blob: cf0694d4ad60f62be703eaec51db142b9c5c4e29 [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;
Linus Torvalds1da177e2005-04-16 15:20:36 -070053extern void 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
Nandita Dukkipati356f0392010-12-20 14:15:56 +000064/* Offer an initial receive window of 10 mss. */
65#define TCP_DEFAULT_INIT_RCVWND 10
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
68#define TCP_MIN_MSS 88U
69
John Heffner5d424d52006-03-20 17:53:41 -080070/* The least MTU to use for probing */
71#define TCP_BASE_MSS 512
72
Linus Torvalds1da177e2005-04-16 15:20:36 -070073/* After receiving this amount of duplicate ACKs fast retransmit starts. */
74#define TCP_FASTRETRANS_THRESH 3
75
76/* Maximal reordering. */
77#define TCP_MAX_REORDERING 127
78
79/* Maximal number of ACKs sent quickly to accelerate slow-start. */
80#define TCP_MAX_QUICKACKS 16U
81
82/* urg_data states */
83#define TCP_URG_VALID 0x0100
84#define TCP_URG_NOTYET 0x0200
85#define TCP_URG_READ 0x0400
86
87#define TCP_RETR1 3 /*
88 * This is how many retries it does before it
89 * tries to figure out if the gateway is
90 * down. Minimal RFC value is 3; it corresponds
91 * to ~3sec-8min depending on RTO.
92 */
93
94#define TCP_RETR2 15 /*
95 * This should take at least
96 * 90 minutes to time out.
97 * RFC1122 says that the limit is 100 sec.
98 * 15 is ~13-30min depending on RTO.
99 */
100
Alex Bergmann6c9ff972012-08-31 02:48:31 +0000101#define TCP_SYN_RETRIES 6 /* This is how many retries are done
102 * when active opening a connection.
103 * RFC1122 says the minimum retry MUST
104 * be at least 180secs. Nevertheless
105 * this value is corresponding to
106 * 63secs of retransmission with the
107 * current initial RTO.
108 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109
Alex Bergmann6c9ff972012-08-31 02:48:31 +0000110#define TCP_SYNACK_RETRIES 5 /* This is how may retries are done
111 * when passive opening a connection.
112 * This is corresponding to 31secs of
113 * retransmission with the current
114 * initial RTO.
115 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
118 * state, about 60 seconds */
119#define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
120 /* BSD style FIN_WAIT2 deadlock breaker.
121 * It used to be 3min, new value is 60sec,
122 * to combine FIN-WAIT-2 timeout with
123 * TIME-WAIT timer.
124 */
125
126#define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
127#if HZ >= 100
128#define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
129#define TCP_ATO_MIN ((unsigned)(HZ/25))
130#else
131#define TCP_DELACK_MIN 4U
132#define TCP_ATO_MIN 4U
133#endif
134#define TCP_RTO_MAX ((unsigned)(120*HZ))
135#define TCP_RTO_MIN ((unsigned)(HZ/5))
Eric Dumazetfd4f2ce2012-04-12 19:48:40 +0000136#define TCP_TIMEOUT_INIT ((unsigned)(1*HZ)) /* RFC6298 2.1 initial RTO value */
Jerry Chu9ad7c042011-06-08 11:08:38 +0000137#define TCP_TIMEOUT_FALLBACK ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value, now
138 * used as a fallback RTO for the
139 * initial data transmission if no
140 * valid RTT sample has been acquired,
141 * most likely due to retrans in 3WHS.
142 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
144#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
145 * for local resources.
146 */
147
148#define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
149#define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
150#define TCP_KEEPALIVE_INTVL (75*HZ)
151
152#define MAX_TCP_KEEPIDLE 32767
153#define MAX_TCP_KEEPINTVL 32767
154#define MAX_TCP_KEEPCNT 127
155#define MAX_TCP_SYNCNT 127
156
157#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158
159#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
160#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
161 * after this time. It should be equal
162 * (or greater than) TCP_TIMEWAIT_LEN
163 * to provide reliability equal to one
164 * provided by timewait state.
165 */
166#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
167 * timestamps. It must be less than
168 * minimal timewait lifetime.
169 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170/*
171 * TCP option
172 */
173
174#define TCPOPT_NOP 1 /* Padding */
175#define TCPOPT_EOL 0 /* End of options */
176#define TCPOPT_MSS 2 /* Segment size negotiating */
177#define TCPOPT_WINDOW 3 /* Window scaling */
178#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
179#define TCPOPT_SACK 5 /* SACK Block */
180#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800181#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
William Allen Simpson435cf552009-12-02 18:17:05 +0000182#define TCPOPT_COOKIE 253 /* Cookie extension (experimental) */
Yuchung Cheng2100c8d2012-07-19 06:43:05 +0000183#define TCPOPT_EXP 254 /* Experimental */
184/* Magic number to be after the option value for sharing TCP
185 * experimental options. See draft-ietf-tcpm-experimental-options-00.txt
186 */
187#define TCPOPT_FASTOPEN_MAGIC 0xF989
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188
189/*
190 * TCP option lengths
191 */
192
193#define TCPOLEN_MSS 4
194#define TCPOLEN_WINDOW 3
195#define TCPOLEN_SACK_PERM 2
196#define TCPOLEN_TIMESTAMP 10
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800197#define TCPOLEN_MD5SIG 18
Yuchung Cheng2100c8d2012-07-19 06:43:05 +0000198#define TCPOLEN_EXP_FASTOPEN_BASE 4
William Allen Simpson435cf552009-12-02 18:17:05 +0000199#define TCPOLEN_COOKIE_BASE 2 /* Cookie-less header extension */
200#define TCPOLEN_COOKIE_PAIR 3 /* Cookie pair header extension */
201#define TCPOLEN_COOKIE_MIN (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MIN)
202#define TCPOLEN_COOKIE_MAX (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203
204/* But this is what stacks really send out. */
205#define TCPOLEN_TSTAMP_ALIGNED 12
206#define TCPOLEN_WSCALE_ALIGNED 4
207#define TCPOLEN_SACKPERM_ALIGNED 4
208#define TCPOLEN_SACK_BASE 2
209#define TCPOLEN_SACK_BASE_ALIGNED 4
210#define TCPOLEN_SACK_PERBLOCK 8
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -0800211#define TCPOLEN_MD5SIG_ALIGNED 20
Adam Langley33ad7982008-07-19 00:04:31 -0700212#define TCPOLEN_MSS_ALIGNED 4
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214/* Flags in tp->nonagle */
215#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
216#define TCP_NAGLE_CORK 2 /* Socket is corked */
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800217#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218
Andreas Petlund36e31b0a2010-02-18 02:47:01 +0000219/* TCP thin-stream limits */
220#define TCP_THIN_LINEAR_RETRIES 6 /* After 6 linear retries, do exp. backoff */
221
David S. Miller7eb38522011-02-05 18:13:45 -0800222/* TCP initial congestion window as per draft-hkchu-tcpm-initcwnd-01 */
David S. Miller442b9632011-02-02 17:05:11 -0800223#define TCP_INIT_CWND 10
224
Yuchung Chengcf60af02012-07-19 06:43:09 +0000225/* Bit Flags for sysctl_tcp_fastopen */
226#define TFO_CLIENT_ENABLE 1
Jerry Chu10467162012-08-31 12:29:11 +0000227#define TFO_SERVER_ENABLE 2
Yuchung Cheng67da22d2012-07-19 06:43:11 +0000228#define TFO_CLIENT_NO_COOKIE 4 /* Data in SYN w/o cookie option */
Yuchung Chengcf60af02012-07-19 06:43:09 +0000229
Jerry Chu10467162012-08-31 12:29:11 +0000230/* Process SYN data but skip cookie validation */
231#define TFO_SERVER_COOKIE_NOT_CHKED 0x100
232/* Accept SYN data w/o any cookie option */
233#define TFO_SERVER_COOKIE_NOT_REQD 0x200
234
235/* Force enable TFO on all listeners, i.e., not requiring the
236 * TCP_FASTOPEN socket option. SOCKOPT1/2 determine how to set max_qlen.
237 */
238#define TFO_SERVER_WO_SOCKOPT1 0x400
239#define TFO_SERVER_WO_SOCKOPT2 0x800
240/* Always create TFO child sockets on a TFO listener even when
241 * cookie/data not present. (For testing purpose!)
242 */
243#define TFO_SERVER_ALWAYS 0x1000
244
Arnaldo Carvalho de Melo295ff7e2005-08-09 20:44:40 -0700245extern struct inet_timewait_death_row tcp_death_row;
246
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247/* sysctl variables for tcp */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248extern int sysctl_tcp_timestamps;
249extern int sysctl_tcp_window_scaling;
250extern int sysctl_tcp_sack;
251extern int sysctl_tcp_fin_timeout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252extern int sysctl_tcp_keepalive_time;
253extern int sysctl_tcp_keepalive_probes;
254extern int sysctl_tcp_keepalive_intvl;
255extern int sysctl_tcp_syn_retries;
256extern int sysctl_tcp_synack_retries;
257extern int sysctl_tcp_retries1;
258extern int sysctl_tcp_retries2;
259extern int sysctl_tcp_orphan_retries;
260extern int sysctl_tcp_syncookies;
Yuchung Cheng2100c8d2012-07-19 06:43:05 +0000261extern int sysctl_tcp_fastopen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262extern int sysctl_tcp_retrans_collapse;
263extern int sysctl_tcp_stdurg;
264extern int sysctl_tcp_rfc1337;
265extern int sysctl_tcp_abort_on_overflow;
266extern int sysctl_tcp_max_orphans;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267extern int sysctl_tcp_fack;
268extern int sysctl_tcp_reordering;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269extern int sysctl_tcp_dsack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270extern int sysctl_tcp_wmem[3];
271extern int sysctl_tcp_rmem[3];
272extern int sysctl_tcp_app_win;
273extern int sysctl_tcp_adv_win_scale;
274extern int sysctl_tcp_tw_reuse;
275extern int sysctl_tcp_frto;
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800276extern int sysctl_tcp_frto_response;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277extern int sysctl_tcp_low_latency;
Chris Leech95937822006-05-23 18:02:55 -0700278extern int sysctl_tcp_dma_copybreak;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279extern int sysctl_tcp_nometrics_save;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280extern int sysctl_tcp_moderate_rcvbuf;
281extern int sysctl_tcp_tso_win_divisor;
John Heffner5d424d52006-03-20 17:53:41 -0800282extern int sysctl_tcp_mtu_probing;
283extern int sysctl_tcp_base_mss;
Rick Jones15d99e02006-03-20 22:40:29 -0800284extern int sysctl_tcp_workaround_signed_windows;
David S. Miller35089bb2006-06-13 22:33:04 -0700285extern int sysctl_tcp_slow_start_after_idle;
John Heffner886236c2007-03-25 19:21:45 -0700286extern int sysctl_tcp_max_ssthresh;
William Allen Simpson519855c2009-12-02 18:14:19 +0000287extern int sysctl_tcp_cookie_size;
Andreas Petlund36e31b0a2010-02-18 02:47:01 +0000288extern int sysctl_tcp_thin_linear_timeouts;
Andreas Petlund7e380172010-02-18 04:48:19 +0000289extern int sysctl_tcp_thin_dupack;
Yuchung Chengeed530b2012-05-02 13:30:03 +0000290extern int sysctl_tcp_early_retrans;
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000291extern int sysctl_tcp_limit_output_bytes;
Eric Dumazet282f23c2012-07-17 10:13:05 +0200292extern int sysctl_tcp_challenge_ack_limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293
Eric Dumazet8d987e52010-11-09 23:24:26 +0000294extern atomic_long_t tcp_memory_allocated;
Eric Dumazet17483762008-11-25 21:16:35 -0800295extern struct percpu_counter tcp_sockets_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296extern int tcp_memory_pressure;
297
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 * The next routines deal with comparing 32 bit unsigned ints
300 * and worry about wraparound (automatic with unsigned arithmetic).
301 */
302
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000303static inline bool before(__u32 seq1, __u32 seq2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304{
Gerrit Renker0d630cc2007-01-04 12:25:16 -0800305 return (__s32)(seq1-seq2) < 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306}
Gerrit Renker9a036b9c2006-12-20 10:25:55 -0800307#define after(seq2, seq1) before(seq1, seq2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
309/* is s2<=s1<=s3 ? */
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000310static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311{
312 return seq3 - seq2 >= seq1 - seq2;
313}
314
Arun Sharmaefcdbf22012-01-30 14:16:06 -0800315static inline bool tcp_out_of_memory(struct sock *sk)
316{
317 if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
318 sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2))
319 return true;
320 return false;
321}
322
David S. Millerad1af0f2010-08-25 02:27:49 -0700323static inline bool tcp_too_many_orphans(struct sock *sk, int shift)
Pavel Emelianove4fd5da2007-05-29 13:19:18 -0700324{
David S. Millerad1af0f2010-08-25 02:27:49 -0700325 struct percpu_counter *ocp = sk->sk_prot->orphan_count;
326 int orphans = percpu_counter_read_positive(ocp);
327
328 if (orphans << shift > sysctl_tcp_max_orphans) {
329 orphans = percpu_counter_sum_positive(ocp);
330 if (orphans << shift > sysctl_tcp_max_orphans)
331 return true;
332 }
David S. Millerad1af0f2010-08-25 02:27:49 -0700333 return false;
Pavel Emelianove4fd5da2007-05-29 13:19:18 -0700334}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335
Arun Sharmaefcdbf22012-01-30 14:16:06 -0800336extern bool tcp_check_oom(struct sock *sk, int shift);
337
Florian Westphala0f82f62009-04-19 09:43:48 +0000338/* syncookies: remember time of last synqueue overflow */
339static inline void tcp_synq_overflow(struct sock *sk)
340{
341 tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies;
342}
343
344/* syncookies: no recent synqueue overflow on this listening socket? */
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000345static inline bool tcp_synq_no_recent_overflow(const struct sock *sk)
Florian Westphala0f82f62009-04-19 09:43:48 +0000346{
347 unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp;
Jerry Chu9ad7c042011-06-08 11:08:38 +0000348 return time_after(jiffies, last_overflow + TCP_TIMEOUT_FALLBACK);
Florian Westphala0f82f62009-04-19 09:43:48 +0000349}
350
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351extern struct proto tcp_prot;
352
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -0700353#define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field)
354#define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
355#define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
356#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 +0000357#define TCP_ADD_STATS(net, field, val) SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358
Glauber Costa4acb4192012-01-30 01:20:17 +0000359extern void tcp_init_mem(struct net *net);
360
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000361extern void tcp_tasklet_init(void);
362
Changli Gao53d31762010-07-10 20:41:06 +0000363extern void tcp_v4_err(struct sk_buff *skb, u32);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364
Changli Gao53d31762010-07-10 20:41:06 +0000365extern void tcp_shutdown (struct sock *sk, int how);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366
David S. Miller160eb5a2012-06-27 22:01:22 -0700367extern void tcp_v4_early_demux(struct sk_buff *skb);
Changli Gao53d31762010-07-10 20:41:06 +0000368extern int tcp_v4_rcv(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369
Changli Gao53d31762010-07-10 20:41:06 +0000370extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
Changli Gao7ba42912010-07-10 20:41:55 +0000371extern int tcp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
372 size_t size);
373extern int tcp_sendpage(struct sock *sk, struct page *page, int offset,
374 size_t size, int flags);
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000375extern void tcp_release_cb(struct sock *sk);
Eric Dumazet6f458df2012-07-20 05:45:50 +0000376extern void tcp_write_timer_handler(struct sock *sk);
377extern void tcp_delack_timer_handler(struct sock *sk);
Changli Gao53d31762010-07-10 20:41:06 +0000378extern int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg);
379extern int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb,
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400380 const struct tcphdr *th, unsigned int len);
Changli Gao53d31762010-07-10 20:41:06 +0000381extern int tcp_rcv_established(struct sock *sk, struct sk_buff *skb,
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400382 const struct tcphdr *th, unsigned int len);
Changli Gao53d31762010-07-10 20:41:06 +0000383extern void tcp_rcv_space_adjust(struct sock *sk);
384extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
385extern int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp);
386extern void tcp_twsk_destructor(struct sock *sk);
387extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
388 struct pipe_inode_info *pipe, size_t len,
389 unsigned int flags);
Jens Axboe9c55e012007-11-06 23:30:13 -0800390
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700391static inline void tcp_dec_quickack_mode(struct sock *sk,
392 const unsigned int pkts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700394 struct inet_connection_sock *icsk = inet_csk(sk);
David S. Millerfc6415bc2005-07-05 15:17:45 -0700395
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700396 if (icsk->icsk_ack.quick) {
397 if (pkts >= icsk->icsk_ack.quick) {
398 icsk->icsk_ack.quick = 0;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700399 /* Leaving quickack mode we deflate ATO. */
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700400 icsk->icsk_ack.ato = TCP_ATO_MIN;
David S. Millerfc6415bc2005-07-05 15:17:45 -0700401 } else
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700402 icsk->icsk_ack.quick -= pkts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 }
404}
405
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -0700406#define TCP_ECN_OK 1
407#define TCP_ECN_QUEUE_CWR 2
408#define TCP_ECN_DEMAND_CWR 4
Eric Dumazet7a269ff2011-09-22 20:02:19 +0000409#define TCP_ECN_SEEN 8
Ilpo Järvinenbdf1ee52007-05-27 02:04:16 -0700410
Eric Dumazetfd2c3ef2009-11-03 03:26:03 +0000411enum tcp_tw_status {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 TCP_TW_SUCCESS = 0,
413 TCP_TW_RST = 1,
414 TCP_TW_ACK = 2,
415 TCP_TW_SYN = 3
416};
417
418
Changli Gao53d31762010-07-10 20:41:06 +0000419extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
420 struct sk_buff *skb,
421 const struct tcphdr *th);
422extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
423 struct request_sock *req,
Jerry Chu83368862012-08-31 12:29:12 +0000424 struct request_sock **prev,
425 bool fastopen);
Changli Gao53d31762010-07-10 20:41:06 +0000426extern int tcp_child_process(struct sock *parent, struct sock *child,
427 struct sk_buff *skb);
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000428extern bool tcp_use_frto(struct sock *sk);
Changli Gao53d31762010-07-10 20:41:06 +0000429extern void tcp_enter_frto(struct sock *sk);
430extern void tcp_enter_loss(struct sock *sk, int how);
431extern void tcp_clear_retrans(struct tcp_sock *tp);
432extern void tcp_update_metrics(struct sock *sk);
David S. Miller4aabd8e2012-07-09 16:07:30 -0700433extern void tcp_init_metrics(struct sock *sk);
David S. Miller51c5d0c2012-07-10 00:49:14 -0700434extern void tcp_metrics_init(void);
David S. Miller81166dd2012-07-10 03:14:24 -0700435extern bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst, bool paws_check);
436extern bool tcp_remember_stamp(struct sock *sk);
437extern bool tcp_tw_remember_stamp(struct inet_timewait_sock *tw);
438extern void tcp_fetch_timewait_stamp(struct sock *sk, struct dst_entry *dst);
David S. Miller4aabd8e2012-07-09 16:07:30 -0700439extern void tcp_disable_fack(struct tcp_sock *tp);
Changli Gao53d31762010-07-10 20:41:06 +0000440extern void tcp_close(struct sock *sk, long timeout);
Neal Cardwell900f65d2012-04-19 09:55:21 +0000441extern void tcp_init_sock(struct sock *sk);
Changli Gao53d31762010-07-10 20:41:06 +0000442extern unsigned int tcp_poll(struct file * file, struct socket *sock,
443 struct poll_table_struct *wait);
444extern int tcp_getsockopt(struct sock *sk, int level, int optname,
445 char __user *optval, int __user *optlen);
446extern int tcp_setsockopt(struct sock *sk, int level, int optname,
447 char __user *optval, unsigned int optlen);
448extern int compat_tcp_getsockopt(struct sock *sk, int level, int optname,
449 char __user *optval, int __user *optlen);
450extern int compat_tcp_setsockopt(struct sock *sk, int level, int optname,
451 char __user *optval, unsigned int optlen);
452extern void tcp_set_keepalive(struct sock *sk, int val);
453extern void tcp_syn_ack_timeout(struct sock *sk, struct request_sock *req);
454extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
455 size_t len, int nonblock, int flags, int *addr_len);
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400456extern void tcp_parse_options(const struct sk_buff *skb,
457 struct tcp_options_received *opt_rx, const u8 **hvpp,
Yuchung Cheng2100c8d2012-07-19 06:43:05 +0000458 int estab, struct tcp_fastopen_cookie *foc);
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400459extern const u8 *tcp_parse_md5sig_option(const struct tcphdr *th);
YOSHIFUJI Hideaki7d5d5522008-04-17 12:29:53 +0900460
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461/*
462 * TCP v4 functions exported for the inet6 API
463 */
464
Changli Gao53d31762010-07-10 20:41:06 +0000465extern void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb);
466extern int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
467extern struct sock * tcp_create_openreq_child(struct sock *sk,
468 struct request_sock *req,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 struct sk_buff *skb);
Changli Gao53d31762010-07-10 20:41:06 +0000470extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
471 struct request_sock *req,
472 struct dst_entry *dst);
473extern int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
474extern int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr,
475 int addr_len);
476extern int tcp_connect(struct sock *sk);
477extern struct sk_buff * tcp_make_synack(struct sock *sk, struct dst_entry *dst,
478 struct request_sock *req,
Jerry Chu83368862012-08-31 12:29:12 +0000479 struct request_values *rvp,
480 struct tcp_fastopen_cookie *foc);
Changli Gao53d31762010-07-10 20:41:06 +0000481extern int tcp_disconnect(struct sock *sk, int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482
Pavel Emelyanov370816a2012-04-19 03:40:01 +0000483void tcp_connect_init(struct sock *sk);
484void tcp_finish_connect(struct sock *sk, struct sk_buff *skb);
Pavel Emelyanov292e8d82012-05-10 01:49:41 +0000485int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size);
Eric Dumazet63d02d12012-08-09 14:11:00 +0000486void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488/* From syncookies.c */
Florian Westphal2051f112008-03-23 22:21:28 -0700489extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
491 struct ip_options *opt);
Eric Dumazete05c82d2011-09-18 21:02:55 -0400492#ifdef CONFIG_SYN_COOKIES
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
494 __u16 *mss);
Eric Dumazete05c82d2011-09-18 21:02:55 -0400495#else
496static inline __u32 cookie_v4_init_sequence(struct sock *sk,
497 struct sk_buff *skb,
498 __u16 *mss)
499{
500 return 0;
501}
502#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Florian Westphal4dfc2812008-04-10 03:12:40 -0700504extern __u32 cookie_init_timestamp(struct request_sock *req);
Hannes Frederic Sowa5d134f12013-01-05 16:10:48 +0000505extern bool cookie_check_timestamp(struct tcp_options_received *opt,
506 struct net *net, bool *ecn_ok);
Florian Westphal4dfc2812008-04-10 03:12:40 -0700507
Glenn Griffinc6aefaf2008-02-07 21:49:26 -0800508/* From net/ipv6/syncookies.c */
509extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
Eric Dumazete05c82d2011-09-18 21:02:55 -0400510#ifdef CONFIG_SYN_COOKIES
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400511extern __u32 cookie_v6_init_sequence(struct sock *sk, const struct sk_buff *skb,
Glenn Griffinc6aefaf2008-02-07 21:49:26 -0800512 __u16 *mss);
Eric Dumazete05c82d2011-09-18 21:02:55 -0400513#else
514static inline __u32 cookie_v6_init_sequence(struct sock *sk,
515 struct sk_buff *skb,
516 __u16 *mss)
517{
518 return 0;
519}
520#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521/* tcp_output.c */
522
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700523extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
524 int nonagle);
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000525extern bool tcp_may_send_now(struct sock *sk);
Yuchung Cheng93b174a2012-12-06 08:45:32 +0000526extern int __tcp_retransmit_skb(struct sock *, struct sk_buff *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000528extern void tcp_retransmit_timer(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529extern void tcp_xmit_retransmit_queue(struct sock *);
530extern void tcp_simple_retransmit(struct sock *);
531extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
David S. Miller6475be12005-09-01 22:47:01 -0700532extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
534extern void tcp_send_probe0(struct sock *);
535extern void tcp_send_partial(struct sock *);
Changli Gao53d31762010-07-10 20:41:06 +0000536extern int tcp_write_wakeup(struct sock *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537extern void tcp_send_fin(struct sock *sk);
Al Virodd0fc662005-10-07 07:46:04 +0100538extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
Changli Gao53d31762010-07-10 20:41:06 +0000539extern int tcp_send_synack(struct sock *);
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000540extern bool tcp_syn_flood_action(struct sock *sk,
541 const struct sk_buff *skb,
542 const char *proto);
David S. Millerc1b4a7e2005-07-05 15:24:38 -0700543extern void tcp_push_one(struct sock *, unsigned int mss_now);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544extern void tcp_send_ack(struct sock *sk);
545extern void tcp_send_delayed_ack(struct sock *sk);
546
David S. Millera762a982005-07-05 15:18:51 -0700547/* tcp_input.c */
548extern void tcp_cwnd_application_limited(struct sock *sk);
Yuchung Cheng750ea2b2012-05-02 13:30:04 +0000549extern void tcp_resume_early_retransmit(struct sock *sk);
550extern void tcp_rearm_rto(struct sock *sk);
Jerry Chu10467162012-08-31 12:29:11 +0000551extern void tcp_reset(struct sock *sk);
David S. Millera762a982005-07-05 15:18:51 -0700552
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553/* tcp_timer.c */
554extern void tcp_init_xmit_timers(struct sock *);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700555static inline void tcp_clear_xmit_timers(struct sock *sk)
556{
557 inet_csk_clear_xmit_timers(sk);
558}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
Ilpo Järvinen0c54b852009-03-14 14:23:05 +0000561extern unsigned int tcp_current_mss(struct sock *sk);
562
563/* Bound MSS / TSO packet size with the half of the window */
564static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
565{
Alexey Kuznetsov01f83d62010-09-15 10:27:52 -0700566 int cutoff;
567
568 /* When peer uses tiny windows, there is no use in packetizing
569 * to sub-MSS pieces for the sake of SWS or making sure there
570 * are enough packets in the pipe for fast recovery.
571 *
572 * On the other hand, for extremely large MSS devices, handling
573 * smaller than MSS windows in this way does make sense.
574 */
575 if (tp->max_window >= 512)
576 cutoff = (tp->max_window >> 1);
577 else
578 cutoff = tp->max_window;
579
580 if (cutoff && pktsize > cutoff)
581 return max_t(int, cutoff, 68U - tp->tcp_header_len);
Ilpo Järvinen0c54b852009-03-14 14:23:05 +0000582 else
583 return pktsize;
584}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
Arnaldo Carvalho de Melo17b085e2005-08-12 12:59:17 -0300586/* tcp.c */
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400587extern void tcp_get_info(const struct sock *, struct tcp_info *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
589/* Read 'sendfile()'-style from a TCP socket */
590typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
591 unsigned int, size_t);
592extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
593 sk_read_actor_t recv_actor);
594
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800595extern void tcp_initialize_rcv_mss(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596
Eric Dumazet67469602012-04-24 07:37:38 +0000597extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
598extern int tcp_mss_to_mtu(struct sock *sk, int mss);
John Heffner5d424d52006-03-20 17:53:41 -0800599extern void tcp_mtup_init(struct sock *sk);
Jerry Chu9ad7c042011-06-08 11:08:38 +0000600extern void tcp_valid_rtt_meas(struct sock *sk, u32 seq_rtt);
Jerry Chu10467162012-08-31 12:29:11 +0000601extern void tcp_init_buffer_space(struct sock *sk);
John Heffner5d424d52006-03-20 17:53:41 -0800602
Damian Lukowskif1ecd5d2009-08-26 00:16:31 +0000603static inline void tcp_bound_rto(const struct sock *sk)
604{
605 if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX)
606 inet_csk(sk)->icsk_rto = TCP_RTO_MAX;
607}
608
609static inline u32 __tcp_set_rto(const struct tcp_sock *tp)
610{
611 return (tp->srtt >> 3) + tp->rttvar;
612}
613
David S. Miller4aabd8e2012-07-09 16:07:30 -0700614extern void tcp_set_rto(struct sock *sk);
615
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800616static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617{
618 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
619 ntohl(TCP_FLAG_ACK) |
620 snd_wnd);
621}
622
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800623static inline void tcp_fast_path_on(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624{
625 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
626}
627
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700628static inline void tcp_fast_path_check(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629{
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700630 struct tcp_sock *tp = tcp_sk(sk);
631
David S. Millerb03efcf2005-07-08 14:57:23 -0700632 if (skb_queue_empty(&tp->out_of_order_queue) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 tp->rcv_wnd &&
634 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
635 !tp->urg_data)
636 tcp_fast_path_on(tp);
637}
638
Satoru SATOH0c266892009-05-04 11:11:01 -0700639/* Compute the actual rto_min value */
640static inline u32 tcp_rto_min(struct sock *sk)
641{
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400642 const struct dst_entry *dst = __sk_dst_get(sk);
Satoru SATOH0c266892009-05-04 11:11:01 -0700643 u32 rto_min = TCP_RTO_MIN;
644
645 if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
646 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
647 return rto_min;
648}
649
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650/* Compute the actual receive window we are currently advertising.
651 * Rcv_nxt can be after the window if our peer push more data
652 * than the offered window.
653 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800654static inline u32 tcp_receive_window(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655{
656 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
657
658 if (win < 0)
659 win = 0;
660 return (u32) win;
661}
662
663/* Choose a new window, without checks for shrinking, and without
664 * scaling applied to the result. The caller does these things
665 * if necessary. This is a "raw" window selection.
666 */
Changli Gao53d31762010-07-10 20:41:06 +0000667extern u32 __tcp_select_window(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668
Pavel Emelyanovee995282012-04-19 03:40:39 +0000669void tcp_send_window_probe(struct sock *sk);
670
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671/* TCP timestamps are only 32-bits, this causes a slight
672 * complication on 64-bit systems since we store a snapshot
Stephen Hemminger31f34262005-11-15 15:17:10 -0800673 * of jiffies in the buffer control blocks below. We decided
674 * to use only the low 32-bits of jiffies and hide the ugly
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 * casts with the following macro.
676 */
677#define tcp_time_stamp ((__u32)(jiffies))
678
Changli Gaoa3433f32010-06-12 14:01:43 +0000679#define tcp_flag_byte(th) (((u_int8_t *)th)[13])
680
681#define TCPHDR_FIN 0x01
682#define TCPHDR_SYN 0x02
683#define TCPHDR_RST 0x04
684#define TCPHDR_PSH 0x08
685#define TCPHDR_ACK 0x10
686#define TCPHDR_URG 0x20
687#define TCPHDR_ECE 0x40
688#define TCPHDR_CWR 0x80
689
Stephen Hemmingercaa20d9a2005-11-10 17:13:47 -0800690/* This is what the send packet queuing engine uses to pass
Eric Dumazetf86586f2010-07-15 21:41:00 -0700691 * TCP per-packet control information to the transmission code.
692 * We also store the host-order sequence numbers in here too.
693 * This is 44 bytes if IPV6 is enabled.
694 * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 */
696struct tcp_skb_cb {
697 union {
698 struct inet_skb_parm h4;
Eric Dumazetdfd56b82011-12-10 09:48:31 +0000699#if IS_ENABLED(CONFIG_IPV6)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 struct inet6_skb_parm h6;
701#endif
702 } header; /* For incoming frames */
703 __u32 seq; /* Starting sequence number */
704 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
705 __u32 when; /* used to compute rtt's */
Eric Dumazet4de075e2011-09-27 13:25:05 -0400706 __u8 tcp_flags; /* TCP header flags. (tcp[13]) */
Neal Cardwellf4f9f6e2012-04-16 07:08:06 +0000707
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 __u8 sacked; /* State flags for SACK/FACK. */
709#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
710#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
711#define TCPCB_LOST 0x04 /* SKB is lost */
712#define TCPCB_TAGBITS 0x07 /* All tag bits */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
714#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
715
Neal Cardwellf4f9f6e2012-04-16 07:08:06 +0000716 __u8 ip_dsfield; /* IPv4 tos or IPv6 dsfield */
717 /* 1 byte hole */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 __u32 ack_seq; /* Sequence number ACK'd */
719};
720
721#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
722
Eric Dumazetbd14b1b2012-05-04 05:14:02 +0000723/* RFC3168 : 6.1.1 SYN packets must not have ECT/ECN bits set
724 *
725 * If we receive a SYN packet with these bits set, it means a network is
726 * playing bad games with TOS bits. In order to avoid possible false congestion
727 * notifications, we disable TCP ECN negociation.
728 */
729static inline void
Hannes Frederic Sowa5d134f12013-01-05 16:10:48 +0000730TCP_ECN_create_request(struct request_sock *req, const struct sk_buff *skb,
731 struct net *net)
Eric Dumazetbd14b1b2012-05-04 05:14:02 +0000732{
733 const struct tcphdr *th = tcp_hdr(skb);
734
Hannes Frederic Sowa5d134f12013-01-05 16:10:48 +0000735 if (net->ipv4.sysctl_tcp_ecn && th->ece && th->cwr &&
Eric Dumazetbd14b1b2012-05-04 05:14:02 +0000736 INET_ECN_is_not_ect(TCP_SKB_CB(skb)->ip_dsfield))
737 inet_rsk(req)->ecn_ok = 1;
738}
739
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740/* Due to TSO, an SKB can be composed of multiple actual
741 * packets. To keep these tracked properly, we use this.
742 */
743static inline int tcp_skb_pcount(const struct sk_buff *skb)
744{
Herbert Xu79671682006-06-22 02:40:14 -0700745 return skb_shinfo(skb)->gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746}
747
748/* This is valid iff tcp_skb_pcount() > 1. */
749static inline int tcp_skb_mss(const struct sk_buff *skb)
750{
Herbert Xu79671682006-06-22 02:40:14 -0700751 return skb_shinfo(skb)->gso_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752}
753
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700754/* Events passed to congestion control interface */
755enum tcp_ca_event {
756 CA_EVENT_TX_START, /* first transmit when no packets in flight */
757 CA_EVENT_CWND_RESTART, /* congestion window restart */
758 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
759 CA_EVENT_FRTO, /* fast recovery timeout */
760 CA_EVENT_LOSS, /* loss timeout */
761 CA_EVENT_FAST_ACK, /* in sequence ack */
762 CA_EVENT_SLOW_ACK, /* other ack */
763};
764
765/*
766 * Interface for adding new TCP congestion control handlers
767 */
768#define TCP_CA_NAME_MAX 16
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800769#define TCP_CA_MAX 128
770#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
771
Stephen Hemminger164891a2007-04-23 22:26:16 -0700772#define TCP_CONG_NON_RESTRICTED 0x1
773#define TCP_CONG_RTT_STAMP 0x2
774
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700775struct tcp_congestion_ops {
776 struct list_head list;
Stephen Hemminger164891a2007-04-23 22:26:16 -0700777 unsigned long flags;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700778
779 /* initialize private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300780 void (*init)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700781 /* cleanup private data (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300782 void (*release)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700783
784 /* return slow start threshold (required) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300785 u32 (*ssthresh)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700786 /* lower bound for congestion window (optional) */
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700787 u32 (*min_cwnd)(const struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700788 /* do new cwnd calculation (required) */
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200789 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700790 /* call before changing ca_state (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300791 void (*set_state)(struct sock *sk, u8 new_state);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700792 /* call when cwnd event occurs (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300793 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700794 /* new value of cwnd after loss (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300795 u32 (*undo_cwnd)(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700796 /* hook for packet ack accounting (optional) */
Stephen Hemminger30cfd0b2007-07-25 23:49:34 -0700797 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
Arnaldo Carvalho de Melo73c1f4a2005-08-12 12:51:49 -0300798 /* get info for inet_diag (optional) */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300799 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700800
801 char name[TCP_CA_NAME_MAX];
802 struct module *owner;
803};
804
805extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
806extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
807
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300808extern void tcp_init_congestion_control(struct sock *sk);
809extern void tcp_cleanup_congestion_control(struct sock *sk);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700810extern int tcp_set_default_congestion_control(const char *name);
811extern void tcp_get_default_congestion_control(char *name);
Stephen Hemminger3ff825b2006-11-09 16:32:06 -0800812extern void tcp_get_available_congestion_control(char *buf, size_t len);
Stephen Hemmingerce7bc3b2006-11-09 16:35:15 -0800813extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
814extern int tcp_set_allowed_congestion_control(char *allowed);
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300815extern int tcp_set_congestion_control(struct sock *sk, const char *name);
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800816extern void tcp_slow_start(struct tcp_sock *tp);
Ilpo Järvinen758ce5c2009-02-28 04:44:37 +0000817extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700818
Stephen Hemminger5f8ef482005-06-23 20:37:36 -0700819extern struct tcp_congestion_ops tcp_init_congestion_ops;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300820extern u32 tcp_reno_ssthresh(struct sock *sk);
Ilpo Järvinenc3a05c62007-12-02 00:47:59 +0200821extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
Stephen Hemminger72dc5b92006-06-05 17:30:08 -0700822extern u32 tcp_reno_min_cwnd(const struct sock *sk);
David S. Millera8acfba2005-06-23 23:45:02 -0700823extern struct tcp_congestion_ops tcp_reno;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700824
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300825static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700826{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300827 struct inet_connection_sock *icsk = inet_csk(sk);
828
829 if (icsk->icsk_ca_ops->set_state)
830 icsk->icsk_ca_ops->set_state(sk, ca_state);
831 icsk->icsk_ca_state = ca_state;
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700832}
833
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300834static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700835{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300836 const struct inet_connection_sock *icsk = inet_csk(sk);
837
838 if (icsk->icsk_ca_ops->cwnd_event)
839 icsk->icsk_ca_ops->cwnd_event(sk, event);
Stephen Hemminger317a76f2005-06-23 12:19:55 -0700840}
841
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300842/* These functions determine how the current flow behaves in respect of SACK
843 * handling. SACK is negotiated with the peer, and therefore it can vary
844 * between different flows.
845 *
846 * tcp_is_sack - SACK enabled
847 * tcp_is_reno - No SACK
848 * tcp_is_fack - FACK enabled, implies SACK enabled
849 */
850static inline int tcp_is_sack(const struct tcp_sock *tp)
851{
852 return tp->rx_opt.sack_ok;
853}
854
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000855static inline bool tcp_is_reno(const struct tcp_sock *tp)
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300856{
857 return !tcp_is_sack(tp);
858}
859
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000860static inline bool tcp_is_fack(const struct tcp_sock *tp)
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300861{
Vijay Subramanianab562222011-12-20 13:23:24 +0000862 return tp->rx_opt.sack_ok & TCP_FACK_ENABLED;
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300863}
864
865static inline void tcp_enable_fack(struct tcp_sock *tp)
866{
Vijay Subramanianab562222011-12-20 13:23:24 +0000867 tp->rx_opt.sack_ok |= TCP_FACK_ENABLED;
Ilpo Järvinene60402d2007-08-09 15:14:46 +0300868}
869
Yuchung Chengeed530b2012-05-02 13:30:03 +0000870/* TCP early-retransmit (ER) is similar to but more conservative than
871 * the thin-dupack feature. Enable ER only if thin-dupack is disabled.
872 */
873static inline void tcp_enable_early_retrans(struct tcp_sock *tp)
874{
875 tp->do_early_retrans = sysctl_tcp_early_retrans &&
876 !sysctl_tcp_thin_dupack && sysctl_tcp_reordering == 3;
Yuchung Cheng750ea2b2012-05-02 13:30:04 +0000877 tp->early_retrans_delayed = 0;
Yuchung Chengeed530b2012-05-02 13:30:03 +0000878}
879
880static inline void tcp_disable_early_retrans(struct tcp_sock *tp)
881{
882 tp->do_early_retrans = 0;
883}
884
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300885static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
886{
887 return tp->sacked_out + tp->lost_out;
888}
889
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890/* This determines how many packets are "in the network" to the best
891 * of our knowledge. In many cases it is conservative, but where
892 * detailed information is available from the receiver (via SACK
893 * blocks etc.) we can make more aggressive calculations.
894 *
895 * Use this for decisions involving congestion control, use just
896 * tp->packets_out to determine if the send queue is empty or not.
897 *
898 * Read this equation as:
899 *
900 * "Packets sent once on transmission queue" MINUS
901 * "Packets left network, but not honestly ACKed yet" PLUS
902 * "Packets fast retransmitted"
903 */
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800904static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905{
Ilpo Järvinen83ae4082007-08-09 14:37:30 +0300906 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907}
908
Ilpo Järvinen0b6a05c2009-09-15 01:30:10 -0700909#define TCP_INFINITE_SSTHRESH 0x7fffffff
910
911static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp)
912{
913 return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH;
914}
915
Yuchung Cheng684bad12012-09-02 17:38:04 +0000916static inline bool tcp_in_cwnd_reduction(const struct sock *sk)
917{
918 return (TCPF_CA_CWR | TCPF_CA_Recovery) &
919 (1 << inet_csk(sk)->icsk_ca_state);
920}
921
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
Yuchung Cheng684bad12012-09-02 17:38:04 +0000923 * The exception is cwnd reduction phase, when cwnd is decreasing towards
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 * ssthresh.
925 */
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300926static inline __u32 tcp_current_ssthresh(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927{
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300928 const struct tcp_sock *tp = tcp_sk(sk);
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400929
Yuchung Cheng684bad12012-09-02 17:38:04 +0000930 if (tcp_in_cwnd_reduction(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 return tp->snd_ssthresh;
932 else
933 return max(tp->snd_ssthresh,
934 ((tp->snd_cwnd >> 1) +
935 (tp->snd_cwnd >> 2)));
936}
937
Ilpo Järvinenb9c45952007-07-27 16:36:17 +0300938/* Use define here intentionally to get WARN_ON location shown at the caller */
939#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940
Ilpo Järvinen3cfe3ba2007-02-27 10:09:49 -0800941extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400942extern __u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943
Neal Cardwell6b5a5c02011-11-21 17:15:14 +0000944/* The maximum number of MSS of available cwnd for which TSO defers
945 * sending if not using sysctl_tcp_tso_win_divisor.
946 */
947static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp)
948{
949 return 3;
950}
951
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952/* Slow start with delack produces 3 packets of burst, so that
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700953 * it is safe "de facto". This will be the default - same as
954 * the default reordering threshold - but if reordering increases,
955 * we must be able to allow cwnd to burst at least this much in order
956 * to not pull it back when holes are filled.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 */
958static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
959{
John Heffnerdd9e0dd2008-04-15 15:26:39 -0700960 return tp->reordering;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961}
962
Ilpo Järvinen90840de2007-12-31 04:48:41 -0800963/* Returns end sequence number of the receiver's advertised window */
964static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
965{
966 return tp->snd_una + tp->snd_wnd;
967}
Eric Dumazeta2a385d2012-05-16 23:15:34 +0000968extern bool tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
Stephen Hemmingerf4805ed2005-11-10 16:53:30 -0800969
Chuck Leverc1bd24b2007-10-23 21:08:54 -0700970static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
Stephen Hemminger40efc6f2006-01-03 16:03:49 -0800971 const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972{
973 if (skb->len < mss)
974 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
975}
976
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700977static inline void tcp_check_probe_timer(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978{
Eric Dumazetcf533ea2011-10-21 05:22:42 -0400979 const struct tcp_sock *tp = tcp_sk(sk);
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700980 const struct inet_connection_sock *icsk = inet_csk(sk);
Ilpo Järvinen9e412ba2007-04-20 22:18:02 -0700981
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -0700982 if (!tp->packets_out && !icsk->icsk_pending)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700983 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
984 icsk->icsk_rto, TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985}
986
Hantzis Fotisee7537b2009-03-02 22:42:02 -0800987static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988{
989 tp->snd_wl1 = seq;
990}
991
Hantzis Fotisee7537b2009-03-02 22:42:02 -0800992static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993{
994 tp->snd_wl1 = seq;
995}
996
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997/*
998 * Calculate(/check) TCP checksum
999 */
Frederik Deweerdtba7808e2007-02-04 20:15:27 -08001000static inline __sum16 tcp_v4_check(int len, __be32 saddr,
1001 __be32 daddr, __wsum base)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002{
1003 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
1004}
1005
Al Virob51655b2006-11-14 21:40:42 -08001006static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007{
Herbert Xufb286bb2005-11-10 13:01:24 -08001008 return __skb_checksum_complete(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009}
1010
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001011static inline bool tcp_checksum_complete(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012{
Herbert Xu60476372007-04-09 11:59:39 -07001013 return !skb_csum_unnecessary(skb) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 __tcp_checksum_complete(skb);
1015}
1016
1017/* Prequeue for VJ style copy to user, combined with checksumming. */
1018
Stephen Hemminger40efc6f2006-01-03 16:03:49 -08001019static inline void tcp_prequeue_init(struct tcp_sock *tp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020{
1021 tp->ucopy.task = NULL;
1022 tp->ucopy.len = 0;
1023 tp->ucopy.memory = 0;
1024 skb_queue_head_init(&tp->ucopy.prequeue);
Chris Leech97fc2f02006-05-23 17:55:33 -07001025#ifdef CONFIG_NET_DMA
1026 tp->ucopy.dma_chan = NULL;
1027 tp->ucopy.wakeup = 0;
1028 tp->ucopy.pinned_list = NULL;
1029 tp->ucopy.dma_cookie = 0;
1030#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031}
1032
1033/* Packet is added to VJ-style prequeue for processing in process
1034 * context, if a reader task is waiting. Apparently, this exciting
1035 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
1036 * failed somewhere. Latency? Burstiness? Well, at least now we will
1037 * see, why it failed. 8)8) --ANK
1038 *
1039 * NOTE: is this not too big to inline?
1040 */
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001041static inline bool tcp_prequeue(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042{
1043 struct tcp_sock *tp = tcp_sk(sk);
1044
Eric Dumazetf5f8d862009-05-07 07:08:38 +00001045 if (sysctl_tcp_low_latency || !tp->ucopy.task)
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001046 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047
Eric Dumazet79ffef12013-02-27 07:05:03 +00001048 if (skb->len <= tcp_hdrlen(skb) &&
1049 skb_queue_len(&tp->ucopy.prequeue) == 0)
1050 return false;
1051
Eric Dumazetf5f8d862009-05-07 07:08:38 +00001052 __skb_queue_tail(&tp->ucopy.prequeue, skb);
1053 tp->ucopy.memory += skb->truesize;
1054 if (tp->ucopy.memory > sk->sk_rcvbuf) {
1055 struct sk_buff *skb1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056
Eric Dumazetf5f8d862009-05-07 07:08:38 +00001057 BUG_ON(sock_owned_by_user(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058
Eric Dumazetf5f8d862009-05-07 07:08:38 +00001059 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
1060 sk_backlog_rcv(sk, skb1);
1061 NET_INC_STATS_BH(sock_net(sk),
1062 LINUX_MIB_TCPPREQUEUEDROPPED);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 }
Eric Dumazetf5f8d862009-05-07 07:08:38 +00001064
1065 tp->ucopy.memory = 0;
1066 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
Eric Dumazetaa395142010-04-20 13:03:51 +00001067 wake_up_interruptible_sync_poll(sk_sleep(sk),
Eric Dumazet7aedec22009-05-07 07:20:39 +00001068 POLLIN | POLLRDNORM | POLLRDBAND);
Eric Dumazetf5f8d862009-05-07 07:08:38 +00001069 if (!inet_csk_ack_scheduled(sk))
1070 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
David S. Miller22f6dac2009-05-08 02:48:30 -07001071 (3 * tcp_rto_min(sk)) / 4,
Eric Dumazetf5f8d862009-05-07 07:08:38 +00001072 TCP_RTO_MAX);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 }
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001074 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075}
1076
1077
1078#undef STATE_TRACE
1079
1080#ifdef STATE_TRACE
1081static const char *statename[]={
1082 "Unused","Established","Syn Sent","Syn Recv",
1083 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
1084 "Close Wait","Last ACK","Listen","Closing"
1085};
1086#endif
Ilpo Järvinen490d5042008-01-12 03:17:20 -08001087extern void tcp_set_state(struct sock *sk, int state);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088
Andi Kleen4ac02ba2007-04-20 17:11:46 -07001089extern void tcp_done(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090
Stephen Hemminger40efc6f2006-01-03 16:03:49 -08001091static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092{
1093 rx_opt->dsack = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 rx_opt->num_sacks = 0;
1095}
1096
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097/* Determine a window scaling and initial window to offer. */
1098extern void tcp_select_initial_window(int __space, __u32 mss,
1099 __u32 *rcv_wnd, __u32 *window_clamp,
laurent chavey31d12922009-12-15 11:15:28 +00001100 int wscale_ok, __u8 *rcv_wscale,
1101 __u32 init_rcv_wnd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102
1103static inline int tcp_win_from_space(int space)
1104{
1105 return sysctl_tcp_adv_win_scale<=0 ?
1106 (space>>(-sysctl_tcp_adv_win_scale)) :
1107 space - (space>>sysctl_tcp_adv_win_scale);
1108}
1109
1110/* Note: caller must be prepared to deal with negative returns */
1111static inline int tcp_space(const struct sock *sk)
1112{
1113 return tcp_win_from_space(sk->sk_rcvbuf -
1114 atomic_read(&sk->sk_rmem_alloc));
1115}
1116
1117static inline int tcp_full_space(const struct sock *sk)
1118{
1119 return tcp_win_from_space(sk->sk_rcvbuf);
1120}
1121
Stephen Hemminger40efc6f2006-01-03 16:03:49 -08001122static inline void tcp_openreq_init(struct request_sock *req,
1123 struct tcp_options_received *rx_opt,
1124 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125{
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001126 struct inet_request_sock *ireq = inet_rsk(req);
1127
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
Florian Westphal4dfc2812008-04-10 03:12:40 -07001129 req->cookie_ts = 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001130 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
Jerry Chu10467162012-08-31 12:29:11 +00001131 tcp_rsk(req)->rcv_nxt = TCP_SKB_CB(skb)->seq + 1;
Neal Cardwell016818d2012-09-22 04:18:55 +00001132 tcp_rsk(req)->snt_synack = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 req->mss = rx_opt->mss_clamp;
1134 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001135 ireq->tstamp_ok = rx_opt->tstamp_ok;
1136 ireq->sack_ok = rx_opt->sack_ok;
1137 ireq->snd_wscale = rx_opt->snd_wscale;
1138 ireq->wscale_ok = rx_opt->wscale_ok;
1139 ireq->acked = 0;
1140 ireq->ecn_ok = 0;
Arnaldo Carvalho de Meloaa8223c2007-04-10 21:04:22 -07001141 ireq->rmt_port = tcp_hdr(skb)->source;
KOVACS Krisztiana3116ac52008-10-01 07:46:49 -07001142 ireq->loc_port = tcp_hdr(skb)->dest;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143}
1144
Neal Cardwell623df482012-09-22 04:18:54 +00001145/* Compute time elapsed between SYNACK and the ACK completing 3WHS */
1146static inline void tcp_synack_rtt_meas(struct sock *sk,
1147 struct request_sock *req)
1148{
1149 if (tcp_rsk(req)->snt_synack)
1150 tcp_valid_rtt_meas(sk,
1151 tcp_time_stamp - tcp_rsk(req)->snt_synack);
1152}
1153
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001154extern void tcp_enter_memory_pressure(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1157{
1158 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1159}
1160
1161static inline int keepalive_time_when(const struct tcp_sock *tp)
1162{
1163 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1164}
1165
Eric Dumazetdf19a622009-08-28 23:48:54 -07001166static inline int keepalive_probes(const struct tcp_sock *tp)
1167{
1168 return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes;
1169}
1170
Flavio Leitner6c37e5d2010-04-26 18:33:27 +00001171static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp)
1172{
1173 const struct inet_connection_sock *icsk = &tp->inet_conn;
1174
1175 return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime,
1176 tcp_time_stamp - tp->rcv_tstamp);
1177}
1178
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001179static inline int tcp_fin_time(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180{
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001181 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1182 const int rto = inet_csk(sk)->icsk_rto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
Arnaldo Carvalho de Melo463c84b2005-08-09 20:10:42 -07001184 if (fin_timeout < (rto << 2) - (rto >> 1))
1185 fin_timeout = (rto << 2) - (rto >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186
1187 return fin_timeout;
1188}
1189
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001190static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt,
1191 int paws_win)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192{
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001193 if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001194 return true;
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001195 if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001196 return true;
Eric Dumazetbc2ce892010-12-16 14:08:34 -08001197 /*
1198 * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0,
1199 * then following tcp messages have valid values. Ignore 0 value,
1200 * or else 'negative' tsval might forbid us to accept their packets.
1201 */
1202 if (!rx_opt->ts_recent)
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001203 return true;
1204 return false;
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001205}
1206
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001207static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt,
1208 int rst)
Ilpo Järvinenc887e6d2009-03-14 14:23:03 +00001209{
1210 if (tcp_paws_check(rx_opt, 0))
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001211 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212
1213 /* RST segments are not recommended to carry timestamp,
1214 and, if they do, it is recommended to ignore PAWS because
1215 "their cleanup function should take precedence over timestamps."
1216 Certainly, it is mistake. It is necessary to understand the reasons
1217 of this constraint to relax it: if peer reboots, clock may go
1218 out-of-sync and half-open connections will not be reset.
1219 Actually, the problem would be not existing if all
1220 the implementations followed draft about maintaining clock
1221 via reboots. Linux-2.2 DOES NOT!
1222
1223 However, we can relax time bounds for RST segments to MSL.
1224 */
James Morris9d729f72007-03-04 16:12:44 -08001225 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001226 return false;
1227 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228}
1229
Pavel Emelyanova9c193292008-07-16 20:21:42 -07001230static inline void tcp_mib_init(struct net *net)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231{
1232 /* See RFC 2012 */
Pavel Emelyanovcf1100a2008-07-16 20:27:38 -07001233 TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1234 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1235 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1236 TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237}
1238
Ilpo Järvinen5af4ec22007-09-20 11:30:48 -07001239/* from STCP */
Ilpo Järvinenef9da472008-09-20 21:25:15 -07001240static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
David S. Miller0800f172007-09-20 11:40:37 -07001241{
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001242 tp->lost_skb_hint = NULL;
1243 tp->scoreboard_skb_hint = NULL;
Ilpo Järvinenef9da472008-09-20 21:25:15 -07001244}
1245
1246static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1247{
1248 tcp_clear_retrans_hints_partial(tp);
Stephen Hemminger6a438bb2005-11-10 17:14:59 -08001249 tp->retransmit_skb_hint = NULL;
Ilpo Järvinenb7689202007-09-20 11:37:19 -07001250}
1251
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001252/* MD5 Signature */
1253struct crypto_hash;
1254
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001255union tcp_md5_addr {
1256 struct in_addr a4;
1257#if IS_ENABLED(CONFIG_IPV6)
1258 struct in6_addr a6;
1259#endif
1260};
1261
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001262/* - key database */
1263struct tcp_md5sig_key {
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001264 struct hlist_node node;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001265 u8 keylen;
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001266 u8 family; /* AF_INET or AF_INET6 */
1267 union tcp_md5_addr addr;
1268 u8 key[TCP_MD5SIG_MAXKEYLEN];
1269 struct rcu_head rcu;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001270};
1271
1272/* - sock block */
1273struct tcp_md5sig_info {
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001274 struct hlist_head head;
Eric Dumazeta8afca02012-01-31 18:45:40 +00001275 struct rcu_head rcu;
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001276};
1277
1278/* - pseudo header */
1279struct tcp4_pseudohdr {
1280 __be32 saddr;
1281 __be32 daddr;
1282 __u8 pad;
1283 __u8 protocol;
1284 __be16 len;
1285};
1286
1287struct tcp6_pseudohdr {
1288 struct in6_addr saddr;
1289 struct in6_addr daddr;
1290 __be32 len;
1291 __be32 protocol; /* including padding */
1292};
1293
1294union tcp_md5sum_block {
1295 struct tcp4_pseudohdr ip4;
Eric Dumazetdfd56b82011-12-10 09:48:31 +00001296#if IS_ENABLED(CONFIG_IPV6)
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001297 struct tcp6_pseudohdr ip6;
1298#endif
1299};
1300
1301/* - pool: digest algorithm, hash description and scratch buffer */
1302struct tcp_md5sig_pool {
1303 struct hash_desc md5_desc;
1304 union tcp_md5sum_block md5_blk;
1305};
1306
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001307/* - functions */
Changli Gao53d31762010-07-10 20:41:06 +00001308extern int tcp_v4_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
Eric Dumazet318cf7a2011-10-24 02:46:04 -04001309 const struct sock *sk,
1310 const struct request_sock *req,
1311 const struct sk_buff *skb);
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001312extern int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
1313 int family, const u8 *newkey,
1314 u8 newkeylen, gfp_t gfp);
1315extern int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr,
1316 int family);
1317extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1318 struct sock *addr_sk);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001319
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +09001320#ifdef CONFIG_TCP_MD5SIG
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001321extern struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1322 const union tcp_md5_addr *addr, int family);
1323#define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_key)
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +09001324#else
Eric Dumazeta915da9b2012-01-31 05:18:33 +00001325static inline struct tcp_md5sig_key *tcp_md5_do_lookup(struct sock *sk,
1326 const union tcp_md5_addr *addr,
1327 int family)
1328{
1329 return NULL;
1330}
YOSHIFUJI Hideaki9501f972008-04-18 12:45:16 +09001331#define tcp_twsk_md5_key(twsk) NULL
1332#endif
1333
Eric Dumazet765cf992011-09-12 20:28:37 +00001334extern struct tcp_md5sig_pool __percpu *tcp_alloc_md5sig_pool(struct sock *);
Changli Gao53d31762010-07-10 20:41:06 +00001335extern void tcp_free_md5sig_pool(void);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001336
Eric Dumazet35790c02010-05-16 00:34:04 -07001337extern struct tcp_md5sig_pool *tcp_get_md5sig_pool(void);
Changli Gao53d31762010-07-10 20:41:06 +00001338extern void tcp_put_md5sig_pool(void);
Eric Dumazet35790c02010-05-16 00:34:04 -07001339
Eric Dumazetca35a0e2011-10-24 01:52:35 -04001340extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, const struct tcphdr *);
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001341extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *,
Eric Dumazet95c96172012-04-15 05:58:06 +00001342 unsigned int header_len);
Adam Langley49a72df2008-07-19 00:01:42 -07001343extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001344 const struct tcp_md5sig_key *key);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001345
Jerry Chu10467162012-08-31 12:29:11 +00001346/* From tcp_fastopen.c */
1347extern void tcp_fastopen_cache_get(struct sock *sk, u16 *mss,
1348 struct tcp_fastopen_cookie *cookie,
1349 int *syn_loss, unsigned long *last_syn_loss);
1350extern void tcp_fastopen_cache_set(struct sock *sk, u16 mss,
1351 struct tcp_fastopen_cookie *cookie,
1352 bool syn_lost);
Yuchung Cheng783237e2012-07-19 06:43:07 +00001353struct tcp_fastopen_request {
1354 /* Fast Open cookie. Size 0 means a cookie request */
1355 struct tcp_fastopen_cookie cookie;
1356 struct msghdr *data; /* data in MSG_FASTOPEN */
1357 u16 copied; /* queued in tcp_connect() */
1358};
Yuchung Cheng783237e2012-07-19 06:43:07 +00001359void tcp_free_fastopen_req(struct tcp_sock *tp);
1360
Jerry Chu10467162012-08-31 12:29:11 +00001361extern struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
1362int tcp_fastopen_reset_cipher(void *key, unsigned int len);
1363void tcp_fastopen_cookie_gen(__be32 addr, struct tcp_fastopen_cookie *foc);
1364
1365#define TCP_FASTOPEN_KEY_LENGTH 16
1366
1367/* Fastopen key context */
1368struct tcp_fastopen_context {
1369 struct crypto_cipher __rcu *tfm;
1370 __u8 key[TCP_FASTOPEN_KEY_LENGTH];
1371 struct rcu_head rcu;
1372};
1373
David S. Millerfe067e82007-03-07 12:12:44 -08001374/* write queue abstraction */
1375static inline void tcp_write_queue_purge(struct sock *sk)
1376{
1377 struct sk_buff *skb;
1378
1379 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001380 sk_wmem_free_skb(sk, skb);
1381 sk_mem_reclaim(sk);
Ilpo Järvinen8818a9d2009-12-02 22:24:02 -08001382 tcp_clear_all_retrans_hints(tcp_sk(sk));
David S. Millerfe067e82007-03-07 12:12:44 -08001383}
1384
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001385static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk)
David S. Millerfe067e82007-03-07 12:12:44 -08001386{
David S. Millercd07a8e2008-09-23 00:50:13 -07001387 return skb_peek(&sk->sk_write_queue);
David S. Millerfe067e82007-03-07 12:12:44 -08001388}
1389
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001390static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk)
David S. Millerfe067e82007-03-07 12:12:44 -08001391{
David S. Millercd07a8e2008-09-23 00:50:13 -07001392 return skb_peek_tail(&sk->sk_write_queue);
David S. Millerfe067e82007-03-07 12:12:44 -08001393}
1394
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001395static inline struct sk_buff *tcp_write_queue_next(const struct sock *sk,
1396 const struct sk_buff *skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001397{
David S. Millercd07a8e2008-09-23 00:50:13 -07001398 return skb_queue_next(&sk->sk_write_queue, skb);
David S. Millerfe067e82007-03-07 12:12:44 -08001399}
1400
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001401static inline struct sk_buff *tcp_write_queue_prev(const struct sock *sk,
1402 const struct sk_buff *skb)
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001403{
1404 return skb_queue_prev(&sk->sk_write_queue, skb);
1405}
1406
David S. Millerfe067e82007-03-07 12:12:44 -08001407#define tcp_for_write_queue(skb, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001408 skb_queue_walk(&(sk)->sk_write_queue, skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001409
1410#define tcp_for_write_queue_from(skb, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001411 skb_queue_walk_from(&(sk)->sk_write_queue, skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001412
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001413#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
David S. Millercd07a8e2008-09-23 00:50:13 -07001414 skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
Ilpo Järvinen234b6862007-12-02 00:48:02 +02001415
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001416static inline struct sk_buff *tcp_send_head(const struct sock *sk)
David S. Millerfe067e82007-03-07 12:12:44 -08001417{
1418 return sk->sk_send_head;
1419}
1420
David S. Millercd07a8e2008-09-23 00:50:13 -07001421static inline bool tcp_skb_is_last(const struct sock *sk,
1422 const struct sk_buff *skb)
1423{
1424 return skb_queue_is_last(&sk->sk_write_queue, skb);
1425}
1426
Eric Dumazetcf533ea2011-10-21 05:22:42 -04001427static inline void tcp_advance_send_head(struct sock *sk, const struct sk_buff *skb)
David S. Millerfe067e82007-03-07 12:12:44 -08001428{
David S. Millercd07a8e2008-09-23 00:50:13 -07001429 if (tcp_skb_is_last(sk, skb))
David S. Millerfe067e82007-03-07 12:12:44 -08001430 sk->sk_send_head = NULL;
David S. Millercd07a8e2008-09-23 00:50:13 -07001431 else
1432 sk->sk_send_head = tcp_write_queue_next(sk, skb);
David S. Millerfe067e82007-03-07 12:12:44 -08001433}
1434
1435static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1436{
1437 if (sk->sk_send_head == skb_unlinked)
1438 sk->sk_send_head = NULL;
1439}
1440
1441static inline void tcp_init_send_head(struct sock *sk)
1442{
1443 sk->sk_send_head = NULL;
1444}
1445
1446static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1447{
1448 __skb_queue_tail(&sk->sk_write_queue, skb);
1449}
1450
1451static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1452{
1453 __tcp_add_write_queue_tail(sk, skb);
1454
1455 /* Queue it, remembering where we must start sending. */
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001456 if (sk->sk_send_head == NULL) {
David S. Millerfe067e82007-03-07 12:12:44 -08001457 sk->sk_send_head = skb;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001458
1459 if (tcp_sk(sk)->highest_sack == NULL)
1460 tcp_sk(sk)->highest_sack = skb;
1461 }
David S. Millerfe067e82007-03-07 12:12:44 -08001462}
1463
1464static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1465{
1466 __skb_queue_head(&sk->sk_write_queue, skb);
1467}
1468
1469/* Insert buff after skb on the write queue of sk. */
1470static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1471 struct sk_buff *buff,
1472 struct sock *sk)
1473{
Gerrit Renker7de6c032008-04-14 00:05:09 -07001474 __skb_queue_after(&sk->sk_write_queue, skb, buff);
David S. Millerfe067e82007-03-07 12:12:44 -08001475}
1476
David S. Miller43f59c82008-09-21 21:28:51 -07001477/* Insert new before skb on the write queue of sk. */
David S. Millerfe067e82007-03-07 12:12:44 -08001478static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1479 struct sk_buff *skb,
1480 struct sock *sk)
1481{
David S. Miller43f59c82008-09-21 21:28:51 -07001482 __skb_queue_before(&sk->sk_write_queue, skb, new);
Ilpo Järvinen6e421412007-11-19 23:24:09 -08001483
1484 if (sk->sk_send_head == skb)
1485 sk->sk_send_head = new;
David S. Millerfe067e82007-03-07 12:12:44 -08001486}
1487
1488static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1489{
1490 __skb_unlink(skb, &sk->sk_write_queue);
1491}
1492
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001493static inline bool tcp_write_queue_empty(struct sock *sk)
David S. Millerfe067e82007-03-07 12:12:44 -08001494{
1495 return skb_queue_empty(&sk->sk_write_queue);
1496}
1497
Krishna Kumar12d50c42009-12-08 22:26:13 +00001498static inline void tcp_push_pending_frames(struct sock *sk)
1499{
1500 if (tcp_send_head(sk)) {
1501 struct tcp_sock *tp = tcp_sk(sk);
1502
1503 __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle);
1504 }
1505}
1506
Neal Cardwellecb97192012-02-27 17:52:52 -05001507/* Start sequence of the skb just after the highest skb with SACKed
1508 * bit, valid only if sacked_out > 0 or when the caller has ensured
1509 * validity by itself.
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001510 */
1511static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1512{
1513 if (!tp->sacked_out)
1514 return tp->snd_una;
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001515
1516 if (tp->highest_sack == NULL)
1517 return tp->snd_nxt;
1518
Ilpo Järvinena47e5a92007-11-15 19:41:46 -08001519 return TCP_SKB_CB(tp->highest_sack)->seq;
1520}
1521
Ilpo Järvinen6859d492007-12-02 00:48:06 +02001522static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1523{
1524 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1525 tcp_write_queue_next(sk, skb);
1526}
1527
1528static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1529{
1530 return tcp_sk(sk)->highest_sack;
1531}
1532
1533static inline void tcp_highest_sack_reset(struct sock *sk)
1534{
1535 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1536}
1537
1538/* Called when old skb is about to be deleted (to be combined with new skb) */
1539static inline void tcp_highest_sack_combine(struct sock *sk,
1540 struct sk_buff *old,
1541 struct sk_buff *new)
1542{
1543 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1544 tcp_sk(sk)->highest_sack = new;
1545}
1546
Andreas Petlund5aa4b322010-02-18 02:45:45 +00001547/* Determines whether this is a thin stream (which may suffer from
1548 * increased latency). Used to trigger latency-reducing mechanisms.
1549 */
Eric Dumazeta2a385d2012-05-16 23:15:34 +00001550static inline bool tcp_stream_is_thin(struct tcp_sock *tp)
Andreas Petlund5aa4b322010-02-18 02:45:45 +00001551{
1552 return tp->packets_out < 4 && !tcp_in_initial_slowstart(tp);
1553}
1554
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555/* /proc */
1556enum tcp_seq_states {
1557 TCP_SEQ_STATE_LISTENING,
1558 TCP_SEQ_STATE_OPENREQ,
1559 TCP_SEQ_STATE_ESTABLISHED,
1560 TCP_SEQ_STATE_TIME_WAIT,
1561};
1562
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00001563int tcp_seq_open(struct inode *inode, struct file *file);
1564
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565struct tcp_seq_afinfo {
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00001566 char *name;
1567 sa_family_t family;
1568 const struct file_operations *seq_fops;
1569 struct seq_operations seq_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570};
1571
1572struct tcp_iter_state {
Denis V. Luneva4146b12008-04-13 22:11:14 -07001573 struct seq_net_private p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 sa_family_t family;
1575 enum tcp_seq_states state;
1576 struct sock *syn_wait_sk;
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06001577 int bucket, offset, sbucket, num;
1578 kuid_t uid;
Tom Herberta8b690f2010-06-07 00:43:42 -07001579 loff_t last_pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580};
1581
Daniel Lezcano6f8b13b2008-03-21 04:14:45 -07001582extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1583extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001585extern struct request_sock_ops tcp_request_sock_ops;
Glenn Griffinc6aefaf2008-02-07 21:49:26 -08001586extern struct request_sock_ops tcp6_request_sock_ops;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001587
Brian Haley7d06b2e2008-06-14 17:04:49 -07001588extern void tcp_v4_destroy_sock(struct sock *sk);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001589
Herbert Xua430a432006-07-08 13:34:56 -07001590extern int tcp_v4_gso_send_check(struct sk_buff *skb);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001591extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb,
1592 netdev_features_t features);
Herbert Xubf296b12008-12-15 23:43:36 -08001593extern struct sk_buff **tcp_gro_receive(struct sk_buff **head,
1594 struct sk_buff *skb);
1595extern struct sk_buff **tcp4_gro_receive(struct sk_buff **head,
1596 struct sk_buff *skb);
1597extern int tcp_gro_complete(struct sk_buff *skb);
1598extern int tcp4_gro_complete(struct sk_buff *skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001599
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001600#ifdef CONFIG_PROC_FS
Changli Gao53d31762010-07-10 20:41:06 +00001601extern int tcp4_proc_init(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001602extern void tcp4_proc_exit(void);
1603#endif
1604
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001605/* TCP af-specific functions */
1606struct tcp_sock_af_ops {
1607#ifdef CONFIG_TCP_MD5SIG
1608 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1609 struct sock *addr_sk);
1610 int (*calc_md5_hash) (char *location,
1611 struct tcp_md5sig_key *md5,
Eric Dumazet318cf7a2011-10-24 02:46:04 -04001612 const struct sock *sk,
1613 const struct request_sock *req,
1614 const struct sk_buff *skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001615 int (*md5_parse) (struct sock *sk,
1616 char __user *optval,
1617 int optlen);
1618#endif
1619};
1620
1621struct tcp_request_sock_ops {
1622#ifdef CONFIG_TCP_MD5SIG
1623 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1624 struct request_sock *req);
John Dykstrae3afe7b2009-07-16 05:04:51 +00001625 int (*calc_md5_hash) (char *location,
1626 struct tcp_md5sig_key *md5,
Eric Dumazet318cf7a2011-10-24 02:46:04 -04001627 const struct sock *sk,
1628 const struct request_sock *req,
1629 const struct sk_buff *skb);
YOSHIFUJI Hideakicfb6eeb2006-11-14 19:07:45 -08001630#endif
1631};
1632
William Allen Simpsonda5c78c2009-12-02 18:12:09 +00001633/* Using SHA1 for now, define some constants.
1634 */
1635#define COOKIE_DIGEST_WORDS (SHA_DIGEST_WORDS)
1636#define COOKIE_MESSAGE_WORDS (SHA_MESSAGE_BYTES / 4)
1637#define COOKIE_WORKSPACE_WORDS (COOKIE_DIGEST_WORDS + COOKIE_MESSAGE_WORDS)
1638
1639extern int tcp_cookie_generator(u32 *bakery);
1640
William Allen Simpson435cf552009-12-02 18:17:05 +00001641/**
1642 * struct tcp_cookie_values - each socket needs extra space for the
1643 * cookies, together with (optional) space for any SYN data.
1644 *
1645 * A tcp_sock contains a pointer to the current value, and this is
1646 * cloned to the tcp_timewait_sock.
1647 *
1648 * @cookie_pair: variable data from the option exchange.
1649 *
1650 * @cookie_desired: user specified tcpct_cookie_desired. Zero
1651 * indicates default (sysctl_tcp_cookie_size).
1652 * After cookie sent, remembers size of cookie.
1653 * Range 0, TCP_COOKIE_MIN to TCP_COOKIE_MAX.
1654 *
1655 * @s_data_desired: user specified tcpct_s_data_desired. When the
1656 * constant payload is specified (@s_data_constant),
1657 * holds its length instead.
1658 * Range 0 to TCP_MSS_DESIRED.
1659 *
1660 * @s_data_payload: constant data that is to be included in the
1661 * payload of SYN or SYNACK segments when the
1662 * cookie option is present.
1663 */
1664struct tcp_cookie_values {
1665 struct kref kref;
1666 u8 cookie_pair[TCP_COOKIE_PAIR_SIZE];
1667 u8 cookie_pair_size;
1668 u8 cookie_desired;
1669 u16 s_data_desired:11,
1670 s_data_constant:1,
1671 s_data_in:1,
1672 s_data_out:1,
1673 s_data_unused:2;
1674 u8 s_data_payload[0];
1675};
1676
1677static inline void tcp_cookie_values_release(struct kref *kref)
1678{
1679 kfree(container_of(kref, struct tcp_cookie_values, kref));
1680}
1681
1682/* The length of constant payload data. Note that s_data_desired is
1683 * overloaded, depending on s_data_constant: either the length of constant
1684 * data (returned here) or the limit on variable data.
1685 */
1686static inline int tcp_s_data_size(const struct tcp_sock *tp)
1687{
1688 return (tp->cookie_values != NULL && tp->cookie_values->s_data_constant)
1689 ? tp->cookie_values->s_data_desired
1690 : 0;
1691}
1692
1693/**
1694 * struct tcp_extend_values - tcp_ipv?.c to tcp_output.c workspace.
1695 *
1696 * As tcp_request_sock has already been extended in other places, the
1697 * only remaining method is to pass stack values along as function
1698 * parameters. These parameters are not needed after sending SYNACK.
1699 *
1700 * @cookie_bakery: cryptographic secret and message workspace.
1701 *
1702 * @cookie_plus: bytes in authenticator/cookie option, copied from
1703 * struct tcp_options_received (above).
1704 */
1705struct tcp_extend_values {
1706 struct request_values rv;
1707 u32 cookie_bakery[COOKIE_WORKSPACE_WORDS];
1708 u8 cookie_plus:6,
1709 cookie_out_never:1,
1710 cookie_in_always:1;
1711};
1712
1713static inline struct tcp_extend_values *tcp_xv(struct request_values *rvp)
1714{
1715 return (struct tcp_extend_values *)rvp;
1716}
1717
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001718extern void tcp_v4_init(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001719extern void tcp_init(void);
1720
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721#endif /* _TCP_H */