bae410e6722c52464e16d6ab7110f8530027147b
[exim.git] / src / src / verify.c
1 /* $Cambridge: exim/src/src/verify.c,v 1.1 2004/10/07 10:39:01 ph10 Exp $ */
2
3 /*************************************************
4 * Exim - an Internet mail transport agent *
5 *************************************************/
6
7 /* Copyright (c) University of Cambridge 1995 - 2004 */
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 options the verification options - these bits are used:
128 vopt_is_recipient => this is a recipient address
129 vopt_callout_no_cache => don't use callout cache
130 vopt_callout_random => do the "random" thing
131 vopt_callout_recipsender => use real sender for recipient
132 vopt_callout_recippmaster => use postmaster for recipient
133 se_mailfrom MAIL FROM address for sender verify; NULL => ""
134 pm_mailfrom if non-NULL, do the postmaster check with this sender
135
136 Returns: OK/FAIL/DEFER
137 */
138
139 static int
140 do_callout(address_item *addr, host_item *host_list, transport_feedback *tf,
141 int callout, int callout_overall, int options, uschar *se_mailfrom,
142 uschar *pm_mailfrom)
143 {
144 BOOL is_recipient = (options & vopt_is_recipient) != 0;
145 BOOL callout_no_cache = (options & vopt_callout_no_cache) != 0;
146 BOOL callout_random = (options & vopt_callout_random) != 0;
147
148 int yield = OK;
149 BOOL done = FALSE;
150 uschar *address_key;
151 uschar *from_address;
152 uschar *random_local_part = NULL;
153 open_db dbblock;
154 open_db *dbm_file = NULL;
155 dbdata_callout_cache new_domain_record;
156 dbdata_callout_cache_address new_address_record;
157 host_item *host;
158 time_t callout_start_time;
159
160 new_domain_record.result = ccache_unknown;
161 new_domain_record.postmaster_result = ccache_unknown;
162 new_domain_record.random_result = ccache_unknown;
163
164 memset(&new_address_record, 0, sizeof(new_address_record));
165
166 /* For a recipient callout, the key used for the address cache record must
167 include the sender address if we are using the real sender in the callout,
168 because that may influence the result of the callout. */
169
170 address_key = addr->address;
171 from_address = US"";
172
173 if (is_recipient)
174 {
175 if ((options & vopt_callout_recipsender) != 0)
176 {
177 address_key = string_sprintf("%s/<%s>", addr->address, sender_address);
178 from_address = sender_address;
179 }
180 else if ((options & vopt_callout_recippmaster) != 0)
181 {
182 address_key = string_sprintf("%s/<postmaster@%s>", addr->address,
183 qualify_domain_sender);
184 from_address = string_sprintf("postmaster@%s", qualify_domain_sender);
185 }
186 }
187
188 /* For a sender callout, we must adjust the key if the mailfrom address is not
189 empty. */
190
191 else
192 {
193 from_address = (se_mailfrom == NULL)? US"" : se_mailfrom;
194 if (from_address[0] != 0)
195 address_key = string_sprintf("%s/<%s>", addr->address, from_address);
196 }
197
198 /* Open the callout cache database, it it exists, for reading only at this
199 stage, unless caching has been disabled. */
200
201 if (callout_no_cache)
202 {
203 HDEBUG(D_verify) debug_printf("callout cache: disabled by no_cache\n");
204 }
205 else if ((dbm_file = dbfn_open(US"callout", O_RDWR, &dbblock, FALSE)) == NULL)
206 {
207 HDEBUG(D_verify) debug_printf("callout cache: not available\n");
208 }
209
210 /* If a cache database is available see if we can avoid the need to do an
211 actual callout by making use of previously-obtained data. */
212
213 if (dbm_file != NULL)
214 {
215 dbdata_callout_cache_address *cache_address_record;
216 dbdata_callout_cache *cache_record = get_callout_cache_record(dbm_file,
217 addr->domain, US"domain",
218 callout_cache_domain_positive_expire,
219 callout_cache_domain_negative_expire);
220
221 /* If an unexpired cache record was found for this domain, see if the callout
222 process can be short-circuited. */
223
224 if (cache_record != NULL)
225 {
226 /* If an early command (up to and including MAIL FROM:<>) was rejected,
227 there is no point carrying on. The callout fails. */
228
229 if (cache_record->result == ccache_reject)
230 {
231 setflag(addr, af_verify_nsfail);
232 HDEBUG(D_verify)
233 debug_printf("callout cache: domain gave initial rejection, or "
234 "does not accept HELO or MAIL FROM:<>\n");
235 setflag(addr, af_verify_nsfail);
236 addr->user_message = US"(result of an earlier callout reused).";
237 yield = FAIL;
238 goto END_CALLOUT;
239 }
240
241 /* If a previous check on a "random" local part was accepted, we assume
242 that the server does not do any checking on local parts. There is therefore
243 no point in doing the callout, because it will always be successful. If a
244 random check previously failed, arrange not to do it again, but preserve
245 the data in the new record. If a random check is required but hasn't been
246 done, skip the remaining cache processing. */
247
248 if (callout_random) switch(cache_record->random_result)
249 {
250 case ccache_accept:
251 HDEBUG(D_verify)
252 debug_printf("callout cache: domain accepts random addresses\n");
253 goto END_CALLOUT; /* Default yield is OK */
254
255 case ccache_reject:
256 HDEBUG(D_verify)
257 debug_printf("callout cache: domain rejects random addresses\n");
258 callout_random = FALSE;
259 new_domain_record.random_result = ccache_reject;
260 new_domain_record.random_stamp = cache_record->random_stamp;
261 break;
262
263 default:
264 HDEBUG(D_verify)
265 debug_printf("callout cache: need to check random address handling "
266 "(not cached or cache expired)\n");
267 goto END_CACHE;
268 }
269
270 /* If a postmaster check is requested, but there was a previous failure,
271 there is again no point in carrying on. If a postmaster check is required,
272 but has not been done before, we are going to have to do a callout, so skip
273 remaining cache processing. */
274
275 if (pm_mailfrom != NULL)
276 {
277 if (cache_record->postmaster_result == ccache_reject)
278 {
279 setflag(addr, af_verify_pmfail);
280 HDEBUG(D_verify)
281 debug_printf("callout cache: domain does not accept "
282 "RCPT TO:<postmaster@domain>\n");
283 yield = FAIL;
284 setflag(addr, af_verify_pmfail);
285 addr->user_message = US"(result of earlier verification reused).";
286 goto END_CALLOUT;
287 }
288 if (cache_record->postmaster_result == ccache_unknown)
289 {
290 HDEBUG(D_verify)
291 debug_printf("callout cache: need to check RCPT "
292 "TO:<postmaster@domain> (not cached or cache expired)\n");
293 goto END_CACHE;
294 }
295
296 /* If cache says OK, set pm_mailfrom NULL to prevent a redundant
297 postmaster check if the address itself has to be checked. Also ensure
298 that the value in the cache record is preserved (with its old timestamp).
299 */
300
301 HDEBUG(D_verify) debug_printf("callout cache: domain accepts RCPT "
302 "TO:<postmaster@domain>\n");
303 pm_mailfrom = NULL;
304 new_domain_record.postmaster_result = ccache_accept;
305 new_domain_record.postmaster_stamp = cache_record->postmaster_stamp;
306 }
307 }
308
309 /* We can't give a result based on information about the domain. See if there
310 is an unexpired cache record for this specific address (combined with the
311 sender address if we are doing a recipient callout with a non-empty sender).
312 */
313
314 cache_address_record = (dbdata_callout_cache_address *)
315 get_callout_cache_record(dbm_file,
316 address_key, US"address",
317 callout_cache_positive_expire,
318 callout_cache_negative_expire);
319
320 if (cache_address_record != NULL)
321 {
322 if (cache_address_record->result == ccache_accept)
323 {
324 HDEBUG(D_verify)
325 debug_printf("callout cache: address record is positive\n");
326 }
327 else
328 {
329 HDEBUG(D_verify)
330 debug_printf("callout cache: address record is negative\n");
331 addr->user_message = US"Previous (cached) callout verification failure";
332 yield = FAIL;
333 }
334 goto END_CALLOUT;
335 }
336
337 /* Close the cache database while we actually do the callout for real. */
338
339 END_CACHE:
340 dbfn_close(dbm_file);
341 dbm_file = NULL;
342 }
343
344 /* The information wasn't available in the cache, so we have to do a real
345 callout and save the result in the cache for next time, unless no_cache is set,
346 or unless we have a previously cached negative random result. If we are to test
347 with a random local part, ensure that such a local part is available. If not,
348 log the fact, but carry on without randomming. */
349
350 if (callout_random && callout_random_local_part != NULL)
351 {
352 random_local_part = expand_string(callout_random_local_part);
353 if (random_local_part == NULL)
354 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand "
355 "callout_random_local_part: %s", expand_string_message);
356 }
357
358 /* Default the overall callout timeout if not set, and record the time we are
359 starting so that we can enforce it. */
360
361 if (callout_overall < 0) callout_overall = 4 * callout;
362 callout_start_time = time(NULL);
363
364 /* Now make connections to the hosts and do real callouts. The list of hosts
365 is passed in as an argument. */
366
367 for (host = host_list; host != NULL && !done; host = host->next)
368 {
369 smtp_inblock inblock;
370 smtp_outblock outblock;
371 int host_af;
372 int port = 25;
373 uschar *helo = US"HELO";
374 uschar *interface = NULL; /* Outgoing interface to use; NULL => any */
375 uschar inbuffer[4096];
376 uschar outbuffer[1024];
377 uschar responsebuffer[4096];
378
379 clearflag(addr, af_verify_pmfail); /* postmaster callout flag */
380 clearflag(addr, af_verify_nsfail); /* null sender callout flag */
381
382 /* Skip this host if we don't have an IP address for it. */
383
384 if (host->address == NULL)
385 {
386 DEBUG(D_verify) debug_printf("no IP address for host name %s: skipping\n",
387 host->name);
388 continue;
389 }
390
391 /* Check the overall callout timeout */
392
393 if (time(NULL) - callout_start_time >= callout_overall)
394 {
395 HDEBUG(D_verify) debug_printf("overall timeout for callout exceeded\n");
396 break;
397 }
398
399 /* Set IPv4 or IPv6 */
400
401 host_af = (Ustrchr(host->address, ':') == NULL)? AF_INET:AF_INET6;
402
403 /* Expand and interpret the interface and port strings. This has to
404 be delayed till now, because they may expand differently for different
405 hosts. If there's a failure, log it, but carry on with the defaults. */
406
407 deliver_host = host->name;
408 deliver_host_address = host->address;
409 if (!smtp_get_interface(tf->interface, host_af, addr, NULL, &interface,
410 US"callout") ||
411 !smtp_get_port(tf->port, addr, &port, US"callout"))
412 log_write(0, LOG_MAIN|LOG_PANIC, "<%s>: %s", addr->address,
413 addr->message);
414 deliver_host = deliver_host_address = NULL;
415
416 /* Set HELO string according to the protocol */
417
418 if (Ustrcmp(tf->protocol, "lmtp") == 0) helo = US"LHLO";
419
420 HDEBUG(D_verify) debug_printf("interface=%s port=%d\n", interface, port);
421
422 /* Set up the buffer for reading SMTP response packets. */
423
424 inblock.buffer = inbuffer;
425 inblock.buffersize = sizeof(inbuffer);
426 inblock.ptr = inbuffer;
427 inblock.ptrend = inbuffer;
428
429 /* Set up the buffer for holding SMTP commands while pipelining */
430
431 outblock.buffer = outbuffer;
432 outblock.buffersize = sizeof(outbuffer);
433 outblock.ptr = outbuffer;
434 outblock.cmd_count = 0;
435 outblock.authenticating = FALSE;
436
437 /* Connect to the host; on failure, just loop for the next one, but we
438 set the error for the last one. */
439
440 inblock.sock = outblock.sock =
441 smtp_connect(host, host_af, port, interface, callout, TRUE);
442 if (inblock.sock < 0)
443 {
444 addr->message = string_sprintf("could not connect to %s [%s]: %s",
445 host->name, host->address, strerror(errno));
446 continue;
447 }
448
449 /* Wait for initial response, and then run the initial SMTP commands. The
450 smtp_write_command() function leaves its command in big_buffer. This is
451 used in error responses. Initialize it in case the connection is
452 rejected. */
453
454 Ustrcpy(big_buffer, "initial connection");
455
456 done =
457 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
458 '2', callout) &&
459
460 smtp_write_command(&outblock, FALSE, "%s %s\r\n", helo,
461 smtp_active_hostname) >= 0 &&
462 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
463 '2', callout) &&
464
465 smtp_write_command(&outblock, FALSE, "MAIL FROM:<%s>\r\n",
466 from_address) >= 0 &&
467 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
468 '2', callout);
469
470 /* If the host gave an initial error, or does not accept HELO or MAIL
471 FROM:<>, arrange to cache this information, but don't record anything for an
472 I/O error or a defer. Do not cache rejections when a non-empty sender has
473 been used, because that blocks the whole domain for all senders. */
474
475 if (!done)
476 {
477 if (errno == 0 && responsebuffer[0] == '5')
478 {
479 setflag(addr, af_verify_nsfail);
480 if (from_address[0] == 0) new_domain_record.result = ccache_reject;
481 }
482 }
483
484 /* Otherwise, proceed to check a "random" address (if required), then the
485 given address, and the postmaster address (if required). Between each check,
486 issue RSET, because some servers accept only one recipient after MAIL
487 FROM:<>. */
488
489 else
490 {
491 new_domain_record.result = ccache_accept;
492
493 /* Do the random local part check first */
494
495 if (random_local_part != NULL)
496 {
497 uschar randombuffer[1024];
498 BOOL random_ok =
499 smtp_write_command(&outblock, FALSE,
500 "RCPT TO:<%.1000s@%.1000s>\r\n", random_local_part,
501 addr->domain) >= 0 &&
502 smtp_read_response(&inblock, randombuffer,
503 sizeof(randombuffer), '2', callout);
504
505 /* Remember when we last did a random test */
506
507 new_domain_record.random_stamp = time(NULL);
508
509 /* If accepted, we aren't going to do any further tests below. */
510
511 if (random_ok)
512 {
513 new_domain_record.random_result = ccache_accept;
514 }
515
516 /* Otherwise, cache a real negative response, and get back to the right
517 state to send RCPT. Unless there's some problem such as a dropped
518 connection, we expect to succeed, because the commands succeeded above. */
519
520 else if (errno == 0)
521 {
522 if (randombuffer[0] == '5')
523 new_domain_record.random_result = ccache_reject;
524
525 done =
526 smtp_write_command(&outblock, FALSE, "RSET\r\n") >= 0 &&
527 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
528 '2', callout) &&
529
530 smtp_write_command(&outblock, FALSE, "MAIL FROM:<>\r\n") >= 0 &&
531 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
532 '2', callout);
533 }
534 else done = FALSE; /* Some timeout/connection problem */
535 } /* Random check */
536
537 /* If the host is accepting all local parts, as determined by the "random"
538 check, we don't need to waste time doing any further checking. */
539
540 if (new_domain_record.random_result != ccache_accept && done)
541 {
542 done =
543 smtp_write_command(&outblock, FALSE, "RCPT TO:<%.1000s>\r\n",
544 addr->address) >= 0 &&
545 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
546 '2', callout);
547
548 if (done)
549 new_address_record.result = ccache_accept;
550 else if (errno == 0 && responsebuffer[0] == '5')
551 new_address_record.result = ccache_reject;
552
553 /* Do postmaster check if requested */
554
555 if (done && pm_mailfrom != NULL)
556 {
557 done =
558 smtp_write_command(&outblock, FALSE, "RSET\r\n") >= 0 &&
559 smtp_read_response(&inblock, responsebuffer,
560 sizeof(responsebuffer), '2', callout) &&
561
562 smtp_write_command(&outblock, FALSE,
563 "MAIL FROM:<%s>\r\n", pm_mailfrom) >= 0 &&
564 smtp_read_response(&inblock, responsebuffer,
565 sizeof(responsebuffer), '2', callout) &&
566
567 smtp_write_command(&outblock, FALSE,
568 "RCPT TO:<postmaster@%.1000s>\r\n", addr->domain) >= 0 &&
569 smtp_read_response(&inblock, responsebuffer,
570 sizeof(responsebuffer), '2', callout);
571
572 new_domain_record.postmaster_stamp = time(NULL);
573
574 if (done)
575 new_domain_record.postmaster_result = ccache_accept;
576 else if (errno == 0 && responsebuffer[0] == '5')
577 {
578 setflag(addr, af_verify_pmfail);
579 new_domain_record.postmaster_result = ccache_reject;
580 }
581 }
582 } /* Random not accepted */
583 } /* MAIL FROM:<> accepted */
584
585 /* For any failure of the main check, other than a negative response, we just
586 close the connection and carry on. We can identify a negative response by the
587 fact that errno is zero. For I/O errors it will be non-zero
588
589 Set up different error texts for logging and for sending back to the caller
590 as an SMTP response. Log in all cases, using a one-line format. For sender
591 callouts, give a full response to the caller, but for recipient callouts,
592 don't give the IP address because this may be an internal host whose identity
593 is not to be widely broadcast. */
594
595 if (!done)
596 {
597 if (errno == ETIMEDOUT)
598 {
599 HDEBUG(D_verify) debug_printf("SMTP timeout\n");
600 }
601 else if (errno == 0)
602 {
603 if (*responsebuffer == 0) Ustrcpy(responsebuffer, US"connection dropped");
604
605 addr->message =
606 string_sprintf("response to \"%s\" from %s [%s] was: %s",
607 big_buffer, host->name, host->address,
608 string_printing(responsebuffer));
609
610 addr->user_message = is_recipient?
611 string_sprintf("Callout verification failed:\n%s", responsebuffer)
612 :
613 string_sprintf("Called: %s\nSent: %s\nResponse: %s",
614 host->address, big_buffer, responsebuffer);
615
616 /* Hard rejection ends the process */
617
618 if (responsebuffer[0] == '5') /* Address rejected */
619 {
620 yield = FAIL;
621 done = TRUE;
622 }
623 }
624 }
625
626 /* End the SMTP conversation and close the connection. */
627
628 (void)smtp_write_command(&outblock, FALSE, "QUIT\r\n");
629 close(inblock.sock);
630 } /* Loop through all hosts, while !done */
631
632 /* If we get here with done == TRUE, a successful callout happened, and yield
633 will be set OK or FAIL according to the response to the RCPT command.
634 Otherwise, we looped through the hosts but couldn't complete the business.
635 However, there may be domain-specific information to cache in both cases.
636
637 The value of the result field in the new_domain record is ccache_unknown if
638 there was an error before or with MAIL FROM:<>, and errno was not zero,
639 implying some kind of I/O error. We don't want to write the cache in that case.
640 Otherwise the value is ccache_accept or ccache_reject. */
641
642 if (!callout_no_cache && new_domain_record.result != ccache_unknown)
643 {
644 if ((dbm_file = dbfn_open(US"callout", O_RDWR|O_CREAT, &dbblock, FALSE))
645 == NULL)
646 {
647 HDEBUG(D_verify) debug_printf("callout cache: not available\n");
648 }
649 else
650 {
651 (void)dbfn_write(dbm_file, addr->domain, &new_domain_record,
652 (int)sizeof(dbdata_callout_cache));
653 HDEBUG(D_verify) debug_printf("wrote callout cache domain record:\n"
654 " result=%d postmaster=%d random=%d\n",
655 new_domain_record.result,
656 new_domain_record.postmaster_result,
657 new_domain_record.random_result);
658 }
659 }
660
661 /* If a definite result was obtained for the callout, cache it unless caching
662 is disabled. */
663
664 if (done)
665 {
666 if (!callout_no_cache && new_address_record.result != ccache_unknown)
667 {
668 if (dbm_file == NULL)
669 dbm_file = dbfn_open(US"callout", O_RDWR|O_CREAT, &dbblock, FALSE);
670 if (dbm_file == NULL)
671 {
672 HDEBUG(D_verify) debug_printf("no callout cache available\n");
673 }
674 else
675 {
676 (void)dbfn_write(dbm_file, address_key, &new_address_record,
677 (int)sizeof(dbdata_callout_cache_address));
678 HDEBUG(D_verify) debug_printf("wrote %s callout cache address record\n",
679 (new_address_record.result == ccache_accept)? "positive" : "negative");
680 }
681 }
682 } /* done */
683
684 /* Failure to connect to any host, or any response other than 2xx or 5xx is a
685 temporary error. If there was only one host, and a response was received, leave
686 it alone if supplying details. Otherwise, give a generic response. */
687
688 else /* !done */
689 {
690 uschar *dullmsg = string_sprintf("Could not complete %s verify callout",
691 is_recipient? "recipient" : "sender");
692 yield = DEFER;
693
694 if (host_list->next != NULL || addr->message == NULL) addr->message = dullmsg;
695
696 addr->user_message = (!smtp_return_error_details)? dullmsg :
697 string_sprintf("%s for <%s>.\n"
698 "The mail server(s) for the domain may be temporarily unreachable, or\n"
699 "they may be permanently unreachable from this server. In the latter case,\n%s",
700 dullmsg, addr->address,
701 is_recipient?
702 "the address will never be accepted."
703 :
704 "you need to change the address or create an MX record for its domain\n"
705 "if it is supposed to be generally accessible from the Internet.\n"
706 "Talk to your mail administrator for details.");
707
708 /* Force a specific error code */
709
710 addr->basic_errno = ERRNO_CALLOUTDEFER;
711 }
712
713 /* Come here from within the cache-reading code on fast-track exit. */
714
715 END_CALLOUT:
716 if (dbm_file != NULL) dbfn_close(dbm_file);
717 return yield;
718 }
719
720
721
722 /*************************************************
723 * Copy error to toplevel address *
724 *************************************************/
725
726 /* This function is used when a verify fails or defers, to ensure that the
727 failure or defer information is in the original toplevel address. This applies
728 when an address is redirected to a single new address, and the failure or
729 deferral happens to the child address.
730
731 Arguments:
732 vaddr the verify address item
733 addr the final address item
734 yield FAIL or DEFER
735
736 Returns: the value of YIELD
737 */
738
739 static int
740 copy_error(address_item *vaddr, address_item *addr, int yield)
741 {
742 if (addr != vaddr)
743 {
744 vaddr->message = addr->message;
745 vaddr->user_message = addr->user_message;
746 vaddr->basic_errno = addr->basic_errno;
747 vaddr->more_errno = addr->more_errno;
748 }
749 return yield;
750 }
751
752
753
754
755 /*************************************************
756 * Verify an email address *
757 *************************************************/
758
759 /* This function is used both for verification (-bv and at other times) and
760 address testing (-bt), which is indicated by address_test_mode being set.
761
762 Arguments:
763 vaddr contains the address to verify; the next field in this block
764 must be NULL
765 f if not NULL, write the result to this file
766 options various option bits:
767 vopt_fake_sender => this sender verify is not for the real
768 sender (it was verify=sender=xxxx or an address from a
769 header line) - rewriting must not change sender_address
770 vopt_is_recipient => this is a recipient address, otherwise
771 it's a sender address - this affects qualification and
772 rewriting and messages from callouts
773 vopt_qualify => qualify an unqualified address; else error
774 vopt_expn => called from SMTP EXPN command
775
776 These ones are used by do_callout() -- the options variable
777 is passed to it.
778
779 vopt_callout_no_cache => don't use callout cache
780 vopt_callout_random => do the "random" thing
781 vopt_callout_recipsender => use real sender for recipient
782 vopt_callout_recippmaster => use postmaster for recipient
783
784 callout if > 0, specifies that callout is required, and gives timeout
785 for individual connections and commands
786 callout_overall if > 0, gives overall timeout for the callout function;
787 if < 0, a default is used (see do_callout())
788 se_mailfrom when callout is requested to verify a sender, use this
789 in MAIL FROM; NULL => ""
790 pm_mailfrom when callout is requested, if non-NULL, do the postmaster
791 thing and use this as the sender address (may be "")
792
793 routed if not NULL, set TRUE if routing succeeded, so we can
794 distinguish between routing failed and callout failed
795
796 Returns: OK address verified
797 FAIL address failed to verify
798 DEFER can't tell at present
799 */
800
801 int
802 verify_address(address_item *vaddr, FILE *f, int options, int callout,
803 int callout_overall, uschar *se_mailfrom, uschar *pm_mailfrom, BOOL *routed)
804 {
805 BOOL allok = TRUE;
806 BOOL full_info = (f == NULL)? FALSE : (debug_selector != 0);
807 BOOL is_recipient = (options & vopt_is_recipient) != 0;
808 BOOL expn = (options & vopt_expn) != 0;
809
810 int i;
811 int yield = OK;
812 int verify_type = expn? v_expn :
813 address_test_mode? v_none :
814 is_recipient? v_recipient : v_sender;
815 address_item *addr_list;
816 address_item *addr_new = NULL;
817 address_item *addr_remote = NULL;
818 address_item *addr_local = NULL;
819 address_item *addr_succeed = NULL;
820 uschar *ko_prefix, *cr;
821 uschar *address = vaddr->address;
822 uschar *save_sender;
823 uschar null_sender[] = { 0 }; /* Ensure writeable memory */
824
825 /* Set up a prefix and suffix for error message which allow us to use the same
826 output statements both in EXPN mode (where an SMTP response is needed) and when
827 debugging with an output file. */
828
829 if (expn)
830 {
831 ko_prefix = US"553 ";
832 cr = US"\r";
833 }
834 else ko_prefix = cr = US"";
835
836 /* Add qualify domain if permitted; otherwise an unqualified address fails. */
837
838 if (parse_find_at(address) == NULL)
839 {
840 if ((options & vopt_qualify) == 0)
841 {
842 if (f != NULL)
843 fprintf(f, "%sA domain is required for \"%s\"%s\n", ko_prefix, address,
844 cr);
845 return FAIL;
846 }
847 address = rewrite_address_qualify(address, is_recipient);
848 }
849
850 DEBUG(D_verify)
851 {
852 debug_printf(">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>\n");
853 debug_printf("%s %s\n", address_test_mode? "Testing" : "Verifying", address);
854 }
855
856 /* Rewrite and report on it. Clear the domain and local part caches - these
857 may have been set by domains and local part tests during an ACL. */
858
859 if (global_rewrite_rules != NULL)
860 {
861 uschar *old = address;
862 address = rewrite_address(address, is_recipient, FALSE,
863 global_rewrite_rules, rewrite_existflags);
864 if (address != old)
865 {
866 for (i = 0; i < (MAX_NAMED_LIST * 2)/32; i++) vaddr->localpart_cache[i] = 0;
867 for (i = 0; i < (MAX_NAMED_LIST * 2)/32; i++) vaddr->domain_cache[i] = 0;
868 if (f != NULL && !expn) fprintf(f, "Address rewritten as: %s\n", address);
869 }
870 }
871
872 /* If this is the real sender address, we must update sender_address at
873 this point, because it may be referred to in the routers. */
874
875 if ((options & (vopt_fake_sender|vopt_is_recipient)) == 0)
876 sender_address = address;
877
878 /* If the address was rewritten to <> no verification can be done, and we have
879 to return OK. This rewriting is permitted only for sender addresses; for other
880 addresses, such rewriting fails. */
881
882 if (address[0] == 0) return OK;
883
884 /* Save a copy of the sender address for re-instating if we change it to <>
885 while verifying a sender address (a nice bit of self-reference there). */
886
887 save_sender = sender_address;
888
889 /* Update the address structure with the possibly qualified and rewritten
890 address. Set it up as the starting address on the chain of new addresses. */
891
892 vaddr->address = address;
893 addr_new = vaddr;
894
895 /* We need a loop, because an address can generate new addresses. We must also
896 cope with generated pipes and files at the top level. (See also the code and
897 comment in deliver.c.) However, it is usually the case that the router for
898 user's .forward files has its verify flag turned off.
899
900 If an address generates more than one child, the loop is used only when
901 full_info is set, and this can only be set locally. Remote enquiries just get
902 information about the top level address, not anything that it generated. */
903
904 while (addr_new != NULL)
905 {
906 int rc;
907 address_item *addr = addr_new;
908
909 addr_new = addr->next;
910 addr->next = NULL;
911
912 DEBUG(D_verify)
913 {
914 debug_printf(">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>\n");
915 debug_printf("Considering %s\n", addr->address);
916 }
917
918 /* Handle generated pipe, file or reply addresses. We don't get these
919 when handling EXPN, as it does only one level of expansion. */
920
921 if (testflag(addr, af_pfr))
922 {
923 allok = FALSE;
924 if (f != NULL)
925 {
926 BOOL allow;
927
928 if (addr->address[0] == '>')
929 {
930 allow = testflag(addr, af_allow_reply);
931 fprintf(f, "%s -> mail %s", addr->parent->address, addr->address + 1);
932 }
933 else
934 {
935 allow = (addr->address[0] == '|')?
936 testflag(addr, af_allow_pipe) : testflag(addr, af_allow_file);
937 fprintf(f, "%s -> %s", addr->parent->address, addr->address);
938 }
939
940 if (addr->basic_errno == ERRNO_BADTRANSPORT)
941 fprintf(f, "\n*** Error in setting up pipe, file, or autoreply:\n"
942 "%s\n", addr->message);
943 else if (allow)
944 fprintf(f, "\n transport = %s\n", addr->transport->name);
945 else
946 fprintf(f, " *** forbidden ***\n");
947 }
948 continue;
949 }
950
951 /* Just in case some router parameter refers to it. */
952
953 return_path = (addr->p.errors_address != NULL)?
954 addr->p.errors_address : sender_address;
955
956 /* Split the address into domain and local part, handling the %-hack if
957 necessary, and then route it. While routing a sender address, set
958 $sender_address to <> because that is what it will be if we were trying to
959 send a bounce to the sender. */
960
961 if (routed != NULL) *routed = FALSE;
962 if ((rc = deliver_split_address(addr)) == OK)
963 {
964 if (!is_recipient) sender_address = null_sender;
965 rc = route_address(addr, &addr_local, &addr_remote, &addr_new,
966 &addr_succeed, verify_type);
967 sender_address = save_sender; /* Put back the real sender */
968 }
969
970 /* If routing an address succeeded, set the flag that remembers, for use when
971 an ACL cached a sender verify (in case a callout fails). Then if routing set
972 up a list of hosts or the transport has a host list, and the callout option
973 is set, and we aren't in a host checking run, do the callout verification,
974 and set another flag that notes that a callout happened. */
975
976 if (rc == OK)
977 {
978 if (routed != NULL) *routed = TRUE;
979 if (callout > 0)
980 {
981 host_item *host_list = addr->host_list;
982
983 /* Default, if no remote transport, to NULL for the interface (=> any),
984 "smtp" for the port, and "smtp" for the protocol. */
985
986 transport_feedback tf = { NULL, US"smtp", US"smtp", NULL, FALSE, FALSE };
987
988 /* If verification yielded a remote transport, we want to use that
989 transport's options, so as to mimic what would happen if we were really
990 sending a message to this address. */
991
992 if (addr->transport != NULL && !addr->transport->info->local)
993 {
994 (void)(addr->transport->setup)(addr->transport, addr, &tf, NULL);
995
996 /* If the transport has hosts and the router does not, or if the
997 transport is configured to override the router's hosts, we must build a
998 host list of the transport's hosts, and find the IP addresses */
999
1000 if (tf.hosts != NULL && (host_list == NULL || tf.hosts_override))
1001 {
1002 uschar *s;
1003
1004 host_list = NULL; /* Ignore the router's hosts */
1005
1006 deliver_domain = addr->domain;
1007 deliver_localpart = addr->local_part;
1008 s = expand_string(tf.hosts);
1009 deliver_domain = deliver_localpart = NULL;
1010
1011 if (s == NULL)
1012 {
1013 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand list of hosts "
1014 "\"%s\" in %s transport for callout: %s", tf.hosts,
1015 addr->transport->name, expand_string_message);
1016 }
1017 else
1018 {
1019 uschar *canonical_name;
1020 host_item *host;
1021 host_build_hostlist(&host_list, s, tf.hosts_randomize);
1022
1023 /* Just ignore failures to find a host address. If we don't manage
1024 to find any addresses, the callout will defer. */
1025
1026 for (host = host_list; host != NULL; host = host->next)
1027 {
1028 if (tf.gethostbyname || string_is_ip_address(host->name, NULL))
1029 (void)host_find_byname(host, NULL, &canonical_name, TRUE);
1030 else
1031 {
1032 int flags = HOST_FIND_BY_A;
1033 if (tf.qualify_single) flags |= HOST_FIND_QUALIFY_SINGLE;
1034 if (tf.search_parents) flags |= HOST_FIND_SEARCH_PARENTS;
1035 (void)host_find_bydns(host, NULL, flags, NULL, NULL, NULL,
1036 &canonical_name, NULL);
1037 }
1038 }
1039 }
1040 }
1041 }
1042
1043 /* Can only do a callout if we have at least one host! */
1044
1045 if (host_list != NULL)
1046 {
1047 HDEBUG(D_verify) debug_printf("Attempting full verification using callout\n");
1048 if (host_checking && !host_checking_callout)
1049 {
1050 HDEBUG(D_verify)
1051 debug_printf("... callout omitted by default when host testing\n"
1052 "(Use -bhc if you want the callouts to happen.)\n");
1053 }
1054 else
1055 {
1056 rc = do_callout(addr, host_list, &tf, callout, callout_overall,
1057 options, se_mailfrom, pm_mailfrom);
1058 }
1059 }
1060 else
1061 {
1062 HDEBUG(D_verify) debug_printf("Cannot do callout: neither router nor "
1063 "transport provided a host list\n");
1064 }
1065 }
1066 }
1067
1068 /* A router may return REROUTED if it has set up a child address as a result
1069 of a change of domain name (typically from widening). In this case we always
1070 want to continue to verify the new child. */
1071
1072 if (rc == REROUTED) continue;
1073
1074 /* Handle hard failures */
1075
1076 if (rc == FAIL)
1077 {
1078 allok = FALSE;
1079 if (f != NULL)
1080 {
1081 fprintf(f, "%s%s %s", ko_prefix, address,
1082 address_test_mode? "is undeliverable" : "failed to verify");
1083 if (!expn && admin_user)
1084 {
1085 if (addr->basic_errno > 0)
1086 fprintf(f, ": %s", strerror(addr->basic_errno));
1087 if (addr->message != NULL)
1088 fprintf(f, ":\n %s", addr->message);
1089 }
1090 fprintf(f, "%s\n", cr);
1091 }
1092
1093 if (!full_info) return copy_error(vaddr, addr, FAIL);
1094 else yield = FAIL;
1095 }
1096
1097 /* Soft failure */
1098
1099 else if (rc == DEFER)
1100 {
1101 allok = FALSE;
1102 if (f != NULL)
1103 {
1104 fprintf(f, "%s%s cannot be resolved at this time", ko_prefix, address);
1105 if (!expn && admin_user)
1106 {
1107 if (addr->basic_errno > 0)
1108 fprintf(f, ":\n %s", strerror(addr->basic_errno));
1109 if (addr->message != NULL)
1110 fprintf(f, ":\n %s", addr->message);
1111 else if (addr->basic_errno <= 0)
1112 fprintf(f, ":\n unknown error");
1113 }
1114
1115 fprintf(f, "%s\n", cr);
1116 }
1117 if (!full_info) return copy_error(vaddr, addr, DEFER);
1118 else if (yield == OK) yield = DEFER;
1119 }
1120
1121 /* If we are handling EXPN, we do not want to continue to route beyond
1122 the top level. */
1123
1124 else if (expn)
1125 {
1126 uschar *ok_prefix = US"250-";
1127 if (addr_new == NULL)
1128 {
1129 if (addr_local == NULL && addr_remote == NULL)
1130 fprintf(f, "250 mail to <%s> is discarded\r\n", address);
1131 else
1132 fprintf(f, "250 <%s>\r\n", address);
1133 }
1134 else while (addr_new != NULL)
1135 {
1136 address_item *addr2 = addr_new;
1137 addr_new = addr2->next;
1138 if (addr_new == NULL) ok_prefix = US"250 ";
1139 fprintf(f, "%s<%s>\r\n", ok_prefix, addr2->address);
1140 }
1141 return OK;
1142 }
1143
1144 /* Successful routing other than EXPN. */
1145
1146 else
1147 {
1148 /* Handle successful routing when short info wanted. Otherwise continue for
1149 other (generated) addresses. Short info is the operational case. Full info
1150 can be requested only when debug_selector != 0 and a file is supplied.
1151
1152 There is a conflict between the use of aliasing as an alternate email
1153 address, and as a sort of mailing list. If an alias turns the incoming
1154 address into just one address (e.g. J.Caesar->jc44) you may well want to
1155 carry on verifying the generated address to ensure it is valid when
1156 checking incoming mail. If aliasing generates multiple addresses, you
1157 probably don't want to do this. Exim therefore treats the generation of
1158 just a single new address as a special case, and continues on to verify the
1159 generated address. */
1160
1161 if (!full_info && /* Stop if short info wanted AND */
1162 (addr_new == NULL || /* No new address OR */
1163 addr_new->next != NULL || /* More than one new address OR */
1164 testflag(addr_new, af_pfr))) /* New address is pfr */
1165 {
1166 if (f != NULL) fprintf(f, "%s %s\n", address,
1167 address_test_mode? "is deliverable" : "verified");
1168
1169 /* If we have carried on to verify a child address, we want the value
1170 of $address_data to be that of the child */
1171
1172 vaddr->p.address_data = addr->p.address_data;
1173 return OK;
1174 }
1175 }
1176 } /* Loop for generated addresses */
1177
1178 /* Display the full results of the successful routing, including any generated
1179 addresses. Control gets here only when full_info is set, which requires f not
1180 to be NULL, and this occurs only when a top-level verify is called with the
1181 debugging switch on.
1182
1183 If there are no local and no remote addresses, and there were no pipes, files,
1184 or autoreplies, and there were no errors or deferments, the message is to be
1185 discarded, usually because of the use of :blackhole: in an alias file. */
1186
1187 if (allok && addr_local == NULL && addr_remote == NULL)
1188 fprintf(f, "mail to %s is discarded\n", address);
1189
1190 else for (addr_list = addr_local, i = 0; i < 2; addr_list = addr_remote, i++)
1191 {
1192 while (addr_list != NULL)
1193 {
1194 address_item *addr = addr_list;
1195 address_item *p = addr->parent;
1196 addr_list = addr->next;
1197
1198 fprintf(f, "%s", CS addr->address);
1199 while (p != NULL)
1200 {
1201 fprintf(f, "\n <-- %s", p->address);
1202 p = p->parent;
1203 }
1204 fprintf(f, "\n ");
1205
1206 /* Show router, and transport */
1207
1208 fprintf(f, "router = %s, ", addr->router->name);
1209 fprintf(f, "transport = %s\n", (addr->transport == NULL)? US"unset" :
1210 addr->transport->name);
1211
1212 /* Show any hosts that are set up by a router unless the transport
1213 is going to override them; fiddle a bit to get a nice format. */
1214
1215 if (addr->host_list != NULL && addr->transport != NULL &&
1216 !addr->transport->overrides_hosts)
1217 {
1218 host_item *h;
1219 int maxlen = 0;
1220 int maxaddlen = 0;
1221 for (h = addr->host_list; h != NULL; h = h->next)
1222 {
1223 int len = Ustrlen(h->name);
1224 if (len > maxlen) maxlen = len;
1225 len = (h->address != NULL)? Ustrlen(h->address) : 7;
1226 if (len > maxaddlen) maxaddlen = len;
1227 }
1228 for (h = addr->host_list; h != NULL; h = h->next)
1229 {
1230 int len = Ustrlen(h->name);
1231 fprintf(f, " host %s ", h->name);
1232 while (len++ < maxlen) fprintf(f, " ");
1233 if (h->address != NULL)
1234 {
1235 fprintf(f, "[%s] ", h->address);
1236 len = Ustrlen(h->address);
1237 }
1238 else if (!addr->transport->info->local) /* Omit [unknown] for local */
1239 {
1240 fprintf(f, "[unknown] ");
1241 len = 7;
1242 }
1243 else len = -3;
1244 while (len++ < maxaddlen) fprintf(f," ");
1245 if (h->mx >= 0) fprintf(f, "MX=%d", h->mx);
1246 if (h->port != PORT_NONE) fprintf(f, " port=%d", h->port);
1247 if (h->status == hstatus_unusable) fprintf(f, " ** unusable **");
1248 fprintf(f, "\n");
1249 }
1250 }
1251 }
1252 }
1253
1254 return yield; /* Will be DEFER or FAIL if any one address has */
1255 }
1256
1257
1258
1259
1260 /*************************************************
1261 * Check headers for syntax errors *
1262 *************************************************/
1263
1264 /* This function checks those header lines that contain addresses, and verifies
1265 that all the addresses therein are syntactially correct.
1266
1267 Arguments:
1268 msgptr where to put an error message
1269
1270 Returns: OK
1271 FAIL
1272 */
1273
1274 int
1275 verify_check_headers(uschar **msgptr)
1276 {
1277 header_line *h;
1278 uschar *colon, *s;
1279
1280 for (h = header_list; h != NULL; h = h->next)
1281 {
1282 if (h->type != htype_from &&
1283 h->type != htype_reply_to &&
1284 h->type != htype_sender &&
1285 h->type != htype_to &&
1286 h->type != htype_cc &&
1287 h->type != htype_bcc)
1288 continue;
1289
1290 colon = Ustrchr(h->text, ':');
1291 s = colon + 1;
1292 while (isspace(*s)) s++;
1293
1294 parse_allow_group = TRUE; /* Allow group syntax */
1295
1296 /* Loop for multiple addresses in the header */
1297
1298 while (*s != 0)
1299 {
1300 uschar *ss = parse_find_address_end(s, FALSE);
1301 uschar *recipient, *errmess;
1302 int terminator = *ss;
1303 int start, end, domain;
1304
1305 /* Temporarily terminate the string at this point, and extract the
1306 operative address within. */
1307
1308 *ss = 0;
1309 recipient = parse_extract_address(s,&errmess,&start,&end,&domain,FALSE);
1310 *ss = terminator;
1311
1312 /* Permit an unqualified address only if the message is local, or if the
1313 sending host is configured to be permitted to send them. */
1314
1315 if (recipient != NULL && domain == 0)
1316 {
1317 if (h->type == htype_from || h->type == htype_sender)
1318 {
1319 if (!allow_unqualified_sender) recipient = NULL;
1320 }
1321 else
1322 {
1323 if (!allow_unqualified_recipient) recipient = NULL;
1324 }
1325 if (recipient == NULL) errmess = US"unqualified address not permitted";
1326 }
1327
1328 /* It's an error if no address could be extracted, except for the special
1329 case of an empty address. */
1330
1331 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
1332 {
1333 uschar *verb = US"is";
1334 uschar *t = ss;
1335 int len;
1336
1337 /* Arrange not to include any white space at the end in the
1338 error message. */
1339
1340 while (t > s && isspace(t[-1])) t--;
1341
1342 /* Add the address which failed to the error message, since in a
1343 header with very many addresses it is sometimes hard to spot
1344 which one is at fault. However, limit the amount of address to
1345 quote - cases have been seen where, for example, a missing double
1346 quote in a humungous To: header creates an "address" that is longer
1347 than string_sprintf can handle. */
1348
1349 len = t - s;
1350 if (len > 1024)
1351 {
1352 len = 1024;
1353 verb = US"begins";
1354 }
1355
1356 *msgptr = string_printing(
1357 string_sprintf("%s: failing address in \"%.*s\" header %s: %.*s",
1358 errmess, colon - h->text, h->text, verb, len, s));
1359
1360 return FAIL;
1361 }
1362
1363 /* Advance to the next address */
1364
1365 s = ss + (terminator? 1:0);
1366 while (isspace(*s)) s++;
1367 } /* Next address */
1368 } /* Next header */
1369
1370 return OK;
1371 }
1372
1373
1374
1375
1376 /*************************************************
1377 * Find if verified sender *
1378 *************************************************/
1379
1380 /* Usually, just a single address is verified as the sender of the message.
1381 However, Exim can be made to verify other addresses as well (often related in
1382 some way), and this is useful in some environments. There may therefore be a
1383 chain of such addresses that have previously been tested. This function finds
1384 whether a given address is on the chain.
1385
1386 Arguments: the address to be verified
1387 Returns: pointer to an address item, or NULL
1388 */
1389
1390 address_item *
1391 verify_checked_sender(uschar *sender)
1392 {
1393 address_item *addr;
1394 for (addr = sender_verified_list; addr != NULL; addr = addr->next)
1395 if (Ustrcmp(sender, addr->address) == 0) break;
1396 return addr;
1397 }
1398
1399
1400
1401
1402
1403 /*************************************************
1404 * Get valid header address *
1405 *************************************************/
1406
1407 /* Scan the originator headers of the message, looking for an address that
1408 verifies successfully. RFC 822 says:
1409
1410 o The "Sender" field mailbox should be sent notices of
1411 any problems in transport or delivery of the original
1412 messages. If there is no "Sender" field, then the
1413 "From" field mailbox should be used.
1414
1415 o If the "Reply-To" field exists, then the reply should
1416 go to the addresses indicated in that field and not to
1417 the address(es) indicated in the "From" field.
1418
1419 So we check a Sender field if there is one, else a Reply_to field, else a From
1420 field. As some strange messages may have more than one of these fields,
1421 especially if they are resent- fields, check all of them if there is more than
1422 one.
1423
1424 Arguments:
1425 user_msgptr points to where to put a user error message
1426 log_msgptr points to where to put a log error message
1427 callout timeout for callout check (passed to verify_address())
1428 callout_overall overall callout timeout (ditto)
1429 se_mailfrom mailfrom for verify; NULL => ""
1430 pm_mailfrom sender for pm callout check (passed to verify_address())
1431 options callout options (passed to verify_address())
1432
1433 If log_msgptr is set to something without setting user_msgptr, the caller
1434 normally uses log_msgptr for both things.
1435
1436 Returns: result of the verification attempt: OK, FAIL, or DEFER;
1437 FAIL is given if no appropriate headers are found
1438 */
1439
1440 int
1441 verify_check_header_address(uschar **user_msgptr, uschar **log_msgptr,
1442 int callout, int callout_overall, uschar *se_mailfrom, uschar *pm_mailfrom,
1443 int options)
1444 {
1445 static int header_types[] = { htype_sender, htype_reply_to, htype_from };
1446 int yield = FAIL;
1447 int i;
1448
1449 for (i = 0; i < 3; i++)
1450 {
1451 header_line *h;
1452 for (h = header_list; h != NULL; h = h->next)
1453 {
1454 int terminator, new_ok;
1455 uschar *s, *ss, *endname;
1456
1457 if (h->type != header_types[i]) continue;
1458 s = endname = Ustrchr(h->text, ':') + 1;
1459
1460 while (*s != 0)
1461 {
1462 address_item *vaddr;
1463
1464 while (isspace(*s) || *s == ',') s++;
1465 if (*s == 0) break; /* End of header */
1466
1467 ss = parse_find_address_end(s, FALSE);
1468
1469 /* The terminator is a comma or end of header, but there may be white
1470 space preceding it (including newline for the last address). Move back
1471 past any white space so we can check against any cached envelope sender
1472 address verifications. */
1473
1474 while (isspace(ss[-1])) ss--;
1475 terminator = *ss;
1476 *ss = 0;
1477
1478 HDEBUG(D_verify) debug_printf("verifying %.*s header address %s\n",
1479 (int)(endname - h->text), h->text, s);
1480
1481 /* See if we have already verified this address as an envelope sender,
1482 and if so, use the previous answer. */
1483
1484 vaddr = verify_checked_sender(s);
1485
1486 if (vaddr != NULL && /* Previously checked */
1487 (callout <= 0 || /* No callout needed; OR */
1488 vaddr->special_action > 256)) /* Callout was done */
1489 {
1490 new_ok = vaddr->special_action & 255;
1491 HDEBUG(D_verify) debug_printf("previously checked as envelope sender\n");
1492 *ss = terminator; /* Restore shortened string */
1493 }
1494
1495 /* Otherwise we run the verification now. We must restore the shortened
1496 string before running the verification, so the headers are correct, in
1497 case there is any rewriting. */
1498
1499 else
1500 {
1501 int start, end, domain;
1502 uschar *address = parse_extract_address(s, log_msgptr, &start,
1503 &end, &domain, FALSE);
1504
1505 *ss = terminator;
1506
1507 /* If verification failed because of a syntax error, fail this
1508 function, and ensure that the failing address gets added to the error
1509 message. */
1510
1511 if (address == NULL)
1512 {
1513 new_ok = FAIL;
1514 if (*log_msgptr != NULL)
1515 {
1516 while (ss > s && isspace(ss[-1])) ss--;
1517 *log_msgptr = string_sprintf("syntax error in '%.*s' header when "
1518 "scanning for sender: %s in \"%.*s\"",
1519 endname - h->text, h->text, *log_msgptr, ss - s, s);
1520 return FAIL;
1521 }
1522 }
1523
1524 /* Else go ahead with the sender verification. But is isn't *the*
1525 sender of the message, so set vopt_fake_sender to stop sender_address
1526 being replaced after rewriting or qualification. */
1527
1528 else
1529 {
1530 vaddr = deliver_make_addr(address, FALSE);
1531 new_ok = verify_address(vaddr, NULL, options | vopt_fake_sender,
1532 callout, callout_overall, se_mailfrom, pm_mailfrom, NULL);
1533 }
1534 }
1535
1536 /* We now have the result, either newly found, or cached. If we are
1537 giving out error details, set a specific user error. This means that the
1538 last of these will be returned to the user if all three fail. We do not
1539 set a log message - the generic one below will be used. */
1540
1541 if (new_ok != OK && smtp_return_error_details)
1542 {
1543 *user_msgptr = string_sprintf("Rejected after DATA: "
1544 "could not verify \"%.*s\" header address\n%s: %s",
1545 endname - h->text, h->text, vaddr->address, vaddr->message);
1546 }
1547
1548 /* Success or defer */
1549
1550 if (new_ok == OK) return OK;
1551 if (new_ok == DEFER) yield = DEFER;
1552
1553 /* Move on to any more addresses in the header */
1554
1555 s = ss;
1556 }
1557 }
1558 }
1559
1560 if (yield == FAIL && *log_msgptr == NULL)
1561 *log_msgptr = US"there is no valid sender in any header line";
1562
1563 if (yield == DEFER && *log_msgptr == NULL)
1564 *log_msgptr = US"all attempts to verify a sender in a header line deferred";
1565
1566 return yield;
1567 }
1568
1569
1570
1571
1572 /*************************************************
1573 * Get RFC 1413 identification *
1574 *************************************************/
1575
1576 /* Attempt to get an id from the sending machine via the RFC 1413 protocol. If
1577 the timeout is set to zero, then the query is not done. There may also be lists
1578 of hosts and nets which are exempt. To guard against malefactors sending
1579 non-printing characters which could, for example, disrupt a message's headers,
1580 make sure the string consists of printing characters only.
1581
1582 Argument:
1583 port the port to connect to; usually this is IDENT_PORT (113), but when
1584 running in the test harness with -bh a different value is used.
1585
1586 Returns: nothing
1587
1588 Side effect: any received ident value is put in sender_ident (NULL otherwise)
1589 */
1590
1591 void
1592 verify_get_ident(int port)
1593 {
1594 int sock, host_af, qlen;
1595 int received_sender_port, received_interface_port, n;
1596 uschar *p;
1597 uschar buffer[2048];
1598
1599 /* Default is no ident. Check whether we want to do an ident check for this
1600 host. */
1601
1602 sender_ident = NULL;
1603 if (rfc1413_query_timeout <= 0 || verify_check_host(&rfc1413_hosts) != OK)
1604 return;
1605
1606 DEBUG(D_ident) debug_printf("doing ident callback\n");
1607
1608 /* Set up a connection to the ident port of the remote host. Bind the local end
1609 to the incoming interface address. If the sender host address is an IPv6
1610 address, the incoming interface address will also be IPv6. */
1611
1612 host_af = (Ustrchr(sender_host_address, ':') == NULL)? AF_INET : AF_INET6;
1613 sock = ip_socket(SOCK_STREAM, host_af);
1614 if (sock < 0) return;
1615
1616 if (ip_bind(sock, host_af, interface_address, 0) < 0)
1617 {
1618 DEBUG(D_ident) debug_printf("bind socket for ident failed: %s\n",
1619 strerror(errno));
1620 goto END_OFF;
1621 }
1622
1623 if (ip_connect(sock, host_af, sender_host_address, port, rfc1413_query_timeout)
1624 < 0)
1625 {
1626 if (errno == ETIMEDOUT && (log_extra_selector & LX_ident_timeout) != 0)
1627 {
1628 log_write(0, LOG_MAIN, "ident connection to %s timed out",
1629 sender_host_address);
1630 }
1631 else
1632 {
1633 DEBUG(D_ident) debug_printf("ident connection to %s failed: %s\n",
1634 sender_host_address, strerror(errno));
1635 }
1636 goto END_OFF;
1637 }
1638
1639 /* Construct and send the query. */
1640
1641 sprintf(CS buffer, "%d , %d\r\n", sender_host_port, interface_port);
1642 qlen = Ustrlen(buffer);
1643 if (send(sock, buffer, qlen, 0) < 0)
1644 {
1645 DEBUG(D_ident) debug_printf("ident send failed: %s\n", strerror(errno));
1646 goto END_OFF;
1647 }
1648
1649 /* Read a response line. We put it into the rest of the buffer, using several
1650 recv() calls if necessary. */
1651
1652 p = buffer + qlen;
1653
1654 for (;;)
1655 {
1656 uschar *pp;
1657 int count;
1658 int size = sizeof(buffer) - (p - buffer);
1659
1660 if (size <= 0) goto END_OFF; /* Buffer filled without seeing \n. */
1661 count = ip_recv(sock, p, size, rfc1413_query_timeout);
1662 if (count <= 0) goto END_OFF; /* Read error or EOF */
1663
1664 /* Scan what we just read, to see if we have reached the terminating \r\n. Be
1665 generous, and accept a plain \n terminator as well. The only illegal
1666 character is 0. */
1667
1668 for (pp = p; pp < p + count; pp++)
1669 {
1670 if (*pp == 0) goto END_OFF; /* Zero octet not allowed */
1671 if (*pp == '\n')
1672 {
1673 if (pp[-1] == '\r') pp--;
1674 *pp = 0;
1675 goto GOT_DATA; /* Break out of both loops */
1676 }
1677 }
1678
1679 /* Reached the end of the data without finding \n. Let the loop continue to
1680 read some more, if there is room. */
1681
1682 p = pp;
1683 }
1684
1685 GOT_DATA:
1686
1687 /* We have received a line of data. Check it carefully. It must start with the
1688 same two port numbers that we sent, followed by data as defined by the RFC. For
1689 example,
1690
1691 12345 , 25 : USERID : UNIX :root
1692
1693 However, the amount of white space may be different to what we sent. In the
1694 "osname" field there may be several sub-fields, comma separated. The data we
1695 actually want to save follows the third colon. Some systems put leading spaces
1696 in it - we discard those. */
1697
1698 if (sscanf(CS buffer + qlen, "%d , %d%n", &received_sender_port,
1699 &received_interface_port, &n) != 2 ||
1700 received_sender_port != sender_host_port ||
1701 received_interface_port != interface_port)
1702 goto END_OFF;
1703
1704 p = buffer + qlen + n;
1705 while(isspace(*p)) p++;
1706 if (*p++ != ':') goto END_OFF;
1707 while(isspace(*p)) p++;
1708 if (Ustrncmp(p, "USERID", 6) != 0) goto END_OFF;
1709 p += 6;
1710 while(isspace(*p)) p++;
1711 if (*p++ != ':') goto END_OFF;
1712 while (*p != 0 && *p != ':') p++;
1713 if (*p++ == 0) goto END_OFF;
1714 while(isspace(*p)) p++;
1715 if (*p == 0) goto END_OFF;
1716
1717 /* The rest of the line is the data we want. We turn it into printing
1718 characters when we save it, so that it cannot mess up the format of any logging
1719 or Received: lines into which it gets inserted. We keep a maximum of 127
1720 characters. */
1721
1722 sender_ident = string_printing(string_copyn(p, 127));
1723 DEBUG(D_ident) debug_printf("sender_ident = %s\n", sender_ident);
1724
1725 END_OFF:
1726 close(sock);
1727 return;
1728 }
1729
1730
1731
1732
1733 /*************************************************
1734 * Match host to a single host-list item *
1735 *************************************************/
1736
1737 /* This function compares a host (name or address) against a single item
1738 from a host list. The host name gets looked up if it is needed and is not
1739 already known. The function is called from verify_check_this_host() via
1740 match_check_list(), which is why most of its arguments are in a single block.
1741
1742 Arguments:
1743 arg the argument block (see below)
1744 ss the host-list item
1745 valueptr where to pass back looked up data, or NULL
1746 error for error message when returning ERROR
1747
1748 The block contains:
1749 host_name the host name or NULL, implying use sender_host_name and
1750 sender_host_aliases, looking them up if required
1751 host_address the host address
1752 host_ipv4 the IPv4 address taken from an IPv6 one
1753
1754 Returns: OK matched
1755 FAIL did not match
1756 DEFER lookup deferred
1757 ERROR failed to find the host name or IP address
1758 unknown lookup type specified
1759 */
1760
1761 static int
1762 check_host(void *arg, uschar *ss, uschar **valueptr, uschar **error)
1763 {
1764 check_host_block *cb = (check_host_block *)arg;
1765 int maskoffset;
1766 BOOL isquery = FALSE;
1767 uschar *semicolon, *t;
1768 uschar **aliases;
1769
1770 /* Optimize for the special case when the pattern is "*". */
1771
1772 if (*ss == '*' && ss[1] == 0) return OK;
1773
1774 /* If the pattern is empty, it matches only in the case when there is no host -
1775 this can occur in ACL checking for SMTP input using the -bs option. In this
1776 situation, the host address is the empty string. */
1777
1778 if (cb->host_address[0] == 0) return (*ss == 0)? OK : FAIL;
1779 if (*ss == 0) return FAIL;
1780
1781 /* If the pattern is precisely "@" then match against the primary host name;
1782 if it's "@[]" match against the local host's IP addresses. */
1783
1784 if (*ss == '@')
1785 {
1786 if (ss[1] == 0) ss = primary_hostname;
1787 else if (Ustrcmp(ss, "@[]") == 0)
1788 {
1789 ip_address_item *ip;
1790 for (ip = host_find_interfaces(); ip != NULL; ip = ip->next)
1791 if (Ustrcmp(ip->address, cb->host_address) == 0) return OK;
1792 return FAIL;
1793 }
1794 }
1795
1796 /* If the pattern is an IP address, optionally followed by a bitmask count, do
1797 a (possibly masked) comparision with the current IP address. */
1798
1799 if (string_is_ip_address(ss, &maskoffset))
1800 return (host_is_in_net(cb->host_address, ss, maskoffset)? OK : FAIL);
1801
1802 /* If the item is of the form net[n]-lookup;<file|query> then it is a lookup on
1803 a masked IP network, in textual form. The net- stuff really only applies to
1804 single-key lookups where the key is implicit. For query-style lookups the key
1805 is specified in the query. From release 4.30, the use of net- for query style
1806 is no longer needed, but we retain it for backward compatibility. */
1807
1808 if (Ustrncmp(ss, "net", 3) == 0 && (semicolon = Ustrchr(ss, ';')) != NULL)
1809 {
1810 int mlen = 0;
1811 for (t = ss + 3; isdigit(*t); t++) mlen = mlen * 10 + *t - '0';
1812 if (*t++ == '-')
1813 {
1814 int insize;
1815 int search_type;
1816 int incoming[4];
1817 void *handle;
1818 uschar *filename, *key, *result;
1819 uschar buffer[64];
1820
1821 /* If no mask was supplied, set a negative value */
1822
1823 if (mlen == 0 && t == ss+4) mlen = -1;
1824
1825 /* Find the search type */
1826
1827 search_type = search_findtype(t, semicolon - t);
1828
1829 if (search_type < 0) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
1830 search_error_message);
1831
1832 /* Adjust parameters for the type of lookup. For a query-style
1833 lookup, there is no file name, and the "key" is just the query. For
1834 a single-key lookup, the key is the current IP address, masked
1835 appropriately, and reconverted to text form, with the mask appended. */
1836
1837 if (mac_islookup(search_type, lookup_querystyle))
1838 {
1839 filename = NULL;
1840 key = semicolon + 1;
1841 }
1842 else
1843 {
1844 insize = host_aton(cb->host_address, incoming);
1845 host_mask(insize, incoming, mlen);
1846 (void)host_nmtoa(insize, incoming, mlen, buffer);
1847 key = buffer;
1848 filename = semicolon + 1;
1849 }
1850
1851 /* Now do the actual lookup; note that there is no search_close() because
1852 of the caching arrangements. */
1853
1854 handle = search_open(filename, search_type, 0, NULL, NULL);
1855 if (handle == NULL) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
1856 search_error_message);
1857 result = search_find(handle, filename, key, -1, NULL, 0, 0, NULL);
1858 if (valueptr != NULL) *valueptr = result;
1859 return (result != NULL)? OK : search_find_defer? DEFER: FAIL;
1860 }
1861 }
1862
1863 /* The pattern is not an IP address or network reference of any kind. That is,
1864 it is a host name pattern. Check the characters of the pattern to see if they
1865 comprise only letters, digits, full stops, and hyphens (the constituents of
1866 domain names). Allow underscores, as they are all too commonly found. Sigh.
1867 Also, if allow_utf8_domains is set, allow top-bit characters. */
1868
1869 for (t = ss; *t != 0; t++)
1870 if (!isalnum(*t) && *t != '.' && *t != '-' && *t != '_' &&
1871 (!allow_utf8_domains || *t < 128)) break;
1872
1873 /* If the pattern is a complete domain name, with no fancy characters, look up
1874 its IP address and match against that. Note that a multi-homed host will add
1875 items to the chain. */
1876
1877 if (*t == 0)
1878 {
1879 int rc;
1880 host_item h;
1881 h.next = NULL;
1882 h.name = ss;
1883 h.address = NULL;
1884 h.mx = MX_NONE;
1885 rc = host_find_byname(&h, NULL, NULL, FALSE);
1886 if (rc == HOST_FOUND || rc == HOST_FOUND_LOCAL)
1887 {
1888 host_item *hh;
1889 for (hh = &h; hh != NULL; hh = hh->next)
1890 {
1891 if (Ustrcmp(hh->address, (Ustrchr(hh->address, ':') == NULL)?
1892 cb->host_ipv4 : cb->host_address) == 0)
1893 return OK;
1894 }
1895 return FAIL;
1896 }
1897 if (rc == HOST_FIND_AGAIN) return DEFER;
1898 *error = string_sprintf("failed to find IP address for %s", ss);
1899 return ERROR;
1900 }
1901
1902 /* Almost all subsequent comparisons require the host name, and can be done
1903 using the general string matching function. When this function is called for
1904 outgoing hosts, the name is always given explicitly. If it is NULL, it means we
1905 must use sender_host_name and its aliases, looking them up if necessary. */
1906
1907 if (cb->host_name != NULL) /* Explicit host name given */
1908 return match_check_string(cb->host_name, ss, -1, TRUE, TRUE, TRUE,
1909 valueptr);
1910
1911 /* Host name not given; in principle we need the sender host name and its
1912 aliases. However, for query-style lookups, we do not need the name if the
1913 query does not contain $sender_host_name. From release 4.23, a reference to
1914 $sender_host_name causes it to be looked up, so we don't need to do the lookup
1915 on spec. */
1916
1917 if ((semicolon = Ustrchr(ss, ';')) != NULL)
1918 {
1919 uschar *affix;
1920 int partial, affixlen, starflags, id;
1921
1922 *semicolon = 0;
1923 id = search_findtype_partial(ss, &partial, &affix, &affixlen, &starflags);
1924 *semicolon=';';
1925
1926 if (id < 0) /* Unknown lookup type */
1927 {
1928 log_write(0, LOG_MAIN|LOG_PANIC, "%s in host list item \"%s\"",
1929 search_error_message, ss);
1930 return DEFER;
1931 }
1932 isquery = mac_islookup(id, lookup_querystyle);
1933 }
1934
1935 if (isquery)
1936 {
1937 switch(match_check_string(US"", ss, -1, TRUE, TRUE, TRUE, valueptr))
1938 {
1939 case OK: return OK;
1940 case DEFER: return DEFER;
1941 default: return FAIL;
1942 }
1943 }
1944
1945 /* Not a query-style lookup; must ensure the host name is present, and then we
1946 do a check on the name and all its aliases. */
1947
1948 if (sender_host_name == NULL)
1949 {
1950 HDEBUG(D_host_lookup)
1951 debug_printf("sender host name required, to match against %s\n", ss);
1952 if (host_lookup_failed || host_name_lookup() != OK)
1953 {
1954 *error = string_sprintf("failed to find host name for %s",
1955 sender_host_address);;
1956 return ERROR;
1957 }
1958 host_build_sender_fullhost();
1959 }
1960
1961 /* Match on the sender host name, using the general matching function */
1962
1963 switch(match_check_string(sender_host_name, ss, -1, TRUE, TRUE, TRUE,
1964 valueptr))
1965 {
1966 case OK: return OK;
1967 case DEFER: return DEFER;
1968 }
1969
1970 /* If there are aliases, try matching on them. */
1971
1972 aliases = sender_host_aliases;
1973 while (*aliases != NULL)
1974 {
1975 switch(match_check_string(*aliases++, ss, -1, TRUE, TRUE, TRUE, valueptr))
1976 {
1977 case OK: return OK;
1978 case DEFER: return DEFER;
1979 }
1980 }
1981 return FAIL;
1982 }
1983
1984
1985
1986
1987 /*************************************************
1988 * Check a specific host matches a host list *
1989 *************************************************/
1990
1991 /* This function is passed a host list containing items in a number of
1992 different formats and the identity of a host. Its job is to determine whether
1993 the given host is in the set of hosts defined by the list. The host name is
1994 passed as a pointer so that it can be looked up if needed and not already
1995 known. This is commonly the case when called from verify_check_host() to check
1996 an incoming connection. When called from elsewhere the host name should usually
1997 be set.
1998
1999 This function is now just a front end to match_check_list(), which runs common
2000 code for scanning a list. We pass it the check_host() function to perform a
2001 single test.
2002
2003 Arguments:
2004 listptr pointer to the host list
2005 cache_bits pointer to cache for named lists, or NULL
2006 host_name the host name or NULL, implying use sender_host_name and
2007 sender_host_aliases, looking them up if required
2008 host_address the IP address
2009 valueptr if not NULL, data from a lookup is passed back here
2010
2011 Returns: OK if the host is in the defined set
2012 FAIL if the host is not in the defined set,
2013 DEFER if a data lookup deferred (not a host lookup)
2014
2015 If the host name was needed in order to make a comparison, and could not be
2016 determined from the IP address, the result is FAIL unless the item
2017 "+allow_unknown" was met earlier in the list, in which case OK is returned. */
2018
2019 int
2020 verify_check_this_host(uschar **listptr, unsigned int *cache_bits,
2021 uschar *host_name, uschar *host_address, uschar **valueptr)
2022 {
2023 unsigned int *local_cache_bits = cache_bits;
2024 check_host_block cb;
2025 cb.host_name = host_name;
2026 cb.host_address = host_address;
2027
2028 if (valueptr != NULL) *valueptr = NULL;
2029
2030 /* If the host address starts off ::ffff: it is an IPv6 address in
2031 IPv4-compatible mode. Find the IPv4 part for checking against IPv4
2032 addresses. */
2033
2034 cb.host_ipv4 = (Ustrncmp(host_address, "::ffff:", 7) == 0)?
2035 host_address + 7 : host_address;
2036
2037 return match_check_list(listptr, 0, &hostlist_anchor, &local_cache_bits,
2038 check_host, &cb, MCL_HOST,
2039 (host_address == sender_host_address)? US"host" : host_address, valueptr);
2040 }
2041
2042
2043
2044
2045 /*************************************************
2046 * Check the remote host matches a list *
2047 *************************************************/
2048
2049 /* This is a front end to verify_check_this_host(), created because checking
2050 the remote host is a common occurrence. With luck, a good compiler will spot
2051 the tail recursion and optimize it. If there's no host address, this is
2052 command-line SMTP input - check against an empty string for the address.
2053
2054 Arguments:
2055 listptr pointer to the host list
2056
2057 Returns: the yield of verify_check_this_host(),
2058 i.e. OK, FAIL, or DEFER
2059 */
2060
2061 int
2062 verify_check_host(uschar **listptr)
2063 {
2064 return verify_check_this_host(listptr, sender_host_cache, NULL,
2065 (sender_host_address == NULL)? US"" : sender_host_address, NULL);
2066 }
2067
2068
2069
2070
2071
2072 /*************************************************
2073 * Invert an IP address for a DNS black list *
2074 *************************************************/
2075
2076 /*
2077 Arguments:
2078 buffer where to put the answer
2079 address the address to invert
2080 */
2081
2082 static void
2083 invert_address(uschar *buffer, uschar *address)
2084 {
2085 int bin[4];
2086 uschar *bptr = buffer;
2087
2088 /* If this is an IPv4 address mapped into IPv6 format, adjust the pointer
2089 to the IPv4 part only. */
2090
2091 if (Ustrncmp(address, "::ffff:", 7) == 0) address += 7;
2092
2093 /* Handle IPv4 address: when HAVE_IPV6 is false, the result of host_aton() is
2094 always 1. */
2095
2096 if (host_aton(address, bin) == 1)
2097 {
2098 int i;
2099 int x = bin[0];
2100 for (i = 0; i < 4; i++)
2101 {
2102 sprintf(CS bptr, "%d.", x & 255);
2103 while (*bptr) bptr++;
2104 x >>= 8;
2105 }
2106 }
2107
2108 /* Handle IPv6 address. Actually, as far as I know, there are no IPv6 addresses
2109 in any DNS black lists, and the format in which they will be looked up is
2110 unknown. This is just a guess. */
2111
2112 #if HAVE_IPV6
2113 else
2114 {
2115 int i, j;
2116 for (j = 3; j >= 0; j--)
2117 {
2118 int x = bin[j];
2119 for (i = 0; i < 8; i++)
2120 {
2121 sprintf(CS bptr, "%x.", x & 15);
2122 while (*bptr) bptr++;
2123 x >>= 4;
2124 }
2125 }
2126 }
2127 #endif
2128 }
2129
2130
2131
2132 /*************************************************
2133 * Check host against DNS black lists *
2134 *************************************************/
2135
2136 /* This function runs checks against a list of DNS black lists, until one
2137 matches. Each item on the list can be of the form
2138
2139 domain=ip-address/key
2140
2141 The domain is the right-most domain that is used for the query, for example,
2142 blackholes.mail-abuse.org. If the IP address is present, there is a match only
2143 if the DNS lookup returns a matching IP address. Several addresses may be
2144 given, comma-separated, for example: x.y.z=127.0.0.1,127.0.0.2.
2145
2146 If no key is given, what is looked up in the domain is the inverted IP address
2147 of the current client host. If a key is given, it is used to construct the
2148 domain for the lookup. For example,
2149
2150 dsn.rfc-ignorant.org/$sender_address_domain
2151
2152 After finding a match in the DNS, the domain is placed in $dnslist_domain, and
2153 then we check for a TXT record for an error message, and if found, save its
2154 value in $dnslist_text. We also cache everything in a tree, to optimize
2155 multiple lookups.
2156
2157 Note: an address for testing RBL is 192.203.178.39
2158 Note: an address for testing DUL is 192.203.178.4
2159 Note: a domain for testing RFCI is example.tld.dsn.rfc-ignorant.org
2160
2161 Arguments:
2162 listptr the domain/address/data list
2163
2164 Returns: OK successful lookup (i.e. the address is on the list), or
2165 lookup deferred after +include_unknown
2166 FAIL name not found, or no data found for the given type, or
2167 lookup deferred after +exclude_unknown (default)
2168 DEFER lookup failure, if +defer_unknown was set
2169 */
2170
2171 int
2172 verify_check_dnsbl(uschar **listptr)
2173 {
2174 int sep = 0;
2175 int defer_return = FAIL;
2176 int old_pool = store_pool;
2177 BOOL invert_result = FALSE;
2178 uschar *list = *listptr;
2179 uschar *domain;
2180 uschar *s;
2181 uschar buffer[1024];
2182 uschar query[256]; /* DNS domain max length */
2183 uschar revadd[128]; /* Long enough for IPv6 address */
2184
2185 /* Indicate that the inverted IP address is not yet set up */
2186
2187 revadd[0] = 0;
2188
2189 /* Loop through all the domains supplied, until something matches */
2190
2191 while ((domain = string_nextinlist(&list, &sep, buffer, sizeof(buffer))) != NULL)
2192 {
2193 BOOL frc;
2194 BOOL bitmask = FALSE;
2195 dns_answer dnsa;
2196 dns_scan dnss;
2197 uschar *iplist;
2198 uschar *key;
2199 tree_node *t;
2200 dnsbl_cache_block *cb;
2201
2202 HDEBUG(D_dnsbl) debug_printf("DNS list check: %s\n", domain);
2203
2204 /* Deal with special values that change the behaviour on defer */
2205
2206 if (domain[0] == '+')
2207 {
2208 if (strcmpic(domain, US"+include_unknown") == 0) defer_return = OK;
2209 else if (strcmpic(domain, US"+exclude_unknown") == 0) defer_return = FAIL;
2210 else if (strcmpic(domain, US"+defer_unknown") == 0) defer_return = DEFER;
2211 else
2212 log_write(0, LOG_MAIN|LOG_PANIC, "unknown item in dnslist (ignored): %s",
2213 domain);
2214 continue;
2215 }
2216
2217 /* See if there's explicit data to be looked up */
2218
2219 key = Ustrchr(domain, '/');
2220 if (key != NULL) *key++ = 0;
2221
2222 /* See if there's a list of addresses supplied after the domain name. This is
2223 introduced by an = or a & character; if preceded by ! we invert the result.
2224 */
2225
2226 iplist = Ustrchr(domain, '=');
2227 if (iplist == NULL)
2228 {
2229 bitmask = TRUE;
2230 iplist = Ustrchr(domain, '&');
2231 }
2232
2233 if (iplist != NULL)
2234 {
2235 if (iplist > domain && iplist[-1] == '!')
2236 {
2237 invert_result = TRUE;
2238 iplist[-1] = 0;
2239 }
2240 *iplist++ = 0;
2241 }
2242
2243 /* Check that what we have left is a sensible domain name. There is no reason
2244 why these domains should in fact use the same syntax as hosts and email
2245 domains, but in practice they seem to. However, there is little point in
2246 actually causing an error here, because that would no doubt hold up incoming
2247 mail. Instead, I'll just log it. */
2248
2249 for (s = domain; *s != 0; s++)
2250 {
2251 if (!isalnum(*s) && *s != '-' && *s != '.')
2252 {
2253 log_write(0, LOG_MAIN, "dnslists domain \"%s\" contains "
2254 "strange characters - is this right?", domain);
2255 break;
2256 }
2257 }
2258
2259 /* Construct the query by adding the domain onto either the sending host
2260 address, or the given key string. */
2261
2262 if (key == NULL)
2263 {
2264 if (sender_host_address == NULL) return FAIL; /* can never match */
2265 if (revadd[0] == 0) invert_address(revadd, sender_host_address);
2266 frc = string_format(query, sizeof(query), "%s%s", revadd, domain);
2267 }
2268 else
2269 {
2270 frc = string_format(query, sizeof(query), "%s.%s", key, domain);
2271 }
2272
2273 if (!frc)
2274 {
2275 log_write(0, LOG_MAIN|LOG_PANIC, "dnslist query is too long "
2276 "(ignored): %s...", query);
2277 continue;
2278 }
2279
2280 /* Look for this query in the cache. */
2281
2282 t = tree_search(dnsbl_cache, query);
2283
2284 /* If not cached from a previous lookup, we must do a DNS lookup, and
2285 cache the result in permanent memory. */
2286
2287 if (t == NULL)
2288 {
2289 store_pool = POOL_PERM;
2290
2291 /* In case this is the first time the DNS resolver is being used. */
2292
2293 dns_init(FALSE, FALSE);
2294
2295 /* Set up a tree entry to cache the lookup */
2296
2297 t = store_get(sizeof(tree_node) + Ustrlen(query));
2298 Ustrcpy(t->name, query);
2299 t->data.ptr = cb = store_get(sizeof(dnsbl_cache_block));
2300 (void)tree_insertnode(&dnsbl_cache, t);
2301
2302 /* Do the DNS loopup . */
2303
2304 HDEBUG(D_dnsbl) debug_printf("new DNS lookup for %s\n", query);
2305 cb->rc = dns_basic_lookup(&dnsa, query, T_A);
2306 cb->text_set = FALSE;
2307 cb->text = NULL;
2308 cb->rhs = NULL;
2309
2310 /* If the lookup succeeded, cache the RHS address. The code allows for
2311 more than one address - this was for complete generality and the possible
2312 use of A6 records. However, A6 records have been reduced to experimental
2313 status (August 2001) and may die out. So they may never get used at all,
2314 let alone in dnsbl records. However, leave the code here, just in case.
2315
2316 Quite apart from one A6 RR generating multiple addresses, there are DNS
2317 lists that return more than one A record, so we must handle multiple
2318 addresses generated in that way as well. */
2319
2320 if (cb->rc == DNS_SUCCEED)
2321 {
2322 dns_record *rr;
2323 dns_address **addrp = &(cb->rhs);
2324 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2325 rr != NULL;
2326 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2327 {
2328 if (rr->type == T_A)
2329 {
2330 dns_address *da = dns_address_from_rr(&dnsa, rr);
2331 if (da != NULL)
2332 {
2333 *addrp = da;
2334 while (da->next != NULL) da = da->next;
2335 addrp = &(da->next);
2336 }
2337 }
2338 }
2339
2340 /* If we didn't find any A records, change the return code. This can
2341 happen when there is a CNAME record but there are no A records for what
2342 it points to. */
2343
2344 if (cb->rhs == NULL) cb->rc = DNS_NODATA;
2345 }
2346
2347 store_pool = old_pool;
2348 }
2349
2350 /* Previous lookup was cached */
2351
2352 else
2353 {
2354 HDEBUG(D_dnsbl) debug_printf("using result of previous DNS lookup\n");
2355 cb = t->data.ptr;
2356 }
2357
2358 /* We now have the result of the DNS lookup, either newly done, or cached
2359 from a previous call. If the lookup succeeded, check against the address
2360 list if there is one. This may be a positive equality list (introduced by
2361 "="), a negative equality list (introduced by "!="), a positive bitmask
2362 list (introduced by "&"), or a negative bitmask list (introduced by "!&").*/
2363
2364 if (cb->rc == DNS_SUCCEED)
2365 {
2366 dns_address *da = NULL;
2367 uschar *addlist = cb->rhs->address;
2368
2369 /* For A and AAAA records, there may be multiple addresses from multiple
2370 records. For A6 records (currently not expected to be used) there may be
2371 multiple addresses from a single record. */
2372
2373 for (da = cb->rhs->next; da != NULL; da = da->next)
2374 addlist = string_sprintf("%s, %s", addlist, da->address);
2375
2376 HDEBUG(D_dnsbl) debug_printf("DNS lookup for %s succeeded (yielding %s)\n",
2377 query, addlist);
2378
2379 /* Address list check; this can be either for equality, or via a bitmask.
2380 In the latter case, all the bits must match. */
2381
2382 if (iplist != NULL)
2383 {
2384 int ipsep = ',';
2385 uschar ip[46];
2386 uschar *ptr = iplist;
2387
2388 while (string_nextinlist(&ptr, &ipsep, ip, sizeof(ip)) != NULL)
2389 {
2390 /* Handle exact matching */
2391 if (!bitmask)
2392 {
2393 for (da = cb->rhs; da != NULL; da = da->next)
2394 {
2395 if (Ustrcmp(CS da->address, ip) == 0) break;
2396 }
2397 }
2398 /* Handle bitmask matching */
2399 else
2400 {
2401 int address[4];
2402 int mask = 0;
2403
2404 /* At present, all known DNS blocking lists use A records, with
2405 IPv4 addresses on the RHS encoding the information they return. I
2406 wonder if this will linger on as the last vestige of IPv4 when IPv6
2407 is ubiquitous? Anyway, for now we use paranoia code to completely
2408 ignore IPv6 addresses. The default mask is 0, which always matches.
2409 We change this only for IPv4 addresses in the list. */
2410
2411 if (host_aton(ip, address) == 1) mask = address[0];
2412
2413 /* Scan the returned addresses, skipping any that are IPv6 */
2414
2415 for (da = cb->rhs; da != NULL; da = da->next)
2416 {
2417 if (host_aton(da->address, address) != 1) continue;
2418 if ((address[0] & mask) == mask) break;
2419 }
2420 }
2421
2422 /* Break out if a match has been found */
2423
2424 if (da != NULL) break;
2425 }
2426
2427 /* If either
2428
2429 (a) No IP address in a positive list matched, or
2430 (b) An IP address in a negative list did match
2431
2432 then behave as if the DNSBL lookup had not succeeded, i.e. the host is
2433 not on the list. */
2434
2435 if (invert_result != (da == NULL))
2436 {
2437 HDEBUG(D_dnsbl)
2438 {
2439 debug_printf("=> but we are not accepting this block class because\n");
2440 debug_printf("=> there was %s match for %c%s\n",
2441 invert_result? "an exclude":"no", bitmask? '&' : '=', iplist);
2442 }
2443 continue; /* With next DNSBL domain */
2444 }
2445 }
2446
2447 /* Either there was no IP list, or the record matched. Look up a TXT record
2448 if it hasn't previously been done. */
2449
2450 if (!cb->text_set)
2451 {
2452 cb->text_set = TRUE;
2453 if (dns_basic_lookup(&dnsa, query, T_TXT) == DNS_SUCCEED)
2454 {
2455 dns_record *rr;
2456 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2457 rr != NULL;
2458 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2459 if (rr->type == T_TXT) break;
2460 if (rr != NULL)
2461 {
2462 int len = (rr->data)[0];
2463 if (len > 511) len = 127;
2464 store_pool = POOL_PERM;
2465 cb->text = string_sprintf("%.*s", len, (const uschar *)(rr->data+1));
2466 store_pool = old_pool;
2467 }
2468 }
2469 }
2470
2471 HDEBUG(D_dnsbl)
2472 {
2473 debug_printf("=> that means %s is listed at %s\n",
2474 (key == NULL)? sender_host_address : key, domain);
2475 }
2476
2477 dnslist_domain = string_copy(domain);
2478 dnslist_value = addlist;
2479 dnslist_text = cb->text;
2480 return OK;
2481 }
2482
2483 /* There was a problem with the DNS lookup */
2484
2485 if (cb->rc != DNS_NOMATCH && cb->rc != DNS_NODATA)
2486 {
2487 log_write(L_dnslist_defer, LOG_MAIN,
2488 "DNS list lookup defer (probably timeout) for %s: %s", query,
2489 (defer_return == OK)? US"assumed in list" :
2490 (defer_return == FAIL)? US"assumed not in list" :
2491 US"returned DEFER");
2492 return defer_return;
2493 }
2494
2495 /* No entry was found in the DNS; continue for next domain */
2496
2497 HDEBUG(D_dnsbl)
2498 {
2499 debug_printf("DNS lookup for %s failed\n", query);
2500 debug_printf("=> that means %s is not listed at %s\n",
2501 (key == NULL)? sender_host_address : key, domain);
2502 }
2503 } /* Continue with next domain */
2504
2505 return FAIL;
2506 }
2507
2508 /* End of verify.c */