Additions to dnsdb lookups: (a) list of domains (b) change output
[exim.git] / src / src / verify.c
CommitLineData
6f0c9a4f 1/* $Cambridge: exim/src/src/verify.c,v 1.6 2004/11/18 11:17:33 ph10 Exp $ */
059ec3d9
PH
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
11caching 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
19typedef 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
29static 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
39Arguments:
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
46Returns: the cache record if a non-expired one exists, else NULL
47*/
48
49static dbdata_callout_cache *
50get_callout_cache_record(open_db *dbm_file, uschar *key, uschar *type,
51 int positive_expire, int negative_expire)
52{
53BOOL negative;
54int length, expire;
55time_t now;
56dbdata_callout_cache *cache_record;
57
58cache_record = dbfn_read_with_length(dbm_file, key, &length);
59
60if (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
67it is a domain record and the postmaster field is negative. */
68
69negative = cache_record->result != ccache_accept ||
70 (type[0] == 'd' && cache_record->postmaster_result == ccache_reject);
71expire = negative? negative_expire : positive_expire;
72now = time(NULL);
73
74if (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
81that doesn't have the postmaster and random timestamps, by looking at the
82length. If so, copy it to a new-style block, replicating the record's
83timestamp. Then check the additional timestamps. (There's no point wasting
84effort if connections are rejected.) */
85
86if (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
103HDEBUG(D_verify) debug_printf("callout cache: found %s record\n", type);
104return 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
114a host list, and a callout has been requested. Callouts are expensive; that is
115why a cache is used to improve the efficiency.
116
117Arguments:
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
4deaf07d
PH
126 callout_overall the overall callout timeout (if < 0 use 4*callout)
127 callout_connect the callout connection timeout (if < 0 use callout)
059ec3d9
PH
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
137Returns: OK/FAIL/DEFER
138*/
139
140static int
141do_callout(address_item *addr, host_item *host_list, transport_feedback *tf,
4deaf07d
PH
142 int callout, int callout_overall, int callout_connect, int options,
143 uschar *se_mailfrom, uschar *pm_mailfrom)
059ec3d9
PH
144{
145BOOL is_recipient = (options & vopt_is_recipient) != 0;
146BOOL callout_no_cache = (options & vopt_callout_no_cache) != 0;
147BOOL callout_random = (options & vopt_callout_random) != 0;
148
149int yield = OK;
150BOOL done = FALSE;
151uschar *address_key;
152uschar *from_address;
153uschar *random_local_part = NULL;
2c7db3f5
PH
154uschar **failure_ptr = is_recipient?
155 &recipient_verify_failure : &sender_verify_failure;
059ec3d9
PH
156open_db dbblock;
157open_db *dbm_file = NULL;
158dbdata_callout_cache new_domain_record;
159dbdata_callout_cache_address new_address_record;
160host_item *host;
161time_t callout_start_time;
162
163new_domain_record.result = ccache_unknown;
164new_domain_record.postmaster_result = ccache_unknown;
165new_domain_record.random_result = ccache_unknown;
166
167memset(&new_address_record, 0, sizeof(new_address_record));
168
169/* For a recipient callout, the key used for the address cache record must
170include the sender address if we are using the real sender in the callout,
171because that may influence the result of the callout. */
172
173address_key = addr->address;
174from_address = US"";
175
176if (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
192empty. */
193
194else
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
202stage, unless caching has been disabled. */
203
204if (callout_no_cache)
205 {
206 HDEBUG(D_verify) debug_printf("callout cache: disabled by no_cache\n");
207 }
208else 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
214actual callout by making use of previously-obtained data. */
215
216if (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;
2c7db3f5 241 *failure_ptr = US"mail";
059ec3d9
PH
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;
2c7db3f5 288 *failure_ptr = US"postmaster";
059ec3d9
PH
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";
2c7db3f5 337 *failure_ptr = US"recipient";
059ec3d9
PH
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
351callout and save the result in the cache for next time, unless no_cache is set,
352or unless we have a previously cached negative random result. If we are to test
353with a random local part, ensure that such a local part is available. If not,
354log the fact, but carry on without randomming. */
355
356if (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
4deaf07d
PH
364/* Default the connect and overall callout timeouts if not set, and record the
365time we are starting so that we can enforce it. */
059ec3d9
PH
366
367if (callout_overall < 0) callout_overall = 4 * callout;
4deaf07d 368if (callout_connect < 0) callout_connect = callout;
059ec3d9
PH
369callout_start_time = time(NULL);
370
371/* Now make connections to the hosts and do real callouts. The list of hosts
372is passed in as an argument. */
373
374for (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
4deaf07d 445 set the error for the last one. Use the callout_connect timeout. */
059ec3d9
PH
446
447 inblock.sock = outblock.sock =
4deaf07d 448 smtp_connect(host, host_af, port, interface, callout_connect, TRUE);
059ec3d9
PH
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 {
2c7db3f5 484 *failure_ptr = US"mail";
059ec3d9
PH
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')
2c7db3f5
PH
559 {
560 *failure_ptr = US"recipient";
059ec3d9 561 new_address_record.result = ccache_reject;
2c7db3f5 562 }
059ec3d9
PH
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 {
2c7db3f5 589 *failure_ptr = US"postmaster";
059ec3d9
PH
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
645will be set OK or FAIL according to the response to the RCPT command.
646Otherwise, we looped through the hosts but couldn't complete the business.
647However, there may be domain-specific information to cache in both cases.
648
649The value of the result field in the new_domain record is ccache_unknown if
650there was an error before or with MAIL FROM:<>, and errno was not zero,
651implying some kind of I/O error. We don't want to write the cache in that case.
652Otherwise the value is ccache_accept or ccache_reject. */
653
654if (!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
674is disabled. */
675
676if (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
697temporary error. If there was only one host, and a response was received, leave
698it alone if supplying details. Otherwise, give a generic response. */
699
700else /* !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
727END_CALLOUT:
728if (dbm_file != NULL) dbfn_close(dbm_file);
729return 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
739failure or defer information is in the original toplevel address. This applies
740when an address is redirected to a single new address, and the failure or
741deferral happens to the child address.
742
743Arguments:
744 vaddr the verify address item
745 addr the final address item
746 yield FAIL or DEFER
747
748Returns: the value of YIELD
749*/
750
751static int
752copy_error(address_item *vaddr, address_item *addr, int yield)
753{
754if (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 }
761return 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
772address testing (-bt), which is indicated by address_test_mode being set.
773
774Arguments:
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
4deaf07d 797 for individual commands
059ec3d9
PH
798 callout_overall if > 0, gives overall timeout for the callout function;
799 if < 0, a default is used (see do_callout())
4deaf07d 800 callout_connect the connection timeout for callouts
059ec3d9
PH
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
809Returns: OK address verified
810 FAIL address failed to verify
811 DEFER can't tell at present
812*/
813
814int
815verify_address(address_item *vaddr, FILE *f, int options, int callout,
4deaf07d
PH
816 int callout_overall, int callout_connect, uschar *se_mailfrom,
817 uschar *pm_mailfrom, BOOL *routed)
059ec3d9
PH
818{
819BOOL allok = TRUE;
820BOOL full_info = (f == NULL)? FALSE : (debug_selector != 0);
821BOOL is_recipient = (options & vopt_is_recipient) != 0;
822BOOL expn = (options & vopt_expn) != 0;
059ec3d9
PH
823int i;
824int yield = OK;
825int verify_type = expn? v_expn :
826 address_test_mode? v_none :
827 is_recipient? v_recipient : v_sender;
828address_item *addr_list;
829address_item *addr_new = NULL;
830address_item *addr_remote = NULL;
831address_item *addr_local = NULL;
832address_item *addr_succeed = NULL;
2c7db3f5
PH
833uschar **failure_ptr = is_recipient?
834 &recipient_verify_failure : &sender_verify_failure;
059ec3d9
PH
835uschar *ko_prefix, *cr;
836uschar *address = vaddr->address;
837uschar *save_sender;
838uschar null_sender[] = { 0 }; /* Ensure writeable memory */
839
2c7db3f5
PH
840/* Clear, just in case */
841
842*failure_ptr = NULL;
843
059ec3d9
PH
844/* Set up a prefix and suffix for error message which allow us to use the same
845output statements both in EXPN mode (where an SMTP response is needed) and when
846debugging with an output file. */
847
848if (expn)
849 {
850 ko_prefix = US"553 ";
851 cr = US"\r";
852 }
853else ko_prefix = cr = US"";
854
855/* Add qualify domain if permitted; otherwise an unqualified address fails. */
856
857if (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);
2c7db3f5 864 *failure_ptr = US"qualify";
059ec3d9
PH
865 return FAIL;
866 }
867 address = rewrite_address_qualify(address, is_recipient);
868 }
869
870DEBUG(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
877may have been set by domains and local part tests during an ACL. */
878
879if (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
893this point, because it may be referred to in the routers. */
894
895if ((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
899to return OK. This rewriting is permitted only for sender addresses; for other
900addresses, such rewriting fails. */
901
902if (address[0] == 0) return OK;
903
904/* Save a copy of the sender address for re-instating if we change it to <>
905while verifying a sender address (a nice bit of self-reference there). */
906
907save_sender = sender_address;
908
909/* Update the address structure with the possibly qualified and rewritten
910address. Set it up as the starting address on the chain of new addresses. */
911
912vaddr->address = address;
913addr_new = vaddr;
914
915/* We need a loop, because an address can generate new addresses. We must also
916cope with generated pipes and files at the top level. (See also the code and
917comment in deliver.c.) However, it is usually the case that the router for
918user's .forward files has its verify flag turned off.
919
920If an address generates more than one child, the loop is used only when
921full_info is set, and this can only be set locally. Remote enquiries just get
922information about the top level address, not anything that it generated. */
923
924while (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;
d8ef3577 1040 host_item *host, *nexthost;
059ec3d9
PH
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
d8ef3577
PH
1044 to find any addresses, the callout will defer. Note that more than
1045 one address may be found for a single host, which will result in
1046 additional host items being inserted into the chain. Hence we must
1047 save the next host first. */
059ec3d9 1048
d8ef3577 1049 for (host = host_list; host != NULL; host = nexthost)
059ec3d9 1050 {
d8ef3577 1051 nexthost = host->next;
059ec3d9
PH
1052 if (tf.gethostbyname || string_is_ip_address(host->name, NULL))
1053 (void)host_find_byname(host, NULL, &canonical_name, TRUE);
1054 else
1055 {
1056 int flags = HOST_FIND_BY_A;
1057 if (tf.qualify_single) flags |= HOST_FIND_QUALIFY_SINGLE;
1058 if (tf.search_parents) flags |= HOST_FIND_SEARCH_PARENTS;
1059 (void)host_find_bydns(host, NULL, flags, NULL, NULL, NULL,
1060 &canonical_name, NULL);
1061 }
1062 }
1063 }
1064 }
1065 }
1066
2c7db3f5
PH
1067 /* Can only do a callout if we have at least one host! If the callout
1068 fails, it will have set ${sender,recipient}_verify_failure. */
059ec3d9
PH
1069
1070 if (host_list != NULL)
1071 {
1072 HDEBUG(D_verify) debug_printf("Attempting full verification using callout\n");
1073 if (host_checking && !host_checking_callout)
1074 {
1075 HDEBUG(D_verify)
1076 debug_printf("... callout omitted by default when host testing\n"
1077 "(Use -bhc if you want the callouts to happen.)\n");
1078 }
1079 else
1080 {
1081 rc = do_callout(addr, host_list, &tf, callout, callout_overall,
4deaf07d 1082 callout_connect, options, se_mailfrom, pm_mailfrom);
059ec3d9
PH
1083 }
1084 }
1085 else
1086 {
1087 HDEBUG(D_verify) debug_printf("Cannot do callout: neither router nor "
1088 "transport provided a host list\n");
1089 }
1090 }
1091 }
2c7db3f5
PH
1092
1093 /* Otherwise, any failure is a routing failure */
1094
1095 else *failure_ptr = US"route";
059ec3d9
PH
1096
1097 /* A router may return REROUTED if it has set up a child address as a result
1098 of a change of domain name (typically from widening). In this case we always
1099 want to continue to verify the new child. */
1100
1101 if (rc == REROUTED) continue;
1102
1103 /* Handle hard failures */
1104
1105 if (rc == FAIL)
1106 {
1107 allok = FALSE;
1108 if (f != NULL)
1109 {
1110 fprintf(f, "%s%s %s", ko_prefix, address,
1111 address_test_mode? "is undeliverable" : "failed to verify");
1112 if (!expn && admin_user)
1113 {
1114 if (addr->basic_errno > 0)
1115 fprintf(f, ": %s", strerror(addr->basic_errno));
1116 if (addr->message != NULL)
1117 fprintf(f, ":\n %s", addr->message);
1118 }
1119 fprintf(f, "%s\n", cr);
1120 }
1121
1122 if (!full_info) return copy_error(vaddr, addr, FAIL);
1123 else yield = FAIL;
1124 }
1125
1126 /* Soft failure */
1127
1128 else if (rc == DEFER)
1129 {
1130 allok = FALSE;
1131 if (f != NULL)
1132 {
1133 fprintf(f, "%s%s cannot be resolved at this time", ko_prefix, address);
1134 if (!expn && admin_user)
1135 {
1136 if (addr->basic_errno > 0)
1137 fprintf(f, ":\n %s", strerror(addr->basic_errno));
1138 if (addr->message != NULL)
1139 fprintf(f, ":\n %s", addr->message);
1140 else if (addr->basic_errno <= 0)
1141 fprintf(f, ":\n unknown error");
1142 }
1143
1144 fprintf(f, "%s\n", cr);
1145 }
1146 if (!full_info) return copy_error(vaddr, addr, DEFER);
1147 else if (yield == OK) yield = DEFER;
1148 }
1149
1150 /* If we are handling EXPN, we do not want to continue to route beyond
1151 the top level. */
1152
1153 else if (expn)
1154 {
1155 uschar *ok_prefix = US"250-";
1156 if (addr_new == NULL)
1157 {
1158 if (addr_local == NULL && addr_remote == NULL)
1159 fprintf(f, "250 mail to <%s> is discarded\r\n", address);
1160 else
1161 fprintf(f, "250 <%s>\r\n", address);
1162 }
1163 else while (addr_new != NULL)
1164 {
1165 address_item *addr2 = addr_new;
1166 addr_new = addr2->next;
1167 if (addr_new == NULL) ok_prefix = US"250 ";
1168 fprintf(f, "%s<%s>\r\n", ok_prefix, addr2->address);
1169 }
1170 return OK;
1171 }
1172
1173 /* Successful routing other than EXPN. */
1174
1175 else
1176 {
1177 /* Handle successful routing when short info wanted. Otherwise continue for
1178 other (generated) addresses. Short info is the operational case. Full info
1179 can be requested only when debug_selector != 0 and a file is supplied.
1180
1181 There is a conflict between the use of aliasing as an alternate email
1182 address, and as a sort of mailing list. If an alias turns the incoming
1183 address into just one address (e.g. J.Caesar->jc44) you may well want to
1184 carry on verifying the generated address to ensure it is valid when
1185 checking incoming mail. If aliasing generates multiple addresses, you
1186 probably don't want to do this. Exim therefore treats the generation of
1187 just a single new address as a special case, and continues on to verify the
1188 generated address. */
1189
1190 if (!full_info && /* Stop if short info wanted AND */
1191 (addr_new == NULL || /* No new address OR */
1192 addr_new->next != NULL || /* More than one new address OR */
1193 testflag(addr_new, af_pfr))) /* New address is pfr */
1194 {
1195 if (f != NULL) fprintf(f, "%s %s\n", address,
1196 address_test_mode? "is deliverable" : "verified");
1197
1198 /* If we have carried on to verify a child address, we want the value
1199 of $address_data to be that of the child */
1200
1201 vaddr->p.address_data = addr->p.address_data;
1202 return OK;
1203 }
1204 }
1205 } /* Loop for generated addresses */
1206
1207/* Display the full results of the successful routing, including any generated
1208addresses. Control gets here only when full_info is set, which requires f not
1209to be NULL, and this occurs only when a top-level verify is called with the
1210debugging switch on.
1211
1212If there are no local and no remote addresses, and there were no pipes, files,
1213or autoreplies, and there were no errors or deferments, the message is to be
1214discarded, usually because of the use of :blackhole: in an alias file. */
1215
1216if (allok && addr_local == NULL && addr_remote == NULL)
1217 fprintf(f, "mail to %s is discarded\n", address);
1218
1219else for (addr_list = addr_local, i = 0; i < 2; addr_list = addr_remote, i++)
1220 {
1221 while (addr_list != NULL)
1222 {
1223 address_item *addr = addr_list;
1224 address_item *p = addr->parent;
1225 addr_list = addr->next;
1226
1227 fprintf(f, "%s", CS addr->address);
1228 while (p != NULL)
1229 {
1230 fprintf(f, "\n <-- %s", p->address);
1231 p = p->parent;
1232 }
1233 fprintf(f, "\n ");
1234
1235 /* Show router, and transport */
1236
1237 fprintf(f, "router = %s, ", addr->router->name);
1238 fprintf(f, "transport = %s\n", (addr->transport == NULL)? US"unset" :
1239 addr->transport->name);
1240
1241 /* Show any hosts that are set up by a router unless the transport
1242 is going to override them; fiddle a bit to get a nice format. */
1243
1244 if (addr->host_list != NULL && addr->transport != NULL &&
1245 !addr->transport->overrides_hosts)
1246 {
1247 host_item *h;
1248 int maxlen = 0;
1249 int maxaddlen = 0;
1250 for (h = addr->host_list; h != NULL; h = h->next)
1251 {
1252 int len = Ustrlen(h->name);
1253 if (len > maxlen) maxlen = len;
1254 len = (h->address != NULL)? Ustrlen(h->address) : 7;
1255 if (len > maxaddlen) maxaddlen = len;
1256 }
1257 for (h = addr->host_list; h != NULL; h = h->next)
1258 {
1259 int len = Ustrlen(h->name);
1260 fprintf(f, " host %s ", h->name);
1261 while (len++ < maxlen) fprintf(f, " ");
1262 if (h->address != NULL)
1263 {
1264 fprintf(f, "[%s] ", h->address);
1265 len = Ustrlen(h->address);
1266 }
1267 else if (!addr->transport->info->local) /* Omit [unknown] for local */
1268 {
1269 fprintf(f, "[unknown] ");
1270 len = 7;
1271 }
1272 else len = -3;
1273 while (len++ < maxaddlen) fprintf(f," ");
1274 if (h->mx >= 0) fprintf(f, "MX=%d", h->mx);
1275 if (h->port != PORT_NONE) fprintf(f, " port=%d", h->port);
1276 if (h->status == hstatus_unusable) fprintf(f, " ** unusable **");
1277 fprintf(f, "\n");
1278 }
1279 }
1280 }
1281 }
1282
2c7db3f5
PH
1283/* Will be DEFER or FAIL if any one address has, only for full_info (which is
1284the -bv or -bt case). */
1285
1286return yield;
059ec3d9
PH
1287}
1288
1289
1290
1291
1292/*************************************************
1293* Check headers for syntax errors *
1294*************************************************/
1295
1296/* This function checks those header lines that contain addresses, and verifies
1297that all the addresses therein are syntactially correct.
1298
1299Arguments:
1300 msgptr where to put an error message
1301
1302Returns: OK
1303 FAIL
1304*/
1305
1306int
1307verify_check_headers(uschar **msgptr)
1308{
1309header_line *h;
1310uschar *colon, *s;
1311
1312for (h = header_list; h != NULL; h = h->next)
1313 {
1314 if (h->type != htype_from &&
1315 h->type != htype_reply_to &&
1316 h->type != htype_sender &&
1317 h->type != htype_to &&
1318 h->type != htype_cc &&
1319 h->type != htype_bcc)
1320 continue;
1321
1322 colon = Ustrchr(h->text, ':');
1323 s = colon + 1;
1324 while (isspace(*s)) s++;
1325
1326 parse_allow_group = TRUE; /* Allow group syntax */
1327
1328 /* Loop for multiple addresses in the header */
1329
1330 while (*s != 0)
1331 {
1332 uschar *ss = parse_find_address_end(s, FALSE);
1333 uschar *recipient, *errmess;
1334 int terminator = *ss;
1335 int start, end, domain;
1336
1337 /* Temporarily terminate the string at this point, and extract the
1338 operative address within. */
1339
1340 *ss = 0;
1341 recipient = parse_extract_address(s,&errmess,&start,&end,&domain,FALSE);
1342 *ss = terminator;
1343
1344 /* Permit an unqualified address only if the message is local, or if the
1345 sending host is configured to be permitted to send them. */
1346
1347 if (recipient != NULL && domain == 0)
1348 {
1349 if (h->type == htype_from || h->type == htype_sender)
1350 {
1351 if (!allow_unqualified_sender) recipient = NULL;
1352 }
1353 else
1354 {
1355 if (!allow_unqualified_recipient) recipient = NULL;
1356 }
1357 if (recipient == NULL) errmess = US"unqualified address not permitted";
1358 }
1359
1360 /* It's an error if no address could be extracted, except for the special
1361 case of an empty address. */
1362
1363 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
1364 {
1365 uschar *verb = US"is";
1366 uschar *t = ss;
1367 int len;
1368
1369 /* Arrange not to include any white space at the end in the
1370 error message. */
1371
1372 while (t > s && isspace(t[-1])) t--;
1373
1374 /* Add the address which failed to the error message, since in a
1375 header with very many addresses it is sometimes hard to spot
1376 which one is at fault. However, limit the amount of address to
1377 quote - cases have been seen where, for example, a missing double
1378 quote in a humungous To: header creates an "address" that is longer
1379 than string_sprintf can handle. */
1380
1381 len = t - s;
1382 if (len > 1024)
1383 {
1384 len = 1024;
1385 verb = US"begins";
1386 }
1387
1388 *msgptr = string_printing(
1389 string_sprintf("%s: failing address in \"%.*s\" header %s: %.*s",
1390 errmess, colon - h->text, h->text, verb, len, s));
1391
1392 return FAIL;
1393 }
1394
1395 /* Advance to the next address */
1396
1397 s = ss + (terminator? 1:0);
1398 while (isspace(*s)) s++;
1399 } /* Next address */
1400 } /* Next header */
1401
1402return OK;
1403}
1404
1405
1406
1407
1408/*************************************************
1409* Find if verified sender *
1410*************************************************/
1411
1412/* Usually, just a single address is verified as the sender of the message.
1413However, Exim can be made to verify other addresses as well (often related in
1414some way), and this is useful in some environments. There may therefore be a
1415chain of such addresses that have previously been tested. This function finds
1416whether a given address is on the chain.
1417
1418Arguments: the address to be verified
1419Returns: pointer to an address item, or NULL
1420*/
1421
1422address_item *
1423verify_checked_sender(uschar *sender)
1424{
1425address_item *addr;
1426for (addr = sender_verified_list; addr != NULL; addr = addr->next)
1427 if (Ustrcmp(sender, addr->address) == 0) break;
1428return addr;
1429}
1430
1431
1432
1433
1434
1435/*************************************************
1436* Get valid header address *
1437*************************************************/
1438
1439/* Scan the originator headers of the message, looking for an address that
1440verifies successfully. RFC 822 says:
1441
1442 o The "Sender" field mailbox should be sent notices of
1443 any problems in transport or delivery of the original
1444 messages. If there is no "Sender" field, then the
1445 "From" field mailbox should be used.
1446
1447 o If the "Reply-To" field exists, then the reply should
1448 go to the addresses indicated in that field and not to
1449 the address(es) indicated in the "From" field.
1450
1451So we check a Sender field if there is one, else a Reply_to field, else a From
1452field. As some strange messages may have more than one of these fields,
1453especially if they are resent- fields, check all of them if there is more than
1454one.
1455
1456Arguments:
1457 user_msgptr points to where to put a user error message
1458 log_msgptr points to where to put a log error message
1459 callout timeout for callout check (passed to verify_address())
1460 callout_overall overall callout timeout (ditto)
4deaf07d 1461 callout_connect connect callout timeout (ditto)
059ec3d9
PH
1462 se_mailfrom mailfrom for verify; NULL => ""
1463 pm_mailfrom sender for pm callout check (passed to verify_address())
1464 options callout options (passed to verify_address())
1465
1466If log_msgptr is set to something without setting user_msgptr, the caller
1467normally uses log_msgptr for both things.
1468
1469Returns: result of the verification attempt: OK, FAIL, or DEFER;
1470 FAIL is given if no appropriate headers are found
1471*/
1472
1473int
1474verify_check_header_address(uschar **user_msgptr, uschar **log_msgptr,
4deaf07d
PH
1475 int callout, int callout_overall, int callout_connect, uschar *se_mailfrom,
1476 uschar *pm_mailfrom, int options)
059ec3d9
PH
1477{
1478static int header_types[] = { htype_sender, htype_reply_to, htype_from };
1479int yield = FAIL;
1480int i;
1481
1482for (i = 0; i < 3; i++)
1483 {
1484 header_line *h;
1485 for (h = header_list; h != NULL; h = h->next)
1486 {
1487 int terminator, new_ok;
1488 uschar *s, *ss, *endname;
1489
1490 if (h->type != header_types[i]) continue;
1491 s = endname = Ustrchr(h->text, ':') + 1;
1492
1493 while (*s != 0)
1494 {
1495 address_item *vaddr;
1496
1497 while (isspace(*s) || *s == ',') s++;
1498 if (*s == 0) break; /* End of header */
1499
1500 ss = parse_find_address_end(s, FALSE);
1501
1502 /* The terminator is a comma or end of header, but there may be white
1503 space preceding it (including newline for the last address). Move back
1504 past any white space so we can check against any cached envelope sender
1505 address verifications. */
1506
1507 while (isspace(ss[-1])) ss--;
1508 terminator = *ss;
1509 *ss = 0;
1510
1511 HDEBUG(D_verify) debug_printf("verifying %.*s header address %s\n",
1512 (int)(endname - h->text), h->text, s);
1513
1514 /* See if we have already verified this address as an envelope sender,
1515 and if so, use the previous answer. */
1516
1517 vaddr = verify_checked_sender(s);
1518
1519 if (vaddr != NULL && /* Previously checked */
1520 (callout <= 0 || /* No callout needed; OR */
1521 vaddr->special_action > 256)) /* Callout was done */
1522 {
1523 new_ok = vaddr->special_action & 255;
1524 HDEBUG(D_verify) debug_printf("previously checked as envelope sender\n");
1525 *ss = terminator; /* Restore shortened string */
1526 }
1527
1528 /* Otherwise we run the verification now. We must restore the shortened
1529 string before running the verification, so the headers are correct, in
1530 case there is any rewriting. */
1531
1532 else
1533 {
1534 int start, end, domain;
1535 uschar *address = parse_extract_address(s, log_msgptr, &start,
1536 &end, &domain, FALSE);
1537
1538 *ss = terminator;
1539
1540 /* If verification failed because of a syntax error, fail this
1541 function, and ensure that the failing address gets added to the error
1542 message. */
1543
1544 if (address == NULL)
1545 {
1546 new_ok = FAIL;
1547 if (*log_msgptr != NULL)
1548 {
1549 while (ss > s && isspace(ss[-1])) ss--;
1550 *log_msgptr = string_sprintf("syntax error in '%.*s' header when "
1551 "scanning for sender: %s in \"%.*s\"",
1552 endname - h->text, h->text, *log_msgptr, ss - s, s);
1553 return FAIL;
1554 }
1555 }
1556
1557 /* Else go ahead with the sender verification. But is isn't *the*
1558 sender of the message, so set vopt_fake_sender to stop sender_address
1559 being replaced after rewriting or qualification. */
1560
1561 else
1562 {
1563 vaddr = deliver_make_addr(address, FALSE);
1564 new_ok = verify_address(vaddr, NULL, options | vopt_fake_sender,
4deaf07d
PH
1565 callout, callout_overall, callout_connect, se_mailfrom,
1566 pm_mailfrom, NULL);
059ec3d9
PH
1567 }
1568 }
1569
1570 /* We now have the result, either newly found, or cached. If we are
1571 giving out error details, set a specific user error. This means that the
1572 last of these will be returned to the user if all three fail. We do not
1573 set a log message - the generic one below will be used. */
1574
1575 if (new_ok != OK && smtp_return_error_details)
1576 {
1577 *user_msgptr = string_sprintf("Rejected after DATA: "
1578 "could not verify \"%.*s\" header address\n%s: %s",
1579 endname - h->text, h->text, vaddr->address, vaddr->message);
1580 }
1581
1582 /* Success or defer */
1583
1584 if (new_ok == OK) return OK;
1585 if (new_ok == DEFER) yield = DEFER;
1586
1587 /* Move on to any more addresses in the header */
1588
1589 s = ss;
1590 }
1591 }
1592 }
1593
1594if (yield == FAIL && *log_msgptr == NULL)
1595 *log_msgptr = US"there is no valid sender in any header line";
1596
1597if (yield == DEFER && *log_msgptr == NULL)
1598 *log_msgptr = US"all attempts to verify a sender in a header line deferred";
1599
1600return yield;
1601}
1602
1603
1604
1605
1606/*************************************************
1607* Get RFC 1413 identification *
1608*************************************************/
1609
1610/* Attempt to get an id from the sending machine via the RFC 1413 protocol. If
1611the timeout is set to zero, then the query is not done. There may also be lists
1612of hosts and nets which are exempt. To guard against malefactors sending
1613non-printing characters which could, for example, disrupt a message's headers,
1614make sure the string consists of printing characters only.
1615
1616Argument:
1617 port the port to connect to; usually this is IDENT_PORT (113), but when
1618 running in the test harness with -bh a different value is used.
1619
1620Returns: nothing
1621
1622Side effect: any received ident value is put in sender_ident (NULL otherwise)
1623*/
1624
1625void
1626verify_get_ident(int port)
1627{
1628int sock, host_af, qlen;
1629int received_sender_port, received_interface_port, n;
1630uschar *p;
1631uschar buffer[2048];
1632
1633/* Default is no ident. Check whether we want to do an ident check for this
1634host. */
1635
1636sender_ident = NULL;
1637if (rfc1413_query_timeout <= 0 || verify_check_host(&rfc1413_hosts) != OK)
1638 return;
1639
1640DEBUG(D_ident) debug_printf("doing ident callback\n");
1641
1642/* Set up a connection to the ident port of the remote host. Bind the local end
1643to the incoming interface address. If the sender host address is an IPv6
1644address, the incoming interface address will also be IPv6. */
1645
1646host_af = (Ustrchr(sender_host_address, ':') == NULL)? AF_INET : AF_INET6;
1647sock = ip_socket(SOCK_STREAM, host_af);
1648if (sock < 0) return;
1649
1650if (ip_bind(sock, host_af, interface_address, 0) < 0)
1651 {
1652 DEBUG(D_ident) debug_printf("bind socket for ident failed: %s\n",
1653 strerror(errno));
1654 goto END_OFF;
1655 }
1656
1657if (ip_connect(sock, host_af, sender_host_address, port, rfc1413_query_timeout)
1658 < 0)
1659 {
1660 if (errno == ETIMEDOUT && (log_extra_selector & LX_ident_timeout) != 0)
1661 {
1662 log_write(0, LOG_MAIN, "ident connection to %s timed out",
1663 sender_host_address);
1664 }
1665 else
1666 {
1667 DEBUG(D_ident) debug_printf("ident connection to %s failed: %s\n",
1668 sender_host_address, strerror(errno));
1669 }
1670 goto END_OFF;
1671 }
1672
1673/* Construct and send the query. */
1674
1675sprintf(CS buffer, "%d , %d\r\n", sender_host_port, interface_port);
1676qlen = Ustrlen(buffer);
1677if (send(sock, buffer, qlen, 0) < 0)
1678 {
1679 DEBUG(D_ident) debug_printf("ident send failed: %s\n", strerror(errno));
1680 goto END_OFF;
1681 }
1682
1683/* Read a response line. We put it into the rest of the buffer, using several
1684recv() calls if necessary. */
1685
1686p = buffer + qlen;
1687
1688for (;;)
1689 {
1690 uschar *pp;
1691 int count;
1692 int size = sizeof(buffer) - (p - buffer);
1693
1694 if (size <= 0) goto END_OFF; /* Buffer filled without seeing \n. */
1695 count = ip_recv(sock, p, size, rfc1413_query_timeout);
1696 if (count <= 0) goto END_OFF; /* Read error or EOF */
1697
1698 /* Scan what we just read, to see if we have reached the terminating \r\n. Be
1699 generous, and accept a plain \n terminator as well. The only illegal
1700 character is 0. */
1701
1702 for (pp = p; pp < p + count; pp++)
1703 {
1704 if (*pp == 0) goto END_OFF; /* Zero octet not allowed */
1705 if (*pp == '\n')
1706 {
1707 if (pp[-1] == '\r') pp--;
1708 *pp = 0;
1709 goto GOT_DATA; /* Break out of both loops */
1710 }
1711 }
1712
1713 /* Reached the end of the data without finding \n. Let the loop continue to
1714 read some more, if there is room. */
1715
1716 p = pp;
1717 }
1718
1719GOT_DATA:
1720
1721/* We have received a line of data. Check it carefully. It must start with the
1722same two port numbers that we sent, followed by data as defined by the RFC. For
1723example,
1724
1725 12345 , 25 : USERID : UNIX :root
1726
1727However, the amount of white space may be different to what we sent. In the
1728"osname" field there may be several sub-fields, comma separated. The data we
1729actually want to save follows the third colon. Some systems put leading spaces
1730in it - we discard those. */
1731
1732if (sscanf(CS buffer + qlen, "%d , %d%n", &received_sender_port,
1733 &received_interface_port, &n) != 2 ||
1734 received_sender_port != sender_host_port ||
1735 received_interface_port != interface_port)
1736 goto END_OFF;
1737
1738p = buffer + qlen + n;
1739while(isspace(*p)) p++;
1740if (*p++ != ':') goto END_OFF;
1741while(isspace(*p)) p++;
1742if (Ustrncmp(p, "USERID", 6) != 0) goto END_OFF;
1743p += 6;
1744while(isspace(*p)) p++;
1745if (*p++ != ':') goto END_OFF;
1746while (*p != 0 && *p != ':') p++;
1747if (*p++ == 0) goto END_OFF;
1748while(isspace(*p)) p++;
1749if (*p == 0) goto END_OFF;
1750
1751/* The rest of the line is the data we want. We turn it into printing
1752characters when we save it, so that it cannot mess up the format of any logging
1753or Received: lines into which it gets inserted. We keep a maximum of 127
1754characters. */
1755
1756sender_ident = string_printing(string_copyn(p, 127));
1757DEBUG(D_ident) debug_printf("sender_ident = %s\n", sender_ident);
1758
1759END_OFF:
1760close(sock);
1761return;
1762}
1763
1764
1765
1766
1767/*************************************************
1768* Match host to a single host-list item *
1769*************************************************/
1770
1771/* This function compares a host (name or address) against a single item
1772from a host list. The host name gets looked up if it is needed and is not
1773already known. The function is called from verify_check_this_host() via
1774match_check_list(), which is why most of its arguments are in a single block.
1775
1776Arguments:
1777 arg the argument block (see below)
1778 ss the host-list item
1779 valueptr where to pass back looked up data, or NULL
1780 error for error message when returning ERROR
1781
1782The block contains:
1783 host_name the host name or NULL, implying use sender_host_name and
1784 sender_host_aliases, looking them up if required
1785 host_address the host address
1786 host_ipv4 the IPv4 address taken from an IPv6 one
1787
1788Returns: OK matched
1789 FAIL did not match
1790 DEFER lookup deferred
1791 ERROR failed to find the host name or IP address
1792 unknown lookup type specified
1793*/
1794
1795static int
1796check_host(void *arg, uschar *ss, uschar **valueptr, uschar **error)
1797{
1798check_host_block *cb = (check_host_block *)arg;
1799int maskoffset;
1800BOOL isquery = FALSE;
1801uschar *semicolon, *t;
1802uschar **aliases;
1803
1804/* Optimize for the special case when the pattern is "*". */
1805
1806if (*ss == '*' && ss[1] == 0) return OK;
1807
1808/* If the pattern is empty, it matches only in the case when there is no host -
1809this can occur in ACL checking for SMTP input using the -bs option. In this
1810situation, the host address is the empty string. */
1811
1812if (cb->host_address[0] == 0) return (*ss == 0)? OK : FAIL;
1813if (*ss == 0) return FAIL;
1814
1815/* If the pattern is precisely "@" then match against the primary host name;
1816if it's "@[]" match against the local host's IP addresses. */
1817
1818if (*ss == '@')
1819 {
1820 if (ss[1] == 0) ss = primary_hostname;
1821 else if (Ustrcmp(ss, "@[]") == 0)
1822 {
1823 ip_address_item *ip;
1824 for (ip = host_find_interfaces(); ip != NULL; ip = ip->next)
1825 if (Ustrcmp(ip->address, cb->host_address) == 0) return OK;
1826 return FAIL;
1827 }
1828 }
1829
1830/* If the pattern is an IP address, optionally followed by a bitmask count, do
1831a (possibly masked) comparision with the current IP address. */
1832
1833if (string_is_ip_address(ss, &maskoffset))
1834 return (host_is_in_net(cb->host_address, ss, maskoffset)? OK : FAIL);
1835
1836/* If the item is of the form net[n]-lookup;<file|query> then it is a lookup on
1837a masked IP network, in textual form. The net- stuff really only applies to
1838single-key lookups where the key is implicit. For query-style lookups the key
1839is specified in the query. From release 4.30, the use of net- for query style
1840is no longer needed, but we retain it for backward compatibility. */
1841
1842if (Ustrncmp(ss, "net", 3) == 0 && (semicolon = Ustrchr(ss, ';')) != NULL)
1843 {
1844 int mlen = 0;
1845 for (t = ss + 3; isdigit(*t); t++) mlen = mlen * 10 + *t - '0';
1846 if (*t++ == '-')
1847 {
1848 int insize;
1849 int search_type;
1850 int incoming[4];
1851 void *handle;
1852 uschar *filename, *key, *result;
1853 uschar buffer[64];
1854
1855 /* If no mask was supplied, set a negative value */
1856
1857 if (mlen == 0 && t == ss+4) mlen = -1;
1858
1859 /* Find the search type */
1860
1861 search_type = search_findtype(t, semicolon - t);
1862
1863 if (search_type < 0) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
1864 search_error_message);
1865
1866 /* Adjust parameters for the type of lookup. For a query-style
1867 lookup, there is no file name, and the "key" is just the query. For
1868 a single-key lookup, the key is the current IP address, masked
6f0c9a4f
PH
1869 appropriately, and reconverted to text form, with the mask appended.
1870 For IPv6 addresses, specify dot separators instead of colons. */
059ec3d9
PH
1871
1872 if (mac_islookup(search_type, lookup_querystyle))
1873 {
1874 filename = NULL;
1875 key = semicolon + 1;
1876 }
1877 else
1878 {
1879 insize = host_aton(cb->host_address, incoming);
1880 host_mask(insize, incoming, mlen);
6f0c9a4f 1881 (void)host_nmtoa(insize, incoming, mlen, buffer, '.');
059ec3d9
PH
1882 key = buffer;
1883 filename = semicolon + 1;
1884 }
1885
1886 /* Now do the actual lookup; note that there is no search_close() because
1887 of the caching arrangements. */
1888
1889 handle = search_open(filename, search_type, 0, NULL, NULL);
1890 if (handle == NULL) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
1891 search_error_message);
1892 result = search_find(handle, filename, key, -1, NULL, 0, 0, NULL);
1893 if (valueptr != NULL) *valueptr = result;
1894 return (result != NULL)? OK : search_find_defer? DEFER: FAIL;
1895 }
1896 }
1897
1898/* The pattern is not an IP address or network reference of any kind. That is,
1899it is a host name pattern. Check the characters of the pattern to see if they
1900comprise only letters, digits, full stops, and hyphens (the constituents of
1901domain names). Allow underscores, as they are all too commonly found. Sigh.
1902Also, if allow_utf8_domains is set, allow top-bit characters. */
1903
1904for (t = ss; *t != 0; t++)
1905 if (!isalnum(*t) && *t != '.' && *t != '-' && *t != '_' &&
1906 (!allow_utf8_domains || *t < 128)) break;
1907
1908/* If the pattern is a complete domain name, with no fancy characters, look up
1909its IP address and match against that. Note that a multi-homed host will add
1910items to the chain. */
1911
1912if (*t == 0)
1913 {
1914 int rc;
1915 host_item h;
1916 h.next = NULL;
1917 h.name = ss;
1918 h.address = NULL;
1919 h.mx = MX_NONE;
1920 rc = host_find_byname(&h, NULL, NULL, FALSE);
1921 if (rc == HOST_FOUND || rc == HOST_FOUND_LOCAL)
1922 {
1923 host_item *hh;
1924 for (hh = &h; hh != NULL; hh = hh->next)
1925 {
1926 if (Ustrcmp(hh->address, (Ustrchr(hh->address, ':') == NULL)?
1927 cb->host_ipv4 : cb->host_address) == 0)
1928 return OK;
1929 }
1930 return FAIL;
1931 }
1932 if (rc == HOST_FIND_AGAIN) return DEFER;
1933 *error = string_sprintf("failed to find IP address for %s", ss);
1934 return ERROR;
1935 }
1936
1937/* Almost all subsequent comparisons require the host name, and can be done
1938using the general string matching function. When this function is called for
1939outgoing hosts, the name is always given explicitly. If it is NULL, it means we
1940must use sender_host_name and its aliases, looking them up if necessary. */
1941
1942if (cb->host_name != NULL) /* Explicit host name given */
1943 return match_check_string(cb->host_name, ss, -1, TRUE, TRUE, TRUE,
1944 valueptr);
1945
1946/* Host name not given; in principle we need the sender host name and its
1947aliases. However, for query-style lookups, we do not need the name if the
1948query does not contain $sender_host_name. From release 4.23, a reference to
1949$sender_host_name causes it to be looked up, so we don't need to do the lookup
1950on spec. */
1951
1952if ((semicolon = Ustrchr(ss, ';')) != NULL)
1953 {
1954 uschar *affix;
1955 int partial, affixlen, starflags, id;
1956
1957 *semicolon = 0;
1958 id = search_findtype_partial(ss, &partial, &affix, &affixlen, &starflags);
1959 *semicolon=';';
1960
1961 if (id < 0) /* Unknown lookup type */
1962 {
1963 log_write(0, LOG_MAIN|LOG_PANIC, "%s in host list item \"%s\"",
1964 search_error_message, ss);
1965 return DEFER;
1966 }
1967 isquery = mac_islookup(id, lookup_querystyle);
1968 }
1969
1970if (isquery)
1971 {
1972 switch(match_check_string(US"", ss, -1, TRUE, TRUE, TRUE, valueptr))
1973 {
1974 case OK: return OK;
1975 case DEFER: return DEFER;
1976 default: return FAIL;
1977 }
1978 }
1979
1980/* Not a query-style lookup; must ensure the host name is present, and then we
1981do a check on the name and all its aliases. */
1982
1983if (sender_host_name == NULL)
1984 {
1985 HDEBUG(D_host_lookup)
1986 debug_printf("sender host name required, to match against %s\n", ss);
1987 if (host_lookup_failed || host_name_lookup() != OK)
1988 {
1989 *error = string_sprintf("failed to find host name for %s",
1990 sender_host_address);;
1991 return ERROR;
1992 }
1993 host_build_sender_fullhost();
1994 }
1995
1996/* Match on the sender host name, using the general matching function */
1997
1998switch(match_check_string(sender_host_name, ss, -1, TRUE, TRUE, TRUE,
1999 valueptr))
2000 {
2001 case OK: return OK;
2002 case DEFER: return DEFER;
2003 }
2004
2005/* If there are aliases, try matching on them. */
2006
2007aliases = sender_host_aliases;
2008while (*aliases != NULL)
2009 {
2010 switch(match_check_string(*aliases++, ss, -1, TRUE, TRUE, TRUE, valueptr))
2011 {
2012 case OK: return OK;
2013 case DEFER: return DEFER;
2014 }
2015 }
2016return FAIL;
2017}
2018
2019
2020
2021
2022/*************************************************
2023* Check a specific host matches a host list *
2024*************************************************/
2025
2026/* This function is passed a host list containing items in a number of
2027different formats and the identity of a host. Its job is to determine whether
2028the given host is in the set of hosts defined by the list. The host name is
2029passed as a pointer so that it can be looked up if needed and not already
2030known. This is commonly the case when called from verify_check_host() to check
2031an incoming connection. When called from elsewhere the host name should usually
2032be set.
2033
2034This function is now just a front end to match_check_list(), which runs common
2035code for scanning a list. We pass it the check_host() function to perform a
2036single test.
2037
2038Arguments:
2039 listptr pointer to the host list
2040 cache_bits pointer to cache for named lists, or NULL
2041 host_name the host name or NULL, implying use sender_host_name and
2042 sender_host_aliases, looking them up if required
2043 host_address the IP address
2044 valueptr if not NULL, data from a lookup is passed back here
2045
2046Returns: OK if the host is in the defined set
2047 FAIL if the host is not in the defined set,
2048 DEFER if a data lookup deferred (not a host lookup)
2049
2050If the host name was needed in order to make a comparison, and could not be
2051determined from the IP address, the result is FAIL unless the item
2052"+allow_unknown" was met earlier in the list, in which case OK is returned. */
2053
2054int
2055verify_check_this_host(uschar **listptr, unsigned int *cache_bits,
2056 uschar *host_name, uschar *host_address, uschar **valueptr)
2057{
d4eb88df 2058int rc;
059ec3d9 2059unsigned int *local_cache_bits = cache_bits;
d4eb88df 2060uschar *save_host_address = deliver_host_address;
059ec3d9
PH
2061check_host_block cb;
2062cb.host_name = host_name;
2063cb.host_address = host_address;
2064
2065if (valueptr != NULL) *valueptr = NULL;
2066
2067/* If the host address starts off ::ffff: it is an IPv6 address in
2068IPv4-compatible mode. Find the IPv4 part for checking against IPv4
2069addresses. */
2070
2071cb.host_ipv4 = (Ustrncmp(host_address, "::ffff:", 7) == 0)?
2072 host_address + 7 : host_address;
2073
d4eb88df
PH
2074/* During the running of the check, put the IP address into $host_address. In
2075the case of calls from the smtp transport, it will already be there. However,
2076in other calls (e.g. when testing ignore_target_hosts), it won't. Just to be on
2077the safe side, any existing setting is preserved, though as I write this
2078(November 2004) I can't see any cases where it is actually needed. */
2079
2080deliver_host_address = host_address;
2081rc = match_check_list(
2082 listptr, /* the list */
2083 0, /* separator character */
2084 &hostlist_anchor, /* anchor pointer */
2085 &local_cache_bits, /* cache pointer */
2086 check_host, /* function for testing */
2087 &cb, /* argument for function */
2088 MCL_HOST, /* type of check */
2089 (host_address == sender_host_address)?
2090 US"host" : host_address, /* text for debugging */
2091 valueptr); /* where to pass back data */
2092deliver_host_address = save_host_address;
2093return rc;
059ec3d9
PH
2094}
2095
2096
2097
2098
2099/*************************************************
2100* Check the remote host matches a list *
2101*************************************************/
2102
2103/* This is a front end to verify_check_this_host(), created because checking
2104the remote host is a common occurrence. With luck, a good compiler will spot
2105the tail recursion and optimize it. If there's no host address, this is
2106command-line SMTP input - check against an empty string for the address.
2107
2108Arguments:
2109 listptr pointer to the host list
2110
2111Returns: the yield of verify_check_this_host(),
2112 i.e. OK, FAIL, or DEFER
2113*/
2114
2115int
2116verify_check_host(uschar **listptr)
2117{
2118return verify_check_this_host(listptr, sender_host_cache, NULL,
2119 (sender_host_address == NULL)? US"" : sender_host_address, NULL);
2120}
2121
2122
2123
2124
2125
2126/*************************************************
2127* Invert an IP address for a DNS black list *
2128*************************************************/
2129
2130/*
2131Arguments:
2132 buffer where to put the answer
2133 address the address to invert
2134*/
2135
2136static void
2137invert_address(uschar *buffer, uschar *address)
2138{
2139int bin[4];
2140uschar *bptr = buffer;
2141
2142/* If this is an IPv4 address mapped into IPv6 format, adjust the pointer
2143to the IPv4 part only. */
2144
2145if (Ustrncmp(address, "::ffff:", 7) == 0) address += 7;
2146
2147/* Handle IPv4 address: when HAVE_IPV6 is false, the result of host_aton() is
2148always 1. */
2149
2150if (host_aton(address, bin) == 1)
2151 {
2152 int i;
2153 int x = bin[0];
2154 for (i = 0; i < 4; i++)
2155 {
2156 sprintf(CS bptr, "%d.", x & 255);
2157 while (*bptr) bptr++;
2158 x >>= 8;
2159 }
2160 }
2161
2162/* Handle IPv6 address. Actually, as far as I know, there are no IPv6 addresses
2163in any DNS black lists, and the format in which they will be looked up is
2164unknown. This is just a guess. */
2165
2166#if HAVE_IPV6
2167else
2168 {
2169 int i, j;
2170 for (j = 3; j >= 0; j--)
2171 {
2172 int x = bin[j];
2173 for (i = 0; i < 8; i++)
2174 {
2175 sprintf(CS bptr, "%x.", x & 15);
2176 while (*bptr) bptr++;
2177 x >>= 4;
2178 }
2179 }
2180 }
2181#endif
2182}
2183
2184
2185
2186/*************************************************
2187* Check host against DNS black lists *
2188*************************************************/
2189
2190/* This function runs checks against a list of DNS black lists, until one
2191matches. Each item on the list can be of the form
2192
2193 domain=ip-address/key
2194
2195The domain is the right-most domain that is used for the query, for example,
2196blackholes.mail-abuse.org. If the IP address is present, there is a match only
2197if the DNS lookup returns a matching IP address. Several addresses may be
2198given, comma-separated, for example: x.y.z=127.0.0.1,127.0.0.2.
2199
2200If no key is given, what is looked up in the domain is the inverted IP address
2201of the current client host. If a key is given, it is used to construct the
2202domain for the lookup. For example,
2203
2204 dsn.rfc-ignorant.org/$sender_address_domain
2205
2206After finding a match in the DNS, the domain is placed in $dnslist_domain, and
2207then we check for a TXT record for an error message, and if found, save its
2208value in $dnslist_text. We also cache everything in a tree, to optimize
2209multiple lookups.
2210
2211Note: an address for testing RBL is 192.203.178.39
2212Note: an address for testing DUL is 192.203.178.4
2213Note: a domain for testing RFCI is example.tld.dsn.rfc-ignorant.org
2214
2215Arguments:
2216 listptr the domain/address/data list
2217
2218Returns: OK successful lookup (i.e. the address is on the list), or
2219 lookup deferred after +include_unknown
2220 FAIL name not found, or no data found for the given type, or
2221 lookup deferred after +exclude_unknown (default)
2222 DEFER lookup failure, if +defer_unknown was set
2223*/
2224
2225int
2226verify_check_dnsbl(uschar **listptr)
2227{
2228int sep = 0;
2229int defer_return = FAIL;
2230int old_pool = store_pool;
2231BOOL invert_result = FALSE;
2232uschar *list = *listptr;
2233uschar *domain;
2234uschar *s;
2235uschar buffer[1024];
2236uschar query[256]; /* DNS domain max length */
2237uschar revadd[128]; /* Long enough for IPv6 address */
2238
2239/* Indicate that the inverted IP address is not yet set up */
2240
2241revadd[0] = 0;
2242
2243/* Loop through all the domains supplied, until something matches */
2244
2245while ((domain = string_nextinlist(&list, &sep, buffer, sizeof(buffer))) != NULL)
2246 {
2247 BOOL frc;
2248 BOOL bitmask = FALSE;
2249 dns_answer dnsa;
2250 dns_scan dnss;
2251 uschar *iplist;
2252 uschar *key;
2253 tree_node *t;
2254 dnsbl_cache_block *cb;
2255
2256 HDEBUG(D_dnsbl) debug_printf("DNS list check: %s\n", domain);
2257
2258 /* Deal with special values that change the behaviour on defer */
2259
2260 if (domain[0] == '+')
2261 {
2262 if (strcmpic(domain, US"+include_unknown") == 0) defer_return = OK;
2263 else if (strcmpic(domain, US"+exclude_unknown") == 0) defer_return = FAIL;
2264 else if (strcmpic(domain, US"+defer_unknown") == 0) defer_return = DEFER;
2265 else
2266 log_write(0, LOG_MAIN|LOG_PANIC, "unknown item in dnslist (ignored): %s",
2267 domain);
2268 continue;
2269 }
2270
2271 /* See if there's explicit data to be looked up */
2272
2273 key = Ustrchr(domain, '/');
2274 if (key != NULL) *key++ = 0;
2275
2276 /* See if there's a list of addresses supplied after the domain name. This is
2277 introduced by an = or a & character; if preceded by ! we invert the result.
2278 */
2279
2280 iplist = Ustrchr(domain, '=');
2281 if (iplist == NULL)
2282 {
2283 bitmask = TRUE;
2284 iplist = Ustrchr(domain, '&');
2285 }
2286
2287 if (iplist != NULL)
2288 {
2289 if (iplist > domain && iplist[-1] == '!')
2290 {
2291 invert_result = TRUE;
2292 iplist[-1] = 0;
2293 }
2294 *iplist++ = 0;
2295 }
2296
2297 /* Check that what we have left is a sensible domain name. There is no reason
2298 why these domains should in fact use the same syntax as hosts and email
2299 domains, but in practice they seem to. However, there is little point in
2300 actually causing an error here, because that would no doubt hold up incoming
2301 mail. Instead, I'll just log it. */
2302
2303 for (s = domain; *s != 0; s++)
2304 {
2305 if (!isalnum(*s) && *s != '-' && *s != '.')
2306 {
2307 log_write(0, LOG_MAIN, "dnslists domain \"%s\" contains "
2308 "strange characters - is this right?", domain);
2309 break;
2310 }
2311 }
2312
2313 /* Construct the query by adding the domain onto either the sending host
2314 address, or the given key string. */
2315
2316 if (key == NULL)
2317 {
2318 if (sender_host_address == NULL) return FAIL; /* can never match */
2319 if (revadd[0] == 0) invert_address(revadd, sender_host_address);
2320 frc = string_format(query, sizeof(query), "%s%s", revadd, domain);
2321 }
2322 else
2323 {
2324 frc = string_format(query, sizeof(query), "%s.%s", key, domain);
2325 }
2326
2327 if (!frc)
2328 {
2329 log_write(0, LOG_MAIN|LOG_PANIC, "dnslist query is too long "
2330 "(ignored): %s...", query);
2331 continue;
2332 }
2333
2334 /* Look for this query in the cache. */
2335
2336 t = tree_search(dnsbl_cache, query);
2337
2338 /* If not cached from a previous lookup, we must do a DNS lookup, and
2339 cache the result in permanent memory. */
2340
2341 if (t == NULL)
2342 {
2343 store_pool = POOL_PERM;
2344
2345 /* In case this is the first time the DNS resolver is being used. */
2346
2347 dns_init(FALSE, FALSE);
2348
2349 /* Set up a tree entry to cache the lookup */
2350
2351 t = store_get(sizeof(tree_node) + Ustrlen(query));
2352 Ustrcpy(t->name, query);
2353 t->data.ptr = cb = store_get(sizeof(dnsbl_cache_block));
2354 (void)tree_insertnode(&dnsbl_cache, t);
2355
2356 /* Do the DNS loopup . */
2357
2358 HDEBUG(D_dnsbl) debug_printf("new DNS lookup for %s\n", query);
2359 cb->rc = dns_basic_lookup(&dnsa, query, T_A);
2360 cb->text_set = FALSE;
2361 cb->text = NULL;
2362 cb->rhs = NULL;
2363
2364 /* If the lookup succeeded, cache the RHS address. The code allows for
2365 more than one address - this was for complete generality and the possible
2366 use of A6 records. However, A6 records have been reduced to experimental
2367 status (August 2001) and may die out. So they may never get used at all,
2368 let alone in dnsbl records. However, leave the code here, just in case.
2369
2370 Quite apart from one A6 RR generating multiple addresses, there are DNS
2371 lists that return more than one A record, so we must handle multiple
2372 addresses generated in that way as well. */
2373
2374 if (cb->rc == DNS_SUCCEED)
2375 {
2376 dns_record *rr;
2377 dns_address **addrp = &(cb->rhs);
2378 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2379 rr != NULL;
2380 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2381 {
2382 if (rr->type == T_A)
2383 {
2384 dns_address *da = dns_address_from_rr(&dnsa, rr);
2385 if (da != NULL)
2386 {
2387 *addrp = da;
2388 while (da->next != NULL) da = da->next;
2389 addrp = &(da->next);
2390 }
2391 }
2392 }
2393
2394 /* If we didn't find any A records, change the return code. This can
2395 happen when there is a CNAME record but there are no A records for what
2396 it points to. */
2397
2398 if (cb->rhs == NULL) cb->rc = DNS_NODATA;
2399 }
2400
2401 store_pool = old_pool;
2402 }
2403
2404 /* Previous lookup was cached */
2405
2406 else
2407 {
2408 HDEBUG(D_dnsbl) debug_printf("using result of previous DNS lookup\n");
2409 cb = t->data.ptr;
2410 }
2411
2412 /* We now have the result of the DNS lookup, either newly done, or cached
2413 from a previous call. If the lookup succeeded, check against the address
2414 list if there is one. This may be a positive equality list (introduced by
2415 "="), a negative equality list (introduced by "!="), a positive bitmask
2416 list (introduced by "&"), or a negative bitmask list (introduced by "!&").*/
2417
2418 if (cb->rc == DNS_SUCCEED)
2419 {
2420 dns_address *da = NULL;
2421 uschar *addlist = cb->rhs->address;
2422
2423 /* For A and AAAA records, there may be multiple addresses from multiple
2424 records. For A6 records (currently not expected to be used) there may be
2425 multiple addresses from a single record. */
2426
2427 for (da = cb->rhs->next; da != NULL; da = da->next)
2428 addlist = string_sprintf("%s, %s", addlist, da->address);
2429
2430 HDEBUG(D_dnsbl) debug_printf("DNS lookup for %s succeeded (yielding %s)\n",
2431 query, addlist);
2432
2433 /* Address list check; this can be either for equality, or via a bitmask.
2434 In the latter case, all the bits must match. */
2435
2436 if (iplist != NULL)
2437 {
2438 int ipsep = ',';
2439 uschar ip[46];
2440 uschar *ptr = iplist;
2441
2442 while (string_nextinlist(&ptr, &ipsep, ip, sizeof(ip)) != NULL)
2443 {
2444 /* Handle exact matching */
2445 if (!bitmask)
2446 {
2447 for (da = cb->rhs; da != NULL; da = da->next)
2448 {
2449 if (Ustrcmp(CS da->address, ip) == 0) break;
2450 }
2451 }
2452 /* Handle bitmask matching */
2453 else
2454 {
2455 int address[4];
2456 int mask = 0;
2457
2458 /* At present, all known DNS blocking lists use A records, with
2459 IPv4 addresses on the RHS encoding the information they return. I
2460 wonder if this will linger on as the last vestige of IPv4 when IPv6
2461 is ubiquitous? Anyway, for now we use paranoia code to completely
2462 ignore IPv6 addresses. The default mask is 0, which always matches.
2463 We change this only for IPv4 addresses in the list. */
2464
2465 if (host_aton(ip, address) == 1) mask = address[0];
2466
2467 /* Scan the returned addresses, skipping any that are IPv6 */
2468
2469 for (da = cb->rhs; da != NULL; da = da->next)
2470 {
2471 if (host_aton(da->address, address) != 1) continue;
2472 if ((address[0] & mask) == mask) break;
2473 }
2474 }
2475
2476 /* Break out if a match has been found */
2477
2478 if (da != NULL) break;
2479 }
2480
2481 /* If either
2482
2483 (a) No IP address in a positive list matched, or
2484 (b) An IP address in a negative list did match
2485
2486 then behave as if the DNSBL lookup had not succeeded, i.e. the host is
2487 not on the list. */
2488
2489 if (invert_result != (da == NULL))
2490 {
2491 HDEBUG(D_dnsbl)
2492 {
2493 debug_printf("=> but we are not accepting this block class because\n");
2494 debug_printf("=> there was %s match for %c%s\n",
2495 invert_result? "an exclude":"no", bitmask? '&' : '=', iplist);
2496 }
2497 continue; /* With next DNSBL domain */
2498 }
2499 }
2500
2501 /* Either there was no IP list, or the record matched. Look up a TXT record
2502 if it hasn't previously been done. */
2503
2504 if (!cb->text_set)
2505 {
2506 cb->text_set = TRUE;
2507 if (dns_basic_lookup(&dnsa, query, T_TXT) == DNS_SUCCEED)
2508 {
2509 dns_record *rr;
2510 for (rr = dns_next_rr(&dnsa, &dnss, RESET_ANSWERS);
2511 rr != NULL;
2512 rr = dns_next_rr(&dnsa, &dnss, RESET_NEXT))
2513 if (rr->type == T_TXT) break;
2514 if (rr != NULL)
2515 {
2516 int len = (rr->data)[0];
2517 if (len > 511) len = 127;
2518 store_pool = POOL_PERM;
2519 cb->text = string_sprintf("%.*s", len, (const uschar *)(rr->data+1));
2520 store_pool = old_pool;
2521 }
2522 }
2523 }
2524
2525 HDEBUG(D_dnsbl)
2526 {
2527 debug_printf("=> that means %s is listed at %s\n",
2528 (key == NULL)? sender_host_address : key, domain);
2529 }
2530
2531 dnslist_domain = string_copy(domain);
2532 dnslist_value = addlist;
2533 dnslist_text = cb->text;
2534 return OK;
2535 }
2536
2537 /* There was a problem with the DNS lookup */
2538
2539 if (cb->rc != DNS_NOMATCH && cb->rc != DNS_NODATA)
2540 {
2541 log_write(L_dnslist_defer, LOG_MAIN,
2542 "DNS list lookup defer (probably timeout) for %s: %s", query,
2543 (defer_return == OK)? US"assumed in list" :
2544 (defer_return == FAIL)? US"assumed not in list" :
2545 US"returned DEFER");
2546 return defer_return;
2547 }
2548
2549 /* No entry was found in the DNS; continue for next domain */
2550
2551 HDEBUG(D_dnsbl)
2552 {
2553 debug_printf("DNS lookup for %s failed\n", query);
2554 debug_printf("=> that means %s is not listed at %s\n",
2555 (key == NULL)? sender_host_address : key, domain);
2556 }
2557 } /* Continue with next domain */
2558
2559return FAIL;
2560}
2561
2562/* End of verify.c */