1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
16 *
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
27 * SUCH DAMAGE.
28 *
29 * $KAME: in6_rmx.c,v 1.11 2001/07/26 06:53:16 jinmei Exp $
30 */
31
32/*-
33 * Copyright 1994, 1995 Massachusetts Institute of Technology
34 *
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
45 * warranty.
46 *
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
58 * SUCH DAMAGE.
59 *
60 */
61
62/*
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
68 * requested.
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.
73 */
74
75#include <sys/cdefs.h>
| 1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
16 *
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
27 * SUCH DAMAGE.
28 *
29 * $KAME: in6_rmx.c,v 1.11 2001/07/26 06:53:16 jinmei Exp $
30 */
31
32/*-
33 * Copyright 1994, 1995 Massachusetts Institute of Technology
34 *
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
45 * warranty.
46 *
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
58 * SUCH DAMAGE.
59 *
60 */
61
62/*
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
68 * requested.
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.
73 */
74
75#include <sys/cdefs.h>
|
76__FBSDID("$FreeBSD: head/sys/netinet6/in6_rmx.c 186119 2008-12-15 06:10:57Z qingli $");
| 76__FBSDID("$FreeBSD: head/sys/netinet6/in6_rmx.c 186215 2008-12-17 10:03:49Z qingli $");
|
77
78#include <sys/param.h>
79#include <sys/systm.h>
80#include <sys/kernel.h>
81#include <sys/lock.h>
82#include <sys/sysctl.h>
83#include <sys/queue.h>
84#include <sys/socket.h>
85#include <sys/socketvar.h>
86#include <sys/mbuf.h>
87#include <sys/rwlock.h>
88#include <sys/syslog.h>
89#include <sys/callout.h>
90#include <sys/vimage.h>
91
92#include <net/if.h>
93#include <net/route.h>
94#include <net/vnet.h>
95
96#include <netinet/in.h>
97#include <netinet/ip_var.h>
98#include <netinet/in_var.h>
99
100#include <netinet/ip6.h>
101#include <netinet6/ip6_var.h>
102
103#include <netinet/icmp6.h>
104#include <netinet6/nd6.h>
105#include <netinet6/vinet6.h>
106
107#include <netinet/tcp.h>
108#include <netinet/tcp_seq.h>
109#include <netinet/tcp_timer.h>
110#include <netinet/tcp_var.h>
111
112extern int in6_inithead(void **head, int off);
113
114#define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */
115
116/*
117 * Do what we need to do when inserting a route.
118 */
119static struct radix_node *
120in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
121 struct radix_node *treenodes)
122{
123 struct rtentry *rt = (struct rtentry *)treenodes;
124 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt);
125 struct radix_node *ret;
126
127 RADIX_NODE_HEAD_WLOCK_ASSERT(head);
128 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
129 rt->rt_flags |= RTF_MULTICAST;
130
131 /*
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
135 * ip_input().
136 *
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
140 * is done above.)
141 *
142 * XXX
143 * should elaborate the code.
144 */
145 if (rt->rt_flags & RTF_HOST) {
146 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr)
147 ->sin6_addr,
148 &sin6->sin6_addr)) {
149 rt->rt_flags |= RTF_LOCAL;
150 }
151 }
152
153 if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp)
154 rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp);
155
156 ret = rn_addroute(v_arg, n_arg, head, treenodes);
157 if (ret == NULL) {
158 struct rtentry *rt2;
159 /*
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.
170 */
171 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED);
172 if (rt2) {
173 if (((rt2->rt_flags & (RTF_HOST|RTF_GATEWAY)) == 0)
174 && rt2->rt_gateway
175 && rt2->rt_gateway->sa_family == AF_LINK
176 && rt2->rt_ifp == rt->rt_ifp) {
177 ret = rt2->rt_nodes;
178 }
179 RTFREE_LOCKED(rt2);
180 }
181 }
182 return (ret);
183}
184
185/*
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
188 * back off again.
189 */
190static struct radix_node *
191in6_matroute(void *v_arg, struct radix_node_head *head)
192{
193 struct radix_node *rn = rn_match(v_arg, head);
194 struct rtentry *rt = (struct rtentry *)rn;
195
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;
200 }
201 }
202 return rn;
203}
204
205SYSCTL_DECL(_net_inet6_ip6);
206
207#ifdef VIMAGE_GLOBALS
208static int rtq_reallyold6;
209static int rtq_minreallyold6;
210static int rtq_toomany6;
211#endif
212
213SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_ip6, IPV6CTL_RTEXPIRE,
214 rtexpire, CTLFLAG_RW, rtq_reallyold6 , 0, "");
215
216SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_ip6, IPV6CTL_RTMINEXPIRE,
217 rtminexpire, CTLFLAG_RW, rtq_minreallyold6 , 0, "");
218
219SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_ip6, IPV6CTL_RTMAXCACHE,
220 rtmaxcache, CTLFLAG_RW, rtq_toomany6 , 0, "");
221
222
| 77
78#include <sys/param.h>
79#include <sys/systm.h>
80#include <sys/kernel.h>
81#include <sys/lock.h>
82#include <sys/sysctl.h>
83#include <sys/queue.h>
84#include <sys/socket.h>
85#include <sys/socketvar.h>
86#include <sys/mbuf.h>
87#include <sys/rwlock.h>
88#include <sys/syslog.h>
89#include <sys/callout.h>
90#include <sys/vimage.h>
91
92#include <net/if.h>
93#include <net/route.h>
94#include <net/vnet.h>
95
96#include <netinet/in.h>
97#include <netinet/ip_var.h>
98#include <netinet/in_var.h>
99
100#include <netinet/ip6.h>
101#include <netinet6/ip6_var.h>
102
103#include <netinet/icmp6.h>
104#include <netinet6/nd6.h>
105#include <netinet6/vinet6.h>
106
107#include <netinet/tcp.h>
108#include <netinet/tcp_seq.h>
109#include <netinet/tcp_timer.h>
110#include <netinet/tcp_var.h>
111
112extern int in6_inithead(void **head, int off);
113
114#define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */
115
116/*
117 * Do what we need to do when inserting a route.
118 */
119static struct radix_node *
120in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
121 struct radix_node *treenodes)
122{
123 struct rtentry *rt = (struct rtentry *)treenodes;
124 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt);
125 struct radix_node *ret;
126
127 RADIX_NODE_HEAD_WLOCK_ASSERT(head);
128 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
129 rt->rt_flags |= RTF_MULTICAST;
130
131 /*
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
135 * ip_input().
136 *
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
140 * is done above.)
141 *
142 * XXX
143 * should elaborate the code.
144 */
145 if (rt->rt_flags & RTF_HOST) {
146 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr)
147 ->sin6_addr,
148 &sin6->sin6_addr)) {
149 rt->rt_flags |= RTF_LOCAL;
150 }
151 }
152
153 if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp)
154 rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp);
155
156 ret = rn_addroute(v_arg, n_arg, head, treenodes);
157 if (ret == NULL) {
158 struct rtentry *rt2;
159 /*
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.
170 */
171 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED);
172 if (rt2) {
173 if (((rt2->rt_flags & (RTF_HOST|RTF_GATEWAY)) == 0)
174 && rt2->rt_gateway
175 && rt2->rt_gateway->sa_family == AF_LINK
176 && rt2->rt_ifp == rt->rt_ifp) {
177 ret = rt2->rt_nodes;
178 }
179 RTFREE_LOCKED(rt2);
180 }
181 }
182 return (ret);
183}
184
185/*
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
188 * back off again.
189 */
190static struct radix_node *
191in6_matroute(void *v_arg, struct radix_node_head *head)
192{
193 struct radix_node *rn = rn_match(v_arg, head);
194 struct rtentry *rt = (struct rtentry *)rn;
195
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;
200 }
201 }
202 return rn;
203}
204
205SYSCTL_DECL(_net_inet6_ip6);
206
207#ifdef VIMAGE_GLOBALS
208static int rtq_reallyold6;
209static int rtq_minreallyold6;
210static int rtq_toomany6;
211#endif
212
213SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_ip6, IPV6CTL_RTEXPIRE,
214 rtexpire, CTLFLAG_RW, rtq_reallyold6 , 0, "");
215
216SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_ip6, IPV6CTL_RTMINEXPIRE,
217 rtminexpire, CTLFLAG_RW, rtq_minreallyold6 , 0, "");
218
219SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_ip6, IPV6CTL_RTMAXCACHE,
220 rtmaxcache, CTLFLAG_RW, rtq_toomany6 , 0, "");
221
222
|
223/*
224 * On last reference drop, mark the route as belong to us so that it can be
225 * timed out.
226 */
227static void
228in6_clsroute(struct radix_node *rn, struct radix_node_head *head)
229{
230 INIT_VNET_INET6(curvnet);
231 struct rtentry *rt = (struct rtentry *)rn;
| |
232
| 223
|
233 RT_LOCK_ASSERT(rt);
234
235 if (!(rt->rt_flags & RTF_UP))
236 return; /* prophylactic measures */
237
238 /*
239 * As requested by David Greenman:
240 * If rtq_reallyold6 is 0, just delete the route without
241 * waiting for a timeout cycle to kill it.
242 */
243 if (V_rtq_reallyold6 != 0) {
244 rt->rt_flags |= RTPRF_OURS;
245 rt->rt_rmx.rmx_expire = time_uptime + V_rtq_reallyold6;
246 } else {
247 rtexpunge(rt);
248 }
249}
250
| |
251struct rtqk_arg {
252 struct radix_node_head *rnh;
253 int mode;
254 int updating;
255 int draining;
256 int killed;
257 int found;
258 time_t nextstop;
259};
260
261/*
262 * Get rid of old routes. When draining, this deletes everything, even when
263 * the timeout is not expired yet. When updating, this makes sure that
264 * nothing has a timeout longer than the current value of rtq_reallyold6.
265 */
266static int
267in6_rtqkill(struct radix_node *rn, void *rock)
268{
269 INIT_VNET_INET6(curvnet);
270 struct rtqk_arg *ap = rock;
271 struct rtentry *rt = (struct rtentry *)rn;
272 int err;
273
274 if (rt->rt_flags & RTPRF_OURS) {
275 ap->found++;
276
277 if (ap->draining || rt->rt_rmx.rmx_expire <= time_uptime) {
278 if (rt->rt_refcnt > 0)
279 panic("rtqkill route really not free");
280
281 err = rtrequest(RTM_DELETE,
282 (struct sockaddr *)rt_key(rt),
283 rt->rt_gateway, rt_mask(rt),
284 rt->rt_flags|RTF_RNH_LOCKED, 0);
285 if (err) {
286 log(LOG_WARNING, "in6_rtqkill: error %d", err);
287 } else {
288 ap->killed++;
289 }
290 } else {
291 if (ap->updating
292 && (rt->rt_rmx.rmx_expire - time_uptime
293 > V_rtq_reallyold6)) {
294 rt->rt_rmx.rmx_expire = time_uptime
295 + V_rtq_reallyold6;
296 }
297 ap->nextstop = lmin(ap->nextstop,
298 rt->rt_rmx.rmx_expire);
299 }
300 }
301
302 return 0;
303}
304
305#define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */
306#ifdef VIMAGE_GLOBALS
307static int rtq_timeout6;
308static struct callout rtq_timer6;
309#endif
310
311static void
312in6_rtqtimo(void *rock)
313{
314 CURVNET_SET_QUIET((struct vnet *) rock);
315 INIT_VNET_NET((struct vnet *) rock);
316 INIT_VNET_INET6((struct vnet *) rock);
317 struct radix_node_head *rnh = rock;
318 struct rtqk_arg arg;
319 struct timeval atv;
320 static time_t last_adjusted_timeout = 0;
321
322 arg.found = arg.killed = 0;
323 arg.rnh = rnh;
324 arg.nextstop = time_uptime + V_rtq_timeout6;
325 arg.draining = arg.updating = 0;
326 RADIX_NODE_HEAD_LOCK(rnh);
327 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
328 RADIX_NODE_HEAD_UNLOCK(rnh);
329
330 /*
331 * Attempt to be somewhat dynamic about this:
332 * If there are ``too many'' routes sitting around taking up space,
333 * then crank down the timeout, and see if we can't make some more
334 * go away. However, we make sure that we will never adjust more
335 * than once in rtq_timeout6 seconds, to keep from cranking down too
336 * hard.
337 */
338 if ((arg.found - arg.killed > V_rtq_toomany6)
339 && (time_uptime - last_adjusted_timeout >= V_rtq_timeout6)
340 && V_rtq_reallyold6 > V_rtq_minreallyold6) {
341 V_rtq_reallyold6 = 2*V_rtq_reallyold6 / 3;
342 if (V_rtq_reallyold6 < V_rtq_minreallyold6) {
343 V_rtq_reallyold6 = V_rtq_minreallyold6;
344 }
345
346 last_adjusted_timeout = time_uptime;
347#ifdef DIAGNOSTIC
348 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold6 to %d",
349 V_rtq_reallyold6);
350#endif
351 arg.found = arg.killed = 0;
352 arg.updating = 1;
353 RADIX_NODE_HEAD_LOCK(rnh);
354 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
355 RADIX_NODE_HEAD_UNLOCK(rnh);
356 }
357
358 atv.tv_usec = 0;
359 atv.tv_sec = arg.nextstop - time_uptime;
360 callout_reset(&V_rtq_timer6, tvtohz(&atv), in6_rtqtimo, rock);
361 CURVNET_RESTORE();
362}
363
364/*
365 * Age old PMTUs.
366 */
367struct mtuex_arg {
368 struct radix_node_head *rnh;
369 time_t nextstop;
370};
371#ifdef VIMAGE_GLOBALS
372static struct callout rtq_mtutimer;
373#endif
374
375static int
376in6_mtuexpire(struct radix_node *rn, void *rock)
377{
378 struct rtentry *rt = (struct rtentry *)rn;
379 struct mtuex_arg *ap = rock;
380
381 /* sanity */
382 if (!rt)
383 panic("rt == NULL in in6_mtuexpire");
384
385 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) {
386 if (rt->rt_rmx.rmx_expire <= time_uptime) {
387 rt->rt_flags |= RTF_PROBEMTU;
388 } else {
389 ap->nextstop = lmin(ap->nextstop,
390 rt->rt_rmx.rmx_expire);
391 }
392 }
393
394 return 0;
395}
396
397#define MTUTIMO_DEFAULT (60*1)
398
399static void
400in6_mtutimo(void *rock)
401{
402 CURVNET_SET_QUIET((struct vnet *) rock);
403 INIT_VNET_NET((struct vnet *) rock);
404 INIT_VNET_INET6((struct vnet *) rock);
405 struct radix_node_head *rnh = rock;
406 struct mtuex_arg arg;
407 struct timeval atv;
408
409 arg.rnh = rnh;
410 arg.nextstop = time_uptime + MTUTIMO_DEFAULT;
411 RADIX_NODE_HEAD_LOCK(rnh);
412 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg);
413 RADIX_NODE_HEAD_UNLOCK(rnh);
414
415 atv.tv_usec = 0;
416 atv.tv_sec = arg.nextstop - time_uptime;
417 if (atv.tv_sec < 0) {
418 printf("invalid mtu expiration time on routing table\n");
419 arg.nextstop = time_uptime + 30; /* last resort */
420 atv.tv_sec = 30;
421 }
422 callout_reset(&V_rtq_mtutimer, tvtohz(&atv), in6_mtutimo, rock);
423 CURVNET_RESTORE();
424}
425
426#if 0
427void
428in6_rtqdrain(void)
429{
430 INIT_VNET_NET(curvnet);
431 struct radix_node_head *rnh = V_rt_tables[AF_INET6];
432 struct rtqk_arg arg;
433
434 arg.found = arg.killed = 0;
435 arg.rnh = rnh;
436 arg.nextstop = 0;
437 arg.draining = 1;
438 arg.updating = 0;
439 RADIX_NODE_HEAD_LOCK(rnh);
440 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
441 RADIX_NODE_HEAD_UNLOCK(rnh);
442}
443#endif
444
445/*
446 * Initialize our routing tree.
447 * XXX MRT When off == 0, we are being called from vfs_export.c
448 * so just set up their table and leave. (we know what the correct
449 * value should be so just use that).. FIX AFTER RELENG_7 is MFC'd
450 * see also comments in in_inithead() vfs_export.c and domain.h
451 */
452int
453in6_inithead(void **head, int off)
454{
455 INIT_VNET_INET6(curvnet);
456 struct radix_node_head *rnh;
457
458 if (!rn_inithead(head, offsetof(struct sockaddr_in6, sin6_addr) << 3))
459 return 0; /* See above */
460
461 if (off == 0) /* See above */
462 return 1; /* only do the rest for the real thing */
463
464 V_rtq_reallyold6 = 60*60; /* one hour is ``really old'' */
465 V_rtq_minreallyold6 = 10; /* never automatically crank down to less */
466 V_rtq_toomany6 = 128; /* 128 cached routes is ``too many'' */
467 V_rtq_timeout6 = RTQ_TIMEOUT;
468
469 rnh = *head;
470 rnh->rnh_addaddr = in6_addroute;
471 rnh->rnh_matchaddr = in6_matroute;
| 224struct rtqk_arg {
225 struct radix_node_head *rnh;
226 int mode;
227 int updating;
228 int draining;
229 int killed;
230 int found;
231 time_t nextstop;
232};
233
234/*
235 * Get rid of old routes. When draining, this deletes everything, even when
236 * the timeout is not expired yet. When updating, this makes sure that
237 * nothing has a timeout longer than the current value of rtq_reallyold6.
238 */
239static int
240in6_rtqkill(struct radix_node *rn, void *rock)
241{
242 INIT_VNET_INET6(curvnet);
243 struct rtqk_arg *ap = rock;
244 struct rtentry *rt = (struct rtentry *)rn;
245 int err;
246
247 if (rt->rt_flags & RTPRF_OURS) {
248 ap->found++;
249
250 if (ap->draining || rt->rt_rmx.rmx_expire <= time_uptime) {
251 if (rt->rt_refcnt > 0)
252 panic("rtqkill route really not free");
253
254 err = rtrequest(RTM_DELETE,
255 (struct sockaddr *)rt_key(rt),
256 rt->rt_gateway, rt_mask(rt),
257 rt->rt_flags|RTF_RNH_LOCKED, 0);
258 if (err) {
259 log(LOG_WARNING, "in6_rtqkill: error %d", err);
260 } else {
261 ap->killed++;
262 }
263 } else {
264 if (ap->updating
265 && (rt->rt_rmx.rmx_expire - time_uptime
266 > V_rtq_reallyold6)) {
267 rt->rt_rmx.rmx_expire = time_uptime
268 + V_rtq_reallyold6;
269 }
270 ap->nextstop = lmin(ap->nextstop,
271 rt->rt_rmx.rmx_expire);
272 }
273 }
274
275 return 0;
276}
277
278#define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */
279#ifdef VIMAGE_GLOBALS
280static int rtq_timeout6;
281static struct callout rtq_timer6;
282#endif
283
284static void
285in6_rtqtimo(void *rock)
286{
287 CURVNET_SET_QUIET((struct vnet *) rock);
288 INIT_VNET_NET((struct vnet *) rock);
289 INIT_VNET_INET6((struct vnet *) rock);
290 struct radix_node_head *rnh = rock;
291 struct rtqk_arg arg;
292 struct timeval atv;
293 static time_t last_adjusted_timeout = 0;
294
295 arg.found = arg.killed = 0;
296 arg.rnh = rnh;
297 arg.nextstop = time_uptime + V_rtq_timeout6;
298 arg.draining = arg.updating = 0;
299 RADIX_NODE_HEAD_LOCK(rnh);
300 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
301 RADIX_NODE_HEAD_UNLOCK(rnh);
302
303 /*
304 * Attempt to be somewhat dynamic about this:
305 * If there are ``too many'' routes sitting around taking up space,
306 * then crank down the timeout, and see if we can't make some more
307 * go away. However, we make sure that we will never adjust more
308 * than once in rtq_timeout6 seconds, to keep from cranking down too
309 * hard.
310 */
311 if ((arg.found - arg.killed > V_rtq_toomany6)
312 && (time_uptime - last_adjusted_timeout >= V_rtq_timeout6)
313 && V_rtq_reallyold6 > V_rtq_minreallyold6) {
314 V_rtq_reallyold6 = 2*V_rtq_reallyold6 / 3;
315 if (V_rtq_reallyold6 < V_rtq_minreallyold6) {
316 V_rtq_reallyold6 = V_rtq_minreallyold6;
317 }
318
319 last_adjusted_timeout = time_uptime;
320#ifdef DIAGNOSTIC
321 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold6 to %d",
322 V_rtq_reallyold6);
323#endif
324 arg.found = arg.killed = 0;
325 arg.updating = 1;
326 RADIX_NODE_HEAD_LOCK(rnh);
327 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
328 RADIX_NODE_HEAD_UNLOCK(rnh);
329 }
330
331 atv.tv_usec = 0;
332 atv.tv_sec = arg.nextstop - time_uptime;
333 callout_reset(&V_rtq_timer6, tvtohz(&atv), in6_rtqtimo, rock);
334 CURVNET_RESTORE();
335}
336
337/*
338 * Age old PMTUs.
339 */
340struct mtuex_arg {
341 struct radix_node_head *rnh;
342 time_t nextstop;
343};
344#ifdef VIMAGE_GLOBALS
345static struct callout rtq_mtutimer;
346#endif
347
348static int
349in6_mtuexpire(struct radix_node *rn, void *rock)
350{
351 struct rtentry *rt = (struct rtentry *)rn;
352 struct mtuex_arg *ap = rock;
353
354 /* sanity */
355 if (!rt)
356 panic("rt == NULL in in6_mtuexpire");
357
358 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) {
359 if (rt->rt_rmx.rmx_expire <= time_uptime) {
360 rt->rt_flags |= RTF_PROBEMTU;
361 } else {
362 ap->nextstop = lmin(ap->nextstop,
363 rt->rt_rmx.rmx_expire);
364 }
365 }
366
367 return 0;
368}
369
370#define MTUTIMO_DEFAULT (60*1)
371
372static void
373in6_mtutimo(void *rock)
374{
375 CURVNET_SET_QUIET((struct vnet *) rock);
376 INIT_VNET_NET((struct vnet *) rock);
377 INIT_VNET_INET6((struct vnet *) rock);
378 struct radix_node_head *rnh = rock;
379 struct mtuex_arg arg;
380 struct timeval atv;
381
382 arg.rnh = rnh;
383 arg.nextstop = time_uptime + MTUTIMO_DEFAULT;
384 RADIX_NODE_HEAD_LOCK(rnh);
385 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg);
386 RADIX_NODE_HEAD_UNLOCK(rnh);
387
388 atv.tv_usec = 0;
389 atv.tv_sec = arg.nextstop - time_uptime;
390 if (atv.tv_sec < 0) {
391 printf("invalid mtu expiration time on routing table\n");
392 arg.nextstop = time_uptime + 30; /* last resort */
393 atv.tv_sec = 30;
394 }
395 callout_reset(&V_rtq_mtutimer, tvtohz(&atv), in6_mtutimo, rock);
396 CURVNET_RESTORE();
397}
398
399#if 0
400void
401in6_rtqdrain(void)
402{
403 INIT_VNET_NET(curvnet);
404 struct radix_node_head *rnh = V_rt_tables[AF_INET6];
405 struct rtqk_arg arg;
406
407 arg.found = arg.killed = 0;
408 arg.rnh = rnh;
409 arg.nextstop = 0;
410 arg.draining = 1;
411 arg.updating = 0;
412 RADIX_NODE_HEAD_LOCK(rnh);
413 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
414 RADIX_NODE_HEAD_UNLOCK(rnh);
415}
416#endif
417
418/*
419 * Initialize our routing tree.
420 * XXX MRT When off == 0, we are being called from vfs_export.c
421 * so just set up their table and leave. (we know what the correct
422 * value should be so just use that).. FIX AFTER RELENG_7 is MFC'd
423 * see also comments in in_inithead() vfs_export.c and domain.h
424 */
425int
426in6_inithead(void **head, int off)
427{
428 INIT_VNET_INET6(curvnet);
429 struct radix_node_head *rnh;
430
431 if (!rn_inithead(head, offsetof(struct sockaddr_in6, sin6_addr) << 3))
432 return 0; /* See above */
433
434 if (off == 0) /* See above */
435 return 1; /* only do the rest for the real thing */
436
437 V_rtq_reallyold6 = 60*60; /* one hour is ``really old'' */
438 V_rtq_minreallyold6 = 10; /* never automatically crank down to less */
439 V_rtq_toomany6 = 128; /* 128 cached routes is ``too many'' */
440 V_rtq_timeout6 = RTQ_TIMEOUT;
441
442 rnh = *head;
443 rnh->rnh_addaddr = in6_addroute;
444 rnh->rnh_matchaddr = in6_matroute;
|
472 rnh->rnh_close = in6_clsroute;
| |
473 callout_init(&V_rtq_timer6, CALLOUT_MPSAFE);
474 in6_rtqtimo(rnh); /* kick off timeout first time */
475 callout_init(&V_rtq_mtutimer, CALLOUT_MPSAFE);
476 in6_mtutimo(rnh); /* kick off timeout first time */
477 return 1;
478}
| 445 callout_init(&V_rtq_timer6, CALLOUT_MPSAFE);
446 in6_rtqtimo(rnh); /* kick off timeout first time */
447 callout_init(&V_rtq_mtutimer, CALLOUT_MPSAFE);
448 in6_mtutimo(rnh); /* kick off timeout first time */
449 return 1;
450}
|