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 int system_bridge_addif(struct device *bridge, struct device *dev)
 456 {
 457         char *oldbr;
 458
 459         oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
 460         if (oldbr && !strcmp(oldbr, bridge->ifname))
 461                 return 0;
 462
 463         return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 464 }
 465
 466 int system_bridge_delif(struct device *bridge, struct device *dev)
 467 {
 468         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 469 }
 470
 471 static int system_if_resolve(struct device *dev)
 472 {
 473         struct ifreq ifr;
 474         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 475         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 476                 return ifr.ifr_ifindex;
 477         else
 478                 return 0;
 479 }
 480
 481 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 482 {
 483         struct ifreq ifr;
 484
 485         memset(&ifr, 0, sizeof(ifr));
 486         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 487         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 488         ifr.ifr_flags |= add;
 489         ifr.ifr_flags &= ~rem;
 490         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 491 }
 492
 493 struct clear_data {
 494         struct nl_msg *msg;
 495         struct device *dev;
 496         int type;
 497         int size;
 498         int af;
 499 };
 500
 501
 502 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 503 {
 504         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 505
 506         return ifa->ifa_index == ifindex;
 507 }
 508
 509 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 510 {
 511         struct rtmsg *r = NLMSG_DATA(hdr);
 512         struct nlattr *tb[__RTA_MAX];
 513
 514         if (r->rtm_protocol == RTPROT_KERNEL &&
 515             r->rtm_family == AF_INET6)
 516                 return false;
 517
 518         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 519         if (!tb[RTA_OIF])
 520                 return false;
 521
 522         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 523 }
 524
 525 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
 526 {
 527         return true;
 528 }
 529
 530 static int cb_clear_event(struct nl_msg *msg, void *arg)
 531 {
 532         struct clear_data *clr = arg;
 533         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 534         bool (*cb)(struct nlmsghdr *, int ifindex);
 535         int type;
 536
 537         switch(clr->type) {
 538         case RTM_GETADDR:
 539                 type = RTM_DELADDR;
 540                 if (hdr->nlmsg_type != RTM_NEWADDR)
 541                         return NL_SKIP;
 542
 543                 cb = check_ifaddr;
 544                 break;
 545         case RTM_GETROUTE:
 546                 type = RTM_DELROUTE;
 547                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 548                         return NL_SKIP;
 549
 550                 cb = check_route;
 551                 break;
 552         case RTM_GETRULE:
 553                 type = RTM_DELRULE;
 554                 if (hdr->nlmsg_type != RTM_NEWRULE)
 555                         return NL_SKIP;
 556
 557                 cb = check_rule;
 558                 break;
 559         default:
 560                 return NL_SKIP;
 561         }
 562
 563         if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
 564                 return NL_SKIP;
 565
 566         if (type == RTM_DELRULE)
 567                 D(SYSTEM, "Remove a rule\n");
 568         else
 569                 D(SYSTEM, "Remove %s from device %s\n",
 570                   type == RTM_DELADDR ? "an address" : "a route",
 571                   clr->dev->ifname);
 572         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 573         hdr = nlmsg_hdr(clr->msg);
 574         hdr->nlmsg_type = type;
 575         hdr->nlmsg_flags = NLM_F_REQUEST;
 576
 577         nl_socket_disable_auto_ack(sock_rtnl);
 578         nl_send_auto_complete(sock_rtnl, clr->msg);
 579         nl_socket_enable_auto_ack(sock_rtnl);
 580
 581         return NL_SKIP;
 582 }
 583
 584 static int
 585 cb_finish_event(struct nl_msg *msg, void *arg)
 586 {
 587         int *pending = arg;
 588         *pending = 0;
 589         return NL_STOP;
 590 }
 591
 592 static int
 593 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 594 {
 595         int *pending = arg;
 596         *pending = err->error;
 597         return NL_STOP;
 598 }
 599
 600 static void
 601 system_if_clear_entries(struct device *dev, int type, int af)
 602 {
 603         struct clear_data clr;
 604         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 605         struct rtmsg rtm = {
 606                 .rtm_family = af,
 607                 .rtm_flags = RTM_F_CLONED,
 608         };
 609         int flags = NLM_F_DUMP;
 610         int pending = 1;
 611
 612         clr.af = af;
 613         clr.dev = dev;
 614         clr.type = type;
 615         switch (type) {
 616         case RTM_GETADDR:
 617         case RTM_GETRULE:
 618                 clr.size = sizeof(struct rtgenmsg);
 619                 break;
 620         case RTM_GETROUTE:
 621                 clr.size = sizeof(struct rtmsg);
 622                 break;
 623         default:
 624                 return;
 625         }
 626
 627         if (!cb)
 628                 return;
 629
 630         clr.msg = nlmsg_alloc_simple(type, flags);
 631         if (!clr.msg)
 632                 goto out;
 633
 634         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 635         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 636         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 637         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 638
 639         nl_send_auto_complete(sock_rtnl, clr.msg);
 640         while (pending > 0)
 641                 nl_recvmsgs(sock_rtnl, cb);
 642
 643         nlmsg_free(clr.msg);
 644 out:
 645         nl_cb_put(cb);
 646 }
 647
 648 /*
 649  * Clear bridge (membership) state and bring down device
 650  */
 651 void system_if_clear_state(struct device *dev)
 652 {
 653         static char buf[256];
 654         char *bridge;
 655
 656         device_set_ifindex(dev, system_if_resolve(dev));
 657         if (dev->external || !dev->ifindex)
 658                 return;
 659
 660         system_if_flags(dev->ifname, 0, IFF_UP);
 661
 662         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 663                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 664                 system_bridge_delbr(dev);
 665                 return;
 666         }
 667
 668         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 669         if (bridge) {
 670                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 671                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 672         }
 673
 674         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 675         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 676         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 677         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 678         system_set_disable_ipv6(dev, "0");
 679 }
 680
 681 static inline unsigned long
 682 sec_to_jiffies(int val)
 683 {
 684         return (unsigned long) val * 100;
 685 }
 686
 687 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 688 {
 689         unsigned long args[4] = {};
 690
 691         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 692                 return -1;
 693
 694         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 695         args[1] = !!cfg->stp;
 696         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 697
 698         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 699         args[1] = sec_to_jiffies(cfg->forward_delay);
 700         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 701
 702         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 703                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 704
 705         args[0] = BRCTL_SET_BRIDGE_PRIORITY;
 706         args[1] = cfg->priority;
 707         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 708
 709         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 710                 args[0] = BRCTL_SET_AGEING_TIME;
 711                 args[1] = sec_to_jiffies(cfg->ageing_time);
 712                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 713         }
 714
 715         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 716                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 717                 args[1] = sec_to_jiffies(cfg->hello_time);
 718                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 719         }
 720
 721         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 722                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 723                 args[1] = sec_to_jiffies(cfg->max_age);
 724                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 725         }
 726
 727         return 0;
 728 }
 729
 730 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
 731 {
 732         struct nl_msg *msg;
 733         struct nlattr *linkinfo, *data;
 734         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
 735         int ifindex = system_if_resolve(dev);
 736         int i, rv;
 737         static const struct {
 738                 const char *name;
 739                 enum macvlan_mode val;
 740         } modes[] = {
 741                 { "private", MACVLAN_MODE_PRIVATE },
 742                 { "vepa", MACVLAN_MODE_VEPA },
 743                 { "bridge", MACVLAN_MODE_BRIDGE },
 744                 { "passthru", MACVLAN_MODE_PASSTHRU },
 745         };
 746
 747         if (ifindex == 0)
 748                 return -ENOENT;
 749
 750         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 751
 752         if (!msg)
 753                 return -1;
 754
 755         nlmsg_append(msg, &iim, sizeof(iim), 0);
 756
 757         if (cfg->flags & MACVLAN_OPT_MACADDR)
 758                 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
 759         nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
 760         nla_put_u32(msg, IFLA_LINK, ifindex);
 761
 762         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 763                 goto nla_put_failure;
 764
 765         nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
 766
 767         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 768                 goto nla_put_failure;
 769
 770         if (cfg->mode) {
 771                 for (i = 0; i < ARRAY_SIZE(modes); i++) {
 772                         if (strcmp(cfg->mode, modes[i].name) != 0)
 773                                 continue;
 774
 775                         nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
 776                         break;
 777                 }
 778         }
 779
 780         nla_nest_end(msg, data);
 781         nla_nest_end(msg, linkinfo);
 782
 783         rv = system_rtnl_call(msg);
 784         if (rv)
 785                 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
 786
 787         return rv;
 788
 789 nla_put_failure:
 790         nlmsg_free(msg);
 791         return -ENOMEM;
 792 }
 793
 794 int system_macvlan_del(struct device *macvlan)
 795 {
 796         struct nl_msg *msg;
 797         struct ifinfomsg iim = {
 798                 .ifi_family = AF_UNSPEC,
 799                 .ifi_index = 0,
 800         };
 801
 802         msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
 803
 804         if (!msg)
 805                 return -1;
 806
 807         nlmsg_append(msg, &iim, sizeof(iim), 0);
 808
 809         nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
 810
 811         system_rtnl_call(msg);
 812
 813         return 0;
 814 }
 815
 816 static int system_vlan(struct device *dev, int id)
 817 {
 818         struct vlan_ioctl_args ifr = {
 819                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 820                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 821         };
 822
 823         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 824
 825         if (id < 0) {
 826                 ifr.cmd = DEL_VLAN_CMD;
 827                 ifr.u.VID = 0;
 828         } else {
 829                 ifr.cmd = ADD_VLAN_CMD;
 830                 ifr.u.VID = id;
 831         }
 832         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 833         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 834 }
 835
 836 int system_vlan_add(struct device *dev, int id)
 837 {
 838         return system_vlan(dev, id);
 839 }
 840
 841 int system_vlan_del(struct device *dev)
 842 {
 843         return system_vlan(dev, -1);
 844 }
 845
 846 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
 847 {
 848         struct nl_msg *msg;
 849         struct nlattr *linkinfo, *data;
 850         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
 851         int ifindex = system_if_resolve(dev);
 852         int rv;
 853
 854         if (ifindex == 0)
 855                 return -ENOENT;
 856
 857         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 858
 859         if (!msg)
 860                 return -1;
 861
 862         nlmsg_append(msg, &iim, sizeof(iim), 0);
 863         nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
 864         nla_put_u32(msg, IFLA_LINK, ifindex);
 865
 866         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 867                 goto nla_put_failure;
 868
 869         nla_put_string(msg, IFLA_INFO_KIND, "vlan");
 870
 871         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 872                 goto nla_put_failure;
 873
 874         nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
 875
 876 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
 877         nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
 878 #else
 879         if(cfg->proto == VLAN_PROTO_8021AD)
 880                 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);
 881 #endif
 882
 883         nla_nest_end(msg, data);
 884         nla_nest_end(msg, linkinfo);
 885
 886         rv = system_rtnl_call(msg);
 887         if (rv)
 888                 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
 889
 890         return rv;
 891
 892 nla_put_failure:
 893         nlmsg_free(msg);
 894         return -ENOMEM;
 895 }
 896
 897 int system_vlandev_del(struct device *vlandev)
 898 {
 899         struct nl_msg *msg;
 900         struct ifinfomsg iim = {
 901                 .ifi_family = AF_UNSPEC,
 902                 .ifi_index = 0,
 903         };
 904
 905         msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
 906
 907         if (!msg)
 908                 return -1;
 909
 910         nlmsg_append(msg, &iim, sizeof(iim), 0);
 911
 912         nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
 913
 914         system_rtnl_call(msg);
 915
 916         return 0;
 917 }
 918
 919 static void
 920 system_if_get_settings(struct device *dev, struct device_settings *s)
 921 {
 922         struct ifreq ifr;
 923         char buf[10];
 924
 925         memset(&ifr, 0, sizeof(ifr));
 926         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 927
 928         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 929                 s->mtu = ifr.ifr_mtu;
 930                 s->flags |= DEV_OPT_MTU;
 931         }
 932
 933         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 934                 s->txqueuelen = ifr.ifr_qlen;
 935                 s->flags |= DEV_OPT_TXQUEUELEN;
 936         }
 937
 938         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 939                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 940                 s->flags |= DEV_OPT_MACADDR;
 941         }
 942
 943         if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
 944                 s->ipv6 = !strtoul(buf, NULL, 0);
 945                 s->flags |= DEV_OPT_IPV6;
 946         }
 947 }
 948
 949 void
 950 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
 951 {
 952         struct ifreq ifr;
 953
 954         if (!apply_mask)
 955                 return;
 956
 957         memset(&ifr, 0, sizeof(ifr));
 958         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 959         if (s->flags & DEV_OPT_MTU & apply_mask) {
 960                 ifr.ifr_mtu = s->mtu;
 961                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 962                         s->flags &= ~DEV_OPT_MTU;
 963         }
 964         if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
 965                 ifr.ifr_qlen = s->txqueuelen;
 966                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 967                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 968         }
 969         if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
 970                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 971                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 972                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 973                         s->flags &= ~DEV_OPT_MACADDR;
 974         }
 975         if (s->flags & DEV_OPT_IPV6 & apply_mask)
 976                 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
 977 }
 978
 979 int system_if_up(struct device *dev)
 980 {
 981         system_if_get_settings(dev, &dev->orig_settings);
 982         system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
 983         device_set_ifindex(dev, system_if_resolve(dev));
 984         return system_if_flags(dev->ifname, IFF_UP, 0);
 985 }
 986
 987 int system_if_down(struct device *dev)
 988 {
 989         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
 990         dev->orig_settings.flags &= dev->settings.flags;
 991         system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
 992         return ret;
 993 }
 994
 995 struct if_check_data {
 996         struct device *dev;
 997         int pending;
 998         int ret;
 999 };
1000
1001 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
1002 {
1003         struct nlmsghdr *nh = nlmsg_hdr(msg);
1004         struct ifinfomsg *ifi = NLMSG_DATA(nh);
1005         struct if_check_data *chk = (struct if_check_data *)arg;
1006
1007         if (nh->nlmsg_type != RTM_NEWLINK)
1008                 return NL_SKIP;
1009
1010         device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1011         device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1012
1013         return NL_OK;
1014 }
1015
1016 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1017 {
1018         struct if_check_data *chk = (struct if_check_data *)arg;
1019         chk->pending = 0;
1020         return NL_STOP;
1021 }
1022
1023 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1024 {
1025         struct if_check_data *chk = (struct if_check_data *)arg;
1026
1027         device_set_present(chk->dev, false);
1028         device_set_link(chk->dev, false);
1029         chk->pending = err->error;
1030
1031         return NL_STOP;
1032 }
1033
1034 int system_if_check(struct device *dev)
1035 {
1036         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1037         struct nl_msg *msg;
1038         struct ifinfomsg ifi = {
1039                 .ifi_family = AF_UNSPEC,
1040                 .ifi_index = 0,
1041         };
1042         struct if_check_data chk = {
1043                 .dev = dev,
1044                 .pending = 1,
1045         };
1046         int ret = 1;
1047
1048         msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1049         if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1050             nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1051                 goto out;
1052
1053         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1054         nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1055         nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1056
1057         nl_send_auto_complete(sock_rtnl, msg);
1058         while (chk.pending > 0)
1059                 nl_recvmsgs(sock_rtnl, cb);
1060
1061         nlmsg_free(msg);
1062         ret = chk.pending;
1063
1064 out:
1065         nl_cb_put(cb);
1066         return ret;
1067 }
1068
1069 struct device *
1070 system_if_get_parent(struct device *dev)
1071 {
1072         char buf[64], *devname;
1073         int ifindex, iflink, len;
1074         FILE *f;
1075
1076         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1077         f = fopen(buf, "r");
1078         if (!f)
1079                 return NULL;
1080
1081         len = fread(buf, 1, sizeof(buf) - 1, f);
1082         fclose(f);
1083
1084         if (len <= 0)
1085                 return NULL;
1086
1087         buf[len] = 0;
1088         iflink = strtoul(buf, NULL, 0);
1089         ifindex = system_if_resolve(dev);
1090         if (!iflink || iflink == ifindex)
1091                 return NULL;
1092
1093         devname = if_indextoname(iflink, buf);
1094         if (!devname)
1095                 return NULL;
1096
1097         return device_get(devname, true);
1098 }
1099
1100 static bool
1101 read_string_file(int dir_fd, const char *file, char *buf, int len)
1102 {
1103         bool ret = false;
1104         char *c;
1105         int fd;
1106
1107         fd = openat(dir_fd, file, O_RDONLY);
1108         if (fd < 0)
1109                 return false;
1110
1111 retry:
1112         len = read(fd, buf, len - 1);
1113         if (len < 0) {
1114                 if (errno == EINTR)
1115                         goto retry;
1116         } else if (len > 0) {
1117                         buf[len] = 0;
1118
1119                         c = strchr(buf, '\n');
1120                         if (c)
1121                                 *c = 0;
1122
1123                         ret = true;
1124         }
1125
1126         close(fd);
1127
1128         return ret;
1129 }
1130
1131 static bool
1132 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1133 {
1134         char buf[64];
1135         bool ret = false;
1136
1137         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1138         if (ret)
1139                 *val = strtoull(buf, NULL, 0);
1140
1141         return ret;
1142 }
1143
1144 /* Assume advertised flags == supported flags */
1145 static const struct {
1146         uint32_t mask;
1147         const char *name;
1148 } ethtool_link_modes[] = {
1149         { ADVERTISED_10baseT_Half, "10H" },
1150         { ADVERTISED_10baseT_Full, "10F" },
1151         { ADVERTISED_100baseT_Half, "100H" },
1152         { ADVERTISED_100baseT_Full, "100F" },
1153         { ADVERTISED_1000baseT_Half, "1000H" },
1154         { ADVERTISED_1000baseT_Full, "1000F" },
1155 };
1156
1157 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1158 {
1159         int i;
1160         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1161                 if (mask & ethtool_link_modes[i].mask)
1162                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1163         }
1164 }
1165
1166 bool
1167 system_if_force_external(const char *ifname)
1168 {
1169         char buf[64];
1170         struct stat s;
1171
1172         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1173         return stat(buf, &s) == 0;
1174 }
1175
1176 int
1177 system_if_dump_info(struct device *dev, struct blob_buf *b)
1178 {
1179         struct ethtool_cmd ecmd;
1180         struct ifreq ifr;
1181         char buf[64], *s;
1182         void *c;
1183         int dir_fd;
1184
1185         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1186         dir_fd = open(buf, O_DIRECTORY);
1187
1188         memset(&ecmd, 0, sizeof(ecmd));
1189         memset(&ifr, 0, sizeof(ifr));
1190         strcpy(ifr.ifr_name, dev->ifname);
1191         ifr.ifr_data = (caddr_t) &ecmd;
1192         ecmd.cmd = ETHTOOL_GSET;
1193
1194         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1195                 c = blobmsg_open_array(b, "link-advertising");
1196                 system_add_link_modes(b, ecmd.advertising);
1197                 blobmsg_close_array(b, c);
1198
1199                 c = blobmsg_open_array(b, "link-supported");
1200                 system_add_link_modes(b, ecmd.supported);
1201                 blobmsg_close_array(b, c);
1202
1203                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1204                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1205                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1206                 blobmsg_add_string_buffer(b);
1207         }
1208
1209         close(dir_fd);
1210         return 0;
1211 }
1212
1213 int
1214 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1215 {
1216         const char *const counters[] = {
1217                 "collisions",     "rx_frame_errors",   "tx_compressed",
1218                 "multicast",      "rx_length_errors",  "tx_dropped",
1219                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
1220                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
1221                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
1222                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
1223                 "rx_errors",      "tx_bytes",          "tx_window_errors",
1224                 "rx_fifo_errors", "tx_carrier_errors",
1225         };
1226         char buf[64];
1227         int stats_dir;
1228         int i;
1229         uint64_t val = 0;
1230
1231         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1232         stats_dir = open(buf, O_DIRECTORY);
1233         if (stats_dir < 0)
1234                 return -1;
1235
1236         for (i = 0; i < ARRAY_SIZE(counters); i++)
1237                 if (read_uint64_file(stats_dir, counters[i], &val))
1238                         blobmsg_add_u64(b, counters[i], val);
1239
1240         close(stats_dir);
1241         return 0;
1242 }
1243
1244 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1245 {
1246         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1247         int alen = v4 ? 4 : 16;
1248         unsigned int flags = 0;
1249         struct ifaddrmsg ifa = {
1250                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1251                 .ifa_prefixlen = addr->mask,
1252                 .ifa_index = dev->ifindex,
1253         };
1254
1255         struct nl_msg *msg;
1256         if (cmd == RTM_NEWADDR)
1257                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1258
1259         msg = nlmsg_alloc_simple(cmd, flags);
1260         if (!msg)
1261                 return -1;
1262
1263         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1264         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1265         if (v4) {
1266                 if (addr->broadcast)
1267                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1268                 if (addr->point_to_point)
1269                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1270         } else {
1271                 time_t now = system_get_rtime();
1272                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1273
1274                 if (addr->preferred_until) {
1275                         int64_t preferred = addr->preferred_until - now;
1276                         if (preferred < 0)
1277                                 preferred = 0;
1278                         else if (preferred > UINT32_MAX)
1279                                 preferred = UINT32_MAX;
1280
1281                         cinfo.ifa_prefered = preferred;
1282                 }
1283
1284                 if (addr->valid_until) {
1285                         int64_t valid = addr->valid_until - now;
1286                         if (valid <= 0)
1287                                 return -1;
1288                         else if (valid > UINT32_MAX)
1289                                 valid = UINT32_MAX;
1290
1291                         cinfo.ifa_valid = valid;
1292                 }
1293
1294                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1295         }
1296
1297         return system_rtnl_call(msg);
1298 }
1299
1300 int system_add_address(struct device *dev, struct device_addr *addr)
1301 {
1302         return system_addr(dev, addr, RTM_NEWADDR);
1303 }
1304
1305 int system_del_address(struct device *dev, struct device_addr *addr)
1306 {
1307         return system_addr(dev, addr, RTM_DELADDR);
1308 }
1309
1310 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1311 {
1312         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1313         bool have_gw;
1314         unsigned int flags = 0;
1315
1316         if (alen == 4)
1317                 have_gw = !!route->nexthop.in.s_addr;
1318         else
1319                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1320                         route->nexthop.in6.s6_addr32[1] ||
1321                         route->nexthop.in6.s6_addr32[2] ||
1322                         route->nexthop.in6.s6_addr32[3];
1323
1324         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1325                         ? route->table : RT_TABLE_MAIN;
1326
1327         struct rtmsg rtm = {
1328                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1329                 .rtm_dst_len = route->mask,
1330                 .rtm_src_len = route->sourcemask,
1331                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1332                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1333                 .rtm_scope = RT_SCOPE_NOWHERE,
1334                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1335                 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1336         };
1337         struct nl_msg *msg;
1338
1339         if (cmd == RTM_NEWROUTE) {
1340                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1341
1342                 if (!dev) { // Add null-route
1343                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1344                         rtm.rtm_type = RTN_UNREACHABLE;
1345                 }
1346                 else
1347                         rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1348         }
1349
1350         if (route->flags & DEVROUTE_TYPE) {
1351                 rtm.rtm_type = route->type;
1352                 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1353                         if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1354                             rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1355                                 rtm.rtm_table = RT_TABLE_LOCAL;
1356                 }
1357
1358                 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT)
1359                         rtm.rtm_scope = RT_SCOPE_HOST;
1360                 else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1361                         rtm.rtm_type == RTN_ANYCAST)
1362                         rtm.rtm_scope = RT_SCOPE_LINK;
1363         }
1364
1365         msg = nlmsg_alloc_simple(cmd, flags);
1366         if (!msg)
1367                 return -1;
1368
1369         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1370
1371         if (route->mask)
1372                 nla_put(msg, RTA_DST, alen, &route->addr);
1373
1374         if (route->sourcemask)
1375                 nla_put(msg, RTA_SRC, alen, &route->source);
1376
1377         if (route->metric > 0)
1378                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1379
1380         if (have_gw)
1381                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1382
1383         if (dev)
1384                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1385
1386         if (table >= 256)
1387                 nla_put_u32(msg, RTA_TABLE, table);
1388
1389         if (route->flags & DEVROUTE_MTU) {
1390                 struct nlattr *metrics;
1391
1392                 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1393                         goto nla_put_failure;
1394
1395                 nla_put_u32(msg, RTAX_MTU, route->mtu);
1396
1397                 nla_nest_end(msg, metrics);
1398         }
1399
1400         return system_rtnl_call(msg);
1401
1402 nla_put_failure:
1403         nlmsg_free(msg);
1404         return -ENOMEM;
1405 }
1406
1407 int system_add_route(struct device *dev, struct device_route *route)
1408 {
1409         return system_rt(dev, route, RTM_NEWROUTE);
1410 }
1411
1412 int system_del_route(struct device *dev, struct device_route *route)
1413 {
1414         return system_rt(dev, route, RTM_DELROUTE);
1415 }
1416
1417 int system_flush_routes(void)
1418 {
1419         const char *names[] = {
1420                 "/proc/sys/net/ipv4/route/flush",
1421                 "/proc/sys/net/ipv6/route/flush"
1422         };
1423         int fd, i;
1424
1425         for (i = 0; i < ARRAY_SIZE(names); i++) {
1426                 fd = open(names[i], O_WRONLY);
1427                 if (fd < 0)
1428                         continue;
1429
1430                 if (write(fd, "-1", 2)) {}
1431                 close(fd);
1432         }
1433         return 0;
1434 }
1435
1436 bool system_resolve_rt_type(const char *type, unsigned int *id)
1437 {
1438         return system_rtn_aton(type, id);
1439 }
1440
1441 bool system_resolve_rt_table(const char *name, unsigned int *id)
1442 {
1443         FILE *f;
1444         char *e, buf[128];
1445         unsigned int n, table = RT_TABLE_UNSPEC;
1446
1447         /* first try to parse table as number */
1448         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1449                 table = n;
1450
1451         /* handle well known aliases */
1452         else if (!strcmp(name, "default"))
1453                 table = RT_TABLE_DEFAULT;
1454         else if (!strcmp(name, "main"))
1455                 table = RT_TABLE_MAIN;
1456         else if (!strcmp(name, "local"))
1457                 table = RT_TABLE_LOCAL;
1458
1459         /* try to look up name in /etc/iproute2/rt_tables */
1460         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1461         {
1462                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1463                 {
1464                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1465                                 continue;
1466
1467                         n = strtoul(e, NULL, 10);
1468                         e = strtok(NULL, " \t\n");
1469
1470                         if (e && !strcmp(e, name))
1471                         {
1472                                 table = n;
1473                                 break;
1474                         }
1475                 }
1476
1477                 fclose(f);
1478         }
1479
1480         if (table == RT_TABLE_UNSPEC)
1481                 return false;
1482
1483         /* do not consider main table special */
1484         if (table == RT_TABLE_MAIN)
1485                 table = RT_TABLE_UNSPEC;
1486
1487         *id = table;
1488         return true;
1489 }
1490
1491 static int system_iprule(struct iprule *rule, int cmd)
1492 {
1493         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1494
1495         struct nl_msg *msg;
1496         struct rtmsg rtm = {
1497                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1498                 .rtm_protocol = RTPROT_STATIC,
1499                 .rtm_scope = RT_SCOPE_UNIVERSE,
1500                 .rtm_table = RT_TABLE_UNSPEC,
1501                 .rtm_type = RTN_UNSPEC,
1502                 .rtm_flags = 0,
1503         };
1504
1505         if (cmd == RTM_NEWRULE) {
1506                 rtm.rtm_type = RTN_UNICAST;
1507                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1508         }
1509
1510         if (rule->invert)
1511                 rtm.rtm_flags |= FIB_RULE_INVERT;
1512
1513         if (rule->flags & IPRULE_SRC)
1514                 rtm.rtm_src_len = rule->src_mask;
1515
1516         if (rule->flags & IPRULE_DEST)
1517                 rtm.rtm_dst_len = rule->dest_mask;
1518
1519         if (rule->flags & IPRULE_TOS)
1520                 rtm.rtm_tos = rule->tos;
1521
1522         if (rule->flags & IPRULE_LOOKUP) {
1523                 if (rule->lookup < 256)
1524                         rtm.rtm_table = rule->lookup;
1525         }
1526
1527         if (rule->flags & IPRULE_ACTION)
1528                 rtm.rtm_type = rule->action;
1529         else if (rule->flags & IPRULE_GOTO)
1530                 rtm.rtm_type = FR_ACT_GOTO;
1531         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1532                 rtm.rtm_type = FR_ACT_NOP;
1533
1534         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1535
1536         if (!msg)
1537                 return -1;
1538
1539         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1540
1541         if (rule->flags & IPRULE_IN)
1542                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1543
1544         if (rule->flags & IPRULE_OUT)
1545                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1546
1547         if (rule->flags & IPRULE_SRC)
1548                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1549
1550         if (rule->flags & IPRULE_DEST)
1551                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1552
1553         if (rule->flags & IPRULE_PRIORITY)
1554                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1555         else if (cmd == RTM_NEWRULE)
1556                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1557
1558         if (rule->flags & IPRULE_FWMARK)
1559                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1560
1561         if (rule->flags & IPRULE_FWMASK)
1562                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1563
1564         if (rule->flags & IPRULE_LOOKUP) {
1565                 if (rule->lookup >= 256)
1566                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1567         }
1568
1569         if (rule->flags & IPRULE_GOTO)
1570                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1571
1572         return system_rtnl_call(msg);
1573 }
1574
1575 int system_add_iprule(struct iprule *rule)
1576 {
1577         return system_iprule(rule, RTM_NEWRULE);
1578 }
1579
1580 int system_del_iprule(struct iprule *rule)
1581 {
1582         return system_iprule(rule, RTM_DELRULE);
1583 }
1584
1585 int system_flush_iprules(void)
1586 {
1587         int rv = 0;
1588         struct iprule rule;
1589
1590         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1591         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1592
1593         memset(&rule, 0, sizeof(rule));
1594
1595
1596         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1597
1598         rule.priority = 0;
1599         rule.lookup = RT_TABLE_LOCAL;
1600         rv |= system_iprule(&rule, RTM_NEWRULE);
1601
1602         rule.priority = 32766;
1603         rule.lookup = RT_TABLE_MAIN;
1604         rv |= system_iprule(&rule, RTM_NEWRULE);
1605
1606         rule.priority = 32767;
1607         rule.lookup = RT_TABLE_DEFAULT;
1608         rv |= system_iprule(&rule, RTM_NEWRULE);
1609
1610
1611         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1612
1613         rule.priority = 0;
1614         rule.lookup = RT_TABLE_LOCAL;
1615         rv |= system_iprule(&rule, RTM_NEWRULE);
1616
1617         rule.priority = 32766;
1618         rule.lookup = RT_TABLE_MAIN;
1619         rv |= system_iprule(&rule, RTM_NEWRULE);
1620
1621         return rv;
1622 }
1623
1624 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1625 {
1626         return system_rtn_aton(action, id);
1627 }
1628
1629 time_t system_get_rtime(void)
1630 {
1631         struct timespec ts;
1632         struct timeval tv;
1633
1634         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1635                 return ts.tv_sec;
1636
1637         if (gettimeofday(&tv, NULL) == 0)
1638                 return tv.tv_sec;
1639
1640         return 0;
1641 }
1642
1643 #ifndef IP_DF
1644 #define IP_DF       0x4000
1645 #endif
1646
1647 static int tunnel_ioctl(const char *name, int cmd, void *p)
1648 {
1649         struct ifreq ifr;
1650
1651         memset(&ifr, 0, sizeof(ifr));
1652         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1653         ifr.ifr_ifru.ifru_data = p;
1654         return ioctl(sock_ioctl, cmd, &ifr);
1655 }
1656
1657 int system_del_ip_tunnel(const char *name)
1658 {
1659         return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1660 }
1661
1662 int system_update_ipv6_mtu(struct device *dev, int mtu)
1663 {
1664         int ret = -1;
1665         char buf[64];
1666         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1667                         dev->ifname);
1668
1669         int fd = open(buf, O_RDWR);
1670         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1671         if (len < 0)
1672                 goto out;
1673
1674         buf[len] = 0;
1675         ret = atoi(buf);
1676
1677         if (!mtu || ret <= mtu)
1678                 goto out;
1679
1680         lseek(fd, 0, SEEK_SET);
1681         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1682                 ret = -1;
1683
1684 out:
1685         close(fd);
1686         return ret;
1687 }
1688
1689 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1690 {
1691         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1692         struct blob_attr *cur;
1693         bool set_df = true;
1694         const char *str;
1695
1696         system_del_ip_tunnel(name);
1697
1698         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1699                 blob_data(attr), blob_len(attr));
1700
1701         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1702                 return -EINVAL;
1703         str = blobmsg_data(cur);
1704
1705         if ((cur = tb[TUNNEL_ATTR_DF]))
1706                 set_df = blobmsg_get_bool(cur);
1707
1708         unsigned int ttl = 0;
1709         if ((cur = tb[TUNNEL_ATTR_TTL])) {
1710                 ttl = blobmsg_get_u32(cur);
1711                 if (ttl > 255 || (!set_df && ttl))
1712                         return -EINVAL;
1713         }
1714
1715         unsigned int link = 0;
1716         if ((cur = tb[TUNNEL_ATTR_LINK])) {
1717                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
1718                 if (!iface)
1719                         return -EINVAL;
1720
1721                 if (iface->l3_dev.dev)
1722                         link = iface->l3_dev.dev->ifindex;
1723         }
1724
1725         if (!strcmp(str, "sit")) {
1726                 struct ip_tunnel_parm p = {
1727                         .link = link,
1728                         .iph = {
1729                                 .version = 4,
1730                                 .ihl = 5,
1731                                 .frag_off = set_df ? htons(IP_DF) : 0,
1732                                 .protocol = IPPROTO_IPV6,
1733                                 .ttl = ttl
1734                         }
1735                 };
1736
1737                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1738                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1739                         return -EINVAL;
1740
1741                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1742                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1743                         return -EINVAL;
1744
1745                 strncpy(p.name, name, sizeof(p.name));
1746                 if (tunnel_ioctl("sit0", SIOCADDTUNNEL, &p) < 0)
1747                         return -1;
1748
1749 #ifdef SIOCADD6RD
1750                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
1751                         unsigned int mask;
1752                         struct ip_tunnel_6rd p6;
1753
1754                         memset(&p6, 0, sizeof(p6));
1755
1756                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
1757                                                 &p6.prefix, &mask) || mask > 128)
1758                                 return -EINVAL;
1759                         p6.prefixlen = mask;
1760
1761                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
1762                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
1763                                                         &p6.relay_prefix, &mask) || mask > 32)
1764                                         return -EINVAL;
1765                                 p6.relay_prefixlen = mask;
1766                         }
1767
1768                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
1769                                 system_del_ip_tunnel(name);
1770                                 return -1;
1771                         }
1772                 }
1773 #endif
1774         } else if (!strcmp(str, "ipip6")) {
1775                 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
1776                                 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1777                 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
1778                 int ret = 0;
1779
1780                 if (!nlm)
1781                         return -1;
1782
1783                 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1784                 nla_put_string(nlm, IFLA_IFNAME, name);
1785
1786                 if (link)
1787                         nla_put_u32(nlm, IFLA_LINK, link);
1788
1789                 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1790                 if (!linkinfo) {
1791                         ret = -ENOMEM;
1792                         goto failure;
1793                 }
1794                 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
1795                 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1796                 if (!infodata) {
1797                         ret = -ENOMEM;
1798                         goto failure;
1799                 }
1800
1801                 if (link)
1802                         nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
1803
1804                 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
1805                 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
1806                 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
1807
1808                 struct in6_addr in6buf;
1809                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1810                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1811                                 ret = -EINVAL;
1812                                 goto failure;
1813                         }
1814                         nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
1815                 }
1816
1817                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1818                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1819                                 ret = -EINVAL;
1820                                 goto failure;
1821                         }
1822                         nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
1823                 }
1824
1825 #ifdef IFLA_IPTUN_FMR_MAX
1826                 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
1827                         struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
1828
1829                         struct blob_attr *fmr;
1830                         unsigned rem, fmrcnt = 0;
1831                         blobmsg_for_each_attr(fmr, cur, rem) {
1832                                 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
1833                                         continue;
1834
1835                                 unsigned ip4len, ip6len, ealen, offset = 6;
1836                                 char ip6buf[48];
1837                                 char ip4buf[16];
1838
1839                                 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
1840                                                 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
1841                                         ret = -EINVAL;
1842                                         goto failure;
1843                                 }
1844
1845                                 struct in6_addr ip6prefix;
1846                                 struct in_addr ip4prefix;
1847                                 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
1848                                                 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
1849                                         ret = -EINVAL;
1850                                         goto failure;
1851                                 }
1852
1853                                 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
1854
1855                                 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
1856                                 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
1857                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
1858                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
1859                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
1860                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
1861
1862                                 nla_nest_end(nlm, rule);
1863                         }
1864
1865                         nla_nest_end(nlm, fmrs);
1866                 }
1867 #endif
1868
1869                 nla_nest_end(nlm, infodata);
1870                 nla_nest_end(nlm, linkinfo);
1871
1872                 return system_rtnl_call(nlm);
1873 failure:
1874                 nlmsg_free(nlm);
1875                 return ret;
1876         }
1877         else
1878                 return -EINVAL;
1879
1880         return 0;
1881 }