Logging: dnssec status on accept & delivery lines
[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 *
fc16abb4 1126add_host_info_for_log(uschar * s, int * sizeptr, int * ptrptr)
059ec3d9 1127{
fc16abb4 1128if (sender_fullhost)
059ec3d9 1129 {
fc16abb4
JH
1130 if (LOGGING(dnssec) && sender_host_dnssec) /*XXX sender_helo_dnssec? */
1131 s = string_cat(s, sizeptr, ptrptr, US" DS");
059ec3d9 1132 s = string_append(s, sizeptr, ptrptr, 2, US" H=", sender_fullhost);
6c6d6e48 1133 if (LOGGING(incoming_interface) && interface_address != NULL)
059ec3d9 1134 {
fc16abb4
JH
1135 s = string_cat(s, sizeptr, ptrptr,
1136 string_sprintf(" I=[%s]:%d", interface_address, interface_port));
059ec3d9
PH
1137 }
1138 }
1139if (sender_ident != NULL)
1140 s = string_append(s, sizeptr, ptrptr, 2, US" U=", sender_ident);
1141if (received_protocol != NULL)
1142 s = string_append(s, sizeptr, ptrptr, 2, US" P=", received_protocol);
1143return s;
1144}
1145
1146
1147
63955bf2 1148#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1149
1150/*************************************************
54cdb463
PH
1151* Run the MIME ACL on a message *
1152*************************************************/
1153
1154/* This code is in a subroutine so that it can be used for both SMTP
1155and non-SMTP messages. It is called with a non-NULL ACL pointer.
1156
1157Arguments:
1158 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1159 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1160 smtp_reply_ptr Where SMTP reply is being built
1161 blackholed_by_ptr Where "blackholed by" message is being built
1162
1163Returns: TRUE to carry on; FALSE to abandon the message
1164*/
1165
1166static BOOL
1167run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1168 uschar **blackholed_by_ptr)
1169{
1170FILE *mbox_file;
1171uschar rfc822_file_path[2048];
1172unsigned long mbox_size;
1173header_line *my_headerlist;
1174uschar *user_msg, *log_msg;
1175int mime_part_count_buffer = -1;
7156b1ef 1176int rc = OK;
54cdb463
PH
1177
1178memset(CS rfc822_file_path,0,2048);
1179
1180/* check if it is a MIME message */
1181my_headerlist = header_list;
4e88a19f
PH
1182while (my_headerlist != NULL)
1183 {
54cdb463 1184 /* skip deleted headers */
4e88a19f
PH
1185 if (my_headerlist->type == '*')
1186 {
54cdb463
PH
1187 my_headerlist = my_headerlist->next;
1188 continue;
4e88a19f
PH
1189 }
1190 if (strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0)
1191 {
54cdb463
PH
1192 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1193 goto DO_MIME_ACL;
4e88a19f 1194 }
54cdb463 1195 my_headerlist = my_headerlist->next;
4e88a19f 1196 }
54cdb463
PH
1197
1198DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1199return TRUE;
1200
1201DO_MIME_ACL:
1202/* make sure the eml mbox file is spooled up */
8544e77a 1203mbox_file = spool_mbox(&mbox_size, NULL);
54cdb463
PH
1204if (mbox_file == NULL) {
1205 /* error while spooling */
1206 log_write(0, LOG_MAIN|LOG_PANIC,
1207 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1208 Uunlink(spool_name);
1209 unspool_mbox();
6f0c431a
PP
1210#ifdef EXPERIMENTAL_DCC
1211 dcc_ok = 0;
1212#endif
a5bd321b 1213 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1214 message_id[0] = 0; /* Indicate no message accepted */
1215 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1216 return FALSE; /* Indicate skip to end of receive function */
1217};
1218
1219mime_is_rfc822 = 0;
1220
1221MIME_ACL_CHECK:
1222mime_part_count = -1;
1223rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1224(void)fclose(mbox_file);
54cdb463 1225
4e88a19f
PH
1226if (Ustrlen(rfc822_file_path) > 0)
1227 {
54cdb463
PH
1228 mime_part_count = mime_part_count_buffer;
1229
4e88a19f
PH
1230 if (unlink(CS rfc822_file_path) == -1)
1231 {
54cdb463
PH
1232 log_write(0, LOG_PANIC,
1233 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1234 goto END_MIME_ACL;
4e88a19f
PH
1235 }
1236 }
54cdb463
PH
1237
1238/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1239if (rc == OK)
1240 {
54cdb463 1241 uschar temp_path[1024];
e8bc7fca
JH
1242 struct dirent * entry;
1243 DIR * tempdir;
54cdb463 1244
e8bc7fca
JH
1245 (void) string_format(temp_path, sizeof(temp_path), "%s/scan/%s",
1246 spool_directory, message_id);
54cdb463 1247
4e88a19f 1248 tempdir = opendir(CS temp_path);
e8bc7fca 1249 for (;;)
4e88a19f 1250 {
e8bc7fca
JH
1251 if (!(entry = readdir(tempdir)))
1252 break;
1253 if (strncmpic(US entry->d_name, US"__rfc822_", 9) == 0)
4e88a19f 1254 {
e8bc7fca
JH
1255 (void) string_format(rfc822_file_path, sizeof(rfc822_file_path),
1256 "%s/scan/%s/%s", spool_directory, message_id, entry->d_name);
1257 debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
1258 rfc822_file_path);
4e88a19f
PH
1259 break;
1260 }
e8bc7fca 1261 }
4e88a19f 1262 closedir(tempdir);
54cdb463 1263
e8bc7fca 1264 if (entry)
4e88a19f 1265 {
e8bc7fca 1266 if ((mbox_file = Ufopen(rfc822_file_path, "rb")))
4e88a19f 1267 {
e8bc7fca
JH
1268 /* set RFC822 expansion variable */
1269 mime_is_rfc822 = 1;
1270 mime_part_count_buffer = mime_part_count;
1271 goto MIME_ACL_CHECK;
4e88a19f 1272 }
e8bc7fca
JH
1273 log_write(0, LOG_PANIC,
1274 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1275 unlink(CS rfc822_file_path);
4e88a19f
PH
1276 }
1277 }
54cdb463
PH
1278
1279END_MIME_ACL:
578d43dc 1280add_acl_headers(ACL_WHERE_MIME, US"MIME");
54cdb463
PH
1281if (rc == DISCARD)
1282 {
1283 recipients_count = 0;
1284 *blackholed_by_ptr = US"MIME ACL";
1285 }
1286else if (rc != OK)
1287 {
1288 Uunlink(spool_name);
1289 unspool_mbox();
6f0c431a
PP
1290#ifdef EXPERIMENTAL_DCC
1291 dcc_ok = 0;
1292#endif
4f6ae5c3
JH
1293 if ( smtp_input
1294 && smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1295 {
85ffcba6 1296 *smtp_yield_ptr = FALSE; /* No more messages after dropped connection */
f4c1088b 1297 *smtp_reply_ptr = US""; /* Indicate reply already sent */
4f6ae5c3 1298 }
54cdb463
PH
1299 message_id[0] = 0; /* Indicate no message accepted */
1300 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1301 }
54cdb463
PH
1302
1303return TRUE;
1304}
1305
63955bf2 1306#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1307
1308
e4bdf652
JH
1309
1310void
1311received_header_gen(void)
1312{
1313uschar *received;
1314uschar *timestamp;
1315header_line *received_header= header_list;
1316
1317timestamp = expand_string(US"${tod_full}");
1318if (recipients_count == 1) received_for = recipients_list[0].address;
1319received = expand_string(received_header_text);
1320received_for = NULL;
1321
d4ff61d1 1322if (!received)
e4bdf652
JH
1323 {
1324 if(spool_name[0] != 0)
1325 Uunlink(spool_name); /* Lose the data file */
1326 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1327 "(received_header_text) failed: %s", string_printing(received_header_text),
1328 expand_string_message);
1329 }
1330
1331/* The first element on the header chain is reserved for the Received header,
1332so all we have to do is fill in the text pointer, and set the type. However, if
1333the result of the expansion is an empty string, we leave the header marked as
1334"old" so as to refrain from adding a Received header. */
1335
1336if (received[0] == 0)
1337 {
1338 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1339 received_header->type = htype_old;
1340 }
1341else
1342 {
1343 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1344 received_header->type = htype_received;
1345 }
1346
1347received_header->slen = Ustrlen(received_header->text);
1348
1349DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1350 received_header->type, received_header->text);
1351}
1352
1353
1354
54cdb463 1355/*************************************************
059ec3d9
PH
1356* Receive message *
1357*************************************************/
1358
1359/* Receive a message on the given input, and put it into a pair of spool files.
1360Either a non-null list of recipients, or the extract flag will be true, or
1361both. The flag sender_local is true for locally generated messages. The flag
1362submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1363suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1364suppress_local_fixups" or -G was passed on the command-line.
1365The flag smtp_input is true if the message is to be
8800895a
PH
1366handled using SMTP conventions about termination and lines starting with dots.
1367For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1368
1369If a message was successfully read, message_id[0] will be non-zero.
1370
1371The general actions of this function are:
1372
1373 . Read the headers of the message (if any) into a chain of store
1374 blocks.
1375
1376 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1377 throw it away, unless the caller is trusted, or unless
1378 active_local_sender_retain is set - which can only happen if
1379 active_local_from_check is false.
059ec3d9
PH
1380
1381 . If recipients are to be extracted from the message, build the
1382 recipients list from the headers, removing any that were on the
1383 original recipients list (unless extract_addresses_remove_arguments is
1384 false), and at the same time, remove any bcc header that may be present.
1385
1386 . Get the spool file for the data, sort out its unique name, open
1387 and lock it (but don't give it the name yet).
1388
1389 . Generate a "Message-Id" header if the message doesn't have one, for
1390 locally-originated messages.
1391
1392 . Generate a "Received" header.
1393
1394 . Ensure the recipients list is fully qualified and rewritten if necessary.
1395
1396 . If there are any rewriting rules, apply them to the sender address
1397 and also to the headers.
1398
1399 . If there is no from: header, generate one, for locally-generated messages
1400 and messages in "submission mode" only.
1401
1402 . If the sender is local, check that from: is correct, and if not, generate
1403 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1404 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1405
1406 . If there is no "date" header, generate one, for locally-originated
1407 or submission mode messages only.
1408
1409 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1410 dot_ends mode, to the data file. Leave it open, to hold the lock.
1411
1412 . Write the envelope and the headers to a new file.
1413
1414 . Set the name for the header file; close it.
1415
1416 . Set the name for the data file; close it.
1417
1418Because this function can potentially be called many times in a single
1419SMTP connection, all store should be got by store_get(), so that it will be
1420automatically retrieved after the message is accepted.
1421
1422FUDGE: It seems that sites on the net send out messages with just LF
1423terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1424we make the CRs optional in all cases.
1425
1426July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1427new regime is now in place in which bare CRs in header lines are turned into LF
1428followed by a space, so as not to terminate the header line.
1429
1430February 2004: A bare LF in a header line in a message whose first line was
1431terminated by CRLF is treated in the same way as a bare CR.
1432
1433Arguments:
1434 extract_recip TRUE if recipients are to be extracted from the message's
1435 headers
1436
1437Returns: TRUE there are more messages to be read (SMTP input)
1438 FALSE there are no more messages to be read (non-SMTP input
1439 or SMTP connection collapsed, or other failure)
1440
1441When reading a message for filter testing, the returned value indicates
1442whether the headers (which is all that is read) were terminated by '.' or
1443not. */
1444
1445BOOL
1446receive_msg(BOOL extract_recip)
1447{
7156b1ef
NM
1448int i;
1449int rc = FAIL;
059ec3d9
PH
1450int msg_size = 0;
1451int process_info_len = Ustrlen(process_info);
1452int error_rc = (error_handling == ERRORS_SENDER)?
1453 errors_sender_rc : EXIT_FAILURE;
1454int header_size = 256;
1455int start, end, domain, size, sptr;
1456int id_resolution;
1457int had_zero = 0;
d677b2f2 1458int prevlines_length = 0;
059ec3d9
PH
1459
1460register int ptr = 0;
1461
1462BOOL contains_resent_headers = FALSE;
1463BOOL extracted_ignored = FALSE;
1464BOOL first_line_ended_crlf = TRUE_UNSET;
1465BOOL smtp_yield = TRUE;
1466BOOL yield = FALSE;
1467
1468BOOL resents_exist = FALSE;
1469uschar *resent_prefix = US"";
1470uschar *blackholed_by = NULL;
04f7d5b9 1471uschar *blackhole_log_msg = US"";
c5430c20 1472enum {NOT_TRIED, TMP_REJ, PERM_REJ, ACCEPTED} cutthrough_done = NOT_TRIED;
059ec3d9
PH
1473
1474flock_t lock_data;
1475error_block *bad_addresses = NULL;
1476
1477uschar *frozen_by = NULL;
1478uschar *queued_by = NULL;
1479
1480uschar *errmsg, *s;
1481struct stat statbuf;
1482
4e88a19f 1483/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1484
1485uschar *smtp_reply = NULL;
4e88a19f 1486uschar *user_msg, *log_msg;
059ec3d9
PH
1487
1488/* Working header pointers */
1489
1490header_line *h, *next;
1491
2cbb4081 1492/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1493
1494BOOL date_header_exists = FALSE;
1495
1496/* Pointers to receive the addresses of headers whose contents we need. */
1497
1498header_line *from_header = NULL;
1499header_line *subject_header = NULL;
1500header_line *msgid_header = NULL;
1501header_line *received_header;
1502
4840604e
TL
1503#ifdef EXPERIMENTAL_DMARC
1504int dmarc_up = 0;
1505#endif /* EXPERIMENTAL_DMARC */
1506
059ec3d9
PH
1507/* Variables for use when building the Received: header. */
1508
059ec3d9
PH
1509uschar *timestamp;
1510int tslen;
1511
1512/* Release any open files that might have been cached while preparing to
1513accept the message - e.g. by verifying addresses - because reading a message
1514might take a fair bit of real time. */
1515
1516search_tidyup();
1517
e4bdf652
JH
1518/* Extracting the recipient list from an input file is incompatible with
1519cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1520to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1521if (extract_recip || !smtp_input)
2e5b33cd 1522 cancel_cutthrough_connection("not smtp input");
e4bdf652 1523
059ec3d9
PH
1524/* Initialize the chain of headers by setting up a place-holder for Received:
1525header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1526pointing to the end of the chain to make adding headers simple. */
1527
1528received_header = header_list = header_last = store_get(sizeof(header_line));
1529header_list->next = NULL;
1530header_list->type = htype_old;
1531header_list->text = NULL;
1532header_list->slen = 0;
1533
1534/* Control block for the next header to be read. */
1535
1536next = store_get(sizeof(header_line));
1537next->text = store_get(header_size);
1538
1539/* Initialize message id to be null (indicating no message read), and the
1540header names list to be the normal list. Indicate there is no data file open
1541yet, initialize the size and warning count, and deal with no size limit. */
1542
1543message_id[0] = 0;
1544data_file = NULL;
1545data_fd = -1;
1546spool_name[0] = 0;
1547message_size = 0;
1548warning_count = 0;
d677b2f2 1549received_count = 1; /* For the one we will add */
059ec3d9
PH
1550
1551if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1552
2e0c1448 1553/* While reading the message, the following counts are computed. */
059ec3d9 1554
d677b2f2
PH
1555message_linecount = body_linecount = body_zerocount =
1556 max_received_linelength = 0;
059ec3d9 1557
80a47a2c
TK
1558#ifndef DISABLE_DKIM
1559/* Call into DKIM to set up the context. */
1560if (smtp_input && !smtp_batched_input && !dkim_disable_verify) dkim_exim_verify_init();
fb2274d4
TK
1561#endif
1562
4840604e
TL
1563#ifdef EXPERIMENTAL_DMARC
1564/* initialize libopendmarc */
1565dmarc_up = dmarc_init();
1566#endif
1567
059ec3d9
PH
1568/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1569ids, and fractions of a second are required. See the comments that precede the
1570message id creation below. */
1571
1572(void)gettimeofday(&message_id_tv, NULL);
1573
1574/* For other uses of the received time we can operate with granularity of one
1575second, and for that we use the global variable received_time. This is for
1576things like ultimate message timeouts. */
1577
1578received_time = message_id_tv.tv_sec;
1579
1580/* If SMTP input, set the special handler for timeouts. The alarm() calls
1581happen in the smtp_getc() function when it refills its buffer. */
1582
1583if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1584
1585/* If not SMTP input, timeout happens only if configured, and we just set a
1586single timeout for the whole message. */
1587
1588else if (receive_timeout > 0)
1589 {
1590 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1591 alarm(receive_timeout);
1592 }
1593
1594/* SIGTERM and SIGINT are caught always. */
1595
1596signal(SIGTERM, data_sigterm_sigint_handler);
1597signal(SIGINT, data_sigterm_sigint_handler);
1598
1599/* Header lines in messages are not supposed to be very long, though when
1600unfolded, to: and cc: headers can take up a lot of store. We must also cope
1601with the possibility of junk being thrown at us. Start by getting 256 bytes for
1602storing the header, and extend this as necessary using string_cat().
1603
1604To cope with total lunacies, impose an upper limit on the length of the header
1605section of the message, as otherwise the store will fill up. We must also cope
1606with the possibility of binary zeros in the data. Hence we cannot use fgets().
1607Folded header lines are joined into one string, leaving the '\n' characters
1608inside them, so that writing them out reproduces the input.
1609
1610Loop for each character of each header; the next structure for chaining the
1611header is set up already, with ptr the offset of the next character in
1612next->text. */
1613
1614for (;;)
1615 {
80a47a2c 1616 int ch = (receive_getc)();
059ec3d9
PH
1617
1618 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1619 SMTP must have a correct "." terminator. */
1620
1621 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1622 {
1623 smtp_reply = handle_lost_connection(US" (header)");
1624 smtp_yield = FALSE;
1625 goto TIDYUP; /* Skip to end of function */
1626 }
1627
1628 /* See if we are at the current header's size limit - there must be at least
1629 four bytes left. This allows for the new character plus a zero, plus two for
1630 extra insertions when we are playing games with dots and carriage returns. If
1631 we are at the limit, extend the text buffer. This could have been done
1632 automatically using string_cat() but because this is a tightish loop storing
1633 only one character at a time, we choose to do it inline. Normally
1634 store_extend() will be able to extend the block; only at the end of a big
1635 store block will a copy be needed. To handle the case of very long headers
1636 (and sometimes lunatic messages can have ones that are 100s of K long) we
1637 call store_release() for strings that have been copied - if the string is at
1638 the start of a block (and therefore the only thing in it, because we aren't
1639 doing any other gets), the block gets freed. We can only do this because we
1640 know there are no other calls to store_get() going on. */
1641
1642 if (ptr >= header_size - 4)
1643 {
1644 int oldsize = header_size;
1645 /* header_size += 256; */
1646 header_size *= 2;
1647 if (!store_extend(next->text, oldsize, header_size))
1648 {
1649 uschar *newtext = store_get(header_size);
1650 memcpy(newtext, next->text, ptr);
1651 store_release(next->text);
1652 next->text = newtext;
1653 }
1654 }
1655
1656 /* Cope with receiving a binary zero. There is dispute about whether
1657 these should be allowed in RFC 822 messages. The middle view is that they
1658 should not be allowed in headers, at least. Exim takes this attitude at
1659 the moment. We can't just stomp on them here, because we don't know that
1660 this line is a header yet. Set a flag to cause scanning later. */
1661
1662 if (ch == 0) had_zero++;
1663
1664 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1665 those from data files use just LF. Treat LF in local SMTP input as a
1666 terminator too. Treat EOF as a line terminator always. */
1667
1668 if (ch == EOF) goto EOL;
1669
1670 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1671 other MTAs accept this. We are therefore forced into this "liberalisation"
1672 too, so we accept LF as a line terminator whatever the source of the message.
1673 However, if the first line of the message ended with a CRLF, we treat a bare
1674 LF specially by inserting a white space after it to ensure that the header
1675 line is not terminated. */
1676
1677 if (ch == '\n')
1678 {
1679 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1680 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1681 goto EOL;
1682 }
1683
1684 /* This is not the end of the line. If this is SMTP input and this is
1685 the first character in the line and it is a "." character, ignore it.
1686 This implements the dot-doubling rule, though header lines starting with
1687 dots aren't exactly common. They are legal in RFC 822, though. If the
1688 following is CRLF or LF, this is the line that that terminates the
1689 entire message. We set message_ended to indicate this has happened (to
1690 prevent further reading), and break out of the loop, having freed the
1691 empty header, and set next = NULL to indicate no data line. */
1692
1693 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1694 {
80a47a2c 1695 ch = (receive_getc)();
059ec3d9
PH
1696 if (ch == '\r')
1697 {
80a47a2c 1698 ch = (receive_getc)();
059ec3d9
PH
1699 if (ch != '\n')
1700 {
80a47a2c 1701 receive_ungetc(ch);
059ec3d9
PH
1702 ch = '\r'; /* Revert to CR */
1703 }
1704 }
1705 if (ch == '\n')
1706 {
1707 message_ended = END_DOT;
1708 store_reset(next);
1709 next = NULL;
1710 break; /* End character-reading loop */
1711 }
1712
1713 /* For non-SMTP input, the dot at the start of the line was really a data
1714 character. What is now in ch is the following character. We guaranteed
1715 enough space for this above. */
1716
1717 if (!smtp_input)
1718 {
1719 next->text[ptr++] = '.';
1720 message_size++;
1721 }
1722 }
1723
1724 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1725 remember this case if this is the first line ending. */
1726
1727 if (ch == '\r')
1728 {
80a47a2c 1729 ch = (receive_getc)();
059ec3d9
PH
1730 if (ch == '\n')
1731 {
1732 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1733 goto EOL;
1734 }
1735
1736 /* Otherwise, put back the character after CR, and turn the bare CR
1737 into LF SP. */
1738
80a47a2c 1739 ch = (receive_ungetc)(ch);
059ec3d9
PH
1740 next->text[ptr++] = '\n';
1741 message_size++;
1742 ch = ' ';
1743 }
1744
1745 /* We have a data character for the header line. */
1746
1747 next->text[ptr++] = ch; /* Add to buffer */
1748 message_size++; /* Total message size so far */
1749
1750 /* Handle failure due to a humungously long header section. The >= allows
1751 for the terminating \n. Add what we have so far onto the headers list so
1752 that it gets reflected in any error message, and back up the just-read
1753 character. */
1754
1755 if (message_size >= header_maxsize)
1756 {
1757 next->text[ptr] = 0;
1758 next->slen = ptr;
1759 next->type = htype_other;
1760 next->next = NULL;
1761 header_last->next = next;
1762 header_last = next;
1763
1764 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1765 "%s (more than %d characters): message abandoned",
1766 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1767
1768 if (smtp_input)
1769 {
1770 smtp_reply = US"552 Message header is ridiculously long";
1771 receive_swallow_smtp();
1772 goto TIDYUP; /* Skip to end of function */
1773 }
1774
1775 else
1776 {
1777 give_local_error(ERRMESS_VLONGHEADER,
1778 string_sprintf("message header longer than %d characters received: "
1779 "message not accepted", header_maxsize), US"", error_rc, stdin,
1780 header_list->next);
1781 /* Does not return */
1782 }
1783 }
1784
1785 continue; /* With next input character */
1786
1787 /* End of header line reached */
1788
1789 EOL:
2e0c1448
PH
1790
1791 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1792
1793 receive_linecount++;
1794 message_linecount++;
059ec3d9 1795
d677b2f2
PH
1796 /* Keep track of maximum line length */
1797
1798 if (ptr - prevlines_length > max_received_linelength)
1799 max_received_linelength = ptr - prevlines_length;
1800 prevlines_length = ptr + 1;
1801
059ec3d9
PH
1802 /* Now put in the terminating newline. There is always space for
1803 at least two more characters. */
1804
1805 next->text[ptr++] = '\n';
1806 message_size++;
1807
1808 /* A blank line signals the end of the headers; release the unwanted
1809 space and set next to NULL to indicate this. */
1810
1811 if (ptr == 1)
1812 {
1813 store_reset(next);
1814 next = NULL;
1815 break;
1816 }
1817
1818 /* There is data in the line; see if the next input character is a
1819 whitespace character. If it is, we have a continuation of this header line.
1820 There is always space for at least one character at this point. */
1821
1822 if (ch != EOF)
1823 {
80a47a2c 1824 int nextch = (receive_getc)();
059ec3d9
PH
1825 if (nextch == ' ' || nextch == '\t')
1826 {
1827 next->text[ptr++] = nextch;
1828 message_size++;
1829 continue; /* Iterate the loop */
1830 }
80a47a2c 1831 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
1832 else ch = EOF; /* Cause main loop to exit at end */
1833 }
1834
1835 /* We have got to the real line end. Terminate the string and release store
1836 beyond it. If it turns out to be a real header, internal binary zeros will
1837 be squashed later. */
1838
1839 next->text[ptr] = 0;
1840 next->slen = ptr;
1841 store_reset(next->text + ptr + 1);
1842
1843 /* Check the running total size against the overall message size limit. We
1844 don't expect to fail here, but if the overall limit is set less than MESSAGE_
1845 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
1846 headers - the code to read the body will then also hit the buffer. */
1847
1848 if (message_size > thismessage_size_limit) break;
1849
1850 /* A line that is not syntactically correct for a header also marks
1851 the end of the headers. In this case, we leave next containing the
1852 first data line. This might actually be several lines because of the
1853 continuation logic applied above, but that doesn't matter.
1854
1855 It turns out that smail, and presumably sendmail, accept leading lines
1856 of the form
1857
1858 From ph10 Fri Jan 5 12:35 GMT 1996
1859
1860 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
1861 find any documentation of this, but for compatibility it had better be
1862 accepted. Exim restricts it to the case of non-smtp messages, and
1863 treats it as an alternative to the -f command line option. Thus it is
1864 ignored except for trusted users or filter testing. Otherwise it is taken
1865 as the sender address, unless -f was used (sendmail compatibility).
1866
1867 It further turns out that some UUCPs generate the From_line in a different
1868 format, e.g.
1869
1870 From ph10 Fri, 7 Jan 97 14:00:00 GMT
1871
1872 The regex for matching these things is now capable of recognizing both
1873 formats (including 2- and 4-digit years in the latter). In fact, the regex
1874 is now configurable, as is the expansion string to fish out the sender.
1875
1876 Even further on it has been discovered that some broken clients send
1877 these lines in SMTP messages. There is now an option to ignore them from
1878 specified hosts or networks. Sigh. */
1879
1880 if (header_last == header_list &&
1881 (!smtp_input
1882 ||
1883 (sender_host_address != NULL &&
1884 verify_check_host(&ignore_fromline_hosts) == OK)
1885 ||
1886 (sender_host_address == NULL && ignore_fromline_local)
1887 ) &&
1888 regex_match_and_setup(regex_From, next->text, 0, -1))
1889 {
1890 if (!sender_address_forced)
1891 {
1892 uschar *uucp_sender = expand_string(uucp_from_sender);
1893 if (uucp_sender == NULL)
1894 {
1895 log_write(0, LOG_MAIN|LOG_PANIC,
1896 "expansion of \"%s\" failed after matching "
1897 "\"From \" line: %s", uucp_from_sender, expand_string_message);
1898 }
1899 else
1900 {
1901 int start, end, domain;
1902 uschar *errmess;
1903 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
1904 &start, &end, &domain, TRUE);
1905 if (newsender != NULL)
1906 {
1907 if (domain == 0 && newsender[0] != 0)
1908 newsender = rewrite_address_qualify(newsender, FALSE);
1909
f05da2e8 1910 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
1911 {
1912 sender_address = newsender;
1913
f05da2e8 1914 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
1915 {
1916 authenticated_sender = NULL;
1917 originator_name = US"";
1918 sender_local = FALSE;
1919 }
1920
f05da2e8 1921 if (filter_test != FTEST_NONE)
059ec3d9
PH
1922 printf("Sender taken from \"From \" line\n");
1923 }
1924 }
1925 }
1926 }
1927 }
1928
1929 /* Not a leading "From " line. Check to see if it is a valid header line.
1930 Header names may contain any non-control characters except space and colon,
1931 amazingly. */
1932
1933 else
1934 {
1935 uschar *p = next->text;
1936
1937 /* If not a valid header line, break from the header reading loop, leaving
1938 next != NULL, indicating that it holds the first line of the body. */
1939
1940 if (isspace(*p)) break;
1941 while (mac_isgraph(*p) && *p != ':') p++;
1942 while (isspace(*p)) p++;
1943 if (*p != ':')
1944 {
1945 body_zerocount = had_zero;
1946 break;
1947 }
1948
1949 /* We have a valid header line. If there were any binary zeroes in
1950 the line, stomp on them here. */
1951
1952 if (had_zero > 0)
1953 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
1954
1955 /* It is perfectly legal to have an empty continuation line
1956 at the end of a header, but it is confusing to humans
1957 looking at such messages, since it looks like a blank line.
1958 Reduce confusion by removing redundant white space at the
1959 end. We know that there is at least one printing character
1960 (the ':' tested for above) so there is no danger of running
1961 off the end. */
1962
1963 p = next->text + ptr - 2;
1964 for (;;)
1965 {
1966 while (*p == ' ' || *p == '\t') p--;
1967 if (*p != '\n') break;
1968 ptr = (p--) - next->text + 1;
1969 message_size -= next->slen - ptr;
1970 next->text[ptr] = 0;
1971 next->slen = ptr;
1972 }
1973
1974 /* Add the header to the chain */
1975
1976 next->type = htype_other;
1977 next->next = NULL;
1978 header_last->next = next;
1979 header_last = next;
1980
1981 /* Check the limit for individual line lengths. This comes after adding to
1982 the chain so that the failing line is reflected if a bounce is generated
1983 (for a local message). */
1984
1985 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
1986 {
1987 log_write(0, LOG_MAIN, "overlong message header line received from "
1988 "%s (more than %d characters): message abandoned",
1989 sender_host_unknown? sender_ident : sender_fullhost,
1990 header_line_maxsize);
1991
1992 if (smtp_input)
1993 {
1994 smtp_reply = US"552 A message header line is too long";
1995 receive_swallow_smtp();
1996 goto TIDYUP; /* Skip to end of function */
1997 }
1998
1999 else
2000 {
2001 give_local_error(ERRMESS_VLONGHDRLINE,
2002 string_sprintf("message header line longer than %d characters "
2003 "received: message not accepted", header_line_maxsize), US"",
2004 error_rc, stdin, header_list->next);
2005 /* Does not return */
2006 }
2007 }
2008
2009 /* Note if any resent- fields exist. */
2010
2011 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
2012 {
2013 resents_exist = TRUE;
2014 resent_prefix = US"Resent-";
2015 }
2016 }
2017
2018 /* The line has been handled. If we have hit EOF, break out of the loop,
2019 indicating no pending data line. */
2020
2021 if (ch == EOF) { next = NULL; break; }
2022
2023 /* Set up for the next header */
2024
2025 header_size = 256;
2026 next = store_get(sizeof(header_line));
2027 next->text = store_get(header_size);
2028 ptr = 0;
2029 had_zero = 0;
d677b2f2 2030 prevlines_length = 0;
059ec3d9
PH
2031 } /* Continue, starting to read the next header */
2032
2033/* At this point, we have read all the headers into a data structure in main
2034store. The first header is still the dummy placeholder for the Received: header
2035we are going to generate a bit later on. If next != NULL, it contains the first
2036data line - which terminated the headers before reaching a blank line (not the
2037normal case). */
2038
2039DEBUG(D_receive)
2040 {
2041 debug_printf(">>Headers received:\n");
2042 for (h = header_list->next; h != NULL; h = h->next)
2043 debug_printf("%s", h->text);
2044 debug_printf("\n");
2045 }
2046
2047/* End of file on any SMTP connection is an error. If an incoming SMTP call
2048is dropped immediately after valid headers, the next thing we will see is EOF.
2049We must test for this specially, as further down the reading of the data is
2050skipped if already at EOF. */
2051
2052if (smtp_input && (receive_feof)())
2053 {
2054 smtp_reply = handle_lost_connection(US" (after header)");
2055 smtp_yield = FALSE;
2056 goto TIDYUP; /* Skip to end of function */
2057 }
2058
2059/* If this is a filter test run and no headers were read, output a warning
2060in case there is a mistake in the test message. */
2061
f05da2e8 2062if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2063 printf("Warning: no message headers read\n");
2064
2065
2066/* Scan the headers to identify them. Some are merely marked for later
2067processing; some are dealt with here. */
2068
2069for (h = header_list->next; h != NULL; h = h->next)
2070 {
2071 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2072 if (is_resent) contains_resent_headers = TRUE;
2073
2074 switch (header_checkname(h, is_resent))
2075 {
059ec3d9 2076 case htype_bcc:
2cbb4081 2077 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
2078 break;
2079
059ec3d9 2080 case htype_cc:
2cbb4081 2081 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
2082 break;
2083
2084 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
2085
2086 case htype_date:
4c69d561 2087 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
2088 break;
2089
2090 /* Same comments as about Return-Path: below. */
2091
2092 case htype_delivery_date:
2093 if (delivery_date_remove) h->type = htype_old;
2094 break;
2095
2096 /* Same comments as about Return-Path: below. */
2097
2098 case htype_envelope_to:
2099 if (envelope_to_remove) h->type = htype_old;
2100 break;
2101
2102 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2103 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2104 header consists of just the login id of the user who called Exim, rewrite
2105 it with the gecos field first. Apply this rule to Resent-From: if there
2106 are resent- fields. */
2107
2108 case htype_from:
2109 h->type = htype_from;
2110 if (!resents_exist || is_resent)
2111 {
2112 from_header = h;
2113 if (!smtp_input)
2114 {
5de8faa3 2115 int len;
059ec3d9
PH
2116 uschar *s = Ustrchr(h->text, ':') + 1;
2117 while (isspace(*s)) s++;
5de8faa3 2118 len = h->slen - (s - h->text) - 1;
e0fccd1d
TF
2119 if (Ustrlen(originator_login) == len &&
2120 strncmpic(s, originator_login, len) == 0)
059ec3d9
PH
2121 {
2122 uschar *name = is_resent? US"Resent-From" : US"From";
2123 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2124 originator_login, qualify_domain_sender);
2125 from_header = header_last;
2126 h->type = htype_old;
2127 DEBUG(D_receive|D_rewrite)
2128 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2129 }
2130 }
2131 }
2132 break;
2133
2134 /* Identify the Message-id: header for generating "in-reply-to" in the
2135 autoreply transport. For incoming logging, save any resent- value. In both
2136 cases, take just the first of any multiples. */
2137
2138 case htype_id:
2139 if (msgid_header == NULL && (!resents_exist || is_resent))
2140 {
2141 msgid_header = h;
2142 h->type = htype_id;
2143 }
2144 break;
2145
2146 /* Flag all Received: headers */
2147
2148 case htype_received:
2149 h->type = htype_received;
2150 received_count++;
2151 break;
2152
2153 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
2154
2155 case htype_reply_to:
2156 h->type = htype_reply_to;
2157 break;
2158
2159 /* The Return-path: header is supposed to be added to messages when
2160 they leave the SMTP system. We shouldn't receive messages that already
2161 contain Return-path. However, since Exim generates Return-path: on
2162 local delivery, resent messages may well contain it. We therefore
2163 provide an option (which defaults on) to remove any Return-path: headers
2164 on input. Removal actually means flagging as "old", which prevents the
2165 header being transmitted with the message. */
2166
2167 case htype_return_path:
2168 if (return_path_remove) h->type = htype_old;
2169
2170 /* If we are testing a mail filter file, use the value of the
2171 Return-Path: header to set up the return_path variable, which is not
2172 otherwise set. However, remove any <> that surround the address
2173 because the variable doesn't have these. */
2174
f05da2e8 2175 if (filter_test != FTEST_NONE)
059ec3d9
PH
2176 {
2177 uschar *start = h->text + 12;
2178 uschar *end = start + Ustrlen(start);
2179 while (isspace(*start)) start++;
2180 while (end > start && isspace(end[-1])) end--;
2181 if (*start == '<' && end[-1] == '>')
2182 {
2183 start++;
2184 end--;
2185 }
2186 return_path = string_copyn(start, end - start);
2187 printf("Return-path taken from \"Return-path:\" header line\n");
2188 }
2189 break;
2190
2191 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2192 and from an untrusted caller and suppress_local_fixups is not set, or if we
2193 are in submission mode for a remote message, mark it "old" so that it will
2194 not be transmitted with the message, unless active_local_sender_retain is
2195 set. (This can only be true if active_local_from_check is false.) If there
2196 are any resent- headers in the message, apply this rule to Resent-Sender:
2197 instead of Sender:. Messages with multiple resent- header sets cannot be
2198 tidily handled. (For this reason, at least one MUA - Pine - turns old
2199 resent- headers into X-resent- headers when resending, leaving just one
2200 set.) */
059ec3d9
PH
2201
2202 case htype_sender:
69358f02 2203 h->type = ((!active_local_sender_retain &&
8800895a
PH
2204 (
2205 (sender_local && !trusted_caller && !suppress_local_fixups)
2206 || submission_mode
2207 )
059ec3d9
PH
2208 ) &&
2209 (!resents_exist||is_resent))?
2210 htype_old : htype_sender;
2211 break;
2212
2213 /* Remember the Subject: header for logging. There is no Resent-Subject */
2214
2215 case htype_subject:
2216 subject_header = h;
2217 break;
2218
2219 /* "To:" gets flagged, and the existence of a recipient header is noted,
2220 whether it's resent- or not. */
2221
2222 case htype_to:
2223 h->type = htype_to;
2224 /****
2225 to_or_cc_header_exists = TRUE;
2226 ****/
2227 break;
2228 }
2229 }
2230
2231/* Extract recipients from the headers if that is required (the -t option).
2232Note that this is documented as being done *before* any address rewriting takes
2233place. There are two possibilities:
2234
2235(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2236recipients already listed are to be REMOVED from the message. Smail 3 works
2237like this. We need to build a non-recipients tree for that list, because in
2238subsequent processing this data is held in a tree and that's what the
2239spool_write_header() function expects. Make sure that non-recipient addresses
2240are fully qualified and rewritten if necessary.
2241
2242(2) According to other sendmail documentation, -t ADDS extracted recipients to
2243those in the command line arguments (and it is rumoured some other MTAs do
2244this). Therefore, there is an option to make Exim behave this way.
2245
2246*** Notes on "Resent-" header lines ***
2247
2248The presence of resent-headers in the message makes -t horribly ambiguous.
2249Experiments with sendmail showed that it uses recipients for all resent-
2250headers, totally ignoring the concept of "sets of resent- headers" as described
2251in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2252with all the addresses in one instance of each header.
2253
2254This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2255error for -t if there were resent- headers in the message. However, after a
2256discussion on the mailing list, I've learned that there are MUAs that use
2257resent- headers with -t, and also that the stuff about sets of resent- headers
2258and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2259message with -t and resent- header lines makes sure that only *its* resent-
2260headers are present; previous ones are often renamed as X-resent- for example.
2261
2262Consequently, Exim has been changed so that, if any resent- header lines are
2263present, the recipients are taken from all of the appropriate resent- lines,
2264and not from the ordinary To:, Cc:, etc. */
2265
2266if (extract_recip)
2267 {
2268 int rcount = 0;
2269 error_block **bnext = &bad_addresses;
2270
2271 if (extract_addresses_remove_arguments)
2272 {
2273 while (recipients_count-- > 0)
2274 {
2275 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2276 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2277 tree_add_nonrecipient(s);
2278 }
2279 recipients_list = NULL;
2280 recipients_count = recipients_list_max = 0;
2281 }
2282
059ec3d9
PH
2283 /* Now scan the headers */
2284
2285 for (h = header_list->next; h != NULL; h = h->next)
2286 {
2287 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2288 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2289 {
2290 uschar *s = Ustrchr(h->text, ':') + 1;
2291 while (isspace(*s)) s++;
2292
1eccaa59
PH
2293 parse_allow_group = TRUE; /* Allow address group syntax */
2294
059ec3d9
PH
2295 while (*s != 0)
2296 {
2297 uschar *ss = parse_find_address_end(s, FALSE);
2298 uschar *recipient, *errmess, *p, *pp;
2299 int start, end, domain;
2300
2301 /* Check on maximum */
2302
2303 if (recipients_max > 0 && ++rcount > recipients_max)
2304 {
2305 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2306 US"message rejected: ", error_rc, stdin, NULL);
2307 /* Does not return */
2308 }
2309
2310 /* Make a copy of the address, and remove any internal newlines. These
2311 may be present as a result of continuations of the header line. The
2312 white space that follows the newline must not be removed - it is part
2313 of the header. */
2314
2315 pp = recipient = store_get(ss - s + 1);
2316 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2317 *pp = 0;
250b6871 2318
8c5d388a 2319#ifdef SUPPORT_I18N
250b6871
JH
2320 {
2321 BOOL b = allow_utf8_domains;
2322 allow_utf8_domains = TRUE;
2323#endif
059ec3d9
PH
2324 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2325 &domain, FALSE);
2326
8c5d388a 2327#ifdef SUPPORT_I18N
250b6871
JH
2328 if (string_is_utf8(recipient))
2329 message_smtputf8 = TRUE;
2330 else
2331 allow_utf8_domains = b;
2332 }
2333#endif
2334
059ec3d9
PH
2335 /* Keep a list of all the bad addresses so we can send a single
2336 error message at the end. However, an empty address is not an error;
2337 just ignore it. This can come from an empty group list like
2338
2339 To: Recipients of list:;
2340
2341 If there are no recipients at all, an error will occur later. */
2342
2343 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2344 {
2345 int len = Ustrlen(s);
2346 error_block *b = store_get(sizeof(error_block));
2347 while (len > 0 && isspace(s[len-1])) len--;
2348 b->next = NULL;
2349 b->text1 = string_printing(string_copyn(s, len));
2350 b->text2 = errmess;
2351 *bnext = b;
2352 bnext = &(b->next);
2353 }
2354
2355 /* If the recipient is already in the nonrecipients tree, it must
2356 have appeared on the command line with the option extract_addresses_
2357 remove_arguments set. Do not add it to the recipients, and keep a note
2358 that this has happened, in order to give a better error if there are
2359 no recipients left. */
2360
2361 else if (recipient != NULL)
2362 {
2363 if (tree_search(tree_nonrecipients, recipient) == NULL)
2364 receive_add_recipient(recipient, -1);
2365 else
2366 extracted_ignored = TRUE;
2367 }
2368
2369 /* Move on past this address */
2370
2371 s = ss + (*ss? 1:0);
2372 while (isspace(*s)) s++;
1eccaa59
PH
2373 } /* Next address */
2374
2375 parse_allow_group = FALSE; /* Reset group syntax flags */
2376 parse_found_group = FALSE;
059ec3d9
PH
2377
2378 /* If this was the bcc: header, mark it "old", which means it
2379 will be kept on the spool, but not transmitted as part of the
2380 message. */
2381
2cbb4081 2382 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2383 } /* For appropriate header line */
2384 } /* For each header line */
2385
059ec3d9
PH
2386 }
2387
2388/* Now build the unique message id. This has changed several times over the
2389lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2390Retaining all the history in the comment has become too unwieldy - read
2391previous release sources if you want it.
2392
2393The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2394The first part is the current time, in seconds. The second part is the current
2395pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2396can hold a number in the range 0-3843. It used to be a computed sequence
2397number, but is now the fractional component of the current time in units of
23981/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2399received, Exim ensures that the timer has ticked at the appropriate level
2400before proceeding, to avoid duplication if the pid happened to be re-used
2401within the same time period. It seems likely that most messages will take at
2402least half a millisecond to be received, so no delay will normally be
2403necessary. At least for some time...
2404
2405There is a modification when localhost_number is set. Formerly this was allowed
2406to be as large as 255. Now it is restricted to the range 0-16, and the final
2407component of the message id becomes (localhost_number * 200) + fractional time
2408in units of 1/200 of a second (i.e. a value in the range 0-3399).
2409
2410Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2411Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2412still allows the tttttt field to hold a large enough number to last for some
2413more decades, and the final two-digit field can hold numbers up to 1295, which
2414is enough for milliseconds (instead of 1/2000 of a second).
2415
2416However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2417pid, so it is probably safe because pids have to be positive. The
2418localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2419final field becomes (localhost_number * 100) + fractional time in centiseconds.
2420
2421Note that string_base62() returns its data in a static storage block, so it
2422must be copied before calling string_base62() again. It always returns exactly
24236 characters.
2424
2425There doesn't seem to be anything in the RFC which requires a message id to
2426start with a letter, but Smail was changed to ensure this. The external form of
2427the message id (as supplied by string expansion) therefore starts with an
2428additional leading 'E'. The spool file names do not include this leading
2429letter and it is not used internally.
2430
2431NOTE: If ever the format of message ids is changed, the regular expression for
2432checking that a string is in this format must be updated in a corresponding
2433way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2434must also be changed to reflect the correct string length. Then, of course,
2435other programs that rely on the message id format will need updating too. */
2436
2437Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2438message_id[6] = '-';
2439Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2440
2441/* Deal with the case where the host number is set. The value of the number was
2442checked when it was read, to ensure it isn't too big. The timing granularity is
2443left in id_resolution so that an appropriate wait can be done after receiving
2444the message, if necessary (we hope it won't be). */
2445
2446if (host_number_string != NULL)
2447 {
2448 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2449 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2450 string_base62((long int)(
2451 host_number * (1000000/id_resolution) +
2452 message_id_tv.tv_usec/id_resolution)) + 4);
2453 }
2454
2455/* Host number not set: final field is just the fractional time at an
2456appropriate resolution. */
2457
2458else
2459 {
2460 id_resolution = (BASE_62 == 62)? 500 : 1000;
2461 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2462 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2463 }
2464
2465/* Add the current message id onto the current process info string if
2466it will fit. */
2467
2468(void)string_format(process_info + process_info_len,
2469 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2470
2471/* If we are using multiple input directories, set up the one for this message
2472to be the least significant base-62 digit of the time of arrival. Otherwise
2473ensure that it is an empty string. */
2474
2475message_subdir[0] = split_spool_directory? message_id[5] : 0;
2476
2477/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2478one, but only for local (without suppress_local_fixups) or submission mode
2479messages. This can be user-configured if required, but we had better flatten
2480any illegal characters therein. */
059ec3d9 2481
8800895a
PH
2482if (msgid_header == NULL &&
2483 ((sender_host_address == NULL && !suppress_local_fixups)
2484 || submission_mode))
059ec3d9
PH
2485 {
2486 uschar *p;
2487 uschar *id_text = US"";
2488 uschar *id_domain = primary_hostname;
2489
2490 /* Permit only letters, digits, dots, and hyphens in the domain */
2491
2492 if (message_id_domain != NULL)
2493 {
2494 uschar *new_id_domain = expand_string(message_id_domain);
2495 if (new_id_domain == NULL)
2496 {
2497 if (!expand_string_forcedfail)
2498 log_write(0, LOG_MAIN|LOG_PANIC,
2499 "expansion of \"%s\" (message_id_header_domain) "
2500 "failed: %s", message_id_domain, expand_string_message);
2501 }
2502 else if (*new_id_domain != 0)
2503 {
2504 id_domain = new_id_domain;
2505 for (p = id_domain; *p != 0; p++)
2506 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2507 }
2508 }
2509
2510 /* Permit all characters except controls and RFC 2822 specials in the
2511 additional text part. */
2512
2513 if (message_id_text != NULL)
2514 {
2515 uschar *new_id_text = expand_string(message_id_text);
2516 if (new_id_text == NULL)
2517 {
2518 if (!expand_string_forcedfail)
2519 log_write(0, LOG_MAIN|LOG_PANIC,
2520 "expansion of \"%s\" (message_id_header_text) "
2521 "failed: %s", message_id_text, expand_string_message);
2522 }
2523 else if (*new_id_text != 0)
2524 {
2525 id_text = new_id_text;
2526 for (p = id_text; *p != 0; p++)
2527 if (mac_iscntrl_or_special(*p)) *p = '-';
2528 }
2529 }
2530
e7e680d6
PP
2531 /* Add the header line
2532 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2533 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2534
e7e680d6 2535 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2536 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2537 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2538 }
2539
2540/* If we are to log recipients, keep a copy of the raw ones before any possible
2541rewriting. Must copy the count, because later ACLs and the local_scan()
2542function may mess with the real recipients. */
2543
6c6d6e48 2544if (LOGGING(received_recipients))
059ec3d9
PH
2545 {
2546 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2547 for (i = 0; i < recipients_count; i++)
2548 raw_recipients[i] = string_copy(recipients_list[i].address);
2549 raw_recipients_count = recipients_count;
2550 }
2551
2552/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2553recipients will get here only if the conditions were right (allow_unqualified_
2554recipient is TRUE). */
2555
2556for (i = 0; i < recipients_count; i++)
2557 recipients_list[i].address =
2558 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2559 global_rewrite_rules, rewrite_existflags);
2560
8800895a
PH
2561/* If there is no From: header, generate one for local (without
2562suppress_local_fixups) or submission_mode messages. If there is no sender
2563address, but the sender is local or this is a local delivery error, use the
2564originator login. This shouldn't happen for genuine bounces, but might happen
2565for autoreplies. The addition of From: must be done *before* checking for the
2566possible addition of a Sender: header, because untrusted_set_sender allows an
2567untrusted user to set anything in the envelope (which might then get info
2568From:) but we still want to ensure a valid Sender: if it is required. */
2569
2570if (from_header == NULL &&
2571 ((sender_host_address == NULL && !suppress_local_fixups)
2572 || submission_mode))
059ec3d9 2573 {
2fe1a124
PH
2574 uschar *oname = US"";
2575
2576 /* Use the originator_name if this is a locally submitted message and the
2577 caller is not trusted. For trusted callers, use it only if -F was used to
2578 force its value or if we have a non-SMTP message for which -f was not used
2579 to set the sender. */
2580
2581 if (sender_host_address == NULL)
2582 {
2583 if (!trusted_caller || sender_name_forced ||
2584 (!smtp_input && !sender_address_forced))
2585 oname = originator_name;
2586 }
2587
2588 /* For non-locally submitted messages, the only time we use the originator
2589 name is when it was forced by the /name= option on control=submission. */
2590
2591 else
2592 {
2593 if (submission_name != NULL) oname = submission_name;
2594 }
2595
059ec3d9
PH
2596 /* Envelope sender is empty */
2597
2598 if (sender_address[0] == 0)
2599 {
87ba3f5f
PH
2600 uschar *fromstart, *fromend;
2601
2602 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2603 oname, (oname[0] == 0)? "" : " <");
2604 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2605
059ec3d9
PH
2606 if (sender_local || local_error_message)
2607 {
87ba3f5f
PH
2608 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2609 local_part_quote(originator_login), qualify_domain_sender,
2610 fromend);
059ec3d9
PH
2611 }
2612 else if (submission_mode && authenticated_id != NULL)
2613 {
2614 if (submission_domain == NULL)
2615 {
87ba3f5f
PH
2616 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2617 local_part_quote(authenticated_id), qualify_domain_sender,
2618 fromend);
059ec3d9
PH
2619 }
2620 else if (submission_domain[0] == 0) /* empty => whole address set */
2621 {
87ba3f5f
PH
2622 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2623 fromend);
059ec3d9
PH
2624 }
2625 else
2626 {
87ba3f5f
PH
2627 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2628 local_part_quote(authenticated_id), submission_domain,
2629 fromend);
059ec3d9
PH
2630 }
2631 from_header = header_last; /* To get it checked for Sender: */
2632 }
2633 }
2634
2635 /* There is a non-null envelope sender. Build the header using the original
2636 sender address, before any rewriting that might have been done while
2637 verifying it. */
2638
2639 else
2640 {
87ba3f5f 2641 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2642 oname,
2643 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2644 (sender_address_unrewritten == NULL)?
2645 sender_address : sender_address_unrewritten,
2fe1a124 2646 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2647
2648 from_header = header_last; /* To get it checked for Sender: */
2649 }
2650 }
2651
2652
8800895a
PH
2653/* If the sender is local (without suppress_local_fixups), or if we are in
2654submission mode and there is an authenticated_id, check that an existing From:
2655is correct, and if not, generate a Sender: header, unless disabled. Any
2656previously-existing Sender: header was removed above. Note that sender_local,
2657as well as being TRUE if the caller of exim is not trusted, is also true if a
2658trusted caller did not supply a -f argument for non-smtp input. To allow
2659trusted callers to forge From: without supplying -f, we have to test explicitly
2660here. If the From: header contains more than one address, then the call to
2661parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2662
2663if (from_header != NULL &&
69358f02 2664 (active_local_from_check &&
8800895a 2665 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2666 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2667 ))
2668 {
2669 BOOL make_sender = TRUE;
2670 int start, end, domain;
2671 uschar *errmess;
2672 uschar *from_address =
2673 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2674 &start, &end, &domain, FALSE);
2675 uschar *generated_sender_address;
2676
2677 if (submission_mode)
2678 {
2679 if (submission_domain == NULL)
2680 {
2681 generated_sender_address = string_sprintf("%s@%s",
2682 local_part_quote(authenticated_id), qualify_domain_sender);
2683 }
2684 else if (submission_domain[0] == 0) /* empty => full address */
2685 {
2686 generated_sender_address = string_sprintf("%s",
2687 authenticated_id);
2688 }
2689 else
2690 {
2691 generated_sender_address = string_sprintf("%s@%s",
2692 local_part_quote(authenticated_id), submission_domain);
2693 }
2694 }
2695 else
2696 generated_sender_address = string_sprintf("%s@%s",
2697 local_part_quote(originator_login), qualify_domain_sender);
2698
2699 /* Remove permitted prefixes and suffixes from the local part of the From:
2700 address before doing the comparison with the generated sender. */
2701
2702 if (from_address != NULL)
2703 {
2704 int slen;
2705 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2706
2707 if (at != NULL) *at = 0;
2708 from_address += route_check_prefix(from_address, local_from_prefix);
2709 slen = route_check_suffix(from_address, local_from_suffix);
2710 if (slen > 0)
2711 {
2712 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2713 from_address += slen;
2714 }
2715 if (at != NULL) *at = '@';
2716
2717 if (strcmpic(generated_sender_address, from_address) == 0 ||
2718 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2719 make_sender = FALSE;
2720 }
2721
2722 /* We have to cause the Sender header to be rewritten if there are
2723 appropriate rewriting rules. */
2724
2725 if (make_sender)
2726 {
2fe1a124 2727 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2728 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2729 generated_sender_address);
2730 else
2731 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2732 resent_prefix,
2733 submission_mode? submission_name : originator_name,
2734 generated_sender_address);
059ec3d9 2735 }
87ba3f5f
PH
2736
2737 /* Ensure that a non-null envelope sender address corresponds to the
2738 submission mode sender address. */
2739
2740 if (submission_mode && sender_address[0] != 0)
2741 {
2742 if (sender_address_unrewritten == NULL)
2743 sender_address_unrewritten = sender_address;
2744 sender_address = generated_sender_address;
089793a4
TF
2745 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2746 log_write(L_address_rewrite, LOG_MAIN,
2747 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2748 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2749 }
059ec3d9
PH
2750 }
2751
059ec3d9
PH
2752/* If there are any rewriting rules, apply them to the sender address, unless
2753it has already been rewritten as part of verification for SMTP input. */
2754
2755if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2756 sender_address[0] != 0)
2757 {
2758 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2759 global_rewrite_rules, rewrite_existflags);
2760 DEBUG(D_receive|D_rewrite)
2761 debug_printf("rewritten sender = %s\n", sender_address);
2762 }
2763
2764
2765/* The headers must be run through rewrite_header(), because it ensures that
2766addresses are fully qualified, as well as applying any rewriting rules that may
2767exist.
2768
2769Qualification of header addresses in a message from a remote host happens only
2770if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2771appropriate. For local messages, qualification always happens, unless -bnq is
2772used to explicitly suppress it. No rewriting is done for an unqualified address
2773that is left untouched.
2774
2775We start at the second header, skipping our own Received:. This rewriting is
2776documented as happening *after* recipient addresses are taken from the headers
2777by the -t command line option. An added Sender: gets rewritten here. */
2778
2779for (h = header_list->next; h != NULL; h = h->next)
2780 {
2781 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2782 rewrite_existflags, TRUE);
2783 if (newh != NULL) h = newh;
2784 }
2785
2786
2787/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2788"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2789"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2790exists, this applies to the set of resent- headers rather than the normal set.
2791
2cbb4081
PH
2792The requirement for a recipient header has been removed in RFC 2822. At this
2793point in the code, earlier versions of Exim added a To: header for locally
2794submitted messages, and an empty Bcc: header for others. In the light of the
2795changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2796
059ec3d9
PH
2797
2798/* If there is no date header, generate one if the message originates locally
8800895a
PH
2799(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2800submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2801to be more confusing if Exim adds one to all remotely-originated messages.
2802As per Message-Id, we prepend if resending, else append.
2803*/
059ec3d9 2804
8800895a
PH
2805if (!date_header_exists &&
2806 ((sender_host_address == NULL && !suppress_local_fixups)
2807 || submission_mode))
e7e680d6
PP
2808 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2809 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2810
2811search_tidyup(); /* Free any cached resources */
2812
2813/* Show the complete set of headers if debugging. Note that the first one (the
2814new Received:) has not yet been set. */
2815
2816DEBUG(D_receive)
2817 {
2818 debug_printf(">>Headers after rewriting and local additions:\n");
2819 for (h = header_list->next; h != NULL; h = h->next)
2820 debug_printf("%c %s", h->type, h->text);
2821 debug_printf("\n");
2822 }
2823
2824/* The headers are now complete in store. If we are running in filter
2825testing mode, that is all this function does. Return TRUE if the message
2826ended with a dot. */
2827
f05da2e8 2828if (filter_test != FTEST_NONE)
059ec3d9
PH
2829 {
2830 process_info[process_info_len] = 0;
2831 return message_ended == END_DOT;
2832 }
2833
817d9f57 2834/* Cutthrough delivery:
5032d1cf
JH
2835We have to create the Received header now rather than at the end of reception,
2836so the timestamp behaviour is a change to the normal case.
2837XXX Ensure this gets documented XXX.
2838Having created it, send the headers to the destination. */
2839if (cutthrough.fd >= 0)
e4bdf652 2840 {
817d9f57
JH
2841 if (received_count > received_headers_max)
2842 {
2e5b33cd 2843 cancel_cutthrough_connection("too many headers");
817d9f57
JH
2844 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2845 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2846 "Too many \"Received\" headers",
2847 sender_address,
2848 (sender_fullhost == NULL)? "" : " H=",
2849 (sender_fullhost == NULL)? US"" : sender_fullhost,
2850 (sender_ident == NULL)? "" : " U=",
2851 (sender_ident == NULL)? US"" : sender_ident);
2852 message_id[0] = 0; /* Indicate no message accepted */
2853 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
2854 goto TIDYUP; /* Skip to end of function */
2855 }
e4bdf652 2856 received_header_gen();
578d43dc 2857 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652
JH
2858 (void) cutthrough_headers_send();
2859 }
61147df4 2860
e4bdf652 2861
059ec3d9
PH
2862/* Open a new spool file for the data portion of the message. We need
2863to access it both via a file descriptor and a stream. Try to make the
2864directory if it isn't there. Note re use of sprintf: spool_directory
2865is checked on input to be < 200 characters long. */
2866
2867sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory, message_subdir,
2868 message_id);
2869data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2870if (data_fd < 0)
2871 {
2872 if (errno == ENOENT)
2873 {
2874 uschar temp[16];
2875 sprintf(CS temp, "input/%s", message_subdir);
2876 if (message_subdir[0] == 0) temp[5] = 0;
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
TK
3652#ifdef EXPERIMENTAL_BRIGHTMAIL
3653if (bmi_run == 1) {
3654 /* rewind data file */
3655 lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3656 bmi_verdicts = bmi_process_message(header_list, data_fd);
3657};
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
3695 {
3696 if ((msg_size = spool_write_header(message_id, SW_RECEIVING, &errmsg)) < 0)
3697 {
3698 log_write(0, LOG_MAIN, "Message abandoned: %s", errmsg);
3699 Uunlink(spool_name); /* Lose the data file */
3700
3701 if (smtp_input)
3702 {
3703 smtp_reply = US"451 Error in writing spool file";
3704 message_id[0] = 0; /* Indicate no message accepted */
3705 goto TIDYUP;
3706 }
3707 else
3708 {
3709 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3710 give_local_error(ERRMESS_IOERR, errmsg, US"", error_rc, data_file,
3711 header_list);
3712 /* Does not return */
3713 }
3714 }
3715 }
3716
3717
3718/* The message has now been successfully received. */
3719
3720receive_messagecount++;
3721
3722/* In SMTP sessions we may receive several in one connection. After each one,
3723we wait for the clock to tick at the level of message-id granularity. This is
3724so that the combination of time+pid is unique, even on systems where the pid
3725can be re-used within our time interval. We can't shorten the interval without
3726re-designing the message-id. See comments above where the message id is
3727created. This is Something For The Future. */
3728
3729message_id_tv.tv_usec = (message_id_tv.tv_usec/id_resolution) * id_resolution;
3730exim_wait_tick(&message_id_tv, id_resolution);
3731
3732/* Add data size to written header size. We do not count the initial file name
3733that is in the file, but we do add one extra for the notional blank line that
3734precedes the data. This total differs from message_size in that it include the
3735added Received: header and any other headers that got created locally. */
3736
3737fflush(data_file);
3738fstat(data_fd, &statbuf);
3739
3740msg_size += statbuf.st_size - SPOOL_DATA_START_OFFSET + 1;
3741
3742/* Generate a "message received" log entry. We do this by building up a dynamic
3743string as required. Since we commonly want to add two items at a time, use a
3744macro to simplify the coding. We log the arrival of a new message while the
3745file is still locked, just in case the machine is *really* fast, and delivers
3746it first! Include any message id that is in the message - since the syntax of a
3747message id is actually an addr-spec, we can use the parse routine to canonicize
3748it. */
3749
3750size = 256;
3751sptr = 0;
3752s = store_get(size);
3753
3754s = string_append(s, &size, &sptr, 2, US"<= ",
3755 (sender_address[0] == 0)? US"<>" : sender_address);
3756if (message_reference != NULL)
3757 s = string_append(s, &size, &sptr, 2, US" R=", message_reference);
3758
3759s = add_host_info_for_log(s, &size, &sptr);
3760