route.c

   1 /*
   2  *   Copyright (C) 2010 Felix Fietkau <nbd@openwrt.org>
   3  *
   4  *   This program is free software; you can redistribute it and/or modify
   5  *   it under the terms of the GNU General Public License v2 as published by
   6  *   the Free Software Foundation.
   7  *
   8  *   This program is distributed in the hope that it will be useful,
   9  *   but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  11  *   GNU General Public License for more details.
  12  *
  13  *   You should have received a copy of the GNU General Public License
  14  *   along with this program; if not, write to the Free Software
  15  *   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA.
  16  */
  17
  18 #include <sys/socket.h>
  19 #include <stdio.h>
  20 #include <string.h>
  21 #include <errno.h>
  22 #include <unistd.h>
  23 #include <fcntl.h>
  24 #include <time.h>
  25
  26 #include <linux/fib_rules.h>
  27
  28 #include "relayd.h"
  29
  30 static struct uloop_fd rtnl_sock;
  31 static unsigned int rtnl_seq, rtnl_dump_seq;
  32 int route_table = 16800;
  33
  34 static void rtnl_flush(void)
  35 {
  36         int fd;
  37
  38         fd = open("/proc/sys/net/ipv4/route/flush", O_WRONLY);
  39         if (fd < 0)
  40                 return;
  41
  42         write(fd, "-1", 2);
  43         close(fd);
  44 }
  45
  46 enum {
  47         RULE_F_ADD = (1 << 0),
  48         RULE_F_DEFGW_WORKAROUND = (1 << 1),
  49 };
  50
  51 static int get_route_table(struct relayd_interface *rif)
  52 {
  53         if (rif)
  54                 return rif->rt_table;
  55         else
  56                 return local_route_table;
  57 }
  58
  59 static void
  60 rtnl_rule_request(struct relayd_interface *rif, int flags)
  61 {
  62         static struct {
  63                 struct nlmsghdr nl;
  64                 struct rtmsg rt;
  65                 struct {
  66                         struct rtattr rta;
  67                         int table;
  68                 } __packed table;
  69                 struct {
  70                         struct rtattr rta;
  71                         char ifname[IFNAMSIZ + 1];
  72                 } __packed dev;
  73         } __packed req = {
  74                 .rt = {
  75                         .rtm_family = AF_INET,
  76                         .rtm_table = RT_TABLE_UNSPEC,
  77                         .rtm_scope = RT_SCOPE_UNIVERSE,
  78                         .rtm_protocol = RTPROT_BOOT,
  79                 },
  80                 .table.rta = {
  81                         .rta_type = FRA_TABLE,
  82                         .rta_len = sizeof(req.table),
  83                 },
  84         };
  85         const char *ifname = "lo";
  86         int padding = sizeof(req.dev.ifname);
  87
  88         if (rif)
  89                 ifname = rif->ifname;
  90
  91         if (!(flags & RULE_F_DEFGW_WORKAROUND)) {
  92                 req.dev.rta.rta_type = FRA_IFNAME;
  93                 padding -= strlen(ifname) + 1;
  94                 strcpy(req.dev.ifname, ifname);
  95                 req.dev.rta.rta_len = sizeof(req.dev.rta) + strlen(ifname) + 1;
  96         } else {
  97                 req.dev.rta.rta_type = FRA_PRIORITY;
  98                 req.dev.rta.rta_len = sizeof(req.dev.rta) + sizeof(uint32_t);
  99                 padding -= sizeof(uint32_t);
 100                 *((uint32_t *) &req.dev.ifname) = 1;
 101         }
 102         req.table.table = get_route_table(rif);
 103         req.nl.nlmsg_len = sizeof(req) - padding;
 104
 105         req.nl.nlmsg_flags = NLM_F_REQUEST;
 106         if (flags & RULE_F_ADD) {
 107                 req.nl.nlmsg_type = RTM_NEWRULE;
 108                 req.nl.nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
 109
 110                 req.rt.rtm_type = RTN_UNICAST;
 111         } else {
 112                 req.nl.nlmsg_type = RTM_DELRULE;
 113                 req.rt.rtm_type = RTN_UNSPEC;
 114         }
 115
 116         send(rtnl_sock.fd, &req, req.nl.nlmsg_len, 0);
 117         rtnl_flush();
 118 }
 119
 120 struct rtnl_addr {
 121         struct rtattr rta;
 122         uint8_t ipaddr[4];
 123 } __packed;
 124
 125 static struct rtnl_addr *
 126 rtnl_add_addr(struct rtnl_addr *addr, int *len, int type, const uint8_t *ipaddr)
 127 {
 128         addr->rta.rta_type = type;
 129         memcpy(addr->ipaddr, ipaddr, 4);
 130         *len += sizeof(*addr);
 131         return addr + 1;
 132 }
 133
 134 static void
 135 rtnl_route_request(struct relayd_interface *rif, struct relayd_host *host,
 136                    struct relayd_route *route, bool add)
 137 {
 138         static struct {
 139                 struct nlmsghdr nl;
 140                 struct rtmsg rt;
 141                 struct {
 142                         struct rtattr rta;
 143                         int table;
 144                 } __packed table;
 145                 struct {
 146                         struct rtattr rta;
 147                         int ifindex;
 148                 } __packed dev;
 149                 struct rtnl_addr addr[3];
 150         } __packed req = {
 151                 .rt = {
 152                         .rtm_family = AF_INET,
 153                         .rtm_dst_len = 32,
 154                         .rtm_table = RT_TABLE_MAIN,
 155                 },
 156                 .table.rta = {
 157                         .rta_type = RTA_TABLE,
 158                         .rta_len = sizeof(req.table),
 159                 },
 160                 .dev.rta = {
 161                         .rta_type = RTA_OIF,
 162                         .rta_len = sizeof(req.dev),
 163                 },
 164                 .addr[0].rta.rta_len = sizeof(struct rtnl_addr),
 165                 .addr[1].rta.rta_len = sizeof(struct rtnl_addr),
 166                 .addr[2].rta.rta_len = sizeof(struct rtnl_addr),
 167         };
 168         int pktlen = sizeof(req) - sizeof(req.addr);
 169         struct rtnl_addr *addr = &req.addr[0];
 170         const char *ifname = "loopback";
 171
 172         req.dev.ifindex = host->rif->sll.sll_ifindex;
 173         req.table.table = get_route_table(rif);
 174
 175         req.nl.nlmsg_flags = NLM_F_REQUEST;
 176         if (add) {
 177                 req.nl.nlmsg_type = RTM_NEWROUTE;
 178                 req.nl.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
 179
 180                 req.rt.rtm_protocol = RTPROT_BOOT;
 181                 if (route) {
 182                         req.rt.rtm_scope = RT_SCOPE_UNIVERSE;
 183                 } else {
 184                         req.rt.rtm_scope = RT_SCOPE_LINK;
 185                 }
 186                 req.rt.rtm_type = RTN_UNICAST;
 187         } else {
 188                 req.nl.nlmsg_type = RTM_DELROUTE;
 189                 req.rt.rtm_scope = RT_SCOPE_NOWHERE;
 190         }
 191
 192         if (rif)
 193                 ifname = rif->ifname;
 194
 195         if (route) {
 196                 DPRINTF(2, "%s: add route to "IP_FMT"/%d via "IP_FMT" (%s)\n", ifname,
 197                         IP_BUF(route->dest), route->mask, IP_BUF(host->ipaddr),
 198                         host->rif->ifname);
 199
 200                 req.rt.rtm_dst_len = route->mask;
 201                 if (route->mask)
 202                         addr = rtnl_add_addr(addr, &pktlen, RTA_DST, route->dest);
 203                 addr = rtnl_add_addr(addr, &pktlen, RTA_GATEWAY, host->ipaddr);
 204         } else {
 205                 DPRINTF(2, "%s: add host route to "IP_FMT" (%s)\n", ifname,
 206                         IP_BUF(host->ipaddr), host->rif->ifname);
 207                 addr = rtnl_add_addr(addr, &pktlen, RTA_DST, host->ipaddr);
 208                 req.rt.rtm_dst_len = 32;
 209         }
 210
 211         /* local route */
 212         if (!rif)
 213                 addr = rtnl_add_addr(addr, &pktlen, RTA_PREFSRC, local_addr);
 214
 215         req.nl.nlmsg_len = pktlen;
 216         if (route)
 217                 rtnl_rule_request(rif, RULE_F_DEFGW_WORKAROUND | RULE_F_ADD);
 218         send(rtnl_sock.fd, &req, pktlen, 0);
 219         if (route)
 220                 rtnl_rule_request(rif, RULE_F_DEFGW_WORKAROUND);
 221         rtnl_flush();
 222 }
 223
 224 void
 225 rtnl_route_set(struct relayd_host *host, struct relayd_route *route, bool add)
 226 {
 227         struct relayd_interface *rif;
 228
 229         list_for_each_entry(rif, &interfaces, list) {
 230                 if (rif == host->rif)
 231                         continue;
 232
 233                 rtnl_route_request(rif, host, route, add);
 234         }
 235         if (local_route_table)
 236                 rtnl_route_request(NULL, host, route, add);
 237 }
 238
 239 void relayd_add_interface_routes(struct relayd_interface *rif)
 240 {
 241         rif->rt_table = route_table++;
 242         rtnl_rule_request(rif, RULE_F_ADD);
 243 }
 244
 245 void relayd_del_interface_routes(struct relayd_interface *rif)
 246 {
 247         rtnl_rule_request(rif, 0);
 248 }
 249
 250 #ifndef NDA_RTA
 251 #define NDA_RTA(r) \
 252     ((struct rtattr*)(((char*)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
 253 #endif
 254
 255 static void rtnl_parse_newneigh(struct nlmsghdr *h)
 256 {
 257         struct relayd_interface *rif = NULL;
 258         struct ndmsg *r = NLMSG_DATA(h);
 259         const uint8_t *lladdr = NULL;
 260         const uint8_t *ipaddr = NULL;
 261         struct rtattr *rta;
 262         int len;
 263
 264         if (r->ndm_family != AF_INET)
 265                 return;
 266
 267         list_for_each_entry(rif, &interfaces, list) {
 268                 if (rif->sll.sll_ifindex == r->ndm_ifindex)
 269                         goto found_interface;
 270         }
 271         return;
 272
 273 found_interface:
 274         len = h->nlmsg_len - NLMSG_LENGTH(sizeof(*r));
 275         for (rta = NDA_RTA(r); RTA_OK(rta, len); rta = RTA_NEXT(rta, len)) {
 276                 switch(rta->rta_type) {
 277                 case NDA_LLADDR:
 278                         lladdr = RTA_DATA(rta);
 279                         break;
 280                 case NDA_DST:
 281                         ipaddr = RTA_DATA(rta);
 282                         break;
 283                 default:
 284                         break;
 285                 }
 286         }
 287
 288         if (!lladdr || !ipaddr || (r->ndm_state & (NUD_INCOMPLETE|NUD_FAILED)))
 289                 return;
 290
 291         if (!memcmp(lladdr, "\x00\x00\x00\x00\x00\x00", ETH_ALEN))
 292                 return;
 293
 294         DPRINTF(1, "%s: Found ARP cache entry for host "IP_FMT" ("MAC_FMT")\n",
 295                 rif->ifname, IP_BUF(ipaddr), MAC_BUF(lladdr));
 296         relayd_refresh_host(rif, lladdr, ipaddr);
 297 }
 298
 299 static void rtnl_parse_packet(void *data, int len)
 300 {
 301         struct nlmsghdr *h;
 302
 303         for (h = data; NLMSG_OK(h, len); h = NLMSG_NEXT(h, len)) {
 304                 if (h->nlmsg_type == NLMSG_DONE ||
 305                     h->nlmsg_type == NLMSG_ERROR)
 306                         return;
 307
 308                 if (h->nlmsg_seq != rtnl_dump_seq)
 309                         continue;
 310
 311                 if (h->nlmsg_type == RTM_NEWNEIGH)
 312                         rtnl_parse_newneigh(h);
 313         }
 314 }
 315
 316 static void rtnl_cb(struct uloop_fd *fd, unsigned int events)
 317 {
 318         struct sockaddr_nl nladdr;
 319         static uint8_t buf[16384];
 320         struct iovec iov = {
 321                 .iov_base = buf,
 322                 .iov_len = sizeof(buf),
 323         };
 324         struct msghdr msg = {
 325                 .msg_name = &nladdr,
 326                 .msg_namelen = sizeof(nladdr),
 327                 .msg_iov = &iov,
 328                 .msg_iovlen = 1,
 329         };
 330
 331         do {
 332                 int len;
 333
 334                 len = recvmsg(rtnl_sock.fd, &msg, 0);
 335                 if (len < 0) {
 336                         if (errno == EINTR)
 337                                 continue;
 338
 339                         return;
 340                 }
 341
 342                 if (!len)
 343                         break;
 344
 345                 if (nladdr.nl_pid != 0)
 346                         continue;
 347
 348                 rtnl_parse_packet(buf, len);
 349         } while (1);
 350 }
 351
 352 static void rtnl_dump_request(int nlmsg_type)
 353 {
 354         static struct {
 355                 struct nlmsghdr nlh;
 356                 struct rtgenmsg g;
 357         } req = {
 358                 .nlh = {
 359                         .nlmsg_len = sizeof(req),
 360                         .nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST,
 361                         .nlmsg_pid = 0,
 362                 },
 363                 .g.rtgen_family = AF_INET,
 364         };
 365         req.nlh.nlmsg_type = nlmsg_type;
 366         req.nlh.nlmsg_seq = rtnl_seq;
 367         send(rtnl_sock.fd, &req, sizeof(req), 0);
 368         rtnl_seq++;
 369 }
 370
 371 int relayd_rtnl_init(void)
 372 {
 373         struct sockaddr_nl snl_local;
 374
 375         rtnl_sock.fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
 376         if (rtnl_sock.fd < 0) {
 377                 perror("socket(AF_NETLINK)");
 378                 return -1;
 379         }
 380
 381         snl_local.nl_family = AF_NETLINK;
 382
 383         if (bind(rtnl_sock.fd, (struct sockaddr *) &snl_local, sizeof(struct sockaddr_nl)) < 0) {
 384                 perror("bind");
 385                 close(rtnl_sock.fd);
 386                 return -1;
 387         }
 388
 389         rtnl_sock.cb = rtnl_cb;
 390         uloop_fd_add(&rtnl_sock, ULOOP_READ | ULOOP_EDGE_TRIGGER);
 391
 392         rtnl_seq = time(NULL);
 393         rtnl_dump_seq = rtnl_seq;
 394         rtnl_dump_request(RTM_GETNEIGH);
 395         rtnl_rule_request(NULL, RULE_F_ADD);
 396
 397         return 0;
 398 }
 399
 400 void relayd_rtnl_done(void)
 401 {
 402         rtnl_rule_request(NULL, 0);
 403         uloop_fd_delete(&rtnl_sock);
 404         close(rtnl_sock.fd);
 405 }