kernel-netlink: Set NODAD flag for virtual IPv6 addresses
[strongswan.git] / src / libcharon / plugins / kernel_netlink / kernel_netlink_net.c
1 /*
2 * Copyright (C) 2008-2016 Tobias Brunner
3 * Copyright (C) 2005-2008 Martin Willi
4 * HSR 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 #ifdef HAVE_LINUX_FIB_RULES_H
48 #include <linux/fib_rules.h>
49 #endif
50
51 #include "kernel_netlink_net.h"
52 #include "kernel_netlink_shared.h"
53
54 #include <daemon.h>
55 #include <utils/debug.h>
56 #include <threading/mutex.h>
57 #include <threading/rwlock.h>
58 #include <threading/rwlock_condvar.h>
59 #include <threading/spinlock.h>
60 #include <collections/hashtable.h>
61 #include <collections/linked_list.h>
62 #include <processing/jobs/callback_job.h>
63
64 /** delay before firing roam events (ms) */
65 #define ROAM_DELAY 100
66
67 /** delay before reinstalling routes (ms) */
68 #define ROUTE_DELAY 100
69
70 /** maximum recursion when searching for addresses in get_route() */
71 #define MAX_ROUTE_RECURSION 2
72
73 #ifndef ROUTING_TABLE
74 #define ROUTING_TABLE 0
75 #endif
76
77 #ifndef ROUTING_TABLE_PRIO
78 #define ROUTING_TABLE_PRIO 0
79 #endif
80
81 ENUM(rt_msg_names, RTM_NEWLINK, RTM_GETRULE,
82 "RTM_NEWLINK",
83 "RTM_DELLINK",
84 "RTM_GETLINK",
85 "RTM_SETLINK",
86 "RTM_NEWADDR",
87 "RTM_DELADDR",
88 "RTM_GETADDR",
89 "31",
90 "RTM_NEWROUTE",
91 "RTM_DELROUTE",
92 "RTM_GETROUTE",
93 "35",
94 "RTM_NEWNEIGH",
95 "RTM_DELNEIGH",
96 "RTM_GETNEIGH",
97 "RTM_NEWRULE",
98 "RTM_DELRULE",
99 "RTM_GETRULE",
100 );
101
102 typedef struct addr_entry_t addr_entry_t;
103
104 /**
105 * IP address in an iface_entry_t
106 */
107 struct addr_entry_t {
108
109 /** the ip address */
110 host_t *ip;
111
112 /** address flags */
113 u_char flags;
114
115 /** scope of the address */
116 u_char scope;
117
118 /** number of times this IP is used, if virtual (i.e. managed by us) */
119 u_int refcount;
120
121 /** TRUE once it is installed, if virtual */
122 bool installed;
123 };
124
125 /**
126 * destroy a addr_entry_t object
127 */
128 static void addr_entry_destroy(addr_entry_t *this)
129 {
130 this->ip->destroy(this->ip);
131 free(this);
132 }
133
134 typedef struct iface_entry_t iface_entry_t;
135
136 /**
137 * A network interface on this system, containing addr_entry_t's
138 */
139 struct iface_entry_t {
140
141 /** interface index */
142 int ifindex;
143
144 /** name of the interface */
145 char ifname[IFNAMSIZ];
146
147 /** interface flags, as in netdevice(7) SIOCGIFFLAGS */
148 u_int flags;
149
150 /** list of addresses as host_t */
151 linked_list_t *addrs;
152
153 /** TRUE if usable by config */
154 bool usable;
155 };
156
157 /**
158 * destroy an interface entry
159 */
160 static void iface_entry_destroy(iface_entry_t *this)
161 {
162 this->addrs->destroy_function(this->addrs, (void*)addr_entry_destroy);
163 free(this);
164 }
165
166 /**
167 * find an interface entry by index
168 */
169 static bool iface_entry_by_index(iface_entry_t *this, int *ifindex)
170 {
171 return this->ifindex == *ifindex;
172 }
173
174 /**
175 * find an interface entry by name
176 */
177 static bool iface_entry_by_name(iface_entry_t *this, char *ifname)
178 {
179 return streq(this->ifname, ifname);
180 }
181
182 /**
183 * check if an interface is up
184 */
185 static inline bool iface_entry_up(iface_entry_t *iface)
186 {
187 return (iface->flags & IFF_UP) == IFF_UP;
188 }
189
190 /**
191 * check if an interface is up and usable
192 */
193 static inline bool iface_entry_up_and_usable(iface_entry_t *iface)
194 {
195 return iface->usable && iface_entry_up(iface);
196 }
197
198 typedef struct addr_map_entry_t addr_map_entry_t;
199
200 /**
201 * Entry that maps an IP address to an interface entry
202 */
203 struct addr_map_entry_t {
204 /** The IP address */
205 host_t *ip;
206
207 /** The address entry for this IP address */
208 addr_entry_t *addr;
209
210 /** The interface this address is installed on */
211 iface_entry_t *iface;
212 };
213
214 /**
215 * Hash a addr_map_entry_t object, all entries with the same IP address
216 * are stored in the same bucket
217 */
218 static u_int addr_map_entry_hash(addr_map_entry_t *this)
219 {
220 return chunk_hash(this->ip->get_address(this->ip));
221 }
222
223 /**
224 * Compare two addr_map_entry_t objects, two entries are equal if they are
225 * installed on the same interface
226 */
227 static bool addr_map_entry_equals(addr_map_entry_t *a, addr_map_entry_t *b)
228 {
229 return a->iface->ifindex == b->iface->ifindex &&
230 a->ip->ip_equals(a->ip, b->ip);
231 }
232
233 /**
234 * Used with get_match this finds an address entry if it is installed on
235 * an up and usable interface
236 */
237 static bool addr_map_entry_match_up_and_usable(addr_map_entry_t *a,
238 addr_map_entry_t *b)
239 {
240 return iface_entry_up_and_usable(b->iface) &&
241 a->ip->ip_equals(a->ip, b->ip);
242 }
243
244 /**
245 * Used with get_match this finds an address entry if it is installed on
246 * any active local interface
247 */
248 static bool addr_map_entry_match_up(addr_map_entry_t *a, addr_map_entry_t *b)
249 {
250 return iface_entry_up(b->iface) && a->ip->ip_equals(a->ip, b->ip);
251 }
252
253 /**
254 * Used with get_match this finds an address entry if it is installed on
255 * any local interface
256 */
257 static bool addr_map_entry_match(addr_map_entry_t *a, addr_map_entry_t *b)
258 {
259 return a->ip->ip_equals(a->ip, b->ip);
260 }
261
262 typedef struct route_entry_t route_entry_t;
263
264 /**
265 * Installed routing entry
266 */
267 struct route_entry_t {
268 /** Name of the interface the route is bound to */
269 char *if_name;
270
271 /** Source ip of the route */
272 host_t *src_ip;
273
274 /** Gateway for this route */
275 host_t *gateway;
276
277 /** Destination net */
278 chunk_t dst_net;
279
280 /** Destination net prefixlen */
281 uint8_t prefixlen;
282 };
283
284 /**
285 * Clone a route_entry_t object.
286 */
287 static route_entry_t *route_entry_clone(route_entry_t *this)
288 {
289 route_entry_t *route;
290
291 INIT(route,
292 .if_name = strdup(this->if_name),
293 .src_ip = this->src_ip->clone(this->src_ip),
294 .gateway = this->gateway ? this->gateway->clone(this->gateway) : NULL,
295 .dst_net = chunk_clone(this->dst_net),
296 .prefixlen = this->prefixlen,
297 );
298 return route;
299 }
300
301 /**
302 * Destroy a route_entry_t object
303 */
304 static void route_entry_destroy(route_entry_t *this)
305 {
306 free(this->if_name);
307 DESTROY_IF(this->src_ip);
308 DESTROY_IF(this->gateway);
309 chunk_free(&this->dst_net);
310 free(this);
311 }
312
313 /**
314 * Hash a route_entry_t object
315 */
316 static u_int route_entry_hash(route_entry_t *this)
317 {
318 return chunk_hash_inc(chunk_from_thing(this->prefixlen),
319 chunk_hash(this->dst_net));
320 }
321
322 /**
323 * Compare two route_entry_t objects
324 */
325 static bool route_entry_equals(route_entry_t *a, route_entry_t *b)
326 {
327 if (a->if_name && b->if_name && streq(a->if_name, b->if_name) &&
328 a->src_ip->ip_equals(a->src_ip, b->src_ip) &&
329 chunk_equals(a->dst_net, b->dst_net) && a->prefixlen == b->prefixlen)
330 {
331 return (!a->gateway && !b->gateway) || (a->gateway && b->gateway &&
332 a->gateway->ip_equals(a->gateway, b->gateway));
333 }
334 return FALSE;
335 }
336
337 typedef struct net_change_t net_change_t;
338
339 /**
340 * Queued network changes
341 */
342 struct net_change_t {
343 /** Name of the interface that got activated (or an IP appeared on) */
344 char *if_name;
345 };
346
347 /**
348 * Destroy a net_change_t object
349 */
350 static void net_change_destroy(net_change_t *this)
351 {
352 free(this->if_name);
353 free(this);
354 }
355
356 /**
357 * Hash a net_change_t object
358 */
359 static u_int net_change_hash(net_change_t *this)
360 {
361 return chunk_hash(chunk_create(this->if_name, strlen(this->if_name)));
362 }
363
364 /**
365 * Compare two net_change_t objects
366 */
367 static bool net_change_equals(net_change_t *a, net_change_t *b)
368 {
369 return streq(a->if_name, b->if_name);
370 }
371
372 typedef struct private_kernel_netlink_net_t private_kernel_netlink_net_t;
373
374 /**
375 * Private variables and functions of kernel_netlink_net class.
376 */
377 struct private_kernel_netlink_net_t {
378 /**
379 * Public part of the kernel_netlink_net_t object.
380 */
381 kernel_netlink_net_t public;
382
383 /**
384 * lock to access various lists and maps
385 */
386 rwlock_t *lock;
387
388 /**
389 * condition variable to signal virtual IP add/removal
390 */
391 rwlock_condvar_t *condvar;
392
393 /**
394 * Cached list of interfaces and its addresses (iface_entry_t)
395 */
396 linked_list_t *ifaces;
397
398 /**
399 * Map for IP addresses to iface_entry_t objects (addr_map_entry_t)
400 */
401 hashtable_t *addrs;
402
403 /**
404 * Map for virtual IP addresses to iface_entry_t objects (addr_map_entry_t)
405 */
406 hashtable_t *vips;
407
408 /**
409 * netlink rt socket (routing)
410 */
411 netlink_socket_t *socket;
412
413 /**
414 * Netlink rt socket to receive address change events
415 */
416 int socket_events;
417
418 /**
419 * earliest time of the next roam event
420 */
421 timeval_t next_roam;
422
423 /**
424 * roam event due to address change
425 */
426 bool roam_address;
427
428 /**
429 * lock to check and update roam event time
430 */
431 spinlock_t *roam_lock;
432
433 /**
434 * routing table to install routes
435 */
436 int routing_table;
437
438 /**
439 * priority of used routing table
440 */
441 int routing_table_prio;
442
443 /**
444 * installed routes
445 */
446 hashtable_t *routes;
447
448 /**
449 * mutex for routes
450 */
451 mutex_t *routes_lock;
452
453 /**
454 * interface changes which may trigger route reinstallation
455 */
456 hashtable_t *net_changes;
457
458 /**
459 * mutex for route reinstallation triggers
460 */
461 mutex_t *net_changes_lock;
462
463 /**
464 * time of last route reinstallation
465 */
466 timeval_t last_route_reinstall;
467
468 /**
469 * whether to react to RTM_NEWROUTE or RTM_DELROUTE events
470 */
471 bool process_route;
472
473 /**
474 * whether to trigger roam events
475 */
476 bool roam_events;
477
478 /**
479 * whether to actually install virtual IPs
480 */
481 bool install_virtual_ip;
482
483 /**
484 * the name of the interface virtual IP addresses are installed on
485 */
486 char *install_virtual_ip_on;
487
488 /**
489 * whether preferred source addresses can be specified for IPv6 routes
490 */
491 bool rta_prefsrc_for_ipv6;
492
493 /**
494 * whether marks can be used in route lookups
495 */
496 bool rta_mark;
497
498 /**
499 * the mark excluded from the routing rule used for virtual IPs
500 */
501 mark_t routing_mark;
502
503 /**
504 * whether to prefer temporary IPv6 addresses over public ones
505 */
506 bool prefer_temporary_addrs;
507
508 /**
509 * list with routing tables to be excluded from route lookup
510 */
511 linked_list_t *rt_exclude;
512
513 /**
514 * MTU to set on installed routes
515 */
516 uint32_t mtu;
517
518 /**
519 * MSS to set on installed routes
520 */
521 uint32_t mss;
522 };
523
524 /**
525 * Forward declaration
526 */
527 static status_t manage_srcroute(private_kernel_netlink_net_t *this,
528 int nlmsg_type, int flags, chunk_t dst_net,
529 uint8_t prefixlen, host_t *gateway,
530 host_t *src_ip, char *if_name);
531
532 /**
533 * Clear the queued network changes.
534 */
535 static void net_changes_clear(private_kernel_netlink_net_t *this)
536 {
537 enumerator_t *enumerator;
538 net_change_t *change;
539
540 enumerator = this->net_changes->create_enumerator(this->net_changes);
541 while (enumerator->enumerate(enumerator, NULL, (void**)&change))
542 {
543 this->net_changes->remove_at(this->net_changes, enumerator);
544 net_change_destroy(change);
545 }
546 enumerator->destroy(enumerator);
547 }
548
549 /**
550 * Act upon queued network changes.
551 */
552 static job_requeue_t reinstall_routes(private_kernel_netlink_net_t *this)
553 {
554 enumerator_t *enumerator;
555 route_entry_t *route;
556
557 this->net_changes_lock->lock(this->net_changes_lock);
558 this->routes_lock->lock(this->routes_lock);
559
560 enumerator = this->routes->create_enumerator(this->routes);
561 while (enumerator->enumerate(enumerator, NULL, (void**)&route))
562 {
563 net_change_t *change, lookup = {
564 .if_name = route->if_name,
565 };
566 /* check if a change for the outgoing interface is queued */
567 change = this->net_changes->get(this->net_changes, &lookup);
568 if (!change)
569 { /* in case src_ip is not on the outgoing interface */
570 if (this->public.interface.get_interface(&this->public.interface,
571 route->src_ip, &lookup.if_name))
572 {
573 if (!streq(lookup.if_name, route->if_name))
574 {
575 change = this->net_changes->get(this->net_changes, &lookup);
576 }
577 free(lookup.if_name);
578 }
579 }
580 if (change)
581 {
582 manage_srcroute(this, RTM_NEWROUTE, NLM_F_CREATE | NLM_F_EXCL,
583 route->dst_net, route->prefixlen, route->gateway,
584 route->src_ip, route->if_name);
585 }
586 }
587 enumerator->destroy(enumerator);
588 this->routes_lock->unlock(this->routes_lock);
589
590 net_changes_clear(this);
591 this->net_changes_lock->unlock(this->net_changes_lock);
592 return JOB_REQUEUE_NONE;
593 }
594
595 /**
596 * Queue route reinstallation caused by network changes for a given interface.
597 *
598 * The route reinstallation is delayed for a while and only done once for
599 * several calls during this delay, in order to avoid doing it too often.
600 * The interface name is freed.
601 */
602 static void queue_route_reinstall(private_kernel_netlink_net_t *this,
603 char *if_name)
604 {
605 net_change_t *update, *found;
606 timeval_t now;
607 job_t *job;
608
609 INIT(update,
610 .if_name = if_name
611 );
612
613 this->net_changes_lock->lock(this->net_changes_lock);
614 found = this->net_changes->put(this->net_changes, update, update);
615 if (found)
616 {
617 net_change_destroy(found);
618 }
619 time_monotonic(&now);
620 if (timercmp(&now, &this->last_route_reinstall, >))
621 {
622 timeval_add_ms(&now, ROUTE_DELAY);
623 this->last_route_reinstall = now;
624
625 job = (job_t*)callback_job_create((callback_job_cb_t)reinstall_routes,
626 this, NULL, NULL);
627 lib->scheduler->schedule_job_ms(lib->scheduler, job, ROUTE_DELAY);
628 }
629 this->net_changes_lock->unlock(this->net_changes_lock);
630 }
631
632 /**
633 * check if the given IP is known as virtual IP and currently installed
634 *
635 * this function will also return TRUE if the virtual IP entry disappeared.
636 * in that case the returned entry will be NULL.
637 *
638 * this->lock must be held when calling this function
639 */
640 static bool is_vip_installed_or_gone(private_kernel_netlink_net_t *this,
641 host_t *ip, addr_map_entry_t **entry)
642 {
643 addr_map_entry_t lookup = {
644 .ip = ip,
645 };
646
647 *entry = this->vips->get_match(this->vips, &lookup,
648 (void*)addr_map_entry_match);
649 if (*entry == NULL)
650 { /* the virtual IP disappeared */
651 return TRUE;
652 }
653 return (*entry)->addr->installed;
654 }
655
656 /**
657 * check if the given IP is known as virtual IP
658 *
659 * this->lock must be held when calling this function
660 */
661 static bool is_known_vip(private_kernel_netlink_net_t *this, host_t *ip)
662 {
663 addr_map_entry_t lookup = {
664 .ip = ip,
665 };
666
667 return this->vips->get_match(this->vips, &lookup,
668 (void*)addr_map_entry_match) != NULL;
669 }
670
671 /**
672 * Add an address map entry
673 */
674 static void addr_map_entry_add(hashtable_t *map, addr_entry_t *addr,
675 iface_entry_t *iface)
676 {
677 addr_map_entry_t *entry;
678
679 INIT(entry,
680 .ip = addr->ip,
681 .addr = addr,
682 .iface = iface,
683 );
684 entry = map->put(map, entry, entry);
685 free(entry);
686 }
687
688 /**
689 * Remove an address map entry
690 */
691 static void addr_map_entry_remove(hashtable_t *map, addr_entry_t *addr,
692 iface_entry_t *iface)
693 {
694 addr_map_entry_t *entry, lookup = {
695 .ip = addr->ip,
696 .addr = addr,
697 .iface = iface,
698 };
699
700 entry = map->remove(map, &lookup);
701 free(entry);
702 }
703
704 /**
705 * Check if an address or net (addr with prefix net bits) is in
706 * subnet (net with net_len net bits)
707 */
708 static bool addr_in_subnet(chunk_t addr, int prefix, chunk_t net, int net_len)
709 {
710 static const u_char mask[] = { 0x00, 0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe };
711 int byte = 0;
712
713 if (net_len == 0)
714 { /* any address matches a /0 network */
715 return TRUE;
716 }
717 if (addr.len != net.len || net_len > 8 * net.len || prefix < net_len)
718 {
719 return FALSE;
720 }
721 /* scan through all bytes in network order */
722 while (net_len > 0)
723 {
724 if (net_len < 8)
725 {
726 return (mask[net_len] & addr.ptr[byte]) == (mask[net_len] & net.ptr[byte]);
727 }
728 else
729 {
730 if (addr.ptr[byte] != net.ptr[byte])
731 {
732 return FALSE;
733 }
734 byte++;
735 net_len -= 8;
736 }
737 }
738 return TRUE;
739 }
740
741 /**
742 * Check if the given address is in subnet (net with net_len net bits)
743 */
744 static bool host_in_subnet(host_t *host, chunk_t net, int net_len)
745 {
746 chunk_t addr;
747
748 addr = host->get_address(host);
749 return addr_in_subnet(addr, addr.len * 8, net, net_len);
750 }
751
752 /**
753 * Determine the type or scope of the given unicast IP address. This is not
754 * the same thing returned in rtm_scope/ifa_scope.
755 *
756 * We use return values as defined in RFC 6724 (referring to RFC 4291).
757 */
758 static u_char get_scope(host_t *ip)
759 {
760 chunk_t addr;
761
762 addr = ip->get_address(ip);
763 switch (addr.len)
764 {
765 case 4:
766 /* we use the mapping defined in RFC 6724, 3.2 */
767 if (addr.ptr[0] == 127)
768 { /* link-local, same as the IPv6 loopback address */
769 return 2;
770 }
771 if (addr.ptr[0] == 169 && addr.ptr[1] == 254)
772 { /* link-local */
773 return 2;
774 }
775 break;
776 case 16:
777 if (IN6_IS_ADDR_LOOPBACK((struct in6_addr*)addr.ptr))
778 { /* link-local, according to RFC 4291, 2.5.3 */
779 return 2;
780 }
781 if (IN6_IS_ADDR_LINKLOCAL((struct in6_addr*)addr.ptr))
782 {
783 return 2;
784 }
785 if (IN6_IS_ADDR_SITELOCAL((struct in6_addr*)addr.ptr))
786 { /* deprecated, according to RFC 4291, 2.5.7 */
787 return 5;
788 }
789 break;
790 default:
791 break;
792 }
793 /* global */
794 return 14;
795 }
796
797 /**
798 * Determine the label of the given unicast IP address.
799 *
800 * We currently only support the default table given in RFC 6724:
801 *
802 * Prefix Precedence Label
803 * ::1/128 50 0
804 * ::/0 40 1
805 * ::ffff:0:0/96 35 4
806 * 2002::/16 30 2
807 * 2001::/32 5 5
808 * fc00::/7 3 13
809 * ::/96 1 3
810 * fec0::/10 1 11
811 * 3ffe::/16 1 12
812 */
813 static u_char get_label(host_t *ip)
814 {
815 struct {
816 chunk_t net;
817 u_char prefix;
818 u_char label;
819 } priorities[] = {
820 /* priority table ordered by prefix */
821 /* ::1/128 */
822 { chunk_from_chars(0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
823 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01), 128, 0 },
824 /* ::ffff:0:0/96 */
825 { chunk_from_chars(0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
826 0x00, 0x00, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00), 96, 4 },
827 /* ::/96 */
828 { chunk_from_chars(0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
829 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00), 96, 3 },
830 /* 2001::/32 */
831 { chunk_from_chars(0x20, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
832 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00), 32, 5 },
833 /* 2002::/16 */
834 { chunk_from_chars(0x20, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
835 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00), 16, 2 },
836 /* 3ffe::/16 */
837 { chunk_from_chars(0x3f, 0xfe, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
838 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00), 16, 12 },
839 /* fec0::/10 */
840 { chunk_from_chars(0xfe, 0xc0, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
841 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00), 10, 11 },
842 /* fc00::/7 */
843 { chunk_from_chars(0xfc, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
844 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00), 7, 13 },
845 };
846 int i;
847
848 for (i = 0; i < countof(priorities); i++)
849 {
850 if (host_in_subnet(ip, priorities[i].net, priorities[i].prefix))
851 {
852 return priorities[i].label;
853 }
854 }
855 /* ::/0 */
856 return 1;
857 }
858
859 /**
860 * Returns the length of the common prefix in bits up to the length of a's
861 * prefix, defined by RFC 6724 as the portion of the address not including the
862 * interface ID, which is 64-bit for most unicast addresses (see RFC 4291).
863 */
864 static u_char common_prefix(host_t *a, host_t *b)
865 {
866 chunk_t aa, ba;
867 u_char byte, bits = 0, match;
868
869 aa = a->get_address(a);
870 ba = b->get_address(b);
871 for (byte = 0; byte < 8; byte++)
872 {
873 if (aa.ptr[byte] != ba.ptr[byte])
874 {
875 match = aa.ptr[byte] ^ ba.ptr[byte];
876 for (bits = 8; match; match >>= 1)
877 {
878 bits--;
879 }
880 break;
881 }
882 }
883 return byte * 8 + bits;
884 }
885
886 /**
887 * Compare two IP addresses and return TRUE if the second address is the better
888 * choice of the two to reach the destination.
889 * For IPv6 we approximately follow RFC 6724.
890 */
891 static bool is_address_better(private_kernel_netlink_net_t *this,
892 addr_entry_t *a, addr_entry_t *b, host_t *d)
893 {
894 u_char sa, sb, sd, la, lb, ld, pa, pb;
895
896 /* rule 2: prefer appropriate scope */
897 if (d)
898 {
899 sa = get_scope(a->ip);
900 sb = get_scope(b->ip);
901 sd = get_scope(d);
902 if (sa < sb)
903 {
904 return sa < sd;
905 }
906 else if (sb < sa)
907 {
908 return sb >= sd;
909 }
910 }
911 if (a->ip->get_family(a->ip) == AF_INET)
912 { /* stop here for IPv4, default to addresses found earlier */
913 return FALSE;
914 }
915 /* rule 3: avoid deprecated addresses (RFC 4862) */
916 if ((a->flags & IFA_F_DEPRECATED) != (b->flags & IFA_F_DEPRECATED))
917 {
918 return a->flags & IFA_F_DEPRECATED;
919 }
920 /* rule 4 is not applicable as we don't know if an address is a home or
921 * care-of addresses.
922 * rule 5 does not apply as we only compare addresses from one interface
923 */
924 /* rule 6: prefer matching label */
925 if (d)
926 {
927 la = get_label(a->ip);
928 lb = get_label(b->ip);
929 ld = get_label(d);
930 if (la == ld && lb != ld)
931 {
932 return FALSE;
933 }
934 else if (lb == ld && la != ld)
935 {
936 return TRUE;
937 }
938 }
939 /* rule 7: prefer temporary addresses (WE REVERSE THIS BY DEFAULT!) */
940 if ((a->flags & IFA_F_TEMPORARY) != (b->flags & IFA_F_TEMPORARY))
941 {
942 if (this->prefer_temporary_addrs)
943 {
944 return b->flags & IFA_F_TEMPORARY;
945 }
946 return a->flags & IFA_F_TEMPORARY;
947 }
948 /* rule 8: use longest matching prefix */
949 if (d)
950 {
951 pa = common_prefix(a->ip, d);
952 pb = common_prefix(b->ip, d);
953 if (pa != pb)
954 {
955 return pb > pa;
956 }
957 }
958 /* default to addresses found earlier */
959 return FALSE;
960 }
961
962 /**
963 * Get a non-virtual IP address on the given interfaces and optionally in a
964 * given subnet.
965 *
966 * If a candidate address is given, we first search for that address and if not
967 * found return the address as above.
968 * Returned host is a clone, has to be freed by caller.
969 *
970 * this->lock must be held when calling this function.
971 */
972 static host_t *get_matching_address(private_kernel_netlink_net_t *this,
973 int *ifindex, int family, chunk_t net,
974 uint8_t mask, host_t *dest,
975 host_t *candidate)
976 {
977 enumerator_t *ifaces, *addrs;
978 iface_entry_t *iface;
979 addr_entry_t *addr, *best = NULL;
980 bool candidate_matched = FALSE;
981
982 ifaces = this->ifaces->create_enumerator(this->ifaces);
983 while (ifaces->enumerate(ifaces, &iface))
984 {
985 if (iface->usable && (!ifindex || iface->ifindex == *ifindex))
986 { /* only use matching interfaces not excluded by config */
987 addrs = iface->addrs->create_enumerator(iface->addrs);
988 while (addrs->enumerate(addrs, &addr))
989 {
990 if (addr->refcount ||
991 addr->ip->get_family(addr->ip) != family)
992 { /* ignore virtual IP addresses and ensure family matches */
993 continue;
994 }
995 if (net.ptr && !host_in_subnet(addr->ip, net, mask))
996 { /* optionally match a subnet */
997 continue;
998 }
999 if (candidate && candidate->ip_equals(candidate, addr->ip))
1000 { /* stop if we find the candidate */
1001 best = addr;
1002 candidate_matched = TRUE;
1003 break;
1004 }
1005 else if (!best || is_address_better(this, best, addr, dest))
1006 {
1007 best = addr;
1008 }
1009 }
1010 addrs->destroy(addrs);
1011 if (ifindex || candidate_matched)
1012 {
1013 break;
1014 }
1015 }
1016 }
1017 ifaces->destroy(ifaces);
1018 return best ? best->ip->clone(best->ip) : NULL;
1019 }
1020
1021 /**
1022 * Get a non-virtual IP address on the given interface.
1023 *
1024 * If a candidate address is given, we first search for that address and if not
1025 * found return the address as above.
1026 * Returned host is a clone, has to be freed by caller.
1027 *
1028 * this->lock must be held when calling this function.
1029 */
1030 static host_t *get_interface_address(private_kernel_netlink_net_t *this,
1031 int ifindex, int family, host_t *dest,
1032 host_t *candidate)
1033 {
1034 return get_matching_address(this, &ifindex, family, chunk_empty, 0, dest,
1035 candidate);
1036 }
1037
1038 /**
1039 * Get a non-virtual IP address in the given subnet.
1040 *
1041 * If a candidate address is given, we first search for that address and if not
1042 * found return the address as above.
1043 * Returned host is a clone, has to be freed by caller.
1044 *
1045 * this->lock must be held when calling this function.
1046 */
1047 static host_t *get_subnet_address(private_kernel_netlink_net_t *this,
1048 int family, chunk_t net, uint8_t mask,
1049 host_t *dest, host_t *candidate)
1050 {
1051 return get_matching_address(this, NULL, family, net, mask, dest, candidate);
1052 }
1053
1054 /**
1055 * callback function that raises the delayed roam event
1056 */
1057 static job_requeue_t roam_event(private_kernel_netlink_net_t *this)
1058 {
1059 bool address;
1060
1061 this->roam_lock->lock(this->roam_lock);
1062 address = this->roam_address;
1063 this->roam_address = FALSE;
1064 this->roam_lock->unlock(this->roam_lock);
1065 charon->kernel->roam(charon->kernel, address);
1066 return JOB_REQUEUE_NONE;
1067 }
1068
1069 /**
1070 * fire a roaming event. we delay it for a bit and fire only one event
1071 * for multiple calls. otherwise we would create too many events.
1072 */
1073 static void fire_roam_event(private_kernel_netlink_net_t *this, bool address)
1074 {
1075 timeval_t now;
1076 job_t *job;
1077
1078 if (!this->roam_events)
1079 {
1080 return;
1081 }
1082
1083 time_monotonic(&now);
1084 this->roam_lock->lock(this->roam_lock);
1085 this->roam_address |= address;
1086 if (!timercmp(&now, &this->next_roam, >))
1087 {
1088 this->roam_lock->unlock(this->roam_lock);
1089 return;
1090 }
1091 timeval_add_ms(&now, ROAM_DELAY);
1092 this->next_roam = now;
1093 this->roam_lock->unlock(this->roam_lock);
1094
1095 job = (job_t*)callback_job_create((callback_job_cb_t)roam_event,
1096 this, NULL, NULL);
1097 lib->scheduler->schedule_job_ms(lib->scheduler, job, ROAM_DELAY);
1098 }
1099
1100 /**
1101 * check if an interface with a given index is up and usable
1102 *
1103 * this->lock must be locked when calling this function
1104 */
1105 static bool is_interface_up_and_usable(private_kernel_netlink_net_t *this,
1106 int index)
1107 {
1108 iface_entry_t *iface;
1109
1110 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_index,
1111 (void**)&iface, &index) == SUCCESS)
1112 {
1113 return iface_entry_up_and_usable(iface);
1114 }
1115 return FALSE;
1116 }
1117
1118 /**
1119 * unregister the current addr_entry_t from the hashtable it is stored in
1120 *
1121 * this->lock must be locked when calling this function
1122 */
1123 static void addr_entry_unregister(addr_entry_t *addr, iface_entry_t *iface,
1124 private_kernel_netlink_net_t *this)
1125 {
1126 if (addr->refcount)
1127 {
1128 addr_map_entry_remove(this->vips, addr, iface);
1129 this->condvar->broadcast(this->condvar);
1130 return;
1131 }
1132 addr_map_entry_remove(this->addrs, addr, iface);
1133 }
1134
1135 /**
1136 * process RTM_NEWLINK/RTM_DELLINK from kernel
1137 */
1138 static void process_link(private_kernel_netlink_net_t *this,
1139 struct nlmsghdr *hdr, bool event)
1140 {
1141 struct ifinfomsg* msg = NLMSG_DATA(hdr);
1142 struct rtattr *rta = IFLA_RTA(msg);
1143 size_t rtasize = IFLA_PAYLOAD (hdr);
1144 enumerator_t *enumerator;
1145 iface_entry_t *current, *entry = NULL;
1146 char *name = NULL;
1147 bool update = FALSE, update_routes = FALSE;
1148
1149 while (RTA_OK(rta, rtasize))
1150 {
1151 switch (rta->rta_type)
1152 {
1153 case IFLA_IFNAME:
1154 name = RTA_DATA(rta);
1155 break;
1156 }
1157 rta = RTA_NEXT(rta, rtasize);
1158 }
1159 if (!name)
1160 {
1161 name = "(unknown)";
1162 }
1163
1164 this->lock->write_lock(this->lock);
1165 switch (hdr->nlmsg_type)
1166 {
1167 case RTM_NEWLINK:
1168 {
1169 if (this->ifaces->find_first(this->ifaces,
1170 (void*)iface_entry_by_index, (void**)&entry,
1171 &msg->ifi_index) != SUCCESS)
1172 {
1173 INIT(entry,
1174 .ifindex = msg->ifi_index,
1175 .addrs = linked_list_create(),
1176 .usable = charon->kernel->is_interface_usable(
1177 charon->kernel, name),
1178 );
1179 this->ifaces->insert_last(this->ifaces, entry);
1180 }
1181 strncpy(entry->ifname, name, IFNAMSIZ);
1182 entry->ifname[IFNAMSIZ-1] = '\0';
1183 if (event && entry->usable)
1184 {
1185 if (!(entry->flags & IFF_UP) && (msg->ifi_flags & IFF_UP))
1186 {
1187 update = update_routes = TRUE;
1188 DBG1(DBG_KNL, "interface %s activated", name);
1189 }
1190 if ((entry->flags & IFF_UP) && !(msg->ifi_flags & IFF_UP))
1191 {
1192 update = TRUE;
1193 DBG1(DBG_KNL, "interface %s deactivated", name);
1194 }
1195 }
1196 entry->flags = msg->ifi_flags;
1197 break;
1198 }
1199 case RTM_DELLINK:
1200 {
1201 enumerator = this->ifaces->create_enumerator(this->ifaces);
1202 while (enumerator->enumerate(enumerator, &current))
1203 {
1204 if (current->ifindex == msg->ifi_index)
1205 {
1206 if (event && current->usable)
1207 {
1208 update = TRUE;
1209 DBG1(DBG_KNL, "interface %s deleted", current->ifname);
1210 }
1211 /* TODO: move virtual IPs installed on this interface to
1212 * another interface? */
1213 this->ifaces->remove_at(this->ifaces, enumerator);
1214 current->addrs->invoke_function(current->addrs,
1215 (void*)addr_entry_unregister, current, this);
1216 iface_entry_destroy(current);
1217 break;
1218 }
1219 }
1220 enumerator->destroy(enumerator);
1221 break;
1222 }
1223 }
1224 this->lock->unlock(this->lock);
1225
1226 if (update_routes && event)
1227 {
1228 queue_route_reinstall(this, strdup(name));
1229 }
1230
1231 if (update && event)
1232 {
1233 fire_roam_event(this, TRUE);
1234 }
1235 }
1236
1237 /**
1238 * process RTM_NEWADDR/RTM_DELADDR from kernel
1239 */
1240 static void process_addr(private_kernel_netlink_net_t *this,
1241 struct nlmsghdr *hdr, bool event)
1242 {
1243 struct ifaddrmsg* msg = NLMSG_DATA(hdr);
1244 struct rtattr *rta = IFA_RTA(msg);
1245 size_t rtasize = IFA_PAYLOAD (hdr);
1246 host_t *host = NULL;
1247 iface_entry_t *iface;
1248 chunk_t local = chunk_empty, address = chunk_empty;
1249 char *route_ifname = NULL;
1250 bool update = FALSE, found = FALSE, changed = FALSE;
1251
1252 while (RTA_OK(rta, rtasize))
1253 {
1254 switch (rta->rta_type)
1255 {
1256 case IFA_LOCAL:
1257 local.ptr = RTA_DATA(rta);
1258 local.len = RTA_PAYLOAD(rta);
1259 break;
1260 case IFA_ADDRESS:
1261 address.ptr = RTA_DATA(rta);
1262 address.len = RTA_PAYLOAD(rta);
1263 break;
1264 }
1265 rta = RTA_NEXT(rta, rtasize);
1266 }
1267
1268 /* For PPP interfaces, we need the IFA_LOCAL address,
1269 * IFA_ADDRESS is the peers address. But IFA_LOCAL is
1270 * not included in all cases (IPv6?), so fallback to IFA_ADDRESS. */
1271 if (local.ptr)
1272 {
1273 host = host_create_from_chunk(msg->ifa_family, local, 0);
1274 }
1275 else if (address.ptr)
1276 {
1277 host = host_create_from_chunk(msg->ifa_family, address, 0);
1278 }
1279
1280 if (host == NULL)
1281 { /* bad family? */
1282 return;
1283 }
1284
1285 this->lock->write_lock(this->lock);
1286 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_index,
1287 (void**)&iface, &msg->ifa_index) == SUCCESS)
1288 {
1289 addr_map_entry_t *entry, lookup = {
1290 .ip = host,
1291 .iface = iface,
1292 };
1293 addr_entry_t *addr;
1294
1295 entry = this->vips->get(this->vips, &lookup);
1296 if (entry)
1297 {
1298 if (hdr->nlmsg_type == RTM_NEWADDR)
1299 { /* mark as installed and signal waiting threads */
1300 entry->addr->installed = TRUE;
1301 }
1302 else
1303 { /* the address was already marked as uninstalled */
1304 addr = entry->addr;
1305 iface->addrs->remove(iface->addrs, addr, NULL);
1306 addr_map_entry_remove(this->vips, addr, iface);
1307 addr_entry_destroy(addr);
1308 }
1309 /* no roam events etc. for virtual IPs */
1310 this->condvar->broadcast(this->condvar);
1311 this->lock->unlock(this->lock);
1312 host->destroy(host);
1313 return;
1314 }
1315 entry = this->addrs->get(this->addrs, &lookup);
1316 if (entry)
1317 {
1318 if (hdr->nlmsg_type == RTM_DELADDR)
1319 {
1320 found = TRUE;
1321 addr = entry->addr;
1322 iface->addrs->remove(iface->addrs, addr, NULL);
1323 if (iface->usable)
1324 {
1325 changed = TRUE;
1326 DBG1(DBG_KNL, "%H disappeared from %s", host,
1327 iface->ifname);
1328 }
1329 addr_map_entry_remove(this->addrs, addr, iface);
1330 addr_entry_destroy(addr);
1331 }
1332 }
1333 else
1334 {
1335 if (hdr->nlmsg_type == RTM_NEWADDR)
1336 {
1337 found = TRUE;
1338 changed = TRUE;
1339 route_ifname = strdup(iface->ifname);
1340 INIT(addr,
1341 .ip = host->clone(host),
1342 .flags = msg->ifa_flags,
1343 .scope = msg->ifa_scope,
1344 );
1345 iface->addrs->insert_last(iface->addrs, addr);
1346 addr_map_entry_add(this->addrs, addr, iface);
1347 if (event && iface->usable)
1348 {
1349 DBG1(DBG_KNL, "%H appeared on %s", host, iface->ifname);
1350 }
1351 }
1352 }
1353 if (found && (iface->flags & IFF_UP))
1354 {
1355 update = TRUE;
1356 }
1357 if (!iface->usable)
1358 { /* ignore events for interfaces excluded by config */
1359 update = changed = FALSE;
1360 }
1361 }
1362 this->lock->unlock(this->lock);
1363
1364 if (update && event && route_ifname)
1365 {
1366 queue_route_reinstall(this, route_ifname);
1367 }
1368 else
1369 {
1370 free(route_ifname);
1371 }
1372 host->destroy(host);
1373
1374 /* send an update to all IKE_SAs */
1375 if (update && event && changed)
1376 {
1377 fire_roam_event(this, TRUE);
1378 }
1379 }
1380
1381 /**
1382 * process RTM_NEWROUTE and RTM_DELROUTE from kernel
1383 */
1384 static void process_route(private_kernel_netlink_net_t *this, struct nlmsghdr *hdr)
1385 {
1386 struct rtmsg* msg = NLMSG_DATA(hdr);
1387 struct rtattr *rta = RTM_RTA(msg);
1388 size_t rtasize = RTM_PAYLOAD(hdr);
1389 uint32_t rta_oif = 0;
1390 host_t *host = NULL;
1391
1392 /* ignore routes added by us or in the local routing table (local addrs) */
1393 if (msg->rtm_table && (msg->rtm_table == this->routing_table ||
1394 msg->rtm_table == RT_TABLE_LOCAL))
1395 {
1396 return;
1397 }
1398 else if (msg->rtm_flags & RTM_F_CLONED)
1399 { /* ignore cached routes, seem to be created a lot for IPv6 */
1400 return;
1401 }
1402
1403 while (RTA_OK(rta, rtasize))
1404 {
1405 switch (rta->rta_type)
1406 {
1407 case RTA_PREFSRC:
1408 DESTROY_IF(host);
1409 host = host_create_from_chunk(msg->rtm_family,
1410 chunk_create(RTA_DATA(rta), RTA_PAYLOAD(rta)), 0);
1411 break;
1412 case RTA_OIF:
1413 if (RTA_PAYLOAD(rta) == sizeof(rta_oif))
1414 {
1415 rta_oif = *(uint32_t*)RTA_DATA(rta);
1416 }
1417 break;
1418 }
1419 rta = RTA_NEXT(rta, rtasize);
1420 }
1421 this->lock->read_lock(this->lock);
1422 if (rta_oif && !is_interface_up_and_usable(this, rta_oif))
1423 { /* ignore route changes for interfaces that are ignored or down */
1424 this->lock->unlock(this->lock);
1425 DESTROY_IF(host);
1426 return;
1427 }
1428 if (!host && rta_oif)
1429 {
1430 host = get_interface_address(this, rta_oif, msg->rtm_family,
1431 NULL, NULL);
1432 }
1433 if (!host || is_known_vip(this, host))
1434 { /* ignore routes added for virtual IPs */
1435 this->lock->unlock(this->lock);
1436 DESTROY_IF(host);
1437 return;
1438 }
1439 this->lock->unlock(this->lock);
1440 fire_roam_event(this, FALSE);
1441 host->destroy(host);
1442 }
1443
1444 /**
1445 * Receives events from kernel
1446 */
1447 static bool receive_events(private_kernel_netlink_net_t *this, int fd,
1448 watcher_event_t event)
1449 {
1450 char response[1536];
1451 struct nlmsghdr *hdr = (struct nlmsghdr*)response;
1452 struct sockaddr_nl addr;
1453 socklen_t addr_len = sizeof(addr);
1454 int len;
1455
1456 len = recvfrom(this->socket_events, response, sizeof(response),
1457 MSG_DONTWAIT, (struct sockaddr*)&addr, &addr_len);
1458 if (len < 0)
1459 {
1460 switch (errno)
1461 {
1462 case EINTR:
1463 /* interrupted, try again */
1464 return TRUE;
1465 case EAGAIN:
1466 /* no data ready, select again */
1467 return TRUE;
1468 default:
1469 DBG1(DBG_KNL, "unable to receive from RT event socket %s (%d)",
1470 strerror(errno), errno);
1471 sleep(1);
1472 return TRUE;
1473 }
1474 }
1475
1476 if (addr.nl_pid != 0)
1477 { /* not from kernel. not interested, try another one */
1478 return TRUE;
1479 }
1480
1481 while (NLMSG_OK(hdr, len))
1482 {
1483 /* looks good so far, dispatch netlink message */
1484 switch (hdr->nlmsg_type)
1485 {
1486 case RTM_NEWADDR:
1487 case RTM_DELADDR:
1488 process_addr(this, hdr, TRUE);
1489 break;
1490 case RTM_NEWLINK:
1491 case RTM_DELLINK:
1492 process_link(this, hdr, TRUE);
1493 break;
1494 case RTM_NEWROUTE:
1495 case RTM_DELROUTE:
1496 if (this->process_route)
1497 {
1498 process_route(this, hdr);
1499 }
1500 break;
1501 default:
1502 break;
1503 }
1504 hdr = NLMSG_NEXT(hdr, len);
1505 }
1506 return TRUE;
1507 }
1508
1509 /** enumerator over addresses */
1510 typedef struct {
1511 private_kernel_netlink_net_t* this;
1512 /** which addresses to enumerate */
1513 kernel_address_type_t which;
1514 } address_enumerator_t;
1515
1516 /**
1517 * cleanup function for address enumerator
1518 */
1519 static void address_enumerator_destroy(address_enumerator_t *data)
1520 {
1521 data->this->lock->unlock(data->this->lock);
1522 free(data);
1523 }
1524
1525 /**
1526 * filter for addresses
1527 */
1528 static bool filter_addresses(address_enumerator_t *data,
1529 addr_entry_t** in, host_t** out)
1530 {
1531 if (!(data->which & ADDR_TYPE_VIRTUAL) && (*in)->refcount)
1532 { /* skip virtual interfaces added by us */
1533 return FALSE;
1534 }
1535 if (!(data->which & ADDR_TYPE_REGULAR) && !(*in)->refcount)
1536 { /* address is regular, but not requested */
1537 return FALSE;
1538 }
1539 if ((*in)->scope >= RT_SCOPE_LINK)
1540 { /* skip addresses with a unusable scope */
1541 return FALSE;
1542 }
1543 *out = (*in)->ip;
1544 return TRUE;
1545 }
1546
1547 /**
1548 * enumerator constructor for interfaces
1549 */
1550 static enumerator_t *create_iface_enumerator(iface_entry_t *iface,
1551 address_enumerator_t *data)
1552 {
1553 return enumerator_create_filter(
1554 iface->addrs->create_enumerator(iface->addrs),
1555 (void*)filter_addresses, data, NULL);
1556 }
1557
1558 /**
1559 * filter for interfaces
1560 */
1561 static bool filter_interfaces(address_enumerator_t *data, iface_entry_t** in,
1562 iface_entry_t** out)
1563 {
1564 if (!(data->which & ADDR_TYPE_IGNORED) && !(*in)->usable)
1565 { /* skip interfaces excluded by config */
1566 return FALSE;
1567 }
1568 if (!(data->which & ADDR_TYPE_LOOPBACK) && ((*in)->flags & IFF_LOOPBACK))
1569 { /* ignore loopback devices */
1570 return FALSE;
1571 }
1572 if (!(data->which & ADDR_TYPE_DOWN) && !((*in)->flags & IFF_UP))
1573 { /* skip interfaces not up */
1574 return FALSE;
1575 }
1576 *out = *in;
1577 return TRUE;
1578 }
1579
1580 METHOD(kernel_net_t, create_address_enumerator, enumerator_t*,
1581 private_kernel_netlink_net_t *this, kernel_address_type_t which)
1582 {
1583 address_enumerator_t *data;
1584
1585 INIT(data,
1586 .this = this,
1587 .which = which,
1588 );
1589
1590 this->lock->read_lock(this->lock);
1591 return enumerator_create_nested(
1592 enumerator_create_filter(
1593 this->ifaces->create_enumerator(this->ifaces),
1594 (void*)filter_interfaces, data, NULL),
1595 (void*)create_iface_enumerator, data,
1596 (void*)address_enumerator_destroy);
1597 }
1598
1599 METHOD(kernel_net_t, get_interface_name, bool,
1600 private_kernel_netlink_net_t *this, host_t* ip, char **name)
1601 {
1602 addr_map_entry_t *entry, lookup = {
1603 .ip = ip,
1604 };
1605
1606 if (ip->is_anyaddr(ip))
1607 {
1608 return FALSE;
1609 }
1610 this->lock->read_lock(this->lock);
1611 /* first try to find it on an up and usable interface */
1612 entry = this->addrs->get_match(this->addrs, &lookup,
1613 (void*)addr_map_entry_match_up_and_usable);
1614 if (entry)
1615 {
1616 if (name)
1617 {
1618 *name = strdup(entry->iface->ifname);
1619 DBG2(DBG_KNL, "%H is on interface %s", ip, *name);
1620 }
1621 this->lock->unlock(this->lock);
1622 return TRUE;
1623 }
1624 /* in a second step, consider virtual IPs installed by us */
1625 entry = this->vips->get_match(this->vips, &lookup,
1626 (void*)addr_map_entry_match_up_and_usable);
1627 if (entry)
1628 {
1629 if (name)
1630 {
1631 *name = strdup(entry->iface->ifname);
1632 DBG2(DBG_KNL, "virtual IP %H is on interface %s", ip, *name);
1633 }
1634 this->lock->unlock(this->lock);
1635 return TRUE;
1636 }
1637 /* maybe it is installed on an ignored interface */
1638 entry = this->addrs->get_match(this->addrs, &lookup,
1639 (void*)addr_map_entry_match_up);
1640 if (!entry)
1641 {
1642 DBG2(DBG_KNL, "%H is not a local address or the interface is down", ip);
1643 }
1644 this->lock->unlock(this->lock);
1645 return FALSE;
1646 }
1647
1648 /**
1649 * get the index of an interface by name
1650 */
1651 static int get_interface_index(private_kernel_netlink_net_t *this, char* name)
1652 {
1653 iface_entry_t *iface;
1654 int ifindex = 0;
1655
1656 DBG2(DBG_KNL, "getting iface index for %s", name);
1657
1658 this->lock->read_lock(this->lock);
1659 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_name,
1660 (void**)&iface, name) == SUCCESS)
1661 {
1662 ifindex = iface->ifindex;
1663 }
1664 this->lock->unlock(this->lock);
1665
1666 if (ifindex == 0)
1667 {
1668 DBG1(DBG_KNL, "unable to get interface index for %s", name);
1669 }
1670 return ifindex;
1671 }
1672
1673 /**
1674 * get the name of an interface by index (allocated)
1675 */
1676 static char *get_interface_name_by_index(private_kernel_netlink_net_t *this,
1677 int index)
1678 {
1679 iface_entry_t *iface;
1680 char *name = NULL;
1681
1682 DBG2(DBG_KNL, "getting iface name for index %d", index);
1683
1684 this->lock->read_lock(this->lock);
1685 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_index,
1686 (void**)&iface, &index) == SUCCESS)
1687 {
1688 name = strdup(iface->ifname);
1689 }
1690 this->lock->unlock(this->lock);
1691
1692 if (!name)
1693 {
1694 DBG1(DBG_KNL, "unable to get interface name for %d", index);
1695 }
1696 return name;
1697 }
1698
1699 /**
1700 * Store information about a route retrieved via RTNETLINK
1701 */
1702 typedef struct {
1703 chunk_t gtw;
1704 chunk_t pref_src;
1705 chunk_t dst;
1706 chunk_t src;
1707 host_t *src_host;
1708 uint8_t dst_len;
1709 uint8_t src_len;
1710 uint32_t table;
1711 uint32_t oif;
1712 uint32_t priority;
1713 } rt_entry_t;
1714
1715 /**
1716 * Free a route entry
1717 */
1718 static void rt_entry_destroy(rt_entry_t *this)
1719 {
1720 DESTROY_IF(this->src_host);
1721 free(this);
1722 }
1723
1724 /**
1725 * Check if the route received with RTM_NEWROUTE is usable based on its type.
1726 */
1727 static bool route_usable(struct nlmsghdr *hdr)
1728 {
1729 struct rtmsg *msg;
1730
1731 msg = NLMSG_DATA(hdr);
1732 switch (msg->rtm_type)
1733 {
1734 case RTN_BLACKHOLE:
1735 case RTN_UNREACHABLE:
1736 case RTN_PROHIBIT:
1737 case RTN_THROW:
1738 return FALSE;
1739 default:
1740 return TRUE;
1741 }
1742 }
1743
1744 /**
1745 * Parse route received with RTM_NEWROUTE. The given rt_entry_t object will be
1746 * reused if not NULL.
1747 *
1748 * Returned chunks point to internal data of the Netlink message.
1749 */
1750 static rt_entry_t *parse_route(struct nlmsghdr *hdr, rt_entry_t *route)
1751 {
1752 struct rtattr *rta;
1753 struct rtmsg *msg;
1754 size_t rtasize;
1755
1756 msg = NLMSG_DATA(hdr);
1757 rta = RTM_RTA(msg);
1758 rtasize = RTM_PAYLOAD(hdr);
1759
1760 if (route)
1761 {
1762 route->gtw = chunk_empty;
1763 route->pref_src = chunk_empty;
1764 route->dst = chunk_empty;
1765 route->dst_len = msg->rtm_dst_len;
1766 route->src = chunk_empty;
1767 route->src_len = msg->rtm_src_len;
1768 route->table = msg->rtm_table;
1769 route->oif = 0;
1770 route->priority = 0;
1771 }
1772 else
1773 {
1774 INIT(route,
1775 .dst_len = msg->rtm_dst_len,
1776 .src_len = msg->rtm_src_len,
1777 .table = msg->rtm_table,
1778 );
1779 }
1780
1781 while (RTA_OK(rta, rtasize))
1782 {
1783 switch (rta->rta_type)
1784 {
1785 case RTA_PREFSRC:
1786 route->pref_src = chunk_create(RTA_DATA(rta), RTA_PAYLOAD(rta));
1787 break;
1788 case RTA_GATEWAY:
1789 route->gtw = chunk_create(RTA_DATA(rta), RTA_PAYLOAD(rta));
1790 break;
1791 case RTA_DST:
1792 route->dst = chunk_create(RTA_DATA(rta), RTA_PAYLOAD(rta));
1793 break;
1794 case RTA_SRC:
1795 route->src = chunk_create(RTA_DATA(rta), RTA_PAYLOAD(rta));
1796 break;
1797 case RTA_OIF:
1798 if (RTA_PAYLOAD(rta) == sizeof(route->oif))
1799 {
1800 route->oif = *(uint32_t*)RTA_DATA(rta);
1801 }
1802 break;
1803 case RTA_PRIORITY:
1804 if (RTA_PAYLOAD(rta) == sizeof(route->priority))
1805 {
1806 route->priority = *(uint32_t*)RTA_DATA(rta);
1807 }
1808 break;
1809 #ifdef HAVE_RTA_TABLE
1810 case RTA_TABLE:
1811 if (RTA_PAYLOAD(rta) == sizeof(route->table))
1812 {
1813 route->table = *(uint32_t*)RTA_DATA(rta);
1814 }
1815 break;
1816 #endif /* HAVE_RTA_TABLE*/
1817 }
1818 rta = RTA_NEXT(rta, rtasize);
1819 }
1820 return route;
1821 }
1822
1823 /**
1824 * Get a route: If "nexthop", the nexthop is returned. source addr otherwise.
1825 */
1826 static host_t *get_route(private_kernel_netlink_net_t *this, host_t *dest,
1827 int prefix, bool nexthop, host_t *candidate,
1828 char **iface, u_int recursion)
1829 {
1830 netlink_buf_t request;
1831 struct nlmsghdr *hdr, *out, *current;
1832 struct rtmsg *msg;
1833 chunk_t chunk;
1834 size_t len;
1835 linked_list_t *routes;
1836 rt_entry_t *route = NULL, *best = NULL;
1837 enumerator_t *enumerator;
1838 host_t *addr = NULL;
1839 bool match_net;
1840 int family;
1841
1842 if (recursion > MAX_ROUTE_RECURSION)
1843 {
1844 return NULL;
1845 }
1846 chunk = dest->get_address(dest);
1847 len = chunk.len * 8;
1848 prefix = prefix < 0 ? len : min(prefix, len);
1849 match_net = prefix != len;
1850
1851 memset(&request, 0, sizeof(request));
1852
1853 family = dest->get_family(dest);
1854 hdr = &request.hdr;
1855 hdr->nlmsg_flags = NLM_F_REQUEST;
1856 hdr->nlmsg_type = RTM_GETROUTE;
1857 hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
1858
1859 msg = NLMSG_DATA(hdr);
1860 msg->rtm_family = family;
1861 if (!match_net && this->rta_mark && this->routing_mark.value)
1862 {
1863 /* if our routing rule excludes packets with a certain mark we can
1864 * get the preferred route without having to dump all routes */
1865 chunk = chunk_from_thing(this->routing_mark.value);
1866 netlink_add_attribute(hdr, RTA_MARK, chunk, sizeof(request));
1867 }
1868 else if (family == AF_INET || this->rta_prefsrc_for_ipv6 ||
1869 this->routing_table || match_net)
1870 { /* kernels prior to 3.0 do not support RTA_PREFSRC for IPv6 routes.
1871 * as we want to ignore routes with virtual IPs we cannot use DUMP
1872 * if these routes are not installed in a separate table */
1873 hdr->nlmsg_flags |= NLM_F_DUMP;
1874 }
1875 if (candidate)
1876 {
1877 chunk = candidate->get_address(candidate);
1878 netlink_add_attribute(hdr, RTA_PREFSRC, chunk, sizeof(request));
1879 }
1880 /* we use this below to match against the routes */
1881 chunk = dest->get_address(dest);
1882 if (!match_net)
1883 {
1884 netlink_add_attribute(hdr, RTA_DST, chunk, sizeof(request));
1885 }
1886
1887 if (this->socket->send(this->socket, hdr, &out, &len) != SUCCESS)
1888 {
1889 DBG2(DBG_KNL, "getting %s to reach %H/%d failed",
1890 nexthop ? "nexthop" : "address", dest, prefix);
1891 return NULL;
1892 }
1893 routes = linked_list_create();
1894 this->lock->read_lock(this->lock);
1895
1896 for (current = out; NLMSG_OK(current, len);
1897 current = NLMSG_NEXT(current, len))
1898 {
1899 switch (current->nlmsg_type)
1900 {
1901 case NLMSG_DONE:
1902 break;
1903 case RTM_NEWROUTE:
1904 {
1905 rt_entry_t *other;
1906 uintptr_t table;
1907
1908 if (!route_usable(current))
1909 {
1910 continue;
1911 }
1912 route = parse_route(current, route);
1913
1914 table = (uintptr_t)route->table;
1915 if (this->rt_exclude->find_first(this->rt_exclude, NULL,
1916 (void**)&table) == SUCCESS)
1917 { /* route is from an excluded routing table */
1918 continue;
1919 }
1920 if (this->routing_table != 0 &&
1921 route->table == this->routing_table)
1922 { /* route is from our own ipsec routing table */
1923 continue;
1924 }
1925 if (route->oif && !is_interface_up_and_usable(this, route->oif))
1926 { /* interface is down */
1927 continue;
1928 }
1929 if (!addr_in_subnet(chunk, prefix, route->dst, route->dst_len))
1930 { /* route destination does not contain dest */
1931 continue;
1932 }
1933 if (route->pref_src.ptr)
1934 { /* verify source address, if any */
1935 host_t *src = host_create_from_chunk(msg->rtm_family,
1936 route->pref_src, 0);
1937 if (src && is_known_vip(this, src))
1938 { /* ignore routes installed by us */
1939 src->destroy(src);
1940 continue;
1941 }
1942 route->src_host = src;
1943 }
1944 /* insert route, sorted by network prefix and priority */
1945 enumerator = routes->create_enumerator(routes);
1946 while (enumerator->enumerate(enumerator, &other))
1947 {
1948 if (route->dst_len > other->dst_len)
1949 {
1950 break;
1951 }
1952 if (route->dst_len == other->dst_len &&
1953 route->priority < other->priority)
1954 {
1955 break;
1956 }
1957 }
1958 routes->insert_before(routes, enumerator, route);
1959 enumerator->destroy(enumerator);
1960 route = NULL;
1961 continue;
1962 }
1963 default:
1964 continue;
1965 }
1966 break;
1967 }
1968 if (route)
1969 {
1970 rt_entry_destroy(route);
1971 }
1972
1973 /* now we have a list of routes matching dest, sorted by net prefix.
1974 * we will look for source addresses for these routes and select the one
1975 * with the preferred source address, if possible */
1976 enumerator = routes->create_enumerator(routes);
1977 while (enumerator->enumerate(enumerator, &route))
1978 {
1979 if (route->src_host)
1980 { /* got a source address with the route, if no preferred source
1981 * is given or it matches we are done, as this is the best route */
1982 if (!candidate || candidate->ip_equals(candidate, route->src_host))
1983 {
1984 best = route;
1985 break;
1986 }
1987 else if (route->oif)
1988 { /* no match yet, maybe it is assigned to the same interface */
1989 host_t *src = get_interface_address(this, route->oif,
1990 msg->rtm_family, dest, candidate);
1991 if (src && src->ip_equals(src, candidate))
1992 {
1993 route->src_host->destroy(route->src_host);
1994 route->src_host = src;
1995 best = route;
1996 break;
1997 }
1998 DESTROY_IF(src);
1999 }
2000 /* no luck yet with the source address. if this is the best (first)
2001 * route we store it as fallback in case we don't find a route with
2002 * the preferred source */
2003 best = best ?: route;
2004 continue;
2005 }
2006 if (route->src.ptr)
2007 { /* no src, but a source selector, try to find a matching address */
2008 route->src_host = get_subnet_address(this, msg->rtm_family,
2009 route->src, route->src_len, dest,
2010 candidate);
2011 if (route->src_host)
2012 { /* we handle this address the same as the one above */
2013 if (!candidate ||
2014 candidate->ip_equals(candidate, route->src_host))
2015 {
2016 best = route;
2017 break;
2018 }
2019 best = best ?: route;
2020 continue;
2021 }
2022 }
2023 if (route->oif)
2024 { /* no src, but an interface - get address from it */
2025 route->src_host = get_interface_address(this, route->oif,
2026 msg->rtm_family, dest, candidate);
2027 if (route->src_host)
2028 { /* more of the same */
2029 if (!candidate ||
2030 candidate->ip_equals(candidate, route->src_host))
2031 {
2032 best = route;
2033 break;
2034 }
2035 best = best ?: route;
2036 continue;
2037 }
2038 }
2039 if (route->gtw.ptr)
2040 { /* no src, no iface, but a gateway - lookup src to reach gtw */
2041 host_t *gtw;
2042
2043 gtw = host_create_from_chunk(msg->rtm_family, route->gtw, 0);
2044 if (gtw && !gtw->ip_equals(gtw, dest))
2045 {
2046 route->src_host = get_route(this, gtw, -1, FALSE, candidate,
2047 iface, recursion + 1);
2048 }
2049 DESTROY_IF(gtw);
2050 if (route->src_host)
2051 { /* more of the same */
2052 if (!candidate ||
2053 candidate->ip_equals(candidate, route->src_host))
2054 {
2055 best = route;
2056 break;
2057 }
2058 best = best ?: route;
2059 }
2060 }
2061 }
2062 enumerator->destroy(enumerator);
2063
2064 if (nexthop)
2065 { /* nexthop lookup, return gateway and oif if any */
2066 if (iface)
2067 {
2068 *iface = NULL;
2069 }
2070 if (best || routes->get_first(routes, (void**)&best) == SUCCESS)
2071 {
2072 addr = host_create_from_chunk(msg->rtm_family, best->gtw, 0);
2073 if (iface && best->oif)
2074 {
2075 *iface = get_interface_name_by_index(this, best->oif);
2076 }
2077 }
2078 if (!addr && !match_net)
2079 { /* fallback to destination address */
2080 addr = dest->clone(dest);
2081 }
2082 }
2083 else
2084 {
2085 if (best)
2086 {
2087 addr = best->src_host->clone(best->src_host);
2088 }
2089 }
2090 this->lock->unlock(this->lock);
2091 routes->destroy_function(routes, (void*)rt_entry_destroy);
2092 free(out);
2093
2094 if (addr)
2095 {
2096 if (nexthop && iface && *iface)
2097 {
2098 DBG2(DBG_KNL, "using %H as nexthop and %s as dev to reach %H/%d",
2099 addr, *iface, dest, prefix);
2100 }
2101 else
2102 {
2103 DBG2(DBG_KNL, "using %H as %s to reach %H/%d", addr,
2104 nexthop ? "nexthop" : "address", dest, prefix);
2105 }
2106 }
2107 else if (!recursion)
2108 {
2109 DBG2(DBG_KNL, "no %s found to reach %H/%d",
2110 nexthop ? "nexthop" : "address", dest, prefix);
2111 }
2112 return addr;
2113 }
2114
2115 METHOD(kernel_net_t, get_source_addr, host_t*,
2116 private_kernel_netlink_net_t *this, host_t *dest, host_t *src)
2117 {
2118 return get_route(this, dest, -1, FALSE, src, NULL, 0);
2119 }
2120
2121 METHOD(kernel_net_t, get_nexthop, host_t*,
2122 private_kernel_netlink_net_t *this, host_t *dest, int prefix, host_t *src,
2123 char **iface)
2124 {
2125 return get_route(this, dest, prefix, TRUE, src, iface, 0);
2126 }
2127
2128 /**
2129 * Manages the creation and deletion of ip addresses on an interface.
2130 * By setting the appropriate nlmsg_type, the ip will be set or unset.
2131 */
2132 static status_t manage_ipaddr(private_kernel_netlink_net_t *this, int nlmsg_type,
2133 int flags, int if_index, host_t *ip, int prefix)
2134 {
2135 netlink_buf_t request;
2136 struct nlmsghdr *hdr;
2137 struct ifaddrmsg *msg;
2138 chunk_t chunk;
2139
2140 memset(&request, 0, sizeof(request));
2141
2142 chunk = ip->get_address(ip);
2143
2144 hdr = &request.hdr;
2145 hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK | flags;
2146 hdr->nlmsg_type = nlmsg_type;
2147 hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct ifaddrmsg));
2148
2149 msg = NLMSG_DATA(hdr);
2150 msg->ifa_family = ip->get_family(ip);
2151 msg->ifa_flags = 0;
2152 msg->ifa_prefixlen = prefix < 0 ? chunk.len * 8 : prefix;
2153 msg->ifa_scope = RT_SCOPE_UNIVERSE;
2154 msg->ifa_index = if_index;
2155
2156 netlink_add_attribute(hdr, IFA_LOCAL, chunk, sizeof(request));
2157
2158 if (ip->get_family(ip) == AF_INET6)
2159 {
2160 msg->ifa_flags |= IFA_F_NODAD;
2161 if (this->rta_prefsrc_for_ipv6)
2162 {
2163 /* if source routes are possible we let the virtual IP get
2164 * deprecated immediately (but mark it as valid forever) so it gets
2165 * only used if forced by our route, and not by the default IPv6
2166 * address selection */
2167 struct ifa_cacheinfo cache = {
2168 .ifa_valid = 0xFFFFFFFF,
2169 .ifa_prefered = 0,
2170 };
2171 netlink_add_attribute(hdr, IFA_CACHEINFO, chunk_from_thing(cache),
2172 sizeof(request));
2173 }
2174 }
2175 return this->socket->send_ack(this->socket, hdr);
2176 }
2177
2178 METHOD(kernel_net_t, add_ip, status_t,
2179 private_kernel_netlink_net_t *this, host_t *virtual_ip, int prefix,
2180 char *iface_name)
2181 {
2182 addr_map_entry_t *entry, lookup = {
2183 .ip = virtual_ip,
2184 };
2185 iface_entry_t *iface = NULL;
2186
2187 if (!this->install_virtual_ip)
2188 { /* disabled by config */
2189 return SUCCESS;
2190 }
2191
2192 this->lock->write_lock(this->lock);
2193 /* the virtual IP might actually be installed as regular IP, in which case
2194 * we don't track it as virtual IP */
2195 entry = this->addrs->get_match(this->addrs, &lookup,
2196 (void*)addr_map_entry_match);
2197 if (!entry)
2198 { /* otherwise it might already be installed as virtual IP */
2199 entry = this->vips->get_match(this->vips, &lookup,
2200 (void*)addr_map_entry_match);
2201 if (entry)
2202 { /* the vip we found can be in one of three states: 1) installed and
2203 * ready, 2) just added by another thread, but not yet confirmed to
2204 * be installed by the kernel, 3) just deleted, but not yet gone.
2205 * Then while we wait below, several things could happen (as we
2206 * release the lock). For instance, the interface could disappear,
2207 * or the IP is finally deleted, and it reappears on a different
2208 * interface. All these cases are handled by the call below. */
2209 while (!is_vip_installed_or_gone(this, virtual_ip, &entry))
2210 {
2211 this->condvar->wait(this->condvar, this->lock);
2212 }
2213 if (entry)
2214 {
2215 entry->addr->refcount++;
2216 }
2217 }
2218 }
2219 if (entry)
2220 {
2221 DBG2(DBG_KNL, "virtual IP %H is already installed on %s", virtual_ip,
2222 entry->iface->ifname);
2223 this->lock->unlock(this->lock);
2224 return SUCCESS;
2225 }
2226 /* try to find the target interface, either by config or via src ip */
2227 if (!this->install_virtual_ip_on ||
2228 this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_name,
2229 (void**)&iface, this->install_virtual_ip_on) != SUCCESS)
2230 {
2231 if (this->ifaces->find_first(this->ifaces, (void*)iface_entry_by_name,
2232 (void**)&iface, iface_name) != SUCCESS)
2233 { /* if we don't find the requested interface we just use the first */
2234 this->ifaces->get_first(this->ifaces, (void**)&iface);
2235 }
2236 }
2237 if (iface)
2238 {
2239 addr_entry_t *addr;
2240 char *ifname;
2241 int ifi;
2242
2243 INIT(addr,
2244 .ip = virtual_ip->clone(virtual_ip),
2245 .refcount = 1,
2246 .scope = RT_SCOPE_UNIVERSE,
2247 );
2248 iface->addrs->insert_last(iface->addrs, addr);
2249 addr_map_entry_add(this->vips, addr, iface);
2250 ifi = iface->ifindex;
2251 this->lock->unlock(this->lock);
2252 if (manage_ipaddr(this, RTM_NEWADDR, NLM_F_CREATE | NLM_F_EXCL,
2253 ifi, virtual_ip, prefix) == SUCCESS)
2254 {
2255 this->lock->write_lock(this->lock);
2256 while (!is_vip_installed_or_gone(this, virtual_ip, &entry))
2257 { /* wait until address appears */
2258 this->condvar->wait(this->condvar, this->lock);
2259 }
2260 if (entry)
2261 { /* we fail if the interface got deleted in the meantime */
2262 ifname = strdup(entry->iface->ifname);
2263 this->lock->unlock(this->lock);
2264 DBG2(DBG_KNL, "virtual IP %H installed on %s",
2265 virtual_ip, ifname);
2266 /* during IKEv1 reauthentication, children get moved from
2267 * old the new SA before the virtual IP is available. This
2268 * kills the route for our virtual IP, reinstall. */
2269 queue_route_reinstall(this, ifname);
2270 return SUCCESS;
2271 }
2272 this->lock->unlock(this->lock);
2273 }
2274 DBG1(DBG_KNL, "adding virtual IP %H failed", virtual_ip);
2275 return FAILED;
2276 }
2277 this->lock->unlock(this->lock);
2278 DBG1(DBG_KNL, "no interface available, unable to install virtual IP %H",
2279 virtual_ip);
2280 return FAILED;
2281 }
2282
2283 METHOD(kernel_net_t, del_ip, status_t,
2284 private_kernel_netlink_net_t *this, host_t *virtual_ip, int prefix,
2285 bool wait)
2286 {
2287 addr_map_entry_t *entry, lookup = {
2288 .ip = virtual_ip,
2289 };
2290
2291 if (!this->install_virtual_ip)
2292 { /* disabled by config */
2293 return SUCCESS;
2294 }
2295
2296 DBG2(DBG_KNL, "deleting virtual IP %H", virtual_ip);
2297
2298 this->lock->write_lock(this->lock);
2299 entry = this->vips->get_match(this->vips, &lookup,
2300 (void*)addr_map_entry_match);
2301 if (!entry)
2302 { /* we didn't install this IP as virtual IP */
2303 entry = this->addrs->get_match(this->addrs, &lookup,
2304 (void*)addr_map_entry_match);
2305 if (entry)
2306 {
2307 DBG2(DBG_KNL, "not deleting existing IP %H on %s", virtual_ip,
2308 entry->iface->ifname);
2309 this->lock->unlock(this->lock);
2310 return SUCCESS;
2311 }
2312 DBG2(DBG_KNL, "virtual IP %H not cached, unable to delete", virtual_ip);
2313 this->lock->unlock(this->lock);
2314 return FAILED;
2315 }
2316 if (entry->addr->refcount == 1)
2317 {
2318 status_t status;
2319 int ifi;
2320
2321 /* we set this flag so that threads calling add_ip will block and wait
2322 * until the entry is gone, also so we can wait below */
2323 entry->addr->installed = FALSE;
2324 ifi = entry->iface->ifindex;
2325 this->lock->unlock(this->lock);
2326 status = manage_ipaddr(this, RTM_DELADDR, 0, ifi, virtual_ip, prefix);
2327 if (status == SUCCESS && wait)
2328 { /* wait until the address is really gone */
2329 this->lock->write_lock(this->lock);
2330 while (is_known_vip(this, virtual_ip))
2331 {
2332 this->condvar->wait(this->condvar, this->lock);
2333 }
2334 this->lock->unlock(this->lock);
2335 }
2336 return status;
2337 }
2338 else
2339 {
2340 entry->addr->refcount--;
2341 }
2342 DBG2(DBG_KNL, "virtual IP %H used by other SAs, not deleting",
2343 virtual_ip);
2344 this->lock->unlock(this->lock);
2345 return SUCCESS;
2346 }
2347
2348 /**
2349 * Manages source routes in the routing table.
2350 * By setting the appropriate nlmsg_type, the route gets added or removed.
2351 */
2352 static status_t manage_srcroute(private_kernel_netlink_net_t *this,
2353 int nlmsg_type, int flags, chunk_t dst_net,
2354 uint8_t prefixlen, host_t *gateway,
2355 host_t *src_ip, char *if_name)
2356 {
2357 netlink_buf_t request;
2358 struct nlmsghdr *hdr;
2359 struct rtmsg *msg;
2360 struct rtattr *rta;
2361 int ifindex;
2362 chunk_t chunk;
2363
2364 /* if route is 0.0.0.0/0, we can't install it, as it would
2365 * overwrite the default route. Instead, we add two routes:
2366 * 0.0.0.0/1 and 128.0.0.0/1 */
2367 if (this->routing_table == 0 && prefixlen == 0)
2368 {
2369 chunk_t half_net;
2370 uint8_t half_prefixlen;
2371 status_t status;
2372
2373 half_net = chunk_alloca(dst_net.len);
2374 memset(half_net.ptr, 0, half_net.len);
2375 half_prefixlen = 1;
2376
2377 status = manage_srcroute(this, nlmsg_type, flags, half_net, half_prefixlen,
2378 gateway, src_ip, if_name);
2379 half_net.ptr[0] |= 0x80;
2380 status = manage_srcroute(this, nlmsg_type, flags, half_net, half_prefixlen,
2381 gateway, src_ip, if_name);
2382 return status;
2383 }
2384
2385 memset(&request, 0, sizeof(request));
2386
2387 hdr = &request.hdr;
2388 hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK | flags;
2389 hdr->nlmsg_type = nlmsg_type;
2390 hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
2391
2392 msg = NLMSG_DATA(hdr);
2393 msg->rtm_family = src_ip->get_family(src_ip);
2394 msg->rtm_dst_len = prefixlen;
2395 msg->rtm_table = this->routing_table;
2396 msg->rtm_protocol = RTPROT_STATIC;
2397 msg->rtm_type = RTN_UNICAST;
2398 msg->rtm_scope = RT_SCOPE_UNIVERSE;
2399
2400 netlink_add_attribute(hdr, RTA_DST, dst_net, sizeof(request));
2401 chunk = src_ip->get_address(src_ip);
2402 netlink_add_attribute(hdr, RTA_PREFSRC, chunk, sizeof(request));
2403 if (gateway && gateway->get_family(gateway) == src_ip->get_family(src_ip))
2404 {
2405 chunk = gateway->get_address(gateway);
2406 netlink_add_attribute(hdr, RTA_GATEWAY, chunk, sizeof(request));
2407 }
2408 ifindex = get_interface_index(this, if_name);
2409 chunk.ptr = (char*)&ifindex;
2410 chunk.len = sizeof(ifindex);
2411 netlink_add_attribute(hdr, RTA_OIF, chunk, sizeof(request));
2412
2413 if (this->mtu || this->mss)
2414 {
2415 chunk = chunk_alloca(RTA_LENGTH((sizeof(struct rtattr) +
2416 sizeof(uint32_t)) * 2));
2417 chunk.len = 0;
2418 rta = (struct rtattr*)chunk.ptr;
2419 if (this->mtu)
2420 {
2421 rta->rta_type = RTAX_MTU;
2422 rta->rta_len = RTA_LENGTH(sizeof(uint32_t));
2423 memcpy(RTA_DATA(rta), &this->mtu, sizeof(uint32_t));
2424 chunk.len = rta->rta_len;
2425 }
2426 if (this->mss)
2427 {
2428 rta = (struct rtattr*)(chunk.ptr + RTA_ALIGN(chunk.len));
2429 rta->rta_type = RTAX_ADVMSS;
2430 rta->rta_len = RTA_LENGTH(sizeof(uint32_t));
2431 memcpy(RTA_DATA(rta), &this->mss, sizeof(uint32_t));
2432 chunk.len = RTA_ALIGN(chunk.len) + rta->rta_len;
2433 }
2434 netlink_add_attribute(hdr, RTA_METRICS, chunk, sizeof(request));
2435 }
2436
2437 return this->socket->send_ack(this->socket, hdr);
2438 }
2439
2440 METHOD(kernel_net_t, add_route, status_t,
2441 private_kernel_netlink_net_t *this, chunk_t dst_net, uint8_t prefixlen,
2442 host_t *gateway, host_t *src_ip, char *if_name)
2443 {
2444 status_t status;
2445 route_entry_t *found, route = {
2446 .dst_net = dst_net,
2447 .prefixlen = prefixlen,
2448 .gateway = gateway,
2449 .src_ip = src_ip,
2450 .if_name = if_name,
2451 };
2452
2453 this->routes_lock->lock(this->routes_lock);
2454 found = this->routes->get(this->routes, &route);
2455 if (found)
2456 {
2457 this->routes_lock->unlock(this->routes_lock);
2458 return ALREADY_DONE;
2459 }
2460 status = manage_srcroute(this, RTM_NEWROUTE, NLM_F_CREATE | NLM_F_EXCL,
2461 dst_net, prefixlen, gateway, src_ip, if_name);
2462 if (status == SUCCESS)
2463 {
2464 found = route_entry_clone(&route);
2465 this->routes->put(this->routes, found, found);
2466 }
2467 this->routes_lock->unlock(this->routes_lock);
2468 return status;
2469 }
2470
2471 METHOD(kernel_net_t, del_route, status_t,
2472 private_kernel_netlink_net_t *this, chunk_t dst_net, uint8_t prefixlen,
2473 host_t *gateway, host_t *src_ip, char *if_name)
2474 {
2475 status_t status;
2476 route_entry_t *found, route = {
2477 .dst_net = dst_net,
2478 .prefixlen = prefixlen,
2479 .gateway = gateway,
2480 .src_ip = src_ip,
2481 .if_name = if_name,
2482 };
2483
2484 this->routes_lock->lock(this->routes_lock);
2485 found = this->routes->get(this->routes, &route);
2486 if (!found)
2487 {
2488 this->routes_lock->unlock(this->routes_lock);
2489 return NOT_FOUND;
2490 }
2491 this->routes->remove(this->routes, found);
2492 route_entry_destroy(found);
2493 status = manage_srcroute(this, RTM_DELROUTE, 0, dst_net, prefixlen,
2494 gateway, src_ip, if_name);
2495 this->routes_lock->unlock(this->routes_lock);
2496 return status;
2497 }
2498
2499 /**
2500 * Initialize a list of local addresses.
2501 */
2502 static status_t init_address_list(private_kernel_netlink_net_t *this)
2503 {
2504 netlink_buf_t request;
2505 struct nlmsghdr *out, *current, *in;
2506 struct rtgenmsg *msg;
2507 size_t len;
2508 enumerator_t *ifaces, *addrs;
2509 iface_entry_t *iface;
2510 addr_entry_t *addr;
2511
2512 DBG2(DBG_KNL, "known interfaces and IP addresses:");
2513
2514 memset(&request, 0, sizeof(request));
2515
2516 in = &request.hdr;
2517 in->nlmsg_len = NLMSG_LENGTH(sizeof(struct rtgenmsg));
2518 in->nlmsg_flags = NLM_F_REQUEST | NLM_F_MATCH | NLM_F_ROOT;
2519 msg = NLMSG_DATA(in);
2520 msg->rtgen_family = AF_UNSPEC;
2521
2522 /* get all links */
2523 in->nlmsg_type = RTM_GETLINK;
2524 if (this->socket->send(this->socket, in, &out, &len) != SUCCESS)
2525 {
2526 return FAILED;
2527 }
2528 current = out;
2529 while (NLMSG_OK(current, len))
2530 {
2531 switch (current->nlmsg_type)
2532 {
2533 case NLMSG_DONE:
2534 break;
2535 case RTM_NEWLINK:
2536 process_link(this, current, FALSE);
2537 /* fall through */
2538 default:
2539 current = NLMSG_NEXT(current, len);
2540 continue;
2541 }
2542 break;
2543 }
2544 free(out);
2545
2546 /* get all interface addresses */
2547 in->nlmsg_type = RTM_GETADDR;
2548 if (this->socket->send(this->socket, in, &out, &len) != SUCCESS)
2549 {
2550 return FAILED;
2551 }
2552 current = out;
2553 while (NLMSG_OK(current, len))
2554 {
2555 switch (current->nlmsg_type)
2556 {
2557 case NLMSG_DONE:
2558 break;
2559 case RTM_NEWADDR:
2560 process_addr(this, current, FALSE);
2561 /* fall through */
2562 default:
2563 current = NLMSG_NEXT(current, len);
2564 continue;
2565 }
2566 break;
2567 }
2568 free(out);
2569
2570 this->lock->read_lock(this->lock);
2571 ifaces = this->ifaces->create_enumerator(this->ifaces);
2572 while (ifaces->enumerate(ifaces, &iface))
2573 {
2574 if (iface_entry_up_and_usable(iface))
2575 {
2576 DBG2(DBG_KNL, " %s", iface->ifname);
2577 addrs = iface->addrs->create_enumerator(iface->addrs);
2578 while (addrs->enumerate(addrs, (void**)&addr))
2579 {
2580 DBG2(DBG_KNL, " %H", addr->ip);
2581 }
2582 addrs->destroy(addrs);
2583 }
2584 }
2585 ifaces->destroy(ifaces);
2586 this->lock->unlock(this->lock);
2587 return SUCCESS;
2588 }
2589
2590 /**
2591 * create or delete a rule to use our routing table
2592 */
2593 static status_t manage_rule(private_kernel_netlink_net_t *this, int nlmsg_type,
2594 int family, uint32_t table, uint32_t prio)
2595 {
2596 netlink_buf_t request;
2597 struct nlmsghdr *hdr;
2598 struct rtmsg *msg;
2599 chunk_t chunk;
2600 char *fwmark;
2601
2602 memset(&request, 0, sizeof(request));
2603 hdr = &request.hdr;
2604 hdr->nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK;
2605 hdr->nlmsg_type = nlmsg_type;
2606 if (nlmsg_type == RTM_NEWRULE)
2607 {
2608 hdr->nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
2609 }
2610 hdr->nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
2611
2612 msg = NLMSG_DATA(hdr);
2613 msg->rtm_table = table;
2614 msg->rtm_family = family;
2615 msg->rtm_protocol = RTPROT_BOOT;
2616 msg->rtm_scope = RT_SCOPE_UNIVERSE;
2617 msg->rtm_type = RTN_UNICAST;
2618
2619 chunk = chunk_from_thing(prio);
2620 netlink_add_attribute(hdr, RTA_PRIORITY, chunk, sizeof(request));
2621
2622 fwmark = lib->settings->get_str(lib->settings,
2623 "%s.plugins.kernel-netlink.fwmark", NULL, lib->ns);
2624 if (fwmark)
2625 {
2626 #ifdef HAVE_LINUX_FIB_RULES_H
2627 mark_t mark;
2628
2629 if (fwmark[0] == '!')
2630 {
2631 msg->rtm_flags |= FIB_RULE_INVERT;
2632 fwmark++;
2633 }
2634 if (mark_from_string(fwmark, &mark))
2635 {
2636 chunk = chunk_from_thing(mark.value);
2637 netlink_add_attribute(hdr, FRA_FWMARK, chunk, sizeof(request));
2638 chunk = chunk_from_thing(mark.mask);
2639 netlink_add_attribute(hdr, FRA_FWMASK, chunk, sizeof(request));
2640 if (msg->rtm_flags & FIB_RULE_INVERT)
2641 {
2642 this->routing_mark = mark;
2643 }
2644 }
2645 #else
2646 DBG1(DBG_KNL, "setting firewall mark on routing rule is not supported");
2647 #endif
2648 }
2649 return this->socket->send_ack(this->socket, hdr);
2650 }
2651
2652 /**
2653 * check for kernel features (currently only via version number)
2654 */
2655 static void check_kernel_features(private_kernel_netlink_net_t *this)
2656 {
2657 struct utsname utsname;
2658 int a, b, c;
2659
2660 if (uname(&utsname) == 0)
2661 {
2662 switch(sscanf(utsname.release, "%d.%d.%d", &a, &b, &c))
2663 {
2664 case 3:
2665 if (a == 2)
2666 {
2667 if (b == 6 && c >= 36)
2668 {
2669 this->rta_mark = TRUE;
2670 }
2671 DBG2(DBG_KNL, "detected Linux %d.%d.%d, no support for "
2672 "RTA_PREFSRC for IPv6 routes", a, b, c);
2673 break;
2674 }
2675 /* fall-through */
2676 case 2:
2677 /* only 3.x+ uses two part version numbers */
2678 this->rta_prefsrc_for_ipv6 = TRUE;
2679 this->rta_mark = TRUE;
2680 break;
2681 default:
2682 break;
2683 }
2684 }
2685 }
2686
2687 /**
2688 * Destroy an address to iface map
2689 */
2690 static void addr_map_destroy(hashtable_t *map)
2691 {
2692 enumerator_t *enumerator;
2693 addr_map_entry_t *addr;
2694
2695 enumerator = map->create_enumerator(map);
2696 while (enumerator->enumerate(enumerator, NULL, (void**)&addr))
2697 {
2698 free(addr);
2699 }
2700 enumerator->destroy(enumerator);
2701 map->destroy(map);
2702 }
2703
2704 METHOD(kernel_net_t, destroy, void,
2705 private_kernel_netlink_net_t *this)
2706 {
2707 enumerator_t *enumerator;
2708 route_entry_t *route;
2709
2710 if (this->routing_table)
2711 {
2712 manage_rule(this, RTM_DELRULE, AF_INET, this->routing_table,
2713 this->routing_table_prio);
2714 manage_rule(this, RTM_DELRULE, AF_INET6, this->routing_table,
2715 this->routing_table_prio);
2716 }
2717 if (this->socket_events > 0)
2718 {
2719 lib->watcher->remove(lib->watcher, this->socket_events);
2720 close(this->socket_events);
2721 }
2722 enumerator = this->routes->create_enumerator(this->routes);
2723 while (enumerator->enumerate(enumerator, NULL, (void**)&route))
2724 {
2725 manage_srcroute(this, RTM_DELROUTE, 0, route->dst_net, route->prefixlen,
2726 route->gateway, route->src_ip, route->if_name);
2727 route_entry_destroy(route);
2728 }
2729 enumerator->destroy(enumerator);
2730 this->routes->destroy(this->routes);
2731 this->routes_lock->destroy(this->routes_lock);
2732 DESTROY_IF(this->socket);
2733
2734 net_changes_clear(this);
2735 this->net_changes->destroy(this->net_changes);
2736 this->net_changes_lock->destroy(this->net_changes_lock);
2737
2738 addr_map_destroy(this->addrs);
2739 addr_map_destroy(this->vips);
2740
2741 this->ifaces->destroy_function(this->ifaces, (void*)iface_entry_destroy);
2742 this->rt_exclude->destroy(this->rt_exclude);
2743 this->roam_lock->destroy(this->roam_lock);
2744 this->condvar->destroy(this->condvar);
2745 this->lock->destroy(this->lock);
2746 free(this);
2747 }
2748
2749 /*
2750 * Described in header.
2751 */
2752 kernel_netlink_net_t *kernel_netlink_net_create()
2753 {
2754 private_kernel_netlink_net_t *this;
2755 enumerator_t *enumerator;
2756 bool register_for_events = TRUE;
2757 char *exclude;
2758
2759 INIT(this,
2760 .public = {
2761 .interface = {
2762 .get_interface = _get_interface_name,
2763 .create_address_enumerator = _create_address_enumerator,
2764 .get_source_addr = _get_source_addr,
2765 .get_nexthop = _get_nexthop,
2766 .add_ip = _add_ip,
2767 .del_ip = _del_ip,
2768 .add_route = _add_route,
2769 .del_route = _del_route,
2770 .destroy = _destroy,
2771 },
2772 },
2773 .socket = netlink_socket_create(NETLINK_ROUTE, rt_msg_names,
2774 lib->settings->get_bool(lib->settings,
2775 "%s.plugins.kernel-netlink.parallel_route", FALSE, lib->ns)),
2776 .rt_exclude = linked_list_create(),
2777 .routes = hashtable_create((hashtable_hash_t)route_entry_hash,
2778 (hashtable_equals_t)route_entry_equals, 16),
2779 .net_changes = hashtable_create(
2780 (hashtable_hash_t)net_change_hash,
2781 (hashtable_equals_t)net_change_equals, 16),
2782 .addrs = hashtable_create(
2783 (hashtable_hash_t)addr_map_entry_hash,
2784 (hashtable_equals_t)addr_map_entry_equals, 16),
2785 .vips = hashtable_create((hashtable_hash_t)addr_map_entry_hash,
2786 (hashtable_equals_t)addr_map_entry_equals, 16),
2787 .routes_lock = mutex_create(MUTEX_TYPE_DEFAULT),
2788 .net_changes_lock = mutex_create(MUTEX_TYPE_DEFAULT),
2789 .ifaces = linked_list_create(),
2790 .lock = rwlock_create(RWLOCK_TYPE_DEFAULT),
2791 .condvar = rwlock_condvar_create(),
2792 .roam_lock = spinlock_create(),
2793 .routing_table = lib->settings->get_int(lib->settings,
2794 "%s.routing_table", ROUTING_TABLE, lib->ns),
2795 .routing_table_prio = lib->settings->get_int(lib->settings,
2796 "%s.routing_table_prio", ROUTING_TABLE_PRIO, lib->ns),
2797 .process_route = lib->settings->get_bool(lib->settings,
2798 "%s.process_route", TRUE, lib->ns),
2799 .install_virtual_ip = lib->settings->get_bool(lib->settings,
2800 "%s.install_virtual_ip", TRUE, lib->ns),
2801 .install_virtual_ip_on = lib->settings->get_str(lib->settings,
2802 "%s.install_virtual_ip_on", NULL, lib->ns),
2803 .prefer_temporary_addrs = lib->settings->get_bool(lib->settings,
2804 "%s.prefer_temporary_addrs", FALSE, lib->ns),
2805 .roam_events = lib->settings->get_bool(lib->settings,
2806 "%s.plugins.kernel-netlink.roam_events", TRUE, lib->ns),
2807 .mtu = lib->settings->get_int(lib->settings,
2808 "%s.plugins.kernel-netlink.mtu", 0, lib->ns),
2809 .mss = lib->settings->get_int(lib->settings,
2810 "%s.plugins.kernel-netlink.mss", 0, lib->ns),
2811 );
2812 timerclear(&this->last_route_reinstall);
2813 timerclear(&this->next_roam);
2814
2815 check_kernel_features(this);
2816
2817 if (streq(lib->ns, "starter"))
2818 { /* starter has no threads, so we do not register for kernel events */
2819 register_for_events = FALSE;
2820 }
2821
2822 exclude = lib->settings->get_str(lib->settings,
2823 "%s.ignore_routing_tables", NULL, lib->ns);
2824 if (exclude)
2825 {
2826 char *token;
2827 uintptr_t table;
2828
2829 enumerator = enumerator_create_token(exclude, " ", " ");
2830 while (enumerator->enumerate(enumerator, &token))
2831 {
2832 errno = 0;
2833 table = strtoul(token, NULL, 10);
2834
2835 if (errno == 0)
2836 {
2837 this->rt_exclude->insert_last(this->rt_exclude, (void*)table);
2838 }
2839 }
2840 enumerator->destroy(enumerator);
2841 }
2842
2843 if (register_for_events)
2844 {
2845 struct sockaddr_nl addr;
2846
2847 memset(&addr, 0, sizeof(addr));
2848 addr.nl_family = AF_NETLINK