2 * Copyright (c) 2011, 2012, 2013, 2014 Nicira, Inc.
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"
21 #include <arpa/inet.h>
22 #include <sys/socket.h>
23 #include <linux/rtnetlink.h>
29 #include "netlink-notifier.h"
30 #include "netlink-socket.h"
32 #include "rtnetlink-link.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. */
60 struct hmap_node node; /* Node in name_map. */
61 uint32_t ifi_index; /* Kernel interface index. */
63 char ifname[IFNAMSIZ]; /* Interface name. */
66 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
68 /* Global change number for route-table, which should be incremented
69 * every time route_table_reset() is called. */
70 static uint64_t rt_change_seq;
72 static unsigned int register_count = 0;
73 static struct nln *nln = NULL;
74 static struct route_table_msg rtmsg;
75 static struct nln_notifier *route_notifier = NULL;
76 static struct nln_notifier *name_notifier = NULL;
78 static bool route_table_valid = false;
79 static bool name_table_valid = false;
80 static struct hmap route_map;
81 static struct hmap name_map;
83 static int route_table_reset(void);
84 static void route_table_handle_msg(const struct route_table_msg *);
85 static bool route_table_parse(struct ofpbuf *, struct route_table_msg *);
86 static void route_table_change(const struct route_table_msg *, void *);
87 static struct route_node *route_node_lookup(const struct route_data *);
88 static struct route_node *route_node_lookup_by_ip(uint32_t ip);
89 static void route_map_clear(void);
90 static uint32_t hash_route_data(const struct route_data *);
92 static void name_table_init(void);
93 static void name_table_uninit(void);
94 static int name_table_reset(void);
95 static void name_table_change(const struct rtnetlink_link_change *, void *);
96 static void name_map_clear(void);
97 static struct name_node *name_node_lookup(int ifi_index);
99 /* Populates 'name' with the name of the interface traffic destined for 'ip'
100 * is likely to egress out of (see route_table_get_ifindex).
102 * Returns true if successful, otherwise false. */
104 route_table_get_name(ovs_be32 ip, char name[IFNAMSIZ])
108 if (!name_table_valid) {
112 if (route_table_get_ifindex(ip, &ifindex)) {
113 struct name_node *nn;
115 nn = name_node_lookup(ifindex);
117 ovs_strlcpy(name, nn->ifname, IFNAMSIZ);
125 /* Populates 'ifindex' with the interface index traffic destined for 'ip' is
126 * likely to egress. There is no hard guarantee that traffic destined for 'ip'
127 * will egress out the specified interface. 'ifindex' may refer to an
128 * interface which is not physical (such as a bridge port).
130 * Returns true if successful, otherwise false. */
132 route_table_get_ifindex(ovs_be32 ip_, int *ifindex)
134 struct route_node *rn;
135 uint32_t ip = ntohl(ip_);
139 if (!route_table_valid) {
143 rn = route_node_lookup_by_ip(ip);
146 *ifindex = rn->rd.rta_oif;
150 /* Choose a default route. */
151 HMAP_FOR_EACH(rn, node, &route_map) {
152 if (rn->rd.rta_dst == 0 && rn->rd.rtm_dst_len == 0) {
153 *ifindex = rn->rd.rta_oif;
162 route_table_get_change_seq(void)
164 return rt_change_seq;
167 /* Users of the route_table module should register themselves with this
168 * function before making any other route_table function calls. */
170 route_table_register(void)
172 if (!register_count) {
174 ovs_assert(!route_notifier);
176 nln = nln_create(NETLINK_ROUTE, RTNLGRP_IPV4_ROUTE,
177 (nln_parse_func *) route_table_parse, &rtmsg);
180 nln_notifier_create(nln, (nln_notify_func *) route_table_change,
183 hmap_init(&route_map);
191 /* Users of the route_table module should unregister themselves with this
192 * function when they will no longer be making any more route_table fuction
195 route_table_unregister(void)
199 if (!register_count) {
200 nln_notifier_destroy(route_notifier);
201 route_notifier = NULL;
206 hmap_destroy(&route_map);
211 /* Run periodically to update the locally maintained routing table. */
213 route_table_run(void)
216 rtnetlink_link_run();
219 if (!route_table_valid) {
225 /* Causes poll_block() to wake up when route_table updates are required. */
227 route_table_wait(void)
230 rtnetlink_link_wait();
236 route_table_reset(void)
239 struct rtgenmsg *rtmsg;
240 uint64_t reply_stub[NL_DUMP_BUFSIZE / 8];
241 struct ofpbuf request, reply, buf;
244 route_table_valid = true;
247 ofpbuf_init(&request, 0);
249 nl_msg_put_nlmsghdr(&request, sizeof *rtmsg, RTM_GETROUTE, NLM_F_REQUEST);
251 rtmsg = ofpbuf_put_zeros(&request, sizeof *rtmsg);
252 rtmsg->rtgen_family = AF_INET;
254 nl_dump_start(&dump, NETLINK_ROUTE, &request);
255 ofpbuf_uninit(&request);
257 ofpbuf_use_stub(&buf, reply_stub, sizeof reply_stub);
258 while (nl_dump_next(&dump, &reply, &buf)) {
259 struct route_table_msg msg;
261 if (route_table_parse(&reply, &msg)) {
262 route_table_handle_msg(&msg);
267 return nl_dump_done(&dump);
272 route_table_parse(struct ofpbuf *buf, struct route_table_msg *change)
276 static const struct nl_policy policy[] = {
277 [RTA_DST] = { .type = NL_A_U32, .optional = true },
278 [RTA_OIF] = { .type = NL_A_U32, .optional = false },
281 struct nlattr *attrs[ARRAY_SIZE(policy)];
283 parsed = nl_policy_parse(buf, NLMSG_HDRLEN + sizeof(struct rtmsg),
284 policy, attrs, ARRAY_SIZE(policy));
287 const struct rtmsg *rtm;
288 const struct nlmsghdr *nlmsg;
290 nlmsg = ofpbuf_data(buf);
291 rtm = ofpbuf_at(buf, NLMSG_HDRLEN, sizeof *rtm);
293 if (rtm->rtm_family != AF_INET) {
294 VLOG_DBG_RL(&rl, "received non AF_INET rtnetlink route message");
298 memset(change, 0, sizeof *change);
299 change->relevant = true;
301 if (rtm->rtm_scope == RT_SCOPE_NOWHERE) {
302 change->relevant = false;
305 if (rtm->rtm_type != RTN_UNICAST &&
306 rtm->rtm_type != RTN_LOCAL) {
307 change->relevant = false;
310 change->nlmsg_type = nlmsg->nlmsg_type;
311 change->rd.rtm_dst_len = rtm->rtm_dst_len;
312 change->rd.rta_oif = nl_attr_get_u32(attrs[RTA_OIF]);
314 if (attrs[RTA_DST]) {
315 change->rd.rta_dst = ntohl(nl_attr_get_be32(attrs[RTA_DST]));
319 VLOG_DBG_RL(&rl, "received unparseable rtnetlink route message");
326 route_table_change(const struct route_table_msg *change OVS_UNUSED,
327 void *aux OVS_UNUSED)
329 route_table_valid = false;
333 route_table_handle_msg(const struct route_table_msg *change)
335 if (change->relevant && change->nlmsg_type == RTM_NEWROUTE &&
336 !route_node_lookup(&change->rd)) {
337 struct route_node *rn;
339 rn = xzalloc(sizeof *rn);
340 memcpy(&rn->rd, &change->rd, sizeof change->rd);
342 hmap_insert(&route_map, &rn->node, hash_route_data(&rn->rd));
346 static struct route_node *
347 route_node_lookup(const struct route_data *rd)
349 struct route_node *rn;
351 HMAP_FOR_EACH_WITH_HASH(rn, node, hash_route_data(rd), &route_map) {
352 if (!memcmp(&rn->rd, rd, sizeof *rd)) {
360 static struct route_node *
361 route_node_lookup_by_ip(uint32_t ip)
364 struct route_node *rn, *rn_ret;
369 HMAP_FOR_EACH(rn, node, &route_map) {
370 uint32_t mask = 0xffffffff << (32 - rn->rd.rtm_dst_len);
372 if (rn->rd.rta_dst == 0 && rn->rd.rtm_dst_len == 0) {
377 if (rn->rd.rtm_dst_len > dst_len &&
378 (ip & mask) == (rn->rd.rta_dst & mask)) {
380 dst_len = rn->rd.rtm_dst_len;
388 route_map_clear(void)
390 struct route_node *rn, *rn_next;
392 HMAP_FOR_EACH_SAFE(rn, rn_next, node, &route_map) {
393 hmap_remove(&route_map, &rn->node);
399 hash_route_data(const struct route_data *rd)
401 return hash_bytes(rd, sizeof *rd, 0);
407 name_table_init(void)
409 hmap_init(&name_map);
410 name_notifier = rtnetlink_link_notifier_create(name_table_change, NULL);
411 name_table_valid = false;
415 name_table_uninit(void)
417 rtnetlink_link_notifier_destroy(name_notifier);
418 name_notifier = NULL;
420 hmap_destroy(&name_map);
424 name_table_reset(void)
427 struct rtgenmsg *rtmsg;
428 uint64_t reply_stub[NL_DUMP_BUFSIZE / 8];
429 struct ofpbuf request, reply, buf;
431 name_table_valid = true;
434 ofpbuf_init(&request, 0);
435 nl_msg_put_nlmsghdr(&request, sizeof *rtmsg, RTM_GETLINK, NLM_F_REQUEST);
436 rtmsg = ofpbuf_put_zeros(&request, sizeof *rtmsg);
437 rtmsg->rtgen_family = AF_INET;
439 nl_dump_start(&dump, NETLINK_ROUTE, &request);
440 ofpbuf_uninit(&request);
442 ofpbuf_use_stub(&buf, reply_stub, sizeof reply_stub);
443 while (nl_dump_next(&dump, &reply, &buf)) {
444 struct rtnetlink_link_change change;
446 if (rtnetlink_link_parse(&reply, &change)
447 && change.nlmsg_type == RTM_NEWLINK
448 && !name_node_lookup(change.ifi_index)) {
449 struct name_node *nn;
451 nn = xzalloc(sizeof *nn);
452 nn->ifi_index = change.ifi_index;
453 ovs_strlcpy(nn->ifname, change.ifname, IFNAMSIZ);
454 hmap_insert(&name_map, &nn->node, hash_int(nn->ifi_index, 0));
458 return nl_dump_done(&dump);
462 name_table_change(const struct rtnetlink_link_change *change OVS_UNUSED,
463 void *aux OVS_UNUSED)
465 /* Changes to interface status can cause routing table changes that some
466 * versions of the linux kernel do not advertise for some reason. */
467 route_table_valid = false;
468 name_table_valid = false;
471 static struct name_node *
472 name_node_lookup(int ifi_index)
474 struct name_node *nn;
476 HMAP_FOR_EACH_WITH_HASH(nn, node, hash_int(ifi_index, 0), &name_map) {
477 if (nn->ifi_index == ifi_index) {
488 struct name_node *nn, *nn_next;
490 HMAP_FOR_EACH_SAFE(nn, nn_next, node, &name_map) {
491 hmap_remove(&name_map, &nn->node);