1 /* $FreeBSD: releng/5.1/sys/netinet6/in6_rmx.c 108250 2002-12-24 03:03:39Z hsu $ */
2 /* $KAME: in6_rmx.c,v 1.11 2001/07/26 06:53:16 jinmei Exp $ */
3
4 /*
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33 /*
34 * Copyright 1994, 1995 Massachusetts Institute of Technology
35 *
36 * Permission to use, copy, modify, and distribute this software and
37 * its documentation for any purpose and without fee is hereby
38 * granted, provided that both the above copyright notice and this
39 * permission notice appear in all copies, that both the above
40 * copyright notice and this permission notice appear in all
41 * supporting documentation, and that the name of M.I.T. not be used
42 * in advertising or publicity pertaining to distribution of the
43 * software without specific, written prior permission. M.I.T. makes
44 * no representations about the suitability of this software for any
45 * purpose. It is provided "as is" without express or implied
46 * warranty.
47 *
48 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS
49 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
50 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
51 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
52 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
53 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
54 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
55 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
56 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
57 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
58 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 */
62
63 /*
64 * This code does two things necessary for the enhanced TCP metrics to
65 * function in a useful manner:
66 * 1) It marks all non-host routes as `cloning', thus ensuring that
67 * every actual reference to such a route actually gets turned
68 * into a reference to a host route to the specific destination
69 * requested.
70 * 2) When such routes lose all their references, it arranges for them
71 * to be deleted in some random collection of circumstances, so that
72 * a large quantity of stale routing data is not kept in kernel memory
73 * indefinitely. See in6_rtqtimo() below for the exact mechanism.
74 */
75
76 #include <sys/param.h>
77 #include <sys/systm.h>
78 #include <sys/kernel.h>
79 #include <sys/sysctl.h>
80 #include <sys/queue.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/mbuf.h>
84 #include <sys/syslog.h>
85
86 #include <net/if.h>
87 #include <net/route.h>
88 #include <netinet/in.h>
89 #include <netinet/ip_var.h>
90 #include <netinet/in_var.h>
91
92 #include <netinet/ip6.h>
93 #include <netinet6/ip6_var.h>
94
95 #include <netinet/icmp6.h>
96
97 #include <netinet/tcp.h>
98 #include <netinet/tcp_seq.h>
99 #include <netinet/tcp_timer.h>
100 #include <netinet/tcp_var.h>
101
102 extern int in6_inithead __P((void **head, int off));
103
104 #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */
105
106 /*
107 * Do what we need to do when inserting a route.
108 */
109 static struct radix_node *
110 in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
111 struct radix_node *treenodes)
112 {
113 struct rtentry *rt = (struct rtentry *)treenodes;
114 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt);
115 struct radix_node *ret;
116
117 /*
118 * For IPv6, all unicast non-host routes are automatically cloning.
119 */
120 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
121 rt->rt_flags |= RTF_MULTICAST;
122
123 if (!(rt->rt_flags & (RTF_HOST | RTF_CLONING | RTF_MULTICAST))) {
124 rt->rt_flags |= RTF_PRCLONING;
125 }
126
127 /*
128 * A little bit of help for both IPv6 output and input:
129 * For local addresses, we make sure that RTF_LOCAL is set,
130 * with the thought that this might one day be used to speed up
131 * ip_input().
132 *
133 * We also mark routes to multicast addresses as such, because
134 * it's easy to do and might be useful (but this is much more
135 * dubious since it's so easy to inspect the address). (This
136 * is done above.)
137 *
138 * XXX
139 * should elaborate the code.
140 */
141 if (rt->rt_flags & RTF_HOST) {
142 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr)
143 ->sin6_addr,
144 &sin6->sin6_addr)) {
145 rt->rt_flags |= RTF_LOCAL;
146 }
147 }
148
149 if (!rt->rt_rmx.rmx_mtu && !(rt->rt_rmx.rmx_locks & RTV_MTU)
150 && rt->rt_ifp)
151 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
152
153 ret = rn_addroute(v_arg, n_arg, head, treenodes);
154 if (ret == NULL && rt->rt_flags & RTF_HOST) {
155 struct rtentry *rt2;
156 /*
157 * We are trying to add a host route, but can't.
158 * Find out if it is because of an
159 * ARP entry and delete it if so.
160 */
161 rt2 = rtalloc1((struct sockaddr *)sin6, 0,
162 RTF_CLONING | RTF_PRCLONING);
163 if (rt2) {
164 if (rt2->rt_flags & RTF_LLINFO &&
165 rt2->rt_flags & RTF_HOST &&
166 rt2->rt_gateway &&
167 rt2->rt_gateway->sa_family == AF_LINK) {
168 rtrequest(RTM_DELETE,
169 (struct sockaddr *)rt_key(rt2),
170 rt2->rt_gateway,
171 rt_mask(rt2), rt2->rt_flags, 0);
172 ret = rn_addroute(v_arg, n_arg, head,
173 treenodes);
174 }
175 RTFREE(rt2);
176 }
177 } else if (ret == NULL && rt->rt_flags & RTF_CLONING) {
178 struct rtentry *rt2;
179 /*
180 * We are trying to add a net route, but can't.
181 * The following case should be allowed, so we'll make a
182 * special check for this:
183 * Two IPv6 addresses with the same prefix is assigned
184 * to a single interrface.
185 * # ifconfig if0 inet6 3ffe:0501::1 prefix 64 alias (*1)
186 * # ifconfig if0 inet6 3ffe:0501::2 prefix 64 alias (*2)
187 * In this case, (*1) and (*2) want to add the same
188 * net route entry, 3ffe:0501:: -> if0.
189 * This case should not raise an error.
190 */
191 rt2 = rtalloc1((struct sockaddr *)sin6, 0,
192 RTF_CLONING | RTF_PRCLONING);
193 if (rt2) {
194 if ((rt2->rt_flags & (RTF_CLONING|RTF_HOST|RTF_GATEWAY))
195 == RTF_CLONING
196 && rt2->rt_gateway
197 && rt2->rt_gateway->sa_family == AF_LINK
198 && rt2->rt_ifp == rt->rt_ifp) {
199 ret = rt2->rt_nodes;
200 }
201 RTFREE(rt2);
202 }
203 }
204 return ret;
205 }
206
207 /*
208 * This code is the inverse of in6_clsroute: on first reference, if we
209 * were managing the route, stop doing so and set the expiration timer
210 * back off again.
211 */
212 static struct radix_node *
213 in6_matroute(void *v_arg, struct radix_node_head *head)
214 {
215 struct radix_node *rn = rn_match(v_arg, head);
216 struct rtentry *rt = (struct rtentry *)rn;
217
218 if (rt && rt->rt_refcnt == 0) { /* this is first reference */
219 if (rt->rt_flags & RTPRF_OURS) {
220 rt->rt_flags &= ~RTPRF_OURS;
221 rt->rt_rmx.rmx_expire = 0;
222 }
223 }
224 return rn;
225 }
226
227 SYSCTL_DECL(_net_inet6_ip6);
228
229 static int rtq_reallyold = 60*60;
230 /* one hour is ``really old'' */
231 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTEXPIRE, rtexpire,
232 CTLFLAG_RW, &rtq_reallyold , 0, "");
233
234 static int rtq_minreallyold = 10;
235 /* never automatically crank down to less */
236 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMINEXPIRE, rtminexpire,
237 CTLFLAG_RW, &rtq_minreallyold , 0, "");
238
239 static int rtq_toomany = 128;
240 /* 128 cached routes is ``too many'' */
241 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMAXCACHE, rtmaxcache,
242 CTLFLAG_RW, &rtq_toomany , 0, "");
243
244
245 /*
246 * On last reference drop, mark the route as belong to us so that it can be
247 * timed out.
248 */
249 static void
250 in6_clsroute(struct radix_node *rn, struct radix_node_head *head)
251 {
252 struct rtentry *rt = (struct rtentry *)rn;
253
254 if (!(rt->rt_flags & RTF_UP))
255 return; /* prophylactic measures */
256
257 if ((rt->rt_flags & (RTF_LLINFO | RTF_HOST)) != RTF_HOST)
258 return;
259
260 if ((rt->rt_flags & (RTF_WASCLONED | RTPRF_OURS)) != RTF_WASCLONED)
261 return;
262
263 /*
264 * As requested by David Greenman:
265 * If rtq_reallyold is 0, just delete the route without
266 * waiting for a timeout cycle to kill it.
267 */
268 if (rtq_reallyold != 0) {
269 rt->rt_flags |= RTPRF_OURS;
270 rt->rt_rmx.rmx_expire = time_second + rtq_reallyold;
271 } else {
272 rtrequest(RTM_DELETE,
273 (struct sockaddr *)rt_key(rt),
274 rt->rt_gateway, rt_mask(rt),
275 rt->rt_flags, 0);
276 }
277 }
278
279 struct rtqk_arg {
280 struct radix_node_head *rnh;
281 int mode;
282 int updating;
283 int draining;
284 int killed;
285 int found;
286 time_t nextstop;
287 };
288
289 /*
290 * Get rid of old routes. When draining, this deletes everything, even when
291 * the timeout is not expired yet. When updating, this makes sure that
292 * nothing has a timeout longer than the current value of rtq_reallyold.
293 */
294 static int
295 in6_rtqkill(struct radix_node *rn, void *rock)
296 {
297 struct rtqk_arg *ap = rock;
298 struct rtentry *rt = (struct rtentry *)rn;
299 int err;
300
301 if (rt->rt_flags & RTPRF_OURS) {
302 ap->found++;
303
304 if (ap->draining || rt->rt_rmx.rmx_expire <= time_second) {
305 if (rt->rt_refcnt > 0)
306 panic("rtqkill route really not free");
307
308 err = rtrequest(RTM_DELETE,
309 (struct sockaddr *)rt_key(rt),
310 rt->rt_gateway, rt_mask(rt),
311 rt->rt_flags, 0);
312 if (err) {
313 log(LOG_WARNING, "in6_rtqkill: error %d", err);
314 } else {
315 ap->killed++;
316 }
317 } else {
318 if (ap->updating
319 && (rt->rt_rmx.rmx_expire - time_second
320 > rtq_reallyold)) {
321 rt->rt_rmx.rmx_expire = time_second
322 + rtq_reallyold;
323 }
324 ap->nextstop = lmin(ap->nextstop,
325 rt->rt_rmx.rmx_expire);
326 }
327 }
328
329 return 0;
330 }
331
332 #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */
333 static int rtq_timeout = RTQ_TIMEOUT;
334
335 static void
336 in6_rtqtimo(void *rock)
337 {
338 struct radix_node_head *rnh = rock;
339 struct rtqk_arg arg;
340 struct timeval atv;
341 static time_t last_adjusted_timeout = 0;
342 int s;
343
344 arg.found = arg.killed = 0;
345 arg.rnh = rnh;
346 arg.nextstop = time_second + rtq_timeout;
347 arg.draining = arg.updating = 0;
348 s = splnet();
349 RADIX_NODE_HEAD_LOCK(rnh);
350 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
351 RADIX_NODE_HEAD_UNLOCK(rnh);
352 splx(s);
353
354 /*
355 * Attempt to be somewhat dynamic about this:
356 * If there are ``too many'' routes sitting around taking up space,
357 * then crank down the timeout, and see if we can't make some more
358 * go away. However, we make sure that we will never adjust more
359 * than once in rtq_timeout seconds, to keep from cranking down too
360 * hard.
361 */
362 if ((arg.found - arg.killed > rtq_toomany)
363 && (time_second - last_adjusted_timeout >= rtq_timeout)
364 && rtq_reallyold > rtq_minreallyold) {
365 rtq_reallyold = 2*rtq_reallyold / 3;
366 if (rtq_reallyold < rtq_minreallyold) {
367 rtq_reallyold = rtq_minreallyold;
368 }
369
370 last_adjusted_timeout = time_second;
371 #ifdef DIAGNOSTIC
372 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold to %d",
373 rtq_reallyold);
374 #endif
375 arg.found = arg.killed = 0;
376 arg.updating = 1;
377 s = splnet();
378 RADIX_NODE_HEAD_LOCK(rnh);
379 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
380 RADIX_NODE_HEAD_UNLOCK(rnh);
381 splx(s);
382 }
383
384 atv.tv_usec = 0;
385 atv.tv_sec = arg.nextstop;
386 timeout(in6_rtqtimo, rock, tvtohz(&atv));
387 }
388
389 /*
390 * Age old PMTUs.
391 */
392 struct mtuex_arg {
393 struct radix_node_head *rnh;
394 time_t nextstop;
395 };
396
397 static int
398 in6_mtuexpire(struct radix_node *rn, void *rock)
399 {
400 struct rtentry *rt = (struct rtentry *)rn;
401 struct mtuex_arg *ap = rock;
402
403 /* sanity */
404 if (!rt)
405 panic("rt == NULL in in6_mtuexpire");
406
407 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) {
408 if (rt->rt_rmx.rmx_expire <= time_second) {
409 rt->rt_flags |= RTF_PROBEMTU;
410 } else {
411 ap->nextstop = lmin(ap->nextstop,
412 rt->rt_rmx.rmx_expire);
413 }
414 }
415
416 return 0;
417 }
418
419 #define MTUTIMO_DEFAULT (60*1)
420
421 static void
422 in6_mtutimo(void *rock)
423 {
424 struct radix_node_head *rnh = rock;
425 struct mtuex_arg arg;
426 struct timeval atv;
427 int s;
428
429 arg.rnh = rnh;
430 arg.nextstop = time_second + MTUTIMO_DEFAULT;
431 s = splnet();
432 RADIX_NODE_HEAD_LOCK(rnh);
433 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg);
434 RADIX_NODE_HEAD_UNLOCK(rnh);
435 splx(s);
436
437 atv.tv_usec = 0;
438 atv.tv_sec = arg.nextstop;
439 if (atv.tv_sec < time_second) {
440 printf("invalid mtu expiration time on routing table\n");
441 arg.nextstop = time_second + 30; /* last resort */
442 }
443 timeout(in6_mtutimo, rock, tvtohz(&atv));
444 }
445
446 #if 0
447 void
448 in6_rtqdrain()
449 {
450 struct radix_node_head *rnh = rt_tables[AF_INET6];
451 struct rtqk_arg arg;
452 int s;
453 arg.found = arg.killed = 0;
454 arg.rnh = rnh;
455 arg.nextstop = 0;
456 arg.draining = 1;
457 arg.updating = 0;
458 s = splnet();
459 RADIX_NODE_HEAD_LOCK(rnh);
460 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
461 RADIX_NODE_HEAD_UNLOCK(rnh);
462 splx(s);
463 }
464 #endif
465
466 /*
467 * Initialize our routing tree.
468 */
469 int
470 in6_inithead(void **head, int off)
471 {
472 struct radix_node_head *rnh;
473
474 if (!rn_inithead(head, off))
475 return 0;
476
477 if (head != (void **)&rt_tables[AF_INET6]) /* BOGUS! */
478 return 1; /* only do this for the real routing table */
479
480 rnh = *head;
481 rnh->rnh_addaddr = in6_addroute;
482 rnh->rnh_matchaddr = in6_matroute;
483 rnh->rnh_close = in6_clsroute;
484 in6_rtqtimo(rnh); /* kick off timeout first time */
485 in6_mtutimo(rnh); /* kick off timeout first time */
486 return 1;
487 }
Cache object: dd10d975371d2f4016214475cf887940
|