Update eximstats to watch out for senders sending 'HELO [IpAddr]'
[exim.git] / src / src / receive.c
CommitLineData
059ec3d9
PH
1/*************************************************
2* Exim - an Internet mail transport agent *
3*************************************************/
4
c4ceed07 5/* Copyright (c) University of Cambridge 1995 - 2012 */
059ec3d9
PH
6/* See the file NOTICE for conditions of use and distribution. */
7
8/* Code for receiving a message and setting up spool files. */
9
059ec3d9
PH
10#include "exim.h"
11
6a8f9482
TK
12#ifdef EXPERIMENTAL_DCC
13extern int dcc_ok;
14#endif
15
059ec3d9
PH
16/*************************************************
17* Local static variables *
18*************************************************/
19
20static FILE *data_file = NULL;
21static int data_fd = -1;
22static uschar spool_name[256];
23
24
25
26/*************************************************
27* Non-SMTP character reading functions *
28*************************************************/
29
30/* These are the default functions that are set up in the variables such as
31receive_getc initially. They just call the standard functions, passing stdin as
32the file. (When SMTP input is occurring, different functions are used by
33changing the pointer variables.) */
34
35int
36stdin_getc(void)
37{
38return getc(stdin);
39}
40
41int
42stdin_ungetc(int c)
43{
44return ungetc(c, stdin);
45}
46
47int
48stdin_feof(void)
49{
50return feof(stdin);
51}
52
53int
54stdin_ferror(void)
55{
56return ferror(stdin);
57}
58
59
60
61
62/*************************************************
63* Check that a set sender is allowed *
64*************************************************/
65
66/* This function is called when a local caller sets an explicit sender address.
67It checks whether this is permitted, which it is for trusted callers.
68Otherwise, it must match the pattern(s) in untrusted_set_sender.
69
70Arguments: the proposed sender address
71Returns: TRUE for a trusted caller
72 TRUE if the address has been set, untrusted_set_sender has been
73 set, and the address matches something in the list
74 FALSE otherwise
75*/
76
77BOOL
78receive_check_set_sender(uschar *newsender)
79{
80uschar *qnewsender;
81if (trusted_caller) return TRUE;
82if (newsender == NULL || untrusted_set_sender == NULL) return FALSE;
83qnewsender = (Ustrchr(newsender, '@') != NULL)?
84 newsender : string_sprintf("%s@%s", newsender, qualify_domain_sender);
85return
86 match_address_list(qnewsender, TRUE, TRUE, &untrusted_set_sender, NULL, -1,
87 0, NULL) == OK;
88}
89
90
91
92
93/*************************************************
5cb8cbc6 94* Read space info for a partition *
059ec3d9
PH
95*************************************************/
96
8e669ac1
PH
97/* This function is called by receive_check_fs() below, and also by string
98expansion for variables such as $spool_space. The field names for the statvfs
5cb8cbc6
PH
99structure are macros, because not all OS have F_FAVAIL and it seems tidier to
100have macros for F_BAVAIL and F_FILES as well. Some kinds of file system do not
101have inodes, and they return -1 for the number available.
059ec3d9 102
5cb8cbc6
PH
103Later: It turns out that some file systems that do not have the concept of
104inodes return 0 rather than -1. Such systems should also return 0 for the total
8e669ac1 105number of inodes, so we require that to be greater than zero before returning
5cb8cbc6 106an inode count.
059ec3d9 107
5cb8cbc6
PH
108Arguments:
109 isspool TRUE for spool partition, FALSE for log partition
110 inodeptr address of int to receive inode count; -1 if there isn't one
8e669ac1 111
5cb8cbc6 112Returns: available on-root space, in kilobytes
8e669ac1
PH
113 -1 for log partition if there isn't one
114
115All values are -1 if the STATFS functions are not available.
059ec3d9
PH
116*/
117
8e669ac1 118int
5cb8cbc6 119receive_statvfs(BOOL isspool, int *inodeptr)
059ec3d9
PH
120{
121#ifdef HAVE_STATFS
059ec3d9 122struct STATVFS statbuf;
5cb8cbc6
PH
123uschar *path;
124uschar *name;
125uschar buffer[1024];
059ec3d9 126
5cb8cbc6 127/* The spool directory must always exist. */
059ec3d9 128
5cb8cbc6 129if (isspool)
059ec3d9 130 {
8e669ac1
PH
131 path = spool_directory;
132 name = US"spool";
133 }
134
059ec3d9
PH
135/* Need to cut down the log file path to the directory, and to ignore any
136appearance of "syslog" in it. */
137
5cb8cbc6 138else
059ec3d9 139 {
059ec3d9 140 int sep = ':'; /* Not variable - outside scripts use */
059ec3d9 141 uschar *p = log_file_path;
8e669ac1 142 name = US"log";
059ec3d9
PH
143
144 /* An empty log_file_path means "use the default". This is the same as an
145 empty item in a list. */
146
147 if (*p == 0) p = US":";
148 while ((path = string_nextinlist(&p, &sep, buffer, sizeof(buffer))) != NULL)
149 {
150 if (Ustrcmp(path, "syslog") != 0) break;
151 }
152
5cb8cbc6
PH
153 if (path == NULL) /* No log files */
154 {
8e669ac1
PH
155 *inodeptr = -1;
156 return -1;
157 }
059ec3d9 158
8e669ac1
PH
159 /* An empty string means use the default, which is in the spool directory.
160 But don't just use the spool directory, as it is possible that the log
5cb8cbc6 161 subdirectory has been symbolically linked elsewhere. */
059ec3d9 162
8e669ac1 163 if (path[0] == 0)
059ec3d9 164 {
5cb8cbc6
PH
165 sprintf(CS buffer, CS"%s/log", CS spool_directory);
166 path = buffer;
8e669ac1
PH
167 }
168 else
059ec3d9 169 {
8e669ac1 170 uschar *cp;
5cb8cbc6 171 if ((cp = Ustrrchr(path, '/')) != NULL) *cp = 0;
8e669ac1 172 }
5cb8cbc6 173 }
8e669ac1 174
8f128379 175/* We now have the path; do the business */
5cb8cbc6
PH
176
177memset(&statbuf, 0, sizeof(statbuf));
178
179if (STATVFS(CS path, &statbuf) != 0)
180 {
181 log_write(0, LOG_MAIN|LOG_PANIC, "cannot accept message: failed to stat "
182 "%s directory %s: %s", name, spool_directory, strerror(errno));
183 smtp_closedown(US"spool or log directory problem");
184 exim_exit(EXIT_FAILURE);
185 }
8e669ac1 186
5cb8cbc6
PH
187*inodeptr = (statbuf.F_FILES > 0)? statbuf.F_FAVAIL : -1;
188
189/* Disks are getting huge. Take care with computing the size in kilobytes. */
8e669ac1 190
5cb8cbc6
PH
191return (int)(((double)statbuf.F_BAVAIL * (double)statbuf.F_FRSIZE)/1024.0);
192
193/* Unable to find partition sizes in this environment. */
194
195#else
196*inodeptr = -1;
197return -1;
198#endif
199}
200
059ec3d9 201
059ec3d9 202
5cb8cbc6
PH
203
204/*************************************************
205* Check space on spool and log partitions *
206*************************************************/
207
208/* This function is called before accepting a message; if any thresholds are
209set, it checks them. If a message_size is supplied, it checks that there is
210enough space for that size plus the threshold - i.e. that the message won't
211reduce the space to the threshold. Not all OS have statvfs(); for those that
212don't, this function always returns TRUE. For some OS the old function and
213struct name statfs is used; that is handled by a macro, defined in exim.h.
214
215Arguments:
216 msg_size the (estimated) size of an incoming message
217
218Returns: FALSE if there isn't enough space, or if the information cannot
219 be obtained
220 TRUE if no check was done or there is enough space
221*/
222
223BOOL
224receive_check_fs(int msg_size)
225{
226int space, inodes;
227
228if (check_spool_space > 0 || msg_size > 0 || check_spool_inodes > 0)
229 {
8e669ac1
PH
230 space = receive_statvfs(TRUE, &inodes);
231
059ec3d9 232 DEBUG(D_receive)
5cb8cbc6
PH
233 debug_printf("spool directory space = %dK inodes = %d "
234 "check_space = %dK inodes = %d msg_size = %d\n",
235 space, inodes, check_spool_space, check_spool_inodes, msg_size);
8e669ac1
PH
236
237 if ((space >= 0 && space < check_spool_space) ||
5cb8cbc6 238 (inodes >= 0 && inodes < check_spool_inodes))
8e669ac1 239 {
5cb8cbc6
PH
240 log_write(0, LOG_MAIN, "spool directory space check failed: space=%d "
241 "inodes=%d", space, inodes);
059ec3d9
PH
242 return FALSE;
243 }
244 }
245
5cb8cbc6
PH
246if (check_log_space > 0 || check_log_inodes > 0)
247 {
8e669ac1
PH
248 space = receive_statvfs(FALSE, &inodes);
249
5cb8cbc6
PH
250 DEBUG(D_receive)
251 debug_printf("log directory space = %dK inodes = %d "
252 "check_space = %dK inodes = %d\n",
253 space, inodes, check_log_space, check_log_inodes);
8e669ac1
PH
254
255 if ((space >= 0 && space < check_log_space) ||
5cb8cbc6 256 (inodes >= 0 && inodes < check_log_inodes))
8e669ac1 257 {
5cb8cbc6
PH
258 log_write(0, LOG_MAIN, "log directory space check failed: space=%d "
259 "inodes=%d", space, inodes);
260 return FALSE;
261 }
8e669ac1
PH
262 }
263
059ec3d9
PH
264return TRUE;
265}
266
267
268
269/*************************************************
270* Bomb out while reading a message *
271*************************************************/
272
273/* The common case of wanting to bomb out is if a SIGTERM or SIGINT is
274received, or if there is a timeout. A rarer case might be if the log files are
275screwed up and Exim can't open them to record a message's arrival. Handling
276that case is done by setting a flag to cause the log functions to call this
277function if there is an ultimate disaster. That is why it is globally
278accessible.
279
8f128379
PH
280Arguments:
281 reason text reason to pass to the not-quit ACL
282 msg default SMTP response to give if in an SMTP session
059ec3d9
PH
283Returns: it doesn't
284*/
285
286void
8f128379 287receive_bomb_out(uschar *reason, uschar *msg)
059ec3d9
PH
288{
289/* If spool_name is set, it contains the name of the data file that is being
290written. Unlink it before closing so that it cannot be picked up by a delivery
291process. Ensure that any header file is also removed. */
292
293if (spool_name[0] != 0)
294 {
295 Uunlink(spool_name);
296 spool_name[Ustrlen(spool_name) - 1] = 'H';
297 Uunlink(spool_name);
298 }
299
300/* Now close the file if it is open, either as a fd or a stream. */
301
f1e894f3
PH
302if (data_file != NULL) (void)fclose(data_file);
303 else if (data_fd >= 0) (void)close(data_fd);
059ec3d9 304
8f128379
PH
305/* Attempt to close down an SMTP connection tidily. For non-batched SMTP, call
306smtp_notquit_exit(), which runs the NOTQUIT ACL, if present, and handles the
307SMTP response. */
059ec3d9
PH
308
309if (smtp_input)
310 {
8f128379
PH
311 if (smtp_batched_input)
312 moan_smtp_batch(NULL, "421 %s - message abandoned", msg); /* No return */
313 smtp_notquit_exit(reason, US"421", US"%s %s - closing connection.",
314 smtp_active_hostname, msg);
059ec3d9
PH
315 }
316
317/* Exit from the program (non-BSMTP cases) */
318
319exim_exit(EXIT_FAILURE);
320}
321
322
323/*************************************************
324* Data read timeout *
325*************************************************/
326
327/* Handler function for timeouts that occur while reading the data that
328comprises a message.
329
330Argument: the signal number
331Returns: nothing
332*/
333
334static void
335data_timeout_handler(int sig)
336{
337uschar *msg = NULL;
338
339sig = sig; /* Keep picky compilers happy */
340
341if (smtp_input)
342 {
343 msg = US"SMTP incoming data timeout";
344 log_write(L_lost_incoming_connection,
345 LOG_MAIN, "SMTP data timeout (message abandoned) on connection "
fed77020
PH
346 "from %s F=<%s>",
347 (sender_fullhost != NULL)? sender_fullhost : US"local process",
348 sender_address);
059ec3d9
PH
349 }
350else
351 {
352 fprintf(stderr, "exim: timed out while reading - message abandoned\n");
353 log_write(L_lost_incoming_connection,
354 LOG_MAIN, "timed out while reading local message");
355 }
356
8f128379 357receive_bomb_out(US"data-timeout", msg); /* Does not return */
059ec3d9
PH
358}
359
360
361
362/*************************************************
363* local_scan() timeout *
364*************************************************/
365
366/* Handler function for timeouts that occur while running a local_scan()
367function.
368
369Argument: the signal number
370Returns: nothing
371*/
372
373static void
374local_scan_timeout_handler(int sig)
375{
376sig = sig; /* Keep picky compilers happy */
377log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function timed out - "
378 "message temporarily rejected (size %d)", message_size);
8f128379
PH
379/* Does not return */
380receive_bomb_out(US"local-scan-timeout", US"local verification problem");
059ec3d9
PH
381}
382
383
384
385/*************************************************
386* local_scan() crashed *
387*************************************************/
388
389/* Handler function for signals that occur while running a local_scan()
390function.
391
392Argument: the signal number
393Returns: nothing
394*/
395
396static void
397local_scan_crash_handler(int sig)
398{
399log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function crashed with "
400 "signal %d - message temporarily rejected (size %d)", sig, message_size);
8f128379
PH
401/* Does not return */
402receive_bomb_out(US"local-scan-error", US"local verification problem");
059ec3d9
PH
403}
404
405
406/*************************************************
407* SIGTERM or SIGINT received *
408*************************************************/
409
410/* Handler for SIGTERM or SIGINT signals that occur while reading the
411data that comprises a message.
412
413Argument: the signal number
414Returns: nothing
415*/
416
417static void
418data_sigterm_sigint_handler(int sig)
419{
420uschar *msg = NULL;
421
422if (smtp_input)
423 {
424 msg = US"Service not available - SIGTERM or SIGINT received";
425 log_write(0, LOG_MAIN, "%s closed after %s", smtp_get_connection_info(),
426 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
427 }
428else
429 {
f05da2e8 430 if (filter_test == FTEST_NONE)
059ec3d9
PH
431 {
432 fprintf(stderr, "\nexim: %s received - message abandoned\n",
433 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
434 log_write(0, LOG_MAIN, "%s received while reading local message",
435 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
436 }
437 }
438
8f128379 439receive_bomb_out(US"signal-exit", msg); /* Does not return */
059ec3d9
PH
440}
441
442
443
444/*************************************************
445* Add new recipient to list *
446*************************************************/
447
448/* This function builds a list of recipient addresses in argc/argv
449format.
450
451Arguments:
452 recipient the next address to add to recipients_list
453 pno parent number for fixed aliases; -1 otherwise
454
455Returns: nothing
456*/
457
458void
459receive_add_recipient(uschar *recipient, int pno)
460{
461if (recipients_count >= recipients_list_max)
462 {
463 recipient_item *oldlist = recipients_list;
464 int oldmax = recipients_list_max;
465 recipients_list_max = recipients_list_max? 2*recipients_list_max : 50;
466 recipients_list = store_get(recipients_list_max * sizeof(recipient_item));
467 if (oldlist != NULL)
468 memcpy(recipients_list, oldlist, oldmax * sizeof(recipient_item));
469 }
470
471recipients_list[recipients_count].address = recipient;
472recipients_list[recipients_count].pno = pno;
8523533c
TK
473#ifdef EXPERIMENTAL_BRIGHTMAIL
474recipients_list[recipients_count].bmi_optin = bmi_current_optin;
475/* reset optin string pointer for next recipient */
476bmi_current_optin = NULL;
477#endif
059ec3d9
PH
478recipients_list[recipients_count++].errors_to = NULL;
479}
480
481
482
483
484/*************************************************
485* Remove a recipient from the list *
486*************************************************/
487
488/* This function is provided for local_scan() to use.
489
490Argument:
491 recipient address to remove
492
493Returns: TRUE if it did remove something; FALSE otherwise
494*/
495
496BOOL
497receive_remove_recipient(uschar *recipient)
498{
499int count;
500DEBUG(D_receive) debug_printf("receive_remove_recipient(\"%s\") called\n",
501 recipient);
502for (count = 0; count < recipients_count; count++)
503 {
504 if (Ustrcmp(recipients_list[count].address, recipient) == 0)
505 {
506 if ((--recipients_count - count) > 0)
507 memmove(recipients_list + count, recipients_list + count + 1,
54cdb463 508 (recipients_count - count)*sizeof(recipient_item));
059ec3d9
PH
509 return TRUE;
510 }
511 }
512return FALSE;
513}
514
515
516
517
518
519/*************************************************
520* Read data portion of a non-SMTP message *
521*************************************************/
522
523/* This function is called to read the remainder of a message (following the
524header) when the input is not from SMTP - we are receiving a local message on
525a standard input stream. The message is always terminated by EOF, and is also
526terminated by a dot on a line by itself if the flag dot_ends is TRUE. Split the
527two cases for maximum efficiency.
528
529Ensure that the body ends with a newline. This will naturally be the case when
530the termination is "\n.\n" but may not be otherwise. The RFC defines messages
531as "sequences of lines" - this of course strictly applies only to SMTP, but
532deliveries into BSD-type mailbox files also require it. Exim used to have a
533flag for doing this at delivery time, but as it was always set for all
534transports, I decided to simplify things by putting the check here instead.
535
536There is at least one MUA (dtmail) that sends CRLF via this interface, and
537other programs are known to do this as well. Exim used to have a option for
538dealing with this: in July 2003, after much discussion, the code has been
539changed to default to treat any of LF, CRLF, and bare CR as line terminators.
540
541However, for the case when a dot on a line by itself terminates a message, the
542only recognized terminating sequences before and after the dot are LF and CRLF.
543Otherwise, having read EOL . CR, you don't know whether to read another
544character or not.
545
546Internally, in messages stored in Exim's spool files, LF is used as the line
547terminator. Under the new regime, bare CRs will no longer appear in these
548files.
549
550Arguments:
551 fout a FILE to which to write the message
552
553Returns: One of the END_xxx values indicating why it stopped reading
554*/
555
556static int
557read_message_data(FILE *fout)
558{
559int ch_state;
560register int ch;
d677b2f2 561register int linelength = 0;
059ec3d9
PH
562
563/* Handle the case when only EOF terminates the message */
564
565if (!dot_ends)
566 {
567 register int last_ch = '\n';
568
80a47a2c 569 for (; (ch = (receive_getc)()) != EOF; last_ch = ch)
059ec3d9
PH
570 {
571 if (ch == 0) body_zerocount++;
572 if (last_ch == '\r' && ch != '\n')
573 {
d677b2f2
PH
574 if (linelength > max_received_linelength)
575 max_received_linelength = linelength;
576 linelength = 0;
059ec3d9
PH
577 if (fputc('\n', fout) == EOF) return END_WERROR;
578 message_size++;
579 body_linecount++;
580 }
581 if (ch == '\r') continue;
582
583 if (fputc(ch, fout) == EOF) return END_WERROR;
d677b2f2
PH
584 if (ch == '\n')
585 {
586 if (linelength > max_received_linelength)
587 max_received_linelength = linelength;
588 linelength = 0;
589 body_linecount++;
590 }
591 else linelength++;
059ec3d9
PH
592 if (++message_size > thismessage_size_limit) return END_SIZE;
593 }
594
595 if (last_ch != '\n')
596 {
d677b2f2
PH
597 if (linelength > max_received_linelength)
598 max_received_linelength = linelength;
059ec3d9
PH
599 if (fputc('\n', fout) == EOF) return END_WERROR;
600 message_size++;
601 body_linecount++;
602 }
603
604 return END_EOF;
605 }
606
607/* Handle the case when a dot on a line on its own, or EOF, terminates. */
608
609ch_state = 1;
610
80a47a2c 611while ((ch = (receive_getc)()) != EOF)
059ec3d9
PH
612 {
613 if (ch == 0) body_zerocount++;
614 switch (ch_state)
615 {
616 case 0: /* Normal state (previous char written) */
617 if (ch == '\n')
d677b2f2
PH
618 {
619 body_linecount++;
620 if (linelength > max_received_linelength)
621 max_received_linelength = linelength;
622 linelength = -1;
623 ch_state = 1;
624 }
059ec3d9
PH
625 else if (ch == '\r')
626 { ch_state = 2; continue; }
627 break;
628
629 case 1: /* After written "\n" */
630 if (ch == '.') { ch_state = 3; continue; }
d677b2f2 631 if (ch != '\n') ch_state = 0; else linelength = -1;
059ec3d9
PH
632 break;
633
634 case 2:
635 body_linecount++; /* After unwritten "\r" */
d677b2f2
PH
636 if (linelength > max_received_linelength)
637 max_received_linelength = linelength;
059ec3d9 638 if (ch == '\n')
d677b2f2
PH
639 {
640 ch_state = 1;
641 linelength = -1;
642 }
059ec3d9
PH
643 else
644 {
645 if (message_size++, fputc('\n', fout) == EOF) return END_WERROR;
646 if (ch == '\r') continue;
647 ch_state = 0;
d677b2f2 648 linelength = 0;
059ec3d9
PH
649 }
650 break;
651
652 case 3: /* After "\n." (\n written, dot not) */
653 if (ch == '\n') return END_DOT;
654 if (ch == '\r') { ch_state = 4; continue; }
655 message_size++;
d677b2f2 656 linelength++;
059ec3d9
PH
657 if (fputc('.', fout) == EOF) return END_WERROR;
658 ch_state = 0;
659 break;
660
661 case 4: /* After "\n.\r" (\n written, rest not) */
662 if (ch == '\n') return END_DOT;
663 message_size += 2;
664 body_linecount++;
665 if (fputs(".\n", fout) == EOF) return END_WERROR;
666 if (ch == '\r') { ch_state = 2; continue; }
667 ch_state = 0;
668 break;
669 }
670
d677b2f2 671 linelength++;
059ec3d9
PH
672 if (fputc(ch, fout) == EOF) return END_WERROR;
673 if (++message_size > thismessage_size_limit) return END_SIZE;
674 }
675
676/* Get here if EOF read. Unless we have just written "\n", we need to ensure
677the message ends with a newline, and we must also write any characters that
678were saved up while testing for an ending dot. */
679
680if (ch_state != 1)
681 {
682 static uschar *ends[] = { US"\n", NULL, US"\n", US".\n", US".\n" };
683 if (fputs(CS ends[ch_state], fout) == EOF) return END_WERROR;
684 message_size += Ustrlen(ends[ch_state]);
685 body_linecount++;
686 }
687
688return END_EOF;
689}
690
691
692
693
694/*************************************************
695* Read data portion of an SMTP message *
696*************************************************/
697
698/* This function is called to read the remainder of an SMTP message (after the
699headers), or to skip over it when an error has occurred. In this case, the
700output file is passed as NULL.
701
702If any line begins with a dot, that character is skipped. The input should only
703be successfully terminated by CR LF . CR LF unless it is local (non-network)
704SMTP, in which case the CRs are optional, but...
705
706FUDGE: It seems that sites on the net send out messages with just LF
707terminators, despite the warnings in the RFCs, and other MTAs handle this. So
708we make the CRs optional in all cases.
709
710July 2003: Bare CRs cause trouble. We now treat them as line terminators as
711well, so that there are no CRs in spooled messages. However, the message
712terminating dot is not recognized between two bare CRs.
713
714Arguments:
715 fout a FILE to which to write the message; NULL if skipping
716
717Returns: One of the END_xxx values indicating why it stopped reading
718*/
719
720static int
721read_message_data_smtp(FILE *fout)
722{
723int ch_state = 0;
e4bdf652 724int ch;
1f5497b2 725register int linelength = 0;
059ec3d9 726
80a47a2c 727while ((ch = (receive_getc)()) != EOF)
059ec3d9
PH
728 {
729 if (ch == 0) body_zerocount++;
730 switch (ch_state)
731 {
732 case 0: /* After LF or CRLF */
733 if (ch == '.')
734 {
735 ch_state = 3;
736 continue; /* Don't ever write . after LF */
737 }
738 ch_state = 1;
739
740 /* Else fall through to handle as normal uschar. */
741
742 case 1: /* Normal state */
743 if (ch == '\n')
744 {
745 ch_state = 0;
746 body_linecount++;
1f5497b2
PH
747 if (linelength > max_received_linelength)
748 max_received_linelength = linelength;
749 linelength = -1;
059ec3d9
PH
750 }
751 else if (ch == '\r')
752 {
753 ch_state = 2;
754 continue;
755 }
756 break;
757
758 case 2: /* After (unwritten) CR */
759 body_linecount++;
1f5497b2
PH
760 if (linelength > max_received_linelength)
761 max_received_linelength = linelength;
762 linelength = -1;
059ec3d9
PH
763 if (ch == '\n')
764 {
765 ch_state = 0;
766 }
767 else
768 {
769 message_size++;
770 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
e4bdf652 771 (void) cutthrough_put_nl();
059ec3d9
PH
772 if (ch != '\r') ch_state = 1; else continue;
773 }
774 break;
775
776 case 3: /* After [CR] LF . */
777 if (ch == '\n')
778 return END_DOT;
779 if (ch == '\r')
780 {
781 ch_state = 4;
782 continue;
783 }
784 ch_state = 1; /* The dot itself is removed */
785 break;
786
787 case 4: /* After [CR] LF . CR */
788 if (ch == '\n') return END_DOT;
789 message_size++;
790 body_linecount++;
791 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
e4bdf652 792 (void) cutthrough_put_nl();
059ec3d9
PH
793 if (ch == '\r')
794 {
795 ch_state = 2;
796 continue;
797 }
798 ch_state = 1;
799 break;
800 }
801
802 /* Add the character to the spool file, unless skipping; then loop for the
803 next. */
804
805 message_size++;
1f5497b2 806 linelength++;
059ec3d9
PH
807 if (fout != NULL)
808 {
809 if (fputc(ch, fout) == EOF) return END_WERROR;
810 if (message_size > thismessage_size_limit) return END_SIZE;
811 }
e4bdf652
JH
812 if(ch == '\n')
813 (void) cutthrough_put_nl();
814 else
815 {
816 uschar c= ch;
817 (void) cutthrough_puts(&c, 1);
818 }
059ec3d9
PH
819 }
820
821/* Fall through here if EOF encountered. This indicates some kind of error,
822since a correct message is terminated by [CR] LF . [CR] LF. */
823
824return END_EOF;
825}
826
827
828
829
830/*************************************************
831* Swallow SMTP message *
832*************************************************/
833
834/* This function is called when there has been some kind of error while reading
835an SMTP message, and the remaining data may need to be swallowed. It is global
836because it is called from smtp_closedown() to shut down an incoming call
837tidily.
838
839Argument: a FILE from which to read the message
840Returns: nothing
841*/
842
843void
844receive_swallow_smtp(void)
845{
846if (message_ended >= END_NOTENDED)
847 message_ended = read_message_data_smtp(NULL);
848}
849
850
851
852/*************************************************
853* Handle lost SMTP connection *
854*************************************************/
855
856/* This function logs connection loss incidents and generates an appropriate
857SMTP response.
858
859Argument: additional data for the message
860Returns: the SMTP response
861*/
862
863static uschar *
864handle_lost_connection(uschar *s)
865{
866log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
867 "%s lost while reading message data%s", smtp_get_connection_info(), s);
868return US"421 Lost incoming connection";
869}
870
871
872
873
874/*************************************************
875* Handle a non-smtp reception error *
876*************************************************/
877
878/* This function is called for various errors during the reception of non-SMTP
879messages. It either sends a message to the sender of the problem message, or it
880writes to the standard error stream.
881
882Arguments:
883 errcode code for moan_to_sender(), identifying the error
884 text1 first message text, passed to moan_to_sender()
885 text2 second message text, used only for stderrr
886 error_rc code to pass to exim_exit if no problem
887 f FILE containing body of message (may be stdin)
888 hptr pointer to instore headers or NULL
889
890Returns: calls exim_exit(), which does not return
891*/
892
893static void
894give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
895 FILE *f, header_line *hptr)
896{
897if (error_handling == ERRORS_SENDER)
898 {
899 error_block eblock;
900 eblock.next = NULL;
901 eblock.text1 = text1;
902 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
903 error_rc = EXIT_FAILURE;
904 }
905else fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 906(void)fclose(f);
059ec3d9
PH
907exim_exit(error_rc);
908}
909
910
911
912/*************************************************
913* Add header lines set up by ACL *
914*************************************************/
915
850635b6
PH
916/* This function is called to add the header lines that were set up by
917statements in an ACL to the list of headers in memory. It is done in two stages
918like this, because when the ACL for RCPT is running, the other headers have not
919yet been received. This function is called twice; once just before running the
920DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
921are visible to the DATA ACL.
059ec3d9
PH
922
923Originally these header lines were added at the end. Now there is support for
924three different places: top, bottom, and after the Received: header(s). There
925will always be at least one Received: header, even if it is marked deleted, and
926even if something else has been put in front of it.
927
928Arguments:
929 acl_name text to identify which ACL
930
931Returns: nothing
932*/
933
934static void
935add_acl_headers(uschar *acl_name)
936{
937header_line *h, *next;
938header_line *last_received = NULL;
e7568d51
TL
939
940if (acl_removed_headers != NULL)
941 {
942 DEBUG(D_receive|D_acl) debug_printf(">>Headers removed by %s ACL:\n", acl_name);
943
944 for (h = header_list; h != NULL; h = h->next)
945 {
e7568d51
TL
946 uschar *list;
947 BOOL include_header;
948
949 if (h->type == htype_old) continue;
950
951 include_header = TRUE;
952 list = acl_removed_headers;
953
954 int sep = ':'; /* This is specified as a colon-separated list */
955 uschar *s;
956 uschar buffer[128];
957 while ((s = string_nextinlist(&list, &sep, buffer, sizeof(buffer)))
958 != NULL)
959 {
960 int len = Ustrlen(s);
961 if (header_testname(h, s, len, FALSE))
962 {
963 h->type = htype_old;
964 DEBUG(D_receive|D_acl) debug_printf(" %s", h->text);
965 }
966 }
967 }
968 acl_removed_headers = NULL;
969 DEBUG(D_receive|D_acl) debug_printf(">>\n");
970 }
059ec3d9 971
71fafd95 972if (acl_added_headers == NULL) return;
059ec3d9
PH
973DEBUG(D_receive|D_acl) debug_printf(">>Headers added by %s ACL:\n", acl_name);
974
71fafd95 975for (h = acl_added_headers; h != NULL; h = next)
059ec3d9
PH
976 {
977 next = h->next;
978
979 switch(h->type)
980 {
981 case htype_add_top:
982 h->next = header_list;
983 header_list = h;
984 DEBUG(D_receive|D_acl) debug_printf(" (at top)");
985 break;
986
987 case htype_add_rec:
988 if (last_received == NULL)
989 {
990 last_received = header_list;
991 while (!header_testname(last_received, US"Received", 8, FALSE))
992 last_received = last_received->next;
993 while (last_received->next != NULL &&
994 header_testname(last_received->next, US"Received", 8, FALSE))
995 last_received = last_received->next;
996 }
997 h->next = last_received->next;
998 last_received->next = h;
999 DEBUG(D_receive|D_acl) debug_printf(" (after Received:)");
1000 break;
1001
8523533c
TK
1002 case htype_add_rfc:
1003 /* add header before any header which is NOT Received: or Resent- */
1004 last_received = header_list;
1005 while ( (last_received->next != NULL) &&
1006 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
1007 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
1008 last_received = last_received->next;
1009 /* last_received now points to the last Received: or Resent-* header
1010 in an uninterrupted chain of those header types (seen from the beginning
1011 of all headers. Our current header must follow it. */
1012 h->next = last_received->next;
1013 last_received->next = h;
8e669ac1 1014 DEBUG(D_receive|D_acl) debug_printf(" (before any non-Received: or Resent-*: header)");
8523533c
TK
1015 break;
1016
059ec3d9
PH
1017 default:
1018 h->next = NULL;
1019 header_last->next = h;
1020 break;
1021 }
1022
1023 if (h->next == NULL) header_last = h;
1024
1025 /* Check for one of the known header types (From:, To:, etc.) though in
1026 practice most added headers are going to be "other". Lower case
1027 identification letters are never stored with the header; they are used
1028 for existence tests when messages are received. So discard any lower case
1029 flag values. */
1030
1031 h->type = header_checkname(h, FALSE);
1032 if (h->type >= 'a') h->type = htype_other;
1033
1034 DEBUG(D_receive|D_acl) debug_printf(" %s", header_last->text);
1035 }
1036
71fafd95 1037acl_added_headers = NULL;
059ec3d9
PH
1038DEBUG(D_receive|D_acl) debug_printf(">>\n");
1039}
1040
1041
1042
1043/*************************************************
1044* Add host information for log line *
1045*************************************************/
1046
1047/* Called for acceptance and rejecting log lines. This adds information about
1048the calling host to a string that is being built dynamically.
1049
1050Arguments:
1051 s the dynamic string
1052 sizeptr points to the size variable
1053 ptrptr points to the pointer variable
1054
1055Returns: the extended string
1056*/
1057
1058static uschar *
1059add_host_info_for_log(uschar *s, int *sizeptr, int *ptrptr)
1060{
1061if (sender_fullhost != NULL)
1062 {
1063 s = string_append(s, sizeptr, ptrptr, 2, US" H=", sender_fullhost);
1064 if ((log_extra_selector & LX_incoming_interface) != 0 &&
1065 interface_address != NULL)
1066 {
1067 uschar *ss = string_sprintf(" I=[%s]:%d", interface_address,
1068 interface_port);
1069 s = string_cat(s, sizeptr, ptrptr, ss, Ustrlen(ss));
1070 }
1071 }
1072if (sender_ident != NULL)
1073 s = string_append(s, sizeptr, ptrptr, 2, US" U=", sender_ident);
1074if (received_protocol != NULL)
1075 s = string_append(s, sizeptr, ptrptr, 2, US" P=", received_protocol);
1076return s;
1077}
1078
1079
1080
63955bf2 1081#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1082
1083/*************************************************
54cdb463
PH
1084* Run the MIME ACL on a message *
1085*************************************************/
1086
1087/* This code is in a subroutine so that it can be used for both SMTP
1088and non-SMTP messages. It is called with a non-NULL ACL pointer.
1089
1090Arguments:
1091 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1092 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1093 smtp_reply_ptr Where SMTP reply is being built
1094 blackholed_by_ptr Where "blackholed by" message is being built
1095
1096Returns: TRUE to carry on; FALSE to abandon the message
1097*/
1098
1099static BOOL
1100run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1101 uschar **blackholed_by_ptr)
1102{
1103FILE *mbox_file;
1104uschar rfc822_file_path[2048];
1105unsigned long mbox_size;
1106header_line *my_headerlist;
1107uschar *user_msg, *log_msg;
1108int mime_part_count_buffer = -1;
7156b1ef 1109int rc = OK;
54cdb463
PH
1110
1111memset(CS rfc822_file_path,0,2048);
1112
1113/* check if it is a MIME message */
1114my_headerlist = header_list;
4e88a19f
PH
1115while (my_headerlist != NULL)
1116 {
54cdb463 1117 /* skip deleted headers */
4e88a19f
PH
1118 if (my_headerlist->type == '*')
1119 {
54cdb463
PH
1120 my_headerlist = my_headerlist->next;
1121 continue;
4e88a19f
PH
1122 }
1123 if (strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0)
1124 {
54cdb463
PH
1125 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1126 goto DO_MIME_ACL;
4e88a19f 1127 }
54cdb463 1128 my_headerlist = my_headerlist->next;
4e88a19f 1129 }
54cdb463
PH
1130
1131DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1132return TRUE;
1133
1134DO_MIME_ACL:
1135/* make sure the eml mbox file is spooled up */
8544e77a 1136mbox_file = spool_mbox(&mbox_size, NULL);
54cdb463
PH
1137if (mbox_file == NULL) {
1138 /* error while spooling */
1139 log_write(0, LOG_MAIN|LOG_PANIC,
1140 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1141 Uunlink(spool_name);
1142 unspool_mbox();
6f0c431a
PP
1143#ifdef EXPERIMENTAL_DCC
1144 dcc_ok = 0;
1145#endif
a5bd321b 1146 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1147 message_id[0] = 0; /* Indicate no message accepted */
1148 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1149 return FALSE; /* Indicate skip to end of receive function */
1150};
1151
1152mime_is_rfc822 = 0;
1153
1154MIME_ACL_CHECK:
1155mime_part_count = -1;
1156rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1157(void)fclose(mbox_file);
54cdb463 1158
4e88a19f
PH
1159if (Ustrlen(rfc822_file_path) > 0)
1160 {
54cdb463
PH
1161 mime_part_count = mime_part_count_buffer;
1162
4e88a19f
PH
1163 if (unlink(CS rfc822_file_path) == -1)
1164 {
54cdb463
PH
1165 log_write(0, LOG_PANIC,
1166 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1167 goto END_MIME_ACL;
4e88a19f
PH
1168 }
1169 }
54cdb463
PH
1170
1171/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1172if (rc == OK)
1173 {
54cdb463
PH
1174 uschar temp_path[1024];
1175 int n;
1176 struct dirent *entry;
1177 DIR *tempdir;
1178
b07e6aa3
PH
1179 (void)string_format(temp_path, 1024, "%s/scan/%s", spool_directory,
1180 message_id);
54cdb463 1181
4e88a19f
PH
1182 tempdir = opendir(CS temp_path);
1183 n = 0;
1184 do
1185 {
1186 entry = readdir(tempdir);
1187 if (entry == NULL) break;
1188 if (strncmpic(US entry->d_name,US"__rfc822_",9) == 0)
1189 {
b07e6aa3 1190 (void)string_format(rfc822_file_path, 2048,"%s/scan/%s/%s", spool_directory, message_id, entry->d_name);
4e88a19f
PH
1191 debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n", rfc822_file_path);
1192 break;
1193 }
1194 } while (1);
1195 closedir(tempdir);
54cdb463 1196
4e88a19f
PH
1197 if (entry != NULL)
1198 {
c58b88df 1199 mbox_file = Ufopen(rfc822_file_path,"rb");
4e88a19f
PH
1200 if (mbox_file == NULL)
1201 {
54cdb463
PH
1202 log_write(0, LOG_PANIC,
1203 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1204 unlink(CS rfc822_file_path);
1205 goto END_MIME_ACL;
4e88a19f 1206 }
54cdb463
PH
1207 /* set RFC822 expansion variable */
1208 mime_is_rfc822 = 1;
1209 mime_part_count_buffer = mime_part_count;
1210 goto MIME_ACL_CHECK;
4e88a19f
PH
1211 }
1212 }
54cdb463
PH
1213
1214END_MIME_ACL:
1215add_acl_headers(US"MIME");
1216if (rc == DISCARD)
1217 {
1218 recipients_count = 0;
1219 *blackholed_by_ptr = US"MIME ACL";
1220 }
1221else if (rc != OK)
1222 {
1223 Uunlink(spool_name);
1224 unspool_mbox();
6f0c431a
PP
1225#ifdef EXPERIMENTAL_DCC
1226 dcc_ok = 0;
1227#endif
54cdb463
PH
1228 if (smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1229 *smtp_yield_ptr = FALSE; /* No more messsages after dropped connection */
1230 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1231 message_id[0] = 0; /* Indicate no message accepted */
1232 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1233 }
54cdb463
PH
1234
1235return TRUE;
1236}
1237
63955bf2 1238#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1239
1240
e4bdf652
JH
1241
1242void
1243received_header_gen(void)
1244{
1245uschar *received;
1246uschar *timestamp;
1247header_line *received_header= header_list;
1248
1249timestamp = expand_string(US"${tod_full}");
1250if (recipients_count == 1) received_for = recipients_list[0].address;
1251received = expand_string(received_header_text);
1252received_for = NULL;
1253
1254if (received == NULL)
1255 {
1256 if(spool_name[0] != 0)
1257 Uunlink(spool_name); /* Lose the data file */
1258 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1259 "(received_header_text) failed: %s", string_printing(received_header_text),
1260 expand_string_message);
1261 }
1262
1263/* The first element on the header chain is reserved for the Received header,
1264so all we have to do is fill in the text pointer, and set the type. However, if
1265the result of the expansion is an empty string, we leave the header marked as
1266"old" so as to refrain from adding a Received header. */
1267
1268if (received[0] == 0)
1269 {
1270 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1271 received_header->type = htype_old;
1272 }
1273else
1274 {
1275 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1276 received_header->type = htype_received;
1277 }
1278
1279received_header->slen = Ustrlen(received_header->text);
1280
1281DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1282 received_header->type, received_header->text);
1283}
1284
1285
1286
54cdb463 1287/*************************************************
059ec3d9
PH
1288* Receive message *
1289*************************************************/
1290
1291/* Receive a message on the given input, and put it into a pair of spool files.
1292Either a non-null list of recipients, or the extract flag will be true, or
1293both. The flag sender_local is true for locally generated messages. The flag
1294submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1295suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1296suppress_local_fixups" or -G was passed on the command-line.
1297The flag smtp_input is true if the message is to be
8800895a
PH
1298handled using SMTP conventions about termination and lines starting with dots.
1299For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1300
1301If a message was successfully read, message_id[0] will be non-zero.
1302
1303The general actions of this function are:
1304
1305 . Read the headers of the message (if any) into a chain of store
1306 blocks.
1307
1308 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1309 throw it away, unless the caller is trusted, or unless
1310 active_local_sender_retain is set - which can only happen if
1311 active_local_from_check is false.
059ec3d9
PH
1312
1313 . If recipients are to be extracted from the message, build the
1314 recipients list from the headers, removing any that were on the
1315 original recipients list (unless extract_addresses_remove_arguments is
1316 false), and at the same time, remove any bcc header that may be present.
1317
1318 . Get the spool file for the data, sort out its unique name, open
1319 and lock it (but don't give it the name yet).
1320
1321 . Generate a "Message-Id" header if the message doesn't have one, for
1322 locally-originated messages.
1323
1324 . Generate a "Received" header.
1325
1326 . Ensure the recipients list is fully qualified and rewritten if necessary.
1327
1328 . If there are any rewriting rules, apply them to the sender address
1329 and also to the headers.
1330
1331 . If there is no from: header, generate one, for locally-generated messages
1332 and messages in "submission mode" only.
1333
1334 . If the sender is local, check that from: is correct, and if not, generate
1335 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1336 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1337
1338 . If there is no "date" header, generate one, for locally-originated
1339 or submission mode messages only.
1340
1341 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1342 dot_ends mode, to the data file. Leave it open, to hold the lock.
1343
1344 . Write the envelope and the headers to a new file.
1345
1346 . Set the name for the header file; close it.
1347
1348 . Set the name for the data file; close it.
1349
1350Because this function can potentially be called many times in a single
1351SMTP connection, all store should be got by store_get(), so that it will be
1352automatically retrieved after the message is accepted.
1353
1354FUDGE: It seems that sites on the net send out messages with just LF
1355terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1356we make the CRs optional in all cases.
1357
1358July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1359new regime is now in place in which bare CRs in header lines are turned into LF
1360followed by a space, so as not to terminate the header line.
1361
1362February 2004: A bare LF in a header line in a message whose first line was
1363terminated by CRLF is treated in the same way as a bare CR.
1364
1365Arguments:
1366 extract_recip TRUE if recipients are to be extracted from the message's
1367 headers
1368
1369Returns: TRUE there are more messages to be read (SMTP input)
1370 FALSE there are no more messages to be read (non-SMTP input
1371 or SMTP connection collapsed, or other failure)
1372
1373When reading a message for filter testing, the returned value indicates
1374whether the headers (which is all that is read) were terminated by '.' or
1375not. */
1376
1377BOOL
1378receive_msg(BOOL extract_recip)
1379{
7156b1ef
NM
1380int i;
1381int rc = FAIL;
059ec3d9
PH
1382int msg_size = 0;
1383int process_info_len = Ustrlen(process_info);
1384int error_rc = (error_handling == ERRORS_SENDER)?
1385 errors_sender_rc : EXIT_FAILURE;
1386int header_size = 256;
1387int start, end, domain, size, sptr;
1388int id_resolution;
1389int had_zero = 0;
d677b2f2 1390int prevlines_length = 0;
059ec3d9
PH
1391
1392register int ptr = 0;
1393
1394BOOL contains_resent_headers = FALSE;
1395BOOL extracted_ignored = FALSE;
1396BOOL first_line_ended_crlf = TRUE_UNSET;
1397BOOL smtp_yield = TRUE;
1398BOOL yield = FALSE;
1399
1400BOOL resents_exist = FALSE;
1401uschar *resent_prefix = US"";
1402uschar *blackholed_by = NULL;
04f7d5b9 1403uschar *blackhole_log_msg = US"";
4ea50c5f 1404int cutthrough_done = 0;
059ec3d9
PH
1405
1406flock_t lock_data;
1407error_block *bad_addresses = NULL;
1408
1409uschar *frozen_by = NULL;
1410uschar *queued_by = NULL;
1411
1412uschar *errmsg, *s;
1413struct stat statbuf;
1414
4e88a19f 1415/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1416
1417uschar *smtp_reply = NULL;
4e88a19f 1418uschar *user_msg, *log_msg;
059ec3d9
PH
1419
1420/* Working header pointers */
1421
1422header_line *h, *next;
1423
2cbb4081 1424/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1425
1426BOOL date_header_exists = FALSE;
1427
1428/* Pointers to receive the addresses of headers whose contents we need. */
1429
1430header_line *from_header = NULL;
1431header_line *subject_header = NULL;
1432header_line *msgid_header = NULL;
1433header_line *received_header;
1434
1435/* Variables for use when building the Received: header. */
1436
059ec3d9
PH
1437uschar *timestamp;
1438int tslen;
1439
1440/* Release any open files that might have been cached while preparing to
1441accept the message - e.g. by verifying addresses - because reading a message
1442might take a fair bit of real time. */
1443
1444search_tidyup();
1445
e4bdf652
JH
1446/* Extracting the recipient list from an input file is incompatible with
1447cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1448to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1449if (extract_recip || !smtp_input)
2e5b33cd 1450 cancel_cutthrough_connection("not smtp input");
e4bdf652 1451
059ec3d9
PH
1452/* Initialize the chain of headers by setting up a place-holder for Received:
1453header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1454pointing to the end of the chain to make adding headers simple. */
1455
1456received_header = header_list = header_last = store_get(sizeof(header_line));
1457header_list->next = NULL;
1458header_list->type = htype_old;
1459header_list->text = NULL;
1460header_list->slen = 0;
1461
1462/* Control block for the next header to be read. */
1463
1464next = store_get(sizeof(header_line));
1465next->text = store_get(header_size);
1466
1467/* Initialize message id to be null (indicating no message read), and the
1468header names list to be the normal list. Indicate there is no data file open
1469yet, initialize the size and warning count, and deal with no size limit. */
1470
1471message_id[0] = 0;
1472data_file = NULL;
1473data_fd = -1;
1474spool_name[0] = 0;
1475message_size = 0;
1476warning_count = 0;
d677b2f2 1477received_count = 1; /* For the one we will add */
059ec3d9
PH
1478
1479if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1480
2e0c1448 1481/* While reading the message, the following counts are computed. */
059ec3d9 1482
d677b2f2
PH
1483message_linecount = body_linecount = body_zerocount =
1484 max_received_linelength = 0;
059ec3d9 1485
80a47a2c
TK
1486#ifndef DISABLE_DKIM
1487/* Call into DKIM to set up the context. */
1488if (smtp_input && !smtp_batched_input && !dkim_disable_verify) dkim_exim_verify_init();
fb2274d4
TK
1489#endif
1490
059ec3d9
PH
1491/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1492ids, and fractions of a second are required. See the comments that precede the
1493message id creation below. */
1494
1495(void)gettimeofday(&message_id_tv, NULL);
1496
1497/* For other uses of the received time we can operate with granularity of one
1498second, and for that we use the global variable received_time. This is for
1499things like ultimate message timeouts. */
1500
1501received_time = message_id_tv.tv_sec;
1502
1503/* If SMTP input, set the special handler for timeouts. The alarm() calls
1504happen in the smtp_getc() function when it refills its buffer. */
1505
1506if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1507
1508/* If not SMTP input, timeout happens only if configured, and we just set a
1509single timeout for the whole message. */
1510
1511else if (receive_timeout > 0)
1512 {
1513 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1514 alarm(receive_timeout);
1515 }
1516
1517/* SIGTERM and SIGINT are caught always. */
1518
1519signal(SIGTERM, data_sigterm_sigint_handler);
1520signal(SIGINT, data_sigterm_sigint_handler);
1521
1522/* Header lines in messages are not supposed to be very long, though when
1523unfolded, to: and cc: headers can take up a lot of store. We must also cope
1524with the possibility of junk being thrown at us. Start by getting 256 bytes for
1525storing the header, and extend this as necessary using string_cat().
1526
1527To cope with total lunacies, impose an upper limit on the length of the header
1528section of the message, as otherwise the store will fill up. We must also cope
1529with the possibility of binary zeros in the data. Hence we cannot use fgets().
1530Folded header lines are joined into one string, leaving the '\n' characters
1531inside them, so that writing them out reproduces the input.
1532
1533Loop for each character of each header; the next structure for chaining the
1534header is set up already, with ptr the offset of the next character in
1535next->text. */
1536
1537for (;;)
1538 {
80a47a2c 1539 int ch = (receive_getc)();
059ec3d9
PH
1540
1541 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1542 SMTP must have a correct "." terminator. */
1543
1544 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1545 {
1546 smtp_reply = handle_lost_connection(US" (header)");
1547 smtp_yield = FALSE;
1548 goto TIDYUP; /* Skip to end of function */
1549 }
1550
1551 /* See if we are at the current header's size limit - there must be at least
1552 four bytes left. This allows for the new character plus a zero, plus two for
1553 extra insertions when we are playing games with dots and carriage returns. If
1554 we are at the limit, extend the text buffer. This could have been done
1555 automatically using string_cat() but because this is a tightish loop storing
1556 only one character at a time, we choose to do it inline. Normally
1557 store_extend() will be able to extend the block; only at the end of a big
1558 store block will a copy be needed. To handle the case of very long headers
1559 (and sometimes lunatic messages can have ones that are 100s of K long) we
1560 call store_release() for strings that have been copied - if the string is at
1561 the start of a block (and therefore the only thing in it, because we aren't
1562 doing any other gets), the block gets freed. We can only do this because we
1563 know there are no other calls to store_get() going on. */
1564
1565 if (ptr >= header_size - 4)
1566 {
1567 int oldsize = header_size;
1568 /* header_size += 256; */
1569 header_size *= 2;
1570 if (!store_extend(next->text, oldsize, header_size))
1571 {
1572 uschar *newtext = store_get(header_size);
1573 memcpy(newtext, next->text, ptr);
1574 store_release(next->text);
1575 next->text = newtext;
1576 }
1577 }
1578
1579 /* Cope with receiving a binary zero. There is dispute about whether
1580 these should be allowed in RFC 822 messages. The middle view is that they
1581 should not be allowed in headers, at least. Exim takes this attitude at
1582 the moment. We can't just stomp on them here, because we don't know that
1583 this line is a header yet. Set a flag to cause scanning later. */
1584
1585 if (ch == 0) had_zero++;
1586
1587 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1588 those from data files use just LF. Treat LF in local SMTP input as a
1589 terminator too. Treat EOF as a line terminator always. */
1590
1591 if (ch == EOF) goto EOL;
1592
1593 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1594 other MTAs accept this. We are therefore forced into this "liberalisation"
1595 too, so we accept LF as a line terminator whatever the source of the message.
1596 However, if the first line of the message ended with a CRLF, we treat a bare
1597 LF specially by inserting a white space after it to ensure that the header
1598 line is not terminated. */
1599
1600 if (ch == '\n')
1601 {
1602 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1603 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1604 goto EOL;
1605 }
1606
1607 /* This is not the end of the line. If this is SMTP input and this is
1608 the first character in the line and it is a "." character, ignore it.
1609 This implements the dot-doubling rule, though header lines starting with
1610 dots aren't exactly common. They are legal in RFC 822, though. If the
1611 following is CRLF or LF, this is the line that that terminates the
1612 entire message. We set message_ended to indicate this has happened (to
1613 prevent further reading), and break out of the loop, having freed the
1614 empty header, and set next = NULL to indicate no data line. */
1615
1616 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1617 {
80a47a2c 1618 ch = (receive_getc)();
059ec3d9
PH
1619 if (ch == '\r')
1620 {
80a47a2c 1621 ch = (receive_getc)();
059ec3d9
PH
1622 if (ch != '\n')
1623 {
80a47a2c 1624 receive_ungetc(ch);
059ec3d9
PH
1625 ch = '\r'; /* Revert to CR */
1626 }
1627 }
1628 if (ch == '\n')
1629 {
1630 message_ended = END_DOT;
1631 store_reset(next);
1632 next = NULL;
1633 break; /* End character-reading loop */
1634 }
1635
1636 /* For non-SMTP input, the dot at the start of the line was really a data
1637 character. What is now in ch is the following character. We guaranteed
1638 enough space for this above. */
1639
1640 if (!smtp_input)
1641 {
1642 next->text[ptr++] = '.';
1643 message_size++;
1644 }
1645 }
1646
1647 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1648 remember this case if this is the first line ending. */
1649
1650 if (ch == '\r')
1651 {
80a47a2c 1652 ch = (receive_getc)();
059ec3d9
PH
1653 if (ch == '\n')
1654 {
1655 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1656 goto EOL;
1657 }
1658
1659 /* Otherwise, put back the character after CR, and turn the bare CR
1660 into LF SP. */
1661
80a47a2c 1662 ch = (receive_ungetc)(ch);
059ec3d9
PH
1663 next->text[ptr++] = '\n';
1664 message_size++;
1665 ch = ' ';
1666 }
1667
1668 /* We have a data character for the header line. */
1669
1670 next->text[ptr++] = ch; /* Add to buffer */
1671 message_size++; /* Total message size so far */
1672
1673 /* Handle failure due to a humungously long header section. The >= allows
1674 for the terminating \n. Add what we have so far onto the headers list so
1675 that it gets reflected in any error message, and back up the just-read
1676 character. */
1677
1678 if (message_size >= header_maxsize)
1679 {
1680 next->text[ptr] = 0;
1681 next->slen = ptr;
1682 next->type = htype_other;
1683 next->next = NULL;
1684 header_last->next = next;
1685 header_last = next;
1686
1687 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1688 "%s (more than %d characters): message abandoned",
1689 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1690
1691 if (smtp_input)
1692 {
1693 smtp_reply = US"552 Message header is ridiculously long";
1694 receive_swallow_smtp();
1695 goto TIDYUP; /* Skip to end of function */
1696 }
1697
1698 else
1699 {
1700 give_local_error(ERRMESS_VLONGHEADER,
1701 string_sprintf("message header longer than %d characters received: "
1702 "message not accepted", header_maxsize), US"", error_rc, stdin,
1703 header_list->next);
1704 /* Does not return */
1705 }
1706 }
1707
1708 continue; /* With next input character */
1709
1710 /* End of header line reached */
1711
1712 EOL:
2e0c1448
PH
1713
1714 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1715
1716 receive_linecount++;
1717 message_linecount++;
059ec3d9 1718
d677b2f2
PH
1719 /* Keep track of maximum line length */
1720
1721 if (ptr - prevlines_length > max_received_linelength)
1722 max_received_linelength = ptr - prevlines_length;
1723 prevlines_length = ptr + 1;
1724
059ec3d9
PH
1725 /* Now put in the terminating newline. There is always space for
1726 at least two more characters. */
1727
1728 next->text[ptr++] = '\n';
1729 message_size++;
1730
1731 /* A blank line signals the end of the headers; release the unwanted
1732 space and set next to NULL to indicate this. */
1733
1734 if (ptr == 1)
1735 {
1736 store_reset(next);
1737 next = NULL;
1738 break;
1739 }
1740
1741 /* There is data in the line; see if the next input character is a
1742 whitespace character. If it is, we have a continuation of this header line.
1743 There is always space for at least one character at this point. */
1744
1745 if (ch != EOF)
1746 {
80a47a2c 1747 int nextch = (receive_getc)();
059ec3d9
PH
1748 if (nextch == ' ' || nextch == '\t')
1749 {
1750 next->text[ptr++] = nextch;
1751 message_size++;
1752 continue; /* Iterate the loop */
1753 }
80a47a2c 1754 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
1755 else ch = EOF; /* Cause main loop to exit at end */
1756 }
1757
1758 /* We have got to the real line end. Terminate the string and release store
1759 beyond it. If it turns out to be a real header, internal binary zeros will
1760 be squashed later. */
1761
1762 next->text[ptr] = 0;
1763 next->slen = ptr;
1764 store_reset(next->text + ptr + 1);
1765
1766 /* Check the running total size against the overall message size limit. We
1767 don't expect to fail here, but if the overall limit is set less than MESSAGE_
1768 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
1769 headers - the code to read the body will then also hit the buffer. */
1770
1771 if (message_size > thismessage_size_limit) break;
1772
1773 /* A line that is not syntactically correct for a header also marks
1774 the end of the headers. In this case, we leave next containing the
1775 first data line. This might actually be several lines because of the
1776 continuation logic applied above, but that doesn't matter.
1777
1778 It turns out that smail, and presumably sendmail, accept leading lines
1779 of the form
1780
1781 From ph10 Fri Jan 5 12:35 GMT 1996
1782
1783 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
1784 find any documentation of this, but for compatibility it had better be
1785 accepted. Exim restricts it to the case of non-smtp messages, and
1786 treats it as an alternative to the -f command line option. Thus it is
1787 ignored except for trusted users or filter testing. Otherwise it is taken
1788 as the sender address, unless -f was used (sendmail compatibility).
1789
1790 It further turns out that some UUCPs generate the From_line in a different
1791 format, e.g.
1792
1793 From ph10 Fri, 7 Jan 97 14:00:00 GMT
1794
1795 The regex for matching these things is now capable of recognizing both
1796 formats (including 2- and 4-digit years in the latter). In fact, the regex
1797 is now configurable, as is the expansion string to fish out the sender.
1798
1799 Even further on it has been discovered that some broken clients send
1800 these lines in SMTP messages. There is now an option to ignore them from
1801 specified hosts or networks. Sigh. */
1802
1803 if (header_last == header_list &&
1804 (!smtp_input
1805 ||
1806 (sender_host_address != NULL &&
1807 verify_check_host(&ignore_fromline_hosts) == OK)
1808 ||
1809 (sender_host_address == NULL && ignore_fromline_local)
1810 ) &&
1811 regex_match_and_setup(regex_From, next->text, 0, -1))
1812 {
1813 if (!sender_address_forced)
1814 {
1815 uschar *uucp_sender = expand_string(uucp_from_sender);
1816 if (uucp_sender == NULL)
1817 {
1818 log_write(0, LOG_MAIN|LOG_PANIC,
1819 "expansion of \"%s\" failed after matching "
1820 "\"From \" line: %s", uucp_from_sender, expand_string_message);
1821 }
1822 else
1823 {
1824 int start, end, domain;
1825 uschar *errmess;
1826 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
1827 &start, &end, &domain, TRUE);
1828 if (newsender != NULL)
1829 {
1830 if (domain == 0 && newsender[0] != 0)
1831 newsender = rewrite_address_qualify(newsender, FALSE);
1832
f05da2e8 1833 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
1834 {
1835 sender_address = newsender;
1836
f05da2e8 1837 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
1838 {
1839 authenticated_sender = NULL;
1840 originator_name = US"";
1841 sender_local = FALSE;
1842 }
1843
f05da2e8 1844 if (filter_test != FTEST_NONE)
059ec3d9
PH
1845 printf("Sender taken from \"From \" line\n");
1846 }
1847 }
1848 }
1849 }
1850 }
1851
1852 /* Not a leading "From " line. Check to see if it is a valid header line.
1853 Header names may contain any non-control characters except space and colon,
1854 amazingly. */
1855
1856 else
1857 {
1858 uschar *p = next->text;
1859
1860 /* If not a valid header line, break from the header reading loop, leaving
1861 next != NULL, indicating that it holds the first line of the body. */
1862
1863 if (isspace(*p)) break;
1864 while (mac_isgraph(*p) && *p != ':') p++;
1865 while (isspace(*p)) p++;
1866 if (*p != ':')
1867 {
1868 body_zerocount = had_zero;
1869 break;
1870 }
1871
1872 /* We have a valid header line. If there were any binary zeroes in
1873 the line, stomp on them here. */
1874
1875 if (had_zero > 0)
1876 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
1877
1878 /* It is perfectly legal to have an empty continuation line
1879 at the end of a header, but it is confusing to humans
1880 looking at such messages, since it looks like a blank line.
1881 Reduce confusion by removing redundant white space at the
1882 end. We know that there is at least one printing character
1883 (the ':' tested for above) so there is no danger of running
1884 off the end. */
1885
1886 p = next->text + ptr - 2;
1887 for (;;)
1888 {
1889 while (*p == ' ' || *p == '\t') p--;
1890 if (*p != '\n') break;
1891 ptr = (p--) - next->text + 1;
1892 message_size -= next->slen - ptr;
1893 next->text[ptr] = 0;
1894 next->slen = ptr;
1895 }
1896
1897 /* Add the header to the chain */
1898
1899 next->type = htype_other;
1900 next->next = NULL;
1901 header_last->next = next;
1902 header_last = next;
1903
1904 /* Check the limit for individual line lengths. This comes after adding to
1905 the chain so that the failing line is reflected if a bounce is generated
1906 (for a local message). */
1907
1908 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
1909 {
1910 log_write(0, LOG_MAIN, "overlong message header line received from "
1911 "%s (more than %d characters): message abandoned",
1912 sender_host_unknown? sender_ident : sender_fullhost,
1913 header_line_maxsize);
1914
1915 if (smtp_input)
1916 {
1917 smtp_reply = US"552 A message header line is too long";
1918 receive_swallow_smtp();
1919 goto TIDYUP; /* Skip to end of function */
1920 }
1921
1922 else
1923 {
1924 give_local_error(ERRMESS_VLONGHDRLINE,
1925 string_sprintf("message header line longer than %d characters "
1926 "received: message not accepted", header_line_maxsize), US"",
1927 error_rc, stdin, header_list->next);
1928 /* Does not return */
1929 }
1930 }
1931
1932 /* Note if any resent- fields exist. */
1933
1934 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
1935 {
1936 resents_exist = TRUE;
1937 resent_prefix = US"Resent-";
1938 }
1939 }
1940
1941 /* The line has been handled. If we have hit EOF, break out of the loop,
1942 indicating no pending data line. */
1943
1944 if (ch == EOF) { next = NULL; break; }
1945
1946 /* Set up for the next header */
1947
1948 header_size = 256;
1949 next = store_get(sizeof(header_line));
1950 next->text = store_get(header_size);
1951 ptr = 0;
1952 had_zero = 0;
d677b2f2 1953 prevlines_length = 0;
059ec3d9
PH
1954 } /* Continue, starting to read the next header */
1955
1956/* At this point, we have read all the headers into a data structure in main
1957store. The first header is still the dummy placeholder for the Received: header
1958we are going to generate a bit later on. If next != NULL, it contains the first
1959data line - which terminated the headers before reaching a blank line (not the
1960normal case). */
1961
1962DEBUG(D_receive)
1963 {
1964 debug_printf(">>Headers received:\n");
1965 for (h = header_list->next; h != NULL; h = h->next)
1966 debug_printf("%s", h->text);
1967 debug_printf("\n");
1968 }
1969
1970/* End of file on any SMTP connection is an error. If an incoming SMTP call
1971is dropped immediately after valid headers, the next thing we will see is EOF.
1972We must test for this specially, as further down the reading of the data is
1973skipped if already at EOF. */
1974
1975if (smtp_input && (receive_feof)())
1976 {
1977 smtp_reply = handle_lost_connection(US" (after header)");
1978 smtp_yield = FALSE;
1979 goto TIDYUP; /* Skip to end of function */
1980 }
1981
1982/* If this is a filter test run and no headers were read, output a warning
1983in case there is a mistake in the test message. */
1984
f05da2e8 1985if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
1986 printf("Warning: no message headers read\n");
1987
1988
1989/* Scan the headers to identify them. Some are merely marked for later
1990processing; some are dealt with here. */
1991
1992for (h = header_list->next; h != NULL; h = h->next)
1993 {
1994 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
1995 if (is_resent) contains_resent_headers = TRUE;
1996
1997 switch (header_checkname(h, is_resent))
1998 {
059ec3d9 1999 case htype_bcc:
2cbb4081 2000 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
2001 break;
2002
059ec3d9 2003 case htype_cc:
2cbb4081 2004 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
2005 break;
2006
2007 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
2008
2009 case htype_date:
4c69d561 2010 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
2011 break;
2012
2013 /* Same comments as about Return-Path: below. */
2014
2015 case htype_delivery_date:
2016 if (delivery_date_remove) h->type = htype_old;
2017 break;
2018
2019 /* Same comments as about Return-Path: below. */
2020
2021 case htype_envelope_to:
2022 if (envelope_to_remove) h->type = htype_old;
2023 break;
2024
2025 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2026 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2027 header consists of just the login id of the user who called Exim, rewrite
2028 it with the gecos field first. Apply this rule to Resent-From: if there
2029 are resent- fields. */
2030
2031 case htype_from:
2032 h->type = htype_from;
2033 if (!resents_exist || is_resent)
2034 {
2035 from_header = h;
2036 if (!smtp_input)
2037 {
5de8faa3 2038 int len;
059ec3d9
PH
2039 uschar *s = Ustrchr(h->text, ':') + 1;
2040 while (isspace(*s)) s++;
5de8faa3 2041 len = h->slen - (s - h->text) - 1;
e0fccd1d
TF
2042 if (Ustrlen(originator_login) == len &&
2043 strncmpic(s, originator_login, len) == 0)
059ec3d9
PH
2044 {
2045 uschar *name = is_resent? US"Resent-From" : US"From";
2046 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2047 originator_login, qualify_domain_sender);
2048 from_header = header_last;
2049 h->type = htype_old;
2050 DEBUG(D_receive|D_rewrite)
2051 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2052 }
2053 }
2054 }
2055 break;
2056
2057 /* Identify the Message-id: header for generating "in-reply-to" in the
2058 autoreply transport. For incoming logging, save any resent- value. In both
2059 cases, take just the first of any multiples. */
2060
2061 case htype_id:
2062 if (msgid_header == NULL && (!resents_exist || is_resent))
2063 {
2064 msgid_header = h;
2065 h->type = htype_id;
2066 }
2067 break;
2068
2069 /* Flag all Received: headers */
2070
2071 case htype_received:
2072 h->type = htype_received;
2073 received_count++;
2074 break;
2075
2076 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
2077
2078 case htype_reply_to:
2079 h->type = htype_reply_to;
2080 break;
2081
2082 /* The Return-path: header is supposed to be added to messages when
2083 they leave the SMTP system. We shouldn't receive messages that already
2084 contain Return-path. However, since Exim generates Return-path: on
2085 local delivery, resent messages may well contain it. We therefore
2086 provide an option (which defaults on) to remove any Return-path: headers
2087 on input. Removal actually means flagging as "old", which prevents the
2088 header being transmitted with the message. */
2089
2090 case htype_return_path:
2091 if (return_path_remove) h->type = htype_old;
2092
2093 /* If we are testing a mail filter file, use the value of the
2094 Return-Path: header to set up the return_path variable, which is not
2095 otherwise set. However, remove any <> that surround the address
2096 because the variable doesn't have these. */
2097
f05da2e8 2098 if (filter_test != FTEST_NONE)
059ec3d9
PH
2099 {
2100 uschar *start = h->text + 12;
2101 uschar *end = start + Ustrlen(start);
2102 while (isspace(*start)) start++;
2103 while (end > start && isspace(end[-1])) end--;
2104 if (*start == '<' && end[-1] == '>')
2105 {
2106 start++;
2107 end--;
2108 }
2109 return_path = string_copyn(start, end - start);
2110 printf("Return-path taken from \"Return-path:\" header line\n");
2111 }
2112 break;
2113
2114 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2115 and from an untrusted caller and suppress_local_fixups is not set, or if we
2116 are in submission mode for a remote message, mark it "old" so that it will
2117 not be transmitted with the message, unless active_local_sender_retain is
2118 set. (This can only be true if active_local_from_check is false.) If there
2119 are any resent- headers in the message, apply this rule to Resent-Sender:
2120 instead of Sender:. Messages with multiple resent- header sets cannot be
2121 tidily handled. (For this reason, at least one MUA - Pine - turns old
2122 resent- headers into X-resent- headers when resending, leaving just one
2123 set.) */
059ec3d9
PH
2124
2125 case htype_sender:
69358f02 2126 h->type = ((!active_local_sender_retain &&
8800895a
PH
2127 (
2128 (sender_local && !trusted_caller && !suppress_local_fixups)
2129 || submission_mode
2130 )
059ec3d9
PH
2131 ) &&
2132 (!resents_exist||is_resent))?
2133 htype_old : htype_sender;
2134 break;
2135
2136 /* Remember the Subject: header for logging. There is no Resent-Subject */
2137
2138 case htype_subject:
2139 subject_header = h;
2140 break;
2141
2142 /* "To:" gets flagged, and the existence of a recipient header is noted,
2143 whether it's resent- or not. */
2144
2145 case htype_to:
2146 h->type = htype_to;
2147 /****
2148 to_or_cc_header_exists = TRUE;
2149 ****/
2150 break;
2151 }
2152 }
2153
2154/* Extract recipients from the headers if that is required (the -t option).
2155Note that this is documented as being done *before* any address rewriting takes
2156place. There are two possibilities:
2157
2158(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2159recipients already listed are to be REMOVED from the message. Smail 3 works
2160like this. We need to build a non-recipients tree for that list, because in
2161subsequent processing this data is held in a tree and that's what the
2162spool_write_header() function expects. Make sure that non-recipient addresses
2163are fully qualified and rewritten if necessary.
2164
2165(2) According to other sendmail documentation, -t ADDS extracted recipients to
2166those in the command line arguments (and it is rumoured some other MTAs do
2167this). Therefore, there is an option to make Exim behave this way.
2168
2169*** Notes on "Resent-" header lines ***
2170
2171The presence of resent-headers in the message makes -t horribly ambiguous.
2172Experiments with sendmail showed that it uses recipients for all resent-
2173headers, totally ignoring the concept of "sets of resent- headers" as described
2174in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2175with all the addresses in one instance of each header.
2176
2177This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2178error for -t if there were resent- headers in the message. However, after a
2179discussion on the mailing list, I've learned that there are MUAs that use
2180resent- headers with -t, and also that the stuff about sets of resent- headers
2181and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2182message with -t and resent- header lines makes sure that only *its* resent-
2183headers are present; previous ones are often renamed as X-resent- for example.
2184
2185Consequently, Exim has been changed so that, if any resent- header lines are
2186present, the recipients are taken from all of the appropriate resent- lines,
2187and not from the ordinary To:, Cc:, etc. */
2188
2189if (extract_recip)
2190 {
2191 int rcount = 0;
2192 error_block **bnext = &bad_addresses;
2193
2194 if (extract_addresses_remove_arguments)
2195 {
2196 while (recipients_count-- > 0)
2197 {
2198 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2199 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2200 tree_add_nonrecipient(s);
2201 }
2202 recipients_list = NULL;
2203 recipients_count = recipients_list_max = 0;
2204 }
2205
059ec3d9
PH
2206 /* Now scan the headers */
2207
2208 for (h = header_list->next; h != NULL; h = h->next)
2209 {
2210 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2211 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2212 {
2213 uschar *s = Ustrchr(h->text, ':') + 1;
2214 while (isspace(*s)) s++;
2215
1eccaa59
PH
2216 parse_allow_group = TRUE; /* Allow address group syntax */
2217
059ec3d9
PH
2218 while (*s != 0)
2219 {
2220 uschar *ss = parse_find_address_end(s, FALSE);
2221 uschar *recipient, *errmess, *p, *pp;
2222 int start, end, domain;
2223
2224 /* Check on maximum */
2225
2226 if (recipients_max > 0 && ++rcount > recipients_max)
2227 {
2228 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2229 US"message rejected: ", error_rc, stdin, NULL);
2230 /* Does not return */
2231 }
2232
2233 /* Make a copy of the address, and remove any internal newlines. These
2234 may be present as a result of continuations of the header line. The
2235 white space that follows the newline must not be removed - it is part
2236 of the header. */
2237
2238 pp = recipient = store_get(ss - s + 1);
2239 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2240 *pp = 0;
2241 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2242 &domain, FALSE);
2243
2244 /* Keep a list of all the bad addresses so we can send a single
2245 error message at the end. However, an empty address is not an error;
2246 just ignore it. This can come from an empty group list like
2247
2248 To: Recipients of list:;
2249
2250 If there are no recipients at all, an error will occur later. */
2251
2252 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2253 {
2254 int len = Ustrlen(s);
2255 error_block *b = store_get(sizeof(error_block));
2256 while (len > 0 && isspace(s[len-1])) len--;
2257 b->next = NULL;
2258 b->text1 = string_printing(string_copyn(s, len));
2259 b->text2 = errmess;
2260 *bnext = b;
2261 bnext = &(b->next);
2262 }
2263
2264 /* If the recipient is already in the nonrecipients tree, it must
2265 have appeared on the command line with the option extract_addresses_
2266 remove_arguments set. Do not add it to the recipients, and keep a note
2267 that this has happened, in order to give a better error if there are
2268 no recipients left. */
2269
2270 else if (recipient != NULL)
2271 {
2272 if (tree_search(tree_nonrecipients, recipient) == NULL)
2273 receive_add_recipient(recipient, -1);
2274 else
2275 extracted_ignored = TRUE;
2276 }
2277
2278 /* Move on past this address */
2279
2280 s = ss + (*ss? 1:0);
2281 while (isspace(*s)) s++;
1eccaa59
PH
2282 } /* Next address */
2283
2284 parse_allow_group = FALSE; /* Reset group syntax flags */
2285 parse_found_group = FALSE;
059ec3d9
PH
2286
2287 /* If this was the bcc: header, mark it "old", which means it
2288 will be kept on the spool, but not transmitted as part of the
2289 message. */
2290
2cbb4081 2291 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2292 } /* For appropriate header line */
2293 } /* For each header line */
2294
059ec3d9
PH
2295 }
2296
2297/* Now build the unique message id. This has changed several times over the
2298lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2299Retaining all the history in the comment has become too unwieldy - read
2300previous release sources if you want it.
2301
2302The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2303The first part is the current time, in seconds. The second part is the current
2304pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2305can hold a number in the range 0-3843. It used to be a computed sequence
2306number, but is now the fractional component of the current time in units of
23071/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2308received, Exim ensures that the timer has ticked at the appropriate level
2309before proceeding, to avoid duplication if the pid happened to be re-used
2310within the same time period. It seems likely that most messages will take at
2311least half a millisecond to be received, so no delay will normally be
2312necessary. At least for some time...
2313
2314There is a modification when localhost_number is set. Formerly this was allowed
2315to be as large as 255. Now it is restricted to the range 0-16, and the final
2316component of the message id becomes (localhost_number * 200) + fractional time
2317in units of 1/200 of a second (i.e. a value in the range 0-3399).
2318
2319Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2320Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2321still allows the tttttt field to hold a large enough number to last for some
2322more decades, and the final two-digit field can hold numbers up to 1295, which
2323is enough for milliseconds (instead of 1/2000 of a second).
2324
2325However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2326pid, so it is probably safe because pids have to be positive. The
2327localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2328final field becomes (localhost_number * 100) + fractional time in centiseconds.
2329
2330Note that string_base62() returns its data in a static storage block, so it
2331must be copied before calling string_base62() again. It always returns exactly
23326 characters.
2333
2334There doesn't seem to be anything in the RFC which requires a message id to
2335start with a letter, but Smail was changed to ensure this. The external form of
2336the message id (as supplied by string expansion) therefore starts with an
2337additional leading 'E'. The spool file names do not include this leading
2338letter and it is not used internally.
2339
2340NOTE: If ever the format of message ids is changed, the regular expression for
2341checking that a string is in this format must be updated in a corresponding
2342way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2343must also be changed to reflect the correct string length. Then, of course,
2344other programs that rely on the message id format will need updating too. */
2345
2346Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2347message_id[6] = '-';
2348Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2349
2350/* Deal with the case where the host number is set. The value of the number was
2351checked when it was read, to ensure it isn't too big. The timing granularity is
2352left in id_resolution so that an appropriate wait can be done after receiving
2353the message, if necessary (we hope it won't be). */
2354
2355if (host_number_string != NULL)
2356 {
2357 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2358 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2359 string_base62((long int)(
2360 host_number * (1000000/id_resolution) +
2361 message_id_tv.tv_usec/id_resolution)) + 4);
2362 }
2363
2364/* Host number not set: final field is just the fractional time at an
2365appropriate resolution. */
2366
2367else
2368 {
2369 id_resolution = (BASE_62 == 62)? 500 : 1000;
2370 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2371 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2372 }
2373
2374/* Add the current message id onto the current process info string if
2375it will fit. */
2376
2377(void)string_format(process_info + process_info_len,
2378 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2379
2380/* If we are using multiple input directories, set up the one for this message
2381to be the least significant base-62 digit of the time of arrival. Otherwise
2382ensure that it is an empty string. */
2383
2384message_subdir[0] = split_spool_directory? message_id[5] : 0;
2385
2386/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2387one, but only for local (without suppress_local_fixups) or submission mode
2388messages. This can be user-configured if required, but we had better flatten
2389any illegal characters therein. */
059ec3d9 2390
8800895a
PH
2391if (msgid_header == NULL &&
2392 ((sender_host_address == NULL && !suppress_local_fixups)
2393 || submission_mode))
059ec3d9
PH
2394 {
2395 uschar *p;
2396 uschar *id_text = US"";
2397 uschar *id_domain = primary_hostname;
2398
2399 /* Permit only letters, digits, dots, and hyphens in the domain */
2400
2401 if (message_id_domain != NULL)
2402 {
2403 uschar *new_id_domain = expand_string(message_id_domain);
2404 if (new_id_domain == NULL)
2405 {
2406 if (!expand_string_forcedfail)
2407 log_write(0, LOG_MAIN|LOG_PANIC,
2408 "expansion of \"%s\" (message_id_header_domain) "
2409 "failed: %s", message_id_domain, expand_string_message);
2410 }
2411 else if (*new_id_domain != 0)
2412 {
2413 id_domain = new_id_domain;
2414 for (p = id_domain; *p != 0; p++)
2415 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2416 }
2417 }
2418
2419 /* Permit all characters except controls and RFC 2822 specials in the
2420 additional text part. */
2421
2422 if (message_id_text != NULL)
2423 {
2424 uschar *new_id_text = expand_string(message_id_text);
2425 if (new_id_text == NULL)
2426 {
2427 if (!expand_string_forcedfail)
2428 log_write(0, LOG_MAIN|LOG_PANIC,
2429 "expansion of \"%s\" (message_id_header_text) "
2430 "failed: %s", message_id_text, expand_string_message);
2431 }
2432 else if (*new_id_text != 0)
2433 {
2434 id_text = new_id_text;
2435 for (p = id_text; *p != 0; p++)
2436 if (mac_iscntrl_or_special(*p)) *p = '-';
2437 }
2438 }
2439
e7e680d6
PP
2440 /* Add the header line
2441 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2442 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2443
e7e680d6 2444 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2445 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2446 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2447 }
2448
2449/* If we are to log recipients, keep a copy of the raw ones before any possible
2450rewriting. Must copy the count, because later ACLs and the local_scan()
2451function may mess with the real recipients. */
2452
2453if ((log_extra_selector & LX_received_recipients) != 0)
2454 {
2455 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2456 for (i = 0; i < recipients_count; i++)
2457 raw_recipients[i] = string_copy(recipients_list[i].address);
2458 raw_recipients_count = recipients_count;
2459 }
2460
2461/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2462recipients will get here only if the conditions were right (allow_unqualified_
2463recipient is TRUE). */
2464
2465for (i = 0; i < recipients_count; i++)
2466 recipients_list[i].address =
2467 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2468 global_rewrite_rules, rewrite_existflags);
2469
8800895a
PH
2470/* If there is no From: header, generate one for local (without
2471suppress_local_fixups) or submission_mode messages. If there is no sender
2472address, but the sender is local or this is a local delivery error, use the
2473originator login. This shouldn't happen for genuine bounces, but might happen
2474for autoreplies. The addition of From: must be done *before* checking for the
2475possible addition of a Sender: header, because untrusted_set_sender allows an
2476untrusted user to set anything in the envelope (which might then get info
2477From:) but we still want to ensure a valid Sender: if it is required. */
2478
2479if (from_header == NULL &&
2480 ((sender_host_address == NULL && !suppress_local_fixups)
2481 || submission_mode))
059ec3d9 2482 {
2fe1a124
PH
2483 uschar *oname = US"";
2484
2485 /* Use the originator_name if this is a locally submitted message and the
2486 caller is not trusted. For trusted callers, use it only if -F was used to
2487 force its value or if we have a non-SMTP message for which -f was not used
2488 to set the sender. */
2489
2490 if (sender_host_address == NULL)
2491 {
2492 if (!trusted_caller || sender_name_forced ||
2493 (!smtp_input && !sender_address_forced))
2494 oname = originator_name;
2495 }
2496
2497 /* For non-locally submitted messages, the only time we use the originator
2498 name is when it was forced by the /name= option on control=submission. */
2499
2500 else
2501 {
2502 if (submission_name != NULL) oname = submission_name;
2503 }
2504
059ec3d9
PH
2505 /* Envelope sender is empty */
2506
2507 if (sender_address[0] == 0)
2508 {
87ba3f5f
PH
2509 uschar *fromstart, *fromend;
2510
2511 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2512 oname, (oname[0] == 0)? "" : " <");
2513 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2514
059ec3d9
PH
2515 if (sender_local || local_error_message)
2516 {
87ba3f5f
PH
2517 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2518 local_part_quote(originator_login), qualify_domain_sender,
2519 fromend);
059ec3d9
PH
2520 }
2521 else if (submission_mode && authenticated_id != NULL)
2522 {
2523 if (submission_domain == NULL)
2524 {
87ba3f5f
PH
2525 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2526 local_part_quote(authenticated_id), qualify_domain_sender,
2527 fromend);
059ec3d9
PH
2528 }
2529 else if (submission_domain[0] == 0) /* empty => whole address set */
2530 {
87ba3f5f
PH
2531 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2532 fromend);
059ec3d9
PH
2533 }
2534 else
2535 {
87ba3f5f
PH
2536 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2537 local_part_quote(authenticated_id), submission_domain,
2538 fromend);
059ec3d9
PH
2539 }
2540 from_header = header_last; /* To get it checked for Sender: */
2541 }
2542 }
2543
2544 /* There is a non-null envelope sender. Build the header using the original
2545 sender address, before any rewriting that might have been done while
2546 verifying it. */
2547
2548 else
2549 {
87ba3f5f 2550 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2551 oname,
2552 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2553 (sender_address_unrewritten == NULL)?
2554 sender_address : sender_address_unrewritten,
2fe1a124 2555 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2556
2557 from_header = header_last; /* To get it checked for Sender: */
2558 }
2559 }
2560
2561
8800895a
PH
2562/* If the sender is local (without suppress_local_fixups), or if we are in
2563submission mode and there is an authenticated_id, check that an existing From:
2564is correct, and if not, generate a Sender: header, unless disabled. Any
2565previously-existing Sender: header was removed above. Note that sender_local,
2566as well as being TRUE if the caller of exim is not trusted, is also true if a
2567trusted caller did not supply a -f argument for non-smtp input. To allow
2568trusted callers to forge From: without supplying -f, we have to test explicitly
2569here. If the From: header contains more than one address, then the call to
2570parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2571
2572if (from_header != NULL &&
69358f02 2573 (active_local_from_check &&
8800895a 2574 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2575 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2576 ))
2577 {
2578 BOOL make_sender = TRUE;
2579 int start, end, domain;
2580 uschar *errmess;
2581 uschar *from_address =
2582 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2583 &start, &end, &domain, FALSE);
2584 uschar *generated_sender_address;
2585
2586 if (submission_mode)
2587 {
2588 if (submission_domain == NULL)
2589 {
2590 generated_sender_address = string_sprintf("%s@%s",
2591 local_part_quote(authenticated_id), qualify_domain_sender);
2592 }
2593 else if (submission_domain[0] == 0) /* empty => full address */
2594 {
2595 generated_sender_address = string_sprintf("%s",
2596 authenticated_id);
2597 }
2598 else
2599 {
2600 generated_sender_address = string_sprintf("%s@%s",
2601 local_part_quote(authenticated_id), submission_domain);
2602 }
2603 }
2604 else
2605 generated_sender_address = string_sprintf("%s@%s",
2606 local_part_quote(originator_login), qualify_domain_sender);
2607
2608 /* Remove permitted prefixes and suffixes from the local part of the From:
2609 address before doing the comparison with the generated sender. */
2610
2611 if (from_address != NULL)
2612 {
2613 int slen;
2614 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2615
2616 if (at != NULL) *at = 0;
2617 from_address += route_check_prefix(from_address, local_from_prefix);
2618 slen = route_check_suffix(from_address, local_from_suffix);
2619 if (slen > 0)
2620 {
2621 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2622 from_address += slen;
2623 }
2624 if (at != NULL) *at = '@';
2625
2626 if (strcmpic(generated_sender_address, from_address) == 0 ||
2627 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2628 make_sender = FALSE;
2629 }
2630
2631 /* We have to cause the Sender header to be rewritten if there are
2632 appropriate rewriting rules. */
2633
2634 if (make_sender)
2635 {
2fe1a124 2636 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2637 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2638 generated_sender_address);
2639 else
2640 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2641 resent_prefix,
2642 submission_mode? submission_name : originator_name,
2643 generated_sender_address);
059ec3d9 2644 }
87ba3f5f
PH
2645
2646 /* Ensure that a non-null envelope sender address corresponds to the
2647 submission mode sender address. */
2648
2649 if (submission_mode && sender_address[0] != 0)
2650 {
2651 if (sender_address_unrewritten == NULL)
2652 sender_address_unrewritten = sender_address;
2653 sender_address = generated_sender_address;
089793a4
TF
2654 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2655 log_write(L_address_rewrite, LOG_MAIN,
2656 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2657 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2658 }
059ec3d9
PH
2659 }
2660
2661
2662/* If there are any rewriting rules, apply them to the sender address, unless
2663it has already been rewritten as part of verification for SMTP input. */
2664
2665if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2666 sender_address[0] != 0)
2667 {
2668 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2669 global_rewrite_rules, rewrite_existflags);
2670 DEBUG(D_receive|D_rewrite)
2671 debug_printf("rewritten sender = %s\n", sender_address);
2672 }
2673
2674
2675/* The headers must be run through rewrite_header(), because it ensures that
2676addresses are fully qualified, as well as applying any rewriting rules that may
2677exist.
2678
2679Qualification of header addresses in a message from a remote host happens only
2680if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2681appropriate. For local messages, qualification always happens, unless -bnq is
2682used to explicitly suppress it. No rewriting is done for an unqualified address
2683that is left untouched.
2684
2685We start at the second header, skipping our own Received:. This rewriting is
2686documented as happening *after* recipient addresses are taken from the headers
2687by the -t command line option. An added Sender: gets rewritten here. */
2688
2689for (h = header_list->next; h != NULL; h = h->next)
2690 {
2691 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2692 rewrite_existflags, TRUE);
2693 if (newh != NULL) h = newh;
2694 }
2695
2696
2697/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2698"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2699"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2700exists, this applies to the set of resent- headers rather than the normal set.
2701
2cbb4081
PH
2702The requirement for a recipient header has been removed in RFC 2822. At this
2703point in the code, earlier versions of Exim added a To: header for locally
2704submitted messages, and an empty Bcc: header for others. In the light of the
2705changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2706
059ec3d9
PH
2707
2708/* If there is no date header, generate one if the message originates locally
8800895a
PH
2709(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2710submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2711to be more confusing if Exim adds one to all remotely-originated messages.
2712As per Message-Id, we prepend if resending, else append.
2713*/
059ec3d9 2714
8800895a
PH
2715if (!date_header_exists &&
2716 ((sender_host_address == NULL && !suppress_local_fixups)
2717 || submission_mode))
e7e680d6
PP
2718 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2719 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2720
2721search_tidyup(); /* Free any cached resources */
2722
2723/* Show the complete set of headers if debugging. Note that the first one (the
2724new Received:) has not yet been set. */
2725
2726DEBUG(D_receive)
2727 {
2728 debug_printf(">>Headers after rewriting and local additions:\n");
2729 for (h = header_list->next; h != NULL; h = h->next)
2730 debug_printf("%c %s", h->type, h->text);
2731 debug_printf("\n");
2732 }
2733
2734/* The headers are now complete in store. If we are running in filter
2735testing mode, that is all this function does. Return TRUE if the message
2736ended with a dot. */
2737
f05da2e8 2738if (filter_test != FTEST_NONE)
059ec3d9
PH
2739 {
2740 process_info[process_info_len] = 0;
2741 return message_ended == END_DOT;
2742 }
2743
817d9f57 2744/* Cutthrough delivery:
e4bdf652
JH
2745 We have to create the Received header now rather than at the end of reception,
2746 so the timestamp behaviour is a change to the normal case.
2747 XXX Ensure this gets documented XXX.
817d9f57 2748 Having created it, send the headers to the destination.
e4bdf652
JH
2749*/
2750if (cutthrough_fd >= 0)
2751 {
817d9f57
JH
2752 if (received_count > received_headers_max)
2753 {
2e5b33cd 2754 cancel_cutthrough_connection("too many headers");
817d9f57
JH
2755 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2756 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2757 "Too many \"Received\" headers",
2758 sender_address,
2759 (sender_fullhost == NULL)? "" : " H=",
2760 (sender_fullhost == NULL)? US"" : sender_fullhost,
2761 (sender_ident == NULL)? "" : " U=",
2762 (sender_ident == NULL)? US"" : sender_ident);
2763 message_id[0] = 0; /* Indicate no message accepted */
2764 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
2765 goto TIDYUP; /* Skip to end of function */
2766 }
e4bdf652
JH
2767 received_header_gen();
2768 add_acl_headers(US"MAIL or RCPT");
2769 (void) cutthrough_headers_send();
2770 }
817d9f57 2771
e4bdf652 2772
059ec3d9
PH
2773/* Open a new spool file for the data portion of the message. We need
2774to access it both via a file descriptor and a stream. Try to make the
2775directory if it isn't there. Note re use of sprintf: spool_directory
2776is checked on input to be < 200 characters long. */
2777
2778sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory, message_subdir,
2779 message_id);
2780data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2781if (data_fd < 0)
2782 {
2783 if (errno == ENOENT)
2784 {
2785 uschar temp[16];
2786 sprintf(CS temp, "input/%s", message_subdir);
2787 if (message_subdir[0] == 0) temp[5] = 0;
2788 (void)directory_make(spool_directory, temp, INPUT_DIRECTORY_MODE, TRUE);
2789 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2790 }
2791 if (data_fd < 0)
2792 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
2793 spool_name, strerror(errno));
2794 }
2795
2796/* Make sure the file's group is the Exim gid, and double-check the mode
2797because the group setting doesn't always get set automatically. */
2798
1ac6b2e7
JH
2799if (fchown(data_fd, exim_uid, exim_gid))
2800 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
2801 "Failed setting ownership on spool file %s: %s",
2802 spool_name, strerror(errno));
ff790e47 2803(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
2804
2805/* We now have data file open. Build a stream for it and lock it. We lock only
2806the first line of the file (containing the message ID) because otherwise there
2807are problems when Exim is run under Cygwin (I'm told). See comments in
2808spool_in.c, where the same locking is done. */
2809
2810data_file = fdopen(data_fd, "w+");
2811lock_data.l_type = F_WRLCK;
2812lock_data.l_whence = SEEK_SET;
2813lock_data.l_start = 0;
2814lock_data.l_len = SPOOL_DATA_START_OFFSET;
2815
2816if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
2817 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
2818 errno, strerror(errno));
2819
2820/* We have an open, locked data file. Write the message id to it to make it
2821self-identifying. Then read the remainder of the input of this message and
2822write it to the data file. If the variable next != NULL, it contains the first
2823data line (which was read as a header but then turned out not to have the right
2824format); write it (remembering that it might contain binary zeros). The result
2825of fwrite() isn't inspected; instead we call ferror() below. */
2826
2827fprintf(data_file, "%s-D\n", message_id);
2828if (next != NULL)
2829 {
2830 uschar *s = next->text;
2831 int len = next->slen;
1ac6b2e7 2832 len = fwrite(s, 1, len, data_file); len = len; /* compiler quietening */
059ec3d9
PH
2833 body_linecount++; /* Assumes only 1 line */
2834 }
2835
2836/* Note that we might already be at end of file, or the logical end of file
2837(indicated by '.'), or might have encountered an error while writing the
2838message id or "next" line. */
2839
2840if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
2841 {
2842 if (smtp_input)
2843 {
2844 message_ended = read_message_data_smtp(data_file);
2845 receive_linecount++; /* The terminating "." line */
2846 }
2847 else message_ended = read_message_data(data_file);
2848
2849 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 2850 message_linecount += body_linecount;
059ec3d9
PH
2851
2852 /* Handle premature termination of SMTP */
2853
2854 if (smtp_input && message_ended == END_EOF)
2855 {
2856 Uunlink(spool_name); /* Lose data file when closed */
2e5b33cd 2857 cancel_cutthrough_connection("sender closed connection");
059ec3d9
PH
2858 message_id[0] = 0; /* Indicate no message accepted */
2859 smtp_reply = handle_lost_connection(US"");
2860 smtp_yield = FALSE;
2861 goto TIDYUP; /* Skip to end of function */
2862 }
2863
2864 /* Handle message that is too big. Don't use host_or_ident() in the log
2865 message; we want to see the ident value even for non-remote messages. */
2866
2867 if (message_ended == END_SIZE)
2868 {
2869 Uunlink(spool_name); /* Lose the data file when closed */
2e5b33cd 2870 cancel_cutthrough_connection("mail too big");
059ec3d9
PH
2871 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2872
2873 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2874 "message too big: read=%d max=%d",
2875 sender_address,
2876 (sender_fullhost == NULL)? "" : " H=",
2877 (sender_fullhost == NULL)? US"" : sender_fullhost,
2878 (sender_ident == NULL)? "" : " U=",
2879 (sender_ident == NULL)? US"" : sender_ident,
2880 message_size,
2881 thismessage_size_limit);
2882
2883 if (smtp_input)
2884 {
2885 smtp_reply = US"552 Message size exceeds maximum permitted";
2886 message_id[0] = 0; /* Indicate no message accepted */
2887 goto TIDYUP; /* Skip to end of function */
2888 }
2889 else
2890 {
2891 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2892 give_local_error(ERRMESS_TOOBIG,
2893 string_sprintf("message too big (max=%d)", thismessage_size_limit),
2894 US"message rejected: ", error_rc, data_file, header_list);
2895 /* Does not return */
2896 }
2897 }
2898 }
2899
2900/* Restore the standard SIGALRM handler for any subsequent processing. (For
2901example, there may be some expansion in an ACL that uses a timer.) */
2902
2903os_non_restarting_signal(SIGALRM, sigalrm_handler);
2904
2905/* The message body has now been read into the data file. Call fflush() to
2906empty the buffers in C, and then call fsync() to get the data written out onto
2907the disk, as fflush() doesn't do this (or at least, it isn't documented as
2908having to do this). If there was an I/O error on either input or output,
2909attempt to send an error message, and unlink the spool file. For non-SMTP input
2910we can then give up. Note that for SMTP input we must swallow the remainder of
2911the input in cases of output errors, since the far end doesn't expect to see
2912anything until the terminating dot line is sent. */
2913
2914if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 2915 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
2916 {
2917 uschar *msg_errno = US strerror(errno);
2918 BOOL input_error = (receive_ferror)() != 0;
2919 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
2920 input_error? "Input read" : "Spool write",
2921 msg_errno,
2922 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
2923
2924 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
2925 Uunlink(spool_name); /* Lose the data file */
2e5b33cd 2926 cancel_cutthrough_connection("error writing spoolfile");
059ec3d9
PH
2927
2928 if (smtp_input)
2929 {
2930 if (input_error)
2931 smtp_reply = US"451 Error while reading input data";
2932 else
2933 {
2934 smtp_reply = US"451 Error while writing spool file";
2935 receive_swallow_smtp();
2936 }
2937 message_id[0] = 0; /* Indicate no message accepted */
2938 goto TIDYUP; /* Skip to end of function */
2939 }
2940
2941 else
2942 {
2943 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2944 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
2945 header_list);
2946 /* Does not return */
2947 }
2948 }
2949
2950
2951/* No I/O errors were encountered while writing the data file. */
2952
2953DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
2954
2955
2956/* If there were any bad addresses extracted by -t, or there were no recipients
2957left after -t, send a message to the sender of this message, or write it to
2958stderr if the error handling option is set that way. Note that there may
2959legitimately be no recipients for an SMTP message if they have all been removed
2960by "discard".
2961
2962We need to rewind the data file in order to read it. In the case of no
2963recipients or stderr error writing, throw the data file away afterwards, and
2964exit. (This can't be SMTP, which always ensures there's at least one
2965syntactically good recipient address.) */
2966
2967if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
2968 {
2969 DEBUG(D_receive)
2970 {
2971 if (recipients_count == 0) debug_printf("*** No recipients\n");
2972 if (bad_addresses != NULL)
2973 {
2974 error_block *eblock = bad_addresses;
2975 debug_printf("*** Bad address(es)\n");
2976 while (eblock != NULL)
2977 {
2978 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
2979 eblock = eblock->next;
2980 }
2981 }
2982 }
2983
2984 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2985
2986 /* If configured to send errors to the sender, but this fails, force
2987 a failure error code. We use a special one for no recipients so that it
2988 can be detected by the autoreply transport. Otherwise error_rc is set to
2989 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
2990 it is EXIT_SUCCESS. */
2991
2992 if (error_handling == ERRORS_SENDER)
2993 {
2994 if (!moan_to_sender(
2995 (bad_addresses == NULL)?
2996 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
2997 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
2998 bad_addresses, header_list, data_file, FALSE))
2999 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
3000 }
3001 else
3002 {
3003 if (bad_addresses == NULL)
3004 {
3005 if (extracted_ignored)
3006 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3007 else
3008 fprintf(stderr, "exim: no recipients in message\n");
3009 }
3010 else
3011 {
3012 fprintf(stderr, "exim: invalid address%s",
3013 (bad_addresses->next == NULL)? ":" : "es:\n");
3014 while (bad_addresses != NULL)
3015 {
3016 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3017 bad_addresses->text2);
3018 bad_addresses = bad_addresses->next;
3019 }
3020 }
3021 }
3022
3023 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3024 {
3025 Uunlink(spool_name);
f1e894f3 3026 (void)fclose(data_file);
059ec3d9
PH
3027 exim_exit(error_rc);
3028 }
3029 }
3030
3031/* Data file successfully written. Generate text for the Received: header by
3032expanding the configured string, and adding a timestamp. By leaving this
3033operation till now, we ensure that the timestamp is the time that message
3034reception was completed. However, this is deliberately done before calling the
3035data ACL and local_scan().
3036
3037This Received: header may therefore be inspected by the data ACL and by code in
3038the local_scan() function. When they have run, we update the timestamp to be
3039the final time of reception.
3040
3041If there is just one recipient, set up its value in the $received_for variable
3042for use when we generate the Received: header.
3043
3044Note: the checking for too many Received: headers is handled by the delivery
3045code. */
e4bdf652 3046/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3047
e4bdf652 3048if (received_header->text == NULL) /* Non-cutthrough case */
059ec3d9 3049 {
e4bdf652 3050 received_header_gen();
059ec3d9 3051
e4bdf652 3052 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3053
e4bdf652
JH
3054 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3055 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3056
e4bdf652
JH
3057 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3058 now, before running the DATA ACL. */
059ec3d9 3059
e4bdf652
JH
3060 add_acl_headers(US"MAIL or RCPT");
3061 }
817d9f57 3062else
e4bdf652
JH
3063 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3064 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3065
3066/* If an ACL is specified for checking things at this stage of reception of a
3067message, run it, unless all the recipients were removed by "discard" in earlier
3068ACLs. That is the only case in which recipients_count can be zero at this
3069stage. Set deliver_datafile to point to the data file so that $message_body and
3070$message_body_end can be extracted if needed. Allow $recipients in expansions.
3071*/
3072
3073deliver_datafile = data_fd;
4e88a19f 3074user_msg = NULL;
059ec3d9 3075
0e20aff9
MH
3076enable_dollar_recipients = TRUE;
3077
059ec3d9
PH
3078if (recipients_count == 0)
3079 {
3080 blackholed_by = recipients_discarded? US"MAIL ACL" : US"RCPT ACL";
3081 }
3082else
3083 {
059ec3d9
PH
3084 /* Handle interactive SMTP messages */
3085
3086 if (smtp_input && !smtp_batched_input)
3087 {
8523533c 3088
80a47a2c
TK
3089#ifndef DISABLE_DKIM
3090 if (!dkim_disable_verify)
3091 {
3092 /* Finish verification, this will log individual signature results to
3093 the mainlog */
3094 dkim_exim_verify_finish();
3095
3096 /* Check if we must run the DKIM ACL */
3097 if ((acl_smtp_dkim != NULL) &&
3098 (dkim_verify_signers != NULL) &&
3099 (dkim_verify_signers[0] != '\0'))
3100 {
3101 uschar *dkim_verify_signers_expanded =
3102 expand_string(dkim_verify_signers);
3103 if (dkim_verify_signers_expanded == NULL)
3104 {
3105 log_write(0, LOG_MAIN|LOG_PANIC,
3106 "expansion of dkim_verify_signers option failed: %s",
3107 expand_string_message);
3108 }
3109 else
3110 {
3111 int sep = 0;
3112 uschar *ptr = dkim_verify_signers_expanded;
3113 uschar *item = NULL;
9e5d6b55
TK
3114 uschar *seen_items = NULL;
3115 int seen_items_size = 0;
3116 int seen_items_offset = 0;
80a47a2c 3117 uschar itembuf[256];
9122af94
TK
3118 /* Default to OK when no items are present */
3119 rc = OK;
80a47a2c
TK
3120 while ((item = string_nextinlist(&ptr, &sep,
3121 itembuf,
3122 sizeof(itembuf))) != NULL)
3123 {
6119d1ea
TK
3124 /* Prevent running ACL for an empty item */
3125 if (!item || (item[0] == '\0')) continue;
9e5d6b55
TK
3126 /* Only run ACL once for each domain or identity, no matter how often it
3127 appears in the expanded list. */
6119d1ea
TK
3128 if (seen_items != NULL)
3129 {
ae9094bf
TK
3130 uschar *seen_item = NULL;
3131 uschar seen_item_buf[256];
4a73449b 3132 uschar *seen_items_list = seen_items;
ae9094bf
TK
3133 int seen_this_item = 0;
3134
3135 while ((seen_item = string_nextinlist(&seen_items_list, &sep,
3136 seen_item_buf,
3137 sizeof(seen_item_buf))) != NULL)
3138 {
3139 if (Ustrcmp(seen_item,item) == 0)
3140 {
3141 seen_this_item = 1;
3142 break;
3143 }
3144 }
3145
3146 if (seen_this_item > 0)
6119d1ea
TK
3147 {
3148 DEBUG(D_receive)
3149 debug_printf("acl_smtp_dkim: skipping signer %s, already seen\n", item);
3150 continue;
3151 }
ae9094bf 3152
4a73449b 3153 seen_items = string_append(seen_items,&seen_items_size,&seen_items_offset,1,":");
6119d1ea
TK
3154 }
3155
4a73449b
TK
3156 seen_items = string_append(seen_items,&seen_items_size,&seen_items_offset,1,item);
3157 seen_items[seen_items_offset] = '\0';
6119d1ea
TK
3158
3159 DEBUG(D_receive)
3160 debug_printf("calling acl_smtp_dkim for dkim_cur_signer=%s\n", item);
3161
80a47a2c
TK
3162 dkim_exim_acl_setup(item);
3163 rc = acl_check(ACL_WHERE_DKIM, NULL, acl_smtp_dkim, &user_msg, &log_msg);
6119d1ea
TK
3164
3165 if (rc != OK)
3166 {
3167 DEBUG(D_receive)
3168 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, skipping remaining items\n", rc, item);
2e5b33cd 3169 cancel_cutthrough_connection("dkim acl not ok");
6119d1ea
TK
3170 break;
3171 }
80a47a2c
TK
3172 }
3173 add_acl_headers(US"DKIM");
3174 if (rc == DISCARD)
3175 {
3176 recipients_count = 0;
3177 blackholed_by = US"DKIM ACL";
3178 if (log_msg != NULL)
3179 blackhole_log_msg = string_sprintf(": %s", log_msg);
3180 }
3181 else if (rc != OK)
3182 {
3183 Uunlink(spool_name);
3184 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
3185 smtp_yield = FALSE; /* No more messsages after dropped connection */
3186 smtp_reply = US""; /* Indicate reply already sent */
3187 message_id[0] = 0; /* Indicate no message accepted */
3188 goto TIDYUP; /* Skip to end of function */
3189 }
3190 }
3191 }
3192 }
3193#endif /* DISABLE_DKIM */
fb2274d4 3194
8523533c 3195#ifdef WITH_CONTENT_SCAN
80a47a2c
TK
3196 if (recipients_count > 0 &&
3197 acl_smtp_mime != NULL &&
54cdb463
PH
3198 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
3199 goto TIDYUP;
8523533c
TK
3200#endif /* WITH_CONTENT_SCAN */
3201
54cdb463
PH
3202 /* Check the recipients count again, as the MIME ACL might have changed
3203 them. */
8523533c 3204
059ec3d9
PH
3205 if (acl_smtp_data != NULL && recipients_count > 0)
3206 {
059ec3d9
PH
3207 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
3208 add_acl_headers(US"DATA");
3209 if (rc == DISCARD)
3210 {
3211 recipients_count = 0;
3212 blackholed_by = US"DATA ACL";
8e669ac1
PH
3213 if (log_msg != NULL)
3214 blackhole_log_msg = string_sprintf(": %s", log_msg);
2e5b33cd 3215 cancel_cutthrough_connection("data acl discard");
059ec3d9
PH
3216 }
3217 else if (rc != OK)
3218 {
3219 Uunlink(spool_name);
2e5b33cd 3220 cancel_cutthrough_connection("data acl not ok");
8523533c
TK
3221#ifdef WITH_CONTENT_SCAN
3222 unspool_mbox();
3223#endif
6f0c431a
PP
3224#ifdef EXPERIMENTAL_DCC
3225 dcc_ok = 0;
3226#endif
059ec3d9
PH
3227 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
3228 smtp_yield = FALSE; /* No more messsages after dropped connection */
3229 smtp_reply = US""; /* Indicate reply already sent */
3230 message_id[0] = 0; /* Indicate no message accepted */
3231 goto TIDYUP; /* Skip to end of function */
3232 }
3233 }
3234 }
3235
3236 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
3237 we cannot take different actions for permanent and temporary rejections. */
3238
54cdb463 3239 else
059ec3d9 3240 {
54cdb463
PH
3241
3242#ifdef WITH_CONTENT_SCAN
3243 if (acl_not_smtp_mime != NULL &&
3244 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
3245 &blackholed_by))
3246 goto TIDYUP;
3247#endif /* WITH_CONTENT_SCAN */
3248
3249 if (acl_not_smtp != NULL)
059ec3d9 3250 {
54cdb463
PH
3251 uschar *user_msg, *log_msg;
3252 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
3253 if (rc == DISCARD)
059ec3d9 3254 {
54cdb463
PH
3255 recipients_count = 0;
3256 blackholed_by = US"non-SMTP ACL";
3257 if (log_msg != NULL)
3258 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 3259 }
54cdb463 3260 else if (rc != OK)
059ec3d9 3261 {
54cdb463
PH
3262 Uunlink(spool_name);
3263#ifdef WITH_CONTENT_SCAN
3264 unspool_mbox();
3265#endif
6f0c431a
PP
3266#ifdef EXPERIMENTAL_DCC
3267 dcc_ok = 0;
3268#endif
6ea85e9a
PH
3269 /* The ACL can specify where rejections are to be logged, possibly
3270 nowhere. The default is main and reject logs. */
3271
3272 if (log_reject_target != 0)
3273 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3274 sender_address, log_msg);
3275
54cdb463
PH
3276 if (user_msg == NULL) user_msg = US"local configuration problem";
3277 if (smtp_batched_input)
3278 {
3279 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3280 /* Does not return */
3281 }
3282 else
3283 {
3284 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3285 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3286 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3287 header_list);
3288 /* Does not return */
3289 }
059ec3d9 3290 }
54cdb463 3291 add_acl_headers(US"non-SMTP");
059ec3d9 3292 }
059ec3d9
PH
3293 }
3294
54cdb463
PH
3295 /* The applicable ACLs have been run */
3296
059ec3d9
PH
3297 if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */
3298 if (queue_only_policy) queued_by = US"ACL";
059ec3d9
PH
3299 }
3300
8523533c
TK
3301#ifdef WITH_CONTENT_SCAN
3302unspool_mbox();
3303#endif
3304
6a8f9482
TK
3305#ifdef EXPERIMENTAL_DCC
3306dcc_ok = 0;
3307#endif
3308
3309
059ec3d9
PH
3310/* The final check on the message is to run the scan_local() function. The
3311version supplied with Exim always accepts, but this is a hook for sysadmins to
3312supply their own checking code. The local_scan() function is run even when all
3313the recipients have been discarded. */
3314
3315lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3316
3317/* Arrange to catch crashes in local_scan(), so that the -D file gets
3318deleted, and the incident gets logged. */
3319
3320os_non_restarting_signal(SIGSEGV, local_scan_crash_handler);
3321os_non_restarting_signal(SIGFPE, local_scan_crash_handler);
3322os_non_restarting_signal(SIGILL, local_scan_crash_handler);
3323os_non_restarting_signal(SIGBUS, local_scan_crash_handler);
3324
3325DEBUG(D_receive) debug_printf("calling local_scan(); timeout=%d\n",
3326 local_scan_timeout);
3327local_scan_data = NULL;
3328
3329os_non_restarting_signal(SIGALRM, local_scan_timeout_handler);
3330if (local_scan_timeout > 0) alarm(local_scan_timeout);
3331rc = local_scan(data_fd, &local_scan_data);
3332alarm(0);
3333os_non_restarting_signal(SIGALRM, sigalrm_handler);
3334
0e20aff9
MH
3335enable_dollar_recipients = FALSE;
3336
059ec3d9
PH
3337store_pool = POOL_MAIN; /* In case changed */
3338DEBUG(D_receive) debug_printf("local_scan() returned %d %s\n", rc,
3339 local_scan_data);
3340
3341os_non_restarting_signal(SIGSEGV, SIG_DFL);
3342os_non_restarting_signal(SIGFPE, SIG_DFL);
3343os_non_restarting_signal(SIGILL, SIG_DFL);
3344os_non_restarting_signal(SIGBUS, SIG_DFL);
3345
3346/* The length check is paranoia against some runaway code, and also because
3347(for a success return) lines in the spool file are read into big_buffer. */
3348
3349if (local_scan_data != NULL)
3350 {
3351 int len = Ustrlen(local_scan_data);
3352 if (len > LOCAL_SCAN_MAX_RETURN) len = LOCAL_SCAN_MAX_RETURN;
3353 local_scan_data = string_copyn(local_scan_data, len);
3354 }
3355
3356if (rc == LOCAL_SCAN_ACCEPT_FREEZE)
3357 {
58eb016e 3358 if (!deliver_freeze) /* ACL might have already frozen */
059ec3d9
PH
3359 {
3360 deliver_freeze = TRUE;
3361 deliver_frozen_at = time(NULL);
3362 frozen_by = US"local_scan()";
3363 }
3364 rc = LOCAL_SCAN_ACCEPT;
3365 }
3366else if (rc == LOCAL_SCAN_ACCEPT_QUEUE)
3367 {
3368 if (!queue_only_policy) /* ACL might have already queued */
3369 {
3370 queue_only_policy = TRUE;
3371 queued_by = US"local_scan()";
3372 }
3373 rc = LOCAL_SCAN_ACCEPT;
3374 }
3375
3376/* Message accepted: remove newlines in local_scan_data because otherwise
3377the spool file gets corrupted. Ensure that all recipients are qualified. */
3378
3379if (rc == LOCAL_SCAN_ACCEPT)
3380 {
3381 if (local_scan_data != NULL)
3382 {
3383 uschar *s;
3384 for (s = local_scan_data; *s != 0; s++) if (*s == '\n') *s = ' ';
3385 }
3386 for (i = 0; i < recipients_count; i++)
3387 {
3388 recipient_item *r = recipients_list + i;
3389 r->address = rewrite_address_qualify(r->address, TRUE);
3390 if (r->errors_to != NULL)
3391 r->errors_to = rewrite_address_qualify(r->errors_to, TRUE);
3392 }
3393 if (recipients_count == 0 && blackholed_by == NULL)
3394 blackholed_by = US"local_scan";
3395 }
3396
3397/* Message rejected: newlines permitted in local_scan_data to generate
3398multiline SMTP responses. */
3399
3400else
3401 {
3402 uschar *istemp = US"";
3403 uschar *s = NULL;
a5bd321b 3404 uschar *smtp_code;
059ec3d9
PH
3405 int size = 0;
3406 int sptr = 0;
059ec3d9
PH
3407
3408 errmsg = local_scan_data;
3409
3410 Uunlink(spool_name); /* Cancel this message */
3411 switch(rc)
3412 {
3413 default:
3414 log_write(0, LOG_MAIN, "invalid return %d from local_scan(). Temporary "
3415 "rejection given", rc);
3416 goto TEMPREJECT;
3417
3418 case LOCAL_SCAN_REJECT_NOLOGHDR:
3419 log_extra_selector &= ~LX_rejected_header;
3420 /* Fall through */
3421
3422 case LOCAL_SCAN_REJECT:
a5bd321b 3423 smtp_code = US"550";
059ec3d9
PH
3424 if (errmsg == NULL) errmsg = US"Administrative prohibition";
3425 break;
3426
3427 case LOCAL_SCAN_TEMPREJECT_NOLOGHDR:
3428 log_extra_selector &= ~LX_rejected_header;
3429 /* Fall through */
3430
3431 case LOCAL_SCAN_TEMPREJECT:
3432 TEMPREJECT:
a5bd321b 3433 smtp_code = US"451";
059ec3d9
PH
3434 if (errmsg == NULL) errmsg = US"Temporary local problem";
3435 istemp = US"temporarily ";
3436 break;
3437 }
3438
3439 s = string_append(s, &size, &sptr, 2, US"F=",
3440 (sender_address[0] == 0)? US"<>" : sender_address);
3441 s = add_host_info_for_log(s, &size, &sptr);
3442 s[sptr] = 0;
3443
3444 log_write(0, LOG_MAIN|LOG_REJECT, "%s %srejected by local_scan(): %.256s",
3445 s, istemp, string_printing(errmsg));
3446
3447 if (smtp_input)
3448 {
3449 if (!smtp_batched_input)
3450 {
a5bd321b 3451 smtp_respond(smtp_code, 3, TRUE, errmsg);
059ec3d9
PH
3452 message_id[0] = 0; /* Indicate no message accepted */
3453 smtp_reply = US""; /* Indicate reply already sent */
3454 goto TIDYUP; /* Skip to end of function */
3455 }
3456 else
3457 {
a5bd321b 3458 moan_smtp_batch(NULL, "%s %s", smtp_code, errmsg);
059ec3d9
PH
3459 /* Does not return */
3460 }
3461 }
3462 else
3463 {
3464 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3465 give_local_error(ERRMESS_LOCAL_SCAN, errmsg,
3466 US"message rejected by local scan code: ", error_rc, data_file,
3467 header_list);
3468 /* Does not return */
3469 }
3470 }
3471
3472/* Reset signal handlers to ignore signals that previously would have caused
3473the message to be abandoned. */
3474
3475signal(SIGTERM, SIG_IGN);
3476signal(SIGINT, SIG_IGN);
3477
e4bdf652 3478
059ec3d9
PH
3479/* Ensure the first time flag is set in the newly-received message. */
3480
3481deliver_firsttime = TRUE;
3482
8523533c
TK
3483#ifdef EXPERIMENTAL_BRIGHTMAIL
3484if (bmi_run == 1) {
3485 /* rewind data file */
3486 lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3487 bmi_verdicts = bmi_process_message(header_list, data_fd);
3488};
3489#endif
3490
059ec3d9
PH
3491/* Update the timstamp in our Received: header to account for any time taken by
3492an ACL or by local_scan(). The new time is the time that all reception
3493processing is complete. */
3494
3495timestamp = expand_string(US"${tod_full}");
3496tslen = Ustrlen(timestamp);
3497
3498memcpy(received_header->text + received_header->slen - tslen - 1,
3499 timestamp, tslen);
3500
3501/* In MUA wrapper mode, ignore queueing actions set by ACL or local_scan() */
3502
3503if (mua_wrapper)
3504 {
3505 deliver_freeze = FALSE;
3506 queue_only_policy = FALSE;
3507 }
3508
3509/* Keep the data file open until we have written the header file, in order to
3510hold onto the lock. In a -bh run, or if the message is to be blackholed, we
3511don't write the header file, and we unlink the data file. If writing the header
3512file fails, we have failed to accept this message. */
3513
3514if (host_checking || blackholed_by != NULL)
3515 {
3516 header_line *h;
3517 Uunlink(spool_name);
3518 msg_size = 0; /* Compute size for log line */
3519 for (h = header_list; h != NULL; h = h->next)
3520 if (h->type != '*') msg_size += h->slen;
3521 }
3522
3523/* Write the -H file */
3524
3525else
3526 {
3527 if ((msg_size = spool_write_header(message_id, SW_RECEIVING, &errmsg)) < 0)
3528 {
3529 log_write(0, LOG_MAIN, "Message abandoned: %s", errmsg);
3530 Uunlink(spool_name); /* Lose the data file */
3531
3532 if (smtp_input)
3533 {
3534 smtp_reply = US"451 Error in writing spool file";
3535 message_id[0] = 0; /* Indicate no message accepted */
3536 goto TIDYUP;
3537 }
3538 else
3539 {
3540 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3541 give_local_error(ERRMESS_IOERR, errmsg, US"", error_rc, data_file,
3542 header_list);
3543 /* Does not return */
3544 }
3545 }
3546 }
3547
3548
3549/* The message has now been successfully received. */
3550
3551receive_messagecount++;
3552
3553/* In SMTP sessions we may receive several in one connection. After each one,
3554we wait for the clock to tick at the level of message-id granularity. This is
3555so that the combination of time+pid is unique, even on systems where the pid
3556can be re-used within our time interval. We can't shorten the interval without
3557re-designing the message-id. See comments above where the message id is
3558created. This is Something For The Future. */
3559
3560message_id_tv.tv_usec = (message_id_tv.tv_usec/id_resolution) * id_resolution;
3561exim_wait_tick(&message_id_tv, id_resolution);
3562
3563/* Add data size to written header size. We do not count the initial file name
3564that is in the file, but we do add one extra for the notional blank line that
3565precedes the data. This total differs from message_size in that it include the
3566added Received: header and any other headers that got created locally. */
3567
3568fflush(data_file);
3569fstat(data_fd, &statbuf);
3570
3571msg_size += statbuf.st_size - SPOOL_DATA_START_OFFSET + 1;
3572
3573/* Generate a "message received" log entry. We do this by building up a dynamic
3574string as required. Since we commonly want to add two items at a time, use a
3575macro to simplify the coding. We log the arrival of a new message while the
3576file is still locked, just in case the machine is *really* fast, and delivers
3577it first! Include any message id that is in the message - since the syntax of a
3578message id is actually an addr-spec, we can use the parse routine to canonicize
3579it. */
3580
3581size = 256;
3582sptr = 0;
3583s = store_get(size);
3584
3585s = string_append(s, &size, &sptr, 2, US"<= ",
3586 (sender_address[0] == 0)? US"<>" : sender_address);
3587if (message_reference != NULL)
3588 s = string_append(s, &size, &sptr, 2, US" R=", message_reference);
3589
3590s = add_host_info_for_log(s, &size, &sptr);
3591
3592#ifdef SUPPORT_TLS
817d9f57
JH
3593if ((log_extra_selector & LX_tls_cipher) != 0 && tls_in.cipher != NULL)
3594 s = string_append(s, &size, &sptr, 2, US" X=", tls_in.cipher);
059ec3d9 3595if ((log_extra_selector & LX_tls_certificate_verified) != 0 &&
817d9f57 3596 tls_in.cipher != NULL)
059ec3d9 3597 s = string_append(s, &size, &sptr, 2, US" CV=",
817d9f57
JH
3598 tls_in.certificate_verified? "yes":"no");
3599if ((log_extra_selector & LX_tls_peerdn) != 0 && tls_in.peerdn != NULL)
48ed62d9 3600 s = string_append(s, &size, &sptr, 3, US" DN=\"",
817d9f57
JH
3601 string_printing(tls_in.peerdn), US"\"");
3602if ((log_extra_selector & LX_tls_sni) != 0 && tls_in.sni != NULL)
3f0945ff 3603 s = string_append(s, &size, &sptr, 3, US" SNI=\"",
817d9f57 3604 string_printing(tls_in.sni), US"\"");
3f0945ff 3605#endif
059ec3d9
PH
3606
3607if (sender_host_authenticated != NULL)
3608 {
3609 s = string_append(s, &size, &sptr, 2, US" A=", sender_host_authenticated);
3610 if (authenticated_id != NULL)
c8e2fc1e 3611 {
059ec3d9 3612 s = string_append(s, &size, &sptr, 2, US":", authenticated_id);
c8e2fc1e
JH
3613 if (log_extra_selector & LX_smtp_mailauth && authenticated_sender != NULL)
3614 s = string_append(s, &size, &sptr, 2, US":", authenticated_sender);
3615 }
059ec3d9
PH
3616 }
3617
3618sprintf(CS big_buffer, "%d", msg_size);
3619s = string_append(s, &size, &sptr, 2, US" S=", big_buffer);
3620
3c0a92dc
JH
3621/* log 8BITMIME mode announced in MAIL_FROM
3622 0 ... no BODY= used
3623 7 ... 7BIT
3624 8 ... 8BITMIME */
c8e2fc1e
JH
3625if (log_extra_selector & LX_8bitmime)
3626 {
3c0a92dc
JH
3627 sprintf(CS big_buffer, "%d", body_8bitmime);
3628 s = string_append(s, &size, &sptr, 2, US" M8S=", big_buffer);
c8e2fc1e 3629 }
3c0a92dc 3630
059ec3d9
PH
3631/* If an addr-spec in a message-id contains a quoted string, it can contain
3632any characters except " \ and CR and so in particular it can contain NL!
3633Therefore, make sure we use a printing-characters only version for the log.
3634Also, allow for domain literals in the message id. */
3635
3636if (msgid_header != NULL)
3637 {
3638 uschar *old_id;
3639 BOOL save_allow_domain_literals = allow_domain_literals;
3640 allow_domain_literals = TRUE;
3641 old_id = parse_extract_address(Ustrchr(msgid_header->text, ':') + 1,
3642 &errmsg, &start, &end, &domain, FALSE);
3643 allow_domain_literals = save_allow_domain_literals;
3644 if (old_id != NULL)
3645 s = string_append(s, &size, &sptr, 2, US" id=", string_printing(old_id));
3646 }
3647
3648/* If subject logging is turned on, create suitable printing-character
3649text. By expanding $h_subject: we make use of the MIME decoding. */
3650
3651if ((log_extra_selector & LX_subject) != 0 && subject_header != NULL)
3652 {
3653 int i;
3654 uschar *p = big_buffer;
3655 uschar *ss = expand_string(US"$h_subject:");
3656
3657 /* Backslash-quote any double quotes or backslashes so as to make a
3658 a C-like string, and turn any non-printers into escape sequences. */
3659
3660 *p++ = '\"';
3661 if (*ss != 0) for (i = 0; i < 100 && ss[i] != 0; i++)
3662 {
3663 if (ss[i] == '\"' || ss[i] == '\\') *p++ = '\\';
3664 *p++ = ss[i];
3665 }
3666 *p++ = '\"';
3667 *p = 0;
3668 s = string_append(s, &size, &sptr, 2, US" T=", string_printing(big_buffer));
3669 }
3670
3671/* Terminate the string: string_cat() and string_append() leave room, but do
3672not put the zero in. */
3673
3674s[sptr] = 0;
3675
059ec3d9
PH
3676/* Create a message log file if message logs are being used and this message is
3677not blackholed. Write the reception stuff to it. We used to leave message log
e4bdf652 3678creation until the first delivery, but this has proved confusing for some
059ec3d9
PH
3679people. */
3680
3681if (message_logs && blackholed_by == NULL)
3682 {
3683 int fd;
3684
3685 sprintf(CS spool_name, "%s/msglog/%s/%s", spool_directory, message_subdir,
3686 message_id);
3687 fd = Uopen(spool_name, O_WRONLY|O_APPEND|O_CREAT, SPOOL_MODE);
3688
3689 if (fd < 0 && errno == ENOENT)
3690 {
3691 uschar temp[16];
3692 sprintf(CS temp, "msglog/%s", message_subdir);
3693 if (message_subdir[0] == 0) temp[6] = 0;
3694 (void)directory_make(spool_directory, temp, MSGLOG_DIRECTORY_MODE, TRUE);
3695 fd = Uopen(spool_name, O_WRONLY|O_APPEND|O_CREAT, SPOOL_MODE);
3696 }
3697
3698 if (fd < 0)
3699 {
3700 log_write(0, LOG_MAIN|LOG_PANIC, "Couldn't open message log %s: %s",
3701 spool_name, strerror(errno));
3702 }
3703
3704 else
3705 {
3706 FILE *message_log = fdopen(fd, "a");
3707 if (message_log == NULL)
3708 {
3709 log_write(0, LOG_MAIN|LOG_PANIC, "Couldn't fdopen message log %s: %s",
3710 spool_name, strerror(errno));
f1e894f3 3711 (void)close(fd);
059ec3d9
PH
3712 }
3713 else
3714 {
3715 uschar *now = tod_stamp(tod_log);
3716 fprintf(message_log, "%s Received from %s\n", now, s+3);
3717 if (deliver_freeze) fprintf(message_log, "%s frozen by %s\n", now,
3718 frozen_by);
3719 if (queue_only_policy) fprintf(message_log,
3720 "%s no immediate delivery: queued by %s\n", now, queued_by);
f1e894f3 3721 (void)fclose(message_log);
059ec3d9
PH
3722 }
3723 }
3724 }
3725
58eb016e
PH
3726/* Everything has now been done for a successful message except logging its
3727arrival, and outputting an SMTP response. While writing to the log, set a flag
3728to cause a call to receive_bomb_out() if the log cannot be opened. */
3729
3730receive_call_bombout = TRUE;
3731
563b63fa
PH
3732/* Before sending an SMTP response in a TCP/IP session, we check to see if the
3733connection has gone away. This can only be done if there is no unconsumed input
3734waiting in the local input buffer. We can test for this by calling
3735receive_smtp_buffered(). RFC 2920 (pipelining) explicitly allows for additional
3736input to be sent following the final dot, so the presence of following input is
3737not an error.
58eb016e 3738
563b63fa
PH
3739If the connection is still present, but there is no unread input for the
3740socket, the result of a select() call will be zero. If, however, the connection
3741has gone away, or if there is pending input, the result of select() will be
3742non-zero. The two cases can be distinguished by trying to read the next input
3743character. If we succeed, we can unread it so that it remains in the local
3744buffer for handling later. If not, the connection has been lost.
58eb016e 3745
563b63fa
PH
3746Of course, since TCP/IP is asynchronous, there is always a chance that the
3747connection will vanish between the time of this test and the sending of the
3748response, but the chance of this happening should be small. */
3749
3750if (smtp_input && sender_host_address != NULL && !sender_host_notsocket &&
3751 !receive_smtp_buffered())
58eb016e
PH
3752 {
3753 struct timeval tv;
3754 fd_set select_check;
3755 FD_ZERO(&select_check);
3756 FD_SET(fileno(smtp_in), &select_check);
3757 tv.tv_sec = 0;
3758 tv.tv_usec = 0;
3759
563b63fa 3760 if (select(fileno(smtp_in) + 1, &select_check, NULL, NULL, &tv) != 0)
58eb016e 3761 {
80a47a2c
TK
3762 int c = (receive_getc)();
3763 if (c != EOF) (receive_ungetc)(c); else
58eb016e 3764 {
563b63fa
PH
3765 uschar *msg = US"SMTP connection lost after final dot";
3766 smtp_reply = US""; /* No attempt to send a response */
3767 smtp_yield = FALSE; /* Nothing more on this connection */
58eb016e 3768
563b63fa 3769 /* Re-use the log line workspace */
58eb016e 3770
563b63fa
PH
3771 sptr = 0;
3772 s = string_cat(s, &size, &sptr, msg, Ustrlen(msg));
3773 s = add_host_info_for_log(s, &size, &sptr);
3774 s[sptr] = 0;
3775 log_write(0, LOG_MAIN, "%s", s);
58eb016e 3776
563b63fa 3777 /* Delete the files for this aborted message. */
58eb016e 3778
563b63fa
PH
3779 sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory,
3780 message_subdir, message_id);
3781 Uunlink(spool_name);
58eb016e 3782
563b63fa
PH
3783 sprintf(CS spool_name, "%s/input/%s/%s-H", spool_directory,
3784 message_subdir, message_id);
3785 Uunlink(spool_name);
58eb016e 3786
563b63fa
PH
3787 sprintf(CS spool_name, "%s/msglog/%s/%s", spool_directory,
3788 message_subdir, message_id);
3789 Uunlink(spool_name);
58eb016e 3790
563b63fa
PH
3791 goto TIDYUP;
3792 }
58eb016e
PH
3793 }
3794 }
3795
3796/* The connection has not gone away; we really are going to take responsibility
3797for this message. */
3798
817d9f57 3799/* Cutthrough - had sender last-dot; assume we've sent (or bufferred) all