de7a36642b71ee5d41c21ee374b5ce5fbf703924
[exim.git] / src / src / verify.c
1 /* $Cambridge: exim/src/src/verify.c,v 1.4 2004/11/11 11:40:36 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 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_random => do the "random" thing
132 vopt_callout_recipsender => use real sender for recipient
133 vopt_callout_recippmaster => use postmaster for recipient
134 se_mailfrom MAIL FROM address for sender verify; NULL => ""
135 pm_mailfrom if non-NULL, do the postmaster check with this sender
136
137 Returns: OK/FAIL/DEFER
138 */
139
140 static int
141 do_callout(address_item *addr, host_item *host_list, transport_feedback *tf,
142 int callout, int callout_overall, int callout_connect, int options,
143 uschar *se_mailfrom, uschar *pm_mailfrom)
144 {
145 BOOL is_recipient = (options & vopt_is_recipient) != 0;
146 BOOL callout_no_cache = (options & vopt_callout_no_cache) != 0;
147 BOOL callout_random = (options & vopt_callout_random) != 0;
148
149 int yield = OK;
150 BOOL done = FALSE;
151 uschar *address_key;
152 uschar *from_address;
153 uschar *random_local_part = NULL;
154 uschar **failure_ptr = is_recipient?
155 &recipient_verify_failure : &sender_verify_failure;
156 open_db dbblock;
157 open_db *dbm_file = NULL;
158 dbdata_callout_cache new_domain_record;
159 dbdata_callout_cache_address new_address_record;
160 host_item *host;
161 time_t callout_start_time;
162
163 new_domain_record.result = ccache_unknown;
164 new_domain_record.postmaster_result = ccache_unknown;
165 new_domain_record.random_result = ccache_unknown;
166
167 memset(&new_address_record, 0, sizeof(new_address_record));
168
169 /* For a recipient callout, the key used for the address cache record must
170 include the sender address if we are using the real sender in the callout,
171 because that may influence the result of the callout. */
172
173 address_key = addr->address;
174 from_address = US"";
175
176 if (is_recipient)
177 {
178 if ((options & vopt_callout_recipsender) != 0)
179 {
180 address_key = string_sprintf("%s/<%s>", addr->address, sender_address);
181 from_address = sender_address;
182 }
183 else if ((options & vopt_callout_recippmaster) != 0)
184 {
185 address_key = string_sprintf("%s/<postmaster@%s>", addr->address,
186 qualify_domain_sender);
187 from_address = string_sprintf("postmaster@%s", qualify_domain_sender);
188 }
189 }
190
191 /* For a sender callout, we must adjust the key if the mailfrom address is not
192 empty. */
193
194 else
195 {
196 from_address = (se_mailfrom == NULL)? US"" : se_mailfrom;
197 if (from_address[0] != 0)
198 address_key = string_sprintf("%s/<%s>", addr->address, from_address);
199 }
200
201 /* Open the callout cache database, it it exists, for reading only at this
202 stage, unless caching has been disabled. */
203
204 if (callout_no_cache)
205 {
206 HDEBUG(D_verify) debug_printf("callout cache: disabled by no_cache\n");
207 }
208 else if ((dbm_file = dbfn_open(US"callout", O_RDWR, &dbblock, FALSE)) == NULL)
209 {
210 HDEBUG(D_verify) debug_printf("callout cache: not available\n");
211 }
212
213 /* If a cache database is available see if we can avoid the need to do an
214 actual callout by making use of previously-obtained data. */
215
216 if (dbm_file != NULL)
217 {
218 dbdata_callout_cache_address *cache_address_record;
219 dbdata_callout_cache *cache_record = get_callout_cache_record(dbm_file,
220 addr->domain, US"domain",
221 callout_cache_domain_positive_expire,
222 callout_cache_domain_negative_expire);
223
224 /* If an unexpired cache record was found for this domain, see if the callout
225 process can be short-circuited. */
226
227 if (cache_record != NULL)
228 {
229 /* If an early command (up to and including MAIL FROM:<>) was rejected,
230 there is no point carrying on. The callout fails. */
231
232 if (cache_record->result == ccache_reject)
233 {
234 setflag(addr, af_verify_nsfail);
235 HDEBUG(D_verify)
236 debug_printf("callout cache: domain gave initial rejection, or "
237 "does not accept HELO or MAIL FROM:<>\n");
238 setflag(addr, af_verify_nsfail);
239 addr->user_message = US"(result of an earlier callout reused).";
240 yield = FAIL;
241 *failure_ptr = US"mail";
242 goto END_CALLOUT;
243 }
244
245 /* If a previous check on a "random" local part was accepted, we assume
246 that the server does not do any checking on local parts. There is therefore
247 no point in doing the callout, because it will always be successful. If a
248 random check previously failed, arrange not to do it again, but preserve
249 the data in the new record. If a random check is required but hasn't been
250 done, skip the remaining cache processing. */
251
252 if (callout_random) switch(cache_record->random_result)
253 {
254 case ccache_accept:
255 HDEBUG(D_verify)
256 debug_printf("callout cache: domain accepts random addresses\n");
257 goto END_CALLOUT; /* Default yield is OK */
258
259 case ccache_reject:
260 HDEBUG(D_verify)
261 debug_printf("callout cache: domain rejects random addresses\n");
262 callout_random = FALSE;
263 new_domain_record.random_result = ccache_reject;
264 new_domain_record.random_stamp = cache_record->random_stamp;
265 break;
266
267 default:
268 HDEBUG(D_verify)
269 debug_printf("callout cache: need to check random address handling "
270 "(not cached or cache expired)\n");
271 goto END_CACHE;
272 }
273
274 /* If a postmaster check is requested, but there was a previous failure,
275 there is again no point in carrying on. If a postmaster check is required,
276 but has not been done before, we are going to have to do a callout, so skip
277 remaining cache processing. */
278
279 if (pm_mailfrom != NULL)
280 {
281 if (cache_record->postmaster_result == ccache_reject)
282 {
283 setflag(addr, af_verify_pmfail);
284 HDEBUG(D_verify)
285 debug_printf("callout cache: domain does not accept "
286 "RCPT TO:<postmaster@domain>\n");
287 yield = FAIL;
288 *failure_ptr = US"postmaster";
289 setflag(addr, af_verify_pmfail);
290 addr->user_message = US"(result of earlier verification reused).";
291 goto END_CALLOUT;
292 }
293 if (cache_record->postmaster_result == ccache_unknown)
294 {
295 HDEBUG(D_verify)
296 debug_printf("callout cache: need to check RCPT "
297 "TO:<postmaster@domain> (not cached or cache expired)\n");
298 goto END_CACHE;
299 }
300
301 /* If cache says OK, set pm_mailfrom NULL to prevent a redundant
302 postmaster check if the address itself has to be checked. Also ensure
303 that the value in the cache record is preserved (with its old timestamp).
304 */
305
306 HDEBUG(D_verify) debug_printf("callout cache: domain accepts RCPT "
307 "TO:<postmaster@domain>\n");
308 pm_mailfrom = NULL;
309 new_domain_record.postmaster_result = ccache_accept;
310 new_domain_record.postmaster_stamp = cache_record->postmaster_stamp;
311 }
312 }
313
314 /* We can't give a result based on information about the domain. See if there
315 is an unexpired cache record for this specific address (combined with the
316 sender address if we are doing a recipient callout with a non-empty sender).
317 */
318
319 cache_address_record = (dbdata_callout_cache_address *)
320 get_callout_cache_record(dbm_file,
321 address_key, US"address",
322 callout_cache_positive_expire,
323 callout_cache_negative_expire);
324
325 if (cache_address_record != NULL)
326 {
327 if (cache_address_record->result == ccache_accept)
328 {
329 HDEBUG(D_verify)
330 debug_printf("callout cache: address record is positive\n");
331 }
332 else
333 {
334 HDEBUG(D_verify)
335 debug_printf("callout cache: address record is negative\n");
336 addr->user_message = US"Previous (cached) callout verification failure";
337 *failure_ptr = US"recipient";
338 yield = FAIL;
339 }
340 goto END_CALLOUT;
341 }
342
343 /* Close the cache database while we actually do the callout for real. */
344
345 END_CACHE:
346 dbfn_close(dbm_file);
347 dbm_file = NULL;
348 }
349
350 /* The information wasn't available in the cache, so we have to do a real
351 callout and save the result in the cache for next time, unless no_cache is set,
352 or unless we have a previously cached negative random result. If we are to test
353 with a random local part, ensure that such a local part is available. If not,
354 log the fact, but carry on without randomming. */
355
356 if (callout_random && callout_random_local_part != NULL)
357 {
358 random_local_part = expand_string(callout_random_local_part);
359 if (random_local_part == NULL)
360 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand "
361 "callout_random_local_part: %s", expand_string_message);
362 }
363
364 /* Default the connect and overall callout timeouts if not set, and record the
365 time we are starting so that we can enforce it. */
366
367 if (callout_overall < 0) callout_overall = 4 * callout;
368 if (callout_connect < 0) callout_connect = callout;
369 callout_start_time = time(NULL);
370
371 /* Now make connections to the hosts and do real callouts. The list of hosts
372 is passed in as an argument. */
373
374 for (host = host_list; host != NULL && !done; host = host->next)
375 {
376 smtp_inblock inblock;
377 smtp_outblock outblock;
378 int host_af;
379 int port = 25;
380 uschar *helo = US"HELO";
381 uschar *interface = NULL; /* Outgoing interface to use; NULL => any */
382 uschar inbuffer[4096];
383 uschar outbuffer[1024];
384 uschar responsebuffer[4096];
385
386 clearflag(addr, af_verify_pmfail); /* postmaster callout flag */
387 clearflag(addr, af_verify_nsfail); /* null sender callout flag */
388
389 /* Skip this host if we don't have an IP address for it. */
390
391 if (host->address == NULL)
392 {
393 DEBUG(D_verify) debug_printf("no IP address for host name %s: skipping\n",
394 host->name);
395 continue;
396 }
397
398 /* Check the overall callout timeout */
399
400 if (time(NULL) - callout_start_time >= callout_overall)
401 {
402 HDEBUG(D_verify) debug_printf("overall timeout for callout exceeded\n");
403 break;
404 }
405
406 /* Set IPv4 or IPv6 */
407
408 host_af = (Ustrchr(host->address, ':') == NULL)? AF_INET:AF_INET6;
409
410 /* Expand and interpret the interface and port strings. This has to
411 be delayed till now, because they may expand differently for different
412 hosts. If there's a failure, log it, but carry on with the defaults. */
413
414 deliver_host = host->name;
415 deliver_host_address = host->address;
416 if (!smtp_get_interface(tf->interface, host_af, addr, NULL, &interface,
417 US"callout") ||
418 !smtp_get_port(tf->port, addr, &port, US"callout"))
419 log_write(0, LOG_MAIN|LOG_PANIC, "<%s>: %s", addr->address,
420 addr->message);
421 deliver_host = deliver_host_address = NULL;
422
423 /* Set HELO string according to the protocol */
424
425 if (Ustrcmp(tf->protocol, "lmtp") == 0) helo = US"LHLO";
426
427 HDEBUG(D_verify) debug_printf("interface=%s port=%d\n", interface, port);
428
429 /* Set up the buffer for reading SMTP response packets. */
430
431 inblock.buffer = inbuffer;
432 inblock.buffersize = sizeof(inbuffer);
433 inblock.ptr = inbuffer;
434 inblock.ptrend = inbuffer;
435
436 /* Set up the buffer for holding SMTP commands while pipelining */
437
438 outblock.buffer = outbuffer;
439 outblock.buffersize = sizeof(outbuffer);
440 outblock.ptr = outbuffer;
441 outblock.cmd_count = 0;
442 outblock.authenticating = FALSE;
443
444 /* Connect to the host; on failure, just loop for the next one, but we
445 set the error for the last one. Use the callout_connect timeout. */
446
447 inblock.sock = outblock.sock =
448 smtp_connect(host, host_af, port, interface, callout_connect, TRUE);
449 if (inblock.sock < 0)
450 {
451 addr->message = string_sprintf("could not connect to %s [%s]: %s",
452 host->name, host->address, strerror(errno));
453 continue;
454 }
455
456 /* Wait for initial response, and then run the initial SMTP commands. The
457 smtp_write_command() function leaves its command in big_buffer. This is
458 used in error responses. Initialize it in case the connection is
459 rejected. */
460
461 Ustrcpy(big_buffer, "initial connection");
462
463 done =
464 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
465 '2', callout) &&
466
467 smtp_write_command(&outblock, FALSE, "%s %s\r\n", helo,
468 smtp_active_hostname) >= 0 &&
469 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
470 '2', callout) &&
471
472 smtp_write_command(&outblock, FALSE, "MAIL FROM:<%s>\r\n",
473 from_address) >= 0 &&
474 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
475 '2', callout);
476
477 /* If the host gave an initial error, or does not accept HELO or MAIL
478 FROM:<>, arrange to cache this information, but don't record anything for an
479 I/O error or a defer. Do not cache rejections when a non-empty sender has
480 been used, because that blocks the whole domain for all senders. */
481
482 if (!done)
483 {
484 *failure_ptr = US"mail";
485 if (errno == 0 && responsebuffer[0] == '5')
486 {
487 setflag(addr, af_verify_nsfail);
488 if (from_address[0] == 0) new_domain_record.result = ccache_reject;
489 }
490 }
491
492 /* Otherwise, proceed to check a "random" address (if required), then the
493 given address, and the postmaster address (if required). Between each check,
494 issue RSET, because some servers accept only one recipient after MAIL
495 FROM:<>. */
496
497 else
498 {
499 new_domain_record.result = ccache_accept;
500
501 /* Do the random local part check first */
502
503 if (random_local_part != NULL)
504 {
505 uschar randombuffer[1024];
506 BOOL random_ok =
507 smtp_write_command(&outblock, FALSE,
508 "RCPT TO:<%.1000s@%.1000s>\r\n", random_local_part,
509 addr->domain) >= 0 &&
510 smtp_read_response(&inblock, randombuffer,
511 sizeof(randombuffer), '2', callout);
512
513 /* Remember when we last did a random test */
514
515 new_domain_record.random_stamp = time(NULL);
516
517 /* If accepted, we aren't going to do any further tests below. */
518
519 if (random_ok)
520 {
521 new_domain_record.random_result = ccache_accept;
522 }
523
524 /* Otherwise, cache a real negative response, and get back to the right
525 state to send RCPT. Unless there's some problem such as a dropped
526 connection, we expect to succeed, because the commands succeeded above. */
527
528 else if (errno == 0)
529 {
530 if (randombuffer[0] == '5')
531 new_domain_record.random_result = ccache_reject;
532
533 done =
534 smtp_write_command(&outblock, FALSE, "RSET\r\n") >= 0 &&
535 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
536 '2', callout) &&
537
538 smtp_write_command(&outblock, FALSE, "MAIL FROM:<>\r\n") >= 0 &&
539 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
540 '2', callout);
541 }
542 else done = FALSE; /* Some timeout/connection problem */
543 } /* Random check */
544
545 /* If the host is accepting all local parts, as determined by the "random"
546 check, we don't need to waste time doing any further checking. */
547
548 if (new_domain_record.random_result != ccache_accept && done)
549 {
550 done =
551 smtp_write_command(&outblock, FALSE, "RCPT TO:<%.1000s>\r\n",
552 addr->address) >= 0 &&
553 smtp_read_response(&inblock, responsebuffer, sizeof(responsebuffer),
554 '2', callout);
555
556 if (done)
557 new_address_record.result = ccache_accept;
558 else if (errno == 0 && responsebuffer[0] == '5')
559 {
560 *failure_ptr = US"recipient";
561 new_address_record.result = ccache_reject;
562 }
563
564 /* Do postmaster check if requested */
565
566 if (done && pm_mailfrom != NULL)
567 {
568 done =
569 smtp_write_command(&outblock, FALSE, "RSET\r\n") >= 0 &&
570 smtp_read_response(&inblock, responsebuffer,
571 sizeof(responsebuffer), '2', callout) &&
572
573 smtp_write_command(&outblock, FALSE,
574 "MAIL FROM:<%s>\r\n", pm_mailfrom) >= 0 &&
575 smtp_read_response(&inblock, responsebuffer,
576 sizeof(responsebuffer), '2', callout) &&
577
578 smtp_write_command(&outblock, FALSE,
579 "RCPT TO:<postmaster@%.1000s>\r\n", addr->domain) >= 0 &&
580 smtp_read_response(&inblock, responsebuffer,
581 sizeof(responsebuffer), '2', callout);
582
583 new_domain_record.postmaster_stamp = time(NULL);
584
585 if (done)
586 new_domain_record.postmaster_result = ccache_accept;
587 else if (errno == 0 && responsebuffer[0] == '5')
588 {
589 *failure_ptr = US"postmaster";
590 setflag(addr, af_verify_pmfail);
591 new_domain_record.postmaster_result = ccache_reject;
592 }
593 }
594 } /* Random not accepted */
595 } /* MAIL FROM:<> accepted */
596
597 /* For any failure of the main check, other than a negative response, we just
598 close the connection and carry on. We can identify a negative response by the
599 fact that errno is zero. For I/O errors it will be non-zero
600
601 Set up different error texts for logging and for sending back to the caller
602 as an SMTP response. Log in all cases, using a one-line format. For sender
603 callouts, give a full response to the caller, but for recipient callouts,
604 don't give the IP address because this may be an internal host whose identity
605 is not to be widely broadcast. */
606
607 if (!done)
608 {
609 if (errno == ETIMEDOUT)
610 {
611 HDEBUG(D_verify) debug_printf("SMTP timeout\n");
612 }
613 else if (errno == 0)
614 {
615 if (*responsebuffer == 0) Ustrcpy(responsebuffer, US"connection dropped");
616
617 addr->message =
618 string_sprintf("response to \"%s\" from %s [%s] was: %s",
619 big_buffer, host->name, host->address,
620 string_printing(responsebuffer));
621
622 addr->user_message = is_recipient?
623 string_sprintf("Callout verification failed:\n%s", responsebuffer)
624 :
625 string_sprintf("Called: %s\nSent: %s\nResponse: %s",
626 host->address, big_buffer, responsebuffer);
627
628 /* Hard rejection ends the process */
629
630 if (responsebuffer[0] == '5') /* Address rejected */
631 {
632 yield = FAIL;
633 done = TRUE;
634 }
635 }
636 }
637
638 /* End the SMTP conversation and close the connection. */
639
640 (void)smtp_write_command(&outblock, FALSE, "QUIT\r\n");
641 close(inblock.sock);
642 } /* Loop through all hosts, while !done */
643
644 /* If we get here with done == TRUE, a successful callout happened, and yield
645 will be set OK or FAIL according to the response to the RCPT command.
646 Otherwise, we looped through the hosts but couldn't complete the business.
647 However, there may be domain-specific information to cache in both cases.
648
649 The value of the result field in the new_domain record is ccache_unknown if
650 there was an error before or with MAIL FROM:<>, and errno was not zero,
651 implying some kind of I/O error. We don't want to write the cache in that case.
652 Otherwise the value is ccache_accept or ccache_reject. */
653
654 if (!callout_no_cache && new_domain_record.result != ccache_unknown)
655 {
656 if ((dbm_file = dbfn_open(US"callout", O_RDWR|O_CREAT, &dbblock, FALSE))
657 == NULL)
658 {
659 HDEBUG(D_verify) debug_printf("callout cache: not available\n");
660 }
661 else
662 {
663 (void)dbfn_write(dbm_file, addr->domain, &new_domain_record,
664 (int)sizeof(dbdata_callout_cache));
665 HDEBUG(D_verify) debug_printf("wrote callout cache domain record:\n"
666 " result=%d postmaster=%d random=%d\n",
667 new_domain_record.result,
668 new_domain_record.postmaster_result,
669 new_domain_record.random_result);
670 }
671 }
672
673 /* If a definite result was obtained for the callout, cache it unless caching
674 is disabled. */
675
676 if (done)
677 {
678 if (!callout_no_cache && new_address_record.result != ccache_unknown)
679 {
680 if (dbm_file == NULL)
681 dbm_file = dbfn_open(US"callout", O_RDWR|O_CREAT, &dbblock, FALSE);
682 if (dbm_file == NULL)
683 {
684 HDEBUG(D_verify) debug_printf("no callout cache available\n");
685 }
686 else
687 {
688 (void)dbfn_write(dbm_file, address_key, &new_address_record,
689 (int)sizeof(dbdata_callout_cache_address));
690 HDEBUG(D_verify) debug_printf("wrote %s callout cache address record\n",
691 (new_address_record.result == ccache_accept)? "positive" : "negative");
692 }
693 }
694 } /* done */
695
696 /* Failure to connect to any host, or any response other than 2xx or 5xx is a
697 temporary error. If there was only one host, and a response was received, leave
698 it alone if supplying details. Otherwise, give a generic response. */
699
700 else /* !done */
701 {
702 uschar *dullmsg = string_sprintf("Could not complete %s verify callout",
703 is_recipient? "recipient" : "sender");
704 yield = DEFER;
705
706 if (host_list->next != NULL || addr->message == NULL) addr->message = dullmsg;
707
708 addr->user_message = (!smtp_return_error_details)? dullmsg :
709 string_sprintf("%s for <%s>.\n"
710 "The mail server(s) for the domain may be temporarily unreachable, or\n"
711 "they may be permanently unreachable from this server. In the latter case,\n%s",
712 dullmsg, addr->address,
713 is_recipient?
714 "the address will never be accepted."
715 :
716 "you need to change the address or create an MX record for its domain\n"
717 "if it is supposed to be generally accessible from the Internet.\n"
718 "Talk to your mail administrator for details.");
719
720 /* Force a specific error code */
721
722 addr->basic_errno = ERRNO_CALLOUTDEFER;
723 }
724
725 /* Come here from within the cache-reading code on fast-track exit. */
726
727 END_CALLOUT:
728 if (dbm_file != NULL) dbfn_close(dbm_file);
729 return yield;
730 }
731
732
733
734 /*************************************************
735 * Copy error to toplevel address *
736 *************************************************/
737
738 /* This function is used when a verify fails or defers, to ensure that the
739 failure or defer information is in the original toplevel address. This applies
740 when an address is redirected to a single new address, and the failure or
741 deferral happens to the child address.
742
743 Arguments:
744 vaddr the verify address item
745 addr the final address item
746 yield FAIL or DEFER
747
748 Returns: the value of YIELD
749 */
750
751 static int
752 copy_error(address_item *vaddr, address_item *addr, int yield)
753 {
754 if (addr != vaddr)
755 {
756 vaddr->message = addr->message;
757 vaddr->user_message = addr->user_message;
758 vaddr->basic_errno = addr->basic_errno;
759 vaddr->more_errno = addr->more_errno;
760 }
761 return yield;
762 }
763
764
765
766
767 /*************************************************
768 * Verify an email address *
769 *************************************************/
770
771 /* This function is used both for verification (-bv and at other times) and
772 address testing (-bt), which is indicated by address_test_mode being set.
773
774 Arguments:
775 vaddr contains the address to verify; the next field in this block
776 must be NULL
777 f if not NULL, write the result to this file
778 options various option bits:
779 vopt_fake_sender => this sender verify is not for the real
780 sender (it was verify=sender=xxxx or an address from a
781 header line) - rewriting must not change sender_address
782 vopt_is_recipient => this is a recipient address, otherwise
783 it's a sender address - this affects qualification and
784 rewriting and messages from callouts
785 vopt_qualify => qualify an unqualified address; else error
786 vopt_expn => called from SMTP EXPN command
787
788 These ones are used by do_callout() -- the options variable
789 is passed to it.
790
791 vopt_callout_no_cache => don't use callout cache
792 vopt_callout_random => do the "random" thing
793 vopt_callout_recipsender => use real sender for recipient
794 vopt_callout_recippmaster => use postmaster for recipient
795
796 callout if > 0, specifies that callout is required, and gives timeout
797 for individual commands
798 callout_overall if > 0, gives overall timeout for the callout function;
799 if < 0, a default is used (see do_callout())
800 callout_connect the connection timeout for callouts
801 se_mailfrom when callout is requested to verify a sender, use this
802 in MAIL FROM; NULL => ""
803 pm_mailfrom when callout is requested, if non-NULL, do the postmaster
804 thing and use this as the sender address (may be "")
805
806 routed if not NULL, set TRUE if routing succeeded, so we can
807 distinguish between routing failed and callout failed
808
809 Returns: OK address verified
810 FAIL address failed to verify
811 DEFER can't tell at present
812 */
813
814 int
815 verify_address(address_item *vaddr, FILE *f, int options, int callout,
816 int callout_overall, int callout_connect, uschar *se_mailfrom,
817 uschar *pm_mailfrom, BOOL *routed)
818 {
819 BOOL allok = TRUE;
820 BOOL full_info = (f == NULL)? FALSE : (debug_selector != 0);
821 BOOL is_recipient = (options & vopt_is_recipient) != 0;
822 BOOL expn = (options & vopt_expn) != 0;
823 int i;
824 int yield = OK;
825 int verify_type = expn? v_expn :
826 address_test_mode? v_none :
827 is_recipient? v_recipient : v_sender;
828 address_item *addr_list;
829 address_item *addr_new = NULL;
830 address_item *addr_remote = NULL;
831 address_item *addr_local = NULL;
832 address_item *addr_succeed = NULL;
833 uschar **failure_ptr = is_recipient?
834 &recipient_verify_failure : &sender_verify_failure;
835 uschar *ko_prefix, *cr;
836 uschar *address = vaddr->address;
837 uschar *save_sender;
838 uschar null_sender[] = { 0 }; /* Ensure writeable memory */
839
840 /* Clear, just in case */
841
842 *failure_ptr = NULL;
843
844 /* Set up a prefix and suffix for error message which allow us to use the same
845 output statements both in EXPN mode (where an SMTP response is needed) and when
846 debugging with an output file. */
847
848 if (expn)
849 {
850 ko_prefix = US"553 ";
851 cr = US"\r";
852 }
853 else ko_prefix = cr = US"";
854
855 /* Add qualify domain if permitted; otherwise an unqualified address fails. */
856
857 if (parse_find_at(address) == NULL)
858 {
859 if ((options & vopt_qualify) == 0)
860 {
861 if (f != NULL)
862 fprintf(f, "%sA domain is required for \"%s\"%s\n", ko_prefix, address,
863 cr);
864 *failure_ptr = US"qualify";
865 return FAIL;
866 }
867 address = rewrite_address_qualify(address, is_recipient);
868 }
869
870 DEBUG(D_verify)
871 {
872 debug_printf(">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>\n");
873 debug_printf("%s %s\n", address_test_mode? "Testing" : "Verifying", address);
874 }
875
876 /* Rewrite and report on it. Clear the domain and local part caches - these
877 may have been set by domains and local part tests during an ACL. */
878
879 if (global_rewrite_rules != NULL)
880 {
881 uschar *old = address;
882 address = rewrite_address(address, is_recipient, FALSE,
883 global_rewrite_rules, rewrite_existflags);
884 if (address != old)
885 {
886 for (i = 0; i < (MAX_NAMED_LIST * 2)/32; i++) vaddr->localpart_cache[i] = 0;
887 for (i = 0; i < (MAX_NAMED_LIST * 2)/32; i++) vaddr->domain_cache[i] = 0;
888 if (f != NULL && !expn) fprintf(f, "Address rewritten as: %s\n", address);
889 }
890 }
891
892 /* If this is the real sender address, we must update sender_address at
893 this point, because it may be referred to in the routers. */
894
895 if ((options & (vopt_fake_sender|vopt_is_recipient)) == 0)
896 sender_address = address;
897
898 /* If the address was rewritten to <> no verification can be done, and we have
899 to return OK. This rewriting is permitted only for sender addresses; for other
900 addresses, such rewriting fails. */
901
902 if (address[0] == 0) return OK;
903
904 /* Save a copy of the sender address for re-instating if we change it to <>
905 while verifying a sender address (a nice bit of self-reference there). */
906
907 save_sender = sender_address;
908
909 /* Update the address structure with the possibly qualified and rewritten
910 address. Set it up as the starting address on the chain of new addresses. */
911
912 vaddr->address = address;
913 addr_new = vaddr;
914
915 /* We need a loop, because an address can generate new addresses. We must also
916 cope with generated pipes and files at the top level. (See also the code and
917 comment in deliver.c.) However, it is usually the case that the router for
918 user's .forward files has its verify flag turned off.
919
920 If an address generates more than one child, the loop is used only when
921 full_info is set, and this can only be set locally. Remote enquiries just get
922 information about the top level address, not anything that it generated. */
923
924 while (addr_new != NULL)
925 {
926 int rc;
927 address_item *addr = addr_new;
928
929 addr_new = addr->next;
930 addr->next = NULL;
931
932 DEBUG(D_verify)
933 {
934 debug_printf(">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>\n");
935 debug_printf("Considering %s\n", addr->address);
936 }
937
938 /* Handle generated pipe, file or reply addresses. We don't get these
939 when handling EXPN, as it does only one level of expansion. */
940
941 if (testflag(addr, af_pfr))
942 {
943 allok = FALSE;
944 if (f != NULL)
945 {
946 BOOL allow;
947
948 if (addr->address[0] == '>')
949 {
950 allow = testflag(addr, af_allow_reply);
951 fprintf(f, "%s -> mail %s", addr->parent->address, addr->address + 1);
952 }
953 else
954 {
955 allow = (addr->address[0] == '|')?
956 testflag(addr, af_allow_pipe) : testflag(addr, af_allow_file);
957 fprintf(f, "%s -> %s", addr->parent->address, addr->address);
958 }
959
960 if (addr->basic_errno == ERRNO_BADTRANSPORT)
961 fprintf(f, "\n*** Error in setting up pipe, file, or autoreply:\n"
962 "%s\n", addr->message);
963 else if (allow)
964 fprintf(f, "\n transport = %s\n", addr->transport->name);
965 else
966 fprintf(f, " *** forbidden ***\n");
967 }
968 continue;
969 }
970
971 /* Just in case some router parameter refers to it. */
972
973 return_path = (addr->p.errors_address != NULL)?
974 addr->p.errors_address : sender_address;
975
976 /* Split the address into domain and local part, handling the %-hack if
977 necessary, and then route it. While routing a sender address, set
978 $sender_address to <> because that is what it will be if we were trying to
979 send a bounce to the sender. */
980
981 if (routed != NULL) *routed = FALSE;
982 if ((rc = deliver_split_address(addr)) == OK)
983 {
984 if (!is_recipient) sender_address = null_sender;
985 rc = route_address(addr, &addr_local, &addr_remote, &addr_new,
986 &addr_succeed, verify_type);
987 sender_address = save_sender; /* Put back the real sender */
988 }
989
990 /* If routing an address succeeded, set the flag that remembers, for use when
991 an ACL cached a sender verify (in case a callout fails). Then if routing set
992 up a list of hosts or the transport has a host list, and the callout option
993 is set, and we aren't in a host checking run, do the callout verification,
994 and set another flag that notes that a callout happened. */
995
996 if (rc == OK)
997 {
998 if (routed != NULL) *routed = TRUE;
999 if (callout > 0)
1000 {
1001 host_item *host_list = addr->host_list;
1002
1003 /* Default, if no remote transport, to NULL for the interface (=> any),
1004 "smtp" for the port, and "smtp" for the protocol. */
1005
1006 transport_feedback tf = { NULL, US"smtp", US"smtp", NULL, FALSE, FALSE };
1007
1008 /* If verification yielded a remote transport, we want to use that
1009 transport's options, so as to mimic what would happen if we were really
1010 sending a message to this address. */
1011
1012 if (addr->transport != NULL && !addr->transport->info->local)
1013 {
1014 (void)(addr->transport->setup)(addr->transport, addr, &tf, NULL);
1015
1016 /* If the transport has hosts and the router does not, or if the
1017 transport is configured to override the router's hosts, we must build a
1018 host list of the transport's hosts, and find the IP addresses */
1019
1020 if (tf.hosts != NULL && (host_list == NULL || tf.hosts_override))
1021 {
1022 uschar *s;
1023
1024 host_list = NULL; /* Ignore the router's hosts */
1025
1026 deliver_domain = addr->domain;
1027 deliver_localpart = addr->local_part;
1028 s = expand_string(tf.hosts);
1029 deliver_domain = deliver_localpart = NULL;
1030
1031 if (s == NULL)
1032 {
1033 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand list of hosts "
1034 "\"%s\" in %s transport for callout: %s", tf.hosts,
1035 addr->transport->name, expand_string_message);
1036 }
1037 else
1038 {
1039 uschar *canonical_name;
1040 host_item *host;
1041 host_build_hostlist(&host_list, s, tf.hosts_randomize);
1042
1043 /* Just ignore failures to find a host address. If we don't manage
1044 to find any addresses, the callout will defer. */
1045
1046 for (host = host_list; host != NULL; host = host->next)
1047 {
1048 if (tf.gethostbyname || string_is_ip_address(host->name, NULL))
1049 (void)host_find_byname(host, NULL, &canonical_name, TRUE);
1050 else
1051 {
1052 int flags = HOST_FIND_BY_A;
1053 if (tf.qualify_single) flags |= HOST_FIND_QUALIFY_SINGLE;
1054 if (tf.search_parents) flags |= HOST_FIND_SEARCH_PARENTS;
1055 (void)host_find_bydns(host, NULL, flags, NULL, NULL, NULL,
1056 &canonical_name, NULL);
1057 }
1058 }
1059 }
1060 }
1061 }
1062
1063 /* Can only do a callout if we have at least one host! If the callout
1064 fails, it will have set ${sender,recipient}_verify_failure. */
1065
1066 if (host_list != NULL)
1067 {
1068 HDEBUG(D_verify) debug_printf("Attempting full verification using callout\n");
1069 if (host_checking && !host_checking_callout)
1070 {
1071 HDEBUG(D_verify)
1072 debug_printf("... callout omitted by default when host testing\n"
1073 "(Use -bhc if you want the callouts to happen.)\n");
1074 }
1075 else
1076 {
1077 rc = do_callout(addr, host_list, &tf, callout, callout_overall,
1078 callout_connect, options, se_mailfrom, pm_mailfrom);
1079 }
1080 }
1081 else
1082 {
1083 HDEBUG(D_verify) debug_printf("Cannot do callout: neither router nor "
1084 "transport provided a host list\n");
1085 }
1086 }
1087 }
1088
1089 /* Otherwise, any failure is a routing failure */
1090
1091 else *failure_ptr = US"route";
1092
1093 /* A router may return REROUTED if it has set up a child address as a result
1094 of a change of domain name (typically from widening). In this case we always
1095 want to continue to verify the new child. */
1096
1097 if (rc == REROUTED) continue;
1098
1099 /* Handle hard failures */
1100
1101 if (rc == FAIL)
1102 {
1103 allok = FALSE;
1104 if (f != NULL)
1105 {
1106 fprintf(f, "%s%s %s", ko_prefix, address,
1107 address_test_mode? "is undeliverable" : "failed to verify");
1108 if (!expn && admin_user)
1109 {
1110 if (addr->basic_errno > 0)
1111 fprintf(f, ": %s", strerror(addr->basic_errno));
1112 if (addr->message != NULL)
1113 fprintf(f, ":\n %s", addr->message);
1114 }
1115 fprintf(f, "%s\n", cr);
1116 }
1117
1118 if (!full_info) return copy_error(vaddr, addr, FAIL);
1119 else yield = FAIL;
1120 }
1121
1122 /* Soft failure */
1123
1124 else if (rc == DEFER)
1125 {
1126 allok = FALSE;
1127 if (f != NULL)
1128 {
1129 fprintf(f, "%s%s cannot be resolved at this time", ko_prefix, address);
1130 if (!expn && admin_user)
1131 {
1132 if (addr->basic_errno > 0)
1133 fprintf(f, ":\n %s", strerror(addr->basic_errno));
1134 if (addr->message != NULL)
1135 fprintf(f, ":\n %s", addr->message);
1136 else if (addr->basic_errno <= 0)
1137 fprintf(f, ":\n unknown error");
1138 }
1139
1140 fprintf(f, "%s\n", cr);
1141 }
1142 if (!full_info) return copy_error(vaddr, addr, DEFER);
1143 else if (yield == OK) yield = DEFER;
1144 }
1145
1146 /* If we are handling EXPN, we do not want to continue to route beyond
1147 the top level. */
1148
1149 else if (expn)
1150 {
1151 uschar *ok_prefix = US"250-";
1152 if (addr_new == NULL)
1153 {
1154 if (addr_local == NULL && addr_remote == NULL)
1155 fprintf(f, "250 mail to <%s> is discarded\r\n", address);
1156 else
1157 fprintf(f, "250 <%s>\r\n", address);
1158 }
1159 else while (addr_new != NULL)
1160 {
1161 address_item *addr2 = addr_new;
1162 addr_new = addr2->next;
1163 if (addr_new == NULL) ok_prefix = US"250 ";
1164 fprintf(f, "%s<%s>\r\n", ok_prefix, addr2->address);
1165 }
1166 return OK;
1167 }
1168
1169 /* Successful routing other than EXPN. */
1170
1171 else
1172 {
1173 /* Handle successful routing when short info wanted. Otherwise continue for
1174 other (generated) addresses. Short info is the operational case. Full info
1175 can be requested only when debug_selector != 0 and a file is supplied.
1176
1177 There is a conflict between the use of aliasing as an alternate email
1178 address, and as a sort of mailing list. If an alias turns the incoming
1179 address into just one address (e.g. J.Caesar->jc44) you may well want to
1180 carry on verifying the generated address to ensure it is valid when
1181 checking incoming mail. If aliasing generates multiple addresses, you
1182 probably don't want to do this. Exim therefore treats the generation of
1183 just a single new address as a special case, and continues on to verify the
1184 generated address. */
1185
1186 if (!full_info && /* Stop if short info wanted AND */
1187 (addr_new == NULL || /* No new address OR */
1188 addr_new->next != NULL || /* More than one new address OR */
1189 testflag(addr_new, af_pfr))) /* New address is pfr */
1190 {
1191 if (f != NULL) fprintf(f, "%s %s\n", address,
1192 address_test_mode? "is deliverable" : "verified");
1193
1194 /* If we have carried on to verify a child address, we want the value
1195 of $address_data to be that of the child */
1196
1197 vaddr->p.address_data = addr->p.address_data;
1198 return OK;
1199 }
1200 }
1201 } /* Loop for generated addresses */
1202
1203 /* Display the full results of the successful routing, including any generated
1204 addresses. Control gets here only when full_info is set, which requires f not
1205 to be NULL, and this occurs only when a top-level verify is called with the
1206 debugging switch on.
1207
1208 If there are no local and no remote addresses, and there were no pipes, files,
1209 or autoreplies, and there were no errors or deferments, the message is to be
1210 discarded, usually because of the use of :blackhole: in an alias file. */
1211
1212 if (allok && addr_local == NULL && addr_remote == NULL)
1213 fprintf(f, "mail to %s is discarded\n", address);
1214
1215 else for (addr_list = addr_local, i = 0; i < 2; addr_list = addr_remote, i++)
1216 {
1217 while (addr_list != NULL)
1218 {
1219 address_item *addr = addr_list;
1220 address_item *p = addr->parent;
1221 addr_list = addr->next;
1222
1223 fprintf(f, "%s", CS addr->address);
1224 while (p != NULL)
1225 {
1226 fprintf(f, "\n <-- %s", p->address);
1227 p = p->parent;
1228 }
1229 fprintf(f, "\n ");
1230
1231 /* Show router, and transport */
1232
1233 fprintf(f, "router = %s, ", addr->router->name);
1234 fprintf(f, "transport = %s\n", (addr->transport == NULL)? US"unset" :
1235 addr->transport->name);
1236
1237 /* Show any hosts that are set up by a router unless the transport
1238 is going to override them; fiddle a bit to get a nice format. */
1239
1240 if (addr->host_list != NULL && addr->transport != NULL &&
1241 !addr->transport->overrides_hosts)
1242 {
1243 host_item *h;
1244 int maxlen = 0;
1245 int maxaddlen = 0;
1246 for (h = addr->host_list; h != NULL; h = h->next)
1247 {
1248 int len = Ustrlen(h->name);
1249 if (len > maxlen) maxlen = len;
1250 len = (h->address != NULL)? Ustrlen(h->address) : 7;
1251 if (len > maxaddlen) maxaddlen = len;
1252 }
1253 for (h = addr->host_list; h != NULL; h = h->next)
1254 {
1255 int len = Ustrlen(h->name);
1256 fprintf(f, " host %s ", h->name);
1257 while (len++ < maxlen) fprintf(f, " ");
1258 if (h->address != NULL)
1259 {
1260 fprintf(f, "[%s] ", h->address);
1261 len = Ustrlen(h->address);
1262 }
1263 else if (!addr->transport->info->local) /* Omit [unknown] for local */
1264 {
1265 fprintf(f, "[unknown] ");
1266 len = 7;
1267 }
1268 else len = -3;
1269 while (len++ < maxaddlen) fprintf(f," ");
1270 if (h->mx >= 0) fprintf(f, "MX=%d", h->mx);
1271 if (h->port != PORT_NONE) fprintf(f, " port=%d", h->port);
1272 if (h->status == hstatus_unusable) fprintf(f, " ** unusable **");
1273 fprintf(f, "\n");
1274 }
1275 }
1276 }
1277 }
1278
1279 /* Will be DEFER or FAIL if any one address has, only for full_info (which is
1280 the -bv or -bt case). */
1281
1282 return yield;
1283 }
1284
1285
1286
1287
1288 /*************************************************
1289 * Check headers for syntax errors *
1290 *************************************************/
1291
1292 /* This function checks those header lines that contain addresses, and verifies
1293 that all the addresses therein are syntactially correct.
1294
1295 Arguments:
1296 msgptr where to put an error message
1297
1298 Returns: OK
1299 FAIL
1300 */
1301
1302 int
1303 verify_check_headers(uschar **msgptr)
1304 {
1305 header_line *h;
1306 uschar *colon, *s;
1307
1308 for (h = header_list; h != NULL; h = h->next)
1309 {
1310 if (h->type != htype_from &&
1311 h->type != htype_reply_to &&
1312 h->type != htype_sender &&
1313 h->type != htype_to &&
1314 h->type != htype_cc &&
1315 h->type != htype_bcc)
1316 continue;
1317
1318 colon = Ustrchr(h->text, ':');
1319 s = colon + 1;
1320 while (isspace(*s)) s++;
1321
1322 parse_allow_group = TRUE; /* Allow group syntax */
1323
1324 /* Loop for multiple addresses in the header */
1325
1326 while (*s != 0)
1327 {
1328 uschar *ss = parse_find_address_end(s, FALSE);
1329 uschar *recipient, *errmess;
1330 int terminator = *ss;
1331 int start, end, domain;
1332
1333 /* Temporarily terminate the string at this point, and extract the
1334 operative address within. */
1335
1336 *ss = 0;
1337 recipient = parse_extract_address(s,&errmess,&start,&end,&domain,FALSE);
1338 *ss = terminator;
1339
1340 /* Permit an unqualified address only if the message is local, or if the
1341 sending host is configured to be permitted to send them. */
1342
1343 if (recipient != NULL && domain == 0)
1344 {
1345 if (h->type == htype_from || h->type == htype_sender)
1346 {
1347 if (!allow_unqualified_sender) recipient = NULL;
1348 }
1349 else
1350 {
1351 if (!allow_unqualified_recipient) recipient = NULL;
1352 }
1353 if (recipient == NULL) errmess = US"unqualified address not permitted";
1354 }
1355
1356 /* It's an error if no address could be extracted, except for the special
1357 case of an empty address. */
1358
1359 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
1360 {
1361 uschar *verb = US"is";
1362 uschar *t = ss;
1363 int len;
1364
1365 /* Arrange not to include any white space at the end in the
1366 error message. */
1367
1368 while (t > s && isspace(t[-1])) t--;
1369
1370 /* Add the address which failed to the error message, since in a
1371 header with very many addresses it is sometimes hard to spot
1372 which one is at fault. However, limit the amount of address to
1373 quote - cases have been seen where, for example, a missing double
1374 quote in a humungous To: header creates an "address" that is longer
1375 than string_sprintf can handle. */
1376
1377 len = t - s;
1378 if (len > 1024)
1379 {
1380 len = 1024;
1381 verb = US"begins";
1382 }
1383
1384 *msgptr = string_printing(
1385 string_sprintf("%s: failing address in \"%.*s\" header %s: %.*s",
1386 errmess, colon - h->text, h->text, verb, len, s));
1387
1388 return FAIL;
1389 }
1390
1391 /* Advance to the next address */
1392
1393 s = ss + (terminator? 1:0);
1394 while (isspace(*s)) s++;
1395 } /* Next address */
1396 } /* Next header */
1397
1398 return OK;
1399 }
1400
1401
1402
1403
1404 /*************************************************
1405 * Find if verified sender *
1406 *************************************************/
1407
1408 /* Usually, just a single address is verified as the sender of the message.
1409 However, Exim can be made to verify other addresses as well (often related in
1410 some way), and this is useful in some environments. There may therefore be a
1411 chain of such addresses that have previously been tested. This function finds
1412 whether a given address is on the chain.
1413
1414 Arguments: the address to be verified
1415 Returns: pointer to an address item, or NULL
1416 */
1417
1418 address_item *
1419 verify_checked_sender(uschar *sender)
1420 {
1421 address_item *addr;
1422 for (addr = sender_verified_list; addr != NULL; addr = addr->next)
1423 if (Ustrcmp(sender, addr->address) == 0) break;
1424 return addr;
1425 }
1426
1427
1428
1429
1430
1431 /*************************************************
1432 * Get valid header address *
1433 *************************************************/
1434
1435 /* Scan the originator headers of the message, looking for an address that
1436 verifies successfully. RFC 822 says:
1437
1438 o The "Sender" field mailbox should be sent notices of
1439 any problems in transport or delivery of the original
1440 messages. If there is no "Sender" field, then the
1441 "From" field mailbox should be used.
1442
1443 o If the "Reply-To" field exists, then the reply should
1444 go to the addresses indicated in that field and not to
1445 the address(es) indicated in the "From" field.
1446
1447 So we check a Sender field if there is one, else a Reply_to field, else a From
1448 field. As some strange messages may have more than one of these fields,
1449 especially if they are resent- fields, check all of them if there is more than
1450 one.
1451
1452 Arguments:
1453 user_msgptr points to where to put a user error message
1454 log_msgptr points to where to put a log error message
1455 callout timeout for callout check (passed to verify_address())
1456 callout_overall overall callout timeout (ditto)
1457 callout_connect connect callout timeout (ditto)
1458 se_mailfrom mailfrom for verify; NULL => ""
1459 pm_mailfrom sender for pm callout check (passed to verify_address())
1460 options callout options (passed to verify_address())
1461
1462 If log_msgptr is set to something without setting user_msgptr, the caller
1463 normally uses log_msgptr for both things.
1464
1465 Returns: result of the verification attempt: OK, FAIL, or DEFER;
1466 FAIL is given if no appropriate headers are found
1467 */
1468
1469 int
1470 verify_check_header_address(uschar **user_msgptr, uschar **log_msgptr,
1471 int callout, int callout_overall, int callout_connect, uschar *se_mailfrom,
1472 uschar *pm_mailfrom, int options)
1473 {
1474 static int header_types[] = { htype_sender, htype_reply_to, htype_from };
1475 int yield = FAIL;
1476 int i;
1477
1478 for (i = 0; i < 3; i++)
1479 {
1480 header_line *h;
1481 for (h = header_list; h != NULL; h = h->next)
1482 {
1483 int terminator, new_ok;
1484 uschar *s, *ss, *endname;
1485
1486 if (h->type != header_types[i]) continue;
1487 s = endname = Ustrchr(h->text, ':') + 1;
1488
1489 while (*s != 0)
1490 {
1491 address_item *vaddr;
1492
1493 while (isspace(*s) || *s == ',') s++;
1494 if (*s == 0) break; /* End of header */
1495
1496 ss = parse_find_address_end(s, FALSE);
1497
1498 /* The terminator is a comma or end of header, but there may be white
1499 space preceding it (including newline for the last address). Move back
1500 past any white space so we can check against any cached envelope sender
1501 address verifications. */
1502
1503 while (isspace(ss[-1])) ss--;
1504 terminator = *ss;
1505 *ss = 0;
1506
1507 HDEBUG(D_verify) debug_printf("verifying %.*s header address %s\n",
1508 (int)(endname - h->text), h->text, s);
1509
1510 /* See if we have already verified this address as an envelope sender,
1511 and if so, use the previous answer. */
1512
1513 vaddr = verify_checked_sender(s);
1514
1515 if (vaddr != NULL && /* Previously checked */
1516 (callout <= 0 || /* No callout needed; OR */
1517 vaddr->special_action > 256)) /* Callout was done */
1518 {
1519 new_ok = vaddr->special_action & 255;
1520 HDEBUG(D_verify) debug_printf("previously checked as envelope sender\n");
1521 *ss = terminator; /* Restore shortened string */
1522 }
1523
1524 /* Otherwise we run the verification now. We must restore the shortened
1525 string before running the verification, so the headers are correct, in
1526 case there is any rewriting. */
1527
1528 else
1529 {
1530 int start, end, domain;
1531 uschar *address = parse_extract_address(s, log_msgptr, &start,
1532 &end, &domain, FALSE);
1533
1534 *ss = terminator;
1535
1536 /* If verification failed because of a syntax error, fail this
1537 function, and ensure that the failing address gets added to the error
1538 message. */
1539
1540 if (address == NULL)
1541 {
1542 new_ok = FAIL;
1543 if (*log_msgptr != NULL)
1544 {
1545 while (ss > s && isspace(ss[-1])) ss--;
1546 *log_msgptr = string_sprintf("syntax error in '%.*s' header when "
1547 "scanning for sender: %s in \"%.*s\"",
1548 endname - h->text, h->text, *log_msgptr, ss - s, s);
1549 return FAIL;
1550 }
1551 }
1552
1553 /* Else go ahead with the sender verification. But is isn't *the*
1554 sender of the message, so set vopt_fake_sender to stop sender_address
1555 being replaced after rewriting or qualification. */
1556
1557 else
1558 {
1559 vaddr = deliver_make_addr(address, FALSE);
1560 new_ok = verify_address(vaddr, NULL, options | vopt_fake_sender,
1561 callout, callout_overall, callout_connect, se_mailfrom,
1562 pm_mailfrom, NULL);
1563 }
1564 }
1565
1566 /* We now have the result, either newly found, or cached. If we are
1567 giving out error details, set a specific user error. This means that the
1568 last of these will be returned to the user if all three fail. We do not
1569 set a log message - the generic one below will be used. */
1570
1571 if (new_ok != OK && smtp_return_error_details)
1572 {
1573 *user_msgptr = string_sprintf("Rejected after DATA: "
1574 "could not verify \"%.*s\" header address\n%s: %s",
1575 endname - h->text, h->text, vaddr->address, vaddr->message);
1576 }
1577
1578 /* Success or defer */
1579
1580 if (new_ok == OK) return OK;
1581 if (new_ok == DEFER) yield = DEFER;
1582
1583 /* Move on to any more addresses in the header */
1584
1585 s = ss;
1586 }
1587 }
1588 }
1589
1590 if (yield == FAIL && *log_msgptr == NULL)
1591 *log_msgptr = US"there is no valid sender in any header line";
1592
1593 if (yield == DEFER && *log_msgptr == NULL)
1594 *log_msgptr = US"all attempts to verify a sender in a header line deferred";
1595
1596 return yield;
1597 }
1598
1599
1600
1601
1602 /*************************************************
1603 * Get RFC 1413 identification *
1604 *************************************************/
1605
1606 /* Attempt to get an id from the sending machine via the RFC 1413 protocol. If
1607 the timeout is set to zero, then the query is not done. There may also be lists
1608 of hosts and nets which are exempt. To guard against malefactors sending
1609 non-printing characters which could, for example, disrupt a message's headers,
1610 make sure the string consists of printing characters only.
1611
1612 Argument:
1613 port the port to connect to; usually this is IDENT_PORT (113), but when
1614 running in the test harness with -bh a different value is used.
1615
1616 Returns: nothing
1617
1618 Side effect: any received ident value is put in sender_ident (NULL otherwise)
1619 */
1620
1621 void
1622 verify_get_ident(int port)
1623 {
1624 int sock, host_af, qlen;
1625 int received_sender_port, received_interface_port, n;
1626 uschar *p;
1627 uschar buffer[2048];
1628
1629 /* Default is no ident. Check whether we want to do an ident check for this
1630 host. */
1631
1632 sender_ident = NULL;
1633 if (rfc1413_query_timeout <= 0 || verify_check_host(&rfc1413_hosts) != OK)
1634 return;
1635
1636 DEBUG(D_ident) debug_printf("doing ident callback\n");
1637
1638 /* Set up a connection to the ident port of the remote host. Bind the local end
1639 to the incoming interface address. If the sender host address is an IPv6
1640 address, the incoming interface address will also be IPv6. */
1641
1642 host_af = (Ustrchr(sender_host_address, ':') == NULL)? AF_INET : AF_INET6;
1643 sock = ip_socket(SOCK_STREAM, host_af);
1644 if (sock < 0) return;
1645
1646 if (ip_bind(sock, host_af, interface_address, 0) < 0)
1647 {
1648 DEBUG(D_ident) debug_printf("bind socket for ident failed: %s\n",
1649 strerror(errno));
1650 goto END_OFF;
1651 }
1652
1653 if (ip_connect(sock, host_af, sender_host_address, port, rfc1413_query_timeout)
1654 < 0)
1655 {
1656 if (errno == ETIMEDOUT && (log_extra_selector & LX_ident_timeout) != 0)
1657 {
1658 log_write(0, LOG_MAIN, "ident connection to %s timed out",
1659 sender_host_address);
1660 }
1661 else
1662 {
1663 DEBUG(D_ident) debug_printf("ident connection to %s failed: %s\n",
1664 sender_host_address, strerror(errno));
1665 }
1666 goto END_OFF;
1667 }
1668
1669 /* Construct and send the query. */
1670
1671 sprintf(CS buffer, "%d , %d\r\n", sender_host_port, interface_port);
1672 qlen = Ustrlen(buffer);
1673 if (send(sock, buffer, qlen, 0) < 0)
1674 {
1675 DEBUG(D_ident) debug_printf("ident send failed: %s\n", strerror(errno));
1676 goto END_OFF;
1677 }
1678
1679 /* Read a response line. We put it into the rest of the buffer, using several
1680 recv() calls if necessary. */
1681
1682 p = buffer + qlen;
1683
1684 for (;;)
1685 {
1686 uschar *pp;
1687 int count;
1688 int size = sizeof(buffer) - (p - buffer);
1689
1690 if (size <= 0) goto END_OFF; /* Buffer filled without seeing \n. */
1691 count = ip_recv(sock, p, size, rfc1413_query_timeout);
1692 if (count <= 0) goto END_OFF; /* Read error or EOF */
1693
1694 /* Scan what we just read, to see if we have reached the terminating \r\n. Be
1695 generous, and accept a plain \n terminator as well. The only illegal
1696 character is 0. */
1697
1698 for (pp = p; pp < p + count; pp++)
1699 {
1700 if (*pp == 0) goto END_OFF; /* Zero octet not allowed */
1701 if (*pp == '\n')
1702 {
1703 if (pp[-1] == '\r') pp--;
1704 *pp = 0;
1705 goto GOT_DATA; /* Break out of both loops */
1706 }
1707 }
1708
1709 /* Reached the end of the data without finding \n. Let the loop continue to
1710 read some more, if there is room. */
1711
1712 p = pp;
1713 }
1714
1715 GOT_DATA:
1716
1717 /* We have received a line of data. Check it carefully. It must start with the
1718 same two port numbers that we sent, followed by data as defined by the RFC. For
1719 example,
1720
1721 12345 , 25 : USERID : UNIX :root
1722
1723 However, the amount of white space may be different to what we sent. In the
1724 "osname" field there may be several sub-fields, comma separated. The data we
1725 actually want to save follows the third colon. Some systems put leading spaces
1726 in it - we discard those. */
1727
1728 if (sscanf(CS buffer + qlen, "%d , %d%n", &received_sender_port,
1729 &received_interface_port, &n) != 2 ||
1730 received_sender_port != sender_host_port ||
1731 received_interface_port != interface_port)
1732 goto END_OFF;
1733
1734 p = buffer + qlen + n;
1735 while(isspace(*p)) p++;
1736 if (*p++ != ':') goto END_OFF;
1737 while(isspace(*p)) p++;
1738 if (Ustrncmp(p, "USERID", 6) != 0) goto END_OFF;
1739 p += 6;
1740 while(isspace(*p)) p++;
1741 if (*p++ != ':') goto END_OFF;
1742 while (*p != 0 && *p != ':') p++;
1743 if (*p++ == 0) goto END_OFF;
1744 while(isspace(*p)) p++;
1745 if (*p == 0) goto END_OFF;
1746
1747 /* The rest of the line is the data we want. We turn it into printing
1748 characters when we save it, so that it cannot mess up the format of any logging
1749 or Received: lines into which it gets inserted. We keep a maximum of 127
1750 characters. */
1751
1752 sender_ident = string_printing(string_copyn(p, 127));
1753 DEBUG(D_ident) debug_printf("sender_ident = %s\n", sender_ident);
1754
1755 END_OFF:
1756 close(sock);
1757 return;
1758 }
1759
1760
1761
1762
1763 /*************************************************
1764 * Match host to a single host-list item *
1765 *************************************************/
1766
1767 /* This function compares a host (name or address) against a single item
1768 from a host list. The host name gets looked up if it is needed and is not
1769 already known. The function is called from verify_check_this_host() via
1770 match_check_list(), which is why most of its arguments are in a single block.
1771
1772 Arguments:
1773 arg the argument block (see below)
1774 ss the host-list item
1775 valueptr where to pass back looked up data, or NULL
1776 error for error message when returning ERROR
1777
1778 The block contains:
1779 host_name the host name or NULL, implying use sender_host_name and
1780 sender_host_aliases, looking them up if required
1781 host_address the host address
1782 host_ipv4 the IPv4 address taken from an IPv6 one
1783
1784 Returns: OK matched
1785 FAIL did not match
1786 DEFER lookup deferred
1787 ERROR failed to find the host name or IP address
1788 unknown lookup type specified
1789 */
1790
1791 static int
1792 check_host(void *arg, uschar *ss, uschar **valueptr, uschar **error)
1793 {
1794 check_host_block *cb = (check_host_block *)arg;
1795 int maskoffset;
1796 BOOL isquery = FALSE;
1797 uschar *semicolon, *t;
1798 uschar **aliases;
1799
1800 /* Optimize for the special case when the pattern is "*". */
1801
1802 if (*ss == '*' && ss[1] == 0) return OK;
1803
1804 /* If the pattern is empty, it matches only in the case when there is no host -
1805 this can occur in ACL checking for SMTP input using the -bs option. In this
1806 situation, the host address is the empty string. */
1807
1808 if (cb->host_address[0] == 0) return (*ss == 0)? OK : FAIL;
1809 if (*ss == 0) return FAIL;
1810
1811 /* If the pattern is precisely "@" then match against the primary host name;
1812 if it's "@[]" match against the local host's IP addresses. */
1813
1814 if (*ss == '@')
1815 {
1816 if (ss[1] == 0) ss = primary_hostname;
1817 else if (Ustrcmp(ss, "@[]") == 0)
1818 {
1819 ip_address_item *ip;
1820 for (ip = host_find_interfaces(); ip != NULL; ip = ip->next)
1821 if (Ustrcmp(ip->address, cb->host_address) == 0) return OK;
1822 return FAIL;
1823 }
1824 }
1825
1826 /* If the pattern is an IP address, optionally followed by a bitmask count, do
1827 a (possibly masked) comparision with the current IP address. */
1828
1829 if (string_is_ip_address(ss, &maskoffset))
1830 return (host_is_in_net(cb->host_address, ss, maskoffset)? OK : FAIL);
1831
1832 /* If the item is of the form net[n]-lookup;<file|query> then it is a lookup on
1833 a masked IP network, in textual form. The net- stuff really only applies to
1834 single-key lookups where the key is implicit. For query-style lookups the key
1835 is specified in the query. From release 4.30, the use of net- for query style
1836 is no longer needed, but we retain it for backward compatibility. */
1837
1838 if (Ustrncmp(ss, "net", 3) == 0 && (semicolon = Ustrchr(ss, ';')) != NULL)
1839 {
1840 int mlen = 0;
1841 for (t = ss + 3; isdigit(*t); t++) mlen = mlen * 10 + *t - '0';
1842 if (*t++ == '-')
1843 {
1844 int insize;
1845 int search_type;
1846 int incoming[4];
1847 void *handle;
1848 uschar *filename, *key, *result;
1849 uschar buffer[64];
1850
1851 /* If no mask was supplied, set a negative value */
1852
1853 if (mlen == 0 && t == ss+4) mlen = -1;
1854
1855 /* Find the search type */
1856
1857 search_type = search_findtype(t, semicolon - t);
1858
1859 if (search_type < 0) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
1860 search_error_message);
1861
1862 /* Adjust parameters for the type of lookup. For a query-style
1863 lookup, there is no file name, and the "key" is just the query. For
1864 a single-key lookup, the key is the current IP address, masked
1865 appropriately, and reconverted to text form, with the mask appended. */
1866
1867 if (mac_islookup(search_type, lookup_querystyle))
1868 {
1869 filename = NULL;
1870 key = semicolon + 1;
1871 }
1872 else
1873 {
1874 insize = host_aton(cb->host_address, incoming);
1875 host_mask(insize, incoming, mlen);
1876 (void)host_nmtoa(insize, incoming, mlen, buffer);
1877 key = buffer;
1878 filename = semicolon + 1;
1879 }
1880
1881 /* Now do the actual lookup; note that there is no search_close() because
1882 of the caching arrangements. */
1883
1884 handle = search_open(filename, search_type, 0, NULL, NULL);
1885 if (handle == NULL) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
1886 search_error_message);
1887 result = search_find(handle, filename, key, -1, NULL, 0, 0, NULL);
1888 if (valueptr != NULL) *valueptr = result;
1889 return (result != NULL)? OK : search_find_defer? DEFER: FAIL;
1890 }
1891 }
1892
1893 /* The pattern is not an IP address or network reference of any kind. That is,
1894 it is a host name pattern. Check the characters of the pattern to see if they
1895 comprise only letters, digits, full stops, and hyphens (the constituents of
1896 domain names). Allow underscores, as they are all too commonly found. Sigh.
1897 Also, if allow_utf8_domains is set, allow top-bit characters. */
1898
1899 for (t = ss; *t != 0; t++)
1900 if (!isalnum(*t) && *t != '.' && *t != '-' && *t != '_' &&
1901 (!allow_utf8_domains || *t < 128)) break;
1902
1903 /* If the pattern is a complete domain name, with no fancy characters, look up
1904 its IP address and match against that. Note that a multi-homed host will add
1905 items to the chain. */
1906
1907 if (*t == 0)
1908 {
1909 int rc;
1910 host_item h;
1911 h.next = NULL;
1912 h.name = ss;
1913 h.address = NULL;
1914 h.mx = MX_NONE;
1915 rc = host_find_byname(&h, NULL, NULL, FALSE);
1916 if (rc == HOST_FOUND || rc == HOST_FOUND_LOCAL)
1917 {
1918 host_item *hh;
1919 for (hh = &h; hh != NULL; hh = hh->next)
1920 {
1921 if (Ustrcmp(hh->address, (Ustrchr(hh->address, ':') == NULL)?
1922 cb->host_ipv4 : cb->host_address) == 0)
1923 return OK;
1924 }
1925 return FAIL;
1926 }
1927 if (rc == HOST_FIND_AGAIN) return DEFER;
1928 *error = string_sprintf("failed to find IP address for %s", ss);
1929 return ERROR;
1930 }
1931
1932 /* Almost all subsequent comparisons require the host name, and can be done
1933 using the general string matching function. When this function is called for
1934 outgoing hosts, the name is always given explicitly. If it is NULL, it means we
1935 must use sender_host_name and its aliases, looking them up if necessary. */
1936
1937 if (cb->host_name != NULL) /* Explicit host name given */
1938 return match_check_string(cb->host_name, ss, -1, TRUE, TRUE, TRUE,
1939 valueptr);
1940
1941 /* Host name not given; in principle we need the sender host name and its
1942 aliases. However, for query-style lookups, we do not need the name if the
1943 query does not contain $sender_host_name. From release 4.23, a reference to
1944 $sender_host_name causes it to be looked up, so we don't need to do the lookup
1945 on spec. */
1946
1947 if ((semicolon = Ustrchr(ss, ';')) != NULL)
1948 {
1949 uschar *affix;
1950 int partial, affixlen, starflags, id;
1951
1952 *semicolon = 0;
1953 id = search_findtype_partial(ss, &partial, &affix, &affixlen, &starflags);
1954 *semicolon=';';
1955
1956 if (id < 0) /* Unknown lookup type */
1957 {
1958 log_write(0, LOG_MAIN|LOG_PANIC, "%s in host list item \"%s\"",
1959 search_error_message, ss);
1960 return DEFER;
1961 }
1962 isquery = mac_islookup(id, lookup_querystyle);
1963 }
1964
1965 if (isquery)
1966 {
1967 switch(match_check_string(US"", ss, -1, TRUE, TRUE, TRUE, valueptr))
1968 {
1969 case OK: return OK;
1970 case DEFER: return DEFER;
1971 default: return FAIL;
1972 }
1973 }
1974
1975 /* Not a query-style lookup; must ensure the host name is present, and then we
1976 do a check on the name and all its aliases. */
1977
1978 if (sender_host_name == NULL)
1979 {
1980 HDEBUG(D_host_lookup)
1981 debug_printf("sender host name required, to match against %s\n", ss);
1982 if (host_lookup_failed || host_name_lookup() != OK)
1983 {
1984 *error = string_sprintf("failed to find host name for %s",
1985 sender_host_address);;
1986 return ERROR;
1987 }
1988 host_build_sender_fullhost();
1989 }
1990
1991 /* Match on the sender host name, using the general matching function */
1992
1993 switch(match_check_string(sender_host_name, ss, -1, TRUE, TRUE, TRUE,
1994 valueptr))
1995 {
1996 case OK: return OK;
1997 case DEFER: return DEFER;
1998 }
1999
2000 /* If there are aliases, try matching on them. */
2001
2002 aliases = sender_host_aliases;
2003 while (*aliases != NULL)
2004 {
2005 switch(match_check_string(*aliases++, ss, -1, TRUE, TRUE, TRUE, valueptr))
2006 {
2007 case OK: return OK;
2008 case DEFER: return DEFER;
2009 }
2010 }
2011 return FAIL;
2012 }
2013
2014
2015
2016
2017 /*************************************************
2018 * Check a specific host matches a host list *
2019 *************************************************/
2020
2021 /* This function is passed a host list containing items in a number of
2022 different formats and the identity of a host. Its job is to determine whether
2023 the given host is in the set of hosts defined by the list. The host name is
2024 passed as a pointer so that it can be looked up if needed and not already
2025 known. This is commonly the case when called from verify_check_host() to check
2026 an incoming connection. When called from elsewhere the host name should usually
2027 be set.
2028
2029 This function is now just a front end to match_check_list(), which runs common
2030 code for scanning a list. We pass it the check_host() function to perform a
2031 single test.
2032
2033 Arguments:
2034 listptr pointer to the host list
2035 cache_bits pointer to cache for named lists, or NULL
2036 host_name the host name or NULL, implying use sender_host_name and
2037 sender_host_aliases, looking them up if required
2038 host_address the IP address
2039 valueptr if not NULL, data from a lookup is passed back here
2040
2041 Returns: OK if the host is in the defined set
2042 FAIL if the host is not in the defined set,
2043 DEFER if a data lookup deferred (not a host lookup)
2044
2045 If the host name was needed in order to make a comparison, and could not be
2046 determined from the IP address, the result is FAIL unless the item
2047 "+allow_unknown" was met earlier in the list, in which case OK is returned. */
2048
2049 int
2050 verify_check_this_host(uschar **listptr, unsigned int *cache_bits,
2051 uschar *host_name, uschar *host_address, uschar **valueptr)
2052 {
2053 int rc;
2054 unsigned int *local_cache_bits = cache_bits;
2055 uschar *save_host_address = deliver_host_address;
2056 check_host_block cb;
2057 cb.host_name = host_name;
2058 cb.host_address = host_address;
2059
2060 if (valueptr != NULL) *valueptr = NULL;
2061
2062 /* If the host address starts off ::ffff: it is an IPv6 address in
2063 IPv4-compatible mode. Find the IPv4 part for checking against IPv4
2064 addresses. */
2065
2066 cb.host_ipv4 = (Ustrncmp(host_address, "::ffff:", 7) == 0)?
2067 host_address + 7 : host_address;
2068
2069 /* During the running of the check, put the IP address into $host_address. In
2070 the case of calls from the smtp transport, it will already be there. However,
2071 in other calls (e.g. when testing ignore_target_hosts), it won't. Just to be on
2072 the safe side, any existing setting is preserved, though as I write this
2073 (November 2004) I can't see any cases where it is actually needed. */
2074
2075 deliver_host_address = host_address;
2076 rc = match_check_list(
2077 listptr, /* the list */
2078 0, /* separator character */
2079 &hostlist_anchor, /* anchor pointer */
2080 &local_cache_bits, /* cache pointer */
2081 check_host, /* function for testing */
2082 &cb, /* argument for function */
2083 MCL_HOST, /* type of check */
2084 (host_address == sender_host_address)?
2085 US"host" : host_address, /* text for debugging */
2086 valueptr); /* where to pass back data */
2087 deliver_host_address = save_host_address;
2088 return rc;
2089 }
2090
2091
2092
2093
2094 /*************************************************
2095 * Check the remote host matches a list *
2096 *************************************************/
2097
2098 /* This is a front end to verify_check_this_host(), created because checking
2099 the remote host is a common occurrence. With luck, a good compiler will spot
2100 the tail recursion and optimize it. If there's no host address, this is
2101 command-line SMTP input - check against an empty string for the address.
2102
2103 Arguments:
2104 listptr pointer to the host list
2105
2106 Returns: the yield of verify_check_this_host(),
2107 i.e. OK, FAIL, or DEFER
2108 */
2109
2110 int
2111 verify_check_host(uschar **listptr)
2112 {
2113 return verify_check_this_host(listptr, sender_host_cache, NULL,
2114 (sender_host_address == NULL)? US"" : sender_host_address, NULL);
2115 }
2116
2117
2118
2119
2120
2121 /*************************************************
2122 * Invert an IP address for a DNS black list *
2123 *************************************************/
2124
2125 /*
2126 Arguments:
2127 buffer where to put the answer
2128 address the address to invert
2129 */
2130
2131 static void
2132 invert_address(uschar *buffer, uschar *address)
2133 {
2134 int bin[4];
2135 uschar *bptr = buffer;
2136
2137 /* If this is an IPv4 address mapped into IPv6 format, adjust the pointer
2138 to the IPv4 part only. */
2139
2140 if (Ustrncmp(address, "::ffff:", 7) == 0) address += 7;
2141
2142 /* Handle IPv4 address: when HAVE_IPV6 is false, the result of host_aton() is
2143 always 1. */
2144
2145 if (host_aton(address, bin) == 1)
2146 {
2147 int i;
2148 int x = bin[0];
2149 for (i = 0; i < 4; i++)
2150 {
2151 sprintf(CS bptr, "%d.", x & 255);
2152 while (*bptr) bptr++;
2153 x >>= 8;
2154 }
2155 }
2156
2157 /* Handle IPv6 address. Actually, as far as I know, there are no IPv6 addresses
2158 in any DNS black lists, and the format in which they will be looked up is
2159 unknown. This is just a guess. */
2160
2161 #if HAVE_IPV6
2162 else
2163 {
2164 int i, j;
2165 for (j = 3; j >= 0; j--)
2166 {
2167 int x = bin[j];
2168 for (i = 0; i < 8; i++)
2169 {
2170 sprintf(CS bptr, "%x.", x & 15);
2171 while (*bptr) bptr++;
2172 x >>= 4;
2173 }
2174 }
2175 }
2176 #endif
2177 }
2178
2179
2180
2181 /*************************************************
2182 * Check host against DNS black lists *
2183 *************************************************/
2184
2185 /* This function runs checks against a list of DNS black lists, until one
2186 matches. Each item on the list can be of the form
2187
2188 domain=ip-address/key
2189
2190 The domain is the right-most domain that is used for the query, for example,
2191 blackholes.mail-abuse.org. If the IP address is present, there is a match only
2192 if the DNS lookup returns a matching IP address. Several addresses may be
2193 given, comma-separated, for example: x.y.z=127.0.0.1,127.0.0.2.
2194
2195 If no key is given, what is looked up in the domain is the inverted IP address
2196 of the current client host. If a key is given, it is used to construct the
2197 domain for the lookup. For example,
2198
2199 dsn.rfc-ignorant.org/$sender_address_domain
2200
2201 After finding a match in the DNS, the domain is placed in $dnslist_domain, and
2202 then we check for a TXT record for an error message, and if found, save its
2203 value in $dnslist_text. We also cache everything in a tree, to optimize
2204 multiple lookups.
2205
2206 Note: an address for testing RBL is 192.203.178.39
2207 Note: an address for testing DUL is 192.203.178.4
2208 Note: a domain for testing RFCI is example.tld.dsn.rfc-ignorant.org
2209
2210 Arguments:
2211 listptr the domain/address/data list
2212
2213 Returns: OK successful lookup (i.e. the address is on the list), or
2214 lookup deferred after +include_unknown
2215 FAIL name not found, or no data found for the given type, or
2216 lookup deferred after +exclude_unknown (default)
2217 DEFER lookup failure, if +defer_unknown was set
2218 */
2219
2220 int
2221 verify_check_dnsbl(uschar **listptr)
2222 {
2223 int sep = 0;
2224 int defer_return = FAIL;
2225 int old_pool = store_pool;
2226 BOOL invert_result = FALSE;
2227 uschar *list = *listptr;
2228 uschar *domain;
2229 uschar *s;
2230 uschar buffer[1024];
2231 uschar query[256]; /* DNS domain max length */
2232 uschar revadd[128]; /* Long enough for IPv6 address */
2233
2234 /* Indicate that the inverted IP address is not yet set up */
2235
2236 revadd[0] = 0;
2237
2238 /* Loop through all the domains supplied, until something matches */
2239
2240 while ((domain = string_nextinlist(&list, &sep, buffer, sizeof(buffer))) != NULL)
2241 {
2242 BOOL frc;
2243 BOOL bitmask = FALSE;
2244 dns_answer dnsa;
2245 dns_scan dnss;
2246 uschar *iplist;
2247 uschar *key;
2248 tree_node *t;
2249 dnsbl_cache_block *cb;
2250
2251 HDEBUG(D_dnsbl) debug_printf("DNS list check: %s\n", domain);
2252
2253 /* Deal with special values that change the behaviour on defer */
2254
2255 if (domain[0] == '+')
2256 {
2257 if (strcmpic(domain, US"+include_unknown") == 0) defer_return = OK;
2258 else if (strcmpic(domain, US"+exclude_unknown") == 0) defer_return = FAIL;
2259 else if (strcmpic(domain, US"+defer_unknown") == 0) defer_return = DEFER;
2260 else
2261 log_write(0, LOG_MAIN|LOG_PANIC, "unknown item in dnslist (ignored): %s",
2262 domain);
2263 continue;
2264 }
2265
2266 /* See if there's explicit data to be looked up */
2267
2268 key = Ustrchr(domain, '/');
2269 if (key != NULL) *key++ = 0;
2270
2271 /* See if there's a list of addresses supplied after the domain name. This is
2272 introduced by an = or a & character; if preceded by ! we invert the result.
2273 */
2274
2275 iplist = Ustrchr(domain, '=');
2276 if (iplist == NULL)
2277 {
2278 bitmask = TRUE;
2279 iplist = Ustrchr(domain, '&');
2280 }
2281
2282 if (iplist != NULL)
2283 {
2284 if (iplist > domain && iplist[-1] == '!')
2285 {
2286 invert_result = TRUE;
2287 iplist[-1] = 0;
2288 }
2289 *iplist++ = 0;
2290 }
2291
2292 /* Check that what we have left is a sensible domain name. There is no reason
2293 why these domains should in fact use the same syntax as hosts and email
2294 domains, but in practice they seem to. However, there is little point in
2295 actually causing an error here, because that would no doubt hold up incoming
2296 mail. Instead, I'll just log it. */
2297
2298 for (s = domain; *s != 0; s++)
2299 {
2300 if (!isalnum(*s) && *s != '-' && *s != '.')
2301 {
2302 log_write(0, LOG_MAIN, "dnslists domain \"%s\" contains "
2303 "strange characters - is this right?", domain);
2304 break;
2305 }
2306 }
2307
2308 /* Construct the query by adding the domain onto either the sending host
2309 address, or the given key string. */
2310
2311 if (key == NULL)
2312 {
2313 if (sender_host_address == NULL) return FAIL; /* can never match */
2314 if (revadd[0] == 0) invert_address(revadd, sender_host_address);
2315 frc = string_format(query, sizeof(query), "%s%s", revadd, domain);
2316 }
2317 else
2318 {
2319 frc = string_format(query, sizeof(query), "%s.%s", key, domain);
2320 }
2321
2322 if (!frc)
2323 {
2324 log_write(0, LOG_MAIN|LOG_PANIC, "dnslist query is too long "
2325 "(ignored): %s...", query);
2326 continue;
2327 }
2328
2329 /* Look for this query in the cache. */
2330
2331 t = tree_search(dnsbl_cache, query);
2332
2333 /* If not cached from a previous lookup, we must do a DNS lookup, and
2334 cache the result in permanent memory. */
2335
2336 if (t == NULL)
2337 {
2338 store_pool = POOL_PERM;
2339
2340 /* In case this is the first time the DNS resolver is being used. */
2341
2342 dns_init(FALSE, FALSE);
2343
2344 /* Set up a tree entry to cache the lookup */
2345
2346 t = store_get(sizeof(tree_node) + Ustrlen(query));
2347 Ustrcpy(t->name, query);
2348 t->data.ptr = cb = store_get(sizeof(dnsbl_cache_block));
2349 (void)tree_insertnode(&dnsbl_cache, t);
2350
2351 /* Do the DNS loopup . */
2352
2353 HDEBUG(D_dnsbl) debug_printf("new DNS lookup for %s\n", query);
2354 cb->rc = dns_basic_lookup(&dnsa, query, T_A);
2355 cb->text_set = FALSE;
2356 cb->text = NULL;
2357 cb->rhs = NULL;
2358
2359 /* If the lookup succeeded, cache the RHS address. The code allows for
2360 more than one address - this was for complete generality and the possible
2361 use of A6 records. However, A6 records have been reduced to experimental
2362 status (August 2001) and may die out. So they may never get used at all,
2363 let alone in dnsbl records. However, leave the code here, just in case.
2364
2365 Quite apart from one A6 RR generating multiple addresses, there are DNS
2366 lists that return more than one A record, so we must handle multiple
2367 addresses generated in that way as well. */
2368
2369 if (cb->rc == DNS_SUCCEED)
2370 {
2371 dns_record *rr;
2372 dns_address **addrp = &(cb->rhs);
2373 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2374 rr != NULL;
2375 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2376 {
2377 if (rr->type == T_A)
2378 {
2379 dns_address *da = dns_address_from_rr(&dnsa, rr);
2380 if (da != NULL)
2381 {
2382 *addrp = da;
2383 while (da->next != NULL) da = da->next;
2384 addrp = &(da->next);
2385 }
2386 }
2387 }
2388
2389 /* If we didn't find any A records, change the return code. This can
2390 happen when there is a CNAME record but there are no A records for what
2391 it points to. */
2392
2393 if (cb->rhs == NULL) cb->rc = DNS_NODATA;
2394 }
2395
2396 store_pool = old_pool;
2397 }
2398
2399 /* Previous lookup was cached */
2400
2401 else
2402 {
2403 HDEBUG(D_dnsbl) debug_printf("using result of previous DNS lookup\n");
2404 cb = t->data.ptr;
2405 }
2406
2407 /* We now have the result of the DNS lookup, either newly done, or cached
2408 from a previous call. If the lookup succeeded, check against the address
2409 list if there is one. This may be a positive equality list (introduced by
2410 "="), a negative equality list (introduced by "!="), a positive bitmask
2411 list (introduced by "&"), or a negative bitmask list (introduced by "!&").*/
2412
2413 if (cb->rc == DNS_SUCCEED)
2414 {
2415 dns_address *da = NULL;
2416 uschar *addlist = cb->rhs->address;
2417
2418 /* For A and AAAA records, there may be multiple addresses from multiple
2419 records. For A6 records (currently not expected to be used) there may be
2420 multiple addresses from a single record. */
2421
2422 for (da = cb->rhs->next; da != NULL; da = da->next)
2423 addlist = string_sprintf("%s, %s", addlist, da->address);
2424
2425 HDEBUG(D_dnsbl) debug_printf("DNS lookup for %s succeeded (yielding %s)\n",
2426 query, addlist);
2427
2428 /* Address list check; this can be either for equality, or via a bitmask.
2429 In the latter case, all the bits must match. */
2430
2431 if (iplist != NULL)
2432 {
2433 int ipsep = ',';
2434 uschar ip[46];
2435 uschar *ptr = iplist;
2436
2437 while (string_nextinlist(&ptr, &ipsep, ip, sizeof(ip)) != NULL)
2438 {
2439 /* Handle exact matching */
2440 if (!bitmask)
2441 {
2442 for (da = cb->rhs; da != NULL; da = da->next)
2443 {
2444 if (Ustrcmp(CS da->address, ip) == 0) break;
2445 }
2446 }
2447 /* Handle bitmask matching */
2448 else
2449 {
2450 int address[4];
2451 int mask = 0;
2452
2453 /* At present, all known DNS blocking lists use A records, with
2454 IPv4 addresses on the RHS encoding the information they return. I
2455 wonder if this will linger on as the last vestige of IPv4 when IPv6
2456 is ubiquitous? Anyway, for now we use paranoia code to completely
2457 ignore IPv6 addresses. The default mask is 0, which always matches.
2458 We change this only for IPv4 addresses in the list. */
2459
2460 if (host_aton(ip, address) == 1) mask = address[0];
2461
2462 /* Scan the returned addresses, skipping any that are IPv6 */
2463
2464 for (da = cb->rhs; da != NULL; da = da->next)
2465 {
2466 if (host_aton(da->address, address) != 1) continue;
2467 if ((address[0] & mask) == mask) break;
2468 }
2469 }
2470
2471 /* Break out if a match has been found */
2472
2473 if (da != NULL) break;
2474 }
2475
2476 /* If either
2477
2478 (a) No IP address in a positive list matched, or
2479 (b) An IP address in a negative list did match
2480
2481 then behave as if the DNSBL lookup had not succeeded, i.e. the host is
2482 not on the list. */
2483
2484 if (invert_result != (da == NULL))
2485 {
2486 HDEBUG(D_dnsbl)
2487 {
2488 debug_printf("=> but we are not accepting this block class because\n");
2489 debug_printf("=> there was %s match for %c%s\n",
2490 invert_result? "an exclude":"no", bitmask? '&' : '=', iplist);
2491 }
2492 continue; /* With next DNSBL domain */
2493 }
2494 }
2495
2496 /* Either there was no IP list, or the record matched. Look up a TXT record
2497 if it hasn't previously been done. */
2498
2499 if (!cb->text_set)
2500 {
2501 cb->text_set = TRUE;
2502 if (dns_basic_lookup(&dnsa, query, T_TXT) == DNS_SUCCEED)
2503 {
2504 dns_record *rr;
2505 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2506 rr != NULL;
2507 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2508 if (rr->type == T_TXT) break;
2509 if (rr != NULL)
2510 {
2511 int len = (rr->data)[0];
2512 if (len > 511) len = 127;
2513 store_pool = POOL_PERM;
2514 cb->text = string_sprintf("%.*s", len, (const uschar *)(rr->data+1));
2515 store_pool = old_pool;
2516 }
2517 }
2518 }
2519
2520 HDEBUG(D_dnsbl)
2521 {
2522 debug_printf("=> that means %s is listed at %s\n",
2523 (key == NULL)? sender_host_address : key, domain);
2524 }
2525
2526 dnslist_domain = string_copy(domain);
2527 dnslist_value = addlist;
2528 dnslist_text = cb->text;
2529 return OK;
2530 }
2531
2532 /* There was a problem with the DNS lookup */
2533
2534 if (cb->rc != DNS_NOMATCH && cb->rc != DNS_NODATA)
2535 {
2536 log_write(L_dnslist_defer, LOG_MAIN,
2537 "DNS list lookup defer (probably timeout) for %s: %s", query,
2538 (defer_return == OK)? US"assumed in list" :
2539 (defer_return == FAIL)? US"assumed not in list" :
2540 US"returned DEFER");
2541 return defer_return;
2542 }
2543
2544 /* No entry was found in the DNS; continue for next domain */
2545
2546 HDEBUG(D_dnsbl)
2547 {
2548 debug_printf("DNS lookup for %s failed\n", query);
2549 debug_printf("=> that means %s is not listed at %s\n",
2550 (key == NULL)? sender_host_address : key, domain);
2551 }
2552 } /* Continue with next domain */
2553
2554 return FAIL;
2555 }
2556
2557 /* End of verify.c */