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 #include <unistd.h>
  19
  20 #include <limits.h>
  21 #include <arpa/inet.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_MAX
  41 };
  42
  43 static const struct blobmsg_policy route_attr[__ROUTE_MAX] = {
  44         [ROUTE_INTERFACE] = { .name = "interface", .type = BLOBMSG_TYPE_STRING },
  45         [ROUTE_TARGET] = { .name = "target", .type = BLOBMSG_TYPE_STRING },
  46         [ROUTE_MASK] = { .name = "netmask", .type = BLOBMSG_TYPE_STRING },
  47         [ROUTE_GATEWAY] = { .name = "gateway", .type = BLOBMSG_TYPE_STRING },
  48         [ROUTE_METRIC] = { .name = "metric", .type = BLOBMSG_TYPE_INT32 },
  49         [ROUTE_MTU] = { .name = "mtu", .type = BLOBMSG_TYPE_INT32 },
  50         [ROUTE_TABLE] = { .name = "table", .type = BLOBMSG_TYPE_STRING },
  51         [ROUTE_VALID] = { .name = "valid", .type = BLOBMSG_TYPE_INT32 },
  52 };
  53
  54 const struct uci_blob_param_list route_attr_list = {
  55         .n_params = __ROUTE_MAX,
  56         .params = route_attr,
  57 };
  58
  59
  60 struct list_head prefixes = LIST_HEAD_INIT(prefixes);
  61 static struct device_prefix *ula_prefix = NULL;
  62 static struct uloop_timeout valid_until_timeout;
  63
  64
  65 static void
  66 clear_if_addr(union if_addr *a, int mask)
  67 {
  68         int m_bytes = (mask + 7) / 8;
  69         uint8_t m_clear = (1 << (m_bytes * 8 - mask)) - 1;
  70         uint8_t *p = (uint8_t *) a;
  71
  72         if (m_bytes < sizeof(a))
  73                 memset(p + m_bytes, 0, sizeof(a) - m_bytes);
  74
  75         p[m_bytes - 1] &= ~m_clear;
  76 }
  77
  78 static bool
  79 match_if_addr(union if_addr *a1, union if_addr *a2, int mask)
  80 {
  81         union if_addr *p1, *p2;
  82
  83         p1 = alloca(sizeof(*a1));
  84         p2 = alloca(sizeof(*a2));
  85
  86         memcpy(p1, a1, sizeof(*a1));
  87         clear_if_addr(p1, mask);
  88         memcpy(p2, a2, sizeof(*a2));
  89         clear_if_addr(p2, mask);
  90
  91         return !memcmp(p1, p2, sizeof(*p1));
  92 }
  93
  94 static int set_ip_source_policy(bool add, bool v6, unsigned int priority,
  95                 const union if_addr *addr, uint8_t mask, struct interface *iface,
  96                 struct interface *in_iface, const char *action)
  97 {
  98         struct iprule rule = {
  99                 .flags = IPRULE_PRIORITY,
 100                 .priority = priority
 101         };
 102
 103         if (addr) {
 104                 rule.flags |= IPRULE_SRC;
 105                 rule.src_addr = *addr;
 106                 rule.src_mask = mask;
 107         }
 108
 109         if (iface) {
 110                 rule.flags |= IPRULE_LOOKUP;
 111                 rule.lookup = (v6) ? iface->ip6table : iface->ip4table;
 112
 113                 if (!rule.lookup)
 114                         return 0;
 115         } else if (action) {
 116                 rule.flags |= IPRULE_ACTION;
 117                 system_resolve_iprule_action(action, &rule.action);
 118         }
 119
 120         if (in_iface && in_iface->l3_dev.dev) {
 121                 rule.flags |= IPRULE_IN;
 122                 strcpy(rule.in_dev, in_iface->l3_dev.dev->ifname);
 123         }
 124
 125         rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
 126
 127         return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
 128 }
 129
 130 static int set_ip_lo_policy(bool add, bool v6, struct interface *iface)
 131 {
 132         struct iprule rule = {
 133                 .flags = IPRULE_IN | IPRULE_LOOKUP | IPRULE_PRIORITY,
 134                 .priority = IPRULE_PRIORITY_NW + iface->l3_dev.dev->ifindex,
 135                 .lookup = (v6) ? iface->ip6table : iface->ip4table,
 136                 .in_dev = "lo"
 137         };
 138
 139         if (!rule.lookup)
 140                 return 0;
 141
 142         rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
 143
 144         return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
 145 }
 146
 147 static bool
 148 __find_ip_addr_target(struct interface_ip_settings *ip, union if_addr *a, bool v6)
 149 {
 150         struct device_addr *addr;
 151
 152         vlist_for_each_element(&ip->addr, addr, node) {
 153                 if (!addr->enabled)
 154                         continue;
 155
 156                 if (v6 != ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
 157                         continue;
 158
 159                 // Handle offlink addresses correctly
 160                 unsigned int mask = addr->mask;
 161                 if ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6 &&
 162                                 (addr->flags & DEVADDR_OFFLINK))
 163                         mask = 128;
 164
 165                 if (!match_if_addr(&addr->addr, a, mask))
 166                         continue;
 167
 168                 return true;
 169         }
 170
 171         return false;
 172 }
 173
 174 static void
 175 __find_ip_route_target(struct interface_ip_settings *ip, union if_addr *a,
 176                        bool v6, struct device_route **res)
 177 {
 178         struct device_route *route;
 179
 180         vlist_for_each_element(&ip->route, route, node) {
 181                 if (!route->enabled)
 182                         continue;
 183
 184                 if (v6 != ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
 185                         continue;
 186
 187                 if (!match_if_addr(&route->addr, a, route->mask))
 188                         continue;
 189
 190                 if (route->flags & DEVROUTE_TABLE)
 191                         continue;
 192
 193                 if (!*res || route->mask < (*res)->mask)
 194                         *res = route;
 195         }
 196 }
 197
 198 static bool
 199 interface_ip_find_addr_target(struct interface *iface, union if_addr *a, bool v6)
 200 {
 201         return __find_ip_addr_target(&iface->proto_ip, a, v6) ||
 202                __find_ip_addr_target(&iface->config_ip, a, v6);
 203 }
 204
 205 static void
 206 interface_ip_find_route_target(struct interface *iface, union if_addr *a,
 207                                bool v6, struct device_route **route)
 208 {
 209         __find_ip_route_target(&iface->proto_ip, a, v6, route);
 210         __find_ip_route_target(&iface->config_ip, a, v6, route);
 211 }
 212
 213 struct interface *
 214 interface_ip_add_target_route(union if_addr *addr, bool v6, struct interface *iface)
 215 {
 216         struct device_route *route, *r_next = NULL;
 217         bool defaultroute_target = false;
 218         int addrsize = v6 ? sizeof(addr->in6) : sizeof(addr->in);
 219
 220         route = calloc(1, sizeof(*route));
 221         if (!route)
 222                 return NULL;
 223
 224         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
 225         route->mask = v6 ? 128 : 32;
 226         if (memcmp(&route->addr, addr, addrsize) == 0)
 227                 defaultroute_target = true;
 228         else
 229                 memcpy(&route->addr, addr, addrsize);
 230
 231         if (iface) {
 232                 /* look for locally addressable target first */
 233                 if (interface_ip_find_addr_target(iface, addr, v6))
 234                         goto done;
 235
 236                 /* do not stop at the first route, let the lookup compare
 237                  * masks to find the best match */
 238                 interface_ip_find_route_target(iface, addr, v6, &r_next);
 239         } else {
 240                 vlist_for_each_element(&interfaces, iface, node) {
 241                         /* look for locally addressable target first */
 242                         if (interface_ip_find_addr_target(iface, addr, v6))
 243                                 goto done;
 244
 245                         /* do not stop at the first route, let the lookup compare
 246                          * masks to find the best match */
 247                         interface_ip_find_route_target(iface, addr, v6, &r_next);
 248                 }
 249         }
 250
 251         if (!r_next) {
 252                 free(route);
 253                 return NULL;
 254         }
 255
 256         iface = r_next->iface;
 257         memcpy(&route->nexthop, &r_next->nexthop, sizeof(route->nexthop));
 258         route->mtu = r_next->mtu;
 259         route->metric = r_next->metric;
 260         route->table = r_next->table;
 261
 262 done:
 263         route->iface = iface;
 264         if (defaultroute_target)
 265                 free(route);
 266         else
 267                 vlist_add(&iface->host_routes, &route->node, route);
 268         return iface;
 269 }
 270
 271 void
 272 interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
 273 {
 274         struct interface_ip_settings *ip;
 275         struct blob_attr *tb[__ROUTE_MAX], *cur;
 276         struct device_route *route;
 277         int af = v6 ? AF_INET6 : AF_INET;
 278         bool is_proto_route = !!iface;
 279
 280         blobmsg_parse(route_attr, __ROUTE_MAX, tb, blobmsg_data(attr), blobmsg_data_len(attr));
 281
 282         if (!iface) {
 283                 if ((cur = tb[ROUTE_INTERFACE]) == NULL)
 284                         return;
 285
 286                 iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
 287                 if (!iface)
 288                         return;
 289
 290                 ip = &iface->config_ip;
 291         } else {
 292                 ip = &iface->proto_ip;
 293         }
 294
 295         route = calloc(1, sizeof(*route));
 296         if (!route)
 297                 return;
 298
 299         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
 300         route->mask = v6 ? 128 : 32;
 301         if ((cur = tb[ROUTE_MASK]) != NULL) {
 302                 route->mask = parse_netmask_string(blobmsg_data(cur), v6);
 303                 if (route->mask > (v6 ? 128 : 32))
 304                         goto error;
 305         }
 306
 307         if ((cur = tb[ROUTE_TARGET]) != NULL) {
 308                 if (!parse_ip_and_netmask(af, blobmsg_data(cur), &route->addr, &route->mask)) {
 309                         DPRINTF("Failed to parse route target: %s\n", (char *) blobmsg_data(cur));
 310                         goto error;
 311                 }
 312         }
 313
 314         if ((cur = tb[ROUTE_GATEWAY]) != NULL) {
 315                 if (!inet_pton(af, blobmsg_data(cur), &route->nexthop)) {
 316                         DPRINTF("Failed to parse route gateway: %s\n", (char *) blobmsg_data(cur));
 317                         goto error;
 318                 }
 319         }
 320
 321         if ((cur = tb[ROUTE_METRIC]) != NULL) {
 322                 route->metric = blobmsg_get_u32(cur);
 323                 route->flags |= DEVROUTE_METRIC;
 324         }
 325
 326         if ((cur = tb[ROUTE_MTU]) != NULL) {
 327                 route->mtu = blobmsg_get_u32(cur);
 328                 route->flags |= DEVROUTE_MTU;
 329         }
 330
 331         // Use source-based routing
 332         if (is_proto_route) {
 333                 route->table = (v6) ? iface->ip6table : iface->ip4table;
 334                 route->flags |= DEVROUTE_SRCTABLE;
 335         }
 336
 337         if ((cur = tb[ROUTE_TABLE]) != NULL) {
 338                 if (!system_resolve_rt_table(blobmsg_data(cur), &route->table)) {
 339                         DPRINTF("Failed to resolve routing table: %s\n", (char *) blobmsg_data(cur));
 340                         goto error;
 341                 }
 342
 343                 if (route->table)
 344                         route->flags |= DEVROUTE_TABLE;
 345         }
 346
 347         if ((cur = tb[ROUTE_VALID]) != NULL) {
 348                 int64_t valid = blobmsg_get_u32(cur);
 349                 int64_t valid_until = valid + (int64_t)system_get_rtime();
 350                 if (valid_until <= LONG_MAX && valid != 0xffffffffLL) // Catch overflow
 351                         route->valid_until = valid_until;
 352         }
 353
 354         vlist_add(&ip->route, &route->node, route);
 355         return;
 356
 357 error:
 358         free(route);
 359 }
 360
 361 static int
 362 addr_cmp(const void *k1, const void *k2, void *ptr)
 363 {
 364         return memcmp(k1, k2, sizeof(struct device_addr) -
 365                       offsetof(struct device_addr, flags));
 366 }
 367
 368 static int
 369 route_cmp(const void *k1, const void *k2, void *ptr)
 370 {
 371         const struct device_route *r1 = k1, *r2 = k2;
 372
 373         if (r1->mask != r2->mask)
 374                 return r2->mask - r1->mask;
 375
 376         if (r1->metric != r2->metric)
 377                 return r1->metric - r2->metric;
 378
 379         if (r1->flags != r2->flags)
 380                 return r2->flags - r1->flags;
 381
 382         return memcmp(&r1->addr, &r2->addr, sizeof(r1->addr));
 383 }
 384
 385 static int
 386 prefix_cmp(const void *k1, const void *k2, void *ptr)
 387 {
 388         return memcmp(k1, k2, offsetof(struct device_prefix, pclass) -
 389                         offsetof(struct device_prefix, addr));
 390 }
 391
 392 static void
 393 interface_handle_subnet_route(struct interface *iface, struct device_addr *addr, bool add)
 394 {
 395         struct device *dev = iface->l3_dev.dev;
 396         struct device_route route;
 397
 398         memset(&route, 0, sizeof(route));
 399         route.iface = iface;
 400         route.flags = addr->flags;
 401         route.mask = addr->mask;
 402         memcpy(&route.addr, &addr->addr, sizeof(route.addr));
 403         clear_if_addr(&route.addr, route.mask);
 404
 405         if (add) {
 406                 route.flags |= DEVADDR_KERNEL;
 407                 system_del_route(dev, &route);
 408
 409                 if (!(addr->flags & DEVADDR_OFFLINK)) {
 410                         route.flags &= ~DEVADDR_KERNEL;
 411                         route.metric = iface->metric;
 412                         system_add_route(dev, &route);
 413                 }
 414         } else {
 415                 if (!(addr->flags & DEVADDR_OFFLINK))
 416                         system_del_route(dev, &route);
 417         }
 418 }
 419
 420 static void
 421 interface_update_proto_addr(struct vlist_tree *tree,
 422                             struct vlist_node *node_new,
 423                             struct vlist_node *node_old)
 424 {
 425         struct interface_ip_settings *ip;
 426         struct interface *iface;
 427         struct device *dev;
 428         struct device_addr *a_new = NULL, *a_old = NULL;
 429         bool replace = false;
 430         bool keep = false;
 431         bool v6 = false;
 432
 433         ip = container_of(tree, struct interface_ip_settings, addr);
 434         iface = ip->iface;
 435         dev = iface->l3_dev.dev;
 436
 437         if (node_new) {
 438                 a_new = container_of(node_new, struct device_addr, node);
 439
 440                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
 441                     !a_new->broadcast) {
 442
 443                         uint32_t mask = ~0;
 444                         uint32_t *a = (uint32_t *) &a_new->addr;
 445
 446                         mask >>= a_new->mask;
 447                         a_new->broadcast = *a | htonl(mask);
 448                 }
 449         }
 450
 451         if (node_old)
 452                 a_old = container_of(node_old, struct device_addr, node);
 453
 454         if (a_new && a_old) {
 455                 keep = true;
 456
 457                 if (a_old->flags != a_new->flags)
 458                         keep = false;
 459
 460                 if (a_old->valid_until != a_new->valid_until ||
 461                                 a_old->preferred_until != a_new->preferred_until)
 462                         replace = true;
 463
 464                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
 465                     a_new->broadcast != a_old->broadcast)
 466                         keep = false;
 467         }
 468
 469         if (node_old) {
 470                 if (!(a_old->flags & DEVADDR_EXTERNAL) && a_old->enabled && !keep) {
 471                         interface_handle_subnet_route(iface, a_old, false);
 472
 473                         if ((a_old->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
 474                                 v6 = true;
 475
 476                         //This is needed for source routing to work correctly. If a device
 477                         //has two connections to a network using the same subnet, adding
 478                         //only the network-rule will cause packets to be routed through the
 479                         //first matching network (source IP matches both masks).
 480                         set_ip_source_policy(false, v6, IPRULE_PRIORITY_ADDR, &a_old->addr,
 481                                 (v6) ? 128 : 32, iface, NULL, NULL);
 482                         set_ip_source_policy(false, v6, IPRULE_PRIORITY_NW, &a_old->addr,
 483                                 a_old->mask, iface, NULL, NULL);
 484
 485                         system_del_address(dev, a_old);
 486                 }
 487                 free(a_old);
 488         }
 489
 490         if (node_new) {
 491                 a_new->enabled = true;
 492                 if (!(a_new->flags & DEVADDR_EXTERNAL) && (!keep || replace)) {
 493                         system_add_address(dev, a_new);
 494
 495                         if (!keep) {
 496                                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
 497                                         v6 = true;
 498
 499                                 set_ip_source_policy(true, v6, IPRULE_PRIORITY_ADDR, &a_new->addr,
 500                                                 (v6) ? 128 : 32, iface, NULL, NULL);
 501                                 set_ip_source_policy(true, v6, IPRULE_PRIORITY_NW, &a_new->addr,
 502                                                 a_new->mask, iface, NULL, NULL);
 503                         }
 504
 505                         if ((a_new->flags & DEVADDR_OFFLINK) || iface->metric)
 506                                 interface_handle_subnet_route(iface, a_new, true);
 507                 }
 508         }
 509 }
 510
 511 static bool
 512 enable_route(struct interface_ip_settings *ip, struct device_route *route)
 513 {
 514         if (ip->no_defaultroute && !route->mask)
 515                 return false;
 516
 517         return ip->enabled;
 518 }
 519
 520 static void
 521 interface_update_proto_route(struct vlist_tree *tree,
 522                              struct vlist_node *node_new,
 523                              struct vlist_node *node_old)
 524 {
 525         struct interface_ip_settings *ip;
 526         struct interface *iface;
 527         struct device *dev;
 528         struct device_route *route_old, *route_new;
 529         bool keep = false;
 530
 531         ip = container_of(tree, struct interface_ip_settings, route);
 532         iface = ip->iface;
 533         dev = iface->l3_dev.dev;
 534
 535         route_old = container_of(node_old, struct device_route, node);
 536         route_new = container_of(node_new, struct device_route, node);
 537
 538         if (node_old && node_new)
 539                 keep = !memcmp(&route_old->nexthop, &route_new->nexthop, sizeof(route_old->nexthop));
 540
 541         if (node_old) {
 542                 if (!(route_old->flags & DEVADDR_EXTERNAL) && route_old->enabled && !keep)
 543                         system_del_route(dev, route_old);
 544                 free(route_old);
 545         }
 546
 547         if (node_new) {
 548                 bool _enabled = enable_route(ip, route_new);
 549
 550                 if (!(route_new->flags & DEVROUTE_METRIC))
 551                         route_new->metric = iface->metric;
 552
 553                 if (!(route_new->flags & DEVADDR_EXTERNAL) && !keep && _enabled)
 554                         system_add_route(dev, route_new);
 555
 556                 route_new->iface = iface;
 557                 route_new->enabled = _enabled;
 558         }
 559 }
 560
 561 static void
 562 interface_update_host_route(struct vlist_tree *tree,
 563                              struct vlist_node *node_new,
 564                              struct vlist_node *node_old)
 565 {
 566         struct interface *iface;
 567         struct device *dev;
 568         struct device_route *route_old, *route_new;
 569
 570         iface = container_of(tree, struct interface, host_routes);
 571         dev = iface->l3_dev.dev;
 572
 573         route_old = container_of(node_old, struct device_route, node);
 574         route_new = container_of(node_new, struct device_route, node);
 575
 576         if (node_old) {
 577                 system_del_route(dev, route_old);
 578                 free(route_old);
 579         }
 580
 581         if (node_new)
 582                 system_add_route(dev, route_new);
 583 }
 584
 585
 586 static void
 587 interface_set_prefix_address(struct device_prefix_assignment *assignment,
 588                 const struct device_prefix *prefix, struct interface *iface, bool add)
 589 {
 590         const struct interface *uplink = prefix->iface;
 591         if (!iface->l3_dev.dev)
 592                 return;
 593
 594         struct device *l3_downlink = iface->l3_dev.dev;
 595
 596         struct device_addr addr;
 597         memset(&addr, 0, sizeof(addr));
 598         addr.addr.in6 = prefix->addr;
 599         addr.addr.in6.s6_addr32[1] |= htonl(assignment->assigned);
 600         addr.addr.in6.s6_addr[15] += 1;
 601         addr.mask = assignment->length;
 602         addr.flags = DEVADDR_INET6;
 603         addr.preferred_until = prefix->preferred_until;
 604         addr.valid_until = prefix->valid_until;
 605
 606         if (!add && assignment->enabled) {
 607                 time_t now = system_get_rtime();
 608                 addr.preferred_until = now;
 609                 if (!addr.valid_until || addr.valid_until - now > 7200)
 610                         addr.valid_until = now + 7200;
 611                 system_add_address(l3_downlink, &addr);
 612                 if (prefix->iface) {
 613                         set_ip_source_policy(false, true, IPRULE_PRIORITY_NW, &addr.addr,
 614                                                         addr.mask, prefix->iface, iface, NULL);
 615
 616                         set_ip_source_policy(false, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 617                                                         addr.mask, NULL, iface, "unreachable");
 618                 }
 619
 620                 assignment->enabled = false;
 621         } else if (add && (iface->state == IFS_UP || iface->state == IFS_SETUP)) {
 622                 system_add_address(l3_downlink, &addr);
 623                 if (prefix->iface && !assignment->enabled) {
 624                         set_ip_source_policy(true, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 625                                         addr.mask, NULL, iface, "unreachable");
 626
 627                         set_ip_source_policy(true, true, IPRULE_PRIORITY_NW, &addr.addr,
 628                                         addr.mask, prefix->iface, iface, NULL);
 629                 }
 630                 if (uplink && uplink->l3_dev.dev) {
 631                         int mtu = system_update_ipv6_mtu(
 632                                         uplink->l3_dev.dev, 0);
 633                         if (mtu > 0)
 634                                 system_update_ipv6_mtu(l3_downlink, mtu);
 635                 }
 636                 assignment->enabled = true;
 637         }
 638 }
 639
 640 static bool interface_prefix_assign(struct list_head *list,
 641                 struct device_prefix_assignment *assign)
 642 {
 643         int32_t current = 0, asize = (1 << (64 - assign->length)) - 1;
 644         struct device_prefix_assignment *c;
 645         list_for_each_entry(c, list, head) {
 646                 if (assign->assigned != -1) {
 647                         if (assign->assigned >= current && assign->assigned + asize < c->assigned) {
 648                                 list_add_tail(&assign->head, &c->head);
 649                                 return true;
 650                         }
 651                 } else if (assign->assigned == -1) {
 652                         current = (current + asize) & (~asize);
 653                         if (current + asize < c->assigned) {
 654                                 assign->assigned = current;
 655                                 list_add_tail(&assign->head, &c->head);
 656                                 return true;
 657                         }
 658                 }
 659                 current = (c->assigned + (1 << (64 - c->length)));
 660         }
 661         return false;
 662 }
 663
 664 static void interface_update_prefix_assignments(struct device_prefix *prefix, bool setup)
 665 {
 666         struct device_prefix_assignment *c;
 667         struct interface *iface;
 668
 669         // Delete all assignments
 670         while (!list_empty(&prefix->assignments)) {
 671                 c = list_first_entry(&prefix->assignments,
 672                                 struct device_prefix_assignment, head);
 673                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 674                         interface_set_prefix_address(c, prefix, iface, false);
 675                 list_del(&c->head);
 676                 free(c);
 677         }
 678
 679         if (!setup)
 680                 return;
 681
 682         // End-of-assignment sentinel
 683         c = malloc(sizeof(*c) + 1);
 684         c->assigned = 1 << (64 - prefix->length);
 685         c->length = 64;
 686         c->name[0] = 0;
 687         list_add(&c->head, &prefix->assignments);
 688
 689         // Excluded prefix
 690         if (prefix->excl_length > 0) {
 691                 const char name[] = "!excluded";
 692                 c = malloc(sizeof(*c) + sizeof(name));
 693                 c->assigned = ntohl(prefix->excl_addr.s6_addr32[1]) &
 694                                 ((1 << (64 - prefix->length)) - 1);
 695                 c->length = prefix->excl_length;
 696                 memcpy(c->name, name, sizeof(name));
 697                 list_add(&c->head, &prefix->assignments);
 698         }
 699
 700         bool assigned_any = false;
 701         struct list_head assign_later = LIST_HEAD_INIT(assign_later);
 702         vlist_for_each_element(&interfaces, iface, node) {
 703                 if (iface->assignment_length < 48 ||
 704                                 iface->assignment_length > 64)
 705                         continue;
 706
 707                 // Test whether there is a matching class
 708                 if (!list_empty(&iface->assignment_classes)) {
 709                         bool found = false;
 710
 711                         struct interface_assignment_class *c;
 712                         list_for_each_entry(c, &iface->assignment_classes, head) {
 713                                 if (!strcmp(c->name, prefix->pclass)) {
 714                                         found = true;
 715                                         break;
 716                                 }
 717                         }
 718
 719                         if (!found)
 720                                 continue;
 721                 }
 722
 723                 size_t namelen = strlen(iface->name) + 1;
 724                 c = malloc(sizeof(*c) + namelen);
 725                 c->length = iface->assignment_length;
 726                 c->assigned = iface->assignment_hint;
 727                 c->enabled = false;
 728                 memcpy(c->name, iface->name, namelen);
 729
 730                 // First process all custom assignments, put all others in later-list
 731                 if (c->assigned == -1 || !interface_prefix_assign(&prefix->assignments, c)) {
 732                         if (c->assigned != -1) {
 733                                 c->assigned = -1;
 734                                 netifd_log_message(L_WARNING, "Failed to assign requested subprefix "
 735                                                 "of size %hhu for %s, trying other\n", c->length, c->name);
 736                         }
 737
 738                         struct list_head *next = &assign_later;
 739                         struct device_prefix_assignment *n;
 740                         list_for_each_entry(n, &assign_later, head) {
 741                                 if (n->length < c->length) {
 742                                         next = &n->head;
 743                                         break;
 744                                 }
 745                         }
 746                         list_add_tail(&c->head, next);
 747                 }
 748
 749                 if (c->assigned != -1)
 750                         assigned_any = true;
 751         }
 752
 753         // Then try to assign all other + failed custom assignments
 754         while (!list_empty(&assign_later)) {
 755                 c = list_first_entry(&assign_later, struct device_prefix_assignment, head);
 756                 list_del(&c->head);
 757
 758                 bool assigned = false;
 759                 do {
 760                         assigned = interface_prefix_assign(&prefix->assignments, c);
 761                 } while (!assigned && ++c->length <= 64);
 762
 763                 if (!assigned) {
 764                         netifd_log_message(L_WARNING, "Failed to assign subprefix "
 765                                         "of size %hhu for %s\n", c->length, c->name);
 766                         free(c);
 767                 } else {
 768                         assigned_any = true;
 769                 }
 770         }
 771
 772         list_for_each_entry(c, &prefix->assignments, head)
 773                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 774                         interface_set_prefix_address(c, prefix, iface, true);
 775
 776         if (!assigned_any)
 777                 netifd_log_message(L_WARNING, "You have delegated IPv6-prefixes but haven't assigned them "
 778                                 "to any interface. Did you forget to set option ip6assign on your lan-interfaces?");
 779 }
 780
 781
 782 void interface_refresh_assignments(bool hint)
 783 {
 784         static bool refresh = false;
 785         if (!hint && refresh) {
 786                 struct device_prefix *p;
 787                 list_for_each_entry(p, &prefixes, head)
 788                         interface_update_prefix_assignments(p, true);
 789         }
 790         refresh = hint;
 791 }
 792
 793
 794 static void
 795 interface_update_prefix(struct vlist_tree *tree,
 796                              struct vlist_node *node_new,
 797                              struct vlist_node *node_old)
 798 {
 799         struct device_prefix *prefix_old, *prefix_new;
 800         prefix_old = container_of(node_old, struct device_prefix, node);
 801         prefix_new = container_of(node_new, struct device_prefix, node);
 802
 803         struct device_route route;
 804         memset(&route, 0, sizeof(route));
 805         route.flags = DEVADDR_INET6;
 806         route.metric = INT32_MAX;
 807         route.mask = (node_new) ? prefix_new->length : prefix_old->length;
 808         route.addr.in6 = (node_new) ? prefix_new->addr : prefix_old->addr;
 809
 810
 811         struct device_prefix_assignment *c;
 812         struct interface *iface;
 813
 814         if (node_old && node_new) {
 815                 // Move assignments and refresh addresses to update valid times
 816                 list_splice(&prefix_old->assignments, &prefix_new->assignments);
 817
 818                 list_for_each_entry(c, &prefix_new->assignments, head)
 819                         if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 820                                 interface_set_prefix_address(c, prefix_new, iface, true);
 821         } else if (node_new) {
 822                 // Set null-route to avoid routing loops
 823                 system_add_route(NULL, &route);
 824                 interface_update_prefix_assignments(prefix_new, true);
 825         } else if (node_old) {
 826                 // Remove null-route
 827                 interface_update_prefix_assignments(prefix_old, false);
 828                 system_del_route(NULL, &route);
 829         }
 830
 831         if (node_old) {
 832                 list_del(&prefix_old->head);
 833                 free(prefix_old);
 834         }
 835
 836         if (node_new)
 837                 list_add(&prefix_new->head, &prefixes);
 838
 839 }
 840
 841 struct device_prefix*
 842 interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
 843                 uint8_t length, time_t valid_until, time_t preferred_until,
 844                 struct in6_addr *excl_addr, uint8_t excl_length, const char *pclass)
 845 {
 846         if (!pclass)
 847                 pclass = (iface) ? iface->name : "local";
 848
 849         struct device_prefix *prefix = calloc(1, sizeof(*prefix) + strlen(pclass) + 1);
 850         prefix->length = length;
 851         prefix->addr = *addr;
 852         prefix->preferred_until = preferred_until;
 853         prefix->valid_until = valid_until;
 854         prefix->iface = iface;
 855         INIT_LIST_HEAD(&prefix->assignments);
 856
 857         if (excl_addr) {
 858                 prefix->excl_addr = *excl_addr;
 859                 prefix->excl_length = excl_length;
 860         }
 861
 862         strcpy(prefix->pclass, pclass);
 863
 864         if (iface)
 865                 vlist_add(&iface->proto_ip.prefix, &prefix->node, &prefix->addr);
 866         else
 867                 interface_update_prefix(NULL, &prefix->node, NULL);
 868
 869         return prefix;
 870 }
 871
 872 void
 873 interface_ip_set_ula_prefix(const char *prefix)
 874 {
 875         char buf[INET6_ADDRSTRLEN + 4] = {0}, *saveptr;
 876         if (prefix)
 877                 strncpy(buf, prefix, sizeof(buf) - 1);
 878         char *prefixaddr = strtok_r(buf, "/", &saveptr);
 879
 880         struct in6_addr addr;
 881         if (!prefixaddr || inet_pton(AF_INET6, prefixaddr, &addr) < 1) {
 882                 if (ula_prefix) {
 883                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
 884                         ula_prefix = NULL;
 885                 }
 886                 return;
 887         }
 888
 889         int length;
 890         char *prefixlen = strtok_r(NULL, ",", &saveptr);
 891         if (!prefixlen || (length = atoi(prefixlen)) < 1 || length > 64)
 892                 return;
 893
 894         if (!ula_prefix || !IN6_ARE_ADDR_EQUAL(&addr, &ula_prefix->addr) ||
 895                         ula_prefix->length != length) {
 896                 if (ula_prefix)
 897                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
 898
 899                 ula_prefix = interface_ip_add_device_prefix(NULL, &addr, length,
 900                                 0, 0, NULL, 0, NULL);
 901         }
 902 }
 903
 904 void
 905 interface_add_dns_server(struct interface_ip_settings *ip, const char *str)
 906 {
 907         struct dns_server *s;
 908
 909         s = calloc(1, sizeof(*s));
 910         if (!s)
 911                 return;
 912
 913         s->af = AF_INET;
 914         if (inet_pton(s->af, str, &s->addr.in))
 915                 goto add;
 916
 917         s->af = AF_INET6;
 918         if (inet_pton(s->af, str, &s->addr.in))
 919                 goto add;
 920
 921         free(s);
 922         return;
 923
 924 add:
 925         D(INTERFACE, "Add IPv%c DNS server: %s\n",
 926           s->af == AF_INET6 ? '6' : '4', str);
 927         vlist_simple_add(&ip->dns_servers, &s->node);
 928 }
 929
 930 void
 931 interface_add_dns_server_list(struct interface_ip_settings *ip, struct blob_attr *list)
 932 {
 933         struct blob_attr *cur;
 934         int rem;
 935
 936         blobmsg_for_each_attr(cur, list, rem) {
 937                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
 938                         continue;
 939
 940                 if (!blobmsg_check_attr(cur, NULL))
 941                         continue;
 942
 943                 interface_add_dns_server(ip, blobmsg_data(cur));
 944         }
 945 }
 946
 947 static void
 948 interface_add_dns_search_domain(struct interface_ip_settings *ip, const char *str)
 949 {
 950         struct dns_search_domain *s;
 951         int len = strlen(str);
 952
 953         s = calloc(1, sizeof(*s) + len + 1);
 954         if (!s)
 955                 return;
 956
 957         D(INTERFACE, "Add DNS search domain: %s\n", str);
 958         memcpy(s->name, str, len);
 959         vlist_simple_add(&ip->dns_search, &s->node);
 960 }
 961
 962 void
 963 interface_add_dns_search_list(struct interface_ip_settings *ip, struct blob_attr *list)
 964 {
 965         struct blob_attr *cur;
 966         int rem;
 967
 968         blobmsg_for_each_attr(cur, list, rem) {
 969                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
 970                         continue;
 971
 972                 if (!blobmsg_check_attr(cur, NULL))
 973                         continue;
 974
 975                 interface_add_dns_search_domain(ip, blobmsg_data(cur));
 976         }
 977 }
 978
 979 static void
 980 write_resolv_conf_entries(FILE *f, struct interface_ip_settings *ip)
 981 {
 982         struct dns_server *s;
 983         struct dns_search_domain *d;
 984         const char *str;
 985         char buf[INET6_ADDRSTRLEN];
 986
 987         vlist_simple_for_each_element(&ip->dns_servers, s, node) {
 988                 str = inet_ntop(s->af, &s->addr, buf, sizeof(buf));
 989                 if (!str)
 990                         continue;
 991
 992                 fprintf(f, "nameserver %s\n", str);
 993         }
 994
 995         vlist_simple_for_each_element(&ip->dns_search, d, node) {
 996                 fprintf(f, "search %s\n", d->name);
 997         }
 998 }
 999
1000 void
1001 interface_write_resolv_conf(void)
1002 {
1003         struct interface *iface;
1004         char *path = alloca(strlen(resolv_conf) + 5);
1005         FILE *f;
1006         uint32_t crcold, crcnew;
1007
1008         sprintf(path, "%s.tmp", resolv_conf);
1009         unlink(path);
1010         f = fopen(path, "w+");
1011         if (!f) {
1012                 D(INTERFACE, "Failed to open %s for writing\n", path);
1013                 return;
1014         }
1015
1016         vlist_for_each_element(&interfaces, iface, node) {
1017                 if (iface->state != IFS_UP)
1018                         continue;
1019
1020                 if (vlist_simple_empty(&iface->proto_ip.dns_search) &&
1021                     vlist_simple_empty(&iface->proto_ip.dns_servers) &&
1022                         vlist_simple_empty(&iface->config_ip.dns_search) &&
1023                     vlist_simple_empty(&iface->config_ip.dns_servers))
1024                         continue;
1025
1026                 fprintf(f, "# Interface %s\n", iface->name);
1027                 write_resolv_conf_entries(f, &iface->config_ip);
1028                 if (!iface->proto_ip.no_dns)
1029                         write_resolv_conf_entries(f, &iface->proto_ip);
1030         }
1031         fflush(f);
1032         rewind(f);
1033         crcnew = crc32_file(f);
1034         fclose(f);
1035
1036         crcold = crcnew + 1;
1037         f = fopen(resolv_conf, "r");
1038         if (f) {
1039                 crcold = crc32_file(f);
1040                 fclose(f);
1041         }
1042
1043         if (crcold == crcnew) {
1044                 unlink(path);
1045         } else if (rename(path, resolv_conf) < 0) {
1046                 D(INTERFACE, "Failed to replace %s\n", resolv_conf);
1047                 unlink(path);
1048         }
1049 }
1050
1051 void interface_ip_set_enabled(struct interface_ip_settings *ip, bool enabled)
1052 {
1053         struct device_addr *addr;
1054         struct device_route *route;
1055         struct device *dev;
1056
1057         ip->enabled = enabled;
1058         dev = ip->iface->l3_dev.dev;
1059         if (!dev)
1060                 return;
1061
1062         vlist_for_each_element(&ip->addr, addr, node) {
1063                 if (addr->enabled == enabled)
1064                         continue;
1065
1066                 if (enabled)
1067                         system_add_address(dev, addr);
1068                 else
1069                         system_del_address(dev, addr);
1070                 addr->enabled = enabled;
1071         }
1072
1073         vlist_for_each_element(&ip->route, route, node) {
1074                 bool _enabled = enabled;
1075
1076                 if (!enable_route(ip, route))
1077                         _enabled = false;
1078
1079                 if (route->enabled == _enabled)
1080                         continue;
1081
1082                 if (_enabled) {
1083                         if (!(route->flags & DEVROUTE_METRIC))
1084                                 route->metric = ip->iface->metric;
1085
1086                         system_add_route(dev, route);
1087                 } else
1088                         system_del_route(dev, route);
1089                 route->enabled = _enabled;
1090         }
1091
1092         struct device_prefix *c;
1093         struct device_prefix_assignment *a;
1094         list_for_each_entry(c, &prefixes, head)
1095                 list_for_each_entry(a, &c->assignments, head)
1096                         if (!strcmp(a->name, ip->iface->name))
1097                                 interface_set_prefix_address(a, c, ip->iface, enabled);
1098
1099         if (ip->iface && ip->iface->l3_dev.dev) {
1100                 set_ip_lo_policy(enabled, true, ip->iface);
1101                 set_ip_lo_policy(enabled, false, ip->iface);
1102
1103                 set_ip_source_policy(enabled, true, IPRULE_PRIORITY_REJECT + ip->iface->l3_dev.dev->ifindex,
1104                         NULL, 0, NULL, ip->iface, "failed_policy");
1105         }
1106 }
1107
1108 void
1109 interface_ip_update_start(struct interface_ip_settings *ip)
1110 {
1111         if (ip != &ip->iface->config_ip) {
1112                 vlist_simple_update(&ip->dns_servers);
1113                 vlist_simple_update(&ip->dns_search);
1114         }
1115         vlist_update(&ip->route);
1116         vlist_update(&ip->addr);
1117         vlist_update(&ip->prefix);
1118 }
1119
1120 void
1121 interface_ip_update_complete(struct interface_ip_settings *ip)
1122 {
1123         vlist_simple_flush(&ip->dns_servers);
1124         vlist_simple_flush(&ip->dns_search);
1125         vlist_flush(&ip->route);
1126         vlist_flush(&ip->addr);
1127         vlist_flush(&ip->prefix);
1128         interface_write_resolv_conf();
1129 }
1130
1131 void
1132 interface_ip_flush(struct interface_ip_settings *ip)
1133 {
1134         if (ip == &ip->iface->proto_ip)
1135                 vlist_flush_all(&ip->iface->host_routes);
1136         vlist_simple_flush_all(&ip->dns_servers);
1137         vlist_simple_flush_all(&ip->dns_search);
1138         vlist_flush_all(&ip->route);
1139         vlist_flush_all(&ip->addr);
1140         vlist_flush_all(&ip->prefix);
1141 }
1142
1143 static void
1144 __interface_ip_init(struct interface_ip_settings *ip, struct interface *iface)
1145 {
1146         ip->iface = iface;
1147         ip->enabled = true;
1148         vlist_simple_init(&ip->dns_search, struct dns_search_domain, node);
1149         vlist_simple_init(&ip->dns_servers, struct dns_server, node);
1150         vlist_init(&ip->route, route_cmp, interface_update_proto_route);
1151         vlist_init(&ip->addr, addr_cmp, interface_update_proto_addr);
1152         vlist_init(&ip->prefix, prefix_cmp, interface_update_prefix);
1153 }
1154
1155 void
1156 interface_ip_init(struct interface *iface)
1157 {
1158         __interface_ip_init(&iface->proto_ip, iface);
1159         __interface_ip_init(&iface->config_ip, iface);
1160         vlist_init(&iface->host_routes, route_cmp, interface_update_host_route);
1161
1162 }
1163
1164 static void
1165 interface_ip_valid_until_handler(struct uloop_timeout *t)
1166 {
1167         time_t now = system_get_rtime();
1168         struct interface *iface;
1169         vlist_for_each_element(&interfaces, iface, node) {
1170                 if (iface->state != IFS_UP)
1171                         continue;
1172
1173                 struct device_addr *addr, *addrp;
1174                 struct device_route *route, *routep;
1175                 struct device_prefix *pref, *prefp;
1176
1177                 vlist_for_each_element_safe(&iface->proto_ip.addr, addr, node, addrp)
1178                         if (addr->valid_until && addr->valid_until < now)
1179                                 vlist_delete(&iface->proto_ip.addr, &addr->node);
1180
1181                 vlist_for_each_element_safe(&iface->proto_ip.route, route, node, routep)
1182                         if (route->valid_until && route->valid_until < now)
1183                                 vlist_delete(&iface->proto_ip.route, &route->node);
1184
1185                 vlist_for_each_element_safe(&iface->proto_ip.prefix, pref, node, prefp)
1186                         if (pref->valid_until && pref->valid_until < now)
1187                                 vlist_delete(&iface->proto_ip.prefix, &pref->node);
1188
1189         }
1190
1191         uloop_timeout_set(t, 1000);
1192 }
1193
1194 static void __init
1195 interface_ip_init_worker(void)
1196 {
1197         valid_until_timeout.cb = interface_ip_valid_until_handler;
1198         uloop_timeout_set(&valid_until_timeout, 1000);
1199 }