system-linux.c

   1 /*
   2  * netifd - network interface daemon
   3  * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
   4  * Copyright (C) 2013 Jo-Philipp Wich <jow@openwrt.org>
   5  * Copyright (C) 2013 Steven Barth <steven@midlink.org>
   6  * Copyright (C) 2014 Gioacchino Mazzurco <gio@eigenlab.org>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2
  10  * as published by the Free Software Foundation
  11  *
  12  * This program is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  */
  17 #define _GNU_SOURCE
  18
  19 #include <sys/socket.h>
  20 #include <sys/ioctl.h>
  21 #include <sys/stat.h>
  22 #include <sys/syscall.h>
  23
  24 #include <net/if.h>
  25 #include <net/if_arp.h>
  26
  27 #include <arpa/inet.h>
  28 #include <netinet/in.h>
  29
  30 #include <linux/rtnetlink.h>
  31 #include <linux/sockios.h>
  32 #include <linux/ip.h>
  33 #include <linux/if_link.h>
  34 #include <linux/if_vlan.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_tunnel.h>
  37 #include <linux/ip6_tunnel.h>
  38 #include <linux/ethtool.h>
  39 #include <linux/fib_rules.h>
  40 #include <linux/version.h>
  41
  42 #ifndef RTN_FAILED_POLICY
  43 #define RTN_FAILED_POLICY 12
  44 #endif
  45
  46 #include <string.h>
  47 #include <fcntl.h>
  48 #include <glob.h>
  49 #include <time.h>
  50
  51 #include <netlink/msg.h>
  52 #include <netlink/attr.h>
  53 #include <netlink/socket.h>
  54 #include <libubox/uloop.h>
  55
  56 #include "netifd.h"
  57 #include "device.h"
  58 #include "system.h"
  59
  60 struct event_socket {
  61         struct uloop_fd uloop;
  62         struct nl_sock *sock;
  63         int bufsize;
  64 };
  65
  66 static int sock_ioctl = -1;
  67 static struct nl_sock *sock_rtnl = NULL;
  68
  69 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
  70 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
  71
  72 static char dev_buf[256];
  73
  74 static void
  75 handler_nl_event(struct uloop_fd *u, unsigned int events)
  76 {
  77         struct event_socket *ev = container_of(u, struct event_socket, uloop);
  78         int err;
  79         socklen_t errlen = sizeof(err);
  80
  81         if (!u->error) {
  82                 nl_recvmsgs_default(ev->sock);
  83                 return;
  84         }
  85
  86         if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
  87                 goto abort;
  88
  89         switch(err) {
  90         case ENOBUFS:
  91                 // Increase rx buffer size on netlink socket
  92                 ev->bufsize *= 2;
  93                 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
  94                         goto abort;
  95
  96                 // Request full dump since some info got dropped
  97                 struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
  98                 nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
  99                 break;
 100
 101         default:
 102                 goto abort;
 103         }
 104         u->error = false;
 105         return;
 106
 107 abort:
 108         uloop_fd_delete(&ev->uloop);
 109         return;
 110 }
 111
 112 static struct nl_sock *
 113 create_socket(int protocol, int groups)
 114 {
 115         struct nl_sock *sock;
 116
 117         sock = nl_socket_alloc();
 118         if (!sock)
 119                 return NULL;
 120
 121         if (groups)
 122                 nl_join_groups(sock, groups);
 123
 124         if (nl_connect(sock, protocol))
 125                 return NULL;
 126
 127         return sock;
 128 }
 129
 130 static bool
 131 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
 132                         uloop_fd_handler cb, int flags)
 133 {
 134         ev->sock = create_socket(protocol, groups);
 135         if (!ev->sock)
 136                 return false;
 137
 138         ev->uloop.fd = nl_socket_get_fd(ev->sock);
 139         ev->uloop.cb = cb;
 140         if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
 141                 return false;
 142
 143         return true;
 144 }
 145
 146 static bool
 147 create_event_socket(struct event_socket *ev, int protocol,
 148                     int (*cb)(struct nl_msg *msg, void *arg))
 149 {
 150         if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
 151                 return false;
 152
 153         // Install the valid custom callback handler
 154         nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
 155
 156         // Disable sequence number checking on event sockets
 157         nl_socket_disable_seq_check(ev->sock);
 158
 159         // Increase rx buffer size to 65K on event sockets
 160         ev->bufsize = 65535;
 161         if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
 162                 return false;
 163
 164         return true;
 165 }
 166
 167 static bool
 168 system_rtn_aton(const char *src, unsigned int *dst)
 169 {
 170         char *e;
 171         unsigned int n;
 172
 173         if (!strcmp(src, "local"))
 174                 n = RTN_LOCAL;
 175         else if (!strcmp(src, "nat"))
 176                 n = RTN_NAT;
 177         else if (!strcmp(src, "broadcast"))
 178                 n = RTN_BROADCAST;
 179         else if (!strcmp(src, "anycast"))
 180                 n = RTN_ANYCAST;
 181         else if (!strcmp(src, "multicast"))
 182                 n = RTN_MULTICAST;
 183         else if (!strcmp(src, "prohibit"))
 184                 n = RTN_PROHIBIT;
 185         else if (!strcmp(src, "unreachable"))
 186                 n = RTN_UNREACHABLE;
 187         else if (!strcmp(src, "blackhole"))
 188                 n = RTN_BLACKHOLE;
 189         else if (!strcmp(src, "xresolve"))
 190                 n = RTN_XRESOLVE;
 191         else if (!strcmp(src, "unicast"))
 192                 n = RTN_UNICAST;
 193         else if (!strcmp(src, "throw"))
 194                 n = RTN_THROW;
 195         else if (!strcmp(src, "failed_policy"))
 196                 n = RTN_FAILED_POLICY;
 197         else {
 198                 n = strtoul(src, &e, 0);
 199                 if (!e || *e || e == src || n > 255)
 200                         return false;
 201         }
 202
 203         *dst = n;
 204         return true;
 205 }
 206
 207 static bool
 208 system_tos_aton(const char *src, unsigned *dst)
 209 {
 210         char *e;
 211
 212         *dst = strtoul(src, &e, 16);
 213         if (e == src || *e || *dst > 255)
 214                 return false;
 215
 216         return true;
 217 }
 218
 219 int system_init(void)
 220 {
 221         static struct event_socket rtnl_event;
 222         static struct event_socket hotplug_event;
 223
 224         sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
 225         system_fd_set_cloexec(sock_ioctl);
 226
 227         // Prepare socket for routing / address control
 228         sock_rtnl = create_socket(NETLINK_ROUTE, 0);
 229         if (!sock_rtnl)
 230                 return -1;
 231
 232         if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
 233                 return -1;
 234
 235         if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
 236                                      handle_hotplug_event, 0))
 237                 return -1;
 238
 239         // Receive network link events form kernel
 240         nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
 241
 242         return 0;
 243 }
 244
 245 static void system_set_sysctl(const char *path, const char *val)
 246 {
 247         int fd;
 248
 249         fd = open(path, O_WRONLY);
 250         if (fd < 0)
 251                 return;
 252
 253         if (write(fd, val, strlen(val))) {}
 254         close(fd);
 255 }
 256
 257 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
 258 {
 259         snprintf(dev_buf, sizeof(dev_buf), path, device);
 260         system_set_sysctl(dev_buf, val);
 261 }
 262
 263 static void system_set_disable_ipv6(struct device *dev, const char *val)
 264 {
 265         system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
 266 }
 267
 268 static void system_set_rpfilter(struct device *dev, const char *val)
 269 {
 270         system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter", dev->ifname, val);
 271 }
 272
 273 static void system_set_acceptlocal(struct device *dev, const char *val)
 274 {
 275         system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/accept_local", dev->ifname, val);
 276 }
 277
 278 static void system_set_igmpversion(struct device *dev, const char *val)
 279 {
 280         system_set_dev_sysctl("/proc/sys/net/ipv4/conf/%s/force_igmp_version", dev->ifname, val);
 281 }
 282
 283 static void system_set_mldversion(struct device *dev, const char *val)
 284 {
 285         system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/force_mld_version", dev->ifname, val);
 286 }
 287
 288 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
 289 {
 290         int fd = -1, ret = -1;
 291
 292         fd = open(path, O_RDONLY);
 293         if (fd < 0)
 294                 goto out;
 295
 296         ssize_t len = read(fd, buf, buf_sz - 1);
 297         if (len < 0)
 298                 goto out;
 299
 300         ret = buf[len] = 0;
 301
 302 out:
 303         if (fd >= 0)
 304                 close(fd);
 305
 306         return ret;
 307 }
 308
 309 static int
 310 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
 311 {
 312         snprintf(dev_buf, sizeof(dev_buf), path, device);
 313         return system_get_sysctl(dev_buf, buf, buf_sz);
 314 }
 315
 316 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
 317 {
 318         return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
 319                         dev->ifname, buf, buf_sz);
 320 }
 321
 322 static int system_get_rpfilter(struct device *dev, char *buf, const size_t buf_sz)
 323 {
 324         return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/rp_filter",
 325                         dev->ifname, buf, buf_sz);
 326 }
 327
 328 static int system_get_acceptlocal(struct device *dev, char *buf, const size_t buf_sz)
 329 {
 330         return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/accept_local",
 331                         dev->ifname, buf, buf_sz);
 332 }
 333
 334 static int system_get_igmpversion(struct device *dev, char *buf, const size_t buf_sz)
 335 {
 336         return system_get_dev_sysctl("/proc/sys/net/ipv4/conf/%s/force_igmp_version",
 337                         dev->ifname, buf, buf_sz);
 338 }
 339
 340 static int system_get_mldversion(struct device *dev, char *buf, const size_t buf_sz)
 341 {
 342         return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/force_mld_version",
 343                         dev->ifname, buf, buf_sz);
 344 }
 345
 346 // Evaluate netlink messages
 347 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 348 {
 349         struct nlmsghdr *nh = nlmsg_hdr(msg);
 350         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 351         struct nlattr *nla[__IFLA_MAX];
 352         int link_state = 0;
 353         char buf[10];
 354
 355         if (nh->nlmsg_type != RTM_NEWLINK)
 356                 goto out;
 357
 358         nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
 359         if (!nla[IFLA_IFNAME])
 360                 goto out;
 361
 362         struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
 363         if (!dev || dev->type->keep_link_status)
 364                 goto out;
 365
 366         if (!system_get_dev_sysctl("/sys/class/net/%s/carrier", dev->ifname, buf, sizeof(buf)))
 367                 link_state = strtoul(buf, NULL, 0);
 368
 369         device_set_link(dev, link_state ? true : false);
 370
 371 out:
 372         return 0;
 373 }
 374
 375 static void
 376 handle_hotplug_msg(char *data, int size)
 377 {
 378         const char *subsystem = NULL, *interface = NULL;
 379         char *cur, *end, *sep;
 380         struct device *dev;
 381         int skip;
 382         bool add;
 383
 384         if (!strncmp(data, "add@", 4))
 385                 add = true;
 386         else if (!strncmp(data, "remove@", 7))
 387                 add = false;
 388         else
 389                 return;
 390
 391         skip = strlen(data) + 1;
 392         end = data + size;
 393
 394         for (cur = data + skip; cur < end; cur += skip) {
 395                 skip = strlen(cur) + 1;
 396
 397                 sep = strchr(cur, '=');
 398                 if (!sep)
 399                         continue;
 400
 401                 *sep = 0;
 402                 if (!strcmp(cur, "INTERFACE"))
 403                         interface = sep + 1;
 404                 else if (!strcmp(cur, "SUBSYSTEM")) {
 405                         subsystem = sep + 1;
 406                         if (strcmp(subsystem, "net") != 0)
 407                                 return;
 408                 }
 409                 if (subsystem && interface)
 410                         goto found;
 411         }
 412         return;
 413
 414 found:
 415         dev = device_get(interface, false);
 416         if (!dev)
 417                 return;
 418
 419         if (dev->type != &simple_device_type)
 420                 return;
 421
 422         if (add && system_if_force_external(dev->ifname))
 423                 return;
 424
 425         device_set_present(dev, add);
 426 }
 427
 428 static void
 429 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
 430 {
 431         struct event_socket *ev = container_of(u, struct event_socket, uloop);
 432         struct sockaddr_nl nla;
 433         unsigned char *buf = NULL;
 434         int size;
 435
 436         while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
 437                 if (nla.nl_pid == 0)
 438                         handle_hotplug_msg((char *) buf, size);
 439
 440                 free(buf);
 441         }
 442 }
 443
 444 static int system_rtnl_call(struct nl_msg *msg)
 445 {
 446         int ret;
 447
 448         ret = nl_send_auto_complete(sock_rtnl, msg);
 449         nlmsg_free(msg);
 450
 451         if (ret < 0)
 452                 return ret;
 453
 454         return nl_wait_for_ack(sock_rtnl);
 455 }
 456
 457 int system_bridge_delbr(struct device *bridge)
 458 {
 459         return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 460 }
 461
 462 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 463 {
 464         struct ifreq ifr;
 465
 466         memset(&ifr, 0, sizeof(ifr));
 467         if (dev)
 468                 ifr.ifr_ifindex = dev->ifindex;
 469         else
 470                 ifr.ifr_data = data;
 471         strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
 472         return ioctl(sock_ioctl, cmd, &ifr);
 473 }
 474
 475 static bool system_is_bridge(const char *name, char *buf, int buflen)
 476 {
 477         struct stat st;
 478
 479         snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
 480         if (stat(buf, &st) < 0)
 481                 return false;
 482
 483         return true;
 484 }
 485
 486 static char *system_get_bridge(const char *name, char *buf, int buflen)
 487 {
 488         char *path;
 489         ssize_t len = -1;
 490         glob_t gl;
 491
 492         snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
 493         if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
 494                 return NULL;
 495
 496         if (gl.gl_pathc > 0)
 497                 len = readlink(gl.gl_pathv[0], buf, buflen);
 498
 499         globfree(&gl);
 500
 501         if (len < 0)
 502                 return NULL;
 503
 504         buf[len] = 0;
 505         path = strrchr(buf, '/');
 506         if (!path)
 507                 return NULL;
 508
 509         return path + 1;
 510 }
 511
 512 static void system_bridge_set_wireless(const char *bridge, const char *dev)
 513 {
 514         snprintf(dev_buf, sizeof(dev_buf),
 515                  "/sys/devices/virtual/net/%s/brif/%s/multicast_to_unicast",
 516                  bridge, dev);
 517         system_set_sysctl(dev_buf, "1");
 518 }
 519
 520 int system_bridge_addif(struct device *bridge, struct device *dev)
 521 {
 522         char *oldbr;
 523         int ret = 0;
 524
 525         oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
 526         if (!oldbr || strcmp(oldbr, bridge->ifname) != 0)
 527                 ret = system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 528
 529         if (dev->wireless)
 530                 system_bridge_set_wireless(bridge->ifname, dev->ifname);
 531
 532         return ret;
 533 }
 534
 535 int system_bridge_delif(struct device *bridge, struct device *dev)
 536 {
 537         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 538 }
 539
 540 int system_if_resolve(struct device *dev)
 541 {
 542         struct ifreq ifr;
 543         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 544         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 545                 return ifr.ifr_ifindex;
 546         else
 547                 return 0;
 548 }
 549
 550 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 551 {
 552         struct ifreq ifr;
 553
 554         memset(&ifr, 0, sizeof(ifr));
 555         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 556         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 557         ifr.ifr_flags |= add;
 558         ifr.ifr_flags &= ~rem;
 559         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 560 }
 561
 562 struct clear_data {
 563         struct nl_msg *msg;
 564         struct device *dev;
 565         int type;
 566         int size;
 567         int af;
 568 };
 569
 570
 571 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 572 {
 573         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 574
 575         return ifa->ifa_index == ifindex;
 576 }
 577
 578 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 579 {
 580         struct rtmsg *r = NLMSG_DATA(hdr);
 581         struct nlattr *tb[__RTA_MAX];
 582
 583         if (r->rtm_protocol == RTPROT_KERNEL &&
 584             r->rtm_family == AF_INET6)
 585                 return false;
 586
 587         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 588         if (!tb[RTA_OIF])
 589                 return false;
 590
 591         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 592 }
 593
 594 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
 595 {
 596         return true;
 597 }
 598
 599 static int cb_clear_event(struct nl_msg *msg, void *arg)
 600 {
 601         struct clear_data *clr = arg;
 602         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 603         bool (*cb)(struct nlmsghdr *, int ifindex);
 604         int type;
 605
 606         switch(clr->type) {
 607         case RTM_GETADDR:
 608                 type = RTM_DELADDR;
 609                 if (hdr->nlmsg_type != RTM_NEWADDR)
 610                         return NL_SKIP;
 611
 612                 cb = check_ifaddr;
 613                 break;
 614         case RTM_GETROUTE:
 615                 type = RTM_DELROUTE;
 616                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 617                         return NL_SKIP;
 618
 619                 cb = check_route;
 620                 break;
 621         case RTM_GETRULE:
 622                 type = RTM_DELRULE;
 623                 if (hdr->nlmsg_type != RTM_NEWRULE)
 624                         return NL_SKIP;
 625
 626                 cb = check_rule;
 627                 break;
 628         default:
 629                 return NL_SKIP;
 630         }
 631
 632         if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
 633                 return NL_SKIP;
 634
 635         if (type == RTM_DELRULE)
 636                 D(SYSTEM, "Remove a rule\n");
 637         else
 638                 D(SYSTEM, "Remove %s from device %s\n",
 639                   type == RTM_DELADDR ? "an address" : "a route",
 640                   clr->dev->ifname);
 641         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 642         hdr = nlmsg_hdr(clr->msg);
 643         hdr->nlmsg_type = type;
 644         hdr->nlmsg_flags = NLM_F_REQUEST;
 645
 646         nl_socket_disable_auto_ack(sock_rtnl);
 647         nl_send_auto_complete(sock_rtnl, clr->msg);
 648         nl_socket_enable_auto_ack(sock_rtnl);
 649
 650         return NL_SKIP;
 651 }
 652
 653 static int
 654 cb_finish_event(struct nl_msg *msg, void *arg)
 655 {
 656         int *pending = arg;
 657         *pending = 0;
 658         return NL_STOP;
 659 }
 660
 661 static int
 662 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 663 {
 664         int *pending = arg;
 665         *pending = err->error;
 666         return NL_STOP;
 667 }
 668
 669 static void
 670 system_if_clear_entries(struct device *dev, int type, int af)
 671 {
 672         struct clear_data clr;
 673         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 674         struct rtmsg rtm = {
 675                 .rtm_family = af,
 676                 .rtm_flags = RTM_F_CLONED,
 677         };
 678         int flags = NLM_F_DUMP;
 679         int pending = 1;
 680
 681         clr.af = af;
 682         clr.dev = dev;
 683         clr.type = type;
 684         switch (type) {
 685         case RTM_GETADDR:
 686         case RTM_GETRULE:
 687                 clr.size = sizeof(struct rtgenmsg);
 688                 break;
 689         case RTM_GETROUTE:
 690                 clr.size = sizeof(struct rtmsg);
 691                 break;
 692         default:
 693                 return;
 694         }
 695
 696         if (!cb)
 697                 return;
 698
 699         clr.msg = nlmsg_alloc_simple(type, flags);
 700         if (!clr.msg)
 701                 goto out;
 702
 703         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 704         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 705         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 706         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 707
 708         nl_send_auto_complete(sock_rtnl, clr.msg);
 709         while (pending > 0)
 710                 nl_recvmsgs(sock_rtnl, cb);
 711
 712         nlmsg_free(clr.msg);
 713 out:
 714         nl_cb_put(cb);
 715 }
 716
 717 /*
 718  * Clear bridge (membership) state and bring down device
 719  */
 720 void system_if_clear_state(struct device *dev)
 721 {
 722         static char buf[256];
 723         char *bridge;
 724
 725         device_set_ifindex(dev, system_if_resolve(dev));
 726         if (dev->external || !dev->ifindex)
 727                 return;
 728
 729         system_if_flags(dev->ifname, 0, IFF_UP);
 730
 731         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 732                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 733                 system_bridge_delbr(dev);
 734                 return;
 735         }
 736
 737         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 738         if (bridge) {
 739                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 740                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 741         }
 742
 743         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 744         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 745         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 746         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 747         system_set_disable_ipv6(dev, "0");
 748 }
 749
 750 static inline unsigned long
 751 sec_to_jiffies(int val)
 752 {
 753         return (unsigned long) val * 100;
 754 }
 755
 756 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 757 {
 758         unsigned long args[4] = {};
 759
 760         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 761                 return -1;
 762
 763         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 764         args[1] = !!cfg->stp;
 765         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 766
 767         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 768         args[1] = sec_to_jiffies(cfg->forward_delay);
 769         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 770
 771         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 772                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 773
 774         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_querier",
 775                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 776
 777         args[0] = BRCTL_SET_BRIDGE_PRIORITY;
 778         args[1] = cfg->priority;
 779         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 780
 781         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 782                 args[0] = BRCTL_SET_AGEING_TIME;
 783                 args[1] = sec_to_jiffies(cfg->ageing_time);
 784                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 785         }
 786
 787         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 788                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 789                 args[1] = sec_to_jiffies(cfg->hello_time);
 790                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 791         }
 792
 793         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 794                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 795                 args[1] = sec_to_jiffies(cfg->max_age);
 796                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 797         }
 798
 799         return 0;
 800 }
 801
 802 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
 803 {
 804         struct nl_msg *msg;
 805         struct nlattr *linkinfo, *data;
 806         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
 807         int i, rv;
 808         static const struct {
 809                 const char *name;
 810                 enum macvlan_mode val;
 811         } modes[] = {
 812                 { "private", MACVLAN_MODE_PRIVATE },
 813                 { "vepa", MACVLAN_MODE_VEPA },
 814                 { "bridge", MACVLAN_MODE_BRIDGE },
 815                 { "passthru", MACVLAN_MODE_PASSTHRU },
 816         };
 817
 818         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 819
 820         if (!msg)
 821                 return -1;
 822
 823         nlmsg_append(msg, &iim, sizeof(iim), 0);
 824
 825         if (cfg->flags & MACVLAN_OPT_MACADDR)
 826                 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
 827         nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
 828         nla_put_u32(msg, IFLA_LINK, dev->ifindex);
 829
 830         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 831                 goto nla_put_failure;
 832
 833         nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
 834
 835         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 836                 goto nla_put_failure;
 837
 838         if (cfg->mode) {
 839                 for (i = 0; i < ARRAY_SIZE(modes); i++) {
 840                         if (strcmp(cfg->mode, modes[i].name) != 0)
 841                                 continue;
 842
 843                         nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
 844                         break;
 845                 }
 846         }
 847
 848         nla_nest_end(msg, data);
 849         nla_nest_end(msg, linkinfo);
 850
 851         rv = system_rtnl_call(msg);
 852         if (rv)
 853                 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
 854
 855         return rv;
 856
 857 nla_put_failure:
 858         nlmsg_free(msg);
 859         return -ENOMEM;
 860 }
 861
 862 static int system_link_del(const char *ifname)
 863 {
 864         struct nl_msg *msg;
 865         struct ifinfomsg iim = {
 866                 .ifi_family = AF_UNSPEC,
 867                 .ifi_index = 0,
 868         };
 869
 870         msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
 871
 872         if (!msg)
 873                 return -1;
 874
 875         nlmsg_append(msg, &iim, sizeof(iim), 0);
 876         nla_put_string(msg, IFLA_IFNAME, ifname);
 877         return system_rtnl_call(msg);
 878 }
 879
 880 int system_macvlan_del(struct device *macvlan)
 881 {
 882         return system_link_del(macvlan->ifname);
 883 }
 884
 885 static int system_vlan(struct device *dev, int id)
 886 {
 887         struct vlan_ioctl_args ifr = {
 888                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 889                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 890         };
 891
 892         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 893
 894         if (id < 0) {
 895                 ifr.cmd = DEL_VLAN_CMD;
 896                 ifr.u.VID = 0;
 897         } else {
 898                 ifr.cmd = ADD_VLAN_CMD;
 899                 ifr.u.VID = id;
 900         }
 901         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 902         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 903 }
 904
 905 int system_vlan_add(struct device *dev, int id)
 906 {
 907         return system_vlan(dev, id);
 908 }
 909
 910 int system_vlan_del(struct device *dev)
 911 {
 912         return system_vlan(dev, -1);
 913 }
 914
 915 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
 916 {
 917         struct nl_msg *msg;
 918         struct nlattr *linkinfo, *data;
 919         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
 920         int rv;
 921
 922         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 923
 924         if (!msg)
 925                 return -1;
 926
 927         nlmsg_append(msg, &iim, sizeof(iim), 0);
 928         nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
 929         nla_put_u32(msg, IFLA_LINK, dev->ifindex);
 930
 931         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 932                 goto nla_put_failure;
 933
 934         nla_put_string(msg, IFLA_INFO_KIND, "vlan");
 935
 936         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 937                 goto nla_put_failure;
 938
 939         nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
 940
 941 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
 942         nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
 943 #else
 944         if(cfg->proto == VLAN_PROTO_8021AD)
 945                 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);
 946 #endif
 947
 948         nla_nest_end(msg, data);
 949         nla_nest_end(msg, linkinfo);
 950
 951         rv = system_rtnl_call(msg);
 952         if (rv)
 953                 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
 954
 955         return rv;
 956
 957 nla_put_failure:
 958         nlmsg_free(msg);
 959         return -ENOMEM;
 960 }
 961
 962 int system_vlandev_del(struct device *vlandev)
 963 {
 964         return system_link_del(vlandev->ifname);
 965 }
 966
 967 static void
 968 system_if_get_settings(struct device *dev, struct device_settings *s)
 969 {
 970         struct ifreq ifr;
 971         char buf[10];
 972
 973         memset(&ifr, 0, sizeof(ifr));
 974         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 975
 976         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 977                 s->mtu = ifr.ifr_mtu;
 978                 s->flags |= DEV_OPT_MTU;
 979         }
 980
 981         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 982                 s->txqueuelen = ifr.ifr_qlen;
 983                 s->flags |= DEV_OPT_TXQUEUELEN;
 984         }
 985
 986         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 987                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 988                 s->flags |= DEV_OPT_MACADDR;
 989         }
 990
 991         if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
 992                 s->ipv6 = !strtoul(buf, NULL, 0);
 993                 s->flags |= DEV_OPT_IPV6;
 994         }
 995
 996         if (ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr) == 0) {
 997                 s->promisc = ifr.ifr_flags & IFF_PROMISC;
 998                 s->flags |= DEV_OPT_PROMISC;
 999         }
1000
1001         if (!system_get_rpfilter(dev, buf, sizeof(buf))) {
1002                 s->rpfilter = strtoul(buf, NULL, 0);
1003                 s->flags |= DEV_OPT_RPFILTER;
1004         }
1005
1006         if (!system_get_acceptlocal(dev, buf, sizeof(buf))) {
1007                 s->acceptlocal = strtoul(buf, NULL, 0);
1008                 s->flags |= DEV_OPT_ACCEPTLOCAL;
1009         }
1010
1011         if (!system_get_igmpversion(dev, buf, sizeof(buf))) {
1012                 s->igmpversion = strtoul(buf, NULL, 0);
1013                 s->flags |= DEV_OPT_IGMPVERSION;
1014         }
1015
1016         if (!system_get_mldversion(dev, buf, sizeof(buf))) {
1017                 s->mldversion = strtoul(buf, NULL, 0);
1018                 s->flags |= DEV_OPT_MLDVERSION;
1019         }
1020 }
1021
1022 void
1023 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
1024 {
1025         struct ifreq ifr;
1026
1027         if (!apply_mask)
1028                 return;
1029
1030         memset(&ifr, 0, sizeof(ifr));
1031         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
1032         if (s->flags & DEV_OPT_MTU & apply_mask) {
1033                 ifr.ifr_mtu = s->mtu;
1034                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
1035                         s->flags &= ~DEV_OPT_MTU;
1036         }
1037         if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
1038                 ifr.ifr_qlen = s->txqueuelen;
1039                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
1040                         s->flags &= ~DEV_OPT_TXQUEUELEN;
1041         }
1042         if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
1043                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
1044                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
1045                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
1046                         s->flags &= ~DEV_OPT_MACADDR;
1047         }
1048         if (s->flags & DEV_OPT_IPV6 & apply_mask)
1049                 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
1050         if (s->flags & DEV_OPT_PROMISC & apply_mask) {
1051                 if (system_if_flags(dev->ifname, s->promisc ? IFF_PROMISC : 0,
1052                                     !s->promisc ? IFF_PROMISC : 0) < 0)
1053                         s->flags &= ~DEV_OPT_PROMISC;
1054         }
1055         if (s->flags & DEV_OPT_RPFILTER & apply_mask) {
1056                 char buf[2];
1057
1058                 snprintf(buf, sizeof(buf), "%d", s->rpfilter);
1059                 system_set_rpfilter(dev, buf);
1060         }
1061         if (s->flags & DEV_OPT_ACCEPTLOCAL & apply_mask)
1062                 system_set_acceptlocal(dev, s->acceptlocal ? "1" : "0");
1063         if (s->flags & DEV_OPT_IGMPVERSION & apply_mask) {
1064                 char buf[2];
1065
1066                 snprintf(buf, sizeof(buf), "%d", s->igmpversion);
1067                 system_set_igmpversion(dev, buf);
1068         }
1069         if (s->flags & DEV_OPT_MLDVERSION & apply_mask) {
1070                 char buf[2];
1071
1072                 snprintf(buf, sizeof(buf), "%d", s->mldversion);
1073                 system_set_mldversion(dev, buf);
1074         }
1075 }
1076
1077 int system_if_up(struct device *dev)
1078 {
1079         system_if_get_settings(dev, &dev->orig_settings);
1080         /* Only keep orig settings based on what needs to be set */
1081         dev->orig_settings.flags &= dev->settings.flags;
1082         system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
1083         return system_if_flags(dev->ifname, IFF_UP, 0);
1084 }
1085
1086 int system_if_down(struct device *dev)
1087 {
1088         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
1089         system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
1090         return ret;
1091 }
1092
1093 struct if_check_data {
1094         struct device *dev;
1095         int pending;
1096         int ret;
1097 };
1098
1099 #ifndef IFF_LOWER_UP
1100 #define IFF_LOWER_UP    0x10000
1101 #endif
1102
1103 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
1104 {
1105         struct nlmsghdr *nh = nlmsg_hdr(msg);
1106         struct ifinfomsg *ifi = NLMSG_DATA(nh);
1107         struct if_check_data *chk = (struct if_check_data *)arg;
1108
1109         if (nh->nlmsg_type != RTM_NEWLINK)
1110                 return NL_SKIP;
1111
1112         device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1113         device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1114
1115         return NL_OK;
1116 }
1117
1118 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1119 {
1120         struct if_check_data *chk = (struct if_check_data *)arg;
1121         chk->pending = 0;
1122         return NL_STOP;
1123 }
1124
1125 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1126 {
1127         struct if_check_data *chk = (struct if_check_data *)arg;
1128
1129         device_set_present(chk->dev, false);
1130         device_set_link(chk->dev, false);
1131         chk->pending = err->error;
1132
1133         return NL_STOP;
1134 }
1135
1136 int system_if_check(struct device *dev)
1137 {
1138         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1139         struct nl_msg *msg;
1140         struct ifinfomsg ifi = {
1141                 .ifi_family = AF_UNSPEC,
1142                 .ifi_index = 0,
1143         };
1144         struct if_check_data chk = {
1145                 .dev = dev,
1146                 .pending = 1,
1147         };
1148         int ret = 1;
1149
1150         msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1151         if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1152             nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1153                 goto out;
1154
1155         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1156         nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1157         nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1158
1159         nl_send_auto_complete(sock_rtnl, msg);
1160         while (chk.pending > 0)
1161                 nl_recvmsgs(sock_rtnl, cb);
1162
1163         nlmsg_free(msg);
1164         ret = chk.pending;
1165
1166 out:
1167         nl_cb_put(cb);
1168         return ret;
1169 }
1170
1171 struct device *
1172 system_if_get_parent(struct device *dev)
1173 {
1174         char buf[64], *devname;
1175         int ifindex, iflink, len;
1176         FILE *f;
1177
1178         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1179         f = fopen(buf, "r");
1180         if (!f)
1181                 return NULL;
1182
1183         len = fread(buf, 1, sizeof(buf) - 1, f);
1184         fclose(f);
1185
1186         if (len <= 0)
1187                 return NULL;
1188
1189         buf[len] = 0;
1190         iflink = strtoul(buf, NULL, 0);
1191         ifindex = system_if_resolve(dev);
1192         if (!iflink || iflink == ifindex)
1193                 return NULL;
1194
1195         devname = if_indextoname(iflink, buf);
1196         if (!devname)
1197                 return NULL;
1198
1199         return device_get(devname, true);
1200 }
1201
1202 static bool
1203 read_string_file(int dir_fd, const char *file, char *buf, int len)
1204 {
1205         bool ret = false;
1206         char *c;
1207         int fd;
1208
1209         fd = openat(dir_fd, file, O_RDONLY);
1210         if (fd < 0)
1211                 return false;
1212
1213 retry:
1214         len = read(fd, buf, len - 1);
1215         if (len < 0) {
1216                 if (errno == EINTR)
1217                         goto retry;
1218         } else if (len > 0) {
1219                         buf[len] = 0;
1220
1221                         c = strchr(buf, '\n');
1222                         if (c)
1223                                 *c = 0;
1224
1225                         ret = true;
1226         }
1227
1228         close(fd);
1229
1230         return ret;
1231 }
1232
1233 static bool
1234 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1235 {
1236         char buf[64];
1237         bool ret = false;
1238
1239         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1240         if (ret)
1241                 *val = strtoull(buf, NULL, 0);
1242
1243         return ret;
1244 }
1245
1246 /* Assume advertised flags == supported flags */
1247 static const struct {
1248         uint32_t mask;
1249         const char *name;
1250 } ethtool_link_modes[] = {
1251         { ADVERTISED_10baseT_Half, "10H" },
1252         { ADVERTISED_10baseT_Full, "10F" },
1253         { ADVERTISED_100baseT_Half, "100H" },
1254         { ADVERTISED_100baseT_Full, "100F" },
1255         { ADVERTISED_1000baseT_Half, "1000H" },
1256         { ADVERTISED_1000baseT_Full, "1000F" },
1257 };
1258
1259 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1260 {
1261         int i;
1262         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1263                 if (mask & ethtool_link_modes[i].mask)
1264                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1265         }
1266 }
1267
1268 bool
1269 system_if_force_external(const char *ifname)
1270 {
1271         char buf[64];
1272         struct stat s;
1273
1274         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1275         return stat(buf, &s) == 0;
1276 }
1277
1278 int
1279 system_if_dump_info(struct device *dev, struct blob_buf *b)
1280 {
1281         struct ethtool_cmd ecmd;
1282         struct ifreq ifr;
1283         char buf[64], *s;
1284         void *c;
1285         int dir_fd;
1286
1287         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1288         dir_fd = open(buf, O_DIRECTORY);
1289
1290         memset(&ecmd, 0, sizeof(ecmd));
1291         memset(&ifr, 0, sizeof(ifr));
1292         strcpy(ifr.ifr_name, dev->ifname);
1293         ifr.ifr_data = (caddr_t) &ecmd;
1294         ecmd.cmd = ETHTOOL_GSET;
1295
1296         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1297                 c = blobmsg_open_array(b, "link-advertising");
1298                 system_add_link_modes(b, ecmd.advertising);
1299                 blobmsg_close_array(b, c);
1300
1301                 c = blobmsg_open_array(b, "link-supported");
1302                 system_add_link_modes(b, ecmd.supported);
1303                 blobmsg_close_array(b, c);
1304
1305                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1306                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1307                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1308                 blobmsg_add_string_buffer(b);
1309         }
1310
1311         close(dir_fd);
1312         return 0;
1313 }
1314
1315 int
1316 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1317 {
1318         const char *const counters[] = {
1319                 "collisions",     "rx_frame_errors",   "tx_compressed",
1320                 "multicast",      "rx_length_errors",  "tx_dropped",
1321                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
1322                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
1323                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
1324                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
1325                 "rx_errors",      "tx_bytes",          "tx_window_errors",
1326                 "rx_fifo_errors", "tx_carrier_errors",
1327         };
1328         char buf[64];
1329         int stats_dir;
1330         int i;
1331         uint64_t val = 0;
1332
1333         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1334         stats_dir = open(buf, O_DIRECTORY);
1335         if (stats_dir < 0)
1336                 return -1;
1337
1338         for (i = 0; i < ARRAY_SIZE(counters); i++)
1339                 if (read_uint64_file(stats_dir, counters[i], &val))
1340                         blobmsg_add_u64(b, counters[i], val);
1341
1342         close(stats_dir);
1343         return 0;
1344 }
1345
1346 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1347 {
1348         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1349         int alen = v4 ? 4 : 16;
1350         unsigned int flags = 0;
1351         struct ifaddrmsg ifa = {
1352                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1353                 .ifa_prefixlen = addr->mask,
1354                 .ifa_index = dev->ifindex,
1355         };
1356
1357         struct nl_msg *msg;
1358         if (cmd == RTM_NEWADDR)
1359                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1360
1361         msg = nlmsg_alloc_simple(cmd, flags);
1362         if (!msg)
1363                 return -1;
1364
1365         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1366         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1367         if (v4) {
1368                 if (addr->broadcast)
1369                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1370                 if (addr->point_to_point)
1371                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1372         } else {
1373                 time_t now = system_get_rtime();
1374                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1375
1376                 if (addr->preferred_until) {
1377                         int64_t preferred = addr->preferred_until - now;
1378                         if (preferred < 0)
1379                                 preferred = 0;
1380                         else if (preferred > UINT32_MAX)
1381                                 preferred = UINT32_MAX;
1382
1383                         cinfo.ifa_prefered = preferred;
1384                 }
1385
1386                 if (addr->valid_until) {
1387                         int64_t valid = addr->valid_until - now;
1388                         if (valid <= 0)
1389                                 return -1;
1390                         else if (valid > UINT32_MAX)
1391                                 valid = UINT32_MAX;
1392
1393                         cinfo.ifa_valid = valid;
1394                 }
1395
1396                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1397         }
1398
1399         return system_rtnl_call(msg);
1400 }
1401
1402 int system_add_address(struct device *dev, struct device_addr *addr)
1403 {
1404         return system_addr(dev, addr, RTM_NEWADDR);
1405 }
1406
1407 int system_del_address(struct device *dev, struct device_addr *addr)
1408 {
1409         return system_addr(dev, addr, RTM_DELADDR);
1410 }
1411
1412 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1413 {
1414         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1415         bool have_gw;
1416         unsigned int flags = 0;
1417
1418         if (alen == 4)
1419                 have_gw = !!route->nexthop.in.s_addr;
1420         else
1421                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1422                         route->nexthop.in6.s6_addr32[1] ||
1423                         route->nexthop.in6.s6_addr32[2] ||
1424                         route->nexthop.in6.s6_addr32[3];
1425
1426         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1427                         ? route->table : RT_TABLE_MAIN;
1428
1429         struct rtmsg rtm = {
1430                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1431                 .rtm_dst_len = route->mask,
1432                 .rtm_src_len = route->sourcemask,
1433                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1434                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1435                 .rtm_scope = RT_SCOPE_NOWHERE,
1436                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1437                 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1438         };
1439         struct nl_msg *msg;
1440
1441         if (cmd == RTM_NEWROUTE) {
1442                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1443
1444                 if (!dev) { // Add null-route
1445                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1446                         rtm.rtm_type = RTN_UNREACHABLE;
1447                 }
1448                 else
1449                         rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1450         }
1451
1452         if (route->flags & DEVROUTE_TYPE) {
1453                 rtm.rtm_type = route->type;
1454                 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1455                         if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1456                             rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1457                                 rtm.rtm_table = RT_TABLE_LOCAL;
1458                 }
1459
1460                 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT) {
1461                         rtm.rtm_scope = RT_SCOPE_HOST;
1462                 } else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1463                                 rtm.rtm_type == RTN_ANYCAST) {
1464                         rtm.rtm_scope = RT_SCOPE_LINK;
1465                 } else if (rtm.rtm_type == RTN_BLACKHOLE || rtm.rtm_type == RTN_UNREACHABLE ||
1466                                 rtm.rtm_type == RTN_PROHIBIT || rtm.rtm_type == RTN_FAILED_POLICY) {
1467                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1468                         dev = NULL;
1469                 }
1470         }
1471
1472         msg = nlmsg_alloc_simple(cmd, flags);
1473         if (!msg)
1474                 return -1;
1475
1476         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1477
1478         if (route->mask)
1479                 nla_put(msg, RTA_DST, alen, &route->addr);
1480
1481         if (route->sourcemask) {
1482                 if (rtm.rtm_family == AF_INET)
1483                         nla_put(msg, RTA_PREFSRC, alen, &route->source);
1484                 else
1485                         nla_put(msg, RTA_SRC, alen, &route->source);
1486         }
1487
1488         if (route->metric > 0)
1489                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1490
1491         if (have_gw)
1492                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1493
1494         if (dev)
1495                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1496
1497         if (table >= 256)
1498                 nla_put_u32(msg, RTA_TABLE, table);
1499
1500         if (route->flags & DEVROUTE_MTU) {
1501                 struct nlattr *metrics;
1502
1503                 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1504                         goto nla_put_failure;
1505
1506                 nla_put_u32(msg, RTAX_MTU, route->mtu);
1507
1508                 nla_nest_end(msg, metrics);
1509         }
1510
1511         return system_rtnl_call(msg);
1512
1513 nla_put_failure:
1514         nlmsg_free(msg);
1515         return -ENOMEM;
1516 }
1517
1518 int system_add_route(struct device *dev, struct device_route *route)
1519 {
1520         return system_rt(dev, route, RTM_NEWROUTE);
1521 }
1522
1523 int system_del_route(struct device *dev, struct device_route *route)
1524 {
1525         return system_rt(dev, route, RTM_DELROUTE);
1526 }
1527
1528 int system_flush_routes(void)
1529 {
1530         const char *names[] = {
1531                 "/proc/sys/net/ipv4/route/flush",
1532                 "/proc/sys/net/ipv6/route/flush"
1533         };
1534         int fd, i;
1535
1536         for (i = 0; i < ARRAY_SIZE(names); i++) {
1537                 fd = open(names[i], O_WRONLY);
1538                 if (fd < 0)
1539                         continue;
1540
1541                 if (write(fd, "-1", 2)) {}
1542                 close(fd);
1543         }
1544         return 0;
1545 }
1546
1547 bool system_resolve_rt_type(const char *type, unsigned int *id)
1548 {
1549         return system_rtn_aton(type, id);
1550 }
1551
1552 bool system_resolve_rt_table(const char *name, unsigned int *id)
1553 {
1554         FILE *f;
1555         char *e, buf[128];
1556         unsigned int n, table = RT_TABLE_UNSPEC;
1557
1558         /* first try to parse table as number */
1559         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1560                 table = n;
1561
1562         /* handle well known aliases */
1563         else if (!strcmp(name, "default"))
1564                 table = RT_TABLE_DEFAULT;
1565         else if (!strcmp(name, "main"))
1566                 table = RT_TABLE_MAIN;
1567         else if (!strcmp(name, "local"))
1568                 table = RT_TABLE_LOCAL;
1569
1570         /* try to look up name in /etc/iproute2/rt_tables */
1571         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1572         {
1573                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1574                 {
1575                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1576                                 continue;
1577
1578                         n = strtoul(e, NULL, 10);
1579                         e = strtok(NULL, " \t\n");
1580
1581                         if (e && !strcmp(e, name))
1582                         {
1583                                 table = n;
1584                                 break;
1585                         }
1586                 }
1587
1588                 fclose(f);
1589         }
1590
1591         if (table == RT_TABLE_UNSPEC)
1592                 return false;
1593
1594         *id = table;
1595         return true;
1596 }
1597
1598 bool system_is_default_rt_table(unsigned int id)
1599 {
1600         return (id == RT_TABLE_MAIN);
1601 }
1602
1603 bool system_resolve_rpfilter(const char *filter, unsigned int *id)
1604 {
1605         char *e;
1606         unsigned int n;
1607
1608         if (!strcmp(filter, "strict"))
1609                 n = 1;
1610         else if (!strcmp(filter, "loose"))
1611                 n = 2;
1612         else {
1613                 n = strtoul(filter, &e, 0);
1614                 if (*e || e == filter || n > 2)
1615                         return false;
1616         }
1617
1618         *id = n;
1619         return true;
1620 }
1621
1622 static int system_iprule(struct iprule *rule, int cmd)
1623 {
1624         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1625
1626         struct nl_msg *msg;
1627         struct rtmsg rtm = {
1628                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1629                 .rtm_protocol = RTPROT_STATIC,
1630                 .rtm_scope = RT_SCOPE_UNIVERSE,
1631                 .rtm_table = RT_TABLE_UNSPEC,
1632                 .rtm_type = RTN_UNSPEC,
1633                 .rtm_flags = 0,
1634         };
1635
1636         if (cmd == RTM_NEWRULE) {
1637                 rtm.rtm_type = RTN_UNICAST;
1638                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1639         }
1640
1641         if (rule->invert)
1642                 rtm.rtm_flags |= FIB_RULE_INVERT;
1643
1644         if (rule->flags & IPRULE_SRC)
1645                 rtm.rtm_src_len = rule->src_mask;
1646
1647         if (rule->flags & IPRULE_DEST)
1648                 rtm.rtm_dst_len = rule->dest_mask;
1649
1650         if (rule->flags & IPRULE_TOS)
1651                 rtm.rtm_tos = rule->tos;
1652
1653         if (rule->flags & IPRULE_LOOKUP) {
1654                 if (rule->lookup < 256)
1655                         rtm.rtm_table = rule->lookup;
1656         }
1657
1658         if (rule->flags & IPRULE_ACTION)
1659                 rtm.rtm_type = rule->action;
1660         else if (rule->flags & IPRULE_GOTO)
1661                 rtm.rtm_type = FR_ACT_GOTO;
1662         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1663                 rtm.rtm_type = FR_ACT_NOP;
1664
1665         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1666
1667         if (!msg)
1668                 return -1;
1669
1670         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1671
1672         if (rule->flags & IPRULE_IN)
1673                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1674
1675         if (rule->flags & IPRULE_OUT)
1676                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1677
1678         if (rule->flags & IPRULE_SRC)
1679                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1680
1681         if (rule->flags & IPRULE_DEST)
1682                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1683
1684         if (rule->flags & IPRULE_PRIORITY)
1685                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1686         else if (cmd == RTM_NEWRULE)
1687                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1688
1689         if (rule->flags & IPRULE_FWMARK)
1690                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1691
1692         if (rule->flags & IPRULE_FWMASK)
1693                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1694
1695         if (rule->flags & IPRULE_LOOKUP) {
1696                 if (rule->lookup >= 256)
1697                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1698         }
1699
1700         if (rule->flags & IPRULE_GOTO)
1701                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1702
1703         return system_rtnl_call(msg);
1704 }
1705
1706 int system_add_iprule(struct iprule *rule)
1707 {
1708         return system_iprule(rule, RTM_NEWRULE);
1709 }
1710
1711 int system_del_iprule(struct iprule *rule)
1712 {
1713         return system_iprule(rule, RTM_DELRULE);
1714 }
1715
1716 int system_flush_iprules(void)
1717 {
1718         int rv = 0;
1719         struct iprule rule;
1720
1721         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1722         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1723
1724         memset(&rule, 0, sizeof(rule));
1725
1726
1727         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1728
1729         rule.priority = 0;
1730         rule.lookup = RT_TABLE_LOCAL;
1731         rv |= system_iprule(&rule, RTM_NEWRULE);
1732
1733         rule.priority = 32766;
1734         rule.lookup = RT_TABLE_MAIN;
1735         rv |= system_iprule(&rule, RTM_NEWRULE);
1736
1737         rule.priority = 32767;
1738         rule.lookup = RT_TABLE_DEFAULT;
1739         rv |= system_iprule(&rule, RTM_NEWRULE);
1740
1741
1742         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1743
1744         rule.priority = 0;
1745         rule.lookup = RT_TABLE_LOCAL;
1746         rv |= system_iprule(&rule, RTM_NEWRULE);
1747
1748         rule.priority = 32766;
1749         rule.lookup = RT_TABLE_MAIN;
1750         rv |= system_iprule(&rule, RTM_NEWRULE);
1751
1752         return rv;
1753 }
1754
1755 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1756 {
1757         return system_rtn_aton(action, id);
1758 }
1759
1760 time_t system_get_rtime(void)
1761 {
1762         struct timespec ts;
1763         struct timeval tv;
1764
1765         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1766                 return ts.tv_sec;
1767
1768         if (gettimeofday(&tv, NULL) == 0)
1769                 return tv.tv_sec;
1770
1771         return 0;
1772 }
1773
1774 #ifndef IP_DF
1775 #define IP_DF       0x4000
1776 #endif
1777
1778 static int tunnel_ioctl(const char *name, int cmd, void *p)
1779 {
1780         struct ifreq ifr;
1781
1782         memset(&ifr, 0, sizeof(ifr));
1783         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1784         ifr.ifr_ifru.ifru_data = p;
1785         return ioctl(sock_ioctl, cmd, &ifr);
1786 }
1787
1788 #ifdef IFLA_IPTUN_MAX
1789 #define IP6_FLOWINFO_TCLASS     htonl(0x0FF00000)
1790 static int system_add_gre_tunnel(const char *name, const char *kind,
1791                                  const unsigned int link, struct blob_attr **tb, bool v6)
1792 {
1793         struct nl_msg *nlm;
1794         struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC, };
1795         struct blob_attr *cur;
1796         uint32_t ikey = 0, okey = 0, flags = 0, flowinfo = 0;
1797         uint16_t iflags = 0, oflags = 0;
1798         uint8_t tos = 0;
1799         int ret = 0, ttl = 64;
1800
1801         nlm = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1802         if (!nlm)
1803                 return -1;
1804
1805         nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1806         nla_put_string(nlm, IFLA_IFNAME, name);
1807
1808         struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1809         if (!linkinfo) {
1810                 ret = -ENOMEM;
1811                 goto failure;
1812         }
1813
1814         nla_put_string(nlm, IFLA_INFO_KIND, kind);
1815         struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1816         if (!infodata) {
1817                 ret = -ENOMEM;
1818                 goto failure;
1819         }
1820
1821         if (link)
1822                 nla_put_u32(nlm, IFLA_GRE_LINK, link);
1823
1824         if ((cur = tb[TUNNEL_ATTR_TTL]))
1825                 ttl = blobmsg_get_u32(cur);
1826
1827         nla_put_u8(nlm, IFLA_GRE_TTL, ttl);
1828
1829         if ((cur = tb[TUNNEL_ATTR_TOS])) {
1830                 char *str = blobmsg_get_string(cur);
1831                 if (strcmp(str, "inherit")) {
1832                         unsigned uval;
1833
1834                         if (!system_tos_aton(str, &uval)) {
1835                                 ret = -EINVAL;
1836                                 goto failure;
1837                         }
1838
1839                         if (v6)
1840                                 flowinfo |= htonl(uval << 20) & IP6_FLOWINFO_TCLASS;
1841                         else
1842                                 tos = uval;
1843                 } else {
1844                         if (v6)
1845                                 flags |= IP6_TNL_F_USE_ORIG_TCLASS;
1846                         else
1847                                 tos = 1;
1848                 }
1849         }
1850
1851         if ((cur = tb[TUNNEL_ATTR_INFO]) && (blobmsg_type(cur) == BLOBMSG_TYPE_STRING)) {
1852                 uint8_t icsum, ocsum, iseqno, oseqno;
1853                 if (sscanf(blobmsg_get_string(cur), "%u,%u,%hhu,%hhu,%hhu,%hhu",
1854                         &ikey, &okey, &icsum, &ocsum, &iseqno, &oseqno) < 6) {
1855                         ret = -EINVAL;
1856                         goto failure;
1857                 }
1858
1859                 if (ikey)
1860                         iflags |= GRE_KEY;
1861
1862                 if (okey)
1863                         oflags |= GRE_KEY;
1864
1865                 if (icsum)
1866                         iflags |= GRE_CSUM;
1867
1868                 if (ocsum)
1869                         oflags |= GRE_CSUM;
1870
1871                 if (iseqno)
1872                         iflags |= GRE_SEQ;
1873
1874                 if (oseqno)
1875                         oflags |= GRE_SEQ;
1876         }
1877
1878         if (v6) {
1879                 struct in6_addr in6buf;
1880                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1881                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1882                                 ret = -EINVAL;
1883                                 goto failure;
1884                         }
1885                         nla_put(nlm, IFLA_GRE_LOCAL, sizeof(in6buf), &in6buf);
1886                 }
1887
1888                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1889                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1890                                 ret = -EINVAL;
1891                                 goto failure;
1892                         }
1893                         nla_put(nlm, IFLA_GRE_REMOTE, sizeof(in6buf), &in6buf);
1894                 }
1895                 nla_put_u8(nlm, IFLA_GRE_ENCAP_LIMIT, 4);
1896
1897                 if (flowinfo)
1898                         nla_put_u32(nlm, IFLA_GRE_FLOWINFO, flowinfo);
1899
1900                 if (flags)
1901                         nla_put_u32(nlm, IFLA_GRE_FLAGS, flags);
1902         } else {
1903                 struct in_addr inbuf;
1904                 bool set_df = true;
1905
1906                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1907                         if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
1908                                 ret = -EINVAL;
1909                                 goto failure;
1910                         }
1911                         nla_put(nlm, IFLA_GRE_LOCAL, sizeof(inbuf), &inbuf);
1912                 }
1913
1914                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1915                         if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
1916                                 ret = -EINVAL;
1917                                 goto failure;
1918                         }
1919                         nla_put(nlm, IFLA_GRE_REMOTE, sizeof(inbuf), &inbuf);
1920
1921                         if (IN_MULTICAST(ntohl(inbuf.s_addr))) {
1922                                 if (!okey) {
1923                                         okey = inbuf.s_addr;
1924                                         oflags |= GRE_KEY;
1925                                 }
1926
1927                                 if (!ikey) {
1928                                         ikey = inbuf.s_addr;
1929                                         iflags |= GRE_KEY;
1930                                 }
1931                         }
1932                 }
1933
1934                 if ((cur = tb[TUNNEL_ATTR_DF]))
1935                         set_df = blobmsg_get_bool(cur);
1936
1937                 /* ttl !=0 and nopmtudisc are incompatible */
1938                 if (ttl && !set_df) {
1939                         ret = -EINVAL;
1940                         goto failure;
1941                 }
1942
1943                 nla_put_u8(nlm, IFLA_GRE_PMTUDISC, set_df ? 1 : 0);
1944
1945                 nla_put_u8(nlm, IFLA_GRE_TOS, tos);
1946         }
1947
1948         if (oflags)
1949                 nla_put_u16(nlm, IFLA_GRE_OFLAGS, oflags);
1950
1951         if (iflags)
1952                 nla_put_u16(nlm, IFLA_GRE_IFLAGS, iflags);
1953
1954         if (okey)
1955                 nla_put_u32(nlm, IFLA_GRE_OKEY, okey);
1956
1957         if (ikey)
1958                 nla_put_u32(nlm, IFLA_GRE_IKEY, ikey);
1959
1960         nla_nest_end(nlm, infodata);
1961         nla_nest_end(nlm, linkinfo);
1962
1963         return system_rtnl_call(nlm);
1964
1965 failure:
1966         nlmsg_free(nlm);
1967         return ret;
1968 }
1969 #endif
1970
1971 static int system_add_proto_tunnel(const char *name, const uint8_t proto, const unsigned int link, struct blob_attr **tb)
1972 {
1973         struct blob_attr *cur;
1974         bool set_df = true;
1975         struct ip_tunnel_parm p  = {
1976                 .link = link,
1977                 .iph = {
1978                         .version = 4,
1979                         .ihl = 5,
1980                         .protocol = proto,
1981                 }
1982         };
1983
1984         if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1985                         inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1986                 return -EINVAL;
1987
1988         if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1989                         inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1990                 return -EINVAL;
1991
1992         if ((cur = tb[TUNNEL_ATTR_DF]))
1993                 set_df = blobmsg_get_bool(cur);
1994
1995         if ((cur = tb[TUNNEL_ATTR_TTL]))
1996                 p.iph.ttl = blobmsg_get_u32(cur);
1997
1998         if ((cur = tb[TUNNEL_ATTR_TOS])) {
1999                 char *str = blobmsg_get_string(cur);
2000                 if (strcmp(str, "inherit")) {
2001                         unsigned uval;
2002
2003                         if (!system_tos_aton(str, &uval))
2004                                 return -EINVAL;
2005
2006                         p.iph.tos = uval;
2007                 } else
2008                         p.iph.tos = 1;
2009         }
2010
2011         p.iph.frag_off = set_df ? htons(IP_DF) : 0;
2012         /* ttl !=0 and nopmtudisc are incompatible */
2013         if (p.iph.ttl && p.iph.frag_off == 0)
2014                 return -EINVAL;
2015
2016         strncpy(p.name, name, sizeof(p.name));
2017
2018         switch (p.iph.protocol) {
2019         case IPPROTO_IPIP:
2020                 return tunnel_ioctl("tunl0", SIOCADDTUNNEL, &p);
2021         case IPPROTO_IPV6:
2022                 return tunnel_ioctl("sit0", SIOCADDTUNNEL, &p);
2023         default:
2024                 break;
2025         }
2026         return -1;
2027 }
2028
2029 static int __system_del_ip_tunnel(const char *name, struct blob_attr **tb)
2030 {
2031         struct blob_attr *cur;
2032         const char *str;
2033
2034         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2035                 return -EINVAL;
2036         str = blobmsg_data(cur);
2037
2038         if (!strcmp(str, "greip") || !strcmp(str, "gretapip") ||
2039             !strcmp(str, "greip6") || !strcmp(str, "gretapip6"))
2040                 return system_link_del(name);
2041         else
2042                 return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
2043 }
2044
2045 int system_del_ip_tunnel(const char *name, struct blob_attr *attr)
2046 {
2047         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2048
2049         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2050                 blob_data(attr), blob_len(attr));
2051
2052         return __system_del_ip_tunnel(name, tb);
2053 }
2054
2055 int system_update_ipv6_mtu(struct device *dev, int mtu)
2056 {
2057         int ret = -1;
2058         char buf[64];
2059         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
2060                         dev->ifname);
2061
2062         int fd = open(buf, O_RDWR);
2063         ssize_t len = read(fd, buf, sizeof(buf) - 1);
2064         if (len < 0)
2065                 goto out;
2066
2067         buf[len] = 0;
2068         ret = atoi(buf);
2069
2070         if (!mtu || ret <= mtu)
2071                 goto out;
2072
2073         lseek(fd, 0, SEEK_SET);
2074         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
2075                 ret = -1;
2076
2077 out:
2078         close(fd);
2079         return ret;
2080 }
2081
2082 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
2083 {
2084         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
2085         struct blob_attr *cur;
2086         const char *str;
2087
2088         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
2089                 blob_data(attr), blob_len(attr));
2090
2091         __system_del_ip_tunnel(name, tb);
2092
2093         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
2094                 return -EINVAL;
2095         str = blobmsg_data(cur);
2096
2097         unsigned int ttl = 0;
2098         if ((cur = tb[TUNNEL_ATTR_TTL])) {
2099                 ttl = blobmsg_get_u32(cur);
2100                 if (ttl > 255)
2101                         return -EINVAL;
2102         }
2103
2104         unsigned int link = 0;
2105         if ((cur = tb[TUNNEL_ATTR_LINK])) {
2106                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
2107                 if (!iface)
2108                         return -EINVAL;
2109
2110                 if (iface->l3_dev.dev)
2111                         link = iface->l3_dev.dev->ifindex;
2112         }
2113
2114         if (!strcmp(str, "sit")) {
2115                 if (system_add_proto_tunnel(name, IPPROTO_IPV6, link, tb) < 0)
2116                         return -1;
2117
2118 #ifdef SIOCADD6RD
2119                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
2120                         unsigned int mask;
2121                         struct ip_tunnel_6rd p6;
2122
2123                         memset(&p6, 0, sizeof(p6));
2124
2125                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
2126                                                 &p6.prefix, &mask) || mask > 128)
2127                                 return -EINVAL;
2128                         p6.prefixlen = mask;
2129
2130                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
2131                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
2132                                                         &p6.relay_prefix, &mask) || mask > 32)
2133                                         return -EINVAL;
2134                                 p6.relay_prefixlen = mask;
2135                         }
2136
2137                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
2138                                 __system_del_ip_tunnel(name, tb);
2139                                 return -1;
2140                         }
2141                 }
2142 #endif
2143 #ifdef IFLA_IPTUN_MAX
2144         } else if (!strcmp(str, "ipip6")) {
2145                 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
2146                                 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
2147                 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
2148                 int ret = 0;
2149
2150                 if (!nlm)
2151                         return -1;
2152
2153                 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
2154                 nla_put_string(nlm, IFLA_IFNAME, name);
2155
2156                 if (link)
2157                         nla_put_u32(nlm, IFLA_LINK, link);
2158
2159                 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
2160                 if (!linkinfo) {
2161                         ret = -ENOMEM;
2162                         goto failure;
2163                 }
2164                 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
2165                 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
2166                 if (!infodata) {
2167                         ret = -ENOMEM;
2168                         goto failure;
2169                 }
2170
2171                 if (link)
2172                         nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
2173
2174                 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
2175                 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
2176                 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
2177
2178                 struct in6_addr in6buf;
2179                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2180                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2181                                 ret = -EINVAL;
2182                                 goto failure;
2183                         }
2184                         nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
2185                 }
2186
2187                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2188                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2189                                 ret = -EINVAL;
2190                                 goto failure;
2191                         }
2192                         nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
2193                 }
2194
2195 #ifdef IFLA_IPTUN_FMR_MAX
2196                 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
2197                         struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
2198
2199                         struct blob_attr *fmr;
2200                         unsigned rem, fmrcnt = 0;
2201                         blobmsg_for_each_attr(fmr, cur, rem) {
2202                                 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
2203                                         continue;
2204
2205                                 unsigned ip4len, ip6len, ealen, offset = 6;
2206                                 char ip6buf[48];
2207                                 char ip4buf[16];
2208
2209                                 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
2210                                                 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
2211                                         ret = -EINVAL;
2212                                         goto failure;
2213                                 }
2214
2215                                 struct in6_addr ip6prefix;
2216                                 struct in_addr ip4prefix;
2217                                 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
2218                                                 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
2219                                         ret = -EINVAL;
2220                                         goto failure;
2221                                 }
2222
2223                                 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
2224
2225                                 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
2226                                 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
2227                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
2228                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
2229                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
2230                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
2231
2232                                 nla_nest_end(nlm, rule);
2233                         }
2234
2235                         nla_nest_end(nlm, fmrs);
2236                 }
2237 #endif
2238
2239                 nla_nest_end(nlm, infodata);
2240                 nla_nest_end(nlm, linkinfo);
2241
2242                 return system_rtnl_call(nlm);
2243 failure:
2244                 nlmsg_free(nlm);
2245                 return ret;
2246         } else if (!strcmp(str, "greip")) {
2247                 return system_add_gre_tunnel(name, "gre", link, tb, false);
2248         } else if (!strcmp(str, "gretapip"))  {
2249                 return system_add_gre_tunnel(name, "gretap", link, tb, false);
2250         } else if (!strcmp(str, "greip6")) {
2251                 return system_add_gre_tunnel(name, "ip6gre", link, tb, true);
2252         } else if (!strcmp(str, "gretapip6")) {
2253                 return system_add_gre_tunnel(name, "ip6gretap", link, tb, true);
2254 #endif
2255         } else if (!strcmp(str, "ipip")) {
2256                 return system_add_proto_tunnel(name, IPPROTO_IPIP, link, tb);
2257         }
2258         else
2259                 return -EINVAL;
2260
2261         return 0;
2262 }