system-linux.c

   1 #define _GNU_SOURCE
   2
   3 #include <sys/socket.h>
   4 #include <sys/ioctl.h>
   5 #include <sys/stat.h>
   6 #include <sys/syscall.h>
   7
   8 #include <linux/rtnetlink.h>
   9 #include <linux/sockios.h>
  10 #include <linux/if_vlan.h>
  11 #include <linux/if_bridge.h>
  12
  13 #include <unistd.h>
  14 #include <string.h>
  15 #include <fcntl.h>
  16 #include <glob.h>
  17
  18 #include <netlink/msg.h>
  19 #include <netlink/attr.h>
  20 #include <netlink/socket.h>
  21 #include <libubox/uloop.h>
  22
  23 #include "netifd.h"
  24 #include "device.h"
  25 #include "system.h"
  26
  27 static int sock_ioctl = -1;
  28 static struct nl_sock *sock_rtnl = NULL;
  29 static struct nl_sock *sock_rtnl_event = NULL;
  30
  31 static void handler_rtnl_event(struct uloop_fd *u, unsigned int events);
  32 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
  33 static struct uloop_fd rtnl_event = {.cb = handler_rtnl_event};
  34 static struct nl_cb *nl_cb_rtnl_event;
  35
  36 int system_init(void)
  37 {
  38         sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
  39         fcntl(sock_ioctl, F_SETFD, fcntl(sock_ioctl, F_GETFD) | FD_CLOEXEC);
  40
  41         // Prepare socket for routing / address control
  42         sock_rtnl = nl_socket_alloc();
  43         if (!sock_rtnl)
  44                 return -1;
  45
  46         if (nl_connect(sock_rtnl, NETLINK_ROUTE))
  47                 goto error_free_sock;
  48
  49         // Prepare socket for link events
  50         nl_cb_rtnl_event = nl_cb_alloc(NL_CB_DEFAULT);
  51         if (!nl_cb_rtnl_event)
  52                 goto error_free_sock;
  53
  54         nl_cb_set(nl_cb_rtnl_event, NL_CB_VALID, NL_CB_CUSTOM,
  55                   cb_rtnl_event, NULL);
  56
  57         sock_rtnl_event = nl_socket_alloc();
  58         if (!sock_rtnl_event)
  59                 goto error_free_cb;
  60
  61         if (nl_connect(sock_rtnl_event, NETLINK_ROUTE))
  62                 goto error_free_event;
  63
  64         // Receive network link events form kernel
  65         nl_socket_add_membership(sock_rtnl_event, RTNLGRP_LINK);
  66
  67         rtnl_event.fd = nl_socket_get_fd(sock_rtnl_event);
  68         uloop_fd_add(&rtnl_event, ULOOP_READ | ULOOP_EDGE_TRIGGER);
  69
  70         return 0;
  71
  72 error_free_event:
  73         nl_socket_free(sock_rtnl_event);
  74         sock_rtnl_event = NULL;
  75 error_free_cb:
  76         nl_cb_put(nl_cb_rtnl_event);
  77         nl_cb_rtnl_event = NULL;
  78 error_free_sock:
  79         nl_socket_free(sock_rtnl);
  80         sock_rtnl = NULL;
  81         return -1;
  82 }
  83
  84 // If socket is ready for reading parse netlink events
  85 static void handler_rtnl_event(struct uloop_fd *u, unsigned int events)
  86 {
  87         nl_recvmsgs(sock_rtnl_event, nl_cb_rtnl_event);
  88 }
  89
  90 // Evaluate netlink messages
  91 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
  92 {
  93         struct nlmsghdr *nh = nlmsg_hdr(msg);
  94         struct ifinfomsg *ifi = NLMSG_DATA(nh);
  95         struct nlattr *nla[__IFLA_MAX];
  96
  97         if (nh->nlmsg_type != RTM_DELLINK && nh->nlmsg_type != RTM_NEWLINK)
  98                 goto out;
  99
 100         nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
 101         if (!nla[IFLA_IFNAME])
 102                 goto out;
 103
 104         struct device *dev = device_get(RTA_DATA(nla[IFLA_IFNAME]), false);
 105         if (!dev)
 106                 goto out;
 107
 108         dev->ifindex = ifi->ifi_index;
 109         device_set_present(dev, (nh->nlmsg_type == RTM_NEWLINK));
 110
 111 out:
 112         return 0;
 113 }
 114
 115 static int system_rtnl_call(struct nl_msg *msg)
 116 {
 117         int s = -(nl_send_auto_complete(sock_rtnl, msg)
 118                         || nl_wait_for_ack(sock_rtnl));
 119         nlmsg_free(msg);
 120         return s;
 121 }
 122
 123 int system_bridge_delbr(struct device *bridge)
 124 {
 125         return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 126 }
 127
 128 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 129 {
 130         struct ifreq ifr;
 131         if (dev)
 132                 ifr.ifr_ifindex = dev->ifindex;
 133         else
 134                 ifr.ifr_data = data;
 135         strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
 136         return ioctl(sock_ioctl, cmd, &ifr);
 137 }
 138
 139 int system_bridge_addif(struct device *bridge, struct device *dev)
 140 {
 141         return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 142 }
 143
 144 int system_bridge_delif(struct device *bridge, struct device *dev)
 145 {
 146         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 147 }
 148
 149 static bool system_is_bridge(const char *name, char *buf, int buflen)
 150 {
 151         struct stat st;
 152
 153         snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
 154         if (stat(buf, &st) < 0)
 155                 return false;
 156
 157         return true;
 158 }
 159
 160 static char *system_get_bridge(const char *name, char *buf, int buflen)
 161 {
 162         char *path;
 163         ssize_t len;
 164         glob_t gl;
 165
 166         snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
 167         if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
 168                 return NULL;
 169
 170         if (gl.gl_pathc == 0)
 171                 return NULL;
 172
 173         len = readlink(gl.gl_pathv[0], buf, buflen);
 174         if (len < 0)
 175                 return NULL;
 176
 177         buf[len] = 0;
 178         path = strrchr(buf, '/');
 179         if (!path)
 180                 return NULL;
 181
 182         return path + 1;
 183 }
 184
 185 static int system_if_resolve(struct device *dev)
 186 {
 187         struct ifreq ifr;
 188         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 189         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 190                 return ifr.ifr_ifindex;
 191         else
 192                 return 0;
 193 }
 194
 195 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 196 {
 197         struct ifreq ifr;
 198         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 199         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 200         ifr.ifr_flags |= add;
 201         ifr.ifr_flags &= ~rem;
 202         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 203 }
 204
 205 struct clear_data {
 206         struct nl_msg *msg;
 207         struct device *dev;
 208         int type;
 209         int size;
 210         int af;
 211 };
 212
 213
 214 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 215 {
 216         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 217
 218         return ifa->ifa_index == ifindex;
 219 }
 220
 221 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 222 {
 223         struct nlattr *tb[__RTA_MAX];
 224
 225         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 226         if (!tb[RTA_OIF])
 227                 return false;
 228
 229         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 230 }
 231
 232 static int cb_clear_event(struct nl_msg *msg, void *arg)
 233 {
 234         struct clear_data *clr = arg;
 235         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 236         bool (*cb)(struct nlmsghdr *, int ifindex);
 237         int type;
 238
 239         switch(clr->type) {
 240         case RTM_GETADDR:
 241                 type = RTM_DELADDR;
 242                 if (hdr->nlmsg_type != RTM_NEWADDR)
 243                         return NL_SKIP;
 244
 245                 cb = check_ifaddr;
 246                 break;
 247         case RTM_GETROUTE:
 248                 type = RTM_DELROUTE;
 249                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 250                         return NL_SKIP;
 251
 252                 cb = check_route;
 253                 break;
 254         default:
 255                 return NL_SKIP;
 256         }
 257
 258         if (!cb(hdr, clr->dev->ifindex))
 259                 return NL_SKIP;
 260
 261         D(SYSTEM, "Remove %s from device %s\n",
 262           type == RTM_DELADDR ? "an address" : "a route",
 263           clr->dev->ifname);
 264         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 265         hdr = nlmsg_hdr(clr->msg);
 266         hdr->nlmsg_type = type;
 267         hdr->nlmsg_flags = NLM_F_REQUEST;
 268
 269         if (!nl_send_auto_complete(sock_rtnl, clr->msg))
 270                 nl_wait_for_ack(sock_rtnl);
 271
 272         return NL_SKIP;
 273 }
 274
 275 static int
 276 cb_finish_event(struct nl_msg *msg, void *arg)
 277 {
 278         int *pending = arg;
 279         *pending = 0;
 280         return NL_STOP;
 281 }
 282
 283 static int
 284 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 285 {
 286         int *pending = arg;
 287         *pending = err->error;
 288         return NL_STOP;
 289 }
 290
 291 static void
 292 system_if_clear_entries(struct device *dev, int type, int af)
 293 {
 294         struct clear_data clr;
 295         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 296         struct rtmsg rtm = {
 297                 .rtm_family = af,
 298                 .rtm_flags = RTM_F_CLONED,
 299         };
 300         int flags = NLM_F_DUMP;
 301         int pending = 1;
 302
 303         clr.af = af;
 304         clr.dev = dev;
 305         clr.type = type;
 306         switch (type) {
 307         case RTM_GETADDR:
 308                 clr.size = sizeof(struct rtgenmsg);
 309                 break;
 310         case RTM_GETROUTE:
 311                 clr.size = sizeof(struct rtmsg);
 312                 break;
 313         default:
 314                 return;
 315         }
 316
 317         if (!cb)
 318                 return;
 319
 320         clr.msg = nlmsg_alloc_simple(type, flags);
 321         if (!clr.msg)
 322                 goto out;
 323
 324         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 325         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 326         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 327         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 328
 329         nl_send_auto_complete(sock_rtnl, clr.msg);
 330         while (pending > 0)
 331                 nl_recvmsgs(sock_rtnl, cb);
 332
 333         nlmsg_free(clr.msg);
 334 out:
 335         nl_cb_put(cb);
 336 }
 337
 338 /*
 339  * Clear bridge (membership) state and bring down device
 340  */
 341 void system_if_clear_state(struct device *dev)
 342 {
 343         char buf[256];
 344         char *bridge;
 345
 346         dev->ifindex = system_if_resolve(dev);
 347         if (!dev->ifindex)
 348                 return;
 349
 350         system_if_flags(dev->ifname, 0, IFF_UP);
 351
 352         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 353                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 354                 system_bridge_delbr(dev);
 355                 return;
 356         }
 357
 358         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 359         if (bridge) {
 360                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 361                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 362         }
 363
 364         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 365         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 366         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 367         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 368 }
 369
 370 static inline unsigned long
 371 sec_to_jiffies(int val)
 372 {
 373         return (unsigned long) val * 100;
 374 }
 375
 376 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 377 {
 378         unsigned long args[4] = {};
 379
 380         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 381                 return -1;
 382
 383         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 384         args[1] = !!cfg->stp;
 385         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 386
 387         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 388         args[1] = sec_to_jiffies(cfg->forward_delay);
 389         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 390
 391         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 392                 args[0] = BRCTL_SET_AGEING_TIME;
 393                 args[1] = sec_to_jiffies(cfg->ageing_time);
 394                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 395         }
 396
 397         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 398                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 399                 args[1] = sec_to_jiffies(cfg->hello_time);
 400                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 401         }
 402
 403         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 404                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 405                 args[1] = sec_to_jiffies(cfg->max_age);
 406                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 407         }
 408
 409         return 0;
 410 }
 411
 412 static int system_vlan(struct device *dev, int id)
 413 {
 414         struct vlan_ioctl_args ifr = {
 415                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 416                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 417         };
 418
 419         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 420
 421         if (id < 0) {
 422                 ifr.cmd = DEL_VLAN_CMD;
 423                 ifr.u.VID = 0;
 424         } else {
 425                 ifr.cmd = ADD_VLAN_CMD;
 426                 ifr.u.VID = id;
 427         }
 428         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 429         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 430 }
 431
 432 int system_vlan_add(struct device *dev, int id)
 433 {
 434         return system_vlan(dev, id);
 435 }
 436
 437 int system_vlan_del(struct device *dev)
 438 {
 439         return system_vlan(dev, -1);
 440 }
 441
 442 int system_if_up(struct device *dev)
 443 {
 444         dev->ifindex = system_if_resolve(dev);
 445         return system_if_flags(dev->ifname, IFF_UP, 0);
 446 }
 447
 448 int system_if_down(struct device *dev)
 449 {
 450         return system_if_flags(dev->ifname, 0, IFF_UP);
 451 }
 452
 453 int system_if_check(struct device *dev)
 454 {
 455         device_set_present(dev, (system_if_resolve(dev) >= 0));
 456         return 0;
 457 }
 458
 459 int system_if_dump_stats(struct device *dev, struct blob_buf *b)
 460 {
 461         const char *const counters[] = {
 462                 "collisions",     "rx_frame_errors",   "tx_compressed",
 463                 "multicast",      "rx_length_errors",  "tx_dropped",
 464                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
 465                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
 466                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
 467                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
 468                 "rx_errors",      "tx_bytes",          "tx_window_errors",
 469                 "rx_fifo_errors", "tx_carrier_errors",
 470         };
 471         char buf[64];
 472         int stats_dir;
 473         int i, fd, len;
 474
 475         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
 476         stats_dir = open(buf, O_DIRECTORY);
 477         if (stats_dir < 0)
 478                 return -1;
 479
 480         for (i = 0; i < ARRAY_SIZE(counters); i++) {
 481                 fd = openat(stats_dir, counters[i], O_RDONLY);
 482                 if (fd < 0)
 483                         continue;
 484
 485 retry:
 486                 len = read(fd, buf, sizeof(buf));
 487                 if (len < 0) {
 488                         if (errno == EINTR)
 489                                 goto retry;
 490                         continue;
 491                 }
 492
 493                 buf[len] = 0;
 494                 blobmsg_add_u32(b, counters[i], strtoul(buf, NULL, 0));
 495                 close(fd);
 496         }
 497
 498         close(stats_dir);
 499         return 0;
 500 }
 501
 502 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
 503 {
 504         int alen = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
 505         struct ifaddrmsg ifa = {
 506                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
 507                 .ifa_prefixlen = addr->mask,
 508                 .ifa_index = dev->ifindex,
 509         };
 510
 511         struct nl_msg *msg = nlmsg_alloc_simple(cmd, 0);
 512         if (!msg)
 513                 return -1;
 514
 515         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
 516         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
 517         return system_rtnl_call(msg);
 518 }
 519
 520 int system_add_address(struct device *dev, struct device_addr *addr)
 521 {
 522         return system_addr(dev, addr, RTM_NEWADDR);
 523 }
 524
 525 int system_del_address(struct device *dev, struct device_addr *addr)
 526 {
 527         return system_addr(dev, addr, RTM_DELADDR);
 528 }
 529
 530 static int system_rt(struct device *dev, struct device_route *route, int cmd)
 531 {
 532         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
 533         bool have_gw;
 534         unsigned int flags = 0;
 535
 536         if (alen == 4)
 537                 have_gw = !!route->nexthop.in.s_addr;
 538         else
 539                 have_gw = route->nexthop.in6.s6_addr32[0] ||
 540                         route->nexthop.in6.s6_addr32[1] ||
 541                         route->nexthop.in6.s6_addr32[2] ||
 542                         route->nexthop.in6.s6_addr32[3];
 543
 544         unsigned char scope = (cmd == RTM_DELROUTE) ? RT_SCOPE_NOWHERE :
 545                         (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
 546
 547         struct rtmsg rtm = {
 548                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
 549                 .rtm_dst_len = route->mask,
 550                 .rtm_table = RT_TABLE_MAIN,
 551                 .rtm_protocol = RTPROT_BOOT,
 552                 .rtm_scope = scope,
 553                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
 554         };
 555
 556         if (cmd == RTM_NEWROUTE)
 557                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
 558
 559         struct nl_msg *msg = nlmsg_alloc_simple(cmd, flags);
 560         if (!msg)
 561                 return -1;
 562
 563         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
 564
 565         if (route->mask)
 566                 nla_put(msg, RTA_DST, alen, &route->addr);
 567
 568         if (have_gw)
 569                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
 570
 571         if (route->flags & DEVADDR_DEVICE)
 572                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
 573
 574         return system_rtnl_call(msg);
 575 }
 576
 577 int system_add_route(struct device *dev, struct device_route *route)
 578 {
 579         return system_rt(dev, route, RTM_NEWROUTE);
 580 }
 581
 582 int system_del_route(struct device *dev, struct device_route *route)
 583 {
 584         return system_rt(dev, route, RTM_DELROUTE);
 585 }
 586
 587 int system_flush_routes(void)
 588 {
 589         const char *names[] = {
 590                 "/proc/sys/net/ipv4/route/flush",
 591                 "/proc/sys/net/ipv6/route/flush"
 592         };
 593         int fd, i;
 594
 595         for (i = 0; i < ARRAY_SIZE(names); i++) {
 596                 fd = open(names[i], O_WRONLY);
 597                 if (fd < 0)
 598                         continue;
 599
 600                 write(fd, "-1", 2);
 601                 close(fd);
 602         }
 603         return 0;
 604 }
 605
 606 time_t system_get_rtime(void)
 607 {
 608         struct timespec ts;
 609         struct timeval tv;
 610
 611         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
 612                 return ts.tv_sec;
 613
 614         if (gettimeofday(&tv, NULL) == 0)
 615                 return tv.tv_sec;
 616
 617         return 0;
 618 }