logging
[exim.git] / src / src / receive.c
CommitLineData
059ec3d9
PH
1/*************************************************
2* Exim - an Internet mail transport agent *
3*************************************************/
4
80fea873 5/* Copyright (c) University of Cambridge 1995 - 2016 */
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
4840604e 16#ifdef EXPERIMENTAL_DMARC
c007c974 17# include "dmarc.h"
4840604e
TL
18#endif /* EXPERIMENTAL_DMARC */
19
059ec3d9
PH
20/*************************************************
21* Local static variables *
22*************************************************/
23
24static FILE *data_file = NULL;
25static int data_fd = -1;
26static uschar spool_name[256];
27
28
29
30/*************************************************
31* Non-SMTP character reading functions *
32*************************************************/
33
34/* These are the default functions that are set up in the variables such as
35receive_getc initially. They just call the standard functions, passing stdin as
36the file. (When SMTP input is occurring, different functions are used by
37changing the pointer variables.) */
38
39int
40stdin_getc(void)
41{
42return getc(stdin);
43}
44
45int
46stdin_ungetc(int c)
47{
48return ungetc(c, stdin);
49}
50
51int
52stdin_feof(void)
53{
54return feof(stdin);
55}
56
57int
58stdin_ferror(void)
59{
60return ferror(stdin);
61}
62
63
64
65
66/*************************************************
67* Check that a set sender is allowed *
68*************************************************/
69
70/* This function is called when a local caller sets an explicit sender address.
71It checks whether this is permitted, which it is for trusted callers.
72Otherwise, it must match the pattern(s) in untrusted_set_sender.
73
74Arguments: the proposed sender address
75Returns: TRUE for a trusted caller
76 TRUE if the address has been set, untrusted_set_sender has been
77 set, and the address matches something in the list
78 FALSE otherwise
79*/
80
81BOOL
82receive_check_set_sender(uschar *newsender)
83{
84uschar *qnewsender;
85if (trusted_caller) return TRUE;
86if (newsender == NULL || untrusted_set_sender == NULL) return FALSE;
87qnewsender = (Ustrchr(newsender, '@') != NULL)?
88 newsender : string_sprintf("%s@%s", newsender, qualify_domain_sender);
89return
55414b25 90 match_address_list(qnewsender, TRUE, TRUE, CUSS &untrusted_set_sender, NULL, -1,
059ec3d9
PH
91 0, NULL) == OK;
92}
93
94
95
96
97/*************************************************
5cb8cbc6 98* Read space info for a partition *
059ec3d9
PH
99*************************************************/
100
8e669ac1
PH
101/* This function is called by receive_check_fs() below, and also by string
102expansion for variables such as $spool_space. The field names for the statvfs
5cb8cbc6
PH
103structure are macros, because not all OS have F_FAVAIL and it seems tidier to
104have macros for F_BAVAIL and F_FILES as well. Some kinds of file system do not
105have inodes, and they return -1 for the number available.
059ec3d9 106
5cb8cbc6
PH
107Later: It turns out that some file systems that do not have the concept of
108inodes return 0 rather than -1. Such systems should also return 0 for the total
8e669ac1 109number of inodes, so we require that to be greater than zero before returning
5cb8cbc6 110an inode count.
059ec3d9 111
5cb8cbc6
PH
112Arguments:
113 isspool TRUE for spool partition, FALSE for log partition
114 inodeptr address of int to receive inode count; -1 if there isn't one
8e669ac1 115
5cb8cbc6 116Returns: available on-root space, in kilobytes
8e669ac1
PH
117 -1 for log partition if there isn't one
118
119All values are -1 if the STATFS functions are not available.
059ec3d9
PH
120*/
121
8e669ac1 122int
5cb8cbc6 123receive_statvfs(BOOL isspool, int *inodeptr)
059ec3d9
PH
124{
125#ifdef HAVE_STATFS
059ec3d9 126struct STATVFS statbuf;
5cb8cbc6
PH
127uschar *path;
128uschar *name;
129uschar buffer[1024];
059ec3d9 130
5cb8cbc6 131/* The spool directory must always exist. */
059ec3d9 132
5cb8cbc6 133if (isspool)
059ec3d9 134 {
8e669ac1
PH
135 path = spool_directory;
136 name = US"spool";
137 }
138
059ec3d9
PH
139/* Need to cut down the log file path to the directory, and to ignore any
140appearance of "syslog" in it. */
141
5cb8cbc6 142else
059ec3d9 143 {
059ec3d9 144 int sep = ':'; /* Not variable - outside scripts use */
55414b25 145 const uschar *p = log_file_path;
8e669ac1 146 name = US"log";
059ec3d9
PH
147
148 /* An empty log_file_path means "use the default". This is the same as an
149 empty item in a list. */
150
151 if (*p == 0) p = US":";
55414b25
JH
152 while ((path = string_nextinlist(&p, &sep, buffer, sizeof(buffer))))
153 if (Ustrcmp(path, "syslog") != 0)
154 break;
059ec3d9 155
5cb8cbc6
PH
156 if (path == NULL) /* No log files */
157 {
8e669ac1
PH
158 *inodeptr = -1;
159 return -1;
160 }
059ec3d9 161
8e669ac1
PH
162 /* An empty string means use the default, which is in the spool directory.
163 But don't just use the spool directory, as it is possible that the log
5cb8cbc6 164 subdirectory has been symbolically linked elsewhere. */
059ec3d9 165
8e669ac1 166 if (path[0] == 0)
059ec3d9 167 {
5cb8cbc6
PH
168 sprintf(CS buffer, CS"%s/log", CS spool_directory);
169 path = buffer;
8e669ac1
PH
170 }
171 else
059ec3d9 172 {
8e669ac1 173 uschar *cp;
5cb8cbc6 174 if ((cp = Ustrrchr(path, '/')) != NULL) *cp = 0;
8e669ac1 175 }
5cb8cbc6 176 }
8e669ac1 177
8f128379 178/* We now have the path; do the business */
5cb8cbc6
PH
179
180memset(&statbuf, 0, sizeof(statbuf));
181
182if (STATVFS(CS path, &statbuf) != 0)
183 {
184 log_write(0, LOG_MAIN|LOG_PANIC, "cannot accept message: failed to stat "
185 "%s directory %s: %s", name, spool_directory, strerror(errno));
186 smtp_closedown(US"spool or log directory problem");
187 exim_exit(EXIT_FAILURE);
188 }
8e669ac1 189
5cb8cbc6
PH
190*inodeptr = (statbuf.F_FILES > 0)? statbuf.F_FAVAIL : -1;
191
192/* Disks are getting huge. Take care with computing the size in kilobytes. */
8e669ac1 193
5cb8cbc6
PH
194return (int)(((double)statbuf.F_BAVAIL * (double)statbuf.F_FRSIZE)/1024.0);
195
196/* Unable to find partition sizes in this environment. */
197
198#else
199*inodeptr = -1;
200return -1;
201#endif
202}
203
059ec3d9 204
059ec3d9 205
5cb8cbc6
PH
206
207/*************************************************
208* Check space on spool and log partitions *
209*************************************************/
210
211/* This function is called before accepting a message; if any thresholds are
212set, it checks them. If a message_size is supplied, it checks that there is
213enough space for that size plus the threshold - i.e. that the message won't
214reduce the space to the threshold. Not all OS have statvfs(); for those that
215don't, this function always returns TRUE. For some OS the old function and
216struct name statfs is used; that is handled by a macro, defined in exim.h.
217
218Arguments:
219 msg_size the (estimated) size of an incoming message
220
221Returns: FALSE if there isn't enough space, or if the information cannot
222 be obtained
223 TRUE if no check was done or there is enough space
224*/
225
226BOOL
227receive_check_fs(int msg_size)
228{
229int space, inodes;
230
231if (check_spool_space > 0 || msg_size > 0 || check_spool_inodes > 0)
232 {
8e669ac1
PH
233 space = receive_statvfs(TRUE, &inodes);
234
059ec3d9 235 DEBUG(D_receive)
5cb8cbc6
PH
236 debug_printf("spool directory space = %dK inodes = %d "
237 "check_space = %dK inodes = %d msg_size = %d\n",
238 space, inodes, check_spool_space, check_spool_inodes, msg_size);
8e669ac1
PH
239
240 if ((space >= 0 && space < check_spool_space) ||
5cb8cbc6 241 (inodes >= 0 && inodes < check_spool_inodes))
8e669ac1 242 {
5cb8cbc6
PH
243 log_write(0, LOG_MAIN, "spool directory space check failed: space=%d "
244 "inodes=%d", space, inodes);
059ec3d9
PH
245 return FALSE;
246 }
247 }
248
5cb8cbc6
PH
249if (check_log_space > 0 || check_log_inodes > 0)
250 {
8e669ac1
PH
251 space = receive_statvfs(FALSE, &inodes);
252
5cb8cbc6
PH
253 DEBUG(D_receive)
254 debug_printf("log directory space = %dK inodes = %d "
255 "check_space = %dK inodes = %d\n",
256 space, inodes, check_log_space, check_log_inodes);
8e669ac1
PH
257
258 if ((space >= 0 && space < check_log_space) ||
5cb8cbc6 259 (inodes >= 0 && inodes < check_log_inodes))
8e669ac1 260 {
5cb8cbc6
PH
261 log_write(0, LOG_MAIN, "log directory space check failed: space=%d "
262 "inodes=%d", space, inodes);
263 return FALSE;
264 }
8e669ac1
PH
265 }
266
059ec3d9
PH
267return TRUE;
268}
269
270
271
272/*************************************************
273* Bomb out while reading a message *
274*************************************************/
275
276/* The common case of wanting to bomb out is if a SIGTERM or SIGINT is
277received, or if there is a timeout. A rarer case might be if the log files are
278screwed up and Exim can't open them to record a message's arrival. Handling
279that case is done by setting a flag to cause the log functions to call this
280function if there is an ultimate disaster. That is why it is globally
281accessible.
282
8f128379
PH
283Arguments:
284 reason text reason to pass to the not-quit ACL
285 msg default SMTP response to give if in an SMTP session
059ec3d9
PH
286Returns: it doesn't
287*/
288
289void
8f128379 290receive_bomb_out(uschar *reason, uschar *msg)
059ec3d9 291{
ead37e6c
PP
292 static BOOL already_bombing_out;
293/* The smtp_notquit_exit() below can call ACLs which can trigger recursive
294timeouts, if someone has something slow in their quit ACL. Since the only
295things we should be doing are to close down cleanly ASAP, on the second
296pass we also close down stuff that might be opened again, before bypassing
297the ACL call and exiting. */
298
059ec3d9
PH
299/* If spool_name is set, it contains the name of the data file that is being
300written. Unlink it before closing so that it cannot be picked up by a delivery
301process. Ensure that any header file is also removed. */
302
ead37e6c 303if (spool_name[0] != '\0')
059ec3d9
PH
304 {
305 Uunlink(spool_name);
306 spool_name[Ustrlen(spool_name) - 1] = 'H';
307 Uunlink(spool_name);
ead37e6c 308 spool_name[0] = '\0';
059ec3d9
PH
309 }
310
311/* Now close the file if it is open, either as a fd or a stream. */
312
ead37e6c
PP
313if (data_file != NULL)
314 {
315 (void)fclose(data_file);
316 data_file = NULL;
317} else if (data_fd >= 0) {
318 (void)close(data_fd);
319 data_fd = -1;
320 }
059ec3d9 321
8f128379
PH
322/* Attempt to close down an SMTP connection tidily. For non-batched SMTP, call
323smtp_notquit_exit(), which runs the NOTQUIT ACL, if present, and handles the
324SMTP response. */
059ec3d9 325
ead37e6c 326if (!already_bombing_out)
059ec3d9 327 {
ead37e6c
PP
328 already_bombing_out = TRUE;
329 if (smtp_input)
330 {
331 if (smtp_batched_input)
332 moan_smtp_batch(NULL, "421 %s - message abandoned", msg); /* No return */
333 smtp_notquit_exit(reason, US"421", US"%s %s - closing connection.",
334 smtp_active_hostname, msg);
335 }
059ec3d9
PH
336 }
337
338/* Exit from the program (non-BSMTP cases) */
339
340exim_exit(EXIT_FAILURE);
341}
342
343
344/*************************************************
345* Data read timeout *
346*************************************************/
347
348/* Handler function for timeouts that occur while reading the data that
349comprises a message.
350
351Argument: the signal number
352Returns: nothing
353*/
354
355static void
356data_timeout_handler(int sig)
357{
358uschar *msg = NULL;
359
360sig = sig; /* Keep picky compilers happy */
361
362if (smtp_input)
363 {
364 msg = US"SMTP incoming data timeout";
365 log_write(L_lost_incoming_connection,
366 LOG_MAIN, "SMTP data timeout (message abandoned) on connection "
fed77020
PH
367 "from %s F=<%s>",
368 (sender_fullhost != NULL)? sender_fullhost : US"local process",
369 sender_address);
059ec3d9
PH
370 }
371else
372 {
373 fprintf(stderr, "exim: timed out while reading - message abandoned\n");
374 log_write(L_lost_incoming_connection,
375 LOG_MAIN, "timed out while reading local message");
376 }
377
8f128379 378receive_bomb_out(US"data-timeout", msg); /* Does not return */
059ec3d9
PH
379}
380
381
382
383/*************************************************
384* local_scan() timeout *
385*************************************************/
386
387/* Handler function for timeouts that occur while running a local_scan()
388function.
389
390Argument: the signal number
391Returns: nothing
392*/
393
394static void
395local_scan_timeout_handler(int sig)
396{
397sig = sig; /* Keep picky compilers happy */
398log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function timed out - "
399 "message temporarily rejected (size %d)", message_size);
8f128379
PH
400/* Does not return */
401receive_bomb_out(US"local-scan-timeout", US"local verification problem");
059ec3d9
PH
402}
403
404
405
406/*************************************************
407* local_scan() crashed *
408*************************************************/
409
410/* Handler function for signals that occur while running a local_scan()
411function.
412
413Argument: the signal number
414Returns: nothing
415*/
416
417static void
418local_scan_crash_handler(int sig)
419{
420log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function crashed with "
421 "signal %d - message temporarily rejected (size %d)", sig, message_size);
8f128379
PH
422/* Does not return */
423receive_bomb_out(US"local-scan-error", US"local verification problem");
059ec3d9
PH
424}
425
426
427/*************************************************
428* SIGTERM or SIGINT received *
429*************************************************/
430
431/* Handler for SIGTERM or SIGINT signals that occur while reading the
432data that comprises a message.
433
434Argument: the signal number
435Returns: nothing
436*/
437
438static void
439data_sigterm_sigint_handler(int sig)
440{
441uschar *msg = NULL;
442
443if (smtp_input)
444 {
445 msg = US"Service not available - SIGTERM or SIGINT received";
446 log_write(0, LOG_MAIN, "%s closed after %s", smtp_get_connection_info(),
447 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
448 }
449else
450 {
f05da2e8 451 if (filter_test == FTEST_NONE)
059ec3d9
PH
452 {
453 fprintf(stderr, "\nexim: %s received - message abandoned\n",
454 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
455 log_write(0, LOG_MAIN, "%s received while reading local message",
456 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
457 }
458 }
459
8f128379 460receive_bomb_out(US"signal-exit", msg); /* Does not return */
059ec3d9
PH
461}
462
463
464
465/*************************************************
466* Add new recipient to list *
467*************************************************/
468
469/* This function builds a list of recipient addresses in argc/argv
470format.
471
472Arguments:
473 recipient the next address to add to recipients_list
474 pno parent number for fixed aliases; -1 otherwise
475
476Returns: nothing
477*/
478
479void
480receive_add_recipient(uschar *recipient, int pno)
481{
482if (recipients_count >= recipients_list_max)
483 {
484 recipient_item *oldlist = recipients_list;
485 int oldmax = recipients_list_max;
486 recipients_list_max = recipients_list_max? 2*recipients_list_max : 50;
487 recipients_list = store_get(recipients_list_max * sizeof(recipient_item));
488 if (oldlist != NULL)
489 memcpy(recipients_list, oldlist, oldmax * sizeof(recipient_item));
490 }
491
492recipients_list[recipients_count].address = recipient;
493recipients_list[recipients_count].pno = pno;
8523533c
TK
494#ifdef EXPERIMENTAL_BRIGHTMAIL
495recipients_list[recipients_count].bmi_optin = bmi_current_optin;
496/* reset optin string pointer for next recipient */
497bmi_current_optin = NULL;
498#endif
6c1c3d1d
WB
499recipients_list[recipients_count].orcpt = NULL;
500recipients_list[recipients_count].dsn_flags = 0;
059ec3d9
PH
501recipients_list[recipients_count++].errors_to = NULL;
502}
503
504
505
506
507/*************************************************
fd98a5c6
JH
508* Send user response message *
509*************************************************/
61147df4 510
fd98a5c6
JH
511/* This function is passed a default response code and a user message. It calls
512smtp_message_code() to check and possibly modify the response code, and then
513calls smtp_respond() to transmit the response. I put this into a function
514just to avoid a lot of repetition.
61147df4
PP
515
516Arguments:
fd98a5c6
JH
517 code the response code
518 user_msg the user message
519
520Returns: nothing
61147df4
PP
521*/
522
8ccd00b1 523#ifndef DISABLE_PRDR
61147df4 524static void
fd98a5c6 525smtp_user_msg(uschar *code, uschar *user_msg)
61147df4 526{
fd98a5c6 527int len = 3;
4f6ae5c3 528smtp_message_code(&code, &len, &user_msg, NULL, TRUE);
fd98a5c6 529smtp_respond(code, len, TRUE, user_msg);
61147df4
PP
530}
531#endif
532
533
534
535
fd98a5c6
JH
536
537/*************************************************
059ec3d9
PH
538* Remove a recipient from the list *
539*************************************************/
540
541/* This function is provided for local_scan() to use.
542
543Argument:
544 recipient address to remove
545
546Returns: TRUE if it did remove something; FALSE otherwise
547*/
548
549BOOL
550receive_remove_recipient(uschar *recipient)
551{
552int count;
553DEBUG(D_receive) debug_printf("receive_remove_recipient(\"%s\") called\n",
554 recipient);
555for (count = 0; count < recipients_count; count++)
556 {
557 if (Ustrcmp(recipients_list[count].address, recipient) == 0)
558 {
559 if ((--recipients_count - count) > 0)
560 memmove(recipients_list + count, recipients_list + count + 1,
54cdb463 561 (recipients_count - count)*sizeof(recipient_item));
059ec3d9
PH
562 return TRUE;
563 }
564 }
565return FALSE;
566}
567
568
569
570
571
572/*************************************************
573* Read data portion of a non-SMTP message *
574*************************************************/
575
576/* This function is called to read the remainder of a message (following the
577header) when the input is not from SMTP - we are receiving a local message on
578a standard input stream. The message is always terminated by EOF, and is also
579terminated by a dot on a line by itself if the flag dot_ends is TRUE. Split the
580two cases for maximum efficiency.
581
582Ensure that the body ends with a newline. This will naturally be the case when
583the termination is "\n.\n" but may not be otherwise. The RFC defines messages
584as "sequences of lines" - this of course strictly applies only to SMTP, but
585deliveries into BSD-type mailbox files also require it. Exim used to have a
586flag for doing this at delivery time, but as it was always set for all
587transports, I decided to simplify things by putting the check here instead.
588
589There is at least one MUA (dtmail) that sends CRLF via this interface, and
590other programs are known to do this as well. Exim used to have a option for
591dealing with this: in July 2003, after much discussion, the code has been
592changed to default to treat any of LF, CRLF, and bare CR as line terminators.
593
594However, for the case when a dot on a line by itself terminates a message, the
595only recognized terminating sequences before and after the dot are LF and CRLF.
596Otherwise, having read EOL . CR, you don't know whether to read another
597character or not.
598
599Internally, in messages stored in Exim's spool files, LF is used as the line
600terminator. Under the new regime, bare CRs will no longer appear in these
601files.
602
603Arguments:
604 fout a FILE to which to write the message
605
606Returns: One of the END_xxx values indicating why it stopped reading
607*/
608
609static int
610read_message_data(FILE *fout)
611{
612int ch_state;
613register int ch;
d677b2f2 614register int linelength = 0;
059ec3d9
PH
615
616/* Handle the case when only EOF terminates the message */
617
618if (!dot_ends)
619 {
620 register int last_ch = '\n';
621
80a47a2c 622 for (; (ch = (receive_getc)()) != EOF; last_ch = ch)
059ec3d9
PH
623 {
624 if (ch == 0) body_zerocount++;
625 if (last_ch == '\r' && ch != '\n')
626 {
d677b2f2
PH
627 if (linelength > max_received_linelength)
628 max_received_linelength = linelength;
629 linelength = 0;
059ec3d9
PH
630 if (fputc('\n', fout) == EOF) return END_WERROR;
631 message_size++;
632 body_linecount++;
633 }
634 if (ch == '\r') continue;
635
636 if (fputc(ch, fout) == EOF) return END_WERROR;
d677b2f2
PH
637 if (ch == '\n')
638 {
639 if (linelength > max_received_linelength)
640 max_received_linelength = linelength;
641 linelength = 0;
642 body_linecount++;
643 }
644 else linelength++;
059ec3d9
PH
645 if (++message_size > thismessage_size_limit) return END_SIZE;
646 }
647
648 if (last_ch != '\n')
649 {
d677b2f2
PH
650 if (linelength > max_received_linelength)
651 max_received_linelength = linelength;
059ec3d9
PH
652 if (fputc('\n', fout) == EOF) return END_WERROR;
653 message_size++;
654 body_linecount++;
655 }
656
657 return END_EOF;
658 }
659
660/* Handle the case when a dot on a line on its own, or EOF, terminates. */
661
662ch_state = 1;
663
80a47a2c 664while ((ch = (receive_getc)()) != EOF)
059ec3d9
PH
665 {
666 if (ch == 0) body_zerocount++;
667 switch (ch_state)
668 {
669 case 0: /* Normal state (previous char written) */
670 if (ch == '\n')
d677b2f2
PH
671 {
672 body_linecount++;
673 if (linelength > max_received_linelength)
674 max_received_linelength = linelength;
675 linelength = -1;
676 ch_state = 1;
677 }
059ec3d9
PH
678 else if (ch == '\r')
679 { ch_state = 2; continue; }
680 break;
681
682 case 1: /* After written "\n" */
683 if (ch == '.') { ch_state = 3; continue; }
6eb02f88 684 if (ch == '\r') { ch_state = 2; continue; }
d677b2f2 685 if (ch != '\n') ch_state = 0; else linelength = -1;
059ec3d9
PH
686 break;
687
688 case 2:
689 body_linecount++; /* After unwritten "\r" */
d677b2f2
PH
690 if (linelength > max_received_linelength)
691 max_received_linelength = linelength;
059ec3d9 692 if (ch == '\n')
d677b2f2
PH
693 {
694 ch_state = 1;
695 linelength = -1;
696 }
059ec3d9
PH
697 else
698 {
699 if (message_size++, fputc('\n', fout) == EOF) return END_WERROR;
700 if (ch == '\r') continue;
701 ch_state = 0;
d677b2f2 702 linelength = 0;
059ec3d9
PH
703 }
704 break;
705
706 case 3: /* After "\n." (\n written, dot not) */
707 if (ch == '\n') return END_DOT;
708 if (ch == '\r') { ch_state = 4; continue; }
709 message_size++;
d677b2f2 710 linelength++;
059ec3d9
PH
711 if (fputc('.', fout) == EOF) return END_WERROR;
712 ch_state = 0;
713 break;
714
715 case 4: /* After "\n.\r" (\n written, rest not) */
716 if (ch == '\n') return END_DOT;
717 message_size += 2;
718 body_linecount++;
719 if (fputs(".\n", fout) == EOF) return END_WERROR;
720 if (ch == '\r') { ch_state = 2; continue; }
721 ch_state = 0;
722 break;
723 }
724
d677b2f2 725 linelength++;
059ec3d9
PH
726 if (fputc(ch, fout) == EOF) return END_WERROR;
727 if (++message_size > thismessage_size_limit) return END_SIZE;
728 }
729
730/* Get here if EOF read. Unless we have just written "\n", we need to ensure
731the message ends with a newline, and we must also write any characters that
732were saved up while testing for an ending dot. */
733
734if (ch_state != 1)
735 {
736 static uschar *ends[] = { US"\n", NULL, US"\n", US".\n", US".\n" };
737 if (fputs(CS ends[ch_state], fout) == EOF) return END_WERROR;
738 message_size += Ustrlen(ends[ch_state]);
739 body_linecount++;
740 }
741
742return END_EOF;
743}
744
745
746
747
748/*************************************************
749* Read data portion of an SMTP message *
750*************************************************/
751
752/* This function is called to read the remainder of an SMTP message (after the
753headers), or to skip over it when an error has occurred. In this case, the
754output file is passed as NULL.
755
756If any line begins with a dot, that character is skipped. The input should only
757be successfully terminated by CR LF . CR LF unless it is local (non-network)
758SMTP, in which case the CRs are optional, but...
759
760FUDGE: It seems that sites on the net send out messages with just LF
761terminators, despite the warnings in the RFCs, and other MTAs handle this. So
762we make the CRs optional in all cases.
763
764July 2003: Bare CRs cause trouble. We now treat them as line terminators as
765well, so that there are no CRs in spooled messages. However, the message
766terminating dot is not recognized between two bare CRs.
767
768Arguments:
769 fout a FILE to which to write the message; NULL if skipping
770
771Returns: One of the END_xxx values indicating why it stopped reading
772*/
773
774static int
775read_message_data_smtp(FILE *fout)
776{
777int ch_state = 0;
e4bdf652 778int ch;
1f5497b2 779register int linelength = 0;
059ec3d9 780
80a47a2c 781while ((ch = (receive_getc)()) != EOF)
059ec3d9
PH
782 {
783 if (ch == 0) body_zerocount++;
784 switch (ch_state)
785 {
786 case 0: /* After LF or CRLF */
787 if (ch == '.')
788 {
789 ch_state = 3;
790 continue; /* Don't ever write . after LF */
791 }
792 ch_state = 1;
793
794 /* Else fall through to handle as normal uschar. */
795
796 case 1: /* Normal state */
797 if (ch == '\n')
798 {
799 ch_state = 0;
800 body_linecount++;
1f5497b2
PH
801 if (linelength > max_received_linelength)
802 max_received_linelength = linelength;
803 linelength = -1;
059ec3d9
PH
804 }
805 else if (ch == '\r')
806 {
807 ch_state = 2;
808 continue;
809 }
810 break;
811
812 case 2: /* After (unwritten) CR */
813 body_linecount++;
1f5497b2
PH
814 if (linelength > max_received_linelength)
815 max_received_linelength = linelength;
816 linelength = -1;
059ec3d9
PH
817 if (ch == '\n')
818 {
819 ch_state = 0;
820 }
821 else
822 {
823 message_size++;
824 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
e4bdf652 825 (void) cutthrough_put_nl();
059ec3d9
PH
826 if (ch != '\r') ch_state = 1; else continue;
827 }
828 break;
829
830 case 3: /* After [CR] LF . */
831 if (ch == '\n')
832 return END_DOT;
833 if (ch == '\r')
834 {
835 ch_state = 4;
836 continue;
837 }
1bc460a6
JH
838 /* The dot was removed at state 3. For a doubled dot, here, reinstate
839 it to cutthrough. The current ch, dot or not, is passed both to cutthrough
840 and to file below. */
841 if (ch == '.')
842 {
843 uschar c= ch;
844 (void) cutthrough_puts(&c, 1);
845 }
846 ch_state = 1;
059ec3d9
PH
847 break;
848
849 case 4: /* After [CR] LF . CR */
850 if (ch == '\n') return END_DOT;
851 message_size++;
852 body_linecount++;
853 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
e4bdf652 854 (void) cutthrough_put_nl();
059ec3d9
PH
855 if (ch == '\r')
856 {
857 ch_state = 2;
858 continue;
859 }
860 ch_state = 1;
861 break;
862 }
863
864 /* Add the character to the spool file, unless skipping; then loop for the
865 next. */
866
867 message_size++;
1f5497b2 868 linelength++;
059ec3d9
PH
869 if (fout != NULL)
870 {
871 if (fputc(ch, fout) == EOF) return END_WERROR;
872 if (message_size > thismessage_size_limit) return END_SIZE;
873 }
e4bdf652
JH
874 if(ch == '\n')
875 (void) cutthrough_put_nl();
876 else
877 {
878 uschar c= ch;
879 (void) cutthrough_puts(&c, 1);
880 }
059ec3d9
PH
881 }
882
883/* Fall through here if EOF encountered. This indicates some kind of error,
884since a correct message is terminated by [CR] LF . [CR] LF. */
885
886return END_EOF;
887}
888
889
890
891
892/*************************************************
893* Swallow SMTP message *
894*************************************************/
895
896/* This function is called when there has been some kind of error while reading
897an SMTP message, and the remaining data may need to be swallowed. It is global
898because it is called from smtp_closedown() to shut down an incoming call
899tidily.
900
901Argument: a FILE from which to read the message
902Returns: nothing
903*/
904
905void
906receive_swallow_smtp(void)
907{
908if (message_ended >= END_NOTENDED)
909 message_ended = read_message_data_smtp(NULL);
910}
911
912
913
914/*************************************************
915* Handle lost SMTP connection *
916*************************************************/
917
918/* This function logs connection loss incidents and generates an appropriate
919SMTP response.
920
921Argument: additional data for the message
922Returns: the SMTP response
923*/
924
925static uschar *
926handle_lost_connection(uschar *s)
927{
928log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
929 "%s lost while reading message data%s", smtp_get_connection_info(), s);
930return US"421 Lost incoming connection";
931}
932
933
934
935
936/*************************************************
937* Handle a non-smtp reception error *
938*************************************************/
939
940/* This function is called for various errors during the reception of non-SMTP
941messages. It either sends a message to the sender of the problem message, or it
942writes to the standard error stream.
943
944Arguments:
945 errcode code for moan_to_sender(), identifying the error
946 text1 first message text, passed to moan_to_sender()
947 text2 second message text, used only for stderrr
948 error_rc code to pass to exim_exit if no problem
949 f FILE containing body of message (may be stdin)
950 hptr pointer to instore headers or NULL
951
952Returns: calls exim_exit(), which does not return
953*/
954
955static void
956give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
957 FILE *f, header_line *hptr)
958{
959if (error_handling == ERRORS_SENDER)
960 {
961 error_block eblock;
962 eblock.next = NULL;
963 eblock.text1 = text1;
37f3dc43 964 eblock.text2 = US"";
059ec3d9
PH
965 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
966 error_rc = EXIT_FAILURE;
967 }
37f3dc43
JH
968else
969 fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 970(void)fclose(f);
059ec3d9
PH
971exim_exit(error_rc);
972}
973
974
975
976/*************************************************
977* Add header lines set up by ACL *
978*************************************************/
979
850635b6
PH
980/* This function is called to add the header lines that were set up by
981statements in an ACL to the list of headers in memory. It is done in two stages
982like this, because when the ACL for RCPT is running, the other headers have not
983yet been received. This function is called twice; once just before running the
984DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
985are visible to the DATA ACL.
059ec3d9
PH
986
987Originally these header lines were added at the end. Now there is support for
988three different places: top, bottom, and after the Received: header(s). There
989will always be at least one Received: header, even if it is marked deleted, and
990even if something else has been put in front of it.
991
992Arguments:
993 acl_name text to identify which ACL
994
995Returns: nothing
996*/
997
998static void
578d43dc 999add_acl_headers(int where, uschar *acl_name)
059ec3d9
PH
1000{
1001header_line *h, *next;
1002header_line *last_received = NULL;
e7568d51 1003
578d43dc
JH
1004switch(where)
1005 {
1006 case ACL_WHERE_DKIM:
1007 case ACL_WHERE_MIME:
af4a1bca 1008 case ACL_WHERE_DATA:
5032d1cf 1009 if (cutthrough.fd >= 0 && (acl_removed_headers || acl_added_headers))
578d43dc
JH
1010 {
1011 log_write(0, LOG_MAIN|LOG_PANIC, "Header modification in data ACLs"
af4a1bca 1012 " will not take effect on cutthrough deliveries");
578d43dc
JH
1013 return;
1014 }
1015 }
1016
e7568d51
TL
1017if (acl_removed_headers != NULL)
1018 {
1019 DEBUG(D_receive|D_acl) debug_printf(">>Headers removed by %s ACL:\n", acl_name);
1020
4a142059 1021 for (h = header_list; h != NULL; h = h->next) if (h->type != htype_old)
e7568d51 1022 {
55414b25 1023 const uschar * list = acl_removed_headers;
e7568d51
TL
1024 int sep = ':'; /* This is specified as a colon-separated list */
1025 uschar *s;
1026 uschar buffer[128];
4a142059
JH
1027
1028 while ((s = string_nextinlist(&list, &sep, buffer, sizeof(buffer))))
1029 if (header_testname(h, s, Ustrlen(s), FALSE))
e7568d51
TL
1030 {
1031 h->type = htype_old;
1032 DEBUG(D_receive|D_acl) debug_printf(" %s", h->text);
1033 }
e7568d51
TL
1034 }
1035 acl_removed_headers = NULL;
1036 DEBUG(D_receive|D_acl) debug_printf(">>\n");
1037 }
059ec3d9 1038
71fafd95 1039if (acl_added_headers == NULL) return;
059ec3d9
PH
1040DEBUG(D_receive|D_acl) debug_printf(">>Headers added by %s ACL:\n", acl_name);
1041
71fafd95 1042for (h = acl_added_headers; h != NULL; h = next)
059ec3d9
PH
1043 {
1044 next = h->next;
1045
1046 switch(h->type)
1047 {
1048 case htype_add_top:
1049 h->next = header_list;
1050 header_list = h;
1051 DEBUG(D_receive|D_acl) debug_printf(" (at top)");
1052 break;
1053
1054 case htype_add_rec:
1055 if (last_received == NULL)
1056 {
1057 last_received = header_list;
1058 while (!header_testname(last_received, US"Received", 8, FALSE))
1059 last_received = last_received->next;
1060 while (last_received->next != NULL &&
1061 header_testname(last_received->next, US"Received", 8, FALSE))
1062 last_received = last_received->next;
1063 }
1064 h->next = last_received->next;
1065 last_received->next = h;
1066 DEBUG(D_receive|D_acl) debug_printf(" (after Received:)");
1067 break;
1068
8523533c
TK
1069 case htype_add_rfc:
1070 /* add header before any header which is NOT Received: or Resent- */
1071 last_received = header_list;
1072 while ( (last_received->next != NULL) &&
1073 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
1074 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
1075 last_received = last_received->next;
1076 /* last_received now points to the last Received: or Resent-* header
1077 in an uninterrupted chain of those header types (seen from the beginning
1078 of all headers. Our current header must follow it. */
1079 h->next = last_received->next;
1080 last_received->next = h;
8e669ac1 1081 DEBUG(D_receive|D_acl) debug_printf(" (before any non-Received: or Resent-*: header)");
8523533c
TK
1082 break;
1083
059ec3d9
PH
1084 default:
1085 h->next = NULL;
1086 header_last->next = h;
1087 break;
1088 }
1089
1090 if (h->next == NULL) header_last = h;
1091
1092 /* Check for one of the known header types (From:, To:, etc.) though in
1093 practice most added headers are going to be "other". Lower case
1094 identification letters are never stored with the header; they are used
1095 for existence tests when messages are received. So discard any lower case
1096 flag values. */
1097
1098 h->type = header_checkname(h, FALSE);
1099 if (h->type >= 'a') h->type = htype_other;
1100
1101 DEBUG(D_receive|D_acl) debug_printf(" %s", header_last->text);
1102 }
1103
71fafd95 1104acl_added_headers = NULL;
059ec3d9
PH
1105DEBUG(D_receive|D_acl) debug_printf(">>\n");
1106}
1107
1108
1109
1110/*************************************************
1111* Add host information for log line *
1112*************************************************/
1113
1114/* Called for acceptance and rejecting log lines. This adds information about
1115the calling host to a string that is being built dynamically.
1116
1117Arguments:
1118 s the dynamic string
1119 sizeptr points to the size variable
1120 ptrptr points to the pointer variable
1121
1122Returns: the extended string
1123*/
1124
1125static uschar *
1126add_host_info_for_log(uschar *s, int *sizeptr, int *ptrptr)
1127{
1128if (sender_fullhost != NULL)
1129 {
1130 s = string_append(s, sizeptr, ptrptr, 2, US" H=", sender_fullhost);
6c6d6e48 1131 if (LOGGING(incoming_interface) && interface_address != NULL)
059ec3d9
PH
1132 {
1133 uschar *ss = string_sprintf(" I=[%s]:%d", interface_address,
1134 interface_port);
c2f669a4 1135 s = string_cat(s, sizeptr, ptrptr, ss);
059ec3d9
PH
1136 }
1137 }
1138if (sender_ident != NULL)
1139 s = string_append(s, sizeptr, ptrptr, 2, US" U=", sender_ident);
1140if (received_protocol != NULL)
1141 s = string_append(s, sizeptr, ptrptr, 2, US" P=", received_protocol);
1142return s;
1143}
1144
1145
1146
63955bf2 1147#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1148
1149/*************************************************
54cdb463
PH
1150* Run the MIME ACL on a message *
1151*************************************************/
1152
1153/* This code is in a subroutine so that it can be used for both SMTP
1154and non-SMTP messages. It is called with a non-NULL ACL pointer.
1155
1156Arguments:
1157 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1158 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1159 smtp_reply_ptr Where SMTP reply is being built
1160 blackholed_by_ptr Where "blackholed by" message is being built
1161
1162Returns: TRUE to carry on; FALSE to abandon the message
1163*/
1164
1165static BOOL
1166run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1167 uschar **blackholed_by_ptr)
1168{
1169FILE *mbox_file;
1170uschar rfc822_file_path[2048];
1171unsigned long mbox_size;
1172header_line *my_headerlist;
1173uschar *user_msg, *log_msg;
1174int mime_part_count_buffer = -1;
7156b1ef 1175int rc = OK;
54cdb463
PH
1176
1177memset(CS rfc822_file_path,0,2048);
1178
1179/* check if it is a MIME message */
1180my_headerlist = header_list;
4e88a19f
PH
1181while (my_headerlist != NULL)
1182 {
54cdb463 1183 /* skip deleted headers */
4e88a19f
PH
1184 if (my_headerlist->type == '*')
1185 {
54cdb463
PH
1186 my_headerlist = my_headerlist->next;
1187 continue;
4e88a19f
PH
1188 }
1189 if (strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0)
1190 {
54cdb463
PH
1191 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1192 goto DO_MIME_ACL;
4e88a19f 1193 }
54cdb463 1194 my_headerlist = my_headerlist->next;
4e88a19f 1195 }
54cdb463
PH
1196
1197DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1198return TRUE;
1199
1200DO_MIME_ACL:
1201/* make sure the eml mbox file is spooled up */
8544e77a 1202mbox_file = spool_mbox(&mbox_size, NULL);
54cdb463
PH
1203if (mbox_file == NULL) {
1204 /* error while spooling */
1205 log_write(0, LOG_MAIN|LOG_PANIC,
1206 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1207 Uunlink(spool_name);
1208 unspool_mbox();
6f0c431a
PP
1209#ifdef EXPERIMENTAL_DCC
1210 dcc_ok = 0;
1211#endif
a5bd321b 1212 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1213 message_id[0] = 0; /* Indicate no message accepted */
1214 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1215 return FALSE; /* Indicate skip to end of receive function */
1216};
1217
1218mime_is_rfc822 = 0;
1219
1220MIME_ACL_CHECK:
1221mime_part_count = -1;
1222rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1223(void)fclose(mbox_file);
54cdb463 1224
4e88a19f
PH
1225if (Ustrlen(rfc822_file_path) > 0)
1226 {
54cdb463
PH
1227 mime_part_count = mime_part_count_buffer;
1228
4e88a19f
PH
1229 if (unlink(CS rfc822_file_path) == -1)
1230 {
54cdb463
PH
1231 log_write(0, LOG_PANIC,
1232 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1233 goto END_MIME_ACL;
4e88a19f
PH
1234 }
1235 }
54cdb463
PH
1236
1237/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1238if (rc == OK)
1239 {
54cdb463 1240 uschar temp_path[1024];
e8bc7fca
JH
1241 struct dirent * entry;
1242 DIR * tempdir;
54cdb463 1243
e8bc7fca
JH
1244 (void) string_format(temp_path, sizeof(temp_path), "%s/scan/%s",
1245 spool_directory, message_id);
54cdb463 1246
4e88a19f 1247 tempdir = opendir(CS temp_path);
e8bc7fca 1248 for (;;)
4e88a19f 1249 {
e8bc7fca
JH
1250 if (!(entry = readdir(tempdir)))
1251 break;
1252 if (strncmpic(US entry->d_name, US"__rfc822_", 9) == 0)
4e88a19f 1253 {
e8bc7fca
JH
1254 (void) string_format(rfc822_file_path, sizeof(rfc822_file_path),
1255 "%s/scan/%s/%s", spool_directory, message_id, entry->d_name);
1256 debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
1257 rfc822_file_path);
4e88a19f
PH
1258 break;
1259 }
e8bc7fca 1260 }
4e88a19f 1261 closedir(tempdir);
54cdb463 1262
e8bc7fca 1263 if (entry)
4e88a19f 1264 {
e8bc7fca 1265 if ((mbox_file = Ufopen(rfc822_file_path, "rb")))
4e88a19f 1266 {
e8bc7fca
JH
1267 /* set RFC822 expansion variable */
1268 mime_is_rfc822 = 1;
1269 mime_part_count_buffer = mime_part_count;
1270 goto MIME_ACL_CHECK;
4e88a19f 1271 }
e8bc7fca
JH
1272 log_write(0, LOG_PANIC,
1273 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1274 unlink(CS rfc822_file_path);
4e88a19f
PH
1275 }
1276 }
54cdb463
PH
1277
1278END_MIME_ACL:
578d43dc 1279add_acl_headers(ACL_WHERE_MIME, US"MIME");
54cdb463
PH
1280if (rc == DISCARD)
1281 {
1282 recipients_count = 0;
1283 *blackholed_by_ptr = US"MIME ACL";
1284 }
1285else if (rc != OK)
1286 {
1287 Uunlink(spool_name);
1288 unspool_mbox();
6f0c431a
PP
1289#ifdef EXPERIMENTAL_DCC
1290 dcc_ok = 0;
1291#endif
4f6ae5c3
JH
1292 if ( smtp_input
1293 && smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1294 {
85ffcba6 1295 *smtp_yield_ptr = FALSE; /* No more messages after dropped connection */
f4c1088b 1296 *smtp_reply_ptr = US""; /* Indicate reply already sent */
4f6ae5c3 1297 }
54cdb463
PH
1298 message_id[0] = 0; /* Indicate no message accepted */
1299 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1300 }
54cdb463
PH
1301
1302return TRUE;
1303}
1304
63955bf2 1305#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1306
1307
e4bdf652
JH
1308
1309void
1310received_header_gen(void)
1311{
1312uschar *received;
1313uschar *timestamp;
1314header_line *received_header= header_list;
1315
1316timestamp = expand_string(US"${tod_full}");
1317if (recipients_count == 1) received_for = recipients_list[0].address;
1318received = expand_string(received_header_text);
1319received_for = NULL;
1320
d4ff61d1 1321if (!received)
e4bdf652
JH
1322 {
1323 if(spool_name[0] != 0)
1324 Uunlink(spool_name); /* Lose the data file */
1325 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1326 "(received_header_text) failed: %s", string_printing(received_header_text),
1327 expand_string_message);
1328 }
1329
1330/* The first element on the header chain is reserved for the Received header,
1331so all we have to do is fill in the text pointer, and set the type. However, if
1332the result of the expansion is an empty string, we leave the header marked as
1333"old" so as to refrain from adding a Received header. */
1334
1335if (received[0] == 0)
1336 {
1337 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1338 received_header->type = htype_old;
1339 }
1340else
1341 {
1342 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1343 received_header->type = htype_received;
1344 }
1345
1346received_header->slen = Ustrlen(received_header->text);
1347
1348DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1349 received_header->type, received_header->text);
1350}
1351
1352
1353
54cdb463 1354/*************************************************
059ec3d9
PH
1355* Receive message *
1356*************************************************/
1357
1358/* Receive a message on the given input, and put it into a pair of spool files.
1359Either a non-null list of recipients, or the extract flag will be true, or
1360both. The flag sender_local is true for locally generated messages. The flag
1361submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1362suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1363suppress_local_fixups" or -G was passed on the command-line.
1364The flag smtp_input is true if the message is to be
8800895a
PH
1365handled using SMTP conventions about termination and lines starting with dots.
1366For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1367
1368If a message was successfully read, message_id[0] will be non-zero.
1369
1370The general actions of this function are:
1371
1372 . Read the headers of the message (if any) into a chain of store
1373 blocks.
1374
1375 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1376 throw it away, unless the caller is trusted, or unless
1377 active_local_sender_retain is set - which can only happen if
1378 active_local_from_check is false.
059ec3d9
PH
1379
1380 . If recipients are to be extracted from the message, build the
1381 recipients list from the headers, removing any that were on the
1382 original recipients list (unless extract_addresses_remove_arguments is
1383 false), and at the same time, remove any bcc header that may be present.
1384
1385 . Get the spool file for the data, sort out its unique name, open
1386 and lock it (but don't give it the name yet).
1387
1388 . Generate a "Message-Id" header if the message doesn't have one, for
1389 locally-originated messages.
1390
1391 . Generate a "Received" header.
1392
1393 . Ensure the recipients list is fully qualified and rewritten if necessary.
1394
1395 . If there are any rewriting rules, apply them to the sender address
1396 and also to the headers.
1397
1398 . If there is no from: header, generate one, for locally-generated messages
1399 and messages in "submission mode" only.
1400
1401 . If the sender is local, check that from: is correct, and if not, generate
1402 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1403 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1404
1405 . If there is no "date" header, generate one, for locally-originated
1406 or submission mode messages only.
1407
1408 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1409 dot_ends mode, to the data file. Leave it open, to hold the lock.
1410
1411 . Write the envelope and the headers to a new file.
1412
1413 . Set the name for the header file; close it.
1414
1415 . Set the name for the data file; close it.
1416
1417Because this function can potentially be called many times in a single
1418SMTP connection, all store should be got by store_get(), so that it will be
1419automatically retrieved after the message is accepted.
1420
1421FUDGE: It seems that sites on the net send out messages with just LF
1422terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1423we make the CRs optional in all cases.
1424
1425July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1426new regime is now in place in which bare CRs in header lines are turned into LF
1427followed by a space, so as not to terminate the header line.
1428
1429February 2004: A bare LF in a header line in a message whose first line was
1430terminated by CRLF is treated in the same way as a bare CR.
1431
1432Arguments:
1433 extract_recip TRUE if recipients are to be extracted from the message's
1434 headers
1435
1436Returns: TRUE there are more messages to be read (SMTP input)
1437 FALSE there are no more messages to be read (non-SMTP input
1438 or SMTP connection collapsed, or other failure)
1439
1440When reading a message for filter testing, the returned value indicates
1441whether the headers (which is all that is read) were terminated by '.' or
1442not. */
1443
1444BOOL
1445receive_msg(BOOL extract_recip)
1446{
7156b1ef
NM
1447int i;
1448int rc = FAIL;
059ec3d9
PH
1449int msg_size = 0;
1450int process_info_len = Ustrlen(process_info);
1451int error_rc = (error_handling == ERRORS_SENDER)?
1452 errors_sender_rc : EXIT_FAILURE;
1453int header_size = 256;
1454int start, end, domain, size, sptr;
1455int id_resolution;
1456int had_zero = 0;
d677b2f2 1457int prevlines_length = 0;
059ec3d9
PH
1458
1459register int ptr = 0;
1460
1461BOOL contains_resent_headers = FALSE;
1462BOOL extracted_ignored = FALSE;
1463BOOL first_line_ended_crlf = TRUE_UNSET;
1464BOOL smtp_yield = TRUE;
1465BOOL yield = FALSE;
1466
1467BOOL resents_exist = FALSE;
1468uschar *resent_prefix = US"";
1469uschar *blackholed_by = NULL;
04f7d5b9 1470uschar *blackhole_log_msg = US"";
c5430c20 1471enum {NOT_TRIED, TMP_REJ, PERM_REJ, ACCEPTED} cutthrough_done = NOT_TRIED;
059ec3d9
PH
1472
1473flock_t lock_data;
1474error_block *bad_addresses = NULL;
1475
1476uschar *frozen_by = NULL;
1477uschar *queued_by = NULL;
1478
1479uschar *errmsg, *s;
1480struct stat statbuf;
1481
4e88a19f 1482/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1483
1484uschar *smtp_reply = NULL;
4e88a19f 1485uschar *user_msg, *log_msg;
059ec3d9
PH
1486
1487/* Working header pointers */
1488
1489header_line *h, *next;
1490
2cbb4081 1491/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1492
1493BOOL date_header_exists = FALSE;
1494
1495/* Pointers to receive the addresses of headers whose contents we need. */
1496
1497header_line *from_header = NULL;
1498header_line *subject_header = NULL;
1499header_line *msgid_header = NULL;
1500header_line *received_header;
1501
4840604e
TL
1502#ifdef EXPERIMENTAL_DMARC
1503int dmarc_up = 0;
1504#endif /* EXPERIMENTAL_DMARC */
1505
059ec3d9
PH
1506/* Variables for use when building the Received: header. */
1507
059ec3d9
PH
1508uschar *timestamp;
1509int tslen;
1510
1511/* Release any open files that might have been cached while preparing to
1512accept the message - e.g. by verifying addresses - because reading a message
1513might take a fair bit of real time. */
1514
1515search_tidyup();
1516
e4bdf652
JH
1517/* Extracting the recipient list from an input file is incompatible with
1518cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1519to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1520if (extract_recip || !smtp_input)
2e5b33cd 1521 cancel_cutthrough_connection("not smtp input");
e4bdf652 1522
059ec3d9
PH
1523/* Initialize the chain of headers by setting up a place-holder for Received:
1524header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1525pointing to the end of the chain to make adding headers simple. */
1526
1527received_header = header_list = header_last = store_get(sizeof(header_line));
1528header_list->next = NULL;
1529header_list->type = htype_old;
1530header_list->text = NULL;
1531header_list->slen = 0;
1532
1533/* Control block for the next header to be read. */
1534
1535next = store_get(sizeof(header_line));
1536next->text = store_get(header_size);
1537
1538/* Initialize message id to be null (indicating no message read), and the
1539header names list to be the normal list. Indicate there is no data file open
1540yet, initialize the size and warning count, and deal with no size limit. */
1541
1542message_id[0] = 0;
1543data_file = NULL;
1544data_fd = -1;
1545spool_name[0] = 0;
1546message_size = 0;
1547warning_count = 0;
d677b2f2 1548received_count = 1; /* For the one we will add */
059ec3d9
PH
1549
1550if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1551
2e0c1448 1552/* While reading the message, the following counts are computed. */
059ec3d9 1553
d677b2f2
PH
1554message_linecount = body_linecount = body_zerocount =
1555 max_received_linelength = 0;
059ec3d9 1556
80a47a2c
TK
1557#ifndef DISABLE_DKIM
1558/* Call into DKIM to set up the context. */
1559if (smtp_input && !smtp_batched_input && !dkim_disable_verify) dkim_exim_verify_init();
fb2274d4
TK
1560#endif
1561
4840604e
TL
1562#ifdef EXPERIMENTAL_DMARC
1563/* initialize libopendmarc */
1564dmarc_up = dmarc_init();
1565#endif
1566
059ec3d9
PH
1567/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1568ids, and fractions of a second are required. See the comments that precede the
1569message id creation below. */
1570
1571(void)gettimeofday(&message_id_tv, NULL);
1572
1573/* For other uses of the received time we can operate with granularity of one
1574second, and for that we use the global variable received_time. This is for
1575things like ultimate message timeouts. */
1576
1577received_time = message_id_tv.tv_sec;
1578
1579/* If SMTP input, set the special handler for timeouts. The alarm() calls
1580happen in the smtp_getc() function when it refills its buffer. */
1581
1582if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1583
1584/* If not SMTP input, timeout happens only if configured, and we just set a
1585single timeout for the whole message. */
1586
1587else if (receive_timeout > 0)
1588 {
1589 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1590 alarm(receive_timeout);
1591 }
1592
1593/* SIGTERM and SIGINT are caught always. */
1594
1595signal(SIGTERM, data_sigterm_sigint_handler);
1596signal(SIGINT, data_sigterm_sigint_handler);
1597
1598/* Header lines in messages are not supposed to be very long, though when
1599unfolded, to: and cc: headers can take up a lot of store. We must also cope
1600with the possibility of junk being thrown at us. Start by getting 256 bytes for
1601storing the header, and extend this as necessary using string_cat().
1602
1603To cope with total lunacies, impose an upper limit on the length of the header
1604section of the message, as otherwise the store will fill up. We must also cope
1605with the possibility of binary zeros in the data. Hence we cannot use fgets().
1606Folded header lines are joined into one string, leaving the '\n' characters
1607inside them, so that writing them out reproduces the input.
1608
1609Loop for each character of each header; the next structure for chaining the
1610header is set up already, with ptr the offset of the next character in
1611next->text. */
1612
1613for (;;)
1614 {
80a47a2c 1615 int ch = (receive_getc)();
059ec3d9
PH
1616
1617 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1618 SMTP must have a correct "." terminator. */
1619
1620 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1621 {
1622 smtp_reply = handle_lost_connection(US" (header)");
1623 smtp_yield = FALSE;
1624 goto TIDYUP; /* Skip to end of function */
1625 }
1626
1627 /* See if we are at the current header's size limit - there must be at least
1628 four bytes left. This allows for the new character plus a zero, plus two for
1629 extra insertions when we are playing games with dots and carriage returns. If
1630 we are at the limit, extend the text buffer. This could have been done
1631 automatically using string_cat() but because this is a tightish loop storing
1632 only one character at a time, we choose to do it inline. Normally
1633 store_extend() will be able to extend the block; only at the end of a big
1634 store block will a copy be needed. To handle the case of very long headers
1635 (and sometimes lunatic messages can have ones that are 100s of K long) we
1636 call store_release() for strings that have been copied - if the string is at
1637 the start of a block (and therefore the only thing in it, because we aren't
1638 doing any other gets), the block gets freed. We can only do this because we
1639 know there are no other calls to store_get() going on. */
1640
1641 if (ptr >= header_size - 4)
1642 {
1643 int oldsize = header_size;
1644 /* header_size += 256; */
1645 header_size *= 2;
1646 if (!store_extend(next->text, oldsize, header_size))
1647 {
1648 uschar *newtext = store_get(header_size);
1649 memcpy(newtext, next->text, ptr);
1650 store_release(next->text);
1651 next->text = newtext;
1652 }
1653 }
1654
1655 /* Cope with receiving a binary zero. There is dispute about whether
1656 these should be allowed in RFC 822 messages. The middle view is that they
1657 should not be allowed in headers, at least. Exim takes this attitude at
1658 the moment. We can't just stomp on them here, because we don't know that
1659 this line is a header yet. Set a flag to cause scanning later. */
1660
1661 if (ch == 0) had_zero++;
1662
1663 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1664 those from data files use just LF. Treat LF in local SMTP input as a
1665 terminator too. Treat EOF as a line terminator always. */
1666
1667 if (ch == EOF) goto EOL;
1668
1669 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1670 other MTAs accept this. We are therefore forced into this "liberalisation"
1671 too, so we accept LF as a line terminator whatever the source of the message.
1672 However, if the first line of the message ended with a CRLF, we treat a bare
1673 LF specially by inserting a white space after it to ensure that the header
1674 line is not terminated. */
1675
1676 if (ch == '\n')
1677 {
1678 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1679 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1680 goto EOL;
1681 }
1682
1683 /* This is not the end of the line. If this is SMTP input and this is
1684 the first character in the line and it is a "." character, ignore it.
1685 This implements the dot-doubling rule, though header lines starting with
1686 dots aren't exactly common. They are legal in RFC 822, though. If the
1687 following is CRLF or LF, this is the line that that terminates the
1688 entire message. We set message_ended to indicate this has happened (to
1689 prevent further reading), and break out of the loop, having freed the
1690 empty header, and set next = NULL to indicate no data line. */
1691
1692 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1693 {
80a47a2c 1694 ch = (receive_getc)();
059ec3d9
PH
1695 if (ch == '\r')
1696 {
80a47a2c 1697 ch = (receive_getc)();
059ec3d9
PH
1698 if (ch != '\n')
1699 {
80a47a2c 1700 receive_ungetc(ch);
059ec3d9
PH
1701 ch = '\r'; /* Revert to CR */
1702 }
1703 }
1704 if (ch == '\n')
1705 {
1706 message_ended = END_DOT;
1707 store_reset(next);
1708 next = NULL;
1709 break; /* End character-reading loop */
1710 }
1711
1712 /* For non-SMTP input, the dot at the start of the line was really a data
1713 character. What is now in ch is the following character. We guaranteed
1714 enough space for this above. */
1715
1716 if (!smtp_input)
1717 {
1718 next->text[ptr++] = '.';
1719 message_size++;
1720 }
1721 }
1722
1723 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1724 remember this case if this is the first line ending. */
1725
1726 if (ch == '\r')
1727 {
80a47a2c 1728 ch = (receive_getc)();
059ec3d9
PH
1729 if (ch == '\n')
1730 {
1731 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1732 goto EOL;
1733 }
1734
1735 /* Otherwise, put back the character after CR, and turn the bare CR
1736 into LF SP. */
1737
80a47a2c 1738 ch = (receive_ungetc)(ch);
059ec3d9
PH
1739 next->text[ptr++] = '\n';
1740 message_size++;
1741 ch = ' ';
1742 }
1743
1744 /* We have a data character for the header line. */
1745
1746 next->text[ptr++] = ch; /* Add to buffer */
1747 message_size++; /* Total message size so far */
1748
1749 /* Handle failure due to a humungously long header section. The >= allows
1750 for the terminating \n. Add what we have so far onto the headers list so
1751 that it gets reflected in any error message, and back up the just-read
1752 character. */
1753
1754 if (message_size >= header_maxsize)
1755 {
1756 next->text[ptr] = 0;
1757 next->slen = ptr;
1758 next->type = htype_other;
1759 next->next = NULL;
1760 header_last->next = next;
1761 header_last = next;
1762
1763 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1764 "%s (more than %d characters): message abandoned",
1765 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1766
1767 if (smtp_input)
1768 {
1769 smtp_reply = US"552 Message header is ridiculously long";
1770 receive_swallow_smtp();
1771 goto TIDYUP; /* Skip to end of function */
1772 }
1773
1774 else
1775 {
1776 give_local_error(ERRMESS_VLONGHEADER,
1777 string_sprintf("message header longer than %d characters received: "
1778 "message not accepted", header_maxsize), US"", error_rc, stdin,
1779 header_list->next);
1780 /* Does not return */
1781 }
1782 }
1783
1784 continue; /* With next input character */
1785
1786 /* End of header line reached */
1787
1788 EOL:
2e0c1448
PH
1789
1790 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1791
1792 receive_linecount++;
1793 message_linecount++;
059ec3d9 1794
d677b2f2
PH
1795 /* Keep track of maximum line length */
1796
1797 if (ptr - prevlines_length > max_received_linelength)
1798 max_received_linelength = ptr - prevlines_length;
1799 prevlines_length = ptr + 1;
1800
059ec3d9
PH
1801 /* Now put in the terminating newline. There is always space for
1802 at least two more characters. */
1803
1804 next->text[ptr++] = '\n';
1805 message_size++;
1806
1807 /* A blank line signals the end of the headers; release the unwanted
1808 space and set next to NULL to indicate this. */
1809
1810 if (ptr == 1)
1811 {
1812 store_reset(next);
1813 next = NULL;
1814 break;
1815 }
1816
1817 /* There is data in the line; see if the next input character is a
1818 whitespace character. If it is, we have a continuation of this header line.
1819 There is always space for at least one character at this point. */
1820
1821 if (ch != EOF)
1822 {
80a47a2c 1823 int nextch = (receive_getc)();
059ec3d9
PH
1824 if (nextch == ' ' || nextch == '\t')
1825 {
1826 next->text[ptr++] = nextch;
1827 message_size++;
1828 continue; /* Iterate the loop */
1829 }
80a47a2c 1830 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
1831 else ch = EOF; /* Cause main loop to exit at end */
1832 }
1833
1834 /* We have got to the real line end. Terminate the string and release store
1835 beyond it. If it turns out to be a real header, internal binary zeros will
1836 be squashed later. */
1837
1838 next->text[ptr] = 0;
1839 next->slen = ptr;
1840 store_reset(next->text + ptr + 1);
1841
1842 /* Check the running total size against the overall message size limit. We
1843 don't expect to fail here, but if the overall limit is set less than MESSAGE_
1844 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
1845 headers - the code to read the body will then also hit the buffer. */
1846
1847 if (message_size > thismessage_size_limit) break;
1848
1849 /* A line that is not syntactically correct for a header also marks
1850 the end of the headers. In this case, we leave next containing the
1851 first data line. This might actually be several lines because of the
1852 continuation logic applied above, but that doesn't matter.
1853
1854 It turns out that smail, and presumably sendmail, accept leading lines
1855 of the form
1856
1857 From ph10 Fri Jan 5 12:35 GMT 1996
1858
1859 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
1860 find any documentation of this, but for compatibility it had better be
1861 accepted. Exim restricts it to the case of non-smtp messages, and
1862 treats it as an alternative to the -f command line option. Thus it is
1863 ignored except for trusted users or filter testing. Otherwise it is taken
1864 as the sender address, unless -f was used (sendmail compatibility).
1865
1866 It further turns out that some UUCPs generate the From_line in a different
1867 format, e.g.
1868
1869 From ph10 Fri, 7 Jan 97 14:00:00 GMT
1870
1871 The regex for matching these things is now capable of recognizing both
1872 formats (including 2- and 4-digit years in the latter). In fact, the regex
1873 is now configurable, as is the expansion string to fish out the sender.
1874
1875 Even further on it has been discovered that some broken clients send
1876 these lines in SMTP messages. There is now an option to ignore them from
1877 specified hosts or networks. Sigh. */
1878
1879 if (header_last == header_list &&
1880 (!smtp_input
1881 ||
1882 (sender_host_address != NULL &&
1883 verify_check_host(&ignore_fromline_hosts) == OK)
1884 ||
1885 (sender_host_address == NULL && ignore_fromline_local)
1886 ) &&
1887 regex_match_and_setup(regex_From, next->text, 0, -1))
1888 {
1889 if (!sender_address_forced)
1890 {
1891 uschar *uucp_sender = expand_string(uucp_from_sender);
1892 if (uucp_sender == NULL)
1893 {
1894 log_write(0, LOG_MAIN|LOG_PANIC,
1895 "expansion of \"%s\" failed after matching "
1896 "\"From \" line: %s", uucp_from_sender, expand_string_message);
1897 }
1898 else
1899 {
1900 int start, end, domain;
1901 uschar *errmess;
1902 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
1903 &start, &end, &domain, TRUE);
1904 if (newsender != NULL)
1905 {
1906 if (domain == 0 && newsender[0] != 0)
1907 newsender = rewrite_address_qualify(newsender, FALSE);
1908
f05da2e8 1909 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
1910 {
1911 sender_address = newsender;
1912
f05da2e8 1913 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
1914 {
1915 authenticated_sender = NULL;
1916 originator_name = US"";
1917 sender_local = FALSE;
1918 }
1919
f05da2e8 1920 if (filter_test != FTEST_NONE)
059ec3d9
PH
1921 printf("Sender taken from \"From \" line\n");
1922 }
1923 }
1924 }
1925 }
1926 }
1927
1928 /* Not a leading "From " line. Check to see if it is a valid header line.
1929 Header names may contain any non-control characters except space and colon,
1930 amazingly. */
1931
1932 else
1933 {
1934 uschar *p = next->text;
1935
1936 /* If not a valid header line, break from the header reading loop, leaving
1937 next != NULL, indicating that it holds the first line of the body. */
1938
1939 if (isspace(*p)) break;
1940 while (mac_isgraph(*p) && *p != ':') p++;
1941 while (isspace(*p)) p++;
1942 if (*p != ':')
1943 {
1944 body_zerocount = had_zero;
1945 break;
1946 }
1947
1948 /* We have a valid header line. If there were any binary zeroes in
1949 the line, stomp on them here. */
1950
1951 if (had_zero > 0)
1952 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
1953
1954 /* It is perfectly legal to have an empty continuation line
1955 at the end of a header, but it is confusing to humans
1956 looking at such messages, since it looks like a blank line.
1957 Reduce confusion by removing redundant white space at the
1958 end. We know that there is at least one printing character
1959 (the ':' tested for above) so there is no danger of running
1960 off the end. */
1961
1962 p = next->text + ptr - 2;
1963 for (;;)
1964 {
1965 while (*p == ' ' || *p == '\t') p--;
1966 if (*p != '\n') break;
1967 ptr = (p--) - next->text + 1;
1968 message_size -= next->slen - ptr;
1969 next->text[ptr] = 0;
1970 next->slen = ptr;
1971 }
1972
1973 /* Add the header to the chain */
1974
1975 next->type = htype_other;
1976 next->next = NULL;
1977 header_last->next = next;
1978 header_last = next;
1979
1980 /* Check the limit for individual line lengths. This comes after adding to
1981 the chain so that the failing line is reflected if a bounce is generated
1982 (for a local message). */
1983
1984 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
1985 {
1986 log_write(0, LOG_MAIN, "overlong message header line received from "
1987 "%s (more than %d characters): message abandoned",
1988 sender_host_unknown? sender_ident : sender_fullhost,
1989 header_line_maxsize);
1990
1991 if (smtp_input)
1992 {
1993 smtp_reply = US"552 A message header line is too long";
1994 receive_swallow_smtp();
1995 goto TIDYUP; /* Skip to end of function */
1996 }
1997
1998 else
1999 {
2000 give_local_error(ERRMESS_VLONGHDRLINE,
2001 string_sprintf("message header line longer than %d characters "
2002 "received: message not accepted", header_line_maxsize), US"",
2003 error_rc, stdin, header_list->next);
2004 /* Does not return */
2005 }
2006 }
2007
2008 /* Note if any resent- fields exist. */
2009
2010 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
2011 {
2012 resents_exist = TRUE;
2013 resent_prefix = US"Resent-";
2014 }
2015 }
2016
2017 /* The line has been handled. If we have hit EOF, break out of the loop,
2018 indicating no pending data line. */
2019
2020 if (ch == EOF) { next = NULL; break; }
2021
2022 /* Set up for the next header */
2023
2024 header_size = 256;
2025 next = store_get(sizeof(header_line));
2026 next->text = store_get(header_size);
2027 ptr = 0;
2028 had_zero = 0;
d677b2f2 2029 prevlines_length = 0;
059ec3d9
PH
2030 } /* Continue, starting to read the next header */
2031
2032/* At this point, we have read all the headers into a data structure in main
2033store. The first header is still the dummy placeholder for the Received: header
2034we are going to generate a bit later on. If next != NULL, it contains the first
2035data line - which terminated the headers before reaching a blank line (not the
2036normal case). */
2037
2038DEBUG(D_receive)
2039 {
2040 debug_printf(">>Headers received:\n");
2041 for (h = header_list->next; h != NULL; h = h->next)
2042 debug_printf("%s", h->text);
2043 debug_printf("\n");
2044 }
2045
2046/* End of file on any SMTP connection is an error. If an incoming SMTP call
2047is dropped immediately after valid headers, the next thing we will see is EOF.
2048We must test for this specially, as further down the reading of the data is
2049skipped if already at EOF. */
2050
2051if (smtp_input && (receive_feof)())
2052 {
2053 smtp_reply = handle_lost_connection(US" (after header)");
2054 smtp_yield = FALSE;
2055 goto TIDYUP; /* Skip to end of function */
2056 }
2057
2058/* If this is a filter test run and no headers were read, output a warning
2059in case there is a mistake in the test message. */
2060
f05da2e8 2061if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2062 printf("Warning: no message headers read\n");
2063
2064
2065/* Scan the headers to identify them. Some are merely marked for later
2066processing; some are dealt with here. */
2067
2068for (h = header_list->next; h != NULL; h = h->next)
2069 {
2070 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2071 if (is_resent) contains_resent_headers = TRUE;
2072
2073 switch (header_checkname(h, is_resent))
2074 {
059ec3d9 2075 case htype_bcc:
2cbb4081 2076 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
2077 break;
2078
059ec3d9 2079 case htype_cc:
2cbb4081 2080 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
2081 break;
2082
2083 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
2084
2085 case htype_date:
4c69d561 2086 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
2087 break;
2088
2089 /* Same comments as about Return-Path: below. */
2090
2091 case htype_delivery_date:
2092 if (delivery_date_remove) h->type = htype_old;
2093 break;
2094
2095 /* Same comments as about Return-Path: below. */
2096
2097 case htype_envelope_to:
2098 if (envelope_to_remove) h->type = htype_old;
2099 break;
2100
2101 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2102 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2103 header consists of just the login id of the user who called Exim, rewrite
2104 it with the gecos field first. Apply this rule to Resent-From: if there
2105 are resent- fields. */
2106
2107 case htype_from:
2108 h->type = htype_from;
2109 if (!resents_exist || is_resent)
2110 {
2111 from_header = h;
2112 if (!smtp_input)
2113 {
5de8faa3 2114 int len;
059ec3d9
PH
2115 uschar *s = Ustrchr(h->text, ':') + 1;
2116 while (isspace(*s)) s++;
5de8faa3 2117 len = h->slen - (s - h->text) - 1;
e0fccd1d
TF
2118 if (Ustrlen(originator_login) == len &&
2119 strncmpic(s, originator_login, len) == 0)
059ec3d9
PH
2120 {
2121 uschar *name = is_resent? US"Resent-From" : US"From";
2122 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2123 originator_login, qualify_domain_sender);
2124 from_header = header_last;
2125 h->type = htype_old;
2126 DEBUG(D_receive|D_rewrite)
2127 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2128 }
2129 }
2130 }
2131 break;
2132
2133 /* Identify the Message-id: header for generating "in-reply-to" in the
2134 autoreply transport. For incoming logging, save any resent- value. In both
2135 cases, take just the first of any multiples. */
2136
2137 case htype_id:
2138 if (msgid_header == NULL && (!resents_exist || is_resent))
2139 {
2140 msgid_header = h;
2141 h->type = htype_id;
2142 }
2143 break;
2144
2145 /* Flag all Received: headers */
2146
2147 case htype_received:
2148 h->type = htype_received;
2149 received_count++;
2150 break;
2151
2152 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
2153
2154 case htype_reply_to:
2155 h->type = htype_reply_to;
2156 break;
2157
2158 /* The Return-path: header is supposed to be added to messages when
2159 they leave the SMTP system. We shouldn't receive messages that already
2160 contain Return-path. However, since Exim generates Return-path: on
2161 local delivery, resent messages may well contain it. We therefore
2162 provide an option (which defaults on) to remove any Return-path: headers
2163 on input. Removal actually means flagging as "old", which prevents the
2164 header being transmitted with the message. */
2165
2166 case htype_return_path:
2167 if (return_path_remove) h->type = htype_old;
2168
2169 /* If we are testing a mail filter file, use the value of the
2170 Return-Path: header to set up the return_path variable, which is not
2171 otherwise set. However, remove any <> that surround the address
2172 because the variable doesn't have these. */
2173
f05da2e8 2174 if (filter_test != FTEST_NONE)
059ec3d9
PH
2175 {
2176 uschar *start = h->text + 12;
2177 uschar *end = start + Ustrlen(start);
2178 while (isspace(*start)) start++;
2179 while (end > start && isspace(end[-1])) end--;
2180 if (*start == '<' && end[-1] == '>')
2181 {
2182 start++;
2183 end--;
2184 }
2185 return_path = string_copyn(start, end - start);
2186 printf("Return-path taken from \"Return-path:\" header line\n");
2187 }
2188 break;
2189
2190 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2191 and from an untrusted caller and suppress_local_fixups is not set, or if we
2192 are in submission mode for a remote message, mark it "old" so that it will
2193 not be transmitted with the message, unless active_local_sender_retain is
2194 set. (This can only be true if active_local_from_check is false.) If there
2195 are any resent- headers in the message, apply this rule to Resent-Sender:
2196 instead of Sender:. Messages with multiple resent- header sets cannot be
2197 tidily handled. (For this reason, at least one MUA - Pine - turns old
2198 resent- headers into X-resent- headers when resending, leaving just one
2199 set.) */
059ec3d9
PH
2200
2201 case htype_sender:
69358f02 2202 h->type = ((!active_local_sender_retain &&
8800895a
PH
2203 (
2204 (sender_local && !trusted_caller && !suppress_local_fixups)
2205 || submission_mode
2206 )
059ec3d9
PH
2207 ) &&
2208 (!resents_exist||is_resent))?
2209 htype_old : htype_sender;
2210 break;
2211
2212 /* Remember the Subject: header for logging. There is no Resent-Subject */
2213
2214 case htype_subject:
2215 subject_header = h;
2216 break;
2217
2218 /* "To:" gets flagged, and the existence of a recipient header is noted,
2219 whether it's resent- or not. */
2220
2221 case htype_to:
2222 h->type = htype_to;
2223 /****
2224 to_or_cc_header_exists = TRUE;
2225 ****/
2226 break;
2227 }
2228 }
2229
2230/* Extract recipients from the headers if that is required (the -t option).
2231Note that this is documented as being done *before* any address rewriting takes
2232place. There are two possibilities:
2233
2234(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2235recipients already listed are to be REMOVED from the message. Smail 3 works
2236like this. We need to build a non-recipients tree for that list, because in
2237subsequent processing this data is held in a tree and that's what the
2238spool_write_header() function expects. Make sure that non-recipient addresses
2239are fully qualified and rewritten if necessary.
2240
2241(2) According to other sendmail documentation, -t ADDS extracted recipients to
2242those in the command line arguments (and it is rumoured some other MTAs do
2243this). Therefore, there is an option to make Exim behave this way.
2244
2245*** Notes on "Resent-" header lines ***
2246
2247The presence of resent-headers in the message makes -t horribly ambiguous.
2248Experiments with sendmail showed that it uses recipients for all resent-
2249headers, totally ignoring the concept of "sets of resent- headers" as described
2250in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2251with all the addresses in one instance of each header.
2252
2253This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2254error for -t if there were resent- headers in the message. However, after a
2255discussion on the mailing list, I've learned that there are MUAs that use
2256resent- headers with -t, and also that the stuff about sets of resent- headers
2257and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2258message with -t and resent- header lines makes sure that only *its* resent-
2259headers are present; previous ones are often renamed as X-resent- for example.
2260
2261Consequently, Exim has been changed so that, if any resent- header lines are
2262present, the recipients are taken from all of the appropriate resent- lines,
2263and not from the ordinary To:, Cc:, etc. */
2264
2265if (extract_recip)
2266 {
2267 int rcount = 0;
2268 error_block **bnext = &bad_addresses;
2269
2270 if (extract_addresses_remove_arguments)
2271 {
2272 while (recipients_count-- > 0)
2273 {
2274 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2275 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2276 tree_add_nonrecipient(s);
2277 }
2278 recipients_list = NULL;
2279 recipients_count = recipients_list_max = 0;
2280 }
2281
059ec3d9
PH
2282 /* Now scan the headers */
2283
2284 for (h = header_list->next; h != NULL; h = h->next)
2285 {
2286 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2287 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2288 {
2289 uschar *s = Ustrchr(h->text, ':') + 1;
2290 while (isspace(*s)) s++;
2291
1eccaa59
PH
2292 parse_allow_group = TRUE; /* Allow address group syntax */
2293
059ec3d9
PH
2294 while (*s != 0)
2295 {
2296 uschar *ss = parse_find_address_end(s, FALSE);
2297 uschar *recipient, *errmess, *p, *pp;
2298 int start, end, domain;
2299
2300 /* Check on maximum */
2301
2302 if (recipients_max > 0 && ++rcount > recipients_max)
2303 {
2304 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2305 US"message rejected: ", error_rc, stdin, NULL);
2306 /* Does not return */
2307 }
2308
2309 /* Make a copy of the address, and remove any internal newlines. These
2310 may be present as a result of continuations of the header line. The
2311 white space that follows the newline must not be removed - it is part
2312 of the header. */
2313
2314 pp = recipient = store_get(ss - s + 1);
2315 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2316 *pp = 0;
250b6871 2317
8c5d388a 2318#ifdef SUPPORT_I18N
250b6871
JH
2319 {
2320 BOOL b = allow_utf8_domains;
2321 allow_utf8_domains = TRUE;
2322#endif
059ec3d9
PH
2323 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2324 &domain, FALSE);
2325
8c5d388a 2326#ifdef SUPPORT_I18N
250b6871
JH
2327 if (string_is_utf8(recipient))
2328 message_smtputf8 = TRUE;
2329 else
2330 allow_utf8_domains = b;
2331 }
2332#endif
2333
059ec3d9
PH
2334 /* Keep a list of all the bad addresses so we can send a single
2335 error message at the end. However, an empty address is not an error;
2336 just ignore it. This can come from an empty group list like
2337
2338 To: Recipients of list:;
2339
2340 If there are no recipients at all, an error will occur later. */
2341
2342 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2343 {
2344 int len = Ustrlen(s);
2345 error_block *b = store_get(sizeof(error_block));
2346 while (len > 0 && isspace(s[len-1])) len--;
2347 b->next = NULL;
2348 b->text1 = string_printing(string_copyn(s, len));
2349 b->text2 = errmess;
2350 *bnext = b;
2351 bnext = &(b->next);
2352 }
2353
2354 /* If the recipient is already in the nonrecipients tree, it must
2355 have appeared on the command line with the option extract_addresses_
2356 remove_arguments set. Do not add it to the recipients, and keep a note
2357 that this has happened, in order to give a better error if there are
2358 no recipients left. */
2359
2360 else if (recipient != NULL)
2361 {
2362 if (tree_search(tree_nonrecipients, recipient) == NULL)
2363 receive_add_recipient(recipient, -1);
2364 else
2365 extracted_ignored = TRUE;
2366 }
2367
2368 /* Move on past this address */
2369
2370 s = ss + (*ss? 1:0);
2371 while (isspace(*s)) s++;
1eccaa59
PH
2372 } /* Next address */
2373
2374 parse_allow_group = FALSE; /* Reset group syntax flags */
2375 parse_found_group = FALSE;
059ec3d9
PH
2376
2377 /* If this was the bcc: header, mark it "old", which means it
2378 will be kept on the spool, but not transmitted as part of the
2379 message. */
2380
2cbb4081 2381 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2382 } /* For appropriate header line */
2383 } /* For each header line */
2384
059ec3d9
PH
2385 }
2386
2387/* Now build the unique message id. This has changed several times over the
2388lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2389Retaining all the history in the comment has become too unwieldy - read
2390previous release sources if you want it.
2391
2392The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2393The first part is the current time, in seconds. The second part is the current
2394pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2395can hold a number in the range 0-3843. It used to be a computed sequence
2396number, but is now the fractional component of the current time in units of
23971/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2398received, Exim ensures that the timer has ticked at the appropriate level
2399before proceeding, to avoid duplication if the pid happened to be re-used
2400within the same time period. It seems likely that most messages will take at
2401least half a millisecond to be received, so no delay will normally be
2402necessary. At least for some time...
2403
2404There is a modification when localhost_number is set. Formerly this was allowed
2405to be as large as 255. Now it is restricted to the range 0-16, and the final
2406component of the message id becomes (localhost_number * 200) + fractional time
2407in units of 1/200 of a second (i.e. a value in the range 0-3399).
2408
2409Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2410Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2411still allows the tttttt field to hold a large enough number to last for some
2412more decades, and the final two-digit field can hold numbers up to 1295, which
2413is enough for milliseconds (instead of 1/2000 of a second).
2414
2415However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2416pid, so it is probably safe because pids have to be positive. The
2417localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2418final field becomes (localhost_number * 100) + fractional time in centiseconds.
2419
2420Note that string_base62() returns its data in a static storage block, so it
2421must be copied before calling string_base62() again. It always returns exactly
24226 characters.
2423
2424There doesn't seem to be anything in the RFC which requires a message id to
2425start with a letter, but Smail was changed to ensure this. The external form of
2426the message id (as supplied by string expansion) therefore starts with an
2427additional leading 'E'. The spool file names do not include this leading
2428letter and it is not used internally.
2429
2430NOTE: If ever the format of message ids is changed, the regular expression for
2431checking that a string is in this format must be updated in a corresponding
2432way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2433must also be changed to reflect the correct string length. Then, of course,
2434other programs that rely on the message id format will need updating too. */
2435
2436Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2437message_id[6] = '-';
2438Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2439
2440/* Deal with the case where the host number is set. The value of the number was
2441checked when it was read, to ensure it isn't too big. The timing granularity is
2442left in id_resolution so that an appropriate wait can be done after receiving
2443the message, if necessary (we hope it won't be). */
2444
2445if (host_number_string != NULL)
2446 {
2447 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2448 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2449 string_base62((long int)(
2450 host_number * (1000000/id_resolution) +
2451 message_id_tv.tv_usec/id_resolution)) + 4);
2452 }
2453
2454/* Host number not set: final field is just the fractional time at an
2455appropriate resolution. */
2456
2457else
2458 {
2459 id_resolution = (BASE_62 == 62)? 500 : 1000;
2460 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2461 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2462 }
2463
2464/* Add the current message id onto the current process info string if
2465it will fit. */
2466
2467(void)string_format(process_info + process_info_len,
2468 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2469
2470/* If we are using multiple input directories, set up the one for this message
2471to be the least significant base-62 digit of the time of arrival. Otherwise
2472ensure that it is an empty string. */
2473
a2da3176 2474message_subdir[0] = split_spool_directory ? message_id[5] : 0;
059ec3d9
PH
2475
2476/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2477one, but only for local (without suppress_local_fixups) or submission mode
2478messages. This can be user-configured if required, but we had better flatten
2479any illegal characters therein. */
059ec3d9 2480
8800895a
PH
2481if (msgid_header == NULL &&
2482 ((sender_host_address == NULL && !suppress_local_fixups)
2483 || submission_mode))
059ec3d9
PH
2484 {
2485 uschar *p;
2486 uschar *id_text = US"";
2487 uschar *id_domain = primary_hostname;
2488
2489 /* Permit only letters, digits, dots, and hyphens in the domain */
2490
2491 if (message_id_domain != NULL)
2492 {
2493 uschar *new_id_domain = expand_string(message_id_domain);
2494 if (new_id_domain == NULL)
2495 {
2496 if (!expand_string_forcedfail)
2497 log_write(0, LOG_MAIN|LOG_PANIC,
2498 "expansion of \"%s\" (message_id_header_domain) "
2499 "failed: %s", message_id_domain, expand_string_message);
2500 }
2501 else if (*new_id_domain != 0)
2502 {
2503 id_domain = new_id_domain;
2504 for (p = id_domain; *p != 0; p++)
2505 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2506 }
2507 }
2508
2509 /* Permit all characters except controls and RFC 2822 specials in the
2510 additional text part. */
2511
2512 if (message_id_text != NULL)
2513 {
2514 uschar *new_id_text = expand_string(message_id_text);
2515 if (new_id_text == NULL)
2516 {
2517 if (!expand_string_forcedfail)
2518 log_write(0, LOG_MAIN|LOG_PANIC,
2519 "expansion of \"%s\" (message_id_header_text) "
2520 "failed: %s", message_id_text, expand_string_message);
2521 }
2522 else if (*new_id_text != 0)
2523 {
2524 id_text = new_id_text;
2525 for (p = id_text; *p != 0; p++)
2526 if (mac_iscntrl_or_special(*p)) *p = '-';
2527 }
2528 }
2529
e7e680d6
PP
2530 /* Add the header line
2531 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2532 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2533
e7e680d6 2534 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2535 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2536 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2537 }
2538
2539/* If we are to log recipients, keep a copy of the raw ones before any possible
2540rewriting. Must copy the count, because later ACLs and the local_scan()
2541function may mess with the real recipients. */
2542
6c6d6e48 2543if (LOGGING(received_recipients))
059ec3d9
PH
2544 {
2545 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2546 for (i = 0; i < recipients_count; i++)
2547 raw_recipients[i] = string_copy(recipients_list[i].address);
2548 raw_recipients_count = recipients_count;
2549 }
2550
2551/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2552recipients will get here only if the conditions were right (allow_unqualified_
2553recipient is TRUE). */
2554
2555for (i = 0; i < recipients_count; i++)
2556 recipients_list[i].address =
2557 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2558 global_rewrite_rules, rewrite_existflags);
2559
8800895a
PH
2560/* If there is no From: header, generate one for local (without
2561suppress_local_fixups) or submission_mode messages. If there is no sender
2562address, but the sender is local or this is a local delivery error, use the
2563originator login. This shouldn't happen for genuine bounces, but might happen
2564for autoreplies. The addition of From: must be done *before* checking for the
2565possible addition of a Sender: header, because untrusted_set_sender allows an
2566untrusted user to set anything in the envelope (which might then get info
2567From:) but we still want to ensure a valid Sender: if it is required. */
2568
2569if (from_header == NULL &&
2570 ((sender_host_address == NULL && !suppress_local_fixups)
2571 || submission_mode))
059ec3d9 2572 {
2fe1a124
PH
2573 uschar *oname = US"";
2574
2575 /* Use the originator_name if this is a locally submitted message and the
2576 caller is not trusted. For trusted callers, use it only if -F was used to
2577 force its value or if we have a non-SMTP message for which -f was not used
2578 to set the sender. */
2579
2580 if (sender_host_address == NULL)
2581 {
2582 if (!trusted_caller || sender_name_forced ||
2583 (!smtp_input && !sender_address_forced))
2584 oname = originator_name;
2585 }
2586
2587 /* For non-locally submitted messages, the only time we use the originator
2588 name is when it was forced by the /name= option on control=submission. */
2589
2590 else
2591 {
2592 if (submission_name != NULL) oname = submission_name;
2593 }
2594
059ec3d9
PH
2595 /* Envelope sender is empty */
2596
2597 if (sender_address[0] == 0)
2598 {
87ba3f5f
PH
2599 uschar *fromstart, *fromend;
2600
2601 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2602 oname, (oname[0] == 0)? "" : " <");
2603 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2604
059ec3d9
PH
2605 if (sender_local || local_error_message)
2606 {
87ba3f5f
PH
2607 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2608 local_part_quote(originator_login), qualify_domain_sender,
2609 fromend);
059ec3d9
PH
2610 }
2611 else if (submission_mode && authenticated_id != NULL)
2612 {
2613 if (submission_domain == NULL)
2614 {
87ba3f5f
PH
2615 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2616 local_part_quote(authenticated_id), qualify_domain_sender,
2617 fromend);
059ec3d9
PH
2618 }
2619 else if (submission_domain[0] == 0) /* empty => whole address set */
2620 {
87ba3f5f
PH
2621 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2622 fromend);
059ec3d9
PH
2623 }
2624 else
2625 {
87ba3f5f
PH
2626 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2627 local_part_quote(authenticated_id), submission_domain,
2628 fromend);
059ec3d9
PH
2629 }
2630 from_header = header_last; /* To get it checked for Sender: */
2631 }
2632 }
2633
2634 /* There is a non-null envelope sender. Build the header using the original
2635 sender address, before any rewriting that might have been done while
2636 verifying it. */
2637
2638 else
2639 {
87ba3f5f 2640 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2641 oname,
2642 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2643 (sender_address_unrewritten == NULL)?
2644 sender_address : sender_address_unrewritten,
2fe1a124 2645 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2646
2647 from_header = header_last; /* To get it checked for Sender: */
2648 }
2649 }
2650
2651
8800895a
PH
2652/* If the sender is local (without suppress_local_fixups), or if we are in
2653submission mode and there is an authenticated_id, check that an existing From:
2654is correct, and if not, generate a Sender: header, unless disabled. Any
2655previously-existing Sender: header was removed above. Note that sender_local,
2656as well as being TRUE if the caller of exim is not trusted, is also true if a
2657trusted caller did not supply a -f argument for non-smtp input. To allow
2658trusted callers to forge From: without supplying -f, we have to test explicitly
2659here. If the From: header contains more than one address, then the call to
2660parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2661
2662if (from_header != NULL &&
69358f02 2663 (active_local_from_check &&
8800895a 2664 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2665 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2666 ))
2667 {
2668 BOOL make_sender = TRUE;
2669 int start, end, domain;
2670 uschar *errmess;
2671 uschar *from_address =
2672 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2673 &start, &end, &domain, FALSE);
2674 uschar *generated_sender_address;
2675
2676 if (submission_mode)
2677 {
2678 if (submission_domain == NULL)
2679 {
2680 generated_sender_address = string_sprintf("%s@%s",
2681 local_part_quote(authenticated_id), qualify_domain_sender);
2682 }
2683 else if (submission_domain[0] == 0) /* empty => full address */
2684 {
2685 generated_sender_address = string_sprintf("%s",
2686 authenticated_id);
2687 }
2688 else
2689 {
2690 generated_sender_address = string_sprintf("%s@%s",
2691 local_part_quote(authenticated_id), submission_domain);
2692 }
2693 }
2694 else
2695 generated_sender_address = string_sprintf("%s@%s",
2696 local_part_quote(originator_login), qualify_domain_sender);
2697
2698 /* Remove permitted prefixes and suffixes from the local part of the From:
2699 address before doing the comparison with the generated sender. */
2700
2701 if (from_address != NULL)
2702 {
2703 int slen;
2704 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2705
2706 if (at != NULL) *at = 0;
2707 from_address += route_check_prefix(from_address, local_from_prefix);
2708 slen = route_check_suffix(from_address, local_from_suffix);
2709 if (slen > 0)
2710 {
2711 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2712 from_address += slen;
2713 }
2714 if (at != NULL) *at = '@';
2715
2716 if (strcmpic(generated_sender_address, from_address) == 0 ||
2717 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2718 make_sender = FALSE;
2719 }
2720
2721 /* We have to cause the Sender header to be rewritten if there are
2722 appropriate rewriting rules. */
2723
2724 if (make_sender)
2725 {
2fe1a124 2726 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2727 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2728 generated_sender_address);
2729 else
2730 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2731 resent_prefix,
2732 submission_mode? submission_name : originator_name,
2733 generated_sender_address);
059ec3d9 2734 }
87ba3f5f
PH
2735
2736 /* Ensure that a non-null envelope sender address corresponds to the
2737 submission mode sender address. */
2738
2739 if (submission_mode && sender_address[0] != 0)
2740 {
2741 if (sender_address_unrewritten == NULL)
2742 sender_address_unrewritten = sender_address;
2743 sender_address = generated_sender_address;
089793a4
TF
2744 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2745 log_write(L_address_rewrite, LOG_MAIN,
2746 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2747 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2748 }
059ec3d9
PH
2749 }
2750
059ec3d9
PH
2751/* If there are any rewriting rules, apply them to the sender address, unless
2752it has already been rewritten as part of verification for SMTP input. */
2753
2754if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2755 sender_address[0] != 0)
2756 {
2757 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2758 global_rewrite_rules, rewrite_existflags);
2759 DEBUG(D_receive|D_rewrite)
2760 debug_printf("rewritten sender = %s\n", sender_address);
2761 }
2762
2763
2764/* The headers must be run through rewrite_header(), because it ensures that
2765addresses are fully qualified, as well as applying any rewriting rules that may
2766exist.
2767
2768Qualification of header addresses in a message from a remote host happens only
2769if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2770appropriate. For local messages, qualification always happens, unless -bnq is
2771used to explicitly suppress it. No rewriting is done for an unqualified address
2772that is left untouched.
2773
2774We start at the second header, skipping our own Received:. This rewriting is
2775documented as happening *after* recipient addresses are taken from the headers
2776by the -t command line option. An added Sender: gets rewritten here. */
2777
2778for (h = header_list->next; h != NULL; h = h->next)
2779 {
2780 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2781 rewrite_existflags, TRUE);
2782 if (newh != NULL) h = newh;
2783 }
2784
2785
2786/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2787"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2788"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2789exists, this applies to the set of resent- headers rather than the normal set.
2790
2cbb4081
PH
2791The requirement for a recipient header has been removed in RFC 2822. At this
2792point in the code, earlier versions of Exim added a To: header for locally
2793submitted messages, and an empty Bcc: header for others. In the light of the
2794changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2795
059ec3d9
PH
2796
2797/* If there is no date header, generate one if the message originates locally
8800895a
PH
2798(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2799submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2800to be more confusing if Exim adds one to all remotely-originated messages.
2801As per Message-Id, we prepend if resending, else append.
2802*/
059ec3d9 2803
8800895a
PH
2804if (!date_header_exists &&
2805 ((sender_host_address == NULL && !suppress_local_fixups)
2806 || submission_mode))
e7e680d6
PP
2807 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2808 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2809
2810search_tidyup(); /* Free any cached resources */
2811
2812/* Show the complete set of headers if debugging. Note that the first one (the
2813new Received:) has not yet been set. */
2814
2815DEBUG(D_receive)
2816 {
2817 debug_printf(">>Headers after rewriting and local additions:\n");
2818 for (h = header_list->next; h != NULL; h = h->next)
2819 debug_printf("%c %s", h->type, h->text);
2820 debug_printf("\n");
2821 }
2822
2823/* The headers are now complete in store. If we are running in filter
2824testing mode, that is all this function does. Return TRUE if the message
2825ended with a dot. */
2826
f05da2e8 2827if (filter_test != FTEST_NONE)
059ec3d9
PH
2828 {
2829 process_info[process_info_len] = 0;
2830 return message_ended == END_DOT;
2831 }
2832
817d9f57 2833/* Cutthrough delivery:
5032d1cf
JH
2834We have to create the Received header now rather than at the end of reception,
2835so the timestamp behaviour is a change to the normal case.
2836XXX Ensure this gets documented XXX.
2837Having created it, send the headers to the destination. */
2838if (cutthrough.fd >= 0)
e4bdf652 2839 {
817d9f57
JH
2840 if (received_count > received_headers_max)
2841 {
2e5b33cd 2842 cancel_cutthrough_connection("too many headers");
817d9f57
JH
2843 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2844 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2845 "Too many \"Received\" headers",
2846 sender_address,
2847 (sender_fullhost == NULL)? "" : " H=",
2848 (sender_fullhost == NULL)? US"" : sender_fullhost,
2849 (sender_ident == NULL)? "" : " U=",
2850 (sender_ident == NULL)? US"" : sender_ident);
2851 message_id[0] = 0; /* Indicate no message accepted */
2852 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
2853 goto TIDYUP; /* Skip to end of function */
2854 }
e4bdf652 2855 received_header_gen();
578d43dc 2856 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652
JH
2857 (void) cutthrough_headers_send();
2858 }
61147df4 2859
e4bdf652 2860
059ec3d9
PH
2861/* Open a new spool file for the data portion of the message. We need
2862to access it both via a file descriptor and a stream. Try to make the
2863directory if it isn't there. Note re use of sprintf: spool_directory
2864is checked on input to be < 200 characters long. */
2865
a2da3176
JH
2866snprintf(CS spool_name, sizeof(spool_name), "%s/input/%s/%s/%s-D",
2867 spool_directory, queue_name, message_subdir, message_id);
2868DEBUG(D_receive) debug_printf("Data file name: %s\n", spool_name);
2869
2870if ((data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE)) < 0)
059ec3d9
PH
2871 {
2872 if (errno == ENOENT)
2873 {
a2da3176
JH
2874 uschar * temp = string_sprintf("input%s%s%s%s",
2875 *queue_name ? "/" : "", queue_name,
2876 *message_subdir ? "/" : "", message_subdir);
059ec3d9
PH
2877 (void)directory_make(spool_directory, temp, INPUT_DIRECTORY_MODE, TRUE);
2878 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2879 }
2880 if (data_fd < 0)
2881 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
2882 spool_name, strerror(errno));
2883 }
2884
2885/* Make sure the file's group is the Exim gid, and double-check the mode
2886because the group setting doesn't always get set automatically. */
2887
1ac6b2e7
JH
2888if (fchown(data_fd, exim_uid, exim_gid))
2889 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
2890 "Failed setting ownership on spool file %s: %s",
2891 spool_name, strerror(errno));
ff790e47 2892(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
2893
2894/* We now have data file open. Build a stream for it and lock it. We lock only
2895the first line of the file (containing the message ID) because otherwise there
2896are problems when Exim is run under Cygwin (I'm told). See comments in
2897spool_in.c, where the same locking is done. */
2898
2899data_file = fdopen(data_fd, "w+");
2900lock_data.l_type = F_WRLCK;
2901lock_data.l_whence = SEEK_SET;
2902lock_data.l_start = 0;
2903lock_data.l_len = SPOOL_DATA_START_OFFSET;
2904
2905if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
2906 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
2907 errno, strerror(errno));
2908
2909/* We have an open, locked data file. Write the message id to it to make it
2910self-identifying. Then read the remainder of the input of this message and
2911write it to the data file. If the variable next != NULL, it contains the first
2912data line (which was read as a header but then turned out not to have the right
2913format); write it (remembering that it might contain binary zeros). The result
2914of fwrite() isn't inspected; instead we call ferror() below. */
2915
2916fprintf(data_file, "%s-D\n", message_id);
2917if (next != NULL)
2918 {
2919 uschar *s = next->text;
2920 int len = next->slen;
1ac6b2e7 2921 len = fwrite(s, 1, len, data_file); len = len; /* compiler quietening */
059ec3d9
PH
2922 body_linecount++; /* Assumes only 1 line */
2923 }
2924
2925/* Note that we might already be at end of file, or the logical end of file
2926(indicated by '.'), or might have encountered an error while writing the
2927message id or "next" line. */
2928
2929if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
2930 {
2931 if (smtp_input)
2932 {
2933 message_ended = read_message_data_smtp(data_file);
2934 receive_linecount++; /* The terminating "." line */
2935 }
2936 else message_ended = read_message_data(data_file);
2937
2938 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 2939 message_linecount += body_linecount;
059ec3d9
PH
2940
2941 /* Handle premature termination of SMTP */
2942
2943 if (smtp_input && message_ended == END_EOF)
2944 {
2945 Uunlink(spool_name); /* Lose data file when closed */
2e5b33cd 2946 cancel_cutthrough_connection("sender closed connection");
059ec3d9
PH
2947 message_id[0] = 0; /* Indicate no message accepted */
2948 smtp_reply = handle_lost_connection(US"");
2949 smtp_yield = FALSE;
2950 goto TIDYUP; /* Skip to end of function */
2951 }
2952
2953 /* Handle message that is too big. Don't use host_or_ident() in the log
2954 message; we want to see the ident value even for non-remote messages. */
2955
2956 if (message_ended == END_SIZE)
2957 {
2958 Uunlink(spool_name); /* Lose the data file when closed */
2e5b33cd 2959 cancel_cutthrough_connection("mail too big");
059ec3d9
PH
2960 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2961
2962 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2963 "message too big: read=%d max=%d",
2964 sender_address,
2965 (sender_fullhost == NULL)? "" : " H=",
2966 (sender_fullhost == NULL)? US"" : sender_fullhost,
2967 (sender_ident == NULL)? "" : " U=",
2968 (sender_ident == NULL)? US"" : sender_ident,
2969 message_size,
2970 thismessage_size_limit);
2971
2972 if (smtp_input)
2973 {
2974 smtp_reply = US"552 Message size exceeds maximum permitted";
2975 message_id[0] = 0; /* Indicate no message accepted */
2976 goto TIDYUP; /* Skip to end of function */
2977 }
2978 else
2979 {
2980 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2981 give_local_error(ERRMESS_TOOBIG,
2982 string_sprintf("message too big (max=%d)", thismessage_size_limit),
2983 US"message rejected: ", error_rc, data_file, header_list);
2984 /* Does not return */
2985 }
2986 }
2987 }
2988
2989/* Restore the standard SIGALRM handler for any subsequent processing. (For
2990example, there may be some expansion in an ACL that uses a timer.) */
2991
2992os_non_restarting_signal(SIGALRM, sigalrm_handler);
2993
2994/* The message body has now been read into the data file. Call fflush() to
2995empty the buffers in C, and then call fsync() to get the data written out onto
2996the disk, as fflush() doesn't do this (or at least, it isn't documented as
2997having to do this). If there was an I/O error on either input or output,
2998attempt to send an error message, and unlink the spool file. For non-SMTP input
2999we can then give up. Note that for SMTP input we must swallow the remainder of
3000the input in cases of output errors, since the far end doesn't expect to see
3001anything until the terminating dot line is sent. */
3002
3003if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 3004 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
3005 {
3006 uschar *msg_errno = US strerror(errno);
3007 BOOL input_error = (receive_ferror)() != 0;
3008 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
3009 input_error? "Input read" : "Spool write",
3010 msg_errno,
3011 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
3012
3013 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
3014 Uunlink(spool_name); /* Lose the data file */
2e5b33cd 3015 cancel_cutthrough_connection("error writing spoolfile");
059ec3d9
PH
3016
3017 if (smtp_input)
3018 {
3019 if (input_error)
3020 smtp_reply = US"451 Error while reading input data";
3021 else
3022 {
3023 smtp_reply = US"451 Error while writing spool file";
3024 receive_swallow_smtp();
3025 }
3026 message_id[0] = 0; /* Indicate no message accepted */
3027 goto TIDYUP; /* Skip to end of function */
3028 }
3029
3030 else
3031 {
3032 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3033 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
3034 header_list);
3035 /* Does not return */
3036 }
3037 }
3038
3039
3040/* No I/O errors were encountered while writing the data file. */
3041
3042DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
3043
3044
3045/* If there were any bad addresses extracted by -t, or there were no recipients
3046left after -t, send a message to the sender of this message, or write it to
3047stderr if the error handling option is set that way. Note that there may
3048legitimately be no recipients for an SMTP message if they have all been removed
3049by "discard".
3050
3051We need to rewind the data file in order to read it. In the case of no
3052recipients or stderr error writing, throw the data file away afterwards, and
3053exit. (This can't be SMTP, which always ensures there's at least one
3054syntactically good recipient address.) */
3055
3056if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
3057 {
3058 DEBUG(D_receive)
3059 {
3060 if (recipients_count == 0) debug_printf("*** No recipients\n");
3061 if (bad_addresses != NULL)
3062 {
3063 error_block *eblock = bad_addresses;
3064 debug_printf("*** Bad address(es)\n");
3065 while (eblock != NULL)
3066 {
3067 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
3068 eblock = eblock->next;
3069 }
3070 }
3071 }
3072
3073 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3074
3075 /* If configured to send errors to the sender, but this fails, force
3076 a failure error code. We use a special one for no recipients so that it
3077 can be detected by the autoreply transport. Otherwise error_rc is set to
3078 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
3079 it is EXIT_SUCCESS. */
3080
3081 if (error_handling == ERRORS_SENDER)
3082 {
3083 if (!moan_to_sender(
3084 (bad_addresses == NULL)?
3085 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
3086 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
3087 bad_addresses, header_list, data_file, FALSE))
3088 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
3089 }
3090 else
3091 {
3092 if (bad_addresses == NULL)
3093 {
3094 if (extracted_ignored)
3095 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3096 else
3097 fprintf(stderr, "exim: no recipients in message\n");
3098 }
3099 else
3100 {
3101 fprintf(stderr, "exim: invalid address%s",
3102 (bad_addresses->next == NULL)? ":" : "es:\n");
3103 while (bad_addresses != NULL)
3104 {
3105 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3106 bad_addresses->text2);
3107 bad_addresses = bad_addresses->next;
3108 }
3109 }
3110 }
3111
3112 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3113 {
3114 Uunlink(spool_name);
f1e894f3 3115 (void)fclose(data_file);
059ec3d9
PH
3116 exim_exit(error_rc);
3117 }
3118 }
3119
3120/* Data file successfully written. Generate text for the Received: header by
3121expanding the configured string, and adding a timestamp. By leaving this
3122operation till now, we ensure that the timestamp is the time that message
3123reception was completed. However, this is deliberately done before calling the
3124data ACL and local_scan().
3125
3126This Received: header may therefore be inspected by the data ACL and by code in
3127the local_scan() function. When they have run, we update the timestamp to be
3128the final time of reception.
3129
3130If there is just one recipient, set up its value in the $received_for variable
3131for use when we generate the Received: header.
3132
3133Note: the checking for too many Received: headers is handled by the delivery
3134code. */
e4bdf652 3135/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3136
e4bdf652 3137if (received_header->text == NULL) /* Non-cutthrough case */
059ec3d9 3138 {
e4bdf652 3139 received_header_gen();
059ec3d9 3140
e4bdf652 3141 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3142
e4bdf652
JH
3143 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3144 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3145
e4bdf652
JH
3146 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3147 now, before running the DATA ACL. */
059ec3d9 3148
578d43dc 3149 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652 3150 }
817d9f57 3151else
e4bdf652
JH
3152 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3153 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3154
3155/* If an ACL is specified for checking things at this stage of reception of a
3156message, run it, unless all the recipients were removed by "discard" in earlier
3157ACLs. That is the only case in which recipients_count can be zero at this
3158stage. Set deliver_datafile to point to the data file so that $message_body and
3159$message_body_end can be extracted if needed. Allow $recipients in expansions.
3160*/
3161
3162deliver_datafile = data_fd;
4e88a19f 3163user_msg = NULL;
059ec3d9 3164
0e20aff9
MH
3165enable_dollar_recipients = TRUE;
3166
059ec3d9
PH
3167if (recipients_count == 0)
3168 {
3169 blackholed_by = recipients_discarded? US"MAIL ACL" : US"RCPT ACL";
3170 }
3171else
3172 {
059ec3d9
PH
3173 /* Handle interactive SMTP messages */
3174
3175 if (smtp_input && !smtp_batched_input)
3176 {
8523533c 3177
80a47a2c
TK
3178#ifndef DISABLE_DKIM
3179 if (!dkim_disable_verify)
3180 {
3181 /* Finish verification, this will log individual signature results to
3182 the mainlog */
3183 dkim_exim_verify_finish();
3184
3185 /* Check if we must run the DKIM ACL */
3186 if ((acl_smtp_dkim != NULL) &&
3187 (dkim_verify_signers != NULL) &&
3188 (dkim_verify_signers[0] != '\0'))
3189 {
3190 uschar *dkim_verify_signers_expanded =
3191 expand_string(dkim_verify_signers);
3192 if (dkim_verify_signers_expanded == NULL)
3193 {
3194 log_write(0, LOG_MAIN|LOG_PANIC,
3195 "expansion of dkim_verify_signers option failed: %s",
3196 expand_string_message);
3197 }
3198 else
3199 {
3200 int sep = 0;
55414b25 3201 const uschar *ptr = dkim_verify_signers_expanded;
80a47a2c 3202 uschar *item = NULL;
9e5d6b55
TK
3203 uschar *seen_items = NULL;
3204 int seen_items_size = 0;
3205 int seen_items_offset = 0;
80a47a2c 3206 uschar itembuf[256];
9122af94
TK
3207 /* Default to OK when no items are present */
3208 rc = OK;
80a47a2c
TK
3209 while ((item = string_nextinlist(&ptr, &sep,
3210 itembuf,
5032d1cf 3211 sizeof(itembuf))))
80a47a2c 3212 {
6119d1ea
TK
3213 /* Prevent running ACL for an empty item */
3214 if (!item || (item[0] == '\0')) continue;
5032d1cf
JH
3215
3216 /* Only run ACL once for each domain or identity,
3217 no matter how often it appears in the expanded list. */
3218 if (seen_items)
6119d1ea 3219 {
ae9094bf
TK
3220 uschar *seen_item = NULL;
3221 uschar seen_item_buf[256];
55414b25 3222 const uschar *seen_items_list = seen_items;
5032d1cf 3223 BOOL seen_this_item = FALSE;
61147df4 3224
ae9094bf
TK
3225 while ((seen_item = string_nextinlist(&seen_items_list, &sep,
3226 seen_item_buf,
5032d1cf
JH
3227 sizeof(seen_item_buf))))
3228 if (Ustrcmp(seen_item,item) == 0)
3229 {
3230 seen_this_item = TRUE;
3231 break;
3232 }
3233
3234 if (seen_this_item)
6119d1ea
TK
3235 {
3236 DEBUG(D_receive)
5032d1cf
JH
3237 debug_printf("acl_smtp_dkim: skipping signer %s, "
3238 "already seen\n", item);
6119d1ea
TK
3239 continue;
3240 }
61147df4 3241
5032d1cf
JH
3242 seen_items = string_append(seen_items, &seen_items_size,
3243 &seen_items_offset, 1, ":");
6119d1ea
TK
3244 }
3245
5032d1cf
JH
3246 seen_items = string_append(seen_items, &seen_items_size,
3247 &seen_items_offset, 1, item);
4a73449b 3248 seen_items[seen_items_offset] = '\0';
6119d1ea
TK
3249
3250 DEBUG(D_receive)
5032d1cf
JH
3251 debug_printf("calling acl_smtp_dkim for dkim_cur_signer=%s\n",
3252 item);
6119d1ea 3253
80a47a2c 3254 dkim_exim_acl_setup(item);
5032d1cf
JH
3255 rc = acl_check(ACL_WHERE_DKIM, NULL, acl_smtp_dkim,
3256 &user_msg, &log_msg);
6119d1ea
TK
3257
3258 if (rc != OK)
5032d1cf
JH
3259 {
3260 DEBUG(D_receive)
3261 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, "
3262 "skipping remaining items\n", rc, item);
3263 cancel_cutthrough_connection("dkim acl not ok");
3264 break;
3265 }
80a47a2c 3266 }
578d43dc 3267 add_acl_headers(ACL_WHERE_DKIM, US"DKIM");
80a47a2c
TK
3268 if (rc == DISCARD)
3269 {
3270 recipients_count = 0;
3271 blackholed_by = US"DKIM ACL";
3272 if (log_msg != NULL)
3273 blackhole_log_msg = string_sprintf(": %s", log_msg);
3274 }
3275 else if (rc != OK)
3276 {
3277 Uunlink(spool_name);
3278 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
85ffcba6 3279 smtp_yield = FALSE; /* No more messages after dropped connection */
80a47a2c
TK
3280 smtp_reply = US""; /* Indicate reply already sent */
3281 message_id[0] = 0; /* Indicate no message accepted */
3282 goto TIDYUP; /* Skip to end of function */
3283 }
3284 }
3285 }
3286 }
4a8ce2d8 3287#endif /* DISABLE_DKIM */
fb2274d4 3288
8523533c 3289#ifdef WITH_CONTENT_SCAN
80a47a2c
TK
3290 if (recipients_count > 0 &&
3291 acl_smtp_mime != NULL &&
54cdb463
PH
3292 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
3293 goto TIDYUP;
8523533c
TK
3294#endif /* WITH_CONTENT_SCAN */
3295
4840604e
TL
3296#ifdef EXPERIMENTAL_DMARC
3297 dmarc_up = dmarc_store_data(from_header);
3298#endif /* EXPERIMENTAL_DMARC */
3299
8ccd00b1
JH
3300#ifndef DISABLE_PRDR
3301 if (prdr_requested && recipients_count > 1 && acl_smtp_data_prdr)
fd98a5c6
JH
3302 {
3303 unsigned int c;
3304 int all_pass = OK;
3305 int all_fail = FAIL;
3306
3307 smtp_printf("353 PRDR content analysis beginning\r\n");
3308 /* Loop through recipients, responses must be in same order received */
3309 for (c = 0; recipients_count > c; c++)
3310 {
3311 uschar * addr= recipients_list[c].address;
3312 uschar * msg= US"PRDR R=<%s> %s";
3313 uschar * code;
3314 DEBUG(D_receive)
3315 debug_printf("PRDR processing recipient %s (%d of %d)\n",
3316 addr, c+1, recipients_count);
3317 rc = acl_check(ACL_WHERE_PRDR, addr,
3318 acl_smtp_data_prdr, &user_msg, &log_msg);
3319
3320 /* If any recipient rejected content, indicate it in final message */
3321 all_pass |= rc;
3322 /* If all recipients rejected, indicate in final message */
3323 all_fail &= rc;
3324
3325 switch (rc)
3326 {
3327 case OK: case DISCARD: code = US"250"; break;
3328 case DEFER: code = US"450"; break;
3329 default: code = US"550"; break;
3330 }
3331 if (user_msg != NULL)
3332 smtp_user_msg(code, user_msg);
3333 else
3334 {
3335 switch (rc)
3336 {
3337 case OK: case DISCARD:
3338 msg = string_sprintf(CS msg, addr, "acceptance"); break;
3339 case DEFER:
3340 msg = string_sprintf(CS msg, addr, "temporary refusal"); break;
3341 default:
3342 msg = string_sprintf(CS msg, addr, "refusal"); break;
3343 }
3344 smtp_user_msg(code, msg);
3345 }
3346 if (log_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, log_msg);
3347 else if (user_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, user_msg);
112b6a93 3348 else log_write(0, LOG_MAIN, "%s", CS msg);
fd98a5c6
JH
3349
3350 if (rc != OK) { receive_remove_recipient(addr); c--; }
3351 }
3352 /* Set up final message, used if data acl gives OK */
3353 smtp_reply = string_sprintf("%s id=%s message %s",
3354 all_fail == FAIL ? US"550" : US"250",
3355 message_id,
3356 all_fail == FAIL
3357 ? US"rejected for all recipients"
3358 : all_pass == OK
3359 ? US"accepted"
3360 : US"accepted for some recipients");
3361 if (recipients_count == 0)
3362 {
3363 message_id[0] = 0; /* Indicate no message accepted */
3364 goto TIDYUP;
3365 }
3366 }
3367 else
3368 prdr_requested = FALSE;
8ccd00b1 3369#endif /* !DISABLE_PRDR */
fd98a5c6 3370
54cdb463
PH
3371 /* Check the recipients count again, as the MIME ACL might have changed
3372 them. */
8523533c 3373
059ec3d9
PH
3374 if (acl_smtp_data != NULL && recipients_count > 0)
3375 {
059ec3d9 3376 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
578d43dc 3377 add_acl_headers(ACL_WHERE_DATA, US"DATA");
059ec3d9
PH
3378 if (rc == DISCARD)
3379 {
3380 recipients_count = 0;
3381 blackholed_by = US"DATA ACL";
8e669ac1
PH
3382 if (log_msg != NULL)
3383 blackhole_log_msg = string_sprintf(": %s", log_msg);
2e5b33cd 3384 cancel_cutthrough_connection("data acl discard");
059ec3d9
PH
3385 }
3386 else if (rc != OK)
3387 {
3388 Uunlink(spool_name);
2e5b33cd 3389 cancel_cutthrough_connection("data acl not ok");
8523533c
TK
3390#ifdef WITH_CONTENT_SCAN
3391 unspool_mbox();
3392#endif
6f0c431a
PP
3393#ifdef EXPERIMENTAL_DCC
3394 dcc_ok = 0;
3395#endif
059ec3d9 3396 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
85ffcba6 3397 smtp_yield = FALSE; /* No more messages after dropped connection */
059ec3d9
PH
3398 smtp_reply = US""; /* Indicate reply already sent */
3399 message_id[0] = 0; /* Indicate no message accepted */
3400 goto TIDYUP; /* Skip to end of function */
3401 }
3402 }
3403 }
3404
3405 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
3406 we cannot take different actions for permanent and temporary rejections. */
3407
54cdb463 3408 else
059ec3d9 3409 {
54cdb463
PH
3410
3411#ifdef WITH_CONTENT_SCAN
3412 if (acl_not_smtp_mime != NULL &&
3413 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
3414 &blackholed_by))
3415 goto TIDYUP;
3416#endif /* WITH_CONTENT_SCAN */
3417
3418 if (acl_not_smtp != NULL)
059ec3d9 3419 {
54cdb463
PH
3420 uschar *user_msg, *log_msg;
3421 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
3422 if (rc == DISCARD)
059ec3d9 3423 {
54cdb463
PH
3424 recipients_count = 0;
3425 blackholed_by = US"non-SMTP ACL";
3426 if (log_msg != NULL)
3427 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 3428 }
54cdb463 3429 else if (rc != OK)
059ec3d9 3430 {
54cdb463
PH
3431 Uunlink(spool_name);
3432#ifdef WITH_CONTENT_SCAN
3433 unspool_mbox();
3434#endif
6f0c431a
PP
3435#ifdef EXPERIMENTAL_DCC
3436 dcc_ok = 0;
3437#endif
6ea85e9a
PH
3438 /* The ACL can specify where rejections are to be logged, possibly
3439 nowhere. The default is main and reject logs. */
3440
3441 if (log_reject_target != 0)
3442 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3443 sender_address, log_msg);
3444
54cdb463
PH
3445 if (user_msg == NULL) user_msg = US"local configuration problem";
3446 if (smtp_batched_input)
3447 {
3448 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3449 /* Does not return */
3450 }
3451 else
3452 {
3453 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3454 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3455 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3456 header_list);
3457 /* Does not return */
3458 }
059ec3d9 3459 }
578d43dc 3460 add_acl_headers(ACL_WHERE_NOTSMTP, US"non-SMTP");
059ec3d9 3461 }
059ec3d9
PH
3462 }
3463
54cdb463
PH
3464 /* The applicable ACLs have been run */
3465
059ec3d9
PH
3466 if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */
3467 if (queue_only_policy) queued_by = US"ACL";
059ec3d9
PH
3468 }
3469
8523533c
TK
3470#ifdef WITH_CONTENT_SCAN
3471unspool_mbox();
3472#endif
3473
6a8f9482
TK
3474#ifdef EXPERIMENTAL_DCC
3475dcc_ok = 0;
3476#endif
3477
3478
059ec3d9
PH
3479/* The final check on the message is to run the scan_local() function. The
3480version supplied with Exim always accepts, but this is a hook for sysadmins to
3481supply their own checking code. The local_scan() function is run even when all
3482the recipients have been discarded. */
3483
3484lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3485
3486/* Arrange to catch crashes in local_scan(), so that the -D file gets
3487deleted, and the incident gets logged. */
3488
3489os_non_restarting_signal(SIGSEGV, local_scan_crash_handler);
3490os_non_restarting_signal(SIGFPE, local_scan_crash_handler);
3491os_non_restarting_signal(SIGILL, local_scan_crash_handler);
3492os_non_restarting_signal(SIGBUS, local_scan_crash_handler);
3493
3494DEBUG(D_receive) debug_printf("calling local_scan(); timeout=%d\n",
3495 local_scan_timeout);
3496local_scan_data = NULL;
3497
3498os_non_restarting_signal(SIGALRM, local_scan_timeout_handler);
3499if (local_scan_timeout > 0) alarm(local_scan_timeout);
3500rc = local_scan(data_fd, &local_scan_data);
3501alarm(0);
3502os_non_restarting_signal(SIGALRM, sigalrm_handler);
3503
0e20aff9
MH
3504enable_dollar_recipients = FALSE;
3505
059ec3d9
PH
3506store_pool = POOL_MAIN; /* In case changed */
3507DEBUG(D_receive) debug_printf("local_scan() returned %d %s\n", rc,
3508 local_scan_data);
3509
3510os_non_restarting_signal(SIGSEGV, SIG_DFL);
3511os_non_restarting_signal(SIGFPE, SIG_DFL);
3512os_non_restarting_signal(SIGILL, SIG_DFL);
3513os_non_restarting_signal(SIGBUS, SIG_DFL);
3514
3515/* The length check is paranoia against some runaway code, and also because
3516(for a success return) lines in the spool file are read into big_buffer. */
3517
3518if (local_scan_data != NULL)
3519 {
3520 int len = Ustrlen(local_scan_data);
3521 if (len > LOCAL_SCAN_MAX_RETURN) len = LOCAL_SCAN_MAX_RETURN;
3522 local_scan_data = string_copyn(local_scan_data, len);
3523 }
3524
3525if (rc == LOCAL_SCAN_ACCEPT_FREEZE)
3526 {
58eb016e 3527 if (!deliver_freeze) /* ACL might have already frozen */
059ec3d9
PH
3528 {
3529 deliver_freeze = TRUE;
3530 deliver_frozen_at = time(NULL);
3531 frozen_by = US"local_scan()";
3532 }
3533 rc = LOCAL_SCAN_ACCEPT;
3534 }
3535else if (rc == LOCAL_SCAN_ACCEPT_QUEUE)
3536 {
3537 if (!queue_only_policy) /* ACL might have already queued */
3538 {
3539 queue_only_policy = TRUE;
3540 queued_by = US"local_scan()";
3541 }
3542 rc = LOCAL_SCAN_ACCEPT;
3543 }
3544
3545/* Message accepted: remove newlines in local_scan_data because otherwise
3546the spool file gets corrupted. Ensure that all recipients are qualified. */
3547
3548if (rc == LOCAL_SCAN_ACCEPT)
3549 {
3550 if (local_scan_data != NULL)
3551 {
3552 uschar *s;
3553 for (s = local_scan_data; *s != 0; s++) if (*s == '\n') *s = ' ';
3554 }
3555 for (i = 0; i < recipients_count; i++)
3556 {
3557 recipient_item *r = recipients_list + i;
3558 r->address = rewrite_address_qualify(r->address, TRUE);
3559 if (r->errors_to != NULL)
3560 r->errors_to = rewrite_address_qualify(r->errors_to, TRUE);
3561 }
3562 if (recipients_count == 0 && blackholed_by == NULL)
3563 blackholed_by = US"local_scan";
3564 }
3565
3566/* Message rejected: newlines permitted in local_scan_data to generate
3567multiline SMTP responses. */
3568
3569else
3570 {
3571 uschar *istemp = US"";
3572 uschar *s = NULL;
a5bd321b 3573 uschar *smtp_code;
059ec3d9
PH
3574 int size = 0;
3575 int sptr = 0;
059ec3d9
PH
3576
3577 errmsg = local_scan_data;
3578
3579 Uunlink(spool_name); /* Cancel this message */
3580 switch(rc)
3581 {
3582 default:
3583 log_write(0, LOG_MAIN, "invalid return %d from local_scan(). Temporary "
3584 "rejection given", rc);
3585 goto TEMPREJECT;
3586
3587 case LOCAL_SCAN_REJECT_NOLOGHDR:
6c6d6e48 3588 BIT_CLEAR(log_selector, log_selector_size, Li_rejected_header);
059ec3d9
PH
3589 /* Fall through */
3590
3591 case LOCAL_SCAN_REJECT:
a5bd321b 3592 smtp_code = US"550";
059ec3d9
PH
3593 if (errmsg == NULL) errmsg = US"Administrative prohibition";
3594 break;
3595
3596 case LOCAL_SCAN_TEMPREJECT_NOLOGHDR:
6c6d6e48 3597 BIT_CLEAR(log_selector, log_selector_size, Li_rejected_header);
059ec3d9
PH
3598 /* Fall through */
3599
3600 case LOCAL_SCAN_TEMPREJECT:
3601 TEMPREJECT:
a5bd321b 3602 smtp_code = US"451";
059ec3d9
PH
3603 if (errmsg == NULL) errmsg = US"Temporary local problem";
3604 istemp = US"temporarily ";
3605 break;
3606 }
3607
3608 s = string_append(s, &size, &sptr, 2, US"F=",
3609 (sender_address[0] == 0)? US"<>" : sender_address);
3610 s = add_host_info_for_log(s, &size, &sptr);
3611 s[sptr] = 0;
3612
3613 log_write(0, LOG_MAIN|LOG_REJECT, "%s %srejected by local_scan(): %.256s",
3614 s, istemp, string_printing(errmsg));
3615
3616 if (smtp_input)
3617 {
3618 if (!smtp_batched_input)
3619 {
a5bd321b 3620 smtp_respond(smtp_code, 3, TRUE, errmsg);
059ec3d9
PH
3621 message_id[0] = 0; /* Indicate no message accepted */
3622 smtp_reply = US""; /* Indicate reply already sent */
3623 goto TIDYUP; /* Skip to end of function */
3624 }
3625 else
3626 {
a5bd321b 3627 moan_smtp_batch(NULL, "%s %s", smtp_code, errmsg);
059ec3d9
PH
3628 /* Does not return */
3629 }
3630 }
3631 else
3632 {
3633 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3634 give_local_error(ERRMESS_LOCAL_SCAN, errmsg,
3635 US"message rejected by local scan code: ", error_rc, data_file,
3636 header_list);
3637 /* Does not return */
3638 }
3639 }
3640
3641/* Reset signal handlers to ignore signals that previously would have caused
3642the message to be abandoned. */
3643
3644signal(SIGTERM, SIG_IGN);
3645signal(SIGINT, SIG_IGN);
3646
e4bdf652 3647
059ec3d9
PH
3648/* Ensure the first time flag is set in the newly-received message. */
3649
3650deliver_firsttime = TRUE;
3651
8523533c 3652#ifdef EXPERIMENTAL_BRIGHTMAIL
a2da3176
JH
3653if (bmi_run == 1)
3654 { /* rewind data file */
8523533c
TK
3655 lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3656 bmi_verdicts = bmi_process_message(header_list, data_fd);
a2da3176 3657 }
8523533c
TK
3658#endif
3659
059ec3d9
PH
3660/* Update the timstamp in our Received: header to account for any time taken by
3661an ACL or by local_scan(). The new time is the time that all reception
3662processing is complete. */
3663
3664timestamp = expand_string(US"${tod_full}");
3665tslen = Ustrlen(timestamp);
3666
3667memcpy(received_header->text + received_header->slen - tslen - 1,
3668 timestamp, tslen);
3669
3670/* In MUA wrapper mode, ignore queueing actions set by ACL or local_scan() */
3671
3672if (mua_wrapper)
3673 {
3674 deliver_freeze = FALSE;
3675 queue_only_policy = FALSE;
3676 }
3677
3678/* Keep the data file open until we have written the header file, in order to
3679hold onto the lock. In a -bh run, or if the message is to be blackholed, we
3680don't write the header file, and we unlink the data file. If writing the header
3681file fails, we have failed to accept this message. */
3682
3683if (host_checking || blackholed_by != NULL)
3684 {
3685 header_line *h;
3686 Uunlink(spool_name);
3687 msg_size = 0; /* Compute size for log line */
3688 for (h = header_list; h != NULL; h = h->next)
3689 if (h->type != '*') msg_size += h->slen;
3690 }
3691
3692/* Write the -H file */
3693
3694else
059ec3d9
PH
3695 if ((msg_size = spool_write_header(message_id, SW_RECEIVING, &errmsg)) < 0)
3696 {
3697 log_write(0, LOG_MAIN, "Message abandoned: %s", errmsg);
3698 Uunlink(spool_name); /* Lose the data file */
3699
3700 if (smtp_input)
3701 {
3702 smtp_reply = US"451 Error in writing spool file";
3703 message_id[0] = 0; /* Indicate no message accepted */
3704 goto TIDYUP;
3705 }
3706 else
3707 {
3708 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3709 give_local_error(ERRMESS_IOERR, errmsg, US"", error_rc, data_file,
3710 header_list);
3711 /* Does not return */
3712 }
3713 }
059ec3d9
PH
3714
3715
3716/* The message has now been successfully received. */
3717
3718receive_messagecount++;
3719
3720/* In SMTP sessions we may receive several in one connection. After each one,
3721we wait for the clock to tick at the level of message-id granularity. This is
3722so that the combination of time+pid is unique, even on systems where the pid
3723can be re-used within our time interval. We can't shorten the interval without
3724re-designing the message-id. See comments above where the message id is
3725created. This is Something For The Future. */
3726
3727message_id_tv.tv_usec = (message_id_tv.tv_usec/id_resolution) * id_resolution;
3728exim_wait_tick(&message_id_tv, id_resolution);
3729
3730/* Add data size to written header size. We do not count the initial file name
3731that is in the file, but we do add one extra for the notional blank line that
3732precedes the data. This total differs from message_size in that it include the
3733added Received: header and any other headers that got created locally. */
3734
3735fflush(data_file);
3736fstat(data_fd, &statbuf);
3737
3738msg_size += statbuf.st_size - SPOOL_DATA_START_OFFSET + 1;
3739
3740/* Generate a "message received" log entry. We do this by building up a dynamic
3741string as required. Since we commonly want to add two items at a time, use a
3742macro to simplify the coding. We log the arrival of a new message while the
3743file is still locked, just in case the machine is *really* fast, and delivers
3744it first! Include any message id that is in the message - since the syntax of a
3745message id is actually an addr-spec, we can use the parse routine to canonicize
3746it. */
3747
3748size = 256;
3749sptr = 0;
3750s = store_get(size);
3751
3752s = string_append(s, &size, &sptr, 2, US"<= ",
3753 (sender_address[0] == 0)? US"<>" : sender_address);
3754if (message_reference != NULL)
3755 s = string_append(s, &size, &sptr, 2, US" R=", message_reference);