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 int system_init(void)
 208 {
 209         static struct event_socket rtnl_event;
 210         static struct event_socket hotplug_event;
 211
 212         sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
 213         system_fd_set_cloexec(sock_ioctl);
 214
 215         // Prepare socket for routing / address control
 216         sock_rtnl = create_socket(NETLINK_ROUTE, 0);
 217         if (!sock_rtnl)
 218                 return -1;
 219
 220         if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
 221                 return -1;
 222
 223         if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
 224                                      handle_hotplug_event, 0))
 225                 return -1;
 226
 227         // Receive network link events form kernel
 228         nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
 229
 230         return 0;
 231 }
 232
 233 static void system_set_sysctl(const char *path, const char *val)
 234 {
 235         int fd;
 236
 237         fd = open(path, O_WRONLY);
 238         if (fd < 0)
 239                 return;
 240
 241         if (write(fd, val, strlen(val))) {}
 242         close(fd);
 243 }
 244
 245 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
 246 {
 247         snprintf(dev_buf, sizeof(dev_buf), path, device);
 248         system_set_sysctl(dev_buf, val);
 249 }
 250
 251 static void system_set_disable_ipv6(struct device *dev, const char *val)
 252 {
 253         system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
 254 }
 255
 256 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
 257 {
 258         int fd = -1, ret = -1;
 259
 260         fd = open(path, O_RDONLY);
 261         if (fd < 0)
 262                 goto out;
 263
 264         ssize_t len = read(fd, buf, buf_sz - 1);
 265         if (len < 0)
 266                 goto out;
 267
 268         ret = buf[len] = 0;
 269
 270 out:
 271         if (fd >= 0)
 272                 close(fd);
 273
 274         return ret;
 275 }
 276
 277 static int
 278 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
 279 {
 280         snprintf(dev_buf, sizeof(dev_buf), path, device);
 281         return system_get_sysctl(dev_buf, buf, buf_sz);
 282 }
 283
 284 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
 285 {
 286         return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
 287                         dev->ifname, buf, buf_sz);
 288 }
 289
 290 #ifndef IFF_LOWER_UP
 291 #define IFF_LOWER_UP    0x10000
 292 #endif
 293
 294 // Evaluate netlink messages
 295 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 296 {
 297         struct nlmsghdr *nh = nlmsg_hdr(msg);
 298         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 299         struct nlattr *nla[__IFLA_MAX];
 300
 301         if (nh->nlmsg_type != RTM_NEWLINK)
 302                 goto out;
 303
 304         nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
 305         if (!nla[IFLA_IFNAME])
 306                 goto out;
 307
 308         struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
 309         if (!dev)
 310                 goto out;
 311
 312         device_set_ifindex(dev, ifi->ifi_index);
 313         device_set_link(dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
 314
 315 out:
 316         return 0;
 317 }
 318
 319 static void
 320 handle_hotplug_msg(char *data, int size)
 321 {
 322         const char *subsystem = NULL, *interface = NULL;
 323         char *cur, *end, *sep;
 324         struct device *dev;
 325         int skip;
 326         bool add;
 327
 328         if (!strncmp(data, "add@", 4))
 329                 add = true;
 330         else if (!strncmp(data, "remove@", 7))
 331                 add = false;
 332         else
 333                 return;
 334
 335         skip = strlen(data) + 1;
 336         end = data + size;
 337
 338         for (cur = data + skip; cur < end; cur += skip) {
 339                 skip = strlen(cur) + 1;
 340
 341                 sep = strchr(cur, '=');
 342                 if (!sep)
 343                         continue;
 344
 345                 *sep = 0;
 346                 if (!strcmp(cur, "INTERFACE"))
 347                         interface = sep + 1;
 348                 else if (!strcmp(cur, "SUBSYSTEM")) {
 349                         subsystem = sep + 1;
 350                         if (strcmp(subsystem, "net") != 0)
 351                                 return;
 352                 }
 353                 if (subsystem && interface)
 354                         goto found;
 355         }
 356         return;
 357
 358 found:
 359         dev = device_get(interface, false);
 360         if (!dev)
 361                 return;
 362
 363         if (dev->type != &simple_device_type)
 364                 return;
 365
 366         if (add && system_if_force_external(dev->ifname))
 367                 return;
 368
 369         device_set_present(dev, add);
 370 }
 371
 372 static void
 373 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
 374 {
 375         struct event_socket *ev = container_of(u, struct event_socket, uloop);
 376         struct sockaddr_nl nla;
 377         unsigned char *buf = NULL;
 378         int size;
 379
 380         while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
 381                 if (nla.nl_pid == 0)
 382                         handle_hotplug_msg((char *) buf, size);
 383
 384                 free(buf);
 385         }
 386 }
 387
 388 static int system_rtnl_call(struct nl_msg *msg)
 389 {
 390         int ret;
 391
 392         ret = nl_send_auto_complete(sock_rtnl, msg);
 393         nlmsg_free(msg);
 394
 395         if (ret < 0)
 396                 return ret;
 397
 398         return nl_wait_for_ack(sock_rtnl);
 399 }
 400
 401 int system_bridge_delbr(struct device *bridge)
 402 {
 403         return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 404 }
 405
 406 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 407 {
 408         struct ifreq ifr;
 409
 410         memset(&ifr, 0, sizeof(ifr));
 411         if (dev)
 412                 ifr.ifr_ifindex = dev->ifindex;
 413         else
 414                 ifr.ifr_data = data;
 415         strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
 416         return ioctl(sock_ioctl, cmd, &ifr);
 417 }
 418
 419 static bool system_is_bridge(const char *name, char *buf, int buflen)
 420 {
 421         struct stat st;
 422
 423         snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
 424         if (stat(buf, &st) < 0)
 425                 return false;
 426
 427         return true;
 428 }
 429
 430 static char *system_get_bridge(const char *name, char *buf, int buflen)
 431 {
 432         char *path;
 433         ssize_t len;
 434         glob_t gl;
 435
 436         snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
 437         if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
 438                 return NULL;
 439
 440         if (gl.gl_pathc == 0)
 441                 return NULL;
 442
 443         len = readlink(gl.gl_pathv[0], buf, buflen);
 444         if (len < 0)
 445                 return NULL;
 446
 447         buf[len] = 0;
 448         path = strrchr(buf, '/');
 449         if (!path)
 450                 return NULL;
 451
 452         return path + 1;
 453 }
 454
 455 static void system_bridge_set_wireless(const char *bridge, const char *dev)
 456 {
 457         snprintf(dev_buf, sizeof(dev_buf),
 458                  "/sys/devices/virtual/net/%s/brif/%s/multicast_to_unicast",
 459                  bridge, dev);
 460         system_set_sysctl(dev_buf, "1");
 461 }
 462
 463 int system_bridge_addif(struct device *bridge, struct device *dev)
 464 {
 465         char *oldbr;
 466         int ret = 0;
 467
 468         oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
 469         if (!oldbr || strcmp(oldbr, bridge->ifname) != 0)
 470                 ret = system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 471
 472         if (dev->wireless)
 473                 system_bridge_set_wireless(bridge->ifname, dev->ifname);
 474
 475         return ret;
 476 }
 477
 478 int system_bridge_delif(struct device *bridge, struct device *dev)
 479 {
 480         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 481 }
 482
 483 static int system_if_resolve(struct device *dev)
 484 {
 485         struct ifreq ifr;
 486         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 487         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 488                 return ifr.ifr_ifindex;
 489         else
 490                 return 0;
 491 }
 492
 493 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 494 {
 495         struct ifreq ifr;
 496
 497         memset(&ifr, 0, sizeof(ifr));
 498         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 499         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 500         ifr.ifr_flags |= add;
 501         ifr.ifr_flags &= ~rem;
 502         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 503 }
 504
 505 struct clear_data {
 506         struct nl_msg *msg;
 507         struct device *dev;
 508         int type;
 509         int size;
 510         int af;
 511 };
 512
 513
 514 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 515 {
 516         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 517
 518         return ifa->ifa_index == ifindex;
 519 }
 520
 521 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 522 {
 523         struct rtmsg *r = NLMSG_DATA(hdr);
 524         struct nlattr *tb[__RTA_MAX];
 525
 526         if (r->rtm_protocol == RTPROT_KERNEL &&
 527             r->rtm_family == AF_INET6)
 528                 return false;
 529
 530         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 531         if (!tb[RTA_OIF])
 532                 return false;
 533
 534         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 535 }
 536
 537 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
 538 {
 539         return true;
 540 }
 541
 542 static int cb_clear_event(struct nl_msg *msg, void *arg)
 543 {
 544         struct clear_data *clr = arg;
 545         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 546         bool (*cb)(struct nlmsghdr *, int ifindex);
 547         int type;
 548
 549         switch(clr->type) {
 550         case RTM_GETADDR:
 551                 type = RTM_DELADDR;
 552                 if (hdr->nlmsg_type != RTM_NEWADDR)
 553                         return NL_SKIP;
 554
 555                 cb = check_ifaddr;
 556                 break;
 557         case RTM_GETROUTE:
 558                 type = RTM_DELROUTE;
 559                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 560                         return NL_SKIP;
 561
 562                 cb = check_route;
 563                 break;
 564         case RTM_GETRULE:
 565                 type = RTM_DELRULE;
 566                 if (hdr->nlmsg_type != RTM_NEWRULE)
 567                         return NL_SKIP;
 568
 569                 cb = check_rule;
 570                 break;
 571         default:
 572                 return NL_SKIP;
 573         }
 574
 575         if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
 576                 return NL_SKIP;
 577
 578         if (type == RTM_DELRULE)
 579                 D(SYSTEM, "Remove a rule\n");
 580         else
 581                 D(SYSTEM, "Remove %s from device %s\n",
 582                   type == RTM_DELADDR ? "an address" : "a route",
 583                   clr->dev->ifname);
 584         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 585         hdr = nlmsg_hdr(clr->msg);
 586         hdr->nlmsg_type = type;
 587         hdr->nlmsg_flags = NLM_F_REQUEST;
 588
 589         nl_socket_disable_auto_ack(sock_rtnl);
 590         nl_send_auto_complete(sock_rtnl, clr->msg);
 591         nl_socket_enable_auto_ack(sock_rtnl);
 592
 593         return NL_SKIP;
 594 }
 595
 596 static int
 597 cb_finish_event(struct nl_msg *msg, void *arg)
 598 {
 599         int *pending = arg;
 600         *pending = 0;
 601         return NL_STOP;
 602 }
 603
 604 static int
 605 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 606 {
 607         int *pending = arg;
 608         *pending = err->error;
 609         return NL_STOP;
 610 }
 611
 612 static void
 613 system_if_clear_entries(struct device *dev, int type, int af)
 614 {
 615         struct clear_data clr;
 616         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 617         struct rtmsg rtm = {
 618                 .rtm_family = af,
 619                 .rtm_flags = RTM_F_CLONED,
 620         };
 621         int flags = NLM_F_DUMP;
 622         int pending = 1;
 623
 624         clr.af = af;
 625         clr.dev = dev;
 626         clr.type = type;
 627         switch (type) {
 628         case RTM_GETADDR:
 629         case RTM_GETRULE:
 630                 clr.size = sizeof(struct rtgenmsg);
 631                 break;
 632         case RTM_GETROUTE:
 633                 clr.size = sizeof(struct rtmsg);
 634                 break;
 635         default:
 636                 return;
 637         }
 638
 639         if (!cb)
 640                 return;
 641
 642         clr.msg = nlmsg_alloc_simple(type, flags);
 643         if (!clr.msg)
 644                 goto out;
 645
 646         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 647         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 648         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 649         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 650
 651         nl_send_auto_complete(sock_rtnl, clr.msg);
 652         while (pending > 0)
 653                 nl_recvmsgs(sock_rtnl, cb);
 654
 655         nlmsg_free(clr.msg);
 656 out:
 657         nl_cb_put(cb);
 658 }
 659
 660 /*
 661  * Clear bridge (membership) state and bring down device
 662  */
 663 void system_if_clear_state(struct device *dev)
 664 {
 665         static char buf[256];
 666         char *bridge;
 667
 668         device_set_ifindex(dev, system_if_resolve(dev));
 669         if (dev->external || !dev->ifindex)
 670                 return;
 671
 672         system_if_flags(dev->ifname, 0, IFF_UP);
 673
 674         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 675                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 676                 system_bridge_delbr(dev);
 677                 return;
 678         }
 679
 680         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 681         if (bridge) {
 682                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 683                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 684         }
 685
 686         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 687         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 688         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 689         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 690         system_set_disable_ipv6(dev, "0");
 691 }
 692
 693 static inline unsigned long
 694 sec_to_jiffies(int val)
 695 {
 696         return (unsigned long) val * 100;
 697 }
 698
 699 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 700 {
 701         unsigned long args[4] = {};
 702
 703         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 704                 return -1;
 705
 706         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 707         args[1] = !!cfg->stp;
 708         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 709
 710         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 711         args[1] = sec_to_jiffies(cfg->forward_delay);
 712         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 713
 714         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 715                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 716
 717         args[0] = BRCTL_SET_BRIDGE_PRIORITY;
 718         args[1] = cfg->priority;
 719         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 720
 721         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 722                 args[0] = BRCTL_SET_AGEING_TIME;
 723                 args[1] = sec_to_jiffies(cfg->ageing_time);
 724                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 725         }
 726
 727         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 728                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 729                 args[1] = sec_to_jiffies(cfg->hello_time);
 730                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 731         }
 732
 733         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 734                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 735                 args[1] = sec_to_jiffies(cfg->max_age);
 736                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 737         }
 738
 739         return 0;
 740 }
 741
 742 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
 743 {
 744         struct nl_msg *msg;
 745         struct nlattr *linkinfo, *data;
 746         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
 747         int ifindex = system_if_resolve(dev);
 748         int i, rv;
 749         static const struct {
 750                 const char *name;
 751                 enum macvlan_mode val;
 752         } modes[] = {
 753                 { "private", MACVLAN_MODE_PRIVATE },
 754                 { "vepa", MACVLAN_MODE_VEPA },
 755                 { "bridge", MACVLAN_MODE_BRIDGE },
 756                 { "passthru", MACVLAN_MODE_PASSTHRU },
 757         };
 758
 759         if (ifindex == 0)
 760                 return -ENOENT;
 761
 762         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 763
 764         if (!msg)
 765                 return -1;
 766
 767         nlmsg_append(msg, &iim, sizeof(iim), 0);
 768
 769         if (cfg->flags & MACVLAN_OPT_MACADDR)
 770                 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
 771         nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
 772         nla_put_u32(msg, IFLA_LINK, ifindex);
 773
 774         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 775                 goto nla_put_failure;
 776
 777         nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
 778
 779         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 780                 goto nla_put_failure;
 781
 782         if (cfg->mode) {
 783                 for (i = 0; i < ARRAY_SIZE(modes); i++) {
 784                         if (strcmp(cfg->mode, modes[i].name) != 0)
 785                                 continue;
 786
 787                         nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
 788                         break;
 789                 }
 790         }
 791
 792         nla_nest_end(msg, data);
 793         nla_nest_end(msg, linkinfo);
 794
 795         rv = system_rtnl_call(msg);
 796         if (rv)
 797                 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
 798
 799         return rv;
 800
 801 nla_put_failure:
 802         nlmsg_free(msg);
 803         return -ENOMEM;
 804 }
 805
 806 static int system_link_del(struct device *dev)
 807 {
 808         struct nl_msg *msg;
 809         struct ifinfomsg iim = {
 810                 .ifi_family = AF_UNSPEC,
 811                 .ifi_index = 0,
 812         };
 813
 814         msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
 815
 816         if (!msg)
 817                 return -1;
 818
 819         nlmsg_append(msg, &iim, sizeof(iim), 0);
 820         nla_put_string(msg, IFLA_IFNAME, dev->ifname);
 821         return system_rtnl_call(msg);
 822 }
 823
 824 int system_macvlan_del(struct device *macvlan)
 825 {
 826         return system_link_del(macvlan);
 827 }
 828
 829 static int system_vlan(struct device *dev, int id)
 830 {
 831         struct vlan_ioctl_args ifr = {
 832                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 833                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 834         };
 835
 836         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 837
 838         if (id < 0) {
 839                 ifr.cmd = DEL_VLAN_CMD;
 840                 ifr.u.VID = 0;
 841         } else {
 842                 ifr.cmd = ADD_VLAN_CMD;
 843                 ifr.u.VID = id;
 844         }
 845         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 846         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 847 }
 848
 849 int system_vlan_add(struct device *dev, int id)
 850 {
 851         return system_vlan(dev, id);
 852 }
 853
 854 int system_vlan_del(struct device *dev)
 855 {
 856         return system_vlan(dev, -1);
 857 }
 858
 859 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
 860 {
 861         struct nl_msg *msg;
 862         struct nlattr *linkinfo, *data;
 863         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
 864         int ifindex = system_if_resolve(dev);
 865         int rv;
 866
 867         if (ifindex == 0)
 868                 return -ENOENT;
 869
 870         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 871
 872         if (!msg)
 873                 return -1;
 874
 875         nlmsg_append(msg, &iim, sizeof(iim), 0);
 876         nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
 877         nla_put_u32(msg, IFLA_LINK, ifindex);
 878
 879         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 880                 goto nla_put_failure;
 881
 882         nla_put_string(msg, IFLA_INFO_KIND, "vlan");
 883
 884         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 885                 goto nla_put_failure;
 886
 887         nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
 888
 889 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
 890         nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
 891 #else
 892         if(cfg->proto == VLAN_PROTO_8021AD)
 893                 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);
 894 #endif
 895
 896         nla_nest_end(msg, data);
 897         nla_nest_end(msg, linkinfo);
 898
 899         rv = system_rtnl_call(msg);
 900         if (rv)
 901                 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
 902
 903         return rv;
 904
 905 nla_put_failure:
 906         nlmsg_free(msg);
 907         return -ENOMEM;
 908 }
 909
 910 int system_vlandev_del(struct device *vlandev)
 911 {
 912         return system_link_del(vlandev);
 913 }
 914
 915 static void
 916 system_if_get_settings(struct device *dev, struct device_settings *s)
 917 {
 918         struct ifreq ifr;
 919         char buf[10];
 920
 921         memset(&ifr, 0, sizeof(ifr));
 922         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 923
 924         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 925                 s->mtu = ifr.ifr_mtu;
 926                 s->flags |= DEV_OPT_MTU;
 927         }
 928
 929         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 930                 s->txqueuelen = ifr.ifr_qlen;
 931                 s->flags |= DEV_OPT_TXQUEUELEN;
 932         }
 933
 934         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 935                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 936                 s->flags |= DEV_OPT_MACADDR;
 937         }
 938
 939         if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
 940                 s->ipv6 = !strtoul(buf, NULL, 0);
 941                 s->flags |= DEV_OPT_IPV6;
 942         }
 943 }
 944
 945 void
 946 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
 947 {
 948         struct ifreq ifr;
 949
 950         if (!apply_mask)
 951                 return;
 952
 953         memset(&ifr, 0, sizeof(ifr));
 954         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 955         if (s->flags & DEV_OPT_MTU & apply_mask) {
 956                 ifr.ifr_mtu = s->mtu;
 957                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 958                         s->flags &= ~DEV_OPT_MTU;
 959         }
 960         if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
 961                 ifr.ifr_qlen = s->txqueuelen;
 962                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 963                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 964         }
 965         if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
 966                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 967                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 968                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 969                         s->flags &= ~DEV_OPT_MACADDR;
 970         }
 971         if (s->flags & DEV_OPT_IPV6 & apply_mask)
 972                 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
 973 }
 974
 975 int system_if_up(struct device *dev)
 976 {
 977         system_if_get_settings(dev, &dev->orig_settings);
 978         system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
 979         device_set_ifindex(dev, system_if_resolve(dev));
 980         return system_if_flags(dev->ifname, IFF_UP, 0);
 981 }
 982
 983 int system_if_down(struct device *dev)
 984 {
 985         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
 986         dev->orig_settings.flags &= dev->settings.flags;
 987         system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
 988         return ret;
 989 }
 990
 991 struct if_check_data {
 992         struct device *dev;
 993         int pending;
 994         int ret;
 995 };
 996
 997 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
 998 {
 999         struct nlmsghdr *nh = nlmsg_hdr(msg);
1000         struct ifinfomsg *ifi = NLMSG_DATA(nh);
1001         struct if_check_data *chk = (struct if_check_data *)arg;
1002
1003         if (nh->nlmsg_type != RTM_NEWLINK)
1004                 return NL_SKIP;
1005
1006         device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1007         device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1008
1009         return NL_OK;
1010 }
1011
1012 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1013 {
1014         struct if_check_data *chk = (struct if_check_data *)arg;
1015         chk->pending = 0;
1016         return NL_STOP;
1017 }
1018
1019 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1020 {
1021         struct if_check_data *chk = (struct if_check_data *)arg;
1022
1023         device_set_present(chk->dev, false);
1024         device_set_link(chk->dev, false);
1025         chk->pending = err->error;
1026
1027         return NL_STOP;
1028 }
1029
1030 int system_if_check(struct device *dev)
1031 {
1032         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1033         struct nl_msg *msg;
1034         struct ifinfomsg ifi = {
1035                 .ifi_family = AF_UNSPEC,
1036                 .ifi_index = 0,
1037         };
1038         struct if_check_data chk = {
1039                 .dev = dev,
1040                 .pending = 1,
1041         };
1042         int ret = 1;
1043
1044         msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1045         if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1046             nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1047                 goto out;
1048
1049         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1050         nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1051         nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1052
1053         nl_send_auto_complete(sock_rtnl, msg);
1054         while (chk.pending > 0)
1055                 nl_recvmsgs(sock_rtnl, cb);
1056
1057         nlmsg_free(msg);
1058         ret = chk.pending;
1059
1060 out:
1061         nl_cb_put(cb);
1062         return ret;
1063 }
1064
1065 struct device *
1066 system_if_get_parent(struct device *dev)
1067 {
1068         char buf[64], *devname;
1069         int ifindex, iflink, len;
1070         FILE *f;
1071
1072         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1073         f = fopen(buf, "r");
1074         if (!f)
1075                 return NULL;
1076
1077         len = fread(buf, 1, sizeof(buf) - 1, f);
1078         fclose(f);
1079
1080         if (len <= 0)
1081                 return NULL;
1082
1083         buf[len] = 0;
1084         iflink = strtoul(buf, NULL, 0);
1085         ifindex = system_if_resolve(dev);
1086         if (!iflink || iflink == ifindex)
1087                 return NULL;
1088
1089         devname = if_indextoname(iflink, buf);
1090         if (!devname)
1091                 return NULL;
1092
1093         return device_get(devname, true);
1094 }
1095
1096 static bool
1097 read_string_file(int dir_fd, const char *file, char *buf, int len)
1098 {
1099         bool ret = false;
1100         char *c;
1101         int fd;
1102
1103         fd = openat(dir_fd, file, O_RDONLY);
1104         if (fd < 0)
1105                 return false;
1106
1107 retry:
1108         len = read(fd, buf, len - 1);
1109         if (len < 0) {
1110                 if (errno == EINTR)
1111                         goto retry;
1112         } else if (len > 0) {
1113                         buf[len] = 0;
1114
1115                         c = strchr(buf, '\n');
1116                         if (c)
1117                                 *c = 0;
1118
1119                         ret = true;
1120         }
1121
1122         close(fd);
1123
1124         return ret;
1125 }
1126
1127 static bool
1128 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1129 {
1130         char buf[64];
1131         bool ret = false;
1132
1133         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1134         if (ret)
1135                 *val = strtoull(buf, NULL, 0);
1136
1137         return ret;
1138 }
1139
1140 /* Assume advertised flags == supported flags */
1141 static const struct {
1142         uint32_t mask;
1143         const char *name;
1144 } ethtool_link_modes[] = {
1145         { ADVERTISED_10baseT_Half, "10H" },
1146         { ADVERTISED_10baseT_Full, "10F" },
1147         { ADVERTISED_100baseT_Half, "100H" },
1148         { ADVERTISED_100baseT_Full, "100F" },
1149         { ADVERTISED_1000baseT_Half, "1000H" },
1150         { ADVERTISED_1000baseT_Full, "1000F" },
1151 };
1152
1153 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1154 {
1155         int i;
1156         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1157                 if (mask & ethtool_link_modes[i].mask)
1158                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1159         }
1160 }
1161
1162 bool
1163 system_if_force_external(const char *ifname)
1164 {
1165         char buf[64];
1166         struct stat s;
1167
1168         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1169         return stat(buf, &s) == 0;
1170 }
1171
1172 int
1173 system_if_dump_info(struct device *dev, struct blob_buf *b)
1174 {
1175         struct ethtool_cmd ecmd;
1176         struct ifreq ifr;
1177         char buf[64], *s;
1178         void *c;
1179         int dir_fd;
1180
1181         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1182         dir_fd = open(buf, O_DIRECTORY);
1183
1184         memset(&ecmd, 0, sizeof(ecmd));
1185         memset(&ifr, 0, sizeof(ifr));
1186         strcpy(ifr.ifr_name, dev->ifname);
1187         ifr.ifr_data = (caddr_t) &ecmd;
1188         ecmd.cmd = ETHTOOL_GSET;
1189
1190         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1191                 c = blobmsg_open_array(b, "link-advertising");
1192                 system_add_link_modes(b, ecmd.advertising);
1193                 blobmsg_close_array(b, c);
1194
1195                 c = blobmsg_open_array(b, "link-supported");
1196                 system_add_link_modes(b, ecmd.supported);
1197                 blobmsg_close_array(b, c);
1198
1199                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1200                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1201                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1202                 blobmsg_add_string_buffer(b);
1203         }
1204
1205         close(dir_fd);
1206         return 0;
1207 }
1208
1209 int
1210 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1211 {
1212         const char *const counters[] = {
1213                 "collisions",     "rx_frame_errors",   "tx_compressed",
1214                 "multicast",      "rx_length_errors",  "tx_dropped",
1215                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
1216                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
1217                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
1218                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
1219                 "rx_errors",      "tx_bytes",          "tx_window_errors",
1220                 "rx_fifo_errors", "tx_carrier_errors",
1221         };
1222         char buf[64];
1223         int stats_dir;
1224         int i;
1225         uint64_t val = 0;
1226
1227         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1228         stats_dir = open(buf, O_DIRECTORY);
1229         if (stats_dir < 0)
1230                 return -1;
1231
1232         for (i = 0; i < ARRAY_SIZE(counters); i++)
1233                 if (read_uint64_file(stats_dir, counters[i], &val))
1234                         blobmsg_add_u64(b, counters[i], val);
1235
1236         close(stats_dir);
1237         return 0;
1238 }
1239
1240 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1241 {
1242         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1243         int alen = v4 ? 4 : 16;
1244         unsigned int flags = 0;
1245         struct ifaddrmsg ifa = {
1246                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1247                 .ifa_prefixlen = addr->mask,
1248                 .ifa_index = dev->ifindex,
1249         };
1250
1251         struct nl_msg *msg;
1252         if (cmd == RTM_NEWADDR)
1253                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1254
1255         msg = nlmsg_alloc_simple(cmd, flags);
1256         if (!msg)
1257                 return -1;
1258
1259         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1260         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1261         if (v4) {
1262                 if (addr->broadcast)
1263                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1264                 if (addr->point_to_point)
1265                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1266         } else {
1267                 time_t now = system_get_rtime();
1268                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1269
1270                 if (addr->preferred_until) {
1271                         int64_t preferred = addr->preferred_until - now;
1272                         if (preferred < 0)
1273                                 preferred = 0;
1274                         else if (preferred > UINT32_MAX)
1275                                 preferred = UINT32_MAX;
1276
1277                         cinfo.ifa_prefered = preferred;
1278                 }
1279
1280                 if (addr->valid_until) {
1281                         int64_t valid = addr->valid_until - now;
1282                         if (valid <= 0)
1283                                 return -1;
1284                         else if (valid > UINT32_MAX)
1285                                 valid = UINT32_MAX;
1286
1287                         cinfo.ifa_valid = valid;
1288                 }
1289
1290                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1291         }
1292
1293         return system_rtnl_call(msg);
1294 }
1295
1296 int system_add_address(struct device *dev, struct device_addr *addr)
1297 {
1298         return system_addr(dev, addr, RTM_NEWADDR);
1299 }
1300
1301 int system_del_address(struct device *dev, struct device_addr *addr)
1302 {
1303         return system_addr(dev, addr, RTM_DELADDR);
1304 }
1305
1306 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1307 {
1308         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1309         bool have_gw;
1310         unsigned int flags = 0;
1311
1312         if (alen == 4)
1313                 have_gw = !!route->nexthop.in.s_addr;
1314         else
1315                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1316                         route->nexthop.in6.s6_addr32[1] ||
1317                         route->nexthop.in6.s6_addr32[2] ||
1318                         route->nexthop.in6.s6_addr32[3];
1319
1320         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1321                         ? route->table : RT_TABLE_MAIN;
1322
1323         struct rtmsg rtm = {
1324                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1325                 .rtm_dst_len = route->mask,
1326                 .rtm_src_len = route->sourcemask,
1327                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1328                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1329                 .rtm_scope = RT_SCOPE_NOWHERE,
1330                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1331                 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1332         };
1333         struct nl_msg *msg;
1334
1335         if (cmd == RTM_NEWROUTE) {
1336                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1337
1338                 if (!dev) { // Add null-route
1339                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1340                         rtm.rtm_type = RTN_UNREACHABLE;
1341                 }
1342                 else
1343                         rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1344         }
1345
1346         if (route->flags & DEVROUTE_TYPE) {
1347                 rtm.rtm_type = route->type;
1348                 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1349                         if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1350                             rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1351                                 rtm.rtm_table = RT_TABLE_LOCAL;
1352                 }
1353
1354                 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT)
1355                         rtm.rtm_scope = RT_SCOPE_HOST;
1356                 else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1357                         rtm.rtm_type == RTN_ANYCAST)
1358                         rtm.rtm_scope = RT_SCOPE_LINK;
1359         }
1360
1361         msg = nlmsg_alloc_simple(cmd, flags);
1362         if (!msg)
1363                 return -1;
1364
1365         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1366
1367         if (route->mask)
1368                 nla_put(msg, RTA_DST, alen, &route->addr);
1369
1370         if (route->sourcemask)
1371                 nla_put(msg, RTA_SRC, alen, &route->source);
1372
1373         if (route->metric > 0)
1374                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1375
1376         if (have_gw)
1377                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1378
1379         if (dev)
1380                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1381
1382         if (table >= 256)
1383                 nla_put_u32(msg, RTA_TABLE, table);
1384
1385         if (route->flags & DEVROUTE_MTU) {
1386                 struct nlattr *metrics;
1387
1388                 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1389                         goto nla_put_failure;
1390
1391                 nla_put_u32(msg, RTAX_MTU, route->mtu);
1392
1393                 nla_nest_end(msg, metrics);
1394         }
1395
1396         return system_rtnl_call(msg);
1397
1398 nla_put_failure:
1399         nlmsg_free(msg);
1400         return -ENOMEM;
1401 }
1402
1403 int system_add_route(struct device *dev, struct device_route *route)
1404 {
1405         return system_rt(dev, route, RTM_NEWROUTE);
1406 }
1407
1408 int system_del_route(struct device *dev, struct device_route *route)
1409 {
1410         return system_rt(dev, route, RTM_DELROUTE);
1411 }
1412
1413 int system_flush_routes(void)
1414 {
1415         const char *names[] = {
1416                 "/proc/sys/net/ipv4/route/flush",
1417                 "/proc/sys/net/ipv6/route/flush"
1418         };
1419         int fd, i;
1420
1421         for (i = 0; i < ARRAY_SIZE(names); i++) {
1422                 fd = open(names[i], O_WRONLY);
1423                 if (fd < 0)
1424                         continue;
1425
1426                 if (write(fd, "-1", 2)) {}
1427                 close(fd);
1428         }
1429         return 0;
1430 }
1431
1432 bool system_resolve_rt_type(const char *type, unsigned int *id)
1433 {
1434         return system_rtn_aton(type, id);
1435 }
1436
1437 bool system_resolve_rt_table(const char *name, unsigned int *id)
1438 {
1439         FILE *f;
1440         char *e, buf[128];
1441         unsigned int n, table = RT_TABLE_UNSPEC;
1442
1443         /* first try to parse table as number */
1444         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1445                 table = n;
1446
1447         /* handle well known aliases */
1448         else if (!strcmp(name, "default"))
1449                 table = RT_TABLE_DEFAULT;
1450         else if (!strcmp(name, "main"))
1451                 table = RT_TABLE_MAIN;
1452         else if (!strcmp(name, "local"))
1453                 table = RT_TABLE_LOCAL;
1454
1455         /* try to look up name in /etc/iproute2/rt_tables */
1456         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1457         {
1458                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1459                 {
1460                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1461                                 continue;
1462
1463                         n = strtoul(e, NULL, 10);
1464                         e = strtok(NULL, " \t\n");
1465
1466                         if (e && !strcmp(e, name))
1467                         {
1468                                 table = n;
1469                                 break;
1470                         }
1471                 }
1472
1473                 fclose(f);
1474         }
1475
1476         if (table == RT_TABLE_UNSPEC)
1477                 return false;
1478
1479         *id = table;
1480         return true;
1481 }
1482
1483 bool system_is_default_rt_table(unsigned int id)
1484 {
1485         return (id == RT_TABLE_MAIN);
1486 }
1487
1488 static int system_iprule(struct iprule *rule, int cmd)
1489 {
1490         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1491
1492         struct nl_msg *msg;
1493         struct rtmsg rtm = {
1494                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1495                 .rtm_protocol = RTPROT_STATIC,
1496                 .rtm_scope = RT_SCOPE_UNIVERSE,
1497                 .rtm_table = RT_TABLE_UNSPEC,
1498                 .rtm_type = RTN_UNSPEC,
1499                 .rtm_flags = 0,
1500         };
1501
1502         if (cmd == RTM_NEWRULE) {
1503                 rtm.rtm_type = RTN_UNICAST;
1504                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1505         }
1506
1507         if (rule->invert)
1508                 rtm.rtm_flags |= FIB_RULE_INVERT;
1509
1510         if (rule->flags & IPRULE_SRC)
1511                 rtm.rtm_src_len = rule->src_mask;
1512
1513         if (rule->flags & IPRULE_DEST)
1514                 rtm.rtm_dst_len = rule->dest_mask;
1515
1516         if (rule->flags & IPRULE_TOS)
1517                 rtm.rtm_tos = rule->tos;
1518
1519         if (rule->flags & IPRULE_LOOKUP) {
1520                 if (rule->lookup < 256)
1521                         rtm.rtm_table = rule->lookup;
1522         }
1523
1524         if (rule->flags & IPRULE_ACTION)
1525                 rtm.rtm_type = rule->action;
1526         else if (rule->flags & IPRULE_GOTO)
1527                 rtm.rtm_type = FR_ACT_GOTO;
1528         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1529                 rtm.rtm_type = FR_ACT_NOP;
1530
1531         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1532
1533         if (!msg)
1534                 return -1;
1535
1536         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1537
1538         if (rule->flags & IPRULE_IN)
1539                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1540
1541         if (rule->flags & IPRULE_OUT)
1542                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1543
1544         if (rule->flags & IPRULE_SRC)
1545                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1546
1547         if (rule->flags & IPRULE_DEST)
1548                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1549
1550         if (rule->flags & IPRULE_PRIORITY)
1551                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1552         else if (cmd == RTM_NEWRULE)
1553                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1554
1555         if (rule->flags & IPRULE_FWMARK)
1556                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1557
1558         if (rule->flags & IPRULE_FWMASK)
1559                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1560
1561         if (rule->flags & IPRULE_LOOKUP) {
1562                 if (rule->lookup >= 256)
1563                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1564         }
1565
1566         if (rule->flags & IPRULE_GOTO)
1567                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1568
1569         return system_rtnl_call(msg);
1570 }
1571
1572 int system_add_iprule(struct iprule *rule)
1573 {
1574         return system_iprule(rule, RTM_NEWRULE);
1575 }
1576
1577 int system_del_iprule(struct iprule *rule)
1578 {
1579         return system_iprule(rule, RTM_DELRULE);
1580 }
1581
1582 int system_flush_iprules(void)
1583 {
1584         int rv = 0;
1585         struct iprule rule;
1586
1587         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1588         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1589
1590         memset(&rule, 0, sizeof(rule));
1591
1592
1593         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1594
1595         rule.priority = 0;
1596         rule.lookup = RT_TABLE_LOCAL;
1597         rv |= system_iprule(&rule, RTM_NEWRULE);
1598
1599         rule.priority = 32766;
1600         rule.lookup = RT_TABLE_MAIN;
1601         rv |= system_iprule(&rule, RTM_NEWRULE);
1602
1603         rule.priority = 32767;
1604         rule.lookup = RT_TABLE_DEFAULT;
1605         rv |= system_iprule(&rule, RTM_NEWRULE);
1606
1607
1608         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1609
1610         rule.priority = 0;
1611         rule.lookup = RT_TABLE_LOCAL;
1612         rv |= system_iprule(&rule, RTM_NEWRULE);
1613
1614         rule.priority = 32766;
1615         rule.lookup = RT_TABLE_MAIN;
1616         rv |= system_iprule(&rule, RTM_NEWRULE);
1617
1618         return rv;
1619 }
1620
1621 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1622 {
1623         return system_rtn_aton(action, id);
1624 }
1625
1626 time_t system_get_rtime(void)
1627 {
1628         struct timespec ts;
1629         struct timeval tv;
1630
1631         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1632                 return ts.tv_sec;
1633
1634         if (gettimeofday(&tv, NULL) == 0)
1635                 return tv.tv_sec;
1636
1637         return 0;
1638 }
1639
1640 #ifndef IP_DF
1641 #define IP_DF       0x4000
1642 #endif
1643
1644 static int tunnel_ioctl(const char *name, int cmd, void *p)
1645 {
1646         struct ifreq ifr;
1647
1648         memset(&ifr, 0, sizeof(ifr));
1649         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1650         ifr.ifr_ifru.ifru_data = p;
1651         return ioctl(sock_ioctl, cmd, &ifr);
1652 }
1653
1654 int system_del_ip_tunnel(const char *name)
1655 {
1656         return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1657 }
1658
1659 int system_update_ipv6_mtu(struct device *dev, int mtu)
1660 {
1661         int ret = -1;
1662         char buf[64];
1663         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1664                         dev->ifname);
1665
1666         int fd = open(buf, O_RDWR);
1667         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1668         if (len < 0)
1669                 goto out;
1670
1671         buf[len] = 0;
1672         ret = atoi(buf);
1673
1674         if (!mtu || ret <= mtu)
1675                 goto out;
1676
1677         lseek(fd, 0, SEEK_SET);
1678         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1679                 ret = -1;
1680
1681 out:
1682         close(fd);
1683         return ret;
1684 }
1685
1686 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1687 {
1688         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1689         struct blob_attr *cur;
1690         bool set_df = true;
1691         const char *str;
1692
1693         system_del_ip_tunnel(name);
1694
1695         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1696                 blob_data(attr), blob_len(attr));
1697
1698         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1699                 return -EINVAL;
1700         str = blobmsg_data(cur);
1701
1702         if ((cur = tb[TUNNEL_ATTR_DF]))
1703                 set_df = blobmsg_get_bool(cur);
1704
1705         unsigned int ttl = 0;
1706         if ((cur = tb[TUNNEL_ATTR_TTL])) {
1707                 ttl = blobmsg_get_u32(cur);
1708                 if (ttl > 255 || (!set_df && ttl))
1709                         return -EINVAL;
1710         }
1711
1712         unsigned int link = 0;
1713         if ((cur = tb[TUNNEL_ATTR_LINK])) {
1714                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
1715                 if (!iface)
1716                         return -EINVAL;
1717
1718                 if (iface->l3_dev.dev)
1719                         link = iface->l3_dev.dev->ifindex;
1720         }
1721
1722         if (!strcmp(str, "sit")) {
1723                 struct ip_tunnel_parm p = {
1724                         .link = link,
1725                         .iph = {
1726                                 .version = 4,
1727                                 .ihl = 5,
1728                                 .frag_off = set_df ? htons(IP_DF) : 0,
1729                                 .protocol = IPPROTO_IPV6,
1730                                 .ttl = ttl
1731                         }
1732                 };
1733
1734                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1735                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1736                         return -EINVAL;
1737
1738                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1739                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1740                         return -EINVAL;
1741
1742                 strncpy(p.name, name, sizeof(p.name));
1743                 if (tunnel_ioctl("sit0", SIOCADDTUNNEL, &p) < 0)
1744                         return -1;
1745
1746 #ifdef SIOCADD6RD
1747                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
1748                         unsigned int mask;
1749                         struct ip_tunnel_6rd p6;
1750
1751                         memset(&p6, 0, sizeof(p6));
1752
1753                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
1754                                                 &p6.prefix, &mask) || mask > 128)
1755                                 return -EINVAL;
1756                         p6.prefixlen = mask;
1757
1758                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
1759                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
1760                                                         &p6.relay_prefix, &mask) || mask > 32)
1761                                         return -EINVAL;
1762                                 p6.relay_prefixlen = mask;
1763                         }
1764
1765                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
1766                                 system_del_ip_tunnel(name);
1767                                 return -1;
1768                         }
1769                 }
1770 #endif
1771         } else if (!strcmp(str, "ipip6")) {
1772                 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
1773                                 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1774                 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
1775                 int ret = 0;
1776
1777                 if (!nlm)
1778                         return -1;
1779
1780                 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1781                 nla_put_string(nlm, IFLA_IFNAME, name);
1782
1783                 if (link)
1784                         nla_put_u32(nlm, IFLA_LINK, link);
1785
1786                 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1787                 if (!linkinfo) {
1788                         ret = -ENOMEM;
1789                         goto failure;
1790                 }
1791                 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
1792                 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1793                 if (!infodata) {
1794                         ret = -ENOMEM;
1795                         goto failure;
1796                 }
1797
1798                 if (link)
1799                         nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
1800
1801                 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
1802                 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
1803                 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
1804
1805                 struct in6_addr in6buf;
1806                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1807                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1808                                 ret = -EINVAL;
1809                                 goto failure;
1810                         }
1811                         nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
1812                 }
1813
1814                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1815                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1816                                 ret = -EINVAL;
1817                                 goto failure;
1818                         }
1819                         nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
1820                 }
1821
1822 #ifdef IFLA_IPTUN_FMR_MAX
1823                 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
1824                         struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
1825
1826                         struct blob_attr *fmr;
1827                         unsigned rem, fmrcnt = 0;
1828                         blobmsg_for_each_attr(fmr, cur, rem) {
1829                                 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
1830                                         continue;
1831
1832                                 unsigned ip4len, ip6len, ealen, offset = 6;
1833                                 char ip6buf[48];
1834                                 char ip4buf[16];
1835
1836                                 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
1837                                                 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
1838                                         ret = -EINVAL;
1839                                         goto failure;
1840                                 }
1841
1842                                 struct in6_addr ip6prefix;
1843                                 struct in_addr ip4prefix;
1844                                 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
1845                                                 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
1846                                         ret = -EINVAL;
1847                                         goto failure;
1848                                 }
1849
1850                                 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
1851
1852                                 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
1853                                 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
1854                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
1855                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
1856                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
1857                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
1858
1859                                 nla_nest_end(nlm, rule);
1860                         }
1861
1862                         nla_nest_end(nlm, fmrs);
1863                 }
1864 #endif
1865
1866                 nla_nest_end(nlm, infodata);
1867                 nla_nest_end(nlm, linkinfo);
1868
1869                 return system_rtnl_call(nlm);
1870 failure:
1871                 nlmsg_free(nlm);
1872                 return ret;
1873         }
1874         else
1875                 return -EINVAL;
1876
1877         return 0;
1878 }