Note about ratelimit resets.
[exim.git] / src / src / receive.c
CommitLineData
563b63fa 1/* $Cambridge: exim/src/src/receive.c,v 1.37 2007/04/16 10:31:58 ph10 Exp $ */
059ec3d9
PH
2
3/*************************************************
4* Exim - an Internet mail transport agent *
5*************************************************/
6
184e8823 7/* Copyright (c) University of Cambridge 1995 - 2007 */
059ec3d9
PH
8/* See the file NOTICE for conditions of use and distribution. */
9
10/* Code for receiving a message and setting up spool files. */
11
059ec3d9
PH
12#include "exim.h"
13
fb2274d4
TK
14#ifdef EXPERIMENTAL_DOMAINKEYS
15#define RECEIVE_GETC dk_receive_getc
16#define RECEIVE_UNGETC dk_receive_ungetc
17#else
18#define RECEIVE_GETC receive_getc
19#define RECEIVE_UNGETC receive_ungetc
20#endif
059ec3d9
PH
21
22/*************************************************
23* Local static variables *
24*************************************************/
25
26static FILE *data_file = NULL;
27static int data_fd = -1;
28static uschar spool_name[256];
29
30
31
32/*************************************************
33* Non-SMTP character reading functions *
34*************************************************/
35
36/* These are the default functions that are set up in the variables such as
37receive_getc initially. They just call the standard functions, passing stdin as
38the file. (When SMTP input is occurring, different functions are used by
39changing the pointer variables.) */
40
41int
42stdin_getc(void)
43{
44return getc(stdin);
45}
46
47int
48stdin_ungetc(int c)
49{
50return ungetc(c, stdin);
51}
52
53int
54stdin_feof(void)
55{
56return feof(stdin);
57}
58
59int
60stdin_ferror(void)
61{
62return ferror(stdin);
63}
64
65
66
67
68/*************************************************
69* Check that a set sender is allowed *
70*************************************************/
71
72/* This function is called when a local caller sets an explicit sender address.
73It checks whether this is permitted, which it is for trusted callers.
74Otherwise, it must match the pattern(s) in untrusted_set_sender.
75
76Arguments: the proposed sender address
77Returns: TRUE for a trusted caller
78 TRUE if the address has been set, untrusted_set_sender has been
79 set, and the address matches something in the list
80 FALSE otherwise
81*/
82
83BOOL
84receive_check_set_sender(uschar *newsender)
85{
86uschar *qnewsender;
87if (trusted_caller) return TRUE;
88if (newsender == NULL || untrusted_set_sender == NULL) return FALSE;
89qnewsender = (Ustrchr(newsender, '@') != NULL)?
90 newsender : string_sprintf("%s@%s", newsender, qualify_domain_sender);
91return
92 match_address_list(qnewsender, TRUE, TRUE, &untrusted_set_sender, NULL, -1,
93 0, NULL) == OK;
94}
95
96
97
98
99/*************************************************
5cb8cbc6 100* Read space info for a partition *
059ec3d9
PH
101*************************************************/
102
8e669ac1
PH
103/* This function is called by receive_check_fs() below, and also by string
104expansion for variables such as $spool_space. The field names for the statvfs
5cb8cbc6
PH
105structure are macros, because not all OS have F_FAVAIL and it seems tidier to
106have macros for F_BAVAIL and F_FILES as well. Some kinds of file system do not
107have inodes, and they return -1 for the number available.
059ec3d9 108
5cb8cbc6
PH
109Later: It turns out that some file systems that do not have the concept of
110inodes return 0 rather than -1. Such systems should also return 0 for the total
8e669ac1 111number of inodes, so we require that to be greater than zero before returning
5cb8cbc6 112an inode count.
059ec3d9 113
5cb8cbc6
PH
114Arguments:
115 isspool TRUE for spool partition, FALSE for log partition
116 inodeptr address of int to receive inode count; -1 if there isn't one
8e669ac1 117
5cb8cbc6 118Returns: available on-root space, in kilobytes
8e669ac1
PH
119 -1 for log partition if there isn't one
120
121All values are -1 if the STATFS functions are not available.
059ec3d9
PH
122*/
123
8e669ac1 124int
5cb8cbc6 125receive_statvfs(BOOL isspool, int *inodeptr)
059ec3d9
PH
126{
127#ifdef HAVE_STATFS
059ec3d9 128struct STATVFS statbuf;
5cb8cbc6
PH
129uschar *path;
130uschar *name;
131uschar buffer[1024];
059ec3d9 132
5cb8cbc6 133/* The spool directory must always exist. */
059ec3d9 134
5cb8cbc6 135if (isspool)
059ec3d9 136 {
8e669ac1
PH
137 path = spool_directory;
138 name = US"spool";
139 }
140
059ec3d9
PH
141/* Need to cut down the log file path to the directory, and to ignore any
142appearance of "syslog" in it. */
143
5cb8cbc6 144else
059ec3d9 145 {
059ec3d9 146 int sep = ':'; /* Not variable - outside scripts use */
059ec3d9 147 uschar *p = log_file_path;
8e669ac1 148 name = US"log";
059ec3d9
PH
149
150 /* An empty log_file_path means "use the default". This is the same as an
151 empty item in a list. */
152
153 if (*p == 0) p = US":";
154 while ((path = string_nextinlist(&p, &sep, buffer, sizeof(buffer))) != NULL)
155 {
156 if (Ustrcmp(path, "syslog") != 0) break;
157 }
158
5cb8cbc6
PH
159 if (path == NULL) /* No log files */
160 {
8e669ac1
PH
161 *inodeptr = -1;
162 return -1;
163 }
059ec3d9 164
8e669ac1
PH
165 /* An empty string means use the default, which is in the spool directory.
166 But don't just use the spool directory, as it is possible that the log
5cb8cbc6 167 subdirectory has been symbolically linked elsewhere. */
059ec3d9 168
8e669ac1 169 if (path[0] == 0)
059ec3d9 170 {
5cb8cbc6
PH
171 sprintf(CS buffer, CS"%s/log", CS spool_directory);
172 path = buffer;
8e669ac1
PH
173 }
174 else
059ec3d9 175 {
8e669ac1 176 uschar *cp;
5cb8cbc6 177 if ((cp = Ustrrchr(path, '/')) != NULL) *cp = 0;
8e669ac1 178 }
5cb8cbc6 179 }
8e669ac1 180
5cb8cbc6
PH
181/* We now have the patch; do the business */
182
183memset(&statbuf, 0, sizeof(statbuf));
184
185if (STATVFS(CS path, &statbuf) != 0)
186 {
187 log_write(0, LOG_MAIN|LOG_PANIC, "cannot accept message: failed to stat "
188 "%s directory %s: %s", name, spool_directory, strerror(errno));
189 smtp_closedown(US"spool or log directory problem");
190 exim_exit(EXIT_FAILURE);
191 }
8e669ac1 192
5cb8cbc6
PH
193*inodeptr = (statbuf.F_FILES > 0)? statbuf.F_FAVAIL : -1;
194
195/* Disks are getting huge. Take care with computing the size in kilobytes. */
8e669ac1 196
5cb8cbc6
PH
197return (int)(((double)statbuf.F_BAVAIL * (double)statbuf.F_FRSIZE)/1024.0);
198
199/* Unable to find partition sizes in this environment. */
200
201#else
202*inodeptr = -1;
203return -1;
204#endif
205}
206
059ec3d9 207
059ec3d9 208
5cb8cbc6
PH
209
210/*************************************************
211* Check space on spool and log partitions *
212*************************************************/
213
214/* This function is called before accepting a message; if any thresholds are
215set, it checks them. If a message_size is supplied, it checks that there is
216enough space for that size plus the threshold - i.e. that the message won't
217reduce the space to the threshold. Not all OS have statvfs(); for those that
218don't, this function always returns TRUE. For some OS the old function and
219struct name statfs is used; that is handled by a macro, defined in exim.h.
220
221Arguments:
222 msg_size the (estimated) size of an incoming message
223
224Returns: FALSE if there isn't enough space, or if the information cannot
225 be obtained
226 TRUE if no check was done or there is enough space
227*/
228
229BOOL
230receive_check_fs(int msg_size)
231{
232int space, inodes;
233
234if (check_spool_space > 0 || msg_size > 0 || check_spool_inodes > 0)
235 {
8e669ac1
PH
236 space = receive_statvfs(TRUE, &inodes);
237
059ec3d9 238 DEBUG(D_receive)
5cb8cbc6
PH
239 debug_printf("spool directory space = %dK inodes = %d "
240 "check_space = %dK inodes = %d msg_size = %d\n",
241 space, inodes, check_spool_space, check_spool_inodes, msg_size);
8e669ac1
PH
242
243 if ((space >= 0 && space < check_spool_space) ||
5cb8cbc6 244 (inodes >= 0 && inodes < check_spool_inodes))
8e669ac1 245 {
5cb8cbc6
PH
246 log_write(0, LOG_MAIN, "spool directory space check failed: space=%d "
247 "inodes=%d", space, inodes);
059ec3d9
PH
248 return FALSE;
249 }
250 }
251
5cb8cbc6
PH
252if (check_log_space > 0 || check_log_inodes > 0)
253 {
8e669ac1
PH
254 space = receive_statvfs(FALSE, &inodes);
255
5cb8cbc6
PH
256 DEBUG(D_receive)
257 debug_printf("log directory space = %dK inodes = %d "
258 "check_space = %dK inodes = %d\n",
259 space, inodes, check_log_space, check_log_inodes);
8e669ac1
PH
260
261 if ((space >= 0 && space < check_log_space) ||
5cb8cbc6 262 (inodes >= 0 && inodes < check_log_inodes))
8e669ac1 263 {
5cb8cbc6
PH
264 log_write(0, LOG_MAIN, "log directory space check failed: space=%d "
265 "inodes=%d", space, inodes);
266 return FALSE;
267 }
8e669ac1
PH
268 }
269
059ec3d9
PH
270return TRUE;
271}
272
273
274
275/*************************************************
276* Bomb out while reading a message *
277*************************************************/
278
279/* The common case of wanting to bomb out is if a SIGTERM or SIGINT is
280received, or if there is a timeout. A rarer case might be if the log files are
281screwed up and Exim can't open them to record a message's arrival. Handling
282that case is done by setting a flag to cause the log functions to call this
283function if there is an ultimate disaster. That is why it is globally
284accessible.
285
286Arguments: SMTP response to give if in an SMTP session
287Returns: it doesn't
288*/
289
290void
291receive_bomb_out(uschar *msg)
292{
293/* If spool_name is set, it contains the name of the data file that is being
294written. Unlink it before closing so that it cannot be picked up by a delivery
295process. Ensure that any header file is also removed. */
296
297if (spool_name[0] != 0)
298 {
299 Uunlink(spool_name);
300 spool_name[Ustrlen(spool_name) - 1] = 'H';
301 Uunlink(spool_name);
302 }
303
304/* Now close the file if it is open, either as a fd or a stream. */
305
f1e894f3
PH
306if (data_file != NULL) (void)fclose(data_file);
307 else if (data_fd >= 0) (void)close(data_fd);
059ec3d9
PH
308
309/* Attempt to close down an SMTP connection tidily. */
310
311if (smtp_input)
312 {
313 if (!smtp_batched_input)
314 {
315 smtp_printf("421 %s %s - closing connection.\r\n", smtp_active_hostname,
316 msg);
317 mac_smtp_fflush();
318 }
319
320 /* Control does not return from moan_smtp_batch(). */
321
322 else moan_smtp_batch(NULL, "421 %s - message abandoned", msg);
323 }
324
325/* Exit from the program (non-BSMTP cases) */
326
327exim_exit(EXIT_FAILURE);
328}
329
330
331/*************************************************
332* Data read timeout *
333*************************************************/
334
335/* Handler function for timeouts that occur while reading the data that
336comprises a message.
337
338Argument: the signal number
339Returns: nothing
340*/
341
342static void
343data_timeout_handler(int sig)
344{
345uschar *msg = NULL;
346
347sig = sig; /* Keep picky compilers happy */
348
349if (smtp_input)
350 {
351 msg = US"SMTP incoming data timeout";
352 log_write(L_lost_incoming_connection,
353 LOG_MAIN, "SMTP data timeout (message abandoned) on connection "
fed77020
PH
354 "from %s F=<%s>",
355 (sender_fullhost != NULL)? sender_fullhost : US"local process",
356 sender_address);
059ec3d9
PH
357 }
358else
359 {
360 fprintf(stderr, "exim: timed out while reading - message abandoned\n");
361 log_write(L_lost_incoming_connection,
362 LOG_MAIN, "timed out while reading local message");
363 }
364
365receive_bomb_out(msg); /* Does not return */
366}
367
368
369
370/*************************************************
371* local_scan() timeout *
372*************************************************/
373
374/* Handler function for timeouts that occur while running a local_scan()
375function.
376
377Argument: the signal number
378Returns: nothing
379*/
380
381static void
382local_scan_timeout_handler(int sig)
383{
384sig = sig; /* Keep picky compilers happy */
385log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function timed out - "
386 "message temporarily rejected (size %d)", message_size);
387receive_bomb_out(US"local verification problem"); /* Does not return */
388}
389
390
391
392/*************************************************
393* local_scan() crashed *
394*************************************************/
395
396/* Handler function for signals that occur while running a local_scan()
397function.
398
399Argument: the signal number
400Returns: nothing
401*/
402
403static void
404local_scan_crash_handler(int sig)
405{
406log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function crashed with "
407 "signal %d - message temporarily rejected (size %d)", sig, message_size);
408receive_bomb_out(US"local verification problem"); /* Does not return */
409}
410
411
412/*************************************************
413* SIGTERM or SIGINT received *
414*************************************************/
415
416/* Handler for SIGTERM or SIGINT signals that occur while reading the
417data that comprises a message.
418
419Argument: the signal number
420Returns: nothing
421*/
422
423static void
424data_sigterm_sigint_handler(int sig)
425{
426uschar *msg = NULL;
427
428if (smtp_input)
429 {
430 msg = US"Service not available - SIGTERM or SIGINT received";
431 log_write(0, LOG_MAIN, "%s closed after %s", smtp_get_connection_info(),
432 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
433 }
434else
435 {
f05da2e8 436 if (filter_test == FTEST_NONE)
059ec3d9
PH
437 {
438 fprintf(stderr, "\nexim: %s received - message abandoned\n",
439 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
440 log_write(0, LOG_MAIN, "%s received while reading local message",
441 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
442 }
443 }
444
445receive_bomb_out(msg); /* Does not return */
446}
447
448
449
450/*************************************************
451* Add new recipient to list *
452*************************************************/
453
454/* This function builds a list of recipient addresses in argc/argv
455format.
456
457Arguments:
458 recipient the next address to add to recipients_list
459 pno parent number for fixed aliases; -1 otherwise
460
461Returns: nothing
462*/
463
464void
465receive_add_recipient(uschar *recipient, int pno)
466{
467if (recipients_count >= recipients_list_max)
468 {
469 recipient_item *oldlist = recipients_list;
470 int oldmax = recipients_list_max;
471 recipients_list_max = recipients_list_max? 2*recipients_list_max : 50;
472 recipients_list = store_get(recipients_list_max * sizeof(recipient_item));
473 if (oldlist != NULL)
474 memcpy(recipients_list, oldlist, oldmax * sizeof(recipient_item));
475 }
476
477recipients_list[recipients_count].address = recipient;
478recipients_list[recipients_count].pno = pno;
8523533c
TK
479#ifdef EXPERIMENTAL_BRIGHTMAIL
480recipients_list[recipients_count].bmi_optin = bmi_current_optin;
481/* reset optin string pointer for next recipient */
482bmi_current_optin = NULL;
483#endif
059ec3d9
PH
484recipients_list[recipients_count++].errors_to = NULL;
485}
486
487
488
489
490/*************************************************
491* Remove a recipient from the list *
492*************************************************/
493
494/* This function is provided for local_scan() to use.
495
496Argument:
497 recipient address to remove
498
499Returns: TRUE if it did remove something; FALSE otherwise
500*/
501
502BOOL
503receive_remove_recipient(uschar *recipient)
504{
505int count;
506DEBUG(D_receive) debug_printf("receive_remove_recipient(\"%s\") called\n",
507 recipient);
508for (count = 0; count < recipients_count; count++)
509 {
510 if (Ustrcmp(recipients_list[count].address, recipient) == 0)
511 {
512 if ((--recipients_count - count) > 0)
513 memmove(recipients_list + count, recipients_list + count + 1,
54cdb463 514 (recipients_count - count)*sizeof(recipient_item));
059ec3d9
PH
515 return TRUE;
516 }
517 }
518return FALSE;
519}
520
521
522
523
524
525/*************************************************
526* Read data portion of a non-SMTP message *
527*************************************************/
528
529/* This function is called to read the remainder of a message (following the
530header) when the input is not from SMTP - we are receiving a local message on
531a standard input stream. The message is always terminated by EOF, and is also
532terminated by a dot on a line by itself if the flag dot_ends is TRUE. Split the
533two cases for maximum efficiency.
534
535Ensure that the body ends with a newline. This will naturally be the case when
536the termination is "\n.\n" but may not be otherwise. The RFC defines messages
537as "sequences of lines" - this of course strictly applies only to SMTP, but
538deliveries into BSD-type mailbox files also require it. Exim used to have a
539flag for doing this at delivery time, but as it was always set for all
540transports, I decided to simplify things by putting the check here instead.
541
542There is at least one MUA (dtmail) that sends CRLF via this interface, and
543other programs are known to do this as well. Exim used to have a option for
544dealing with this: in July 2003, after much discussion, the code has been
545changed to default to treat any of LF, CRLF, and bare CR as line terminators.
546
547However, for the case when a dot on a line by itself terminates a message, the
548only recognized terminating sequences before and after the dot are LF and CRLF.
549Otherwise, having read EOL . CR, you don't know whether to read another
550character or not.
551
552Internally, in messages stored in Exim's spool files, LF is used as the line
553terminator. Under the new regime, bare CRs will no longer appear in these
554files.
555
556Arguments:
557 fout a FILE to which to write the message
558
559Returns: One of the END_xxx values indicating why it stopped reading
560*/
561
562static int
563read_message_data(FILE *fout)
564{
565int ch_state;
566register int ch;
567
568/* Handle the case when only EOF terminates the message */
569
570if (!dot_ends)
571 {
572 register int last_ch = '\n';
573
fb2274d4 574 for (; (ch = (RECEIVE_GETC)()) != EOF; last_ch = ch)
059ec3d9
PH
575 {
576 if (ch == 0) body_zerocount++;
577 if (last_ch == '\r' && ch != '\n')
578 {
579 if (fputc('\n', fout) == EOF) return END_WERROR;
580 message_size++;
581 body_linecount++;
582 }
583 if (ch == '\r') continue;
584
585 if (fputc(ch, fout) == EOF) return END_WERROR;
586 if (ch == '\n') body_linecount++;
587 if (++message_size > thismessage_size_limit) return END_SIZE;
588 }
589
590 if (last_ch != '\n')
591 {
592 if (fputc('\n', fout) == EOF) return END_WERROR;
593 message_size++;
594 body_linecount++;
595 }
596
597 return END_EOF;
598 }
599
600/* Handle the case when a dot on a line on its own, or EOF, terminates. */
601
602ch_state = 1;
603
fb2274d4 604while ((ch = (RECEIVE_GETC)()) != EOF)
059ec3d9
PH
605 {
606 if (ch == 0) body_zerocount++;
607 switch (ch_state)
608 {
609 case 0: /* Normal state (previous char written) */
610 if (ch == '\n')
611 { body_linecount++; ch_state = 1; }
612 else if (ch == '\r')
613 { ch_state = 2; continue; }
614 break;
615
616 case 1: /* After written "\n" */
617 if (ch == '.') { ch_state = 3; continue; }
618 if (ch != '\n') ch_state = 0;
619 break;
620
621 case 2:
622 body_linecount++; /* After unwritten "\r" */
623 if (ch == '\n')
624 { ch_state = 1; }
625 else
626 {
627 if (message_size++, fputc('\n', fout) == EOF) return END_WERROR;
628 if (ch == '\r') continue;
629 ch_state = 0;
630 }
631 break;
632
633 case 3: /* After "\n." (\n written, dot not) */
634 if (ch == '\n') return END_DOT;
635 if (ch == '\r') { ch_state = 4; continue; }
636 message_size++;
637 if (fputc('.', fout) == EOF) return END_WERROR;
638 ch_state = 0;
639 break;
640
641 case 4: /* After "\n.\r" (\n written, rest not) */
642 if (ch == '\n') return END_DOT;
643 message_size += 2;
644 body_linecount++;
645 if (fputs(".\n", fout) == EOF) return END_WERROR;
646 if (ch == '\r') { ch_state = 2; continue; }
647 ch_state = 0;
648 break;
649 }
650
651 if (fputc(ch, fout) == EOF) return END_WERROR;
652 if (++message_size > thismessage_size_limit) return END_SIZE;
653 }
654
655/* Get here if EOF read. Unless we have just written "\n", we need to ensure
656the message ends with a newline, and we must also write any characters that
657were saved up while testing for an ending dot. */
658
659if (ch_state != 1)
660 {
661 static uschar *ends[] = { US"\n", NULL, US"\n", US".\n", US".\n" };
662 if (fputs(CS ends[ch_state], fout) == EOF) return END_WERROR;
663 message_size += Ustrlen(ends[ch_state]);
664 body_linecount++;
665 }
666
667return END_EOF;
668}
669
670
671
672
673/*************************************************
674* Read data portion of an SMTP message *
675*************************************************/
676
677/* This function is called to read the remainder of an SMTP message (after the
678headers), or to skip over it when an error has occurred. In this case, the
679output file is passed as NULL.
680
681If any line begins with a dot, that character is skipped. The input should only
682be successfully terminated by CR LF . CR LF unless it is local (non-network)
683SMTP, in which case the CRs are optional, but...
684
685FUDGE: It seems that sites on the net send out messages with just LF
686terminators, despite the warnings in the RFCs, and other MTAs handle this. So
687we make the CRs optional in all cases.
688
689July 2003: Bare CRs cause trouble. We now treat them as line terminators as
690well, so that there are no CRs in spooled messages. However, the message
691terminating dot is not recognized between two bare CRs.
692
693Arguments:
694 fout a FILE to which to write the message; NULL if skipping
695
696Returns: One of the END_xxx values indicating why it stopped reading
697*/
698
699static int
700read_message_data_smtp(FILE *fout)
701{
702int ch_state = 0;
703register int ch;
704
fb2274d4 705while ((ch = (RECEIVE_GETC)()) != EOF)
059ec3d9
PH
706 {
707 if (ch == 0) body_zerocount++;
708 switch (ch_state)
709 {
710 case 0: /* After LF or CRLF */
711 if (ch == '.')
712 {
713 ch_state = 3;
714 continue; /* Don't ever write . after LF */
715 }
716 ch_state = 1;
717
718 /* Else fall through to handle as normal uschar. */
719
720 case 1: /* Normal state */
721 if (ch == '\n')
722 {
723 ch_state = 0;
724 body_linecount++;
725 }
726 else if (ch == '\r')
727 {
728 ch_state = 2;
729 continue;
730 }
731 break;
732
733 case 2: /* After (unwritten) CR */
734 body_linecount++;
735 if (ch == '\n')
736 {
737 ch_state = 0;
738 }
739 else
740 {
741 message_size++;
742 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
743 if (ch != '\r') ch_state = 1; else continue;
744 }
745 break;
746
747 case 3: /* After [CR] LF . */
748 if (ch == '\n')
749 return END_DOT;
750 if (ch == '\r')
751 {
752 ch_state = 4;
753 continue;
754 }
755 ch_state = 1; /* The dot itself is removed */
756 break;
757
758 case 4: /* After [CR] LF . CR */
759 if (ch == '\n') return END_DOT;
760 message_size++;
761 body_linecount++;
762 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
763 if (ch == '\r')
764 {
765 ch_state = 2;
766 continue;
767 }
768 ch_state = 1;
769 break;
770 }
771
772 /* Add the character to the spool file, unless skipping; then loop for the
773 next. */
774
775 message_size++;
776 if (fout != NULL)
777 {
778 if (fputc(ch, fout) == EOF) return END_WERROR;
779 if (message_size > thismessage_size_limit) return END_SIZE;
780 }
781 }
782
783/* Fall through here if EOF encountered. This indicates some kind of error,
784since a correct message is terminated by [CR] LF . [CR] LF. */
785
786return END_EOF;
787}
788
789
790
791
792/*************************************************
793* Swallow SMTP message *
794*************************************************/
795
796/* This function is called when there has been some kind of error while reading
797an SMTP message, and the remaining data may need to be swallowed. It is global
798because it is called from smtp_closedown() to shut down an incoming call
799tidily.
800
801Argument: a FILE from which to read the message
802Returns: nothing
803*/
804
805void
806receive_swallow_smtp(void)
807{
808if (message_ended >= END_NOTENDED)
809 message_ended = read_message_data_smtp(NULL);
810}
811
812
813
814/*************************************************
815* Handle lost SMTP connection *
816*************************************************/
817
818/* This function logs connection loss incidents and generates an appropriate
819SMTP response.
820
821Argument: additional data for the message
822Returns: the SMTP response
823*/
824
825static uschar *
826handle_lost_connection(uschar *s)
827{
828log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
829 "%s lost while reading message data%s", smtp_get_connection_info(), s);
830return US"421 Lost incoming connection";
831}
832
833
834
835
836/*************************************************
837* Handle a non-smtp reception error *
838*************************************************/
839
840/* This function is called for various errors during the reception of non-SMTP
841messages. It either sends a message to the sender of the problem message, or it
842writes to the standard error stream.
843
844Arguments:
845 errcode code for moan_to_sender(), identifying the error
846 text1 first message text, passed to moan_to_sender()
847 text2 second message text, used only for stderrr
848 error_rc code to pass to exim_exit if no problem
849 f FILE containing body of message (may be stdin)
850 hptr pointer to instore headers or NULL
851
852Returns: calls exim_exit(), which does not return
853*/
854
855static void
856give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
857 FILE *f, header_line *hptr)
858{
859if (error_handling == ERRORS_SENDER)
860 {
861 error_block eblock;
862 eblock.next = NULL;
863 eblock.text1 = text1;
864 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
865 error_rc = EXIT_FAILURE;
866 }
867else fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 868(void)fclose(f);
059ec3d9
PH
869exim_exit(error_rc);
870}
871
872
873
874/*************************************************
875* Add header lines set up by ACL *
876*************************************************/
877
850635b6
PH
878/* This function is called to add the header lines that were set up by
879statements in an ACL to the list of headers in memory. It is done in two stages
880like this, because when the ACL for RCPT is running, the other headers have not
881yet been received. This function is called twice; once just before running the
882DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
883are visible to the DATA ACL.
059ec3d9
PH
884
885Originally these header lines were added at the end. Now there is support for
886three different places: top, bottom, and after the Received: header(s). There
887will always be at least one Received: header, even if it is marked deleted, and
888even if something else has been put in front of it.
889
890Arguments:
891 acl_name text to identify which ACL
892
893Returns: nothing
894*/
895
896static void
897add_acl_headers(uschar *acl_name)
898{
899header_line *h, *next;
900header_line *last_received = NULL;
901
71fafd95 902if (acl_added_headers == NULL) return;
059ec3d9
PH
903DEBUG(D_receive|D_acl) debug_printf(">>Headers added by %s ACL:\n", acl_name);
904
71fafd95 905for (h = acl_added_headers; h != NULL; h = next)
059ec3d9
PH
906 {
907 next = h->next;
908
909 switch(h->type)
910 {
911 case htype_add_top:
912 h->next = header_list;
913 header_list = h;
914 DEBUG(D_receive|D_acl) debug_printf(" (at top)");
915 break;
916
917 case htype_add_rec:
918 if (last_received == NULL)
919 {
920 last_received = header_list;
921 while (!header_testname(last_received, US"Received", 8, FALSE))
922 last_received = last_received->next;
923 while (last_received->next != NULL &&
924 header_testname(last_received->next, US"Received", 8, FALSE))
925 last_received = last_received->next;
926 }
927 h->next = last_received->next;
928 last_received->next = h;
929 DEBUG(D_receive|D_acl) debug_printf(" (after Received:)");
930 break;
931
8523533c
TK
932 case htype_add_rfc:
933 /* add header before any header which is NOT Received: or Resent- */
934 last_received = header_list;
935 while ( (last_received->next != NULL) &&
936 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
937 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
938 last_received = last_received->next;
939 /* last_received now points to the last Received: or Resent-* header
940 in an uninterrupted chain of those header types (seen from the beginning
941 of all headers. Our current header must follow it. */
942 h->next = last_received->next;
943 last_received->next = h;
8e669ac1 944 DEBUG(D_receive|D_acl) debug_printf(" (before any non-Received: or Resent-*: header)");
8523533c
TK
945 break;
946
059ec3d9
PH
947 default:
948 h->next = NULL;
949 header_last->next = h;
950 break;
951 }
952
953 if (h->next == NULL) header_last = h;
954
955 /* Check for one of the known header types (From:, To:, etc.) though in
956 practice most added headers are going to be "other". Lower case
957 identification letters are never stored with the header; they are used
958 for existence tests when messages are received. So discard any lower case
959 flag values. */
960
961 h->type = header_checkname(h, FALSE);
962 if (h->type >= 'a') h->type = htype_other;
963
964 DEBUG(D_receive|D_acl) debug_printf(" %s", header_last->text);
965 }
966
71fafd95 967acl_added_headers = NULL;
059ec3d9
PH
968DEBUG(D_receive|D_acl) debug_printf(">>\n");
969}
970
971
972
973/*************************************************
974* Add host information for log line *
975*************************************************/
976
977/* Called for acceptance and rejecting log lines. This adds information about
978the calling host to a string that is being built dynamically.
979
980Arguments:
981 s the dynamic string
982 sizeptr points to the size variable
983 ptrptr points to the pointer variable
984
985Returns: the extended string
986*/
987
988static uschar *
989add_host_info_for_log(uschar *s, int *sizeptr, int *ptrptr)
990{
991if (sender_fullhost != NULL)
992 {
993 s = string_append(s, sizeptr, ptrptr, 2, US" H=", sender_fullhost);
994 if ((log_extra_selector & LX_incoming_interface) != 0 &&
995 interface_address != NULL)
996 {
997 uschar *ss = string_sprintf(" I=[%s]:%d", interface_address,
998 interface_port);
999 s = string_cat(s, sizeptr, ptrptr, ss, Ustrlen(ss));
1000 }
1001 }
1002if (sender_ident != NULL)
1003 s = string_append(s, sizeptr, ptrptr, 2, US" U=", sender_ident);
1004if (received_protocol != NULL)
1005 s = string_append(s, sizeptr, ptrptr, 2, US" P=", received_protocol);
1006return s;
1007}
1008
1009
1010
63955bf2 1011#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1012
1013/*************************************************
54cdb463
PH
1014* Run the MIME ACL on a message *
1015*************************************************/
1016
1017/* This code is in a subroutine so that it can be used for both SMTP
1018and non-SMTP messages. It is called with a non-NULL ACL pointer.
1019
1020Arguments:
1021 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1022 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1023 smtp_reply_ptr Where SMTP reply is being built
1024 blackholed_by_ptr Where "blackholed by" message is being built
1025
1026Returns: TRUE to carry on; FALSE to abandon the message
1027*/
1028
1029static BOOL
1030run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1031 uschar **blackholed_by_ptr)
1032{
1033FILE *mbox_file;
1034uschar rfc822_file_path[2048];
1035unsigned long mbox_size;
1036header_line *my_headerlist;
1037uschar *user_msg, *log_msg;
1038int mime_part_count_buffer = -1;
1039int rc;
1040
1041memset(CS rfc822_file_path,0,2048);
1042
1043/* check if it is a MIME message */
1044my_headerlist = header_list;
4e88a19f
PH
1045while (my_headerlist != NULL)
1046 {
54cdb463 1047 /* skip deleted headers */
4e88a19f
PH
1048 if (my_headerlist->type == '*')
1049 {
54cdb463
PH
1050 my_headerlist = my_headerlist->next;
1051 continue;
4e88a19f
PH
1052 }
1053 if (strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0)
1054 {
54cdb463
PH
1055 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1056 goto DO_MIME_ACL;
4e88a19f 1057 }
54cdb463 1058 my_headerlist = my_headerlist->next;
4e88a19f 1059 }
54cdb463
PH
1060
1061DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1062return TRUE;
1063
1064DO_MIME_ACL:
1065/* make sure the eml mbox file is spooled up */
1066mbox_file = spool_mbox(&mbox_size);
1067if (mbox_file == NULL) {
1068 /* error while spooling */
1069 log_write(0, LOG_MAIN|LOG_PANIC,
1070 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1071 Uunlink(spool_name);
1072 unspool_mbox();
a5bd321b 1073 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1074 message_id[0] = 0; /* Indicate no message accepted */
1075 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1076 return FALSE; /* Indicate skip to end of receive function */
1077};
1078
1079mime_is_rfc822 = 0;
1080
1081MIME_ACL_CHECK:
1082mime_part_count = -1;
1083rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1084(void)fclose(mbox_file);
54cdb463 1085
4e88a19f
PH
1086if (Ustrlen(rfc822_file_path) > 0)
1087 {
54cdb463
PH
1088 mime_part_count = mime_part_count_buffer;
1089
4e88a19f
PH
1090 if (unlink(CS rfc822_file_path) == -1)
1091 {
54cdb463
PH
1092 log_write(0, LOG_PANIC,
1093 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1094 goto END_MIME_ACL;
4e88a19f
PH
1095 }
1096 }
54cdb463
PH
1097
1098/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1099if (rc == OK)
1100 {
54cdb463
PH
1101 uschar temp_path[1024];
1102 int n;
1103 struct dirent *entry;
1104 DIR *tempdir;
1105
b07e6aa3
PH
1106 (void)string_format(temp_path, 1024, "%s/scan/%s", spool_directory,
1107 message_id);
54cdb463 1108
4e88a19f
PH
1109 tempdir = opendir(CS temp_path);
1110 n = 0;
1111 do
1112 {
1113 entry = readdir(tempdir);
1114 if (entry == NULL) break;
1115 if (strncmpic(US entry->d_name,US"__rfc822_",9) == 0)
1116 {
b07e6aa3 1117 (void)string_format(rfc822_file_path, 2048,"%s/scan/%s/%s", spool_directory, message_id, entry->d_name);
4e88a19f
PH
1118 debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n", rfc822_file_path);
1119 break;
1120 }
1121 } while (1);
1122 closedir(tempdir);
54cdb463 1123
4e88a19f
PH
1124 if (entry != NULL)
1125 {
c58b88df 1126 mbox_file = Ufopen(rfc822_file_path,"rb");
4e88a19f
PH
1127 if (mbox_file == NULL)
1128 {
54cdb463
PH
1129 log_write(0, LOG_PANIC,
1130 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1131 unlink(CS rfc822_file_path);
1132 goto END_MIME_ACL;
4e88a19f 1133 }
54cdb463
PH
1134 /* set RFC822 expansion variable */
1135 mime_is_rfc822 = 1;
1136 mime_part_count_buffer = mime_part_count;
1137 goto MIME_ACL_CHECK;
4e88a19f
PH
1138 }
1139 }
54cdb463
PH
1140
1141END_MIME_ACL:
1142add_acl_headers(US"MIME");
1143if (rc == DISCARD)
1144 {
1145 recipients_count = 0;
1146 *blackholed_by_ptr = US"MIME ACL";
1147 }
1148else if (rc != OK)
1149 {
1150 Uunlink(spool_name);
1151 unspool_mbox();
1152 if (smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1153 *smtp_yield_ptr = FALSE; /* No more messsages after dropped connection */
1154 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1155 message_id[0] = 0; /* Indicate no message accepted */
1156 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1157 }
54cdb463
PH
1158
1159return TRUE;
1160}
1161
63955bf2 1162#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1163
1164
1165/*************************************************
059ec3d9
PH
1166* Receive message *
1167*************************************************/
1168
1169/* Receive a message on the given input, and put it into a pair of spool files.
1170Either a non-null list of recipients, or the extract flag will be true, or
1171both. The flag sender_local is true for locally generated messages. The flag
1172submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a
PH
1173suppress_local_fixups is true if an ACL has obeyed "control =
1174suppress_local_fixups". The flag smtp_input is true if the message is to be
1175handled using SMTP conventions about termination and lines starting with dots.
1176For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1177
1178If a message was successfully read, message_id[0] will be non-zero.
1179
1180The general actions of this function are:
1181
1182 . Read the headers of the message (if any) into a chain of store
1183 blocks.
1184
1185 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1186 throw it away, unless the caller is trusted, or unless
1187 active_local_sender_retain is set - which can only happen if
1188 active_local_from_check is false.
059ec3d9
PH
1189
1190 . If recipients are to be extracted from the message, build the
1191 recipients list from the headers, removing any that were on the
1192 original recipients list (unless extract_addresses_remove_arguments is
1193 false), and at the same time, remove any bcc header that may be present.
1194
1195 . Get the spool file for the data, sort out its unique name, open
1196 and lock it (but don't give it the name yet).
1197
1198 . Generate a "Message-Id" header if the message doesn't have one, for
1199 locally-originated messages.
1200
1201 . Generate a "Received" header.
1202
1203 . Ensure the recipients list is fully qualified and rewritten if necessary.
1204
1205 . If there are any rewriting rules, apply them to the sender address
1206 and also to the headers.
1207
1208 . If there is no from: header, generate one, for locally-generated messages
1209 and messages in "submission mode" only.
1210
1211 . If the sender is local, check that from: is correct, and if not, generate
1212 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1213 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1214
1215 . If there is no "date" header, generate one, for locally-originated
1216 or submission mode messages only.
1217
1218 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1219 dot_ends mode, to the data file. Leave it open, to hold the lock.
1220
1221 . Write the envelope and the headers to a new file.
1222
1223 . Set the name for the header file; close it.
1224
1225 . Set the name for the data file; close it.
1226
1227Because this function can potentially be called many times in a single
1228SMTP connection, all store should be got by store_get(), so that it will be
1229automatically retrieved after the message is accepted.
1230
1231FUDGE: It seems that sites on the net send out messages with just LF
1232terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1233we make the CRs optional in all cases.
1234
1235July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1236new regime is now in place in which bare CRs in header lines are turned into LF
1237followed by a space, so as not to terminate the header line.
1238
1239February 2004: A bare LF in a header line in a message whose first line was
1240terminated by CRLF is treated in the same way as a bare CR.
1241
1242Arguments:
1243 extract_recip TRUE if recipients are to be extracted from the message's
1244 headers
1245
1246Returns: TRUE there are more messages to be read (SMTP input)
1247 FALSE there are no more messages to be read (non-SMTP input
1248 or SMTP connection collapsed, or other failure)
1249
1250When reading a message for filter testing, the returned value indicates
1251whether the headers (which is all that is read) were terminated by '.' or
1252not. */
1253
1254BOOL
1255receive_msg(BOOL extract_recip)
1256{
1257int i, rc;
1258int msg_size = 0;
1259int process_info_len = Ustrlen(process_info);
1260int error_rc = (error_handling == ERRORS_SENDER)?
1261 errors_sender_rc : EXIT_FAILURE;
1262int header_size = 256;
1263int start, end, domain, size, sptr;
1264int id_resolution;
1265int had_zero = 0;
1266
1267register int ptr = 0;
1268
1269BOOL contains_resent_headers = FALSE;
1270BOOL extracted_ignored = FALSE;
1271BOOL first_line_ended_crlf = TRUE_UNSET;
1272BOOL smtp_yield = TRUE;
1273BOOL yield = FALSE;
1274
1275BOOL resents_exist = FALSE;
1276uschar *resent_prefix = US"";
1277uschar *blackholed_by = NULL;
04f7d5b9 1278uschar *blackhole_log_msg = US"";
059ec3d9
PH
1279
1280flock_t lock_data;
1281error_block *bad_addresses = NULL;
1282
1283uschar *frozen_by = NULL;
1284uschar *queued_by = NULL;
1285
1286uschar *errmsg, *s;
1287struct stat statbuf;
1288
4e88a19f 1289/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1290
1291uschar *smtp_reply = NULL;
4e88a19f 1292uschar *user_msg, *log_msg;
059ec3d9
PH
1293
1294/* Working header pointers */
1295
1296header_line *h, *next;
1297
2cbb4081 1298/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1299
1300BOOL date_header_exists = FALSE;
1301
1302/* Pointers to receive the addresses of headers whose contents we need. */
1303
1304header_line *from_header = NULL;
1305header_line *subject_header = NULL;
1306header_line *msgid_header = NULL;
1307header_line *received_header;
1308
1309/* Variables for use when building the Received: header. */
1310
1311uschar *received;
1312uschar *timestamp;
1313int tslen;
1314
1315/* Release any open files that might have been cached while preparing to
1316accept the message - e.g. by verifying addresses - because reading a message
1317might take a fair bit of real time. */
1318
1319search_tidyup();
1320
1321/* Initialize the chain of headers by setting up a place-holder for Received:
1322header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1323pointing to the end of the chain to make adding headers simple. */
1324
1325received_header = header_list = header_last = store_get(sizeof(header_line));
1326header_list->next = NULL;
1327header_list->type = htype_old;
1328header_list->text = NULL;
1329header_list->slen = 0;
1330
1331/* Control block for the next header to be read. */
1332
1333next = store_get(sizeof(header_line));
1334next->text = store_get(header_size);
1335
1336/* Initialize message id to be null (indicating no message read), and the
1337header names list to be the normal list. Indicate there is no data file open
1338yet, initialize the size and warning count, and deal with no size limit. */
1339
1340message_id[0] = 0;
1341data_file = NULL;
1342data_fd = -1;
1343spool_name[0] = 0;
1344message_size = 0;
1345warning_count = 0;
1346received_count = 1; /* For the one we will add */
1347
1348if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1349
2e0c1448 1350/* While reading the message, the following counts are computed. */
059ec3d9 1351
2e0c1448 1352message_linecount = body_linecount = body_zerocount = 0;
059ec3d9 1353
fb2274d4
TK
1354#ifdef EXPERIMENTAL_DOMAINKEYS
1355/* Call into DK to set up the context. Check if DK is to be run are carried out
1356 inside dk_exim_verify_init(). */
1357dk_exim_verify_init();
1358#endif
1359
059ec3d9
PH
1360/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1361ids, and fractions of a second are required. See the comments that precede the
1362message id creation below. */
1363
1364(void)gettimeofday(&message_id_tv, NULL);
1365
1366/* For other uses of the received time we can operate with granularity of one
1367second, and for that we use the global variable received_time. This is for
1368things like ultimate message timeouts. */
1369
1370received_time = message_id_tv.tv_sec;
1371
1372/* If SMTP input, set the special handler for timeouts. The alarm() calls
1373happen in the smtp_getc() function when it refills its buffer. */
1374
1375if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1376
1377/* If not SMTP input, timeout happens only if configured, and we just set a
1378single timeout for the whole message. */
1379
1380else if (receive_timeout > 0)
1381 {
1382 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1383 alarm(receive_timeout);
1384 }
1385
1386/* SIGTERM and SIGINT are caught always. */
1387
1388signal(SIGTERM, data_sigterm_sigint_handler);
1389signal(SIGINT, data_sigterm_sigint_handler);
1390
1391/* Header lines in messages are not supposed to be very long, though when
1392unfolded, to: and cc: headers can take up a lot of store. We must also cope
1393with the possibility of junk being thrown at us. Start by getting 256 bytes for
1394storing the header, and extend this as necessary using string_cat().
1395
1396To cope with total lunacies, impose an upper limit on the length of the header
1397section of the message, as otherwise the store will fill up. We must also cope
1398with the possibility of binary zeros in the data. Hence we cannot use fgets().
1399Folded header lines are joined into one string, leaving the '\n' characters
1400inside them, so that writing them out reproduces the input.
1401
1402Loop for each character of each header; the next structure for chaining the
1403header is set up already, with ptr the offset of the next character in
1404next->text. */
1405
1406for (;;)
1407 {
fb2274d4 1408 int ch = (RECEIVE_GETC)();
059ec3d9
PH
1409
1410 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1411 SMTP must have a correct "." terminator. */
1412
1413 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1414 {
1415 smtp_reply = handle_lost_connection(US" (header)");
1416 smtp_yield = FALSE;
1417 goto TIDYUP; /* Skip to end of function */
1418 }
1419
1420 /* See if we are at the current header's size limit - there must be at least
1421 four bytes left. This allows for the new character plus a zero, plus two for
1422 extra insertions when we are playing games with dots and carriage returns. If
1423 we are at the limit, extend the text buffer. This could have been done
1424 automatically using string_cat() but because this is a tightish loop storing
1425 only one character at a time, we choose to do it inline. Normally
1426 store_extend() will be able to extend the block; only at the end of a big
1427 store block will a copy be needed. To handle the case of very long headers
1428 (and sometimes lunatic messages can have ones that are 100s of K long) we
1429 call store_release() for strings that have been copied - if the string is at
1430 the start of a block (and therefore the only thing in it, because we aren't
1431 doing any other gets), the block gets freed. We can only do this because we
1432 know there are no other calls to store_get() going on. */
1433
1434 if (ptr >= header_size - 4)
1435 {
1436 int oldsize = header_size;
1437 /* header_size += 256; */
1438 header_size *= 2;
1439 if (!store_extend(next->text, oldsize, header_size))
1440 {
1441 uschar *newtext = store_get(header_size);
1442 memcpy(newtext, next->text, ptr);
1443 store_release(next->text);
1444 next->text = newtext;
1445 }
1446 }
1447
1448 /* Cope with receiving a binary zero. There is dispute about whether
1449 these should be allowed in RFC 822 messages. The middle view is that they
1450 should not be allowed in headers, at least. Exim takes this attitude at
1451 the moment. We can't just stomp on them here, because we don't know that
1452 this line is a header yet. Set a flag to cause scanning later. */
1453
1454 if (ch == 0) had_zero++;
1455
1456 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1457 those from data files use just LF. Treat LF in local SMTP input as a
1458 terminator too. Treat EOF as a line terminator always. */
1459
1460 if (ch == EOF) goto EOL;
1461
1462 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1463 other MTAs accept this. We are therefore forced into this "liberalisation"
1464 too, so we accept LF as a line terminator whatever the source of the message.
1465 However, if the first line of the message ended with a CRLF, we treat a bare
1466 LF specially by inserting a white space after it to ensure that the header
1467 line is not terminated. */
1468
1469 if (ch == '\n')
1470 {
1471 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
fb2274d4 1472 else if (first_line_ended_crlf) RECEIVE_UNGETC(' ');
059ec3d9
PH
1473 goto EOL;
1474 }
1475
1476 /* This is not the end of the line. If this is SMTP input and this is
1477 the first character in the line and it is a "." character, ignore it.
1478 This implements the dot-doubling rule, though header lines starting with
1479 dots aren't exactly common. They are legal in RFC 822, though. If the
1480 following is CRLF or LF, this is the line that that terminates the
1481 entire message. We set message_ended to indicate this has happened (to
1482 prevent further reading), and break out of the loop, having freed the
1483 empty header, and set next = NULL to indicate no data line. */
1484
1485 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1486 {
fb2274d4 1487 ch = (RECEIVE_GETC)();
059ec3d9
PH
1488 if (ch == '\r')
1489 {
fb2274d4 1490 ch = (RECEIVE_GETC)();
059ec3d9
PH
1491 if (ch != '\n')
1492 {
fb2274d4 1493 RECEIVE_UNGETC(ch);
059ec3d9
PH
1494 ch = '\r'; /* Revert to CR */
1495 }
1496 }
1497 if (ch == '\n')
1498 {
1499 message_ended = END_DOT;
1500 store_reset(next);
1501 next = NULL;
1502 break; /* End character-reading loop */
1503 }
1504
1505 /* For non-SMTP input, the dot at the start of the line was really a data
1506 character. What is now in ch is the following character. We guaranteed
1507 enough space for this above. */
1508
1509 if (!smtp_input)
1510 {
1511 next->text[ptr++] = '.';
1512 message_size++;
1513 }
1514 }
1515
1516 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1517 remember this case if this is the first line ending. */
1518
1519 if (ch == '\r')
1520 {
fb2274d4 1521 ch = (RECEIVE_GETC)();
059ec3d9
PH
1522 if (ch == '\n')
1523 {
1524 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1525 goto EOL;
1526 }
1527
1528 /* Otherwise, put back the character after CR, and turn the bare CR
1529 into LF SP. */
1530
fb2274d4 1531 ch = (RECEIVE_UNGETC)(ch);
059ec3d9
PH
1532 next->text[ptr++] = '\n';
1533 message_size++;
1534 ch = ' ';
1535 }
1536
1537 /* We have a data character for the header line. */
1538
1539 next->text[ptr++] = ch; /* Add to buffer */
1540 message_size++; /* Total message size so far */
1541
1542 /* Handle failure due to a humungously long header section. The >= allows
1543 for the terminating \n. Add what we have so far onto the headers list so
1544 that it gets reflected in any error message, and back up the just-read
1545 character. */
1546
1547 if (message_size >= header_maxsize)
1548 {
1549 next->text[ptr] = 0;
1550 next->slen = ptr;
1551 next->type = htype_other;
1552 next->next = NULL;
1553 header_last->next = next;
1554 header_last = next;
1555
1556 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1557 "%s (more than %d characters): message abandoned",
1558 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1559
1560 if (smtp_input)
1561 {
1562 smtp_reply = US"552 Message header is ridiculously long";
1563 receive_swallow_smtp();
1564 goto TIDYUP; /* Skip to end of function */
1565 }
1566
1567 else
1568 {
1569 give_local_error(ERRMESS_VLONGHEADER,
1570 string_sprintf("message header longer than %d characters received: "
1571 "message not accepted", header_maxsize), US"", error_rc, stdin,
1572 header_list->next);
1573 /* Does not return */
1574 }
1575 }
1576
1577 continue; /* With next input character */
1578
1579 /* End of header line reached */
1580
1581 EOL:
2e0c1448
PH
1582
1583 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1584
1585 receive_linecount++;
1586 message_linecount++;
059ec3d9
PH
1587
1588 /* Now put in the terminating newline. There is always space for
1589 at least two more characters. */
1590
1591 next->text[ptr++] = '\n';
1592 message_size++;
1593
1594 /* A blank line signals the end of the headers; release the unwanted
1595 space and set next to NULL to indicate this. */
1596
1597 if (ptr == 1)
1598 {
1599 store_reset(next);
1600 next = NULL;
1601 break;
1602 }
1603
1604 /* There is data in the line; see if the next input character is a
1605 whitespace character. If it is, we have a continuation of this header line.
1606 There is always space for at least one character at this point. */
1607
1608 if (ch != EOF)
1609 {
fb2274d4 1610 int nextch = (RECEIVE_GETC)();
059ec3d9
PH
1611 if (nextch == ' ' || nextch == '\t')
1612 {
1613 next->text[ptr++] = nextch;
1614 message_size++;
1615 continue; /* Iterate the loop */
1616 }
fb2274d4 1617 else if (nextch != EOF) (RECEIVE_UNGETC)(nextch); /* For next time */
059ec3d9
PH
1618 else ch = EOF; /* Cause main loop to exit at end */
1619 }
1620
1621 /* We have got to the real line end. Terminate the string and release store
1622 beyond it. If it turns out to be a real header, internal binary zeros will
1623 be squashed later. */
1624
1625 next->text[ptr] = 0;
1626 next->slen = ptr;
1627 store_reset(next->text + ptr + 1);
1628
1629 /* Check the running total size against the overall message size limit. We
1630 don't expect to fail here, but if the overall limit is set less than MESSAGE_
1631 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
1632 headers - the code to read the body will then also hit the buffer. */
1633
1634 if (message_size > thismessage_size_limit) break;
1635
1636 /* A line that is not syntactically correct for a header also marks
1637 the end of the headers. In this case, we leave next containing the
1638 first data line. This might actually be several lines because of the
1639 continuation logic applied above, but that doesn't matter.
1640
1641 It turns out that smail, and presumably sendmail, accept leading lines
1642 of the form
1643
1644 From ph10 Fri Jan 5 12:35 GMT 1996
1645
1646 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
1647 find any documentation of this, but for compatibility it had better be
1648 accepted. Exim restricts it to the case of non-smtp messages, and
1649 treats it as an alternative to the -f command line option. Thus it is
1650 ignored except for trusted users or filter testing. Otherwise it is taken
1651 as the sender address, unless -f was used (sendmail compatibility).
1652
1653 It further turns out that some UUCPs generate the From_line in a different
1654 format, e.g.
1655
1656 From ph10 Fri, 7 Jan 97 14:00:00 GMT
1657
1658 The regex for matching these things is now capable of recognizing both
1659 formats (including 2- and 4-digit years in the latter). In fact, the regex
1660 is now configurable, as is the expansion string to fish out the sender.
1661
1662 Even further on it has been discovered that some broken clients send
1663 these lines in SMTP messages. There is now an option to ignore them from
1664 specified hosts or networks. Sigh. */
1665
1666 if (header_last == header_list &&
1667 (!smtp_input
1668 ||
1669 (sender_host_address != NULL &&
1670 verify_check_host(&ignore_fromline_hosts) == OK)
1671 ||
1672 (sender_host_address == NULL && ignore_fromline_local)
1673 ) &&
1674 regex_match_and_setup(regex_From, next->text, 0, -1))
1675 {
1676 if (!sender_address_forced)
1677 {
1678 uschar *uucp_sender = expand_string(uucp_from_sender);
1679 if (uucp_sender == NULL)
1680 {
1681 log_write(0, LOG_MAIN|LOG_PANIC,
1682 "expansion of \"%s\" failed after matching "
1683 "\"From \" line: %s", uucp_from_sender, expand_string_message);
1684 }
1685 else
1686 {
1687 int start, end, domain;
1688 uschar *errmess;
1689 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
1690 &start, &end, &domain, TRUE);
1691 if (newsender != NULL)
1692 {
1693 if (domain == 0 && newsender[0] != 0)
1694 newsender = rewrite_address_qualify(newsender, FALSE);
1695
f05da2e8 1696 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
1697 {
1698 sender_address = newsender;
1699
f05da2e8 1700 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
1701 {
1702 authenticated_sender = NULL;
1703 originator_name = US"";
1704 sender_local = FALSE;
1705 }
1706
f05da2e8 1707 if (filter_test != FTEST_NONE)
059ec3d9
PH
1708 printf("Sender taken from \"From \" line\n");
1709 }
1710 }
1711 }
1712 }
1713 }
1714
1715 /* Not a leading "From " line. Check to see if it is a valid header line.
1716 Header names may contain any non-control characters except space and colon,
1717 amazingly. */
1718
1719 else
1720 {
1721 uschar *p = next->text;
1722
1723 /* If not a valid header line, break from the header reading loop, leaving
1724 next != NULL, indicating that it holds the first line of the body. */
1725
1726 if (isspace(*p)) break;
1727 while (mac_isgraph(*p) && *p != ':') p++;
1728 while (isspace(*p)) p++;
1729 if (*p != ':')
1730 {
1731 body_zerocount = had_zero;
1732 break;
1733 }
1734
1735 /* We have a valid header line. If there were any binary zeroes in
1736 the line, stomp on them here. */
1737
1738 if (had_zero > 0)
1739 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
1740
1741 /* It is perfectly legal to have an empty continuation line
1742 at the end of a header, but it is confusing to humans
1743 looking at such messages, since it looks like a blank line.
1744 Reduce confusion by removing redundant white space at the
1745 end. We know that there is at least one printing character
1746 (the ':' tested for above) so there is no danger of running
1747 off the end. */
1748
1749 p = next->text + ptr - 2;
1750 for (;;)
1751 {
1752 while (*p == ' ' || *p == '\t') p--;
1753 if (*p != '\n') break;
1754 ptr = (p--) - next->text + 1;
1755 message_size -= next->slen - ptr;
1756 next->text[ptr] = 0;
1757 next->slen = ptr;
1758 }
1759
1760 /* Add the header to the chain */
1761
1762 next->type = htype_other;
1763 next->next = NULL;
1764 header_last->next = next;
1765 header_last = next;
1766
1767 /* Check the limit for individual line lengths. This comes after adding to
1768 the chain so that the failing line is reflected if a bounce is generated
1769 (for a local message). */
1770
1771 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
1772 {
1773 log_write(0, LOG_MAIN, "overlong message header line received from "
1774 "%s (more than %d characters): message abandoned",
1775 sender_host_unknown? sender_ident : sender_fullhost,
1776 header_line_maxsize);
1777
1778 if (smtp_input)
1779 {
1780 smtp_reply = US"552 A message header line is too long";
1781 receive_swallow_smtp();
1782 goto TIDYUP; /* Skip to end of function */
1783 }
1784
1785 else
1786 {
1787 give_local_error(ERRMESS_VLONGHDRLINE,
1788 string_sprintf("message header line longer than %d characters "
1789 "received: message not accepted", header_line_maxsize), US"",
1790 error_rc, stdin, header_list->next);
1791 /* Does not return */
1792 }
1793 }
1794
1795 /* Note if any resent- fields exist. */
1796
1797 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
1798 {
1799 resents_exist = TRUE;
1800 resent_prefix = US"Resent-";
1801 }
1802 }
1803
1804 /* The line has been handled. If we have hit EOF, break out of the loop,
1805 indicating no pending data line. */
1806
1807 if (ch == EOF) { next = NULL; break; }
1808
1809 /* Set up for the next header */
1810
1811 header_size = 256;
1812 next = store_get(sizeof(header_line));
1813 next->text = store_get(header_size);
1814 ptr = 0;
1815 had_zero = 0;
1816 } /* Continue, starting to read the next header */
1817
1818/* At this point, we have read all the headers into a data structure in main
1819store. The first header is still the dummy placeholder for the Received: header
1820we are going to generate a bit later on. If next != NULL, it contains the first
1821data line - which terminated the headers before reaching a blank line (not the
1822normal case). */
1823
1824DEBUG(D_receive)
1825 {
1826 debug_printf(">>Headers received:\n");
1827 for (h = header_list->next; h != NULL; h = h->next)
1828 debug_printf("%s", h->text);
1829 debug_printf("\n");
1830 }
1831
1832/* End of file on any SMTP connection is an error. If an incoming SMTP call
1833is dropped immediately after valid headers, the next thing we will see is EOF.
1834We must test for this specially, as further down the reading of the data is
1835skipped if already at EOF. */
1836
1837if (smtp_input && (receive_feof)())
1838 {
1839 smtp_reply = handle_lost_connection(US" (after header)");
1840 smtp_yield = FALSE;
1841 goto TIDYUP; /* Skip to end of function */
1842 }
1843
1844/* If this is a filter test run and no headers were read, output a warning
1845in case there is a mistake in the test message. */
1846
f05da2e8 1847if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
1848 printf("Warning: no message headers read\n");
1849
1850
1851/* Scan the headers to identify them. Some are merely marked for later
1852processing; some are dealt with here. */
1853
1854for (h = header_list->next; h != NULL; h = h->next)
1855 {
1856 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
1857 if (is_resent) contains_resent_headers = TRUE;
1858
1859 switch (header_checkname(h, is_resent))
1860 {
059ec3d9 1861 case htype_bcc:
2cbb4081 1862 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
1863 break;
1864
059ec3d9 1865 case htype_cc:
2cbb4081 1866 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
1867 break;
1868
1869 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
1870
1871 case htype_date:
1872 date_header_exists = !resents_exist || is_resent;
1873 break;
1874
1875 /* Same comments as about Return-Path: below. */
1876
1877 case htype_delivery_date:
1878 if (delivery_date_remove) h->type = htype_old;
1879 break;
1880
1881 /* Same comments as about Return-Path: below. */
1882
1883 case htype_envelope_to:
1884 if (envelope_to_remove) h->type = htype_old;
1885 break;
1886
1887 /* Mark all "From:" headers so they get rewritten. Save the one that is to
1888 be used for Sender: checking. For Sendmail compatibility, if the "From:"
1889 header consists of just the login id of the user who called Exim, rewrite
1890 it with the gecos field first. Apply this rule to Resent-From: if there
1891 are resent- fields. */
1892
1893 case htype_from:
1894 h->type = htype_from;
1895 if (!resents_exist || is_resent)
1896 {
1897 from_header = h;
1898 if (!smtp_input)
1899 {
1900 uschar *s = Ustrchr(h->text, ':') + 1;
1901 while (isspace(*s)) s++;
1902 if (strncmpic(s, originator_login, h->slen - (s - h->text) - 1) == 0)
1903 {
1904 uschar *name = is_resent? US"Resent-From" : US"From";
1905 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
1906 originator_login, qualify_domain_sender);
1907 from_header = header_last;
1908 h->type = htype_old;
1909 DEBUG(D_receive|D_rewrite)
1910 debug_printf("rewrote \"%s:\" header using gecos\n", name);
1911 }
1912 }
1913 }
1914 break;
1915
1916 /* Identify the Message-id: header for generating "in-reply-to" in the
1917 autoreply transport. For incoming logging, save any resent- value. In both
1918 cases, take just the first of any multiples. */
1919
1920 case htype_id:
1921 if (msgid_header == NULL && (!resents_exist || is_resent))
1922 {
1923 msgid_header = h;
1924 h->type = htype_id;
1925 }
1926 break;
1927
1928 /* Flag all Received: headers */
1929
1930 case htype_received:
1931 h->type = htype_received;
1932 received_count++;
1933 break;
1934
1935 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
1936
1937 case htype_reply_to:
1938 h->type = htype_reply_to;
1939 break;
1940
1941 /* The Return-path: header is supposed to be added to messages when
1942 they leave the SMTP system. We shouldn't receive messages that already
1943 contain Return-path. However, since Exim generates Return-path: on
1944 local delivery, resent messages may well contain it. We therefore
1945 provide an option (which defaults on) to remove any Return-path: headers
1946 on input. Removal actually means flagging as "old", which prevents the
1947 header being transmitted with the message. */
1948
1949 case htype_return_path:
1950 if (return_path_remove) h->type = htype_old;
1951
1952 /* If we are testing a mail filter file, use the value of the
1953 Return-Path: header to set up the return_path variable, which is not
1954 otherwise set. However, remove any <> that surround the address
1955 because the variable doesn't have these. */
1956
f05da2e8 1957 if (filter_test != FTEST_NONE)
059ec3d9
PH
1958 {
1959 uschar *start = h->text + 12;
1960 uschar *end = start + Ustrlen(start);
1961 while (isspace(*start)) start++;
1962 while (end > start && isspace(end[-1])) end--;
1963 if (*start == '<' && end[-1] == '>')
1964 {
1965 start++;
1966 end--;
1967 }
1968 return_path = string_copyn(start, end - start);
1969 printf("Return-path taken from \"Return-path:\" header line\n");
1970 }
1971 break;
1972
1973 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
1974 and from an untrusted caller and suppress_local_fixups is not set, or if we
1975 are in submission mode for a remote message, mark it "old" so that it will
1976 not be transmitted with the message, unless active_local_sender_retain is
1977 set. (This can only be true if active_local_from_check is false.) If there
1978 are any resent- headers in the message, apply this rule to Resent-Sender:
1979 instead of Sender:. Messages with multiple resent- header sets cannot be
1980 tidily handled. (For this reason, at least one MUA - Pine - turns old
1981 resent- headers into X-resent- headers when resending, leaving just one
1982 set.) */
059ec3d9
PH
1983
1984 case htype_sender:
69358f02 1985 h->type = ((!active_local_sender_retain &&
8800895a
PH
1986 (
1987 (sender_local && !trusted_caller && !suppress_local_fixups)
1988 || submission_mode
1989 )
059ec3d9
PH
1990 ) &&
1991 (!resents_exist||is_resent))?
1992 htype_old : htype_sender;
1993 break;
1994
1995 /* Remember the Subject: header for logging. There is no Resent-Subject */
1996
1997 case htype_subject:
1998 subject_header = h;
1999 break;
2000
2001 /* "To:" gets flagged, and the existence of a recipient header is noted,
2002 whether it's resent- or not. */
2003
2004 case htype_to:
2005 h->type = htype_to;
2006 /****
2007 to_or_cc_header_exists = TRUE;
2008 ****/
2009 break;
2010 }
2011 }
2012
2013/* Extract recipients from the headers if that is required (the -t option).
2014Note that this is documented as being done *before* any address rewriting takes
2015place. There are two possibilities:
2016
2017(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2018recipients already listed are to be REMOVED from the message. Smail 3 works
2019like this. We need to build a non-recipients tree for that list, because in
2020subsequent processing this data is held in a tree and that's what the
2021spool_write_header() function expects. Make sure that non-recipient addresses
2022are fully qualified and rewritten if necessary.
2023
2024(2) According to other sendmail documentation, -t ADDS extracted recipients to
2025those in the command line arguments (and it is rumoured some other MTAs do
2026this). Therefore, there is an option to make Exim behave this way.
2027
2028*** Notes on "Resent-" header lines ***
2029
2030The presence of resent-headers in the message makes -t horribly ambiguous.
2031Experiments with sendmail showed that it uses recipients for all resent-
2032headers, totally ignoring the concept of "sets of resent- headers" as described
2033in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2034with all the addresses in one instance of each header.
2035
2036This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2037error for -t if there were resent- headers in the message. However, after a
2038discussion on the mailing list, I've learned that there are MUAs that use
2039resent- headers with -t, and also that the stuff about sets of resent- headers
2040and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2041message with -t and resent- header lines makes sure that only *its* resent-
2042headers are present; previous ones are often renamed as X-resent- for example.
2043
2044Consequently, Exim has been changed so that, if any resent- header lines are
2045present, the recipients are taken from all of the appropriate resent- lines,
2046and not from the ordinary To:, Cc:, etc. */
2047
2048if (extract_recip)
2049 {
2050 int rcount = 0;
2051 error_block **bnext = &bad_addresses;
2052
2053 if (extract_addresses_remove_arguments)
2054 {
2055 while (recipients_count-- > 0)
2056 {
2057 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2058 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2059 tree_add_nonrecipient(s);
2060 }
2061 recipients_list = NULL;
2062 recipients_count = recipients_list_max = 0;
2063 }
2064
059ec3d9
PH
2065 /* Now scan the headers */
2066
2067 for (h = header_list->next; h != NULL; h = h->next)
2068 {
2069 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2070 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2071 {
2072 uschar *s = Ustrchr(h->text, ':') + 1;
2073 while (isspace(*s)) s++;
2074
1eccaa59
PH
2075 parse_allow_group = TRUE; /* Allow address group syntax */
2076
059ec3d9
PH
2077 while (*s != 0)
2078 {
2079 uschar *ss = parse_find_address_end(s, FALSE);
2080 uschar *recipient, *errmess, *p, *pp;
2081 int start, end, domain;
2082
2083 /* Check on maximum */
2084
2085 if (recipients_max > 0 && ++rcount > recipients_max)
2086 {
2087 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2088 US"message rejected: ", error_rc, stdin, NULL);
2089 /* Does not return */
2090 }
2091
2092 /* Make a copy of the address, and remove any internal newlines. These
2093 may be present as a result of continuations of the header line. The
2094 white space that follows the newline must not be removed - it is part
2095 of the header. */
2096
2097 pp = recipient = store_get(ss - s + 1);
2098 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2099 *pp = 0;
2100 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2101 &domain, FALSE);
2102
2103 /* Keep a list of all the bad addresses so we can send a single
2104 error message at the end. However, an empty address is not an error;
2105 just ignore it. This can come from an empty group list like
2106
2107 To: Recipients of list:;
2108
2109 If there are no recipients at all, an error will occur later. */
2110
2111 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2112 {
2113 int len = Ustrlen(s);
2114 error_block *b = store_get(sizeof(error_block));
2115 while (len > 0 && isspace(s[len-1])) len--;
2116 b->next = NULL;
2117 b->text1 = string_printing(string_copyn(s, len));
2118 b->text2 = errmess;
2119 *bnext = b;
2120 bnext = &(b->next);
2121 }
2122
2123 /* If the recipient is already in the nonrecipients tree, it must
2124 have appeared on the command line with the option extract_addresses_
2125 remove_arguments set. Do not add it to the recipients, and keep a note
2126 that this has happened, in order to give a better error if there are
2127 no recipients left. */
2128
2129 else if (recipient != NULL)
2130 {
2131 if (tree_search(tree_nonrecipients, recipient) == NULL)
2132 receive_add_recipient(recipient, -1);
2133 else
2134 extracted_ignored = TRUE;
2135 }
2136
2137 /* Move on past this address */
2138
2139 s = ss + (*ss? 1:0);
2140 while (isspace(*s)) s++;
1eccaa59
PH
2141 } /* Next address */
2142
2143 parse_allow_group = FALSE; /* Reset group syntax flags */
2144 parse_found_group = FALSE;
059ec3d9
PH
2145
2146 /* If this was the bcc: header, mark it "old", which means it
2147 will be kept on the spool, but not transmitted as part of the
2148 message. */
2149
2cbb4081 2150 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2151 } /* For appropriate header line */
2152 } /* For each header line */
2153
059ec3d9
PH
2154 }
2155
2156/* Now build the unique message id. This has changed several times over the
2157lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2158Retaining all the history in the comment has become too unwieldy - read
2159previous release sources if you want it.
2160
2161The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2162The first part is the current time, in seconds. The second part is the current
2163pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2164can hold a number in the range 0-3843. It used to be a computed sequence
2165number, but is now the fractional component of the current time in units of
21661/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2167received, Exim ensures that the timer has ticked at the appropriate level
2168before proceeding, to avoid duplication if the pid happened to be re-used
2169within the same time period. It seems likely that most messages will take at
2170least half a millisecond to be received, so no delay will normally be
2171necessary. At least for some time...
2172
2173There is a modification when localhost_number is set. Formerly this was allowed
2174to be as large as 255. Now it is restricted to the range 0-16, and the final
2175component of the message id becomes (localhost_number * 200) + fractional time
2176in units of 1/200 of a second (i.e. a value in the range 0-3399).
2177
2178Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2179Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2180still allows the tttttt field to hold a large enough number to last for some
2181more decades, and the final two-digit field can hold numbers up to 1295, which
2182is enough for milliseconds (instead of 1/2000 of a second).
2183
2184However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2185pid, so it is probably safe because pids have to be positive. The
2186localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2187final field becomes (localhost_number * 100) + fractional time in centiseconds.
2188
2189Note that string_base62() returns its data in a static storage block, so it
2190must be copied before calling string_base62() again. It always returns exactly
21916 characters.
2192
2193There doesn't seem to be anything in the RFC which requires a message id to
2194start with a letter, but Smail was changed to ensure this. The external form of
2195the message id (as supplied by string expansion) therefore starts with an
2196additional leading 'E'. The spool file names do not include this leading
2197letter and it is not used internally.
2198
2199NOTE: If ever the format of message ids is changed, the regular expression for
2200checking that a string is in this format must be updated in a corresponding
2201way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2202must also be changed to reflect the correct string length. Then, of course,
2203other programs that rely on the message id format will need updating too. */
2204
2205Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2206message_id[6] = '-';
2207Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2208
2209/* Deal with the case where the host number is set. The value of the number was
2210checked when it was read, to ensure it isn't too big. The timing granularity is
2211left in id_resolution so that an appropriate wait can be done after receiving
2212the message, if necessary (we hope it won't be). */
2213
2214if (host_number_string != NULL)
2215 {
2216 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2217 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2218 string_base62((long int)(
2219 host_number * (1000000/id_resolution) +
2220 message_id_tv.tv_usec/id_resolution)) + 4);
2221 }
2222
2223/* Host number not set: final field is just the fractional time at an
2224appropriate resolution. */
2225
2226else
2227 {
2228 id_resolution = (BASE_62 == 62)? 500 : 1000;
2229 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2230 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2231 }
2232
2233/* Add the current message id onto the current process info string if
2234it will fit. */
2235
2236(void)string_format(process_info + process_info_len,
2237 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2238
2239/* If we are using multiple input directories, set up the one for this message
2240to be the least significant base-62 digit of the time of arrival. Otherwise
2241ensure that it is an empty string. */
2242
2243message_subdir[0] = split_spool_directory? message_id[5] : 0;
2244
2245/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2246one, but only for local (without suppress_local_fixups) or submission mode
2247messages. This can be user-configured if required, but we had better flatten
2248any illegal characters therein. */
059ec3d9 2249
8800895a
PH
2250if (msgid_header == NULL &&
2251 ((sender_host_address == NULL && !suppress_local_fixups)
2252 || submission_mode))
059ec3d9
PH
2253 {
2254 uschar *p;
2255 uschar *id_text = US"";
2256 uschar *id_domain = primary_hostname;
2257
2258 /* Permit only letters, digits, dots, and hyphens in the domain */
2259
2260 if (message_id_domain != NULL)
2261 {
2262 uschar *new_id_domain = expand_string(message_id_domain);
2263 if (new_id_domain == NULL)
2264 {
2265 if (!expand_string_forcedfail)
2266 log_write(0, LOG_MAIN|LOG_PANIC,
2267 "expansion of \"%s\" (message_id_header_domain) "
2268 "failed: %s", message_id_domain, expand_string_message);
2269 }
2270 else if (*new_id_domain != 0)
2271 {
2272 id_domain = new_id_domain;
2273 for (p = id_domain; *p != 0; p++)
2274 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2275 }
2276 }
2277
2278 /* Permit all characters except controls and RFC 2822 specials in the
2279 additional text part. */
2280
2281 if (message_id_text != NULL)
2282 {
2283 uschar *new_id_text = expand_string(message_id_text);
2284 if (new_id_text == NULL)
2285 {
2286 if (!expand_string_forcedfail)
2287 log_write(0, LOG_MAIN|LOG_PANIC,
2288 "expansion of \"%s\" (message_id_header_text) "
2289 "failed: %s", message_id_text, expand_string_message);
2290 }
2291 else if (*new_id_text != 0)
2292 {
2293 id_text = new_id_text;
2294 for (p = id_text; *p != 0; p++)
2295 if (mac_iscntrl_or_special(*p)) *p = '-';
2296 }
2297 }
2298
2299 /* Add the header line */
2300
2301 header_add(htype_id, "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix,
2302 message_id_external, (*id_text == 0)? "" : ".", id_text, id_domain);
2303 }
2304
2305/* If we are to log recipients, keep a copy of the raw ones before any possible
2306rewriting. Must copy the count, because later ACLs and the local_scan()
2307function may mess with the real recipients. */
2308
2309if ((log_extra_selector & LX_received_recipients) != 0)
2310 {
2311 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2312 for (i = 0; i < recipients_count; i++)
2313 raw_recipients[i] = string_copy(recipients_list[i].address);
2314 raw_recipients_count = recipients_count;
2315 }
2316
2317/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2318recipients will get here only if the conditions were right (allow_unqualified_
2319recipient is TRUE). */
2320
2321for (i = 0; i < recipients_count; i++)
2322 recipients_list[i].address =
2323 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2324 global_rewrite_rules, rewrite_existflags);
2325
8800895a
PH
2326/* If there is no From: header, generate one for local (without
2327suppress_local_fixups) or submission_mode messages. If there is no sender
2328address, but the sender is local or this is a local delivery error, use the
2329originator login. This shouldn't happen for genuine bounces, but might happen
2330for autoreplies. The addition of From: must be done *before* checking for the
2331possible addition of a Sender: header, because untrusted_set_sender allows an
2332untrusted user to set anything in the envelope (which might then get info
2333From:) but we still want to ensure a valid Sender: if it is required. */
2334
2335if (from_header == NULL &&
2336 ((sender_host_address == NULL && !suppress_local_fixups)
2337 || submission_mode))
059ec3d9 2338 {
2fe1a124
PH
2339 uschar *oname = US"";
2340
2341 /* Use the originator_name if this is a locally submitted message and the
2342 caller is not trusted. For trusted callers, use it only if -F was used to
2343 force its value or if we have a non-SMTP message for which -f was not used
2344 to set the sender. */
2345
2346 if (sender_host_address == NULL)
2347 {
2348 if (!trusted_caller || sender_name_forced ||
2349 (!smtp_input && !sender_address_forced))
2350 oname = originator_name;
2351 }
2352
2353 /* For non-locally submitted messages, the only time we use the originator
2354 name is when it was forced by the /name= option on control=submission. */
2355
2356 else
2357 {
2358 if (submission_name != NULL) oname = submission_name;
2359 }
2360
059ec3d9
PH
2361 /* Envelope sender is empty */
2362
2363 if (sender_address[0] == 0)
2364 {
87ba3f5f
PH
2365 uschar *fromstart, *fromend;
2366
2367 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2368 oname, (oname[0] == 0)? "" : " <");
2369 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2370
059ec3d9
PH
2371 if (sender_local || local_error_message)
2372 {
87ba3f5f
PH
2373 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2374 local_part_quote(originator_login), qualify_domain_sender,
2375 fromend);
059ec3d9
PH
2376 }
2377 else if (submission_mode && authenticated_id != NULL)
2378 {
2379 if (submission_domain == NULL)
2380 {
87ba3f5f
PH
2381 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2382 local_part_quote(authenticated_id), qualify_domain_sender,
2383 fromend);
059ec3d9
PH
2384 }
2385 else if (submission_domain[0] == 0) /* empty => whole address set */
2386 {
87ba3f5f
PH
2387 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2388 fromend);
059ec3d9
PH
2389 }
2390 else
2391 {
87ba3f5f
PH
2392 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2393 local_part_quote(authenticated_id), submission_domain,
2394 fromend);
059ec3d9
PH
2395 }
2396 from_header = header_last; /* To get it checked for Sender: */
2397 }
2398 }
2399
2400 /* There is a non-null envelope sender. Build the header using the original
2401 sender address, before any rewriting that might have been done while
2402 verifying it. */
2403
2404 else
2405 {
87ba3f5f 2406 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2407 oname,
2408 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2409 (sender_address_unrewritten == NULL)?
2410 sender_address : sender_address_unrewritten,
2fe1a124 2411 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2412
2413 from_header = header_last; /* To get it checked for Sender: */
2414 }
2415 }
2416
2417
8800895a
PH
2418/* If the sender is local (without suppress_local_fixups), or if we are in
2419submission mode and there is an authenticated_id, check that an existing From:
2420is correct, and if not, generate a Sender: header, unless disabled. Any
2421previously-existing Sender: header was removed above. Note that sender_local,
2422as well as being TRUE if the caller of exim is not trusted, is also true if a
2423trusted caller did not supply a -f argument for non-smtp input. To allow
2424trusted callers to forge From: without supplying -f, we have to test explicitly
2425here. If the From: header contains more than one address, then the call to
2426parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2427
2428if (from_header != NULL &&
69358f02 2429 (active_local_from_check &&
8800895a 2430 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2431 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2432 ))
2433 {
2434 BOOL make_sender = TRUE;
2435 int start, end, domain;
2436 uschar *errmess;
2437 uschar *from_address =
2438 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2439 &start, &end, &domain, FALSE);
2440 uschar *generated_sender_address;
2441
2442 if (submission_mode)
2443 {
2444 if (submission_domain == NULL)
2445 {
2446 generated_sender_address = string_sprintf("%s@%s",
2447 local_part_quote(authenticated_id), qualify_domain_sender);
2448 }
2449 else if (submission_domain[0] == 0) /* empty => full address */
2450 {
2451 generated_sender_address = string_sprintf("%s",
2452 authenticated_id);
2453 }
2454 else
2455 {
2456 generated_sender_address = string_sprintf("%s@%s",
2457 local_part_quote(authenticated_id), submission_domain);
2458 }
2459 }
2460 else
2461 generated_sender_address = string_sprintf("%s@%s",
2462 local_part_quote(originator_login), qualify_domain_sender);
2463
2464 /* Remove permitted prefixes and suffixes from the local part of the From:
2465 address before doing the comparison with the generated sender. */
2466
2467 if (from_address != NULL)
2468 {
2469 int slen;
2470 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2471
2472 if (at != NULL) *at = 0;
2473 from_address += route_check_prefix(from_address, local_from_prefix);
2474 slen = route_check_suffix(from_address, local_from_suffix);
2475 if (slen > 0)
2476 {
2477 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2478 from_address += slen;
2479 }
2480 if (at != NULL) *at = '@';
2481
2482 if (strcmpic(generated_sender_address, from_address) == 0 ||
2483 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2484 make_sender = FALSE;
2485 }
2486
2487 /* We have to cause the Sender header to be rewritten if there are
2488 appropriate rewriting rules. */
2489
2490 if (make_sender)
2491 {
2fe1a124 2492 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2493 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2494 generated_sender_address);
2495 else
2496 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2497 resent_prefix,
2498 submission_mode? submission_name : originator_name,
2499 generated_sender_address);
059ec3d9 2500 }
87ba3f5f
PH
2501
2502 /* Ensure that a non-null envelope sender address corresponds to the
2503 submission mode sender address. */
2504
2505 if (submission_mode && sender_address[0] != 0)
2506 {
2507 if (sender_address_unrewritten == NULL)
2508 sender_address_unrewritten = sender_address;
2509 sender_address = generated_sender_address;
2510 log_write(L_address_rewrite, LOG_MAIN,
2511 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2512 sender_address_unrewritten, generated_sender_address);
2513 }
059ec3d9
PH
2514 }
2515
2516
2517/* If there are any rewriting rules, apply them to the sender address, unless
2518it has already been rewritten as part of verification for SMTP input. */
2519
2520if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2521 sender_address[0] != 0)
2522 {
2523 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2524 global_rewrite_rules, rewrite_existflags);
2525 DEBUG(D_receive|D_rewrite)
2526 debug_printf("rewritten sender = %s\n", sender_address);
2527 }
2528
2529
2530/* The headers must be run through rewrite_header(), because it ensures that
2531addresses are fully qualified, as well as applying any rewriting rules that may
2532exist.
2533
2534Qualification of header addresses in a message from a remote host happens only
2535if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2536appropriate. For local messages, qualification always happens, unless -bnq is
2537used to explicitly suppress it. No rewriting is done for an unqualified address
2538that is left untouched.
2539
2540We start at the second header, skipping our own Received:. This rewriting is
2541documented as happening *after* recipient addresses are taken from the headers
2542by the -t command line option. An added Sender: gets rewritten here. */
2543
2544for (h = header_list->next; h != NULL; h = h->next)
2545 {
2546 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2547 rewrite_existflags, TRUE);
2548 if (newh != NULL) h = newh;
2549 }
2550
2551
2552/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2553"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2554"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2555exists, this applies to the set of resent- headers rather than the normal set.
2556
2cbb4081
PH
2557The requirement for a recipient header has been removed in RFC 2822. At this
2558point in the code, earlier versions of Exim added a To: header for locally
2559submitted messages, and an empty Bcc: header for others. In the light of the
2560changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2561
059ec3d9
PH
2562
2563/* If there is no date header, generate one if the message originates locally
8800895a
PH
2564(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2565submission mode flag is set. Messages without Date: are not valid, but it seems
2566to be more confusing if Exim adds one to all remotely-originated messages. */
059ec3d9 2567
8800895a
PH
2568if (!date_header_exists &&
2569 ((sender_host_address == NULL && !suppress_local_fixups)
2570 || submission_mode))
059ec3d9
PH
2571 header_add(htype_other, "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
2572
2573search_tidyup(); /* Free any cached resources */
2574
2575/* Show the complete set of headers if debugging. Note that the first one (the
2576new Received:) has not yet been set. */
2577
2578DEBUG(D_receive)
2579 {
2580 debug_printf(">>Headers after rewriting and local additions:\n");
2581 for (h = header_list->next; h != NULL; h = h->next)
2582 debug_printf("%c %s", h->type, h->text);
2583 debug_printf("\n");
2584 }
2585
2586/* The headers are now complete in store. If we are running in filter
2587testing mode, that is all this function does. Return TRUE if the message
2588ended with a dot. */
2589
f05da2e8 2590if (filter_test != FTEST_NONE)
059ec3d9
PH
2591 {
2592 process_info[process_info_len] = 0;
2593 return message_ended == END_DOT;
2594 }
2595
2596/* Open a new spool file for the data portion of the message. We need
2597to access it both via a file descriptor and a stream. Try to make the
2598directory if it isn't there. Note re use of sprintf: spool_directory
2599is checked on input to be < 200 characters long. */
2600
2601sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory, message_subdir,
2602 message_id);
2603data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2604if (data_fd < 0)
2605 {
2606 if (errno == ENOENT)
2607 {
2608 uschar temp[16];
2609 sprintf(CS temp, "input/%s", message_subdir);
2610 if (message_subdir[0] == 0) temp[5] = 0;
2611 (void)directory_make(spool_directory, temp, INPUT_DIRECTORY_MODE, TRUE);
2612 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2613 }
2614 if (data_fd < 0)
2615 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
2616 spool_name, strerror(errno));
2617 }
2618
2619/* Make sure the file's group is the Exim gid, and double-check the mode
2620because the group setting doesn't always get set automatically. */
2621
ff790e47
PH
2622(void)fchown(data_fd, exim_uid, exim_gid);
2623(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
2624
2625/* We now have data file open. Build a stream for it and lock it. We lock only
2626the first line of the file (containing the message ID) because otherwise there
2627are problems when Exim is run under Cygwin (I'm told). See comments in
2628spool_in.c, where the same locking is done. */
2629
2630data_file = fdopen(data_fd, "w+");
2631lock_data.l_type = F_WRLCK;
2632lock_data.l_whence = SEEK_SET;
2633lock_data.l_start = 0;
2634lock_data.l_len = SPOOL_DATA_START_OFFSET;
2635
2636if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
2637 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
2638 errno, strerror(errno));
2639
2640/* We have an open, locked data file. Write the message id to it to make it
2641self-identifying. Then read the remainder of the input of this message and
2642write it to the data file. If the variable next != NULL, it contains the first
2643data line (which was read as a header but then turned out not to have the right
2644format); write it (remembering that it might contain binary zeros). The result
2645of fwrite() isn't inspected; instead we call ferror() below. */
2646
2647fprintf(data_file, "%s-D\n", message_id);
2648if (next != NULL)
2649 {
2650 uschar *s = next->text;
2651 int len = next->slen;
f1e894f3 2652 (void)fwrite(s, 1, len, data_file);
059ec3d9
PH
2653 body_linecount++; /* Assumes only 1 line */
2654 }
2655
2656/* Note that we might already be at end of file, or the logical end of file
2657(indicated by '.'), or might have encountered an error while writing the
2658message id or "next" line. */
2659
2660if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
2661 {
2662 if (smtp_input)
2663 {
2664 message_ended = read_message_data_smtp(data_file);
2665 receive_linecount++; /* The terminating "." line */
2666 }
2667 else message_ended = read_message_data(data_file);
2668
2669 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 2670 message_linecount += body_linecount;
059ec3d9
PH
2671
2672 /* Handle premature termination of SMTP */
2673
2674 if (smtp_input && message_ended == END_EOF)
2675 {
2676 Uunlink(spool_name); /* Lose data file when closed */
2677 message_id[0] = 0; /* Indicate no message accepted */
2678 smtp_reply = handle_lost_connection(US"");
2679 smtp_yield = FALSE;
2680 goto TIDYUP; /* Skip to end of function */
2681 }
2682
2683 /* Handle message that is too big. Don't use host_or_ident() in the log
2684 message; we want to see the ident value even for non-remote messages. */
2685
2686 if (message_ended == END_SIZE)
2687 {
2688 Uunlink(spool_name); /* Lose the data file when closed */
2689 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2690
2691 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2692 "message too big: read=%d max=%d",
2693 sender_address,
2694 (sender_fullhost == NULL)? "" : " H=",
2695 (sender_fullhost == NULL)? US"" : sender_fullhost,
2696 (sender_ident == NULL)? "" : " U=",
2697 (sender_ident == NULL)? US"" : sender_ident,
2698 message_size,
2699 thismessage_size_limit);
2700
2701 if (smtp_input)
2702 {
2703 smtp_reply = US"552 Message size exceeds maximum permitted";
2704 message_id[0] = 0; /* Indicate no message accepted */
2705 goto TIDYUP; /* Skip to end of function */
2706 }
2707 else
2708 {
2709 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2710 give_local_error(ERRMESS_TOOBIG,
2711 string_sprintf("message too big (max=%d)", thismessage_size_limit),
2712 US"message rejected: ", error_rc, data_file, header_list);
2713 /* Does not return */
2714 }
2715 }
2716 }
2717
2718/* Restore the standard SIGALRM handler for any subsequent processing. (For
2719example, there may be some expansion in an ACL that uses a timer.) */
2720
2721os_non_restarting_signal(SIGALRM, sigalrm_handler);
2722
2723/* The message body has now been read into the data file. Call fflush() to
2724empty the buffers in C, and then call fsync() to get the data written out onto
2725the disk, as fflush() doesn't do this (or at least, it isn't documented as
2726having to do this). If there was an I/O error on either input or output,
2727attempt to send an error message, and unlink the spool file. For non-SMTP input
2728we can then give up. Note that for SMTP input we must swallow the remainder of
2729the input in cases of output errors, since the far end doesn't expect to see
2730anything until the terminating dot line is sent. */
2731
2732if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 2733 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
2734 {
2735 uschar *msg_errno = US strerror(errno);
2736 BOOL input_error = (receive_ferror)() != 0;
2737 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
2738 input_error? "Input read" : "Spool write",
2739 msg_errno,
2740 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
2741
2742 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
2743 Uunlink(spool_name); /* Lose the data file */
2744
2745 if (smtp_input)
2746 {
2747 if (input_error)
2748 smtp_reply = US"451 Error while reading input data";
2749 else
2750 {
2751 smtp_reply = US"451 Error while writing spool file";
2752 receive_swallow_smtp();
2753 }
2754 message_id[0] = 0; /* Indicate no message accepted */
2755 goto TIDYUP; /* Skip to end of function */
2756 }
2757
2758 else
2759 {
2760 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2761 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
2762 header_list);
2763 /* Does not return */
2764 }
2765 }
2766
2767
2768/* No I/O errors were encountered while writing the data file. */
2769
2770DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
2771
2772
2773/* If there were any bad addresses extracted by -t, or there were no recipients
2774left after -t, send a message to the sender of this message, or write it to
2775stderr if the error handling option is set that way. Note that there may
2776legitimately be no recipients for an SMTP message if they have all been removed
2777by "discard".
2778
2779We need to rewind the data file in order to read it. In the case of no
2780recipients or stderr error writing, throw the data file away afterwards, and
2781exit. (This can't be SMTP, which always ensures there's at least one
2782syntactically good recipient address.) */
2783
2784if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
2785 {
2786 DEBUG(D_receive)
2787 {
2788 if (recipients_count == 0) debug_printf("*** No recipients\n");
2789 if (bad_addresses != NULL)
2790 {
2791 error_block *eblock = bad_addresses;
2792 debug_printf("*** Bad address(es)\n");
2793 while (eblock != NULL)
2794 {
2795 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
2796 eblock = eblock->next;
2797 }
2798 }
2799 }
2800
2801 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2802
2803 /* If configured to send errors to the sender, but this fails, force
2804 a failure error code. We use a special one for no recipients so that it
2805 can be detected by the autoreply transport. Otherwise error_rc is set to
2806 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
2807 it is EXIT_SUCCESS. */
2808
2809 if (error_handling == ERRORS_SENDER)
2810 {
2811 if (!moan_to_sender(
2812 (bad_addresses == NULL)?
2813 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
2814 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
2815 bad_addresses, header_list, data_file, FALSE))
2816 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
2817 }
2818 else
2819 {
2820 if (bad_addresses == NULL)
2821 {
2822 if (extracted_ignored)
2823 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
2824 else
2825 fprintf(stderr, "exim: no recipients in message\n");
2826 }
2827 else
2828 {
2829 fprintf(stderr, "exim: invalid address%s",
2830 (bad_addresses->next == NULL)? ":" : "es:\n");
2831 while (bad_addresses != NULL)
2832 {
2833 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
2834 bad_addresses->text2);
2835 bad_addresses = bad_addresses->next;
2836 }
2837 }
2838 }
2839
2840 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
2841 {
2842 Uunlink(spool_name);
f1e894f3 2843 (void)fclose(data_file);
059ec3d9
PH
2844 exim_exit(error_rc);
2845 }
2846 }
2847
2848/* Data file successfully written. Generate text for the Received: header by
2849expanding the configured string, and adding a timestamp. By leaving this
2850operation till now, we ensure that the timestamp is the time that message
2851reception was completed. However, this is deliberately done before calling the
2852data ACL and local_scan().
2853
2854This Received: header may therefore be inspected by the data ACL and by code in
2855the local_scan() function. When they have run, we update the timestamp to be
2856the final time of reception.
2857
2858If there is just one recipient, set up its value in the $received_for variable
2859for use when we generate the Received: header.
2860
2861Note: the checking for too many Received: headers is handled by the delivery
2862code. */
2863
2864timestamp = expand_string(US"${tod_full}");
2865if (recipients_count == 1) received_for = recipients_list[0].address;
2866received = expand_string(received_header_text);
2867received_for = NULL;
2868
2869if (received == NULL)
2870 {
2871 Uunlink(spool_name); /* Lose the data file */
2872 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
2873 "(received_header_text) failed: %s", string_printing(received_header_text),
2874 expand_string_message);
2875 }
2876
2877/* The first element on the header chain is reserved for the Received header,
2878so all we have to do is fill in the text pointer, and set the type. However, if
2879the result of the expansion is an empty string, we leave the header marked as
2880"old" so as to refrain from adding a Received header. */
2881
2882if (received[0] == 0)
2883 {
2884 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
2885 received_header->type = htype_old;
2886 }
2887else
2888 {
2889 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
2890 received_header->type = htype_received;
2891 }
2892
2893received_header->slen = Ustrlen(received_header->text);
2894
2895DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
2896 received_header->type, received_header->text);
2897
2898/* Set the value of message_body_size for the DATA ACL and for local_scan() */
2899
2900message_body_size = (fstat(data_fd, &statbuf) == 0)?
2901 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
2902
2903/* If an ACL from any RCPT commands set up any warning headers to add, do so
2904now, before running the DATA ACL. */
2905
2906add_acl_headers(US"MAIL or RCPT");
2907
2908/* If an ACL is specified for checking things at this stage of reception of a
2909message, run it, unless all the recipients were removed by "discard" in earlier
2910ACLs. That is the only case in which recipients_count can be zero at this
2911stage. Set deliver_datafile to point to the data file so that $message_body and
2912$message_body_end can be extracted if needed. Allow $recipients in expansions.
2913*/
2914
2915deliver_datafile = data_fd;
4e88a19f 2916user_msg = NULL;
059ec3d9 2917
0e20aff9
MH
2918enable_dollar_recipients = TRUE;
2919
059ec3d9
PH
2920if (recipients_count == 0)
2921 {
2922 blackholed_by = recipients_discarded? US"MAIL ACL" : US"RCPT ACL";
2923 }
2924else
2925 {
059ec3d9
PH
2926 /* Handle interactive SMTP messages */
2927
2928 if (smtp_input && !smtp_batched_input)
2929 {
8523533c 2930
fb2274d4
TK
2931#ifdef EXPERIMENTAL_DOMAINKEYS
2932 dk_exim_verify_finish();
2933#endif
2934
8523533c 2935#ifdef WITH_CONTENT_SCAN
54cdb463
PH
2936 if (acl_smtp_mime != NULL &&
2937 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
2938 goto TIDYUP;
8523533c
TK
2939#endif /* WITH_CONTENT_SCAN */
2940
54cdb463
PH
2941 /* Check the recipients count again, as the MIME ACL might have changed
2942 them. */
8523533c 2943
059ec3d9
PH
2944 if (acl_smtp_data != NULL && recipients_count > 0)
2945 {
059ec3d9
PH
2946 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
2947 add_acl_headers(US"DATA");
2948 if (rc == DISCARD)
2949 {
2950 recipients_count = 0;
2951 blackholed_by = US"DATA ACL";
8e669ac1
PH
2952 if (log_msg != NULL)
2953 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9
PH
2954 }
2955 else if (rc != OK)
2956 {
2957 Uunlink(spool_name);
8523533c
TK
2958#ifdef WITH_CONTENT_SCAN
2959 unspool_mbox();
2960#endif
059ec3d9
PH
2961 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
2962 smtp_yield = FALSE; /* No more messsages after dropped connection */
2963 smtp_reply = US""; /* Indicate reply already sent */
2964 message_id[0] = 0; /* Indicate no message accepted */
2965 goto TIDYUP; /* Skip to end of function */
2966 }
2967 }
2968 }
2969
2970 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
2971 we cannot take different actions for permanent and temporary rejections. */
2972
54cdb463 2973 else
059ec3d9 2974 {
54cdb463
PH
2975
2976#ifdef WITH_CONTENT_SCAN
2977 if (acl_not_smtp_mime != NULL &&
2978 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
2979 &blackholed_by))
2980 goto TIDYUP;
2981#endif /* WITH_CONTENT_SCAN */
2982
2983 if (acl_not_smtp != NULL)
059ec3d9 2984 {
54cdb463
PH
2985 uschar *user_msg, *log_msg;
2986 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
2987 if (rc == DISCARD)
059ec3d9 2988 {
54cdb463
PH
2989 recipients_count = 0;
2990 blackholed_by = US"non-SMTP ACL";
2991 if (log_msg != NULL)
2992 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 2993 }
54cdb463 2994 else if (rc != OK)
059ec3d9 2995 {
54cdb463
PH
2996 Uunlink(spool_name);
2997#ifdef WITH_CONTENT_SCAN
2998 unspool_mbox();
2999#endif
6ea85e9a
PH
3000 /* The ACL can specify where rejections are to be logged, possibly
3001 nowhere. The default is main and reject logs. */
3002
3003 if (log_reject_target != 0)
3004 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3005 sender_address, log_msg);
3006
54cdb463
PH
3007 if (user_msg == NULL) user_msg = US"local configuration problem";
3008 if (smtp_batched_input)
3009 {
3010 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3011 /* Does not return */
3012 }
3013 else
3014 {
3015 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3016 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3017 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3018 header_list);
3019 /* Does not return */
3020 }
059ec3d9 3021 }
54cdb463 3022 add_acl_headers(US"non-SMTP");
059ec3d9 3023 }
059ec3d9
PH
3024 }
3025
54cdb463
PH
3026 /* The applicable ACLs have been run */
3027
059ec3d9
PH
3028 if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */
3029 if (queue_only_policy) queued_by = US"ACL";
059ec3d9
PH
3030 }
3031
8523533c
TK
3032#ifdef WITH_CONTENT_SCAN
3033unspool_mbox();
3034#endif
3035
059ec3d9
PH
3036/* The final check on the message is to run the scan_local() function. The
3037version supplied with Exim always accepts, but this is a hook for sysadmins to
3038supply their own checking code. The local_scan() function is run even when all
3039the recipients have been discarded. */
3040
3041lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3042
3043/* Arrange to catch crashes in local_scan(), so that the -D file gets
3044deleted, and the incident gets logged. */
3045
3046os_non_restarting_signal(SIGSEGV, local_scan_crash_handler);
3047os_non_restarting_signal(SIGFPE, local_scan_crash_handler);
3048os_non_restarting_signal(SIGILL, local_scan_crash_handler);
3049os_non_restarting_signal(SIGBUS, local_scan_crash_handler);
3050
3051DEBUG(D_receive) debug_printf("calling local_scan(); timeout=%d\n",
3052 local_scan_timeout);
3053local_scan_data = NULL;
3054
3055os_non_restarting_signal(SIGALRM, local_scan_timeout_handler);
3056if (local_scan_timeout > 0) alarm(local_scan_timeout);
3057rc = local_scan(data_fd, &local_scan_data);
3058alarm(0);
3059os_non_restarting_signal(SIGALRM, sigalrm_handler);
3060
0e20aff9
MH
3061enable_dollar_recipients = FALSE;
3062
059ec3d9
PH
3063store_pool = POOL_MAIN; /* In case changed */
3064DEBUG(D_receive) debug_printf("local_scan() returned %d %s\n", rc,
3065 local_scan_data);
3066
3067os_non_restarting_signal(SIGSEGV, SIG_DFL);
3068os_non_restarting_signal(SIGFPE, SIG_DFL);
3069os_non_restarting_signal(SIGILL, SIG_DFL);
3070os_non_restarting_signal(SIGBUS, SIG_DFL);
3071
3072/* The length check is paranoia against some runaway code, and also because
3073(for a success return) lines in the spool file are read into big_buffer. */
3074
3075if (local_scan_data != NULL)
3076 {
3077 int len = Ustrlen(local_scan_data);
3078 if (len > LOCAL_SCAN_MAX_RETURN) len = LOCAL_SCAN_MAX_RETURN;
3079 local_scan_data = string_copyn(local_scan_data, len);
3080 }
3081
3082if (rc == LOCAL_SCAN_ACCEPT_FREEZE)
3083 {
58eb016e 3084 if (!deliver_freeze) /* ACL might have already frozen */
059ec3d9
PH
3085 {
3086 deliver_freeze = TRUE;
3087 deliver_frozen_at = time(NULL);
3088 frozen_by = US"local_scan()";
3089 }
3090 rc = LOCAL_SCAN_ACCEPT;
3091 }
3092else if (rc == LOCAL_SCAN_ACCEPT_QUEUE)
3093 {
3094 if (!queue_only_policy) /* ACL might have already queued */
3095 {
3096 queue_only_policy = TRUE;
3097 queued_by = US"local_scan()";
3098 }
3099 rc = LOCAL_SCAN_ACCEPT;
3100 }
3101
3102/* Message accepted: remove newlines in local_scan_data because otherwise
3103the spool file gets corrupted. Ensure that all recipients are qualified. */
3104
3105if (rc == LOCAL_SCAN_ACCEPT)
3106 {
3107 if (local_scan_data != NULL)
3108 {
3109 uschar *s;
3110 for (s = local_scan_data; *s != 0; s++) if (*s == '\n') *s = ' ';
3111 }
3112 for (i = 0; i < recipients_count; i++)
3113 {
3114 recipient_item *r = recipients_list + i;
3115 r->address = rewrite_address_qualify(r->address, TRUE);
3116 if (r->errors_to != NULL)
3117 r->errors_to = rewrite_address_qualify(r->errors_to, TRUE);
3118 }
3119 if (recipients_count == 0 && blackholed_by == NULL)
3120 blackholed_by = US"local_scan";
3121 }
3122
3123/* Message rejected: newlines permitted in local_scan_data to generate
3124multiline SMTP responses. */
3125
3126else
3127 {
3128 uschar *istemp = US"";
3129 uschar *s = NULL;
a5bd321b 3130 uschar *smtp_code;
059ec3d9
PH
3131 int size = 0;
3132 int sptr = 0;
059ec3d9
PH
3133
3134 errmsg = local_scan_data;
3135
3136 Uunlink(spool_name); /* Cancel this message */
3137 switch(rc)
3138 {
3139 default:
3140 log_write(0, LOG_MAIN, "invalid return %d from local_scan(). Temporary "
3141 "rejection given", rc);
3142 goto TEMPREJECT;
3143
3144 case LOCAL_SCAN_REJECT_NOLOGHDR:
3145 log_extra_selector &= ~LX_rejected_header;
3146 /* Fall through */
3147
3148 case LOCAL_SCAN_REJECT:
a5bd321b 3149 smtp_code = US"550";
059ec3d9
PH
3150 if (errmsg == NULL) errmsg = US"Administrative prohibition";
3151 break;
3152
3153 case LOCAL_SCAN_TEMPREJECT_NOLOGHDR:
3154 log_extra_selector &= ~LX_rejected_header;
3155 /* Fall through */
3156
3157 case LOCAL_SCAN_TEMPREJECT:
3158 TEMPREJECT:
a5bd321b 3159 smtp_code = US"451";
059ec3d9
PH
3160 if (errmsg == NULL) errmsg = US"Temporary local problem";
3161 istemp = US"temporarily ";
3162 break;
3163 }
3164
3165 s = string_append(s, &size, &sptr, 2, US"F=",
3166 (sender_address[0] == 0)? US"<>" : sender_address);
3167 s = add_host_info_for_log(s, &size, &sptr);
3168 s[sptr] = 0;
3169
3170 log_write(0, LOG_MAIN|LOG_REJECT, "%s %srejected by local_scan(): %.256s",
3171 s, istemp, string_printing(errmsg));
3172
3173 if (smtp_input)
3174 {
3175 if (!smtp_batched_input)
3176 {
a5bd321b 3177 smtp_respond(smtp_code, 3, TRUE, errmsg);
059ec3d9
PH
3178 message_id[0] = 0; /* Indicate no message accepted */
3179 smtp_reply = US""; /* Indicate reply already sent */
3180 goto TIDYUP; /* Skip to end of function */
3181 }
3182 else
3183 {
a5bd321b 3184 moan_smtp_batch(NULL, "%s %s", smtp_code, errmsg);
059ec3d9
PH
3185 /* Does not return */
3186 }
3187 }
3188 else
3189 {
3190 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3191 give_local_error(ERRMESS_LOCAL_SCAN, errmsg,
3192 US"message rejected by local scan code: ", error_rc, data_file,
3193 header_list);
3194 /* Does not return */
3195 }
3196 }
3197
3198/* Reset signal handlers to ignore signals that previously would have caused
3199the message to be abandoned. */
3200
3201signal(SIGTERM, SIG_IGN);
3202signal(SIGINT, SIG_IGN);
3203
3204/* Ensure the first time flag is set in the newly-received message. */
3205
3206deliver_firsttime = TRUE;
3207
8523533c
TK
3208#ifdef EXPERIMENTAL_BRIGHTMAIL
3209if (bmi_run == 1) {
3210 /* rewind data file */
3211 lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3212 bmi_verdicts = bmi_process_message(header_list, data_fd);
3213};
3214#endif
3215
059ec3d9
PH
3216/* Update the timstamp in our Received: header to account for any time taken by
3217an ACL or by local_scan(). The new time is the time that all reception
3218processing is complete. */
3219
3220timestamp = expand_string(US"${tod_full}");
3221tslen = Ustrlen(timestamp);
3222
3223memcpy(received_header->text + received_header->slen - tslen - 1,
3224 timestamp, tslen);
3225
3226/* In MUA wrapper mode, ignore queueing actions set by ACL or local_scan() */
3227
3228if (mua_wrapper)
3229 {
3230 deliver_freeze = FALSE;
3231 queue_only_policy = FALSE;
3232 }
3233
3234/* Keep the data file open until we have written the header file, in order to
3235hold onto the lock. In a -bh run, or if the message is to be blackholed, we
3236don't write the header file, and we unlink the data file. If writing the header
3237file fails, we have failed to accept this message. */
3238
3239if (host_checking || blackholed_by != NULL)
3240 {
3241 header_line *h;
3242 Uunlink(spool_name);
3243 msg_size = 0; /* Compute size for log line */
3244 for (h = header_list; h != NULL; h = h->next)
3245 if (h->type != '*') msg_size += h->slen;
3246 }
3247
3248/* Write the -H file */
3249
3250else
3251 {
3252 if ((msg_size = spool_write_header(message_id, SW_RECEIVING, &errmsg)) < 0)
3253 {
3254 log_write(0, LOG_MAIN, "Message abandoned: %s", errmsg);
3255 Uunlink(spool_name); /* Lose the data file */
3256
3257 if (smtp_input)
3258 {
3259 smtp_reply = US"451 Error in writing spool file";
3260 message_id[0] = 0; /* Indicate no message accepted */
3261 goto TIDYUP;
3262 }
3263 else
3264 {
3265 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3266 give_local_error(ERRMESS_IOERR, errmsg, US"", error_rc, data_file,
3267 header_list);
3268 /* Does not return */
3269 }
3270 }
3271 }
3272
3273
3274/* The message has now been successfully received. */
3275
3276receive_messagecount++;
3277
3278/* In SMTP sessions we may receive several in one connection. After each one,
3279we wait for the clock to tick at the level of message-id granularity. This is
3280so that the combination of time+pid is unique, even on systems where the pid
3281can be re-used within our time interval. We can't shorten the interval without
3282re-designing the message-id. See comments above where the message id is
3283created. This is Something For The Future. */
3284
3285message_id_tv.tv_usec = (message_id_tv.tv_usec/id_resolution) * id_resolution;
3286exim_wait_tick(&message_id_tv, id_resolution);
3287
3288/* Add data size to written header size. We do not count the initial file name
3289that is in the file, but we do add one extra for the notional blank line that
3290precedes the data. This total differs from message_size in that it include the
3291added Received: header and any other headers that got created locally. */
3292
3293fflush(data_file);
3294fstat(data_fd, &statbuf);
3295
3296msg_size += statbuf.st_size - SPOOL_DATA_START_OFFSET + 1;
3297
3298/* Generate a "message received" log entry. We do this by building up a dynamic
3299string as required. Since we commonly want to add two items at a time, use a
3300macro to simplify the coding. We log the arrival of a new message while the
3301file is still locked, just in case the machine is *really* fast, and delivers
3302it first! Include any message id that is in the message - since the syntax of a
3303message id is actually an addr-spec, we can use the parse routine to canonicize
3304it. */
3305
3306size = 256;
3307sptr = 0;
3308s = store_get(size);
3309
3310s = string_append(s, &size, &sptr, 2, US"<= ",
3311 (sender_address[0] == 0)? US"<>" : sender_address);
3312if (message_reference != NULL)
3313 s = string_append(s, &size, &sptr, 2, US" R=", message_reference);
3314
3315s = add_host_info_for_log(s, &size, &sptr);
3316
3317#ifdef SUPPORT_TLS
3318if ((log_extra_selector & LX_tls_cipher) != 0 && tls_cipher != NULL)
3319 s = string_append(s, &size, &sptr, 2, US" X=", tls_cipher);
3320if ((log_extra_selector & LX_tls_certificate_verified) != 0 &&
3321 tls_cipher != NULL)
3322 s = string_append(s, &size, &sptr, 2, US" CV=",
3323 tls_certificate_verified? "yes":"no");
3324if ((log_extra_selector & LX_tls_peerdn) != 0 && tls_peerdn != NULL)
3325 s = string_append(s, &size, &sptr, 3, US" DN=\"", tls_peerdn, US"\"");
3326#endif
3327
3328if (sender_host_authenticated != NULL)
3329 {
3330 s = string_append(s, &size, &sptr, 2, US" A=", sender_host_authenticated);
3331 if (authenticated_id != NULL)
3332 s = string_append(s, &size, &sptr, 2, US":", authenticated_id);
3333 }
3334
3335sprintf(CS big_buffer, "%d", msg_size);
3336s = string_append(s, &size, &sptr, 2, US" S=", big_buffer);
3337
3338/* If an addr-spec in a message-id contains a quoted string, it can contain
3339any characters except " \ and CR and so in particular it can contain NL!
3340Therefore, make sure we use a printing-characters only version for the log.
3341Also, allow for domain literals in the message id. */
3342
3343if (msgid_header != NULL)
3344 {
3345 uschar *old_id;
3346 BOOL save_allow_domain_literals = allow_domain_literals;
3347 allow_domain_literals = TRUE;
3348 old_id = parse_extract_address(Ustrchr(msgid_header->text, ':') + 1,
3349 &errmsg, &start, &end, &domain, FALSE);
3350 allow_domain_literals = save_allow_domain_literals;
3351 if (old_id != NULL)
3352 s = string_append(s, &size, &sptr, 2, US" id=", string_printing(old_id));
3353 }
3354
3355/* If subject logging is turned on, create suitable printing-character
3356text. By expanding $h_subject: we make use of the MIME decoding. */
3357
3358if ((log_extra_selector & LX_subject) != 0 && subject_header != NULL)
3359 {
3360 int i;
3361 uschar *p = big_buffer;
3362 uschar *ss = expand_string(US"$h_subject:");
3363
3364 /* Backslash-quote any double quotes or backslashes so as to make a
3365 a C-like string, and turn any non-printers into escape sequences. */
3366
3367 *p++ = '\"';
3368 if (*ss != 0) for (i = 0; i < 100 && ss[i] != 0; i++)
3369 {
3370 if (ss[i] == '\"' || ss[i] == '\\') *p++ = '\\';
3371 *p++ = ss[i];
3372 }
3373 *p++ = '\"';
3374 *p = 0;
3375 s = string_append(s, &size, &sptr, 2, US" T=", string_printing(big_buffer));
3376 }
3377
3378/* Terminate the string: string_cat() and string_append() leave room, but do
3379not put the zero in. */
3380
3381s[sptr] = 0;
3382
059ec3d9
PH
3383/* Create a message log file if message logs are being used and this message is
3384not blackholed. Write the reception stuff to it. We used to leave message log
3385creation until the first delivery, but this has proved confusing for somep
3386people. */
3387
3388if (message_logs && blackholed_by == NULL)
3389 {
3390 int fd;
3391
3392 sprintf(CS spool_name, "%s/msglog/%s/%s", spool_directory, message_subdir,
3393 message_id);
3394 fd = Uopen(spool_name, O_WRONLY|O_APPEND|O_CREAT, SPOOL_MODE);
3395
3396 if (fd < 0 && errno == ENOENT)
3397 {
3398 uschar temp[16];
3399 sprintf(CS temp, "msglog/%s", message_subdir);
3400 if (message_subdir[0] == 0) temp[6] = 0;
3401 (void)directory_make(spool_directory, temp, MSGLOG_DIRECTORY_MODE, TRUE);
3402 fd = Uopen(spool_name, O_WRONLY|O_APPEND|O_CREAT, SPOOL_MODE);
3403 }
3404
3405 if (fd < 0)
3406 {
3407 log_write(0, LOG_MAIN|LOG_PANIC, "Couldn't open message log %s: %s",
3408 spool_name, strerror(errno));
3409 }
3410
3411 else
3412 {
3413 FILE *message_log = fdopen(fd, "a");
3414 if (message_log == NULL)
3415 {
3416 log_write(0, LOG_MAIN|LOG_PANIC, "Couldn't fdopen message log %s: %s",
3417 spool_name, strerror(errno));
f1e894f3 3418 (void)close(fd);
059ec3d9
PH
3419 }
3420 else
3421 {
3422 uschar *now = tod_stamp(tod_log);
3423 fprintf(message_log, "%s Received from %s\n", now, s+3);
3424 if (deliver_freeze) fprintf(message_log, "%s frozen by %s\n", now,
3425 frozen_by);
3426 if (queue_only_policy) fprintf(message_log,
3427 "%s no immediate delivery: queued by %s\n", now, queued_by);
f1e894f3 3428 (void)fclose(message_log);
059ec3d9
PH
3429 }
3430 }
3431 }
3432
58eb016e
PH
3433/* Everything has now been done for a successful message except logging its
3434arrival, and outputting an SMTP response. While writing to the log, set a flag
3435to cause a call to receive_bomb_out() if the log cannot be opened. */
3436
3437receive_call_bombout = TRUE;
3438
563b63fa
PH
3439/* Before sending an SMTP response in a TCP/IP session, we check to see if the
3440connection has gone away. This can only be done if there is no unconsumed input
3441waiting in the local input buffer. We can test for this by calling
3442receive_smtp_buffered(). RFC 2920 (pipelining) explicitly allows for additional
3443input to be sent following the final dot, so the presence of following input is
3444not an error.
58eb016e 3445
563b63fa
PH
3446If the connection is still present, but there is no unread input for the
3447socket, the result of a select() call will be zero. If, however, the connection
3448has gone away, or if there is pending input, the result of select() will be
3449non-zero. The two cases can be distinguished by trying to read the next input
3450character. If we succeed, we can unread it so that it remains in the local
3451buffer for handling later. If not, the connection has been lost.
58eb016e 3452
563b63fa
PH
3453Of course, since TCP/IP is asynchronous, there is always a chance that the
3454connection will vanish between the time of this test and the sending of the
3455response, but the chance of this happening should be small. */
3456
3457if (smtp_input && sender_host_address != NULL && !sender_host_notsocket &&
3458 !receive_smtp_buffered())
58eb016e
PH
3459 {
3460 struct timeval tv;
3461 fd_set select_check;
3462 FD_ZERO(&select_check);
3463 FD_SET(fileno(smtp_in), &select_check);
3464 tv.tv_sec = 0;
3465 tv.tv_usec = 0;
3466
563b63fa 3467 if (select(fileno(smtp_in) + 1, &select_check, NULL, NULL, &tv) != 0)
58eb016e 3468 {
563b63fa
PH
3469 int c = (RECEIVE_GETC)();
3470 if (c != EOF) (RECEIVE_UNGETC)(c); else
58eb016e 3471 {
563b63fa
PH
3472 uschar *msg = US"SMTP connection lost after final dot";
3473 smtp_reply = US""; /* No attempt to send a response */
3474 smtp_yield = FALSE; /* Nothing more on this connection */
58eb016e 3475
563b63fa 3476 /* Re-use the log line workspace */
58eb016e 3477
563b63fa
PH
3478 sptr = 0;
3479 s = string_cat(s, &size, &sptr, msg, Ustrlen(msg));
3480 s = add_host_info_for_log(s, &size, &sptr);
3481 s[sptr] = 0;
3482 log_write(0, LOG_MAIN, "%s", s);
58eb016e 3483
563b63fa 3484 /* Delete the files for this aborted message. */
58eb016e 3485
563b63fa
PH
3486 sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory,
3487 message_subdir, message_id);
3488 Uunlink(spool_name);
58eb016e 3489
563b63fa
PH
3490 sprintf(CS spool_name, "%s/input/%s/%s-H", spool_directory,
3491 message_subdir, message_id);
3492 Uunlink(spool_name);
58eb016e 3493
563b63fa
PH
3494 sprintf(CS spool_name, "%s/msglog/%s/%s", spool_directory,
3495 message_subdir, message_id);
3496 Uunlink(spool_name);
58eb016e 3497
563b63fa
PH
3498 goto TIDYUP;
3499 }
58eb016e
PH
3500 }
3501 }
3502
3503/* The connection has not gone away; we really are going to take responsibility
3504for this message. */
3505
3506log_write(0, LOG_MAIN |
3507 (((log_extra_selector & LX_received_recipients) != 0)? LOG_RECIPIENTS : 0) |
3508 (((log_extra_selector & LX_received_sender) != 0)? LOG_SENDER : 0),
3509 "%s", s);
3510receive_call_bombout = FALSE;
3511
3512/* Log any control actions taken by an ACL or local_scan(). */
3513
3514if (deliver_freeze) log_write(0, LOG_MAIN, "frozen by %s", frozen_by);
3515if (queue_only_policy) log_write(L_delay_delivery, LOG_MAIN,
3516 "no immediate delivery: queued by %s", queued_by);
3517
059ec3d9
PH
3518store_reset(s); /* The store for the main log message can be reused */
3519
3520/* If the message is frozen, and freeze_tell is set, do the telling. */
3521
3522if (deliver_freeze && freeze_tell != NULL && freeze_tell[0] != 0)
3523 {
3524 moan_tell_someone(freeze_tell, NULL, US"Message frozen on arrival",
3525 "Message %s was frozen on arrival by %s.\nThe sender is <%s>.\n",
3526 message_id, frozen_by, sender_address);
3527 }
3528
3529
3530/* Either a message has been successfully received and written to the two spool
3531files, or an error in writing the spool has occurred for an SMTP message, or
58eb016e
PH
3532an SMTP message has been rejected for policy reasons. (For a non-SMTP message
3533we will have already given up because there's no point in carrying on!) In
3534either event, we must now close (and thereby unlock) the data file. In the
059ec3d9
PH
3535successful case, this leaves the message on the spool, ready for delivery. In
3536the error case, the spool file will be deleted. Then tidy up store, interact
3537with an SMTP call if necessary, and return.
3538
3539A fflush() was done earlier in the expectation that any write errors on the
3540data file will be flushed(!) out thereby. Nevertheless, it is theoretically
3541possible for fclose() to fail - but what to do? What has happened to the lock
3542if this happens? */
3543
3544TIDYUP:
f1e894f3
PH
3545process_info[process_info_len] = 0; /* Remove message id */
3546if (data_file != NULL) (void)fclose(data_file); /* Frees the lock */
059ec3d9
PH
3547
3548/* Now reset signal handlers to their defaults */
3549
3550signal(SIGTERM, SIG_DFL);
3551signal(SIGINT, SIG_DFL);
3552
3553/* Tell an SMTP caller the state of play, and arrange to return the SMTP return
3554value, which defaults TRUE - meaning there may be more incoming messages from
3555this connection. For non-SMTP callers (where there is only ever one message),
3556the default is FALSE. */
3557
3558if (smtp_input)
3559 {
3560 yield = smtp_yield;
3561
3562 /* Handle interactive SMTP callers. After several kinds of error, smtp_reply
58eb016e
PH
3563 is set to the response that should be sent. When it is NULL, we generate
3564 default responses. After an ACL error or local_scan() error, the response has
3565 already been sent, and smtp_reply is an empty string to indicate this. */
059ec3d9
PH
3566
3567 if (!smtp_batched_input)
3568 {
3569 if (smtp_reply == NULL)
3570 {
29aba418 3571 if (fake_response != OK)
a5bd321b
PH
3572 smtp_respond((fake_response == DEFER)? US"450" : US"550", 3, TRUE,
3573 fake_response_text);
4e88a19f
PH
3574
3575 /* An OK response is required; use "message" text if present. */
3576
3577 else if (user_msg != NULL)
3578 {
3579 uschar *code = US"250";
3580 int len = 3;
3581 smtp_message_code(&code, &len, &user_msg, NULL);
3582 smtp_respond(code, len, TRUE, user_msg);
3583 }
3584
3585 /* Default OK response */
3586
8e669ac1
PH
3587 else
3588 smtp_printf("250 OK id=%s\r\n", message_id);
059ec3d9
PH
3589 if (host_checking)
3590 fprintf(stdout,
3591 "\n**** SMTP testing: that is not a real message id!\n\n");
3592 }
4e88a19f 3593
58eb016e 3594 /* smtp_reply is set non-empty */
4e88a19f 3595
8523533c
TK
3596 else if (smtp_reply[0] != 0)
3597 {
29aba418 3598 if (fake_response != OK && (smtp_reply[0] == '2'))
a5bd321b
PH
3599 smtp_respond((fake_response == DEFER)? US"450" : US"550", 3, TRUE,
3600 fake_response_text);
8e669ac1 3601 else
6951ac6c
PH
3602 smtp_printf("%.1024s\r\n", smtp_reply);
3603 }
059ec3d9
PH
3604 }
3605
3606 /* For batched SMTP, generate an error message on failure, and do
3607 nothing on success. The function moan_smtp_batch() does not return -
3608 it exits from the program with a non-zero return code. */
3609
3610 else if (smtp_reply != NULL) moan_smtp_batch(NULL, "%s", smtp_reply);
3611 }
3612
3613
3614/* If blackholing, we can immediately log this message's sad fate. The data
3615file has already been unlinked, and the header file was never written to disk.
3616We must now indicate that nothing was received, to prevent a delivery from
3617starting. */
3618
3619if (blackholed_by != NULL)
3620 {
3621 uschar *detail = (local_scan_data != NULL)?
3622 string_printing(local_scan_data) :
3623 string_sprintf("(%s discarded recipients)", blackholed_by);
04f7d5b9 3624 log_write(0, LOG_MAIN, "=> blackhole %s%s", detail, blackhole_log_msg);
059ec3d9
PH
3625 log_write(0, LOG_MAIN, "Completed");
3626 message_id[0] = 0;
3627 }
3628
3629/* Reset headers so that logging of rejects for a subsequent message doesn't
3630include them. It is also important to set header_last = NULL before exiting
3631from this function, as this prevents certain rewrites that might happen during
3632subsequent verifying (of another incoming message) from trying to add headers
3633when they shouldn't. */
3634
3635header_list = header_last = NULL;
3636
3637return yield; /* TRUE if more messages (SMTP only) */
3638}
3639
3640/* End of receive.c */