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