blob: 56fc366da6d5183b536648e949769a118ba33677 [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#define FASTRETRANS_DEBUG 1
22
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <linux/list.h>
24#include <linux/tcp.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050025#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#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>
Glauber Costa180d8cd2011-12-11 21:47:02 +000048#include <linux/memcontrol.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049
Arnaldo Carvalho de Melo6e04e022005-08-09 20:07:35 -070050extern struct inet_hashinfo tcp_hashinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -070051
Eric Dumazetdd24c002008-11-25 21:17:14 -080052extern struct percpu_counter tcp_orphan_count;
Joe Perches5c9f3022013-09-23 11:33:32 -070053void tcp_time_wait(struct sock *sk, int state, int timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -070054
Linus Torvalds1da177e2005-04-16 15:20:36 -070055#define MAX_TCP_HEADER (128 + MAX_HEADER)
Adam Langley33ad7982008-07-19 00:04:31 -070056#define MAX_TCP_OPTION_SPACE 40
Linus Torvalds1da177e2005-04-16 15:20:36 -070057
58/*
59 * Never offer a window over 32767 without using window scaling. Some
60 * poor stacks do signed 16bit maths!
61 */
62#define MAX_TCP_WINDOW 32767U
63
64/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
65#define TCP_MIN_MSS 88U
66
John Heffner5d424d52006-03-20 17:53:41 -080067/* The least MTU to use for probing */
68#define TCP_BASE_MSS 512
69
Linus Torvalds1da177e2005-04-16 15:20:36 -070070/* After receiving this amount of duplicate ACKs fast retransmit starts. */
71#define TCP_FASTRETRANS_THRESH 3
72
73/* Maximal reordering. */
74#define TCP_MAX_REORDERING 127
75
76/* Maximal number of ACKs sent quickly to accelerate slow-start. */
77#define TCP_MAX_QUICKACKS 16U
78
79/* urg_data states */
80#define TCP_URG_VALID 0x0100
81#define TCP_URG_NOTYET 0x0200
82#define TCP_URG_READ 0x0400
83
84#define TCP_RETR1 3 /*
85 * This is how many retries it does before it
86 * tries to figure out if the gateway is
87 * down. Minimal RFC value is 3; it corresponds
88 * to ~3sec-8min depending on RTO.
89 */
90
91#define TCP_RETR2 15 /*
92 * This should take at least
93 * 90 minutes to time out.
94 * RFC1122 says that the limit is 100 sec.
95 * 15 is ~13-30min depending on RTO.
96 */
97
Alex Bergmann6c9ff972012-08-31 02:48:31 +000098#define TCP_SYN_RETRIES 6 /* This is how many retries are done
99 * when active opening a connection.
100 * RFC1122 says the minimum retry MUST
101 * be at least 180secs. Nevertheless
102 * this value is corresponding to
103 * 63secs of retransmission with the
104 * current initial RTO.
105 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
Alex Bergmann6c9ff972012-08-31 02:48:31 +0000107#define TCP_SYNACK_RETRIES 5 /* This is how may retries are done
108 * when passive opening a connection.
109 * This is corresponding to 31secs of
110 * retransmission with the current
111 * initial RTO.
112 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
115 * state, about 60 seconds */
116#define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
117 /* BSD style FIN_WAIT2 deadlock breaker.
118 * It used to be 3min, new value is 60sec,
119 * to combine FIN-WAIT-2 timeout with
120 * TIME-WAIT timer.
121 */
122
123#define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
124#if HZ >= 100
125#define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
126#define TCP_ATO_MIN ((unsigned)(HZ/25))
127#else
128#define TCP_DELACK_MIN 4U
129#define TCP_ATO_MIN 4U
130#endif
131#define TCP_RTO_MAX ((unsigned)(120*HZ))
132#define TCP_RTO_MIN ((unsigned)(HZ/5))
Eric Dumazetfd4f2ce2012-04-12 19:48:40 +0000133#define TCP_TIMEOUT_INIT ((unsigned)(1*HZ)) /* RFC6298 2.1 initial RTO value */
Jerry Chu9ad7c042011-06-08 11:08:38 +0000134#define TCP_TIMEOUT_FALLBACK ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value, now
135 * used as a fallback RTO for the
136 * initial data transmission if no
137 * valid RTT sample has been acquired,
138 * most likely due to retrans in 3WHS.
139 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140
141#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
142 * for local resources.
143 */
144
145#define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
146#define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
147#define TCP_KEEPALIVE_INTVL (75*HZ)
148
149#define MAX_TCP_KEEPIDLE 32767
150#define MAX_TCP_KEEPINTVL 32767
151#define MAX_TCP_KEEPCNT 127
152#define MAX_TCP_SYNCNT 127
153
154#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
156#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
157#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
158 * after this time. It should be equal
159 * (or greater than) TCP_TIMEWAIT_LEN
160 * to provide reliability equal to one
161 * provided by timewait state.
162 */
163#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
164 * timestamps. It must be less than
165 * minimal timewait lifetime.
166 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167/*
168 * TCP option
169 */
170
171#define TCPOPT_NOP 1 /* Padding */
172#define TCPOPT_EOL 0 /* End of options */
173#define TCPOPT_MSS 2 /* Segment size negotiating */
174#define TCPOPT_WINDOW 3 /* Window scaling */
175#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
176#define TCPOPT_SACK 5 /* SACK Block */
177#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800178#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
Yuchung Cheng2100c8d2012-07-19 06:43:05 +0000179#define TCPOPT_EXP 254 /* Experimental */
180/* Magic number to be after the option value for sharing TCP
181 * experimental options. See draft-ietf-tcpm-experimental-options-00.txt
182 */
183#define TCPOPT_FASTOPEN_MAGIC 0xF989
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184
185/*
186 * TCP option lengths
187 */
188
189#define TCPOLEN_MSS 4
190#define TCPOLEN_WINDOW 3
191#define TCPOLEN_SACK_PERM 2
192#define TCPOLEN_TIMESTAMP 10
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800193#define TCPOLEN_MD5SIG 18
Yuchung Cheng2100c8d2012-07-19 06:43:05 +0000194#define TCPOLEN_EXP_FASTOPEN_BASE 4
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195
196/* But this is what stacks really send out. */
197#define TCPOLEN_TSTAMP_ALIGNED 12
198#define TCPOLEN_WSCALE_ALIGNED 4
199#define TCPOLEN_SACKPERM_ALIGNED 4
200#define TCPOLEN_SACK_BASE 2
201#define TCPOLEN_SACK_BASE_ALIGNED 4
202#define TCPOLEN_SACK_PERBLOCK 8
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800203#define TCPOLEN_MD5SIG_ALIGNED 20
Adam Langley33ad7982008-07-19 00:04:31 -0700204#define TCPOLEN_MSS_ALIGNED 4
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206/* Flags in tp->nonagle */
207#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
208#define TCP_NAGLE_CORK 2 /* Socket is corked */
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800209#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210
Andreas Petlund36e31b0a2010-02-18 02:47:01 +0000211/* TCP thin-stream limits */
212#define TCP_THIN_LINEAR_RETRIES 6 /* After 6 linear retries, do exp. backoff */
213
David S. Miller7eb38522011-02-05 18:13:45 -0800214/* TCP initial congestion window as per draft-hkchu-tcpm-initcwnd-01 */
David S. Miller442b9632011-02-02 17:05:11 -0800215#define TCP_INIT_CWND 10
216
Yuchung Chengcf60af02012-07-19 06:43:09 +0000217/* Bit Flags for sysctl_tcp_fastopen */
218#define TFO_CLIENT_ENABLE 1
Jerry Chu10467162012-08-31 12:29:11 +0000219#define TFO_SERVER_ENABLE 2
Yuchung Cheng67da22d2012-07-19 06:43:11 +0000220#define TFO_CLIENT_NO_COOKIE 4 /* Data in SYN w/o cookie option */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000221
Jerry Chu10467162012-08-31 12:29:11 +0000222/* Process SYN data but skip cookie validation */
223#define TFO_SERVER_COOKIE_NOT_CHKED 0x100
224/* Accept SYN data w/o any cookie option */
225#define TFO_SERVER_COOKIE_NOT_REQD 0x200
226
227/* Force enable TFO on all listeners, i.e., not requiring the
228 * TCP_FASTOPEN socket option. SOCKOPT1/2 determine how to set max_qlen.
229 */
230#define TFO_SERVER_WO_SOCKOPT1 0x400
231#define TFO_SERVER_WO_SOCKOPT2 0x800
232/* Always create TFO child sockets on a TFO listener even when
233 * cookie/data not present. (For testing purpose!)
234 */
235#define TFO_SERVER_ALWAYS 0x1000
236
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -0700237extern struct inet_timewait_death_row tcp_death_row;
238
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239/* sysctl variables for tcp */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240extern int sysctl_tcp_timestamps;
241extern int sysctl_tcp_window_scaling;
242extern int sysctl_tcp_sack;
243extern int sysctl_tcp_fin_timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244extern int sysctl_tcp_keepalive_time;
245extern int sysctl_tcp_keepalive_probes;
246extern int sysctl_tcp_keepalive_intvl;
247extern int sysctl_tcp_syn_retries;
248extern int sysctl_tcp_synack_retries;
249extern int sysctl_tcp_retries1;
250extern int sysctl_tcp_retries2;
251extern int sysctl_tcp_orphan_retries;
252extern int sysctl_tcp_syncookies;
Yuchung Cheng2100c8d2012-07-19 06:43:05 +0000253extern int sysctl_tcp_fastopen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254extern int sysctl_tcp_retrans_collapse;
255extern int sysctl_tcp_stdurg;
256extern int sysctl_tcp_rfc1337;
257extern int sysctl_tcp_abort_on_overflow;
258extern int sysctl_tcp_max_orphans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259extern int sysctl_tcp_fack;
260extern int sysctl_tcp_reordering;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261extern int sysctl_tcp_dsack;
Eric W. Biedermana4fe34b2013-10-19 16:25:36 -0700262extern long sysctl_tcp_mem[3];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263extern int sysctl_tcp_wmem[3];
264extern int sysctl_tcp_rmem[3];
265extern int sysctl_tcp_app_win;
266extern int sysctl_tcp_adv_win_scale;
267extern int sysctl_tcp_tw_reuse;
268extern int sysctl_tcp_frto;
269extern int sysctl_tcp_low_latency;
Chris Leech95937822006-05-23 18:02:55 -0700270extern int sysctl_tcp_dma_copybreak;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271extern int sysctl_tcp_nometrics_save;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272extern int sysctl_tcp_moderate_rcvbuf;
273extern int sysctl_tcp_tso_win_divisor;
John Heffner5d424d52006-03-20 17:53:41 -0800274extern int sysctl_tcp_mtu_probing;
275extern int sysctl_tcp_base_mss;
Rick Jones15d99e02006-03-20 22:40:29 -0800276extern int sysctl_tcp_workaround_signed_windows;
David S. Miller35089bb2006-06-13 22:33:04 -0700277extern int sysctl_tcp_slow_start_after_idle;
Andreas Petlund36e31b0a2010-02-18 02:47:01 +0000278extern int sysctl_tcp_thin_linear_timeouts;
Andreas Petlund7e380172010-02-18 04:48:19 +0000279extern int sysctl_tcp_thin_dupack;
Yuchung Chengeed530b2012-05-02 13:30:03 +0000280extern int sysctl_tcp_early_retrans;
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000281extern int sysctl_tcp_limit_output_bytes;
Eric Dumazet282f23c2012-07-17 10:13:05 +0200282extern int sysctl_tcp_challenge_ack_limit;
Eric Dumazetc9bee3b72013-07-22 20:27:07 -0700283extern unsigned int sysctl_tcp_notsent_lowat;
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700284extern int sysctl_tcp_min_tso_segs;
Eric Dumazetf54b3112013-12-05 22:36:05 -0800285extern int sysctl_tcp_autocorking;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
Eric Dumazet8d987e52010-11-09 23:24:26 +0000287extern atomic_long_t tcp_memory_allocated;
Eric Dumazet17483762008-11-25 21:16:35 -0800288extern struct percpu_counter tcp_sockets_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289extern int tcp_memory_pressure;
290
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 * The next routines deal with comparing 32 bit unsigned ints
293 * and worry about wraparound (automatic with unsigned arithmetic).
294 */
295
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000296static inline bool before(__u32 seq1, __u32 seq2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297{
Gerrit Renker0d630cc2007-01-04 12:25:16 -0800298 return (__s32)(seq1-seq2) < 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299}
Gerrit Renker9a036b9c2006-12-20 10:25:55 -0800300#define after(seq2, seq1) before(seq1, seq2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301
302/* is s2<=s1<=s3 ? */
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000303static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304{
305 return seq3 - seq2 >= seq1 - seq2;
306}
307
Arun Sharmaefcdbf22012-01-30 14:16:06 -0800308static inline bool tcp_out_of_memory(struct sock *sk)
309{
310 if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
311 sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2))
312 return true;
313 return false;
314}
315
David S. Millerad1af0f2010-08-25 02:27:49 -0700316static inline bool tcp_too_many_orphans(struct sock *sk, int shift)
Pavel Emelianove4fd5da2007-05-29 13:19:18 -0700317{
David S. Millerad1af0f2010-08-25 02:27:49 -0700318 struct percpu_counter *ocp = sk->sk_prot->orphan_count;
319 int orphans = percpu_counter_read_positive(ocp);
320
321 if (orphans << shift > sysctl_tcp_max_orphans) {
322 orphans = percpu_counter_sum_positive(ocp);
323 if (orphans << shift > sysctl_tcp_max_orphans)
324 return true;
325 }
David S. Millerad1af0f2010-08-25 02:27:49 -0700326 return false;
Pavel Emelianove4fd5da2007-05-29 13:19:18 -0700327}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
Joe Perches5c9f3022013-09-23 11:33:32 -0700329bool tcp_check_oom(struct sock *sk, int shift);
Arun Sharmaefcdbf22012-01-30 14:16:06 -0800330
Florian Westphala0f82f62009-04-19 09:43:48 +0000331/* syncookies: remember time of last synqueue overflow */
332static inline void tcp_synq_overflow(struct sock *sk)
333{
334 tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies;
335}
336
337/* syncookies: no recent synqueue overflow on this listening socket? */
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000338static inline bool tcp_synq_no_recent_overflow(const struct sock *sk)
Florian Westphala0f82f62009-04-19 09:43:48 +0000339{
340 unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp;
Jerry Chu9ad7c042011-06-08 11:08:38 +0000341 return time_after(jiffies, last_overflow + TCP_TIMEOUT_FALLBACK);
Florian Westphala0f82f62009-04-19 09:43:48 +0000342}
343
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344extern struct proto tcp_prot;
345
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -0700346#define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field)
347#define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
348#define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
349#define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val)
Tom Herbertaa2ea052010-04-22 07:00:24 +0000350#define TCP_ADD_STATS(net, field, val) SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351
Joe Perches5c9f3022013-09-23 11:33:32 -0700352void tcp_tasklet_init(void);
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000353
Joe Perches5c9f3022013-09-23 11:33:32 -0700354void tcp_v4_err(struct sk_buff *skb, u32);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355
Joe Perches5c9f3022013-09-23 11:33:32 -0700356void tcp_shutdown(struct sock *sk, int how);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357
Joe Perches5c9f3022013-09-23 11:33:32 -0700358void tcp_v4_early_demux(struct sk_buff *skb);
359int tcp_v4_rcv(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360
Joe Perches5c9f3022013-09-23 11:33:32 -0700361int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
362int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
363 size_t size);
364int tcp_sendpage(struct sock *sk, struct page *page, int offset, size_t size,
365 int flags);
366void tcp_release_cb(struct sock *sk);
367void tcp_wfree(struct sk_buff *skb);
368void tcp_write_timer_handler(struct sock *sk);
369void tcp_delack_timer_handler(struct sock *sk);
370int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg);
371int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
372 const struct tcphdr *th, unsigned int len);
373void tcp_rcv_established(struct sock *sk, struct sk_buff *skb,
374 const struct tcphdr *th, unsigned int len);
375void tcp_rcv_space_adjust(struct sock *sk);
376void tcp_cleanup_rbuf(struct sock *sk, int copied);
377int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp);
378void tcp_twsk_destructor(struct sock *sk);
379ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
380 struct pipe_inode_info *pipe, size_t len,
381 unsigned int flags);
Jens Axboe9c55e012007-11-06 23:30:13 -0800382
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700383static inline void tcp_dec_quickack_mode(struct sock *sk,
384 const unsigned int pkts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700386 struct inet_connection_sock *icsk = inet_csk(sk);
David S. Millerfc6415bc2005-07-05 15:17:45 -0700387
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700388 if (icsk->icsk_ack.quick) {
389 if (pkts >= icsk->icsk_ack.quick) {
390 icsk->icsk_ack.quick = 0;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700391 /* Leaving quickack mode we deflate ATO. */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700392 icsk->icsk_ack.ato = TCP_ATO_MIN;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700393 } else
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700394 icsk->icsk_ack.quick -= pkts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 }
396}
397
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -0700398#define TCP_ECN_OK 1
399#define TCP_ECN_QUEUE_CWR 2
400#define TCP_ECN_DEMAND_CWR 4
Eric Dumazet7a269ff2011-09-22 20:02:19 +0000401#define TCP_ECN_SEEN 8
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -0700402
Eric Dumazetfd2c3ef2009-11-03 03:26:03 +0000403enum tcp_tw_status {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 TCP_TW_SUCCESS = 0,
405 TCP_TW_RST = 1,
406 TCP_TW_ACK = 2,
407 TCP_TW_SYN = 3
408};
409
410
Joe Perches5c9f3022013-09-23 11:33:32 -0700411enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
412 struct sk_buff *skb,
413 const struct tcphdr *th);
414struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb,
415 struct request_sock *req, struct request_sock **prev,
416 bool fastopen);
417int tcp_child_process(struct sock *parent, struct sock *child,
418 struct sk_buff *skb);
419void tcp_enter_loss(struct sock *sk, int how);
420void tcp_clear_retrans(struct tcp_sock *tp);
421void tcp_update_metrics(struct sock *sk);
422void tcp_init_metrics(struct sock *sk);
423void tcp_metrics_init(void);
424bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst,
425 bool paws_check);
426bool tcp_remember_stamp(struct sock *sk);
427bool tcp_tw_remember_stamp(struct inet_timewait_sock *tw);
428void tcp_fetch_timewait_stamp(struct sock *sk, struct dst_entry *dst);
429void tcp_disable_fack(struct tcp_sock *tp);
430void tcp_close(struct sock *sk, long timeout);
431void tcp_init_sock(struct sock *sk);
432unsigned int tcp_poll(struct file *file, struct socket *sock,
433 struct poll_table_struct *wait);
434int tcp_getsockopt(struct sock *sk, int level, int optname,
435 char __user *optval, int __user *optlen);
436int tcp_setsockopt(struct sock *sk, int level, int optname,
437 char __user *optval, unsigned int optlen);
438int compat_tcp_getsockopt(struct sock *sk, int level, int optname,
Changli Gao53d31762010-07-10 20:41:06 +0000439 char __user *optval, int __user *optlen);
Joe Perches5c9f3022013-09-23 11:33:32 -0700440int compat_tcp_setsockopt(struct sock *sk, int level, int optname,
Changli Gao53d31762010-07-10 20:41:06 +0000441 char __user *optval, unsigned int optlen);
Joe Perches5c9f3022013-09-23 11:33:32 -0700442void tcp_set_keepalive(struct sock *sk, int val);
443void tcp_syn_ack_timeout(struct sock *sk, struct request_sock *req);
444int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
445 size_t len, int nonblock, int flags, int *addr_len);
446void tcp_parse_options(const struct sk_buff *skb,
447 struct tcp_options_received *opt_rx,
448 int estab, struct tcp_fastopen_cookie *foc);
449const u8 *tcp_parse_md5sig_option(const struct tcphdr *th);
YOSHIFUJI Hideaki7d5d5522008-04-17 12:29:53 +0900450
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451/*
452 * TCP v4 functions exported for the inet6 API
453 */
454
Joe Perches5c9f3022013-09-23 11:33:32 -0700455void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb);
456int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
457struct sock *tcp_create_openreq_child(struct sock *sk,
458 struct request_sock *req,
459 struct sk_buff *skb);
460struct sock *tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
461 struct request_sock *req,
462 struct dst_entry *dst);
463int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
464int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
465int tcp_connect(struct sock *sk);
466struct sk_buff *tcp_make_synack(struct sock *sk, struct dst_entry *dst,
467 struct request_sock *req,
468 struct tcp_fastopen_cookie *foc);
469int tcp_disconnect(struct sock *sk, int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
Pavel Emelyanov370816a2012-04-19 03:40:01 +0000471void tcp_finish_connect(struct sock *sk, struct sk_buff *skb);
Pavel Emelyanov292e8d82012-05-10 01:49:41 +0000472int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size);
Eric Dumazet63d02d12012-08-09 14:11:00 +0000473void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475/* From syncookies.c */
Joe Perches5c9f3022013-09-23 11:33:32 -0700476int __cookie_v4_check(const struct iphdr *iph, const struct tcphdr *th,
477 u32 cookie);
478struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
479 struct ip_options *opt);
Eric Dumazete05c82d2011-09-18 21:02:55 -0400480#ifdef CONFIG_SYN_COOKIES
Florian Westphal8c27bd72013-09-20 22:32:55 +0200481#include <linux/ktime.h>
482
483/* Syncookies use a monotonic timer which increments every 64 seconds.
484 * This counter is used both as a hash input and partially encoded into
485 * the cookie value. A cookie is only validated further if the delta
486 * between the current counter value and the encoded one is less than this,
487 * i.e. a sent cookie is valid only at most for 128 seconds (or less if
488 * the counter advances immediately after a cookie is generated).
489 */
490#define MAX_SYNCOOKIE_AGE 2
491
492static inline u32 tcp_cookie_time(void)
493{
494 struct timespec now;
495 getnstimeofday(&now);
496 return now.tv_sec >> 6; /* 64 seconds granularity */
497}
498
Joe Perches5c9f3022013-09-23 11:33:32 -0700499u32 __cookie_v4_init_sequence(const struct iphdr *iph, const struct tcphdr *th,
500 u16 *mssp);
501__u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb, __u16 *mss);
Eric Dumazete05c82d2011-09-18 21:02:55 -0400502#else
503static inline __u32 cookie_v4_init_sequence(struct sock *sk,
504 struct sk_buff *skb,
505 __u16 *mss)
506{
507 return 0;
508}
509#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510
Joe Perches5c9f3022013-09-23 11:33:32 -0700511__u32 cookie_init_timestamp(struct request_sock *req);
512bool cookie_check_timestamp(struct tcp_options_received *opt, struct net *net,
513 bool *ecn_ok);
Florian Westphal4dfc2812008-04-10 03:12:40 -0700514
Glenn Griffinc6aefaf2008-02-07 21:49:26 -0800515/* From net/ipv6/syncookies.c */
Joe Perches5c9f3022013-09-23 11:33:32 -0700516int __cookie_v6_check(const struct ipv6hdr *iph, const struct tcphdr *th,
517 u32 cookie);
518struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
Eric Dumazete05c82d2011-09-18 21:02:55 -0400519#ifdef CONFIG_SYN_COOKIES
Joe Perches5c9f3022013-09-23 11:33:32 -0700520u32 __cookie_v6_init_sequence(const struct ipv6hdr *iph,
521 const struct tcphdr *th, u16 *mssp);
522__u32 cookie_v6_init_sequence(struct sock *sk, const struct sk_buff *skb,
523 __u16 *mss);
Eric Dumazete05c82d2011-09-18 21:02:55 -0400524#else
525static inline __u32 cookie_v6_init_sequence(struct sock *sk,
526 struct sk_buff *skb,
527 __u16 *mss)
528{
529 return 0;
530}
531#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532/* tcp_output.c */
533
Joe Perches5c9f3022013-09-23 11:33:32 -0700534void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
535 int nonagle);
536bool tcp_may_send_now(struct sock *sk);
537int __tcp_retransmit_skb(struct sock *, struct sk_buff *);
538int tcp_retransmit_skb(struct sock *, struct sk_buff *);
539void tcp_retransmit_timer(struct sock *sk);
540void tcp_xmit_retransmit_queue(struct sock *);
541void tcp_simple_retransmit(struct sock *);
542int tcp_trim_head(struct sock *, struct sk_buff *, u32);
543int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Joe Perches5c9f3022013-09-23 11:33:32 -0700545void tcp_send_probe0(struct sock *);
546void tcp_send_partial(struct sock *);
547int tcp_write_wakeup(struct sock *);
548void tcp_send_fin(struct sock *sk);
549void tcp_send_active_reset(struct sock *sk, gfp_t priority);
550int tcp_send_synack(struct sock *);
551bool tcp_syn_flood_action(struct sock *sk, const struct sk_buff *skb,
552 const char *proto);
553void tcp_push_one(struct sock *, unsigned int mss_now);
554void tcp_send_ack(struct sock *sk);
555void tcp_send_delayed_ack(struct sock *sk);
556void tcp_send_loss_probe(struct sock *sk);
557bool tcp_schedule_loss_probe(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558
David S. Millera762a982005-07-05 15:18:51 -0700559/* tcp_input.c */
Joe Perches5c9f3022013-09-23 11:33:32 -0700560void tcp_cwnd_application_limited(struct sock *sk);
561void tcp_resume_early_retransmit(struct sock *sk);
562void tcp_rearm_rto(struct sock *sk);
563void tcp_reset(struct sock *sk);
David S. Millera762a982005-07-05 15:18:51 -0700564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565/* tcp_timer.c */
Joe Perches5c9f3022013-09-23 11:33:32 -0700566void tcp_init_xmit_timers(struct sock *);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700567static inline void tcp_clear_xmit_timers(struct sock *sk)
568{
569 inet_csk_clear_xmit_timers(sk);
570}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
Joe Perches5c9f3022013-09-23 11:33:32 -0700572unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
573unsigned int tcp_current_mss(struct sock *sk);
Ilpo Järvinen0c54b852009-03-14 14:23:05 +0000574
575/* Bound MSS / TSO packet size with the half of the window */
576static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
577{
Alexey Kuznetsov01f83d62010-09-15 10:27:52 -0700578 int cutoff;
579
580 /* When peer uses tiny windows, there is no use in packetizing
581 * to sub-MSS pieces for the sake of SWS or making sure there
582 * are enough packets in the pipe for fast recovery.
583 *
584 * On the other hand, for extremely large MSS devices, handling
585 * smaller than MSS windows in this way does make sense.
586 */
587 if (tp->max_window >= 512)
588 cutoff = (tp->max_window >> 1);
589 else
590 cutoff = tp->max_window;
591
592 if (cutoff && pktsize > cutoff)
593 return max_t(int, cutoff, 68U - tp->tcp_header_len);
Ilpo Järvinen0c54b852009-03-14 14:23:05 +0000594 else
595 return pktsize;
596}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597
Arnaldo Carvalho de Melo17b085e2005-08-12 12:59:17 -0300598/* tcp.c */
Joe Perches5c9f3022013-09-23 11:33:32 -0700599void tcp_get_info(const struct sock *, struct tcp_info *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601/* Read 'sendfile()'-style from a TCP socket */
602typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
603 unsigned int, size_t);
Joe Perches5c9f3022013-09-23 11:33:32 -0700604int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
605 sk_read_actor_t recv_actor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
Joe Perches5c9f3022013-09-23 11:33:32 -0700607void tcp_initialize_rcv_mss(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608
Joe Perches5c9f3022013-09-23 11:33:32 -0700609int tcp_mtu_to_mss(struct sock *sk, int pmtu);
610int tcp_mss_to_mtu(struct sock *sk, int mss);
611void tcp_mtup_init(struct sock *sk);
612void tcp_init_buffer_space(struct sock *sk);
John Heffner5d424d52006-03-20 17:53:41 -0800613
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000614static inline void tcp_bound_rto(const struct sock *sk)
615{
616 if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX)
617 inet_csk(sk)->icsk_rto = TCP_RTO_MAX;
618}
619
620static inline u32 __tcp_set_rto(const struct tcp_sock *tp)
621{
622 return (tp->srtt >> 3) + tp->rttvar;
623}
624
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800625static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626{
627 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
628 ntohl(TCP_FLAG_ACK) |
629 snd_wnd);
630}
631
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800632static inline void tcp_fast_path_on(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633{
634 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
635}
636
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700637static inline void tcp_fast_path_check(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700639 struct tcp_sock *tp = tcp_sk(sk);
640
David S. Millerb03efcf2005-07-08 14:57:23 -0700641 if (skb_queue_empty(&tp->out_of_order_queue) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 tp->rcv_wnd &&
643 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
644 !tp->urg_data)
645 tcp_fast_path_on(tp);
646}
647
Satoru SATOH0c266892009-05-04 11:11:01 -0700648/* Compute the actual rto_min value */
649static inline u32 tcp_rto_min(struct sock *sk)
650{
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400651 const struct dst_entry *dst = __sk_dst_get(sk);
Satoru SATOH0c266892009-05-04 11:11:01 -0700652 u32 rto_min = TCP_RTO_MIN;
653
654 if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
655 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
656 return rto_min;
657}
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659/* Compute the actual receive window we are currently advertising.
660 * Rcv_nxt can be after the window if our peer push more data
661 * than the offered window.
662 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800663static inline u32 tcp_receive_window(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664{
665 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
666
667 if (win < 0)
668 win = 0;
669 return (u32) win;
670}
671
672/* Choose a new window, without checks for shrinking, and without
673 * scaling applied to the result. The caller does these things
674 * if necessary. This is a "raw" window selection.
675 */
Joe Perches5c9f3022013-09-23 11:33:32 -0700676u32 __tcp_select_window(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
Pavel Emelyanovee995282012-04-19 03:40:39 +0000678void tcp_send_window_probe(struct sock *sk);
679
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680/* TCP timestamps are only 32-bits, this causes a slight
681 * complication on 64-bit systems since we store a snapshot
Stephen Hemminger31f34262005-11-15 15:17:10 -0800682 * of jiffies in the buffer control blocks below. We decided
683 * to use only the low 32-bits of jiffies and hide the ugly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 * casts with the following macro.
685 */
686#define tcp_time_stamp ((__u32)(jiffies))
687
Changli Gaoa3433f32010-06-12 14:01:43 +0000688#define tcp_flag_byte(th) (((u_int8_t *)th)[13])
689
690#define TCPHDR_FIN 0x01
691#define TCPHDR_SYN 0x02
692#define TCPHDR_RST 0x04
693#define TCPHDR_PSH 0x08
694#define TCPHDR_ACK 0x10
695#define TCPHDR_URG 0x20
696#define TCPHDR_ECE 0x40
697#define TCPHDR_CWR 0x80
698
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800699/* This is what the send packet queuing engine uses to pass
Eric Dumazetf86586f2010-07-15 21:41:00 -0700700 * TCP per-packet control information to the transmission code.
701 * We also store the host-order sequence numbers in here too.
702 * This is 44 bytes if IPV6 is enabled.
703 * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 */
705struct tcp_skb_cb {
706 union {
707 struct inet_skb_parm h4;
Eric Dumazetdfd56b82011-12-10 09:48:31 +0000708#if IS_ENABLED(CONFIG_IPV6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 struct inet6_skb_parm h6;
710#endif
711 } header; /* For incoming frames */
712 __u32 seq; /* Starting sequence number */
713 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
714 __u32 when; /* used to compute rtt's */
Eric Dumazet4de075e2011-09-27 13:25:05 -0400715 __u8 tcp_flags; /* TCP header flags. (tcp[13]) */
Neal Cardwellf4f9f6e2012-04-16 07:08:06 +0000716
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 __u8 sacked; /* State flags for SACK/FACK. */
718#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
719#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
720#define TCPCB_LOST 0x04 /* SKB is lost */
721#define TCPCB_TAGBITS 0x07 /* All tag bits */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
723#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
724
Neal Cardwellf4f9f6e2012-04-16 07:08:06 +0000725 __u8 ip_dsfield; /* IPv4 tos or IPv6 dsfield */
726 /* 1 byte hole */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 __u32 ack_seq; /* Sequence number ACK'd */
728};
729
730#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
731
Eric Dumazetbd14b1b2012-05-04 05:14:02 +0000732/* RFC3168 : 6.1.1 SYN packets must not have ECT/ECN bits set
733 *
734 * If we receive a SYN packet with these bits set, it means a network is
735 * playing bad games with TOS bits. In order to avoid possible false congestion
736 * notifications, we disable TCP ECN negociation.
737 */
738static inline void
Hannes Frederic Sowa5d134f12013-01-05 16:10:48 +0000739TCP_ECN_create_request(struct request_sock *req, const struct sk_buff *skb,
740 struct net *net)
Eric Dumazetbd14b1b2012-05-04 05:14:02 +0000741{
742 const struct tcphdr *th = tcp_hdr(skb);
743
Hannes Frederic Sowa5d134f12013-01-05 16:10:48 +0000744 if (net->ipv4.sysctl_tcp_ecn && th->ece && th->cwr &&
Eric Dumazetbd14b1b2012-05-04 05:14:02 +0000745 INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield))
746 inet_rsk(req)->ecn_ok = 1;
747}
748
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749/* Due to TSO, an SKB can be composed of multiple actual
750 * packets. To keep these tracked properly, we use this.
751 */
752static inline int tcp_skb_pcount(const struct sk_buff *skb)
753{
Herbert Xu79671682006-06-22 02:40:14 -0700754 return skb_shinfo(skb)->gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755}
756
757/* This is valid iff tcp_skb_pcount() > 1. */
758static inline int tcp_skb_mss(const struct sk_buff *skb)
759{
Herbert Xu79671682006-06-22 02:40:14 -0700760 return skb_shinfo(skb)->gso_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761}
762
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700763/* Events passed to congestion control interface */
764enum tcp_ca_event {
765 CA_EVENT_TX_START, /* first transmit when no packets in flight */
766 CA_EVENT_CWND_RESTART, /* congestion window restart */
767 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700768 CA_EVENT_LOSS, /* loss timeout */
769 CA_EVENT_FAST_ACK, /* in sequence ack */
770 CA_EVENT_SLOW_ACK, /* other ack */
771};
772
773/*
774 * Interface for adding new TCP congestion control handlers
775 */
776#define TCP_CA_NAME_MAX 16
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800777#define TCP_CA_MAX 128
778#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
779
Stephen Hemminger164891a2007-04-23 22:26:16 -0700780#define TCP_CONG_NON_RESTRICTED 0x1
781#define TCP_CONG_RTT_STAMP 0x2
782
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700783struct tcp_congestion_ops {
784 struct list_head list;
Stephen Hemminger164891a2007-04-23 22:26:16 -0700785 unsigned long flags;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700786
787 /* initialize private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300788 void (*init)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700789 /* cleanup private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300790 void (*release)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700791
792 /* return slow start threshold (required) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300793 u32 (*ssthresh)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700794 /* lower bound for congestion window (optional) */
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700795 u32 (*min_cwnd)(const struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700796 /* do new cwnd calculation (required) */
Yuchung Cheng9f9843a72013-10-31 11:07:31 -0700797 void (*cong_avoid)(struct sock *sk, u32 ack, u32 acked, u32 in_flight);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700798 /* call before changing ca_state (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300799 void (*set_state)(struct sock *sk, u8 new_state);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700800 /* call when cwnd event occurs (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300801 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700802 /* new value of cwnd after loss (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300803 u32 (*undo_cwnd)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700804 /* hook for packet ack accounting (optional) */
Stephen Hemminger30cfd0b2007-07-25 23:49:34 -0700805 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
Arnaldo Carvalho de Melo73c1f4a2005-08-12 12:51:49 -0300806 /* get info for inet_diag (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300807 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700808
809 char name[TCP_CA_NAME_MAX];
810 struct module *owner;
811};
812
Joe Perches5c9f3022013-09-23 11:33:32 -0700813int tcp_register_congestion_control(struct tcp_congestion_ops *type);
814void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700815
Joe Perches5c9f3022013-09-23 11:33:32 -0700816void tcp_init_congestion_control(struct sock *sk);
817void tcp_cleanup_congestion_control(struct sock *sk);
818int tcp_set_default_congestion_control(const char *name);
819void tcp_get_default_congestion_control(char *name);
820void tcp_get_available_congestion_control(char *buf, size_t len);
821void tcp_get_allowed_congestion_control(char *buf, size_t len);
822int tcp_set_allowed_congestion_control(char *allowed);
823int tcp_set_congestion_control(struct sock *sk, const char *name);
Yuchung Cheng9f9843a72013-10-31 11:07:31 -0700824int tcp_slow_start(struct tcp_sock *tp, u32 acked);
Joe Perches5c9f3022013-09-23 11:33:32 -0700825void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700826
Stephen Hemminger5f8ef482005-06-23 20:37:36 -0700827extern struct tcp_congestion_ops tcp_init_congestion_ops;
Joe Perches5c9f3022013-09-23 11:33:32 -0700828u32 tcp_reno_ssthresh(struct sock *sk);
Yuchung Cheng9f9843a72013-10-31 11:07:31 -0700829void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 acked, u32 in_flight);
Joe Perches5c9f3022013-09-23 11:33:32 -0700830u32 tcp_reno_min_cwnd(const struct sock *sk);
David S. Millera8acfba2005-06-23 23:45:02 -0700831extern struct tcp_congestion_ops tcp_reno;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700832
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300833static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700834{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300835 struct inet_connection_sock *icsk = inet_csk(sk);
836
837 if (icsk->icsk_ca_ops->set_state)
838 icsk->icsk_ca_ops->set_state(sk, ca_state);
839 icsk->icsk_ca_state = ca_state;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700840}
841
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300842static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700843{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300844 const struct inet_connection_sock *icsk = inet_csk(sk);
845
846 if (icsk->icsk_ca_ops->cwnd_event)
847 icsk->icsk_ca_ops->cwnd_event(sk, event);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700848}
849
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300850/* These functions determine how the current flow behaves in respect of SACK
851 * handling. SACK is negotiated with the peer, and therefore it can vary
852 * between different flows.
853 *
854 * tcp_is_sack - SACK enabled
855 * tcp_is_reno - No SACK
856 * tcp_is_fack - FACK enabled, implies SACK enabled
857 */
858static inline int tcp_is_sack(const struct tcp_sock *tp)
859{
860 return tp->rx_opt.sack_ok;
861}
862
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000863static inline bool tcp_is_reno(const struct tcp_sock *tp)
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300864{
865 return !tcp_is_sack(tp);
866}
867
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000868static inline bool tcp_is_fack(const struct tcp_sock *tp)
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300869{
Vijay Subramanianab562222011-12-20 13:23:24 +0000870 return tp->rx_opt.sack_ok & TCP_FACK_ENABLED;
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300871}
872
873static inline void tcp_enable_fack(struct tcp_sock *tp)
874{
Vijay Subramanianab562222011-12-20 13:23:24 +0000875 tp->rx_opt.sack_ok |= TCP_FACK_ENABLED;
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300876}
877
Yuchung Chengeed530b2012-05-02 13:30:03 +0000878/* TCP early-retransmit (ER) is similar to but more conservative than
879 * the thin-dupack feature. Enable ER only if thin-dupack is disabled.
880 */
881static inline void tcp_enable_early_retrans(struct tcp_sock *tp)
882{
883 tp->do_early_retrans = sysctl_tcp_early_retrans &&
Nandita Dukkipati6ba8a3b2013-03-11 10:00:43 +0000884 sysctl_tcp_early_retrans < 4 && !sysctl_tcp_thin_dupack &&
885 sysctl_tcp_reordering == 3;
Yuchung Chengeed530b2012-05-02 13:30:03 +0000886}
887
888static inline void tcp_disable_early_retrans(struct tcp_sock *tp)
889{
890 tp->do_early_retrans = 0;
891}
892
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300893static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
894{
895 return tp->sacked_out + tp->lost_out;
896}
897
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898/* This determines how many packets are "in the network" to the best
899 * of our knowledge. In many cases it is conservative, but where
900 * detailed information is available from the receiver (via SACK
901 * blocks etc.) we can make more aggressive calculations.
902 *
903 * Use this for decisions involving congestion control, use just
904 * tp->packets_out to determine if the send queue is empty or not.
905 *
906 * Read this equation as:
907 *
908 * "Packets sent once on transmission queue" MINUS
909 * "Packets left network, but not honestly ACKed yet" PLUS
910 * "Packets fast retransmitted"
911 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800912static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913{
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300914 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915}
916
Ilpo Järvinen0b6a05c2009-09-15 01:30:10 -0700917#define TCP_INFINITE_SSTHRESH 0x7fffffff
918
919static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp)
920{
921 return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH;
922}
923
Yuchung Cheng684bad12012-09-02 17:38:04 +0000924static inline bool tcp_in_cwnd_reduction(const struct sock *sk)
925{
926 return (TCPF_CA_CWR | TCPF_CA_Recovery) &
927 (1 << inet_csk(sk)->icsk_ca_state);
928}
929
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
Yuchung Cheng684bad12012-09-02 17:38:04 +0000931 * The exception is cwnd reduction phase, when cwnd is decreasing towards
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 * ssthresh.
933 */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300934static inline __u32 tcp_current_ssthresh(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300936 const struct tcp_sock *tp = tcp_sk(sk);
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400937
Yuchung Cheng684bad12012-09-02 17:38:04 +0000938 if (tcp_in_cwnd_reduction(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 return tp->snd_ssthresh;
940 else
941 return max(tp->snd_ssthresh,
942 ((tp->snd_cwnd >> 1) +
943 (tp->snd_cwnd >> 2)));
944}
945
Ilpo Järvinenb9c45952007-07-27 16:36:17 +0300946/* Use define here intentionally to get WARN_ON location shown at the caller */
947#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948
Joe Perches5c9f3022013-09-23 11:33:32 -0700949void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
950__u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951
Neal Cardwell6b5a5c02011-11-21 17:15:14 +0000952/* The maximum number of MSS of available cwnd for which TSO defers
953 * sending if not using sysctl_tcp_tso_win_divisor.
954 */
955static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp)
956{
957 return 3;
958}
959
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960/* Slow start with delack produces 3 packets of burst, so that
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700961 * it is safe "de facto". This will be the default - same as
962 * the default reordering threshold - but if reordering increases,
963 * we must be able to allow cwnd to burst at least this much in order
964 * to not pull it back when holes are filled.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 */
966static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
967{
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700968 return tp->reordering;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969}
970
Ilpo Järvinen90840de2007-12-31 04:48:41 -0800971/* Returns end sequence number of the receiver's advertised window */
972static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
973{
974 return tp->snd_una + tp->snd_wnd;
975}
Joe Perches5c9f3022013-09-23 11:33:32 -0700976bool tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800977
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700978static inline void tcp_check_probe_timer(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979{
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400980 const struct tcp_sock *tp = tcp_sk(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700981 const struct inet_connection_sock *icsk = inet_csk(sk);
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700982
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700983 if (!tp->packets_out && !icsk->icsk_pending)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700984 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
985 icsk->icsk_rto, TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986}
987
Hantzis Fotisee7537b2009-03-02 22:42:02 -0800988static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989{
990 tp->snd_wl1 = seq;
991}
992
Hantzis Fotisee7537b2009-03-02 22:42:02 -0800993static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994{
995 tp->snd_wl1 = seq;
996}
997
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998/*
999 * Calculate(/check) TCP checksum
1000 */
Frederik Deweerdtba7808e2007-02-04 20:15:27 -08001001static inline __sum16 tcp_v4_check(int len, __be32 saddr,
1002 __be32 daddr, __wsum base)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003{
1004 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
1005}
1006
Al Virob51655b2006-11-14 21:40:42 -08001007static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008{
Herbert Xufb286bb2005-11-10 13:01:24 -08001009 return __skb_checksum_complete(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010}
1011
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001012static inline bool tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013{
Herbert Xu60476372007-04-09 11:59:39 -07001014 return !skb_csum_unnecessary(skb) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 __tcp_checksum_complete(skb);
1016}
1017
1018/* Prequeue for VJ style copy to user, combined with checksumming. */
1019
Stephen Hemminger40efc6f2006-01-03 16:03:49 -08001020static inline void tcp_prequeue_init(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021{
1022 tp->ucopy.task = NULL;
1023 tp->ucopy.len = 0;
1024 tp->ucopy.memory = 0;
1025 skb_queue_head_init(&tp->ucopy.prequeue);
Chris Leech97fc2f02006-05-23 17:55:33 -07001026#ifdef CONFIG_NET_DMA
1027 tp->ucopy.dma_chan = NULL;
1028 tp->ucopy.wakeup = 0;
1029 tp->ucopy.pinned_list = NULL;
1030 tp->ucopy.dma_cookie = 0;
1031#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032}
1033
Joe Perches5c9f3022013-09-23 11:33:32 -07001034bool tcp_prequeue(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035
1036#undef STATE_TRACE
1037
1038#ifdef STATE_TRACE
1039static const char *statename[]={
1040 "Unused","Established","Syn Sent","Syn Recv",
1041 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
1042 "Close Wait","Last ACK","Listen","Closing"
1043};
1044#endif
Joe Perches5c9f3022013-09-23 11:33:32 -07001045void tcp_set_state(struct sock *sk, int state);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046
Joe Perches5c9f3022013-09-23 11:33:32 -07001047void tcp_done(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048
Stephen Hemminger40efc6f2006-01-03 16:03:49 -08001049static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050{
1051 rx_opt->dsack = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 rx_opt->num_sacks = 0;
1053}
1054
Joe Perches5c9f3022013-09-23 11:33:32 -07001055u32 tcp_default_init_rwnd(u32 mss);
Yuchung Cheng85f16522013-06-11 15:35:32 -07001056
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057/* Determine a window scaling and initial window to offer. */
Joe Perches5c9f3022013-09-23 11:33:32 -07001058void tcp_select_initial_window(int __space, __u32 mss, __u32 *rcv_wnd,
1059 __u32 *window_clamp, int wscale_ok,
1060 __u8 *rcv_wscale, __u32 init_rcv_wnd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061
1062static inline int tcp_win_from_space(int space)
1063{
1064 return sysctl_tcp_adv_win_scale<=0 ?
1065 (space>>(-sysctl_tcp_adv_win_scale)) :
1066 space - (space>>sysctl_tcp_adv_win_scale);
1067}
1068
1069/* Note: caller must be prepared to deal with negative returns */
1070static inline int tcp_space(const struct sock *sk)
1071{
1072 return tcp_win_from_space(sk->sk_rcvbuf -
1073 atomic_read(&sk->sk_rmem_alloc));
1074}
1075
1076static inline int tcp_full_space(const struct sock *sk)
1077{
1078 return tcp_win_from_space(sk->sk_rcvbuf);
1079}
1080
Stephen Hemminger40efc6f2006-01-03 16:03:49 -08001081static inline void tcp_openreq_init(struct request_sock *req,
1082 struct tcp_options_received *rx_opt,
1083 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001085 struct inet_request_sock *ireq = inet_rsk(req);
1086
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
Florian Westphal4dfc2812008-04-10 03:12:40 -07001088 req->cookie_ts = 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001089 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
Jerry Chu10467162012-08-31 12:29:11 +00001090 tcp_rsk(req)->rcv_nxt = TCP_SKB_CB(skb)->seq + 1;
Neal Cardwell016818d2012-09-22 04:18:55 +00001091 tcp_rsk(req)->snt_synack = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 req->mss = rx_opt->mss_clamp;
1093 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001094 ireq->tstamp_ok = rx_opt->tstamp_ok;
1095 ireq->sack_ok = rx_opt->sack_ok;
1096 ireq->snd_wscale = rx_opt->snd_wscale;
1097 ireq->wscale_ok = rx_opt->wscale_ok;
1098 ireq->acked = 0;
1099 ireq->ecn_ok = 0;
Eric Dumazet634fb9792013-10-09 15:21:29 -07001100 ireq->ir_rmt_port = tcp_hdr(skb)->source;
Eric Dumazetb44084c2013-10-10 00:04:37 -07001101 ireq->ir_num = ntohs(tcp_hdr(skb)->dest);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102}
1103
Joe Perches5c9f3022013-09-23 11:33:32 -07001104void tcp_enter_memory_pressure(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1107{
1108 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1109}
1110
1111static inline int keepalive_time_when(const struct tcp_sock *tp)
1112{
1113 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1114}
1115
Eric Dumazetdf19a622009-08-28 23:48:54 -07001116static inline int keepalive_probes(const struct tcp_sock *tp)
1117{
1118 return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes;
1119}
1120
Flavio Leitner6c37e5d2010-04-26 18:33:27 +00001121static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp)
1122{
1123 const struct inet_connection_sock *icsk = &tp->inet_conn;
1124
1125 return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime,
1126 tcp_time_stamp - tp->rcv_tstamp);
1127}
1128
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001129static inline int tcp_fin_time(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001131 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1132 const int rto = inet_csk(sk)->icsk_rto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001134 if (fin_timeout < (rto << 2) - (rto >> 1))
1135 fin_timeout = (rto << 2) - (rto >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
1137 return fin_timeout;
1138}
1139
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001140static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt,
1141 int paws_win)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142{
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001143 if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001144 return true;
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001145 if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001146 return true;
Eric Dumazetbc2ce892010-12-16 14:08:34 -08001147 /*
1148 * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0,
1149 * then following tcp messages have valid values. Ignore 0 value,
1150 * or else 'negative' tsval might forbid us to accept their packets.
1151 */
1152 if (!rx_opt->ts_recent)
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001153 return true;
1154 return false;
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001155}
1156
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001157static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt,
1158 int rst)
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001159{
1160 if (tcp_paws_check(rx_opt, 0))
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001161 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162
1163 /* RST segments are not recommended to carry timestamp,
1164 and, if they do, it is recommended to ignore PAWS because
1165 "their cleanup function should take precedence over timestamps."
1166 Certainly, it is mistake. It is necessary to understand the reasons
1167 of this constraint to relax it: if peer reboots, clock may go
1168 out-of-sync and half-open connections will not be reset.
1169 Actually, the problem would be not existing if all
1170 the implementations followed draft about maintaining clock
1171 via reboots. Linux-2.2 DOES NOT!
1172
1173 However, we can relax time bounds for RST segments to MSL.
1174 */
James Morris9d729f72007-03-04 16:12:44 -08001175 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001176 return false;
1177 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178}
1179
Pavel Emelyanova9c193292008-07-16 20:21:42 -07001180static inline void tcp_mib_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181{
1182 /* See RFC 2012 */
Pavel Emelyanovcf1100a2008-07-16 20:27:38 -07001183 TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1184 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1185 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1186 TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187}
1188
Ilpo Järvinen5af4ec22007-09-20 11:30:48 -07001189/* from STCP */
Ilpo Järvinenef9da472008-09-20 21:25:15 -07001190static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
David S. Miller0800f172007-09-20 11:40:37 -07001191{
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001192 tp->lost_skb_hint = NULL;
Ilpo Järvinenef9da472008-09-20 21:25:15 -07001193}
1194
1195static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1196{
1197 tcp_clear_retrans_hints_partial(tp);
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001198 tp->retransmit_skb_hint = NULL;
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001199}
1200
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001201/* MD5 Signature */
1202struct crypto_hash;
1203
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001204union tcp_md5_addr {
1205 struct in_addr a4;
1206#if IS_ENABLED(CONFIG_IPV6)
1207 struct in6_addr a6;
1208#endif
1209};
1210
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001211/* - key database */
1212struct tcp_md5sig_key {
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001213 struct hlist_node node;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001214 u8 keylen;
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001215 u8 family; /* AF_INET or AF_INET6 */
1216 union tcp_md5_addr addr;
1217 u8 key[TCP_MD5SIG_MAXKEYLEN];
1218 struct rcu_head rcu;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001219};
1220
1221/* - sock block */
1222struct tcp_md5sig_info {
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001223 struct hlist_head head;
Eric Dumazeta8afca02012-01-31 18:45:40 +00001224 struct rcu_head rcu;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001225};
1226
1227/* - pseudo header */
1228struct tcp4_pseudohdr {
1229 __be32 saddr;
1230 __be32 daddr;
1231 __u8 pad;
1232 __u8 protocol;
1233 __be16 len;
1234};
1235
1236struct tcp6_pseudohdr {
1237 struct in6_addr saddr;
1238 struct in6_addr daddr;
1239 __be32 len;
1240 __be32 protocol; /* including padding */
1241};
1242
1243union tcp_md5sum_block {
1244 struct tcp4_pseudohdr ip4;
Eric Dumazetdfd56b82011-12-10 09:48:31 +00001245#if IS_ENABLED(CONFIG_IPV6)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001246 struct tcp6_pseudohdr ip6;
1247#endif
1248};
1249
1250/* - pool: digest algorithm, hash description and scratch buffer */
1251struct tcp_md5sig_pool {
1252 struct hash_desc md5_desc;
1253 union tcp_md5sum_block md5_blk;
1254};
1255
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001256/* - functions */
Joe Perches5c9f3022013-09-23 11:33:32 -07001257int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
1258 const struct sock *sk, const struct request_sock *req,
1259 const struct sk_buff *skb);
1260int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
1261 int family, const u8 *newkey, u8 newkeylen, gfp_t gfp);
1262int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr,
1263 int family);
1264struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001265 struct sock *addr_sk);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001266
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +09001267#ifdef CONFIG_TCP_MD5SIG
Joe Perches5c9f3022013-09-23 11:33:32 -07001268struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1269 const union tcp_md5_addr *addr,
1270 int family);
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001271#define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_key)
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +09001272#else
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001273static inline struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1274 const union tcp_md5_addr *addr,
1275 int family)
1276{
1277 return NULL;
1278}
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +09001279#define tcp_twsk_md5_key(twsk) NULL
1280#endif
1281
Joe Perches5c9f3022013-09-23 11:33:32 -07001282bool tcp_alloc_md5sig_pool(void);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001283
Joe Perches5c9f3022013-09-23 11:33:32 -07001284struct tcp_md5sig_pool *tcp_get_md5sig_pool(void);
Eric Dumazet71cea172013-05-20 06:52:26 +00001285static inline void tcp_put_md5sig_pool(void)
1286{
1287 local_bh_enable();
1288}
Eric Dumazet35790c02010-05-16 00:34:04 -07001289
Joe Perches5c9f3022013-09-23 11:33:32 -07001290int tcp_md5_hash_header(struct tcp_md5sig_pool *, const struct tcphdr *);
1291int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *,
1292 unsigned int header_len);
1293int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
1294 const struct tcp_md5sig_key *key);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001295
Jerry Chu10467162012-08-31 12:29:11 +00001296/* From tcp_fastopen.c */
Joe Perches5c9f3022013-09-23 11:33:32 -07001297void tcp_fastopen_cache_get(struct sock *sk, u16 *mss,
1298 struct tcp_fastopen_cookie *cookie, int *syn_loss,
1299 unsigned long *last_syn_loss);
1300void tcp_fastopen_cache_set(struct sock *sk, u16 mss,
1301 struct tcp_fastopen_cookie *cookie, bool syn_lost);
Yuchung Cheng783237e2012-07-19 06:43:07 +00001302struct tcp_fastopen_request {
1303 /* Fast Open cookie. Size 0 means a cookie request */
1304 struct tcp_fastopen_cookie cookie;
1305 struct msghdr *data; /* data in MSG_FASTOPEN */
1306 u16 copied; /* queued in tcp_connect() */
1307};
Yuchung Cheng783237e2012-07-19 06:43:07 +00001308void tcp_free_fastopen_req(struct tcp_sock *tp);
1309
Jerry Chu10467162012-08-31 12:29:11 +00001310extern struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
1311int tcp_fastopen_reset_cipher(void *key, unsigned int len);
Joe Perches5c9f3022013-09-23 11:33:32 -07001312void tcp_fastopen_cookie_gen(__be32 src, __be32 dst,
1313 struct tcp_fastopen_cookie *foc);
Hannes Frederic Sowa222e83d2013-10-19 21:48:58 +02001314void tcp_fastopen_init_key_once(bool publish);
Jerry Chu10467162012-08-31 12:29:11 +00001315#define TCP_FASTOPEN_KEY_LENGTH 16
1316
1317/* Fastopen key context */
1318struct tcp_fastopen_context {
Eric Dumazet7ae86392013-06-25 01:21:06 -07001319 struct crypto_cipher *tfm;
1320 __u8 key[TCP_FASTOPEN_KEY_LENGTH];
1321 struct rcu_head rcu;
Jerry Chu10467162012-08-31 12:29:11 +00001322};
1323
David S. Millerfe067e82007-03-07 12:12:44 -08001324/* write queue abstraction */
1325static inline void tcp_write_queue_purge(struct sock *sk)
1326{
1327 struct sk_buff *skb;
1328
1329 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001330 sk_wmem_free_skb(sk, skb);
1331 sk_mem_reclaim(sk);
Ilpo Järvinen8818a9d2009-12-02 22:24:02 -08001332 tcp_clear_all_retrans_hints(tcp_sk(sk));
David S. Millerfe067e82007-03-07 12:12:44 -08001333}
1334
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001335static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk)
David S. Millerfe067e82007-03-07 12:12:44 -08001336{
David S. Millercd07a8e2008-09-23 00:50:13 -07001337 return skb_peek(&sk->sk_write_queue);
David S. Millerfe067e82007-03-07 12:12:44 -08001338}
1339
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001340static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk)
David S. Millerfe067e82007-03-07 12:12:44 -08001341{
David S. Millercd07a8e2008-09-23 00:50:13 -07001342 return skb_peek_tail(&sk->sk_write_queue);
David S. Millerfe067e82007-03-07 12:12:44 -08001343}
1344
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001345static inline struct sk_buff *tcp_write_queue_next(const struct sock *sk,
1346 const struct sk_buff *skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001347{
David S. Millercd07a8e2008-09-23 00:50:13 -07001348 return skb_queue_next(&sk->sk_write_queue, skb);
David S. Millerfe067e82007-03-07 12:12:44 -08001349}
1350
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001351static inline struct sk_buff *tcp_write_queue_prev(const struct sock *sk,
1352 const struct sk_buff *skb)
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001353{
1354 return skb_queue_prev(&sk->sk_write_queue, skb);
1355}
1356
David S. Millerfe067e82007-03-07 12:12:44 -08001357#define tcp_for_write_queue(skb, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001358 skb_queue_walk(&(sk)->sk_write_queue, skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001359
1360#define tcp_for_write_queue_from(skb, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001361 skb_queue_walk_from(&(sk)->sk_write_queue, skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001362
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001363#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001364 skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001365
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001366static inline struct sk_buff *tcp_send_head(const struct sock *sk)
David S. Millerfe067e82007-03-07 12:12:44 -08001367{
1368 return sk->sk_send_head;
1369}
1370
David S. Millercd07a8e2008-09-23 00:50:13 -07001371static inline bool tcp_skb_is_last(const struct sock *sk,
1372 const struct sk_buff *skb)
1373{
1374 return skb_queue_is_last(&sk->sk_write_queue, skb);
1375}
1376
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001377static inline void tcp_advance_send_head(struct sock *sk, const struct sk_buff *skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001378{
David S. Millercd07a8e2008-09-23 00:50:13 -07001379 if (tcp_skb_is_last(sk, skb))
David S. Millerfe067e82007-03-07 12:12:44 -08001380 sk->sk_send_head = NULL;
David S. Millercd07a8e2008-09-23 00:50:13 -07001381 else
1382 sk->sk_send_head = tcp_write_queue_next(sk, skb);
David S. Millerfe067e82007-03-07 12:12:44 -08001383}
1384
1385static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1386{
1387 if (sk->sk_send_head == skb_unlinked)
1388 sk->sk_send_head = NULL;
1389}
1390
1391static inline void tcp_init_send_head(struct sock *sk)
1392{
1393 sk->sk_send_head = NULL;
1394}
1395
1396static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1397{
1398 __skb_queue_tail(&sk->sk_write_queue, skb);
1399}
1400
1401static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1402{
1403 __tcp_add_write_queue_tail(sk, skb);
1404
1405 /* Queue it, remembering where we must start sending. */
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001406 if (sk->sk_send_head == NULL) {
David S. Millerfe067e82007-03-07 12:12:44 -08001407 sk->sk_send_head = skb;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001408
1409 if (tcp_sk(sk)->highest_sack == NULL)
1410 tcp_sk(sk)->highest_sack = skb;
1411 }
David S. Millerfe067e82007-03-07 12:12:44 -08001412}
1413
1414static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1415{
1416 __skb_queue_head(&sk->sk_write_queue, skb);
1417}
1418
1419/* Insert buff after skb on the write queue of sk. */
1420static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1421 struct sk_buff *buff,
1422 struct sock *sk)
1423{
Gerrit Renker7de6c032008-04-14 00:05:09 -07001424 __skb_queue_after(&sk->sk_write_queue, skb, buff);
David S. Millerfe067e82007-03-07 12:12:44 -08001425}
1426
David S. Miller43f59c82008-09-21 21:28:51 -07001427/* Insert new before skb on the write queue of sk. */
David S. Millerfe067e82007-03-07 12:12:44 -08001428static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1429 struct sk_buff *skb,
1430 struct sock *sk)
1431{
David S. Miller43f59c82008-09-21 21:28:51 -07001432 __skb_queue_before(&sk->sk_write_queue, skb, new);
Ilpo Järvinen6e421412007-11-19 23:24:09 -08001433
1434 if (sk->sk_send_head == skb)
1435 sk->sk_send_head = new;
David S. Millerfe067e82007-03-07 12:12:44 -08001436}
1437
1438static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1439{
1440 __skb_unlink(skb, &sk->sk_write_queue);
1441}
1442
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001443static inline bool tcp_write_queue_empty(struct sock *sk)
David S. Millerfe067e82007-03-07 12:12:44 -08001444{
1445 return skb_queue_empty(&sk->sk_write_queue);
1446}
1447
Krishna Kumar12d50c42009-12-08 22:26:13 +00001448static inline void tcp_push_pending_frames(struct sock *sk)
1449{
1450 if (tcp_send_head(sk)) {
1451 struct tcp_sock *tp = tcp_sk(sk);
1452
1453 __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle);
1454 }
1455}
1456
Neal Cardwellecb97192012-02-27 17:52:52 -05001457/* Start sequence of the skb just after the highest skb with SACKed
1458 * bit, valid only if sacked_out > 0 or when the caller has ensured
1459 * validity by itself.
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001460 */
1461static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1462{
1463 if (!tp->sacked_out)
1464 return tp->snd_una;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001465
1466 if (tp->highest_sack == NULL)
1467 return tp->snd_nxt;
1468
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001469 return TCP_SKB_CB(tp->highest_sack)->seq;
1470}
1471
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001472static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1473{
1474 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1475 tcp_write_queue_next(sk, skb);
1476}
1477
1478static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1479{
1480 return tcp_sk(sk)->highest_sack;
1481}
1482
1483static inline void tcp_highest_sack_reset(struct sock *sk)
1484{
1485 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1486}
1487
1488/* Called when old skb is about to be deleted (to be combined with new skb) */
1489static inline void tcp_highest_sack_combine(struct sock *sk,
1490 struct sk_buff *old,
1491 struct sk_buff *new)
1492{
1493 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1494 tcp_sk(sk)->highest_sack = new;
1495}
1496
Andreas Petlund5aa4b322010-02-18 02:45:45 +00001497/* Determines whether this is a thin stream (which may suffer from
1498 * increased latency). Used to trigger latency-reducing mechanisms.
1499 */
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001500static inline bool tcp_stream_is_thin(struct tcp_sock *tp)
Andreas Petlund5aa4b322010-02-18 02:45:45 +00001501{
1502 return tp->packets_out < 4 && !tcp_in_initial_slowstart(tp);
1503}
1504
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505/* /proc */
1506enum tcp_seq_states {
1507 TCP_SEQ_STATE_LISTENING,
1508 TCP_SEQ_STATE_OPENREQ,
1509 TCP_SEQ_STATE_ESTABLISHED,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510};
1511
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00001512int tcp_seq_open(struct inode *inode, struct file *file);
1513
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514struct tcp_seq_afinfo {
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00001515 char *name;
1516 sa_family_t family;
1517 const struct file_operations *seq_fops;
1518 struct seq_operations seq_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519};
1520
1521struct tcp_iter_state {
Denis V. Luneva4146b12008-04-13 22:11:14 -07001522 struct seq_net_private p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 sa_family_t family;
1524 enum tcp_seq_states state;
1525 struct sock *syn_wait_sk;
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06001526 int bucket, offset, sbucket, num;
1527 kuid_t uid;
Tom Herberta8b690f2010-06-07 00:43:42 -07001528 loff_t last_pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529};
1530
Joe Perches5c9f3022013-09-23 11:33:32 -07001531int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1532void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001534extern struct request_sock_ops tcp_request_sock_ops;
Glenn Griffinc6aefaf2008-02-07 21:49:26 -08001535extern struct request_sock_ops tcp6_request_sock_ops;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001536
Joe Perches5c9f3022013-09-23 11:33:32 -07001537void tcp_v4_destroy_sock(struct sock *sk);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001538
Eric Dumazet28be6e02013-10-18 10:36:17 -07001539struct sk_buff *tcp_gso_segment(struct sk_buff *skb,
Joe Perches5c9f3022013-09-23 11:33:32 -07001540 netdev_features_t features);
1541struct sk_buff **tcp_gro_receive(struct sk_buff **head, struct sk_buff *skb);
1542int tcp_gro_complete(struct sk_buff *skb);
Daniel Borkmann28850dc2013-06-07 05:11:46 +00001543
Joe Perches5c9f3022013-09-23 11:33:32 -07001544void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001545
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001546static inline u32 tcp_notsent_lowat(const struct tcp_sock *tp)
1547{
1548 return tp->notsent_lowat ?: sysctl_tcp_notsent_lowat;
1549}
1550
1551static inline bool tcp_stream_memory_free(const struct sock *sk)
1552{
1553 const struct tcp_sock *tp = tcp_sk(sk);
1554 u32 notsent_bytes = tp->write_seq - tp->snd_nxt;
1555
1556 return notsent_bytes < tcp_notsent_lowat(tp);
1557}
1558
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001559#ifdef CONFIG_PROC_FS
Joe Perches5c9f3022013-09-23 11:33:32 -07001560int tcp4_proc_init(void);
1561void tcp4_proc_exit(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001562#endif
1563
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001564/* TCP af-specific functions */
1565struct tcp_sock_af_ops {
1566#ifdef CONFIG_TCP_MD5SIG
1567 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1568 struct sock *addr_sk);
1569 int (*calc_md5_hash) (char *location,
1570 struct tcp_md5sig_key *md5,
Eric Dumazet318cf7a2011-10-24 02:46:04 -04001571 const struct sock *sk,
1572 const struct request_sock *req,
1573 const struct sk_buff *skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001574 int (*md5_parse) (struct sock *sk,
1575 char __user *optval,
1576 int optlen);
1577#endif
1578};
1579
1580struct tcp_request_sock_ops {
1581#ifdef CONFIG_TCP_MD5SIG
1582 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1583 struct request_sock *req);
John Dykstrae3afe7b2009-07-16 05:04:51 +00001584 int (*calc_md5_hash) (char *location,
1585 struct tcp_md5sig_key *md5,
Eric Dumazet318cf7a2011-10-24 02:46:04 -04001586 const struct sock *sk,
1587 const struct request_sock *req,
1588 const struct sk_buff *skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001589#endif
1590};
1591
Joe Perches5c9f3022013-09-23 11:33:32 -07001592int tcpv4_offload_init(void);
Daniel Borkmann28850dc2013-06-07 05:11:46 +00001593
Joe Perches5c9f3022013-09-23 11:33:32 -07001594void tcp_v4_init(void);
1595void tcp_init(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001596
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597#endif /* _TCP_H */