interface-ip: unify handling of interface metric/table for routes, fixes handling...
[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 #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 static void
278 interface_set_route_info(struct interface *iface, struct device_route *route)
279 {
280 bool v6 = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET6);
281
282 if (!iface)
283 return;
284
285 if (!(route->flags & DEVROUTE_METRIC))
286 route->metric = iface->metric;
287
288 if (!(route->flags & DEVROUTE_TABLE)) {
289 route->table = (v6) ? iface->ip6table : iface->ip4table;
290 if (route->table)
291 route->flags |= DEVROUTE_SRCTABLE;
292 }
293 }
294
295 void
296 interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
297 {
298 struct interface_ip_settings *ip;
299 struct blob_attr *tb[__ROUTE_MAX], *cur;
300 struct device_route *route;
301 int af = v6 ? AF_INET6 : AF_INET;
302
303 blobmsg_parse(route_attr, __ROUTE_MAX, tb, blobmsg_data(attr), blobmsg_data_len(attr));
304
305 if (!iface) {
306 if ((cur = tb[ROUTE_INTERFACE]) == NULL)
307 return;
308
309 iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
310 if (!iface)
311 return;
312
313 ip = &iface->config_ip;
314 } else {
315 ip = &iface->proto_ip;
316 }
317
318 route = calloc(1, sizeof(*route));
319 if (!route)
320 return;
321
322 route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
323 route->mask = v6 ? 128 : 32;
324 if ((cur = tb[ROUTE_MASK]) != NULL) {
325 route->mask = parse_netmask_string(blobmsg_data(cur), v6);
326 if (route->mask > (v6 ? 128 : 32))
327 goto error;
328 }
329
330 if ((cur = tb[ROUTE_TARGET]) != NULL) {
331 if (!parse_ip_and_netmask(af, blobmsg_data(cur), &route->addr, &route->mask)) {
332 DPRINTF("Failed to parse route target: %s\n", (char *) blobmsg_data(cur));
333 goto error;
334 }
335 }
336
337 if ((cur = tb[ROUTE_GATEWAY]) != NULL) {
338 if (!inet_pton(af, blobmsg_data(cur), &route->nexthop)) {
339 DPRINTF("Failed to parse route gateway: %s\n", (char *) blobmsg_data(cur));
340 goto error;
341 }
342 }
343
344 if ((cur = tb[ROUTE_METRIC]) != NULL) {
345 route->metric = blobmsg_get_u32(cur);
346 route->flags |= DEVROUTE_METRIC;
347 }
348
349 if ((cur = tb[ROUTE_MTU]) != NULL) {
350 route->mtu = blobmsg_get_u32(cur);
351 route->flags |= DEVROUTE_MTU;
352 }
353
354 // Use source-based routing
355 if ((cur = tb[ROUTE_SOURCE]) != NULL) {
356 char *saveptr, *source = alloca(blobmsg_data_len(cur));
357 memcpy(source, blobmsg_data(cur), blobmsg_data_len(cur));
358
359 const char *addr = strtok_r(source, "/", &saveptr);
360 const char *mask = strtok_r(NULL, "/", &saveptr);
361
362 if (!addr || inet_pton(af, addr, &route->source) < 1) {
363 DPRINTF("Failed to parse route source: %s\n", addr);
364 goto error;
365 }
366
367 route->sourcemask = (mask) ? atoi(mask) : ((af == AF_INET6) ? 128 : 32);
368 }
369
370 if ((cur = tb[ROUTE_ONLINK]) != NULL && blobmsg_get_bool(cur))
371 route->flags |= DEVROUTE_ONLINK;
372
373 if ((cur = tb[ROUTE_TABLE]) != NULL) {
374 if (!system_resolve_rt_table(blobmsg_data(cur), &route->table)) {
375 DPRINTF("Failed to resolve routing table: %s\n", (char *) blobmsg_data(cur));
376 goto error;
377 }
378
379 /* only set the table flag if not using the main (default) table */
380 if (system_is_default_rt_table(route->table))
381 route->table = 0;
382
383 if (route->table)
384 route->flags |= DEVROUTE_TABLE;
385 }
386
387 if ((cur = tb[ROUTE_VALID]) != NULL) {
388 int64_t valid = blobmsg_get_u32(cur);
389 int64_t valid_until = valid + (int64_t)system_get_rtime();
390 if (valid_until <= LONG_MAX && valid != 0xffffffffLL) // Catch overflow
391 route->valid_until = valid_until;
392 }
393
394 if ((cur = tb[ROUTE_TYPE]) != NULL) {
395 if (!system_resolve_rt_type(blobmsg_data(cur), &route->type)) {
396 DPRINTF("Failed to resolve routing type: %s\n", (char *) blobmsg_data(cur));
397 goto error;
398 }
399 route->flags |= DEVROUTE_TYPE;
400 }
401
402 interface_set_route_info(iface, route);
403 vlist_add(&ip->route, &route->node, route);
404 return;
405
406 error:
407 free(route);
408 }
409
410 static int
411 addr_cmp(const void *k1, const void *k2, void *ptr)
412 {
413 return memcmp(k1, k2, sizeof(struct device_addr) -
414 offsetof(struct device_addr, flags));
415 }
416
417 static int
418 route_cmp(const void *k1, const void *k2, void *ptr)
419 {
420 const struct device_route *r1 = k1, *r2 = k2;
421
422 if (r1->mask != r2->mask)
423 return r2->mask - r1->mask;
424
425 if (r1->metric != r2->metric)
426 return r1->metric - r2->metric;
427
428 if (r1->flags != r2->flags)
429 return r2->flags - r1->flags;
430
431 if (r1->sourcemask != r2->sourcemask)
432 return r1->sourcemask - r2->sourcemask;
433
434 if (r1->table != r2->table)
435 return r1->table - r2->table;
436
437 int maskcmp = memcmp(&r1->source, &r2->source, sizeof(r1->source));
438 if (maskcmp)
439 return maskcmp;
440
441 return memcmp(&r1->addr, &r2->addr, sizeof(r1->addr));
442 }
443
444 static int
445 prefix_cmp(const void *k1, const void *k2, void *ptr)
446 {
447 return memcmp(k1, k2, offsetof(struct device_prefix, pclass) -
448 offsetof(struct device_prefix, addr));
449 }
450
451 static void
452 interface_handle_subnet_route(struct interface *iface, struct device_addr *addr, bool add)
453 {
454 struct device *dev = iface->l3_dev.dev;
455 struct device_route *r = &addr->subnet;
456
457 if (addr->flags & DEVADDR_OFFLINK)
458 return;
459
460 if (!add) {
461 if (!addr->subnet.iface)
462 return;
463
464 system_del_route(dev, r);
465 memset(r, 0, sizeof(*r));
466 return;
467 }
468
469 r->iface = iface;
470 r->flags = addr->flags;
471 r->mask = addr->mask;
472 memcpy(&r->addr, &addr->addr, sizeof(r->addr));
473 clear_if_addr(&r->addr, r->mask);
474
475 r->flags |= DEVADDR_KERNEL;
476 system_del_route(dev, r);
477
478 r->flags &= ~DEVADDR_KERNEL;
479 interface_set_route_info(iface, r);
480
481 system_add_route(dev, r);
482 }
483
484 static void
485 interface_update_proto_addr(struct vlist_tree *tree,
486 struct vlist_node *node_new,
487 struct vlist_node *node_old)
488 {
489 struct interface_ip_settings *ip;
490 struct interface *iface;
491 struct device *dev;
492 struct device_addr *a_new = NULL, *a_old = NULL;
493 bool replace = false;
494 bool keep = false;
495 bool v6 = false;
496
497 ip = container_of(tree, struct interface_ip_settings, addr);
498 iface = ip->iface;
499 dev = iface->l3_dev.dev;
500
501 if (!node_new || !node_old)
502 iface->updated |= IUF_ADDRESS;
503
504 if (node_new) {
505 a_new = container_of(node_new, struct device_addr, node);
506
507 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
508 !a_new->broadcast) {
509
510 /* /31 and /32 addressing need 255.255.255.255
511 * as broadcast address. */
512 if (a_new->mask >= 31) {
513 a_new->broadcast = (uint32_t) ~0;
514 } else {
515 uint32_t mask = ~0;
516 uint32_t *a = (uint32_t *) &a_new->addr;
517
518 mask >>= a_new->mask;
519 a_new->broadcast = *a | htonl(mask);
520 }
521 }
522 }
523
524 if (node_old)
525 a_old = container_of(node_old, struct device_addr, node);
526
527 if (a_new && a_old) {
528 keep = true;
529
530 if (a_old->flags != a_new->flags || a_old->failed)
531 keep = false;
532
533 if (a_old->valid_until != a_new->valid_until ||
534 a_old->preferred_until != a_new->preferred_until)
535 replace = true;
536
537 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
538 a_new->broadcast != a_old->broadcast)
539 keep = false;
540 }
541
542 if (node_old) {
543 if (a_old->enabled && !keep) {
544 if ((a_old->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
545 v6 = true;
546
547 //This is needed for source routing to work correctly. If a device
548 //has two connections to a network using the same subnet, adding
549 //only the network-rule will cause packets to be routed through the
550 //first matching network (source IP matches both masks).
551 if (a_old->policy_table)
552 set_ip_source_policy(false, v6, IPRULE_PRIORITY_ADDR, &a_old->addr,
553 (v6) ? 128 : 32, a_old->policy_table, NULL, NULL);
554
555 if (!(a_old->flags & DEVADDR_EXTERNAL)) {
556 interface_handle_subnet_route(iface, a_old, false);
557 system_del_address(dev, a_old);
558 }
559 }
560 free(a_old->pclass);
561 free(a_old);
562 }
563
564 if (node_new) {
565 a_new->enabled = true;
566
567 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
568 v6 = true;
569
570 a_new->policy_table = (v6) ? iface->ip6table : iface->ip4table;
571
572 if (!keep || replace) {
573 if (!(a_new->flags & DEVADDR_EXTERNAL)) {
574 if (system_add_address(dev, a_new))
575 a_new->failed = true;
576
577 if (iface->metric || a_new->policy_table)
578 interface_handle_subnet_route(iface, a_new, true);
579 }
580
581 if (!keep) {
582 if (a_new->policy_table)
583 set_ip_source_policy(true, v6, IPRULE_PRIORITY_ADDR, &a_new->addr,
584 (v6) ? 128 : 32, a_new->policy_table, NULL, NULL);
585 }
586 }
587 }
588 }
589
590 static bool
591 enable_route(struct interface_ip_settings *ip, struct device_route *route)
592 {
593 if (ip->no_defaultroute && !route->mask)
594 return false;
595
596 return ip->enabled;
597 }
598
599 static void
600 interface_update_proto_route(struct vlist_tree *tree,
601 struct vlist_node *node_new,
602 struct vlist_node *node_old)
603 {
604 struct interface_ip_settings *ip;
605 struct interface *iface;
606 struct device *dev;
607 struct device_route *route_old, *route_new;
608 bool keep = false;
609
610 ip = container_of(tree, struct interface_ip_settings, route);
611 iface = ip->iface;
612 dev = iface->l3_dev.dev;
613
614 if (!node_new || !node_old)
615 iface->updated |= IUF_ROUTE;
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 && node_new)
621 keep = !memcmp(&route_old->nexthop, &route_new->nexthop, sizeof(route_old->nexthop)) &&
622 (route_old->mtu == route_new->mtu) && (route_old->type == route_new->type) &&
623 !route_old->failed;
624
625 if (node_old) {
626 if (!(route_old->flags & DEVADDR_EXTERNAL) && route_old->enabled && !keep)
627 system_del_route(dev, route_old);
628
629 free(route_old);
630 }
631
632 if (node_new) {
633 bool _enabled = enable_route(ip, route_new);
634
635 if (!(route_new->flags & DEVADDR_EXTERNAL) && !keep && _enabled)
636 if (system_add_route(dev, route_new))
637 route_new->failed = true;
638
639 route_new->iface = iface;
640 route_new->enabled = _enabled;
641 }
642 }
643
644 static void
645 interface_update_host_route(struct vlist_tree *tree,
646 struct vlist_node *node_new,
647 struct vlist_node *node_old)
648 {
649 struct interface *iface;
650 struct device *dev;
651 struct device_route *route_old, *route_new;
652
653 iface = container_of(tree, struct interface, host_routes);
654 dev = iface->l3_dev.dev;
655
656 route_old = container_of(node_old, struct device_route, node);
657 route_new = container_of(node_new, struct device_route, node);
658
659 if (node_old) {
660 system_del_route(dev, route_old);
661 free(route_old);
662 }
663
664 if (node_new) {
665 if (system_add_route(dev, route_new))
666 route_new->failed = true;
667 }
668 }
669
670 static void
671 random_ifaceid(struct in6_addr *addr)
672 {
673 static bool initialized = false;
674 struct timeval t;
675
676 if (!initialized) {
677 long int seed = 0;
678 gettimeofday(&t, NULL);
679 seed = t.tv_sec ^ t.tv_usec ^ getpid();
680 srand48(seed);
681 initialized = true;
682 }
683 addr->s6_addr32[2] = (uint32_t)mrand48();
684 addr->s6_addr32[3] = (uint32_t)mrand48();
685 }
686
687 static void
688 eui64_ifaceid(struct interface *iface, struct in6_addr *addr)
689 {
690 /* get mac address */
691 uint8_t *macaddr = iface->l3_dev.dev->settings.macaddr;
692 uint8_t *ifaceid = addr->s6_addr + 8;
693 memcpy(ifaceid,macaddr,3);
694 memcpy(ifaceid + 5,macaddr + 3, 3);
695 ifaceid[3] = 0xff;
696 ifaceid[4] = 0xfe;
697 ifaceid[0] ^= 0x02;
698 }
699
700 static void
701 generate_ifaceid(struct interface *iface, struct in6_addr *addr)
702 {
703 /* generate new iface id */
704 switch (iface->assignment_iface_id_selection) {
705 case IFID_FIXED:
706 /* fixed */
707 /* copy host part from assignment_fixed_iface_id */
708 memcpy(addr->s6_addr + 8, iface->assignment_fixed_iface_id.s6_addr + 8, 8);
709 break;
710 case IFID_RANDOM:
711 /* randomize last 64 bits */
712 random_ifaceid(addr);
713 break;
714 case IFID_EUI64:
715 /* eui64 */
716 eui64_ifaceid(iface, addr);
717 break;
718 }
719 }
720
721 static void
722 interface_set_prefix_address(struct device_prefix_assignment *assignment,
723 const struct device_prefix *prefix, struct interface *iface, bool add)
724 {
725 const struct interface *uplink = prefix->iface;
726 if (!iface->l3_dev.dev)
727 return;
728
729 struct device *l3_downlink = iface->l3_dev.dev;
730
731 struct device_addr addr;
732 struct device_route route;
733 memset(&addr, 0, sizeof(addr));
734 memset(&route, 0, sizeof(route));
735
736 if (IN6_IS_ADDR_UNSPECIFIED(&assignment->addr)) {
737 addr.addr.in6 = prefix->addr;
738 addr.addr.in6.s6_addr32[1] |= htonl(assignment->assigned);
739 generate_ifaceid(iface, &addr.addr.in6);
740 assignment->addr = addr.addr.in6;
741 }
742 else
743 addr.addr.in6 = assignment->addr;
744
745 addr.mask = assignment->length;
746 addr.flags = DEVADDR_INET6 | DEVADDR_OFFLINK;
747 addr.preferred_until = prefix->preferred_until;
748 addr.valid_until = prefix->valid_until;
749
750 route.flags = DEVADDR_INET6;
751 route.mask = addr.mask < 64 ? 64 : addr.mask;
752 route.addr = addr.addr;
753 clear_if_addr(&route.addr, route.mask);
754 interface_set_route_info(iface, &route);
755
756 if (!add && assignment->enabled) {
757 time_t now = system_get_rtime();
758 addr.preferred_until = now;
759 if (!addr.valid_until || addr.valid_until - now > 7200)
760 addr.valid_until = now + 7200;
761
762 if (prefix->iface) {
763 if (prefix->iface->ip6table)
764 set_ip_source_policy(false, true, IPRULE_PRIORITY_NW, &addr.addr,
765 addr.mask, prefix->iface->ip6table, iface, NULL);
766
767 set_ip_source_policy(false, true, IPRULE_PRIORITY_REJECT, &addr.addr,
768 addr.mask, 0, iface, "unreachable");
769 }
770
771 system_del_route(l3_downlink, &route);
772 system_add_address(l3_downlink, &addr);
773
774 assignment->enabled = false;
775 } else if (add && (iface->state == IFS_UP || iface->state == IFS_SETUP) &&
776 !system_add_address(l3_downlink, &addr)) {
777
778 if (prefix->iface && !assignment->enabled) {
779 set_ip_source_policy(true, true, IPRULE_PRIORITY_REJECT, &addr.addr,
780 addr.mask, 0, iface, "unreachable");
781
782 if (prefix->iface->ip6table)
783 set_ip_source_policy(true, true, IPRULE_PRIORITY_NW, &addr.addr,
784 addr.mask, prefix->iface->ip6table, iface, NULL);
785 }
786
787 route.metric = iface->metric;
788 system_add_route(l3_downlink, &route);
789
790 if (uplink && uplink->l3_dev.dev && !(l3_downlink->settings.flags & DEV_OPT_MTU6)) {
791 int mtu = system_update_ipv6_mtu(uplink->l3_dev.dev, 0);
792 int mtu_old = system_update_ipv6_mtu(l3_downlink, 0);
793
794 if (mtu > 0 && mtu_old > mtu)
795 system_update_ipv6_mtu(l3_downlink, mtu);
796 }
797
798 assignment->enabled = true;
799 }
800 }
801
802 static bool interface_prefix_assign(struct list_head *list,
803 struct device_prefix_assignment *assign)
804 {
805 int32_t current = 0, asize = (1 << (64 - assign->length)) - 1;
806 struct device_prefix_assignment *c;
807 list_for_each_entry(c, list, head) {
808 if (assign->assigned != -1) {
809 if (assign->assigned >= current && assign->assigned + asize < c->assigned) {
810 list_add_tail(&assign->head, &c->head);
811 return true;
812 }
813 } else if (assign->assigned == -1) {
814 current = (current + asize) & (~asize);
815 if (current + asize < c->assigned) {
816 assign->assigned = current;
817 list_add_tail(&assign->head, &c->head);
818 return true;
819 }
820 }
821 current = (c->assigned + (1 << (64 - c->length)));
822 }
823 return false;
824 }
825
826 static void interface_update_prefix_assignments(struct device_prefix *prefix, bool setup)
827 {
828 struct device_prefix_assignment *c;
829 struct interface *iface;
830
831 // Delete all assignments
832 while (!list_empty(&prefix->assignments)) {
833 c = list_first_entry(&prefix->assignments,
834 struct device_prefix_assignment, head);
835 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
836 interface_set_prefix_address(c, prefix, iface, false);
837 list_del(&c->head);
838 free(c);
839 }
840
841 if (!setup)
842 return;
843
844 // End-of-assignment sentinel
845 c = malloc(sizeof(*c) + 1);
846 c->assigned = 1 << (64 - prefix->length);
847 c->length = 64;
848 c->name[0] = 0;
849 c->addr = in6addr_any;
850 list_add(&c->head, &prefix->assignments);
851
852 // Excluded prefix
853 if (prefix->excl_length > 0) {
854 const char name[] = "!excluded";
855 c = malloc(sizeof(*c) + sizeof(name));
856 c->assigned = ntohl(prefix->excl_addr.s6_addr32[1]) &
857 ((1 << (64 - prefix->length)) - 1);
858 c->length = prefix->excl_length;
859 c->addr = in6addr_any;
860 memcpy(c->name, name, sizeof(name));
861 list_add(&c->head, &prefix->assignments);
862 }
863
864 bool assigned_any = false;
865 struct list_head assign_later = LIST_HEAD_INIT(assign_later);
866 vlist_for_each_element(&interfaces, iface, node) {
867 if (iface->assignment_length < 48 ||
868 iface->assignment_length > 64)
869 continue;
870
871 // Test whether there is a matching class
872 if (!list_empty(&iface->assignment_classes)) {
873 bool found = false;
874
875 struct interface_assignment_class *c;
876 list_for_each_entry(c, &iface->assignment_classes, head) {
877 if (!strcmp(c->name, prefix->pclass)) {
878 found = true;
879 break;
880 }
881 }
882
883 if (!found)
884 continue;
885 }
886
887 size_t namelen = strlen(iface->name) + 1;
888 c = malloc(sizeof(*c) + namelen);
889 c->length = iface->assignment_length;
890 c->assigned = iface->assignment_hint;
891 c->addr = in6addr_any;
892 c->enabled = false;
893 memcpy(c->name, iface->name, namelen);
894
895 // First process all custom assignments, put all others in later-list
896 if (c->assigned == -1 || !interface_prefix_assign(&prefix->assignments, c)) {
897 if (c->assigned != -1) {
898 c->assigned = -1;
899 netifd_log_message(L_WARNING, "Failed to assign requested subprefix "
900 "of size %hhu for %s, trying other\n", c->length, c->name);
901 }
902
903 struct list_head *next = &assign_later;
904 struct device_prefix_assignment *n;
905 list_for_each_entry(n, &assign_later, head) {
906 if (n->length < c->length) {
907 next = &n->head;
908 break;
909 }
910 }
911 list_add_tail(&c->head, next);
912 }
913
914 if (c->assigned != -1)
915 assigned_any = true;
916 }
917
918 // Then try to assign all other + failed custom assignments
919 while (!list_empty(&assign_later)) {
920 c = list_first_entry(&assign_later, struct device_prefix_assignment, head);
921 list_del(&c->head);
922
923 bool assigned = false;
924 do {
925 assigned = interface_prefix_assign(&prefix->assignments, c);
926 } while (!assigned && ++c->length <= 64);
927
928 if (!assigned) {
929 netifd_log_message(L_WARNING, "Failed to assign subprefix "
930 "of size %hhu for %s\n", c->length, c->name);
931 free(c);
932 } else {
933 assigned_any = true;
934 }
935 }
936
937 list_for_each_entry(c, &prefix->assignments, head)
938 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
939 interface_set_prefix_address(c, prefix, iface, true);
940
941 if (!assigned_any)
942 netifd_log_message(L_WARNING, "You have delegated IPv6-prefixes but haven't assigned them "
943 "to any interface. Did you forget to set option ip6assign on your lan-interfaces?");
944 }
945
946
947 void interface_refresh_assignments(bool hint)
948 {
949 static bool refresh = false;
950 if (!hint && refresh) {
951 struct device_prefix *p;
952 list_for_each_entry(p, &prefixes, head)
953 interface_update_prefix_assignments(p, true);
954 }
955 refresh = hint;
956 }
957
958
959 static void
960 interface_update_prefix(struct vlist_tree *tree,
961 struct vlist_node *node_new,
962 struct vlist_node *node_old)
963 {
964 struct device_prefix *prefix_old, *prefix_new;
965 prefix_old = container_of(node_old, struct device_prefix, node);
966 prefix_new = container_of(node_new, struct device_prefix, node);
967
968 struct interface_ip_settings *ip = container_of(tree, struct interface_ip_settings, prefix);
969 if (tree && (!node_new || !node_old))
970 ip->iface->updated |= IUF_PREFIX;
971
972 struct device_route route;
973 memset(&route, 0, sizeof(route));
974 route.flags = DEVADDR_INET6;
975 route.metric = INT32_MAX;
976 route.mask = (node_new) ? prefix_new->length : prefix_old->length;
977 route.addr.in6 = (node_new) ? prefix_new->addr : prefix_old->addr;
978
979
980 struct device_prefix_assignment *c;
981 struct interface *iface;
982
983 if (node_old && node_new) {
984 // Move assignments and refresh addresses to update valid times
985 list_splice(&prefix_old->assignments, &prefix_new->assignments);
986
987 list_for_each_entry(c, &prefix_new->assignments, head)
988 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
989 interface_set_prefix_address(c, prefix_new, iface, true);
990 } else if (node_new) {
991 // Set null-route to avoid routing loops
992 system_add_route(NULL, &route);
993
994 if (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation)
995 interface_update_prefix_assignments(prefix_new, true);
996 } else if (node_old) {
997 // Remove null-route
998 interface_update_prefix_assignments(prefix_old, false);
999 system_del_route(NULL, &route);
1000 }
1001
1002 if (node_old) {
1003 if (prefix_old->head.next)
1004 list_del(&prefix_old->head);
1005 free(prefix_old);
1006 }
1007
1008 if (node_new && (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation))
1009 list_add(&prefix_new->head, &prefixes);
1010
1011 }
1012
1013 struct device_prefix*
1014 interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
1015 uint8_t length, time_t valid_until, time_t preferred_until,
1016 struct in6_addr *excl_addr, uint8_t excl_length, const char *pclass)
1017 {
1018 if (!pclass)
1019 pclass = (iface) ? iface->name : "local";
1020
1021 struct device_prefix *prefix = calloc(1, sizeof(*prefix) + strlen(pclass) + 1);
1022 prefix->length = length;
1023 prefix->addr = *addr;
1024 prefix->preferred_until = preferred_until;
1025 prefix->valid_until = valid_until;
1026 prefix->iface = iface;
1027 INIT_LIST_HEAD(&prefix->assignments);
1028
1029 if (excl_addr) {
1030 prefix->excl_addr = *excl_addr;
1031 prefix->excl_length = excl_length;
1032 }
1033
1034 strcpy(prefix->pclass, pclass);
1035
1036 if (iface)
1037 vlist_add(&iface->proto_ip.prefix, &prefix->node, &prefix->addr);
1038 else
1039 interface_update_prefix(NULL, &prefix->node, NULL);
1040
1041 return prefix;
1042 }
1043
1044 void
1045 interface_ip_set_ula_prefix(const char *prefix)
1046 {
1047 char buf[INET6_ADDRSTRLEN + 4] = {0}, *saveptr;
1048 if (prefix)
1049 strncpy(buf, prefix, sizeof(buf) - 1);
1050 char *prefixaddr = strtok_r(buf, "/", &saveptr);
1051
1052 struct in6_addr addr;
1053 if (!prefixaddr || inet_pton(AF_INET6, prefixaddr, &addr) < 1) {
1054 if (ula_prefix) {
1055 interface_update_prefix(NULL, NULL, &ula_prefix->node);
1056 ula_prefix = NULL;
1057 }
1058 return;
1059 }
1060
1061 int length;
1062 char *prefixlen = strtok_r(NULL, ",", &saveptr);
1063 if (!prefixlen || (length = atoi(prefixlen)) < 1 || length > 64)
1064 return;
1065
1066 if (!ula_prefix || !IN6_ARE_ADDR_EQUAL(&addr, &ula_prefix->addr) ||
1067 ula_prefix->length != length) {
1068 if (ula_prefix)
1069 interface_update_prefix(NULL, NULL, &ula_prefix->node);
1070
1071 ula_prefix = interface_ip_add_device_prefix(NULL, &addr, length,
1072 0, 0, NULL, 0, NULL);
1073 }
1074 }
1075
1076 void
1077 interface_add_dns_server(struct interface_ip_settings *ip, const char *str)
1078 {
1079 struct dns_server *s;
1080
1081 s = calloc(1, sizeof(*s));
1082 if (!s)
1083 return;
1084
1085 s->af = AF_INET;
1086 if (inet_pton(s->af, str, &s->addr.in))
1087 goto add;
1088
1089 s->af = AF_INET6;
1090 if (inet_pton(s->af, str, &s->addr.in))
1091 goto add;
1092
1093 free(s);
1094 return;
1095
1096 add:
1097 D(INTERFACE, "Add IPv%c DNS server: %s\n",
1098 s->af == AF_INET6 ? '6' : '4', str);
1099 vlist_simple_add(&ip->dns_servers, &s->node);
1100 }
1101
1102 void
1103 interface_add_dns_server_list(struct interface_ip_settings *ip, struct blob_attr *list)
1104 {
1105 struct blob_attr *cur;
1106 int rem;
1107
1108 blobmsg_for_each_attr(cur, list, rem) {
1109 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1110 continue;
1111
1112 if (!blobmsg_check_attr(cur, NULL))
1113 continue;
1114
1115 interface_add_dns_server(ip, blobmsg_data(cur));
1116 }
1117 }
1118
1119 static void
1120 interface_add_dns_search_domain(struct interface_ip_settings *ip, const char *str)
1121 {
1122 struct dns_search_domain *s;
1123 int len = strlen(str);
1124
1125 s = calloc(1, sizeof(*s) + len + 1);
1126 if (!s)
1127 return;
1128
1129 D(INTERFACE, "Add DNS search domain: %s\n", str);
1130 memcpy(s->name, str, len);
1131 vlist_simple_add(&ip->dns_search, &s->node);
1132 }
1133
1134 void
1135 interface_add_dns_search_list(struct interface_ip_settings *ip, struct blob_attr *list)
1136 {
1137 struct blob_attr *cur;
1138 int rem;
1139
1140 blobmsg_for_each_attr(cur, list, rem) {
1141 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1142 continue;
1143
1144 if (!blobmsg_check_attr(cur, NULL))
1145 continue;
1146
1147 interface_add_dns_search_domain(ip, blobmsg_data(cur));
1148 }
1149 }
1150
1151 static void
1152 write_resolv_conf_entries(FILE *f, struct interface_ip_settings *ip, const char *dev)
1153 {
1154 struct dns_server *s;
1155 struct dns_search_domain *d;
1156 const char *str;
1157 char buf[INET6_ADDRSTRLEN];
1158
1159 vlist_simple_for_each_element(&ip->dns_servers, s, node) {
1160 str = inet_ntop(s->af, &s->addr, buf, sizeof(buf));
1161 if (!str)
1162 continue;
1163
1164 if (s->af == AF_INET6 && IN6_IS_ADDR_LINKLOCAL(&s->addr.in6))
1165 fprintf(f, "nameserver %s%%%s\n", str, dev);
1166 else
1167 fprintf(f, "nameserver %s\n", str);
1168 }
1169
1170 vlist_simple_for_each_element(&ip->dns_search, d, node) {
1171 fprintf(f, "search %s\n", d->name);
1172 }
1173 }
1174
1175 void
1176 interface_write_resolv_conf(void)
1177 {
1178 struct interface *iface;
1179 char *path = alloca(strlen(resolv_conf) + 5);
1180 FILE *f;
1181 uint32_t crcold, crcnew;
1182
1183 sprintf(path, "%s.tmp", resolv_conf);
1184 unlink(path);
1185 f = fopen(path, "w+");
1186 if (!f) {
1187 D(INTERFACE, "Failed to open %s for writing\n", path);
1188 return;
1189 }
1190
1191 vlist_for_each_element(&interfaces, iface, node) {
1192 if (iface->state != IFS_UP)
1193 continue;
1194
1195 if (vlist_simple_empty(&iface->proto_ip.dns_search) &&
1196 vlist_simple_empty(&iface->proto_ip.dns_servers) &&
1197 vlist_simple_empty(&iface->config_ip.dns_search) &&
1198 vlist_simple_empty(&iface->config_ip.dns_servers))
1199 continue;
1200
1201 fprintf(f, "# Interface %s\n", iface->name);
1202 write_resolv_conf_entries(f, &iface->config_ip, iface->ifname);
1203 if (!iface->proto_ip.no_dns)
1204 write_resolv_conf_entries(f, &iface->proto_ip, iface->ifname);
1205 }
1206 fflush(f);
1207 rewind(f);
1208 crcnew = crc32_file(f);
1209 fclose(f);
1210
1211 crcold = crcnew + 1;
1212 f = fopen(resolv_conf, "r");
1213 if (f) {
1214 crcold = crc32_file(f);
1215 fclose(f);
1216 }
1217
1218 if (crcold == crcnew) {
1219 unlink(path);
1220 } else if (rename(path, resolv_conf) < 0) {
1221 D(INTERFACE, "Failed to replace %s\n", resolv_conf);
1222 unlink(path);
1223 }
1224 }
1225
1226 void interface_ip_set_enabled(struct interface_ip_settings *ip, bool enabled)
1227 {
1228 struct device_addr *addr;
1229 struct device_route *route;
1230 struct device *dev;
1231 struct interface *iface;
1232
1233 ip->enabled = enabled;
1234 iface = ip->iface;
1235 dev = iface->l3_dev.dev;
1236 if (!dev)
1237 return;
1238
1239 vlist_for_each_element(&ip->addr, addr, node) {
1240 bool v6 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6) ? true : false;
1241
1242 if (addr->enabled == enabled)
1243 continue;
1244
1245 if (enabled) {
1246 system_add_address(dev, addr);
1247
1248 addr->policy_table = (v6) ? iface->ip6table : iface->ip4table;
1249 if (iface->metric || addr->policy_table)
1250 interface_handle_subnet_route(iface, addr, true);
1251
1252 if (addr->policy_table)
1253 set_ip_source_policy(true, v6, IPRULE_PRIORITY_ADDR, &addr->addr,
1254 (v6) ? 128 : 32, addr->policy_table, NULL, NULL);
1255 } else {
1256 interface_handle_subnet_route(iface, addr, false);
1257 system_del_address(dev, addr);
1258
1259 if (addr->policy_table)
1260 set_ip_source_policy(false, v6, IPRULE_PRIORITY_ADDR, &addr->addr,
1261 (v6) ? 128 : 32, addr->policy_table, NULL, NULL);
1262 }
1263 addr->enabled = enabled;
1264 }
1265
1266 vlist_for_each_element(&ip->route, route, node) {
1267 bool _enabled = enabled;
1268
1269 if (!enable_route(ip, route))
1270 _enabled = false;
1271
1272 if (route->enabled == _enabled)
1273 continue;
1274
1275 if (_enabled) {
1276 interface_set_route_info(ip->iface, route);
1277
1278 if (system_add_route(dev, route))
1279 route->failed = true;
1280 } else
1281 system_del_route(dev, route);
1282 route->enabled = _enabled;
1283 }
1284
1285 struct device_prefix *c;
1286 struct device_prefix_assignment *a;
1287 list_for_each_entry(c, &prefixes, head)
1288 list_for_each_entry(a, &c->assignments, head)
1289 if (!strcmp(a->name, ip->iface->name))
1290 interface_set_prefix_address(a, c, ip->iface, enabled);
1291
1292 if (ip->iface && ip->iface->l3_dev.dev) {
1293 set_ip_lo_policy(enabled, true, ip->iface);
1294 set_ip_lo_policy(enabled, false, ip->iface);
1295
1296 set_ip_source_policy(enabled, true, IPRULE_PRIORITY_REJECT + ip->iface->l3_dev.dev->ifindex,
1297 NULL, 0, 0, ip->iface, "failed_policy");
1298 }
1299 }
1300
1301 void
1302 interface_ip_update_start(struct interface_ip_settings *ip)
1303 {
1304 if (ip != &ip->iface->config_ip) {
1305 vlist_simple_update(&ip->dns_servers);
1306 vlist_simple_update(&ip->dns_search);
1307 }
1308 vlist_update(&ip->route);
1309 vlist_update(&ip->addr);
1310 vlist_update(&ip->prefix);
1311 }
1312
1313 void
1314 interface_ip_update_complete(struct interface_ip_settings *ip)
1315 {
1316 vlist_simple_flush(&ip->dns_servers);
1317 vlist_simple_flush(&ip->dns_search);
1318 vlist_flush(&ip->route);
1319 vlist_flush(&ip->addr);
1320 vlist_flush(&ip->prefix);
1321 interface_write_resolv_conf();
1322 }
1323
1324 void
1325 interface_ip_flush(struct interface_ip_settings *ip)
1326 {
1327 if (ip == &ip->iface->proto_ip)
1328 vlist_flush_all(&ip->iface->host_routes);
1329 vlist_simple_flush_all(&ip->dns_servers);
1330 vlist_simple_flush_all(&ip->dns_search);
1331 vlist_flush_all(&ip->route);
1332 vlist_flush_all(&ip->addr);
1333 vlist_flush_all(&ip->prefix);
1334 }
1335
1336 static void
1337 __interface_ip_init(struct interface_ip_settings *ip, struct interface *iface)
1338 {
1339 ip->iface = iface;
1340 ip->enabled = true;
1341 vlist_simple_init(&ip->dns_search, struct dns_search_domain, node);
1342 vlist_simple_init(&ip->dns_servers, struct dns_server, node);
1343 vlist_init(&ip->route, route_cmp, interface_update_proto_route);
1344 vlist_init(&ip->addr, addr_cmp, interface_update_proto_addr);
1345 vlist_init(&ip->prefix, prefix_cmp, interface_update_prefix);
1346 }
1347
1348 void
1349 interface_ip_init(struct interface *iface)
1350 {
1351 __interface_ip_init(&iface->proto_ip, iface);
1352 __interface_ip_init(&iface->config_ip, iface);
1353 vlist_init(&iface->host_routes, route_cmp, interface_update_host_route);
1354 }
1355
1356 static void
1357 interface_ip_valid_until_handler(struct uloop_timeout *t)
1358 {
1359 time_t now = system_get_rtime();
1360 struct interface *iface;
1361 vlist_for_each_element(&interfaces, iface, node) {
1362 if (iface->state != IFS_UP)
1363 continue;
1364
1365 struct device_addr *addr, *addrp;
1366 struct device_route *route, *routep;
1367 struct device_prefix *pref, *prefp;
1368
1369 vlist_for_each_element_safe(&iface->proto_ip.addr, addr, node, addrp)
1370 if (addr->valid_until && addr->valid_until < now)
1371 vlist_delete(&iface->proto_ip.addr, &addr->node);
1372
1373 vlist_for_each_element_safe(&iface->proto_ip.route, route, node, routep)
1374 if (route->valid_until && route->valid_until < now)
1375 vlist_delete(&iface->proto_ip.route, &route->node);
1376
1377 vlist_for_each_element_safe(&iface->proto_ip.prefix, pref, node, prefp)
1378 if (pref->valid_until && pref->valid_until < now)
1379 vlist_delete(&iface->proto_ip.prefix, &pref->node);
1380
1381 }
1382
1383 uloop_timeout_set(t, 1000);
1384 }
1385
1386 static void __init
1387 interface_ip_init_worker(void)
1388 {
1389 valid_until_timeout.cb = interface_ip_valid_until_handler;
1390 uloop_timeout_set(&valid_until_timeout, 1000);
1391 }