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>
47 #include <sys/malloc.h>
48 #include <sys/mutex.h>
49 #include <sys/socket.h>
50 #include <sys/domain.h>
51 #include <sys/kernel.h>
52 #include <sys/syslog.h>
53 #include <net/radix.h>
54 #include <net/radix_mpath.h>
55 #include <sys/rmlock.h>
56 #include <net/route.h>
57 #include <net/route_var.h>
59 #include <net/if_var.h>
62 * give some jitter to hash, to avoid synchronization between routers
64 static uint32_t hashjitter;
67 rt_mpath_capable(struct rib_head *rnh)
70 return rnh->rnh_multipath;
74 rn_mpath_capable(struct radix_head *rh)
77 return (rt_mpath_capable((struct rib_head *)rh));
81 rn_mpath_next(struct radix_node *rn)
83 struct radix_node *next;
87 next = rn->rn_dupedkey;
88 if (rn->rn_mask == next->rn_mask)
95 rn_mpath_count(struct radix_node *rn)
101 rt = (struct rtentry *)rn;
103 rn = rn_mpath_next(rn);
109 rt_mpath_matchgate(struct rtentry *rt, struct sockaddr *gate)
111 struct radix_node *rn;
113 if (!gate || !rt->rt_gateway)
116 /* beyond here, we use rn as the master copy */
117 rn = (struct radix_node *)rt;
119 rt = (struct rtentry *)rn;
121 * we are removing an address alias that has
122 * the same prefix as another address
123 * we need to compare the interface address because
124 * rt_gateway is a special sockadd_dl structure
126 if (rt->rt_gateway->sa_family == AF_LINK) {
127 if (!memcmp(rt->rt_ifa->ifa_addr, gate, gate->sa_len))
132 * Check for other options:
133 * 1) Routes with 'real' IPv4/IPv6 gateway
134 * 2) Loopback host routes (another AF_LINK/sockadd_dl check)
136 if (rt->rt_gateway->sa_len == gate->sa_len &&
137 !memcmp(rt->rt_gateway, gate, gate->sa_len))
139 } while ((rn = rn_mpath_next(rn)) != NULL);
141 return (struct rtentry *)rn;
145 * go through the chain and unlink "rt" from the list
146 * the caller will free "rt"
149 rt_mpath_deldup(struct rtentry *headrt, struct rtentry *rt)
151 struct radix_node *t, *tt;
155 t = (struct radix_node *)headrt;
156 tt = rn_mpath_next(t);
158 if (tt == (struct radix_node *)rt) {
159 t->rn_dupedkey = tt->rn_dupedkey;
160 tt->rn_dupedkey = NULL;
161 tt->rn_flags &= ~RNF_ACTIVE;
162 tt[1].rn_flags &= ~RNF_ACTIVE;
166 tt = rn_mpath_next((struct radix_node *)t);
172 * check if we have the same key/mask/gateway on the table already.
173 * Assume @rt rt_key host bits are cleared according to @netmask
176 rt_mpath_conflict(struct rib_head *rnh, struct rtentry *rt,
177 struct sockaddr *netmask)
179 struct radix_node *rn, *rn1;
182 rn = (struct radix_node *)rt;
183 rn1 = rnh->rnh_lookup(rt_key(rt), netmask, &rnh->head);
184 if (!rn1 || rn1->rn_flags & RNF_ROOT)
187 /* key/mask are the same. compare gateway for all multipaths */
189 rt1 = (struct rtentry *)rn1;
191 /* sanity: no use in comparing the same thing */
195 if (rt1->rt_gateway->sa_family == AF_LINK) {
196 if (rt1->rt_ifa->ifa_addr->sa_len != rt->rt_ifa->ifa_addr->sa_len ||
197 bcmp(rt1->rt_ifa->ifa_addr, rt->rt_ifa->ifa_addr,
198 rt1->rt_ifa->ifa_addr->sa_len))
201 if (rt1->rt_gateway->sa_len != rt->rt_gateway->sa_len ||
202 bcmp(rt1->rt_gateway, rt->rt_gateway,
203 rt1->rt_gateway->sa_len))
207 /* all key/mask/gateway are the same. conflicting entry. */
209 } while ((rn1 = rn_mpath_next(rn1)) != NULL);
215 rt_mpath_selectrte(struct rtentry *rte, uint32_t hash)
217 struct radix_node *rn0, *rn;
218 uint32_t total_weight;
222 /* beyond here, we use rn as the master copy */
223 rn0 = rn = (struct radix_node *)rte;
226 /* gw selection by Modulo-N Hash (RFC2991) XXX need improvement? */
227 total_weight = rn_mpath_count(rn0);
229 hash %= total_weight;
230 for (weight = abs((int32_t)hash);
231 rt != NULL && weight >= rt->rt_weight;
232 weight -= (rt == NULL) ? 0 : rt->rt_weight) {
234 /* stay within the multipath routes */
235 if (rn->rn_dupedkey && rn->rn_mask != rn->rn_dupedkey->rn_mask)
237 rn = rn->rn_dupedkey;
238 rt = (struct rtentry *)rn;
245 rt_mpath_select(struct rtentry *rte, uint32_t hash)
247 if (rn_mpath_next((struct radix_node *)rte) == NULL)
250 return (rt_mpath_selectrte(rte, hash));
254 rtalloc_mpath_fib(struct route *ro, uint32_t hash, u_int fibnum)
259 * XXX we don't attempt to lookup cached route again; what should
260 * be done for sendto(3) case?
262 if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP)
263 && RT_LINK_IS_UP(ro->ro_rt->rt_ifp))
265 ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, 0, fibnum);
267 /* if the route does not exist or it is not multipath, don't care */
268 if (ro->ro_rt == NULL)
270 if (rn_mpath_next((struct radix_node *)ro->ro_rt) == NULL) {
271 RT_UNLOCK(ro->ro_rt);
275 rt = rt_mpath_selectrte(ro->ro_rt, hash);
276 /* XXX try filling rt_gwroute and avoid unreachable gw */
278 /* gw selection has failed - there must be only zero weight routes */
280 RT_UNLOCK(ro->ro_rt);
284 if (ro->ro_rt != rt) {
285 RTFREE_LOCKED(ro->ro_rt);
288 RT_ADDREF(ro->ro_rt);
291 RT_UNLOCK(ro->ro_rt);
295 rt_mpath_init_rnh(struct rib_head *rnh)
298 rnh->rnh_multipath = 1;
305 hashjitter = arc4random();
307 SYSINIT(mpath_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, mpath_init, NULL);