2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * $KAME: in6_rmx.c,v 1.11 2001/07/26 06:53:16 jinmei Exp $
33 * Copyright 1994, 1995 Massachusetts Institute of Technology
35 * Permission to use, copy, modify, and distribute this software and
36 * its documentation for any purpose and without fee is hereby
37 * granted, provided that both the above copyright notice and this
38 * permission notice appear in all copies, that both the above
39 * copyright notice and this permission notice appear in all
40 * supporting documentation, and that the name of M.I.T. not be used
41 * in advertising or publicity pertaining to distribution of the
42 * software without specific, written prior permission. M.I.T. makes
43 * no representations about the suitability of this software for any
44 * purpose. It is provided "as is" without express or implied
47 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS
48 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
49 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
50 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
51 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
52 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
53 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
54 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
55 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
56 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
57 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * This code does two things necessary for the enhanced TCP metrics to
64 * function in a useful manner:
65 * 1) It marks all non-host routes as `cloning', thus ensuring that
66 * every actual reference to such a route actually gets turned
67 * into a reference to a host route to the specific destination
69 * 2) When such routes lose all their references, it arranges for them
70 * to be deleted in some random collection of circumstances, so that
71 * a large quantity of stale routing data is not kept in kernel memory
72 * indefinitely. See in6_rtqtimo() below for the exact mechanism.
75 #include <sys/cdefs.h>
76 __FBSDID("$FreeBSD$");
78 #include <sys/param.h>
79 #include <sys/systm.h>
80 #include <sys/kernel.h>
81 #include <sys/sysctl.h>
82 #include <sys/queue.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
86 #include <sys/syslog.h>
87 #include <sys/callout.h>
90 #include <net/route.h>
91 #include <netinet/in.h>
92 #include <netinet/ip_var.h>
93 #include <netinet/in_var.h>
95 #include <netinet/ip6.h>
96 #include <netinet6/ip6_var.h>
98 #include <netinet/icmp6.h>
99 #include <netinet6/nd6.h>
101 #include <netinet/tcp.h>
102 #include <netinet/tcp_seq.h>
103 #include <netinet/tcp_timer.h>
104 #include <netinet/tcp_var.h>
106 extern int in6_inithead(void **head, int off);
108 #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */
111 * Do what we need to do when inserting a route.
113 static struct radix_node *
114 in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
115 struct radix_node *treenodes)
117 struct rtentry *rt = (struct rtentry *)treenodes;
118 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt);
119 struct radix_node *ret;
121 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
122 rt->rt_flags |= RTF_MULTICAST;
125 * A little bit of help for both IPv6 output and input:
126 * For local addresses, we make sure that RTF_LOCAL is set,
127 * with the thought that this might one day be used to speed up
130 * We also mark routes to multicast addresses as such, because
131 * it's easy to do and might be useful (but this is much more
132 * dubious since it's so easy to inspect the address). (This
136 * should elaborate the code.
138 if (rt->rt_flags & RTF_HOST) {
139 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr)
142 rt->rt_flags |= RTF_LOCAL;
146 if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp)
147 rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp);
149 ret = rn_addroute(v_arg, n_arg, head, treenodes);
150 if (ret == NULL && rt->rt_flags & RTF_HOST) {
153 * We are trying to add a host route, but can't.
154 * Find out if it is because of an
155 * ARP entry and delete it if so.
157 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED|RTF_CLONING);
159 if (rt2->rt_flags & RTF_LLINFO &&
160 rt2->rt_flags & RTF_HOST &&
162 rt2->rt_gateway->sa_family == AF_LINK) {
165 ret = rn_addroute(v_arg, n_arg, head,
170 } else if (ret == NULL && rt->rt_flags & RTF_CLONING) {
173 * We are trying to add a net route, but can't.
174 * The following case should be allowed, so we'll make a
175 * special check for this:
176 * Two IPv6 addresses with the same prefix is assigned
177 * to a single interrface.
178 * # ifconfig if0 inet6 3ffe:0501::1 prefix 64 alias (*1)
179 * # ifconfig if0 inet6 3ffe:0501::2 prefix 64 alias (*2)
180 * In this case, (*1) and (*2) want to add the same
181 * net route entry, 3ffe:0501:: -> if0.
182 * This case should not raise an error.
184 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED|RTF_CLONING);
186 if ((rt2->rt_flags & (RTF_CLONING|RTF_HOST|RTF_GATEWAY))
189 && rt2->rt_gateway->sa_family == AF_LINK
190 && rt2->rt_ifp == rt->rt_ifp) {
200 * This code is the inverse of in6_clsroute: on first reference, if we
201 * were managing the route, stop doing so and set the expiration timer
204 static struct radix_node *
205 in6_matroute(void *v_arg, struct radix_node_head *head)
207 struct radix_node *rn = rn_match(v_arg, head);
208 struct rtentry *rt = (struct rtentry *)rn;
210 if (rt && rt->rt_refcnt == 0) { /* this is first reference */
211 if (rt->rt_flags & RTPRF_OURS) {
212 rt->rt_flags &= ~RTPRF_OURS;
213 rt->rt_rmx.rmx_expire = 0;
219 SYSCTL_DECL(_net_inet6_ip6);
221 static int rtq_reallyold = 60*60;
222 /* one hour is ``really old'' */
223 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTEXPIRE, rtexpire,
224 CTLFLAG_RW, &rtq_reallyold , 0, "");
226 static int rtq_minreallyold = 10;
227 /* never automatically crank down to less */
228 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMINEXPIRE, rtminexpire,
229 CTLFLAG_RW, &rtq_minreallyold , 0, "");
231 static int rtq_toomany = 128;
232 /* 128 cached routes is ``too many'' */
233 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMAXCACHE, rtmaxcache,
234 CTLFLAG_RW, &rtq_toomany , 0, "");
238 * On last reference drop, mark the route as belong to us so that it can be
242 in6_clsroute(struct radix_node *rn, struct radix_node_head *head)
244 struct rtentry *rt = (struct rtentry *)rn;
248 if (!(rt->rt_flags & RTF_UP))
249 return; /* prophylactic measures */
251 if ((rt->rt_flags & (RTF_LLINFO | RTF_HOST)) != RTF_HOST)
254 if ((rt->rt_flags & (RTF_WASCLONED | RTPRF_OURS)) != RTF_WASCLONED)
258 * As requested by David Greenman:
259 * If rtq_reallyold is 0, just delete the route without
260 * waiting for a timeout cycle to kill it.
262 if (rtq_reallyold != 0) {
263 rt->rt_flags |= RTPRF_OURS;
264 rt->rt_rmx.rmx_expire = time_uptime + rtq_reallyold;
271 struct radix_node_head *rnh;
281 * Get rid of old routes. When draining, this deletes everything, even when
282 * the timeout is not expired yet. When updating, this makes sure that
283 * nothing has a timeout longer than the current value of rtq_reallyold.
286 in6_rtqkill(struct radix_node *rn, void *rock)
288 struct rtqk_arg *ap = rock;
289 struct rtentry *rt = (struct rtentry *)rn;
292 RADIX_NODE_HEAD_LOCK_ASSERT(ap->rnh);
294 if (rt->rt_flags & RTPRF_OURS) {
297 if (ap->draining || rt->rt_rmx.rmx_expire <= time_uptime) {
298 if (rt->rt_refcnt > 0)
299 panic("rtqkill route really not free");
301 err = rtrequest(RTM_DELETE,
302 (struct sockaddr *)rt_key(rt),
303 rt->rt_gateway, rt_mask(rt),
304 rt->rt_flags|RTF_RNH_LOCKED, 0);
306 log(LOG_WARNING, "in6_rtqkill: error %d", err);
312 && (rt->rt_rmx.rmx_expire - time_uptime
314 rt->rt_rmx.rmx_expire = time_uptime
317 ap->nextstop = lmin(ap->nextstop,
318 rt->rt_rmx.rmx_expire);
325 #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */
326 static int rtq_timeout = RTQ_TIMEOUT;
327 static struct callout rtq_timer;
330 in6_rtqtimo(void *rock)
332 struct radix_node_head *rnh = rock;
335 static time_t last_adjusted_timeout = 0;
337 arg.found = arg.killed = 0;
339 arg.nextstop = time_uptime + rtq_timeout;
340 arg.draining = arg.updating = 0;
341 RADIX_NODE_HEAD_LOCK(rnh);
342 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
343 RADIX_NODE_HEAD_UNLOCK(rnh);
346 * Attempt to be somewhat dynamic about this:
347 * If there are ``too many'' routes sitting around taking up space,
348 * then crank down the timeout, and see if we can't make some more
349 * go away. However, we make sure that we will never adjust more
350 * than once in rtq_timeout seconds, to keep from cranking down too
353 if ((arg.found - arg.killed > rtq_toomany)
354 && (time_uptime - last_adjusted_timeout >= rtq_timeout)
355 && rtq_reallyold > rtq_minreallyold) {
356 rtq_reallyold = 2*rtq_reallyold / 3;
357 if (rtq_reallyold < rtq_minreallyold) {
358 rtq_reallyold = rtq_minreallyold;
361 last_adjusted_timeout = time_uptime;
363 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold to %d",
366 arg.found = arg.killed = 0;
368 RADIX_NODE_HEAD_LOCK(rnh);
369 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
370 RADIX_NODE_HEAD_UNLOCK(rnh);
374 atv.tv_sec = arg.nextstop - time_uptime;
375 callout_reset(&rtq_timer, tvtohz(&atv), in6_rtqtimo, rock);
382 struct radix_node_head *rnh;
385 static struct callout rtq_mtutimer;
388 in6_mtuexpire(struct radix_node *rn, void *rock)
390 struct rtentry *rt = (struct rtentry *)rn;
391 struct mtuex_arg *ap = rock;
395 panic("rt == NULL in in6_mtuexpire");
397 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) {
398 if (rt->rt_rmx.rmx_expire <= time_uptime) {
399 rt->rt_flags |= RTF_PROBEMTU;
401 ap->nextstop = lmin(ap->nextstop,
402 rt->rt_rmx.rmx_expire);
409 #define MTUTIMO_DEFAULT (60*1)
412 in6_mtutimo(void *rock)
414 struct radix_node_head *rnh = rock;
415 struct mtuex_arg arg;
419 arg.nextstop = time_uptime + MTUTIMO_DEFAULT;
420 RADIX_NODE_HEAD_LOCK(rnh);
421 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg);
422 RADIX_NODE_HEAD_UNLOCK(rnh);
425 atv.tv_sec = arg.nextstop - time_uptime;
426 if (atv.tv_sec < 0) {
427 printf("invalid mtu expiration time on routing table\n");
428 arg.nextstop = time_uptime + 30; /* last resort */
431 callout_reset(&rtq_mtutimer, tvtohz(&atv), in6_mtutimo, rock);
438 struct radix_node_head *rnh = rt_tables[AF_INET6];
441 arg.found = arg.killed = 0;
446 RADIX_NODE_HEAD_LOCK(rnh);
447 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
448 RADIX_NODE_HEAD_UNLOCK(rnh);
453 * Initialize our routing tree.
454 * XXX MRT When off == 0, we are being called from vfs_export.c
455 * so just set up their table and leave. (we know what the correct
456 * value should be so just use that).. FIX AFTER RELENG_7 is MFC'd
457 * see also comments in in_inithead() vfs_export.c and domain.h
460 in6_inithead(void **head, int off)
462 struct radix_node_head *rnh;
464 if (!rn_inithead(head, offsetof(struct sockaddr_in6, sin6_addr) << 3))
465 return 0; /* See above */
467 if (off == 0) /* See above */
468 return 1; /* only do the rest for the real thing */
471 rnh->rnh_addaddr = in6_addroute;
472 rnh->rnh_matchaddr = in6_matroute;
473 rnh->rnh_close = in6_clsroute;
474 callout_init(&rtq_timer, CALLOUT_MPSAFE);
475 in6_rtqtimo(rnh); /* kick off timeout first time */
476 callout_init(&rtq_mtutimer, CALLOUT_MPSAFE);
477 in6_mtutimo(rnh); /* kick off timeout first time */