Added support for SQLite.
[exim.git] / src / src / verify.c
1 /* $Cambridge: exim/src/src/verify.c,v 1.24 2005/08/01 13:20:28 ph10 Exp $ */
2
3 /*************************************************
4 * Exim - an Internet mail transport agent *
5 *************************************************/
6
7 /* Copyright (c) University of Cambridge 1995 - 2005 */
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 BOOL done = FALSE;
152 uschar *address_key;
153 uschar *from_address;
154 uschar *random_local_part = NULL;
155 uschar **failure_ptr = is_recipient?
156 &recipient_verify_failure : &sender_verify_failure;
157 open_db dbblock;
158 open_db *dbm_file = NULL;
159 dbdata_callout_cache new_domain_record;
160 dbdata_callout_cache_address new_address_record;
161 host_item *host;
162 time_t callout_start_time;
163
164 new_domain_record.result = ccache_unknown;
165 new_domain_record.postmaster_result = ccache_unknown;
166 new_domain_record.random_result = ccache_unknown;
167
168 memset(&new_address_record, 0, sizeof(new_address_record));
169
170 /* For a recipient callout, the key used for the address cache record must
171 include the sender address if we are using the real sender in the callout,
172 because that may influence the result of the callout. */
173
174 address_key = addr->address;
175 from_address = US"";
176
177 if (is_recipient)
178 {
179 if ((options & vopt_callout_recipsender) != 0)
180 {
181 address_key = string_sprintf("%s/<%s>", addr->address, sender_address);
182 from_address = sender_address;
183 }
184 else if ((options & vopt_callout_recippmaster) != 0)
185 {
186 address_key = string_sprintf("%s/<postmaster@%s>", addr->address,
187 qualify_domain_sender);
188 from_address = string_sprintf("postmaster@%s", qualify_domain_sender);
189 }
190 }
191
192 /* For a sender callout, we must adjust the key if the mailfrom address is not
193 empty. */
194
195 else
196 {
197 from_address = (se_mailfrom == NULL)? US"" : se_mailfrom;
198 if (from_address[0] != 0)
199 address_key = string_sprintf("%s/<%s>", addr->address, from_address);
200 }
201
202 /* Open the callout cache database, it it exists, for reading only at this
203 stage, unless caching has been disabled. */
204
205 if (callout_no_cache)
206 {
207 HDEBUG(D_verify) debug_printf("callout cache: disabled by no_cache\n");
208 }
209 else if ((dbm_file = dbfn_open(US"callout", O_RDWR, &dbblock, FALSE)) == NULL)
210 {
211 HDEBUG(D_verify) debug_printf("callout cache: not available\n");
212 }
213
214 /* If a cache database is available see if we can avoid the need to do an
215 actual callout by making use of previously-obtained data. */
216
217 if (dbm_file != NULL)
218 {
219 dbdata_callout_cache_address *cache_address_record;
220 dbdata_callout_cache *cache_record = get_callout_cache_record(dbm_file,
221 addr->domain, US"domain",
222 callout_cache_domain_positive_expire,
223 callout_cache_domain_negative_expire);
224
225 /* If an unexpired cache record was found for this domain, see if the callout
226 process can be short-circuited. */
227
228 if (cache_record != NULL)
229 {
230 /* If an early command (up to and including MAIL FROM:<>) was rejected,
231 there is no point carrying on. The callout fails. */
232
233 if (cache_record->result == ccache_reject)
234 {
235 setflag(addr, af_verify_nsfail);
236 HDEBUG(D_verify)
237 debug_printf("callout cache: domain gave initial rejection, or "
238 "does not accept HELO or MAIL FROM:<>\n");
239 setflag(addr, af_verify_nsfail);
240 addr->user_message = US"(result of an earlier callout reused).";
241 yield = FAIL;
242 *failure_ptr = US"mail";
243 goto END_CALLOUT;
244 }
245
246 /* If a previous check on a "random" local part was accepted, we assume
247 that the server does not do any checking on local parts. There is therefore
248 no point in doing the callout, because it will always be successful. If a
249 random check previously failed, arrange not to do it again, but preserve
250 the data in the new record. If a random check is required but hasn't been
251 done, skip the remaining cache processing. */
252
253 if (callout_random) switch(cache_record->random_result)
254 {
255 case ccache_accept:
256 HDEBUG(D_verify)
257 debug_printf("callout cache: domain accepts random addresses\n");
258 goto END_CALLOUT; /* Default yield is OK */
259
260 case ccache_reject:
261 HDEBUG(D_verify)
262 debug_printf("callout cache: domain rejects random addresses\n");
263 callout_random = FALSE;
264 new_domain_record.random_result = ccache_reject;
265 new_domain_record.random_stamp = cache_record->random_stamp;
266 break;
267
268 default:
269 HDEBUG(D_verify)
270 debug_printf("callout cache: need to check random address handling "
271 "(not cached or cache expired)\n");
272 goto END_CACHE;
273 }
274
275 /* If a postmaster check is requested, but there was a previous failure,
276 there is again no point in carrying on. If a postmaster check is required,
277 but has not been done before, we are going to have to do a callout, so skip
278 remaining cache processing. */
279
280 if (pm_mailfrom != NULL)
281 {
282 if (cache_record->postmaster_result == ccache_reject)
283 {
284 setflag(addr, af_verify_pmfail);
285 HDEBUG(D_verify)
286 debug_printf("callout cache: domain does not accept "
287 "RCPT TO:<postmaster@domain>\n");
288 yield = FAIL;
289 *failure_ptr = US"postmaster";
290 setflag(addr, af_verify_pmfail);
291 addr->user_message = US"(result of earlier verification reused).";
292 goto END_CALLOUT;
293 }
294 if (cache_record->postmaster_result == ccache_unknown)
295 {
296 HDEBUG(D_verify)
297 debug_printf("callout cache: need to check RCPT "
298 "TO:<postmaster@domain> (not cached or cache expired)\n");
299 goto END_CACHE;
300 }
301
302 /* If cache says OK, set pm_mailfrom NULL to prevent a redundant
303 postmaster check if the address itself has to be checked. Also ensure
304 that the value in the cache record is preserved (with its old timestamp).
305 */
306
307 HDEBUG(D_verify) debug_printf("callout cache: domain accepts RCPT "
308 "TO:<postmaster@domain>\n");
309 pm_mailfrom = NULL;
310 new_domain_record.postmaster_result = ccache_accept;
311 new_domain_record.postmaster_stamp = cache_record->postmaster_stamp;
312 }
313 }
314
315 /* We can't give a result based on information about the domain. See if there
316 is an unexpired cache record for this specific address (combined with the
317 sender address if we are doing a recipient callout with a non-empty sender).
318 */
319
320 cache_address_record = (dbdata_callout_cache_address *)
321 get_callout_cache_record(dbm_file,
322 address_key, US"address",
323 callout_cache_positive_expire,
324 callout_cache_negative_expire);
325
326 if (cache_address_record != NULL)
327 {
328 if (cache_address_record->result == ccache_accept)
329 {
330 HDEBUG(D_verify)
331 debug_printf("callout cache: address record is positive\n");
332 }
333 else
334 {
335 HDEBUG(D_verify)
336 debug_printf("callout cache: address record is negative\n");
337 addr->user_message = US"Previous (cached) callout verification failure";
338 *failure_ptr = US"recipient";
339 yield = FAIL;
340 }
341 goto END_CALLOUT;
342 }
343
344 /* Close the cache database while we actually do the callout for real. */
345
346 END_CACHE:
347 dbfn_close(dbm_file);
348 dbm_file = NULL;
349 }
350
351 /* The information wasn't available in the cache, so we have to do a real
352 callout and save the result in the cache for next time, unless no_cache is set,
353 or unless we have a previously cached negative random result. If we are to test
354 with a random local part, ensure that such a local part is available. If not,
355 log the fact, but carry on without randomming. */
356
357 if (callout_random && callout_random_local_part != NULL)
358 {
359 random_local_part = expand_string(callout_random_local_part);
360 if (random_local_part == NULL)
361 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand "
362 "callout_random_local_part: %s", expand_string_message);
363 }
364
365 /* Default the connect and overall callout timeouts if not set, and record the
366 time we are starting so that we can enforce it. */
367
368 if (callout_overall < 0) callout_overall = 4 * callout;
369 if (callout_connect < 0) callout_connect = callout;
370 callout_start_time = time(NULL);
371
372 /* Now make connections to the hosts and do real callouts. The list of hosts
373 is passed in as an argument. */
374
375 for (host = host_list; host != NULL && !done; host = host->next)
376 {
377 smtp_inblock inblock;
378 smtp_outblock outblock;
379 int host_af;
380 int port = 25;
381 BOOL send_quit = TRUE;
382 uschar *helo = US"HELO";
383 uschar *interface = NULL; /* Outgoing interface to use; NULL => any */
384 uschar inbuffer[4096];
385 uschar outbuffer[1024];
386 uschar responsebuffer[4096];
387
388 clearflag(addr, af_verify_pmfail); /* postmaster callout flag */
389 clearflag(addr, af_verify_nsfail); /* null sender callout flag */
390
391 /* Skip this host if we don't have an IP address for it. */
392
393 if (host->address == NULL)
394 {
395 DEBUG(D_verify) debug_printf("no IP address for host name %s: skipping\n",
396 host->name);
397 continue;
398 }
399
400 /* Check the overall callout timeout */
401
402 if (time(NULL) - callout_start_time >= callout_overall)
403 {
404 HDEBUG(D_verify) debug_printf("overall timeout for callout exceeded\n");
405 break;
406 }
407
408 /* Set IPv4 or IPv6 */
409
410 host_af = (Ustrchr(host->address, ':') == NULL)? AF_INET:AF_INET6;
411
412 /* Expand and interpret the interface and port strings. This has to
413 be delayed till now, because they may expand differently for different
414 hosts. If there's a failure, log it, but carry on with the defaults. */
415
416 deliver_host = host->name;
417 deliver_host_address = host->address;
418 if (!smtp_get_interface(tf->interface, host_af, addr, NULL, &interface,
419 US"callout") ||
420 !smtp_get_port(tf->port, addr, &port, US"callout"))
421 log_write(0, LOG_MAIN|LOG_PANIC, "<%s>: %s", addr->address,
422 addr->message);
423 deliver_host = deliver_host_address = NULL;
424
425 /* Set HELO string according to the protocol */
426
427 if (Ustrcmp(tf->protocol, "lmtp") == 0) helo = US"LHLO";
428
429 HDEBUG(D_verify) debug_printf("interface=%s port=%d\n", interface, port);
430
431 /* Set up the buffer for reading SMTP response packets. */
432
433 inblock.buffer = inbuffer;
434 inblock.buffersize = sizeof(inbuffer);
435 inblock.ptr = inbuffer;
436 inblock.ptrend = inbuffer;
437
438 /* Set up the buffer for holding SMTP commands while pipelining */
439
440 outblock.buffer = outbuffer;
441 outblock.buffersize = sizeof(outbuffer);
442 outblock.ptr = outbuffer;
443 outblock.cmd_count = 0;
444 outblock.authenticating = FALSE;
445
446 /* Connect to the host; on failure, just loop for the next one, but we
447 set the error for the last one. Use the callout_connect timeout. */
448
449 inblock.sock = outblock.sock =
450 smtp_connect(host, host_af, port, interface, callout_connect, TRUE);
451 if (inblock.sock < 0)
452 {
453 addr->message = string_sprintf("could not connect to %s [%s]: %s",
454 host->name, host->address, strerror(errno));
455 continue;
456 }
457
458 /* Wait for initial response, and then run the initial SMTP commands. The
459 smtp_write_command() function leaves its command in big_buffer. This is
460 used in error responses. Initialize it in case the connection is
461 rejected. */
462
463 Ustrcpy(big_buffer, "initial connection");
464
465 done =
466 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
467 '2', callout) &&
468
469 smtp_write_command(&outblock, FALSE, "%s %s\r\n", helo,
470 smtp_active_hostname) >= 0 &&
471 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
472 '2', callout) &&
473
474 smtp_write_command(&outblock, FALSE, "MAIL FROM:<%s>\r\n",
475 from_address) >= 0 &&
476 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
477 '2', callout);
478
479 /* If the host gave an initial error, or does not accept HELO or MAIL
480 FROM:<>, arrange to cache this information, but don't record anything for an
481 I/O error or a defer. Do not cache rejections when a non-empty sender has
482 been used, because that blocks the whole domain for all senders. */
483
484 if (!done)
485 {
486 *failure_ptr = US"mail";
487 if (errno == 0 && responsebuffer[0] == '5')
488 {
489 setflag(addr, af_verify_nsfail);
490 if (from_address[0] == 0) new_domain_record.result = ccache_reject;
491 }
492 }
493
494 /* Otherwise, proceed to check a "random" address (if required), then the
495 given address, and the postmaster address (if required). Between each check,
496 issue RSET, because some servers accept only one recipient after MAIL
497 FROM:<>. */
498
499 else
500 {
501 new_domain_record.result = ccache_accept;
502
503 /* Do the random local part check first */
504
505 if (random_local_part != NULL)
506 {
507 uschar randombuffer[1024];
508 BOOL random_ok =
509 smtp_write_command(&outblock, FALSE,
510 "RCPT TO:<%.1000s@%.1000s>\r\n", random_local_part,
511 addr->domain) >= 0 &&
512 smtp_read_response(&inblock, randombuffer,
513 sizeof(randombuffer), '2', callout);
514
515 /* Remember when we last did a random test */
516
517 new_domain_record.random_stamp = time(NULL);
518
519 /* If accepted, we aren't going to do any further tests below. */
520
521 if (random_ok)
522 {
523 new_domain_record.random_result = ccache_accept;
524 }
525
526 /* Otherwise, cache a real negative response, and get back to the right
527 state to send RCPT. Unless there's some problem such as a dropped
528 connection, we expect to succeed, because the commands succeeded above. */
529
530 else if (errno == 0)
531 {
532 if (randombuffer[0] == '5')
533 new_domain_record.random_result = ccache_reject;
534
535 done =
536 smtp_write_command(&outblock, FALSE, "RSET\r\n") >= 0 &&
537 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
538 '2', callout) &&
539
540 smtp_write_command(&outblock, FALSE, "MAIL FROM:<%s>\r\n",
541 from_address) >= 0 &&
542 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
543 '2', callout);
544 }
545 else done = FALSE; /* Some timeout/connection problem */
546 } /* Random check */
547
548 /* If the host is accepting all local parts, as determined by the "random"
549 check, we don't need to waste time doing any further checking. */
550
551 if (new_domain_record.random_result != ccache_accept && done)
552 {
553 /* Get the rcpt_include_affixes flag from the transport if there is one,
554 but assume FALSE if there is not. */
555
556 done =
557 smtp_write_command(&outblock, FALSE, "RCPT TO:<%.1000s>\r\n",
558 transport_rcpt_address(addr,
559 (addr->transport == NULL)? FALSE :
560 addr->transport->rcpt_include_affixes)) >= 0 &&
561 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
562 '2', callout);
563
564 if (done)
565 new_address_record.result = ccache_accept;
566 else if (errno == 0 && responsebuffer[0] == '5')
567 {
568 *failure_ptr = US"recipient";
569 new_address_record.result = ccache_reject;
570 }
571
572 /* Do postmaster check if requested; if a full check is required, we
573 check for RCPT TO:<postmaster> (no domain) in accordance with RFC 821. */
574
575 if (done && pm_mailfrom != NULL)
576 {
577 done =
578 smtp_write_command(&outblock, FALSE, "RSET\r\n") >= 0 &&
579 smtp_read_response(&inblock, responsebuffer,
580 sizeof(responsebuffer), '2', callout) &&
581
582 smtp_write_command(&outblock, FALSE,
583 "MAIL FROM:<%s>\r\n", pm_mailfrom) >= 0 &&
584 smtp_read_response(&inblock, responsebuffer,
585 sizeof(responsebuffer), '2', callout) &&
586
587 /* First try using the current domain */
588
589 ((
590 smtp_write_command(&outblock, FALSE,
591 "RCPT TO:<postmaster@%.1000s>\r\n", addr->domain) >= 0 &&
592 smtp_read_response(&inblock, responsebuffer,
593 sizeof(responsebuffer), '2', callout)
594 )
595
596 ||
597
598 /* If that doesn't work, and a full check is requested,
599 try without the domain. */
600
601 (
602 (options & vopt_callout_fullpm) != 0 &&
603 smtp_write_command(&outblock, FALSE,
604 "RCPT TO:<postmaster>\r\n") >= 0 &&
605 smtp_read_response(&inblock, responsebuffer,
606 sizeof(responsebuffer), '2', callout)
607 ));
608
609 /* Sort out the cache record */
610
611 new_domain_record.postmaster_stamp = time(NULL);
612
613 if (done)
614 new_domain_record.postmaster_result = ccache_accept;
615 else if (errno == 0 && responsebuffer[0] == '5')
616 {
617 *failure_ptr = US"postmaster";
618 setflag(addr, af_verify_pmfail);
619 new_domain_record.postmaster_result = ccache_reject;
620 }
621 }
622 } /* Random not accepted */
623 } /* MAIL FROM: accepted */
624
625 /* For any failure of the main check, other than a negative response, we just
626 close the connection and carry on. We can identify a negative response by the
627 fact that errno is zero. For I/O errors it will be non-zero
628
629 Set up different error texts for logging and for sending back to the caller
630 as an SMTP response. Log in all cases, using a one-line format. For sender
631 callouts, give a full response to the caller, but for recipient callouts,
632 don't give the IP address because this may be an internal host whose identity
633 is not to be widely broadcast. */
634
635 if (!done)
636 {
637 if (errno == ETIMEDOUT)
638 {
639 HDEBUG(D_verify) debug_printf("SMTP timeout\n");
640 send_quit = FALSE;
641 }
642 else if (errno == 0)
643 {
644 if (*responsebuffer == 0) Ustrcpy(responsebuffer, US"connection dropped");
645
646 addr->message =
647 string_sprintf("response to \"%s\" from %s [%s] was: %s",
648 big_buffer, host->name, host->address,
649 string_printing(responsebuffer));
650
651 addr->user_message = is_recipient?
652 string_sprintf("Callout verification failed:\n%s", responsebuffer)
653 :
654 string_sprintf("Called: %s\nSent: %s\nResponse: %s",
655 host->address, big_buffer, responsebuffer);
656
657 /* Hard rejection ends the process */
658
659 if (responsebuffer[0] == '5') /* Address rejected */
660 {
661 yield = FAIL;
662 done = TRUE;
663 }
664 }
665 }
666
667 /* End the SMTP conversation and close the connection. */
668
669 if (send_quit) (void)smtp_write_command(&outblock, FALSE, "QUIT\r\n");
670 (void)close(inblock.sock);
671 } /* Loop through all hosts, while !done */
672
673 /* If we get here with done == TRUE, a successful callout happened, and yield
674 will be set OK or FAIL according to the response to the RCPT command.
675 Otherwise, we looped through the hosts but couldn't complete the business.
676 However, there may be domain-specific information to cache in both cases.
677
678 The value of the result field in the new_domain record is ccache_unknown if
679 there was an error before or with MAIL FROM:, and errno was not zero,
680 implying some kind of I/O error. We don't want to write the cache in that case.
681 Otherwise the value is ccache_accept or ccache_reject. */
682
683 if (!callout_no_cache && new_domain_record.result != ccache_unknown)
684 {
685 if ((dbm_file = dbfn_open(US"callout", O_RDWR|O_CREAT, &dbblock, FALSE))
686 == NULL)
687 {
688 HDEBUG(D_verify) debug_printf("callout cache: not available\n");
689 }
690 else
691 {
692 (void)dbfn_write(dbm_file, addr->domain, &new_domain_record,
693 (int)sizeof(dbdata_callout_cache));
694 HDEBUG(D_verify) debug_printf("wrote callout cache domain record:\n"
695 " result=%d postmaster=%d random=%d\n",
696 new_domain_record.result,
697 new_domain_record.postmaster_result,
698 new_domain_record.random_result);
699 }
700 }
701
702 /* If a definite result was obtained for the callout, cache it unless caching
703 is disabled. */
704
705 if (done)
706 {
707 if (!callout_no_cache && new_address_record.result != ccache_unknown)
708 {
709 if (dbm_file == NULL)
710 dbm_file = dbfn_open(US"callout", O_RDWR|O_CREAT, &dbblock, FALSE);
711 if (dbm_file == NULL)
712 {
713 HDEBUG(D_verify) debug_printf("no callout cache available\n");
714 }
715 else
716 {
717 (void)dbfn_write(dbm_file, address_key, &new_address_record,
718 (int)sizeof(dbdata_callout_cache_address));
719 HDEBUG(D_verify) debug_printf("wrote %s callout cache address record\n",
720 (new_address_record.result == ccache_accept)? "positive" : "negative");
721 }
722 }
723 } /* done */
724
725 /* Failure to connect to any host, or any response other than 2xx or 5xx is a
726 temporary error. If there was only one host, and a response was received, leave
727 it alone if supplying details. Otherwise, give a generic response. */
728
729 else /* !done */
730 {
731 uschar *dullmsg = string_sprintf("Could not complete %s verify callout",
732 is_recipient? "recipient" : "sender");
733 yield = DEFER;
734
735 if (host_list->next != NULL || addr->message == NULL) addr->message = dullmsg;
736
737 addr->user_message = (!smtp_return_error_details)? dullmsg :
738 string_sprintf("%s for <%s>.\n"
739 "The mail server(s) for the domain may be temporarily unreachable, or\n"
740 "they may be permanently unreachable from this server. In the latter case,\n%s",
741 dullmsg, addr->address,
742 is_recipient?
743 "the address will never be accepted."
744 :
745 "you need to change the address or create an MX record for its domain\n"
746 "if it is supposed to be generally accessible from the Internet.\n"
747 "Talk to your mail administrator for details.");
748
749 /* Force a specific error code */
750
751 addr->basic_errno = ERRNO_CALLOUTDEFER;
752 }
753
754 /* Come here from within the cache-reading code on fast-track exit. */
755
756 END_CALLOUT:
757 if (dbm_file != NULL) dbfn_close(dbm_file);
758 return yield;
759 }
760
761
762
763 /*************************************************
764 * Copy error to toplevel address *
765 *************************************************/
766
767 /* This function is used when a verify fails or defers, to ensure that the
768 failure or defer information is in the original toplevel address. This applies
769 when an address is redirected to a single new address, and the failure or
770 deferral happens to the child address.
771
772 Arguments:
773 vaddr the verify address item
774 addr the final address item
775 yield FAIL or DEFER
776
777 Returns: the value of YIELD
778 */
779
780 static int
781 copy_error(address_item *vaddr, address_item *addr, int yield)
782 {
783 if (addr != vaddr)
784 {
785 vaddr->message = addr->message;
786 vaddr->user_message = addr->user_message;
787 vaddr->basic_errno = addr->basic_errno;
788 vaddr->more_errno = addr->more_errno;
789 }
790 return yield;
791 }
792
793
794
795
796 /*************************************************
797 * Verify an email address *
798 *************************************************/
799
800 /* This function is used both for verification (-bv and at other times) and
801 address testing (-bt), which is indicated by address_test_mode being set.
802
803 Arguments:
804 vaddr contains the address to verify; the next field in this block
805 must be NULL
806 f if not NULL, write the result to this file
807 options various option bits:
808 vopt_fake_sender => this sender verify is not for the real
809 sender (it was verify=sender=xxxx or an address from a
810 header line) - rewriting must not change sender_address
811 vopt_is_recipient => this is a recipient address, otherwise
812 it's a sender address - this affects qualification and
813 rewriting and messages from callouts
814 vopt_qualify => qualify an unqualified address; else error
815 vopt_expn => called from SMTP EXPN command
816 vopt_success_on_redirect => when a new address is generated
817 the verification instantly succeeds
818
819 These ones are used by do_callout() -- the options variable
820 is passed to it.
821
822 vopt_callout_fullpm => if postmaster check, do full one
823 vopt_callout_no_cache => don't use callout cache
824 vopt_callout_random => do the "random" thing
825 vopt_callout_recipsender => use real sender for recipient
826 vopt_callout_recippmaster => use postmaster for recipient
827
828 callout if > 0, specifies that callout is required, and gives timeout
829 for individual commands
830 callout_overall if > 0, gives overall timeout for the callout function;
831 if < 0, a default is used (see do_callout())
832 callout_connect the connection timeout for callouts
833 se_mailfrom when callout is requested to verify a sender, use this
834 in MAIL FROM; NULL => ""
835 pm_mailfrom when callout is requested, if non-NULL, do the postmaster
836 thing and use this as the sender address (may be "")
837
838 routed if not NULL, set TRUE if routing succeeded, so we can
839 distinguish between routing failed and callout failed
840
841 Returns: OK address verified
842 FAIL address failed to verify
843 DEFER can't tell at present
844 */
845
846 int
847 verify_address(address_item *vaddr, FILE *f, int options, int callout,
848 int callout_overall, int callout_connect, uschar *se_mailfrom,
849 uschar *pm_mailfrom, BOOL *routed)
850 {
851 BOOL allok = TRUE;
852 BOOL full_info = (f == NULL)? FALSE : (debug_selector != 0);
853 BOOL is_recipient = (options & vopt_is_recipient) != 0;
854 BOOL expn = (options & vopt_expn) != 0;
855 BOOL success_on_redirect = (options & vopt_success_on_redirect) != 0;
856 int i;
857 int yield = OK;
858 int verify_type = expn? v_expn :
859 address_test_mode? v_none :
860 is_recipient? v_recipient : v_sender;
861 address_item *addr_list;
862 address_item *addr_new = NULL;
863 address_item *addr_remote = NULL;
864 address_item *addr_local = NULL;
865 address_item *addr_succeed = NULL;
866 uschar **failure_ptr = is_recipient?
867 &recipient_verify_failure : &sender_verify_failure;
868 uschar *ko_prefix, *cr;
869 uschar *address = vaddr->address;
870 uschar *save_sender;
871 uschar null_sender[] = { 0 }; /* Ensure writeable memory */
872
873 /* Clear, just in case */
874
875 *failure_ptr = NULL;
876
877 /* Set up a prefix and suffix for error message which allow us to use the same
878 output statements both in EXPN mode (where an SMTP response is needed) and when
879 debugging with an output file. */
880
881 if (expn)
882 {
883 ko_prefix = US"553 ";
884 cr = US"\r";
885 }
886 else ko_prefix = cr = US"";
887
888 /* Add qualify domain if permitted; otherwise an unqualified address fails. */
889
890 if (parse_find_at(address) == NULL)
891 {
892 if ((options & vopt_qualify) == 0)
893 {
894 if (f != NULL)
895 fprintf(f, "%sA domain is required for \"%s\"%s\n", ko_prefix, address,
896 cr);
897 *failure_ptr = US"qualify";
898 return FAIL;
899 }
900 address = rewrite_address_qualify(address, is_recipient);
901 }
902
903 DEBUG(D_verify)
904 {
905 debug_printf(">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>\n");
906 debug_printf("%s %s\n", address_test_mode? "Testing" : "Verifying", address);
907 }
908
909 /* Rewrite and report on it. Clear the domain and local part caches - these
910 may have been set by domains and local part tests during an ACL. */
911
912 if (global_rewrite_rules != NULL)
913 {
914 uschar *old = address;
915 address = rewrite_address(address, is_recipient, FALSE,
916 global_rewrite_rules, rewrite_existflags);
917 if (address != old)
918 {
919 for (i = 0; i < (MAX_NAMED_LIST * 2)/32; i++) vaddr->localpart_cache[i] = 0;
920 for (i = 0; i < (MAX_NAMED_LIST * 2)/32; i++) vaddr->domain_cache[i] = 0;
921 if (f != NULL && !expn) fprintf(f, "Address rewritten as: %s\n", address);
922 }
923 }
924
925 /* If this is the real sender address, we must update sender_address at
926 this point, because it may be referred to in the routers. */
927
928 if ((options & (vopt_fake_sender|vopt_is_recipient)) == 0)
929 sender_address = address;
930
931 /* If the address was rewritten to <> no verification can be done, and we have
932 to return OK. This rewriting is permitted only for sender addresses; for other
933 addresses, such rewriting fails. */
934
935 if (address[0] == 0) return OK;
936
937 /* Save a copy of the sender address for re-instating if we change it to <>
938 while verifying a sender address (a nice bit of self-reference there). */
939
940 save_sender = sender_address;
941
942 /* Update the address structure with the possibly qualified and rewritten
943 address. Set it up as the starting address on the chain of new addresses. */
944
945 vaddr->address = address;
946 addr_new = vaddr;
947
948 /* We need a loop, because an address can generate new addresses. We must also
949 cope with generated pipes and files at the top level. (See also the code and
950 comment in deliver.c.) However, it is usually the case that the router for
951 user's .forward files has its verify flag turned off.
952
953 If an address generates more than one child, the loop is used only when
954 full_info is set, and this can only be set locally. Remote enquiries just get
955 information about the top level address, not anything that it generated. */
956
957 while (addr_new != NULL)
958 {
959 int rc;
960 address_item *addr = addr_new;
961
962 addr_new = addr->next;
963 addr->next = NULL;
964
965 DEBUG(D_verify)
966 {
967 debug_printf(">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>\n");
968 debug_printf("Considering %s\n", addr->address);
969 }
970
971 /* Handle generated pipe, file or reply addresses. We don't get these
972 when handling EXPN, as it does only one level of expansion. */
973
974 if (testflag(addr, af_pfr))
975 {
976 allok = FALSE;
977 if (f != NULL)
978 {
979 BOOL allow;
980
981 if (addr->address[0] == '>')
982 {
983 allow = testflag(addr, af_allow_reply);
984 fprintf(f, "%s -> mail %s", addr->parent->address, addr->address + 1);
985 }
986 else
987 {
988 allow = (addr->address[0] == '|')?
989 testflag(addr, af_allow_pipe) : testflag(addr, af_allow_file);
990 fprintf(f, "%s -> %s", addr->parent->address, addr->address);
991 }
992
993 if (addr->basic_errno == ERRNO_BADTRANSPORT)
994 fprintf(f, "\n*** Error in setting up pipe, file, or autoreply:\n"
995 "%s\n", addr->message);
996 else if (allow)
997 fprintf(f, "\n transport = %s\n", addr->transport->name);
998 else
999 fprintf(f, " *** forbidden ***\n");
1000 }
1001 continue;
1002 }
1003
1004 /* Just in case some router parameter refers to it. */
1005
1006 return_path = (addr->p.errors_address != NULL)?
1007 addr->p.errors_address : sender_address;
1008
1009 /* Split the address into domain and local part, handling the %-hack if
1010 necessary, and then route it. While routing a sender address, set
1011 $sender_address to <> because that is what it will be if we were trying to
1012 send a bounce to the sender. */
1013
1014 if (routed != NULL) *routed = FALSE;
1015 if ((rc = deliver_split_address(addr)) == OK)
1016 {
1017 if (!is_recipient) sender_address = null_sender;
1018 rc = route_address(addr, &addr_local, &addr_remote, &addr_new,
1019 &addr_succeed, verify_type);
1020 sender_address = save_sender; /* Put back the real sender */
1021 }
1022
1023 /* If routing an address succeeded, set the flag that remembers, for use when
1024 an ACL cached a sender verify (in case a callout fails). Then if routing set
1025 up a list of hosts or the transport has a host list, and the callout option
1026 is set, and we aren't in a host checking run, do the callout verification,
1027 and set another flag that notes that a callout happened. */
1028
1029 if (rc == OK)
1030 {
1031 if (routed != NULL) *routed = TRUE;
1032 if (callout > 0)
1033 {
1034 host_item *host_list = addr->host_list;
1035
1036 /* Default, if no remote transport, to NULL for the interface (=> any),
1037 "smtp" for the port, and "smtp" for the protocol. */
1038
1039 transport_feedback tf = { NULL, US"smtp", US"smtp", NULL, FALSE, FALSE };
1040
1041 /* If verification yielded a remote transport, we want to use that
1042 transport's options, so as to mimic what would happen if we were really
1043 sending a message to this address. */
1044
1045 if (addr->transport != NULL && !addr->transport->info->local)
1046 {
1047 (void)(addr->transport->setup)(addr->transport, addr, &tf, NULL);
1048
1049 /* If the transport has hosts and the router does not, or if the
1050 transport is configured to override the router's hosts, we must build a
1051 host list of the transport's hosts, and find the IP addresses */
1052
1053 if (tf.hosts != NULL && (host_list == NULL || tf.hosts_override))
1054 {
1055 uschar *s;
1056
1057 host_list = NULL; /* Ignore the router's hosts */
1058
1059 deliver_domain = addr->domain;
1060 deliver_localpart = addr->local_part;
1061 s = expand_string(tf.hosts);
1062 deliver_domain = deliver_localpart = NULL;
1063
1064 if (s == NULL)
1065 {
1066 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand list of hosts "
1067 "\"%s\" in %s transport for callout: %s", tf.hosts,
1068 addr->transport->name, expand_string_message);
1069 }
1070 else
1071 {
1072 uschar *canonical_name;
1073 host_item *host, *nexthost;
1074 host_build_hostlist(&host_list, s, tf.hosts_randomize);
1075
1076 /* Just ignore failures to find a host address. If we don't manage
1077 to find any addresses, the callout will defer. Note that more than
1078 one address may be found for a single host, which will result in
1079 additional host items being inserted into the chain. Hence we must
1080 save the next host first. */
1081
1082 for (host = host_list; host != NULL; host = nexthost)
1083 {
1084 nexthost = host->next;
1085 if (tf.gethostbyname ||
1086 string_is_ip_address(host->name, NULL) > 0)
1087 (void)host_find_byname(host, NULL, &canonical_name, TRUE);
1088 else
1089 {
1090 int flags = HOST_FIND_BY_A;
1091 if (tf.qualify_single) flags |= HOST_FIND_QUALIFY_SINGLE;
1092 if (tf.search_parents) flags |= HOST_FIND_SEARCH_PARENTS;
1093 (void)host_find_bydns(host, NULL, flags, NULL, NULL, NULL,
1094 &canonical_name, NULL);
1095 }
1096 }
1097 }
1098 }
1099 }
1100
1101 /* Can only do a callout if we have at least one host! If the callout
1102 fails, it will have set ${sender,recipient}_verify_failure. */
1103
1104 if (host_list != NULL)
1105 {
1106 HDEBUG(D_verify) debug_printf("Attempting full verification using callout\n");
1107 if (host_checking && !host_checking_callout)
1108 {
1109 HDEBUG(D_verify)
1110 debug_printf("... callout omitted by default when host testing\n"
1111 "(Use -bhc if you want the callouts to happen.)\n");
1112 }
1113 else
1114 {
1115 rc = do_callout(addr, host_list, &tf, callout, callout_overall,
1116 callout_connect, options, se_mailfrom, pm_mailfrom);
1117 }
1118 }
1119 else
1120 {
1121 HDEBUG(D_verify) debug_printf("Cannot do callout: neither router nor "
1122 "transport provided a host list\n");
1123 }
1124 }
1125 }
1126
1127 /* Otherwise, any failure is a routing failure */
1128
1129 else *failure_ptr = US"route";
1130
1131 /* A router may return REROUTED if it has set up a child address as a result
1132 of a change of domain name (typically from widening). In this case we always
1133 want to continue to verify the new child. */
1134
1135 if (rc == REROUTED) continue;
1136
1137 /* Handle hard failures */
1138
1139 if (rc == FAIL)
1140 {
1141 allok = FALSE;
1142 if (f != NULL)
1143 {
1144 fprintf(f, "%s%s %s", ko_prefix, address,
1145 address_test_mode? "is undeliverable" : "failed to verify");
1146 if (!expn && admin_user)
1147 {
1148 if (addr->basic_errno > 0)
1149 fprintf(f, ": %s", strerror(addr->basic_errno));
1150 if (addr->message != NULL)
1151 fprintf(f, ":\n %s", addr->message);
1152 }
1153 fprintf(f, "%s\n", cr);
1154 }
1155
1156 if (!full_info) return copy_error(vaddr, addr, FAIL);
1157 else yield = FAIL;
1158 }
1159
1160 /* Soft failure */
1161
1162 else if (rc == DEFER)
1163 {
1164 allok = FALSE;
1165 if (f != NULL)
1166 {
1167 fprintf(f, "%s%s cannot be resolved at this time", ko_prefix, address);
1168 if (!expn && admin_user)
1169 {
1170 if (addr->basic_errno > 0)
1171 fprintf(f, ":\n %s", strerror(addr->basic_errno));
1172 if (addr->message != NULL)
1173 fprintf(f, ":\n %s", addr->message);
1174 else if (addr->basic_errno <= 0)
1175 fprintf(f, ":\n unknown error");
1176 }
1177
1178 fprintf(f, "%s\n", cr);
1179 }
1180 if (!full_info) return copy_error(vaddr, addr, DEFER);
1181 else if (yield == OK) yield = DEFER;
1182 }
1183
1184 /* If we are handling EXPN, we do not want to continue to route beyond
1185 the top level. */
1186
1187 else if (expn)
1188 {
1189 uschar *ok_prefix = US"250-";
1190 if (addr_new == NULL)
1191 {
1192 if (addr_local == NULL && addr_remote == NULL)
1193 fprintf(f, "250 mail to <%s> is discarded\r\n", address);
1194 else
1195 fprintf(f, "250 <%s>\r\n", address);
1196 }
1197 else while (addr_new != NULL)
1198 {
1199 address_item *addr2 = addr_new;
1200 addr_new = addr2->next;
1201 if (addr_new == NULL) ok_prefix = US"250 ";
1202 fprintf(f, "%s<%s>\r\n", ok_prefix, addr2->address);
1203 }
1204 return OK;
1205 }
1206
1207 /* Successful routing other than EXPN. */
1208
1209 else
1210 {
1211 /* Handle successful routing when short info wanted. Otherwise continue for
1212 other (generated) addresses. Short info is the operational case. Full info
1213 can be requested only when debug_selector != 0 and a file is supplied.
1214
1215 There is a conflict between the use of aliasing as an alternate email
1216 address, and as a sort of mailing list. If an alias turns the incoming
1217 address into just one address (e.g. J.Caesar->jc44) you may well want to
1218 carry on verifying the generated address to ensure it is valid when
1219 checking incoming mail. If aliasing generates multiple addresses, you
1220 probably don't want to do this. Exim therefore treats the generation of
1221 just a single new address as a special case, and continues on to verify the
1222 generated address. */
1223
1224 if (!full_info && /* Stop if short info wanted AND */
1225 (((addr_new == NULL || /* No new address OR */
1226 addr_new->next != NULL || /* More than one new address OR */
1227 testflag(addr_new, af_pfr))) /* New address is pfr */
1228 || /* OR */
1229 (addr_new != NULL && /* At least one new address AND */
1230 success_on_redirect))) /* success_on_redirect is set */
1231 {
1232 if (f != NULL) fprintf(f, "%s %s\n", address,
1233 address_test_mode? "is deliverable" : "verified");
1234
1235 /* If we have carried on to verify a child address, we want the value
1236 of $address_data to be that of the child */
1237
1238 vaddr->p.address_data = addr->p.address_data;
1239 return OK;
1240 }
1241 }
1242 } /* Loop for generated addresses */
1243
1244 /* Display the full results of the successful routing, including any generated
1245 addresses. Control gets here only when full_info is set, which requires f not
1246 to be NULL, and this occurs only when a top-level verify is called with the
1247 debugging switch on.
1248
1249 If there are no local and no remote addresses, and there were no pipes, files,
1250 or autoreplies, and there were no errors or deferments, the message is to be
1251 discarded, usually because of the use of :blackhole: in an alias file. */
1252
1253 if (allok && addr_local == NULL && addr_remote == NULL)
1254 fprintf(f, "mail to %s is discarded\n", address);
1255
1256 else for (addr_list = addr_local, i = 0; i < 2; addr_list = addr_remote, i++)
1257 {
1258 while (addr_list != NULL)
1259 {
1260 address_item *addr = addr_list;
1261 address_item *p = addr->parent;
1262 addr_list = addr->next;
1263
1264 fprintf(f, "%s", CS addr->address);
1265 #ifdef EXPERIMENTAL_SRS
1266 if(addr->p.srs_sender)
1267 fprintf(f, " [srs = %s]", addr->p.srs_sender);
1268 #endif
1269 while (p != NULL)
1270 {
1271 fprintf(f, "\n <-- %s", p->address);
1272 p = p->parent;
1273 }
1274 fprintf(f, "\n ");
1275
1276 /* Show router, and transport */
1277
1278 fprintf(f, "router = %s, ", addr->router->name);
1279 fprintf(f, "transport = %s\n", (addr->transport == NULL)? US"unset" :
1280 addr->transport->name);
1281
1282 /* Show any hosts that are set up by a router unless the transport
1283 is going to override them; fiddle a bit to get a nice format. */
1284
1285 if (addr->host_list != NULL && addr->transport != NULL &&
1286 !addr->transport->overrides_hosts)
1287 {
1288 host_item *h;
1289 int maxlen = 0;
1290 int maxaddlen = 0;
1291 for (h = addr->host_list; h != NULL; h = h->next)
1292 {
1293 int len = Ustrlen(h->name);
1294 if (len > maxlen) maxlen = len;
1295 len = (h->address != NULL)? Ustrlen(h->address) : 7;
1296 if (len > maxaddlen) maxaddlen = len;
1297 }
1298 for (h = addr->host_list; h != NULL; h = h->next)
1299 {
1300 int len = Ustrlen(h->name);
1301 fprintf(f, " host %s ", h->name);
1302 while (len++ < maxlen) fprintf(f, " ");
1303 if (h->address != NULL)
1304 {
1305 fprintf(f, "[%s] ", h->address);
1306 len = Ustrlen(h->address);
1307 }
1308 else if (!addr->transport->info->local) /* Omit [unknown] for local */
1309 {
1310 fprintf(f, "[unknown] ");
1311 len = 7;
1312 }
1313 else len = -3;
1314 while (len++ < maxaddlen) fprintf(f," ");
1315 if (h->mx >= 0) fprintf(f, "MX=%d", h->mx);
1316 if (h->port != PORT_NONE) fprintf(f, " port=%d", h->port);
1317 if (h->status == hstatus_unusable) fprintf(f, " ** unusable **");
1318 fprintf(f, "\n");
1319 }
1320 }
1321 }
1322 }
1323
1324 /* Will be DEFER or FAIL if any one address has, only for full_info (which is
1325 the -bv or -bt case). */
1326
1327 return yield;
1328 }
1329
1330
1331
1332
1333 /*************************************************
1334 * Check headers for syntax errors *
1335 *************************************************/
1336
1337 /* This function checks those header lines that contain addresses, and verifies
1338 that all the addresses therein are syntactially correct.
1339
1340 Arguments:
1341 msgptr where to put an error message
1342
1343 Returns: OK
1344 FAIL
1345 */
1346
1347 int
1348 verify_check_headers(uschar **msgptr)
1349 {
1350 header_line *h;
1351 uschar *colon, *s;
1352
1353 for (h = header_list; h != NULL; h = h->next)
1354 {
1355 if (h->type != htype_from &&
1356 h->type != htype_reply_to &&
1357 h->type != htype_sender &&
1358 h->type != htype_to &&
1359 h->type != htype_cc &&
1360 h->type != htype_bcc)
1361 continue;
1362
1363 colon = Ustrchr(h->text, ':');
1364 s = colon + 1;
1365 while (isspace(*s)) s++;
1366
1367 parse_allow_group = TRUE; /* Allow group syntax */
1368
1369 /* Loop for multiple addresses in the header */
1370
1371 while (*s != 0)
1372 {
1373 uschar *ss = parse_find_address_end(s, FALSE);
1374 uschar *recipient, *errmess;
1375 int terminator = *ss;
1376 int start, end, domain;
1377
1378 /* Temporarily terminate the string at this point, and extract the
1379 operative address within. */
1380
1381 *ss = 0;
1382 recipient = parse_extract_address(s,&errmess,&start,&end,&domain,FALSE);
1383 *ss = terminator;
1384
1385 /* Permit an unqualified address only if the message is local, or if the
1386 sending host is configured to be permitted to send them. */
1387
1388 if (recipient != NULL && domain == 0)
1389 {
1390 if (h->type == htype_from || h->type == htype_sender)
1391 {
1392 if (!allow_unqualified_sender) recipient = NULL;
1393 }
1394 else
1395 {
1396 if (!allow_unqualified_recipient) recipient = NULL;
1397 }
1398 if (recipient == NULL) errmess = US"unqualified address not permitted";
1399 }
1400
1401 /* It's an error if no address could be extracted, except for the special
1402 case of an empty address. */
1403
1404 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
1405 {
1406 uschar *verb = US"is";
1407 uschar *t = ss;
1408 int len;
1409
1410 /* Arrange not to include any white space at the end in the
1411 error message. */
1412
1413 while (t > s && isspace(t[-1])) t--;
1414
1415 /* Add the address which failed to the error message, since in a
1416 header with very many addresses it is sometimes hard to spot
1417 which one is at fault. However, limit the amount of address to
1418 quote - cases have been seen where, for example, a missing double
1419 quote in a humungous To: header creates an "address" that is longer
1420 than string_sprintf can handle. */
1421
1422 len = t - s;
1423 if (len > 1024)
1424 {
1425 len = 1024;
1426 verb = US"begins";
1427 }
1428
1429 *msgptr = string_printing(
1430 string_sprintf("%s: failing address in \"%.*s\" header %s: %.*s",
1431 errmess, colon - h->text, h->text, verb, len, s));
1432
1433 return FAIL;
1434 }
1435
1436 /* Advance to the next address */
1437
1438 s = ss + (terminator? 1:0);
1439 while (isspace(*s)) s++;
1440 } /* Next address */
1441 } /* Next header */
1442
1443 return OK;
1444 }
1445
1446
1447
1448
1449 /*************************************************
1450 * Find if verified sender *
1451 *************************************************/
1452
1453 /* Usually, just a single address is verified as the sender of the message.
1454 However, Exim can be made to verify other addresses as well (often related in
1455 some way), and this is useful in some environments. There may therefore be a
1456 chain of such addresses that have previously been tested. This function finds
1457 whether a given address is on the chain.
1458
1459 Arguments: the address to be verified
1460 Returns: pointer to an address item, or NULL
1461 */
1462
1463 address_item *
1464 verify_checked_sender(uschar *sender)
1465 {
1466 address_item *addr;
1467 for (addr = sender_verified_list; addr != NULL; addr = addr->next)
1468 if (Ustrcmp(sender, addr->address) == 0) break;
1469 return addr;
1470 }
1471
1472
1473
1474
1475
1476 /*************************************************
1477 * Get valid header address *
1478 *************************************************/
1479
1480 /* Scan the originator headers of the message, looking for an address that
1481 verifies successfully. RFC 822 says:
1482
1483 o The "Sender" field mailbox should be sent notices of
1484 any problems in transport or delivery of the original
1485 messages. If there is no "Sender" field, then the
1486 "From" field mailbox should be used.
1487
1488 o If the "Reply-To" field exists, then the reply should
1489 go to the addresses indicated in that field and not to
1490 the address(es) indicated in the "From" field.
1491
1492 So we check a Sender field if there is one, else a Reply_to field, else a From
1493 field. As some strange messages may have more than one of these fields,
1494 especially if they are resent- fields, check all of them if there is more than
1495 one.
1496
1497 Arguments:
1498 user_msgptr points to where to put a user error message
1499 log_msgptr points to where to put a log error message
1500 callout timeout for callout check (passed to verify_address())
1501 callout_overall overall callout timeout (ditto)
1502 callout_connect connect callout timeout (ditto)
1503 se_mailfrom mailfrom for verify; NULL => ""
1504 pm_mailfrom sender for pm callout check (passed to verify_address())
1505 options callout options (passed to verify_address())
1506 verrno where to put the address basic_errno
1507
1508 If log_msgptr is set to something without setting user_msgptr, the caller
1509 normally uses log_msgptr for both things.
1510
1511 Returns: result of the verification attempt: OK, FAIL, or DEFER;
1512 FAIL is given if no appropriate headers are found
1513 */
1514
1515 int
1516 verify_check_header_address(uschar **user_msgptr, uschar **log_msgptr,
1517 int callout, int callout_overall, int callout_connect, uschar *se_mailfrom,
1518 uschar *pm_mailfrom, int options, int *verrno)
1519 {
1520 static int header_types[] = { htype_sender, htype_reply_to, htype_from };
1521 int yield = FAIL;
1522 int i;
1523
1524 for (i = 0; i < 3; i++)
1525 {
1526 header_line *h;
1527 for (h = header_list; h != NULL; h = h->next)
1528 {
1529 int terminator, new_ok;
1530 uschar *s, *ss, *endname;
1531
1532 if (h->type != header_types[i]) continue;
1533 s = endname = Ustrchr(h->text, ':') + 1;
1534
1535 while (*s != 0)
1536 {
1537 address_item *vaddr;
1538
1539 while (isspace(*s) || *s == ',') s++;
1540 if (*s == 0) break; /* End of header */
1541
1542 ss = parse_find_address_end(s, FALSE);
1543
1544 /* The terminator is a comma or end of header, but there may be white
1545 space preceding it (including newline for the last address). Move back
1546 past any white space so we can check against any cached envelope sender
1547 address verifications. */
1548
1549 while (isspace(ss[-1])) ss--;
1550 terminator = *ss;
1551 *ss = 0;
1552
1553 HDEBUG(D_verify) debug_printf("verifying %.*s header address %s\n",
1554 (int)(endname - h->text), h->text, s);
1555
1556 /* See if we have already verified this address as an envelope sender,
1557 and if so, use the previous answer. */
1558
1559 vaddr = verify_checked_sender(s);
1560
1561 if (vaddr != NULL && /* Previously checked */
1562 (callout <= 0 || /* No callout needed; OR */
1563 vaddr->special_action > 256)) /* Callout was done */
1564 {
1565 new_ok = vaddr->special_action & 255;
1566 HDEBUG(D_verify) debug_printf("previously checked as envelope sender\n");
1567 *ss = terminator; /* Restore shortened string */
1568 }
1569
1570 /* Otherwise we run the verification now. We must restore the shortened
1571 string before running the verification, so the headers are correct, in
1572 case there is any rewriting. */
1573
1574 else
1575 {
1576 int start, end, domain;
1577 uschar *address = parse_extract_address(s, log_msgptr, &start,
1578 &end, &domain, FALSE);
1579
1580 *ss = terminator;
1581
1582 /* If verification failed because of a syntax error, fail this
1583 function, and ensure that the failing address gets added to the error
1584 message. */
1585
1586 if (address == NULL)
1587 {
1588 new_ok = FAIL;
1589 if (*log_msgptr != NULL)
1590 {
1591 while (ss > s && isspace(ss[-1])) ss--;
1592 *log_msgptr = string_sprintf("syntax error in '%.*s' header when "
1593 "scanning for sender: %s in \"%.*s\"",
1594 endname - h->text, h->text, *log_msgptr, ss - s, s);
1595 return FAIL;
1596 }
1597 }
1598
1599 /* Else go ahead with the sender verification. But it isn't *the*
1600 sender of the message, so set vopt_fake_sender to stop sender_address
1601 being replaced after rewriting or qualification. */
1602
1603 else
1604 {
1605 vaddr = deliver_make_addr(address, FALSE);
1606 new_ok = verify_address(vaddr, NULL, options | vopt_fake_sender,
1607 callout, callout_overall, callout_connect, se_mailfrom,
1608 pm_mailfrom, NULL);
1609 }
1610 }
1611
1612 /* We now have the result, either newly found, or cached. If we are
1613 giving out error details, set a specific user error. This means that the
1614 last of these will be returned to the user if all three fail. We do not
1615 set a log message - the generic one below will be used. */
1616
1617 if (new_ok != OK)
1618 {
1619 *verrno = vaddr->basic_errno;
1620 if (smtp_return_error_details)
1621 {
1622 *user_msgptr = string_sprintf("Rejected after DATA: "
1623 "could not verify \"%.*s\" header address\n%s: %s",
1624 endname - h->text, h->text, vaddr->address, vaddr->message);
1625 }
1626 }
1627
1628 /* Success or defer */
1629
1630 if (new_ok == OK) return OK;
1631 if (new_ok == DEFER) yield = DEFER;
1632
1633 /* Move on to any more addresses in the header */
1634
1635 s = ss;
1636 }
1637 }
1638 }
1639
1640 if (yield == FAIL && *log_msgptr == NULL)
1641 *log_msgptr = US"there is no valid sender in any header line";
1642
1643 if (yield == DEFER && *log_msgptr == NULL)
1644 *log_msgptr = US"all attempts to verify a sender in a header line deferred";
1645
1646 return yield;
1647 }
1648
1649
1650
1651
1652 /*************************************************
1653 * Get RFC 1413 identification *
1654 *************************************************/
1655
1656 /* Attempt to get an id from the sending machine via the RFC 1413 protocol. If
1657 the timeout is set to zero, then the query is not done. There may also be lists
1658 of hosts and nets which are exempt. To guard against malefactors sending
1659 non-printing characters which could, for example, disrupt a message's headers,
1660 make sure the string consists of printing characters only.
1661
1662 Argument:
1663 port the port to connect to; usually this is IDENT_PORT (113), but when
1664 running in the test harness with -bh a different value is used.
1665
1666 Returns: nothing
1667
1668 Side effect: any received ident value is put in sender_ident (NULL otherwise)
1669 */
1670
1671 void
1672 verify_get_ident(int port)
1673 {
1674 int sock, host_af, qlen;
1675 int received_sender_port, received_interface_port, n;
1676 uschar *p;
1677 uschar buffer[2048];
1678
1679 /* Default is no ident. Check whether we want to do an ident check for this
1680 host. */
1681
1682 sender_ident = NULL;
1683 if (rfc1413_query_timeout <= 0 || verify_check_host(&rfc1413_hosts) != OK)
1684 return;
1685
1686 DEBUG(D_ident) debug_printf("doing ident callback\n");
1687
1688 /* Set up a connection to the ident port of the remote host. Bind the local end
1689 to the incoming interface address. If the sender host address is an IPv6
1690 address, the incoming interface address will also be IPv6. */
1691
1692 host_af = (Ustrchr(sender_host_address, ':') == NULL)? AF_INET : AF_INET6;
1693 sock = ip_socket(SOCK_STREAM, host_af);
1694 if (sock < 0) return;
1695
1696 if (ip_bind(sock, host_af, interface_address, 0) < 0)
1697 {
1698 DEBUG(D_ident) debug_printf("bind socket for ident failed: %s\n",
1699 strerror(errno));
1700 goto END_OFF;
1701 }
1702
1703 if (ip_connect(sock, host_af, sender_host_address, port, rfc1413_query_timeout)
1704 < 0)
1705 {
1706 if (errno == ETIMEDOUT && (log_extra_selector & LX_ident_timeout) != 0)
1707 {
1708 log_write(0, LOG_MAIN, "ident connection to %s timed out",
1709 sender_host_address);
1710 }
1711 else
1712 {
1713 DEBUG(D_ident) debug_printf("ident connection to %s failed: %s\n",
1714 sender_host_address, strerror(errno));
1715 }
1716 goto END_OFF;
1717 }
1718
1719 /* Construct and send the query. */
1720
1721 sprintf(CS buffer, "%d , %d\r\n", sender_host_port, interface_port);
1722 qlen = Ustrlen(buffer);
1723 if (send(sock, buffer, qlen, 0) < 0)
1724 {
1725 DEBUG(D_ident) debug_printf("ident send failed: %s\n", strerror(errno));
1726 goto END_OFF;
1727 }
1728
1729 /* Read a response line. We put it into the rest of the buffer, using several
1730 recv() calls if necessary. */
1731
1732 p = buffer + qlen;
1733
1734 for (;;)
1735 {
1736 uschar *pp;
1737 int count;
1738 int size = sizeof(buffer) - (p - buffer);
1739
1740 if (size <= 0) goto END_OFF; /* Buffer filled without seeing \n. */
1741 count = ip_recv(sock, p, size, rfc1413_query_timeout);
1742 if (count <= 0) goto END_OFF; /* Read error or EOF */
1743
1744 /* Scan what we just read, to see if we have reached the terminating \r\n. Be
1745 generous, and accept a plain \n terminator as well. The only illegal
1746 character is 0. */
1747
1748 for (pp = p; pp < p + count; pp++)
1749 {
1750 if (*pp == 0) goto END_OFF; /* Zero octet not allowed */
1751 if (*pp == '\n')
1752 {
1753 if (pp[-1] == '\r') pp--;
1754 *pp = 0;
1755 goto GOT_DATA; /* Break out of both loops */
1756 }
1757 }
1758
1759 /* Reached the end of the data without finding \n. Let the loop continue to
1760 read some more, if there is room. */
1761
1762 p = pp;
1763 }
1764
1765 GOT_DATA:
1766
1767 /* We have received a line of data. Check it carefully. It must start with the
1768 same two port numbers that we sent, followed by data as defined by the RFC. For
1769 example,
1770
1771 12345 , 25 : USERID : UNIX :root
1772
1773 However, the amount of white space may be different to what we sent. In the
1774 "osname" field there may be several sub-fields, comma separated. The data we
1775 actually want to save follows the third colon. Some systems put leading spaces
1776 in it - we discard those. */
1777
1778 if (sscanf(CS buffer + qlen, "%d , %d%n", &received_sender_port,
1779 &received_interface_port, &n) != 2 ||
1780 received_sender_port != sender_host_port ||
1781 received_interface_port != interface_port)
1782 goto END_OFF;
1783
1784 p = buffer + qlen + n;
1785 while(isspace(*p)) p++;
1786 if (*p++ != ':') goto END_OFF;
1787 while(isspace(*p)) p++;
1788 if (Ustrncmp(p, "USERID", 6) != 0) goto END_OFF;
1789 p += 6;
1790 while(isspace(*p)) p++;
1791 if (*p++ != ':') goto END_OFF;
1792 while (*p != 0 && *p != ':') p++;
1793 if (*p++ == 0) goto END_OFF;
1794 while(isspace(*p)) p++;
1795 if (*p == 0) goto END_OFF;
1796
1797 /* The rest of the line is the data we want. We turn it into printing
1798 characters when we save it, so that it cannot mess up the format of any logging
1799 or Received: lines into which it gets inserted. We keep a maximum of 127
1800 characters. */
1801
1802 sender_ident = string_printing(string_copyn(p, 127));
1803 DEBUG(D_ident) debug_printf("sender_ident = %s\n", sender_ident);
1804
1805 END_OFF:
1806 (void)close(sock);
1807 return;
1808 }
1809
1810
1811
1812
1813 /*************************************************
1814 * Match host to a single host-list item *
1815 *************************************************/
1816
1817 /* This function compares a host (name or address) against a single item
1818 from a host list. The host name gets looked up if it is needed and is not
1819 already known. The function is called from verify_check_this_host() via
1820 match_check_list(), which is why most of its arguments are in a single block.
1821
1822 Arguments:
1823 arg the argument block (see below)
1824 ss the host-list item
1825 valueptr where to pass back looked up data, or NULL
1826 error for error message when returning ERROR
1827
1828 The block contains:
1829 host_name (a) the host name, or
1830 (b) NULL, implying use sender_host_name and
1831 sender_host_aliases, looking them up if required, or
1832 (c) the empty string, meaning that only IP address matches
1833 are permitted
1834 host_address the host address
1835 host_ipv4 the IPv4 address taken from an IPv6 one
1836
1837 Returns: OK matched
1838 FAIL did not match
1839 DEFER lookup deferred
1840 ERROR (a) failed to find the host name or IP address, or
1841 (b) unknown lookup type specified, or
1842 (c) host name encountered when only IP addresses are
1843 being matched
1844 */
1845
1846 int
1847 check_host(void *arg, uschar *ss, uschar **valueptr, uschar **error)
1848 {
1849 check_host_block *cb = (check_host_block *)arg;
1850 int mlen = -1;
1851 int maskoffset;
1852 BOOL iplookup = FALSE;
1853 BOOL isquery = FALSE;
1854 BOOL isiponly = cb->host_name != NULL && cb->host_name[0] == 0;
1855 uschar *t = ss;
1856 uschar *semicolon;
1857 uschar **aliases;
1858
1859 /* Optimize for the special case when the pattern is "*". */
1860
1861 if (*ss == '*' && ss[1] == 0) return OK;
1862
1863 /* If the pattern is empty, it matches only in the case when there is no host -
1864 this can occur in ACL checking for SMTP input using the -bs option. In this
1865 situation, the host address is the empty string. */
1866
1867 if (cb->host_address[0] == 0) return (*ss == 0)? OK : FAIL;
1868 if (*ss == 0) return FAIL;
1869
1870 /* If the pattern is precisely "@" then match against the primary host name,
1871 provided that host name matching is permitted; if it's "@[]" match against the
1872 local host's IP addresses. */
1873
1874 if (*ss == '@')
1875 {
1876 if (ss[1] == 0)
1877 {
1878 if (isiponly) return ERROR;
1879 ss = primary_hostname;
1880 }
1881 else if (Ustrcmp(ss, "@[]") == 0)
1882 {
1883 ip_address_item *ip;
1884 for (ip = host_find_interfaces(); ip != NULL; ip = ip->next)
1885 if (Ustrcmp(ip->address, cb->host_address) == 0) return OK;
1886 return FAIL;
1887 }
1888 }
1889
1890 /* If the pattern is an IP address, optionally followed by a bitmask count, do
1891 a (possibly masked) comparision with the current IP address. */
1892
1893 if (string_is_ip_address(ss, &maskoffset) > 0)
1894 return (host_is_in_net(cb->host_address, ss, maskoffset)? OK : FAIL);
1895
1896 /* See if there is a semicolon in the pattern */
1897
1898 semicolon = Ustrchr(ss, ';');
1899
1900 /* If we are doing an IP address only match, then all lookups must be IP
1901 address lookups. */
1902
1903 if (isiponly)
1904 {
1905 iplookup = semicolon != NULL;
1906 }
1907
1908 /* Otherwise, if the item is of the form net[n]-lookup;<file|query> then it is
1909 a lookup on a masked IP network, in textual form. The net- stuff really only
1910 applies to single-key lookups where the key is implicit. For query-style
1911 lookups the key is specified in the query. From release 4.30, the use of net-
1912 for query style is no longer needed, but we retain it for backward
1913 compatibility. */
1914
1915 else if (Ustrncmp(ss, "net", 3) == 0 && semicolon != NULL)
1916 {
1917 mlen = 0;
1918 for (t = ss + 3; isdigit(*t); t++) mlen = mlen * 10 + *t - '0';
1919 if (mlen == 0 && t == ss+3) mlen = -1; /* No mask supplied */
1920 iplookup = (*t++ == '-');
1921 }
1922
1923 /* Do the IP address lookup if that is indeed what we have */
1924
1925 if (iplookup)
1926 {
1927 int insize;
1928 int search_type;
1929 int incoming[4];
1930 void *handle;
1931 uschar *filename, *key, *result;
1932 uschar buffer[64];
1933
1934 /* Find the search type */
1935
1936 search_type = search_findtype(t, semicolon - t);
1937
1938 if (search_type < 0) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
1939 search_error_message);
1940
1941 /* Adjust parameters for the type of lookup. For a query-style lookup, there
1942 is no file name, and the "key" is just the query. For query-style with a file
1943 name, we have to fish the file off the start of the query. For a single-key
1944 lookup, the key is the current IP address, masked appropriately, and
1945 reconverted to text form, with the mask appended. For IPv6 addresses, specify
1946 dot separators instead of colons. */
1947
1948 if (mac_islookup(search_type, lookup_absfilequery))
1949 {
1950 filename = semicolon + 1;
1951 key = filename;
1952 while (*key != 0 && !isspace(*key)) key++;
1953 filename = string_copyn(filename, key - filename);
1954 while (isspace(*key)) key++;
1955 }
1956 else if (mac_islookup(search_type, lookup_querystyle))
1957 {
1958 filename = NULL;
1959 key = semicolon + 1;
1960 }
1961 else
1962 {
1963 insize = host_aton(cb->host_address, incoming);
1964 host_mask(insize, incoming, mlen);
1965 (void)host_nmtoa(insize, incoming, mlen, buffer, '.');
1966 key = buffer;
1967 filename = semicolon + 1;
1968 }
1969
1970 /* Now do the actual lookup; note that there is no search_close() because
1971 of the caching arrangements. */
1972
1973 handle = search_open(filename, search_type, 0, NULL, NULL);
1974 if (handle == NULL) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
1975 search_error_message);
1976 result = search_find(handle, filename, key, -1, NULL, 0, 0, NULL);
1977 if (valueptr != NULL) *valueptr = result;
1978 return (result != NULL)? OK : search_find_defer? DEFER: FAIL;
1979 }
1980
1981 /* The pattern is not an IP address or network reference of any kind. That is,
1982 it is a host name pattern. If this is an IP only match, there's an error in the
1983 host list. */
1984
1985 if (isiponly)
1986 {
1987 *error = US"cannot match host name in match_ip list";
1988 return ERROR;
1989 }
1990
1991 /* Check the characters of the pattern to see if they comprise only letters,
1992 digits, full stops, and hyphens (the constituents of domain names). Allow
1993 underscores, as they are all too commonly found. Sigh. Also, if
1994 allow_utf8_domains is set, allow top-bit characters. */
1995
1996 for (t = ss; *t != 0; t++)
1997 if (!isalnum(*t) && *t != '.' && *t != '-' && *t != '_' &&
1998 (!allow_utf8_domains || *t < 128)) break;
1999
2000 /* If the pattern is a complete domain name, with no fancy characters, look up
2001 its IP address and match against that. Note that a multi-homed host will add
2002 items to the chain. */
2003
2004 if (*t == 0)
2005 {
2006 int rc;
2007 host_item h;
2008 h.next = NULL;
2009 h.name = ss;
2010 h.address = NULL;
2011 h.mx = MX_NONE;
2012 rc = host_find_byname(&h, NULL, NULL, FALSE);
2013 if (rc == HOST_FOUND || rc == HOST_FOUND_LOCAL)
2014 {
2015 host_item *hh;
2016 for (hh = &h; hh != NULL; hh = hh->next)
2017 {
2018 if (Ustrcmp(hh->address, (Ustrchr(hh->address, ':') == NULL)?
2019 cb->host_ipv4 : cb->host_address) == 0)
2020 return OK;
2021 }
2022 return FAIL;
2023 }
2024 if (rc == HOST_FIND_AGAIN) return DEFER;
2025 *error = string_sprintf("failed to find IP address for %s", ss);
2026 return ERROR;
2027 }
2028
2029 /* Almost all subsequent comparisons require the host name, and can be done
2030 using the general string matching function. When this function is called for
2031 outgoing hosts, the name is always given explicitly. If it is NULL, it means we
2032 must use sender_host_name and its aliases, looking them up if necessary. */
2033
2034 if (cb->host_name != NULL) /* Explicit host name given */
2035 return match_check_string(cb->host_name, ss, -1, TRUE, TRUE, TRUE,
2036 valueptr);
2037
2038 /* Host name not given; in principle we need the sender host name and its
2039 aliases. However, for query-style lookups, we do not need the name if the
2040 query does not contain $sender_host_name. From release 4.23, a reference to
2041 $sender_host_name causes it to be looked up, so we don't need to do the lookup
2042 on spec. */
2043
2044 if ((semicolon = Ustrchr(ss, ';')) != NULL)
2045 {
2046 uschar *affix;
2047 int partial, affixlen, starflags, id;
2048
2049 *semicolon = 0;
2050 id = search_findtype_partial(ss, &partial, &affix, &affixlen, &starflags);
2051 *semicolon=';';
2052
2053 if (id < 0) /* Unknown lookup type */
2054 {
2055 log_write(0, LOG_MAIN|LOG_PANIC, "%s in host list item \"%s\"",
2056 search_error_message, ss);
2057 return DEFER;
2058 }
2059 isquery = mac_islookup(id, lookup_querystyle|lookup_absfilequery);
2060 }
2061
2062 if (isquery)
2063 {
2064 switch(match_check_string(US"", ss, -1, TRUE, TRUE, TRUE, valueptr))
2065 {
2066 case OK: return OK;
2067 case DEFER: return DEFER;
2068 default: return FAIL;
2069 }
2070 }
2071
2072 /* Not a query-style lookup; must ensure the host name is present, and then we
2073 do a check on the name and all its aliases. */
2074
2075 if (sender_host_name == NULL)
2076 {
2077 HDEBUG(D_host_lookup)
2078 debug_printf("sender host name required, to match against %s\n", ss);
2079 if (host_lookup_failed || host_name_lookup() != OK)
2080 {
2081 *error = string_sprintf("failed to find host name for %s",
2082 sender_host_address);;
2083 return ERROR;
2084 }
2085 host_build_sender_fullhost();
2086 }
2087
2088 /* Match on the sender host name, using the general matching function */
2089
2090 switch(match_check_string(sender_host_name, ss, -1, TRUE, TRUE, TRUE,
2091 valueptr))
2092 {
2093 case OK: return OK;
2094 case DEFER: return DEFER;
2095 }
2096
2097 /* If there are aliases, try matching on them. */
2098
2099 aliases = sender_host_aliases;
2100 while (*aliases != NULL)
2101 {
2102 switch(match_check_string(*aliases++, ss, -1, TRUE, TRUE, TRUE, valueptr))
2103 {
2104 case OK: return OK;
2105 case DEFER: return DEFER;
2106 }
2107 }
2108 return FAIL;
2109 }
2110
2111
2112
2113
2114 /*************************************************
2115 * Check a specific host matches a host list *
2116 *************************************************/
2117
2118 /* This function is passed a host list containing items in a number of
2119 different formats and the identity of a host. Its job is to determine whether
2120 the given host is in the set of hosts defined by the list. The host name is
2121 passed as a pointer so that it can be looked up if needed and not already
2122 known. This is commonly the case when called from verify_check_host() to check
2123 an incoming connection. When called from elsewhere the host name should usually
2124 be set.
2125
2126 This function is now just a front end to match_check_list(), which runs common
2127 code for scanning a list. We pass it the check_host() function to perform a
2128 single test.
2129
2130 Arguments:
2131 listptr pointer to the host list
2132 cache_bits pointer to cache for named lists, or NULL
2133 host_name the host name or NULL, implying use sender_host_name and
2134 sender_host_aliases, looking them up if required
2135 host_address the IP address
2136 valueptr if not NULL, data from a lookup is passed back here
2137
2138 Returns: OK if the host is in the defined set
2139 FAIL if the host is not in the defined set,
2140 DEFER if a data lookup deferred (not a host lookup)
2141
2142 If the host name was needed in order to make a comparison, and could not be
2143 determined from the IP address, the result is FAIL unless the item
2144 "+allow_unknown" was met earlier in the list, in which case OK is returned. */
2145
2146 int
2147 verify_check_this_host(uschar **listptr, unsigned int *cache_bits,
2148 uschar *host_name, uschar *host_address, uschar **valueptr)
2149 {
2150 int rc;
2151 unsigned int *local_cache_bits = cache_bits;
2152 uschar *save_host_address = deliver_host_address;
2153 check_host_block cb;
2154 cb.host_name = host_name;
2155 cb.host_address = host_address;
2156
2157 if (valueptr != NULL) *valueptr = NULL;
2158
2159 /* If the host address starts off ::ffff: it is an IPv6 address in
2160 IPv4-compatible mode. Find the IPv4 part for checking against IPv4
2161 addresses. */
2162
2163 cb.host_ipv4 = (Ustrncmp(host_address, "::ffff:", 7) == 0)?
2164 host_address + 7 : host_address;
2165
2166 /* During the running of the check, put the IP address into $host_address. In
2167 the case of calls from the smtp transport, it will already be there. However,
2168 in other calls (e.g. when testing ignore_target_hosts), it won't. Just to be on
2169 the safe side, any existing setting is preserved, though as I write this
2170 (November 2004) I can't see any cases where it is actually needed. */
2171
2172 deliver_host_address = host_address;
2173 rc = match_check_list(
2174 listptr, /* the list */
2175 0, /* separator character */
2176 &hostlist_anchor, /* anchor pointer */
2177 &local_cache_bits, /* cache pointer */
2178 check_host, /* function for testing */
2179 &cb, /* argument for function */
2180 MCL_HOST, /* type of check */
2181 (host_address == sender_host_address)?
2182 US"host" : host_address, /* text for debugging */
2183 valueptr); /* where to pass back data */
2184 deliver_host_address = save_host_address;
2185 return rc;
2186 }
2187
2188
2189
2190
2191 /*************************************************
2192 * Check the remote host matches a list *
2193 *************************************************/
2194
2195 /* This is a front end to verify_check_this_host(), created because checking
2196 the remote host is a common occurrence. With luck, a good compiler will spot
2197 the tail recursion and optimize it. If there's no host address, this is
2198 command-line SMTP input - check against an empty string for the address.
2199
2200 Arguments:
2201 listptr pointer to the host list
2202
2203 Returns: the yield of verify_check_this_host(),
2204 i.e. OK, FAIL, or DEFER
2205 */
2206
2207 int
2208 verify_check_host(uschar **listptr)
2209 {
2210 return verify_check_this_host(listptr, sender_host_cache, NULL,
2211 (sender_host_address == NULL)? US"" : sender_host_address, NULL);
2212 }
2213
2214
2215
2216
2217
2218 /*************************************************
2219 * Invert an IP address for a DNS black list *
2220 *************************************************/
2221
2222 /*
2223 Arguments:
2224 buffer where to put the answer
2225 address the address to invert
2226 */
2227
2228 static void
2229 invert_address(uschar *buffer, uschar *address)
2230 {
2231 int bin[4];
2232 uschar *bptr = buffer;
2233
2234 /* If this is an IPv4 address mapped into IPv6 format, adjust the pointer
2235 to the IPv4 part only. */
2236
2237 if (Ustrncmp(address, "::ffff:", 7) == 0) address += 7;
2238
2239 /* Handle IPv4 address: when HAVE_IPV6 is false, the result of host_aton() is
2240 always 1. */
2241
2242 if (host_aton(address, bin) == 1)
2243 {
2244 int i;
2245 int x = bin[0];
2246 for (i = 0; i < 4; i++)
2247 {
2248 sprintf(CS bptr, "%d.", x & 255);
2249 while (*bptr) bptr++;
2250 x >>= 8;
2251 }
2252 }
2253
2254 /* Handle IPv6 address. Actually, as far as I know, there are no IPv6 addresses
2255 in any DNS black lists, and the format in which they will be looked up is
2256 unknown. This is just a guess. */
2257
2258 #if HAVE_IPV6
2259 else
2260 {
2261 int i, j;
2262 for (j = 3; j >= 0; j--)
2263 {
2264 int x = bin[j];
2265 for (i = 0; i < 8; i++)
2266 {
2267 sprintf(CS bptr, "%x.", x & 15);
2268 while (*bptr) bptr++;
2269 x >>= 4;
2270 }
2271 }
2272 }
2273 #endif
2274 }
2275
2276
2277
2278 /*************************************************
2279 * Perform a single dnsbl lookup *
2280 *************************************************/
2281
2282 /* This function is called from verify_check_dnsbl() below.
2283
2284 Arguments:
2285 domain the outer dnsbl domain (for debug message)
2286 keydomain the current keydomain (for debug message)
2287 query the domain to be looked up
2288 iplist the list of matching IP addresses
2289 bitmask true if bitmask matching is wanted
2290 invert_result true if result to be inverted
2291 defer_return what to return for a defer
2292
2293 Returns: OK if lookup succeeded
2294 FAIL if not
2295 */
2296
2297 static int
2298 one_check_dnsbl(uschar *domain, uschar *keydomain, uschar *query,
2299 uschar *iplist, BOOL bitmask, BOOL invert_result, int defer_return)
2300 {
2301 dns_answer dnsa;
2302 dns_scan dnss;
2303 tree_node *t;
2304 dnsbl_cache_block *cb;
2305 int old_pool = store_pool;
2306
2307 /* Look for this query in the cache. */
2308
2309 t = tree_search(dnsbl_cache, query);
2310
2311 /* If not cached from a previous lookup, we must do a DNS lookup, and
2312 cache the result in permanent memory. */
2313
2314 if (t == NULL)
2315 {
2316 store_pool = POOL_PERM;
2317
2318 /* Set up a tree entry to cache the lookup */
2319
2320 t = store_get(sizeof(tree_node) + Ustrlen(query));
2321 Ustrcpy(t->name, query);
2322 t->data.ptr = cb = store_get(sizeof(dnsbl_cache_block));
2323 (void)tree_insertnode(&dnsbl_cache, t);
2324
2325 /* Do the DNS loopup . */
2326
2327 HDEBUG(D_dnsbl) debug_printf("new DNS lookup for %s\n", query);
2328 cb->rc = dns_basic_lookup(&dnsa, query, T_A);
2329 cb->text_set = FALSE;
2330 cb->text = NULL;
2331 cb->rhs = NULL;
2332
2333 /* If the lookup succeeded, cache the RHS address. The code allows for
2334 more than one address - this was for complete generality and the possible
2335 use of A6 records. However, A6 records have been reduced to experimental
2336 status (August 2001) and may die out. So they may never get used at all,
2337 let alone in dnsbl records. However, leave the code here, just in case.
2338
2339 Quite apart from one A6 RR generating multiple addresses, there are DNS
2340 lists that return more than one A record, so we must handle multiple
2341 addresses generated in that way as well. */
2342
2343 if (cb->rc == DNS_SUCCEED)
2344 {
2345 dns_record *rr;
2346 dns_address **addrp = &(cb->rhs);
2347 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2348 rr != NULL;
2349 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2350 {
2351 if (rr->type == T_A)
2352 {
2353 dns_address *da = dns_address_from_rr(&dnsa, rr);
2354 if (da != NULL)
2355 {
2356 *addrp = da;
2357 while (da->next != NULL) da = da->next;
2358 addrp = &(da->next);
2359 }
2360 }
2361 }
2362
2363 /* If we didn't find any A records, change the return code. This can
2364 happen when there is a CNAME record but there are no A records for what
2365 it points to. */
2366
2367 if (cb->rhs == NULL) cb->rc = DNS_NODATA;
2368 }
2369
2370 store_pool = old_pool;
2371 }
2372
2373 /* Previous lookup was cached */
2374
2375 else
2376 {
2377 HDEBUG(D_dnsbl) debug_printf("using result of previous DNS lookup\n");
2378 cb = t->data.ptr;
2379 }
2380
2381 /* We now have the result of the DNS lookup, either newly done, or cached
2382 from a previous call. If the lookup succeeded, check against the address
2383 list if there is one. This may be a positive equality list (introduced by
2384 "="), a negative equality list (introduced by "!="), a positive bitmask
2385 list (introduced by "&"), or a negative bitmask list (introduced by "!&").*/
2386
2387 if (cb->rc == DNS_SUCCEED)
2388 {
2389 dns_address *da = NULL;
2390 uschar *addlist = cb->rhs->address;
2391
2392 /* For A and AAAA records, there may be multiple addresses from multiple
2393 records. For A6 records (currently not expected to be used) there may be
2394 multiple addresses from a single record. */
2395
2396 for (da = cb->rhs->next; da != NULL; da = da->next)
2397 addlist = string_sprintf("%s, %s", addlist, da->address);
2398
2399 HDEBUG(D_dnsbl) debug_printf("DNS lookup for %s succeeded (yielding %s)\n",
2400 query, addlist);
2401
2402 /* Address list check; this can be either for equality, or via a bitmask.
2403 In the latter case, all the bits must match. */
2404
2405 if (iplist != NULL)
2406 {
2407 int ipsep = ',';
2408 uschar ip[46];
2409 uschar *ptr = iplist;
2410
2411 while (string_nextinlist(&ptr, &ipsep, ip, sizeof(ip)) != NULL)
2412 {
2413 /* Handle exact matching */
2414 if (!bitmask)
2415 {
2416 for (da = cb->rhs; da != NULL; da = da->next)
2417 {
2418 if (Ustrcmp(CS da->address, ip) == 0) break;
2419 }
2420 }
2421 /* Handle bitmask matching */
2422 else
2423 {
2424 int address[4];
2425 int mask = 0;
2426
2427 /* At present, all known DNS blocking lists use A records, with
2428 IPv4 addresses on the RHS encoding the information they return. I
2429 wonder if this will linger on as the last vestige of IPv4 when IPv6
2430 is ubiquitous? Anyway, for now we use paranoia code to completely
2431 ignore IPv6 addresses. The default mask is 0, which always matches.
2432 We change this only for IPv4 addresses in the list. */
2433
2434 if (host_aton(ip, address) == 1) mask = address[0];
2435
2436 /* Scan the returned addresses, skipping any that are IPv6 */
2437
2438 for (da = cb->rhs; da != NULL; da = da->next)
2439 {
2440 if (host_aton(da->address, address) != 1) continue;
2441 if ((address[0] & mask) == mask) break;
2442 }
2443 }
2444
2445 /* Break out if a match has been found */
2446
2447 if (da != NULL) break;
2448 }
2449
2450 /* If either
2451
2452 (a) No IP address in a positive list matched, or
2453 (b) An IP address in a negative list did match
2454
2455 then behave as if the DNSBL lookup had not succeeded, i.e. the host is
2456 not on the list. */
2457
2458 if (invert_result != (da == NULL))
2459 {
2460 HDEBUG(D_dnsbl)
2461 {
2462 debug_printf("=> but we are not accepting this block class because\n");
2463 debug_printf("=> there was %s match for %c%s\n",
2464 invert_result? "an exclude":"no", bitmask? '&' : '=', iplist);
2465 }
2466 return FAIL;
2467 }
2468 }
2469
2470 /* Either there was no IP list, or the record matched. Look up a TXT record
2471 if it hasn't previously been done. */
2472
2473 if (!cb->text_set)
2474 {
2475 cb->text_set = TRUE;
2476 if (dns_basic_lookup(&dnsa, query, T_TXT) == DNS_SUCCEED)
2477 {
2478 dns_record *rr;
2479 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2480 rr != NULL;
2481 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2482 if (rr->type == T_TXT) break;
2483 if (rr != NULL)
2484 {
2485 int len = (rr->data)[0];
2486 if (len > 511) len = 127;
2487 store_pool = POOL_PERM;
2488 cb->text = string_sprintf("%.*s", len, (const uschar *)(rr->data+1));
2489 store_pool = old_pool;
2490 }
2491 }
2492 }
2493
2494 dnslist_value = addlist;
2495 dnslist_text = cb->text;
2496 return OK;
2497 }
2498
2499 /* There was a problem with the DNS lookup */
2500
2501 if (cb->rc != DNS_NOMATCH && cb->rc != DNS_NODATA)
2502 {
2503 log_write(L_dnslist_defer, LOG_MAIN,
2504 "DNS list lookup defer (probably timeout) for %s: %s", query,
2505 (defer_return == OK)? US"assumed in list" :
2506 (defer_return == FAIL)? US"assumed not in list" :
2507 US"returned DEFER");
2508 return defer_return;
2509 }
2510
2511 /* No entry was found in the DNS; continue for next domain */
2512
2513 HDEBUG(D_dnsbl)
2514 {
2515 debug_printf("DNS lookup for %s failed\n", query);
2516 debug_printf("=> that means %s is not listed at %s\n",
2517 keydomain, domain);
2518 }
2519
2520 return FAIL;
2521 }
2522
2523
2524
2525
2526 /*************************************************
2527 * Check host against DNS black lists *
2528 *************************************************/
2529
2530 /* This function runs checks against a list of DNS black lists, until one
2531 matches. Each item on the list can be of the form
2532
2533 domain=ip-address/key
2534
2535 The domain is the right-most domain that is used for the query, for example,
2536 blackholes.mail-abuse.org. If the IP address is present, there is a match only
2537 if the DNS lookup returns a matching IP address. Several addresses may be
2538 given, comma-separated, for example: x.y.z=127.0.0.1,127.0.0.2.
2539
2540 If no key is given, what is looked up in the domain is the inverted IP address
2541 of the current client host. If a key is given, it is used to construct the
2542 domain for the lookup. For example,
2543
2544 dsn.rfc-ignorant.org/$sender_address_domain
2545
2546 After finding a match in the DNS, the domain is placed in $dnslist_domain, and
2547 then we check for a TXT record for an error message, and if found, save its
2548 value in $dnslist_text. We also cache everything in a tree, to optimize
2549 multiple lookups.
2550
2551 Note: an address for testing RBL is 192.203.178.39
2552 Note: an address for testing DUL is 192.203.178.4
2553 Note: a domain for testing RFCI is example.tld.dsn.rfc-ignorant.org
2554
2555 Arguments:
2556 listptr the domain/address/data list
2557
2558 Returns: OK successful lookup (i.e. the address is on the list), or
2559 lookup deferred after +include_unknown
2560 FAIL name not found, or no data found for the given type, or
2561 lookup deferred after +exclude_unknown (default)
2562 DEFER lookup failure, if +defer_unknown was set
2563 */
2564
2565 int
2566 verify_check_dnsbl(uschar **listptr)
2567 {
2568 int sep = 0;
2569 int defer_return = FAIL;
2570 BOOL invert_result = FALSE;
2571 uschar *list = *listptr;
2572 uschar *domain;
2573 uschar *s;
2574 uschar buffer[1024];
2575 uschar query[256]; /* DNS domain max length */
2576 uschar revadd[128]; /* Long enough for IPv6 address */
2577
2578 /* Indicate that the inverted IP address is not yet set up */
2579
2580 revadd[0] = 0;
2581
2582 /* In case this is the first time the DNS resolver is being used. */
2583
2584 dns_init(FALSE, FALSE);
2585
2586 /* Loop through all the domains supplied, until something matches */
2587
2588 while ((domain = string_nextinlist(&list, &sep, buffer, sizeof(buffer))) != NULL)
2589 {
2590 int rc;
2591 BOOL frc;
2592 BOOL bitmask = FALSE;
2593 uschar *iplist;
2594 uschar *key;
2595
2596 HDEBUG(D_dnsbl) debug_printf("DNS list check: %s\n", domain);
2597
2598 /* Deal with special values that change the behaviour on defer */
2599
2600 if (domain[0] == '+')
2601 {
2602 if (strcmpic(domain, US"+include_unknown") == 0) defer_return = OK;
2603 else if (strcmpic(domain, US"+exclude_unknown") == 0) defer_return = FAIL;
2604 else if (strcmpic(domain, US"+defer_unknown") == 0) defer_return = DEFER;
2605 else
2606 log_write(0, LOG_MAIN|LOG_PANIC, "unknown item in dnslist (ignored): %s",
2607 domain);
2608 continue;
2609 }
2610
2611 /* See if there's explicit data to be looked up */
2612
2613 key = Ustrchr(domain, '/');
2614 if (key != NULL) *key++ = 0;
2615
2616 /* See if there's a list of addresses supplied after the domain name. This is
2617 introduced by an = or a & character; if preceded by ! we invert the result.
2618 */
2619
2620 iplist = Ustrchr(domain, '=');
2621 if (iplist == NULL)
2622 {
2623 bitmask = TRUE;
2624 iplist = Ustrchr(domain, '&');
2625 }
2626
2627 if (iplist != NULL)
2628 {
2629 if (iplist > domain && iplist[-1] == '!')
2630 {
2631 invert_result = TRUE;
2632 iplist[-1] = 0;
2633 }
2634 *iplist++ = 0;
2635 }
2636
2637 /* Check that what we have left is a sensible domain name. There is no reason
2638 why these domains should in fact use the same syntax as hosts and email
2639 domains, but in practice they seem to. However, there is little point in
2640 actually causing an error here, because that would no doubt hold up incoming
2641 mail. Instead, I'll just log it. */
2642
2643 for (s = domain; *s != 0; s++)
2644 {
2645 if (!isalnum(*s) && *s != '-' && *s != '.')
2646 {
2647 log_write(0, LOG_MAIN, "dnslists domain \"%s\" contains "
2648 "strange characters - is this right?", domain);
2649 break;
2650 }
2651 }
2652
2653 /* If there is no key string, construct the query by adding the domain name
2654 onto the inverted host address, and perform a single DNS lookup. */
2655
2656 if (key == NULL)
2657 {
2658 if (sender_host_address == NULL) return FAIL; /* can never match */
2659 if (revadd[0] == 0) invert_address(revadd, sender_host_address);
2660 frc = string_format(query, sizeof(query), "%s%s", revadd, domain);
2661
2662 if (!frc)
2663 {
2664 log_write(0, LOG_MAIN|LOG_PANIC, "dnslist query is too long "
2665 "(ignored): %s...", query);
2666 continue;
2667 }
2668
2669 rc = one_check_dnsbl(domain, sender_host_address, query, iplist, bitmask,
2670 invert_result, defer_return);
2671
2672 if (rc == OK)
2673 {
2674 dnslist_domain = string_copy(domain);
2675 HDEBUG(D_dnsbl) debug_printf("=> that means %s is listed at %s\n",
2676 sender_host_address, domain);
2677 }
2678
2679 if (rc != FAIL) return rc; /* OK or DEFER */
2680 }
2681
2682 /* If there is a key string, it can be a list of domains or IP addresses to
2683 be concatenated with the main domain. */
2684
2685 else
2686 {
2687 int keysep = 0;
2688 BOOL defer = FALSE;
2689 uschar *keydomain;
2690 uschar keybuffer[256];
2691
2692 while ((keydomain = string_nextinlist(&key, &keysep, keybuffer,
2693 sizeof(keybuffer))) != NULL)
2694 {
2695 if (string_is_ip_address(keydomain, NULL) > 0)
2696 {
2697 uschar keyrevadd[128];
2698 invert_address(keyrevadd, keydomain);
2699 frc = string_format(query, sizeof(query), "%s%s", keyrevadd, domain);
2700 }
2701 else
2702 {
2703 frc = string_format(query, sizeof(query), "%s.%s", keydomain, domain);
2704 }
2705
2706 if (!frc)
2707 {
2708 log_write(0, LOG_MAIN|LOG_PANIC, "dnslist query is too long "
2709 "(ignored): %s...", query);
2710 continue;
2711 }
2712
2713 rc = one_check_dnsbl(domain, keydomain, query, iplist, bitmask,
2714 invert_result, defer_return);
2715
2716 if (rc == OK)
2717 {
2718 dnslist_domain = string_copy(domain);
2719 HDEBUG(D_dnsbl) debug_printf("=> that means %s is listed at %s\n",
2720 keydomain, domain);
2721 return OK;
2722 }
2723
2724 /* If the lookup deferred, remember this fact. We keep trying the rest
2725 of the list to see if we get a useful result, and if we don't, we return
2726 DEFER at the end. */
2727
2728 if (rc == DEFER) defer = TRUE;
2729 } /* continue with next keystring domain/address */
2730
2731 if (defer) return DEFER;
2732 }
2733 } /* continue with next dnsdb outer domain */
2734
2735 return FAIL;
2736 }
2737
2738 /* End of verify.c */