interface-ip.c

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