projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
tcp md5sig: Share most of hash calcucaltion bits between IPv4 and IPv6.
[linux-2.6-block.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>
c6aefafb 32#include <linux/cryptohash.h>
3f421baa
ACM		 33
34#include <net/inet_connection_sock.h>
295ff7ed 35#include <net/inet_timewait_sock.h>
77d8bf9c 36#include <net/inet_hashtables.h>
1da177e4 37#include <net/checksum.h>
2e6599cb 38#include <net/request_sock.h>
1da177e4
LT		 39#include <net/sock.h>
40#include <net/snmp.h>
41#include <net/ip.h>
c752f073 42#include <net/tcp_states.h>
bdf1ee5d 43#include <net/inet_ecn.h>
c752f073 44
1da177e4
LT		 45#include <linux/seq_file.h>
46
6e04e021 47extern struct inet_hashinfo tcp_hashinfo;
1da177e4 48
1da177e4 49extern atomic_t tcp_orphan_count;
1da177e4 50extern void tcp_time_wait(struct sock *sk, int state, int timeo);
1da177e4 51
1da177e4
LT		 52#define MAX_TCP_HEADER (128 + MAX_HEADER)
53
54/*
55 * Never offer a window over 32767 without using window scaling. Some
56 * poor stacks do signed 16bit maths!
57 */
58#define MAX_TCP_WINDOW 32767U
59
60/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
61#define TCP_MIN_MSS 88U
62
63/* Minimal RCV_MSS. */
64#define TCP_MIN_RCVMSS 536U
65
5d424d5a
JH		 66/* The least MTU to use for probing */
67#define TCP_BASE_MSS 512
68
1da177e4
LT		 69/* After receiving this amount of duplicate ACKs fast retransmit starts. */
70#define TCP_FASTRETRANS_THRESH 3
71
72/* Maximal reordering. */
73#define TCP_MAX_REORDERING 127
74
75/* Maximal number of ACKs sent quickly to accelerate slow-start. */
76#define TCP_MAX_QUICKACKS 16U
77
78/* urg_data states */
79#define TCP_URG_VALID 0x0100
80#define TCP_URG_NOTYET 0x0200
81#define TCP_URG_READ 0x0400
82
83#define TCP_RETR1 3 /*
84 * This is how many retries it does before it
85 * tries to figure out if the gateway is
86 * down. Minimal RFC value is 3; it corresponds
87 * to ~3sec-8min depending on RTO.
88 */
89
90#define TCP_RETR2 15 /*
91 * This should take at least
92 * 90 minutes to time out.
93 * RFC1122 says that the limit is 100 sec.
94 * 15 is ~13-30min depending on RTO.
95 */
96
97#define TCP_SYN_RETRIES 5 /* number of times to retry active opening a
caa20d9a 98 * connection: ~180sec is RFC minimum */
1da177e4
LT		 99
100#define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
caa20d9a 101 * connection: ~180sec is RFC minimum */
1da177e4
LT		 102
103
104#define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned
105 * socket. 7 is ~50sec-16min.
106 */
107
108
109#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
110 * state, about 60 seconds */
111#define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
112 /* BSD style FIN_WAIT2 deadlock breaker.
113 * It used to be 3min, new value is 60sec,
114 * to combine FIN-WAIT-2 timeout with
115 * TIME-WAIT timer.
116 */
117
118#define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
119#if HZ >= 100
120#define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
121#define TCP_ATO_MIN ((unsigned)(HZ/25))
122#else
123#define TCP_DELACK_MIN 4U
124#define TCP_ATO_MIN 4U
125#endif
126#define TCP_RTO_MAX ((unsigned)(120*HZ))
127#define TCP_RTO_MIN ((unsigned)(HZ/5))
128#define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */
129
130#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
131 * for local resources.
132 */
133
134#define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
135#define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
136#define TCP_KEEPALIVE_INTVL (75*HZ)
137
138#define MAX_TCP_KEEPIDLE 32767
139#define MAX_TCP_KEEPINTVL 32767
140#define MAX_TCP_KEEPCNT 127
141#define MAX_TCP_SYNCNT 127
ec3c0982 142#define MAX_TCP_ACCEPT_DEFERRED 65535
1da177e4
LT		 143
144#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
1da177e4
LT		 145
146#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
147#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
148 * after this time. It should be equal
149 * (or greater than) TCP_TIMEWAIT_LEN
150 * to provide reliability equal to one
151 * provided by timewait state.
152 */
153#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
154 * timestamps. It must be less than
155 * minimal timewait lifetime.
156 */
1da177e4
LT		 157/*
158 * TCP option
159 */
160
161#define TCPOPT_NOP 1 /* Padding */
162#define TCPOPT_EOL 0 /* End of options */
163#define TCPOPT_MSS 2 /* Segment size negotiating */
164#define TCPOPT_WINDOW 3 /* Window scaling */
165#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
166#define TCPOPT_SACK 5 /* SACK Block */
167#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
cfb6eeb4 168#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
1da177e4
LT		 169
170/*
171 * TCP option lengths
172 */
173
174#define TCPOLEN_MSS 4
175#define TCPOLEN_WINDOW 3
176#define TCPOLEN_SACK_PERM 2
177#define TCPOLEN_TIMESTAMP 10
cfb6eeb4 178#define TCPOLEN_MD5SIG 18
1da177e4
LT		 179
180/* But this is what stacks really send out. */
181#define TCPOLEN_TSTAMP_ALIGNED 12
182#define TCPOLEN_WSCALE_ALIGNED 4
183#define TCPOLEN_SACKPERM_ALIGNED 4
184#define TCPOLEN_SACK_BASE 2
185#define TCPOLEN_SACK_BASE_ALIGNED 4
186#define TCPOLEN_SACK_PERBLOCK 8
cfb6eeb4 187#define TCPOLEN_MD5SIG_ALIGNED 20
1da177e4 188
1da177e4
LT		 189/* Flags in tp->nonagle */
190#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
191#define TCP_NAGLE_CORK 2 /* Socket is corked */
caa20d9a 192#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
1da177e4 193
295ff7ed
ACM		 194extern struct inet_timewait_death_row tcp_death_row;
195
1da177e4 196/* sysctl variables for tcp */
1da177e4
LT		 197extern int sysctl_tcp_timestamps;
198extern int sysctl_tcp_window_scaling;
199extern int sysctl_tcp_sack;
200extern int sysctl_tcp_fin_timeout;
1da177e4
LT		 201extern int sysctl_tcp_keepalive_time;
202extern int sysctl_tcp_keepalive_probes;
203extern int sysctl_tcp_keepalive_intvl;
204extern int sysctl_tcp_syn_retries;
205extern int sysctl_tcp_synack_retries;
206extern int sysctl_tcp_retries1;
207extern int sysctl_tcp_retries2;
208extern int sysctl_tcp_orphan_retries;
209extern int sysctl_tcp_syncookies;
210extern int sysctl_tcp_retrans_collapse;
211extern int sysctl_tcp_stdurg;
212extern int sysctl_tcp_rfc1337;
213extern int sysctl_tcp_abort_on_overflow;
214extern int sysctl_tcp_max_orphans;
1da177e4
LT		 215extern int sysctl_tcp_fack;
216extern int sysctl_tcp_reordering;
217extern int sysctl_tcp_ecn;
218extern int sysctl_tcp_dsack;
219extern int sysctl_tcp_mem[3];
220extern int sysctl_tcp_wmem[3];
221extern int sysctl_tcp_rmem[3];
222extern int sysctl_tcp_app_win;
223extern int sysctl_tcp_adv_win_scale;
224extern int sysctl_tcp_tw_reuse;
225extern int sysctl_tcp_frto;
3cfe3baa 226extern int sysctl_tcp_frto_response;
1da177e4 227extern int sysctl_tcp_low_latency;
95937825 228extern int sysctl_tcp_dma_copybreak;
1da177e4 229extern int sysctl_tcp_nometrics_save;
1da177e4
LT		 230extern int sysctl_tcp_moderate_rcvbuf;
231extern int sysctl_tcp_tso_win_divisor;
9772efb9 232extern int sysctl_tcp_abc;
5d424d5a
JH		 233extern int sysctl_tcp_mtu_probing;
234extern int sysctl_tcp_base_mss;
15d99e02 235extern int sysctl_tcp_workaround_signed_windows;
35089bb2 236extern int sysctl_tcp_slow_start_after_idle;
886236c1 237extern int sysctl_tcp_max_ssthresh;
1da177e4
LT		 238
239extern atomic_t tcp_memory_allocated;
240extern atomic_t tcp_sockets_allocated;
241extern int tcp_memory_pressure;
242
1da177e4
LT		 243/*
244 * The next routines deal with comparing 32 bit unsigned ints
245 * and worry about wraparound (automatic with unsigned arithmetic).
246 */
247
248static inline int before(__u32 seq1, __u32 seq2)
249{
0d630cc0 250 return (__s32)(seq1-seq2) < 0;
1da177e4 251}
9a036b9c 252#define after(seq2, seq1) before(seq1, seq2)
1da177e4
LT		 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
e4fd5da3
PE		 260static inline int tcp_too_many_orphans(struct sock *sk, int num)
261{
262 return (num > sysctl_tcp_max_orphans) ||
263 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
264 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
265}
1da177e4
LT		 266
267extern struct proto tcp_prot;
268
269DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
270#define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field)
271#define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field)
272#define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field)
273#define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field)
274#define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val)
275#define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val)
276
1da177e4
LT		 277extern void tcp_v4_err(struct sk_buff *skb, u32);
278
279extern void tcp_shutdown (struct sock *sk, int how);
280
281extern int tcp_v4_rcv(struct sk_buff *skb);
282
283extern int tcp_v4_remember_stamp(struct sock *sk);
284
8feaf0c0 285extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
1da177e4 286
3516ffb0 287extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
1da177e4
LT		 288 struct msghdr *msg, size_t size);
289extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
290
291extern int tcp_ioctl(struct sock *sk,
292 int cmd,
293 unsigned long arg);
294
295extern int tcp_rcv_state_process(struct sock *sk,
296 struct sk_buff *skb,
297 struct tcphdr *th,
298 unsigned len);
299
300extern int tcp_rcv_established(struct sock *sk,
301 struct sk_buff *skb,
302 struct tcphdr *th,
303 unsigned len);
304
305extern void tcp_rcv_space_adjust(struct sock *sk);
306
0e4b4992
CL		 307extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
308
6d6ee43e
ACM		 309extern int tcp_twsk_unique(struct sock *sk,
310 struct sock *sktw, void *twp);
311
cfb6eeb4
YH		 312extern void tcp_twsk_destructor(struct sock *sk);
313
9c55e01c
JA		 314extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
315 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
316
463c84b9
ACM		 317static inline void tcp_dec_quickack_mode(struct sock *sk,
318 const unsigned int pkts)
1da177e4 319{
463c84b9 320 struct inet_connection_sock *icsk = inet_csk(sk);
fc6415bc 321
463c84b9
ACM		 322 if (icsk->icsk_ack.quick) {
323 if (pkts >= icsk->icsk_ack.quick) {
324 icsk->icsk_ack.quick = 0;
fc6415bc 325 /* Leaving quickack mode we deflate ATO. */
463c84b9 326 icsk->icsk_ack.ato = TCP_ATO_MIN;
fc6415bc 327 } else
463c84b9 328 icsk->icsk_ack.quick -= pkts;
1da177e4
LT		 329 }
330}
331
463c84b9 332extern void tcp_enter_quickack_mode(struct sock *sk);
1da177e4 333
1da177e4
LT		 334static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
335{
336 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
337}
338
bdf1ee5d
IJ		 339#define TCP_ECN_OK 1
340#define TCP_ECN_QUEUE_CWR 2
341#define TCP_ECN_DEMAND_CWR 4
342
343static __inline__ void
344TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
345{
346 if (sysctl_tcp_ecn && th->ece && th->cwr)
347 inet_rsk(req)->ecn_ok = 1;
348}
349
1da177e4
LT		 350enum tcp_tw_status
351{
352 TCP_TW_SUCCESS = 0,
353 TCP_TW_RST = 1,
354 TCP_TW_ACK = 2,
355 TCP_TW_SYN = 3
356};
357
358
8feaf0c0 359extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
1da177e4 360 struct sk_buff *skb,
8feaf0c0 361 const struct tcphdr *th);
1da177e4
LT		 362
363extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
60236fdd
ACM		 364 struct request_sock *req,
365 struct request_sock **prev);
1da177e4
LT		 366extern int tcp_child_process(struct sock *parent,
367 struct sock *child,
368 struct sk_buff *skb);
46d0de4e 369extern int tcp_use_frto(struct sock *sk);
1da177e4
LT		 370extern void tcp_enter_frto(struct sock *sk);
371extern void tcp_enter_loss(struct sock *sk, int how);
372extern void tcp_clear_retrans(struct tcp_sock *tp);
373extern void tcp_update_metrics(struct sock *sk);
374
375extern void tcp_close(struct sock *sk,
376 long timeout);
1da177e4
LT		 377extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
378
379extern int tcp_getsockopt(struct sock *sk, int level,
380 int optname,
381 char __user *optval,
382 int __user *optlen);
383extern int tcp_setsockopt(struct sock *sk, int level,
384 int optname, char __user *optval,
385 int optlen);
3fdadf7d
DM		 386extern int compat_tcp_getsockopt(struct sock *sk,
387 int level, int optname,
388 char __user *optval, int __user *optlen);
389extern int compat_tcp_setsockopt(struct sock *sk,
390 int level, int optname,
391 char __user *optval, int optlen);
1da177e4
LT		 392extern void tcp_set_keepalive(struct sock *sk, int val);
393extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
394 struct msghdr *msg,
395 size_t len, int nonblock,
396 int flags, int *addr_len);
397
1da177e4
LT		 398extern void tcp_parse_options(struct sk_buff *skb,
399 struct tcp_options_received *opt_rx,
400 int estab);
401
7d5d5525
YH		 402extern u8 *tcp_parse_md5sig_option(struct tcphdr *th);
403
1da177e4
LT		 404/*
405 * TCP v4 functions exported for the inet6 API
406 */
407
8292a17a 408extern void tcp_v4_send_check(struct sock *sk, int len,
1da177e4
LT		 409 struct sk_buff *skb);
410
411extern int tcp_v4_conn_request(struct sock *sk,
412 struct sk_buff *skb);
413
414extern struct sock * tcp_create_openreq_child(struct sock *sk,
60236fdd 415 struct request_sock *req,
1da177e4
LT		 416 struct sk_buff *skb);
417
418extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
419 struct sk_buff *skb,
60236fdd 420 struct request_sock *req,
1da177e4
LT		 421 struct dst_entry *dst);
422
423extern int tcp_v4_do_rcv(struct sock *sk,
424 struct sk_buff *skb);
425
426extern int tcp_v4_connect(struct sock *sk,
427 struct sockaddr *uaddr,
428 int addr_len);
429
430extern int tcp_connect(struct sock *sk);
431
432extern struct sk_buff * tcp_make_synack(struct sock *sk,
433 struct dst_entry *dst,
60236fdd 434 struct request_sock *req);
1da177e4
LT		 435
436extern int tcp_disconnect(struct sock *sk, int flags);
437
438extern void tcp_unhash(struct sock *sk);
439
1da177e4 440/* From syncookies.c */
2051f11f 441extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
1da177e4
LT		 442extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
443 struct ip_options *opt);
444extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
445 __u16 *mss);
446
4dfc2817
FW		 447extern __u32 cookie_init_timestamp(struct request_sock *req);
448extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt);
449
c6aefafb
GG		 450/* From net/ipv6/syncookies.c */
451extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
452extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb,
453 __u16 *mss);
454
1da177e4
LT		 455/* tcp_output.c */
456
9e412ba7
IJ		 457extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
458 int nonagle);
459extern int tcp_may_send_now(struct sock *sk);
1da177e4
LT		 460extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
461extern void tcp_xmit_retransmit_queue(struct sock *);
462extern void tcp_simple_retransmit(struct sock *);
463extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
6475be16 464extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
1da177e4
LT		 465
466extern void tcp_send_probe0(struct sock *);
467extern void tcp_send_partial(struct sock *);
468extern int tcp_write_wakeup(struct sock *);
469extern void tcp_send_fin(struct sock *sk);
dd0fc66f 470extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
1da177e4 471extern int tcp_send_synack(struct sock *);
c1b4a7e6 472extern void tcp_push_one(struct sock *, unsigned int mss_now);
1da177e4
LT		 473extern void tcp_send_ack(struct sock *sk);
474extern void tcp_send_delayed_ack(struct sock *sk);
475
a762a980
DM		 476/* tcp_input.c */
477extern void tcp_cwnd_application_limited(struct sock *sk);
478
1da177e4
LT		 479/* tcp_timer.c */
480extern void tcp_init_xmit_timers(struct sock *);
463c84b9
ACM		 481static inline void tcp_clear_xmit_timers(struct sock *sk)
482{
483 inet_csk_clear_xmit_timers(sk);
484}
1da177e4 485
1da177e4
LT		 486extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
487extern unsigned int tcp_current_mss(struct sock *sk, int large);
488
17b085ea 489/* tcp.c */
1da177e4
LT		 490extern void tcp_get_info(struct sock *, struct tcp_info *);
491
492/* Read 'sendfile()'-style from a TCP socket */
493typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
494 unsigned int, size_t);
495extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
496 sk_read_actor_t recv_actor);
497
40efc6fa 498extern void tcp_initialize_rcv_mss(struct sock *sk);
1da177e4 499
5d424d5a
JH		 500extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
501extern int tcp_mss_to_mtu(struct sock *sk, int mss);
502extern void tcp_mtup_init(struct sock *sk);
503
40efc6fa 504static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
1da177e4
LT		 505{
506 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
507 ntohl(TCP_FLAG_ACK) |
508 snd_wnd);
509}
510
40efc6fa 511static inline void tcp_fast_path_on(struct tcp_sock *tp)
1da177e4
LT		 512{
513 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
514}
515
9e412ba7 516static inline void tcp_fast_path_check(struct sock *sk)
1da177e4 517{
9e412ba7
IJ		 518 struct tcp_sock *tp = tcp_sk(sk);
519
b03efcfb 520 if (skb_queue_empty(&tp->out_of_order_queue) &&
1da177e4
LT		 521 tp->rcv_wnd &&
522 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
523 !tp->urg_data)
524 tcp_fast_path_on(tp);
525}
526
527/* Compute the actual receive window we are currently advertising.
528 * Rcv_nxt can be after the window if our peer push more data
529 * than the offered window.
530 */
40efc6fa 531static inline u32 tcp_receive_window(const struct tcp_sock *tp)
1da177e4
LT		 532{
533 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
534
535 if (win < 0)
536 win = 0;
537 return (u32) win;
538}
539
540/* Choose a new window, without checks for shrinking, and without
541 * scaling applied to the result. The caller does these things
542 * if necessary. This is a "raw" window selection.
543 */
544extern u32 __tcp_select_window(struct sock *sk);
545
546/* TCP timestamps are only 32-bits, this causes a slight
547 * complication on 64-bit systems since we store a snapshot
31f34269
SH		 548 * of jiffies in the buffer control blocks below. We decided
549 * to use only the low 32-bits of jiffies and hide the ugly
1da177e4
LT		 550 * casts with the following macro.
551 */
552#define tcp_time_stamp ((__u32)(jiffies))
553
caa20d9a 554/* This is what the send packet queuing engine uses to pass
1da177e4
LT		 555 * TCP per-packet control information to the transmission
556 * code. We also store the host-order sequence numbers in
557 * here too. This is 36 bytes on 32-bit architectures,
558 * 40 bytes on 64-bit machines, if this grows please adjust
559 * skbuff.h:skbuff->cb[xxx] size appropriately.
560 */
561struct tcp_skb_cb {
562 union {
563 struct inet_skb_parm h4;
564#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
565 struct inet6_skb_parm h6;
566#endif
567 } header; /* For incoming frames */
568 __u32 seq; /* Starting sequence number */
569 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
570 __u32 when; /* used to compute rtt's */
571 __u8 flags; /* TCP header flags. */
572
573 /* NOTE: These must match up to the flags byte in a
574 * real TCP header.
575 */
576#define TCPCB_FLAG_FIN 0x01
577#define TCPCB_FLAG_SYN 0x02
578#define TCPCB_FLAG_RST 0x04
579#define TCPCB_FLAG_PSH 0x08
580#define TCPCB_FLAG_ACK 0x10
581#define TCPCB_FLAG_URG 0x20
582#define TCPCB_FLAG_ECE 0x40
583#define TCPCB_FLAG_CWR 0x80
584
585 __u8 sacked; /* State flags for SACK/FACK. */
586#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
587#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
588#define TCPCB_LOST 0x04 /* SKB is lost */
589#define TCPCB_TAGBITS 0x07 /* All tag bits */
590
591#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
592#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
593
1da177e4
LT		 594 __u16 urg_ptr; /* Valid w/URG flags is set. */
595 __u32 ack_seq; /* Sequence number ACK'd */
596};
597
598#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
599
1da177e4
LT		 600/* Due to TSO, an SKB can be composed of multiple actual
601 * packets. To keep these tracked properly, we use this.
602 */
603static inline int tcp_skb_pcount(const struct sk_buff *skb)
604{
7967168c 605 return skb_shinfo(skb)->gso_segs;
1da177e4
LT		 606}
607
608/* This is valid iff tcp_skb_pcount() > 1. */
609static inline int tcp_skb_mss(const struct sk_buff *skb)
610{
7967168c 611 return skb_shinfo(skb)->gso_size;
1da177e4
LT		 612}
613
af610b4c 614static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr)
1da177e4
LT		 615{
616 if (*count) {
af610b4c 617 *count -= decr;
1da177e4
LT		 618 if ((int)*count < 0)
619 *count = 0;
620 }
621}
622
af610b4c
IJ		 623static inline void tcp_dec_pcount_approx(__u32 *count,
624 const struct sk_buff *skb)
625{
626 tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb));
627}
628
317a76f9
SH		 629/* Events passed to congestion control interface */
630enum tcp_ca_event {
631 CA_EVENT_TX_START, /* first transmit when no packets in flight */
632 CA_EVENT_CWND_RESTART, /* congestion window restart */
633 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
634 CA_EVENT_FRTO, /* fast recovery timeout */
635 CA_EVENT_LOSS, /* loss timeout */
636 CA_EVENT_FAST_ACK, /* in sequence ack */
637 CA_EVENT_SLOW_ACK, /* other ack */
638};
639
640/*
641 * Interface for adding new TCP congestion control handlers
642 */
643#define TCP_CA_NAME_MAX 16
3ff825b2
SH		 644#define TCP_CA_MAX 128
645#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
646
164891aa
SH		 647#define TCP_CONG_NON_RESTRICTED 0x1
648#define TCP_CONG_RTT_STAMP 0x2
649
317a76f9
SH		 650struct tcp_congestion_ops {
651 struct list_head list;
164891aa 652 unsigned long flags;
317a76f9
SH		 653
654 /* initialize private data (optional) */
6687e988 655 void (*init)(struct sock *sk);
317a76f9 656 /* cleanup private data (optional) */
6687e988 657 void (*release)(struct sock *sk);
317a76f9
SH		 658
659 /* return slow start threshold (required) */
6687e988 660 u32 (*ssthresh)(struct sock *sk);
317a76f9 661 /* lower bound for congestion window (optional) */
72dc5b92 662 u32 (*min_cwnd)(const struct sock *sk);
317a76f9 663 /* do new cwnd calculation (required) */
c3a05c60 664 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
317a76f9 665 /* call before changing ca_state (optional) */
6687e988 666 void (*set_state)(struct sock *sk, u8 new_state);
317a76f9 667 /* call when cwnd event occurs (optional) */
6687e988 668 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
317a76f9 669 /* new value of cwnd after loss (optional) */
6687e988 670 u32 (*undo_cwnd)(struct sock *sk);
317a76f9 671 /* hook for packet ack accounting (optional) */
30cfd0ba 672 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
73c1f4a0 673 /* get info for inet_diag (optional) */
6687e988 674 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
317a76f9
SH		 675
676 char name[TCP_CA_NAME_MAX];
677 struct module *owner;
678};
679
680extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
681extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
682
6687e988
ACM		 683extern void tcp_init_congestion_control(struct sock *sk);
684extern void tcp_cleanup_congestion_control(struct sock *sk);
317a76f9
SH		 685extern int tcp_set_default_congestion_control(const char *name);
686extern void tcp_get_default_congestion_control(char *name);
3ff825b2 687extern void tcp_get_available_congestion_control(char *buf, size_t len);
ce7bc3bf
SH		 688extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
689extern int tcp_set_allowed_congestion_control(char *allowed);
6687e988 690extern int tcp_set_congestion_control(struct sock *sk, const char *name);
40efc6fa 691extern void tcp_slow_start(struct tcp_sock *tp);
317a76f9 692
5f8ef48d 693extern struct tcp_congestion_ops tcp_init_congestion_ops;
6687e988 694extern u32 tcp_reno_ssthresh(struct sock *sk);
c3a05c60 695extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
72dc5b92 696extern u32 tcp_reno_min_cwnd(const struct sock *sk);
a8acfbac 697extern struct tcp_congestion_ops tcp_reno;
317a76f9 698
6687e988 699static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
317a76f9 700{
6687e988
ACM		 701 struct inet_connection_sock *icsk = inet_csk(sk);
702
703 if (icsk->icsk_ca_ops->set_state)
704 icsk->icsk_ca_ops->set_state(sk, ca_state);
705 icsk->icsk_ca_state = ca_state;
317a76f9
SH		 706}
707
6687e988 708static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
317a76f9 709{
6687e988
ACM		 710 const struct inet_connection_sock *icsk = inet_csk(sk);
711
712 if (icsk->icsk_ca_ops->cwnd_event)
713 icsk->icsk_ca_ops->cwnd_event(sk, event);
317a76f9
SH		 714}
715
e60402d0
IJ		 716/* These functions determine how the current flow behaves in respect of SACK
717 * handling. SACK is negotiated with the peer, and therefore it can vary
718 * between different flows.
719 *
720 * tcp_is_sack - SACK enabled
721 * tcp_is_reno - No SACK
722 * tcp_is_fack - FACK enabled, implies SACK enabled
723 */
724static inline int tcp_is_sack(const struct tcp_sock *tp)
725{
726 return tp->rx_opt.sack_ok;
727}
728
729static inline int tcp_is_reno(const struct tcp_sock *tp)
730{
731 return !tcp_is_sack(tp);
732}
733
734static inline int tcp_is_fack(const struct tcp_sock *tp)
735{
736 return tp->rx_opt.sack_ok & 2;
737}
738
739static inline void tcp_enable_fack(struct tcp_sock *tp)
740{
741 tp->rx_opt.sack_ok |= 2;
742}
743
83ae4088
IJ		 744static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
745{
746 return tp->sacked_out + tp->lost_out;
747}
748
1da177e4
LT		 749/* This determines how many packets are "in the network" to the best
750 * of our knowledge. In many cases it is conservative, but where
751 * detailed information is available from the receiver (via SACK
752 * blocks etc.) we can make more aggressive calculations.
753 *
754 * Use this for decisions involving congestion control, use just
755 * tp->packets_out to determine if the send queue is empty or not.
756 *
757 * Read this equation as:
758 *
759 * "Packets sent once on transmission queue" MINUS
760 * "Packets left network, but not honestly ACKed yet" PLUS
761 * "Packets fast retransmitted"
762 */
40efc6fa 763static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
1da177e4 764{
83ae4088 765 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
1da177e4
LT		 766}
767
882bebaa
IJ		 768extern int tcp_limit_reno_sacked(struct tcp_sock *tp);
769
1da177e4
LT		 770/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
771 * The exception is rate halving phase, when cwnd is decreasing towards
772 * ssthresh.
773 */
6687e988 774static inline __u32 tcp_current_ssthresh(const struct sock *sk)
1da177e4 775{
6687e988
ACM		 776 const struct tcp_sock *tp = tcp_sk(sk);
777 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
1da177e4
LT		 778 return tp->snd_ssthresh;
779 else
780 return max(tp->snd_ssthresh,
781 ((tp->snd_cwnd >> 1) +
782 (tp->snd_cwnd >> 2)));
783}
784
b9c4595b
IJ		 785/* Use define here intentionally to get WARN_ON location shown at the caller */
786#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
1da177e4 787
3cfe3baa 788extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
1da177e4
LT		 789extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
790
791/* Slow start with delack produces 3 packets of burst, so that
dd9e0dda
JH		 792 * it is safe "de facto". This will be the default - same as
793 * the default reordering threshold - but if reordering increases,
794 * we must be able to allow cwnd to burst at least this much in order
795 * to not pull it back when holes are filled.
1da177e4
LT		 796 */
797static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
798{
dd9e0dda 799 return tp->reordering;
1da177e4
LT		 800}
801
90840def
IJ		 802/* Returns end sequence number of the receiver's advertised window */
803static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
804{
805 return tp->snd_una + tp->snd_wnd;
806}
cea14e0e 807extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
f4805ede 808
c1bd24b7 809static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
40efc6fa 810 const struct sk_buff *skb)
1da177e4
LT		 811{
812 if (skb->len < mss)
813 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
814}
815
9e412ba7 816static inline void tcp_check_probe_timer(struct sock *sk)
1da177e4 817{
9e412ba7 818 struct tcp_sock *tp = tcp_sk(sk);
463c84b9 819 const struct inet_connection_sock *icsk = inet_csk(sk);
9e412ba7 820
463c84b9 821 if (!tp->packets_out && !icsk->icsk_pending)
3f421baa
ACM		 822 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
823 icsk->icsk_rto, TCP_RTO_MAX);
1da177e4
LT		 824}
825
9e412ba7 826static inline void tcp_push_pending_frames(struct sock *sk)
1da177e4 827{
9e412ba7
IJ		 828 struct tcp_sock *tp = tcp_sk(sk);
829
830 __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle);
1da177e4
LT		 831}
832
40efc6fa 833static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT		 834{
835 tp->snd_wl1 = seq;
836}
837
40efc6fa 838static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT		 839{
840 tp->snd_wl1 = seq;
841}
842
1da177e4
LT		 843/*
844 * Calculate(/check) TCP checksum
845 */
ba7808ea
FD		 846static inline __sum16 tcp_v4_check(int len, __be32 saddr,
847 __be32 daddr, __wsum base)
1da177e4
LT		 848{
849 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
850}
851
b51655b9 852static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1da177e4 853{
fb286bb2 854 return __skb_checksum_complete(skb);
1da177e4
LT		 855}
856
40efc6fa 857static inline int tcp_checksum_complete(struct sk_buff *skb)
1da177e4 858{
60476372 859 return !skb_csum_unnecessary(skb) &&
1da177e4
LT		 860 __tcp_checksum_complete(skb);
861}
862
863/* Prequeue for VJ style copy to user, combined with checksumming. */
864
40efc6fa 865static inline void tcp_prequeue_init(struct tcp_sock *tp)
1da177e4
LT		 866{
867 tp->ucopy.task = NULL;
868 tp->ucopy.len = 0;
869 tp->ucopy.memory = 0;
870 skb_queue_head_init(&tp->ucopy.prequeue);
97fc2f08
CL		 871#ifdef CONFIG_NET_DMA
872 tp->ucopy.dma_chan = NULL;
873 tp->ucopy.wakeup = 0;
874 tp->ucopy.pinned_list = NULL;
875 tp->ucopy.dma_cookie = 0;
876#endif
1da177e4
LT		 877}
878
879/* Packet is added to VJ-style prequeue for processing in process
880 * context, if a reader task is waiting. Apparently, this exciting
881 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
882 * failed somewhere. Latency? Burstiness? Well, at least now we will
883 * see, why it failed. 8)8) --ANK
884 *
885 * NOTE: is this not too big to inline?
886 */
40efc6fa 887static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
1da177e4
LT		 888{
889 struct tcp_sock *tp = tcp_sk(sk);
890
891 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
892 __skb_queue_tail(&tp->ucopy.prequeue, skb);
893 tp->ucopy.memory += skb->truesize;
894 if (tp->ucopy.memory > sk->sk_rcvbuf) {
895 struct sk_buff *skb1;
896
897 BUG_ON(sock_owned_by_user(sk));
898
899 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
900 sk->sk_backlog_rcv(sk, skb1);
901 NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED);
902 }
903
904 tp->ucopy.memory = 0;
905 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
906 wake_up_interruptible(sk->sk_sleep);
463c84b9
ACM		 907 if (!inet_csk_ack_scheduled(sk))
908 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
3f421baa
ACM		 909 (3 * TCP_RTO_MIN) / 4,
910 TCP_RTO_MAX);
1da177e4
LT		 911 }
912 return 1;
913 }
914 return 0;
915}
916
917
918#undef STATE_TRACE
919
920#ifdef STATE_TRACE
921static const char *statename[]={
922 "Unused","Established","Syn Sent","Syn Recv",
923 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
924 "Close Wait","Last ACK","Listen","Closing"
925};
926#endif
490d5046 927extern void tcp_set_state(struct sock *sk, int state);
1da177e4 928
4ac02bab 929extern void tcp_done(struct sock *sk);
1da177e4 930
40efc6fa 931static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1da177e4
LT		 932{
933 rx_opt->dsack = 0;
934 rx_opt->eff_sacks = 0;
935 rx_opt->num_sacks = 0;
936}
937
1da177e4
LT		 938/* Determine a window scaling and initial window to offer. */
939extern void tcp_select_initial_window(int __space, __u32 mss,
940 __u32 *rcv_wnd, __u32 *window_clamp,
941 int wscale_ok, __u8 *rcv_wscale);
942
943static inline int tcp_win_from_space(int space)
944{
945 return sysctl_tcp_adv_win_scale<=0 ?
946 (space>>(-sysctl_tcp_adv_win_scale)) :
947 space - (space>>sysctl_tcp_adv_win_scale);
948}
949
950/* Note: caller must be prepared to deal with negative returns */
951static inline int tcp_space(const struct sock *sk)
952{
953 return tcp_win_from_space(sk->sk_rcvbuf -
954 atomic_read(&sk->sk_rmem_alloc));
955}
956
957static inline int tcp_full_space(const struct sock *sk)
958{
959 return tcp_win_from_space(sk->sk_rcvbuf);
960}
961
40efc6fa
SH		 962static inline void tcp_openreq_init(struct request_sock *req,
963 struct tcp_options_received *rx_opt,
964 struct sk_buff *skb)
1da177e4 965{
2e6599cb
ACM		 966 struct inet_request_sock *ireq = inet_rsk(req);
967
1da177e4 968 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
4dfc2817 969 req->cookie_ts = 0;
2e6599cb 970 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1da177e4
LT		 971 req->mss = rx_opt->mss_clamp;
972 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
2e6599cb
ACM		 973 ireq->tstamp_ok = rx_opt->tstamp_ok;
974 ireq->sack_ok = rx_opt->sack_ok;
975 ireq->snd_wscale = rx_opt->snd_wscale;
976 ireq->wscale_ok = rx_opt->wscale_ok;
977 ireq->acked = 0;
978 ireq->ecn_ok = 0;
aa8223c7 979 ireq->rmt_port = tcp_hdr(skb)->source;
1da177e4
LT		 980}
981
982extern void tcp_enter_memory_pressure(void);
983
1da177e4
LT		 984static inline int keepalive_intvl_when(const struct tcp_sock *tp)
985{
986 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
987}
988
989static inline int keepalive_time_when(const struct tcp_sock *tp)
990{
991 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
992}
993
463c84b9 994static inline int tcp_fin_time(const struct sock *sk)
1da177e4 995{
463c84b9
ACM		 996 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
997 const int rto = inet_csk(sk)->icsk_rto;
1da177e4 998
463c84b9
ACM		 999 if (fin_timeout < (rto << 2) - (rto >> 1))
1000 fin_timeout = (rto << 2) - (rto >> 1);
1da177e4
LT		 1001
1002 return fin_timeout;
1003}
1004
1005static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1006{
1007 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1008 return 0;
9d729f72 1009 if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
1da177e4
LT		 1010 return 0;
1011
1012 /* RST segments are not recommended to carry timestamp,
1013 and, if they do, it is recommended to ignore PAWS because
1014 "their cleanup function should take precedence over timestamps."
1015 Certainly, it is mistake. It is necessary to understand the reasons
1016 of this constraint to relax it: if peer reboots, clock may go
1017 out-of-sync and half-open connections will not be reset.
1018 Actually, the problem would be not existing if all
1019 the implementations followed draft about maintaining clock
1020 via reboots. Linux-2.2 DOES NOT!
1021
1022 However, we can relax time bounds for RST segments to MSL.
1023 */
9d729f72 1024 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1da177e4
LT		 1025 return 0;
1026 return 1;
1027}
1028
1da177e4
LT		 1029#define TCP_CHECK_TIMER(sk) do { } while (0)
1030
1da177e4
LT		 1031static inline void tcp_mib_init(void)
1032{
1033 /* See RFC 2012 */
1034 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1035 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1036 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1037 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1038}
1039
5af4ec23 1040/* from STCP */
0800f170
DM		 1041static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1042{
6a438bbe
SH		 1043 tp->lost_skb_hint = NULL;
1044 tp->scoreboard_skb_hint = NULL;
1045 tp->retransmit_skb_hint = NULL;
1046 tp->forward_skb_hint = NULL;
b7689205
IJ		 1047}
1048
0800f170
DM		 1049static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1050{
b7689205 1051 tcp_clear_retrans_hints_partial(tp);
6a438bbe
SH		 1052}
1053
cfb6eeb4
YH		 1054/* MD5 Signature */
1055struct crypto_hash;
1056
1057/* - key database */
1058struct tcp_md5sig_key {
1059 u8 *key;
1060 u8 keylen;
1061};
1062
1063struct tcp4_md5sig_key {
f8ab18d2 1064 struct tcp_md5sig_key base;
cfb6eeb4
YH		 1065 __be32 addr;
1066};
1067
1068struct tcp6_md5sig_key {
f8ab18d2 1069 struct tcp_md5sig_key base;
cfb6eeb4
YH		 1070#if 0
1071 u32 scope_id; /* XXX */
1072#endif
1073 struct in6_addr addr;
1074};
1075
1076/* - sock block */
1077struct tcp_md5sig_info {
1078 struct tcp4_md5sig_key *keys4;
1079#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1080 struct tcp6_md5sig_key *keys6;
1081 u32 entries6;
1082 u32 alloced6;
1083#endif
1084 u32 entries4;
1085 u32 alloced4;
1086};
1087
1088/* - pseudo header */
1089struct tcp4_pseudohdr {
1090 __be32 saddr;
1091 __be32 daddr;
1092 __u8 pad;
1093 __u8 protocol;
1094 __be16 len;
1095};
1096
1097struct tcp6_pseudohdr {
1098 struct in6_addr saddr;
1099 struct in6_addr daddr;
1100 __be32 len;
1101 __be32 protocol; /* including padding */
1102};
1103
1104union tcp_md5sum_block {
1105 struct tcp4_pseudohdr ip4;
1106#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1107 struct tcp6_pseudohdr ip6;
1108#endif
1109};
1110
1111/* - pool: digest algorithm, hash description and scratch buffer */
1112struct tcp_md5sig_pool {
1113 struct hash_desc md5_desc;
1114 union tcp_md5sum_block md5_blk;
1115};
1116
1117#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1118
1119/* - functions */
8d26d76d
YH		 1120extern int tcp_calc_md5_hash(char *md5_hash,
1121 struct tcp_md5sig_key *key,
1122 int bplen,
1123 struct tcphdr *th,
1124 unsigned int tcplen,
1125 struct tcp_md5sig_pool *hp);
1126
cfb6eeb4
YH		 1127extern int tcp_v4_calc_md5_hash(char *md5_hash,
1128 struct tcp_md5sig_key *key,
1129 struct sock *sk,
1130 struct dst_entry *dst,
1131 struct request_sock *req,
1132 struct tcphdr *th,
9cb5734e 1133 unsigned int tcplen);
cfb6eeb4
YH		 1134extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1135 struct sock *addr_sk);
1136
1137extern int tcp_v4_md5_do_add(struct sock *sk,
1138 __be32 addr,
1139 u8 *newkey,
1140 u8 newkeylen);
1141
1142extern int tcp_v4_md5_do_del(struct sock *sk,
8e5200f5 1143 __be32 addr);
cfb6eeb4
YH		 1144
1145extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1146extern void tcp_free_md5sig_pool(void);
1147
1148extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1149extern void __tcp_put_md5sig_pool(void);
1150
1151static inline
1152struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1153{
1154 int cpu = get_cpu();
1155 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1156 if (!ret)
1157 put_cpu();
1158 return ret;
1159}
1160
1161static inline void tcp_put_md5sig_pool(void)
1162{
1163 __tcp_put_md5sig_pool();
1164 put_cpu();
1165}
1166
fe067e8a
DM		 1167/* write queue abstraction */
1168static inline void tcp_write_queue_purge(struct sock *sk)
1169{
1170 struct sk_buff *skb;
1171
1172 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
3ab224be
HA		 1173 sk_wmem_free_skb(sk, skb);
1174 sk_mem_reclaim(sk);
fe067e8a
DM		 1175}
1176
1177static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1178{
1179 struct sk_buff *skb = sk->sk_write_queue.next;
1180 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1181 return NULL;
1182 return skb;
1183}
1184
1185static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1186{
1187 struct sk_buff *skb = sk->sk_write_queue.prev;
1188 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1189 return NULL;
1190 return skb;
1191}
1192
1193static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1194{
1195 return skb->next;
1196}
1197
1198#define tcp_for_write_queue(skb, sk) \
1199 for (skb = (sk)->sk_write_queue.next; \
1200 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1201 skb = skb->next)
1202
1203#define tcp_for_write_queue_from(skb, sk) \
1204 for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\
1205 skb = skb->next)
1206
234b6860
IJ		 1207#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
1208 for (tmp = skb->next; \
1209 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1210 skb = tmp, tmp = skb->next)
1211
fe067e8a
DM		 1212static inline struct sk_buff *tcp_send_head(struct sock *sk)
1213{
1214 return sk->sk_send_head;
1215}
1216
1217static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1218{
1219 sk->sk_send_head = skb->next;
1220 if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue)
1221 sk->sk_send_head = NULL;
1222}
1223
1224static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1225{
1226 if (sk->sk_send_head == skb_unlinked)
1227 sk->sk_send_head = NULL;
1228}
1229
1230static inline void tcp_init_send_head(struct sock *sk)
1231{
1232 sk->sk_send_head = NULL;
1233}
1234
1235static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1236{
1237 __skb_queue_tail(&sk->sk_write_queue, skb);
1238}
1239
1240static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1241{
1242 __tcp_add_write_queue_tail(sk, skb);
1243
1244 /* Queue it, remembering where we must start sending. */
6859d494 1245 if (sk->sk_send_head == NULL) {
fe067e8a 1246 sk->sk_send_head = skb;
6859d494
IJ		 1247
1248 if (tcp_sk(sk)->highest_sack == NULL)
1249 tcp_sk(sk)->highest_sack = skb;
1250 }
fe067e8a
DM		 1251}
1252
1253static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1254{
1255 __skb_queue_head(&sk->sk_write_queue, skb);
1256}
1257
1258/* Insert buff after skb on the write queue of sk. */
1259static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1260 struct sk_buff *buff,
1261 struct sock *sk)
1262{
7de6c033 1263 __skb_queue_after(&sk->sk_write_queue, skb, buff);
fe067e8a
DM		 1264}
1265
1266/* Insert skb between prev and next on the write queue of sk. */
1267static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1268 struct sk_buff *skb,
1269 struct sock *sk)
1270{
1271 __skb_insert(new, skb->prev, skb, &sk->sk_write_queue);