1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
5 /* Copyright (c) University of Cambridge 1995 - 2018 */
6 /* Copyright (c) The Exim Maintainers 2020 */
7 /* See the file NOTICE for conditions of use and distribution. */
9 /* Functions for matching strings */
15 /* Argument block for the check_string() function. This is used for general
16 strings, domains, and local parts. */
18 typedef struct check_string_block {
19 const uschar *origsubject; /* caseful; keep these two first, in */
20 const uschar *subject; /* step with the block below */
28 /* Argument block for the check_address() function. This is used for whole
31 typedef struct check_address_block {
32 const uschar *origaddress; /* caseful; keep these two first, in */
33 uschar *address; /* step with the block above */
36 } check_address_block;
40 /*************************************************
41 * Generalized string match *
42 *************************************************/
44 /* This function does a single match of a subject against a pattern, and
45 optionally sets up the numeric variables according to what it matched. It is
46 called from match_isinlist() via match_check_list() when scanning a list, and
47 from match_check_string() when testing just a single item. The subject and
48 options arguments are passed in a check_string_block so as to make it easier to
49 pass them through match_check_list.
51 The possible types of pattern are:
53 . regular expression - starts with ^
54 . tail match - starts with *
55 . lookup - starts with search type
56 . if at_is_special is set in the argument block:
57 @ matches the primary host name
58 @[] matches a local IP address in brackets
59 @mx_any matches any domain with an MX to the local host
60 @mx_primary matches any domain with a primary MX to the local host
61 @mx_secondary matches any domain with a secondary MX to the local host
62 . literal - anything else
64 Any of the @mx_xxx options can be followed by "/ignore=<list>" where <list> is
65 a list of IP addresses that are to be ignored (typically 127.0.0.1).
68 arg check_string_block pointer - see below
69 pattern the pattern to be matched
70 valueptr if not NULL, and a lookup is done, return the result here
71 instead of discarding it; else set it to point to NULL
72 error for error messages (not used in this function; it never
75 Contents of the argument block:
76 origsubject the subject in its original casing
77 subject the subject string to be checked, lowercased if caseless
78 expand_setup if < 0, don't set up any numeric expansion variables;
79 if = 0, set $0 to whole subject, and either
80 $1 to what matches * or
81 $1, $2, ... to r.e. bracketed items
82 if > 0, don't set $0, but do set either
83 $n to what matches *, or
84 $n, $n+1, ... to r.e. bracketed items
85 (where n = expand_setup)
86 use_partial if FALSE, override any partial- search types
87 caseless TRUE for caseless matching where possible
88 at_is_special enable special handling of items starting with @
90 Returns: OK if matched
92 DEFER if lookup deferred
96 check_string(void *arg, const uschar *pattern, const uschar **valueptr, uschar **error)
98 const check_string_block *cb = arg;
99 int search_type, partial, affixlen, starflags;
100 int expand_setup = cb->expand_setup;
101 const uschar * affix, * opts;
103 uschar *filename = NULL;
104 uschar *keyquery, *result, *semicolon;
107 error = error; /* Keep clever compilers from complaining */
109 if (valueptr) *valueptr = NULL;
111 /* For regular expressions, use cb->origsubject rather than cb->subject so that
112 it works if the pattern uses (?-i) to turn off case-independence, overriding
115 s = string_copy(pattern[0] == '^' ? cb->origsubject : cb->subject);
117 /* If required to set up $0, initialize the data but don't turn on by setting
118 expand_nmax until the match is assured. */
121 if (expand_setup == 0)
123 expand_nstring[0] = s; /* $0 (might be) the matched subject in full */
124 expand_nlength[0] = Ustrlen(s);
126 else if (expand_setup > 0) expand_setup--;
128 /* Regular expression match: compile, match, and set up $ variables if
131 if (pattern[0] == '^')
133 const pcre * re = regex_must_compile(pattern, cb->caseless, FALSE);
135 ? pcre_exec(re, NULL, CCS s, Ustrlen(s), 0, PCRE_EOPT, NULL, 0) < 0
136 : !regex_match_and_setup(re, s, 0, expand_setup)
139 if (valueptr) *valueptr = pattern; /* "value" gets the RE */
145 if (pattern[0] == '*')
147 int slen = Ustrlen(s);
148 int patlen; /* Sun compiler doesn't like non-constant initializer */
150 patlen = Ustrlen(++pattern);
151 if (patlen > slen) return FAIL;
153 ? strncmpic(s + slen - patlen, pattern, patlen) != 0
154 : Ustrncmp(s + slen - patlen, pattern, patlen) != 0)
156 if (expand_setup >= 0)
158 expand_nstring[++expand_setup] = s; /* write a $n, the matched subject variable-part */
159 expand_nlength[expand_setup] = slen - patlen;
160 expand_nmax = expand_setup; /* commit also $0, the matched subject */
162 if (valueptr) *valueptr = pattern - 1; /* "value" gets the (original) pattern */
166 /* Match a special item starting with @ if so enabled. On its own, "@" matches
167 the primary host name - implement this by changing the pattern. For the other
168 cases we have to do some more work. If we don't recognize a special pattern,
169 just fall through - the match will fail. */
171 if (cb->at_is_special && pattern[0] == '@')
175 pattern = primary_hostname;
176 goto NOT_AT_SPECIAL; /* Handle as exact string match */
179 if (Ustrcmp(pattern, "@[]") == 0)
181 int slen = Ustrlen(s);
182 if (s[0] != '[' && s[slen-1] != ']') return FAIL; /*XXX should this be || ? */
183 for (ip_address_item * ip = host_find_interfaces(); ip; ip = ip->next)
184 if (Ustrncmp(ip->address, s+1, slen - 2) == 0
185 && ip->address[slen - 2] == 0)
187 if (expand_setup >= 0) expand_nmax = expand_setup; /* commit $0, the IP addr */
188 if (valueptr) *valueptr = pattern; /* "value" gets the pattern */
194 if (strncmpic(pattern, US"@mx_", 4) == 0)
200 BOOL removed = FALSE;
201 const uschar *ss = pattern + 4;
202 const uschar *ignore_target_hosts = NULL;
204 if (strncmpic(ss, US"any", 3) == 0) ss += 3;
205 else if (strncmpic(ss, US"primary", 7) == 0)
210 else if (strncmpic(ss, US"secondary", 9) == 0)
215 else goto NOT_AT_SPECIAL;
217 if (strncmpic(ss, US"/ignore=", 8) == 0) ignore_target_hosts = ss + 8;
218 else if (*ss) goto NOT_AT_SPECIAL;
224 rc = host_find_bydns(&h,
226 HOST_FIND_BY_MX, /* search only for MX, not SRV or A */
227 NULL, /* service name not relevant */
228 NULL, /* srv_fail_domains not relevant */
229 NULL, /* mx_fail_domains not relevant */
230 NULL, /* no dnssec request/require XXX ? */
231 NULL, /* no feedback FQDN */
232 &removed); /* feedback if local removed */
234 if (rc == HOST_FIND_AGAIN)
236 search_error_message = string_sprintf("DNS lookup of \"%s\" deferred", s);
240 if ((rc != HOST_FOUND_LOCAL || secy) && (prim || !removed))
243 if (expand_setup >= 0) expand_nmax = expand_setup; /* commit $0, the matched subject */
244 if (valueptr) *valueptr = pattern; /* "value" gets the patterm */
247 /*** The above line used to be the following line, but this is incorrect,
248 because host_find_bydns() may return HOST_NOT_FOUND if it removed some MX
249 hosts, but the remaining ones were non-existent. All we are interested in
250 is whether or not it removed some hosts.
252 return (rc == HOST_FOUND && removed)? OK : FAIL;
257 /* Escape point from code for specials that start with "@" */
261 /* This is an exact string match if there is no semicolon in the pattern. */
263 if ((semicolon = Ustrchr(pattern, ';')) == NULL)
265 if (cb->caseless ? strcmpic(s, pattern) != 0 : Ustrcmp(s, pattern) != 0)
267 if (expand_setup >= 0) expand_nmax = expand_setup; /* Original code! $0 gets the matched subject */
268 if (valueptr) *valueptr = pattern; /* "value" gets the pattern */
272 /* Otherwise we have a lookup item. The lookup type, including partial, etc. is
273 the part of the string preceding the semicolon. */
276 search_type = search_findtype_partial(pattern, &partial, &affix, &affixlen,
279 if (search_type < 0) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
280 search_error_message);
282 /* Partial matching is not appropriate for certain lookups (e.g. when looking
283 up user@domain for sender rejection). There's a flag to disable it. */
285 if (!cb->use_partial) partial = -1;
287 /* Set the parameters for the three different kinds of lookup. */
289 keyquery = semicolon + 1;
290 Uskip_whitespace(&keyquery);
292 if (mac_islookup(search_type, lookup_absfilequery))
295 while (*keyquery && !isspace(*keyquery)) keyquery++;
296 filename = string_copyn(filename, keyquery - filename);
297 Uskip_whitespace(&keyquery);
300 else if (!mac_islookup(search_type, lookup_querystyle))
306 /* Now do the actual lookup; throw away the data returned unless it was asked
307 for; partial matching is all handled inside search_find(). Note that there is
308 no search_close() because of the caching arrangements. */
310 if (!(handle = search_open(filename, search_type, 0, NULL, NULL)))
311 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s", search_error_message);
312 result = search_find(handle, filename, keyquery, partial, affix, affixlen,
313 starflags, &expand_setup, opts);
315 if (!result) return f.search_find_defer ? DEFER : FAIL;
316 if (valueptr) *valueptr = result;
318 expand_nmax = expand_setup;
324 /*************************************************
325 * Public interface to check_string() *
326 *************************************************/
328 /* This function is called from several places where is it most convenient to
329 pass the arguments individually. It places them in a check_string_block
330 structure, and then calls check_string().
333 s the subject string to be checked
334 pattern the pattern to check it against
335 expand_setup expansion setup option (see check_string())
336 use_partial if FALSE, override any partial- search types
337 caseless TRUE for caseless matching where possible
338 at_is_special TRUE to recognize @, @[], etc.
339 valueptr if not NULL, and a file lookup was done, return the result
340 here instead of discarding it; else set it to point to NULL
342 Returns: OK if matched
344 DEFER if lookup deferred
348 match_check_string(const uschar *s, const uschar *pattern, int expand_setup,
349 BOOL use_partial, BOOL caseless, BOOL at_is_special, const uschar **valueptr)
351 check_string_block cb;
353 cb.subject = caseless ? string_copylc(s) : string_copy(s);
354 cb.expand_setup = expand_setup;
355 cb.use_partial = use_partial;
356 cb.caseless = caseless;
357 cb.at_is_special = at_is_special;
358 return check_string(&cb, pattern, valueptr, NULL);
363 /*************************************************
364 * Get key string from check block *
365 *************************************************/
367 /* When caching the data from a lookup for a named list, we have to save the
368 key that was found, because other lookups of different keys on the same list
369 may occur. This function has knowledge of the different lookup types, and
370 extracts the appropriate key.
374 type MCL_STRING, MCL_DOMAIN, MCL_HOST, MCL_ADDRESS, or MCL_LOCALPART
377 static const uschar *
378 get_check_key(void *arg, int type)
385 return ((check_string_block *)arg)->subject;
388 return ((check_host_block *)arg)->host_address;
391 return ((check_address_block *)arg)->address;
393 return US""; /* In practice, should never happen */
398 /*************************************************
399 * Scan list and run matching function *
400 *************************************************/
402 /* This function scans a list of patterns, and runs a matching function for
403 each item in the list. It is called from the functions that match domains,
404 local parts, hosts, and addresses, because its overall structure is the same in
405 all cases. However, the details of each particular match is different, so it
406 calls back to a given function do perform an actual match.
408 We can't quite keep the different types anonymous here because they permit
409 different special cases. A pity.
411 If a list item starts with !, that implies negation if the subject matches the
412 rest of the item (ignoring white space after the !). The result when the end of
413 the list is reached is FALSE unless the last item on the list is negated, in
414 which case it is TRUE. A file name in the list causes its lines to be
415 interpolated as if items in the list. An item starting with + is a named
416 sublist, obtained by searching the tree pointed to by anchorptr, with possible
417 cached match results in cache_bits.
420 listptr pointer to the pointer to the list
421 sep separator character for string_nextinlist();
422 normally zero for a standard list;
423 sometimes UCHAR_MAX+1 for single items;
424 anchorptr -> tree of named items, or NULL if no named items
425 cache_ptr pointer to pointer to cache bits for named items, or
426 pointer to NULL if not caching; may get set NULL if an
427 uncacheable named list is encountered
428 func function to call back to do one test
429 arg pointer to pass to the function; the string to be matched is
430 in the structure it points to
431 type MCL_STRING, MCL_DOMAIN, MCL_HOST, MCL_ADDRESS, or MCL_LOCALPART
432 these are used for some special handling
433 MCL_NOEXPAND (whose value is greater than any of them) may
434 be added to any value to suppress expansion of the list
435 name string to use in debugging info
436 valueptr where to pass back data from a lookup
438 Returns: OK if matched a non-negated item
439 OK if hit end of list after a negated item
440 FAIL if expansion force-failed
441 FAIL if matched a negated item
442 FAIL if hit end of list after a non-negated item
443 DEFER if a something deferred or expansion failed
447 match_check_list(const uschar **listptr, int sep, tree_node **anchorptr,
448 unsigned int **cache_ptr, int (*func)(void *,const uschar *,const uschar **,uschar **),
449 void *arg, int type, const uschar *name, const uschar **valueptr)
452 unsigned int *original_cache_bits = *cache_ptr;
453 BOOL include_unknown = FALSE;
454 BOOL ignore_unknown = FALSE;
455 BOOL include_defer = FALSE;
456 BOOL ignore_defer = FALSE;
461 /* Save time by not scanning for the option name when we don't need it. */
465 uschar *listname = readconf_find_option(listptr);
466 if (listname[0] != 0) ot = string_sprintf("%s in %s?", name, listname);
469 /* If the list is empty, the answer is no. Skip the debugging output for
474 HDEBUG(D_lists) if (ot) debug_printf("%s no (option unset)\n", ot);
478 /* Expand the list before we scan it. A forced expansion gives the answer
479 "not in list"; other expansion errors cause DEFER to be returned. However,
480 if the type value is greater than or equal to than MCL_NOEXPAND, do not expand
483 if (type >= MCL_NOEXPAND)
486 type -= MCL_NOEXPAND; /* Remove the "no expand" flag */
490 /* If we are searching a domain list, and $domain is not set, set it to the
491 subject that is being sought for the duration of the expansion. */
493 if (type == MCL_DOMAIN && !deliver_domain)
495 check_string_block *cb = (check_string_block *)arg;
496 deliver_domain = string_copy(cb->subject);
497 list = expand_cstring(*listptr);
498 deliver_domain = NULL;
501 list = expand_cstring(*listptr);
505 if (f.expand_string_forcedfail)
507 HDEBUG(D_lists) debug_printf("expansion of \"%s\" forced failure: "
508 "assume not in this list\n", *listptr);
511 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand \"%s\" while checking "
512 "a list: %s", *listptr, expand_string_message);
517 /* For an unnamed list, use the expanded version in comments */
519 HDEBUG(D_any) if (!ot) ot = string_sprintf("%s in \"%s\"?", name, list);
521 /* Now scan the list and process each item in turn, until one of them matches,
522 or we hit an error. */
524 while ((sss = string_nextinlist(&list, &sep, NULL, 0)))
528 /* Address lists may contain +caseful, to restore caseful matching of the
529 local part. We have to know the layout of the control block, unfortunately.
530 The lower cased address is in a temporary buffer, so we just copy the local
531 part back to the start of it (if a local part exists). */
533 if (type == MCL_ADDRESS)
535 if (Ustrcmp(ss, "+caseful") == 0)
537 check_address_block *cb = (check_address_block *)arg;
538 uschar *at = Ustrrchr(cb->origaddress, '@');
541 Ustrncpy(cb->address, cb->origaddress, at - cb->origaddress);
542 cb->caseless = FALSE;
547 /* Similar processing for local parts */
549 else if (type == MCL_LOCALPART)
551 if (Ustrcmp(ss, "+caseful") == 0)
553 check_string_block *cb = (check_string_block *)arg;
554 Ustrcpy(US cb->subject, cb->origsubject);
555 cb->caseless = FALSE;
560 /* If the host item is "+include_unknown" or "+ignore_unknown", remember it
561 in case there's a subsequent failed reverse lookup. There is similar
562 processing for "defer". */
564 else if (type == MCL_HOST && *ss == '+')
566 if (Ustrcmp(ss, "+include_unknown") == 0)
568 include_unknown = TRUE;
569 ignore_unknown = FALSE;
572 if (Ustrcmp(ss, "+ignore_unknown") == 0)
574 ignore_unknown = TRUE;
575 include_unknown = FALSE;
578 if (Ustrcmp(ss, "+include_defer") == 0)
580 include_defer = TRUE;
581 ignore_defer = FALSE;
584 if (Ustrcmp(ss, "+ignore_defer") == 0)
587 include_defer = FALSE;
592 /* Starting with ! specifies a negative item. It is theoretically possible
593 for a local part to start with !. In that case, a regex has to be used. */
598 while (isspace((*(++ss))));
603 /* If the item does not begin with '/', it might be a + item for a named
604 list. Otherwise, it is just a single list entry that has to be matched.
605 We recognize '+' only when supplied with a tree of named lists. */
609 if (*ss == '+' && anchorptr)
614 unsigned int *use_cache_bits = original_cache_bits;
615 uschar *cached = US"";
619 if (!(t = tree_search(*anchorptr, ss+1)))
621 log_write(0, LOG_MAIN|LOG_PANIC, "unknown named%s list \"%s\"",
622 type == MCL_DOMAIN ? " domain" :
623 type == MCL_HOST ? " host" :
624 type == MCL_ADDRESS ? " address" :
625 type == MCL_LOCALPART ? " local part" : "",
631 /* If the list number is negative, it means that this list is not
632 cacheable because it contains expansion items. */
634 if (nb->number < 0) use_cache_bits = NULL;
636 /* If we have got a cache pointer, get the bits. This is not an "else"
637 because the pointer may be NULL from the start if caching is not
642 offset = (nb->number)/16;
643 shift = ((nb->number)%16)*2;
644 bits = use_cache_bits[offset] & (3 << shift);
647 /* Not previously tested or no cache - run the full test */
651 switch (match_check_list(&(nb->string), 0, anchorptr, &use_cache_bits,
652 func, arg, type, name, valueptr))
654 case OK: bits = 1; break;
655 case FAIL: bits = 3; break;
656 case DEFER: goto DEFER_RETURN;
659 /* If this list was uncacheable, or a sublist turned out to be
660 uncacheable, the value of use_cache_bits will now be NULL, even if it
661 wasn't before. Ensure that this is passed up to the next level.
662 Otherwise, remember the result of the search in the cache. */
668 use_cache_bits[offset] |= bits << shift;
672 int old_pool = store_pool;
673 namedlist_cacheblock *p;
675 /* Cached data for hosts persists over more than one message,
676 so we use the permanent store pool */
678 store_pool = POOL_PERM;
679 p = store_get(sizeof(namedlist_cacheblock), FALSE);
680 p->key = string_copy(get_check_key(arg, type));
683 p->data = *valueptr ? string_copy(*valueptr) : NULL;
684 store_pool = old_pool;
686 p->next = nb->cache_data;
689 DEBUG(D_lists) debug_printf("data from lookup saved for "
690 "cache for %s: key '%s' value '%s'\n", ss, p->key, *valueptr);
695 /* Previously cached; to find a lookup value, search a chain of values
696 and compare keys. Typically, there is only one such, but it is possible
697 for different keys to have matched the same named list. */
701 DEBUG(D_lists) debug_printf("cached %s match for %s\n",
702 (bits & (-bits)) == bits ? "yes" : "no", ss);
704 cached = US" - cached";
707 const uschar *key = get_check_key(arg, type);
709 for (namedlist_cacheblock * p = nb->cache_data; p; p = p->next)
710 if (Ustrcmp(key, p->key) == 0)
715 DEBUG(D_lists) debug_printf("cached lookup data = %s\n", *valueptr);
719 /* Result of test is indicated by value in bits. For each test, we
720 have 00 => untested, 01 => tested yes, 11 => tested no. */
722 if ((bits & (-bits)) == bits) /* Only one of the two bits is set */
724 HDEBUG(D_lists) debug_printf("%s %s (matched \"%s\"%s)\n", ot,
725 (yield == OK)? "yes" : "no", sss, cached);
730 /* Run the provided function to do the individual test. */
734 uschar * error = NULL;
735 switch ((func)(arg, ss, valueptr, &error))
738 HDEBUG(D_lists) debug_printf("%s %s (matched \"%s\")\n", ot,
739 (yield == OK)? "yes" : "no", sss);
744 error = string_sprintf("DNS lookup of \"%s\" deferred", ss);
747 HDEBUG(D_lists) debug_printf("%s: item ignored by +ignore_defer\n",
753 log_write(0, LOG_MAIN, "%s: accepted by +include_defer", error);
756 if (!search_error_message) search_error_message = error;
759 /* The ERROR return occurs when checking hosts, when either a forward
760 or reverse lookup has failed. It can also occur in a match_ip list if a
761 non-IP address item is encountered. The error string gives details of
767 HDEBUG(D_lists) debug_printf("%s: item ignored by +ignore_unknown\n",
772 HDEBUG(D_lists) debug_printf("%s %s (%s)\n", ot,
773 include_unknown? "yes":"no", error);
774 if (!include_unknown)
776 if (LOGGING(unknown_in_list))
777 log_write(0, LOG_MAIN, "list matching forced to fail: %s", error);
780 log_write(0, LOG_MAIN, "%s: accepted by +include_unknown", error);
787 /* If the item is a file name, we read the file and do a match attempt
788 on each line in the file, including possibly more negation processing. */
792 int file_yield = yield; /* In case empty file */
793 uschar * filename = ss;
794 FILE * f = Ufopen(filename, "rb");
795 uschar filebuffer[1024];
797 /* ot will be null in non-debugging cases, and anyway, we get better
798 wording by reworking it. */
802 uschar * listname = readconf_find_option(listptr);
803 if (listname[0] == 0)
804 listname = string_sprintf("\"%s\"", *listptr);
805 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
806 string_open_failed(errno, "%s when checking %s", sss, listname));
809 /* Trailing comments are introduced by #, but in an address list or local
810 part list, the # must be preceded by white space or the start of the line,
811 because the # character is a legal character in local parts. */
813 while (Ufgets(filebuffer, sizeof(filebuffer), f) != NULL)
816 uschar *sss = filebuffer;
818 while ((ss = Ustrchr(sss, '#')) != NULL)
820 if ((type != MCL_ADDRESS && type != MCL_LOCALPART) ||
821 ss == filebuffer || isspace(ss[-1]))
829 ss = filebuffer + Ustrlen(filebuffer); /* trailing space */
830 while (ss > filebuffer && isspace(ss[-1])) ss--;
834 while (isspace(*ss)) ss++; /* leading space */
836 if (*ss == 0) continue; /* ignore empty */
838 file_yield = yield; /* positive yield */
839 sss = ss; /* for debugging */
841 if (*ss == '!') /* negation */
843 file_yield = (file_yield == OK)? FAIL : OK;
844 while (isspace((*(++ss))));
847 switch ((func)(arg, ss, valueptr, &error))
851 HDEBUG(D_lists) debug_printf("%s %s (matched \"%s\" in %s)\n", ot,
852 yield == OK ? "yes" : "no", sss, filename);
854 /* The "pattern" being matched came from the file; we use a stack-local.
855 Copy it to allocated memory now we know it matched. */
857 if (valueptr) *valueptr = string_copy(ss);
862 error = string_sprintf("DNS lookup of %s deferred", ss);
865 HDEBUG(D_lists) debug_printf("%s: item ignored by +ignore_defer\n",
872 log_write(0, LOG_MAIN, "%s: accepted by +include_defer", error);
877 case ERROR: /* host name lookup failed - this can only */
878 if (ignore_unknown) /* be for an incoming host (not outgoing) */
880 HDEBUG(D_lists) debug_printf("%s: item ignored by +ignore_unknown\n",
885 HDEBUG(D_lists) debug_printf("%s %s (%s)\n", ot,
886 include_unknown? "yes":"no", error);
888 if (!include_unknown)
890 if (LOGGING(unknown_in_list))
891 log_write(0, LOG_MAIN, "list matching forced to fail: %s", error);
894 log_write(0, LOG_MAIN, "%s: accepted by +include_unknown", error);
900 /* At the end of the file, leave the yield setting at the final setting
901 for the file, in case this is the last item in the list. */
906 } /* Loop for the next item on the top-level list */
908 /* End of list reached: if the last item was negated yield OK, else FAIL. */
911 debug_printf("%s %s (end of list)\n", ot, yield == OK ? "no":"yes");
912 return yield == OK ? FAIL : OK;
914 /* Something deferred */
917 HDEBUG(D_lists) debug_printf("%s list match deferred for %s\n", ot, sss);
922 /*************************************************
923 * Match in colon-separated list *
924 *************************************************/
926 /* This function is used for domain lists and local part lists. It is not used
927 for host lists or address lists, which have additional interpretation of the
928 patterns. Some calls of it set sep > UCHAR_MAX in order to use its matching
929 facilities on single items. When this is done, it arranges to set the numerical
930 variables as a result of the match.
932 This function is now just a short interface to match_check_list(), which does
933 list scanning in a general way. A good compiler will optimize the tail
937 s string to search for
938 listptr ptr to ptr to colon separated list of patterns, or NULL
939 sep a separator value for the list (see string_nextinlist())
940 anchorptr ptr to tree for named items, or NULL if no named items
941 cache_bits ptr to cache_bits for ditto, or NULL if not caching
942 type MCL_DOMAIN when matching a domain list
943 MCL_LOCALPART when matching a local part list (address lists
944 have their own function)
945 MCL_STRING for others (e.g. list of ciphers)
946 MCL_NOEXPAND (whose value is greater than any of them) may
947 be added to any value to suppress expansion of the list
948 caseless TRUE for (mostly) caseless matching - passed directly to
950 valueptr pointer to where any lookup data is to be passed back,
951 or NULL (just passed on to match_check_string)
953 Returns: OK if matched a non-negated item
954 OK if hit end of list after a negated item
955 FAIL if expansion force-failed
956 FAIL if matched a negated item
957 FAIL if hit end of list after a non-negated item
958 DEFER if a lookup deferred
962 match_isinlist(const uschar *s, const uschar **listptr, int sep,
963 tree_node **anchorptr,
964 unsigned int *cache_bits, int type, BOOL caseless, const uschar **valueptr)
966 unsigned int *local_cache_bits = cache_bits;
967 check_string_block cb;
969 cb.subject = caseless ? string_copylc(s) : string_copy(s);
970 cb.at_is_special = FALSE;
971 switch (type & ~MCL_NOEXPAND)
973 case MCL_DOMAIN: cb.at_is_special = TRUE; /*FALLTHROUGH*/
974 case MCL_LOCALPART: cb.expand_setup = 0; break;
975 default: cb.expand_setup = sep > UCHAR_MAX ? 0 : -1; break;
977 cb.use_partial = TRUE;
978 cb.caseless = caseless;
979 if (valueptr) *valueptr = NULL;
980 return match_check_list(listptr, sep, anchorptr, &local_cache_bits,
981 check_string, &cb, type, s, valueptr);
986 /*************************************************
987 * Match address to single address-list item *
988 *************************************************/
990 /* This function matches an address to an item from an address list. It is
991 called from match_address_list() via match_check_list(). That is why most of
992 its arguments are in an indirect block.
995 arg the argument block (see below)
996 pattern the pattern to match
997 valueptr where to return a value
998 error for error messages (not used in this function; it never
1001 The argument block contains:
1002 address the start of the subject address; when called from retry.c
1003 it may be *@domain if the local part isn't relevant
1004 origaddress the original, un-case-forced address (not used here, but used
1005 in match_check_list() when +caseful is encountered)
1006 expand_setup controls setting up of $n variables
1007 caseless TRUE for caseless local part matching
1009 Returns: OK for a match
1011 DEFER if a lookup deferred
1015 check_address(void *arg, const uschar *pattern, const uschar **valueptr, uschar **error)
1017 check_address_block *cb = (check_address_block *)arg;
1018 check_string_block csb;
1021 unsigned int *null = NULL;
1022 const uschar *listptr;
1023 uschar *subject = cb->address;
1025 uschar *pdomain, *sdomain;
1027 error = error; /* Keep clever compilers from complaining */
1029 DEBUG(D_lists) debug_printf("address match test: subject=%s pattern=%s\n",
1032 /* Find the subject's domain */
1034 sdomain = Ustrrchr(subject, '@');
1036 /* The only case where a subject may not have a domain is if the subject is
1037 empty. Otherwise, a subject with no domain is a serious configuration error. */
1039 if (sdomain == NULL && *subject != 0)
1041 log_write(0, LOG_MAIN|LOG_PANIC, "no @ found in the subject of an "
1042 "address list match: subject=\"%s\" pattern=\"%s\"", subject, pattern);
1046 /* Handle a regular expression, which must match the entire incoming address.
1047 This may be the empty address. */
1049 if (*pattern == '^')
1050 return match_check_string(subject, pattern, cb->expand_setup, TRUE,
1051 cb->caseless, FALSE, NULL);
1053 /* Handle a pattern that is just a lookup. Skip over possible lookup names
1054 (letters, digits, hyphens). Skip over a possible * or *@ at the end. Then we
1055 must have a semicolon for it to be a lookup. */
1057 for (s = pattern; isalnum(*s) || *s == '-'; s++);
1061 /* If it is a straight lookup, do a lookup for the whole address. This may be
1062 the empty address. Partial matching doesn't make sense here, so we ignore it,
1063 but write a panic log entry. However, *@ matching will be honoured. */
1067 if (Ustrncmp(pattern, "partial-", 8) == 0)
1068 log_write(0, LOG_MAIN|LOG_PANIC, "partial matching is not applicable to "
1069 "whole-address lookups: ignored \"partial-\" in \"%s\"", pattern);
1070 return match_check_string(subject, pattern, -1, FALSE, cb->caseless, FALSE,
1074 /* For the remaining cases, an empty subject matches only an empty pattern,
1075 because other patterns expect to have a local part and a domain to match
1078 if (*subject == 0) return (*pattern == 0)? OK : FAIL;
1080 /* If the pattern starts with "@@" we have a split lookup, where the domain is
1081 looked up to obtain a list of local parts. If the subject's local part is just
1082 "*" (called from retry) the match always fails. */
1084 if (pattern[0] == '@' && pattern[1] == '@')
1088 uschar buffer[1024];
1090 if (sdomain == subject + 1 && *subject == '*') return FAIL;
1092 /* Loop for handling chains. The last item in any list may be of the form
1093 ">name" in order to chain on to another list. */
1095 for (const uschar * key = sdomain + 1; key && watchdog-- > 0; )
1099 if ((rc = match_check_string(key, pattern + 2, -1, TRUE, FALSE, FALSE,
1100 CUSS &list)) != OK) return rc;
1102 /* Check for chaining from the last item; set up the next key if one
1105 ss = Ustrrchr(list, ':');
1106 if (ss == NULL) ss = list; else ss++;
1107 while (isspace(*ss)) ss++;
1111 while (isspace(*ss)) ss++;
1112 key = string_copy(ss);
1116 /* Look up the local parts provided by the list; negation is permitted.
1117 If a local part has to begin with !, a regex can be used. */
1119 while ((ss = string_nextinlist(CUSS &list, &sep, buffer, sizeof(buffer))))
1126 while (isspace((*(++ss))));
1128 else local_yield = OK;
1131 rc = match_check_string(subject, ss, -1, TRUE, cb->caseless, FALSE,
1146 /* End of chain loop; panic if too many times */
1149 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Loop detected in lookup of "
1150 "local part of %s in %s", subject, pattern);
1152 /* Otherwise the local part check has failed, so the whole match
1159 /* We get here if the pattern is not a lookup or a regular expression. If it
1160 contains an @ there is both a local part and a domain. */
1162 pdomain = Ustrrchr(pattern, '@');
1163 if (pdomain != NULL)
1167 /* If the domain in the pattern is empty or one of the special cases [] or
1168 mx_{any,primary,secondary}, and the local part in the pattern ends in "@",
1169 we have a pattern of the form <something>@@, <something>@@[], or
1170 <something>@@mx_{any,primary,secondary}. These magic "domains" are
1171 automatically interpreted in match_check_string. We just need to arrange that
1172 the leading @ is included in the domain. */
1174 if (pdomain > pattern && pdomain[-1] == '@' &&
1176 Ustrcmp(pdomain+1, "[]") == 0 ||
1177 Ustrcmp(pdomain+1, "mx_any") == 0 ||
1178 Ustrcmp(pdomain+1, "mx_primary") == 0 ||
1179 Ustrcmp(pdomain+1, "mx_secondary") == 0))
1182 pllen = pdomain - pattern;
1183 sllen = sdomain - subject;
1185 /* Compare the local parts in the subject and the pattern */
1187 if (*pattern == '*')
1189 int cllen = pllen - 1;
1190 if (sllen < cllen) return FAIL;
1192 ? strncmpic(subject+sllen-cllen, pattern + 1, cllen) != 0
1193 : Ustrncmp(subject+sllen-cllen, pattern + 1, cllen) != 0)
1195 if (cb->expand_setup > 0)
1197 expand_nstring[cb->expand_setup] = subject;
1198 expand_nlength[cb->expand_setup] = sllen - cllen;
1204 if (sllen != pllen) return FAIL;
1206 ? strncmpic(subject, pattern, sllen) != 0
1207 : Ustrncmp(subject, pattern, sllen) != 0) return FAIL;
1211 /* If the local part matched, or was not being checked, check the domain using
1212 the generalized function, which supports file lookups (which may defer). The
1213 original code read as follows:
1215 return match_check_string(sdomain + 1,
1216 pdomain ? pdomain + 1 : pattern,
1217 cb->expand_setup + expand_inc, TRUE, cb->caseless, TRUE, NULL);
1219 This supported only literal domains and *.x.y patterns. In order to allow for
1220 named domain lists (so that you can right, for example, "senders=+xxxx"), it
1221 was changed to use the list scanning function. */
1223 csb.origsubject = sdomain + 1;
1224 csb.subject = cb->caseless ? string_copylc(sdomain+1) : string_copy(sdomain+1);
1225 csb.expand_setup = cb->expand_setup + expand_inc;
1226 csb.use_partial = TRUE;
1227 csb.caseless = cb->caseless;
1228 csb.at_is_special = TRUE;
1230 listptr = pdomain ? pdomain + 1 : pattern;
1231 if (valueptr) *valueptr = NULL;
1233 return match_check_list(
1234 &listptr, /* list of one item */
1235 UCHAR_MAX+1, /* impossible separator; single item */
1236 &domainlist_anchor, /* it's a domain list */
1237 &null, /* ptr to NULL means no caching */
1238 check_string, /* the function to do one test */
1239 &csb, /* its data */
1240 MCL_DOMAIN + MCL_NOEXPAND, /* domain list; don't expand */
1241 csb.subject, /* string for messages */
1242 valueptr); /* where to pass back lookup data */
1248 /*************************************************
1249 * Test whether address matches address list *
1250 *************************************************/
1252 /* This function is given an address and a list of things to match it against.
1253 The list may contain individual addresses, regular expressions, lookup
1254 specifications, and indirection via bare files. Negation is supported. The
1255 address to check can consist of just a domain, which will then match only
1256 domain items or items specified as *@domain.
1258 Domains are always lower cased before the match. Local parts are also lower
1259 cased unless "caseless" is false. The work of actually scanning the list is
1260 done by match_check_list(), with an appropriate block of arguments and a
1261 callback to check_address(). During caseless matching, it will recognize
1262 +caseful and revert to caseful matching.
1265 address address to test
1266 caseless TRUE to start in caseless state
1267 expand TRUE to allow list expansion
1268 listptr list to check against
1269 cache_bits points to cache bits for named address lists, or NULL
1270 expand_setup controls setting up of $n variables - passed through
1271 to check_address (q.v.)
1272 sep separator character for the list;
1273 may be 0 to get separator from the list;
1274 may be UCHAR_MAX+1 for one-item list
1275 valueptr where to return a lookup value, or NULL
1277 Returns: OK for a positive match, or end list after a negation;
1278 FAIL for a negative match, or end list after non-negation;
1279 DEFER if a lookup deferred
1283 match_address_list(const uschar *address, BOOL caseless, BOOL expand,
1284 const uschar **listptr, unsigned int *cache_bits, int expand_setup, int sep,
1285 const uschar **valueptr)
1287 check_address_block ab;
1288 unsigned int *local_cache_bits = cache_bits;
1291 /* RFC 2505 recommends that for spam checking, local parts should be caselessly
1292 compared. Therefore, Exim now forces the entire address into lower case here,
1293 provided that "caseless" is set. (It is FALSE for calls for matching rewriting
1294 patterns.) Otherwise just the domain is lower cases. A magic item "+caseful" in
1295 the list can be used to restore a caseful copy of the local part from the
1297 Limit the subject address size to avoid mem-exhastion attacks. The size chosen
1298 is historical (we used to use big_buffer her). */
1300 if ((len = Ustrlen(address)) > BIG_BUFFER_SIZE) len = BIG_BUFFER_SIZE;
1301 ab.address = string_copyn(address, len);
1303 for (uschar * p = ab.address + len - 1; p >= ab.address; p--)
1305 if (!caseless && *p == '@') break;
1309 /* If expand_setup is zero, we need to set up $0 to the whole thing, in
1310 case there is a match. Can't use the built-in facilities of match_check_string
1311 (via check_address), as we may just be calling that for part of the address
1314 if (expand_setup == 0)
1316 expand_nstring[0] = string_copy(address);
1317 expand_nlength[0] = Ustrlen(address);
1321 /* Set up the data to be passed ultimately to check_address. */
1323 ab.origaddress = address;
1324 /* ab.address is above */
1325 ab.expand_setup = expand_setup;
1326 ab.caseless = caseless;
1328 return match_check_list(listptr, sep, &addresslist_anchor, &local_cache_bits,
1329 check_address, &ab, MCL_ADDRESS + (expand? 0:MCL_NOEXPAND), address,
1333 /* Simpler version of match_address_list; always caseless, expanding,
1334 no cache bits, no value-return.
1337 address address to test
1338 listptr list to check against
1339 sep separator character for the list;
1340 may be 0 to get separator from the list;
1341 may be UCHAR_MAX+1 for one-item list
1343 Returns: OK for a positive match, or end list after a negation;
1344 FAIL for a negative match, or end list after non-negation;
1345 DEFER if a lookup deferred
1349 match_address_list_basic(const uschar *address, const uschar **listptr, int sep)
1351 return match_address_list(address, TRUE, TRUE, listptr, NULL, -1, sep, NULL);
1354 /* End of match.c */