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                 /* only set the table flag if not using the main (default) table */
 369                 if (system_is_default_rt_table(route->table))
 370                         route->table = 0;
 371
 372                 if (route->table)
 373                         route->flags |= DEVROUTE_TABLE;
 374         }
 375
 376         if ((cur = tb[ROUTE_VALID]) != NULL) {
 377                 int64_t valid = blobmsg_get_u32(cur);
 378                 int64_t valid_until = valid + (int64_t)system_get_rtime();
 379                 if (valid_until <= LONG_MAX && valid != 0xffffffffLL) // Catch overflow
 380                         route->valid_until = valid_until;
 381         }
 382
 383         if ((cur = tb[ROUTE_TYPE]) != NULL) {
 384                 if (!system_resolve_rt_type(blobmsg_data(cur), &route->type)) {
 385                         DPRINTF("Failed to resolve routing type: %s\n", (char *) blobmsg_data(cur));
 386                         goto error;
 387                 }
 388                 route->flags |= DEVROUTE_TYPE;
 389         }
 390
 391         vlist_add(&ip->route, &route->node, route);
 392         return;
 393
 394 error:
 395         free(route);
 396 }
 397
 398 static int
 399 addr_cmp(const void *k1, const void *k2, void *ptr)
 400 {
 401         return memcmp(k1, k2, sizeof(struct device_addr) -
 402                       offsetof(struct device_addr, flags));
 403 }
 404
 405 static int
 406 route_cmp(const void *k1, const void *k2, void *ptr)
 407 {
 408         const struct device_route *r1 = k1, *r2 = k2;
 409
 410         if (r1->mask != r2->mask)
 411                 return r2->mask - r1->mask;
 412
 413         if (r1->metric != r2->metric)
 414                 return r1->metric - r2->metric;
 415
 416         if (r1->flags != r2->flags)
 417                 return r2->flags - r1->flags;
 418
 419         if (r1->sourcemask != r2->sourcemask)
 420                 return r1->sourcemask - r2->sourcemask;
 421
 422         if (r1->table != r2->table)
 423                 return r1->table - r2->table;
 424
 425         int maskcmp = memcmp(&r1->source, &r2->source, sizeof(r1->source));
 426         if (maskcmp)
 427                 return maskcmp;
 428
 429         return memcmp(&r1->addr, &r2->addr, sizeof(r1->addr));
 430 }
 431
 432 static int
 433 prefix_cmp(const void *k1, const void *k2, void *ptr)
 434 {
 435         return memcmp(k1, k2, offsetof(struct device_prefix, pclass) -
 436                         offsetof(struct device_prefix, addr));
 437 }
 438
 439 static void
 440 interface_handle_subnet_route(struct interface *iface, struct device_addr *addr, bool add)
 441 {
 442         struct device *dev = iface->l3_dev.dev;
 443         struct device_route route;
 444
 445         memset(&route, 0, sizeof(route));
 446         route.iface = iface;
 447         route.flags = addr->flags;
 448         route.mask = addr->mask;
 449         memcpy(&route.addr, &addr->addr, sizeof(route.addr));
 450         clear_if_addr(&route.addr, route.mask);
 451
 452         if (add) {
 453                 route.flags |= DEVADDR_KERNEL;
 454                 system_del_route(dev, &route);
 455
 456                 if (!(addr->flags & DEVADDR_OFFLINK)) {
 457                         route.flags &= ~DEVADDR_KERNEL;
 458                         route.metric = iface->metric;
 459                         system_add_route(dev, &route);
 460                 }
 461         } else {
 462                 if (!(addr->flags & DEVADDR_OFFLINK))
 463                         system_del_route(dev, &route);
 464         }
 465 }
 466
 467 static void
 468 interface_update_proto_addr(struct vlist_tree *tree,
 469                             struct vlist_node *node_new,
 470                             struct vlist_node *node_old)
 471 {
 472         struct interface_ip_settings *ip;
 473         struct interface *iface;
 474         struct device *dev;
 475         struct device_addr *a_new = NULL, *a_old = NULL;
 476         bool replace = false;
 477         bool keep = false;
 478         bool v6 = false;
 479
 480         ip = container_of(tree, struct interface_ip_settings, addr);
 481         iface = ip->iface;
 482         dev = iface->l3_dev.dev;
 483
 484         if (!node_new || !node_old)
 485                 iface->updated |= IUF_ADDRESS;
 486
 487         if (node_new) {
 488                 a_new = container_of(node_new, struct device_addr, node);
 489
 490                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
 491                     !a_new->broadcast) {
 492
 493                         uint32_t mask = ~0;
 494                         uint32_t *a = (uint32_t *) &a_new->addr;
 495
 496                         mask >>= a_new->mask;
 497                         a_new->broadcast = *a | htonl(mask);
 498                 }
 499         }
 500
 501         if (node_old)
 502                 a_old = container_of(node_old, struct device_addr, node);
 503
 504         if (a_new && a_old) {
 505                 keep = true;
 506
 507                 if (a_old->flags != a_new->flags || a_old->failed)
 508                         keep = false;
 509
 510                 if (a_old->valid_until != a_new->valid_until ||
 511                                 a_old->preferred_until != a_new->preferred_until)
 512                         replace = true;
 513
 514                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
 515                     a_new->broadcast != a_old->broadcast)
 516                         keep = false;
 517         }
 518
 519         if (node_old) {
 520                 if (!(a_old->flags & DEVADDR_EXTERNAL) && a_old->enabled && !keep) {
 521                         interface_handle_subnet_route(iface, a_old, false);
 522
 523                         if ((a_old->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
 524                                 v6 = true;
 525
 526                         unsigned int table = (v6) ? iface->ip6table : iface->ip4table;
 527
 528                         //This is needed for source routing to work correctly. If a device
 529                         //has two connections to a network using the same subnet, adding
 530                         //only the network-rule will cause packets to be routed through the
 531                         //first matching network (source IP matches both masks).
 532                         if (table) {
 533                                 set_ip_source_policy(false, v6, IPRULE_PRIORITY_ADDR, &a_old->addr,
 534                                                 (v6) ? 128 : 32, table, NULL, NULL);
 535                                 set_ip_source_policy(false, v6, IPRULE_PRIORITY_NW, &a_old->addr,
 536                                                 a_old->mask, table, NULL, NULL);
 537                         }
 538
 539                         system_del_address(dev, a_old);
 540                 }
 541                 free(a_old->pclass);
 542                 free(a_old);
 543         }
 544
 545         if (node_new) {
 546                 a_new->enabled = true;
 547                 if (!(a_new->flags & DEVADDR_EXTERNAL) && (!keep || replace)) {
 548                         if (system_add_address(dev, a_new))
 549                                 a_new->failed = true;
 550
 551                         if (!keep) {
 552                                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
 553                                         v6 = true;
 554
 555                                 unsigned int table = (v6) ? iface->ip6table : iface->ip4table;
 556
 557                                 if (table) {
 558                                         set_ip_source_policy(true, v6, IPRULE_PRIORITY_ADDR, &a_new->addr,
 559                                                         (v6) ? 128 : 32, table, NULL, NULL);
 560                                         set_ip_source_policy(true, v6, IPRULE_PRIORITY_NW, &a_new->addr,
 561                                                         a_new->mask, table, NULL, NULL);
 562                                 }
 563                         }
 564
 565                         if ((a_new->flags & DEVADDR_OFFLINK) || iface->metric)
 566                                 interface_handle_subnet_route(iface, a_new, true);
 567                 }
 568         }
 569 }
 570
 571 static bool
 572 enable_route(struct interface_ip_settings *ip, struct device_route *route)
 573 {
 574         if (ip->no_defaultroute && !route->mask)
 575                 return false;
 576
 577         return ip->enabled;
 578 }
 579
 580 static void
 581 interface_update_proto_route(struct vlist_tree *tree,
 582                              struct vlist_node *node_new,
 583                              struct vlist_node *node_old)
 584 {
 585         struct interface_ip_settings *ip;
 586         struct interface *iface;
 587         struct device *dev;
 588         struct device_route *route_old, *route_new;
 589         bool keep = false;
 590
 591         ip = container_of(tree, struct interface_ip_settings, route);
 592         iface = ip->iface;
 593         dev = iface->l3_dev.dev;
 594
 595         if (!node_new || !node_old)
 596                 iface->updated |= IUF_ROUTE;
 597
 598         route_old = container_of(node_old, struct device_route, node);
 599         route_new = container_of(node_new, struct device_route, node);
 600
 601         if (node_old && node_new)
 602                 keep = !memcmp(&route_old->nexthop, &route_new->nexthop, sizeof(route_old->nexthop)) &&
 603                         (route_old->mtu == route_new->mtu) && (route_old->type == route_new->type) &&
 604                         !route_old->failed;
 605
 606         if (node_old) {
 607                 if (!(route_old->flags & DEVADDR_EXTERNAL) && route_old->enabled && !keep)
 608                         system_del_route(dev, route_old);
 609
 610                 free(route_old);
 611         }
 612
 613         if (node_new) {
 614                 bool _enabled = enable_route(ip, route_new);
 615
 616                 if (!(route_new->flags & DEVADDR_EXTERNAL) && !keep && _enabled)
 617                         if (system_add_route(dev, route_new))
 618                                 route_new->failed = true;
 619
 620                 route_new->iface = iface;
 621                 route_new->enabled = _enabled;
 622         }
 623 }
 624
 625 static void
 626 interface_update_host_route(struct vlist_tree *tree,
 627                              struct vlist_node *node_new,
 628                              struct vlist_node *node_old)
 629 {
 630         struct interface *iface;
 631         struct device *dev;
 632         struct device_route *route_old, *route_new;
 633
 634         iface = container_of(tree, struct interface, host_routes);
 635         dev = iface->l3_dev.dev;
 636
 637         route_old = container_of(node_old, struct device_route, node);
 638         route_new = container_of(node_new, struct device_route, node);
 639
 640         if (node_old) {
 641                 system_del_route(dev, route_old);
 642                 free(route_old);
 643         }
 644
 645         if (node_new) {
 646                 if (system_add_route(dev, route_new))
 647                         route_new->failed = true;
 648         }
 649 }
 650
 651
 652 static void
 653 interface_set_prefix_address(struct device_prefix_assignment *assignment,
 654                 const struct device_prefix *prefix, struct interface *iface, bool add)
 655 {
 656         const struct interface *uplink = prefix->iface;
 657         if (!iface->l3_dev.dev)
 658                 return;
 659
 660         struct device *l3_downlink = iface->l3_dev.dev;
 661
 662         struct device_addr addr;
 663         memset(&addr, 0, sizeof(addr));
 664         addr.addr.in6 = prefix->addr;
 665         addr.addr.in6.s6_addr32[1] |= htonl(assignment->assigned);
 666         addr.addr.in6.s6_addr[15] += 1;
 667         addr.mask = assignment->length;
 668         addr.flags = DEVADDR_INET6;
 669         addr.preferred_until = prefix->preferred_until;
 670         addr.valid_until = prefix->valid_until;
 671
 672         if (!add && assignment->enabled) {
 673                 time_t now = system_get_rtime();
 674                 addr.preferred_until = now;
 675                 if (!addr.valid_until || addr.valid_until - now > 7200)
 676                         addr.valid_until = now + 7200;
 677                 system_del_address(l3_downlink, &addr); // Work around dangling prefix routes
 678                 system_add_address(l3_downlink, &addr);
 679                 if (prefix->iface) {
 680                         if (prefix->iface->ip6table)
 681                                 set_ip_source_policy(false, true, IPRULE_PRIORITY_NW, &addr.addr,
 682                                                 addr.mask, prefix->iface->ip6table, iface, NULL);
 683
 684                         set_ip_source_policy(false, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 685                                                         addr.mask, 0, iface, "unreachable");
 686                 }
 687
 688                 assignment->enabled = false;
 689         } else if (add && (iface->state == IFS_UP || iface->state == IFS_SETUP) &&
 690                         !system_add_address(l3_downlink, &addr)) {
 691                 if (prefix->iface && !assignment->enabled) {
 692                         set_ip_source_policy(true, true, IPRULE_PRIORITY_REJECT, &addr.addr,
 693                                         addr.mask, 0, iface, "unreachable");
 694
 695                         if (prefix->iface->ip6table)
 696                                 set_ip_source_policy(true, true, IPRULE_PRIORITY_NW, &addr.addr,
 697                                                 addr.mask, prefix->iface->ip6table, iface, NULL);
 698                 }
 699                 if (uplink && uplink->l3_dev.dev) {
 700                         int mtu = system_update_ipv6_mtu(
 701                                         uplink->l3_dev.dev, 0);
 702                         if (mtu > 0)
 703                                 system_update_ipv6_mtu(l3_downlink, mtu);
 704                 }
 705                 assignment->enabled = true;
 706         }
 707 }
 708
 709 static bool interface_prefix_assign(struct list_head *list,
 710                 struct device_prefix_assignment *assign)
 711 {
 712         int32_t current = 0, asize = (1 << (64 - assign->length)) - 1;
 713         struct device_prefix_assignment *c;
 714         list_for_each_entry(c, list, head) {
 715                 if (assign->assigned != -1) {
 716                         if (assign->assigned >= current && assign->assigned + asize < c->assigned) {
 717                                 list_add_tail(&assign->head, &c->head);
 718                                 return true;
 719                         }
 720                 } else if (assign->assigned == -1) {
 721                         current = (current + asize) & (~asize);
 722                         if (current + asize < c->assigned) {
 723                                 assign->assigned = current;
 724                                 list_add_tail(&assign->head, &c->head);
 725                                 return true;
 726                         }
 727                 }
 728                 current = (c->assigned + (1 << (64 - c->length)));
 729         }
 730         return false;
 731 }
 732
 733 static void interface_update_prefix_assignments(struct device_prefix *prefix, bool setup)
 734 {
 735         struct device_prefix_assignment *c;
 736         struct interface *iface;
 737
 738         // Delete all assignments
 739         while (!list_empty(&prefix->assignments)) {
 740                 c = list_first_entry(&prefix->assignments,
 741                                 struct device_prefix_assignment, head);
 742                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 743                         interface_set_prefix_address(c, prefix, iface, false);
 744                 list_del(&c->head);
 745                 free(c);
 746         }
 747
 748         if (!setup)
 749                 return;
 750
 751         // End-of-assignment sentinel
 752         c = malloc(sizeof(*c) + 1);
 753         c->assigned = 1 << (64 - prefix->length);
 754         c->length = 64;
 755         c->name[0] = 0;
 756         list_add(&c->head, &prefix->assignments);
 757
 758         // Excluded prefix
 759         if (prefix->excl_length > 0) {
 760                 const char name[] = "!excluded";
 761                 c = malloc(sizeof(*c) + sizeof(name));
 762                 c->assigned = ntohl(prefix->excl_addr.s6_addr32[1]) &
 763                                 ((1 << (64 - prefix->length)) - 1);
 764                 c->length = prefix->excl_length;
 765                 memcpy(c->name, name, sizeof(name));
 766                 list_add(&c->head, &prefix->assignments);
 767         }
 768
 769         bool assigned_any = false;
 770         struct list_head assign_later = LIST_HEAD_INIT(assign_later);
 771         vlist_for_each_element(&interfaces, iface, node) {
 772                 if (iface->assignment_length < 48 ||
 773                                 iface->assignment_length > 64)
 774                         continue;
 775
 776                 // Test whether there is a matching class
 777                 if (!list_empty(&iface->assignment_classes)) {
 778                         bool found = false;
 779
 780                         struct interface_assignment_class *c;
 781                         list_for_each_entry(c, &iface->assignment_classes, head) {
 782                                 if (!strcmp(c->name, prefix->pclass)) {
 783                                         found = true;
 784                                         break;
 785                                 }
 786                         }
 787
 788                         if (!found)
 789                                 continue;
 790                 }
 791
 792                 size_t namelen = strlen(iface->name) + 1;
 793                 c = malloc(sizeof(*c) + namelen);
 794                 c->length = iface->assignment_length;
 795                 c->assigned = iface->assignment_hint;
 796                 c->enabled = false;
 797                 memcpy(c->name, iface->name, namelen);
 798
 799                 // First process all custom assignments, put all others in later-list
 800                 if (c->assigned == -1 || !interface_prefix_assign(&prefix->assignments, c)) {
 801                         if (c->assigned != -1) {
 802                                 c->assigned = -1;
 803                                 netifd_log_message(L_WARNING, "Failed to assign requested subprefix "
 804                                                 "of size %hhu for %s, trying other\n", c->length, c->name);
 805                         }
 806
 807                         struct list_head *next = &assign_later;
 808                         struct device_prefix_assignment *n;
 809                         list_for_each_entry(n, &assign_later, head) {
 810                                 if (n->length < c->length) {
 811                                         next = &n->head;
 812                                         break;
 813                                 }
 814                         }
 815                         list_add_tail(&c->head, next);
 816                 }
 817
 818                 if (c->assigned != -1)
 819                         assigned_any = true;
 820         }
 821
 822         // Then try to assign all other + failed custom assignments
 823         while (!list_empty(&assign_later)) {
 824                 c = list_first_entry(&assign_later, struct device_prefix_assignment, head);
 825                 list_del(&c->head);
 826
 827                 bool assigned = false;
 828                 do {
 829                         assigned = interface_prefix_assign(&prefix->assignments, c);
 830                 } while (!assigned && ++c->length <= 64);
 831
 832                 if (!assigned) {
 833                         netifd_log_message(L_WARNING, "Failed to assign subprefix "
 834                                         "of size %hhu for %s\n", c->length, c->name);
 835                         free(c);
 836                 } else {
 837                         assigned_any = true;
 838                 }
 839         }
 840
 841         list_for_each_entry(c, &prefix->assignments, head)
 842                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 843                         interface_set_prefix_address(c, prefix, iface, true);
 844
 845         if (!assigned_any)
 846                 netifd_log_message(L_WARNING, "You have delegated IPv6-prefixes but haven't assigned them "
 847                                 "to any interface. Did you forget to set option ip6assign on your lan-interfaces?");
 848 }
 849
 850
 851 void interface_refresh_assignments(bool hint)
 852 {
 853         static bool refresh = false;
 854         if (!hint && refresh) {
 855                 struct device_prefix *p;
 856                 list_for_each_entry(p, &prefixes, head)
 857                         interface_update_prefix_assignments(p, true);
 858         }
 859         refresh = hint;
 860 }
 861
 862
 863 static void
 864 interface_update_prefix(struct vlist_tree *tree,
 865                              struct vlist_node *node_new,
 866                              struct vlist_node *node_old)
 867 {
 868         struct device_prefix *prefix_old, *prefix_new;
 869         prefix_old = container_of(node_old, struct device_prefix, node);
 870         prefix_new = container_of(node_new, struct device_prefix, node);
 871
 872         struct interface_ip_settings *ip = container_of(tree, struct interface_ip_settings, prefix);
 873         if (tree && (!node_new || !node_old))
 874                 ip->iface->updated |= IUF_PREFIX;
 875
 876         struct device_route route;
 877         memset(&route, 0, sizeof(route));
 878         route.flags = DEVADDR_INET6;
 879         route.metric = INT32_MAX;
 880         route.mask = (node_new) ? prefix_new->length : prefix_old->length;
 881         route.addr.in6 = (node_new) ? prefix_new->addr : prefix_old->addr;
 882
 883
 884         struct device_prefix_assignment *c;
 885         struct interface *iface;
 886
 887         if (node_old && node_new) {
 888                 // Move assignments and refresh addresses to update valid times
 889                 list_splice(&prefix_old->assignments, &prefix_new->assignments);
 890
 891                 list_for_each_entry(c, &prefix_new->assignments, head)
 892                         if ((iface = vlist_find(&interfaces, c->name, iface, node)))
 893                                 interface_set_prefix_address(c, prefix_new, iface, true);
 894         } else if (node_new) {
 895                 // Set null-route to avoid routing loops
 896                 system_add_route(NULL, &route);
 897
 898                 if (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation)
 899                         interface_update_prefix_assignments(prefix_new, true);
 900         } else if (node_old) {
 901                 // Remove null-route
 902                 interface_update_prefix_assignments(prefix_old, false);
 903                 system_del_route(NULL, &route);
 904         }
 905
 906         if (node_old) {
 907                 if (prefix_old->head.next)
 908                         list_del(&prefix_old->head);
 909                 free(prefix_old);
 910         }
 911
 912         if (node_new && (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation))
 913                 list_add(&prefix_new->head, &prefixes);
 914
 915 }
 916
 917 struct device_prefix*
 918 interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
 919                 uint8_t length, time_t valid_until, time_t preferred_until,
 920                 struct in6_addr *excl_addr, uint8_t excl_length, const char *pclass)
 921 {
 922         if (!pclass)
 923                 pclass = (iface) ? iface->name : "local";
 924
 925         struct device_prefix *prefix = calloc(1, sizeof(*prefix) + strlen(pclass) + 1);
 926         prefix->length = length;
 927         prefix->addr = *addr;
 928         prefix->preferred_until = preferred_until;
 929         prefix->valid_until = valid_until;
 930         prefix->iface = iface;
 931         INIT_LIST_HEAD(&prefix->assignments);
 932
 933         if (excl_addr) {
 934                 prefix->excl_addr = *excl_addr;
 935                 prefix->excl_length = excl_length;
 936         }
 937
 938         strcpy(prefix->pclass, pclass);
 939
 940         if (iface)
 941                 vlist_add(&iface->proto_ip.prefix, &prefix->node, &prefix->addr);
 942         else
 943                 interface_update_prefix(NULL, &prefix->node, NULL);
 944
 945         return prefix;
 946 }
 947
 948 void
 949 interface_ip_set_ula_prefix(const char *prefix)
 950 {
 951         char buf[INET6_ADDRSTRLEN + 4] = {0}, *saveptr;
 952         if (prefix)
 953                 strncpy(buf, prefix, sizeof(buf) - 1);
 954         char *prefixaddr = strtok_r(buf, "/", &saveptr);
 955
 956         struct in6_addr addr;
 957         if (!prefixaddr || inet_pton(AF_INET6, prefixaddr, &addr) < 1) {
 958                 if (ula_prefix) {
 959                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
 960                         ula_prefix = NULL;
 961                 }
 962                 return;
 963         }
 964
 965         int length;
 966         char *prefixlen = strtok_r(NULL, ",", &saveptr);
 967         if (!prefixlen || (length = atoi(prefixlen)) < 1 || length > 64)
 968                 return;
 969
 970         if (!ula_prefix || !IN6_ARE_ADDR_EQUAL(&addr, &ula_prefix->addr) ||
 971                         ula_prefix->length != length) {
 972                 if (ula_prefix)
 973                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
 974
 975                 ula_prefix = interface_ip_add_device_prefix(NULL, &addr, length,
 976                                 0, 0, NULL, 0, NULL);
 977         }
 978 }
 979
 980 void
 981 interface_add_dns_server(struct interface_ip_settings *ip, const char *str)
 982 {
 983         struct dns_server *s;
 984
 985         s = calloc(1, sizeof(*s));
 986         if (!s)
 987                 return;
 988
 989         s->af = AF_INET;
 990         if (inet_pton(s->af, str, &s->addr.in))
 991                 goto add;
 992
 993         s->af = AF_INET6;
 994         if (inet_pton(s->af, str, &s->addr.in))
 995                 goto add;
 996
 997         free(s);
 998         return;
 999
1000 add:
1001         D(INTERFACE, "Add IPv%c DNS server: %s\n",
1002           s->af == AF_INET6 ? '6' : '4', str);
1003         vlist_simple_add(&ip->dns_servers, &s->node);
1004 }
1005
1006 void
1007 interface_add_dns_server_list(struct interface_ip_settings *ip, struct blob_attr *list)
1008 {
1009         struct blob_attr *cur;
1010         int rem;
1011
1012         blobmsg_for_each_attr(cur, list, rem) {
1013                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1014                         continue;
1015
1016                 if (!blobmsg_check_attr(cur, NULL))
1017                         continue;
1018
1019                 interface_add_dns_server(ip, blobmsg_data(cur));
1020         }
1021 }
1022
1023 static void
1024 interface_add_dns_search_domain(struct interface_ip_settings *ip, const char *str)
1025 {
1026         struct dns_search_domain *s;
1027         int len = strlen(str);
1028
1029         s = calloc(1, sizeof(*s) + len + 1);
1030         if (!s)
1031                 return;
1032
1033         D(INTERFACE, "Add DNS search domain: %s\n", str);
1034         memcpy(s->name, str, len);
1035         vlist_simple_add(&ip->dns_search, &s->node);
1036 }
1037
1038 void
1039 interface_add_dns_search_list(struct interface_ip_settings *ip, struct blob_attr *list)
1040 {
1041         struct blob_attr *cur;
1042         int rem;
1043
1044         blobmsg_for_each_attr(cur, list, rem) {
1045                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1046                         continue;
1047
1048                 if (!blobmsg_check_attr(cur, NULL))
1049                         continue;
1050
1051                 interface_add_dns_search_domain(ip, blobmsg_data(cur));
1052         }
1053 }
1054
1055 static void
1056 write_resolv_conf_entries(FILE *f, struct interface_ip_settings *ip, const char *dev)
1057 {
1058         struct dns_server *s;
1059         struct dns_search_domain *d;
1060         const char *str;
1061         char buf[INET6_ADDRSTRLEN];
1062
1063         vlist_simple_for_each_element(&ip->dns_servers, s, node) {
1064                 str = inet_ntop(s->af, &s->addr, buf, sizeof(buf));
1065                 if (!str)
1066                         continue;
1067
1068                 if (s->af == AF_INET6 && IN6_IS_ADDR_LINKLOCAL(&s->addr.in6))
1069                         fprintf(f, "nameserver %s%%%s\n", str, dev);
1070                 else
1071                         fprintf(f, "nameserver %s\n", str);
1072         }
1073
1074         vlist_simple_for_each_element(&ip->dns_search, d, node) {
1075                 fprintf(f, "search %s\n", d->name);
1076         }
1077 }
1078
1079 void
1080 interface_write_resolv_conf(void)
1081 {
1082         struct interface *iface;
1083         char *path = alloca(strlen(resolv_conf) + 5);
1084         FILE *f;
1085         uint32_t crcold, crcnew;
1086
1087         sprintf(path, "%s.tmp", resolv_conf);
1088         unlink(path);
1089         f = fopen(path, "w+");
1090         if (!f) {
1091                 D(INTERFACE, "Failed to open %s for writing\n", path);
1092                 return;
1093         }
1094
1095         vlist_for_each_element(&interfaces, iface, node) {
1096                 if (iface->state != IFS_UP)
1097                         continue;
1098
1099                 if (vlist_simple_empty(&iface->proto_ip.dns_search) &&
1100                     vlist_simple_empty(&iface->proto_ip.dns_servers) &&
1101                         vlist_simple_empty(&iface->config_ip.dns_search) &&
1102                     vlist_simple_empty(&iface->config_ip.dns_servers))
1103                         continue;
1104
1105                 fprintf(f, "# Interface %s\n", iface->name);
1106                 write_resolv_conf_entries(f, &iface->config_ip, iface->ifname);
1107                 if (!iface->proto_ip.no_dns)
1108                         write_resolv_conf_entries(f, &iface->proto_ip, iface->ifname);
1109         }
1110         fflush(f);
1111         rewind(f);
1112         crcnew = crc32_file(f);
1113         fclose(f);
1114
1115         crcold = crcnew + 1;
1116         f = fopen(resolv_conf, "r");
1117         if (f) {
1118                 crcold = crc32_file(f);
1119                 fclose(f);
1120         }
1121
1122         if (crcold == crcnew) {
1123                 unlink(path);
1124         } else if (rename(path, resolv_conf) < 0) {
1125                 D(INTERFACE, "Failed to replace %s\n", resolv_conf);
1126                 unlink(path);
1127         }
1128 }
1129
1130 void interface_ip_set_enabled(struct interface_ip_settings *ip, bool enabled)
1131 {
1132         struct device_addr *addr;
1133         struct device_route *route;
1134         struct device *dev;
1135
1136         ip->enabled = enabled;
1137         dev = ip->iface->l3_dev.dev;
1138         if (!dev)
1139                 return;
1140
1141         vlist_for_each_element(&ip->addr, addr, node) {
1142                 if (addr->enabled == enabled)
1143                         continue;
1144
1145                 if (enabled)
1146                         system_add_address(dev, addr);
1147                 else
1148                         system_del_address(dev, addr);
1149                 addr->enabled = enabled;
1150         }
1151
1152         vlist_for_each_element(&ip->route, route, node) {
1153                 bool _enabled = enabled;
1154
1155                 if (!enable_route(ip, route))
1156                         _enabled = false;
1157
1158                 if (route->enabled == _enabled)
1159                         continue;
1160
1161                 if (_enabled) {
1162                         if (!(route->flags & DEVROUTE_METRIC))
1163                                 route->metric = ip->iface->metric;
1164
1165                         if (system_add_route(dev, route))
1166                                 route->failed = true;
1167                 } else
1168                         system_del_route(dev, route);
1169                 route->enabled = _enabled;
1170         }
1171
1172         struct device_prefix *c;
1173         struct device_prefix_assignment *a;
1174         list_for_each_entry(c, &prefixes, head)
1175                 list_for_each_entry(a, &c->assignments, head)
1176                         if (!strcmp(a->name, ip->iface->name))
1177                                 interface_set_prefix_address(a, c, ip->iface, enabled);
1178
1179         if (ip->iface && ip->iface->l3_dev.dev) {
1180                 set_ip_lo_policy(enabled, true, ip->iface);
1181                 set_ip_lo_policy(enabled, false, ip->iface);
1182
1183                 set_ip_source_policy(enabled, true, IPRULE_PRIORITY_REJECT + ip->iface->l3_dev.dev->ifindex,
1184                         NULL, 0, 0, ip->iface, "failed_policy");
1185         }
1186 }
1187
1188 void
1189 interface_ip_update_start(struct interface_ip_settings *ip)
1190 {
1191         if (ip != &ip->iface->config_ip) {
1192                 vlist_simple_update(&ip->dns_servers);
1193                 vlist_simple_update(&ip->dns_search);
1194         }
1195         vlist_update(&ip->route);
1196         vlist_update(&ip->addr);
1197         vlist_update(&ip->prefix);
1198 }
1199
1200 void
1201 interface_ip_update_complete(struct interface_ip_settings *ip)
1202 {
1203         vlist_simple_flush(&ip->dns_servers);
1204         vlist_simple_flush(&ip->dns_search);
1205         vlist_flush(&ip->route);
1206         vlist_flush(&ip->addr);
1207         vlist_flush(&ip->prefix);
1208         interface_write_resolv_conf();
1209 }
1210
1211 void
1212 interface_ip_flush(struct interface_ip_settings *ip)
1213 {
1214         if (ip == &ip->iface->proto_ip)
1215                 vlist_flush_all(&ip->iface->host_routes);
1216         vlist_simple_flush_all(&ip->dns_servers);
1217         vlist_simple_flush_all(&ip->dns_search);
1218         vlist_flush_all(&ip->route);
1219         vlist_flush_all(&ip->addr);
1220         vlist_flush_all(&ip->prefix);
1221 }
1222
1223 static void
1224 __interface_ip_init(struct interface_ip_settings *ip, struct interface *iface)
1225 {
1226         ip->iface = iface;
1227         ip->enabled = true;
1228         vlist_simple_init(&ip->dns_search, struct dns_search_domain, node);
1229         vlist_simple_init(&ip->dns_servers, struct dns_server, node);
1230         vlist_init(&ip->route, route_cmp, interface_update_proto_route);
1231         vlist_init(&ip->addr, addr_cmp, interface_update_proto_addr);
1232         vlist_init(&ip->prefix, prefix_cmp, interface_update_prefix);
1233 }
1234
1235 void
1236 interface_ip_init(struct interface *iface)
1237 {
1238         __interface_ip_init(&iface->proto_ip, iface);
1239         __interface_ip_init(&iface->config_ip, iface);
1240         vlist_init(&iface->host_routes, route_cmp, interface_update_host_route);
1241 }
1242
1243 static void
1244 interface_ip_valid_until_handler(struct uloop_timeout *t)
1245 {
1246         time_t now = system_get_rtime();
1247         struct interface *iface;
1248         vlist_for_each_element(&interfaces, iface, node) {
1249                 if (iface->state != IFS_UP)
1250                         continue;
1251
1252                 struct device_addr *addr, *addrp;
1253                 struct device_route *route, *routep;
1254                 struct device_prefix *pref, *prefp;
1255
1256                 vlist_for_each_element_safe(&iface->proto_ip.addr, addr, node, addrp)
1257                         if (addr->valid_until && addr->valid_until < now)
1258                                 vlist_delete(&iface->proto_ip.addr, &addr->node);
1259
1260                 vlist_for_each_element_safe(&iface->proto_ip.route, route, node, routep)
1261                         if (route->valid_until && route->valid_until < now)
1262                                 vlist_delete(&iface->proto_ip.route, &route->node);
1263
1264                 vlist_for_each_element_safe(&iface->proto_ip.prefix, pref, node, prefp)
1265                         if (pref->valid_until && pref->valid_until < now)
1266                                 vlist_delete(&iface->proto_ip.prefix, &pref->node);
1267
1268         }
1269
1270         uloop_timeout_set(t, 1000);
1271 }
1272
1273 static void __init
1274 interface_ip_init_worker(void)
1275 {
1276         valid_until_timeout.cb = interface_ip_valid_until_handler;
1277         uloop_timeout_set(&valid_until_timeout, 1000);
1278 }