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         if (dev->external)
 615                 return;
 616
 617         device_set_ifindex(dev, system_if_resolve(dev));
 618         if (!dev->ifindex)
 619                 return;
 620
 621         system_if_flags(dev->ifname, 0, IFF_UP);
 622
 623         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 624                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 625                 system_bridge_delbr(dev);
 626                 return;
 627         }
 628
 629         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 630         if (bridge) {
 631                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 632                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 633         }
 634
 635         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 636         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 637         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 638         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 639         system_set_disable_ipv6(dev, "0");
 640 }
 641
 642 static inline unsigned long
 643 sec_to_jiffies(int val)
 644 {
 645         return (unsigned long) val * 100;
 646 }
 647
 648 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 649 {
 650         unsigned long args[4] = {};
 651
 652         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 653                 return -1;
 654
 655         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 656         args[1] = !!cfg->stp;
 657         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 658
 659         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 660         args[1] = sec_to_jiffies(cfg->forward_delay);
 661         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 662
 663         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 664                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 665
 666         args[0] = BRCTL_SET_BRIDGE_PRIORITY;
 667         args[1] = cfg->priority;
 668         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 669
 670         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 671                 args[0] = BRCTL_SET_AGEING_TIME;
 672                 args[1] = sec_to_jiffies(cfg->ageing_time);
 673                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 674         }
 675
 676         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 677                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 678                 args[1] = sec_to_jiffies(cfg->hello_time);
 679                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 680         }
 681
 682         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 683                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 684                 args[1] = sec_to_jiffies(cfg->max_age);
 685                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 686         }
 687
 688         return 0;
 689 }
 690
 691 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
 692 {
 693         struct nl_msg *msg;
 694         struct nlattr *linkinfo, *data;
 695         struct ifinfomsg iim = { .ifi_family = AF_INET };
 696         int ifindex = system_if_resolve(dev);
 697         int i, rv;
 698         static const struct {
 699                 const char *name;
 700                 enum macvlan_mode val;
 701         } modes[] = {
 702                 { "private", MACVLAN_MODE_PRIVATE },
 703                 { "vepa", MACVLAN_MODE_VEPA },
 704                 { "bridge", MACVLAN_MODE_BRIDGE },
 705                 { "passthru", MACVLAN_MODE_PASSTHRU },
 706         };
 707
 708         if (ifindex == 0)
 709                 return -ENOENT;
 710
 711         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 712
 713         if (!msg)
 714                 return -1;
 715
 716         nlmsg_append(msg, &iim, sizeof(iim), 0);
 717
 718         if (cfg->flags & MACVLAN_OPT_MACADDR)
 719                 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
 720         nla_put(msg, IFLA_IFNAME, IFNAMSIZ, macvlan->ifname);
 721         nla_put_u32(msg, IFLA_LINK, ifindex);
 722
 723         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 724                 goto nla_put_failure;
 725
 726         nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
 727
 728         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 729                 goto nla_put_failure;
 730
 731         if (cfg->mode) {
 732                 for (i = 0; i < ARRAY_SIZE(modes); i++) {
 733                         if (strcmp(cfg->mode, modes[i].name) != 0)
 734                                 continue;
 735
 736                         nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
 737                         break;
 738                 }
 739         }
 740
 741         nla_nest_end(msg, data);
 742         nla_nest_end(msg, linkinfo);
 743
 744         rv = system_rtnl_call(msg);
 745         if (rv)
 746                 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
 747
 748         return rv;
 749
 750 nla_put_failure:
 751         nlmsg_free(msg);
 752         return -ENOMEM;
 753 }
 754
 755 int system_macvlan_del(struct device *macvlan)
 756 {
 757         struct nl_msg *msg;
 758         struct ifinfomsg iim;
 759
 760         iim.ifi_family = AF_INET;
 761         iim.ifi_index  = 0;
 762
 763         msg = nlmsg_alloc_simple(RTM_DELLINK, 0);
 764
 765         if (!msg)
 766                 return -1;
 767
 768         nlmsg_append(msg, &iim, sizeof(iim), 0);
 769
 770         nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
 771         nla_put(msg, IFLA_IFNAME, sizeof(macvlan->ifname), macvlan->ifname);
 772
 773         system_rtnl_call(msg);
 774
 775         return 0;
 776 }
 777
 778 static int system_vlan(struct device *dev, int id)
 779 {
 780         struct vlan_ioctl_args ifr = {
 781                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 782                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 783         };
 784
 785         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 786
 787         if (id < 0) {
 788                 ifr.cmd = DEL_VLAN_CMD;
 789                 ifr.u.VID = 0;
 790         } else {
 791                 ifr.cmd = ADD_VLAN_CMD;
 792                 ifr.u.VID = id;
 793         }
 794         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 795         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 796 }
 797
 798 int system_vlan_add(struct device *dev, int id)
 799 {
 800         return system_vlan(dev, id);
 801 }
 802
 803 int system_vlan_del(struct device *dev)
 804 {
 805         return system_vlan(dev, -1);
 806 }
 807
 808 static void
 809 system_if_get_settings(struct device *dev, struct device_settings *s)
 810 {
 811         struct ifreq ifr;
 812         char buf[10];
 813
 814         memset(&ifr, 0, sizeof(ifr));
 815         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 816
 817         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 818                 s->mtu = ifr.ifr_mtu;
 819                 s->flags |= DEV_OPT_MTU;
 820         }
 821
 822         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 823                 s->txqueuelen = ifr.ifr_qlen;
 824                 s->flags |= DEV_OPT_TXQUEUELEN;
 825         }
 826
 827         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 828                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 829                 s->flags |= DEV_OPT_MACADDR;
 830         }
 831
 832         if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
 833                 s->ipv6 = !strtoul(buf, NULL, 0);
 834                 s->flags |= DEV_OPT_IPV6;
 835         }
 836 }
 837
 838 void
 839 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
 840 {
 841         struct ifreq ifr;
 842
 843         if (!apply_mask)
 844                 return;
 845
 846         memset(&ifr, 0, sizeof(ifr));
 847         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 848         if (s->flags & DEV_OPT_MTU & apply_mask) {
 849                 ifr.ifr_mtu = s->mtu;
 850                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 851                         s->flags &= ~DEV_OPT_MTU;
 852         }
 853         if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
 854                 ifr.ifr_qlen = s->txqueuelen;
 855                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 856                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 857         }
 858         if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
 859                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 860                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 861                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 862                         s->flags &= ~DEV_OPT_MACADDR;
 863         }
 864         if (s->flags & DEV_OPT_IPV6 & apply_mask)
 865                 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
 866 }
 867
 868 int system_if_up(struct device *dev)
 869 {
 870         system_if_get_settings(dev, &dev->orig_settings);
 871         system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
 872         device_set_ifindex(dev, system_if_resolve(dev));
 873         return system_if_flags(dev->ifname, IFF_UP, 0);
 874 }
 875
 876 int system_if_down(struct device *dev)
 877 {
 878         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
 879         dev->orig_settings.flags &= dev->settings.flags;
 880         system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
 881         return ret;
 882 }
 883
 884 struct if_check_data {
 885         struct device *dev;
 886         int pending;
 887         int ret;
 888 };
 889
 890 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
 891 {
 892         struct nlmsghdr *nh = nlmsg_hdr(msg);
 893         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 894         struct if_check_data *chk = (struct if_check_data *)arg;
 895
 896         if (nh->nlmsg_type != RTM_NEWLINK)
 897                 return NL_SKIP;
 898
 899         device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
 900         device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
 901
 902         return NL_OK;
 903 }
 904
 905 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
 906 {
 907         struct if_check_data *chk = (struct if_check_data *)arg;
 908         chk->pending = 0;
 909         return NL_STOP;
 910 }
 911
 912 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 913 {
 914         struct if_check_data *chk = (struct if_check_data *)arg;
 915
 916         device_set_present(chk->dev, false);
 917         device_set_link(chk->dev, false);
 918         chk->pending = err->error;
 919
 920         return NL_STOP;
 921 }
 922
 923 int system_if_check(struct device *dev)
 924 {
 925         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 926         struct nl_msg *msg;
 927         struct ifinfomsg ifi = {
 928                 .ifi_family = AF_UNSPEC,
 929                 .ifi_index = 0,
 930         };
 931         struct if_check_data chk = {
 932                 .dev = dev,
 933                 .pending = 1,
 934         };
 935         int ret = 1;
 936
 937         msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
 938         if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
 939             nla_put_string(msg, IFLA_IFNAME, dev->ifname))
 940                 goto out;
 941
 942         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
 943         nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
 944         nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
 945
 946         nl_send_auto_complete(sock_rtnl, msg);
 947         while (chk.pending > 0)
 948                 nl_recvmsgs(sock_rtnl, cb);
 949
 950         nlmsg_free(msg);
 951         ret = chk.pending;
 952
 953 out:
 954         nl_cb_put(cb);
 955         return ret;
 956 }
 957
 958 struct device *
 959 system_if_get_parent(struct device *dev)
 960 {
 961         char buf[64], *devname;
 962         int ifindex, iflink, len;
 963         FILE *f;
 964
 965         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
 966         f = fopen(buf, "r");
 967         if (!f)
 968                 return NULL;
 969
 970         len = fread(buf, 1, sizeof(buf) - 1, f);
 971         fclose(f);
 972
 973         if (len <= 0)
 974                 return NULL;
 975
 976         buf[len] = 0;
 977         iflink = strtoul(buf, NULL, 0);
 978         ifindex = system_if_resolve(dev);
 979         if (!iflink || iflink == ifindex)
 980                 return NULL;
 981
 982         devname = if_indextoname(iflink, buf);
 983         if (!devname)
 984                 return NULL;
 985
 986         return device_get(devname, true);
 987 }
 988
 989 static bool
 990 read_string_file(int dir_fd, const char *file, char *buf, int len)
 991 {
 992         bool ret = false;
 993         char *c;
 994         int fd;
 995
 996         fd = openat(dir_fd, file, O_RDONLY);
 997         if (fd < 0)
 998                 return false;
 999
1000 retry:
1001         len = read(fd, buf, len - 1);
1002         if (len < 0) {
1003                 if (errno == EINTR)
1004                         goto retry;
1005         } else if (len > 0) {
1006                         buf[len] = 0;
1007
1008                         c = strchr(buf, '\n');
1009                         if (c)
1010                                 *c = 0;
1011
1012                         ret = true;
1013         }
1014
1015         close(fd);
1016
1017         return ret;
1018 }
1019
1020 static bool
1021 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1022 {
1023         char buf[64];
1024         bool ret = false;
1025
1026         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1027         if (ret)
1028                 *val = strtoull(buf, NULL, 0);
1029
1030         return ret;
1031 }
1032
1033 /* Assume advertised flags == supported flags */
1034 static const struct {
1035         uint32_t mask;
1036         const char *name;
1037 } ethtool_link_modes[] = {
1038         { ADVERTISED_10baseT_Half, "10H" },
1039         { ADVERTISED_10baseT_Full, "10F" },
1040         { ADVERTISED_100baseT_Half, "100H" },
1041         { ADVERTISED_100baseT_Full, "100F" },
1042         { ADVERTISED_1000baseT_Half, "1000H" },
1043         { ADVERTISED_1000baseT_Full, "1000F" },
1044 };
1045
1046 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1047 {
1048         int i;
1049         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1050                 if (mask & ethtool_link_modes[i].mask)
1051                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1052         }
1053 }
1054
1055 bool
1056 system_if_force_external(const char *ifname)
1057 {
1058         char buf[64];
1059         struct stat s;
1060
1061         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1062         return stat(buf, &s) == 0;
1063 }
1064
1065 int
1066 system_if_dump_info(struct device *dev, struct blob_buf *b)
1067 {
1068         struct ethtool_cmd ecmd;
1069         struct ifreq ifr;
1070         char buf[64], *s;
1071         void *c;
1072         int dir_fd;
1073
1074         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1075         dir_fd = open(buf, O_DIRECTORY);
1076
1077         memset(&ecmd, 0, sizeof(ecmd));
1078         memset(&ifr, 0, sizeof(ifr));
1079         strcpy(ifr.ifr_name, dev->ifname);
1080         ifr.ifr_data = (caddr_t) &ecmd;
1081         ecmd.cmd = ETHTOOL_GSET;
1082
1083         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1084                 c = blobmsg_open_array(b, "link-advertising");
1085                 system_add_link_modes(b, ecmd.advertising);
1086                 blobmsg_close_array(b, c);
1087
1088                 c = blobmsg_open_array(b, "link-supported");
1089                 system_add_link_modes(b, ecmd.supported);
1090                 blobmsg_close_array(b, c);
1091
1092                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1093                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1094                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1095                 blobmsg_add_string_buffer(b);
1096         }
1097
1098         close(dir_fd);
1099         return 0;
1100 }
1101
1102 int
1103 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1104 {
1105         const char *const counters[] = {
1106                 "collisions",     "rx_frame_errors",   "tx_compressed",
1107                 "multicast",      "rx_length_errors",  "tx_dropped",
1108                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
1109                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
1110                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
1111                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
1112                 "rx_errors",      "tx_bytes",          "tx_window_errors",
1113                 "rx_fifo_errors", "tx_carrier_errors",
1114         };
1115         char buf[64];
1116         int stats_dir;
1117         int i;
1118         uint64_t val = 0;
1119
1120         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1121         stats_dir = open(buf, O_DIRECTORY);
1122         if (stats_dir < 0)
1123                 return -1;
1124
1125         for (i = 0; i < ARRAY_SIZE(counters); i++)
1126                 if (read_uint64_file(stats_dir, counters[i], &val))
1127                         blobmsg_add_u64(b, counters[i], val);
1128
1129         close(stats_dir);
1130         return 0;
1131 }
1132
1133 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1134 {
1135         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1136         int alen = v4 ? 4 : 16;
1137         unsigned int flags = 0;
1138         struct ifaddrmsg ifa = {
1139                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1140                 .ifa_prefixlen = addr->mask,
1141                 .ifa_index = dev->ifindex,
1142         };
1143
1144         struct nl_msg *msg;
1145         if (cmd == RTM_NEWADDR)
1146                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1147
1148         msg = nlmsg_alloc_simple(cmd, flags);
1149         if (!msg)
1150                 return -1;
1151
1152         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1153         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1154         if (v4) {
1155                 if (addr->broadcast)
1156                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1157                 if (addr->point_to_point)
1158                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1159         } else {
1160                 time_t now = system_get_rtime();
1161                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1162
1163                 if (addr->preferred_until) {
1164                         int64_t preferred = addr->preferred_until - now;
1165                         if (preferred < 0)
1166                                 preferred = 0;
1167                         else if (preferred > UINT32_MAX)
1168                                 preferred = UINT32_MAX;
1169
1170                         cinfo.ifa_prefered = preferred;
1171                 }
1172
1173                 if (addr->valid_until) {
1174                         int64_t valid = addr->valid_until - now;
1175                         if (valid <= 0)
1176                                 return -1;
1177                         else if (valid > UINT32_MAX)
1178                                 valid = UINT32_MAX;
1179
1180                         cinfo.ifa_valid = valid;
1181                 }
1182
1183                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1184         }
1185
1186         return system_rtnl_call(msg);
1187 }
1188
1189 int system_add_address(struct device *dev, struct device_addr *addr)
1190 {
1191         return system_addr(dev, addr, RTM_NEWADDR);
1192 }
1193
1194 int system_del_address(struct device *dev, struct device_addr *addr)
1195 {
1196         return system_addr(dev, addr, RTM_DELADDR);
1197 }
1198
1199 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1200 {
1201         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1202         bool have_gw;
1203         unsigned int flags = 0;
1204
1205         if (alen == 4)
1206                 have_gw = !!route->nexthop.in.s_addr;
1207         else
1208                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1209                         route->nexthop.in6.s6_addr32[1] ||
1210                         route->nexthop.in6.s6_addr32[2] ||
1211                         route->nexthop.in6.s6_addr32[3];
1212
1213         unsigned char scope = (cmd == RTM_DELROUTE) ? RT_SCOPE_NOWHERE :
1214                         (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1215
1216         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1217                         ? route->table : RT_TABLE_MAIN;
1218
1219         struct rtmsg rtm = {
1220                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1221                 .rtm_dst_len = route->mask,
1222                 .rtm_src_len = route->sourcemask,
1223                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1224                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1225                 .rtm_scope = scope,
1226                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1227         };
1228         struct nl_msg *msg;
1229
1230         if (cmd == RTM_NEWROUTE) {
1231                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1232
1233                 if (!dev) { // Add null-route
1234                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1235                         rtm.rtm_type = RTN_UNREACHABLE;
1236                 }
1237         }
1238
1239         msg = nlmsg_alloc_simple(cmd, flags);
1240         if (!msg)
1241                 return -1;
1242
1243         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1244
1245         if (route->mask)
1246                 nla_put(msg, RTA_DST, alen, &route->addr);
1247
1248         if (route->sourcemask)
1249                 nla_put(msg, RTA_SRC, alen, &route->source);
1250
1251         if (route->metric > 0)
1252                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1253
1254         if (have_gw)
1255                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1256
1257         if (dev)
1258                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1259
1260         if (table >= 256)
1261                 nla_put_u32(msg, RTA_TABLE, table);
1262
1263         return system_rtnl_call(msg);
1264 }
1265
1266 int system_add_route(struct device *dev, struct device_route *route)
1267 {
1268         return system_rt(dev, route, RTM_NEWROUTE);
1269 }
1270
1271 int system_del_route(struct device *dev, struct device_route *route)
1272 {
1273         return system_rt(dev, route, RTM_DELROUTE);
1274 }
1275
1276 int system_flush_routes(void)
1277 {
1278         const char *names[] = {
1279                 "/proc/sys/net/ipv4/route/flush",
1280                 "/proc/sys/net/ipv6/route/flush"
1281         };
1282         int fd, i;
1283
1284         for (i = 0; i < ARRAY_SIZE(names); i++) {
1285                 fd = open(names[i], O_WRONLY);
1286                 if (fd < 0)
1287                         continue;
1288
1289                 if (write(fd, "-1", 2)) {}
1290                 close(fd);
1291         }
1292         return 0;
1293 }
1294
1295 bool system_resolve_rt_table(const char *name, unsigned int *id)
1296 {
1297         FILE *f;
1298         char *e, buf[128];
1299         unsigned int n, table = RT_TABLE_UNSPEC;
1300
1301         /* first try to parse table as number */
1302         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1303                 table = n;
1304
1305         /* handle well known aliases */
1306         else if (!strcmp(name, "default"))
1307                 table = RT_TABLE_DEFAULT;
1308         else if (!strcmp(name, "main"))
1309                 table = RT_TABLE_MAIN;
1310         else if (!strcmp(name, "local"))
1311                 table = RT_TABLE_LOCAL;
1312
1313         /* try to look up name in /etc/iproute2/rt_tables */
1314         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1315         {
1316                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1317                 {
1318                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1319                                 continue;
1320
1321                         n = strtoul(e, NULL, 10);
1322                         e = strtok(NULL, " \t\n");
1323
1324                         if (e && !strcmp(e, name))
1325                         {
1326                                 table = n;
1327                                 break;
1328                         }
1329                 }
1330
1331                 fclose(f);
1332         }
1333
1334         if (table == RT_TABLE_UNSPEC)
1335                 return false;
1336
1337         /* do not consider main table special */
1338         if (table == RT_TABLE_MAIN)
1339                 table = RT_TABLE_UNSPEC;
1340
1341         *id = table;
1342         return true;
1343 }
1344
1345 static int system_iprule(struct iprule *rule, int cmd)
1346 {
1347         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1348
1349         struct nl_msg *msg;
1350         struct rtmsg rtm = {
1351                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1352                 .rtm_protocol = RTPROT_STATIC,
1353                 .rtm_scope = RT_SCOPE_UNIVERSE,
1354                 .rtm_table = RT_TABLE_UNSPEC,
1355                 .rtm_type = RTN_UNSPEC,
1356                 .rtm_flags = 0,
1357         };
1358
1359         if (cmd == RTM_NEWRULE) {
1360                 rtm.rtm_type = RTN_UNICAST;
1361                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1362         }
1363
1364         if (rule->invert)
1365                 rtm.rtm_flags |= FIB_RULE_INVERT;
1366
1367         if (rule->flags & IPRULE_SRC)
1368                 rtm.rtm_src_len = rule->src_mask;
1369
1370         if (rule->flags & IPRULE_DEST)
1371                 rtm.rtm_dst_len = rule->dest_mask;
1372
1373         if (rule->flags & IPRULE_TOS)
1374                 rtm.rtm_tos = rule->tos;
1375
1376         if (rule->flags & IPRULE_LOOKUP) {
1377                 if (rule->lookup < 256)
1378                         rtm.rtm_table = rule->lookup;
1379         }
1380
1381         if (rule->flags & IPRULE_ACTION)
1382                 rtm.rtm_type = rule->action;
1383         else if (rule->flags & IPRULE_GOTO)
1384                 rtm.rtm_type = FR_ACT_GOTO;
1385         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1386                 rtm.rtm_type = FR_ACT_NOP;
1387
1388         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1389
1390         if (!msg)
1391                 return -1;
1392
1393         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1394
1395         if (rule->flags & IPRULE_IN)
1396                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1397
1398         if (rule->flags & IPRULE_OUT)
1399                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1400
1401         if (rule->flags & IPRULE_SRC)
1402                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1403
1404         if (rule->flags & IPRULE_DEST)
1405                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1406
1407         if (rule->flags & IPRULE_PRIORITY)
1408                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1409         else if (cmd == RTM_NEWRULE)
1410                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1411
1412         if (rule->flags & IPRULE_FWMARK)
1413                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1414
1415         if (rule->flags & IPRULE_FWMASK)
1416                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1417
1418         if (rule->flags & IPRULE_LOOKUP) {
1419                 if (rule->lookup >= 256)
1420                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1421         }
1422
1423         if (rule->flags & IPRULE_GOTO)
1424                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1425
1426         return system_rtnl_call(msg);
1427 }
1428
1429 int system_add_iprule(struct iprule *rule)
1430 {
1431         return system_iprule(rule, RTM_NEWRULE);
1432 }
1433
1434 int system_del_iprule(struct iprule *rule)
1435 {
1436         return system_iprule(rule, RTM_DELRULE);
1437 }
1438
1439 int system_flush_iprules(void)
1440 {
1441         int rv = 0;
1442         struct iprule rule;
1443
1444         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1445         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1446
1447         memset(&rule, 0, sizeof(rule));
1448
1449
1450         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1451
1452         rule.priority = 0;
1453         rule.lookup = RT_TABLE_LOCAL;
1454         rv |= system_iprule(&rule, RTM_NEWRULE);
1455
1456         rule.priority = 32766;
1457         rule.lookup = RT_TABLE_MAIN;
1458         rv |= system_iprule(&rule, RTM_NEWRULE);
1459
1460         rule.priority = 32767;
1461         rule.lookup = RT_TABLE_DEFAULT;
1462         rv |= system_iprule(&rule, RTM_NEWRULE);
1463
1464
1465         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1466
1467         rule.priority = 0;
1468         rule.lookup = RT_TABLE_LOCAL;
1469         rv |= system_iprule(&rule, RTM_NEWRULE);
1470
1471         rule.priority = 32766;
1472         rule.lookup = RT_TABLE_MAIN;
1473         rv |= system_iprule(&rule, RTM_NEWRULE);
1474
1475         return rv;
1476 }
1477
1478 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1479 {
1480         char *e;
1481         unsigned int n;
1482
1483         if (!strcmp(action, "local"))
1484                 n = RTN_LOCAL;
1485         else if (!strcmp(action, "nat"))
1486                 n = RTN_NAT;
1487         else if (!strcmp(action, "broadcast"))
1488                 n = RTN_BROADCAST;
1489         else if (!strcmp(action, "anycast"))
1490                 n = RTN_ANYCAST;
1491         else if (!strcmp(action, "multicast"))
1492                 n = RTN_MULTICAST;
1493         else if (!strcmp(action, "prohibit"))
1494                 n = RTN_PROHIBIT;
1495         else if (!strcmp(action, "unreachable"))
1496                 n = RTN_UNREACHABLE;
1497         else if (!strcmp(action, "blackhole"))
1498                 n = RTN_BLACKHOLE;
1499         else if (!strcmp(action, "xresolve"))
1500                 n = RTN_XRESOLVE;
1501         else if (!strcmp(action, "unicast"))
1502                 n = RTN_UNICAST;
1503         else if (!strcmp(action, "throw"))
1504                 n = RTN_THROW;
1505         else if (!strcmp(action, "failed_policy"))
1506                 n = RTN_FAILED_POLICY;
1507         else {
1508                 n = strtoul(action, &e, 0);
1509                 if (!e || *e || e == action || n > 255)
1510                         return false;
1511         }
1512
1513         *id = n;
1514         return true;
1515 }
1516
1517 time_t system_get_rtime(void)
1518 {
1519         struct timespec ts;
1520         struct timeval tv;
1521
1522         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1523                 return ts.tv_sec;
1524
1525         if (gettimeofday(&tv, NULL) == 0)
1526                 return tv.tv_sec;
1527
1528         return 0;
1529 }
1530
1531 #ifndef IP_DF
1532 #define IP_DF       0x4000
1533 #endif
1534
1535 static int tunnel_ioctl(const char *name, int cmd, void *p)
1536 {
1537         struct ifreq ifr;
1538
1539         memset(&ifr, 0, sizeof(ifr));
1540         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1541         ifr.ifr_ifru.ifru_data = p;
1542         return ioctl(sock_ioctl, cmd, &ifr);
1543 }
1544
1545 int system_del_ip_tunnel(const char *name)
1546 {
1547         return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1548 }
1549
1550 int system_update_ipv6_mtu(struct device *dev, int mtu)
1551 {
1552         int ret = -1;
1553         char buf[64];
1554         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1555                         dev->ifname);
1556
1557         int fd = open(buf, O_RDWR);
1558         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1559         if (len < 0)
1560                 goto out;
1561
1562         buf[len] = 0;
1563         ret = atoi(buf);
1564
1565         if (!mtu || ret <= mtu)
1566                 goto out;
1567
1568         lseek(fd, 0, SEEK_SET);
1569         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1570                 ret = -1;
1571
1572 out:
1573         close(fd);
1574         return ret;
1575 }
1576
1577 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1578 {
1579         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1580         struct blob_attr *cur;
1581         bool set_df = true;
1582         const char *str;
1583
1584         system_del_ip_tunnel(name);
1585
1586         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1587                 blob_data(attr), blob_len(attr));
1588
1589         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1590                 return -EINVAL;
1591         str = blobmsg_data(cur);
1592
1593         if ((cur = tb[TUNNEL_ATTR_DF]))
1594                 set_df = blobmsg_get_bool(cur);
1595
1596         unsigned int ttl = 0;
1597         if ((cur = tb[TUNNEL_ATTR_TTL])) {
1598                 ttl = blobmsg_get_u32(cur);
1599                 if (ttl > 255 || (!set_df && ttl))
1600                         return -EINVAL;
1601         }
1602
1603         unsigned int link = 0;
1604         if ((cur = tb[TUNNEL_ATTR_LINK])) {
1605                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
1606                 if (!iface)
1607                         return -EINVAL;
1608
1609                 if (iface->l3_dev.dev)
1610                         link = iface->l3_dev.dev->ifindex;
1611         }
1612
1613         if (!strcmp(str, "sit")) {
1614                 struct ip_tunnel_parm p = {
1615                         .link = link,
1616                         .iph = {
1617                                 .version = 4,
1618                                 .ihl = 5,
1619                                 .frag_off = set_df ? htons(IP_DF) : 0,
1620                                 .protocol = IPPROTO_IPV6,
1621                                 .ttl = ttl
1622                         }
1623                 };
1624
1625                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1626                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1627                         return -EINVAL;
1628
1629                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1630                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1631                         return -EINVAL;
1632
1633                 strncpy(p.name, name, sizeof(p.name));
1634                 if (tunnel_ioctl("sit0", SIOCADDTUNNEL, &p) < 0)
1635                         return -1;
1636
1637 #ifdef SIOCADD6RD
1638                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
1639                         unsigned int mask;
1640                         struct ip_tunnel_6rd p6;
1641
1642                         memset(&p6, 0, sizeof(p6));
1643
1644                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
1645                                                 &p6.prefix, &mask) || mask > 128)
1646                                 return -EINVAL;
1647                         p6.prefixlen = mask;
1648
1649                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
1650                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
1651                                                         &p6.relay_prefix, &mask) || mask > 32)
1652                                         return -EINVAL;
1653                                 p6.relay_prefixlen = mask;
1654                         }
1655
1656                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
1657                                 system_del_ip_tunnel(name);
1658                                 return -1;
1659                         }
1660                 }
1661 #endif
1662         } else if (!strcmp(str, "ipip6")) {
1663                 struct ip6_tnl_parm p = {
1664                         .link = link,
1665                         .proto = IPPROTO_IPIP,
1666                         .hop_limit = (ttl) ? ttl : 64,
1667                         .encap_limit = 4,
1668                 };
1669
1670                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1671                                 inet_pton(AF_INET6, blobmsg_data(cur), &p.laddr) < 1)
1672                         return -EINVAL;
1673
1674                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1675                                 inet_pton(AF_INET6, blobmsg_data(cur), &p.raddr) < 1)
1676                         return -EINVAL;
1677
1678                 strncpy(p.name, name, sizeof(p.name));
1679                 if (tunnel_ioctl("ip6tnl0", SIOCADDTUNNEL, &p) < 0)
1680                         return -1;
1681         } else
1682                 return -EINVAL;
1683
1684         return 0;
1685 }