interface-ip.c

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