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_MAX
  42 };
  43
  44 static const struct blobmsg_policy route_attr[__ROUTE_MAX] = {
  45         [ROUTE_INTERFACE] = { .name = "interface", .type = BLOBMSG_TYPE_STRING },
  46         [ROUTE_TARGET] = { .name = "target", .type = BLOBMSG_TYPE_STRING },
  47         [ROUTE_MASK] = { .name = "netmask", .type = BLOBMSG_TYPE_STRING },
  48         [ROUTE_GATEWAY] = { .name = "gateway", .type = BLOBMSG_TYPE_STRING },
  49         [ROUTE_METRIC] = { .name = "metric", .type = BLOBMSG_TYPE_INT32 },
  50         [ROUTE_MTU] = { .name = "mtu", .type = BLOBMSG_TYPE_INT32 },
  51         [ROUTE_TABLE] = { .name = "table", .type = BLOBMSG_TYPE_STRING },
  52         [ROUTE_VALID] = { .name = "valid", .type = BLOBMSG_TYPE_INT32 },
  53         [ROUTE_SOURCE] = { .name = "source", .type = BLOBMSG_TYPE_STRING },
  54 };
  55
  56 const struct uci_blob_param_list route_attr_list = {
  57         .n_params = __ROUTE_MAX,
  58         .params = route_attr,
  59 };
  60
  61
  62 struct list_head prefixes = LIST_HEAD_INIT(prefixes);
  63 static struct device_prefix *ula_prefix = NULL;
  64 static struct uloop_timeout valid_until_timeout;
  65
  66
  67 static void
  68 clear_if_addr(union if_addr *a, int mask)
  69 {
  70         int m_bytes = (mask + 7) / 8;
  71         uint8_t m_clear = (1 << (m_bytes * 8 - mask)) - 1;
  72         uint8_t *p = (uint8_t *) a;
  73
  74         if (m_bytes < sizeof(a))
  75                 memset(p + m_bytes, 0, sizeof(a) - m_bytes);
  76
  77         p[m_bytes - 1] &= ~m_clear;
  78 }
  79
  80 static bool
  81 match_if_addr(union if_addr *a1, union if_addr *a2, int mask)
  82 {
  83         union if_addr *p1, *p2;
  84
  85         p1 = alloca(sizeof(*a1));
  86         p2 = alloca(sizeof(*a2));
  87
  88         memcpy(p1, a1, sizeof(*a1));
  89         clear_if_addr(p1, mask);
  90         memcpy(p2, a2, sizeof(*a2));
  91         clear_if_addr(p2, mask);
  92
  93         return !memcmp(p1, p2, sizeof(*p1));
  94 }
  95
  96 static int set_ip_source_policy(bool add, bool v6, unsigned int priority,
  97                 const union if_addr *addr, uint8_t mask, unsigned int table,
  98                 struct interface *in_iface, const char *action)
  99 {
 100         struct iprule rule = {
 101                 .flags = IPRULE_PRIORITY,
 102                 .priority = priority
 103         };
 104
 105         if (addr) {
 106                 rule.flags |= IPRULE_SRC;
 107                 rule.src_addr = *addr;
 108                 rule.src_mask = mask;
 109         }
 110
 111         if (table) {
 112                 rule.flags |= IPRULE_LOOKUP;
 113                 rule.lookup = table;
 114
 115                 if (!rule.lookup)
 116                         return 0;
 117         } else if (action) {
 118                 rule.flags |= IPRULE_ACTION;
 119                 system_resolve_iprule_action(action, &rule.action);
 120         }
 121
 122         if (in_iface && in_iface->l3_dev.dev) {
 123                 rule.flags |= IPRULE_IN;
 124                 strcpy(rule.in_dev, in_iface->l3_dev.dev->ifname);
 125         }
 126
 127         rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
 128
 129         return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
 130 }
 131
 132 static int set_ip_lo_policy(bool add, bool v6, struct interface *iface)
 133 {
 134         struct iprule rule = {
 135                 .flags = IPRULE_IN | IPRULE_LOOKUP | IPRULE_PRIORITY,
 136                 .priority = IPRULE_PRIORITY_NW + iface->l3_dev.dev->ifindex,
 137                 .lookup = (v6) ? iface->ip6table : iface->ip4table,
 138                 .in_dev = "lo"
 139         };
 140
 141         if (!rule.lookup)
 142                 return 0;
 143
 144         rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
 145
 146         return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
 147 }
 148
 149 static bool
 150 __find_ip_addr_target(struct interface_ip_settings *ip, union if_addr *a, bool v6)
 151 {
 152         struct device_addr *addr;
 153
 154         vlist_for_each_element(&ip->addr, addr, node) {
 155                 if (!addr->enabled)
 156                         continue;
 157
 158                 if (v6 != ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
 159                         continue;
 160
 161                 // Handle offlink addresses correctly
 162                 unsigned int mask = addr->mask;
 163                 if ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6 &&
 164                                 (addr->flags & DEVADDR_OFFLINK))
 165                         mask = 128;
 166
 167                 if (!match_if_addr(&addr->addr, a, mask))
 168                         continue;
 169
 170                 return true;
 171         }
 172
 173         return false;
 174 }
 175
 176 static void
 177 __find_ip_route_target(struct interface_ip_settings *ip, union if_addr *a,
 178                        bool v6, struct device_route **res)
 179 {
 180         struct device_route *route;
 181
 182         vlist_for_each_element(&ip->route, route, node) {
 183                 if (!route->enabled)
 184                         continue;
 185
 186                 if (v6 != ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
 187                         continue;
 188
 189                 if (!match_if_addr(&route->addr, a, route->mask))
 190                         continue;
 191
 192                 if (route->flags & DEVROUTE_TABLE)
 193                         continue;
 194
 195                 if (!*res || route->mask < (*res)->mask)
 196                         *res = route;
 197         }
 198 }
 199
 200 static bool
 201 interface_ip_find_addr_target(struct interface *iface, union if_addr *a, bool v6)
 202 {
 203         return __find_ip_addr_target(&iface->proto_ip, a, v6) ||
 204                __find_ip_addr_target(&iface->config_ip, a, v6);
 205 }
 206
 207 static void
 208 interface_ip_find_route_target(struct interface *iface, union if_addr *a,
 209                                bool v6, struct device_route **route)
 210 {
 211         __find_ip_route_target(&iface->proto_ip, a, v6, route);
 212         __find_ip_route_target(&iface->config_ip, a, v6, route);
 213 }
 214
 215 struct interface *
 216 interface_ip_add_target_route(union if_addr *addr, bool v6, struct interface *iface)
 217 {
 218         struct device_route *route, *r_next = NULL;
 219         bool defaultroute_target = false;
 220         int addrsize = v6 ? sizeof(addr->in6) : sizeof(addr->in);
 221
 222         route = calloc(1, sizeof(*route));
 223         if (!route)
 224                 return NULL;
 225
 226         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
 227         route->mask = v6 ? 128 : 32;
 228         if (memcmp(&route->addr, addr, addrsize) == 0)
 229                 defaultroute_target = true;
 230         else
 231                 memcpy(&route->addr, addr, addrsize);
 232
 233         if (iface) {
 234                 /* look for locally addressable target first */
 235                 if (interface_ip_find_addr_target(iface, addr, v6))
 236                         goto done;
 237
 238                 /* do not stop at the first route, let the lookup compare
 239                  * masks to find the best match */
 240                 interface_ip_find_route_target(iface, addr, v6, &r_next);
 241         } else {
 242                 vlist_for_each_element(&interfaces, iface, node) {
 243                         /* look for locally addressable target first */
 244                         if (interface_ip_find_addr_target(iface, addr, v6))
 245                                 goto done;
 246
 247                         /* do not stop at the first route, let the lookup compare
 248                          * masks to find the best match */
 249                         interface_ip_find_route_target(iface, addr, v6, &r_next);
 250                 }
 251         }
 252
 253         if (!r_next) {
 254                 free(route);
 255                 return NULL;
 256         }
 257
 258         iface = r_next->iface;
 259         memcpy(&route->nexthop, &r_next->nexthop, sizeof(route->nexthop));
 260         route->mtu = r_next->mtu;
 261         route->metric = r_next->metric;
 262         route->table = r_next->table;
 263
 264 done:
 265         route->iface = iface;
 266         if (defaultroute_target)
 267                 free(route);
 268         else
 269                 vlist_add(&iface->host_routes, &route->node, route);
 270         return iface;
 271 }
 272
 273 void
 274 interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
 275 {
 276         struct interface_ip_settings *ip;
 277         struct blob_attr *tb[__ROUTE_MAX], *cur;
 278         struct device_route *route;
 279         int af = v6 ? AF_INET6 : AF_INET;
 280         bool is_proto_route = !!iface;
 281
 282         blobmsg_parse(route_attr, __ROUTE_MAX, tb, blobmsg_data(attr), blobmsg_data_len(attr));
 283
 284         if (!iface) {
 285                 if ((cur = tb[ROUTE_INTERFACE]) == NULL)
 286                         return;
 287
 288                 iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
 289                 if (!iface)
 290                         return;
 291
 292                 ip = &iface->config_ip;
 293         } else {
 294                 ip = &iface->proto_ip;
 295         }
 296
 297         route = calloc(1, sizeof(*route));
 298         if (!route)
 299                 return;
 300
 301         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
 302         route->mask = v6 ? 128 : 32;
 303         if ((cur = tb[ROUTE_MASK]) != NULL) {
 304                 route->mask = parse_netmask_string(blobmsg_data(cur), v6);
 305                 if (route->mask > (v6 ? 128 : 32))
 306                         goto error;
 307         }
 308
 309         if ((cur = tb[ROUTE_TARGET]) != NULL) {
 310                 if (!parse_ip_and_netmask(af, blobmsg_data(cur), &route->addr, &route->mask)) {
 311                         DPRINTF("Failed to parse route target: %s\n", (char *) blobmsg_data(cur));
 312                         goto error;
 313                 }
 314         }
 315
 316         if ((cur = tb[ROUTE_GATEWAY]) != NULL) {
 317                 if (!inet_pton(af, blobmsg_data(cur), &route->nexthop)) {
 318                         DPRINTF("Failed to parse route gateway: %s\n", (char *) blobmsg_data(cur));
 319                         goto error;
 320                 }
 321         }
 322
 323         if ((cur = tb[ROUTE_METRIC]) != NULL) {
 324                 route->metric = blobmsg_get_u32(cur);
 325                 route->flags |= DEVROUTE_METRIC;
 326         } else
 327                 route->metric = iface->metric;
 328
 329         if ((cur = tb[ROUTE_MTU]) != NULL) {
 330                 route->mtu = blobmsg_get_u32(cur);
 331                 route->flags |= DEVROUTE_MTU;
 332         }
 333
 334         // Use source-based routing
 335         if ((cur = tb[ROUTE_SOURCE]) != NULL) {
 336                 char *saveptr, *source = alloca(blobmsg_data_len(cur));
 337                 memcpy(source, blobmsg_data(cur), blobmsg_data_len(cur));
 338
 339                 const char *addr = strtok_r(source, "/", &saveptr);
 340                 const char *mask = strtok_r(NULL, "/", &saveptr);
 341
 342                 if (!addr || inet_pton(af, addr, &route->source) < 1) {
 343                         DPRINTF("Failed to parse route source: %s\n", addr);
 344                         goto error;
 345                 }
 346
 347                 route->sourcemask = (mask) ? atoi(mask) : ((af == AF_INET6) ? 128 : 32);
 348         }
 349
 350         if (is_proto_route) {
 351                 route->table = (v6) ? iface->ip6table : iface->ip4table;
 352                 route->flags |= DEVROUTE_SRCTABLE;
 353         }
 354
 355         if ((cur = tb[ROUTE_TABLE]) != NULL) {
 356                 if (!system_resolve_rt_table(blobmsg_data(cur), &route->table)) {
 357                         DPRINTF("Failed to resolve routing table: %s\n", (char *) blobmsg_data(cur));
 358                         goto error;
 359                 }
 360
 361                 if (route->table)
 362                         route->flags |= DEVROUTE_TABLE;
 363         }
 364
 365         if ((cur = tb[ROUTE_VALID]) != NULL) {
 366                 int64_t valid = blobmsg_get_u32(cur);
 367                 int64_t valid_until = valid + (int64_t)system_get_rtime();
 368                 if (valid_until <= LONG_MAX && valid != 0xffffffffLL) // Catch overflow
 369                         route->valid_until = valid_until;
 370         }
 371
 372         vlist_add(&ip->route, &route->node, route);
 373         return;
 374
 375 error:
 376         free(route);
 377 }
 378
 379 static int
 380 addr_cmp(const void *k1, const void *k2, void *ptr)
 381 {
 382         return memcmp(k1, k2, sizeof(struct device_addr) -
 383                       offsetof(struct device_addr, flags));
 384 }
 385
 386 static int
 387 route_cmp(const void *k1, const void *k2, void *ptr)
 388 {
 389         const struct device_route *r1 = k1, *r2 = k2;
 390
 391         if (r1->mask != r2->mask)
 392                 return r2->mask - r1->mask;
 393
 394         if (r1->metric != r2->metric)
 395                 return r1->metric - r2->metric;
 396
 397         if (r1->flags != r2->flags)
 398                 return r2->flags - r1->flags;
 399
 400         if (r1->sourcemask != r2->sourcemask)
 401                 return r1->sourcemask - r2->sourcemask;
 402
 403         int maskcmp = memcmp(&r1->source, &r2->source, sizeof(r1->source));
 404         if (maskcmp)
 405                 return maskcmp;
 406
 407         return memcmp(&r1->addr, &r2->addr, sizeof(r1->addr));
 408 }
 409
 410 static int
 411 prefix_cmp(const void *k1, const void *k2, void *ptr)
 412 {
 413         return memcmp(k1, k2, offsetof(struct device_prefix, pclass) -
 414                         offsetof(struct device_prefix, addr));
 415 }
 416
 417 static void
 418 interface_handle_subnet_route(struct interface *iface, struct device_addr *addr, bool add)
 419 {
 420         struct device *dev = iface->l3_dev.dev;
 421         struct device_route route;
 422
 423         memset(&route, 0, sizeof(route));
 424         route.iface = iface;
 425         route.flags = addr->flags;
 426         route.mask = addr->mask;
 427         memcpy(&route.addr, &addr->addr, sizeof(route.addr));
 428         clear_if_addr(&route.addr, route.mask);
 429
 430         if (add) {
 431                 route.flags |= DEVADDR_KERNEL;
 432                 system_del_route(dev, &route);
 433
 434                 if (!(addr->flags & DEVADDR_OFFLINK)) {
 435                         route.flags &= ~DEVADDR_KERNEL;
 436                         route.metric = iface->metric;
 437                         system_add_route(dev, &route);
 438                 }
 439         } else {
 440                 if (!(addr->flags & DEVADDR_OFFLINK))
 441                         system_del_route(dev, &route);
 442         }
 443 }
 444
 445 static void
 446 interface_update_proto_addr(struct vlist_tree *tree,
 447                             struct vlist_node *node_new,
 448                             struct vlist_node *node_old)
 449 {
 450         struct interface_ip_settings *ip;
 451         struct interface *iface;
 452         struct device *dev;
 453         struct device_addr *a_new = NULL, *a_old = NULL;
 454         bool replace = false;
 455         bool keep = false;
 456         bool v6 = false;
 457
 458         ip = container_of(tree, struct interface_ip_settings, addr);
 459         iface = ip->iface;
 460         dev = iface->l3_dev.dev;
 461
 462         if (!node_new || !node_old)
 463                 iface->updated |= IUF_ADDRESS;
 464
 465         if (node_new) {
 466                 a_new = container_of(node_new, struct device_addr, node);
 467
 468                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
 469                     !a_new->broadcast) {
 470
 471                         uint32_t mask = ~0;
 472                         uint32_t *a = (uint32_t *) &a_new->addr;
 473
 474                         mask >>= a_new->mask;
 475                         a_new->broadcast = *a | htonl(mask);
 476                 }
 477         }
 478
 479         if (node_old)
 480                 a_old = container_of(node_old, struct device_addr, node);
 481
 482         if (a_new && a_old) {
 483                 keep = true;
 484
 485                 if (a_old->flags != a_new->flags || a_old->failed)
 486                         keep = false;
 487
 488                 if (a_old->valid_until != a_new->valid_until ||
 489                                 a_old->preferred_until != a_new->preferred_until)
 490                         replace = true;
 491
 492                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
 493                     a_new->broadcast != a_old->broadcast)
 494                         keep = false;
 495         }
 496
 497         if (node_old) {
 498                 if (!(a_old->flags & DEVADDR_EXTERNAL) && a_old->enabled && !keep) {
 499                         interface_handle_subnet_route(iface, a_old, false);
 500
 501                         if ((a_old->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
 502                                 v6 = true;
 503
 504                         unsigned int table = (v6) ? iface->ip6table : iface->ip4table;
 505
 506                         //This is needed for source routing to work correctly. If a device
 507                         //has two connections to a network using the same subnet, adding
 508                         //only the network-rule will cause packets to be routed through the
 509                         //first matching network (source IP matches both masks).
 510                         if (table) {
 511                                 set_ip_source_policy(false, v6, IPRULE_PRIORITY_ADDR, &a_old->addr,
 512                                                 (v6) ? 128 : 32, table, NULL, NULL);
 513                                 set_ip_source_policy(false, v6, IPRULE_PRIORITY_NW, &a_old->addr,
 514                                                 a_old->mask, table, NULL, NULL);
 515                         }
 516
 517                         system_del_address(dev, a_old);
 518                 }
 519                 free(a_old->pclass);
 520                 free(a_old);
 521         }
 522
 523         if (node_new) {
 524                 a_new->enabled = true;
 525                 if (!(a_new->flags & DEVADDR_EXTERNAL) && (!keep || replace)) {
 526                         if (system_add_address(dev, a_new))
 527                                 a_new->failed = true;
 528
 529                         if (!keep) {
 530                                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
 531                                         v6 = true;
 532
 533                                 unsigned int table = (v6) ? iface->ip6table : iface->ip4table;
 534
 535                                 if (table) {
 536                                         set_ip_source_policy(true, v6, IPRULE_PRIORITY_ADDR, &a_new->addr,
 537                                                         (v6) ? 128 : 32, table, NULL, NULL);
 538                                         set_ip_source_policy(true, v6, IPRULE_PRIORITY_NW, &a_new->addr,
 539                                                         a_new->mask, table, NULL, NULL);
 540                                 }
 541                         }
 542
 543                         if ((a_new->flags & DEVADDR_OFFLINK) || iface->metric)
 544                                 interface_handle_subnet_route(iface, a_new, true);
 545                 }
 546         }
 547 }
 548
 549 static bool
 550 enable_route(struct interface_ip_settings *ip, struct device_route *route)
 551 {
 552         if (ip->no_defaultroute && !route->mask)
 553                 return false;
 554
 555         return ip->enabled;
 556 }
 557
 558 static void
 559 interface_update_proto_route(struct vlist_tree *tree,
 560                              struct vlist_node *node_new,
 561                              struct vlist_node *node_old)
 562 {
 563         struct interface_ip_settings *ip;
 564         struct interface *iface;
 565         struct device *dev;
 566         struct device_route *route_old, *route_new;
 567         bool keep = false;
 568
 569         ip = container_of(tree, struct interface_ip_settings, route);
 570         iface = ip->iface;
 571         dev = iface->l3_dev.dev;
 572
 573         if (!node_new || !node_old)
 574                 iface->updated |= IUF_ROUTE;
 575
 576         route_old = container_of(node_old, struct device_route, node);
 577         route_new = container_of(node_new, struct device_route, node);
 578
 579         if (node_old && node_new)
 580                 keep = !memcmp(&route_old->nexthop, &route_new->nexthop, sizeof(route_old->nexthop)) &&
 581                         (route_old->table == route_new->table) && !route_old->failed;
 582
 583         if (node_old) {
 584                 if (!(route_old->flags & DEVADDR_EXTERNAL) && route_old->enabled && !keep)
 585                         system_del_route(dev, route_old);
 586
 587                 free(route_old);
 588         }
 589
 590         if (node_new) {
 591                 bool _enabled = enable_route(ip, route_new);
 592
 593                 if (!(route_new->flags & DEVADDR_EXTERNAL) && !keep && _enabled)
 594                         if (system_add_route(dev, route_new))
 595                                 route_new->failed = true;
 596
 597                 route_new->iface = iface;
 598                 route_new->enabled = _enabled;
 599         }
 600 }
 601
 602 static void
 603 interface_update_host_route(struct vlist_tree *tree,
 604                              struct vlist_node *node_new,
 605                              struct vlist_node *node_old)
 606 {
 607         struct interface *iface;
 608         struct device *dev;
 609         struct device_route *route_old, *route_new;
 610
 611         iface = container_of(tree, struct interface, host_routes);
 612         dev = iface->l3_dev.dev;
 613
 614         route_old = container_of(node_old, struct device_route, node);
 615         route_new = container_of(node_new, struct device_route, node);
 616
 617         if (node_old) {
 618                 system_del_route(dev, route_old);
 619                 free(route_old);
 620         }
 621
 622         if (node_new)
 623                 system_add_route(dev, route_new);
 624 }
 625
 626
 627 static void
 628 interface_set_prefix_address(struct device_prefix_assignment *assignment,
 629                 const struct device_prefix *prefix, struct interface *iface, bool add)
 630 {
 631         const struct interface *uplink = prefix->iface;
 632         if (!iface->l3_dev.dev)
 633                 return;
 634
 635         struct device *l3_downlink = iface->l3_dev.dev;
 636
 637         struct device_addr addr;
 638         memset(&addr, 0, sizeof(addr));
 639         addr.addr.in6 = prefix->addr;
 640         addr.addr.in6.s6_addr32[1] |= htonl(assignment->assigned);
 641         addr.addr.in6.s6_addr[15] += 1;
 642         addr.mask = assignment->length;
 643         addr.flags = DEVADDR_INET6;
 644         addr.preferred_until = prefix->preferred_until;
 645         addr.valid_until = prefix->valid_until;
 646
 647         if (!add && assignment->enabled) {
 648                 time_t now = system_get_rtime();
 649                 addr.preferred_until = now;
 650                 if (!addr.valid_until || addr.valid_until - now > 7200)
 651                         addr.valid_until = now + 7200;
 652                 system_del_address(l3_downlink, &addr); // Work around dangling prefix routes
 653                 system_add_address(l3_downlink, &addr);
 654                 if (prefix->iface) {
 655                         if (prefix->iface->ip6table)
 656                                 set_ip_source_policy(false, true, IPRULE_PRIORITY_NW, &addr.addr,
 657                                                 addr.mask, prefix->iface->ip6table, iface, NULL);
 658
 659                         set_ip_source_policy(false, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 660                                                         addr.mask, 0, iface, "unreachable");
 661                 }
 662
 663                 assignment->enabled = false;
 664         } else if (add && (iface->state == IFS_UP || iface->state == IFS_SETUP)) {
 665                 system_add_address(l3_downlink, &addr);
 666                 if (prefix->iface && !assignment->enabled) {
 667                         set_ip_source_policy(true, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 668                                         addr.mask, 0, iface, "unreachable");
 669
 670                         if (prefix->iface->ip6table)
 671                                 set_ip_source_policy(true, true, IPRULE_PRIORITY_NW, &addr.addr,
 672                                                 addr.mask, prefix->iface->ip6table, iface, NULL);
 673                 }
 674                 if (uplink && uplink->l3_dev.dev) {
 675                         int mtu = system_update_ipv6_mtu(
 676                                         uplink->l3_dev.dev, 0);
 677                         if (mtu > 0)
 678                                 system_update_ipv6_mtu(l3_downlink, mtu);
 679                 }
 680                 assignment->enabled = true;
 681         }
 682 }
 683
 684 static bool interface_prefix_assign(struct list_head *list,
 685                 struct device_prefix_assignment *assign)
 686 {
 687         int32_t current = 0, asize = (1 << (64 - assign->length)) - 1;
 688         struct device_prefix_assignment *c;
 689         list_for_each_entry(c, list, head) {
 690                 if (assign->assigned != -1) {
 691                         if (assign->assigned >= current && assign->assigned + asize < c->assigned) {
 692                                 list_add_tail(&assign->head, &c->head);
 693                                 return true;
 694                         }
 695                 } else if (assign->assigned == -1) {
 696                         current = (current + asize) & (~asize);
 697                         if (current + asize < c->assigned) {
 698                                 assign->assigned = current;
 699                                 list_add_tail(&assign->head, &c->head);
 700                                 return true;
 701                         }
 702                 }
 703                 current = (c->assigned + (1 << (64 - c->length)));
 704         }
 705         return false;
 706 }
 707
 708 static void interface_update_prefix_assignments(struct device_prefix *prefix, bool setup)
 709 {
 710         struct device_prefix_assignment *c;
 711         struct interface *iface;
 712
 713         // Delete all assignments
 714         while (!list_empty(&prefix->assignments)) {
 715                 c = list_first_entry(&prefix->assignments,
 716                                 struct device_prefix_assignment, head);
 717                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 718                         interface_set_prefix_address(c, prefix, iface, false);
 719                 list_del(&c->head);
 720                 free(c);
 721         }
 722
 723         if (!setup)
 724                 return;
 725
 726         // End-of-assignment sentinel
 727         c = malloc(sizeof(*c) + 1);
 728         c->assigned = 1 << (64 - prefix->length);
 729         c->length = 64;
 730         c->name[0] = 0;
 731         list_add(&c->head, &prefix->assignments);
 732
 733         // Excluded prefix
 734         if (prefix->excl_length > 0) {
 735                 const char name[] = "!excluded";
 736                 c = malloc(sizeof(*c) + sizeof(name));
 737                 c->assigned = ntohl(prefix->excl_addr.s6_addr32[1]) &
 738                                 ((1 << (64 - prefix->length)) - 1);
 739                 c->length = prefix->excl_length;
 740                 memcpy(c->name, name, sizeof(name));
 741                 list_add(&c->head, &prefix->assignments);
 742         }
 743
 744         bool assigned_any = false;
 745         struct list_head assign_later = LIST_HEAD_INIT(assign_later);
 746         vlist_for_each_element(&interfaces, iface, node) {
 747                 if (iface->assignment_length < 48 ||
 748                                 iface->assignment_length > 64)
 749                         continue;
 750
 751                 // Test whether there is a matching class
 752                 if (!list_empty(&iface->assignment_classes)) {
 753                         bool found = false;
 754
 755                         struct interface_assignment_class *c;
 756                         list_for_each_entry(c, &iface->assignment_classes, head) {
 757                                 if (!strcmp(c->name, prefix->pclass)) {
 758                                         found = true;
 759                                         break;
 760                                 }
 761                         }
 762
 763                         if (!found)
 764                                 continue;
 765                 }
 766
 767                 size_t namelen = strlen(iface->name) + 1;
 768                 c = malloc(sizeof(*c) + namelen);
 769                 c->length = iface->assignment_length;
 770                 c->assigned = iface->assignment_hint;
 771                 c->enabled = false;
 772                 memcpy(c->name, iface->name, namelen);
 773
 774                 // First process all custom assignments, put all others in later-list
 775                 if (c->assigned == -1 || !interface_prefix_assign(&prefix->assignments, c)) {
 776                         if (c->assigned != -1) {
 777                                 c->assigned = -1;
 778                                 netifd_log_message(L_WARNING, "Failed to assign requested subprefix "
 779                                                 "of size %hhu for %s, trying other\n", c->length, c->name);
 780                         }
 781
 782                         struct list_head *next = &assign_later;
 783                         struct device_prefix_assignment *n;
 784                         list_for_each_entry(n, &assign_later, head) {
 785                                 if (n->length < c->length) {
 786                                         next = &n->head;
 787                                         break;
 788                                 }
 789                         }
 790                         list_add_tail(&c->head, next);
 791                 }
 792
 793                 if (c->assigned != -1)
 794                         assigned_any = true;
 795         }
 796
 797         // Then try to assign all other + failed custom assignments
 798         while (!list_empty(&assign_later)) {
 799                 c = list_first_entry(&assign_later, struct device_prefix_assignment, head);
 800                 list_del(&c->head);
 801
 802                 bool assigned = false;
 803                 do {
 804                         assigned = interface_prefix_assign(&prefix->assignments, c);
 805                 } while (!assigned && ++c->length <= 64);
 806
 807                 if (!assigned) {
 808                         netifd_log_message(L_WARNING, "Failed to assign subprefix "
 809                                         "of size %hhu for %s\n", c->length, c->name);
 810                         free(c);
 811                 } else {
 812                         assigned_any = true;
 813                 }
 814         }
 815
 816         list_for_each_entry(c, &prefix->assignments, head)
 817                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 818                         interface_set_prefix_address(c, prefix, iface, true);
 819
 820         if (!assigned_any)
 821                 netifd_log_message(L_WARNING, "You have delegated IPv6-prefixes but haven't assigned them "
 822                                 "to any interface. Did you forget to set option ip6assign on your lan-interfaces?");
 823 }
 824
 825
 826 void interface_refresh_assignments(bool hint)
 827 {
 828         static bool refresh = false;
 829         if (!hint && refresh) {
 830                 struct device_prefix *p;
 831                 list_for_each_entry(p, &prefixes, head)
 832                         interface_update_prefix_assignments(p, true);
 833         }
 834         refresh = hint;
 835 }
 836
 837
 838 static void
 839 interface_update_prefix(struct vlist_tree *tree,
 840                              struct vlist_node *node_new,
 841                              struct vlist_node *node_old)
 842 {
 843         struct device_prefix *prefix_old, *prefix_new;
 844         prefix_old = container_of(node_old, struct device_prefix, node);
 845         prefix_new = container_of(node_new, struct device_prefix, node);
 846
 847         struct interface_ip_settings *ip = container_of(tree, struct interface_ip_settings, prefix);
 848         if (tree && (!node_new || !node_old))
 849                 ip->iface->updated |= IUF_PREFIX;
 850
 851         struct device_route route;
 852         memset(&route, 0, sizeof(route));
 853         route.flags = DEVADDR_INET6;
 854         route.metric = INT32_MAX;
 855         route.mask = (node_new) ? prefix_new->length : prefix_old->length;
 856         route.addr.in6 = (node_new) ? prefix_new->addr : prefix_old->addr;
 857
 858
 859         struct device_prefix_assignment *c;
 860         struct interface *iface;
 861
 862         if (node_old && node_new) {
 863                 // Move assignments and refresh addresses to update valid times
 864                 list_splice(&prefix_old->assignments, &prefix_new->assignments);
 865
 866                 list_for_each_entry(c, &prefix_new->assignments, head)
 867                         if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 868                                 interface_set_prefix_address(c, prefix_new, iface, true);
 869         } else if (node_new) {
 870                 // Set null-route to avoid routing loops
 871                 system_add_route(NULL, &route);
 872
 873                 if (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation)
 874                         interface_update_prefix_assignments(prefix_new, true);
 875         } else if (node_old) {
 876                 // Remove null-route
 877                 interface_update_prefix_assignments(prefix_old, false);
 878                 system_del_route(NULL, &route);
 879         }
 880
 881         if (node_old) {
 882                 if (prefix_old->head.next)
 883                         list_del(&prefix_old->head);
 884                 free(prefix_old);
 885         }
 886
 887         if (node_new && (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation))
 888                 list_add(&prefix_new->head, &prefixes);
 889
 890 }
 891
 892 struct device_prefix*
 893 interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
 894                 uint8_t length, time_t valid_until, time_t preferred_until,
 895                 struct in6_addr *excl_addr, uint8_t excl_length, const char *pclass)
 896 {
 897         if (!pclass)
 898                 pclass = (iface) ? iface->name : "local";
 899
 900         struct device_prefix *prefix = calloc(1, sizeof(*prefix) + strlen(pclass) + 1);
 901         prefix->length = length;
 902         prefix->addr = *addr;
 903         prefix->preferred_until = preferred_until;
 904         prefix->valid_until = valid_until;
 905         prefix->iface = iface;
 906         INIT_LIST_HEAD(&prefix->assignments);
 907
 908         if (excl_addr) {
 909                 prefix->excl_addr = *excl_addr;
 910                 prefix->excl_length = excl_length;
 911         }
 912
 913         strcpy(prefix->pclass, pclass);
 914
 915         if (iface)
 916                 vlist_add(&iface->proto_ip.prefix, &prefix->node, &prefix->addr);
 917         else
 918                 interface_update_prefix(NULL, &prefix->node, NULL);
 919
 920         return prefix;
 921 }
 922
 923 void
 924 interface_ip_set_ula_prefix(const char *prefix)
 925 {
 926         char buf[INET6_ADDRSTRLEN + 4] = {0}, *saveptr;
 927         if (prefix)
 928                 strncpy(buf, prefix, sizeof(buf) - 1);
 929         char *prefixaddr = strtok_r(buf, "/", &saveptr);
 930
 931         struct in6_addr addr;
 932         if (!prefixaddr || inet_pton(AF_INET6, prefixaddr, &addr) < 1) {
 933                 if (ula_prefix) {
 934                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
 935                         ula_prefix = NULL;
 936                 }
 937                 return;
 938         }
 939
 940         int length;
 941         char *prefixlen = strtok_r(NULL, ",", &saveptr);
 942         if (!prefixlen || (length = atoi(prefixlen)) < 1 || length > 64)
 943                 return;
 944
 945         if (!ula_prefix || !IN6_ARE_ADDR_EQUAL(&addr, &ula_prefix->addr) ||
 946                         ula_prefix->length != length) {
 947                 if (ula_prefix)
 948                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
 949
 950                 ula_prefix = interface_ip_add_device_prefix(NULL, &addr, length,
 951                                 0, 0, NULL, 0, NULL);
 952         }
 953 }
 954
 955 void
 956 interface_add_dns_server(struct interface_ip_settings *ip, const char *str)
 957 {
 958         struct dns_server *s;
 959
 960         s = calloc(1, sizeof(*s));
 961         if (!s)
 962                 return;
 963
 964         s->af = AF_INET;
 965         if (inet_pton(s->af, str, &s->addr.in))
 966                 goto add;
 967
 968         s->af = AF_INET6;
 969         if (inet_pton(s->af, str, &s->addr.in))
 970                 goto add;
 971
 972         free(s);
 973         return;
 974
 975 add:
 976         D(INTERFACE, "Add IPv%c DNS server: %s\n",
 977           s->af == AF_INET6 ? '6' : '4', str);
 978         vlist_simple_add(&ip->dns_servers, &s->node);
 979 }
 980
 981 void
 982 interface_add_dns_server_list(struct interface_ip_settings *ip, struct blob_attr *list)
 983 {
 984         struct blob_attr *cur;
 985         int rem;
 986
 987         blobmsg_for_each_attr(cur, list, rem) {
 988                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
 989                         continue;
 990
 991                 if (!blobmsg_check_attr(cur, NULL))
 992                         continue;
 993
 994                 interface_add_dns_server(ip, blobmsg_data(cur));
 995         }
 996 }
 997
 998 static void
 999 interface_add_dns_search_domain(struct interface_ip_settings *ip, const char *str)
1000 {
1001         struct dns_search_domain *s;
1002         int len = strlen(str);
1003
1004         s = calloc(1, sizeof(*s) + len + 1);
1005         if (!s)
1006                 return;
1007
1008         D(INTERFACE, "Add DNS search domain: %s\n", str);
1009         memcpy(s->name, str, len);
1010         vlist_simple_add(&ip->dns_search, &s->node);
1011 }
1012
1013 void
1014 interface_add_dns_search_list(struct interface_ip_settings *ip, struct blob_attr *list)
1015 {
1016         struct blob_attr *cur;
1017         int rem;
1018
1019         blobmsg_for_each_attr(cur, list, rem) {
1020                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1021                         continue;
1022
1023                 if (!blobmsg_check_attr(cur, NULL))
1024                         continue;
1025
1026                 interface_add_dns_search_domain(ip, blobmsg_data(cur));
1027         }
1028 }
1029
1030 static void
1031 write_resolv_conf_entries(FILE *f, struct interface_ip_settings *ip, const char *dev)
1032 {
1033         struct dns_server *s;
1034         struct dns_search_domain *d;
1035         const char *str;
1036         char buf[INET6_ADDRSTRLEN];
1037
1038         vlist_simple_for_each_element(&ip->dns_servers, s, node) {
1039                 str = inet_ntop(s->af, &s->addr, buf, sizeof(buf));
1040                 if (!str)
1041                         continue;
1042
1043                 if (s->af == AF_INET6 && IN6_IS_ADDR_LINKLOCAL(&s->addr))
1044                         fprintf(f, "nameserver %s%%%s\n", str, dev);
1045                 else
1046                         fprintf(f, "nameserver %s\n", str);
1047         }
1048
1049         vlist_simple_for_each_element(&ip->dns_search, d, node) {
1050                 fprintf(f, "search %s\n", d->name);
1051         }
1052 }
1053
1054 void
1055 interface_write_resolv_conf(void)
1056 {
1057         struct interface *iface;
1058         char *path = alloca(strlen(resolv_conf) + 5);
1059         FILE *f;
1060         uint32_t crcold, crcnew;
1061
1062         sprintf(path, "%s.tmp", resolv_conf);
1063         unlink(path);
1064         f = fopen(path, "w+");
1065         if (!f) {
1066                 D(INTERFACE, "Failed to open %s for writing\n", path);
1067                 return;
1068         }
1069
1070         vlist_for_each_element(&interfaces, iface, node) {
1071                 if (iface->state != IFS_UP)
1072                         continue;
1073
1074                 if (vlist_simple_empty(&iface->proto_ip.dns_search) &&
1075                     vlist_simple_empty(&iface->proto_ip.dns_servers) &&
1076                         vlist_simple_empty(&iface->config_ip.dns_search) &&
1077                     vlist_simple_empty(&iface->config_ip.dns_servers))
1078                         continue;
1079
1080                 fprintf(f, "# Interface %s\n", iface->name);
1081                 write_resolv_conf_entries(f, &iface->config_ip, iface->ifname);
1082                 if (!iface->proto_ip.no_dns)
1083                         write_resolv_conf_entries(f, &iface->proto_ip, iface->ifname);
1084         }
1085         fflush(f);
1086         rewind(f);
1087         crcnew = crc32_file(f);
1088         fclose(f);
1089
1090         crcold = crcnew + 1;
1091         f = fopen(resolv_conf, "r");
1092         if (f) {
1093                 crcold = crc32_file(f);
1094                 fclose(f);
1095         }
1096
1097         if (crcold == crcnew) {
1098                 unlink(path);
1099         } else if (rename(path, resolv_conf) < 0) {
1100                 D(INTERFACE, "Failed to replace %s\n", resolv_conf);
1101                 unlink(path);
1102         }
1103 }
1104
1105 void interface_ip_set_enabled(struct interface_ip_settings *ip, bool enabled)
1106 {
1107         struct device_addr *addr;
1108         struct device_route *route;
1109         struct device *dev;
1110
1111         ip->enabled = enabled;
1112         dev = ip->iface->l3_dev.dev;
1113         if (!dev)
1114                 return;
1115
1116         vlist_for_each_element(&ip->addr, addr, node) {
1117                 if (addr->enabled == enabled)
1118                         continue;
1119
1120                 if (enabled)
1121                         system_add_address(dev, addr);
1122                 else
1123                         system_del_address(dev, addr);
1124                 addr->enabled = enabled;
1125         }
1126
1127         vlist_for_each_element(&ip->route, route, node) {
1128                 bool _enabled = enabled;
1129
1130                 if (!enable_route(ip, route))
1131                         _enabled = false;
1132
1133                 if (route->enabled == _enabled)
1134                         continue;
1135
1136                 if (_enabled) {
1137                         if (!(route->flags & DEVROUTE_METRIC))
1138                                 route->metric = ip->iface->metric;
1139
1140                         system_add_route(dev, route);
1141                 } else
1142                         system_del_route(dev, route);
1143                 route->enabled = _enabled;
1144         }
1145
1146         struct device_prefix *c;
1147         struct device_prefix_assignment *a;
1148         list_for_each_entry(c, &prefixes, head)
1149                 list_for_each_entry(a, &c->assignments, head)
1150                         if (!strcmp(a->name, ip->iface->name))
1151                                 interface_set_prefix_address(a, c, ip->iface, enabled);
1152
1153         if (ip->iface && ip->iface->l3_dev.dev) {
1154                 set_ip_lo_policy(enabled, true, ip->iface);
1155                 set_ip_lo_policy(enabled, false, ip->iface);
1156
1157                 set_ip_source_policy(enabled, true, IPRULE_PRIORITY_REJECT + ip->iface->l3_dev.dev->ifindex,
1158                         NULL, 0, 0, ip->iface, "failed_policy");
1159         }
1160 }
1161
1162 void
1163 interface_ip_update_start(struct interface_ip_settings *ip)
1164 {
1165         if (ip != &ip->iface->config_ip) {
1166                 vlist_simple_update(&ip->dns_servers);
1167                 vlist_simple_update(&ip->dns_search);
1168         }
1169         vlist_update(&ip->route);
1170         vlist_update(&ip->addr);
1171         vlist_update(&ip->prefix);
1172 }
1173
1174 void
1175 interface_ip_update_complete(struct interface_ip_settings *ip)
1176 {
1177         vlist_simple_flush(&ip->dns_servers);
1178         vlist_simple_flush(&ip->dns_search);
1179         vlist_flush(&ip->route);
1180         vlist_flush(&ip->addr);
1181         vlist_flush(&ip->prefix);
1182         interface_write_resolv_conf();
1183 }
1184
1185 void
1186 interface_ip_flush(struct interface_ip_settings *ip)
1187 {
1188         if (ip == &ip->iface->proto_ip)
1189                 vlist_flush_all(&ip->iface->host_routes);
1190         vlist_simple_flush_all(&ip->dns_servers);
1191         vlist_simple_flush_all(&ip->dns_search);
1192         vlist_flush_all(&ip->route);
1193         vlist_flush_all(&ip->addr);
1194         vlist_flush_all(&ip->prefix);
1195 }
1196
1197 static void
1198 __interface_ip_init(struct interface_ip_settings *ip, struct interface *iface)
1199 {
1200         ip->iface = iface;
1201         ip->enabled = true;
1202         vlist_simple_init(&ip->dns_search, struct dns_search_domain, node);
1203         vlist_simple_init(&ip->dns_servers, struct dns_server, node);
1204         vlist_init(&ip->route, route_cmp, interface_update_proto_route);
1205         vlist_init(&ip->addr, addr_cmp, interface_update_proto_addr);
1206         vlist_init(&ip->prefix, prefix_cmp, interface_update_prefix);
1207 }
1208
1209 void
1210 interface_ip_init(struct interface *iface)
1211 {
1212         __interface_ip_init(&iface->proto_ip, iface);
1213         __interface_ip_init(&iface->config_ip, iface);
1214         vlist_init(&iface->host_routes, route_cmp, interface_update_host_route);
1215
1216 }
1217
1218 static void
1219 interface_ip_valid_until_handler(struct uloop_timeout *t)
1220 {
1221         time_t now = system_get_rtime();
1222         struct interface *iface;
1223         vlist_for_each_element(&interfaces, iface, node) {
1224                 if (iface->state != IFS_UP)
1225                         continue;
1226
1227                 struct device_addr *addr, *addrp;
1228                 struct device_route *route, *routep;
1229                 struct device_prefix *pref, *prefp;
1230
1231                 vlist_for_each_element_safe(&iface->proto_ip.addr, addr, node, addrp)
1232                         if (addr->valid_until && addr->valid_until < now)
1233                                 vlist_delete(&iface->proto_ip.addr, &addr->node);
1234
1235                 vlist_for_each_element_safe(&iface->proto_ip.route, route, node, routep)
1236                         if (route->valid_until && route->valid_until < now)
1237                                 vlist_delete(&iface->proto_ip.route, &route->node);
1238
1239                 vlist_for_each_element_safe(&iface->proto_ip.prefix, pref, node, prefp)
1240                         if (pref->valid_until && pref->valid_until < now)
1241                                 vlist_delete(&iface->proto_ip.prefix, &pref->node);
1242
1243         }
1244
1245         uloop_timeout_set(t, 1000);
1246 }
1247
1248 static void __init
1249 interface_ip_init_worker(void)
1250 {
1251         valid_until_timeout.cb = interface_ip_valid_until_handler;
1252         uloop_timeout_set(&valid_until_timeout, 1000);
1253 }