system-linux.c

   1 /*
   2  * netifd - network interface daemon
   3  * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
   4  * Copyright (C) 2013 Jo-Philipp Wich <jow@openwrt.org>
   5  * Copyright (C) 2013 Steven Barth <steven@midlink.org>
   6  * Copyright (C) 2014 Gioacchino Mazzurco <gio@eigenlab.org>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2
  10  * as published by the Free Software Foundation
  11  *
  12  * This program is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  */
  17 #define _GNU_SOURCE
  18
  19 #include <sys/socket.h>
  20 #include <sys/ioctl.h>
  21 #include <sys/stat.h>
  22 #include <sys/syscall.h>
  23
  24 #include <net/if.h>
  25 #include <net/if_arp.h>
  26
  27 #include <arpa/inet.h>
  28 #include <netinet/in.h>
  29
  30 #include <linux/rtnetlink.h>
  31 #include <linux/sockios.h>
  32 #include <linux/ip.h>
  33 #include <linux/if_link.h>
  34 #include <linux/if_vlan.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_tunnel.h>
  37 #include <linux/ip6_tunnel.h>
  38 #include <linux/ethtool.h>
  39 #include <linux/fib_rules.h>
  40 #include <linux/version.h>
  41
  42 #ifndef RTN_FAILED_POLICY
  43 #define RTN_FAILED_POLICY 12
  44 #endif
  45
  46 #include <string.h>
  47 #include <fcntl.h>
  48 #include <glob.h>
  49 #include <time.h>
  50
  51 #include <netlink/msg.h>
  52 #include <netlink/attr.h>
  53 #include <netlink/socket.h>
  54 #include <libubox/uloop.h>
  55
  56 #include "netifd.h"
  57 #include "device.h"
  58 #include "system.h"
  59
  60 struct event_socket {
  61         struct uloop_fd uloop;
  62         struct nl_sock *sock;
  63         int bufsize;
  64 };
  65
  66 static int sock_ioctl = -1;
  67 static struct nl_sock *sock_rtnl = NULL;
  68
  69 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
  70 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
  71
  72 static char dev_buf[256];
  73
  74 static void
  75 handler_nl_event(struct uloop_fd *u, unsigned int events)
  76 {
  77         struct event_socket *ev = container_of(u, struct event_socket, uloop);
  78         int err;
  79         socklen_t errlen = sizeof(err);
  80
  81         if (!u->error) {
  82                 nl_recvmsgs_default(ev->sock);
  83                 return;
  84         }
  85
  86         if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
  87                 goto abort;
  88
  89         switch(err) {
  90         case ENOBUFS:
  91                 // Increase rx buffer size on netlink socket
  92                 ev->bufsize *= 2;
  93                 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
  94                         goto abort;
  95
  96                 // Request full dump since some info got dropped
  97                 struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
  98                 nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
  99                 break;
 100
 101         default:
 102                 goto abort;
 103         }
 104         u->error = false;
 105         return;
 106
 107 abort:
 108         uloop_fd_delete(&ev->uloop);
 109         return;
 110 }
 111
 112 static struct nl_sock *
 113 create_socket(int protocol, int groups)
 114 {
 115         struct nl_sock *sock;
 116
 117         sock = nl_socket_alloc();
 118         if (!sock)
 119                 return NULL;
 120
 121         if (groups)
 122                 nl_join_groups(sock, groups);
 123
 124         if (nl_connect(sock, protocol))
 125                 return NULL;
 126
 127         return sock;
 128 }
 129
 130 static bool
 131 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
 132                         uloop_fd_handler cb, int flags)
 133 {
 134         ev->sock = create_socket(protocol, groups);
 135         if (!ev->sock)
 136                 return false;
 137
 138         ev->uloop.fd = nl_socket_get_fd(ev->sock);
 139         ev->uloop.cb = cb;
 140         if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
 141                 return false;
 142
 143         return true;
 144 }
 145
 146 static bool
 147 create_event_socket(struct event_socket *ev, int protocol,
 148                     int (*cb)(struct nl_msg *msg, void *arg))
 149 {
 150         if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
 151                 return false;
 152
 153         // Install the valid custom callback handler
 154         nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
 155
 156         // Disable sequence number checking on event sockets
 157         nl_socket_disable_seq_check(ev->sock);
 158
 159         // Increase rx buffer size to 65K on event sockets
 160         ev->bufsize = 65535;
 161         if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
 162                 return false;
 163
 164         return true;
 165 }
 166
 167 static bool
 168 system_rtn_aton(const char *src, unsigned int *dst)
 169 {
 170         char *e;
 171         unsigned int n;
 172
 173         if (!strcmp(src, "local"))
 174                 n = RTN_LOCAL;
 175         else if (!strcmp(src, "nat"))
 176                 n = RTN_NAT;
 177         else if (!strcmp(src, "broadcast"))
 178                 n = RTN_BROADCAST;
 179         else if (!strcmp(src, "anycast"))
 180                 n = RTN_ANYCAST;
 181         else if (!strcmp(src, "multicast"))
 182                 n = RTN_MULTICAST;
 183         else if (!strcmp(src, "prohibit"))
 184                 n = RTN_PROHIBIT;
 185         else if (!strcmp(src, "unreachable"))
 186                 n = RTN_UNREACHABLE;
 187         else if (!strcmp(src, "blackhole"))
 188                 n = RTN_BLACKHOLE;
 189         else if (!strcmp(src, "xresolve"))
 190                 n = RTN_XRESOLVE;
 191         else if (!strcmp(src, "unicast"))
 192                 n = RTN_UNICAST;
 193         else if (!strcmp(src, "throw"))
 194                 n = RTN_THROW;
 195         else if (!strcmp(src, "failed_policy"))
 196                 n = RTN_FAILED_POLICY;
 197         else {
 198                 n = strtoul(src, &e, 0);
 199                 if (!e || *e || e == src || n > 255)
 200                         return false;
 201         }
 202
 203         *dst = n;
 204         return true;
 205 }
 206
 207 static bool
 208 system_tos_aton(const char *src, unsigned *dst)
 209 {
 210         char *e;
 211
 212         *dst = strtoul(src, &e, 16);
 213         if (e == src || *e || *dst > 255)
 214                 return false;
 215
 216         return true;
 217 }
 218
 219 int system_init(void)
 220 {
 221         static struct event_socket rtnl_event;
 222         static struct event_socket hotplug_event;
 223
 224         sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
 225         system_fd_set_cloexec(sock_ioctl);
 226
 227         // Prepare socket for routing / address control
 228         sock_rtnl = create_socket(NETLINK_ROUTE, 0);
 229         if (!sock_rtnl)
 230                 return -1;
 231
 232         if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
 233                 return -1;
 234
 235         if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
 236                                      handle_hotplug_event, 0))
 237                 return -1;
 238
 239         // Receive network link events form kernel
 240         nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
 241
 242         return 0;
 243 }
 244
 245 static void system_set_sysctl(const char *path, const char *val)
 246 {
 247         int fd;
 248
 249         fd = open(path, O_WRONLY);
 250         if (fd < 0)
 251                 return;
 252
 253         if (write(fd, val, strlen(val))) {}
 254         close(fd);
 255 }
 256
 257 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
 258 {
 259         snprintf(dev_buf, sizeof(dev_buf), path, device);
 260         system_set_sysctl(dev_buf, val);
 261 }
 262
 263 static void system_set_disable_ipv6(struct device *dev, const char *val)
 264 {
 265         system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
 266 }
 267
 268 static void system_set_rpfilter(struct device *dev, const char *val)
 269 {
 270         system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter", dev->ifname, val);
 271 }
 272
 273 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
 274 {
 275         int fd = -1, ret = -1;
 276
 277         fd = open(path, O_RDONLY);
 278         if (fd < 0)
 279                 goto out;
 280
 281         ssize_t len = read(fd, buf, buf_sz - 1);
 282         if (len < 0)
 283                 goto out;
 284
 285         ret = buf[len] = 0;
 286
 287 out:
 288         if (fd >= 0)
 289                 close(fd);
 290
 291         return ret;
 292 }
 293
 294 static int
 295 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
 296 {
 297         snprintf(dev_buf, sizeof(dev_buf), path, device);
 298         return system_get_sysctl(dev_buf, buf, buf_sz);
 299 }
 300
 301 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
 302 {
 303         return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
 304                         dev->ifname, buf, buf_sz);
 305 }
 306
 307 static int system_get_rpfilter(struct device *dev, char *buf, const size_t buf_sz)
 308 {
 309         return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter",
 310                         dev->ifname, buf, buf_sz);
 311 }
 312
 313 // Evaluate netlink messages
 314 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 315 {
 316         struct nlmsghdr *nh = nlmsg_hdr(msg);
 317         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 318         struct nlattr *nla[__IFLA_MAX];
 319         int link_state = 0;
 320         char buf[10];
 321
 322         if (nh->nlmsg_type != RTM_NEWLINK)
 323                 goto out;
 324
 325         nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
 326         if (!nla[IFLA_IFNAME])
 327                 goto out;
 328
 329         struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
 330         if (!dev || dev->type->keep_link_status)
 331                 goto out;
 332
 333         if (!system_get_dev_sysctl("/sys/class/net/%s/carrier", dev->ifname, buf, sizeof(buf)))
 334                 link_state = strtoul(buf, NULL, 0);
 335
 336         device_set_link(dev, link_state ? true : false);
 337
 338 out:
 339         return 0;
 340 }
 341
 342 static void
 343 handle_hotplug_msg(char *data, int size)
 344 {
 345         const char *subsystem = NULL, *interface = NULL;
 346         char *cur, *end, *sep;
 347         struct device *dev;
 348         int skip;
 349         bool add;
 350
 351         if (!strncmp(data, "add@", 4))
 352                 add = true;
 353         else if (!strncmp(data, "remove@", 7))
 354                 add = false;
 355         else
 356                 return;
 357
 358         skip = strlen(data) + 1;
 359         end = data + size;
 360
 361         for (cur = data + skip; cur < end; cur += skip) {
 362                 skip = strlen(cur) + 1;
 363
 364                 sep = strchr(cur, '=');
 365                 if (!sep)
 366                         continue;
 367
 368                 *sep = 0;
 369                 if (!strcmp(cur, "INTERFACE"))
 370                         interface = sep + 1;
 371                 else if (!strcmp(cur, "SUBSYSTEM")) {
 372                         subsystem = sep + 1;
 373                         if (strcmp(subsystem, "net") != 0)
 374                                 return;
 375                 }
 376                 if (subsystem && interface)
 377                         goto found;
 378         }
 379         return;
 380
 381 found:
 382         dev = device_get(interface, false);
 383         if (!dev)
 384                 return;
 385
 386         if (dev->type != &simple_device_type)
 387                 return;
 388
 389         if (add && system_if_force_external(dev->ifname))
 390                 return;
 391
 392         device_set_present(dev, add);
 393 }
 394
 395 static void
 396 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
 397 {
 398         struct event_socket *ev = container_of(u, struct event_socket, uloop);
 399         struct sockaddr_nl nla;
 400         unsigned char *buf = NULL;
 401         int size;
 402
 403         while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
 404                 if (nla.nl_pid == 0)
 405                         handle_hotplug_msg((char *) buf, size);
 406
 407                 free(buf);
 408         }
 409 }
 410
 411 static int system_rtnl_call(struct nl_msg *msg)
 412 {
 413         int ret;
 414
 415         ret = nl_send_auto_complete(sock_rtnl, msg);
 416         nlmsg_free(msg);
 417
 418         if (ret < 0)
 419                 return ret;
 420
 421         return nl_wait_for_ack(sock_rtnl);
 422 }
 423
 424 int system_bridge_delbr(struct device *bridge)
 425 {
 426         return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 427 }
 428
 429 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 430 {
 431         struct ifreq ifr;
 432
 433         memset(&ifr, 0, sizeof(ifr));
 434         if (dev)
 435                 ifr.ifr_ifindex = dev->ifindex;
 436         else
 437                 ifr.ifr_data = data;
 438         strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
 439         return ioctl(sock_ioctl, cmd, &ifr);
 440 }
 441
 442 static bool system_is_bridge(const char *name, char *buf, int buflen)
 443 {
 444         struct stat st;
 445
 446         snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
 447         if (stat(buf, &st) < 0)
 448                 return false;
 449
 450         return true;
 451 }
 452
 453 static char *system_get_bridge(const char *name, char *buf, int buflen)
 454 {
 455         char *path;
 456         ssize_t len = -1;
 457         glob_t gl;
 458
 459         snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
 460         if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
 461                 return NULL;
 462
 463         if (gl.gl_pathc > 0)
 464                 len = readlink(gl.gl_pathv[0], buf, buflen);
 465
 466         globfree(&gl);
 467
 468         if (len < 0)
 469                 return NULL;
 470
 471         buf[len] = 0;
 472         path = strrchr(buf, '/');
 473         if (!path)
 474                 return NULL;
 475
 476         return path + 1;
 477 }
 478
 479 static void system_bridge_set_wireless(const char *bridge, const char *dev)
 480 {
 481         snprintf(dev_buf, sizeof(dev_buf),
 482                  "/sys/devices/virtual/net/%s/brif/%s/multicast_to_unicast",
 483                  bridge, dev);
 484         system_set_sysctl(dev_buf, "1");
 485 }
 486
 487 int system_bridge_addif(struct device *bridge, struct device *dev)
 488 {
 489         char *oldbr;
 490         int ret = 0;
 491
 492         oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
 493         if (!oldbr || strcmp(oldbr, bridge->ifname) != 0)
 494                 ret = system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 495
 496         if (dev->wireless)
 497                 system_bridge_set_wireless(bridge->ifname, dev->ifname);
 498
 499         return ret;
 500 }
 501
 502 int system_bridge_delif(struct device *bridge, struct device *dev)
 503 {
 504         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 505 }
 506
 507 int system_if_resolve(struct device *dev)
 508 {
 509         struct ifreq ifr;
 510         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 511         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 512                 return ifr.ifr_ifindex;
 513         else
 514                 return 0;
 515 }
 516
 517 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 518 {
 519         struct ifreq ifr;
 520
 521         memset(&ifr, 0, sizeof(ifr));
 522         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 523         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 524         ifr.ifr_flags |= add;
 525         ifr.ifr_flags &= ~rem;
 526         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 527 }
 528
 529 struct clear_data {
 530         struct nl_msg *msg;
 531         struct device *dev;
 532         int type;
 533         int size;
 534         int af;
 535 };
 536
 537
 538 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 539 {
 540         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 541
 542         return ifa->ifa_index == ifindex;
 543 }
 544
 545 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 546 {
 547         struct rtmsg *r = NLMSG_DATA(hdr);
 548         struct nlattr *tb[__RTA_MAX];
 549
 550         if (r->rtm_protocol == RTPROT_KERNEL &&
 551             r->rtm_family == AF_INET6)
 552                 return false;
 553
 554         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 555         if (!tb[RTA_OIF])
 556                 return false;
 557
 558         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 559 }
 560
 561 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
 562 {
 563         return true;
 564 }
 565
 566 static int cb_clear_event(struct nl_msg *msg, void *arg)
 567 {
 568         struct clear_data *clr = arg;
 569         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 570         bool (*cb)(struct nlmsghdr *, int ifindex);
 571         int type;
 572
 573         switch(clr->type) {
 574         case RTM_GETADDR:
 575                 type = RTM_DELADDR;
 576                 if (hdr->nlmsg_type != RTM_NEWADDR)
 577                         return NL_SKIP;
 578
 579                 cb = check_ifaddr;
 580                 break;
 581         case RTM_GETROUTE:
 582                 type = RTM_DELROUTE;
 583                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 584                         return NL_SKIP;
 585
 586                 cb = check_route;
 587                 break;
 588         case RTM_GETRULE:
 589                 type = RTM_DELRULE;
 590                 if (hdr->nlmsg_type != RTM_NEWRULE)
 591                         return NL_SKIP;
 592
 593                 cb = check_rule;
 594                 break;
 595         default:
 596                 return NL_SKIP;
 597         }
 598
 599         if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
 600                 return NL_SKIP;
 601
 602         if (type == RTM_DELRULE)
 603                 D(SYSTEM, "Remove a rule\n");
 604         else
 605                 D(SYSTEM, "Remove %s from device %s\n",
 606                   type == RTM_DELADDR ? "an address" : "a route",
 607                   clr->dev->ifname);
 608         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 609         hdr = nlmsg_hdr(clr->msg);
 610         hdr->nlmsg_type = type;
 611         hdr->nlmsg_flags = NLM_F_REQUEST;
 612
 613         nl_socket_disable_auto_ack(sock_rtnl);
 614         nl_send_auto_complete(sock_rtnl, clr->msg);
 615         nl_socket_enable_auto_ack(sock_rtnl);
 616
 617         return NL_SKIP;
 618 }
 619
 620 static int
 621 cb_finish_event(struct nl_msg *msg, void *arg)
 622 {
 623         int *pending = arg;
 624         *pending = 0;
 625         return NL_STOP;
 626 }
 627
 628 static int
 629 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 630 {
 631         int *pending = arg;
 632         *pending = err->error;
 633         return NL_STOP;
 634 }
 635
 636 static void
 637 system_if_clear_entries(struct device *dev, int type, int af)
 638 {
 639         struct clear_data clr;
 640         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 641         struct rtmsg rtm = {
 642                 .rtm_family = af,
 643                 .rtm_flags = RTM_F_CLONED,
 644         };
 645         int flags = NLM_F_DUMP;
 646         int pending = 1;
 647
 648         clr.af = af;
 649         clr.dev = dev;
 650         clr.type = type;
 651         switch (type) {
 652         case RTM_GETADDR:
 653         case RTM_GETRULE:
 654                 clr.size = sizeof(struct rtgenmsg);
 655                 break;
 656         case RTM_GETROUTE:
 657                 clr.size = sizeof(struct rtmsg);
 658                 break;
 659         default:
 660                 return;
 661         }
 662
 663         if (!cb)
 664                 return;
 665
 666         clr.msg = nlmsg_alloc_simple(type, flags);
 667         if (!clr.msg)
 668                 goto out;
 669
 670         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 671         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 672         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 673         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 674
 675         nl_send_auto_complete(sock_rtnl, clr.msg);
 676         while (pending > 0)
 677                 nl_recvmsgs(sock_rtnl, cb);
 678
 679         nlmsg_free(clr.msg);
 680 out:
 681         nl_cb_put(cb);
 682 }
 683
 684 /*
 685  * Clear bridge (membership) state and bring down device
 686  */
 687 void system_if_clear_state(struct device *dev)
 688 {
 689         static char buf[256];
 690         char *bridge;
 691
 692         device_set_ifindex(dev, system_if_resolve(dev));
 693         if (dev->external || !dev->ifindex)
 694                 return;
 695
 696         system_if_flags(dev->ifname, 0, IFF_UP);
 697
 698         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 699                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 700                 system_bridge_delbr(dev);
 701                 return;
 702         }
 703
 704         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 705         if (bridge) {
 706                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 707                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 708         }
 709
 710         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 711         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 712         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 713         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 714         system_set_disable_ipv6(dev, "0");
 715 }
 716
 717 static inline unsigned long
 718 sec_to_jiffies(int val)
 719 {
 720         return (unsigned long) val * 100;
 721 }
 722
 723 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 724 {
 725         unsigned long args[4] = {};
 726
 727         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 728                 return -1;
 729
 730         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 731         args[1] = !!cfg->stp;
 732         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 733
 734         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 735         args[1] = sec_to_jiffies(cfg->forward_delay);
 736         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 737
 738         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 739                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 740
 741         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_querier",
 742                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 743
 744         args[0] = BRCTL_SET_BRIDGE_PRIORITY;
 745         args[1] = cfg->priority;
 746         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 747
 748         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 749                 args[0] = BRCTL_SET_AGEING_TIME;
 750                 args[1] = sec_to_jiffies(cfg->ageing_time);
 751                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 752         }
 753
 754         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 755                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 756                 args[1] = sec_to_jiffies(cfg->hello_time);
 757                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 758         }
 759
 760         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 761                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 762                 args[1] = sec_to_jiffies(cfg->max_age);
 763                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 764         }
 765
 766         return 0;
 767 }
 768
 769 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
 770 {
 771         struct nl_msg *msg;
 772         struct nlattr *linkinfo, *data;
 773         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
 774         int i, rv;
 775         static const struct {
 776                 const char *name;
 777                 enum macvlan_mode val;
 778         } modes[] = {
 779                 { "private", MACVLAN_MODE_PRIVATE },
 780                 { "vepa", MACVLAN_MODE_VEPA },
 781                 { "bridge", MACVLAN_MODE_BRIDGE },
 782                 { "passthru", MACVLAN_MODE_PASSTHRU },
 783         };
 784
 785         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 786
 787         if (!msg)
 788                 return -1;
 789
 790         nlmsg_append(msg, &iim, sizeof(iim), 0);
 791
 792         if (cfg->flags & MACVLAN_OPT_MACADDR)
 793                 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
 794         nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
 795         nla_put_u32(msg, IFLA_LINK, dev->ifindex);
 796
 797         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 798                 goto nla_put_failure;
 799
 800         nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
 801
 802         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 803                 goto nla_put_failure;
 804
 805         if (cfg->mode) {
 806                 for (i = 0; i < ARRAY_SIZE(modes); i++) {
 807                         if (strcmp(cfg->mode, modes[i].name) != 0)
 808                                 continue;
 809
 810                         nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
 811                         break;
 812                 }
 813         }
 814
 815         nla_nest_end(msg, data);
 816         nla_nest_end(msg, linkinfo);
 817
 818         rv = system_rtnl_call(msg);
 819         if (rv)
 820                 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
 821
 822         return rv;
 823
 824 nla_put_failure:
 825         nlmsg_free(msg);
 826         return -ENOMEM;
 827 }
 828
 829 static int system_link_del(const char *ifname)
 830 {
 831         struct nl_msg *msg;
 832         struct ifinfomsg iim = {
 833                 .ifi_family = AF_UNSPEC,
 834                 .ifi_index = 0,
 835         };
 836
 837         msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
 838
 839         if (!msg)
 840                 return -1;
 841
 842         nlmsg_append(msg, &iim, sizeof(iim), 0);
 843         nla_put_string(msg, IFLA_IFNAME, ifname);
 844         return system_rtnl_call(msg);
 845 }
 846
 847 int system_macvlan_del(struct device *macvlan)
 848 {
 849         return system_link_del(macvlan->ifname);
 850 }
 851
 852 static int system_vlan(struct device *dev, int id)
 853 {
 854         struct vlan_ioctl_args ifr = {
 855                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 856                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 857         };
 858
 859         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 860
 861         if (id < 0) {
 862                 ifr.cmd = DEL_VLAN_CMD;
 863                 ifr.u.VID = 0;
 864         } else {
 865                 ifr.cmd = ADD_VLAN_CMD;
 866                 ifr.u.VID = id;
 867         }
 868         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 869         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 870 }
 871
 872 int system_vlan_add(struct device *dev, int id)
 873 {
 874         return system_vlan(dev, id);
 875 }
 876
 877 int system_vlan_del(struct device *dev)
 878 {
 879         return system_vlan(dev, -1);
 880 }
 881
 882 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
 883 {
 884         struct nl_msg *msg;
 885         struct nlattr *linkinfo, *data;
 886         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
 887         int rv;
 888
 889         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 890
 891         if (!msg)
 892                 return -1;
 893
 894         nlmsg_append(msg, &iim, sizeof(iim), 0);
 895         nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
 896         nla_put_u32(msg, IFLA_LINK, dev->ifindex);
 897
 898         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 899                 goto nla_put_failure;
 900
 901         nla_put_string(msg, IFLA_INFO_KIND, "vlan");
 902
 903         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 904                 goto nla_put_failure;
 905
 906         nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
 907
 908 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
 909         nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
 910 #else
 911         if(cfg->proto == VLAN_PROTO_8021AD)
 912                 netifd_log_message(L_WARNING, "%s Your kernel is older than linux 3.10.0, 802.1ad is not supported defaulting to 802.1q", vlandev->type->name);
 913 #endif
 914
 915         nla_nest_end(msg, data);
 916         nla_nest_end(msg, linkinfo);
 917
 918         rv = system_rtnl_call(msg);
 919         if (rv)
 920                 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
 921
 922         return rv;
 923
 924 nla_put_failure:
 925         nlmsg_free(msg);
 926         return -ENOMEM;
 927 }
 928
 929 int system_vlandev_del(struct device *vlandev)
 930 {
 931         return system_link_del(vlandev->ifname);
 932 }
 933
 934 static void
 935 system_if_get_settings(struct device *dev, struct device_settings *s)
 936 {
 937         struct ifreq ifr;
 938         char buf[10];
 939
 940         memset(&ifr, 0, sizeof(ifr));
 941         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 942
 943         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 944                 s->mtu = ifr.ifr_mtu;
 945                 s->flags |= DEV_OPT_MTU;
 946         }
 947
 948         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 949                 s->txqueuelen = ifr.ifr_qlen;
 950                 s->flags |= DEV_OPT_TXQUEUELEN;
 951         }
 952
 953         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 954                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 955                 s->flags |= DEV_OPT_MACADDR;
 956         }
 957
 958         if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
 959                 s->ipv6 = !strtoul(buf, NULL, 0);
 960                 s->flags |= DEV_OPT_IPV6;
 961         }
 962
 963         if (ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr) == 0) {
 964                 s->promisc = ifr.ifr_flags & IFF_PROMISC;
 965                 s->flags |= DEV_OPT_PROMISC;
 966         }
 967
 968         if (!system_get_rpfilter(dev, buf, sizeof(buf))) {
 969                 s->rpfilter = strtoul(buf, NULL, 0);
 970                 s->flags |= DEV_OPT_RPFILTER;
 971         }
 972 }
 973
 974 void
 975 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
 976 {
 977         struct ifreq ifr;
 978
 979         if (!apply_mask)
 980                 return;
 981
 982         memset(&ifr, 0, sizeof(ifr));
 983         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 984         if (s->flags & DEV_OPT_MTU & apply_mask) {
 985                 ifr.ifr_mtu = s->mtu;
 986                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 987                         s->flags &= ~DEV_OPT_MTU;
 988         }
 989         if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
 990                 ifr.ifr_qlen = s->txqueuelen;
 991                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 992                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 993         }
 994         if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
 995                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 996                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 997                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 998                         s->flags &= ~DEV_OPT_MACADDR;
 999         }
1000         if (s->flags & DEV_OPT_IPV6 & apply_mask)
1001                 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
1002         if (s->flags & DEV_OPT_PROMISC & apply_mask) {
1003                 if (system_if_flags(dev->ifname, s->promisc ? IFF_PROMISC : 0,
1004                                     !s->promisc ? IFF_PROMISC : 0) < 0)
1005                         s->flags &= ~DEV_OPT_PROMISC;
1006         }
1007         if (s->flags & DEV_OPT_RPFILTER & apply_mask) {
1008                 char buf[2];
1009
1010                 snprintf(buf, sizeof(buf), "%d", s->rpfilter);
1011                 system_set_rpfilter(dev, buf);
1012         }
1013 }
1014
1015 int system_if_up(struct device *dev)
1016 {
1017         system_if_get_settings(dev, &dev->orig_settings);
1018         system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
1019         return system_if_flags(dev->ifname, IFF_UP, 0);
1020 }
1021
1022 int system_if_down(struct device *dev)
1023 {
1024         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
1025         dev->orig_settings.flags &= dev->settings.flags;
1026         system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
1027         return ret;
1028 }
1029
1030 struct if_check_data {
1031         struct device *dev;
1032         int pending;
1033         int ret;
1034 };
1035
1036 #ifndef IFF_LOWER_UP
1037 #define IFF_LOWER_UP    0x10000
1038 #endif
1039
1040 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
1041 {
1042         struct nlmsghdr *nh = nlmsg_hdr(msg);
1043         struct ifinfomsg *ifi = NLMSG_DATA(nh);
1044         struct if_check_data *chk = (struct if_check_data *)arg;
1045
1046         if (nh->nlmsg_type != RTM_NEWLINK)
1047                 return NL_SKIP;
1048
1049         device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1050         device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1051
1052         return NL_OK;
1053 }
1054
1055 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1056 {
1057         struct if_check_data *chk = (struct if_check_data *)arg;
1058         chk->pending = 0;
1059         return NL_STOP;
1060 }
1061
1062 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1063 {
1064         struct if_check_data *chk = (struct if_check_data *)arg;
1065
1066         device_set_present(chk->dev, false);
1067         device_set_link(chk->dev, false);
1068         chk->pending = err->error;
1069
1070         return NL_STOP;
1071 }
1072
1073 int system_if_check(struct device *dev)
1074 {
1075         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1076         struct nl_msg *msg;
1077         struct ifinfomsg ifi = {
1078                 .ifi_family = AF_UNSPEC,
1079                 .ifi_index = 0,
1080         };
1081         struct if_check_data chk = {
1082                 .dev = dev,
1083                 .pending = 1,
1084         };
1085         int ret = 1;
1086
1087         msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1088         if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1089             nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1090                 goto out;
1091
1092         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1093         nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1094         nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1095
1096         nl_send_auto_complete(sock_rtnl, msg);
1097         while (chk.pending > 0)
1098                 nl_recvmsgs(sock_rtnl, cb);
1099
1100         nlmsg_free(msg);
1101         ret = chk.pending;
1102
1103 out:
1104         nl_cb_put(cb);
1105         return ret;
1106 }
1107
1108 struct device *
1109 system_if_get_parent(struct device *dev)
1110 {
1111         char buf[64], *devname;
1112         int ifindex, iflink, len;
1113         FILE *f;
1114
1115         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1116         f = fopen(buf, "r");
1117         if (!f)
1118                 return NULL;
1119
1120         len = fread(buf, 1, sizeof(buf) - 1, f);
1121         fclose(f);
1122
1123         if (len <= 0)
1124                 return NULL;
1125
1126         buf[len] = 0;
1127         iflink = strtoul(buf, NULL, 0);
1128         ifindex = system_if_resolve(dev);
1129         if (!iflink || iflink == ifindex)
1130                 return NULL;
1131
1132         devname = if_indextoname(iflink, buf);
1133         if (!devname)
1134                 return NULL;
1135
1136         return device_get(devname, true);
1137 }
1138
1139 static bool
1140 read_string_file(int dir_fd, const char *file, char *buf, int len)
1141 {
1142         bool ret = false;
1143         char *c;
1144         int fd;
1145
1146         fd = openat(dir_fd, file, O_RDONLY);
1147         if (fd < 0)
1148                 return false;
1149
1150 retry:
1151         len = read(fd, buf, len - 1);
1152         if (len < 0) {
1153                 if (errno == EINTR)
1154                         goto retry;
1155         } else if (len > 0) {
1156                         buf[len] = 0;
1157
1158                         c = strchr(buf, '\n');
1159                         if (c)
1160                                 *c = 0;
1161
1162                         ret = true;
1163         }
1164
1165         close(fd);
1166
1167         return ret;
1168 }
1169
1170 static bool
1171 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1172 {
1173         char buf[64];
1174         bool ret = false;
1175
1176         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1177         if (ret)
1178                 *val = strtoull(buf, NULL, 0);
1179
1180         return ret;
1181 }
1182
1183 /* Assume advertised flags == supported flags */
1184 static const struct {
1185         uint32_t mask;
1186         const char *name;
1187 } ethtool_link_modes[] = {
1188         { ADVERTISED_10baseT_Half, "10H" },
1189         { ADVERTISED_10baseT_Full, "10F" },
1190         { ADVERTISED_100baseT_Half, "100H" },
1191         { ADVERTISED_100baseT_Full, "100F" },
1192         { ADVERTISED_1000baseT_Half, "1000H" },
1193         { ADVERTISED_1000baseT_Full, "1000F" },
1194 };
1195
1196 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1197 {
1198         int i;
1199         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1200                 if (mask & ethtool_link_modes[i].mask)
1201                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1202         }
1203 }
1204
1205 bool
1206 system_if_force_external(const char *ifname)
1207 {
1208         char buf[64];
1209         struct stat s;
1210
1211         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1212         return stat(buf, &s) == 0;
1213 }
1214
1215 int
1216 system_if_dump_info(struct device *dev, struct blob_buf *b)
1217 {
1218         struct ethtool_cmd ecmd;
1219         struct ifreq ifr;
1220         char buf[64], *s;
1221         void *c;
1222         int dir_fd;
1223
1224         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1225         dir_fd = open(buf, O_DIRECTORY);
1226
1227         memset(&ecmd, 0, sizeof(ecmd));
1228         memset(&ifr, 0, sizeof(ifr));
1229         strcpy(ifr.ifr_name, dev->ifname);
1230         ifr.ifr_data = (caddr_t) &ecmd;
1231         ecmd.cmd = ETHTOOL_GSET;
1232
1233         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1234                 c = blobmsg_open_array(b, "link-advertising");
1235                 system_add_link_modes(b, ecmd.advertising);
1236                 blobmsg_close_array(b, c);
1237
1238                 c = blobmsg_open_array(b, "link-supported");
1239                 system_add_link_modes(b, ecmd.supported);
1240                 blobmsg_close_array(b, c);
1241
1242                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1243                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1244                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1245                 blobmsg_add_string_buffer(b);
1246         }
1247
1248         close(dir_fd);
1249         return 0;
1250 }
1251
1252 int
1253 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1254 {
1255         const char *const counters[] = {
1256                 "collisions",     "rx_frame_errors",   "tx_compressed",
1257                 "multicast",      "rx_length_errors",  "tx_dropped",
1258                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
1259                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
1260                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
1261                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
1262                 "rx_errors",      "tx_bytes",          "tx_window_errors",
1263                 "rx_fifo_errors", "tx_carrier_errors",
1264         };
1265         char buf[64];
1266         int stats_dir;
1267         int i;
1268         uint64_t val = 0;
1269
1270         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1271         stats_dir = open(buf, O_DIRECTORY);
1272         if (stats_dir < 0)
1273                 return -1;
1274
1275         for (i = 0; i < ARRAY_SIZE(counters); i++)
1276                 if (read_uint64_file(stats_dir, counters[i], &val))
1277                         blobmsg_add_u64(b, counters[i], val);
1278
1279         close(stats_dir);
1280         return 0;
1281 }
1282
1283 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1284 {
1285         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1286         int alen = v4 ? 4 : 16;
1287         unsigned int flags = 0;
1288         struct ifaddrmsg ifa = {
1289                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1290                 .ifa_prefixlen = addr->mask,
1291                 .ifa_index = dev->ifindex,
1292         };
1293
1294         struct nl_msg *msg;
1295         if (cmd == RTM_NEWADDR)
1296                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1297
1298         msg = nlmsg_alloc_simple(cmd, flags);
1299         if (!msg)
1300                 return -1;
1301
1302         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1303         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1304         if (v4) {
1305                 if (addr->broadcast)
1306                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1307                 if (addr->point_to_point)
1308                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1309         } else {
1310                 time_t now = system_get_rtime();
1311                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1312
1313                 if (addr->preferred_until) {
1314                         int64_t preferred = addr->preferred_until - now;
1315                         if (preferred < 0)
1316                                 preferred = 0;
1317                         else if (preferred > UINT32_MAX)
1318                                 preferred = UINT32_MAX;
1319
1320                         cinfo.ifa_prefered = preferred;
1321                 }
1322
1323                 if (addr->valid_until) {
1324                         int64_t valid = addr->valid_until - now;
1325                         if (valid <= 0)
1326                                 return -1;
1327                         else if (valid > UINT32_MAX)
1328                                 valid = UINT32_MAX;
1329
1330                         cinfo.ifa_valid = valid;
1331                 }
1332
1333                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1334         }
1335
1336         return system_rtnl_call(msg);
1337 }
1338
1339 int system_add_address(struct device *dev, struct device_addr *addr)
1340 {
1341         return system_addr(dev, addr, RTM_NEWADDR);
1342 }
1343
1344 int system_del_address(struct device *dev, struct device_addr *addr)
1345 {
1346         return system_addr(dev, addr, RTM_DELADDR);
1347 }
1348
1349 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1350 {
1351         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1352         bool have_gw;
1353         unsigned int flags = 0;
1354
1355         if (alen == 4)
1356                 have_gw = !!route->nexthop.in.s_addr;
1357         else
1358                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1359                         route->nexthop.in6.s6_addr32[1] ||
1360                         route->nexthop.in6.s6_addr32[2] ||
1361                         route->nexthop.in6.s6_addr32[3];
1362
1363         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1364                         ? route->table : RT_TABLE_MAIN;
1365
1366         struct rtmsg rtm = {
1367                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1368                 .rtm_dst_len = route->mask,
1369                 .rtm_src_len = route->sourcemask,
1370                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1371                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1372                 .rtm_scope = RT_SCOPE_NOWHERE,
1373                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1374                 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1375         };
1376         struct nl_msg *msg;
1377
1378         if (cmd == RTM_NEWROUTE) {
1379                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1380
1381                 if (!dev) { // Add null-route
1382                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1383                         rtm.rtm_type = RTN_UNREACHABLE;
1384                 }
1385                 else
1386                         rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1387         }
1388
1389         if (route->flags & DEVROUTE_TYPE) {
1390                 rtm.rtm_type = route->type;
1391                 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1392                         if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1393                             rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1394                                 rtm.rtm_table = RT_TABLE_LOCAL;
1395                 }
1396
1397                 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT)
1398                         rtm.rtm_scope = RT_SCOPE_HOST;
1399                 else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1400                         rtm.rtm_type == RTN_ANYCAST)
1401                         rtm.rtm_scope = RT_SCOPE_LINK;
1402         }
1403
1404         msg = nlmsg_alloc_simple(cmd, flags);
1405         if (!msg)
1406                 return -1;
1407
1408         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1409
1410         if (route->mask)
1411                 nla_put(msg, RTA_DST, alen, &route->addr);
1412
1413         if (route->sourcemask)
1414                 nla_put(msg, RTA_SRC, alen, &route->source);
1415
1416         if (route->metric > 0)
1417                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1418
1419         if (have_gw)
1420                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1421
1422         if (dev)
1423                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1424
1425         if (table >= 256)
1426                 nla_put_u32(msg, RTA_TABLE, table);
1427
1428         if (route->flags & DEVROUTE_MTU) {
1429                 struct nlattr *metrics;
1430
1431                 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1432                         goto nla_put_failure;
1433
1434                 nla_put_u32(msg, RTAX_MTU, route->mtu);
1435
1436                 nla_nest_end(msg, metrics);
1437         }
1438
1439         return system_rtnl_call(msg);
1440
1441 nla_put_failure:
1442         nlmsg_free(msg);
1443         return -ENOMEM;
1444 }
1445
1446 int system_add_route(struct device *dev, struct device_route *route)
1447 {
1448         return system_rt(dev, route, RTM_NEWROUTE);
1449 }
1450
1451 int system_del_route(struct device *dev, struct device_route *route)
1452 {
1453         return system_rt(dev, route, RTM_DELROUTE);
1454 }
1455
1456 int system_flush_routes(void)
1457 {
1458         const char *names[] = {
1459                 "/proc/sys/net/ipv4/route/flush",
1460                 "/proc/sys/net/ipv6/route/flush"
1461         };
1462         int fd, i;
1463
1464         for (i = 0; i < ARRAY_SIZE(names); i++) {
1465                 fd = open(names[i], O_WRONLY);
1466                 if (fd < 0)
1467                         continue;
1468
1469                 if (write(fd, "-1", 2)) {}
1470                 close(fd);
1471         }
1472         return 0;
1473 }
1474
1475 bool system_resolve_rt_type(const char *type, unsigned int *id)
1476 {
1477         return system_rtn_aton(type, id);
1478 }
1479
1480 bool system_resolve_rt_table(const char *name, unsigned int *id)
1481 {
1482         FILE *f;
1483         char *e, buf[128];
1484         unsigned int n, table = RT_TABLE_UNSPEC;
1485
1486         /* first try to parse table as number */
1487         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1488                 table = n;
1489
1490         /* handle well known aliases */
1491         else if (!strcmp(name, "default"))
1492                 table = RT_TABLE_DEFAULT;
1493         else if (!strcmp(name, "main"))
1494                 table = RT_TABLE_MAIN;
1495         else if (!strcmp(name, "local"))
1496                 table = RT_TABLE_LOCAL;
1497
1498         /* try to look up name in /etc/iproute2/rt_tables */
1499         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1500         {
1501                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1502                 {
1503                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1504                                 continue;
1505
1506                         n = strtoul(e, NULL, 10);
1507                         e = strtok(NULL, " \t\n");
1508
1509                         if (e && !strcmp(e, name))
1510                         {
1511                                 table = n;
1512                                 break;
1513                         }
1514                 }
1515
1516                 fclose(f);
1517         }
1518
1519         if (table == RT_TABLE_UNSPEC)
1520                 return false;
1521
1522         *id = table;
1523         return true;
1524 }
1525
1526 bool system_is_default_rt_table(unsigned int id)
1527 {
1528         return (id == RT_TABLE_MAIN);
1529 }
1530
1531 bool system_resolve_rpfilter(const char *filter, unsigned int *id)
1532 {
1533         char *e;
1534         unsigned int n;
1535
1536         if (!strcmp(filter, "strict"))
1537                 n = 1;
1538         else if (!strcmp(filter, "loose"))
1539                 n = 2;
1540         else {
1541                 n = strtoul(filter, &e, 0);
1542                 if (*e || e == filter || n > 2)
1543                         return false;
1544         }
1545
1546         *id = n;
1547         return true;
1548 }
1549
1550 static int system_iprule(struct iprule *rule, int cmd)
1551 {
1552         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1553
1554         struct nl_msg *msg;
1555         struct rtmsg rtm = {
1556                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1557                 .rtm_protocol = RTPROT_STATIC,
1558                 .rtm_scope = RT_SCOPE_UNIVERSE,
1559                 .rtm_table = RT_TABLE_UNSPEC,
1560                 .rtm_type = RTN_UNSPEC,
1561                 .rtm_flags = 0,
1562         };
1563
1564         if (cmd == RTM_NEWRULE) {
1565                 rtm.rtm_type = RTN_UNICAST;
1566                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1567         }
1568
1569         if (rule->invert)
1570                 rtm.rtm_flags |= FIB_RULE_INVERT;
1571
1572         if (rule->flags & IPRULE_SRC)
1573                 rtm.rtm_src_len = rule->src_mask;
1574
1575         if (rule->flags & IPRULE_DEST)
1576                 rtm.rtm_dst_len = rule->dest_mask;
1577
1578         if (rule->flags & IPRULE_TOS)
1579                 rtm.rtm_tos = rule->tos;
1580
1581         if (rule->flags & IPRULE_LOOKUP) {
1582                 if (rule->lookup < 256)
1583                         rtm.rtm_table = rule->lookup;
1584         }
1585
1586         if (rule->flags & IPRULE_ACTION)
1587                 rtm.rtm_type = rule->action;
1588         else if (rule->flags & IPRULE_GOTO)
1589                 rtm.rtm_type = FR_ACT_GOTO;
1590         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1591                 rtm.rtm_type = FR_ACT_NOP;
1592
1593         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1594
1595         if (!msg)
1596                 return -1;
1597
1598         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1599
1600         if (rule->flags & IPRULE_IN)
1601                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1602
1603         if (rule->flags & IPRULE_OUT)
1604                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1605
1606         if (rule->flags & IPRULE_SRC)
1607                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1608
1609         if (rule->flags & IPRULE_DEST)
1610                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1611
1612         if (rule->flags & IPRULE_PRIORITY)
1613                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1614         else if (cmd == RTM_NEWRULE)
1615                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1616
1617         if (rule->flags & IPRULE_FWMARK)
1618                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1619
1620         if (rule->flags & IPRULE_FWMASK)
1621                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1622
1623         if (rule->flags & IPRULE_LOOKUP) {
1624                 if (rule->lookup >= 256)
1625                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1626         }
1627
1628         if (rule->flags & IPRULE_GOTO)
1629                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1630
1631         return system_rtnl_call(msg);
1632 }
1633
1634 int system_add_iprule(struct iprule *rule)
1635 {
1636         return system_iprule(rule, RTM_NEWRULE);
1637 }
1638
1639 int system_del_iprule(struct iprule *rule)
1640 {
1641         return system_iprule(rule, RTM_DELRULE);
1642 }
1643
1644 int system_flush_iprules(void)
1645 {
1646         int rv = 0;
1647         struct iprule rule;
1648
1649         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1650         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1651
1652         memset(&rule, 0, sizeof(rule));
1653
1654
1655         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1656
1657         rule.priority = 0;
1658         rule.lookup = RT_TABLE_LOCAL;
1659         rv |= system_iprule(&rule, RTM_NEWRULE);
1660
1661         rule.priority = 32766;
1662         rule.lookup = RT_TABLE_MAIN;
1663         rv |= system_iprule(&rule, RTM_NEWRULE);
1664
1665         rule.priority = 32767;
1666         rule.lookup = RT_TABLE_DEFAULT;
1667         rv |= system_iprule(&rule, RTM_NEWRULE);
1668
1669
1670         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1671
1672         rule.priority = 0;
1673         rule.lookup = RT_TABLE_LOCAL;
1674         rv |= system_iprule(&rule, RTM_NEWRULE);
1675
1676         rule.priority = 32766;
1677         rule.lookup = RT_TABLE_MAIN;
1678         rv |= system_iprule(&rule, RTM_NEWRULE);
1679
1680         return rv;
1681 }
1682
1683 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1684 {
1685         return system_rtn_aton(action, id);
1686 }
1687
1688 time_t system_get_rtime(void)
1689 {
1690         struct timespec ts;
1691         struct timeval tv;
1692
1693         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1694                 return ts.tv_sec;
1695
1696         if (gettimeofday(&tv, NULL) == 0)
1697                 return tv.tv_sec;
1698
1699         return 0;
1700 }
1701
1702 #ifndef IP_DF
1703 #define IP_DF       0x4000
1704 #endif
1705
1706 static int tunnel_ioctl(const char *name, int cmd, void *p)
1707 {
1708         struct ifreq ifr;
1709
1710         memset(&ifr, 0, sizeof(ifr));
1711         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1712         ifr.ifr_ifru.ifru_data = p;
1713         return ioctl(sock_ioctl, cmd, &ifr);
1714 }
1715
1716 #ifdef IFLA_IPTUN_MAX
1717 #define IP6_FLOWINFO_TCLASS     htonl(0x0FF00000)
1718 static int system_add_gre_tunnel(const char *name, const char *kind,
1719                                  const unsigned int link, struct blob_attr **tb, bool v6)
1720 {
1721         struct nl_msg *nlm;
1722         struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC, };
1723         struct blob_attr *cur;
1724         uint32_t ikey = 0, okey = 0, flags = 0, flowinfo = 0;
1725         uint16_t iflags = 0, oflags = 0;
1726         uint8_t tos = 0;
1727         int ret = 0, ttl = 64;
1728
1729         nlm = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1730         if (!nlm)
1731                 return -1;
1732
1733         nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1734         nla_put_string(nlm, IFLA_IFNAME, name);
1735
1736         struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1737         if (!linkinfo) {
1738                 ret = -ENOMEM;
1739                 goto failure;
1740         }
1741
1742         nla_put_string(nlm, IFLA_INFO_KIND, kind);
1743         struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1744         if (!infodata) {
1745                 ret = -ENOMEM;
1746                 goto failure;
1747         }
1748
1749         if (link)
1750                 nla_put_u32(nlm, IFLA_GRE_LINK, link);
1751
1752         if ((cur = tb[TUNNEL_ATTR_TTL]))
1753                 ttl = blobmsg_get_u32(cur);
1754
1755         nla_put_u8(nlm, IFLA_GRE_TTL, ttl);
1756
1757         if ((cur = tb[TUNNEL_ATTR_TOS])) {
1758                 char *str = blobmsg_get_string(cur);
1759                 if (strcmp(str, "inherit")) {
1760                         unsigned uval;
1761
1762                         if (!system_tos_aton(str, &uval)) {
1763                                 ret = -EINVAL;
1764                                 goto failure;
1765                         }
1766
1767                         if (v6)
1768                                 flowinfo |= htonl(uval << 20) & IP6_FLOWINFO_TCLASS;
1769                         else
1770                                 tos = uval;
1771                 } else {
1772                         if (v6)
1773                                 flags |= IP6_TNL_F_USE_ORIG_TCLASS;
1774                         else
1775                                 tos = 1;
1776                 }
1777         }
1778
1779         if ((cur = tb[TUNNEL_ATTR_INFO]) && (blobmsg_type(cur) == BLOBMSG_TYPE_STRING)) {
1780                 uint8_t icsum, ocsum, iseqno, oseqno;
1781                 if (sscanf(blobmsg_get_string(cur), "%u,%u,%hhu,%hhu,%hhu,%hhu",
1782                         &ikey, &okey, &icsum, &ocsum, &iseqno, &oseqno) < 6) {
1783                         ret = -EINVAL;
1784                         goto failure;
1785                 }
1786
1787                 if (ikey)
1788                         iflags |= GRE_KEY;
1789
1790                 if (okey)
1791                         oflags |= GRE_KEY;
1792
1793                 if (icsum)
1794                         iflags |= GRE_CSUM;
1795
1796                 if (ocsum)
1797                         oflags |= GRE_CSUM;
1798
1799                 if (iseqno)
1800                         iflags |= GRE_SEQ;
1801
1802                 if (oseqno)
1803                         oflags |= GRE_SEQ;
1804         }
1805
1806         if (v6) {
1807                 struct in6_addr in6buf;
1808                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1809                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1810                                 ret = -EINVAL;
1811                                 goto failure;
1812                         }
1813                         nla_put(nlm, IFLA_GRE_LOCAL, sizeof(in6buf), &in6buf);
1814                 }
1815
1816                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1817                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1818                                 ret = -EINVAL;
1819                                 goto failure;
1820                         }
1821                         nla_put(nlm, IFLA_GRE_REMOTE, sizeof(in6buf), &in6buf);
1822                 }
1823                 nla_put_u8(nlm, IFLA_GRE_ENCAP_LIMIT, 4);
1824
1825                 if (flowinfo)
1826                         nla_put_u32(nlm, IFLA_GRE_FLOWINFO, flowinfo);
1827
1828                 if (flags)
1829                         nla_put_u32(nlm, IFLA_GRE_FLAGS, flags);
1830         } else {
1831                 struct in_addr inbuf;
1832                 bool set_df = true;
1833
1834                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1835                         if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
1836                                 ret = -EINVAL;
1837                                 goto failure;
1838                         }
1839                         nla_put(nlm, IFLA_GRE_LOCAL, sizeof(inbuf), &inbuf);
1840                 }
1841
1842                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1843                         if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
1844                                 ret = -EINVAL;
1845                                 goto failure;
1846                         }
1847                         nla_put(nlm, IFLA_GRE_REMOTE, sizeof(inbuf), &inbuf);
1848
1849                         if (IN_MULTICAST(ntohl(inbuf.s_addr))) {
1850                                 if (!okey) {
1851                                         okey = inbuf.s_addr;
1852                                         oflags |= GRE_KEY;
1853                                 }
1854
1855                                 if (!ikey) {
1856                                         ikey = inbuf.s_addr;
1857                                         iflags |= GRE_KEY;
1858                                 }
1859                         }
1860                 }
1861
1862                 if ((cur = tb[TUNNEL_ATTR_DF]))
1863                         set_df = blobmsg_get_bool(cur);
1864
1865                 /* ttl !=0 and nopmtudisc are incompatible */
1866                 if (ttl && !set_df) {
1867                         ret = -EINVAL;
1868                         goto failure;
1869                 }
1870
1871                 nla_put_u8(nlm, IFLA_GRE_PMTUDISC, set_df ? 1 : 0);
1872
1873                 nla_put_u8(nlm, IFLA_GRE_TOS, tos);
1874         }
1875
1876         if (oflags)
1877                 nla_put_u16(nlm, IFLA_GRE_OFLAGS, oflags);
1878
1879         if (iflags)
1880                 nla_put_u16(nlm, IFLA_GRE_IFLAGS, iflags);
1881
1882         if (okey)
1883                 nla_put_u32(nlm, IFLA_GRE_OKEY, okey);
1884
1885         if (ikey)
1886                 nla_put_u32(nlm, IFLA_GRE_IKEY, ikey);
1887
1888         nla_nest_end(nlm, infodata);
1889         nla_nest_end(nlm, linkinfo);
1890
1891         return system_rtnl_call(nlm);
1892
1893 failure:
1894         nlmsg_free(nlm);
1895         return ret;
1896 }
1897 #endif
1898
1899 static int system_add_proto_tunnel(const char *name, const uint8_t proto, const unsigned int link, struct blob_attr **tb)
1900 {
1901         struct blob_attr *cur;
1902         bool set_df = true;
1903         struct ip_tunnel_parm p  = {
1904                 .link = link,
1905                 .iph = {
1906                         .version = 4,
1907                         .ihl = 5,
1908                         .protocol = proto,
1909                 }
1910         };
1911
1912         if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1913                         inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1914                 return -EINVAL;
1915
1916         if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1917                         inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1918                 return -EINVAL;
1919
1920         if ((cur = tb[TUNNEL_ATTR_DF]))
1921                 set_df = blobmsg_get_bool(cur);
1922
1923         if ((cur = tb[TUNNEL_ATTR_TTL]))
1924                 p.iph.ttl = blobmsg_get_u32(cur);
1925
1926         if ((cur = tb[TUNNEL_ATTR_TOS])) {
1927                 char *str = blobmsg_get_string(cur);
1928                 if (strcmp(str, "inherit")) {
1929                         unsigned uval;
1930
1931                         if (!system_tos_aton(str, &uval))
1932                                 return -EINVAL;
1933
1934                         p.iph.tos = uval;
1935                 } else
1936                         p.iph.tos = 1;
1937         }
1938
1939         p.iph.frag_off = set_df ? htons(IP_DF) : 0;
1940         /* ttl !=0 and nopmtudisc are incompatible */
1941         if (p.iph.ttl && p.iph.frag_off == 0)
1942                 return -EINVAL;
1943
1944         strncpy(p.name, name, sizeof(p.name));
1945
1946         switch (p.iph.protocol) {
1947         case IPPROTO_IPIP:
1948                 return tunnel_ioctl("tunl0", SIOCADDTUNNEL, &p);
1949         case IPPROTO_IPV6:
1950                 return tunnel_ioctl("sit0", SIOCADDTUNNEL, &p);
1951         default:
1952                 break;
1953         }
1954         return -1;
1955 }
1956
1957 static int __system_del_ip_tunnel(const char *name, struct blob_attr **tb)
1958 {
1959         struct blob_attr *cur;
1960         const char *str;
1961
1962         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1963                 return -EINVAL;
1964         str = blobmsg_data(cur);
1965
1966         if (!strcmp(str, "greip") || !strcmp(str, "gretapip") ||
1967             !strcmp(str, "greip6") || !strcmp(str, "gretapip6"))
1968                 return system_link_del(name);
1969         else
1970                 return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1971 }
1972
1973 int system_del_ip_tunnel(const char *name, struct blob_attr *attr)
1974 {
1975         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1976
1977         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1978                 blob_data(attr), blob_len(attr));
1979
1980         return __system_del_ip_tunnel(name, tb);
1981 }
1982
1983 int system_update_ipv6_mtu(struct device *dev, int mtu)
1984 {
1985         int ret = -1;
1986         char buf[64];
1987         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1988                         dev->ifname);
1989
1990         int fd = open(buf, O_RDWR);
1991         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1992         if (len < 0)
1993                 goto out;
1994
1995         buf[len] = 0;
1996         ret = atoi(buf);
1997
1998         if (!mtu || ret <= mtu)
1999                 goto out;
2000
2001         lseek(fd, 0, SEEK_SET);
2002         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
2003                 ret = -1;
2004
2005 out:
2006         close(fd);
2007         return ret;
2008 }
2009
2010 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
2011 {
2012         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2013         struct blob_attr *cur;
2014         const char *str;
2015
2016         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2017                 blob_data(attr), blob_len(attr));
2018
2019         __system_del_ip_tunnel(name, tb);
2020
2021         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2022                 return -EINVAL;
2023         str = blobmsg_data(cur);
2024
2025         unsigned int ttl = 0;
2026         if ((cur = tb[TUNNEL_ATTR_TTL])) {
2027                 ttl = blobmsg_get_u32(cur);
2028                 if (ttl > 255)
2029                         return -EINVAL;
2030         }
2031
2032         unsigned int link = 0;
2033         if ((cur = tb[TUNNEL_ATTR_LINK])) {
2034                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
2035                 if (!iface)
2036                         return -EINVAL;
2037
2038                 if (iface->l3_dev.dev)
2039                         link = iface->l3_dev.dev->ifindex;
2040         }
2041
2042         if (!strcmp(str, "sit")) {
2043                 if (system_add_proto_tunnel(name, IPPROTO_IPV6, link, tb) < 0)
2044                         return -1;
2045
2046 #ifdef SIOCADD6RD
2047                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
2048                         unsigned int mask;
2049                         struct ip_tunnel_6rd p6;
2050
2051                         memset(&p6, 0, sizeof(p6));
2052
2053                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
2054                                                 &p6.prefix, &mask) || mask > 128)
2055                                 return -EINVAL;
2056                         p6.prefixlen = mask;
2057
2058                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
2059                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
2060                                                         &p6.relay_prefix, &mask) || mask > 32)
2061                                         return -EINVAL;
2062                                 p6.relay_prefixlen = mask;
2063                         }
2064
2065                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
2066                                 __system_del_ip_tunnel(name, tb);
2067                                 return -1;
2068                         }
2069                 }
2070 #endif
2071 #ifdef IFLA_IPTUN_MAX
2072         } else if (!strcmp(str, "ipip6")) {
2073                 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
2074                                 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
2075                 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
2076                 int ret = 0;
2077
2078                 if (!nlm)
2079                         return -1;
2080
2081                 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
2082                 nla_put_string(nlm, IFLA_IFNAME, name);
2083
2084                 if (link)
2085                         nla_put_u32(nlm, IFLA_LINK, link);
2086
2087                 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
2088                 if (!linkinfo) {
2089                         ret = -ENOMEM;
2090                         goto failure;
2091                 }
2092                 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
2093                 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
2094                 if (!infodata) {
2095                         ret = -ENOMEM;
2096                         goto failure;
2097                 }
2098
2099                 if (link)
2100                         nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
2101
2102                 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
2103                 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
2104                 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
2105
2106                 struct in6_addr in6buf;
2107                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2108                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2109                                 ret = -EINVAL;
2110                                 goto failure;
2111                         }
2112                         nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
2113                 }
2114
2115                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2116                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2117                                 ret = -EINVAL;
2118                                 goto failure;
2119                         }
2120                         nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
2121                 }
2122
2123 #ifdef IFLA_IPTUN_FMR_MAX
2124                 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
2125                         struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
2126
2127                         struct blob_attr *fmr;
2128                         unsigned rem, fmrcnt = 0;
2129                         blobmsg_for_each_attr(fmr, cur, rem) {
2130                                 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
2131                                         continue;
2132
2133                                 unsigned ip4len, ip6len, ealen, offset = 6;
2134                                 char ip6buf[48];
2135                                 char ip4buf[16];
2136
2137                                 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
2138                                                 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
2139                                         ret = -EINVAL;
2140                                         goto failure;
2141                                 }
2142
2143                                 struct in6_addr ip6prefix;
2144                                 struct in_addr ip4prefix;
2145                                 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
2146                                                 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
2147                                         ret = -EINVAL;
2148                                         goto failure;
2149                                 }
2150
2151                                 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
2152
2153                                 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
2154                                 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
2155                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
2156                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
2157                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
2158                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
2159
2160                                 nla_nest_end(nlm, rule);
2161                         }
2162
2163                         nla_nest_end(nlm, fmrs);
2164                 }
2165 #endif
2166
2167                 nla_nest_end(nlm, infodata);
2168                 nla_nest_end(nlm, linkinfo);
2169
2170                 return system_rtnl_call(nlm);
2171 failure:
2172                 nlmsg_free(nlm);
2173                 return ret;
2174         } else if (!strcmp(str, "greip")) {
2175                 return system_add_gre_tunnel(name, "gre", link, tb, false);
2176         } else if (!strcmp(str, "gretapip"))  {
2177                 return system_add_gre_tunnel(name, "gretap", link, tb, false);
2178         } else if (!strcmp(str, "greip6")) {
2179                 return system_add_gre_tunnel(name, "ip6gre", link, tb, true);
2180         } else if (!strcmp(str, "gretapip6")) {
2181                 return system_add_gre_tunnel(name, "ip6gretap", link, tb, true);
2182 #endif
2183         } else if (!strcmp(str, "ipip")) {
2184                 return system_add_proto_tunnel(name, IPPROTO_IPIP, link, tb);
2185         }
2186         else
2187                 return -EINVAL;
2188
2189         return 0;
2190 }