a6a7dca3189b0e15dc2a2ab03b61ad28b021bf46
[project/netifd.git] / interface-ip.c
1 /*
2  * netifd - network interface daemon
3  * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
4  * Copyright (C) 2012 Steven Barth <steven@midlink.org>
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License version 2
8  * as published by the Free Software Foundation
9  *
10  * This program is distributed in the hope that it will be useful,
11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
13  * GNU General Public License for more details.
14  */
15 #include <string.h>
16 #include <stdlib.h>
17 #include <stdio.h>
18
19 #include <limits.h>
20 #include <arpa/inet.h>
21
22 #include "netifd.h"
23 #include "device.h"
24 #include "interface.h"
25 #include "interface-ip.h"
26 #include "proto.h"
27 #include "ubus.h"
28 #include "system.h"
29
30 enum {
31         ROUTE_INTERFACE,
32         ROUTE_TARGET,
33         ROUTE_MASK,
34         ROUTE_GATEWAY,
35         ROUTE_METRIC,
36         ROUTE_MTU,
37         ROUTE_VALID,
38         ROUTE_TABLE,
39         ROUTE_SOURCE,
40         __ROUTE_MAX
41 };
42
43 static const struct blobmsg_policy route_attr[__ROUTE_MAX] = {
44         [ROUTE_INTERFACE] = { .name = "interface", .type = BLOBMSG_TYPE_STRING },
45         [ROUTE_TARGET] = { .name = "target", .type = BLOBMSG_TYPE_STRING },
46         [ROUTE_MASK] = { .name = "netmask", .type = BLOBMSG_TYPE_STRING },
47         [ROUTE_GATEWAY] = { .name = "gateway", .type = BLOBMSG_TYPE_STRING },
48         [ROUTE_METRIC] = { .name = "metric", .type = BLOBMSG_TYPE_INT32 },
49         [ROUTE_MTU] = { .name = "mtu", .type = BLOBMSG_TYPE_INT32 },
50         [ROUTE_TABLE] = { .name = "table", .type = BLOBMSG_TYPE_STRING },
51         [ROUTE_VALID] = { .name = "valid", .type = BLOBMSG_TYPE_INT32 },
52         [ROUTE_SOURCE] = { .name = "source", .type = BLOBMSG_TYPE_STRING },
53 };
54
55 const struct uci_blob_param_list route_attr_list = {
56         .n_params = __ROUTE_MAX,
57         .params = route_attr,
58 };
59
60
61 struct list_head prefixes = LIST_HEAD_INIT(prefixes);
62 static struct device_prefix *ula_prefix = NULL;
63 static struct uloop_timeout valid_until_timeout;
64
65
66 static void
67 clear_if_addr(union if_addr *a, int mask)
68 {
69         int m_bytes = (mask + 7) / 8;
70         uint8_t m_clear = (1 << (m_bytes * 8 - mask)) - 1;
71         uint8_t *p = (uint8_t *) a;
72
73         if (m_bytes < sizeof(a))
74                 memset(p + m_bytes, 0, sizeof(a) - m_bytes);
75
76         p[m_bytes - 1] &= ~m_clear;
77 }
78
79 static bool
80 match_if_addr(union if_addr *a1, union if_addr *a2, int mask)
81 {
82         union if_addr *p1, *p2;
83
84         p1 = alloca(sizeof(*a1));
85         p2 = alloca(sizeof(*a2));
86
87         memcpy(p1, a1, sizeof(*a1));
88         clear_if_addr(p1, mask);
89         memcpy(p2, a2, sizeof(*a2));
90         clear_if_addr(p2, mask);
91
92         return !memcmp(p1, p2, sizeof(*p1));
93 }
94
95 static int set_ip_source_policy(bool add, bool v6, unsigned int priority,
96                 const union if_addr *addr, uint8_t mask, unsigned int table,
97                 struct interface *in_iface, const char *action)
98 {
99         struct iprule rule = {
100                 .flags = IPRULE_PRIORITY,
101                 .priority = priority
102         };
103
104         if (addr) {
105                 rule.flags |= IPRULE_SRC;
106                 rule.src_addr = *addr;
107                 rule.src_mask = mask;
108         }
109
110         if (table) {
111                 rule.flags |= IPRULE_LOOKUP;
112                 rule.lookup = table;
113
114                 if (!rule.lookup)
115                         return 0;
116         } else if (action) {
117                 rule.flags |= IPRULE_ACTION;
118                 system_resolve_iprule_action(action, &rule.action);
119         }
120
121         if (in_iface && in_iface->l3_dev.dev) {
122                 rule.flags |= IPRULE_IN;
123                 strcpy(rule.in_dev, in_iface->l3_dev.dev->ifname);
124         }
125
126         rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
127
128         return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
129 }
130
131 static int set_ip_lo_policy(bool add, bool v6, struct interface *iface)
132 {
133         struct iprule rule = {
134                 .flags = IPRULE_IN | IPRULE_LOOKUP | IPRULE_PRIORITY,
135                 .priority = IPRULE_PRIORITY_NW + iface->l3_dev.dev->ifindex,
136                 .lookup = (v6) ? iface->ip6table : iface->ip4table,
137                 .in_dev = "lo"
138         };
139
140         if (!rule.lookup)
141                 return 0;
142
143         rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
144
145         return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
146 }
147
148 static bool
149 __find_ip_addr_target(struct interface_ip_settings *ip, union if_addr *a, bool v6)
150 {
151         struct device_addr *addr;
152
153         vlist_for_each_element(&ip->addr, addr, node) {
154                 if (!addr->enabled)
155                         continue;
156
157                 if (v6 != ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
158                         continue;
159
160                 // Handle offlink addresses correctly
161                 unsigned int mask = addr->mask;
162                 if ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6 &&
163                                 (addr->flags & DEVADDR_OFFLINK))
164                         mask = 128;
165
166                 if (!match_if_addr(&addr->addr, a, mask))
167                         continue;
168
169                 return true;
170         }
171
172         return false;
173 }
174
175 static void
176 __find_ip_route_target(struct interface_ip_settings *ip, union if_addr *a,
177                        bool v6, struct device_route **res)
178 {
179         struct device_route *route;
180
181         vlist_for_each_element(&ip->route, route, node) {
182                 if (!route->enabled)
183                         continue;
184
185                 if (v6 != ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
186                         continue;
187
188                 if (!match_if_addr(&route->addr, a, route->mask))
189                         continue;
190
191                 if (route->flags & DEVROUTE_TABLE)
192                         continue;
193
194                 if (!*res || route->mask < (*res)->mask)
195                         *res = route;
196         }
197 }
198
199 static bool
200 interface_ip_find_addr_target(struct interface *iface, union if_addr *a, bool v6)
201 {
202         return __find_ip_addr_target(&iface->proto_ip, a, v6) ||
203                __find_ip_addr_target(&iface->config_ip, a, v6);
204 }
205
206 static void
207 interface_ip_find_route_target(struct interface *iface, union if_addr *a,
208                                bool v6, struct device_route **route)
209 {
210         __find_ip_route_target(&iface->proto_ip, a, v6, route);
211         __find_ip_route_target(&iface->config_ip, a, v6, route);
212 }
213
214 struct interface *
215 interface_ip_add_target_route(union if_addr *addr, bool v6, struct interface *iface)
216 {
217         struct device_route *route, *r_next = NULL;
218         bool defaultroute_target = false;
219         int addrsize = v6 ? sizeof(addr->in6) : sizeof(addr->in);
220
221         route = calloc(1, sizeof(*route));
222         if (!route)
223                 return NULL;
224
225         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
226         route->mask = v6 ? 128 : 32;
227         if (memcmp(&route->addr, addr, addrsize) == 0)
228                 defaultroute_target = true;
229         else
230                 memcpy(&route->addr, addr, addrsize);
231
232         if (iface) {
233                 /* look for locally addressable target first */
234                 if (interface_ip_find_addr_target(iface, addr, v6))
235                         goto done;
236
237                 /* do not stop at the first route, let the lookup compare
238                  * masks to find the best match */
239                 interface_ip_find_route_target(iface, addr, v6, &r_next);
240         } else {
241                 vlist_for_each_element(&interfaces, iface, node) {
242                         /* look for locally addressable target first */
243                         if (interface_ip_find_addr_target(iface, addr, v6))
244                                 goto done;
245
246                         /* do not stop at the first route, let the lookup compare
247                          * masks to find the best match */
248                         interface_ip_find_route_target(iface, addr, v6, &r_next);
249                 }
250         }
251
252         if (!r_next) {
253                 free(route);
254                 return NULL;
255         }
256
257         iface = r_next->iface;
258         memcpy(&route->nexthop, &r_next->nexthop, sizeof(route->nexthop));
259         route->mtu = r_next->mtu;
260         route->metric = r_next->metric;
261         route->table = r_next->table;
262
263 done:
264         route->iface = iface;
265         if (defaultroute_target)
266                 free(route);
267         else
268                 vlist_add(&iface->host_routes, &route->node, route);
269         return iface;
270 }
271
272 void
273 interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
274 {
275         struct interface_ip_settings *ip;
276         struct blob_attr *tb[__ROUTE_MAX], *cur;
277         struct device_route *route;
278         int af = v6 ? AF_INET6 : AF_INET;
279         bool is_proto_route = !!iface;
280
281         blobmsg_parse(route_attr, __ROUTE_MAX, tb, blobmsg_data(attr), blobmsg_data_len(attr));
282
283         if (!iface) {
284                 if ((cur = tb[ROUTE_INTERFACE]) == NULL)
285                         return;
286
287                 iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
288                 if (!iface)
289                         return;
290
291                 ip = &iface->config_ip;
292         } else {
293                 ip = &iface->proto_ip;
294         }
295
296         route = calloc(1, sizeof(*route));
297         if (!route)
298                 return;
299
300         route->flags = v6 ? DEVADDR_INET6 : DEVADDR_INET4;
301         route->mask = v6 ? 128 : 32;
302         if ((cur = tb[ROUTE_MASK]) != NULL) {
303                 route->mask = parse_netmask_string(blobmsg_data(cur), v6);
304                 if (route->mask > (v6 ? 128 : 32))
305                         goto error;
306         }
307
308         if ((cur = tb[ROUTE_TARGET]) != NULL) {
309                 if (!parse_ip_and_netmask(af, blobmsg_data(cur), &route->addr, &route->mask)) {
310                         DPRINTF("Failed to parse route target: %s\n", (char *) blobmsg_data(cur));
311                         goto error;
312                 }
313         }
314
315         if ((cur = tb[ROUTE_GATEWAY]) != NULL) {
316                 if (!inet_pton(af, blobmsg_data(cur), &route->nexthop)) {
317                         DPRINTF("Failed to parse route gateway: %s\n", (char *) blobmsg_data(cur));
318                         goto error;
319                 }
320         }
321
322         if ((cur = tb[ROUTE_METRIC]) != NULL) {
323                 route->metric = blobmsg_get_u32(cur);
324                 route->flags |= DEVROUTE_METRIC;
325         }
326
327         if ((cur = tb[ROUTE_MTU]) != NULL) {
328                 route->mtu = blobmsg_get_u32(cur);
329                 route->flags |= DEVROUTE_MTU;
330         }
331
332         // Use source-based routing
333         if ((cur = tb[ROUTE_SOURCE]) != NULL) {
334                 char *saveptr, *source = alloca(blobmsg_data_len(cur));
335                 memcpy(source, blobmsg_data(cur), blobmsg_data_len(cur));
336
337                 const char *addr = strtok_r(source, "/", &saveptr);
338                 const char *mask = strtok_r(NULL, "/", &saveptr);
339
340                 if (inet_pton(af, addr, &route->source) < 1) {
341                         DPRINTF("Failed to parse route source: %s\n", addr);
342                         goto error;
343                 }
344
345                 route->sourcemask = atoi(mask);
346         }
347
348         if (is_proto_route) {
349                 route->table = (v6) ? iface->ip6table : iface->ip4table;
350                 route->flags |= DEVROUTE_SRCTABLE;
351         }
352
353         if ((cur = tb[ROUTE_TABLE]) != NULL) {
354                 if (!system_resolve_rt_table(blobmsg_data(cur), &route->table)) {
355                         DPRINTF("Failed to resolve routing table: %s\n", (char *) blobmsg_data(cur));
356                         goto error;
357                 }
358
359                 if (route->table)
360                         route->flags |= DEVROUTE_TABLE;
361         }
362
363         if ((cur = tb[ROUTE_VALID]) != NULL) {
364                 int64_t valid = blobmsg_get_u32(cur);
365                 int64_t valid_until = valid + (int64_t)system_get_rtime();
366                 if (valid_until <= LONG_MAX && valid != 0xffffffffLL) // Catch overflow
367                         route->valid_until = valid_until;
368         }
369
370         vlist_add(&ip->route, &route->node, route);
371         return;
372
373 error:
374         free(route);
375 }
376
377 static int
378 addr_cmp(const void *k1, const void *k2, void *ptr)
379 {
380         return memcmp(k1, k2, sizeof(struct device_addr) -
381                       offsetof(struct device_addr, flags));
382 }
383
384 static int
385 route_cmp(const void *k1, const void *k2, void *ptr)
386 {
387         const struct device_route *r1 = k1, *r2 = k2;
388
389         if (r1->mask != r2->mask)
390                 return r2->mask - r1->mask;
391
392         if (r1->metric != r2->metric)
393                 return r1->metric - r2->metric;
394
395         if (r1->flags != r2->flags)
396                 return r2->flags - r1->flags;
397
398         if (r1->sourcemask != r2->sourcemask)
399                 return r1->sourcemask - r2->sourcemask;
400
401         int maskcmp = memcmp(&r1->source, &r2->source, sizeof(r1->source));
402         if (maskcmp)
403                 return maskcmp;
404
405         return memcmp(&r1->addr, &r2->addr, sizeof(r1->addr));
406 }
407
408 static int
409 prefix_cmp(const void *k1, const void *k2, void *ptr)
410 {
411         return memcmp(k1, k2, offsetof(struct device_prefix, pclass) -
412                         offsetof(struct device_prefix, addr));
413 }
414
415 static void
416 interface_handle_subnet_route(struct interface *iface, struct device_addr *addr, bool add)
417 {
418         struct device *dev = iface->l3_dev.dev;
419         struct device_route route;
420
421         memset(&route, 0, sizeof(route));
422         route.iface = iface;
423         route.flags = addr->flags;
424         route.mask = addr->mask;
425         memcpy(&route.addr, &addr->addr, sizeof(route.addr));
426         clear_if_addr(&route.addr, route.mask);
427
428         if (add) {
429                 route.flags |= DEVADDR_KERNEL;
430                 system_del_route(dev, &route);
431
432                 if (!(addr->flags & DEVADDR_OFFLINK)) {
433                         route.flags &= ~DEVADDR_KERNEL;
434                         route.metric = iface->metric;
435                         system_add_route(dev, &route);
436                 }
437         } else {
438                 if (!(addr->flags & DEVADDR_OFFLINK))
439                         system_del_route(dev, &route);
440         }
441 }
442
443 static void
444 interface_update_proto_addr(struct vlist_tree *tree,
445                             struct vlist_node *node_new,
446                             struct vlist_node *node_old)
447 {
448         struct interface_ip_settings *ip;
449         struct interface *iface;
450         struct device *dev;
451         struct device_addr *a_new = NULL, *a_old = NULL;
452         bool replace = false;
453         bool keep = false;
454         bool v6 = false;
455
456         ip = container_of(tree, struct interface_ip_settings, addr);
457         iface = ip->iface;
458         dev = iface->l3_dev.dev;
459
460         if (!node_new || !node_old)
461                 iface->updated |= IUF_ADDRESS;
462
463         if (node_new) {
464                 a_new = container_of(node_new, struct device_addr, node);
465
466                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
467                     !a_new->broadcast) {
468
469                         uint32_t mask = ~0;
470                         uint32_t *a = (uint32_t *) &a_new->addr;
471
472                         mask >>= a_new->mask;
473                         a_new->broadcast = *a | htonl(mask);
474                 }
475         }
476
477         if (node_old)
478                 a_old = container_of(node_old, struct device_addr, node);
479
480         if (a_new && a_old) {
481                 keep = true;
482
483                 if (a_old->flags != a_new->flags)
484                         keep = false;
485
486                 if (a_old->valid_until != a_new->valid_until ||
487                                 a_old->preferred_until != a_new->preferred_until)
488                         replace = true;
489
490                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET4 &&
491                     a_new->broadcast != a_old->broadcast)
492                         keep = false;
493         }
494
495         if (node_old) {
496                 if (!(a_old->flags & DEVADDR_EXTERNAL) && a_old->enabled && !keep) {
497                         interface_handle_subnet_route(iface, a_old, false);
498
499                         if ((a_old->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
500                                 v6 = true;
501
502                         //This is needed for source routing to work correctly. If a device
503                         //has two connections to a network using the same subnet, adding
504                         //only the network-rule will cause packets to be routed through the
505                         //first matching network (source IP matches both masks).
506                         set_ip_source_policy(false, v6, IPRULE_PRIORITY_ADDR, &a_old->addr,
507                                 (v6) ? 128 : 32, (v6) ? iface->ip6table : iface->ip4table,
508                                                 NULL, NULL);
509                         set_ip_source_policy(false, v6, IPRULE_PRIORITY_NW, &a_old->addr,
510                                 a_old->mask, (v6) ? iface->ip6table : iface->ip4table, NULL, NULL);
511
512                         system_del_address(dev, a_old);
513                 }
514                 free(a_old->pclass);
515                 free(a_old);
516         }
517
518         if (node_new) {
519                 a_new->enabled = true;
520                 if (!(a_new->flags & DEVADDR_EXTERNAL) && (!keep || replace)) {
521                         system_add_address(dev, a_new);
522
523                         if (!keep) {
524                                 if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
525                                         v6 = true;
526
527                                 set_ip_source_policy(true, v6, IPRULE_PRIORITY_ADDR, &a_new->addr,
528                                         (v6) ? 128 : 32, (v6) ? iface->ip6table : iface->ip4table,
529                                                         NULL, NULL);
530                                 set_ip_source_policy(true, v6, IPRULE_PRIORITY_NW, &a_new->addr,
531                                         a_new->mask, (v6) ? iface->ip6table : iface->ip4table,
532                                                         NULL, NULL);
533                         }
534
535                         if ((a_new->flags & DEVADDR_OFFLINK) || iface->metric)
536                                 interface_handle_subnet_route(iface, a_new, true);
537                 }
538         }
539 }
540
541 static bool
542 enable_route(struct interface_ip_settings *ip, struct device_route *route)
543 {
544         if (ip->no_defaultroute && !route->mask)
545                 return false;
546
547         return ip->enabled;
548 }
549
550 static void
551 interface_update_proto_route(struct vlist_tree *tree,
552                              struct vlist_node *node_new,
553                              struct vlist_node *node_old)
554 {
555         struct interface_ip_settings *ip;
556         struct interface *iface;
557         struct device *dev;
558         struct device_route *route_old, *route_new;
559         bool keep = false;
560
561         ip = container_of(tree, struct interface_ip_settings, route);
562         iface = ip->iface;
563         dev = iface->l3_dev.dev;
564
565         if (!node_new || !node_old)
566                 iface->updated |= IUF_ROUTE;
567
568         route_old = container_of(node_old, struct device_route, node);
569         route_new = container_of(node_new, struct device_route, node);
570
571         if (node_old && node_new)
572                 keep = !memcmp(&route_old->nexthop, &route_new->nexthop, sizeof(route_old->nexthop)) &&
573                         (route_old->table == route_new->table);
574
575         if (node_old) {
576                 if (!(route_old->flags & DEVADDR_EXTERNAL) && route_old->enabled && !keep)
577                         system_del_route(dev, route_old);
578
579                 free(route_old);
580         }
581
582         if (node_new) {
583                 bool _enabled = enable_route(ip, route_new);
584
585                 if (!(route_new->flags & DEVROUTE_METRIC))
586                         route_new->metric = iface->metric;
587
588                 if (!(route_new->flags & DEVADDR_EXTERNAL) && !keep && _enabled)
589                         system_add_route(dev, route_new);
590
591                 route_new->iface = iface;
592                 route_new->enabled = _enabled;
593         }
594 }
595
596 static void
597 interface_update_host_route(struct vlist_tree *tree,
598                              struct vlist_node *node_new,
599                              struct vlist_node *node_old)
600 {
601         struct interface *iface;
602         struct device *dev;
603         struct device_route *route_old, *route_new;
604
605         iface = container_of(tree, struct interface, host_routes);
606         dev = iface->l3_dev.dev;
607
608         route_old = container_of(node_old, struct device_route, node);
609         route_new = container_of(node_new, struct device_route, node);
610
611         if (node_old) {
612                 system_del_route(dev, route_old);
613                 free(route_old);
614         }
615
616         if (node_new)
617                 system_add_route(dev, route_new);
618 }
619
620
621 static void
622 interface_set_prefix_address(struct device_prefix_assignment *assignment,
623                 const struct device_prefix *prefix, struct interface *iface, bool add)
624 {
625         const struct interface *uplink = prefix->iface;
626         if (!iface->l3_dev.dev)
627                 return;
628
629         struct device *l3_downlink = iface->l3_dev.dev;
630
631         struct device_addr addr;
632         memset(&addr, 0, sizeof(addr));
633         addr.addr.in6 = prefix->addr;
634         addr.addr.in6.s6_addr32[1] |= htonl(assignment->assigned);
635         addr.addr.in6.s6_addr[15] += 1;
636         addr.mask = assignment->length;
637         addr.flags = DEVADDR_INET6;
638         addr.preferred_until = prefix->preferred_until;
639         addr.valid_until = prefix->valid_until;
640
641         if (!add && assignment->enabled) {
642                 time_t now = system_get_rtime();
643                 addr.preferred_until = now;
644                 if (!addr.valid_until || addr.valid_until - now > 7200)
645                         addr.valid_until = now + 7200;
646                 system_add_address(l3_downlink, &addr);
647                 if (prefix->iface) {
648                         set_ip_source_policy(false, true, IPRULE_PRIORITY_NW, &addr.addr,
649                                         addr.mask, prefix->iface->ip6table, iface, NULL);
650
651                         set_ip_source_policy(false, true, IPRULE_PRIORITY_REJECT, &addr.addr,
652                                                         addr.mask, 0, iface, "unreachable");
653                 }
654
655                 assignment->enabled = false;
656         } else if (add && (iface->state == IFS_UP || iface->state == IFS_SETUP)) {
657                 system_add_address(l3_downlink, &addr);
658                 if (prefix->iface && !assignment->enabled) {
659                         set_ip_source_policy(true, true, IPRULE_PRIORITY_REJECT, &addr.addr,
660                                         addr.mask, 0, iface, "unreachable");
661
662                         set_ip_source_policy(true, true, IPRULE_PRIORITY_NW, &addr.addr,
663                                         addr.mask, prefix->iface->ip6table, iface, NULL);
664                 }
665                 if (uplink && uplink->l3_dev.dev) {
666                         int mtu = system_update_ipv6_mtu(
667                                         uplink->l3_dev.dev, 0);
668                         if (mtu > 0)
669                                 system_update_ipv6_mtu(l3_downlink, mtu);
670                 }
671                 assignment->enabled = true;
672         }
673 }
674
675 static bool interface_prefix_assign(struct list_head *list,
676                 struct device_prefix_assignment *assign)
677 {
678         int32_t current = 0, asize = (1 << (64 - assign->length)) - 1;
679         struct device_prefix_assignment *c;
680         list_for_each_entry(c, list, head) {
681                 if (assign->assigned != -1) {
682                         if (assign->assigned >= current && assign->assigned + asize < c->assigned) {
683                                 list_add_tail(&assign->head, &c->head);
684                                 return true;
685                         }
686                 } else if (assign->assigned == -1) {
687                         current = (current + asize) & (~asize);
688                         if (current + asize < c->assigned) {
689                                 assign->assigned = current;
690                                 list_add_tail(&assign->head, &c->head);
691                                 return true;
692                         }
693                 }
694                 current = (c->assigned + (1 << (64 - c->length)));
695         }
696         return false;
697 }
698
699 static void interface_update_prefix_assignments(struct device_prefix *prefix, bool setup)
700 {
701         struct device_prefix_assignment *c;
702         struct interface *iface;
703
704         // Delete all assignments
705         while (!list_empty(&prefix->assignments)) {
706                 c = list_first_entry(&prefix->assignments,
707                                 struct device_prefix_assignment, head);
708                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
709                         interface_set_prefix_address(c, prefix, iface, false);
710                 list_del(&c->head);
711                 free(c);
712         }
713
714         if (!setup)
715                 return;
716
717         // End-of-assignment sentinel
718         c = malloc(sizeof(*c) + 1);
719         c->assigned = 1 << (64 - prefix->length);
720         c->length = 64;
721         c->name[0] = 0;
722         list_add(&c->head, &prefix->assignments);
723
724         // Excluded prefix
725         if (prefix->excl_length > 0) {
726                 const char name[] = "!excluded";
727                 c = malloc(sizeof(*c) + sizeof(name));
728                 c->assigned = ntohl(prefix->excl_addr.s6_addr32[1]) &
729                                 ((1 << (64 - prefix->length)) - 1);
730                 c->length = prefix->excl_length;
731                 memcpy(c->name, name, sizeof(name));
732                 list_add(&c->head, &prefix->assignments);
733         }
734
735         bool assigned_any = false;
736         struct list_head assign_later = LIST_HEAD_INIT(assign_later);
737         vlist_for_each_element(&interfaces, iface, node) {
738                 if (iface->assignment_length < 48 ||
739                                 iface->assignment_length > 64)
740                         continue;
741
742                 // Test whether there is a matching class
743                 if (!list_empty(&iface->assignment_classes)) {
744                         bool found = false;
745
746                         struct interface_assignment_class *c;
747                         list_for_each_entry(c, &iface->assignment_classes, head) {
748                                 if (!strcmp(c->name, prefix->pclass)) {
749                                         found = true;
750                                         break;
751                                 }
752                         }
753
754                         if (!found)
755                                 continue;
756                 }
757
758                 size_t namelen = strlen(iface->name) + 1;
759                 c = malloc(sizeof(*c) + namelen);
760                 c->length = iface->assignment_length;
761                 c->assigned = iface->assignment_hint;
762                 c->enabled = false;
763                 memcpy(c->name, iface->name, namelen);
764
765                 // First process all custom assignments, put all others in later-list
766                 if (c->assigned == -1 || !interface_prefix_assign(&prefix->assignments, c)) {
767                         if (c->assigned != -1) {
768                                 c->assigned = -1;
769                                 netifd_log_message(L_WARNING, "Failed to assign requested subprefix "
770                                                 "of size %hhu for %s, trying other\n", c->length, c->name);
771                         }
772
773                         struct list_head *next = &assign_later;
774                         struct device_prefix_assignment *n;
775                         list_for_each_entry(n, &assign_later, head) {
776                                 if (n->length < c->length) {
777                                         next = &n->head;
778                                         break;
779                                 }
780                         }
781                         list_add_tail(&c->head, next);
782                 }
783
784                 if (c->assigned != -1)
785                         assigned_any = true;
786         }
787
788         // Then try to assign all other + failed custom assignments
789         while (!list_empty(&assign_later)) {
790                 c = list_first_entry(&assign_later, struct device_prefix_assignment, head);
791                 list_del(&c->head);
792
793                 bool assigned = false;
794                 do {
795                         assigned = interface_prefix_assign(&prefix->assignments, c);
796                 } while (!assigned && ++c->length <= 64);
797
798                 if (!assigned) {
799                         netifd_log_message(L_WARNING, "Failed to assign subprefix "
800                                         "of size %hhu for %s\n", c->length, c->name);
801                         free(c);
802                 } else {
803                         assigned_any = true;
804                 }
805         }
806
807         list_for_each_entry(c, &prefix->assignments, head)
808                 if ((iface = vlist_find(&interfaces, c->name, iface, node)))
809                         interface_set_prefix_address(c, prefix, iface, true);
810
811         if (!assigned_any)
812                 netifd_log_message(L_WARNING, "You have delegated IPv6-prefixes but haven't assigned them "
813                                 "to any interface. Did you forget to set option ip6assign on your lan-interfaces?");
814 }
815
816
817 void interface_refresh_assignments(bool hint)
818 {
819         static bool refresh = false;
820         if (!hint && refresh) {
821                 struct device_prefix *p;
822                 list_for_each_entry(p, &prefixes, head)
823                         interface_update_prefix_assignments(p, true);
824         }
825         refresh = hint;
826 }
827
828
829 static void
830 interface_update_prefix(struct vlist_tree *tree,
831                              struct vlist_node *node_new,
832                              struct vlist_node *node_old)
833 {
834         struct device_prefix *prefix_old, *prefix_new;
835         prefix_old = container_of(node_old, struct device_prefix, node);
836         prefix_new = container_of(node_new, struct device_prefix, node);
837
838         struct interface_ip_settings *ip = container_of(tree, struct interface_ip_settings, prefix);
839         if (tree && (!node_new || !node_old))
840                 ip->iface->updated |= IUF_PREFIX;
841
842         struct device_route route;
843         memset(&route, 0, sizeof(route));
844         route.flags = DEVADDR_INET6;
845         route.metric = INT32_MAX;
846         route.mask = (node_new) ? prefix_new->length : prefix_old->length;
847         route.addr.in6 = (node_new) ? prefix_new->addr : prefix_old->addr;
848
849
850         struct device_prefix_assignment *c;
851         struct interface *iface;
852
853         if (node_old && node_new) {
854                 // Move assignments and refresh addresses to update valid times
855                 list_splice(&prefix_old->assignments, &prefix_new->assignments);
856
857                 list_for_each_entry(c, &prefix_new->assignments, head)
858                         if ((iface = vlist_find(&interfaces, c->name, iface, node)))
859                                 interface_set_prefix_address(c, prefix_new, iface, true);
860         } else if (node_new) {
861                 // Set null-route to avoid routing loops
862                 system_add_route(NULL, &route);
863
864                 if (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation)
865                         interface_update_prefix_assignments(prefix_new, true);
866         } else if (node_old) {
867                 // Remove null-route
868                 interface_update_prefix_assignments(prefix_old, false);
869                 system_del_route(NULL, &route);
870         }
871
872         if (node_old) {
873                 if (prefix_old->head.next)
874                         list_del(&prefix_old->head);
875                 free(prefix_old);
876         }
877
878         if (node_new && (!prefix_new->iface || !prefix_new->iface->proto_ip.no_delegation))
879                 list_add(&prefix_new->head, &prefixes);
880
881 }
882
883 struct device_prefix*
884 interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
885                 uint8_t length, time_t valid_until, time_t preferred_until,
886                 struct in6_addr *excl_addr, uint8_t excl_length, const char *pclass)
887 {
888         if (!pclass)
889                 pclass = (iface) ? iface->name : "local";
890
891         struct device_prefix *prefix = calloc(1, sizeof(*prefix) + strlen(pclass) + 1);
892         prefix->length = length;
893         prefix->addr = *addr;
894         prefix->preferred_until = preferred_until;
895         prefix->valid_until = valid_until;
896         prefix->iface = iface;
897         INIT_LIST_HEAD(&prefix->assignments);
898
899         if (excl_addr) {
900                 prefix->excl_addr = *excl_addr;
901                 prefix->excl_length = excl_length;
902         }
903
904         strcpy(prefix->pclass, pclass);
905
906         if (iface)
907                 vlist_add(&iface->proto_ip.prefix, &prefix->node, &prefix->addr);
908         else
909                 interface_update_prefix(NULL, &prefix->node, NULL);
910
911         return prefix;
912 }
913
914 void
915 interface_ip_set_ula_prefix(const char *prefix)
916 {
917         char buf[INET6_ADDRSTRLEN + 4] = {0}, *saveptr;
918         if (prefix)
919                 strncpy(buf, prefix, sizeof(buf) - 1);
920         char *prefixaddr = strtok_r(buf, "/", &saveptr);
921
922         struct in6_addr addr;
923         if (!prefixaddr || inet_pton(AF_INET6, prefixaddr, &addr) < 1) {
924                 if (ula_prefix) {
925                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
926                         ula_prefix = NULL;
927                 }
928                 return;
929         }
930
931         int length;
932         char *prefixlen = strtok_r(NULL, ",", &saveptr);
933         if (!prefixlen || (length = atoi(prefixlen)) < 1 || length > 64)
934                 return;
935
936         if (!ula_prefix || !IN6_ARE_ADDR_EQUAL(&addr, &ula_prefix->addr) ||
937                         ula_prefix->length != length) {
938                 if (ula_prefix)
939                         interface_update_prefix(NULL, NULL, &ula_prefix->node);
940
941                 ula_prefix = interface_ip_add_device_prefix(NULL, &addr, length,
942                                 0, 0, NULL, 0, NULL);
943         }
944 }
945
946 void
947 interface_add_dns_server(struct interface_ip_settings *ip, const char *str)
948 {
949         struct dns_server *s;
950
951         s = calloc(1, sizeof(*s));
952         if (!s)
953                 return;
954
955         s->af = AF_INET;
956         if (inet_pton(s->af, str, &s->addr.in))
957                 goto add;
958
959         s->af = AF_INET6;
960         if (inet_pton(s->af, str, &s->addr.in))
961                 goto add;
962
963         free(s);
964         return;
965
966 add:
967         D(INTERFACE, "Add IPv%c DNS server: %s\n",
968           s->af == AF_INET6 ? '6' : '4', str);
969         vlist_simple_add(&ip->dns_servers, &s->node);
970 }
971
972 void
973 interface_add_dns_server_list(struct interface_ip_settings *ip, struct blob_attr *list)
974 {
975         struct blob_attr *cur;
976         int rem;
977
978         blobmsg_for_each_attr(cur, list, rem) {
979                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
980                         continue;
981
982                 if (!blobmsg_check_attr(cur, NULL))
983                         continue;
984
985                 interface_add_dns_server(ip, blobmsg_data(cur));
986         }
987 }
988
989 static void
990 interface_add_dns_search_domain(struct interface_ip_settings *ip, const char *str)
991 {
992         struct dns_search_domain *s;
993         int len = strlen(str);
994
995         s = calloc(1, sizeof(*s) + len + 1);
996         if (!s)
997                 return;
998
999         D(INTERFACE, "Add DNS search domain: %s\n", str);
1000         memcpy(s->name, str, len);
1001         vlist_simple_add(&ip->dns_search, &s->node);
1002 }
1003
1004 void
1005 interface_add_dns_search_list(struct interface_ip_settings *ip, struct blob_attr *list)
1006 {
1007         struct blob_attr *cur;
1008         int rem;
1009
1010         blobmsg_for_each_attr(cur, list, rem) {
1011                 if (blobmsg_type(cur) != BLOBMSG_TYPE_STRING)
1012                         continue;
1013
1014                 if (!blobmsg_check_attr(cur, NULL))
1015                         continue;
1016
1017                 interface_add_dns_search_domain(ip, blobmsg_data(cur));
1018         }
1019 }
1020
1021 static void
1022 write_resolv_conf_entries(FILE *f, struct interface_ip_settings *ip)
1023 {
1024         struct dns_server *s;
1025         struct dns_search_domain *d;
1026         const char *str;
1027         char buf[INET6_ADDRSTRLEN];
1028
1029         vlist_simple_for_each_element(&ip->dns_servers, s, node) {
1030                 str = inet_ntop(s->af, &s->addr, buf, sizeof(buf));
1031                 if (!str)
1032                         continue;
1033
1034                 fprintf(f, "nameserver %s\n", str);
1035         }
1036
1037         vlist_simple_for_each_element(&ip->dns_search, d, node) {
1038                 fprintf(f, "search %s\n", d->name);
1039         }
1040 }
1041
1042 void
1043 interface_write_resolv_conf(void)
1044 {
1045         struct interface *iface;
1046         char *path = alloca(strlen(resolv_conf) + 5);
1047         FILE *f;
1048         uint32_t crcold, crcnew;
1049
1050         sprintf(path, "%s.tmp", resolv_conf);
1051         unlink(path);
1052         f = fopen(path, "w+");
1053         if (!f) {
1054                 D(INTERFACE, "Failed to open %s for writing\n", path);
1055                 return;
1056         }
1057
1058         vlist_for_each_element(&interfaces, iface, node) {
1059                 if (iface->state != IFS_UP)
1060                         continue;
1061
1062                 if (vlist_simple_empty(&iface->proto_ip.dns_search) &&
1063                     vlist_simple_empty(&iface->proto_ip.dns_servers) &&
1064                         vlist_simple_empty(&iface->config_ip.dns_search) &&
1065                     vlist_simple_empty(&iface->config_ip.dns_servers))
1066                         continue;
1067
1068                 fprintf(f, "# Interface %s\n", iface->name);
1069                 write_resolv_conf_entries(f, &iface->config_ip);
1070                 if (!iface->proto_ip.no_dns)
1071                         write_resolv_conf_entries(f, &iface->proto_ip);
1072         }
1073         fflush(f);
1074         rewind(f);
1075         crcnew = crc32_file(f);
1076         fclose(f);
1077
1078         crcold = crcnew + 1;
1079         f = fopen(resolv_conf, "r");
1080         if (f) {
1081                 crcold = crc32_file(f);
1082                 fclose(f);
1083         }
1084
1085         if (crcold == crcnew) {
1086                 unlink(path);
1087         } else if (rename(path, resolv_conf) < 0) {
1088                 D(INTERFACE, "Failed to replace %s\n", resolv_conf);
1089                 unlink(path);
1090         }
1091 }
1092
1093 void interface_ip_set_enabled(struct interface_ip_settings *ip, bool enabled)
1094 {
1095         struct device_addr *addr;
1096         struct device_route *route;
1097         struct device *dev;
1098
1099         ip->enabled = enabled;
1100         dev = ip->iface->l3_dev.dev;
1101         if (!dev)
1102                 return;
1103
1104         vlist_for_each_element(&ip->addr, addr, node) {
1105                 if (addr->enabled == enabled)
1106                         continue;
1107
1108                 if (enabled)
1109                         system_add_address(dev, addr);
1110                 else
1111                         system_del_address(dev, addr);
1112                 addr->enabled = enabled;
1113         }
1114
1115         vlist_for_each_element(&ip->route, route, node) {
1116                 bool _enabled = enabled;
1117
1118                 if (!enable_route(ip, route))
1119                         _enabled = false;
1120
1121                 if (route->enabled == _enabled)
1122                         continue;
1123
1124                 if (_enabled) {
1125                         if (!(route->flags & DEVROUTE_METRIC))
1126                                 route->metric = ip->iface->metric;
1127
1128                         system_add_route(dev, route);
1129                 } else
1130                         system_del_route(dev, route);
1131                 route->enabled = _enabled;
1132         }
1133
1134         struct device_prefix *c;
1135         struct device_prefix_assignment *a;
1136         list_for_each_entry(c, &prefixes, head)
1137                 list_for_each_entry(a, &c->assignments, head)
1138                         if (!strcmp(a->name, ip->iface->name))
1139                                 interface_set_prefix_address(a, c, ip->iface, enabled);
1140
1141         if (ip->iface && ip->iface->l3_dev.dev) {
1142                 set_ip_lo_policy(enabled, true, ip->iface);
1143                 set_ip_lo_policy(enabled, false, ip->iface);
1144
1145                 set_ip_source_policy(enabled, true, IPRULE_PRIORITY_REJECT + ip->iface->l3_dev.dev->ifindex,
1146                         NULL, 0, 0, ip->iface, "failed_policy");
1147         }
1148 }
1149
1150 void
1151 interface_ip_update_start(struct interface_ip_settings *ip)
1152 {
1153         if (ip != &ip->iface->config_ip) {
1154                 vlist_simple_update(&ip->dns_servers);
1155                 vlist_simple_update(&ip->dns_search);
1156         }
1157         vlist_update(&ip->route);
1158         vlist_update(&ip->addr);
1159         vlist_update(&ip->prefix);
1160 }
1161
1162 void
1163 interface_ip_update_complete(struct interface_ip_settings *ip)
1164 {
1165         vlist_simple_flush(&ip->dns_servers);
1166         vlist_simple_flush(&ip->dns_search);
1167         vlist_flush(&ip->route);
1168         vlist_flush(&ip->addr);
1169         vlist_flush(&ip->prefix);
1170         interface_write_resolv_conf();
1171 }
1172
1173 void
1174 interface_ip_flush(struct interface_ip_settings *ip)
1175 {
1176         if (ip == &ip->iface->proto_ip)
1177                 vlist_flush_all(&ip->iface->host_routes);
1178         vlist_simple_flush_all(&ip->dns_servers);
1179         vlist_simple_flush_all(&ip->dns_search);
1180         vlist_flush_all(&ip->route);
1181         vlist_flush_all(&ip->addr);
1182         vlist_flush_all(&ip->prefix);
1183 }
1184
1185 static void
1186 __interface_ip_init(struct interface_ip_settings *ip, struct interface *iface)
1187 {
1188         ip->iface = iface;
1189         ip->enabled = true;
1190         vlist_simple_init(&ip->dns_search, struct dns_search_domain, node);
1191         vlist_simple_init(&ip->dns_servers, struct dns_server, node);
1192         vlist_init(&ip->route, route_cmp, interface_update_proto_route);
1193         vlist_init(&ip->addr, addr_cmp, interface_update_proto_addr);
1194         vlist_init(&ip->prefix, prefix_cmp, interface_update_prefix);
1195 }
1196
1197 void
1198 interface_ip_init(struct interface *iface)
1199 {
1200         __interface_ip_init(&iface->proto_ip, iface);
1201         __interface_ip_init(&iface->config_ip, iface);
1202         vlist_init(&iface->host_routes, route_cmp, interface_update_host_route);
1203
1204 }
1205
1206 static void
1207 interface_ip_valid_until_handler(struct uloop_timeout *t)
1208 {
1209         time_t now = system_get_rtime();
1210         struct interface *iface;
1211         vlist_for_each_element(&interfaces, iface, node) {
1212                 if (iface->state != IFS_UP)
1213                         continue;
1214
1215                 struct device_addr *addr, *addrp;
1216                 struct device_route *route, *routep;
1217                 struct device_prefix *pref, *prefp;
1218
1219                 vlist_for_each_element_safe(&iface->proto_ip.addr, addr, node, addrp)
1220                         if (addr->valid_until && addr->valid_until < now)
1221                                 vlist_delete(&iface->proto_ip.addr, &addr->node);
1222
1223                 vlist_for_each_element_safe(&iface->proto_ip.route, route, node, routep)
1224                         if (route->valid_until && route->valid_until < now)
1225                                 vlist_delete(&iface->proto_ip.route, &route->node);
1226
1227                 vlist_for_each_element_safe(&iface->proto_ip.prefix, pref, node, prefp)
1228                         if (pref->valid_until && pref->valid_until < now)
1229                                 vlist_delete(&iface->proto_ip.prefix, &pref->node);
1230
1231         }
1232
1233         uloop_timeout_set(t, 1000);
1234 }
1235
1236 static void __init
1237 interface_ip_init_worker(void)
1238 {
1239         valid_until_timeout.cb = interface_ip_valid_until_handler;
1240         uloop_timeout_set(&valid_until_timeout, 1000);
1241 }