Commit | Line | Data |
---|---|---|
2fe58dfd SE |
1 | /* User-kernel network link */ |
2 | ||
baa06aeb SE |
3 | /* Each netlink device is actually a router, with its own IP address. |
4 | We do things like decreasing the TTL and recalculating the header | |
5 | checksum, generating ICMP, responding to pings, etc. */ | |
2fe58dfd SE |
6 | |
7 | /* This is where we have the anti-spoofing paranoia - before sending a | |
8 | packet to the kernel we check that the tunnel it came over could | |
9 | reasonably have produced it. */ | |
10 | ||
8689b3a9 | 11 | #include "secnet.h" |
2fe58dfd | 12 | #include "util.h" |
9d3a4132 | 13 | #include "netlink.h" |
2fe58dfd | 14 | |
4efd681a SE |
15 | /* Generic IP checksum routine */ |
16 | static inline uint16_t ip_csum(uint8_t *iph,uint32_t count) | |
2fe58dfd | 17 | { |
4efd681a SE |
18 | register uint32_t sum=0; |
19 | ||
20 | while (count>1) { | |
21 | sum+=ntohs(*(uint16_t *)iph); | |
22 | iph+=2; | |
23 | count-=2; | |
24 | } | |
25 | if(count>0) | |
26 | sum+=*(uint8_t *)iph; | |
27 | while (sum>>16) | |
28 | sum=(sum&0xffff)+(sum>>16); | |
29 | return htons(~sum); | |
2fe58dfd SE |
30 | } |
31 | ||
4efd681a SE |
32 | #ifdef i386 |
33 | /* | |
34 | * This is a version of ip_compute_csum() optimized for IP headers, | |
35 | * which always checksum on 4 octet boundaries. | |
36 | * | |
37 | * By Jorge Cwik <jorge@laser.satlink.net>, adapted for linux by | |
38 | * Arnt Gulbrandsen. | |
39 | */ | |
40 | static inline uint16_t ip_fast_csum(uint8_t *iph, uint32_t ihl) { | |
41 | uint32_t sum; | |
42 | ||
43 | __asm__ __volatile__(" | |
44 | movl (%1), %0 | |
45 | subl $4, %2 | |
46 | jbe 2f | |
47 | addl 4(%1), %0 | |
48 | adcl 8(%1), %0 | |
49 | adcl 12(%1), %0 | |
50 | 1: adcl 16(%1), %0 | |
51 | lea 4(%1), %1 | |
52 | decl %2 | |
53 | jne 1b | |
54 | adcl $0, %0 | |
55 | movl %0, %2 | |
56 | shrl $16, %0 | |
57 | addw %w2, %w0 | |
58 | adcl $0, %0 | |
59 | notl %0 | |
60 | 2: | |
61 | " | |
62 | /* Since the input registers which are loaded with iph and ipl | |
63 | are modified, we must also specify them as outputs, or gcc | |
64 | will assume they contain their original values. */ | |
65 | : "=r" (sum), "=r" (iph), "=r" (ihl) | |
66 | : "1" (iph), "2" (ihl)); | |
67 | return sum; | |
68 | } | |
69 | #else | |
70 | static inline uint16_t ip_fast_csum(uint8_t *iph, uint32_t ihl) | |
2fe58dfd | 71 | { |
4efd681a SE |
72 | return ip_csum(iph,ihl*4); |
73 | } | |
74 | #endif | |
75 | ||
76 | struct iphdr { | |
77 | #if defined (WORDS_BIGENDIAN) | |
78 | uint8_t version:4, | |
79 | ihl:4; | |
80 | #else | |
81 | uint8_t ihl:4, | |
82 | version:4; | |
83 | #endif | |
84 | uint8_t tos; | |
85 | uint16_t tot_len; | |
86 | uint16_t id; | |
87 | uint16_t frag_off; | |
88 | uint8_t ttl; | |
89 | uint8_t protocol; | |
90 | uint16_t check; | |
91 | uint32_t saddr; | |
92 | uint32_t daddr; | |
93 | /* The options start here. */ | |
94 | }; | |
95 | ||
96 | struct icmphdr { | |
97 | struct iphdr iph; | |
98 | uint8_t type; | |
99 | uint8_t code; | |
100 | uint16_t check; | |
101 | union { | |
102 | uint32_t unused; | |
103 | struct { | |
104 | uint8_t pointer; | |
105 | uint8_t unused1; | |
106 | uint16_t unused2; | |
107 | } pprob; | |
108 | uint32_t gwaddr; | |
109 | struct { | |
110 | uint16_t id; | |
111 | uint16_t seq; | |
112 | } echo; | |
113 | } d; | |
114 | }; | |
115 | ||
70dc107b SE |
116 | static void netlink_packet_deliver(struct netlink *st, |
117 | struct netlink_client *client, | |
118 | struct buffer_if *buf); | |
4efd681a SE |
119 | |
120 | static struct icmphdr *netlink_icmp_tmpl(struct netlink *st, | |
121 | uint32_t dest,uint16_t len) | |
122 | { | |
123 | struct icmphdr *h; | |
124 | ||
125 | BUF_ALLOC(&st->icmp,"netlink_icmp_tmpl"); | |
126 | buffer_init(&st->icmp,st->max_start_pad); | |
127 | h=buf_append(&st->icmp,sizeof(*h)); | |
128 | ||
129 | h->iph.version=4; | |
130 | h->iph.ihl=5; | |
131 | h->iph.tos=0; | |
132 | h->iph.tot_len=htons(len+(h->iph.ihl*4)+8); | |
133 | h->iph.id=0; | |
134 | h->iph.frag_off=0; | |
135 | h->iph.ttl=255; | |
136 | h->iph.protocol=1; | |
137 | h->iph.saddr=htonl(st->secnet_address); | |
138 | h->iph.daddr=htonl(dest); | |
139 | h->iph.check=0; | |
140 | h->iph.check=ip_fast_csum((uint8_t *)&h->iph,h->iph.ihl); | |
141 | h->check=0; | |
142 | h->d.unused=0; | |
143 | ||
144 | return h; | |
145 | } | |
146 | ||
147 | /* Fill in the ICMP checksum field correctly */ | |
148 | static void netlink_icmp_csum(struct icmphdr *h) | |
149 | { | |
150 | uint32_t len; | |
151 | ||
152 | len=ntohs(h->iph.tot_len)-(4*h->iph.ihl); | |
153 | h->check=0; | |
154 | h->check=ip_csum(&h->type,len); | |
155 | } | |
156 | ||
157 | /* RFC1122: | |
158 | * An ICMP error message MUST NOT be sent as the result of | |
159 | * receiving: | |
160 | * | |
161 | * * an ICMP error message, or | |
162 | * | |
163 | * * a datagram destined to an IP broadcast or IP multicast | |
164 | * address, or | |
165 | * | |
166 | * * a datagram sent as a link-layer broadcast, or | |
167 | * | |
168 | * * a non-initial fragment, or | |
169 | * | |
170 | * * a datagram whose source address does not define a single | |
171 | * host -- e.g., a zero address, a loopback address, a | |
172 | * broadcast address, a multicast address, or a Class E | |
173 | * address. | |
174 | */ | |
175 | static bool_t netlink_icmp_may_reply(struct buffer_if *buf) | |
176 | { | |
177 | struct iphdr *iph; | |
178 | uint32_t source; | |
179 | ||
180 | iph=(struct iphdr *)buf->start; | |
181 | if (iph->protocol==1) return False; /* Overly-broad; we may reply to | |
182 | eg. icmp echo-request */ | |
183 | /* How do we spot broadcast destination addresses? */ | |
184 | if (ntohs(iph->frag_off)&0x1fff) return False; /* Non-initial fragment */ | |
185 | source=ntohl(iph->saddr); | |
186 | if (source==0) return False; | |
187 | if ((source&0xff000000)==0x7f000000) return False; | |
188 | /* How do we spot broadcast source addresses? */ | |
189 | if ((source&0xf0000000)==0xe0000000) return False; /* Multicast */ | |
190 | if ((source&0xf0000000)==0xf0000000) return False; /* Class E */ | |
191 | return True; | |
192 | } | |
193 | ||
194 | /* How much of the original IP packet do we include in its ICMP | |
195 | response? The header plus up to 64 bits. */ | |
196 | static uint16_t netlink_icmp_reply_len(struct buffer_if *buf) | |
197 | { | |
198 | struct iphdr *iph=(struct iphdr *)buf->start; | |
199 | uint16_t hlen,plen; | |
200 | ||
201 | hlen=iph->ihl*4; | |
202 | /* We include the first 8 bytes of the packet data, provided they exist */ | |
203 | hlen+=8; | |
204 | plen=ntohs(iph->tot_len); | |
205 | return (hlen>plen?plen:hlen); | |
206 | } | |
207 | ||
70dc107b SE |
208 | /* client indicates where the packet we're constructing a response to |
209 | comes from. NULL indicates the host. */ | |
4efd681a | 210 | static void netlink_icmp_simple(struct netlink *st, struct buffer_if *buf, |
70dc107b | 211 | struct netlink_client *client, |
4efd681a SE |
212 | uint8_t type, uint8_t code) |
213 | { | |
214 | struct iphdr *iph=(struct iphdr *)buf->start; | |
215 | struct icmphdr *h; | |
216 | uint16_t len; | |
217 | ||
218 | if (netlink_icmp_may_reply(buf)) { | |
219 | len=netlink_icmp_reply_len(buf); | |
220 | h=netlink_icmp_tmpl(st,ntohl(iph->saddr),len); | |
221 | h->type=type; h->code=code; | |
222 | memcpy(buf_append(&st->icmp,len),buf->start,len); | |
223 | netlink_icmp_csum(h); | |
70dc107b | 224 | netlink_packet_deliver(st,NULL,&st->icmp); |
4efd681a SE |
225 | BUF_ASSERT_FREE(&st->icmp); |
226 | } | |
227 | } | |
228 | ||
229 | /* | |
230 | * RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the | |
231 | * checksum. | |
232 | * | |
233 | * Is the datagram acceptable? | |
234 | * | |
235 | * 1. Length at least the size of an ip header | |
236 | * 2. Version of 4 | |
237 | * 3. Checksums correctly. | |
238 | * 4. Doesn't have a bogus length | |
239 | */ | |
240 | static bool_t netlink_check(struct netlink *st, struct buffer_if *buf) | |
241 | { | |
242 | struct iphdr *iph=(struct iphdr *)buf->start; | |
243 | uint32_t len; | |
244 | ||
9d3a4132 SE |
245 | if (iph->ihl < 5 || iph->version != 4) return False; |
246 | if (buf->size < iph->ihl*4) return False; | |
247 | if (ip_fast_csum((uint8_t *)iph, iph->ihl)!=0) return False; | |
4efd681a SE |
248 | len=ntohs(iph->tot_len); |
249 | /* There should be no padding */ | |
9d3a4132 | 250 | if (buf->size!=len || len<(iph->ihl<<2)) return False; |
4efd681a SE |
251 | /* XXX check that there's no source route specified */ |
252 | return True; | |
253 | } | |
254 | ||
70dc107b SE |
255 | /* Deliver a packet. "client" points to the _origin_ of the packet, not |
256 | its destination. (May be used when sending ICMP response - avoid | |
257 | asymmetric routing.) */ | |
258 | static void netlink_packet_deliver(struct netlink *st, | |
259 | struct netlink_client *client, | |
260 | struct buffer_if *buf) | |
4efd681a SE |
261 | { |
262 | struct iphdr *iph=(struct iphdr *)buf->start; | |
263 | uint32_t dest=ntohl(iph->daddr); | |
70dc107b SE |
264 | uint32_t source=ntohl(iph->saddr); |
265 | uint32_t best_quality; | |
266 | int best_match; | |
267 | int i; | |
2fe58dfd | 268 | |
4efd681a | 269 | BUF_ASSERT_USED(buf); |
2fe58dfd | 270 | |
4efd681a SE |
271 | if (dest==st->secnet_address) { |
272 | Message(M_ERROR,"%s: trying to deliver a packet to myself!\n"); | |
273 | BUF_FREE(buf); | |
2fe58dfd SE |
274 | return; |
275 | } | |
4efd681a | 276 | |
3454dce4 SE |
277 | /* XXX we're going to need an extra value 'allow_route' for the |
278 | source of the packet. It's always True for packets from the | |
279 | host. For packets from tunnels, we consult the client | |
280 | options. If !allow_route and the destination is a tunnel that | |
281 | also doesn't allow routing, we must reject the packet with an | |
282 | 'administratively prohibited' or something similar ICMP. */ | |
70dc107b SE |
283 | if (!client) { |
284 | /* Origin of packet is host or secnet. Might be for a tunnel. */ | |
285 | best_quality=0; | |
286 | best_match=-1; | |
287 | for (i=0; i<st->n_routes; i++) { | |
9d3a4132 | 288 | if (st->routes[i].up && subnet_match(&st->routes[i].net,dest)) { |
70dc107b SE |
289 | if (st->routes[i].c->link_quality>best_quality |
290 | || best_quality==0) { | |
291 | best_quality=st->routes[i].c->link_quality; | |
292 | best_match=i; | |
293 | /* If quality isn't perfect we may wish to | |
294 | consider kicking the tunnel with a 0-length | |
295 | packet to prompt it to perform a key setup. | |
296 | Then it'll eventually decide it's up or | |
297 | down. */ | |
298 | /* If quality is perfect we don't need to search | |
299 | any more. */ | |
300 | if (best_quality>=MAXIMUM_LINK_QUALITY) break; | |
301 | } | |
302 | } | |
303 | } | |
304 | if (best_match==-1) { | |
305 | /* Not going down a tunnel. Might be for the host. | |
306 | XXX think about this - only situation should be if we're | |
307 | sending ICMP. */ | |
308 | if (source!=st->secnet_address) { | |
309 | Message(M_ERROR,"netlink_packet_deliver: outgoing packet " | |
310 | "from host that won't fit down any of our tunnels!\n"); | |
9d3a4132 SE |
311 | /* XXX I think this could also occur if a soft tunnel just |
312 | went down, but still had packets queued in the kernel. */ | |
70dc107b SE |
313 | BUF_FREE(buf); |
314 | } else { | |
315 | st->deliver_to_host(st->dst,NULL,buf); | |
316 | BUF_ASSERT_FREE(buf); | |
317 | } | |
318 | } else { | |
319 | if (best_quality>0) { | |
320 | st->routes[best_match].c->deliver( | |
321 | st->routes[best_match].c->dst, | |
322 | st->routes[best_match].c, buf); | |
4efd681a SE |
323 | BUF_ASSERT_FREE(buf); |
324 | } else { | |
325 | /* Generate ICMP destination unreachable */ | |
70dc107b | 326 | netlink_icmp_simple(st,buf,client,3,0); /* client==NULL */ |
4efd681a SE |
327 | BUF_FREE(buf); |
328 | } | |
70dc107b SE |
329 | } |
330 | } else { /* client is set */ | |
331 | /* We know the origin is a tunnel - packet must be for the host */ | |
3454dce4 SE |
332 | /* XXX THIS IS NOT NECESSARILY TRUE, AND NEEDS FIXING */ |
333 | /* THIS FUNCTION MUST JUST DELIVER THE PACKET: IT MUST ASSUME | |
334 | THE PACKET HAS ALREADY BEEN CHECKED */ | |
70dc107b SE |
335 | if (subnet_matches_list(&st->networks,dest)) { |
336 | st->deliver_to_host(st->dst,NULL,buf); | |
337 | BUF_ASSERT_FREE(buf); | |
338 | } else { | |
339 | Message(M_ERROR,"%s: packet from tunnel %s can't be delivered " | |
340 | "to the host\n",st->name,client->name); | |
341 | netlink_icmp_simple(st,buf,client,3,0); | |
342 | BUF_FREE(buf); | |
2fe58dfd SE |
343 | } |
344 | } | |
70dc107b | 345 | BUF_ASSERT_FREE(buf); |
4efd681a SE |
346 | } |
347 | ||
70dc107b SE |
348 | static void netlink_packet_forward(struct netlink *st, |
349 | struct netlink_client *client, | |
350 | struct buffer_if *buf) | |
4efd681a SE |
351 | { |
352 | struct iphdr *iph=(struct iphdr *)buf->start; | |
353 | ||
354 | BUF_ASSERT_USED(buf); | |
355 | ||
356 | /* Packet has already been checked */ | |
357 | if (iph->ttl<=1) { | |
358 | /* Generate ICMP time exceeded */ | |
70dc107b | 359 | netlink_icmp_simple(st,buf,client,11,0); |
4efd681a SE |
360 | BUF_FREE(buf); |
361 | return; | |
362 | } | |
363 | iph->ttl--; | |
364 | iph->check=0; | |
365 | iph->check=ip_fast_csum((uint8_t *)iph,iph->ihl); | |
366 | ||
70dc107b | 367 | netlink_packet_deliver(st,client,buf); |
4efd681a SE |
368 | BUF_ASSERT_FREE(buf); |
369 | } | |
370 | ||
9d3a4132 | 371 | /* Deal with packets addressed explicitly to us */ |
70dc107b SE |
372 | static void netlink_packet_local(struct netlink *st, |
373 | struct netlink_client *client, | |
374 | struct buffer_if *buf) | |
4efd681a SE |
375 | { |
376 | struct icmphdr *h; | |
377 | ||
378 | h=(struct icmphdr *)buf->start; | |
379 | ||
380 | if ((ntohs(h->iph.frag_off)&0xbfff)!=0) { | |
9d3a4132 SE |
381 | Message(M_WARNING,"%s: fragmented packet addressed to secnet; " |
382 | "ignoring it\n",st->name); | |
4efd681a SE |
383 | BUF_FREE(buf); |
384 | return; | |
385 | } | |
386 | ||
387 | if (h->iph.protocol==1) { | |
388 | /* It's ICMP */ | |
389 | if (h->type==8 && h->code==0) { | |
390 | /* ICMP echo-request. Special case: we re-use the buffer | |
391 | to construct the reply. */ | |
392 | h->type=0; | |
393 | h->iph.daddr=h->iph.saddr; | |
394 | h->iph.saddr=htonl(st->secnet_address); | |
395 | h->iph.ttl=255; /* Be nice and bump it up again... */ | |
396 | h->iph.check=0; | |
397 | h->iph.check=ip_fast_csum((uint8_t *)h,h->iph.ihl); | |
398 | netlink_icmp_csum(h); | |
70dc107b | 399 | netlink_packet_deliver(st,NULL,buf); |
4efd681a SE |
400 | return; |
401 | } | |
402 | Message(M_WARNING,"%s: unknown incoming ICMP\n",st->name); | |
403 | } else { | |
404 | /* Send ICMP protocol unreachable */ | |
70dc107b | 405 | netlink_icmp_simple(st,buf,client,3,2); |
4efd681a SE |
406 | BUF_FREE(buf); |
407 | return; | |
408 | } | |
409 | ||
410 | BUF_FREE(buf); | |
411 | } | |
412 | ||
9d3a4132 SE |
413 | /* If cid==NULL packet is from host, otherwise cid specifies which tunnel |
414 | it came from. */ | |
415 | static void netlink_incoming(void *sst, void *cid, struct buffer_if *buf) | |
4efd681a SE |
416 | { |
417 | struct netlink *st=sst; | |
9d3a4132 | 418 | struct netlink_client *client=cid; |
4efd681a SE |
419 | uint32_t source,dest; |
420 | struct iphdr *iph; | |
421 | ||
422 | BUF_ASSERT_USED(buf); | |
423 | if (!netlink_check(st,buf)) { | |
9d3a4132 SE |
424 | Message(M_WARNING,"%s: bad IP packet from %s\n", |
425 | st->name,client?client->name:"host"); | |
4efd681a SE |
426 | BUF_FREE(buf); |
427 | return; | |
428 | } | |
429 | iph=(struct iphdr *)buf->start; | |
430 | ||
431 | source=ntohl(iph->saddr); | |
432 | dest=ntohl(iph->daddr); | |
433 | ||
9d3a4132 SE |
434 | /* Check source */ |
435 | if (client) { | |
436 | /* Check that the packet source is in 'nets' and its destination is | |
437 | in st->networks */ | |
438 | if (!subnet_matches_list(client->networks,source)) { | |
439 | string_t s,d; | |
440 | s=ipaddr_to_string(source); | |
441 | d=ipaddr_to_string(dest); | |
442 | Message(M_WARNING,"%s: packet from tunnel %s with bad " | |
443 | "source address (s=%s,d=%s)\n",st->name,client->name,s,d); | |
444 | free(s); free(d); | |
445 | BUF_FREE(buf); | |
446 | return; | |
447 | } | |
448 | } else { | |
449 | if (!subnet_matches_list(&st->networks,source)) { | |
450 | string_t s,d; | |
451 | s=ipaddr_to_string(source); | |
452 | d=ipaddr_to_string(dest); | |
453 | Message(M_WARNING,"%s: outgoing packet with bad source address " | |
454 | "(s=%s,d=%s)\n",st->name,s,d); | |
455 | free(s); free(d); | |
456 | BUF_FREE(buf); | |
457 | return; | |
458 | } | |
4efd681a | 459 | } |
9d3a4132 SE |
460 | /* (st->secnet_address needs checking before matching destination |
461 | addresses) */ | |
2fe58dfd | 462 | if (dest==st->secnet_address) { |
9d3a4132 | 463 | netlink_packet_local(st,client,buf); |
4efd681a | 464 | BUF_ASSERT_FREE(buf); |
2fe58dfd SE |
465 | return; |
466 | } | |
9d3a4132 | 467 | if (client) { |
3454dce4 | 468 | /* Check for free routing */ |
9d3a4132 SE |
469 | if (!subnet_matches_list(&st->networks,dest)) { |
470 | string_t s,d; | |
471 | s=ipaddr_to_string(source); | |
472 | d=ipaddr_to_string(dest); | |
473 | Message(M_WARNING,"%s: incoming packet from tunnel %s " | |
474 | "with bad destination address " | |
475 | "(s=%s,d=%s)\n",st->name,client->name,s,d); | |
476 | free(s); free(d); | |
477 | BUF_FREE(buf); | |
478 | return; | |
479 | } | |
2fe58dfd | 480 | } |
70dc107b | 481 | netlink_packet_forward(st,client,buf); |
4efd681a SE |
482 | BUF_ASSERT_FREE(buf); |
483 | } | |
484 | ||
9d3a4132 SE |
485 | static void netlink_set_softlinks(struct netlink *st, struct netlink_client *c, |
486 | bool_t up) | |
4efd681a | 487 | { |
9d3a4132 | 488 | uint32_t i; |
4efd681a | 489 | |
9d3a4132 SE |
490 | if (!st->routes) return; /* Table has not yet been created */ |
491 | for (i=0; i<st->n_routes; i++) { | |
492 | if (!st->routes[i].hard && st->routes[i].c==c) { | |
493 | st->routes[i].up=up; | |
494 | st->set_route(st->dst,&st->routes[i]); | |
495 | } | |
4efd681a | 496 | } |
4efd681a SE |
497 | } |
498 | ||
70dc107b | 499 | static void netlink_set_quality(void *sst, void *cid, uint32_t quality) |
4efd681a | 500 | { |
9d3a4132 | 501 | struct netlink *st=sst; |
4efd681a SE |
502 | struct netlink_client *c=cid; |
503 | ||
70dc107b | 504 | c->link_quality=quality; |
9d3a4132 SE |
505 | if (c->link_quality==LINK_QUALITY_DOWN) { |
506 | netlink_set_softlinks(st,c,False); | |
507 | } else { | |
508 | netlink_set_softlinks(st,c,True); | |
509 | } | |
4efd681a SE |
510 | } |
511 | ||
512 | static void *netlink_regnets(void *sst, struct subnet_list *nets, | |
513 | netlink_deliver_fn *deliver, void *dst, | |
514 | uint32_t max_start_pad, uint32_t max_end_pad, | |
3454dce4 | 515 | uint32_t options, string_t client_name) |
4efd681a SE |
516 | { |
517 | struct netlink *st=sst; | |
518 | struct netlink_client *c; | |
519 | ||
520 | Message(M_DEBUG_CONFIG,"netlink_regnets: request for %d networks, " | |
521 | "max_start_pad=%d, max_end_pad=%d\n", | |
522 | nets->entries,max_start_pad,max_end_pad); | |
523 | ||
3454dce4 | 524 | if ((options&NETLINK_OPTION_SOFTROUTE) && !st->set_route) { |
9d3a4132 SE |
525 | Message(M_ERROR,"%s: this netlink device does not support " |
526 | "soft routes.\n"); | |
527 | return NULL; | |
528 | } | |
529 | ||
3454dce4 | 530 | if (options&NETLINK_OPTION_SOFTROUTE) { |
9d3a4132 | 531 | /* XXX for now we assume that soft routes require root privilege; |
3454dce4 | 532 | this may not always be true. The device driver can tell us. */ |
9d3a4132 SE |
533 | require_root_privileges=True; |
534 | require_root_privileges_explanation="netlink: soft routes"; | |
535 | } | |
536 | ||
9d3a4132 SE |
537 | /* Check that nets do not intersect st->exclude_remote_networks; |
538 | refuse to register if they do. */ | |
baa06aeb SE |
539 | if (subnet_lists_intersect(&st->exclude_remote_networks,nets)) { |
540 | Message(M_ERROR,"%s: site %s specifies networks that " | |
541 | "intersect with the explicitly excluded remote networks\n", | |
542 | st->name,client_name); | |
543 | return False; | |
544 | } | |
545 | ||
4efd681a SE |
546 | c=safe_malloc(sizeof(*c),"netlink_regnets"); |
547 | c->networks=nets; | |
548 | c->deliver=deliver; | |
549 | c->dst=dst; | |
550 | c->name=client_name; /* XXX copy it? */ | |
3454dce4 | 551 | c->options=options; |
70dc107b | 552 | c->link_quality=LINK_QUALITY_DOWN; |
4efd681a SE |
553 | c->next=st->clients; |
554 | st->clients=c; | |
555 | if (max_start_pad > st->max_start_pad) st->max_start_pad=max_start_pad; | |
556 | if (max_end_pad > st->max_end_pad) st->max_end_pad=max_end_pad; | |
70dc107b | 557 | st->n_routes+=nets->entries; |
4efd681a SE |
558 | |
559 | return c; | |
560 | } | |
561 | ||
9d3a4132 SE |
562 | static void netlink_dump_routes(struct netlink *st) |
563 | { | |
564 | int i; | |
565 | string_t net; | |
566 | ||
567 | Message(M_INFO,"%s: routing table:\n",st->name); | |
568 | for (i=0; i<st->n_routes; i++) { | |
569 | net=subnet_to_string(&st->routes[i].net); | |
3454dce4 SE |
570 | Message(M_INFO,"%s -> tunnel %s (%s,%s route,%s)\n",net, |
571 | st->routes[i].c->name, | |
9d3a4132 | 572 | st->routes[i].hard?"hard":"soft", |
3454dce4 | 573 | st->routes[i].allow_route?"free":"restricted", |
9d3a4132 SE |
574 | st->routes[i].up?"up":"down"); |
575 | free(net); | |
576 | } | |
577 | Message(M_INFO,"%s/32 -> netlink \"%s\"\n", | |
578 | ipaddr_to_string(st->secnet_address),st->name); | |
579 | for (i=0; i<st->networks.entries; i++) { | |
580 | net=subnet_to_string(&st->networks.list[i]); | |
581 | Message(M_INFO,"%s -> host\n",net); | |
582 | free(net); | |
583 | } | |
584 | } | |
585 | ||
70dc107b SE |
586 | static int netlink_compare_route_specificity(const void *ap, const void *bp) |
587 | { | |
588 | const struct netlink_route *a=ap; | |
589 | const struct netlink_route *b=bp; | |
590 | ||
591 | if (a->net.len==b->net.len) return 0; | |
592 | if (a->net.len<b->net.len) return 1; | |
593 | return -1; | |
594 | } | |
595 | ||
596 | static void netlink_phase_hook(void *sst, uint32_t new_phase) | |
597 | { | |
598 | struct netlink *st=sst; | |
599 | struct netlink_client *c; | |
600 | uint32_t i,j; | |
601 | ||
602 | /* All the networks serviced by the various tunnels should now | |
603 | * have been registered. We build a routing table by sorting the | |
604 | * routes into most-specific-first order. */ | |
605 | st->routes=safe_malloc(st->n_routes*sizeof(*st->routes), | |
606 | "netlink_phase_hook"); | |
607 | /* Fill the table */ | |
608 | i=0; | |
609 | for (c=st->clients; c; c=c->next) { | |
610 | for (j=0; j<c->networks->entries; j++) { | |
611 | st->routes[i].net=c->networks->list[j]; | |
612 | st->routes[i].c=c; | |
3454dce4 SE |
613 | /* Hard routes are always up; |
614 | soft routes default to down */ | |
615 | st->routes[i].up=c->options&NETLINK_OPTION_SOFTROUTE?False:True; | |
9d3a4132 | 616 | st->routes[i].kup=False; |
3454dce4 SE |
617 | st->routes[i].hard=c->options&NETLINK_OPTION_SOFTROUTE?False:True; |
618 | st->routes[i].allow_route=c->options&NETLINK_OPTION_ALLOW_ROUTE? | |
619 | True:False; | |
70dc107b SE |
620 | i++; |
621 | } | |
622 | } | |
623 | /* ASSERT i==st->n_routes */ | |
624 | if (i!=st->n_routes) { | |
625 | fatal("netlink: route count error: expected %d got %d\n", | |
626 | st->n_routes,i); | |
627 | } | |
628 | /* Sort the table in descending order of specificity */ | |
629 | qsort(st->routes,st->n_routes,sizeof(*st->routes), | |
630 | netlink_compare_route_specificity); | |
9d3a4132 SE |
631 | |
632 | netlink_dump_routes(st); | |
70dc107b SE |
633 | } |
634 | ||
9d3a4132 SE |
635 | netlink_deliver_fn *netlink_init(struct netlink *st, |
636 | void *dst, struct cloc loc, | |
637 | dict_t *dict, string_t description, | |
638 | netlink_route_fn *set_route, | |
639 | netlink_deliver_fn *to_host) | |
4efd681a SE |
640 | { |
641 | st->dst=dst; | |
642 | st->cl.description=description; | |
643 | st->cl.type=CL_NETLINK; | |
644 | st->cl.apply=NULL; | |
645 | st->cl.interface=&st->ops; | |
646 | st->ops.st=st; | |
647 | st->ops.regnets=netlink_regnets; | |
9d3a4132 | 648 | st->ops.deliver=netlink_incoming; |
70dc107b | 649 | st->ops.set_quality=netlink_set_quality; |
4efd681a SE |
650 | st->max_start_pad=0; |
651 | st->max_end_pad=0; | |
652 | st->clients=NULL; | |
9d3a4132 | 653 | st->set_route=set_route; |
4efd681a SE |
654 | st->deliver_to_host=to_host; |
655 | ||
656 | st->name=dict_read_string(dict,"name",False,"netlink",loc); | |
657 | if (!st->name) st->name=description; | |
658 | dict_read_subnet_list(dict, "networks", True, "netlink", loc, | |
659 | &st->networks); | |
baa06aeb SE |
660 | dict_read_subnet_list(dict, "exclude-remote-networks", False, "netlink", |
661 | loc, &st->exclude_remote_networks); | |
b2a56f7c SE |
662 | /* secnet-address does not have to be in local-networks; |
663 | however, it should be advertised in the 'sites' file for the | |
baa06aeb | 664 | local site. */ |
4efd681a SE |
665 | st->secnet_address=string_to_ipaddr( |
666 | dict_find_item(dict,"secnet-address", True, "netlink", loc),"netlink"); | |
4efd681a SE |
667 | st->mtu=dict_read_number(dict, "mtu", False, "netlink", loc, DEFAULT_MTU); |
668 | buffer_new(&st->icmp,ICMP_BUFSIZE); | |
70dc107b SE |
669 | st->n_routes=0; |
670 | st->routes=NULL; | |
671 | ||
672 | add_hook(PHASE_SETUP,netlink_phase_hook,st); | |
4efd681a | 673 | |
9d3a4132 | 674 | return netlink_incoming; |
2fe58dfd SE |
675 | } |
676 | ||
9d3a4132 | 677 | /* No connection to the kernel at all... */ |
2fe58dfd | 678 | |
9d3a4132 | 679 | struct null { |
4efd681a | 680 | struct netlink nl; |
4efd681a | 681 | }; |
2fe58dfd | 682 | |
9d3a4132 | 683 | static bool_t null_set_route(void *sst, struct netlink_route *route) |
4efd681a | 684 | { |
9d3a4132 SE |
685 | struct null *st=sst; |
686 | string_t t; | |
4efd681a | 687 | |
9d3a4132 SE |
688 | if (route->up!=route->kup) { |
689 | t=subnet_to_string(&route->net); | |
690 | Message(M_INFO,"%s: setting route %s to state %s\n",st->nl.name, | |
691 | t, route->up?"up":"down"); | |
692 | free(t); | |
693 | route->kup=route->up; | |
694 | return True; | |
2fe58dfd | 695 | } |
9d3a4132 | 696 | return False; |
2fe58dfd | 697 | } |
9d3a4132 | 698 | |
2fe58dfd SE |
699 | static void null_deliver(void *sst, void *cid, struct buffer_if *buf) |
700 | { | |
701 | return; | |
702 | } | |
703 | ||
704 | static list_t *null_apply(closure_t *self, struct cloc loc, dict_t *context, | |
705 | list_t *args) | |
706 | { | |
707 | struct null *st; | |
4efd681a SE |
708 | item_t *item; |
709 | dict_t *dict; | |
2fe58dfd | 710 | |
4efd681a | 711 | st=safe_malloc(sizeof(*st),"null_apply"); |
2fe58dfd | 712 | |
4efd681a SE |
713 | item=list_elem(args,0); |
714 | if (!item || item->type!=t_dict) | |
715 | cfgfatal(loc,"null-netlink","parameter must be a dictionary\n"); | |
716 | ||
717 | dict=item->data.dict; | |
718 | ||
9d3a4132 SE |
719 | netlink_init(&st->nl,st,loc,dict,"null-netlink",null_set_route, |
720 | null_deliver); | |
4efd681a SE |
721 | |
722 | return new_closure(&st->nl.cl); | |
2fe58dfd SE |
723 | } |
724 | ||
725 | init_module netlink_module; | |
726 | void netlink_module(dict_t *dict) | |
727 | { | |
4efd681a | 728 | add_closure(dict,"null-netlink",null_apply); |
2fe58dfd | 729 | } |