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
  22 #include "netifd.h"
  23 #include "device.h"
  24 #include "interface.h"
  25 #include "interface-ip.h"
  26 #include "proto.h"
  27 #include "ubus.h"
  28 #include "system.h"
  29
  30 enum {
  31         ROUTE_INTERFACE,
  32         ROUTE_TARGET,
  33         ROUTE_MASK,
  34         ROUTE_GATEWAY,
  35         ROUTE_METRIC,
  36         ROUTE_MTU,
  37         ROUTE_VALID,
  38         ROUTE_TABLE,
  39         ROUTE_SOURCE,
  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         [ROUTE_SOURCE] = { .name = "source", .type = BLOBMSG_TYPE_STRING },
  53 };
  54
  55 const struct uci_blob_param_list route_attr_list = {
  56         .n_params = __ROUTE_MAX,
  57         .params = route_attr,
  58 };
  59
  60
  61 struct list_head prefixes = LIST_HEAD_INIT(prefixes);
  62 static struct list_head source_tables = LIST_HEAD_INIT(source_tables);
  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 struct device_source_table*
 201 find_source_table(const struct device_route *route)
 202 {
 203         struct device_source_table key = {
 204                 .v6 = (route->flags & DEVADDR_FAMILY) == DEVADDR_INET6,
 205                 .addr = route->source,
 206                 .mask = route->sourcemask
 207         };
 208         struct device_source_table *c;
 209         list_for_each_entry(c, &source_tables, head)
 210                 if (!memcmp(&c->v6, &key.v6, sizeof(key) -
 211                                 offsetof(struct device_source_table, v6)))
 212                         return c;
 213         return NULL;
 214 }
 215
 216 static uint32_t
 217 get_source_table(const struct device_route *route)
 218 {
 219         if (route->table || route->sourcemask == 0)
 220                 return route->table;
 221
 222         struct device_source_table *tbl = find_source_table(route);
 223
 224         if (!tbl) {
 225                 tbl = calloc(1, sizeof(*tbl));
 226                 tbl->addr = route->source;
 227                 tbl->mask = route->sourcemask;
 228                 tbl->v6 = (route->flags & DEVADDR_FAMILY) == DEVADDR_INET6;
 229                 tbl->table = IPRULE_PRIORITY_SOURCE | (((~tbl->mask) & 0x7f) << 20);
 230
 231                 struct list_head *before = NULL;
 232                 struct device_source_table *c;
 233                 list_for_each_entry(c, &source_tables, head) {
 234                         if (c->table > tbl->table) {
 235                                 before = &c->head;
 236                                 break;
 237                         } else if (c->table == tbl->table) {
 238                                 ++tbl->table;
 239                         }
 240                 }
 241
 242                 if (!before)
 243                         before = &source_tables;
 244
 245                 list_add_tail(&tbl->head, before);
 246                 set_ip_source_policy(true, tbl->v6, tbl->table, &tbl->addr,
 247                                 tbl->mask, tbl->table, NULL, NULL);
 248         }
 249
 250         ++tbl->refcount;
 251         return tbl->table;
 252 }
 253
 254 static void
 255 put_source_table(const struct device_route *route)
 256 {
 257         struct device_source_table *tbl = find_source_table(route);
 258         if (tbl && tbl->table == route->table && --tbl->refcount == 0) {
 259                 set_ip_source_policy(false, tbl->v6, tbl->table, &tbl->addr,
 260                                 tbl->mask, tbl->table, NULL, NULL);
 261                 list_del(&tbl->head);
 262                 free(tbl);
 263         }
 264 }
 265
 266 static bool
 267 interface_ip_find_addr_target(struct interface *iface, union if_addr *a, bool v6)
 268 {
 269         return __find_ip_addr_target(&iface->proto_ip, a, v6) ||
 270                __find_ip_addr_target(&iface->config_ip, a, v6);
 271 }
 272
 273 static void
 274 interface_ip_find_route_target(struct interface *iface, union if_addr *a,
 275                                bool v6, struct device_route **route)
 276 {
 277         __find_ip_route_target(&iface->proto_ip, a, v6, route);
 278         __find_ip_route_target(&iface->config_ip, a, v6, route);
 279 }
 280
 281 struct interface *
 282 interface_ip_add_target_route(union if_addr *addr, bool v6, struct interface *iface)
 283 {
 284         struct device_route *route, *r_next = NULL;
 285         bool defaultroute_target = false;
 286         int addrsize = v6 ? sizeof(addr->in6) : sizeof(addr->in);
 287
 288         route = calloc(1, sizeof(*route));
 289         if (!route)
 290                 return NULL;
 291
 292         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
 293         route->mask = v6 ? 128 : 32;
 294         if (memcmp(&route->addr, addr, addrsize) == 0)
 295                 defaultroute_target = true;
 296         else
 297                 memcpy(&route->addr, addr, addrsize);
 298
 299         if (iface) {
 300                 /* look for locally addressable target first */
 301                 if (interface_ip_find_addr_target(iface, addr, v6))
 302                         goto done;
 303
 304                 /* do not stop at the first route, let the lookup compare
 305                  * masks to find the best match */
 306                 interface_ip_find_route_target(iface, addr, v6, &r_next);
 307         } else {
 308                 vlist_for_each_element(&interfaces, iface, node) {
 309                         /* look for locally addressable target first */
 310                         if (interface_ip_find_addr_target(iface, addr, v6))
 311                                 goto done;
 312
 313                         /* do not stop at the first route, let the lookup compare
 314                          * masks to find the best match */
 315                         interface_ip_find_route_target(iface, addr, v6, &r_next);
 316                 }
 317         }
 318
 319         if (!r_next) {
 320                 free(route);
 321                 return NULL;
 322         }
 323
 324         iface = r_next->iface;
 325         memcpy(&route->nexthop, &r_next->nexthop, sizeof(route->nexthop));
 326         route->mtu = r_next->mtu;
 327         route->metric = r_next->metric;
 328         route->table = r_next->table;
 329
 330 done:
 331         route->iface = iface;
 332         if (defaultroute_target)
 333                 free(route);
 334         else
 335                 vlist_add(&iface->host_routes, &route->node, route);
 336         return iface;
 337 }
 338
 339 void
 340 interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
 341 {
 342         struct interface_ip_settings *ip;
 343         struct blob_attr *tb[__ROUTE_MAX], *cur;
 344         struct device_route *route;
 345         int af = v6 ? AF_INET6 : AF_INET;
 346         bool is_proto_route = !!iface;
 347
 348         blobmsg_parse(route_attr, __ROUTE_MAX, tb, blobmsg_data(attr), blobmsg_data_len(attr));
 349
 350         if (!iface) {
 351                 if ((cur = tb[ROUTE_INTERFACE]) == NULL)
 352                         return;
 353
 354                 iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
 355                 if (!iface)
 356                         return;
 357
 358                 ip = &iface->config_ip;
 359         } else {
 360                 ip = &iface->proto_ip;
 361         }
 362
 363         route = calloc(1, sizeof(*route));
 364         if (!route)
 365                 return;
 366
 367         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
 368         route->mask = v6 ? 128 : 32;
 369         if ((cur = tb[ROUTE_MASK]) != NULL) {
 370                 route->mask = parse_netmask_string(blobmsg_data(cur), v6);
 371                 if (route->mask > (v6 ? 128 : 32))
 372                         goto error;
 373         }
 374
 375         if ((cur = tb[ROUTE_TARGET]) != NULL) {
 376                 if (!parse_ip_and_netmask(af, blobmsg_data(cur), &route->addr, &route->mask)) {
 377                         DPRINTF("Failed to parse route target: %s\n", (char *) blobmsg_data(cur));
 378                         goto error;
 379                 }
 380         }
 381
 382         if ((cur = tb[ROUTE_GATEWAY]) != NULL) {
 383                 if (!inet_pton(af, blobmsg_data(cur), &route->nexthop)) {
 384                         DPRINTF("Failed to parse route gateway: %s\n", (char *) blobmsg_data(cur));
 385                         goto error;
 386                 }
 387         }
 388
 389         if ((cur = tb[ROUTE_METRIC]) != NULL) {
 390                 route->metric = blobmsg_get_u32(cur);
 391                 route->flags |= DEVROUTE_METRIC;
 392         }
 393
 394         if ((cur = tb[ROUTE_MTU]) != NULL) {
 395                 route->mtu = blobmsg_get_u32(cur);
 396                 route->flags |= DEVROUTE_MTU;
 397         }
 398
 399         // Use source-based routing
 400         if ((cur = tb[ROUTE_SOURCE]) != NULL) {
 401                 char *saveptr, *source = alloca(blobmsg_data_len(cur));
 402                 memcpy(source, blobmsg_data(cur), blobmsg_data_len(cur));
 403
 404                 const char *addr = strtok_r(source, "/", &saveptr);
 405                 const char *mask = strtok_r(NULL, "/", &saveptr);
 406
 407                 if (inet_pton(af, addr, &route->source) < 1) {
 408                         DPRINTF("Failed to parse route source: %s\n", addr);
 409                         goto error;
 410                 }
 411
 412                 route->sourcemask = atoi(mask);
 413         } else if (is_proto_route) {
 414                 route->table = (v6) ? iface->ip6table : iface->ip4table;
 415                 route->flags |= DEVROUTE_SRCTABLE;
 416         }
 417
 418         if ((cur = tb[ROUTE_TABLE]) != NULL) {
 419                 if (!system_resolve_rt_table(blobmsg_data(cur), &route->table)) {
 420                         DPRINTF("Failed to resolve routing table: %s\n", (char *) blobmsg_data(cur));
 421                         goto error;
 422                 }
 423
 424                 if (route->table)
 425                         route->flags |= DEVROUTE_TABLE;
 426         }
 427
 428         if ((cur = tb[ROUTE_VALID]) != NULL) {
 429                 int64_t valid = blobmsg_get_u32(cur);
 430                 int64_t valid_until = valid + (int64_t)system_get_rtime();
 431                 if (valid_until <= LONG_MAX && valid != 0xffffffffLL) // Catch overflow
 432                         route->valid_until = valid_until;
 433         }
 434
 435         if (route->sourcemask) {
 436                 route->table = get_source_table(route);
 437                 route->flags |= DEVROUTE_SRCTABLE;
 438         }
 439
 440         vlist_add(&ip->route, &route->node, route);
 441         return;
 442
 443 error:
 444         free(route);
 445 }
 446
 447 static int
 448 addr_cmp(const void *k1, const void *k2, void *ptr)
 449 {
 450         return memcmp(k1, k2, sizeof(struct device_addr) -
 451                       offsetof(struct device_addr, flags));
 452 }
 453
 454 static int
 455 route_cmp(const void *k1, const void *k2, void *ptr)
 456 {
 457         const struct device_route *r1 = k1, *r2 = k2;
 458
 459         if (r1->mask != r2->mask)
 460                 return r2->mask - r1->mask;
 461
 462         if (r1->metric != r2->metric)
 463                 return r1->metric - r2->metric;
 464
 465         if (r1->flags != r2->flags)
 466                 return r2->flags - r1->flags;
 467
 468         if (r1->sourcemask != r2->sourcemask)
 469                 return r1->sourcemask - r2->sourcemask;
 470
 471         int maskcmp = memcmp(&r1->source, &r2->source, sizeof(r1->source));
 472         if (maskcmp)
 473                 return maskcmp;
 474
 475         return memcmp(&r1->addr, &r2->addr, sizeof(r1->addr));
 476 }
 477
 478 static int
 479 prefix_cmp(const void *k1, const void *k2, void *ptr)
 480 {
 481         return memcmp(k1, k2, offsetof(struct device_prefix, pclass) -
 482                         offsetof(struct device_prefix, addr));
 483 }
 484
 485 static void
 486 interface_handle_subnet_route(struct interface *iface, struct device_addr *addr, bool add)
 487 {
 488         struct device *dev = iface->l3_dev.dev;
 489         struct device_route route;
 490
 491         memset(&route, 0, sizeof(route));
 492         route.iface = iface;
 493         route.flags = addr->flags;
 494         route.mask = addr->mask;
 495         memcpy(&route.addr, &addr->addr, sizeof(route.addr));
 496         clear_if_addr(&route.addr, route.mask);
 497
 498         if (add) {
 499                 route.flags |= DEVADDR_KERNEL;
 500                 system_del_route(dev, &route);
 501
 502                 if (!(addr->flags & DEVADDR_OFFLINK)) {
 503                         route.flags &= ~DEVADDR_KERNEL;
 504                         route.metric = iface->metric;
 505                         system_add_route(dev, &route);
 506                 }
 507         } else {
 508                 if (!(addr->flags & DEVADDR_OFFLINK))
 509                         system_del_route(dev, &route);
 510         }
 511 }
 512
 513 static void
 514 interface_update_proto_addr(struct vlist_tree *tree,
 515                             struct vlist_node *node_new,
 516                             struct vlist_node *node_old)
 517 {
 518         struct interface_ip_settings *ip;
 519         struct interface *iface;
 520         struct device *dev;
 521         struct device_addr *a_new = NULL, *a_old = NULL;
 522         bool replace = false;
 523         bool keep = false;
 524         bool v6 = false;
 525
 526         ip = container_of(tree, struct interface_ip_settings, addr);
 527         iface = ip->iface;
 528         dev = iface->l3_dev.dev;
 529
 530         if (node_new) {
 531                 a_new = container_of(node_new, struct device_addr, node);
 532
 533                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
 534                     !a_new->broadcast) {
 535
 536                         uint32_t mask = ~0;
 537                         uint32_t *a = (uint32_t *) &a_new->addr;
 538
 539                         mask >>= a_new->mask;
 540                         a_new->broadcast = *a | htonl(mask);
 541                 }
 542         }
 543
 544         if (node_old)
 545                 a_old = container_of(node_old, struct device_addr, node);
 546
 547         if (a_new && a_old) {
 548                 keep = true;
 549
 550                 if (a_old->flags != a_new->flags)
 551                         keep = false;
 552
 553                 if (a_old->valid_until != a_new->valid_until ||
 554                                 a_old->preferred_until != a_new->preferred_until)
 555                         replace = true;
 556
 557                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
 558                     a_new->broadcast != a_old->broadcast)
 559                         keep = false;
 560         }
 561
 562         if (node_old) {
 563                 if (!(a_old->flags & DEVADDR_EXTERNAL) && a_old->enabled && !keep) {
 564                         interface_handle_subnet_route(iface, a_old, false);
 565
 566                         if ((a_old->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
 567                                 v6 = true;
 568
 569                         //This is needed for source routing to work correctly. If a device
 570                         //has two connections to a network using the same subnet, adding
 571                         //only the network-rule will cause packets to be routed through the
 572                         //first matching network (source IP matches both masks).
 573                         set_ip_source_policy(false, v6, IPRULE_PRIORITY_ADDR, &a_old->addr,
 574                                 (v6) ? 128 : 32, (v6) ? iface->ip6table : iface->ip4table,
 575                                                 NULL, NULL);
 576                         set_ip_source_policy(false, v6, IPRULE_PRIORITY_NW, &a_old->addr,
 577                                 a_old->mask, (v6) ? iface->ip6table : iface->ip4table, NULL, NULL);
 578
 579                         system_del_address(dev, a_old);
 580                 }
 581                 free(a_old->pclass);
 582                 free(a_old);
 583         }
 584
 585         if (node_new) {
 586                 a_new->enabled = true;
 587                 if (!(a_new->flags & DEVADDR_EXTERNAL) && (!keep || replace)) {
 588                         system_add_address(dev, a_new);
 589
 590                         if (!keep) {
 591                                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
 592                                         v6 = true;
 593
 594                                 set_ip_source_policy(true, v6, IPRULE_PRIORITY_ADDR, &a_new->addr,
 595                                         (v6) ? 128 : 32, (v6) ? iface->ip6table : iface->ip4table,
 596                                                         NULL, NULL);
 597                                 set_ip_source_policy(true, v6, IPRULE_PRIORITY_NW, &a_new->addr,
 598                                         a_new->mask, (v6) ? iface->ip6table : iface->ip4table,
 599                                                         NULL, NULL);
 600                         }
 601
 602                         if ((a_new->flags & DEVADDR_OFFLINK) || iface->metric)
 603                                 interface_handle_subnet_route(iface, a_new, true);
 604                 }
 605         }
 606 }
 607
 608 static bool
 609 enable_route(struct interface_ip_settings *ip, struct device_route *route)
 610 {
 611         if (ip->no_defaultroute && !route->mask)
 612                 return false;
 613
 614         return ip->enabled;
 615 }
 616
 617 static void
 618 interface_update_proto_route(struct vlist_tree *tree,
 619                              struct vlist_node *node_new,
 620                              struct vlist_node *node_old)
 621 {
 622         struct interface_ip_settings *ip;
 623         struct interface *iface;
 624         struct device *dev;
 625         struct device_route *route_old, *route_new;
 626         bool keep = false;
 627
 628         ip = container_of(tree, struct interface_ip_settings, route);
 629         iface = ip->iface;
 630         dev = iface->l3_dev.dev;
 631
 632         route_old = container_of(node_old, struct device_route, node);
 633         route_new = container_of(node_new, struct device_route, node);
 634
 635         if (node_old && node_new)
 636                 keep = !memcmp(&route_old->nexthop, &route_new->nexthop, sizeof(route_old->nexthop));
 637
 638         if (node_old) {
 639                 if (!(route_old->flags & DEVADDR_EXTERNAL) && route_old->enabled && !keep)
 640                         system_del_route(dev, route_old);
 641
 642                 put_source_table(route_old);
 643                 free(route_old);
 644         }
 645
 646         if (node_new) {
 647                 bool _enabled = enable_route(ip, route_new);
 648
 649                 if (!(route_new->flags & DEVROUTE_METRIC))
 650                         route_new->metric = iface->metric;
 651
 652                 if (!(route_new->flags & DEVADDR_EXTERNAL) && !keep && _enabled)
 653                         system_add_route(dev, route_new);
 654
 655                 route_new->iface = iface;
 656                 route_new->enabled = _enabled;
 657         }
 658 }
 659
 660 static void
 661 interface_update_host_route(struct vlist_tree *tree,
 662                              struct vlist_node *node_new,
 663                              struct vlist_node *node_old)
 664 {
 665         struct interface *iface;
 666         struct device *dev;
 667         struct device_route *route_old, *route_new;
 668
 669         iface = container_of(tree, struct interface, host_routes);
 670         dev = iface->l3_dev.dev;
 671
 672         route_old = container_of(node_old, struct device_route, node);
 673         route_new = container_of(node_new, struct device_route, node);
 674
 675         if (node_old) {
 676                 system_del_route(dev, route_old);
 677                 free(route_old);
 678         }
 679
 680         if (node_new)
 681                 system_add_route(dev, route_new);
 682 }
 683
 684
 685 static void
 686 interface_set_prefix_address(struct device_prefix_assignment *assignment,
 687                 const struct device_prefix *prefix, struct interface *iface, bool add)
 688 {
 689         const struct interface *uplink = prefix->iface;
 690         if (!iface->l3_dev.dev)
 691                 return;
 692
 693         struct device *l3_downlink = iface->l3_dev.dev;
 694
 695         struct device_addr addr;
 696         memset(&addr, 0, sizeof(addr));
 697         addr.addr.in6 = prefix->addr;
 698         addr.addr.in6.s6_addr32[1] |= htonl(assignment->assigned);
 699         addr.addr.in6.s6_addr[15] += 1;
 700         addr.mask = assignment->length;
 701         addr.flags = DEVADDR_INET6;
 702         addr.preferred_until = prefix->preferred_until;
 703         addr.valid_until = prefix->valid_until;
 704
 705         if (!add && assignment->enabled) {
 706                 time_t now = system_get_rtime();
 707                 addr.preferred_until = now;
 708                 if (!addr.valid_until || addr.valid_until - now > 7200)
 709                         addr.valid_until = now + 7200;
 710                 system_add_address(l3_downlink, &addr);
 711                 if (prefix->iface) {
 712                         set_ip_source_policy(false, true, IPRULE_PRIORITY_NW, &addr.addr,
 713                                         addr.mask, prefix->iface->ip6table, iface, NULL);
 714
 715                         set_ip_source_policy(false, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 716                                                         addr.mask, 0, iface, "unreachable");
 717                 }
 718
 719                 assignment->enabled = false;
 720         } else if (add && (iface->state == IFS_UP || iface->state == IFS_SETUP)) {
 721                 system_add_address(l3_downlink, &addr);
 722                 if (prefix->iface && !assignment->enabled) {
 723                         set_ip_source_policy(true, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 724                                         addr.mask, 0, iface, "unreachable");
 725
 726                         set_ip_source_policy(true, true, IPRULE_PRIORITY_NW, &addr.addr,
 727                                         addr.mask, prefix->iface->ip6table, iface, NULL);
 728                 }
 729                 if (uplink && uplink->l3_dev.dev) {
 730                         int mtu = system_update_ipv6_mtu(
 731                                         uplink->l3_dev.dev, 0);
 732                         if (mtu > 0)
 733                                 system_update_ipv6_mtu(l3_downlink, mtu);
 734                 }
 735                 assignment->enabled = true;
 736         }
 737 }
 738
 739 static bool interface_prefix_assign(struct list_head *list,
 740                 struct device_prefix_assignment *assign)
 741 {
 742         int32_t current = 0, asize = (1 << (64 - assign->length)) - 1;
 743         struct device_prefix_assignment *c;
 744         list_for_each_entry(c, list, head) {
 745                 if (assign->assigned != -1) {
 746                         if (assign->assigned >= current && assign->assigned + asize < c->assigned) {
 747                                 list_add_tail(&assign->head, &c->head);
 748                                 return true;
 749                         }
 750                 } else if (assign->assigned == -1) {
 751                         current = (current + asize) & (~asize);
 752                         if (current + asize < c->assigned) {
 753                                 assign->assigned = current;
 754                                 list_add_tail(&assign->head, &c->head);
 755                                 return true;
 756                         }
 757                 }
 758                 current = (c->assigned + (1 << (64 - c->length)));
 759         }
 760         return false;
 761 }
 762
 763 static void interface_update_prefix_assignments(struct device_prefix *prefix, bool setup)
 764 {
 765         struct device_prefix_assignment *c;
 766         struct interface *iface;
 767
 768         // Delete all assignments
 769         while (!list_empty(&prefix->assignments)) {
 770                 c = list_first_entry(&prefix->assignments,
 771                                 struct device_prefix_assignment, head);
 772                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 773                         interface_set_prefix_address(c, prefix, iface, false);
 774                 list_del(&c->head);
 775                 free(c);
 776         }
 777
 778         if (!setup)
 779                 return;
 780
 781         // End-of-assignment sentinel
 782         c = malloc(sizeof(*c) + 1);
 783         c->assigned = 1 << (64 - prefix->length);
 784         c->length = 64;
 785         c->name[0] = 0;
 786         list_add(&c->head, &prefix->assignments);
 787
 788         // Excluded prefix
 789         if (prefix->excl_length > 0) {
 790                 const char name[] = "!excluded";
 791                 c = malloc(sizeof(*c) + sizeof(name));
 792                 c->assigned = ntohl(prefix->excl_addr.s6_addr32[1]) &
 793                                 ((1 << (64 - prefix->length)) - 1);
 794                 c->length = prefix->excl_length;
 795                 memcpy(c->name, name, sizeof(name));
 796                 list_add(&c->head, &prefix->assignments);
 797         }
 798
 799         bool assigned_any = false;
 800         struct list_head assign_later = LIST_HEAD_INIT(assign_later);
 801         vlist_for_each_element(&interfaces, iface, node) {
 802                 if (iface->assignment_length < 48 ||
 803                                 iface->assignment_length > 64)
 804                         continue;
 805
 806                 // Test whether there is a matching class
 807                 if (!list_empty(&iface->assignment_classes)) {
 808                         bool found = false;
 809
 810                         struct interface_assignment_class *c;
 811                         list_for_each_entry(c, &iface->assignment_classes, head) {
 812                                 if (!strcmp(c->name, prefix->pclass)) {
 813                                         found = true;
 814                                         break;
 815                                 }
 816                         }
 817
 818                         if (!found)
 819                                 continue;
 820                 }
 821
 822                 size_t namelen = strlen(iface->name) + 1;
 823                 c = malloc(sizeof(*c) + namelen);
 824                 c->length = iface->assignment_length;
 825                 c->assigned = iface->assignment_hint;
 826                 c->enabled = false;
 827                 memcpy(c->name, iface->name, namelen);
 828
 829                 // First process all custom assignments, put all others in later-list
 830                 if (c->assigned == -1 || !interface_prefix_assign(&prefix->assignments, c)) {
 831                         if (c->assigned != -1) {
 832                                 c->assigned = -1;
 833                                 netifd_log_message(L_WARNING, "Failed to assign requested subprefix "
 834                                                 "of size %hhu for %s, trying other\n", c->length, c->name);
 835                         }
 836
 837                         struct list_head *next = &assign_later;
 838                         struct device_prefix_assignment *n;
 839                         list_for_each_entry(n, &assign_later, head) {
 840                                 if (n->length < c->length) {
 841                                         next = &n->head;
 842                                         break;
 843                                 }
 844                         }
 845                         list_add_tail(&c->head, next);
 846                 }
 847
 848                 if (c->assigned != -1)
 849                         assigned_any = true;
 850         }
 851
 852         // Then try to assign all other + failed custom assignments
 853         while (!list_empty(&assign_later)) {
 854                 c = list_first_entry(&assign_later, struct device_prefix_assignment, head);
 855                 list_del(&c->head);
 856
 857                 bool assigned = false;
 858                 do {
 859                         assigned = interface_prefix_assign(&prefix->assignments, c);
 860                 } while (!assigned && ++c->length <= 64);
 861
 862                 if (!assigned) {
 863                         netifd_log_message(L_WARNING, "Failed to assign subprefix "
 864                                         "of size %hhu for %s\n", c->length, c->name);
 865                         free(c);
 866                 } else {
 867                         assigned_any = true;
 868                 }
 869         }
 870
 871         list_for_each_entry(c, &prefix->assignments, head)
 872                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 873                         interface_set_prefix_address(c, prefix, iface, true);
 874
 875         if (!assigned_any)
 876                 netifd_log_message(L_WARNING, "You have delegated IPv6-prefixes but haven't assigned them "
 877                                 "to any interface. Did you forget to set option ip6assign on your lan-interfaces?");
 878 }
 879
 880
 881 void interface_refresh_assignments(bool hint)
 882 {
 883         static bool refresh = false;
 884         if (!hint && refresh) {
 885                 struct device_prefix *p;
 886                 list_for_each_entry(p, &prefixes, head)
 887                         interface_update_prefix_assignments(p, true);
 888         }
 889         refresh = hint;
 890 }
 891
 892
 893 static void
 894 interface_update_prefix(struct vlist_tree *tree,
 895                              struct vlist_node *node_new,
 896                              struct vlist_node *node_old)
 897 {
 898         struct device_prefix *prefix_old, *prefix_new;
 899         prefix_old = container_of(node_old, struct device_prefix, node);
 900         prefix_new = container_of(node_new, struct device_prefix, node);
 901
 902         struct device_route route;
 903         memset(&route, 0, sizeof(route));
 904         route.flags = DEVADDR_INET6;
 905         route.metric = INT32_MAX;
 906         route.mask = (node_new) ? prefix_new->length : prefix_old->length;
 907         route.addr.in6 = (node_new) ? prefix_new->addr : prefix_old->addr;
 908
 909
 910         struct device_prefix_assignment *c;
 911         struct interface *iface;
 912
 913         if (node_old && node_new) {
 914                 // Move assignments and refresh addresses to update valid times
 915                 list_splice(&prefix_old->assignments, &prefix_new->assignments);
 916
 917                 list_for_each_entry(c, &prefix_new->assignments, head)
 918                         if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 919                                 interface_set_prefix_address(c, prefix_new, iface, true);
 920         } else if (node_new) {
 921                 // Set null-route to avoid routing loops
 922                 system_add_route(NULL, &route);
 923
 924                 if (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation)
 925                         interface_update_prefix_assignments(prefix_new, true);
 926         } else if (node_old) {
 927                 // Remove null-route
 928                 interface_update_prefix_assignments(prefix_old, false);
 929                 system_del_route(NULL, &route);
 930         }
 931
 932         if (node_old) {
 933                 if (prefix_old->head.next)
 934                         list_del(&prefix_old->head);
 935                 free(prefix_old);
 936         }
 937
 938         if (node_new && (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation))
 939                 list_add(&prefix_new->head, &prefixes);
 940
 941 }
 942
 943 struct device_prefix*
 944 interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
 945                 uint8_t length, time_t valid_until, time_t preferred_until,
 946                 struct in6_addr *excl_addr, uint8_t excl_length, const char *pclass)
 947 {
 948         if (!pclass)
 949                 pclass = (iface) ? iface->name : "local";
 950
 951         struct device_prefix *prefix = calloc(1, sizeof(*prefix) + strlen(pclass) + 1);
 952         prefix->length = length;
 953         prefix->addr = *addr;
 954         prefix->preferred_until = preferred_until;
 955         prefix->valid_until = valid_until;
 956         prefix->iface = iface;
 957         INIT_LIST_HEAD(&prefix->assignments);
 958
 959         if (excl_addr) {
 960                 prefix->excl_addr = *excl_addr;
 961                 prefix->excl_length = excl_length;
 962         }
 963
 964         strcpy(prefix->pclass, pclass);
 965
 966         if (iface)
 967                 vlist_add(&iface->proto_ip.prefix, &prefix->node, &prefix->addr);
 968         else
 969                 interface_update_prefix(NULL, &prefix->node, NULL);
 970
 971         return prefix;
 972 }
 973
 974 void
 975 interface_ip_set_ula_prefix(const char *prefix)
 976 {
 977         char buf[INET6_ADDRSTRLEN + 4] = {0}, *saveptr;
 978         if (prefix)
 979                 strncpy(buf, prefix, sizeof(buf) - 1);
 980         char *prefixaddr = strtok_r(buf, "/", &saveptr);
 981
 982         struct in6_addr addr;
 983         if (!prefixaddr || inet_pton(AF_INET6, prefixaddr, &addr) < 1) {
 984                 if (ula_prefix) {
 985                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
 986                         ula_prefix = NULL;
 987                 }
 988                 return;
 989         }
 990
 991         int length;
 992         char *prefixlen = strtok_r(NULL, ",", &saveptr);
 993         if (!prefixlen || (length = atoi(prefixlen)) < 1 || length > 64)
 994                 return;
 995
 996         if (!ula_prefix || !IN6_ARE_ADDR_EQUAL(&addr, &ula_prefix->addr) ||
 997                         ula_prefix->length != length) {
 998                 if (ula_prefix)
 999                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
1000
1001                 ula_prefix = interface_ip_add_device_prefix(NULL, &addr, length,
1002                                 0, 0, NULL, 0, NULL);
1003         }
1004 }
1005
1006 void
1007 interface_add_dns_server(struct interface_ip_settings *ip, const char *str)
1008 {
1009         struct dns_server *s;
1010
1011         s = calloc(1, sizeof(*s));
1012         if (!s)
1013                 return;
1014
1015         s->af = AF_INET;
1016         if (inet_pton(s->af, str, &s->addr.in))
1017                 goto add;
1018
1019         s->af = AF_INET6;
1020         if (inet_pton(s->af, str, &s->addr.in))
1021                 goto add;
1022
1023         free(s);
1024         return;
1025
1026 add:
1027         D(INTERFACE, "Add IPv%c DNS server: %s\n",
1028           s->af == AF_INET6 ? '6' : '4', str);
1029         vlist_simple_add(&ip->dns_servers, &s->node);
1030 }
1031
1032 void
1033 interface_add_dns_server_list(struct interface_ip_settings *ip, struct blob_attr *list)
1034 {
1035         struct blob_attr *cur;
1036         int rem;
1037
1038         blobmsg_for_each_attr(cur, list, rem) {
1039                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1040                         continue;
1041
1042                 if (!blobmsg_check_attr(cur, NULL))
1043                         continue;
1044
1045                 interface_add_dns_server(ip, blobmsg_data(cur));
1046         }
1047 }
1048
1049 static void
1050 interface_add_dns_search_domain(struct interface_ip_settings *ip, const char *str)
1051 {
1052         struct dns_search_domain *s;
1053         int len = strlen(str);
1054
1055         s = calloc(1, sizeof(*s) + len + 1);
1056         if (!s)
1057                 return;
1058
1059         D(INTERFACE, "Add DNS search domain: %s\n", str);
1060         memcpy(s->name, str, len);
1061         vlist_simple_add(&ip->dns_search, &s->node);
1062 }
1063
1064 void
1065 interface_add_dns_search_list(struct interface_ip_settings *ip, struct blob_attr *list)
1066 {
1067         struct blob_attr *cur;
1068         int rem;
1069
1070         blobmsg_for_each_attr(cur, list, rem) {
1071                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1072                         continue;
1073
1074                 if (!blobmsg_check_attr(cur, NULL))
1075                         continue;
1076
1077                 interface_add_dns_search_domain(ip, blobmsg_data(cur));
1078         }
1079 }
1080
1081 static void
1082 write_resolv_conf_entries(FILE *f, struct interface_ip_settings *ip)
1083 {
1084         struct dns_server *s;
1085         struct dns_search_domain *d;
1086         const char *str;
1087         char buf[INET6_ADDRSTRLEN];
1088
1089         vlist_simple_for_each_element(&ip->dns_servers, s, node) {
1090                 str = inet_ntop(s->af, &s->addr, buf, sizeof(buf));
1091                 if (!str)
1092                         continue;
1093
1094                 fprintf(f, "nameserver %s\n", str);
1095         }
1096
1097         vlist_simple_for_each_element(&ip->dns_search, d, node) {
1098                 fprintf(f, "search %s\n", d->name);
1099         }
1100 }
1101
1102 void
1103 interface_write_resolv_conf(void)
1104 {
1105         struct interface *iface;
1106         char *path = alloca(strlen(resolv_conf) + 5);
1107         FILE *f;
1108         uint32_t crcold, crcnew;
1109
1110         sprintf(path, "%s.tmp", resolv_conf);
1111         unlink(path);
1112         f = fopen(path, "w+");
1113         if (!f) {
1114                 D(INTERFACE, "Failed to open %s for writing\n", path);
1115                 return;
1116         }
1117
1118         vlist_for_each_element(&interfaces, iface, node) {
1119                 if (iface->state != IFS_UP)
1120                         continue;
1121
1122                 if (vlist_simple_empty(&iface->proto_ip.dns_search) &&
1123                     vlist_simple_empty(&iface->proto_ip.dns_servers) &&
1124                         vlist_simple_empty(&iface->config_ip.dns_search) &&
1125                     vlist_simple_empty(&iface->config_ip.dns_servers))
1126                         continue;
1127
1128                 fprintf(f, "# Interface %s\n", iface->name);
1129                 write_resolv_conf_entries(f, &iface->config_ip);
1130                 if (!iface->proto_ip.no_dns)
1131                         write_resolv_conf_entries(f, &iface->proto_ip);
1132         }
1133         fflush(f);
1134         rewind(f);
1135         crcnew = crc32_file(f);
1136         fclose(f);
1137
1138         crcold = crcnew + 1;
1139         f = fopen(resolv_conf, "r");
1140         if (f) {
1141                 crcold = crc32_file(f);
1142                 fclose(f);
1143         }
1144
1145         if (crcold == crcnew) {
1146                 unlink(path);
1147         } else if (rename(path, resolv_conf) < 0) {
1148                 D(INTERFACE, "Failed to replace %s\n", resolv_conf);
1149                 unlink(path);
1150         }
1151 }
1152
1153 void interface_ip_set_enabled(struct interface_ip_settings *ip, bool enabled)
1154 {
1155         struct device_addr *addr;
1156         struct device_route *route;
1157         struct device *dev;
1158
1159         ip->enabled = enabled;
1160         dev = ip->iface->l3_dev.dev;
1161         if (!dev)
1162                 return;
1163
1164         vlist_for_each_element(&ip->addr, addr, node) {
1165                 if (addr->enabled == enabled)
1166                         continue;
1167
1168                 if (enabled)
1169                         system_add_address(dev, addr);
1170                 else
1171                         system_del_address(dev, addr);
1172                 addr->enabled = enabled;
1173         }
1174
1175         vlist_for_each_element(&ip->route, route, node) {
1176                 bool _enabled = enabled;
1177
1178                 if (!enable_route(ip, route))
1179                         _enabled = false;
1180
1181                 if (route->enabled == _enabled)
1182                         continue;
1183
1184                 if (_enabled) {
1185                         if (!(route->flags & DEVROUTE_METRIC))
1186                                 route->metric = ip->iface->metric;
1187
1188                         system_add_route(dev, route);
1189                 } else
1190                         system_del_route(dev, route);
1191                 route->enabled = _enabled;
1192         }
1193
1194         struct device_prefix *c;
1195         struct device_prefix_assignment *a;
1196         list_for_each_entry(c, &prefixes, head)
1197                 list_for_each_entry(a, &c->assignments, head)
1198                         if (!strcmp(a->name, ip->iface->name))
1199                                 interface_set_prefix_address(a, c, ip->iface, enabled);
1200
1201         if (ip->iface && ip->iface->l3_dev.dev) {
1202                 set_ip_lo_policy(enabled, true, ip->iface);
1203                 set_ip_lo_policy(enabled, false, ip->iface);
1204
1205                 set_ip_source_policy(enabled, true, IPRULE_PRIORITY_REJECT + ip->iface->l3_dev.dev->ifindex,
1206                         NULL, 0, 0, ip->iface, "failed_policy");
1207         }
1208 }
1209
1210 void
1211 interface_ip_update_start(struct interface_ip_settings *ip)
1212 {
1213         if (ip != &ip->iface->config_ip) {
1214                 vlist_simple_update(&ip->dns_servers);
1215                 vlist_simple_update(&ip->dns_search);
1216         }
1217         vlist_update(&ip->route);
1218         vlist_update(&ip->addr);
1219         vlist_update(&ip->prefix);
1220 }
1221
1222 void
1223 interface_ip_update_complete(struct interface_ip_settings *ip)
1224 {
1225         vlist_simple_flush(&ip->dns_servers);
1226         vlist_simple_flush(&ip->dns_search);
1227         vlist_flush(&ip->route);
1228         vlist_flush(&ip->addr);
1229         vlist_flush(&ip->prefix);
1230         interface_write_resolv_conf();
1231 }
1232
1233 void
1234 interface_ip_flush(struct interface_ip_settings *ip)
1235 {
1236         if (ip == &ip->iface->proto_ip)
1237                 vlist_flush_all(&ip->iface->host_routes);
1238         vlist_simple_flush_all(&ip->dns_servers);
1239         vlist_simple_flush_all(&ip->dns_search);
1240         vlist_flush_all(&ip->route);
1241         vlist_flush_all(&ip->addr);
1242         vlist_flush_all(&ip->prefix);
1243 }
1244
1245 static void
1246 __interface_ip_init(struct interface_ip_settings *ip, struct interface *iface)
1247 {
1248         ip->iface = iface;
1249         ip->enabled = true;
1250         vlist_simple_init(&ip->dns_search, struct dns_search_domain, node);
1251         vlist_simple_init(&ip->dns_servers, struct dns_server, node);
1252         vlist_init(&ip->route, route_cmp, interface_update_proto_route);
1253         vlist_init(&ip->addr, addr_cmp, interface_update_proto_addr);
1254         vlist_init(&ip->prefix, prefix_cmp, interface_update_prefix);
1255 }
1256
1257 void
1258 interface_ip_init(struct interface *iface)
1259 {
1260         __interface_ip_init(&iface->proto_ip, iface);
1261         __interface_ip_init(&iface->config_ip, iface);
1262         vlist_init(&iface->host_routes, route_cmp, interface_update_host_route);
1263
1264 }
1265
1266 static void
1267 interface_ip_valid_until_handler(struct uloop_timeout *t)
1268 {
1269         time_t now = system_get_rtime();
1270         struct interface *iface;
1271         vlist_for_each_element(&interfaces, iface, node) {
1272                 if (iface->state != IFS_UP)
1273                         continue;
1274
1275                 struct device_addr *addr, *addrp;
1276                 struct device_route *route, *routep;
1277                 struct device_prefix *pref, *prefp;
1278
1279                 vlist_for_each_element_safe(&iface->proto_ip.addr, addr, node, addrp)
1280                         if (addr->valid_until && addr->valid_until < now)
1281                                 vlist_delete(&iface->proto_ip.addr, &addr->node);
1282
1283                 vlist_for_each_element_safe(&iface->proto_ip.route, route, node, routep)
1284                         if (route->valid_until && route->valid_until < now)
1285                                 vlist_delete(&iface->proto_ip.route, &route->node);
1286
1287                 vlist_for_each_element_safe(&iface->proto_ip.prefix, pref, node, prefp)
1288                         if (pref->valid_until && pref->valid_until < now)
1289                                 vlist_delete(&iface->proto_ip.prefix, &pref->node);
1290
1291         }
1292
1293         uloop_timeout_set(t, 1000);
1294 }
1295
1296 static void __init
1297 interface_ip_init_worker(void)
1298 {
1299         valid_until_timeout.cb = interface_ip_valid_until_handler;
1300         uloop_timeout_set(&valid_until_timeout, 1000);
1301 }