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