lib/route-table.c

   1 /*
   2  * Copyright (c) 2011, 2012, 2013, 2014, 2017 Nicira, Inc.
   3  *
   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:
   7  *
   8  *     http://www.apache.org/licenses/LICENSE-2.0
   9  *
  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.
  15  */
  16
  17 #include <config.h>
  18
  19 #include "route-table.h"
  20
  21 #include <errno.h>
  22 #include <sys/types.h>
  23 #include <netinet/in.h>
  24 #include <arpa/inet.h>
  25 #include <sys/socket.h>
  26 #include <linux/rtnetlink.h>
  27 #include <net/if.h>
  28
  29 #include "hash.h"
  30 #include "netdev.h"
  31 #include "netlink.h"
  32 #include "netlink-notifier.h"
  33 #include "netlink-socket.h"
  34 #include "openvswitch/ofpbuf.h"
  35 #include "ovs-router.h"
  36 #include "packets.h"
  37 #include "rtnetlink.h"
  38 #include "tnl-ports.h"
  39 #include "openvswitch/vlog.h"
  40
  41 /* Linux 2.6.36 added RTA_MARK, so define it just in case we're building with
  42  * old headers.  (We can't test for it with #ifdef because it's an enum.) */
  43 #define RTA_MARK 16
  44
  45 VLOG_DEFINE_THIS_MODULE(route_table);
  46
  47 struct route_data {
  48     /* Copied from struct rtmsg. */
  49     unsigned char rtm_dst_len;
  50     bool local;
  51
  52     /* Extracted from Netlink attributes. */
  53     struct in6_addr rta_dst; /* 0 if missing. */
  54     struct in6_addr rta_gw;
  55     char ifname[IFNAMSIZ]; /* Interface name. */
  56     uint32_t mark;
  57 };
  58
  59 /* A digested version of a route message sent down by the kernel to indicate
  60  * that a route has changed. */
  61 struct route_table_msg {
  62     bool relevant;        /* Should this message be processed? */
  63     int nlmsg_type;       /* e.g. RTM_NEWROUTE, RTM_DELROUTE. */
  64     struct route_data rd; /* Data parsed from this message. */
  65 };
  66
  67 static struct ovs_mutex route_table_mutex = OVS_MUTEX_INITIALIZER;
  68 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
  69
  70 /* Global change number for route-table, which should be incremented
  71  * every time route_table_reset() is called.  */
  72 static uint64_t rt_change_seq;
  73
  74 static struct nln *nln = NULL;
  75 static struct route_table_msg rtmsg;
  76 static struct nln_notifier *route_notifier = NULL;
  77 static struct nln_notifier *route6_notifier = NULL;
  78 static struct nln_notifier *name_notifier = NULL;
  79
  80 static bool route_table_valid = false;
  81
  82 static int route_table_reset(void);
  83 static void route_table_handle_msg(const struct route_table_msg *);
  84 static int route_table_parse(struct ofpbuf *, struct route_table_msg *);
  85 static void route_table_change(const struct route_table_msg *, void *);
  86 static void route_map_clear(void);
  87
  88 static void name_table_init(void);
  89 static void name_table_change(const struct rtnetlink_change *, void *);
  90
  91 uint64_t
  92 route_table_get_change_seq(void)
  93 {
  94     return rt_change_seq;
  95 }
  96
  97 /* Users of the route_table module should register themselves with this
  98  * function before making any other route_table function calls. */
  99 void
 100 route_table_init(void)
 101     OVS_EXCLUDED(route_table_mutex)
 102 {
 103     ovs_mutex_lock(&route_table_mutex);
 104     ovs_assert(!nln);
 105     ovs_assert(!route_notifier);
 106     ovs_assert(!route6_notifier);
 107
 108     ovs_router_init();
 109     nln = nln_create(NETLINK_ROUTE, (nln_parse_func *) route_table_parse,
 110                      &rtmsg);
 111
 112     route_notifier =
 113         nln_notifier_create(nln, RTNLGRP_IPV4_ROUTE,
 114                             (nln_notify_func *) route_table_change, NULL);
 115     route6_notifier =
 116         nln_notifier_create(nln, RTNLGRP_IPV6_ROUTE,
 117                             (nln_notify_func *) route_table_change, NULL);
 118
 119     route_table_reset();
 120     name_table_init();
 121
 122     ovs_mutex_unlock(&route_table_mutex);
 123 }
 124
 125 /* Run periodically to update the locally maintained routing table. */
 126 void
 127 route_table_run(void)
 128     OVS_EXCLUDED(route_table_mutex)
 129 {
 130     ovs_mutex_lock(&route_table_mutex);
 131     if (nln) {
 132         rtnetlink_run();
 133         nln_run(nln);
 134
 135         if (!route_table_valid) {
 136             route_table_reset();
 137         }
 138     }
 139     ovs_mutex_unlock(&route_table_mutex);
 140 }
 141
 142 /* Causes poll_block() to wake up when route_table updates are required. */
 143 void
 144 route_table_wait(void)
 145     OVS_EXCLUDED(route_table_mutex)
 146 {
 147     ovs_mutex_lock(&route_table_mutex);
 148     if (nln) {
 149         rtnetlink_wait();
 150         nln_wait(nln);
 151     }
 152     ovs_mutex_unlock(&route_table_mutex);
 153 }
 154
 155 static int
 156 route_table_reset(void)
 157 {
 158     struct nl_dump dump;
 159     struct rtgenmsg *rtgenmsg;
 160     uint64_t reply_stub[NL_DUMP_BUFSIZE / 8];
 161     struct ofpbuf request, reply, buf;
 162
 163     route_map_clear();
 164     netdev_get_addrs_list_flush();
 165     route_table_valid = true;
 166     rt_change_seq++;
 167
 168     ofpbuf_init(&request, 0);
 169
 170     nl_msg_put_nlmsghdr(&request, sizeof *rtgenmsg, RTM_GETROUTE,
 171                         NLM_F_REQUEST);
 172
 173     rtgenmsg = ofpbuf_put_zeros(&request, sizeof *rtgenmsg);
 174     rtgenmsg->rtgen_family = AF_UNSPEC;
 175
 176     nl_dump_start(&dump, NETLINK_ROUTE, &request);
 177     ofpbuf_uninit(&request);
 178
 179     ofpbuf_use_stub(&buf, reply_stub, sizeof reply_stub);
 180     while (nl_dump_next(&dump, &reply, &buf)) {
 181         struct route_table_msg msg;
 182
 183         if (route_table_parse(&reply, &msg)) {
 184             route_table_handle_msg(&msg);
 185         }
 186     }
 187     ofpbuf_uninit(&buf);
 188
 189     return nl_dump_done(&dump);
 190 }
 191
 192 /* Return RTNLGRP_IPV4_ROUTE or RTNLGRP_IPV6_ROUTE on success, 0 on parse
 193  * error. */
 194 static int
 195 route_table_parse(struct ofpbuf *buf, struct route_table_msg *change)
 196 {
 197     bool parsed, ipv4 = false;
 198
 199     static const struct nl_policy policy[] = {
 200         [RTA_DST] = { .type = NL_A_U32, .optional = true  },
 201         [RTA_OIF] = { .type = NL_A_U32, .optional = true },
 202         [RTA_GATEWAY] = { .type = NL_A_U32, .optional = true },
 203         [RTA_MARK] = { .type = NL_A_U32, .optional = true },
 204     };
 205
 206     static const struct nl_policy policy6[] = {
 207         [RTA_DST] = { .type = NL_A_IPV6, .optional = true },
 208         [RTA_OIF] = { .type = NL_A_U32, .optional = true },
 209         [RTA_MARK] = { .type = NL_A_U32, .optional = true },
 210         [RTA_GATEWAY] = { .type = NL_A_IPV6, .optional = true },
 211     };
 212
 213     struct nlattr *attrs[ARRAY_SIZE(policy)];
 214     const struct rtmsg *rtm;
 215
 216     rtm = ofpbuf_at(buf, NLMSG_HDRLEN, sizeof *rtm);
 217
 218     if (rtm->rtm_family == AF_INET) {
 219         parsed = nl_policy_parse(buf, NLMSG_HDRLEN + sizeof(struct rtmsg),
 220                                  policy, attrs, ARRAY_SIZE(policy));
 221         ipv4 = true;
 222     } else if (rtm->rtm_family == AF_INET6) {
 223         parsed = nl_policy_parse(buf, NLMSG_HDRLEN + sizeof(struct rtmsg),
 224                                  policy6, attrs, ARRAY_SIZE(policy6));
 225     } else {
 226         VLOG_DBG_RL(&rl, "received non AF_INET rtnetlink route message");
 227         return 0;
 228     }
 229
 230     if (parsed) {
 231         const struct nlmsghdr *nlmsg;
 232         int rta_oif;      /* Output interface index. */
 233
 234         nlmsg = buf->data;
 235
 236         memset(change, 0, sizeof *change);
 237         change->relevant = true;
 238
 239         if (rtm->rtm_scope == RT_SCOPE_NOWHERE) {
 240             change->relevant = false;
 241         }
 242
 243         if (rtm->rtm_type != RTN_UNICAST &&
 244             rtm->rtm_type != RTN_LOCAL) {
 245             change->relevant = false;
 246         }
 247         change->nlmsg_type     = nlmsg->nlmsg_type;
 248         change->rd.rtm_dst_len = rtm->rtm_dst_len + (ipv4 ? 96 : 0);
 249         change->rd.local = rtm->rtm_type == RTN_LOCAL;
 250         if (attrs[RTA_OIF]) {
 251             rta_oif = nl_attr_get_u32(attrs[RTA_OIF]);
 252
 253             if (!if_indextoname(rta_oif, change->rd.ifname)) {
 254                 int error = errno;
 255
 256                 VLOG_DBG_RL(&rl, "Could not find interface name[%u]: %s",
 257                             rta_oif, ovs_strerror(error));
 258                 if (error == ENXIO) {
 259                     change->relevant = false;
 260                 } else {
 261                     return 0;
 262                 }
 263             }
 264         }
 265
 266         if (attrs[RTA_DST]) {
 267             if (ipv4) {
 268                 ovs_be32 dst;
 269                 dst = nl_attr_get_be32(attrs[RTA_DST]);
 270                 in6_addr_set_mapped_ipv4(&change->rd.rta_dst, dst);
 271             } else {
 272                 change->rd.rta_dst = nl_attr_get_in6_addr(attrs[RTA_DST]);
 273             }
 274         } else if (ipv4) {
 275             in6_addr_set_mapped_ipv4(&change->rd.rta_dst, 0);
 276         }
 277         if (attrs[RTA_GATEWAY]) {
 278             if (ipv4) {
 279                 ovs_be32 gw;
 280                 gw = nl_attr_get_be32(attrs[RTA_GATEWAY]);
 281                 in6_addr_set_mapped_ipv4(&change->rd.rta_gw, gw);
 282             } else {
 283                 change->rd.rta_gw = nl_attr_get_in6_addr(attrs[RTA_GATEWAY]);
 284             }
 285         }
 286         if (attrs[RTA_MARK]) {
 287             change->rd.mark = nl_attr_get_u32(attrs[RTA_MARK]);
 288         }
 289     } else {
 290         VLOG_DBG_RL(&rl, "received unparseable rtnetlink route message");
 291         return 0;
 292     }
 293
 294     /* Success. */
 295     return ipv4 ? RTNLGRP_IPV4_ROUTE : RTNLGRP_IPV6_ROUTE;
 296 }
 297
 298 static void
 299 route_table_change(const struct route_table_msg *change OVS_UNUSED,
 300                    void *aux OVS_UNUSED)
 301 {
 302     route_table_valid = false;
 303 }
 304
 305 static void
 306 route_table_handle_msg(const struct route_table_msg *change)
 307 {
 308     if (change->relevant && change->nlmsg_type == RTM_NEWROUTE) {
 309         const struct route_data *rd = &change->rd;
 310
 311         ovs_router_insert(rd->mark, &rd->rta_dst, rd->rtm_dst_len,
 312                           rd->local, rd->ifname, &rd->rta_gw);
 313     }
 314 }
 315
 316 static void
 317 route_map_clear(void)
 318 {
 319     ovs_router_flush();
 320 }
 321
 322 bool
 323 route_table_fallback_lookup(const struct in6_addr *ip6_dst OVS_UNUSED,
 324                             char name[] OVS_UNUSED,
 325                             struct in6_addr *gw6)
 326 {
 327     *gw6 = in6addr_any;
 328     return false;
 329 }
 330
 331 \f
 332 /* name_table . */
 333
 334 static void
 335 name_table_init(void)
 336 {
 337     name_notifier = rtnetlink_notifier_create(name_table_change, NULL);
 338 }
 339
 340
 341 static void
 342 name_table_change(const struct rtnetlink_change *change,
 343                   void *aux OVS_UNUSED)
 344 {
 345     /* Changes to interface status can cause routing table changes that some
 346      * versions of the linux kernel do not advertise for some reason. */
 347     route_table_valid = false;
 348
 349     if (change && change->nlmsg_type == RTM_DELLINK) {
 350         if (change->ifname) {
 351             tnl_port_map_delete_ipdev(change->ifname);
 352         }
 353     }
 354 }