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>
82 #include <sys/sysctl.h>
83 #include <sys/queue.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
87 #include <sys/rwlock.h>
88 #include <sys/syslog.h>
89 #include <sys/callout.h>
92 #include <net/route.h>
94 #include <netinet/in.h>
95 #include <netinet/ip_var.h>
96 #include <netinet/in_var.h>
98 #include <netinet/ip6.h>
99 #include <netinet6/ip6_var.h>
101 #include <netinet/icmp6.h>
102 #include <netinet6/nd6.h>
104 #include <netinet/tcp.h>
105 #include <netinet/tcp_seq.h>
106 #include <netinet/tcp_timer.h>
107 #include <netinet/tcp_var.h>
109 extern int in6_inithead(void **head, int off);
111 extern int in6_detachhead(void **head, int off);
114 #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */
117 * Do what we need to do when inserting a route.
119 static struct radix_node *
120 in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
121 struct radix_node *treenodes)
123 struct rtentry *rt = (struct rtentry *)treenodes;
124 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt);
125 struct radix_node *ret;
127 RADIX_NODE_HEAD_WLOCK_ASSERT(head);
128 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
129 rt->rt_flags |= RTF_MULTICAST;
132 * A little bit of help for both IPv6 output and input:
133 * For local addresses, we make sure that RTF_LOCAL is set,
134 * with the thought that this might one day be used to speed up
137 * We also mark routes to multicast addresses as such, because
138 * it's easy to do and might be useful (but this is much more
139 * dubious since it's so easy to inspect the address). (This
143 * should elaborate the code.
145 if (rt->rt_flags & RTF_HOST) {
146 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr)
149 rt->rt_flags |= RTF_LOCAL;
153 if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp)
154 rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp);
156 ret = rn_addroute(v_arg, n_arg, head, treenodes);
160 * We are trying to add a net route, but can't.
161 * The following case should be allowed, so we'll make a
162 * special check for this:
163 * Two IPv6 addresses with the same prefix is assigned
164 * to a single interrface.
165 * # ifconfig if0 inet6 3ffe:0501::1 prefix 64 alias (*1)
166 * # ifconfig if0 inet6 3ffe:0501::2 prefix 64 alias (*2)
167 * In this case, (*1) and (*2) want to add the same
168 * net route entry, 3ffe:0501:: -> if0.
169 * This case should not raise an error.
171 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED);
173 if (((rt2->rt_flags & (RTF_HOST|RTF_GATEWAY)) == 0)
175 && rt2->rt_gateway->sa_family == AF_LINK
176 && rt2->rt_ifp == rt->rt_ifp) {
186 * This code is the inverse of in6_clsroute: on first reference, if we
187 * were managing the route, stop doing so and set the expiration timer
190 static struct radix_node *
191 in6_matroute(void *v_arg, struct radix_node_head *head)
193 struct radix_node *rn = rn_match(v_arg, head);
194 struct rtentry *rt = (struct rtentry *)rn;
196 if (rt && rt->rt_refcnt == 0) { /* this is first reference */
197 if (rt->rt_flags & RTPRF_OURS) {
198 rt->rt_flags &= ~RTPRF_OURS;
199 rt->rt_rmx.rmx_expire = 0;
205 SYSCTL_DECL(_net_inet6_ip6);
207 static VNET_DEFINE(int, rtq_reallyold6) = 60*60;
208 /* one hour is ``really old'' */
209 #define V_rtq_reallyold6 VNET(rtq_reallyold6)
210 SYSCTL_VNET_INT(_net_inet6_ip6, IPV6CTL_RTEXPIRE, rtexpire, CTLFLAG_RW,
211 &VNET_NAME(rtq_reallyold6) , 0, "");
213 static VNET_DEFINE(int, rtq_minreallyold6) = 10;
214 /* never automatically crank down to less */
215 #define V_rtq_minreallyold6 VNET(rtq_minreallyold6)
216 SYSCTL_VNET_INT(_net_inet6_ip6, IPV6CTL_RTMINEXPIRE, rtminexpire, CTLFLAG_RW,
217 &VNET_NAME(rtq_minreallyold6) , 0, "");
219 static VNET_DEFINE(int, rtq_toomany6) = 128;
220 /* 128 cached routes is ``too many'' */
221 #define V_rtq_toomany6 VNET(rtq_toomany6)
222 SYSCTL_VNET_INT(_net_inet6_ip6, IPV6CTL_RTMAXCACHE, rtmaxcache, CTLFLAG_RW,
223 &VNET_NAME(rtq_toomany6) , 0, "");
226 struct radix_node_head *rnh;
236 * Get rid of old routes. When draining, this deletes everything, even when
237 * the timeout is not expired yet. When updating, this makes sure that
238 * nothing has a timeout longer than the current value of rtq_reallyold6.
241 in6_rtqkill(struct radix_node *rn, void *rock)
243 struct rtqk_arg *ap = rock;
244 struct rtentry *rt = (struct rtentry *)rn;
247 RADIX_NODE_HEAD_WLOCK_ASSERT(ap->rnh);
249 if (rt->rt_flags & RTPRF_OURS) {
252 if (ap->draining || rt->rt_rmx.rmx_expire <= time_uptime) {
253 if (rt->rt_refcnt > 0)
254 panic("rtqkill route really not free");
256 err = rtrequest(RTM_DELETE,
257 (struct sockaddr *)rt_key(rt),
258 rt->rt_gateway, rt_mask(rt),
259 rt->rt_flags|RTF_RNH_LOCKED, 0);
261 log(LOG_WARNING, "in6_rtqkill: error %d", err);
267 && (rt->rt_rmx.rmx_expire - time_uptime
268 > V_rtq_reallyold6)) {
269 rt->rt_rmx.rmx_expire = time_uptime
272 ap->nextstop = lmin(ap->nextstop,
273 rt->rt_rmx.rmx_expire);
280 #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */
281 static VNET_DEFINE(int, rtq_timeout6) = RTQ_TIMEOUT;
282 static VNET_DEFINE(struct callout, rtq_timer6);
284 #define V_rtq_timeout6 VNET(rtq_timeout6)
285 #define V_rtq_timer6 VNET(rtq_timer6)
288 in6_rtqtimo(void *rock)
290 CURVNET_SET_QUIET((struct vnet *) rock);
291 struct radix_node_head *rnh;
294 static time_t last_adjusted_timeout = 0;
296 rnh = rt_tables_get_rnh(0, AF_INET6);
301 arg.found = arg.killed = 0;
303 arg.nextstop = time_uptime + V_rtq_timeout6;
304 arg.draining = arg.updating = 0;
305 RADIX_NODE_HEAD_LOCK(rnh);
306 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
307 RADIX_NODE_HEAD_UNLOCK(rnh);
310 * Attempt to be somewhat dynamic about this:
311 * If there are ``too many'' routes sitting around taking up space,
312 * then crank down the timeout, and see if we can't make some more
313 * go away. However, we make sure that we will never adjust more
314 * than once in rtq_timeout6 seconds, to keep from cranking down too
317 if ((arg.found - arg.killed > V_rtq_toomany6)
318 && (time_uptime - last_adjusted_timeout >= V_rtq_timeout6)
319 && V_rtq_reallyold6 > V_rtq_minreallyold6) {
320 V_rtq_reallyold6 = 2*V_rtq_reallyold6 / 3;
321 if (V_rtq_reallyold6 < V_rtq_minreallyold6) {
322 V_rtq_reallyold6 = V_rtq_minreallyold6;
325 last_adjusted_timeout = time_uptime;
327 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold6 to %d",
330 arg.found = arg.killed = 0;
332 RADIX_NODE_HEAD_LOCK(rnh);
333 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
334 RADIX_NODE_HEAD_UNLOCK(rnh);
338 atv.tv_sec = arg.nextstop - time_uptime;
339 callout_reset(&V_rtq_timer6, tvtohz(&atv), in6_rtqtimo, rock);
347 struct radix_node_head *rnh;
350 static VNET_DEFINE(struct callout, rtq_mtutimer);
351 #define V_rtq_mtutimer VNET(rtq_mtutimer)
354 in6_mtuexpire(struct radix_node *rn, void *rock)
356 struct rtentry *rt = (struct rtentry *)rn;
357 struct mtuex_arg *ap = rock;
361 panic("rt == NULL in in6_mtuexpire");
363 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) {
364 if (rt->rt_rmx.rmx_expire <= time_uptime) {
365 rt->rt_flags |= RTF_PROBEMTU;
367 ap->nextstop = lmin(ap->nextstop,
368 rt->rt_rmx.rmx_expire);
375 #define MTUTIMO_DEFAULT (60*1)
378 in6_mtutimo(void *rock)
380 CURVNET_SET_QUIET((struct vnet *) rock);
381 struct radix_node_head *rnh;
382 struct mtuex_arg arg;
385 rnh = rt_tables_get_rnh(0, AF_INET6);
391 arg.nextstop = time_uptime + MTUTIMO_DEFAULT;
392 RADIX_NODE_HEAD_LOCK(rnh);
393 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg);
394 RADIX_NODE_HEAD_UNLOCK(rnh);
397 atv.tv_sec = arg.nextstop - time_uptime;
398 if (atv.tv_sec < 0) {
399 printf("invalid mtu expiration time on routing table\n");
400 arg.nextstop = time_uptime + 30; /* last resort */
403 callout_reset(&V_rtq_mtutimer, tvtohz(&atv), in6_mtutimo, rock);
408 * Initialize our routing tree.
409 * XXX MRT When off == 0, we are being called from vfs_export.c
410 * so just set up their table and leave. (we know what the correct
411 * value should be so just use that).. FIX AFTER RELENG_7 is MFC'd
412 * see also comments in in_inithead() vfs_export.c and domain.h
415 in6_inithead(void **head, int off)
417 struct radix_node_head *rnh;
419 if (!rn_inithead(head, offsetof(struct sockaddr_in6, sin6_addr) << 3))
420 return 0; /* See above */
422 if (off == 0) /* See above */
423 return 1; /* only do the rest for the real thing */
426 KASSERT(rnh == rt_tables_get_rnh(0, AF_INET6), ("rnh?"));
427 rnh->rnh_addaddr = in6_addroute;
428 rnh->rnh_matchaddr = in6_matroute;
429 callout_init(&V_rtq_timer6, CALLOUT_MPSAFE);
430 callout_init(&V_rtq_mtutimer, CALLOUT_MPSAFE);
431 in6_rtqtimo(curvnet); /* kick off timeout first time */
432 in6_mtutimo(curvnet); /* kick off timeout first time */
438 in6_detachhead(void **head, int off)
441 callout_drain(&V_rtq_timer6);
442 callout_drain(&V_rtq_mtutimer);