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 #define NLMSG_ALIGNTO        4U
  31 #define NLMSG_ALIGN(len) ( ((len)+NLMSG_ALIGNTO-1) & ~(NLMSG_ALIGNTO-1) )
  32
  33 static struct uloop_fd rtnl_sock;
  34 static unsigned int rtnl_seq, rtnl_dump_seq;
  35 int route_table = 16800;
  36
  37 static void rtnl_flush(void)
  38 {
  39         int fd;
  40
  41         fd = open("/proc/sys/net/ipv4/route/flush", O_WRONLY);
  42         if (fd < 0)
  43                 return;
  44
  45         write(fd, "-1", 2);
  46         close(fd);
  47 }
  48
  49 enum {
  50         RULE_F_ADD = (1 << 0),
  51         RULE_F_DEFGW_WORKAROUND = (1 << 1),
  52 };
  53
  54 static int get_route_table(struct relayd_interface *rif)
  55 {
  56         if (rif)
  57                 return rif->rt_table;
  58         else
  59                 return local_route_table;
  60 }
  61
  62 static void
  63 rtnl_rule_request(struct relayd_interface *rif, int flags)
  64 {
  65         struct {
  66                 struct nlmsghdr nl;
  67                 struct rtmsg rt;
  68                 struct {
  69                         struct rtattr rta;
  70                         int table;
  71                 } __packed table;
  72                 struct {
  73                         struct rtattr rta;
  74                         int prio;
  75                 } __packed prio;
  76                 struct {
  77                         struct rtattr rta;
  78                         char ifname[IFNAMSIZ + 1];
  79                 } __packed dev;
  80         } __packed req = {
  81                 .rt = {
  82                         .rtm_family = AF_INET,
  83                         .rtm_table = RT_TABLE_UNSPEC,
  84                         .rtm_scope = RT_SCOPE_UNIVERSE,
  85                         .rtm_protocol = RTPROT_BOOT,
  86                 },
  87                 .prio = {
  88                         .rta.rta_type = FRA_PRIORITY,
  89                         .rta.rta_len = sizeof(req.prio),
  90                         .prio = 2,
  91                 },
  92                 .table.rta = {
  93                         .rta_type = FRA_TABLE,
  94                         .rta_len = sizeof(req.table),
  95                 },
  96         };
  97         const char *ifname = "lo";
  98         int padding = sizeof(req.dev.ifname);
  99
 100         if (rif)
 101                 ifname = rif->ifname;
 102
 103         if (!(flags & RULE_F_DEFGW_WORKAROUND)) {
 104                 int len = strlen(ifname) + 1;
 105                 req.dev.rta.rta_type = FRA_IFNAME;
 106                 padding -= NLMSG_ALIGN(len);
 107                 strcpy(req.dev.ifname, ifname);
 108                 req.dev.rta.rta_len = sizeof(req.dev.rta) + len;
 109         } else {
 110                 padding = sizeof(req.dev);
 111                 req.prio.prio--;
 112         }
 113         req.table.table = get_route_table(rif);
 114         req.nl.nlmsg_len = sizeof(req) - padding;
 115
 116         req.nl.nlmsg_flags = NLM_F_REQUEST;
 117         if (flags & RULE_F_ADD) {
 118                 req.nl.nlmsg_type = RTM_NEWRULE;
 119                 req.nl.nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
 120
 121                 req.rt.rtm_type = RTN_UNICAST;
 122         } else {
 123                 req.nl.nlmsg_type = RTM_DELRULE;
 124                 req.rt.rtm_type = RTN_UNSPEC;
 125         }
 126
 127         send(rtnl_sock.fd, &req, req.nl.nlmsg_len, 0);
 128         rtnl_flush();
 129 }
 130
 131 struct rtnl_addr {
 132         struct rtattr rta;
 133         uint8_t ipaddr[4];
 134 } __packed;
 135
 136 static struct rtnl_addr *
 137 rtnl_add_addr(struct rtnl_addr *addr, int *len, int type, const uint8_t *ipaddr)
 138 {
 139         addr->rta.rta_type = type;
 140         memcpy(addr->ipaddr, ipaddr, 4);
 141         *len += sizeof(*addr);
 142         return addr + 1;
 143 }
 144
 145 static void
 146 rtnl_route_request(struct relayd_interface *rif, struct relayd_host *host,
 147                    struct relayd_route *route, bool add)
 148 {
 149         static struct {
 150                 struct nlmsghdr nl;
 151                 struct rtmsg rt;
 152                 struct {
 153                         struct rtattr rta;
 154                         int table;
 155                 } __packed table;
 156                 struct {
 157                         struct rtattr rta;
 158                         int ifindex;
 159                 } __packed dev;
 160                 struct rtnl_addr addr[3];
 161         } __packed req = {
 162                 .rt = {
 163                         .rtm_family = AF_INET,
 164                         .rtm_dst_len = 32,
 165                         .rtm_table = RT_TABLE_MAIN,
 166                 },
 167                 .table.rta = {
 168                         .rta_type = RTA_TABLE,
 169                         .rta_len = sizeof(req.table),
 170                 },
 171                 .dev.rta = {
 172                         .rta_type = RTA_OIF,
 173                         .rta_len = sizeof(req.dev),
 174                 },
 175                 .addr[0].rta.rta_len = sizeof(struct rtnl_addr),
 176                 .addr[1].rta.rta_len = sizeof(struct rtnl_addr),
 177                 .addr[2].rta.rta_len = sizeof(struct rtnl_addr),
 178         };
 179         int pktlen = sizeof(req) - sizeof(req.addr);
 180         struct rtnl_addr *addr = &req.addr[0];
 181         const char *ifname = "loopback";
 182
 183         req.dev.ifindex = host->rif->sll.sll_ifindex;
 184         req.table.table = get_route_table(rif);
 185
 186         req.nl.nlmsg_flags = NLM_F_REQUEST;
 187         if (add) {
 188                 req.nl.nlmsg_type = RTM_NEWROUTE;
 189                 req.nl.nlmsg_flags |= NLM_F_CREATE | NLM_F_REPLACE;
 190
 191                 req.rt.rtm_protocol = RTPROT_BOOT;
 192                 if (route) {
 193                         req.rt.rtm_scope = RT_SCOPE_UNIVERSE;
 194                 } else {
 195                         req.rt.rtm_scope = RT_SCOPE_LINK;
 196                 }
 197                 req.rt.rtm_type = RTN_UNICAST;
 198         } else {
 199                 req.nl.nlmsg_type = RTM_DELROUTE;
 200                 req.rt.rtm_scope = RT_SCOPE_NOWHERE;
 201         }
 202
 203         if (rif)
 204                 ifname = rif->ifname;
 205
 206         if (route) {
 207                 DPRINTF(2, "%s: add route to "IP_FMT"/%d via "IP_FMT" (%s)\n", ifname,
 208                         IP_BUF(route->dest), route->mask, IP_BUF(host->ipaddr),
 209                         host->rif->ifname);
 210
 211                 req.rt.rtm_dst_len = route->mask;
 212                 if (route->mask)
 213                         addr = rtnl_add_addr(addr, &pktlen, RTA_DST, route->dest);
 214                 addr = rtnl_add_addr(addr, &pktlen, RTA_GATEWAY, host->ipaddr);
 215         } else {
 216                 DPRINTF(2, "%s: add host route to "IP_FMT" (%s)\n", ifname,
 217                         IP_BUF(host->ipaddr), host->rif->ifname);
 218                 addr = rtnl_add_addr(addr, &pktlen, RTA_DST, host->ipaddr);
 219                 req.rt.rtm_dst_len = 32;
 220         }
 221
 222         /* local route */
 223         if (!rif)
 224                 addr = rtnl_add_addr(addr, &pktlen, RTA_PREFSRC, local_addr);
 225
 226         req.nl.nlmsg_len = pktlen;
 227         if (route)
 228                 rtnl_rule_request(rif, RULE_F_DEFGW_WORKAROUND | RULE_F_ADD);
 229         send(rtnl_sock.fd, &req, pktlen, 0);
 230         if (route)
 231                 rtnl_rule_request(rif, RULE_F_DEFGW_WORKAROUND);
 232         rtnl_flush();
 233 }
 234
 235 void
 236 rtnl_route_set(struct relayd_host *host, struct relayd_route *route, bool add)
 237 {
 238         struct relayd_interface *rif;
 239
 240         list_for_each_entry(rif, &interfaces, list) {
 241                 if (rif == host->rif)
 242                         continue;
 243
 244                 rtnl_route_request(rif, host, route, add);
 245         }
 246         if (local_route_table)
 247                 rtnl_route_request(NULL, host, route, add);
 248 }
 249
 250 void relayd_add_interface_routes(struct relayd_interface *rif)
 251 {
 252         rif->rt_table = route_table++;
 253         rtnl_rule_request(rif, RULE_F_ADD);
 254 }
 255
 256 void relayd_del_interface_routes(struct relayd_interface *rif)
 257 {
 258         rtnl_rule_request(rif, 0);
 259 }
 260
 261 #ifndef NDA_RTA
 262 #define NDA_RTA(r) \
 263     ((struct rtattr*)(((char*)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
 264 #endif
 265
 266 static void rtnl_parse_newneigh(struct nlmsghdr *h)
 267 {
 268         struct relayd_interface *rif = NULL;
 269         struct ndmsg *r = NLMSG_DATA(h);
 270         const uint8_t *lladdr = NULL;
 271         const uint8_t *ipaddr = NULL;
 272         struct rtattr *rta;
 273         int len;
 274
 275         if (r->ndm_family != AF_INET)
 276                 return;
 277
 278         list_for_each_entry(rif, &interfaces, list) {
 279                 if (rif->sll.sll_ifindex == r->ndm_ifindex)
 280                         goto found_interface;
 281         }
 282         return;
 283
 284 found_interface:
 285         len = h->nlmsg_len - NLMSG_LENGTH(sizeof(*r));
 286         for (rta = NDA_RTA(r); RTA_OK(rta, len); rta = RTA_NEXT(rta, len)) {
 287                 switch(rta->rta_type) {
 288                 case NDA_LLADDR:
 289                         lladdr = RTA_DATA(rta);
 290                         break;
 291                 case NDA_DST:
 292                         ipaddr = RTA_DATA(rta);
 293                         break;
 294                 default:
 295                         break;
 296                 }
 297         }
 298
 299         if (!lladdr || !ipaddr || (r->ndm_state & (NUD_INCOMPLETE|NUD_FAILED)))
 300                 return;
 301
 302         if (!memcmp(lladdr, "\x00\x00\x00\x00\x00\x00", ETH_ALEN))
 303                 return;
 304
 305         DPRINTF(1, "%s: Found ARP cache entry for host "IP_FMT" ("MAC_FMT")\n",
 306                 rif->ifname, IP_BUF(ipaddr), MAC_BUF(lladdr));
 307         relayd_refresh_host(rif, lladdr, ipaddr);
 308 }
 309
 310 static void rtnl_parse_packet(void *data, int len)
 311 {
 312         struct nlmsghdr *h;
 313
 314         for (h = data; NLMSG_OK(h, len); h = NLMSG_NEXT(h, len)) {
 315                 if (h->nlmsg_type == NLMSG_DONE ||
 316                     h->nlmsg_type == NLMSG_ERROR)
 317                         return;
 318
 319                 if (h->nlmsg_seq != rtnl_dump_seq)
 320                         continue;
 321
 322                 if (h->nlmsg_type == RTM_NEWNEIGH)
 323                         rtnl_parse_newneigh(h);
 324         }
 325 }
 326
 327 static void rtnl_cb(struct uloop_fd *fd, unsigned int events)
 328 {
 329         struct sockaddr_nl nladdr;
 330         static uint8_t buf[16384];
 331         struct iovec iov = {
 332                 .iov_base = buf,
 333                 .iov_len = sizeof(buf),
 334         };
 335         struct msghdr msg = {
 336                 .msg_name = &nladdr,
 337                 .msg_namelen = sizeof(nladdr),
 338                 .msg_iov = &iov,
 339                 .msg_iovlen = 1,
 340         };
 341
 342         do {
 343                 int len;
 344
 345                 len = recvmsg(rtnl_sock.fd, &msg, 0);
 346                 if (len < 0) {
 347                         if (errno == EINTR)
 348                                 continue;
 349
 350                         return;
 351                 }
 352
 353                 if (!len)
 354                         break;
 355
 356                 if (nladdr.nl_pid != 0)
 357                         continue;
 358
 359                 rtnl_parse_packet(buf, len);
 360         } while (1);
 361 }
 362
 363 static void rtnl_dump_request(int nlmsg_type)
 364 {
 365         static struct {
 366                 struct nlmsghdr nlh;
 367                 struct rtgenmsg g;
 368         } req = {
 369                 .nlh = {
 370                         .nlmsg_len = sizeof(req),
 371                         .nlmsg_flags = NLM_F_ROOT|NLM_F_MATCH|NLM_F_REQUEST,
 372                         .nlmsg_pid = 0,
 373                 },
 374                 .g.rtgen_family = AF_INET,
 375         };
 376         req.nlh.nlmsg_type = nlmsg_type;
 377         req.nlh.nlmsg_seq = rtnl_seq;
 378         send(rtnl_sock.fd, &req, sizeof(req), 0);
 379         rtnl_seq++;
 380 }
 381
 382 int relayd_rtnl_init(void)
 383 {
 384         struct sockaddr_nl snl_local = {};
 385
 386         rtnl_sock.fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
 387         if (rtnl_sock.fd < 0) {
 388                 perror("socket(AF_NETLINK)");
 389                 return -1;
 390         }
 391
 392         snl_local.nl_family = AF_NETLINK;
 393
 394         if (bind(rtnl_sock.fd, (struct sockaddr *) &snl_local, sizeof(struct sockaddr_nl)) < 0) {
 395                 perror("bind");
 396                 close(rtnl_sock.fd);
 397                 return -1;
 398         }
 399
 400         rtnl_sock.cb = rtnl_cb;
 401         uloop_fd_add(&rtnl_sock, ULOOP_READ | ULOOP_EDGE_TRIGGER);
 402
 403         rtnl_seq = time(NULL);
 404         rtnl_dump_seq = rtnl_seq;
 405         rtnl_dump_request(RTM_GETNEIGH);
 406         rtnl_rule_request(NULL, RULE_F_ADD);
 407
 408         return 0;
 409 }
 410
 411 void relayd_rtnl_done(void)
 412 {
 413         rtnl_rule_request(NULL, 0);
 414         uloop_fd_delete(&rtnl_sock);
 415         close(rtnl_sock.fd);
 416 }