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