Merge branch 'netlink-align'
[strongswan.git] / src / libhydra / plugins / kernel_netlink / kernel_netlink_net.c
1 /*
2 * Copyright (C) 2008-2012 Tobias Brunner
3 * Copyright (C) 2005-2008 Martin Willi
4 * Hochschule fuer Technik Rapperswil
5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the
8 * Free Software Foundation; either version 2 of the License, or (at your
9 * option) any later version. See <http://www.fsf.org/copyleft/gpl.txt>.
10 *
11 * This program is distributed in the hope that it will be useful, but
12 * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
13 * or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
14 * for more details.
15 */
16
17 /*
18 * Copyright (C) 2010 secunet Security Networks AG
19 * Copyright (C) 2010 Thomas Egerer
20 *
21 * Permission is hereby granted, free of charge, to any person obtaining a copy
22 * of this software and associated documentation files (the "Software"), to deal
23 * in the Software without restriction, including without limitation the rights
24 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
25 * copies of the Software, and to permit persons to whom the Software is
26 * furnished to do so, subject to the following conditions:
27 *
28 * The above copyright notice and this permission notice shall be included in
29 * all copies or substantial portions of the Software.
30 *
31 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
32 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
33 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
34 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
35 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
36 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
37 * THE SOFTWARE.
38 */
39
40 #include <sys/socket.h>
41 #include <sys/utsname.h>
42 #include <linux/netlink.h>
43 #include <linux/rtnetlink.h>
44 #include <unistd.h>
45 #include <errno.h>
46 #include <net/if.h>
47
48 #include "kernel_netlink_net.h"
49 #include "kernel_netlink_shared.h"
50
51 #include <hydra.h>
52 #include <utils/debug.h>
53 #include <threading/thread.h>
54 #include <threading/mutex.h>
55 #include <threading/rwlock.h>
56 #include <threading/rwlock_condvar.h>
57 #include <threading/spinlock.h>
58 #include <collections/hashtable.h>
59 #include <collections/linked_list.h>
60 #include <processing/jobs/callback_job.h>
61
62 /** delay before firing roam events (ms) */
63 #define ROAM_DELAY 100
64
65 /** delay before reinstalling routes (ms) */
66 #define ROUTE_DELAY 100
67
68 /** maximum recursion when searching for addresses in get_route() */
69 #define MAX_ROUTE_RECURSION 2
70
71 typedef struct addr_entry_t addr_entry_t;
72
73 /**
74 * IP address in an iface_entry_t
75 */
76 struct addr_entry_t {
77
78 /** the ip address */
79 host_t *ip;
80
81 /** scope of the address */
82 u_char scope;
83
84 /** number of times this IP is used, if virtual (i.e. managed by us) */
85 u_int refcount;
86
87 /** TRUE once it is installed, if virtual */
88 bool installed;
89 };
90
91 /**
92 * destroy a addr_entry_t object
93 */
94 static void addr_entry_destroy(addr_entry_t *this)
95 {
96 this->ip->destroy(this->ip);
97 free(this);
98 }
99
100 typedef struct iface_entry_t iface_entry_t;
101
102 /**
103 * A network interface on this system, containing addr_entry_t's
104 */
105 struct iface_entry_t {
106
107 /** interface index */
108 int ifindex;
109
110 /** name of the interface */
111 char ifname[IFNAMSIZ];
112
113 /** interface flags, as in netdevice(7) SIOCGIFFLAGS */
114 u_int flags;
115
116 /** list of addresses as host_t */
117 linked_list_t *addrs;
118
119 /** TRUE if usable by config */
120 bool usable;
121 };
122
123 /**
124 * destroy an interface entry
125 */
126 static void iface_entry_destroy(iface_entry_t *this)
127 {
128 this->addrs->destroy_function(this->addrs, (void*)addr_entry_destroy);
129 free(this);
130 }
131
132 /**
133 * find an interface entry by index
134 */
135 static bool iface_entry_by_index(iface_entry_t *this, int *ifindex)
136 {
137 return this->ifindex == *ifindex;
138 }
139
140 /**
141 * find an interface entry by name
142 */
143 static bool iface_entry_by_name(iface_entry_t *this, char *ifname)
144 {
145 return streq(this->ifname, ifname);
146 }
147
148 /**
149 * check if an interface is up
150 */
151 static inline bool iface_entry_up(iface_entry_t *iface)
152 {
153 return (iface->flags & IFF_UP) == IFF_UP;
154 }
155
156 /**
157 * check if an interface is up and usable
158 */
159 static inline bool iface_entry_up_and_usable(iface_entry_t *iface)
160 {
161 return iface->usable && iface_entry_up(iface);
162 }
163
164 typedef struct addr_map_entry_t addr_map_entry_t;
165
166 /**
167 * Entry that maps an IP address to an interface entry
168 */
169 struct addr_map_entry_t {
170 /** The IP address */
171 host_t *ip;
172
173 /** The address entry for this IP address */
174 addr_entry_t *addr;
175
176 /** The interface this address is installed on */
177 iface_entry_t *iface;
178 };
179
180 /**
181 * Hash a addr_map_entry_t object, all entries with the same IP address
182 * are stored in the same bucket
183 */
184 static u_int addr_map_entry_hash(addr_map_entry_t *this)
185 {
186 return chunk_hash(this->ip->get_address(this->ip));
187 }
188
189 /**
190 * Compare two addr_map_entry_t objects, two entries are equal if they are
191 * installed on the same interface
192 */
193 static bool addr_map_entry_equals(addr_map_entry_t *a, addr_map_entry_t *b)
194 {
195 return a->iface->ifindex == b->iface->ifindex &&
196 a->ip->ip_equals(a->ip, b->ip);
197 }
198
199 /**
200 * Used with get_match this finds an address entry if it is installed on
201 * an up and usable interface
202 */
203 static bool addr_map_entry_match_up_and_usable(addr_map_entry_t *a,
204 addr_map_entry_t *b)
205 {
206 return iface_entry_up_and_usable(b->iface) &&
207 a->ip->ip_equals(a->ip, b->ip);
208 }
209
210 /**
211 * Used with get_match this finds an address entry if it is installed on
212 * any active local interface
213 */
214 static bool addr_map_entry_match_up(addr_map_entry_t *a, addr_map_entry_t *b)
215 {
216 return iface_entry_up(b->iface) && a->ip->ip_equals(a->ip, b->ip);
217 }
218
219 /**
220 * Used with get_match this finds an address entry if it is installed on
221 * any local interface
222 */
223 static bool addr_map_entry_match(addr_map_entry_t *a, addr_map_entry_t *b)
224 {
225 return a->ip->ip_equals(a->ip, b->ip);
226 }
227
228 typedef struct route_entry_t route_entry_t;
229
230 /**
231 * Installed routing entry
232 */
233 struct route_entry_t {
234 /** Name of the interface the route is bound to */
235 char *if_name;
236
237 /** Source ip of the route */
238 host_t *src_ip;
239
240 /** Gateway for this route */
241 host_t *gateway;
242
243 /** Destination net */
244 chunk_t dst_net;
245
246 /** Destination net prefixlen */
247 u_int8_t prefixlen;
248 };
249
250 /**
251 * Clone a route_entry_t object.
252 */
253 static route_entry_t *route_entry_clone(route_entry_t *this)
254 {
255 route_entry_t *route;
256
257 INIT(route,
258 .if_name = strdup(this->if_name),
259 .src_ip = this->src_ip->clone(this->src_ip),
260 .gateway = this->gateway->clone(this->gateway),
261 .dst_net = chunk_clone(this->dst_net),
262 .prefixlen = this->prefixlen,
263 );
264 return route;
265 }
266
267 /**
268 * Destroy a route_entry_t object
269 */
270 static void route_entry_destroy(route_entry_t *this)
271 {
272 free(this->if_name);
273 DESTROY_IF(this->src_ip);
274 DESTROY_IF(this->gateway);
275 chunk_free(&this->dst_net);
276 free(this);
277 }
278
279 /**
280 * Hash a route_entry_t object
281 */
282 static u_int route_entry_hash(route_entry_t *this)
283 {
284 return chunk_hash_inc(chunk_from_thing(this->prefixlen),
285 chunk_hash(this->dst_net));
286 }
287
288 /**
289 * Compare two route_entry_t objects
290 */
291 static bool route_entry_equals(route_entry_t *a, route_entry_t *b)
292 {
293 return a->if_name && b->if_name && streq(a->if_name, b->if_name) &&
294 a->src_ip->ip_equals(a->src_ip, b->src_ip) &&
295 a->gateway->ip_equals(a->gateway, b->gateway) &&
296 chunk_equals(a->dst_net, b->dst_net) && a->prefixlen == b->prefixlen;
297 }
298
299 typedef struct net_change_t net_change_t;
300
301 /**
302 * Queued network changes
303 */
304 struct net_change_t {
305 /** Name of the interface that got activated (or an IP appeared on) */
306 char *if_name;
307 };
308
309 /**
310 * Destroy a net_change_t object
311 */
312 static void net_change_destroy(net_change_t *this)
313 {
314 free(this->if_name);
315 free(this);
316 }
317
318 /**
319 * Hash a net_change_t object
320 */
321 static u_int net_change_hash(net_change_t *this)
322 {
323 return chunk_hash(chunk_create(this->if_name, strlen(this->if_name)));
324 }
325
326 /**
327 * Compare two net_change_t objects
328 */
329 static bool net_change_equals(net_change_t *a, net_change_t *b)
330 {
331 return streq(a->if_name, b->if_name);
332 }
333
334 typedef struct private_kernel_netlink_net_t private_kernel_netlink_net_t;
335
336 /**
337 * Private variables and functions of kernel_netlink_net class.
338 */
339 struct private_kernel_netlink_net_t {
340 /**
341 * Public part of the kernel_netlink_net_t object.
342 */
343 kernel_netlink_net_t public;
344
345 /**
346 * lock to access various lists and maps
347 */
348 rwlock_t *lock;
349
350 /**
351 * condition variable to signal virtual IP add/removal
352 */
353 rwlock_condvar_t *condvar;
354
355 /**
356 * Cached list of interfaces and its addresses (iface_entry_t)
357 */
358 linked_list_t *ifaces;
359
360 /**
361 * Map for IP addresses to iface_entry_t objects (addr_map_entry_t)
362 */
363 hashtable_t *addrs;
364
365 /**
366 * Map for virtual IP addresses to iface_entry_t objects (addr_map_entry_t)
367 */
368 hashtable_t *vips;
369
370 /**
371 * netlink rt socket (routing)
372 */
373 netlink_socket_t *socket;
374
375 /**
376 * Netlink rt socket to receive address change events
377 */
378 int socket_events;
379
380 /**
381 * earliest time of the next roam event
382 */
383 timeval_t next_roam;
384
385 /**
386 * lock to check and update roam event time
387 */
388 spinlock_t *roam_lock;
389
390 /**
391 * routing table to install routes
392 */
393 int routing_table;
394
395 /**
396 * priority of used routing table
397 */
398 int routing_table_prio;
399
400 /**
401 * installed routes
402 */
403 hashtable_t *routes;
404
405 /**
406 * mutex for routes
407 */
408 mutex_t *routes_lock;
409
410 /**
411 * interface changes which may trigger route reinstallation
412 */
413 hashtable_t *net_changes;
414
415 /**
416 * mutex for route reinstallation triggers
417 */
418 mutex_t *net_changes_lock;
419
420 /**
421 * time of last route reinstallation
422 */
423 timeval_t last_route_reinstall;
424
425 /**
426 * whether to react to RTM_NEWROUTE or RTM_DELROUTE events
427 */
428 bool process_route;
429
430 /**
431 * whether to actually install virtual IPs
432 */
433 bool install_virtual_ip;
434
435 /**
436 * the name of the interface virtual IP addresses are installed on
437 */
438 char *install_virtual_ip_on;
439
440 /**
441 * whether preferred source addresses can be specified for IPv6 routes
442 */
443 bool rta_prefsrc_for_ipv6;
444
445 /**
446 * list with routing tables to be excluded from route lookup
447 */
448 linked_list_t *rt_exclude;
449 };
450
451 /**
452 * Forward declaration
453 */
454 static status_t manage_srcroute(private_kernel_netlink_net_t *this,
455 int nlmsg_type, int flags, chunk_t dst_net,
456 u_int8_t prefixlen, host_t *gateway,
457 host_t *src_ip, char *if_name);
458
459 /**
460 * Clear the queued network changes.
461 */
462 static void net_changes_clear(private_kernel_netlink_net_t *this)
463 {
464 enumerator_t *enumerator;
465 net_change_t *change;
466
467 enumerator = this->net_changes->create_enumerator(this->net_changes);
468 while (enumerator->enumerate(enumerator, NULL, (void**)&change))
469 {
470 this->net_changes->remove_at(this->net_changes, enumerator);
471 net_change_destroy(change);
472 }
473 enumerator->destroy(enumerator);
474 }
475
476 /**
477 * Act upon queued network changes.
478 */
479 static job_requeue_t reinstall_routes(private_kernel_netlink_net_t *this)
480 {
481 enumerator_t *enumerator;
482 route_entry_t *route;
483
484 this->net_changes_lock->lock(this->net_changes_lock);
485 this->routes_lock->lock(this->routes_lock);
486
487 enumerator = this->routes->create_enumerator(this->routes);
488 while (enumerator->enumerate(enumerator, NULL, (void**)&route))
489 {
490 net_change_t *change, lookup = {
491 .if_name = route->if_name,
492 };
493 /* check if a change for the outgoing interface is queued */
494 change = this->net_changes->get(this->net_changes, &lookup);
495 if (!change)
496 { /* in case src_ip is not on the outgoing interface */
497 if (this->public.interface.get_interface(&this->public.interface,
498 route->src_ip, &lookup.if_name))
499 {
500 if (!streq(lookup.if_name, route->if_name))
501 {
502 change = this->net_changes->get(this->net_changes, &lookup);
503 }
504 free(lookup.if_name);
505 }
506 }
507 if (change)
508 {
509 manage_srcroute(this, RTM_NEWROUTE, NLM_F_CREATE | NLM_F_EXCL,
510 route->dst_net, route->prefixlen, route->gateway,
511 route->src_ip, route->if_name);
512 }
513 }
514 enumerator->destroy(enumerator);
515 this->routes_lock->unlock(this->routes_lock);
516
517 net_changes_clear(this);
518 this->net_changes_lock->unlock(this->net_changes_lock);
519 return JOB_REQUEUE_NONE;
520 }
521
522 /**
523 * Queue route reinstallation caused by network changes for a given interface.
524 *
525 * The route reinstallation is delayed for a while and only done once for
526 * several calls during this delay, in order to avoid doing it too often.
527 * The interface name is freed.
528 */
529 static void queue_route_reinstall(private_kernel_netlink_net_t *this,
530 char *if_name)
531 {
532 net_change_t *update, *found;
533 timeval_t now;
534 job_t *job;
535
536 INIT(update,
537 .if_name = if_name
538 );
539
540 this->net_changes_lock->lock(this->net_changes_lock);
541 found = this->net_changes->put(this->net_changes, update, update);
542 if (found)
543 {
544 net_change_destroy(found);
545 }
546 time_monotonic(&now);
547 if (timercmp(&now, &this->last_route_reinstall, >))
548 {
549 timeval_add_ms(&now, ROUTE_DELAY);
550 this->last_route_reinstall = now;
551
552 job = (job_t*)callback_job_create((callback_job_cb_t)reinstall_routes,
553 this, NULL, NULL);
554 lib->scheduler->schedule_job_ms(lib->scheduler, job, ROUTE_DELAY);
555 }
556 this->net_changes_lock->unlock(this->net_changes_lock);
557 }
558
559 /**
560 * check if the given IP is known as virtual IP and currently installed
561 *
562 * this function will also return TRUE if the virtual IP entry disappeared.
563 * in that case the returned entry will be NULL.
564 *
565 * this->lock must be held when calling this function
566 */
567 static bool is_vip_installed_or_gone(private_kernel_netlink_net_t *this,
568 host_t *ip, addr_map_entry_t **entry)
569 {
570 addr_map_entry_t lookup = {
571 .ip = ip,
572 };
573
574 *entry = this->vips->get_match(this->vips, &lookup,
575 (void*)addr_map_entry_match);
576 if (*entry == NULL)
577 { /* the virtual IP disappeared */
578 return TRUE;
579 }
580 return (*entry)->addr->installed;
581 }
582
583 /**
584 * check if the given IP is known as virtual IP
585 *
586 * this->lock must be held when calling this function
587 */
588 static bool is_known_vip(private_kernel_netlink_net_t *this, host_t *ip)
589 {
590 addr_map_entry_t lookup = {
591 .ip = ip,
592 };
593
594 return this->vips->get_match(this->vips, &lookup,
595 (void*)addr_map_entry_match) != NULL;
596 }
597
598 /**
599 * Add an address map entry
600 */
601 static void addr_map_entry_add(hashtable_t *map, addr_entry_t *addr,
602 iface_entry_t *iface)
603 {
604 addr_map_entry_t *entry;
605
606 INIT(entry,
607 .ip = addr->ip,
608 .addr = addr,
609 .iface = iface,
610 );
611 entry = map->put(map, entry, entry);
612 free(entry);
613 }
614
615 /**
616 * Remove an address map entry
617 */
618 static void addr_map_entry_remove(hashtable_t *map, addr_entry_t *addr,
619 iface_entry_t *iface)
620 {
621 addr_map_entry_t *entry, lookup = {
622 .ip = addr->ip,
623 .addr = addr,
624 .iface = iface,
625 };
626
627 entry = map->remove(map, &lookup);
628 free(entry);
629 }
630
631 /**
632 * get the first non-virtual ip address on the given interface.
633 * if a candidate address is given, we first search for that address and if not
634 * found return the address as above.
635 * returned host is a clone, has to be freed by caller.
636 *
637 * this->lock must be held when calling this function
638 */
639 static host_t *get_interface_address(private_kernel_netlink_net_t *this,
640 int ifindex, int family, host_t *candidate)
641 {
642 iface_entry_t *iface;
643 enumerator_t *addrs;
644 addr_entry_t *addr;
645 host_t *ip = NULL;
646
647 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_index,
648 (void**)&iface, &ifindex) == SUCCESS)
649 {
650 if (iface->usable)
651 { /* only use interfaces not excluded by config */
652 addrs = iface->addrs->create_enumerator(iface->addrs);
653 while (addrs->enumerate(addrs, &addr))
654 {
655 if (addr->refcount)
656 { /* ignore virtual IP addresses */
657 continue;
658 }
659 if (addr->ip->get_family(addr->ip) == family)
660 {
661 if (!candidate || candidate->ip_equals(candidate, addr->ip))
662 { /* stop at the first address if we don't search for a
663 * candidate or if the candidate matches */
664 ip = addr->ip;
665 break;
666 }
667 else if (!ip)
668 { /* store the first address as fallback if candidate is
669 * not found */
670 ip = addr->ip;
671 }
672 }
673 }
674 addrs->destroy(addrs);
675 }
676 }
677 return ip ? ip->clone(ip) : NULL;
678 }
679
680 /**
681 * callback function that raises the delayed roam event
682 */
683 static job_requeue_t roam_event(uintptr_t address)
684 {
685 hydra->kernel_interface->roam(hydra->kernel_interface, address != 0);
686 return JOB_REQUEUE_NONE;
687 }
688
689 /**
690 * fire a roaming event. we delay it for a bit and fire only one event
691 * for multiple calls. otherwise we would create too many events.
692 */
693 static void fire_roam_event(private_kernel_netlink_net_t *this, bool address)
694 {
695 timeval_t now;
696 job_t *job;
697
698 time_monotonic(&now);
699 this->roam_lock->lock(this->roam_lock);
700 if (!timercmp(&now, &this->next_roam, >))
701 {
702 this->roam_lock->unlock(this->roam_lock);
703 return;
704 }
705 timeval_add_ms(&now, ROAM_DELAY);
706 this->next_roam = now;
707 this->roam_lock->unlock(this->roam_lock);
708
709 job = (job_t*)callback_job_create((callback_job_cb_t)roam_event,
710 (void*)(uintptr_t)(address ? 1 : 0),
711 NULL, NULL);
712 lib->scheduler->schedule_job_ms(lib->scheduler, job, ROAM_DELAY);
713 }
714
715 /**
716 * check if an interface with a given index is up and usable
717 *
718 * this->lock must be locked when calling this function
719 */
720 static bool is_interface_up_and_usable(private_kernel_netlink_net_t *this,
721 int index)
722 {
723 iface_entry_t *iface;
724
725 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_index,
726 (void**)&iface, &index) == SUCCESS)
727 {
728 return iface_entry_up_and_usable(iface);
729 }
730 return FALSE;
731 }
732
733 /**
734 * unregister the current addr_entry_t from the hashtable it is stored in
735 *
736 * this->lock must be locked when calling this function
737 */
738 static void addr_entry_unregister(addr_entry_t *addr, iface_entry_t *iface,
739 private_kernel_netlink_net_t *this)
740 {
741 if (addr->refcount)
742 {
743 addr_map_entry_remove(this->vips, addr, iface);
744 this->condvar->broadcast(this->condvar);
745 return;
746 }
747 addr_map_entry_remove(this->addrs, addr, iface);
748 }
749
750 /**
751 * process RTM_NEWLINK/RTM_DELLINK from kernel
752 */
753 static void process_link(private_kernel_netlink_net_t *this,
754 struct nlmsghdr *hdr, bool event)
755 {
756 struct ifinfomsg* msg = (struct ifinfomsg*)(NLMSG_DATA(hdr));
757 struct rtattr *rta = IFLA_RTA(msg);
758 size_t rtasize = IFLA_PAYLOAD (hdr);
759 enumerator_t *enumerator;
760 iface_entry_t *current, *entry = NULL;
761 char *name = NULL;
762 bool update = FALSE, update_routes = FALSE;
763
764 while (RTA_OK(rta, rtasize))
765 {
766 switch (rta->rta_type)
767 {
768 case IFLA_IFNAME:
769 name = RTA_DATA(rta);
770 break;
771 }
772 rta = RTA_NEXT(rta, rtasize);
773 }
774 if (!name)
775 {
776 name = "(unknown)";
777 }
778
779 this->lock->write_lock(this->lock);
780 switch (hdr->nlmsg_type)
781 {
782 case RTM_NEWLINK:
783 {
784 if (this->ifaces->find_first(this->ifaces,
785 (void*)iface_entry_by_index, (void**)&entry,
786 &msg->ifi_index) != SUCCESS)
787 {
788 INIT(entry,
789 .ifindex = msg->ifi_index,
790 .addrs = linked_list_create(),
791 .usable = hydra->kernel_interface->is_interface_usable(
792 hydra->kernel_interface, name),
793 );
794 this->ifaces->insert_last(this->ifaces, entry);
795 }
796 strncpy(entry->ifname, name, IFNAMSIZ);
797 entry->ifname[IFNAMSIZ-1] = '\0';
798 if (event && entry->usable)
799 {
800 if (!(entry->flags & IFF_UP) && (msg->ifi_flags & IFF_UP))
801 {
802 update = update_routes = TRUE;
803 DBG1(DBG_KNL, "interface %s activated", name);
804 }
805 if ((entry->flags & IFF_UP) && !(msg->ifi_flags & IFF_UP))
806 {
807 update = TRUE;
808 DBG1(DBG_KNL, "interface %s deactivated", name);
809 }
810 }
811 entry->flags = msg->ifi_flags;
812 break;
813 }
814 case RTM_DELLINK:
815 {
816 enumerator = this->ifaces->create_enumerator(this->ifaces);
817 while (enumerator->enumerate(enumerator, &current))
818 {
819 if (current->ifindex == msg->ifi_index)
820 {
821 if (event && current->usable)
822 {
823 update = TRUE;
824 DBG1(DBG_KNL, "interface %s deleted", current->ifname);
825 }
826 /* TODO: move virtual IPs installed on this interface to
827 * another interface? */
828 this->ifaces->remove_at(this->ifaces, enumerator);
829 current->addrs->invoke_function(current->addrs,
830 (void*)addr_entry_unregister, current, this);
831 iface_entry_destroy(current);
832 break;
833 }
834 }
835 enumerator->destroy(enumerator);
836 break;
837 }
838 }
839 this->lock->unlock(this->lock);
840
841 if (update_routes && event)
842 {
843 queue_route_reinstall(this, strdup(name));
844 }
845
846 if (update && event)
847 {
848 fire_roam_event(this, TRUE);
849 }
850 }
851
852 /**
853 * process RTM_NEWADDR/RTM_DELADDR from kernel
854 */
855 static void process_addr(private_kernel_netlink_net_t *this,
856 struct nlmsghdr *hdr, bool event)
857 {
858 struct ifaddrmsg* msg = (struct ifaddrmsg*)(NLMSG_DATA(hdr));
859 struct rtattr *rta = IFA_RTA(msg);
860 size_t rtasize = IFA_PAYLOAD (hdr);
861 host_t *host = NULL;
862 iface_entry_t *iface;
863 chunk_t local = chunk_empty, address = chunk_empty;
864 char *route_ifname = NULL;
865 bool update = FALSE, found = FALSE, changed = FALSE;
866
867 while (RTA_OK(rta, rtasize))
868 {
869 switch (rta->rta_type)
870 {
871 case IFA_LOCAL:
872 local.ptr = RTA_DATA(rta);
873 local.len = RTA_PAYLOAD(rta);
874 break;
875 case IFA_ADDRESS:
876 address.ptr = RTA_DATA(rta);
877 address.len = RTA_PAYLOAD(rta);
878 break;
879 }
880 rta = RTA_NEXT(rta, rtasize);
881 }
882
883 /* For PPP interfaces, we need the IFA_LOCAL address,
884 * IFA_ADDRESS is the peers address. But IFA_LOCAL is
885 * not included in all cases (IPv6?), so fallback to IFA_ADDRESS. */
886 if (local.ptr)
887 {
888 host = host_create_from_chunk(msg->ifa_family, local, 0);
889 }
890 else if (address.ptr)
891 {
892 host = host_create_from_chunk(msg->ifa_family, address, 0);
893 }
894
895 if (host == NULL)
896 { /* bad family? */
897 return;
898 }
899
900 this->lock->write_lock(this->lock);
901 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_index,
902 (void**)&iface, &msg->ifa_index) == SUCCESS)
903 {
904 addr_map_entry_t *entry, lookup = {
905 .ip = host,
906 .iface = iface,
907 };
908 addr_entry_t *addr;
909
910 entry = this->vips->get(this->vips, &lookup);
911 if (entry)
912 {
913 if (hdr->nlmsg_type == RTM_NEWADDR)
914 { /* mark as installed and signal waiting threads */
915 entry->addr->installed = TRUE;
916 }
917 else
918 { /* the address was already marked as uninstalled */
919 addr = entry->addr;
920 iface->addrs->remove(iface->addrs, addr, NULL);
921 addr_map_entry_remove(this->vips, addr, iface);
922 addr_entry_destroy(addr);
923 }
924 /* no roam events etc. for virtual IPs */
925 this->condvar->broadcast(this->condvar);
926 this->lock->unlock(this->lock);
927 host->destroy(host);
928 return;
929 }
930 entry = this->addrs->get(this->addrs, &lookup);
931 if (entry)
932 {
933 if (hdr->nlmsg_type == RTM_DELADDR)
934 {
935 found = TRUE;
936 addr = entry->addr;
937 iface->addrs->remove(iface->addrs, addr, NULL);
938 if (iface->usable)
939 {
940 changed = TRUE;
941 DBG1(DBG_KNL, "%H disappeared from %s", host,
942 iface->ifname);
943 }
944 addr_map_entry_remove(this->addrs, addr, iface);
945 addr_entry_destroy(addr);
946 }
947 }
948 else
949 {
950 if (hdr->nlmsg_type == RTM_NEWADDR)
951 {
952 found = TRUE;
953 changed = TRUE;
954 route_ifname = strdup(iface->ifname);
955 INIT(addr,
956 .ip = host->clone(host),
957 .scope = msg->ifa_scope,
958 );
959 iface->addrs->insert_last(iface->addrs, addr);
960 addr_map_entry_add(this->addrs, addr, iface);
961 if (event && iface->usable)
962 {
963 DBG1(DBG_KNL, "%H appeared on %s", host, iface->ifname);
964 }
965 }
966 }
967 if (found && (iface->flags & IFF_UP))
968 {
969 update = TRUE;
970 }
971 if (!iface->usable)
972 { /* ignore events for interfaces excluded by config */
973 update = changed = FALSE;
974 }
975 }
976 this->lock->unlock(this->lock);
977
978 if (update && event && route_ifname)
979 {
980 queue_route_reinstall(this, route_ifname);
981 }
982 else
983 {
984 free(route_ifname);
985 }
986 host->destroy(host);
987
988 /* send an update to all IKE_SAs */
989 if (update && event && changed)
990 {
991 fire_roam_event(this, TRUE);
992 }
993 }
994
995 /**
996 * process RTM_NEWROUTE and RTM_DELROUTE from kernel
997 */
998 static void process_route(private_kernel_netlink_net_t *this, struct nlmsghdr *hdr)
999 {
1000 struct rtmsg* msg = (struct rtmsg*)(NLMSG_DATA(hdr));
1001 struct rtattr *rta = RTM_RTA(msg);
1002 size_t rtasize = RTM_PAYLOAD(hdr);
1003 u_int32_t rta_oif = 0;
1004 host_t *host = NULL;
1005
1006 /* ignore routes added by us or in the local routing table (local addrs) */
1007 if (msg->rtm_table && (msg->rtm_table == this->routing_table ||
1008 msg->rtm_table == RT_TABLE_LOCAL))
1009 {
1010 return;
1011 }
1012 else if (msg->rtm_flags & RTM_F_CLONED)
1013 { /* ignore cached routes, seem to be created a lot for IPv6 */
1014 return;
1015 }
1016
1017 while (RTA_OK(rta, rtasize))
1018 {
1019 switch (rta->rta_type)
1020 {
1021 case RTA_PREFSRC:
1022 DESTROY_IF(host);
1023 host = host_create_from_chunk(msg->rtm_family,
1024 chunk_create(RTA_DATA(rta), RTA_PAYLOAD(rta)), 0);
1025 break;
1026 case RTA_OIF:
1027 if (RTA_PAYLOAD(rta) == sizeof(rta_oif))
1028 {
1029 rta_oif = *(u_int32_t*)RTA_DATA(rta);
1030 }
1031 break;
1032 }
1033 rta = RTA_NEXT(rta, rtasize);
1034 }
1035 this->lock->read_lock(this->lock);
1036 if (rta_oif && !is_interface_up_and_usable(this, rta_oif))
1037 { /* ignore route changes for interfaces that are ignored or down */
1038 this->lock->unlock(this->lock);
1039 DESTROY_IF(host);
1040 return;
1041 }
1042 if (!host && rta_oif)
1043 {
1044 host = get_interface_address(this, rta_oif, msg->rtm_family, NULL);
1045 }
1046 if (!host || is_known_vip(this, host))
1047 { /* ignore routes added for virtual IPs */
1048 this->lock->unlock(this->lock);
1049 DESTROY_IF(host);
1050 return;
1051 }
1052 this->lock->unlock(this->lock);
1053 fire_roam_event(this, FALSE);
1054 host->destroy(host);
1055 }
1056
1057 /**
1058 * Receives events from kernel
1059 */
1060 static job_requeue_t receive_events(private_kernel_netlink_net_t *this)
1061 {
1062 char response[1024];
1063 struct nlmsghdr *hdr = (struct nlmsghdr*)response;
1064 struct sockaddr_nl addr;
1065 socklen_t addr_len = sizeof(addr);
1066 int len;
1067 bool oldstate;
1068
1069 oldstate = thread_cancelability(TRUE);
1070 len = recvfrom(this->socket_events, response, sizeof(response), 0,
1071 (struct sockaddr*)&addr, &addr_len);
1072 thread_cancelability(oldstate);
1073
1074 if (len < 0)
1075 {
1076 switch (errno)
1077 {
1078 case EINTR:
1079 /* interrupted, try again */
1080 return JOB_REQUEUE_DIRECT;
1081 case EAGAIN:
1082 /* no data ready, select again */
1083 return JOB_REQUEUE_DIRECT;
1084 default:
1085 DBG1(DBG_KNL, "unable to receive from rt event socket");
1086 sleep(1);
1087 return JOB_REQUEUE_FAIR;
1088 }
1089 }
1090
1091 if (addr.nl_pid != 0)
1092 { /* not from kernel. not interested, try another one */
1093 return JOB_REQUEUE_DIRECT;
1094 }
1095
1096 while (NLMSG_OK(hdr, len))
1097 {
1098 /* looks good so far, dispatch netlink message */
1099 switch (hdr->nlmsg_type)
1100 {
1101 case RTM_NEWADDR:
1102 case RTM_DELADDR:
1103 process_addr(this, hdr, TRUE);
1104 break;
1105 case RTM_NEWLINK:
1106 case RTM_DELLINK:
1107 process_link(this, hdr, TRUE);
1108 break;
1109 case RTM_NEWROUTE:
1110 case RTM_DELROUTE:
1111 if (this->process_route)
1112 {
1113 process_route(this, hdr);
1114 }
1115 break;
1116 default:
1117 break;
1118 }
1119 hdr = NLMSG_NEXT(hdr, len);
1120 }
1121 return JOB_REQUEUE_DIRECT;
1122 }
1123
1124 /** enumerator over addresses */
1125 typedef struct {
1126 private_kernel_netlink_net_t* this;
1127 /** which addresses to enumerate */
1128 kernel_address_type_t which;
1129 } address_enumerator_t;
1130
1131 /**
1132 * cleanup function for address enumerator
1133 */
1134 static void address_enumerator_destroy(address_enumerator_t *data)
1135 {
1136 data->this->lock->unlock(data->this->lock);
1137 free(data);
1138 }
1139
1140 /**
1141 * filter for addresses
1142 */
1143 static bool filter_addresses(address_enumerator_t *data,
1144 addr_entry_t** in, host_t** out)
1145 {
1146 if (!(data->which & ADDR_TYPE_VIRTUAL) && (*in)->refcount)
1147 { /* skip virtual interfaces added by us */
1148 return FALSE;
1149 }
1150 if ((*in)->scope >= RT_SCOPE_LINK)
1151 { /* skip addresses with a unusable scope */
1152 return FALSE;
1153 }
1154 *out = (*in)->ip;
1155 return TRUE;
1156 }
1157
1158 /**
1159 * enumerator constructor for interfaces
1160 */
1161 static enumerator_t *create_iface_enumerator(iface_entry_t *iface,
1162 address_enumerator_t *data)
1163 {
1164 return enumerator_create_filter(
1165 iface->addrs->create_enumerator(iface->addrs),
1166 (void*)filter_addresses, data, NULL);
1167 }
1168
1169 /**
1170 * filter for interfaces
1171 */
1172 static bool filter_interfaces(address_enumerator_t *data, iface_entry_t** in,
1173 iface_entry_t** out)
1174 {
1175 if (!(data->which & ADDR_TYPE_IGNORED) && !(*in)->usable)
1176 { /* skip interfaces excluded by config */
1177 return FALSE;
1178 }
1179 if (!(data->which & ADDR_TYPE_LOOPBACK) && ((*in)->flags & IFF_LOOPBACK))
1180 { /* ignore loopback devices */
1181 return FALSE;
1182 }
1183 if (!(data->which & ADDR_TYPE_DOWN) && !((*in)->flags & IFF_UP))
1184 { /* skip interfaces not up */
1185 return FALSE;
1186 }
1187 *out = *in;
1188 return TRUE;
1189 }
1190
1191 METHOD(kernel_net_t, create_address_enumerator, enumerator_t*,
1192 private_kernel_netlink_net_t *this, kernel_address_type_t which)
1193 {
1194 address_enumerator_t *data = malloc_thing(address_enumerator_t);
1195 data->this = this;
1196 data->which = which;
1197
1198 this->lock->read_lock(this->lock);
1199 return enumerator_create_nested(
1200 enumerator_create_filter(
1201 this->ifaces->create_enumerator(this->ifaces),
1202 (void*)filter_interfaces, data, NULL),
1203 (void*)create_iface_enumerator, data,
1204 (void*)address_enumerator_destroy);
1205 }
1206
1207 METHOD(kernel_net_t, get_interface_name, bool,
1208 private_kernel_netlink_net_t *this, host_t* ip, char **name)
1209 {
1210 addr_map_entry_t *entry, lookup = {
1211 .ip = ip,
1212 };
1213
1214 if (ip->is_anyaddr(ip))
1215 {
1216 return FALSE;
1217 }
1218 this->lock->read_lock(this->lock);
1219 /* first try to find it on an up and usable interface */
1220 entry = this->addrs->get_match(this->addrs, &lookup,
1221 (void*)addr_map_entry_match_up_and_usable);
1222 if (entry)
1223 {
1224 if (name)
1225 {
1226 *name = strdup(entry->iface->ifname);
1227 DBG2(DBG_KNL, "%H is on interface %s", ip, *name);
1228 }
1229 this->lock->unlock(this->lock);
1230 return TRUE;
1231 }
1232 /* in a second step, consider virtual IPs installed by us */
1233 entry = this->vips->get_match(this->vips, &lookup,
1234 (void*)addr_map_entry_match_up_and_usable);
1235 if (entry)
1236 {
1237 if (name)
1238 {
1239 *name = strdup(entry->iface->ifname);
1240 DBG2(DBG_KNL, "virtual %H is on interface %s", ip, *name);
1241 }
1242 this->lock->unlock(this->lock);
1243 return TRUE;
1244 }
1245 /* maybe it is installed on an ignored interface */
1246 entry = this->addrs->get_match(this->addrs, &lookup,
1247 (void*)addr_map_entry_match_up);
1248 if (!entry)
1249 {
1250 DBG2(DBG_KNL, "%H is not a local address or the interface is down", ip);
1251 }
1252 this->lock->unlock(this->lock);
1253 return FALSE;
1254 }
1255
1256 /**
1257 * get the index of an interface by name
1258 */
1259 static int get_interface_index(private_kernel_netlink_net_t *this, char* name)
1260 {
1261 iface_entry_t *iface;
1262 int ifindex = 0;
1263
1264 DBG2(DBG_KNL, "getting iface index for %s", name);
1265
1266 this->lock->read_lock(this->lock);
1267 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_name,
1268 (void**)&iface, name) == SUCCESS)
1269 {
1270 ifindex = iface->ifindex;
1271 }
1272 this->lock->unlock(this->lock);
1273
1274 if (ifindex == 0)
1275 {
1276 DBG1(DBG_KNL, "unable to get interface index for %s", name);
1277 }
1278 return ifindex;
1279 }
1280
1281 /**
1282 * check if an address (chunk) addr is in subnet (net with net_len net bits)
1283 */
1284 static bool addr_in_subnet(chunk_t addr, chunk_t net, int net_len)
1285 {
1286 static const u_char mask[] = { 0x00, 0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe };
1287 int byte = 0;
1288
1289 if (net_len == 0)
1290 { /* any address matches a /0 network */
1291 return TRUE;
1292 }
1293 if (addr.len != net.len || net_len > 8 * net.len )
1294 {
1295 return FALSE;
1296 }
1297 /* scan through all bytes in network order */
1298 while (net_len > 0)
1299 {
1300 if (net_len < 8)
1301 {
1302 return (mask[net_len] & addr.ptr[byte]) == (mask[net_len] & net.ptr[byte]);
1303 }
1304 else
1305 {
1306 if (addr.ptr[byte] != net.ptr[byte])
1307 {
1308 return FALSE;
1309 }
1310 byte++;
1311 net_len -= 8;
1312 }
1313 }
1314 return TRUE;
1315 }
1316
1317 /**
1318 * Store information about a route retrieved via RTNETLINK
1319 */
1320 typedef struct {
1321 chunk_t gtw;
1322 chunk_t src;
1323 chunk_t dst;
1324 host_t *src_host;
1325 u_int8_t dst_len;
1326 u_int32_t table;
1327 u_int32_t oif;
1328 } rt_entry_t;
1329
1330 /**
1331 * Free a route entry
1332 */
1333 static void rt_entry_destroy(rt_entry_t *this)
1334 {
1335 DESTROY_IF(this->src_host);
1336 free(this);
1337 }
1338
1339 /**
1340 * Parse route received with RTM_NEWROUTE. The given rt_entry_t object will be
1341 * reused if not NULL.
1342 *
1343 * Returned chunks point to internal data of the Netlink message.
1344 */
1345 static rt_entry_t *parse_route(struct nlmsghdr *hdr, rt_entry_t *route)
1346 {
1347 struct rtattr *rta;
1348 struct rtmsg *msg;
1349 size_t rtasize;
1350
1351 msg = (struct rtmsg*)(NLMSG_DATA(hdr));
1352 rta = RTM_RTA(msg);
1353 rtasize = RTM_PAYLOAD(hdr);
1354
1355 if (route)
1356 {
1357 route->gtw = chunk_empty;
1358 route->src = chunk_empty;
1359 route->dst = chunk_empty;
1360 route->dst_len = msg->rtm_dst_len;
1361 route->table = msg->rtm_table;
1362 route->oif = 0;
1363 }
1364 else
1365 {
1366 INIT(route,
1367 .dst_len = msg->rtm_dst_len,
1368 .table = msg->rtm_table,
1369 );
1370 }
1371
1372 while (RTA_OK(rta, rtasize))
1373 {
1374 switch (rta->rta_type)
1375 {
1376 case RTA_PREFSRC:
1377 route->src = chunk_create(RTA_DATA(rta), RTA_PAYLOAD(rta));
1378 break;
1379 case RTA_GATEWAY:
1380 route->gtw = chunk_create(RTA_DATA(rta), RTA_PAYLOAD(rta));
1381 break;
1382 case RTA_DST:
1383 route->dst = chunk_create(RTA_DATA(rta), RTA_PAYLOAD(rta));
1384 break;
1385 case RTA_OIF:
1386 if (RTA_PAYLOAD(rta) == sizeof(route->oif))
1387 {
1388 route->oif = *(u_int32_t*)RTA_DATA(rta);
1389 }
1390 break;
1391 #ifdef HAVE_RTA_TABLE
1392 case RTA_TABLE:
1393 if (RTA_PAYLOAD(rta) == sizeof(route->table))
1394 {
1395 route->table = *(u_int32_t*)RTA_DATA(rta);
1396 }
1397 break;
1398 #endif /* HAVE_RTA_TABLE*/
1399 }
1400 rta = RTA_NEXT(rta, rtasize);
1401 }
1402 return route;
1403 }
1404
1405 /**
1406 * Get a route: If "nexthop", the nexthop is returned. source addr otherwise.
1407 */
1408 static host_t *get_route(private_kernel_netlink_net_t *this, host_t *dest,
1409 bool nexthop, host_t *candidate, u_int recursion)
1410 {
1411 netlink_buf_t request;
1412 struct nlmsghdr *hdr, *out, *current;
1413 struct rtmsg *msg;
1414 chunk_t chunk;
1415 size_t len;
1416 linked_list_t *routes;
1417 rt_entry_t *route = NULL, *best = NULL;
1418 enumerator_t *enumerator;
1419 host_t *addr = NULL;
1420
1421 if (recursion > MAX_ROUTE_RECURSION)
1422 {
1423 return NULL;
1424 }
1425
1426 memset(&request, 0, sizeof(request));
1427
1428 hdr = (struct nlmsghdr*)request;
1429 hdr->nlmsg_flags = NLM_F_REQUEST;
1430 if (dest->get_family(dest) == AF_INET || this->rta_prefsrc_for_ipv6 ||
1431 this->routing_table)
1432 { /* kernels prior to 3.0 do not support RTA_PREFSRC for IPv6 routes.
1433 * as we want to ignore routes with virtual IPs we cannot use DUMP
1434 * if these routes are not installed in a separate table */
1435 hdr->nlmsg_flags |= NLM_F_DUMP;
1436 }
1437 hdr->nlmsg_type = RTM_GETROUTE;
1438 hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
1439
1440 msg = (struct rtmsg*)NLMSG_DATA(hdr);
1441 msg->rtm_family = dest->get_family(dest);
1442 if (candidate)
1443 {
1444 chunk = candidate->get_address(candidate);
1445 netlink_add_attribute(hdr, RTA_PREFSRC, chunk, sizeof(request));
1446 }
1447 chunk = dest->get_address(dest);
1448 netlink_add_attribute(hdr, RTA_DST, chunk, sizeof(request));
1449
1450 if (this->socket->send(this->socket, hdr, &out, &len) != SUCCESS)
1451 {
1452 DBG2(DBG_KNL, "getting %s to reach %H failed",
1453 nexthop ? "nexthop" : "address", dest);
1454 return NULL;
1455 }
1456 routes = linked_list_create();
1457 this->lock->read_lock(this->lock);
1458
1459 for (current = out; NLMSG_OK(current, len);
1460 current = NLMSG_NEXT(current, len))
1461 {
1462 switch (current->nlmsg_type)
1463 {
1464 case NLMSG_DONE:
1465 break;
1466 case RTM_NEWROUTE:
1467 {
1468 rt_entry_t *other;
1469 uintptr_t table;
1470
1471 route = parse_route(current, route);
1472
1473 table = (uintptr_t)route->table;
1474 if (this->rt_exclude->find_first(this->rt_exclude, NULL,
1475 (void**)&table) == SUCCESS)
1476 { /* route is from an excluded routing table */
1477 continue;
1478 }
1479 if (this->routing_table != 0 &&
1480 route->table == this->routing_table)
1481 { /* route is from our own ipsec routing table */
1482 continue;
1483 }
1484 if (route->oif && !is_interface_up_and_usable(this, route->oif))
1485 { /* interface is down */
1486 continue;
1487 }
1488 if (!addr_in_subnet(chunk, route->dst, route->dst_len))
1489 { /* route destination does not contain dest */
1490 continue;
1491 }
1492 if (route->src.ptr)
1493 { /* verify source address, if any */
1494 host_t *src = host_create_from_chunk(msg->rtm_family,
1495 route->src, 0);
1496 if (src && is_known_vip(this, src))
1497 { /* ignore routes installed by us */
1498 src->destroy(src);
1499 continue;
1500 }
1501 route->src_host = src;
1502 }
1503 /* insert route, sorted by decreasing network prefix */
1504 enumerator = routes->create_enumerator(routes);
1505 while (enumerator->enumerate(enumerator, &other))
1506 {
1507 if (route->dst_len > other->dst_len)
1508 {
1509 break;
1510 }
1511 }
1512 routes->insert_before(routes, enumerator, route);
1513 enumerator->destroy(enumerator);
1514 route = NULL;
1515 continue;
1516 }
1517 default:
1518 continue;
1519 }
1520 break;
1521 }
1522 if (route)
1523 {
1524 rt_entry_destroy(route);
1525 }
1526
1527 /* now we have a list of routes matching dest, sorted by net prefix.
1528 * we will look for source addresses for these routes and select the one
1529 * with the preferred source address, if possible */
1530 enumerator = routes->create_enumerator(routes);
1531 while (enumerator->enumerate(enumerator, &route))
1532 {
1533 if (route->src_host)
1534 { /* got a source address with the route, if no preferred source
1535 * is given or it matches we are done, as this is the best route */
1536 if (!candidate || candidate->ip_equals(candidate, route->src_host))
1537 {
1538 best = route;
1539 break;
1540 }
1541 else if (route->oif)
1542 { /* no match yet, maybe it is assigned to the same interface */
1543 host_t *src = get_interface_address(this, route->oif,
1544 msg->rtm_family, candidate);
1545 if (src && src->ip_equals(src, candidate))
1546 {
1547 route->src_host->destroy(route->src_host);
1548 route->src_host = src;
1549 best = route;
1550 break;
1551 }
1552 DESTROY_IF(src);
1553 }
1554 /* no luck yet with the source address. if this is the best (first)
1555 * route we store it as fallback in case we don't find a route with
1556 * the preferred source */
1557 best = best ?: route;
1558 continue;
1559 }
1560 if (route->oif)
1561 { /* no src, but an interface - get address from it */
1562 route->src_host = get_interface_address(this, route->oif,
1563 msg->rtm_family, candidate);
1564 if (route->src_host)
1565 { /* we handle this address the same as the one above */
1566 if (!candidate ||
1567 candidate->ip_equals(candidate, route->src_host))
1568 {
1569 best = route;
1570 break;
1571 }
1572 best = best ?: route;
1573 continue;
1574 }
1575 }
1576 if (route->gtw.ptr)
1577 { /* no src, no iface, but a gateway - lookup src to reach gtw */
1578 host_t *gtw;
1579
1580 gtw = host_create_from_chunk(msg->rtm_family, route->gtw, 0);
1581 if (gtw && !gtw->ip_equals(gtw, dest))
1582 {
1583 route->src_host = get_route(this, gtw, FALSE, candidate,
1584 recursion + 1);
1585 }
1586 DESTROY_IF(gtw);
1587 if (route->src_host)
1588 { /* more of the same */
1589 if (!candidate ||
1590 candidate->ip_equals(candidate, route->src_host))
1591 {
1592 best = route;
1593 break;
1594 }
1595 best = best ?: route;
1596 }
1597 }
1598 }
1599 enumerator->destroy(enumerator);
1600
1601 if (nexthop)
1602 { /* nexthop lookup, return gateway if any */
1603 if (best || routes->get_first(routes, (void**)&best) == SUCCESS)
1604 {
1605 addr = host_create_from_chunk(msg->rtm_family, best->gtw, 0);
1606 }
1607 addr = addr ?: dest->clone(dest);
1608 }
1609 else
1610 {
1611 if (best)
1612 {
1613 addr = best->src_host->clone(best->src_host);
1614 }
1615 }
1616 this->lock->unlock(this->lock);
1617 routes->destroy_function(routes, (void*)rt_entry_destroy);
1618 free(out);
1619
1620 if (addr)
1621 {
1622 DBG2(DBG_KNL, "using %H as %s to reach %H", addr,
1623 nexthop ? "nexthop" : "address", dest);
1624 }
1625 else if (!recursion)
1626 {
1627 DBG2(DBG_KNL, "no %s found to reach %H",
1628 nexthop ? "nexthop" : "address", dest);
1629 }
1630 return addr;
1631 }
1632
1633 METHOD(kernel_net_t, get_source_addr, host_t*,
1634 private_kernel_netlink_net_t *this, host_t *dest, host_t *src)
1635 {
1636 return get_route(this, dest, FALSE, src, 0);
1637 }
1638
1639 METHOD(kernel_net_t, get_nexthop, host_t*,
1640 private_kernel_netlink_net_t *this, host_t *dest, host_t *src)
1641 {
1642 return get_route(this, dest, TRUE, src, 0);
1643 }
1644
1645 /**
1646 * Manages the creation and deletion of ip addresses on an interface.
1647 * By setting the appropriate nlmsg_type, the ip will be set or unset.
1648 */
1649 static status_t manage_ipaddr(private_kernel_netlink_net_t *this, int nlmsg_type,
1650 int flags, int if_index, host_t *ip, int prefix)
1651 {
1652 netlink_buf_t request;
1653 struct nlmsghdr *hdr;
1654 struct ifaddrmsg *msg;
1655 chunk_t chunk;
1656
1657 memset(&request, 0, sizeof(request));
1658
1659 chunk = ip->get_address(ip);
1660
1661 hdr = (struct nlmsghdr*)request;
1662 hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK | flags;
1663 hdr->nlmsg_type = nlmsg_type;
1664 hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg));
1665
1666 msg = (struct ifaddrmsg*)NLMSG_DATA(hdr);
1667 msg->ifa_family = ip->get_family(ip);
1668 msg->ifa_flags = 0;
1669 msg->ifa_prefixlen = prefix < 0 ? chunk.len * 8 : prefix;
1670 msg->ifa_scope = RT_SCOPE_UNIVERSE;
1671 msg->ifa_index = if_index;
1672
1673 netlink_add_attribute(hdr, IFA_LOCAL, chunk, sizeof(request));
1674
1675 return this->socket->send_ack(this->socket, hdr);
1676 }
1677
1678 METHOD(kernel_net_t, add_ip, status_t,
1679 private_kernel_netlink_net_t *this, host_t *virtual_ip, int prefix,
1680 char *iface_name)
1681 {
1682 addr_map_entry_t *entry, lookup = {
1683 .ip = virtual_ip,
1684 };
1685 iface_entry_t *iface = NULL;
1686
1687 if (!this->install_virtual_ip)
1688 { /* disabled by config */
1689 return SUCCESS;
1690 }
1691
1692 this->lock->write_lock(this->lock);
1693 /* the virtual IP might actually be installed as regular IP, in which case
1694 * we don't track it as virtual IP */
1695 entry = this->addrs->get_match(this->addrs, &lookup,
1696 (void*)addr_map_entry_match);
1697 if (!entry)
1698 { /* otherwise it might already be installed as virtual IP */
1699 entry = this->vips->get_match(this->vips, &lookup,
1700 (void*)addr_map_entry_match);
1701 if (entry)
1702 { /* the vip we found can be in one of three states: 1) installed and
1703 * ready, 2) just added by another thread, but not yet confirmed to
1704 * be installed by the kernel, 3) just deleted, but not yet gone.
1705 * Then while we wait below, several things could happen (as we
1706 * release the lock). For instance, the interface could disappear,
1707 * or the IP is finally deleted, and it reappears on a different
1708 * interface. All these cases are handled by the call below. */
1709 while (!is_vip_installed_or_gone(this, virtual_ip, &entry))
1710 {
1711 this->condvar->wait(this->condvar, this->lock);
1712 }
1713 if (entry)
1714 {
1715 entry->addr->refcount++;
1716 }
1717 }
1718 }
1719 if (entry)
1720 {
1721 DBG2(DBG_KNL, "virtual IP %H is already installed on %s", virtual_ip,
1722 entry->iface->ifname);
1723 this->lock->unlock(this->lock);
1724 return SUCCESS;
1725 }
1726 /* try to find the target interface, either by config or via src ip */
1727 if (!this->install_virtual_ip_on ||
1728 this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_name,
1729 (void**)&iface, this->install_virtual_ip_on) != SUCCESS)
1730 {
1731 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_name,
1732 (void**)&iface, iface_name) != SUCCESS)
1733 { /* if we don't find the requested interface we just use the first */
1734 this->ifaces->get_first(this->ifaces, (void**)&iface);
1735 }
1736 }
1737 if (iface)
1738 {
1739 addr_entry_t *addr;
1740
1741 INIT(addr,
1742 .ip = virtual_ip->clone(virtual_ip),
1743 .refcount = 1,
1744 .scope = RT_SCOPE_UNIVERSE,
1745 );
1746 iface->addrs->insert_last(iface->addrs, addr);
1747 addr_map_entry_add(this->vips, addr, iface);
1748 if (manage_ipaddr(this, RTM_NEWADDR, NLM_F_CREATE | NLM_F_EXCL,
1749 iface->ifindex, virtual_ip, prefix) == SUCCESS)
1750 {
1751 while (!is_vip_installed_or_gone(this, virtual_ip, &entry))
1752 { /* wait until address appears */
1753 this->condvar->wait(this->condvar, this->lock);
1754 }
1755 if (entry)
1756 { /* we fail if the interface got deleted in the meantime */
1757 DBG2(DBG_KNL, "virtual IP %H installed on %s", virtual_ip,
1758 entry->iface->ifname);
1759 this->lock->unlock(this->lock);
1760 /* during IKEv1 reauthentication, children get moved from
1761 * old the new SA before the virtual IP is available. This
1762 * kills the route for our virtual IP, reinstall. */
1763 queue_route_reinstall(this, strdup(entry->iface->ifname));
1764 return SUCCESS;
1765 }
1766 }
1767 this->lock->unlock(this->lock);
1768 DBG1(DBG_KNL, "adding virtual IP %H failed", virtual_ip);
1769 return FAILED;
1770 }
1771 this->lock->unlock(this->lock);
1772 DBG1(DBG_KNL, "no interface available, unable to install virtual IP %H",
1773 virtual_ip);
1774 return FAILED;
1775 }
1776
1777 METHOD(kernel_net_t, del_ip, status_t,
1778 private_kernel_netlink_net_t *this, host_t *virtual_ip, int prefix,
1779 bool wait)
1780 {
1781 addr_map_entry_t *entry, lookup = {
1782 .ip = virtual_ip,
1783 };
1784
1785 if (!this->install_virtual_ip)
1786 { /* disabled by config */
1787 return SUCCESS;
1788 }
1789
1790 DBG2(DBG_KNL, "deleting virtual IP %H", virtual_ip);
1791
1792 this->lock->write_lock(this->lock);
1793 entry = this->vips->get_match(this->vips, &lookup,
1794 (void*)addr_map_entry_match);
1795 if (!entry)
1796 { /* we didn't install this IP as virtual IP */
1797 entry = this->addrs->get_match(this->addrs, &lookup,
1798 (void*)addr_map_entry_match);
1799 if (entry)
1800 {
1801 DBG2(DBG_KNL, "not deleting existing IP %H on %s", virtual_ip,
1802 entry->iface->ifname);
1803 this->lock->unlock(this->lock);
1804 return SUCCESS;
1805 }
1806 DBG2(DBG_KNL, "virtual IP %H not cached, unable to delete", virtual_ip);
1807 this->lock->unlock(this->lock);
1808 return FAILED;
1809 }
1810 if (entry->addr->refcount == 1)
1811 {
1812 status_t status;
1813
1814 /* we set this flag so that threads calling add_ip will block and wait
1815 * until the entry is gone, also so we can wait below */
1816 entry->addr->installed = FALSE;
1817 status = manage_ipaddr(this, RTM_DELADDR, 0, entry->iface->ifindex,
1818 virtual_ip, prefix);
1819 if (status == SUCCESS && wait)
1820 { /* wait until the address is really gone */
1821 while (is_known_vip(this, virtual_ip))
1822 {
1823 this->condvar->wait(this->condvar, this->lock);
1824 }
1825 }
1826 this->lock->unlock(this->lock);
1827 return status;
1828 }
1829 else
1830 {
1831 entry->addr->refcount--;
1832 }
1833 DBG2(DBG_KNL, "virtual IP %H used by other SAs, not deleting",
1834 virtual_ip);
1835 this->lock->unlock(this->lock);
1836 return SUCCESS;
1837 }
1838
1839 /**
1840 * Manages source routes in the routing table.
1841 * By setting the appropriate nlmsg_type, the route gets added or removed.
1842 */
1843 static status_t manage_srcroute(private_kernel_netlink_net_t *this,
1844 int nlmsg_type, int flags, chunk_t dst_net,
1845 u_int8_t prefixlen, host_t *gateway,
1846 host_t *src_ip, char *if_name)
1847 {
1848 netlink_buf_t request;
1849 struct nlmsghdr *hdr;
1850 struct rtmsg *msg;
1851 int ifindex;
1852 chunk_t chunk;
1853
1854 /* if route is 0.0.0.0/0, we can't install it, as it would
1855 * overwrite the default route. Instead, we add two routes:
1856 * 0.0.0.0/1 and 128.0.0.0/1 */
1857 if (this->routing_table == 0 && prefixlen == 0)
1858 {
1859 chunk_t half_net;
1860 u_int8_t half_prefixlen;
1861 status_t status;
1862
1863 half_net = chunk_alloca(dst_net.len);
1864 memset(half_net.ptr, 0, half_net.len);
1865 half_prefixlen = 1;
1866
1867 status = manage_srcroute(this, nlmsg_type, flags, half_net, half_prefixlen,
1868 gateway, src_ip, if_name);
1869 half_net.ptr[0] |= 0x80;
1870 status = manage_srcroute(this, nlmsg_type, flags, half_net, half_prefixlen,
1871 gateway, src_ip, if_name);
1872 return status;
1873 }
1874
1875 memset(&request, 0, sizeof(request));
1876
1877 hdr = (struct nlmsghdr*)request;
1878 hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK | flags;
1879 hdr->nlmsg_type = nlmsg_type;
1880 hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
1881
1882 msg = (struct rtmsg*)NLMSG_DATA(hdr);
1883 msg->rtm_family = src_ip->get_family(src_ip);
1884 msg->rtm_dst_len = prefixlen;
1885 msg->rtm_table = this->routing_table;
1886 msg->rtm_protocol = RTPROT_STATIC;
1887 msg->rtm_type = RTN_UNICAST;
1888 msg->rtm_scope = RT_SCOPE_UNIVERSE;
1889
1890 netlink_add_attribute(hdr, RTA_DST, dst_net, sizeof(request));
1891 chunk = src_ip->get_address(src_ip);
1892 netlink_add_attribute(hdr, RTA_PREFSRC, chunk, sizeof(request));
1893 if (gateway && gateway->get_family(gateway) == src_ip->get_family(src_ip))
1894 {
1895 chunk = gateway->get_address(gateway);
1896 netlink_add_attribute(hdr, RTA_GATEWAY, chunk, sizeof(request));
1897 }
1898 ifindex = get_interface_index(this, if_name);
1899 chunk.ptr = (char*)&ifindex;
1900 chunk.len = sizeof(ifindex);
1901 netlink_add_attribute(hdr, RTA_OIF, chunk, sizeof(request));
1902
1903 return this->socket->send_ack(this->socket, hdr);
1904 }
1905
1906 METHOD(kernel_net_t, add_route, status_t,
1907 private_kernel_netlink_net_t *this, chunk_t dst_net, u_int8_t prefixlen,
1908 host_t *gateway, host_t *src_ip, char *if_name)
1909 {
1910 status_t status;
1911 route_entry_t *found, route = {
1912 .dst_net = dst_net,
1913 .prefixlen = prefixlen,
1914 .gateway = gateway,
1915 .src_ip = src_ip,
1916 .if_name = if_name,
1917 };
1918
1919 this->routes_lock->lock(this->routes_lock);
1920 found = this->routes->get(this->routes, &route);
1921 if (found)
1922 {
1923 this->routes_lock->unlock(this->routes_lock);
1924 return ALREADY_DONE;
1925 }
1926 found = route_entry_clone(&route);
1927 this->routes->put(this->routes, found, found);
1928 status = manage_srcroute(this, RTM_NEWROUTE, NLM_F_CREATE | NLM_F_EXCL,
1929 dst_net, prefixlen, gateway, src_ip, if_name);
1930 this->routes_lock->unlock(this->routes_lock);
1931 return status;
1932 }
1933
1934 METHOD(kernel_net_t, del_route, status_t,
1935 private_kernel_netlink_net_t *this, chunk_t dst_net, u_int8_t prefixlen,
1936 host_t *gateway, host_t *src_ip, char *if_name)
1937 {
1938 status_t status;
1939 route_entry_t *found, route = {
1940 .dst_net = dst_net,
1941 .prefixlen = prefixlen,
1942 .gateway = gateway,
1943 .src_ip = src_ip,
1944 .if_name = if_name,
1945 };
1946
1947 this->routes_lock->lock(this->routes_lock);
1948 found = this->routes->get(this->routes, &route);
1949 if (!found)
1950 {
1951 this->routes_lock->unlock(this->routes_lock);
1952 return NOT_FOUND;
1953 }
1954 this->routes->remove(this->routes, found);
1955 route_entry_destroy(found);
1956 status = manage_srcroute(this, RTM_DELROUTE, 0, dst_net, prefixlen,
1957 gateway, src_ip, if_name);
1958 this->routes_lock->unlock(this->routes_lock);
1959 return status;
1960 }
1961
1962 /**
1963 * Initialize a list of local addresses.
1964 */
1965 static status_t init_address_list(private_kernel_netlink_net_t *this)
1966 {
1967 netlink_buf_t request;
1968 struct nlmsghdr *out, *current, *in;
1969 struct rtgenmsg *msg;
1970 size_t len;
1971 enumerator_t *ifaces, *addrs;
1972 iface_entry_t *iface;
1973 addr_entry_t *addr;
1974
1975 DBG2(DBG_KNL, "known interfaces and IP addresses:");
1976
1977 memset(&request, 0, sizeof(request));
1978
1979 in = (struct nlmsghdr*)&request;
1980 in->nlmsg_len = NLMSG_LENGTH(sizeof(struct rtgenmsg));
1981 in->nlmsg_flags = NLM_F_REQUEST | NLM_F_MATCH | NLM_F_ROOT;
1982 msg = (struct rtgenmsg*)NLMSG_DATA(in);
1983 msg->rtgen_family = AF_UNSPEC;
1984
1985 /* get all links */
1986 in->nlmsg_type = RTM_GETLINK;
1987 if (this->socket->send(this->socket, in, &out, &len) != SUCCESS)
1988 {
1989 return FAILED;
1990 }
1991 current = out;
1992 while (NLMSG_OK(current, len))
1993 {
1994 switch (current->nlmsg_type)
1995 {
1996 case NLMSG_DONE:
1997 break;
1998 case RTM_NEWLINK:
1999 process_link(this, current, FALSE);
2000 /* fall through */
2001 default:
2002 current = NLMSG_NEXT(current, len);
2003 continue;
2004 }
2005 break;
2006 }
2007 free(out);
2008
2009 /* get all interface addresses */
2010 in->nlmsg_type = RTM_GETADDR;
2011 if (this->socket->send(this->socket, in, &out, &len) != SUCCESS)
2012 {
2013 return FAILED;
2014 }
2015 current = out;
2016 while (NLMSG_OK(current, len))
2017 {
2018 switch (current->nlmsg_type)
2019 {
2020 case NLMSG_DONE:
2021 break;
2022 case RTM_NEWADDR:
2023 process_addr(this, current, FALSE);
2024 /* fall through */
2025 default:
2026 current = NLMSG_NEXT(current, len);
2027 continue;
2028 }
2029 break;
2030 }
2031 free(out);
2032
2033 this->lock->read_lock(this->lock);
2034 ifaces = this->ifaces->create_enumerator(this->ifaces);
2035 while (ifaces->enumerate(ifaces, &iface))
2036 {
2037 if (iface_entry_up_and_usable(iface))
2038 {
2039 DBG2(DBG_KNL, " %s", iface->ifname);
2040 addrs = iface->addrs->create_enumerator(iface->addrs);
2041 while (addrs->enumerate(addrs, (void**)&addr))
2042 {
2043 DBG2(DBG_KNL, " %H", addr->ip);
2044 }
2045 addrs->destroy(addrs);
2046 }
2047 }
2048 ifaces->destroy(ifaces);
2049 this->lock->unlock(this->lock);
2050 return SUCCESS;
2051 }
2052
2053 /**
2054 * create or delete a rule to use our routing table
2055 */
2056 static status_t manage_rule(private_kernel_netlink_net_t *this, int nlmsg_type,
2057 int family, u_int32_t table, u_int32_t prio)
2058 {
2059 netlink_buf_t request;
2060 struct nlmsghdr *hdr;
2061 struct rtmsg *msg;
2062 chunk_t chunk;
2063
2064 memset(&request, 0, sizeof(request));
2065 hdr = (struct nlmsghdr*)request;
2066 hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
2067 hdr->nlmsg_type = nlmsg_type;
2068 if (nlmsg_type == RTM_NEWRULE)
2069 {
2070 hdr->nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
2071 }
2072 hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
2073
2074 msg = (struct rtmsg*)NLMSG_DATA(hdr);
2075 msg->rtm_table = table;
2076 msg->rtm_family = family;
2077 msg->rtm_protocol = RTPROT_BOOT;
2078 msg->rtm_scope = RT_SCOPE_UNIVERSE;
2079 msg->rtm_type = RTN_UNICAST;
2080
2081 chunk = chunk_from_thing(prio);
2082 netlink_add_attribute(hdr, RTA_PRIORITY, chunk, sizeof(request));
2083
2084 return this->socket->send_ack(this->socket, hdr);
2085 }
2086
2087 /**
2088 * check for kernel features (currently only via version number)
2089 */
2090 static void check_kernel_features(private_kernel_netlink_net_t *this)
2091 {
2092 struct utsname utsname;
2093 int a, b, c;
2094
2095 if (uname(&utsname) == 0)
2096 {
2097 switch(sscanf(utsname.release, "%d.%d.%d", &a, &b, &c))
2098 {
2099 case 3:
2100 if (a == 2)
2101 {
2102 DBG2(DBG_KNL, "detected Linux %d.%d.%d, no support for "
2103 "RTA_PREFSRC for IPv6 routes", a, b, c);
2104 break;
2105 }
2106 /* fall-through */
2107 case 2:
2108 /* only 3.x+ uses two part version numbers */
2109 this->rta_prefsrc_for_ipv6 = TRUE;
2110 break;
2111 default:
2112 break;
2113 }
2114 }
2115 }
2116
2117 /**
2118 * Destroy an address to iface map
2119 */
2120 static void addr_map_destroy(hashtable_t *map)
2121 {
2122 enumerator_t *enumerator;
2123 addr_map_entry_t *addr;
2124
2125 enumerator = map->create_enumerator(map);
2126 while (enumerator->enumerate(enumerator, NULL, (void**)&addr))
2127 {
2128 free(addr);
2129 }
2130 enumerator->destroy(enumerator);
2131 map->destroy(map);
2132 }
2133
2134 METHOD(kernel_net_t, destroy, void,
2135 private_kernel_netlink_net_t *this)
2136 {
2137 enumerator_t *enumerator;
2138 route_entry_t *route;
2139
2140 if (this->routing_table)
2141 {
2142 manage_rule(this, RTM_DELRULE, AF_INET, this->routing_table,
2143 this->routing_table_prio);
2144 manage_rule(this, RTM_DELRULE, AF_INET6, this->routing_table,
2145 this->routing_table_prio);
2146 }
2147 if (this->socket_events > 0)
2148 {
2149 close(this->socket_events);
2150 }
2151 enumerator = this->routes->create_enumerator(this->routes);
2152 while (enumerator->enumerate(enumerator, NULL, (void**)&route))
2153 {
2154 manage_srcroute(this, RTM_DELROUTE, 0, route->dst_net, route->prefixlen,
2155 route->gateway, route->src_ip, route->if_name);
2156 route_entry_destroy(route);
2157 }
2158 enumerator->destroy(enumerator);
2159 this->routes->destroy(this->routes);
2160 this->routes_lock->destroy(this->routes_lock);
2161 DESTROY_IF(this->socket);
2162
2163 net_changes_clear(this);
2164 this->net_changes->destroy(this->net_changes);
2165 this->net_changes_lock->destroy(this->net_changes_lock);
2166
2167 addr_map_destroy(this->addrs);
2168 addr_map_destroy(this->vips);
2169
2170 this->ifaces->destroy_function(this->ifaces, (void*)iface_entry_destroy);
2171 this->rt_exclude->destroy(this->rt_exclude);
2172 this->roam_lock->destroy(this->roam_lock);
2173 this->condvar->destroy(this->condvar);
2174 this->lock->destroy(this->lock);
2175 free(this);
2176 }
2177
2178 /*
2179 * Described in header.
2180 */
2181 kernel_netlink_net_t *kernel_netlink_net_create()
2182 {
2183 private_kernel_netlink_net_t *this;
2184 enumerator_t *enumerator;
2185 bool register_for_events = TRUE;
2186 char *exclude;
2187
2188 INIT(this,
2189 .public = {
2190 .interface = {
2191 .get_interface = _get_interface_name,
2192 .create_address_enumerator = _create_address_enumerator,
2193 .get_source_addr = _get_source_addr,
2194 .get_nexthop = _get_nexthop,
2195 .add_ip = _add_ip,
2196 .del_ip = _del_ip,
2197 .add_route = _add_route,
2198 .del_route = _del_route,
2199 .destroy = _destroy,
2200 },
2201 },
2202 .socket = netlink_socket_create(NETLINK_ROUTE),
2203 .rt_exclude = linked_list_create(),
2204 .routes = hashtable_create((hashtable_hash_t)route_entry_hash,
2205 (hashtable_equals_t)route_entry_equals, 16),
2206 .net_changes = hashtable_create(
2207 (hashtable_hash_t)net_change_hash,
2208 (hashtable_equals_t)net_change_equals, 16),
2209 .addrs = hashtable_create(
2210 (hashtable_hash_t)addr_map_entry_hash,
2211 (hashtable_equals_t)addr_map_entry_equals, 16),
2212 .vips = hashtable_create((hashtable_hash_t)addr_map_entry_hash,
2213 (hashtable_equals_t)addr_map_entry_equals, 16),
2214 .routes_lock = mutex_create(MUTEX_TYPE_DEFAULT),
2215 .net_changes_lock = mutex_create(MUTEX_TYPE_DEFAULT),
2216 .ifaces = linked_list_create(),
2217 .lock = rwlock_create(RWLOCK_TYPE_DEFAULT),
2218 .condvar = rwlock_condvar_create(),
2219 .roam_lock = spinlock_create(),
2220 .routing_table = lib->settings->get_int(lib->settings,
2221 "%s.routing_table", ROUTING_TABLE, hydra->daemon),
2222 .routing_table_prio = lib->settings->get_int(lib->settings,
2223 "%s.routing_table_prio", ROUTING_TABLE_PRIO, hydra->daemon),
2224 .process_route = lib->settings->get_bool(lib->settings,
2225 "%s.process_route", TRUE, hydra->daemon),
2226 .install_virtual_ip = lib->settings->get_bool(lib->settings,
2227 "%s.install_virtual_ip", TRUE, hydra->daemon),
2228 .install_virtual_ip_on = lib->settings->get_str(lib->settings,
2229 "%s.install_virtual_ip_on", NULL, hydra->daemon),
2230 );
2231 timerclear(&this->last_route_reinstall);
2232 timerclear(&this->next_roam);
2233
2234 check_kernel_features(this);
2235
2236 if (streq(hydra->daemon, "starter"))
2237 { /* starter has no threads, so we do not register for kernel events */
2238 register_for_events = FALSE;
2239 }
2240
2241 exclude = lib->settings->get_str(lib->settings,
2242 "%s.ignore_routing_tables", NULL, hydra->daemon);
2243 if (exclude)
2244 {
2245 char *token;
2246 uintptr_t table;
2247
2248 enumerator = enumerator_create_token(exclude, " ", " ");
2249 while (enumerator->enumerate(enumerator, &token))
2250 {
2251 errno = 0;
2252 table = strtoul(token, NULL, 10);
2253
2254 if (errno == 0)
2255 {
2256 this->rt_exclude->insert_last(this->rt_exclude, (void*)table);
2257 }
2258 }
2259 enumerator->destroy(enumerator);
2260 }
2261
2262 if (register_for_events)
2263 {
2264 struct sockaddr_nl addr;
2265
2266 memset(&addr, 0, sizeof(addr));
2267 addr.nl_family = AF_NETLINK;
2268
2269 /* create and bind RT socket for events (address/interface/route changes) */
2270 this->socket_events = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
2271 if (this->socket_events < 0)
2272 {
2273 DBG1(DBG_KNL, "unable to create RT event socket");
2274 destroy(this);
2275 return NULL;
2276 }
2277 addr.nl_groups = RTMGRP_IPV4_IFADDR | RTMGRP_IPV6_IFADDR |
2278 RTMGRP_IPV4_ROUTE | RTMGRP_IPV6_ROUTE | RTMGRP_LINK;
2279 if (bind(this->socket_events, (struct sockaddr*)&addr, sizeof(addr)))
2280 {
2281 DBG1(DBG_KNL, "unable to bind RT event socket");
2282 destroy(this);
2283 return NULL;
2284 }
2285
2286 lib->processor->queue_job(lib->processor,
2287 (job_t*)callback_job_create_with_prio(
2288 (callback_job_cb_t)receive_events, this, NULL,
2289 (callback_job_cancel_t)return_false, JOB_PRIO_CRITICAL));
2290 }
2291
2292 if (init_address_list(this) != SUCCESS)
2293 {
2294 DBG1(DBG_KNL, "unable to get interface list");
2295 destroy(this);
2296 return NULL;
2297 }
2298
2299 if (this->routing_table)
2300 {
2301 if (manage_rule(this, RTM_NEWRULE, AF_INET, this->routing_table,
2302 this->routing_table_prio) != SUCCESS)
2303 {
2304 DBG1(DBG_KNL, "unable to create IPv4 routing table rule");
2305 }
2306 if (manage_rule(this, RTM_NEWRULE, AF_INET6, this->routing_table,
2307 this->routing_table_prio) != SUCCESS)
2308 {
2309 DBG1(DBG_KNL, "unable to create IPv6 routing table rule");
2310 }
2311 }
2312
2313 return &this->public;
2314 }