Guard smtp_user_msg() with EXPERIMENTAL_PRDR check.
[exim.git] / src / src / receive.c
CommitLineData
059ec3d9
PH
1/*************************************************
2* Exim - an Internet mail transport agent *
3*************************************************/
4
c4ceed07 5/* Copyright (c) University of Cambridge 1995 - 2012 */
059ec3d9
PH
6/* See the file NOTICE for conditions of use and distribution. */
7
8/* Code for receiving a message and setting up spool files. */
9
059ec3d9
PH
10#include "exim.h"
11
6a8f9482
TK
12#ifdef EXPERIMENTAL_DCC
13extern int dcc_ok;
14#endif
15
059ec3d9
PH
16/*************************************************
17* Local static variables *
18*************************************************/
19
20static FILE *data_file = NULL;
21static int data_fd = -1;
22static uschar spool_name[256];
23
24
25
26/*************************************************
27* Non-SMTP character reading functions *
28*************************************************/
29
30/* These are the default functions that are set up in the variables such as
31receive_getc initially. They just call the standard functions, passing stdin as
32the file. (When SMTP input is occurring, different functions are used by
33changing the pointer variables.) */
34
35int
36stdin_getc(void)
37{
38return getc(stdin);
39}
40
41int
42stdin_ungetc(int c)
43{
44return ungetc(c, stdin);
45}
46
47int
48stdin_feof(void)
49{
50return feof(stdin);
51}
52
53int
54stdin_ferror(void)
55{
56return ferror(stdin);
57}
58
59
60
61
62/*************************************************
63* Check that a set sender is allowed *
64*************************************************/
65
66/* This function is called when a local caller sets an explicit sender address.
67It checks whether this is permitted, which it is for trusted callers.
68Otherwise, it must match the pattern(s) in untrusted_set_sender.
69
70Arguments: the proposed sender address
71Returns: TRUE for a trusted caller
72 TRUE if the address has been set, untrusted_set_sender has been
73 set, and the address matches something in the list
74 FALSE otherwise
75*/
76
77BOOL
78receive_check_set_sender(uschar *newsender)
79{
80uschar *qnewsender;
81if (trusted_caller) return TRUE;
82if (newsender == NULL || untrusted_set_sender == NULL) return FALSE;
83qnewsender = (Ustrchr(newsender, '@') != NULL)?
84 newsender : string_sprintf("%s@%s", newsender, qualify_domain_sender);
85return
86 match_address_list(qnewsender, TRUE, TRUE, &untrusted_set_sender, NULL, -1,
87 0, NULL) == OK;
88}
89
90
91
92
93/*************************************************
5cb8cbc6 94* Read space info for a partition *
059ec3d9
PH
95*************************************************/
96
8e669ac1
PH
97/* This function is called by receive_check_fs() below, and also by string
98expansion for variables such as $spool_space. The field names for the statvfs
5cb8cbc6
PH
99structure are macros, because not all OS have F_FAVAIL and it seems tidier to
100have macros for F_BAVAIL and F_FILES as well. Some kinds of file system do not
101have inodes, and they return -1 for the number available.
059ec3d9 102
5cb8cbc6
PH
103Later: It turns out that some file systems that do not have the concept of
104inodes return 0 rather than -1. Such systems should also return 0 for the total
8e669ac1 105number of inodes, so we require that to be greater than zero before returning
5cb8cbc6 106an inode count.
059ec3d9 107
5cb8cbc6
PH
108Arguments:
109 isspool TRUE for spool partition, FALSE for log partition
110 inodeptr address of int to receive inode count; -1 if there isn't one
8e669ac1 111
5cb8cbc6 112Returns: available on-root space, in kilobytes
8e669ac1
PH
113 -1 for log partition if there isn't one
114
115All values are -1 if the STATFS functions are not available.
059ec3d9
PH
116*/
117
8e669ac1 118int
5cb8cbc6 119receive_statvfs(BOOL isspool, int *inodeptr)
059ec3d9
PH
120{
121#ifdef HAVE_STATFS
059ec3d9 122struct STATVFS statbuf;
5cb8cbc6
PH
123uschar *path;
124uschar *name;
125uschar buffer[1024];
059ec3d9 126
5cb8cbc6 127/* The spool directory must always exist. */
059ec3d9 128
5cb8cbc6 129if (isspool)
059ec3d9 130 {
8e669ac1
PH
131 path = spool_directory;
132 name = US"spool";
133 }
134
059ec3d9
PH
135/* Need to cut down the log file path to the directory, and to ignore any
136appearance of "syslog" in it. */
137
5cb8cbc6 138else
059ec3d9 139 {
059ec3d9 140 int sep = ':'; /* Not variable - outside scripts use */
059ec3d9 141 uschar *p = log_file_path;
8e669ac1 142 name = US"log";
059ec3d9
PH
143
144 /* An empty log_file_path means "use the default". This is the same as an
145 empty item in a list. */
146
147 if (*p == 0) p = US":";
148 while ((path = string_nextinlist(&p, &sep, buffer, sizeof(buffer))) != NULL)
149 {
150 if (Ustrcmp(path, "syslog") != 0) break;
151 }
152
5cb8cbc6
PH
153 if (path == NULL) /* No log files */
154 {
8e669ac1
PH
155 *inodeptr = -1;
156 return -1;
157 }
059ec3d9 158
8e669ac1
PH
159 /* An empty string means use the default, which is in the spool directory.
160 But don't just use the spool directory, as it is possible that the log
5cb8cbc6 161 subdirectory has been symbolically linked elsewhere. */
059ec3d9 162
8e669ac1 163 if (path[0] == 0)
059ec3d9 164 {
5cb8cbc6
PH
165 sprintf(CS buffer, CS"%s/log", CS spool_directory);
166 path = buffer;
8e669ac1
PH
167 }
168 else
059ec3d9 169 {
8e669ac1 170 uschar *cp;
5cb8cbc6 171 if ((cp = Ustrrchr(path, '/')) != NULL) *cp = 0;
8e669ac1 172 }
5cb8cbc6 173 }
8e669ac1 174
8f128379 175/* We now have the path; do the business */
5cb8cbc6
PH
176
177memset(&statbuf, 0, sizeof(statbuf));
178
179if (STATVFS(CS path, &statbuf) != 0)
180 {
181 log_write(0, LOG_MAIN|LOG_PANIC, "cannot accept message: failed to stat "
182 "%s directory %s: %s", name, spool_directory, strerror(errno));
183 smtp_closedown(US"spool or log directory problem");
184 exim_exit(EXIT_FAILURE);
185 }
8e669ac1 186
5cb8cbc6
PH
187*inodeptr = (statbuf.F_FILES > 0)? statbuf.F_FAVAIL : -1;
188
189/* Disks are getting huge. Take care with computing the size in kilobytes. */
8e669ac1 190
5cb8cbc6
PH
191return (int)(((double)statbuf.F_BAVAIL * (double)statbuf.F_FRSIZE)/1024.0);
192
193/* Unable to find partition sizes in this environment. */
194
195#else
196*inodeptr = -1;
197return -1;
198#endif
199}
200
059ec3d9 201
059ec3d9 202
5cb8cbc6
PH
203
204/*************************************************
205* Check space on spool and log partitions *
206*************************************************/
207
208/* This function is called before accepting a message; if any thresholds are
209set, it checks them. If a message_size is supplied, it checks that there is
210enough space for that size plus the threshold - i.e. that the message won't
211reduce the space to the threshold. Not all OS have statvfs(); for those that
212don't, this function always returns TRUE. For some OS the old function and
213struct name statfs is used; that is handled by a macro, defined in exim.h.
214
215Arguments:
216 msg_size the (estimated) size of an incoming message
217
218Returns: FALSE if there isn't enough space, or if the information cannot
219 be obtained
220 TRUE if no check was done or there is enough space
221*/
222
223BOOL
224receive_check_fs(int msg_size)
225{
226int space, inodes;
227
228if (check_spool_space > 0 || msg_size > 0 || check_spool_inodes > 0)
229 {
8e669ac1
PH
230 space = receive_statvfs(TRUE, &inodes);
231
059ec3d9 232 DEBUG(D_receive)
5cb8cbc6
PH
233 debug_printf("spool directory space = %dK inodes = %d "
234 "check_space = %dK inodes = %d msg_size = %d\n",
235 space, inodes, check_spool_space, check_spool_inodes, msg_size);
8e669ac1
PH
236
237 if ((space >= 0 && space < check_spool_space) ||
5cb8cbc6 238 (inodes >= 0 && inodes < check_spool_inodes))
8e669ac1 239 {
5cb8cbc6
PH
240 log_write(0, LOG_MAIN, "spool directory space check failed: space=%d "
241 "inodes=%d", space, inodes);
059ec3d9
PH
242 return FALSE;
243 }
244 }
245
5cb8cbc6
PH
246if (check_log_space > 0 || check_log_inodes > 0)
247 {
8e669ac1
PH
248 space = receive_statvfs(FALSE, &inodes);
249
5cb8cbc6
PH
250 DEBUG(D_receive)
251 debug_printf("log directory space = %dK inodes = %d "
252 "check_space = %dK inodes = %d\n",
253 space, inodes, check_log_space, check_log_inodes);
8e669ac1
PH
254
255 if ((space >= 0 && space < check_log_space) ||
5cb8cbc6 256 (inodes >= 0 && inodes < check_log_inodes))
8e669ac1 257 {
5cb8cbc6
PH
258 log_write(0, LOG_MAIN, "log directory space check failed: space=%d "
259 "inodes=%d", space, inodes);
260 return FALSE;
261 }
8e669ac1
PH
262 }
263
059ec3d9
PH
264return TRUE;
265}
266
267
268
269/*************************************************
270* Bomb out while reading a message *
271*************************************************/
272
273/* The common case of wanting to bomb out is if a SIGTERM or SIGINT is
274received, or if there is a timeout. A rarer case might be if the log files are
275screwed up and Exim can't open them to record a message's arrival. Handling
276that case is done by setting a flag to cause the log functions to call this
277function if there is an ultimate disaster. That is why it is globally
278accessible.
279
8f128379
PH
280Arguments:
281 reason text reason to pass to the not-quit ACL
282 msg default SMTP response to give if in an SMTP session
059ec3d9
PH
283Returns: it doesn't
284*/
285
286void
8f128379 287receive_bomb_out(uschar *reason, uschar *msg)
059ec3d9 288{
ead37e6c
PP
289 static BOOL already_bombing_out;
290/* The smtp_notquit_exit() below can call ACLs which can trigger recursive
291timeouts, if someone has something slow in their quit ACL. Since the only
292things we should be doing are to close down cleanly ASAP, on the second
293pass we also close down stuff that might be opened again, before bypassing
294the ACL call and exiting. */
295
059ec3d9
PH
296/* If spool_name is set, it contains the name of the data file that is being
297written. Unlink it before closing so that it cannot be picked up by a delivery
298process. Ensure that any header file is also removed. */
299
ead37e6c 300if (spool_name[0] != '\0')
059ec3d9
PH
301 {
302 Uunlink(spool_name);
303 spool_name[Ustrlen(spool_name) - 1] = 'H';
304 Uunlink(spool_name);
ead37e6c 305 spool_name[0] = '\0';
059ec3d9
PH
306 }
307
308/* Now close the file if it is open, either as a fd or a stream. */
309
ead37e6c
PP
310if (data_file != NULL)
311 {
312 (void)fclose(data_file);
313 data_file = NULL;
314} else if (data_fd >= 0) {
315 (void)close(data_fd);
316 data_fd = -1;
317 }
059ec3d9 318
8f128379
PH
319/* Attempt to close down an SMTP connection tidily. For non-batched SMTP, call
320smtp_notquit_exit(), which runs the NOTQUIT ACL, if present, and handles the
321SMTP response. */
059ec3d9 322
ead37e6c 323if (!already_bombing_out)
059ec3d9 324 {
ead37e6c
PP
325 already_bombing_out = TRUE;
326 if (smtp_input)
327 {
328 if (smtp_batched_input)
329 moan_smtp_batch(NULL, "421 %s - message abandoned", msg); /* No return */
330 smtp_notquit_exit(reason, US"421", US"%s %s - closing connection.",
331 smtp_active_hostname, msg);
332 }
059ec3d9
PH
333 }
334
335/* Exit from the program (non-BSMTP cases) */
336
337exim_exit(EXIT_FAILURE);
338}
339
340
341/*************************************************
342* Data read timeout *
343*************************************************/
344
345/* Handler function for timeouts that occur while reading the data that
346comprises a message.
347
348Argument: the signal number
349Returns: nothing
350*/
351
352static void
353data_timeout_handler(int sig)
354{
355uschar *msg = NULL;
356
357sig = sig; /* Keep picky compilers happy */
358
359if (smtp_input)
360 {
361 msg = US"SMTP incoming data timeout";
362 log_write(L_lost_incoming_connection,
363 LOG_MAIN, "SMTP data timeout (message abandoned) on connection "
fed77020
PH
364 "from %s F=<%s>",
365 (sender_fullhost != NULL)? sender_fullhost : US"local process",
366 sender_address);
059ec3d9
PH
367 }
368else
369 {
370 fprintf(stderr, "exim: timed out while reading - message abandoned\n");
371 log_write(L_lost_incoming_connection,
372 LOG_MAIN, "timed out while reading local message");
373 }
374
8f128379 375receive_bomb_out(US"data-timeout", msg); /* Does not return */
059ec3d9
PH
376}
377
378
379
380/*************************************************
381* local_scan() timeout *
382*************************************************/
383
384/* Handler function for timeouts that occur while running a local_scan()
385function.
386
387Argument: the signal number
388Returns: nothing
389*/
390
391static void
392local_scan_timeout_handler(int sig)
393{
394sig = sig; /* Keep picky compilers happy */
395log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function timed out - "
396 "message temporarily rejected (size %d)", message_size);
8f128379
PH
397/* Does not return */
398receive_bomb_out(US"local-scan-timeout", US"local verification problem");
059ec3d9
PH
399}
400
401
402
403/*************************************************
404* local_scan() crashed *
405*************************************************/
406
407/* Handler function for signals that occur while running a local_scan()
408function.
409
410Argument: the signal number
411Returns: nothing
412*/
413
414static void
415local_scan_crash_handler(int sig)
416{
417log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function crashed with "
418 "signal %d - message temporarily rejected (size %d)", sig, message_size);
8f128379
PH
419/* Does not return */
420receive_bomb_out(US"local-scan-error", US"local verification problem");
059ec3d9
PH
421}
422
423
424/*************************************************
425* SIGTERM or SIGINT received *
426*************************************************/
427
428/* Handler for SIGTERM or SIGINT signals that occur while reading the
429data that comprises a message.
430
431Argument: the signal number
432Returns: nothing
433*/
434
435static void
436data_sigterm_sigint_handler(int sig)
437{
438uschar *msg = NULL;
439
440if (smtp_input)
441 {
442 msg = US"Service not available - SIGTERM or SIGINT received";
443 log_write(0, LOG_MAIN, "%s closed after %s", smtp_get_connection_info(),
444 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
445 }
446else
447 {
f05da2e8 448 if (filter_test == FTEST_NONE)
059ec3d9
PH
449 {
450 fprintf(stderr, "\nexim: %s received - message abandoned\n",
451 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
452 log_write(0, LOG_MAIN, "%s received while reading local message",
453 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
454 }
455 }
456
8f128379 457receive_bomb_out(US"signal-exit", msg); /* Does not return */
059ec3d9
PH
458}
459
460
461
462/*************************************************
463* Add new recipient to list *
464*************************************************/
465
466/* This function builds a list of recipient addresses in argc/argv
467format.
468
469Arguments:
470 recipient the next address to add to recipients_list
471 pno parent number for fixed aliases; -1 otherwise
472
473Returns: nothing
474*/
475
476void
477receive_add_recipient(uschar *recipient, int pno)
478{
479if (recipients_count >= recipients_list_max)
480 {
481 recipient_item *oldlist = recipients_list;
482 int oldmax = recipients_list_max;
483 recipients_list_max = recipients_list_max? 2*recipients_list_max : 50;
484 recipients_list = store_get(recipients_list_max * sizeof(recipient_item));
485 if (oldlist != NULL)
486 memcpy(recipients_list, oldlist, oldmax * sizeof(recipient_item));
487 }
488
489recipients_list[recipients_count].address = recipient;
490recipients_list[recipients_count].pno = pno;
8523533c
TK
491#ifdef EXPERIMENTAL_BRIGHTMAIL
492recipients_list[recipients_count].bmi_optin = bmi_current_optin;
493/* reset optin string pointer for next recipient */
494bmi_current_optin = NULL;
495#endif
059ec3d9
PH
496recipients_list[recipients_count++].errors_to = NULL;
497}
498
499
500
501
502/*************************************************
fd98a5c6
JH
503* Send user response message *
504*************************************************/
61147df4 505
fd98a5c6
JH
506/* This function is passed a default response code and a user message. It calls
507smtp_message_code() to check and possibly modify the response code, and then
508calls smtp_respond() to transmit the response. I put this into a function
509just to avoid a lot of repetition.
61147df4
PP
510
511Arguments:
fd98a5c6
JH
512 code the response code
513 user_msg the user message
514
515Returns: nothing
61147df4
PP
516*/
517
518#ifdef EXPERIMENTAL_PRDR
519static void
fd98a5c6 520smtp_user_msg(uschar *code, uschar *user_msg)
61147df4 521{
fd98a5c6
JH
522int len = 3;
523smtp_message_code(&code, &len, &user_msg, NULL);
524smtp_respond(code, len, TRUE, user_msg);
61147df4
PP
525}
526#endif
527
528
529
530
fd98a5c6
JH
531
532/*************************************************
059ec3d9
PH
533* Remove a recipient from the list *
534*************************************************/
535
536/* This function is provided for local_scan() to use.
537
538Argument:
539 recipient address to remove
540
541Returns: TRUE if it did remove something; FALSE otherwise
542*/
543
544BOOL
545receive_remove_recipient(uschar *recipient)
546{
547int count;
548DEBUG(D_receive) debug_printf("receive_remove_recipient(\"%s\") called\n",
549 recipient);
550for (count = 0; count < recipients_count; count++)
551 {
552 if (Ustrcmp(recipients_list[count].address, recipient) == 0)
553 {
554 if ((--recipients_count - count) > 0)
555 memmove(recipients_list + count, recipients_list + count + 1,
54cdb463 556 (recipients_count - count)*sizeof(recipient_item));
059ec3d9
PH
557 return TRUE;
558 }
559 }
560return FALSE;
561}
562
563
564
565
566
567/*************************************************
568* Read data portion of a non-SMTP message *
569*************************************************/
570
571/* This function is called to read the remainder of a message (following the
572header) when the input is not from SMTP - we are receiving a local message on
573a standard input stream. The message is always terminated by EOF, and is also
574terminated by a dot on a line by itself if the flag dot_ends is TRUE. Split the
575two cases for maximum efficiency.
576
577Ensure that the body ends with a newline. This will naturally be the case when
578the termination is "\n.\n" but may not be otherwise. The RFC defines messages
579as "sequences of lines" - this of course strictly applies only to SMTP, but
580deliveries into BSD-type mailbox files also require it. Exim used to have a
581flag for doing this at delivery time, but as it was always set for all
582transports, I decided to simplify things by putting the check here instead.
583
584There is at least one MUA (dtmail) that sends CRLF via this interface, and
585other programs are known to do this as well. Exim used to have a option for
586dealing with this: in July 2003, after much discussion, the code has been
587changed to default to treat any of LF, CRLF, and bare CR as line terminators.
588
589However, for the case when a dot on a line by itself terminates a message, the
590only recognized terminating sequences before and after the dot are LF and CRLF.
591Otherwise, having read EOL . CR, you don't know whether to read another
592character or not.
593
594Internally, in messages stored in Exim's spool files, LF is used as the line
595terminator. Under the new regime, bare CRs will no longer appear in these
596files.
597
598Arguments:
599 fout a FILE to which to write the message
600
601Returns: One of the END_xxx values indicating why it stopped reading
602*/
603
604static int
605read_message_data(FILE *fout)
606{
607int ch_state;
608register int ch;
d677b2f2 609register int linelength = 0;
059ec3d9
PH
610
611/* Handle the case when only EOF terminates the message */
612
613if (!dot_ends)
614 {
615 register int last_ch = '\n';
616
80a47a2c 617 for (; (ch = (receive_getc)()) != EOF; last_ch = ch)
059ec3d9
PH
618 {
619 if (ch == 0) body_zerocount++;
620 if (last_ch == '\r' && ch != '\n')
621 {
d677b2f2
PH
622 if (linelength > max_received_linelength)
623 max_received_linelength = linelength;
624 linelength = 0;
059ec3d9
PH
625 if (fputc('\n', fout) == EOF) return END_WERROR;
626 message_size++;
627 body_linecount++;
628 }
629 if (ch == '\r') continue;
630
631 if (fputc(ch, fout) == EOF) return END_WERROR;
d677b2f2
PH
632 if (ch == '\n')
633 {
634 if (linelength > max_received_linelength)
635 max_received_linelength = linelength;
636 linelength = 0;
637 body_linecount++;
638 }
639 else linelength++;
059ec3d9
PH
640 if (++message_size > thismessage_size_limit) return END_SIZE;
641 }
642
643 if (last_ch != '\n')
644 {
d677b2f2
PH
645 if (linelength > max_received_linelength)
646 max_received_linelength = linelength;
059ec3d9
PH
647 if (fputc('\n', fout) == EOF) return END_WERROR;
648 message_size++;
649 body_linecount++;
650 }
651
652 return END_EOF;
653 }
654
655/* Handle the case when a dot on a line on its own, or EOF, terminates. */
656
657ch_state = 1;
658
80a47a2c 659while ((ch = (receive_getc)()) != EOF)
059ec3d9
PH
660 {
661 if (ch == 0) body_zerocount++;
662 switch (ch_state)
663 {
664 case 0: /* Normal state (previous char written) */
665 if (ch == '\n')
d677b2f2
PH
666 {
667 body_linecount++;
668 if (linelength > max_received_linelength)
669 max_received_linelength = linelength;
670 linelength = -1;
671 ch_state = 1;
672 }
059ec3d9
PH
673 else if (ch == '\r')
674 { ch_state = 2; continue; }
675 break;
676
677 case 1: /* After written "\n" */
678 if (ch == '.') { ch_state = 3; continue; }
d677b2f2 679 if (ch != '\n') ch_state = 0; else linelength = -1;
059ec3d9
PH
680 break;
681
682 case 2:
683 body_linecount++; /* After unwritten "\r" */
d677b2f2
PH
684 if (linelength > max_received_linelength)
685 max_received_linelength = linelength;
059ec3d9 686 if (ch == '\n')
d677b2f2
PH
687 {
688 ch_state = 1;
689 linelength = -1;
690 }
059ec3d9
PH
691 else
692 {
693 if (message_size++, fputc('\n', fout) == EOF) return END_WERROR;
694 if (ch == '\r') continue;
695 ch_state = 0;
d677b2f2 696 linelength = 0;
059ec3d9
PH
697 }
698 break;
699
700 case 3: /* After "\n." (\n written, dot not) */
701 if (ch == '\n') return END_DOT;
702 if (ch == '\r') { ch_state = 4; continue; }
703 message_size++;
d677b2f2 704 linelength++;
059ec3d9
PH
705 if (fputc('.', fout) == EOF) return END_WERROR;
706 ch_state = 0;
707 break;
708
709 case 4: /* After "\n.\r" (\n written, rest not) */
710 if (ch == '\n') return END_DOT;
711 message_size += 2;
712 body_linecount++;
713 if (fputs(".\n", fout) == EOF) return END_WERROR;
714 if (ch == '\r') { ch_state = 2; continue; }
715 ch_state = 0;
716 break;
717 }
718
d677b2f2 719 linelength++;
059ec3d9
PH
720 if (fputc(ch, fout) == EOF) return END_WERROR;
721 if (++message_size > thismessage_size_limit) return END_SIZE;
722 }
723
724/* Get here if EOF read. Unless we have just written "\n", we need to ensure
725the message ends with a newline, and we must also write any characters that
726were saved up while testing for an ending dot. */
727
728if (ch_state != 1)
729 {
730 static uschar *ends[] = { US"\n", NULL, US"\n", US".\n", US".\n" };
731 if (fputs(CS ends[ch_state], fout) == EOF) return END_WERROR;
732 message_size += Ustrlen(ends[ch_state]);
733 body_linecount++;
734 }
735
736return END_EOF;
737}
738
739
740
741
742/*************************************************
743* Read data portion of an SMTP message *
744*************************************************/
745
746/* This function is called to read the remainder of an SMTP message (after the
747headers), or to skip over it when an error has occurred. In this case, the
748output file is passed as NULL.
749
750If any line begins with a dot, that character is skipped. The input should only
751be successfully terminated by CR LF . CR LF unless it is local (non-network)
752SMTP, in which case the CRs are optional, but...
753
754FUDGE: It seems that sites on the net send out messages with just LF
755terminators, despite the warnings in the RFCs, and other MTAs handle this. So
756we make the CRs optional in all cases.
757
758July 2003: Bare CRs cause trouble. We now treat them as line terminators as
759well, so that there are no CRs in spooled messages. However, the message
760terminating dot is not recognized between two bare CRs.
761
762Arguments:
763 fout a FILE to which to write the message; NULL if skipping
764
765Returns: One of the END_xxx values indicating why it stopped reading
766*/
767
768static int
769read_message_data_smtp(FILE *fout)
770{
771int ch_state = 0;
e4bdf652 772int ch;
1f5497b2 773register int linelength = 0;
059ec3d9 774
80a47a2c 775while ((ch = (receive_getc)()) != EOF)
059ec3d9
PH
776 {
777 if (ch == 0) body_zerocount++;
778 switch (ch_state)
779 {
780 case 0: /* After LF or CRLF */
781 if (ch == '.')
782 {
783 ch_state = 3;
784 continue; /* Don't ever write . after LF */
785 }
786 ch_state = 1;
787
788 /* Else fall through to handle as normal uschar. */
789
790 case 1: /* Normal state */
791 if (ch == '\n')
792 {
793 ch_state = 0;
794 body_linecount++;
1f5497b2
PH
795 if (linelength > max_received_linelength)
796 max_received_linelength = linelength;
797 linelength = -1;
059ec3d9
PH
798 }
799 else if (ch == '\r')
800 {
801 ch_state = 2;
802 continue;
803 }
804 break;
805
806 case 2: /* After (unwritten) CR */
807 body_linecount++;
1f5497b2
PH
808 if (linelength > max_received_linelength)
809 max_received_linelength = linelength;
810 linelength = -1;
059ec3d9
PH
811 if (ch == '\n')
812 {
813 ch_state = 0;
814 }
815 else
816 {
817 message_size++;
818 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
e4bdf652 819 (void) cutthrough_put_nl();
059ec3d9
PH
820 if (ch != '\r') ch_state = 1; else continue;
821 }
822 break;
823
824 case 3: /* After [CR] LF . */
825 if (ch == '\n')
826 return END_DOT;
827 if (ch == '\r')
828 {
829 ch_state = 4;
830 continue;
831 }
832 ch_state = 1; /* The dot itself is removed */
833 break;
834
835 case 4: /* After [CR] LF . CR */
836 if (ch == '\n') return END_DOT;
837 message_size++;
838 body_linecount++;
839 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
e4bdf652 840 (void) cutthrough_put_nl();
059ec3d9
PH
841 if (ch == '\r')
842 {
843 ch_state = 2;
844 continue;
845 }
846 ch_state = 1;
847 break;
848 }
849
850 /* Add the character to the spool file, unless skipping; then loop for the
851 next. */
852
853 message_size++;
1f5497b2 854 linelength++;
059ec3d9
PH
855 if (fout != NULL)
856 {
857 if (fputc(ch, fout) == EOF) return END_WERROR;
858 if (message_size > thismessage_size_limit) return END_SIZE;
859 }
e4bdf652
JH
860 if(ch == '\n')
861 (void) cutthrough_put_nl();
862 else
863 {
864 uschar c= ch;
865 (void) cutthrough_puts(&c, 1);
866 }
059ec3d9
PH
867 }
868
869/* Fall through here if EOF encountered. This indicates some kind of error,
870since a correct message is terminated by [CR] LF . [CR] LF. */
871
872return END_EOF;
873}
874
875
876
877
878/*************************************************
879* Swallow SMTP message *
880*************************************************/
881
882/* This function is called when there has been some kind of error while reading
883an SMTP message, and the remaining data may need to be swallowed. It is global
884because it is called from smtp_closedown() to shut down an incoming call
885tidily.
886
887Argument: a FILE from which to read the message
888Returns: nothing
889*/
890
891void
892receive_swallow_smtp(void)
893{
894if (message_ended >= END_NOTENDED)
895 message_ended = read_message_data_smtp(NULL);
896}
897
898
899
900/*************************************************
901* Handle lost SMTP connection *
902*************************************************/
903
904/* This function logs connection loss incidents and generates an appropriate
905SMTP response.
906
907Argument: additional data for the message
908Returns: the SMTP response
909*/
910
911static uschar *
912handle_lost_connection(uschar *s)
913{
914log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
915 "%s lost while reading message data%s", smtp_get_connection_info(), s);
916return US"421 Lost incoming connection";
917}
918
919
920
921
922/*************************************************
923* Handle a non-smtp reception error *
924*************************************************/
925
926/* This function is called for various errors during the reception of non-SMTP
927messages. It either sends a message to the sender of the problem message, or it
928writes to the standard error stream.
929
930Arguments:
931 errcode code for moan_to_sender(), identifying the error
932 text1 first message text, passed to moan_to_sender()
933 text2 second message text, used only for stderrr
934 error_rc code to pass to exim_exit if no problem
935 f FILE containing body of message (may be stdin)
936 hptr pointer to instore headers or NULL
937
938Returns: calls exim_exit(), which does not return
939*/
940
941static void
942give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
943 FILE *f, header_line *hptr)
944{
945if (error_handling == ERRORS_SENDER)
946 {
947 error_block eblock;
948 eblock.next = NULL;
949 eblock.text1 = text1;
950 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
951 error_rc = EXIT_FAILURE;
952 }
953else fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 954(void)fclose(f);
059ec3d9
PH
955exim_exit(error_rc);
956}
957
958
959
960/*************************************************
961* Add header lines set up by ACL *
962*************************************************/
963
850635b6
PH
964/* This function is called to add the header lines that were set up by
965statements in an ACL to the list of headers in memory. It is done in two stages
966like this, because when the ACL for RCPT is running, the other headers have not
967yet been received. This function is called twice; once just before running the
968DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
969are visible to the DATA ACL.
059ec3d9
PH
970
971Originally these header lines were added at the end. Now there is support for
972three different places: top, bottom, and after the Received: header(s). There
973will always be at least one Received: header, even if it is marked deleted, and
974even if something else has been put in front of it.
975
976Arguments:
977 acl_name text to identify which ACL
978
979Returns: nothing
980*/
981
982static void
983add_acl_headers(uschar *acl_name)
984{
985header_line *h, *next;
986header_line *last_received = NULL;
e7568d51
TL
987
988if (acl_removed_headers != NULL)
989 {
990 DEBUG(D_receive|D_acl) debug_printf(">>Headers removed by %s ACL:\n", acl_name);
991
992 for (h = header_list; h != NULL; h = h->next)
993 {
e7568d51
TL
994 uschar *list;
995 BOOL include_header;
996
997 if (h->type == htype_old) continue;
998
999 include_header = TRUE;
1000 list = acl_removed_headers;
1001
1002 int sep = ':'; /* This is specified as a colon-separated list */
1003 uschar *s;
1004 uschar buffer[128];
1005 while ((s = string_nextinlist(&list, &sep, buffer, sizeof(buffer)))
1006 != NULL)
1007 {
1008 int len = Ustrlen(s);
1009 if (header_testname(h, s, len, FALSE))
1010 {
1011 h->type = htype_old;
1012 DEBUG(D_receive|D_acl) debug_printf(" %s", h->text);
1013 }
1014 }
1015 }
1016 acl_removed_headers = NULL;
1017 DEBUG(D_receive|D_acl) debug_printf(">>\n");
1018 }
059ec3d9 1019
71fafd95 1020if (acl_added_headers == NULL) return;
059ec3d9
PH
1021DEBUG(D_receive|D_acl) debug_printf(">>Headers added by %s ACL:\n", acl_name);
1022
71fafd95 1023for (h = acl_added_headers; h != NULL; h = next)
059ec3d9
PH
1024 {
1025 next = h->next;
1026
1027 switch(h->type)
1028 {
1029 case htype_add_top:
1030 h->next = header_list;
1031 header_list = h;
1032 DEBUG(D_receive|D_acl) debug_printf(" (at top)");
1033 break;
1034
1035 case htype_add_rec:
1036 if (last_received == NULL)
1037 {
1038 last_received = header_list;
1039 while (!header_testname(last_received, US"Received", 8, FALSE))
1040 last_received = last_received->next;
1041 while (last_received->next != NULL &&
1042 header_testname(last_received->next, US"Received", 8, FALSE))
1043 last_received = last_received->next;
1044 }
1045 h->next = last_received->next;
1046 last_received->next = h;
1047 DEBUG(D_receive|D_acl) debug_printf(" (after Received:)");
1048 break;
1049
8523533c
TK
1050 case htype_add_rfc:
1051 /* add header before any header which is NOT Received: or Resent- */
1052 last_received = header_list;
1053 while ( (last_received->next != NULL) &&
1054 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
1055 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
1056 last_received = last_received->next;
1057 /* last_received now points to the last Received: or Resent-* header
1058 in an uninterrupted chain of those header types (seen from the beginning
1059 of all headers. Our current header must follow it. */
1060 h->next = last_received->next;
1061 last_received->next = h;
8e669ac1 1062 DEBUG(D_receive|D_acl) debug_printf(" (before any non-Received: or Resent-*: header)");
8523533c
TK
1063 break;
1064
059ec3d9
PH
1065 default:
1066 h->next = NULL;
1067 header_last->next = h;
1068 break;
1069 }
1070
1071 if (h->next == NULL) header_last = h;
1072
1073 /* Check for one of the known header types (From:, To:, etc.) though in
1074 practice most added headers are going to be "other". Lower case
1075 identification letters are never stored with the header; they are used
1076 for existence tests when messages are received. So discard any lower case
1077 flag values. */
1078
1079 h->type = header_checkname(h, FALSE);
1080 if (h->type >= 'a') h->type = htype_other;
1081
1082 DEBUG(D_receive|D_acl) debug_printf(" %s", header_last->text);
1083 }
1084
71fafd95 1085acl_added_headers = NULL;
059ec3d9
PH
1086DEBUG(D_receive|D_acl) debug_printf(">>\n");
1087}
1088
1089
1090
1091/*************************************************
1092* Add host information for log line *
1093*************************************************/
1094
1095/* Called for acceptance and rejecting log lines. This adds information about
1096the calling host to a string that is being built dynamically.
1097
1098Arguments:
1099 s the dynamic string
1100 sizeptr points to the size variable
1101 ptrptr points to the pointer variable
1102
1103Returns: the extended string
1104*/
1105
1106static uschar *
1107add_host_info_for_log(uschar *s, int *sizeptr, int *ptrptr)
1108{
1109if (sender_fullhost != NULL)
1110 {
1111 s = string_append(s, sizeptr, ptrptr, 2, US" H=", sender_fullhost);
1112 if ((log_extra_selector & LX_incoming_interface) != 0 &&
1113 interface_address != NULL)
1114 {
1115 uschar *ss = string_sprintf(" I=[%s]:%d", interface_address,
1116 interface_port);
1117 s = string_cat(s, sizeptr, ptrptr, ss, Ustrlen(ss));
1118 }
1119 }
1120if (sender_ident != NULL)
1121 s = string_append(s, sizeptr, ptrptr, 2, US" U=", sender_ident);
1122if (received_protocol != NULL)
1123 s = string_append(s, sizeptr, ptrptr, 2, US" P=", received_protocol);
1124return s;
1125}
1126
1127
1128
63955bf2 1129#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1130
1131/*************************************************
54cdb463
PH
1132* Run the MIME ACL on a message *
1133*************************************************/
1134
1135/* This code is in a subroutine so that it can be used for both SMTP
1136and non-SMTP messages. It is called with a non-NULL ACL pointer.
1137
1138Arguments:
1139 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1140 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1141 smtp_reply_ptr Where SMTP reply is being built
1142 blackholed_by_ptr Where "blackholed by" message is being built
1143
1144Returns: TRUE to carry on; FALSE to abandon the message
1145*/
1146
1147static BOOL
1148run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1149 uschar **blackholed_by_ptr)
1150{
1151FILE *mbox_file;
1152uschar rfc822_file_path[2048];
1153unsigned long mbox_size;
1154header_line *my_headerlist;
1155uschar *user_msg, *log_msg;
1156int mime_part_count_buffer = -1;
7156b1ef 1157int rc = OK;
54cdb463
PH
1158
1159memset(CS rfc822_file_path,0,2048);
1160
1161/* check if it is a MIME message */
1162my_headerlist = header_list;
4e88a19f
PH
1163while (my_headerlist != NULL)
1164 {
54cdb463 1165 /* skip deleted headers */
4e88a19f
PH
1166 if (my_headerlist->type == '*')
1167 {
54cdb463
PH
1168 my_headerlist = my_headerlist->next;
1169 continue;
4e88a19f
PH
1170 }
1171 if (strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0)
1172 {
54cdb463
PH
1173 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1174 goto DO_MIME_ACL;
4e88a19f 1175 }
54cdb463 1176 my_headerlist = my_headerlist->next;
4e88a19f 1177 }
54cdb463
PH
1178
1179DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1180return TRUE;
1181
1182DO_MIME_ACL:
1183/* make sure the eml mbox file is spooled up */
8544e77a 1184mbox_file = spool_mbox(&mbox_size, NULL);
54cdb463
PH
1185if (mbox_file == NULL) {
1186 /* error while spooling */
1187 log_write(0, LOG_MAIN|LOG_PANIC,
1188 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1189 Uunlink(spool_name);
1190 unspool_mbox();
6f0c431a
PP
1191#ifdef EXPERIMENTAL_DCC
1192 dcc_ok = 0;
1193#endif
a5bd321b 1194 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1195 message_id[0] = 0; /* Indicate no message accepted */
1196 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1197 return FALSE; /* Indicate skip to end of receive function */
1198};
1199
1200mime_is_rfc822 = 0;
1201
1202MIME_ACL_CHECK:
1203mime_part_count = -1;
1204rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1205(void)fclose(mbox_file);
54cdb463 1206
4e88a19f
PH
1207if (Ustrlen(rfc822_file_path) > 0)
1208 {
54cdb463
PH
1209 mime_part_count = mime_part_count_buffer;
1210
4e88a19f
PH
1211 if (unlink(CS rfc822_file_path) == -1)
1212 {
54cdb463
PH
1213 log_write(0, LOG_PANIC,
1214 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1215 goto END_MIME_ACL;
4e88a19f
PH
1216 }
1217 }
54cdb463
PH
1218
1219/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1220if (rc == OK)
1221 {
54cdb463
PH
1222 uschar temp_path[1024];
1223 int n;
1224 struct dirent *entry;
1225 DIR *tempdir;
1226
b07e6aa3
PH
1227 (void)string_format(temp_path, 1024, "%s/scan/%s", spool_directory,
1228 message_id);
54cdb463 1229
4e88a19f
PH
1230 tempdir = opendir(CS temp_path);
1231 n = 0;
1232 do
1233 {
1234 entry = readdir(tempdir);
1235 if (entry == NULL) break;
1236 if (strncmpic(US entry->d_name,US"__rfc822_",9) == 0)
1237 {
b07e6aa3 1238 (void)string_format(rfc822_file_path, 2048,"%s/scan/%s/%s", spool_directory, message_id, entry->d_name);
4e88a19f
PH
1239 debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n", rfc822_file_path);
1240 break;
1241 }
1242 } while (1);
1243 closedir(tempdir);
54cdb463 1244
4e88a19f
PH
1245 if (entry != NULL)
1246 {
c58b88df 1247 mbox_file = Ufopen(rfc822_file_path,"rb");
4e88a19f
PH
1248 if (mbox_file == NULL)
1249 {
54cdb463
PH
1250 log_write(0, LOG_PANIC,
1251 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1252 unlink(CS rfc822_file_path);
1253 goto END_MIME_ACL;
4e88a19f 1254 }
54cdb463
PH
1255 /* set RFC822 expansion variable */
1256 mime_is_rfc822 = 1;
1257 mime_part_count_buffer = mime_part_count;
1258 goto MIME_ACL_CHECK;
4e88a19f
PH
1259 }
1260 }
54cdb463
PH
1261
1262END_MIME_ACL:
1263add_acl_headers(US"MIME");
1264if (rc == DISCARD)
1265 {
1266 recipients_count = 0;
1267 *blackholed_by_ptr = US"MIME ACL";
1268 }
1269else if (rc != OK)
1270 {
1271 Uunlink(spool_name);
1272 unspool_mbox();
6f0c431a
PP
1273#ifdef EXPERIMENTAL_DCC
1274 dcc_ok = 0;
1275#endif
54cdb463
PH
1276 if (smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1277 *smtp_yield_ptr = FALSE; /* No more messsages after dropped connection */
1278 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1279 message_id[0] = 0; /* Indicate no message accepted */
1280 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1281 }
54cdb463
PH
1282
1283return TRUE;
1284}
1285
63955bf2 1286#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1287
1288
e4bdf652
JH
1289
1290void
1291received_header_gen(void)
1292{
1293uschar *received;
1294uschar *timestamp;
1295header_line *received_header= header_list;
1296
1297timestamp = expand_string(US"${tod_full}");
1298if (recipients_count == 1) received_for = recipients_list[0].address;
1299received = expand_string(received_header_text);
1300received_for = NULL;
1301
1302if (received == NULL)
1303 {
1304 if(spool_name[0] != 0)
1305 Uunlink(spool_name); /* Lose the data file */
1306 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1307 "(received_header_text) failed: %s", string_printing(received_header_text),
1308 expand_string_message);
1309 }
1310
1311/* The first element on the header chain is reserved for the Received header,
1312so all we have to do is fill in the text pointer, and set the type. However, if
1313the result of the expansion is an empty string, we leave the header marked as
1314"old" so as to refrain from adding a Received header. */
1315
1316if (received[0] == 0)
1317 {
1318 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1319 received_header->type = htype_old;
1320 }
1321else
1322 {
1323 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1324 received_header->type = htype_received;
1325 }
1326
1327received_header->slen = Ustrlen(received_header->text);
1328
1329DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1330 received_header->type, received_header->text);
1331}
1332
1333
1334
54cdb463 1335/*************************************************
059ec3d9
PH
1336* Receive message *
1337*************************************************/
1338
1339/* Receive a message on the given input, and put it into a pair of spool files.
1340Either a non-null list of recipients, or the extract flag will be true, or
1341both. The flag sender_local is true for locally generated messages. The flag
1342submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1343suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1344suppress_local_fixups" or -G was passed on the command-line.
1345The flag smtp_input is true if the message is to be
8800895a
PH
1346handled using SMTP conventions about termination and lines starting with dots.
1347For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1348
1349If a message was successfully read, message_id[0] will be non-zero.
1350
1351The general actions of this function are:
1352
1353 . Read the headers of the message (if any) into a chain of store
1354 blocks.
1355
1356 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1357 throw it away, unless the caller is trusted, or unless
1358 active_local_sender_retain is set - which can only happen if
1359 active_local_from_check is false.
059ec3d9
PH
1360
1361 . If recipients are to be extracted from the message, build the
1362 recipients list from the headers, removing any that were on the
1363 original recipients list (unless extract_addresses_remove_arguments is
1364 false), and at the same time, remove any bcc header that may be present.
1365
1366 . Get the spool file for the data, sort out its unique name, open
1367 and lock it (but don't give it the name yet).
1368
1369 . Generate a "Message-Id" header if the message doesn't have one, for
1370 locally-originated messages.
1371
1372 . Generate a "Received" header.
1373
1374 . Ensure the recipients list is fully qualified and rewritten if necessary.
1375
1376 . If there are any rewriting rules, apply them to the sender address
1377 and also to the headers.
1378
1379 . If there is no from: header, generate one, for locally-generated messages
1380 and messages in "submission mode" only.
1381
1382 . If the sender is local, check that from: is correct, and if not, generate
1383 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1384 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1385
1386 . If there is no "date" header, generate one, for locally-originated
1387 or submission mode messages only.
1388
1389 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1390 dot_ends mode, to the data file. Leave it open, to hold the lock.
1391
1392 . Write the envelope and the headers to a new file.
1393
1394 . Set the name for the header file; close it.
1395
1396 . Set the name for the data file; close it.
1397
1398Because this function can potentially be called many times in a single
1399SMTP connection, all store should be got by store_get(), so that it will be
1400automatically retrieved after the message is accepted.
1401
1402FUDGE: It seems that sites on the net send out messages with just LF
1403terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1404we make the CRs optional in all cases.
1405
1406July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1407new regime is now in place in which bare CRs in header lines are turned into LF
1408followed by a space, so as not to terminate the header line.
1409
1410February 2004: A bare LF in a header line in a message whose first line was
1411terminated by CRLF is treated in the same way as a bare CR.
1412
1413Arguments:
1414 extract_recip TRUE if recipients are to be extracted from the message's
1415 headers
1416
1417Returns: TRUE there are more messages to be read (SMTP input)
1418 FALSE there are no more messages to be read (non-SMTP input
1419 or SMTP connection collapsed, or other failure)
1420
1421When reading a message for filter testing, the returned value indicates
1422whether the headers (which is all that is read) were terminated by '.' or
1423not. */
1424
1425BOOL
1426receive_msg(BOOL extract_recip)
1427{
7156b1ef
NM
1428int i;
1429int rc = FAIL;
059ec3d9
PH
1430int msg_size = 0;
1431int process_info_len = Ustrlen(process_info);
1432int error_rc = (error_handling == ERRORS_SENDER)?
1433 errors_sender_rc : EXIT_FAILURE;
1434int header_size = 256;
1435int start, end, domain, size, sptr;
1436int id_resolution;
1437int had_zero = 0;
d677b2f2 1438int prevlines_length = 0;
059ec3d9
PH
1439
1440register int ptr = 0;
1441
1442BOOL contains_resent_headers = FALSE;
1443BOOL extracted_ignored = FALSE;
1444BOOL first_line_ended_crlf = TRUE_UNSET;
1445BOOL smtp_yield = TRUE;
1446BOOL yield = FALSE;
1447
1448BOOL resents_exist = FALSE;
1449uschar *resent_prefix = US"";
1450uschar *blackholed_by = NULL;
04f7d5b9 1451uschar *blackhole_log_msg = US"";
4ea50c5f 1452int cutthrough_done = 0;
059ec3d9
PH
1453
1454flock_t lock_data;
1455error_block *bad_addresses = NULL;
1456
1457uschar *frozen_by = NULL;
1458uschar *queued_by = NULL;
1459
1460uschar *errmsg, *s;
1461struct stat statbuf;
1462
4e88a19f 1463/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1464
1465uschar *smtp_reply = NULL;
4e88a19f 1466uschar *user_msg, *log_msg;
059ec3d9
PH
1467
1468/* Working header pointers */
1469
1470header_line *h, *next;
1471
2cbb4081 1472/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1473
1474BOOL date_header_exists = FALSE;
1475
1476/* Pointers to receive the addresses of headers whose contents we need. */
1477
1478header_line *from_header = NULL;
1479header_line *subject_header = NULL;
1480header_line *msgid_header = NULL;
1481header_line *received_header;
1482
1483/* Variables for use when building the Received: header. */
1484
059ec3d9
PH
1485uschar *timestamp;
1486int tslen;
1487
1488/* Release any open files that might have been cached while preparing to
1489accept the message - e.g. by verifying addresses - because reading a message
1490might take a fair bit of real time. */
1491
1492search_tidyup();
1493
e4bdf652
JH
1494/* Extracting the recipient list from an input file is incompatible with
1495cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1496to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1497if (extract_recip || !smtp_input)
2e5b33cd 1498 cancel_cutthrough_connection("not smtp input");
e4bdf652 1499
059ec3d9
PH
1500/* Initialize the chain of headers by setting up a place-holder for Received:
1501header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1502pointing to the end of the chain to make adding headers simple. */
1503
1504received_header = header_list = header_last = store_get(sizeof(header_line));
1505header_list->next = NULL;
1506header_list->type = htype_old;
1507header_list->text = NULL;
1508header_list->slen = 0;
1509
1510/* Control block for the next header to be read. */
1511
1512next = store_get(sizeof(header_line));
1513next->text = store_get(header_size);
1514
1515/* Initialize message id to be null (indicating no message read), and the
1516header names list to be the normal list. Indicate there is no data file open
1517yet, initialize the size and warning count, and deal with no size limit. */
1518
1519message_id[0] = 0;
1520data_file = NULL;
1521data_fd = -1;
1522spool_name[0] = 0;
1523message_size = 0;
1524warning_count = 0;
d677b2f2 1525received_count = 1; /* For the one we will add */
059ec3d9
PH
1526
1527if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1528
2e0c1448 1529/* While reading the message, the following counts are computed. */
059ec3d9 1530
d677b2f2
PH
1531message_linecount = body_linecount = body_zerocount =
1532 max_received_linelength = 0;
059ec3d9 1533
80a47a2c
TK
1534#ifndef DISABLE_DKIM
1535/* Call into DKIM to set up the context. */
1536if (smtp_input && !smtp_batched_input && !dkim_disable_verify) dkim_exim_verify_init();
fb2274d4
TK
1537#endif
1538
059ec3d9
PH
1539/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1540ids, and fractions of a second are required. See the comments that precede the
1541message id creation below. */
1542
1543(void)gettimeofday(&message_id_tv, NULL);
1544
1545/* For other uses of the received time we can operate with granularity of one
1546second, and for that we use the global variable received_time. This is for
1547things like ultimate message timeouts. */
1548
1549received_time = message_id_tv.tv_sec;
1550
1551/* If SMTP input, set the special handler for timeouts. The alarm() calls
1552happen in the smtp_getc() function when it refills its buffer. */
1553
1554if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1555
1556/* If not SMTP input, timeout happens only if configured, and we just set a
1557single timeout for the whole message. */
1558
1559else if (receive_timeout > 0)
1560 {
1561 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1562 alarm(receive_timeout);
1563 }
1564
1565/* SIGTERM and SIGINT are caught always. */
1566
1567signal(SIGTERM, data_sigterm_sigint_handler);
1568signal(SIGINT, data_sigterm_sigint_handler);
1569
1570/* Header lines in messages are not supposed to be very long, though when
1571unfolded, to: and cc: headers can take up a lot of store. We must also cope
1572with the possibility of junk being thrown at us. Start by getting 256 bytes for
1573storing the header, and extend this as necessary using string_cat().
1574
1575To cope with total lunacies, impose an upper limit on the length of the header
1576section of the message, as otherwise the store will fill up. We must also cope
1577with the possibility of binary zeros in the data. Hence we cannot use fgets().
1578Folded header lines are joined into one string, leaving the '\n' characters
1579inside them, so that writing them out reproduces the input.
1580
1581Loop for each character of each header; the next structure for chaining the
1582header is set up already, with ptr the offset of the next character in
1583next->text. */
1584
1585for (;;)
1586 {
80a47a2c 1587 int ch = (receive_getc)();
059ec3d9
PH
1588
1589 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1590 SMTP must have a correct "." terminator. */
1591
1592 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1593 {
1594 smtp_reply = handle_lost_connection(US" (header)");
1595 smtp_yield = FALSE;
1596 goto TIDYUP; /* Skip to end of function */
1597 }
1598
1599 /* See if we are at the current header's size limit - there must be at least
1600 four bytes left. This allows for the new character plus a zero, plus two for
1601 extra insertions when we are playing games with dots and carriage returns. If
1602 we are at the limit, extend the text buffer. This could have been done
1603 automatically using string_cat() but because this is a tightish loop storing
1604 only one character at a time, we choose to do it inline. Normally
1605 store_extend() will be able to extend the block; only at the end of a big
1606 store block will a copy be needed. To handle the case of very long headers
1607 (and sometimes lunatic messages can have ones that are 100s of K long) we
1608 call store_release() for strings that have been copied - if the string is at
1609 the start of a block (and therefore the only thing in it, because we aren't
1610 doing any other gets), the block gets freed. We can only do this because we
1611 know there are no other calls to store_get() going on. */
1612
1613 if (ptr >= header_size - 4)
1614 {
1615 int oldsize = header_size;
1616 /* header_size += 256; */
1617 header_size *= 2;
1618 if (!store_extend(next->text, oldsize, header_size))
1619 {
1620 uschar *newtext = store_get(header_size);
1621 memcpy(newtext, next->text, ptr);
1622 store_release(next->text);
1623 next->text = newtext;
1624 }
1625 }
1626
1627 /* Cope with receiving a binary zero. There is dispute about whether
1628 these should be allowed in RFC 822 messages. The middle view is that they
1629 should not be allowed in headers, at least. Exim takes this attitude at
1630 the moment. We can't just stomp on them here, because we don't know that
1631 this line is a header yet. Set a flag to cause scanning later. */
1632
1633 if (ch == 0) had_zero++;
1634
1635 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1636 those from data files use just LF. Treat LF in local SMTP input as a
1637 terminator too. Treat EOF as a line terminator always. */
1638
1639 if (ch == EOF) goto EOL;
1640
1641 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1642 other MTAs accept this. We are therefore forced into this "liberalisation"
1643 too, so we accept LF as a line terminator whatever the source of the message.
1644 However, if the first line of the message ended with a CRLF, we treat a bare
1645 LF specially by inserting a white space after it to ensure that the header
1646 line is not terminated. */
1647
1648 if (ch == '\n')
1649 {
1650 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1651 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1652 goto EOL;
1653 }
1654
1655 /* This is not the end of the line. If this is SMTP input and this is
1656 the first character in the line and it is a "." character, ignore it.
1657 This implements the dot-doubling rule, though header lines starting with
1658 dots aren't exactly common. They are legal in RFC 822, though. If the
1659 following is CRLF or LF, this is the line that that terminates the
1660 entire message. We set message_ended to indicate this has happened (to
1661 prevent further reading), and break out of the loop, having freed the
1662 empty header, and set next = NULL to indicate no data line. */
1663
1664 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1665 {
80a47a2c 1666 ch = (receive_getc)();
059ec3d9
PH
1667 if (ch == '\r')
1668 {
80a47a2c 1669 ch = (receive_getc)();
059ec3d9
PH
1670 if (ch != '\n')
1671 {
80a47a2c 1672 receive_ungetc(ch);
059ec3d9
PH
1673 ch = '\r'; /* Revert to CR */
1674 }
1675 }
1676 if (ch == '\n')
1677 {
1678 message_ended = END_DOT;
1679 store_reset(next);
1680 next = NULL;
1681 break; /* End character-reading loop */
1682 }
1683
1684 /* For non-SMTP input, the dot at the start of the line was really a data
1685 character. What is now in ch is the following character. We guaranteed
1686 enough space for this above. */
1687
1688 if (!smtp_input)
1689 {
1690 next->text[ptr++] = '.';
1691 message_size++;
1692 }
1693 }
1694
1695 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1696 remember this case if this is the first line ending. */
1697
1698 if (ch == '\r')
1699 {
80a47a2c 1700 ch = (receive_getc)();
059ec3d9
PH
1701 if (ch == '\n')
1702 {
1703 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1704 goto EOL;
1705 }
1706
1707 /* Otherwise, put back the character after CR, and turn the bare CR
1708 into LF SP. */
1709
80a47a2c 1710 ch = (receive_ungetc)(ch);
059ec3d9
PH
1711 next->text[ptr++] = '\n';
1712 message_size++;
1713 ch = ' ';
1714 }
1715
1716 /* We have a data character for the header line. */
1717
1718 next->text[ptr++] = ch; /* Add to buffer */
1719 message_size++; /* Total message size so far */
1720
1721 /* Handle failure due to a humungously long header section. The >= allows
1722 for the terminating \n. Add what we have so far onto the headers list so
1723 that it gets reflected in any error message, and back up the just-read
1724 character. */
1725
1726 if (message_size >= header_maxsize)
1727 {
1728 next->text[ptr] = 0;
1729 next->slen = ptr;
1730 next->type = htype_other;
1731 next->next = NULL;
1732 header_last->next = next;
1733 header_last = next;
1734
1735 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1736 "%s (more than %d characters): message abandoned",
1737 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1738
1739 if (smtp_input)
1740 {
1741 smtp_reply = US"552 Message header is ridiculously long";
1742 receive_swallow_smtp();
1743 goto TIDYUP; /* Skip to end of function */
1744 }
1745
1746 else
1747 {
1748 give_local_error(ERRMESS_VLONGHEADER,
1749 string_sprintf("message header longer than %d characters received: "
1750 "message not accepted", header_maxsize), US"", error_rc, stdin,
1751 header_list->next);
1752 /* Does not return */
1753 }
1754 }
1755
1756 continue; /* With next input character */
1757
1758 /* End of header line reached */
1759
1760 EOL:
2e0c1448
PH
1761
1762 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1763
1764 receive_linecount++;
1765 message_linecount++;
059ec3d9 1766
d677b2f2
PH
1767 /* Keep track of maximum line length */
1768
1769 if (ptr - prevlines_length > max_received_linelength)
1770 max_received_linelength = ptr - prevlines_length;
1771 prevlines_length = ptr + 1;
1772
059ec3d9
PH
1773 /* Now put in the terminating newline. There is always space for
1774 at least two more characters. */
1775
1776 next->text[ptr++] = '\n';
1777 message_size++;
1778
1779 /* A blank line signals the end of the headers; release the unwanted
1780 space and set next to NULL to indicate this. */
1781
1782 if (ptr == 1)
1783 {
1784 store_reset(next);
1785 next = NULL;
1786 break;
1787 }
1788
1789 /* There is data in the line; see if the next input character is a
1790 whitespace character. If it is, we have a continuation of this header line.
1791 There is always space for at least one character at this point. */
1792
1793 if (ch != EOF)
1794 {
80a47a2c 1795 int nextch = (receive_getc)();
059ec3d9
PH
1796 if (nextch == ' ' || nextch == '\t')
1797 {
1798 next->text[ptr++] = nextch;
1799 message_size++;
1800 continue; /* Iterate the loop */
1801 }
80a47a2c 1802 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
1803 else ch = EOF; /* Cause main loop to exit at end */
1804 }
1805
1806 /* We have got to the real line end. Terminate the string and release store
1807 beyond it. If it turns out to be a real header, internal binary zeros will
1808 be squashed later. */
1809
1810 next->text[ptr] = 0;
1811 next->slen = ptr;
1812 store_reset(next->text + ptr + 1);
1813
1814 /* Check the running total size against the overall message size limit. We
1815 don't expect to fail here, but if the overall limit is set less than MESSAGE_
1816 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
1817 headers - the code to read the body will then also hit the buffer. */
1818
1819 if (message_size > thismessage_size_limit) break;
1820
1821 /* A line that is not syntactically correct for a header also marks
1822 the end of the headers. In this case, we leave next containing the
1823 first data line. This might actually be several lines because of the
1824 continuation logic applied above, but that doesn't matter.
1825
1826 It turns out that smail, and presumably sendmail, accept leading lines
1827 of the form
1828
1829 From ph10 Fri Jan 5 12:35 GMT 1996
1830
1831 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
1832 find any documentation of this, but for compatibility it had better be
1833 accepted. Exim restricts it to the case of non-smtp messages, and
1834 treats it as an alternative to the -f command line option. Thus it is
1835 ignored except for trusted users or filter testing. Otherwise it is taken
1836 as the sender address, unless -f was used (sendmail compatibility).
1837
1838 It further turns out that some UUCPs generate the From_line in a different
1839 format, e.g.
1840
1841 From ph10 Fri, 7 Jan 97 14:00:00 GMT
1842
1843 The regex for matching these things is now capable of recognizing both
1844 formats (including 2- and 4-digit years in the latter). In fact, the regex
1845 is now configurable, as is the expansion string to fish out the sender.
1846
1847 Even further on it has been discovered that some broken clients send
1848 these lines in SMTP messages. There is now an option to ignore them from
1849 specified hosts or networks. Sigh. */
1850
1851 if (header_last == header_list &&
1852 (!smtp_input
1853 ||
1854 (sender_host_address != NULL &&
1855 verify_check_host(&ignore_fromline_hosts) == OK)
1856 ||
1857 (sender_host_address == NULL && ignore_fromline_local)
1858 ) &&
1859 regex_match_and_setup(regex_From, next->text, 0, -1))
1860 {
1861 if (!sender_address_forced)
1862 {
1863 uschar *uucp_sender = expand_string(uucp_from_sender);
1864 if (uucp_sender == NULL)
1865 {
1866 log_write(0, LOG_MAIN|LOG_PANIC,
1867 "expansion of \"%s\" failed after matching "
1868 "\"From \" line: %s", uucp_from_sender, expand_string_message);
1869 }
1870 else
1871 {
1872 int start, end, domain;
1873 uschar *errmess;
1874 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
1875 &start, &end, &domain, TRUE);
1876 if (newsender != NULL)
1877 {
1878 if (domain == 0 && newsender[0] != 0)
1879 newsender = rewrite_address_qualify(newsender, FALSE);
1880
f05da2e8 1881 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
1882 {
1883 sender_address = newsender;
1884
f05da2e8 1885 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
1886 {
1887 authenticated_sender = NULL;
1888 originator_name = US"";
1889 sender_local = FALSE;
1890 }
1891
f05da2e8 1892 if (filter_test != FTEST_NONE)
059ec3d9
PH
1893 printf("Sender taken from \"From \" line\n");
1894 }
1895 }
1896 }
1897 }
1898 }
1899
1900 /* Not a leading "From " line. Check to see if it is a valid header line.
1901 Header names may contain any non-control characters except space and colon,
1902 amazingly. */
1903
1904 else
1905 {
1906 uschar *p = next->text;
1907
1908 /* If not a valid header line, break from the header reading loop, leaving
1909 next != NULL, indicating that it holds the first line of the body. */
1910
1911 if (isspace(*p)) break;
1912 while (mac_isgraph(*p) && *p != ':') p++;
1913 while (isspace(*p)) p++;
1914 if (*p != ':')
1915 {
1916 body_zerocount = had_zero;
1917 break;
1918 }
1919
1920 /* We have a valid header line. If there were any binary zeroes in
1921 the line, stomp on them here. */
1922
1923 if (had_zero > 0)
1924 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
1925
1926 /* It is perfectly legal to have an empty continuation line
1927 at the end of a header, but it is confusing to humans
1928 looking at such messages, since it looks like a blank line.
1929 Reduce confusion by removing redundant white space at the
1930 end. We know that there is at least one printing character
1931 (the ':' tested for above) so there is no danger of running
1932 off the end. */
1933
1934 p = next->text + ptr - 2;
1935 for (;;)
1936 {
1937 while (*p == ' ' || *p == '\t') p--;
1938 if (*p != '\n') break;
1939 ptr = (p--) - next->text + 1;
1940 message_size -= next->slen - ptr;
1941 next->text[ptr] = 0;
1942 next->slen = ptr;
1943 }
1944
1945 /* Add the header to the chain */
1946
1947 next->type = htype_other;
1948 next->next = NULL;
1949 header_last->next = next;
1950 header_last = next;
1951
1952 /* Check the limit for individual line lengths. This comes after adding to
1953 the chain so that the failing line is reflected if a bounce is generated
1954 (for a local message). */
1955
1956 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
1957 {
1958 log_write(0, LOG_MAIN, "overlong message header line received from "
1959 "%s (more than %d characters): message abandoned",
1960 sender_host_unknown? sender_ident : sender_fullhost,
1961 header_line_maxsize);
1962
1963 if (smtp_input)
1964 {
1965 smtp_reply = US"552 A message header line is too long";
1966 receive_swallow_smtp();
1967 goto TIDYUP; /* Skip to end of function */
1968 }
1969
1970 else
1971 {
1972 give_local_error(ERRMESS_VLONGHDRLINE,
1973 string_sprintf("message header line longer than %d characters "
1974 "received: message not accepted", header_line_maxsize), US"",
1975 error_rc, stdin, header_list->next);
1976 /* Does not return */
1977 }
1978 }
1979
1980 /* Note if any resent- fields exist. */
1981
1982 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
1983 {
1984 resents_exist = TRUE;
1985 resent_prefix = US"Resent-";
1986 }
1987 }
1988
1989 /* The line has been handled. If we have hit EOF, break out of the loop,
1990 indicating no pending data line. */
1991
1992 if (ch == EOF) { next = NULL; break; }
1993
1994 /* Set up for the next header */
1995
1996 header_size = 256;
1997 next = store_get(sizeof(header_line));
1998 next->text = store_get(header_size);
1999 ptr = 0;
2000 had_zero = 0;
d677b2f2 2001 prevlines_length = 0;
059ec3d9
PH
2002 } /* Continue, starting to read the next header */
2003
2004/* At this point, we have read all the headers into a data structure in main
2005store. The first header is still the dummy placeholder for the Received: header
2006we are going to generate a bit later on. If next != NULL, it contains the first
2007data line - which terminated the headers before reaching a blank line (not the
2008normal case). */
2009
2010DEBUG(D_receive)
2011 {
2012 debug_printf(">>Headers received:\n");
2013 for (h = header_list->next; h != NULL; h = h->next)
2014 debug_printf("%s", h->text);
2015 debug_printf("\n");
2016 }
2017
2018/* End of file on any SMTP connection is an error. If an incoming SMTP call
2019is dropped immediately after valid headers, the next thing we will see is EOF.
2020We must test for this specially, as further down the reading of the data is
2021skipped if already at EOF. */
2022
2023if (smtp_input && (receive_feof)())
2024 {
2025 smtp_reply = handle_lost_connection(US" (after header)");
2026 smtp_yield = FALSE;
2027 goto TIDYUP; /* Skip to end of function */
2028 }
2029
2030/* If this is a filter test run and no headers were read, output a warning
2031in case there is a mistake in the test message. */
2032
f05da2e8 2033if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2034 printf("Warning: no message headers read\n");
2035
2036
2037/* Scan the headers to identify them. Some are merely marked for later
2038processing; some are dealt with here. */
2039
2040for (h = header_list->next; h != NULL; h = h->next)
2041 {
2042 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2043 if (is_resent) contains_resent_headers = TRUE;
2044
2045 switch (header_checkname(h, is_resent))
2046 {
059ec3d9 2047 case htype_bcc:
2cbb4081 2048 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
2049 break;
2050
059ec3d9 2051 case htype_cc:
2cbb4081 2052 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
2053 break;
2054
2055 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
2056
2057 case htype_date:
4c69d561 2058 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
2059 break;
2060
2061 /* Same comments as about Return-Path: below. */
2062
2063 case htype_delivery_date:
2064 if (delivery_date_remove) h->type = htype_old;
2065 break;
2066
2067 /* Same comments as about Return-Path: below. */
2068
2069 case htype_envelope_to:
2070 if (envelope_to_remove) h->type = htype_old;
2071 break;
2072
2073 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2074 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2075 header consists of just the login id of the user who called Exim, rewrite
2076 it with the gecos field first. Apply this rule to Resent-From: if there
2077 are resent- fields. */
2078
2079 case htype_from:
2080 h->type = htype_from;
2081 if (!resents_exist || is_resent)
2082 {
2083 from_header = h;
2084 if (!smtp_input)
2085 {
5de8faa3 2086 int len;
059ec3d9
PH
2087 uschar *s = Ustrchr(h->text, ':') + 1;
2088 while (isspace(*s)) s++;
5de8faa3 2089 len = h->slen - (s - h->text) - 1;
e0fccd1d
TF
2090 if (Ustrlen(originator_login) == len &&
2091 strncmpic(s, originator_login, len) == 0)
059ec3d9
PH
2092 {
2093 uschar *name = is_resent? US"Resent-From" : US"From";
2094 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2095 originator_login, qualify_domain_sender);
2096 from_header = header_last;
2097 h->type = htype_old;
2098 DEBUG(D_receive|D_rewrite)
2099 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2100 }
2101 }
2102 }
2103 break;
2104
2105 /* Identify the Message-id: header for generating "in-reply-to" in the
2106 autoreply transport. For incoming logging, save any resent- value. In both
2107 cases, take just the first of any multiples. */
2108
2109 case htype_id:
2110 if (msgid_header == NULL && (!resents_exist || is_resent))
2111 {
2112 msgid_header = h;
2113 h->type = htype_id;
2114 }
2115 break;
2116
2117 /* Flag all Received: headers */
2118
2119 case htype_received:
2120 h->type = htype_received;
2121 received_count++;
2122 break;
2123
2124 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
2125
2126 case htype_reply_to:
2127 h->type = htype_reply_to;
2128 break;
2129
2130 /* The Return-path: header is supposed to be added to messages when
2131 they leave the SMTP system. We shouldn't receive messages that already
2132 contain Return-path. However, since Exim generates Return-path: on
2133 local delivery, resent messages may well contain it. We therefore
2134 provide an option (which defaults on) to remove any Return-path: headers
2135 on input. Removal actually means flagging as "old", which prevents the
2136 header being transmitted with the message. */
2137
2138 case htype_return_path:
2139 if (return_path_remove) h->type = htype_old;
2140
2141 /* If we are testing a mail filter file, use the value of the
2142 Return-Path: header to set up the return_path variable, which is not
2143 otherwise set. However, remove any <> that surround the address
2144 because the variable doesn't have these. */
2145
f05da2e8 2146 if (filter_test != FTEST_NONE)
059ec3d9
PH
2147 {
2148 uschar *start = h->text + 12;
2149 uschar *end = start + Ustrlen(start);
2150 while (isspace(*start)) start++;
2151 while (end > start && isspace(end[-1])) end--;
2152 if (*start == '<' && end[-1] == '>')
2153 {
2154 start++;
2155 end--;
2156 }
2157 return_path = string_copyn(start, end - start);
2158 printf("Return-path taken from \"Return-path:\" header line\n");
2159 }
2160 break;
2161
2162 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2163 and from an untrusted caller and suppress_local_fixups is not set, or if we
2164 are in submission mode for a remote message, mark it "old" so that it will
2165 not be transmitted with the message, unless active_local_sender_retain is
2166 set. (This can only be true if active_local_from_check is false.) If there
2167 are any resent- headers in the message, apply this rule to Resent-Sender:
2168 instead of Sender:. Messages with multiple resent- header sets cannot be
2169 tidily handled. (For this reason, at least one MUA - Pine - turns old
2170 resent- headers into X-resent- headers when resending, leaving just one
2171 set.) */
059ec3d9
PH
2172
2173 case htype_sender:
69358f02 2174 h->type = ((!active_local_sender_retain &&
8800895a
PH
2175 (
2176 (sender_local && !trusted_caller && !suppress_local_fixups)
2177 || submission_mode
2178 )
059ec3d9
PH
2179 ) &&
2180 (!resents_exist||is_resent))?
2181 htype_old : htype_sender;
2182 break;
2183
2184 /* Remember the Subject: header for logging. There is no Resent-Subject */
2185
2186 case htype_subject:
2187 subject_header = h;
2188 break;
2189
2190 /* "To:" gets flagged, and the existence of a recipient header is noted,
2191 whether it's resent- or not. */
2192
2193 case htype_to:
2194 h->type = htype_to;
2195 /****
2196 to_or_cc_header_exists = TRUE;
2197 ****/
2198 break;
2199 }
2200 }
2201
2202/* Extract recipients from the headers if that is required (the -t option).
2203Note that this is documented as being done *before* any address rewriting takes
2204place. There are two possibilities:
2205
2206(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2207recipients already listed are to be REMOVED from the message. Smail 3 works
2208like this. We need to build a non-recipients tree for that list, because in
2209subsequent processing this data is held in a tree and that's what the
2210spool_write_header() function expects. Make sure that non-recipient addresses
2211are fully qualified and rewritten if necessary.
2212
2213(2) According to other sendmail documentation, -t ADDS extracted recipients to
2214those in the command line arguments (and it is rumoured some other MTAs do
2215this). Therefore, there is an option to make Exim behave this way.
2216
2217*** Notes on "Resent-" header lines ***
2218
2219The presence of resent-headers in the message makes -t horribly ambiguous.
2220Experiments with sendmail showed that it uses recipients for all resent-
2221headers, totally ignoring the concept of "sets of resent- headers" as described
2222in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2223with all the addresses in one instance of each header.
2224
2225This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2226error for -t if there were resent- headers in the message. However, after a
2227discussion on the mailing list, I've learned that there are MUAs that use
2228resent- headers with -t, and also that the stuff about sets of resent- headers
2229and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2230message with -t and resent- header lines makes sure that only *its* resent-
2231headers are present; previous ones are often renamed as X-resent- for example.
2232
2233Consequently, Exim has been changed so that, if any resent- header lines are
2234present, the recipients are taken from all of the appropriate resent- lines,
2235and not from the ordinary To:, Cc:, etc. */
2236
2237if (extract_recip)
2238 {
2239 int rcount = 0;
2240 error_block **bnext = &bad_addresses;
2241
2242 if (extract_addresses_remove_arguments)
2243 {
2244 while (recipients_count-- > 0)
2245 {
2246 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2247 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2248 tree_add_nonrecipient(s);
2249 }
2250 recipients_list = NULL;
2251 recipients_count = recipients_list_max = 0;
2252 }
2253
059ec3d9
PH
2254 /* Now scan the headers */
2255
2256 for (h = header_list->next; h != NULL; h = h->next)
2257 {
2258 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2259 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2260 {
2261 uschar *s = Ustrchr(h->text, ':') + 1;
2262 while (isspace(*s)) s++;
2263
1eccaa59
PH
2264 parse_allow_group = TRUE; /* Allow address group syntax */
2265
059ec3d9
PH
2266 while (*s != 0)
2267 {
2268 uschar *ss = parse_find_address_end(s, FALSE);
2269 uschar *recipient, *errmess, *p, *pp;
2270 int start, end, domain;
2271
2272 /* Check on maximum */
2273
2274 if (recipients_max > 0 && ++rcount > recipients_max)
2275 {
2276 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2277 US"message rejected: ", error_rc, stdin, NULL);
2278 /* Does not return */
2279 }
2280
2281 /* Make a copy of the address, and remove any internal newlines. These
2282 may be present as a result of continuations of the header line. The
2283 white space that follows the newline must not be removed - it is part
2284 of the header. */
2285
2286 pp = recipient = store_get(ss - s + 1);
2287 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2288 *pp = 0;
2289 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2290 &domain, FALSE);
2291
2292 /* Keep a list of all the bad addresses so we can send a single
2293 error message at the end. However, an empty address is not an error;
2294 just ignore it. This can come from an empty group list like
2295
2296 To: Recipients of list:;
2297
2298 If there are no recipients at all, an error will occur later. */
2299
2300 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2301 {
2302 int len = Ustrlen(s);
2303 error_block *b = store_get(sizeof(error_block));
2304 while (len > 0 && isspace(s[len-1])) len--;
2305 b->next = NULL;
2306 b->text1 = string_printing(string_copyn(s, len));
2307 b->text2 = errmess;
2308 *bnext = b;
2309 bnext = &(b->next);
2310 }
2311
2312 /* If the recipient is already in the nonrecipients tree, it must
2313 have appeared on the command line with the option extract_addresses_
2314 remove_arguments set. Do not add it to the recipients, and keep a note
2315 that this has happened, in order to give a better error if there are
2316 no recipients left. */
2317
2318 else if (recipient != NULL)
2319 {
2320 if (tree_search(tree_nonrecipients, recipient) == NULL)
2321 receive_add_recipient(recipient, -1);
2322 else
2323 extracted_ignored = TRUE;
2324 }
2325
2326 /* Move on past this address */
2327
2328 s = ss + (*ss? 1:0);
2329 while (isspace(*s)) s++;
1eccaa59
PH
2330 } /* Next address */
2331
2332 parse_allow_group = FALSE; /* Reset group syntax flags */
2333 parse_found_group = FALSE;
059ec3d9
PH
2334
2335 /* If this was the bcc: header, mark it "old", which means it
2336 will be kept on the spool, but not transmitted as part of the
2337 message. */
2338
2cbb4081 2339 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2340 } /* For appropriate header line */
2341 } /* For each header line */
2342
059ec3d9
PH
2343 }
2344
2345/* Now build the unique message id. This has changed several times over the
2346lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2347Retaining all the history in the comment has become too unwieldy - read
2348previous release sources if you want it.
2349
2350The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2351The first part is the current time, in seconds. The second part is the current
2352pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2353can hold a number in the range 0-3843. It used to be a computed sequence
2354number, but is now the fractional component of the current time in units of
23551/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2356received, Exim ensures that the timer has ticked at the appropriate level
2357before proceeding, to avoid duplication if the pid happened to be re-used
2358within the same time period. It seems likely that most messages will take at
2359least half a millisecond to be received, so no delay will normally be
2360necessary. At least for some time...
2361
2362There is a modification when localhost_number is set. Formerly this was allowed
2363to be as large as 255. Now it is restricted to the range 0-16, and the final
2364component of the message id becomes (localhost_number * 200) + fractional time
2365in units of 1/200 of a second (i.e. a value in the range 0-3399).
2366
2367Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2368Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2369still allows the tttttt field to hold a large enough number to last for some
2370more decades, and the final two-digit field can hold numbers up to 1295, which
2371is enough for milliseconds (instead of 1/2000 of a second).
2372
2373However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2374pid, so it is probably safe because pids have to be positive. The
2375localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2376final field becomes (localhost_number * 100) + fractional time in centiseconds.
2377
2378Note that string_base62() returns its data in a static storage block, so it
2379must be copied before calling string_base62() again. It always returns exactly
23806 characters.
2381
2382There doesn't seem to be anything in the RFC which requires a message id to
2383start with a letter, but Smail was changed to ensure this. The external form of
2384the message id (as supplied by string expansion) therefore starts with an
2385additional leading 'E'. The spool file names do not include this leading
2386letter and it is not used internally.
2387
2388NOTE: If ever the format of message ids is changed, the regular expression for
2389checking that a string is in this format must be updated in a corresponding
2390way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2391must also be changed to reflect the correct string length. Then, of course,
2392other programs that rely on the message id format will need updating too. */
2393
2394Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2395message_id[6] = '-';
2396Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2397
2398/* Deal with the case where the host number is set. The value of the number was
2399checked when it was read, to ensure it isn't too big. The timing granularity is
2400left in id_resolution so that an appropriate wait can be done after receiving
2401the message, if necessary (we hope it won't be). */
2402
2403if (host_number_string != NULL)
2404 {
2405 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2406 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2407 string_base62((long int)(
2408 host_number * (1000000/id_resolution) +
2409 message_id_tv.tv_usec/id_resolution)) + 4);
2410 }
2411
2412/* Host number not set: final field is just the fractional time at an
2413appropriate resolution. */
2414
2415else
2416 {
2417 id_resolution = (BASE_62 == 62)? 500 : 1000;
2418 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2419 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2420 }
2421
2422/* Add the current message id onto the current process info string if
2423it will fit. */
2424
2425(void)string_format(process_info + process_info_len,
2426 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2427
2428/* If we are using multiple input directories, set up the one for this message
2429to be the least significant base-62 digit of the time of arrival. Otherwise
2430ensure that it is an empty string. */
2431
2432message_subdir[0] = split_spool_directory? message_id[5] : 0;
2433
2434/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2435one, but only for local (without suppress_local_fixups) or submission mode
2436messages. This can be user-configured if required, but we had better flatten
2437any illegal characters therein. */
059ec3d9 2438
8800895a
PH
2439if (msgid_header == NULL &&
2440 ((sender_host_address == NULL && !suppress_local_fixups)
2441 || submission_mode))
059ec3d9
PH
2442 {
2443 uschar *p;
2444 uschar *id_text = US"";
2445 uschar *id_domain = primary_hostname;
2446
2447 /* Permit only letters, digits, dots, and hyphens in the domain */
2448
2449 if (message_id_domain != NULL)
2450 {
2451 uschar *new_id_domain = expand_string(message_id_domain);
2452 if (new_id_domain == NULL)
2453 {
2454 if (!expand_string_forcedfail)
2455 log_write(0, LOG_MAIN|LOG_PANIC,
2456 "expansion of \"%s\" (message_id_header_domain) "
2457 "failed: %s", message_id_domain, expand_string_message);
2458 }
2459 else if (*new_id_domain != 0)
2460 {
2461 id_domain = new_id_domain;
2462 for (p = id_domain; *p != 0; p++)
2463 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2464 }
2465 }
2466
2467 /* Permit all characters except controls and RFC 2822 specials in the
2468 additional text part. */
2469
2470 if (message_id_text != NULL)
2471 {
2472 uschar *new_id_text = expand_string(message_id_text);
2473 if (new_id_text == NULL)
2474 {
2475 if (!expand_string_forcedfail)
2476 log_write(0, LOG_MAIN|LOG_PANIC,
2477 "expansion of \"%s\" (message_id_header_text) "
2478 "failed: %s", message_id_text, expand_string_message);
2479 }
2480 else if (*new_id_text != 0)
2481 {
2482 id_text = new_id_text;
2483 for (p = id_text; *p != 0; p++)
2484 if (mac_iscntrl_or_special(*p)) *p = '-';
2485 }
2486 }
2487
e7e680d6
PP
2488 /* Add the header line
2489 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2490 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2491
e7e680d6 2492 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2493 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2494 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2495 }
2496
2497/* If we are to log recipients, keep a copy of the raw ones before any possible
2498rewriting. Must copy the count, because later ACLs and the local_scan()
2499function may mess with the real recipients. */
2500
2501if ((log_extra_selector & LX_received_recipients) != 0)
2502 {
2503 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2504 for (i = 0; i < recipients_count; i++)
2505 raw_recipients[i] = string_copy(recipients_list[i].address);
2506 raw_recipients_count = recipients_count;
2507 }
2508
2509/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2510recipients will get here only if the conditions were right (allow_unqualified_
2511recipient is TRUE). */
2512
2513for (i = 0; i < recipients_count; i++)
2514 recipients_list[i].address =
2515 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2516 global_rewrite_rules, rewrite_existflags);
2517
8800895a
PH
2518/* If there is no From: header, generate one for local (without
2519suppress_local_fixups) or submission_mode messages. If there is no sender
2520address, but the sender is local or this is a local delivery error, use the
2521originator login. This shouldn't happen for genuine bounces, but might happen
2522for autoreplies. The addition of From: must be done *before* checking for the
2523possible addition of a Sender: header, because untrusted_set_sender allows an
2524untrusted user to set anything in the envelope (which might then get info
2525From:) but we still want to ensure a valid Sender: if it is required. */
2526
2527if (from_header == NULL &&
2528 ((sender_host_address == NULL && !suppress_local_fixups)
2529 || submission_mode))
059ec3d9 2530 {
2fe1a124
PH
2531 uschar *oname = US"";
2532
2533 /* Use the originator_name if this is a locally submitted message and the
2534 caller is not trusted. For trusted callers, use it only if -F was used to
2535 force its value or if we have a non-SMTP message for which -f was not used
2536 to set the sender. */
2537
2538 if (sender_host_address == NULL)
2539 {
2540 if (!trusted_caller || sender_name_forced ||
2541 (!smtp_input && !sender_address_forced))
2542 oname = originator_name;
2543 }
2544
2545 /* For non-locally submitted messages, the only time we use the originator
2546 name is when it was forced by the /name= option on control=submission. */
2547
2548 else
2549 {
2550 if (submission_name != NULL) oname = submission_name;
2551 }
2552
059ec3d9
PH
2553 /* Envelope sender is empty */
2554
2555 if (sender_address[0] == 0)
2556 {
87ba3f5f
PH
2557 uschar *fromstart, *fromend;
2558
2559 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2560 oname, (oname[0] == 0)? "" : " <");
2561 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2562
059ec3d9
PH
2563 if (sender_local || local_error_message)
2564 {
87ba3f5f
PH
2565 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2566 local_part_quote(originator_login), qualify_domain_sender,
2567 fromend);
059ec3d9
PH
2568 }
2569 else if (submission_mode && authenticated_id != NULL)
2570 {
2571 if (submission_domain == NULL)
2572 {
87ba3f5f
PH
2573 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2574 local_part_quote(authenticated_id), qualify_domain_sender,
2575 fromend);
059ec3d9
PH
2576 }
2577 else if (submission_domain[0] == 0) /* empty => whole address set */
2578 {
87ba3f5f
PH
2579 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2580 fromend);
059ec3d9
PH
2581 }
2582 else
2583 {
87ba3f5f
PH
2584 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2585 local_part_quote(authenticated_id), submission_domain,
2586 fromend);
059ec3d9
PH
2587 }
2588 from_header = header_last; /* To get it checked for Sender: */
2589 }
2590 }
2591
2592 /* There is a non-null envelope sender. Build the header using the original
2593 sender address, before any rewriting that might have been done while
2594 verifying it. */
2595
2596 else
2597 {
87ba3f5f 2598 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2599 oname,
2600 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2601 (sender_address_unrewritten == NULL)?
2602 sender_address : sender_address_unrewritten,
2fe1a124 2603 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2604
2605 from_header = header_last; /* To get it checked for Sender: */
2606 }
2607 }
2608
2609
8800895a
PH
2610/* If the sender is local (without suppress_local_fixups), or if we are in
2611submission mode and there is an authenticated_id, check that an existing From:
2612is correct, and if not, generate a Sender: header, unless disabled. Any
2613previously-existing Sender: header was removed above. Note that sender_local,
2614as well as being TRUE if the caller of exim is not trusted, is also true if a
2615trusted caller did not supply a -f argument for non-smtp input. To allow
2616trusted callers to forge From: without supplying -f, we have to test explicitly
2617here. If the From: header contains more than one address, then the call to
2618parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2619
2620if (from_header != NULL &&
69358f02 2621 (active_local_from_check &&
8800895a 2622 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2623 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2624 ))
2625 {
2626 BOOL make_sender = TRUE;
2627 int start, end, domain;
2628 uschar *errmess;
2629 uschar *from_address =
2630 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2631 &start, &end, &domain, FALSE);
2632 uschar *generated_sender_address;
2633
2634 if (submission_mode)
2635 {
2636 if (submission_domain == NULL)
2637 {
2638 generated_sender_address = string_sprintf("%s@%s",
2639 local_part_quote(authenticated_id), qualify_domain_sender);
2640 }
2641 else if (submission_domain[0] == 0) /* empty => full address */
2642 {
2643 generated_sender_address = string_sprintf("%s",
2644 authenticated_id);
2645 }
2646 else
2647 {
2648 generated_sender_address = string_sprintf("%s@%s",
2649 local_part_quote(authenticated_id), submission_domain);
2650 }
2651 }
2652 else
2653 generated_sender_address = string_sprintf("%s@%s",
2654 local_part_quote(originator_login), qualify_domain_sender);
2655
2656 /* Remove permitted prefixes and suffixes from the local part of the From:
2657 address before doing the comparison with the generated sender. */
2658
2659 if (from_address != NULL)
2660 {
2661 int slen;
2662 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2663
2664 if (at != NULL) *at = 0;
2665 from_address += route_check_prefix(from_address, local_from_prefix);
2666 slen = route_check_suffix(from_address, local_from_suffix);
2667 if (slen > 0)
2668 {
2669 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2670 from_address += slen;
2671 }
2672 if (at != NULL) *at = '@';
2673
2674 if (strcmpic(generated_sender_address, from_address) == 0 ||
2675 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2676 make_sender = FALSE;
2677 }
2678
2679 /* We have to cause the Sender header to be rewritten if there are
2680 appropriate rewriting rules. */
2681
2682 if (make_sender)
2683 {
2fe1a124 2684 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2685 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2686 generated_sender_address);
2687 else
2688 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2689 resent_prefix,
2690 submission_mode? submission_name : originator_name,
2691 generated_sender_address);
059ec3d9 2692 }
87ba3f5f
PH
2693
2694 /* Ensure that a non-null envelope sender address corresponds to the
2695 submission mode sender address. */
2696
2697 if (submission_mode && sender_address[0] != 0)
2698 {
2699 if (sender_address_unrewritten == NULL)
2700 sender_address_unrewritten = sender_address;
2701 sender_address = generated_sender_address;
089793a4
TF
2702 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2703 log_write(L_address_rewrite, LOG_MAIN,
2704 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2705 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2706 }
059ec3d9
PH
2707 }
2708
2709
2710/* If there are any rewriting rules, apply them to the sender address, unless
2711it has already been rewritten as part of verification for SMTP input. */
2712
2713if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2714 sender_address[0] != 0)
2715 {
2716 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2717 global_rewrite_rules, rewrite_existflags);
2718 DEBUG(D_receive|D_rewrite)
2719 debug_printf("rewritten sender = %s\n", sender_address);
2720 }
2721
2722
2723/* The headers must be run through rewrite_header(), because it ensures that
2724addresses are fully qualified, as well as applying any rewriting rules that may
2725exist.
2726
2727Qualification of header addresses in a message from a remote host happens only
2728if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2729appropriate. For local messages, qualification always happens, unless -bnq is
2730used to explicitly suppress it. No rewriting is done for an unqualified address
2731that is left untouched.
2732
2733We start at the second header, skipping our own Received:. This rewriting is
2734documented as happening *after* recipient addresses are taken from the headers
2735by the -t command line option. An added Sender: gets rewritten here. */
2736
2737for (h = header_list->next; h != NULL; h = h->next)
2738 {
2739 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2740 rewrite_existflags, TRUE);
2741 if (newh != NULL) h = newh;
2742 }
2743
2744
2745/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2746"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2747"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2748exists, this applies to the set of resent- headers rather than the normal set.
2749
2cbb4081
PH
2750The requirement for a recipient header has been removed in RFC 2822. At this
2751point in the code, earlier versions of Exim added a To: header for locally
2752submitted messages, and an empty Bcc: header for others. In the light of the
2753changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2754
059ec3d9
PH
2755
2756/* If there is no date header, generate one if the message originates locally
8800895a
PH
2757(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2758submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2759to be more confusing if Exim adds one to all remotely-originated messages.
2760As per Message-Id, we prepend if resending, else append.
2761*/
059ec3d9 2762
8800895a
PH
2763if (!date_header_exists &&
2764 ((sender_host_address == NULL && !suppress_local_fixups)
2765 || submission_mode))
e7e680d6
PP
2766 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2767 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2768
2769search_tidyup(); /* Free any cached resources */
2770
2771/* Show the complete set of headers if debugging. Note that the first one (the
2772new Received:) has not yet been set. */
2773
2774DEBUG(D_receive)
2775 {
2776 debug_printf(">>Headers after rewriting and local additions:\n");
2777 for (h = header_list->next; h != NULL; h = h->next)
2778 debug_printf("%c %s", h->type, h->text);
2779 debug_printf("\n");
2780 }
2781
2782/* The headers are now complete in store. If we are running in filter
2783testing mode, that is all this function does. Return TRUE if the message
2784ended with a dot. */
2785
f05da2e8 2786if (filter_test != FTEST_NONE)
059ec3d9
PH
2787 {
2788 process_info[process_info_len] = 0;
2789 return message_ended == END_DOT;
2790 }
2791
817d9f57 2792/* Cutthrough delivery:
e4bdf652
JH
2793 We have to create the Received header now rather than at the end of reception,
2794 so the timestamp behaviour is a change to the normal case.
2795 XXX Ensure this gets documented XXX.
817d9f57 2796 Having created it, send the headers to the destination.
e4bdf652
JH
2797*/
2798if (cutthrough_fd >= 0)
2799 {
817d9f57
JH
2800 if (received_count > received_headers_max)
2801 {
2e5b33cd 2802 cancel_cutthrough_connection("too many headers");
817d9f57
JH
2803 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2804 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2805 "Too many \"Received\" headers",
2806 sender_address,
2807 (sender_fullhost == NULL)? "" : " H=",
2808 (sender_fullhost == NULL)? US"" : sender_fullhost,
2809 (sender_ident == NULL)? "" : " U=",
2810 (sender_ident == NULL)? US"" : sender_ident);
2811 message_id[0] = 0; /* Indicate no message accepted */
2812 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
2813 goto TIDYUP; /* Skip to end of function */
2814 }
e4bdf652
JH
2815 received_header_gen();
2816 add_acl_headers(US"MAIL or RCPT");
2817 (void) cutthrough_headers_send();
2818 }
61147df4 2819
e4bdf652 2820
059ec3d9
PH
2821/* Open a new spool file for the data portion of the message. We need
2822to access it both via a file descriptor and a stream. Try to make the
2823directory if it isn't there. Note re use of sprintf: spool_directory
2824is checked on input to be < 200 characters long. */
2825
2826sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory, message_subdir,
2827 message_id);
2828data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2829if (data_fd < 0)
2830 {
2831 if (errno == ENOENT)
2832 {
2833 uschar temp[16];
2834 sprintf(CS temp, "input/%s", message_subdir);
2835 if (message_subdir[0] == 0) temp[5] = 0;
2836 (void)directory_make(spool_directory, temp, INPUT_DIRECTORY_MODE, TRUE);
2837 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2838 }
2839 if (data_fd < 0)
2840 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
2841 spool_name, strerror(errno));
2842 }
2843
2844/* Make sure the file's group is the Exim gid, and double-check the mode
2845because the group setting doesn't always get set automatically. */
2846
1ac6b2e7
JH
2847if (fchown(data_fd, exim_uid, exim_gid))
2848 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
2849 "Failed setting ownership on spool file %s: %s",
2850 spool_name, strerror(errno));
ff790e47 2851(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
2852
2853/* We now have data file open. Build a stream for it and lock it. We lock only
2854the first line of the file (containing the message ID) because otherwise there
2855are problems when Exim is run under Cygwin (I'm told). See comments in
2856spool_in.c, where the same locking is done. */
2857
2858data_file = fdopen(data_fd, "w+");
2859lock_data.l_type = F_WRLCK;
2860lock_data.l_whence = SEEK_SET;
2861lock_data.l_start = 0;
2862lock_data.l_len = SPOOL_DATA_START_OFFSET;
2863
2864if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
2865 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
2866 errno, strerror(errno));
2867
2868/* We have an open, locked data file. Write the message id to it to make it
2869self-identifying. Then read the remainder of the input of this message and
2870write it to the data file. If the variable next != NULL, it contains the first
2871data line (which was read as a header but then turned out not to have the right
2872format); write it (remembering that it might contain binary zeros). The result
2873of fwrite() isn't inspected; instead we call ferror() below. */
2874
2875fprintf(data_file, "%s-D\n", message_id);
2876if (next != NULL)
2877 {
2878 uschar *s = next->text;
2879 int len = next->slen;
1ac6b2e7 2880 len = fwrite(s, 1, len, data_file); len = len; /* compiler quietening */
059ec3d9
PH
2881 body_linecount++; /* Assumes only 1 line */
2882 }
2883
2884/* Note that we might already be at end of file, or the logical end of file
2885(indicated by '.'), or might have encountered an error while writing the
2886message id or "next" line. */
2887
2888if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
2889 {
2890 if (smtp_input)
2891 {
2892 message_ended = read_message_data_smtp(data_file);
2893 receive_linecount++; /* The terminating "." line */
2894 }
2895 else message_ended = read_message_data(data_file);
2896
2897 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 2898 message_linecount += body_linecount;
059ec3d9
PH
2899
2900 /* Handle premature termination of SMTP */
2901
2902 if (smtp_input && message_ended == END_EOF)
2903 {
2904 Uunlink(spool_name); /* Lose data file when closed */
2e5b33cd 2905 cancel_cutthrough_connection("sender closed connection");
059ec3d9
PH
2906 message_id[0] = 0; /* Indicate no message accepted */
2907 smtp_reply = handle_lost_connection(US"");
2908 smtp_yield = FALSE;
2909 goto TIDYUP; /* Skip to end of function */
2910 }
2911
2912 /* Handle message that is too big. Don't use host_or_ident() in the log
2913 message; we want to see the ident value even for non-remote messages. */
2914
2915 if (message_ended == END_SIZE)
2916 {
2917 Uunlink(spool_name); /* Lose the data file when closed */
2e5b33cd 2918 cancel_cutthrough_connection("mail too big");
059ec3d9
PH
2919 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2920
2921 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2922 "message too big: read=%d max=%d",
2923 sender_address,
2924 (sender_fullhost == NULL)? "" : " H=",
2925 (sender_fullhost == NULL)? US"" : sender_fullhost,
2926 (sender_ident == NULL)? "" : " U=",
2927 (sender_ident == NULL)? US"" : sender_ident,
2928 message_size,
2929 thismessage_size_limit);
2930
2931 if (smtp_input)
2932 {
2933 smtp_reply = US"552 Message size exceeds maximum permitted";
2934 message_id[0] = 0; /* Indicate no message accepted */
2935 goto TIDYUP; /* Skip to end of function */
2936 }
2937 else
2938 {
2939 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2940 give_local_error(ERRMESS_TOOBIG,
2941 string_sprintf("message too big (max=%d)", thismessage_size_limit),
2942 US"message rejected: ", error_rc, data_file, header_list);
2943 /* Does not return */
2944 }
2945 }
2946 }
2947
2948/* Restore the standard SIGALRM handler for any subsequent processing. (For
2949example, there may be some expansion in an ACL that uses a timer.) */
2950
2951os_non_restarting_signal(SIGALRM, sigalrm_handler);
2952
2953/* The message body has now been read into the data file. Call fflush() to
2954empty the buffers in C, and then call fsync() to get the data written out onto
2955the disk, as fflush() doesn't do this (or at least, it isn't documented as
2956having to do this). If there was an I/O error on either input or output,
2957attempt to send an error message, and unlink the spool file. For non-SMTP input
2958we can then give up. Note that for SMTP input we must swallow the remainder of
2959the input in cases of output errors, since the far end doesn't expect to see
2960anything until the terminating dot line is sent. */
2961
2962if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 2963 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
2964 {
2965 uschar *msg_errno = US strerror(errno);
2966 BOOL input_error = (receive_ferror)() != 0;
2967 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
2968 input_error? "Input read" : "Spool write",
2969 msg_errno,
2970 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
2971
2972 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
2973 Uunlink(spool_name); /* Lose the data file */
2e5b33cd 2974 cancel_cutthrough_connection("error writing spoolfile");
059ec3d9
PH
2975
2976 if (smtp_input)
2977 {
2978 if (input_error)
2979 smtp_reply = US"451 Error while reading input data";
2980 else
2981 {
2982 smtp_reply = US"451 Error while writing spool file";
2983 receive_swallow_smtp();
2984 }
2985 message_id[0] = 0; /* Indicate no message accepted */
2986 goto TIDYUP; /* Skip to end of function */
2987 }
2988
2989 else
2990 {
2991 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2992 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
2993 header_list);
2994 /* Does not return */
2995 }
2996 }
2997
2998
2999/* No I/O errors were encountered while writing the data file. */
3000
3001DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
3002
3003
3004/* If there were any bad addresses extracted by -t, or there were no recipients
3005left after -t, send a message to the sender of this message, or write it to
3006stderr if the error handling option is set that way. Note that there may
3007legitimately be no recipients for an SMTP message if they have all been removed
3008by "discard".
3009
3010We need to rewind the data file in order to read it. In the case of no
3011recipients or stderr error writing, throw the data file away afterwards, and
3012exit. (This can't be SMTP, which always ensures there's at least one
3013syntactically good recipient address.) */
3014
3015if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
3016 {
3017 DEBUG(D_receive)
3018 {
3019 if (recipients_count == 0) debug_printf("*** No recipients\n");
3020 if (bad_addresses != NULL)
3021 {
3022 error_block *eblock = bad_addresses;
3023 debug_printf("*** Bad address(es)\n");
3024 while (eblock != NULL)
3025 {
3026 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
3027 eblock = eblock->next;
3028 }
3029 }
3030 }
3031
3032 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3033
3034 /* If configured to send errors to the sender, but this fails, force
3035 a failure error code. We use a special one for no recipients so that it
3036 can be detected by the autoreply transport. Otherwise error_rc is set to
3037 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
3038 it is EXIT_SUCCESS. */
3039
3040 if (error_handling == ERRORS_SENDER)
3041 {
3042 if (!moan_to_sender(
3043 (bad_addresses == NULL)?
3044 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
3045 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
3046 bad_addresses, header_list, data_file, FALSE))
3047 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
3048 }
3049 else
3050 {
3051 if (bad_addresses == NULL)
3052 {
3053 if (extracted_ignored)
3054 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3055 else
3056 fprintf(stderr, "exim: no recipients in message\n");
3057 }
3058 else
3059 {
3060 fprintf(stderr, "exim: invalid address%s",
3061 (bad_addresses->next == NULL)? ":" : "es:\n");
3062 while (bad_addresses != NULL)
3063 {
3064 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3065 bad_addresses->text2);
3066 bad_addresses = bad_addresses->next;
3067 }
3068 }
3069 }
3070
3071 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3072 {
3073 Uunlink(spool_name);
f1e894f3 3074 (void)fclose(data_file);
059ec3d9
PH
3075 exim_exit(error_rc);
3076 }
3077 }
3078
3079/* Data file successfully written. Generate text for the Received: header by
3080expanding the configured string, and adding a timestamp. By leaving this
3081operation till now, we ensure that the timestamp is the time that message
3082reception was completed. However, this is deliberately done before calling the
3083data ACL and local_scan().
3084
3085This Received: header may therefore be inspected by the data ACL and by code in
3086the local_scan() function. When they have run, we update the timestamp to be
3087the final time of reception.
3088
3089If there is just one recipient, set up its value in the $received_for variable
3090for use when we generate the Received: header.
3091
3092Note: the checking for too many Received: headers is handled by the delivery
3093code. */
e4bdf652 3094/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3095
e4bdf652 3096if (received_header->text == NULL) /* Non-cutthrough case */
059ec3d9 3097 {
e4bdf652 3098 received_header_gen();
059ec3d9 3099
e4bdf652 3100 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3101
e4bdf652
JH
3102 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3103 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3104
e4bdf652
JH
3105 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3106 now, before running the DATA ACL. */
059ec3d9 3107
e4bdf652
JH
3108 add_acl_headers(US"MAIL or RCPT");
3109 }
817d9f57 3110else
e4bdf652
JH
3111 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3112 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3113
3114/* If an ACL is specified for checking things at this stage of reception of a
3115message, run it, unless all the recipients were removed by "discard" in earlier
3116ACLs. That is the only case in which recipients_count can be zero at this
3117stage. Set deliver_datafile to point to the data file so that $message_body and
3118$message_body_end can be extracted if needed. Allow $recipients in expansions.
3119*/
3120
3121deliver_datafile = data_fd;
4e88a19f 3122user_msg = NULL;
059ec3d9 3123
0e20aff9
MH
3124enable_dollar_recipients = TRUE;
3125
059ec3d9
PH
3126if (recipients_count == 0)
3127 {
3128 blackholed_by = recipients_discarded? US"MAIL ACL" : US"RCPT ACL";
3129 }
3130else
3131 {
059ec3d9
PH
3132 /* Handle interactive SMTP messages */
3133
3134 if (smtp_input && !smtp_batched_input)
3135 {
8523533c 3136
80a47a2c
TK
3137#ifndef DISABLE_DKIM
3138 if (!dkim_disable_verify)
3139 {
3140 /* Finish verification, this will log individual signature results to
3141 the mainlog */
3142 dkim_exim_verify_finish();
3143
3144 /* Check if we must run the DKIM ACL */
3145 if ((acl_smtp_dkim != NULL) &&
3146 (dkim_verify_signers != NULL) &&
3147 (dkim_verify_signers[0] != '\0'))
3148 {
3149 uschar *dkim_verify_signers_expanded =
3150 expand_string(dkim_verify_signers);
3151 if (dkim_verify_signers_expanded == NULL)
3152 {
3153 log_write(0, LOG_MAIN|LOG_PANIC,
3154 "expansion of dkim_verify_signers option failed: %s",
3155 expand_string_message);
3156 }
3157 else
3158 {
3159 int sep = 0;
3160 uschar *ptr = dkim_verify_signers_expanded;
3161 uschar *item = NULL;
9e5d6b55
TK
3162 uschar *seen_items = NULL;
3163 int seen_items_size = 0;
3164 int seen_items_offset = 0;
80a47a2c 3165 uschar itembuf[256];
9122af94
TK
3166 /* Default to OK when no items are present */
3167 rc = OK;
80a47a2c
TK
3168 while ((item = string_nextinlist(&ptr, &sep,
3169 itembuf,
3170 sizeof(itembuf))) != NULL)
3171 {
6119d1ea
TK
3172 /* Prevent running ACL for an empty item */
3173 if (!item || (item[0] == '\0')) continue;
9e5d6b55
TK
3174 /* Only run ACL once for each domain or identity, no matter how often it
3175 appears in the expanded list. */
6119d1ea
TK
3176 if (seen_items != NULL)
3177 {
ae9094bf
TK
3178 uschar *seen_item = NULL;
3179 uschar seen_item_buf[256];
4a73449b 3180 uschar *seen_items_list = seen_items;
ae9094bf 3181 int seen_this_item = 0;
61147df4 3182
ae9094bf
TK
3183 while ((seen_item = string_nextinlist(&seen_items_list, &sep,
3184 seen_item_buf,
3185 sizeof(seen_item_buf))) != NULL)
3186 {
3187 if (Ustrcmp(seen_item,item) == 0)
3188 {
3189 seen_this_item = 1;
3190 break;
61147df4 3191 }
ae9094bf
TK
3192 }
3193
3194 if (seen_this_item > 0)
6119d1ea
TK
3195 {
3196 DEBUG(D_receive)
3197 debug_printf("acl_smtp_dkim: skipping signer %s, already seen\n", item);
3198 continue;
3199 }
61147df4 3200
4a73449b 3201 seen_items = string_append(seen_items,&seen_items_size,&seen_items_offset,1,":");
6119d1ea
TK
3202 }
3203
4a73449b
TK
3204 seen_items = string_append(seen_items,&seen_items_size,&seen_items_offset,1,item);
3205 seen_items[seen_items_offset] = '\0';
6119d1ea
TK
3206
3207 DEBUG(D_receive)
3208 debug_printf("calling acl_smtp_dkim for dkim_cur_signer=%s\n", item);
3209
80a47a2c
TK
3210 dkim_exim_acl_setup(item);
3211 rc = acl_check(ACL_WHERE_DKIM, NULL, acl_smtp_dkim, &user_msg, &log_msg);
6119d1ea
TK
3212
3213 if (rc != OK)
3214 {
3215 DEBUG(D_receive)
3216 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, skipping remaining items\n", rc, item);
2e5b33cd 3217 cancel_cutthrough_connection("dkim acl not ok");
6119d1ea
TK
3218 break;
3219 }
80a47a2c
TK
3220 }
3221 add_acl_headers(US"DKIM");
3222 if (rc == DISCARD)
3223 {
3224 recipients_count = 0;
3225 blackholed_by = US"DKIM ACL";
3226 if (log_msg != NULL)
3227 blackhole_log_msg = string_sprintf(": %s", log_msg);
3228 }
3229 else if (rc != OK)
3230 {
3231 Uunlink(spool_name);
3232 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
3233 smtp_yield = FALSE; /* No more messsages after dropped connection */
3234 smtp_reply = US""; /* Indicate reply already sent */
3235 message_id[0] = 0; /* Indicate no message accepted */
3236 goto TIDYUP; /* Skip to end of function */
3237 }
3238 }
3239 }
3240 }
3241#endif /* DISABLE_DKIM */
fb2274d4 3242
8523533c 3243#ifdef WITH_CONTENT_SCAN
80a47a2c
TK
3244 if (recipients_count > 0 &&
3245 acl_smtp_mime != NULL &&
54cdb463
PH
3246 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
3247 goto TIDYUP;
8523533c
TK
3248#endif /* WITH_CONTENT_SCAN */
3249
fd98a5c6
JH
3250#ifdef EXPERIMENTAL_PRDR
3251 if (prdr_requested && recipients_count > 1 && acl_smtp_data_prdr != NULL )
3252 {
3253 unsigned int c;
3254 int all_pass = OK;
3255 int all_fail = FAIL;
3256
3257 smtp_printf("353 PRDR content analysis beginning\r\n");
3258 /* Loop through recipients, responses must be in same order received */
3259 for (c = 0; recipients_count > c; c++)
3260 {
3261 uschar * addr= recipients_list[c].address;
3262 uschar * msg= US"PRDR R=<%s> %s";
3263 uschar * code;
3264 DEBUG(D_receive)
3265 debug_printf("PRDR processing recipient %s (%d of %d)\n",
3266 addr, c+1, recipients_count);
3267 rc = acl_check(ACL_WHERE_PRDR, addr,
3268 acl_smtp_data_prdr, &user_msg, &log_msg);
3269
3270 /* If any recipient rejected content, indicate it in final message */
3271 all_pass |= rc;
3272 /* If all recipients rejected, indicate in final message */
3273 all_fail &= rc;
3274
3275 switch (rc)
3276 {
3277 case OK: case DISCARD: code = US"250"; break;
3278 case DEFER: code = US"450"; break;
3279 default: code = US"550"; break;
3280 }
3281 if (user_msg != NULL)
3282 smtp_user_msg(code, user_msg);
3283 else
3284 {
3285 switch (rc)
3286 {
3287 case OK: case DISCARD:
3288 msg = string_sprintf(CS msg, addr, "acceptance"); break;
3289 case DEFER:
3290 msg = string_sprintf(CS msg, addr, "temporary refusal"); break;
3291 default:
3292 msg = string_sprintf(CS msg, addr, "refusal"); break;
3293 }
3294 smtp_user_msg(code, msg);
3295 }
3296 if (log_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, log_msg);
3297 else if (user_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, user_msg);
3298 else log_write(0, LOG_MAIN, CS msg);
3299
3300 if (rc != OK) { receive_remove_recipient(addr); c--; }
3301 }
3302 /* Set up final message, used if data acl gives OK */
3303 smtp_reply = string_sprintf("%s id=%s message %s",
3304 all_fail == FAIL ? US"550" : US"250",
3305 message_id,
3306 all_fail == FAIL
3307 ? US"rejected for all recipients"
3308 : all_pass == OK
3309 ? US"accepted"
3310 : US"accepted for some recipients");
3311 if (recipients_count == 0)
3312 {
3313 message_id[0] = 0; /* Indicate no message accepted */
3314 goto TIDYUP;
3315 }
3316 }
3317 else
3318 prdr_requested = FALSE;
3319#endif /* EXPERIMENTAL_PRDR */
3320
54cdb463
PH
3321 /* Check the recipients count again, as the MIME ACL might have changed
3322 them. */
8523533c 3323
059ec3d9
PH
3324 if (acl_smtp_data != NULL && recipients_count > 0)
3325 {
059ec3d9
PH
3326 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
3327 add_acl_headers(US"DATA");
3328 if (rc == DISCARD)
3329 {
3330 recipients_count = 0;
3331 blackholed_by = US"DATA ACL";
8e669ac1
PH
3332 if (log_msg != NULL)
3333 blackhole_log_msg = string_sprintf(": %s", log_msg);
2e5b33cd 3334 cancel_cutthrough_connection("data acl discard");
059ec3d9
PH
3335 }
3336 else if (rc != OK)
3337 {
3338 Uunlink(spool_name);
2e5b33cd 3339 cancel_cutthrough_connection("data acl not ok");
8523533c
TK
3340#ifdef WITH_CONTENT_SCAN
3341 unspool_mbox();
3342#endif
6f0c431a
PP
3343#ifdef EXPERIMENTAL_DCC
3344 dcc_ok = 0;
3345#endif
059ec3d9
PH
3346 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
3347 smtp_yield = FALSE; /* No more messsages after dropped connection */
3348 smtp_reply = US""; /* Indicate reply already sent */
3349 message_id[0] = 0; /* Indicate no message accepted */
3350 goto TIDYUP; /* Skip to end of function */
3351 }
3352 }
3353 }
3354
3355 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
3356 we cannot take different actions for permanent and temporary rejections. */
3357
54cdb463 3358 else
059ec3d9 3359 {
54cdb463
PH
3360
3361#ifdef WITH_CONTENT_SCAN
3362 if (acl_not_smtp_mime != NULL &&
3363 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
3364 &blackholed_by))
3365 goto TIDYUP;
3366#endif /* WITH_CONTENT_SCAN */
3367
3368 if (acl_not_smtp != NULL)
059ec3d9 3369 {
54cdb463
PH
3370 uschar *user_msg, *log_msg;
3371 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
3372 if (rc == DISCARD)
059ec3d9 3373 {
54cdb463
PH
3374 recipients_count = 0;
3375 blackholed_by = US"non-SMTP ACL";
3376 if (log_msg != NULL)
3377 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 3378 }
54cdb463 3379 else if (rc != OK)
059ec3d9 3380 {
54cdb463
PH
3381 Uunlink(spool_name);
3382#ifdef WITH_CONTENT_SCAN
3383 unspool_mbox();
3384#endif
6f0c431a
PP
3385#ifdef EXPERIMENTAL_DCC
3386 dcc_ok = 0;
3387#endif
6ea85e9a
PH
3388 /* The ACL can specify where rejections are to be logged, possibly
3389 nowhere. The default is main and reject logs. */
3390
3391 if (log_reject_target != 0)
3392 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3393 sender_address, log_msg);
3394
54cdb463
PH
3395 if (user_msg == NULL) user_msg = US"local configuration problem";
3396 if (smtp_batched_input)
3397 {
3398 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3399 /* Does not return */
3400 }
3401 else
3402 {
3403 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3404 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3405 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3406 header_list);
3407 /* Does not return */
3408 }
059ec3d9 3409 }
54cdb463 3410 add_acl_headers(US"non-SMTP");
059ec3d9 3411 }
059ec3d9
PH
3412 }
3413
54cdb463
PH
3414 /* The applicable ACLs have been run */
3415
059ec3d9
PH
3416 if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */
3417 if (queue_only_policy) queued_by = US"ACL";
059ec3d9
PH
3418 }
3419
8523533c
TK
3420#ifdef WITH_CONTENT_SCAN
3421unspool_mbox();
3422#endif
3423
6a8f9482
TK
3424#ifdef EXPERIMENTAL_DCC
3425dcc_ok = 0;
3426#endif
3427
3428
059ec3d9
PH
3429/* The final check on the message is to run the scan_local() function. The
3430version supplied with Exim always accepts, but this is a hook for sysadmins to
3431supply their own checking code. The local_scan() function is run even when all
3432the recipients have been discarded. */
3433
3434lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3435
3436/* Arrange to catch crashes in local_scan(), so that the -D file gets
3437deleted, and the incident gets logged. */
3438
3439os_non_restarting_signal(SIGSEGV, local_scan_crash_handler);
3440os_non_restarting_signal(SIGFPE, local_scan_crash_handler);
3441os_non_restarting_signal(SIGILL, local_scan_crash_handler);
3442os_non_restarting_signal(SIGBUS, local_scan_crash_handler);
3443
3444DEBUG(D_receive) debug_printf("calling local_scan(); timeout=%d\n",
3445 local_scan_timeout);
3446local_scan_data = NULL;
3447
3448os_non_restarting_signal(SIGALRM, local_scan_timeout_handler);
3449if (local_scan_timeout > 0) alarm(local_scan_timeout);
3450rc = local_scan(data_fd, &local_scan_data);
3451alarm(0);
3452os_non_restarting_signal(SIGALRM, sigalrm_handler);
3453
0e20aff9
MH
3454enable_dollar_recipients = FALSE;
3455
059ec3d9
PH
3456store_pool = POOL_MAIN; /* In case changed */
3457DEBUG(D_receive) debug_printf("local_scan() returned %d %s\n", rc,
3458 local_scan_data);
3459
3460os_non_restarting_signal(SIGSEGV, SIG_DFL);
3461os_non_restarting_signal(SIGFPE, SIG_DFL);
3462os_non_restarting_signal(SIGILL, SIG_DFL);
3463os_non_restarting_signal(SIGBUS, SIG_DFL);
3464
3465/* The length check is paranoia against some runaway code, and also because
3466(for a success return) lines in the spool file are read into big_buffer. */
3467
3468if (local_scan_data != NULL)
3469 {
3470 int len = Ustrlen(local_scan_data);
3471 if (len > LOCAL_SCAN_MAX_RETURN) len = LOCAL_SCAN_MAX_RETURN;
3472 local_scan_data = string_copyn(local_scan_data, len);
3473 }
3474
3475if (rc == LOCAL_SCAN_ACCEPT_FREEZE)
3476 {
58eb016e 3477 if (!deliver_freeze) /* ACL might have already frozen */
059ec3d9
PH
3478 {
3479 deliver_freeze = TRUE;
3480 deliver_frozen_at = time(NULL);
3481 frozen_by = US"local_scan()";
3482 }
3483 rc = LOCAL_SCAN_ACCEPT;
3484 }
3485else if (rc == LOCAL_SCAN_ACCEPT_QUEUE)
3486 {
3487 if (!queue_only_policy) /* ACL might have already queued */
3488 {
3489 queue_only_policy = TRUE;
3490 queued_by = US"local_scan()";
3491 }
3492 rc = LOCAL_SCAN_ACCEPT;
3493 }
3494
3495/* Message accepted: remove newlines in local_scan_data because otherwise
3496the spool file gets corrupted. Ensure that all recipients are qualified. */
3497
3498if (rc == LOCAL_SCAN_ACCEPT)
3499 {
3500 if (local_scan_data != NULL)
3501 {
3502 uschar *s;
3503 for (s = local_scan_data; *s != 0; s++) if (*s == '\n') *s = ' ';
3504 }
3505 for (i = 0; i < recipients_count; i++)
3506 {
3507 recipient_item *r = recipients_list + i;
3508 r->address = rewrite_address_qualify(r->address, TRUE);
3509 if (r->errors_to != NULL)
3510 r->errors_to = rewrite_address_qualify(r->errors_to, TRUE);
3511 }
3512 if (recipients_count == 0 && blackholed_by == NULL)
3513 blackholed_by = US"local_scan";
3514 }
3515
3516/* Message rejected: newlines permitted in local_scan_data to generate
3517multiline SMTP responses. */
3518
3519else
3520 {
3521 uschar *istemp = US"";
3522 uschar *s = NULL;
a5bd321b 3523 uschar *smtp_code;
059ec3d9
PH
3524 int size = 0;
3525 int sptr = 0;
059ec3d9
PH
3526
3527 errmsg = local_scan_data;
3528
3529 Uunlink(spool_name); /* Cancel this message */
3530 switch(rc)
3531 {
3532 default:
3533 log_write(0, LOG_MAIN, "invalid return %d from local_scan(). Temporary "
3534 "rejection given", rc);
3535 goto TEMPREJECT;
3536
3537 case LOCAL_SCAN_REJECT_NOLOGHDR:
3538 log_extra_selector &= ~LX_rejected_header;
3539 /* Fall through */
3540
3541 case LOCAL_SCAN_REJECT:
a5bd321b 3542 smtp_code = US"550";
059ec3d9
PH
3543 if (errmsg == NULL) errmsg = US"Administrative prohibition";
3544 break;
3545
3546 case LOCAL_SCAN_TEMPREJECT_NOLOGHDR:
3547 log_extra_selector &= ~LX_rejected_header;
3548 /* Fall through */
3549
3550 case LOCAL_SCAN_TEMPREJECT:
3551 TEMPREJECT:
a5bd321b 3552 smtp_code = US"451";
059ec3d9
PH
3553 if (errmsg == NULL) errmsg = US"Temporary local problem";
3554 istemp = US"temporarily ";
3555 break;
3556 }
3557
3558 s = string_append(s, &size, &sptr, 2, US"F=",
3559 (sender_address[0] == 0)? US"<>" : sender_address);
3560 s = add_host_info_for_log(s, &size, &sptr);
3561 s[sptr] = 0;
3562
3563 log_write(0, LOG_MAIN|LOG_REJECT, "%s %srejected by local_scan(): %.256s",
3564 s, istemp, string_printing(errmsg));
3565
3566 if (smtp_input)
3567 {
3568 if (!smtp_batched_input)
3569 {
a5bd321b 3570 smtp_respond(smtp_code, 3, TRUE, errmsg);
059ec3d9
PH
3571 message_id[0] = 0; /* Indicate no message accepted */
3572 smtp_reply = US""; /* Indicate reply already sent */
3573 goto TIDYUP; /* Skip to end of function */
3574 }
3575 else
3576 {
a5bd321b 3577 moan_smtp_batch(NULL, "%s %s", smtp_code, errmsg);
059ec3d9
PH
3578 /* Does not return */
3579 }
3580 }
3581 else
3582 {
3583 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3584 give_local_error(ERRMESS_LOCAL_SCAN, errmsg,
3585 US"message rejected by local scan code: ", error_rc, data_file,
3586 header_list);
3587 /* Does not return */
3588 }
3589 }
3590
3591/* Reset signal handlers to ignore signals that previously would have caused
3592the message to be abandoned. */
3593
3594signal(SIGTERM, SIG_IGN);
3595signal(SIGINT, SIG_IGN);
3596
e4bdf652 3597
059ec3d9
PH
3598/* Ensure the first time flag is set in the newly-received message. */
3599
3600deliver_firsttime = TRUE;
3601
8523533c
TK
3602#ifdef EXPERIMENTAL_BRIGHTMAIL
3603if (bmi_run == 1) {
3604 /* rewind data file */
3605 lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3606 bmi_verdicts = bmi_process_message(header_list, data_fd);
3607};
3608#endif
3609
059ec3d9
PH
3610/* Update the timstamp in our Received: header to account for any time taken by
3611an ACL or by local_scan(). The new time is the time that all reception
3612processing is complete. */
3613
3614timestamp = expand_string(US"${tod_full}");
3615tslen = Ustrlen(timestamp);
3616
3617memcpy(received_header->text + received_header->slen - tslen - 1,
3618 timestamp, tslen);
3619
3620/* In MUA wrapper mode, ignore queueing actions set by ACL or local_scan() */
3621
3622if (mua_wrapper)
3623 {
3624 deliver_freeze = FALSE;
3625 queue_only_policy = FALSE;
3626 }
3627
3628/* Keep the data file open until we have written the header file, in order to
3629hold onto the lock. In a -bh run, or if the message is to be blackholed, we
3630don't write the header file, and we unlink the data file. If writing the header
3631file fails, we have failed to accept this message. */
3632
3633if (host_checking || blackholed_by != NULL)
3634 {
3635 header_line *h;
3636 Uunlink(spool_name);
3637 msg_size = 0; /* Compute size for log line */
3638 for (h = header_list; h != NULL; h = h->next)
3639 if (h->type != '*') msg_size += h->slen;
3640 }
3641
3642/* Write the -H file */
3643
3644else
3645 {
3646 if ((msg_size = spool_write_header(message_id, SW_RECEIVING, &errmsg)) < 0)
3647 {
3648 log_write(0, LOG_MAIN, "Message abandoned: %s", errmsg);
3649 Uunlink(spool_name); /* Lose the data file */
3650
3651 if (smtp_input)
3652 {
3653 smtp_reply = US"451 Error in writing spool file";
3654 message_id[0] = 0; /* Indicate no message accepted */
3655 goto TIDYUP;
3656 }
3657 else
3658 {
3659 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3660 give_local_error(ERRMESS_IOERR, errmsg, US"", error_rc, data_file,
3661 header_list);
3662 /* Does not return */
3663 }
3664 }
3665 }
3666
3667
3668/* The message has now been successfully received. */
3669
3670receive_messagecount++;
3671
3672/* In SMTP sessions we may receive several in one connection. After each one,
3673we wait for the clock to tick at the level of message-id granularity. This is
3674so that the combination of time+pid is unique, even on systems where the pid
3675can be re-used within our time interval. We can't shorten the interval without
3676re-designing the message-id. See comments above where the message id is
3677created. This is Something For The Future. */
3678
3679message_id_tv.tv_usec = (message_id_tv.tv_usec/id_resolution) * id_resolution;
3680exim_wait_tick(&message_id_tv, id_resolution);
3681
3682/* Add data size to written header size. We do not count the initial file name
3683that is in the file, but we do add one extra for the notional blank line that
3684precedes the data. This total differs from message_size in that it include the
3685added Received: header and any other headers that got created locally. */
3686
3687fflush(data_file);
3688fstat(data_fd, &statbuf);
3689
3690msg_size += statbuf.st_size - SPOOL_DATA_START_OFFSET + 1;
3691
3692/* Generate a "message received" log entry. We do this by building up a dynamic
3693string as required. Since we commonly want to add two items at a time, use a
3694macro to simplify the coding. We log the arrival of a new message while the
3695file is still locked, just in case the machine is *really* fast, and delivers
3696it first! Include any message id that is in the message - since the syntax of a
3697message id is actually an addr-spec, we can use the parse routine to canonicize
3698it. */
3699
3700size = 256;
3701sptr = 0;
3702s = store_get(size);
3703
3704s = string_append(s, &size, &sptr, 2, US"<= ",
3705 (sender_address[0] == 0)? US"<>" : sender_address);
3706if (message_reference != NULL)
3707 s = string_append(s, &size, &sptr, 2, US" R=", message_reference);
3708
3709s = add_host_info_for_log(s, &size, &sptr);
3710
3711#ifdef SUPPORT_TLS
817d9f57
JH
3712if ((log_extra_selector & LX_tls_cipher) != 0 && tls_in.cipher != NULL)
3713 s = string_append(s, &size, &sptr, 2, US" X=", tls_in.cipher);
059ec3d9 3714if ((log_extra_selector & LX_tls_certificate_verified) != 0 &&
817d9f57 3715 tls_in.cipher != NULL)
059ec3d9 3716 s = string_append(s, &size, &sptr, 2, US" CV=",
817d9f57
JH
3717 tls_in.certificate_verified? "yes":"no");
3718if ((log_extra_selector & LX_tls_peerdn) != 0 && tls_in.peerdn != NULL)
48ed62d9 3719 s = string_append(s, &size, &sptr, 3, US" DN=\"",
817d9f57
JH
3720 string_printing(tls_in.peerdn), US"\"");
3721if ((log_extra_selector & LX_tls_sni) != 0 && tls_in.sni != NULL)
3f0945ff 3722 s = string_append(s, &size, &sptr, 3, US" SNI=\"",
817d9f57 3723 string_printing(tls_in.sni), US"\"");
3f0945ff 3724#endif
059ec3d9
PH
3725
3726if (sender_host_authenticated != NULL)
3727 {
3728 s = string_append(s, &size, &sptr, 2, US" A=", sender_host_authenticated);
3729 if (authenticated_id != NULL)
c8e2fc1e 3730 {
059ec3d9 3731 s = string_append(s, &size, &sptr, 2, US":", authenticated_id);
c8e2fc1e
JH
3732 if (log_extra_selector & LX_smtp_mailauth && authenticated_sender != NULL)
3733 s = string_append(s, &size, &sptr, 2, US":", authenticated_sender);
3734 }
059ec3d9
PH
3735 }
3736
fd98a5c6
JH
3737#ifdef EXPERIMENTAL_PRDR
3738if (prdr_requested)
3739 s = string_append(s, &size, &sptr, 1, US" PRDR");
3740#endif
3741
059ec3d9
PH
3742sprintf(CS big_buffer, "%d", msg_size);
3743s = string_append(s, &size, &sptr, 2, US" S=", big_buffer);
3744
3c0a92dc
JH
3745/* log 8BITMIME mode announced in MAIL_FROM
3746 0 ... no BODY= used
3747 7 ... 7BIT
3748 8 ... 8BITMIME */
c8e2fc1e
JH
3749if (log_extra_selector & LX_8bitmime)
3750 {
3c0a92dc
JH
3751 sprintf(CS big_buffer, "%d", body_8bitmime);
3752 s = string_append(s, &size, &sptr, 2, US" M8S=", big_buffer);
c8e2fc1e 3753 }
3c0a92dc 3754
059ec3d9
PH
3755/* If an addr-spec in a message-id contains a quoted string, it can contain
3756any characters except " \ and CR and so in particular it can contain NL!
3757Therefore, make sure we use a printing-characters only version for the log.
3758Also, allow for domain literals in the message id. */
3759
3760if (msgid_header != NULL)
3761 {
3762 uschar *old_id;
3763 BOOL save_allow_domain_literals = allow_domain_literals;
3764 allow_domain_literals = TRUE;
3765 old_id = parse_extract_address(Ustrchr(msgid_header->text, ':') + 1,
3766 &errmsg, &start, &end, &domain, FALSE);
3767 allow_domain_literals = save_allow_domain_literals;
3768 if (old_id != NULL)
3769 s = string_append(s, &size, &sptr, 2, US" id=", string_printing(old_id));
3770 }
3771
3772/* If subject logging is turned on, create suitable printing-character
3773text. By expanding $h_subject: we make use of the MIME decoding. */
3774
3775if ((log_extra_selector & LX_subject) != 0 && subject_header != NULL)
3776 {
3777 int i;
3778 uschar *p = big_buffer;
3779 uschar *ss = expand_string(US"$h_subject:");
3780
3781 /* Backslash-quote any double quotes or backslashes so as to make a
3782 a C-like string, and turn any non-printers into escape sequences. */
3783
3784 *p++ = '\"';
3785 if (*ss != 0) for (i = 0; i < 100 && ss[i] != 0; i++)
3786 {
3787 if (ss[i] == '\"' || ss[i] == '\\') *p++ = '\\';
3788 *p++ = ss[i];
3789 }
3790 *p++ = '\"';
3791 *p = 0;
3792 s = string_append(s, &size, &sptr, 2, US" T=", string_printing(big_buffer));
3793 }
3794
3795/* Terminate the string: string_cat() and string_append() leave room, but do
3796not put the zero in. */
3797
3798s[sptr] = 0;
3799
059ec3d9
PH
3800/* Create a message log file if message logs are being used and this message is
3801not blackholed. Write the reception stuff to it. We used to leave message log
e4bdf652 3802creation until the first delivery, but this has proved confusing for some
059ec3d9
PH
3803people. */
3804
3805if (message_logs && blackholed_by == NULL)
3806 {
3807 int fd;
3808