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