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 = search_args(search_type, s, semicolon+1, &filename, opts);
291 /* Now do the actual lookup; throw away the data returned unless it was asked
292 for; partial matching is all handled inside search_find(). Note that there is
293 no search_close() because of the caching arrangements. */
295 if (!(handle = search_open(filename, search_type, 0, NULL, NULL)))
296 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s", search_error_message);
297 result = search_find(handle, filename, keyquery, partial, affix, affixlen,
298 starflags, &expand_setup, opts);
300 if (!result) return f.search_find_defer ? DEFER : FAIL;
301 if (valueptr) *valueptr = result;
303 expand_nmax = expand_setup;
309 /*************************************************
310 * Public interface to check_string() *
311 *************************************************/
313 /* This function is called from several places where is it most convenient to
314 pass the arguments individually. It places them in a check_string_block
315 structure, and then calls check_string().
318 s the subject string to be checked
319 pattern the pattern to check it against
320 expand_setup expansion setup option (see check_string())
321 use_partial if FALSE, override any partial- search types
322 caseless TRUE for caseless matching where possible
323 at_is_special TRUE to recognize @, @[], etc.
324 valueptr if not NULL, and a file lookup was done, return the result
325 here instead of discarding it; else set it to point to NULL
327 Returns: OK if matched
329 DEFER if lookup deferred
333 match_check_string(const uschar *s, const uschar *pattern, int expand_setup,
334 BOOL use_partial, BOOL caseless, BOOL at_is_special, const uschar **valueptr)
336 check_string_block cb;
338 cb.subject = caseless ? string_copylc(s) : string_copy(s);
339 cb.expand_setup = expand_setup;
340 cb.use_partial = use_partial;
341 cb.caseless = caseless;
342 cb.at_is_special = at_is_special;
343 return check_string(&cb, pattern, valueptr, NULL);
348 /*************************************************
349 * Get key string from check block *
350 *************************************************/
352 /* When caching the data from a lookup for a named list, we have to save the
353 key that was found, because other lookups of different keys on the same list
354 may occur. This function has knowledge of the different lookup types, and
355 extracts the appropriate key.
359 type MCL_STRING, MCL_DOMAIN, MCL_HOST, MCL_ADDRESS, or MCL_LOCALPART
362 static const uschar *
363 get_check_key(void *arg, int type)
370 return ((check_string_block *)arg)->subject;
373 return ((check_host_block *)arg)->host_address;
376 return ((check_address_block *)arg)->address;
378 return US""; /* In practice, should never happen */
383 /*************************************************
384 * Scan list and run matching function *
385 *************************************************/
387 /* This function scans a list of patterns, and runs a matching function for
388 each item in the list. It is called from the functions that match domains,
389 local parts, hosts, and addresses, because its overall structure is the same in
390 all cases. However, the details of each particular match is different, so it
391 calls back to a given function do perform an actual match.
393 We can't quite keep the different types anonymous here because they permit
394 different special cases. A pity.
396 If a list item starts with !, that implies negation if the subject matches the
397 rest of the item (ignoring white space after the !). The result when the end of
398 the list is reached is FALSE unless the last item on the list is negated, in
399 which case it is TRUE. A file name in the list causes its lines to be
400 interpolated as if items in the list. An item starting with + is a named
401 sublist, obtained by searching the tree pointed to by anchorptr, with possible
402 cached match results in cache_bits.
405 listptr pointer to the pointer to the list
406 sep separator character for string_nextinlist();
407 normally zero for a standard list;
408 sometimes UCHAR_MAX+1 for single items;
409 anchorptr -> tree of named items, or NULL if no named items
410 cache_ptr pointer to pointer to cache bits for named items, or
411 pointer to NULL if not caching; may get set NULL if an
412 uncacheable named list is encountered
413 func function to call back to do one test
414 arg pointer to pass to the function; the string to be matched is
415 in the structure it points to
416 type MCL_STRING, MCL_DOMAIN, MCL_HOST, MCL_ADDRESS, or MCL_LOCALPART
417 these are used for some special handling
418 MCL_NOEXPAND (whose value is greater than any of them) may
419 be added to any value to suppress expansion of the list
420 name string to use in debugging info
421 valueptr where to pass back data from a lookup
423 Returns: OK if matched a non-negated item
424 OK if hit end of list after a negated item
425 FAIL if expansion force-failed
426 FAIL if matched a negated item
427 FAIL if hit end of list after a non-negated item
428 DEFER if a something deferred or expansion failed
432 match_check_list(const uschar **listptr, int sep, tree_node **anchorptr,
433 unsigned int **cache_ptr, int (*func)(void *,const uschar *,const uschar **,uschar **),
434 void *arg, int type, const uschar *name, const uschar **valueptr)
437 unsigned int *original_cache_bits = *cache_ptr;
438 BOOL include_unknown = FALSE;
439 BOOL ignore_unknown = FALSE;
440 BOOL include_defer = FALSE;
441 BOOL ignore_defer = FALSE;
446 /* Save time by not scanning for the option name when we don't need it. */
450 uschar *listname = readconf_find_option(listptr);
451 if (listname[0] != 0) ot = string_sprintf("%s in %s?", name, listname);
454 /* If the list is empty, the answer is no. Skip the debugging output for
459 HDEBUG(D_lists) if (ot) debug_printf("%s no (option unset)\n", ot);
463 /* Expand the list before we scan it. A forced expansion gives the answer
464 "not in list"; other expansion errors cause DEFER to be returned. However,
465 if the type value is greater than or equal to than MCL_NOEXPAND, do not expand
468 if (type >= MCL_NOEXPAND)
471 type -= MCL_NOEXPAND; /* Remove the "no expand" flag */
475 /* If we are searching a domain list, and $domain is not set, set it to the
476 subject that is being sought for the duration of the expansion. */
478 if (type == MCL_DOMAIN && !deliver_domain)
480 check_string_block *cb = (check_string_block *)arg;
481 deliver_domain = string_copy(cb->subject);
482 list = expand_cstring(*listptr);
483 deliver_domain = NULL;
486 list = expand_cstring(*listptr);
490 if (f.expand_string_forcedfail)
492 HDEBUG(D_lists) debug_printf("expansion of \"%s\" forced failure: "
493 "assume not in this list\n", *listptr);
496 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand \"%s\" while checking "
497 "a list: %s", *listptr, expand_string_message);
502 /* For an unnamed list, use the expanded version in comments */
504 HDEBUG(D_any) if (!ot) ot = string_sprintf("%s in \"%s\"?", name, list);
506 /* Now scan the list and process each item in turn, until one of them matches,
507 or we hit an error. */
509 while ((sss = string_nextinlist(&list, &sep, NULL, 0)))
513 /* Address lists may contain +caseful, to restore caseful matching of the
514 local part. We have to know the layout of the control block, unfortunately.
515 The lower cased address is in a temporary buffer, so we just copy the local
516 part back to the start of it (if a local part exists). */
518 if (type == MCL_ADDRESS)
520 if (Ustrcmp(ss, "+caseful") == 0)
522 check_address_block *cb = (check_address_block *)arg;
523 uschar *at = Ustrrchr(cb->origaddress, '@');
526 Ustrncpy(cb->address, cb->origaddress, at - cb->origaddress);
527 cb->caseless = FALSE;
532 /* Similar processing for local parts */
534 else if (type == MCL_LOCALPART)
536 if (Ustrcmp(ss, "+caseful") == 0)
538 check_string_block *cb = (check_string_block *)arg;
539 Ustrcpy(US cb->subject, cb->origsubject);
540 cb->caseless = FALSE;
545 /* If the host item is "+include_unknown" or "+ignore_unknown", remember it
546 in case there's a subsequent failed reverse lookup. There is similar
547 processing for "defer". */
549 else if (type == MCL_HOST && *ss == '+')
551 if (Ustrcmp(ss, "+include_unknown") == 0)
553 include_unknown = TRUE;
554 ignore_unknown = FALSE;
557 if (Ustrcmp(ss, "+ignore_unknown") == 0)
559 ignore_unknown = TRUE;
560 include_unknown = FALSE;
563 if (Ustrcmp(ss, "+include_defer") == 0)
565 include_defer = TRUE;
566 ignore_defer = FALSE;
569 if (Ustrcmp(ss, "+ignore_defer") == 0)
572 include_defer = FALSE;
577 /* Starting with ! specifies a negative item. It is theoretically possible
578 for a local part to start with !. In that case, a regex has to be used. */
583 while (isspace((*(++ss))));
588 /* If the item does not begin with '/', it might be a + item for a named
589 list. Otherwise, it is just a single list entry that has to be matched.
590 We recognize '+' only when supplied with a tree of named lists. */
594 if (*ss == '+' && anchorptr)
599 unsigned int *use_cache_bits = original_cache_bits;
600 uschar *cached = US"";
604 if (!(t = tree_search(*anchorptr, ss+1)))
606 log_write(0, LOG_MAIN|LOG_PANIC, "unknown named%s list \"%s\"",
607 type == MCL_DOMAIN ? " domain" :
608 type == MCL_HOST ? " host" :
609 type == MCL_ADDRESS ? " address" :
610 type == MCL_LOCALPART ? " local part" : "",
616 /* If the list number is negative, it means that this list is not
617 cacheable because it contains expansion items. */
619 if (nb->number < 0) use_cache_bits = NULL;
621 /* If we have got a cache pointer, get the bits. This is not an "else"
622 because the pointer may be NULL from the start if caching is not
627 offset = (nb->number)/16;
628 shift = ((nb->number)%16)*2;
629 bits = use_cache_bits[offset] & (3 << shift);
632 /* Not previously tested or no cache - run the full test */
636 switch (match_check_list(&(nb->string), 0, anchorptr, &use_cache_bits,
637 func, arg, type, name, valueptr))
639 case OK: bits = 1; break;
640 case FAIL: bits = 3; break;
641 case DEFER: goto DEFER_RETURN;
644 /* If this list was uncacheable, or a sublist turned out to be
645 uncacheable, the value of use_cache_bits will now be NULL, even if it
646 wasn't before. Ensure that this is passed up to the next level.
647 Otherwise, remember the result of the search in the cache. */
653 use_cache_bits[offset] |= bits << shift;
657 int old_pool = store_pool;
658 namedlist_cacheblock *p;
660 /* Cached data for hosts persists over more than one message,
661 so we use the permanent store pool */
663 store_pool = POOL_PERM;
664 p = store_get(sizeof(namedlist_cacheblock), FALSE);
665 p->key = string_copy(get_check_key(arg, type));
668 p->data = *valueptr ? string_copy(*valueptr) : NULL;
669 store_pool = old_pool;
671 p->next = nb->cache_data;
674 DEBUG(D_lists) debug_printf("data from lookup saved for "
675 "cache for %s: key '%s' value '%s'\n", ss, p->key, *valueptr);
680 /* Previously cached; to find a lookup value, search a chain of values
681 and compare keys. Typically, there is only one such, but it is possible
682 for different keys to have matched the same named list. */
686 DEBUG(D_lists) debug_printf("cached %s match for %s\n",
687 (bits & (-bits)) == bits ? "yes" : "no", ss);
689 cached = US" - cached";
692 const uschar *key = get_check_key(arg, type);
694 for (namedlist_cacheblock * p = nb->cache_data; p; p = p->next)
695 if (Ustrcmp(key, p->key) == 0)
700 DEBUG(D_lists) debug_printf("cached lookup data = %s\n", *valueptr);
704 /* Result of test is indicated by value in bits. For each test, we
705 have 00 => untested, 01 => tested yes, 11 => tested no. */
707 if ((bits & (-bits)) == bits) /* Only one of the two bits is set */
709 HDEBUG(D_lists) debug_printf("%s %s (matched \"%s\"%s)\n", ot,
710 (yield == OK)? "yes" : "no", sss, cached);
715 /* Run the provided function to do the individual test. */
719 uschar * error = NULL;
720 switch ((func)(arg, ss, valueptr, &error))
723 HDEBUG(D_lists) debug_printf("%s %s (matched \"%s\")\n", ot,
724 (yield == OK)? "yes" : "no", sss);
729 error = string_sprintf("DNS lookup of \"%s\" deferred", ss);
732 HDEBUG(D_lists) debug_printf("%s: item ignored by +ignore_defer\n",
738 log_write(0, LOG_MAIN, "%s: accepted by +include_defer", error);
741 if (!search_error_message) search_error_message = error;
744 /* The ERROR return occurs when checking hosts, when either a forward
745 or reverse lookup has failed. It can also occur in a match_ip list if a
746 non-IP address item is encountered. The error string gives details of
752 HDEBUG(D_lists) debug_printf("%s: item ignored by +ignore_unknown\n",
757 HDEBUG(D_lists) debug_printf("%s %s (%s)\n", ot,
758 include_unknown? "yes":"no", error);
759 if (!include_unknown)
761 if (LOGGING(unknown_in_list))
762 log_write(0, LOG_MAIN, "list matching forced to fail: %s", error);
765 log_write(0, LOG_MAIN, "%s: accepted by +include_unknown", error);
772 /* If the item is a file name, we read the file and do a match attempt
773 on each line in the file, including possibly more negation processing. */
777 int file_yield = yield; /* In case empty file */
778 uschar * filename = ss;
779 FILE * f = Ufopen(filename, "rb");
780 uschar filebuffer[1024];
782 /* ot will be null in non-debugging cases, and anyway, we get better
783 wording by reworking it. */
787 uschar * listname = readconf_find_option(listptr);
788 if (listname[0] == 0)
789 listname = string_sprintf("\"%s\"", *listptr);
790 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
791 string_open_failed(errno, "%s when checking %s", sss, listname));
794 /* Trailing comments are introduced by #, but in an address list or local
795 part list, the # must be preceded by white space or the start of the line,
796 because the # character is a legal character in local parts. */
798 while (Ufgets(filebuffer, sizeof(filebuffer), f) != NULL)
801 uschar *sss = filebuffer;
803 while ((ss = Ustrchr(sss, '#')) != NULL)
805 if ((type != MCL_ADDRESS && type != MCL_LOCALPART) ||
806 ss == filebuffer || isspace(ss[-1]))
814 ss = filebuffer + Ustrlen(filebuffer); /* trailing space */
815 while (ss > filebuffer && isspace(ss[-1])) ss--;
819 while (isspace(*ss)) ss++; /* leading space */
821 if (*ss == 0) continue; /* ignore empty */
823 file_yield = yield; /* positive yield */
824 sss = ss; /* for debugging */
826 if (*ss == '!') /* negation */
828 file_yield = (file_yield == OK)? FAIL : OK;
829 while (isspace((*(++ss))));
832 switch ((func)(arg, ss, valueptr, &error))
836 HDEBUG(D_lists) debug_printf("%s %s (matched \"%s\" in %s)\n", ot,
837 yield == OK ? "yes" : "no", sss, filename);
842 error = string_sprintf("DNS lookup of %s deferred", ss);
845 HDEBUG(D_lists) debug_printf("%s: item ignored by +ignore_defer\n",
852 log_write(0, LOG_MAIN, "%s: accepted by +include_defer", error);
857 case ERROR: /* host name lookup failed - this can only */
858 if (ignore_unknown) /* be for an incoming host (not outgoing) */
860 HDEBUG(D_lists) debug_printf("%s: item ignored by +ignore_unknown\n",
865 HDEBUG(D_lists) debug_printf("%s %s (%s)\n", ot,
866 include_unknown? "yes":"no", error);
868 if (!include_unknown)
870 if (LOGGING(unknown_in_list))
871 log_write(0, LOG_MAIN, "list matching forced to fail: %s", error);
874 log_write(0, LOG_MAIN, "%s: accepted by +include_unknown", error);
880 /* At the end of the file, leave the yield setting at the final setting
881 for the file, in case this is the last item in the list. */
886 } /* Loop for the next item on the top-level list */
888 /* End of list reached: if the last item was negated yield OK, else FAIL. */
891 debug_printf("%s %s (end of list)\n", ot, yield == OK ? "no":"yes");
892 return yield == OK ? FAIL : OK;
894 /* Something deferred */
897 HDEBUG(D_lists) debug_printf("%s list match deferred for %s\n", ot, sss);
902 /*************************************************
903 * Match in colon-separated list *
904 *************************************************/
906 /* This function is used for domain lists and local part lists. It is not used
907 for host lists or address lists, which have additional interpretation of the
908 patterns. Some calls of it set sep > UCHAR_MAX in order to use its matching
909 facilities on single items. When this is done, it arranges to set the numerical
910 variables as a result of the match.
912 This function is now just a short interface to match_check_list(), which does
913 list scanning in a general way. A good compiler will optimize the tail
917 s string to search for
918 listptr ptr to ptr to colon separated list of patterns, or NULL
919 sep a separator value for the list (see string_nextinlist())
920 anchorptr ptr to tree for named items, or NULL if no named items
921 cache_bits ptr to cache_bits for ditto, or NULL if not caching
922 type MCL_DOMAIN when matching a domain list
923 MCL_LOCALPART when matching a local part list (address lists
924 have their own function)
925 MCL_STRING for others (e.g. list of ciphers)
926 MCL_NOEXPAND (whose value is greater than any of them) may
927 be added to any value to suppress expansion of the list
928 caseless TRUE for (mostly) caseless matching - passed directly to
930 valueptr pointer to where any lookup data is to be passed back,
931 or NULL (just passed on to match_check_string)
933 Returns: OK if matched a non-negated item
934 OK if hit end of list after a negated item
935 FAIL if expansion force-failed
936 FAIL if matched a negated item
937 FAIL if hit end of list after a non-negated item
938 DEFER if a lookup deferred
942 match_isinlist(const uschar *s, const uschar **listptr, int sep,
943 tree_node **anchorptr,
944 unsigned int *cache_bits, int type, BOOL caseless, const uschar **valueptr)
946 unsigned int *local_cache_bits = cache_bits;
947 check_string_block cb;
949 cb.subject = caseless ? string_copylc(s) : string_copy(s);
950 cb.at_is_special = FALSE;
951 switch (type & ~MCL_NOEXPAND)
953 case MCL_DOMAIN: cb.at_is_special = TRUE; /*FALLTHROUGH*/
954 case MCL_LOCALPART: cb.expand_setup = 0; break;
955 default: cb.expand_setup = sep > UCHAR_MAX ? 0 : -1; break;
957 cb.use_partial = TRUE;
958 cb.caseless = caseless;
959 if (valueptr) *valueptr = NULL;
960 return match_check_list(listptr, sep, anchorptr, &local_cache_bits,
961 check_string, &cb, type, s, valueptr);
966 /*************************************************
967 * Match address to single address-list item *
968 *************************************************/
970 /* This function matches an address to an item from an address list. It is
971 called from match_address_list() via match_check_list(). That is why most of
972 its arguments are in an indirect block.
975 arg the argument block (see below)
976 pattern the pattern to match
977 valueptr where to return a value
978 error for error messages (not used in this function; it never
981 The argument block contains:
982 address the start of the subject address; when called from retry.c
983 it may be *@domain if the local part isn't relevant
984 origaddress the original, un-case-forced address (not used here, but used
985 in match_check_list() when +caseful is encountered)
986 expand_setup controls setting up of $n variables
987 caseless TRUE for caseless local part matching
989 Returns: OK for a match
991 DEFER if a lookup deferred
995 check_address(void *arg, const uschar *pattern, const uschar **valueptr, uschar **error)
997 check_address_block *cb = (check_address_block *)arg;
998 check_string_block csb;
1001 unsigned int *null = NULL;
1002 const uschar *listptr;
1003 uschar *subject = cb->address;
1005 uschar *pdomain, *sdomain;
1007 error = error; /* Keep clever compilers from complaining */
1009 DEBUG(D_lists) debug_printf("address match test: subject=%s pattern=%s\n",
1012 /* Find the subject's domain */
1014 sdomain = Ustrrchr(subject, '@');
1016 /* The only case where a subject may not have a domain is if the subject is
1017 empty. Otherwise, a subject with no domain is a serious configuration error. */
1019 if (sdomain == NULL && *subject != 0)
1021 log_write(0, LOG_MAIN|LOG_PANIC, "no @ found in the subject of an "
1022 "address list match: subject=\"%s\" pattern=\"%s\"", subject, pattern);
1026 /* Handle a regular expression, which must match the entire incoming address.
1027 This may be the empty address. */
1029 if (*pattern == '^')
1030 return match_check_string(subject, pattern, cb->expand_setup, TRUE,
1031 cb->caseless, FALSE, NULL);
1033 /* Handle a pattern that is just a lookup. Skip over possible lookup names
1034 (letters, digits, hyphens). Skip over a possible * or *@ at the end. Then we
1035 must have a semicolon for it to be a lookup. */
1037 for (s = pattern; isalnum(*s) || *s == '-'; s++);
1041 /* If it is a straight lookup, do a lookup for the whole address. This may be
1042 the empty address. Partial matching doesn't make sense here, so we ignore it,
1043 but write a panic log entry. However, *@ matching will be honoured. */
1047 if (Ustrncmp(pattern, "partial-", 8) == 0)
1048 log_write(0, LOG_MAIN|LOG_PANIC, "partial matching is not applicable to "
1049 "whole-address lookups: ignored \"partial-\" in \"%s\"", pattern);
1050 return match_check_string(subject, pattern, -1, FALSE, cb->caseless, FALSE,
1054 /* For the remaining cases, an empty subject matches only an empty pattern,
1055 because other patterns expect to have a local part and a domain to match
1058 if (*subject == 0) return (*pattern == 0)? OK : FAIL;
1060 /* If the pattern starts with "@@" we have a split lookup, where the domain is
1061 looked up to obtain a list of local parts. If the subject's local part is just
1062 "*" (called from retry) the match always fails. */
1064 if (pattern[0] == '@' && pattern[1] == '@')
1068 uschar buffer[1024];
1070 if (sdomain == subject + 1 && *subject == '*') return FAIL;
1072 /* Loop for handling chains. The last item in any list may be of the form
1073 ">name" in order to chain on to another list. */
1075 for (const uschar * key = sdomain + 1; key && watchdog-- > 0; )
1079 if ((rc = match_check_string(key, pattern + 2, -1, TRUE, FALSE, FALSE,
1080 CUSS &list)) != OK) return rc;
1082 /* Check for chaining from the last item; set up the next key if one
1085 ss = Ustrrchr(list, ':');
1086 if (ss == NULL) ss = list; else ss++;
1087 while (isspace(*ss)) ss++;
1091 while (isspace(*ss)) ss++;
1092 key = string_copy(ss);
1096 /* Look up the local parts provided by the list; negation is permitted.
1097 If a local part has to begin with !, a regex can be used. */
1099 while ((ss = string_nextinlist(CUSS &list, &sep, buffer, sizeof(buffer))))
1106 while (isspace((*(++ss))));
1108 else local_yield = OK;
1111 rc = match_check_string(subject, ss, -1, TRUE, cb->caseless, FALSE,
1126 /* End of chain loop; panic if too many times */
1129 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Loop detected in lookup of "
1130 "local part of %s in %s", subject, pattern);
1132 /* Otherwise the local part check has failed, so the whole match
1139 /* We get here if the pattern is not a lookup or a regular expression. If it
1140 contains an @ there is both a local part and a domain. */
1142 pdomain = Ustrrchr(pattern, '@');
1143 if (pdomain != NULL)
1147 /* If the domain in the pattern is empty or one of the special cases [] or
1148 mx_{any,primary,secondary}, and the local part in the pattern ends in "@",
1149 we have a pattern of the form <something>@@, <something>@@[], or
1150 <something>@@mx_{any,primary,secondary}. These magic "domains" are
1151 automatically interpreted in match_check_string. We just need to arrange that
1152 the leading @ is included in the domain. */
1154 if (pdomain > pattern && pdomain[-1] == '@' &&
1156 Ustrcmp(pdomain+1, "[]") == 0 ||
1157 Ustrcmp(pdomain+1, "mx_any") == 0 ||
1158 Ustrcmp(pdomain+1, "mx_primary") == 0 ||
1159 Ustrcmp(pdomain+1, "mx_secondary") == 0))
1162 pllen = pdomain - pattern;
1163 sllen = sdomain - subject;
1165 /* Compare the local parts in the subject and the pattern */
1167 if (*pattern == '*')
1169 int cllen = pllen - 1;
1170 if (sllen < cllen) return FAIL;
1172 ? strncmpic(subject+sllen-cllen, pattern + 1, cllen) != 0
1173 : Ustrncmp(subject+sllen-cllen, pattern + 1, cllen) != 0)
1175 if (cb->expand_setup > 0)
1177 expand_nstring[cb->expand_setup] = subject;
1178 expand_nlength[cb->expand_setup] = sllen - cllen;
1184 if (sllen != pllen) return FAIL;
1186 ? strncmpic(subject, pattern, sllen) != 0
1187 : Ustrncmp(subject, pattern, sllen) != 0) return FAIL;
1191 /* If the local part matched, or was not being checked, check the domain using
1192 the generalized function, which supports file lookups (which may defer). The
1193 original code read as follows:
1195 return match_check_string(sdomain + 1,
1196 pdomain ? pdomain + 1 : pattern,
1197 cb->expand_setup + expand_inc, TRUE, cb->caseless, TRUE, NULL);
1199 This supported only literal domains and *.x.y patterns. In order to allow for
1200 named domain lists (so that you can right, for example, "senders=+xxxx"), it
1201 was changed to use the list scanning function. */
1203 csb.origsubject = sdomain + 1;
1204 csb.subject = cb->caseless ? string_copylc(sdomain+1) : string_copy(sdomain+1);
1205 csb.expand_setup = cb->expand_setup + expand_inc;
1206 csb.use_partial = TRUE;
1207 csb.caseless = cb->caseless;
1208 csb.at_is_special = TRUE;
1210 listptr = pdomain ? pdomain + 1 : pattern;
1211 if (valueptr) *valueptr = NULL;
1213 return match_check_list(
1214 &listptr, /* list of one item */
1215 UCHAR_MAX+1, /* impossible separator; single item */
1216 &domainlist_anchor, /* it's a domain list */
1217 &null, /* ptr to NULL means no caching */
1218 check_string, /* the function to do one test */
1219 &csb, /* its data */
1220 MCL_DOMAIN + MCL_NOEXPAND, /* domain list; don't expand */
1221 csb.subject, /* string for messages */
1222 valueptr); /* where to pass back lookup data */
1228 /*************************************************
1229 * Test whether address matches address list *
1230 *************************************************/
1232 /* This function is given an address and a list of things to match it against.
1233 The list may contain individual addresses, regular expressions, lookup
1234 specifications, and indirection via bare files. Negation is supported. The
1235 address to check can consist of just a domain, which will then match only
1236 domain items or items specified as *@domain.
1238 Domains are always lower cased before the match. Local parts are also lower
1239 cased unless "caseless" is false. The work of actually scanning the list is
1240 done by match_check_list(), with an appropriate block of arguments and a
1241 callback to check_address(). During caseless matching, it will recognize
1242 +caseful and revert to caseful matching.
1245 address address to test
1246 caseless TRUE to start in caseless state
1247 expand TRUE to allow list expansion
1248 listptr list to check against
1249 cache_bits points to cache bits for named address lists, or NULL
1250 expand_setup controls setting up of $n variables - passed through
1251 to check_address (q.v.)
1252 sep separator character for the list;
1253 may be 0 to get separator from the list;
1254 may be UCHAR_MAX+1 for one-item list
1255 valueptr where to return a lookup value, or NULL
1257 Returns: OK for a positive match, or end list after a negation;
1258 FAIL for a negative match, or end list after non-negation;
1259 DEFER if a lookup deferred
1263 match_address_list(const uschar *address, BOOL caseless, BOOL expand,
1264 const uschar **listptr, unsigned int *cache_bits, int expand_setup, int sep,
1265 const uschar **valueptr)
1267 check_address_block ab;
1268 unsigned int *local_cache_bits = cache_bits;
1271 /* RFC 2505 recommends that for spam checking, local parts should be caselessly
1272 compared. Therefore, Exim now forces the entire address into lower case here,
1273 provided that "caseless" is set. (It is FALSE for calls for matching rewriting
1274 patterns.) Otherwise just the domain is lower cases. A magic item "+caseful" in
1275 the list can be used to restore a caseful copy of the local part from the
1276 original address. */
1278 if ((len = Ustrlen(address)) > 255) len = 255;
1279 ab.address = string_copyn(address, len);
1281 for (uschar * p = ab.address + len - 1; p >= ab.address; p--)
1283 if (!caseless && *p == '@') break;
1287 /* If expand_setup is zero, we need to set up $0 to the whole thing, in
1288 case there is a match. Can't use the built-in facilities of match_check_string
1289 (via check_address), as we may just be calling that for part of the address
1292 if (expand_setup == 0)
1294 expand_nstring[0] = string_copy(address);
1295 expand_nlength[0] = Ustrlen(address);
1299 /* Set up the data to be passed ultimately to check_address. */
1301 ab.origaddress = address;
1302 /* ab.address is above */
1303 ab.expand_setup = expand_setup;
1304 ab.caseless = caseless;
1306 return match_check_list(listptr, sep, &addresslist_anchor, &local_cache_bits,
1307 check_address, &ab, MCL_ADDRESS + (expand? 0:MCL_NOEXPAND), address,
1311 /* Simpler version of match_address_list; always caseless, expanding,
1312 no cache bits, no value-return.
1315 address address to test
1316 listptr list to check against
1317 sep separator character for the list;
1318 may be 0 to get separator from the list;
1319 may be UCHAR_MAX+1 for one-item list
1321 Returns: OK for a positive match, or end list after a negation;
1322 FAIL for a negative match, or end list after non-negation;
1323 DEFER if a lookup deferred
1327 match_address_list_basic(const uschar *address, const uschar **listptr, int sep)
1329 return match_address_list(address, TRUE, TRUE, listptr, NULL, -1, sep, NULL);
1332 /* End of match.c */