]> bbs.cooldavid.org Git - net-next-2.6.git/blame - include/net/tcp.h
[NET]: Annotate __skb_checksum_complete() and friends.
[net-next-2.6.git] / include / net / tcp.h
CommitLineData
1da177e4
LT
1/*
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 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 *
13 * This program is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU General Public License
15 * as published by the Free Software Foundation; either version
16 * 2 of the License, or (at your option) any later version.
17 */
18#ifndef _TCP_H
19#define _TCP_H
20
21#define TCP_DEBUG 1
22#define FASTRETRANS_DEBUG 1
23
1da177e4
LT
24#include <linux/list.h>
25#include <linux/tcp.h>
26#include <linux/slab.h>
27#include <linux/cache.h>
28#include <linux/percpu.h>
fb286bb2 29#include <linux/skbuff.h>
97fc2f08 30#include <linux/dmaengine.h>
cfb6eeb4 31#include <linux/crypto.h>
3f421baa
ACM
32
33#include <net/inet_connection_sock.h>
295ff7ed 34#include <net/inet_timewait_sock.h>
77d8bf9c 35#include <net/inet_hashtables.h>
1da177e4 36#include <net/checksum.h>
2e6599cb 37#include <net/request_sock.h>
1da177e4
LT
38#include <net/sock.h>
39#include <net/snmp.h>
40#include <net/ip.h>
c752f073
ACM
41#include <net/tcp_states.h>
42
1da177e4
LT
43#include <linux/seq_file.h>
44
6e04e021 45extern struct inet_hashinfo tcp_hashinfo;
1da177e4 46
1da177e4 47extern atomic_t tcp_orphan_count;
1da177e4 48extern void tcp_time_wait(struct sock *sk, int state, int timeo);
1da177e4 49
1da177e4
LT
50#define MAX_TCP_HEADER (128 + MAX_HEADER)
51
52/*
53 * Never offer a window over 32767 without using window scaling. Some
54 * poor stacks do signed 16bit maths!
55 */
56#define MAX_TCP_WINDOW 32767U
57
58/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
59#define TCP_MIN_MSS 88U
60
61/* Minimal RCV_MSS. */
62#define TCP_MIN_RCVMSS 536U
63
5d424d5a
JH
64/* The least MTU to use for probing */
65#define TCP_BASE_MSS 512
66
1da177e4
LT
67/* After receiving this amount of duplicate ACKs fast retransmit starts. */
68#define TCP_FASTRETRANS_THRESH 3
69
70/* Maximal reordering. */
71#define TCP_MAX_REORDERING 127
72
73/* Maximal number of ACKs sent quickly to accelerate slow-start. */
74#define TCP_MAX_QUICKACKS 16U
75
76/* urg_data states */
77#define TCP_URG_VALID 0x0100
78#define TCP_URG_NOTYET 0x0200
79#define TCP_URG_READ 0x0400
80
81#define TCP_RETR1 3 /*
82 * This is how many retries it does before it
83 * tries to figure out if the gateway is
84 * down. Minimal RFC value is 3; it corresponds
85 * to ~3sec-8min depending on RTO.
86 */
87
88#define TCP_RETR2 15 /*
89 * This should take at least
90 * 90 minutes to time out.
91 * RFC1122 says that the limit is 100 sec.
92 * 15 is ~13-30min depending on RTO.
93 */
94
95#define TCP_SYN_RETRIES 5 /* number of times to retry active opening a
caa20d9a 96 * connection: ~180sec is RFC minimum */
1da177e4
LT
97
98#define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
caa20d9a 99 * connection: ~180sec is RFC minimum */
1da177e4
LT
100
101
102#define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned
103 * socket. 7 is ~50sec-16min.
104 */
105
106
107#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
108 * state, about 60 seconds */
109#define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
110 /* BSD style FIN_WAIT2 deadlock breaker.
111 * It used to be 3min, new value is 60sec,
112 * to combine FIN-WAIT-2 timeout with
113 * TIME-WAIT timer.
114 */
115
116#define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
117#if HZ >= 100
118#define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
119#define TCP_ATO_MIN ((unsigned)(HZ/25))
120#else
121#define TCP_DELACK_MIN 4U
122#define TCP_ATO_MIN 4U
123#endif
124#define TCP_RTO_MAX ((unsigned)(120*HZ))
125#define TCP_RTO_MIN ((unsigned)(HZ/5))
126#define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */
127
128#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
129 * for local resources.
130 */
131
132#define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
133#define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
134#define TCP_KEEPALIVE_INTVL (75*HZ)
135
136#define MAX_TCP_KEEPIDLE 32767
137#define MAX_TCP_KEEPINTVL 32767
138#define MAX_TCP_KEEPCNT 127
139#define MAX_TCP_SYNCNT 127
140
141#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
1da177e4
LT
142
143#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
144#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
145 * after this time. It should be equal
146 * (or greater than) TCP_TIMEWAIT_LEN
147 * to provide reliability equal to one
148 * provided by timewait state.
149 */
150#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
151 * timestamps. It must be less than
152 * minimal timewait lifetime.
153 */
1da177e4
LT
154/*
155 * TCP option
156 */
157
158#define TCPOPT_NOP 1 /* Padding */
159#define TCPOPT_EOL 0 /* End of options */
160#define TCPOPT_MSS 2 /* Segment size negotiating */
161#define TCPOPT_WINDOW 3 /* Window scaling */
162#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
163#define TCPOPT_SACK 5 /* SACK Block */
164#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
cfb6eeb4 165#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
1da177e4
LT
166
167/*
168 * TCP option lengths
169 */
170
171#define TCPOLEN_MSS 4
172#define TCPOLEN_WINDOW 3
173#define TCPOLEN_SACK_PERM 2
174#define TCPOLEN_TIMESTAMP 10
cfb6eeb4 175#define TCPOLEN_MD5SIG 18
1da177e4
LT
176
177/* But this is what stacks really send out. */
178#define TCPOLEN_TSTAMP_ALIGNED 12
179#define TCPOLEN_WSCALE_ALIGNED 4
180#define TCPOLEN_SACKPERM_ALIGNED 4
181#define TCPOLEN_SACK_BASE 2
182#define TCPOLEN_SACK_BASE_ALIGNED 4
183#define TCPOLEN_SACK_PERBLOCK 8
cfb6eeb4 184#define TCPOLEN_MD5SIG_ALIGNED 20
1da177e4 185
1da177e4
LT
186/* Flags in tp->nonagle */
187#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
188#define TCP_NAGLE_CORK 2 /* Socket is corked */
caa20d9a 189#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
1da177e4 190
295ff7ed
ACM
191extern struct inet_timewait_death_row tcp_death_row;
192
1da177e4 193/* sysctl variables for tcp */
1da177e4
LT
194extern int sysctl_tcp_timestamps;
195extern int sysctl_tcp_window_scaling;
196extern int sysctl_tcp_sack;
197extern int sysctl_tcp_fin_timeout;
1da177e4
LT
198extern int sysctl_tcp_keepalive_time;
199extern int sysctl_tcp_keepalive_probes;
200extern int sysctl_tcp_keepalive_intvl;
201extern int sysctl_tcp_syn_retries;
202extern int sysctl_tcp_synack_retries;
203extern int sysctl_tcp_retries1;
204extern int sysctl_tcp_retries2;
205extern int sysctl_tcp_orphan_retries;
206extern int sysctl_tcp_syncookies;
207extern int sysctl_tcp_retrans_collapse;
208extern int sysctl_tcp_stdurg;
209extern int sysctl_tcp_rfc1337;
210extern int sysctl_tcp_abort_on_overflow;
211extern int sysctl_tcp_max_orphans;
1da177e4
LT
212extern int sysctl_tcp_fack;
213extern int sysctl_tcp_reordering;
214extern int sysctl_tcp_ecn;
215extern int sysctl_tcp_dsack;
216extern int sysctl_tcp_mem[3];
217extern int sysctl_tcp_wmem[3];
218extern int sysctl_tcp_rmem[3];
219extern int sysctl_tcp_app_win;
220extern int sysctl_tcp_adv_win_scale;
221extern int sysctl_tcp_tw_reuse;
222extern int sysctl_tcp_frto;
223extern int sysctl_tcp_low_latency;
95937825 224extern int sysctl_tcp_dma_copybreak;
1da177e4 225extern int sysctl_tcp_nometrics_save;
1da177e4
LT
226extern int sysctl_tcp_moderate_rcvbuf;
227extern int sysctl_tcp_tso_win_divisor;
9772efb9 228extern int sysctl_tcp_abc;
5d424d5a
JH
229extern int sysctl_tcp_mtu_probing;
230extern int sysctl_tcp_base_mss;
15d99e02 231extern int sysctl_tcp_workaround_signed_windows;
35089bb2 232extern int sysctl_tcp_slow_start_after_idle;
1da177e4
LT
233
234extern atomic_t tcp_memory_allocated;
235extern atomic_t tcp_sockets_allocated;
236extern int tcp_memory_pressure;
237
1da177e4
LT
238/*
239 * The next routines deal with comparing 32 bit unsigned ints
240 * and worry about wraparound (automatic with unsigned arithmetic).
241 */
242
243static inline int before(__u32 seq1, __u32 seq2)
244{
245 return (__s32)(seq1-seq2) < 0;
246}
247
248static inline int after(__u32 seq1, __u32 seq2)
249{
250 return (__s32)(seq2-seq1) < 0;
251}
252
253
254/* is s2<=s1<=s3 ? */
255static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
256{
257 return seq3 - seq2 >= seq1 - seq2;
258}
259
260
261extern struct proto tcp_prot;
262
263DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
264#define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field)
265#define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field)
266#define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field)
267#define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field)
268#define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val)
269#define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val)
270
1da177e4
LT
271extern void tcp_v4_err(struct sk_buff *skb, u32);
272
273extern void tcp_shutdown (struct sock *sk, int how);
274
275extern int tcp_v4_rcv(struct sk_buff *skb);
276
277extern int tcp_v4_remember_stamp(struct sock *sk);
278
8feaf0c0 279extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
1da177e4
LT
280
281extern int tcp_sendmsg(struct kiocb *iocb, struct sock *sk,
282 struct msghdr *msg, size_t size);
283extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
284
285extern int tcp_ioctl(struct sock *sk,
286 int cmd,
287 unsigned long arg);
288
289extern int tcp_rcv_state_process(struct sock *sk,
290 struct sk_buff *skb,
291 struct tcphdr *th,
292 unsigned len);
293
294extern int tcp_rcv_established(struct sock *sk,
295 struct sk_buff *skb,
296 struct tcphdr *th,
297 unsigned len);
298
299extern void tcp_rcv_space_adjust(struct sock *sk);
300
0e4b4992
CL
301extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
302
6d6ee43e
ACM
303extern int tcp_twsk_unique(struct sock *sk,
304 struct sock *sktw, void *twp);
305
cfb6eeb4
YH
306extern void tcp_twsk_destructor(struct sock *sk);
307
463c84b9
ACM
308static inline void tcp_dec_quickack_mode(struct sock *sk,
309 const unsigned int pkts)
1da177e4 310{
463c84b9 311 struct inet_connection_sock *icsk = inet_csk(sk);
fc6415bc 312
463c84b9
ACM
313 if (icsk->icsk_ack.quick) {
314 if (pkts >= icsk->icsk_ack.quick) {
315 icsk->icsk_ack.quick = 0;
fc6415bc 316 /* Leaving quickack mode we deflate ATO. */
463c84b9 317 icsk->icsk_ack.ato = TCP_ATO_MIN;
fc6415bc 318 } else
463c84b9 319 icsk->icsk_ack.quick -= pkts;
1da177e4
LT
320 }
321}
322
463c84b9 323extern void tcp_enter_quickack_mode(struct sock *sk);
1da177e4 324
1da177e4
LT
325static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
326{
327 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
328}
329
330enum tcp_tw_status
331{
332 TCP_TW_SUCCESS = 0,
333 TCP_TW_RST = 1,
334 TCP_TW_ACK = 2,
335 TCP_TW_SYN = 3
336};
337
338
8feaf0c0 339extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
1da177e4 340 struct sk_buff *skb,
8feaf0c0 341 const struct tcphdr *th);
1da177e4
LT
342
343extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
60236fdd
ACM
344 struct request_sock *req,
345 struct request_sock **prev);
1da177e4
LT
346extern int tcp_child_process(struct sock *parent,
347 struct sock *child,
348 struct sk_buff *skb);
349extern void tcp_enter_frto(struct sock *sk);
350extern void tcp_enter_loss(struct sock *sk, int how);
351extern void tcp_clear_retrans(struct tcp_sock *tp);
352extern void tcp_update_metrics(struct sock *sk);
353
354extern void tcp_close(struct sock *sk,
355 long timeout);
1da177e4
LT
356extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
357
358extern int tcp_getsockopt(struct sock *sk, int level,
359 int optname,
360 char __user *optval,
361 int __user *optlen);
362extern int tcp_setsockopt(struct sock *sk, int level,
363 int optname, char __user *optval,
364 int optlen);
3fdadf7d
DM
365extern int compat_tcp_getsockopt(struct sock *sk,
366 int level, int optname,
367 char __user *optval, int __user *optlen);
368extern int compat_tcp_setsockopt(struct sock *sk,
369 int level, int optname,
370 char __user *optval, int optlen);
1da177e4
LT
371extern void tcp_set_keepalive(struct sock *sk, int val);
372extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
373 struct msghdr *msg,
374 size_t len, int nonblock,
375 int flags, int *addr_len);
376
1da177e4
LT
377extern void tcp_parse_options(struct sk_buff *skb,
378 struct tcp_options_received *opt_rx,
379 int estab);
380
381/*
382 * TCP v4 functions exported for the inet6 API
383 */
384
8292a17a 385extern void tcp_v4_send_check(struct sock *sk, int len,
1da177e4
LT
386 struct sk_buff *skb);
387
388extern int tcp_v4_conn_request(struct sock *sk,
389 struct sk_buff *skb);
390
391extern struct sock * tcp_create_openreq_child(struct sock *sk,
60236fdd 392 struct request_sock *req,
1da177e4
LT
393 struct sk_buff *skb);
394
395extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
396 struct sk_buff *skb,
60236fdd 397 struct request_sock *req,
1da177e4
LT
398 struct dst_entry *dst);
399
400extern int tcp_v4_do_rcv(struct sock *sk,
401 struct sk_buff *skb);
402
403extern int tcp_v4_connect(struct sock *sk,
404 struct sockaddr *uaddr,
405 int addr_len);
406
407extern int tcp_connect(struct sock *sk);
408
409extern struct sk_buff * tcp_make_synack(struct sock *sk,
410 struct dst_entry *dst,
60236fdd 411 struct request_sock *req);
1da177e4
LT
412
413extern int tcp_disconnect(struct sock *sk, int flags);
414
415extern void tcp_unhash(struct sock *sk);
416
1da177e4
LT
417/* From syncookies.c */
418extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
419 struct ip_options *opt);
420extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
421 __u16 *mss);
422
423/* tcp_output.c */
424
f6302d1d 425extern void __tcp_push_pending_frames(struct sock *sk, struct tcp_sock *tp,
a2e2a59c 426 unsigned int cur_mss, int nonagle);
f6302d1d 427extern int tcp_may_send_now(struct sock *sk, struct tcp_sock *tp);
1da177e4
LT
428extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
429extern void tcp_xmit_retransmit_queue(struct sock *);
430extern void tcp_simple_retransmit(struct sock *);
431extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
6475be16 432extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
1da177e4
LT
433
434extern void tcp_send_probe0(struct sock *);
435extern void tcp_send_partial(struct sock *);
436extern int tcp_write_wakeup(struct sock *);
437extern void tcp_send_fin(struct sock *sk);
dd0fc66f 438extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
1da177e4 439extern int tcp_send_synack(struct sock *);
c1b4a7e6 440extern void tcp_push_one(struct sock *, unsigned int mss_now);
1da177e4
LT
441extern void tcp_send_ack(struct sock *sk);
442extern void tcp_send_delayed_ack(struct sock *sk);
443
a762a980
DM
444/* tcp_input.c */
445extern void tcp_cwnd_application_limited(struct sock *sk);
446
1da177e4
LT
447/* tcp_timer.c */
448extern void tcp_init_xmit_timers(struct sock *);
463c84b9
ACM
449static inline void tcp_clear_xmit_timers(struct sock *sk)
450{
451 inet_csk_clear_xmit_timers(sk);
452}
1da177e4 453
1da177e4
LT
454extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
455extern unsigned int tcp_current_mss(struct sock *sk, int large);
456
17b085ea 457/* tcp.c */
1da177e4
LT
458extern void tcp_get_info(struct sock *, struct tcp_info *);
459
460/* Read 'sendfile()'-style from a TCP socket */
461typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
462 unsigned int, size_t);
463extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
464 sk_read_actor_t recv_actor);
465
40efc6fa 466extern void tcp_initialize_rcv_mss(struct sock *sk);
1da177e4 467
5d424d5a
JH
468extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
469extern int tcp_mss_to_mtu(struct sock *sk, int mss);
470extern void tcp_mtup_init(struct sock *sk);
471
40efc6fa 472static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
1da177e4
LT
473{
474 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
475 ntohl(TCP_FLAG_ACK) |
476 snd_wnd);
477}
478
40efc6fa 479static inline void tcp_fast_path_on(struct tcp_sock *tp)
1da177e4
LT
480{
481 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
482}
483
484static inline void tcp_fast_path_check(struct sock *sk, struct tcp_sock *tp)
485{
b03efcfb 486 if (skb_queue_empty(&tp->out_of_order_queue) &&
1da177e4
LT
487 tp->rcv_wnd &&
488 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
489 !tp->urg_data)
490 tcp_fast_path_on(tp);
491}
492
493/* Compute the actual receive window we are currently advertising.
494 * Rcv_nxt can be after the window if our peer push more data
495 * than the offered window.
496 */
40efc6fa 497static inline u32 tcp_receive_window(const struct tcp_sock *tp)
1da177e4
LT
498{
499 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
500
501 if (win < 0)
502 win = 0;
503 return (u32) win;
504}
505
506/* Choose a new window, without checks for shrinking, and without
507 * scaling applied to the result. The caller does these things
508 * if necessary. This is a "raw" window selection.
509 */
510extern u32 __tcp_select_window(struct sock *sk);
511
512/* TCP timestamps are only 32-bits, this causes a slight
513 * complication on 64-bit systems since we store a snapshot
31f34269
SH
514 * of jiffies in the buffer control blocks below. We decided
515 * to use only the low 32-bits of jiffies and hide the ugly
1da177e4
LT
516 * casts with the following macro.
517 */
518#define tcp_time_stamp ((__u32)(jiffies))
519
caa20d9a 520/* This is what the send packet queuing engine uses to pass
1da177e4
LT
521 * TCP per-packet control information to the transmission
522 * code. We also store the host-order sequence numbers in
523 * here too. This is 36 bytes on 32-bit architectures,
524 * 40 bytes on 64-bit machines, if this grows please adjust
525 * skbuff.h:skbuff->cb[xxx] size appropriately.
526 */
527struct tcp_skb_cb {
528 union {
529 struct inet_skb_parm h4;
530#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
531 struct inet6_skb_parm h6;
532#endif
533 } header; /* For incoming frames */
534 __u32 seq; /* Starting sequence number */
535 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
536 __u32 when; /* used to compute rtt's */
537 __u8 flags; /* TCP header flags. */
538
539 /* NOTE: These must match up to the flags byte in a
540 * real TCP header.
541 */
542#define TCPCB_FLAG_FIN 0x01
543#define TCPCB_FLAG_SYN 0x02
544#define TCPCB_FLAG_RST 0x04
545#define TCPCB_FLAG_PSH 0x08
546#define TCPCB_FLAG_ACK 0x10
547#define TCPCB_FLAG_URG 0x20
548#define TCPCB_FLAG_ECE 0x40
549#define TCPCB_FLAG_CWR 0x80
550
551 __u8 sacked; /* State flags for SACK/FACK. */
552#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
553#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
554#define TCPCB_LOST 0x04 /* SKB is lost */
555#define TCPCB_TAGBITS 0x07 /* All tag bits */
556
557#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
558#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
559
caa20d9a 560#define TCPCB_URG 0x20 /* Urgent pointer advanced here */
1da177e4
LT
561
562#define TCPCB_AT_TAIL (TCPCB_URG)
563
564 __u16 urg_ptr; /* Valid w/URG flags is set. */
565 __u32 ack_seq; /* Sequence number ACK'd */
566};
567
568#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
569
570#include <net/tcp_ecn.h>
571
572/* Due to TSO, an SKB can be composed of multiple actual
573 * packets. To keep these tracked properly, we use this.
574 */
575static inline int tcp_skb_pcount(const struct sk_buff *skb)
576{
7967168c 577 return skb_shinfo(skb)->gso_segs;
1da177e4
LT
578}
579
580/* This is valid iff tcp_skb_pcount() > 1. */
581static inline int tcp_skb_mss(const struct sk_buff *skb)
582{
7967168c 583 return skb_shinfo(skb)->gso_size;
1da177e4
LT
584}
585
586static inline void tcp_dec_pcount_approx(__u32 *count,
587 const struct sk_buff *skb)
588{
589 if (*count) {
590 *count -= tcp_skb_pcount(skb);
591 if ((int)*count < 0)
592 *count = 0;
593 }
594}
595
596static inline void tcp_packets_out_inc(struct sock *sk,
597 struct tcp_sock *tp,
598 const struct sk_buff *skb)
599{
600 int orig = tp->packets_out;
601
602 tp->packets_out += tcp_skb_pcount(skb);
603 if (!orig)
3f421baa
ACM
604 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
605 inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
1da177e4
LT
606}
607
608static inline void tcp_packets_out_dec(struct tcp_sock *tp,
609 const struct sk_buff *skb)
610{
611 tp->packets_out -= tcp_skb_pcount(skb);
612}
613
317a76f9
SH
614/* Events passed to congestion control interface */
615enum tcp_ca_event {
616 CA_EVENT_TX_START, /* first transmit when no packets in flight */
617 CA_EVENT_CWND_RESTART, /* congestion window restart */
618 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
619 CA_EVENT_FRTO, /* fast recovery timeout */
620 CA_EVENT_LOSS, /* loss timeout */
621 CA_EVENT_FAST_ACK, /* in sequence ack */
622 CA_EVENT_SLOW_ACK, /* other ack */
623};
624
625/*
626 * Interface for adding new TCP congestion control handlers
627 */
628#define TCP_CA_NAME_MAX 16
3ff825b2
SH
629#define TCP_CA_MAX 128
630#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
631
317a76f9
SH
632struct tcp_congestion_ops {
633 struct list_head list;
ce7bc3bf 634 int non_restricted;
317a76f9
SH
635
636 /* initialize private data (optional) */
6687e988 637 void (*init)(struct sock *sk);
317a76f9 638 /* cleanup private data (optional) */
6687e988 639 void (*release)(struct sock *sk);
317a76f9
SH
640
641 /* return slow start threshold (required) */
6687e988 642 u32 (*ssthresh)(struct sock *sk);
317a76f9 643 /* lower bound for congestion window (optional) */
72dc5b92 644 u32 (*min_cwnd)(const struct sock *sk);
317a76f9 645 /* do new cwnd calculation (required) */
6687e988 646 void (*cong_avoid)(struct sock *sk, u32 ack,
317a76f9
SH
647 u32 rtt, u32 in_flight, int good_ack);
648 /* round trip time sample per acked packet (optional) */
6687e988 649 void (*rtt_sample)(struct sock *sk, u32 usrtt);
317a76f9 650 /* call before changing ca_state (optional) */
6687e988 651 void (*set_state)(struct sock *sk, u8 new_state);
317a76f9 652 /* call when cwnd event occurs (optional) */
6687e988 653 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
317a76f9 654 /* new value of cwnd after loss (optional) */
6687e988 655 u32 (*undo_cwnd)(struct sock *sk);
317a76f9 656 /* hook for packet ack accounting (optional) */
6687e988 657 void (*pkts_acked)(struct sock *sk, u32 num_acked);
73c1f4a0 658 /* get info for inet_diag (optional) */
6687e988 659 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
317a76f9
SH
660
661 char name[TCP_CA_NAME_MAX];
662 struct module *owner;
663};
664
665extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
666extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
667
6687e988
ACM
668extern void tcp_init_congestion_control(struct sock *sk);
669extern void tcp_cleanup_congestion_control(struct sock *sk);
317a76f9
SH
670extern int tcp_set_default_congestion_control(const char *name);
671extern void tcp_get_default_congestion_control(char *name);
3ff825b2 672extern void tcp_get_available_congestion_control(char *buf, size_t len);
ce7bc3bf
SH
673extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
674extern int tcp_set_allowed_congestion_control(char *allowed);
6687e988 675extern int tcp_set_congestion_control(struct sock *sk, const char *name);
40efc6fa 676extern void tcp_slow_start(struct tcp_sock *tp);
317a76f9 677
5f8ef48d 678extern struct tcp_congestion_ops tcp_init_congestion_ops;
6687e988
ACM
679extern u32 tcp_reno_ssthresh(struct sock *sk);
680extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack,
317a76f9 681 u32 rtt, u32 in_flight, int flag);
72dc5b92 682extern u32 tcp_reno_min_cwnd(const struct sock *sk);
a8acfbac 683extern struct tcp_congestion_ops tcp_reno;
317a76f9 684
6687e988 685static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
317a76f9 686{
6687e988
ACM
687 struct inet_connection_sock *icsk = inet_csk(sk);
688
689 if (icsk->icsk_ca_ops->set_state)
690 icsk->icsk_ca_ops->set_state(sk, ca_state);
691 icsk->icsk_ca_state = ca_state;
317a76f9
SH
692}
693
6687e988 694static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
317a76f9 695{
6687e988
ACM
696 const struct inet_connection_sock *icsk = inet_csk(sk);
697
698 if (icsk->icsk_ca_ops->cwnd_event)
699 icsk->icsk_ca_ops->cwnd_event(sk, event);
317a76f9
SH
700}
701
1da177e4
LT
702/* This determines how many packets are "in the network" to the best
703 * of our knowledge. In many cases it is conservative, but where
704 * detailed information is available from the receiver (via SACK
705 * blocks etc.) we can make more aggressive calculations.
706 *
707 * Use this for decisions involving congestion control, use just
708 * tp->packets_out to determine if the send queue is empty or not.
709 *
710 * Read this equation as:
711 *
712 * "Packets sent once on transmission queue" MINUS
713 * "Packets left network, but not honestly ACKed yet" PLUS
714 * "Packets fast retransmitted"
715 */
40efc6fa 716static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
1da177e4
LT
717{
718 return (tp->packets_out - tp->left_out + tp->retrans_out);
719}
720
1da177e4
LT
721/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
722 * The exception is rate halving phase, when cwnd is decreasing towards
723 * ssthresh.
724 */
6687e988 725static inline __u32 tcp_current_ssthresh(const struct sock *sk)
1da177e4 726{
6687e988
ACM
727 const struct tcp_sock *tp = tcp_sk(sk);
728 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
1da177e4
LT
729 return tp->snd_ssthresh;
730 else
731 return max(tp->snd_ssthresh,
732 ((tp->snd_cwnd >> 1) +
733 (tp->snd_cwnd >> 2)));
734}
735
736static inline void tcp_sync_left_out(struct tcp_sock *tp)
737{
738 if (tp->rx_opt.sack_ok &&
739 (tp->sacked_out >= tp->packets_out - tp->lost_out))
740 tp->sacked_out = tp->packets_out - tp->lost_out;
741 tp->left_out = tp->sacked_out + tp->lost_out;
742}
743
40efc6fa 744extern void tcp_enter_cwr(struct sock *sk);
1da177e4
LT
745extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
746
747/* Slow start with delack produces 3 packets of burst, so that
748 * it is safe "de facto".
749 */
750static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
751{
752 return 3;
753}
754
f4805ede
SH
755/* RFC2861 Check whether we are limited by application or congestion window
756 * This is the inverse of cwnd check in tcp_tso_should_defer
757 */
758static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
759{
760 const struct tcp_sock *tp = tcp_sk(sk);
761 u32 left;
762
763 if (in_flight >= tp->snd_cwnd)
764 return 1;
765
bcd76111 766 if (!sk_can_gso(sk))
f4805ede
SH
767 return 0;
768
769 left = tp->snd_cwnd - in_flight;
770 if (sysctl_tcp_tso_win_divisor)
771 return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd;
772 else
773 return left <= tcp_max_burst(tp);
774}
775
40efc6fa
SH
776static inline void tcp_minshall_update(struct tcp_sock *tp, int mss,
777 const struct sk_buff *skb)
1da177e4
LT
778{
779 if (skb->len < mss)
780 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
781}
782
40efc6fa 783static inline void tcp_check_probe_timer(struct sock *sk, struct tcp_sock *tp)
1da177e4 784{
463c84b9
ACM
785 const struct inet_connection_sock *icsk = inet_csk(sk);
786 if (!tp->packets_out && !icsk->icsk_pending)
3f421baa
ACM
787 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
788 icsk->icsk_rto, TCP_RTO_MAX);
1da177e4
LT
789}
790
40efc6fa
SH
791static inline void tcp_push_pending_frames(struct sock *sk,
792 struct tcp_sock *tp)
1da177e4
LT
793{
794 __tcp_push_pending_frames(sk, tp, tcp_current_mss(sk, 1), tp->nonagle);
795}
796
40efc6fa 797static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT
798{
799 tp->snd_wl1 = seq;
800}
801
40efc6fa 802static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT
803{
804 tp->snd_wl1 = seq;
805}
806
1da177e4
LT
807/*
808 * Calculate(/check) TCP checksum
809 */
6b11687e
AV
810static inline __sum16 tcp_v4_check(struct tcphdr *th, int len,
811 __be32 saddr, __be32 daddr,
812 __wsum base)
1da177e4
LT
813{
814 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
815}
816
b51655b9 817static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1da177e4 818{
fb286bb2 819 return __skb_checksum_complete(skb);
1da177e4
LT
820}
821
40efc6fa 822static inline int tcp_checksum_complete(struct sk_buff *skb)
1da177e4
LT
823{
824 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
825 __tcp_checksum_complete(skb);
826}
827
828/* Prequeue for VJ style copy to user, combined with checksumming. */
829
40efc6fa 830static inline void tcp_prequeue_init(struct tcp_sock *tp)
1da177e4
LT
831{
832 tp->ucopy.task = NULL;
833 tp->ucopy.len = 0;
834 tp->ucopy.memory = 0;
835 skb_queue_head_init(&tp->ucopy.prequeue);
97fc2f08
CL
836#ifdef CONFIG_NET_DMA
837 tp->ucopy.dma_chan = NULL;
838 tp->ucopy.wakeup = 0;
839 tp->ucopy.pinned_list = NULL;
840 tp->ucopy.dma_cookie = 0;
841#endif
1da177e4
LT
842}
843
844/* Packet is added to VJ-style prequeue for processing in process
845 * context, if a reader task is waiting. Apparently, this exciting
846 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
847 * failed somewhere. Latency? Burstiness? Well, at least now we will
848 * see, why it failed. 8)8) --ANK
849 *
850 * NOTE: is this not too big to inline?
851 */
40efc6fa 852static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
1da177e4
LT
853{
854 struct tcp_sock *tp = tcp_sk(sk);
855
856 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
857 __skb_queue_tail(&tp->ucopy.prequeue, skb);
858 tp->ucopy.memory += skb->truesize;
859 if (tp->ucopy.memory > sk->sk_rcvbuf) {
860 struct sk_buff *skb1;
861
862 BUG_ON(sock_owned_by_user(sk));
863
864 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
865 sk->sk_backlog_rcv(sk, skb1);
866 NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED);
867 }
868
869 tp->ucopy.memory = 0;
870 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
871 wake_up_interruptible(sk->sk_sleep);
463c84b9
ACM
872 if (!inet_csk_ack_scheduled(sk))
873 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
3f421baa
ACM
874 (3 * TCP_RTO_MIN) / 4,
875 TCP_RTO_MAX);
1da177e4
LT
876 }
877 return 1;
878 }
879 return 0;
880}
881
882
883#undef STATE_TRACE
884
885#ifdef STATE_TRACE
886static const char *statename[]={
887 "Unused","Established","Syn Sent","Syn Recv",
888 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
889 "Close Wait","Last ACK","Listen","Closing"
890};
891#endif
892
40efc6fa 893static inline void tcp_set_state(struct sock *sk, int state)
1da177e4
LT
894{
895 int oldstate = sk->sk_state;
896
897 switch (state) {
898 case TCP_ESTABLISHED:
899 if (oldstate != TCP_ESTABLISHED)
900 TCP_INC_STATS(TCP_MIB_CURRESTAB);
901 break;
902
903 case TCP_CLOSE:
904 if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED)
905 TCP_INC_STATS(TCP_MIB_ESTABRESETS);
906
907 sk->sk_prot->unhash(sk);
463c84b9 908 if (inet_csk(sk)->icsk_bind_hash &&
1da177e4 909 !(sk->sk_userlocks & SOCK_BINDPORT_LOCK))
2d8c4ce5 910 inet_put_port(&tcp_hashinfo, sk);
1da177e4
LT
911 /* fall through */
912 default:
913 if (oldstate==TCP_ESTABLISHED)
914 TCP_DEC_STATS(TCP_MIB_CURRESTAB);
915 }
916
917 /* Change state AFTER socket is unhashed to avoid closed
918 * socket sitting in hash tables.
919 */
920 sk->sk_state = state;
921
922#ifdef STATE_TRACE
923 SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n",sk, statename[oldstate],statename[state]);
924#endif
925}
926
40efc6fa 927static inline void tcp_done(struct sock *sk)
1da177e4 928{
3687b1dc
WY
929 if(sk->sk_state == TCP_SYN_SENT || sk->sk_state == TCP_SYN_RECV)
930 TCP_INC_STATS_BH(TCP_MIB_ATTEMPTFAILS);
931
1da177e4
LT
932 tcp_set_state(sk, TCP_CLOSE);
933 tcp_clear_xmit_timers(sk);
934
935 sk->sk_shutdown = SHUTDOWN_MASK;
936
937 if (!sock_flag(sk, SOCK_DEAD))
938 sk->sk_state_change(sk);
939 else
0a5578cf 940 inet_csk_destroy_sock(sk);
1da177e4
LT
941}
942
40efc6fa 943static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1da177e4
LT
944{
945 rx_opt->dsack = 0;
946 rx_opt->eff_sacks = 0;
947 rx_opt->num_sacks = 0;
948}
949
1da177e4
LT
950/* Determine a window scaling and initial window to offer. */
951extern void tcp_select_initial_window(int __space, __u32 mss,
952 __u32 *rcv_wnd, __u32 *window_clamp,
953 int wscale_ok, __u8 *rcv_wscale);
954
955static inline int tcp_win_from_space(int space)
956{
957 return sysctl_tcp_adv_win_scale<=0 ?
958 (space>>(-sysctl_tcp_adv_win_scale)) :
959 space - (space>>sysctl_tcp_adv_win_scale);
960}
961
962/* Note: caller must be prepared to deal with negative returns */
963static inline int tcp_space(const struct sock *sk)
964{
965 return tcp_win_from_space(sk->sk_rcvbuf -
966 atomic_read(&sk->sk_rmem_alloc));
967}
968
969static inline int tcp_full_space(const struct sock *sk)
970{
971 return tcp_win_from_space(sk->sk_rcvbuf);
972}
973
40efc6fa
SH
974static inline void tcp_openreq_init(struct request_sock *req,
975 struct tcp_options_received *rx_opt,
976 struct sk_buff *skb)
1da177e4 977{
2e6599cb
ACM
978 struct inet_request_sock *ireq = inet_rsk(req);
979
1da177e4 980 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
2e6599cb 981 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1da177e4
LT
982 req->mss = rx_opt->mss_clamp;
983 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
2e6599cb
ACM
984 ireq->tstamp_ok = rx_opt->tstamp_ok;
985 ireq->sack_ok = rx_opt->sack_ok;
986 ireq->snd_wscale = rx_opt->snd_wscale;
987 ireq->wscale_ok = rx_opt->wscale_ok;
988 ireq->acked = 0;
989 ireq->ecn_ok = 0;
990 ireq->rmt_port = skb->h.th->source;
1da177e4
LT
991}
992
993extern void tcp_enter_memory_pressure(void);
994
1da177e4
LT
995static inline int keepalive_intvl_when(const struct tcp_sock *tp)
996{
997 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
998}
999
1000static inline int keepalive_time_when(const struct tcp_sock *tp)
1001{
1002 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1003}
1004
463c84b9 1005static inline int tcp_fin_time(const struct sock *sk)
1da177e4 1006{
463c84b9
ACM
1007 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1008 const int rto = inet_csk(sk)->icsk_rto;
1da177e4 1009
463c84b9
ACM
1010 if (fin_timeout < (rto << 2) - (rto >> 1))
1011 fin_timeout = (rto << 2) - (rto >> 1);
1da177e4
LT
1012
1013 return fin_timeout;
1014}
1015
1016static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1017{
1018 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1019 return 0;
1020 if (xtime.tv_sec >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
1021 return 0;
1022
1023 /* RST segments are not recommended to carry timestamp,
1024 and, if they do, it is recommended to ignore PAWS because
1025 "their cleanup function should take precedence over timestamps."
1026 Certainly, it is mistake. It is necessary to understand the reasons
1027 of this constraint to relax it: if peer reboots, clock may go
1028 out-of-sync and half-open connections will not be reset.
1029 Actually, the problem would be not existing if all
1030 the implementations followed draft about maintaining clock
1031 via reboots. Linux-2.2 DOES NOT!
1032
1033 However, we can relax time bounds for RST segments to MSL.
1034 */
1035 if (rst && xtime.tv_sec >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1036 return 0;
1037 return 1;
1038}
1039
1da177e4
LT
1040#define TCP_CHECK_TIMER(sk) do { } while (0)
1041
1042static inline int tcp_use_frto(const struct sock *sk)
1043{
1044 const struct tcp_sock *tp = tcp_sk(sk);
1045
1046 /* F-RTO must be activated in sysctl and there must be some
1047 * unsent new data, and the advertised window should allow
1048 * sending it.
1049 */
1050 return (sysctl_tcp_frto && sk->sk_send_head &&
1051 !after(TCP_SKB_CB(sk->sk_send_head)->end_seq,
1052 tp->snd_una + tp->snd_wnd));
1053}
1054
1055static inline void tcp_mib_init(void)
1056{
1057 /* See RFC 2012 */
1058 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1059 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1060 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1061 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1062}
1063
6a438bbe
SH
1064/*from STCP */
1065static inline void clear_all_retrans_hints(struct tcp_sock *tp){
1066 tp->lost_skb_hint = NULL;
1067 tp->scoreboard_skb_hint = NULL;
1068 tp->retransmit_skb_hint = NULL;
1069 tp->forward_skb_hint = NULL;
1070 tp->fastpath_skb_hint = NULL;
1071}
1072
cfb6eeb4
YH
1073/* MD5 Signature */
1074struct crypto_hash;
1075
1076/* - key database */
1077struct tcp_md5sig_key {
1078 u8 *key;
1079 u8 keylen;
1080};
1081
1082struct tcp4_md5sig_key {
1083 u8 *key;
1084 u16 keylen;
1085 __be32 addr;
1086};
1087
1088struct tcp6_md5sig_key {
1089 u8 *key;
1090 u16 keylen;
1091#if 0
1092 u32 scope_id; /* XXX */
1093#endif
1094 struct in6_addr addr;
1095};
1096
1097/* - sock block */
1098struct tcp_md5sig_info {
1099 struct tcp4_md5sig_key *keys4;
1100#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1101 struct tcp6_md5sig_key *keys6;
1102 u32 entries6;
1103 u32 alloced6;
1104#endif
1105 u32 entries4;
1106 u32 alloced4;
1107};
1108
1109/* - pseudo header */
1110struct tcp4_pseudohdr {
1111 __be32 saddr;
1112 __be32 daddr;
1113 __u8 pad;
1114 __u8 protocol;
1115 __be16 len;
1116};
1117
1118struct tcp6_pseudohdr {
1119 struct in6_addr saddr;
1120 struct in6_addr daddr;
1121 __be32 len;
1122 __be32 protocol; /* including padding */
1123};
1124
1125union tcp_md5sum_block {
1126 struct tcp4_pseudohdr ip4;
1127#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1128 struct tcp6_pseudohdr ip6;
1129#endif
1130};
1131
1132/* - pool: digest algorithm, hash description and scratch buffer */
1133struct tcp_md5sig_pool {
1134 struct hash_desc md5_desc;
1135 union tcp_md5sum_block md5_blk;
1136};
1137
1138#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1139
1140/* - functions */
1141extern int tcp_v4_calc_md5_hash(char *md5_hash,
1142 struct tcp_md5sig_key *key,
1143 struct sock *sk,
1144 struct dst_entry *dst,
1145 struct request_sock *req,
1146 struct tcphdr *th,
1147 int protocol, int tcplen);
1148extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1149 struct sock *addr_sk);
1150
1151extern int tcp_v4_md5_do_add(struct sock *sk,
1152 __be32 addr,
1153 u8 *newkey,
1154 u8 newkeylen);
1155
1156extern int tcp_v4_md5_do_del(struct sock *sk,
1157 u32 addr);
1158
1159extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1160extern void tcp_free_md5sig_pool(void);
1161
1162extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1163extern void __tcp_put_md5sig_pool(void);
1164
1165static inline
1166struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1167{
1168 int cpu = get_cpu();
1169 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1170 if (!ret)
1171 put_cpu();
1172 return ret;
1173}
1174
1175static inline void tcp_put_md5sig_pool(void)
1176{
1177 __tcp_put_md5sig_pool();
1178 put_cpu();
1179}
1180
1da177e4
LT
1181/* /proc */
1182enum tcp_seq_states {
1183 TCP_SEQ_STATE_LISTENING,
1184 TCP_SEQ_STATE_OPENREQ,
1185 TCP_SEQ_STATE_ESTABLISHED,
1186 TCP_SEQ_STATE_TIME_WAIT,
1187};
1188
1189struct tcp_seq_afinfo {
1190 struct module *owner;
1191 char *name;
1192 sa_family_t family;
1193 int (*seq_show) (struct seq_file *m, void *v);
1194 struct file_operations *seq_fops;
1195};
1196
1197struct tcp_iter_state {
1198 sa_family_t family;
1199 enum tcp_seq_states state;
1200 struct sock *syn_wait_sk;
1201 int bucket, sbucket, num, uid;
1202 struct seq_operations seq_ops;
1203};
1204
1205extern int tcp_proc_register(struct tcp_seq_afinfo *afinfo);
1206extern void tcp_proc_unregister(struct tcp_seq_afinfo *afinfo);
1207
20380731
ACM
1208extern struct request_sock_ops tcp_request_sock_ops;
1209
1210extern int tcp_v4_destroy_sock(struct sock *sk);
1211
a430a43d 1212extern int tcp_v4_gso_send_check(struct sk_buff *skb);
576a30eb 1213extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
f4c50d99 1214
20380731
ACM
1215#ifdef CONFIG_PROC_FS
1216extern int tcp4_proc_init(void);
1217extern void tcp4_proc_exit(void);
1218#endif
1219
cfb6eeb4
YH
1220/* TCP af-specific functions */
1221struct tcp_sock_af_ops {
1222#ifdef CONFIG_TCP_MD5SIG
1223 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1224 struct sock *addr_sk);
1225 int (*calc_md5_hash) (char *location,
1226 struct tcp_md5sig_key *md5,
1227 struct sock *sk,
1228 struct dst_entry *dst,
1229 struct request_sock *req,
1230 struct tcphdr *th,
1231 int protocol, int len);
1232 int (*md5_add) (struct sock *sk,
1233 struct sock *addr_sk,
1234 u8 *newkey,
1235 u8 len);
1236 int (*md5_parse) (struct sock *sk,
1237 char __user *optval,
1238 int optlen);
1239#endif
1240};
1241
1242struct tcp_request_sock_ops {
1243#ifdef CONFIG_TCP_MD5SIG
1244 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1245 struct request_sock *req);
1246#endif
1247};
1248
20380731
ACM
1249extern void tcp_v4_init(struct net_proto_family *ops);
1250extern void tcp_init(void);
1251
1da177e4 1252#endif /* _TCP_H */