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