1 |
158 |
chris |
/*
|
2 |
|
|
* Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
|
3 |
|
|
* The Regents of the University of California. All rights reserved.
|
4 |
|
|
*
|
5 |
|
|
* Redistribution and use in source and binary forms, with or without
|
6 |
|
|
* modification, are permitted provided that the following conditions
|
7 |
|
|
* are met:
|
8 |
|
|
* 1. Redistributions of source code must retain the above copyright
|
9 |
|
|
* notice, this list of conditions and the following disclaimer.
|
10 |
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
11 |
|
|
* notice, this list of conditions and the following disclaimer in the
|
12 |
|
|
* documentation and/or other materials provided with the distribution.
|
13 |
|
|
* 3. All advertising materials mentioning features or use of this software
|
14 |
|
|
* must display the following acknowledgement:
|
15 |
|
|
* This product includes software developed by the University of
|
16 |
|
|
* California, Berkeley and its contributors.
|
17 |
|
|
* 4. Neither the name of the University nor the names of its contributors
|
18 |
|
|
* may be used to endorse or promote products derived from this software
|
19 |
|
|
* without specific prior written permission.
|
20 |
|
|
*
|
21 |
|
|
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
|
22 |
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
23 |
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
24 |
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
|
25 |
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
26 |
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
27 |
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
28 |
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
29 |
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
30 |
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
31 |
|
|
* SUCH DAMAGE.
|
32 |
|
|
*
|
33 |
|
|
* @(#)tcp_timer.c 8.2 (Berkeley) 5/24/95
|
34 |
208 |
chris |
* $Id: tcp_timer.c,v 1.2 2001-09-27 12:01:56 chris Exp $
|
35 |
158 |
chris |
*/
|
36 |
|
|
|
37 |
|
|
#include "opt_tcpdebug.h"
|
38 |
|
|
|
39 |
|
|
#ifndef TUBA_INCLUDE
|
40 |
|
|
#include <sys/param.h>
|
41 |
|
|
#include <sys/queue.h>
|
42 |
|
|
#include <sys/systm.h>
|
43 |
|
|
#include <sys/kernel.h>
|
44 |
|
|
#include <sys/sysctl.h>
|
45 |
|
|
#include <sys/malloc.h>
|
46 |
|
|
#include <sys/mbuf.h>
|
47 |
|
|
#include <sys/socket.h>
|
48 |
|
|
#include <sys/socketvar.h>
|
49 |
|
|
#include <sys/protosw.h>
|
50 |
|
|
#include <sys/errno.h>
|
51 |
|
|
|
52 |
|
|
#include <machine/cpu.h> /* before tcp_seq.h, for tcp_random18() */
|
53 |
|
|
|
54 |
|
|
#include <net/if.h>
|
55 |
|
|
#include <net/route.h>
|
56 |
|
|
|
57 |
|
|
#include <netinet/in.h>
|
58 |
|
|
#include <netinet/in_systm.h>
|
59 |
|
|
#include <netinet/ip.h>
|
60 |
|
|
#include <netinet/in_pcb.h>
|
61 |
|
|
#include <netinet/ip_var.h>
|
62 |
|
|
#include <netinet/tcp.h>
|
63 |
|
|
#include <netinet/tcp_fsm.h>
|
64 |
|
|
#include <netinet/tcp_seq.h>
|
65 |
|
|
#include <netinet/tcp_timer.h>
|
66 |
|
|
#include <netinet/tcp_var.h>
|
67 |
|
|
#include <netinet/tcpip.h>
|
68 |
|
|
#ifdef TCPDEBUG
|
69 |
|
|
#include <netinet/tcp_debug.h>
|
70 |
|
|
#endif
|
71 |
|
|
|
72 |
|
|
int tcp_keepinit = TCPTV_KEEP_INIT;
|
73 |
|
|
SYSCTL_INT(_net_inet_tcp, TCPCTL_KEEPINIT, keepinit,
|
74 |
|
|
CTLFLAG_RW, &tcp_keepinit , 0, "");
|
75 |
|
|
|
76 |
|
|
int tcp_keepidle = TCPTV_KEEP_IDLE;
|
77 |
|
|
SYSCTL_INT(_net_inet_tcp, TCPCTL_KEEPIDLE, keepidle,
|
78 |
|
|
CTLFLAG_RW, &tcp_keepidle , 0, "");
|
79 |
|
|
|
80 |
|
|
static int tcp_keepintvl = TCPTV_KEEPINTVL;
|
81 |
|
|
SYSCTL_INT(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl,
|
82 |
|
|
CTLFLAG_RW, &tcp_keepintvl , 0, "");
|
83 |
|
|
|
84 |
|
|
static int always_keepalive = 0;
|
85 |
|
|
SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive,
|
86 |
|
|
CTLFLAG_RW, &always_keepalive , 0, "");
|
87 |
|
|
|
88 |
|
|
static int tcp_keepcnt = TCPTV_KEEPCNT;
|
89 |
|
|
/* max idle probes */
|
90 |
|
|
static int tcp_maxpersistidle = TCPTV_KEEP_IDLE;
|
91 |
|
|
/* max idle time in persist */
|
92 |
|
|
int tcp_maxidle;
|
93 |
|
|
#else /* TUBA_INCLUDE */
|
94 |
|
|
|
95 |
|
|
static int tcp_maxpersistidle;
|
96 |
|
|
#endif /* TUBA_INCLUDE */
|
97 |
|
|
|
98 |
|
|
/*
|
99 |
|
|
* Fast timeout routine for processing delayed acks
|
100 |
|
|
*/
|
101 |
|
|
void
|
102 |
|
|
tcp_fasttimo()
|
103 |
|
|
{
|
104 |
|
|
register struct inpcb *inp;
|
105 |
|
|
register struct tcpcb *tp;
|
106 |
|
|
int s;
|
107 |
|
|
|
108 |
|
|
s = splnet();
|
109 |
|
|
|
110 |
|
|
for (inp = tcb.lh_first; inp != NULL; inp = inp->inp_list.le_next) {
|
111 |
|
|
if ((tp = (struct tcpcb *)inp->inp_ppcb) &&
|
112 |
|
|
(tp->t_flags & TF_DELACK)) {
|
113 |
|
|
tp->t_flags &= ~TF_DELACK;
|
114 |
|
|
tp->t_flags |= TF_ACKNOW;
|
115 |
|
|
tcpstat.tcps_delack++;
|
116 |
|
|
(void) tcp_output(tp);
|
117 |
|
|
}
|
118 |
|
|
}
|
119 |
|
|
splx(s);
|
120 |
|
|
}
|
121 |
|
|
|
122 |
|
|
/*
|
123 |
|
|
* Tcp protocol timeout routine called every 500 ms.
|
124 |
|
|
* Updates the timers in all active tcb's and
|
125 |
|
|
* causes finite state machine actions if timers expire.
|
126 |
|
|
*/
|
127 |
|
|
void
|
128 |
|
|
tcp_slowtimo()
|
129 |
|
|
{
|
130 |
|
|
register struct inpcb *ip, *ipnxt;
|
131 |
|
|
register struct tcpcb *tp;
|
132 |
|
|
register int i;
|
133 |
|
|
int s;
|
134 |
|
|
#ifdef TCPDEBUG
|
135 |
|
|
int ostate;
|
136 |
|
|
#endif
|
137 |
|
|
|
138 |
|
|
s = splnet();
|
139 |
|
|
|
140 |
|
|
tcp_maxidle = tcp_keepcnt * tcp_keepintvl;
|
141 |
|
|
|
142 |
|
|
ip = tcb.lh_first;
|
143 |
|
|
if (ip == NULL) {
|
144 |
|
|
splx(s);
|
145 |
|
|
return;
|
146 |
|
|
}
|
147 |
|
|
/*
|
148 |
|
|
* Search through tcb's and update active timers.
|
149 |
|
|
*/
|
150 |
|
|
for (; ip != NULL; ip = ipnxt) {
|
151 |
|
|
ipnxt = ip->inp_list.le_next;
|
152 |
|
|
tp = intotcpcb(ip);
|
153 |
|
|
if (tp == 0 || tp->t_state == TCPS_LISTEN)
|
154 |
|
|
continue;
|
155 |
|
|
for (i = 0; i < TCPT_NTIMERS; i++) {
|
156 |
|
|
if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
|
157 |
|
|
#ifdef TCPDEBUG
|
158 |
|
|
ostate = tp->t_state;
|
159 |
|
|
#endif
|
160 |
|
|
tp = tcp_timers(tp, i);
|
161 |
|
|
if (tp == NULL)
|
162 |
|
|
goto tpgone;
|
163 |
|
|
#ifdef TCPDEBUG
|
164 |
|
|
if (tp->t_inpcb->inp_socket->so_options
|
165 |
|
|
& SO_DEBUG)
|
166 |
|
|
tcp_trace(TA_USER, ostate, tp,
|
167 |
|
|
(struct tcpiphdr *)0,
|
168 |
|
|
PRU_SLOWTIMO);
|
169 |
|
|
#endif
|
170 |
|
|
}
|
171 |
|
|
}
|
172 |
|
|
tp->t_idle++;
|
173 |
|
|
tp->t_duration++;
|
174 |
|
|
if (tp->t_rtt)
|
175 |
|
|
tp->t_rtt++;
|
176 |
|
|
tpgone:
|
177 |
|
|
;
|
178 |
|
|
}
|
179 |
|
|
tcp_iss += TCP_ISSINCR/PR_SLOWHZ; /* increment iss */
|
180 |
|
|
#ifdef TCP_COMPAT_42
|
181 |
|
|
if ((int)tcp_iss < 0)
|
182 |
|
|
tcp_iss = TCP_ISSINCR; /* XXX */
|
183 |
|
|
#endif
|
184 |
|
|
tcp_now++; /* for timestamps */
|
185 |
|
|
splx(s);
|
186 |
|
|
}
|
187 |
|
|
#ifndef TUBA_INCLUDE
|
188 |
|
|
|
189 |
|
|
/*
|
190 |
|
|
* Cancel all timers for TCP tp.
|
191 |
|
|
*/
|
192 |
|
|
void
|
193 |
|
|
tcp_canceltimers(tp)
|
194 |
|
|
struct tcpcb *tp;
|
195 |
|
|
{
|
196 |
|
|
register int i;
|
197 |
|
|
|
198 |
|
|
for (i = 0; i < TCPT_NTIMERS; i++)
|
199 |
|
|
tp->t_timer[i] = 0;
|
200 |
|
|
}
|
201 |
|
|
|
202 |
|
|
int tcp_backoff[TCP_MAXRXTSHIFT + 1] =
|
203 |
|
|
{ 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
|
204 |
|
|
|
205 |
|
|
static int tcp_totbackoff = 511; /* sum of tcp_backoff[] */
|
206 |
|
|
|
207 |
|
|
/*
|
208 |
|
|
* TCP timer processing.
|
209 |
|
|
*/
|
210 |
|
|
struct tcpcb *
|
211 |
|
|
tcp_timers(tp, timer)
|
212 |
|
|
register struct tcpcb *tp;
|
213 |
|
|
int timer;
|
214 |
|
|
{
|
215 |
|
|
register int rexmt;
|
216 |
|
|
|
217 |
|
|
switch (timer) {
|
218 |
|
|
|
219 |
|
|
/*
|
220 |
|
|
* 2 MSL timeout in shutdown went off. If we're closed but
|
221 |
|
|
* still waiting for peer to close and connection has been idle
|
222 |
|
|
* too long, or if 2MSL time is up from TIME_WAIT, delete connection
|
223 |
|
|
* control block. Otherwise, check again in a bit.
|
224 |
|
|
*/
|
225 |
|
|
case TCPT_2MSL:
|
226 |
|
|
if (tp->t_state != TCPS_TIME_WAIT &&
|
227 |
|
|
tp->t_idle <= tcp_maxidle)
|
228 |
|
|
tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
|
229 |
|
|
else
|
230 |
|
|
tp = tcp_close(tp);
|
231 |
|
|
break;
|
232 |
|
|
|
233 |
|
|
/*
|
234 |
|
|
* Retransmission timer went off. Message has not
|
235 |
|
|
* been acked within retransmit interval. Back off
|
236 |
|
|
* to a longer retransmit interval and retransmit one segment.
|
237 |
|
|
*/
|
238 |
|
|
case TCPT_REXMT:
|
239 |
|
|
if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
|
240 |
|
|
tp->t_rxtshift = TCP_MAXRXTSHIFT;
|
241 |
|
|
tcpstat.tcps_timeoutdrop++;
|
242 |
|
|
tp = tcp_drop(tp, tp->t_softerror ?
|
243 |
|
|
tp->t_softerror : ETIMEDOUT);
|
244 |
|
|
break;
|
245 |
|
|
}
|
246 |
|
|
tcpstat.tcps_rexmttimeo++;
|
247 |
|
|
rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
|
248 |
|
|
TCPT_RANGESET(tp->t_rxtcur, rexmt,
|
249 |
|
|
tp->t_rttmin, TCPTV_REXMTMAX);
|
250 |
|
|
tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
|
251 |
|
|
/*
|
252 |
|
|
* If losing, let the lower level know and try for
|
253 |
|
|
* a better route. Also, if we backed off this far,
|
254 |
|
|
* our srtt estimate is probably bogus. Clobber it
|
255 |
|
|
* so we'll take the next rtt measurement as our srtt;
|
256 |
|
|
* move the current srtt into rttvar to keep the current
|
257 |
|
|
* retransmit times until then.
|
258 |
|
|
*/
|
259 |
|
|
if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
|
260 |
|
|
in_losing(tp->t_inpcb);
|
261 |
|
|
tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
|
262 |
|
|
tp->t_srtt = 0;
|
263 |
|
|
}
|
264 |
|
|
tp->snd_nxt = tp->snd_una;
|
265 |
|
|
/*
|
266 |
|
|
* Force a segment to be sent.
|
267 |
|
|
*/
|
268 |
|
|
tp->t_flags |= TF_ACKNOW;
|
269 |
|
|
/*
|
270 |
|
|
* If timing a segment in this window, stop the timer.
|
271 |
|
|
*/
|
272 |
|
|
tp->t_rtt = 0;
|
273 |
|
|
/*
|
274 |
|
|
* Close the congestion window down to one segment
|
275 |
|
|
* (we'll open it by one segment for each ack we get).
|
276 |
|
|
* Since we probably have a window's worth of unacked
|
277 |
|
|
* data accumulated, this "slow start" keeps us from
|
278 |
|
|
* dumping all that data as back-to-back packets (which
|
279 |
|
|
* might overwhelm an intermediate gateway).
|
280 |
|
|
*
|
281 |
|
|
* There are two phases to the opening: Initially we
|
282 |
|
|
* open by one mss on each ack. This makes the window
|
283 |
|
|
* size increase exponentially with time. If the
|
284 |
|
|
* window is larger than the path can handle, this
|
285 |
|
|
* exponential growth results in dropped packet(s)
|
286 |
|
|
* almost immediately. To get more time between
|
287 |
|
|
* drops but still "push" the network to take advantage
|
288 |
|
|
* of improving conditions, we switch from exponential
|
289 |
|
|
* to linear window opening at some threshhold size.
|
290 |
|
|
* For a threshhold, we use half the current window
|
291 |
|
|
* size, truncated to a multiple of the mss.
|
292 |
|
|
*
|
293 |
|
|
* (the minimum cwnd that will give us exponential
|
294 |
|
|
* growth is 2 mss. We don't allow the threshhold
|
295 |
|
|
* to go below this.)
|
296 |
|
|
*/
|
297 |
|
|
{
|
298 |
|
|
u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
|
299 |
|
|
if (win < 2)
|
300 |
|
|
win = 2;
|
301 |
|
|
tp->snd_cwnd = tp->t_maxseg;
|
302 |
|
|
tp->snd_ssthresh = win * tp->t_maxseg;
|
303 |
|
|
tp->t_dupacks = 0;
|
304 |
|
|
}
|
305 |
|
|
(void) tcp_output(tp);
|
306 |
|
|
break;
|
307 |
|
|
|
308 |
|
|
/*
|
309 |
|
|
* Persistance timer into zero window.
|
310 |
|
|
* Force a byte to be output, if possible.
|
311 |
|
|
*/
|
312 |
|
|
case TCPT_PERSIST:
|
313 |
|
|
tcpstat.tcps_persisttimeo++;
|
314 |
|
|
/*
|
315 |
|
|
* Hack: if the peer is dead/unreachable, we do not
|
316 |
|
|
* time out if the window is closed. After a full
|
317 |
|
|
* backoff, drop the connection if the idle time
|
318 |
|
|
* (no responses to probes) reaches the maximum
|
319 |
|
|
* backoff that we would use if retransmitting.
|
320 |
|
|
*/
|
321 |
|
|
if (tp->t_rxtshift == TCP_MAXRXTSHIFT) {
|
322 |
|
|
u_long maxidle = TCP_REXMTVAL(tp);
|
323 |
|
|
if (maxidle < tp->t_rttmin)
|
324 |
|
|
maxidle = tp->t_rttmin;
|
325 |
|
|
maxidle *= tcp_totbackoff;
|
326 |
|
|
if (tp->t_idle >= tcp_maxpersistidle ||
|
327 |
|
|
tp->t_idle >= maxidle) {
|
328 |
|
|
tcpstat.tcps_persistdrop++;
|
329 |
|
|
tp = tcp_drop(tp, ETIMEDOUT);
|
330 |
|
|
break;
|
331 |
|
|
}
|
332 |
|
|
}
|
333 |
|
|
tcp_setpersist(tp);
|
334 |
|
|
tp->t_force = 1;
|
335 |
|
|
(void) tcp_output(tp);
|
336 |
|
|
tp->t_force = 0;
|
337 |
|
|
break;
|
338 |
|
|
|
339 |
|
|
/*
|
340 |
|
|
* Keep-alive timer went off; send something
|
341 |
|
|
* or drop connection if idle for too long.
|
342 |
|
|
*/
|
343 |
|
|
case TCPT_KEEP:
|
344 |
|
|
tcpstat.tcps_keeptimeo++;
|
345 |
|
|
if (tp->t_state < TCPS_ESTABLISHED)
|
346 |
|
|
goto dropit;
|
347 |
|
|
if ((always_keepalive ||
|
348 |
|
|
tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE) &&
|
349 |
|
|
tp->t_state <= TCPS_CLOSING) {
|
350 |
|
|
if (tp->t_idle >= tcp_keepidle + tcp_maxidle)
|
351 |
|
|
goto dropit;
|
352 |
|
|
/*
|
353 |
|
|
* Send a packet designed to force a response
|
354 |
|
|
* if the peer is up and reachable:
|
355 |
|
|
* either an ACK if the connection is still alive,
|
356 |
|
|
* or an RST if the peer has closed the connection
|
357 |
|
|
* due to timeout or reboot.
|
358 |
|
|
* Using sequence number tp->snd_una-1
|
359 |
|
|
* causes the transmitted zero-length segment
|
360 |
|
|
* to lie outside the receive window;
|
361 |
|
|
* by the protocol spec, this requires the
|
362 |
|
|
* correspondent TCP to respond.
|
363 |
|
|
*/
|
364 |
|
|
tcpstat.tcps_keepprobe++;
|
365 |
|
|
#ifdef TCP_COMPAT_42
|
366 |
|
|
/*
|
367 |
|
|
* The keepalive packet must have nonzero length
|
368 |
|
|
* to get a 4.2 host to respond.
|
369 |
|
|
*/
|
370 |
|
|
tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
|
371 |
|
|
tp->rcv_nxt - 1, tp->snd_una - 1, 0);
|
372 |
|
|
#else
|
373 |
|
|
tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
|
374 |
|
|
tp->rcv_nxt, tp->snd_una - 1, 0);
|
375 |
|
|
#endif
|
376 |
|
|
tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
|
377 |
|
|
} else
|
378 |
|
|
tp->t_timer[TCPT_KEEP] = tcp_keepidle;
|
379 |
|
|
break;
|
380 |
|
|
dropit:
|
381 |
|
|
tcpstat.tcps_keepdrops++;
|
382 |
|
|
tp = tcp_drop(tp, ETIMEDOUT);
|
383 |
|
|
break;
|
384 |
|
|
}
|
385 |
|
|
return (tp);
|
386 |
|
|
}
|
387 |
|
|
#endif /* TUBA_INCLUDE */
|