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