]> bbs.cooldavid.org Git - net-next-2.6.git/blame - include/net/tcp.h
syncookies: do not store rcv_wscale in tcp timestamp
[net-next-2.6.git] / include / net / tcp.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the TCP module.
7 *
8 * Version: @(#)tcp.h 1.0.5 05/23/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 *
13 * This program is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU General Public License
15 * as published by the Free Software Foundation; either version
16 * 2 of the License, or (at your option) any later version.
17 */
18#ifndef _TCP_H
19#define _TCP_H
20
21#define TCP_DEBUG 1
22#define FASTRETRANS_DEBUG 1
23
1da177e4
LT
24#include <linux/list.h>
25#include <linux/tcp.h>
26#include <linux/slab.h>
27#include <linux/cache.h>
28#include <linux/percpu.h>
fb286bb2 29#include <linux/skbuff.h>
97fc2f08 30#include <linux/dmaengine.h>
cfb6eeb4 31#include <linux/crypto.h>
c6aefafb 32#include <linux/cryptohash.h>
435cf559 33#include <linux/kref.h>
3f421baa
ACM
34
35#include <net/inet_connection_sock.h>
295ff7ed 36#include <net/inet_timewait_sock.h>
77d8bf9c 37#include <net/inet_hashtables.h>
1da177e4 38#include <net/checksum.h>
2e6599cb 39#include <net/request_sock.h>
1da177e4
LT
40#include <net/sock.h>
41#include <net/snmp.h>
42#include <net/ip.h>
c752f073 43#include <net/tcp_states.h>
bdf1ee5d 44#include <net/inet_ecn.h>
0c266898 45#include <net/dst.h>
c752f073 46
1da177e4
LT
47#include <linux/seq_file.h>
48
6e04e021 49extern struct inet_hashinfo tcp_hashinfo;
1da177e4 50
dd24c001 51extern struct percpu_counter tcp_orphan_count;
1da177e4 52extern void tcp_time_wait(struct sock *sk, int state, int timeo);
1da177e4 53
1da177e4 54#define MAX_TCP_HEADER (128 + MAX_HEADER)
33ad798c 55#define MAX_TCP_OPTION_SPACE 40
1da177e4
LT
56
57/*
58 * Never offer a window over 32767 without using window scaling. Some
59 * poor stacks do signed 16bit maths!
60 */
61#define MAX_TCP_WINDOW 32767U
62
63/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
64#define TCP_MIN_MSS 88U
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
142
143#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
1da177e4
LT
144
145#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
146#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
147 * after this time. It should be equal
148 * (or greater than) TCP_TIMEWAIT_LEN
149 * to provide reliability equal to one
150 * provided by timewait state.
151 */
152#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
153 * timestamps. It must be less than
154 * minimal timewait lifetime.
155 */
1da177e4
LT
156/*
157 * TCP option
158 */
159
160#define TCPOPT_NOP 1 /* Padding */
161#define TCPOPT_EOL 0 /* End of options */
162#define TCPOPT_MSS 2 /* Segment size negotiating */
163#define TCPOPT_WINDOW 3 /* Window scaling */
164#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
165#define TCPOPT_SACK 5 /* SACK Block */
166#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
cfb6eeb4 167#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
435cf559 168#define TCPOPT_COOKIE 253 /* Cookie extension (experimental) */
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
435cf559
WAS
179#define TCPOLEN_COOKIE_BASE 2 /* Cookie-less header extension */
180#define TCPOLEN_COOKIE_PAIR 3 /* Cookie pair header extension */
181#define TCPOLEN_COOKIE_MIN (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MIN)
182#define TCPOLEN_COOKIE_MAX (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MAX)
1da177e4
LT
183
184/* But this is what stacks really send out. */
185#define TCPOLEN_TSTAMP_ALIGNED 12
186#define TCPOLEN_WSCALE_ALIGNED 4
187#define TCPOLEN_SACKPERM_ALIGNED 4
188#define TCPOLEN_SACK_BASE 2
189#define TCPOLEN_SACK_BASE_ALIGNED 4
190#define TCPOLEN_SACK_PERBLOCK 8
cfb6eeb4 191#define TCPOLEN_MD5SIG_ALIGNED 20
33ad798c 192#define TCPOLEN_MSS_ALIGNED 4
1da177e4 193
1da177e4
LT
194/* Flags in tp->nonagle */
195#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
196#define TCP_NAGLE_CORK 2 /* Socket is corked */
caa20d9a 197#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
1da177e4 198
36e31b0a
AP
199/* TCP thin-stream limits */
200#define TCP_THIN_LINEAR_RETRIES 6 /* After 6 linear retries, do exp. backoff */
201
295ff7ed
ACM
202extern struct inet_timewait_death_row tcp_death_row;
203
1da177e4 204/* sysctl variables for tcp */
1da177e4
LT
205extern int sysctl_tcp_timestamps;
206extern int sysctl_tcp_window_scaling;
207extern int sysctl_tcp_sack;
208extern int sysctl_tcp_fin_timeout;
1da177e4
LT
209extern int sysctl_tcp_keepalive_time;
210extern int sysctl_tcp_keepalive_probes;
211extern int sysctl_tcp_keepalive_intvl;
212extern int sysctl_tcp_syn_retries;
213extern int sysctl_tcp_synack_retries;
214extern int sysctl_tcp_retries1;
215extern int sysctl_tcp_retries2;
216extern int sysctl_tcp_orphan_retries;
217extern int sysctl_tcp_syncookies;
218extern int sysctl_tcp_retrans_collapse;
219extern int sysctl_tcp_stdurg;
220extern int sysctl_tcp_rfc1337;
221extern int sysctl_tcp_abort_on_overflow;
222extern int sysctl_tcp_max_orphans;
1da177e4
LT
223extern int sysctl_tcp_fack;
224extern int sysctl_tcp_reordering;
225extern int sysctl_tcp_ecn;
226extern int sysctl_tcp_dsack;
227extern int sysctl_tcp_mem[3];
228extern int sysctl_tcp_wmem[3];
229extern int sysctl_tcp_rmem[3];
230extern int sysctl_tcp_app_win;
231extern int sysctl_tcp_adv_win_scale;
232extern int sysctl_tcp_tw_reuse;
233extern int sysctl_tcp_frto;
3cfe3baa 234extern int sysctl_tcp_frto_response;
1da177e4 235extern int sysctl_tcp_low_latency;
95937825 236extern int sysctl_tcp_dma_copybreak;
1da177e4 237extern int sysctl_tcp_nometrics_save;
1da177e4
LT
238extern int sysctl_tcp_moderate_rcvbuf;
239extern int sysctl_tcp_tso_win_divisor;
9772efb9 240extern int sysctl_tcp_abc;
5d424d5a
JH
241extern int sysctl_tcp_mtu_probing;
242extern int sysctl_tcp_base_mss;
15d99e02 243extern int sysctl_tcp_workaround_signed_windows;
35089bb2 244extern int sysctl_tcp_slow_start_after_idle;
886236c1 245extern int sysctl_tcp_max_ssthresh;
519855c5 246extern int sysctl_tcp_cookie_size;
36e31b0a 247extern int sysctl_tcp_thin_linear_timeouts;
7e380175 248extern int sysctl_tcp_thin_dupack;
1da177e4
LT
249
250extern atomic_t tcp_memory_allocated;
1748376b 251extern struct percpu_counter tcp_sockets_allocated;
1da177e4
LT
252extern int tcp_memory_pressure;
253
1da177e4
LT
254/*
255 * The next routines deal with comparing 32 bit unsigned ints
256 * and worry about wraparound (automatic with unsigned arithmetic).
257 */
258
259static inline int before(__u32 seq1, __u32 seq2)
260{
0d630cc0 261 return (__s32)(seq1-seq2) < 0;
1da177e4 262}
9a036b9c 263#define after(seq2, seq1) before(seq1, seq2)
1da177e4
LT
264
265/* is s2<=s1<=s3 ? */
266static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
267{
268 return seq3 - seq2 >= seq1 - seq2;
269}
270
e4fd5da3
PE
271static inline int tcp_too_many_orphans(struct sock *sk, int num)
272{
273 return (num > sysctl_tcp_max_orphans) ||
274 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
275 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
276}
1da177e4 277
a0f82f64
FW
278/* syncookies: remember time of last synqueue overflow */
279static inline void tcp_synq_overflow(struct sock *sk)
280{
281 tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies;
282}
283
284/* syncookies: no recent synqueue overflow on this listening socket? */
285static inline int tcp_synq_no_recent_overflow(const struct sock *sk)
286{
287 unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp;
288 return time_after(jiffies, last_overflow + TCP_TIMEOUT_INIT);
289}
290
1da177e4
LT
291extern struct proto tcp_prot;
292
57ef42d5
PE
293#define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field)
294#define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
295#define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
296#define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val)
aa2ea058 297#define TCP_ADD_STATS(net, field, val) SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val)
1da177e4 298
1da177e4
LT
299extern void tcp_v4_err(struct sk_buff *skb, u32);
300
301extern void tcp_shutdown (struct sock *sk, int how);
302
303extern int tcp_v4_rcv(struct sk_buff *skb);
304
305extern int tcp_v4_remember_stamp(struct sock *sk);
306
8feaf0c0 307extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
1da177e4 308
3516ffb0 309extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
1da177e4
LT
310 struct msghdr *msg, size_t size);
311extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
312
313extern int tcp_ioctl(struct sock *sk,
314 int cmd,
315 unsigned long arg);
316
317extern int tcp_rcv_state_process(struct sock *sk,
318 struct sk_buff *skb,
319 struct tcphdr *th,
320 unsigned len);
321
322extern int tcp_rcv_established(struct sock *sk,
323 struct sk_buff *skb,
324 struct tcphdr *th,
325 unsigned len);
326
327extern void tcp_rcv_space_adjust(struct sock *sk);
328
0e4b4992
CL
329extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
330
6d6ee43e
ACM
331extern int tcp_twsk_unique(struct sock *sk,
332 struct sock *sktw, void *twp);
333
cfb6eeb4
YH
334extern void tcp_twsk_destructor(struct sock *sk);
335
9c55e01c
JA
336extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
337 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
338
463c84b9
ACM
339static inline void tcp_dec_quickack_mode(struct sock *sk,
340 const unsigned int pkts)
1da177e4 341{
463c84b9 342 struct inet_connection_sock *icsk = inet_csk(sk);
fc6415bc 343
463c84b9
ACM
344 if (icsk->icsk_ack.quick) {
345 if (pkts >= icsk->icsk_ack.quick) {
346 icsk->icsk_ack.quick = 0;
fc6415bc 347 /* Leaving quickack mode we deflate ATO. */
463c84b9 348 icsk->icsk_ack.ato = TCP_ATO_MIN;
fc6415bc 349 } else
463c84b9 350 icsk->icsk_ack.quick -= pkts;
1da177e4
LT
351 }
352}
353
463c84b9 354extern void tcp_enter_quickack_mode(struct sock *sk);
1da177e4 355
bdf1ee5d
IJ
356#define TCP_ECN_OK 1
357#define TCP_ECN_QUEUE_CWR 2
358#define TCP_ECN_DEMAND_CWR 4
359
360static __inline__ void
361TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
362{
363 if (sysctl_tcp_ecn && th->ece && th->cwr)
364 inet_rsk(req)->ecn_ok = 1;
365}
366
fd2c3ef7 367enum tcp_tw_status {
1da177e4
LT
368 TCP_TW_SUCCESS = 0,
369 TCP_TW_RST = 1,
370 TCP_TW_ACK = 2,
371 TCP_TW_SYN = 3
372};
373
374
8feaf0c0 375extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
1da177e4 376 struct sk_buff *skb,
8feaf0c0 377 const struct tcphdr *th);
1da177e4
LT
378
379extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
60236fdd
ACM
380 struct request_sock *req,
381 struct request_sock **prev);
1da177e4
LT
382extern int tcp_child_process(struct sock *parent,
383 struct sock *child,
384 struct sk_buff *skb);
46d0de4e 385extern int tcp_use_frto(struct sock *sk);
1da177e4
LT
386extern void tcp_enter_frto(struct sock *sk);
387extern void tcp_enter_loss(struct sock *sk, int how);
388extern void tcp_clear_retrans(struct tcp_sock *tp);
389extern void tcp_update_metrics(struct sock *sk);
390
391extern void tcp_close(struct sock *sk,
392 long timeout);
1da177e4
LT
393extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
394
395extern int tcp_getsockopt(struct sock *sk, int level,
396 int optname,
397 char __user *optval,
398 int __user *optlen);
399extern int tcp_setsockopt(struct sock *sk, int level,
400 int optname, char __user *optval,
b7058842 401 unsigned int optlen);
3fdadf7d
DM
402extern int compat_tcp_getsockopt(struct sock *sk,
403 int level, int optname,
404 char __user *optval, int __user *optlen);
405extern int compat_tcp_setsockopt(struct sock *sk,
406 int level, int optname,
b7058842 407 char __user *optval, unsigned int optlen);
1da177e4 408extern void tcp_set_keepalive(struct sock *sk, int val);
72659ecc
OP
409extern void tcp_syn_ack_timeout(struct sock *sk,
410 struct request_sock *req);
1da177e4
LT
411extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
412 struct msghdr *msg,
413 size_t len, int nonblock,
414 int flags, int *addr_len);
415
1da177e4
LT
416extern void tcp_parse_options(struct sk_buff *skb,
417 struct tcp_options_received *opt_rx,
4957faad 418 u8 **hvpp,
bb5b7c11 419 int estab);
1da177e4 420
7d5d5525
YH
421extern u8 *tcp_parse_md5sig_option(struct tcphdr *th);
422
1da177e4
LT
423/*
424 * TCP v4 functions exported for the inet6 API
425 */
426
bb296246 427extern void tcp_v4_send_check(struct sock *sk,
1da177e4
LT
428 struct sk_buff *skb);
429
430extern int tcp_v4_conn_request(struct sock *sk,
431 struct sk_buff *skb);
432
433extern struct sock * tcp_create_openreq_child(struct sock *sk,
60236fdd 434 struct request_sock *req,
1da177e4
LT
435 struct sk_buff *skb);
436
437extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
438 struct sk_buff *skb,
60236fdd 439 struct request_sock *req,
1da177e4
LT
440 struct dst_entry *dst);
441
442extern int tcp_v4_do_rcv(struct sock *sk,
443 struct sk_buff *skb);
444
445extern int tcp_v4_connect(struct sock *sk,
446 struct sockaddr *uaddr,
447 int addr_len);
448
449extern int tcp_connect(struct sock *sk);
450
451extern struct sk_buff * tcp_make_synack(struct sock *sk,
452 struct dst_entry *dst,
e6b4d113
WAS
453 struct request_sock *req,
454 struct request_values *rvp);
1da177e4
LT
455
456extern int tcp_disconnect(struct sock *sk, int flags);
457
1da177e4 458
1da177e4 459/* From syncookies.c */
2051f11f 460extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
1da177e4
LT
461extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
462 struct ip_options *opt);
463extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
464 __u16 *mss);
465
4dfc2817 466extern __u32 cookie_init_timestamp(struct request_sock *req);
8c763681 467extern bool cookie_check_timestamp(struct tcp_options_received *tcp_opt);
4dfc2817 468
c6aefafb
GG
469/* From net/ipv6/syncookies.c */
470extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
471extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb,
472 __u16 *mss);
473
1da177e4
LT
474/* tcp_output.c */
475
9e412ba7
IJ
476extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
477 int nonagle);
478extern int tcp_may_send_now(struct sock *sk);
1da177e4 479extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
f1ecd5d9 480extern void tcp_retransmit_timer(struct sock *sk);
1da177e4
LT
481extern void tcp_xmit_retransmit_queue(struct sock *);
482extern void tcp_simple_retransmit(struct sock *);
483extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
6475be16 484extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
1da177e4
LT
485
486extern void tcp_send_probe0(struct sock *);
487extern void tcp_send_partial(struct sock *);
488extern int tcp_write_wakeup(struct sock *);
489extern void tcp_send_fin(struct sock *sk);
dd0fc66f 490extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
1da177e4 491extern int tcp_send_synack(struct sock *);
c1b4a7e6 492extern void tcp_push_one(struct sock *, unsigned int mss_now);
1da177e4
LT
493extern void tcp_send_ack(struct sock *sk);
494extern void tcp_send_delayed_ack(struct sock *sk);
495
a762a980
DM
496/* tcp_input.c */
497extern void tcp_cwnd_application_limited(struct sock *sk);
498
1da177e4
LT
499/* tcp_timer.c */
500extern void tcp_init_xmit_timers(struct sock *);
463c84b9
ACM
501static inline void tcp_clear_xmit_timers(struct sock *sk)
502{
503 inet_csk_clear_xmit_timers(sk);
504}
1da177e4 505
1da177e4 506extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
0c54b85f
IJ
507extern unsigned int tcp_current_mss(struct sock *sk);
508
509/* Bound MSS / TSO packet size with the half of the window */
510static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
511{
512 if (tp->max_window && pktsize > (tp->max_window >> 1))
513 return max(tp->max_window >> 1, 68U - tp->tcp_header_len);
514 else
515 return pktsize;
516}
1da177e4 517
17b085ea 518/* tcp.c */
1da177e4
LT
519extern void tcp_get_info(struct sock *, struct tcp_info *);
520
521/* Read 'sendfile()'-style from a TCP socket */
522typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
523 unsigned int, size_t);
524extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
525 sk_read_actor_t recv_actor);
526
40efc6fa 527extern void tcp_initialize_rcv_mss(struct sock *sk);
1da177e4 528
5d424d5a
JH
529extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
530extern int tcp_mss_to_mtu(struct sock *sk, int mss);
531extern void tcp_mtup_init(struct sock *sk);
532
f1ecd5d9
DL
533static inline void tcp_bound_rto(const struct sock *sk)
534{
535 if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX)
536 inet_csk(sk)->icsk_rto = TCP_RTO_MAX;
537}
538
539static inline u32 __tcp_set_rto(const struct tcp_sock *tp)
540{
541 return (tp->srtt >> 3) + tp->rttvar;
542}
543
40efc6fa 544static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
1da177e4
LT
545{
546 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
547 ntohl(TCP_FLAG_ACK) |
548 snd_wnd);
549}
550
40efc6fa 551static inline void tcp_fast_path_on(struct tcp_sock *tp)
1da177e4
LT
552{
553 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
554}
555
9e412ba7 556static inline void tcp_fast_path_check(struct sock *sk)
1da177e4 557{
9e412ba7
IJ
558 struct tcp_sock *tp = tcp_sk(sk);
559
b03efcfb 560 if (skb_queue_empty(&tp->out_of_order_queue) &&
1da177e4
LT
561 tp->rcv_wnd &&
562 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
563 !tp->urg_data)
564 tcp_fast_path_on(tp);
565}
566
0c266898
SS
567/* Compute the actual rto_min value */
568static inline u32 tcp_rto_min(struct sock *sk)
569{
570 struct dst_entry *dst = __sk_dst_get(sk);
571 u32 rto_min = TCP_RTO_MIN;
572
573 if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
574 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
575 return rto_min;
576}
577
1da177e4
LT
578/* Compute the actual receive window we are currently advertising.
579 * Rcv_nxt can be after the window if our peer push more data
580 * than the offered window.
581 */
40efc6fa 582static inline u32 tcp_receive_window(const struct tcp_sock *tp)
1da177e4
LT
583{
584 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
585
586 if (win < 0)
587 win = 0;
588 return (u32) win;
589}
590
591/* Choose a new window, without checks for shrinking, and without
592 * scaling applied to the result. The caller does these things
593 * if necessary. This is a "raw" window selection.
594 */
595extern u32 __tcp_select_window(struct sock *sk);
596
597/* TCP timestamps are only 32-bits, this causes a slight
598 * complication on 64-bit systems since we store a snapshot
31f34269
SH
599 * of jiffies in the buffer control blocks below. We decided
600 * to use only the low 32-bits of jiffies and hide the ugly
1da177e4
LT
601 * casts with the following macro.
602 */
603#define tcp_time_stamp ((__u32)(jiffies))
604
a3433f35
CG
605#define tcp_flag_byte(th) (((u_int8_t *)th)[13])
606
607#define TCPHDR_FIN 0x01
608#define TCPHDR_SYN 0x02
609#define TCPHDR_RST 0x04
610#define TCPHDR_PSH 0x08
611#define TCPHDR_ACK 0x10
612#define TCPHDR_URG 0x20
613#define TCPHDR_ECE 0x40
614#define TCPHDR_CWR 0x80
615
caa20d9a 616/* This is what the send packet queuing engine uses to pass
1da177e4
LT
617 * TCP per-packet control information to the transmission
618 * code. We also store the host-order sequence numbers in
619 * here too. This is 36 bytes on 32-bit architectures,
620 * 40 bytes on 64-bit machines, if this grows please adjust
621 * skbuff.h:skbuff->cb[xxx] size appropriately.
622 */
623struct tcp_skb_cb {
624 union {
625 struct inet_skb_parm h4;
626#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
627 struct inet6_skb_parm h6;
628#endif
629 } header; /* For incoming frames */
630 __u32 seq; /* Starting sequence number */
631 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
632 __u32 when; /* used to compute rtt's */
633 __u8 flags; /* TCP header flags. */
1da177e4
LT
634 __u8 sacked; /* State flags for SACK/FACK. */
635#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
636#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
637#define TCPCB_LOST 0x04 /* SKB is lost */
638#define TCPCB_TAGBITS 0x07 /* All tag bits */
639
640#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
641#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
642
1da177e4
LT
643 __u32 ack_seq; /* Sequence number ACK'd */
644};
645
646#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
647
1da177e4
LT
648/* Due to TSO, an SKB can be composed of multiple actual
649 * packets. To keep these tracked properly, we use this.
650 */
651static inline int tcp_skb_pcount(const struct sk_buff *skb)
652{
7967168c 653 return skb_shinfo(skb)->gso_segs;
1da177e4
LT
654}
655
656/* This is valid iff tcp_skb_pcount() > 1. */
657static inline int tcp_skb_mss(const struct sk_buff *skb)
658{
7967168c 659 return skb_shinfo(skb)->gso_size;
1da177e4
LT
660}
661
317a76f9
SH
662/* Events passed to congestion control interface */
663enum tcp_ca_event {
664 CA_EVENT_TX_START, /* first transmit when no packets in flight */
665 CA_EVENT_CWND_RESTART, /* congestion window restart */
666 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
667 CA_EVENT_FRTO, /* fast recovery timeout */
668 CA_EVENT_LOSS, /* loss timeout */
669 CA_EVENT_FAST_ACK, /* in sequence ack */
670 CA_EVENT_SLOW_ACK, /* other ack */
671};
672
673/*
674 * Interface for adding new TCP congestion control handlers
675 */
676#define TCP_CA_NAME_MAX 16
3ff825b2
SH
677#define TCP_CA_MAX 128
678#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
679
164891aa
SH
680#define TCP_CONG_NON_RESTRICTED 0x1
681#define TCP_CONG_RTT_STAMP 0x2
682
317a76f9
SH
683struct tcp_congestion_ops {
684 struct list_head list;
164891aa 685 unsigned long flags;
317a76f9
SH
686
687 /* initialize private data (optional) */
6687e988 688 void (*init)(struct sock *sk);
317a76f9 689 /* cleanup private data (optional) */
6687e988 690 void (*release)(struct sock *sk);
317a76f9
SH
691
692 /* return slow start threshold (required) */
6687e988 693 u32 (*ssthresh)(struct sock *sk);
317a76f9 694 /* lower bound for congestion window (optional) */
72dc5b92 695 u32 (*min_cwnd)(const struct sock *sk);
317a76f9 696 /* do new cwnd calculation (required) */
c3a05c60 697 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
317a76f9 698 /* call before changing ca_state (optional) */
6687e988 699 void (*set_state)(struct sock *sk, u8 new_state);
317a76f9 700 /* call when cwnd event occurs (optional) */
6687e988 701 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
317a76f9 702 /* new value of cwnd after loss (optional) */
6687e988 703 u32 (*undo_cwnd)(struct sock *sk);
317a76f9 704 /* hook for packet ack accounting (optional) */
30cfd0ba 705 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
73c1f4a0 706 /* get info for inet_diag (optional) */
6687e988 707 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
317a76f9
SH
708
709 char name[TCP_CA_NAME_MAX];
710 struct module *owner;
711};
712
713extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
714extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
715
6687e988
ACM
716extern void tcp_init_congestion_control(struct sock *sk);
717extern void tcp_cleanup_congestion_control(struct sock *sk);
317a76f9
SH
718extern int tcp_set_default_congestion_control(const char *name);
719extern void tcp_get_default_congestion_control(char *name);
3ff825b2 720extern void tcp_get_available_congestion_control(char *buf, size_t len);
ce7bc3bf
SH
721extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
722extern int tcp_set_allowed_congestion_control(char *allowed);
6687e988 723extern int tcp_set_congestion_control(struct sock *sk, const char *name);
40efc6fa 724extern void tcp_slow_start(struct tcp_sock *tp);
758ce5c8 725extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w);
317a76f9 726
5f8ef48d 727extern struct tcp_congestion_ops tcp_init_congestion_ops;
6687e988 728extern u32 tcp_reno_ssthresh(struct sock *sk);
c3a05c60 729extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
72dc5b92 730extern u32 tcp_reno_min_cwnd(const struct sock *sk);
a8acfbac 731extern struct tcp_congestion_ops tcp_reno;
317a76f9 732
6687e988 733static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
317a76f9 734{
6687e988
ACM
735 struct inet_connection_sock *icsk = inet_csk(sk);
736
737 if (icsk->icsk_ca_ops->set_state)
738 icsk->icsk_ca_ops->set_state(sk, ca_state);
739 icsk->icsk_ca_state = ca_state;
317a76f9
SH
740}
741
6687e988 742static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
317a76f9 743{
6687e988
ACM
744 const struct inet_connection_sock *icsk = inet_csk(sk);
745
746 if (icsk->icsk_ca_ops->cwnd_event)
747 icsk->icsk_ca_ops->cwnd_event(sk, event);
317a76f9
SH
748}
749
e60402d0
IJ
750/* These functions determine how the current flow behaves in respect of SACK
751 * handling. SACK is negotiated with the peer, and therefore it can vary
752 * between different flows.
753 *
754 * tcp_is_sack - SACK enabled
755 * tcp_is_reno - No SACK
756 * tcp_is_fack - FACK enabled, implies SACK enabled
757 */
758static inline int tcp_is_sack(const struct tcp_sock *tp)
759{
760 return tp->rx_opt.sack_ok;
761}
762
763static inline int tcp_is_reno(const struct tcp_sock *tp)
764{
765 return !tcp_is_sack(tp);
766}
767
768static inline int tcp_is_fack(const struct tcp_sock *tp)
769{
770 return tp->rx_opt.sack_ok & 2;
771}
772
773static inline void tcp_enable_fack(struct tcp_sock *tp)
774{
775 tp->rx_opt.sack_ok |= 2;
776}
777
83ae4088
IJ
778static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
779{
780 return tp->sacked_out + tp->lost_out;
781}
782
1da177e4
LT
783/* This determines how many packets are "in the network" to the best
784 * of our knowledge. In many cases it is conservative, but where
785 * detailed information is available from the receiver (via SACK
786 * blocks etc.) we can make more aggressive calculations.
787 *
788 * Use this for decisions involving congestion control, use just
789 * tp->packets_out to determine if the send queue is empty or not.
790 *
791 * Read this equation as:
792 *
793 * "Packets sent once on transmission queue" MINUS
794 * "Packets left network, but not honestly ACKed yet" PLUS
795 * "Packets fast retransmitted"
796 */
40efc6fa 797static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
1da177e4 798{
83ae4088 799 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
1da177e4
LT
800}
801
0b6a05c1
IJ
802#define TCP_INFINITE_SSTHRESH 0x7fffffff
803
804static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp)
805{
806 return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH;
807}
808
1da177e4
LT
809/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
810 * The exception is rate halving phase, when cwnd is decreasing towards
811 * ssthresh.
812 */
6687e988 813static inline __u32 tcp_current_ssthresh(const struct sock *sk)
1da177e4 814{
6687e988
ACM
815 const struct tcp_sock *tp = tcp_sk(sk);
816 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
1da177e4
LT
817 return tp->snd_ssthresh;
818 else
819 return max(tp->snd_ssthresh,
820 ((tp->snd_cwnd >> 1) +
821 (tp->snd_cwnd >> 2)));
822}
823
b9c4595b
IJ
824/* Use define here intentionally to get WARN_ON location shown at the caller */
825#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
1da177e4 826
3cfe3baa 827extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
1da177e4
LT
828extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
829
830/* Slow start with delack produces 3 packets of burst, so that
dd9e0dda
JH
831 * it is safe "de facto". This will be the default - same as
832 * the default reordering threshold - but if reordering increases,
833 * we must be able to allow cwnd to burst at least this much in order
834 * to not pull it back when holes are filled.
1da177e4
LT
835 */
836static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
837{
dd9e0dda 838 return tp->reordering;
1da177e4
LT
839}
840
90840def
IJ
841/* Returns end sequence number of the receiver's advertised window */
842static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
843{
844 return tp->snd_una + tp->snd_wnd;
845}
cea14e0e 846extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
f4805ede 847
c1bd24b7 848static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
40efc6fa 849 const struct sk_buff *skb)
1da177e4
LT
850{
851 if (skb->len < mss)
852 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
853}
854
9e412ba7 855static inline void tcp_check_probe_timer(struct sock *sk)
1da177e4 856{
9e412ba7 857 struct tcp_sock *tp = tcp_sk(sk);
463c84b9 858 const struct inet_connection_sock *icsk = inet_csk(sk);
9e412ba7 859
463c84b9 860 if (!tp->packets_out && !icsk->icsk_pending)
3f421baa
ACM
861 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
862 icsk->icsk_rto, TCP_RTO_MAX);
1da177e4
LT
863}
864
ee7537b6 865static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
1da177e4
LT
866{
867 tp->snd_wl1 = seq;
868}
869
ee7537b6 870static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
1da177e4
LT
871{
872 tp->snd_wl1 = seq;
873}
874
1da177e4
LT
875/*
876 * Calculate(/check) TCP checksum
877 */
ba7808ea
FD
878static inline __sum16 tcp_v4_check(int len, __be32 saddr,
879 __be32 daddr, __wsum base)
1da177e4
LT
880{
881 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
882}
883
b51655b9 884static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1da177e4 885{
fb286bb2 886 return __skb_checksum_complete(skb);
1da177e4
LT
887}
888
40efc6fa 889static inline int tcp_checksum_complete(struct sk_buff *skb)
1da177e4 890{
60476372 891 return !skb_csum_unnecessary(skb) &&
1da177e4
LT
892 __tcp_checksum_complete(skb);
893}
894
895/* Prequeue for VJ style copy to user, combined with checksumming. */
896
40efc6fa 897static inline void tcp_prequeue_init(struct tcp_sock *tp)
1da177e4
LT
898{
899 tp->ucopy.task = NULL;
900 tp->ucopy.len = 0;
901 tp->ucopy.memory = 0;
902 skb_queue_head_init(&tp->ucopy.prequeue);
97fc2f08
CL
903#ifdef CONFIG_NET_DMA
904 tp->ucopy.dma_chan = NULL;
905 tp->ucopy.wakeup = 0;
906 tp->ucopy.pinned_list = NULL;
907 tp->ucopy.dma_cookie = 0;
908#endif
1da177e4
LT
909}
910
911/* Packet is added to VJ-style prequeue for processing in process
912 * context, if a reader task is waiting. Apparently, this exciting
913 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
914 * failed somewhere. Latency? Burstiness? Well, at least now we will
915 * see, why it failed. 8)8) --ANK
916 *
917 * NOTE: is this not too big to inline?
918 */
40efc6fa 919static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
1da177e4
LT
920{
921 struct tcp_sock *tp = tcp_sk(sk);
922
f5f8d86b
ED
923 if (sysctl_tcp_low_latency || !tp->ucopy.task)
924 return 0;
925
926 __skb_queue_tail(&tp->ucopy.prequeue, skb);
927 tp->ucopy.memory += skb->truesize;
928 if (tp->ucopy.memory > sk->sk_rcvbuf) {
929 struct sk_buff *skb1;
930
931 BUG_ON(sock_owned_by_user(sk));
932
933 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
934 sk_backlog_rcv(sk, skb1);
935 NET_INC_STATS_BH(sock_net(sk),
936 LINUX_MIB_TCPPREQUEUEDROPPED);
1da177e4 937 }
f5f8d86b
ED
938
939 tp->ucopy.memory = 0;
940 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
aa395145 941 wake_up_interruptible_sync_poll(sk_sleep(sk),
7aedec2a 942 POLLIN | POLLRDNORM | POLLRDBAND);
f5f8d86b
ED
943 if (!inet_csk_ack_scheduled(sk))
944 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
22f6dacd 945 (3 * tcp_rto_min(sk)) / 4,
f5f8d86b 946 TCP_RTO_MAX);
1da177e4 947 }
f5f8d86b 948 return 1;
1da177e4
LT
949}
950
951
952#undef STATE_TRACE
953
954#ifdef STATE_TRACE
955static const char *statename[]={
956 "Unused","Established","Syn Sent","Syn Recv",
957 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
958 "Close Wait","Last ACK","Listen","Closing"
959};
960#endif
490d5046 961extern void tcp_set_state(struct sock *sk, int state);
1da177e4 962
4ac02bab 963extern void tcp_done(struct sock *sk);
1da177e4 964
40efc6fa 965static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1da177e4
LT
966{
967 rx_opt->dsack = 0;
1da177e4
LT
968 rx_opt->num_sacks = 0;
969}
970
1da177e4
LT
971/* Determine a window scaling and initial window to offer. */
972extern void tcp_select_initial_window(int __space, __u32 mss,
973 __u32 *rcv_wnd, __u32 *window_clamp,
31d12926 974 int wscale_ok, __u8 *rcv_wscale,
975 __u32 init_rcv_wnd);
1da177e4
LT
976
977static inline int tcp_win_from_space(int space)
978{
979 return sysctl_tcp_adv_win_scale<=0 ?
980 (space>>(-sysctl_tcp_adv_win_scale)) :
981 space - (space>>sysctl_tcp_adv_win_scale);
982}
983
984/* Note: caller must be prepared to deal with negative returns */
985static inline int tcp_space(const struct sock *sk)
986{
987 return tcp_win_from_space(sk->sk_rcvbuf -
988 atomic_read(&sk->sk_rmem_alloc));
989}
990
991static inline int tcp_full_space(const struct sock *sk)
992{
993 return tcp_win_from_space(sk->sk_rcvbuf);
994}
995
40efc6fa
SH
996static inline void tcp_openreq_init(struct request_sock *req,
997 struct tcp_options_received *rx_opt,
998 struct sk_buff *skb)
1da177e4 999{
2e6599cb
ACM
1000 struct inet_request_sock *ireq = inet_rsk(req);
1001
1da177e4 1002 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
4dfc2817 1003 req->cookie_ts = 0;
2e6599cb 1004 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1da177e4
LT
1005 req->mss = rx_opt->mss_clamp;
1006 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
2e6599cb
ACM
1007 ireq->tstamp_ok = rx_opt->tstamp_ok;
1008 ireq->sack_ok = rx_opt->sack_ok;
1009 ireq->snd_wscale = rx_opt->snd_wscale;
1010 ireq->wscale_ok = rx_opt->wscale_ok;
1011 ireq->acked = 0;
1012 ireq->ecn_ok = 0;
aa8223c7 1013 ireq->rmt_port = tcp_hdr(skb)->source;
a3116ac5 1014 ireq->loc_port = tcp_hdr(skb)->dest;
1da177e4
LT
1015}
1016
5c52ba17 1017extern void tcp_enter_memory_pressure(struct sock *sk);
1da177e4 1018
1da177e4
LT
1019static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1020{
1021 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1022}
1023
1024static inline int keepalive_time_when(const struct tcp_sock *tp)
1025{
1026 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1027}
1028
df19a626
ED
1029static inline int keepalive_probes(const struct tcp_sock *tp)
1030{
1031 return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes;
1032}
1033
6c37e5de
FL
1034static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp)
1035{
1036 const struct inet_connection_sock *icsk = &tp->inet_conn;
1037
1038 return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime,
1039 tcp_time_stamp - tp->rcv_tstamp);
1040}
1041
463c84b9 1042static inline int tcp_fin_time(const struct sock *sk)
1da177e4 1043{
463c84b9
ACM
1044 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1045 const int rto = inet_csk(sk)->icsk_rto;
1da177e4 1046
463c84b9
ACM
1047 if (fin_timeout < (rto << 2) - (rto >> 1))
1048 fin_timeout = (rto << 2) - (rto >> 1);
1da177e4
LT
1049
1050 return fin_timeout;
1051}
1052
c887e6d2
IJ
1053static inline int tcp_paws_check(const struct tcp_options_received *rx_opt,
1054 int paws_win)
1da177e4 1055{
c887e6d2
IJ
1056 if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
1057 return 1;
1058 if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
1059 return 1;
1060
1061 return 0;
1062}
1063
1064static inline int tcp_paws_reject(const struct tcp_options_received *rx_opt,
1065 int rst)
1066{
1067 if (tcp_paws_check(rx_opt, 0))
1da177e4
LT
1068 return 0;
1069
1070 /* RST segments are not recommended to carry timestamp,
1071 and, if they do, it is recommended to ignore PAWS because
1072 "their cleanup function should take precedence over timestamps."
1073 Certainly, it is mistake. It is necessary to understand the reasons
1074 of this constraint to relax it: if peer reboots, clock may go
1075 out-of-sync and half-open connections will not be reset.
1076 Actually, the problem would be not existing if all
1077 the implementations followed draft about maintaining clock
1078 via reboots. Linux-2.2 DOES NOT!
1079
1080 However, we can relax time bounds for RST segments to MSL.
1081 */
9d729f72 1082 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1da177e4
LT
1083 return 0;
1084 return 1;
1085}
1086
1da177e4
LT
1087#define TCP_CHECK_TIMER(sk) do { } while (0)
1088
a9c19329 1089static inline void tcp_mib_init(struct net *net)
1da177e4
LT
1090{
1091 /* See RFC 2012 */
cf1100a7
PE
1092 TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1093 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1094 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1095 TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
1da177e4
LT
1096}
1097
5af4ec23 1098/* from STCP */
ef9da47c 1099static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
0800f170 1100{
6a438bbe
SH
1101 tp->lost_skb_hint = NULL;
1102 tp->scoreboard_skb_hint = NULL;
ef9da47c
IJ
1103}
1104
1105static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1106{
1107 tcp_clear_retrans_hints_partial(tp);
6a438bbe 1108 tp->retransmit_skb_hint = NULL;
b7689205
IJ
1109}
1110
cfb6eeb4
YH
1111/* MD5 Signature */
1112struct crypto_hash;
1113
1114/* - key database */
1115struct tcp_md5sig_key {
1116 u8 *key;
1117 u8 keylen;
1118};
1119
1120struct tcp4_md5sig_key {
f8ab18d2 1121 struct tcp_md5sig_key base;
cfb6eeb4
YH
1122 __be32 addr;
1123};
1124
1125struct tcp6_md5sig_key {
f8ab18d2 1126 struct tcp_md5sig_key base;
cfb6eeb4
YH
1127#if 0
1128 u32 scope_id; /* XXX */
1129#endif
1130 struct in6_addr addr;
1131};
1132
1133/* - sock block */
1134struct tcp_md5sig_info {
1135 struct tcp4_md5sig_key *keys4;
1136#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1137 struct tcp6_md5sig_key *keys6;
1138 u32 entries6;
1139 u32 alloced6;
1140#endif
1141 u32 entries4;
1142 u32 alloced4;
1143};
1144
1145/* - pseudo header */
1146struct tcp4_pseudohdr {
1147 __be32 saddr;
1148 __be32 daddr;
1149 __u8 pad;
1150 __u8 protocol;
1151 __be16 len;
1152};
1153
1154struct tcp6_pseudohdr {
1155 struct in6_addr saddr;
1156 struct in6_addr daddr;
1157 __be32 len;
1158 __be32 protocol; /* including padding */
1159};
1160
1161union tcp_md5sum_block {
1162 struct tcp4_pseudohdr ip4;
1163#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1164 struct tcp6_pseudohdr ip6;
1165#endif
1166};
1167
1168/* - pool: digest algorithm, hash description and scratch buffer */
1169struct tcp_md5sig_pool {
1170 struct hash_desc md5_desc;
1171 union tcp_md5sum_block md5_blk;
1172};
1173
1174#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1175
1176/* - functions */
49a72dfb
AL
1177extern int tcp_v4_md5_hash_skb(char *md5_hash,
1178 struct tcp_md5sig_key *key,
1179 struct sock *sk,
1180 struct request_sock *req,
1181 struct sk_buff *skb);
1182
cfb6eeb4
YH
1183extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1184 struct sock *addr_sk);
1185
1186extern int tcp_v4_md5_do_add(struct sock *sk,
1187 __be32 addr,
1188 u8 *newkey,
1189 u8 newkeylen);
1190
1191extern int tcp_v4_md5_do_del(struct sock *sk,
8e5200f5 1192 __be32 addr);
cfb6eeb4 1193
9501f972
YH
1194#ifdef CONFIG_TCP_MD5SIG
1195#define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_keylen ? \
1196 &(struct tcp_md5sig_key) { \
1197 .key = (twsk)->tw_md5_key, \
1198 .keylen = (twsk)->tw_md5_keylen, \
1199 } : NULL)
1200#else
1201#define tcp_twsk_md5_key(twsk) NULL
1202#endif
1203
7d720c3e 1204extern struct tcp_md5sig_pool * __percpu *tcp_alloc_md5sig_pool(struct sock *);
cfb6eeb4
YH
1205extern void tcp_free_md5sig_pool(void);
1206
35790c04
ED
1207extern struct tcp_md5sig_pool *tcp_get_md5sig_pool(void);
1208extern void tcp_put_md5sig_pool(void);
1209
49a72dfb
AL
1210extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, struct tcphdr *);
1211extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, struct sk_buff *,
1212 unsigned header_len);
1213extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
1214 struct tcp_md5sig_key *key);
cfb6eeb4 1215
fe067e8a
DM
1216/* write queue abstraction */
1217static inline void tcp_write_queue_purge(struct sock *sk)
1218{
1219 struct sk_buff *skb;
1220
1221 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
3ab224be
HA
1222 sk_wmem_free_skb(sk, skb);
1223 sk_mem_reclaim(sk);
8818a9d8 1224 tcp_clear_all_retrans_hints(tcp_sk(sk));
fe067e8a
DM
1225}
1226
1227static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1228{
cd07a8ea 1229 return skb_peek(&sk->sk_write_queue);
fe067e8a
DM
1230}
1231
1232static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1233{
cd07a8ea 1234 return skb_peek_tail(&sk->sk_write_queue);
fe067e8a
DM
1235}
1236
1237static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1238{
cd07a8ea 1239 return skb_queue_next(&sk->sk_write_queue, skb);
fe067e8a
DM
1240}
1241
832d11c5
IJ
1242static inline struct sk_buff *tcp_write_queue_prev(struct sock *sk, struct sk_buff *skb)
1243{
1244 return skb_queue_prev(&sk->sk_write_queue, skb);
1245}
1246
fe067e8a 1247#define tcp_for_write_queue(skb, sk) \
cd07a8ea 1248 skb_queue_walk(&(sk)->sk_write_queue, skb)
fe067e8a
DM
1249
1250#define tcp_for_write_queue_from(skb, sk) \
cd07a8ea 1251 skb_queue_walk_from(&(sk)->sk_write_queue, skb)
fe067e8a 1252
234b6860 1253#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
cd07a8ea 1254 skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
234b6860 1255
fe067e8a
DM
1256static inline struct sk_buff *tcp_send_head(struct sock *sk)
1257{
1258 return sk->sk_send_head;
1259}
1260
cd07a8ea
DM
1261static inline bool tcp_skb_is_last(const struct sock *sk,
1262 const struct sk_buff *skb)
1263{
1264 return skb_queue_is_last(&sk->sk_write_queue, skb);
1265}
1266
fe067e8a
DM
1267static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1268{
cd07a8ea 1269 if (tcp_skb_is_last(sk, skb))
fe067e8a 1270 sk->sk_send_head = NULL;
cd07a8ea
DM
1271 else
1272 sk->sk_send_head = tcp_write_queue_next(sk, skb);
fe067e8a
DM
1273}
1274
1275static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1276{
1277 if (sk->sk_send_head == skb_unlinked)
1278 sk->sk_send_head = NULL;
1279}
1280
1281static inline void tcp_init_send_head(struct sock *sk)
1282{
1283 sk->sk_send_head = NULL;
1284}
1285
1286static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1287{
1288 __skb_queue_tail(&sk->sk_write_queue, skb);
1289}
1290
1291static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1292{
1293 __tcp_add_write_queue_tail(sk, skb);
1294
1295 /* Queue it, remembering where we must start sending. */
6859d494 1296 if (sk->sk_send_head == NULL) {
fe067e8a 1297 sk->sk_send_head = skb;
6859d494
IJ
1298
1299 if (tcp_sk(sk)->highest_sack == NULL)
1300 tcp_sk(sk)->highest_sack = skb;
1301 }
fe067e8a
DM
1302}
1303
1304static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1305{
1306 __skb_queue_head(&sk->sk_write_queue, skb);
1307}
1308
1309/* Insert buff after skb on the write queue of sk. */
1310static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1311 struct sk_buff *buff,
1312 struct sock *sk)
1313{
7de6c033 1314 __skb_queue_after(&sk->sk_write_queue, skb, buff);
fe067e8a
DM
1315}
1316
43f59c89 1317/* Insert new before skb on the write queue of sk. */
fe067e8a
DM
1318static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1319 struct sk_buff *skb,
1320 struct sock *sk)
1321{
43f59c89 1322 __skb_queue_before(&sk->sk_write_queue, skb, new);