system-linux.c

   1 /*
   2  * netifd - network interface daemon
   3  * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License version 2
   7  * as published by the Free Software Foundation
   8  *
   9  * This program is distributed in the hope that it will be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12  * GNU General Public License for more details.
  13  */
  14 #define _GNU_SOURCE
  15
  16 #include <sys/socket.h>
  17 #include <sys/ioctl.h>
  18 #include <sys/stat.h>
  19 #include <sys/syscall.h>
  20
  21 #include <net/if.h>
  22 #include <net/if_arp.h>
  23
  24 #include <arpa/inet.h>
  25 #include <netinet/in.h>
  26
  27 #include <linux/rtnetlink.h>
  28 #include <linux/sockios.h>
  29 #include <linux/ip.h>
  30 #include <linux/if_vlan.h>
  31 #include <linux/if_bridge.h>
  32 #include <linux/if_tunnel.h>
  33 #include <linux/ethtool.h>
  34
  35 #include <unistd.h>
  36 #include <string.h>
  37 #include <fcntl.h>
  38 #include <glob.h>
  39 #include <time.h>
  40
  41 #include <netlink/msg.h>
  42 #include <netlink/attr.h>
  43 #include <netlink/socket.h>
  44 #include <libubox/uloop.h>
  45
  46 #include "netifd.h"
  47 #include "device.h"
  48 #include "system.h"
  49
  50 struct event_socket {
  51         struct uloop_fd uloop;
  52         struct nl_sock *sock;
  53         struct nl_cb *cb;
  54 };
  55
  56 static int sock_ioctl = -1;
  57 static struct nl_sock *sock_rtnl = NULL;
  58
  59 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
  60 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
  61
  62 static char dev_buf[256];
  63
  64 static void
  65 handler_nl_event(struct uloop_fd *u, unsigned int events)
  66 {
  67         struct event_socket *ev = container_of(u, struct event_socket, uloop);
  68         nl_recvmsgs(ev->sock, ev->cb);
  69 }
  70
  71 static struct nl_sock *
  72 create_socket(int protocol, int groups)
  73 {
  74         struct nl_sock *sock;
  75
  76         sock = nl_socket_alloc();
  77         if (!sock)
  78                 return NULL;
  79
  80         if (groups)
  81                 nl_join_groups(sock, groups);
  82
  83         if (nl_connect(sock, protocol))
  84                 return NULL;
  85
  86         return sock;
  87 }
  88
  89 static bool
  90 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
  91                         uloop_fd_handler cb)
  92 {
  93         ev->sock = create_socket(protocol, groups);
  94         if (!ev->sock)
  95                 return false;
  96
  97         ev->uloop.fd = nl_socket_get_fd(ev->sock);
  98         ev->uloop.cb = cb;
  99         uloop_fd_add(&ev->uloop, ULOOP_READ | ULOOP_EDGE_TRIGGER);
 100         return true;
 101 }
 102
 103 static bool
 104 create_event_socket(struct event_socket *ev, int protocol,
 105                     int (*cb)(struct nl_msg *msg, void *arg))
 106 {
 107         // Prepare socket for link events
 108         ev->cb = nl_cb_alloc(NL_CB_DEFAULT);
 109         if (!ev->cb)
 110                 return false;
 111
 112         nl_cb_set(ev->cb, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
 113
 114         return create_raw_event_socket(ev, protocol, 0, handler_nl_event);
 115 }
 116
 117 int system_init(void)
 118 {
 119         static struct event_socket rtnl_event;
 120         static struct event_socket hotplug_event;
 121
 122         sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
 123         fcntl(sock_ioctl, F_SETFD, fcntl(sock_ioctl, F_GETFD) | FD_CLOEXEC);
 124
 125         // Prepare socket for routing / address control
 126         sock_rtnl = create_socket(NETLINK_ROUTE, 0);
 127         if (!sock_rtnl)
 128                 return -1;
 129
 130         if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
 131                 return -1;
 132
 133         if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
 134                                      handle_hotplug_event))
 135                 return -1;
 136
 137         // Receive network link events form kernel
 138         nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
 139
 140         return 0;
 141 }
 142
 143 static void system_set_sysctl(const char *path, const char *val)
 144 {
 145         int fd;
 146
 147         fd = open(path, O_WRONLY);
 148         if (fd < 0)
 149                 return;
 150
 151         write(fd, val, strlen(val));
 152         close(fd);
 153 }
 154
 155 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
 156 {
 157         snprintf(dev_buf, sizeof(dev_buf), path, device);
 158         system_set_sysctl(dev_buf, val);
 159 }
 160
 161 static void system_set_disable_ipv6(struct device *dev, const char *val)
 162 {
 163         system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
 164 }
 165
 166 // Evaluate netlink messages
 167 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 168 {
 169         struct nlmsghdr *nh = nlmsg_hdr(msg);
 170         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 171         struct nlattr *nla[__IFLA_MAX];
 172
 173         if (nh->nlmsg_type != RTM_DELLINK && nh->nlmsg_type != RTM_NEWLINK)
 174                 goto out;
 175
 176         nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
 177         if (!nla[IFLA_IFNAME])
 178                 goto out;
 179
 180         struct device *dev = device_get(RTA_DATA(nla[IFLA_IFNAME]), false);
 181         if (!dev)
 182                 goto out;
 183
 184         dev->ifindex = ifi->ifi_index;
 185         /* TODO: parse link status */
 186
 187 out:
 188         return 0;
 189 }
 190
 191 static void
 192 handle_hotplug_msg(char *data, int size)
 193 {
 194         const char *subsystem = NULL, *interface = NULL;
 195         char *cur, *end, *sep;
 196         struct device *dev;
 197         int skip;
 198         bool add;
 199
 200         if (!strncmp(data, "add@", 4))
 201                 add = true;
 202         else if (!strncmp(data, "remove@", 7))
 203                 add = false;
 204         else
 205                 return;
 206
 207         skip = strlen(data) + 1;
 208         end = data + size;
 209
 210         for (cur = data + skip; cur < end; cur += skip) {
 211                 skip = strlen(cur) + 1;
 212
 213                 sep = strchr(cur, '=');
 214                 if (!sep)
 215                         continue;
 216
 217                 *sep = 0;
 218                 if (!strcmp(cur, "INTERFACE"))
 219                         interface = sep + 1;
 220                 else if (!strcmp(cur, "SUBSYSTEM")) {
 221                         subsystem = sep + 1;
 222                         if (strcmp(subsystem, "net") != 0)
 223                                 return;
 224                 }
 225                 if (subsystem && interface)
 226                         goto found;
 227         }
 228         return;
 229
 230 found:
 231         dev = device_get(interface, false);
 232         if (!dev)
 233                 return;
 234
 235         if (dev->type != &simple_device_type)
 236                 return;
 237
 238         if (add && system_if_force_external(dev->ifname))
 239                 return;
 240
 241         device_set_present(dev, add);
 242 }
 243
 244 static void
 245 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
 246 {
 247         struct event_socket *ev = container_of(u, struct event_socket, uloop);
 248         struct sockaddr_nl nla;
 249         unsigned char *buf = NULL;
 250         int size;
 251
 252         while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
 253                 if (nla.nl_pid == 0)
 254                         handle_hotplug_msg((char *) buf, size);
 255
 256                 free(buf);
 257         }
 258 }
 259
 260 static int system_rtnl_call(struct nl_msg *msg)
 261 {
 262         int ret;
 263
 264         ret = nl_send_auto_complete(sock_rtnl, msg);
 265         nlmsg_free(msg);
 266
 267         if (ret < 0)
 268                 return ret;
 269
 270         return nl_wait_for_ack(sock_rtnl);
 271 }
 272
 273 int system_bridge_delbr(struct device *bridge)
 274 {
 275         return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 276 }
 277
 278 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 279 {
 280         struct ifreq ifr;
 281
 282         memset(&ifr, 0, sizeof(ifr));
 283         if (dev)
 284                 ifr.ifr_ifindex = dev->ifindex;
 285         else
 286                 ifr.ifr_data = data;
 287         strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
 288         return ioctl(sock_ioctl, cmd, &ifr);
 289 }
 290
 291 static bool system_is_bridge(const char *name, char *buf, int buflen)
 292 {
 293         struct stat st;
 294
 295         snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
 296         if (stat(buf, &st) < 0)
 297                 return false;
 298
 299         return true;
 300 }
 301
 302 static char *system_get_bridge(const char *name, char *buf, int buflen)
 303 {
 304         char *path;
 305         ssize_t len;
 306         glob_t gl;
 307
 308         snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
 309         if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
 310                 return NULL;
 311
 312         if (gl.gl_pathc == 0)
 313                 return NULL;
 314
 315         len = readlink(gl.gl_pathv[0], buf, buflen);
 316         if (len < 0)
 317                 return NULL;
 318
 319         buf[len] = 0;
 320         path = strrchr(buf, '/');
 321         if (!path)
 322                 return NULL;
 323
 324         return path + 1;
 325 }
 326
 327 int system_bridge_addif(struct device *bridge, struct device *dev)
 328 {
 329         char *oldbr;
 330
 331         system_set_disable_ipv6(dev, "1");
 332         oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
 333         if (oldbr && !strcmp(oldbr, bridge->ifname))
 334                 return 0;
 335
 336         return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 337 }
 338
 339 int system_bridge_delif(struct device *bridge, struct device *dev)
 340 {
 341         system_set_disable_ipv6(dev, "0");
 342         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 343 }
 344
 345 static int system_if_resolve(struct device *dev)
 346 {
 347         struct ifreq ifr;
 348         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 349         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 350                 return ifr.ifr_ifindex;
 351         else
 352                 return 0;
 353 }
 354
 355 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 356 {
 357         struct ifreq ifr;
 358
 359         memset(&ifr, 0, sizeof(ifr));
 360         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 361         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 362         ifr.ifr_flags |= add;
 363         ifr.ifr_flags &= ~rem;
 364         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 365 }
 366
 367 struct clear_data {
 368         struct nl_msg *msg;
 369         struct device *dev;
 370         int type;
 371         int size;
 372         int af;
 373 };
 374
 375
 376 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 377 {
 378         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 379
 380         return ifa->ifa_index == ifindex;
 381 }
 382
 383 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 384 {
 385         struct rtmsg *r = NLMSG_DATA(hdr);
 386         struct nlattr *tb[__RTA_MAX];
 387
 388         if (r->rtm_protocol == RTPROT_KERNEL &&
 389             r->rtm_family == AF_INET6)
 390                 return false;
 391
 392         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 393         if (!tb[RTA_OIF])
 394                 return false;
 395
 396         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 397 }
 398
 399 static int cb_clear_event(struct nl_msg *msg, void *arg)
 400 {
 401         struct clear_data *clr = arg;
 402         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 403         bool (*cb)(struct nlmsghdr *, int ifindex);
 404         int type;
 405
 406         switch(clr->type) {
 407         case RTM_GETADDR:
 408                 type = RTM_DELADDR;
 409                 if (hdr->nlmsg_type != RTM_NEWADDR)
 410                         return NL_SKIP;
 411
 412                 cb = check_ifaddr;
 413                 break;
 414         case RTM_GETROUTE:
 415                 type = RTM_DELROUTE;
 416                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 417                         return NL_SKIP;
 418
 419                 cb = check_route;
 420                 break;
 421         default:
 422                 return NL_SKIP;
 423         }
 424
 425         if (!cb(hdr, clr->dev->ifindex))
 426                 return NL_SKIP;
 427
 428         D(SYSTEM, "Remove %s from device %s\n",
 429           type == RTM_DELADDR ? "an address" : "a route",
 430           clr->dev->ifname);
 431         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 432         hdr = nlmsg_hdr(clr->msg);
 433         hdr->nlmsg_type = type;
 434         hdr->nlmsg_flags = NLM_F_REQUEST;
 435
 436         if (!nl_send_auto_complete(sock_rtnl, clr->msg))
 437                 nl_wait_for_ack(sock_rtnl);
 438
 439         return NL_SKIP;
 440 }
 441
 442 static int
 443 cb_finish_event(struct nl_msg *msg, void *arg)
 444 {
 445         int *pending = arg;
 446         *pending = 0;
 447         return NL_STOP;
 448 }
 449
 450 static int
 451 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 452 {
 453         int *pending = arg;
 454         *pending = err->error;
 455         return NL_STOP;
 456 }
 457
 458 static void
 459 system_if_clear_entries(struct device *dev, int type, int af)
 460 {
 461         struct clear_data clr;
 462         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 463         struct rtmsg rtm = {
 464                 .rtm_family = af,
 465                 .rtm_flags = RTM_F_CLONED,
 466         };
 467         int flags = NLM_F_DUMP;
 468         int pending = 1;
 469
 470         clr.af = af;
 471         clr.dev = dev;
 472         clr.type = type;
 473         switch (type) {
 474         case RTM_GETADDR:
 475                 clr.size = sizeof(struct rtgenmsg);
 476                 break;
 477         case RTM_GETROUTE:
 478                 clr.size = sizeof(struct rtmsg);
 479                 break;
 480         default:
 481                 return;
 482         }
 483
 484         if (!cb)
 485                 return;
 486
 487         clr.msg = nlmsg_alloc_simple(type, flags);
 488         if (!clr.msg)
 489                 goto out;
 490
 491         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 492         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 493         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 494         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 495
 496         nl_send_auto_complete(sock_rtnl, clr.msg);
 497         while (pending > 0)
 498                 nl_recvmsgs(sock_rtnl, cb);
 499
 500         nlmsg_free(clr.msg);
 501 out:
 502         nl_cb_put(cb);
 503 }
 504
 505 /*
 506  * Clear bridge (membership) state and bring down device
 507  */
 508 void system_if_clear_state(struct device *dev)
 509 {
 510         static char buf[256];
 511         char *bridge;
 512
 513         if (dev->external)
 514                 return;
 515
 516         dev->ifindex = system_if_resolve(dev);
 517         if (!dev->ifindex)
 518                 return;
 519
 520         system_if_flags(dev->ifname, 0, IFF_UP);
 521
 522         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 523                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 524                 system_bridge_delbr(dev);
 525                 return;
 526         }
 527
 528         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 529         if (bridge) {
 530                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 531                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 532         }
 533
 534         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 535         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 536         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 537         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 538         system_set_disable_ipv6(dev, "0");
 539 }
 540
 541 static inline unsigned long
 542 sec_to_jiffies(int val)
 543 {
 544         return (unsigned long) val * 100;
 545 }
 546
 547 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 548 {
 549         unsigned long args[4] = {};
 550
 551         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 552                 return -1;
 553
 554         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 555         args[1] = !!cfg->stp;
 556         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 557
 558         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 559         args[1] = sec_to_jiffies(cfg->forward_delay);
 560         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 561
 562         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 563                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 564
 565         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 566                 args[0] = BRCTL_SET_AGEING_TIME;
 567                 args[1] = sec_to_jiffies(cfg->ageing_time);
 568                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 569         }
 570
 571         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 572                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 573                 args[1] = sec_to_jiffies(cfg->hello_time);
 574                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 575         }
 576
 577         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 578                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 579                 args[1] = sec_to_jiffies(cfg->max_age);
 580                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 581         }
 582
 583         return 0;
 584 }
 585
 586 static int system_vlan(struct device *dev, int id)
 587 {
 588         struct vlan_ioctl_args ifr = {
 589                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 590                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 591         };
 592
 593         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 594
 595         if (id < 0) {
 596                 ifr.cmd = DEL_VLAN_CMD;
 597                 ifr.u.VID = 0;
 598         } else {
 599                 ifr.cmd = ADD_VLAN_CMD;
 600                 ifr.u.VID = id;
 601         }
 602         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 603         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 604 }
 605
 606 int system_vlan_add(struct device *dev, int id)
 607 {
 608         return system_vlan(dev, id);
 609 }
 610
 611 int system_vlan_del(struct device *dev)
 612 {
 613         return system_vlan(dev, -1);
 614 }
 615
 616 static void
 617 system_if_get_settings(struct device *dev, struct device_settings *s)
 618 {
 619         struct ifreq ifr;
 620
 621         memset(&ifr, 0, sizeof(ifr));
 622         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 623
 624         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 625                 s->mtu = ifr.ifr_mtu;
 626                 s->flags |= DEV_OPT_MTU;
 627         }
 628
 629         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 630                 s->txqueuelen = ifr.ifr_qlen;
 631                 s->flags |= DEV_OPT_TXQUEUELEN;
 632         }
 633
 634         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 635                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 636                 s->flags |= DEV_OPT_MACADDR;
 637         }
 638 }
 639
 640 void
 641 system_if_apply_settings(struct device *dev, struct device_settings *s)
 642 {
 643         struct ifreq ifr;
 644
 645         memset(&ifr, 0, sizeof(ifr));
 646         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 647         if (s->flags & DEV_OPT_MTU) {
 648                 ifr.ifr_mtu = s->mtu;
 649                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 650                         s->flags &= ~DEV_OPT_MTU;
 651         }
 652         if (s->flags & DEV_OPT_TXQUEUELEN) {
 653                 ifr.ifr_qlen = s->txqueuelen;
 654                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 655                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 656         }
 657         if ((s->flags & DEV_OPT_MACADDR) && !dev->external) {
 658                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 659                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 660                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 661                         s->flags &= ~DEV_OPT_MACADDR;
 662         }
 663 }
 664
 665 int system_if_up(struct device *dev)
 666 {
 667         system_if_get_settings(dev, &dev->orig_settings);
 668         system_if_apply_settings(dev, &dev->settings);
 669         dev->ifindex = system_if_resolve(dev);
 670         return system_if_flags(dev->ifname, IFF_UP, 0);
 671 }
 672
 673 int system_if_down(struct device *dev)
 674 {
 675         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
 676         dev->orig_settings.flags &= dev->settings.flags;
 677         system_if_apply_settings(dev, &dev->orig_settings);
 678         return ret;
 679 }
 680
 681 int system_if_check(struct device *dev)
 682 {
 683         device_set_present(dev, (system_if_resolve(dev) > 0));
 684         return 0;
 685 }
 686
 687 struct device *
 688 system_if_get_parent(struct device *dev)
 689 {
 690         char buf[64], *devname;
 691         int ifindex, iflink, len;
 692         FILE *f;
 693
 694         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
 695         f = fopen(buf, "r");
 696         if (!f)
 697                 return NULL;
 698
 699         len = fread(buf, 1, sizeof(buf) - 1, f);
 700         fclose(f);
 701
 702         if (len <= 0)
 703                 return NULL;
 704
 705         buf[len] = 0;
 706         iflink = strtoul(buf, NULL, 0);
 707         ifindex = system_if_resolve(dev);
 708         if (!iflink || iflink == ifindex)
 709                 return NULL;
 710
 711         devname = if_indextoname(iflink, buf);
 712         if (!devname)
 713                 return NULL;
 714
 715         return device_get(devname, true);
 716 }
 717
 718 static bool
 719 read_string_file(int dir_fd, const char *file, char *buf, int len)
 720 {
 721         bool ret = false;
 722         char *c;
 723         int fd;
 724
 725         fd = openat(dir_fd, file, O_RDONLY);
 726         if (fd < 0)
 727                 return false;
 728
 729 retry:
 730         len = read(fd, buf, len - 1);
 731         if (len < 0) {
 732                 if (errno == EINTR)
 733                         goto retry;
 734         } else if (len > 0) {
 735                         buf[len] = 0;
 736
 737                         c = strchr(buf, '\n');
 738                         if (c)
 739                                 *c = 0;
 740
 741                         ret = true;
 742         }
 743
 744         close(fd);
 745
 746         return ret;
 747 }
 748
 749 static bool
 750 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
 751 {
 752         char buf[64];
 753         bool ret = false;
 754
 755         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
 756         if (ret)
 757                 *val = strtoull(buf, NULL, 0);
 758
 759         return ret;
 760 }
 761
 762 /* Assume advertised flags == supported flags */
 763 static const struct {
 764         uint32_t mask;
 765         const char *name;
 766 } ethtool_link_modes[] = {
 767         { ADVERTISED_10baseT_Half, "10H" },
 768         { ADVERTISED_10baseT_Full, "10F" },
 769         { ADVERTISED_100baseT_Half, "100H" },
 770         { ADVERTISED_100baseT_Full, "100F" },
 771         { ADVERTISED_1000baseT_Half, "1000H" },
 772         { ADVERTISED_1000baseT_Full, "1000F" },
 773 };
 774
 775 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
 776 {
 777         int i;
 778         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
 779                 if (mask & ethtool_link_modes[i].mask)
 780                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
 781         }
 782 }
 783
 784 bool
 785 system_if_force_external(const char *ifname)
 786 {
 787         char buf[64];
 788         struct stat s;
 789
 790         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
 791         return stat(buf, &s) == 0;
 792 }
 793
 794 int
 795 system_if_dump_info(struct device *dev, struct blob_buf *b)
 796 {
 797         struct ethtool_cmd ecmd;
 798         struct ifreq ifr;
 799         char buf[64], *s;
 800         void *c;
 801         int dir_fd;
 802         uint64_t val = 0;
 803
 804         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
 805         dir_fd = open(buf, O_DIRECTORY);
 806
 807         if (read_uint64_file(dir_fd, "carrier", &val))
 808                 blobmsg_add_u8(b, "link", !!val);
 809
 810         memset(&ecmd, 0, sizeof(ecmd));
 811         memset(&ifr, 0, sizeof(ifr));
 812         strcpy(ifr.ifr_name, dev->ifname);
 813         ifr.ifr_data = (caddr_t) &ecmd;
 814         ecmd.cmd = ETHTOOL_GSET;
 815
 816         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
 817                 c = blobmsg_open_array(b, "link-advertising");
 818                 system_add_link_modes(b, ecmd.advertising);
 819                 blobmsg_close_array(b, c);
 820
 821                 c = blobmsg_open_array(b, "link-supported");
 822                 system_add_link_modes(b, ecmd.supported);
 823                 blobmsg_close_array(b, c);
 824
 825                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
 826                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
 827                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
 828                 blobmsg_add_string_buffer(b);
 829         }
 830
 831         close(dir_fd);
 832         return 0;
 833 }
 834
 835 int
 836 system_if_dump_stats(struct device *dev, struct blob_buf *b)
 837 {
 838         const char *const counters[] = {
 839                 "collisions",     "rx_frame_errors",   "tx_compressed",
 840                 "multicast",      "rx_length_errors",  "tx_dropped",
 841                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
 842                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
 843                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
 844                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
 845                 "rx_errors",      "tx_bytes",          "tx_window_errors",
 846                 "rx_fifo_errors", "tx_carrier_errors",
 847         };
 848         char buf[64];
 849         int stats_dir;
 850         int i;
 851         uint64_t val = 0;
 852
 853         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
 854         stats_dir = open(buf, O_DIRECTORY);
 855         if (stats_dir < 0)
 856                 return -1;
 857
 858         for (i = 0; i < ARRAY_SIZE(counters); i++)
 859                 if (read_uint64_file(stats_dir, counters[i], &val))
 860                         blobmsg_add_u64(b, counters[i], val);
 861
 862         close(stats_dir);
 863         return 0;
 864 }
 865
 866 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
 867 {
 868         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
 869         int alen = v4 ? 4 : 16;
 870         unsigned int flags = 0;
 871         struct ifaddrmsg ifa = {
 872                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
 873                 .ifa_prefixlen = addr->mask,
 874                 .ifa_index = dev->ifindex,
 875         };
 876
 877         struct nl_msg *msg;
 878         if (cmd == RTM_NEWADDR)
 879                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
 880
 881         msg = nlmsg_alloc_simple(cmd, flags);
 882         if (!msg)
 883                 return -1;
 884
 885         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
 886         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
 887         if (v4) {
 888                 if (addr->broadcast)
 889                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
 890                 if (addr->point_to_point)
 891                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
 892         } else {
 893                 time_t now = system_get_rtime();
 894                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
 895
 896                 if (addr->preferred_until) {
 897                         int preferred = addr->preferred_until - now;
 898                         if (preferred < 0)
 899                                 preferred = 0;
 900
 901                         cinfo.ifa_prefered = preferred;
 902                 }
 903
 904                 if (addr->valid_until) {
 905                         int valid = addr->valid_until - now;
 906                         if (valid <= 0)
 907                                 return -1;
 908
 909                         cinfo.ifa_valid = valid;
 910                 }
 911
 912                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
 913         }
 914
 915         return system_rtnl_call(msg);
 916 }
 917
 918 int system_add_address(struct device *dev, struct device_addr *addr)
 919 {
 920         return system_addr(dev, addr, RTM_NEWADDR);
 921 }
 922
 923 int system_del_address(struct device *dev, struct device_addr *addr)
 924 {
 925         return system_addr(dev, addr, RTM_DELADDR);
 926 }
 927
 928 static int system_rt(struct device *dev, struct device_route *route, int cmd)
 929 {
 930         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
 931         bool have_gw;
 932         unsigned int flags = 0;
 933
 934         if (alen == 4)
 935                 have_gw = !!route->nexthop.in.s_addr;
 936         else
 937                 have_gw = route->nexthop.in6.s6_addr32[0] ||
 938                         route->nexthop.in6.s6_addr32[1] ||
 939                         route->nexthop.in6.s6_addr32[2] ||
 940                         route->nexthop.in6.s6_addr32[3];
 941
 942         unsigned char scope = (cmd == RTM_DELROUTE) ? RT_SCOPE_NOWHERE :
 943                         (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
 944
 945         unsigned int table = (route->flags & DEVROUTE_TABLE) ? route->table : RT_TABLE_MAIN;
 946
 947         struct rtmsg rtm = {
 948                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
 949                 .rtm_dst_len = route->mask,
 950                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
 951                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
 952                 .rtm_scope = scope,
 953                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
 954         };
 955         struct nl_msg *msg;
 956
 957         if (cmd == RTM_NEWROUTE) {
 958                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
 959
 960                 if (!dev) { // Add null-route
 961                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
 962                         rtm.rtm_type = RTN_UNREACHABLE;
 963                 }
 964         }
 965
 966         msg = nlmsg_alloc_simple(cmd, flags);
 967         if (!msg)
 968                 return -1;
 969
 970         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
 971
 972         if (route->mask)
 973                 nla_put(msg, RTA_DST, alen, &route->addr);
 974
 975         if (route->metric > 0)
 976                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
 977
 978         if (have_gw)
 979                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
 980
 981         if (dev)
 982                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
 983
 984         if (table >= 256)
 985                 nla_put_u32(msg, RTA_TABLE, table);
 986
 987         return system_rtnl_call(msg);
 988 }
 989
 990 int system_add_route(struct device *dev, struct device_route *route)
 991 {
 992         return system_rt(dev, route, RTM_NEWROUTE);
 993 }
 994
 995 int system_del_route(struct device *dev, struct device_route *route)
 996 {
 997         return system_rt(dev, route, RTM_DELROUTE);
 998 }
 999
1000 int system_flush_routes(void)
1001 {
1002         const char *names[] = {
1003                 "/proc/sys/net/ipv4/route/flush",
1004                 "/proc/sys/net/ipv6/route/flush"
1005         };
1006         int fd, i;
1007
1008         for (i = 0; i < ARRAY_SIZE(names); i++) {
1009                 fd = open(names[i], O_WRONLY);
1010                 if (fd < 0)
1011                         continue;
1012
1013                 write(fd, "-1", 2);
1014                 close(fd);
1015         }
1016         return 0;
1017 }
1018
1019 bool system_resolve_rt_table(const char *name, unsigned int *id)
1020 {
1021         FILE *f;
1022         char *e, buf[128];
1023         unsigned int n, table = RT_TABLE_UNSPEC;
1024
1025         /* first try to parse table as number */
1026         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1027                 table = n;
1028
1029         /* handle well known aliases */
1030         else if (!strcmp(name, "default"))
1031                 table = RT_TABLE_DEFAULT;
1032         else if (!strcmp(name, "main"))
1033                 table = RT_TABLE_MAIN;
1034         else if (!strcmp(name, "local"))
1035                 table = RT_TABLE_LOCAL;
1036
1037         /* try to look up name in /etc/iproute2/rt_tables */
1038         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1039         {
1040                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1041                 {
1042                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1043                                 continue;
1044
1045                         n = strtoul(e, NULL, 10);
1046                         e = strtok(NULL, " \t\n");
1047
1048                         if (e && !strcmp(e, name))
1049                         {
1050                                 table = n;
1051                                 break;
1052                         }
1053                 }
1054
1055                 fclose(f);
1056         }
1057
1058         if (table == RT_TABLE_UNSPEC)
1059                 return false;
1060
1061         /* do not consider main table special */
1062         if (table == RT_TABLE_MAIN)
1063                 table = RT_TABLE_UNSPEC;
1064
1065         *id = table;
1066         return true;
1067 }
1068
1069 time_t system_get_rtime(void)
1070 {
1071         struct timespec ts;
1072         struct timeval tv;
1073
1074         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1075                 return ts.tv_sec;
1076
1077         if (gettimeofday(&tv, NULL) == 0)
1078                 return tv.tv_sec;
1079
1080         return 0;
1081 }
1082
1083 #ifndef IP_DF
1084 #define IP_DF       0x4000
1085 #endif
1086
1087 static void tunnel_parm_init(struct ip_tunnel_parm *p)
1088 {
1089         memset(p, 0, sizeof(*p));
1090         p->iph.version = 4;
1091         p->iph.ihl = 5;
1092         p->iph.frag_off = htons(IP_DF);
1093 }
1094
1095 static int tunnel_ioctl(const char *name, int cmd, void *p)
1096 {
1097         struct ifreq ifr;
1098
1099         memset(&ifr, 0, sizeof(ifr));
1100         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1101         ifr.ifr_ifru.ifru_data = p;
1102         return ioctl(sock_ioctl, cmd, &ifr);
1103 }
1104
1105 int system_del_ip_tunnel(const char *name)
1106 {
1107         struct ip_tunnel_parm p;
1108
1109         tunnel_parm_init(&p);
1110         return tunnel_ioctl(name, SIOCDELTUNNEL, &p);
1111 }
1112
1113 int system_update_ipv6_mtu(struct device *dev, int mtu)
1114 {
1115         int ret = -1;
1116         char buf[64];
1117         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1118                         dev->ifname);
1119
1120         int fd = open(buf, O_RDWR);
1121         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1122         if (len < 0)
1123                 goto out;
1124
1125         buf[len] = 0;
1126         ret = atoi(buf);
1127
1128         if (!mtu || ret <= mtu)
1129                 goto out;
1130
1131         lseek(fd, 0, SEEK_SET);
1132         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1133                 ret = -1;
1134
1135 out:
1136         close(fd);
1137         return ret;
1138 }
1139
1140 static int parse_ipaddr(struct blob_attr *attr, __be32 *addr)
1141 {
1142         if (!attr)
1143                 return 1;
1144
1145         return inet_pton(AF_INET, blobmsg_data(attr), (void *) addr);
1146 }
1147
1148
1149 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1150 {
1151         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1152         struct blob_attr *cur;
1153         struct ip_tunnel_parm p;
1154         const char *base, *str;
1155         bool is_sit;
1156
1157         system_del_ip_tunnel(name);
1158
1159         tunnel_parm_init(&p);
1160
1161         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1162                 blob_data(attr), blob_len(attr));
1163
1164         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1165                 return -EINVAL;
1166         str = blobmsg_data(cur);
1167         is_sit = !strcmp(str, "sit");
1168
1169         if (is_sit) {
1170                 p.iph.protocol = IPPROTO_IPV6;
1171                 base = "sit0";
1172         } else
1173                 return -EINVAL;
1174
1175         if (!parse_ipaddr(tb[TUNNEL_ATTR_LOCAL], &p.iph.saddr))
1176                 return -EINVAL;
1177
1178         if (!parse_ipaddr(tb[TUNNEL_ATTR_REMOTE], &p.iph.daddr))
1179                 return -EINVAL;
1180
1181         if ((cur = tb[TUNNEL_ATTR_TTL])) {
1182                 unsigned int val = blobmsg_get_u32(cur);
1183
1184                 if (val > 255)
1185                         return -EINVAL;
1186
1187                 p.iph.ttl = val;
1188         }
1189
1190         strncpy(p.name, name, sizeof(p.name));
1191         if (tunnel_ioctl(base, SIOCADDTUNNEL, &p) < 0)
1192                 return -1;
1193
1194 #ifdef SIOCADD6RD
1195         cur = tb[TUNNEL_ATTR_6RD_PREFIX];
1196         if (cur && is_sit) {
1197                 unsigned int mask;
1198                 struct ip_tunnel_6rd p6;
1199
1200                 memset(&p6, 0, sizeof(p6));
1201
1202                 if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
1203                                         &p6.prefix, &mask) || mask > 128)
1204                         return -EINVAL;
1205                 p6.prefixlen = mask;
1206
1207                 if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
1208                         if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
1209                                                 &p6.relay_prefix, &mask) || mask > 32)
1210                                 return -EINVAL;
1211                         p6.relay_prefixlen = mask;
1212                 }
1213
1214                 if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
1215                         system_del_ip_tunnel(name);
1216                         return -1;
1217                 }
1218         }
1219 #endif
1220
1221         return 0;
1222 }