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 | |
a0b107b8 IJ |
109 | #ifdef NETLINK_DEBUG |
110 | #define MDEBUG(...) Message(M_DEBUG, __VA_ARGS__) | |
111 | #else /* !NETLINK_DEBUG */ | |
112 | #define MDEBUG(...) ((void)0) | |
113 | #endif /* !NETLINK_DEBUG */ | |
114 | ||
ff05a229 SE |
115 | #define ICMP_TYPE_ECHO_REPLY 0 |
116 | ||
117 | #define ICMP_TYPE_UNREACHABLE 3 | |
118 | #define ICMP_CODE_NET_UNREACHABLE 0 | |
119 | #define ICMP_CODE_PROTOCOL_UNREACHABLE 2 | |
120 | #define ICMP_CODE_FRAGMENTATION_REQUIRED 4 | |
121 | #define ICMP_CODE_NET_PROHIBITED 13 | |
122 | ||
123 | #define ICMP_TYPE_ECHO_REQUEST 8 | |
124 | ||
125 | #define ICMP_TYPE_TIME_EXCEEDED 11 | |
126 | #define ICMP_CODE_TTL_EXCEEDED 0 | |
127 | ||
4efd681a | 128 | /* Generic IP checksum routine */ |
211cd627 | 129 | static inline uint16_t ip_csum(const uint8_t *iph,int32_t count) |
2fe58dfd | 130 | { |
4efd681a SE |
131 | register uint32_t sum=0; |
132 | ||
133 | while (count>1) { | |
134 | sum+=ntohs(*(uint16_t *)iph); | |
135 | iph+=2; | |
136 | count-=2; | |
137 | } | |
138 | if(count>0) | |
139 | sum+=*(uint8_t *)iph; | |
140 | while (sum>>16) | |
141 | sum=(sum&0xffff)+(sum>>16); | |
142 | return htons(~sum); | |
2fe58dfd SE |
143 | } |
144 | ||
4efd681a SE |
145 | #ifdef i386 |
146 | /* | |
147 | * This is a version of ip_compute_csum() optimized for IP headers, | |
148 | * which always checksum on 4 octet boundaries. | |
149 | * | |
150 | * By Jorge Cwik <jorge@laser.satlink.net>, adapted for linux by | |
151 | * Arnt Gulbrandsen. | |
152 | */ | |
211cd627 | 153 | static inline uint16_t ip_fast_csum(const uint8_t *iph, int32_t ihl) { |
4efd681a SE |
154 | uint32_t sum; |
155 | ||
20d324b6 SE |
156 | __asm__ __volatile__( |
157 | "movl (%1), %0 ;\n" | |
158 | "subl $4, %2 ;\n" | |
159 | "jbe 2f ;\n" | |
160 | "addl 4(%1), %0 ;\n" | |
161 | "adcl 8(%1), %0 ;\n" | |
162 | "adcl 12(%1), %0 ;\n" | |
163 | "1: adcl 16(%1), %0 ;\n" | |
164 | "lea 4(%1), %1 ;\n" | |
165 | "decl %2 ;\n" | |
166 | "jne 1b ;\n" | |
167 | "adcl $0, %0 ;\n" | |
168 | "movl %0, %2 ;\n" | |
169 | "shrl $16, %0 ;\n" | |
170 | "addw %w2, %w0 ;\n" | |
171 | "adcl $0, %0 ;\n" | |
172 | "notl %0 ;\n" | |
173 | "2: ;\n" | |
4efd681a SE |
174 | /* Since the input registers which are loaded with iph and ipl |
175 | are modified, we must also specify them as outputs, or gcc | |
176 | will assume they contain their original values. */ | |
177 | : "=r" (sum), "=r" (iph), "=r" (ihl) | |
20d324b6 SE |
178 | : "1" (iph), "2" (ihl) |
179 | : "memory"); | |
4efd681a SE |
180 | return sum; |
181 | } | |
182 | #else | |
1caa23ff | 183 | static inline uint16_t ip_fast_csum(uint8_t *iph, int32_t ihl) |
2fe58dfd | 184 | { |
1caa23ff | 185 | assert(ihl < INT_MAX/4); |
4efd681a SE |
186 | return ip_csum(iph,ihl*4); |
187 | } | |
188 | #endif | |
189 | ||
190 | struct iphdr { | |
191 | #if defined (WORDS_BIGENDIAN) | |
192 | uint8_t version:4, | |
193 | ihl:4; | |
194 | #else | |
195 | uint8_t ihl:4, | |
196 | version:4; | |
197 | #endif | |
198 | uint8_t tos; | |
199 | uint16_t tot_len; | |
200 | uint16_t id; | |
a6768d7c | 201 | uint16_t frag; |
eff13010 IJ |
202 | #define IPHDR_FRAG_OFF ((uint16_t)0x1fff) |
203 | #define IPHDR_FRAG_MORE ((uint16_t)0x2000) | |
204 | #define IPHDR_FRAG_DONT ((uint16_t)0x4000) | |
205 | /* reserved 0x8000 */ | |
4efd681a SE |
206 | uint8_t ttl; |
207 | uint8_t protocol; | |
208 | uint16_t check; | |
209 | uint32_t saddr; | |
210 | uint32_t daddr; | |
211 | /* The options start here. */ | |
212 | }; | |
213 | ||
214 | struct icmphdr { | |
215 | struct iphdr iph; | |
216 | uint8_t type; | |
217 | uint8_t code; | |
218 | uint16_t check; | |
cfd79482 | 219 | union icmpinfofield { |
4efd681a SE |
220 | uint32_t unused; |
221 | struct { | |
222 | uint8_t pointer; | |
223 | uint8_t unused1; | |
224 | uint16_t unused2; | |
225 | } pprob; | |
226 | uint32_t gwaddr; | |
227 | struct { | |
228 | uint16_t id; | |
229 | uint16_t seq; | |
230 | } echo; | |
231 | } d; | |
232 | }; | |
cfd79482 IJ |
233 | |
234 | static const union icmpinfofield icmp_noinfo; | |
4efd681a | 235 | |
70dc107b SE |
236 | static void netlink_packet_deliver(struct netlink *st, |
237 | struct netlink_client *client, | |
238 | struct buffer_if *buf); | |
4efd681a | 239 | |
ff05a229 SE |
240 | /* XXX RFC1812 4.3.2.5: |
241 | All other ICMP error messages (Destination Unreachable, | |
242 | Redirect, Time Exceeded, and Parameter Problem) SHOULD have their | |
243 | precedence value set to 6 (INTERNETWORK CONTROL) or 7 (NETWORK | |
244 | CONTROL). The IP Precedence value for these error messages MAY be | |
245 | settable. | |
246 | */ | |
4efd681a SE |
247 | static struct icmphdr *netlink_icmp_tmpl(struct netlink *st, |
248 | uint32_t dest,uint16_t len) | |
249 | { | |
250 | struct icmphdr *h; | |
251 | ||
252 | BUF_ALLOC(&st->icmp,"netlink_icmp_tmpl"); | |
3abd18e8 | 253 | buffer_init(&st->icmp,calculate_max_start_pad()); |
4efd681a SE |
254 | h=buf_append(&st->icmp,sizeof(*h)); |
255 | ||
256 | h->iph.version=4; | |
257 | h->iph.ihl=5; | |
258 | h->iph.tos=0; | |
259 | h->iph.tot_len=htons(len+(h->iph.ihl*4)+8); | |
260 | h->iph.id=0; | |
a6768d7c | 261 | h->iph.frag=0; |
ff05a229 | 262 | h->iph.ttl=255; /* XXX should be configurable */ |
4efd681a SE |
263 | h->iph.protocol=1; |
264 | h->iph.saddr=htonl(st->secnet_address); | |
265 | h->iph.daddr=htonl(dest); | |
266 | h->iph.check=0; | |
267 | h->iph.check=ip_fast_csum((uint8_t *)&h->iph,h->iph.ihl); | |
268 | h->check=0; | |
269 | h->d.unused=0; | |
270 | ||
271 | return h; | |
272 | } | |
273 | ||
274 | /* Fill in the ICMP checksum field correctly */ | |
275 | static void netlink_icmp_csum(struct icmphdr *h) | |
276 | { | |
1caa23ff | 277 | int32_t len; |
4efd681a SE |
278 | |
279 | len=ntohs(h->iph.tot_len)-(4*h->iph.ihl); | |
280 | h->check=0; | |
281 | h->check=ip_csum(&h->type,len); | |
282 | } | |
283 | ||
284 | /* RFC1122: | |
285 | * An ICMP error message MUST NOT be sent as the result of | |
286 | * receiving: | |
287 | * | |
288 | * * an ICMP error message, or | |
289 | * | |
290 | * * a datagram destined to an IP broadcast or IP multicast | |
291 | * address, or | |
292 | * | |
293 | * * a datagram sent as a link-layer broadcast, or | |
294 | * | |
295 | * * a non-initial fragment, or | |
296 | * | |
297 | * * a datagram whose source address does not define a single | |
298 | * host -- e.g., a zero address, a loopback address, a | |
299 | * broadcast address, a multicast address, or a Class E | |
300 | * address. | |
301 | */ | |
302 | static bool_t netlink_icmp_may_reply(struct buffer_if *buf) | |
303 | { | |
304 | struct iphdr *iph; | |
8dea8d37 | 305 | struct icmphdr *icmph; |
4efd681a SE |
306 | uint32_t source; |
307 | ||
975820aa | 308 | if (buf->size < (int)sizeof(struct icmphdr)) return False; |
4efd681a | 309 | iph=(struct iphdr *)buf->start; |
8dea8d37 SE |
310 | icmph=(struct icmphdr *)buf->start; |
311 | if (iph->protocol==1) { | |
312 | switch(icmph->type) { | |
686b7f1d IJ |
313 | /* Based on http://www.iana.org/assignments/icmp-parameters/icmp-parameters.xhtml#icmp-parameters-types |
314 | * as retrieved Thu, 20 Mar 2014 00:16:44 +0000. | |
315 | * Deprecated, reserved, unassigned and experimental | |
316 | * options are treated as not safe to reply to. | |
317 | */ | |
318 | case 0: /* Echo Reply */ | |
319 | case 8: /* Echo */ | |
320 | case 13: /* Timestamp */ | |
321 | case 14: /* Timestamp Reply */ | |
322 | return True; | |
323 | default: | |
8dea8d37 SE |
324 | return False; |
325 | } | |
326 | } | |
4efd681a | 327 | /* How do we spot broadcast destination addresses? */ |
a6768d7c | 328 | if (ntohs(iph->frag)&IPHDR_FRAG_OFF) return False; |
4efd681a SE |
329 | source=ntohl(iph->saddr); |
330 | if (source==0) return False; | |
331 | if ((source&0xff000000)==0x7f000000) return False; | |
332 | /* How do we spot broadcast source addresses? */ | |
333 | if ((source&0xf0000000)==0xe0000000) return False; /* Multicast */ | |
334 | if ((source&0xf0000000)==0xf0000000) return False; /* Class E */ | |
335 | return True; | |
336 | } | |
337 | ||
338 | /* How much of the original IP packet do we include in its ICMP | |
339 | response? The header plus up to 64 bits. */ | |
ff05a229 SE |
340 | |
341 | /* XXX TODO RFC1812: | |
342 | 4.3.2.3 Original Message Header | |
343 | ||
344 | Historically, every ICMP error message has included the Internet | |
345 | header and at least the first 8 data bytes of the datagram that | |
346 | triggered the error. This is no longer adequate, due to the use of | |
347 | IP-in-IP tunneling and other technologies. Therefore, the ICMP | |
348 | datagram SHOULD contain as much of the original datagram as possible | |
349 | without the length of the ICMP datagram exceeding 576 bytes. The | |
350 | returned IP header (and user data) MUST be identical to that which | |
351 | was received, except that the router is not required to undo any | |
352 | modifications to the IP header that are normally performed in | |
353 | forwarding that were performed before the error was detected (e.g., | |
354 | decrementing the TTL, or updating options). Note that the | |
355 | requirements of Section [4.3.3.5] supersede this requirement in some | |
356 | cases (i.e., for a Parameter Problem message, if the problem is in a | |
357 | modified field, the router must undo the modification). See Section | |
358 | [4.3.3.5]). | |
359 | */ | |
4efd681a SE |
360 | static uint16_t netlink_icmp_reply_len(struct buffer_if *buf) |
361 | { | |
975820aa | 362 | if (buf->size < (int)sizeof(struct iphdr)) return 0; |
4efd681a SE |
363 | struct iphdr *iph=(struct iphdr *)buf->start; |
364 | uint16_t hlen,plen; | |
365 | ||
366 | hlen=iph->ihl*4; | |
367 | /* We include the first 8 bytes of the packet data, provided they exist */ | |
368 | hlen+=8; | |
369 | plen=ntohs(iph->tot_len); | |
370 | return (hlen>plen?plen:hlen); | |
371 | } | |
372 | ||
70dc107b SE |
373 | /* client indicates where the packet we're constructing a response to |
374 | comes from. NULL indicates the host. */ | |
4efd681a | 375 | static void netlink_icmp_simple(struct netlink *st, struct buffer_if *buf, |
cfd79482 IJ |
376 | uint8_t type, uint8_t code, |
377 | union icmpinfofield info) | |
4efd681a | 378 | { |
4efd681a SE |
379 | struct icmphdr *h; |
380 | uint16_t len; | |
381 | ||
382 | if (netlink_icmp_may_reply(buf)) { | |
975820aa | 383 | struct iphdr *iph=(struct iphdr *)buf->start; |
4efd681a SE |
384 | len=netlink_icmp_reply_len(buf); |
385 | h=netlink_icmp_tmpl(st,ntohl(iph->saddr),len); | |
cfd79482 | 386 | h->type=type; h->code=code; h->d=info; |
4efd681a SE |
387 | memcpy(buf_append(&st->icmp,len),buf->start,len); |
388 | netlink_icmp_csum(h); | |
70dc107b | 389 | netlink_packet_deliver(st,NULL,&st->icmp); |
4efd681a SE |
390 | BUF_ASSERT_FREE(&st->icmp); |
391 | } | |
392 | } | |
393 | ||
394 | /* | |
395 | * RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the | |
396 | * checksum. | |
ff05a229 | 397 | * RFC1812: 4.2.2.5 MUST discard messages containing invalid checksums. |
4efd681a SE |
398 | * |
399 | * Is the datagram acceptable? | |
400 | * | |
401 | * 1. Length at least the size of an ip header | |
402 | * 2. Version of 4 | |
403 | * 3. Checksums correctly. | |
404 | * 4. Doesn't have a bogus length | |
405 | */ | |
d714da29 IJ |
406 | static bool_t netlink_check(struct netlink *st, struct buffer_if *buf, |
407 | char *errmsgbuf, int errmsgbuflen) | |
4efd681a | 408 | { |
d714da29 IJ |
409 | #define BAD(...) do{ \ |
410 | snprintf(errmsgbuf,errmsgbuflen,__VA_ARGS__); \ | |
411 | return False; \ | |
412 | }while(0) | |
413 | ||
975820aa | 414 | if (buf->size < (int)sizeof(struct iphdr)) BAD("len %"PRIu32"",buf->size); |
4efd681a | 415 | struct iphdr *iph=(struct iphdr *)buf->start; |
1caa23ff | 416 | int32_t len; |
4efd681a | 417 | |
d714da29 IJ |
418 | if (iph->ihl < 5) BAD("ihl %u",iph->ihl); |
419 | if (iph->version != 4) BAD("version %u",iph->version); | |
420 | if (buf->size < iph->ihl*4) BAD("size %"PRId32"<%u*4",buf->size,iph->ihl); | |
421 | if (ip_fast_csum((uint8_t *)iph, iph->ihl)!=0) BAD("csum"); | |
4efd681a SE |
422 | len=ntohs(iph->tot_len); |
423 | /* There should be no padding */ | |
d714da29 IJ |
424 | if (buf->size!=len) BAD("len %"PRId32"!=%"PRId32,buf->size,len); |
425 | if (len<(iph->ihl<<2)) BAD("len %"PRId32"<(%u<<2)",len,iph->ihl); | |
4efd681a SE |
426 | /* XXX check that there's no source route specified */ |
427 | return True; | |
d714da29 IJ |
428 | |
429 | #undef BAD | |
4efd681a SE |
430 | } |
431 | ||
7b6abafa | 432 | /* Deliver a packet _to_ client; used after we have decided |
55bc97e6 IJ |
433 | * what to do with it (and just to check that the client has |
434 | * actually registered a delivery function with us). */ | |
7b6abafa IJ |
435 | static void netlink_client_deliver(struct netlink *st, |
436 | struct netlink_client *client, | |
437 | uint32_t source, uint32_t dest, | |
438 | struct buffer_if *buf) | |
439 | { | |
55bc97e6 IJ |
440 | if (!client->deliver) { |
441 | string_t s,d; | |
442 | s=ipaddr_to_string(source); | |
443 | d=ipaddr_to_string(dest); | |
444 | Message(M_ERR,"%s: dropping %s->%s, client not registered\n", | |
445 | st->name,s,d); | |
446 | free(s); free(d); | |
447 | BUF_FREE(buf); | |
448 | return; | |
449 | } | |
7b6abafa IJ |
450 | client->deliver(client->dst, buf); |
451 | client->outcount++; | |
452 | } | |
453 | ||
f928f069 IJ |
454 | /* Deliver a packet to the host; used after we have decided that that |
455 | * is what to do with it. */ | |
456 | static void netlink_host_deliver(struct netlink *st, | |
457 | uint32_t source, uint32_t dest, | |
458 | struct buffer_if *buf) | |
459 | { | |
460 | st->deliver_to_host(st->dst,buf); | |
461 | st->outcount++; | |
462 | } | |
463 | ||
469fd1d9 | 464 | /* Deliver a packet. "client" is the _origin_ of the packet, not its |
d3fe100d SE |
465 | destination, and is NULL for packets from the host and packets |
466 | generated internally in secnet. */ | |
70dc107b SE |
467 | static void netlink_packet_deliver(struct netlink *st, |
468 | struct netlink_client *client, | |
469 | struct buffer_if *buf) | |
4efd681a | 470 | { |
975820aa IJ |
471 | if (buf->size < (int)sizeof(struct iphdr)) { |
472 | Message(M_ERR,"%s: trying to deliver a too-short packet" | |
473 | " from %s!\n",st->name, client?client->name:"(local)"); | |
474 | BUF_FREE(buf); | |
475 | return; | |
476 | } | |
477 | ||
4efd681a SE |
478 | struct iphdr *iph=(struct iphdr *)buf->start; |
479 | uint32_t dest=ntohl(iph->daddr); | |
70dc107b SE |
480 | uint32_t source=ntohl(iph->saddr); |
481 | uint32_t best_quality; | |
469fd1d9 SE |
482 | bool_t allow_route=False; |
483 | bool_t found_allowed=False; | |
70dc107b SE |
484 | int best_match; |
485 | int i; | |
2fe58dfd | 486 | |
4efd681a | 487 | BUF_ASSERT_USED(buf); |
2fe58dfd | 488 | |
4efd681a | 489 | if (dest==st->secnet_address) { |
4f5e39ec | 490 | Message(M_ERR,"%s: trying to deliver a packet to myself!\n",st->name); |
4efd681a | 491 | BUF_FREE(buf); |
2fe58dfd SE |
492 | return; |
493 | } | |
4efd681a | 494 | |
d3fe100d | 495 | /* Packets from the host (client==NULL) may always be routed. Packets |
469fd1d9 SE |
496 | from clients with the allow_route option will also be routed. */ |
497 | if (!client || (client && (client->options & OPT_ALLOWROUTE))) | |
498 | allow_route=True; | |
499 | ||
500 | /* If !allow_route, we check the routing table anyway, and if | |
501 | there's a suitable route with OPT_ALLOWROUTE set we use it. If | |
502 | there's a suitable route, but none with OPT_ALLOWROUTE set then | |
503 | we generate ICMP 'communication with destination network | |
504 | administratively prohibited'. */ | |
505 | ||
506 | best_quality=0; | |
507 | best_match=-1; | |
d3fe100d SE |
508 | for (i=0; i<st->n_clients; i++) { |
509 | if (st->routes[i]->up && | |
510 | ipset_contains_addr(st->routes[i]->networks,dest)) { | |
469fd1d9 SE |
511 | /* It's an available route to the correct destination. But is |
512 | it better than the one we already have? */ | |
513 | ||
514 | /* If we have already found an allowed route then we don't | |
515 | bother looking at routes we're not allowed to use. If | |
516 | we don't yet have an allowed route we'll consider any. */ | |
517 | if (!allow_route && found_allowed) { | |
d3fe100d | 518 | if (!(st->routes[i]->options&OPT_ALLOWROUTE)) continue; |
70dc107b | 519 | } |
469fd1d9 | 520 | |
d3fe100d | 521 | if (st->routes[i]->link_quality>best_quality |
469fd1d9 | 522 | || best_quality==0) { |
d3fe100d | 523 | best_quality=st->routes[i]->link_quality; |
469fd1d9 | 524 | best_match=i; |
d3fe100d | 525 | if (st->routes[i]->options&OPT_ALLOWROUTE) |
469fd1d9 SE |
526 | found_allowed=True; |
527 | /* If quality isn't perfect we may wish to | |
528 | consider kicking the tunnel with a 0-length | |
529 | packet to prompt it to perform a key setup. | |
530 | Then it'll eventually decide it's up or | |
531 | down. */ | |
532 | /* If quality is perfect and we're allowed to use the | |
533 | route we don't need to search any more. */ | |
534 | if (best_quality>=MAXIMUM_LINK_QUALITY && | |
535 | (allow_route || found_allowed)) break; | |
4efd681a | 536 | } |
70dc107b | 537 | } |
469fd1d9 SE |
538 | } |
539 | if (best_match==-1) { | |
540 | /* The packet's not going down a tunnel. It might (ought to) | |
541 | be for the host. */ | |
794f2398 | 542 | if (ipset_contains_addr(st->networks,dest)) { |
f928f069 | 543 | netlink_host_deliver(st,source,dest,buf); |
70dc107b SE |
544 | BUF_ASSERT_FREE(buf); |
545 | } else { | |
469fd1d9 SE |
546 | string_t s,d; |
547 | s=ipaddr_to_string(source); | |
548 | d=ipaddr_to_string(dest); | |
ff05a229 | 549 | Message(M_DEBUG,"%s: don't know where to deliver packet " |
469fd1d9 SE |
550 | "(s=%s, d=%s)\n", st->name, s, d); |
551 | free(s); free(d); | |
311e11e4 | 552 | netlink_icmp_simple(st,buf,ICMP_TYPE_UNREACHABLE, |
cfd79482 | 553 | ICMP_CODE_NET_UNREACHABLE, icmp_noinfo); |
70dc107b | 554 | BUF_FREE(buf); |
2fe58dfd | 555 | } |
469fd1d9 SE |
556 | } else { |
557 | if (!allow_route && | |
d3fe100d | 558 | !(st->routes[best_match]->options&OPT_ALLOWROUTE)) { |
469fd1d9 SE |
559 | string_t s,d; |
560 | s=ipaddr_to_string(source); | |
561 | d=ipaddr_to_string(dest); | |
562 | /* We have a usable route but aren't allowed to use it. | |
563 | Generate ICMP destination unreachable: communication | |
564 | with destination network administratively prohibited */ | |
565 | Message(M_NOTICE,"%s: denied forwarding for packet (s=%s, d=%s)\n", | |
566 | st->name,s,d); | |
567 | free(s); free(d); | |
568 | ||
311e11e4 | 569 | netlink_icmp_simple(st,buf,ICMP_TYPE_UNREACHABLE, |
cfd79482 | 570 | ICMP_CODE_NET_PROHIBITED, icmp_noinfo); |
469fd1d9 | 571 | BUF_FREE(buf); |
469fd1d9 | 572 | } else { |
ea7ec970 SE |
573 | if (best_quality>0) { |
574 | /* XXX Fragment if required */ | |
7b6abafa IJ |
575 | netlink_client_deliver(st,st->routes[best_match], |
576 | source,dest,buf); | |
ea7ec970 SE |
577 | BUF_ASSERT_FREE(buf); |
578 | } else { | |
579 | /* Generate ICMP destination unreachable */ | |
311e11e4 | 580 | netlink_icmp_simple(st,buf, |
cfd79482 IJ |
581 | ICMP_TYPE_UNREACHABLE, |
582 | ICMP_CODE_NET_UNREACHABLE, | |
583 | icmp_noinfo); | |
ea7ec970 SE |
584 | BUF_FREE(buf); |
585 | } | |
469fd1d9 | 586 | } |
2fe58dfd | 587 | } |
70dc107b | 588 | BUF_ASSERT_FREE(buf); |
4efd681a SE |
589 | } |
590 | ||
70dc107b SE |
591 | static void netlink_packet_forward(struct netlink *st, |
592 | struct netlink_client *client, | |
593 | struct buffer_if *buf) | |
4efd681a | 594 | { |
975820aa | 595 | if (buf->size < (int)sizeof(struct iphdr)) return; |
4efd681a SE |
596 | struct iphdr *iph=(struct iphdr *)buf->start; |
597 | ||
598 | BUF_ASSERT_USED(buf); | |
599 | ||
600 | /* Packet has already been checked */ | |
601 | if (iph->ttl<=1) { | |
602 | /* Generate ICMP time exceeded */ | |
311e11e4 | 603 | netlink_icmp_simple(st,buf,ICMP_TYPE_TIME_EXCEEDED, |
cfd79482 | 604 | ICMP_CODE_TTL_EXCEEDED,icmp_noinfo); |
4efd681a SE |
605 | BUF_FREE(buf); |
606 | return; | |
607 | } | |
608 | iph->ttl--; | |
609 | iph->check=0; | |
610 | iph->check=ip_fast_csum((uint8_t *)iph,iph->ihl); | |
611 | ||
70dc107b | 612 | netlink_packet_deliver(st,client,buf); |
4efd681a SE |
613 | BUF_ASSERT_FREE(buf); |
614 | } | |
615 | ||
9d3a4132 | 616 | /* Deal with packets addressed explicitly to us */ |
70dc107b SE |
617 | static void netlink_packet_local(struct netlink *st, |
618 | struct netlink_client *client, | |
619 | struct buffer_if *buf) | |
4efd681a SE |
620 | { |
621 | struct icmphdr *h; | |
622 | ||
469fd1d9 SE |
623 | st->localcount++; |
624 | ||
975820aa IJ |
625 | if (buf->size < (int)sizeof(struct icmphdr)) { |
626 | Message(M_WARNING,"%s: short packet addressed to secnet; " | |
627 | "ignoring it\n",st->name); | |
628 | BUF_FREE(buf); | |
629 | return; | |
630 | } | |
4efd681a SE |
631 | h=(struct icmphdr *)buf->start; |
632 | ||
a6768d7c | 633 | if ((ntohs(h->iph.frag)&(IPHDR_FRAG_OFF|IPHDR_FRAG_MORE))!=0) { |
9d3a4132 SE |
634 | Message(M_WARNING,"%s: fragmented packet addressed to secnet; " |
635 | "ignoring it\n",st->name); | |
4efd681a SE |
636 | BUF_FREE(buf); |
637 | return; | |
638 | } | |
639 | ||
640 | if (h->iph.protocol==1) { | |
641 | /* It's ICMP */ | |
ff05a229 | 642 | if (h->type==ICMP_TYPE_ECHO_REQUEST && h->code==0) { |
4efd681a SE |
643 | /* ICMP echo-request. Special case: we re-use the buffer |
644 | to construct the reply. */ | |
ff05a229 | 645 | h->type=ICMP_TYPE_ECHO_REPLY; |
4efd681a SE |
646 | h->iph.daddr=h->iph.saddr; |
647 | h->iph.saddr=htonl(st->secnet_address); | |
ff05a229 | 648 | h->iph.ttl=255; |
4efd681a SE |
649 | h->iph.check=0; |
650 | h->iph.check=ip_fast_csum((uint8_t *)h,h->iph.ihl); | |
651 | netlink_icmp_csum(h); | |
70dc107b | 652 | netlink_packet_deliver(st,NULL,buf); |
4efd681a SE |
653 | return; |
654 | } | |
655 | Message(M_WARNING,"%s: unknown incoming ICMP\n",st->name); | |
656 | } else { | |
657 | /* Send ICMP protocol unreachable */ | |
311e11e4 | 658 | netlink_icmp_simple(st,buf,ICMP_TYPE_UNREACHABLE, |
cfd79482 | 659 | ICMP_CODE_PROTOCOL_UNREACHABLE,icmp_noinfo); |
4efd681a SE |
660 | BUF_FREE(buf); |
661 | return; | |
662 | } | |
663 | ||
664 | BUF_FREE(buf); | |
665 | } | |
666 | ||
9d3a4132 SE |
667 | /* If cid==NULL packet is from host, otherwise cid specifies which tunnel |
668 | it came from. */ | |
469fd1d9 SE |
669 | static void netlink_incoming(struct netlink *st, struct netlink_client *client, |
670 | struct buffer_if *buf) | |
4efd681a | 671 | { |
4efd681a SE |
672 | uint32_t source,dest; |
673 | struct iphdr *iph; | |
d714da29 | 674 | char errmsgbuf[50]; |
a28d65a5 | 675 | const char *sourcedesc=client?client->name:"host"; |
4efd681a SE |
676 | |
677 | BUF_ASSERT_USED(buf); | |
a28d65a5 | 678 | |
d714da29 IJ |
679 | if (!netlink_check(st,buf,errmsgbuf,sizeof(errmsgbuf))) { |
680 | Message(M_WARNING,"%s: bad IP packet from %s: %s\n", | |
a28d65a5 | 681 | st->name,sourcedesc, |
d714da29 | 682 | errmsgbuf); |
4efd681a SE |
683 | BUF_FREE(buf); |
684 | return; | |
685 | } | |
975820aa | 686 | assert(buf->size >= (int)sizeof(struct icmphdr)); |
4efd681a SE |
687 | iph=(struct iphdr *)buf->start; |
688 | ||
689 | source=ntohl(iph->saddr); | |
690 | dest=ntohl(iph->daddr); | |
691 | ||
d3fe100d SE |
692 | /* Check source. If we don't like the source, there's no point |
693 | generating ICMP because we won't know how to get it to the | |
694 | source of the packet. */ | |
9d3a4132 | 695 | if (client) { |
c6f79b17 SE |
696 | /* Check that the packet source is appropriate for the tunnel |
697 | it came down */ | |
794f2398 | 698 | if (!ipset_contains_addr(client->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: packet from tunnel %s with bad " | |
703 | "source address (s=%s,d=%s)\n",st->name,client->name,s,d); | |
704 | free(s); free(d); | |
705 | BUF_FREE(buf); | |
706 | return; | |
707 | } | |
708 | } else { | |
c6f79b17 SE |
709 | /* Check that the packet originates in our configured local |
710 | network, and hasn't been forwarded from elsewhere or | |
711 | generated with the wrong source address */ | |
794f2398 | 712 | if (!ipset_contains_addr(st->networks,source)) { |
9d3a4132 SE |
713 | string_t s,d; |
714 | s=ipaddr_to_string(source); | |
715 | d=ipaddr_to_string(dest); | |
716 | Message(M_WARNING,"%s: outgoing packet with bad source address " | |
717 | "(s=%s,d=%s)\n",st->name,s,d); | |
718 | free(s); free(d); | |
719 | BUF_FREE(buf); | |
720 | return; | |
721 | } | |
4efd681a | 722 | } |
c6f79b17 | 723 | |
794f2398 SE |
724 | /* If this is a point-to-point device we don't examine the |
725 | destination address at all; we blindly send it down our | |
726 | one-and-only registered tunnel, or to the host, depending on | |
d3fe100d SE |
727 | where it came from. It's up to external software to check |
728 | address validity and generate ICMP, etc. */ | |
c6f79b17 SE |
729 | if (st->ptp) { |
730 | if (client) { | |
f928f069 | 731 | netlink_host_deliver(st,source,dest,buf); |
c6f79b17 | 732 | } else { |
7b6abafa | 733 | netlink_client_deliver(st,st->clients,source,dest,buf); |
c6f79b17 SE |
734 | } |
735 | BUF_ASSERT_FREE(buf); | |
736 | return; | |
737 | } | |
738 | ||
d3fe100d SE |
739 | /* st->secnet_address needs checking before matching destination |
740 | addresses */ | |
2fe58dfd | 741 | if (dest==st->secnet_address) { |
9d3a4132 | 742 | netlink_packet_local(st,client,buf); |
4efd681a | 743 | BUF_ASSERT_FREE(buf); |
2fe58dfd SE |
744 | return; |
745 | } | |
70dc107b | 746 | netlink_packet_forward(st,client,buf); |
4efd681a SE |
747 | BUF_ASSERT_FREE(buf); |
748 | } | |
749 | ||
469fd1d9 SE |
750 | static void netlink_inst_incoming(void *sst, struct buffer_if *buf) |
751 | { | |
752 | struct netlink_client *c=sst; | |
753 | struct netlink *st=c->nst; | |
754 | ||
755 | netlink_incoming(st,c,buf); | |
756 | } | |
757 | ||
758 | static void netlink_dev_incoming(void *sst, struct buffer_if *buf) | |
759 | { | |
760 | struct netlink *st=sst; | |
761 | ||
762 | netlink_incoming(st,NULL,buf); | |
763 | } | |
764 | ||
d3fe100d | 765 | static void netlink_set_quality(void *sst, uint32_t quality) |
4efd681a | 766 | { |
d3fe100d SE |
767 | struct netlink_client *c=sst; |
768 | struct netlink *st=c->nst; | |
4efd681a | 769 | |
d3fe100d SE |
770 | c->link_quality=quality; |
771 | c->up=(c->link_quality==LINK_QUALITY_DOWN)?False:True; | |
772 | if (c->options&OPT_SOFTROUTE) { | |
773 | st->set_routes(st->dst,c); | |
4efd681a | 774 | } |
4efd681a SE |
775 | } |
776 | ||
d3fe100d SE |
777 | static void netlink_output_subnets(struct netlink *st, uint32_t loglevel, |
778 | struct subnet_list *snets) | |
4efd681a | 779 | { |
1caa23ff | 780 | int32_t i; |
d3fe100d | 781 | string_t net; |
4efd681a | 782 | |
d3fe100d SE |
783 | for (i=0; i<snets->entries; i++) { |
784 | net=subnet_to_string(snets->list[i]); | |
785 | Message(loglevel,"%s ",net); | |
786 | free(net); | |
9d3a4132 | 787 | } |
4efd681a SE |
788 | } |
789 | ||
042a8da9 | 790 | static void netlink_dump_routes(struct netlink *st, bool_t requested) |
9d3a4132 SE |
791 | { |
792 | int i; | |
793 | string_t net; | |
042a8da9 | 794 | uint32_t c=M_INFO; |
9d3a4132 | 795 | |
042a8da9 | 796 | if (requested) c=M_WARNING; |
469fd1d9 SE |
797 | if (st->ptp) { |
798 | net=ipaddr_to_string(st->secnet_address); | |
34d3bf4c | 799 | Message(c,"%s: point-to-point (remote end is %s); routes: ", |
469fd1d9 | 800 | st->name, net); |
9d3a4132 | 801 | free(net); |
d3fe100d | 802 | netlink_output_subnets(st,c,st->clients->subnets); |
469fd1d9 SE |
803 | Message(c,"\n"); |
804 | } else { | |
805 | Message(c,"%s: routing table:\n",st->name); | |
d3fe100d SE |
806 | for (i=0; i<st->n_clients; i++) { |
807 | netlink_output_subnets(st,c,st->routes[i]->subnets); | |
ff05a229 | 808 | Message(c,"-> tunnel %s (%s,mtu %d,%s routes,%s," |
ea7ec970 | 809 | "quality %d,use %d,pri %lu)\n", |
d3fe100d | 810 | st->routes[i]->name, |
ff05a229 SE |
811 | st->routes[i]->up?"up":"down", |
812 | st->routes[i]->mtu, | |
d3fe100d SE |
813 | st->routes[i]->options&OPT_SOFTROUTE?"soft":"hard", |
814 | st->routes[i]->options&OPT_ALLOWROUTE?"free":"restricted", | |
d3fe100d | 815 | st->routes[i]->link_quality, |
ea7ec970 SE |
816 | st->routes[i]->outcount, |
817 | (unsigned long)st->routes[i]->priority); | |
469fd1d9 SE |
818 | } |
819 | net=ipaddr_to_string(st->secnet_address); | |
820 | Message(c,"%s/32 -> netlink \"%s\" (use %d)\n", | |
821 | net,st->name,st->localcount); | |
9d3a4132 | 822 | free(net); |
794f2398 SE |
823 | for (i=0; i<st->subnets->entries; i++) { |
824 | net=subnet_to_string(st->subnets->list[i]); | |
825 | Message(c,"%s ",net); | |
469fd1d9 SE |
826 | free(net); |
827 | } | |
794f2398 SE |
828 | if (i>0) |
829 | Message(c,"-> host (use %d)\n",st->outcount); | |
9d3a4132 SE |
830 | } |
831 | } | |
832 | ||
d3fe100d SE |
833 | /* ap is a pointer to a member of the routes array */ |
834 | static int netlink_compare_client_priority(const void *ap, const void *bp) | |
70dc107b | 835 | { |
d3fe100d SE |
836 | const struct netlink_client *const*a=ap; |
837 | const struct netlink_client *const*b=bp; | |
70dc107b | 838 | |
d3fe100d SE |
839 | if ((*a)->priority==(*b)->priority) return 0; |
840 | if ((*a)->priority<(*b)->priority) return 1; | |
70dc107b SE |
841 | return -1; |
842 | } | |
843 | ||
844 | static void netlink_phase_hook(void *sst, uint32_t new_phase) | |
845 | { | |
846 | struct netlink *st=sst; | |
847 | struct netlink_client *c; | |
1caa23ff | 848 | int32_t i; |
70dc107b SE |
849 | |
850 | /* All the networks serviced by the various tunnels should now | |
851 | * have been registered. We build a routing table by sorting the | |
d3fe100d | 852 | * clients by priority. */ |
bb9d0561 IJ |
853 | st->routes=safe_malloc_ary(sizeof(*st->routes),st->n_clients, |
854 | "netlink_phase_hook"); | |
70dc107b SE |
855 | /* Fill the table */ |
856 | i=0; | |
59230b9b IJ |
857 | for (c=st->clients; c; c=c->next) { |
858 | assert(i<INT_MAX); | |
d3fe100d | 859 | st->routes[i++]=c; |
59230b9b | 860 | } |
d3fe100d SE |
861 | /* Sort the table in descending order of priority */ |
862 | qsort(st->routes,st->n_clients,sizeof(*st->routes), | |
863 | netlink_compare_client_priority); | |
9d3a4132 | 864 | |
042a8da9 SE |
865 | netlink_dump_routes(st,False); |
866 | } | |
867 | ||
868 | static void netlink_signal_handler(void *sst, int signum) | |
869 | { | |
870 | struct netlink *st=sst; | |
871 | Message(M_INFO,"%s: route dump requested by SIGUSR1\n",st->name); | |
872 | netlink_dump_routes(st,True); | |
70dc107b SE |
873 | } |
874 | ||
1caa23ff | 875 | static void netlink_inst_set_mtu(void *sst, int32_t new_mtu) |
d3fe100d SE |
876 | { |
877 | struct netlink_client *c=sst; | |
878 | ||
879 | c->mtu=new_mtu; | |
880 | } | |
881 | ||
469fd1d9 | 882 | static void netlink_inst_reg(void *sst, netlink_deliver_fn *deliver, |
3abd18e8 | 883 | void *dst) |
469fd1d9 SE |
884 | { |
885 | struct netlink_client *c=sst; | |
469fd1d9 | 886 | |
469fd1d9 SE |
887 | c->deliver=deliver; |
888 | c->dst=dst; | |
889 | } | |
890 | ||
891 | static struct flagstr netlink_option_table[]={ | |
892 | { "soft", OPT_SOFTROUTE }, | |
893 | { "allow-route", OPT_ALLOWROUTE }, | |
894 | { NULL, 0} | |
895 | }; | |
896 | /* This is the routine that gets called when the closure that's | |
897 | returned by an invocation of a netlink device closure (eg. tun, | |
898 | userv-ipif) is invoked. It's used to create routes and pass in | |
899 | information about them; the closure it returns is used by site | |
900 | code. */ | |
901 | static closure_t *netlink_inst_create(struct netlink *st, | |
902 | struct cloc loc, dict_t *dict) | |
903 | { | |
904 | struct netlink_client *c; | |
905 | string_t name; | |
794f2398 | 906 | struct ipset *networks; |
1caa23ff IJ |
907 | uint32_t options,priority; |
908 | int32_t mtu; | |
794f2398 | 909 | list_t *l; |
469fd1d9 SE |
910 | |
911 | name=dict_read_string(dict, "name", True, st->name, loc); | |
912 | ||
794f2398 SE |
913 | l=dict_lookup(dict,"routes"); |
914 | if (!l) | |
915 | cfgfatal(loc,st->name,"required parameter \"routes\" not found\n"); | |
916 | networks=string_list_to_ipset(l,loc,st->name,"routes"); | |
469fd1d9 SE |
917 | options=string_list_to_word(dict_lookup(dict,"options"), |
918 | netlink_option_table,st->name); | |
919 | ||
d3fe100d SE |
920 | priority=dict_read_number(dict,"priority",False,st->name,loc,0); |
921 | mtu=dict_read_number(dict,"mtu",False,st->name,loc,0); | |
922 | ||
923 | if ((options&OPT_SOFTROUTE) && !st->set_routes) { | |
469fd1d9 SE |
924 | cfgfatal(loc,st->name,"this netlink device does not support " |
925 | "soft routes.\n"); | |
926 | return NULL; | |
927 | } | |
928 | ||
929 | if (options&OPT_SOFTROUTE) { | |
930 | /* XXX for now we assume that soft routes require root privilege; | |
931 | this may not always be true. The device driver can tell us. */ | |
932 | require_root_privileges=True; | |
933 | require_root_privileges_explanation="netlink: soft routes"; | |
934 | if (st->ptp) { | |
935 | cfgfatal(loc,st->name,"point-to-point netlinks do not support " | |
936 | "soft routes.\n"); | |
937 | return NULL; | |
938 | } | |
939 | } | |
940 | ||
794f2398 SE |
941 | /* Check that nets are a subset of st->remote_networks; |
942 | refuse to register if they are not. */ | |
943 | if (!ipset_is_subset(st->remote_networks,networks)) { | |
944 | cfgfatal(loc,st->name,"routes are not allowed\n"); | |
469fd1d9 SE |
945 | return NULL; |
946 | } | |
947 | ||
948 | c=safe_malloc(sizeof(*c),"netlink_inst_create"); | |
949 | c->cl.description=name; | |
950 | c->cl.type=CL_NETLINK; | |
951 | c->cl.apply=NULL; | |
952 | c->cl.interface=&c->ops; | |
953 | c->ops.st=c; | |
954 | c->ops.reg=netlink_inst_reg; | |
955 | c->ops.deliver=netlink_inst_incoming; | |
956 | c->ops.set_quality=netlink_set_quality; | |
d3fe100d | 957 | c->ops.set_mtu=netlink_inst_set_mtu; |
469fd1d9 SE |
958 | c->nst=st; |
959 | ||
960 | c->networks=networks; | |
794f2398 | 961 | c->subnets=ipset_to_subnet_list(networks); |
d3fe100d | 962 | c->priority=priority; |
469fd1d9 SE |
963 | c->deliver=NULL; |
964 | c->dst=NULL; | |
965 | c->name=name; | |
f208b9a9 | 966 | c->link_quality=LINK_QUALITY_UNUSED; |
d3fe100d SE |
967 | c->mtu=mtu?mtu:st->mtu; |
968 | c->options=options; | |
969 | c->outcount=0; | |
970 | c->up=False; | |
971 | c->kup=False; | |
469fd1d9 SE |
972 | c->next=st->clients; |
973 | st->clients=c; | |
59230b9b | 974 | assert(st->n_clients < INT_MAX); |
d3fe100d | 975 | st->n_clients++; |
469fd1d9 SE |
976 | |
977 | return &c->cl; | |
978 | } | |
979 | ||
980 | static list_t *netlink_inst_apply(closure_t *self, struct cloc loc, | |
981 | dict_t *context, list_t *args) | |
982 | { | |
983 | struct netlink *st=self->interface; | |
984 | ||
985 | dict_t *dict; | |
986 | item_t *item; | |
987 | closure_t *cl; | |
988 | ||
469fd1d9 SE |
989 | item=list_elem(args,0); |
990 | if (!item || item->type!=t_dict) { | |
991 | cfgfatal(loc,st->name,"must have a dictionary argument\n"); | |
992 | } | |
993 | dict=item->data.dict; | |
994 | ||
995 | cl=netlink_inst_create(st,loc,dict); | |
996 | ||
997 | return new_closure(cl); | |
998 | } | |
999 | ||
9d3a4132 SE |
1000 | netlink_deliver_fn *netlink_init(struct netlink *st, |
1001 | void *dst, struct cloc loc, | |
fe5e9cc4 | 1002 | dict_t *dict, cstring_t description, |
d3fe100d | 1003 | netlink_route_fn *set_routes, |
9d3a4132 | 1004 | netlink_deliver_fn *to_host) |
4efd681a | 1005 | { |
c6f79b17 | 1006 | item_t *sa, *ptpa; |
794f2398 | 1007 | list_t *l; |
c6f79b17 | 1008 | |
4efd681a SE |
1009 | st->dst=dst; |
1010 | st->cl.description=description; | |
469fd1d9 SE |
1011 | st->cl.type=CL_PURE; |
1012 | st->cl.apply=netlink_inst_apply; | |
1013 | st->cl.interface=st; | |
4efd681a | 1014 | st->clients=NULL; |
d3fe100d SE |
1015 | st->routes=NULL; |
1016 | st->n_clients=0; | |
1017 | st->set_routes=set_routes; | |
4efd681a SE |
1018 | st->deliver_to_host=to_host; |
1019 | ||
794f2398 | 1020 | st->name=dict_read_string(dict,"name",False,description,loc); |
4efd681a | 1021 | if (!st->name) st->name=description; |
794f2398 SE |
1022 | l=dict_lookup(dict,"networks"); |
1023 | if (l) | |
1024 | st->networks=string_list_to_ipset(l,loc,st->name,"networks"); | |
1025 | else { | |
4f5e39ec SE |
1026 | struct ipset *empty; |
1027 | empty=ipset_new(); | |
1028 | st->networks=ipset_complement(empty); | |
1029 | ipset_free(empty); | |
794f2398 SE |
1030 | } |
1031 | l=dict_lookup(dict,"remote-networks"); | |
1032 | if (l) { | |
1033 | st->remote_networks=string_list_to_ipset(l,loc,st->name, | |
1034 | "remote-networks"); | |
1035 | } else { | |
1036 | struct ipset *empty; | |
1037 | empty=ipset_new(); | |
1038 | st->remote_networks=ipset_complement(empty); | |
1039 | ipset_free(empty); | |
1040 | } | |
1041 | ||
c6f79b17 | 1042 | sa=dict_find_item(dict,"secnet-address",False,"netlink",loc); |
469fd1d9 | 1043 | ptpa=dict_find_item(dict,"ptp-address",False,"netlink",loc); |
c6f79b17 SE |
1044 | if (sa && ptpa) { |
1045 | cfgfatal(loc,st->name,"you may not specify secnet-address and " | |
1046 | "ptp-address in the same netlink device\n"); | |
1047 | } | |
1048 | if (!(sa || ptpa)) { | |
1049 | cfgfatal(loc,st->name,"you must specify secnet-address or " | |
1050 | "ptp-address for this netlink device\n"); | |
1051 | } | |
1052 | if (sa) { | |
794f2398 | 1053 | st->secnet_address=string_item_to_ipaddr(sa,"netlink"); |
c6f79b17 SE |
1054 | st->ptp=False; |
1055 | } else { | |
794f2398 | 1056 | st->secnet_address=string_item_to_ipaddr(ptpa,"netlink"); |
c6f79b17 SE |
1057 | st->ptp=True; |
1058 | } | |
d3fe100d SE |
1059 | /* To be strictly correct we could subtract secnet_address from |
1060 | networks here. It shouldn't make any practical difference, | |
794f2398 SE |
1061 | though, and will make the route dump look complicated... */ |
1062 | st->subnets=ipset_to_subnet_list(st->networks); | |
4efd681a SE |
1063 | st->mtu=dict_read_number(dict, "mtu", False, "netlink", loc, DEFAULT_MTU); |
1064 | buffer_new(&st->icmp,ICMP_BUFSIZE); | |
469fd1d9 SE |
1065 | st->outcount=0; |
1066 | st->localcount=0; | |
70dc107b SE |
1067 | |
1068 | add_hook(PHASE_SETUP,netlink_phase_hook,st); | |
042a8da9 | 1069 | request_signal_notification(SIGUSR1, netlink_signal_handler, st); |
4efd681a | 1070 | |
469fd1d9 SE |
1071 | /* If we're point-to-point then we return a CL_NETLINK directly, |
1072 | rather than a CL_NETLINK_OLD or pure closure (depending on | |
1073 | compatibility). This CL_NETLINK is for our one and only | |
1074 | client. Our cl.apply function is NULL. */ | |
1075 | if (st->ptp) { | |
1076 | closure_t *cl; | |
1077 | cl=netlink_inst_create(st,loc,dict); | |
1078 | st->cl=*cl; | |
1079 | } | |
1080 | return netlink_dev_incoming; | |
2fe58dfd SE |
1081 | } |
1082 | ||
9d3a4132 | 1083 | /* No connection to the kernel at all... */ |
2fe58dfd | 1084 | |
9d3a4132 | 1085 | struct null { |
4efd681a | 1086 | struct netlink nl; |
4efd681a | 1087 | }; |
2fe58dfd | 1088 | |
d3fe100d | 1089 | static bool_t null_set_route(void *sst, struct netlink_client *routes) |
4efd681a | 1090 | { |
9d3a4132 | 1091 | struct null *st=sst; |
d3fe100d SE |
1092 | |
1093 | if (routes->up!=routes->kup) { | |
1094 | Message(M_INFO,"%s: setting routes for tunnel %s to state %s\n", | |
1095 | st->nl.name,routes->name, | |
1096 | routes->up?"up":"down"); | |
1097 | routes->kup=routes->up; | |
9d3a4132 | 1098 | return True; |
2fe58dfd | 1099 | } |
9d3a4132 | 1100 | return False; |
2fe58dfd | 1101 | } |
9d3a4132 | 1102 | |
469fd1d9 | 1103 | static void null_deliver(void *sst, struct buffer_if *buf) |
2fe58dfd SE |
1104 | { |
1105 | return; | |
1106 | } | |
1107 | ||
1108 | static list_t *null_apply(closure_t *self, struct cloc loc, dict_t *context, | |
1109 | list_t *args) | |
1110 | { | |
1111 | struct null *st; | |
4efd681a SE |
1112 | item_t *item; |
1113 | dict_t *dict; | |
2fe58dfd | 1114 | |
4efd681a | 1115 | st=safe_malloc(sizeof(*st),"null_apply"); |
2fe58dfd | 1116 | |
4efd681a SE |
1117 | item=list_elem(args,0); |
1118 | if (!item || item->type!=t_dict) | |
1119 | cfgfatal(loc,"null-netlink","parameter must be a dictionary\n"); | |
1120 | ||
1121 | dict=item->data.dict; | |
1122 | ||
9d3a4132 SE |
1123 | netlink_init(&st->nl,st,loc,dict,"null-netlink",null_set_route, |
1124 | null_deliver); | |
4efd681a SE |
1125 | |
1126 | return new_closure(&st->nl.cl); | |
2fe58dfd SE |
1127 | } |
1128 | ||
2fe58dfd SE |
1129 | void netlink_module(dict_t *dict) |
1130 | { | |
4efd681a | 1131 | add_closure(dict,"null-netlink",null_apply); |
2fe58dfd | 1132 | } |