Commit | Line | Data |
---|---|---|
d677b2f2 | 1 | /* $Cambridge: exim/src/src/receive.c,v 1.38 2007/06/22 14:38:58 ph10 Exp $ */ |
059ec3d9 PH |
2 | |
3 | /************************************************* | |
4 | * Exim - an Internet mail transport agent * | |
5 | *************************************************/ | |
6 | ||
184e8823 | 7 | /* Copyright (c) University of Cambridge 1995 - 2007 */ |
059ec3d9 PH |
8 | /* See the file NOTICE for conditions of use and distribution. */ |
9 | ||
10 | /* Code for receiving a message and setting up spool files. */ | |
11 | ||
059ec3d9 PH |
12 | #include "exim.h" |
13 | ||
fb2274d4 TK |
14 | #ifdef EXPERIMENTAL_DOMAINKEYS |
15 | #define RECEIVE_GETC dk_receive_getc | |
16 | #define RECEIVE_UNGETC dk_receive_ungetc | |
17 | #else | |
18 | #define RECEIVE_GETC receive_getc | |
19 | #define RECEIVE_UNGETC receive_ungetc | |
20 | #endif | |
059ec3d9 PH |
21 | |
22 | /************************************************* | |
23 | * Local static variables * | |
24 | *************************************************/ | |
25 | ||
26 | static FILE *data_file = NULL; | |
27 | static int data_fd = -1; | |
28 | static uschar spool_name[256]; | |
29 | ||
30 | ||
31 | ||
32 | /************************************************* | |
33 | * Non-SMTP character reading functions * | |
34 | *************************************************/ | |
35 | ||
36 | /* These are the default functions that are set up in the variables such as | |
37 | receive_getc initially. They just call the standard functions, passing stdin as | |
38 | the file. (When SMTP input is occurring, different functions are used by | |
39 | changing the pointer variables.) */ | |
40 | ||
41 | int | |
42 | stdin_getc(void) | |
43 | { | |
44 | return getc(stdin); | |
45 | } | |
46 | ||
47 | int | |
48 | stdin_ungetc(int c) | |
49 | { | |
50 | return ungetc(c, stdin); | |
51 | } | |
52 | ||
53 | int | |
54 | stdin_feof(void) | |
55 | { | |
56 | return feof(stdin); | |
57 | } | |
58 | ||
59 | int | |
60 | stdin_ferror(void) | |
61 | { | |
62 | return ferror(stdin); | |
63 | } | |
64 | ||
65 | ||
66 | ||
67 | ||
68 | /************************************************* | |
69 | * Check that a set sender is allowed * | |
70 | *************************************************/ | |
71 | ||
72 | /* This function is called when a local caller sets an explicit sender address. | |
73 | It checks whether this is permitted, which it is for trusted callers. | |
74 | Otherwise, it must match the pattern(s) in untrusted_set_sender. | |
75 | ||
76 | Arguments: the proposed sender address | |
77 | Returns: TRUE for a trusted caller | |
78 | TRUE if the address has been set, untrusted_set_sender has been | |
79 | set, and the address matches something in the list | |
80 | FALSE otherwise | |
81 | */ | |
82 | ||
83 | BOOL | |
84 | receive_check_set_sender(uschar *newsender) | |
85 | { | |
86 | uschar *qnewsender; | |
87 | if (trusted_caller) return TRUE; | |
88 | if (newsender == NULL || untrusted_set_sender == NULL) return FALSE; | |
89 | qnewsender = (Ustrchr(newsender, '@') != NULL)? | |
90 | newsender : string_sprintf("%s@%s", newsender, qualify_domain_sender); | |
91 | return | |
92 | match_address_list(qnewsender, TRUE, TRUE, &untrusted_set_sender, NULL, -1, | |
93 | 0, NULL) == OK; | |
94 | } | |
95 | ||
96 | ||
97 | ||
98 | ||
99 | /************************************************* | |
5cb8cbc6 | 100 | * Read space info for a partition * |
059ec3d9 PH |
101 | *************************************************/ |
102 | ||
8e669ac1 PH |
103 | /* This function is called by receive_check_fs() below, and also by string |
104 | expansion for variables such as $spool_space. The field names for the statvfs | |
5cb8cbc6 PH |
105 | structure are macros, because not all OS have F_FAVAIL and it seems tidier to |
106 | have macros for F_BAVAIL and F_FILES as well. Some kinds of file system do not | |
107 | have inodes, and they return -1 for the number available. | |
059ec3d9 | 108 | |
5cb8cbc6 PH |
109 | Later: It turns out that some file systems that do not have the concept of |
110 | inodes return 0 rather than -1. Such systems should also return 0 for the total | |
8e669ac1 | 111 | number of inodes, so we require that to be greater than zero before returning |
5cb8cbc6 | 112 | an inode count. |
059ec3d9 | 113 | |
5cb8cbc6 PH |
114 | Arguments: |
115 | isspool TRUE for spool partition, FALSE for log partition | |
116 | inodeptr address of int to receive inode count; -1 if there isn't one | |
8e669ac1 | 117 | |
5cb8cbc6 | 118 | Returns: available on-root space, in kilobytes |
8e669ac1 PH |
119 | -1 for log partition if there isn't one |
120 | ||
121 | All values are -1 if the STATFS functions are not available. | |
059ec3d9 PH |
122 | */ |
123 | ||
8e669ac1 | 124 | int |
5cb8cbc6 | 125 | receive_statvfs(BOOL isspool, int *inodeptr) |
059ec3d9 PH |
126 | { |
127 | #ifdef HAVE_STATFS | |
059ec3d9 | 128 | struct STATVFS statbuf; |
5cb8cbc6 PH |
129 | uschar *path; |
130 | uschar *name; | |
131 | uschar buffer[1024]; | |
059ec3d9 | 132 | |
5cb8cbc6 | 133 | /* The spool directory must always exist. */ |
059ec3d9 | 134 | |
5cb8cbc6 | 135 | if (isspool) |
059ec3d9 | 136 | { |
8e669ac1 PH |
137 | path = spool_directory; |
138 | name = US"spool"; | |
139 | } | |
140 | ||
059ec3d9 PH |
141 | /* Need to cut down the log file path to the directory, and to ignore any |
142 | appearance of "syslog" in it. */ | |
143 | ||
5cb8cbc6 | 144 | else |
059ec3d9 | 145 | { |
059ec3d9 | 146 | int sep = ':'; /* Not variable - outside scripts use */ |
059ec3d9 | 147 | uschar *p = log_file_path; |
8e669ac1 | 148 | name = US"log"; |
059ec3d9 PH |
149 | |
150 | /* An empty log_file_path means "use the default". This is the same as an | |
151 | empty item in a list. */ | |
152 | ||
153 | if (*p == 0) p = US":"; | |
154 | while ((path = string_nextinlist(&p, &sep, buffer, sizeof(buffer))) != NULL) | |
155 | { | |
156 | if (Ustrcmp(path, "syslog") != 0) break; | |
157 | } | |
158 | ||
5cb8cbc6 PH |
159 | if (path == NULL) /* No log files */ |
160 | { | |
8e669ac1 PH |
161 | *inodeptr = -1; |
162 | return -1; | |
163 | } | |
059ec3d9 | 164 | |
8e669ac1 PH |
165 | /* An empty string means use the default, which is in the spool directory. |
166 | But don't just use the spool directory, as it is possible that the log | |
5cb8cbc6 | 167 | subdirectory has been symbolically linked elsewhere. */ |
059ec3d9 | 168 | |
8e669ac1 | 169 | if (path[0] == 0) |
059ec3d9 | 170 | { |
5cb8cbc6 PH |
171 | sprintf(CS buffer, CS"%s/log", CS spool_directory); |
172 | path = buffer; | |
8e669ac1 PH |
173 | } |
174 | else | |
059ec3d9 | 175 | { |
8e669ac1 | 176 | uschar *cp; |
5cb8cbc6 | 177 | if ((cp = Ustrrchr(path, '/')) != NULL) *cp = 0; |
8e669ac1 | 178 | } |
5cb8cbc6 | 179 | } |
8e669ac1 | 180 | |
5cb8cbc6 PH |
181 | /* We now have the patch; do the business */ |
182 | ||
183 | memset(&statbuf, 0, sizeof(statbuf)); | |
184 | ||
185 | if (STATVFS(CS path, &statbuf) != 0) | |
186 | { | |
187 | log_write(0, LOG_MAIN|LOG_PANIC, "cannot accept message: failed to stat " | |
188 | "%s directory %s: %s", name, spool_directory, strerror(errno)); | |
189 | smtp_closedown(US"spool or log directory problem"); | |
190 | exim_exit(EXIT_FAILURE); | |
191 | } | |
8e669ac1 | 192 | |
5cb8cbc6 PH |
193 | *inodeptr = (statbuf.F_FILES > 0)? statbuf.F_FAVAIL : -1; |
194 | ||
195 | /* Disks are getting huge. Take care with computing the size in kilobytes. */ | |
8e669ac1 | 196 | |
5cb8cbc6 PH |
197 | return (int)(((double)statbuf.F_BAVAIL * (double)statbuf.F_FRSIZE)/1024.0); |
198 | ||
199 | /* Unable to find partition sizes in this environment. */ | |
200 | ||
201 | #else | |
202 | *inodeptr = -1; | |
203 | return -1; | |
204 | #endif | |
205 | } | |
206 | ||
059ec3d9 | 207 | |
059ec3d9 | 208 | |
5cb8cbc6 PH |
209 | |
210 | /************************************************* | |
211 | * Check space on spool and log partitions * | |
212 | *************************************************/ | |
213 | ||
214 | /* This function is called before accepting a message; if any thresholds are | |
215 | set, it checks them. If a message_size is supplied, it checks that there is | |
216 | enough space for that size plus the threshold - i.e. that the message won't | |
217 | reduce the space to the threshold. Not all OS have statvfs(); for those that | |
218 | don't, this function always returns TRUE. For some OS the old function and | |
219 | struct name statfs is used; that is handled by a macro, defined in exim.h. | |
220 | ||
221 | Arguments: | |
222 | msg_size the (estimated) size of an incoming message | |
223 | ||
224 | Returns: FALSE if there isn't enough space, or if the information cannot | |
225 | be obtained | |
226 | TRUE if no check was done or there is enough space | |
227 | */ | |
228 | ||
229 | BOOL | |
230 | receive_check_fs(int msg_size) | |
231 | { | |
232 | int space, inodes; | |
233 | ||
234 | if (check_spool_space > 0 || msg_size > 0 || check_spool_inodes > 0) | |
235 | { | |
8e669ac1 PH |
236 | space = receive_statvfs(TRUE, &inodes); |
237 | ||
059ec3d9 | 238 | DEBUG(D_receive) |
5cb8cbc6 PH |
239 | debug_printf("spool directory space = %dK inodes = %d " |
240 | "check_space = %dK inodes = %d msg_size = %d\n", | |
241 | space, inodes, check_spool_space, check_spool_inodes, msg_size); | |
8e669ac1 PH |
242 | |
243 | if ((space >= 0 && space < check_spool_space) || | |
5cb8cbc6 | 244 | (inodes >= 0 && inodes < check_spool_inodes)) |
8e669ac1 | 245 | { |
5cb8cbc6 PH |
246 | log_write(0, LOG_MAIN, "spool directory space check failed: space=%d " |
247 | "inodes=%d", space, inodes); | |
059ec3d9 PH |
248 | return FALSE; |
249 | } | |
250 | } | |
251 | ||
5cb8cbc6 PH |
252 | if (check_log_space > 0 || check_log_inodes > 0) |
253 | { | |
8e669ac1 PH |
254 | space = receive_statvfs(FALSE, &inodes); |
255 | ||
5cb8cbc6 PH |
256 | DEBUG(D_receive) |
257 | debug_printf("log directory space = %dK inodes = %d " | |
258 | "check_space = %dK inodes = %d\n", | |
259 | space, inodes, check_log_space, check_log_inodes); | |
8e669ac1 PH |
260 | |
261 | if ((space >= 0 && space < check_log_space) || | |
5cb8cbc6 | 262 | (inodes >= 0 && inodes < check_log_inodes)) |
8e669ac1 | 263 | { |
5cb8cbc6 PH |
264 | log_write(0, LOG_MAIN, "log directory space check failed: space=%d " |
265 | "inodes=%d", space, inodes); | |
266 | return FALSE; | |
267 | } | |
8e669ac1 PH |
268 | } |
269 | ||
059ec3d9 PH |
270 | return TRUE; |
271 | } | |
272 | ||
273 | ||
274 | ||
275 | /************************************************* | |
276 | * Bomb out while reading a message * | |
277 | *************************************************/ | |
278 | ||
279 | /* The common case of wanting to bomb out is if a SIGTERM or SIGINT is | |
280 | received, or if there is a timeout. A rarer case might be if the log files are | |
281 | screwed up and Exim can't open them to record a message's arrival. Handling | |
282 | that case is done by setting a flag to cause the log functions to call this | |
283 | function if there is an ultimate disaster. That is why it is globally | |
284 | accessible. | |
285 | ||
286 | Arguments: SMTP response to give if in an SMTP session | |
287 | Returns: it doesn't | |
288 | */ | |
289 | ||
290 | void | |
291 | receive_bomb_out(uschar *msg) | |
292 | { | |
293 | /* If spool_name is set, it contains the name of the data file that is being | |
294 | written. Unlink it before closing so that it cannot be picked up by a delivery | |
295 | process. Ensure that any header file is also removed. */ | |
296 | ||
297 | if (spool_name[0] != 0) | |
298 | { | |
299 | Uunlink(spool_name); | |
300 | spool_name[Ustrlen(spool_name) - 1] = 'H'; | |
301 | Uunlink(spool_name); | |
302 | } | |
303 | ||
304 | /* Now close the file if it is open, either as a fd or a stream. */ | |
305 | ||
f1e894f3 PH |
306 | if (data_file != NULL) (void)fclose(data_file); |
307 | else if (data_fd >= 0) (void)close(data_fd); | |
059ec3d9 PH |
308 | |
309 | /* Attempt to close down an SMTP connection tidily. */ | |
310 | ||
311 | if (smtp_input) | |
312 | { | |
313 | if (!smtp_batched_input) | |
314 | { | |
315 | smtp_printf("421 %s %s - closing connection.\r\n", smtp_active_hostname, | |
316 | msg); | |
317 | mac_smtp_fflush(); | |
318 | } | |
319 | ||
320 | /* Control does not return from moan_smtp_batch(). */ | |
321 | ||
322 | else moan_smtp_batch(NULL, "421 %s - message abandoned", msg); | |
323 | } | |
324 | ||
325 | /* Exit from the program (non-BSMTP cases) */ | |
326 | ||
327 | exim_exit(EXIT_FAILURE); | |
328 | } | |
329 | ||
330 | ||
331 | /************************************************* | |
332 | * Data read timeout * | |
333 | *************************************************/ | |
334 | ||
335 | /* Handler function for timeouts that occur while reading the data that | |
336 | comprises a message. | |
337 | ||
338 | Argument: the signal number | |
339 | Returns: nothing | |
340 | */ | |
341 | ||
342 | static void | |
343 | data_timeout_handler(int sig) | |
344 | { | |
345 | uschar *msg = NULL; | |
346 | ||
347 | sig = sig; /* Keep picky compilers happy */ | |
348 | ||
349 | if (smtp_input) | |
350 | { | |
351 | msg = US"SMTP incoming data timeout"; | |
352 | log_write(L_lost_incoming_connection, | |
353 | LOG_MAIN, "SMTP data timeout (message abandoned) on connection " | |
fed77020 PH |
354 | "from %s F=<%s>", |
355 | (sender_fullhost != NULL)? sender_fullhost : US"local process", | |
356 | sender_address); | |
059ec3d9 PH |
357 | } |
358 | else | |
359 | { | |
360 | fprintf(stderr, "exim: timed out while reading - message abandoned\n"); | |
361 | log_write(L_lost_incoming_connection, | |
362 | LOG_MAIN, "timed out while reading local message"); | |
363 | } | |
364 | ||
365 | receive_bomb_out(msg); /* Does not return */ | |
366 | } | |
367 | ||
368 | ||
369 | ||
370 | /************************************************* | |
371 | * local_scan() timeout * | |
372 | *************************************************/ | |
373 | ||
374 | /* Handler function for timeouts that occur while running a local_scan() | |
375 | function. | |
376 | ||
377 | Argument: the signal number | |
378 | Returns: nothing | |
379 | */ | |
380 | ||
381 | static void | |
382 | local_scan_timeout_handler(int sig) | |
383 | { | |
384 | sig = sig; /* Keep picky compilers happy */ | |
385 | log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function timed out - " | |
386 | "message temporarily rejected (size %d)", message_size); | |
387 | receive_bomb_out(US"local verification problem"); /* Does not return */ | |
388 | } | |
389 | ||
390 | ||
391 | ||
392 | /************************************************* | |
393 | * local_scan() crashed * | |
394 | *************************************************/ | |
395 | ||
396 | /* Handler function for signals that occur while running a local_scan() | |
397 | function. | |
398 | ||
399 | Argument: the signal number | |
400 | Returns: nothing | |
401 | */ | |
402 | ||
403 | static void | |
404 | local_scan_crash_handler(int sig) | |
405 | { | |
406 | log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function crashed with " | |
407 | "signal %d - message temporarily rejected (size %d)", sig, message_size); | |
408 | receive_bomb_out(US"local verification problem"); /* Does not return */ | |
409 | } | |
410 | ||
411 | ||
412 | /************************************************* | |
413 | * SIGTERM or SIGINT received * | |
414 | *************************************************/ | |
415 | ||
416 | /* Handler for SIGTERM or SIGINT signals that occur while reading the | |
417 | data that comprises a message. | |
418 | ||
419 | Argument: the signal number | |
420 | Returns: nothing | |
421 | */ | |
422 | ||
423 | static void | |
424 | data_sigterm_sigint_handler(int sig) | |
425 | { | |
426 | uschar *msg = NULL; | |
427 | ||
428 | if (smtp_input) | |
429 | { | |
430 | msg = US"Service not available - SIGTERM or SIGINT received"; | |
431 | log_write(0, LOG_MAIN, "%s closed after %s", smtp_get_connection_info(), | |
432 | (sig == SIGTERM)? "SIGTERM" : "SIGINT"); | |
433 | } | |
434 | else | |
435 | { | |
f05da2e8 | 436 | if (filter_test == FTEST_NONE) |
059ec3d9 PH |
437 | { |
438 | fprintf(stderr, "\nexim: %s received - message abandoned\n", | |
439 | (sig == SIGTERM)? "SIGTERM" : "SIGINT"); | |
440 | log_write(0, LOG_MAIN, "%s received while reading local message", | |
441 | (sig == SIGTERM)? "SIGTERM" : "SIGINT"); | |
442 | } | |
443 | } | |
444 | ||
445 | receive_bomb_out(msg); /* Does not return */ | |
446 | } | |
447 | ||
448 | ||
449 | ||
450 | /************************************************* | |
451 | * Add new recipient to list * | |
452 | *************************************************/ | |
453 | ||
454 | /* This function builds a list of recipient addresses in argc/argv | |
455 | format. | |
456 | ||
457 | Arguments: | |
458 | recipient the next address to add to recipients_list | |
459 | pno parent number for fixed aliases; -1 otherwise | |
460 | ||
461 | Returns: nothing | |
462 | */ | |
463 | ||
464 | void | |
465 | receive_add_recipient(uschar *recipient, int pno) | |
466 | { | |
467 | if (recipients_count >= recipients_list_max) | |
468 | { | |
469 | recipient_item *oldlist = recipients_list; | |
470 | int oldmax = recipients_list_max; | |
471 | recipients_list_max = recipients_list_max? 2*recipients_list_max : 50; | |
472 | recipients_list = store_get(recipients_list_max * sizeof(recipient_item)); | |
473 | if (oldlist != NULL) | |
474 | memcpy(recipients_list, oldlist, oldmax * sizeof(recipient_item)); | |
475 | } | |
476 | ||
477 | recipients_list[recipients_count].address = recipient; | |
478 | recipients_list[recipients_count].pno = pno; | |
8523533c TK |
479 | #ifdef EXPERIMENTAL_BRIGHTMAIL |
480 | recipients_list[recipients_count].bmi_optin = bmi_current_optin; | |
481 | /* reset optin string pointer for next recipient */ | |
482 | bmi_current_optin = NULL; | |
483 | #endif | |
059ec3d9 PH |
484 | recipients_list[recipients_count++].errors_to = NULL; |
485 | } | |
486 | ||
487 | ||
488 | ||
489 | ||
490 | /************************************************* | |
491 | * Remove a recipient from the list * | |
492 | *************************************************/ | |
493 | ||
494 | /* This function is provided for local_scan() to use. | |
495 | ||
496 | Argument: | |
497 | recipient address to remove | |
498 | ||
499 | Returns: TRUE if it did remove something; FALSE otherwise | |
500 | */ | |
501 | ||
502 | BOOL | |
503 | receive_remove_recipient(uschar *recipient) | |
504 | { | |
505 | int count; | |
506 | DEBUG(D_receive) debug_printf("receive_remove_recipient(\"%s\") called\n", | |
507 | recipient); | |
508 | for (count = 0; count < recipients_count; count++) | |
509 | { | |
510 | if (Ustrcmp(recipients_list[count].address, recipient) == 0) | |
511 | { | |
512 | if ((--recipients_count - count) > 0) | |
513 | memmove(recipients_list + count, recipients_list + count + 1, | |
54cdb463 | 514 | (recipients_count - count)*sizeof(recipient_item)); |
059ec3d9 PH |
515 | return TRUE; |
516 | } | |
517 | } | |
518 | return FALSE; | |
519 | } | |
520 | ||
521 | ||
522 | ||
523 | ||
524 | ||
525 | /************************************************* | |
526 | * Read data portion of a non-SMTP message * | |
527 | *************************************************/ | |
528 | ||
529 | /* This function is called to read the remainder of a message (following the | |
530 | header) when the input is not from SMTP - we are receiving a local message on | |
531 | a standard input stream. The message is always terminated by EOF, and is also | |
532 | terminated by a dot on a line by itself if the flag dot_ends is TRUE. Split the | |
533 | two cases for maximum efficiency. | |
534 | ||
535 | Ensure that the body ends with a newline. This will naturally be the case when | |
536 | the termination is "\n.\n" but may not be otherwise. The RFC defines messages | |
537 | as "sequences of lines" - this of course strictly applies only to SMTP, but | |
538 | deliveries into BSD-type mailbox files also require it. Exim used to have a | |
539 | flag for doing this at delivery time, but as it was always set for all | |
540 | transports, I decided to simplify things by putting the check here instead. | |
541 | ||
542 | There is at least one MUA (dtmail) that sends CRLF via this interface, and | |
543 | other programs are known to do this as well. Exim used to have a option for | |
544 | dealing with this: in July 2003, after much discussion, the code has been | |
545 | changed to default to treat any of LF, CRLF, and bare CR as line terminators. | |
546 | ||
547 | However, for the case when a dot on a line by itself terminates a message, the | |
548 | only recognized terminating sequences before and after the dot are LF and CRLF. | |
549 | Otherwise, having read EOL . CR, you don't know whether to read another | |
550 | character or not. | |
551 | ||
552 | Internally, in messages stored in Exim's spool files, LF is used as the line | |
553 | terminator. Under the new regime, bare CRs will no longer appear in these | |
554 | files. | |
555 | ||
556 | Arguments: | |
557 | fout a FILE to which to write the message | |
558 | ||
559 | Returns: One of the END_xxx values indicating why it stopped reading | |
560 | */ | |
561 | ||
562 | static int | |
563 | read_message_data(FILE *fout) | |
564 | { | |
565 | int ch_state; | |
566 | register int ch; | |
d677b2f2 | 567 | register int linelength = 0; |
059ec3d9 PH |
568 | |
569 | /* Handle the case when only EOF terminates the message */ | |
570 | ||
571 | if (!dot_ends) | |
572 | { | |
573 | register int last_ch = '\n'; | |
574 | ||
fb2274d4 | 575 | for (; (ch = (RECEIVE_GETC)()) != EOF; last_ch = ch) |
059ec3d9 PH |
576 | { |
577 | if (ch == 0) body_zerocount++; | |
578 | if (last_ch == '\r' && ch != '\n') | |
579 | { | |
d677b2f2 PH |
580 | if (linelength > max_received_linelength) |
581 | max_received_linelength = linelength; | |
582 | linelength = 0; | |
059ec3d9 PH |
583 | if (fputc('\n', fout) == EOF) return END_WERROR; |
584 | message_size++; | |
585 | body_linecount++; | |
586 | } | |
587 | if (ch == '\r') continue; | |
588 | ||
589 | if (fputc(ch, fout) == EOF) return END_WERROR; | |
d677b2f2 PH |
590 | if (ch == '\n') |
591 | { | |
592 | if (linelength > max_received_linelength) | |
593 | max_received_linelength = linelength; | |
594 | linelength = 0; | |
595 | body_linecount++; | |
596 | } | |
597 | else linelength++; | |
059ec3d9 PH |
598 | if (++message_size > thismessage_size_limit) return END_SIZE; |
599 | } | |
600 | ||
601 | if (last_ch != '\n') | |
602 | { | |
d677b2f2 PH |
603 | if (linelength > max_received_linelength) |
604 | max_received_linelength = linelength; | |
059ec3d9 PH |
605 | if (fputc('\n', fout) == EOF) return END_WERROR; |
606 | message_size++; | |
607 | body_linecount++; | |
608 | } | |
609 | ||
610 | return END_EOF; | |
611 | } | |
612 | ||
613 | /* Handle the case when a dot on a line on its own, or EOF, terminates. */ | |
614 | ||
615 | ch_state = 1; | |
616 | ||
fb2274d4 | 617 | while ((ch = (RECEIVE_GETC)()) != EOF) |
059ec3d9 PH |
618 | { |
619 | if (ch == 0) body_zerocount++; | |
620 | switch (ch_state) | |
621 | { | |
622 | case 0: /* Normal state (previous char written) */ | |
623 | if (ch == '\n') | |
d677b2f2 PH |
624 | { |
625 | body_linecount++; | |
626 | if (linelength > max_received_linelength) | |
627 | max_received_linelength = linelength; | |
628 | linelength = -1; | |
629 | ch_state = 1; | |
630 | } | |
059ec3d9 PH |
631 | else if (ch == '\r') |
632 | { ch_state = 2; continue; } | |
633 | break; | |
634 | ||
635 | case 1: /* After written "\n" */ | |
636 | if (ch == '.') { ch_state = 3; continue; } | |
d677b2f2 | 637 | if (ch != '\n') ch_state = 0; else linelength = -1; |
059ec3d9 PH |
638 | break; |
639 | ||
640 | case 2: | |
641 | body_linecount++; /* After unwritten "\r" */ | |
d677b2f2 PH |
642 | if (linelength > max_received_linelength) |
643 | max_received_linelength = linelength; | |
059ec3d9 | 644 | if (ch == '\n') |
d677b2f2 PH |
645 | { |
646 | ch_state = 1; | |
647 | linelength = -1; | |
648 | } | |
059ec3d9 PH |
649 | else |
650 | { | |
651 | if (message_size++, fputc('\n', fout) == EOF) return END_WERROR; | |
652 | if (ch == '\r') continue; | |
653 | ch_state = 0; | |
d677b2f2 | 654 | linelength = 0; |
059ec3d9 PH |
655 | } |
656 | break; | |
657 | ||
658 | case 3: /* After "\n." (\n written, dot not) */ | |
659 | if (ch == '\n') return END_DOT; | |
660 | if (ch == '\r') { ch_state = 4; continue; } | |
661 | message_size++; | |
d677b2f2 | 662 | linelength++; |
059ec3d9 PH |
663 | if (fputc('.', fout) == EOF) return END_WERROR; |
664 | ch_state = 0; | |
665 | break; | |
666 | ||
667 | case 4: /* After "\n.\r" (\n written, rest not) */ | |
668 | if (ch == '\n') return END_DOT; | |
669 | message_size += 2; | |
670 | body_linecount++; | |
671 | if (fputs(".\n", fout) == EOF) return END_WERROR; | |
672 | if (ch == '\r') { ch_state = 2; continue; } | |
673 | ch_state = 0; | |
674 | break; | |
675 | } | |
676 | ||
d677b2f2 | 677 | linelength++; |
059ec3d9 PH |
678 | if (fputc(ch, fout) == EOF) return END_WERROR; |
679 | if (++message_size > thismessage_size_limit) return END_SIZE; | |
680 | } | |
681 | ||
682 | /* Get here if EOF read. Unless we have just written "\n", we need to ensure | |
683 | the message ends with a newline, and we must also write any characters that | |
684 | were saved up while testing for an ending dot. */ | |
685 | ||
686 | if (ch_state != 1) | |
687 | { | |
688 | static uschar *ends[] = { US"\n", NULL, US"\n", US".\n", US".\n" }; | |
689 | if (fputs(CS ends[ch_state], fout) == EOF) return END_WERROR; | |
690 | message_size += Ustrlen(ends[ch_state]); | |
691 | body_linecount++; | |
692 | } | |
693 | ||
694 | return END_EOF; | |
695 | } | |
696 | ||
697 | ||
698 | ||
699 | ||
700 | /************************************************* | |
701 | * Read data portion of an SMTP message * | |
702 | *************************************************/ | |
703 | ||
704 | /* This function is called to read the remainder of an SMTP message (after the | |
705 | headers), or to skip over it when an error has occurred. In this case, the | |
706 | output file is passed as NULL. | |
707 | ||
708 | If any line begins with a dot, that character is skipped. The input should only | |
709 | be successfully terminated by CR LF . CR LF unless it is local (non-network) | |
710 | SMTP, in which case the CRs are optional, but... | |
711 | ||
712 | FUDGE: It seems that sites on the net send out messages with just LF | |
713 | terminators, despite the warnings in the RFCs, and other MTAs handle this. So | |
714 | we make the CRs optional in all cases. | |
715 | ||
716 | July 2003: Bare CRs cause trouble. We now treat them as line terminators as | |
717 | well, so that there are no CRs in spooled messages. However, the message | |
718 | terminating dot is not recognized between two bare CRs. | |
719 | ||
720 | Arguments: | |
721 | fout a FILE to which to write the message; NULL if skipping | |
722 | ||
723 | Returns: One of the END_xxx values indicating why it stopped reading | |
724 | */ | |
725 | ||
726 | static int | |
727 | read_message_data_smtp(FILE *fout) | |
728 | { | |
729 | int ch_state = 0; | |
730 | register int ch; | |
731 | ||
fb2274d4 | 732 | while ((ch = (RECEIVE_GETC)()) != EOF) |
059ec3d9 PH |
733 | { |
734 | if (ch == 0) body_zerocount++; | |
735 | switch (ch_state) | |
736 | { | |
737 | case 0: /* After LF or CRLF */ | |
738 | if (ch == '.') | |
739 | { | |
740 | ch_state = 3; | |
741 | continue; /* Don't ever write . after LF */ | |
742 | } | |
743 | ch_state = 1; | |
744 | ||
745 | /* Else fall through to handle as normal uschar. */ | |
746 | ||
747 | case 1: /* Normal state */ | |
748 | if (ch == '\n') | |
749 | { | |
750 | ch_state = 0; | |
751 | body_linecount++; | |
752 | } | |
753 | else if (ch == '\r') | |
754 | { | |
755 | ch_state = 2; | |
756 | continue; | |
757 | } | |
758 | break; | |
759 | ||
760 | case 2: /* After (unwritten) CR */ | |
761 | body_linecount++; | |
762 | if (ch == '\n') | |
763 | { | |
764 | ch_state = 0; | |
765 | } | |
766 | else | |
767 | { | |
768 | message_size++; | |
769 | if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR; | |
770 | if (ch != '\r') ch_state = 1; else continue; | |
771 | } | |
772 | break; | |
773 | ||
774 | case 3: /* After [CR] LF . */ | |
775 | if (ch == '\n') | |
776 | return END_DOT; | |
777 | if (ch == '\r') | |
778 | { | |
779 | ch_state = 4; | |
780 | continue; | |
781 | } | |
782 | ch_state = 1; /* The dot itself is removed */ | |
783 | break; | |
784 | ||
785 | case 4: /* After [CR] LF . CR */ | |
786 | if (ch == '\n') return END_DOT; | |
787 | message_size++; | |
788 | body_linecount++; | |
789 | if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR; | |
790 | if (ch == '\r') | |
791 | { | |
792 | ch_state = 2; | |
793 | continue; | |
794 | } | |
795 | ch_state = 1; | |
796 | break; | |
797 | } | |
798 | ||
799 | /* Add the character to the spool file, unless skipping; then loop for the | |
800 | next. */ | |
801 | ||
802 | message_size++; | |
803 | if (fout != NULL) | |
804 | { | |
805 | if (fputc(ch, fout) == EOF) return END_WERROR; | |
806 | if (message_size > thismessage_size_limit) return END_SIZE; | |
807 | } | |
808 | } | |
809 | ||
810 | /* Fall through here if EOF encountered. This indicates some kind of error, | |
811 | since a correct message is terminated by [CR] LF . [CR] LF. */ | |
812 | ||
813 | return END_EOF; | |
814 | } | |
815 | ||
816 | ||
817 | ||
818 | ||
819 | /************************************************* | |
820 | * Swallow SMTP message * | |
821 | *************************************************/ | |
822 | ||
823 | /* This function is called when there has been some kind of error while reading | |
824 | an SMTP message, and the remaining data may need to be swallowed. It is global | |
825 | because it is called from smtp_closedown() to shut down an incoming call | |
826 | tidily. | |
827 | ||
828 | Argument: a FILE from which to read the message | |
829 | Returns: nothing | |
830 | */ | |
831 | ||
832 | void | |
833 | receive_swallow_smtp(void) | |
834 | { | |
835 | if (message_ended >= END_NOTENDED) | |
836 | message_ended = read_message_data_smtp(NULL); | |
837 | } | |
838 | ||
839 | ||
840 | ||
841 | /************************************************* | |
842 | * Handle lost SMTP connection * | |
843 | *************************************************/ | |
844 | ||
845 | /* This function logs connection loss incidents and generates an appropriate | |
846 | SMTP response. | |
847 | ||
848 | Argument: additional data for the message | |
849 | Returns: the SMTP response | |
850 | */ | |
851 | ||
852 | static uschar * | |
853 | handle_lost_connection(uschar *s) | |
854 | { | |
855 | log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN, | |
856 | "%s lost while reading message data%s", smtp_get_connection_info(), s); | |
857 | return US"421 Lost incoming connection"; | |
858 | } | |
859 | ||
860 | ||
861 | ||
862 | ||
863 | /************************************************* | |
864 | * Handle a non-smtp reception error * | |
865 | *************************************************/ | |
866 | ||
867 | /* This function is called for various errors during the reception of non-SMTP | |
868 | messages. It either sends a message to the sender of the problem message, or it | |
869 | writes to the standard error stream. | |
870 | ||
871 | Arguments: | |
872 | errcode code for moan_to_sender(), identifying the error | |
873 | text1 first message text, passed to moan_to_sender() | |
874 | text2 second message text, used only for stderrr | |
875 | error_rc code to pass to exim_exit if no problem | |
876 | f FILE containing body of message (may be stdin) | |
877 | hptr pointer to instore headers or NULL | |
878 | ||
879 | Returns: calls exim_exit(), which does not return | |
880 | */ | |
881 | ||
882 | static void | |
883 | give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc, | |
884 | FILE *f, header_line *hptr) | |
885 | { | |
886 | if (error_handling == ERRORS_SENDER) | |
887 | { | |
888 | error_block eblock; | |
889 | eblock.next = NULL; | |
890 | eblock.text1 = text1; | |
891 | if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE)) | |
892 | error_rc = EXIT_FAILURE; | |
893 | } | |
894 | else fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */ | |
f1e894f3 | 895 | (void)fclose(f); |
059ec3d9 PH |
896 | exim_exit(error_rc); |
897 | } | |
898 | ||
899 | ||
900 | ||
901 | /************************************************* | |
902 | * Add header lines set up by ACL * | |
903 | *************************************************/ | |
904 | ||
850635b6 PH |
905 | /* This function is called to add the header lines that were set up by |
906 | statements in an ACL to the list of headers in memory. It is done in two stages | |
907 | like this, because when the ACL for RCPT is running, the other headers have not | |
908 | yet been received. This function is called twice; once just before running the | |
909 | DATA ACL, and once after. This is so that header lines added by MAIL or RCPT | |
910 | are visible to the DATA ACL. | |
059ec3d9 PH |
911 | |
912 | Originally these header lines were added at the end. Now there is support for | |
913 | three different places: top, bottom, and after the Received: header(s). There | |
914 | will always be at least one Received: header, even if it is marked deleted, and | |
915 | even if something else has been put in front of it. | |
916 | ||
917 | Arguments: | |
918 | acl_name text to identify which ACL | |
919 | ||
920 | Returns: nothing | |
921 | */ | |
922 | ||
923 | static void | |
924 | add_acl_headers(uschar *acl_name) | |
925 | { | |
926 | header_line *h, *next; | |
927 | header_line *last_received = NULL; | |
928 | ||
71fafd95 | 929 | if (acl_added_headers == NULL) return; |
059ec3d9 PH |
930 | DEBUG(D_receive|D_acl) debug_printf(">>Headers added by %s ACL:\n", acl_name); |
931 | ||
71fafd95 | 932 | for (h = acl_added_headers; h != NULL; h = next) |
059ec3d9 PH |
933 | { |
934 | next = h->next; | |
935 | ||
936 | switch(h->type) | |
937 | { | |
938 | case htype_add_top: | |
939 | h->next = header_list; | |
940 | header_list = h; | |
941 | DEBUG(D_receive|D_acl) debug_printf(" (at top)"); | |
942 | break; | |
943 | ||
944 | case htype_add_rec: | |
945 | if (last_received == NULL) | |
946 | { | |
947 | last_received = header_list; | |
948 | while (!header_testname(last_received, US"Received", 8, FALSE)) | |
949 | last_received = last_received->next; | |
950 | while (last_received->next != NULL && | |
951 | header_testname(last_received->next, US"Received", 8, FALSE)) | |
952 | last_received = last_received->next; | |
953 | } | |
954 | h->next = last_received->next; | |
955 | last_received->next = h; | |
956 | DEBUG(D_receive|D_acl) debug_printf(" (after Received:)"); | |
957 | break; | |
958 | ||
8523533c TK |
959 | case htype_add_rfc: |
960 | /* add header before any header which is NOT Received: or Resent- */ | |
961 | last_received = header_list; | |
962 | while ( (last_received->next != NULL) && | |
963 | ( (header_testname(last_received->next, US"Received", 8, FALSE)) || | |
964 | (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) ) | |
965 | last_received = last_received->next; | |
966 | /* last_received now points to the last Received: or Resent-* header | |
967 | in an uninterrupted chain of those header types (seen from the beginning | |
968 | of all headers. Our current header must follow it. */ | |
969 | h->next = last_received->next; | |
970 | last_received->next = h; | |
8e669ac1 | 971 | DEBUG(D_receive|D_acl) debug_printf(" (before any non-Received: or Resent-*: header)"); |
8523533c TK |
972 | break; |
973 | ||
059ec3d9 PH |
974 | default: |
975 | h->next = NULL; | |
976 | header_last->next = h; | |
977 | break; | |
978 | } | |
979 | ||
980 | if (h->next == NULL) header_last = h; | |
981 | ||
982 | /* Check for one of the known header types (From:, To:, etc.) though in | |
983 | practice most added headers are going to be "other". Lower case | |
984 | identification letters are never stored with the header; they are used | |
985 | for existence tests when messages are received. So discard any lower case | |
986 | flag values. */ | |
987 | ||
988 | h->type = header_checkname(h, FALSE); | |
989 | if (h->type >= 'a') h->type = htype_other; | |
990 | ||
991 | DEBUG(D_receive|D_acl) debug_printf(" %s", header_last->text); | |
992 | } | |
993 | ||
71fafd95 | 994 | acl_added_headers = NULL; |
059ec3d9 PH |
995 | DEBUG(D_receive|D_acl) debug_printf(">>\n"); |
996 | } | |
997 | ||
998 | ||
999 | ||
1000 | /************************************************* | |
1001 | * Add host information for log line * | |
1002 | *************************************************/ | |
1003 | ||
1004 | /* Called for acceptance and rejecting log lines. This adds information about | |
1005 | the calling host to a string that is being built dynamically. | |
1006 | ||
1007 | Arguments: | |
1008 | s the dynamic string | |
1009 | sizeptr points to the size variable | |
1010 | ptrptr points to the pointer variable | |
1011 | ||
1012 | Returns: the extended string | |
1013 | */ | |
1014 | ||
1015 | static uschar * | |
1016 | add_host_info_for_log(uschar *s, int *sizeptr, int *ptrptr) | |
1017 | { | |
1018 | if (sender_fullhost != NULL) | |
1019 | { | |
1020 | s = string_append(s, sizeptr, ptrptr, 2, US" H=", sender_fullhost); | |
1021 | if ((log_extra_selector & LX_incoming_interface) != 0 && | |
1022 | interface_address != NULL) | |
1023 | { | |
1024 | uschar *ss = string_sprintf(" I=[%s]:%d", interface_address, | |
1025 | interface_port); | |
1026 | s = string_cat(s, sizeptr, ptrptr, ss, Ustrlen(ss)); | |
1027 | } | |
1028 | } | |
1029 | if (sender_ident != NULL) | |
1030 | s = string_append(s, sizeptr, ptrptr, 2, US" U=", sender_ident); | |
1031 | if (received_protocol != NULL) | |
1032 | s = string_append(s, sizeptr, ptrptr, 2, US" P=", received_protocol); | |
1033 | return s; | |
1034 | } | |
1035 | ||
1036 | ||
1037 | ||
63955bf2 | 1038 | #ifdef WITH_CONTENT_SCAN |
059ec3d9 | 1039 | |
54cdb463 PH |
1040 | /************************************************* |
1041 | * Run the MIME ACL on a message * | |
1042 | *************************************************/ | |
1043 | ||
1044 | /* This code is in a subroutine so that it can be used for both SMTP | |
1045 | and non-SMTP messages. It is called with a non-NULL ACL pointer. | |
1046 | ||
1047 | Arguments: | |
1048 | acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime) | |
1049 | smtp_yield_ptr Set FALSE to kill messages after dropped connection | |
1050 | smtp_reply_ptr Where SMTP reply is being built | |
1051 | blackholed_by_ptr Where "blackholed by" message is being built | |
1052 | ||
1053 | Returns: TRUE to carry on; FALSE to abandon the message | |
1054 | */ | |
1055 | ||
1056 | static BOOL | |
1057 | run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr, | |
1058 | uschar **blackholed_by_ptr) | |
1059 | { | |
1060 | FILE *mbox_file; | |
1061 | uschar rfc822_file_path[2048]; | |
1062 | unsigned long mbox_size; | |
1063 | header_line *my_headerlist; | |
1064 | uschar *user_msg, *log_msg; | |
1065 | int mime_part_count_buffer = -1; | |
1066 | int rc; | |
1067 | ||
1068 | memset(CS rfc822_file_path,0,2048); | |
1069 | ||
1070 | /* check if it is a MIME message */ | |
1071 | my_headerlist = header_list; | |
4e88a19f PH |
1072 | while (my_headerlist != NULL) |
1073 | { | |
54cdb463 | 1074 | /* skip deleted headers */ |
4e88a19f PH |
1075 | if (my_headerlist->type == '*') |
1076 | { | |
54cdb463 PH |
1077 | my_headerlist = my_headerlist->next; |
1078 | continue; | |
4e88a19f PH |
1079 | } |
1080 | if (strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0) | |
1081 | { | |
54cdb463 PH |
1082 | DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n"); |
1083 | goto DO_MIME_ACL; | |
4e88a19f | 1084 | } |
54cdb463 | 1085 | my_headerlist = my_headerlist->next; |
4e88a19f | 1086 | } |
54cdb463 PH |
1087 | |
1088 | DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n"); | |
1089 | return TRUE; | |
1090 | ||
1091 | DO_MIME_ACL: | |
1092 | /* make sure the eml mbox file is spooled up */ | |
1093 | mbox_file = spool_mbox(&mbox_size); | |
1094 | if (mbox_file == NULL) { | |
1095 | /* error while spooling */ | |
1096 | log_write(0, LOG_MAIN|LOG_PANIC, | |
1097 | "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected."); | |
1098 | Uunlink(spool_name); | |
1099 | unspool_mbox(); | |
a5bd321b | 1100 | smtp_respond(US"451", 3, TRUE, US"temporary local problem"); |
54cdb463 PH |
1101 | message_id[0] = 0; /* Indicate no message accepted */ |
1102 | *smtp_reply_ptr = US""; /* Indicate reply already sent */ | |
1103 | return FALSE; /* Indicate skip to end of receive function */ | |
1104 | }; | |
1105 | ||
1106 | mime_is_rfc822 = 0; | |
1107 | ||
1108 | MIME_ACL_CHECK: | |
1109 | mime_part_count = -1; | |
1110 | rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg); | |
f1e894f3 | 1111 | (void)fclose(mbox_file); |
54cdb463 | 1112 | |
4e88a19f PH |
1113 | if (Ustrlen(rfc822_file_path) > 0) |
1114 | { | |
54cdb463 PH |
1115 | mime_part_count = mime_part_count_buffer; |
1116 | ||
4e88a19f PH |
1117 | if (unlink(CS rfc822_file_path) == -1) |
1118 | { | |
54cdb463 PH |
1119 | log_write(0, LOG_PANIC, |
1120 | "acl_smtp_mime: can't unlink RFC822 spool file, skipping."); | |
1121 | goto END_MIME_ACL; | |
4e88a19f PH |
1122 | } |
1123 | } | |
54cdb463 PH |
1124 | |
1125 | /* check if we must check any message/rfc822 attachments */ | |
4e88a19f PH |
1126 | if (rc == OK) |
1127 | { | |
54cdb463 PH |
1128 | uschar temp_path[1024]; |
1129 | int n; | |
1130 | struct dirent *entry; | |
1131 | DIR *tempdir; | |
1132 | ||
b07e6aa3 PH |
1133 | (void)string_format(temp_path, 1024, "%s/scan/%s", spool_directory, |
1134 | message_id); | |
54cdb463 | 1135 | |
4e88a19f PH |
1136 | tempdir = opendir(CS temp_path); |
1137 | n = 0; | |
1138 | do | |
1139 | { | |
1140 | entry = readdir(tempdir); | |
1141 | if (entry == NULL) break; | |
1142 | if (strncmpic(US entry->d_name,US"__rfc822_",9) == 0) | |
1143 | { | |
b07e6aa3 | 1144 | (void)string_format(rfc822_file_path, 2048,"%s/scan/%s/%s", spool_directory, message_id, entry->d_name); |
4e88a19f PH |
1145 | debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n", rfc822_file_path); |
1146 | break; | |
1147 | } | |
1148 | } while (1); | |
1149 | closedir(tempdir); | |
54cdb463 | 1150 | |
4e88a19f PH |
1151 | if (entry != NULL) |
1152 | { | |
c58b88df | 1153 | mbox_file = Ufopen(rfc822_file_path,"rb"); |
4e88a19f PH |
1154 | if (mbox_file == NULL) |
1155 | { | |
54cdb463 PH |
1156 | log_write(0, LOG_PANIC, |
1157 | "acl_smtp_mime: can't open RFC822 spool file, skipping."); | |
1158 | unlink(CS rfc822_file_path); | |
1159 | goto END_MIME_ACL; | |
4e88a19f | 1160 | } |
54cdb463 PH |
1161 | /* set RFC822 expansion variable */ |
1162 | mime_is_rfc822 = 1; | |
1163 | mime_part_count_buffer = mime_part_count; | |
1164 | goto MIME_ACL_CHECK; | |
4e88a19f PH |
1165 | } |
1166 | } | |
54cdb463 PH |
1167 | |
1168 | END_MIME_ACL: | |
1169 | add_acl_headers(US"MIME"); | |
1170 | if (rc == DISCARD) | |
1171 | { | |
1172 | recipients_count = 0; | |
1173 | *blackholed_by_ptr = US"MIME ACL"; | |
1174 | } | |
1175 | else if (rc != OK) | |
1176 | { | |
1177 | Uunlink(spool_name); | |
1178 | unspool_mbox(); | |
1179 | if (smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0) | |
1180 | *smtp_yield_ptr = FALSE; /* No more messsages after dropped connection */ | |
1181 | *smtp_reply_ptr = US""; /* Indicate reply already sent */ | |
1182 | message_id[0] = 0; /* Indicate no message accepted */ | |
1183 | return FALSE; /* Cause skip to end of receive function */ | |
4e88a19f | 1184 | } |
54cdb463 PH |
1185 | |
1186 | return TRUE; | |
1187 | } | |
1188 | ||
63955bf2 | 1189 | #endif /* WITH_CONTENT_SCAN */ |
54cdb463 PH |
1190 | |
1191 | ||
059ec3d9 PH |
1192 | /************************************************* |
1193 | * Receive message * | |
1194 | *************************************************/ | |
1195 | ||
1196 | /* Receive a message on the given input, and put it into a pair of spool files. | |
1197 | Either a non-null list of recipients, or the extract flag will be true, or | |
1198 | both. The flag sender_local is true for locally generated messages. The flag | |
1199 | submission_mode is true if an ACL has obeyed "control = submission". The flag | |
8800895a PH |
1200 | suppress_local_fixups is true if an ACL has obeyed "control = |
1201 | suppress_local_fixups". The flag smtp_input is true if the message is to be | |
1202 | handled using SMTP conventions about termination and lines starting with dots. | |
1203 | For non-SMTP messages, dot_ends is true for dot-terminated messages. | |
059ec3d9 PH |
1204 | |
1205 | If a message was successfully read, message_id[0] will be non-zero. | |
1206 | ||
1207 | The general actions of this function are: | |
1208 | ||
1209 | . Read the headers of the message (if any) into a chain of store | |
1210 | blocks. | |
1211 | ||
1212 | . If there is a "sender:" header and the message is locally originated, | |
69358f02 PH |
1213 | throw it away, unless the caller is trusted, or unless |
1214 | active_local_sender_retain is set - which can only happen if | |
1215 | active_local_from_check is false. | |
059ec3d9 PH |
1216 | |
1217 | . If recipients are to be extracted from the message, build the | |
1218 | recipients list from the headers, removing any that were on the | |
1219 | original recipients list (unless extract_addresses_remove_arguments is | |
1220 | false), and at the same time, remove any bcc header that may be present. | |
1221 | ||
1222 | . Get the spool file for the data, sort out its unique name, open | |
1223 | and lock it (but don't give it the name yet). | |
1224 | ||
1225 | . Generate a "Message-Id" header if the message doesn't have one, for | |
1226 | locally-originated messages. | |
1227 | ||
1228 | . Generate a "Received" header. | |
1229 | ||
1230 | . Ensure the recipients list is fully qualified and rewritten if necessary. | |
1231 | ||
1232 | . If there are any rewriting rules, apply them to the sender address | |
1233 | and also to the headers. | |
1234 | ||
1235 | . If there is no from: header, generate one, for locally-generated messages | |
1236 | and messages in "submission mode" only. | |
1237 | ||
1238 | . If the sender is local, check that from: is correct, and if not, generate | |
1239 | a Sender: header, unless message comes from a trusted caller, or this | |
69358f02 | 1240 | feature is disabled by active_local_from_check being false. |
059ec3d9 PH |
1241 | |
1242 | . If there is no "date" header, generate one, for locally-originated | |
1243 | or submission mode messages only. | |
1244 | ||
1245 | . Copy the rest of the input, or up to a terminating "." if in SMTP or | |
1246 | dot_ends mode, to the data file. Leave it open, to hold the lock. | |
1247 | ||
1248 | . Write the envelope and the headers to a new file. | |
1249 | ||
1250 | . Set the name for the header file; close it. | |
1251 | ||
1252 | . Set the name for the data file; close it. | |
1253 | ||
1254 | Because this function can potentially be called many times in a single | |
1255 | SMTP connection, all store should be got by store_get(), so that it will be | |
1256 | automatically retrieved after the message is accepted. | |
1257 | ||
1258 | FUDGE: It seems that sites on the net send out messages with just LF | |
1259 | terminators, despite the warnings in the RFCs, and other MTAs handle this. So | |
1260 | we make the CRs optional in all cases. | |
1261 | ||
1262 | July 2003: Bare CRs in messages, especially in header lines, cause trouble. A | |
1263 | new regime is now in place in which bare CRs in header lines are turned into LF | |
1264 | followed by a space, so as not to terminate the header line. | |
1265 | ||
1266 | February 2004: A bare LF in a header line in a message whose first line was | |
1267 | terminated by CRLF is treated in the same way as a bare CR. | |
1268 | ||
1269 | Arguments: | |
1270 | extract_recip TRUE if recipients are to be extracted from the message's | |
1271 | headers | |
1272 | ||
1273 | Returns: TRUE there are more messages to be read (SMTP input) | |
1274 | FALSE there are no more messages to be read (non-SMTP input | |
1275 | or SMTP connection collapsed, or other failure) | |
1276 | ||
1277 | When reading a message for filter testing, the returned value indicates | |
1278 | whether the headers (which is all that is read) were terminated by '.' or | |
1279 | not. */ | |
1280 | ||
1281 | BOOL | |
1282 | receive_msg(BOOL extract_recip) | |
1283 | { | |
1284 | int i, rc; | |
1285 | int msg_size = 0; | |
1286 | int process_info_len = Ustrlen(process_info); | |
1287 | int error_rc = (error_handling == ERRORS_SENDER)? | |
1288 | errors_sender_rc : EXIT_FAILURE; | |
1289 | int header_size = 256; | |
1290 | int start, end, domain, size, sptr; | |
1291 | int id_resolution; | |
1292 | int had_zero = 0; | |
d677b2f2 | 1293 | int prevlines_length = 0; |
059ec3d9 PH |
1294 | |
1295 | register int ptr = 0; | |
1296 | ||
1297 | BOOL contains_resent_headers = FALSE; | |
1298 | BOOL extracted_ignored = FALSE; | |
1299 | BOOL first_line_ended_crlf = TRUE_UNSET; | |
1300 | BOOL smtp_yield = TRUE; | |
1301 | BOOL yield = FALSE; | |
1302 | ||
1303 | BOOL resents_exist = FALSE; | |
1304 | uschar *resent_prefix = US""; | |
1305 | uschar *blackholed_by = NULL; | |
04f7d5b9 | 1306 | uschar *blackhole_log_msg = US""; |
059ec3d9 PH |
1307 | |
1308 | flock_t lock_data; | |
1309 | error_block *bad_addresses = NULL; | |
1310 | ||
1311 | uschar *frozen_by = NULL; | |
1312 | uschar *queued_by = NULL; | |
1313 | ||
1314 | uschar *errmsg, *s; | |
1315 | struct stat statbuf; | |
1316 | ||
4e88a19f | 1317 | /* Final message to give to SMTP caller, and messages from ACLs */ |
059ec3d9 PH |
1318 | |
1319 | uschar *smtp_reply = NULL; | |
4e88a19f | 1320 | uschar *user_msg, *log_msg; |
059ec3d9 PH |
1321 | |
1322 | /* Working header pointers */ | |
1323 | ||
1324 | header_line *h, *next; | |
1325 | ||
2cbb4081 | 1326 | /* Flags for noting the existence of certain headers (only one left) */ |
059ec3d9 PH |
1327 | |
1328 | BOOL date_header_exists = FALSE; | |
1329 | ||
1330 | /* Pointers to receive the addresses of headers whose contents we need. */ | |
1331 | ||
1332 | header_line *from_header = NULL; | |
1333 | header_line *subject_header = NULL; | |
1334 | header_line *msgid_header = NULL; | |
1335 | header_line *received_header; | |
1336 | ||
1337 | /* Variables for use when building the Received: header. */ | |
1338 | ||
1339 | uschar *received; | |
1340 | uschar *timestamp; | |
1341 | int tslen; | |
1342 | ||
1343 | /* Release any open files that might have been cached while preparing to | |
1344 | accept the message - e.g. by verifying addresses - because reading a message | |
1345 | might take a fair bit of real time. */ | |
1346 | ||
1347 | search_tidyup(); | |
1348 | ||
1349 | /* Initialize the chain of headers by setting up a place-holder for Received: | |
1350 | header. Temporarily mark it as "old", i.e. not to be used. We keep header_last | |
1351 | pointing to the end of the chain to make adding headers simple. */ | |
1352 | ||
1353 | received_header = header_list = header_last = store_get(sizeof(header_line)); | |
1354 | header_list->next = NULL; | |
1355 | header_list->type = htype_old; | |
1356 | header_list->text = NULL; | |
1357 | header_list->slen = 0; | |
1358 | ||
1359 | /* Control block for the next header to be read. */ | |
1360 | ||
1361 | next = store_get(sizeof(header_line)); | |
1362 | next->text = store_get(header_size); | |
1363 | ||
1364 | /* Initialize message id to be null (indicating no message read), and the | |
1365 | header names list to be the normal list. Indicate there is no data file open | |
1366 | yet, initialize the size and warning count, and deal with no size limit. */ | |
1367 | ||
1368 | message_id[0] = 0; | |
1369 | data_file = NULL; | |
1370 | data_fd = -1; | |
1371 | spool_name[0] = 0; | |
1372 | message_size = 0; | |
1373 | warning_count = 0; | |
d677b2f2 | 1374 | received_count = 1; /* For the one we will add */ |
059ec3d9 PH |
1375 | |
1376 | if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX; | |
1377 | ||
2e0c1448 | 1378 | /* While reading the message, the following counts are computed. */ |
059ec3d9 | 1379 | |
d677b2f2 PH |
1380 | message_linecount = body_linecount = body_zerocount = |
1381 | max_received_linelength = 0; | |
059ec3d9 | 1382 | |
fb2274d4 TK |
1383 | #ifdef EXPERIMENTAL_DOMAINKEYS |
1384 | /* Call into DK to set up the context. Check if DK is to be run are carried out | |
1385 | inside dk_exim_verify_init(). */ | |
1386 | dk_exim_verify_init(); | |
1387 | #endif | |
1388 | ||
059ec3d9 PH |
1389 | /* Remember the time of reception. Exim uses time+pid for uniqueness of message |
1390 | ids, and fractions of a second are required. See the comments that precede the | |
1391 | message id creation below. */ | |
1392 | ||
1393 | (void)gettimeofday(&message_id_tv, NULL); | |
1394 | ||
1395 | /* For other uses of the received time we can operate with granularity of one | |
1396 | second, and for that we use the global variable received_time. This is for | |
1397 | things like ultimate message timeouts. */ | |
1398 | ||
1399 | received_time = message_id_tv.tv_sec; | |
1400 | ||
1401 | /* If SMTP input, set the special handler for timeouts. The alarm() calls | |
1402 | happen in the smtp_getc() function when it refills its buffer. */ | |
1403 | ||
1404 | if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler); | |
1405 | ||
1406 | /* If not SMTP input, timeout happens only if configured, and we just set a | |
1407 | single timeout for the whole message. */ | |
1408 | ||
1409 | else if (receive_timeout > 0) | |
1410 | { | |
1411 | os_non_restarting_signal(SIGALRM, data_timeout_handler); | |
1412 | alarm(receive_timeout); | |
1413 | } | |
1414 | ||
1415 | /* SIGTERM and SIGINT are caught always. */ | |
1416 | ||
1417 | signal(SIGTERM, data_sigterm_sigint_handler); | |
1418 | signal(SIGINT, data_sigterm_sigint_handler); | |
1419 | ||
1420 | /* Header lines in messages are not supposed to be very long, though when | |
1421 | unfolded, to: and cc: headers can take up a lot of store. We must also cope | |
1422 | with the possibility of junk being thrown at us. Start by getting 256 bytes for | |
1423 | storing the header, and extend this as necessary using string_cat(). | |
1424 | ||
1425 | To cope with total lunacies, impose an upper limit on the length of the header | |
1426 | section of the message, as otherwise the store will fill up. We must also cope | |
1427 | with the possibility of binary zeros in the data. Hence we cannot use fgets(). | |
1428 | Folded header lines are joined into one string, leaving the '\n' characters | |
1429 | inside them, so that writing them out reproduces the input. | |
1430 | ||
1431 | Loop for each character of each header; the next structure for chaining the | |
1432 | header is set up already, with ptr the offset of the next character in | |
1433 | next->text. */ | |
1434 | ||
1435 | for (;;) | |
1436 | { | |
fb2274d4 | 1437 | int ch = (RECEIVE_GETC)(); |
059ec3d9 PH |
1438 | |
1439 | /* If we hit EOF on a SMTP connection, it's an error, since incoming | |
1440 | SMTP must have a correct "." terminator. */ | |
1441 | ||
1442 | if (ch == EOF && smtp_input /* && !smtp_batched_input */) | |
1443 | { | |
1444 | smtp_reply = handle_lost_connection(US" (header)"); | |
1445 | smtp_yield = FALSE; | |
1446 | goto TIDYUP; /* Skip to end of function */ | |
1447 | } | |
1448 | ||
1449 | /* See if we are at the current header's size limit - there must be at least | |
1450 | four bytes left. This allows for the new character plus a zero, plus two for | |
1451 | extra insertions when we are playing games with dots and carriage returns. If | |
1452 | we are at the limit, extend the text buffer. This could have been done | |
1453 | automatically using string_cat() but because this is a tightish loop storing | |
1454 | only one character at a time, we choose to do it inline. Normally | |
1455 | store_extend() will be able to extend the block; only at the end of a big | |
1456 | store block will a copy be needed. To handle the case of very long headers | |
1457 | (and sometimes lunatic messages can have ones that are 100s of K long) we | |
1458 | call store_release() for strings that have been copied - if the string is at | |
1459 | the start of a block (and therefore the only thing in it, because we aren't | |
1460 | doing any other gets), the block gets freed. We can only do this because we | |
1461 | know there are no other calls to store_get() going on. */ | |
1462 | ||
1463 | if (ptr >= header_size - 4) | |
1464 | { | |
1465 | int oldsize = header_size; | |
1466 | /* header_size += 256; */ | |
1467 | header_size *= 2; | |
1468 | if (!store_extend(next->text, oldsize, header_size)) | |
1469 | { | |
1470 | uschar *newtext = store_get(header_size); | |
1471 | memcpy(newtext, next->text, ptr); | |
1472 | store_release(next->text); | |
1473 | next->text = newtext; | |
1474 | } | |
1475 | } | |
1476 | ||
1477 | /* Cope with receiving a binary zero. There is dispute about whether | |
1478 | these should be allowed in RFC 822 messages. The middle view is that they | |
1479 | should not be allowed in headers, at least. Exim takes this attitude at | |
1480 | the moment. We can't just stomp on them here, because we don't know that | |
1481 | this line is a header yet. Set a flag to cause scanning later. */ | |
1482 | ||
1483 | if (ch == 0) had_zero++; | |
1484 | ||
1485 | /* Test for termination. Lines in remote SMTP are terminated by CRLF, while | |
1486 | those from data files use just LF. Treat LF in local SMTP input as a | |
1487 | terminator too. Treat EOF as a line terminator always. */ | |
1488 | ||
1489 | if (ch == EOF) goto EOL; | |
1490 | ||
1491 | /* FUDGE: There are sites out there that don't send CRs before their LFs, and | |
1492 | other MTAs accept this. We are therefore forced into this "liberalisation" | |
1493 | too, so we accept LF as a line terminator whatever the source of the message. | |
1494 | However, if the first line of the message ended with a CRLF, we treat a bare | |
1495 | LF specially by inserting a white space after it to ensure that the header | |
1496 | line is not terminated. */ | |
1497 | ||
1498 | if (ch == '\n') | |
1499 | { | |
1500 | if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE; | |
fb2274d4 | 1501 | else if (first_line_ended_crlf) RECEIVE_UNGETC(' '); |
059ec3d9 PH |
1502 | goto EOL; |
1503 | } | |
1504 | ||
1505 | /* This is not the end of the line. If this is SMTP input and this is | |
1506 | the first character in the line and it is a "." character, ignore it. | |
1507 | This implements the dot-doubling rule, though header lines starting with | |
1508 | dots aren't exactly common. They are legal in RFC 822, though. If the | |
1509 | following is CRLF or LF, this is the line that that terminates the | |
1510 | entire message. We set message_ended to indicate this has happened (to | |
1511 | prevent further reading), and break out of the loop, having freed the | |
1512 | empty header, and set next = NULL to indicate no data line. */ | |
1513 | ||
1514 | if (ptr == 0 && ch == '.' && (smtp_input || dot_ends)) | |
1515 | { | |
fb2274d4 | 1516 | ch = (RECEIVE_GETC)(); |
059ec3d9 PH |
1517 | if (ch == '\r') |
1518 | { | |
fb2274d4 | 1519 | ch = (RECEIVE_GETC)(); |
059ec3d9 PH |
1520 | if (ch != '\n') |
1521 | { | |
fb2274d4 | 1522 | RECEIVE_UNGETC(ch); |
059ec3d9 PH |
1523 | ch = '\r'; /* Revert to CR */ |
1524 | } | |
1525 | } | |
1526 | if (ch == '\n') | |
1527 | { | |
1528 | message_ended = END_DOT; | |
1529 | store_reset(next); | |
1530 | next = NULL; | |
1531 | break; /* End character-reading loop */ | |
1532 | } | |
1533 | ||
1534 | /* For non-SMTP input, the dot at the start of the line was really a data | |
1535 | character. What is now in ch is the following character. We guaranteed | |
1536 | enough space for this above. */ | |
1537 | ||
1538 | if (!smtp_input) | |
1539 | { | |
1540 | next->text[ptr++] = '.'; | |
1541 | message_size++; | |
1542 | } | |
1543 | } | |
1544 | ||
1545 | /* If CR is immediately followed by LF, end the line, ignoring the CR, and | |
1546 | remember this case if this is the first line ending. */ | |
1547 | ||
1548 | if (ch == '\r') | |
1549 | { | |
fb2274d4 | 1550 | ch = (RECEIVE_GETC)(); |
059ec3d9 PH |
1551 | if (ch == '\n') |
1552 | { | |
1553 | if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE; | |
1554 | goto EOL; | |
1555 | } | |
1556 | ||
1557 | /* Otherwise, put back the character after CR, and turn the bare CR | |
1558 | into LF SP. */ | |
1559 | ||
fb2274d4 | 1560 | ch = (RECEIVE_UNGETC)(ch); |
059ec3d9 PH |
1561 | next->text[ptr++] = '\n'; |
1562 | message_size++; | |
1563 | ch = ' '; | |
1564 | } | |
1565 | ||
1566 | /* We have a data character for the header line. */ | |
1567 | ||
1568 | next->text[ptr++] = ch; /* Add to buffer */ | |
1569 | message_size++; /* Total message size so far */ | |
1570 | ||
1571 | /* Handle failure due to a humungously long header section. The >= allows | |
1572 | for the terminating \n. Add what we have so far onto the headers list so | |
1573 | that it gets reflected in any error message, and back up the just-read | |
1574 | character. */ | |
1575 | ||
1576 | if (message_size >= header_maxsize) | |
1577 | { | |
1578 | next->text[ptr] = 0; | |
1579 | next->slen = ptr; | |
1580 | next->type = htype_other; | |
1581 | next->next = NULL; | |
1582 | header_last->next = next; | |
1583 | header_last = next; | |
1584 | ||
1585 | log_write(0, LOG_MAIN, "ridiculously long message header received from " | |
1586 | "%s (more than %d characters): message abandoned", | |
1587 | sender_host_unknown? sender_ident : sender_fullhost, header_maxsize); | |
1588 | ||
1589 | if (smtp_input) | |
1590 | { | |
1591 | smtp_reply = US"552 Message header is ridiculously long"; | |
1592 | receive_swallow_smtp(); | |
1593 | goto TIDYUP; /* Skip to end of function */ | |
1594 | } | |
1595 | ||
1596 | else | |
1597 | { | |
1598 | give_local_error(ERRMESS_VLONGHEADER, | |
1599 | string_sprintf("message header longer than %d characters received: " | |
1600 | "message not accepted", header_maxsize), US"", error_rc, stdin, | |
1601 | header_list->next); | |
1602 | /* Does not return */ | |
1603 | } | |
1604 | } | |
1605 | ||
1606 | continue; /* With next input character */ | |
1607 | ||
1608 | /* End of header line reached */ | |
1609 | ||
1610 | EOL: | |
2e0c1448 PH |
1611 | |
1612 | /* Keep track of lines for BSMTP errors and overall message_linecount. */ | |
1613 | ||
1614 | receive_linecount++; | |
1615 | message_linecount++; | |
059ec3d9 | 1616 | |
d677b2f2 PH |
1617 | /* Keep track of maximum line length */ |
1618 | ||
1619 | if (ptr - prevlines_length > max_received_linelength) | |
1620 | max_received_linelength = ptr - prevlines_length; | |
1621 | prevlines_length = ptr + 1; | |
1622 | ||
059ec3d9 PH |
1623 | /* Now put in the terminating newline. There is always space for |
1624 | at least two more characters. */ | |
1625 | ||
1626 | next->text[ptr++] = '\n'; | |
1627 | message_size++; | |
1628 | ||
1629 | /* A blank line signals the end of the headers; release the unwanted | |
1630 | space and set next to NULL to indicate this. */ | |
1631 | ||
1632 | if (ptr == 1) | |
1633 | { | |
1634 | store_reset(next); | |
1635 | next = NULL; | |
1636 | break; | |
1637 | } | |
1638 | ||
1639 | /* There is data in the line; see if the next input character is a | |
1640 | whitespace character. If it is, we have a continuation of this header line. | |
1641 | There is always space for at least one character at this point. */ | |
1642 | ||
1643 | if (ch != EOF) | |
1644 | { | |
fb2274d4 | 1645 | int nextch = (RECEIVE_GETC)(); |
059ec3d9 PH |
1646 | if (nextch == ' ' || nextch == '\t') |
1647 | { | |
1648 | next->text[ptr++] = nextch; | |
1649 | message_size++; | |
1650 | continue; /* Iterate the loop */ | |
1651 | } | |
fb2274d4 | 1652 | else if (nextch != EOF) (RECEIVE_UNGETC)(nextch); /* For next time */ |
059ec3d9 PH |
1653 | else ch = EOF; /* Cause main loop to exit at end */ |
1654 | } | |
1655 | ||
1656 | /* We have got to the real line end. Terminate the string and release store | |
1657 | beyond it. If it turns out to be a real header, internal binary zeros will | |
1658 | be squashed later. */ | |
1659 | ||
1660 | next->text[ptr] = 0; | |
1661 | next->slen = ptr; | |
1662 | store_reset(next->text + ptr + 1); | |
1663 | ||
1664 | /* Check the running total size against the overall message size limit. We | |
1665 | don't expect to fail here, but if the overall limit is set less than MESSAGE_ | |
1666 | MAXSIZE and a big header is sent, we want to catch it. Just stop reading | |
1667 | headers - the code to read the body will then also hit the buffer. */ | |
1668 | ||
1669 | if (message_size > thismessage_size_limit) break; | |
1670 | ||
1671 | /* A line that is not syntactically correct for a header also marks | |
1672 | the end of the headers. In this case, we leave next containing the | |
1673 | first data line. This might actually be several lines because of the | |
1674 | continuation logic applied above, but that doesn't matter. | |
1675 | ||
1676 | It turns out that smail, and presumably sendmail, accept leading lines | |
1677 | of the form | |
1678 | ||
1679 | From ph10 Fri Jan 5 12:35 GMT 1996 | |
1680 | ||
1681 | in messages. The "mail" command on Solaris 2 sends such lines. I cannot | |
1682 | find any documentation of this, but for compatibility it had better be | |
1683 | accepted. Exim restricts it to the case of non-smtp messages, and | |
1684 | treats it as an alternative to the -f command line option. Thus it is | |
1685 | ignored except for trusted users or filter testing. Otherwise it is taken | |
1686 | as the sender address, unless -f was used (sendmail compatibility). | |
1687 | ||
1688 | It further turns out that some UUCPs generate the From_line in a different | |
1689 | format, e.g. | |
1690 | ||
1691 | From ph10 Fri, 7 Jan 97 14:00:00 GMT | |
1692 | ||
1693 | The regex for matching these things is now capable of recognizing both | |
1694 | formats (including 2- and 4-digit years in the latter). In fact, the regex | |
1695 | is now configurable, as is the expansion string to fish out the sender. | |
1696 | ||
1697 | Even further on it has been discovered that some broken clients send | |
1698 | these lines in SMTP messages. There is now an option to ignore them from | |
1699 | specified hosts or networks. Sigh. */ | |
1700 | ||
1701 | if (header_last == header_list && | |
1702 | (!smtp_input | |
1703 | || | |
1704 | (sender_host_address != NULL && | |
1705 | verify_check_host(&ignore_fromline_hosts) == OK) | |
1706 | || | |
1707 | (sender_host_address == NULL && ignore_fromline_local) | |
1708 | ) && | |
1709 | regex_match_and_setup(regex_From, next->text, 0, -1)) | |
1710 | { | |
1711 | if (!sender_address_forced) | |
1712 | { | |
1713 | uschar *uucp_sender = expand_string(uucp_from_sender); | |
1714 | if (uucp_sender == NULL) | |
1715 | { | |
1716 | log_write(0, LOG_MAIN|LOG_PANIC, | |
1717 | "expansion of \"%s\" failed after matching " | |
1718 | "\"From \" line: %s", uucp_from_sender, expand_string_message); | |
1719 | } | |
1720 | else | |
1721 | { | |
1722 | int start, end, domain; | |
1723 | uschar *errmess; | |
1724 | uschar *newsender = parse_extract_address(uucp_sender, &errmess, | |
1725 | &start, &end, &domain, TRUE); | |
1726 | if (newsender != NULL) | |
1727 | { | |
1728 | if (domain == 0 && newsender[0] != 0) | |
1729 | newsender = rewrite_address_qualify(newsender, FALSE); | |
1730 | ||
f05da2e8 | 1731 | if (filter_test != FTEST_NONE || receive_check_set_sender(newsender)) |
059ec3d9 PH |
1732 | { |
1733 | sender_address = newsender; | |
1734 | ||
f05da2e8 | 1735 | if (trusted_caller || filter_test != FTEST_NONE) |
059ec3d9 PH |
1736 | { |
1737 | authenticated_sender = NULL; | |
1738 | originator_name = US""; | |
1739 | sender_local = FALSE; | |
1740 | } | |
1741 | ||
f05da2e8 | 1742 | if (filter_test != FTEST_NONE) |
059ec3d9 PH |
1743 | printf("Sender taken from \"From \" line\n"); |
1744 | } | |
1745 | } | |
1746 | } | |
1747 | } | |
1748 | } | |
1749 | ||
1750 | /* Not a leading "From " line. Check to see if it is a valid header line. | |
1751 | Header names may contain any non-control characters except space and colon, | |
1752 | amazingly. */ | |
1753 | ||
1754 | else | |
1755 | { | |
1756 | uschar *p = next->text; | |
1757 | ||
1758 | /* If not a valid header line, break from the header reading loop, leaving | |
1759 | next != NULL, indicating that it holds the first line of the body. */ | |
1760 | ||
1761 | if (isspace(*p)) break; | |
1762 | while (mac_isgraph(*p) && *p != ':') p++; | |
1763 | while (isspace(*p)) p++; | |
1764 | if (*p != ':') | |
1765 | { | |
1766 | body_zerocount = had_zero; | |
1767 | break; | |
1768 | } | |
1769 | ||
1770 | /* We have a valid header line. If there were any binary zeroes in | |
1771 | the line, stomp on them here. */ | |
1772 | ||
1773 | if (had_zero > 0) | |
1774 | for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?'; | |
1775 | ||
1776 | /* It is perfectly legal to have an empty continuation line | |
1777 | at the end of a header, but it is confusing to humans | |
1778 | looking at such messages, since it looks like a blank line. | |
1779 | Reduce confusion by removing redundant white space at the | |
1780 | end. We know that there is at least one printing character | |
1781 | (the ':' tested for above) so there is no danger of running | |
1782 | off the end. */ | |
1783 | ||
1784 | p = next->text + ptr - 2; | |
1785 | for (;;) | |
1786 | { | |
1787 | while (*p == ' ' || *p == '\t') p--; | |
1788 | if (*p != '\n') break; | |
1789 | ptr = (p--) - next->text + 1; | |
1790 | message_size -= next->slen - ptr; | |
1791 | next->text[ptr] = 0; | |
1792 | next->slen = ptr; | |
1793 | } | |
1794 | ||
1795 | /* Add the header to the chain */ | |
1796 | ||
1797 | next->type = htype_other; | |
1798 | next->next = NULL; | |
1799 | header_last->next = next; | |
1800 | header_last = next; | |
1801 | ||
1802 | /* Check the limit for individual line lengths. This comes after adding to | |
1803 | the chain so that the failing line is reflected if a bounce is generated | |
1804 | (for a local message). */ | |
1805 | ||
1806 | if (header_line_maxsize > 0 && next->slen > header_line_maxsize) | |
1807 | { | |
1808 | log_write(0, LOG_MAIN, "overlong message header line received from " | |
1809 | "%s (more than %d characters): message abandoned", | |
1810 | sender_host_unknown? sender_ident : sender_fullhost, | |
1811 | header_line_maxsize); | |
1812 | ||
1813 | if (smtp_input) | |
1814 | { | |
1815 | smtp_reply = US"552 A message header line is too long"; | |
1816 | receive_swallow_smtp(); | |
1817 | goto TIDYUP; /* Skip to end of function */ | |
1818 | } | |
1819 | ||
1820 | else | |
1821 | { | |
1822 | give_local_error(ERRMESS_VLONGHDRLINE, | |
1823 | string_sprintf("message header line longer than %d characters " | |
1824 | "received: message not accepted", header_line_maxsize), US"", | |
1825 | error_rc, stdin, header_list->next); | |
1826 | /* Does not return */ | |
1827 | } | |
1828 | } | |
1829 | ||
1830 | /* Note if any resent- fields exist. */ | |
1831 | ||
1832 | if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0) | |
1833 | { | |
1834 | resents_exist = TRUE; | |
1835 | resent_prefix = US"Resent-"; | |
1836 | } | |
1837 | } | |
1838 | ||
1839 | /* The line has been handled. If we have hit EOF, break out of the loop, | |
1840 | indicating no pending data line. */ | |
1841 | ||
1842 | if (ch == EOF) { next = NULL; break; } | |
1843 | ||
1844 | /* Set up for the next header */ | |
1845 | ||
1846 | header_size = 256; | |
1847 | next = store_get(sizeof(header_line)); | |
1848 | next->text = store_get(header_size); | |
1849 | ptr = 0; | |
1850 | had_zero = 0; | |
d677b2f2 | 1851 | prevlines_length = 0; |
059ec3d9 PH |
1852 | } /* Continue, starting to read the next header */ |
1853 | ||
1854 | /* At this point, we have read all the headers into a data structure in main | |
1855 | store. The first header is still the dummy placeholder for the Received: header | |
1856 | we are going to generate a bit later on. If next != NULL, it contains the first | |
1857 | data line - which terminated the headers before reaching a blank line (not the | |
1858 | normal case). */ | |
1859 | ||
1860 | DEBUG(D_receive) | |
1861 | { | |
1862 | debug_printf(">>Headers received:\n"); | |
1863 | for (h = header_list->next; h != NULL; h = h->next) | |
1864 | debug_printf("%s", h->text); | |
1865 | debug_printf("\n"); | |
1866 | } | |
1867 | ||
1868 | /* End of file on any SMTP connection is an error. If an incoming SMTP call | |
1869 | is dropped immediately after valid headers, the next thing we will see is EOF. | |
1870 | We must test for this specially, as further down the reading of the data is | |
1871 | skipped if already at EOF. */ | |
1872 | ||
1873 | if (smtp_input && (receive_feof)()) | |
1874 | { | |
1875 | smtp_reply = handle_lost_connection(US" (after header)"); | |
1876 | smtp_yield = FALSE; | |
1877 | goto TIDYUP; /* Skip to end of function */ | |
1878 | } | |
1879 | ||
1880 | /* If this is a filter test run and no headers were read, output a warning | |
1881 | in case there is a mistake in the test message. */ | |
1882 | ||
f05da2e8 | 1883 | if (filter_test != FTEST_NONE && header_list->next == NULL) |
059ec3d9 PH |
1884 | printf("Warning: no message headers read\n"); |
1885 | ||
1886 | ||
1887 | /* Scan the headers to identify them. Some are merely marked for later | |
1888 | processing; some are dealt with here. */ | |
1889 | ||
1890 | for (h = header_list->next; h != NULL; h = h->next) | |
1891 | { | |
1892 | BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0; | |
1893 | if (is_resent) contains_resent_headers = TRUE; | |
1894 | ||
1895 | switch (header_checkname(h, is_resent)) | |
1896 | { | |
059ec3d9 | 1897 | case htype_bcc: |
2cbb4081 | 1898 | h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */ |
059ec3d9 PH |
1899 | break; |
1900 | ||
059ec3d9 | 1901 | case htype_cc: |
2cbb4081 | 1902 | h->type = htype_cc; /* Both Cc: and Resent-Cc: */ |
059ec3d9 PH |
1903 | break; |
1904 | ||
1905 | /* Record whether a Date: or Resent-Date: header exists, as appropriate. */ | |
1906 | ||
1907 | case htype_date: | |
1908 | date_header_exists = !resents_exist || is_resent; | |
1909 | break; | |
1910 | ||
1911 | /* Same comments as about Return-Path: below. */ | |
1912 | ||
1913 | case htype_delivery_date: | |
1914 | if (delivery_date_remove) h->type = htype_old; | |
1915 | break; | |
1916 | ||
1917 | /* Same comments as about Return-Path: below. */ | |
1918 | ||
1919 | case htype_envelope_to: | |
1920 | if (envelope_to_remove) h->type = htype_old; | |
1921 | break; | |
1922 | ||
1923 | /* Mark all "From:" headers so they get rewritten. Save the one that is to | |
1924 | be used for Sender: checking. For Sendmail compatibility, if the "From:" | |
1925 | header consists of just the login id of the user who called Exim, rewrite | |
1926 | it with the gecos field first. Apply this rule to Resent-From: if there | |
1927 | are resent- fields. */ | |
1928 | ||
1929 | case htype_from: | |
1930 | h->type = htype_from; | |
1931 | if (!resents_exist || is_resent) | |
1932 | { | |
1933 | from_header = h; | |
1934 | if (!smtp_input) | |
1935 | { | |
1936 | uschar *s = Ustrchr(h->text, ':') + 1; | |
1937 | while (isspace(*s)) s++; | |
1938 | if (strncmpic(s, originator_login, h->slen - (s - h->text) - 1) == 0) | |
1939 | { | |
1940 | uschar *name = is_resent? US"Resent-From" : US"From"; | |
1941 | header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name, | |
1942 | originator_login, qualify_domain_sender); | |
1943 | from_header = header_last; | |
1944 | h->type = htype_old; | |
1945 | DEBUG(D_receive|D_rewrite) | |
1946 | debug_printf("rewrote \"%s:\" header using gecos\n", name); | |
1947 | } | |
1948 | } | |
1949 | } | |
1950 | break; | |
1951 | ||
1952 | /* Identify the Message-id: header for generating "in-reply-to" in the | |
1953 | autoreply transport. For incoming logging, save any resent- value. In both | |
1954 | cases, take just the first of any multiples. */ | |
1955 | ||
1956 | case htype_id: | |
1957 | if (msgid_header == NULL && (!resents_exist || is_resent)) | |
1958 | { | |
1959 | msgid_header = h; | |
1960 | h->type = htype_id; | |
1961 | } | |
1962 | break; | |
1963 | ||
1964 | /* Flag all Received: headers */ | |
1965 | ||
1966 | case htype_received: | |
1967 | h->type = htype_received; | |
1968 | received_count++; | |
1969 | break; | |
1970 | ||
1971 | /* "Reply-to:" is just noted (there is no resent-reply-to field) */ | |
1972 | ||
1973 | case htype_reply_to: | |
1974 | h->type = htype_reply_to; | |
1975 | break; | |
1976 | ||
1977 | /* The Return-path: header is supposed to be added to messages when | |
1978 | they leave the SMTP system. We shouldn't receive messages that already | |
1979 | contain Return-path. However, since Exim generates Return-path: on | |
1980 | local delivery, resent messages may well contain it. We therefore | |
1981 | provide an option (which defaults on) to remove any Return-path: headers | |
1982 | on input. Removal actually means flagging as "old", which prevents the | |
1983 | header being transmitted with the message. */ | |
1984 | ||
1985 | case htype_return_path: | |
1986 | if (return_path_remove) h->type = htype_old; | |
1987 | ||
1988 | /* If we are testing a mail filter file, use the value of the | |
1989 | Return-Path: header to set up the return_path variable, which is not | |
1990 | otherwise set. However, remove any <> that surround the address | |
1991 | because the variable doesn't have these. */ | |
1992 | ||
f05da2e8 | 1993 | if (filter_test != FTEST_NONE) |
059ec3d9 PH |
1994 | { |
1995 | uschar *start = h->text + 12; | |
1996 | uschar *end = start + Ustrlen(start); | |
1997 | while (isspace(*start)) start++; | |
1998 | while (end > start && isspace(end[-1])) end--; | |
1999 | if (*start == '<' && end[-1] == '>') | |
2000 | { | |
2001 | start++; | |
2002 | end--; | |
2003 | } | |
2004 | return_path = string_copyn(start, end - start); | |
2005 | printf("Return-path taken from \"Return-path:\" header line\n"); | |
2006 | } | |
2007 | break; | |
2008 | ||
2009 | /* If there is a "Sender:" header and the message is locally originated, | |
8800895a PH |
2010 | and from an untrusted caller and suppress_local_fixups is not set, or if we |
2011 | are in submission mode for a remote message, mark it "old" so that it will | |
2012 | not be transmitted with the message, unless active_local_sender_retain is | |
2013 | set. (This can only be true if active_local_from_check is false.) If there | |
2014 | are any resent- headers in the message, apply this rule to Resent-Sender: | |
2015 | instead of Sender:. Messages with multiple resent- header sets cannot be | |
2016 | tidily handled. (For this reason, at least one MUA - Pine - turns old | |
2017 | resent- headers into X-resent- headers when resending, leaving just one | |
2018 | set.) */ | |
059ec3d9 PH |
2019 | |
2020 | case htype_sender: | |
69358f02 | 2021 | h->type = ((!active_local_sender_retain && |
8800895a PH |
2022 | ( |
2023 | (sender_local && !trusted_caller && !suppress_local_fixups) | |
2024 | || submission_mode | |
2025 | ) | |
059ec3d9 PH |
2026 | ) && |
2027 | (!resents_exist||is_resent))? | |
2028 | htype_old : htype_sender; | |
2029 | break; | |
2030 | ||
2031 | /* Remember the Subject: header for logging. There is no Resent-Subject */ | |
2032 | ||
2033 | case htype_subject: | |
2034 | subject_header = h; | |
2035 | break; | |
2036 | ||
2037 | /* "To:" gets flagged, and the existence of a recipient header is noted, | |
2038 | whether it's resent- or not. */ | |
2039 | ||
2040 | case htype_to: | |
2041 | h->type = htype_to; | |
2042 | /**** | |
2043 | to_or_cc_header_exists = TRUE; | |
2044 | ****/ | |
2045 | break; | |
2046 | } | |
2047 | } | |
2048 | ||
2049 | /* Extract recipients from the headers if that is required (the -t option). | |
2050 | Note that this is documented as being done *before* any address rewriting takes | |
2051 | place. There are two possibilities: | |
2052 | ||
2053 | (1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any | |
2054 | recipients already listed are to be REMOVED from the message. Smail 3 works | |
2055 | like this. We need to build a non-recipients tree for that list, because in | |
2056 | subsequent processing this data is held in a tree and that's what the | |
2057 | spool_write_header() function expects. Make sure that non-recipient addresses | |
2058 | are fully qualified and rewritten if necessary. | |
2059 | ||
2060 | (2) According to other sendmail documentation, -t ADDS extracted recipients to | |
2061 | those in the command line arguments (and it is rumoured some other MTAs do | |
2062 | this). Therefore, there is an option to make Exim behave this way. | |
2063 | ||
2064 | *** Notes on "Resent-" header lines *** | |
2065 | ||
2066 | The presence of resent-headers in the message makes -t horribly ambiguous. | |
2067 | Experiments with sendmail showed that it uses recipients for all resent- | |
2068 | headers, totally ignoring the concept of "sets of resent- headers" as described | |
2069 | in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set | |
2070 | with all the addresses in one instance of each header. | |
2071 | ||
2072 | This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an | |
2073 | error for -t if there were resent- headers in the message. However, after a | |
2074 | discussion on the mailing list, I've learned that there are MUAs that use | |
2075 | resent- headers with -t, and also that the stuff about sets of resent- headers | |
2076 | and their ordering in RFC 2822 is generally ignored. An MUA that submits a | |
2077 | message with -t and resent- header lines makes sure that only *its* resent- | |
2078 | headers are present; previous ones are often renamed as X-resent- for example. | |
2079 | ||
2080 | Consequently, Exim has been changed so that, if any resent- header lines are | |
2081 | present, the recipients are taken from all of the appropriate resent- lines, | |
2082 | and not from the ordinary To:, Cc:, etc. */ | |
2083 | ||
2084 | if (extract_recip) | |
2085 | { | |
2086 | int rcount = 0; | |
2087 | error_block **bnext = &bad_addresses; | |
2088 | ||
2089 | if (extract_addresses_remove_arguments) | |
2090 | { | |
2091 | while (recipients_count-- > 0) | |
2092 | { | |
2093 | uschar *s = rewrite_address(recipients_list[recipients_count].address, | |
2094 | TRUE, TRUE, global_rewrite_rules, rewrite_existflags); | |
2095 | tree_add_nonrecipient(s); | |
2096 | } | |
2097 | recipients_list = NULL; | |
2098 | recipients_count = recipients_list_max = 0; | |
2099 | } | |
2100 | ||
059ec3d9 PH |
2101 | /* Now scan the headers */ |
2102 | ||
2103 | for (h = header_list->next; h != NULL; h = h->next) | |
2104 | { | |
2105 | if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) && | |
2106 | (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0)) | |
2107 | { | |
2108 | uschar *s = Ustrchr(h->text, ':') + 1; | |
2109 | while (isspace(*s)) s++; | |
2110 | ||
1eccaa59 PH |
2111 | parse_allow_group = TRUE; /* Allow address group syntax */ |
2112 | ||
059ec3d9 PH |
2113 | while (*s != 0) |
2114 | { | |
2115 | uschar *ss = parse_find_address_end(s, FALSE); | |
2116 | uschar *recipient, *errmess, *p, *pp; | |
2117 | int start, end, domain; | |
2118 | ||
2119 | /* Check on maximum */ | |
2120 | ||
2121 | if (recipients_max > 0 && ++rcount > recipients_max) | |
2122 | { | |
2123 | give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients", | |
2124 | US"message rejected: ", error_rc, stdin, NULL); | |
2125 | /* Does not return */ | |
2126 | } | |
2127 | ||
2128 | /* Make a copy of the address, and remove any internal newlines. These | |
2129 | may be present as a result of continuations of the header line. The | |
2130 | white space that follows the newline must not be removed - it is part | |
2131 | of the header. */ | |
2132 | ||
2133 | pp = recipient = store_get(ss - s + 1); | |
2134 | for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p; | |
2135 | *pp = 0; | |
2136 | recipient = parse_extract_address(recipient, &errmess, &start, &end, | |
2137 | &domain, FALSE); | |
2138 | ||
2139 | /* Keep a list of all the bad addresses so we can send a single | |
2140 | error message at the end. However, an empty address is not an error; | |
2141 | just ignore it. This can come from an empty group list like | |
2142 | ||
2143 | To: Recipients of list:; | |
2144 | ||
2145 | If there are no recipients at all, an error will occur later. */ | |
2146 | ||
2147 | if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0) | |
2148 | { | |
2149 | int len = Ustrlen(s); | |
2150 | error_block *b = store_get(sizeof(error_block)); | |
2151 | while (len > 0 && isspace(s[len-1])) len--; | |
2152 | b->next = NULL; | |
2153 | b->text1 = string_printing(string_copyn(s, len)); | |
2154 | b->text2 = errmess; | |
2155 | *bnext = b; | |
2156 | bnext = &(b->next); | |
2157 | } | |
2158 | ||
2159 | /* If the recipient is already in the nonrecipients tree, it must | |
2160 | have appeared on the command line with the option extract_addresses_ | |
2161 | remove_arguments set. Do not add it to the recipients, and keep a note | |
2162 | that this has happened, in order to give a better error if there are | |
2163 | no recipients left. */ | |
2164 | ||
2165 | else if (recipient != NULL) | |
2166 | { | |
2167 | if (tree_search(tree_nonrecipients, recipient) == NULL) | |
2168 | receive_add_recipient(recipient, -1); | |
2169 | else | |
2170 | extracted_ignored = TRUE; | |
2171 | } | |
2172 | ||
2173 | /* Move on past this address */ | |
2174 | ||
2175 | s = ss + (*ss? 1:0); | |
2176 | while (isspace(*s)) s++; | |
1eccaa59 PH |
2177 | } /* Next address */ |
2178 | ||
2179 | parse_allow_group = FALSE; /* Reset group syntax flags */ | |
2180 | parse_found_group = FALSE; | |
059ec3d9 PH |
2181 | |
2182 | /* If this was the bcc: header, mark it "old", which means it | |
2183 | will be kept on the spool, but not transmitted as part of the | |
2184 | message. */ | |
2185 | ||
2cbb4081 | 2186 | if (h->type == htype_bcc) h->type = htype_old; |
059ec3d9 PH |
2187 | } /* For appropriate header line */ |
2188 | } /* For each header line */ | |
2189 | ||
059ec3d9 PH |
2190 | } |
2191 | ||
2192 | /* Now build the unique message id. This has changed several times over the | |
2193 | lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003). | |
2194 | Retaining all the history in the comment has become too unwieldy - read | |
2195 | previous release sources if you want it. | |
2196 | ||
2197 | The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62. | |
2198 | The first part is the current time, in seconds. The second part is the current | |
2199 | pid. Both are large enough to hold 32-bit numbers in base 62. The third part | |
2200 | can hold a number in the range 0-3843. It used to be a computed sequence | |
2201 | number, but is now the fractional component of the current time in units of | |
2202 | 1/2000 of a second (i.e. a value in the range 0-1999). After a message has been | |
2203 | received, Exim ensures that the timer has ticked at the appropriate level | |
2204 | before proceeding, to avoid duplication if the pid happened to be re-used | |
2205 | within the same time period. It seems likely that most messages will take at | |
2206 | least half a millisecond to be received, so no delay will normally be | |
2207 | necessary. At least for some time... | |
2208 | ||
2209 | There is a modification when localhost_number is set. Formerly this was allowed | |
2210 | to be as large as 255. Now it is restricted to the range 0-16, and the final | |
2211 | component of the message id becomes (localhost_number * 200) + fractional time | |
2212 | in units of 1/200 of a second (i.e. a value in the range 0-3399). | |
2213 | ||
2214 | Some not-really-Unix operating systems use case-insensitive file names (Darwin, | |
2215 | Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this | |
2216 | still allows the tttttt field to hold a large enough number to last for some | |
2217 | more decades, and the final two-digit field can hold numbers up to 1295, which | |
2218 | is enough for milliseconds (instead of 1/2000 of a second). | |
2219 | ||
2220 | However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit | |
2221 | pid, so it is probably safe because pids have to be positive. The | |
2222 | localhost_number is restricted to 0-10 for these hosts, and when it is set, the | |
2223 | final field becomes (localhost_number * 100) + fractional time in centiseconds. | |
2224 | ||
2225 | Note that string_base62() returns its data in a static storage block, so it | |
2226 | must be copied before calling string_base62() again. It always returns exactly | |
2227 | 6 characters. | |
2228 | ||
2229 | There doesn't seem to be anything in the RFC which requires a message id to | |
2230 | start with a letter, but Smail was changed to ensure this. The external form of | |
2231 | the message id (as supplied by string expansion) therefore starts with an | |
2232 | additional leading 'E'. The spool file names do not include this leading | |
2233 | letter and it is not used internally. | |
2234 | ||
2235 | NOTE: If ever the format of message ids is changed, the regular expression for | |
2236 | checking that a string is in this format must be updated in a corresponding | |
2237 | way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH | |
2238 | must also be changed to reflect the correct string length. Then, of course, | |
2239 | other programs that rely on the message id format will need updating too. */ | |
2240 | ||
2241 | Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6); | |
2242 | message_id[6] = '-'; | |
2243 | Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6); | |
2244 | ||
2245 | /* Deal with the case where the host number is set. The value of the number was | |
2246 | checked when it was read, to ensure it isn't too big. The timing granularity is | |
2247 | left in id_resolution so that an appropriate wait can be done after receiving | |
2248 | the message, if necessary (we hope it won't be). */ | |
2249 | ||
2250 | if (host_number_string != NULL) | |
2251 | { | |
2252 | id_resolution = (BASE_62 == 62)? 5000 : 10000; | |
2253 | sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s", | |
2254 | string_base62((long int)( | |
2255 | host_number * (1000000/id_resolution) + | |
2256 | message_id_tv.tv_usec/id_resolution)) + 4); | |
2257 | } | |
2258 | ||
2259 | /* Host number not set: final field is just the fractional time at an | |
2260 | appropriate resolution. */ | |
2261 | ||
2262 | else | |
2263 | { | |
2264 | id_resolution = (BASE_62 == 62)? 500 : 1000; | |
2265 | sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s", | |
2266 | string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4); | |
2267 | } | |
2268 | ||
2269 | /* Add the current message id onto the current process info string if | |
2270 | it will fit. */ | |
2271 | ||
2272 | (void)string_format(process_info + process_info_len, | |
2273 | PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id); | |
2274 | ||
2275 | /* If we are using multiple input directories, set up the one for this message | |
2276 | to be the least significant base-62 digit of the time of arrival. Otherwise | |
2277 | ensure that it is an empty string. */ | |
2278 | ||
2279 | message_subdir[0] = split_spool_directory? message_id[5] : 0; | |
2280 | ||
2281 | /* Now that we have the message-id, if there is no message-id: header, generate | |
8800895a PH |
2282 | one, but only for local (without suppress_local_fixups) or submission mode |
2283 | messages. This can be user-configured if required, but we had better flatten | |
2284 | any illegal characters therein. */ | |
059ec3d9 | 2285 | |
8800895a PH |
2286 | if (msgid_header == NULL && |
2287 | ((sender_host_address == NULL && !suppress_local_fixups) | |
2288 | || submission_mode)) | |
059ec3d9 PH |
2289 | { |
2290 | uschar *p; | |
2291 | uschar *id_text = US""; | |
2292 | uschar *id_domain = primary_hostname; | |
2293 | ||
2294 | /* Permit only letters, digits, dots, and hyphens in the domain */ | |
2295 | ||
2296 | if (message_id_domain != NULL) | |
2297 | { | |
2298 | uschar *new_id_domain = expand_string(message_id_domain); | |
2299 | if (new_id_domain == NULL) | |
2300 | { | |
2301 | if (!expand_string_forcedfail) | |
2302 | log_write(0, LOG_MAIN|LOG_PANIC, | |
2303 | "expansion of \"%s\" (message_id_header_domain) " | |
2304 | "failed: %s", message_id_domain, expand_string_message); | |
2305 | } | |
2306 | else if (*new_id_domain != 0) | |
2307 | { | |
2308 | id_domain = new_id_domain; | |
2309 | for (p = id_domain; *p != 0; p++) | |
2310 | if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */ | |
2311 | } | |
2312 | } | |
2313 | ||
2314 | /* Permit all characters except controls and RFC 2822 specials in the | |
2315 | additional text part. */ | |
2316 | ||
2317 | if (message_id_text != NULL) | |
2318 | { | |
2319 | uschar *new_id_text = expand_string(message_id_text); | |
2320 | if (new_id_text == NULL) | |
2321 | { | |
2322 | if (!expand_string_forcedfail) | |
2323 | log_write(0, LOG_MAIN|LOG_PANIC, | |
2324 | "expansion of \"%s\" (message_id_header_text) " | |
2325 | "failed: %s", message_id_text, expand_string_message); | |
2326 | } | |
2327 | else if (*new_id_text != 0) | |
2328 | { | |
2329 | id_text = new_id_text; | |
2330 | for (p = id_text; *p != 0; p++) | |
2331 | if (mac_iscntrl_or_special(*p)) *p = '-'; | |
2332 | } | |
2333 | } | |
2334 | ||
2335 | /* Add the header line */ | |
2336 | ||
2337 | header_add(htype_id, "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, | |
2338 | message_id_external, (*id_text == 0)? "" : ".", id_text, id_domain); | |
2339 | } | |
2340 | ||
2341 | /* If we are to log recipients, keep a copy of the raw ones before any possible | |
2342 | rewriting. Must copy the count, because later ACLs and the local_scan() | |
2343 | function may mess with the real recipients. */ | |
2344 | ||
2345 | if ((log_extra_selector & LX_received_recipients) != 0) | |
2346 | { | |
2347 | raw_recipients = store_get(recipients_count * sizeof(uschar *)); | |
2348 | for (i = 0; i < recipients_count; i++) | |
2349 | raw_recipients[i] = string_copy(recipients_list[i].address); | |
2350 | raw_recipients_count = recipients_count; | |
2351 | } | |
2352 | ||
2353 | /* Ensure the recipients list is fully qualified and rewritten. Unqualified | |
2354 | recipients will get here only if the conditions were right (allow_unqualified_ | |
2355 | recipient is TRUE). */ | |
2356 | ||
2357 | for (i = 0; i < recipients_count; i++) | |
2358 | recipients_list[i].address = | |
2359 | rewrite_address(recipients_list[i].address, TRUE, TRUE, | |
2360 | global_rewrite_rules, rewrite_existflags); | |
2361 | ||
8800895a PH |
2362 | /* If there is no From: header, generate one for local (without |
2363 | suppress_local_fixups) or submission_mode messages. If there is no sender | |
2364 | address, but the sender is local or this is a local delivery error, use the | |
2365 | originator login. This shouldn't happen for genuine bounces, but might happen | |
2366 | for autoreplies. The addition of From: must be done *before* checking for the | |
2367 | possible addition of a Sender: header, because untrusted_set_sender allows an | |
2368 | untrusted user to set anything in the envelope (which might then get info | |
2369 | From:) but we still want to ensure a valid Sender: if it is required. */ | |
2370 | ||
2371 | if (from_header == NULL && | |
2372 | ((sender_host_address == NULL && !suppress_local_fixups) | |
2373 | || submission_mode)) | |
059ec3d9 | 2374 | { |
2fe1a124 PH |
2375 | uschar *oname = US""; |
2376 | ||
2377 | /* Use the originator_name if this is a locally submitted message and the | |
2378 | caller is not trusted. For trusted callers, use it only if -F was used to | |
2379 | force its value or if we have a non-SMTP message for which -f was not used | |
2380 | to set the sender. */ | |
2381 | ||
2382 | if (sender_host_address == NULL) | |
2383 | { | |
2384 | if (!trusted_caller || sender_name_forced || | |
2385 | (!smtp_input && !sender_address_forced)) | |
2386 | oname = originator_name; | |
2387 | } | |
2388 | ||
2389 | /* For non-locally submitted messages, the only time we use the originator | |
2390 | name is when it was forced by the /name= option on control=submission. */ | |
2391 | ||
2392 | else | |
2393 | { | |
2394 | if (submission_name != NULL) oname = submission_name; | |
2395 | } | |
2396 | ||
059ec3d9 PH |
2397 | /* Envelope sender is empty */ |
2398 | ||
2399 | if (sender_address[0] == 0) | |
2400 | { | |
87ba3f5f PH |
2401 | uschar *fromstart, *fromend; |
2402 | ||
2403 | fromstart = string_sprintf("%sFrom: %s%s", resent_prefix, | |
2fe1a124 PH |
2404 | oname, (oname[0] == 0)? "" : " <"); |
2405 | fromend = (oname[0] == 0)? US"" : US">"; | |
87ba3f5f | 2406 | |
059ec3d9 PH |
2407 | if (sender_local || local_error_message) |
2408 | { | |
87ba3f5f PH |
2409 | header_add(htype_from, "%s%s@%s%s\n", fromstart, |
2410 | local_part_quote(originator_login), qualify_domain_sender, | |
2411 | fromend); | |
059ec3d9 PH |
2412 | } |
2413 | else if (submission_mode && authenticated_id != NULL) | |
2414 | { | |
2415 | if (submission_domain == NULL) | |
2416 | { | |
87ba3f5f PH |
2417 | header_add(htype_from, "%s%s@%s%s\n", fromstart, |
2418 | local_part_quote(authenticated_id), qualify_domain_sender, | |
2419 | fromend); | |
059ec3d9 PH |
2420 | } |
2421 | else if (submission_domain[0] == 0) /* empty => whole address set */ | |
2422 | { | |
87ba3f5f PH |
2423 | header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id, |
2424 | fromend); | |
059ec3d9 PH |
2425 | } |
2426 | else | |
2427 | { | |
87ba3f5f PH |
2428 | header_add(htype_from, "%s%s@%s%s\n", fromstart, |
2429 | local_part_quote(authenticated_id), submission_domain, | |
2430 | fromend); | |
059ec3d9 PH |
2431 | } |
2432 | from_header = header_last; /* To get it checked for Sender: */ | |
2433 | } | |
2434 | } | |
2435 | ||
2436 | /* There is a non-null envelope sender. Build the header using the original | |
2437 | sender address, before any rewriting that might have been done while | |
2438 | verifying it. */ | |
2439 | ||
2440 | else | |
2441 | { | |
87ba3f5f | 2442 | header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix, |
2fe1a124 PH |
2443 | oname, |
2444 | (oname[0] == 0)? "" : " <", | |
87ba3f5f PH |
2445 | (sender_address_unrewritten == NULL)? |
2446 | sender_address : sender_address_unrewritten, | |
2fe1a124 | 2447 | (oname[0] == 0)? "" : ">"); |
059ec3d9 PH |
2448 | |
2449 | from_header = header_last; /* To get it checked for Sender: */ | |
2450 | } | |
2451 | } | |
2452 | ||
2453 | ||
8800895a PH |
2454 | /* If the sender is local (without suppress_local_fixups), or if we are in |
2455 | submission mode and there is an authenticated_id, check that an existing From: | |
2456 | is correct, and if not, generate a Sender: header, unless disabled. Any | |
2457 | previously-existing Sender: header was removed above. Note that sender_local, | |
2458 | as well as being TRUE if the caller of exim is not trusted, is also true if a | |
2459 | trusted caller did not supply a -f argument for non-smtp input. To allow | |
2460 | trusted callers to forge From: without supplying -f, we have to test explicitly | |
2461 | here. If the From: header contains more than one address, then the call to | |
2462 | parse_extract_address fails, and a Sender: header is inserted, as required. */ | |
059ec3d9 PH |
2463 | |
2464 | if (from_header != NULL && | |
69358f02 | 2465 | (active_local_from_check && |
8800895a | 2466 | ((sender_local && !trusted_caller && !suppress_local_fixups) || |
69358f02 | 2467 | (submission_mode && authenticated_id != NULL)) |
059ec3d9 PH |
2468 | )) |
2469 | { | |
2470 | BOOL make_sender = TRUE; | |
2471 | int start, end, domain; | |
2472 | uschar *errmess; | |
2473 | uschar *from_address = | |
2474 | parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess, | |
2475 | &start, &end, &domain, FALSE); | |
2476 | uschar *generated_sender_address; | |
2477 | ||
2478 | if (submission_mode) | |
2479 | { | |
2480 | if (submission_domain == NULL) | |
2481 | { | |
2482 | generated_sender_address = string_sprintf("%s@%s", | |
2483 | local_part_quote(authenticated_id), qualify_domain_sender); | |
2484 | } | |
2485 | else if (submission_domain[0] == 0) /* empty => full address */ | |
2486 | { | |
2487 | generated_sender_address = string_sprintf("%s", | |
2488 | authenticated_id); | |
2489 | } | |
2490 | else | |
2491 | { | |
2492 | generated_sender_address = string_sprintf("%s@%s", | |
2493 | local_part_quote(authenticated_id), submission_domain); | |
2494 | } | |
2495 | } | |
2496 | else | |
2497 | generated_sender_address = string_sprintf("%s@%s", | |
2498 | local_part_quote(originator_login), qualify_domain_sender); | |
2499 | ||
2500 | /* Remove permitted prefixes and suffixes from the local part of the From: | |
2501 | address before doing the comparison with the generated sender. */ | |
2502 | ||
2503 | if (from_address != NULL) | |
2504 | { | |
2505 | int slen; | |
2506 | uschar *at = (domain == 0)? NULL : from_address + domain - 1; | |
2507 | ||
2508 | if (at != NULL) *at = 0; | |
2509 | from_address += route_check_prefix(from_address, local_from_prefix); | |
2510 | slen = route_check_suffix(from_address, local_from_suffix); | |
2511 | if (slen > 0) | |
2512 | { | |
2513 | memmove(from_address+slen, from_address, Ustrlen(from_address)-slen); | |
2514 | from_address += slen; | |
2515 | } | |
2516 | if (at != NULL) *at = '@'; | |
2517 | ||
2518 | if (strcmpic(generated_sender_address, from_address) == 0 || | |
2519 | (domain == 0 && strcmpic(from_address, originator_login) == 0)) | |
2520 | make_sender = FALSE; | |
2521 | } | |
2522 | ||
2523 | /* We have to cause the Sender header to be rewritten if there are | |
2524 | appropriate rewriting rules. */ | |
2525 | ||
2526 | if (make_sender) | |
2527 | { | |
2fe1a124 | 2528 | if (submission_mode && submission_name == NULL) |
059ec3d9 PH |
2529 | header_add(htype_sender, "%sSender: %s\n", resent_prefix, |
2530 | generated_sender_address); | |
2531 | else | |
2532 | header_add(htype_sender, "%sSender: %s <%s>\n", | |
2fe1a124 PH |
2533 | resent_prefix, |
2534 | submission_mode? submission_name : originator_name, | |
2535 | generated_sender_address); | |
059ec3d9 | 2536 | } |
87ba3f5f PH |
2537 | |
2538 | /* Ensure that a non-null envelope sender address corresponds to the | |
2539 | submission mode sender address. */ | |
2540 | ||
2541 | if (submission_mode && sender_address[0] != 0) | |
2542 | { | |
2543 | if (sender_address_unrewritten == NULL) | |
2544 | sender_address_unrewritten = sender_address; | |
2545 | sender_address = generated_sender_address; | |
2546 | log_write(L_address_rewrite, LOG_MAIN, | |
2547 | "\"%s\" from env-from rewritten as \"%s\" by submission mode", | |
2548 | sender_address_unrewritten, generated_sender_address); | |
2549 | } | |
059ec3d9 PH |
2550 | } |
2551 | ||
2552 | ||
2553 | /* If there are any rewriting rules, apply them to the sender address, unless | |
2554 | it has already been rewritten as part of verification for SMTP input. */ | |
2555 | ||
2556 | if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL && | |
2557 | sender_address[0] != 0) | |
2558 | { | |
2559 | sender_address = rewrite_address(sender_address, FALSE, TRUE, | |
2560 | global_rewrite_rules, rewrite_existflags); | |
2561 | DEBUG(D_receive|D_rewrite) | |
2562 | debug_printf("rewritten sender = %s\n", sender_address); | |
2563 | } | |
2564 | ||
2565 | ||
2566 | /* The headers must be run through rewrite_header(), because it ensures that | |
2567 | addresses are fully qualified, as well as applying any rewriting rules that may | |
2568 | exist. | |
2569 | ||
2570 | Qualification of header addresses in a message from a remote host happens only | |
2571 | if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as | |
2572 | appropriate. For local messages, qualification always happens, unless -bnq is | |
2573 | used to explicitly suppress it. No rewriting is done for an unqualified address | |
2574 | that is left untouched. | |
2575 | ||
2576 | We start at the second header, skipping our own Received:. This rewriting is | |
2577 | documented as happening *after* recipient addresses are taken from the headers | |
2578 | by the -t command line option. An added Sender: gets rewritten here. */ | |
2579 | ||
2580 | for (h = header_list->next; h != NULL; h = h->next) | |
2581 | { | |
2582 | header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules, | |
2583 | rewrite_existflags, TRUE); | |
2584 | if (newh != NULL) h = newh; | |
2585 | } | |
2586 | ||
2587 | ||
2588 | /* An RFC 822 (sic) message is not legal unless it has at least one of "to", | |
2cbb4081 | 2589 | "cc", or "bcc". Note that although the minimal examples in RFC 822 show just |
059ec3d9 PH |
2590 | "to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header |
2591 | exists, this applies to the set of resent- headers rather than the normal set. | |
2592 | ||
2cbb4081 PH |
2593 | The requirement for a recipient header has been removed in RFC 2822. At this |
2594 | point in the code, earlier versions of Exim added a To: header for locally | |
2595 | submitted messages, and an empty Bcc: header for others. In the light of the | |
2596 | changes in RFC 2822, this was dropped in November 2003. */ | |
059ec3d9 | 2597 | |
059ec3d9 PH |
2598 | |
2599 | /* If there is no date header, generate one if the message originates locally | |
8800895a PH |
2600 | (i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the |
2601 | submission mode flag is set. Messages without Date: are not valid, but it seems | |
2602 | to be more confusing if Exim adds one to all remotely-originated messages. */ | |
059ec3d9 | 2603 | |
8800895a PH |
2604 | if (!date_header_exists && |
2605 | ((sender_host_address == NULL && !suppress_local_fixups) | |
2606 | || submission_mode)) | |
059ec3d9 PH |
2607 | header_add(htype_other, "%sDate: %s\n", resent_prefix, tod_stamp(tod_full)); |
2608 | ||
2609 | search_tidyup(); /* Free any cached resources */ | |
2610 | ||
2611 | /* Show the complete set of headers if debugging. Note that the first one (the | |
2612 | new Received:) has not yet been set. */ | |
2613 | ||
2614 | DEBUG(D_receive) | |
2615 | { | |
2616 | debug_printf(">>Headers after rewriting and local additions:\n"); | |
2617 | for (h = header_list->next; h != NULL; h = h->next) | |
2618 | debug_printf("%c %s", h->type, h->text); | |
2619 | debug_printf("\n"); | |
2620 | } | |
2621 | ||
2622 | /* The headers are now complete in store. If we are running in filter | |
2623 | testing mode, that is all this function does. Return TRUE if the message | |
2624 | ended with a dot. */ | |
2625 | ||
f05da2e8 | 2626 | if (filter_test != FTEST_NONE) |
059ec3d9 PH |
2627 | { |
2628 | process_info[process_info_len] = 0; | |
2629 | return message_ended == END_DOT; | |
2630 | } | |
2631 | ||
2632 | /* Open a new spool file for the data portion of the message. We need | |
2633 | to access it both via a file descriptor and a stream. Try to make the | |
2634 | directory if it isn't there. Note re use of sprintf: spool_directory | |
2635 | is checked on input to be < 200 characters long. */ | |
2636 | ||
2637 | sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory, message_subdir, | |
2638 | message_id); | |
2639 | data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE); | |
2640 | if (data_fd < 0) | |
2641 | { | |
2642 | if (errno == ENOENT) | |
2643 | { | |
2644 | uschar temp[16]; | |
2645 | sprintf(CS temp, "input/%s", message_subdir); | |
2646 | if (message_subdir[0] == 0) temp[5] = 0; | |
2647 | (void)directory_make(spool_directory, temp, INPUT_DIRECTORY_MODE, TRUE); | |
2648 | data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE); | |
2649 | } | |
2650 | if (data_fd < 0) | |
2651 | log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s", | |
2652 | spool_name, strerror(errno)); | |
2653 | } | |
2654 | ||
2655 | /* Make sure the file's group is the Exim gid, and double-check the mode | |
2656 | because the group setting doesn't always get set automatically. */ | |
2657 | ||
ff790e47 PH |
2658 | (void)fchown(data_fd, exim_uid, exim_gid); |
2659 | (void)fchmod(data_fd, SPOOL_MODE); | |
059ec3d9 PH |
2660 | |
2661 | /* We now have data file open. Build a stream for it and lock it. We lock only | |
2662 | the first line of the file (containing the message ID) because otherwise there | |
2663 | are problems when Exim is run under Cygwin (I'm told). See comments in | |
2664 | spool_in.c, where the same locking is done. */ | |
2665 | ||
2666 | data_file = fdopen(data_fd, "w+"); | |
2667 | lock_data.l_type = F_WRLCK; | |
2668 | lock_data.l_whence = SEEK_SET; | |
2669 | lock_data.l_start = 0; | |
2670 | lock_data.l_len = SPOOL_DATA_START_OFFSET; | |
2671 | ||
2672 | if (fcntl(data_fd, F_SETLK, &lock_data) < 0) | |
2673 | log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name, | |
2674 | errno, strerror(errno)); | |
2675 | ||
2676 | /* We have an open, locked data file. Write the message id to it to make it | |
2677 | self-identifying. Then read the remainder of the input of this message and | |
2678 | write it to the data file. If the variable next != NULL, it contains the first | |
2679 | data line (which was read as a header but then turned out not to have the right | |
2680 | format); write it (remembering that it might contain binary zeros). The result | |
2681 | of fwrite() isn't inspected; instead we call ferror() below. */ | |
2682 | ||
2683 | fprintf(data_file, "%s-D\n", message_id); | |
2684 | if (next != NULL) | |
2685 | { | |
2686 | uschar *s = next->text; | |
2687 | int len = next->slen; | |
f1e894f3 | 2688 | (void)fwrite(s, 1, len, data_file); |
059ec3d9 PH |
2689 | body_linecount++; /* Assumes only 1 line */ |
2690 | } | |
2691 | ||
2692 | /* Note that we might already be at end of file, or the logical end of file | |
2693 | (indicated by '.'), or might have encountered an error while writing the | |
2694 | message id or "next" line. */ | |
2695 | ||
2696 | if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT) | |
2697 | { | |
2698 | if (smtp_input) | |
2699 | { | |
2700 | message_ended = read_message_data_smtp(data_file); | |
2701 | receive_linecount++; /* The terminating "." line */ | |
2702 | } | |
2703 | else message_ended = read_message_data(data_file); | |
2704 | ||
2705 | receive_linecount += body_linecount; /* For BSMTP errors mainly */ | |
2e0c1448 | 2706 | message_linecount += body_linecount; |
059ec3d9 PH |
2707 | |
2708 | /* Handle premature termination of SMTP */ | |
2709 | ||
2710 | if (smtp_input && message_ended == END_EOF) | |
2711 | { | |
2712 | Uunlink(spool_name); /* Lose data file when closed */ | |
2713 | message_id[0] = 0; /* Indicate no message accepted */ | |
2714 | smtp_reply = handle_lost_connection(US""); | |
2715 | smtp_yield = FALSE; | |
2716 | goto TIDYUP; /* Skip to end of function */ | |
2717 | } | |
2718 | ||
2719 | /* Handle message that is too big. Don't use host_or_ident() in the log | |
2720 | message; we want to see the ident value even for non-remote messages. */ | |
2721 | ||
2722 | if (message_ended == END_SIZE) | |
2723 | { | |
2724 | Uunlink(spool_name); /* Lose the data file when closed */ | |
2725 | if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */ | |
2726 | ||
2727 | log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: " | |
2728 | "message too big: read=%d max=%d", | |
2729 | sender_address, | |
2730 | (sender_fullhost == NULL)? "" : " H=", | |
2731 | (sender_fullhost == NULL)? US"" : sender_fullhost, | |
2732 | (sender_ident == NULL)? "" : " U=", | |
2733 | (sender_ident == NULL)? US"" : sender_ident, | |
2734 | message_size, | |
2735 | thismessage_size_limit); | |
2736 | ||
2737 | if (smtp_input) | |
2738 | { | |
2739 | smtp_reply = US"552 Message size exceeds maximum permitted"; | |
2740 | message_id[0] = 0; /* Indicate no message accepted */ | |
2741 | goto TIDYUP; /* Skip to end of function */ | |
2742 | } | |
2743 | else | |
2744 | { | |
2745 | fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET); | |
2746 | give_local_error(ERRMESS_TOOBIG, | |
2747 | string_sprintf("message too big (max=%d)", thismessage_size_limit), | |
2748 | US"message rejected: ", error_rc, data_file, header_list); | |
2749 | /* Does not return */ | |
2750 | } | |
2751 | } | |
2752 | } | |
2753 | ||
2754 | /* Restore the standard SIGALRM handler for any subsequent processing. (For | |
2755 | example, there may be some expansion in an ACL that uses a timer.) */ | |
2756 | ||
2757 | os_non_restarting_signal(SIGALRM, sigalrm_handler); | |
2758 | ||
2759 | /* The message body has now been read into the data file. Call fflush() to | |
2760 | empty the buffers in C, and then call fsync() to get the data written out onto | |
2761 | the disk, as fflush() doesn't do this (or at least, it isn't documented as | |
2762 | having to do this). If there was an I/O error on either input or output, | |
2763 | attempt to send an error message, and unlink the spool file. For non-SMTP input | |
2764 | we can then give up. Note that for SMTP input we must swallow the remainder of | |
2765 | the input in cases of output errors, since the far end doesn't expect to see | |
2766 | anything until the terminating dot line is sent. */ | |
2767 | ||
2768 | if (fflush(data_file) == EOF || ferror(data_file) || | |
54fc8428 | 2769 | EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)()) |
059ec3d9 PH |
2770 | { |
2771 | uschar *msg_errno = US strerror(errno); | |
2772 | BOOL input_error = (receive_ferror)() != 0; | |
2773 | uschar *msg = string_sprintf("%s error (%s) while receiving message from %s", | |
2774 | input_error? "Input read" : "Spool write", | |
2775 | msg_errno, | |
2776 | (sender_fullhost != NULL)? sender_fullhost : sender_ident); | |
2777 | ||
2778 | log_write(0, LOG_MAIN, "Message abandoned: %s", msg); | |
2779 | Uunlink(spool_name); /* Lose the data file */ | |
2780 | ||
2781 | if (smtp_input) | |
2782 | { | |
2783 | if (input_error) | |
2784 | smtp_reply = US"451 Error while reading input data"; | |
2785 | else | |
2786 | { | |
2787 | smtp_reply = US"451 Error while writing spool file"; | |
2788 | receive_swallow_smtp(); | |
2789 | } | |
2790 | message_id[0] = 0; /* Indicate no message accepted */ | |
2791 | goto TIDYUP; /* Skip to end of function */ | |
2792 | } | |
2793 | ||
2794 | else | |
2795 | { | |
2796 | fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET); | |
2797 | give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file, | |
2798 | header_list); | |
2799 | /* Does not return */ | |
2800 | } | |
2801 | } | |
2802 | ||
2803 | ||
2804 | /* No I/O errors were encountered while writing the data file. */ | |
2805 | ||
2806 | DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id); | |
2807 | ||
2808 | ||
2809 | /* If there were any bad addresses extracted by -t, or there were no recipients | |
2810 | left after -t, send a message to the sender of this message, or write it to | |
2811 | stderr if the error handling option is set that way. Note that there may | |
2812 | legitimately be no recipients for an SMTP message if they have all been removed | |
2813 | by "discard". | |
2814 | ||
2815 | We need to rewind the data file in order to read it. In the case of no | |
2816 | recipients or stderr error writing, throw the data file away afterwards, and | |
2817 | exit. (This can't be SMTP, which always ensures there's at least one | |
2818 | syntactically good recipient address.) */ | |
2819 | ||
2820 | if (extract_recip && (bad_addresses != NULL || recipients_count == 0)) | |
2821 | { | |
2822 | DEBUG(D_receive) | |
2823 | { | |
2824 | if (recipients_count == 0) debug_printf("*** No recipients\n"); | |
2825 | if (bad_addresses != NULL) | |
2826 | { | |
2827 | error_block *eblock = bad_addresses; | |
2828 | debug_printf("*** Bad address(es)\n"); | |
2829 | while (eblock != NULL) | |
2830 | { | |
2831 | debug_printf(" %s: %s\n", eblock->text1, eblock->text2); | |
2832 | eblock = eblock->next; | |
2833 | } | |
2834 | } | |
2835 | } | |
2836 | ||
2837 | fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET); | |
2838 | ||
2839 | /* If configured to send errors to the sender, but this fails, force | |
2840 | a failure error code. We use a special one for no recipients so that it | |
2841 | can be detected by the autoreply transport. Otherwise error_rc is set to | |
2842 | errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case | |
2843 | it is EXIT_SUCCESS. */ | |
2844 | ||
2845 | if (error_handling == ERRORS_SENDER) | |
2846 | { | |
2847 | if (!moan_to_sender( | |
2848 | (bad_addresses == NULL)? | |
2849 | (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) : | |
2850 | (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS, | |
2851 | bad_addresses, header_list, data_file, FALSE)) | |
2852 | error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE; | |
2853 | } | |
2854 | else | |
2855 | { | |
2856 | if (bad_addresses == NULL) | |
2857 | { | |
2858 | if (extracted_ignored) | |
2859 | fprintf(stderr, "exim: all -t recipients overridden by command line\n"); | |
2860 | else | |
2861 | fprintf(stderr, "exim: no recipients in message\n"); | |
2862 | } | |
2863 | else | |
2864 | { | |
2865 | fprintf(stderr, "exim: invalid address%s", | |
2866 | (bad_addresses->next == NULL)? ":" : "es:\n"); | |
2867 | while (bad_addresses != NULL) | |
2868 | { | |
2869 | fprintf(stderr, " %s: %s\n", bad_addresses->text1, | |
2870 | bad_addresses->text2); | |
2871 | bad_addresses = bad_addresses->next; | |
2872 | } | |
2873 | } | |
2874 | } | |
2875 | ||
2876 | if (recipients_count == 0 || error_handling == ERRORS_STDERR) | |
2877 | { | |
2878 | Uunlink(spool_name); | |
f1e894f3 | 2879 | (void)fclose(data_file); |
059ec3d9 PH |
2880 | exim_exit(error_rc); |
2881 | } | |
2882 | } | |
2883 | ||
2884 | /* Data file successfully written. Generate text for the Received: header by | |
2885 | expanding the configured string, and adding a timestamp. By leaving this | |
2886 | operation till now, we ensure that the timestamp is the time that message | |
2887 | reception was completed. However, this is deliberately done before calling the | |
2888 | data ACL and local_scan(). | |
2889 | ||
2890 | This Received: header may therefore be inspected by the data ACL and by code in | |
2891 | the local_scan() function. When they have run, we update the timestamp to be | |
2892 | the final time of reception. | |
2893 | ||
2894 | If there is just one recipient, set up its value in the $received_for variable | |
2895 | for use when we generate the Received: header. | |
2896 | ||
2897 | Note: the checking for too many Received: headers is handled by the delivery | |
2898 | code. */ | |
2899 | ||
2900 | timestamp = expand_string(US"${tod_full}"); | |
2901 | if (recipients_count == 1) received_for = recipients_list[0].address; | |
2902 | received = expand_string(received_header_text); | |
2903 | received_for = NULL; | |
2904 | ||
2905 | if (received == NULL) | |
2906 | { | |
2907 | Uunlink(spool_name); /* Lose the data file */ | |
2908 | log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" " | |
2909 | "(received_header_text) failed: %s", string_printing(received_header_text), | |
2910 | expand_string_message); | |
2911 | } | |
2912 | ||
2913 | /* The first element on the header chain is reserved for the Received header, | |
2914 | so all we have to do is fill in the text pointer, and set the type. However, if | |
2915 | the result of the expansion is an empty string, we leave the header marked as | |
2916 | "old" so as to refrain from adding a Received header. */ | |
2917 | ||
2918 | if (received[0] == 0) | |
2919 | { | |
2920 | received_header->text = string_sprintf("Received: ; %s\n", timestamp); | |
2921 | received_header->type = htype_old; | |
2922 | } | |
2923 | else | |
2924 | { | |
2925 | received_header->text = string_sprintf("%s; %s\n", received, timestamp); | |
2926 | received_header->type = htype_received; | |
2927 | } | |
2928 | ||
2929 | received_header->slen = Ustrlen(received_header->text); | |
2930 | ||
2931 | DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s", | |
2932 | received_header->type, received_header->text); | |
2933 | ||
2934 | /* Set the value of message_body_size for the DATA ACL and for local_scan() */ | |
2935 | ||
2936 | message_body_size = (fstat(data_fd, &statbuf) == 0)? | |
2937 | statbuf.st_size - SPOOL_DATA_START_OFFSET : -1; | |
2938 | ||
2939 | /* If an ACL from any RCPT commands set up any warning headers to add, do so | |
2940 | now, before running the DATA ACL. */ | |
2941 | ||
2942 | add_acl_headers(US"MAIL or RCPT"); | |
2943 | ||
2944 | /* If an ACL is specified for checking things at this stage of reception of a | |
2945 | message, run it, unless all the recipients were removed by "discard" in earlier | |
2946 | ACLs. That is the only case in which recipients_count can be zero at this | |
2947 | stage. Set deliver_datafile to point to the data file so that $message_body and | |
2948 | $message_body_end can be extracted if needed. Allow $recipients in expansions. | |
2949 | */ | |
2950 | ||
2951 | deliver_datafile = data_fd; | |
4e88a19f | 2952 | user_msg = NULL; |
059ec3d9 | 2953 | |
0e20aff9 MH |
2954 | enable_dollar_recipients = TRUE; |
2955 | ||
059ec3d9 PH |
2956 | if (recipients_count == 0) |
2957 | { | |
2958 | blackholed_by = recipients_discarded? US"MAIL ACL" : US"RCPT ACL"; | |
2959 | } | |
2960 | else | |
2961 | { | |
059ec3d9 PH |
2962 | /* Handle interactive SMTP messages */ |
2963 | ||
2964 | if (smtp_input && !smtp_batched_input) | |
2965 | { | |
8523533c | 2966 | |
fb2274d4 TK |
2967 | #ifdef EXPERIMENTAL_DOMAINKEYS |
2968 | dk_exim_verify_finish(); | |
2969 | #endif | |
2970 | ||
8523533c | 2971 | #ifdef WITH_CONTENT_SCAN |
54cdb463 PH |
2972 | if (acl_smtp_mime != NULL && |
2973 | !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by)) | |
2974 | goto TIDYUP; | |
8523533c TK |
2975 | #endif /* WITH_CONTENT_SCAN */ |
2976 | ||
54cdb463 PH |
2977 | /* Check the recipients count again, as the MIME ACL might have changed |
2978 | them. */ | |
8523533c | 2979 | |
059ec3d9 PH |
2980 | if (acl_smtp_data != NULL && recipients_count > 0) |
2981 | { | |
059ec3d9 PH |
2982 | rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg); |
2983 | add_acl_headers(US"DATA"); | |
2984 | if (rc == DISCARD) | |
2985 | { | |
2986 | recipients_count = 0; | |
2987 | blackholed_by = US"DATA ACL"; | |
8e669ac1 PH |
2988 | if (log_msg != NULL) |
2989 | blackhole_log_msg = string_sprintf(": %s", log_msg); | |
059ec3d9 PH |
2990 | } |
2991 | else if (rc != OK) | |
2992 | { | |
2993 | Uunlink(spool_name); | |
8523533c TK |
2994 | #ifdef WITH_CONTENT_SCAN |
2995 | unspool_mbox(); | |
2996 | #endif | |
059ec3d9 PH |
2997 | if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0) |
2998 | smtp_yield = FALSE; /* No more messsages after dropped connection */ | |
2999 | smtp_reply = US""; /* Indicate reply already sent */ | |
3000 | message_id[0] = 0; /* Indicate no message accepted */ | |
3001 | goto TIDYUP; /* Skip to end of function */ | |
3002 | } | |
3003 | } | |
3004 | } | |
3005 | ||
3006 | /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that | |
3007 | we cannot take different actions for permanent and temporary rejections. */ | |
3008 | ||
54cdb463 | 3009 | else |
059ec3d9 | 3010 | { |
54cdb463 PH |
3011 | |
3012 | #ifdef WITH_CONTENT_SCAN | |
3013 | if (acl_not_smtp_mime != NULL && | |
3014 | !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply, | |
3015 | &blackholed_by)) | |
3016 | goto TIDYUP; | |
3017 | #endif /* WITH_CONTENT_SCAN */ | |
3018 | ||
3019 | if (acl_not_smtp != NULL) | |
059ec3d9 | 3020 | { |
54cdb463 PH |
3021 | uschar *user_msg, *log_msg; |
3022 | rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg); | |
3023 | if (rc == DISCARD) | |
059ec3d9 | 3024 | { |
54cdb463 PH |
3025 | recipients_count = 0; |
3026 | blackholed_by = US"non-SMTP ACL"; | |
3027 | if (log_msg != NULL) | |
3028 | blackhole_log_msg = string_sprintf(": %s", log_msg); | |
059ec3d9 | 3029 | } |
54cdb463 | 3030 | else if (rc != OK) |
059ec3d9 | 3031 | { |
54cdb463 PH |
3032 | Uunlink(spool_name); |
3033 | #ifdef WITH_CONTENT_SCAN | |
3034 | unspool_mbox(); | |
3035 | #endif | |
6ea85e9a PH |
3036 | /* The ACL can specify where rejections are to be logged, possibly |
3037 | nowhere. The default is main and reject logs. */ | |
3038 | ||
3039 | if (log_reject_target != 0) | |
3040 | log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s", | |
3041 | sender_address, log_msg); | |
3042 | ||
54cdb463 PH |
3043 | if (user_msg == NULL) user_msg = US"local configuration problem"; |
3044 | if (smtp_batched_input) | |
3045 | { | |
3046 | moan_smtp_batch(NULL, "%d %s", 550, user_msg); | |
3047 | /* Does not return */ | |
3048 | } | |
3049 | else | |
3050 | { | |
3051 | fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET); | |
3052 | give_local_error(ERRMESS_LOCAL_ACL, user_msg, | |
3053 | US"message rejected by non-SMTP ACL: ", error_rc, data_file, | |
3054 | header_list); | |
3055 | /* Does not return */ | |
3056 | } | |
059ec3d9 | 3057 | } |
54cdb463 | 3058 | add_acl_headers(US"non-SMTP"); |
059ec3d9 | 3059 | } |
059ec3d9 PH |
3060 | } |
3061 | ||
54cdb463 PH |
3062 | /* The applicable ACLs have been run */ |
3063 | ||
059ec3d9 PH |
3064 | if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */ |
3065 | if (queue_only_policy) queued_by = US"ACL"; | |
059ec3d9 PH |
3066 | } |
3067 | ||
8523533c TK |
3068 | #ifdef WITH_CONTENT_SCAN |
3069 | unspool_mbox(); | |
3070 | #endif | |
3071 | ||
059ec3d9 PH |
3072 | /* The final check on the message is to run the scan_local() function. The |
3073 | version supplied with Exim always accepts, but this is a hook for sysadmins to | |
3074 | supply their own checking code. The local_scan() function is run even when all | |
3075 | the recipients have been discarded. */ | |
3076 | ||
3077 | lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET); | |
3078 | ||
3079 | /* Arrange to catch crashes in local_scan(), so that the -D file gets | |
3080 | deleted, and the incident gets logged. */ | |
3081 | ||
3082 | os_non_restarting_signal(SIGSEGV, local_scan_crash_handler); | |
3083 | os_non_restarting_signal(SIGFPE, local_scan_crash_handler); | |
3084 | os_non_restarting_signal(SIGILL, local_scan_crash_handler); | |
3085 | os_non_restarting_signal(SIGBUS, local_scan_crash_handler); | |
3086 | ||
3087 | DEBUG(D_receive) debug_printf("calling local_scan(); timeout=%d\n", | |
3088 | local_scan_timeout); | |
3089 | local_scan_data = NULL; | |
3090 | ||
3091 | os_non_restarting_signal(SIGALRM, local_scan_timeout_handler); | |
3092 | if (local_scan_timeout > 0) alarm(local_scan_timeout); | |
3093 | rc = local_scan(data_fd, &local_scan_data); | |
3094 | alarm(0); | |
3095 | os_non_restarting_signal(SIGALRM, sigalrm_handler); | |
3096 | ||
0e20aff9 MH |
3097 | enable_dollar_recipients = FALSE; |
3098 | ||
059ec3d9 PH |
3099 | store_pool = POOL_MAIN; /* In case changed */ |
3100 | DEBUG(D_receive) debug_printf("local_scan() returned %d %s\n", rc, | |
3101 | local_scan_data); | |
3102 | ||
3103 | os_non_restarting_signal(SIGSEGV, SIG_DFL); | |
3104 | os_non_restarting_signal(SIGFPE, SIG_DFL); | |
3105 | os_non_restarting_signal(SIGILL, SIG_DFL); | |
3106 | os_non_restarting_signal(SIGBUS, SIG_DFL); | |
3107 | ||
3108 | /* The length check is paranoia against some runaway code, and also because | |
3109 | (for a success return) lines in the spool file are read into big_buffer. */ | |
3110 | ||
3111 | if (local_scan_data != NULL) | |
3112 | { | |
3113 | int len = Ustrlen(local_scan_data); | |
3114 | if (len > LOCAL_SCAN_MAX_RETURN) len = LOCAL_SCAN_MAX_RETURN; | |
3115 | local_scan_data = string_copyn(local_scan_data, len); | |
3116 | } | |
3117 | ||
3118 | if (rc == LOCAL_SCAN_ACCEPT_FREEZE) | |
3119 | { | |
58eb016e | 3120 | if (!deliver_freeze) /* ACL might have already frozen */ |
059ec3d9 PH |
3121 | { |
3122 | deliver_freeze = TRUE; | |
3123 | deliver_frozen_at = time(NULL); | |
3124 | frozen_by = US"local_scan()"; | |
3125 | } | |
3126 | rc = LOCAL_SCAN_ACCEPT; | |
3127 | } | |
3128 | else if (rc == LOCAL_SCAN_ACCEPT_QUEUE) | |
3129 | { | |
3130 | if (!queue_only_policy) /* ACL might have already queued */ | |
3131 | { | |
3132 | queue_only_policy = TRUE; | |
3133 | queued_by = US"local_scan()"; | |
3134 | } | |
3135 | rc = LOCAL_SCAN_ACCEPT; | |
3136 | } | |
3137 | ||
3138 | /* Message accepted: remove newlines in local_scan_data because otherwise | |
3139 | the spool file gets corrupted. Ensure that all recipients are qualified. */ | |
3140 | ||
3141 | if (rc == LOCAL_SCAN_ACCEPT) | |
3142 | { | |
3143 | if (local_scan_data != NULL) | |
3144 | { | |
3145 | uschar *s; | |
3146 | for (s = local_scan_data; *s != 0; s++) if (*s == '\n') *s = ' '; | |
3147 | } | |
3148 | for (i = 0; i < recipients_count; i++) | |
3149 | { | |
3150 | recipient_item *r = recipients_list + i; | |
3151 | r->address = rewrite_address_qualify(r->address, TRUE); | |
3152 | if (r->errors_to != NULL) | |
3153 | r->errors_to = rewrite_address_qualify(r->errors_to, TRUE); | |
3154 | } | |
3155 | if (recipients_count == 0 && blackholed_by == NULL) | |
3156 | blackholed_by = US"local_scan"; | |
3157 | } | |
3158 | ||
3159 | /* Message rejected: newlines permitted in local_scan_data to generate | |
3160 | multiline SMTP responses. */ | |
3161 | ||
3162 | else | |
3163 | { | |
3164 | uschar *istemp = US""; | |
3165 | uschar *s = NULL; | |
a5bd321b | 3166 | uschar *smtp_code; |
059ec3d9 PH |
3167 | int size = 0; |
3168 | int sptr = 0; | |
059ec3d9 PH |
3169 | |
3170 | errmsg = local_scan_data; | |
3171 | ||
3172 | Uunlink(spool_name); /* Cancel this message */ | |
3173 | switch(rc) | |
3174 | { | |
3175 | default: | |
3176 | log_write(0, LOG_MAIN, "invalid return %d from local_scan(). Temporary " | |
3177 | "rejection given", rc); | |
3178 | goto TEMPREJECT; | |
3179 | ||
3180 | case LOCAL_SCAN_REJECT_NOLOGHDR: | |
3181 | log_extra_selector &= ~LX_rejected_header; | |
3182 | /* Fall through */ | |
3183 | ||
3184 | case LOCAL_SCAN_REJECT: | |
a5bd321b | 3185 | smtp_code = US"550"; |
059ec3d9 PH |
3186 | if (errmsg == NULL) errmsg = US"Administrative prohibition"; |
3187 | break; | |
3188 | ||
3189 | case LOCAL_SCAN_TEMPREJECT_NOLOGHDR: | |
3190 | log_extra_selector &= ~LX_rejected_header; | |
3191 | /* Fall through */ | |
3192 | ||
3193 | case LOCAL_SCAN_TEMPREJECT: | |
3194 | TEMPREJECT: | |
a5bd321b | 3195 | smtp_code = US"451"; |
059ec3d9 PH |
3196 | if (errmsg == NULL) errmsg = US"Temporary local problem"; |
3197 | istemp = US"temporarily "; | |
3198 | break; | |
3199 | } | |
3200 | ||
3201 | s = string_append(s, &size, &sptr, 2, US"F=", | |
3202 | (sender_address[0] == 0)? US"<>" : sender_address); | |
3203 | s = add_host_info_for_log(s, &size, &sptr); | |
3204 | s[sptr] = 0; | |
3205 | ||
3206 | log_write(0, LOG_MAIN|LOG_REJECT, "%s %srejected by local_scan(): %.256s", | |
3207 | s, istemp, string_printing(errmsg)); | |
3208 | ||
3209 | if (smtp_input) | |
3210 | { | |
3211 | if (!smtp_batched_input) | |
3212 | { | |
a5bd321b | 3213 | smtp_respond(smtp_code, 3, TRUE, errmsg); |
059ec3d9 PH |
3214 | message_id[0] = 0; /* Indicate no message accepted */ |
3215 | smtp_reply = US""; /* Indicate reply already sent */ | |
3216 | goto TIDYUP; /* Skip to end of function */ | |
3217 | } | |
3218 | else | |
3219 | { | |
a5bd321b | 3220 | moan_smtp_batch(NULL, "%s %s", smtp_code, errmsg); |
059ec3d9 PH |
3221 | /* Does not return */ |
3222 | } | |
3223 | } | |
3224 | else | |
3225 | { | |
3226 | fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET); | |
3227 | give_local_error(ERRMESS_LOCAL_SCAN, errmsg, | |
3228 | US"message rejected by local scan code: ", error_rc, data_file, | |
3229 | header_list); | |
3230 | /* Does not return */ | |
3231 | } | |
3232 | } | |
3233 | ||
3234 | /* Reset signal handlers to ignore signals that previously would have caused | |
3235 | the message to be abandoned. */ | |
3236 | ||
3237 | signal(SIGTERM, SIG_IGN); | |
3238 | signal(SIGINT, SIG_IGN); | |
3239 | ||
3240 | /* Ensure the first time flag is set in the newly-received message. */ | |
3241 | ||
3242 | deliver_firsttime = TRUE; | |
3243 | ||
8523533c TK |
3244 | #ifdef EXPERIMENTAL_BRIGHTMAIL |
3245 | if (bmi_run == 1) { | |
3246 | /* rewind data file */ | |
3247 | lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET); | |
3248 | bmi_verdicts = bmi_process_message(header_list, data_fd); | |
3249 | }; | |
3250 | #endif | |
3251 | ||
059ec3d9 PH |
3252 | /* Update the timstamp in our Received: header to account for any time taken by |
3253 | an ACL or by local_scan(). The new time is the time that all reception | |
3254 | processing is complete. */ | |
3255 | ||
3256 | timestamp = expand_string(US"${tod_full}"); | |
3257 | tslen = Ustrlen(timestamp); | |
3258 | ||
3259 | memcpy(received_header->text + received_header->slen - tslen - 1, | |
3260 | timestamp, tslen); | |
3261 | ||
3262 | /* In MUA wrapper mode, ignore queueing actions set by ACL or local_scan() */ | |
3263 | ||
3264 | if (mua_wrapper) | |
3265 | { | |
3266 | deliver_freeze = FALSE; | |
3267 | queue_only_policy = FALSE; | |
3268 | } | |
3269 | ||
3270 | /* Keep the data file open until we have written the header file, in order to | |
3271 | hold onto the lock. In a -bh run, or if the message is to be blackholed, we | |
3272 | don't write the header file, and we unlink the data file. If writing the header | |
3273 | file fails, we have failed to accept this message. */ | |
3274 | ||
3275 | if (host_checking || blackholed_by != NULL) | |
3276 | { | |
3277 | header_line *h; | |
3278 | Uunlink(spool_name); | |
3279 | msg_size = 0; /* Compute size for log line */ | |
3280 | for (h = header_list; h != NULL; h = h->next) | |
3281 | if (h->type != '*') msg_size += h->slen; | |
3282 | } | |
3283 | ||
3284 | /* Write the -H file */ | |
3285 | ||
3286 | else | |
3287 | { | |
3288 | if ((msg_size = spool_write_header(message_id, SW_RECEIVING, &errmsg)) < 0) | |
3289 | { | |
3290 | log_write(0, LOG_MAIN, "Message abandoned: %s", errmsg); | |
3291 | Uunlink(spool_name); /* Lose the data file */ | |
3292 | ||
3293 | if (smtp_input) | |
3294 | { | |
3295 | smtp_reply = US"451 Error in writing spool file"; | |
3296 | message_id[0] = 0; /* Indicate no message accepted */ | |
3297 | goto TIDYUP; | |
3298 | } | |
3299 | else | |
3300 | { | |
3301 | fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET); | |
3302 | give_local_error(ERRMESS_IOERR, errmsg, US"", error_rc, data_file, | |
3303 | header_list); | |
3304 | /* Does not return */ | |
3305 | } | |
3306 | } | |
3307 | } | |
3308 | ||
3309 | ||
3310 | /* The message has now been successfully received. */ | |
3311 | ||
3312 | receive_messagecount++; | |
3313 | ||
3314 | /* In SMTP sessions we may receive several in one connection. After each one, | |
3315 | we wait for the clock to tick at the level of message-id granularity. This is | |
3316 | so that the combination of time+pid is unique, even on systems where the pid | |
3317 | can be re-used within our time interval. We can't shorten the interval without | |
3318 | re-designing the message-id. See comments above where the message id is | |
3319 | created. This is Something For The Future. */ | |
3320 | ||
3321 | message_id_tv.tv_usec = (message_id_tv.tv_usec/id_resolution) * id_resolution; | |
3322 | exim_wait_tick(&message_id_tv, id_resolution); | |
3323 | ||
3324 | /* Add data size to written header size. We do not count the initial file name | |
3325 | that is in the file, but we do add one extra for the notional blank line that | |
3326 | precedes the data. This total differs from message_size in that it include the | |
3327 | added Received: header and any other headers that got created locally. */ | |
3328 | ||
3329 | fflush(data_file); | |
3330 | fstat(data_fd, &statbuf); | |
3331 | ||
3332 | msg_size += statbuf.st_size - SPOOL_DATA_START_OFFSET + 1; | |
3333 | ||
3334 | /* Generate a "message received" log entry. We do this by building up a dynamic | |
3335 | string as required. Since we commonly want to add two items at a time, use a | |
3336 | macro to simplify the coding. We log the arrival of a new message while the | |
3337 | file is still locked, just in case the machine is *really* fast, and delivers | |
3338 | it first! Include any message id that is in the message - since the syntax of a | |
3339 | message id is actually an addr-spec, we can use the parse routine to canonicize | |
3340 | it. */ | |
3341 | ||
3342 | size = 256; | |
3343 | sptr = 0; | |
3344 | s = store_get(size); | |
3345 | ||
3346 | s = string_append(s, &size, &sptr, 2, US"<= ", | |
3347 | (sender_address[0] == 0)? US"<>" : sender_address); | |
3348 | if (message_reference != NULL) | |
3349 | s = string_append(s, &size, &sptr, 2, US" R=", message_reference); | |
3350 | ||
3351 | s = add_host_info_for_log(s, &size, &sptr); | |
3352 | ||
3353 | #ifdef SUPPORT_TLS | |
3354 | if ((log_extra_selector & LX_tls_cipher) != 0 && tls_cipher != NULL) | |
3355 | s = string_append(s, &size, &sptr, 2, US" X=", tls_cipher); | |
3356 | if ((log_extra_selector & LX_tls_certificate_verified) != 0 && | |
3357 | tls_cipher != NULL) | |
3358 | s = string_append(s, &size, &sptr, 2, US" CV=", | |
3359 | tls_certificate_verified? "yes":"no"); | |
3360 | if ((log_extra_selector & LX_tls_peerdn) != 0 && tls_peerdn != NULL) | |
3361 | s = string_append(s, &size, &sptr, 3, US" DN=\"", tls_peerdn, US"\""); | |
3362 | #endif | |
3363 | ||
3364 | if (sender_host_authenticated != NULL) | |
3365 | { | |
3366 | s = string_append(s, &size, &sptr, 2, US" A=", sender_host_authenticated); | |
3367 | if (authenticated_id != NULL) | |
3368 | s = string_append(s, &size, &sptr, 2, US":", authenticated_id); | |
3369 | } | |
3370 | ||
3371 | sprintf(CS big_buffer, "%d", msg_size); | |
3372 | s = string_append(s, &size, &sptr, 2, US" S=", big_buffer); | |
3373 | ||
3374 | /* If an addr-spec in a message-id contains a quoted string, it can contain | |
3375 | any characters except " \ and CR and so in particular it can contain NL! | |
3376 | Therefore, make sure we use a printing-characters only version for the log. | |
3377 | Also, allow for domain literals in the message id. */ | |
3378 | ||
3379 | if (msgid_header != NULL) | |
3380 | { | |
3381 | uschar *old_id; | |
3382 | BOOL save_allow_domain_literals = allow_domain_literals; | |
3383 | allow_domain_literals = TRUE; | |
3384 | old_id = parse_extract_address(Ustrchr(msgid_header->text, ':') + 1, | |
3385 | &errmsg, &start, &end, &domain, FALSE); | |
3386 | allow_domain_literals = save_allow_domain_literals; | |
3387 | if (old_id != NULL) | |
3388 | s = string_append(s, &size, &sptr, 2, US" id=", string_printing(old_id)); | |
3389 | } | |
3390 | ||
3391 | /* If subject logging is turned on, create suitable printing-character | |
3392 | text. By expanding $h_subject: we make use of the MIME decoding. */ | |
3393 | ||
3394 | if ((log_extra_selector & LX_subject) != 0 && subject_header != NULL) | |
3395 | { | |
3396 | int i; | |
3397 | uschar *p = big_buffer; | |
3398 | uschar *ss = expand_string(US"$h_subject:"); | |
3399 | ||
3400 | /* Backslash-quote any double quotes or backslashes so as to make a | |
3401 | a C-like string, and turn any non-printers into escape sequences. */ | |
3402 | ||
3403 | *p++ = '\"'; | |
3404 | if (*ss != 0) for (i = 0; i < 100 && ss[i] != 0; i++) | |
3405 | { | |
3406 | if (ss[i] == '\"' || ss[i] == '\\') *p++ = '\\'; | |
3407 | *p++ = ss[i]; | |
3408 | } | |
3409 | *p++ = '\"'; | |
3410 | *p = 0; | |
3411 | s = string_append(s, &size, &sptr, 2, US" T=", string_printing(big_buffer)); | |
3412 | } | |
3413 | ||
3414 | /* Terminate the string: string_cat() and string_append() leave room, but do | |
3415 | not put the zero in. */ | |
3416 | ||
3417 | s[sptr] = 0; | |
3418 | ||
059ec3d9 PH |
3419 | /* Create a message log file if message logs are being used and this message is |
3420 | not blackholed. Write the reception stuff to it. We used to leave message log | |
3421 | creation until the first delivery, but this has proved confusing for somep | |
3422 | people. */ | |
3423 | ||
3424 | if (message_logs && blackholed_by == NULL) | |
3425 | { | |
3426 | int fd; | |
3427 | ||
3428 | sprintf(CS spool_name, "%s/msglog/%s/%s", spool_directory, message_subdir, | |
3429 | message_id); | |
3430 | fd = Uopen(spool_name, O_WRONLY|O_APPEND|O_CREAT, SPOOL_MODE); | |
3431 | ||
3432 | if (fd < 0 && errno == ENOENT) | |
3433 | { | |
3434 | uschar temp[16]; | |
3435 | sprintf(CS temp, "msglog/%s", message_subdir); | |
3436 | if (message_subdir[0] == 0) temp[6] = 0; | |
3437 | (void)directory_make(spool_directory, temp, MSGLOG_DIRECTORY_MODE, TRUE); | |
3438 | fd = Uopen(spool_name, O_WRONLY|O_APPEND|O_CREAT, SPOOL_MODE); | |
3439 | } | |
3440 | ||
3441 | if (fd < 0) | |
3442 | { | |
3443 | log_write(0, LOG_MAIN|LOG_PANIC, "Couldn't open message log %s: %s", | |
3444 | spool_name, strerror(errno)); | |
3445 | } | |
3446 | ||
3447 | else | |
3448 | { | |
3449 | FILE *message_log = fdopen(fd, "a"); | |
3450 | if (message_log == NULL) | |
3451 | { | |
3452 | log_write(0, LOG_MAIN|LOG_PANIC, "Couldn't fdopen message log %s: %s", | |
3453 | spool_name, strerror(errno)); | |
f1e894f3 | 3454 | (void)close(fd); |
059ec3d9 PH |
3455 | } |
3456 | else | |
3457 | { | |
3458 | uschar *now = tod_stamp(tod_log); | |
3459 | fprintf(message_log, "%s Received from %s\n", now, s+3); | |
3460 | if (deliver_freeze) fprintf(message_log, "%s frozen by %s\n", now, | |
3461 | frozen_by); | |
3462 | if (queue_only_policy) fprintf(message_log, | |
3463 | "%s no immediate delivery: queued by %s\n", now, queued_by); | |
f1e894f3 | 3464 | (void)fclose(message_log); |
059ec3d9 PH |
3465 | } |
3466 | } | |
3467 | } | |
3468 | ||
58eb016e PH |
3469 | /* Everything has now been done for a successful message except logging its |
3470 | arrival, and outputting an SMTP response. While writing to the log, set a flag | |
3471 | to cause a call to receive_bomb_out() if the log cannot be opened. */ | |
3472 | ||
3473 | receive_call_bombout = TRUE; | |
3474 | ||
563b63fa PH |
3475 | /* Before sending an SMTP response in a TCP/IP session, we check to see if the |
3476 | connection has gone away. This can only be done if there is no unconsumed input | |
3477 | waiting in the local input buffer. We can test for this by calling | |
3478 | receive_smtp_buffered(). RFC 2920 (pipelining) explicitly allows for additional | |
3479 | input to be sent following the final dot, so the presence of following input is | |
3480 | not an error. | |
58eb016e | 3481 | |
563b63fa PH |
3482 | If the connection is still present, but there is no unread input for the |
3483 | socket, the result of a select() call will be zero. If, however, the connection | |
3484 | has gone away, or if there is pending input, the result of select() will be | |
3485 | non-zero. The two cases can be distinguished by trying to read the next input | |
3486 | character. If we succeed, we can unread it so that it remains in the local | |
3487 | buffer for handling later. If not, the connection has been lost. | |
58eb016e | 3488 | |
563b63fa PH |
3489 | Of course, since TCP/IP is asynchronous, there is always a chance that the |
3490 | connection will vanish between the time of this test and the sending of the | |
3491 | response, but the chance of this happening should be small. */ | |
3492 | ||
3493 | if (smtp_input && sender_host_address != NULL && !sender_host_notsocket && | |
3494 | !receive_smtp_buffered()) | |
58eb016e PH |
3495 | { |
3496 | struct timeval tv; | |
3497 | fd_set select_check; | |
3498 | FD_ZERO(&select_check); | |
3499 | FD_SET(fileno(smtp_in), &select_check); | |
3500 | tv.tv_sec = 0; | |
3501 | tv.tv_usec = 0; | |
3502 | ||
563b63fa | 3503 | if (select(fileno(smtp_in) + 1, &select_check, NULL, NULL, &tv) != 0) |
58eb016e | 3504 | { |
563b63fa PH |
3505 | int c = (RECEIVE_GETC)(); |
3506 | if (c != EOF) (RECEIVE_UNGETC)(c); else | |
58eb016e | 3507 | { |
563b63fa PH |
3508 | uschar *msg = US"SMTP connection lost after final dot"; |
3509 | smtp_reply = US""; /* No attempt to send a response */ | |
3510 | smtp_yield = FALSE; /* Nothing more on this connection */ | |
58eb016e | 3511 | |
563b63fa | 3512 | /* Re-use the log line workspace */ |
58eb016e | 3513 | |
563b63fa PH |
3514 | sptr = 0; |
3515 | s = string_cat(s, &size, &sptr, msg, Ustrlen(msg)); | |
3516 | s = add_host_info_for_log(s, &size, &sptr); | |
3517 | s[sptr] = 0; | |
3518 | log_write(0, LOG_MAIN, "%s", s); | |
58eb016e | 3519 | |
563b63fa | 3520 | /* Delete the files for this aborted message. */ |
58eb016e | 3521 | |
563b63fa PH |
3522 | sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory, |
3523 | message_subdir, message_id); | |
3524 | Uunlink(spool_name); | |
58eb016e | 3525 | |
563b63fa PH |
3526 | sprintf(CS spool_name, "%s/input/%s/%s-H", spool_directory, |
3527 | message_subdir, message_id); | |
3528 | Uunlink(spool_name); | |
58eb016e | 3529 | |
563b63fa PH |
3530 | sprintf(CS spool_name, "%s/msglog/%s/%s", spool_directory, |
3531 | message_subdir, message_id); | |
3532 | Uunlink(spool_name); | |
58eb016e | 3533 | |
563b63fa PH |
3534 | goto TIDYUP; |
3535 | } | |
58eb016e PH |
3536 | } |
3537 | } | |
3538 | ||
3539 | /* The connection has not gone away; we really are going to take responsibility | |
3540 | for this message. */ | |
3541 | ||
3542 | log_write(0, LOG_MAIN | | |
3543 | (((log_extra_selector & LX_received_recipients) != 0)? LOG_RECIPIENTS : 0) | | |
3544 | (((log_extra_selector & LX_received_sender) != 0)? LOG_SENDER : 0), | |
3545 | "%s", s); | |
3546 | receive_call_bombout = FALSE; | |
3547 | ||
3548 | /* Log any control actions taken by an ACL or local_scan(). */ | |
3549 | ||
3550 | if (deliver_freeze) log_write(0, LOG_MAIN, "frozen by %s", frozen_by); | |
3551 | if (queue_only_policy) log_write(L_delay_delivery, LOG_MAIN, | |
3552 | "no immediate delivery: queued by %s", queued_by); | |
3553 | ||
059ec3d9 PH |
3554 | store_reset(s); /* The store for the main log message can be reused */ |
3555 | ||
3556 | /* If the message is frozen, and freeze_tell is set, do the telling. */ | |
3557 | ||
3558 | if (deliver_freeze && freeze_tell != NULL && freeze_tell[0] != 0) | |
3559 | { | |
3560 | moan_tell_someone(freeze_tell, NULL, US"Message frozen on arrival", | |
3561 | "Message %s was frozen on arrival by %s.\nThe sender is <%s>.\n", | |
3562 | message_id, frozen_by, sender_address); | |
3563 | } | |
3564 | ||
3565 | ||
3566 | /* Either a message has been successfully received and written to the two spool | |
3567 | files, or an error in writing the spool has occurred for an SMTP message, or | |
58eb016e PH |
3568 | an SMTP message has been rejected for policy reasons. (For a non-SMTP message |
3569 | we will have already given up because there's no point in carrying on!) In | |
3570 | either event, we must now close (and thereby unlock) the data file. In the | |
059ec3d9 PH |
3571 | successful case, this leaves the message on the spool, ready for delivery. In |
3572 | the error case, the spool file will be deleted. Then tidy up store, interact | |
3573 | with an SMTP call if necessary, and return. | |
3574 | ||
3575 | A fflush() was done earlier in the expectation that any write errors on the | |
3576 | data file will be flushed(!) out thereby. Nevertheless, it is theoretically | |
3577 | possible for fclose() to fail - but what to do? What has happened to the lock | |
3578 | if this happens? */ | |
3579 | ||
3580 | TIDYUP: | |
f1e894f3 PH |
3581 | process_info[process_info_len] = 0; /* Remove message id */ |
3582 | if (data_file != NULL) (void)fclose(data_file); /* Frees the lock */ | |
059ec3d9 PH |
3583 | |
3584 | /* Now reset signal handlers to their defaults */ | |
3585 | ||
3586 | signal(SIGTERM, SIG_DFL); | |
3587 | signal(SIGINT, SIG_DFL); | |
3588 | ||
3589 | /* Tell an SMTP caller the state of play, and arrange to return the SMTP return | |
3590 | value, which defaults TRUE - meaning there may be more incoming messages from | |
3591 | this connection. For non-SMTP callers (where there is only ever one message), | |
3592 | the default is FALSE. */ | |
3593 | ||
3594 | if (smtp_input) | |
3595 | { | |
3596 | yield = smtp_yield; | |
3597 | ||
3598 | /* Handle interactive SMTP callers. After several kinds of error, smtp_reply | |
58eb016e PH |
3599 | is set to the response that should be sent. When it is NULL, we generate |
3600 | default responses. After an ACL error or local_scan() error, the response has | |
3601 | already been sent, and smtp_reply is an empty string to indicate this. */ | |
059ec3d9 PH |
3602 | |
3603 | if (!smtp_batched_input) | |
3604 | { | |
3605 | if (smtp_reply == NULL) | |
3606 | { | |
29aba418 | 3607 | if (fake_response != OK) |
a5bd321b PH |
3608 | smtp_respond((fake_response == DEFER)? US"450" : US"550", 3, TRUE, |
3609 | fake_response_text); | |
4e88a19f PH |
3610 | |
3611 | /* An OK response is required; use "message" text if present. */ | |
3612 | ||
3613 | else if (user_msg != NULL) | |
3614 | { | |
3615 | uschar *code = US"250"; | |
3616 | int len = 3; | |
3617 | smtp_message_code(&code, &len, &user_msg, NULL); | |
3618 | smtp_respond(code, len, TRUE, user_msg); | |
3619 | } | |
3620 | ||
3621 | /* Default OK response */ | |
3622 | ||
8e669ac1 PH |
3623 | else |
3624 | smtp_printf("250 OK id=%s\r\n", message_id); | |
059ec3d9 PH |
3625 | if (host_checking) |
3626 | fprintf(stdout, | |
3627 | "\n**** SMTP testing: that is not a real message id!\n\n"); | |
3628 | } | |
4e88a19f | 3629 | |
58eb016e | 3630 | /* smtp_reply is set non-empty */ |
4e88a19f | 3631 | |
8523533c TK |
3632 | else if (smtp_reply[0] != 0) |
3633 | { | |
29aba418 | 3634 | if (fake_response != OK && (smtp_reply[0] == '2')) |
a5bd321b PH |
3635 | smtp_respond((fake_response == DEFER)? US"450" : US"550", 3, TRUE, |
3636 | fake_response_text); | |
8e669ac1 | 3637 | else |
6951ac6c PH |
3638 | smtp_printf("%.1024s\r\n", smtp_reply); |
3639 | } | |
059ec3d9 PH |
3640 | } |
3641 | ||
3642 | /* For batched SMTP, generate an error message on failure, and do | |
3643 | nothing on success. The function moan_smtp_batch() does not return - | |
3644 | it exits from the program with a non-zero return code. */ | |
3645 | ||
3646 | else if (smtp_reply != NULL) moan_smtp_batch(NULL, "%s", smtp_reply); | |
3647 | } | |
3648 | ||
3649 | ||
3650 | /* If blackholing, we can immediately log this message's sad fate. The data | |
3651 | file has already been unlinked, and the header file was never written to disk. | |
3652 | We must now indicate that nothing was received, to prevent a delivery from | |
3653 | starting. */ | |
3654 | ||
3655 | if (blackholed_by != NULL) | |
3656 | { | |
3657 | uschar *detail = (local_scan_data != NULL)? | |
3658 | string_printing(local_scan_data) : | |
3659 | string_sprintf("(%s discarded recipients)", blackholed_by); | |
04f7d5b9 | 3660 | log_write(0, LOG_MAIN, "=> blackhole %s%s", detail, blackhole_log_msg); |
059ec3d9 PH |
3661 | log_write(0, LOG_MAIN, "Completed"); |
3662 | message_id[0] = 0; | |
3663 | } | |
3664 | ||
3665 | /* Reset headers so that logging of rejects for a subsequent message doesn't | |
3666 | include them. It is also important to set header_last = NULL before exiting | |
3667 | from this function, as this prevents certain rewrites that might happen during | |
3668 | subsequent verifying (of another incoming message) from trying to add headers | |
3669 | when they shouldn't. */ | |
3670 | ||
3671 | header_list = header_last = NULL; | |
3672 | ||
3673 | return yield; /* TRUE if more messages (SMTP only) */ | |
3674 | } | |
3675 | ||
3676 | /* End of receive.c */ |