Commit | Line | Data |
---|---|---|
059ec3d9 PH |
1 | /************************************************* |
2 | * Exim - an Internet mail transport agent * | |
3 | *************************************************/ | |
4 | ||
f9ba5e22 | 5 | /* Copyright (c) University of Cambridge 1995 - 2018 */ |
1e1ddfac | 6 | /* Copyright (c) The Exim Maintainers 2020 */ |
059ec3d9 PH |
7 | /* See the file NOTICE for conditions of use and distribution. */ |
8 | ||
9 | /* Functions for parsing addresses */ | |
10 | ||
11 | ||
12 | #include "exim.h" | |
13 | ||
14 | ||
15 | static uschar *last_comment_position; | |
16 | ||
17 | ||
18 | ||
19 | /* In stand-alone mode, provide a replacement for deliver_make_addr() | |
20 | and rewrite_address[_qualify]() so as to avoid having to drag in too much | |
21 | redundant apparatus. */ | |
22 | ||
23 | #ifdef STAND_ALONE | |
24 | ||
25 | address_item *deliver_make_addr(uschar *address, BOOL copy) | |
26 | { | |
f3ebb786 | 27 | address_item *addr = store_get(sizeof(address_item), FALSE); |
059ec3d9 PH |
28 | addr->next = NULL; |
29 | addr->parent = NULL; | |
30 | addr->address = address; | |
31 | return addr; | |
32 | } | |
33 | ||
34 | uschar *rewrite_address(uschar *recipient, BOOL dummy1, BOOL dummy2, rewrite_rule | |
35 | *dummy3, int dummy4) | |
36 | { | |
37 | return recipient; | |
38 | } | |
39 | ||
40 | uschar *rewrite_address_qualify(uschar *recipient, BOOL dummy1) | |
41 | { | |
42 | return recipient; | |
43 | } | |
44 | ||
45 | #endif | |
46 | ||
47 | ||
48 | ||
49 | ||
50 | /************************************************* | |
51 | * Find the end of an address * | |
52 | *************************************************/ | |
53 | ||
54 | /* Scan over a string looking for the termination of an address at a comma, | |
55 | or end of the string. It's the source-routed addresses which cause much pain | |
56 | here. Although Exim ignores source routes, it must recognize such addresses, so | |
57 | we cannot get rid of this logic. | |
58 | ||
59 | Argument: | |
60 | s pointer to the start of an address | |
61 | nl_ends if TRUE, '\n' terminates an address | |
62 | ||
63 | Returns: pointer past the end of the address | |
64 | (i.e. points to null or comma) | |
65 | */ | |
66 | ||
67 | uschar * | |
68 | parse_find_address_end(uschar *s, BOOL nl_ends) | |
69 | { | |
70 | BOOL source_routing = *s == '@'; | |
71 | int no_term = source_routing? 1 : 0; | |
72 | ||
73 | while (*s != 0 && (*s != ',' || no_term > 0) && (*s != '\n' || !nl_ends)) | |
74 | { | |
75 | /* Skip single quoted characters. Strictly these should not occur outside | |
76 | quoted strings in RFC 822 addresses, but they can in RFC 821 addresses. Pity | |
77 | about the lack of consistency, isn't it? */ | |
78 | ||
79 | if (*s == '\\' && s[1] != 0) s += 2; | |
80 | ||
81 | /* Skip quoted items that are not inside brackets. Note that | |
82 | quoted pairs are allowed inside quoted strings. */ | |
83 | ||
84 | else if (*s == '\"') | |
85 | { | |
86 | while (*(++s) != 0 && (*s != '\n' || !nl_ends)) | |
87 | { | |
88 | if (*s == '\\' && s[1] != 0) s++; | |
89 | else if (*s == '\"') { s++; break; } | |
90 | } | |
91 | } | |
92 | ||
93 | /* Skip comments, which may include nested brackets, but quotes | |
94 | are not recognized inside comments, though quoted pairs are. */ | |
95 | ||
96 | else if (*s == '(') | |
97 | { | |
98 | int level = 1; | |
99 | while (*(++s) != 0 && (*s != '\n' || !nl_ends)) | |
100 | { | |
101 | if (*s == '\\' && s[1] != 0) s++; | |
102 | else if (*s == '(') level++; | |
103 | else if (*s == ')' && --level <= 0) { s++; break; } | |
104 | } | |
105 | } | |
106 | ||
107 | /* Non-special character; just advance. Passing the colon in a source | |
108 | routed address means that any subsequent comma or colon may terminate unless | |
109 | inside angle brackets. */ | |
110 | ||
111 | else | |
112 | { | |
113 | if (*s == '<') | |
114 | { | |
115 | source_routing = s[1] == '@'; | |
116 | no_term = source_routing? 2 : 1; | |
117 | } | |
118 | else if (*s == '>') no_term--; | |
119 | else if (source_routing && *s == ':') no_term--; | |
120 | s++; | |
121 | } | |
122 | } | |
123 | ||
124 | return s; | |
125 | } | |
126 | ||
127 | ||
128 | ||
129 | /************************************************* | |
130 | * Find last @ in an address * | |
131 | *************************************************/ | |
132 | ||
133 | /* This function is used when we have something that may not qualified. If we | |
134 | know it's qualified, searching for the rightmost '@' is sufficient. Here we | |
135 | have to be a bit more clever than just a plain search, in order to handle | |
136 | unqualified local parts like "thing@thong" correctly. Since quotes may not | |
137 | legally be part of a domain name, we can give up on hitting the first quote | |
138 | when searching from the right. Now that the parsing also permits the RFC 821 | |
139 | form of address, where quoted-pairs are allowed in unquoted local parts, we | |
140 | must take care to handle that too. | |
141 | ||
142 | Argument: pointer to an address, possibly unqualified | |
143 | Returns: pointer to the last @ in an address, or NULL if none | |
144 | */ | |
145 | ||
146 | uschar * | |
147 | parse_find_at(uschar *s) | |
148 | { | |
149 | uschar *t = s + Ustrlen(s); | |
150 | while (--t >= s) | |
151 | { | |
152 | if (*t == '@') | |
153 | { | |
154 | int backslash_count = 0; | |
155 | uschar *tt = t - 1; | |
156 | while (tt > s && *tt-- == '\\') backslash_count++; | |
157 | if ((backslash_count & 1) == 0) return t; | |
158 | } | |
159 | else if (*t == '\"') return NULL; | |
160 | } | |
161 | return NULL; | |
162 | } | |
163 | ||
164 | ||
165 | ||
166 | ||
167 | /*************************************************************************** | |
168 | * In all the functions below that read a particular object type from * | |
169 | * the input, return the new value of the pointer s (the first argument), * | |
170 | * and put the object into the store pointed to by t (the second argument), * | |
171 | * adding a terminating zero. If no object is found, t will point to zero * | |
172 | * on return. * | |
173 | ***************************************************************************/ | |
174 | ||
175 | ||
176 | /************************************************* | |
177 | * Skip white space and comment * | |
178 | *************************************************/ | |
179 | ||
180 | /* Algorithm: | |
181 | (1) Skip spaces. | |
182 | (2) If uschar not '(', return. | |
183 | (3) Skip till matching ')', not counting any characters | |
184 | escaped with '\'. | |
185 | (4) Move past ')' and goto (1). | |
186 | ||
187 | The start of the last potential comment position is remembered to | |
188 | make it possible to ignore comments at the end of compound items. | |
189 | ||
190 | Argument: current character pointer | |
4c04137d | 191 | Returns: new character pointer |
059ec3d9 PH |
192 | */ |
193 | ||
194 | static uschar * | |
195 | skip_comment(uschar *s) | |
196 | { | |
197 | last_comment_position = s; | |
198 | while (*s) | |
199 | { | |
200 | int c, level; | |
137ae145 JH |
201 | |
202 | if (Uskip_whitespace(&s) != '(') break; | |
059ec3d9 | 203 | level = 1; |
137ae145 | 204 | while((c = *(++s))) |
059ec3d9 PH |
205 | { |
206 | if (c == '(') level++; | |
207 | else if (c == ')') { if (--level <= 0) { s++; break; } } | |
208 | else if (c == '\\' && s[1] != 0) s++; | |
209 | } | |
210 | } | |
211 | return s; | |
212 | } | |
213 | ||
214 | ||
215 | ||
216 | /************************************************* | |
217 | * Read a domain * | |
218 | *************************************************/ | |
219 | ||
220 | /* A domain is a sequence of subdomains, separated by dots. See comments below | |
221 | for detailed syntax of the subdomains. | |
222 | ||
223 | If allow_domain_literals is TRUE, a "domain" may also be an IP address enclosed | |
224 | in []. Make sure the output is set to the null string if there is a syntax | |
225 | error as well as if there is no domain at all. | |
226 | ||
227 | Arguments: | |
228 | s current character pointer | |
229 | t where to put the domain | |
230 | errorptr put error message here on failure (*t will be 0 on exit) | |
231 | ||
232 | Returns: new character pointer | |
233 | */ | |
234 | ||
235 | static uschar * | |
236 | read_domain(uschar *s, uschar *t, uschar **errorptr) | |
237 | { | |
238 | uschar *tt = t; | |
239 | s = skip_comment(s); | |
240 | ||
241 | /* Handle domain literals if permitted. An RFC 822 domain literal may contain | |
242 | any character except [ ] \, including linear white space, and may contain | |
243 | quoted characters. However, RFC 821 restricts literals to being dot-separated | |
244 | 3-digit numbers, and we make the obvious extension for IPv6. Go for a sequence | |
309bd837 PH |
245 | of digits, dots, hex digits, and colons here; later this will be checked for |
246 | being a syntactically valid IP address if it ever gets to a router. | |
059ec3d9 | 247 | |
309bd837 PH |
248 | Allow both the formal IPv6 form, with IPV6: at the start, and the informal form |
249 | without it, and accept IPV4: as well, 'cause someone will use it sooner or | |
250 | later. */ | |
059ec3d9 PH |
251 | |
252 | if (*s == '[') | |
253 | { | |
254 | *t++ = *s++; | |
255 | ||
059ec3d9 PH |
256 | if (strncmpic(s, US"IPv6:", 5) == 0 || strncmpic(s, US"IPv4:", 5) == 0) |
257 | { | |
258 | memcpy(t, s, 5); | |
259 | t += 5; | |
260 | s += 5; | |
261 | } | |
262 | while (*s == '.' || *s == ':' || isxdigit(*s)) *t++ = *s++; | |
263 | ||
059ec3d9 PH |
264 | if (*s == ']') *t++ = *s++; else |
265 | { | |
266 | *errorptr = US"malformed domain literal"; | |
267 | *tt = 0; | |
268 | } | |
269 | ||
270 | if (!allow_domain_literals) | |
271 | { | |
272 | *errorptr = US"domain literals not allowed"; | |
273 | *tt = 0; | |
274 | } | |
275 | *t = 0; | |
276 | return skip_comment(s); | |
277 | } | |
278 | ||
279 | /* Handle a proper domain, which is a sequence of dot-separated atoms. Remove | |
280 | trailing dots if strip_trailing_dot is set. A subdomain is an atom. | |
281 | ||
282 | An atom is a sequence of any characters except specials, space, and controls. | |
283 | The specials are ( ) < > @ , ; : \ " . [ and ]. This is the rule for RFC 822 | |
284 | and its successor (RFC 2822). However, RFC 821 and its successor (RFC 2821) is | |
285 | tighter, allowing only letters, digits, and hyphens, not starting with a | |
286 | hyphen. | |
287 | ||
288 | There used to be a global flag that got set when checking addresses that came | |
289 | in over SMTP and which should therefore should be checked according to the | |
290 | stricter rule. However, it seems silly to make the distinction, because I don't | |
291 | suppose anybody ever uses local domains that are 822-compliant and not | |
292 | 821-compliant. Furthermore, Exim now has additional data on the spool file line | |
293 | after an address (after "one_time" processing), and it makes use of a # | |
294 | character to delimit it. When I wrote that code, I forgot about this 822-domain | |
295 | stuff, and assumed # could never appear in a domain. | |
296 | ||
297 | So the old code is now cut out for Release 4.11 onwards, on 09-Aug-02. In a few | |
298 | years, when we are sure this isn't actually causing trouble, throw it away. | |
299 | ||
300 | March 2003: the story continues: There is a camp that is arguing for the use of | |
301 | UTF-8 in domain names as the way to internationalization, and other MTAs | |
302 | support this. Therefore, we now have a flag that permits the use of characters | |
303 | with values greater than 127, encoded in UTF-8, in subdomains, so that Exim can | |
304 | be used experimentally in this way. */ | |
305 | ||
306 | for (;;) | |
307 | { | |
308 | uschar *tsave = t; | |
309 | ||
310 | /********************* | |
311 | if (rfc821_domains) | |
312 | { | |
313 | if (*s != '-') while (isalnum(*s) || *s == '-') *t++ = *s++; | |
314 | } | |
315 | else | |
316 | while (!mac_iscntrl_or_special(*s)) *t++ = *s++; | |
317 | *********************/ | |
318 | ||
319 | if (*s != '-') | |
320 | { | |
321 | /* Only letters, digits, and hyphens */ | |
322 | ||
323 | if (!allow_utf8_domains) | |
324 | { | |
325 | while (isalnum(*s) || *s == '-') *t++ = *s++; | |
326 | } | |
327 | ||
328 | /* Permit legal UTF-8 characters to be included */ | |
329 | ||
330 | else for(;;) | |
331 | { | |
332 | int i, d; | |
333 | if (isalnum(*s) || *s == '-') /* legal ascii characters */ | |
334 | { | |
335 | *t++ = *s++; | |
336 | continue; | |
337 | } | |
338 | if ((*s & 0xc0) != 0xc0) break; /* not start of UTF-8 character */ | |
339 | d = *s << 2; | |
340 | for (i = 1; i < 6; i++) /* i is the number of additional bytes */ | |
341 | { | |
342 | if ((d & 0x80) == 0) break; | |
343 | d <<= 1; | |
344 | } | |
345 | if (i == 6) goto BAD_UTF8; /* invalid UTF-8 */ | |
346 | *t++ = *s++; /* leading UTF-8 byte */ | |
347 | while (i-- > 0) /* copy and check remainder */ | |
348 | { | |
349 | if ((*s & 0xc0) != 0x80) | |
350 | { | |
351 | BAD_UTF8: | |
352 | *errorptr = US"invalid UTF-8 byte sequence"; | |
353 | *tt = 0; | |
354 | return s; | |
355 | } | |
356 | *t++ = *s++; | |
357 | } | |
358 | } /* End of loop for UTF-8 character */ | |
359 | } /* End of subdomain */ | |
360 | ||
361 | s = skip_comment(s); | |
362 | *t = 0; | |
363 | ||
364 | if (t == tsave) /* empty component */ | |
365 | { | |
366 | if (strip_trailing_dot && t > tt && *s != '.') t[-1] = 0; else | |
367 | { | |
368 | *errorptr = US"domain missing or malformed"; | |
369 | *tt = 0; | |
370 | } | |
371 | return s; | |
372 | } | |
373 | ||
374 | if (*s != '.') break; | |
375 | *t++ = *s++; | |
376 | s = skip_comment(s); | |
377 | } | |
378 | ||
379 | return s; | |
380 | } | |
381 | ||
382 | ||
383 | ||
384 | /************************************************* | |
385 | * Read a local-part * | |
386 | *************************************************/ | |
387 | ||
388 | /* A local-part is a sequence of words, separated by periods. A null word | |
389 | between dots is not strictly allowed but apparently many mailers permit it, | |
390 | so, sigh, better be compatible. Even accept a trailing dot... | |
391 | ||
392 | A <word> is either a quoted string, or an <atom>, which is a sequence | |
393 | of any characters except specials, space, and controls. The specials are | |
394 | ( ) < > @ , ; : \ " . [ and ]. In RFC 822, a single quoted character, (a | |
395 | quoted-pair) is not allowed in a word. However, in RFC 821, it is permitted in | |
396 | the local part of an address. Rather than have separate parsing functions for | |
397 | the different cases, take the liberal attitude always. At least one MUA is | |
398 | happy to recognize this case; I don't know how many other programs do. | |
399 | ||
400 | Arguments: | |
401 | s current character pointer | |
402 | t where to put the local part | |
403 | error where to point error text | |
404 | allow_null TRUE if an empty local part is not an error | |
405 | ||
406 | Returns: new character pointer | |
407 | */ | |
408 | ||
409 | static uschar * | |
410 | read_local_part(uschar *s, uschar *t, uschar **error, BOOL allow_null) | |
411 | { | |
412 | uschar *tt = t; | |
413 | *error = NULL; | |
414 | for (;;) | |
415 | { | |
416 | int c; | |
417 | uschar *tsave = t; | |
418 | s = skip_comment(s); | |
419 | ||
420 | /* Handle a quoted string */ | |
421 | ||
422 | if (*s == '\"') | |
423 | { | |
424 | *t++ = '\"'; | |
db57e575 | 425 | while ((c = *++s) && c != '\"') |
059ec3d9 PH |
426 | { |
427 | *t++ = c; | |
db57e575 | 428 | if (c == '\\' && s[1]) *t++ = *++s; |
059ec3d9 PH |
429 | } |
430 | if (c == '\"') | |
431 | { | |
432 | s++; | |
433 | *t++ = '\"'; | |
434 | } | |
435 | else | |
436 | { | |
437 | *error = US"unmatched doublequote in local part"; | |
438 | return s; | |
439 | } | |
440 | } | |
441 | ||
442 | /* Handle an atom, but allow quoted pairs within it. */ | |
443 | ||
444 | else while (!mac_iscntrl_or_special(*s) || *s == '\\') | |
445 | { | |
446 | c = *t++ = *s++; | |
db57e575 | 447 | if (c == '\\' && *s) *t++ = *s++; |
059ec3d9 PH |
448 | } |
449 | ||
450 | /* Terminate the word and skip subsequent comment */ | |
451 | ||
452 | *t = 0; | |
453 | s = skip_comment(s); | |
454 | ||
455 | /* If we have read a null component at this point, give an error unless it is | |
456 | terminated by a dot - an extension to RFC 822 - or if it is the first | |
457 | component of the local part and an empty local part is permitted, in which | |
458 | case just return normally. */ | |
459 | ||
460 | if (t == tsave && *s != '.') | |
461 | { | |
462 | if (t == tt && !allow_null) | |
463 | *error = US"missing or malformed local part"; | |
464 | return s; | |
465 | } | |
466 | ||
467 | /* Anything other than a dot terminates the local part. Treat multiple dots | |
468 | as a single dot, as this seems to be a common extension. */ | |
469 | ||
470 | if (*s != '.') break; | |
471 | do { *t++ = *s++; } while (*s == '.'); | |
472 | } | |
473 | ||
474 | return s; | |
475 | } | |
476 | ||
477 | ||
478 | /************************************************* | |
479 | * Read route part of route-addr * | |
480 | *************************************************/ | |
481 | ||
482 | /* The pointer is at the initial "@" on entry. Return it following the | |
483 | terminating colon. Exim no longer supports the use of source routes, but it is | |
484 | required to accept the syntax. | |
485 | ||
486 | Arguments: | |
487 | s current character pointer | |
488 | t where to put the route | |
489 | errorptr where to put an error message | |
490 | ||
491 | Returns: new character pointer | |
492 | */ | |
493 | ||
494 | static uschar * | |
495 | read_route(uschar *s, uschar *t, uschar **errorptr) | |
496 | { | |
497 | BOOL commas = FALSE; | |
498 | *errorptr = NULL; | |
499 | ||
500 | while (*s == '@') | |
501 | { | |
502 | *t++ = '@'; | |
503 | s = read_domain(s+1, t, errorptr); | |
504 | if (*t == 0) return s; | |
505 | t += Ustrlen((const uschar *)t); | |
506 | if (*s != ',') break; | |
507 | *t++ = *s++; | |
508 | commas = TRUE; | |
509 | s = skip_comment(s); | |
510 | } | |
511 | ||
512 | if (*s == ':') *t++ = *s++; | |
513 | ||
514 | /* If there is no colon, and there were no commas, the most likely error | |
515 | is in fact a missing local part in the address rather than a missing colon | |
516 | after the route. */ | |
517 | ||
518 | else *errorptr = commas? | |
519 | US"colon expected after route list" : | |
520 | US"no local part"; | |
521 | ||
522 | /* Terminate the route and return */ | |
523 | ||
524 | *t = 0; | |
525 | return skip_comment(s); | |
526 | } | |
527 | ||
528 | ||
529 | ||
530 | /************************************************* | |
531 | * Read addr-spec * | |
532 | *************************************************/ | |
533 | ||
534 | /* Addr-spec is local-part@domain. We make the domain optional - | |
535 | the expected terminator for the whole thing is passed to check this. | |
536 | This function is called only when we know we have a route-addr. | |
537 | ||
538 | Arguments: | |
539 | s current character pointer | |
540 | t where to put the addr-spec | |
541 | term expected terminator (0 or >) | |
542 | errorptr where to put an error message | |
543 | domainptr set to point to the start of the domain | |
544 | ||
545 | Returns: new character pointer | |
546 | */ | |
547 | ||
548 | static uschar * | |
549 | read_addr_spec(uschar *s, uschar *t, int term, uschar **errorptr, | |
550 | uschar **domainptr) | |
551 | { | |
552 | s = read_local_part(s, t, errorptr, FALSE); | |
553 | if (*errorptr == NULL) | |
059ec3d9 | 554 | if (*s != term) |
059ec3d9 PH |
555 | if (*s != '@') |
556 | *errorptr = string_sprintf("\"@\" or \".\" expected after \"%s\"", t); | |
557 | else | |
558 | { | |
559 | t += Ustrlen((const uschar *)t); | |
560 | *t++ = *s++; | |
561 | *domainptr = t; | |
562 | s = read_domain(s, t, errorptr); | |
563 | } | |
059ec3d9 PH |
564 | return s; |
565 | } | |
566 | ||
567 | ||
568 | ||
569 | /************************************************* | |
570 | * Extract operative address * | |
571 | *************************************************/ | |
572 | ||
573 | /* This function extracts an operative address from a full RFC822 mailbox and | |
574 | returns it in a piece of dynamic store. We take the easy way and get a piece | |
575 | of store the same size as the input, and then copy into it whatever is | |
576 | necessary. If we cannot find a valid address (syntax error), return NULL, and | |
577 | point the error pointer to the reason. The arguments "start" and "end" are used | |
578 | to return the offsets of the first and one past the last characters in the | |
579 | original mailbox of the address that has been extracted, to aid in re-writing. | |
580 | The argument "domain" is set to point to the first character after "@" in the | |
581 | final part of the returned address, or zero if there is no @. | |
582 | ||
583 | Exim no longer supports the use of source routed addresses (those of the form | |
584 | @domain,...:route_addr). It recognizes the syntax, but collapses such addresses | |
585 | down to their final components. Formerly, collapse_source_routes had to be set | |
586 | to achieve this effect. RFC 1123 allows collapsing with MAY, while the revision | |
587 | of RFC 821 had increased this to SHOULD, so I've gone for it, because it makes | |
588 | a lot of code elsewhere in Exim much simpler. | |
589 | ||
590 | There are some special fudges here for handling RFC 822 group address notation | |
591 | which may appear in certain headers. If the flag parse_allow_group is set | |
592 | TRUE and parse_found_group is FALSE when this function is called, an address | |
593 | which is the start of a group (i.e. preceded by a phrase and a colon) is | |
594 | recognized; the phrase is ignored and the flag parse_found_group is set. If | |
1eccaa59 PH |
595 | this flag is TRUE at the end of an address, and if an extraneous semicolon is |
596 | found, it is ignored and the flag is cleared. | |
597 | ||
598 | This logic is used only when scanning through addresses in headers, either to | |
599 | fulfil the -t option, or for rewriting, or for checking header syntax. Because | |
600 | the group "state" has to be remembered between multiple calls of this function, | |
601 | the variables parse_{allow,found}_group are global. It is important to ensure | |
602 | that they are reset to FALSE at the end of scanning a header's list of | |
603 | addresses. | |
059ec3d9 PH |
604 | |
605 | Arguments: | |
606 | mailbox points to the RFC822 mailbox | |
607 | errorptr where to point an error message | |
608 | start set to start offset in mailbox | |
609 | end set to end offset in mailbox | |
610 | domain set to domain offset in result, or 0 if no domain present | |
611 | allow_null allow <> if TRUE | |
612 | ||
613 | Returns: points to the extracted address, or NULL on error | |
614 | */ | |
615 | ||
616 | #define FAILED(s) { *errorptr = s; goto PARSE_FAILED; } | |
617 | ||
618 | uschar * | |
619 | parse_extract_address(uschar *mailbox, uschar **errorptr, int *start, int *end, | |
620 | int *domain, BOOL allow_null) | |
621 | { | |
f3ebb786 | 622 | uschar *yield = store_get(Ustrlen(mailbox) + 1, is_tainted(mailbox)); |
059ec3d9 | 623 | uschar *startptr, *endptr; |
5903c6ff JH |
624 | uschar *s = US mailbox; |
625 | uschar *t = US yield; | |
059ec3d9 PH |
626 | |
627 | *domain = 0; | |
628 | ||
629 | /* At the start of the string we expect either an addr-spec or a phrase | |
630 | preceding a <route-addr>. If groups are allowed, we might also find a phrase | |
631 | preceding a colon and an address. If we find an initial word followed by | |
632 | a dot, strict interpretation of the RFC would cause it to be taken | |
633 | as the start of an addr-spec. However, many mailers break the rules | |
634 | and use addresses of the form "a.n.other <ano@somewhere>" and so we | |
635 | allow this case. */ | |
636 | ||
637 | RESTART: /* Come back here after passing a group name */ | |
638 | ||
639 | s = skip_comment(s); | |
640 | startptr = s; /* In case addr-spec */ | |
641 | s = read_local_part(s, t, errorptr, TRUE); /* Dot separated words */ | |
db57e575 | 642 | if (*errorptr) goto PARSE_FAILED; |
059ec3d9 PH |
643 | |
644 | /* If the terminator is neither < nor @ then the format of the address | |
645 | must either be a bare local-part (we are now at the end), or a phrase | |
646 | followed by a route-addr (more words must follow). */ | |
647 | ||
648 | if (*s != '@' && *s != '<') | |
649 | { | |
650 | if (*s == 0 || *s == ';') | |
651 | { | |
652 | if (*t == 0) FAILED(US"empty address"); | |
653 | endptr = last_comment_position; | |
654 | goto PARSE_SUCCEEDED; /* Bare local part */ | |
655 | } | |
656 | ||
657 | /* Expect phrase route-addr, or phrase : if groups permitted, but allow | |
658 | dots in the phrase; complete the loop only when '<' or ':' is encountered - | |
659 | end of string will produce a null local_part and therefore fail. We don't | |
660 | need to keep updating t, as the phrase isn't to be kept. */ | |
661 | ||
8768d548 | 662 | while (*s != '<' && (!f.parse_allow_group || *s != ':')) |
059ec3d9 PH |
663 | { |
664 | s = read_local_part(s, t, errorptr, FALSE); | |
b6040544 | 665 | if (*errorptr) |
059ec3d9 PH |
666 | { |
667 | *errorptr = string_sprintf("%s (expected word or \"<\")", *errorptr); | |
668 | goto PARSE_FAILED; | |
669 | } | |
670 | } | |
671 | ||
672 | if (*s == ':') | |
673 | { | |
8768d548 JH |
674 | f.parse_found_group = TRUE; |
675 | f.parse_allow_group = FALSE; | |
059ec3d9 PH |
676 | s++; |
677 | goto RESTART; | |
678 | } | |
679 | ||
680 | /* Assert *s == '<' */ | |
681 | } | |
682 | ||
683 | /* At this point the next character is either '@' or '<'. If it is '@', only a | |
684 | single local-part has previously been read. An angle bracket signifies the | |
685 | start of an <addr-spec>. Throw away anything we have saved so far before | |
686 | processing it. Note that this is "if" rather than "else if" because it's also | |
687 | used after reading a preceding phrase. | |
688 | ||
689 | There are a lot of broken sendmails out there that put additional pairs of <> | |
b6040544 JH |
690 | round <route-addr>s. If strip_excess_angle_brackets is set, allow a limited |
691 | number of them, as long as they match. */ | |
059ec3d9 PH |
692 | |
693 | if (*s == '<') | |
694 | { | |
695 | uschar *domainptr = yield; | |
696 | BOOL source_routed = FALSE; | |
697 | int bracket_count = 1; | |
698 | ||
699 | s++; | |
b6040544 JH |
700 | if (strip_excess_angle_brackets) while (*s == '<') |
701 | { | |
702 | if(bracket_count++ > 5) FAILED(US"angle-brackets nested too deep"); | |
703 | s++; | |
704 | } | |
059ec3d9 PH |
705 | |
706 | t = yield; | |
707 | startptr = s; | |
708 | s = skip_comment(s); | |
709 | ||
710 | /* Read an optional series of routes, each of which is a domain. They | |
711 | are separated by commas and terminated by a colon. However, we totally ignore | |
712 | such routes (RFC 1123 says we MAY, and the revision of RFC 821 says we | |
713 | SHOULD). */ | |
714 | ||
715 | if (*s == '@') | |
716 | { | |
717 | s = read_route(s, t, errorptr); | |
b6040544 | 718 | if (*errorptr) goto PARSE_FAILED; |
059ec3d9 PH |
719 | *t = 0; /* Ensure route is ignored - probably overkill */ |
720 | source_routed = TRUE; | |
721 | } | |
722 | ||
723 | /* Now an addr-spec, terminated by '>'. If there is no preceding route, | |
724 | we must allow an empty addr-spec if allow_null is TRUE, to permit the | |
725 | address "<>" in some circumstances. A source-routed address MUST have | |
726 | a domain in the final part. */ | |
727 | ||
728 | if (allow_null && !source_routed && *s == '>') | |
729 | { | |
730 | *t = 0; | |
731 | *errorptr = NULL; | |
732 | } | |
733 | else | |
734 | { | |
735 | s = read_addr_spec(s, t, '>', errorptr, &domainptr); | |
b6040544 | 736 | if (*errorptr) goto PARSE_FAILED; |
059ec3d9 PH |
737 | *domain = domainptr - yield; |
738 | if (source_routed && *domain == 0) | |
739 | FAILED(US"domain missing in source-routed address"); | |
740 | } | |
741 | ||
742 | endptr = s; | |
743 | if (*errorptr != NULL) goto PARSE_FAILED; | |
744 | while (bracket_count-- > 0) if (*s++ != '>') | |
745 | { | |
b6040544 JH |
746 | *errorptr = s[-1] == 0 |
747 | ? US"'>' missing at end of address" | |
748 | : string_sprintf("malformed address: %.32s may not follow %.*s", | |
bb07bcd3 | 749 | s-1, (int)(s - US mailbox - 1), mailbox); |
059ec3d9 PH |
750 | goto PARSE_FAILED; |
751 | } | |
752 | ||
753 | s = skip_comment(s); | |
754 | } | |
755 | ||
756 | /* Hitting '@' after the first local-part means we have definitely got an | |
757 | addr-spec, on a strict reading of the RFC, and the rest of the string | |
758 | should be the domain. However, for flexibility we allow for a route-address | |
759 | not enclosed in <> as well, which is indicated by an empty first local | |
760 | part preceding '@'. The source routing is, however, ignored. */ | |
761 | ||
762 | else if (*t == 0) | |
763 | { | |
764 | uschar *domainptr = yield; | |
765 | s = read_route(s, t, errorptr); | |
766 | if (*errorptr != NULL) goto PARSE_FAILED; | |
767 | *t = 0; /* Ensure route is ignored - probably overkill */ | |
768 | s = read_addr_spec(s, t, 0, errorptr, &domainptr); | |
769 | if (*errorptr != NULL) goto PARSE_FAILED; | |
770 | *domain = domainptr - yield; | |
771 | endptr = last_comment_position; | |
772 | if (*domain == 0) FAILED(US"domain missing in source-routed address"); | |
773 | } | |
774 | ||
775 | /* This is the strict case of local-part@domain. */ | |
776 | ||
777 | else | |
778 | { | |
779 | t += Ustrlen((const uschar *)t); | |
780 | *t++ = *s++; | |
781 | *domain = t - yield; | |
782 | s = read_domain(s, t, errorptr); | |
783 | if (*t == 0) goto PARSE_FAILED; | |
784 | endptr = last_comment_position; | |
785 | } | |
786 | ||
787 | /* Use goto to get here from the bare local part case. Arrive by falling | |
788 | through for other cases. Endptr may have been moved over whitespace, so | |
789 | move it back past white space if necessary. */ | |
790 | ||
791 | PARSE_SUCCEEDED: | |
792 | if (*s != 0) | |
793 | { | |
8768d548 | 794 | if (f.parse_found_group && *s == ';') |
059ec3d9 | 795 | { |
8768d548 JH |
796 | f.parse_found_group = FALSE; |
797 | f.parse_allow_group = TRUE; | |
059ec3d9 PH |
798 | } |
799 | else | |
800 | { | |
7ade712c | 801 | *errorptr = string_sprintf("malformed address: %.32s may not follow %.*s", |
bb07bcd3 | 802 | s, (int)(s - US mailbox), mailbox); |
059ec3d9 PH |
803 | goto PARSE_FAILED; |
804 | } | |
805 | } | |
5903c6ff | 806 | *start = startptr - US mailbox; /* Return offsets */ |
059ec3d9 | 807 | while (isspace(endptr[-1])) endptr--; |
5903c6ff | 808 | *end = endptr - US mailbox; |
059ec3d9 PH |
809 | |
810 | /* Although this code has no limitation on the length of address extracted, | |
811 | other parts of Exim may have limits, and in any case, RFC 2821 limits local | |
812 | parts to 64 and domains to 255, so we do a check here, giving an error if the | |
813 | address is ridiculously long. */ | |
814 | ||
815 | if (*end - *start > ADDRESS_MAXLENGTH) | |
816 | { | |
817 | *errorptr = string_sprintf("address is ridiculously long: %.64s...", yield); | |
818 | return NULL; | |
819 | } | |
820 | ||
9d4319df | 821 | return yield; |
059ec3d9 PH |
822 | |
823 | /* Use goto (via the macro FAILED) to get to here from a variety of places. | |
824 | We might have an empty address in a group - the caller can choose to ignore | |
825 | this. We must, however, keep the flags correct. */ | |
826 | ||
827 | PARSE_FAILED: | |
8768d548 | 828 | if (f.parse_found_group && *s == ';') |
059ec3d9 | 829 | { |
8768d548 JH |
830 | f.parse_found_group = FALSE; |
831 | f.parse_allow_group = TRUE; | |
059ec3d9 PH |
832 | } |
833 | return NULL; | |
834 | } | |
835 | ||
836 | #undef FAILED | |
837 | ||
838 | ||
839 | ||
840 | /************************************************* | |
841 | * Quote according to RFC 2047 * | |
842 | *************************************************/ | |
843 | ||
844 | /* This function is used for quoting text in headers according to RFC 2047. | |
845 | If the only characters that strictly need quoting are spaces, we return the | |
846 | original string, unmodified. If a quoted string is too long for the buffer, it | |
847 | is truncated. (This shouldn't happen: this is normally handling short strings.) | |
848 | ||
8e669ac1 PH |
849 | Hmmph. As always, things get perverted for other uses. This function was |
850 | originally for the "phrase" part of addresses. Now it is being used for much | |
851 | longer texts in ACLs and via the ${rfc2047: expansion item. This means we have | |
14702f5b PH |
852 | to check for overlong "encoded-word"s and split them. November 2004. |
853 | ||
059ec3d9 PH |
854 | Arguments: |
855 | string the string to quote - already checked to contain non-printing | |
856 | chars | |
857 | len the length of the string | |
858 | charset the name of the character set; NULL => iso-8859-1 | |
859 | buffer the buffer to put the answer in | |
860 | buffer_size the size of the buffer | |
46218253 PH |
861 | fold if TRUE, a newline is inserted before the separating space when |
862 | more than one encoded-word is generated | |
059ec3d9 PH |
863 | |
864 | Returns: pointer to the original string, if no quoting needed, or | |
865 | pointer to buffer containing the quoted string, or | |
866 | a pointer to "String too long" if the buffer can't even hold | |
867 | the introduction | |
868 | */ | |
869 | ||
55414b25 JH |
870 | const uschar * |
871 | parse_quote_2047(const uschar *string, int len, uschar *charset, uschar *buffer, | |
46218253 | 872 | int buffer_size, BOOL fold) |
059ec3d9 | 873 | { |
55414b25 | 874 | const uschar *s = string; |
14702f5b PH |
875 | uschar *p, *t; |
876 | int hlen; | |
059ec3d9 | 877 | BOOL coded = FALSE; |
86ae49a6 | 878 | BOOL first_byte = FALSE; |
059ec3d9 | 879 | |
94759fce | 880 | if (!charset) charset = US"iso-8859-1"; |
059ec3d9 PH |
881 | |
882 | /* We don't expect this to fail! */ | |
883 | ||
884 | if (!string_format(buffer, buffer_size, "=?%s?Q?", charset)) | |
885 | return US"String too long"; | |
886 | ||
14702f5b PH |
887 | hlen = Ustrlen(buffer); |
888 | t = buffer + hlen; | |
889 | p = buffer; | |
890 | ||
059ec3d9 PH |
891 | for (; len > 0; len--) |
892 | { | |
893 | int ch = *s++; | |
14702f5b | 894 | if (t > buffer + buffer_size - hlen - 8) break; |
8e669ac1 | 895 | |
86ae49a6 | 896 | if ((t - p > 67) && !first_byte) |
14702f5b PH |
897 | { |
898 | *t++ = '?'; | |
899 | *t++ = '='; | |
46218253 | 900 | if (fold) *t++ = '\n'; |
14702f5b PH |
901 | *t++ = ' '; |
902 | p = t; | |
903 | Ustrncpy(p, buffer, hlen); | |
904 | t += hlen; | |
8e669ac1 PH |
905 | } |
906 | ||
059ec3d9 PH |
907 | if (ch < 33 || ch > 126 || |
908 | Ustrchr("?=()<>@,;:\\\".[]_", ch) != NULL) | |
909 | { | |
86ae49a6 NM |
910 | if (ch == ' ') |
911 | { | |
912 | *t++ = '_'; | |
913 | first_byte = FALSE; | |
94431adb | 914 | } |
86ae49a6 | 915 | else |
059ec3d9 | 916 | { |
5976eb99 | 917 | t += sprintf(CS t, "=%02X", ch); |
059ec3d9 | 918 | coded = TRUE; |
86ae49a6 | 919 | first_byte = !first_byte; |
059ec3d9 PH |
920 | } |
921 | } | |
86ae49a6 | 922 | else { *t++ = ch; first_byte = FALSE; } |
059ec3d9 | 923 | } |
8e669ac1 | 924 | |
14702f5b | 925 | *t++ = '?'; |
8e669ac1 | 926 | *t++ = '='; |
14702f5b | 927 | *t = 0; |
8e669ac1 | 928 | |
94759fce | 929 | return coded ? buffer : string; |
059ec3d9 PH |
930 | } |
931 | ||
932 | ||
933 | ||
934 | ||
935 | /************************************************* | |
936 | * Fix up an RFC 822 "phrase" * | |
937 | *************************************************/ | |
938 | ||
939 | /* This function is called to repair any syntactic defects in the "phrase" part | |
940 | of an RFC822 address. In particular, it is applied to the user's name as read | |
941 | from the passwd file when accepting a local message, and to the data from the | |
942 | -F option. | |
943 | ||
944 | If the string contains existing quoted strings or comments containing | |
945 | freestanding quotes, then we just quote those bits that need quoting - | |
946 | otherwise it would get awfully messy and probably not look good. If not, we | |
947 | quote the whole thing if necessary. Thus | |
948 | ||
949 | John Q. Smith => "John Q. Smith" | |
950 | John "Jack" Smith => John "Jack" Smith | |
951 | John "Jack" Q. Smith => John "Jack" "Q." Smith | |
952 | John (Jack) Q. Smith => "John (Jack) Q. Smith" | |
953 | John ("Jack") Q. Smith => John ("Jack") "Q." Smith | |
954 | but | |
955 | John (\"Jack\") Q. Smith => "John (\"Jack\") Q. Smith" | |
956 | ||
957 | Sheesh! This is tedious code. It is a great pity that the syntax of RFC822 is | |
958 | the way it is... | |
959 | ||
960 | August 2000: Additional code added: | |
961 | ||
962 | Previously, non-printing characters were turned into question marks, which do | |
963 | not need to be quoted. | |
964 | ||
965 | Now, a different tactic is used if there are any non-printing ASCII | |
966 | characters. The encoding method from RFC 2047 is used, assuming iso-8859-1 as | |
967 | the character set. | |
968 | ||
969 | We *could* use this for all cases, getting rid of the messy original code, | |
970 | but leave it for now. It would complicate simple cases like "John Q. Smith". | |
971 | ||
972 | The result is passed back in the buffer; it is usually going to be added to | |
973 | some other string. In order to be sure there is going to be no overflow, | |
974 | restrict the length of the input to 1/4 of the buffer size - this allows for | |
975 | every single character to be quoted or encoded without overflowing, and that | |
976 | wouldn't happen because of amalgamation. If the phrase is too long, return a | |
977 | fixed string. | |
978 | ||
979 | Arguments: | |
980 | phrase an RFC822 phrase | |
981 | len the length of the phrase | |
982 | buffer a buffer to put the result in | |
983 | buffer_size the size of the buffer | |
984 | ||
985 | Returns: the fixed RFC822 phrase | |
986 | */ | |
987 | ||
55414b25 JH |
988 | const uschar * |
989 | parse_fix_phrase(const uschar *phrase, int len, uschar *buffer, int buffer_size) | |
059ec3d9 PH |
990 | { |
991 | int ch, i; | |
992 | BOOL quoted = FALSE; | |
55414b25 JH |
993 | const uschar *s, *end; |
994 | uschar *t, *yield; | |
059ec3d9 PH |
995 | |
996 | while (len > 0 && isspace(*phrase)) { phrase++; len--; } | |
997 | if (len > buffer_size/4) return US"Name too long"; | |
998 | ||
999 | /* See if there are any non-printing characters, and if so, use the RFC 2047 | |
1000 | encoding for the whole thing. */ | |
1001 | ||
1002 | for (i = 0, s = phrase; i < len; i++, s++) | |
1003 | if ((*s < 32 && *s != '\t') || *s > 126) break; | |
1004 | ||
1005 | if (i < len) return parse_quote_2047(phrase, len, headers_charset, buffer, | |
46218253 | 1006 | buffer_size, FALSE); |
059ec3d9 PH |
1007 | |
1008 | /* No non-printers; use the RFC 822 quoting rules */ | |
1009 | ||
1010 | s = phrase; | |
1011 | end = s + len; | |
1012 | yield = t = buffer + 1; | |
1013 | ||
1014 | while (s < end) | |
1015 | { | |
1016 | ch = *s++; | |
1017 | ||
1018 | /* Copy over quoted strings, remembering we encountered one */ | |
1019 | ||
1020 | if (ch == '\"') | |
1021 | { | |
1022 | *t++ = '\"'; | |
1023 | while (s < end && (ch = *s++) != '\"') | |
1024 | { | |
1025 | *t++ = ch; | |
1026 | if (ch == '\\' && s < end) *t++ = *s++; | |
1027 | } | |
1028 | *t++ = '\"'; | |
1029 | if (s >= end) break; | |
1030 | quoted = TRUE; | |
1031 | } | |
1032 | ||
1033 | /* Copy over comments, noting if they contain freestanding quote | |
1034 | characters */ | |
1035 | ||
1036 | else if (ch == '(') | |
1037 | { | |
1038 | int level = 1; | |
1039 | *t++ = '('; | |
1040 | while (s < end) | |
1041 | { | |
1042 | ch = *s++; | |
1043 | *t++ = ch; | |
1044 | if (ch == '(') level++; | |
1045 | else if (ch == ')') { if (--level <= 0) break; } | |
1046 | else if (ch == '\\' && s < end) *t++ = *s++ & 127; | |
1047 | else if (ch == '\"') quoted = TRUE; | |
1048 | } | |
1049 | if (ch == 0) | |
1050 | { | |
1051 | while (level--) *t++ = ')'; | |
1052 | break; | |
1053 | } | |
1054 | } | |
1055 | ||
1056 | /* Handle special characters that need to be quoted */ | |
1057 | ||
1058 | else if (Ustrchr(")<>@,;:\\.[]", ch) != NULL) | |
1059 | { | |
1060 | /* If hit previous quotes just make one quoted "word" */ | |
1061 | ||
1062 | if (quoted) | |
1063 | { | |
1064 | uschar *tt = t++; | |
1065 | while (*(--tt) != ' ' && *tt != '\"' && *tt != ')') tt[1] = *tt; | |
1066 | tt[1] = '\"'; | |
1067 | *t++ = ch; | |
1068 | while (s < end) | |
1069 | { | |
1070 | ch = *s++; | |
1071 | if (ch == ' ' || ch == '\"') { s--; break; } else *t++ = ch; | |
1072 | } | |
1073 | *t++ = '\"'; | |
1074 | } | |
1075 | ||
1076 | /* Else quote the whole string so far, and the rest up to any following | |
1077 | quotes. We must treat anything following a backslash as a literal. */ | |
1078 | ||
1079 | else | |
1080 | { | |
1081 | BOOL escaped = (ch == '\\'); | |
1082 | *(--yield) = '\"'; | |
1083 | *t++ = ch; | |
1084 | ||
1085 | /* Now look for the end or a quote */ | |
1086 | ||
1087 | while (s < end) | |
1088 | { | |
1089 | ch = *s++; | |
1090 | ||
1091 | /* Handle escaped pairs */ | |
1092 | ||
1093 | if (escaped) | |
1094 | { | |
1095 | *t++ = ch; | |
1096 | escaped = FALSE; | |
1097 | } | |
1098 | ||
1099 | else if (ch == '\\') | |
1100 | { | |
1101 | *t++ = ch; | |
1102 | escaped = TRUE; | |
1103 | } | |
1104 | ||
1105 | /* If hit subsequent quotes, insert our quote before any trailing | |
1106 | spaces and back up to re-handle the quote in the outer loop. */ | |
1107 | ||
1108 | else if (ch == '\"') | |
1109 | { | |
1110 | int count = 0; | |
1111 | while (t[-1] == ' ') { t--; count++; } | |
1112 | *t++ = '\"'; | |
1113 | while (count-- > 0) *t++ = ' '; | |
1114 | s--; | |
1115 | break; | |
1116 | } | |
1117 | ||
1118 | /* If hit a subsequent comment, check it for unescaped quotes, | |
1119 | and if so, end our quote before it. */ | |
1120 | ||
1121 | else if (ch == '(') | |
1122 | { | |
55414b25 | 1123 | const uschar *ss = s; /* uschar after '(' */ |
059ec3d9 PH |
1124 | int level = 1; |
1125 | while(ss < end) | |
1126 | { | |
1127 | ch = *ss++; | |
1128 | if (ch == '(') level++; | |
1129 | else if (ch == ')') { if (--level <= 0) break; } | |
1130 | else if (ch == '\\' && ss+1 < end) ss++; | |
1131 | else if (ch == '\"') { quoted = TRUE; break; } | |
1132 | } | |
1133 | ||
1134 | /* Comment contains unescaped quotes; end our quote before | |
1135 | the start of the comment. */ | |
1136 | ||
1137 | if (quoted) | |
1138 | { | |
1139 | int count = 0; | |
1140 | while (t[-1] == ' ') { t--; count++; } | |
1141 | *t++ = '\"'; | |
1142 | while (count-- > 0) *t++ = ' '; | |
1143 | break; | |
1144 | } | |
1145 | ||
1146 | /* Comment does not contain unescaped quotes; include it in | |
1147 | our quote. */ | |
1148 | ||
1149 | else | |
1150 | { | |
1151 | if (ss >= end) ss--; | |
1152 | *t++ = '('; | |
1153 | Ustrncpy(t, s, ss-s); | |
1154 | t += ss-s; | |
1155 | s = ss; | |
1156 | } | |
1157 | } | |
1158 | ||
1159 | /* Not a comment or quote; include this character in our quotes. */ | |
1160 | ||
1161 | else *t++ = ch; | |
1162 | } | |
1163 | } | |
1164 | ||
1165 | /* Add a final quote if we hit the end of the string. */ | |
1166 | ||
1167 | if (s >= end) *t++ = '\"'; | |
1168 | } | |
1169 | ||
1170 | /* Non-special character; just copy it over */ | |
1171 | ||
1172 | else *t++ = ch; | |
1173 | } | |
1174 | ||
1175 | *t = 0; | |
1176 | return yield; | |
1177 | } | |
1178 | ||
1179 | ||
1180 | /************************************************* | |
1181 | * Extract addresses from a list * | |
1182 | *************************************************/ | |
1183 | ||
1184 | /* This function is called by the redirect router to scan a string containing a | |
1185 | list of addresses separated by commas (with optional white space) or by | |
1186 | newlines, and to generate a chain of address items from them. In other words, | |
1187 | to unpick data from an alias or .forward file. | |
1188 | ||
1189 | The SunOS5 documentation for alias files is not very clear on the syntax; it | |
1190 | does not say that either a comma or a newline can be used for separation. | |
1191 | However, that is the way Smail does it, so we follow suit. | |
1192 | ||
1193 | If a # character is encountered in a white space position, then characters from | |
1194 | there to the next newline are skipped. | |
1195 | ||
1196 | If an unqualified address begins with '\', just skip that character. This gives | |
1197 | compatibility with Sendmail's use of \ to prevent looping. Exim has its own | |
1198 | loop prevention scheme which handles other cases too - see the code in | |
1199 | route_address(). | |
1200 | ||
1201 | An "address" can be a specification of a file or a pipe; the latter may often | |
1202 | need to be quoted because it may contain spaces, but we don't want to retain | |
1203 | the quotes. Quotes may appear in normal addresses too, and should be retained. | |
1204 | We can distinguish between these cases, because in addresses, quotes are used | |
1205 | only for parts of the address, not the whole thing. Therefore, we remove quotes | |
1206 | from items when they entirely enclose them, but not otherwise. | |
1207 | ||
1208 | An "address" can also be of the form :include:pathname to include a list of | |
1209 | addresses contained in the specified file. | |
1210 | ||
1211 | Any unqualified addresses are qualified with and rewritten if necessary, via | |
1212 | the rewrite_address() function. | |
1213 | ||
1214 | Arguments: | |
1215 | s the list of addresses (typically a complete | |
1216 | .forward file or a list of entries in an alias file) | |
1217 | options option bits for permitting or denying various special cases; | |
1218 | not all bits are relevant here - some are for filter | |
1219 | files; those we use here are: | |
1220 | RDO_DEFER | |
1221 | RDO_FREEZE | |
1222 | RDO_FAIL | |
1223 | RDO_BLACKHOLE | |
1224 | RDO_REWRITE | |
1225 | RDO_INCLUDE | |
1226 | anchor where to hang the chain of newly-created addresses. This | |
1227 | should be initialized to NULL. | |
1228 | error where to return an error text | |
1229 | incoming domain domain of the incoming address; used to qualify unqualified | |
1230 | local parts preceded by \ | |
1231 | directory if NULL, no checks are done on :include: files | |
1232 | otherwise, included file names must start with the given | |
1233 | directory | |
1234 | syntax_errors if not NULL, it carries on after syntax errors in addresses, | |
1235 | building up a list of errors as error blocks chained on | |
1236 | here. | |
1237 | ||
1238 | Returns: FF_DELIVERED addresses extracted | |
1239 | FF_NOTDELIVERED no addresses extracted, but no errors | |
1240 | FF_BLACKHOLE :blackhole: | |
1241 | FF_DEFER :defer: | |
1242 | FF_FAIL :fail: | |
1243 | FF_INCLUDEFAIL some problem with :include:; *error set | |
1244 | FF_ERROR other problems; *error is set | |
1245 | */ | |
1246 | ||
1247 | int | |
1248 | parse_forward_list(uschar *s, int options, address_item **anchor, | |
55414b25 | 1249 | uschar **error, const uschar *incoming_domain, uschar *directory, |
059ec3d9 PH |
1250 | error_block **syntax_errors) |
1251 | { | |
1252 | int count = 0; | |
1253 | ||
1254 | DEBUG(D_route) debug_printf("parse_forward_list: %s\n", s); | |
1255 | ||
1256 | for (;;) | |
1257 | { | |
1258 | int len; | |
1259 | int special = 0; | |
1260 | int specopt = 0; | |
1261 | int specbit = 0; | |
1262 | uschar *ss, *nexts; | |
1263 | address_item *addr; | |
1264 | BOOL inquote = FALSE; | |
1265 | ||
1266 | for (;;) | |
1267 | { | |
1268 | while (isspace(*s) || *s == ',') s++; | |
1269 | if (*s == '#') { while (*s != 0 && *s != '\n') s++; } else break; | |
1270 | } | |
1271 | ||
1272 | /* When we reach the end of the list, we return FF_DELIVERED if any child | |
1273 | addresses have been generated. If nothing has been generated, there are two | |
1274 | possibilities: either the list is really empty, or there were syntax errors | |
1275 | that are being skipped. (If syntax errors are not being skipped, an FF_ERROR | |
1276 | return is generated on hitting a syntax error and we don't get here.) For a | |
1277 | truly empty list we return FF_NOTDELIVERED so that the router can decline. | |
1278 | However, if the list is empty only because syntax errors were skipped, we | |
1279 | return FF_DELIVERED. */ | |
1280 | ||
7d99cba1 | 1281 | if (!*s) |
059ec3d9 | 1282 | { |
7d99cba1 JH |
1283 | return (count > 0 || (syntax_errors && *syntax_errors)) |
1284 | ? FF_DELIVERED : FF_NOTDELIVERED; | |
059ec3d9 PH |
1285 | |
1286 | /* This previous code returns FF_ERROR if nothing is generated but a | |
1287 | syntax error has been skipped. I now think it is the wrong approach, but | |
1288 | have left this here just in case, and for the record. */ | |
1289 | ||
1290 | #ifdef NEVER | |
1291 | if (count > 0) return FF_DELIVERED; /* Something was generated */ | |
1292 | ||
1293 | if (syntax_errors == NULL || /* Not skipping syntax errors, or */ | |
1294 | *syntax_errors == NULL) /* we didn't actually skip any */ | |
1295 | return FF_NOTDELIVERED; | |
1296 | ||
1297 | *error = string_sprintf("no addresses generated: syntax error in %s: %s", | |
1298 | (*syntax_errors)->text2, (*syntax_errors)->text1); | |
1299 | return FF_ERROR; | |
1300 | #endif | |
1301 | ||
1302 | } | |
1303 | ||
1304 | /* Find the end of the next address. Quoted strings in addresses may contain | |
1305 | escaped characters; I haven't found a proper specification of .forward or | |
1306 | alias files that mentions the quoting properties, but it seems right to do | |
1307 | the escaping thing in all cases, so use the function that finds the end of an | |
1308 | address. However, don't let a quoted string extend over the end of a line. */ | |
1309 | ||
1310 | ss = parse_find_address_end(s, TRUE); | |
1311 | ||
1312 | /* Remember where we finished, for starting the next one. */ | |
1313 | ||
1314 | nexts = ss; | |
1315 | ||
1316 | /* Remove any trailing spaces; we know there's at least one non-space. */ | |
1317 | ||
1318 | while (isspace((ss[-1]))) ss--; | |
1319 | ||
1320 | /* We now have s->start and ss->end of the next address. Remove quotes | |
1321 | if they completely enclose, remembering the address started with a quote | |
1322 | for handling pipes and files. Another round of removal of leading and | |
1323 | trailing spaces is then required. */ | |
1324 | ||
1325 | if (*s == '\"' && ss[-1] == '\"') | |
1326 | { | |
1327 | s++; | |
1328 | ss--; | |
1329 | inquote = TRUE; | |
1330 | while (s < ss && isspace(*s)) s++; | |
1331 | while (ss > s && isspace((ss[-1]))) ss--; | |
1332 | } | |
1333 | ||
1334 | /* Set up the length of the address. */ | |
1335 | ||
1336 | len = ss - s; | |
1337 | ||
1338 | DEBUG(D_route) | |
1339 | { | |
1340 | int save = s[len]; | |
1341 | s[len] = 0; | |
1342 | debug_printf("extract item: %s\n", s); | |
1343 | s[len] = save; | |
1344 | } | |
1345 | ||
1346 | /* Handle special addresses if permitted. If the address is :unknown: | |
1347 | ignore it - this is for backward compatibility with old alias files. You | |
1348 | don't need to use it nowadays - just generate an empty string. For :defer:, | |
1349 | :blackhole:, or :fail: we have to set up the error message and give up right | |
1350 | away. */ | |
1351 | ||
1352 | if (Ustrncmp(s, ":unknown:", len) == 0) | |
1353 | { | |
1354 | s = nexts; | |
1355 | continue; | |
1356 | } | |
1357 | ||
1358 | if (Ustrncmp(s, ":defer:", 7) == 0) | |
1359 | { special = FF_DEFER; specopt = RDO_DEFER; } /* specbit is 0 */ | |
1360 | else if (Ustrncmp(s, ":blackhole:", 11) == 0) | |
1361 | { special = FF_BLACKHOLE; specopt = specbit = RDO_BLACKHOLE; } | |
1362 | else if (Ustrncmp(s, ":fail:", 6) == 0) | |
1363 | { special = FF_FAIL; specopt = RDO_FAIL; } /* specbit is 0 */ | |
1364 | ||
1365 | if (special != 0) | |
1366 | { | |
1367 | uschar *ss = Ustrchr(s+1, ':') + 1; | |
1368 | if ((options & specopt) == specbit) | |
1369 | { | |
1370 | *error = string_sprintf("\"%.*s\" is not permitted", len, s); | |
1371 | return FF_ERROR; | |
1372 | } | |
1373 | while (*ss != 0 && isspace(*ss)) ss++; | |
1374 | while (s[len] != 0 && s[len] != '\n') len++; | |
1375 | s[len] = 0; | |
1376 | *error = string_copy(ss); | |
1377 | return special; | |
1378 | } | |
1379 | ||
1380 | /* If the address is of the form :include:pathname, read the file, and call | |
1381 | this function recursively to extract the addresses from it. If directory is | |
1382 | NULL, do no checks. Otherwise, insist that the file name starts with the | |
1383 | given directory and is a regular file. */ | |
1384 | ||
1385 | if (Ustrncmp(s, ":include:", 9) == 0) | |
1386 | { | |
1387 | uschar *filebuf; | |
1388 | uschar filename[256]; | |
1389 | uschar *t = s+9; | |
1390 | int flen = len - 9; | |
1391 | int frc; | |
1392 | struct stat statbuf; | |
1393 | address_item *last; | |
1394 | FILE *f; | |
1395 | ||
1396 | while (flen > 0 && isspace(*t)) { t++; flen--; } | |
1397 | ||
1398 | if (flen <= 0) | |
1399 | { | |
f3ebb786 | 1400 | *error = US"file name missing after :include:"; |
059ec3d9 PH |
1401 | return FF_ERROR; |
1402 | } | |
1403 | ||
1404 | if (flen > 255) | |
1405 | { | |
1406 | *error = string_sprintf("included file name \"%s\" is too long", t); | |
1407 | return FF_ERROR; | |
1408 | } | |
1409 | ||
1410 | Ustrncpy(filename, t, flen); | |
1411 | filename[flen] = 0; | |
1412 | ||
1413 | /* Insist on absolute path */ | |
1414 | ||
7d99cba1 | 1415 | if (filename[0] != '/') |
059ec3d9 PH |
1416 | { |
1417 | *error = string_sprintf("included file \"%s\" is not an absolute path", | |
1418 | filename); | |
1419 | return FF_ERROR; | |
1420 | } | |
1421 | ||
1422 | /* Check if include is permitted */ | |
1423 | ||
7d99cba1 | 1424 | if (options & RDO_INCLUDE) |
059ec3d9 PH |
1425 | { |
1426 | *error = US"included files not permitted"; | |
1427 | return FF_ERROR; | |
1428 | } | |
1429 | ||
7d99cba1 JH |
1430 | if (is_tainted(filename)) |
1431 | { | |
1432 | *error = string_sprintf("Tainted name '%s' for included file not permitted\n", | |
1433 | filename); | |
1434 | return FF_ERROR; | |
1435 | } | |
1436 | ||
059ec3d9 PH |
1437 | /* Check file name if required */ |
1438 | ||
10c50704 | 1439 | if (directory) |
059ec3d9 PH |
1440 | { |
1441 | int len = Ustrlen(directory); | |
1442 | uschar *p = filename + len; | |
1443 | ||
1444 | if (Ustrncmp(filename, directory, len) != 0 || *p != '/') | |
1445 | { | |
1446 | *error = string_sprintf("included file %s is not in directory %s", | |
1447 | filename, directory); | |
1448 | return FF_ERROR; | |
1449 | } | |
1450 | ||
10c50704 JH |
1451 | #ifdef EXIM_HAVE_OPENAT |
1452 | /* It is necessary to check that every component inside the directory | |
1453 | is NOT a symbolic link, in order to keep the file inside the directory. | |
1454 | This is mighty tedious. We open the directory and openat every component, | |
1455 | with a flag that fails symlinks. */ | |
1456 | ||
1457 | { | |
9e21ce8f | 1458 | int fd = exim_open2(CS directory, O_RDONLY); |
10c50704 JH |
1459 | if (fd < 0) |
1460 | { | |
1461 | *error = string_sprintf("failed to open directory %s", directory); | |
1462 | return FF_ERROR; | |
1463 | } | |
1464 | while (*p) | |
1465 | { | |
1466 | uschar temp; | |
1467 | int fd2; | |
1468 | uschar * q = p; | |
1469 | ||
1470 | while (*++p && *p != '/') ; | |
1471 | temp = *p; | |
1472 | *p = '\0'; | |
1473 | ||
9e21ce8f | 1474 | fd2 = exim_openat(fd, CS q, O_RDONLY|O_NOFOLLOW); |
ad6f5499 JH |
1475 | close(fd); |
1476 | *p = temp; | |
1477 | if (fd2 < 0) | |
10c50704 JH |
1478 | { |
1479 | *error = string_sprintf("failed to open %s (component of included " | |
1480 | "file); could be symbolic link", filename); | |
1481 | return FF_ERROR; | |
1482 | } | |
10c50704 | 1483 | fd = fd2; |
10c50704 JH |
1484 | } |
1485 | f = fdopen(fd, "rb"); | |
1486 | } | |
1487 | #else | |
059ec3d9 PH |
1488 | /* It is necessary to check that every component inside the directory |
1489 | is NOT a symbolic link, in order to keep the file inside the directory. | |
1490 | This is mighty tedious. It is also not totally foolproof in that it | |
1491 | leaves the possibility of a race attack, but I don't know how to do | |
1492 | any better. */ | |
1493 | ||
10c50704 | 1494 | while (*p) |
059ec3d9 PH |
1495 | { |
1496 | int temp; | |
10c50704 | 1497 | while (*++p && *p != '/'); |
059ec3d9 PH |
1498 | temp = *p; |
1499 | *p = 0; | |
1500 | if (Ulstat(filename, &statbuf) != 0) | |
1501 | { | |
1502 | *error = string_sprintf("failed to stat %s (component of included " | |
1503 | "file)", filename); | |
1504 | *p = temp; | |
1505 | return FF_ERROR; | |
1506 | } | |
1507 | ||
1508 | *p = temp; | |
1509 | ||
1510 | if ((statbuf.st_mode & S_IFMT) == S_IFLNK) | |
1511 | { | |
1512 | *error = string_sprintf("included file %s in the %s directory " | |
1513 | "involves a symbolic link", filename, directory); | |
1514 | return FF_ERROR; | |
1515 | } | |
1516 | } | |
10c50704 | 1517 | #endif |
059ec3d9 PH |
1518 | } |
1519 | ||
10c50704 JH |
1520 | #ifdef EXIM_HAVE_OPENAT |
1521 | else | |
1522 | #endif | |
1523 | /* Open and stat the file */ | |
1524 | f = Ufopen(filename, "rb"); | |
059ec3d9 | 1525 | |
10c50704 | 1526 | if (!f) |
059ec3d9 PH |
1527 | { |
1528 | *error = string_open_failed(errno, "included file %s", filename); | |
1529 | return FF_INCLUDEFAIL; | |
1530 | } | |
1531 | ||
1532 | if (fstat(fileno(f), &statbuf) != 0) | |
1533 | { | |
1534 | *error = string_sprintf("failed to stat included file %s: %s", | |
1535 | filename, strerror(errno)); | |
f1e894f3 | 1536 | (void)fclose(f); |
059ec3d9 PH |
1537 | return FF_INCLUDEFAIL; |
1538 | } | |
1539 | ||
1540 | /* If directory was checked, double check that we opened a regular file */ | |
1541 | ||
10c50704 | 1542 | if (directory && (statbuf.st_mode & S_IFMT) != S_IFREG) |
059ec3d9 PH |
1543 | { |
1544 | *error = string_sprintf("included file %s is not a regular file in " | |
1545 | "the %s directory", filename, directory); | |
1546 | return FF_ERROR; | |
1547 | } | |
1548 | ||
1549 | /* Get a buffer and read the contents */ | |
1550 | ||
1551 | if (statbuf.st_size > MAX_INCLUDE_SIZE) | |
1552 | { | |
1553 | *error = string_sprintf("included file %s is too big (max %d)", | |
1554 | filename, MAX_INCLUDE_SIZE); | |
1555 | return FF_ERROR; | |
1556 | } | |
1557 | ||
f3ebb786 | 1558 | filebuf = store_get(statbuf.st_size + 1, is_tainted(filename)); |
059ec3d9 PH |
1559 | if (fread(filebuf, 1, statbuf.st_size, f) != statbuf.st_size) |
1560 | { | |
1561 | *error = string_sprintf("error while reading included file %s: %s", | |
1562 | filename, strerror(errno)); | |
f1e894f3 | 1563 | (void)fclose(f); |
059ec3d9 PH |
1564 | return FF_ERROR; |
1565 | } | |
1566 | filebuf[statbuf.st_size] = 0; | |
f1e894f3 | 1567 | (void)fclose(f); |
059ec3d9 PH |
1568 | |
1569 | addr = NULL; | |
1570 | frc = parse_forward_list(filebuf, options, &addr, | |
1571 | error, incoming_domain, directory, syntax_errors); | |
1572 | if (frc != FF_DELIVERED && frc != FF_NOTDELIVERED) return frc; | |
1573 | ||
10c50704 | 1574 | if (addr) |
059ec3d9 | 1575 | { |
10c50704 | 1576 | for (last = addr; last->next; last = last->next) count++; |
059ec3d9 PH |
1577 | last->next = *anchor; |
1578 | *anchor = addr; | |
1579 | count++; | |
1580 | } | |
1581 | } | |
1582 | ||
1583 | /* Else (not :include:) ensure address is syntactically correct and fully | |
1584 | qualified if not a pipe or a file, removing a leading \ if present on an | |
1585 | unqualified address. For pipes and files we must handle quoting. It's | |
1586 | not quite clear exactly what to do for partially quoted things, but the | |
1587 | common case of having the whole thing in quotes is straightforward. If this | |
1588 | was the case, inquote will have been set TRUE above and the quotes removed. | |
1589 | ||
1590 | There is a possible ambiguity over addresses whose local parts start with | |
1591 | a vertical bar or a slash, and the latter do in fact occur, thanks to X.400. | |
1592 | Consider a .forward file that contains the line | |
1593 | ||
1594 | /X=xxx/Y=xxx/OU=xxx/@some.gate.way | |
1595 | ||
1596 | Is this a file or an X.400 address? Does it make any difference if it is in | |
1597 | quotes? On the grounds that file names of this type are rare, Exim treats | |
1598 | something that parses as an RFC 822 address and has a domain as an address | |
1599 | rather than a file or a pipe. This is also how an address such as the above | |
1600 | would be treated if it came in from outside. */ | |
1601 | ||
1602 | else | |
1603 | { | |
1604 | int start, end, domain; | |
1605 | uschar *recipient = NULL; | |
1606 | int save = s[len]; | |
1607 | s[len] = 0; | |
1608 | ||
1609 | /* If it starts with \ and the rest of it parses as a valid mail address | |
1610 | without a domain, carry on with that address, but qualify it with the | |
1611 | incoming domain. Otherwise arrange for the address to fall through, | |
1612 | causing an error message on the re-parse. */ | |
1613 | ||
1614 | if (*s == '\\') | |
1615 | { | |
1616 | recipient = | |
1617 | parse_extract_address(s+1, error, &start, &end, &domain, FALSE); | |
5fcc791a JH |
1618 | if (recipient) |
1619 | recipient = domain != 0 ? NULL : | |
059ec3d9 PH |
1620 | string_sprintf("%s@%s", recipient, incoming_domain); |
1621 | } | |
1622 | ||
1623 | /* Try parsing the item as an address. */ | |
1624 | ||
5fcc791a | 1625 | if (!recipient) recipient = |
059ec3d9 PH |
1626 | parse_extract_address(s, error, &start, &end, &domain, FALSE); |
1627 | ||
1628 | /* If item starts with / or | and is not a valid address, or there | |
1629 | is no domain, treat it as a file or pipe. If it was a quoted item, | |
1630 | remove the quoting occurrences of \ within it. */ | |
1631 | ||
1632 | if ((*s == '|' || *s == '/') && (recipient == NULL || domain == 0)) | |
1633 | { | |
f3ebb786 | 1634 | uschar *t = store_get(Ustrlen(s) + 1, is_tainted(s)); |
059ec3d9 PH |
1635 | uschar *p = t; |
1636 | uschar *q = s; | |
1637 | while (*q != 0) | |
1638 | { | |
1639 | if (inquote) | |
1640 | { | |
1641 | *p++ = (*q == '\\')? *(++q) : *q; | |
1642 | q++; | |
1643 | } | |
1644 | else *p++ = *q++; | |
1645 | } | |
1646 | *p = 0; | |
1647 | addr = deliver_make_addr(t, TRUE); | |
1648 | setflag(addr, af_pfr); /* indicates pipe/file/reply */ | |
1649 | if (*s != '|') setflag(addr, af_file); /* indicates file */ | |
1650 | } | |
1651 | ||
1652 | /* Item must be an address. Complain if not, else qualify, rewrite and set | |
1653 | up the control block. It appears that people are in the habit of using | |
1654 | empty addresses but with comments as a way of putting comments into | |
1655 | alias and forward files. Therefore, ignore the error "empty address". | |
1656 | Mailing lists might want to tolerate syntax errors; there is therefore | |
1657 | an option to do so. */ | |
1658 | ||
1659 | else | |
1660 | { | |
1661 | if (recipient == NULL) | |
1662 | { | |
1663 | if (Ustrcmp(*error, "empty address") == 0) | |
1664 | { | |
1665 | *error = NULL; | |
1666 | s[len] = save; | |
1667 | s = nexts; | |
1668 | continue; | |
1669 | } | |
1670 | ||
1671 | if (syntax_errors != NULL) | |
1672 | { | |
f3ebb786 | 1673 | error_block *e = store_get(sizeof(error_block), FALSE); |
059ec3d9 PH |
1674 | error_block *last = *syntax_errors; |
1675 | if (last == NULL) *syntax_errors = e; else | |
1676 | { | |
1677 | while (last->next != NULL) last = last->next; | |
1678 | last->next = e; | |
1679 | } | |
1680 | e->next = NULL; | |
1681 | e->text1 = *error; | |
1682 | e->text2 = string_copy(s); | |
1683 | s[len] = save; | |
1684 | s = nexts; | |
1685 | continue; | |
1686 | } | |
1687 | else | |
1688 | { | |
1689 | *error = string_sprintf("%s in \"%s\"", *error, s); | |
1690 | s[len] = save; /* _after_ using it for *error */ | |
1691 | return FF_ERROR; | |
1692 | } | |
1693 | } | |
1694 | ||
1695 | /* Address was successfully parsed. Rewrite, and then make an address | |
1696 | block. */ | |
1697 | ||
1698 | recipient = ((options & RDO_REWRITE) != 0)? | |
1699 | rewrite_address(recipient, TRUE, FALSE, global_rewrite_rules, | |
1700 | rewrite_existflags) : | |
1701 | rewrite_address_qualify(recipient, TRUE); | |
1702 | addr = deliver_make_addr(recipient, TRUE); /* TRUE => copy recipient */ | |
1703 | } | |
1704 | ||
1705 | /* Restore the final character in the original data, and add to the | |
1706 | output chain. */ | |
1707 | ||
1708 | s[len] = save; | |
1709 | addr->next = *anchor; | |
1710 | *anchor = addr; | |
1711 | count++; | |
1712 | } | |
1713 | ||
1714 | /* Advance pointer for the next address */ | |
1715 | ||
1716 | s = nexts; | |
1717 | } | |
1718 | } | |
1719 | ||
1720 | ||
30dba1e6 PH |
1721 | /************************************************* |
1722 | * Extract a Message-ID * | |
1723 | *************************************************/ | |
1724 | ||
1725 | /* This function is used to extract message ids from In-Reply-To: and | |
1726 | References: header lines. | |
1727 | ||
1728 | Arguments: | |
1729 | str pointer to the start of the message-id | |
1730 | yield put pointer to the message id (in dynamic memory) here | |
1731 | error put error message here on failure | |
1732 | ||
1733 | Returns: points after the processed message-id or NULL on error | |
1734 | */ | |
1735 | ||
1736 | uschar * | |
1737 | parse_message_id(uschar *str, uschar **yield, uschar **error) | |
1738 | { | |
1739 | uschar *domain = NULL; | |
1740 | uschar *id; | |
f3ebb786 | 1741 | rmark reset_point; |
30dba1e6 PH |
1742 | |
1743 | str = skip_comment(str); | |
1744 | if (*str != '<') | |
1745 | { | |
1746 | *error = US"Missing '<' before message-id"; | |
1747 | return NULL; | |
1748 | } | |
1749 | ||
1750 | /* Getting a block the size of the input string will definitely be sufficient | |
1751 | for the answer, but it may also be very long if we are processing a header | |
1752 | line. Therefore, take care to release unwanted store afterwards. */ | |
1753 | ||
f3ebb786 JH |
1754 | reset_point = store_mark(); |
1755 | id = *yield = store_get(Ustrlen(str) + 1, is_tainted(str)); | |
30dba1e6 PH |
1756 | *id++ = *str++; |
1757 | ||
1758 | str = read_addr_spec(str, id, '>', error, &domain); | |
1759 | ||
f3ebb786 | 1760 | if (!*error) |
30dba1e6 PH |
1761 | { |
1762 | if (*str != '>') *error = US"Missing '>' after message-id"; | |
1763 | else if (domain == NULL) *error = US"domain missing in message-id"; | |
1764 | } | |
1765 | ||
f3ebb786 | 1766 | if (*error) |
30dba1e6 | 1767 | { |
f3ebb786 | 1768 | store_reset(reset_point); |
30dba1e6 PH |
1769 | return NULL; |
1770 | } | |
1771 | ||
f3ebb786 | 1772 | while (*id) id++; |
30dba1e6 PH |
1773 | *id++ = *str++; |
1774 | *id++ = 0; | |
f3ebb786 | 1775 | store_release_above(id); |
30dba1e6 PH |
1776 | |
1777 | str = skip_comment(str); | |
1778 | return str; | |
63ac05ee MH |
1779 | } |
1780 | ||
1781 | ||
1782 | /************************************************* | |
1783 | * Parse a fixed digit number * | |
1784 | *************************************************/ | |
1785 | ||
1786 | /* Parse a string containing an ASCII encoded fixed digits number | |
1787 | ||
1788 | Arguments: | |
1789 | str pointer to the start of the ASCII encoded number | |
1790 | n pointer to the resulting value | |
1791 | digits number of required digits | |
1792 | ||
1793 | Returns: points after the processed date or NULL on error | |
1794 | */ | |
1795 | ||
1796 | static uschar * | |
1797 | parse_number(uschar *str, int *n, int digits) | |
1798 | { | |
1799 | *n=0; | |
1800 | while (digits--) | |
1801 | { | |
1802 | if (*str<'0' || *str>'9') return NULL; | |
1803 | *n=10*(*n)+(*str++-'0'); | |
1804 | } | |
1805 | return str; | |
1806 | } | |
1807 | ||
1808 | ||
1809 | /************************************************* | |
1810 | * Parse a RFC 2822 day of week * | |
1811 | *************************************************/ | |
1812 | ||
1813 | /* Parse the day of the week from a RFC 2822 date, but do not | |
1814 | decode it, because it is only for humans. | |
1815 | ||
1816 | Arguments: | |
1817 | str pointer to the start of the day of the week | |
1818 | ||
1819 | Returns: points after the parsed day or NULL on error | |
1820 | */ | |
1821 | ||
1822 | static uschar * | |
1823 | parse_day_of_week(uschar *str) | |
1824 | { | |
1825 | /* | |
1826 | day-of-week = ([FWS] day-name) / obs-day-of-week | |
1827 | ||
1828 | day-name = "Mon" / "Tue" / "Wed" / "Thu" / | |
1829 | "Fri" / "Sat" / "Sun" | |
1830 | ||
1831 | obs-day-of-week = [CFWS] day-name [CFWS] | |
1832 | */ | |
1833 | ||
5ca6d115 | 1834 | static const uschar *day_name[7]={ US"mon", US"tue", US"wed", US"thu", US"fri", US"sat", US"sun" }; |
63ac05ee MH |
1835 | int i; |
1836 | uschar day[4]; | |
1837 | ||
1838 | str=skip_comment(str); | |
1839 | for (i=0; i<3; ++i) | |
1840 | { | |
1841 | if ((day[i]=tolower(*str))=='\0') return NULL; | |
1842 | ++str; | |
1843 | } | |
1844 | day[3]='\0'; | |
5ca6d115 | 1845 | for (i=0; i<7; ++i) if (Ustrcmp(day,day_name[i])==0) break; |
63ac05ee MH |
1846 | if (i==7) return NULL; |
1847 | str=skip_comment(str); | |
1848 | return str; | |
1849 | } | |
1850 | ||
1851 | ||
1852 | /************************************************* | |
1853 | * Parse a RFC 2822 date * | |
1854 | *************************************************/ | |
1855 | ||
1856 | /* Parse the date part of a RFC 2822 date-time, extracting the | |
1857 | day, month and year. | |
1858 | ||
1859 | Arguments: | |
1860 | str pointer to the start of the date | |
1861 | d pointer to the resulting day | |
1862 | m pointer to the resulting month | |
1863 | y pointer to the resulting year | |
1864 | ||
1865 | Returns: points after the processed date or NULL on error | |
1866 | */ | |
1867 | ||
1868 | static uschar * | |
1869 | parse_date(uschar *str, int *d, int *m, int *y) | |
1870 | { | |
1871 | /* | |
1872 | date = day month year | |
1873 | ||
1874 | year = 4*DIGIT / obs-year | |
1875 | ||
1876 | obs-year = [CFWS] 2*DIGIT [CFWS] | |
1877 | ||
1878 | month = (FWS month-name FWS) / obs-month | |
1879 | ||
1880 | month-name = "Jan" / "Feb" / "Mar" / "Apr" / | |
1881 | "May" / "Jun" / "Jul" / "Aug" / | |
1882 | "Sep" / "Oct" / "Nov" / "Dec" | |
1883 | ||
1884 | obs-month = CFWS month-name CFWS | |
1885 | ||
1886 | day = ([FWS] 1*2DIGIT) / obs-day | |
1887 | ||
1888 | obs-day = [CFWS] 1*2DIGIT [CFWS] | |
1889 | */ | |
1890 | ||
1891 | uschar *c,*n; | |
5ca6d115 | 1892 | static const uschar *month_name[]={ US"jan", US"feb", US"mar", US"apr", US"may", US"jun", US"jul", US"aug", US"sep", US"oct", US"nov", US"dec" }; |
63ac05ee MH |
1893 | int i; |
1894 | uschar month[4]; | |
1895 | ||
1896 | str=skip_comment(str); | |
1897 | if ((str=parse_number(str,d,1))==NULL) return NULL; | |
1898 | if (*str>='0' && *str<='9') *d=10*(*d)+(*str++-'0'); | |
1899 | c=skip_comment(str); | |
1900 | if (c==str) return NULL; | |
1901 | else str=c; | |
1902 | for (i=0; i<3; ++i) if ((month[i]=tolower(*(str+i)))=='\0') return NULL; | |
1903 | month[3]='\0'; | |
1904 | for (i=0; i<12; ++i) if (Ustrcmp(month,month_name[i])==0) break; | |
1905 | if (i==12) return NULL; | |
1906 | str+=3; | |
1907 | *m=i; | |
1908 | c=skip_comment(str); | |
1909 | if (c==str) return NULL; | |
1910 | else str=c; | |
1911 | if ((n=parse_number(str,y,4))) | |
1912 | { | |
1913 | str=n; | |
1914 | if (*y<1900) return NULL; | |
1915 | *y=*y-1900; | |
1916 | } | |
1917 | else if ((n=parse_number(str,y,2))) | |
1918 | { | |
1919 | str=skip_comment(n); | |
1920 | while (*(str-1)==' ' || *(str-1)=='\t') --str; /* match last FWS later */ | |
1921 | if (*y<50) *y+=100; | |
1922 | } | |
1923 | else return NULL; | |
1924 | return str; | |
1925 | } | |
1926 | ||
1927 | ||
1928 | /************************************************* | |
1929 | * Parse a RFC 2822 Time * | |
1930 | *************************************************/ | |
1931 | ||
1932 | /* Parse the time part of a RFC 2822 date-time, extracting the | |
1933 | hour, minute, second and timezone. | |
1934 | ||
1935 | Arguments: | |
1936 | str pointer to the start of the time | |
1937 | h pointer to the resulting hour | |
1938 | m pointer to the resulting minute | |
1939 | s pointer to the resulting second | |
1940 | z pointer to the resulting timezone (offset in seconds) | |
1941 | ||
1942 | Returns: points after the processed time or NULL on error | |
1943 | */ | |
1944 | ||
1945 | static uschar * | |
1946 | parse_time(uschar *str, int *h, int *m, int *s, int *z) | |
1947 | { | |
1948 | /* | |
1949 | time = time-of-day FWS zone | |
1950 | ||
1951 | time-of-day = hour ":" minute [ ":" second ] | |
1952 | ||
1953 | hour = 2DIGIT / obs-hour | |
1954 | ||
1955 | obs-hour = [CFWS] 2DIGIT [CFWS] | |
1956 | ||
1957 | minute = 2DIGIT / obs-minute | |
1958 | ||
1959 | obs-minute = [CFWS] 2DIGIT [CFWS] | |
1960 | ||
1961 | second = 2DIGIT / obs-second | |
1962 | ||
1963 | obs-second = [CFWS] 2DIGIT [CFWS] | |
1964 | ||
1965 | zone = (( "+" / "-" ) 4DIGIT) / obs-zone | |
1966 | ||
1967 | obs-zone = "UT" / "GMT" / ; Universal Time | |
1968 | ; North American UT | |
1969 | ; offsets | |
1970 | "EST" / "EDT" / ; Eastern: - 5/ - 4 | |
1971 | "CST" / "CDT" / ; Central: - 6/ - 5 | |
1972 | "MST" / "MDT" / ; Mountain: - 7/ - 6 | |
1973 | "PST" / "PDT" / ; Pacific: - 8/ - 7 | |
1974 | ||
1975 | %d65-73 / ; Military zones - "A" | |
1976 | %d75-90 / ; through "I" and "K" | |
1977 | %d97-105 / ; through "Z", both | |
1978 | %d107-122 ; upper and lower case | |
1979 | */ | |
1980 | ||
1981 | uschar *c; | |
1982 | ||
1983 | str=skip_comment(str); | |
1984 | if ((str=parse_number(str,h,2))==NULL) return NULL; | |
1985 | str=skip_comment(str); | |
1986 | if (*str!=':') return NULL; | |
1987 | ++str; | |
1988 | str=skip_comment(str); | |
1989 | if ((str=parse_number(str,m,2))==NULL) return NULL; | |
1990 | c=skip_comment(str); | |
1991 | if (*str==':') | |
1992 | { | |
1993 | ++str; | |
1994 | str=skip_comment(str); | |
1995 | if ((str=parse_number(str,s,2))==NULL) return NULL; | |
1996 | c=skip_comment(str); | |
1997 | } | |
1998 | if (c==str) return NULL; | |
1999 | else str=c; | |
2000 | if (*str=='+' || *str=='-') | |
2001 | { | |
2002 | int neg; | |
2003 | ||
2004 | neg=(*str=='-'); | |
2005 | ++str; | |
2006 | if ((str=parse_number(str,z,4))==NULL) return NULL; | |
2007 | *z=(*z/100)*3600+(*z%100)*60; | |
2008 | if (neg) *z=-*z; | |
2009 | } | |
2010 | else | |
2011 | { | |
2012 | char zone[5]; | |
2013 | struct { const char *name; int off; } zone_name[10]= | |
2014 | { {"gmt",0}, {"ut",0}, {"est",-5}, {"edt",-4}, {"cst",-6}, {"cdt",-5}, {"mst",-7}, {"mdt",-6}, {"pst",-8}, {"pdt",-7}}; | |
2015 | int i,j; | |
2016 | ||
2017 | for (i=0; i<4; ++i) | |
2018 | { | |
2019 | zone[i]=tolower(*(str+i)); | |
2020 | if (zone[i]<'a' || zone[i]>'z') break; | |
2021 | } | |
2022 | zone[i]='\0'; | |
2023 | for (j=0; j<10 && strcmp(zone,zone_name[j].name); ++j); | |
2024 | /* Besides zones named in the grammar, RFC 2822 says other alphabetic */ | |
2025 | /* time zones should be treated as unknown offsets. */ | |
2026 | if (j<10) | |
2027 | { | |
2028 | *z=zone_name[j].off*3600; | |
2029 | str+=i; | |
2030 | } | |
2031 | else if (zone[0]<'a' || zone[1]>'z') return 0; | |
2032 | else | |
2033 | { | |
2034 | while ((*str>='a' && *str<='z') || (*str>='A' && *str<='Z')) ++str; | |
2035 | *z=0; | |
2036 | } | |
2037 | } | |
2038 | return str; | |
2039 | } | |
2040 | ||
2041 | ||
2042 | /************************************************* | |
2043 | * Parse a RFC 2822 date-time * | |
2044 | *************************************************/ | |
2045 | ||
2046 | /* Parse a RFC 2822 date-time and return it in seconds since the epoch. | |
2047 | ||
2048 | Arguments: | |
2049 | str pointer to the start of the date-time | |
2050 | t pointer to the parsed time | |
2051 | ||
2052 | Returns: points after the processed date-time or NULL on error | |
2053 | */ | |
2054 | ||
2055 | uschar * | |
2056 | parse_date_time(uschar *str, time_t *t) | |
2057 | { | |
2058 | /* | |
2059 | date-time = [ day-of-week "," ] date FWS time [CFWS] | |
2060 | */ | |
2061 | ||
2062 | struct tm tm; | |
2063 | int zone; | |
2064 | extern char **environ; | |
2065 | char **old_environ; | |
2066 | static char gmt0[]="TZ=GMT0"; | |
2067 | static char *gmt_env[]={ gmt0, (char*)0 }; | |
ed2a4866 | 2068 | uschar *try; |
63ac05ee | 2069 | |
ed2a4866 | 2070 | if ((try=parse_day_of_week(str))) |
63ac05ee | 2071 | { |
ed2a4866 | 2072 | str=try; |
63ac05ee MH |
2073 | if (*str!=',') return 0; |
2074 | ++str; | |
2075 | } | |
2076 | if ((str=parse_date(str,&tm.tm_mday,&tm.tm_mon,&tm.tm_year))==NULL) return NULL; | |
2077 | if (*str!=' ' && *str!='\t') return NULL; | |
2078 | while (*str==' ' || *str=='\t') ++str; | |
2079 | if ((str=parse_time(str,&tm.tm_hour,&tm.tm_min,&tm.tm_sec,&zone))==NULL) return NULL; | |
2080 | tm.tm_isdst=0; | |
2081 | old_environ=environ; | |
2082 | environ=gmt_env; | |
2083 | *t=mktime(&tm); | |
2084 | environ=old_environ; | |
2085 | if (*t==-1) return NULL; | |
2086 | *t-=zone; | |
2087 | str=skip_comment(str); | |
2088 | return str; | |
30dba1e6 PH |
2089 | } |
2090 | ||
2091 | ||
2092 | ||
2093 | ||
059ec3d9 PH |
2094 | /************************************************* |
2095 | ************************************************** | |
2096 | * Stand-alone test program * | |
2097 | ************************************************** | |
2098 | *************************************************/ | |
2099 | ||
2100 | #if defined STAND_ALONE | |
2101 | int main(void) | |
2102 | { | |
2103 | int start, end, domain; | |
2104 | uschar buffer[1024]; | |
2105 | uschar outbuff[1024]; | |
2106 | ||
2107 | big_buffer = store_malloc(big_buffer_size); | |
2108 | ||
2109 | /* strip_trailing_dot = TRUE; */ | |
2110 | allow_domain_literals = TRUE; | |
2111 | ||
2112 | printf("Testing parse_fix_phrase\n"); | |
2113 | ||
2114 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
2115 | { | |
2116 | buffer[Ustrlen(buffer)-1] = 0; | |
2117 | if (buffer[0] == 0) break; | |
2118 | printf("%s\n", CS parse_fix_phrase(buffer, Ustrlen(buffer), outbuff, | |
2119 | sizeof(outbuff))); | |
2120 | } | |
2121 | ||
2122 | printf("Testing parse_extract_address without group syntax and without UTF-8\n"); | |
2123 | ||
2124 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
2125 | { | |
2126 | uschar *out; | |
2127 | uschar *errmess; | |
2128 | buffer[Ustrlen(buffer) - 1] = 0; | |
2129 | if (buffer[0] == 0) break; | |
2130 | out = parse_extract_address(buffer, &errmess, &start, &end, &domain, FALSE); | |
5fcc791a JH |
2131 | if (!out) |
2132 | printf("*** bad address: %s\n", errmess); | |
2133 | else | |
059ec3d9 PH |
2134 | { |
2135 | uschar extract[1024]; | |
2136 | Ustrncpy(extract, buffer+start, end-start); | |
2137 | extract[end-start] = 0; | |
2138 | printf("%s %d %d %d \"%s\"\n", out, start, end, domain, extract); | |
2139 | } | |
2140 | } | |
2141 | ||
2142 | printf("Testing parse_extract_address without group syntax but with UTF-8\n"); | |
2143 | ||
2144 | allow_utf8_domains = TRUE; | |
2145 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
2146 | { | |
2147 | uschar *out; | |
2148 | uschar *errmess; | |
2149 | buffer[Ustrlen(buffer) - 1] = 0; | |
2150 | if (buffer[0] == 0) break; | |
2151 | out = parse_extract_address(buffer, &errmess, &start, &end, &domain, FALSE); | |
5fcc791a JH |
2152 | if (!out) |
2153 | printf("*** bad address: %s\n", errmess); | |
2154 | else | |
059ec3d9 PH |
2155 | { |
2156 | uschar extract[1024]; | |
2157 | Ustrncpy(extract, buffer+start, end-start); | |
2158 | extract[end-start] = 0; | |
2159 | printf("%s %d %d %d \"%s\"\n", out, start, end, domain, extract); | |
2160 | } | |
2161 | } | |
2162 | allow_utf8_domains = FALSE; | |
2163 | ||
2164 | printf("Testing parse_extract_address with group syntax\n"); | |
2165 | ||
8768d548 | 2166 | f.parse_allow_group = TRUE; |
059ec3d9 PH |
2167 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) |
2168 | { | |
2169 | uschar *out; | |
2170 | uschar *errmess; | |
2171 | uschar *s; | |
2172 | buffer[Ustrlen(buffer) - 1] = 0; | |
2173 | if (buffer[0] == 0) break; | |
2174 | s = buffer; | |
5fcc791a | 2175 | while (*s) |
059ec3d9 PH |
2176 | { |
2177 | uschar *ss = parse_find_address_end(s, FALSE); | |
2178 | int terminator = *ss; | |
2179 | *ss = 0; | |
2180 | out = parse_extract_address(buffer, &errmess, &start, &end, &domain, FALSE); | |
2181 | *ss = terminator; | |
2182 | ||
5fcc791a JH |
2183 | if (!out) |
2184 | printf("*** bad address: %s\n", errmess); | |
2185 | else | |
059ec3d9 PH |
2186 | { |
2187 | uschar extract[1024]; | |
2188 | Ustrncpy(extract, buffer+start, end-start); | |
2189 | extract[end-start] = 0; | |
2190 | printf("%s %d %d %d \"%s\"\n", out, start, end, domain, extract); | |
2191 | } | |
2192 | ||
2193 | s = ss + (terminator? 1:0); | |
137ae145 | 2194 | Uskip_whitespace(&s); |
059ec3d9 PH |
2195 | } |
2196 | } | |
2197 | ||
2198 | printf("Testing parse_find_at\n"); | |
2199 | ||
2200 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
2201 | { | |
2202 | uschar *s; | |
2203 | buffer[Ustrlen(buffer)-1] = 0; | |
2204 | if (buffer[0] == 0) break; | |
2205 | s = parse_find_at(buffer); | |
2206 | if (s == NULL) printf("no @ found\n"); | |
2207 | else printf("offset = %d\n", s - buffer); | |
2208 | } | |
2209 | ||
2210 | printf("Testing parse_extract_addresses\n"); | |
2211 | ||
2212 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
2213 | { | |
2214 | uschar *errmess; | |
2215 | int extracted; | |
2216 | address_item *anchor = NULL; | |
2217 | buffer[Ustrlen(buffer) - 1] = 0; | |
2218 | if (buffer[0] == 0) break; | |
2219 | if ((extracted = parse_forward_list(buffer, -1, &anchor, | |
2220 | &errmess, US"incoming.domain", NULL, NULL)) == FF_DELIVERED) | |
2221 | { | |
2222 | while (anchor != NULL) | |
2223 | { | |
2224 | address_item *addr = anchor; | |
2225 | anchor = anchor->next; | |
2226 | printf("%d %s\n", testflag(addr, af_pfr), addr->address); | |
2227 | } | |
2228 | } | |
2229 | else printf("Failed: %d %s\n", extracted, errmess); | |
2230 | } | |
2231 | ||
30dba1e6 PH |
2232 | printf("Testing parse_message_id\n"); |
2233 | ||
2234 | while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) | |
2235 | { | |
2236 | uschar *s, *t, *errmess; | |
2237 | buffer[Ustrlen(buffer) - 1] = 0; | |
2238 | if (buffer[0] == 0) break; | |
2239 | s = buffer; | |
2240 | while (*s != 0) | |
2241 | { | |
2242 | s = parse_message_id(s, &t, &errmess); | |
2243 | if (errmess != NULL) | |
2244 | { | |
2245 | printf("Failed: %s\n", errmess); | |
2246 | break; | |
2247 | } | |
2248 | printf("%s\n", t); | |
2249 | } | |
2250 | } | |
2251 | ||
059ec3d9 PH |
2252 | return 0; |
2253 | } | |
2254 | ||
2255 | #endif | |
2256 | ||
2257 | /* End of parse.c */ |