inline void tcp_push_pending_frames(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700862 struct tcp_sock *tp = tcp_sk(sk);
863
Ilpo Järvinen0c54b852009-03-14 14:23:05 +0000864 __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865}
866
Hantzis Fotisee7537b2009-03-02 22:42:02 -0800867static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868{
869 tp->snd_wl1 = seq;
870}
871
Hantzis Fotisee7537b2009-03-02 22:42:02 -0800872static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873{
874 tp->snd_wl1 = seq;
875}
876
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877/*
878 * Calculate(/check) TCP checksum
879 */
Frederik Deweerdtba7808e2007-02-04 20:15:27 -0800880static inline __sum16 tcp_v4_check(int len, __be32 saddr,
881 __be32 daddr, __wsum base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882{
883 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
884}
885
Al Virob51655b2006-11-14 21:40:42 -0800886static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887{
Herbert Xufb286bb2005-11-10 13:01:24 -0800888 return __skb_checksum_complete(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889}
890
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800891static inline int tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892{
Herbert Xu60476372007-04-09 11:59:39 -0700893 return !skb_csum_unnecessary(skb) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 __tcp_checksum_complete(skb);
895}
896
897/* Prequeue for VJ style copy to user, combined with checksumming. */
898
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800899static inline void tcp_prequeue_init(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900{
901 tp->ucopy.task = NULL;
902 tp->ucopy.len = 0;
903 tp->ucopy.memory = 0;
904 skb_queue_head_init(&tp->ucopy.prequeue);
Chris Leech97fc2f02006-05-23 17:55:33 -0700905#ifdef CONFIG_NET_DMA
906 tp->ucopy.dma_chan = NULL;
907 tp->ucopy.wakeup = 0;
908 tp->ucopy.pinned_list = NULL;
909 tp->ucopy.dma_cookie = 0;
910#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911}
912
913/* Packet is added to VJ-style prequeue for processing in process
914 * context, if a reader task is waiting. Apparently, this exciting
915 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
916 * failed somewhere. Latency? Burstiness? Well, at least now we will
917 * see, why it failed. 8)8) --ANK
918 *
919 * NOTE: is this not too big to inline?
920 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800921static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922{
923 struct tcp_sock *tp = tcp_sk(sk);
924
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000925 if (sysctl_tcp_low_latency || !tp->ucopy.task)
926 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000928 __skb_queue_tail(&tp->ucopy.prequeue, skb);
929 tp->ucopy.memory += skb->truesize;
930 if (tp->ucopy.memory > sk->sk_rcvbuf) {
931 struct sk_buff *skb1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000933 BUG_ON(sock_owned_by_user(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000935 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
936 sk_backlog_rcv(sk, skb1);
937 NET_INC_STATS_BH(sock_net(sk),
938 LINUX_MIB_TCPPREQUEUEDROPPED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 }
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000940
941 tp->ucopy.memory = 0;
942 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
Eric Dumazet7aedec22009-05-07 07:20:39 +0000943 wake_up_interruptible_poll(sk->sk_sleep,
944 POLLIN | POLLRDNORM | POLLRDBAND);
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000945 if (!inet_csk_ack_scheduled(sk))
946 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
David S. Miller22f6dac2009-05-08 02:48:30 -0700947 (3 * tcp_rto_min(sk)) / 4,
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000948 TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 }
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000950 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951}
952
953
954#undef STATE_TRACE
955
956#ifdef STATE_TRACE
957static const char *statename[]={
958 "Unused","Established","Syn Sent","Syn Recv",
959 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
960 "Close Wait","Last ACK","Listen","Closing"
961};
962#endif
Ilpo Järvinen490d5042008-01-12 03:17:20 -0800963extern void tcp_set_state(struct sock *sk, int state);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964
Andi Kleen4ac02ba2007-04-20 17:11:46 -0700965extern void tcp_done(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800967static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968{
969 rx_opt->dsack = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 rx_opt->num_sacks = 0;
971}
972
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973/* Determine a window scaling and initial window to offer. */
974extern void tcp_select_initial_window(int __space, __u32 mss,
975 __u32 *rcv_wnd, __u32 *window_clamp,
976 int wscale_ok, __u8 *rcv_wscale);
977
978static inline int tcp_win_from_space(int space)
979{
980 return sysctl_tcp_adv_win_scale<=0 ?
981 (space>>(-sysctl_tcp_adv_win_scale)) :
982 space - (space>>sysctl_tcp_adv_win_scale);
983}
984
985/* Note: caller must be prepared to deal with negative returns */
986static inline int tcp_space(const struct sock *sk)
987{
988 return tcp_win_from_space(sk->sk_rcvbuf -
989 atomic_read(&sk->sk_rmem_alloc));
990}
991
992static inline int tcp_full_space(const struct sock *sk)
993{
994 return tcp_win_from_space(sk->sk_rcvbuf);
995}
996
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800997static inline void tcp_openreq_init(struct request_sock *req,
998 struct tcp_options_received *rx_opt,
999 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001001 struct inet_request_sock *ireq = inet_rsk(req);
1002
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
Florian Westphal4dfc2812008-04-10 03:12:40 -07001004 req->cookie_ts = 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001005 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 req->mss = rx_opt->mss_clamp;
1007 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001008 ireq->tstamp_ok = rx_opt->tstamp_ok;
1009 ireq->sack_ok = rx_opt->sack_ok;
1010 ireq->snd_wscale = rx_opt->snd_wscale;
1011 ireq->wscale_ok = rx_opt->wscale_ok;
1012 ireq->acked = 0;
1013 ireq->ecn_ok = 0;
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001014 ireq->rmt_port = tcp_hdr(skb)->source;
KOVACS Krisztiana3116ac52008-10-01 07:46:49 -07001015 ireq->loc_port = tcp_hdr(skb)->dest;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016}
1017
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001018extern void tcp_enter_memory_pressure(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1021{
1022 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1023}
1024
1025static inline int keepalive_time_when(const struct tcp_sock *tp)
1026{
1027 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1028}
1029
Eric Dumazetdf19a622009-08-28 23:48:54 -07001030static inline int keepalive_probes(const struct tcp_sock *tp)
1031{
1032 return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes;
1033}
1034
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001035static inline int tcp_fin_time(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001037 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1038 const int rto = inet_csk(sk)->icsk_rto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001040 if (fin_timeout < (rto << 2) - (rto >> 1))
1041 fin_timeout = (rto << 2) - (rto >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042
1043 return fin_timeout;
1044}
1045
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001046static inline int tcp_paws_check(const struct tcp_options_received *rx_opt,
1047 int paws_win)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048{
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001049 if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
1050 return 1;
1051 if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
1052 return 1;
1053
1054 return 0;
1055}
1056
1057static inline int tcp_paws_reject(const struct tcp_options_received *rx_opt,
1058 int rst)
1059{
1060 if (tcp_paws_check(rx_opt, 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 return 0;
1062
1063 /* RST segments are not recommended to carry timestamp,
1064 and, if they do, it is recommended to ignore PAWS because
1065 "their cleanup function should take precedence over timestamps."
1066 Certainly, it is mistake. It is necessary to understand the reasons
1067 of this constraint to relax it: if peer reboots, clock may go
1068 out-of-sync and half-open connections will not be reset.
1069 Actually, the problem would be not existing if all
1070 the implementations followed draft about maintaining clock
1071 via reboots. Linux-2.2 DOES NOT!
1072
1073 However, we can relax time bounds for RST segments to MSL.
1074 */
James Morris9d729f72007-03-04 16:12:44 -08001075 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 return 0;
1077 return 1;
1078}
1079
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080#define TCP_CHECK_TIMER(sk) do { } while (0)
1081
Pavel Emelyanova9c193292008-07-16 20:21:42 -07001082static inline void tcp_mib_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083{
1084 /* See RFC 2012 */
Pavel Emelyanovcf1100a2008-07-16 20:27:38 -07001085 TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1086 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1087 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1088 TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089}
1090
Ilpo Järvinen5af4ec22007-09-20 11:30:48 -07001091/* from STCP */
Ilpo Järvinenef9da472008-09-20 21:25:15 -07001092static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
David S. Miller0800f172007-09-20 11:40:37 -07001093{
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001094 tp->lost_skb_hint = NULL;
1095 tp->scoreboard_skb_hint = NULL;
Ilpo Järvinenef9da472008-09-20 21:25:15 -07001096}
1097
1098static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1099{
1100 tcp_clear_retrans_hints_partial(tp);
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001101 tp->retransmit_skb_hint = NULL;
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001102}
1103
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001104/* MD5 Signature */
1105struct crypto_hash;
1106
1107/* - key database */
1108struct tcp_md5sig_key {
1109 u8 *key;
1110 u8 keylen;
1111};
1112
1113struct tcp4_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001114 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001115 __be32 addr;
1116};
1117
1118struct tcp6_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001119 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001120#if 0
1121 u32 scope_id; /* XXX */
1122#endif
1123 struct in6_addr addr;
1124};
1125
1126/* - sock block */
1127struct tcp_md5sig_info {
1128 struct tcp4_md5sig_key *keys4;
1129#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1130 struct tcp6_md5sig_key *keys6;
1131 u32 entries6;
1132 u32 alloced6;
1133#endif
1134 u32 entries4;
1135 u32 alloced4;
1136};
1137
1138/* - pseudo header */
1139struct tcp4_pseudohdr {
1140 __be32 saddr;
1141 __be32 daddr;
1142 __u8 pad;
1143 __u8 protocol;
1144 __be16 len;
1145};
1146
1147struct tcp6_pseudohdr {
1148 struct in6_addr saddr;
1149 struct in6_addr daddr;
1150 __be32 len;
1151 __be32 protocol; /* including padding */
1152};
1153
1154union tcp_md5sum_block {
1155 struct tcp4_pseudohdr ip4;
1156#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1157 struct tcp6_pseudohdr ip6;
1158#endif
1159};
1160
1161/* - pool: digest algorithm, hash description and scratch buffer */
1162struct tcp_md5sig_pool {
1163 struct hash_desc md5_desc;
1164 union tcp_md5sum_block md5_blk;
1165};
1166
1167#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1168
1169/* - functions */
Adam Langley49a72df2008-07-19 00:01:42 -07001170extern int tcp_v4_md5_hash_skb(char *md5_hash,
1171 struct tcp_md5sig_key *key,
1172 struct sock *sk,
1173 struct request_sock *req,
1174 struct sk_buff *skb);
YOSHIFUJI Hideaki8d26d762008-04-17 13:19:16 +09001175
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001176extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1177 struct sock *addr_sk);
1178
1179extern int tcp_v4_md5_do_add(struct sock *sk,
1180 __be32 addr,
1181 u8 *newkey,
1182 u8 newkeylen);
1183
1184extern int tcp_v4_md5_do_del(struct sock *sk,
Al Viro8e5200f2006-11-20 18:06:37 -08001185 __be32 addr);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001186
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +09001187#ifdef CONFIG_TCP_MD5SIG
1188#define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_keylen ? \
1189 &(struct tcp_md5sig_key) { \
1190 .key = (twsk)->tw_md5_key, \
1191 .keylen = (twsk)->tw_md5_keylen, \
1192 } : NULL)
1193#else
1194#define tcp_twsk_md5_key(twsk) NULL
1195#endif
1196
Wu Fengguangaa133072009-09-02 23:45:45 -07001197extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(struct sock *);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001198extern void tcp_free_md5sig_pool(void);
1199
1200extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1201extern void __tcp_put_md5sig_pool(void);
Adam Langley49a72df2008-07-19 00:01:42 -07001202extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, struct tcphdr *);
1203extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, struct sk_buff *,
1204 unsigned header_len);
1205extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
1206 struct tcp_md5sig_key *key);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001207
1208static inline
1209struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1210{
1211 int cpu = get_cpu();
1212 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1213 if (!ret)
1214 put_cpu();
1215 return ret;
1216}
1217
1218static inline void tcp_put_md5sig_pool(void)
1219{
1220 __tcp_put_md5sig_pool();
1221 put_cpu();
1222}
1223
David S. Millerfe067e82007-03-07 12:12:44 -08001224/* write queue abstraction */
1225static inline void tcp_write_queue_purge(struct sock *sk)
1226{
1227 struct sk_buff *skb;
1228
1229 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001230 sk_wmem_free_skb(sk, skb);
1231 sk_mem_reclaim(sk);
Ilpo Järvinen8818a9d2009-12-02 22:24:02 -08001232 tcp_clear_all_retrans_hints(tcp_sk(sk));
David S. Millerfe067e82007-03-07 12:12:44 -08001233}
1234
1235static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1236{
David S. Millercd07a8e2008-09-23 00:50:13 -07001237 return skb_peek(&sk->sk_write_queue);
David S. Millerfe067e82007-03-07 12:12:44 -08001238}
1239
1240static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1241{
David S. Millercd07a8e2008-09-23 00:50:13 -07001242 return skb_peek_tail(&sk->sk_write_queue);
David S. Millerfe067e82007-03-07 12:12:44 -08001243}
1244
1245static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1246{
David S. Millercd07a8e2008-09-23 00:50:13 -07001247 return skb_queue_next(&sk->sk_write_queue, skb);
David S. Millerfe067e82007-03-07 12:12:44 -08001248}
1249
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001250static inline struct sk_buff *tcp_write_queue_prev(struct sock *sk, struct sk_buff *skb)
1251{
1252 return skb_queue_prev(&sk->sk_write_queue, skb);
1253}
1254
David S. Millerfe067e82007-03-07 12:12:44 -08001255#define tcp_for_write_queue(skb, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001256 skb_queue_walk(&(sk)->sk_write_queue, skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001257
1258#define tcp_for_write_queue_from(skb, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001259 skb_queue_walk_from(&(sk)->sk_write_queue, skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001260
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001261#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001262 skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001263
Damian Lukowski5152fc72009-09-01 10:24:00 +00001264/* This function calculates a "timeout" which is equivalent to the timeout of a
André Goddard Rosaaf901ca2009-11-14 13:09:05 -02001265 * TCP connection after "boundary" unsuccessful, exponentially backed-off
Damian Lukowski5152fc72009-09-01 10:24:00 +00001266 * retransmissions with an initial RTO of TCP_RTO_MIN.
1267 */
Damian Lukowski6fa12c82009-08-26 00:16:34 +00001268static inline bool retransmits_timed_out(const struct sock *sk,
1269 unsigned int boundary)
1270{
Damian Lukowski5152fc72009-09-01 10:24:00 +00001271 unsigned int timeout, linear_backoff_thresh;
1272
Damian Lukowski6fa12c82009-08-26 00:16:34 +00001273 if (!inet_csk(sk)->icsk_retransmits)
1274 return false;
1275
Damian Lukowski5152fc72009-09-01 10:24:00 +00001276 linear_backoff_thresh = ilog2(TCP_RTO_MAX/TCP_RTO_MIN);
Damian Lukowski6fa12c82009-08-26 00:16:34 +00001277
Damian Lukowski5152fc72009-09-01 10:24:00 +00001278 if (boundary <= linear_backoff_thresh)
1279 timeout = ((2 << boundary) - 1) * TCP_RTO_MIN;
Damian Lukowski6fa12c82009-08-26 00:16:34 +00001280 else
Damian Lukowski5152fc72009-09-01 10:24:00 +00001281 timeout = ((2 << linear_backoff_thresh) - 1) * TCP_RTO_MIN +
1282 (boundary - linear_backoff_thresh) * TCP_RTO_MAX;
Damian Lukowski6fa12c82009-08-26 00:16:34 +00001283
Damian Lukowski5152fc72009-09-01 10:24:00 +00001284 return (tcp_time_stamp - tcp_sk(sk)->retrans_stamp) >= timeout;
Damian Lukowski6fa12c82009-08-26 00:16:34 +00001285}
1286
David S. Millerfe067e82007-03-07 12:12:44 -08001287static inline struct sk_buff *tcp_send_head(struct sock *sk)
1288{
1289 return sk->sk_send_head;
1290}
1291
David S. Millercd07a8e2008-09-23 00:50:13 -07001292static inline bool tcp_skb_is_last(const struct sock *sk,
1293 const struct sk_buff *skb)
1294{
1295 return skb_queue_is_last(&sk->sk_write_queue, skb);
1296}
1297
David S. Millerfe067e82007-03-07 12:12:44 -08001298static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1299{
David S. Millercd07a8e2008-09-23 00:50:13 -07001300 if (tcp_skb_is_last(sk, skb))
David S. Millerfe067e82007-03-07 12:12:44 -08001301 sk->sk_send_head = NULL;
David S. Millercd07a8e2008-09-23 00:50:13 -07001302 else
1303 sk->sk_send_head = tcp_write_queue_next(sk, skb);
David S. Millerfe067e82007-03-07 12:12:44 -08001304}
1305
1306static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1307{
1308 if (sk->sk_send_head == skb_unlinked)
1309 sk->sk_send_head = NULL;
1310}
1311
1312static inline void tcp_init_send_head(struct sock *sk)
1313{
1314 sk->sk_send_head = NULL;
1315}
1316
1317static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1318{
1319 __skb_queue_tail(&sk->sk_write_queue, skb);
1320}
1321
1322static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1323{
1324 __tcp_add_write_queue_tail(sk, skb);
1325
1326 /* Queue it, remembering where we must start sending. */
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001327 if (sk->sk_send_head == NULL) {
David S. Millerfe067e82007-03-07 12:12:44 -08001328 sk->sk_send_head = skb;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001329
1330 if (tcp_sk(sk)->highest_sack == NULL)
1331 tcp_sk(sk)->highest_sack = skb;
1332 }
David S. Millerfe067e82007-03-07 12:12:44 -08001333}
1334
1335static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1336{
1337 __skb_queue_head(&sk->sk_write_queue, skb);
1338}
1339
1340/* Insert buff after skb on the write queue of sk. */
1341static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1342 struct sk_buff *buff,
1343 struct sock *sk)
1344{
Gerrit Renker7de6c032008-04-14 00:05:09 -07001345 __skb_queue_after(&sk->sk_write_queue, skb, buff);
David S. Millerfe067e82007-03-07 12:12:44 -08001346}
1347
David S. Miller43f59c82008-09-21 21:28:51 -07001348/* Insert new before skb on the write queue of sk. */
David S. Millerfe067e82007-03-07 12:12:44 -08001349static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1350 struct sk_buff *skb,
1351 struct sock *sk)
1352{
David S. Miller43f59c82008-09-21 21:28:51 -07001353 __skb_queue_before(&sk->sk_write_queue, skb, new);
Ilpo Järvinen6e421412007-11-19 23:24:09 -08001354
1355 if (sk->sk_send_head == skb)
1356 sk->sk_send_head = new;
David S. Millerfe067e82007-03-07 12:12:44 -08001357}
1358
1359static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1360{
1361 __skb_unlink(skb, &sk->sk_write_queue);
1362}
1363
David S. Millerfe067e82007-03-07 12:12:44 -08001364static inline int tcp_write_queue_empty(struct sock *sk)
1365{
1366 return skb_queue_empty(&sk->sk_write_queue);
1367}
1368
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001369/* Start sequence of the highest skb with SACKed bit, valid only if
1370 * sacked > 0 or when the caller has ensured validity by itself.
1371 */
1372static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1373{
1374 if (!tp->sacked_out)
1375 return tp->snd_una;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001376
1377 if (tp->highest_sack == NULL)
1378 return tp->snd_nxt;
1379
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001380 return TCP_SKB_CB(tp->highest_sack)->seq;
1381}
1382
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001383static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1384{
1385 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1386 tcp_write_queue_next(sk, skb);
1387}
1388
1389static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1390{
1391 return tcp_sk(sk)->highest_sack;
1392}
1393
1394static inline void tcp_highest_sack_reset(struct sock *sk)
1395{
1396 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1397}
1398
1399/* Called when old skb is about to be deleted (to be combined with new skb) */
1400static inline void tcp_highest_sack_combine(struct sock *sk,
1401 struct sk_buff *old,
1402 struct sk_buff *new)
1403{
1404 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1405 tcp_sk(sk)->highest_sack = new;
1406}
1407
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408/* /proc */
1409enum tcp_seq_states {
1410 TCP_SEQ_STATE_LISTENING,
1411 TCP_SEQ_STATE_OPENREQ,
1412 TCP_SEQ_STATE_ESTABLISHED,
1413 TCP_SEQ_STATE_TIME_WAIT,
1414};
1415
1416struct tcp_seq_afinfo {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 char *name;
1418 sa_family_t family;
Denis V. Lunev68fcadd2008-04-13 22:13:30 -07001419 struct file_operations seq_fops;
Denis V. Lunev9427c4b2008-04-13 22:12:13 -07001420 struct seq_operations seq_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421};
1422
1423struct tcp_iter_state {
Denis V. Luneva4146b12008-04-13 22:11:14 -07001424 struct seq_net_private p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 sa_family_t family;
1426 enum tcp_seq_states state;
1427 struct sock *syn_wait_sk;
1428 int bucket, sbucket, num, uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429};
1430
Daniel Lezcano6f8b13b2008-03-21 04:14:45 -07001431extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1432extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001434extern struct request_sock_ops tcp_request_sock_ops;
Glenn Griffinc6aefaf2008-02-07 21:49:26 -08001435extern struct request_sock_ops tcp6_request_sock_ops;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001436
Brian Haley7d06b2e2008-06-14 17:04:49 -07001437extern void tcp_v4_destroy_sock(struct sock *sk);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001438
Herbert Xua430a432006-07-08 13:34:56 -07001439extern int tcp_v4_gso_send_check(struct sk_buff *skb);
Herbert Xu576a30e2006-06-27 13:22:38 -07001440extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
Herbert Xubf296b12008-12-15 23:43:36 -08001441extern struct sk_buff **tcp_gro_receive(struct sk_buff **head,
1442 struct sk_buff *skb);
1443extern struct sk_buff **tcp4_gro_receive(struct sk_buff **head,
1444 struct sk_buff *skb);
1445extern int tcp_gro_complete(struct sk_buff *skb);
1446extern int tcp4_gro_complete(struct sk_buff *skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001447
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001448#ifdef CONFIG_PROC_FS
1449extern int tcp4_proc_init(void);
1450extern void tcp4_proc_exit(void);
1451#endif
1452
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001453/* TCP af-specific functions */
1454struct tcp_sock_af_ops {
1455#ifdef CONFIG_TCP_MD5SIG
1456 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1457 struct sock *addr_sk);
1458 int (*calc_md5_hash) (char *location,
1459 struct tcp_md5sig_key *md5,
1460 struct sock *sk,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001461 struct request_sock *req,
Adam Langley49a72df2008-07-19 00:01:42 -07001462 struct sk_buff *skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001463 int (*md5_add) (struct sock *sk,
1464 struct sock *addr_sk,
1465 u8 *newkey,
1466 u8 len);
1467 int (*md5_parse) (struct sock *sk,
1468 char __user *optval,
1469 int optlen);
1470#endif
1471};
1472
1473struct tcp_request_sock_ops {
1474#ifdef CONFIG_TCP_MD5SIG
1475 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1476 struct request_sock *req);
John Dykstrae3afe7b2009-07-16 05:04:51 +00001477 int (*calc_md5_hash) (char *location,
1478 struct tcp_md5sig_key *md5,
1479 struct sock *sk,
1480 struct request_sock *req,
1481 struct sk_buff *skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001482#endif
1483};
1484
William Allen Simpsonda5c78c2009-12-02 18:12:09 +00001485/* Using SHA1 for now, define some constants.
1486 */
1487#define COOKIE_DIGEST_WORDS (SHA_DIGEST_WORDS)
1488#define COOKIE_MESSAGE_WORDS (SHA_MESSAGE_BYTES / 4)
1489#define COOKIE_WORKSPACE_WORDS (COOKIE_DIGEST_WORDS + COOKIE_MESSAGE_WORDS)
1490
1491extern int tcp_cookie_generator(u32 *bakery);
1492
William Allen Simpson435cf552009-12-02 18:17:05 +00001493/**
1494 * struct tcp_cookie_values - each socket needs extra space for the
1495 * cookies, together with (optional) space for any SYN data.
1496 *
1497 * A tcp_sock contains a pointer to the current value, and this is
1498 * cloned to the tcp_timewait_sock.
1499 *
1500 * @cookie_pair: variable data from the option exchange.
1501 *
1502 * @cookie_desired: user specified tcpct_cookie_desired. Zero
1503 * indicates default (sysctl_tcp_cookie_size).
1504 * After cookie sent, remembers size of cookie.
1505 * Range 0, TCP_COOKIE_MIN to TCP_COOKIE_MAX.
1506 *
1507 * @s_data_desired: user specified tcpct_s_data_desired. When the
1508 * constant payload is specified (@s_data_constant),
1509 * holds its length instead.
1510 * Range 0 to TCP_MSS_DESIRED.
1511 *
1512 * @s_data_payload: constant data that is to be included in the
1513 * payload of SYN or SYNACK segments when the
1514 * cookie option is present.
1515 */
1516struct tcp_cookie_values {
1517 struct kref kref;
1518 u8 cookie_pair[TCP_COOKIE_PAIR_SIZE];
1519 u8 cookie_pair_size;
1520 u8 cookie_desired;
1521 u16 s_data_desired:11,
1522 s_data_constant:1,
1523 s_data_in:1,
1524 s_data_out:1,
1525 s_data_unused:2;
1526 u8 s_data_payload[0];
1527};
1528
1529static inline void tcp_cookie_values_release(struct kref *kref)
1530{
1531 kfree(container_of(kref, struct tcp_cookie_values, kref));
1532}
1533
1534/* The length of constant payload data. Note that s_data_desired is
1535 * overloaded, depending on s_data_constant: either the length of constant
1536 * data (returned here) or the limit on variable data.
1537 */
1538static inline int tcp_s_data_size(const struct tcp_sock *tp)
1539{
1540 return (tp->cookie_values != NULL && tp->cookie_values->s_data_constant)
1541 ? tp->cookie_values->s_data_desired
1542 : 0;
1543}
1544
1545/**
1546 * struct tcp_extend_values - tcp_ipv?.c to tcp_output.c workspace.
1547 *
1548 * As tcp_request_sock has already been extended in other places, the
1549 * only remaining method is to pass stack values along as function
1550 * parameters. These parameters are not needed after sending SYNACK.
1551 *
1552 * @cookie_bakery: cryptographic secret and message workspace.
1553 *
1554 * @cookie_plus: bytes in authenticator/cookie option, copied from
1555 * struct tcp_options_received (above).
1556 */
1557struct tcp_extend_values {
1558 struct request_values rv;
1559 u32 cookie_bakery[COOKIE_WORKSPACE_WORDS];
1560 u8 cookie_plus:6,
1561 cookie_out_never:1,
1562 cookie_in_always:1;
1563};
1564
1565static inline struct tcp_extend_values *tcp_xv(struct request_values *rvp)
1566{
1567 return (struct tcp_extend_values *)rvp;
1568}
1569
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001570extern void tcp_v4_init(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001571extern void tcp_init(void);
1572
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573#endif /* _TCP_H */