1 /* User-kernel network link */
3 /* See RFCs 791, 792, 1123 and 1812 */
5 /* The netlink device is actually a router. Tunnels are unnumbered
6 point-to-point lines (RFC1812 section 2.2.7); the router has a
7 single address (the 'router-id'). */
9 /* This is where we currently have the anti-spoofing paranoia - before
10 sending a packet to the kernel we check that the tunnel it came
11 over could reasonably have produced it. */
14 /* Points to note from RFC1812 (which may require changes in this
17 3.3.4 Maximum Transmission Unit - MTU
19 The MTU of each logical interface MUST be configurable within the
20 range of legal MTUs for the interface.
22 Many Link Layer protocols define a maximum frame size that may be
23 sent. In such cases, a router MUST NOT allow an MTU to be set which
24 would allow sending of frames larger than those allowed by the Link
25 Layer protocol. However, a router SHOULD be willing to receive a
26 packet as large as the maximum frame size even if that is larger than
29 4.2.1 A router SHOULD count datagrams discarded.
31 4.2.2.1 Source route options - we probably should implement processing
32 of source routes, even though mostly the security policy will prevent
35 5.3.13.4 Source Route Options
37 A router MUST implement support for source route options in forwarded
38 packets. A router MAY implement a configuration option that, when
39 enabled, causes all source-routed packets to be discarded. However,
40 such an option MUST NOT be enabled by default.
42 5.3.13.5 Record Route Option
44 Routers MUST support the Record Route option in forwarded packets.
46 A router MAY provide a configuration option that, if enabled, will
47 cause the router to ignore (i.e., pass through unchanged) Record
48 Route options in forwarded packets. If provided, such an option MUST
49 default to enabling the record-route. This option should not affect
50 the processing of Record Route options in datagrams received by the
51 router itself (in particular, Record Route options in ICMP echo
52 requests will still be processed according to Section [4.3.3.6]).
54 5.3.13.6 Timestamp Option
56 Routers MUST support the timestamp option in forwarded packets. A
57 timestamp value MUST follow the rules given [INTRO:2].
59 If the flags field = 3 (timestamp and prespecified address), the
60 router MUST add its timestamp if the next prespecified address
61 matches any of the router's IP addresses. It is not necessary that
62 the prespecified address be either the address of the interface on
63 which the packet arrived or the address of the interface over which
67 4.2.2.7 Fragmentation: RFC 791 Section 3.2
69 Fragmentation, as described in [INTERNET:1], MUST be supported by a
72 4.2.2.8 Reassembly: RFC 791 Section 3.2
74 As specified in the corresponding section of [INTRO:2], a router MUST
75 support reassembly of datagrams that it delivers to itself.
77 4.2.2.9 Time to Live: RFC 791 Section 3.2
79 Note in particular that a router MUST NOT check the TTL of a packet
80 except when forwarding it.
82 A router MUST NOT discard a datagram just because it was received
83 with TTL equal to zero or one; if it is to the router and otherwise
84 valid, the router MUST attempt to receive it.
86 On messages the router originates, the IP layer MUST provide a means
87 for the transport layer to set the TTL field of every datagram that
88 is sent. When a fixed TTL value is used, it MUST be configurable.
91 8.1 The Simple Network Management Protocol - SNMP
92 8.1.1 SNMP Protocol Elements
94 Routers MUST be manageable by SNMP [MGT:3]. The SNMP MUST operate
95 using UDP/IP as its transport and network protocols.
109 #define OPT_SOFTROUTE 1
110 #define OPT_ALLOWROUTE 2
112 #define ICMP_TYPE_ECHO_REPLY 0
114 #define ICMP_TYPE_UNREACHABLE 3
115 #define ICMP_CODE_NET_UNREACHABLE 0
116 #define ICMP_CODE_PROTOCOL_UNREACHABLE 2
117 #define ICMP_CODE_FRAGMENTATION_REQUIRED 4
118 #define ICMP_CODE_NET_PROHIBITED 13
120 #define ICMP_TYPE_ECHO_REQUEST 8
122 #define ICMP_TYPE_TIME_EXCEEDED 11
123 #define ICMP_CODE_TTL_EXCEEDED 0
125 /* Generic IP checksum routine */
126 static inline uint16_t ip_csum(uint8_t *iph,int32_t count)
128 register uint32_t sum=0;
131 sum+=ntohs(*(uint16_t *)iph);
136 sum+=*(uint8_t *)iph;
138 sum=(sum&0xffff)+(sum>>16);
144 * This is a version of ip_compute_csum() optimized for IP headers,
145 * which always checksum on 4 octet boundaries.
147 * By Jorge Cwik <jorge@laser.satlink.net>, adapted for linux by
150 static inline uint16_t ip_fast_csum(uint8_t *iph, int32_t ihl) {
153 __asm__ __volatile__(
159 "adcl 12(%1), %0 ;\n"
160 "1: adcl 16(%1), %0 ;\n"
171 /* Since the input registers which are loaded with iph and ipl
172 are modified, we must also specify them as outputs, or gcc
173 will assume they contain their original values. */
174 : "=r" (sum), "=r" (iph), "=r" (ihl)
175 : "1" (iph), "2" (ihl)
180 static inline uint16_t ip_fast_csum(uint8_t *iph, int32_t ihl)
182 assert(ihl < INT_MAX/4);
183 return ip_csum(iph,ihl*4);
188 #if defined (WORDS_BIGENDIAN)
204 /* The options start here. */
227 static void netlink_packet_deliver(struct netlink *st,
228 struct netlink_client *client,
229 struct buffer_if *buf);
231 /* XXX RFC1812 4.3.2.5:
232 All other ICMP error messages (Destination Unreachable,
233 Redirect, Time Exceeded, and Parameter Problem) SHOULD have their
234 precedence value set to 6 (INTERNETWORK CONTROL) or 7 (NETWORK
235 CONTROL). The IP Precedence value for these error messages MAY be
238 static struct icmphdr *netlink_icmp_tmpl(struct netlink *st,
239 uint32_t dest,uint16_t len)
243 BUF_ALLOC(&st->icmp,"netlink_icmp_tmpl");
244 buffer_init(&st->icmp,st->max_start_pad);
245 h=buf_append(&st->icmp,sizeof(*h));
250 h->iph.tot_len=htons(len+(h->iph.ihl*4)+8);
253 h->iph.ttl=255; /* XXX should be configurable */
255 h->iph.saddr=htonl(st->secnet_address);
256 h->iph.daddr=htonl(dest);
258 h->iph.check=ip_fast_csum((uint8_t *)&h->iph,h->iph.ihl);
265 /* Fill in the ICMP checksum field correctly */
266 static void netlink_icmp_csum(struct icmphdr *h)
270 len=ntohs(h->iph.tot_len)-(4*h->iph.ihl);
272 h->check=ip_csum(&h->type,len);
276 * An ICMP error message MUST NOT be sent as the result of
279 * * an ICMP error message, or
281 * * a datagram destined to an IP broadcast or IP multicast
284 * * a datagram sent as a link-layer broadcast, or
286 * * a non-initial fragment, or
288 * * a datagram whose source address does not define a single
289 * host -- e.g., a zero address, a loopback address, a
290 * broadcast address, a multicast address, or a Class E
293 static bool_t netlink_icmp_may_reply(struct buffer_if *buf)
296 struct icmphdr *icmph;
299 iph=(struct iphdr *)buf->start;
300 icmph=(struct icmphdr *)buf->start;
301 if (iph->protocol==1) {
302 switch(icmph->type) {
303 case 3: /* Destination unreachable */
304 case 11: /* Time Exceeded */
305 case 12: /* Parameter Problem */
309 /* How do we spot broadcast destination addresses? */
310 if (ntohs(iph->frag_off)&0x1fff) return False; /* Non-initial fragment */
311 source=ntohl(iph->saddr);
312 if (source==0) return False;
313 if ((source&0xff000000)==0x7f000000) return False;
314 /* How do we spot broadcast source addresses? */
315 if ((source&0xf0000000)==0xe0000000) return False; /* Multicast */
316 if ((source&0xf0000000)==0xf0000000) return False; /* Class E */
320 /* How much of the original IP packet do we include in its ICMP
321 response? The header plus up to 64 bits. */
324 4.3.2.3 Original Message Header
326 Historically, every ICMP error message has included the Internet
327 header and at least the first 8 data bytes of the datagram that
328 triggered the error. This is no longer adequate, due to the use of
329 IP-in-IP tunneling and other technologies. Therefore, the ICMP
330 datagram SHOULD contain as much of the original datagram as possible
331 without the length of the ICMP datagram exceeding 576 bytes. The
332 returned IP header (and user data) MUST be identical to that which
333 was received, except that the router is not required to undo any
334 modifications to the IP header that are normally performed in
335 forwarding that were performed before the error was detected (e.g.,
336 decrementing the TTL, or updating options). Note that the
337 requirements of Section [4.3.3.5] supersede this requirement in some
338 cases (i.e., for a Parameter Problem message, if the problem is in a
339 modified field, the router must undo the modification). See Section
342 static uint16_t netlink_icmp_reply_len(struct buffer_if *buf)
344 struct iphdr *iph=(struct iphdr *)buf->start;
348 /* We include the first 8 bytes of the packet data, provided they exist */
350 plen=ntohs(iph->tot_len);
351 return (hlen>plen?plen:hlen);
354 /* client indicates where the packet we're constructing a response to
355 comes from. NULL indicates the host. */
356 static void netlink_icmp_simple(struct netlink *st, struct buffer_if *buf,
357 struct netlink_client *client,
358 uint8_t type, uint8_t code)
360 struct iphdr *iph=(struct iphdr *)buf->start;
364 if (netlink_icmp_may_reply(buf)) {
365 len=netlink_icmp_reply_len(buf);
366 h=netlink_icmp_tmpl(st,ntohl(iph->saddr),len);
367 h->type=type; h->code=code;
368 memcpy(buf_append(&st->icmp,len),buf->start,len);
369 netlink_icmp_csum(h);
370 netlink_packet_deliver(st,NULL,&st->icmp);
371 BUF_ASSERT_FREE(&st->icmp);
376 * RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the
378 * RFC1812: 4.2.2.5 MUST discard messages containing invalid checksums.
380 * Is the datagram acceptable?
382 * 1. Length at least the size of an ip header
384 * 3. Checksums correctly.
385 * 4. Doesn't have a bogus length
387 static bool_t netlink_check(struct netlink *st, struct buffer_if *buf)
389 struct iphdr *iph=(struct iphdr *)buf->start;
392 if (iph->ihl < 5 || iph->version != 4) return False;
393 if (buf->size < iph->ihl*4) return False;
394 if (ip_fast_csum((uint8_t *)iph, iph->ihl)!=0) return False;
395 len=ntohs(iph->tot_len);
396 /* There should be no padding */
397 if (buf->size!=len || len<(iph->ihl<<2)) return False;
398 /* XXX check that there's no source route specified */
402 /* Deliver a packet. "client" is the _origin_ of the packet, not its
403 destination, and is NULL for packets from the host and packets
404 generated internally in secnet. */
405 static void netlink_packet_deliver(struct netlink *st,
406 struct netlink_client *client,
407 struct buffer_if *buf)
409 struct iphdr *iph=(struct iphdr *)buf->start;
410 uint32_t dest=ntohl(iph->daddr);
411 uint32_t source=ntohl(iph->saddr);
412 uint32_t best_quality;
413 bool_t allow_route=False;
414 bool_t found_allowed=False;
418 BUF_ASSERT_USED(buf);
420 if (dest==st->secnet_address) {
421 Message(M_ERR,"%s: trying to deliver a packet to myself!\n",st->name);
426 /* Packets from the host (client==NULL) may always be routed. Packets
427 from clients with the allow_route option will also be routed. */
428 if (!client || (client && (client->options & OPT_ALLOWROUTE)))
431 /* If !allow_route, we check the routing table anyway, and if
432 there's a suitable route with OPT_ALLOWROUTE set we use it. If
433 there's a suitable route, but none with OPT_ALLOWROUTE set then
434 we generate ICMP 'communication with destination network
435 administratively prohibited'. */
439 for (i=0; i<st->n_clients; i++) {
440 if (st->routes[i]->up &&
441 ipset_contains_addr(st->routes[i]->networks,dest)) {
442 /* It's an available route to the correct destination. But is
443 it better than the one we already have? */
445 /* If we have already found an allowed route then we don't
446 bother looking at routes we're not allowed to use. If
447 we don't yet have an allowed route we'll consider any. */
448 if (!allow_route && found_allowed) {
449 if (!(st->routes[i]->options&OPT_ALLOWROUTE)) continue;
452 if (st->routes[i]->link_quality>best_quality
453 || best_quality==0) {
454 best_quality=st->routes[i]->link_quality;
456 if (st->routes[i]->options&OPT_ALLOWROUTE)
458 /* If quality isn't perfect we may wish to
459 consider kicking the tunnel with a 0-length
460 packet to prompt it to perform a key setup.
461 Then it'll eventually decide it's up or
463 /* If quality is perfect and we're allowed to use the
464 route we don't need to search any more. */
465 if (best_quality>=MAXIMUM_LINK_QUALITY &&
466 (allow_route || found_allowed)) break;
470 if (best_match==-1) {
471 /* The packet's not going down a tunnel. It might (ought to)
473 if (ipset_contains_addr(st->networks,dest)) {
474 st->deliver_to_host(st->dst,buf);
476 BUF_ASSERT_FREE(buf);
479 s=ipaddr_to_string(source);
480 d=ipaddr_to_string(dest);
481 Message(M_DEBUG,"%s: don't know where to deliver packet "
482 "(s=%s, d=%s)\n", st->name, s, d);
484 netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE,
485 ICMP_CODE_NET_UNREACHABLE);
490 !(st->routes[best_match]->options&OPT_ALLOWROUTE)) {
492 s=ipaddr_to_string(source);
493 d=ipaddr_to_string(dest);
494 /* We have a usable route but aren't allowed to use it.
495 Generate ICMP destination unreachable: communication
496 with destination network administratively prohibited */
497 Message(M_NOTICE,"%s: denied forwarding for packet (s=%s, d=%s)\n",
501 netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE,
502 ICMP_CODE_NET_PROHIBITED);
505 if (best_quality>0) {
506 /* XXX Fragment if required */
507 st->routes[best_match]->deliver(
508 st->routes[best_match]->dst, buf);
509 st->routes[best_match]->outcount++;
510 BUF_ASSERT_FREE(buf);
512 /* Generate ICMP destination unreachable */
513 netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE,
514 ICMP_CODE_NET_UNREACHABLE); /* client==NULL */
519 BUF_ASSERT_FREE(buf);
522 static void netlink_packet_forward(struct netlink *st,
523 struct netlink_client *client,
524 struct buffer_if *buf)
526 struct iphdr *iph=(struct iphdr *)buf->start;
528 BUF_ASSERT_USED(buf);
530 /* Packet has already been checked */
532 /* Generate ICMP time exceeded */
533 netlink_icmp_simple(st,buf,client,ICMP_TYPE_TIME_EXCEEDED,
534 ICMP_CODE_TTL_EXCEEDED);
540 iph->check=ip_fast_csum((uint8_t *)iph,iph->ihl);
542 netlink_packet_deliver(st,client,buf);
543 BUF_ASSERT_FREE(buf);
546 /* Deal with packets addressed explicitly to us */
547 static void netlink_packet_local(struct netlink *st,
548 struct netlink_client *client,
549 struct buffer_if *buf)
555 h=(struct icmphdr *)buf->start;
557 if ((ntohs(h->iph.frag_off)&0xbfff)!=0) {
558 Message(M_WARNING,"%s: fragmented packet addressed to secnet; "
559 "ignoring it\n",st->name);
564 if (h->iph.protocol==1) {
566 if (h->type==ICMP_TYPE_ECHO_REQUEST && h->code==0) {
567 /* ICMP echo-request. Special case: we re-use the buffer
568 to construct the reply. */
569 h->type=ICMP_TYPE_ECHO_REPLY;
570 h->iph.daddr=h->iph.saddr;
571 h->iph.saddr=htonl(st->secnet_address);
574 h->iph.check=ip_fast_csum((uint8_t *)h,h->iph.ihl);
575 netlink_icmp_csum(h);
576 netlink_packet_deliver(st,NULL,buf);
579 Message(M_WARNING,"%s: unknown incoming ICMP\n",st->name);
581 /* Send ICMP protocol unreachable */
582 netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE,
583 ICMP_CODE_PROTOCOL_UNREACHABLE);
591 /* If cid==NULL packet is from host, otherwise cid specifies which tunnel
593 static void netlink_incoming(struct netlink *st, struct netlink_client *client,
594 struct buffer_if *buf)
596 uint32_t source,dest;
599 BUF_ASSERT_USED(buf);
600 if (!netlink_check(st,buf)) {
601 Message(M_WARNING,"%s: bad IP packet from %s\n",
602 st->name,client?client->name:"host");
606 iph=(struct iphdr *)buf->start;
608 source=ntohl(iph->saddr);
609 dest=ntohl(iph->daddr);
611 /* Check source. If we don't like the source, there's no point
612 generating ICMP because we won't know how to get it to the
613 source of the packet. */
615 /* Check that the packet source is appropriate for the tunnel
617 if (!ipset_contains_addr(client->networks,source)) {
619 s=ipaddr_to_string(source);
620 d=ipaddr_to_string(dest);
621 Message(M_WARNING,"%s: packet from tunnel %s with bad "
622 "source address (s=%s,d=%s)\n",st->name,client->name,s,d);
628 /* Check that the packet originates in our configured local
629 network, and hasn't been forwarded from elsewhere or
630 generated with the wrong source address */
631 if (!ipset_contains_addr(st->networks,source)) {
633 s=ipaddr_to_string(source);
634 d=ipaddr_to_string(dest);
635 Message(M_WARNING,"%s: outgoing packet with bad source address "
636 "(s=%s,d=%s)\n",st->name,s,d);
643 /* If this is a point-to-point device we don't examine the
644 destination address at all; we blindly send it down our
645 one-and-only registered tunnel, or to the host, depending on
646 where it came from. It's up to external software to check
647 address validity and generate ICMP, etc. */
650 st->deliver_to_host(st->dst,buf);
652 st->clients->deliver(st->clients->dst,buf);
654 BUF_ASSERT_FREE(buf);
658 /* st->secnet_address needs checking before matching destination
660 if (dest==st->secnet_address) {
661 netlink_packet_local(st,client,buf);
662 BUF_ASSERT_FREE(buf);
665 netlink_packet_forward(st,client,buf);
666 BUF_ASSERT_FREE(buf);
669 static void netlink_inst_incoming(void *sst, struct buffer_if *buf)
671 struct netlink_client *c=sst;
672 struct netlink *st=c->nst;
674 netlink_incoming(st,c,buf);
677 static void netlink_dev_incoming(void *sst, struct buffer_if *buf)
679 struct netlink *st=sst;
681 netlink_incoming(st,NULL,buf);
684 static void netlink_set_quality(void *sst, uint32_t quality)
686 struct netlink_client *c=sst;
687 struct netlink *st=c->nst;
689 c->link_quality=quality;
690 c->up=(c->link_quality==LINK_QUALITY_DOWN)?False:True;
691 if (c->options&OPT_SOFTROUTE) {
692 st->set_routes(st->dst,c);
696 static void netlink_output_subnets(struct netlink *st, uint32_t loglevel,
697 struct subnet_list *snets)
702 for (i=0; i<snets->entries; i++) {
703 net=subnet_to_string(snets->list[i]);
704 Message(loglevel,"%s ",net);
709 static void netlink_dump_routes(struct netlink *st, bool_t requested)
715 if (requested) c=M_WARNING;
717 net=ipaddr_to_string(st->secnet_address);
718 Message(c,"%s: point-to-point (remote end is %s); routes:\n",
721 netlink_output_subnets(st,c,st->clients->subnets);
724 Message(c,"%s: routing table:\n",st->name);
725 for (i=0; i<st->n_clients; i++) {
726 netlink_output_subnets(st,c,st->routes[i]->subnets);
727 Message(c,"-> tunnel %s (%s,mtu %d,%s routes,%s,"
728 "quality %d,use %d,pri %lu)\n",
730 st->routes[i]->up?"up":"down",
732 st->routes[i]->options&OPT_SOFTROUTE?"soft":"hard",
733 st->routes[i]->options&OPT_ALLOWROUTE?"free":"restricted",
734 st->routes[i]->link_quality,
735 st->routes[i]->outcount,
736 (unsigned long)st->routes[i]->priority);
738 net=ipaddr_to_string(st->secnet_address);
739 Message(c,"%s/32 -> netlink \"%s\" (use %d)\n",
740 net,st->name,st->localcount);
742 for (i=0; i<st->subnets->entries; i++) {
743 net=subnet_to_string(st->subnets->list[i]);
744 Message(c,"%s ",net);
748 Message(c,"-> host (use %d)\n",st->outcount);
752 /* ap is a pointer to a member of the routes array */
753 static int netlink_compare_client_priority(const void *ap, const void *bp)
755 const struct netlink_client *const*a=ap;
756 const struct netlink_client *const*b=bp;
758 if ((*a)->priority==(*b)->priority) return 0;
759 if ((*a)->priority<(*b)->priority) return 1;
763 static void netlink_phase_hook(void *sst, uint32_t new_phase)
765 struct netlink *st=sst;
766 struct netlink_client *c;
769 /* All the networks serviced by the various tunnels should now
770 * have been registered. We build a routing table by sorting the
771 * clients by priority. */
772 st->routes=safe_malloc_ary(sizeof(*st->routes),st->n_clients,
773 "netlink_phase_hook");
776 for (c=st->clients; c; c=c->next) {
780 /* Sort the table in descending order of priority */
781 qsort(st->routes,st->n_clients,sizeof(*st->routes),
782 netlink_compare_client_priority);
784 netlink_dump_routes(st,False);
787 static void netlink_signal_handler(void *sst, int signum)
789 struct netlink *st=sst;
790 Message(M_INFO,"%s: route dump requested by SIGUSR1\n",st->name);
791 netlink_dump_routes(st,True);
794 static void netlink_inst_output_config(void *sst, struct buffer_if *buf)
796 /* struct netlink_client *c=sst; */
797 /* struct netlink *st=c->nst; */
799 /* For now we don't output anything */
800 BUF_ASSERT_USED(buf);
803 static bool_t netlink_inst_check_config(void *sst, struct buffer_if *buf)
805 /* struct netlink_client *c=sst; */
806 /* struct netlink *st=c->nst; */
808 BUF_ASSERT_USED(buf);
809 /* We need to eat all of the configuration information from the buffer
810 for backward compatibility. */
815 static void netlink_inst_set_mtu(void *sst, int32_t new_mtu)
817 struct netlink_client *c=sst;
822 static void netlink_inst_reg(void *sst, netlink_deliver_fn *deliver,
823 void *dst, int32_t max_start_pad,
826 struct netlink_client *c=sst;
827 struct netlink *st=c->nst;
829 if (max_start_pad > st->max_start_pad) st->max_start_pad=max_start_pad;
830 if (max_end_pad > st->max_end_pad) st->max_end_pad=max_end_pad;
835 static struct flagstr netlink_option_table[]={
836 { "soft", OPT_SOFTROUTE },
837 { "allow-route", OPT_ALLOWROUTE },
840 /* This is the routine that gets called when the closure that's
841 returned by an invocation of a netlink device closure (eg. tun,
842 userv-ipif) is invoked. It's used to create routes and pass in
843 information about them; the closure it returns is used by site
845 static closure_t *netlink_inst_create(struct netlink *st,
846 struct cloc loc, dict_t *dict)
848 struct netlink_client *c;
850 struct ipset *networks;
851 uint32_t options,priority;
855 name=dict_read_string(dict, "name", True, st->name, loc);
857 l=dict_lookup(dict,"routes");
859 cfgfatal(loc,st->name,"required parameter \"routes\" not found\n");
860 networks=string_list_to_ipset(l,loc,st->name,"routes");
861 options=string_list_to_word(dict_lookup(dict,"options"),
862 netlink_option_table,st->name);
864 priority=dict_read_number(dict,"priority",False,st->name,loc,0);
865 mtu=dict_read_number(dict,"mtu",False,st->name,loc,0);
867 if ((options&OPT_SOFTROUTE) && !st->set_routes) {
868 cfgfatal(loc,st->name,"this netlink device does not support "
873 if (options&OPT_SOFTROUTE) {
874 /* XXX for now we assume that soft routes require root privilege;
875 this may not always be true. The device driver can tell us. */
876 require_root_privileges=True;
877 require_root_privileges_explanation="netlink: soft routes";
879 cfgfatal(loc,st->name,"point-to-point netlinks do not support "
885 /* Check that nets are a subset of st->remote_networks;
886 refuse to register if they are not. */
887 if (!ipset_is_subset(st->remote_networks,networks)) {
888 cfgfatal(loc,st->name,"routes are not allowed\n");
892 c=safe_malloc(sizeof(*c),"netlink_inst_create");
893 c->cl.description=name;
894 c->cl.type=CL_NETLINK;
896 c->cl.interface=&c->ops;
898 c->ops.reg=netlink_inst_reg;
899 c->ops.deliver=netlink_inst_incoming;
900 c->ops.set_quality=netlink_set_quality;
901 c->ops.output_config=netlink_inst_output_config;
902 c->ops.check_config=netlink_inst_check_config;
903 c->ops.set_mtu=netlink_inst_set_mtu;
906 c->networks=networks;
907 c->subnets=ipset_to_subnet_list(networks);
908 c->priority=priority;
912 c->link_quality=LINK_QUALITY_UNUSED;
913 c->mtu=mtu?mtu:st->mtu;
920 assert(st->n_clients < INT_MAX);
926 static list_t *netlink_inst_apply(closure_t *self, struct cloc loc,
927 dict_t *context, list_t *args)
929 struct netlink *st=self->interface;
935 item=list_elem(args,0);
936 if (!item || item->type!=t_dict) {
937 cfgfatal(loc,st->name,"must have a dictionary argument\n");
939 dict=item->data.dict;
941 cl=netlink_inst_create(st,loc,dict);
943 return new_closure(cl);
946 netlink_deliver_fn *netlink_init(struct netlink *st,
947 void *dst, struct cloc loc,
948 dict_t *dict, cstring_t description,
949 netlink_route_fn *set_routes,
950 netlink_deliver_fn *to_host)
956 st->cl.description=description;
958 st->cl.apply=netlink_inst_apply;
965 st->set_routes=set_routes;
966 st->deliver_to_host=to_host;
968 st->name=dict_read_string(dict,"name",False,description,loc);
969 if (!st->name) st->name=description;
970 l=dict_lookup(dict,"networks");
972 st->networks=string_list_to_ipset(l,loc,st->name,"networks");
976 st->networks=ipset_complement(empty);
979 l=dict_lookup(dict,"remote-networks");
981 st->remote_networks=string_list_to_ipset(l,loc,st->name,
986 st->remote_networks=ipset_complement(empty);
990 sa=dict_find_item(dict,"secnet-address",False,"netlink",loc);
991 ptpa=dict_find_item(dict,"ptp-address",False,"netlink",loc);
993 cfgfatal(loc,st->name,"you may not specify secnet-address and "
994 "ptp-address in the same netlink device\n");
997 cfgfatal(loc,st->name,"you must specify secnet-address or "
998 "ptp-address for this netlink device\n");
1001 st->secnet_address=string_item_to_ipaddr(sa,"netlink");
1004 st->secnet_address=string_item_to_ipaddr(ptpa,"netlink");
1007 /* To be strictly correct we could subtract secnet_address from
1008 networks here. It shouldn't make any practical difference,
1009 though, and will make the route dump look complicated... */
1010 st->subnets=ipset_to_subnet_list(st->networks);
1011 st->mtu=dict_read_number(dict, "mtu", False, "netlink", loc, DEFAULT_MTU);
1012 buffer_new(&st->icmp,ICMP_BUFSIZE);
1016 add_hook(PHASE_SETUP,netlink_phase_hook,st);
1017 request_signal_notification(SIGUSR1, netlink_signal_handler, st);
1019 /* If we're point-to-point then we return a CL_NETLINK directly,
1020 rather than a CL_NETLINK_OLD or pure closure (depending on
1021 compatibility). This CL_NETLINK is for our one and only
1022 client. Our cl.apply function is NULL. */
1025 cl=netlink_inst_create(st,loc,dict);
1028 return netlink_dev_incoming;
1031 /* No connection to the kernel at all... */
1037 static bool_t null_set_route(void *sst, struct netlink_client *routes)
1039 struct null *st=sst;
1041 if (routes->up!=routes->kup) {
1042 Message(M_INFO,"%s: setting routes for tunnel %s to state %s\n",
1043 st->nl.name,routes->name,
1044 routes->up?"up":"down");
1045 routes->kup=routes->up;
1051 static void null_deliver(void *sst, struct buffer_if *buf)
1056 static list_t *null_apply(closure_t *self, struct cloc loc, dict_t *context,
1063 st=safe_malloc(sizeof(*st),"null_apply");
1065 item=list_elem(args,0);
1066 if (!item || item->type!=t_dict)
1067 cfgfatal(loc,"null-netlink","parameter must be a dictionary\n");
1069 dict=item->data.dict;
1071 netlink_init(&st->nl,st,loc,dict,"null-netlink",null_set_route,
1074 return new_closure(&st->nl.cl);
1077 void netlink_module(dict_t *dict)
1079 add_closure(dict,"null-netlink",null_apply);