/*
- * Copyright (c) 2011 Nicira Networks.
+ * Copyright (c) 2011, 2012, 2013, 2014 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#include "route-table.h"
-#include <assert.h>
+#include <errno.h>
#include <arpa/inet.h>
#include <sys/socket.h>
#include <linux/rtnetlink.h>
#include <net/if.h>
#include "hash.h"
-#include "hmap.h"
#include "netlink.h"
#include "netlink-notifier.h"
#include "netlink-socket.h"
#include "ofpbuf.h"
-#include "rtnetlink-link.h"
-#include "vlog.h"
+#include "ovs-router.h"
+#include "packets.h"
+#include "rtnetlink.h"
+#include "openvswitch/vlog.h"
VLOG_DEFINE_THIS_MODULE(route_table);
unsigned char rtm_dst_len;
/* Extracted from Netlink attributes. */
- uint32_t rta_dst; /* Destination in host byte order. 0 if missing. */
- int rta_oif; /* Output interface index. */
+ struct in6_addr rta_dst; /* 0 if missing. */
+ struct in6_addr rta_gw;
+ char ifname[IFNAMSIZ]; /* Interface name. */
};
/* A digested version of a route message sent down by the kernel to indicate
struct route_data rd; /* Data parsed from this message. */
};
-struct route_node {
- struct hmap_node node; /* Node in route_map. */
- struct route_data rd; /* Data associated with this node. */
-};
-
-struct name_node {
- struct hmap_node node; /* Node in name_map. */
- uint32_t ifi_index; /* Kernel interface index. */
-
- char ifname[IFNAMSIZ]; /* Interface name. */
-};
-
+static struct ovs_mutex route_table_mutex = OVS_MUTEX_INITIALIZER;
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
-static unsigned int register_count = 0;
+/* Global change number for route-table, which should be incremented
+ * every time route_table_reset() is called. */
+static uint64_t rt_change_seq;
+
static struct nln *nln = NULL;
+static struct nln *nln6 = NULL;
static struct route_table_msg rtmsg;
static struct nln_notifier *route_notifier = NULL;
+static struct nln_notifier *route6_notifier = NULL;
static struct nln_notifier *name_notifier = NULL;
static bool route_table_valid = false;
-static bool name_table_valid = false;
-static struct hmap route_map;
-static struct hmap name_map;
static int route_table_reset(void);
static void route_table_handle_msg(const struct route_table_msg *);
static bool route_table_parse(struct ofpbuf *, struct route_table_msg *);
static void route_table_change(const struct route_table_msg *, void *);
-static struct route_node *route_node_lookup(const struct route_data *);
-static struct route_node *route_node_lookup_by_ip(uint32_t ip);
static void route_map_clear(void);
-static uint32_t hash_route_data(const struct route_data *);
static void name_table_init(void);
-static void name_table_uninit(void);
-static int name_table_reset(void);
-static void name_table_change(const struct rtnetlink_link_change *, void *);
-static void name_map_clear(void);
-static struct name_node *name_node_lookup(int ifi_index);
-
-/* Populates 'name' with the name of the interface traffic destined for 'ip'
- * is likely to egress out of (see route_table_get_ifindex).
- *
- * Returns true if successful, otherwise false. */
-bool
-route_table_get_name(ovs_be32 ip, char name[IFNAMSIZ])
-{
- int ifindex;
-
- if (!name_table_valid) {
- name_table_reset();
- }
-
- if (route_table_get_ifindex(ip, &ifindex)) {
- struct name_node *nn;
-
- nn = name_node_lookup(ifindex);
- if (nn) {
- ovs_strlcpy(name, nn->ifname, IFNAMSIZ);
- return true;
- }
- }
-
- return false;
-}
+static void name_table_change(const struct rtnetlink_change *, void *);
-/* Populates 'ifindex' with the interface index traffic destined for 'ip' is
- * likely to egress. There is no hard guarantee that traffic destined for 'ip'
- * will egress out the specified interface. 'ifindex' may refer to an
- * interface which is not physical (such as a bridge port).
- *
- * Returns true if successful, otherwise false. */
-bool
-route_table_get_ifindex(ovs_be32 ip_, int *ifindex)
+uint64_t
+route_table_get_change_seq(void)
{
- struct route_node *rn;
- uint32_t ip = ntohl(ip_);
-
- *ifindex = 0;
-
- if (!route_table_valid) {
- route_table_reset();
- }
-
- rn = route_node_lookup_by_ip(ip);
-
- if (rn) {
- *ifindex = rn->rd.rta_oif;
- return true;
- }
-
- /* Choose a default route. */
- HMAP_FOR_EACH(rn, node, &route_map) {
- if (rn->rd.rta_dst == 0 && rn->rd.rtm_dst_len == 0) {
- *ifindex = rn->rd.rta_oif;
- return true;
- }
- }
-
- return false;
+ return rt_change_seq;
}
/* Users of the route_table module should register themselves with this
* function before making any other route_table function calls. */
void
-route_table_register(void)
-{
- if (!register_count) {
- assert(!nln);
- assert(!route_notifier);
-
- nln = nln_create(NETLINK_ROUTE, RTNLGRP_IPV4_ROUTE,
- (nln_parse_func *) route_table_parse, &rtmsg);
-
- route_notifier =
- nln_notifier_create(nln, (nln_notify_func *) route_table_change,
- NULL);
-
- hmap_init(&route_map);
- route_table_reset();
- name_table_init();
- }
-
- register_count++;
-}
-
-/* Users of the route_table module should unregister themselves with this
- * function when they will no longer be making any more route_table fuction
- * calls. */
-void
-route_table_unregister(void)
+route_table_init(void)
+ OVS_EXCLUDED(route_table_mutex)
{
- register_count--;
-
- if (!register_count) {
- nln_notifier_destroy(route_notifier);
- route_notifier = NULL;
- nln_destroy(nln);
- nln = NULL;
-
- route_map_clear();
- hmap_destroy(&route_map);
- name_table_uninit();
- }
+ ovs_mutex_lock(&route_table_mutex);
+ ovs_assert(!nln);
+ ovs_assert(!nln6);
+ ovs_assert(!route_notifier);
+ ovs_assert(!route6_notifier);
+
+ ovs_router_init();
+ nln = nln_create(NETLINK_ROUTE, RTNLGRP_IPV4_ROUTE,
+ (nln_parse_func *) route_table_parse, &rtmsg);
+ nln6 = nln_create(NETLINK_ROUTE, RTNLGRP_IPV6_ROUTE,
+ (nln_parse_func *) route_table_parse, &rtmsg);
+
+ route_notifier =
+ nln_notifier_create(nln, (nln_notify_func *) route_table_change,
+ NULL);
+ route6_notifier =
+ nln_notifier_create(nln6, (nln_notify_func *) route_table_change,
+ NULL);
+
+ route_table_reset();
+ name_table_init();
+
+ ovs_mutex_unlock(&route_table_mutex);
}
/* Run periodically to update the locally maintained routing table. */
void
route_table_run(void)
+ OVS_EXCLUDED(route_table_mutex)
{
- if (nln) {
- rtnetlink_link_run();
- nln_run(nln);
+ ovs_mutex_lock(&route_table_mutex);
+ if (nln || nln6) {
+ rtnetlink_run();
+ if (nln) {
+ nln_run(nln);
+ }
+ if (nln6) {
+ nln_run(nln6);
+ }
+
+ if (!route_table_valid) {
+ route_table_reset();
+ }
}
+ ovs_mutex_unlock(&route_table_mutex);
}
/* Causes poll_block() to wake up when route_table updates are required. */
void
route_table_wait(void)
+ OVS_EXCLUDED(route_table_mutex)
{
- if (nln) {
- rtnetlink_link_wait();
- nln_wait(nln);
+ ovs_mutex_lock(&route_table_mutex);
+ if (nln || nln6) {
+ rtnetlink_wait();
+ if (nln) {
+ nln_wait(nln);
+ }
+ if (nln6) {
+ nln_wait(nln6);
+ }
}
+ ovs_mutex_unlock(&route_table_mutex);
}
static int
route_table_reset(void)
{
- int error;
struct nl_dump dump;
struct rtgenmsg *rtmsg;
- struct ofpbuf request, reply;
- static struct nl_sock *rtnl_sock;
+ uint64_t reply_stub[NL_DUMP_BUFSIZE / 8];
+ struct ofpbuf request, reply, buf;
route_map_clear();
route_table_valid = true;
-
- error = nl_sock_create(NETLINK_ROUTE, &rtnl_sock);
- if (error) {
- VLOG_WARN_RL(&rl, "failed to reset routing table, "
- "cannot create RTNETLINK_ROUTE socket");
- return error;
- }
+ rt_change_seq++;
ofpbuf_init(&request, 0);
nl_msg_put_nlmsghdr(&request, sizeof *rtmsg, RTM_GETROUTE, NLM_F_REQUEST);
rtmsg = ofpbuf_put_zeros(&request, sizeof *rtmsg);
- rtmsg->rtgen_family = AF_INET;
+ rtmsg->rtgen_family = AF_UNSPEC;
- nl_dump_start(&dump, rtnl_sock, &request);
+ nl_dump_start(&dump, NETLINK_ROUTE, &request);
+ ofpbuf_uninit(&request);
- while (nl_dump_next(&dump, &reply)) {
+ ofpbuf_use_stub(&buf, reply_stub, sizeof reply_stub);
+ while (nl_dump_next(&dump, &reply, &buf)) {
struct route_table_msg msg;
if (route_table_parse(&reply, &msg)) {
route_table_handle_msg(&msg);
}
}
+ ofpbuf_uninit(&buf);
- error = nl_dump_done(&dump);
- nl_sock_destroy(rtnl_sock);
-
- return error;
+ return nl_dump_done(&dump);
}
-
static bool
route_table_parse(struct ofpbuf *buf, struct route_table_msg *change)
{
- bool parsed;
+ bool parsed, ipv4 = false;
static const struct nl_policy policy[] = {
[RTA_DST] = { .type = NL_A_U32, .optional = true },
[RTA_OIF] = { .type = NL_A_U32, .optional = false },
+ [RTA_GATEWAY] = { .type = NL_A_U32, .optional = true },
};
- static struct nlattr *attrs[ARRAY_SIZE(policy)];
+ static const struct nl_policy policy6[] = {
+ [RTA_DST] = { .type = NL_A_IPV6, .optional = true },
+ [RTA_OIF] = { .type = NL_A_U32, .optional = true },
+ [RTA_GATEWAY] = { .type = NL_A_IPV6, .optional = true },
+ };
+
+ struct nlattr *attrs[ARRAY_SIZE(policy)];
+ const struct rtmsg *rtm;
- parsed = nl_policy_parse(buf, NLMSG_HDRLEN + sizeof(struct rtmsg),
- policy, attrs, ARRAY_SIZE(policy));
+ rtm = ofpbuf_at(buf, NLMSG_HDRLEN, sizeof *rtm);
+
+ if (rtm->rtm_family == AF_INET) {
+ parsed = nl_policy_parse(buf, NLMSG_HDRLEN + sizeof(struct rtmsg),
+ policy, attrs, ARRAY_SIZE(policy));
+ ipv4 = true;
+ } else if (rtm->rtm_family == AF_INET6) {
+ parsed = nl_policy_parse(buf, NLMSG_HDRLEN + sizeof(struct rtmsg),
+ policy6, attrs, ARRAY_SIZE(policy6));
+ } else {
+ VLOG_DBG_RL(&rl, "received non AF_INET rtnetlink route message");
+ return false;
+ }
if (parsed) {
- const struct rtmsg *rtm;
const struct nlmsghdr *nlmsg;
+ int rta_oif; /* Output interface index. */
nlmsg = buf->data;
- rtm = (const struct rtmsg *) ((const char *) buf->data + NLMSG_HDRLEN);
-
- if (rtm->rtm_family != AF_INET) {
- VLOG_DBG_RL(&rl, "received non AF_INET rtnetlink route message");
- return false;
- }
memset(change, 0, sizeof *change);
change->relevant = true;
rtm->rtm_type != RTN_LOCAL) {
change->relevant = false;
}
-
change->nlmsg_type = nlmsg->nlmsg_type;
- change->rd.rtm_dst_len = rtm->rtm_dst_len;
- change->rd.rta_oif = nl_attr_get_u32(attrs[RTA_OIF]);
+ change->rd.rtm_dst_len = rtm->rtm_dst_len + (ipv4 ? 96 : 0);
+ if (attrs[RTA_OIF]) {
+ rta_oif = nl_attr_get_u32(attrs[RTA_OIF]);
+
+ if (!if_indextoname(rta_oif, change->rd.ifname)) {
+ int error = errno;
+
+ VLOG_DBG_RL(&rl, "Could not find interface name[%u]: %s",
+ rta_oif, ovs_strerror(error));
+ return false;
+ }
+ }
if (attrs[RTA_DST]) {
- change->rd.rta_dst = ntohl(nl_attr_get_be32(attrs[RTA_DST]));
+ if (ipv4) {
+ ovs_be32 dst;
+ dst = nl_attr_get_be32(attrs[RTA_DST]);
+ in6_addr_set_mapped_ipv4(&change->rd.rta_dst, dst);
+ } else {
+ change->rd.rta_dst = nl_attr_get_in6_addr(attrs[RTA_DST]);
+ }
+ } else if (ipv4) {
+ in6_addr_set_mapped_ipv4(&change->rd.rta_dst, 0);
}
+ if (attrs[RTA_GATEWAY]) {
+ if (ipv4) {
+ ovs_be32 gw;
+ gw = nl_attr_get_be32(attrs[RTA_GATEWAY]);
+ in6_addr_set_mapped_ipv4(&change->rd.rta_gw, gw);
+ } else {
+ change->rd.rta_gw = nl_attr_get_in6_addr(attrs[RTA_GATEWAY]);
+ }
+ }
+
} else {
VLOG_DBG_RL(&rl, "received unparseable rtnetlink route message");
static void
route_table_handle_msg(const struct route_table_msg *change)
{
- if (change->relevant && change->nlmsg_type == RTM_NEWROUTE &&
- !route_node_lookup(&change->rd)) {
- struct route_node *rn;
-
- rn = xzalloc(sizeof *rn);
- memcpy(&rn->rd, &change->rd, sizeof change->rd);
-
- hmap_insert(&route_map, &rn->node, hash_route_data(&rn->rd));
- }
-}
-
-static struct route_node *
-route_node_lookup(const struct route_data *rd)
-{
- struct route_node *rn;
+ if (change->relevant && change->nlmsg_type == RTM_NEWROUTE) {
+ const struct route_data *rd = &change->rd;
- HMAP_FOR_EACH_WITH_HASH(rn, node, hash_route_data(rd), &route_map) {
- if (!memcmp(&rn->rd, rd, sizeof *rd)) {
- return rn;
- }
+ ovs_router_insert(&rd->rta_dst, rd->rtm_dst_len,
+ rd->ifname, &rd->rta_gw);
}
-
- return NULL;
-}
-
-static struct route_node *
-route_node_lookup_by_ip(uint32_t ip)
-{
- int dst_len;
- struct route_node *rn, *rn_ret;
-
- dst_len = -1;
- rn_ret = NULL;
-
- HMAP_FOR_EACH(rn, node, &route_map) {
- uint32_t mask = 0xffffffff << (32 - rn->rd.rtm_dst_len);
-
- if (rn->rd.rta_dst == 0 && rn->rd.rtm_dst_len == 0) {
- /* Default route. */
- continue;
- }
-
- if (rn->rd.rtm_dst_len > dst_len &&
- (ip & mask) == (rn->rd.rta_dst & mask)) {
- rn_ret = rn;
- dst_len = rn->rd.rtm_dst_len;
- }
- }
-
- return rn_ret;
}
static void
route_map_clear(void)
{
- struct route_node *rn, *rn_next;
-
- HMAP_FOR_EACH_SAFE(rn, rn_next, node, &route_map) {
- hmap_remove(&route_map, &rn->node);
- free(rn);
- }
+ ovs_router_flush();
}
-static uint32_t
-hash_route_data(const struct route_data *rd)
+bool
+route_table_fallback_lookup(ovs_be32 ip_dst OVS_UNUSED,
+ char output_bridge[] OVS_UNUSED,
+ ovs_be32 *gw)
{
- return hash_bytes(rd, sizeof *rd, 0);
+ *gw = 0;
+ return false;
}
+
\f
/* name_table . */
static void
name_table_init(void)
{
- hmap_init(&name_map);
- name_notifier = rtnetlink_link_notifier_create(name_table_change, NULL);
- name_table_valid = false;
-}
-
-static void
-name_table_uninit(void)
-{
- rtnetlink_link_notifier_destroy(name_notifier);
- name_notifier = NULL;
- name_map_clear();
- hmap_destroy(&name_map);
+ name_notifier = rtnetlink_notifier_create(name_table_change, NULL);
}
-static int
-name_table_reset(void)
-{
- int error;
- struct nl_dump dump;
- struct rtgenmsg *rtmsg;
- struct ofpbuf request, reply;
- static struct nl_sock *rtnl_sock;
-
- name_table_valid = true;
- name_map_clear();
- error = nl_sock_create(NETLINK_ROUTE, &rtnl_sock);
- if (error) {
- VLOG_WARN_RL(&rl, "failed to create NETLINK_ROUTE socket");
- return error;
- }
-
- ofpbuf_init(&request, 0);
- nl_msg_put_nlmsghdr(&request, sizeof *rtmsg, RTM_GETLINK, NLM_F_REQUEST);
- rtmsg = ofpbuf_put_zeros(&request, sizeof *rtmsg);
- rtmsg->rtgen_family = AF_INET;
-
- nl_dump_start(&dump, rtnl_sock, &request);
- while (nl_dump_next(&dump, &reply)) {
- struct rtnetlink_link_change change;
-
- if (rtnetlink_link_parse(&reply, &change)
- && change.nlmsg_type == RTM_NEWLINK
- && !name_node_lookup(change.ifi_index)) {
- struct name_node *nn;
-
- nn = xzalloc(sizeof *nn);
- nn->ifi_index = change.ifi_index;
- ovs_strlcpy(nn->ifname, change.ifname, IFNAMSIZ);
- hmap_insert(&name_map, &nn->node, hash_int(nn->ifi_index, 0));
- }
- }
- nl_sock_destroy(rtnl_sock);
- return nl_dump_done(&dump);
-}
static void
-name_table_change(const struct rtnetlink_link_change *change OVS_UNUSED,
+name_table_change(const struct rtnetlink_change *change OVS_UNUSED,
void *aux OVS_UNUSED)
{
/* Changes to interface status can cause routing table changes that some
* versions of the linux kernel do not advertise for some reason. */
route_table_valid = false;
- name_table_valid = false;
-}
-
-static struct name_node *
-name_node_lookup(int ifi_index)
-{
- struct name_node *nn;
-
- HMAP_FOR_EACH_WITH_HASH(nn, node, hash_int(ifi_index, 0), &name_map) {
- if (nn->ifi_index == ifi_index) {
- return nn;
- }
- }
-
- return NULL;
-}
-
-static void
-name_map_clear(void)
-{
- struct name_node *nn, *nn_next;
-
- HMAP_FOR_EACH_SAFE(nn, nn_next, node, &name_map) {
- hmap_remove(&name_map, &nn->node);
- free(nn);
- }
}