interface-ip.c

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