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 static int system_link_del(struct device *dev)
 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         nla_put_string(msg, IFLA_IFNAME, dev->ifname);
 809         return system_rtnl_call(msg);
 810 }
 811
 812 int system_macvlan_del(struct device *macvlan)
 813 {
 814         return system_link_del(macvlan);
 815 }
 816
 817 static int system_vlan(struct device *dev, int id)
 818 {
 819         struct vlan_ioctl_args ifr = {
 820                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 821                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 822         };
 823
 824         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 825
 826         if (id < 0) {
 827                 ifr.cmd = DEL_VLAN_CMD;
 828                 ifr.u.VID = 0;
 829         } else {
 830                 ifr.cmd = ADD_VLAN_CMD;
 831                 ifr.u.VID = id;
 832         }
 833         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 834         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 835 }
 836
 837 int system_vlan_add(struct device *dev, int id)
 838 {
 839         return system_vlan(dev, id);
 840 }
 841
 842 int system_vlan_del(struct device *dev)
 843 {
 844         return system_vlan(dev, -1);
 845 }
 846
 847 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
 848 {
 849         struct nl_msg *msg;
 850         struct nlattr *linkinfo, *data;
 851         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
 852         int ifindex = system_if_resolve(dev);
 853         int rv;
 854
 855         if (ifindex == 0)
 856                 return -ENOENT;
 857
 858         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 859
 860         if (!msg)
 861                 return -1;
 862
 863         nlmsg_append(msg, &iim, sizeof(iim), 0);
 864         nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
 865         nla_put_u32(msg, IFLA_LINK, ifindex);
 866
 867         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 868                 goto nla_put_failure;
 869
 870         nla_put_string(msg, IFLA_INFO_KIND, "vlan");
 871
 872         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 873                 goto nla_put_failure;
 874
 875         nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
 876
 877 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
 878         nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
 879 #else
 880         if(cfg->proto == VLAN_PROTO_8021AD)
 881                 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);
 882 #endif
 883
 884         nla_nest_end(msg, data);
 885         nla_nest_end(msg, linkinfo);
 886
 887         rv = system_rtnl_call(msg);
 888         if (rv)
 889                 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
 890
 891         return rv;
 892
 893 nla_put_failure:
 894         nlmsg_free(msg);
 895         return -ENOMEM;
 896 }
 897
 898 int system_vlandev_del(struct device *vlandev)
 899 {
 900         return system_link_del(vlandev);
 901 }
 902
 903 static void
 904 system_if_get_settings(struct device *dev, struct device_settings *s)
 905 {
 906         struct ifreq ifr;
 907         char buf[10];
 908
 909         memset(&ifr, 0, sizeof(ifr));
 910         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 911
 912         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 913                 s->mtu = ifr.ifr_mtu;
 914                 s->flags |= DEV_OPT_MTU;
 915         }
 916
 917         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 918                 s->txqueuelen = ifr.ifr_qlen;
 919                 s->flags |= DEV_OPT_TXQUEUELEN;
 920         }
 921
 922         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 923                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 924                 s->flags |= DEV_OPT_MACADDR;
 925         }
 926
 927         if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
 928                 s->ipv6 = !strtoul(buf, NULL, 0);
 929                 s->flags |= DEV_OPT_IPV6;
 930         }
 931 }
 932
 933 void
 934 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
 935 {
 936         struct ifreq ifr;
 937
 938         if (!apply_mask)
 939                 return;
 940
 941         memset(&ifr, 0, sizeof(ifr));
 942         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 943         if (s->flags & DEV_OPT_MTU & apply_mask) {
 944                 ifr.ifr_mtu = s->mtu;
 945                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 946                         s->flags &= ~DEV_OPT_MTU;
 947         }
 948         if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
 949                 ifr.ifr_qlen = s->txqueuelen;
 950                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 951                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 952         }
 953         if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
 954                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 955                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 956                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 957                         s->flags &= ~DEV_OPT_MACADDR;
 958         }
 959         if (s->flags & DEV_OPT_IPV6 & apply_mask)
 960                 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
 961 }
 962
 963 int system_if_up(struct device *dev)
 964 {
 965         system_if_get_settings(dev, &dev->orig_settings);
 966         system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
 967         device_set_ifindex(dev, system_if_resolve(dev));
 968         return system_if_flags(dev->ifname, IFF_UP, 0);
 969 }
 970
 971 int system_if_down(struct device *dev)
 972 {
 973         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
 974         dev->orig_settings.flags &= dev->settings.flags;
 975         system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
 976         return ret;
 977 }
 978
 979 struct if_check_data {
 980         struct device *dev;
 981         int pending;
 982         int ret;
 983 };
 984
 985 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
 986 {
 987         struct nlmsghdr *nh = nlmsg_hdr(msg);
 988         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 989         struct if_check_data *chk = (struct if_check_data *)arg;
 990
 991         if (nh->nlmsg_type != RTM_NEWLINK)
 992                 return NL_SKIP;
 993
 994         device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
 995         device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
 996
 997         return NL_OK;
 998 }
 999
1000 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1001 {
1002         struct if_check_data *chk = (struct if_check_data *)arg;
1003         chk->pending = 0;
1004         return NL_STOP;
1005 }
1006
1007 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1008 {
1009         struct if_check_data *chk = (struct if_check_data *)arg;
1010
1011         device_set_present(chk->dev, false);
1012         device_set_link(chk->dev, false);
1013         chk->pending = err->error;
1014
1015         return NL_STOP;
1016 }
1017
1018 int system_if_check(struct device *dev)
1019 {
1020         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1021         struct nl_msg *msg;
1022         struct ifinfomsg ifi = {
1023                 .ifi_family = AF_UNSPEC,
1024                 .ifi_index = 0,
1025         };
1026         struct if_check_data chk = {
1027                 .dev = dev,
1028                 .pending = 1,
1029         };
1030         int ret = 1;
1031
1032         msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1033         if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1034             nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1035                 goto out;
1036
1037         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1038         nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1039         nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1040
1041         nl_send_auto_complete(sock_rtnl, msg);
1042         while (chk.pending > 0)
1043                 nl_recvmsgs(sock_rtnl, cb);
1044
1045         nlmsg_free(msg);
1046         ret = chk.pending;
1047
1048 out:
1049         nl_cb_put(cb);
1050         return ret;
1051 }
1052
1053 struct device *
1054 system_if_get_parent(struct device *dev)
1055 {
1056         char buf[64], *devname;
1057         int ifindex, iflink, len;
1058         FILE *f;
1059
1060         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1061         f = fopen(buf, "r");
1062         if (!f)
1063                 return NULL;
1064
1065         len = fread(buf, 1, sizeof(buf) - 1, f);
1066         fclose(f);
1067
1068         if (len <= 0)
1069                 return NULL;
1070
1071         buf[len] = 0;
1072         iflink = strtoul(buf, NULL, 0);
1073         ifindex = system_if_resolve(dev);
1074         if (!iflink || iflink == ifindex)
1075                 return NULL;
1076
1077         devname = if_indextoname(iflink, buf);
1078         if (!devname)
1079                 return NULL;
1080
1081         return device_get(devname, true);
1082 }
1083
1084 static bool
1085 read_string_file(int dir_fd, const char *file, char *buf, int len)
1086 {
1087         bool ret = false;
1088         char *c;
1089         int fd;
1090
1091         fd = openat(dir_fd, file, O_RDONLY);
1092         if (fd < 0)
1093                 return false;
1094
1095 retry:
1096         len = read(fd, buf, len - 1);
1097         if (len < 0) {
1098                 if (errno == EINTR)
1099                         goto retry;
1100         } else if (len > 0) {
1101                         buf[len] = 0;
1102
1103                         c = strchr(buf, '\n');
1104                         if (c)
1105                                 *c = 0;
1106
1107                         ret = true;
1108         }
1109
1110         close(fd);
1111
1112         return ret;
1113 }
1114
1115 static bool
1116 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1117 {
1118         char buf[64];
1119         bool ret = false;
1120
1121         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1122         if (ret)
1123                 *val = strtoull(buf, NULL, 0);
1124
1125         return ret;
1126 }
1127
1128 /* Assume advertised flags == supported flags */
1129 static const struct {
1130         uint32_t mask;
1131         const char *name;
1132 } ethtool_link_modes[] = {
1133         { ADVERTISED_10baseT_Half, "10H" },
1134         { ADVERTISED_10baseT_Full, "10F" },
1135         { ADVERTISED_100baseT_Half, "100H" },
1136         { ADVERTISED_100baseT_Full, "100F" },
1137         { ADVERTISED_1000baseT_Half, "1000H" },
1138         { ADVERTISED_1000baseT_Full, "1000F" },
1139 };
1140
1141 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1142 {
1143         int i;
1144         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1145                 if (mask & ethtool_link_modes[i].mask)
1146                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1147         }
1148 }
1149
1150 bool
1151 system_if_force_external(const char *ifname)
1152 {
1153         char buf[64];
1154         struct stat s;
1155
1156         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1157         return stat(buf, &s) == 0;
1158 }
1159
1160 int
1161 system_if_dump_info(struct device *dev, struct blob_buf *b)
1162 {
1163         struct ethtool_cmd ecmd;
1164         struct ifreq ifr;
1165         char buf[64], *s;
1166         void *c;
1167         int dir_fd;
1168
1169         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1170         dir_fd = open(buf, O_DIRECTORY);
1171
1172         memset(&ecmd, 0, sizeof(ecmd));
1173         memset(&ifr, 0, sizeof(ifr));
1174         strcpy(ifr.ifr_name, dev->ifname);
1175         ifr.ifr_data = (caddr_t) &ecmd;
1176         ecmd.cmd = ETHTOOL_GSET;
1177
1178         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1179                 c = blobmsg_open_array(b, "link-advertising");
1180                 system_add_link_modes(b, ecmd.advertising);
1181                 blobmsg_close_array(b, c);
1182
1183                 c = blobmsg_open_array(b, "link-supported");
1184                 system_add_link_modes(b, ecmd.supported);
1185                 blobmsg_close_array(b, c);
1186
1187                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1188                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1189                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1190                 blobmsg_add_string_buffer(b);
1191         }
1192
1193         close(dir_fd);
1194         return 0;
1195 }
1196
1197 int
1198 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1199 {
1200         const char *const counters[] = {
1201                 "collisions",     "rx_frame_errors",   "tx_compressed",
1202                 "multicast",      "rx_length_errors",  "tx_dropped",
1203                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
1204                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
1205                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
1206                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
1207                 "rx_errors",      "tx_bytes",          "tx_window_errors",
1208                 "rx_fifo_errors", "tx_carrier_errors",
1209         };
1210         char buf[64];
1211         int stats_dir;
1212         int i;
1213         uint64_t val = 0;
1214
1215         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1216         stats_dir = open(buf, O_DIRECTORY);
1217         if (stats_dir < 0)
1218                 return -1;
1219
1220         for (i = 0; i < ARRAY_SIZE(counters); i++)
1221                 if (read_uint64_file(stats_dir, counters[i], &val))
1222                         blobmsg_add_u64(b, counters[i], val);
1223
1224         close(stats_dir);
1225         return 0;
1226 }
1227
1228 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1229 {
1230         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1231         int alen = v4 ? 4 : 16;
1232         unsigned int flags = 0;
1233         struct ifaddrmsg ifa = {
1234                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1235                 .ifa_prefixlen = addr->mask,
1236                 .ifa_index = dev->ifindex,
1237         };
1238
1239         struct nl_msg *msg;
1240         if (cmd == RTM_NEWADDR)
1241                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1242
1243         msg = nlmsg_alloc_simple(cmd, flags);
1244         if (!msg)
1245                 return -1;
1246
1247         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1248         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1249         if (v4) {
1250                 if (addr->broadcast)
1251                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1252                 if (addr->point_to_point)
1253                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1254         } else {
1255                 time_t now = system_get_rtime();
1256                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1257
1258                 if (addr->preferred_until) {
1259                         int64_t preferred = addr->preferred_until - now;
1260                         if (preferred < 0)
1261                                 preferred = 0;
1262                         else if (preferred > UINT32_MAX)
1263                                 preferred = UINT32_MAX;
1264
1265                         cinfo.ifa_prefered = preferred;
1266                 }
1267
1268                 if (addr->valid_until) {
1269                         int64_t valid = addr->valid_until - now;
1270                         if (valid <= 0)
1271                                 return -1;
1272                         else if (valid > UINT32_MAX)
1273                                 valid = UINT32_MAX;
1274
1275                         cinfo.ifa_valid = valid;
1276                 }
1277
1278                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1279         }
1280
1281         return system_rtnl_call(msg);
1282 }
1283
1284 int system_add_address(struct device *dev, struct device_addr *addr)
1285 {
1286         return system_addr(dev, addr, RTM_NEWADDR);
1287 }
1288
1289 int system_del_address(struct device *dev, struct device_addr *addr)
1290 {
1291         return system_addr(dev, addr, RTM_DELADDR);
1292 }
1293
1294 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1295 {
1296         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1297         bool have_gw;
1298         unsigned int flags = 0;
1299
1300         if (alen == 4)
1301                 have_gw = !!route->nexthop.in.s_addr;
1302         else
1303                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1304                         route->nexthop.in6.s6_addr32[1] ||
1305                         route->nexthop.in6.s6_addr32[2] ||
1306                         route->nexthop.in6.s6_addr32[3];
1307
1308         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1309                         ? route->table : RT_TABLE_MAIN;
1310
1311         struct rtmsg rtm = {
1312                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1313                 .rtm_dst_len = route->mask,
1314                 .rtm_src_len = route->sourcemask,
1315                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1316                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1317                 .rtm_scope = RT_SCOPE_NOWHERE,
1318                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1319                 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1320         };
1321         struct nl_msg *msg;
1322
1323         if (cmd == RTM_NEWROUTE) {
1324                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1325
1326                 if (!dev) { // Add null-route
1327                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1328                         rtm.rtm_type = RTN_UNREACHABLE;
1329                 }
1330                 else
1331                         rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1332         }
1333
1334         if (route->flags & DEVROUTE_TYPE) {
1335                 rtm.rtm_type = route->type;
1336                 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1337                         if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1338                             rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1339                                 rtm.rtm_table = RT_TABLE_LOCAL;
1340                 }
1341
1342                 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT)
1343                         rtm.rtm_scope = RT_SCOPE_HOST;
1344                 else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1345                         rtm.rtm_type == RTN_ANYCAST)
1346                         rtm.rtm_scope = RT_SCOPE_LINK;
1347         }
1348
1349         msg = nlmsg_alloc_simple(cmd, flags);
1350         if (!msg)
1351                 return -1;
1352
1353         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1354
1355         if (route->mask)
1356                 nla_put(msg, RTA_DST, alen, &route->addr);
1357
1358         if (route->sourcemask)
1359                 nla_put(msg, RTA_SRC, alen, &route->source);
1360
1361         if (route->metric > 0)
1362                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1363
1364         if (have_gw)
1365                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1366
1367         if (dev)
1368                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1369
1370         if (table >= 256)
1371                 nla_put_u32(msg, RTA_TABLE, table);
1372
1373         if (route->flags & DEVROUTE_MTU) {
1374                 struct nlattr *metrics;
1375
1376                 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1377                         goto nla_put_failure;
1378
1379                 nla_put_u32(msg, RTAX_MTU, route->mtu);
1380
1381                 nla_nest_end(msg, metrics);
1382         }
1383
1384         return system_rtnl_call(msg);
1385
1386 nla_put_failure:
1387         nlmsg_free(msg);
1388         return -ENOMEM;
1389 }
1390
1391 int system_add_route(struct device *dev, struct device_route *route)
1392 {
1393         return system_rt(dev, route, RTM_NEWROUTE);
1394 }
1395
1396 int system_del_route(struct device *dev, struct device_route *route)
1397 {
1398         return system_rt(dev, route, RTM_DELROUTE);
1399 }
1400
1401 int system_flush_routes(void)
1402 {
1403         const char *names[] = {
1404                 "/proc/sys/net/ipv4/route/flush",
1405                 "/proc/sys/net/ipv6/route/flush"
1406         };
1407         int fd, i;
1408
1409         for (i = 0; i < ARRAY_SIZE(names); i++) {
1410                 fd = open(names[i], O_WRONLY);
1411                 if (fd < 0)
1412                         continue;
1413
1414                 if (write(fd, "-1", 2)) {}
1415                 close(fd);
1416         }
1417         return 0;
1418 }
1419
1420 bool system_resolve_rt_type(const char *type, unsigned int *id)
1421 {
1422         return system_rtn_aton(type, id);
1423 }
1424
1425 bool system_resolve_rt_table(const char *name, unsigned int *id)
1426 {
1427         FILE *f;
1428         char *e, buf[128];
1429         unsigned int n, table = RT_TABLE_UNSPEC;
1430
1431         /* first try to parse table as number */
1432         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1433                 table = n;
1434
1435         /* handle well known aliases */
1436         else if (!strcmp(name, "default"))
1437                 table = RT_TABLE_DEFAULT;
1438         else if (!strcmp(name, "main"))
1439                 table = RT_TABLE_MAIN;
1440         else if (!strcmp(name, "local"))
1441                 table = RT_TABLE_LOCAL;
1442
1443         /* try to look up name in /etc/iproute2/rt_tables */
1444         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1445         {
1446                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1447                 {
1448                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1449                                 continue;
1450
1451                         n = strtoul(e, NULL, 10);
1452                         e = strtok(NULL, " \t\n");
1453
1454                         if (e && !strcmp(e, name))
1455                         {
1456                                 table = n;
1457                                 break;
1458                         }
1459                 }
1460
1461                 fclose(f);
1462         }
1463
1464         if (table == RT_TABLE_UNSPEC)
1465                 return false;
1466
1467         /* do not consider main table special */
1468         if (table == RT_TABLE_MAIN)
1469                 table = RT_TABLE_UNSPEC;
1470
1471         *id = table;
1472         return true;
1473 }
1474
1475 static int system_iprule(struct iprule *rule, int cmd)
1476 {
1477         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1478
1479         struct nl_msg *msg;
1480         struct rtmsg rtm = {
1481                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1482                 .rtm_protocol = RTPROT_STATIC,
1483                 .rtm_scope = RT_SCOPE_UNIVERSE,
1484                 .rtm_table = RT_TABLE_UNSPEC,
1485                 .rtm_type = RTN_UNSPEC,
1486                 .rtm_flags = 0,
1487         };
1488
1489         if (cmd == RTM_NEWRULE) {
1490                 rtm.rtm_type = RTN_UNICAST;
1491                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1492         }
1493
1494         if (rule->invert)
1495                 rtm.rtm_flags |= FIB_RULE_INVERT;
1496
1497         if (rule->flags & IPRULE_SRC)
1498                 rtm.rtm_src_len = rule->src_mask;
1499
1500         if (rule->flags & IPRULE_DEST)
1501                 rtm.rtm_dst_len = rule->dest_mask;
1502
1503         if (rule->flags & IPRULE_TOS)
1504                 rtm.rtm_tos = rule->tos;
1505
1506         if (rule->flags & IPRULE_LOOKUP) {
1507                 if (rule->lookup < 256)
1508                         rtm.rtm_table = rule->lookup;
1509         }
1510
1511         if (rule->flags & IPRULE_ACTION)
1512                 rtm.rtm_type = rule->action;
1513         else if (rule->flags & IPRULE_GOTO)
1514                 rtm.rtm_type = FR_ACT_GOTO;
1515         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1516                 rtm.rtm_type = FR_ACT_NOP;
1517
1518         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1519
1520         if (!msg)
1521                 return -1;
1522
1523         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1524
1525         if (rule->flags & IPRULE_IN)
1526                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1527
1528         if (rule->flags & IPRULE_OUT)
1529                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1530
1531         if (rule->flags & IPRULE_SRC)
1532                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1533
1534         if (rule->flags & IPRULE_DEST)
1535                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1536
1537         if (rule->flags & IPRULE_PRIORITY)
1538                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1539         else if (cmd == RTM_NEWRULE)
1540                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1541
1542         if (rule->flags & IPRULE_FWMARK)
1543                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1544
1545         if (rule->flags & IPRULE_FWMASK)
1546                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1547
1548         if (rule->flags & IPRULE_LOOKUP) {
1549                 if (rule->lookup >= 256)
1550                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1551         }
1552
1553         if (rule->flags & IPRULE_GOTO)
1554                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1555
1556         return system_rtnl_call(msg);
1557 }
1558
1559 int system_add_iprule(struct iprule *rule)
1560 {
1561         return system_iprule(rule, RTM_NEWRULE);
1562 }
1563
1564 int system_del_iprule(struct iprule *rule)
1565 {
1566         return system_iprule(rule, RTM_DELRULE);
1567 }
1568
1569 int system_flush_iprules(void)
1570 {
1571         int rv = 0;
1572         struct iprule rule;
1573
1574         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1575         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1576
1577         memset(&rule, 0, sizeof(rule));
1578
1579
1580         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1581
1582         rule.priority = 0;
1583         rule.lookup = RT_TABLE_LOCAL;
1584         rv |= system_iprule(&rule, RTM_NEWRULE);
1585
1586         rule.priority = 32766;
1587         rule.lookup = RT_TABLE_MAIN;
1588         rv |= system_iprule(&rule, RTM_NEWRULE);
1589
1590         rule.priority = 32767;
1591         rule.lookup = RT_TABLE_DEFAULT;
1592         rv |= system_iprule(&rule, RTM_NEWRULE);
1593
1594
1595         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1596
1597         rule.priority = 0;
1598         rule.lookup = RT_TABLE_LOCAL;
1599         rv |= system_iprule(&rule, RTM_NEWRULE);
1600
1601         rule.priority = 32766;
1602         rule.lookup = RT_TABLE_MAIN;
1603         rv |= system_iprule(&rule, RTM_NEWRULE);
1604
1605         return rv;
1606 }
1607
1608 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1609 {
1610         return system_rtn_aton(action, id);
1611 }
1612
1613 time_t system_get_rtime(void)
1614 {
1615         struct timespec ts;
1616         struct timeval tv;
1617
1618         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1619                 return ts.tv_sec;
1620
1621         if (gettimeofday(&tv, NULL) == 0)
1622                 return tv.tv_sec;
1623
1624         return 0;
1625 }
1626
1627 #ifndef IP_DF
1628 #define IP_DF       0x4000
1629 #endif
1630
1631 static int tunnel_ioctl(const char *name, int cmd, void *p)
1632 {
1633         struct ifreq ifr;
1634
1635         memset(&ifr, 0, sizeof(ifr));
1636         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1637         ifr.ifr_ifru.ifru_data = p;
1638         return ioctl(sock_ioctl, cmd, &ifr);
1639 }
1640
1641 int system_del_ip_tunnel(const char *name)
1642 {
1643         return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1644 }
1645
1646 int system_update_ipv6_mtu(struct device *dev, int mtu)
1647 {
1648         int ret = -1;
1649         char buf[64];
1650         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1651                         dev->ifname);
1652
1653         int fd = open(buf, O_RDWR);
1654         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1655         if (len < 0)
1656                 goto out;
1657
1658         buf[len] = 0;
1659         ret = atoi(buf);
1660
1661         if (!mtu || ret <= mtu)
1662                 goto out;
1663
1664         lseek(fd, 0, SEEK_SET);
1665         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1666                 ret = -1;
1667
1668 out:
1669         close(fd);
1670         return ret;
1671 }
1672
1673 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1674 {
1675         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1676         struct blob_attr *cur;
1677         bool set_df = true;
1678         const char *str;
1679
1680         system_del_ip_tunnel(name);
1681
1682         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1683                 blob_data(attr), blob_len(attr));
1684
1685         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1686                 return -EINVAL;
1687         str = blobmsg_data(cur);
1688
1689         if ((cur = tb[TUNNEL_ATTR_DF]))
1690                 set_df = blobmsg_get_bool(cur);
1691
1692         unsigned int ttl = 0;
1693         if ((cur = tb[TUNNEL_ATTR_TTL])) {
1694                 ttl = blobmsg_get_u32(cur);
1695                 if (ttl > 255 || (!set_df && ttl))
1696                         return -EINVAL;
1697         }
1698
1699         unsigned int link = 0;
1700         if ((cur = tb[TUNNEL_ATTR_LINK])) {
1701                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
1702                 if (!iface)
1703                         return -EINVAL;
1704
1705                 if (iface->l3_dev.dev)
1706                         link = iface->l3_dev.dev->ifindex;
1707         }
1708
1709         if (!strcmp(str, "sit")) {
1710                 struct ip_tunnel_parm p = {
1711                         .link = link,
1712                         .iph = {
1713                                 .version = 4,
1714                                 .ihl = 5,
1715                                 .frag_off = set_df ? htons(IP_DF) : 0,
1716                                 .protocol = IPPROTO_IPV6,
1717                                 .ttl = ttl
1718                         }
1719                 };
1720
1721                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1722                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1723                         return -EINVAL;
1724
1725                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1726                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1727                         return -EINVAL;
1728
1729                 strncpy(p.name, name, sizeof(p.name));
1730                 if (tunnel_ioctl("sit0", SIOCADDTUNNEL, &p) < 0)
1731                         return -1;
1732
1733 #ifdef SIOCADD6RD
1734                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
1735                         unsigned int mask;
1736                         struct ip_tunnel_6rd p6;
1737
1738                         memset(&p6, 0, sizeof(p6));
1739
1740                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
1741                                                 &p6.prefix, &mask) || mask > 128)
1742                                 return -EINVAL;
1743                         p6.prefixlen = mask;
1744
1745                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
1746                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
1747                                                         &p6.relay_prefix, &mask) || mask > 32)
1748                                         return -EINVAL;
1749                                 p6.relay_prefixlen = mask;
1750                         }
1751
1752                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
1753                                 system_del_ip_tunnel(name);
1754                                 return -1;
1755                         }
1756                 }
1757 #endif
1758         } else if (!strcmp(str, "ipip6")) {
1759                 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
1760                                 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1761                 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
1762                 int ret = 0;
1763
1764                 if (!nlm)
1765                         return -1;
1766
1767                 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1768                 nla_put_string(nlm, IFLA_IFNAME, name);
1769
1770                 if (link)
1771                         nla_put_u32(nlm, IFLA_LINK, link);
1772
1773                 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1774                 if (!linkinfo) {
1775                         ret = -ENOMEM;
1776                         goto failure;
1777                 }
1778                 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
1779                 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1780                 if (!infodata) {
1781                         ret = -ENOMEM;
1782                         goto failure;
1783                 }
1784
1785                 if (link)
1786                         nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
1787
1788                 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
1789                 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
1790                 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
1791
1792                 struct in6_addr in6buf;
1793                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1794                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1795                                 ret = -EINVAL;
1796                                 goto failure;
1797                         }
1798                         nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
1799                 }
1800
1801                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1802                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1803                                 ret = -EINVAL;
1804                                 goto failure;
1805                         }
1806                         nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
1807                 }
1808
1809 #ifdef IFLA_IPTUN_FMR_MAX
1810                 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
1811                         struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
1812
1813                         struct blob_attr *fmr;
1814                         unsigned rem, fmrcnt = 0;
1815                         blobmsg_for_each_attr(fmr, cur, rem) {
1816                                 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
1817                                         continue;
1818
1819                                 unsigned ip4len, ip6len, ealen, offset = 6;
1820                                 char ip6buf[48];
1821                                 char ip4buf[16];
1822
1823                                 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
1824                                                 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
1825                                         ret = -EINVAL;
1826                                         goto failure;
1827                                 }
1828
1829                                 struct in6_addr ip6prefix;
1830                                 struct in_addr ip4prefix;
1831                                 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
1832                                                 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
1833                                         ret = -EINVAL;
1834                                         goto failure;
1835                                 }
1836
1837                                 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
1838
1839                                 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
1840                                 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
1841                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
1842                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
1843                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
1844                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
1845
1846                                 nla_nest_end(nlm, rule);
1847                         }
1848
1849                         nla_nest_end(nlm, fmrs);
1850                 }
1851 #endif
1852
1853                 nla_nest_end(nlm, infodata);
1854                 nla_nest_end(nlm, linkinfo);
1855
1856                 return system_rtnl_call(nlm);
1857 failure:
1858                 nlmsg_free(nlm);
1859                 return ret;
1860         }
1861         else
1862                 return -EINVAL;
1863
1864         return 0;
1865 }