[NET]: Fix comments for register_netdev().
[GitHub/mt8127/android_kernel_alcatel_ttab.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;
3cfe3baa 223extern int sysctl_tcp_frto_response;
1da177e4 224extern int sysctl_tcp_low_latency;
95937825 225extern int sysctl_tcp_dma_copybreak;
1da177e4 226extern int sysctl_tcp_nometrics_save;
1da177e4
LT
227extern int sysctl_tcp_moderate_rcvbuf;
228extern int sysctl_tcp_tso_win_divisor;
9772efb9 229extern int sysctl_tcp_abc;
5d424d5a
JH
230extern int sysctl_tcp_mtu_probing;
231extern int sysctl_tcp_base_mss;
15d99e02 232extern int sysctl_tcp_workaround_signed_windows;
35089bb2 233extern int sysctl_tcp_slow_start_after_idle;
886236c1 234extern int sysctl_tcp_max_ssthresh;
1da177e4
LT
235
236extern atomic_t tcp_memory_allocated;
237extern atomic_t tcp_sockets_allocated;
238extern int tcp_memory_pressure;
239
1da177e4
LT
240/*
241 * The next routines deal with comparing 32 bit unsigned ints
242 * and worry about wraparound (automatic with unsigned arithmetic).
243 */
244
245static inline int before(__u32 seq1, __u32 seq2)
246{
0d630cc0 247 return (__s32)(seq1-seq2) < 0;
1da177e4 248}
9a036b9c 249#define after(seq2, seq1) before(seq1, seq2)
1da177e4
LT
250
251/* is s2<=s1<=s3 ? */
252static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
253{
254 return seq3 - seq2 >= seq1 - seq2;
255}
256
257
258extern struct proto tcp_prot;
259
260DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
261#define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field)
262#define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field)
263#define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field)
264#define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field)
265#define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val)
266#define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val)
267
1da177e4
LT
268extern void tcp_v4_err(struct sk_buff *skb, u32);
269
270extern void tcp_shutdown (struct sock *sk, int how);
271
272extern int tcp_v4_rcv(struct sk_buff *skb);
273
274extern int tcp_v4_remember_stamp(struct sock *sk);
275
8feaf0c0 276extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
1da177e4
LT
277
278extern int tcp_sendmsg(struct kiocb *iocb, struct sock *sk,
279 struct msghdr *msg, size_t size);
280extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
281
282extern int tcp_ioctl(struct sock *sk,
283 int cmd,
284 unsigned long arg);
285
286extern int tcp_rcv_state_process(struct sock *sk,
287 struct sk_buff *skb,
288 struct tcphdr *th,
289 unsigned len);
290
291extern int tcp_rcv_established(struct sock *sk,
292 struct sk_buff *skb,
293 struct tcphdr *th,
294 unsigned len);
295
296extern void tcp_rcv_space_adjust(struct sock *sk);
297
0e4b4992
CL
298extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
299
6d6ee43e
ACM
300extern int tcp_twsk_unique(struct sock *sk,
301 struct sock *sktw, void *twp);
302
cfb6eeb4
YH
303extern void tcp_twsk_destructor(struct sock *sk);
304
463c84b9
ACM
305static inline void tcp_dec_quickack_mode(struct sock *sk,
306 const unsigned int pkts)
1da177e4 307{
463c84b9 308 struct inet_connection_sock *icsk = inet_csk(sk);
fc6415bc 309
463c84b9
ACM
310 if (icsk->icsk_ack.quick) {
311 if (pkts >= icsk->icsk_ack.quick) {
312 icsk->icsk_ack.quick = 0;
fc6415bc 313 /* Leaving quickack mode we deflate ATO. */
463c84b9 314 icsk->icsk_ack.ato = TCP_ATO_MIN;
fc6415bc 315 } else
463c84b9 316 icsk->icsk_ack.quick -= pkts;
1da177e4
LT
317 }
318}
319
463c84b9 320extern void tcp_enter_quickack_mode(struct sock *sk);
1da177e4 321
1da177e4
LT
322static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
323{
324 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
325}
326
327enum tcp_tw_status
328{
329 TCP_TW_SUCCESS = 0,
330 TCP_TW_RST = 1,
331 TCP_TW_ACK = 2,
332 TCP_TW_SYN = 3
333};
334
335
8feaf0c0 336extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
1da177e4 337 struct sk_buff *skb,
8feaf0c0 338 const struct tcphdr *th);
1da177e4
LT
339
340extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
60236fdd
ACM
341 struct request_sock *req,
342 struct request_sock **prev);
1da177e4
LT
343extern int tcp_child_process(struct sock *parent,
344 struct sock *child,
345 struct sk_buff *skb);
46d0de4e 346extern int tcp_use_frto(struct sock *sk);
1da177e4
LT
347extern void tcp_enter_frto(struct sock *sk);
348extern void tcp_enter_loss(struct sock *sk, int how);
349extern void tcp_clear_retrans(struct tcp_sock *tp);
350extern void tcp_update_metrics(struct sock *sk);
351
352extern void tcp_close(struct sock *sk,
353 long timeout);
1da177e4
LT
354extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
355
356extern int tcp_getsockopt(struct sock *sk, int level,
357 int optname,
358 char __user *optval,
359 int __user *optlen);
360extern int tcp_setsockopt(struct sock *sk, int level,
361 int optname, char __user *optval,
362 int optlen);
3fdadf7d
DM
363extern int compat_tcp_getsockopt(struct sock *sk,
364 int level, int optname,
365 char __user *optval, int __user *optlen);
366extern int compat_tcp_setsockopt(struct sock *sk,
367 int level, int optname,
368 char __user *optval, int optlen);
1da177e4
LT
369extern void tcp_set_keepalive(struct sock *sk, int val);
370extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
371 struct msghdr *msg,
372 size_t len, int nonblock,
373 int flags, int *addr_len);
374
1da177e4
LT
375extern void tcp_parse_options(struct sk_buff *skb,
376 struct tcp_options_received *opt_rx,
377 int estab);
378
379/*
380 * TCP v4 functions exported for the inet6 API
381 */
382
8292a17a 383extern void tcp_v4_send_check(struct sock *sk, int len,
1da177e4
LT
384 struct sk_buff *skb);
385
386extern int tcp_v4_conn_request(struct sock *sk,
387 struct sk_buff *skb);
388
389extern struct sock * tcp_create_openreq_child(struct sock *sk,
60236fdd 390 struct request_sock *req,
1da177e4
LT
391 struct sk_buff *skb);
392
393extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
394 struct sk_buff *skb,
60236fdd 395 struct request_sock *req,
1da177e4
LT
396 struct dst_entry *dst);
397
398extern int tcp_v4_do_rcv(struct sock *sk,
399 struct sk_buff *skb);
400
401extern int tcp_v4_connect(struct sock *sk,
402 struct sockaddr *uaddr,
403 int addr_len);
404
405extern int tcp_connect(struct sock *sk);
406
407extern struct sk_buff * tcp_make_synack(struct sock *sk,
408 struct dst_entry *dst,
60236fdd 409 struct request_sock *req);
1da177e4
LT
410
411extern int tcp_disconnect(struct sock *sk, int flags);
412
413extern void tcp_unhash(struct sock *sk);
414
1da177e4
LT
415/* From syncookies.c */
416extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
417 struct ip_options *opt);
418extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
419 __u16 *mss);
420
421/* tcp_output.c */
422
f6302d1d 423extern void __tcp_push_pending_frames(struct sock *sk, struct tcp_sock *tp,
a2e2a59c 424 unsigned int cur_mss, int nonagle);
f6302d1d 425extern int tcp_may_send_now(struct sock *sk, struct tcp_sock *tp);
1da177e4
LT
426extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
427extern void tcp_xmit_retransmit_queue(struct sock *);
428extern void tcp_simple_retransmit(struct sock *);
429extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
6475be16 430extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
1da177e4
LT
431
432extern void tcp_send_probe0(struct sock *);
433extern void tcp_send_partial(struct sock *);
434extern int tcp_write_wakeup(struct sock *);
435extern void tcp_send_fin(struct sock *sk);
dd0fc66f 436extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
1da177e4 437extern int tcp_send_synack(struct sock *);
c1b4a7e6 438extern void tcp_push_one(struct sock *, unsigned int mss_now);
1da177e4
LT
439extern void tcp_send_ack(struct sock *sk);
440extern void tcp_send_delayed_ack(struct sock *sk);
441
a762a980
DM
442/* tcp_input.c */
443extern void tcp_cwnd_application_limited(struct sock *sk);
444
1da177e4
LT
445/* tcp_timer.c */
446extern void tcp_init_xmit_timers(struct sock *);
463c84b9
ACM
447static inline void tcp_clear_xmit_timers(struct sock *sk)
448{
449 inet_csk_clear_xmit_timers(sk);
450}
1da177e4 451
1da177e4
LT
452extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
453extern unsigned int tcp_current_mss(struct sock *sk, int large);
454
17b085ea 455/* tcp.c */
1da177e4
LT
456extern void tcp_get_info(struct sock *, struct tcp_info *);
457
458/* Read 'sendfile()'-style from a TCP socket */
459typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
460 unsigned int, size_t);
461extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
462 sk_read_actor_t recv_actor);
463
40efc6fa 464extern void tcp_initialize_rcv_mss(struct sock *sk);
1da177e4 465
5d424d5a
JH
466extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
467extern int tcp_mss_to_mtu(struct sock *sk, int mss);
468extern void tcp_mtup_init(struct sock *sk);
469
40efc6fa 470static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
1da177e4
LT
471{
472 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
473 ntohl(TCP_FLAG_ACK) |
474 snd_wnd);
475}
476
40efc6fa 477static inline void tcp_fast_path_on(struct tcp_sock *tp)
1da177e4
LT
478{
479 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
480}
481
482static inline void tcp_fast_path_check(struct sock *sk, struct tcp_sock *tp)
483{
b03efcfb 484 if (skb_queue_empty(&tp->out_of_order_queue) &&
1da177e4
LT
485 tp->rcv_wnd &&
486 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
487 !tp->urg_data)
488 tcp_fast_path_on(tp);
489}
490
491/* Compute the actual receive window we are currently advertising.
492 * Rcv_nxt can be after the window if our peer push more data
493 * than the offered window.
494 */
40efc6fa 495static inline u32 tcp_receive_window(const struct tcp_sock *tp)
1da177e4
LT
496{
497 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
498
499 if (win < 0)
500 win = 0;
501 return (u32) win;
502}
503
504/* Choose a new window, without checks for shrinking, and without
505 * scaling applied to the result. The caller does these things
506 * if necessary. This is a "raw" window selection.
507 */
508extern u32 __tcp_select_window(struct sock *sk);
509
510/* TCP timestamps are only 32-bits, this causes a slight
511 * complication on 64-bit systems since we store a snapshot
31f34269
SH
512 * of jiffies in the buffer control blocks below. We decided
513 * to use only the low 32-bits of jiffies and hide the ugly
1da177e4
LT
514 * casts with the following macro.
515 */
516#define tcp_time_stamp ((__u32)(jiffies))
517
caa20d9a 518/* This is what the send packet queuing engine uses to pass
1da177e4
LT
519 * TCP per-packet control information to the transmission
520 * code. We also store the host-order sequence numbers in
521 * here too. This is 36 bytes on 32-bit architectures,
522 * 40 bytes on 64-bit machines, if this grows please adjust
523 * skbuff.h:skbuff->cb[xxx] size appropriately.
524 */
525struct tcp_skb_cb {
526 union {
527 struct inet_skb_parm h4;
528#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
529 struct inet6_skb_parm h6;
530#endif
531 } header; /* For incoming frames */
532 __u32 seq; /* Starting sequence number */
533 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
534 __u32 when; /* used to compute rtt's */
535 __u8 flags; /* TCP header flags. */
536
537 /* NOTE: These must match up to the flags byte in a
538 * real TCP header.
539 */
540#define TCPCB_FLAG_FIN 0x01
541#define TCPCB_FLAG_SYN 0x02
542#define TCPCB_FLAG_RST 0x04
543#define TCPCB_FLAG_PSH 0x08
544#define TCPCB_FLAG_ACK 0x10
545#define TCPCB_FLAG_URG 0x20
546#define TCPCB_FLAG_ECE 0x40
547#define TCPCB_FLAG_CWR 0x80
548
549 __u8 sacked; /* State flags for SACK/FACK. */
550#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
551#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
552#define TCPCB_LOST 0x04 /* SKB is lost */
553#define TCPCB_TAGBITS 0x07 /* All tag bits */
554
555#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
556#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
557
caa20d9a 558#define TCPCB_URG 0x20 /* Urgent pointer advanced here */
1da177e4
LT
559
560#define TCPCB_AT_TAIL (TCPCB_URG)
561
562 __u16 urg_ptr; /* Valid w/URG flags is set. */
563 __u32 ack_seq; /* Sequence number ACK'd */
564};
565
566#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
567
568#include <net/tcp_ecn.h>
569
570/* Due to TSO, an SKB can be composed of multiple actual
571 * packets. To keep these tracked properly, we use this.
572 */
573static inline int tcp_skb_pcount(const struct sk_buff *skb)
574{
7967168c 575 return skb_shinfo(skb)->gso_segs;
1da177e4
LT
576}
577
578/* This is valid iff tcp_skb_pcount() > 1. */
579static inline int tcp_skb_mss(const struct sk_buff *skb)
580{
7967168c 581 return skb_shinfo(skb)->gso_size;
1da177e4
LT
582}
583
584static inline void tcp_dec_pcount_approx(__u32 *count,
585 const struct sk_buff *skb)
586{
587 if (*count) {
588 *count -= tcp_skb_pcount(skb);
589 if ((int)*count < 0)
590 *count = 0;
591 }
592}
593
594static inline void tcp_packets_out_inc(struct sock *sk,
595 struct tcp_sock *tp,
596 const struct sk_buff *skb)
597{
598 int orig = tp->packets_out;
599
600 tp->packets_out += tcp_skb_pcount(skb);
601 if (!orig)
3f421baa
ACM
602 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
603 inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
1da177e4
LT
604}
605
606static inline void tcp_packets_out_dec(struct tcp_sock *tp,
607 const struct sk_buff *skb)
608{
609 tp->packets_out -= tcp_skb_pcount(skb);
610}
611
317a76f9
SH
612/* Events passed to congestion control interface */
613enum tcp_ca_event {
614 CA_EVENT_TX_START, /* first transmit when no packets in flight */
615 CA_EVENT_CWND_RESTART, /* congestion window restart */
616 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
617 CA_EVENT_FRTO, /* fast recovery timeout */
618 CA_EVENT_LOSS, /* loss timeout */
619 CA_EVENT_FAST_ACK, /* in sequence ack */
620 CA_EVENT_SLOW_ACK, /* other ack */
621};
622
623/*
624 * Interface for adding new TCP congestion control handlers
625 */
626#define TCP_CA_NAME_MAX 16
3ff825b2
SH
627#define TCP_CA_MAX 128
628#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
629
317a76f9
SH
630struct tcp_congestion_ops {
631 struct list_head list;
ce7bc3bf 632 int non_restricted;
317a76f9
SH
633
634 /* initialize private data (optional) */
6687e988 635 void (*init)(struct sock *sk);
317a76f9 636 /* cleanup private data (optional) */
6687e988 637 void (*release)(struct sock *sk);
317a76f9
SH
638
639 /* return slow start threshold (required) */
6687e988 640 u32 (*ssthresh)(struct sock *sk);
317a76f9 641 /* lower bound for congestion window (optional) */
72dc5b92 642 u32 (*min_cwnd)(const struct sock *sk);
317a76f9 643 /* do new cwnd calculation (required) */
6687e988 644 void (*cong_avoid)(struct sock *sk, u32 ack,
317a76f9
SH
645 u32 rtt, u32 in_flight, int good_ack);
646 /* round trip time sample per acked packet (optional) */
6687e988 647 void (*rtt_sample)(struct sock *sk, u32 usrtt);
317a76f9 648 /* call before changing ca_state (optional) */
6687e988 649 void (*set_state)(struct sock *sk, u8 new_state);
317a76f9 650 /* call when cwnd event occurs (optional) */
6687e988 651 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
317a76f9 652 /* new value of cwnd after loss (optional) */
6687e988 653 u32 (*undo_cwnd)(struct sock *sk);
317a76f9 654 /* hook for packet ack accounting (optional) */
6687e988 655 void (*pkts_acked)(struct sock *sk, u32 num_acked);
73c1f4a0 656 /* get info for inet_diag (optional) */
6687e988 657 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
317a76f9
SH
658
659 char name[TCP_CA_NAME_MAX];
660 struct module *owner;
661};
662
663extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
664extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
665
6687e988
ACM
666extern void tcp_init_congestion_control(struct sock *sk);
667extern void tcp_cleanup_congestion_control(struct sock *sk);
317a76f9
SH
668extern int tcp_set_default_congestion_control(const char *name);
669extern void tcp_get_default_congestion_control(char *name);
3ff825b2 670extern void tcp_get_available_congestion_control(char *buf, size_t len);
ce7bc3bf
SH
671extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
672extern int tcp_set_allowed_congestion_control(char *allowed);
6687e988 673extern int tcp_set_congestion_control(struct sock *sk, const char *name);
40efc6fa 674extern void tcp_slow_start(struct tcp_sock *tp);
317a76f9 675
5f8ef48d 676extern struct tcp_congestion_ops tcp_init_congestion_ops;
6687e988
ACM
677extern u32 tcp_reno_ssthresh(struct sock *sk);
678extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack,
317a76f9 679 u32 rtt, u32 in_flight, int flag);
72dc5b92 680extern u32 tcp_reno_min_cwnd(const struct sock *sk);
a8acfbac 681extern struct tcp_congestion_ops tcp_reno;
317a76f9 682
6687e988 683static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
317a76f9 684{
6687e988
ACM
685 struct inet_connection_sock *icsk = inet_csk(sk);
686
687 if (icsk->icsk_ca_ops->set_state)
688 icsk->icsk_ca_ops->set_state(sk, ca_state);
689 icsk->icsk_ca_state = ca_state;
317a76f9
SH
690}
691
6687e988 692static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
317a76f9 693{
6687e988
ACM
694 const struct inet_connection_sock *icsk = inet_csk(sk);
695
696 if (icsk->icsk_ca_ops->cwnd_event)
697 icsk->icsk_ca_ops->cwnd_event(sk, event);
317a76f9
SH
698}
699
1da177e4
LT
700/* This determines how many packets are "in the network" to the best
701 * of our knowledge. In many cases it is conservative, but where
702 * detailed information is available from the receiver (via SACK
703 * blocks etc.) we can make more aggressive calculations.
704 *
705 * Use this for decisions involving congestion control, use just
706 * tp->packets_out to determine if the send queue is empty or not.
707 *
708 * Read this equation as:
709 *
710 * "Packets sent once on transmission queue" MINUS
711 * "Packets left network, but not honestly ACKed yet" PLUS
712 * "Packets fast retransmitted"
713 */
40efc6fa 714static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
1da177e4
LT
715{
716 return (tp->packets_out - tp->left_out + tp->retrans_out);
717}
718
1da177e4
LT
719/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
720 * The exception is rate halving phase, when cwnd is decreasing towards
721 * ssthresh.
722 */
6687e988 723static inline __u32 tcp_current_ssthresh(const struct sock *sk)
1da177e4 724{
6687e988
ACM
725 const struct tcp_sock *tp = tcp_sk(sk);
726 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
1da177e4
LT
727 return tp->snd_ssthresh;
728 else
729 return max(tp->snd_ssthresh,
730 ((tp->snd_cwnd >> 1) +
731 (tp->snd_cwnd >> 2)));
732}
733
734static inline void tcp_sync_left_out(struct tcp_sock *tp)
735{
736 if (tp->rx_opt.sack_ok &&
737 (tp->sacked_out >= tp->packets_out - tp->lost_out))
738 tp->sacked_out = tp->packets_out - tp->lost_out;
739 tp->left_out = tp->sacked_out + tp->lost_out;
740}
741
3cfe3baa 742extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
1da177e4
LT
743extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
744
745/* Slow start with delack produces 3 packets of burst, so that
746 * it is safe "de facto".
747 */
748static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
749{
750 return 3;
751}
752
f4805ede
SH
753/* RFC2861 Check whether we are limited by application or congestion window
754 * This is the inverse of cwnd check in tcp_tso_should_defer
755 */
756static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
757{
758 const struct tcp_sock *tp = tcp_sk(sk);
759 u32 left;
760
761 if (in_flight >= tp->snd_cwnd)
762 return 1;
763
bcd76111 764 if (!sk_can_gso(sk))
f4805ede
SH
765 return 0;
766
767 left = tp->snd_cwnd - in_flight;
768 if (sysctl_tcp_tso_win_divisor)
769 return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd;
770 else
771 return left <= tcp_max_burst(tp);
772}
773
40efc6fa
SH
774static inline void tcp_minshall_update(struct tcp_sock *tp, int mss,
775 const struct sk_buff *skb)
1da177e4
LT
776{
777 if (skb->len < mss)
778 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
779}
780
40efc6fa 781static inline void tcp_check_probe_timer(struct sock *sk, struct tcp_sock *tp)
1da177e4 782{
463c84b9
ACM
783 const struct inet_connection_sock *icsk = inet_csk(sk);
784 if (!tp->packets_out && !icsk->icsk_pending)
3f421baa
ACM
785 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
786 icsk->icsk_rto, TCP_RTO_MAX);
1da177e4
LT
787}
788
40efc6fa
SH
789static inline void tcp_push_pending_frames(struct sock *sk,
790 struct tcp_sock *tp)
1da177e4
LT
791{
792 __tcp_push_pending_frames(sk, tp, tcp_current_mss(sk, 1), tp->nonagle);
793}
794
40efc6fa 795static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT
796{
797 tp->snd_wl1 = seq;
798}
799
40efc6fa 800static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT
801{
802 tp->snd_wl1 = seq;
803}
804
1da177e4
LT
805/*
806 * Calculate(/check) TCP checksum
807 */
ba7808ea
FD
808static inline __sum16 tcp_v4_check(int len, __be32 saddr,
809 __be32 daddr, __wsum base)
1da177e4
LT
810{
811 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
812}
813
b51655b9 814static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1da177e4 815{
fb286bb2 816 return __skb_checksum_complete(skb);
1da177e4
LT
817}
818
40efc6fa 819static inline int tcp_checksum_complete(struct sk_buff *skb)
1da177e4 820{
60476372 821 return !skb_csum_unnecessary(skb) &&
1da177e4
LT
822 __tcp_checksum_complete(skb);
823}
824
825/* Prequeue for VJ style copy to user, combined with checksumming. */
826
40efc6fa 827static inline void tcp_prequeue_init(struct tcp_sock *tp)
1da177e4
LT
828{
829 tp->ucopy.task = NULL;
830 tp->ucopy.len = 0;
831 tp->ucopy.memory = 0;
832 skb_queue_head_init(&tp->ucopy.prequeue);
97fc2f08
CL
833#ifdef CONFIG_NET_DMA
834 tp->ucopy.dma_chan = NULL;
835 tp->ucopy.wakeup = 0;
836 tp->ucopy.pinned_list = NULL;
837 tp->ucopy.dma_cookie = 0;
838#endif
1da177e4
LT
839}
840
841/* Packet is added to VJ-style prequeue for processing in process
842 * context, if a reader task is waiting. Apparently, this exciting
843 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
844 * failed somewhere. Latency? Burstiness? Well, at least now we will
845 * see, why it failed. 8)8) --ANK
846 *
847 * NOTE: is this not too big to inline?
848 */
40efc6fa 849static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
1da177e4
LT
850{
851 struct tcp_sock *tp = tcp_sk(sk);
852
853 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
854 __skb_queue_tail(&tp->ucopy.prequeue, skb);
855 tp->ucopy.memory += skb->truesize;
856 if (tp->ucopy.memory > sk->sk_rcvbuf) {
857 struct sk_buff *skb1;
858
859 BUG_ON(sock_owned_by_user(sk));
860
861 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
862 sk->sk_backlog_rcv(sk, skb1);
863 NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED);
864 }
865
866 tp->ucopy.memory = 0;
867 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
868 wake_up_interruptible(sk->sk_sleep);
463c84b9
ACM
869 if (!inet_csk_ack_scheduled(sk))
870 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
3f421baa
ACM
871 (3 * TCP_RTO_MIN) / 4,
872 TCP_RTO_MAX);
1da177e4
LT
873 }
874 return 1;
875 }
876 return 0;
877}
878
879
880#undef STATE_TRACE
881
882#ifdef STATE_TRACE
883static const char *statename[]={
884 "Unused","Established","Syn Sent","Syn Recv",
885 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
886 "Close Wait","Last ACK","Listen","Closing"
887};
888#endif
889
40efc6fa 890static inline void tcp_set_state(struct sock *sk, int state)
1da177e4
LT
891{
892 int oldstate = sk->sk_state;
893
894 switch (state) {
895 case TCP_ESTABLISHED:
896 if (oldstate != TCP_ESTABLISHED)
897 TCP_INC_STATS(TCP_MIB_CURRESTAB);
898 break;
899
900 case TCP_CLOSE:
901 if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED)
902 TCP_INC_STATS(TCP_MIB_ESTABRESETS);
903
904 sk->sk_prot->unhash(sk);
463c84b9 905 if (inet_csk(sk)->icsk_bind_hash &&
1da177e4 906 !(sk->sk_userlocks & SOCK_BINDPORT_LOCK))
2d8c4ce5 907 inet_put_port(&tcp_hashinfo, sk);
1da177e4
LT
908 /* fall through */
909 default:
910 if (oldstate==TCP_ESTABLISHED)
911 TCP_DEC_STATS(TCP_MIB_CURRESTAB);
912 }
913
914 /* Change state AFTER socket is unhashed to avoid closed
915 * socket sitting in hash tables.
916 */
917 sk->sk_state = state;
918
919#ifdef STATE_TRACE
920 SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n",sk, statename[oldstate],statename[state]);
921#endif
922}
923
4ac02bab 924extern void tcp_done(struct sock *sk);
1da177e4 925
40efc6fa 926static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1da177e4
LT
927{
928 rx_opt->dsack = 0;
929 rx_opt->eff_sacks = 0;
930 rx_opt->num_sacks = 0;
931}
932
1da177e4
LT
933/* Determine a window scaling and initial window to offer. */
934extern void tcp_select_initial_window(int __space, __u32 mss,
935 __u32 *rcv_wnd, __u32 *window_clamp,
936 int wscale_ok, __u8 *rcv_wscale);
937
938static inline int tcp_win_from_space(int space)
939{
940 return sysctl_tcp_adv_win_scale<=0 ?
941 (space>>(-sysctl_tcp_adv_win_scale)) :
942 space - (space>>sysctl_tcp_adv_win_scale);
943}
944
945/* Note: caller must be prepared to deal with negative returns */
946static inline int tcp_space(const struct sock *sk)
947{
948 return tcp_win_from_space(sk->sk_rcvbuf -
949 atomic_read(&sk->sk_rmem_alloc));
950}
951
952static inline int tcp_full_space(const struct sock *sk)
953{
954 return tcp_win_from_space(sk->sk_rcvbuf);
955}
956
40efc6fa
SH
957static inline void tcp_openreq_init(struct request_sock *req,
958 struct tcp_options_received *rx_opt,
959 struct sk_buff *skb)
1da177e4 960{
2e6599cb
ACM
961 struct inet_request_sock *ireq = inet_rsk(req);
962
1da177e4 963 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
2e6599cb 964 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1da177e4
LT
965 req->mss = rx_opt->mss_clamp;
966 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
2e6599cb
ACM
967 ireq->tstamp_ok = rx_opt->tstamp_ok;
968 ireq->sack_ok = rx_opt->sack_ok;
969 ireq->snd_wscale = rx_opt->snd_wscale;
970 ireq->wscale_ok = rx_opt->wscale_ok;
971 ireq->acked = 0;
972 ireq->ecn_ok = 0;
aa8223c7 973 ireq->rmt_port = tcp_hdr(skb)->source;
1da177e4
LT
974}
975
976extern void tcp_enter_memory_pressure(void);
977
1da177e4
LT
978static inline int keepalive_intvl_when(const struct tcp_sock *tp)
979{
980 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
981}
982
983static inline int keepalive_time_when(const struct tcp_sock *tp)
984{
985 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
986}
987
463c84b9 988static inline int tcp_fin_time(const struct sock *sk)
1da177e4 989{
463c84b9
ACM
990 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
991 const int rto = inet_csk(sk)->icsk_rto;
1da177e4 992
463c84b9
ACM
993 if (fin_timeout < (rto << 2) - (rto >> 1))
994 fin_timeout = (rto << 2) - (rto >> 1);
1da177e4
LT
995
996 return fin_timeout;
997}
998
999static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1000{
1001 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1002 return 0;
9d729f72 1003 if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
1da177e4
LT
1004 return 0;
1005
1006 /* RST segments are not recommended to carry timestamp,
1007 and, if they do, it is recommended to ignore PAWS because
1008 "their cleanup function should take precedence over timestamps."
1009 Certainly, it is mistake. It is necessary to understand the reasons
1010 of this constraint to relax it: if peer reboots, clock may go
1011 out-of-sync and half-open connections will not be reset.
1012 Actually, the problem would be not existing if all
1013 the implementations followed draft about maintaining clock
1014 via reboots. Linux-2.2 DOES NOT!
1015
1016 However, we can relax time bounds for RST segments to MSL.
1017 */
9d729f72 1018 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1da177e4
LT
1019 return 0;
1020 return 1;
1021}
1022
1da177e4
LT
1023#define TCP_CHECK_TIMER(sk) do { } while (0)
1024
1da177e4
LT
1025static inline void tcp_mib_init(void)
1026{
1027 /* See RFC 2012 */
1028 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1029 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1030 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1031 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1032}
1033
6a438bbe
SH
1034/*from STCP */
1035static inline void clear_all_retrans_hints(struct tcp_sock *tp){
1036 tp->lost_skb_hint = NULL;
1037 tp->scoreboard_skb_hint = NULL;
1038 tp->retransmit_skb_hint = NULL;
1039 tp->forward_skb_hint = NULL;
1040 tp->fastpath_skb_hint = NULL;
1041}
1042
cfb6eeb4
YH
1043/* MD5 Signature */
1044struct crypto_hash;
1045
1046/* - key database */
1047struct tcp_md5sig_key {
1048 u8 *key;
1049 u8 keylen;
1050};
1051
1052struct tcp4_md5sig_key {
1053 u8 *key;
1054 u16 keylen;
1055 __be32 addr;
1056};
1057
1058struct tcp6_md5sig_key {
1059 u8 *key;
1060 u16 keylen;
1061#if 0
1062 u32 scope_id; /* XXX */
1063#endif
1064 struct in6_addr addr;
1065};
1066
1067/* - sock block */
1068struct tcp_md5sig_info {
1069 struct tcp4_md5sig_key *keys4;
1070#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1071 struct tcp6_md5sig_key *keys6;
1072 u32 entries6;
1073 u32 alloced6;
1074#endif
1075 u32 entries4;
1076 u32 alloced4;
1077};
1078
1079/* - pseudo header */
1080struct tcp4_pseudohdr {
1081 __be32 saddr;
1082 __be32 daddr;
1083 __u8 pad;
1084 __u8 protocol;
1085 __be16 len;
1086};
1087
1088struct tcp6_pseudohdr {
1089 struct in6_addr saddr;
1090 struct in6_addr daddr;
1091 __be32 len;
1092 __be32 protocol; /* including padding */
1093};
1094
1095union tcp_md5sum_block {
1096 struct tcp4_pseudohdr ip4;
1097#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1098 struct tcp6_pseudohdr ip6;
1099#endif
1100};
1101
1102/* - pool: digest algorithm, hash description and scratch buffer */
1103struct tcp_md5sig_pool {
1104 struct hash_desc md5_desc;
1105 union tcp_md5sum_block md5_blk;
1106};
1107
1108#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1109
1110/* - functions */
1111extern int tcp_v4_calc_md5_hash(char *md5_hash,
1112 struct tcp_md5sig_key *key,
1113 struct sock *sk,
1114 struct dst_entry *dst,
1115 struct request_sock *req,
1116 struct tcphdr *th,
1117 int protocol, int tcplen);
1118extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1119 struct sock *addr_sk);
1120
1121extern int tcp_v4_md5_do_add(struct sock *sk,
1122 __be32 addr,
1123 u8 *newkey,
1124 u8 newkeylen);
1125
1126extern int tcp_v4_md5_do_del(struct sock *sk,
8e5200f5 1127 __be32 addr);
cfb6eeb4
YH
1128
1129extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1130extern void tcp_free_md5sig_pool(void);
1131
1132extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1133extern void __tcp_put_md5sig_pool(void);
1134
1135static inline
1136struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1137{
1138 int cpu = get_cpu();
1139 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1140 if (!ret)
1141 put_cpu();
1142 return ret;
1143}
1144
1145static inline void tcp_put_md5sig_pool(void)
1146{
1147 __tcp_put_md5sig_pool();
1148 put_cpu();
1149}
1150
fe067e8a
DM
1151/* write queue abstraction */
1152static inline void tcp_write_queue_purge(struct sock *sk)
1153{
1154 struct sk_buff *skb;
1155
1156 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
1157 sk_stream_free_skb(sk, skb);
1158 sk_stream_mem_reclaim(sk);
1159}
1160
1161static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1162{
1163 struct sk_buff *skb = sk->sk_write_queue.next;
1164 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1165 return NULL;
1166 return skb;
1167}
1168
1169static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1170{
1171 struct sk_buff *skb = sk->sk_write_queue.prev;
1172 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1173 return NULL;
1174 return skb;
1175}
1176
1177static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1178{
1179 return skb->next;
1180}
1181
1182#define tcp_for_write_queue(skb, sk) \
1183 for (skb = (sk)->sk_write_queue.next; \
1184 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1185 skb = skb->next)
1186
1187#define tcp_for_write_queue_from(skb, sk) \
1188 for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\
1189 skb = skb->next)
1190
1191static inline struct sk_buff *tcp_send_head(struct sock *sk)
1192{
1193 return sk->sk_send_head;
1194}
1195
1196static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1197{
1198 sk->sk_send_head = skb->next;
1199 if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue)
1200 sk->sk_send_head = NULL;
1201}
1202
1203static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1204{
1205 if (sk->sk_send_head == skb_unlinked)
1206 sk->sk_send_head = NULL;
1207}
1208
1209static inline void tcp_init_send_head(struct sock *sk)
1210{
1211 sk->sk_send_head = NULL;
1212}
1213
1214static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1215{
1216 __skb_queue_tail(&sk->sk_write_queue, skb);
1217}
1218
1219static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1220{
1221 __tcp_add_write_queue_tail(sk, skb);
1222
1223 /* Queue it, remembering where we must start sending. */
1224 if (sk->sk_send_head == NULL)
1225 sk->sk_send_head = skb;
1226}
1227
1228static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1229{
1230 __skb_queue_head(&sk->sk_write_queue, skb);
1231}
1232
1233/* Insert buff after skb on the write queue of sk. */
1234static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1235 struct sk_buff *buff,
1236 struct sock *sk)
1237{
1238 __skb_append(skb, buff, &sk->sk_write_queue);
1239}
1240
1241/* Insert skb between prev and next on the write queue of sk. */
1242static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1243 struct sk_buff *skb,
1244 struct sock *sk)
1245{
1246 __skb_insert(new, skb->prev, skb, &sk->sk_write_queue);
1247}
1248
1249static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1250{
1251 __skb_unlink(skb, &sk->sk_write_queue);
1252}
1253
1254static inline int tcp_skb_is_last(const struct sock *sk,
1255 const struct sk_buff *skb)
1256{
1257 return skb->next == (struct sk_buff *)&sk->sk_write_queue;
1258}
1259
1260static inline int tcp_write_queue_empty(struct sock *sk)
1261{
1262 return skb_queue_empty(&sk->sk_write_queue);
1263}
1264
1da177e4
LT
1265/* /proc */
1266enum tcp_seq_states {
1267 TCP_SEQ_STATE_LISTENING,
1268 TCP_SEQ_STATE_OPENREQ,
1269 TCP_SEQ_STATE_ESTABLISHED,
1270 TCP_SEQ_STATE_TIME_WAIT,
1271};
1272
1273struct tcp_seq_afinfo {
1274 struct module *owner;
1275 char *name;
1276 sa_family_t family;
1277 int (*seq_show) (struct seq_file *m, void *v);
1278 struct file_operations *seq_fops;
1279};
1280
1281struct tcp_iter_state {
1282 sa_family_t family;
1283 enum tcp_seq_states state;
1284 struct sock *syn_wait_sk;
1285 int bucket, sbucket, num, uid;
1286 struct seq_operations seq_ops;
1287};
1288
1289extern int tcp_proc_register(struct tcp_seq_afinfo *afinfo);
1290extern void tcp_proc_unregister(struct tcp_seq_afinfo *afinfo);
1291
20380731
ACM
1292extern struct request_sock_ops tcp_request_sock_ops;
1293
1294extern int tcp_v4_destroy_sock(struct sock *sk);
1295
a430a43d 1296extern int tcp_v4_gso_send_check(struct sk_buff *skb);
576a30eb 1297extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
f4c50d99 1298
20380731
ACM
1299#ifdef CONFIG_PROC_FS
1300extern int tcp4_proc_init(void);
1301extern void tcp4_proc_exit(void);
1302#endif
1303
cfb6eeb4
YH
1304/* TCP af-specific functions */
1305struct tcp_sock_af_ops {
1306#ifdef CONFIG_TCP_MD5SIG
1307 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1308 struct sock *addr_sk);
1309 int (*calc_md5_hash) (char *location,
1310 struct tcp_md5sig_key *md5,
1311 struct sock *sk,
1312 struct dst_entry *dst,
1313 struct request_sock *req,
1314 struct tcphdr *th,
1315 int protocol, int len);
1316 int (*md5_add) (struct sock *sk,
1317 struct sock *addr_sk,
1318 u8 *newkey,
1319 u8 len);
1320 int (*md5_parse) (struct sock *sk,
1321 char __user *optval,
1322 int optlen);
1323#endif
1324};
1325
1326struct tcp_request_sock_ops {
1327#ifdef CONFIG_TCP_MD5SIG
1328 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1329 struct request_sock *req);
1330#endif
1331};
1332
20380731
ACM
1333extern void tcp_v4_init(struct net_proto_family *ops);
1334extern void tcp_init(void);
1335
1da177e4 1336#endif /* _TCP_H */