interface-ip.c

   1 /*
   2  * netifd - network interface daemon
   3  * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
   4  * Copyright (C) 2012 Steven Barth <steven@midlink.org>
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2
   8  * as published by the Free Software Foundation
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13  * GNU General Public License for more details.
  14  */
  15 #include <string.h>
  16 #include <stdlib.h>
  17 #include <stdio.h>
  18
  19 #include <limits.h>
  20 #include <arpa/inet.h>
  21 #include <netinet/in.h>
  22
  23 #include "netifd.h"
  24 #include "device.h"
  25 #include "interface.h"
  26 #include "interface-ip.h"
  27 #include "proto.h"
  28 #include "ubus.h"
  29 #include "system.h"
  30
  31 enum {
  32         ROUTE_INTERFACE,
  33         ROUTE_TARGET,
  34         ROUTE_MASK,
  35         ROUTE_GATEWAY,
  36         ROUTE_METRIC,
  37         ROUTE_MTU,
  38         ROUTE_VALID,
  39         ROUTE_TABLE,
  40         ROUTE_SOURCE,
  41         ROUTE_ONLINK,
  42         ROUTE_TYPE,
  43         ROUTE_PROTO,
  44         __ROUTE_MAX
  45 };
  46
  47 static const struct blobmsg_policy route_attr[__ROUTE_MAX] = {
  48         [ROUTE_INTERFACE] = { .name = "interface", .type = BLOBMSG_TYPE_STRING },
  49         [ROUTE_TARGET] = { .name = "target", .type = BLOBMSG_TYPE_STRING },
  50         [ROUTE_MASK] = { .name = "netmask", .type = BLOBMSG_TYPE_STRING },
  51         [ROUTE_GATEWAY] = { .name = "gateway", .type = BLOBMSG_TYPE_STRING },
  52         [ROUTE_METRIC] = { .name = "metric", .type = BLOBMSG_TYPE_INT32 },
  53         [ROUTE_MTU] = { .name = "mtu", .type = BLOBMSG_TYPE_INT32 },
  54         [ROUTE_TABLE] = { .name = "table", .type = BLOBMSG_TYPE_STRING },
  55         [ROUTE_VALID] = { .name = "valid", .type = BLOBMSG_TYPE_INT32 },
  56         [ROUTE_SOURCE] = { .name = "source", .type = BLOBMSG_TYPE_STRING },
  57         [ROUTE_ONLINK] = { .name = "onlink", .type = BLOBMSG_TYPE_BOOL },
  58         [ROUTE_TYPE] = { .name = "type", .type = BLOBMSG_TYPE_STRING },
  59         [ROUTE_PROTO] = { .name = "proto", .type = BLOBMSG_TYPE_STRING },
  60 };
  61
  62 const struct uci_blob_param_list route_attr_list = {
  63         .n_params = __ROUTE_MAX,
  64         .params = route_attr,
  65 };
  66
  67
  68 struct list_head prefixes = LIST_HEAD_INIT(prefixes);
  69 static struct device_prefix *ula_prefix = NULL;
  70 static struct uloop_timeout valid_until_timeout;
  71
  72
  73 static void
  74 clear_if_addr(union if_addr *a, int mask)
  75 {
  76         int m_bytes = (mask + 7) / 8;
  77         uint8_t m_clear = (1 << (m_bytes * 8 - mask)) - 1;
  78         uint8_t *p = (uint8_t *) a;
  79
  80         if (m_bytes < sizeof(*a))
  81                 memset(p + m_bytes, 0, sizeof(*a) - m_bytes);
  82
  83         p[m_bytes - 1] &= ~m_clear;
  84 }
  85
  86 static bool
  87 match_if_addr(union if_addr *a1, union if_addr *a2, int mask)
  88 {
  89         union if_addr *p1, *p2;
  90
  91         p1 = alloca(sizeof(*a1));
  92         p2 = alloca(sizeof(*a2));
  93
  94         memcpy(p1, a1, sizeof(*a1));
  95         clear_if_addr(p1, mask);
  96         memcpy(p2, a2, sizeof(*a2));
  97         clear_if_addr(p2, mask);
  98
  99         return !memcmp(p1, p2, sizeof(*p1));
 100 }
 101
 102 static int set_ip_source_policy(bool add, bool v6, unsigned int priority,
 103                 const union if_addr *addr, uint8_t mask, unsigned int table,
 104                 struct interface *in_iface, const char *action, bool src)
 105 {
 106         struct iprule rule = {
 107                 .flags = IPRULE_PRIORITY,
 108                 .priority = priority
 109         };
 110
 111         if (addr) {
 112                 if (src) {
 113                         rule.flags |= IPRULE_SRC;
 114                         rule.src_addr = *addr;
 115                         rule.src_mask = mask;
 116                 } else {
 117                         rule.flags |= IPRULE_DEST;
 118                         rule.dest_addr = *addr;
 119                         rule.dest_mask = mask;
 120                 }
 121         }
 122
 123         if (table) {
 124                 rule.flags |= IPRULE_LOOKUP;
 125                 rule.lookup = table;
 126
 127                 if (!rule.lookup)
 128                         return 0;
 129         } else if (action) {
 130                 rule.flags |= IPRULE_ACTION;
 131                 system_resolve_iprule_action(action, &rule.action);
 132         }
 133
 134         if (in_iface && in_iface->l3_dev.dev) {
 135                 rule.flags |= IPRULE_IN;
 136                 strcpy(rule.in_dev, in_iface->l3_dev.dev->ifname);
 137         }
 138
 139         rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
 140
 141         return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
 142 }
 143
 144 static int set_ip_lo_policy(bool add, bool v6, struct interface *iface)
 145 {
 146         struct iprule rule = {
 147                 .flags = IPRULE_IN | IPRULE_LOOKUP | IPRULE_PRIORITY,
 148                 .priority = IPRULE_PRIORITY_NW + iface->l3_dev.dev->ifindex,
 149                 .lookup = (v6) ? iface->ip6table : iface->ip4table,
 150                 .in_dev = "lo"
 151         };
 152
 153         if (!rule.lookup)
 154                 return 0;
 155
 156         rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
 157
 158         return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
 159 }
 160
 161 static bool
 162 __find_ip_addr_target(struct interface_ip_settings *ip, union if_addr *a, bool v6)
 163 {
 164         struct device_addr *addr;
 165
 166         vlist_for_each_element(&ip->addr, addr, node) {
 167                 if (!addr->enabled)
 168                         continue;
 169
 170                 if (v6 != ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
 171                         continue;
 172
 173                 /* Handle offlink addresses correctly */
 174                 unsigned int mask = addr->mask;
 175                 if ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6 &&
 176                                 (addr->flags & DEVADDR_OFFLINK))
 177                         mask = 128;
 178
 179                 if (!match_if_addr(&addr->addr, a, mask))
 180                         continue;
 181
 182                 return true;
 183         }
 184
 185         return false;
 186 }
 187
 188 static void
 189 __find_ip_route_target(struct interface_ip_settings *ip, union if_addr *a,
 190                        bool v6, struct device_route **res)
 191 {
 192         struct device_route *route;
 193
 194         vlist_for_each_element(&ip->route, route, node) {
 195                 if (!route->enabled)
 196                         continue;
 197
 198                 if (v6 != ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
 199                         continue;
 200
 201                 if (!match_if_addr(&route->addr, a, route->mask))
 202                         continue;
 203
 204                 if (route->flags & DEVROUTE_TABLE)
 205                         continue;
 206
 207                 if (!*res || route->mask > (*res)->mask)
 208                         *res = route;
 209         }
 210 }
 211
 212 static bool
 213 interface_ip_find_addr_target(struct interface *iface, union if_addr *a, bool v6)
 214 {
 215         return __find_ip_addr_target(&iface->proto_ip, a, v6) ||
 216                __find_ip_addr_target(&iface->config_ip, a, v6);
 217 }
 218
 219 static void
 220 interface_ip_find_route_target(struct interface *iface, union if_addr *a,
 221                                bool v6, struct device_route **route)
 222 {
 223         __find_ip_route_target(&iface->proto_ip, a, v6, route);
 224         __find_ip_route_target(&iface->config_ip, a, v6, route);
 225 }
 226
 227 struct interface *
 228 interface_ip_add_target_route(union if_addr *addr, bool v6, struct interface *iface)
 229 {
 230         struct device_route *route, *r_next = NULL;
 231         bool defaultroute_target = false;
 232         union if_addr addr_zero;
 233         int addrsize = v6 ? sizeof(addr->in6) : sizeof(addr->in);
 234
 235         memset(&addr_zero, 0, sizeof(addr_zero));
 236         if (memcmp(&addr_zero, addr, addrsize) == 0)
 237                 defaultroute_target = true;
 238
 239         if (iface) {
 240                 /* look for locally addressable target first */
 241                 if (interface_ip_find_addr_target(iface, addr, v6))
 242                         return iface;
 243
 244                 /* do not stop at the first route, let the lookup compare
 245                  * masks to find the best match */
 246                 interface_ip_find_route_target(iface, addr, v6, &r_next);
 247         } else {
 248                 vlist_for_each_element(&interfaces, iface, node) {
 249                         /* look for locally addressable target first */
 250                         if (interface_ip_find_addr_target(iface, addr, v6))
 251                                 return iface;
 252
 253                         /* do not stop at the first route, let the lookup compare
 254                          * masks to find the best match */
 255                         interface_ip_find_route_target(iface, addr, v6, &r_next);
 256                 }
 257         }
 258
 259         if (!r_next)
 260                 return NULL;
 261
 262         iface = r_next->iface;
 263         if (defaultroute_target)
 264                 return iface;
 265
 266         route = calloc(1, sizeof(*route));
 267         if (!route)
 268                 return NULL;
 269
 270         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
 271         route->mask = v6 ? 128 : 32;
 272         memcpy(&route->addr, addr, addrsize);
 273         memcpy(&route->nexthop, &r_next->nexthop, sizeof(route->nexthop));
 274         route->mtu = r_next->mtu;
 275         route->metric = r_next->metric;
 276         route->table = r_next->table;
 277         route->iface = iface;
 278         vlist_add(&iface->host_routes, &route->node, route);
 279
 280         return iface;
 281 }
 282
 283 static void
 284 interface_set_route_info(struct interface *iface, struct device_route *route)
 285 {
 286         bool v6 = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET6);
 287
 288         if (!iface)
 289                 return;
 290
 291         if (!(route->flags & DEVROUTE_METRIC))
 292                 route->metric = iface->metric;
 293
 294         if (!(route->flags & DEVROUTE_TABLE)) {
 295                 route->table = (v6) ? iface->ip6table : iface->ip4table;
 296                 if (route->table)
 297                         route->flags |= DEVROUTE_SRCTABLE;
 298         }
 299 }
 300
 301 void
 302 interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
 303 {
 304         struct interface_ip_settings *ip;
 305         struct blob_attr *tb[__ROUTE_MAX], *cur;
 306         struct device_route *route;
 307         int af = v6 ? AF_INET6 : AF_INET;
 308
 309         blobmsg_parse(route_attr, __ROUTE_MAX, tb, blobmsg_data(attr), blobmsg_data_len(attr));
 310
 311         if (!iface) {
 312                 if ((cur = tb[ROUTE_INTERFACE]) == NULL)
 313                         return;
 314
 315                 iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
 316                 if (!iface)
 317                         return;
 318
 319                 ip = &iface->config_ip;
 320         } else {
 321                 ip = &iface->proto_ip;
 322         }
 323
 324         route = calloc(1, sizeof(*route));
 325         if (!route)
 326                 return;
 327
 328         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
 329         route->mask = v6 ? 128 : 32;
 330         if ((cur = tb[ROUTE_MASK]) != NULL) {
 331                 route->mask = parse_netmask_string(blobmsg_data(cur), v6);
 332                 if (route->mask > (v6 ? 128 : 32))
 333                         goto error;
 334         }
 335
 336         if ((cur = tb[ROUTE_TARGET]) != NULL) {
 337                 if (!parse_ip_and_netmask(af, blobmsg_data(cur), &route->addr, &route->mask)) {
 338                         DPRINTF("Failed to parse route target: %s\n", (char *) blobmsg_data(cur));
 339                         goto error;
 340                 }
 341         }
 342
 343         if ((cur = tb[ROUTE_GATEWAY]) != NULL) {
 344                 if (!inet_pton(af, blobmsg_data(cur), &route->nexthop)) {
 345                         DPRINTF("Failed to parse route gateway: %s\n", (char *) blobmsg_data(cur));
 346                         goto error;
 347                 }
 348         }
 349
 350         if ((cur = tb[ROUTE_METRIC]) != NULL) {
 351                 route->metric = blobmsg_get_u32(cur);
 352                 route->flags |= DEVROUTE_METRIC;
 353         }
 354
 355         if ((cur = tb[ROUTE_MTU]) != NULL) {
 356                 route->mtu = blobmsg_get_u32(cur);
 357                 route->flags |= DEVROUTE_MTU;
 358         }
 359
 360         /* Use source-based routing */
 361         if ((cur = tb[ROUTE_SOURCE]) != NULL) {
 362                 char *saveptr, *source = alloca(blobmsg_data_len(cur));
 363                 memcpy(source, blobmsg_data(cur), blobmsg_data_len(cur));
 364
 365                 const char *addr = strtok_r(source, "/", &saveptr);
 366                 const char *mask = strtok_r(NULL, "/", &saveptr);
 367
 368                 if (!addr || inet_pton(af, addr, &route->source) < 1) {
 369                         DPRINTF("Failed to parse route source: %s\n", addr ? addr : "NULL");
 370                         goto error;
 371                 }
 372
 373                 route->sourcemask = (mask) ? atoi(mask) : ((af == AF_INET6) ? 128 : 32);
 374         }
 375
 376         if ((cur = tb[ROUTE_ONLINK]) != NULL && blobmsg_get_bool(cur))
 377                 route->flags |= DEVROUTE_ONLINK;
 378
 379         if ((cur = tb[ROUTE_TABLE]) != NULL) {
 380                 if (!system_resolve_rt_table(blobmsg_data(cur), &route->table)) {
 381                         DPRINTF("Failed to resolve routing table: %s\n", (char *) blobmsg_data(cur));
 382                         goto error;
 383                 }
 384
 385                 /* only set the table flag if not using the main (default) table */
 386                 if (system_is_default_rt_table(route->table))
 387                         route->table = 0;
 388
 389                 if (route->table)
 390                         route->flags |= DEVROUTE_TABLE;
 391         }
 392
 393         if ((cur = tb[ROUTE_VALID]) != NULL) {
 394                 int64_t valid = blobmsg_get_u32(cur);
 395                 int64_t valid_until = valid + (int64_t)system_get_rtime();
 396                 if (valid_until <= LONG_MAX && valid != 0xffffffffLL) /* Catch overflow */
 397                         route->valid_until = valid_until;
 398         }
 399
 400         if ((cur = tb[ROUTE_TYPE]) != NULL) {
 401                 if (!system_resolve_rt_type(blobmsg_data(cur), &route->type)) {
 402                         DPRINTF("Failed to resolve routing type: %s\n", (char *) blobmsg_data(cur));
 403                         goto error;
 404                 }
 405                 route->flags |= DEVROUTE_TYPE;
 406         }
 407
 408         if ((cur = tb[ROUTE_PROTO]) != NULL) {
 409                 if (!system_resolve_rt_proto(blobmsg_data(cur), &route->proto)) {
 410                         DPRINTF("Failed to resolve proto type: %s\n", (char *) blobmsg_data(cur));
 411                         goto error;
 412                 }
 413                 route->flags |= DEVROUTE_PROTO;
 414         }
 415
 416         interface_set_route_info(iface, route);
 417         vlist_add(&ip->route, &route->node, route);
 418         return;
 419
 420 error:
 421         free(route);
 422 }
 423
 424 static int
 425 addr_cmp(const void *k1, const void *k2, void *ptr)
 426 {
 427         return memcmp(k1, k2, sizeof(struct device_addr) -
 428                       offsetof(struct device_addr, flags));
 429 }
 430
 431 static int
 432 route_cmp(const void *k1, const void *k2, void *ptr)
 433 {
 434         const struct device_route *r1 = k1, *r2 = k2;
 435
 436         if (r1->mask != r2->mask)
 437                 return r2->mask - r1->mask;
 438
 439         if (r1->metric != r2->metric)
 440                 return r1->metric - r2->metric;
 441
 442         if (r1->flags != r2->flags)
 443                 return r2->flags - r1->flags;
 444
 445         if (r1->sourcemask != r2->sourcemask)
 446                 return r1->sourcemask - r2->sourcemask;
 447
 448         if (r1->table != r2->table)
 449                 return r1->table - r2->table;
 450
 451         int maskcmp = memcmp(&r1->source, &r2->source, sizeof(r1->source));
 452         if (maskcmp)
 453                 return maskcmp;
 454
 455         return memcmp(&r1->addr, &r2->addr, sizeof(r1->addr));
 456 }
 457
 458 static int
 459 prefix_cmp(const void *k1, const void *k2, void *ptr)
 460 {
 461         return memcmp(k1, k2, offsetof(struct device_prefix, pclass) -
 462                         offsetof(struct device_prefix, addr));
 463 }
 464
 465 static void
 466 interface_handle_subnet_route(struct interface *iface, struct device_addr *addr, bool add)
 467 {
 468         struct device *dev = iface->l3_dev.dev;
 469         struct device_route *r = &addr->subnet;
 470
 471         if (addr->flags & DEVADDR_OFFLINK)
 472                 return;
 473
 474         if (!add) {
 475                 if (!addr->subnet.iface)
 476                         return;
 477
 478                 system_del_route(dev, r);
 479                 memset(r, 0, sizeof(*r));
 480                 return;
 481         }
 482
 483         r->iface = iface;
 484         r->flags = addr->flags;
 485         r->mask = addr->mask;
 486         memcpy(&r->addr, &addr->addr, sizeof(r->addr));
 487         clear_if_addr(&r->addr, r->mask);
 488
 489         if (!system_resolve_rt_proto("kernel", &r->proto))
 490                 return;
 491
 492         r->flags |= DEVROUTE_PROTO;
 493         system_del_route(dev, r);
 494
 495         r->flags &= ~DEVROUTE_PROTO;
 496         interface_set_route_info(iface, r);
 497
 498         system_add_route(dev, r);
 499 }
 500
 501 static void
 502 interface_add_addr_rules(struct device_addr *addr, bool enabled)
 503 {
 504         bool v6 = (addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6;
 505
 506         set_ip_source_policy(enabled, v6, IPRULE_PRIORITY_ADDR, &addr->addr,
 507                              (v6) ? 128 : 32, addr->policy_table, NULL, NULL,
 508                              true);
 509         set_ip_source_policy(enabled, v6, IPRULE_PRIORITY_ADDR_MASK,
 510                              &addr->addr, addr->mask, addr->policy_table, NULL,
 511                              NULL, false);
 512 }
 513
 514 static void
 515 interface_update_proto_addr(struct vlist_tree *tree,
 516                             struct vlist_node *node_new,
 517                             struct vlist_node *node_old)
 518 {
 519         struct interface_ip_settings *ip;
 520         struct interface *iface;
 521         struct device *dev;
 522         struct device_addr *a_new = NULL, *a_old = NULL;
 523         bool replace = false;
 524         bool keep = false;
 525         bool v6 = false;
 526
 527         ip = container_of(tree, struct interface_ip_settings, addr);
 528         iface = ip->iface;
 529         dev = iface->l3_dev.dev;
 530
 531         if (!node_new || !node_old)
 532                 iface->updated |= IUF_ADDRESS;
 533
 534         if (node_new) {
 535                 a_new = container_of(node_new, struct device_addr, node);
 536
 537                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
 538                     !a_new->broadcast) {
 539
 540                         /* /31 and /32 addressing need 255.255.255.255
 541                          * as broadcast address. */
 542                         if (a_new->mask >= 31) {
 543                                 a_new->broadcast = (uint32_t) ~0;
 544                         } else {
 545                                 uint32_t mask = ~0;
 546                                 uint32_t *a = (uint32_t *) &a_new->addr;
 547
 548                                 mask >>= a_new->mask;
 549                                 a_new->broadcast = *a | htonl(mask);
 550                         }
 551                 }
 552         }
 553
 554         if (node_old)
 555                 a_old = container_of(node_old, struct device_addr, node);
 556
 557         if (a_new && a_old) {
 558                 keep = true;
 559
 560                 if (a_old->flags != a_new->flags || a_old->failed)
 561                         keep = false;
 562
 563                 if (a_old->valid_until != a_new->valid_until ||
 564                                 a_old->preferred_until != a_new->preferred_until)
 565                         replace = true;
 566
 567                 if (((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4) &&
 568                      (a_new->broadcast != a_old->broadcast ||
 569                       a_new->point_to_point != a_old->point_to_point))
 570                         keep = false;
 571         }
 572
 573         if (node_old) {
 574                 if (a_old->enabled && !keep) {
 575                         /*
 576                          * This is needed for source routing to work correctly. If a device
 577                          * has two connections to a network using the same subnet, adding
 578                          * only the network-rule will cause packets to be routed through the
 579                          * first matching network (source IP matches both masks)
 580                          */
 581                         if (a_old->policy_table)
 582                                 interface_add_addr_rules(a_old, false);
 583
 584                         if (!(a_old->flags & DEVADDR_EXTERNAL)) {
 585                                 interface_handle_subnet_route(iface, a_old, false);
 586                                 system_del_address(dev, a_old);
 587                         }
 588                 }
 589                 free(a_old->pclass);
 590                 free(a_old);
 591         }
 592
 593         if (node_new) {
 594                 a_new->enabled = true;
 595
 596                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
 597                                 v6 = true;
 598
 599                 a_new->policy_table = (v6) ? iface->ip6table : iface->ip4table;
 600
 601                 if (!keep || replace) {
 602                         if (!(a_new->flags & DEVADDR_EXTERNAL)) {
 603                                 if (system_add_address(dev, a_new))
 604                                         a_new->failed = true;
 605
 606                                 if (iface->metric || a_new->policy_table)
 607                                         interface_handle_subnet_route(iface, a_new, true);
 608                         }
 609
 610                         if (!keep) {
 611                                 if (a_new->policy_table)
 612                                         interface_add_addr_rules(a_new, true);
 613                         }
 614                 }
 615         }
 616 }
 617
 618 static bool
 619 enable_route(struct interface_ip_settings *ip, struct device_route *route)
 620 {
 621         if (ip->no_defaultroute && !route->mask)
 622                 return false;
 623
 624         return ip->enabled;
 625 }
 626
 627 static void
 628 interface_update_proto_route(struct vlist_tree *tree,
 629                              struct vlist_node *node_new,
 630                              struct vlist_node *node_old)
 631 {
 632         struct interface_ip_settings *ip;
 633         struct interface *iface;
 634         struct device *dev;
 635         struct device_route *route_old, *route_new;
 636         bool keep = false;
 637
 638         ip = container_of(tree, struct interface_ip_settings, route);
 639         iface = ip->iface;
 640         dev = iface->l3_dev.dev;
 641
 642         if (!node_new || !node_old)
 643                 iface->updated |= IUF_ROUTE;
 644
 645         route_old = container_of(node_old, struct device_route, node);
 646         route_new = container_of(node_new, struct device_route, node);
 647
 648         if (node_old && node_new)
 649                 keep = !memcmp(&route_old->nexthop, &route_new->nexthop, sizeof(route_old->nexthop)) &&
 650                         (route_old->mtu == route_new->mtu) && (route_old->type == route_new->type) &&
 651                         (route_old->proto == route_new->proto) && !route_old->failed;
 652
 653         if (node_old) {
 654                 if (!(route_old->flags & DEVADDR_EXTERNAL) && route_old->enabled && !keep)
 655                         system_del_route(dev, route_old);
 656
 657                 free(route_old);
 658         }
 659
 660         if (node_new) {
 661                 bool _enabled = enable_route(ip, route_new);
 662
 663                 if (!(route_new->flags & DEVADDR_EXTERNAL) && !keep && _enabled)
 664                         if (system_add_route(dev, route_new))
 665                                 route_new->failed = true;
 666
 667                 route_new->iface = iface;
 668                 route_new->enabled = _enabled;
 669         }
 670 }
 671
 672 static void
 673 interface_update_host_route(struct vlist_tree *tree,
 674                              struct vlist_node *node_new,
 675                              struct vlist_node *node_old)
 676 {
 677         struct interface *iface;
 678         struct device *dev;
 679         struct device_route *route_old, *route_new;
 680
 681         iface = container_of(tree, struct interface, host_routes);
 682         dev = iface->l3_dev.dev;
 683
 684         route_old = container_of(node_old, struct device_route, node);
 685         route_new = container_of(node_new, struct device_route, node);
 686
 687         if (node_old) {
 688                 system_del_route(dev, route_old);
 689                 free(route_old);
 690         }
 691
 692         if (node_new) {
 693                 if (system_add_route(dev, route_new))
 694                         route_new->failed = true;
 695         }
 696 }
 697
 698 static void
 699 random_ifaceid(struct in6_addr *addr)
 700 {
 701         static bool initialized = false;
 702         struct timeval t;
 703
 704         if (!initialized) {
 705                 long int seed = 0;
 706                 gettimeofday(&t, NULL);
 707                 seed = t.tv_sec ^ t.tv_usec ^ getpid();
 708                 srand48(seed);
 709                 initialized = true;
 710         }
 711         addr->s6_addr32[2] = (uint32_t)mrand48();
 712         addr->s6_addr32[3] = (uint32_t)mrand48();
 713 }
 714
 715 static bool
 716 eui64_ifaceid(struct interface *iface, struct in6_addr *addr)
 717 {
 718         struct device_settings st;
 719
 720         device_merge_settings(iface->l3_dev.dev, &st);
 721
 722         if (!(st.flags & DEV_OPT_MACADDR))
 723                 return false;
 724
 725         /* get mac address */
 726         uint8_t *ifaceid = addr->s6_addr + 8;
 727         memcpy(ifaceid, st.macaddr, 3);
 728         memcpy(ifaceid + 5, st.macaddr + 3, 3);
 729         ifaceid[3] = 0xff;
 730         ifaceid[4] = 0xfe;
 731         ifaceid[0] ^= 0x02;
 732
 733         return true;
 734 }
 735
 736 static bool
 737 generate_ifaceid(struct interface *iface, struct in6_addr *addr)
 738 {
 739         bool ret = true;
 740
 741         /* generate new iface id */
 742         switch (iface->assignment_iface_id_selection) {
 743         case IFID_FIXED:
 744                 /* fixed */
 745                 /* copy host part from assignment_fixed_iface_id */
 746                 memcpy(addr->s6_addr + 8, iface->assignment_fixed_iface_id.s6_addr + 8, 8);
 747                 break;
 748         case IFID_RANDOM:
 749                 /* randomize last 64 bits */
 750                 random_ifaceid(addr);
 751                 break;
 752         case IFID_EUI64:
 753                 /* eui64 */
 754                 ret = eui64_ifaceid(iface, addr);
 755                 break;
 756         default:
 757                 ret = false;
 758                 break;
 759         }
 760         return ret;
 761 }
 762
 763 static void
 764 interface_set_prefix_address(struct device_prefix_assignment *assignment,
 765                 const struct device_prefix *prefix, struct interface *iface, bool add)
 766 {
 767         const struct interface *uplink = prefix->iface;
 768         if (!iface->l3_dev.dev)
 769                 return;
 770
 771         struct device *l3_downlink = iface->l3_dev.dev;
 772
 773         struct device_addr addr;
 774         struct device_route route;
 775         memset(&addr, 0, sizeof(addr));
 776         memset(&route, 0, sizeof(route));
 777
 778         addr.addr.in6 = assignment->addr;
 779         addr.mask = assignment->length;
 780         addr.flags = DEVADDR_INET6 | DEVADDR_OFFLINK;
 781         addr.preferred_until = prefix->preferred_until;
 782         addr.valid_until = prefix->valid_until;
 783
 784         route.flags = DEVADDR_INET6;
 785         route.mask = addr.mask < 64 ? 64 : addr.mask;
 786         route.addr = addr.addr;
 787
 788         if (!add && assignment->enabled) {
 789                 time_t now = system_get_rtime();
 790
 791                 addr.preferred_until = now;
 792                 if (!addr.valid_until || addr.valid_until - now > 7200)
 793                         addr.valid_until = now + 7200;
 794
 795                 if (iface->ip6table)
 796                         set_ip_source_policy(false, true, IPRULE_PRIORITY_ADDR_MASK, &addr.addr,
 797                                         addr.mask < 64 ? 64 : addr.mask, iface->ip6table, NULL, NULL, false);
 798
 799                 if (prefix->iface) {
 800                         if (prefix->iface->ip6table)
 801                                 set_ip_source_policy(false, true, IPRULE_PRIORITY_NW, &addr.addr,
 802                                                 addr.mask, prefix->iface->ip6table, iface, NULL, true);
 803
 804                         set_ip_source_policy(false, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 805                                                         addr.mask, 0, iface, "unreachable", true);
 806                 }
 807
 808                 clear_if_addr(&route.addr, route.mask);
 809                 interface_set_route_info(iface, &route);
 810
 811                 system_del_route(l3_downlink, &route);
 812                 system_add_address(l3_downlink, &addr);
 813
 814                 assignment->addr = in6addr_any;
 815                 assignment->enabled = false;
 816         } else if (add && (iface->state == IFS_UP || iface->state == IFS_SETUP)) {
 817                 if (IN6_IS_ADDR_UNSPECIFIED(&addr.addr.in6)) {
 818                         addr.addr.in6 = prefix->addr;
 819                         addr.addr.in6.s6_addr32[1] |= htonl(assignment->assigned);
 820                         if (!generate_ifaceid(iface, &addr.addr.in6))
 821                                 return;
 822
 823                         assignment->addr = addr.addr.in6;
 824                         route.addr = addr.addr;
 825                 }
 826
 827                 if (system_add_address(l3_downlink, &addr))
 828                         return;
 829
 830                 if (!assignment->enabled) {
 831                         if (iface->ip6table)
 832                                 set_ip_source_policy(true, true, IPRULE_PRIORITY_ADDR_MASK, &addr.addr,
 833                                                 addr.mask < 64 ? 64 : addr.mask, iface->ip6table, NULL, NULL, false);
 834
 835                         if (prefix->iface) {
 836                                 set_ip_source_policy(true, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 837                                                 addr.mask, 0, iface, "unreachable", true);
 838
 839                                 if (prefix->iface->ip6table)
 840                                         set_ip_source_policy(true, true, IPRULE_PRIORITY_NW, &addr.addr,
 841                                                         addr.mask, prefix->iface->ip6table, iface, NULL, true);
 842                         }
 843                 }
 844
 845                 clear_if_addr(&route.addr, route.mask);
 846                 interface_set_route_info(iface, &route);
 847
 848                 system_add_route(l3_downlink, &route);
 849
 850                 if (uplink && uplink->l3_dev.dev && !(l3_downlink->settings.flags & DEV_OPT_MTU6)) {
 851                         int mtu = system_update_ipv6_mtu(uplink->l3_dev.dev, 0);
 852                         int mtu_old = system_update_ipv6_mtu(l3_downlink, 0);
 853
 854                         if (mtu > 0 && mtu_old != mtu) {
 855                                 if (system_update_ipv6_mtu(l3_downlink, mtu) < 0)
 856                                         netifd_log_message(L_WARNING, "Failed to set IPv6 mtu to %d"
 857                                                         "on interface '%s'\n", mtu, iface->name);
 858                         }
 859                 }
 860
 861                 assignment->enabled = true;
 862         }
 863 }
 864
 865 static bool interface_prefix_assign(struct list_head *list,
 866                 struct device_prefix_assignment *assign)
 867 {
 868         int32_t current = 0, asize = (1 << (64 - assign->length)) - 1;
 869         struct device_prefix_assignment *c;
 870         list_for_each_entry(c, list, head) {
 871                 if (assign->assigned != -1) {
 872                         if (assign->assigned >= current && assign->assigned + asize < c->assigned) {
 873                                 list_add_tail(&assign->head, &c->head);
 874                                 return true;
 875                         }
 876                 } else if (assign->assigned == -1) {
 877                         current = (current + asize) & (~asize);
 878                         if (current + asize < c->assigned) {
 879                                 assign->assigned = current;
 880                                 list_add_tail(&assign->head, &c->head);
 881                                 return true;
 882                         }
 883                 }
 884                 current = (c->assigned + (1 << (64 - c->length)));
 885         }
 886         return false;
 887 }
 888
 889 /*
 890  * Sorting of assignment entries:
 891  * Primary on assignment length: smallest assignment first
 892  * Secondary on assignment weight: highest weight first
 893  * Finally alphabetical order of interface names
 894  */
 895 static int prefix_assignment_cmp(const void *k1, const void *k2, void *ptr)
 896 {
 897         const struct device_prefix_assignment *a1 = k1, *a2 = k2;
 898
 899         if (a1->length != a2->length)
 900                 return a1->length - a2->length;
 901
 902         if (a1->weight != a2->weight)
 903                 return a2->weight - a1->weight;
 904
 905         return strcmp(a1->name, a2->name);
 906 }
 907
 908 static void interface_update_prefix_assignments(struct device_prefix *prefix, bool setup)
 909 {
 910         struct device_prefix_assignment *c;
 911         struct interface *iface;
 912
 913         /* Delete all assignments */
 914         while (!list_empty(&prefix->assignments)) {
 915                 c = list_first_entry(&prefix->assignments,
 916                                 struct device_prefix_assignment, head);
 917                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 918                         interface_set_prefix_address(c, prefix, iface, false);
 919                 list_del(&c->head);
 920                 free(c);
 921         }
 922
 923         if (!setup)
 924                 return;
 925
 926         /* End-of-assignment sentinel */
 927         c = malloc(sizeof(*c) + 1);
 928         if (!c)
 929                 return;
 930
 931         c->assigned = 1 << (64 - prefix->length);
 932         c->length = 64;
 933         c->name[0] = 0;
 934         c->addr = in6addr_any;
 935         list_add(&c->head, &prefix->assignments);
 936
 937         /* Excluded prefix */
 938         if (prefix->excl_length > 0) {
 939                 const char name[] = "!excluded";
 940                 c = malloc(sizeof(*c) + sizeof(name));
 941                 if (c) {
 942                         c->assigned = ntohl(prefix->excl_addr.s6_addr32[1]) &
 943                                         ((1 << (64 - prefix->length)) - 1);
 944                         c->length = prefix->excl_length;
 945                         c->addr = in6addr_any;
 946                         memcpy(c->name, name, sizeof(name));
 947                         list_add(&c->head, &prefix->assignments);
 948                 }
 949         }
 950
 951         bool assigned_any = false;
 952         struct {
 953                 struct avl_node node;
 954         } *entry, *n_entry;
 955         struct avl_tree assign_later;
 956
 957         avl_init(&assign_later, prefix_assignment_cmp, false, NULL);
 958
 959         vlist_for_each_element(&interfaces, iface, node) {
 960                 if (iface->assignment_length < 48 ||
 961                                 iface->assignment_length > 64)
 962                         continue;
 963
 964                 /* Test whether there is a matching class */
 965                 if (!list_empty(&iface->assignment_classes)) {
 966                         bool found = false;
 967
 968                         struct interface_assignment_class *c;
 969                         list_for_each_entry(c, &iface->assignment_classes, head) {
 970                                 if (!strcmp(c->name, prefix->pclass)) {
 971                                         found = true;
 972                                         break;
 973                                 }
 974                         }
 975
 976                         if (!found)
 977                                 continue;
 978                 }
 979
 980                 size_t namelen = strlen(iface->name) + 1;
 981                 c = malloc(sizeof(*c) + namelen);
 982                 if (!c)
 983                         continue;
 984
 985                 c->length = iface->assignment_length;
 986                 c->assigned = iface->assignment_hint;
 987                 c->weight = iface->assignment_weight;
 988                 c->addr = in6addr_any;
 989                 c->enabled = false;
 990                 memcpy(c->name, iface->name, namelen);
 991
 992                 /* First process all custom assignments, put all others in later-list */
 993                 if (c->assigned == -1 || !interface_prefix_assign(&prefix->assignments, c)) {
 994                         if (c->assigned != -1) {
 995                                 c->assigned = -1;
 996                                 netifd_log_message(L_WARNING, "Failed to assign requested subprefix "
 997                                                 "of size %hhu for %s, trying other\n", c->length, c->name);
 998                         }
 999
1000                         entry = calloc(1, sizeof(*entry));
1001                         if (!entry) {
1002                                 free(c);
1003                                 continue;
1004                         }
1005
1006                         entry->node.key = c;
1007                         avl_insert(&assign_later, &entry->node);
1008                 }
1009
1010                 if (c->assigned != -1)
1011                         assigned_any = true;
1012         }
1013
1014         /* Then try to assign all other + failed custom assignments */
1015         avl_for_each_element_safe(&assign_later, entry, node, n_entry) {
1016                 bool assigned = false;
1017
1018                 c = (struct device_prefix_assignment *)entry->node.key;
1019                 avl_delete(&assign_later, &entry->node);
1020
1021                 do {
1022                         assigned = interface_prefix_assign(&prefix->assignments, c);
1023                 } while (!assigned && ++c->length <= 64);
1024
1025                 if (!assigned) {
1026                         netifd_log_message(L_WARNING, "Failed to assign subprefix "
1027                                         "of size %hhu for %s\n", c->length, c->name);
1028                         free(c);
1029                 } else
1030                         assigned_any = true;
1031
1032                 free(entry);
1033         }
1034
1035         list_for_each_entry(c, &prefix->assignments, head)
1036                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
1037                         interface_set_prefix_address(c, prefix, iface, true);
1038
1039         if (!assigned_any)
1040                 netifd_log_message(L_WARNING, "You have delegated IPv6-prefixes but haven't assigned them "
1041                                 "to any interface. Did you forget to set option ip6assign on your lan-interfaces?");
1042 }
1043
1044
1045 void interface_refresh_assignments(bool hint)
1046 {
1047         static bool refresh = false;
1048         if (!hint && refresh) {
1049                 struct device_prefix *p;
1050                 list_for_each_entry(p, &prefixes, head)
1051                         interface_update_prefix_assignments(p, true);
1052         }
1053         refresh = hint;
1054 }
1055
1056 void interface_update_prefix_delegation(struct interface_ip_settings *ip)
1057 {
1058         struct device_prefix *prefix;
1059
1060         vlist_for_each_element(&ip->prefix, prefix, node) {
1061                 interface_update_prefix_assignments(prefix, !ip->no_delegation);
1062
1063                 if (ip->no_delegation) {
1064                         if (prefix->head.next)
1065                                 list_del(&prefix->head);
1066                 } else
1067                         list_add(&prefix->head, &prefixes);
1068         }
1069 }
1070
1071 static void
1072 interface_update_prefix(struct vlist_tree *tree,
1073                              struct vlist_node *node_new,
1074                              struct vlist_node *node_old)
1075 {
1076         struct device_prefix *prefix_old, *prefix_new;
1077         prefix_old = container_of(node_old, struct device_prefix, node);
1078         prefix_new = container_of(node_new, struct device_prefix, node);
1079
1080         struct interface_ip_settings *ip = container_of(tree, struct interface_ip_settings, prefix);
1081         if (tree && (!node_new || !node_old))
1082                 ip->iface->updated |= IUF_PREFIX;
1083
1084         struct device_route route;
1085         memset(&route, 0, sizeof(route));
1086         route.flags = DEVADDR_INET6;
1087         route.metric = INT32_MAX;
1088         route.mask = (node_new) ? prefix_new->length : prefix_old->length;
1089         route.addr.in6 = (node_new) ? prefix_new->addr : prefix_old->addr;
1090
1091
1092         struct device_prefix_assignment *c;
1093         struct interface *iface;
1094
1095         if (node_old && node_new) {
1096                 /* Move assignments and refresh addresses to update valid times */
1097                 list_splice(&prefix_old->assignments, &prefix_new->assignments);
1098
1099                 list_for_each_entry(c, &prefix_new->assignments, head)
1100                         if ((iface = vlist_find(&interfaces, c->name, iface, node)))
1101                                 interface_set_prefix_address(c, prefix_new, iface, true);
1102
1103                 if (prefix_new->preferred_until != prefix_old->preferred_until ||
1104                                 prefix_new->valid_until != prefix_old->valid_until)
1105                         ip->iface->updated |= IUF_PREFIX;
1106         } else if (node_new) {
1107                 /* Set null-route to avoid routing loops */
1108                 system_add_route(NULL, &route);
1109
1110                 if (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation)
1111                         interface_update_prefix_assignments(prefix_new, true);
1112         } else if (node_old) {
1113                 /* Remove null-route */
1114                 interface_update_prefix_assignments(prefix_old, false);
1115                 system_del_route(NULL, &route);
1116         }
1117
1118         if (node_old) {
1119                 if (prefix_old->head.next)
1120                         list_del(&prefix_old->head);
1121                 free(prefix_old);
1122         }
1123
1124         if (node_new && (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation))
1125                 list_add(&prefix_new->head, &prefixes);
1126
1127 }
1128
1129 struct device_prefix*
1130 interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
1131                 uint8_t length, time_t valid_until, time_t preferred_until,
1132                 struct in6_addr *excl_addr, uint8_t excl_length, const char *pclass)
1133 {
1134         if (!pclass)
1135                 pclass = (iface) ? iface->name : "local";
1136
1137         struct device_prefix *prefix = calloc(1, sizeof(*prefix) + strlen(pclass) + 1);
1138         if (!prefix)
1139                 return NULL;
1140
1141         prefix->length = length;
1142         prefix->addr = *addr;
1143         prefix->preferred_until = preferred_until;
1144         prefix->valid_until = valid_until;
1145         prefix->iface = iface;
1146         INIT_LIST_HEAD(&prefix->assignments);
1147
1148         if (excl_addr) {
1149                 prefix->excl_addr = *excl_addr;
1150                 prefix->excl_length = excl_length;
1151         }
1152
1153         strcpy(prefix->pclass, pclass);
1154
1155         if (iface)
1156                 vlist_add(&iface->proto_ip.prefix, &prefix->node, &prefix->addr);
1157         else
1158                 interface_update_prefix(NULL, &prefix->node, NULL);
1159
1160         return prefix;
1161 }
1162
1163 void
1164 interface_ip_set_ula_prefix(const char *prefix)
1165 {
1166         char buf[INET6_ADDRSTRLEN + 4] = {0}, *saveptr;
1167         if (prefix)
1168                 strncpy(buf, prefix, sizeof(buf) - 1);
1169         char *prefixaddr = strtok_r(buf, "/", &saveptr);
1170
1171         struct in6_addr addr;
1172         if (!prefixaddr || inet_pton(AF_INET6, prefixaddr, &addr) < 1) {
1173                 if (ula_prefix) {
1174                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
1175                         ula_prefix = NULL;
1176                 }
1177                 return;
1178         }
1179
1180         int length;
1181         char *prefixlen = strtok_r(NULL, ",", &saveptr);
1182         if (!prefixlen || (length = atoi(prefixlen)) < 1 || length > 64)
1183                 return;
1184
1185         if (!ula_prefix || !IN6_ARE_ADDR_EQUAL(&addr, &ula_prefix->addr) ||
1186                         ula_prefix->length != length) {
1187                 if (ula_prefix)
1188                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
1189
1190                 ula_prefix = interface_ip_add_device_prefix(NULL, &addr, length,
1191                                 0, 0, NULL, 0, NULL);
1192         }
1193 }
1194
1195 static void
1196 interface_add_dns_server(struct interface_ip_settings *ip, const char *str)
1197 {
1198         struct dns_server *s;
1199
1200         s = calloc(1, sizeof(*s));
1201         if (!s)
1202                 return;
1203
1204         s->af = AF_INET;
1205         if (inet_pton(s->af, str, &s->addr.in))
1206                 goto add;
1207
1208         s->af = AF_INET6;
1209         if (inet_pton(s->af, str, &s->addr.in))
1210                 goto add;
1211
1212         free(s);
1213         return;
1214
1215 add:
1216         D(INTERFACE, "Add IPv%c DNS server: %s\n",
1217           s->af == AF_INET6 ? '6' : '4', str);
1218         vlist_simple_add(&ip->dns_servers, &s->node);
1219 }
1220
1221 void
1222 interface_add_dns_server_list(struct interface_ip_settings *ip, struct blob_attr *list)
1223 {
1224         struct blob_attr *cur;
1225         int rem;
1226
1227         blobmsg_for_each_attr(cur, list, rem) {
1228                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1229                         continue;
1230
1231                 if (!blobmsg_check_attr(cur, NULL))
1232                         continue;
1233
1234                 interface_add_dns_server(ip, blobmsg_data(cur));
1235         }
1236 }
1237
1238 static void
1239 interface_add_dns_search_domain(struct interface_ip_settings *ip, const char *str)
1240 {
1241         struct dns_search_domain *s;
1242         int len = strlen(str);
1243
1244         s = calloc(1, sizeof(*s) + len + 1);
1245         if (!s)
1246                 return;
1247
1248         D(INTERFACE, "Add DNS search domain: %s\n", str);
1249         memcpy(s->name, str, len);
1250         vlist_simple_add(&ip->dns_search, &s->node);
1251 }
1252
1253 void
1254 interface_add_dns_search_list(struct interface_ip_settings *ip, struct blob_attr *list)
1255 {
1256         struct blob_attr *cur;
1257         int rem;
1258
1259         blobmsg_for_each_attr(cur, list, rem) {
1260                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1261                         continue;
1262
1263                 if (!blobmsg_check_attr(cur, NULL))
1264                         continue;
1265
1266                 interface_add_dns_search_domain(ip, blobmsg_data(cur));
1267         }
1268 }
1269
1270 static void
1271 write_resolv_conf_entries(FILE *f, struct interface_ip_settings *ip, const char *dev)
1272 {
1273         struct dns_server *s;
1274         struct dns_search_domain *d;
1275         const char *str;
1276         char buf[INET6_ADDRSTRLEN];
1277
1278         vlist_simple_for_each_element(&ip->dns_servers, s, node) {
1279                 str = inet_ntop(s->af, &s->addr, buf, sizeof(buf));
1280                 if (!str)
1281                         continue;
1282
1283                 if (s->af == AF_INET6 && IN6_IS_ADDR_LINKLOCAL(&s->addr.in6))
1284                         fprintf(f, "nameserver %s%%%s\n", str, dev);
1285                 else
1286                         fprintf(f, "nameserver %s\n", str);
1287         }
1288
1289         vlist_simple_for_each_element(&ip->dns_search, d, node) {
1290                 fprintf(f, "search %s\n", d->name);
1291         }
1292 }
1293
1294 /* Sorting of interface resolver entries :               */
1295 /* Primary on interface dns_metric : lowest metric first */
1296 /* Secondary on interface metric : lowest metric first   */
1297 /* Finally alphabetical order of interface names         */
1298 static int resolv_conf_iface_cmp(const void *k1, const void *k2, void *ptr)
1299 {
1300         const struct interface *iface1 = k1, *iface2 = k2;
1301
1302         if (iface1->dns_metric != iface2->dns_metric)
1303                 return iface1->dns_metric - iface2->dns_metric;
1304
1305         if (iface1->metric != iface2->metric)
1306                 return iface1->metric - iface2->metric;
1307
1308         return strcmp(iface1->name, iface2->name);
1309 }
1310
1311 static void
1312 __interface_write_dns_entries(FILE *f)
1313 {
1314         struct interface *iface;
1315         struct {
1316                 struct avl_node node;
1317         } *entry, *n_entry;
1318         struct avl_tree resolv_conf_iface_entries;
1319
1320         avl_init(&resolv_conf_iface_entries, resolv_conf_iface_cmp, false, NULL);
1321
1322         vlist_for_each_element(&interfaces, iface, node) {
1323                 if (iface->state != IFS_UP)
1324                         continue;
1325
1326                 if (vlist_simple_empty(&iface->proto_ip.dns_search) &&
1327                     vlist_simple_empty(&iface->proto_ip.dns_servers) &&
1328                     vlist_simple_empty(&iface->config_ip.dns_search) &&
1329                     vlist_simple_empty(&iface->config_ip.dns_servers))
1330                         continue;
1331
1332                 entry = calloc(1, sizeof(*entry));
1333                 if (!entry)
1334                         continue;
1335
1336                 entry->node.key = iface;
1337                 avl_insert(&resolv_conf_iface_entries, &entry->node);
1338         }
1339
1340         avl_for_each_element(&resolv_conf_iface_entries, entry, node) {
1341                 iface = (struct interface *)entry->node.key;
1342                 struct device *dev = iface->l3_dev.dev;
1343
1344                 fprintf(f, "# Interface %s\n", iface->name);
1345
1346                 write_resolv_conf_entries(f, &iface->config_ip, dev->ifname);
1347
1348                 if (!iface->proto_ip.no_dns)
1349                         write_resolv_conf_entries(f, &iface->proto_ip, dev->ifname);
1350         }
1351
1352         avl_remove_all_elements(&resolv_conf_iface_entries, entry, node, n_entry)
1353                 free(entry);
1354 }
1355
1356 void
1357 interface_write_resolv_conf(void)
1358 {
1359         char *path = alloca(strlen(resolv_conf) + 5);
1360         FILE *f;
1361         uint32_t crcold, crcnew;
1362
1363         sprintf(path, "%s.tmp", resolv_conf);
1364         unlink(path);
1365         f = fopen(path, "w+");
1366         if (!f) {
1367                 D(INTERFACE, "Failed to open %s for writing\n", path);
1368                 return;
1369         }
1370
1371         __interface_write_dns_entries(f);
1372
1373         fflush(f);
1374         rewind(f);
1375         crcnew = crc32_file(f);
1376         fclose(f);
1377
1378         crcold = crcnew + 1;
1379         f = fopen(resolv_conf, "r");
1380         if (f) {
1381                 crcold = crc32_file(f);
1382                 fclose(f);
1383         }
1384
1385         if (crcold == crcnew) {
1386                 unlink(path);
1387         } else if (rename(path, resolv_conf) < 0) {
1388                 D(INTERFACE, "Failed to replace %s\n", resolv_conf);
1389                 unlink(path);
1390         }
1391 }
1392
1393 void interface_ip_set_enabled(struct interface_ip_settings *ip, bool enabled)
1394 {
1395         struct device_addr *addr;
1396         struct device_route *route;
1397         struct device *dev;
1398         struct interface *iface;
1399
1400         ip->enabled = enabled;
1401         iface = ip->iface;
1402         dev = iface->l3_dev.dev;
1403         if (!dev)
1404                 return;
1405
1406         vlist_for_each_element(&ip->addr, addr, node) {
1407                 bool v6 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6) ? true : false;
1408
1409                 if (addr->flags & DEVADDR_EXTERNAL)
1410                         continue;
1411
1412                 if (addr->enabled == enabled)
1413                         continue;
1414
1415                 if (enabled) {
1416                         system_add_address(dev, addr);
1417
1418                         addr->policy_table = (v6) ? iface->ip6table : iface->ip4table;
1419                         if (iface->metric || addr->policy_table)
1420                                 interface_handle_subnet_route(iface, addr, true);
1421
1422                         if (addr->policy_table)
1423                                 interface_add_addr_rules(addr, true);
1424                 } else {
1425                         interface_handle_subnet_route(iface, addr, false);
1426                         system_del_address(dev, addr);
1427
1428                         if (addr->policy_table)
1429                                 interface_add_addr_rules(addr, false);
1430                 }
1431                 addr->enabled = enabled;
1432         }
1433
1434         vlist_for_each_element(&ip->route, route, node) {
1435                 bool _enabled = enabled;
1436
1437                 if (route->flags & DEVADDR_EXTERNAL)
1438                         continue;
1439
1440                 if (!enable_route(ip, route))
1441                         _enabled = false;
1442
1443                 if (route->enabled == _enabled)
1444                         continue;
1445
1446                 if (_enabled) {
1447                         interface_set_route_info(ip->iface, route);
1448
1449                         if (system_add_route(dev, route))
1450                                 route->failed = true;
1451                 } else
1452                         system_del_route(dev, route);
1453                 route->enabled = _enabled;
1454         }
1455
1456         struct device_prefix *c;
1457         struct device_prefix_assignment *a;
1458         list_for_each_entry(c, &prefixes, head)
1459                 list_for_each_entry(a, &c->assignments, head)
1460                         if (!strcmp(a->name, ip->iface->name))
1461                                 interface_set_prefix_address(a, c, ip->iface, enabled);
1462
1463         if (ip->iface->policy_rules_set != enabled &&
1464             ip->iface->l3_dev.dev) {
1465                 set_ip_lo_policy(enabled, true, ip->iface);
1466                 set_ip_lo_policy(enabled, false, ip->iface);
1467
1468                 set_ip_source_policy(enabled, true, IPRULE_PRIORITY_REJECT + ip->iface->l3_dev.dev->ifindex,
1469                         NULL, 0, 0, ip->iface, "failed_policy", true);
1470                 ip->iface->policy_rules_set = enabled;
1471         }
1472 }
1473
1474 void
1475 interface_ip_update_start(struct interface_ip_settings *ip)
1476 {
1477         if (ip != &ip->iface->config_ip) {
1478                 vlist_simple_update(&ip->dns_servers);
1479                 vlist_simple_update(&ip->dns_search);
1480         }
1481         vlist_update(&ip->route);
1482         vlist_update(&ip->addr);
1483         vlist_update(&ip->prefix);
1484 }
1485
1486 void
1487 interface_ip_update_complete(struct interface_ip_settings *ip)
1488 {
1489         vlist_simple_flush(&ip->dns_servers);
1490         vlist_simple_flush(&ip->dns_search);
1491         vlist_flush(&ip->route);
1492         vlist_flush(&ip->addr);
1493         vlist_flush(&ip->prefix);
1494         interface_write_resolv_conf();
1495 }
1496
1497 void
1498 interface_ip_flush(struct interface_ip_settings *ip)
1499 {
1500         if (ip == &ip->iface->proto_ip)
1501                 vlist_flush_all(&ip->iface->host_routes);
1502         vlist_simple_flush_all(&ip->dns_servers);
1503         vlist_simple_flush_all(&ip->dns_search);
1504         vlist_flush_all(&ip->route);
1505         vlist_flush_all(&ip->addr);
1506         vlist_flush_all(&ip->prefix);
1507 }
1508
1509 static void
1510 __interface_ip_init(struct interface_ip_settings *ip, struct interface *iface)
1511 {
1512         ip->iface = iface;
1513         ip->enabled = true;
1514         vlist_simple_init(&ip->dns_search, struct dns_search_domain, node);
1515         vlist_simple_init(&ip->dns_servers, struct dns_server, node);
1516         vlist_init(&ip->route, route_cmp, interface_update_proto_route);
1517         vlist_init(&ip->addr, addr_cmp, interface_update_proto_addr);
1518         vlist_init(&ip->prefix, prefix_cmp, interface_update_prefix);
1519 }
1520
1521 void
1522 interface_ip_init(struct interface *iface)
1523 {
1524         __interface_ip_init(&iface->proto_ip, iface);
1525         __interface_ip_init(&iface->config_ip, iface);
1526         vlist_init(&iface->host_routes, route_cmp, interface_update_host_route);
1527 }
1528
1529 static void
1530 interface_ip_valid_until_handler(struct uloop_timeout *t)
1531 {
1532         time_t now = system_get_rtime();
1533         struct interface *iface;
1534         vlist_for_each_element(&interfaces, iface, node) {
1535                 if (iface->state != IFS_UP)
1536                         continue;
1537
1538                 struct device_addr *addr, *addrp;
1539                 struct device_route *route, *routep;
1540                 struct device_prefix *pref, *prefp;
1541
1542                 vlist_for_each_element_safe(&iface->proto_ip.addr, addr, node, addrp)
1543                         if (addr->valid_until && addr->valid_until < now)
1544                                 vlist_delete(&iface->proto_ip.addr, &addr->node);
1545
1546                 vlist_for_each_element_safe(&iface->proto_ip.route, route, node, routep)
1547                         if (route->valid_until && route->valid_until < now)
1548                                 vlist_delete(&iface->proto_ip.route, &route->node);
1549
1550                 vlist_for_each_element_safe(&iface->proto_ip.prefix, pref, node, prefp)
1551                         if (pref->valid_until && pref->valid_until < now)
1552                                 vlist_delete(&iface->proto_ip.prefix, &pref->node);
1553
1554         }
1555
1556         uloop_timeout_set(t, 1000);
1557 }
1558
1559 static void __init
1560 interface_ip_init_worker(void)
1561 {
1562         valid_until_timeout.cb = interface_ip_valid_until_handler;
1563         uloop_timeout_set(&valid_until_timeout, 1000);
1564 }