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