blob: 19f4150f4d4d166380f95079c1877af28e6d62a5 [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>
Satoru SATOH0c266892009-05-04 11:11:01 -070044#include <net/dst.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070045
Linus Torvalds1da177e2005-04-16 15:20:36 -070046#include <linux/seq_file.h>
47
Arnaldo Carvalho de Melo6e04e022005-08-09 20:07:35 -070048extern struct inet_hashinfo tcp_hashinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -070049
Eric Dumazetdd24c002008-11-25 21:17:14 -080050extern struct percpu_counter tcp_orphan_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -070051extern void tcp_time_wait(struct sock *sk, int state, int timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#define MAX_TCP_HEADER (128 + MAX_HEADER)
Adam Langley33ad7982008-07-19 00:04:31 -070054#define MAX_TCP_OPTION_SPACE 40
Linus Torvalds1da177e2005-04-16 15:20:36 -070055
56/*
57 * Never offer a window over 32767 without using window scaling. Some
58 * poor stacks do signed 16bit maths!
59 */
60#define MAX_TCP_WINDOW 32767U
61
62/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
63#define TCP_MIN_MSS 88U
64
65/* Minimal RCV_MSS. */
66#define TCP_MIN_RCVMSS 536U
67
John Heffner5d424d52006-03-20 17:53:41 -080068/* The least MTU to use for probing */
69#define TCP_BASE_MSS 512
70
Linus Torvalds1da177e2005-04-16 15:20:36 -070071/* After receiving this amount of duplicate ACKs fast retransmit starts. */
72#define TCP_FASTRETRANS_THRESH 3
73
74/* Maximal reordering. */
75#define TCP_MAX_REORDERING 127
76
77/* Maximal number of ACKs sent quickly to accelerate slow-start. */
78#define TCP_MAX_QUICKACKS 16U
79
80/* urg_data states */
81#define TCP_URG_VALID 0x0100
82#define TCP_URG_NOTYET 0x0200
83#define TCP_URG_READ 0x0400
84
85#define TCP_RETR1 3 /*
86 * This is how many retries it does before it
87 * tries to figure out if the gateway is
88 * down. Minimal RFC value is 3; it corresponds
89 * to ~3sec-8min depending on RTO.
90 */
91
92#define TCP_RETR2 15 /*
93 * This should take at least
94 * 90 minutes to time out.
95 * RFC1122 says that the limit is 100 sec.
96 * 15 is ~13-30min depending on RTO.
97 */
98
99#define TCP_SYN_RETRIES 5 /* number of times to retry active opening a
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800100 * connection: ~180sec is RFC minimum */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
102#define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800103 * connection: ~180sec is RFC minimum */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104
105
106#define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned
107 * socket. 7 is ~50sec-16min.
108 */
109
110
111#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
112 * state, about 60 seconds */
113#define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
114 /* BSD style FIN_WAIT2 deadlock breaker.
115 * It used to be 3min, new value is 60sec,
116 * to combine FIN-WAIT-2 timeout with
117 * TIME-WAIT timer.
118 */
119
120#define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
121#if HZ >= 100
122#define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
123#define TCP_ATO_MIN ((unsigned)(HZ/25))
124#else
125#define TCP_DELACK_MIN 4U
126#define TCP_ATO_MIN 4U
127#endif
128#define TCP_RTO_MAX ((unsigned)(120*HZ))
129#define TCP_RTO_MIN ((unsigned)(HZ/5))
130#define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */
131
132#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
133 * for local resources.
134 */
135
136#define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
137#define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
138#define TCP_KEEPALIVE_INTVL (75*HZ)
139
140#define MAX_TCP_KEEPIDLE 32767
141#define MAX_TCP_KEEPINTVL 32767
142#define MAX_TCP_KEEPCNT 127
143#define MAX_TCP_SYNCNT 127
144
145#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146
147#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
148#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
149 * after this time. It should be equal
150 * (or greater than) TCP_TIMEWAIT_LEN
151 * to provide reliability equal to one
152 * provided by timewait state.
153 */
154#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
155 * timestamps. It must be less than
156 * minimal timewait lifetime.
157 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158/*
159 * TCP option
160 */
161
162#define TCPOPT_NOP 1 /* Padding */
163#define TCPOPT_EOL 0 /* End of options */
164#define TCPOPT_MSS 2 /* Segment size negotiating */
165#define TCPOPT_WINDOW 3 /* Window scaling */
166#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
167#define TCPOPT_SACK 5 /* SACK Block */
168#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800169#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
171/*
172 * TCP option lengths
173 */
174
175#define TCPOLEN_MSS 4
176#define TCPOLEN_WINDOW 3
177#define TCPOLEN_SACK_PERM 2
178#define TCPOLEN_TIMESTAMP 10
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800179#define TCPOLEN_MD5SIG 18
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180
181/* But this is what stacks really send out. */
182#define TCPOLEN_TSTAMP_ALIGNED 12
183#define TCPOLEN_WSCALE_ALIGNED 4
184#define TCPOLEN_SACKPERM_ALIGNED 4
185#define TCPOLEN_SACK_BASE 2
186#define TCPOLEN_SACK_BASE_ALIGNED 4
187#define TCPOLEN_SACK_PERBLOCK 8
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800188#define TCPOLEN_MD5SIG_ALIGNED 20
Adam Langley33ad7982008-07-19 00:04:31 -0700189#define TCPOLEN_MSS_ALIGNED 4
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191/* Flags in tp->nonagle */
192#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
193#define TCP_NAGLE_CORK 2 /* Socket is corked */
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800194#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -0700196extern struct inet_timewait_death_row tcp_death_row;
197
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198/* sysctl variables for tcp */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199extern int sysctl_tcp_timestamps;
200extern int sysctl_tcp_window_scaling;
201extern int sysctl_tcp_sack;
202extern int sysctl_tcp_fin_timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203extern int sysctl_tcp_keepalive_time;
204extern int sysctl_tcp_keepalive_probes;
205extern int sysctl_tcp_keepalive_intvl;
206extern int sysctl_tcp_syn_retries;
207extern int sysctl_tcp_synack_retries;
208extern int sysctl_tcp_retries1;
209extern int sysctl_tcp_retries2;
210extern int sysctl_tcp_orphan_retries;
211extern int sysctl_tcp_syncookies;
212extern int sysctl_tcp_retrans_collapse;
213extern int sysctl_tcp_stdurg;
214extern int sysctl_tcp_rfc1337;
215extern int sysctl_tcp_abort_on_overflow;
216extern int sysctl_tcp_max_orphans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217extern int sysctl_tcp_fack;
218extern int sysctl_tcp_reordering;
219extern int sysctl_tcp_ecn;
220extern int sysctl_tcp_dsack;
221extern int sysctl_tcp_mem[3];
222extern int sysctl_tcp_wmem[3];
223extern int sysctl_tcp_rmem[3];
224extern int sysctl_tcp_app_win;
225extern int sysctl_tcp_adv_win_scale;
226extern int sysctl_tcp_tw_reuse;
227extern int sysctl_tcp_frto;
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800228extern int sysctl_tcp_frto_response;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229extern int sysctl_tcp_low_latency;
Chris Leech95937822006-05-23 18:02:55 -0700230extern int sysctl_tcp_dma_copybreak;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231extern int sysctl_tcp_nometrics_save;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232extern int sysctl_tcp_moderate_rcvbuf;
233extern int sysctl_tcp_tso_win_divisor;
Stephen Hemminger9772efb2005-11-10 17:09:53 -0800234extern int sysctl_tcp_abc;
John Heffner5d424d52006-03-20 17:53:41 -0800235extern int sysctl_tcp_mtu_probing;
236extern int sysctl_tcp_base_mss;
Rick Jones15d99e02006-03-20 22:40:29 -0800237extern int sysctl_tcp_workaround_signed_windows;
David S. Miller35089bb2006-06-13 22:33:04 -0700238extern int sysctl_tcp_slow_start_after_idle;
John Heffner886236c2007-03-25 19:21:45 -0700239extern int sysctl_tcp_max_ssthresh;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
241extern atomic_t tcp_memory_allocated;
Eric Dumazet17483762008-11-25 21:16:35 -0800242extern struct percpu_counter tcp_sockets_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243extern int tcp_memory_pressure;
244
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 * The next routines deal with comparing 32 bit unsigned ints
247 * and worry about wraparound (automatic with unsigned arithmetic).
248 */
249
250static inline int before(__u32 seq1, __u32 seq2)
251{
Gerrit Renker0d630cc2007-01-04 12:25:16 -0800252 return (__s32)(seq1-seq2) < 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253}
Gerrit Renker9a036b9c2006-12-20 10:25:55 -0800254#define after(seq2, seq1) before(seq1, seq2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
256/* is s2<=s1<=s3 ? */
257static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
258{
259 return seq3 - seq2 >= seq1 - seq2;
260}
261
Pavel Emelianove4fd5da2007-05-29 13:19:18 -0700262static inline int tcp_too_many_orphans(struct sock *sk, int num)
263{
264 return (num > sysctl_tcp_max_orphans) ||
265 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
266 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
267}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
Florian Westphala0f82f62009-04-19 09:43:48 +0000269/* syncookies: remember time of last synqueue overflow */
270static inline void tcp_synq_overflow(struct sock *sk)
271{
272 tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies;
273}
274
275/* syncookies: no recent synqueue overflow on this listening socket? */
276static inline int tcp_synq_no_recent_overflow(const struct sock *sk)
277{
278 unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp;
279 return time_after(jiffies, last_overflow + TCP_TIMEOUT_INIT);
280}
281
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282extern struct proto tcp_prot;
283
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -0700284#define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field)
285#define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
286#define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
287#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 -0700288
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289extern void tcp_v4_err(struct sk_buff *skb, u32);
290
291extern void tcp_shutdown (struct sock *sk, int how);
292
293extern int tcp_v4_rcv(struct sk_buff *skb);
294
295extern int tcp_v4_remember_stamp(struct sock *sk);
296
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700297extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298
David S. Miller3516ffb2007-08-02 19:23:56 -0700299extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 struct msghdr *msg, size_t size);
301extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
302
303extern int tcp_ioctl(struct sock *sk,
304 int cmd,
305 unsigned long arg);
306
307extern int tcp_rcv_state_process(struct sock *sk,
308 struct sk_buff *skb,
309 struct tcphdr *th,
310 unsigned len);
311
312extern int tcp_rcv_established(struct sock *sk,
313 struct sk_buff *skb,
314 struct tcphdr *th,
315 unsigned len);
316
317extern void tcp_rcv_space_adjust(struct sock *sk);
318
Chris Leech0e4b4992006-05-23 18:00:16 -0700319extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
320
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800321extern int tcp_twsk_unique(struct sock *sk,
322 struct sock *sktw, void *twp);
323
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800324extern void tcp_twsk_destructor(struct sock *sk);
325
Jens Axboe9c55e012007-11-06 23:30:13 -0800326extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
327 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
328
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700329static inline void tcp_dec_quickack_mode(struct sock *sk,
330 const unsigned int pkts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700332 struct inet_connection_sock *icsk = inet_csk(sk);
David S. Millerfc6415bc2005-07-05 15:17:45 -0700333
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700334 if (icsk->icsk_ack.quick) {
335 if (pkts >= icsk->icsk_ack.quick) {
336 icsk->icsk_ack.quick = 0;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700337 /* Leaving quickack mode we deflate ATO. */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700338 icsk->icsk_ack.ato = TCP_ATO_MIN;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700339 } else
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700340 icsk->icsk_ack.quick -= pkts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 }
342}
343
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700344extern void tcp_enter_quickack_mode(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
347{
348 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
349}
350
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -0700351#define TCP_ECN_OK 1
352#define TCP_ECN_QUEUE_CWR 2
353#define TCP_ECN_DEMAND_CWR 4
354
355static __inline__ void
356TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
357{
358 if (sysctl_tcp_ecn && th->ece && th->cwr)
359 inet_rsk(req)->ecn_ok = 1;
360}
361
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362enum tcp_tw_status
363{
364 TCP_TW_SUCCESS = 0,
365 TCP_TW_RST = 1,
366 TCP_TW_ACK = 2,
367 TCP_TW_SYN = 3
368};
369
370
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700371extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 struct sk_buff *skb,
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700373 const struct tcphdr *th);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374
375extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700376 struct request_sock *req,
377 struct request_sock **prev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378extern int tcp_child_process(struct sock *parent,
379 struct sock *child,
380 struct sk_buff *skb);
Ilpo Järvinen46d0de42007-02-21 23:10:39 -0800381extern int tcp_use_frto(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382extern void tcp_enter_frto(struct sock *sk);
383extern void tcp_enter_loss(struct sock *sk, int how);
384extern void tcp_clear_retrans(struct tcp_sock *tp);
385extern void tcp_update_metrics(struct sock *sk);
386
387extern void tcp_close(struct sock *sk,
388 long timeout);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
390
391extern int tcp_getsockopt(struct sock *sk, int level,
392 int optname,
393 char __user *optval,
394 int __user *optlen);
395extern int tcp_setsockopt(struct sock *sk, int level,
396 int optname, char __user *optval,
397 int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800398extern int compat_tcp_getsockopt(struct sock *sk,
399 int level, int optname,
400 char __user *optval, int __user *optlen);
401extern int compat_tcp_setsockopt(struct sock *sk,
402 int level, int optname,
403 char __user *optval, int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404extern void tcp_set_keepalive(struct sock *sk, int val);
405extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
406 struct msghdr *msg,
407 size_t len, int nonblock,
408 int flags, int *addr_len);
409
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410extern void tcp_parse_options(struct sk_buff *skb,
411 struct tcp_options_received *opt_rx,
412 int estab);
413
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,
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700446 struct request_sock *req);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447
448extern int tcp_disconnect(struct sock *sk, int flags);
449
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451/* From syncookies.c */
Florian Westphal2051f112008-03-23 22:21:28 -0700452extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
454 struct ip_options *opt);
455extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
456 __u16 *mss);
457
Florian Westphal4dfc2812008-04-10 03:12:40 -0700458extern __u32 cookie_init_timestamp(struct request_sock *req);
459extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt);
460
Glenn Griffinc6aefaf2008-02-07 21:49:26 -0800461/* From net/ipv6/syncookies.c */
462extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
463extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb,
464 __u16 *mss);
465
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466/* tcp_output.c */
467
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700468extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
469 int nonagle);
470extern int tcp_may_send_now(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
472extern void tcp_xmit_retransmit_queue(struct sock *);
473extern void tcp_simple_retransmit(struct sock *);
474extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
David S. Miller6475be12005-09-01 22:47:01 -0700475extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476
477extern void tcp_send_probe0(struct sock *);
478extern void tcp_send_partial(struct sock *);
479extern int tcp_write_wakeup(struct sock *);
480extern void tcp_send_fin(struct sock *sk);
Al Virodd0fc662005-10-07 07:46:04 +0100481extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482extern int tcp_send_synack(struct sock *);
David S. Millerc1b4a7e2005-07-05 15:24:38 -0700483extern void tcp_push_one(struct sock *, unsigned int mss_now);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484extern void tcp_send_ack(struct sock *sk);
485extern void tcp_send_delayed_ack(struct sock *sk);
486
David S. Millera762a982005-07-05 15:18:51 -0700487/* tcp_input.c */
488extern void tcp_cwnd_application_limited(struct sock *sk);
489
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490/* tcp_timer.c */
491extern void tcp_init_xmit_timers(struct sock *);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700492static inline void tcp_clear_xmit_timers(struct sock *sk)
493{
494 inet_csk_clear_xmit_timers(sk);
495}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
Ilpo Järvinen0c54b852009-03-14 14:23:05 +0000498extern unsigned int tcp_current_mss(struct sock *sk);
499
500/* Bound MSS / TSO packet size with the half of the window */
501static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
502{
503 if (tp->max_window && pktsize > (tp->max_window >> 1))
504 return max(tp->max_window >> 1, 68U - tp->tcp_header_len);
505 else
506 return pktsize;
507}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508
Arnaldo Carvalho de Melo17b085e2005-08-12 12:59:17 -0300509/* tcp.c */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510extern void tcp_get_info(struct sock *, struct tcp_info *);
511
512/* Read 'sendfile()'-style from a TCP socket */
513typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
514 unsigned int, size_t);
515extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
516 sk_read_actor_t recv_actor);
517
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800518extern void tcp_initialize_rcv_mss(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519
John Heffner5d424d52006-03-20 17:53:41 -0800520extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
521extern int tcp_mss_to_mtu(struct sock *sk, int mss);
522extern void tcp_mtup_init(struct sock *sk);
523
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800524static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525{
526 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
527 ntohl(TCP_FLAG_ACK) |
528 snd_wnd);
529}
530
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800531static inline void tcp_fast_path_on(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532{
533 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
534}
535
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700536static inline void tcp_fast_path_check(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700538 struct tcp_sock *tp = tcp_sk(sk);
539
David S. Millerb03efcf2005-07-08 14:57:23 -0700540 if (skb_queue_empty(&tp->out_of_order_queue) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 tp->rcv_wnd &&
542 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
543 !tp->urg_data)
544 tcp_fast_path_on(tp);
545}
546
Satoru SATOH0c266892009-05-04 11:11:01 -0700547/* Compute the actual rto_min value */
548static inline u32 tcp_rto_min(struct sock *sk)
549{
550 struct dst_entry *dst = __sk_dst_get(sk);
551 u32 rto_min = TCP_RTO_MIN;
552
553 if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
554 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
555 return rto_min;
556}
557
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558/* Compute the actual receive window we are currently advertising.
559 * Rcv_nxt can be after the window if our peer push more data
560 * than the offered window.
561 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800562static inline u32 tcp_receive_window(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563{
564 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
565
566 if (win < 0)
567 win = 0;
568 return (u32) win;
569}
570
571/* Choose a new window, without checks for shrinking, and without
572 * scaling applied to the result. The caller does these things
573 * if necessary. This is a "raw" window selection.
574 */
575extern u32 __tcp_select_window(struct sock *sk);
576
577/* TCP timestamps are only 32-bits, this causes a slight
578 * complication on 64-bit systems since we store a snapshot
Stephen Hemminger31f34262005-11-15 15:17:10 -0800579 * of jiffies in the buffer control blocks below. We decided
580 * to use only the low 32-bits of jiffies and hide the ugly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 * casts with the following macro.
582 */
583#define tcp_time_stamp ((__u32)(jiffies))
584
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800585/* This is what the send packet queuing engine uses to pass
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 * TCP per-packet control information to the transmission
587 * code. We also store the host-order sequence numbers in
588 * here too. This is 36 bytes on 32-bit architectures,
589 * 40 bytes on 64-bit machines, if this grows please adjust
590 * skbuff.h:skbuff->cb[xxx] size appropriately.
591 */
592struct tcp_skb_cb {
593 union {
594 struct inet_skb_parm h4;
595#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
596 struct inet6_skb_parm h6;
597#endif
598 } header; /* For incoming frames */
599 __u32 seq; /* Starting sequence number */
600 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
601 __u32 when; /* used to compute rtt's */
602 __u8 flags; /* TCP header flags. */
603
604 /* NOTE: These must match up to the flags byte in a
605 * real TCP header.
606 */
607#define TCPCB_FLAG_FIN 0x01
608#define TCPCB_FLAG_SYN 0x02
609#define TCPCB_FLAG_RST 0x04
610#define TCPCB_FLAG_PSH 0x08
611#define TCPCB_FLAG_ACK 0x10
612#define TCPCB_FLAG_URG 0x20
613#define TCPCB_FLAG_ECE 0x40
614#define TCPCB_FLAG_CWR 0x80
615
616 __u8 sacked; /* State flags for SACK/FACK. */
617#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
618#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
619#define TCPCB_LOST 0x04 /* SKB is lost */
620#define TCPCB_TAGBITS 0x07 /* All tag bits */
621
622#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
623#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
624
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 __u32 ack_seq; /* Sequence number ACK'd */
626};
627
628#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
629
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630/* Due to TSO, an SKB can be composed of multiple actual
631 * packets. To keep these tracked properly, we use this.
632 */
633static inline int tcp_skb_pcount(const struct sk_buff *skb)
634{
Herbert Xu79671682006-06-22 02:40:14 -0700635 return skb_shinfo(skb)->gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636}
637
638/* This is valid iff tcp_skb_pcount() > 1. */
639static inline int tcp_skb_mss(const struct sk_buff *skb)
640{
Herbert Xu79671682006-06-22 02:40:14 -0700641 return skb_shinfo(skb)->gso_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642}
643
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700644/* Events passed to congestion control interface */
645enum tcp_ca_event {
646 CA_EVENT_TX_START, /* first transmit when no packets in flight */
647 CA_EVENT_CWND_RESTART, /* congestion window restart */
648 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
649 CA_EVENT_FRTO, /* fast recovery timeout */
650 CA_EVENT_LOSS, /* loss timeout */
651 CA_EVENT_FAST_ACK, /* in sequence ack */
652 CA_EVENT_SLOW_ACK, /* other ack */
653};
654
655/*
656 * Interface for adding new TCP congestion control handlers
657 */
658#define TCP_CA_NAME_MAX 16
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800659#define TCP_CA_MAX 128
660#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
661
Stephen Hemminger164891a2007-04-23 22:26:16 -0700662#define TCP_CONG_NON_RESTRICTED 0x1
663#define TCP_CONG_RTT_STAMP 0x2
664
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700665struct tcp_congestion_ops {
666 struct list_head list;
Stephen Hemminger164891a2007-04-23 22:26:16 -0700667 unsigned long flags;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700668
669 /* initialize private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300670 void (*init)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700671 /* cleanup private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300672 void (*release)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700673
674 /* return slow start threshold (required) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300675 u32 (*ssthresh)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700676 /* lower bound for congestion window (optional) */
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700677 u32 (*min_cwnd)(const struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700678 /* do new cwnd calculation (required) */
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200679 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700680 /* call before changing ca_state (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300681 void (*set_state)(struct sock *sk, u8 new_state);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700682 /* call when cwnd event occurs (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300683 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700684 /* new value of cwnd after loss (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300685 u32 (*undo_cwnd)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700686 /* hook for packet ack accounting (optional) */
Stephen Hemminger30cfd0b2007-07-25 23:49:34 -0700687 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
Arnaldo Carvalho de Melo73c1f4a2005-08-12 12:51:49 -0300688 /* get info for inet_diag (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300689 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700690
691 char name[TCP_CA_NAME_MAX];
692 struct module *owner;
693};
694
695extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
696extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
697
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300698extern void tcp_init_congestion_control(struct sock *sk);
699extern void tcp_cleanup_congestion_control(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700700extern int tcp_set_default_congestion_control(const char *name);
701extern void tcp_get_default_congestion_control(char *name);
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800702extern void tcp_get_available_congestion_control(char *buf, size_t len);
Stephen Hemmingerce7bc3b2006-11-09 16:35:15 -0800703extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
704extern int tcp_set_allowed_congestion_control(char *allowed);
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300705extern int tcp_set_congestion_control(struct sock *sk, const char *name);
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800706extern void tcp_slow_start(struct tcp_sock *tp);
Ilpo Järvinen758ce5c2009-02-28 04:44:37 +0000707extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700708
Stephen Hemminger5f8ef482005-06-23 20:37:36 -0700709extern struct tcp_congestion_ops tcp_init_congestion_ops;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300710extern u32 tcp_reno_ssthresh(struct sock *sk);
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200711extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700712extern u32 tcp_reno_min_cwnd(const struct sock *sk);
David S. Millera8acfba2005-06-23 23:45:02 -0700713extern struct tcp_congestion_ops tcp_reno;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700714
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300715static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700716{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300717 struct inet_connection_sock *icsk = inet_csk(sk);
718
719 if (icsk->icsk_ca_ops->set_state)
720 icsk->icsk_ca_ops->set_state(sk, ca_state);
721 icsk->icsk_ca_state = ca_state;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700722}
723
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300724static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700725{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300726 const struct inet_connection_sock *icsk = inet_csk(sk);
727
728 if (icsk->icsk_ca_ops->cwnd_event)
729 icsk->icsk_ca_ops->cwnd_event(sk, event);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700730}
731
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300732/* These functions determine how the current flow behaves in respect of SACK
733 * handling. SACK is negotiated with the peer, and therefore it can vary
734 * between different flows.
735 *
736 * tcp_is_sack - SACK enabled
737 * tcp_is_reno - No SACK
738 * tcp_is_fack - FACK enabled, implies SACK enabled
739 */
740static inline int tcp_is_sack(const struct tcp_sock *tp)
741{
742 return tp->rx_opt.sack_ok;
743}
744
745static inline int tcp_is_reno(const struct tcp_sock *tp)
746{
747 return !tcp_is_sack(tp);
748}
749
750static inline int tcp_is_fack(const struct tcp_sock *tp)
751{
752 return tp->rx_opt.sack_ok & 2;
753}
754
755static inline void tcp_enable_fack(struct tcp_sock *tp)
756{
757 tp->rx_opt.sack_ok |= 2;
758}
759
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300760static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
761{
762 return tp->sacked_out + tp->lost_out;
763}
764
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765/* This determines how many packets are "in the network" to the best
766 * of our knowledge. In many cases it is conservative, but where
767 * detailed information is available from the receiver (via SACK
768 * blocks etc.) we can make more aggressive calculations.
769 *
770 * Use this for decisions involving congestion control, use just
771 * tp->packets_out to determine if the send queue is empty or not.
772 *
773 * Read this equation as:
774 *
775 * "Packets sent once on transmission queue" MINUS
776 * "Packets left network, but not honestly ACKed yet" PLUS
777 * "Packets fast retransmitted"
778 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800779static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780{
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300781 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782}
783
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
785 * The exception is rate halving phase, when cwnd is decreasing towards
786 * ssthresh.
787 */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300788static inline __u32 tcp_current_ssthresh(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300790 const struct tcp_sock *tp = tcp_sk(sk);
791 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 return tp->snd_ssthresh;
793 else
794 return max(tp->snd_ssthresh,
795 ((tp->snd_cwnd >> 1) +
796 (tp->snd_cwnd >> 2)));
797}
798
Ilpo Järvinenb9c45952007-07-27 16:36:17 +0300799/* Use define here intentionally to get WARN_ON location shown at the caller */
800#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800802extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
804
805/* Slow start with delack produces 3 packets of burst, so that
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700806 * it is safe "de facto". This will be the default - same as
807 * the default reordering threshold - but if reordering increases,
808 * we must be able to allow cwnd to burst at least this much in order
809 * to not pull it back when holes are filled.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 */
811static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
812{
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700813 return tp->reordering;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814}
815
Ilpo Järvinen90840de2007-12-31 04:48:41 -0800816/* Returns end sequence number of the receiver's advertised window */
817static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
818{
819 return tp->snd_una + tp->snd_wnd;
820}
Ilpo Järvinencea14e02008-01-12 03:19:12 -0800821extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800822
Chuck Leverc1bd24b2007-10-23 21:08:54 -0700823static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800824 const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825{
826 if (skb->len < mss)
827 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
828}
829
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700830static inline void tcp_check_probe_timer(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700832 struct tcp_sock *tp = tcp_sk(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700833 const struct inet_connection_sock *icsk = inet_csk(sk);
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700834
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700835 if (!tp->packets_out && !icsk->icsk_pending)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700836 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
837 icsk->icsk_rto, TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838}
839
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700840static inline void tcp_push_pending_frames(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700842 struct tcp_sock *tp = tcp_sk(sk);
843
Ilpo Järvinen0c54b852009-03-14 14:23:05 +0000844 __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845}
846
Hantzis Fotisee7537b2009-03-02 22:42:02 -0800847static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848{
849 tp->snd_wl1 = seq;
850}
851
Hantzis Fotisee7537b2009-03-02 22:42:02 -0800852static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853{
854 tp->snd_wl1 = seq;
855}
856
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857/*
858 * Calculate(/check) TCP checksum
859 */
Frederik Deweerdtba7808e2007-02-04 20:15:27 -0800860static inline __sum16 tcp_v4_check(int len, __be32 saddr,
861 __be32 daddr, __wsum base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862{
863 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
864}
865
Al Virob51655b2006-11-14 21:40:42 -0800866static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867{
Herbert Xufb286bb2005-11-10 13:01:24 -0800868 return __skb_checksum_complete(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869}
870
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800871static inline int tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872{
Herbert Xu60476372007-04-09 11:59:39 -0700873 return !skb_csum_unnecessary(skb) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 __tcp_checksum_complete(skb);
875}
876
877/* Prequeue for VJ style copy to user, combined with checksumming. */
878
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800879static inline void tcp_prequeue_init(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880{
881 tp->ucopy.task = NULL;
882 tp->ucopy.len = 0;
883 tp->ucopy.memory = 0;
884 skb_queue_head_init(&tp->ucopy.prequeue);
Chris Leech97fc2f02006-05-23 17:55:33 -0700885#ifdef CONFIG_NET_DMA
886 tp->ucopy.dma_chan = NULL;
887 tp->ucopy.wakeup = 0;
888 tp->ucopy.pinned_list = NULL;
889 tp->ucopy.dma_cookie = 0;
890#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891}
892
893/* Packet is added to VJ-style prequeue for processing in process
894 * context, if a reader task is waiting. Apparently, this exciting
895 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
896 * failed somewhere. Latency? Burstiness? Well, at least now we will
897 * see, why it failed. 8)8) --ANK
898 *
899 * NOTE: is this not too big to inline?
900 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800901static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902{
903 struct tcp_sock *tp = tcp_sk(sk);
904
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000905 if (sysctl_tcp_low_latency || !tp->ucopy.task)
906 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000908 __skb_queue_tail(&tp->ucopy.prequeue, skb);
909 tp->ucopy.memory += skb->truesize;
910 if (tp->ucopy.memory > sk->sk_rcvbuf) {
911 struct sk_buff *skb1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000913 BUG_ON(sock_owned_by_user(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000915 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
916 sk_backlog_rcv(sk, skb1);
917 NET_INC_STATS_BH(sock_net(sk),
918 LINUX_MIB_TCPPREQUEUEDROPPED);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 }
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000920
921 tp->ucopy.memory = 0;
922 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
Eric Dumazet7aedec22009-05-07 07:20:39 +0000923 wake_up_interruptible_poll(sk->sk_sleep,
924 POLLIN | POLLRDNORM | POLLRDBAND);
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000925 if (!inet_csk_ack_scheduled(sk))
926 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
David S. Miller22f6dac2009-05-08 02:48:30 -0700927 (3 * tcp_rto_min(sk)) / 4,
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000928 TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 }
Eric Dumazetf5f8d862009-05-07 07:08:38 +0000930 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931}
932
933
934#undef STATE_TRACE
935
936#ifdef STATE_TRACE
937static const char *statename[]={
938 "Unused","Established","Syn Sent","Syn Recv",
939 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
940 "Close Wait","Last ACK","Listen","Closing"
941};
942#endif
Ilpo Järvinen490d5042008-01-12 03:17:20 -0800943extern void tcp_set_state(struct sock *sk, int state);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944
Andi Kleen4ac02ba2007-04-20 17:11:46 -0700945extern void tcp_done(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800947static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948{
949 rx_opt->dsack = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 rx_opt->num_sacks = 0;
951}
952
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953/* Determine a window scaling and initial window to offer. */
954extern void tcp_select_initial_window(int __space, __u32 mss,
955 __u32 *rcv_wnd, __u32 *window_clamp,
956 int wscale_ok, __u8 *rcv_wscale);
957
958static inline int tcp_win_from_space(int space)
959{
960 return sysctl_tcp_adv_win_scale<=0 ?
961 (space>>(-sysctl_tcp_adv_win_scale)) :
962 space - (space>>sysctl_tcp_adv_win_scale);
963}
964
965/* Note: caller must be prepared to deal with negative returns */
966static inline int tcp_space(const struct sock *sk)
967{
968 return tcp_win_from_space(sk->sk_rcvbuf -
969 atomic_read(&sk->sk_rmem_alloc));
970}
971
972static inline int tcp_full_space(const struct sock *sk)
973{
974 return tcp_win_from_space(sk->sk_rcvbuf);
975}
976
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800977static inline void tcp_openreq_init(struct request_sock *req,
978 struct tcp_options_received *rx_opt,
979 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700981 struct inet_request_sock *ireq = inet_rsk(req);
982
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
Florian Westphal4dfc2812008-04-10 03:12:40 -0700984 req->cookie_ts = 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700985 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 req->mss = rx_opt->mss_clamp;
987 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700988 ireq->tstamp_ok = rx_opt->tstamp_ok;
989 ireq->sack_ok = rx_opt->sack_ok;
990 ireq->snd_wscale = rx_opt->snd_wscale;
991 ireq->wscale_ok = rx_opt->wscale_ok;
992 ireq->acked = 0;
993 ireq->ecn_ok = 0;
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -0700994 ireq->rmt_port = tcp_hdr(skb)->source;
KOVACS Krisztiana3116ac52008-10-01 07:46:49 -0700995 ireq->loc_port = tcp_hdr(skb)->dest;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996}
997
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700998extern void tcp_enter_memory_pressure(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1001{
1002 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1003}
1004
1005static inline int keepalive_time_when(const struct tcp_sock *tp)
1006{
1007 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1008}
1009
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001010static inline int tcp_fin_time(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001012 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1013 const int rto = inet_csk(sk)->icsk_rto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001015 if (fin_timeout < (rto << 2) - (rto >> 1))
1016 fin_timeout = (rto << 2) - (rto >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017
1018 return fin_timeout;
1019}
1020
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001021static inline int tcp_paws_check(const struct tcp_options_received *rx_opt,
1022 int paws_win)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023{
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001024 if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
1025 return 1;
1026 if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
1027 return 1;
1028
1029 return 0;
1030}
1031
1032static inline int tcp_paws_reject(const struct tcp_options_received *rx_opt,
1033 int rst)
1034{
1035 if (tcp_paws_check(rx_opt, 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 return 0;
1037
1038 /* RST segments are not recommended to carry timestamp,
1039 and, if they do, it is recommended to ignore PAWS because
1040 "their cleanup function should take precedence over timestamps."
1041 Certainly, it is mistake. It is necessary to understand the reasons
1042 of this constraint to relax it: if peer reboots, clock may go
1043 out-of-sync and half-open connections will not be reset.
1044 Actually, the problem would be not existing if all
1045 the implementations followed draft about maintaining clock
1046 via reboots. Linux-2.2 DOES NOT!
1047
1048 However, we can relax time bounds for RST segments to MSL.
1049 */
James Morris9d729f72007-03-04 16:12:44 -08001050 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 return 0;
1052 return 1;
1053}
1054
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055#define TCP_CHECK_TIMER(sk) do { } while (0)
1056
Pavel Emelyanova9c193292008-07-16 20:21:42 -07001057static inline void tcp_mib_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058{
1059 /* See RFC 2012 */
Pavel Emelyanovcf1100a2008-07-16 20:27:38 -07001060 TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1061 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1062 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1063 TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064}
1065
Ilpo Järvinen5af4ec22007-09-20 11:30:48 -07001066/* from STCP */
Ilpo Järvinenef9da472008-09-20 21:25:15 -07001067static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
David S. Miller0800f172007-09-20 11:40:37 -07001068{
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001069 tp->lost_skb_hint = NULL;
1070 tp->scoreboard_skb_hint = NULL;
Ilpo Järvinenef9da472008-09-20 21:25:15 -07001071}
1072
1073static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1074{
1075 tcp_clear_retrans_hints_partial(tp);
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001076 tp->retransmit_skb_hint = NULL;
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001077}
1078
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001079/* MD5 Signature */
1080struct crypto_hash;
1081
1082/* - key database */
1083struct tcp_md5sig_key {
1084 u8 *key;
1085 u8 keylen;
1086};
1087
1088struct tcp4_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001089 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001090 __be32 addr;
1091};
1092
1093struct tcp6_md5sig_key {
David S. Millerf8ab18d2007-09-28 15:18:35 -07001094 struct tcp_md5sig_key base;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001095#if 0
1096 u32 scope_id; /* XXX */
1097#endif
1098 struct in6_addr addr;
1099};
1100
1101/* - sock block */
1102struct tcp_md5sig_info {
1103 struct tcp4_md5sig_key *keys4;
1104#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1105 struct tcp6_md5sig_key *keys6;
1106 u32 entries6;
1107 u32 alloced6;
1108#endif
1109 u32 entries4;
1110 u32 alloced4;
1111};
1112
1113/* - pseudo header */
1114struct tcp4_pseudohdr {
1115 __be32 saddr;
1116 __be32 daddr;
1117 __u8 pad;
1118 __u8 protocol;
1119 __be16 len;
1120};
1121
1122struct tcp6_pseudohdr {
1123 struct in6_addr saddr;
1124 struct in6_addr daddr;
1125 __be32 len;
1126 __be32 protocol; /* including padding */
1127};
1128
1129union tcp_md5sum_block {
1130 struct tcp4_pseudohdr ip4;
1131#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1132 struct tcp6_pseudohdr ip6;
1133#endif
1134};
1135
1136/* - pool: digest algorithm, hash description and scratch buffer */
1137struct tcp_md5sig_pool {
1138 struct hash_desc md5_desc;
1139 union tcp_md5sum_block md5_blk;
1140};
1141
1142#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1143
1144/* - functions */
Adam Langley49a72df2008-07-19 00:01:42 -07001145extern int tcp_v4_md5_hash_skb(char *md5_hash,
1146 struct tcp_md5sig_key *key,
1147 struct sock *sk,
1148 struct request_sock *req,
1149 struct sk_buff *skb);
YOSHIFUJI Hideaki8d26d762008-04-17 13:19:16 +09001150
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001151extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1152 struct sock *addr_sk);
1153
1154extern int tcp_v4_md5_do_add(struct sock *sk,
1155 __be32 addr,
1156 u8 *newkey,
1157 u8 newkeylen);
1158
1159extern int tcp_v4_md5_do_del(struct sock *sk,
Al Viro8e5200f2006-11-20 18:06:37 -08001160 __be32 addr);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001161
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +09001162#ifdef CONFIG_TCP_MD5SIG
1163#define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_keylen ? \
1164 &(struct tcp_md5sig_key) { \
1165 .key = (twsk)->tw_md5_key, \
1166 .keylen = (twsk)->tw_md5_keylen, \
1167 } : NULL)
1168#else
1169#define tcp_twsk_md5_key(twsk) NULL
1170#endif
1171
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001172extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1173extern void tcp_free_md5sig_pool(void);
1174
1175extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1176extern void __tcp_put_md5sig_pool(void);
Adam Langley49a72df2008-07-19 00:01:42 -07001177extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, struct tcphdr *);
1178extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, struct sk_buff *,
1179 unsigned header_len);
1180extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
1181 struct tcp_md5sig_key *key);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001182
1183static inline
1184struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1185{
1186 int cpu = get_cpu();
1187 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1188 if (!ret)
1189 put_cpu();
1190 return ret;
1191}
1192
1193static inline void tcp_put_md5sig_pool(void)
1194{
1195 __tcp_put_md5sig_pool();
1196 put_cpu();
1197}
1198
David S. Millerfe067e82007-03-07 12:12:44 -08001199/* write queue abstraction */
1200static inline void tcp_write_queue_purge(struct sock *sk)
1201{
1202 struct sk_buff *skb;
1203
1204 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001205 sk_wmem_free_skb(sk, skb);
1206 sk_mem_reclaim(sk);
David S. Millerfe067e82007-03-07 12:12:44 -08001207}
1208
1209static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1210{
David S. Millercd07a8e2008-09-23 00:50:13 -07001211 return skb_peek(&sk->sk_write_queue);
David S. Millerfe067e82007-03-07 12:12:44 -08001212}
1213
1214static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1215{
David S. Millercd07a8e2008-09-23 00:50:13 -07001216 return skb_peek_tail(&sk->sk_write_queue);
David S. Millerfe067e82007-03-07 12:12:44 -08001217}
1218
1219static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1220{
David S. Millercd07a8e2008-09-23 00:50:13 -07001221 return skb_queue_next(&sk->sk_write_queue, skb);
David S. Millerfe067e82007-03-07 12:12:44 -08001222}
1223
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001224static inline struct sk_buff *tcp_write_queue_prev(struct sock *sk, struct sk_buff *skb)
1225{
1226 return skb_queue_prev(&sk->sk_write_queue, skb);
1227}
1228
David S. Millerfe067e82007-03-07 12:12:44 -08001229#define tcp_for_write_queue(skb, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001230 skb_queue_walk(&(sk)->sk_write_queue, skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001231
1232#define tcp_for_write_queue_from(skb, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001233 skb_queue_walk_from(&(sk)->sk_write_queue, skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001234
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001235#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001236 skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001237
David S. Millerfe067e82007-03-07 12:12:44 -08001238static inline struct sk_buff *tcp_send_head(struct sock *sk)
1239{
1240 return sk->sk_send_head;
1241}
1242
David S. Millercd07a8e2008-09-23 00:50:13 -07001243static inline bool tcp_skb_is_last(const struct sock *sk,
1244 const struct sk_buff *skb)
1245{
1246 return skb_queue_is_last(&sk->sk_write_queue, skb);
1247}
1248
David S. Millerfe067e82007-03-07 12:12:44 -08001249static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1250{
David S. Millercd07a8e2008-09-23 00:50:13 -07001251 if (tcp_skb_is_last(sk, skb))
David S. Millerfe067e82007-03-07 12:12:44 -08001252 sk->sk_send_head = NULL;
David S. Millercd07a8e2008-09-23 00:50:13 -07001253 else
1254 sk->sk_send_head = tcp_write_queue_next(sk, skb);
David S. Millerfe067e82007-03-07 12:12:44 -08001255}
1256
1257static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1258{
1259 if (sk->sk_send_head == skb_unlinked)
1260 sk->sk_send_head = NULL;
1261}
1262
1263static inline void tcp_init_send_head(struct sock *sk)
1264{
1265 sk->sk_send_head = NULL;
1266}
1267
1268static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1269{
1270 __skb_queue_tail(&sk->sk_write_queue, skb);
1271}
1272
1273static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1274{
1275 __tcp_add_write_queue_tail(sk, skb);
1276
1277 /* Queue it, remembering where we must start sending. */
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001278 if (sk->sk_send_head == NULL) {
David S. Millerfe067e82007-03-07 12:12:44 -08001279 sk->sk_send_head = skb;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001280
1281 if (tcp_sk(sk)->highest_sack == NULL)
1282 tcp_sk(sk)->highest_sack = skb;
1283 }
David S. Millerfe067e82007-03-07 12:12:44 -08001284}
1285
1286static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1287{
1288 __skb_queue_head(&sk->sk_write_queue, skb);
1289}
1290
1291/* Insert buff after skb on the write queue of sk. */
1292static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1293 struct sk_buff *buff,
1294 struct sock *sk)
1295{
Gerrit Renker7de6c032008-04-14 00:05:09 -07001296 __skb_queue_after(&sk->sk_write_queue, skb, buff);
David S. Millerfe067e82007-03-07 12:12:44 -08001297}
1298
David S. Miller43f59c82008-09-21 21:28:51 -07001299/* Insert new before skb on the write queue of sk. */
David S. Millerfe067e82007-03-07 12:12:44 -08001300static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1301 struct sk_buff *skb,
1302 struct sock *sk)
1303{
David S. Miller43f59c82008-09-21 21:28:51 -07001304 __skb_queue_before(&sk->sk_write_queue, skb, new);
Ilpo Järvinen6e421412007-11-19 23:24:09 -08001305
1306 if (sk->sk_send_head == skb)
1307 sk->sk_send_head = new;
David S. Millerfe067e82007-03-07 12:12:44 -08001308}
1309
1310static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1311{
1312 __skb_unlink(skb, &sk->sk_write_queue);
1313}
1314
David S. Millerfe067e82007-03-07 12:12:44 -08001315static inline int tcp_write_queue_empty(struct sock *sk)
1316{
1317 return skb_queue_empty(&sk->sk_write_queue);
1318}
1319
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001320/* Start sequence of the highest skb with SACKed bit, valid only if
1321 * sacked > 0 or when the caller has ensured validity by itself.
1322 */
1323static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1324{
1325 if (!tp->sacked_out)
1326 return tp->snd_una;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001327
1328 if (tp->highest_sack == NULL)
1329 return tp->snd_nxt;
1330
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001331 return TCP_SKB_CB(tp->highest_sack)->seq;
1332}
1333
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001334static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1335{
1336 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1337 tcp_write_queue_next(sk, skb);
1338}
1339
1340static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1341{
1342 return tcp_sk(sk)->highest_sack;
1343}
1344
1345static inline void tcp_highest_sack_reset(struct sock *sk)
1346{
1347 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1348}
1349
1350/* Called when old skb is about to be deleted (to be combined with new skb) */
1351static inline void tcp_highest_sack_combine(struct sock *sk,
1352 struct sk_buff *old,
1353 struct sk_buff *new)
1354{
1355 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1356 tcp_sk(sk)->highest_sack = new;
1357}
1358
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359/* /proc */
1360enum tcp_seq_states {
1361 TCP_SEQ_STATE_LISTENING,
1362 TCP_SEQ_STATE_OPENREQ,
1363 TCP_SEQ_STATE_ESTABLISHED,
1364 TCP_SEQ_STATE_TIME_WAIT,
1365};
1366
1367struct tcp_seq_afinfo {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 char *name;
1369 sa_family_t family;
Denis V. Lunev68fcadd2008-04-13 22:13:30 -07001370 struct file_operations seq_fops;
Denis V. Lunev9427c4b2008-04-13 22:12:13 -07001371 struct seq_operations seq_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372};
1373
1374struct tcp_iter_state {
Denis V. Luneva4146b12008-04-13 22:11:14 -07001375 struct seq_net_private p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 sa_family_t family;
1377 enum tcp_seq_states state;
1378 struct sock *syn_wait_sk;
1379 int bucket, sbucket, num, uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380};
1381
Daniel Lezcano6f8b13b2008-03-21 04:14:45 -07001382extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1383extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001385extern struct request_sock_ops tcp_request_sock_ops;
Glenn Griffinc6aefaf2008-02-07 21:49:26 -08001386extern struct request_sock_ops tcp6_request_sock_ops;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001387
Brian Haley7d06b2e2008-06-14 17:04:49 -07001388extern void tcp_v4_destroy_sock(struct sock *sk);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001389
Herbert Xua430a432006-07-08 13:34:56 -07001390extern int tcp_v4_gso_send_check(struct sk_buff *skb);
Herbert Xu576a30e2006-06-27 13:22:38 -07001391extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
Herbert Xubf296b12008-12-15 23:43:36 -08001392extern struct sk_buff **tcp_gro_receive(struct sk_buff **head,
1393 struct sk_buff *skb);
1394extern struct sk_buff **tcp4_gro_receive(struct sk_buff **head,
1395 struct sk_buff *skb);
1396extern int tcp_gro_complete(struct sk_buff *skb);
1397extern int tcp4_gro_complete(struct sk_buff *skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001398
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001399#ifdef CONFIG_PROC_FS
1400extern int tcp4_proc_init(void);
1401extern void tcp4_proc_exit(void);
1402#endif
1403
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001404/* TCP af-specific functions */
1405struct tcp_sock_af_ops {
1406#ifdef CONFIG_TCP_MD5SIG
1407 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1408 struct sock *addr_sk);
1409 int (*calc_md5_hash) (char *location,
1410 struct tcp_md5sig_key *md5,
1411 struct sock *sk,
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001412 struct request_sock *req,
Adam Langley49a72df2008-07-19 00:01:42 -07001413 struct sk_buff *skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001414 int (*md5_add) (struct sock *sk,
1415 struct sock *addr_sk,
1416 u8 *newkey,
1417 u8 len);
1418 int (*md5_parse) (struct sock *sk,
1419 char __user *optval,
1420 int optlen);
1421#endif
1422};
1423
1424struct tcp_request_sock_ops {
1425#ifdef CONFIG_TCP_MD5SIG
1426 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1427 struct request_sock *req);
1428#endif
1429};
1430
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001431extern void tcp_v4_init(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001432extern void tcp_init(void);
1433
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434#endif /* _TCP_H */