?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
K2.6 patches and update.
[tomato.git] / release / src-rt / linux / linux-2.6 / include / net / tcp.h
blob0b6a4d8025aafb72ed90639e42adf74c41fc0af5
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 *
10 * Authors: Ross Biro
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
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>
29 #include <linux/skbuff.h>
30 #include <linux/dmaengine.h>
31 #include <linux/crypto.h>
32
33 #include <net/inet_connection_sock.h>
34 #include <net/inet_timewait_sock.h>
35 #include <net/inet_hashtables.h>
36 #include <net/checksum.h>
37 #include <net/request_sock.h>
38 #include <net/sock.h>
39 #include <net/snmp.h>
40 #include <net/ip.h>
41 #include <net/tcp_states.h>
42 #include <net/dst.h>
43
44 #include <linux/seq_file.h>
45
46 extern struct inet_hashinfo tcp_hashinfo;
47
48 extern atomic_t tcp_orphan_count;
49 extern void tcp_time_wait(struct sock *sk, int state, int timeo);
50
51 #define MAX_TCP_HEADER (128 + MAX_HEADER)
52
53 /*
54 * Never offer a window over 32767 without using window scaling. Some
55 * poor stacks do signed 16bit maths!
56 */
57 #define MAX_TCP_WINDOW 32767U
58
59 /* Minimal accepted MSS. It is (60+60+8) - (20+20). */
60 #define TCP_MIN_MSS 88U
61
62 /* Minimal RCV_MSS. */
63 #define TCP_MIN_RCVMSS 536U
64
65 /* The least MTU to use for probing */
66 #define TCP_BASE_MSS 512
67
68 /* After receiving this amount of duplicate ACKs fast retransmit starts. */
69 #define TCP_FASTRETRANS_THRESH 3
70
71 /* Maximal reordering. */
72 #define TCP_MAX_REORDERING 127
73
74 /* Maximal number of ACKs sent quickly to accelerate slow-start. */
75 #define TCP_MAX_QUICKACKS 16U
76
77 /* urg_data states */
78 #define TCP_URG_VALID 0x0100
79 #define TCP_URG_NOTYET 0x0200
80 #define TCP_URG_READ 0x0400
81
82 #define TCP_RETR1 3 /*
83 * This is how many retries it does before it
84 * tries to figure out if the gateway is
85 * down. Minimal RFC value is 3; it corresponds
86 * to ~3sec-8min depending on RTO.
87 */
88
89 #define TCP_RETR2 15 /*
90 * This should take at least
91 * 90 minutes to time out.
92 * RFC1122 says that the limit is 100 sec.
93 * 15 is ~13-30min depending on RTO.
94 */
95
96 #define TCP_SYN_RETRIES 5 /* number of times to retry active opening a
97 * connection: ~180sec is RFC minimum */
98
99 #define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
100 * connection: ~180sec is RFC minimum */
101
102
103 #define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned
104 * socket. 7 is ~50sec-16min.
105 */
106
107
108 #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
109 * state, about 60 seconds */
110 #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
111 /* BSD style FIN_WAIT2 deadlock breaker.
112 * It used to be 3min, new value is 60sec,
113 * to combine FIN-WAIT-2 timeout with
114 * TIME-WAIT timer.
115 */
116
117 #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
118 #if HZ >= 100
119 #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
120 #define TCP_ATO_MIN ((unsigned)(HZ/25))
121 #else
122 #define TCP_DELACK_MIN 4U
123 #define TCP_ATO_MIN 4U
124 #endif
125 #define TCP_RTO_MAX ((unsigned)(120*HZ))
126 #define TCP_RTO_MIN ((unsigned)(HZ/5))
127 #define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */
128
129 #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
130 * for local resources.
131 */
132
133 #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
134 #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
135 #define TCP_KEEPALIVE_INTVL (75*HZ)
136
137 #define MAX_TCP_KEEPIDLE 32767
138 #define MAX_TCP_KEEPINTVL 32767
139 #define MAX_TCP_KEEPCNT 127
140 #define MAX_TCP_SYNCNT 127
141
142 #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
143
144 #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
145 #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
146 * after this time. It should be equal
147 * (or greater than) TCP_TIMEWAIT_LEN
148 * to provide reliability equal to one
149 * provided by timewait state.
150 */
151 #define TCP_PAWS_WINDOW 1 /* Replay window for per-host
152 * timestamps. It must be less than
153 * minimal timewait lifetime.
154 */
155 /*
156 * TCP option
157 */
158
159 #define TCPOPT_NOP 1 /* Padding */
160 #define TCPOPT_EOL 0 /* End of options */
161 #define TCPOPT_MSS 2 /* Segment size negotiating */
162 #define TCPOPT_WINDOW 3 /* Window scaling */
163 #define TCPOPT_SACK_PERM 4 /* SACK Permitted */
164 #define TCPOPT_SACK 5 /* SACK Block */
165 #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
166 #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
167
168 /*
169 * TCP option lengths
170 */
171
172 #define TCPOLEN_MSS 4
173 #define TCPOLEN_WINDOW 3
174 #define TCPOLEN_SACK_PERM 2
175 #define TCPOLEN_TIMESTAMP 10
176 #define TCPOLEN_MD5SIG 18
177
178 /* But this is what stacks really send out. */
179 #define TCPOLEN_TSTAMP_ALIGNED 12
180 #define TCPOLEN_WSCALE_ALIGNED 4
181 #define TCPOLEN_SACKPERM_ALIGNED 4
182 #define TCPOLEN_SACK_BASE 2
183 #define TCPOLEN_SACK_BASE_ALIGNED 4
184 #define TCPOLEN_SACK_PERBLOCK 8
185 #define TCPOLEN_MD5SIG_ALIGNED 20
186
187 /* Flags in tp->nonagle */
188 #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
189 #define TCP_NAGLE_CORK 2 /* Socket is corked */
190 #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
191
192 extern struct inet_timewait_death_row tcp_death_row;
193
194 /* sysctl variables for tcp */
195 extern int sysctl_tcp_timestamps;
196 extern int sysctl_tcp_window_scaling;
197 extern int sysctl_tcp_sack;
198 extern int sysctl_tcp_fin_timeout;
199 extern int sysctl_tcp_keepalive_time;
200 extern int sysctl_tcp_keepalive_probes;
201 extern int sysctl_tcp_keepalive_intvl;
202 extern int sysctl_tcp_syn_retries;
203 extern int sysctl_tcp_synack_retries;
204 extern int sysctl_tcp_retries1;
205 extern int sysctl_tcp_retries2;
206 extern int sysctl_tcp_orphan_retries;
207 extern int sysctl_tcp_syncookies;
208 extern int sysctl_tcp_retrans_collapse;
209 extern int sysctl_tcp_stdurg;
210 extern int sysctl_tcp_rfc1337;
211 extern int sysctl_tcp_abort_on_overflow;
212 extern int sysctl_tcp_max_orphans;
213 extern int sysctl_tcp_fack;
214 extern int sysctl_tcp_reordering;
215 extern int sysctl_tcp_ecn;
216 extern int sysctl_tcp_dsack;
217 extern int sysctl_tcp_mem[3];
218 extern int sysctl_tcp_wmem[3];
219 extern int sysctl_tcp_rmem[3];
220 extern int sysctl_tcp_app_win;
221 extern int sysctl_tcp_adv_win_scale;
222 extern int sysctl_tcp_tw_reuse;
223 extern int sysctl_tcp_frto;
224 extern int sysctl_tcp_frto_response;
225 extern int sysctl_tcp_low_latency;
226 extern int sysctl_tcp_dma_copybreak;
227 extern int sysctl_tcp_nometrics_save;
228 extern int sysctl_tcp_moderate_rcvbuf;
229 extern int sysctl_tcp_tso_win_divisor;
230 extern int sysctl_tcp_abc;
231 extern int sysctl_tcp_mtu_probing;
232 extern int sysctl_tcp_base_mss;
233 extern int sysctl_tcp_workaround_signed_windows;
234 extern int sysctl_tcp_slow_start_after_idle;
235 extern int sysctl_tcp_max_ssthresh;
236
237 extern atomic_t tcp_memory_allocated;
238 extern atomic_t tcp_sockets_allocated;
239 extern int tcp_memory_pressure;
240
241 /*
242 * The next routines deal with comparing 32 bit unsigned ints
243 * and worry about wraparound (automatic with unsigned arithmetic).
244 */
245
246 static inline int before(__u32 seq1, __u32 seq2)
247 {
248 return (__s32)(seq1-seq2) < 0;
249 }
250 #define after(seq2, seq1) before(seq1, seq2)
251
252 /* is s2<=s1<=s3 ? */
253 static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
254 {
255 return seq3 - seq2 >= seq1 - seq2;
256 }
257
258 static inline int tcp_too_many_orphans(struct sock *sk, int num)
259 {
260 return (num > sysctl_tcp_max_orphans) ||
261 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
262 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
263 }
264
265 extern struct proto tcp_prot;
266
267 DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
268 #define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field)
269 #define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field)
270 #define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field)
271 #define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field)
272 #define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val)
273 #define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val)
274
275 extern void tcp_v4_err(struct sk_buff *skb, u32);
276
277 extern void tcp_shutdown (struct sock *sk, int how);
278
279 extern int tcp_v4_rcv(struct sk_buff *skb);
280
281 extern int tcp_v4_remember_stamp(struct sock *sk);
282
283 extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
284
285 extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
286 struct msghdr *msg, size_t size);
287 extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
288
289 extern int tcp_ioctl(struct sock *sk,
290 int cmd,
291 unsigned long arg);
292
293 extern int tcp_rcv_state_process(struct sock *sk,
294 struct sk_buff *skb,
295 struct tcphdr *th,
296 unsigned len);
297
298 extern int tcp_rcv_established(struct sock *sk,
299 struct sk_buff *skb,
300 struct tcphdr *th,
301 unsigned len);
302
303 extern void tcp_rcv_space_adjust(struct sock *sk);
304
305 extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
306
307 extern int tcp_twsk_unique(struct sock *sk,
308 struct sock *sktw, void *twp);
309
310 extern void tcp_twsk_destructor(struct sock *sk);
311
312 extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
313 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
314
315 static inline void tcp_dec_quickack_mode(struct sock *sk,
316 const unsigned int pkts)
317 {
318 struct inet_connection_sock *icsk = inet_csk(sk);
319
320 if (icsk->icsk_ack.quick) {
321 if (pkts >= icsk->icsk_ack.quick) {
322 icsk->icsk_ack.quick = 0;
323 /* Leaving quickack mode we deflate ATO. */
324 icsk->icsk_ack.ato = TCP_ATO_MIN;
325 } else
326 icsk->icsk_ack.quick -= pkts;
327 }
328 }
329
330 extern void tcp_enter_quickack_mode(struct sock *sk);
331
332 static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
333 {
334 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
335 }
336
337 enum tcp_tw_status
338 {
339 TCP_TW_SUCCESS = 0,
340 TCP_TW_RST = 1,
341 TCP_TW_ACK = 2,
342 TCP_TW_SYN = 3
343 };
344
345
346 extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
347 struct sk_buff *skb,
348 const struct tcphdr *th);
349
350 extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
351 struct request_sock *req,
352 struct request_sock **prev);
353 extern int tcp_child_process(struct sock *parent,
354 struct sock *child,
355 struct sk_buff *skb);
356 extern int tcp_use_frto(struct sock *sk);
357 extern void tcp_enter_frto(struct sock *sk);
358 extern void tcp_enter_loss(struct sock *sk, int how);
359 extern void tcp_clear_retrans(struct tcp_sock *tp);
360 extern void tcp_update_metrics(struct sock *sk);
361
362 extern void tcp_close(struct sock *sk,
363 long timeout);
364 extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
365
366 extern int tcp_getsockopt(struct sock *sk, int level,
367 int optname,
368 char __user *optval,
369 int __user *optlen);
370 extern int tcp_setsockopt(struct sock *sk, int level,
371 int optname, char __user *optval,
372 int optlen);
373 extern int compat_tcp_getsockopt(struct sock *sk,
374 int level, int optname,
375 char __user *optval, int __user *optlen);
376 extern int compat_tcp_setsockopt(struct sock *sk,
377 int level, int optname,
378 char __user *optval, int optlen);
379 extern void tcp_set_keepalive(struct sock *sk, int val);
380 extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
381 struct msghdr *msg,
382 size_t len, int nonblock,
383 int flags, int *addr_len);
384
385 extern void tcp_parse_options(struct sk_buff *skb,
386 struct tcp_options_received *opt_rx,
387 int estab);
388
389 /*
390 * TCP v4 functions exported for the inet6 API
391 */
392
393 extern void tcp_v4_send_check(struct sock *sk, int len,
394 struct sk_buff *skb);
395
396 extern int tcp_v4_conn_request(struct sock *sk,
397 struct sk_buff *skb);
398
399 extern struct sock * tcp_create_openreq_child(struct sock *sk,
400 struct request_sock *req,
401 struct sk_buff *skb);
402
403 extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
404 struct sk_buff *skb,
405 struct request_sock *req,
406 struct dst_entry *dst);
407
408 extern int tcp_v4_do_rcv(struct sock *sk,
409 struct sk_buff *skb);
410
411 extern int tcp_v4_connect(struct sock *sk,
412 struct sockaddr *uaddr,
413 int addr_len);
414
415 extern int tcp_connect(struct sock *sk);
416
417 extern struct sk_buff * tcp_make_synack(struct sock *sk,
418 struct dst_entry *dst,
419 struct request_sock *req);
420
421 extern int tcp_disconnect(struct sock *sk, int flags);
422
423 extern void tcp_unhash(struct sock *sk);
424
425 /* From syncookies.c */
426 extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
427 struct ip_options *opt);
428 extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
429 __u16 *mss);
430
431 /* tcp_output.c */
432
433 extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
434 int nonagle);
435 extern int tcp_may_send_now(struct sock *sk);
436 extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
437 extern void tcp_xmit_retransmit_queue(struct sock *);
438 extern void tcp_simple_retransmit(struct sock *);
439 extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
440 extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
441
442 extern void tcp_send_probe0(struct sock *);
443 extern void tcp_send_partial(struct sock *);
444 extern int tcp_write_wakeup(struct sock *);
445 extern void tcp_send_fin(struct sock *sk);
446 extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
447 extern int tcp_send_synack(struct sock *);
448 extern void tcp_push_one(struct sock *, unsigned int mss_now);
449 extern void tcp_send_ack(struct sock *sk);
450 extern void tcp_send_delayed_ack(struct sock *sk);
451
452 /* tcp_input.c */
453 extern void tcp_cwnd_application_limited(struct sock *sk);
454
455 /* tcp_timer.c */
456 extern void tcp_init_xmit_timers(struct sock *);
457 static inline void tcp_clear_xmit_timers(struct sock *sk)
458 {
459 inet_csk_clear_xmit_timers(sk);
460 }
461
462 extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
463 extern unsigned int tcp_current_mss(struct sock *sk, int large);
464
465 /* tcp.c */
466 extern void tcp_get_info(struct sock *, struct tcp_info *);
467
468 /* Read 'sendfile()'-style from a TCP socket */
469 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
470 unsigned int, size_t);
471 extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
472 sk_read_actor_t recv_actor);
473
474 extern void tcp_initialize_rcv_mss(struct sock *sk);
475
476 extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
477 extern int tcp_mss_to_mtu(struct sock *sk, int mss);
478 extern void tcp_mtup_init(struct sock *sk);
479
480 static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
481 {
482 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
483 ntohl(TCP_FLAG_ACK) |
484 snd_wnd);
485 }
486
487 static inline void tcp_fast_path_on(struct tcp_sock *tp)
488 {
489 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
490 }
491
492 static inline void tcp_fast_path_check(struct sock *sk)
493 {
494 struct tcp_sock *tp = tcp_sk(sk);
495
496 if (skb_queue_empty(&tp->out_of_order_queue) &&
497 tp->rcv_wnd &&
498 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
499 !tp->urg_data)
500 tcp_fast_path_on(tp);
501 }
502
503 /* Compute the actual rto_min value */
504 static inline u32 tcp_rto_min(struct sock *sk)
505 {
506 struct dst_entry *dst = __sk_dst_get(sk);
507 u32 rto_min = TCP_RTO_MIN;
508
509 if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
510 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
511 return rto_min;
512 }
513
514 /* Compute the actual receive window we are currently advertising.
515 * Rcv_nxt can be after the window if our peer push more data
516 * than the offered window.
517 */
518 static inline u32 tcp_receive_window(const struct tcp_sock *tp)
519 {
520 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
521
522 if (win < 0)
523 win = 0;
524 return (u32) win;
525 }
526
527 /* Choose a new window, without checks for shrinking, and without
528 * scaling applied to the result. The caller does these things
529 * if necessary. This is a "raw" window selection.
530 */
531 extern u32 __tcp_select_window(struct sock *sk);
532
533 /* TCP timestamps are only 32-bits, this causes a slight
534 * complication on 64-bit systems since we store a snapshot
535 * of jiffies in the buffer control blocks below. We decided
536 * to use only the low 32-bits of jiffies and hide the ugly
537 * casts with the following macro.
538 */
539 #define tcp_time_stamp ((__u32)(jiffies))
540
541 /* This is what the send packet queuing engine uses to pass
542 * TCP per-packet control information to the transmission
543 * code. We also store the host-order sequence numbers in
544 * here too. This is 36 bytes on 32-bit architectures,
545 * 40 bytes on 64-bit machines, if this grows please adjust
546 * skbuff.h:skbuff->cb[xxx] size appropriately.
547 */
548 struct tcp_skb_cb {
549 union {
550 struct inet_skb_parm h4;
551 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
552 struct inet6_skb_parm h6;
553 #endif
554 } header; /* For incoming frames */
555 __u32 seq; /* Starting sequence number */
556 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
557 __u32 when; /* used to compute rtt's */
558 __u8 flags; /* TCP header flags. */
559
560 /* NOTE: These must match up to the flags byte in a
561 * real TCP header.
562 */
563 #define TCPCB_FLAG_FIN 0x01
564 #define TCPCB_FLAG_SYN 0x02
565 #define TCPCB_FLAG_RST 0x04
566 #define TCPCB_FLAG_PSH 0x08
567 #define TCPCB_FLAG_ACK 0x10
568 #define TCPCB_FLAG_URG 0x20
569 #define TCPCB_FLAG_ECE 0x40
570 #define TCPCB_FLAG_CWR 0x80
571
572 __u8 sacked; /* State flags for SACK/FACK. */
573 #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
574 #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
575 #define TCPCB_LOST 0x04 /* SKB is lost */
576 #define TCPCB_TAGBITS 0x07 /* All tag bits */
577
578 #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
579 #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
580
581 #define TCPCB_URG 0x20 /* Urgent pointer advanced here */
582
583 #define TCPCB_AT_TAIL (TCPCB_URG)
584
585 __u16 urg_ptr; /* Valid w/URG flags is set. */
586 __u32 ack_seq; /* Sequence number ACK'd */
587 };
588
589 #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
590
591 #include <net/tcp_ecn.h>
592
593 /* Due to TSO, an SKB can be composed of multiple actual
594 * packets. To keep these tracked properly, we use this.
595 */
596 static inline int tcp_skb_pcount(const struct sk_buff *skb)
597 {
598 return skb_shinfo(skb)->gso_segs;
599 }
600
601 /* This is valid iff tcp_skb_pcount() > 1. */
602 static inline int tcp_skb_mss(const struct sk_buff *skb)
603 {
604 return skb_shinfo(skb)->gso_size;
605 }
606
607 static inline void tcp_dec_pcount_approx(__u32 *count,
608 const struct sk_buff *skb)
609 {
610 if (*count) {
611 *count -= tcp_skb_pcount(skb);
612 if ((int)*count < 0)
613 *count = 0;
614 }
615 }
616
617 static inline void tcp_packets_out_inc(struct sock *sk,
618 const struct sk_buff *skb)
619 {
620 struct tcp_sock *tp = tcp_sk(sk);
621 int orig = tp->packets_out;
622
623 tp->packets_out += tcp_skb_pcount(skb);
624 if (!orig)
625 inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
626 inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
627 }
628
629 static inline void tcp_packets_out_dec(struct tcp_sock *tp,
630 const struct sk_buff *skb)
631 {
632 tp->packets_out -= tcp_skb_pcount(skb);
633 }
634
635 /* Events passed to congestion control interface */
636 enum tcp_ca_event {
637 CA_EVENT_TX_START, /* first transmit when no packets in flight */
638 CA_EVENT_CWND_RESTART, /* congestion window restart */
639 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
640 CA_EVENT_FRTO, /* fast recovery timeout */
641 CA_EVENT_LOSS, /* loss timeout */
642 CA_EVENT_FAST_ACK, /* in sequence ack */
643 CA_EVENT_SLOW_ACK, /* other ack */
644 };
645
646 /*
647 * Interface for adding new TCP congestion control handlers
648 */
649 #define TCP_CA_NAME_MAX 16
650 #define TCP_CA_MAX 128
651 #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
652
653 #define TCP_CONG_NON_RESTRICTED 0x1
654 #define TCP_CONG_RTT_STAMP 0x2
655
656 struct tcp_congestion_ops {
657 struct list_head list;
658 unsigned long flags;
659
660 /* initialize private data (optional) */
661 void (*init)(struct sock *sk);
662 /* cleanup private data (optional) */
663 void (*release)(struct sock *sk);
664
665 /* return slow start threshold (required) */
666 u32 (*ssthresh)(struct sock *sk);
667 /* lower bound for congestion window (optional) */
668 u32 (*min_cwnd)(const struct sock *sk);
669 /* do new cwnd calculation (required) */
670 void (*cong_avoid)(struct sock *sk, u32 ack,
671 u32 rtt, u32 in_flight, int good_ack);
672 /* call before changing ca_state (optional) */
673 void (*set_state)(struct sock *sk, u8 new_state);
674 /* call when cwnd event occurs (optional) */
675 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
676 /* new value of cwnd after loss (optional) */
677 u32 (*undo_cwnd)(struct sock *sk);
678 /* hook for packet ack accounting (optional) */
679 void (*pkts_acked)(struct sock *sk, u32 num_acked, ktime_t last);
680 /* get info for inet_diag (optional) */
681 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
682
683 char name[TCP_CA_NAME_MAX];
684 struct module *owner;
685 };
686
687 extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
688 extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
689
690 extern void tcp_init_congestion_control(struct sock *sk);
691 extern void tcp_cleanup_congestion_control(struct sock *sk);
692 extern int tcp_set_default_congestion_control(const char *name);
693 extern void tcp_get_default_congestion_control(char *name);
694 extern void tcp_get_available_congestion_control(char *buf, size_t len);
695 extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
696 extern int tcp_set_allowed_congestion_control(char *allowed);
697 extern int tcp_set_congestion_control(struct sock *sk, const char *name);
698 extern void tcp_slow_start(struct tcp_sock *tp);
699
700 extern struct tcp_congestion_ops tcp_init_congestion_ops;
701 extern u32 tcp_reno_ssthresh(struct sock *sk);
702 extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack,
703 u32 rtt, u32 in_flight, int flag);
704 extern u32 tcp_reno_min_cwnd(const struct sock *sk);
705 extern struct tcp_congestion_ops tcp_reno;
706
707 static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
708 {
709 struct inet_connection_sock *icsk = inet_csk(sk);
710
711 if (icsk->icsk_ca_ops->set_state)
712 icsk->icsk_ca_ops->set_state(sk, ca_state);
713 icsk->icsk_ca_state = ca_state;
714 }
715
716 static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
717 {
718 const struct inet_connection_sock *icsk = inet_csk(sk);
719
720 if (icsk->icsk_ca_ops->cwnd_event)
721 icsk->icsk_ca_ops->cwnd_event(sk, event);
722 }
723
724 /* This determines how many packets are "in the network" to the best
725 * of our knowledge. In many cases it is conservative, but where
726 * detailed information is available from the receiver (via SACK
727 * blocks etc.) we can make more aggressive calculations.
728 *
729 * Use this for decisions involving congestion control, use just
730 * tp->packets_out to determine if the send queue is empty or not.
731 *
732 * Read this equation as:
733 *
734 * "Packets sent once on transmission queue" MINUS
735 * "Packets left network, but not honestly ACKed yet" PLUS
736 * "Packets fast retransmitted"
737 */
738 static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
739 {
740 return (tp->packets_out - tp->left_out + tp->retrans_out);
741 }
742
743 /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
744 * The exception is rate halving phase, when cwnd is decreasing towards
745 * ssthresh.
746 */
747 static inline __u32 tcp_current_ssthresh(const struct sock *sk)
748 {
749 const struct tcp_sock *tp = tcp_sk(sk);
750 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
751 return tp->snd_ssthresh;
752 else
753 return max(tp->snd_ssthresh,
754 ((tp->snd_cwnd >> 1) +
755 (tp->snd_cwnd >> 2)));
756 }
757
758 static inline void tcp_sync_left_out(struct tcp_sock *tp)
759 {
760 WARN_ON(tp->rx_opt.sack_ok &&
761 (tp->sacked_out + tp->lost_out > tp->packets_out));
762 tp->left_out = tp->sacked_out + tp->lost_out;
763 }
764
765 extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
766 extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
767
768 /* Slow start with delack produces 3 packets of burst, so that
769 * it is safe "de facto".
770 */
771 static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
772 {
773 return 3;
774 }
775
776 /* RFC2861 Check whether we are limited by application or congestion window
777 * This is the inverse of cwnd check in tcp_tso_should_defer
778 */
779 static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
780 {
781 const struct tcp_sock *tp = tcp_sk(sk);
782 u32 left;
783
784 if (in_flight >= tp->snd_cwnd)
785 return 1;
786
787 if (!sk_can_gso(sk))
788 return 0;
789
790 left = tp->snd_cwnd - in_flight;
791 if (sysctl_tcp_tso_win_divisor)
792 return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd;
793 else
794 return left <= tcp_max_burst(tp);
795 }
796
797 static inline void tcp_minshall_update(struct tcp_sock *tp, int mss,
798 const struct sk_buff *skb)
799 {
800 if (skb->len < mss)
801 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
802 }
803
804 static inline void tcp_check_probe_timer(struct sock *sk)
805 {
806 struct tcp_sock *tp = tcp_sk(sk);
807 const struct inet_connection_sock *icsk = inet_csk(sk);
808
809 if (!tp->packets_out && !icsk->icsk_pending)
810 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
811 icsk->icsk_rto, TCP_RTO_MAX);
812 }
813
814 static inline void tcp_push_pending_frames(struct sock *sk)
815 {
816 struct tcp_sock *tp = tcp_sk(sk);
817
818 __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle);
819 }
820
821 static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
822 {
823 tp->snd_wl1 = seq;
824 }
825
826 static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
827 {
828 tp->snd_wl1 = seq;
829 }
830
831 /*
832 * Calculate(/check) TCP checksum
833 */
834 static inline __sum16 tcp_v4_check(int len, __be32 saddr,
835 __be32 daddr, __wsum base)
836 {
837 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
838 }
839
840 static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
841 {
842 return __skb_checksum_complete(skb);
843 }
844
845 static inline int tcp_checksum_complete(struct sk_buff *skb)
846 {
847 return !skb_csum_unnecessary(skb) &&
848 __tcp_checksum_complete(skb);
849 }
850
851 /* Prequeue for VJ style copy to user, combined with checksumming. */
852
853 static inline void tcp_prequeue_init(struct tcp_sock *tp)
854 {
855 tp->ucopy.task = NULL;
856 tp->ucopy.len = 0;
857 tp->ucopy.memory = 0;
858 skb_queue_head_init(&tp->ucopy.prequeue);
859 #ifdef CONFIG_NET_DMA
860 tp->ucopy.dma_chan = NULL;
861 tp->ucopy.wakeup = 0;
862 tp->ucopy.pinned_list = NULL;
863 tp->ucopy.dma_cookie = 0;
864 #endif
865 }
866
867 /* Packet is added to VJ-style prequeue for processing in process
868 * context, if a reader task is waiting. Apparently, this exciting
869 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
870 * failed somewhere. Latency? Burstiness? Well, at least now we will
871 * see, why it failed. 8)8) --ANK
872 *
873 * NOTE: is this not too big to inline?
874 */
875 static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
876 {
877 struct tcp_sock *tp = tcp_sk(sk);
878
879 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
880 __skb_queue_tail(&tp->ucopy.prequeue, skb);
881 tp->ucopy.memory += skb->truesize;
882 if (tp->ucopy.memory > sk->sk_rcvbuf) {
883 struct sk_buff *skb1;
884
885 BUG_ON(sock_owned_by_user(sk));
886
887 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
888 sk->sk_backlog_rcv(sk, skb1);
889 NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED);
890 }
891
892 tp->ucopy.memory = 0;
893 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
894 wake_up_interruptible(sk->sk_sleep);
895 if (!inet_csk_ack_scheduled(sk))
896 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
897 (3 * tcp_rto_min(sk)) / 4,
898 TCP_RTO_MAX);
899 }
900 return 1;
901 }
902 return 0;
903 }
904
905
906 #undef STATE_TRACE
907
908 #ifdef STATE_TRACE
909 static const char *statename[]={
910 "Unused","Established","Syn Sent","Syn Recv",
911 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
912 "Close Wait","Last ACK","Listen","Closing"
913 };
914 #endif
915 extern void tcp_set_state(struct sock *sk, int state);
916
917 extern void tcp_done(struct sock *sk);
918
919 static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
920 {
921 rx_opt->dsack = 0;
922 rx_opt->eff_sacks = 0;
923 rx_opt->num_sacks = 0;
924 }
925
926 /* Determine a window scaling and initial window to offer. */
927 extern void tcp_select_initial_window(int __space, __u32 mss,
928 __u32 *rcv_wnd, __u32 *window_clamp,
929 int wscale_ok, __u8 *rcv_wscale);
930
931 static inline int tcp_win_from_space(int space)
932 {
933 return sysctl_tcp_adv_win_scale<=0 ?
934 (space>>(-sysctl_tcp_adv_win_scale)) :
935 space - (space>>sysctl_tcp_adv_win_scale);
936 }
937
938 /* Note: caller must be prepared to deal with negative returns */
939 static inline int tcp_space(const struct sock *sk)
940 {
941 return tcp_win_from_space(sk->sk_rcvbuf -
942 atomic_read(&sk->sk_rmem_alloc));
943 }
944
945 static inline int tcp_full_space(const struct sock *sk)
946 {
947 return tcp_win_from_space(sk->sk_rcvbuf);
948 }
949
950 static inline void tcp_openreq_init(struct request_sock *req,
951 struct tcp_options_received *rx_opt,
952 struct sk_buff *skb)
953 {
954 struct inet_request_sock *ireq = inet_rsk(req);
955
956 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
957 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
958 req->mss = rx_opt->mss_clamp;
959 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
960 ireq->tstamp_ok = rx_opt->tstamp_ok;
961 ireq->sack_ok = rx_opt->sack_ok;
962 ireq->snd_wscale = rx_opt->snd_wscale;
963 ireq->wscale_ok = rx_opt->wscale_ok;
964 ireq->acked = 0;
965 ireq->ecn_ok = 0;
966 ireq->rmt_port = tcp_hdr(skb)->source;
967 }
968
969 extern void tcp_enter_memory_pressure(void);
970
971 static inline int keepalive_intvl_when(const struct tcp_sock *tp)
972 {
973 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
974 }
975
976 static inline int keepalive_time_when(const struct tcp_sock *tp)
977 {
978 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
979 }
980
981 static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp)
982 {
983 const struct inet_connection_sock *icsk = &tp->inet_conn;
984
985 return min_t(u32, tcp_time_stamp - icsk->icsk_ack.lrcvtime,
986 tcp_time_stamp - tp->rcv_tstamp);
987 }
988
989 static inline int tcp_fin_time(const struct sock *sk)
990 {
991 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
992 const int rto = inet_csk(sk)->icsk_rto;
993
994 if (fin_timeout < (rto << 2) - (rto >> 1))
995 fin_timeout = (rto << 2) - (rto >> 1);
996
997 return fin_timeout;
998 }
999
1000 static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1001 {
1002 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1003 return 0;
1004 if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
1005 return 0;
1006
1007 /* RST segments are not recommended to carry timestamp,
1008 and, if they do, it is recommended to ignore PAWS because
1009 "their cleanup function should take precedence over timestamps."
1010 Certainly, it is mistake. It is necessary to understand the reasons
1011 of this constraint to relax it: if peer reboots, clock may go
1012 out-of-sync and half-open connections will not be reset.
1013 Actually, the problem would be not existing if all
1014 the implementations followed draft about maintaining clock
1015 via reboots. Linux-2.2 DOES NOT!
1016
1017 However, we can relax time bounds for RST segments to MSL.
1018 */
1019 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1020 return 0;
1021 return 1;
1022 }
1023
1024 #define TCP_CHECK_TIMER(sk) do { } while (0)
1025
1026 static inline void tcp_mib_init(void)
1027 {
1028 /* See RFC 2012 */
1029 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1030 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1031 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1032 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1033 }
1034
1035 /*from STCP */
1036 static inline void clear_all_retrans_hints(struct tcp_sock *tp){
1037 tp->lost_skb_hint = NULL;
1038 tp->scoreboard_skb_hint = NULL;
1039 tp->retransmit_skb_hint = NULL;
1040 tp->forward_skb_hint = NULL;
1041 tp->fastpath_skb_hint = NULL;
1042 }
1043
1044 /* MD5 Signature */
1045 struct crypto_hash;
1046
1047 /* - key database */
1048 struct tcp_md5sig_key {
1049 u8 *key;
1050 u8 keylen;
1051 };
1052
1053 struct tcp4_md5sig_key {
1054 struct tcp_md5sig_key base;
1055 __be32 addr;
1056 };
1057
1058 struct tcp6_md5sig_key {
1059 struct tcp_md5sig_key base;
1060 #if 0
1061 u32 scope_id; /* XXX */
1062 #endif
1063 struct in6_addr addr;
1064 };
1065
1066 /* - sock block */
1067 struct tcp_md5sig_info {
1068 struct tcp4_md5sig_key *keys4;
1069 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1070 struct tcp6_md5sig_key *keys6;
1071 u32 entries6;
1072 u32 alloced6;
1073 #endif
1074 u32 entries4;
1075 u32 alloced4;
1076 };
1077
1078 /* - pseudo header */
1079 struct tcp4_pseudohdr {
1080 __be32 saddr;
1081 __be32 daddr;
1082 __u8 pad;
1083 __u8 protocol;
1084 __be16 len;
1085 };
1086
1087 struct tcp6_pseudohdr {
1088 struct in6_addr saddr;
1089 struct in6_addr daddr;
1090 __be32 len;
1091 __be32 protocol; /* including padding */
1092 };
1093
1094 union tcp_md5sum_block {
1095 struct tcp4_pseudohdr ip4;
1096 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1097 struct tcp6_pseudohdr ip6;
1098 #endif
1099 };
1100
1101 /* - pool: digest algorithm, hash description and scratch buffer */
1102 struct tcp_md5sig_pool {
1103 struct hash_desc md5_desc;
1104 union tcp_md5sum_block md5_blk;
1105 };
1106
1107 #define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1108
1109 /* - functions */
1110 extern int tcp_v4_calc_md5_hash(char *md5_hash,
1111 struct tcp_md5sig_key *key,
1112 struct sock *sk,
1113 struct dst_entry *dst,
1114 struct request_sock *req,
1115 struct tcphdr *th,
1116 int protocol, int tcplen);
1117 extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1118 struct sock *addr_sk);
1119
1120 extern int tcp_v4_md5_do_add(struct sock *sk,
1121 __be32 addr,
1122 u8 *newkey,
1123 u8 newkeylen);
1124
1125 extern int tcp_v4_md5_do_del(struct sock *sk,
1126 __be32 addr);
1127
1128 extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1129 extern void tcp_free_md5sig_pool(void);
1130
1131 extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1132 extern void __tcp_put_md5sig_pool(void);
1133
1134 static inline
1135 struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1136 {
1137 int cpu = get_cpu();
1138 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1139 if (!ret)
1140 put_cpu();
1141 return ret;
1142 }
1143
1144 static inline void tcp_put_md5sig_pool(void)
1145 {
1146 __tcp_put_md5sig_pool();
1147 put_cpu();
1148 }
1149
1150 /* write queue abstraction */
1151 static inline void tcp_write_queue_purge(struct sock *sk)
1152 {
1153 struct sk_buff *skb;
1154
1155 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
1156 sk_stream_free_skb(sk, skb);
1157 sk_stream_mem_reclaim(sk);
1158 }
1159
1160 static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1161 {
1162 struct sk_buff *skb = sk->sk_write_queue.next;
1163 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1164 return NULL;
1165 return skb;
1166 }
1167
1168 static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1169 {
1170 struct sk_buff *skb = sk->sk_write_queue.prev;
1171 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1172 return NULL;
1173 return skb;
1174 }
1175
1176 static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1177 {
1178 return skb->next;
1179 }
1180
1181 #define tcp_for_write_queue(skb, sk) \
1182 for (skb = (sk)->sk_write_queue.next; \
1183 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1184 skb = skb->next)
1185
1186 #define tcp_for_write_queue_from(skb, sk) \
1187 for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\
1188 skb = skb->next)
1189
1190 static inline struct sk_buff *tcp_send_head(struct sock *sk)
1191 {
1192 return sk->sk_send_head;
1193 }
1194
1195 static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1196 {
1197 struct tcp_sock *tp = tcp_sk(sk);
1198
1199 sk->sk_send_head = skb->next;
1200 if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue)
1201 sk->sk_send_head = NULL;
1202 /* Don't override Nagle indefinately with F-RTO */
1203 if (tp->frto_counter == 2)
1204 tp->frto_counter = 3;
1205 }
1206
1207 static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1208 {
1209 if (sk->sk_send_head == skb_unlinked)
1210 sk->sk_send_head = NULL;
1211 }
1212
1213 static inline void tcp_init_send_head(struct sock *sk)
1214 {
1215 sk->sk_send_head = NULL;
1216 }
1217
1218 static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1219 {
1220 __skb_queue_tail(&sk->sk_write_queue, skb);
1221 }
1222
1223 static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1224 {
1225 __tcp_add_write_queue_tail(sk, skb);
1226
1227 /* Queue it, remembering where we must start sending. */
1228 if (sk->sk_send_head == NULL)
1229 sk->sk_send_head = skb;
1230 }
1231
1232 static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1233 {
1234 __skb_queue_head(&sk->sk_write_queue, skb);
1235 }
1236
1237 /* Insert buff after skb on the write queue of sk. */
1238 static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1239 struct sk_buff *buff,
1240 struct sock *sk)
1241 {
1242 __skb_append(skb, buff, &sk->sk_write_queue);
1243 }
1244
1245 /* Insert skb between prev and next on the write queue of sk. */
1246 static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1247 struct sk_buff *skb,
1248 struct sock *sk)
1249 {
1250 __skb_insert(new, skb->prev, skb, &sk->sk_write_queue);
1251
1252 if (sk->sk_send_head == skb)
1253 sk->sk_send_head = new;
1254 }
1255
1256 static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1257 {
1258 __skb_unlink(skb, &sk->sk_write_queue);
1259 }
1260
1261 static inline int tcp_skb_is_last(const struct sock *sk,
1262 const struct sk_buff *skb)
1263 {
1264 return skb->next == (struct sk_buff *)&sk->sk_write_queue;
1265 }
1266
1267 static inline int tcp_write_queue_empty(struct sock *sk)
1268 {
1269 return skb_queue_empty(&sk->sk_write_queue);
1270 }
1271
1272 /* /proc */
1273 enum tcp_seq_states {
1274 TCP_SEQ_STATE_LISTENING,
1275 TCP_SEQ_STATE_OPENREQ,
1276 TCP_SEQ_STATE_ESTABLISHED,
1277 TCP_SEQ_STATE_TIME_WAIT,
1278 };
1279
1280 struct tcp_seq_afinfo {
1281 struct module *owner;
1282 char *name;
1283 sa_family_t family;
1284 struct file_operations seq_fops;
1285 struct seq_operations seq_ops;
1286 };
1287
1288 struct tcp_iter_state {
1289 sa_family_t family;
1290 enum tcp_seq_states state;
1291 struct sock *syn_wait_sk;
1292 int bucket, sbucket, num, uid;
1293 };
1294
1295 extern int tcp_proc_register(struct tcp_seq_afinfo *afinfo);
1296 extern void tcp_proc_unregister(struct tcp_seq_afinfo *afinfo);
1297
1298 extern struct request_sock_ops tcp_request_sock_ops;
1299
1300 extern int tcp_v4_destroy_sock(struct sock *sk);
1301
1302 extern int tcp_v4_gso_send_check(struct sk_buff *skb);
1303 extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
1304
1305 #ifdef CONFIG_PROC_FS
1306 extern int tcp4_proc_init(void);
1307 extern void tcp4_proc_exit(void);
1308 #endif
1309
1310 /* TCP af-specific functions */
1311 struct tcp_sock_af_ops {
1312 #ifdef CONFIG_TCP_MD5SIG
1313 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1314 struct sock *addr_sk);
1315 int (*calc_md5_hash) (char *location,
1316 struct tcp_md5sig_key *md5,
1317 struct sock *sk,
1318 struct dst_entry *dst,
1319 struct request_sock *req,
1320 struct tcphdr *th,
1321 int protocol, int len);
1322 int (*md5_add) (struct sock *sk,
1323 struct sock *addr_sk,
1324 u8 *newkey,
1325 u8 len);
1326 int (*md5_parse) (struct sock *sk,
1327 char __user *optval,
1328 int optlen);
1329 #endif
1330 };
1331
1332 struct tcp_request_sock_ops {
1333 #ifdef CONFIG_TCP_MD5SIG
1334 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1335 struct request_sock *req);
1336 #endif
1337 };
1338
1339 extern void tcp_v4_init(struct net_proto_family *ops);
1340 extern void tcp_init(void);
1341
1342 #endif /* _TCP_H */
Tomato firmware and modifications.