]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/net/tcp.h
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[TCP] YeAH-TCP: algorithm implementation
[mirror_ubuntu-artful-kernel.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{
0d630cc0 245 return (__s32)(seq1-seq2) < 0;
1da177e4 246}
9a036b9c 247#define after(seq2, seq1) before(seq1, seq2)
1da177e4
LT		 248
249/* is s2<=s1<=s3 ? */
250static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
251{
252 return seq3 - seq2 >= seq1 - seq2;
253}
254
255
256extern struct proto tcp_prot;
257
258DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
259#define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field)
260#define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field)
261#define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field)
262#define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field)
263#define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val)
264#define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val)
265
1da177e4
LT		 266extern void tcp_v4_err(struct sk_buff *skb, u32);
267
268extern void tcp_shutdown (struct sock *sk, int how);
269
270extern int tcp_v4_rcv(struct sk_buff *skb);
271
272extern int tcp_v4_remember_stamp(struct sock *sk);
273
8feaf0c0 274extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
1da177e4
LT		 275
276extern int tcp_sendmsg(struct kiocb *iocb, struct sock *sk,
277 struct msghdr *msg, size_t size);
278extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
279
280extern int tcp_ioctl(struct sock *sk,
281 int cmd,
282 unsigned long arg);
283
284extern int tcp_rcv_state_process(struct sock *sk,
285 struct sk_buff *skb,
286 struct tcphdr *th,
287 unsigned len);
288
289extern int tcp_rcv_established(struct sock *sk,
290 struct sk_buff *skb,
291 struct tcphdr *th,
292 unsigned len);
293
294extern void tcp_rcv_space_adjust(struct sock *sk);
295
0e4b4992
CL		 296extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
297
6d6ee43e
ACM		 298extern int tcp_twsk_unique(struct sock *sk,
299 struct sock *sktw, void *twp);
300
cfb6eeb4
YH		 301extern void tcp_twsk_destructor(struct sock *sk);
302
463c84b9
ACM		 303static inline void tcp_dec_quickack_mode(struct sock *sk,
304 const unsigned int pkts)
1da177e4 305{
463c84b9 306 struct inet_connection_sock *icsk = inet_csk(sk);
fc6415bc 307
463c84b9
ACM		 308 if (icsk->icsk_ack.quick) {
309 if (pkts >= icsk->icsk_ack.quick) {
310 icsk->icsk_ack.quick = 0;
fc6415bc 311 /* Leaving quickack mode we deflate ATO. */
463c84b9 312 icsk->icsk_ack.ato = TCP_ATO_MIN;
fc6415bc 313 } else
463c84b9 314 icsk->icsk_ack.quick -= pkts;
1da177e4
LT		 315 }
316}
317
463c84b9 318extern void tcp_enter_quickack_mode(struct sock *sk);
1da177e4 319
1da177e4
LT		 320static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
321{
322 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
323}
324
325enum tcp_tw_status
326{
327 TCP_TW_SUCCESS = 0,
328 TCP_TW_RST = 1,
329 TCP_TW_ACK = 2,
330 TCP_TW_SYN = 3
331};
332
333
8feaf0c0 334extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
1da177e4 335 struct sk_buff *skb,
8feaf0c0 336 const struct tcphdr *th);
1da177e4
LT		 337
338extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
60236fdd
ACM		 339 struct request_sock *req,
340 struct request_sock **prev);
1da177e4
LT		 341extern int tcp_child_process(struct sock *parent,
342 struct sock *child,
343 struct sk_buff *skb);
46d0de4e 344extern int tcp_use_frto(struct sock *sk);
1da177e4
LT		 345extern void tcp_enter_frto(struct sock *sk);
346extern void tcp_enter_loss(struct sock *sk, int how);
347extern void tcp_clear_retrans(struct tcp_sock *tp);
348extern void tcp_update_metrics(struct sock *sk);
349
350extern void tcp_close(struct sock *sk,
351 long timeout);
1da177e4
LT		 352extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
353
354extern int tcp_getsockopt(struct sock *sk, int level,
355 int optname,
356 char __user *optval,
357 int __user *optlen);
358extern int tcp_setsockopt(struct sock *sk, int level,
359 int optname, char __user *optval,
360 int optlen);
3fdadf7d
DM		 361extern int compat_tcp_getsockopt(struct sock *sk,
362 int level, int optname,
363 char __user *optval, int __user *optlen);
364extern int compat_tcp_setsockopt(struct sock *sk,
365 int level, int optname,
366 char __user *optval, int optlen);
1da177e4
LT		 367extern void tcp_set_keepalive(struct sock *sk, int val);
368extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
369 struct msghdr *msg,
370 size_t len, int nonblock,
371 int flags, int *addr_len);
372
1da177e4
LT		 373extern void tcp_parse_options(struct sk_buff *skb,
374 struct tcp_options_received *opt_rx,
375 int estab);
376
377/*
378 * TCP v4 functions exported for the inet6 API
379 */
380
8292a17a 381extern void tcp_v4_send_check(struct sock *sk, int len,
1da177e4
LT		 382 struct sk_buff *skb);
383
384extern int tcp_v4_conn_request(struct sock *sk,
385 struct sk_buff *skb);
386
387extern struct sock * tcp_create_openreq_child(struct sock *sk,
60236fdd 388 struct request_sock *req,
1da177e4
LT		 389 struct sk_buff *skb);
390
391extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
392 struct sk_buff *skb,
60236fdd 393 struct request_sock *req,
1da177e4
LT		 394 struct dst_entry *dst);
395
396extern int tcp_v4_do_rcv(struct sock *sk,
397 struct sk_buff *skb);
398
399extern int tcp_v4_connect(struct sock *sk,
400 struct sockaddr *uaddr,
401 int addr_len);
402
403extern int tcp_connect(struct sock *sk);
404
405extern struct sk_buff * tcp_make_synack(struct sock *sk,
406 struct dst_entry *dst,
60236fdd 407 struct request_sock *req);
1da177e4
LT		 408
409extern int tcp_disconnect(struct sock *sk, int flags);
410
411extern void tcp_unhash(struct sock *sk);
412
1da177e4
LT		 413/* From syncookies.c */
414extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
415 struct ip_options *opt);
416extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
417 __u16 *mss);
418
419/* tcp_output.c */
420
f6302d1d 421extern void __tcp_push_pending_frames(struct sock *sk, struct tcp_sock *tp,
a2e2a59c 422 unsigned int cur_mss, int nonagle);
f6302d1d 423extern int tcp_may_send_now(struct sock *sk, struct tcp_sock *tp);
1da177e4
LT		 424extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
425extern void tcp_xmit_retransmit_queue(struct sock *);
426extern void tcp_simple_retransmit(struct sock *);
427extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
6475be16 428extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
1da177e4
LT		 429
430extern void tcp_send_probe0(struct sock *);
431extern void tcp_send_partial(struct sock *);
432extern int tcp_write_wakeup(struct sock *);
433extern void tcp_send_fin(struct sock *sk);
dd0fc66f 434extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
1da177e4 435extern int tcp_send_synack(struct sock *);
c1b4a7e6 436extern void tcp_push_one(struct sock *, unsigned int mss_now);
1da177e4
LT		 437extern void tcp_send_ack(struct sock *sk);
438extern void tcp_send_delayed_ack(struct sock *sk);
439
a762a980
DM		 440/* tcp_input.c */
441extern void tcp_cwnd_application_limited(struct sock *sk);
442
1da177e4
LT		 443/* tcp_timer.c */
444extern void tcp_init_xmit_timers(struct sock *);
463c84b9
ACM		 445static inline void tcp_clear_xmit_timers(struct sock *sk)
446{
447 inet_csk_clear_xmit_timers(sk);
448}
1da177e4 449
1da177e4
LT		 450extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
451extern unsigned int tcp_current_mss(struct sock *sk, int large);
452
17b085ea 453/* tcp.c */
1da177e4
LT		 454extern void tcp_get_info(struct sock *, struct tcp_info *);
455
456/* Read 'sendfile()'-style from a TCP socket */
457typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
458 unsigned int, size_t);
459extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
460 sk_read_actor_t recv_actor);
461
40efc6fa 462extern void tcp_initialize_rcv_mss(struct sock *sk);
1da177e4 463
5d424d5a
JH		 464extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
465extern int tcp_mss_to_mtu(struct sock *sk, int mss);
466extern void tcp_mtup_init(struct sock *sk);
467
40efc6fa 468static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
1da177e4
LT		 469{
470 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
471 ntohl(TCP_FLAG_ACK) |
472 snd_wnd);
473}
474
40efc6fa 475static inline void tcp_fast_path_on(struct tcp_sock *tp)
1da177e4
LT		 476{
477 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
478}
479
480static inline void tcp_fast_path_check(struct sock *sk, struct tcp_sock *tp)
481{
b03efcfb 482 if (skb_queue_empty(&tp->out_of_order_queue) &&
1da177e4
LT		 483 tp->rcv_wnd &&
484 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
485 !tp->urg_data)
486 tcp_fast_path_on(tp);
487}
488
489/* Compute the actual receive window we are currently advertising.
490 * Rcv_nxt can be after the window if our peer push more data
491 * than the offered window.
492 */
40efc6fa 493static inline u32 tcp_receive_window(const struct tcp_sock *tp)
1da177e4
LT		 494{
495 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
496
497 if (win < 0)
498 win = 0;
499 return (u32) win;
500}
501
502/* Choose a new window, without checks for shrinking, and without
503 * scaling applied to the result. The caller does these things
504 * if necessary. This is a "raw" window selection.
505 */
506extern u32 __tcp_select_window(struct sock *sk);
507
508/* TCP timestamps are only 32-bits, this causes a slight
509 * complication on 64-bit systems since we store a snapshot
31f34269
SH		 510 * of jiffies in the buffer control blocks below. We decided
511 * to use only the low 32-bits of jiffies and hide the ugly
1da177e4
LT		 512 * casts with the following macro.
513 */
514#define tcp_time_stamp ((__u32)(jiffies))
515
caa20d9a 516/* This is what the send packet queuing engine uses to pass
1da177e4
LT		 517 * TCP per-packet control information to the transmission
518 * code. We also store the host-order sequence numbers in
519 * here too. This is 36 bytes on 32-bit architectures,
520 * 40 bytes on 64-bit machines, if this grows please adjust
521 * skbuff.h:skbuff->cb[xxx] size appropriately.
522 */
523struct tcp_skb_cb {
524 union {
525 struct inet_skb_parm h4;
526#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
527 struct inet6_skb_parm h6;
528#endif
529 } header; /* For incoming frames */
530 __u32 seq; /* Starting sequence number */
531 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
532 __u32 when; /* used to compute rtt's */
533 __u8 flags; /* TCP header flags. */
534
535 /* NOTE: These must match up to the flags byte in a
536 * real TCP header.
537 */
538#define TCPCB_FLAG_FIN 0x01
539#define TCPCB_FLAG_SYN 0x02
540#define TCPCB_FLAG_RST 0x04
541#define TCPCB_FLAG_PSH 0x08
542#define TCPCB_FLAG_ACK 0x10
543#define TCPCB_FLAG_URG 0x20
544#define TCPCB_FLAG_ECE 0x40
545#define TCPCB_FLAG_CWR 0x80
546
547 __u8 sacked; /* State flags for SACK/FACK. */
548#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
549#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
550#define TCPCB_LOST 0x04 /* SKB is lost */
551#define TCPCB_TAGBITS 0x07 /* All tag bits */
552
553#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
554#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
555
caa20d9a 556#define TCPCB_URG 0x20 /* Urgent pointer advanced here */
1da177e4
LT		 557
558#define TCPCB_AT_TAIL (TCPCB_URG)
559
560 __u16 urg_ptr; /* Valid w/URG flags is set. */
561 __u32 ack_seq; /* Sequence number ACK'd */
562};
563
564#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
565
566#include <net/tcp_ecn.h>
567
568/* Due to TSO, an SKB can be composed of multiple actual
569 * packets. To keep these tracked properly, we use this.
570 */
571static inline int tcp_skb_pcount(const struct sk_buff *skb)
572{
7967168c 573 return skb_shinfo(skb)->gso_segs;
1da177e4
LT		 574}
575
576/* This is valid iff tcp_skb_pcount() > 1. */
577static inline int tcp_skb_mss(const struct sk_buff *skb)
578{
7967168c 579 return skb_shinfo(skb)->gso_size;
1da177e4
LT		 580}
581
582static inline void tcp_dec_pcount_approx(__u32 *count,
583 const struct sk_buff *skb)
584{
585 if (*count) {
586 *count -= tcp_skb_pcount(skb);
587 if ((int)*count < 0)
588 *count = 0;
589 }
590}
591
592static inline void tcp_packets_out_inc(struct sock *sk,
593 struct tcp_sock *tp,
594 const struct sk_buff *skb)
595{
596 int orig = tp->packets_out;
597
598 tp->packets_out += tcp_skb_pcount(skb);
599 if (!orig)
3f421baa
ACM		 600 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
601 inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
1da177e4
LT		 602}
603
604static inline void tcp_packets_out_dec(struct tcp_sock *tp,
605 const struct sk_buff *skb)
606{
607 tp->packets_out -= tcp_skb_pcount(skb);
608}
609
317a76f9
SH		 610/* Events passed to congestion control interface */
611enum tcp_ca_event {
612 CA_EVENT_TX_START, /* first transmit when no packets in flight */
613 CA_EVENT_CWND_RESTART, /* congestion window restart */
614 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
615 CA_EVENT_FRTO, /* fast recovery timeout */
616 CA_EVENT_LOSS, /* loss timeout */
617 CA_EVENT_FAST_ACK, /* in sequence ack */
618 CA_EVENT_SLOW_ACK, /* other ack */
619};
620
621/*
622 * Interface for adding new TCP congestion control handlers
623 */
624#define TCP_CA_NAME_MAX 16
3ff825b2
SH		 625#define TCP_CA_MAX 128
626#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
627
317a76f9
SH		 628struct tcp_congestion_ops {
629 struct list_head list;
ce7bc3bf 630 int non_restricted;
317a76f9
SH		 631
632 /* initialize private data (optional) */
6687e988 633 void (*init)(struct sock *sk);
317a76f9 634 /* cleanup private data (optional) */
6687e988 635 void (*release)(struct sock *sk);
317a76f9
SH		 636
637 /* return slow start threshold (required) */
6687e988 638 u32 (*ssthresh)(struct sock *sk);
317a76f9 639 /* lower bound for congestion window (optional) */
72dc5b92 640 u32 (*min_cwnd)(const struct sock *sk);
317a76f9 641 /* do new cwnd calculation (required) */
6687e988 642 void (*cong_avoid)(struct sock *sk, u32 ack,
317a76f9
SH		 643 u32 rtt, u32 in_flight, int good_ack);
644 /* round trip time sample per acked packet (optional) */
6687e988 645 void (*rtt_sample)(struct sock *sk, u32 usrtt);
317a76f9 646 /* call before changing ca_state (optional) */
6687e988 647 void (*set_state)(struct sock *sk, u8 new_state);
317a76f9 648 /* call when cwnd event occurs (optional) */
6687e988 649 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
317a76f9 650 /* new value of cwnd after loss (optional) */
6687e988 651 u32 (*undo_cwnd)(struct sock *sk);
317a76f9 652 /* hook for packet ack accounting (optional) */
6687e988 653 void (*pkts_acked)(struct sock *sk, u32 num_acked);
73c1f4a0 654 /* get info for inet_diag (optional) */
6687e988 655 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
317a76f9
SH		 656
657 char name[TCP_CA_NAME_MAX];
658 struct module *owner;
659};
660
661extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
662extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
663
6687e988
ACM		 664extern void tcp_init_congestion_control(struct sock *sk);
665extern void tcp_cleanup_congestion_control(struct sock *sk);
317a76f9
SH		 666extern int tcp_set_default_congestion_control(const char *name);
667extern void tcp_get_default_congestion_control(char *name);
3ff825b2 668extern void tcp_get_available_congestion_control(char *buf, size_t len);
ce7bc3bf
SH		 669extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
670extern int tcp_set_allowed_congestion_control(char *allowed);
6687e988 671extern int tcp_set_congestion_control(struct sock *sk, const char *name);
40efc6fa 672extern void tcp_slow_start(struct tcp_sock *tp);
317a76f9 673
5f8ef48d 674extern struct tcp_congestion_ops tcp_init_congestion_ops;
6687e988
ACM		 675extern u32 tcp_reno_ssthresh(struct sock *sk);
676extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack,
317a76f9 677 u32 rtt, u32 in_flight, int flag);
72dc5b92 678extern u32 tcp_reno_min_cwnd(const struct sock *sk);
a8acfbac 679extern struct tcp_congestion_ops tcp_reno;
317a76f9 680
6687e988 681static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
317a76f9 682{
6687e988
ACM		 683 struct inet_connection_sock *icsk = inet_csk(sk);
684
685 if (icsk->icsk_ca_ops->set_state)
686 icsk->icsk_ca_ops->set_state(sk, ca_state);
687 icsk->icsk_ca_state = ca_state;
317a76f9
SH		 688}
689
6687e988 690static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
317a76f9 691{
6687e988
ACM		 692 const struct inet_connection_sock *icsk = inet_csk(sk);
693
694 if (icsk->icsk_ca_ops->cwnd_event)
695 icsk->icsk_ca_ops->cwnd_event(sk, event);
317a76f9
SH		 696}
697
1da177e4
LT		 698/* This determines how many packets are "in the network" to the best
699 * of our knowledge. In many cases it is conservative, but where
700 * detailed information is available from the receiver (via SACK
701 * blocks etc.) we can make more aggressive calculations.
702 *
703 * Use this for decisions involving congestion control, use just
704 * tp->packets_out to determine if the send queue is empty or not.
705 *
706 * Read this equation as:
707 *
708 * "Packets sent once on transmission queue" MINUS
709 * "Packets left network, but not honestly ACKed yet" PLUS
710 * "Packets fast retransmitted"
711 */
40efc6fa 712static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
1da177e4
LT		 713{
714 return (tp->packets_out - tp->left_out + tp->retrans_out);
715}
716
1da177e4
LT		 717/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
718 * The exception is rate halving phase, when cwnd is decreasing towards
719 * ssthresh.
720 */
6687e988 721static inline __u32 tcp_current_ssthresh(const struct sock *sk)
1da177e4 722{
6687e988
ACM		 723 const struct tcp_sock *tp = tcp_sk(sk);
724 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
1da177e4
LT		 725 return tp->snd_ssthresh;
726 else
727 return max(tp->snd_ssthresh,
728 ((tp->snd_cwnd >> 1) +
729 (tp->snd_cwnd >> 2)));
730}
731
732static inline void tcp_sync_left_out(struct tcp_sock *tp)
733{
734 if (tp->rx_opt.sack_ok &&
735 (tp->sacked_out >= tp->packets_out - tp->lost_out))
736 tp->sacked_out = tp->packets_out - tp->lost_out;
737 tp->left_out = tp->sacked_out + tp->lost_out;
738}
739
40efc6fa 740extern void tcp_enter_cwr(struct sock *sk);
1da177e4
LT		 741extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
742
743/* Slow start with delack produces 3 packets of burst, so that
744 * it is safe "de facto".
745 */
746static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
747{
748 return 3;
749}
750
f4805ede
SH		 751/* RFC2861 Check whether we are limited by application or congestion window
752 * This is the inverse of cwnd check in tcp_tso_should_defer
753 */
754static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
755{
756 const struct tcp_sock *tp = tcp_sk(sk);
757 u32 left;
758
759 if (in_flight >= tp->snd_cwnd)
760 return 1;
761
bcd76111 762 if (!sk_can_gso(sk))
f4805ede
SH		 763 return 0;
764
765 left = tp->snd_cwnd - in_flight;
766 if (sysctl_tcp_tso_win_divisor)
767 return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd;
768 else
769 return left <= tcp_max_burst(tp);
770}
771
40efc6fa
SH		 772static inline void tcp_minshall_update(struct tcp_sock *tp, int mss,
773 const struct sk_buff *skb)
1da177e4
LT		 774{
775 if (skb->len < mss)
776 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
777}
778
40efc6fa 779static inline void tcp_check_probe_timer(struct sock *sk, struct tcp_sock *tp)
1da177e4 780{
463c84b9
ACM		 781 const struct inet_connection_sock *icsk = inet_csk(sk);
782 if (!tp->packets_out && !icsk->icsk_pending)
3f421baa
ACM		 783 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
784 icsk->icsk_rto, TCP_RTO_MAX);
1da177e4
LT		 785}
786
40efc6fa
SH		 787static inline void tcp_push_pending_frames(struct sock *sk,
788 struct tcp_sock *tp)
1da177e4
LT		 789{
790 __tcp_push_pending_frames(sk, tp, tcp_current_mss(sk, 1), tp->nonagle);
791}
792
40efc6fa 793static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT		 794{
795 tp->snd_wl1 = seq;
796}
797
40efc6fa 798static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT		 799{
800 tp->snd_wl1 = seq;
801}
802
1da177e4
LT		 803/*
804 * Calculate(/check) TCP checksum
805 */
ba7808ea
FD		 806static inline __sum16 tcp_v4_check(int len, __be32 saddr,
807 __be32 daddr, __wsum base)
1da177e4
LT		 808{
809 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
810}
811
b51655b9 812static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1da177e4 813{
fb286bb2 814 return __skb_checksum_complete(skb);
1da177e4
LT		 815}
816
40efc6fa 817static inline int tcp_checksum_complete(struct sk_buff *skb)
1da177e4
LT		 818{
819 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
820 __tcp_checksum_complete(skb);
821}
822
823/* Prequeue for VJ style copy to user, combined with checksumming. */
824
40efc6fa 825static inline void tcp_prequeue_init(struct tcp_sock *tp)
1da177e4
LT		 826{
827 tp->ucopy.task = NULL;
828 tp->ucopy.len = 0;
829 tp->ucopy.memory = 0;
830 skb_queue_head_init(&tp->ucopy.prequeue);
97fc2f08
CL		 831#ifdef CONFIG_NET_DMA
832 tp->ucopy.dma_chan = NULL;
833 tp->ucopy.wakeup = 0;
834 tp->ucopy.pinned_list = NULL;
835 tp->ucopy.dma_cookie = 0;
836#endif
1da177e4
LT		 837}
838
839/* Packet is added to VJ-style prequeue for processing in process
840 * context, if a reader task is waiting. Apparently, this exciting
841 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
842 * failed somewhere. Latency? Burstiness? Well, at least now we will
843 * see, why it failed. 8)8) --ANK
844 *
845 * NOTE: is this not too big to inline?
846 */
40efc6fa 847static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
1da177e4
LT		 848{
849 struct tcp_sock *tp = tcp_sk(sk);
850
851 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
852 __skb_queue_tail(&tp->ucopy.prequeue, skb);
853 tp->ucopy.memory += skb->truesize;
854 if (tp->ucopy.memory > sk->sk_rcvbuf) {
855 struct sk_buff *skb1;
856
857 BUG_ON(sock_owned_by_user(sk));
858
859 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
860 sk->sk_backlog_rcv(sk, skb1);
861 NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED);
862 }
863
864 tp->ucopy.memory = 0;
865 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
866 wake_up_interruptible(sk->sk_sleep);
463c84b9
ACM		 867 if (!inet_csk_ack_scheduled(sk))
868 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
3f421baa
ACM		 869 (3 * TCP_RTO_MIN) / 4,
870 TCP_RTO_MAX);
1da177e4
LT		 871 }
872 return 1;
873 }
874 return 0;
875}
876
877
878#undef STATE_TRACE
879
880#ifdef STATE_TRACE
881static const char *statename[]={
882 "Unused","Established","Syn Sent","Syn Recv",
883 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
884 "Close Wait","Last ACK","Listen","Closing"
885};
886#endif
887
40efc6fa 888static inline void tcp_set_state(struct sock *sk, int state)
1da177e4
LT		 889{
890 int oldstate = sk->sk_state;
891
892 switch (state) {
893 case TCP_ESTABLISHED:
894 if (oldstate != TCP_ESTABLISHED)
895 TCP_INC_STATS(TCP_MIB_CURRESTAB);
896 break;
897
898 case TCP_CLOSE:
899 if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED)
900 TCP_INC_STATS(TCP_MIB_ESTABRESETS);
901
902 sk->sk_prot->unhash(sk);
463c84b9 903 if (inet_csk(sk)->icsk_bind_hash &&
1da177e4 904 !(sk->sk_userlocks & SOCK_BINDPORT_LOCK))
2d8c4ce5 905 inet_put_port(&tcp_hashinfo, sk);
1da177e4
LT		 906 /* fall through */
907 default:
908 if (oldstate==TCP_ESTABLISHED)
909 TCP_DEC_STATS(TCP_MIB_CURRESTAB);
910 }
911
912 /* Change state AFTER socket is unhashed to avoid closed
913 * socket sitting in hash tables.
914 */
915 sk->sk_state = state;
916
917#ifdef STATE_TRACE
918 SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n",sk, statename[oldstate],statename[state]);
919#endif
920}
921
40efc6fa 922static inline void tcp_done(struct sock *sk)
1da177e4 923{
3687b1dc
WY		 924 if(sk->sk_state == TCP_SYN_SENT || sk->sk_state == TCP_SYN_RECV)
925 TCP_INC_STATS_BH(TCP_MIB_ATTEMPTFAILS);
926
1da177e4
LT		 927 tcp_set_state(sk, TCP_CLOSE);
928 tcp_clear_xmit_timers(sk);
929
930 sk->sk_shutdown = SHUTDOWN_MASK;
931
932 if (!sock_flag(sk, SOCK_DEAD))
933 sk->sk_state_change(sk);
934 else
0a5578cf 935 inet_csk_destroy_sock(sk);
1da177e4
LT		 936}
937
40efc6fa 938static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1da177e4
LT		 939{
940 rx_opt->dsack = 0;
941 rx_opt->eff_sacks = 0;
942 rx_opt->num_sacks = 0;
943}
944
1da177e4
LT		 945/* Determine a window scaling and initial window to offer. */
946extern void tcp_select_initial_window(int __space, __u32 mss,
947 __u32 *rcv_wnd, __u32 *window_clamp,
948 int wscale_ok, __u8 *rcv_wscale);
949
950static inline int tcp_win_from_space(int space)
951{
952 return sysctl_tcp_adv_win_scale<=0 ?
953 (space>>(-sysctl_tcp_adv_win_scale)) :
954 space - (space>>sysctl_tcp_adv_win_scale);
955}
956
957/* Note: caller must be prepared to deal with negative returns */
958static inline int tcp_space(const struct sock *sk)
959{
960 return tcp_win_from_space(sk->sk_rcvbuf -
961 atomic_read(&sk->sk_rmem_alloc));
962}
963
964static inline int tcp_full_space(const struct sock *sk)
965{
966 return tcp_win_from_space(sk->sk_rcvbuf);
967}
968
40efc6fa
SH		 969static inline void tcp_openreq_init(struct request_sock *req,
970 struct tcp_options_received *rx_opt,
971 struct sk_buff *skb)
1da177e4 972{
2e6599cb
ACM		 973 struct inet_request_sock *ireq = inet_rsk(req);
974
1da177e4 975 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
2e6599cb 976 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1da177e4
LT		 977 req->mss = rx_opt->mss_clamp;
978 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
2e6599cb
ACM		 979 ireq->tstamp_ok = rx_opt->tstamp_ok;
980 ireq->sack_ok = rx_opt->sack_ok;
981 ireq->snd_wscale = rx_opt->snd_wscale;
982 ireq->wscale_ok = rx_opt->wscale_ok;
983 ireq->acked = 0;
984 ireq->ecn_ok = 0;
985 ireq->rmt_port = skb->h.th->source;
1da177e4
LT		 986}
987
988extern void tcp_enter_memory_pressure(void);
989
1da177e4
LT		 990static inline int keepalive_intvl_when(const struct tcp_sock *tp)
991{
992 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
993}
994
995static inline int keepalive_time_when(const struct tcp_sock *tp)
996{
997 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
998}
999
463c84b9 1000static inline int tcp_fin_time(const struct sock *sk)
1da177e4 1001{
463c84b9
ACM		 1002 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1003 const int rto = inet_csk(sk)->icsk_rto;
1da177e4 1004
463c84b9
ACM		 1005 if (fin_timeout < (rto << 2) - (rto >> 1))
1006 fin_timeout = (rto << 2) - (rto >> 1);
1da177e4
LT		 1007
1008 return fin_timeout;
1009}
1010
1011static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1012{
1013 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1014 return 0;
1015 if (xtime.tv_sec >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
1016 return 0;
1017
1018 /* RST segments are not recommended to carry timestamp,
1019 and, if they do, it is recommended to ignore PAWS because
1020 "their cleanup function should take precedence over timestamps."
1021 Certainly, it is mistake. It is necessary to understand the reasons
1022 of this constraint to relax it: if peer reboots, clock may go
1023 out-of-sync and half-open connections will not be reset.
1024 Actually, the problem would be not existing if all
1025 the implementations followed draft about maintaining clock
1026 via reboots. Linux-2.2 DOES NOT!
1027
1028 However, we can relax time bounds for RST segments to MSL.
1029 */
1030 if (rst && xtime.tv_sec >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1031 return 0;
1032 return 1;
1033}
1034
1da177e4
LT		 1035#define TCP_CHECK_TIMER(sk) do { } while (0)
1036
1da177e4
LT		 1037static inline void tcp_mib_init(void)
1038{
1039 /* See RFC 2012 */
1040 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1041 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1042 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1043 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1044}
1045
6a438bbe
SH		 1046/*from STCP */
1047static inline void clear_all_retrans_hints(struct tcp_sock *tp){
1048 tp->lost_skb_hint = NULL;
1049 tp->scoreboard_skb_hint = NULL;
1050 tp->retransmit_skb_hint = NULL;
1051 tp->forward_skb_hint = NULL;
1052 tp->fastpath_skb_hint = NULL;
1053}
1054
cfb6eeb4
YH		 1055/* MD5 Signature */
1056struct crypto_hash;
1057
1058/* - key database */
1059struct tcp_md5sig_key {
1060 u8 *key;
1061 u8 keylen;
1062};
1063
1064struct tcp4_md5sig_key {
1065 u8 *key;
1066 u16 keylen;
1067 __be32 addr;
1068};
1069
1070struct tcp6_md5sig_key {
1071 u8 *key;
1072 u16 keylen;
1073#if 0
1074 u32 scope_id; /* XXX */
1075#endif
1076 struct in6_addr addr;
1077};
1078
1079/* - sock block */
1080struct tcp_md5sig_info {
1081 struct tcp4_md5sig_key *keys4;
1082#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1083 struct tcp6_md5sig_key *keys6;
1084 u32 entries6;
1085 u32 alloced6;
1086#endif
1087 u32 entries4;
1088 u32 alloced4;
1089};
1090
1091/* - pseudo header */
1092struct tcp4_pseudohdr {
1093 __be32 saddr;
1094 __be32 daddr;
1095 __u8 pad;
1096 __u8 protocol;
1097 __be16 len;
1098};
1099
1100struct tcp6_pseudohdr {
1101 struct in6_addr saddr;
1102 struct in6_addr daddr;
1103 __be32 len;
1104 __be32 protocol; /* including padding */
1105};
1106
1107union tcp_md5sum_block {
1108 struct tcp4_pseudohdr ip4;
1109#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1110 struct tcp6_pseudohdr ip6;
1111#endif
1112};
1113
1114/* - pool: digest algorithm, hash description and scratch buffer */
1115struct tcp_md5sig_pool {
1116 struct hash_desc md5_desc;
1117 union tcp_md5sum_block md5_blk;
1118};
1119
1120#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1121
1122/* - functions */
1123extern int tcp_v4_calc_md5_hash(char *md5_hash,
1124 struct tcp_md5sig_key *key,
1125 struct sock *sk,
1126 struct dst_entry *dst,
1127 struct request_sock *req,
1128 struct tcphdr *th,
1129 int protocol, int tcplen);
1130extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1131 struct sock *addr_sk);
1132
1133extern int tcp_v4_md5_do_add(struct sock *sk,
1134 __be32 addr,
1135 u8 *newkey,
1136 u8 newkeylen);
1137
1138extern int tcp_v4_md5_do_del(struct sock *sk,
8e5200f5 1139 __be32 addr);
cfb6eeb4
YH		 1140
1141extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1142extern void tcp_free_md5sig_pool(void);
1143
1144extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1145extern void __tcp_put_md5sig_pool(void);
1146
1147static inline
1148struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1149{
1150 int cpu = get_cpu();
1151 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1152 if (!ret)
1153 put_cpu();
1154 return ret;
1155}
1156
1157static inline void tcp_put_md5sig_pool(void)
1158{
1159 __tcp_put_md5sig_pool();
1160 put_cpu();
1161}
1162
1da177e4
LT		 1163/* /proc */
1164enum tcp_seq_states {
1165 TCP_SEQ_STATE_LISTENING,
1166 TCP_SEQ_STATE_OPENREQ,
1167 TCP_SEQ_STATE_ESTABLISHED,
1168 TCP_SEQ_STATE_TIME_WAIT,
1169};
1170
1171struct tcp_seq_afinfo {
1172 struct module *owner;
1173 char *name;
1174 sa_family_t family;
1175 int (*seq_show) (struct seq_file *m, void *v);
1176 struct file_operations *seq_fops;
1177};
1178
1179struct tcp_iter_state {
1180 sa_family_t family;
1181 enum tcp_seq_states state;
1182 struct sock *syn_wait_sk;
1183 int bucket, sbucket, num, uid;
1184 struct seq_operations seq_ops;
1185};
1186
1187extern int tcp_proc_register(struct tcp_seq_afinfo *afinfo);
1188extern void tcp_proc_unregister(struct tcp_seq_afinfo *afinfo);
1189
20380731
ACM		 1190extern struct request_sock_ops tcp_request_sock_ops;
1191
1192extern int tcp_v4_destroy_sock(struct sock *sk);
1193
a430a43d 1194extern int tcp_v4_gso_send_check(struct sk_buff *skb);
576a30eb 1195extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
f4c50d99 1196
20380731
ACM		 1197#ifdef CONFIG_PROC_FS
1198extern int tcp4_proc_init(void);
1199extern void tcp4_proc_exit(void);
1200#endif
1201
cfb6eeb4
YH		 1202/* TCP af-specific functions */
1203struct tcp_sock_af_ops {
1204#ifdef CONFIG_TCP_MD5SIG
1205 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1206 struct sock *addr_sk);
1207 int (*calc_md5_hash) (char *location,
1208 struct tcp_md5sig_key *md5,
1209 struct sock *sk,
1210 struct dst_entry *dst,
1211 struct request_sock *req,
1212 struct tcphdr *th,
1213 int protocol, int len);
1214 int (*md5_add) (struct sock *sk,
1215 struct sock *addr_sk,
1216 u8 *newkey,
1217 u8 len);
1218 int (*md5_parse) (struct sock *sk,
1219 char __user *optval,
1220 int optlen);
1221#endif
1222};
1223
1224struct tcp_request_sock_ops {
1225#ifdef CONFIG_TCP_MD5SIG
1226 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1227 struct request_sock *req);
1228#endif
1229};
1230
20380731
ACM		 1231extern void tcp_v4_init(struct net_proto_family *ops);
1232extern void tcp_init(void);
1233
1da177e4 1234#endif /* _TCP_H */
Ubuntu Artful kernel mirror