2 * Copyright (c) 2011 Nicira Networks.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
19 #include "route-table.h"
22 #include <arpa/inet.h>
23 #include <sys/socket.h>
24 #include <linux/rtnetlink.h>
30 #include "netlink-socket.h"
32 #include "rtnetlink.h"
35 VLOG_DEFINE_THIS_MODULE(route_table);
38 /* Copied from struct rtmsg. */
39 unsigned char rtm_dst_len;
41 /* Extracted from Netlink attributes. */
42 uint32_t rta_dst; /* Destination in host byte order. 0 if missing. */
43 int rta_oif; /* Output interface index. */
46 /* A digested version of a route message sent down by the kernel to indicate
47 * that a route has changed. */
48 struct route_table_msg {
49 bool relevant; /* Should this message be processed? */
50 int nlmsg_type; /* e.g. RTM_NEWROUTE, RTM_DELROUTE. */
51 struct route_data rd; /* Data parsed from this message. */
55 struct hmap_node node; /* Node in route_map. */
56 struct route_data rd; /* Data associated with this node. */
59 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
61 static unsigned int register_count = 0;
62 static struct rtnetlink *rtn = NULL;
63 static struct route_table_msg rtmsg;
64 static struct rtnetlink_notifier notifier;
65 static struct hmap route_map;
67 static int route_table_reset(void);
68 static bool route_table_parse(struct ofpbuf *, struct route_table_msg *);
69 static void route_table_change(const struct route_table_msg *, void *);
70 static struct route_node *route_node_lookup(const struct route_data *);
71 static struct route_node *route_node_lookup_by_ip(uint32_t ip);
72 static void route_map_clear(void);
73 static uint32_t hash_route_data(const struct route_data *);
75 /* Populates 'ifindex' with the interface index traffic destined for 'ip' is
76 * likely to egress. There is no hard guarantee that traffic destined for 'ip'
77 * will egress out the specified interface. 'ifindex' may refer to an
78 * interface which is not physical (such as a bridge port).
80 * Returns true if successful, otherwise false. */
82 route_table_get_ifindex(ovs_be32 ip_, int *ifindex)
84 struct route_node *rn;
85 uint32_t ip = ntohl(ip_);
89 rn = route_node_lookup_by_ip(ip);
92 *ifindex = rn->rd.rta_oif;
96 /* Choose a default route. */
97 HMAP_FOR_EACH(rn, node, &route_map) {
98 if (rn->rd.rta_dst == 0 && rn->rd.rtm_dst_len == 0) {
99 *ifindex = rn->rd.rta_oif;
107 /* Users of the route_table module should register themselves with this
108 * function before making any other route_table function calls. */
110 route_table_register(void)
112 if (!register_count) {
113 rtnetlink_parse_func *pf;
114 rtnetlink_notify_func *nf;
118 pf = (rtnetlink_parse_func *) route_table_parse;
119 nf = (rtnetlink_notify_func *) route_table_change;
121 rtn = rtnetlink_create(RTNLGRP_IPV4_ROUTE, pf, &rtmsg);
122 rtnetlink_notifier_register(rtn, ¬ifier, nf, NULL);
124 hmap_init(&route_map);
131 /* Users of the route_table module should unregister themselves with this
132 * function when they will no longer be making any more route_table fuction
135 route_table_unregister(void)
139 if (!register_count) {
140 rtnetlink_destroy(rtn);
144 hmap_destroy(&route_map);
148 /* Run periodically to update the locally maintained routing table. */
150 route_table_run(void)
153 rtnetlink_notifier_run(rtn);
157 /* Causes poll_block() to wake up when route_table updates are required. */
159 route_table_wait(void)
162 rtnetlink_notifier_wait(rtn);
167 route_table_reset(void)
171 struct rtgenmsg *rtmsg;
172 struct ofpbuf request, reply;
173 static struct nl_sock *rtnl_sock;
177 error = nl_sock_create(NETLINK_ROUTE, 0, 0, 0, &rtnl_sock);
179 VLOG_WARN_RL(&rl, "failed to reset routing table, "
180 "cannot create RTNETLINK_ROUTE socket");
184 ofpbuf_init(&request, 0);
186 nl_msg_put_nlmsghdr(&request, sizeof *rtmsg, RTM_GETROUTE, NLM_F_REQUEST);
188 rtmsg = ofpbuf_put_zeros(&request, sizeof *rtmsg);
189 rtmsg->rtgen_family = AF_INET;
191 nl_dump_start(&dump, rtnl_sock, &request);
193 while (nl_dump_next(&dump, &reply)) {
194 struct route_table_msg msg;
196 if (route_table_parse(&reply, &msg)) {
197 route_table_change(&msg, NULL);
201 error = nl_dump_done(&dump);
202 nl_sock_destroy(rtnl_sock);
209 route_table_parse(struct ofpbuf *buf, struct route_table_msg *change)
213 static const struct nl_policy policy[] = {
214 [RTA_DST] = { .type = NL_A_U32, .optional = true },
215 [RTA_OIF] = { .type = NL_A_U32, .optional = false },
218 static struct nlattr *attrs[ARRAY_SIZE(policy)];
220 parsed = nl_policy_parse(buf, NLMSG_HDRLEN + sizeof(struct rtmsg),
221 policy, attrs, ARRAY_SIZE(policy));
224 const struct rtmsg *rtm;
225 const struct nlmsghdr *nlmsg;
228 rtm = (const struct rtmsg *) ((const char *) buf->data + NLMSG_HDRLEN);
230 if (rtm->rtm_family != AF_INET) {
231 VLOG_DBG_RL(&rl, "received non AF_INET rtnetlink route message");
235 memset(change, 0, sizeof *change);
236 change->relevant = true;
238 if (rtm->rtm_scope == RT_SCOPE_NOWHERE) {
239 change->relevant = false;
242 if (rtm->rtm_type != RTN_UNICAST &&
243 rtm->rtm_type != RTN_LOCAL) {
244 change->relevant = false;
247 change->nlmsg_type = nlmsg->nlmsg_type;
248 change->rd.rtm_dst_len = rtm->rtm_dst_len;
249 change->rd.rta_oif = nl_attr_get_u32(attrs[RTA_OIF]);
251 if (attrs[RTA_DST]) {
252 change->rd.rta_dst = ntohl(nl_attr_get_be32(attrs[RTA_DST]));
256 VLOG_DBG_RL(&rl, "received unparseable rtnetlink route message");
263 route_table_change(const struct route_table_msg *change, void *aux OVS_UNUSED)
266 VLOG_DBG_RL(&rl, "received NULL change message");
268 } else if (!change->relevant) {
269 VLOG_DBG_RL(&rl, "ignoring irrelevant change message");
270 } else if (change->nlmsg_type == RTM_NEWROUTE) {
271 if (!route_node_lookup(&change->rd)) {
272 struct route_node *rn;
274 rn = xzalloc(sizeof *rn);
275 memcpy(&rn->rd, &change->rd, sizeof change->rd);
277 hmap_insert(&route_map, &rn->node, hash_route_data(&rn->rd));
279 VLOG_DBG_RL(&rl, "skipping insertion of duplicate route entry");
281 } else if (change->nlmsg_type == RTM_DELROUTE) {
282 struct route_node *rn;
284 rn = route_node_lookup(&change->rd);
287 hmap_remove(&route_map, &rn->node);
290 VLOG_DBG_RL(&rl, "skipping deletion of non-existent route entry");
295 static struct route_node *
296 route_node_lookup(const struct route_data *rd)
298 struct route_node *rn;
300 HMAP_FOR_EACH_WITH_HASH(rn, node, hash_route_data(rd), &route_map) {
301 if (!memcmp(&rn->rd, rd, sizeof *rd)) {
309 static struct route_node *
310 route_node_lookup_by_ip(uint32_t ip)
313 struct route_node *rn, *rn_ret;
318 HMAP_FOR_EACH(rn, node, &route_map) {
319 uint32_t mask = 0xffffffff << (32 - rn->rd.rtm_dst_len);
321 if (rn->rd.rta_dst == 0 && rn->rd.rtm_dst_len == 0) {
326 if (rn->rd.rtm_dst_len > dst_len &&
327 (ip & mask) == (rn->rd.rta_dst & mask)) {
329 dst_len = rn->rd.rtm_dst_len;
337 route_map_clear(void)
339 struct route_node *rn, *rn_next;
341 HMAP_FOR_EACH_SAFE(rn, rn_next, node, &route_map) {
342 hmap_remove(&route_map, &rn->node);
348 hash_route_data(const struct route_data *rd)
350 return hash_bytes(rd, sizeof *rd, 0);