1 /* $KAME: radix_mpath.c,v 1.17 2004/11/08 10:29:39 itojun Exp $ */
4 * SPDX-License-Identifier: BSD-3-Clause
6 * Copyright (C) 2001 WIDE Project.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * THE AUTHORS DO NOT GUARANTEE THAT THIS SOFTWARE DOES NOT INFRINGE
33 * ANY OTHERS' INTELLECTUAL PROPERTIES. IN NO EVENT SHALL THE AUTHORS
34 * BE LIABLE FOR ANY INFRINGEMENT OF ANY OTHERS' INTELLECTUAL
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
42 #include "opt_inet6.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/malloc.h>
47 #include <sys/socket.h>
48 #include <sys/domain.h>
49 #include <sys/syslog.h>
50 #include <net/radix.h>
51 #include <net/radix_mpath.h>
52 #include <net/route.h>
53 #include <net/route_var.h>
55 #include <net/if_var.h>
58 * give some jitter to hash, to avoid synchronization between routers
60 static uint32_t hashjitter;
63 rt_mpath_capable(struct rib_head *rnh)
66 return rnh->rnh_multipath;
70 rn_mpath_capable(struct radix_head *rh)
73 return (rt_mpath_capable((struct rib_head *)rh));
77 rn_mpath_next(struct radix_node *rn)
79 struct radix_node *next;
83 next = rn->rn_dupedkey;
84 if (rn->rn_mask == next->rn_mask)
91 rn_mpath_count(struct radix_node *rn)
97 rt = (struct rtentry *)rn;
99 rn = rn_mpath_next(rn);
105 rt_mpath_matchgate(struct rtentry *rt, struct sockaddr *gate)
107 struct radix_node *rn;
109 if (!gate || !rt->rt_gateway)
112 /* beyond here, we use rn as the master copy */
113 rn = (struct radix_node *)rt;
115 rt = (struct rtentry *)rn;
117 * we are removing an address alias that has
118 * the same prefix as another address
119 * we need to compare the interface address because
120 * rt_gateway is a special sockadd_dl structure
122 if (rt->rt_gateway->sa_family == AF_LINK) {
123 if (!memcmp(rt->rt_ifa->ifa_addr, gate, gate->sa_len))
128 * Check for other options:
129 * 1) Routes with 'real' IPv4/IPv6 gateway
130 * 2) Loopback host routes (another AF_LINK/sockadd_dl check)
132 if (rt->rt_gateway->sa_len == gate->sa_len &&
133 !memcmp(rt->rt_gateway, gate, gate->sa_len))
135 } while ((rn = rn_mpath_next(rn)) != NULL);
137 return (struct rtentry *)rn;
141 * go through the chain and unlink "rt" from the list
142 * the caller will free "rt"
145 rt_mpath_deldup(struct rtentry *headrt, struct rtentry *rt)
147 struct radix_node *t, *tt;
151 t = (struct radix_node *)headrt;
152 tt = rn_mpath_next(t);
154 if (tt == (struct radix_node *)rt) {
155 t->rn_dupedkey = tt->rn_dupedkey;
156 tt->rn_dupedkey = NULL;
157 tt->rn_flags &= ~RNF_ACTIVE;
158 tt[1].rn_flags &= ~RNF_ACTIVE;
162 tt = rn_mpath_next((struct radix_node *)t);
168 * check if we have the same key/mask/gateway on the table already.
169 * Assume @rt rt_key host bits are cleared according to @netmask
172 rt_mpath_conflict(struct rib_head *rnh, struct rtentry *rt,
173 struct sockaddr *netmask)
175 struct radix_node *rn, *rn1;
178 rn = (struct radix_node *)rt;
179 rn1 = rnh->rnh_lookup(rt_key(rt), netmask, &rnh->head);
180 if (!rn1 || rn1->rn_flags & RNF_ROOT)
183 /* key/mask are the same. compare gateway for all multipaths */
185 rt1 = (struct rtentry *)rn1;
187 /* sanity: no use in comparing the same thing */
191 if (rt1->rt_gateway->sa_family == AF_LINK) {
192 if (rt1->rt_ifa->ifa_addr->sa_len != rt->rt_ifa->ifa_addr->sa_len ||
193 bcmp(rt1->rt_ifa->ifa_addr, rt->rt_ifa->ifa_addr,
194 rt1->rt_ifa->ifa_addr->sa_len))
197 if (rt1->rt_gateway->sa_len != rt->rt_gateway->sa_len ||
198 bcmp(rt1->rt_gateway, rt->rt_gateway,
199 rt1->rt_gateway->sa_len))
203 /* all key/mask/gateway are the same. conflicting entry. */
205 } while ((rn1 = rn_mpath_next(rn1)) != NULL);
210 static struct rtentry *
211 rt_mpath_selectrte(struct rtentry *rte, uint32_t hash)
213 struct radix_node *rn0, *rn;
214 uint32_t total_weight;
218 /* beyond here, we use rn as the master copy */
219 rn0 = rn = (struct radix_node *)rte;
222 /* gw selection by Modulo-N Hash (RFC2991) XXX need improvement? */
223 total_weight = rn_mpath_count(rn0);
225 hash %= total_weight;
226 for (weight = abs((int32_t)hash);
227 rt != NULL && weight >= rt->rt_weight;
228 weight -= (rt == NULL) ? 0 : rt->rt_weight) {
230 /* stay within the multipath routes */
231 if (rn->rn_dupedkey && rn->rn_mask != rn->rn_dupedkey->rn_mask)
233 rn = rn->rn_dupedkey;
234 rt = (struct rtentry *)rn;
241 rt_mpath_select(struct rtentry *rte, uint32_t hash)
243 if (rn_mpath_next((struct radix_node *)rte) == NULL)
246 return (rt_mpath_selectrte(rte, hash));
250 rtalloc_mpath_fib(struct route *ro, uint32_t hash, u_int fibnum)
255 * XXX we don't attempt to lookup cached route again; what should
256 * be done for sendto(3) case?
258 if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP)
259 && RT_LINK_IS_UP(ro->ro_rt->rt_ifp))
261 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, 0, fibnum);
263 /* if the route does not exist or it is not multipath, don't care */
264 if (ro->ro_rt == NULL)
266 if (rn_mpath_next((struct radix_node *)ro->ro_rt) == NULL) {
267 RT_UNLOCK(ro->ro_rt);
271 rt = rt_mpath_selectrte(ro->ro_rt, hash);
272 /* XXX try filling rt_gwroute and avoid unreachable gw */
274 /* gw selection has failed - there must be only zero weight routes */
276 RT_UNLOCK(ro->ro_rt);
280 if (ro->ro_rt != rt) {
281 RTFREE_LOCKED(ro->ro_rt);
284 RT_ADDREF(ro->ro_rt);
287 RT_UNLOCK(ro->ro_rt);
290 extern int in6_inithead(void **head, int off);
291 extern int in_inithead(void **head, int off);
295 rn4_mpath_inithead(void **head, int off)
297 struct rib_head *rnh;
299 hashjitter = arc4random();
300 if (in_inithead(head, off) == 1) {
301 rnh = (struct rib_head *)*head;
302 rnh->rnh_multipath = 1;
311 rn6_mpath_inithead(void **head, int off)
313 struct rib_head *rnh;
315 hashjitter = arc4random();
316 if (in6_inithead(head, off) == 1) {
317 rnh = (struct rib_head *)*head;
318 rnh->rnh_multipath = 1;