2 /* $KAME: in6_rmx.c,v 1.11 2001/07/26 06:53:16 jinmei Exp $ */
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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.
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
34 * Copyright 1994, 1995 Massachusetts Institute of Technology
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
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
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
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.
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>
84 #include <sys/syslog.h>
85 #include <sys/callout.h>
88 #include <net/route.h>
89 #include <netinet/in.h>
90 #include <netinet/ip_var.h>
91 #include <netinet/in_var.h>
93 #include <netinet/ip6.h>
94 #include <netinet6/ip6_var.h>
96 #include <netinet/icmp6.h>
98 #include <netinet/tcp.h>
99 #include <netinet/tcp_seq.h>
100 #include <netinet/tcp_timer.h>
101 #include <netinet/tcp_var.h>
103 extern int in6_inithead __P((void **head, int off));
105 #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */
108 * Do what we need to do when inserting a route.
110 static struct radix_node *
111 in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
112 struct radix_node *treenodes)
114 struct rtentry *rt = (struct rtentry *)treenodes;
115 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt);
116 struct radix_node *ret;
119 * For IPv6, all unicast non-host routes are automatically cloning.
121 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
122 rt->rt_flags |= RTF_MULTICAST;
124 if (!(rt->rt_flags & (RTF_HOST | RTF_CLONING | RTF_MULTICAST))) {
125 rt->rt_flags |= RTF_PRCLONING;
129 * A little bit of help for both IPv6 output and input:
130 * For local addresses, we make sure that RTF_LOCAL is set,
131 * with the thought that this might one day be used to speed up
134 * We also mark routes to multicast addresses as such, because
135 * it's easy to do and might be useful (but this is much more
136 * dubious since it's so easy to inspect the address). (This
140 * should elaborate the code.
142 if (rt->rt_flags & RTF_HOST) {
143 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr)
146 rt->rt_flags |= RTF_LOCAL;
150 if (!rt->rt_rmx.rmx_mtu && !(rt->rt_rmx.rmx_locks & RTV_MTU)
152 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
154 ret = rn_addroute(v_arg, n_arg, head, treenodes);
155 if (ret == NULL && rt->rt_flags & RTF_HOST) {
158 * We are trying to add a host route, but can't.
159 * Find out if it is because of an
160 * ARP entry and delete it if so.
162 rt2 = rtalloc1((struct sockaddr *)sin6, 0,
163 RTF_CLONING | RTF_PRCLONING);
165 if (rt2->rt_flags & RTF_LLINFO &&
166 rt2->rt_flags & RTF_HOST &&
168 rt2->rt_gateway->sa_family == AF_LINK) {
169 /* NB: must unlock to avoid recursion */
171 rtrequest(RTM_DELETE,
172 (struct sockaddr *)rt_key(rt2),
174 rt_mask(rt2), rt2->rt_flags, 0);
175 ret = rn_addroute(v_arg, n_arg, head,
181 } else if (ret == NULL && rt->rt_flags & RTF_CLONING) {
184 * We are trying to add a net route, but can't.
185 * The following case should be allowed, so we'll make a
186 * special check for this:
187 * Two IPv6 addresses with the same prefix is assigned
188 * to a single interrface.
189 * # ifconfig if0 inet6 3ffe:0501::1 prefix 64 alias (*1)
190 * # ifconfig if0 inet6 3ffe:0501::2 prefix 64 alias (*2)
191 * In this case, (*1) and (*2) want to add the same
192 * net route entry, 3ffe:0501:: -> if0.
193 * This case should not raise an error.
195 rt2 = rtalloc1((struct sockaddr *)sin6, 0,
196 RTF_CLONING | RTF_PRCLONING);
198 if ((rt2->rt_flags & (RTF_CLONING|RTF_HOST|RTF_GATEWAY))
201 && rt2->rt_gateway->sa_family == AF_LINK
202 && rt2->rt_ifp == rt->rt_ifp) {
212 * This code is the inverse of in6_clsroute: on first reference, if we
213 * were managing the route, stop doing so and set the expiration timer
216 static struct radix_node *
217 in6_matroute(void *v_arg, struct radix_node_head *head)
219 struct radix_node *rn = rn_match(v_arg, head);
220 struct rtentry *rt = (struct rtentry *)rn;
222 if (rt && rt->rt_refcnt == 0) { /* this is first reference */
223 if (rt->rt_flags & RTPRF_OURS) {
224 rt->rt_flags &= ~RTPRF_OURS;
225 rt->rt_rmx.rmx_expire = 0;
231 SYSCTL_DECL(_net_inet6_ip6);
233 static int rtq_reallyold = 60*60;
234 /* one hour is ``really old'' */
235 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTEXPIRE, rtexpire,
236 CTLFLAG_RW, &rtq_reallyold , 0, "");
238 static int rtq_minreallyold = 10;
239 /* never automatically crank down to less */
240 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMINEXPIRE, rtminexpire,
241 CTLFLAG_RW, &rtq_minreallyold , 0, "");
243 static int rtq_toomany = 128;
244 /* 128 cached routes is ``too many'' */
245 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMAXCACHE, rtmaxcache,
246 CTLFLAG_RW, &rtq_toomany , 0, "");
250 * On last reference drop, mark the route as belong to us so that it can be
254 in6_clsroute(struct radix_node *rn, struct radix_node_head *head)
256 struct rtentry *rt = (struct rtentry *)rn;
260 if (!(rt->rt_flags & RTF_UP))
261 return; /* prophylactic measures */
263 if ((rt->rt_flags & (RTF_LLINFO | RTF_HOST)) != RTF_HOST)
266 if ((rt->rt_flags & (RTF_WASCLONED | RTPRF_OURS)) != RTF_WASCLONED)
270 * As requested by David Greenman:
271 * If rtq_reallyold is 0, just delete the route without
272 * waiting for a timeout cycle to kill it.
274 if (rtq_reallyold != 0) {
275 rt->rt_flags |= RTPRF_OURS;
276 rt->rt_rmx.rmx_expire = time_second + rtq_reallyold;
278 /* NB: must unlock to avoid recursion */
280 rtrequest(RTM_DELETE,
281 (struct sockaddr *)rt_key(rt),
282 rt->rt_gateway, rt_mask(rt),
289 struct radix_node_head *rnh;
299 * Get rid of old routes. When draining, this deletes everything, even when
300 * the timeout is not expired yet. When updating, this makes sure that
301 * nothing has a timeout longer than the current value of rtq_reallyold.
304 in6_rtqkill(struct radix_node *rn, void *rock)
306 struct rtqk_arg *ap = rock;
307 struct rtentry *rt = (struct rtentry *)rn;
310 if (rt->rt_flags & RTPRF_OURS) {
313 if (ap->draining || rt->rt_rmx.rmx_expire <= time_second) {
314 if (rt->rt_refcnt > 0)
315 panic("rtqkill route really not free");
317 err = rtrequest(RTM_DELETE,
318 (struct sockaddr *)rt_key(rt),
319 rt->rt_gateway, rt_mask(rt),
322 log(LOG_WARNING, "in6_rtqkill: error %d", err);
328 && (rt->rt_rmx.rmx_expire - time_second
330 rt->rt_rmx.rmx_expire = time_second
333 ap->nextstop = lmin(ap->nextstop,
334 rt->rt_rmx.rmx_expire);
341 #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */
342 static int rtq_timeout = RTQ_TIMEOUT;
343 static struct callout rtq_timer;
346 in6_rtqtimo(void *rock)
348 struct radix_node_head *rnh = rock;
351 static time_t last_adjusted_timeout = 0;
353 arg.found = arg.killed = 0;
355 arg.nextstop = time_second + rtq_timeout;
356 arg.draining = arg.updating = 0;
357 RADIX_NODE_HEAD_LOCK(rnh);
358 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
359 RADIX_NODE_HEAD_UNLOCK(rnh);
362 * Attempt to be somewhat dynamic about this:
363 * If there are ``too many'' routes sitting around taking up space,
364 * then crank down the timeout, and see if we can't make some more
365 * go away. However, we make sure that we will never adjust more
366 * than once in rtq_timeout seconds, to keep from cranking down too
369 if ((arg.found - arg.killed > rtq_toomany)
370 && (time_second - last_adjusted_timeout >= rtq_timeout)
371 && rtq_reallyold > rtq_minreallyold) {
372 rtq_reallyold = 2*rtq_reallyold / 3;
373 if (rtq_reallyold < rtq_minreallyold) {
374 rtq_reallyold = rtq_minreallyold;
377 last_adjusted_timeout = time_second;
379 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold to %d",
382 arg.found = arg.killed = 0;
384 RADIX_NODE_HEAD_LOCK(rnh);
385 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
386 RADIX_NODE_HEAD_UNLOCK(rnh);
390 atv.tv_sec = arg.nextstop;
391 callout_reset(&rtq_timer, tvtohz(&atv), in6_rtqtimo, rock);
398 struct radix_node_head *rnh;
401 static struct callout rtq_mtutimer;
404 in6_mtuexpire(struct radix_node *rn, void *rock)
406 struct rtentry *rt = (struct rtentry *)rn;
407 struct mtuex_arg *ap = rock;
411 panic("rt == NULL in in6_mtuexpire");
413 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) {
414 if (rt->rt_rmx.rmx_expire <= time_second) {
415 rt->rt_flags |= RTF_PROBEMTU;
417 ap->nextstop = lmin(ap->nextstop,
418 rt->rt_rmx.rmx_expire);
425 #define MTUTIMO_DEFAULT (60*1)
428 in6_mtutimo(void *rock)
430 struct radix_node_head *rnh = rock;
431 struct mtuex_arg arg;
435 arg.nextstop = time_second + MTUTIMO_DEFAULT;
436 RADIX_NODE_HEAD_LOCK(rnh);
437 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg);
438 RADIX_NODE_HEAD_UNLOCK(rnh);
441 atv.tv_sec = arg.nextstop;
442 if (atv.tv_sec < time_second) {
443 printf("invalid mtu expiration time on routing table\n");
444 arg.nextstop = time_second + 30; /* last resort */
446 callout_reset(&rtq_mtutimer, tvtohz(&atv), in6_mtutimo, rock);
453 struct radix_node_head *rnh = rt_tables[AF_INET6];
456 arg.found = arg.killed = 0;
461 RADIX_NODE_HEAD_LOCK(rnh);
462 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
463 RADIX_NODE_HEAD_UNLOCK(rnh);
468 * Initialize our routing tree.
471 in6_inithead(void **head, int off)
473 struct radix_node_head *rnh;
475 if (!rn_inithead(head, off))
478 if (head != (void **)&rt_tables[AF_INET6]) /* BOGUS! */
479 return 1; /* only do this for the real routing table */
482 rnh->rnh_addaddr = in6_addroute;
483 rnh->rnh_matchaddr = in6_matroute;
484 rnh->rnh_close = in6_clsroute;
485 callout_init(&rtq_timer, CALLOUT_MPSAFE);
486 in6_rtqtimo(rnh); /* kick off timeout first time */
487 callout_init(&rtq_mtutimer, CALLOUT_MPSAFE);
488 in6_mtutimo(rnh); /* kick off timeout first time */