Commit | Line | Data |
---|---|---|
2fe58dfd SE |
1 | /* User-kernel network link */ |
2 | ||
ff05a229 | 3 | /* See RFCs 791, 792, 1123 and 1812 */ |
2fe58dfd | 4 | |
ff05a229 SE |
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'). */ | |
8 | ||
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. */ | |
12 | ||
13 | ||
14 | /* Points to note from RFC1812 (which may require changes in this | |
15 | file): | |
16 | ||
17 | 3.3.4 Maximum Transmission Unit - MTU | |
18 | ||
19 | The MTU of each logical interface MUST be configurable within the | |
20 | range of legal MTUs for the interface. | |
21 | ||
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 | |
27 | the MTU. | |
28 | ||
29 | 4.2.1 A router SHOULD count datagrams discarded. | |
30 | ||
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 | |
33 | their use. | |
34 | ||
35 | 5.3.13.4 Source Route Options | |
36 | ||
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. | |
41 | ||
42 | 5.3.13.5 Record Route Option | |
43 | ||
44 | Routers MUST support the Record Route option in forwarded packets. | |
45 | ||
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]). | |
53 | ||
54 | 5.3.13.6 Timestamp Option | |
55 | ||
56 | Routers MUST support the timestamp option in forwarded packets. A | |
57 | timestamp value MUST follow the rules given [INTRO:2]. | |
58 | ||
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 | |
64 | it will be sent. | |
65 | ||
66 | ||
67 | 4.2.2.7 Fragmentation: RFC 791 Section 3.2 | |
68 | ||
69 | Fragmentation, as described in [INTERNET:1], MUST be supported by a | |
70 | router. | |
71 | ||
72 | 4.2.2.8 Reassembly: RFC 791 Section 3.2 | |
73 | ||
74 | As specified in the corresponding section of [INTRO:2], a router MUST | |
75 | support reassembly of datagrams that it delivers to itself. | |
76 | ||
77 | 4.2.2.9 Time to Live: RFC 791 Section 3.2 | |
78 | ||
79 | Note in particular that a router MUST NOT check the TTL of a packet | |
80 | except when forwarding it. | |
81 | ||
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. | |
85 | ||
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. | |
89 | ||
90 | ||
91 | 8.1 The Simple Network Management Protocol - SNMP | |
92 | 8.1.1 SNMP Protocol Elements | |
93 | ||
94 | Routers MUST be manageable by SNMP [MGT:3]. The SNMP MUST operate | |
95 | using UDP/IP as its transport and network protocols. | |
96 | ||
97 | ||
98 | */ | |
2fe58dfd | 99 | |
3b83c932 | 100 | #include <string.h> |
59230b9b IJ |
101 | #include <assert.h> |
102 | #include <limits.h> | |
8689b3a9 | 103 | #include "secnet.h" |
2fe58dfd | 104 | #include "util.h" |
7138d0c5 | 105 | #include "ipaddr.h" |
9d3a4132 | 106 | #include "netlink.h" |
042a8da9 | 107 | #include "process.h" |
2fe58dfd | 108 | |
ff05a229 SE |
109 | #define ICMP_TYPE_ECHO_REPLY 0 |
110 | ||
111 | #define ICMP_TYPE_UNREACHABLE 3 | |
112 | #define ICMP_CODE_NET_UNREACHABLE 0 | |
113 | #define ICMP_CODE_PROTOCOL_UNREACHABLE 2 | |
114 | #define ICMP_CODE_FRAGMENTATION_REQUIRED 4 | |
115 | #define ICMP_CODE_NET_PROHIBITED 13 | |
116 | ||
117 | #define ICMP_TYPE_ECHO_REQUEST 8 | |
118 | ||
119 | #define ICMP_TYPE_TIME_EXCEEDED 11 | |
120 | #define ICMP_CODE_TTL_EXCEEDED 0 | |
121 | ||
4efd681a | 122 | /* Generic IP checksum routine */ |
211cd627 | 123 | static inline uint16_t ip_csum(const uint8_t *iph,int32_t count) |
2fe58dfd | 124 | { |
4efd681a SE |
125 | register uint32_t sum=0; |
126 | ||
127 | while (count>1) { | |
128 | sum+=ntohs(*(uint16_t *)iph); | |
129 | iph+=2; | |
130 | count-=2; | |
131 | } | |
132 | if(count>0) | |
133 | sum+=*(uint8_t *)iph; | |
134 | while (sum>>16) | |
135 | sum=(sum&0xffff)+(sum>>16); | |
136 | return htons(~sum); | |
2fe58dfd SE |
137 | } |
138 | ||
4efd681a SE |
139 | #ifdef i386 |
140 | /* | |
141 | * This is a version of ip_compute_csum() optimized for IP headers, | |
142 | * which always checksum on 4 octet boundaries. | |
143 | * | |
144 | * By Jorge Cwik <jorge@laser.satlink.net>, adapted for linux by | |
145 | * Arnt Gulbrandsen. | |
146 | */ | |
211cd627 | 147 | static inline uint16_t ip_fast_csum(const uint8_t *iph, int32_t ihl) { |
4efd681a SE |
148 | uint32_t sum; |
149 | ||
20d324b6 SE |
150 | __asm__ __volatile__( |
151 | "movl (%1), %0 ;\n" | |
152 | "subl $4, %2 ;\n" | |
153 | "jbe 2f ;\n" | |
154 | "addl 4(%1), %0 ;\n" | |
155 | "adcl 8(%1), %0 ;\n" | |
156 | "adcl 12(%1), %0 ;\n" | |
157 | "1: adcl 16(%1), %0 ;\n" | |
158 | "lea 4(%1), %1 ;\n" | |
159 | "decl %2 ;\n" | |
160 | "jne 1b ;\n" | |
161 | "adcl $0, %0 ;\n" | |
162 | "movl %0, %2 ;\n" | |
163 | "shrl $16, %0 ;\n" | |
164 | "addw %w2, %w0 ;\n" | |
165 | "adcl $0, %0 ;\n" | |
166 | "notl %0 ;\n" | |
167 | "2: ;\n" | |
4efd681a SE |
168 | /* Since the input registers which are loaded with iph and ipl |
169 | are modified, we must also specify them as outputs, or gcc | |
170 | will assume they contain their original values. */ | |
171 | : "=r" (sum), "=r" (iph), "=r" (ihl) | |
20d324b6 SE |
172 | : "1" (iph), "2" (ihl) |
173 | : "memory"); | |
4efd681a SE |
174 | return sum; |
175 | } | |
176 | #else | |
1caa23ff | 177 | static inline uint16_t ip_fast_csum(uint8_t *iph, int32_t ihl) |
2fe58dfd | 178 | { |
1caa23ff | 179 | assert(ihl < INT_MAX/4); |
4efd681a SE |
180 | return ip_csum(iph,ihl*4); |
181 | } | |
182 | #endif | |
183 | ||
184 | struct iphdr { | |
185 | #if defined (WORDS_BIGENDIAN) | |
186 | uint8_t version:4, | |
187 | ihl:4; | |
188 | #else | |
189 | uint8_t ihl:4, | |
190 | version:4; | |
191 | #endif | |
192 | uint8_t tos; | |
193 | uint16_t tot_len; | |
194 | uint16_t id; | |
a6768d7c | 195 | uint16_t frag; |
eff13010 IJ |
196 | #define IPHDR_FRAG_OFF ((uint16_t)0x1fff) |
197 | #define IPHDR_FRAG_MORE ((uint16_t)0x2000) | |
198 | #define IPHDR_FRAG_DONT ((uint16_t)0x4000) | |
199 | /* reserved 0x8000 */ | |
4efd681a SE |
200 | uint8_t ttl; |
201 | uint8_t protocol; | |
202 | uint16_t check; | |
203 | uint32_t saddr; | |
204 | uint32_t daddr; | |
205 | /* The options start here. */ | |
206 | }; | |
207 | ||
208 | struct icmphdr { | |
209 | struct iphdr iph; | |
210 | uint8_t type; | |
211 | uint8_t code; | |
212 | uint16_t check; | |
cfd79482 | 213 | union icmpinfofield { |
4efd681a SE |
214 | uint32_t unused; |
215 | struct { | |
216 | uint8_t pointer; | |
217 | uint8_t unused1; | |
218 | uint16_t unused2; | |
219 | } pprob; | |
220 | uint32_t gwaddr; | |
221 | struct { | |
222 | uint16_t id; | |
223 | uint16_t seq; | |
224 | } echo; | |
225 | } d; | |
226 | }; | |
cfd79482 IJ |
227 | |
228 | static const union icmpinfofield icmp_noinfo; | |
4efd681a | 229 | |
70dc107b SE |
230 | static void netlink_packet_deliver(struct netlink *st, |
231 | struct netlink_client *client, | |
232 | struct buffer_if *buf); | |
4efd681a | 233 | |
ff05a229 SE |
234 | /* XXX RFC1812 4.3.2.5: |
235 | All other ICMP error messages (Destination Unreachable, | |
236 | Redirect, Time Exceeded, and Parameter Problem) SHOULD have their | |
237 | precedence value set to 6 (INTERNETWORK CONTROL) or 7 (NETWORK | |
238 | CONTROL). The IP Precedence value for these error messages MAY be | |
239 | settable. | |
240 | */ | |
4efd681a SE |
241 | static struct icmphdr *netlink_icmp_tmpl(struct netlink *st, |
242 | uint32_t dest,uint16_t len) | |
243 | { | |
244 | struct icmphdr *h; | |
245 | ||
246 | BUF_ALLOC(&st->icmp,"netlink_icmp_tmpl"); | |
3abd18e8 | 247 | buffer_init(&st->icmp,calculate_max_start_pad()); |
4efd681a SE |
248 | h=buf_append(&st->icmp,sizeof(*h)); |
249 | ||
250 | h->iph.version=4; | |
251 | h->iph.ihl=5; | |
252 | h->iph.tos=0; | |
253 | h->iph.tot_len=htons(len+(h->iph.ihl*4)+8); | |
254 | h->iph.id=0; | |
a6768d7c | 255 | h->iph.frag=0; |
ff05a229 | 256 | h->iph.ttl=255; /* XXX should be configurable */ |
4efd681a SE |
257 | h->iph.protocol=1; |
258 | h->iph.saddr=htonl(st->secnet_address); | |
259 | h->iph.daddr=htonl(dest); | |
260 | h->iph.check=0; | |
261 | h->iph.check=ip_fast_csum((uint8_t *)&h->iph,h->iph.ihl); | |
262 | h->check=0; | |
263 | h->d.unused=0; | |
264 | ||
265 | return h; | |
266 | } | |
267 | ||
268 | /* Fill in the ICMP checksum field correctly */ | |
269 | static void netlink_icmp_csum(struct icmphdr *h) | |
270 | { | |
1caa23ff | 271 | int32_t len; |
4efd681a SE |
272 | |
273 | len=ntohs(h->iph.tot_len)-(4*h->iph.ihl); | |
274 | h->check=0; | |
275 | h->check=ip_csum(&h->type,len); | |
276 | } | |
277 | ||
278 | /* RFC1122: | |
279 | * An ICMP error message MUST NOT be sent as the result of | |
280 | * receiving: | |
281 | * | |
282 | * * an ICMP error message, or | |
283 | * | |
284 | * * a datagram destined to an IP broadcast or IP multicast | |
285 | * address, or | |
286 | * | |
287 | * * a datagram sent as a link-layer broadcast, or | |
288 | * | |
289 | * * a non-initial fragment, or | |
290 | * | |
291 | * * a datagram whose source address does not define a single | |
292 | * host -- e.g., a zero address, a loopback address, a | |
293 | * broadcast address, a multicast address, or a Class E | |
294 | * address. | |
295 | */ | |
296 | static bool_t netlink_icmp_may_reply(struct buffer_if *buf) | |
297 | { | |
298 | struct iphdr *iph; | |
8dea8d37 | 299 | struct icmphdr *icmph; |
4efd681a SE |
300 | uint32_t source; |
301 | ||
975820aa | 302 | if (buf->size < (int)sizeof(struct icmphdr)) return False; |
4efd681a | 303 | iph=(struct iphdr *)buf->start; |
8dea8d37 SE |
304 | icmph=(struct icmphdr *)buf->start; |
305 | if (iph->protocol==1) { | |
306 | switch(icmph->type) { | |
686b7f1d IJ |
307 | /* Based on http://www.iana.org/assignments/icmp-parameters/icmp-parameters.xhtml#icmp-parameters-types |
308 | * as retrieved Thu, 20 Mar 2014 00:16:44 +0000. | |
309 | * Deprecated, reserved, unassigned and experimental | |
310 | * options are treated as not safe to reply to. | |
311 | */ | |
312 | case 0: /* Echo Reply */ | |
313 | case 8: /* Echo */ | |
314 | case 13: /* Timestamp */ | |
315 | case 14: /* Timestamp Reply */ | |
316 | return True; | |
317 | default: | |
8dea8d37 SE |
318 | return False; |
319 | } | |
320 | } | |
4efd681a | 321 | /* How do we spot broadcast destination addresses? */ |
a6768d7c | 322 | if (ntohs(iph->frag)&IPHDR_FRAG_OFF) return False; |
4efd681a SE |
323 | source=ntohl(iph->saddr); |
324 | if (source==0) return False; | |
325 | if ((source&0xff000000)==0x7f000000) return False; | |
326 | /* How do we spot broadcast source addresses? */ | |
327 | if ((source&0xf0000000)==0xe0000000) return False; /* Multicast */ | |
328 | if ((source&0xf0000000)==0xf0000000) return False; /* Class E */ | |
329 | return True; | |
330 | } | |
331 | ||
332 | /* How much of the original IP packet do we include in its ICMP | |
333 | response? The header plus up to 64 bits. */ | |
ff05a229 SE |
334 | |
335 | /* XXX TODO RFC1812: | |
336 | 4.3.2.3 Original Message Header | |
337 | ||
338 | Historically, every ICMP error message has included the Internet | |
339 | header and at least the first 8 data bytes of the datagram that | |
340 | triggered the error. This is no longer adequate, due to the use of | |
341 | IP-in-IP tunneling and other technologies. Therefore, the ICMP | |
342 | datagram SHOULD contain as much of the original datagram as possible | |
343 | without the length of the ICMP datagram exceeding 576 bytes. The | |
344 | returned IP header (and user data) MUST be identical to that which | |
345 | was received, except that the router is not required to undo any | |
346 | modifications to the IP header that are normally performed in | |
347 | forwarding that were performed before the error was detected (e.g., | |
348 | decrementing the TTL, or updating options). Note that the | |
349 | requirements of Section [4.3.3.5] supersede this requirement in some | |
350 | cases (i.e., for a Parameter Problem message, if the problem is in a | |
351 | modified field, the router must undo the modification). See Section | |
352 | [4.3.3.5]). | |
353 | */ | |
4efd681a SE |
354 | static uint16_t netlink_icmp_reply_len(struct buffer_if *buf) |
355 | { | |
975820aa | 356 | if (buf->size < (int)sizeof(struct iphdr)) return 0; |
4efd681a SE |
357 | struct iphdr *iph=(struct iphdr *)buf->start; |
358 | uint16_t hlen,plen; | |
359 | ||
360 | hlen=iph->ihl*4; | |
361 | /* We include the first 8 bytes of the packet data, provided they exist */ | |
362 | hlen+=8; | |
363 | plen=ntohs(iph->tot_len); | |
364 | return (hlen>plen?plen:hlen); | |
365 | } | |
366 | ||
70dc107b SE |
367 | /* client indicates where the packet we're constructing a response to |
368 | comes from. NULL indicates the host. */ | |
4efd681a | 369 | static void netlink_icmp_simple(struct netlink *st, struct buffer_if *buf, |
70dc107b | 370 | struct netlink_client *client, |
cfd79482 IJ |
371 | uint8_t type, uint8_t code, |
372 | union icmpinfofield info) | |
4efd681a | 373 | { |
4efd681a SE |
374 | struct icmphdr *h; |
375 | uint16_t len; | |
376 | ||
377 | if (netlink_icmp_may_reply(buf)) { | |
975820aa | 378 | struct iphdr *iph=(struct iphdr *)buf->start; |
4efd681a SE |
379 | len=netlink_icmp_reply_len(buf); |
380 | h=netlink_icmp_tmpl(st,ntohl(iph->saddr),len); | |
cfd79482 | 381 | h->type=type; h->code=code; h->d=info; |
4efd681a SE |
382 | memcpy(buf_append(&st->icmp,len),buf->start,len); |
383 | netlink_icmp_csum(h); | |
70dc107b | 384 | netlink_packet_deliver(st,NULL,&st->icmp); |
4efd681a SE |
385 | BUF_ASSERT_FREE(&st->icmp); |
386 | } | |
387 | } | |
388 | ||
389 | /* | |
390 | * RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the | |
391 | * checksum. | |
ff05a229 | 392 | * RFC1812: 4.2.2.5 MUST discard messages containing invalid checksums. |
4efd681a SE |
393 | * |
394 | * Is the datagram acceptable? | |
395 | * | |
396 | * 1. Length at least the size of an ip header | |
397 | * 2. Version of 4 | |
398 | * 3. Checksums correctly. | |
399 | * 4. Doesn't have a bogus length | |
400 | */ | |
d714da29 IJ |
401 | static bool_t netlink_check(struct netlink *st, struct buffer_if *buf, |
402 | char *errmsgbuf, int errmsgbuflen) | |
4efd681a | 403 | { |
d714da29 IJ |
404 | #define BAD(...) do{ \ |
405 | snprintf(errmsgbuf,errmsgbuflen,__VA_ARGS__); \ | |
406 | return False; \ | |
407 | }while(0) | |
408 | ||
975820aa | 409 | if (buf->size < (int)sizeof(struct iphdr)) BAD("len %"PRIu32"",buf->size); |
4efd681a | 410 | struct iphdr *iph=(struct iphdr *)buf->start; |
1caa23ff | 411 | int32_t len; |
4efd681a | 412 | |
d714da29 IJ |
413 | if (iph->ihl < 5) BAD("ihl %u",iph->ihl); |
414 | if (iph->version != 4) BAD("version %u",iph->version); | |
415 | if (buf->size < iph->ihl*4) BAD("size %"PRId32"<%u*4",buf->size,iph->ihl); | |
416 | if (ip_fast_csum((uint8_t *)iph, iph->ihl)!=0) BAD("csum"); | |
4efd681a SE |
417 | len=ntohs(iph->tot_len); |
418 | /* There should be no padding */ | |
d714da29 IJ |
419 | if (buf->size!=len) BAD("len %"PRId32"!=%"PRId32,buf->size,len); |
420 | if (len<(iph->ihl<<2)) BAD("len %"PRId32"<(%u<<2)",len,iph->ihl); | |
4efd681a SE |
421 | /* XXX check that there's no source route specified */ |
422 | return True; | |
d714da29 IJ |
423 | |
424 | #undef BAD | |
4efd681a SE |
425 | } |
426 | ||
7b6abafa | 427 | /* Deliver a packet _to_ client; used after we have decided |
55bc97e6 IJ |
428 | * what to do with it (and just to check that the client has |
429 | * actually registered a delivery function with us). */ | |
7b6abafa IJ |
430 | static void netlink_client_deliver(struct netlink *st, |
431 | struct netlink_client *client, | |
432 | uint32_t source, uint32_t dest, | |
433 | struct buffer_if *buf) | |
434 | { | |
55bc97e6 IJ |
435 | if (!client->deliver) { |
436 | string_t s,d; | |
437 | s=ipaddr_to_string(source); | |
438 | d=ipaddr_to_string(dest); | |
439 | Message(M_ERR,"%s: dropping %s->%s, client not registered\n", | |
440 | st->name,s,d); | |
441 | free(s); free(d); | |
442 | BUF_FREE(buf); | |
443 | return; | |
444 | } | |
7b6abafa IJ |
445 | client->deliver(client->dst, buf); |
446 | client->outcount++; | |
447 | } | |
448 | ||
469fd1d9 | 449 | /* Deliver a packet. "client" is the _origin_ of the packet, not its |
d3fe100d SE |
450 | destination, and is NULL for packets from the host and packets |
451 | generated internally in secnet. */ | |
70dc107b SE |
452 | static void netlink_packet_deliver(struct netlink *st, |
453 | struct netlink_client *client, | |
454 | struct buffer_if *buf) | |
4efd681a | 455 | { |
975820aa IJ |
456 | if (buf->size < (int)sizeof(struct iphdr)) { |
457 | Message(M_ERR,"%s: trying to deliver a too-short packet" | |
458 | " from %s!\n",st->name, client?client->name:"(local)"); | |
459 | BUF_FREE(buf); | |
460 | return; | |
461 | } | |
462 | ||
4efd681a SE |
463 | struct iphdr *iph=(struct iphdr *)buf->start; |
464 | uint32_t dest=ntohl(iph->daddr); | |
70dc107b SE |
465 | uint32_t source=ntohl(iph->saddr); |
466 | uint32_t best_quality; | |
469fd1d9 SE |
467 | bool_t allow_route=False; |
468 | bool_t found_allowed=False; | |
70dc107b SE |
469 | int best_match; |
470 | int i; | |
2fe58dfd | 471 | |
4efd681a | 472 | BUF_ASSERT_USED(buf); |
2fe58dfd | 473 | |
4efd681a | 474 | if (dest==st->secnet_address) { |
4f5e39ec | 475 | Message(M_ERR,"%s: trying to deliver a packet to myself!\n",st->name); |
4efd681a | 476 | BUF_FREE(buf); |
2fe58dfd SE |
477 | return; |
478 | } | |
4efd681a | 479 | |
d3fe100d | 480 | /* Packets from the host (client==NULL) may always be routed. Packets |
469fd1d9 SE |
481 | from clients with the allow_route option will also be routed. */ |
482 | if (!client || (client && (client->options & OPT_ALLOWROUTE))) | |
483 | allow_route=True; | |
484 | ||
485 | /* If !allow_route, we check the routing table anyway, and if | |
486 | there's a suitable route with OPT_ALLOWROUTE set we use it. If | |
487 | there's a suitable route, but none with OPT_ALLOWROUTE set then | |
488 | we generate ICMP 'communication with destination network | |
489 | administratively prohibited'. */ | |
490 | ||
491 | best_quality=0; | |
492 | best_match=-1; | |
d3fe100d SE |
493 | for (i=0; i<st->n_clients; i++) { |
494 | if (st->routes[i]->up && | |
495 | ipset_contains_addr(st->routes[i]->networks,dest)) { | |
469fd1d9 SE |
496 | /* It's an available route to the correct destination. But is |
497 | it better than the one we already have? */ | |
498 | ||
499 | /* If we have already found an allowed route then we don't | |
500 | bother looking at routes we're not allowed to use. If | |
501 | we don't yet have an allowed route we'll consider any. */ | |
502 | if (!allow_route && found_allowed) { | |
d3fe100d | 503 | if (!(st->routes[i]->options&OPT_ALLOWROUTE)) continue; |
70dc107b | 504 | } |
469fd1d9 | 505 | |
d3fe100d | 506 | if (st->routes[i]->link_quality>best_quality |
469fd1d9 | 507 | || best_quality==0) { |
d3fe100d | 508 | best_quality=st->routes[i]->link_quality; |
469fd1d9 | 509 | best_match=i; |
d3fe100d | 510 | if (st->routes[i]->options&OPT_ALLOWROUTE) |
469fd1d9 SE |
511 | found_allowed=True; |
512 | /* If quality isn't perfect we may wish to | |
513 | consider kicking the tunnel with a 0-length | |
514 | packet to prompt it to perform a key setup. | |
515 | Then it'll eventually decide it's up or | |
516 | down. */ | |
517 | /* If quality is perfect and we're allowed to use the | |
518 | route we don't need to search any more. */ | |
519 | if (best_quality>=MAXIMUM_LINK_QUALITY && | |
520 | (allow_route || found_allowed)) break; | |
4efd681a | 521 | } |
70dc107b | 522 | } |
469fd1d9 SE |
523 | } |
524 | if (best_match==-1) { | |
525 | /* The packet's not going down a tunnel. It might (ought to) | |
526 | be for the host. */ | |
794f2398 | 527 | if (ipset_contains_addr(st->networks,dest)) { |
469fd1d9 SE |
528 | st->deliver_to_host(st->dst,buf); |
529 | st->outcount++; | |
70dc107b SE |
530 | BUF_ASSERT_FREE(buf); |
531 | } else { | |
469fd1d9 SE |
532 | string_t s,d; |
533 | s=ipaddr_to_string(source); | |
534 | d=ipaddr_to_string(dest); | |
ff05a229 | 535 | Message(M_DEBUG,"%s: don't know where to deliver packet " |
469fd1d9 SE |
536 | "(s=%s, d=%s)\n", st->name, s, d); |
537 | free(s); free(d); | |
ff05a229 | 538 | netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE, |
cfd79482 | 539 | ICMP_CODE_NET_UNREACHABLE, icmp_noinfo); |
70dc107b | 540 | BUF_FREE(buf); |
2fe58dfd | 541 | } |
469fd1d9 SE |
542 | } else { |
543 | if (!allow_route && | |
d3fe100d | 544 | !(st->routes[best_match]->options&OPT_ALLOWROUTE)) { |
469fd1d9 SE |
545 | string_t s,d; |
546 | s=ipaddr_to_string(source); | |
547 | d=ipaddr_to_string(dest); | |
548 | /* We have a usable route but aren't allowed to use it. | |
549 | Generate ICMP destination unreachable: communication | |
550 | with destination network administratively prohibited */ | |
551 | Message(M_NOTICE,"%s: denied forwarding for packet (s=%s, d=%s)\n", | |
552 | st->name,s,d); | |
553 | free(s); free(d); | |
554 | ||
ff05a229 | 555 | netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE, |
cfd79482 | 556 | ICMP_CODE_NET_PROHIBITED, icmp_noinfo); |
469fd1d9 | 557 | BUF_FREE(buf); |
469fd1d9 | 558 | } else { |
ea7ec970 SE |
559 | if (best_quality>0) { |
560 | /* XXX Fragment if required */ | |
7b6abafa IJ |
561 | netlink_client_deliver(st,st->routes[best_match], |
562 | source,dest,buf); | |
ea7ec970 SE |
563 | BUF_ASSERT_FREE(buf); |
564 | } else { | |
565 | /* Generate ICMP destination unreachable */ | |
cfd79482 IJ |
566 | netlink_icmp_simple(st,buf,client,/* client==NULL */ |
567 | ICMP_TYPE_UNREACHABLE, | |
568 | ICMP_CODE_NET_UNREACHABLE, | |
569 | icmp_noinfo); | |
ea7ec970 SE |
570 | BUF_FREE(buf); |
571 | } | |
469fd1d9 | 572 | } |
2fe58dfd | 573 | } |
70dc107b | 574 | BUF_ASSERT_FREE(buf); |
4efd681a SE |
575 | } |
576 | ||
70dc107b SE |
577 | static void netlink_packet_forward(struct netlink *st, |
578 | struct netlink_client *client, | |
579 | struct buffer_if *buf) | |
4efd681a | 580 | { |
975820aa | 581 | if (buf->size < (int)sizeof(struct iphdr)) return; |
4efd681a SE |
582 | struct iphdr *iph=(struct iphdr *)buf->start; |
583 | ||
584 | BUF_ASSERT_USED(buf); | |
585 | ||
586 | /* Packet has already been checked */ | |
587 | if (iph->ttl<=1) { | |
588 | /* Generate ICMP time exceeded */ | |
ff05a229 | 589 | netlink_icmp_simple(st,buf,client,ICMP_TYPE_TIME_EXCEEDED, |
cfd79482 | 590 | ICMP_CODE_TTL_EXCEEDED,icmp_noinfo); |
4efd681a SE |
591 | BUF_FREE(buf); |
592 | return; | |
593 | } | |
594 | iph->ttl--; | |
595 | iph->check=0; | |
596 | iph->check=ip_fast_csum((uint8_t *)iph,iph->ihl); | |
597 | ||
70dc107b | 598 | netlink_packet_deliver(st,client,buf); |
4efd681a SE |
599 | BUF_ASSERT_FREE(buf); |
600 | } | |
601 | ||
9d3a4132 | 602 | /* Deal with packets addressed explicitly to us */ |
70dc107b SE |
603 | static void netlink_packet_local(struct netlink *st, |
604 | struct netlink_client *client, | |
605 | struct buffer_if *buf) | |
4efd681a SE |
606 | { |
607 | struct icmphdr *h; | |
608 | ||
469fd1d9 SE |
609 | st->localcount++; |
610 | ||
975820aa IJ |
611 | if (buf->size < (int)sizeof(struct icmphdr)) { |
612 | Message(M_WARNING,"%s: short packet addressed to secnet; " | |
613 | "ignoring it\n",st->name); | |
614 | BUF_FREE(buf); | |
615 | return; | |
616 | } | |
4efd681a SE |
617 | h=(struct icmphdr *)buf->start; |
618 | ||
a6768d7c | 619 | if ((ntohs(h->iph.frag)&(IPHDR_FRAG_OFF|IPHDR_FRAG_MORE))!=0) { |
9d3a4132 SE |
620 | Message(M_WARNING,"%s: fragmented packet addressed to secnet; " |
621 | "ignoring it\n",st->name); | |
4efd681a SE |
622 | BUF_FREE(buf); |
623 | return; | |
624 | } | |
625 | ||
626 | if (h->iph.protocol==1) { | |
627 | /* It's ICMP */ | |
ff05a229 | 628 | if (h->type==ICMP_TYPE_ECHO_REQUEST && h->code==0) { |
4efd681a SE |
629 | /* ICMP echo-request. Special case: we re-use the buffer |
630 | to construct the reply. */ | |
ff05a229 | 631 | h->type=ICMP_TYPE_ECHO_REPLY; |
4efd681a SE |
632 | h->iph.daddr=h->iph.saddr; |
633 | h->iph.saddr=htonl(st->secnet_address); | |
ff05a229 | 634 | h->iph.ttl=255; |
4efd681a SE |
635 | h->iph.check=0; |
636 | h->iph.check=ip_fast_csum((uint8_t *)h,h->iph.ihl); | |
637 | netlink_icmp_csum(h); | |
70dc107b | 638 | netlink_packet_deliver(st,NULL,buf); |
4efd681a SE |
639 | return; |
640 | } | |
641 | Message(M_WARNING,"%s: unknown incoming ICMP\n",st->name); | |
642 | } else { | |
643 | /* Send ICMP protocol unreachable */ | |
ff05a229 | 644 | netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE, |
cfd79482 | 645 | ICMP_CODE_PROTOCOL_UNREACHABLE,icmp_noinfo); |
4efd681a SE |
646 | BUF_FREE(buf); |
647 | return; | |
648 | } | |
649 | ||
650 | BUF_FREE(buf); | |
651 | } | |
652 | ||
9d3a4132 SE |
653 | /* If cid==NULL packet is from host, otherwise cid specifies which tunnel |
654 | it came from. */ | |
469fd1d9 SE |
655 | static void netlink_incoming(struct netlink *st, struct netlink_client *client, |
656 | struct buffer_if *buf) | |
4efd681a | 657 | { |
4efd681a SE |
658 | uint32_t source,dest; |
659 | struct iphdr *iph; | |
d714da29 | 660 | char errmsgbuf[50]; |
a28d65a5 | 661 | const char *sourcedesc=client?client->name:"host"; |
4efd681a SE |
662 | |
663 | BUF_ASSERT_USED(buf); | |
a28d65a5 | 664 | |
d714da29 IJ |
665 | if (!netlink_check(st,buf,errmsgbuf,sizeof(errmsgbuf))) { |
666 | Message(M_WARNING,"%s: bad IP packet from %s: %s\n", | |
a28d65a5 | 667 | st->name,sourcedesc, |
d714da29 | 668 | errmsgbuf); |
4efd681a SE |
669 | BUF_FREE(buf); |
670 | return; | |
671 | } | |
975820aa | 672 | assert(buf->size >= (int)sizeof(struct icmphdr)); |
4efd681a SE |
673 | iph=(struct iphdr *)buf->start; |
674 | ||
675 | source=ntohl(iph->saddr); | |
676 | dest=ntohl(iph->daddr); | |
677 | ||
d3fe100d SE |
678 | /* Check source. If we don't like the source, there's no point |
679 | generating ICMP because we won't know how to get it to the | |
680 | source of the packet. */ | |
9d3a4132 | 681 | if (client) { |
c6f79b17 SE |
682 | /* Check that the packet source is appropriate for the tunnel |
683 | it came down */ | |
794f2398 | 684 | if (!ipset_contains_addr(client->networks,source)) { |
9d3a4132 SE |
685 | string_t s,d; |
686 | s=ipaddr_to_string(source); | |
687 | d=ipaddr_to_string(dest); | |
688 | Message(M_WARNING,"%s: packet from tunnel %s with bad " | |
689 | "source address (s=%s,d=%s)\n",st->name,client->name,s,d); | |
690 | free(s); free(d); | |
691 | BUF_FREE(buf); | |
692 | return; | |
693 | } | |
694 | } else { | |
c6f79b17 SE |
695 | /* Check that the packet originates in our configured local |
696 | network, and hasn't been forwarded from elsewhere or | |
697 | generated with the wrong source address */ | |
794f2398 | 698 | if (!ipset_contains_addr(st->networks,source)) { |
9d3a4132 SE |
699 | string_t s,d; |
700 | s=ipaddr_to_string(source); | |
701 | d=ipaddr_to_string(dest); | |
702 | Message(M_WARNING,"%s: outgoing packet with bad source address " | |
703 | "(s=%s,d=%s)\n",st->name,s,d); | |
704 | free(s); free(d); | |
705 | BUF_FREE(buf); | |
706 | return; | |
707 | } | |
4efd681a | 708 | } |
c6f79b17 | 709 | |
794f2398 SE |
710 | /* If this is a point-to-point device we don't examine the |
711 | destination address at all; we blindly send it down our | |
712 | one-and-only registered tunnel, or to the host, depending on | |
d3fe100d SE |
713 | where it came from. It's up to external software to check |
714 | address validity and generate ICMP, etc. */ | |
c6f79b17 SE |
715 | if (st->ptp) { |
716 | if (client) { | |
469fd1d9 | 717 | st->deliver_to_host(st->dst,buf); |
c6f79b17 | 718 | } else { |
7b6abafa | 719 | netlink_client_deliver(st,st->clients,source,dest,buf); |
c6f79b17 SE |
720 | } |
721 | BUF_ASSERT_FREE(buf); | |
722 | return; | |
723 | } | |
724 | ||
d3fe100d SE |
725 | /* st->secnet_address needs checking before matching destination |
726 | addresses */ | |
2fe58dfd | 727 | if (dest==st->secnet_address) { |
9d3a4132 | 728 | netlink_packet_local(st,client,buf); |
4efd681a | 729 | BUF_ASSERT_FREE(buf); |
2fe58dfd SE |
730 | return; |
731 | } | |
70dc107b | 732 | netlink_packet_forward(st,client,buf); |
4efd681a SE |
733 | BUF_ASSERT_FREE(buf); |
734 | } | |
735 | ||
469fd1d9 SE |
736 | static void netlink_inst_incoming(void *sst, struct buffer_if *buf) |
737 | { | |
738 | struct netlink_client *c=sst; | |
739 | struct netlink *st=c->nst; | |
740 | ||
741 | netlink_incoming(st,c,buf); | |
742 | } | |
743 | ||
744 | static void netlink_dev_incoming(void *sst, struct buffer_if *buf) | |
745 | { | |
746 | struct netlink *st=sst; | |
747 | ||
748 | netlink_incoming(st,NULL,buf); | |
749 | } | |
750 | ||
d3fe100d | 751 | static void netlink_set_quality(void *sst, uint32_t quality) |
4efd681a | 752 | { |
d3fe100d SE |
753 | struct netlink_client *c=sst; |
754 | struct netlink *st=c->nst; | |
4efd681a | 755 | |
d3fe100d SE |
756 | c->link_quality=quality; |
757 | c->up=(c->link_quality==LINK_QUALITY_DOWN)?False:True; | |
758 | if (c->options&OPT_SOFTROUTE) { | |
759 | st->set_routes(st->dst,c); | |
4efd681a | 760 | } |
4efd681a SE |
761 | } |
762 | ||
d3fe100d SE |
763 | static void netlink_output_subnets(struct netlink *st, uint32_t loglevel, |
764 | struct subnet_list *snets) | |
4efd681a | 765 | { |
1caa23ff | 766 | int32_t i; |
d3fe100d | 767 | string_t net; |
4efd681a | 768 | |
d3fe100d SE |
769 | for (i=0; i<snets->entries; i++) { |
770 | net=subnet_to_string(snets->list[i]); | |
771 | Message(loglevel,"%s ",net); | |
772 | free(net); | |
9d3a4132 | 773 | } |
4efd681a SE |
774 | } |
775 | ||
042a8da9 | 776 | static void netlink_dump_routes(struct netlink *st, bool_t requested) |
9d3a4132 SE |
777 | { |
778 | int i; | |
779 | string_t net; | |
042a8da9 | 780 | uint32_t c=M_INFO; |
9d3a4132 | 781 | |
042a8da9 | 782 | if (requested) c=M_WARNING; |
469fd1d9 SE |
783 | if (st->ptp) { |
784 | net=ipaddr_to_string(st->secnet_address); | |
34d3bf4c | 785 | Message(c,"%s: point-to-point (remote end is %s); routes: ", |
469fd1d9 | 786 | st->name, net); |
9d3a4132 | 787 | free(net); |
d3fe100d | 788 | netlink_output_subnets(st,c,st->clients->subnets); |
469fd1d9 SE |
789 | Message(c,"\n"); |
790 | } else { | |
791 | Message(c,"%s: routing table:\n",st->name); | |
d3fe100d SE |
792 | for (i=0; i<st->n_clients; i++) { |
793 | netlink_output_subnets(st,c,st->routes[i]->subnets); | |
ff05a229 | 794 | Message(c,"-> tunnel %s (%s,mtu %d,%s routes,%s," |
ea7ec970 | 795 | "quality %d,use %d,pri %lu)\n", |
d3fe100d | 796 | st->routes[i]->name, |
ff05a229 SE |
797 | st->routes[i]->up?"up":"down", |
798 | st->routes[i]->mtu, | |
d3fe100d SE |
799 | st->routes[i]->options&OPT_SOFTROUTE?"soft":"hard", |
800 | st->routes[i]->options&OPT_ALLOWROUTE?"free":"restricted", | |
d3fe100d | 801 | st->routes[i]->link_quality, |
ea7ec970 SE |
802 | st->routes[i]->outcount, |
803 | (unsigned long)st->routes[i]->priority); | |
469fd1d9 SE |
804 | } |
805 | net=ipaddr_to_string(st->secnet_address); | |
806 | Message(c,"%s/32 -> netlink \"%s\" (use %d)\n", | |
807 | net,st->name,st->localcount); | |
9d3a4132 | 808 | free(net); |
794f2398 SE |
809 | for (i=0; i<st->subnets->entries; i++) { |
810 | net=subnet_to_string(st->subnets->list[i]); | |
811 | Message(c,"%s ",net); | |
469fd1d9 SE |
812 | free(net); |
813 | } | |
794f2398 SE |
814 | if (i>0) |
815 | Message(c,"-> host (use %d)\n",st->outcount); | |
9d3a4132 SE |
816 | } |
817 | } | |
818 | ||
d3fe100d SE |
819 | /* ap is a pointer to a member of the routes array */ |
820 | static int netlink_compare_client_priority(const void *ap, const void *bp) | |
70dc107b | 821 | { |
d3fe100d SE |
822 | const struct netlink_client *const*a=ap; |
823 | const struct netlink_client *const*b=bp; | |
70dc107b | 824 | |
d3fe100d SE |
825 | if ((*a)->priority==(*b)->priority) return 0; |
826 | if ((*a)->priority<(*b)->priority) return 1; | |
70dc107b SE |
827 | return -1; |
828 | } | |
829 | ||
830 | static void netlink_phase_hook(void *sst, uint32_t new_phase) | |
831 | { | |
832 | struct netlink *st=sst; | |
833 | struct netlink_client *c; | |
1caa23ff | 834 | int32_t i; |
70dc107b SE |
835 | |
836 | /* All the networks serviced by the various tunnels should now | |
837 | * have been registered. We build a routing table by sorting the | |
d3fe100d | 838 | * clients by priority. */ |
bb9d0561 IJ |
839 | st->routes=safe_malloc_ary(sizeof(*st->routes),st->n_clients, |
840 | "netlink_phase_hook"); | |
70dc107b SE |
841 | /* Fill the table */ |
842 | i=0; | |
59230b9b IJ |
843 | for (c=st->clients; c; c=c->next) { |
844 | assert(i<INT_MAX); | |
d3fe100d | 845 | st->routes[i++]=c; |
59230b9b | 846 | } |
d3fe100d SE |
847 | /* Sort the table in descending order of priority */ |
848 | qsort(st->routes,st->n_clients,sizeof(*st->routes), | |
849 | netlink_compare_client_priority); | |
9d3a4132 | 850 | |
042a8da9 SE |
851 | netlink_dump_routes(st,False); |
852 | } | |
853 | ||
854 | static void netlink_signal_handler(void *sst, int signum) | |
855 | { | |
856 | struct netlink *st=sst; | |
857 | Message(M_INFO,"%s: route dump requested by SIGUSR1\n",st->name); | |
858 | netlink_dump_routes(st,True); | |
70dc107b SE |
859 | } |
860 | ||
1caa23ff | 861 | static void netlink_inst_set_mtu(void *sst, int32_t new_mtu) |
d3fe100d SE |
862 | { |
863 | struct netlink_client *c=sst; | |
864 | ||
865 | c->mtu=new_mtu; | |
866 | } | |
867 | ||
469fd1d9 | 868 | static void netlink_inst_reg(void *sst, netlink_deliver_fn *deliver, |
3abd18e8 | 869 | void *dst) |
469fd1d9 SE |
870 | { |
871 | struct netlink_client *c=sst; | |
469fd1d9 | 872 | |
469fd1d9 SE |
873 | c->deliver=deliver; |
874 | c->dst=dst; | |
875 | } | |
876 | ||
877 | static struct flagstr netlink_option_table[]={ | |
878 | { "soft", OPT_SOFTROUTE }, | |
879 | { "allow-route", OPT_ALLOWROUTE }, | |
880 | { NULL, 0} | |
881 | }; | |
882 | /* This is the routine that gets called when the closure that's | |
883 | returned by an invocation of a netlink device closure (eg. tun, | |
884 | userv-ipif) is invoked. It's used to create routes and pass in | |
885 | information about them; the closure it returns is used by site | |
886 | code. */ | |
887 | static closure_t *netlink_inst_create(struct netlink *st, | |
888 | struct cloc loc, dict_t *dict) | |
889 | { | |
890 | struct netlink_client *c; | |
891 | string_t name; | |
794f2398 | 892 | struct ipset *networks; |
1caa23ff IJ |
893 | uint32_t options,priority; |
894 | int32_t mtu; | |
794f2398 | 895 | list_t *l; |
469fd1d9 SE |
896 | |
897 | name=dict_read_string(dict, "name", True, st->name, loc); | |
898 | ||
794f2398 SE |
899 | l=dict_lookup(dict,"routes"); |
900 | if (!l) | |
901 | cfgfatal(loc,st->name,"required parameter \"routes\" not found\n"); | |
902 | networks=string_list_to_ipset(l,loc,st->name,"routes"); | |
469fd1d9 SE |
903 | options=string_list_to_word(dict_lookup(dict,"options"), |
904 | netlink_option_table,st->name); | |
905 | ||
d3fe100d SE |
906 | priority=dict_read_number(dict,"priority",False,st->name,loc,0); |
907 | mtu=dict_read_number(dict,"mtu",False,st->name,loc,0); | |
908 | ||
909 | if ((options&OPT_SOFTROUTE) && !st->set_routes) { | |
469fd1d9 SE |
910 | cfgfatal(loc,st->name,"this netlink device does not support " |
911 | "soft routes.\n"); | |
912 | return NULL; | |
913 | } | |
914 | ||
915 | if (options&OPT_SOFTROUTE) { | |
916 | /* XXX for now we assume that soft routes require root privilege; | |
917 | this may not always be true. The device driver can tell us. */ | |
918 | require_root_privileges=True; | |
919 | require_root_privileges_explanation="netlink: soft routes"; | |
920 | if (st->ptp) { | |
921 | cfgfatal(loc,st->name,"point-to-point netlinks do not support " | |
922 | "soft routes.\n"); | |
923 | return NULL; | |
924 | } | |
925 | } | |
926 | ||
794f2398 SE |
927 | /* Check that nets are a subset of st->remote_networks; |
928 | refuse to register if they are not. */ | |
929 | if (!ipset_is_subset(st->remote_networks,networks)) { | |
930 | cfgfatal(loc,st->name,"routes are not allowed\n"); | |
469fd1d9 SE |
931 | return NULL; |
932 | } | |
933 | ||
934 | c=safe_malloc(sizeof(*c),"netlink_inst_create"); | |
935 | c->cl.description=name; | |
936 | c->cl.type=CL_NETLINK; | |
937 | c->cl.apply=NULL; | |
938 | c->cl.interface=&c->ops; | |
939 | c->ops.st=c; | |
940 | c->ops.reg=netlink_inst_reg; | |
941 | c->ops.deliver=netlink_inst_incoming; | |
942 | c->ops.set_quality=netlink_set_quality; | |
d3fe100d | 943 | c->ops.set_mtu=netlink_inst_set_mtu; |
469fd1d9 SE |
944 | c->nst=st; |
945 | ||
946 | c->networks=networks; | |
794f2398 | 947 | c->subnets=ipset_to_subnet_list(networks); |
d3fe100d | 948 | c->priority=priority; |
469fd1d9 SE |
949 | c->deliver=NULL; |
950 | c->dst=NULL; | |
951 | c->name=name; | |
f208b9a9 | 952 | c->link_quality=LINK_QUALITY_UNUSED; |
d3fe100d SE |
953 | c->mtu=mtu?mtu:st->mtu; |
954 | c->options=options; | |
955 | c->outcount=0; | |
956 | c->up=False; | |
957 | c->kup=False; | |
469fd1d9 SE |
958 | c->next=st->clients; |
959 | st->clients=c; | |
59230b9b | 960 | assert(st->n_clients < INT_MAX); |
d3fe100d | 961 | st->n_clients++; |
469fd1d9 SE |
962 | |
963 | return &c->cl; | |
964 | } | |
965 | ||
966 | static list_t *netlink_inst_apply(closure_t *self, struct cloc loc, | |
967 | dict_t *context, list_t *args) | |
968 | { | |
969 | struct netlink *st=self->interface; | |
970 | ||
971 | dict_t *dict; | |
972 | item_t *item; | |
973 | closure_t *cl; | |
974 | ||
469fd1d9 SE |
975 | item=list_elem(args,0); |
976 | if (!item || item->type!=t_dict) { | |
977 | cfgfatal(loc,st->name,"must have a dictionary argument\n"); | |
978 | } | |
979 | dict=item->data.dict; | |
980 | ||
981 | cl=netlink_inst_create(st,loc,dict); | |
982 | ||
983 | return new_closure(cl); | |
984 | } | |
985 | ||
9d3a4132 SE |
986 | netlink_deliver_fn *netlink_init(struct netlink *st, |
987 | void *dst, struct cloc loc, | |
fe5e9cc4 | 988 | dict_t *dict, cstring_t description, |
d3fe100d | 989 | netlink_route_fn *set_routes, |
9d3a4132 | 990 | netlink_deliver_fn *to_host) |
4efd681a | 991 | { |
c6f79b17 | 992 | item_t *sa, *ptpa; |
794f2398 | 993 | list_t *l; |
c6f79b17 | 994 | |
4efd681a SE |
995 | st->dst=dst; |
996 | st->cl.description=description; | |
469fd1d9 SE |
997 | st->cl.type=CL_PURE; |
998 | st->cl.apply=netlink_inst_apply; | |
999 | st->cl.interface=st; | |
4efd681a | 1000 | st->clients=NULL; |
d3fe100d SE |
1001 | st->routes=NULL; |
1002 | st->n_clients=0; | |
1003 | st->set_routes=set_routes; | |
4efd681a SE |
1004 | st->deliver_to_host=to_host; |
1005 | ||
794f2398 | 1006 | st->name=dict_read_string(dict,"name",False,description,loc); |
4efd681a | 1007 | if (!st->name) st->name=description; |
794f2398 SE |
1008 | l=dict_lookup(dict,"networks"); |
1009 | if (l) | |
1010 | st->networks=string_list_to_ipset(l,loc,st->name,"networks"); | |
1011 | else { | |
4f5e39ec SE |
1012 | struct ipset *empty; |
1013 | empty=ipset_new(); | |
1014 | st->networks=ipset_complement(empty); | |
1015 | ipset_free(empty); | |
794f2398 SE |
1016 | } |
1017 | l=dict_lookup(dict,"remote-networks"); | |
1018 | if (l) { | |
1019 | st->remote_networks=string_list_to_ipset(l,loc,st->name, | |
1020 | "remote-networks"); | |
1021 | } else { | |
1022 | struct ipset *empty; | |
1023 | empty=ipset_new(); | |
1024 | st->remote_networks=ipset_complement(empty); | |
1025 | ipset_free(empty); | |
1026 | } | |
1027 | ||
c6f79b17 | 1028 | sa=dict_find_item(dict,"secnet-address",False,"netlink",loc); |
469fd1d9 | 1029 | ptpa=dict_find_item(dict,"ptp-address",False,"netlink",loc); |
c6f79b17 SE |
1030 | if (sa && ptpa) { |
1031 | cfgfatal(loc,st->name,"you may not specify secnet-address and " | |
1032 | "ptp-address in the same netlink device\n"); | |
1033 | } | |
1034 | if (!(sa || ptpa)) { | |
1035 | cfgfatal(loc,st->name,"you must specify secnet-address or " | |
1036 | "ptp-address for this netlink device\n"); | |
1037 | } | |
1038 | if (sa) { | |
794f2398 | 1039 | st->secnet_address=string_item_to_ipaddr(sa,"netlink"); |
c6f79b17 SE |
1040 | st->ptp=False; |
1041 | } else { | |
794f2398 | 1042 | st->secnet_address=string_item_to_ipaddr(ptpa,"netlink"); |
c6f79b17 SE |
1043 | st->ptp=True; |
1044 | } | |
d3fe100d SE |
1045 | /* To be strictly correct we could subtract secnet_address from |
1046 | networks here. It shouldn't make any practical difference, | |
794f2398 SE |
1047 | though, and will make the route dump look complicated... */ |
1048 | st->subnets=ipset_to_subnet_list(st->networks); | |
4efd681a SE |
1049 | st->mtu=dict_read_number(dict, "mtu", False, "netlink", loc, DEFAULT_MTU); |
1050 | buffer_new(&st->icmp,ICMP_BUFSIZE); | |
469fd1d9 SE |
1051 | st->outcount=0; |
1052 | st->localcount=0; | |
70dc107b SE |
1053 | |
1054 | add_hook(PHASE_SETUP,netlink_phase_hook,st); | |
042a8da9 | 1055 | request_signal_notification(SIGUSR1, netlink_signal_handler, st); |
4efd681a | 1056 | |
469fd1d9 SE |
1057 | /* If we're point-to-point then we return a CL_NETLINK directly, |
1058 | rather than a CL_NETLINK_OLD or pure closure (depending on | |
1059 | compatibility). This CL_NETLINK is for our one and only | |
1060 | client. Our cl.apply function is NULL. */ | |
1061 | if (st->ptp) { | |
1062 | closure_t *cl; | |
1063 | cl=netlink_inst_create(st,loc,dict); | |
1064 | st->cl=*cl; | |
1065 | } | |
1066 | return netlink_dev_incoming; | |
2fe58dfd SE |
1067 | } |
1068 | ||
9d3a4132 | 1069 | /* No connection to the kernel at all... */ |
2fe58dfd | 1070 | |
9d3a4132 | 1071 | struct null { |
4efd681a | 1072 | struct netlink nl; |
4efd681a | 1073 | }; |
2fe58dfd | 1074 | |
d3fe100d | 1075 | static bool_t null_set_route(void *sst, struct netlink_client *routes) |
4efd681a | 1076 | { |
9d3a4132 | 1077 | struct null *st=sst; |
d3fe100d SE |
1078 | |
1079 | if (routes->up!=routes->kup) { | |
1080 | Message(M_INFO,"%s: setting routes for tunnel %s to state %s\n", | |
1081 | st->nl.name,routes->name, | |
1082 | routes->up?"up":"down"); | |
1083 | routes->kup=routes->up; | |
9d3a4132 | 1084 | return True; |
2fe58dfd | 1085 | } |
9d3a4132 | 1086 | return False; |
2fe58dfd | 1087 | } |
9d3a4132 | 1088 | |
469fd1d9 | 1089 | static void null_deliver(void *sst, struct buffer_if *buf) |
2fe58dfd SE |
1090 | { |
1091 | return; | |
1092 | } | |
1093 | ||
1094 | static list_t *null_apply(closure_t *self, struct cloc loc, dict_t *context, | |
1095 | list_t *args) | |
1096 | { | |
1097 | struct null *st; | |
4efd681a SE |
1098 | item_t *item; |
1099 | dict_t *dict; | |
2fe58dfd | 1100 | |
4efd681a | 1101 | st=safe_malloc(sizeof(*st),"null_apply"); |
2fe58dfd | 1102 | |
4efd681a SE |
1103 | item=list_elem(args,0); |
1104 | if (!item || item->type!=t_dict) | |
1105 | cfgfatal(loc,"null-netlink","parameter must be a dictionary\n"); | |
1106 | ||
1107 | dict=item->data.dict; | |
1108 | ||
9d3a4132 SE |
1109 | netlink_init(&st->nl,st,loc,dict,"null-netlink",null_set_route, |
1110 | null_deliver); | |
4efd681a SE |
1111 | |
1112 | return new_closure(&st->nl.cl); | |
2fe58dfd SE |
1113 | } |
1114 | ||
2fe58dfd SE |
1115 | void netlink_module(dict_t *dict) |
1116 | { | |
4efd681a | 1117 | add_closure(dict,"null-netlink",null_apply); |
2fe58dfd | 1118 | } |