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 "opt_route.h"
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/kernel.h>
84 #include <sys/sysctl.h>
85 #include <sys/queue.h>
86 #include <sys/socket.h>
87 #include <sys/socketvar.h>
89 #include <sys/rwlock.h>
90 #include <sys/syslog.h>
91 #include <sys/callout.h>
92 #include <sys/vimage.h>
95 #include <net/route.h>
98 #include <netinet/in.h>
99 #include <netinet/ip_var.h>
100 #include <netinet/in_var.h>
102 #include <netinet/ip6.h>
103 #include <netinet6/ip6_var.h>
105 #include <netinet/icmp6.h>
106 #include <netinet6/nd6.h>
107 #include <netinet6/vinet6.h>
109 #include <netinet/tcp.h>
110 #include <netinet/tcp_seq.h>
111 #include <netinet/tcp_timer.h>
112 #include <netinet/tcp_var.h>
114 extern int in6_inithead(void **head, int off);
116 #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */
119 * Do what we need to do when inserting a route.
121 static struct radix_node *
122 in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
123 struct radix_node *treenodes)
125 struct rtentry *rt = (struct rtentry *)treenodes;
126 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt);
127 struct radix_node *ret;
129 RADIX_NODE_HEAD_WLOCK_ASSERT(head);
130 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
131 rt->rt_flags |= RTF_MULTICAST;
134 * A little bit of help for both IPv6 output and input:
135 * For local addresses, we make sure that RTF_LOCAL is set,
136 * with the thought that this might one day be used to speed up
139 * We also mark routes to multicast addresses as such, because
140 * it's easy to do and might be useful (but this is much more
141 * dubious since it's so easy to inspect the address). (This
145 * should elaborate the code.
147 if (rt->rt_flags & RTF_HOST) {
148 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr)
151 rt->rt_flags |= RTF_LOCAL;
155 if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp)
156 rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp);
158 ret = rn_addroute(v_arg, n_arg, head, treenodes);
162 * We are trying to add a net route, but can't.
163 * The following case should be allowed, so we'll make a
164 * special check for this:
165 * Two IPv6 addresses with the same prefix is assigned
166 * to a single interrface.
167 * # ifconfig if0 inet6 3ffe:0501::1 prefix 64 alias (*1)
168 * # ifconfig if0 inet6 3ffe:0501::2 prefix 64 alias (*2)
169 * In this case, (*1) and (*2) want to add the same
170 * net route entry, 3ffe:0501:: -> if0.
171 * This case should not raise an error.
173 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED);
175 if (((rt2->rt_flags & (RTF_HOST|RTF_GATEWAY)) == 0)
177 && rt2->rt_gateway->sa_family == AF_LINK
178 && rt2->rt_ifp == rt->rt_ifp) {
188 * This code is the inverse of in6_clsroute: on first reference, if we
189 * were managing the route, stop doing so and set the expiration timer
192 static struct radix_node *
193 in6_matroute(void *v_arg, struct radix_node_head *head)
195 struct radix_node *rn = rn_match(v_arg, head);
196 struct rtentry *rt = (struct rtentry *)rn;
198 if (rt && rt->rt_refcnt == 0) { /* this is first reference */
199 if (rt->rt_flags & RTPRF_OURS) {
200 rt->rt_flags &= ~RTPRF_OURS;
201 rt->rt_rmx.rmx_expire = 0;
207 SYSCTL_DECL(_net_inet6_ip6);
209 #ifdef VIMAGE_GLOBALS
210 static int rtq_reallyold6;
211 static int rtq_minreallyold6;
212 static int rtq_toomany6;
215 SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_ip6, IPV6CTL_RTEXPIRE,
216 rtexpire, CTLFLAG_RW, rtq_reallyold6 , 0, "");
218 SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_ip6, IPV6CTL_RTMINEXPIRE,
219 rtminexpire, CTLFLAG_RW, rtq_minreallyold6 , 0, "");
221 SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_ip6, IPV6CTL_RTMAXCACHE,
222 rtmaxcache, CTLFLAG_RW, rtq_toomany6 , 0, "");
227 struct radix_node_head *rnh;
237 * Get rid of old routes. When draining, this deletes everything, even when
238 * the timeout is not expired yet. When updating, this makes sure that
239 * nothing has a timeout longer than the current value of rtq_reallyold6.
242 in6_rtqkill(struct radix_node *rn, void *rock)
244 INIT_VNET_INET6(curvnet);
245 struct rtqk_arg *ap = rock;
246 struct rtentry *rt = (struct rtentry *)rn;
249 RADIX_NODE_HEAD_WLOCK_ASSERT(ap->rnh);
251 if (rt->rt_flags & RTPRF_OURS) {
254 if (ap->draining || rt->rt_rmx.rmx_expire <= time_uptime) {
255 if (rt->rt_refcnt > 0)
256 panic("rtqkill route really not free");
258 err = rtrequest(RTM_DELETE,
259 (struct sockaddr *)rt_key(rt),
260 rt->rt_gateway, rt_mask(rt),
261 rt->rt_flags|RTF_RNH_LOCKED, 0);
263 log(LOG_WARNING, "in6_rtqkill: error %d", err);
269 && (rt->rt_rmx.rmx_expire - time_uptime
270 > V_rtq_reallyold6)) {
271 rt->rt_rmx.rmx_expire = time_uptime
274 ap->nextstop = lmin(ap->nextstop,
275 rt->rt_rmx.rmx_expire);
282 #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */
283 #ifdef VIMAGE_GLOBALS
284 static int rtq_timeout6;
285 static struct callout rtq_timer6;
289 in6_rtqtimo(void *rock)
291 CURVNET_SET_QUIET((struct vnet *) rock);
292 INIT_VNET_NET((struct vnet *) rock);
293 INIT_VNET_INET6((struct vnet *) rock);
294 struct radix_node_head *rnh = rock;
297 static time_t last_adjusted_timeout = 0;
299 arg.found = arg.killed = 0;
301 arg.nextstop = time_uptime + V_rtq_timeout6;
302 arg.draining = arg.updating = 0;
303 RADIX_NODE_HEAD_LOCK(rnh);
304 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
305 RADIX_NODE_HEAD_UNLOCK(rnh);
308 * Attempt to be somewhat dynamic about this:
309 * If there are ``too many'' routes sitting around taking up space,
310 * then crank down the timeout, and see if we can't make some more
311 * go away. However, we make sure that we will never adjust more
312 * than once in rtq_timeout6 seconds, to keep from cranking down too
315 if ((arg.found - arg.killed > V_rtq_toomany6)
316 && (time_uptime - last_adjusted_timeout >= V_rtq_timeout6)
317 && V_rtq_reallyold6 > V_rtq_minreallyold6) {
318 V_rtq_reallyold6 = 2*V_rtq_reallyold6 / 3;
319 if (V_rtq_reallyold6 < V_rtq_minreallyold6) {
320 V_rtq_reallyold6 = V_rtq_minreallyold6;
323 last_adjusted_timeout = time_uptime;
325 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold6 to %d",
328 arg.found = arg.killed = 0;
330 RADIX_NODE_HEAD_LOCK(rnh);
331 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
332 RADIX_NODE_HEAD_UNLOCK(rnh);
336 atv.tv_sec = arg.nextstop - time_uptime;
337 callout_reset(&V_rtq_timer6, tvtohz(&atv), in6_rtqtimo, rock);
345 struct radix_node_head *rnh;
348 #ifdef VIMAGE_GLOBALS
349 static struct callout rtq_mtutimer;
353 in6_mtuexpire(struct radix_node *rn, void *rock)
355 struct rtentry *rt = (struct rtentry *)rn;
356 struct mtuex_arg *ap = rock;
360 panic("rt == NULL in in6_mtuexpire");
362 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) {
363 if (rt->rt_rmx.rmx_expire <= time_uptime) {
364 rt->rt_flags |= RTF_PROBEMTU;
366 ap->nextstop = lmin(ap->nextstop,
367 rt->rt_rmx.rmx_expire);
374 #define MTUTIMO_DEFAULT (60*1)
377 in6_mtutimo(void *rock)
379 CURVNET_SET_QUIET((struct vnet *) rock);
380 INIT_VNET_NET((struct vnet *) rock);
381 INIT_VNET_INET6((struct vnet *) rock);
382 struct radix_node_head *rnh = rock;
383 struct mtuex_arg arg;
387 arg.nextstop = time_uptime + MTUTIMO_DEFAULT;
388 RADIX_NODE_HEAD_LOCK(rnh);
389 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg);
390 RADIX_NODE_HEAD_UNLOCK(rnh);
393 atv.tv_sec = arg.nextstop - time_uptime;
394 if (atv.tv_sec < 0) {
395 printf("invalid mtu expiration time on routing table\n");
396 arg.nextstop = time_uptime + 30; /* last resort */
399 callout_reset(&V_rtq_mtutimer, tvtohz(&atv), in6_mtutimo, rock);
407 INIT_VNET_NET(curvnet);
408 struct radix_node_head *rnh = V_rt_tables[AF_INET6];
411 arg.found = arg.killed = 0;
416 RADIX_NODE_HEAD_LOCK(rnh);
417 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
418 RADIX_NODE_HEAD_UNLOCK(rnh);
423 * Initialize our routing tree.
424 * XXX MRT When off == 0, we are being called from vfs_export.c
425 * so just set up their table and leave. (we know what the correct
426 * value should be so just use that).. FIX AFTER RELENG_7 is MFC'd
427 * see also comments in in_inithead() vfs_export.c and domain.h
430 in6_inithead(void **head, int off)
432 INIT_VNET_INET6(curvnet);
433 struct radix_node_head *rnh;
435 if (!rn_inithead(head, offsetof(struct sockaddr_in6, sin6_addr) << 3))
436 return 0; /* See above */
438 if (off == 0) /* See above */
439 return 1; /* only do the rest for the real thing */
441 V_rtq_reallyold6 = 60*60; /* one hour is ``really old'' */
442 V_rtq_minreallyold6 = 10; /* never automatically crank down to less */
443 V_rtq_toomany6 = 128; /* 128 cached routes is ``too many'' */
444 V_rtq_timeout6 = RTQ_TIMEOUT;
447 rnh->rnh_addaddr = in6_addroute;
448 rnh->rnh_matchaddr = in6_matroute;
449 callout_init(&V_rtq_timer6, CALLOUT_MPSAFE);
450 in6_rtqtimo(rnh); /* kick off timeout first time */
451 callout_init(&V_rtq_mtutimer, CALLOUT_MPSAFE);
452 in6_mtutimo(rnh); /* kick off timeout first time */