interface-ip: Set route table when enabling interface ip settings
[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 bool
137 __find_ip_addr_target(struct interface_ip_settings *ip, union if_addr *a, bool v6)
138 {
139         struct device_addr *addr;
140
141         vlist_for_each_element(&ip->addr, addr, node) {
142                 if (!addr->enabled)
143                         continue;
144
145                 if (v6 != ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
146                         continue;
147
148                 // Handle offlink addresses correctly
149                 unsigned int mask = addr->mask;
150                 if ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6 &&
151                                 (addr->flags & DEVADDR_OFFLINK))
152                         mask = 128;
153
154                 if (!match_if_addr(&addr->addr, a, mask))
155                         continue;
156
157                 return true;
158         }
159
160         return false;
161 }
162
163 static void
164 __find_ip_route_target(struct interface_ip_settings *ip, union if_addr *a,
165                        bool v6, struct device_route **res)
166 {
167         struct device_route *route;
168
169         vlist_for_each_element(&ip->route, route, node) {
170                 if (!route->enabled)
171                         continue;
172
173                 if (v6 != ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
174                         continue;
175
176                 if (!match_if_addr(&route->addr, a, route->mask))
177                         continue;
178
179                 if (route->flags & DEVROUTE_TABLE)
180                         continue;
181
182                 if (!*res || route->mask < (*res)->mask)
183                         *res = route;
184         }
185 }
186
187 static bool
188 interface_ip_find_addr_target(struct interface *iface, union if_addr *a, bool v6)
189 {
190         return __find_ip_addr_target(&iface->proto_ip, a, v6) ||
191                __find_ip_addr_target(&iface->config_ip, a, v6);
192 }
193
194 static void
195 interface_ip_find_route_target(struct interface *iface, union if_addr *a,
196                                bool v6, struct device_route **route)
197 {
198         __find_ip_route_target(&iface->proto_ip, a, v6, route);
199         __find_ip_route_target(&iface->config_ip, a, v6, route);
200 }
201
202 struct interface *
203 interface_ip_add_target_route(union if_addr *addr, bool v6, struct interface *iface)
204 {
205         struct device_route *route, *r_next = NULL;
206         bool defaultroute_target = false;
207         int addrsize = v6 ? sizeof(addr->in6) : sizeof(addr->in);
208
209         route = calloc(1, sizeof(*route));
210         if (!route)
211                 return NULL;
212
213         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
214         route->mask = v6 ? 128 : 32;
215         if (memcmp(&route->addr, addr, addrsize) == 0)
216                 defaultroute_target = true;
217         else
218                 memcpy(&route->addr, addr, addrsize);
219
220         if (iface) {
221                 /* look for locally addressable target first */
222                 if (interface_ip_find_addr_target(iface, addr, v6))
223                         goto done;
224
225                 /* do not stop at the first route, let the lookup compare
226                  * masks to find the best match */
227                 interface_ip_find_route_target(iface, addr, v6, &r_next);
228         } else {
229                 vlist_for_each_element(&interfaces, iface, node) {
230                         /* look for locally addressable target first */
231                         if (interface_ip_find_addr_target(iface, addr, v6))
232                                 goto done;
233
234                         /* do not stop at the first route, let the lookup compare
235                          * masks to find the best match */
236                         interface_ip_find_route_target(iface, addr, v6, &r_next);
237                 }
238         }
239
240         if (!r_next) {
241                 free(route);
242                 return NULL;
243         }
244
245         iface = r_next->iface;
246         memcpy(&route->nexthop, &r_next->nexthop, sizeof(route->nexthop));
247         route->mtu = r_next->mtu;
248         route->metric = r_next->metric;
249         route->table = r_next->table;
250
251 done:
252         route->iface = iface;
253         if (defaultroute_target)
254                 free(route);
255         else
256                 vlist_add(&iface->host_routes, &route->node, route);
257         return iface;
258 }
259
260 void
261 interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
262 {
263         struct interface_ip_settings *ip;
264         struct blob_attr *tb[__ROUTE_MAX], *cur;
265         struct device_route *route;
266         int af = v6 ? AF_INET6 : AF_INET;
267         bool is_proto_route = !!iface;
268
269         blobmsg_parse(route_attr, __ROUTE_MAX, tb, blobmsg_data(attr), blobmsg_data_len(attr));
270
271         if (!iface) {
272                 if ((cur = tb[ROUTE_INTERFACE]) == NULL)
273                         return;
274
275                 iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
276                 if (!iface)
277                         return;
278
279                 ip = &iface->config_ip;
280         } else {
281                 ip = &iface->proto_ip;
282         }
283
284         route = calloc(1, sizeof(*route));
285         if (!route)
286                 return;
287
288         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
289         route->mask = v6 ? 128 : 32;
290         if ((cur = tb[ROUTE_MASK]) != NULL) {
291                 route->mask = parse_netmask_string(blobmsg_data(cur), v6);
292                 if (route->mask > (v6 ? 128 : 32))
293                         goto error;
294         }
295
296         if ((cur = tb[ROUTE_TARGET]) != NULL) {
297                 if (!parse_ip_and_netmask(af, blobmsg_data(cur), &route->addr, &route->mask)) {
298                         DPRINTF("Failed to parse route target: %s\n", (char *) blobmsg_data(cur));
299                         goto error;
300                 }
301         }
302
303         if ((cur = tb[ROUTE_GATEWAY]) != NULL) {
304                 if (!inet_pton(af, blobmsg_data(cur), &route->nexthop)) {
305                         DPRINTF("Failed to parse route gateway: %s\n", (char *) blobmsg_data(cur));
306                         goto error;
307                 }
308         }
309
310         if ((cur = tb[ROUTE_METRIC]) != NULL) {
311                 route->metric = blobmsg_get_u32(cur);
312                 route->flags |= DEVROUTE_METRIC;
313         } else
314                 route->metric = iface->metric;
315
316         if ((cur = tb[ROUTE_MTU]) != NULL) {
317                 route->mtu = blobmsg_get_u32(cur);
318                 route->flags |= DEVROUTE_MTU;
319         }
320
321         // Use source-based routing
322         if ((cur = tb[ROUTE_SOURCE]) != NULL) {
323                 char *saveptr, *source = alloca(blobmsg_data_len(cur));
324                 memcpy(source, blobmsg_data(cur), blobmsg_data_len(cur));
325
326                 const char *addr = strtok_r(source, "/", &saveptr);
327                 const char *mask = strtok_r(NULL, "/", &saveptr);
328
329                 if (!addr || inet_pton(af, addr, &route->source) < 1) {
330                         DPRINTF("Failed to parse route source: %s\n", addr);
331                         goto error;
332                 }
333
334                 route->sourcemask = (mask) ? atoi(mask) : ((af == AF_INET6) ? 128 : 32);
335         }
336
337         if ((cur = tb[ROUTE_ONLINK]) != NULL && blobmsg_get_bool(cur))
338                 route->flags |= DEVROUTE_ONLINK;
339
340         if (is_proto_route) {
341                 route->table = (v6) ? iface->ip6table : iface->ip4table;
342                 route->flags |= DEVROUTE_SRCTABLE;
343         }
344
345         if ((cur = tb[ROUTE_TABLE]) != NULL) {
346                 if (!system_resolve_rt_table(blobmsg_data(cur), &route->table)) {
347                         DPRINTF("Failed to resolve routing table: %s\n", (char *) blobmsg_data(cur));
348                         goto error;
349                 }
350
351                 /* only set the table flag if not using the main (default) table */
352                 if (system_is_default_rt_table(route->table))
353                         route->table = 0;
354
355                 if (route->table)
356                         route->flags |= DEVROUTE_TABLE;
357         }
358
359         if ((cur = tb[ROUTE_VALID]) != NULL) {
360                 int64_t valid = blobmsg_get_u32(cur);
361                 int64_t valid_until = valid + (int64_t)system_get_rtime();
362                 if (valid_until <= LONG_MAX && valid != 0xffffffffLL) // Catch overflow
363                         route->valid_until = valid_until;
364         }
365
366         if ((cur = tb[ROUTE_TYPE]) != NULL) {
367                 if (!system_resolve_rt_type(blobmsg_data(cur), &route->type)) {
368                         DPRINTF("Failed to resolve routing type: %s\n", (char *) blobmsg_data(cur));
369                         goto error;
370                 }
371                 route->flags |= DEVROUTE_TYPE;
372         }
373
374         vlist_add(&ip->route, &route->node, route);
375         return;
376
377 error:
378         free(route);
379 }
380
381 static int
382 addr_cmp(const void *k1, const void *k2, void *ptr)
383 {
384         return memcmp(k1, k2, sizeof(struct device_addr) -
385                       offsetof(struct device_addr, flags));
386 }
387
388 static int
389 route_cmp(const void *k1, const void *k2, void *ptr)
390 {
391         const struct device_route *r1 = k1, *r2 = k2;
392
393         if (r1->mask != r2->mask)
394                 return r2->mask - r1->mask;
395
396         if (r1->metric != r2->metric)
397                 return r1->metric - r2->metric;
398
399         if (r1->flags != r2->flags)
400                 return r2->flags - r1->flags;
401
402         if (r1->sourcemask != r2->sourcemask)
403                 return r1->sourcemask - r2->sourcemask;
404
405         if (r1->table != r2->table)
406                 return r1->table - r2->table;
407
408         int maskcmp = memcmp(&r1->source, &r2->source, sizeof(r1->source));
409         if (maskcmp)
410                 return maskcmp;
411
412         return memcmp(&r1->addr, &r2->addr, sizeof(r1->addr));
413 }
414
415 static int
416 prefix_cmp(const void *k1, const void *k2, void *ptr)
417 {
418         return memcmp(k1, k2, offsetof(struct device_prefix, pclass) -
419                         offsetof(struct device_prefix, addr));
420 }
421
422 static void
423 interface_handle_subnet_route(struct interface *iface, struct device_addr *addr, bool add)
424 {
425         struct device *dev = iface->l3_dev.dev;
426         struct device_route route;
427
428         if (addr->flags & DEVADDR_OFFLINK)
429                 return;
430
431         memset(&route, 0, sizeof(route));
432         route.iface = iface;
433         route.flags = addr->flags;
434         route.mask = addr->mask;
435         memcpy(&route.addr, &addr->addr, sizeof(route.addr));
436         clear_if_addr(&route.addr, route.mask);
437
438         if (add) {
439                 route.flags |= DEVADDR_KERNEL;
440                 system_del_route(dev, &route);
441
442                 route.flags &= ~DEVADDR_KERNEL;
443                 route.metric = iface->metric;
444                 system_add_route(dev, &route);
445         } else {
446                 system_del_route(dev, &route);
447         }
448 }
449
450 static void
451 interface_update_proto_addr(struct vlist_tree *tree,
452                             struct vlist_node *node_new,
453                             struct vlist_node *node_old)
454 {
455         struct interface_ip_settings *ip;
456         struct interface *iface;
457         struct device *dev;
458         struct device_addr *a_new = NULL, *a_old = NULL;
459         bool replace = false;
460         bool keep = false;
461         bool v6 = false;
462
463         ip = container_of(tree, struct interface_ip_settings, addr);
464         iface = ip->iface;
465         dev = iface->l3_dev.dev;
466
467         if (!node_new || !node_old)
468                 iface->updated |= IUF_ADDRESS;
469
470         if (node_new) {
471                 a_new = container_of(node_new, struct device_addr, node);
472
473                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
474                     !a_new->broadcast) {
475
476                         uint32_t mask = ~0;
477                         uint32_t *a = (uint32_t *) &a_new->addr;
478
479                         mask >>= a_new->mask;
480                         a_new->broadcast = *a | htonl(mask);
481                 }
482         }
483
484         if (node_old)
485                 a_old = container_of(node_old, struct device_addr, node);
486
487         if (a_new && a_old) {
488                 keep = true;
489
490                 if (a_old->flags != a_new->flags || a_old->failed)
491                         keep = false;
492
493                 if (a_old->valid_until != a_new->valid_until ||
494                                 a_old->preferred_until != a_new->preferred_until)
495                         replace = true;
496
497                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
498                     a_new->broadcast != a_old->broadcast)
499                         keep = false;
500         }
501
502         if (node_old) {
503                 if (a_old->enabled && !keep) {
504                         if ((a_old->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
505                                 v6 = true;
506
507                         unsigned int table = (v6) ? iface->ip6table : iface->ip4table;
508
509                         //This is needed for source routing to work correctly. If a device
510                         //has two connections to a network using the same subnet, adding
511                         //only the network-rule will cause packets to be routed through the
512                         //first matching network (source IP matches both masks).
513                         if (table) {
514                                 set_ip_source_policy(false, v6, IPRULE_PRIORITY_ADDR, &a_old->addr,
515                                                 (v6) ? 128 : 32, table, NULL, NULL);
516
517                                 if (a_old->mask != ((v6) ? 128 : 32))
518                                         set_ip_source_policy(false, v6, IPRULE_PRIORITY_NW, &a_old->addr,
519                                                         a_old->mask, table, NULL, NULL);
520                         }
521
522                         if (!(a_old->flags & DEVADDR_EXTERNAL)) {
523                                 interface_handle_subnet_route(iface, a_old, false);
524                                 system_del_address(dev, a_old);
525                         }
526                 }
527                 free(a_old->pclass);
528                 free(a_old);
529         }
530
531         if (node_new) {
532                 a_new->enabled = true;
533                 if (!keep || replace) {
534                         if (!(a_new->flags & DEVADDR_EXTERNAL)) {
535                                 if (system_add_address(dev, a_new))
536                                         a_new->failed = true;
537
538                                 if (iface->metric)
539                                         interface_handle_subnet_route(iface, a_new, true);
540                         }
541
542                         if (!keep) {
543                                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
544                                         v6 = true;
545
546                                 unsigned int table = (v6) ? iface->ip6table : iface->ip4table;
547
548                                 if (table) {
549                                         set_ip_source_policy(true, v6, IPRULE_PRIORITY_ADDR, &a_new->addr,
550                                                         (v6) ? 128 : 32, table, NULL, NULL);
551
552                                         if (a_new->mask != ((v6) ? 128 : 32))
553                                                 set_ip_source_policy(true, v6, IPRULE_PRIORITY_NW, &a_new->addr,
554                                                                 a_new->mask, table, NULL, NULL);
555                                 }
556                         }
557                 }
558         }
559 }
560
561 static bool
562 enable_route(struct interface_ip_settings *ip, struct device_route *route)
563 {
564         if (ip->no_defaultroute && !route->mask)
565                 return false;
566
567         return ip->enabled;
568 }
569
570 static void
571 interface_update_proto_route(struct vlist_tree *tree,
572                              struct vlist_node *node_new,
573                              struct vlist_node *node_old)
574 {
575         struct interface_ip_settings *ip;
576         struct interface *iface;
577         struct device *dev;
578         struct device_route *route_old, *route_new;
579         bool keep = false;
580
581         ip = container_of(tree, struct interface_ip_settings, route);
582         iface = ip->iface;
583         dev = iface->l3_dev.dev;
584
585         if (!node_new || !node_old)
586                 iface->updated |= IUF_ROUTE;
587
588         route_old = container_of(node_old, struct device_route, node);
589         route_new = container_of(node_new, struct device_route, node);
590
591         if (node_old && node_new)
592                 keep = !memcmp(&route_old->nexthop, &route_new->nexthop, sizeof(route_old->nexthop)) &&
593                         (route_old->mtu == route_new->mtu) && (route_old->type == route_new->type) &&
594                         (route_old->valid_until == route_new->valid_until) && !route_old->failed;
595
596         if (node_old) {
597                 if (!(route_old->flags & DEVADDR_EXTERNAL) && route_old->enabled && !keep)
598                         system_del_route(dev, route_old);
599
600                 free(route_old);
601         }
602
603         if (node_new) {
604                 bool _enabled = enable_route(ip, route_new);
605
606                 if (!(route_new->flags & DEVADDR_EXTERNAL) && !keep && _enabled)
607                         if (system_add_route(dev, route_new))
608                                 route_new->failed = true;
609
610                 route_new->iface = iface;
611                 route_new->enabled = _enabled;
612         }
613 }
614
615 static void
616 interface_update_host_route(struct vlist_tree *tree,
617                              struct vlist_node *node_new,
618                              struct vlist_node *node_old)
619 {
620         struct interface *iface;
621         struct device *dev;
622         struct device_route *route_old, *route_new;
623
624         iface = container_of(tree, struct interface, host_routes);
625         dev = iface->l3_dev.dev;
626
627         route_old = container_of(node_old, struct device_route, node);
628         route_new = container_of(node_new, struct device_route, node);
629
630         if (node_old) {
631                 system_del_route(dev, route_old);
632                 free(route_old);
633         }
634
635         if (node_new) {
636                 if (system_add_route(dev, route_new))
637                         route_new->failed = true;
638         }
639 }
640
641 static void
642 random_ifaceid(struct in6_addr *addr)
643 {
644         static bool initialized = false;
645         struct timeval t;
646
647         if (!initialized) {
648                 long int seed = 0;
649                 gettimeofday(&t, NULL);
650                 seed = t.tv_sec ^ t.tv_usec ^ getpid();
651                 srand48(seed);
652                 initialized = true;
653         }
654         addr->s6_addr32[2] = (uint32_t)mrand48();
655         addr->s6_addr32[3] = (uint32_t)mrand48();
656 }
657
658 static void
659 eui64_ifaceid(struct interface *iface, struct in6_addr *addr)
660 {
661         /* get mac address */
662         uint8_t *macaddr = iface->l3_dev.dev->settings.macaddr;
663         uint8_t *ifaceid = addr->s6_addr + 8;
664         memcpy(ifaceid,macaddr,3);
665         memcpy(ifaceid + 5,macaddr + 3, 3);
666         ifaceid[3] = 0xff;
667         ifaceid[4] = 0xfe;
668         ifaceid[0] ^= 0x02;
669 }
670
671 static void
672 generate_ifaceid(struct interface *iface, struct in6_addr *addr)
673 {
674         /* generate new iface id */
675         switch (iface->assignment_iface_id_selection) {
676         case IFID_FIXED:
677                 /* fixed */
678                 /* copy host part from assignment_fixed_iface_id */
679                 memcpy(addr->s6_addr + 8, iface->assignment_fixed_iface_id.s6_addr + 8, 8);
680                 break;
681         case IFID_RANDOM:
682                 /* randomize last 64 bits */
683                 random_ifaceid(addr);
684                 break;
685         case IFID_EUI64:
686                 /* eui64 */
687                 eui64_ifaceid(iface, addr);
688                 break;
689         }
690 }
691
692 static void
693 interface_set_prefix_address(struct device_prefix_assignment *assignment,
694                 const struct device_prefix *prefix, struct interface *iface, bool add)
695 {
696         const struct interface *uplink = prefix->iface;
697         if (!iface->l3_dev.dev)
698                 return;
699
700         struct device *l3_downlink = iface->l3_dev.dev;
701
702         struct device_addr addr;
703         struct device_route route;
704         memset(&addr, 0, sizeof(addr));
705         memset(&route, 0, sizeof(route));
706
707         if (IN6_IS_ADDR_UNSPECIFIED(&assignment->addr)) {
708                 addr.addr.in6 = prefix->addr;
709                 addr.addr.in6.s6_addr32[1] |= htonl(assignment->assigned);
710                 generate_ifaceid(iface, &addr.addr.in6);
711                 assignment->addr = addr.addr.in6;
712         }
713         else
714                 addr.addr.in6 = assignment->addr;
715
716         addr.mask = assignment->length;
717         addr.flags = DEVADDR_INET6 | DEVADDR_OFFLINK;
718         addr.preferred_until = prefix->preferred_until;
719         addr.valid_until = prefix->valid_until;
720
721         route.flags = DEVADDR_INET6;
722         route.mask = addr.mask < 64 ? 64 : addr.mask;
723         route.addr = addr.addr;
724         clear_if_addr(&route.addr, route.mask);
725
726         if (!add && assignment->enabled) {
727                 time_t now = system_get_rtime();
728                 addr.preferred_until = now;
729                 if (!addr.valid_until || addr.valid_until - now > 7200)
730                         addr.valid_until = now + 7200;
731
732                 if (prefix->iface) {
733                         if (prefix->iface->ip6table)
734                                 set_ip_source_policy(false, true, IPRULE_PRIORITY_NW, &addr.addr,
735                                                 addr.mask, prefix->iface->ip6table, iface, NULL);
736
737                         set_ip_source_policy(false, true, IPRULE_PRIORITY_REJECT, &addr.addr,
738                                                         addr.mask, 0, iface, "unreachable");
739                 }
740
741                 system_del_route(l3_downlink, &route);
742                 system_add_address(l3_downlink, &addr);
743
744                 assignment->enabled = false;
745         } else if (add && (iface->state == IFS_UP || iface->state == IFS_SETUP) &&
746                         !system_add_address(l3_downlink, &addr)) {
747
748                 if (prefix->iface && !assignment->enabled) {
749                         set_ip_source_policy(true, true, IPRULE_PRIORITY_REJECT, &addr.addr,
750                                         addr.mask, 0, iface, "unreachable");
751
752                         if (prefix->iface->ip6table)
753                                 set_ip_source_policy(true, true, IPRULE_PRIORITY_NW, &addr.addr,
754                                                 addr.mask, prefix->iface->ip6table, iface, NULL);
755                 }
756
757                 route.metric = iface->metric;
758                 system_add_route(l3_downlink, &route);
759
760                 if (uplink && uplink->l3_dev.dev && !(l3_downlink->settings.flags & DEV_OPT_MTU6)) {
761                         int mtu = system_update_ipv6_mtu(uplink->l3_dev.dev, 0);
762                         int mtu_old = system_update_ipv6_mtu(l3_downlink, 0);
763
764                         if (mtu > 0 && mtu_old > mtu)
765                                 system_update_ipv6_mtu(l3_downlink, mtu);
766                 }
767
768                 assignment->enabled = true;
769         }
770 }
771
772 static bool interface_prefix_assign(struct list_head *list,
773                 struct device_prefix_assignment *assign)
774 {
775         int32_t current = 0, asize = (1 << (64 - assign->length)) - 1;
776         struct device_prefix_assignment *c;
777         list_for_each_entry(c, list, head) {
778                 if (assign->assigned != -1) {
779                         if (assign->assigned >= current && assign->assigned + asize < c->assigned) {
780                                 list_add_tail(&assign->head, &c->head);
781                                 return true;
782                         }
783                 } else if (assign->assigned == -1) {
784                         current = (current + asize) & (~asize);
785                         if (current + asize < c->assigned) {
786                                 assign->assigned = current;
787                                 list_add_tail(&assign->head, &c->head);
788                                 return true;
789                         }
790                 }
791                 current = (c->assigned + (1 << (64 - c->length)));
792         }
793         return false;
794 }
795
796 static void interface_update_prefix_assignments(struct device_prefix *prefix, bool setup)
797 {
798         struct device_prefix_assignment *c;
799         struct interface *iface;
800
801         // Delete all assignments
802         while (!list_empty(&prefix->assignments)) {
803                 c = list_first_entry(&prefix->assignments,
804                                 struct device_prefix_assignment, head);
805                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
806                         interface_set_prefix_address(c, prefix, iface, false);
807                 list_del(&c->head);
808                 free(c);
809         }
810
811         if (!setup)
812                 return;
813
814         // End-of-assignment sentinel
815         c = malloc(sizeof(*c) + 1);
816         c->assigned = 1 << (64 - prefix->length);
817         c->length = 64;
818         c->name[0] = 0;
819         c->addr = in6addr_any;
820         list_add(&c->head, &prefix->assignments);
821
822         // Excluded prefix
823         if (prefix->excl_length > 0) {
824                 const char name[] = "!excluded";
825                 c = malloc(sizeof(*c) + sizeof(name));
826                 c->assigned = ntohl(prefix->excl_addr.s6_addr32[1]) &
827                                 ((1 << (64 - prefix->length)) - 1);
828                 c->length = prefix->excl_length;
829                 c->addr = in6addr_any;
830                 memcpy(c->name, name, sizeof(name));
831                 list_add(&c->head, &prefix->assignments);
832         }
833
834         bool assigned_any = false;
835         struct list_head assign_later = LIST_HEAD_INIT(assign_later);
836         vlist_for_each_element(&interfaces, iface, node) {
837                 if (iface->assignment_length < 48 ||
838                                 iface->assignment_length > 64)
839                         continue;
840
841                 // Test whether there is a matching class
842                 if (!list_empty(&iface->assignment_classes)) {
843                         bool found = false;
844
845                         struct interface_assignment_class *c;
846                         list_for_each_entry(c, &iface->assignment_classes, head) {
847                                 if (!strcmp(c->name, prefix->pclass)) {
848                                         found = true;
849                                         break;
850                                 }
851                         }
852
853                         if (!found)
854                                 continue;
855                 }
856
857                 size_t namelen = strlen(iface->name) + 1;
858                 c = malloc(sizeof(*c) + namelen);
859                 c->length = iface->assignment_length;
860                 c->assigned = iface->assignment_hint;
861                 c->addr = in6addr_any;
862                 c->enabled = false;
863                 memcpy(c->name, iface->name, namelen);
864
865                 // First process all custom assignments, put all others in later-list
866                 if (c->assigned == -1 || !interface_prefix_assign(&prefix->assignments, c)) {
867                         if (c->assigned != -1) {
868                                 c->assigned = -1;
869                                 netifd_log_message(L_WARNING, "Failed to assign requested subprefix "
870                                                 "of size %hhu for %s, trying other\n", c->length, c->name);
871                         }
872
873                         struct list_head *next = &assign_later;
874                         struct device_prefix_assignment *n;
875                         list_for_each_entry(n, &assign_later, head) {
876                                 if (n->length < c->length) {
877                                         next = &n->head;
878                                         break;
879                                 }
880                         }
881                         list_add_tail(&c->head, next);
882                 }
883
884                 if (c->assigned != -1)
885                         assigned_any = true;
886         }
887
888         // Then try to assign all other + failed custom assignments
889         while (!list_empty(&assign_later)) {
890                 c = list_first_entry(&assign_later, struct device_prefix_assignment, head);
891                 list_del(&c->head);
892
893                 bool assigned = false;
894                 do {
895                         assigned = interface_prefix_assign(&prefix->assignments, c);
896                 } while (!assigned && ++c->length <= 64);
897
898                 if (!assigned) {
899                         netifd_log_message(L_WARNING, "Failed to assign subprefix "
900                                         "of size %hhu for %s\n", c->length, c->name);
901                         free(c);
902                 } else {
903                         assigned_any = true;
904                 }
905         }
906
907         list_for_each_entry(c, &prefix->assignments, head)
908                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
909                         interface_set_prefix_address(c, prefix, iface, true);
910
911         if (!assigned_any)
912                 netifd_log_message(L_WARNING, "You have delegated IPv6-prefixes but haven't assigned them "
913                                 "to any interface. Did you forget to set option ip6assign on your lan-interfaces?");
914 }
915
916
917 void interface_refresh_assignments(bool hint)
918 {
919         static bool refresh = false;
920         if (!hint && refresh) {
921                 struct device_prefix *p;
922                 list_for_each_entry(p, &prefixes, head)
923                         interface_update_prefix_assignments(p, true);
924         }
925         refresh = hint;
926 }
927
928
929 static void
930 interface_update_prefix(struct vlist_tree *tree,
931                              struct vlist_node *node_new,
932                              struct vlist_node *node_old)
933 {
934         struct device_prefix *prefix_old, *prefix_new;
935         prefix_old = container_of(node_old, struct device_prefix, node);
936         prefix_new = container_of(node_new, struct device_prefix, node);
937
938         struct interface_ip_settings *ip = container_of(tree, struct interface_ip_settings, prefix);
939         if (tree && (!node_new || !node_old))
940                 ip->iface->updated |= IUF_PREFIX;
941
942         struct device_route route;
943         memset(&route, 0, sizeof(route));
944         route.flags = DEVADDR_INET6;
945         route.metric = INT32_MAX;
946         route.mask = (node_new) ? prefix_new->length : prefix_old->length;
947         route.addr.in6 = (node_new) ? prefix_new->addr : prefix_old->addr;
948
949
950         struct device_prefix_assignment *c;
951         struct interface *iface;
952
953         if (node_old && node_new) {
954                 // Move assignments and refresh addresses to update valid times
955                 list_splice(&prefix_old->assignments, &prefix_new->assignments);
956
957                 list_for_each_entry(c, &prefix_new->assignments, head)
958                         if ((iface = vlist_find(&interfaces, c->name, iface, node)))
959                                 interface_set_prefix_address(c, prefix_new, iface, true);
960         } else if (node_new) {
961                 // Set null-route to avoid routing loops
962                 system_add_route(NULL, &route);
963
964                 if (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation)
965                         interface_update_prefix_assignments(prefix_new, true);
966         } else if (node_old) {
967                 // Remove null-route
968                 interface_update_prefix_assignments(prefix_old, false);
969                 system_del_route(NULL, &route);
970         }
971
972         if (node_old) {
973                 if (prefix_old->head.next)
974                         list_del(&prefix_old->head);
975                 free(prefix_old);
976         }
977
978         if (node_new && (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation))
979                 list_add(&prefix_new->head, &prefixes);
980
981 }
982
983 struct device_prefix*
984 interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
985                 uint8_t length, time_t valid_until, time_t preferred_until,
986                 struct in6_addr *excl_addr, uint8_t excl_length, const char *pclass)
987 {
988         if (!pclass)
989                 pclass = (iface) ? iface->name : "local";
990
991         struct device_prefix *prefix = calloc(1, sizeof(*prefix) + strlen(pclass) + 1);
992         prefix->length = length;
993         prefix->addr = *addr;
994         prefix->preferred_until = preferred_until;
995         prefix->valid_until = valid_until;
996         prefix->iface = iface;
997         INIT_LIST_HEAD(&prefix->assignments);
998
999         if (excl_addr) {
1000                 prefix->excl_addr = *excl_addr;
1001                 prefix->excl_length = excl_length;
1002         }
1003
1004         strcpy(prefix->pclass, pclass);
1005
1006         if (iface)
1007                 vlist_add(&iface->proto_ip.prefix, &prefix->node, &prefix->addr);
1008         else
1009                 interface_update_prefix(NULL, &prefix->node, NULL);
1010
1011         return prefix;
1012 }
1013
1014 void
1015 interface_ip_set_ula_prefix(const char *prefix)
1016 {
1017         char buf[INET6_ADDRSTRLEN + 4] = {0}, *saveptr;
1018         if (prefix)
1019                 strncpy(buf, prefix, sizeof(buf) - 1);
1020         char *prefixaddr = strtok_r(buf, "/", &saveptr);
1021
1022         struct in6_addr addr;
1023         if (!prefixaddr || inet_pton(AF_INET6, prefixaddr, &addr) < 1) {
1024                 if (ula_prefix) {
1025                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
1026                         ula_prefix = NULL;
1027                 }
1028                 return;
1029         }
1030
1031         int length;
1032         char *prefixlen = strtok_r(NULL, ",", &saveptr);
1033         if (!prefixlen || (length = atoi(prefixlen)) < 1 || length > 64)
1034                 return;
1035
1036         if (!ula_prefix || !IN6_ARE_ADDR_EQUAL(&addr, &ula_prefix->addr) ||
1037                         ula_prefix->length != length) {
1038                 if (ula_prefix)
1039                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
1040
1041                 ula_prefix = interface_ip_add_device_prefix(NULL, &addr, length,
1042                                 0, 0, NULL, 0, NULL);
1043         }
1044 }
1045
1046 void
1047 interface_add_dns_server(struct interface_ip_settings *ip, const char *str)
1048 {
1049         struct dns_server *s;
1050
1051         s = calloc(1, sizeof(*s));
1052         if (!s)
1053                 return;
1054
1055         s->af = AF_INET;
1056         if (inet_pton(s->af, str, &s->addr.in))
1057                 goto add;
1058
1059         s->af = AF_INET6;
1060         if (inet_pton(s->af, str, &s->addr.in))
1061                 goto add;
1062
1063         free(s);
1064         return;
1065
1066 add:
1067         D(INTERFACE, "Add IPv%c DNS server: %s\n",
1068           s->af == AF_INET6 ? '6' : '4', str);
1069         vlist_simple_add(&ip->dns_servers, &s->node);
1070 }
1071
1072 void
1073 interface_add_dns_server_list(struct interface_ip_settings *ip, struct blob_attr *list)
1074 {
1075         struct blob_attr *cur;
1076         int rem;
1077
1078         blobmsg_for_each_attr(cur, list, rem) {
1079                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1080                         continue;
1081
1082                 if (!blobmsg_check_attr(cur, NULL))
1083                         continue;
1084
1085                 interface_add_dns_server(ip, blobmsg_data(cur));
1086         }
1087 }
1088
1089 static void
1090 interface_add_dns_search_domain(struct interface_ip_settings *ip, const char *str)
1091 {
1092         struct dns_search_domain *s;
1093         int len = strlen(str);
1094
1095         s = calloc(1, sizeof(*s) + len + 1);
1096         if (!s)
1097                 return;
1098
1099         D(INTERFACE, "Add DNS search domain: %s\n", str);
1100         memcpy(s->name, str, len);
1101         vlist_simple_add(&ip->dns_search, &s->node);
1102 }
1103
1104 void
1105 interface_add_dns_search_list(struct interface_ip_settings *ip, struct blob_attr *list)
1106 {
1107         struct blob_attr *cur;
1108         int rem;
1109
1110         blobmsg_for_each_attr(cur, list, rem) {
1111                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1112                         continue;
1113
1114                 if (!blobmsg_check_attr(cur, NULL))
1115                         continue;
1116
1117                 interface_add_dns_search_domain(ip, blobmsg_data(cur));
1118         }
1119 }
1120
1121 static void
1122 write_resolv_conf_entries(FILE *f, struct interface_ip_settings *ip, const char *dev)
1123 {
1124         struct dns_server *s;
1125         struct dns_search_domain *d;
1126         const char *str;
1127         char buf[INET6_ADDRSTRLEN];
1128
1129         vlist_simple_for_each_element(&ip->dns_servers, s, node) {
1130                 str = inet_ntop(s->af, &s->addr, buf, sizeof(buf));
1131                 if (!str)
1132                         continue;
1133
1134                 if (s->af == AF_INET6 && IN6_IS_ADDR_LINKLOCAL(&s->addr.in6))
1135                         fprintf(f, "nameserver %s%%%s\n", str, dev);
1136                 else
1137                         fprintf(f, "nameserver %s\n", str);
1138         }
1139
1140         vlist_simple_for_each_element(&ip->dns_search, d, node) {
1141                 fprintf(f, "search %s\n", d->name);
1142         }
1143 }
1144
1145 void
1146 interface_write_resolv_conf(void)
1147 {
1148         struct interface *iface;
1149         char *path = alloca(strlen(resolv_conf) + 5);
1150         FILE *f;
1151         uint32_t crcold, crcnew;
1152
1153         sprintf(path, "%s.tmp", resolv_conf);
1154         unlink(path);
1155         f = fopen(path, "w+");
1156         if (!f) {
1157                 D(INTERFACE, "Failed to open %s for writing\n", path);
1158                 return;
1159         }
1160
1161         vlist_for_each_element(&interfaces, iface, node) {
1162                 if (iface->state != IFS_UP)
1163                         continue;
1164
1165                 if (vlist_simple_empty(&iface->proto_ip.dns_search) &&
1166                     vlist_simple_empty(&iface->proto_ip.dns_servers) &&
1167                         vlist_simple_empty(&iface->config_ip.dns_search) &&
1168                     vlist_simple_empty(&iface->config_ip.dns_servers))
1169                         continue;
1170
1171                 fprintf(f, "# Interface %s\n", iface->name);
1172                 write_resolv_conf_entries(f, &iface->config_ip, iface->ifname);
1173                 if (!iface->proto_ip.no_dns)
1174                         write_resolv_conf_entries(f, &iface->proto_ip, iface->ifname);
1175         }
1176         fflush(f);
1177         rewind(f);
1178         crcnew = crc32_file(f);
1179         fclose(f);
1180
1181         crcold = crcnew + 1;
1182         f = fopen(resolv_conf, "r");
1183         if (f) {
1184                 crcold = crc32_file(f);
1185                 fclose(f);
1186         }
1187
1188         if (crcold == crcnew) {
1189                 unlink(path);
1190         } else if (rename(path, resolv_conf) < 0) {
1191                 D(INTERFACE, "Failed to replace %s\n", resolv_conf);
1192                 unlink(path);
1193         }
1194 }
1195
1196 void interface_ip_set_enabled(struct interface_ip_settings *ip, bool enabled)
1197 {
1198         struct device_addr *addr;
1199         struct device_route *route;
1200         struct device *dev;
1201         struct interface *iface;
1202
1203         ip->enabled = enabled;
1204         iface = ip->iface;
1205         dev = iface->l3_dev.dev;
1206         if (!dev)
1207                 return;
1208
1209         vlist_for_each_element(&ip->addr, addr, node) {
1210                 if (addr->enabled == enabled)
1211                         continue;
1212
1213                 if (enabled) {
1214                         system_add_address(dev, addr);
1215                         if (iface->metric)
1216                                 interface_handle_subnet_route(iface, addr, true);
1217                 } else {
1218                         interface_handle_subnet_route(iface, addr, false);
1219                         system_del_address(dev, addr);
1220                 }
1221                 addr->enabled = enabled;
1222         }
1223
1224         vlist_for_each_element(&ip->route, route, node) {
1225                 bool _enabled = enabled;
1226
1227                 if (!enable_route(ip, route))
1228                         _enabled = false;
1229
1230                 if (route->enabled == _enabled)
1231                         continue;
1232
1233                 if (_enabled) {
1234                         if (!(route->flags & DEVROUTE_METRIC))
1235                                 route->metric = ip->iface->metric;
1236
1237                         if (!(route->flags & DEVROUTE_TABLE)) {
1238                                 route->flags &= ~DEVROUTE_SRCTABLE;
1239                                 route->table = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET6) ?
1240                                                         iface->ip6table : iface->ip4table;
1241
1242                                 if (route->table)
1243                                         route->flags |= DEVROUTE_SRCTABLE;
1244                         }
1245
1246                         if (system_add_route(dev, route))
1247                                 route->failed = true;
1248                 } else
1249                         system_del_route(dev, route);
1250                 route->enabled = _enabled;
1251         }
1252
1253         struct device_prefix *c;
1254         struct device_prefix_assignment *a;
1255         list_for_each_entry(c, &prefixes, head)
1256                 list_for_each_entry(a, &c->assignments, head)
1257                         if (!strcmp(a->name, ip->iface->name))
1258                                 interface_set_prefix_address(a, c, ip->iface, enabled);
1259
1260         if (ip->iface && ip->iface->l3_dev.dev)
1261                 set_ip_source_policy(enabled, true, IPRULE_PRIORITY_REJECT + ip->iface->l3_dev.dev->ifindex,
1262                         NULL, 0, 0, ip->iface, "failed_policy");
1263 }
1264
1265 void
1266 interface_ip_update_start(struct interface_ip_settings *ip)
1267 {
1268         if (ip != &ip->iface->config_ip) {
1269                 vlist_simple_update(&ip->dns_servers);
1270                 vlist_simple_update(&ip->dns_search);
1271         }
1272         vlist_update(&ip->route);
1273         vlist_update(&ip->addr);
1274         vlist_update(&ip->prefix);
1275 }
1276
1277 void
1278 interface_ip_update_complete(struct interface_ip_settings *ip)
1279 {
1280         vlist_simple_flush(&ip->dns_servers);
1281         vlist_simple_flush(&ip->dns_search);
1282         vlist_flush(&ip->route);
1283         vlist_flush(&ip->addr);
1284         vlist_flush(&ip->prefix);
1285         interface_write_resolv_conf();
1286 }
1287
1288 void
1289 interface_ip_flush(struct interface_ip_settings *ip)
1290 {
1291         if (ip == &ip->iface->proto_ip)
1292                 vlist_flush_all(&ip->iface->host_routes);
1293         vlist_simple_flush_all(&ip->dns_servers);
1294         vlist_simple_flush_all(&ip->dns_search);
1295         vlist_flush_all(&ip->route);
1296         vlist_flush_all(&ip->addr);
1297         vlist_flush_all(&ip->prefix);
1298 }
1299
1300 static void
1301 __interface_ip_init(struct interface_ip_settings *ip, struct interface *iface)
1302 {
1303         ip->iface = iface;
1304         ip->enabled = true;
1305         vlist_simple_init(&ip->dns_search, struct dns_search_domain, node);
1306         vlist_simple_init(&ip->dns_servers, struct dns_server, node);
1307         vlist_init(&ip->route, route_cmp, interface_update_proto_route);
1308         vlist_init(&ip->addr, addr_cmp, interface_update_proto_addr);
1309         vlist_init(&ip->prefix, prefix_cmp, interface_update_prefix);
1310 }
1311
1312 void
1313 interface_ip_init(struct interface *iface)
1314 {
1315         __interface_ip_init(&iface->proto_ip, iface);
1316         __interface_ip_init(&iface->config_ip, iface);
1317         vlist_init(&iface->host_routes, route_cmp, interface_update_host_route);
1318 }
1319
1320 static void
1321 interface_ip_valid_until_handler(struct uloop_timeout *t)
1322 {
1323         time_t now = system_get_rtime();
1324         struct interface *iface;
1325         vlist_for_each_element(&interfaces, iface, node) {
1326                 if (iface->state != IFS_UP)
1327                         continue;
1328
1329                 struct device_addr *addr, *addrp;
1330                 struct device_route *route, *routep;
1331                 struct device_prefix *pref, *prefp;
1332
1333                 vlist_for_each_element_safe(&iface->proto_ip.addr, addr, node, addrp)
1334                         if (addr->valid_until && addr->valid_until < now)
1335                                 vlist_delete(&iface->proto_ip.addr, &addr->node);
1336
1337                 vlist_for_each_element_safe(&iface->proto_ip.route, route, node, routep)
1338                         if (route->valid_until && route->valid_until < now)
1339                                 vlist_delete(&iface->proto_ip.route, &route->node);
1340
1341                 vlist_for_each_element_safe(&iface->proto_ip.prefix, pref, node, prefp)
1342                         if (pref->valid_until && pref->valid_until < now)
1343                                 vlist_delete(&iface->proto_ip.prefix, &pref->node);
1344
1345         }
1346
1347         uloop_timeout_set(t, 1000);
1348 }
1349
1350 static void __init
1351 interface_ip_init_worker(void)
1352 {
1353         valid_until_timeout.cb = interface_ip_valid_until_handler;
1354         uloop_timeout_set(&valid_until_timeout, 1000);
1355 }