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  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2
   9  * as published by the Free Software Foundation
  10  *
  11  * This program is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14  * GNU General Public License for more details.
  15  */
  16 #define _GNU_SOURCE
  17
  18 #include <sys/socket.h>
  19 #include <sys/ioctl.h>
  20 #include <sys/stat.h>
  21 #include <sys/syscall.h>
  22
  23 #include <net/if.h>
  24 #include <net/if_arp.h>
  25
  26 #include <arpa/inet.h>
  27 #include <netinet/in.h>
  28
  29 #include <linux/rtnetlink.h>
  30 #include <linux/sockios.h>
  31 #include <linux/ip.h>
  32 #include <linux/if_link.h>
  33 #include <linux/if_vlan.h>
  34 #include <linux/if_bridge.h>
  35 #include <linux/if_tunnel.h>
  36 #include <linux/ip6_tunnel.h>
  37 #include <linux/ethtool.h>
  38 #include <linux/fib_rules.h>
  39
  40 #ifndef RTN_FAILED_POLICY
  41 #define RTN_FAILED_POLICY 12
  42 #endif
  43
  44 #include <string.h>
  45 #include <fcntl.h>
  46 #include <glob.h>
  47 #include <time.h>
  48
  49 #include <netlink/msg.h>
  50 #include <netlink/attr.h>
  51 #include <netlink/socket.h>
  52 #include <libubox/uloop.h>
  53
  54 #include "netifd.h"
  55 #include "device.h"
  56 #include "system.h"
  57
  58 struct event_socket {
  59         struct uloop_fd uloop;
  60         struct nl_sock *sock;
  61         int bufsize;
  62 };
  63
  64 static int sock_ioctl = -1;
  65 static struct nl_sock *sock_rtnl = NULL;
  66
  67 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
  68 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
  69
  70 static char dev_buf[256];
  71
  72 static void
  73 handler_nl_event(struct uloop_fd *u, unsigned int events)
  74 {
  75         struct event_socket *ev = container_of(u, struct event_socket, uloop);
  76         int err;
  77         socklen_t errlen = sizeof(err);
  78
  79         if (!u->error) {
  80                 nl_recvmsgs_default(ev->sock);
  81                 return;
  82         }
  83
  84         if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
  85                 goto abort;
  86
  87         switch(err) {
  88         case ENOBUFS:
  89                 // Increase rx buffer size on netlink socket
  90                 ev->bufsize *= 2;
  91                 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
  92                         goto abort;
  93
  94                 // Request full dump since some info got dropped
  95                 struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
  96                 nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
  97                 break;
  98
  99         default:
 100                 goto abort;
 101         }
 102         u->error = false;
 103         return;
 104
 105 abort:
 106         uloop_fd_delete(&ev->uloop);
 107         return;
 108 }
 109
 110 static struct nl_sock *
 111 create_socket(int protocol, int groups)
 112 {
 113         struct nl_sock *sock;
 114
 115         sock = nl_socket_alloc();
 116         if (!sock)
 117                 return NULL;
 118
 119         if (groups)
 120                 nl_join_groups(sock, groups);
 121
 122         if (nl_connect(sock, protocol))
 123                 return NULL;
 124
 125         return sock;
 126 }
 127
 128 static bool
 129 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
 130                         uloop_fd_handler cb, int flags)
 131 {
 132         ev->sock = create_socket(protocol, groups);
 133         if (!ev->sock)
 134                 return false;
 135
 136         ev->uloop.fd = nl_socket_get_fd(ev->sock);
 137         ev->uloop.cb = cb;
 138         if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
 139                 return false;
 140
 141         return true;
 142 }
 143
 144 static bool
 145 create_event_socket(struct event_socket *ev, int protocol,
 146                     int (*cb)(struct nl_msg *msg, void *arg))
 147 {
 148         if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
 149                 return false;
 150
 151         // Install the valid custom callback handler
 152         nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
 153
 154         // Disable sequence number checking on event sockets
 155         nl_socket_disable_seq_check(ev->sock);
 156
 157         // Increase rx buffer size to 65K on event sockets
 158         ev->bufsize = 65535;
 159         if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
 160                 return false;
 161
 162         return true;
 163 }
 164
 165 int system_init(void)
 166 {
 167         static struct event_socket rtnl_event;
 168         static struct event_socket hotplug_event;
 169
 170         sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
 171         system_fd_set_cloexec(sock_ioctl);
 172
 173         // Prepare socket for routing / address control
 174         sock_rtnl = create_socket(NETLINK_ROUTE, 0);
 175         if (!sock_rtnl)
 176                 return -1;
 177
 178         if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
 179                 return -1;
 180
 181         if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
 182                                      handle_hotplug_event, 0))
 183                 return -1;
 184
 185         // Receive network link events form kernel
 186         nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
 187
 188         return 0;
 189 }
 190
 191 static void system_set_sysctl(const char *path, const char *val)
 192 {
 193         int fd;
 194
 195         fd = open(path, O_WRONLY);
 196         if (fd < 0)
 197                 return;
 198
 199         if (write(fd, val, strlen(val))) {}
 200         close(fd);
 201 }
 202
 203 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
 204 {
 205         snprintf(dev_buf, sizeof(dev_buf), path, device);
 206         system_set_sysctl(dev_buf, val);
 207 }
 208
 209 static void system_set_disable_ipv6(struct device *dev, const char *val)
 210 {
 211         system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
 212 }
 213
 214 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
 215 {
 216         int fd = -1, ret = -1;
 217
 218         fd = open(path, O_RDONLY);
 219         if (fd < 0)
 220                 goto out;
 221
 222         ssize_t len = read(fd, buf, buf_sz - 1);
 223         if (len < 0)
 224                 goto out;
 225
 226         ret = buf[len] = 0;
 227
 228 out:
 229         if (fd >= 0)
 230                 close(fd);
 231
 232         return ret;
 233 }
 234
 235 static int
 236 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
 237 {
 238         snprintf(dev_buf, sizeof(dev_buf), path, device);
 239         return system_get_sysctl(dev_buf, buf, buf_sz);
 240 }
 241
 242 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
 243 {
 244         return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
 245                         dev->ifname, buf, buf_sz);
 246 }
 247
 248 #ifndef IFF_LOWER_UP
 249 #define IFF_LOWER_UP    0x10000
 250 #endif
 251
 252 // Evaluate netlink messages
 253 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 254 {
 255         struct nlmsghdr *nh = nlmsg_hdr(msg);
 256         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 257         struct nlattr *nla[__IFLA_MAX];
 258
 259         if (nh->nlmsg_type != RTM_NEWLINK)
 260                 goto out;
 261
 262         nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
 263         if (!nla[IFLA_IFNAME])
 264                 goto out;
 265
 266         struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
 267         if (!dev)
 268                 goto out;
 269
 270         device_set_ifindex(dev, ifi->ifi_index);
 271         device_set_link(dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
 272
 273 out:
 274         return 0;
 275 }
 276
 277 static void
 278 handle_hotplug_msg(char *data, int size)
 279 {
 280         const char *subsystem = NULL, *interface = NULL;
 281         char *cur, *end, *sep;
 282         struct device *dev;
 283         int skip;
 284         bool add;
 285
 286         if (!strncmp(data, "add@", 4))
 287                 add = true;
 288         else if (!strncmp(data, "remove@", 7))
 289                 add = false;
 290         else
 291                 return;
 292
 293         skip = strlen(data) + 1;
 294         end = data + size;
 295
 296         for (cur = data + skip; cur < end; cur += skip) {
 297                 skip = strlen(cur) + 1;
 298
 299                 sep = strchr(cur, '=');
 300                 if (!sep)
 301                         continue;
 302
 303                 *sep = 0;
 304                 if (!strcmp(cur, "INTERFACE"))
 305                         interface = sep + 1;
 306                 else if (!strcmp(cur, "SUBSYSTEM")) {
 307                         subsystem = sep + 1;
 308                         if (strcmp(subsystem, "net") != 0)
 309                                 return;
 310                 }
 311                 if (subsystem && interface)
 312                         goto found;
 313         }
 314         return;
 315
 316 found:
 317         dev = device_get(interface, false);
 318         if (!dev)
 319                 return;
 320
 321         if (dev->type != &simple_device_type)
 322                 return;
 323
 324         if (add && system_if_force_external(dev->ifname))
 325                 return;
 326
 327         device_set_present(dev, add);
 328 }
 329
 330 static void
 331 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
 332 {
 333         struct event_socket *ev = container_of(u, struct event_socket, uloop);
 334         struct sockaddr_nl nla;
 335         unsigned char *buf = NULL;
 336         int size;
 337
 338         while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
 339                 if (nla.nl_pid == 0)
 340                         handle_hotplug_msg((char *) buf, size);
 341
 342                 free(buf);
 343         }
 344 }
 345
 346 static int system_rtnl_call(struct nl_msg *msg)
 347 {
 348         int ret;
 349
 350         ret = nl_send_auto_complete(sock_rtnl, msg);
 351         nlmsg_free(msg);
 352
 353         if (ret < 0)
 354                 return ret;
 355
 356         return nl_wait_for_ack(sock_rtnl);
 357 }
 358
 359 int system_bridge_delbr(struct device *bridge)
 360 {
 361         return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 362 }
 363
 364 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 365 {
 366         struct ifreq ifr;
 367
 368         memset(&ifr, 0, sizeof(ifr));
 369         if (dev)
 370                 ifr.ifr_ifindex = dev->ifindex;
 371         else
 372                 ifr.ifr_data = data;
 373         strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
 374         return ioctl(sock_ioctl, cmd, &ifr);
 375 }
 376
 377 static bool system_is_bridge(const char *name, char *buf, int buflen)
 378 {
 379         struct stat st;
 380
 381         snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
 382         if (stat(buf, &st) < 0)
 383                 return false;
 384
 385         return true;
 386 }
 387
 388 static char *system_get_bridge(const char *name, char *buf, int buflen)
 389 {
 390         char *path;
 391         ssize_t len;
 392         glob_t gl;
 393
 394         snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
 395         if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
 396                 return NULL;
 397
 398         if (gl.gl_pathc == 0)
 399                 return NULL;
 400
 401         len = readlink(gl.gl_pathv[0], buf, buflen);
 402         if (len < 0)
 403                 return NULL;
 404
 405         buf[len] = 0;
 406         path = strrchr(buf, '/');
 407         if (!path)
 408                 return NULL;
 409
 410         return path + 1;
 411 }
 412
 413 int system_bridge_addif(struct device *bridge, struct device *dev)
 414 {
 415         char *oldbr;
 416
 417         oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
 418         if (oldbr && !strcmp(oldbr, bridge->ifname))
 419                 return 0;
 420
 421         return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 422 }
 423
 424 int system_bridge_delif(struct device *bridge, struct device *dev)
 425 {
 426         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 427 }
 428
 429 static int system_if_resolve(struct device *dev)
 430 {
 431         struct ifreq ifr;
 432         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 433         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 434                 return ifr.ifr_ifindex;
 435         else
 436                 return 0;
 437 }
 438
 439 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 440 {
 441         struct ifreq ifr;
 442
 443         memset(&ifr, 0, sizeof(ifr));
 444         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 445         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 446         ifr.ifr_flags |= add;
 447         ifr.ifr_flags &= ~rem;
 448         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 449 }
 450
 451 struct clear_data {
 452         struct nl_msg *msg;
 453         struct device *dev;
 454         int type;
 455         int size;
 456         int af;
 457 };
 458
 459
 460 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 461 {
 462         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 463
 464         return ifa->ifa_index == ifindex;
 465 }
 466
 467 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 468 {
 469         struct rtmsg *r = NLMSG_DATA(hdr);
 470         struct nlattr *tb[__RTA_MAX];
 471
 472         if (r->rtm_protocol == RTPROT_KERNEL &&
 473             r->rtm_family == AF_INET6)
 474                 return false;
 475
 476         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 477         if (!tb[RTA_OIF])
 478                 return false;
 479
 480         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 481 }
 482
 483 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
 484 {
 485         return true;
 486 }
 487
 488 static int cb_clear_event(struct nl_msg *msg, void *arg)
 489 {
 490         struct clear_data *clr = arg;
 491         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 492         bool (*cb)(struct nlmsghdr *, int ifindex);
 493         int type;
 494
 495         switch(clr->type) {
 496         case RTM_GETADDR:
 497                 type = RTM_DELADDR;
 498                 if (hdr->nlmsg_type != RTM_NEWADDR)
 499                         return NL_SKIP;
 500
 501                 cb = check_ifaddr;
 502                 break;
 503         case RTM_GETROUTE:
 504                 type = RTM_DELROUTE;
 505                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 506                         return NL_SKIP;
 507
 508                 cb = check_route;
 509                 break;
 510         case RTM_GETRULE:
 511                 type = RTM_DELRULE;
 512                 if (hdr->nlmsg_type != RTM_NEWRULE)
 513                         return NL_SKIP;
 514
 515                 cb = check_rule;
 516                 break;
 517         default:
 518                 return NL_SKIP;
 519         }
 520
 521         if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
 522                 return NL_SKIP;
 523
 524         if (type == RTM_DELRULE)
 525                 D(SYSTEM, "Remove a rule\n");
 526         else
 527                 D(SYSTEM, "Remove %s from device %s\n",
 528                   type == RTM_DELADDR ? "an address" : "a route",
 529                   clr->dev->ifname);
 530         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 531         hdr = nlmsg_hdr(clr->msg);
 532         hdr->nlmsg_type = type;
 533         hdr->nlmsg_flags = NLM_F_REQUEST;
 534
 535         nl_socket_disable_auto_ack(sock_rtnl);
 536         nl_send_auto_complete(sock_rtnl, clr->msg);
 537         nl_socket_enable_auto_ack(sock_rtnl);
 538
 539         return NL_SKIP;
 540 }
 541
 542 static int
 543 cb_finish_event(struct nl_msg *msg, void *arg)
 544 {
 545         int *pending = arg;
 546         *pending = 0;
 547         return NL_STOP;
 548 }
 549
 550 static int
 551 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 552 {
 553         int *pending = arg;
 554         *pending = err->error;
 555         return NL_STOP;
 556 }
 557
 558 static void
 559 system_if_clear_entries(struct device *dev, int type, int af)
 560 {
 561         struct clear_data clr;
 562         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 563         struct rtmsg rtm = {
 564                 .rtm_family = af,
 565                 .rtm_flags = RTM_F_CLONED,
 566         };
 567         int flags = NLM_F_DUMP;
 568         int pending = 1;
 569
 570         clr.af = af;
 571         clr.dev = dev;
 572         clr.type = type;
 573         switch (type) {
 574         case RTM_GETADDR:
 575         case RTM_GETRULE:
 576                 clr.size = sizeof(struct rtgenmsg);
 577                 break;
 578         case RTM_GETROUTE:
 579                 clr.size = sizeof(struct rtmsg);
 580                 break;
 581         default:
 582                 return;
 583         }
 584
 585         if (!cb)
 586                 return;
 587
 588         clr.msg = nlmsg_alloc_simple(type, flags);
 589         if (!clr.msg)
 590                 goto out;
 591
 592         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 593         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 594         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 595         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 596
 597         nl_send_auto_complete(sock_rtnl, clr.msg);
 598         while (pending > 0)
 599                 nl_recvmsgs(sock_rtnl, cb);
 600
 601         nlmsg_free(clr.msg);
 602 out:
 603         nl_cb_put(cb);
 604 }
 605
 606 /*
 607  * Clear bridge (membership) state and bring down device
 608  */
 609 void system_if_clear_state(struct device *dev)
 610 {
 611         static char buf[256];
 612         char *bridge;
 613
 614         device_set_ifindex(dev, system_if_resolve(dev));
 615         if (dev->external || !dev->ifindex)
 616                 return;
 617
 618         system_if_flags(dev->ifname, 0, IFF_UP);
 619
 620         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 621                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 622                 system_bridge_delbr(dev);
 623                 return;
 624         }
 625
 626         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 627         if (bridge) {
 628                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 629                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 630         }
 631
 632         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 633         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 634         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 635         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 636         system_set_disable_ipv6(dev, "0");
 637 }
 638
 639 static inline unsigned long
 640 sec_to_jiffies(int val)
 641 {
 642         return (unsigned long) val * 100;
 643 }
 644
 645 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 646 {
 647         unsigned long args[4] = {};
 648
 649         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 650                 return -1;
 651
 652         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 653         args[1] = !!cfg->stp;
 654         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 655
 656         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 657         args[1] = sec_to_jiffies(cfg->forward_delay);
 658         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 659
 660         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 661                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 662
 663         args[0] = BRCTL_SET_BRIDGE_PRIORITY;
 664         args[1] = cfg->priority;
 665         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 666
 667         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 668                 args[0] = BRCTL_SET_AGEING_TIME;
 669                 args[1] = sec_to_jiffies(cfg->ageing_time);
 670                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 671         }
 672
 673         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 674                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 675                 args[1] = sec_to_jiffies(cfg->hello_time);
 676                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 677         }
 678
 679         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 680                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 681                 args[1] = sec_to_jiffies(cfg->max_age);
 682                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 683         }
 684
 685         return 0;
 686 }
 687
 688 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
 689 {
 690         struct nl_msg *msg;
 691         struct nlattr *linkinfo, *data;
 692         struct ifinfomsg iim = { .ifi_family = AF_INET };
 693         int ifindex = system_if_resolve(dev);
 694         int i, rv;
 695         static const struct {
 696                 const char *name;
 697                 enum macvlan_mode val;
 698         } modes[] = {
 699                 { "private", MACVLAN_MODE_PRIVATE },
 700                 { "vepa", MACVLAN_MODE_VEPA },
 701                 { "bridge", MACVLAN_MODE_BRIDGE },
 702                 { "passthru", MACVLAN_MODE_PASSTHRU },
 703         };
 704
 705         if (ifindex == 0)
 706                 return -ENOENT;
 707
 708         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 709
 710         if (!msg)
 711                 return -1;
 712
 713         nlmsg_append(msg, &iim, sizeof(iim), 0);
 714
 715         if (cfg->flags & MACVLAN_OPT_MACADDR)
 716                 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
 717         nla_put(msg, IFLA_IFNAME, IFNAMSIZ, macvlan->ifname);
 718         nla_put_u32(msg, IFLA_LINK, ifindex);
 719
 720         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 721                 goto nla_put_failure;
 722
 723         nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
 724
 725         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 726                 goto nla_put_failure;
 727
 728         if (cfg->mode) {
 729                 for (i = 0; i < ARRAY_SIZE(modes); i++) {
 730                         if (strcmp(cfg->mode, modes[i].name) != 0)
 731                                 continue;
 732
 733                         nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
 734                         break;
 735                 }
 736         }
 737
 738         nla_nest_end(msg, data);
 739         nla_nest_end(msg, linkinfo);
 740
 741         rv = system_rtnl_call(msg);
 742         if (rv)
 743                 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
 744
 745         return rv;
 746
 747 nla_put_failure:
 748         nlmsg_free(msg);
 749         return -ENOMEM;
 750 }
 751
 752 int system_macvlan_del(struct device *macvlan)
 753 {
 754         struct nl_msg *msg;
 755         struct ifinfomsg iim;
 756
 757         iim.ifi_family = AF_INET;
 758         iim.ifi_index  = 0;
 759
 760         msg = nlmsg_alloc_simple(RTM_DELLINK, 0);
 761
 762         if (!msg)
 763                 return -1;
 764
 765         nlmsg_append(msg, &iim, sizeof(iim), 0);
 766
 767         nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
 768         nla_put(msg, IFLA_IFNAME, sizeof(macvlan->ifname), macvlan->ifname);
 769
 770         system_rtnl_call(msg);
 771
 772         return 0;
 773 }
 774
 775 static int system_vlan(struct device *dev, int id)
 776 {
 777         struct vlan_ioctl_args ifr = {
 778                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 779                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 780         };
 781
 782         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 783
 784         if (id < 0) {
 785                 ifr.cmd = DEL_VLAN_CMD;
 786                 ifr.u.VID = 0;
 787         } else {
 788                 ifr.cmd = ADD_VLAN_CMD;
 789                 ifr.u.VID = id;
 790         }
 791         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 792         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 793 }
 794
 795 int system_vlan_add(struct device *dev, int id)
 796 {
 797         return system_vlan(dev, id);
 798 }
 799
 800 int system_vlan_del(struct device *dev)
 801 {
 802         return system_vlan(dev, -1);
 803 }
 804
 805 static void
 806 system_if_get_settings(struct device *dev, struct device_settings *s)
 807 {
 808         struct ifreq ifr;
 809         char buf[10];
 810
 811         memset(&ifr, 0, sizeof(ifr));
 812         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 813
 814         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 815                 s->mtu = ifr.ifr_mtu;
 816                 s->flags |= DEV_OPT_MTU;
 817         }
 818
 819         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 820                 s->txqueuelen = ifr.ifr_qlen;
 821                 s->flags |= DEV_OPT_TXQUEUELEN;
 822         }
 823
 824         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 825                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 826                 s->flags |= DEV_OPT_MACADDR;
 827         }
 828
 829         if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
 830                 s->ipv6 = !strtoul(buf, NULL, 0);
 831                 s->flags |= DEV_OPT_IPV6;
 832         }
 833 }
 834
 835 void
 836 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
 837 {
 838         struct ifreq ifr;
 839
 840         if (!apply_mask)
 841                 return;
 842
 843         memset(&ifr, 0, sizeof(ifr));
 844         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 845         if (s->flags & DEV_OPT_MTU & apply_mask) {
 846                 ifr.ifr_mtu = s->mtu;
 847                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 848                         s->flags &= ~DEV_OPT_MTU;
 849         }
 850         if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
 851                 ifr.ifr_qlen = s->txqueuelen;
 852                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 853                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 854         }
 855         if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
 856                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 857                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 858                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 859                         s->flags &= ~DEV_OPT_MACADDR;
 860         }
 861         if (s->flags & DEV_OPT_IPV6 & apply_mask)
 862                 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
 863 }
 864
 865 int system_if_up(struct device *dev)
 866 {
 867         system_if_get_settings(dev, &dev->orig_settings);
 868         system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
 869         device_set_ifindex(dev, system_if_resolve(dev));
 870         return system_if_flags(dev->ifname, IFF_UP, 0);
 871 }
 872
 873 int system_if_down(struct device *dev)
 874 {
 875         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
 876         dev->orig_settings.flags &= dev->settings.flags;
 877         system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
 878         return ret;
 879 }
 880
 881 struct if_check_data {
 882         struct device *dev;
 883         int pending;
 884         int ret;
 885 };
 886
 887 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
 888 {
 889         struct nlmsghdr *nh = nlmsg_hdr(msg);
 890         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 891         struct if_check_data *chk = (struct if_check_data *)arg;
 892
 893         if (nh->nlmsg_type != RTM_NEWLINK)
 894                 return NL_SKIP;
 895
 896         device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
 897         device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
 898
 899         return NL_OK;
 900 }
 901
 902 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
 903 {
 904         struct if_check_data *chk = (struct if_check_data *)arg;
 905         chk->pending = 0;
 906         return NL_STOP;
 907 }
 908
 909 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 910 {
 911         struct if_check_data *chk = (struct if_check_data *)arg;
 912
 913         device_set_present(chk->dev, false);
 914         device_set_link(chk->dev, false);
 915         chk->pending = err->error;
 916
 917         return NL_STOP;
 918 }
 919
 920 int system_if_check(struct device *dev)
 921 {
 922         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 923         struct nl_msg *msg;
 924         struct ifinfomsg ifi = {
 925                 .ifi_family = AF_UNSPEC,
 926                 .ifi_index = 0,
 927         };
 928         struct if_check_data chk = {
 929                 .dev = dev,
 930                 .pending = 1,
 931         };
 932         int ret = 1;
 933
 934         msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
 935         if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
 936             nla_put_string(msg, IFLA_IFNAME, dev->ifname))
 937                 goto out;
 938
 939         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
 940         nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
 941         nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
 942
 943         nl_send_auto_complete(sock_rtnl, msg);
 944         while (chk.pending > 0)
 945                 nl_recvmsgs(sock_rtnl, cb);
 946
 947         nlmsg_free(msg);
 948         ret = chk.pending;
 949
 950 out:
 951         nl_cb_put(cb);
 952         return ret;
 953 }
 954
 955 struct device *
 956 system_if_get_parent(struct device *dev)
 957 {
 958         char buf[64], *devname;
 959         int ifindex, iflink, len;
 960         FILE *f;
 961
 962         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
 963         f = fopen(buf, "r");
 964         if (!f)
 965                 return NULL;
 966
 967         len = fread(buf, 1, sizeof(buf) - 1, f);
 968         fclose(f);
 969
 970         if (len <= 0)
 971                 return NULL;
 972
 973         buf[len] = 0;
 974         iflink = strtoul(buf, NULL, 0);
 975         ifindex = system_if_resolve(dev);
 976         if (!iflink || iflink == ifindex)
 977                 return NULL;
 978
 979         devname = if_indextoname(iflink, buf);
 980         if (!devname)
 981                 return NULL;
 982
 983         return device_get(devname, true);
 984 }
 985
 986 static bool
 987 read_string_file(int dir_fd, const char *file, char *buf, int len)
 988 {
 989         bool ret = false;
 990         char *c;
 991         int fd;
 992
 993         fd = openat(dir_fd, file, O_RDONLY);
 994         if (fd < 0)
 995                 return false;
 996
 997 retry:
 998         len = read(fd, buf, len - 1);
 999         if (len < 0) {
1000                 if (errno == EINTR)
1001                         goto retry;
1002         } else if (len > 0) {
1003                         buf[len] = 0;
1004
1005                         c = strchr(buf, '\n');
1006                         if (c)
1007                                 *c = 0;
1008
1009                         ret = true;
1010         }
1011
1012         close(fd);
1013
1014         return ret;
1015 }
1016
1017 static bool
1018 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1019 {
1020         char buf[64];
1021         bool ret = false;
1022
1023         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1024         if (ret)
1025                 *val = strtoull(buf, NULL, 0);
1026
1027         return ret;
1028 }
1029
1030 /* Assume advertised flags == supported flags */
1031 static const struct {
1032         uint32_t mask;
1033         const char *name;
1034 } ethtool_link_modes[] = {
1035         { ADVERTISED_10baseT_Half, "10H" },
1036         { ADVERTISED_10baseT_Full, "10F" },
1037         { ADVERTISED_100baseT_Half, "100H" },
1038         { ADVERTISED_100baseT_Full, "100F" },
1039         { ADVERTISED_1000baseT_Half, "1000H" },
1040         { ADVERTISED_1000baseT_Full, "1000F" },
1041 };
1042
1043 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1044 {
1045         int i;
1046         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1047                 if (mask & ethtool_link_modes[i].mask)
1048                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1049         }
1050 }
1051
1052 bool
1053 system_if_force_external(const char *ifname)
1054 {
1055         char buf[64];
1056         struct stat s;
1057
1058         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1059         return stat(buf, &s) == 0;
1060 }
1061
1062 int
1063 system_if_dump_info(struct device *dev, struct blob_buf *b)
1064 {
1065         struct ethtool_cmd ecmd;
1066         struct ifreq ifr;
1067         char buf[64], *s;
1068         void *c;
1069         int dir_fd;
1070
1071         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1072         dir_fd = open(buf, O_DIRECTORY);
1073
1074         memset(&ecmd, 0, sizeof(ecmd));
1075         memset(&ifr, 0, sizeof(ifr));
1076         strcpy(ifr.ifr_name, dev->ifname);
1077         ifr.ifr_data = (caddr_t) &ecmd;
1078         ecmd.cmd = ETHTOOL_GSET;
1079
1080         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1081                 c = blobmsg_open_array(b, "link-advertising");
1082                 system_add_link_modes(b, ecmd.advertising);
1083                 blobmsg_close_array(b, c);
1084
1085                 c = blobmsg_open_array(b, "link-supported");
1086                 system_add_link_modes(b, ecmd.supported);
1087                 blobmsg_close_array(b, c);
1088
1089                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1090                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1091                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1092                 blobmsg_add_string_buffer(b);
1093         }
1094
1095         close(dir_fd);
1096         return 0;
1097 }
1098
1099 int
1100 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1101 {
1102         const char *const counters[] = {
1103                 "collisions",     "rx_frame_errors",   "tx_compressed",
1104                 "multicast",      "rx_length_errors",  "tx_dropped",
1105                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
1106                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
1107                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
1108                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
1109                 "rx_errors",      "tx_bytes",          "tx_window_errors",
1110                 "rx_fifo_errors", "tx_carrier_errors",
1111         };
1112         char buf[64];
1113         int stats_dir;
1114         int i;
1115         uint64_t val = 0;
1116
1117         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1118         stats_dir = open(buf, O_DIRECTORY);
1119         if (stats_dir < 0)
1120                 return -1;
1121
1122         for (i = 0; i < ARRAY_SIZE(counters); i++)
1123                 if (read_uint64_file(stats_dir, counters[i], &val))
1124                         blobmsg_add_u64(b, counters[i], val);
1125
1126         close(stats_dir);
1127         return 0;
1128 }
1129
1130 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1131 {
1132         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1133         int alen = v4 ? 4 : 16;
1134         unsigned int flags = 0;
1135         struct ifaddrmsg ifa = {
1136                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1137                 .ifa_prefixlen = addr->mask,
1138                 .ifa_index = dev->ifindex,
1139         };
1140
1141         struct nl_msg *msg;
1142         if (cmd == RTM_NEWADDR)
1143                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1144
1145         msg = nlmsg_alloc_simple(cmd, flags);
1146         if (!msg)
1147                 return -1;
1148
1149         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1150         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1151         if (v4) {
1152                 if (addr->broadcast)
1153                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1154                 if (addr->point_to_point)
1155                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1156         } else {
1157                 time_t now = system_get_rtime();
1158                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1159
1160                 if (addr->preferred_until) {
1161                         int64_t preferred = addr->preferred_until - now;
1162                         if (preferred < 0)
1163                                 preferred = 0;
1164                         else if (preferred > UINT32_MAX)
1165                                 preferred = UINT32_MAX;
1166
1167                         cinfo.ifa_prefered = preferred;
1168                 }
1169
1170                 if (addr->valid_until) {
1171                         int64_t valid = addr->valid_until - now;
1172                         if (valid <= 0)
1173                                 return -1;
1174                         else if (valid > UINT32_MAX)
1175                                 valid = UINT32_MAX;
1176
1177                         cinfo.ifa_valid = valid;
1178                 }
1179
1180                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1181         }
1182
1183         return system_rtnl_call(msg);
1184 }
1185
1186 int system_add_address(struct device *dev, struct device_addr *addr)
1187 {
1188         return system_addr(dev, addr, RTM_NEWADDR);
1189 }
1190
1191 int system_del_address(struct device *dev, struct device_addr *addr)
1192 {
1193         return system_addr(dev, addr, RTM_DELADDR);
1194 }
1195
1196 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1197 {
1198         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1199         bool have_gw;
1200         unsigned int flags = 0;
1201
1202         if (alen == 4)
1203                 have_gw = !!route->nexthop.in.s_addr;
1204         else
1205                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1206                         route->nexthop.in6.s6_addr32[1] ||
1207                         route->nexthop.in6.s6_addr32[2] ||
1208                         route->nexthop.in6.s6_addr32[3];
1209
1210         unsigned char scope = (cmd == RTM_DELROUTE) ? RT_SCOPE_NOWHERE :
1211                         (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1212
1213         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1214                         ? route->table : RT_TABLE_MAIN;
1215
1216         struct rtmsg rtm = {
1217                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1218                 .rtm_dst_len = route->mask,
1219                 .rtm_src_len = route->sourcemask,
1220                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1221                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1222                 .rtm_scope = scope,
1223                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1224         };
1225         struct nl_msg *msg;
1226
1227         if (cmd == RTM_NEWROUTE) {
1228                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1229
1230                 if (!dev) { // Add null-route
1231                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1232                         rtm.rtm_type = RTN_UNREACHABLE;
1233                 }
1234         }
1235
1236         msg = nlmsg_alloc_simple(cmd, flags);
1237         if (!msg)
1238                 return -1;
1239
1240         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1241
1242         if (route->mask)
1243                 nla_put(msg, RTA_DST, alen, &route->addr);
1244
1245         if (route->sourcemask)
1246                 nla_put(msg, RTA_SRC, alen, &route->source);
1247
1248         if (route->metric > 0)
1249                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1250
1251         if (have_gw)
1252                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1253
1254         if (dev)
1255                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1256
1257         if (table >= 256)
1258                 nla_put_u32(msg, RTA_TABLE, table);
1259
1260         if (route->flags & DEVROUTE_MTU) {
1261                 struct nlattr *metrics;
1262
1263                 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1264                         goto nla_put_failure;
1265
1266                 nla_put_u32(msg, RTAX_MTU, route->mtu);
1267
1268                 nla_nest_end(msg, metrics);
1269         }
1270
1271         return system_rtnl_call(msg);
1272
1273 nla_put_failure:
1274         nlmsg_free(msg);
1275         return -ENOMEM;
1276 }
1277
1278 int system_add_route(struct device *dev, struct device_route *route)
1279 {
1280         return system_rt(dev, route, RTM_NEWROUTE);
1281 }
1282
1283 int system_del_route(struct device *dev, struct device_route *route)
1284 {
1285         return system_rt(dev, route, RTM_DELROUTE);
1286 }
1287
1288 int system_flush_routes(void)
1289 {
1290         const char *names[] = {
1291                 "/proc/sys/net/ipv4/route/flush",
1292                 "/proc/sys/net/ipv6/route/flush"
1293         };
1294         int fd, i;
1295
1296         for (i = 0; i < ARRAY_SIZE(names); i++) {
1297                 fd = open(names[i], O_WRONLY);
1298                 if (fd < 0)
1299                         continue;
1300
1301                 if (write(fd, "-1", 2)) {}
1302                 close(fd);
1303         }
1304         return 0;
1305 }
1306
1307 bool system_resolve_rt_table(const char *name, unsigned int *id)
1308 {
1309         FILE *f;
1310         char *e, buf[128];
1311         unsigned int n, table = RT_TABLE_UNSPEC;
1312
1313         /* first try to parse table as number */
1314         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1315                 table = n;
1316
1317         /* handle well known aliases */
1318         else if (!strcmp(name, "default"))
1319                 table = RT_TABLE_DEFAULT;
1320         else if (!strcmp(name, "main"))
1321                 table = RT_TABLE_MAIN;
1322         else if (!strcmp(name, "local"))
1323                 table = RT_TABLE_LOCAL;
1324
1325         /* try to look up name in /etc/iproute2/rt_tables */
1326         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1327         {
1328                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1329                 {
1330                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1331                                 continue;
1332
1333                         n = strtoul(e, NULL, 10);
1334                         e = strtok(NULL, " \t\n");
1335
1336                         if (e && !strcmp(e, name))
1337                         {
1338                                 table = n;
1339                                 break;
1340                         }
1341                 }
1342
1343                 fclose(f);
1344         }
1345
1346         if (table == RT_TABLE_UNSPEC)
1347                 return false;
1348
1349         /* do not consider main table special */
1350         if (table == RT_TABLE_MAIN)
1351                 table = RT_TABLE_UNSPEC;
1352
1353         *id = table;
1354         return true;
1355 }
1356
1357 static int system_iprule(struct iprule *rule, int cmd)
1358 {
1359         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1360
1361         struct nl_msg *msg;
1362         struct rtmsg rtm = {
1363                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1364                 .rtm_protocol = RTPROT_STATIC,
1365                 .rtm_scope = RT_SCOPE_UNIVERSE,
1366                 .rtm_table = RT_TABLE_UNSPEC,
1367                 .rtm_type = RTN_UNSPEC,
1368                 .rtm_flags = 0,
1369         };
1370
1371         if (cmd == RTM_NEWRULE) {
1372                 rtm.rtm_type = RTN_UNICAST;
1373                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1374         }
1375
1376         if (rule->invert)
1377                 rtm.rtm_flags |= FIB_RULE_INVERT;
1378
1379         if (rule->flags & IPRULE_SRC)
1380                 rtm.rtm_src_len = rule->src_mask;
1381
1382         if (rule->flags & IPRULE_DEST)
1383                 rtm.rtm_dst_len = rule->dest_mask;
1384
1385         if (rule->flags & IPRULE_TOS)
1386                 rtm.rtm_tos = rule->tos;
1387
1388         if (rule->flags & IPRULE_LOOKUP) {
1389                 if (rule->lookup < 256)
1390                         rtm.rtm_table = rule->lookup;
1391         }
1392
1393         if (rule->flags & IPRULE_ACTION)
1394                 rtm.rtm_type = rule->action;
1395         else if (rule->flags & IPRULE_GOTO)
1396                 rtm.rtm_type = FR_ACT_GOTO;
1397         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1398                 rtm.rtm_type = FR_ACT_NOP;
1399
1400         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1401
1402         if (!msg)
1403                 return -1;
1404
1405         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1406
1407         if (rule->flags & IPRULE_IN)
1408                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1409
1410         if (rule->flags & IPRULE_OUT)
1411                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1412
1413         if (rule->flags & IPRULE_SRC)
1414                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1415
1416         if (rule->flags & IPRULE_DEST)
1417                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1418
1419         if (rule->flags & IPRULE_PRIORITY)
1420                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1421         else if (cmd == RTM_NEWRULE)
1422                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1423
1424         if (rule->flags & IPRULE_FWMARK)
1425                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1426
1427         if (rule->flags & IPRULE_FWMASK)
1428                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1429
1430         if (rule->flags & IPRULE_LOOKUP) {
1431                 if (rule->lookup >= 256)
1432                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1433         }
1434
1435         if (rule->flags & IPRULE_GOTO)
1436                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1437
1438         return system_rtnl_call(msg);
1439 }
1440
1441 int system_add_iprule(struct iprule *rule)
1442 {
1443         return system_iprule(rule, RTM_NEWRULE);
1444 }
1445
1446 int system_del_iprule(struct iprule *rule)
1447 {
1448         return system_iprule(rule, RTM_DELRULE);
1449 }
1450
1451 int system_flush_iprules(void)
1452 {
1453         int rv = 0;
1454         struct iprule rule;
1455
1456         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1457         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1458
1459         memset(&rule, 0, sizeof(rule));
1460
1461
1462         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1463
1464         rule.priority = 0;
1465         rule.lookup = RT_TABLE_LOCAL;
1466         rv |= system_iprule(&rule, RTM_NEWRULE);
1467
1468         rule.priority = 32766;
1469         rule.lookup = RT_TABLE_MAIN;
1470         rv |= system_iprule(&rule, RTM_NEWRULE);
1471
1472         rule.priority = 32767;
1473         rule.lookup = RT_TABLE_DEFAULT;
1474         rv |= system_iprule(&rule, RTM_NEWRULE);
1475
1476
1477         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1478
1479         rule.priority = 0;
1480         rule.lookup = RT_TABLE_LOCAL;
1481         rv |= system_iprule(&rule, RTM_NEWRULE);
1482
1483         rule.priority = 32766;
1484         rule.lookup = RT_TABLE_MAIN;
1485         rv |= system_iprule(&rule, RTM_NEWRULE);
1486
1487         return rv;
1488 }
1489
1490 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1491 {
1492         char *e;
1493         unsigned int n;
1494
1495         if (!strcmp(action, "local"))
1496                 n = RTN_LOCAL;
1497         else if (!strcmp(action, "nat"))
1498                 n = RTN_NAT;
1499         else if (!strcmp(action, "broadcast"))
1500                 n = RTN_BROADCAST;
1501         else if (!strcmp(action, "anycast"))
1502                 n = RTN_ANYCAST;
1503         else if (!strcmp(action, "multicast"))
1504                 n = RTN_MULTICAST;
1505         else if (!strcmp(action, "prohibit"))
1506                 n = RTN_PROHIBIT;
1507         else if (!strcmp(action, "unreachable"))
1508                 n = RTN_UNREACHABLE;
1509         else if (!strcmp(action, "blackhole"))
1510                 n = RTN_BLACKHOLE;
1511         else if (!strcmp(action, "xresolve"))
1512                 n = RTN_XRESOLVE;
1513         else if (!strcmp(action, "unicast"))
1514                 n = RTN_UNICAST;
1515         else if (!strcmp(action, "throw"))
1516                 n = RTN_THROW;
1517         else if (!strcmp(action, "failed_policy"))
1518                 n = RTN_FAILED_POLICY;
1519         else {
1520                 n = strtoul(action, &e, 0);
1521                 if (!e || *e || e == action || n > 255)
1522                         return false;
1523         }
1524
1525         *id = n;
1526         return true;
1527 }
1528
1529 time_t system_get_rtime(void)
1530 {
1531         struct timespec ts;
1532         struct timeval tv;
1533
1534         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1535                 return ts.tv_sec;
1536
1537         if (gettimeofday(&tv, NULL) == 0)
1538                 return tv.tv_sec;
1539
1540         return 0;
1541 }
1542
1543 #ifndef IP_DF
1544 #define IP_DF       0x4000
1545 #endif
1546
1547 static int tunnel_ioctl(const char *name, int cmd, void *p)
1548 {
1549         struct ifreq ifr;
1550
1551         memset(&ifr, 0, sizeof(ifr));
1552         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1553         ifr.ifr_ifru.ifru_data = p;
1554         return ioctl(sock_ioctl, cmd, &ifr);
1555 }
1556
1557 int system_del_ip_tunnel(const char *name)
1558 {
1559         return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1560 }
1561
1562 int system_update_ipv6_mtu(struct device *dev, int mtu)
1563 {
1564         int ret = -1;
1565         char buf[64];
1566         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1567                         dev->ifname);
1568
1569         int fd = open(buf, O_RDWR);
1570         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1571         if (len < 0)
1572                 goto out;
1573
1574         buf[len] = 0;
1575         ret = atoi(buf);
1576
1577         if (!mtu || ret <= mtu)
1578                 goto out;
1579
1580         lseek(fd, 0, SEEK_SET);
1581         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1582                 ret = -1;
1583
1584 out:
1585         close(fd);
1586         return ret;
1587 }
1588
1589 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1590 {
1591         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1592         struct blob_attr *cur;
1593         bool set_df = true;
1594         const char *str;
1595
1596         system_del_ip_tunnel(name);
1597
1598         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1599                 blob_data(attr), blob_len(attr));
1600
1601         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1602                 return -EINVAL;
1603         str = blobmsg_data(cur);
1604
1605         if ((cur = tb[TUNNEL_ATTR_DF]))
1606                 set_df = blobmsg_get_bool(cur);
1607
1608         unsigned int ttl = 0;
1609         if ((cur = tb[TUNNEL_ATTR_TTL])) {
1610                 ttl = blobmsg_get_u32(cur);
1611                 if (ttl > 255 || (!set_df && ttl))
1612                         return -EINVAL;
1613         }
1614
1615         unsigned int link = 0;
1616         if ((cur = tb[TUNNEL_ATTR_LINK])) {
1617                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
1618                 if (!iface)
1619                         return -EINVAL;
1620
1621                 if (iface->l3_dev.dev)
1622                         link = iface->l3_dev.dev->ifindex;
1623         }
1624
1625         if (!strcmp(str, "sit")) {
1626                 struct ip_tunnel_parm p = {
1627                         .link = link,
1628                         .iph = {
1629                                 .version = 4,
1630                                 .ihl = 5,
1631                                 .frag_off = set_df ? htons(IP_DF) : 0,
1632                                 .protocol = IPPROTO_IPV6,
1633                                 .ttl = ttl
1634                         }
1635                 };
1636
1637                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1638                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1639                         return -EINVAL;
1640
1641                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1642                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1643                         return -EINVAL;
1644
1645                 strncpy(p.name, name, sizeof(p.name));
1646                 if (tunnel_ioctl("sit0", SIOCADDTUNNEL, &p) < 0)
1647                         return -1;
1648
1649 #ifdef SIOCADD6RD
1650                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
1651                         unsigned int mask;
1652                         struct ip_tunnel_6rd p6;
1653
1654                         memset(&p6, 0, sizeof(p6));
1655
1656                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
1657                                                 &p6.prefix, &mask) || mask > 128)
1658                                 return -EINVAL;
1659                         p6.prefixlen = mask;
1660
1661                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
1662                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
1663                                                         &p6.relay_prefix, &mask) || mask > 32)
1664                                         return -EINVAL;
1665                                 p6.relay_prefixlen = mask;
1666                         }
1667
1668                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
1669                                 system_del_ip_tunnel(name);
1670                                 return -1;
1671                         }
1672                 }
1673 #endif
1674         } else if (!strcmp(str, "ipip6")) {
1675                 struct ip6_tnl_parm p = {
1676                         .link = link,
1677                         .proto = IPPROTO_IPIP,
1678                         .hop_limit = (ttl) ? ttl : 64,
1679                         .encap_limit = 4,
1680                 };
1681
1682                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1683                                 inet_pton(AF_INET6, blobmsg_data(cur), &p.laddr) < 1)
1684                         return -EINVAL;
1685
1686                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1687                                 inet_pton(AF_INET6, blobmsg_data(cur), &p.raddr) < 1)
1688                         return -EINVAL;
1689
1690                 strncpy(p.name, name, sizeof(p.name));
1691                 if (tunnel_ioctl("ip6tnl0", SIOCADDTUNNEL, &p) < 0)
1692                         return -1;
1693         } else
1694                 return -EINVAL;
1695
1696         return 0;
1697 }