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 <unistd.h>
  45 #include <string.h>
  46 #include <fcntl.h>
  47 #include <glob.h>
  48 #include <time.h>
  49
  50 #include <netlink/msg.h>
  51 #include <netlink/attr.h>
  52 #include <netlink/socket.h>
  53 #include <libubox/uloop.h>
  54
  55 #include "netifd.h"
  56 #include "device.h"
  57 #include "system.h"
  58
  59 struct event_socket {
  60         struct uloop_fd uloop;
  61         struct nl_sock *sock;
  62         struct nl_cb *cb;
  63 };
  64
  65 static int sock_ioctl = -1;
  66 static struct nl_sock *sock_rtnl = NULL;
  67
  68 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
  69 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
  70
  71 static char dev_buf[256];
  72
  73 static void
  74 handler_nl_event(struct uloop_fd *u, unsigned int events)
  75 {
  76         struct event_socket *ev = container_of(u, struct event_socket, uloop);
  77         nl_recvmsgs(ev->sock, ev->cb);
  78 }
  79
  80 static struct nl_sock *
  81 create_socket(int protocol, int groups)
  82 {
  83         struct nl_sock *sock;
  84
  85         sock = nl_socket_alloc();
  86         if (!sock)
  87                 return NULL;
  88
  89         if (groups)
  90                 nl_join_groups(sock, groups);
  91
  92         if (nl_connect(sock, protocol))
  93                 return NULL;
  94
  95         return sock;
  96 }
  97
  98 static bool
  99 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
 100                         uloop_fd_handler cb)
 101 {
 102         ev->sock = create_socket(protocol, groups);
 103         if (!ev->sock)
 104                 return false;
 105
 106         ev->uloop.fd = nl_socket_get_fd(ev->sock);
 107         ev->uloop.cb = cb;
 108         uloop_fd_add(&ev->uloop, ULOOP_READ | ULOOP_EDGE_TRIGGER);
 109         return true;
 110 }
 111
 112 static bool
 113 create_event_socket(struct event_socket *ev, int protocol,
 114                     int (*cb)(struct nl_msg *msg, void *arg))
 115 {
 116         // Prepare socket for link events
 117         ev->cb = nl_cb_alloc(NL_CB_DEFAULT);
 118         if (!ev->cb)
 119                 return false;
 120
 121         nl_cb_set(ev->cb, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
 122
 123         return create_raw_event_socket(ev, protocol, 0, handler_nl_event);
 124 }
 125
 126 int system_init(void)
 127 {
 128         static struct event_socket rtnl_event;
 129         static struct event_socket hotplug_event;
 130
 131         sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
 132         fcntl(sock_ioctl, F_SETFD, fcntl(sock_ioctl, F_GETFD) | FD_CLOEXEC);
 133
 134         // Prepare socket for routing / address control
 135         sock_rtnl = create_socket(NETLINK_ROUTE, 0);
 136         if (!sock_rtnl)
 137                 return -1;
 138
 139         if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
 140                 return -1;
 141
 142         if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
 143                                      handle_hotplug_event))
 144                 return -1;
 145
 146         // Receive network link events form kernel
 147         nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
 148
 149         return 0;
 150 }
 151
 152 static void system_set_sysctl(const char *path, const char *val)
 153 {
 154         int fd;
 155
 156         fd = open(path, O_WRONLY);
 157         if (fd < 0)
 158                 return;
 159
 160         write(fd, val, strlen(val));
 161         close(fd);
 162 }
 163
 164 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
 165 {
 166         snprintf(dev_buf, sizeof(dev_buf), path, device);
 167         system_set_sysctl(dev_buf, val);
 168 }
 169
 170 static void system_set_disable_ipv6(struct device *dev, const char *val)
 171 {
 172         system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
 173 }
 174
 175 // Evaluate netlink messages
 176 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 177 {
 178         struct nlmsghdr *nh = nlmsg_hdr(msg);
 179         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 180         struct nlattr *nla[__IFLA_MAX];
 181
 182         if (nh->nlmsg_type != RTM_DELLINK && nh->nlmsg_type != RTM_NEWLINK)
 183                 goto out;
 184
 185         nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
 186         if (!nla[IFLA_IFNAME])
 187                 goto out;
 188
 189         struct device *dev = device_get(RTA_DATA(nla[IFLA_IFNAME]), false);
 190         if (!dev)
 191                 goto out;
 192
 193         dev->ifindex = ifi->ifi_index;
 194         /* TODO: parse link status */
 195
 196 out:
 197         return 0;
 198 }
 199
 200 static void
 201 handle_hotplug_msg(char *data, int size)
 202 {
 203         const char *subsystem = NULL, *interface = NULL;
 204         char *cur, *end, *sep;
 205         struct device *dev;
 206         int skip;
 207         bool add;
 208
 209         if (!strncmp(data, "add@", 4))
 210                 add = true;
 211         else if (!strncmp(data, "remove@", 7))
 212                 add = false;
 213         else
 214                 return;
 215
 216         skip = strlen(data) + 1;
 217         end = data + size;
 218
 219         for (cur = data + skip; cur < end; cur += skip) {
 220                 skip = strlen(cur) + 1;
 221
 222                 sep = strchr(cur, '=');
 223                 if (!sep)
 224                         continue;
 225
 226                 *sep = 0;
 227                 if (!strcmp(cur, "INTERFACE"))
 228                         interface = sep + 1;
 229                 else if (!strcmp(cur, "SUBSYSTEM")) {
 230                         subsystem = sep + 1;
 231                         if (strcmp(subsystem, "net") != 0)
 232                                 return;
 233                 }
 234                 if (subsystem && interface)
 235                         goto found;
 236         }
 237         return;
 238
 239 found:
 240         dev = device_get(interface, false);
 241         if (!dev)
 242                 return;
 243
 244         if (dev->type != &simple_device_type)
 245                 return;
 246
 247         if (add && system_if_force_external(dev->ifname))
 248                 return;
 249
 250         device_set_present(dev, add);
 251 }
 252
 253 static void
 254 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
 255 {
 256         struct event_socket *ev = container_of(u, struct event_socket, uloop);
 257         struct sockaddr_nl nla;
 258         unsigned char *buf = NULL;
 259         int size;
 260
 261         while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
 262                 if (nla.nl_pid == 0)
 263                         handle_hotplug_msg((char *) buf, size);
 264
 265                 free(buf);
 266         }
 267 }
 268
 269 static int system_rtnl_call(struct nl_msg *msg)
 270 {
 271         int ret;
 272
 273         ret = nl_send_auto_complete(sock_rtnl, msg);
 274         nlmsg_free(msg);
 275
 276         if (ret < 0)
 277                 return ret;
 278
 279         return nl_wait_for_ack(sock_rtnl);
 280 }
 281
 282 int system_bridge_delbr(struct device *bridge)
 283 {
 284         return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 285 }
 286
 287 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 288 {
 289         struct ifreq ifr;
 290
 291         memset(&ifr, 0, sizeof(ifr));
 292         if (dev)
 293                 ifr.ifr_ifindex = dev->ifindex;
 294         else
 295                 ifr.ifr_data = data;
 296         strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
 297         return ioctl(sock_ioctl, cmd, &ifr);
 298 }
 299
 300 static bool system_is_bridge(const char *name, char *buf, int buflen)
 301 {
 302         struct stat st;
 303
 304         snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
 305         if (stat(buf, &st) < 0)
 306                 return false;
 307
 308         return true;
 309 }
 310
 311 static char *system_get_bridge(const char *name, char *buf, int buflen)
 312 {
 313         char *path;
 314         ssize_t len;
 315         glob_t gl;
 316
 317         snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
 318         if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
 319                 return NULL;
 320
 321         if (gl.gl_pathc == 0)
 322                 return NULL;
 323
 324         len = readlink(gl.gl_pathv[0], buf, buflen);
 325         if (len < 0)
 326                 return NULL;
 327
 328         buf[len] = 0;
 329         path = strrchr(buf, '/');
 330         if (!path)
 331                 return NULL;
 332
 333         return path + 1;
 334 }
 335
 336 int system_bridge_addif(struct device *bridge, struct device *dev)
 337 {
 338         char *oldbr;
 339
 340         system_set_disable_ipv6(dev, "1");
 341         oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
 342         if (oldbr && !strcmp(oldbr, bridge->ifname))
 343                 return 0;
 344
 345         return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 346 }
 347
 348 int system_bridge_delif(struct device *bridge, struct device *dev)
 349 {
 350         system_set_disable_ipv6(dev, "0");
 351         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 352 }
 353
 354 static int system_if_resolve(struct device *dev)
 355 {
 356         struct ifreq ifr;
 357         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 358         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 359                 return ifr.ifr_ifindex;
 360         else
 361                 return 0;
 362 }
 363
 364 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 365 {
 366         struct ifreq ifr;
 367
 368         memset(&ifr, 0, sizeof(ifr));
 369         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 370         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 371         ifr.ifr_flags |= add;
 372         ifr.ifr_flags &= ~rem;
 373         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 374 }
 375
 376 struct clear_data {
 377         struct nl_msg *msg;
 378         struct device *dev;
 379         int type;
 380         int size;
 381         int af;
 382 };
 383
 384
 385 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 386 {
 387         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 388
 389         return ifa->ifa_index == ifindex;
 390 }
 391
 392 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 393 {
 394         struct rtmsg *r = NLMSG_DATA(hdr);
 395         struct nlattr *tb[__RTA_MAX];
 396
 397         if (r->rtm_protocol == RTPROT_KERNEL &&
 398             r->rtm_family == AF_INET6)
 399                 return false;
 400
 401         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 402         if (!tb[RTA_OIF])
 403                 return false;
 404
 405         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 406 }
 407
 408 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
 409 {
 410         return true;
 411 }
 412
 413 static int cb_clear_event(struct nl_msg *msg, void *arg)
 414 {
 415         struct clear_data *clr = arg;
 416         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 417         bool (*cb)(struct nlmsghdr *, int ifindex);
 418         int type;
 419
 420         switch(clr->type) {
 421         case RTM_GETADDR:
 422                 type = RTM_DELADDR;
 423                 if (hdr->nlmsg_type != RTM_NEWADDR)
 424                         return NL_SKIP;
 425
 426                 cb = check_ifaddr;
 427                 break;
 428         case RTM_GETROUTE:
 429                 type = RTM_DELROUTE;
 430                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 431                         return NL_SKIP;
 432
 433                 cb = check_route;
 434                 break;
 435         case RTM_GETRULE:
 436                 type = RTM_DELRULE;
 437                 if (hdr->nlmsg_type != RTM_NEWRULE)
 438                         return NL_SKIP;
 439
 440                 cb = check_rule;
 441                 break;
 442         default:
 443                 return NL_SKIP;
 444         }
 445
 446         if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
 447                 return NL_SKIP;
 448
 449         if (type == RTM_DELRULE)
 450                 D(SYSTEM, "Remove a rule\n");
 451         else
 452                 D(SYSTEM, "Remove %s from device %s\n",
 453                   type == RTM_DELADDR ? "an address" : "a route",
 454                   clr->dev->ifname);
 455         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 456         hdr = nlmsg_hdr(clr->msg);
 457         hdr->nlmsg_type = type;
 458         hdr->nlmsg_flags = NLM_F_REQUEST;
 459
 460         if (!nl_send_auto_complete(sock_rtnl, clr->msg))
 461                 nl_wait_for_ack(sock_rtnl);
 462
 463         return NL_SKIP;
 464 }
 465
 466 static int
 467 cb_finish_event(struct nl_msg *msg, void *arg)
 468 {
 469         int *pending = arg;
 470         *pending = 0;
 471         return NL_STOP;
 472 }
 473
 474 static int
 475 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 476 {
 477         int *pending = arg;
 478         *pending = err->error;
 479         return NL_STOP;
 480 }
 481
 482 static void
 483 system_if_clear_entries(struct device *dev, int type, int af)
 484 {
 485         struct clear_data clr;
 486         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 487         struct rtmsg rtm = {
 488                 .rtm_family = af,
 489                 .rtm_flags = RTM_F_CLONED,
 490         };
 491         int flags = NLM_F_DUMP;
 492         int pending = 1;
 493
 494         clr.af = af;
 495         clr.dev = dev;
 496         clr.type = type;
 497         switch (type) {
 498         case RTM_GETADDR:
 499         case RTM_GETRULE:
 500                 clr.size = sizeof(struct rtgenmsg);
 501                 break;
 502         case RTM_GETROUTE:
 503                 clr.size = sizeof(struct rtmsg);
 504                 break;
 505         default:
 506                 return;
 507         }
 508
 509         if (!cb)
 510                 return;
 511
 512         clr.msg = nlmsg_alloc_simple(type, flags);
 513         if (!clr.msg)
 514                 goto out;
 515
 516         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 517         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 518         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 519         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 520
 521         nl_send_auto_complete(sock_rtnl, clr.msg);
 522         while (pending > 0)
 523                 nl_recvmsgs(sock_rtnl, cb);
 524
 525         nlmsg_free(clr.msg);
 526 out:
 527         nl_cb_put(cb);
 528 }
 529
 530 /*
 531  * Clear bridge (membership) state and bring down device
 532  */
 533 void system_if_clear_state(struct device *dev)
 534 {
 535         static char buf[256];
 536         char *bridge;
 537
 538         if (dev->external)
 539                 return;
 540
 541         dev->ifindex = system_if_resolve(dev);
 542         if (!dev->ifindex)
 543                 return;
 544
 545         system_if_flags(dev->ifname, 0, IFF_UP);
 546
 547         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 548                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 549                 system_bridge_delbr(dev);
 550                 return;
 551         }
 552
 553         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 554         if (bridge) {
 555                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 556                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 557         }
 558
 559         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 560         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 561         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 562         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 563         system_set_disable_ipv6(dev, "0");
 564 }
 565
 566 static inline unsigned long
 567 sec_to_jiffies(int val)
 568 {
 569         return (unsigned long) val * 100;
 570 }
 571
 572 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 573 {
 574         unsigned long args[4] = {};
 575
 576         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 577                 return -1;
 578
 579         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 580         args[1] = !!cfg->stp;
 581         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 582
 583         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 584         args[1] = sec_to_jiffies(cfg->forward_delay);
 585         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 586
 587         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 588                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 589
 590         args[0] = BRCTL_SET_BRIDGE_PRIORITY;
 591         args[1] = cfg->priority;
 592         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 593
 594         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 595                 args[0] = BRCTL_SET_AGEING_TIME;
 596                 args[1] = sec_to_jiffies(cfg->ageing_time);
 597                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 598         }
 599
 600         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 601                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 602                 args[1] = sec_to_jiffies(cfg->hello_time);
 603                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 604         }
 605
 606         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 607                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 608                 args[1] = sec_to_jiffies(cfg->max_age);
 609                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 610         }
 611
 612         return 0;
 613 }
 614
 615 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
 616 {
 617         struct nl_msg *msg;
 618         struct nlattr *linkinfo, *data;
 619         struct ifinfomsg iim = { .ifi_family = AF_INET };
 620         int ifindex = system_if_resolve(dev);
 621         int i, rv;
 622         static const struct {
 623                 const char *name;
 624                 enum macvlan_mode val;
 625         } modes[] = {
 626                 { "private", MACVLAN_MODE_PRIVATE },
 627                 { "vepa", MACVLAN_MODE_VEPA },
 628                 { "bridge", MACVLAN_MODE_BRIDGE },
 629                 { "passthru", MACVLAN_MODE_PASSTHRU },
 630         };
 631
 632         if (ifindex == 0)
 633                 return -ENOENT;
 634
 635         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 636
 637         if (!msg)
 638                 return -1;
 639
 640         nlmsg_append(msg, &iim, sizeof(iim), 0);
 641
 642         if (cfg->flags & MACVLAN_OPT_MACADDR)
 643                 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
 644         nla_put(msg, IFLA_IFNAME, IFNAMSIZ, macvlan->ifname);
 645         nla_put_u32(msg, IFLA_LINK, ifindex);
 646
 647         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 648                 goto nla_put_failure;
 649
 650         nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
 651
 652         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 653                 goto nla_put_failure;
 654
 655         if (cfg->mode) {
 656                 for (i = 0; i < ARRAY_SIZE(modes); i++) {
 657                         if (strcmp(cfg->mode, modes[i].name) != 0)
 658                                 continue;
 659
 660                         nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
 661                         break;
 662                 }
 663         }
 664
 665         nla_nest_end(msg, data);
 666         nla_nest_end(msg, linkinfo);
 667
 668         rv = system_rtnl_call(msg);
 669         if (rv)
 670                 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
 671
 672         return rv;
 673
 674 nla_put_failure:
 675         nlmsg_free(msg);
 676         return -ENOMEM;
 677 }
 678
 679 int system_macvlan_del(struct device *macvlan)
 680 {
 681         struct nl_msg *msg;
 682         struct ifinfomsg iim;
 683
 684         iim.ifi_family = AF_INET;
 685         iim.ifi_index  = 0;
 686
 687         msg = nlmsg_alloc_simple(RTM_DELLINK, 0);
 688
 689         if (!msg)
 690                 return -1;
 691
 692         nlmsg_append(msg, &iim, sizeof(iim), 0);
 693
 694         nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
 695         nla_put(msg, IFLA_IFNAME, sizeof(macvlan->ifname), macvlan->ifname);
 696
 697         system_rtnl_call(msg);
 698
 699         return 0;
 700 }
 701
 702 static int system_vlan(struct device *dev, int id)
 703 {
 704         struct vlan_ioctl_args ifr = {
 705                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 706                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 707         };
 708
 709         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 710
 711         if (id < 0) {
 712                 ifr.cmd = DEL_VLAN_CMD;
 713                 ifr.u.VID = 0;
 714         } else {
 715                 ifr.cmd = ADD_VLAN_CMD;
 716                 ifr.u.VID = id;
 717         }
 718         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 719         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 720 }
 721
 722 int system_vlan_add(struct device *dev, int id)
 723 {
 724         return system_vlan(dev, id);
 725 }
 726
 727 int system_vlan_del(struct device *dev)
 728 {
 729         return system_vlan(dev, -1);
 730 }
 731
 732 static void
 733 system_if_get_settings(struct device *dev, struct device_settings *s)
 734 {
 735         struct ifreq ifr;
 736
 737         memset(&ifr, 0, sizeof(ifr));
 738         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 739
 740         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 741                 s->mtu = ifr.ifr_mtu;
 742                 s->flags |= DEV_OPT_MTU;
 743         }
 744
 745         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 746                 s->txqueuelen = ifr.ifr_qlen;
 747                 s->flags |= DEV_OPT_TXQUEUELEN;
 748         }
 749
 750         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 751                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 752                 s->flags |= DEV_OPT_MACADDR;
 753         }
 754 }
 755
 756 void
 757 system_if_apply_settings(struct device *dev, struct device_settings *s)
 758 {
 759         struct ifreq ifr;
 760
 761         memset(&ifr, 0, sizeof(ifr));
 762         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 763         if (s->flags & DEV_OPT_MTU) {
 764                 ifr.ifr_mtu = s->mtu;
 765                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 766                         s->flags &= ~DEV_OPT_MTU;
 767         }
 768         if (s->flags & DEV_OPT_TXQUEUELEN) {
 769                 ifr.ifr_qlen = s->txqueuelen;
 770                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 771                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 772         }
 773         if ((s->flags & DEV_OPT_MACADDR) && !dev->external) {
 774                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 775                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 776                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 777                         s->flags &= ~DEV_OPT_MACADDR;
 778         }
 779 }
 780
 781 int system_if_up(struct device *dev)
 782 {
 783         system_if_get_settings(dev, &dev->orig_settings);
 784         system_if_apply_settings(dev, &dev->settings);
 785         dev->ifindex = system_if_resolve(dev);
 786         return system_if_flags(dev->ifname, IFF_UP, 0);
 787 }
 788
 789 int system_if_down(struct device *dev)
 790 {
 791         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
 792         dev->orig_settings.flags &= dev->settings.flags;
 793         system_if_apply_settings(dev, &dev->orig_settings);
 794         return ret;
 795 }
 796
 797 int system_if_check(struct device *dev)
 798 {
 799         device_set_present(dev, (system_if_resolve(dev) > 0));
 800         return 0;
 801 }
 802
 803 struct device *
 804 system_if_get_parent(struct device *dev)
 805 {
 806         char buf[64], *devname;
 807         int ifindex, iflink, len;
 808         FILE *f;
 809
 810         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
 811         f = fopen(buf, "r");
 812         if (!f)
 813                 return NULL;
 814
 815         len = fread(buf, 1, sizeof(buf) - 1, f);
 816         fclose(f);
 817
 818         if (len <= 0)
 819                 return NULL;
 820
 821         buf[len] = 0;
 822         iflink = strtoul(buf, NULL, 0);
 823         ifindex = system_if_resolve(dev);
 824         if (!iflink || iflink == ifindex)
 825                 return NULL;
 826
 827         devname = if_indextoname(iflink, buf);
 828         if (!devname)
 829                 return NULL;
 830
 831         return device_get(devname, true);
 832 }
 833
 834 static bool
 835 read_string_file(int dir_fd, const char *file, char *buf, int len)
 836 {
 837         bool ret = false;
 838         char *c;
 839         int fd;
 840
 841         fd = openat(dir_fd, file, O_RDONLY);
 842         if (fd < 0)
 843                 return false;
 844
 845 retry:
 846         len = read(fd, buf, len - 1);
 847         if (len < 0) {
 848                 if (errno == EINTR)
 849                         goto retry;
 850         } else if (len > 0) {
 851                         buf[len] = 0;
 852
 853                         c = strchr(buf, '\n');
 854                         if (c)
 855                                 *c = 0;
 856
 857                         ret = true;
 858         }
 859
 860         close(fd);
 861
 862         return ret;
 863 }
 864
 865 static bool
 866 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
 867 {
 868         char buf[64];
 869         bool ret = false;
 870
 871         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
 872         if (ret)
 873                 *val = strtoull(buf, NULL, 0);
 874
 875         return ret;
 876 }
 877
 878 /* Assume advertised flags == supported flags */
 879 static const struct {
 880         uint32_t mask;
 881         const char *name;
 882 } ethtool_link_modes[] = {
 883         { ADVERTISED_10baseT_Half, "10H" },
 884         { ADVERTISED_10baseT_Full, "10F" },
 885         { ADVERTISED_100baseT_Half, "100H" },
 886         { ADVERTISED_100baseT_Full, "100F" },
 887         { ADVERTISED_1000baseT_Half, "1000H" },
 888         { ADVERTISED_1000baseT_Full, "1000F" },
 889 };
 890
 891 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
 892 {
 893         int i;
 894         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
 895                 if (mask & ethtool_link_modes[i].mask)
 896                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
 897         }
 898 }
 899
 900 bool
 901 system_if_force_external(const char *ifname)
 902 {
 903         char buf[64];
 904         struct stat s;
 905
 906         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
 907         return stat(buf, &s) == 0;
 908 }
 909
 910 int
 911 system_if_dump_info(struct device *dev, struct blob_buf *b)
 912 {
 913         struct ethtool_cmd ecmd;
 914         struct ifreq ifr;
 915         char buf[64], *s;
 916         void *c;
 917         int dir_fd;
 918         uint64_t val = 0;
 919
 920         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
 921         dir_fd = open(buf, O_DIRECTORY);
 922
 923         if (read_uint64_file(dir_fd, "carrier", &val))
 924                 blobmsg_add_u8(b, "link", !!val);
 925
 926         memset(&ecmd, 0, sizeof(ecmd));
 927         memset(&ifr, 0, sizeof(ifr));
 928         strcpy(ifr.ifr_name, dev->ifname);
 929         ifr.ifr_data = (caddr_t) &ecmd;
 930         ecmd.cmd = ETHTOOL_GSET;
 931
 932         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
 933                 c = blobmsg_open_array(b, "link-advertising");
 934                 system_add_link_modes(b, ecmd.advertising);
 935                 blobmsg_close_array(b, c);
 936
 937                 c = blobmsg_open_array(b, "link-supported");
 938                 system_add_link_modes(b, ecmd.supported);
 939                 blobmsg_close_array(b, c);
 940
 941                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
 942                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
 943                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
 944                 blobmsg_add_string_buffer(b);
 945         }
 946
 947         close(dir_fd);
 948         return 0;
 949 }
 950
 951 int
 952 system_if_dump_stats(struct device *dev, struct blob_buf *b)
 953 {
 954         const char *const counters[] = {
 955                 "collisions",     "rx_frame_errors",   "tx_compressed",
 956                 "multicast",      "rx_length_errors",  "tx_dropped",
 957                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
 958                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
 959                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
 960                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
 961                 "rx_errors",      "tx_bytes",          "tx_window_errors",
 962                 "rx_fifo_errors", "tx_carrier_errors",
 963         };
 964         char buf[64];
 965         int stats_dir;
 966         int i;
 967         uint64_t val = 0;
 968
 969         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
 970         stats_dir = open(buf, O_DIRECTORY);
 971         if (stats_dir < 0)
 972                 return -1;
 973
 974         for (i = 0; i < ARRAY_SIZE(counters); i++)
 975                 if (read_uint64_file(stats_dir, counters[i], &val))
 976                         blobmsg_add_u64(b, counters[i], val);
 977
 978         close(stats_dir);
 979         return 0;
 980 }
 981
 982 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
 983 {
 984         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
 985         int alen = v4 ? 4 : 16;
 986         unsigned int flags = 0;
 987         struct ifaddrmsg ifa = {
 988                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
 989                 .ifa_prefixlen = addr->mask,
 990                 .ifa_index = dev->ifindex,
 991         };
 992
 993         struct nl_msg *msg;
 994         if (cmd == RTM_NEWADDR)
 995                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
 996
 997         msg = nlmsg_alloc_simple(cmd, flags);
 998         if (!msg)
 999                 return -1;
1000
1001         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1002         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1003         if (v4) {
1004                 if (addr->broadcast)
1005                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1006                 if (addr->point_to_point)
1007                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1008         } else {
1009                 time_t now = system_get_rtime();
1010                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1011
1012                 if (addr->preferred_until) {
1013                         int64_t preferred = addr->preferred_until - now;
1014                         if (preferred < 0)
1015                                 preferred = 0;
1016                         else if (preferred > UINT32_MAX)
1017                                 preferred = UINT32_MAX;
1018
1019                         cinfo.ifa_prefered = preferred;
1020                 }
1021
1022                 if (addr->valid_until) {
1023                         int64_t valid = addr->valid_until - now;
1024                         if (valid <= 0)
1025                                 return -1;
1026                         else if (valid > UINT32_MAX)
1027                                 valid = UINT32_MAX;
1028
1029                         cinfo.ifa_valid = valid;
1030                 }
1031
1032                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1033         }
1034
1035         return system_rtnl_call(msg);
1036 }
1037
1038 int system_add_address(struct device *dev, struct device_addr *addr)
1039 {
1040         return system_addr(dev, addr, RTM_NEWADDR);
1041 }
1042
1043 int system_del_address(struct device *dev, struct device_addr *addr)
1044 {
1045         return system_addr(dev, addr, RTM_DELADDR);
1046 }
1047
1048 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1049 {
1050         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1051         bool have_gw;
1052         unsigned int flags = 0;
1053
1054         if (alen == 4)
1055                 have_gw = !!route->nexthop.in.s_addr;
1056         else
1057                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1058                         route->nexthop.in6.s6_addr32[1] ||
1059                         route->nexthop.in6.s6_addr32[2] ||
1060                         route->nexthop.in6.s6_addr32[3];
1061
1062         unsigned char scope = (cmd == RTM_DELROUTE) ? RT_SCOPE_NOWHERE :
1063                         (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1064
1065         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1066                         ? route->table : RT_TABLE_MAIN;
1067
1068         struct rtmsg rtm = {
1069                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1070                 .rtm_dst_len = route->mask,
1071                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1072                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1073                 .rtm_scope = scope,
1074                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1075         };
1076         struct nl_msg *msg;
1077
1078         if (cmd == RTM_NEWROUTE) {
1079                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1080
1081                 if (!dev) { // Add null-route
1082                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1083                         rtm.rtm_type = RTN_UNREACHABLE;
1084                 }
1085         }
1086
1087         msg = nlmsg_alloc_simple(cmd, flags);
1088         if (!msg)
1089                 return -1;
1090
1091         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1092
1093         if (route->mask)
1094                 nla_put(msg, RTA_DST, alen, &route->addr);
1095
1096         if (route->metric > 0)
1097                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1098
1099         if (have_gw)
1100                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1101
1102         if (dev)
1103                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1104
1105         if (table >= 256)
1106                 nla_put_u32(msg, RTA_TABLE, table);
1107
1108         return system_rtnl_call(msg);
1109 }
1110
1111 int system_add_route(struct device *dev, struct device_route *route)
1112 {
1113         return system_rt(dev, route, RTM_NEWROUTE);
1114 }
1115
1116 int system_del_route(struct device *dev, struct device_route *route)
1117 {
1118         return system_rt(dev, route, RTM_DELROUTE);
1119 }
1120
1121 int system_flush_routes(void)
1122 {
1123         const char *names[] = {
1124                 "/proc/sys/net/ipv4/route/flush",
1125                 "/proc/sys/net/ipv6/route/flush"
1126         };
1127         int fd, i;
1128
1129         for (i = 0; i < ARRAY_SIZE(names); i++) {
1130                 fd = open(names[i], O_WRONLY);
1131                 if (fd < 0)
1132                         continue;
1133
1134                 write(fd, "-1", 2);
1135                 close(fd);
1136         }
1137         return 0;
1138 }
1139
1140 bool system_resolve_rt_table(const char *name, unsigned int *id)
1141 {
1142         FILE *f;
1143         char *e, buf[128];
1144         unsigned int n, table = RT_TABLE_UNSPEC;
1145
1146         /* first try to parse table as number */
1147         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1148                 table = n;
1149
1150         /* handle well known aliases */
1151         else if (!strcmp(name, "default"))
1152                 table = RT_TABLE_DEFAULT;
1153         else if (!strcmp(name, "main"))
1154                 table = RT_TABLE_MAIN;
1155         else if (!strcmp(name, "local"))
1156                 table = RT_TABLE_LOCAL;
1157
1158         /* try to look up name in /etc/iproute2/rt_tables */
1159         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1160         {
1161                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1162                 {
1163                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1164                                 continue;
1165
1166                         n = strtoul(e, NULL, 10);
1167                         e = strtok(NULL, " \t\n");
1168
1169                         if (e && !strcmp(e, name))
1170                         {
1171                                 table = n;
1172                                 break;
1173                         }
1174                 }
1175
1176                 fclose(f);
1177         }
1178
1179         if (table == RT_TABLE_UNSPEC)
1180                 return false;
1181
1182         /* do not consider main table special */
1183         if (table == RT_TABLE_MAIN)
1184                 table = RT_TABLE_UNSPEC;
1185
1186         *id = table;
1187         return true;
1188 }
1189
1190 static int system_iprule(struct iprule *rule, int cmd)
1191 {
1192         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1193
1194         struct nl_msg *msg;
1195         struct rtmsg rtm = {
1196                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1197                 .rtm_protocol = RTPROT_STATIC,
1198                 .rtm_scope = RT_SCOPE_UNIVERSE,
1199                 .rtm_table = RT_TABLE_UNSPEC,
1200                 .rtm_type = RTN_UNSPEC,
1201                 .rtm_flags = 0,
1202         };
1203
1204         if (cmd == RTM_NEWRULE) {
1205                 rtm.rtm_type = RTN_UNICAST;
1206                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1207         }
1208
1209         if (rule->invert)
1210                 rtm.rtm_flags |= FIB_RULE_INVERT;
1211
1212         if (rule->flags & IPRULE_SRC)
1213                 rtm.rtm_src_len = rule->src_mask;
1214
1215         if (rule->flags & IPRULE_DEST)
1216                 rtm.rtm_dst_len = rule->dest_mask;
1217
1218         if (rule->flags & IPRULE_TOS)
1219                 rtm.rtm_tos = rule->tos;
1220
1221         if (rule->flags & IPRULE_LOOKUP) {
1222                 if (rule->lookup < 256)
1223                         rtm.rtm_table = rule->lookup;
1224         }
1225
1226         if (rule->flags & IPRULE_ACTION)
1227                 rtm.rtm_type = rule->action;
1228         else if (rule->flags & IPRULE_GOTO)
1229                 rtm.rtm_type = FR_ACT_GOTO;
1230         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1231                 rtm.rtm_type = FR_ACT_NOP;
1232
1233         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1234
1235         if (!msg)
1236                 return -1;
1237
1238         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1239
1240         if (rule->flags & IPRULE_IN)
1241                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1242
1243         if (rule->flags & IPRULE_OUT)
1244                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1245
1246         if (rule->flags & IPRULE_SRC)
1247                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1248
1249         if (rule->flags & IPRULE_DEST)
1250                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1251
1252         if (rule->flags & IPRULE_PRIORITY)
1253                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1254         else if (cmd == RTM_NEWRULE)
1255                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1256
1257         if (rule->flags & IPRULE_FWMARK)
1258                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1259
1260         if (rule->flags & IPRULE_FWMASK)
1261                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1262
1263         if (rule->flags & IPRULE_LOOKUP) {
1264                 if (rule->lookup >= 256)
1265                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1266         }
1267
1268         if (rule->flags & IPRULE_GOTO)
1269                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1270
1271         return system_rtnl_call(msg);
1272 }
1273
1274 int system_add_iprule(struct iprule *rule)
1275 {
1276         return system_iprule(rule, RTM_NEWRULE);
1277 }
1278
1279 int system_del_iprule(struct iprule *rule)
1280 {
1281         return system_iprule(rule, RTM_DELRULE);
1282 }
1283
1284 int system_flush_iprules(void)
1285 {
1286         int rv = 0;
1287         struct iprule rule;
1288
1289         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1290         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1291
1292         memset(&rule, 0, sizeof(rule));
1293
1294
1295         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1296
1297         rule.priority = 0;
1298         rule.lookup = RT_TABLE_LOCAL;
1299         rv |= system_iprule(&rule, RTM_NEWRULE);
1300
1301         rule.priority = 32766;
1302         rule.lookup = RT_TABLE_MAIN;
1303         rv |= system_iprule(&rule, RTM_NEWRULE);
1304
1305         rule.priority = 32767;
1306         rule.lookup = RT_TABLE_DEFAULT;
1307         rv |= system_iprule(&rule, RTM_NEWRULE);
1308
1309
1310         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1311
1312         rule.priority = 0;
1313         rule.lookup = RT_TABLE_LOCAL;
1314         rv |= system_iprule(&rule, RTM_NEWRULE);
1315
1316         rule.priority = 32766;
1317         rule.lookup = RT_TABLE_MAIN;
1318         rv |= system_iprule(&rule, RTM_NEWRULE);
1319
1320         return rv;
1321 }
1322
1323 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1324 {
1325         char *e;
1326         unsigned int n;
1327
1328         if (!strcmp(action, "local"))
1329                 n = RTN_LOCAL;
1330         else if (!strcmp(action, "nat"))
1331                 n = RTN_NAT;
1332         else if (!strcmp(action, "broadcast"))
1333                 n = RTN_BROADCAST;
1334         else if (!strcmp(action, "anycast"))
1335                 n = RTN_ANYCAST;
1336         else if (!strcmp(action, "multicast"))
1337                 n = RTN_MULTICAST;
1338         else if (!strcmp(action, "prohibit"))
1339                 n = RTN_PROHIBIT;
1340         else if (!strcmp(action, "unreachable"))
1341                 n = RTN_UNREACHABLE;
1342         else if (!strcmp(action, "blackhole"))
1343                 n = RTN_BLACKHOLE;
1344         else if (!strcmp(action, "xresolve"))
1345                 n = RTN_XRESOLVE;
1346         else if (!strcmp(action, "unicast"))
1347                 n = RTN_UNICAST;
1348         else if (!strcmp(action, "throw"))
1349                 n = RTN_THROW;
1350         else if (!strcmp(action, "failed_policy"))
1351                 n = RTN_FAILED_POLICY;
1352         else {
1353                 n = strtoul(action, &e, 0);
1354                 if (!e || *e || e == action || n > 255)
1355                         return false;
1356         }
1357
1358         *id = n;
1359         return true;
1360 }
1361
1362 time_t system_get_rtime(void)
1363 {
1364         struct timespec ts;
1365         struct timeval tv;
1366
1367         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1368                 return ts.tv_sec;
1369
1370         if (gettimeofday(&tv, NULL) == 0)
1371                 return tv.tv_sec;
1372
1373         return 0;
1374 }
1375
1376 #ifndef IP_DF
1377 #define IP_DF       0x4000
1378 #endif
1379
1380 static int tunnel_ioctl(const char *name, int cmd, void *p)
1381 {
1382         struct ifreq ifr;
1383
1384         memset(&ifr, 0, sizeof(ifr));
1385         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1386         ifr.ifr_ifru.ifru_data = p;
1387         return ioctl(sock_ioctl, cmd, &ifr);
1388 }
1389
1390 int system_del_ip_tunnel(const char *name)
1391 {
1392         return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1393 }
1394
1395 int system_update_ipv6_mtu(struct device *dev, int mtu)
1396 {
1397         int ret = -1;
1398         char buf[64];
1399         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1400                         dev->ifname);
1401
1402         int fd = open(buf, O_RDWR);
1403         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1404         if (len < 0)
1405                 goto out;
1406
1407         buf[len] = 0;
1408         ret = atoi(buf);
1409
1410         if (!mtu || ret <= mtu)
1411                 goto out;
1412
1413         lseek(fd, 0, SEEK_SET);
1414         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1415                 ret = -1;
1416
1417 out:
1418         close(fd);
1419         return ret;
1420 }
1421
1422
1423 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1424 {
1425         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1426         struct blob_attr *cur;
1427         const char *str;
1428
1429         system_del_ip_tunnel(name);
1430
1431         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1432                 blob_data(attr), blob_len(attr));
1433
1434         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1435                 return -EINVAL;
1436         str = blobmsg_data(cur);
1437
1438         unsigned int ttl = 0;
1439         if ((cur = tb[TUNNEL_ATTR_TTL]) && (ttl = blobmsg_get_u32(cur)) > 255)
1440                 return -EINVAL;
1441
1442         unsigned int link = 0;
1443         if ((cur = tb[TUNNEL_ATTR_LINK])) {
1444                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
1445                 if (!iface)
1446                         return -EINVAL;
1447
1448                 if (iface->l3_dev.dev)
1449                         link = iface->l3_dev.dev->ifindex;
1450         }
1451
1452
1453         if (!strcmp(str, "sit")) {
1454                 struct ip_tunnel_parm p = {
1455                         .link = link,
1456                         .iph = {
1457                                 .version = 4,
1458                                 .ihl = 5,
1459                                 .frag_off = htons(IP_DF),
1460                                 .protocol = IPPROTO_IPV6,
1461                                 .ttl = ttl
1462                         }
1463                 };
1464
1465                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1466                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1467                         return -EINVAL;
1468
1469                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1470                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1471                         return -EINVAL;
1472
1473                 strncpy(p.name, name, sizeof(p.name));
1474                 if (tunnel_ioctl("sit0", SIOCADDTUNNEL, &p) < 0)
1475                         return -1;
1476
1477 #ifdef SIOCADD6RD
1478                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
1479                         unsigned int mask;
1480                         struct ip_tunnel_6rd p6;
1481
1482                         memset(&p6, 0, sizeof(p6));
1483
1484                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
1485                                                 &p6.prefix, &mask) || mask > 128)
1486                                 return -EINVAL;
1487                         p6.prefixlen = mask;
1488
1489                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
1490                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
1491                                                         &p6.relay_prefix, &mask) || mask > 32)
1492                                         return -EINVAL;
1493                                 p6.relay_prefixlen = mask;
1494                         }
1495
1496                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
1497                                 system_del_ip_tunnel(name);
1498                                 return -1;
1499                         }
1500                 }
1501 #endif
1502         } else if (!strcmp(str, "ipip6")) {
1503                 struct ip6_tnl_parm p = {
1504                         .link = link,
1505                         .proto = IPPROTO_IPIP,
1506                         .hop_limit = (ttl) ? ttl : 64,
1507                         .encap_limit = 4,
1508                 };
1509
1510                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1511                                 inet_pton(AF_INET6, blobmsg_data(cur), &p.laddr) < 1)
1512                         return -EINVAL;
1513
1514                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1515                                 inet_pton(AF_INET6, blobmsg_data(cur), &p.raddr) < 1)
1516                         return -EINVAL;
1517
1518                 strncpy(p.name, name, sizeof(p.name));
1519                 if (tunnel_ioctl("ip6tnl0", SIOCADDTUNNEL, &p) < 0)
1520                         return -1;
1521         } else
1522                 return -EINVAL;
1523
1524         return 0;
1525 }