4d0fe69d2f2a526bc03485ac37e663ee2ebd3f7c
[exim.git] / src / src / verify.c
1 /* $Cambridge: exim/src/src/verify.c,v 1.41 2006/10/03 15:11:22 ph10 Exp $ */
2
3 /*************************************************
4 * Exim - an Internet mail transport agent *
5 *************************************************/
6
7 /* Copyright (c) University of Cambridge 1995 - 2006 */
8 /* See the file NOTICE for conditions of use and distribution. */
9
10 /* Functions concerned with verifying things. The original code for callout
11 caching was contributed by Kevin Fleming (but I hacked it around a bit). */
12
13
14 #include "exim.h"
15
16
17 /* Structure for caching DNSBL lookups */
18
19 typedef struct dnsbl_cache_block {
20 dns_address *rhs;
21 uschar *text;
22 int rc;
23 BOOL text_set;
24 } dnsbl_cache_block;
25
26
27 /* Anchor for DNSBL cache */
28
29 static tree_node *dnsbl_cache = NULL;
30
31
32
33 /*************************************************
34 * Retrieve a callout cache record *
35 *************************************************/
36
37 /* If a record exists, check whether it has expired.
38
39 Arguments:
40 dbm_file an open hints file
41 key the record key
42 type "address" or "domain"
43 positive_expire expire time for positive records
44 negative_expire expire time for negative records
45
46 Returns: the cache record if a non-expired one exists, else NULL
47 */
48
49 static dbdata_callout_cache *
50 get_callout_cache_record(open_db *dbm_file, uschar *key, uschar *type,
51 int positive_expire, int negative_expire)
52 {
53 BOOL negative;
54 int length, expire;
55 time_t now;
56 dbdata_callout_cache *cache_record;
57
58 cache_record = dbfn_read_with_length(dbm_file, key, &length);
59
60 if (cache_record == NULL)
61 {
62 HDEBUG(D_verify) debug_printf("callout cache: no %s record found\n", type);
63 return NULL;
64 }
65
66 /* We treat a record as "negative" if its result field is not positive, or if
67 it is a domain record and the postmaster field is negative. */
68
69 negative = cache_record->result != ccache_accept ||
70 (type[0] == 'd' && cache_record->postmaster_result == ccache_reject);
71 expire = negative? negative_expire : positive_expire;
72 now = time(NULL);
73
74 if (now - cache_record->time_stamp > expire)
75 {
76 HDEBUG(D_verify) debug_printf("callout cache: %s record expired\n", type);
77 return NULL;
78 }
79
80 /* If this is a non-reject domain record, check for the obsolete format version
81 that doesn't have the postmaster and random timestamps, by looking at the
82 length. If so, copy it to a new-style block, replicating the record's
83 timestamp. Then check the additional timestamps. (There's no point wasting
84 effort if connections are rejected.) */
85
86 if (type[0] == 'd' && cache_record->result != ccache_reject)
87 {
88 if (length == sizeof(dbdata_callout_cache_obs))
89 {
90 dbdata_callout_cache *new = store_get(sizeof(dbdata_callout_cache));
91 memcpy(new, cache_record, length);
92 new->postmaster_stamp = new->random_stamp = new->time_stamp;
93 cache_record = new;
94 }
95
96 if (now - cache_record->postmaster_stamp > expire)
97 cache_record->postmaster_result = ccache_unknown;
98
99 if (now - cache_record->random_stamp > expire)
100 cache_record->random_result = ccache_unknown;
101 }
102
103 HDEBUG(D_verify) debug_printf("callout cache: found %s record\n", type);
104 return cache_record;
105 }
106
107
108
109 /*************************************************
110 * Do callout verification for an address *
111 *************************************************/
112
113 /* This function is called from verify_address() when the address has routed to
114 a host list, and a callout has been requested. Callouts are expensive; that is
115 why a cache is used to improve the efficiency.
116
117 Arguments:
118 addr the address that's been routed
119 host_list the list of hosts to try
120 tf the transport feedback block
121
122 ifstring "interface" option from transport, or NULL
123 portstring "port" option from transport, or NULL
124 protocolstring "protocol" option from transport, or NULL
125 callout the per-command callout timeout
126 callout_overall the overall callout timeout (if < 0 use 4*callout)
127 callout_connect the callout connection timeout (if < 0 use callout)
128 options the verification options - these bits are used:
129 vopt_is_recipient => this is a recipient address
130 vopt_callout_no_cache => don't use callout cache
131 vopt_callout_fullpm => if postmaster check, do full one
132 vopt_callout_random => do the "random" thing
133 vopt_callout_recipsender => use real sender for recipient
134 vopt_callout_recippmaster => use postmaster for recipient
135 se_mailfrom MAIL FROM address for sender verify; NULL => ""
136 pm_mailfrom if non-NULL, do the postmaster check with this sender
137
138 Returns: OK/FAIL/DEFER
139 */
140
141 static int
142 do_callout(address_item *addr, host_item *host_list, transport_feedback *tf,
143 int callout, int callout_overall, int callout_connect, int options,
144 uschar *se_mailfrom, uschar *pm_mailfrom)
145 {
146 BOOL is_recipient = (options & vopt_is_recipient) != 0;
147 BOOL callout_no_cache = (options & vopt_callout_no_cache) != 0;
148 BOOL callout_random = (options & vopt_callout_random) != 0;
149
150 int yield = OK;
151 int old_domain_cache_result = ccache_accept;
152 BOOL done = FALSE;
153 uschar *address_key;
154 uschar *from_address;
155 uschar *random_local_part = NULL;
156 uschar *save_deliver_domain = deliver_domain;
157 uschar **failure_ptr = is_recipient?
158 &recipient_verify_failure : &sender_verify_failure;
159 open_db dbblock;
160 open_db *dbm_file = NULL;
161 dbdata_callout_cache new_domain_record;
162 dbdata_callout_cache_address new_address_record;
163 host_item *host;
164 time_t callout_start_time;
165
166 new_domain_record.result = ccache_unknown;
167 new_domain_record.postmaster_result = ccache_unknown;
168 new_domain_record.random_result = ccache_unknown;
169
170 memset(&new_address_record, 0, sizeof(new_address_record));
171
172 /* For a recipient callout, the key used for the address cache record must
173 include the sender address if we are using the real sender in the callout,
174 because that may influence the result of the callout. */
175
176 address_key = addr->address;
177 from_address = US"";
178
179 if (is_recipient)
180 {
181 if ((options & vopt_callout_recipsender) != 0)
182 {
183 address_key = string_sprintf("%s/<%s>", addr->address, sender_address);
184 from_address = sender_address;
185 }
186 else if ((options & vopt_callout_recippmaster) != 0)
187 {
188 address_key = string_sprintf("%s/<postmaster@%s>", addr->address,
189 qualify_domain_sender);
190 from_address = string_sprintf("postmaster@%s", qualify_domain_sender);
191 }
192 }
193
194 /* For a sender callout, we must adjust the key if the mailfrom address is not
195 empty. */
196
197 else
198 {
199 from_address = (se_mailfrom == NULL)? US"" : se_mailfrom;
200 if (from_address[0] != 0)
201 address_key = string_sprintf("%s/<%s>", addr->address, from_address);
202 }
203
204 /* Open the callout cache database, it it exists, for reading only at this
205 stage, unless caching has been disabled. */
206
207 if (callout_no_cache)
208 {
209 HDEBUG(D_verify) debug_printf("callout cache: disabled by no_cache\n");
210 }
211 else if ((dbm_file = dbfn_open(US"callout", O_RDWR, &dbblock, FALSE)) == NULL)
212 {
213 HDEBUG(D_verify) debug_printf("callout cache: not available\n");
214 }
215
216 /* If a cache database is available see if we can avoid the need to do an
217 actual callout by making use of previously-obtained data. */
218
219 if (dbm_file != NULL)
220 {
221 dbdata_callout_cache_address *cache_address_record;
222 dbdata_callout_cache *cache_record = get_callout_cache_record(dbm_file,
223 addr->domain, US"domain",
224 callout_cache_domain_positive_expire,
225 callout_cache_domain_negative_expire);
226
227 /* If an unexpired cache record was found for this domain, see if the callout
228 process can be short-circuited. */
229
230 if (cache_record != NULL)
231 {
232 /* In most cases, if an early command (up to and including MAIL FROM:<>)
233 was rejected, there is no point carrying on. The callout fails. However, if
234 we are doing a recipient verification with use_sender or use_postmaster
235 set, a previous failure of MAIL FROM:<> doesn't count, because this time we
236 will be using a non-empty sender. We have to remember this situation so as
237 not to disturb the cached domain value if this whole verification succeeds
238 (we don't want it turning into "accept"). */
239
240 old_domain_cache_result = cache_record->result;
241
242 if (cache_record->result == ccache_reject ||
243 (*from_address == 0 && cache_record->result == ccache_reject_mfnull))
244 {
245 setflag(addr, af_verify_nsfail);
246 HDEBUG(D_verify)
247 debug_printf("callout cache: domain gave initial rejection, or "
248 "does not accept HELO or MAIL FROM:<>\n");
249 setflag(addr, af_verify_nsfail);
250 addr->user_message = US"(result of an earlier callout reused).";
251 yield = FAIL;
252 *failure_ptr = US"mail";
253 goto END_CALLOUT;
254 }
255
256 /* If a previous check on a "random" local part was accepted, we assume
257 that the server does not do any checking on local parts. There is therefore
258 no point in doing the callout, because it will always be successful. If a
259 random check previously failed, arrange not to do it again, but preserve
260 the data in the new record. If a random check is required but hasn't been
261 done, skip the remaining cache processing. */
262
263 if (callout_random) switch(cache_record->random_result)
264 {
265 case ccache_accept:
266 HDEBUG(D_verify)
267 debug_printf("callout cache: domain accepts random addresses\n");
268 goto END_CALLOUT; /* Default yield is OK */
269
270 case ccache_reject:
271 HDEBUG(D_verify)
272 debug_printf("callout cache: domain rejects random addresses\n");
273 callout_random = FALSE;
274 new_domain_record.random_result = ccache_reject;
275 new_domain_record.random_stamp = cache_record->random_stamp;
276 break;
277
278 default:
279 HDEBUG(D_verify)
280 debug_printf("callout cache: need to check random address handling "
281 "(not cached or cache expired)\n");
282 goto END_CACHE;
283 }
284
285 /* If a postmaster check is requested, but there was a previous failure,
286 there is again no point in carrying on. If a postmaster check is required,
287 but has not been done before, we are going to have to do a callout, so skip
288 remaining cache processing. */
289
290 if (pm_mailfrom != NULL)
291 {
292 if (cache_record->postmaster_result == ccache_reject)
293 {
294 setflag(addr, af_verify_pmfail);
295 HDEBUG(D_verify)
296 debug_printf("callout cache: domain does not accept "
297 "RCPT TO:<postmaster@domain>\n");
298 yield = FAIL;
299 *failure_ptr = US"postmaster";
300 setflag(addr, af_verify_pmfail);
301 addr->user_message = US"(result of earlier verification reused).";
302 goto END_CALLOUT;
303 }
304 if (cache_record->postmaster_result == ccache_unknown)
305 {
306 HDEBUG(D_verify)
307 debug_printf("callout cache: need to check RCPT "
308 "TO:<postmaster@domain> (not cached or cache expired)\n");
309 goto END_CACHE;
310 }
311
312 /* If cache says OK, set pm_mailfrom NULL to prevent a redundant
313 postmaster check if the address itself has to be checked. Also ensure
314 that the value in the cache record is preserved (with its old timestamp).
315 */
316
317 HDEBUG(D_verify) debug_printf("callout cache: domain accepts RCPT "
318 "TO:<postmaster@domain>\n");
319 pm_mailfrom = NULL;
320 new_domain_record.postmaster_result = ccache_accept;
321 new_domain_record.postmaster_stamp = cache_record->postmaster_stamp;
322 }
323 }
324
325 /* We can't give a result based on information about the domain. See if there
326 is an unexpired cache record for this specific address (combined with the
327 sender address if we are doing a recipient callout with a non-empty sender).
328 */
329
330 cache_address_record = (dbdata_callout_cache_address *)
331 get_callout_cache_record(dbm_file,
332 address_key, US"address",
333 callout_cache_positive_expire,
334 callout_cache_negative_expire);
335
336 if (cache_address_record != NULL)
337 {
338 if (cache_address_record->result == ccache_accept)
339 {
340 HDEBUG(D_verify)
341 debug_printf("callout cache: address record is positive\n");
342 }
343 else
344 {
345 HDEBUG(D_verify)
346 debug_printf("callout cache: address record is negative\n");
347 addr->user_message = US"Previous (cached) callout verification failure";
348 *failure_ptr = US"recipient";
349 yield = FAIL;
350 }
351 goto END_CALLOUT;
352 }
353
354 /* Close the cache database while we actually do the callout for real. */
355
356 END_CACHE:
357 dbfn_close(dbm_file);
358 dbm_file = NULL;
359 }
360
361 /* The information wasn't available in the cache, so we have to do a real
362 callout and save the result in the cache for next time, unless no_cache is set,
363 or unless we have a previously cached negative random result. If we are to test
364 with a random local part, ensure that such a local part is available. If not,
365 log the fact, but carry on without randomming. */
366
367 if (callout_random && callout_random_local_part != NULL)
368 {
369 random_local_part = expand_string(callout_random_local_part);
370 if (random_local_part == NULL)
371 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand "
372 "callout_random_local_part: %s", expand_string_message);
373 }
374
375 /* Default the connect and overall callout timeouts if not set, and record the
376 time we are starting so that we can enforce it. */
377
378 if (callout_overall < 0) callout_overall = 4 * callout;
379 if (callout_connect < 0) callout_connect = callout;
380 callout_start_time = time(NULL);
381
382 /* Now make connections to the hosts and do real callouts. The list of hosts
383 is passed in as an argument. */
384
385 for (host = host_list; host != NULL && !done; host = host->next)
386 {
387 smtp_inblock inblock;
388 smtp_outblock outblock;
389 int host_af;
390 int port = 25;
391 BOOL send_quit = TRUE;
392 uschar *active_hostname = smtp_active_hostname;
393 uschar *helo = US"HELO";
394 uschar *interface = NULL; /* Outgoing interface to use; NULL => any */
395 uschar inbuffer[4096];
396 uschar outbuffer[1024];
397 uschar responsebuffer[4096];
398
399 clearflag(addr, af_verify_pmfail); /* postmaster callout flag */
400 clearflag(addr, af_verify_nsfail); /* null sender callout flag */
401
402 /* Skip this host if we don't have an IP address for it. */
403
404 if (host->address == NULL)
405 {
406 DEBUG(D_verify) debug_printf("no IP address for host name %s: skipping\n",
407 host->name);
408 continue;
409 }
410
411 /* Check the overall callout timeout */
412
413 if (time(NULL) - callout_start_time >= callout_overall)
414 {
415 HDEBUG(D_verify) debug_printf("overall timeout for callout exceeded\n");
416 break;
417 }
418
419 /* Set IPv4 or IPv6 */
420
421 host_af = (Ustrchr(host->address, ':') == NULL)? AF_INET:AF_INET6;
422
423 /* Expand and interpret the interface and port strings. The latter will not
424 be used if there is a host-specific port (e.g. from a manualroute router).
425 This has to be delayed till now, because they may expand differently for
426 different hosts. If there's a failure, log it, but carry on with the
427 defaults. */
428
429 deliver_host = host->name;
430 deliver_host_address = host->address;
431 deliver_domain = addr->domain;
432
433 if (!smtp_get_interface(tf->interface, host_af, addr, NULL, &interface,
434 US"callout") ||
435 !smtp_get_port(tf->port, addr, &port, US"callout"))
436 log_write(0, LOG_MAIN|LOG_PANIC, "<%s>: %s", addr->address,
437 addr->message);
438
439 /* Expand the helo_data string to find the host name to use. */
440
441 if (tf->helo_data != NULL)
442 {
443 uschar *s = expand_string(tf->helo_data);
444 if (active_hostname == NULL)
445 log_write(0, LOG_MAIN|LOG_PANIC, "<%s>: failed to expand transport's "
446 "helo_data value for callout: %s", expand_string_message);
447 else active_hostname = s;
448 }
449
450 deliver_host = deliver_host_address = NULL;
451 deliver_domain = save_deliver_domain;
452
453 /* Set HELO string according to the protocol */
454
455 if (Ustrcmp(tf->protocol, "lmtp") == 0) helo = US"LHLO";
456
457 HDEBUG(D_verify) debug_printf("interface=%s port=%d\n", interface, port);
458
459 /* Set up the buffer for reading SMTP response packets. */
460
461 inblock.buffer = inbuffer;
462 inblock.buffersize = sizeof(inbuffer);
463 inblock.ptr = inbuffer;
464 inblock.ptrend = inbuffer;
465
466 /* Set up the buffer for holding SMTP commands while pipelining */
467
468 outblock.buffer = outbuffer;
469 outblock.buffersize = sizeof(outbuffer);
470 outblock.ptr = outbuffer;
471 outblock.cmd_count = 0;
472 outblock.authenticating = FALSE;
473
474 /* Connect to the host; on failure, just loop for the next one, but we
475 set the error for the last one. Use the callout_connect timeout. */
476
477 inblock.sock = outblock.sock =
478 smtp_connect(host, host_af, port, interface, callout_connect, TRUE);
479 if (inblock.sock < 0)
480 {
481 addr->message = string_sprintf("could not connect to %s [%s]: %s",
482 host->name, host->address, strerror(errno));
483 continue;
484 }
485
486 /* Wait for initial response, and send HELO. The smtp_write_command()
487 function leaves its command in big_buffer. This is used in error responses.
488 Initialize it in case the connection is rejected. */
489
490 Ustrcpy(big_buffer, "initial connection");
491
492 done =
493 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
494 '2', callout) &&
495 smtp_write_command(&outblock, FALSE, "%s %s\r\n", helo,
496 active_hostname) >= 0 &&
497 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
498 '2', callout);
499
500 /* Failure to accept HELO is cached; this blocks the whole domain for all
501 senders. I/O errors and defer responses are not cached. */
502
503 if (!done)
504 {
505 *failure_ptr = US"mail"; /* At or before MAIL */
506 if (errno == 0 && responsebuffer[0] == '5')
507 {
508 setflag(addr, af_verify_nsfail);
509 new_domain_record.result = ccache_reject;
510 }
511 }
512
513 /* Send the MAIL command */
514
515 else done =
516 smtp_write_command(&outblock, FALSE, "MAIL FROM:<%s>\r\n",
517 from_address) >= 0 &&
518 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
519 '2', callout);
520
521 /* If the host does not accept MAIL FROM:<>, arrange to cache this
522 information, but again, don't record anything for an I/O error or a defer. Do
523 not cache rejections of MAIL when a non-empty sender has been used, because
524 that blocks the whole domain for all senders. */
525
526 if (!done)
527 {
528 *failure_ptr = US"mail"; /* At or before MAIL */
529 if (errno == 0 && responsebuffer[0] == '5')
530 {
531 setflag(addr, af_verify_nsfail);
532 if (from_address[0] == 0)
533 new_domain_record.result = ccache_reject_mfnull;
534 }
535 }
536
537 /* Otherwise, proceed to check a "random" address (if required), then the
538 given address, and the postmaster address (if required). Between each check,
539 issue RSET, because some servers accept only one recipient after MAIL
540 FROM:<>.
541
542 Before doing this, set the result in the domain cache record to "accept",
543 unless its previous value was ccache_reject_mfnull. In that case, the domain
544 rejects MAIL FROM:<> and we want to continue to remember that. When that is
545 the case, we have got here only in the case of a recipient verification with
546 a non-null sender. */
547
548 else
549 {
550 new_domain_record.result =
551 (old_domain_cache_result == ccache_reject_mfnull)?
552 ccache_reject_mfnull: ccache_accept;
553
554 /* Do the random local part check first */
555
556 if (random_local_part != NULL)
557 {
558 uschar randombuffer[1024];
559 BOOL random_ok =
560 smtp_write_command(&outblock, FALSE,
561 "RCPT TO:<%.1000s@%.1000s>\r\n", random_local_part,
562 addr->domain) >= 0 &&
563 smtp_read_response(&inblock, randombuffer,
564 sizeof(randombuffer), '2', callout);
565
566 /* Remember when we last did a random test */
567
568 new_domain_record.random_stamp = time(NULL);
569
570 /* If accepted, we aren't going to do any further tests below. */
571
572 if (random_ok)
573 {
574 new_domain_record.random_result = ccache_accept;
575 }
576
577 /* Otherwise, cache a real negative response, and get back to the right
578 state to send RCPT. Unless there's some problem such as a dropped
579 connection, we expect to succeed, because the commands succeeded above. */
580
581 else if (errno == 0)
582 {
583 if (randombuffer[0] == '5')
584 new_domain_record.random_result = ccache_reject;
585
586 done =
587 smtp_write_command(&outblock, FALSE, "RSET\r\n") >= 0 &&
588 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
589 '2', callout) &&
590
591 smtp_write_command(&outblock, FALSE, "MAIL FROM:<%s>\r\n",
592 from_address) >= 0 &&
593 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
594 '2', callout);
595 }
596 else done = FALSE; /* Some timeout/connection problem */
597 } /* Random check */
598
599 /* If the host is accepting all local parts, as determined by the "random"
600 check, we don't need to waste time doing any further checking. */
601
602 if (new_domain_record.random_result != ccache_accept && done)
603 {
604 /* Get the rcpt_include_affixes flag from the transport if there is one,
605 but assume FALSE if there is not. */
606
607 done =
608 smtp_write_command(&outblock, FALSE, "RCPT TO:<%.1000s>\r\n",
609 transport_rcpt_address(addr,
610 (addr->transport == NULL)? FALSE :
611 addr->transport->rcpt_include_affixes)) >= 0 &&
612 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
613 '2', callout);
614
615 if (done)
616 new_address_record.result = ccache_accept;
617 else if (errno == 0 && responsebuffer[0] == '5')
618 {
619 *failure_ptr = US"recipient";
620 new_address_record.result = ccache_reject;
621 }
622
623 /* Do postmaster check if requested; if a full check is required, we
624 check for RCPT TO:<postmaster> (no domain) in accordance with RFC 821. */
625
626 if (done && pm_mailfrom != NULL)
627 {
628 done =
629 smtp_write_command(&outblock, FALSE, "RSET\r\n") >= 0 &&
630 smtp_read_response(&inblock, responsebuffer,
631 sizeof(responsebuffer), '2', callout) &&
632
633 smtp_write_command(&outblock, FALSE,
634 "MAIL FROM:<%s>\r\n", pm_mailfrom) >= 0 &&
635 smtp_read_response(&inblock, responsebuffer,
636 sizeof(responsebuffer), '2', callout) &&
637
638 /* First try using the current domain */
639
640 ((
641 smtp_write_command(&outblock, FALSE,
642 "RCPT TO:<postmaster@%.1000s>\r\n", addr->domain) >= 0 &&
643 smtp_read_response(&inblock, responsebuffer,
644 sizeof(responsebuffer), '2', callout)
645 )
646
647 ||
648
649 /* If that doesn't work, and a full check is requested,
650 try without the domain. */
651
652 (
653 (options & vopt_callout_fullpm) != 0 &&
654 smtp_write_command(&outblock, FALSE,
655 "RCPT TO:<postmaster>\r\n") >= 0 &&
656 smtp_read_response(&inblock, responsebuffer,
657 sizeof(responsebuffer), '2', callout)
658 ));
659
660 /* Sort out the cache record */
661
662 new_domain_record.postmaster_stamp = time(NULL);
663
664 if (done)
665 new_domain_record.postmaster_result = ccache_accept;
666 else if (errno == 0 && responsebuffer[0] == '5')
667 {
668 *failure_ptr = US"postmaster";
669 setflag(addr, af_verify_pmfail);
670 new_domain_record.postmaster_result = ccache_reject;
671 }
672 }
673 } /* Random not accepted */
674 } /* MAIL FROM: accepted */
675
676 /* For any failure of the main check, other than a negative response, we just
677 close the connection and carry on. We can identify a negative response by the
678 fact that errno is zero. For I/O errors it will be non-zero
679
680 Set up different error texts for logging and for sending back to the caller
681 as an SMTP response. Log in all cases, using a one-line format. For sender
682 callouts, give a full response to the caller, but for recipient callouts,
683 don't give the IP address because this may be an internal host whose identity
684 is not to be widely broadcast. */
685
686 if (!done)
687 {
688 if (errno == ETIMEDOUT)
689 {
690 HDEBUG(D_verify) debug_printf("SMTP timeout\n");
691 send_quit = FALSE;
692 }
693 else if (errno == 0)
694 {
695 if (*responsebuffer == 0) Ustrcpy(responsebuffer, US"connection dropped");
696
697 addr->message =
698 string_sprintf("response to \"%s\" from %s [%s] was: %s",
699 big_buffer, host->name, host->address,
700 string_printing(responsebuffer));
701
702 addr->user_message = is_recipient?
703 string_sprintf("Callout verification failed:\n%s", responsebuffer)
704 :
705 string_sprintf("Called: %s\nSent: %s\nResponse: %s",
706 host->address, big_buffer, responsebuffer);
707
708 /* Hard rejection ends the process */
709
710 if (responsebuffer[0] == '5') /* Address rejected */
711 {
712 yield = FAIL;
713 done = TRUE;
714 }
715 }
716 }
717
718 /* End the SMTP conversation and close the connection. */
719
720 if (send_quit) (void)smtp_write_command(&outblock, FALSE, "QUIT\r\n");
721 (void)close(inblock.sock);
722 } /* Loop through all hosts, while !done */
723
724 /* If we get here with done == TRUE, a successful callout happened, and yield
725 will be set OK or FAIL according to the response to the RCPT command.
726 Otherwise, we looped through the hosts but couldn't complete the business.
727 However, there may be domain-specific information to cache in both cases.
728
729 The value of the result field in the new_domain record is ccache_unknown if
730 there was an error before or with MAIL FROM:, and errno was not zero,
731 implying some kind of I/O error. We don't want to write the cache in that case.
732 Otherwise the value is ccache_accept, ccache_reject, or ccache_reject_mfnull. */
733
734 if (!callout_no_cache && new_domain_record.result != ccache_unknown)
735 {
736 if ((dbm_file = dbfn_open(US"callout", O_RDWR|O_CREAT, &dbblock, FALSE))
737 == NULL)
738 {
739 HDEBUG(D_verify) debug_printf("callout cache: not available\n");
740 }
741 else
742 {
743 (void)dbfn_write(dbm_file, addr->domain, &new_domain_record,
744 (int)sizeof(dbdata_callout_cache));
745 HDEBUG(D_verify) debug_printf("wrote callout cache domain record:\n"
746 " result=%d postmaster=%d random=%d\n",
747 new_domain_record.result,
748 new_domain_record.postmaster_result,
749 new_domain_record.random_result);
750 }
751 }
752
753 /* If a definite result was obtained for the callout, cache it unless caching
754 is disabled. */
755
756 if (done)
757 {
758 if (!callout_no_cache && new_address_record.result != ccache_unknown)
759 {
760 if (dbm_file == NULL)
761 dbm_file = dbfn_open(US"callout", O_RDWR|O_CREAT, &dbblock, FALSE);
762 if (dbm_file == NULL)
763 {
764 HDEBUG(D_verify) debug_printf("no callout cache available\n");
765 }
766 else
767 {
768 (void)dbfn_write(dbm_file, address_key, &new_address_record,
769 (int)sizeof(dbdata_callout_cache_address));
770 HDEBUG(D_verify) debug_printf("wrote %s callout cache address record\n",
771 (new_address_record.result == ccache_accept)? "positive" : "negative");
772 }
773 }
774 } /* done */
775
776 /* Failure to connect to any host, or any response other than 2xx or 5xx is a
777 temporary error. If there was only one host, and a response was received, leave
778 it alone if supplying details. Otherwise, give a generic response. */
779
780 else /* !done */
781 {
782 uschar *dullmsg = string_sprintf("Could not complete %s verify callout",
783 is_recipient? "recipient" : "sender");
784 yield = DEFER;
785
786 if (host_list->next != NULL || addr->message == NULL) addr->message = dullmsg;
787
788 addr->user_message = (!smtp_return_error_details)? dullmsg :
789 string_sprintf("%s for <%s>.\n"
790 "The mail server(s) for the domain may be temporarily unreachable, or\n"
791 "they may be permanently unreachable from this server. In the latter case,\n%s",
792 dullmsg, addr->address,
793 is_recipient?
794 "the address will never be accepted."
795 :
796 "you need to change the address or create an MX record for its domain\n"
797 "if it is supposed to be generally accessible from the Internet.\n"
798 "Talk to your mail administrator for details.");
799
800 /* Force a specific error code */
801
802 addr->basic_errno = ERRNO_CALLOUTDEFER;
803 }
804
805 /* Come here from within the cache-reading code on fast-track exit. */
806
807 END_CALLOUT:
808 if (dbm_file != NULL) dbfn_close(dbm_file);
809 return yield;
810 }
811
812
813
814 /*************************************************
815 * Copy error to toplevel address *
816 *************************************************/
817
818 /* This function is used when a verify fails or defers, to ensure that the
819 failure or defer information is in the original toplevel address. This applies
820 when an address is redirected to a single new address, and the failure or
821 deferral happens to the child address.
822
823 Arguments:
824 vaddr the verify address item
825 addr the final address item
826 yield FAIL or DEFER
827
828 Returns: the value of YIELD
829 */
830
831 static int
832 copy_error(address_item *vaddr, address_item *addr, int yield)
833 {
834 if (addr != vaddr)
835 {
836 vaddr->message = addr->message;
837 vaddr->user_message = addr->user_message;
838 vaddr->basic_errno = addr->basic_errno;
839 vaddr->more_errno = addr->more_errno;
840 vaddr->p.address_data = addr->p.address_data;
841 }
842 return yield;
843 }
844
845
846
847
848 /*************************************************
849 * Verify an email address *
850 *************************************************/
851
852 /* This function is used both for verification (-bv and at other times) and
853 address testing (-bt), which is indicated by address_test_mode being set.
854
855 Arguments:
856 vaddr contains the address to verify; the next field in this block
857 must be NULL
858 f if not NULL, write the result to this file
859 options various option bits:
860 vopt_fake_sender => this sender verify is not for the real
861 sender (it was verify=sender=xxxx or an address from a
862 header line) - rewriting must not change sender_address
863 vopt_is_recipient => this is a recipient address, otherwise
864 it's a sender address - this affects qualification and
865 rewriting and messages from callouts
866 vopt_qualify => qualify an unqualified address; else error
867 vopt_expn => called from SMTP EXPN command
868 vopt_success_on_redirect => when a new address is generated
869 the verification instantly succeeds
870
871 These ones are used by do_callout() -- the options variable
872 is passed to it.
873
874 vopt_callout_fullpm => if postmaster check, do full one
875 vopt_callout_no_cache => don't use callout cache
876 vopt_callout_random => do the "random" thing
877 vopt_callout_recipsender => use real sender for recipient
878 vopt_callout_recippmaster => use postmaster for recipient
879
880 callout if > 0, specifies that callout is required, and gives timeout
881 for individual commands
882 callout_overall if > 0, gives overall timeout for the callout function;
883 if < 0, a default is used (see do_callout())
884 callout_connect the connection timeout for callouts
885 se_mailfrom when callout is requested to verify a sender, use this
886 in MAIL FROM; NULL => ""
887 pm_mailfrom when callout is requested, if non-NULL, do the postmaster
888 thing and use this as the sender address (may be "")
889
890 routed if not NULL, set TRUE if routing succeeded, so we can
891 distinguish between routing failed and callout failed
892
893 Returns: OK address verified
894 FAIL address failed to verify
895 DEFER can't tell at present
896 */
897
898 int
899 verify_address(address_item *vaddr, FILE *f, int options, int callout,
900 int callout_overall, int callout_connect, uschar *se_mailfrom,
901 uschar *pm_mailfrom, BOOL *routed)
902 {
903 BOOL allok = TRUE;
904 BOOL full_info = (f == NULL)? FALSE : (debug_selector != 0);
905 BOOL is_recipient = (options & vopt_is_recipient) != 0;
906 BOOL expn = (options & vopt_expn) != 0;
907 BOOL success_on_redirect = (options & vopt_success_on_redirect) != 0;
908 int i;
909 int yield = OK;
910 int verify_type = expn? v_expn :
911 address_test_mode? v_none :
912 is_recipient? v_recipient : v_sender;
913 address_item *addr_list;
914 address_item *addr_new = NULL;
915 address_item *addr_remote = NULL;
916 address_item *addr_local = NULL;
917 address_item *addr_succeed = NULL;
918 uschar **failure_ptr = is_recipient?
919 &recipient_verify_failure : &sender_verify_failure;
920 uschar *ko_prefix, *cr;
921 uschar *address = vaddr->address;
922 uschar *save_sender;
923 uschar null_sender[] = { 0 }; /* Ensure writeable memory */
924
925 /* Clear, just in case */
926
927 *failure_ptr = NULL;
928
929 /* Set up a prefix and suffix for error message which allow us to use the same
930 output statements both in EXPN mode (where an SMTP response is needed) and when
931 debugging with an output file. */
932
933 if (expn)
934 {
935 ko_prefix = US"553 ";
936 cr = US"\r";
937 }
938 else ko_prefix = cr = US"";
939
940 /* Add qualify domain if permitted; otherwise an unqualified address fails. */
941
942 if (parse_find_at(address) == NULL)
943 {
944 if ((options & vopt_qualify) == 0)
945 {
946 if (f != NULL)
947 fprintf(f, "%sA domain is required for \"%s\"%s\n", ko_prefix, address,
948 cr);
949 *failure_ptr = US"qualify";
950 return FAIL;
951 }
952 address = rewrite_address_qualify(address, is_recipient);
953 }
954
955 DEBUG(D_verify)
956 {
957 debug_printf(">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>\n");
958 debug_printf("%s %s\n", address_test_mode? "Testing" : "Verifying", address);
959 }
960
961 /* Rewrite and report on it. Clear the domain and local part caches - these
962 may have been set by domains and local part tests during an ACL. */
963
964 if (global_rewrite_rules != NULL)
965 {
966 uschar *old = address;
967 address = rewrite_address(address, is_recipient, FALSE,
968 global_rewrite_rules, rewrite_existflags);
969 if (address != old)
970 {
971 for (i = 0; i < (MAX_NAMED_LIST * 2)/32; i++) vaddr->localpart_cache[i] = 0;
972 for (i = 0; i < (MAX_NAMED_LIST * 2)/32; i++) vaddr->domain_cache[i] = 0;
973 if (f != NULL && !expn) fprintf(f, "Address rewritten as: %s\n", address);
974 }
975 }
976
977 /* If this is the real sender address, we must update sender_address at
978 this point, because it may be referred to in the routers. */
979
980 if ((options & (vopt_fake_sender|vopt_is_recipient)) == 0)
981 sender_address = address;
982
983 /* If the address was rewritten to <> no verification can be done, and we have
984 to return OK. This rewriting is permitted only for sender addresses; for other
985 addresses, such rewriting fails. */
986
987 if (address[0] == 0) return OK;
988
989 /* Save a copy of the sender address for re-instating if we change it to <>
990 while verifying a sender address (a nice bit of self-reference there). */
991
992 save_sender = sender_address;
993
994 /* Update the address structure with the possibly qualified and rewritten
995 address. Set it up as the starting address on the chain of new addresses. */
996
997 vaddr->address = address;
998 addr_new = vaddr;
999
1000 /* We need a loop, because an address can generate new addresses. We must also
1001 cope with generated pipes and files at the top level. (See also the code and
1002 comment in deliver.c.) However, it is usually the case that the router for
1003 user's .forward files has its verify flag turned off.
1004
1005 If an address generates more than one child, the loop is used only when
1006 full_info is set, and this can only be set locally. Remote enquiries just get
1007 information about the top level address, not anything that it generated. */
1008
1009 while (addr_new != NULL)
1010 {
1011 int rc;
1012 uschar *show_address;
1013 address_item *addr = addr_new;
1014
1015 addr_new = addr->next;
1016 addr->next = NULL;
1017
1018 /* When full_info is set, child addresses are displayed in top-level
1019 messages. Otherwise, we show only the top level address. */
1020
1021 show_address = full_info? addr->address : address;
1022
1023 DEBUG(D_verify)
1024 {
1025 debug_printf(">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>\n");
1026 debug_printf("Considering %s\n", addr->address);
1027 }
1028
1029 /* Handle generated pipe, file or reply addresses. We don't get these
1030 when handling EXPN, as it does only one level of expansion. */
1031
1032 if (testflag(addr, af_pfr))
1033 {
1034 allok = FALSE;
1035 if (f != NULL)
1036 {
1037 BOOL allow;
1038
1039 if (addr->address[0] == '>')
1040 {
1041 allow = testflag(addr, af_allow_reply);
1042 fprintf(f, "%s -> mail %s", addr->parent->address, addr->address + 1);
1043 }
1044 else
1045 {
1046 allow = (addr->address[0] == '|')?
1047 testflag(addr, af_allow_pipe) : testflag(addr, af_allow_file);
1048 fprintf(f, "%s -> %s", addr->parent->address, addr->address);
1049 }
1050
1051 if (addr->basic_errno == ERRNO_BADTRANSPORT)
1052 fprintf(f, "\n*** Error in setting up pipe, file, or autoreply:\n"
1053 "%s\n", addr->message);
1054 else if (allow)
1055 fprintf(f, "\n transport = %s\n", addr->transport->name);
1056 else
1057 fprintf(f, " *** forbidden ***\n");
1058 }
1059 continue;
1060 }
1061
1062 /* Just in case some router parameter refers to it. */
1063
1064 return_path = (addr->p.errors_address != NULL)?
1065 addr->p.errors_address : sender_address;
1066
1067 /* Split the address into domain and local part, handling the %-hack if
1068 necessary, and then route it. While routing a sender address, set
1069 $sender_address to <> because that is what it will be if we were trying to
1070 send a bounce to the sender. */
1071
1072 if (routed != NULL) *routed = FALSE;
1073 if ((rc = deliver_split_address(addr)) == OK)
1074 {
1075 if (!is_recipient) sender_address = null_sender;
1076 rc = route_address(addr, &addr_local, &addr_remote, &addr_new,
1077 &addr_succeed, verify_type);
1078 sender_address = save_sender; /* Put back the real sender */
1079 }
1080
1081 /* If routing an address succeeded, set the flag that remembers, for use when
1082 an ACL cached a sender verify (in case a callout fails). Then if routing set
1083 up a list of hosts or the transport has a host list, and the callout option
1084 is set, and we aren't in a host checking run, do the callout verification,
1085 and set another flag that notes that a callout happened. */
1086
1087 if (rc == OK)
1088 {
1089 if (routed != NULL) *routed = TRUE;
1090 if (callout > 0)
1091 {
1092 host_item *host_list = addr->host_list;
1093
1094 /* Make up some data for use in the case where there is no remote
1095 transport. */
1096
1097 transport_feedback tf = {
1098 NULL, /* interface (=> any) */
1099 US"smtp", /* port */
1100 US"smtp", /* protocol */
1101 NULL, /* hosts */
1102 US"$smtp_active_hostname", /* helo_data */
1103 FALSE, /* hosts_override */
1104 FALSE, /* hosts_randomize */
1105 FALSE, /* gethostbyname */
1106 TRUE, /* qualify_single */
1107 FALSE /* search_parents */
1108 };
1109
1110 /* If verification yielded a remote transport, we want to use that
1111 transport's options, so as to mimic what would happen if we were really
1112 sending a message to this address. */
1113
1114 if (addr->transport != NULL && !addr->transport->info->local)
1115 {
1116 (void)(addr->transport->setup)(addr->transport, addr, &tf, 0, 0, NULL);
1117
1118 /* If the transport has hosts and the router does not, or if the
1119 transport is configured to override the router's hosts, we must build a
1120 host list of the transport's hosts, and find the IP addresses */
1121
1122 if (tf.hosts != NULL && (host_list == NULL || tf.hosts_override))
1123 {
1124 uschar *s;
1125 uschar *save_deliver_domain = deliver_domain;
1126 uschar *save_deliver_localpart = deliver_localpart;
1127
1128 host_list = NULL; /* Ignore the router's hosts */
1129
1130 deliver_domain = addr->domain;
1131 deliver_localpart = addr->local_part;
1132 s = expand_string(tf.hosts);
1133 deliver_domain = save_deliver_domain;
1134 deliver_localpart = save_deliver_localpart;
1135
1136 if (s == NULL)
1137 {
1138 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand list of hosts "
1139 "\"%s\" in %s transport for callout: %s", tf.hosts,
1140 addr->transport->name, expand_string_message);
1141 }
1142 else
1143 {
1144 uschar *canonical_name;
1145 host_item *host, *nexthost;
1146 host_build_hostlist(&host_list, s, tf.hosts_randomize);
1147
1148 /* Just ignore failures to find a host address. If we don't manage
1149 to find any addresses, the callout will defer. Note that more than
1150 one address may be found for a single host, which will result in
1151 additional host items being inserted into the chain. Hence we must
1152 save the next host first. */
1153
1154 for (host = host_list; host != NULL; host = nexthost)
1155 {
1156 nexthost = host->next;
1157 if (tf.gethostbyname ||
1158 string_is_ip_address(host->name, NULL) != 0)
1159 (void)host_find_byname(host, NULL, &canonical_name, TRUE);
1160 else
1161 {
1162 int flags = HOST_FIND_BY_A;
1163 if (tf.qualify_single) flags |= HOST_FIND_QUALIFY_SINGLE;
1164 if (tf.search_parents) flags |= HOST_FIND_SEARCH_PARENTS;
1165 (void)host_find_bydns(host, NULL, flags, NULL, NULL, NULL,
1166 &canonical_name, NULL);
1167 }
1168 }
1169 }
1170 }
1171 }
1172
1173 /* Can only do a callout if we have at least one host! If the callout
1174 fails, it will have set ${sender,recipient}_verify_failure. */
1175
1176 if (host_list != NULL)
1177 {
1178 HDEBUG(D_verify) debug_printf("Attempting full verification using callout\n");
1179 if (host_checking && !host_checking_callout)
1180 {
1181 HDEBUG(D_verify)
1182 debug_printf("... callout omitted by default when host testing\n"
1183 "(Use -bhc if you want the callouts to happen.)\n");
1184 }
1185 else
1186 {
1187 rc = do_callout(addr, host_list, &tf, callout, callout_overall,
1188 callout_connect, options, se_mailfrom, pm_mailfrom);
1189 }
1190 }
1191 else
1192 {
1193 HDEBUG(D_verify) debug_printf("Cannot do callout: neither router nor "
1194 "transport provided a host list\n");
1195 }
1196 }
1197 }
1198
1199 /* Otherwise, any failure is a routing failure */
1200
1201 else *failure_ptr = US"route";
1202
1203 /* A router may return REROUTED if it has set up a child address as a result
1204 of a change of domain name (typically from widening). In this case we always
1205 want to continue to verify the new child. */
1206
1207 if (rc == REROUTED) continue;
1208
1209 /* Handle hard failures */
1210
1211 if (rc == FAIL)
1212 {
1213 allok = FALSE;
1214 if (f != NULL)
1215 {
1216 address_item *p = addr->parent;
1217
1218 fprintf(f, "%s%s %s", ko_prefix, show_address,
1219 address_test_mode? "is undeliverable" : "failed to verify");
1220 if (!expn && admin_user)
1221 {
1222 if (addr->basic_errno > 0)
1223 fprintf(f, ": %s", strerror(addr->basic_errno));
1224 if (addr->message != NULL)
1225 fprintf(f, ": %s", addr->message);
1226 }
1227
1228 /* Show parents iff doing full info */
1229
1230 if (full_info) while (p != NULL)
1231 {
1232 fprintf(f, "%s\n <-- %s", cr, p->address);
1233 p = p->parent;
1234 }
1235 fprintf(f, "%s\n", cr);
1236 }
1237
1238 if (!full_info) return copy_error(vaddr, addr, FAIL);
1239 else yield = FAIL;
1240 }
1241
1242 /* Soft failure */
1243
1244 else if (rc == DEFER)
1245 {
1246 allok = FALSE;
1247 if (f != NULL)
1248 {
1249 address_item *p = addr->parent;
1250 fprintf(f, "%s%s cannot be resolved at this time", ko_prefix,
1251 show_address);
1252 if (!expn && admin_user)
1253 {
1254 if (addr->basic_errno > 0)
1255 fprintf(f, ": %s", strerror(addr->basic_errno));
1256 if (addr->message != NULL)
1257 fprintf(f, ": %s", addr->message);
1258 else if (addr->basic_errno <= 0)
1259 fprintf(f, ": unknown error");
1260 }
1261
1262 /* Show parents iff doing full info */
1263
1264 if (full_info) while (p != NULL)
1265 {
1266 fprintf(f, "%s\n <-- %s", cr, p->address);
1267 p = p->parent;
1268 }
1269 fprintf(f, "%s\n", cr);
1270 }
1271
1272 if (!full_info) return copy_error(vaddr, addr, DEFER);
1273 else if (yield == OK) yield = DEFER;
1274 }
1275
1276 /* If we are handling EXPN, we do not want to continue to route beyond
1277 the top level (whose address is in "address"). */
1278
1279 else if (expn)
1280 {
1281 uschar *ok_prefix = US"250-";
1282 if (addr_new == NULL)
1283 {
1284 if (addr_local == NULL && addr_remote == NULL)
1285 fprintf(f, "250 mail to <%s> is discarded\r\n", address);
1286 else
1287 fprintf(f, "250 <%s>\r\n", address);
1288 }
1289 else while (addr_new != NULL)
1290 {
1291 address_item *addr2 = addr_new;
1292 addr_new = addr2->next;
1293 if (addr_new == NULL) ok_prefix = US"250 ";
1294 fprintf(f, "%s<%s>\r\n", ok_prefix, addr2->address);
1295 }
1296 return OK;
1297 }
1298
1299 /* Successful routing other than EXPN. */
1300
1301 else
1302 {
1303 /* Handle successful routing when short info wanted. Otherwise continue for
1304 other (generated) addresses. Short info is the operational case. Full info
1305 can be requested only when debug_selector != 0 and a file is supplied.
1306
1307 There is a conflict between the use of aliasing as an alternate email
1308 address, and as a sort of mailing list. If an alias turns the incoming
1309 address into just one address (e.g. J.Caesar->jc44) you may well want to
1310 carry on verifying the generated address to ensure it is valid when
1311 checking incoming mail. If aliasing generates multiple addresses, you
1312 probably don't want to do this. Exim therefore treats the generation of
1313 just a single new address as a special case, and continues on to verify the
1314 generated address. */
1315
1316 if (!full_info && /* Stop if short info wanted AND */
1317 (((addr_new == NULL || /* No new address OR */
1318 addr_new->next != NULL || /* More than one new address OR */
1319 testflag(addr_new, af_pfr))) /* New address is pfr */
1320 || /* OR */
1321 (addr_new != NULL && /* At least one new address AND */
1322 success_on_redirect))) /* success_on_redirect is set */
1323 {
1324 if (f != NULL) fprintf(f, "%s %s\n", show_address,
1325 address_test_mode? "is deliverable" : "verified");
1326
1327 /* If we have carried on to verify a child address, we want the value
1328 of $address_data to be that of the child */
1329
1330 vaddr->p.address_data = addr->p.address_data;
1331 return OK;
1332 }
1333 }
1334 } /* Loop for generated addresses */
1335
1336 /* Display the full results of the successful routing, including any generated
1337 addresses. Control gets here only when full_info is set, which requires f not
1338 to be NULL, and this occurs only when a top-level verify is called with the
1339 debugging switch on.
1340
1341 If there are no local and no remote addresses, and there were no pipes, files,
1342 or autoreplies, and there were no errors or deferments, the message is to be
1343 discarded, usually because of the use of :blackhole: in an alias file. */
1344
1345 if (allok && addr_local == NULL && addr_remote == NULL)
1346 {
1347 fprintf(f, "mail to %s is discarded\n", address);
1348 return yield;
1349 }
1350
1351 for (addr_list = addr_local, i = 0; i < 2; addr_list = addr_remote, i++)
1352 {
1353 while (addr_list != NULL)
1354 {
1355 address_item *addr = addr_list;
1356 address_item *p = addr->parent;
1357 addr_list = addr->next;
1358
1359 fprintf(f, "%s", CS addr->address);
1360 #ifdef EXPERIMENTAL_SRS
1361 if(addr->p.srs_sender)
1362 fprintf(f, " [srs = %s]", addr->p.srs_sender);
1363 #endif
1364
1365 /* If the address is a duplicate, show something about it. */
1366
1367 if (!testflag(addr, af_pfr))
1368 {
1369 tree_node *tnode;
1370 if ((tnode = tree_search(tree_duplicates, addr->unique)) != NULL)
1371 fprintf(f, " [duplicate, would not be delivered]");
1372 else tree_add_duplicate(addr->unique, addr);
1373 }
1374
1375 /* Now show its parents */
1376
1377 while (p != NULL)
1378 {
1379 fprintf(f, "\n <-- %s", p->address);
1380 p = p->parent;
1381 }
1382 fprintf(f, "\n ");
1383
1384 /* Show router, and transport */
1385
1386 fprintf(f, "router = %s, ", addr->router->name);
1387 fprintf(f, "transport = %s\n", (addr->transport == NULL)? US"unset" :
1388 addr->transport->name);
1389
1390 /* Show any hosts that are set up by a router unless the transport
1391 is going to override them; fiddle a bit to get a nice format. */
1392
1393 if (addr->host_list != NULL && addr->transport != NULL &&
1394 !addr->transport->overrides_hosts)
1395 {
1396 host_item *h;
1397 int maxlen = 0;
1398 int maxaddlen = 0;
1399 for (h = addr->host_list; h != NULL; h = h->next)
1400 {
1401 int len = Ustrlen(h->name);
1402 if (len > maxlen) maxlen = len;
1403 len = (h->address != NULL)? Ustrlen(h->address) : 7;
1404 if (len > maxaddlen) maxaddlen = len;
1405 }
1406 for (h = addr->host_list; h != NULL; h = h->next)
1407 {
1408 int len = Ustrlen(h->name);
1409 fprintf(f, " host %s ", h->name);
1410 while (len++ < maxlen) fprintf(f, " ");
1411 if (h->address != NULL)
1412 {
1413 fprintf(f, "[%s] ", h->address);
1414 len = Ustrlen(h->address);
1415 }
1416 else if (!addr->transport->info->local) /* Omit [unknown] for local */
1417 {
1418 fprintf(f, "[unknown] ");
1419 len = 7;
1420 }
1421 else len = -3;
1422 while (len++ < maxaddlen) fprintf(f," ");
1423 if (h->mx >= 0) fprintf(f, "MX=%d", h->mx);
1424 if (h->port != PORT_NONE) fprintf(f, " port=%d", h->port);
1425 if (h->status == hstatus_unusable) fprintf(f, " ** unusable **");
1426 fprintf(f, "\n");
1427 }
1428 }
1429 }
1430 }
1431
1432 /* Will be DEFER or FAIL if any one address has, only for full_info (which is
1433 the -bv or -bt case). */
1434
1435 return yield;
1436 }
1437
1438
1439
1440
1441 /*************************************************
1442 * Check headers for syntax errors *
1443 *************************************************/
1444
1445 /* This function checks those header lines that contain addresses, and verifies
1446 that all the addresses therein are syntactially correct.
1447
1448 Arguments:
1449 msgptr where to put an error message
1450
1451 Returns: OK
1452 FAIL
1453 */
1454
1455 int
1456 verify_check_headers(uschar **msgptr)
1457 {
1458 header_line *h;
1459 uschar *colon, *s;
1460
1461 for (h = header_list; h != NULL; h = h->next)
1462 {
1463 if (h->type != htype_from &&
1464 h->type != htype_reply_to &&
1465 h->type != htype_sender &&
1466 h->type != htype_to &&
1467 h->type != htype_cc &&
1468 h->type != htype_bcc)
1469 continue;
1470
1471 colon = Ustrchr(h->text, ':');
1472 s = colon + 1;
1473 while (isspace(*s)) s++;
1474
1475 parse_allow_group = TRUE; /* Allow group syntax */
1476
1477 /* Loop for multiple addresses in the header */
1478
1479 while (*s != 0)
1480 {
1481 uschar *ss = parse_find_address_end(s, FALSE);
1482 uschar *recipient, *errmess;
1483 int terminator = *ss;
1484 int start, end, domain;
1485
1486 /* Temporarily terminate the string at this point, and extract the
1487 operative address within. */
1488
1489 *ss = 0;
1490 recipient = parse_extract_address(s,&errmess,&start,&end,&domain,FALSE);
1491 *ss = terminator;
1492
1493 /* Permit an unqualified address only if the message is local, or if the
1494 sending host is configured to be permitted to send them. */
1495
1496 if (recipient != NULL && domain == 0)
1497 {
1498 if (h->type == htype_from || h->type == htype_sender)
1499 {
1500 if (!allow_unqualified_sender) recipient = NULL;
1501 }
1502 else
1503 {
1504 if (!allow_unqualified_recipient) recipient = NULL;
1505 }
1506 if (recipient == NULL) errmess = US"unqualified address not permitted";
1507 }
1508
1509 /* It's an error if no address could be extracted, except for the special
1510 case of an empty address. */
1511
1512 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
1513 {
1514 uschar *verb = US"is";
1515 uschar *t = ss;
1516 uschar *tt = colon;
1517 int len;
1518
1519 /* Arrange not to include any white space at the end in the
1520 error message or the header name. */
1521
1522 while (t > s && isspace(t[-1])) t--;
1523 while (tt > h->text && isspace(tt[-1])) tt--;
1524
1525 /* Add the address that failed to the error message, since in a
1526 header with very many addresses it is sometimes hard to spot
1527 which one is at fault. However, limit the amount of address to
1528 quote - cases have been seen where, for example, a missing double
1529 quote in a humungous To: header creates an "address" that is longer
1530 than string_sprintf can handle. */
1531
1532 len = t - s;
1533 if (len > 1024)
1534 {
1535 len = 1024;
1536 verb = US"begins";
1537 }
1538
1539 *msgptr = string_printing(
1540 string_sprintf("%s: failing address in \"%.*s:\" header %s: %.*s",
1541 errmess, tt - h->text, h->text, verb, len, s));
1542
1543 return FAIL;
1544 }
1545
1546 /* Advance to the next address */
1547
1548 s = ss + (terminator? 1:0);
1549 while (isspace(*s)) s++;
1550 } /* Next address */
1551 } /* Next header */
1552
1553 return OK;
1554 }
1555
1556
1557
1558 /*************************************************
1559 * Check for blind recipients *
1560 *************************************************/
1561
1562 /* This function checks that every (envelope) recipient is mentioned in either
1563 the To: or Cc: header lines, thus detecting blind carbon copies.
1564
1565 There are two ways of scanning that could be used: either scan the header lines
1566 and tick off the recipients, or scan the recipients and check the header lines.
1567 The original proposed patch did the former, but I have chosen to do the latter,
1568 because (a) it requires no memory and (b) will use fewer resources when there
1569 are many addresses in To: and/or Cc: and only one or two envelope recipients.
1570
1571 Arguments: none
1572 Returns: OK if there are no blind recipients
1573 FAIL if there is at least one blind recipient
1574 */
1575
1576 int
1577 verify_check_notblind(void)
1578 {
1579 int i;
1580 for (i = 0; i < recipients_count; i++)
1581 {
1582 header_line *h;
1583 BOOL found = FALSE;
1584 uschar *address = recipients_list[i].address;
1585
1586 for (h = header_list; !found && h != NULL; h = h->next)
1587 {
1588 uschar *colon, *s;
1589
1590 if (h->type != htype_to && h->type != htype_cc) continue;
1591
1592 colon = Ustrchr(h->text, ':');
1593 s = colon + 1;
1594 while (isspace(*s)) s++;
1595
1596 parse_allow_group = TRUE; /* Allow group syntax */
1597
1598 /* Loop for multiple addresses in the header */
1599
1600 while (*s != 0)
1601 {
1602 uschar *ss = parse_find_address_end(s, FALSE);
1603 uschar *recipient,*errmess;
1604 int terminator = *ss;
1605 int start, end, domain;
1606
1607 /* Temporarily terminate the string at this point, and extract the
1608 operative address within. */
1609
1610 *ss = 0;
1611 recipient = parse_extract_address(s,&errmess,&start,&end,&domain,FALSE);
1612 *ss = terminator;
1613
1614 /* If we found a valid recipient that has a domain, compare it with the
1615 envelope recipient. Local parts are compared case-sensitively, domains
1616 case-insensitively. By comparing from the start with length "domain", we
1617 include the "@" at the end, which ensures that we are comparing the whole
1618 local part of each address. */
1619
1620 if (recipient != NULL && domain != 0)
1621 {
1622 found = Ustrncmp(recipient, address, domain) == 0 &&
1623 strcmpic(recipient + domain, address + domain) == 0;
1624 if (found) break;
1625 }
1626
1627 /* Advance to the next address */
1628
1629 s = ss + (terminator? 1:0);
1630 while (isspace(*s)) s++;
1631 } /* Next address */
1632 } /* Next header (if found is false) */
1633
1634 if (!found) return FAIL;
1635 } /* Next recipient */
1636
1637 return OK;
1638 }
1639
1640
1641
1642 /*************************************************
1643 * Find if verified sender *
1644 *************************************************/
1645
1646 /* Usually, just a single address is verified as the sender of the message.
1647 However, Exim can be made to verify other addresses as well (often related in
1648 some way), and this is useful in some environments. There may therefore be a
1649 chain of such addresses that have previously been tested. This function finds
1650 whether a given address is on the chain.
1651
1652 Arguments: the address to be verified
1653 Returns: pointer to an address item, or NULL
1654 */
1655
1656 address_item *
1657 verify_checked_sender(uschar *sender)
1658 {
1659 address_item *addr;
1660 for (addr = sender_verified_list; addr != NULL; addr = addr->next)
1661 if (Ustrcmp(sender, addr->address) == 0) break;
1662 return addr;
1663 }
1664
1665
1666
1667
1668
1669 /*************************************************
1670 * Get valid header address *
1671 *************************************************/
1672
1673 /* Scan the originator headers of the message, looking for an address that
1674 verifies successfully. RFC 822 says:
1675
1676 o The "Sender" field mailbox should be sent notices of
1677 any problems in transport or delivery of the original
1678 messages. If there is no "Sender" field, then the
1679 "From" field mailbox should be used.
1680
1681 o If the "Reply-To" field exists, then the reply should
1682 go to the addresses indicated in that field and not to
1683 the address(es) indicated in the "From" field.
1684
1685 So we check a Sender field if there is one, else a Reply_to field, else a From
1686 field. As some strange messages may have more than one of these fields,
1687 especially if they are resent- fields, check all of them if there is more than
1688 one.
1689
1690 Arguments:
1691 user_msgptr points to where to put a user error message
1692 log_msgptr points to where to put a log error message
1693 callout timeout for callout check (passed to verify_address())
1694 callout_overall overall callout timeout (ditto)
1695 callout_connect connect callout timeout (ditto)
1696 se_mailfrom mailfrom for verify; NULL => ""
1697 pm_mailfrom sender for pm callout check (passed to verify_address())
1698 options callout options (passed to verify_address())
1699 verrno where to put the address basic_errno
1700
1701 If log_msgptr is set to something without setting user_msgptr, the caller
1702 normally uses log_msgptr for both things.
1703
1704 Returns: result of the verification attempt: OK, FAIL, or DEFER;
1705 FAIL is given if no appropriate headers are found
1706 */
1707
1708 int
1709 verify_check_header_address(uschar **user_msgptr, uschar **log_msgptr,
1710 int callout, int callout_overall, int callout_connect, uschar *se_mailfrom,
1711 uschar *pm_mailfrom, int options, int *verrno)
1712 {
1713 static int header_types[] = { htype_sender, htype_reply_to, htype_from };
1714 int yield = FAIL;
1715 int i;
1716
1717 for (i = 0; i < 3; i++)
1718 {
1719 header_line *h;
1720 for (h = header_list; h != NULL; h = h->next)
1721 {
1722 int terminator, new_ok;
1723 uschar *s, *ss, *endname;
1724
1725 if (h->type != header_types[i]) continue;
1726 s = endname = Ustrchr(h->text, ':') + 1;
1727
1728 while (*s != 0)
1729 {
1730 address_item *vaddr;
1731
1732 while (isspace(*s) || *s == ',') s++;
1733 if (*s == 0) break; /* End of header */
1734
1735 ss = parse_find_address_end(s, FALSE);
1736
1737 /* The terminator is a comma or end of header, but there may be white
1738 space preceding it (including newline for the last address). Move back
1739 past any white space so we can check against any cached envelope sender
1740 address verifications. */
1741
1742 while (isspace(ss[-1])) ss--;
1743 terminator = *ss;
1744 *ss = 0;
1745
1746 HDEBUG(D_verify) debug_printf("verifying %.*s header address %s\n",
1747 (int)(endname - h->text), h->text, s);
1748
1749 /* See if we have already verified this address as an envelope sender,
1750 and if so, use the previous answer. */
1751
1752 vaddr = verify_checked_sender(s);
1753
1754 if (vaddr != NULL && /* Previously checked */
1755 (callout <= 0 || /* No callout needed; OR */
1756 vaddr->special_action > 256)) /* Callout was done */
1757 {
1758 new_ok = vaddr->special_action & 255;
1759 HDEBUG(D_verify) debug_printf("previously checked as envelope sender\n");
1760 *ss = terminator; /* Restore shortened string */
1761 }
1762
1763 /* Otherwise we run the verification now. We must restore the shortened
1764 string before running the verification, so the headers are correct, in
1765 case there is any rewriting. */
1766
1767 else
1768 {
1769 int start, end, domain;
1770 uschar *address = parse_extract_address(s, log_msgptr, &start,
1771 &end, &domain, FALSE);
1772
1773 *ss = terminator;
1774
1775 /* If verification failed because of a syntax error, fail this
1776 function, and ensure that the failing address gets added to the error
1777 message. */
1778
1779 if (address == NULL)
1780 {
1781 new_ok = FAIL;
1782 if (*log_msgptr != NULL)
1783 {
1784 while (ss > s && isspace(ss[-1])) ss--;
1785 *log_msgptr = string_sprintf("syntax error in '%.*s' header when "
1786 "scanning for sender: %s in \"%.*s\"",
1787 endname - h->text, h->text, *log_msgptr, ss - s, s);
1788 return FAIL;
1789 }
1790 }
1791
1792 /* Else go ahead with the sender verification. But it isn't *the*
1793 sender of the message, so set vopt_fake_sender to stop sender_address
1794 being replaced after rewriting or qualification. */
1795
1796 else
1797 {
1798 vaddr = deliver_make_addr(address, FALSE);
1799 new_ok = verify_address(vaddr, NULL, options | vopt_fake_sender,
1800 callout, callout_overall, callout_connect, se_mailfrom,
1801 pm_mailfrom, NULL);
1802 }
1803 }
1804
1805 /* We now have the result, either newly found, or cached. If we are
1806 giving out error details, set a specific user error. This means that the
1807 last of these will be returned to the user if all three fail. We do not
1808 set a log message - the generic one below will be used. */
1809
1810 if (new_ok != OK)
1811 {
1812 *verrno = vaddr->basic_errno;
1813 if (smtp_return_error_details)
1814 {
1815 *user_msgptr = string_sprintf("Rejected after DATA: "
1816 "could not verify \"%.*s\" header address\n%s: %s",
1817 endname - h->text, h->text, vaddr->address, vaddr->message);
1818 }
1819 }
1820
1821 /* Success or defer */
1822
1823 if (new_ok == OK) return OK;
1824 if (new_ok == DEFER) yield = DEFER;
1825
1826 /* Move on to any more addresses in the header */
1827
1828 s = ss;
1829 }
1830 }
1831 }
1832
1833 if (yield == FAIL && *log_msgptr == NULL)
1834 *log_msgptr = US"there is no valid sender in any header line";
1835
1836 if (yield == DEFER && *log_msgptr == NULL)
1837 *log_msgptr = US"all attempts to verify a sender in a header line deferred";
1838
1839 return yield;
1840 }
1841
1842
1843
1844
1845 /*************************************************
1846 * Get RFC 1413 identification *
1847 *************************************************/
1848
1849 /* Attempt to get an id from the sending machine via the RFC 1413 protocol. If
1850 the timeout is set to zero, then the query is not done. There may also be lists
1851 of hosts and nets which are exempt. To guard against malefactors sending
1852 non-printing characters which could, for example, disrupt a message's headers,
1853 make sure the string consists of printing characters only.
1854
1855 Argument:
1856 port the port to connect to; usually this is IDENT_PORT (113), but when
1857 running in the test harness with -bh a different value is used.
1858
1859 Returns: nothing
1860
1861 Side effect: any received ident value is put in sender_ident (NULL otherwise)
1862 */
1863
1864 void
1865 verify_get_ident(int port)
1866 {
1867 int sock, host_af, qlen;
1868 int received_sender_port, received_interface_port, n;
1869 uschar *p;
1870 uschar buffer[2048];
1871
1872 /* Default is no ident. Check whether we want to do an ident check for this
1873 host. */
1874
1875 sender_ident = NULL;
1876 if (rfc1413_query_timeout <= 0 || verify_check_host(&rfc1413_hosts) != OK)
1877 return;
1878
1879 DEBUG(D_ident) debug_printf("doing ident callback\n");
1880
1881 /* Set up a connection to the ident port of the remote host. Bind the local end
1882 to the incoming interface address. If the sender host address is an IPv6
1883 address, the incoming interface address will also be IPv6. */
1884
1885 host_af = (Ustrchr(sender_host_address, ':') == NULL)? AF_INET : AF_INET6;
1886 sock = ip_socket(SOCK_STREAM, host_af);
1887 if (sock < 0) return;
1888
1889 if (ip_bind(sock, host_af, interface_address, 0) < 0)
1890 {
1891 DEBUG(D_ident) debug_printf("bind socket for ident failed: %s\n",
1892 strerror(errno));
1893 goto END_OFF;
1894 }
1895
1896 if (ip_connect(sock, host_af, sender_host_address, port, rfc1413_query_timeout)
1897 < 0)
1898 {
1899 if (errno == ETIMEDOUT && (log_extra_selector & LX_ident_timeout) != 0)
1900 {
1901 log_write(0, LOG_MAIN, "ident connection to %s timed out",
1902 sender_host_address);
1903 }
1904 else
1905 {
1906 DEBUG(D_ident) debug_printf("ident connection to %s failed: %s\n",
1907 sender_host_address, strerror(errno));
1908 }
1909 goto END_OFF;
1910 }
1911
1912 /* Construct and send the query. */
1913
1914 sprintf(CS buffer, "%d , %d\r\n", sender_host_port, interface_port);
1915 qlen = Ustrlen(buffer);
1916 if (send(sock, buffer, qlen, 0) < 0)
1917 {
1918 DEBUG(D_ident) debug_printf("ident send failed: %s\n", strerror(errno));
1919 goto END_OFF;
1920 }
1921
1922 /* Read a response line. We put it into the rest of the buffer, using several
1923 recv() calls if necessary. */
1924
1925 p = buffer + qlen;
1926
1927 for (;;)
1928 {
1929 uschar *pp;
1930 int count;
1931 int size = sizeof(buffer) - (p - buffer);
1932
1933 if (size <= 0) goto END_OFF; /* Buffer filled without seeing \n. */
1934 count = ip_recv(sock, p, size, rfc1413_query_timeout);
1935 if (count <= 0) goto END_OFF; /* Read error or EOF */
1936
1937 /* Scan what we just read, to see if we have reached the terminating \r\n. Be
1938 generous, and accept a plain \n terminator as well. The only illegal
1939 character is 0. */
1940
1941 for (pp = p; pp < p + count; pp++)
1942 {
1943 if (*pp == 0) goto END_OFF; /* Zero octet not allowed */
1944 if (*pp == '\n')
1945 {
1946 if (pp[-1] == '\r') pp--;
1947 *pp = 0;
1948 goto GOT_DATA; /* Break out of both loops */
1949 }
1950 }
1951
1952 /* Reached the end of the data without finding \n. Let the loop continue to
1953 read some more, if there is room. */
1954
1955 p = pp;
1956 }
1957
1958 GOT_DATA:
1959
1960 /* We have received a line of data. Check it carefully. It must start with the
1961 same two port numbers that we sent, followed by data as defined by the RFC. For
1962 example,
1963
1964 12345 , 25 : USERID : UNIX :root
1965
1966 However, the amount of white space may be different to what we sent. In the
1967 "osname" field there may be several sub-fields, comma separated. The data we
1968 actually want to save follows the third colon. Some systems put leading spaces
1969 in it - we discard those. */
1970
1971 if (sscanf(CS buffer + qlen, "%d , %d%n", &received_sender_port,
1972 &received_interface_port, &n) != 2 ||
1973 received_sender_port != sender_host_port ||
1974 received_interface_port != interface_port)
1975 goto END_OFF;
1976
1977 p = buffer + qlen + n;
1978 while(isspace(*p)) p++;
1979 if (*p++ != ':') goto END_OFF;
1980 while(isspace(*p)) p++;
1981 if (Ustrncmp(p, "USERID", 6) != 0) goto END_OFF;
1982 p += 6;
1983 while(isspace(*p)) p++;
1984 if (*p++ != ':') goto END_OFF;
1985 while (*p != 0 && *p != ':') p++;
1986 if (*p++ == 0) goto END_OFF;
1987 while(isspace(*p)) p++;
1988 if (*p == 0) goto END_OFF;
1989
1990 /* The rest of the line is the data we want. We turn it into printing
1991 characters when we save it, so that it cannot mess up the format of any logging
1992 or Received: lines into which it gets inserted. We keep a maximum of 127
1993 characters. */
1994
1995 sender_ident = string_printing(string_copyn(p, 127));
1996 DEBUG(D_ident) debug_printf("sender_ident = %s\n", sender_ident);
1997
1998 END_OFF:
1999 (void)close(sock);
2000 return;
2001 }
2002
2003
2004
2005
2006 /*************************************************
2007 * Match host to a single host-list item *
2008 *************************************************/
2009
2010 /* This function compares a host (name or address) against a single item
2011 from a host list. The host name gets looked up if it is needed and is not
2012 already known. The function is called from verify_check_this_host() via
2013 match_check_list(), which is why most of its arguments are in a single block.
2014
2015 Arguments:
2016 arg the argument block (see below)
2017 ss the host-list item
2018 valueptr where to pass back looked up data, or NULL
2019 error for error message when returning ERROR
2020
2021 The block contains:
2022 host_name (a) the host name, or
2023 (b) NULL, implying use sender_host_name and
2024 sender_host_aliases, looking them up if required, or
2025 (c) the empty string, meaning that only IP address matches
2026 are permitted
2027 host_address the host address
2028 host_ipv4 the IPv4 address taken from an IPv6 one
2029
2030 Returns: OK matched
2031 FAIL did not match
2032 DEFER lookup deferred
2033 ERROR (a) failed to find the host name or IP address, or
2034 (b) unknown lookup type specified, or
2035 (c) host name encountered when only IP addresses are
2036 being matched
2037 */
2038
2039 int
2040 check_host(void *arg, uschar *ss, uschar **valueptr, uschar **error)
2041 {
2042 check_host_block *cb = (check_host_block *)arg;
2043 int mlen = -1;
2044 int maskoffset;
2045 BOOL iplookup = FALSE;
2046 BOOL isquery = FALSE;
2047 BOOL isiponly = cb->host_name != NULL && cb->host_name[0] == 0;
2048 uschar *t;
2049 uschar *semicolon;
2050 uschar **aliases;
2051
2052 /* Optimize for the special case when the pattern is "*". */
2053
2054 if (*ss == '*' && ss[1] == 0) return OK;
2055
2056 /* If the pattern is empty, it matches only in the case when there is no host -
2057 this can occur in ACL checking for SMTP input using the -bs option. In this
2058 situation, the host address is the empty string. */
2059
2060 if (cb->host_address[0] == 0) return (*ss == 0)? OK : FAIL;
2061 if (*ss == 0) return FAIL;
2062
2063 /* If the pattern is precisely "@" then match against the primary host name,
2064 provided that host name matching is permitted; if it's "@[]" match against the
2065 local host's IP addresses. */
2066
2067 if (*ss == '@')
2068 {
2069 if (ss[1] == 0)
2070 {
2071 if (isiponly) return ERROR;
2072 ss = primary_hostname;
2073 }
2074 else if (Ustrcmp(ss, "@[]") == 0)
2075 {
2076 ip_address_item *ip;
2077 for (ip = host_find_interfaces(); ip != NULL; ip = ip->next)
2078 if (Ustrcmp(ip->address, cb->host_address) == 0) return OK;
2079 return FAIL;
2080 }
2081 }
2082
2083 /* If the pattern is an IP address, optionally followed by a bitmask count, do
2084 a (possibly masked) comparision with the current IP address. */
2085
2086 if (string_is_ip_address(ss, &maskoffset) != 0)
2087 return (host_is_in_net(cb->host_address, ss, maskoffset)? OK : FAIL);
2088
2089 /* The pattern is not an IP address. A common error that people make is to omit
2090 one component of an IPv4 address, either by accident, or believing that, for
2091 example, 1.2.3/24 is the same as 1.2.3.0/24, or 1.2.3 is the same as 1.2.3.0,
2092 which it isn't. (Those applications that do accept 1.2.3 as an IP address
2093 interpret it as 1.2.0.3 because the final component becomes 16-bit - this is an
2094 ancient specification.) To aid in debugging these cases, we give a specific
2095 error if the pattern contains only digits and dots or contains a slash preceded
2096 only by digits and dots (a slash at the start indicates a file name and of
2097 course slashes may be present in lookups, but not preceded only by digits and
2098 dots). */
2099
2100 for (t = ss; isdigit(*t) || *t == '.'; t++);
2101 if (*t == 0 || (*t == '/' && t != ss))
2102 {
2103 *error = US"malformed IPv4 address or address mask";
2104 return ERROR;
2105 }
2106
2107 /* See if there is a semicolon in the pattern */
2108
2109 semicolon = Ustrchr(ss, ';');
2110
2111 /* If we are doing an IP address only match, then all lookups must be IP
2112 address lookups, even if there is no "net-". */
2113
2114 if (isiponly)
2115 {
2116 iplookup = semicolon != NULL;
2117 }
2118
2119 /* Otherwise, if the item is of the form net[n]-lookup;<file|query> then it is
2120 a lookup on a masked IP network, in textual form. We obey this code even if we
2121 have already set iplookup, so as to skip over the "net-" prefix and to set the
2122 mask length. The net- stuff really only applies to single-key lookups where the
2123 key is implicit. For query-style lookups the key is specified in the query.
2124 From release 4.30, the use of net- for query style is no longer needed, but we
2125 retain it for backward compatibility. */
2126
2127 if (Ustrncmp(ss, "net", 3) == 0 && semicolon != NULL)
2128 {
2129 mlen = 0;
2130 for (t = ss + 3; isdigit(*t); t++) mlen = mlen * 10 + *t - '0';
2131 if (mlen == 0 && t == ss+3) mlen = -1; /* No mask supplied */
2132 iplookup = (*t++ == '-');
2133 }
2134 else t = ss;
2135
2136 /* Do the IP address lookup if that is indeed what we have */
2137
2138 if (iplookup)
2139 {
2140 int insize;
2141 int search_type;
2142 int incoming[4];
2143 void *handle;
2144 uschar *filename, *key, *result;
2145 uschar buffer[64];
2146
2147 /* Find the search type */
2148
2149 search_type = search_findtype(t, semicolon - t);
2150
2151 if (search_type < 0) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
2152 search_error_message);
2153
2154 /* Adjust parameters for the type of lookup. For a query-style lookup, there
2155 is no file name, and the "key" is just the query. For query-style with a file
2156 name, we have to fish the file off the start of the query. For a single-key
2157 lookup, the key is the current IP address, masked appropriately, and
2158 reconverted to text form, with the mask appended. For IPv6 addresses, specify
2159 dot separators instead of colons. */
2160
2161 if (mac_islookup(search_type, lookup_absfilequery))
2162 {
2163 filename = semicolon + 1;
2164 key = filename;
2165 while (*key != 0 && !isspace(*key)) key++;
2166 filename = string_copyn(filename, key - filename);
2167 while (isspace(*key)) key++;
2168 }
2169 else if (mac_islookup(search_type, lookup_querystyle))
2170 {
2171 filename = NULL;
2172 key = semicolon + 1;
2173 }
2174 else
2175 {
2176 insize = host_aton(cb->host_address, incoming);
2177 host_mask(insize, incoming, mlen);
2178 (void)host_nmtoa(insize, incoming, mlen, buffer, '.');
2179 key = buffer;
2180 filename = semicolon + 1;
2181 }
2182
2183 /* Now do the actual lookup; note that there is no search_close() because
2184 of the caching arrangements. */
2185
2186 handle = search_open(filename, search_type, 0, NULL, NULL);
2187 if (handle == NULL) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
2188 search_error_message);
2189 result = search_find(handle, filename, key, -1, NULL, 0, 0, NULL);
2190 if (valueptr != NULL) *valueptr = result;
2191 return (result != NULL)? OK : search_find_defer? DEFER: FAIL;
2192 }
2193
2194 /* The pattern is not an IP address or network reference of any kind. That is,
2195 it is a host name pattern. If this is an IP only match, there's an error in the
2196 host list. */
2197
2198 if (isiponly)
2199 {
2200 *error = US"cannot match host name in match_ip list";
2201 return ERROR;
2202 }
2203
2204 /* Check the characters of the pattern to see if they comprise only letters,
2205 digits, full stops, and hyphens (the constituents of domain names). Allow
2206 underscores, as they are all too commonly found. Sigh. Also, if
2207 allow_utf8_domains is set, allow top-bit characters. */
2208
2209 for (t = ss; *t != 0; t++)
2210 if (!isalnum(*t) && *t != '.' && *t != '-' && *t != '_' &&
2211 (!allow_utf8_domains || *t < 128)) break;
2212
2213 /* If the pattern is a complete domain name, with no fancy characters, look up
2214 its IP address and match against that. Note that a multi-homed host will add
2215 items to the chain. */
2216
2217 if (*t == 0)
2218 {
2219 int rc;
2220 host_item h;
2221 h.next = NULL;
2222 h.name = ss;
2223 h.address = NULL;
2224 h.mx = MX_NONE;
2225
2226 rc = host_find_byname(&h, NULL, NULL, FALSE);
2227 if (rc == HOST_FOUND || rc == HOST_FOUND_LOCAL)
2228 {
2229 host_item *hh;
2230 for (hh = &h; hh != NULL; hh = hh->next)
2231 {
2232 if (host_is_in_net(hh->address, cb->host_address, 0)) return OK;
2233 }
2234 return FAIL;
2235 }
2236 if (rc == HOST_FIND_AGAIN) return DEFER;
2237 *error = string_sprintf("failed to find IP address for %s", ss);
2238 return ERROR;
2239 }
2240
2241 /* Almost all subsequent comparisons require the host name, and can be done
2242 using the general string matching function. When this function is called for
2243 outgoing hosts, the name is always given explicitly. If it is NULL, it means we
2244 must use sender_host_name and its aliases, looking them up if necessary. */
2245
2246 if (cb->host_name != NULL) /* Explicit host name given */
2247 return match_check_string(cb->host_name, ss, -1, TRUE, TRUE, TRUE,
2248 valueptr);
2249
2250 /* Host name not given; in principle we need the sender host name and its
2251 aliases. However, for query-style lookups, we do not need the name if the
2252 query does not contain $sender_host_name. From release 4.23, a reference to
2253 $sender_host_name causes it to be looked up, so we don't need to do the lookup
2254 on spec. */
2255
2256 if ((semicolon = Ustrchr(ss, ';')) != NULL)
2257 {
2258 uschar *affix;
2259 int partial, affixlen, starflags, id;
2260
2261 *semicolon = 0;
2262 id = search_findtype_partial(ss, &partial, &affix, &affixlen, &starflags);
2263 *semicolon=';';
2264
2265 if (id < 0) /* Unknown lookup type */
2266 {
2267 log_write(0, LOG_MAIN|LOG_PANIC, "%s in host list item \"%s\"",
2268 search_error_message, ss);
2269 return DEFER;
2270 }
2271 isquery = mac_islookup(id, lookup_querystyle|lookup_absfilequery);
2272 }
2273
2274 if (isquery)
2275 {
2276 switch(match_check_string(US"", ss, -1, TRUE, TRUE, TRUE, valueptr))
2277 {
2278 case OK: return OK;
2279 case DEFER: return DEFER;
2280 default: return FAIL;
2281 }
2282 }
2283
2284 /* Not a query-style lookup; must ensure the host name is present, and then we
2285 do a check on the name and all its aliases. */
2286
2287 if (sender_host_name == NULL)
2288 {
2289 HDEBUG(D_host_lookup)
2290 debug_printf("sender host name required, to match against %s\n", ss);
2291 if (host_lookup_failed || host_name_lookup() != OK)
2292 {
2293 *error = string_sprintf("failed to find host name for %s",
2294 sender_host_address);;
2295 return ERROR;
2296 }
2297 host_build_sender_fullhost();
2298 }
2299
2300 /* Match on the sender host name, using the general matching function */
2301
2302 switch(match_check_string(sender_host_name, ss, -1, TRUE, TRUE, TRUE,
2303 valueptr))
2304 {
2305 case OK: return OK;
2306 case DEFER: return DEFER;
2307 }
2308
2309 /* If there are aliases, try matching on them. */
2310
2311 aliases = sender_host_aliases;
2312 while (*aliases != NULL)
2313 {
2314 switch(match_check_string(*aliases++, ss, -1, TRUE, TRUE, TRUE, valueptr))
2315 {
2316 case OK: return OK;
2317 case DEFER: return DEFER;
2318 }
2319 }
2320 return FAIL;
2321 }
2322
2323
2324
2325
2326 /*************************************************
2327 * Check a specific host matches a host list *
2328 *************************************************/
2329
2330 /* This function is passed a host list containing items in a number of
2331 different formats and the identity of a host. Its job is to determine whether
2332 the given host is in the set of hosts defined by the list. The host name is
2333 passed as a pointer so that it can be looked up if needed and not already
2334 known. This is commonly the case when called from verify_check_host() to check
2335 an incoming connection. When called from elsewhere the host name should usually
2336 be set.
2337
2338 This function is now just a front end to match_check_list(), which runs common
2339 code for scanning a list. We pass it the check_host() function to perform a
2340 single test.
2341
2342 Arguments:
2343 listptr pointer to the host list
2344 cache_bits pointer to cache for named lists, or NULL
2345 host_name the host name or NULL, implying use sender_host_name and
2346 sender_host_aliases, looking them up if required
2347 host_address the IP address
2348 valueptr if not NULL, data from a lookup is passed back here
2349
2350 Returns: OK if the host is in the defined set
2351 FAIL if the host is not in the defined set,
2352 DEFER if a data lookup deferred (not a host lookup)
2353
2354 If the host name was needed in order to make a comparison, and could not be
2355 determined from the IP address, the result is FAIL unless the item
2356 "+allow_unknown" was met earlier in the list, in which case OK is returned. */
2357
2358 int
2359 verify_check_this_host(uschar **listptr, unsigned int *cache_bits,
2360 uschar *host_name, uschar *host_address, uschar **valueptr)
2361 {
2362 int rc;
2363 unsigned int *local_cache_bits = cache_bits;
2364 uschar *save_host_address = deliver_host_address;
2365 check_host_block cb;
2366 cb.host_name = host_name;
2367 cb.host_address = host_address;
2368
2369 if (valueptr != NULL) *valueptr = NULL;
2370
2371 /* If the host address starts off ::ffff: it is an IPv6 address in
2372 IPv4-compatible mode. Find the IPv4 part for checking against IPv4
2373 addresses. */
2374
2375 cb.host_ipv4 = (Ustrncmp(host_address, "::ffff:", 7) == 0)?
2376 host_address + 7 : host_address;
2377
2378 /* During the running of the check, put the IP address into $host_address. In
2379 the case of calls from the smtp transport, it will already be there. However,
2380 in other calls (e.g. when testing ignore_target_hosts), it won't. Just to be on
2381 the safe side, any existing setting is preserved, though as I write this
2382 (November 2004) I can't see any cases where it is actually needed. */
2383
2384 deliver_host_address = host_address;
2385 rc = match_check_list(
2386 listptr, /* the list */
2387 0, /* separator character */
2388 &hostlist_anchor, /* anchor pointer */
2389 &local_cache_bits, /* cache pointer */
2390 check_host, /* function for testing */
2391 &cb, /* argument for function */
2392 MCL_HOST, /* type of check */
2393 (host_address == sender_host_address)?
2394 US"host" : host_address, /* text for debugging */
2395 valueptr); /* where to pass back data */
2396 deliver_host_address = save_host_address;
2397 return rc;
2398 }
2399
2400
2401
2402
2403 /*************************************************
2404 * Check the remote host matches a list *
2405 *************************************************/
2406
2407 /* This is a front end to verify_check_this_host(), created because checking
2408 the remote host is a common occurrence. With luck, a good compiler will spot
2409 the tail recursion and optimize it. If there's no host address, this is
2410 command-line SMTP input - check against an empty string for the address.
2411
2412 Arguments:
2413 listptr pointer to the host list
2414
2415 Returns: the yield of verify_check_this_host(),
2416 i.e. OK, FAIL, or DEFER
2417 */
2418
2419 int
2420 verify_check_host(uschar **listptr)
2421 {
2422 return verify_check_this_host(listptr, sender_host_cache, NULL,
2423 (sender_host_address == NULL)? US"" : sender_host_address, NULL);
2424 }
2425
2426
2427
2428
2429
2430 /*************************************************
2431 * Invert an IP address for a DNS black list *
2432 *************************************************/
2433
2434 /*
2435 Arguments:
2436 buffer where to put the answer
2437 address the address to invert
2438 */
2439
2440 static void
2441 invert_address(uschar *buffer, uschar *address)
2442 {
2443 int bin[4];
2444 uschar *bptr = buffer;
2445
2446 /* If this is an IPv4 address mapped into IPv6 format, adjust the pointer
2447 to the IPv4 part only. */
2448
2449 if (Ustrncmp(address, "::ffff:", 7) == 0) address += 7;
2450
2451 /* Handle IPv4 address: when HAVE_IPV6 is false, the result of host_aton() is
2452 always 1. */
2453
2454 if (host_aton(address, bin) == 1)
2455 {
2456 int i;
2457 int x = bin[0];
2458 for (i = 0; i < 4; i++)
2459 {
2460 sprintf(CS bptr, "%d.", x & 255);
2461 while (*bptr) bptr++;
2462 x >>= 8;
2463 }
2464 }
2465
2466 /* Handle IPv6 address. Actually, as far as I know, there are no IPv6 addresses
2467 in any DNS black lists, and the format in which they will be looked up is
2468 unknown. This is just a guess. */
2469
2470 #if HAVE_IPV6
2471 else
2472 {
2473 int i, j;
2474 for (j = 3; j >= 0; j--)
2475 {
2476 int x = bin[j];
2477 for (i = 0; i < 8; i++)
2478 {
2479 sprintf(CS bptr, "%x.", x & 15);
2480 while (*bptr) bptr++;
2481 x >>= 4;
2482 }
2483 }
2484 }
2485 #endif
2486
2487 /* Remove trailing period -- this is needed so that both arbitrary
2488 dnsbl keydomains and inverted addresses may be combined with the
2489 same format string, "%s.%s" */
2490
2491 *(--bptr) = 0;
2492 }
2493
2494
2495
2496 /*************************************************
2497 * Perform a single dnsbl lookup *
2498 *************************************************/
2499
2500 /* This function is called from verify_check_dnsbl() below. It is also called
2501 recursively from within itself when domain and domain_txt are different
2502 pointers, in order to get the TXT record from the alternate domain.
2503
2504 Arguments:
2505 domain the outer dnsbl domain
2506 domain_txt alternate domain to lookup TXT record on success; when the
2507 same domain is to be used, domain_txt == domain (that is,
2508 the pointers must be identical, not just the text)
2509 keydomain the current keydomain (for debug message)
2510 prepend subdomain to lookup (like keydomain, but
2511 reversed if IP address)
2512 iplist the list of matching IP addresses, or NULL for "any"
2513 bitmask true if bitmask matching is wanted
2514 invert_result true if result to be inverted
2515 defer_return what to return for a defer
2516
2517 Returns: OK if lookup succeeded
2518 FAIL if not
2519 */
2520
2521 static int
2522 one_check_dnsbl(uschar *domain, uschar *domain_txt, uschar *keydomain,
2523 uschar *prepend, uschar *iplist, BOOL bitmask, BOOL invert_result,
2524 int defer_return)
2525 {
2526 dns_answer dnsa;
2527 dns_scan dnss;
2528 tree_node *t;
2529 dnsbl_cache_block *cb;
2530 int old_pool = store_pool;
2531 uschar query[256]; /* DNS domain max length */
2532
2533 /* Construct the specific query domainname */
2534
2535 if (!string_format(query, sizeof(query), "%s.%s", prepend, domain))
2536 {
2537 log_write(0, LOG_MAIN|LOG_PANIC, "dnslist query is too long "
2538 "(ignored): %s...", query);
2539 return FAIL;
2540 }
2541
2542 /* Look for this query in the cache. */
2543
2544 t = tree_search(dnsbl_cache, query);
2545
2546 /* If not cached from a previous lookup, we must do a DNS lookup, and
2547 cache the result in permanent memory. */
2548
2549 if (t == NULL)
2550 {
2551 store_pool = POOL_PERM;
2552
2553 /* Set up a tree entry to cache the lookup */
2554
2555 t = store_get(sizeof(tree_node) + Ustrlen(query));
2556 Ustrcpy(t->name, query);
2557 t->data.ptr = cb = store_get(sizeof(dnsbl_cache_block));
2558 (void)tree_insertnode(&dnsbl_cache, t);
2559
2560 /* Do the DNS loopup . */
2561
2562 HDEBUG(D_dnsbl) debug_printf("new DNS lookup for %s\n", query);
2563 cb->rc = dns_basic_lookup(&dnsa, query, T_A);
2564 cb->text_set = FALSE;
2565 cb->text = NULL;
2566 cb->rhs = NULL;
2567
2568 /* If the lookup succeeded, cache the RHS address. The code allows for
2569 more than one address - this was for complete generality and the possible
2570 use of A6 records. However, A6 records have been reduced to experimental
2571 status (August 2001) and may die out. So they may never get used at all,
2572 let alone in dnsbl records. However, leave the code here, just in case.
2573
2574 Quite apart from one A6 RR generating multiple addresses, there are DNS
2575 lists that return more than one A record, so we must handle multiple
2576 addresses generated in that way as well. */
2577
2578 if (cb->rc == DNS_SUCCEED)
2579 {
2580 dns_record *rr;
2581 dns_address **addrp = &(cb->rhs);
2582 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2583 rr != NULL;
2584 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2585 {
2586 if (rr->type == T_A)
2587 {
2588 dns_address *da = dns_address_from_rr(&dnsa, rr);
2589 if (da != NULL)
2590 {
2591 *addrp = da;
2592 while (da->next != NULL) da = da->next;
2593 addrp = &(da->next);
2594 }
2595 }
2596 }
2597
2598 /* If we didn't find any A records, change the return code. This can
2599 happen when there is a CNAME record but there are no A records for what
2600 it points to. */
2601
2602 if (cb->rhs == NULL) cb->rc = DNS_NODATA;
2603 }
2604
2605 store_pool = old_pool;
2606 }
2607
2608 /* Previous lookup was cached */
2609
2610 else
2611 {
2612 HDEBUG(D_dnsbl) debug_printf("using result of previous DNS lookup\n");
2613 cb = t->data.ptr;
2614 }
2615
2616 /* We now have the result of the DNS lookup, either newly done, or cached
2617 from a previous call. If the lookup succeeded, check against the address
2618 list if there is one. This may be a positive equality list (introduced by
2619 "="), a negative equality list (introduced by "!="), a positive bitmask
2620 list (introduced by "&"), or a negative bitmask list (introduced by "!&").*/
2621
2622 if (cb->rc == DNS_SUCCEED)
2623 {
2624 dns_address *da = NULL;
2625 uschar *addlist = cb->rhs->address;
2626
2627 /* For A and AAAA records, there may be multiple addresses from multiple
2628 records. For A6 records (currently not expected to be used) there may be
2629 multiple addresses from a single record. */
2630
2631 for (da = cb->rhs->next; da != NULL; da = da->next)
2632 addlist = string_sprintf("%s, %s", addlist, da->address);
2633
2634 HDEBUG(D_dnsbl) debug_printf("DNS lookup for %s succeeded (yielding %s)\n",
2635 query, addlist);
2636
2637 /* Address list check; this can be either for equality, or via a bitmask.
2638 In the latter case, all the bits must match. */
2639
2640 if (iplist != NULL)
2641 {
2642 int ipsep = ',';
2643 uschar ip[46];
2644 uschar *ptr = iplist;
2645
2646 while (string_nextinlist(&ptr, &ipsep, ip, sizeof(ip)) != NULL)
2647 {
2648 /* Handle exact matching */
2649 if (!bitmask)
2650 {
2651 for (da = cb->rhs; da != NULL; da = da->next)
2652 {
2653 if (Ustrcmp(CS da->address, ip) == 0) break;
2654 }
2655 }
2656 /* Handle bitmask matching */
2657 else
2658 {
2659 int address[4];
2660 int mask = 0;
2661
2662 /* At present, all known DNS blocking lists use A records, with
2663 IPv4 addresses on the RHS encoding the information they return. I
2664 wonder if this will linger on as the last vestige of IPv4 when IPv6
2665 is ubiquitous? Anyway, for now we use paranoia code to completely
2666 ignore IPv6 addresses. The default mask is 0, which always matches.
2667 We change this only for IPv4 addresses in the list. */
2668
2669 if (host_aton(ip, address) == 1) mask = address[0];
2670
2671 /* Scan the returned addresses, skipping any that are IPv6 */
2672
2673 for (da = cb->rhs; da != NULL; da = da->next)
2674 {
2675 if (host_aton(da->address, address) != 1) continue;
2676 if ((address[0] & mask) == mask) break;
2677 }
2678 }
2679
2680 /* Break out if a match has been found */
2681
2682 if (da != NULL) break;
2683 }
2684
2685 /* If either
2686
2687 (a) No IP address in a positive list matched, or
2688 (b) An IP address in a negative list did match
2689
2690 then behave as if the DNSBL lookup had not succeeded, i.e. the host is
2691 not on the list. */
2692
2693 if (invert_result != (da == NULL))
2694 {
2695 HDEBUG(D_dnsbl)
2696 {
2697 debug_printf("=> but we are not accepting this block class because\n");
2698 debug_printf("=> there was %s match for %c%s\n",
2699 invert_result? "an exclude":"no", bitmask? '&' : '=', iplist);
2700 }
2701 return FAIL;
2702 }
2703 }
2704
2705 /* Either there was no IP list, or the record matched, implying that the
2706 domain is on the list. We now want to find a corresponding TXT record. If an
2707 alternate domain is specified for the TXT record, call this function
2708 recursively to look that up; this has the side effect of re-checking that
2709 there is indeed an A record at the alternate domain. */
2710
2711 if (domain_txt != domain)
2712 return one_check_dnsbl(domain_txt, domain_txt, keydomain, prepend, NULL,
2713 FALSE, invert_result, defer_return);
2714
2715 /* If there is no alternate domain, look up a TXT record in the main domain
2716 if it has not previously been cached. */
2717
2718 if (!cb->text_set)
2719 {
2720 cb->text_set = TRUE;
2721 if (dns_basic_lookup(&dnsa, query, T_TXT) == DNS_SUCCEED)
2722 {
2723 dns_record *rr;
2724 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2725 rr != NULL;
2726 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2727 if (rr->type == T_TXT) break;
2728 if (rr != NULL)
2729 {
2730 int len = (rr->data)[0];
2731 if (len > 511) len = 127;
2732 store_pool = POOL_PERM;
2733 cb->text = string_sprintf("%.*s", len, (const uschar *)(rr->data+1));
2734 store_pool = old_pool;
2735 }
2736 }
2737 }
2738
2739 dnslist_value = addlist;
2740 dnslist_text = cb->text;
2741 return OK;
2742 }
2743
2744 /* There was a problem with the DNS lookup */
2745
2746 if (cb->rc != DNS_NOMATCH && cb->rc != DNS_NODATA)
2747 {
2748 log_write(L_dnslist_defer, LOG_MAIN,
2749 "DNS list lookup defer (probably timeout) for %s: %s", query,
2750 (defer_return == OK)? US"assumed in list" :
2751 (defer_return == FAIL)? US"assumed not in list" :
2752 US"returned DEFER");
2753 return defer_return;
2754 }
2755
2756 /* No entry was found in the DNS; continue for next domain */
2757
2758 HDEBUG(D_dnsbl)
2759 {
2760 debug_printf("DNS lookup for %s failed\n", query);
2761 debug_printf("=> that means %s is not listed at %s\n",
2762 keydomain, domain);
2763 }
2764
2765 return FAIL;
2766 }
2767
2768
2769
2770
2771 /*************************************************
2772 * Check host against DNS black lists *
2773 *************************************************/
2774
2775 /* This function runs checks against a list of DNS black lists, until one
2776 matches. Each item on the list can be of the form
2777
2778 domain=ip-address/key
2779
2780 The domain is the right-most domain that is used for the query, for example,
2781 blackholes.mail-abuse.org. If the IP address is present, there is a match only
2782 if the DNS lookup returns a matching IP address. Several addresses may be
2783 given, comma-separated, for example: x.y.z=127.0.0.1,127.0.0.2.
2784
2785 If no key is given, what is looked up in the domain is the inverted IP address
2786 of the current client host. If a key is given, it is used to construct the
2787 domain for the lookup. For example:
2788
2789 dsn.rfc-ignorant.org/$sender_address_domain
2790
2791 After finding a match in the DNS, the domain is placed in $dnslist_domain, and
2792 then we check for a TXT record for an error message, and if found, save its
2793 value in $dnslist_text. We also cache everything in a tree, to optimize
2794 multiple lookups.
2795
2796 The TXT record is normally looked up in the same domain as the A record, but
2797 when many lists are combined in a single DNS domain, this will not be a very
2798 specific message. It is possible to specify a different domain for looking up
2799 TXT records; this is given before the main domain, comma-separated. For
2800 example:
2801
2802 dnslists = http.dnsbl.sorbs.net,dnsbl.sorbs.net=127.0.0.2 : \
2803 socks.dnsbl.sorbs.net,dnsbl.sorbs.net=127.0.0.3
2804
2805 The caching ensures that only one lookup in dnsbl.sorbs.net is done.
2806
2807 Note: an address for testing RBL is 192.203.178.39
2808 Note: an address for testing DUL is 192.203.178.4
2809 Note: a domain for testing RFCI is example.tld.dsn.rfc-ignorant.org
2810
2811 Arguments:
2812 listptr the domain/address/data list
2813
2814 Returns: OK successful lookup (i.e. the address is on the list), or
2815 lookup deferred after +include_unknown
2816 FAIL name not found, or no data found for the given type, or
2817 lookup deferred after +exclude_unknown (default)
2818 DEFER lookup failure, if +defer_unknown was set
2819 */
2820
2821 int
2822 verify_check_dnsbl(uschar **listptr)
2823 {
2824 int sep = 0;
2825 int defer_return = FAIL;
2826 BOOL invert_result = FALSE;
2827 uschar *list = *listptr;
2828 uschar *domain;
2829 uschar *s;
2830 uschar buffer[1024];
2831 uschar revadd[128]; /* Long enough for IPv6 address */
2832
2833 /* Indicate that the inverted IP address is not yet set up */
2834
2835 revadd[0] = 0;
2836
2837 /* In case this is the first time the DNS resolver is being used. */
2838
2839 dns_init(FALSE, FALSE);
2840
2841 /* Loop through all the domains supplied, until something matches */
2842
2843 while ((domain = string_nextinlist(&list, &sep, buffer, sizeof(buffer))) != NULL)
2844 {
2845 int rc;
2846 BOOL bitmask = FALSE;
2847 uschar *domain_txt;
2848 uschar *comma;
2849 uschar *iplist;
2850 uschar *key;
2851
2852 HDEBUG(D_dnsbl) debug_printf("DNS list check: %s\n", domain);
2853
2854 /* Deal with special values that change the behaviour on defer */
2855
2856 if (domain[0] == '+')
2857 {
2858 if (strcmpic(domain, US"+include_unknown") == 0) defer_return = OK;
2859 else if (strcmpic(domain, US"+exclude_unknown") == 0) defer_return = FAIL;
2860 else if (strcmpic(domain, US"+defer_unknown") == 0) defer_return = DEFER;
2861 else
2862 log_write(0, LOG_MAIN|LOG_PANIC, "unknown item in dnslist (ignored): %s",
2863 domain);
2864 continue;
2865 }
2866
2867 /* See if there's explicit data to be looked up */
2868
2869 key = Ustrchr(domain, '/');
2870 if (key != NULL) *key++ = 0;
2871
2872 /* See if there's a list of addresses supplied after the domain name. This is
2873 introduced by an = or a & character; if preceded by ! we invert the result.
2874 */
2875
2876 iplist = Ustrchr(domain, '=');
2877 if (iplist == NULL)
2878 {
2879 bitmask = TRUE;
2880 iplist = Ustrchr(domain, '&');
2881 }
2882
2883 if (iplist != NULL)
2884 {
2885 if (iplist > domain && iplist[-1] == '!')
2886 {
2887 invert_result = TRUE;
2888 iplist[-1] = 0;
2889 }
2890 *iplist++ = 0;
2891 }
2892
2893 /* If there is a comma in the domain, it indicates that a second domain for
2894 looking up TXT records is provided, before the main domain. Otherwise we must
2895 set domain_txt == domain. */
2896
2897 domain_txt = domain;
2898 comma = Ustrchr(domain, ',');
2899 if (comma != NULL)
2900 {
2901 *comma++ = 0;
2902 domain = comma;
2903 }
2904
2905 /* Check that what we have left is a sensible domain name. There is no reason
2906 why these domains should in fact use the same syntax as hosts and email
2907 domains, but in practice they seem to. However, there is little point in
2908 actually causing an error here, because that would no doubt hold up incoming
2909 mail. Instead, I'll just log it. */
2910
2911 for (s = domain; *s != 0; s++)
2912 {
2913 if (!isalnum(*s) && *s != '-' && *s != '.')
2914 {
2915 log_write(0, LOG_MAIN, "dnslists domain \"%s\" contains "
2916 "strange characters - is this right?", domain);
2917 break;
2918 }
2919 }
2920
2921 /* Check the alternate domain if present */
2922
2923 if (domain_txt != domain) for (s = domain_txt; *s != 0; s++)
2924 {
2925 if (!isalnum(*s) && *s != '-' && *s != '.')
2926 {
2927 log_write(0, LOG_MAIN, "dnslists domain \"%s\" contains "
2928 "strange characters - is this right?", domain_txt);
2929 break;
2930 }
2931 }
2932
2933 /* If there is no key string, construct the query by adding the domain name
2934 onto the inverted host address, and perform a single DNS lookup. */
2935
2936 if (key == NULL)
2937 {
2938 if (sender_host_address == NULL) return FAIL; /* can never match */
2939 if (revadd[0] == 0) invert_address(revadd, sender_host_address);
2940 rc = one_check_dnsbl(domain, domain_txt, sender_host_address, revadd,
2941 iplist, bitmask, invert_result, defer_return);
2942 if (rc == OK)
2943 {
2944 dnslist_domain = string_copy(domain_txt);
2945 HDEBUG(D_dnsbl) debug_printf("=> that means %s is listed at %s\n",
2946 sender_host_address, dnslist_domain);
2947 }
2948 if (rc != FAIL) return rc; /* OK or DEFER */
2949 }
2950
2951 /* If there is a key string, it can be a list of domains or IP addresses to
2952 be concatenated with the main domain. */
2953
2954 else
2955 {
2956 int keysep = 0;
2957 BOOL defer = FALSE;
2958 uschar *keydomain;
2959 uschar keybuffer[256];
2960 uschar keyrevadd[128];
2961
2962 while ((keydomain = string_nextinlist(&key, &keysep, keybuffer,
2963 sizeof(keybuffer))) != NULL)
2964 {
2965 uschar *prepend = keydomain;
2966
2967 if (string_is_ip_address(keydomain, NULL) != 0)
2968 {
2969 invert_address(keyrevadd, keydomain);
2970 prepend = keyrevadd;
2971 }
2972
2973 rc = one_check_dnsbl(domain, domain_txt, keydomain, prepend, iplist,
2974 bitmask, invert_result, defer_return);
2975
2976 if (rc == OK)
2977 {
2978 dnslist_domain = string_copy(domain_txt);
2979 HDEBUG(D_dnsbl) debug_printf("=> that means %s is listed at %s\n",
2980 keydomain, dnslist_domain);
2981 return OK;
2982 }
2983
2984 /* If the lookup deferred, remember this fact. We keep trying the rest
2985 of the list to see if we get a useful result, and if we don't, we return
2986 DEFER at the end. */
2987
2988 if (rc == DEFER) defer = TRUE;
2989 } /* continue with next keystring domain/address */
2990
2991 if (defer) return DEFER;
2992 }
2993 } /* continue with next dnsdb outer domain */
2994
2995 return FAIL;
2996 }
2997
2998 /* End of verify.c */