system-linux.c

   1 /*
   2  * netifd - network interface daemon
   3  * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
   4  * Copyright (C) 2013 Jo-Philipp Wich <jow@openwrt.org>
   5  * Copyright (C) 2013 Steven Barth <steven@midlink.org>
   6  * Copyright (C) 2014 Gioacchino Mazzurco <gio@eigenlab.org>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2
  10  * as published by the Free Software Foundation
  11  *
  12  * This program is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  */
  17 #define _GNU_SOURCE
  18
  19 #include <sys/socket.h>
  20 #include <sys/ioctl.h>
  21 #include <sys/stat.h>
  22 #include <sys/syscall.h>
  23
  24 #include <net/if.h>
  25 #include <net/if_arp.h>
  26
  27 #include <arpa/inet.h>
  28 #include <netinet/in.h>
  29
  30 #include <linux/rtnetlink.h>
  31 #include <linux/sockios.h>
  32 #include <linux/ip.h>
  33 #include <linux/if_link.h>
  34 #include <linux/if_vlan.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_tunnel.h>
  37 #include <linux/ip6_tunnel.h>
  38 #include <linux/ethtool.h>
  39 #include <linux/fib_rules.h>
  40 #include <linux/version.h>
  41
  42 #ifndef RTN_FAILED_POLICY
  43 #define RTN_FAILED_POLICY 12
  44 #endif
  45
  46 #include <string.h>
  47 #include <fcntl.h>
  48 #include <glob.h>
  49 #include <time.h>
  50
  51 #include <netlink/msg.h>
  52 #include <netlink/attr.h>
  53 #include <netlink/socket.h>
  54 #include <libubox/uloop.h>
  55
  56 #include "netifd.h"
  57 #include "device.h"
  58 #include "system.h"
  59
  60 struct event_socket {
  61         struct uloop_fd uloop;
  62         struct nl_sock *sock;
  63         int bufsize;
  64 };
  65
  66 static int sock_ioctl = -1;
  67 static struct nl_sock *sock_rtnl = NULL;
  68
  69 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
  70 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
  71
  72 static char dev_buf[256];
  73
  74 static void
  75 handler_nl_event(struct uloop_fd *u, unsigned int events)
  76 {
  77         struct event_socket *ev = container_of(u, struct event_socket, uloop);
  78         int err;
  79         socklen_t errlen = sizeof(err);
  80
  81         if (!u->error) {
  82                 nl_recvmsgs_default(ev->sock);
  83                 return;
  84         }
  85
  86         if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
  87                 goto abort;
  88
  89         switch(err) {
  90         case ENOBUFS:
  91                 // Increase rx buffer size on netlink socket
  92                 ev->bufsize *= 2;
  93                 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
  94                         goto abort;
  95
  96                 // Request full dump since some info got dropped
  97                 struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
  98                 nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
  99                 break;
 100
 101         default:
 102                 goto abort;
 103         }
 104         u->error = false;
 105         return;
 106
 107 abort:
 108         uloop_fd_delete(&ev->uloop);
 109         return;
 110 }
 111
 112 static struct nl_sock *
 113 create_socket(int protocol, int groups)
 114 {
 115         struct nl_sock *sock;
 116
 117         sock = nl_socket_alloc();
 118         if (!sock)
 119                 return NULL;
 120
 121         if (groups)
 122                 nl_join_groups(sock, groups);
 123
 124         if (nl_connect(sock, protocol))
 125                 return NULL;
 126
 127         return sock;
 128 }
 129
 130 static bool
 131 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
 132                         uloop_fd_handler cb, int flags)
 133 {
 134         ev->sock = create_socket(protocol, groups);
 135         if (!ev->sock)
 136                 return false;
 137
 138         ev->uloop.fd = nl_socket_get_fd(ev->sock);
 139         ev->uloop.cb = cb;
 140         if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
 141                 return false;
 142
 143         return true;
 144 }
 145
 146 static bool
 147 create_event_socket(struct event_socket *ev, int protocol,
 148                     int (*cb)(struct nl_msg *msg, void *arg))
 149 {
 150         if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
 151                 return false;
 152
 153         // Install the valid custom callback handler
 154         nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
 155
 156         // Disable sequence number checking on event sockets
 157         nl_socket_disable_seq_check(ev->sock);
 158
 159         // Increase rx buffer size to 65K on event sockets
 160         ev->bufsize = 65535;
 161         if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
 162                 return false;
 163
 164         return true;
 165 }
 166
 167 int system_init(void)
 168 {
 169         static struct event_socket rtnl_event;
 170         static struct event_socket hotplug_event;
 171
 172         sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
 173         system_fd_set_cloexec(sock_ioctl);
 174
 175         // Prepare socket for routing / address control
 176         sock_rtnl = create_socket(NETLINK_ROUTE, 0);
 177         if (!sock_rtnl)
 178                 return -1;
 179
 180         if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
 181                 return -1;
 182
 183         if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
 184                                      handle_hotplug_event, 0))
 185                 return -1;
 186
 187         // Receive network link events form kernel
 188         nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
 189
 190         return 0;
 191 }
 192
 193 static void system_set_sysctl(const char *path, const char *val)
 194 {
 195         int fd;
 196
 197         fd = open(path, O_WRONLY);
 198         if (fd < 0)
 199                 return;
 200
 201         if (write(fd, val, strlen(val))) {}
 202         close(fd);
 203 }
 204
 205 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
 206 {
 207         snprintf(dev_buf, sizeof(dev_buf), path, device);
 208         system_set_sysctl(dev_buf, val);
 209 }
 210
 211 static void system_set_disable_ipv6(struct device *dev, const char *val)
 212 {
 213         system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
 214 }
 215
 216 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
 217 {
 218         int fd = -1, ret = -1;
 219
 220         fd = open(path, O_RDONLY);
 221         if (fd < 0)
 222                 goto out;
 223
 224         ssize_t len = read(fd, buf, buf_sz - 1);
 225         if (len < 0)
 226                 goto out;
 227
 228         ret = buf[len] = 0;
 229
 230 out:
 231         if (fd >= 0)
 232                 close(fd);
 233
 234         return ret;
 235 }
 236
 237 static int
 238 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
 239 {
 240         snprintf(dev_buf, sizeof(dev_buf), path, device);
 241         return system_get_sysctl(dev_buf, buf, buf_sz);
 242 }
 243
 244 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
 245 {
 246         return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
 247                         dev->ifname, buf, buf_sz);
 248 }
 249
 250 #ifndef IFF_LOWER_UP
 251 #define IFF_LOWER_UP    0x10000
 252 #endif
 253
 254 // Evaluate netlink messages
 255 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 256 {
 257         struct nlmsghdr *nh = nlmsg_hdr(msg);
 258         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 259         struct nlattr *nla[__IFLA_MAX];
 260
 261         if (nh->nlmsg_type != RTM_NEWLINK)
 262                 goto out;
 263
 264         nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
 265         if (!nla[IFLA_IFNAME])
 266                 goto out;
 267
 268         struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
 269         if (!dev)
 270                 goto out;
 271
 272         device_set_ifindex(dev, ifi->ifi_index);
 273         device_set_link(dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
 274
 275 out:
 276         return 0;
 277 }
 278
 279 static void
 280 handle_hotplug_msg(char *data, int size)
 281 {
 282         const char *subsystem = NULL, *interface = NULL;
 283         char *cur, *end, *sep;
 284         struct device *dev;
 285         int skip;
 286         bool add;
 287
 288         if (!strncmp(data, "add@", 4))
 289                 add = true;
 290         else if (!strncmp(data, "remove@", 7))
 291                 add = false;
 292         else
 293                 return;
 294
 295         skip = strlen(data) + 1;
 296         end = data + size;
 297
 298         for (cur = data + skip; cur < end; cur += skip) {
 299                 skip = strlen(cur) + 1;
 300
 301                 sep = strchr(cur, '=');
 302                 if (!sep)
 303                         continue;
 304
 305                 *sep = 0;
 306                 if (!strcmp(cur, "INTERFACE"))
 307                         interface = sep + 1;
 308                 else if (!strcmp(cur, "SUBSYSTEM")) {
 309                         subsystem = sep + 1;
 310                         if (strcmp(subsystem, "net") != 0)
 311                                 return;
 312                 }
 313                 if (subsystem && interface)
 314                         goto found;
 315         }
 316         return;
 317
 318 found:
 319         dev = device_get(interface, false);
 320         if (!dev)
 321                 return;
 322
 323         if (dev->type != &simple_device_type)
 324                 return;
 325
 326         if (add && system_if_force_external(dev->ifname))
 327                 return;
 328
 329         device_set_present(dev, add);
 330 }
 331
 332 static void
 333 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
 334 {
 335         struct event_socket *ev = container_of(u, struct event_socket, uloop);
 336         struct sockaddr_nl nla;
 337         unsigned char *buf = NULL;
 338         int size;
 339
 340         while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
 341                 if (nla.nl_pid == 0)
 342                         handle_hotplug_msg((char *) buf, size);
 343
 344                 free(buf);
 345         }
 346 }
 347
 348 static int system_rtnl_call(struct nl_msg *msg)
 349 {
 350         int ret;
 351
 352         ret = nl_send_auto_complete(sock_rtnl, msg);
 353         nlmsg_free(msg);
 354
 355         if (ret < 0)
 356                 return ret;
 357
 358         return nl_wait_for_ack(sock_rtnl);
 359 }
 360
 361 int system_bridge_delbr(struct device *bridge)
 362 {
 363         return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 364 }
 365
 366 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 367 {
 368         struct ifreq ifr;
 369
 370         memset(&ifr, 0, sizeof(ifr));
 371         if (dev)
 372                 ifr.ifr_ifindex = dev->ifindex;
 373         else
 374                 ifr.ifr_data = data;
 375         strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
 376         return ioctl(sock_ioctl, cmd, &ifr);
 377 }
 378
 379 static bool system_is_bridge(const char *name, char *buf, int buflen)
 380 {
 381         struct stat st;
 382
 383         snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
 384         if (stat(buf, &st) < 0)
 385                 return false;
 386
 387         return true;
 388 }
 389
 390 static char *system_get_bridge(const char *name, char *buf, int buflen)
 391 {
 392         char *path;
 393         ssize_t len;
 394         glob_t gl;
 395
 396         snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
 397         if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
 398                 return NULL;
 399
 400         if (gl.gl_pathc == 0)
 401                 return NULL;
 402
 403         len = readlink(gl.gl_pathv[0], buf, buflen);
 404         if (len < 0)
 405                 return NULL;
 406
 407         buf[len] = 0;
 408         path = strrchr(buf, '/');
 409         if (!path)
 410                 return NULL;
 411
 412         return path + 1;
 413 }
 414
 415 int system_bridge_addif(struct device *bridge, struct device *dev)
 416 {
 417         char *oldbr;
 418
 419         oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
 420         if (oldbr && !strcmp(oldbr, bridge->ifname))
 421                 return 0;
 422
 423         return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 424 }
 425
 426 int system_bridge_delif(struct device *bridge, struct device *dev)
 427 {
 428         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 429 }
 430
 431 static int system_if_resolve(struct device *dev)
 432 {
 433         struct ifreq ifr;
 434         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 435         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 436                 return ifr.ifr_ifindex;
 437         else
 438                 return 0;
 439 }
 440
 441 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 442 {
 443         struct ifreq ifr;
 444
 445         memset(&ifr, 0, sizeof(ifr));
 446         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 447         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 448         ifr.ifr_flags |= add;
 449         ifr.ifr_flags &= ~rem;
 450         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 451 }
 452
 453 struct clear_data {
 454         struct nl_msg *msg;
 455         struct device *dev;
 456         int type;
 457         int size;
 458         int af;
 459 };
 460
 461
 462 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 463 {
 464         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 465
 466         return ifa->ifa_index == ifindex;
 467 }
 468
 469 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 470 {
 471         struct rtmsg *r = NLMSG_DATA(hdr);
 472         struct nlattr *tb[__RTA_MAX];
 473
 474         if (r->rtm_protocol == RTPROT_KERNEL &&
 475             r->rtm_family == AF_INET6)
 476                 return false;
 477
 478         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 479         if (!tb[RTA_OIF])
 480                 return false;
 481
 482         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 483 }
 484
 485 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
 486 {
 487         return true;
 488 }
 489
 490 static int cb_clear_event(struct nl_msg *msg, void *arg)
 491 {
 492         struct clear_data *clr = arg;
 493         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 494         bool (*cb)(struct nlmsghdr *, int ifindex);
 495         int type;
 496
 497         switch(clr->type) {
 498         case RTM_GETADDR:
 499                 type = RTM_DELADDR;
 500                 if (hdr->nlmsg_type != RTM_NEWADDR)
 501                         return NL_SKIP;
 502
 503                 cb = check_ifaddr;
 504                 break;
 505         case RTM_GETROUTE:
 506                 type = RTM_DELROUTE;
 507                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 508                         return NL_SKIP;
 509
 510                 cb = check_route;
 511                 break;
 512         case RTM_GETRULE:
 513                 type = RTM_DELRULE;
 514                 if (hdr->nlmsg_type != RTM_NEWRULE)
 515                         return NL_SKIP;
 516
 517                 cb = check_rule;
 518                 break;
 519         default:
 520                 return NL_SKIP;
 521         }
 522
 523         if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
 524                 return NL_SKIP;
 525
 526         if (type == RTM_DELRULE)
 527                 D(SYSTEM, "Remove a rule\n");
 528         else
 529                 D(SYSTEM, "Remove %s from device %s\n",
 530                   type == RTM_DELADDR ? "an address" : "a route",
 531                   clr->dev->ifname);
 532         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 533         hdr = nlmsg_hdr(clr->msg);
 534         hdr->nlmsg_type = type;
 535         hdr->nlmsg_flags = NLM_F_REQUEST;
 536
 537         nl_socket_disable_auto_ack(sock_rtnl);
 538         nl_send_auto_complete(sock_rtnl, clr->msg);
 539         nl_socket_enable_auto_ack(sock_rtnl);
 540
 541         return NL_SKIP;
 542 }
 543
 544 static int
 545 cb_finish_event(struct nl_msg *msg, void *arg)
 546 {
 547         int *pending = arg;
 548         *pending = 0;
 549         return NL_STOP;
 550 }
 551
 552 static int
 553 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 554 {
 555         int *pending = arg;
 556         *pending = err->error;
 557         return NL_STOP;
 558 }
 559
 560 static void
 561 system_if_clear_entries(struct device *dev, int type, int af)
 562 {
 563         struct clear_data clr;
 564         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 565         struct rtmsg rtm = {
 566                 .rtm_family = af,
 567                 .rtm_flags = RTM_F_CLONED,
 568         };
 569         int flags = NLM_F_DUMP;
 570         int pending = 1;
 571
 572         clr.af = af;
 573         clr.dev = dev;
 574         clr.type = type;
 575         switch (type) {
 576         case RTM_GETADDR:
 577         case RTM_GETRULE:
 578                 clr.size = sizeof(struct rtgenmsg);
 579                 break;
 580         case RTM_GETROUTE:
 581                 clr.size = sizeof(struct rtmsg);
 582                 break;
 583         default:
 584                 return;
 585         }
 586
 587         if (!cb)
 588                 return;
 589
 590         clr.msg = nlmsg_alloc_simple(type, flags);
 591         if (!clr.msg)
 592                 goto out;
 593
 594         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 595         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 596         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 597         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 598
 599         nl_send_auto_complete(sock_rtnl, clr.msg);
 600         while (pending > 0)
 601                 nl_recvmsgs(sock_rtnl, cb);
 602
 603         nlmsg_free(clr.msg);
 604 out:
 605         nl_cb_put(cb);
 606 }
 607
 608 /*
 609  * Clear bridge (membership) state and bring down device
 610  */
 611 void system_if_clear_state(struct device *dev)
 612 {
 613         static char buf[256];
 614         char *bridge;
 615
 616         device_set_ifindex(dev, system_if_resolve(dev));
 617         if (dev->external || !dev->ifindex)
 618                 return;
 619
 620         system_if_flags(dev->ifname, 0, IFF_UP);
 621
 622         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 623                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 624                 system_bridge_delbr(dev);
 625                 return;
 626         }
 627
 628         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 629         if (bridge) {
 630                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 631                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 632         }
 633
 634         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 635         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 636         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 637         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 638         system_set_disable_ipv6(dev, "0");
 639 }
 640
 641 static inline unsigned long
 642 sec_to_jiffies(int val)
 643 {
 644         return (unsigned long) val * 100;
 645 }
 646
 647 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 648 {
 649         unsigned long args[4] = {};
 650
 651         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 652                 return -1;
 653
 654         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 655         args[1] = !!cfg->stp;
 656         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 657
 658         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 659         args[1] = sec_to_jiffies(cfg->forward_delay);
 660         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 661
 662         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 663                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 664
 665         args[0] = BRCTL_SET_BRIDGE_PRIORITY;
 666         args[1] = cfg->priority;
 667         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 668
 669         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 670                 args[0] = BRCTL_SET_AGEING_TIME;
 671                 args[1] = sec_to_jiffies(cfg->ageing_time);
 672                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 673         }
 674
 675         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 676                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 677                 args[1] = sec_to_jiffies(cfg->hello_time);
 678                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 679         }
 680
 681         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 682                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 683                 args[1] = sec_to_jiffies(cfg->max_age);
 684                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 685         }
 686
 687         return 0;
 688 }
 689
 690 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
 691 {
 692         struct nl_msg *msg;
 693         struct nlattr *linkinfo, *data;
 694         struct ifinfomsg iim = { .ifi_family = AF_INET };
 695         int ifindex = system_if_resolve(dev);
 696         int i, rv;
 697         static const struct {
 698                 const char *name;
 699                 enum macvlan_mode val;
 700         } modes[] = {
 701                 { "private", MACVLAN_MODE_PRIVATE },
 702                 { "vepa", MACVLAN_MODE_VEPA },
 703                 { "bridge", MACVLAN_MODE_BRIDGE },
 704                 { "passthru", MACVLAN_MODE_PASSTHRU },
 705         };
 706
 707         if (ifindex == 0)
 708                 return -ENOENT;
 709
 710         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 711
 712         if (!msg)
 713                 return -1;
 714
 715         nlmsg_append(msg, &iim, sizeof(iim), 0);
 716
 717         if (cfg->flags & MACVLAN_OPT_MACADDR)
 718                 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
 719         nla_put(msg, IFLA_IFNAME, IFNAMSIZ, macvlan->ifname);
 720         nla_put_u32(msg, IFLA_LINK, ifindex);
 721
 722         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 723                 goto nla_put_failure;
 724
 725         nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
 726
 727         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 728                 goto nla_put_failure;
 729
 730         if (cfg->mode) {
 731                 for (i = 0; i < ARRAY_SIZE(modes); i++) {
 732                         if (strcmp(cfg->mode, modes[i].name) != 0)
 733                                 continue;
 734
 735                         nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
 736                         break;
 737                 }
 738         }
 739
 740         nla_nest_end(msg, data);
 741         nla_nest_end(msg, linkinfo);
 742
 743         rv = system_rtnl_call(msg);
 744         if (rv)
 745                 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
 746
 747         return rv;
 748
 749 nla_put_failure:
 750         nlmsg_free(msg);
 751         return -ENOMEM;
 752 }
 753
 754 int system_macvlan_del(struct device *macvlan)
 755 {
 756         struct nl_msg *msg;
 757         struct ifinfomsg iim;
 758
 759         iim.ifi_family = AF_INET;
 760         iim.ifi_index  = 0;
 761
 762         msg = nlmsg_alloc_simple(RTM_DELLINK, 0);
 763
 764         if (!msg)
 765                 return -1;
 766
 767         nlmsg_append(msg, &iim, sizeof(iim), 0);
 768
 769         nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
 770         nla_put(msg, IFLA_IFNAME, sizeof(macvlan->ifname), macvlan->ifname);
 771
 772         system_rtnl_call(msg);
 773
 774         return 0;
 775 }
 776
 777 static int system_vlan(struct device *dev, int id)
 778 {
 779         struct vlan_ioctl_args ifr = {
 780                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 781                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 782         };
 783
 784         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 785
 786         if (id < 0) {
 787                 ifr.cmd = DEL_VLAN_CMD;
 788                 ifr.u.VID = 0;
 789         } else {
 790                 ifr.cmd = ADD_VLAN_CMD;
 791                 ifr.u.VID = id;
 792         }
 793         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 794         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 795 }
 796
 797 int system_vlan_add(struct device *dev, int id)
 798 {
 799         return system_vlan(dev, id);
 800 }
 801
 802 int system_vlan_del(struct device *dev)
 803 {
 804         return system_vlan(dev, -1);
 805 }
 806
 807 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
 808 {
 809         struct nl_msg *msg;
 810         struct nlattr *linkinfo, *data;
 811         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
 812         int ifindex = system_if_resolve(dev);
 813         int rv;
 814
 815         if (ifindex == 0)
 816                 return -ENOENT;
 817
 818         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 819
 820         if (!msg)
 821                 return -1;
 822
 823         nlmsg_append(msg, &iim, sizeof(iim), 0);
 824         nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
 825         nla_put_u32(msg, IFLA_LINK, ifindex);
 826
 827         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 828                 goto nla_put_failure;
 829
 830         nla_put_string(msg, IFLA_INFO_KIND, "vlan");
 831
 832         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 833                 goto nla_put_failure;
 834
 835         nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
 836
 837 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
 838         nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
 839 #else
 840         if(cfg->proto == VLAN_PROTO_8021AD)
 841                 netifd_log_message(L_WARNING, "%s Your kernel is older than linux 3.10.0, 802.1ad is not supported defaulting to 802.1q", vlandev->type->name);
 842 #endif
 843
 844         nla_nest_end(msg, data);
 845         nla_nest_end(msg, linkinfo);
 846
 847         rv = system_rtnl_call(msg);
 848         if (rv)
 849                 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
 850
 851         return rv;
 852
 853 nla_put_failure:
 854         nlmsg_free(msg);
 855         return -ENOMEM;
 856 }
 857
 858 int system_vlandev_del(struct device *vlandev)
 859 {
 860         struct nl_msg *msg;
 861         struct ifinfomsg iim = {
 862                 .ifi_family = AF_UNSPEC,
 863                 .ifi_index = 0,
 864         };
 865
 866         msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
 867
 868         if (!msg)
 869                 return -1;
 870
 871         nlmsg_append(msg, &iim, sizeof(iim), 0);
 872
 873         nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
 874
 875         system_rtnl_call(msg);
 876
 877         return 0;
 878 }
 879
 880 static void
 881 system_if_get_settings(struct device *dev, struct device_settings *s)
 882 {
 883         struct ifreq ifr;
 884         char buf[10];
 885
 886         memset(&ifr, 0, sizeof(ifr));
 887         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 888
 889         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 890                 s->mtu = ifr.ifr_mtu;
 891                 s->flags |= DEV_OPT_MTU;
 892         }
 893
 894         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 895                 s->txqueuelen = ifr.ifr_qlen;
 896                 s->flags |= DEV_OPT_TXQUEUELEN;
 897         }
 898
 899         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 900                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 901                 s->flags |= DEV_OPT_MACADDR;
 902         }
 903
 904         if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
 905                 s->ipv6 = !strtoul(buf, NULL, 0);
 906                 s->flags |= DEV_OPT_IPV6;
 907         }
 908 }
 909
 910 void
 911 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
 912 {
 913         struct ifreq ifr;
 914
 915         if (!apply_mask)
 916                 return;
 917
 918         memset(&ifr, 0, sizeof(ifr));
 919         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 920         if (s->flags & DEV_OPT_MTU & apply_mask) {
 921                 ifr.ifr_mtu = s->mtu;
 922                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 923                         s->flags &= ~DEV_OPT_MTU;
 924         }
 925         if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
 926                 ifr.ifr_qlen = s->txqueuelen;
 927                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 928                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 929         }
 930         if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
 931                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 932                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 933                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 934                         s->flags &= ~DEV_OPT_MACADDR;
 935         }
 936         if (s->flags & DEV_OPT_IPV6 & apply_mask)
 937                 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
 938 }
 939
 940 int system_if_up(struct device *dev)
 941 {
 942         system_if_get_settings(dev, &dev->orig_settings);
 943         system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
 944         device_set_ifindex(dev, system_if_resolve(dev));
 945         return system_if_flags(dev->ifname, IFF_UP, 0);
 946 }
 947
 948 int system_if_down(struct device *dev)
 949 {
 950         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
 951         dev->orig_settings.flags &= dev->settings.flags;
 952         system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
 953         return ret;
 954 }
 955
 956 struct if_check_data {
 957         struct device *dev;
 958         int pending;
 959         int ret;
 960 };
 961
 962 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
 963 {
 964         struct nlmsghdr *nh = nlmsg_hdr(msg);
 965         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 966         struct if_check_data *chk = (struct if_check_data *)arg;
 967
 968         if (nh->nlmsg_type != RTM_NEWLINK)
 969                 return NL_SKIP;
 970
 971         device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
 972         device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
 973
 974         return NL_OK;
 975 }
 976
 977 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
 978 {
 979         struct if_check_data *chk = (struct if_check_data *)arg;
 980         chk->pending = 0;
 981         return NL_STOP;
 982 }
 983
 984 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 985 {
 986         struct if_check_data *chk = (struct if_check_data *)arg;
 987
 988         device_set_present(chk->dev, false);
 989         device_set_link(chk->dev, false);
 990         chk->pending = err->error;
 991
 992         return NL_STOP;
 993 }
 994
 995 int system_if_check(struct device *dev)
 996 {
 997         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 998         struct nl_msg *msg;
 999         struct ifinfomsg ifi = {
1000                 .ifi_family = AF_UNSPEC,
1001                 .ifi_index = 0,
1002         };
1003         struct if_check_data chk = {
1004                 .dev = dev,
1005                 .pending = 1,
1006         };
1007         int ret = 1;
1008
1009         msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1010         if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1011             nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1012                 goto out;
1013
1014         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1015         nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1016         nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1017
1018         nl_send_auto_complete(sock_rtnl, msg);
1019         while (chk.pending > 0)
1020                 nl_recvmsgs(sock_rtnl, cb);
1021
1022         nlmsg_free(msg);
1023         ret = chk.pending;
1024
1025 out:
1026         nl_cb_put(cb);
1027         return ret;
1028 }
1029
1030 struct device *
1031 system_if_get_parent(struct device *dev)
1032 {
1033         char buf[64], *devname;
1034         int ifindex, iflink, len;
1035         FILE *f;
1036
1037         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1038         f = fopen(buf, "r");
1039         if (!f)
1040                 return NULL;
1041
1042         len = fread(buf, 1, sizeof(buf) - 1, f);
1043         fclose(f);
1044
1045         if (len <= 0)
1046                 return NULL;
1047
1048         buf[len] = 0;
1049         iflink = strtoul(buf, NULL, 0);
1050         ifindex = system_if_resolve(dev);
1051         if (!iflink || iflink == ifindex)
1052                 return NULL;
1053
1054         devname = if_indextoname(iflink, buf);
1055         if (!devname)
1056                 return NULL;
1057
1058         return device_get(devname, true);
1059 }
1060
1061 static bool
1062 read_string_file(int dir_fd, const char *file, char *buf, int len)
1063 {
1064         bool ret = false;
1065         char *c;
1066         int fd;
1067
1068         fd = openat(dir_fd, file, O_RDONLY);
1069         if (fd < 0)
1070                 return false;
1071
1072 retry:
1073         len = read(fd, buf, len - 1);
1074         if (len < 0) {
1075                 if (errno == EINTR)
1076                         goto retry;
1077         } else if (len > 0) {
1078                         buf[len] = 0;
1079
1080                         c = strchr(buf, '\n');
1081                         if (c)
1082                                 *c = 0;
1083
1084                         ret = true;
1085         }
1086
1087         close(fd);
1088
1089         return ret;
1090 }
1091
1092 static bool
1093 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1094 {
1095         char buf[64];
1096         bool ret = false;
1097
1098         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1099         if (ret)
1100                 *val = strtoull(buf, NULL, 0);
1101
1102         return ret;
1103 }
1104
1105 /* Assume advertised flags == supported flags */
1106 static const struct {
1107         uint32_t mask;
1108         const char *name;
1109 } ethtool_link_modes[] = {
1110         { ADVERTISED_10baseT_Half, "10H" },
1111         { ADVERTISED_10baseT_Full, "10F" },
1112         { ADVERTISED_100baseT_Half, "100H" },
1113         { ADVERTISED_100baseT_Full, "100F" },
1114         { ADVERTISED_1000baseT_Half, "1000H" },
1115         { ADVERTISED_1000baseT_Full, "1000F" },
1116 };
1117
1118 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1119 {
1120         int i;
1121         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1122                 if (mask & ethtool_link_modes[i].mask)
1123                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1124         }
1125 }
1126
1127 bool
1128 system_if_force_external(const char *ifname)
1129 {
1130         char buf[64];
1131         struct stat s;
1132
1133         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1134         return stat(buf, &s) == 0;
1135 }
1136
1137 int
1138 system_if_dump_info(struct device *dev, struct blob_buf *b)
1139 {
1140         struct ethtool_cmd ecmd;
1141         struct ifreq ifr;
1142         char buf[64], *s;
1143         void *c;
1144         int dir_fd;
1145
1146         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1147         dir_fd = open(buf, O_DIRECTORY);
1148
1149         memset(&ecmd, 0, sizeof(ecmd));
1150         memset(&ifr, 0, sizeof(ifr));
1151         strcpy(ifr.ifr_name, dev->ifname);
1152         ifr.ifr_data = (caddr_t) &ecmd;
1153         ecmd.cmd = ETHTOOL_GSET;
1154
1155         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1156                 c = blobmsg_open_array(b, "link-advertising");
1157                 system_add_link_modes(b, ecmd.advertising);
1158                 blobmsg_close_array(b, c);
1159
1160                 c = blobmsg_open_array(b, "link-supported");
1161                 system_add_link_modes(b, ecmd.supported);
1162                 blobmsg_close_array(b, c);
1163
1164                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1165                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1166                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1167                 blobmsg_add_string_buffer(b);
1168         }
1169
1170         close(dir_fd);
1171         return 0;
1172 }
1173
1174 int
1175 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1176 {
1177         const char *const counters[] = {
1178                 "collisions",     "rx_frame_errors",   "tx_compressed",
1179                 "multicast",      "rx_length_errors",  "tx_dropped",
1180                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
1181                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
1182                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
1183                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
1184                 "rx_errors",      "tx_bytes",          "tx_window_errors",
1185                 "rx_fifo_errors", "tx_carrier_errors",
1186         };
1187         char buf[64];
1188         int stats_dir;
1189         int i;
1190         uint64_t val = 0;
1191
1192         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1193         stats_dir = open(buf, O_DIRECTORY);
1194         if (stats_dir < 0)
1195                 return -1;
1196
1197         for (i = 0; i < ARRAY_SIZE(counters); i++)
1198                 if (read_uint64_file(stats_dir, counters[i], &val))
1199                         blobmsg_add_u64(b, counters[i], val);
1200
1201         close(stats_dir);
1202         return 0;
1203 }
1204
1205 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1206 {
1207         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1208         int alen = v4 ? 4 : 16;
1209         unsigned int flags = 0;
1210         struct ifaddrmsg ifa = {
1211                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1212                 .ifa_prefixlen = addr->mask,
1213                 .ifa_index = dev->ifindex,
1214         };
1215
1216         struct nl_msg *msg;
1217         if (cmd == RTM_NEWADDR)
1218                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1219
1220         msg = nlmsg_alloc_simple(cmd, flags);
1221         if (!msg)
1222                 return -1;
1223
1224         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1225         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1226         if (v4) {
1227                 if (addr->broadcast)
1228                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1229                 if (addr->point_to_point)
1230                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1231         } else {
1232                 time_t now = system_get_rtime();
1233                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1234
1235                 if (addr->preferred_until) {
1236                         int64_t preferred = addr->preferred_until - now;
1237                         if (preferred < 0)
1238                                 preferred = 0;
1239                         else if (preferred > UINT32_MAX)
1240                                 preferred = UINT32_MAX;
1241
1242                         cinfo.ifa_prefered = preferred;
1243                 }
1244
1245                 if (addr->valid_until) {
1246                         int64_t valid = addr->valid_until - now;
1247                         if (valid <= 0)
1248                                 return -1;
1249                         else if (valid > UINT32_MAX)
1250                                 valid = UINT32_MAX;
1251
1252                         cinfo.ifa_valid = valid;
1253                 }
1254
1255                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1256         }
1257
1258         return system_rtnl_call(msg);
1259 }
1260
1261 int system_add_address(struct device *dev, struct device_addr *addr)
1262 {
1263         return system_addr(dev, addr, RTM_NEWADDR);
1264 }
1265
1266 int system_del_address(struct device *dev, struct device_addr *addr)
1267 {
1268         return system_addr(dev, addr, RTM_DELADDR);
1269 }
1270
1271 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1272 {
1273         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1274         bool have_gw;
1275         unsigned int flags = 0;
1276
1277         if (alen == 4)
1278                 have_gw = !!route->nexthop.in.s_addr;
1279         else
1280                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1281                         route->nexthop.in6.s6_addr32[1] ||
1282                         route->nexthop.in6.s6_addr32[2] ||
1283                         route->nexthop.in6.s6_addr32[3];
1284
1285         unsigned char scope = (cmd == RTM_DELROUTE) ? RT_SCOPE_NOWHERE :
1286                         (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1287
1288         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1289                         ? route->table : RT_TABLE_MAIN;
1290
1291         struct rtmsg rtm = {
1292                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1293                 .rtm_dst_len = route->mask,
1294                 .rtm_src_len = route->sourcemask,
1295                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1296                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1297                 .rtm_scope = scope,
1298                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1299                 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1300         };
1301         struct nl_msg *msg;
1302
1303         if (cmd == RTM_NEWROUTE) {
1304                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1305
1306                 if (!dev) { // Add null-route
1307                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1308                         rtm.rtm_type = RTN_UNREACHABLE;
1309                 }
1310         }
1311
1312         msg = nlmsg_alloc_simple(cmd, flags);
1313         if (!msg)
1314                 return -1;
1315
1316         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1317
1318         if (route->mask)
1319                 nla_put(msg, RTA_DST, alen, &route->addr);
1320
1321         if (route->sourcemask)
1322                 nla_put(msg, RTA_SRC, alen, &route->source);
1323
1324         if (route->metric > 0)
1325                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1326
1327         if (have_gw)
1328                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1329
1330         if (dev)
1331                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1332
1333         if (table >= 256)
1334                 nla_put_u32(msg, RTA_TABLE, table);
1335
1336         if (route->flags & DEVROUTE_MTU) {
1337                 struct nlattr *metrics;
1338
1339                 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1340                         goto nla_put_failure;
1341
1342                 nla_put_u32(msg, RTAX_MTU, route->mtu);
1343
1344                 nla_nest_end(msg, metrics);
1345         }
1346
1347         return system_rtnl_call(msg);
1348
1349 nla_put_failure:
1350         nlmsg_free(msg);
1351         return -ENOMEM;
1352 }
1353
1354 int system_add_route(struct device *dev, struct device_route *route)
1355 {
1356         return system_rt(dev, route, RTM_NEWROUTE);
1357 }
1358
1359 int system_del_route(struct device *dev, struct device_route *route)
1360 {
1361         return system_rt(dev, route, RTM_DELROUTE);
1362 }
1363
1364 int system_flush_routes(void)
1365 {
1366         const char *names[] = {
1367                 "/proc/sys/net/ipv4/route/flush",
1368                 "/proc/sys/net/ipv6/route/flush"
1369         };
1370         int fd, i;
1371
1372         for (i = 0; i < ARRAY_SIZE(names); i++) {
1373                 fd = open(names[i], O_WRONLY);
1374                 if (fd < 0)
1375                         continue;
1376
1377                 if (write(fd, "-1", 2)) {}
1378                 close(fd);
1379         }
1380         return 0;
1381 }
1382
1383 bool system_resolve_rt_table(const char *name, unsigned int *id)
1384 {
1385         FILE *f;
1386         char *e, buf[128];
1387         unsigned int n, table = RT_TABLE_UNSPEC;
1388
1389         /* first try to parse table as number */
1390         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1391                 table = n;
1392
1393         /* handle well known aliases */
1394         else if (!strcmp(name, "default"))
1395                 table = RT_TABLE_DEFAULT;
1396         else if (!strcmp(name, "main"))
1397                 table = RT_TABLE_MAIN;
1398         else if (!strcmp(name, "local"))
1399                 table = RT_TABLE_LOCAL;
1400
1401         /* try to look up name in /etc/iproute2/rt_tables */
1402         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1403         {
1404                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1405                 {
1406                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1407                                 continue;
1408
1409                         n = strtoul(e, NULL, 10);
1410                         e = strtok(NULL, " \t\n");
1411
1412                         if (e && !strcmp(e, name))
1413                         {
1414                                 table = n;
1415                                 break;
1416                         }
1417                 }
1418
1419                 fclose(f);
1420         }
1421
1422         if (table == RT_TABLE_UNSPEC)
1423                 return false;
1424
1425         /* do not consider main table special */
1426         if (table == RT_TABLE_MAIN)
1427                 table = RT_TABLE_UNSPEC;
1428
1429         *id = table;
1430         return true;
1431 }
1432
1433 static int system_iprule(struct iprule *rule, int cmd)
1434 {
1435         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1436
1437         struct nl_msg *msg;
1438         struct rtmsg rtm = {
1439                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1440                 .rtm_protocol = RTPROT_STATIC,
1441                 .rtm_scope = RT_SCOPE_UNIVERSE,
1442                 .rtm_table = RT_TABLE_UNSPEC,
1443                 .rtm_type = RTN_UNSPEC,
1444                 .rtm_flags = 0,
1445         };
1446
1447         if (cmd == RTM_NEWRULE) {
1448                 rtm.rtm_type = RTN_UNICAST;
1449                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1450         }
1451
1452         if (rule->invert)
1453                 rtm.rtm_flags |= FIB_RULE_INVERT;
1454
1455         if (rule->flags & IPRULE_SRC)
1456                 rtm.rtm_src_len = rule->src_mask;
1457
1458         if (rule->flags & IPRULE_DEST)
1459                 rtm.rtm_dst_len = rule->dest_mask;
1460
1461         if (rule->flags & IPRULE_TOS)
1462                 rtm.rtm_tos = rule->tos;
1463
1464         if (rule->flags & IPRULE_LOOKUP) {
1465                 if (rule->lookup < 256)
1466                         rtm.rtm_table = rule->lookup;
1467         }
1468
1469         if (rule->flags & IPRULE_ACTION)
1470                 rtm.rtm_type = rule->action;
1471         else if (rule->flags & IPRULE_GOTO)
1472                 rtm.rtm_type = FR_ACT_GOTO;
1473         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1474                 rtm.rtm_type = FR_ACT_NOP;
1475
1476         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1477
1478         if (!msg)
1479                 return -1;
1480
1481         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1482
1483         if (rule->flags & IPRULE_IN)
1484                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1485
1486         if (rule->flags & IPRULE_OUT)
1487                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1488
1489         if (rule->flags & IPRULE_SRC)
1490                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1491
1492         if (rule->flags & IPRULE_DEST)
1493                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1494
1495         if (rule->flags & IPRULE_PRIORITY)
1496                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1497         else if (cmd == RTM_NEWRULE)
1498                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1499
1500         if (rule->flags & IPRULE_FWMARK)
1501                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1502
1503         if (rule->flags & IPRULE_FWMASK)
1504                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1505
1506         if (rule->flags & IPRULE_LOOKUP) {
1507                 if (rule->lookup >= 256)
1508                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1509         }
1510
1511         if (rule->flags & IPRULE_GOTO)
1512                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1513
1514         return system_rtnl_call(msg);
1515 }
1516
1517 int system_add_iprule(struct iprule *rule)
1518 {
1519         return system_iprule(rule, RTM_NEWRULE);
1520 }
1521
1522 int system_del_iprule(struct iprule *rule)
1523 {
1524         return system_iprule(rule, RTM_DELRULE);
1525 }
1526
1527 int system_flush_iprules(void)
1528 {
1529         int rv = 0;
1530         struct iprule rule;
1531
1532         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1533         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1534
1535         memset(&rule, 0, sizeof(rule));
1536
1537
1538         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1539
1540         rule.priority = 0;
1541         rule.lookup = RT_TABLE_LOCAL;
1542         rv |= system_iprule(&rule, RTM_NEWRULE);
1543
1544         rule.priority = 32766;
1545         rule.lookup = RT_TABLE_MAIN;
1546         rv |= system_iprule(&rule, RTM_NEWRULE);
1547
1548         rule.priority = 32767;
1549         rule.lookup = RT_TABLE_DEFAULT;
1550         rv |= system_iprule(&rule, RTM_NEWRULE);
1551
1552
1553         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1554
1555         rule.priority = 0;
1556         rule.lookup = RT_TABLE_LOCAL;
1557         rv |= system_iprule(&rule, RTM_NEWRULE);
1558
1559         rule.priority = 32766;
1560         rule.lookup = RT_TABLE_MAIN;
1561         rv |= system_iprule(&rule, RTM_NEWRULE);
1562
1563         return rv;
1564 }
1565
1566 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1567 {
1568         char *e;
1569         unsigned int n;
1570
1571         if (!strcmp(action, "local"))
1572                 n = RTN_LOCAL;
1573         else if (!strcmp(action, "nat"))
1574                 n = RTN_NAT;
1575         else if (!strcmp(action, "broadcast"))
1576                 n = RTN_BROADCAST;
1577         else if (!strcmp(action, "anycast"))
1578                 n = RTN_ANYCAST;
1579         else if (!strcmp(action, "multicast"))
1580                 n = RTN_MULTICAST;
1581         else if (!strcmp(action, "prohibit"))
1582                 n = RTN_PROHIBIT;
1583         else if (!strcmp(action, "unreachable"))
1584                 n = RTN_UNREACHABLE;
1585         else if (!strcmp(action, "blackhole"))
1586                 n = RTN_BLACKHOLE;
1587         else if (!strcmp(action, "xresolve"))
1588                 n = RTN_XRESOLVE;
1589         else if (!strcmp(action, "unicast"))
1590                 n = RTN_UNICAST;
1591         else if (!strcmp(action, "throw"))
1592                 n = RTN_THROW;
1593         else if (!strcmp(action, "failed_policy"))
1594                 n = RTN_FAILED_POLICY;
1595         else {
1596                 n = strtoul(action, &e, 0);
1597                 if (!e || *e || e == action || n > 255)
1598                         return false;
1599         }
1600
1601         *id = n;
1602         return true;
1603 }
1604
1605 time_t system_get_rtime(void)
1606 {
1607         struct timespec ts;
1608         struct timeval tv;
1609
1610         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1611                 return ts.tv_sec;
1612
1613         if (gettimeofday(&tv, NULL) == 0)
1614                 return tv.tv_sec;
1615
1616         return 0;
1617 }
1618
1619 #ifndef IP_DF
1620 #define IP_DF       0x4000
1621 #endif
1622
1623 static int tunnel_ioctl(const char *name, int cmd, void *p)
1624 {
1625         struct ifreq ifr;
1626
1627         memset(&ifr, 0, sizeof(ifr));
1628         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1629         ifr.ifr_ifru.ifru_data = p;
1630         return ioctl(sock_ioctl, cmd, &ifr);
1631 }
1632
1633 int system_del_ip_tunnel(const char *name)
1634 {
1635         return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1636 }
1637
1638 int system_update_ipv6_mtu(struct device *dev, int mtu)
1639 {
1640         int ret = -1;
1641         char buf[64];
1642         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1643                         dev->ifname);
1644
1645         int fd = open(buf, O_RDWR);
1646         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1647         if (len < 0)
1648                 goto out;
1649
1650         buf[len] = 0;
1651         ret = atoi(buf);
1652
1653         if (!mtu || ret <= mtu)
1654                 goto out;
1655
1656         lseek(fd, 0, SEEK_SET);
1657         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1658                 ret = -1;
1659
1660 out:
1661         close(fd);
1662         return ret;
1663 }
1664
1665 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1666 {
1667         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1668         struct blob_attr *cur;
1669         bool set_df = true;
1670         const char *str;
1671
1672         system_del_ip_tunnel(name);
1673
1674         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1675                 blob_data(attr), blob_len(attr));
1676
1677         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1678                 return -EINVAL;
1679         str = blobmsg_data(cur);
1680
1681         if ((cur = tb[TUNNEL_ATTR_DF]))
1682                 set_df = blobmsg_get_bool(cur);
1683
1684         unsigned int ttl = 0;
1685         if ((cur = tb[TUNNEL_ATTR_TTL])) {
1686                 ttl = blobmsg_get_u32(cur);
1687                 if (ttl > 255 || (!set_df && ttl))
1688                         return -EINVAL;
1689         }
1690
1691         unsigned int link = 0;
1692         if ((cur = tb[TUNNEL_ATTR_LINK])) {
1693                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
1694                 if (!iface)
1695                         return -EINVAL;
1696
1697                 if (iface->l3_dev.dev)
1698                         link = iface->l3_dev.dev->ifindex;
1699         }
1700
1701         if (!strcmp(str, "sit")) {
1702                 struct ip_tunnel_parm p = {
1703                         .link = link,
1704                         .iph = {
1705                                 .version = 4,
1706                                 .ihl = 5,
1707                                 .frag_off = set_df ? htons(IP_DF) : 0,
1708                                 .protocol = IPPROTO_IPV6,
1709                                 .ttl = ttl
1710                         }
1711                 };
1712
1713                 if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1714                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1715                         return -EINVAL;
1716
1717                 if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1718                                 inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1719                         return -EINVAL;
1720
1721                 strncpy(p.name, name, sizeof(p.name));
1722                 if (tunnel_ioctl("sit0", SIOCADDTUNNEL, &p) < 0)
1723                         return -1;
1724
1725 #ifdef SIOCADD6RD
1726                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
1727                         unsigned int mask;
1728                         struct ip_tunnel_6rd p6;
1729
1730                         memset(&p6, 0, sizeof(p6));
1731
1732                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
1733                                                 &p6.prefix, &mask) || mask > 128)
1734                                 return -EINVAL;
1735                         p6.prefixlen = mask;
1736
1737                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
1738                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
1739                                                         &p6.relay_prefix, &mask) || mask > 32)
1740                                         return -EINVAL;
1741                                 p6.relay_prefixlen = mask;
1742                         }
1743
1744                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
1745                                 system_del_ip_tunnel(name);
1746                                 return -1;
1747                         }
1748                 }
1749 #endif
1750         } else if (!strcmp(str, "ipip6")) {
1751                 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
1752                                 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1753
1754                 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
1755                 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1756                 nla_put_string(nlm, IFLA_IFNAME, name);
1757
1758                 if (link)
1759                         nla_put_u32(nlm, IFLA_LINK, link);
1760
1761                 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1762                 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
1763                 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1764
1765                 if (link)
1766                         nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
1767
1768                 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
1769                 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
1770                 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
1771
1772                 struct in6_addr in6buf;
1773                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1774                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1)
1775                                 return -EINVAL;
1776                         nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
1777                 }
1778
1779                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1780                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1)
1781                                 return -EINVAL;
1782                         nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
1783                 }
1784
1785 #ifdef IFLA_IPTUN_FMR_MAX
1786                 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
1787                         struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
1788
1789                         struct blob_attr *fmr;
1790                         unsigned rem, fmrcnt = 0;
1791                         blobmsg_for_each_attr(fmr, cur, rem) {
1792                                 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
1793                                         continue;
1794
1795                                 unsigned ip4len, ip6len, ealen, offset = 6;
1796                                 char ip6buf[48];
1797                                 char ip4buf[16];
1798
1799                                 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
1800                                                 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5)
1801                                         return -EINVAL;
1802
1803                                 struct in6_addr ip6prefix;
1804                                 struct in_addr ip4prefix;
1805                                 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
1806                                                 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1)
1807                                         return -EINVAL;
1808
1809                                 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
1810
1811                                 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
1812                                 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
1813                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
1814                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
1815                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
1816                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
1817
1818                                 nla_nest_end(nlm, rule);
1819                         }
1820
1821                         nla_nest_end(nlm, fmrs);
1822                 }
1823 #endif
1824
1825                 nla_nest_end(nlm, infodata);
1826                 nla_nest_end(nlm, linkinfo);
1827
1828                 return system_rtnl_call(nlm);
1829         } else
1830                 return -EINVAL;
1831
1832         return 0;
1833 }