2 * Copyright 1994, 1995 Massachusetts Institute of Technology
4 * Permission to use, copy, modify, and distribute this software and
5 * its documentation for any purpose and without fee is hereby
6 * granted, provided that both the above copyright notice and this
7 * permission notice appear in all copies, that both the above
8 * copyright notice and this permission notice appear in all
9 * supporting documentation, and that the name of M.I.T. not be used
10 * in advertising or publicity pertaining to distribution of the
11 * software without specific, written prior permission. M.I.T. makes
12 * no representations about the suitability of this software for any
13 * purpose. It is provided "as is" without express or implied
16 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS
17 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
18 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
19 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
20 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
22 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
23 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
24 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
25 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
26 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/kernel.h>
36 #include <sys/sysctl.h>
37 #include <sys/socket.h>
39 #include <sys/syslog.h>
40 #include <sys/callout.h>
43 #include <net/route.h>
46 #include <netinet/in.h>
47 #include <netinet/in_var.h>
48 #include <netinet/ip.h>
49 #include <netinet/ip_icmp.h>
50 #include <netinet/ip_var.h>
52 extern int in_inithead(void **head, int off);
54 extern int in_detachhead(void **head, int off);
57 #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */
60 * Do what we need to do when inserting a route.
62 static struct radix_node *
63 in_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
64 struct radix_node *treenodes)
66 struct rtentry *rt = (struct rtentry *)treenodes;
67 struct sockaddr_in *sin = (struct sockaddr_in *)rt_key(rt);
69 RADIX_NODE_HEAD_WLOCK_ASSERT(head);
71 * A little bit of help for both IP output and input:
72 * For host routes, we make sure that RTF_BROADCAST
73 * is set for anything that looks like a broadcast address.
74 * This way, we can avoid an expensive call to in_broadcast()
75 * in ip_output() most of the time (because the route passed
76 * to ip_output() is almost always a host route).
78 * We also do the same for local addresses, with the thought
79 * that this might one day be used to speed up ip_input().
81 * We also mark routes to multicast addresses as such, because
82 * it's easy to do and might be useful (but this is much more
83 * dubious since it's so easy to inspect the address).
85 if (rt->rt_flags & RTF_HOST) {
86 if (in_broadcast(sin->sin_addr, rt->rt_ifp)) {
87 rt->rt_flags |= RTF_BROADCAST;
88 } else if (satosin(rt->rt_ifa->ifa_addr)->sin_addr.s_addr ==
89 sin->sin_addr.s_addr) {
90 rt->rt_flags |= RTF_LOCAL;
93 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
94 rt->rt_flags |= RTF_MULTICAST;
96 if (rt->rt_mtu == 0 && rt->rt_ifp != NULL)
97 rt->rt_mtu = rt->rt_ifp->if_mtu;
99 return (rn_addroute(v_arg, n_arg, head, treenodes));
103 * This code is the inverse of in_clsroute: on first reference, if we
104 * were managing the route, stop doing so and set the expiration timer
107 static struct radix_node *
108 in_matroute(void *v_arg, struct radix_node_head *head)
110 struct radix_node *rn = rn_match(v_arg, head);
111 struct rtentry *rt = (struct rtentry *)rn;
115 if (rt->rt_flags & RTPRF_OURS) {
116 rt->rt_flags &= ~RTPRF_OURS;
124 static VNET_DEFINE(int, rtq_reallyold) = 60*60; /* one hour is "really old" */
125 #define V_rtq_reallyold VNET(rtq_reallyold)
126 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_RTEXPIRE, rtexpire, CTLFLAG_RW,
127 &VNET_NAME(rtq_reallyold), 0,
128 "Default expiration time on dynamically learned routes");
130 /* never automatically crank down to less */
131 static VNET_DEFINE(int, rtq_minreallyold) = 10;
132 #define V_rtq_minreallyold VNET(rtq_minreallyold)
133 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_RTMINEXPIRE, rtminexpire, CTLFLAG_RW,
134 &VNET_NAME(rtq_minreallyold), 0,
135 "Minimum time to attempt to hold onto dynamically learned routes");
137 /* 128 cached routes is "too many" */
138 static VNET_DEFINE(int, rtq_toomany) = 128;
139 #define V_rtq_toomany VNET(rtq_toomany)
140 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_RTMAXCACHE, rtmaxcache, CTLFLAG_RW,
141 &VNET_NAME(rtq_toomany), 0,
142 "Upper limit on dynamically learned routes");
145 * On last reference drop, mark the route as belong to us so that it can be
149 in_clsroute(struct radix_node *rn, struct radix_node_head *head)
151 struct rtentry *rt = (struct rtentry *)rn;
155 if (!(rt->rt_flags & RTF_UP))
156 return; /* prophylactic measures */
158 if (rt->rt_flags & RTPRF_OURS)
161 if (!(rt->rt_flags & RTF_DYNAMIC))
165 * If rtq_reallyold is 0, just delete the route without
166 * waiting for a timeout cycle to kill it.
168 if (V_rtq_reallyold != 0) {
169 rt->rt_flags |= RTPRF_OURS;
170 rt->rt_expire = time_uptime + V_rtq_reallyold;
177 struct radix_node_head *rnh;
186 * Get rid of old routes. When draining, this deletes everything, even when
187 * the timeout is not expired yet. When updating, this makes sure that
188 * nothing has a timeout longer than the current value of rtq_reallyold.
191 in_rtqkill(struct radix_node *rn, void *rock)
193 struct rtqk_arg *ap = rock;
194 struct rtentry *rt = (struct rtentry *)rn;
197 RADIX_NODE_HEAD_WLOCK_ASSERT(ap->rnh);
199 if (rt->rt_flags & RTPRF_OURS) {
202 if (ap->draining || rt->rt_expire <= time_uptime) {
203 if (rt->rt_refcnt > 0)
204 panic("rtqkill route really not free");
206 err = in_rtrequest(RTM_DELETE,
207 (struct sockaddr *)rt_key(rt),
208 rt->rt_gateway, rt_mask(rt),
209 rt->rt_flags | RTF_RNH_LOCKED, 0,
212 log(LOG_WARNING, "in_rtqkill: error %d\n", err);
218 (rt->rt_expire - time_uptime > V_rtq_reallyold))
219 rt->rt_expire = time_uptime + V_rtq_reallyold;
220 ap->nextstop = lmin(ap->nextstop, rt->rt_expire);
227 #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */
228 static VNET_DEFINE(int, rtq_timeout) = RTQ_TIMEOUT;
229 static VNET_DEFINE(struct callout, rtq_timer);
231 #define V_rtq_timeout VNET(rtq_timeout)
232 #define V_rtq_timer VNET(rtq_timer)
234 static void in_rtqtimo_one(void *rock);
237 in_rtqtimo(void *rock)
239 CURVNET_SET((struct vnet *) rock);
244 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
245 newrock = rt_tables_get_rnh(fibnum, AF_INET);
247 in_rtqtimo_one(newrock);
250 atv.tv_sec = V_rtq_timeout;
251 callout_reset(&V_rtq_timer, tvtohz(&atv), in_rtqtimo, rock);
256 in_rtqtimo_one(void *rock)
258 struct radix_node_head *rnh = rock;
260 static time_t last_adjusted_timeout = 0;
262 arg.found = arg.killed = 0;
264 arg.nextstop = time_uptime + V_rtq_timeout;
265 arg.draining = arg.updating = 0;
266 RADIX_NODE_HEAD_LOCK(rnh);
267 rnh->rnh_walktree(rnh, in_rtqkill, &arg);
268 RADIX_NODE_HEAD_UNLOCK(rnh);
271 * Attempt to be somewhat dynamic about this:
272 * If there are ``too many'' routes sitting around taking up space,
273 * then crank down the timeout, and see if we can't make some more
274 * go away. However, we make sure that we will never adjust more
275 * than once in rtq_timeout seconds, to keep from cranking down too
278 if ((arg.found - arg.killed > V_rtq_toomany) &&
279 (time_uptime - last_adjusted_timeout >= V_rtq_timeout) &&
280 V_rtq_reallyold > V_rtq_minreallyold) {
281 V_rtq_reallyold = 2 * V_rtq_reallyold / 3;
282 if (V_rtq_reallyold < V_rtq_minreallyold) {
283 V_rtq_reallyold = V_rtq_minreallyold;
286 last_adjusted_timeout = time_uptime;
288 log(LOG_DEBUG, "in_rtqtimo: adjusted rtq_reallyold to %d\n",
291 arg.found = arg.killed = 0;
293 RADIX_NODE_HEAD_LOCK(rnh);
294 rnh->rnh_walktree(rnh, in_rtqkill, &arg);
295 RADIX_NODE_HEAD_UNLOCK(rnh);
303 VNET_ITERATOR_DECL(vnet_iter);
304 struct radix_node_head *rnh;
308 VNET_LIST_RLOCK_NOSLEEP();
309 VNET_FOREACH(vnet_iter) {
310 CURVNET_SET(vnet_iter);
312 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
313 rnh = rt_tables_get_rnh(fibnum, AF_INET);
314 arg.found = arg.killed = 0;
319 RADIX_NODE_HEAD_LOCK(rnh);
320 rnh->rnh_walktree(rnh, in_rtqkill, &arg);
321 RADIX_NODE_HEAD_UNLOCK(rnh);
325 VNET_LIST_RUNLOCK_NOSLEEP();
329 in_setmatchfunc(struct radix_node_head *rnh, int val)
332 rnh->rnh_matchaddr = (val != 0) ? rn_match : in_matroute;
335 static int _in_rt_was_here;
337 * Initialize our routing tree.
340 in_inithead(void **head, int off)
342 struct radix_node_head *rnh;
345 * This can be called from vfs_export.c too in which case 'off'
346 * will be 0. We know the correct value so just use that and
347 * return directly if it was 0.
348 * This is a hack that replaces an even worse hack on a bad hack
349 * on a bad design. After RELENG_7 this should be fixed but that
350 * will change the ABI, so for now do it this way.
352 if (!rn_inithead(head, 32))
355 if (off == 0) /* XXX MRT see above */
356 return 1; /* only do the rest for a real routing table */
359 rnh->rnh_addaddr = in_addroute;
360 in_setmatchfunc(rnh, V_drop_redirect);
361 rnh->rnh_close = in_clsroute;
362 if (_in_rt_was_here == 0 ) {
363 callout_init(&V_rtq_timer, CALLOUT_MPSAFE);
364 callout_reset(&V_rtq_timer, 1, in_rtqtimo, curvnet);
372 in_detachhead(void **head, int off)
375 callout_drain(&V_rtq_timer);
376 return (rn_detachhead(head));
381 * This zaps old routes when the interface goes down or interface
382 * address is deleted. In the latter case, it deletes static routes
383 * that point to this address. If we don't do this, we may end up
384 * using the old address in the future. The ones we always want to
385 * get rid of are things like ARP entries, since the user might down
386 * the interface, walk over to a completely different network, and
389 struct in_ifadown_arg {
395 in_ifadownkill(struct radix_node *rn, void *xap)
397 struct in_ifadown_arg *ap = xap;
398 struct rtentry *rt = (struct rtentry *)rn;
401 if (rt->rt_ifa == ap->ifa &&
402 (ap->del || !(rt->rt_flags & RTF_STATIC))) {
404 * Aquire a reference so that it can later be freed
405 * as the refcount would be 0 here in case of at least
410 * Disconnect it from the tree and permit protocols
415 * At this point it is an rttrash node, and in case
416 * the above is the only reference we must free it.
417 * If we do not noone will have a pointer and the
418 * rtentry will be leaked forever.
419 * In case someone else holds a reference, we are
420 * fine as we only decrement the refcount. In that
421 * case if the other entity calls RT_REMREF, we
422 * will still be leaking but at least we tried.
432 in_ifadown(struct ifaddr *ifa, int delete)
434 struct in_ifadown_arg arg;
435 struct radix_node_head *rnh;
438 if (ifa->ifa_addr->sa_family != AF_INET)
441 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
442 rnh = rt_tables_get_rnh(fibnum, AF_INET);
445 RADIX_NODE_HEAD_LOCK(rnh);
446 rnh->rnh_walktree(rnh, in_ifadownkill, &arg);
447 RADIX_NODE_HEAD_UNLOCK(rnh);
448 ifa->ifa_flags &= ~IFA_ROUTE; /* XXXlocking? */
454 * inet versions of rt functions. These have fib extensions and
455 * for now will just reference the _fib variants.
456 * eventually this order will be reversed,
459 in_rtalloc_ign(struct route *ro, u_long ignflags, u_int fibnum)
461 rtalloc_ign_fib(ro, ignflags, fibnum);
465 in_rtrequest( int req,
466 struct sockaddr *dst,
467 struct sockaddr *gateway,
468 struct sockaddr *netmask,
470 struct rtentry **ret_nrt,
473 return (rtrequest_fib(req, dst, gateway, netmask,
474 flags, ret_nrt, fibnum));
478 in_rtalloc1(struct sockaddr *dst, int report, u_long ignflags, u_int fibnum)
480 return (rtalloc1_fib(dst, report, ignflags, fibnum));
484 in_rtredirect(struct sockaddr *dst,
485 struct sockaddr *gateway,
486 struct sockaddr *netmask,
488 struct sockaddr *src,
491 rtredirect_fib(dst, gateway, netmask, flags, src, fibnum);
495 in_rtalloc(struct route *ro, u_int fibnum)
497 rtalloc_ign_fib(ro, 0UL, fibnum);
501 int in_rt_getifa(struct rt_addrinfo *, u_int fibnum);
502 int in_rtioctl(u_long, caddr_t, u_int);
503 int in_rtrequest1(int, struct rt_addrinfo *, struct rtentry **, u_int);