Test for proper parsing of optional MAIL FROM args.
[exim.git] / src / src / receive.c
CommitLineData
059ec3d9
PH
1/*************************************************
2* Exim - an Internet mail transport agent *
3*************************************************/
4
c4ceed07 5/* Copyright (c) University of Cambridge 1995 - 2012 */
059ec3d9
PH
6/* See the file NOTICE for conditions of use and distribution. */
7
8/* Code for receiving a message and setting up spool files. */
9
059ec3d9
PH
10#include "exim.h"
11
6a8f9482
TK
12#ifdef EXPERIMENTAL_DCC
13extern int dcc_ok;
14#endif
15
059ec3d9
PH
16/*************************************************
17* Local static variables *
18*************************************************/
19
20static FILE *data_file = NULL;
21static int data_fd = -1;
22static uschar spool_name[256];
23
24
25
26/*************************************************
27* Non-SMTP character reading functions *
28*************************************************/
29
30/* These are the default functions that are set up in the variables such as
31receive_getc initially. They just call the standard functions, passing stdin as
32the file. (When SMTP input is occurring, different functions are used by
33changing the pointer variables.) */
34
35int
36stdin_getc(void)
37{
38return getc(stdin);
39}
40
41int
42stdin_ungetc(int c)
43{
44return ungetc(c, stdin);
45}
46
47int
48stdin_feof(void)
49{
50return feof(stdin);
51}
52
53int
54stdin_ferror(void)
55{
56return ferror(stdin);
57}
58
59
60
61
62/*************************************************
63* Check that a set sender is allowed *
64*************************************************/
65
66/* This function is called when a local caller sets an explicit sender address.
67It checks whether this is permitted, which it is for trusted callers.
68Otherwise, it must match the pattern(s) in untrusted_set_sender.
69
70Arguments: the proposed sender address
71Returns: TRUE for a trusted caller
72 TRUE if the address has been set, untrusted_set_sender has been
73 set, and the address matches something in the list
74 FALSE otherwise
75*/
76
77BOOL
78receive_check_set_sender(uschar *newsender)
79{
80uschar *qnewsender;
81if (trusted_caller) return TRUE;
82if (newsender == NULL || untrusted_set_sender == NULL) return FALSE;
83qnewsender = (Ustrchr(newsender, '@') != NULL)?
84 newsender : string_sprintf("%s@%s", newsender, qualify_domain_sender);
85return
86 match_address_list(qnewsender, TRUE, TRUE, &untrusted_set_sender, NULL, -1,
87 0, NULL) == OK;
88}
89
90
91
92
93/*************************************************
5cb8cbc6 94* Read space info for a partition *
059ec3d9
PH
95*************************************************/
96
8e669ac1
PH
97/* This function is called by receive_check_fs() below, and also by string
98expansion for variables such as $spool_space. The field names for the statvfs
5cb8cbc6
PH
99structure are macros, because not all OS have F_FAVAIL and it seems tidier to
100have macros for F_BAVAIL and F_FILES as well. Some kinds of file system do not
101have inodes, and they return -1 for the number available.
059ec3d9 102
5cb8cbc6
PH
103Later: It turns out that some file systems that do not have the concept of
104inodes return 0 rather than -1. Such systems should also return 0 for the total
8e669ac1 105number of inodes, so we require that to be greater than zero before returning
5cb8cbc6 106an inode count.
059ec3d9 107
5cb8cbc6
PH
108Arguments:
109 isspool TRUE for spool partition, FALSE for log partition
110 inodeptr address of int to receive inode count; -1 if there isn't one
8e669ac1 111
5cb8cbc6 112Returns: available on-root space, in kilobytes
8e669ac1
PH
113 -1 for log partition if there isn't one
114
115All values are -1 if the STATFS functions are not available.
059ec3d9
PH
116*/
117
8e669ac1 118int
5cb8cbc6 119receive_statvfs(BOOL isspool, int *inodeptr)
059ec3d9
PH
120{
121#ifdef HAVE_STATFS
059ec3d9 122struct STATVFS statbuf;
5cb8cbc6
PH
123uschar *path;
124uschar *name;
125uschar buffer[1024];
059ec3d9 126
5cb8cbc6 127/* The spool directory must always exist. */
059ec3d9 128
5cb8cbc6 129if (isspool)
059ec3d9 130 {
8e669ac1
PH
131 path = spool_directory;
132 name = US"spool";
133 }
134
059ec3d9
PH
135/* Need to cut down the log file path to the directory, and to ignore any
136appearance of "syslog" in it. */
137
5cb8cbc6 138else
059ec3d9 139 {
059ec3d9 140 int sep = ':'; /* Not variable - outside scripts use */
059ec3d9 141 uschar *p = log_file_path;
8e669ac1 142 name = US"log";
059ec3d9
PH
143
144 /* An empty log_file_path means "use the default". This is the same as an
145 empty item in a list. */
146
147 if (*p == 0) p = US":";
148 while ((path = string_nextinlist(&p, &sep, buffer, sizeof(buffer))) != NULL)
149 {
150 if (Ustrcmp(path, "syslog") != 0) break;
151 }
152
5cb8cbc6
PH
153 if (path == NULL) /* No log files */
154 {
8e669ac1
PH
155 *inodeptr = -1;
156 return -1;
157 }
059ec3d9 158
8e669ac1
PH
159 /* An empty string means use the default, which is in the spool directory.
160 But don't just use the spool directory, as it is possible that the log
5cb8cbc6 161 subdirectory has been symbolically linked elsewhere. */
059ec3d9 162
8e669ac1 163 if (path[0] == 0)
059ec3d9 164 {
5cb8cbc6
PH
165 sprintf(CS buffer, CS"%s/log", CS spool_directory);
166 path = buffer;
8e669ac1
PH
167 }
168 else
059ec3d9 169 {
8e669ac1 170 uschar *cp;
5cb8cbc6 171 if ((cp = Ustrrchr(path, '/')) != NULL) *cp = 0;
8e669ac1 172 }
5cb8cbc6 173 }
8e669ac1 174
8f128379 175/* We now have the path; do the business */
5cb8cbc6
PH
176
177memset(&statbuf, 0, sizeof(statbuf));
178
179if (STATVFS(CS path, &statbuf) != 0)
180 {
181 log_write(0, LOG_MAIN|LOG_PANIC, "cannot accept message: failed to stat "
182 "%s directory %s: %s", name, spool_directory, strerror(errno));
183 smtp_closedown(US"spool or log directory problem");
184 exim_exit(EXIT_FAILURE);
185 }
8e669ac1 186
5cb8cbc6
PH
187*inodeptr = (statbuf.F_FILES > 0)? statbuf.F_FAVAIL : -1;
188
189/* Disks are getting huge. Take care with computing the size in kilobytes. */
8e669ac1 190
5cb8cbc6
PH
191return (int)(((double)statbuf.F_BAVAIL * (double)statbuf.F_FRSIZE)/1024.0);
192
193/* Unable to find partition sizes in this environment. */
194
195#else
196*inodeptr = -1;
197return -1;
198#endif
199}
200
059ec3d9 201
059ec3d9 202
5cb8cbc6
PH
203
204/*************************************************
205* Check space on spool and log partitions *
206*************************************************/
207
208/* This function is called before accepting a message; if any thresholds are
209set, it checks them. If a message_size is supplied, it checks that there is
210enough space for that size plus the threshold - i.e. that the message won't
211reduce the space to the threshold. Not all OS have statvfs(); for those that
212don't, this function always returns TRUE. For some OS the old function and
213struct name statfs is used; that is handled by a macro, defined in exim.h.
214
215Arguments:
216 msg_size the (estimated) size of an incoming message
217
218Returns: FALSE if there isn't enough space, or if the information cannot
219 be obtained
220 TRUE if no check was done or there is enough space
221*/
222
223BOOL
224receive_check_fs(int msg_size)
225{
226int space, inodes;
227
228if (check_spool_space > 0 || msg_size > 0 || check_spool_inodes > 0)
229 {
8e669ac1
PH
230 space = receive_statvfs(TRUE, &inodes);
231
059ec3d9 232 DEBUG(D_receive)
5cb8cbc6
PH
233 debug_printf("spool directory space = %dK inodes = %d "
234 "check_space = %dK inodes = %d msg_size = %d\n",
235 space, inodes, check_spool_space, check_spool_inodes, msg_size);
8e669ac1
PH
236
237 if ((space >= 0 && space < check_spool_space) ||
5cb8cbc6 238 (inodes >= 0 && inodes < check_spool_inodes))
8e669ac1 239 {
5cb8cbc6
PH
240 log_write(0, LOG_MAIN, "spool directory space check failed: space=%d "
241 "inodes=%d", space, inodes);
059ec3d9
PH
242 return FALSE;
243 }
244 }
245
5cb8cbc6
PH
246if (check_log_space > 0 || check_log_inodes > 0)
247 {
8e669ac1
PH
248 space = receive_statvfs(FALSE, &inodes);
249
5cb8cbc6
PH
250 DEBUG(D_receive)
251 debug_printf("log directory space = %dK inodes = %d "
252 "check_space = %dK inodes = %d\n",
253 space, inodes, check_log_space, check_log_inodes);
8e669ac1
PH
254
255 if ((space >= 0 && space < check_log_space) ||
5cb8cbc6 256 (inodes >= 0 && inodes < check_log_inodes))
8e669ac1 257 {
5cb8cbc6
PH
258 log_write(0, LOG_MAIN, "log directory space check failed: space=%d "
259 "inodes=%d", space, inodes);
260 return FALSE;
261 }
8e669ac1
PH
262 }
263
059ec3d9
PH
264return TRUE;
265}
266
267
268
269/*************************************************
270* Bomb out while reading a message *
271*************************************************/
272
273/* The common case of wanting to bomb out is if a SIGTERM or SIGINT is
274received, or if there is a timeout. A rarer case might be if the log files are
275screwed up and Exim can't open them to record a message's arrival. Handling
276that case is done by setting a flag to cause the log functions to call this
277function if there is an ultimate disaster. That is why it is globally
278accessible.
279
8f128379
PH
280Arguments:
281 reason text reason to pass to the not-quit ACL
282 msg default SMTP response to give if in an SMTP session
059ec3d9
PH
283Returns: it doesn't
284*/
285
286void
8f128379 287receive_bomb_out(uschar *reason, uschar *msg)
059ec3d9
PH
288{
289/* If spool_name is set, it contains the name of the data file that is being
290written. Unlink it before closing so that it cannot be picked up by a delivery
291process. Ensure that any header file is also removed. */
292
293if (spool_name[0] != 0)
294 {
295 Uunlink(spool_name);
296 spool_name[Ustrlen(spool_name) - 1] = 'H';
297 Uunlink(spool_name);
298 }
299
300/* Now close the file if it is open, either as a fd or a stream. */
301
f1e894f3
PH
302if (data_file != NULL) (void)fclose(data_file);
303 else if (data_fd >= 0) (void)close(data_fd);
059ec3d9 304
8f128379
PH
305/* Attempt to close down an SMTP connection tidily. For non-batched SMTP, call
306smtp_notquit_exit(), which runs the NOTQUIT ACL, if present, and handles the
307SMTP response. */
059ec3d9
PH
308
309if (smtp_input)
310 {
8f128379
PH
311 if (smtp_batched_input)
312 moan_smtp_batch(NULL, "421 %s - message abandoned", msg); /* No return */
313 smtp_notquit_exit(reason, US"421", US"%s %s - closing connection.",
314 smtp_active_hostname, msg);
059ec3d9
PH
315 }
316
317/* Exit from the program (non-BSMTP cases) */
318
319exim_exit(EXIT_FAILURE);
320}
321
322
323/*************************************************
324* Data read timeout *
325*************************************************/
326
327/* Handler function for timeouts that occur while reading the data that
328comprises a message.
329
330Argument: the signal number
331Returns: nothing
332*/
333
334static void
335data_timeout_handler(int sig)
336{
337uschar *msg = NULL;
338
339sig = sig; /* Keep picky compilers happy */
340
341if (smtp_input)
342 {
343 msg = US"SMTP incoming data timeout";
344 log_write(L_lost_incoming_connection,
345 LOG_MAIN, "SMTP data timeout (message abandoned) on connection "
fed77020
PH
346 "from %s F=<%s>",
347 (sender_fullhost != NULL)? sender_fullhost : US"local process",
348 sender_address);
059ec3d9
PH
349 }
350else
351 {
352 fprintf(stderr, "exim: timed out while reading - message abandoned\n");
353 log_write(L_lost_incoming_connection,
354 LOG_MAIN, "timed out while reading local message");
355 }
356
8f128379 357receive_bomb_out(US"data-timeout", msg); /* Does not return */
059ec3d9
PH
358}
359
360
361
362/*************************************************
363* local_scan() timeout *
364*************************************************/
365
366/* Handler function for timeouts that occur while running a local_scan()
367function.
368
369Argument: the signal number
370Returns: nothing
371*/
372
373static void
374local_scan_timeout_handler(int sig)
375{
376sig = sig; /* Keep picky compilers happy */
377log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function timed out - "
378 "message temporarily rejected (size %d)", message_size);
8f128379
PH
379/* Does not return */
380receive_bomb_out(US"local-scan-timeout", US"local verification problem");
059ec3d9
PH
381}
382
383
384
385/*************************************************
386* local_scan() crashed *
387*************************************************/
388
389/* Handler function for signals that occur while running a local_scan()
390function.
391
392Argument: the signal number
393Returns: nothing
394*/
395
396static void
397local_scan_crash_handler(int sig)
398{
399log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function crashed with "
400 "signal %d - message temporarily rejected (size %d)", sig, message_size);
8f128379
PH
401/* Does not return */
402receive_bomb_out(US"local-scan-error", US"local verification problem");
059ec3d9
PH
403}
404
405
406/*************************************************
407* SIGTERM or SIGINT received *
408*************************************************/
409
410/* Handler for SIGTERM or SIGINT signals that occur while reading the
411data that comprises a message.
412
413Argument: the signal number
414Returns: nothing
415*/
416
417static void
418data_sigterm_sigint_handler(int sig)
419{
420uschar *msg = NULL;
421
422if (smtp_input)
423 {
424 msg = US"Service not available - SIGTERM or SIGINT received";
425 log_write(0, LOG_MAIN, "%s closed after %s", smtp_get_connection_info(),
426 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
427 }
428else
429 {
f05da2e8 430 if (filter_test == FTEST_NONE)
059ec3d9
PH
431 {
432 fprintf(stderr, "\nexim: %s received - message abandoned\n",
433 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
434 log_write(0, LOG_MAIN, "%s received while reading local message",
435 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
436 }
437 }
438
8f128379 439receive_bomb_out(US"signal-exit", msg); /* Does not return */
059ec3d9
PH
440}
441
442
443
444/*************************************************
445* Add new recipient to list *
446*************************************************/
447
448/* This function builds a list of recipient addresses in argc/argv
449format.
450
451Arguments:
452 recipient the next address to add to recipients_list
453 pno parent number for fixed aliases; -1 otherwise
454
455Returns: nothing
456*/
457
458void
459receive_add_recipient(uschar *recipient, int pno)
460{
461if (recipients_count >= recipients_list_max)
462 {
463 recipient_item *oldlist = recipients_list;
464 int oldmax = recipients_list_max;
465 recipients_list_max = recipients_list_max? 2*recipients_list_max : 50;
466 recipients_list = store_get(recipients_list_max * sizeof(recipient_item));
467 if (oldlist != NULL)
468 memcpy(recipients_list, oldlist, oldmax * sizeof(recipient_item));
469 }
470
471recipients_list[recipients_count].address = recipient;
472recipients_list[recipients_count].pno = pno;
8523533c
TK
473#ifdef EXPERIMENTAL_BRIGHTMAIL
474recipients_list[recipients_count].bmi_optin = bmi_current_optin;
475/* reset optin string pointer for next recipient */
476bmi_current_optin = NULL;
477#endif
059ec3d9
PH
478recipients_list[recipients_count++].errors_to = NULL;
479}
480
481
482
483
484/*************************************************
485* Remove a recipient from the list *
486*************************************************/
487
488/* This function is provided for local_scan() to use.
489
490Argument:
491 recipient address to remove
492
493Returns: TRUE if it did remove something; FALSE otherwise
494*/
495
496BOOL
497receive_remove_recipient(uschar *recipient)
498{
499int count;
500DEBUG(D_receive) debug_printf("receive_remove_recipient(\"%s\") called\n",
501 recipient);
502for (count = 0; count < recipients_count; count++)
503 {
504 if (Ustrcmp(recipients_list[count].address, recipient) == 0)
505 {
506 if ((--recipients_count - count) > 0)
507 memmove(recipients_list + count, recipients_list + count + 1,
54cdb463 508 (recipients_count - count)*sizeof(recipient_item));
059ec3d9
PH
509 return TRUE;
510 }
511 }
512return FALSE;
513}
514
515
516
517
518
519/*************************************************
520* Read data portion of a non-SMTP message *
521*************************************************/
522
523/* This function is called to read the remainder of a message (following the
524header) when the input is not from SMTP - we are receiving a local message on
525a standard input stream. The message is always terminated by EOF, and is also
526terminated by a dot on a line by itself if the flag dot_ends is TRUE. Split the
527two cases for maximum efficiency.
528
529Ensure that the body ends with a newline. This will naturally be the case when
530the termination is "\n.\n" but may not be otherwise. The RFC defines messages
531as "sequences of lines" - this of course strictly applies only to SMTP, but
532deliveries into BSD-type mailbox files also require it. Exim used to have a
533flag for doing this at delivery time, but as it was always set for all
534transports, I decided to simplify things by putting the check here instead.
535
536There is at least one MUA (dtmail) that sends CRLF via this interface, and
537other programs are known to do this as well. Exim used to have a option for
538dealing with this: in July 2003, after much discussion, the code has been
539changed to default to treat any of LF, CRLF, and bare CR as line terminators.
540
541However, for the case when a dot on a line by itself terminates a message, the
542only recognized terminating sequences before and after the dot are LF and CRLF.
543Otherwise, having read EOL . CR, you don't know whether to read another
544character or not.
545
546Internally, in messages stored in Exim's spool files, LF is used as the line
547terminator. Under the new regime, bare CRs will no longer appear in these
548files.
549
550Arguments:
551 fout a FILE to which to write the message
552
553Returns: One of the END_xxx values indicating why it stopped reading
554*/
555
556static int
557read_message_data(FILE *fout)
558{
559int ch_state;
560register int ch;
d677b2f2 561register int linelength = 0;
059ec3d9
PH
562
563/* Handle the case when only EOF terminates the message */
564
565if (!dot_ends)
566 {
567 register int last_ch = '\n';
568
80a47a2c 569 for (; (ch = (receive_getc)()) != EOF; last_ch = ch)
059ec3d9
PH
570 {
571 if (ch == 0) body_zerocount++;
572 if (last_ch == '\r' && ch != '\n')
573 {
d677b2f2
PH
574 if (linelength > max_received_linelength)
575 max_received_linelength = linelength;
576 linelength = 0;
059ec3d9
PH
577 if (fputc('\n', fout) == EOF) return END_WERROR;
578 message_size++;
579 body_linecount++;
580 }
581 if (ch == '\r') continue;
582
583 if (fputc(ch, fout) == EOF) return END_WERROR;
d677b2f2
PH
584 if (ch == '\n')
585 {
586 if (linelength > max_received_linelength)
587 max_received_linelength = linelength;
588 linelength = 0;
589 body_linecount++;
590 }
591 else linelength++;
059ec3d9
PH
592 if (++message_size > thismessage_size_limit) return END_SIZE;
593 }
594
595 if (last_ch != '\n')
596 {
d677b2f2
PH
597 if (linelength > max_received_linelength)
598 max_received_linelength = linelength;
059ec3d9
PH
599 if (fputc('\n', fout) == EOF) return END_WERROR;
600 message_size++;
601 body_linecount++;
602 }
603
604 return END_EOF;
605 }
606
607/* Handle the case when a dot on a line on its own, or EOF, terminates. */
608
609ch_state = 1;
610
80a47a2c 611while ((ch = (receive_getc)()) != EOF)
059ec3d9
PH
612 {
613 if (ch == 0) body_zerocount++;
614 switch (ch_state)
615 {
616 case 0: /* Normal state (previous char written) */
617 if (ch == '\n')
d677b2f2
PH
618 {
619 body_linecount++;
620 if (linelength > max_received_linelength)
621 max_received_linelength = linelength;
622 linelength = -1;
623 ch_state = 1;
624 }
059ec3d9
PH
625 else if (ch == '\r')
626 { ch_state = 2; continue; }
627 break;
628
629 case 1: /* After written "\n" */
630 if (ch == '.') { ch_state = 3; continue; }
d677b2f2 631 if (ch != '\n') ch_state = 0; else linelength = -1;
059ec3d9
PH
632 break;
633
634 case 2:
635 body_linecount++; /* After unwritten "\r" */
d677b2f2
PH
636 if (linelength > max_received_linelength)
637 max_received_linelength = linelength;
059ec3d9 638 if (ch == '\n')
d677b2f2
PH
639 {
640 ch_state = 1;
641 linelength = -1;
642 }
059ec3d9
PH
643 else
644 {
645 if (message_size++, fputc('\n', fout) == EOF) return END_WERROR;
646 if (ch == '\r') continue;
647 ch_state = 0;
d677b2f2 648 linelength = 0;
059ec3d9
PH
649 }
650 break;
651
652 case 3: /* After "\n." (\n written, dot not) */
653 if (ch == '\n') return END_DOT;
654 if (ch == '\r') { ch_state = 4; continue; }
655 message_size++;
d677b2f2 656 linelength++;
059ec3d9
PH
657 if (fputc('.', fout) == EOF) return END_WERROR;
658 ch_state = 0;
659 break;
660
661 case 4: /* After "\n.\r" (\n written, rest not) */
662 if (ch == '\n') return END_DOT;
663 message_size += 2;
664 body_linecount++;
665 if (fputs(".\n", fout) == EOF) return END_WERROR;
666 if (ch == '\r') { ch_state = 2; continue; }
667 ch_state = 0;
668 break;
669 }
670
d677b2f2 671 linelength++;
059ec3d9
PH
672 if (fputc(ch, fout) == EOF) return END_WERROR;
673 if (++message_size > thismessage_size_limit) return END_SIZE;
674 }
675
676/* Get here if EOF read. Unless we have just written "\n", we need to ensure
677the message ends with a newline, and we must also write any characters that
678were saved up while testing for an ending dot. */
679
680if (ch_state != 1)
681 {
682 static uschar *ends[] = { US"\n", NULL, US"\n", US".\n", US".\n" };
683 if (fputs(CS ends[ch_state], fout) == EOF) return END_WERROR;
684 message_size += Ustrlen(ends[ch_state]);
685 body_linecount++;
686 }
687
688return END_EOF;
689}
690
691
692
693
694/*************************************************
695* Read data portion of an SMTP message *
696*************************************************/
697
698/* This function is called to read the remainder of an SMTP message (after the
699headers), or to skip over it when an error has occurred. In this case, the
700output file is passed as NULL.
701
702If any line begins with a dot, that character is skipped. The input should only
703be successfully terminated by CR LF . CR LF unless it is local (non-network)
704SMTP, in which case the CRs are optional, but...
705
706FUDGE: It seems that sites on the net send out messages with just LF
707terminators, despite the warnings in the RFCs, and other MTAs handle this. So
708we make the CRs optional in all cases.
709
710July 2003: Bare CRs cause trouble. We now treat them as line terminators as
711well, so that there are no CRs in spooled messages. However, the message
712terminating dot is not recognized between two bare CRs.
713
714Arguments:
715 fout a FILE to which to write the message; NULL if skipping
716
717Returns: One of the END_xxx values indicating why it stopped reading
718*/
719
720static int
721read_message_data_smtp(FILE *fout)
722{
723int ch_state = 0;
724register int ch;
1f5497b2 725register int linelength = 0;
059ec3d9 726
80a47a2c 727while ((ch = (receive_getc)()) != EOF)
059ec3d9
PH
728 {
729 if (ch == 0) body_zerocount++;
730 switch (ch_state)
731 {
732 case 0: /* After LF or CRLF */
733 if (ch == '.')
734 {
735 ch_state = 3;
736 continue; /* Don't ever write . after LF */
737 }
738 ch_state = 1;
739
740 /* Else fall through to handle as normal uschar. */
741
742 case 1: /* Normal state */
743 if (ch == '\n')
744 {
745 ch_state = 0;
746 body_linecount++;
1f5497b2
PH
747 if (linelength > max_received_linelength)
748 max_received_linelength = linelength;
749 linelength = -1;
059ec3d9
PH
750 }
751 else if (ch == '\r')
752 {
753 ch_state = 2;
754 continue;
755 }
756 break;
757
758 case 2: /* After (unwritten) CR */
759 body_linecount++;
1f5497b2
PH
760 if (linelength > max_received_linelength)
761 max_received_linelength = linelength;
762 linelength = -1;
059ec3d9
PH
763 if (ch == '\n')
764 {
765 ch_state = 0;
766 }
767 else
768 {
769 message_size++;
770 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
771 if (ch != '\r') ch_state = 1; else continue;
772 }
773 break;
774
775 case 3: /* After [CR] LF . */
776 if (ch == '\n')
777 return END_DOT;
778 if (ch == '\r')
779 {
780 ch_state = 4;
781 continue;
782 }
783 ch_state = 1; /* The dot itself is removed */
784 break;
785
786 case 4: /* After [CR] LF . CR */
787 if (ch == '\n') return END_DOT;
788 message_size++;
789 body_linecount++;
790 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
791 if (ch == '\r')
792 {
793 ch_state = 2;
794 continue;
795 }
796 ch_state = 1;
797 break;
798 }
799
800 /* Add the character to the spool file, unless skipping; then loop for the
801 next. */
802
803 message_size++;
1f5497b2 804 linelength++;
059ec3d9
PH
805 if (fout != NULL)
806 {
807 if (fputc(ch, fout) == EOF) return END_WERROR;
808 if (message_size > thismessage_size_limit) return END_SIZE;
809 }
810 }
811
812/* Fall through here if EOF encountered. This indicates some kind of error,
813since a correct message is terminated by [CR] LF . [CR] LF. */
814
815return END_EOF;
816}
817
818
819
820
821/*************************************************
822* Swallow SMTP message *
823*************************************************/
824
825/* This function is called when there has been some kind of error while reading
826an SMTP message, and the remaining data may need to be swallowed. It is global
827because it is called from smtp_closedown() to shut down an incoming call
828tidily.
829
830Argument: a FILE from which to read the message
831Returns: nothing
832*/
833
834void
835receive_swallow_smtp(void)
836{
837if (message_ended >= END_NOTENDED)
838 message_ended = read_message_data_smtp(NULL);
839}
840
841
842
843/*************************************************
844* Handle lost SMTP connection *
845*************************************************/
846
847/* This function logs connection loss incidents and generates an appropriate
848SMTP response.
849
850Argument: additional data for the message
851Returns: the SMTP response
852*/
853
854static uschar *
855handle_lost_connection(uschar *s)
856{
857log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
858 "%s lost while reading message data%s", smtp_get_connection_info(), s);
859return US"421 Lost incoming connection";
860}
861
862
863
864
865/*************************************************
866* Handle a non-smtp reception error *
867*************************************************/
868
869/* This function is called for various errors during the reception of non-SMTP
870messages. It either sends a message to the sender of the problem message, or it
871writes to the standard error stream.
872
873Arguments:
874 errcode code for moan_to_sender(), identifying the error
875 text1 first message text, passed to moan_to_sender()
876 text2 second message text, used only for stderrr
877 error_rc code to pass to exim_exit if no problem
878 f FILE containing body of message (may be stdin)
879 hptr pointer to instore headers or NULL
880
881Returns: calls exim_exit(), which does not return
882*/
883
884static void
885give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
886 FILE *f, header_line *hptr)
887{
888if (error_handling == ERRORS_SENDER)
889 {
890 error_block eblock;
891 eblock.next = NULL;
892 eblock.text1 = text1;
893 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
894 error_rc = EXIT_FAILURE;
895 }
896else fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 897(void)fclose(f);
059ec3d9
PH
898exim_exit(error_rc);
899}
900
901
902
903/*************************************************
904* Add header lines set up by ACL *
905*************************************************/
906
850635b6
PH
907/* This function is called to add the header lines that were set up by
908statements in an ACL to the list of headers in memory. It is done in two stages
909like this, because when the ACL for RCPT is running, the other headers have not
910yet been received. This function is called twice; once just before running the
911DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
912are visible to the DATA ACL.
059ec3d9
PH
913
914Originally these header lines were added at the end. Now there is support for
915three different places: top, bottom, and after the Received: header(s). There
916will always be at least one Received: header, even if it is marked deleted, and
917even if something else has been put in front of it.
918
919Arguments:
920 acl_name text to identify which ACL
921
922Returns: nothing
923*/
924
925static void
926add_acl_headers(uschar *acl_name)
927{
928header_line *h, *next;
929header_line *last_received = NULL;
930
71fafd95 931if (acl_added_headers == NULL) return;
059ec3d9
PH
932DEBUG(D_receive|D_acl) debug_printf(">>Headers added by %s ACL:\n", acl_name);
933
71fafd95 934for (h = acl_added_headers; h != NULL; h = next)
059ec3d9
PH
935 {
936 next = h->next;
937
938 switch(h->type)
939 {
940 case htype_add_top:
941 h->next = header_list;
942 header_list = h;
943 DEBUG(D_receive|D_acl) debug_printf(" (at top)");
944 break;
945
946 case htype_add_rec:
947 if (last_received == NULL)
948 {
949 last_received = header_list;
950 while (!header_testname(last_received, US"Received", 8, FALSE))
951 last_received = last_received->next;
952 while (last_received->next != NULL &&
953 header_testname(last_received->next, US"Received", 8, FALSE))
954 last_received = last_received->next;
955 }
956 h->next = last_received->next;
957 last_received->next = h;
958 DEBUG(D_receive|D_acl) debug_printf(" (after Received:)");
959 break;
960
8523533c
TK
961 case htype_add_rfc:
962 /* add header before any header which is NOT Received: or Resent- */
963 last_received = header_list;
964 while ( (last_received->next != NULL) &&
965 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
966 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
967 last_received = last_received->next;
968 /* last_received now points to the last Received: or Resent-* header
969 in an uninterrupted chain of those header types (seen from the beginning
970 of all headers. Our current header must follow it. */
971 h->next = last_received->next;
972 last_received->next = h;
8e669ac1 973 DEBUG(D_receive|D_acl) debug_printf(" (before any non-Received: or Resent-*: header)");
8523533c
TK
974 break;
975
059ec3d9
PH
976 default:
977 h->next = NULL;
978 header_last->next = h;
979 break;
980 }
981
982 if (h->next == NULL) header_last = h;
983
984 /* Check for one of the known header types (From:, To:, etc.) though in
985 practice most added headers are going to be "other". Lower case
986 identification letters are never stored with the header; they are used
987 for existence tests when messages are received. So discard any lower case
988 flag values. */
989
990 h->type = header_checkname(h, FALSE);
991 if (h->type >= 'a') h->type = htype_other;
992
993 DEBUG(D_receive|D_acl) debug_printf(" %s", header_last->text);
994 }
995
71fafd95 996acl_added_headers = NULL;
059ec3d9
PH
997DEBUG(D_receive|D_acl) debug_printf(">>\n");
998}
999
1000
1001
1002/*************************************************
1003* Add host information for log line *
1004*************************************************/
1005
1006/* Called for acceptance and rejecting log lines. This adds information about
1007the calling host to a string that is being built dynamically.
1008
1009Arguments:
1010 s the dynamic string
1011 sizeptr points to the size variable
1012 ptrptr points to the pointer variable
1013
1014Returns: the extended string
1015*/
1016
1017static uschar *
1018add_host_info_for_log(uschar *s, int *sizeptr, int *ptrptr)
1019{
1020if (sender_fullhost != NULL)
1021 {
1022 s = string_append(s, sizeptr, ptrptr, 2, US" H=", sender_fullhost);
1023 if ((log_extra_selector & LX_incoming_interface) != 0 &&
1024 interface_address != NULL)
1025 {
1026 uschar *ss = string_sprintf(" I=[%s]:%d", interface_address,
1027 interface_port);
1028 s = string_cat(s, sizeptr, ptrptr, ss, Ustrlen(ss));
1029 }
1030 }
1031if (sender_ident != NULL)
1032 s = string_append(s, sizeptr, ptrptr, 2, US" U=", sender_ident);
1033if (received_protocol != NULL)
1034 s = string_append(s, sizeptr, ptrptr, 2, US" P=", received_protocol);
1035return s;
1036}
1037
1038
1039
63955bf2 1040#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1041
1042/*************************************************
54cdb463
PH
1043* Run the MIME ACL on a message *
1044*************************************************/
1045
1046/* This code is in a subroutine so that it can be used for both SMTP
1047and non-SMTP messages. It is called with a non-NULL ACL pointer.
1048
1049Arguments:
1050 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1051 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1052 smtp_reply_ptr Where SMTP reply is being built
1053 blackholed_by_ptr Where "blackholed by" message is being built
1054
1055Returns: TRUE to carry on; FALSE to abandon the message
1056*/
1057
1058static BOOL
1059run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1060 uschar **blackholed_by_ptr)
1061{
1062FILE *mbox_file;
1063uschar rfc822_file_path[2048];
1064unsigned long mbox_size;
1065header_line *my_headerlist;
1066uschar *user_msg, *log_msg;
1067int mime_part_count_buffer = -1;
7156b1ef 1068int rc = OK;
54cdb463
PH
1069
1070memset(CS rfc822_file_path,0,2048);
1071
1072/* check if it is a MIME message */
1073my_headerlist = header_list;
4e88a19f
PH
1074while (my_headerlist != NULL)
1075 {
54cdb463 1076 /* skip deleted headers */
4e88a19f
PH
1077 if (my_headerlist->type == '*')
1078 {
54cdb463
PH
1079 my_headerlist = my_headerlist->next;
1080 continue;
4e88a19f
PH
1081 }
1082 if (strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0)
1083 {
54cdb463
PH
1084 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1085 goto DO_MIME_ACL;
4e88a19f 1086 }
54cdb463 1087 my_headerlist = my_headerlist->next;
4e88a19f 1088 }
54cdb463
PH
1089
1090DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1091return TRUE;
1092
1093DO_MIME_ACL:
1094/* make sure the eml mbox file is spooled up */
8544e77a 1095mbox_file = spool_mbox(&mbox_size, NULL);
54cdb463
PH
1096if (mbox_file == NULL) {
1097 /* error while spooling */
1098 log_write(0, LOG_MAIN|LOG_PANIC,
1099 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1100 Uunlink(spool_name);
1101 unspool_mbox();
6f0c431a
PP
1102#ifdef EXPERIMENTAL_DCC
1103 dcc_ok = 0;
1104#endif
a5bd321b 1105 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1106 message_id[0] = 0; /* Indicate no message accepted */
1107 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1108 return FALSE; /* Indicate skip to end of receive function */
1109};
1110
1111mime_is_rfc822 = 0;
1112
1113MIME_ACL_CHECK:
1114mime_part_count = -1;
1115rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1116(void)fclose(mbox_file);
54cdb463 1117
4e88a19f
PH
1118if (Ustrlen(rfc822_file_path) > 0)
1119 {
54cdb463
PH
1120 mime_part_count = mime_part_count_buffer;
1121
4e88a19f
PH
1122 if (unlink(CS rfc822_file_path) == -1)
1123 {
54cdb463
PH
1124 log_write(0, LOG_PANIC,
1125 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1126 goto END_MIME_ACL;
4e88a19f
PH
1127 }
1128 }
54cdb463
PH
1129
1130/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1131if (rc == OK)
1132 {
54cdb463
PH
1133 uschar temp_path[1024];
1134 int n;
1135 struct dirent *entry;
1136 DIR *tempdir;
1137
b07e6aa3
PH
1138 (void)string_format(temp_path, 1024, "%s/scan/%s", spool_directory,
1139 message_id);
54cdb463 1140
4e88a19f
PH
1141 tempdir = opendir(CS temp_path);
1142 n = 0;
1143 do
1144 {
1145 entry = readdir(tempdir);
1146 if (entry == NULL) break;
1147 if (strncmpic(US entry->d_name,US"__rfc822_",9) == 0)
1148 {
b07e6aa3 1149 (void)string_format(rfc822_file_path, 2048,"%s/scan/%s/%s", spool_directory, message_id, entry->d_name);
4e88a19f
PH
1150 debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n", rfc822_file_path);
1151 break;
1152 }
1153 } while (1);
1154 closedir(tempdir);
54cdb463 1155
4e88a19f
PH
1156 if (entry != NULL)
1157 {
c58b88df 1158 mbox_file = Ufopen(rfc822_file_path,"rb");
4e88a19f
PH
1159 if (mbox_file == NULL)
1160 {
54cdb463
PH
1161 log_write(0, LOG_PANIC,
1162 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1163 unlink(CS rfc822_file_path);
1164 goto END_MIME_ACL;
4e88a19f 1165 }
54cdb463
PH
1166 /* set RFC822 expansion variable */
1167 mime_is_rfc822 = 1;
1168 mime_part_count_buffer = mime_part_count;
1169 goto MIME_ACL_CHECK;
4e88a19f
PH
1170 }
1171 }
54cdb463
PH
1172
1173END_MIME_ACL:
1174add_acl_headers(US"MIME");
1175if (rc == DISCARD)
1176 {
1177 recipients_count = 0;
1178 *blackholed_by_ptr = US"MIME ACL";
1179 }
1180else if (rc != OK)
1181 {
1182 Uunlink(spool_name);
1183 unspool_mbox();
6f0c431a
PP
1184#ifdef EXPERIMENTAL_DCC
1185 dcc_ok = 0;
1186#endif
54cdb463
PH
1187 if (smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1188 *smtp_yield_ptr = FALSE; /* No more messsages after dropped connection */
1189 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1190 message_id[0] = 0; /* Indicate no message accepted */
1191 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1192 }
54cdb463
PH
1193
1194return TRUE;
1195}
1196
63955bf2 1197#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1198
1199
1200/*************************************************
059ec3d9
PH
1201* Receive message *
1202*************************************************/
1203
1204/* Receive a message on the given input, and put it into a pair of spool files.
1205Either a non-null list of recipients, or the extract flag will be true, or
1206both. The flag sender_local is true for locally generated messages. The flag
1207submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1208suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1209suppress_local_fixups" or -G was passed on the command-line.
1210The flag smtp_input is true if the message is to be
8800895a
PH
1211handled using SMTP conventions about termination and lines starting with dots.
1212For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1213
1214If a message was successfully read, message_id[0] will be non-zero.
1215
1216The general actions of this function are:
1217
1218 . Read the headers of the message (if any) into a chain of store
1219 blocks.
1220
1221 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1222 throw it away, unless the caller is trusted, or unless
1223 active_local_sender_retain is set - which can only happen if
1224 active_local_from_check is false.
059ec3d9
PH
1225
1226 . If recipients are to be extracted from the message, build the
1227 recipients list from the headers, removing any that were on the
1228 original recipients list (unless extract_addresses_remove_arguments is
1229 false), and at the same time, remove any bcc header that may be present.
1230
1231 . Get the spool file for the data, sort out its unique name, open
1232 and lock it (but don't give it the name yet).
1233
1234 . Generate a "Message-Id" header if the message doesn't have one, for
1235 locally-originated messages.
1236
1237 . Generate a "Received" header.
1238
1239 . Ensure the recipients list is fully qualified and rewritten if necessary.
1240
1241 . If there are any rewriting rules, apply them to the sender address
1242 and also to the headers.
1243
1244 . If there is no from: header, generate one, for locally-generated messages
1245 and messages in "submission mode" only.
1246
1247 . If the sender is local, check that from: is correct, and if not, generate
1248 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1249 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1250
1251 . If there is no "date" header, generate one, for locally-originated
1252 or submission mode messages only.
1253
1254 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1255 dot_ends mode, to the data file. Leave it open, to hold the lock.
1256
1257 . Write the envelope and the headers to a new file.
1258
1259 . Set the name for the header file; close it.
1260
1261 . Set the name for the data file; close it.
1262
1263Because this function can potentially be called many times in a single
1264SMTP connection, all store should be got by store_get(), so that it will be
1265automatically retrieved after the message is accepted.
1266
1267FUDGE: It seems that sites on the net send out messages with just LF
1268terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1269we make the CRs optional in all cases.
1270
1271July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1272new regime is now in place in which bare CRs in header lines are turned into LF
1273followed by a space, so as not to terminate the header line.
1274
1275February 2004: A bare LF in a header line in a message whose first line was
1276terminated by CRLF is treated in the same way as a bare CR.
1277
1278Arguments:
1279 extract_recip TRUE if recipients are to be extracted from the message's
1280 headers
1281
1282Returns: TRUE there are more messages to be read (SMTP input)
1283 FALSE there are no more messages to be read (non-SMTP input
1284 or SMTP connection collapsed, or other failure)
1285
1286When reading a message for filter testing, the returned value indicates
1287whether the headers (which is all that is read) were terminated by '.' or
1288not. */
1289
1290BOOL
1291receive_msg(BOOL extract_recip)
1292{
7156b1ef
NM
1293int i;
1294int rc = FAIL;
059ec3d9
PH
1295int msg_size = 0;
1296int process_info_len = Ustrlen(process_info);
1297int error_rc = (error_handling == ERRORS_SENDER)?
1298 errors_sender_rc : EXIT_FAILURE;
1299int header_size = 256;
1300int start, end, domain, size, sptr;
1301int id_resolution;
1302int had_zero = 0;
d677b2f2 1303int prevlines_length = 0;
059ec3d9
PH
1304
1305register int ptr = 0;
1306
1307BOOL contains_resent_headers = FALSE;
1308BOOL extracted_ignored = FALSE;
1309BOOL first_line_ended_crlf = TRUE_UNSET;
1310BOOL smtp_yield = TRUE;
1311BOOL yield = FALSE;
1312
1313BOOL resents_exist = FALSE;
1314uschar *resent_prefix = US"";
1315uschar *blackholed_by = NULL;
04f7d5b9 1316uschar *blackhole_log_msg = US"";
059ec3d9
PH
1317
1318flock_t lock_data;
1319error_block *bad_addresses = NULL;
1320
1321uschar *frozen_by = NULL;
1322uschar *queued_by = NULL;
1323
1324uschar *errmsg, *s;
1325struct stat statbuf;
1326
4e88a19f 1327/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1328
1329uschar *smtp_reply = NULL;
4e88a19f 1330uschar *user_msg, *log_msg;
059ec3d9
PH
1331
1332/* Working header pointers */
1333
1334header_line *h, *next;
1335
2cbb4081 1336/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1337
1338BOOL date_header_exists = FALSE;
1339
1340/* Pointers to receive the addresses of headers whose contents we need. */
1341
1342header_line *from_header = NULL;
1343header_line *subject_header = NULL;
1344header_line *msgid_header = NULL;
1345header_line *received_header;
1346
1347/* Variables for use when building the Received: header. */
1348
1349uschar *received;
1350uschar *timestamp;
1351int tslen;
1352
1353/* Release any open files that might have been cached while preparing to
1354accept the message - e.g. by verifying addresses - because reading a message
1355might take a fair bit of real time. */
1356
1357search_tidyup();
1358
1359/* Initialize the chain of headers by setting up a place-holder for Received:
1360header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1361pointing to the end of the chain to make adding headers simple. */
1362
1363received_header = header_list = header_last = store_get(sizeof(header_line));
1364header_list->next = NULL;
1365header_list->type = htype_old;
1366header_list->text = NULL;
1367header_list->slen = 0;
1368
1369/* Control block for the next header to be read. */
1370
1371next = store_get(sizeof(header_line));
1372next->text = store_get(header_size);
1373
1374/* Initialize message id to be null (indicating no message read), and the
1375header names list to be the normal list. Indicate there is no data file open
1376yet, initialize the size and warning count, and deal with no size limit. */
1377
1378message_id[0] = 0;
1379data_file = NULL;
1380data_fd = -1;
1381spool_name[0] = 0;
1382message_size = 0;
1383warning_count = 0;
d677b2f2 1384received_count = 1; /* For the one we will add */
059ec3d9
PH
1385
1386if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1387
2e0c1448 1388/* While reading the message, the following counts are computed. */
059ec3d9 1389
d677b2f2
PH
1390message_linecount = body_linecount = body_zerocount =
1391 max_received_linelength = 0;
059ec3d9 1392
80a47a2c
TK
1393#ifndef DISABLE_DKIM
1394/* Call into DKIM to set up the context. */
1395if (smtp_input && !smtp_batched_input && !dkim_disable_verify) dkim_exim_verify_init();
fb2274d4
TK
1396#endif
1397
059ec3d9
PH
1398/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1399ids, and fractions of a second are required. See the comments that precede the
1400message id creation below. */
1401
1402(void)gettimeofday(&message_id_tv, NULL);
1403
1404/* For other uses of the received time we can operate with granularity of one
1405second, and for that we use the global variable received_time. This is for
1406things like ultimate message timeouts. */
1407
1408received_time = message_id_tv.tv_sec;
1409
1410/* If SMTP input, set the special handler for timeouts. The alarm() calls
1411happen in the smtp_getc() function when it refills its buffer. */
1412
1413if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1414
1415/* If not SMTP input, timeout happens only if configured, and we just set a
1416single timeout for the whole message. */
1417
1418else if (receive_timeout > 0)
1419 {
1420 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1421 alarm(receive_timeout);
1422 }
1423
1424/* SIGTERM and SIGINT are caught always. */
1425
1426signal(SIGTERM, data_sigterm_sigint_handler);
1427signal(SIGINT, data_sigterm_sigint_handler);
1428
1429/* Header lines in messages are not supposed to be very long, though when
1430unfolded, to: and cc: headers can take up a lot of store. We must also cope
1431with the possibility of junk being thrown at us. Start by getting 256 bytes for
1432storing the header, and extend this as necessary using string_cat().
1433
1434To cope with total lunacies, impose an upper limit on the length of the header
1435section of the message, as otherwise the store will fill up. We must also cope
1436with the possibility of binary zeros in the data. Hence we cannot use fgets().
1437Folded header lines are joined into one string, leaving the '\n' characters
1438inside them, so that writing them out reproduces the input.
1439
1440Loop for each character of each header; the next structure for chaining the
1441header is set up already, with ptr the offset of the next character in
1442next->text. */
1443
1444for (;;)
1445 {
80a47a2c 1446 int ch = (receive_getc)();
059ec3d9
PH
1447
1448 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1449 SMTP must have a correct "." terminator. */
1450
1451 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1452 {
1453 smtp_reply = handle_lost_connection(US" (header)");
1454 smtp_yield = FALSE;
1455 goto TIDYUP; /* Skip to end of function */
1456 }
1457
1458 /* See if we are at the current header's size limit - there must be at least
1459 four bytes left. This allows for the new character plus a zero, plus two for
1460 extra insertions when we are playing games with dots and carriage returns. If
1461 we are at the limit, extend the text buffer. This could have been done
1462 automatically using string_cat() but because this is a tightish loop storing
1463 only one character at a time, we choose to do it inline. Normally
1464 store_extend() will be able to extend the block; only at the end of a big
1465 store block will a copy be needed. To handle the case of very long headers
1466 (and sometimes lunatic messages can have ones that are 100s of K long) we
1467 call store_release() for strings that have been copied - if the string is at
1468 the start of a block (and therefore the only thing in it, because we aren't
1469 doing any other gets), the block gets freed. We can only do this because we
1470 know there are no other calls to store_get() going on. */
1471
1472 if (ptr >= header_size - 4)
1473 {
1474 int oldsize = header_size;
1475 /* header_size += 256; */
1476 header_size *= 2;
1477 if (!store_extend(next->text, oldsize, header_size))
1478 {
1479 uschar *newtext = store_get(header_size);
1480 memcpy(newtext, next->text, ptr);
1481 store_release(next->text);
1482 next->text = newtext;
1483 }
1484 }
1485
1486 /* Cope with receiving a binary zero. There is dispute about whether
1487 these should be allowed in RFC 822 messages. The middle view is that they
1488 should not be allowed in headers, at least. Exim takes this attitude at
1489 the moment. We can't just stomp on them here, because we don't know that
1490 this line is a header yet. Set a flag to cause scanning later. */
1491
1492 if (ch == 0) had_zero++;
1493
1494 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1495 those from data files use just LF. Treat LF in local SMTP input as a
1496 terminator too. Treat EOF as a line terminator always. */
1497
1498 if (ch == EOF) goto EOL;
1499
1500 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1501 other MTAs accept this. We are therefore forced into this "liberalisation"
1502 too, so we accept LF as a line terminator whatever the source of the message.
1503 However, if the first line of the message ended with a CRLF, we treat a bare
1504 LF specially by inserting a white space after it to ensure that the header
1505 line is not terminated. */
1506
1507 if (ch == '\n')
1508 {
1509 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1510 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1511 goto EOL;
1512 }
1513
1514 /* This is not the end of the line. If this is SMTP input and this is
1515 the first character in the line and it is a "." character, ignore it.
1516 This implements the dot-doubling rule, though header lines starting with
1517 dots aren't exactly common. They are legal in RFC 822, though. If the
1518 following is CRLF or LF, this is the line that that terminates the
1519 entire message. We set message_ended to indicate this has happened (to
1520 prevent further reading), and break out of the loop, having freed the
1521 empty header, and set next = NULL to indicate no data line. */
1522
1523 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1524 {
80a47a2c 1525 ch = (receive_getc)();
059ec3d9
PH
1526 if (ch == '\r')
1527 {
80a47a2c 1528 ch = (receive_getc)();
059ec3d9
PH
1529 if (ch != '\n')
1530 {
80a47a2c 1531 receive_ungetc(ch);
059ec3d9
PH
1532 ch = '\r'; /* Revert to CR */
1533 }
1534 }
1535 if (ch == '\n')
1536 {
1537 message_ended = END_DOT;
1538 store_reset(next);
1539 next = NULL;
1540 break; /* End character-reading loop */
1541 }
1542
1543 /* For non-SMTP input, the dot at the start of the line was really a data
1544 character. What is now in ch is the following character. We guaranteed
1545 enough space for this above. */
1546
1547 if (!smtp_input)
1548 {
1549 next->text[ptr++] = '.';
1550 message_size++;
1551 }
1552 }
1553
1554 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1555 remember this case if this is the first line ending. */
1556
1557 if (ch == '\r')
1558 {
80a47a2c 1559 ch = (receive_getc)();
059ec3d9
PH
1560 if (ch == '\n')
1561 {
1562 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1563 goto EOL;
1564 }
1565
1566 /* Otherwise, put back the character after CR, and turn the bare CR
1567 into LF SP. */
1568
80a47a2c 1569 ch = (receive_ungetc)(ch);
059ec3d9
PH
1570 next->text[ptr++] = '\n';
1571 message_size++;
1572 ch = ' ';
1573 }
1574
1575 /* We have a data character for the header line. */
1576
1577 next->text[ptr++] = ch; /* Add to buffer */
1578 message_size++; /* Total message size so far */
1579
1580 /* Handle failure due to a humungously long header section. The >= allows
1581 for the terminating \n. Add what we have so far onto the headers list so
1582 that it gets reflected in any error message, and back up the just-read
1583 character. */
1584
1585 if (message_size >= header_maxsize)
1586 {
1587 next->text[ptr] = 0;
1588 next->slen = ptr;
1589 next->type = htype_other;
1590 next->next = NULL;
1591 header_last->next = next;
1592 header_last = next;
1593
1594 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1595 "%s (more than %d characters): message abandoned",
1596 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1597
1598 if (smtp_input)
1599 {
1600 smtp_reply = US"552 Message header is ridiculously long";
1601 receive_swallow_smtp();
1602 goto TIDYUP; /* Skip to end of function */
1603 }
1604
1605 else
1606 {
1607 give_local_error(ERRMESS_VLONGHEADER,
1608 string_sprintf("message header longer than %d characters received: "
1609 "message not accepted", header_maxsize), US"", error_rc, stdin,
1610 header_list->next);
1611 /* Does not return */
1612 }
1613 }
1614
1615 continue; /* With next input character */
1616
1617 /* End of header line reached */
1618
1619 EOL:
2e0c1448
PH
1620
1621 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1622
1623 receive_linecount++;
1624 message_linecount++;
059ec3d9 1625
d677b2f2
PH
1626 /* Keep track of maximum line length */
1627
1628 if (ptr - prevlines_length > max_received_linelength)
1629 max_received_linelength = ptr - prevlines_length;
1630 prevlines_length = ptr + 1;
1631
059ec3d9
PH
1632 /* Now put in the terminating newline. There is always space for
1633 at least two more characters. */
1634
1635 next->text[ptr++] = '\n';
1636 message_size++;
1637
1638 /* A blank line signals the end of the headers; release the unwanted
1639 space and set next to NULL to indicate this. */
1640
1641 if (ptr == 1)
1642 {
1643 store_reset(next);
1644 next = NULL;
1645 break;
1646 }
1647
1648 /* There is data in the line; see if the next input character is a
1649 whitespace character. If it is, we have a continuation of this header line.
1650 There is always space for at least one character at this point. */
1651
1652 if (ch != EOF)
1653 {
80a47a2c 1654 int nextch = (receive_getc)();
059ec3d9
PH
1655 if (nextch == ' ' || nextch == '\t')
1656 {
1657 next->text[ptr++] = nextch;
1658 message_size++;
1659 continue; /* Iterate the loop */
1660 }
80a47a2c 1661 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
1662 else ch = EOF; /* Cause main loop to exit at end */
1663 }
1664
1665 /* We have got to the real line end. Terminate the string and release store
1666 beyond it. If it turns out to be a real header, internal binary zeros will
1667 be squashed later. */
1668
1669 next->text[ptr] = 0;
1670 next->slen = ptr;
1671 store_reset(next->text + ptr + 1);
1672
1673 /* Check the running total size against the overall message size limit. We
1674 don't expect to fail here, but if the overall limit is set less than MESSAGE_
1675 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
1676 headers - the code to read the body will then also hit the buffer. */
1677
1678 if (message_size > thismessage_size_limit) break;
1679
1680 /* A line that is not syntactically correct for a header also marks
1681 the end of the headers. In this case, we leave next containing the
1682 first data line. This might actually be several lines because of the
1683 continuation logic applied above, but that doesn't matter.
1684
1685 It turns out that smail, and presumably sendmail, accept leading lines
1686 of the form
1687
1688 From ph10 Fri Jan 5 12:35 GMT 1996
1689
1690 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
1691 find any documentation of this, but for compatibility it had better be
1692 accepted. Exim restricts it to the case of non-smtp messages, and
1693 treats it as an alternative to the -f command line option. Thus it is
1694 ignored except for trusted users or filter testing. Otherwise it is taken
1695 as the sender address, unless -f was used (sendmail compatibility).
1696
1697 It further turns out that some UUCPs generate the From_line in a different
1698 format, e.g.
1699
1700 From ph10 Fri, 7 Jan 97 14:00:00 GMT
1701
1702 The regex for matching these things is now capable of recognizing both
1703 formats (including 2- and 4-digit years in the latter). In fact, the regex
1704 is now configurable, as is the expansion string to fish out the sender.
1705
1706 Even further on it has been discovered that some broken clients send
1707 these lines in SMTP messages. There is now an option to ignore them from
1708 specified hosts or networks. Sigh. */
1709
1710 if (header_last == header_list &&
1711 (!smtp_input
1712 ||
1713 (sender_host_address != NULL &&
1714 verify_check_host(&ignore_fromline_hosts) == OK)
1715 ||
1716 (sender_host_address == NULL && ignore_fromline_local)
1717 ) &&
1718 regex_match_and_setup(regex_From, next->text, 0, -1))
1719 {
1720 if (!sender_address_forced)
1721 {
1722 uschar *uucp_sender = expand_string(uucp_from_sender);
1723 if (uucp_sender == NULL)
1724 {
1725 log_write(0, LOG_MAIN|LOG_PANIC,
1726 "expansion of \"%s\" failed after matching "
1727 "\"From \" line: %s", uucp_from_sender, expand_string_message);
1728 }
1729 else
1730 {
1731 int start, end, domain;
1732 uschar *errmess;
1733 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
1734 &start, &end, &domain, TRUE);
1735 if (newsender != NULL)
1736 {
1737 if (domain == 0 && newsender[0] != 0)
1738 newsender = rewrite_address_qualify(newsender, FALSE);
1739
f05da2e8 1740 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
1741 {
1742 sender_address = newsender;
1743
f05da2e8 1744 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
1745 {
1746 authenticated_sender = NULL;
1747 originator_name = US"";
1748 sender_local = FALSE;
1749 }
1750
f05da2e8 1751 if (filter_test != FTEST_NONE)
059ec3d9
PH
1752 printf("Sender taken from \"From \" line\n");
1753 }
1754 }
1755 }
1756 }
1757 }
1758
1759 /* Not a leading "From " line. Check to see if it is a valid header line.
1760 Header names may contain any non-control characters except space and colon,
1761 amazingly. */
1762
1763 else
1764 {
1765 uschar *p = next->text;
1766
1767 /* If not a valid header line, break from the header reading loop, leaving
1768 next != NULL, indicating that it holds the first line of the body. */
1769
1770 if (isspace(*p)) break;
1771 while (mac_isgraph(*p) && *p != ':') p++;
1772 while (isspace(*p)) p++;
1773 if (*p != ':')
1774 {
1775 body_zerocount = had_zero;
1776 break;
1777 }
1778
1779 /* We have a valid header line. If there were any binary zeroes in
1780 the line, stomp on them here. */
1781
1782 if (had_zero > 0)
1783 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
1784
1785 /* It is perfectly legal to have an empty continuation line
1786 at the end of a header, but it is confusing to humans
1787 looking at such messages, since it looks like a blank line.
1788 Reduce confusion by removing redundant white space at the
1789 end. We know that there is at least one printing character
1790 (the ':' tested for above) so there is no danger of running
1791 off the end. */
1792
1793 p = next->text + ptr - 2;
1794 for (;;)
1795 {
1796 while (*p == ' ' || *p == '\t') p--;
1797 if (*p != '\n') break;
1798 ptr = (p--) - next->text + 1;
1799 message_size -= next->slen - ptr;
1800 next->text[ptr] = 0;
1801 next->slen = ptr;
1802 }
1803
1804 /* Add the header to the chain */
1805
1806 next->type = htype_other;
1807 next->next = NULL;
1808 header_last->next = next;
1809 header_last = next;
1810
1811 /* Check the limit for individual line lengths. This comes after adding to
1812 the chain so that the failing line is reflected if a bounce is generated
1813 (for a local message). */
1814
1815 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
1816 {
1817 log_write(0, LOG_MAIN, "overlong message header line received from "
1818 "%s (more than %d characters): message abandoned",
1819 sender_host_unknown? sender_ident : sender_fullhost,
1820 header_line_maxsize);
1821
1822 if (smtp_input)
1823 {
1824 smtp_reply = US"552 A message header line is too long";
1825 receive_swallow_smtp();
1826 goto TIDYUP; /* Skip to end of function */
1827 }
1828
1829 else
1830 {
1831 give_local_error(ERRMESS_VLONGHDRLINE,
1832 string_sprintf("message header line longer than %d characters "
1833 "received: message not accepted", header_line_maxsize), US"",
1834 error_rc, stdin, header_list->next);
1835 /* Does not return */
1836 }
1837 }
1838
1839 /* Note if any resent- fields exist. */
1840
1841 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
1842 {
1843 resents_exist = TRUE;
1844 resent_prefix = US"Resent-";
1845 }
1846 }
1847
1848 /* The line has been handled. If we have hit EOF, break out of the loop,
1849 indicating no pending data line. */
1850
1851 if (ch == EOF) { next = NULL; break; }
1852
1853 /* Set up for the next header */
1854
1855 header_size = 256;
1856 next = store_get(sizeof(header_line));
1857 next->text = store_get(header_size);
1858 ptr = 0;
1859 had_zero = 0;
d677b2f2 1860 prevlines_length = 0;
059ec3d9
PH
1861 } /* Continue, starting to read the next header */
1862
1863/* At this point, we have read all the headers into a data structure in main
1864store. The first header is still the dummy placeholder for the Received: header
1865we are going to generate a bit later on. If next != NULL, it contains the first
1866data line - which terminated the headers before reaching a blank line (not the
1867normal case). */
1868
1869DEBUG(D_receive)
1870 {
1871 debug_printf(">>Headers received:\n");
1872 for (h = header_list->next; h != NULL; h = h->next)
1873 debug_printf("%s", h->text);
1874 debug_printf("\n");
1875 }
1876
1877/* End of file on any SMTP connection is an error. If an incoming SMTP call
1878is dropped immediately after valid headers, the next thing we will see is EOF.
1879We must test for this specially, as further down the reading of the data is
1880skipped if already at EOF. */
1881
1882if (smtp_input && (receive_feof)())
1883 {
1884 smtp_reply = handle_lost_connection(US" (after header)");
1885 smtp_yield = FALSE;
1886 goto TIDYUP; /* Skip to end of function */
1887 }
1888
1889/* If this is a filter test run and no headers were read, output a warning
1890in case there is a mistake in the test message. */
1891
f05da2e8 1892if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
1893 printf("Warning: no message headers read\n");
1894
1895
1896/* Scan the headers to identify them. Some are merely marked for later
1897processing; some are dealt with here. */
1898
1899for (h = header_list->next; h != NULL; h = h->next)
1900 {
1901 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
1902 if (is_resent) contains_resent_headers = TRUE;
1903
1904 switch (header_checkname(h, is_resent))
1905 {
059ec3d9 1906 case htype_bcc:
2cbb4081 1907 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
1908 break;
1909
059ec3d9 1910 case htype_cc:
2cbb4081 1911 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
1912 break;
1913
1914 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
1915
1916 case htype_date:
4c69d561 1917 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
1918 break;
1919
1920 /* Same comments as about Return-Path: below. */
1921
1922 case htype_delivery_date:
1923 if (delivery_date_remove) h->type = htype_old;
1924 break;
1925
1926 /* Same comments as about Return-Path: below. */
1927
1928 case htype_envelope_to:
1929 if (envelope_to_remove) h->type = htype_old;
1930 break;
1931
1932 /* Mark all "From:" headers so they get rewritten. Save the one that is to
1933 be used for Sender: checking. For Sendmail compatibility, if the "From:"
1934 header consists of just the login id of the user who called Exim, rewrite
1935 it with the gecos field first. Apply this rule to Resent-From: if there
1936 are resent- fields. */
1937
1938 case htype_from:
1939 h->type = htype_from;
1940 if (!resents_exist || is_resent)
1941 {
1942 from_header = h;
1943 if (!smtp_input)
1944 {
1945 uschar *s = Ustrchr(h->text, ':') + 1;
1946 while (isspace(*s)) s++;
1947 if (strncmpic(s, originator_login, h->slen - (s - h->text) - 1) == 0)
1948 {
1949 uschar *name = is_resent? US"Resent-From" : US"From";
1950 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
1951 originator_login, qualify_domain_sender);
1952 from_header = header_last;
1953 h->type = htype_old;
1954 DEBUG(D_receive|D_rewrite)
1955 debug_printf("rewrote \"%s:\" header using gecos\n", name);
1956 }
1957 }
1958 }
1959 break;
1960
1961 /* Identify the Message-id: header for generating "in-reply-to" in the
1962 autoreply transport. For incoming logging, save any resent- value. In both
1963 cases, take just the first of any multiples. */
1964
1965 case htype_id:
1966 if (msgid_header == NULL && (!resents_exist || is_resent))
1967 {
1968 msgid_header = h;
1969 h->type = htype_id;
1970 }
1971 break;
1972
1973 /* Flag all Received: headers */
1974
1975 case htype_received:
1976 h->type = htype_received;
1977 received_count++;
1978 break;
1979
1980 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
1981
1982 case htype_reply_to:
1983 h->type = htype_reply_to;
1984 break;
1985
1986 /* The Return-path: header is supposed to be added to messages when
1987 they leave the SMTP system. We shouldn't receive messages that already
1988 contain Return-path. However, since Exim generates Return-path: on
1989 local delivery, resent messages may well contain it. We therefore
1990 provide an option (which defaults on) to remove any Return-path: headers
1991 on input. Removal actually means flagging as "old", which prevents the
1992 header being transmitted with the message. */
1993
1994 case htype_return_path:
1995 if (return_path_remove) h->type = htype_old;
1996
1997 /* If we are testing a mail filter file, use the value of the
1998 Return-Path: header to set up the return_path variable, which is not
1999 otherwise set. However, remove any <> that surround the address
2000 because the variable doesn't have these. */
2001
f05da2e8 2002 if (filter_test != FTEST_NONE)
059ec3d9
PH
2003 {
2004 uschar *start = h->text + 12;
2005 uschar *end = start + Ustrlen(start);
2006 while (isspace(*start)) start++;
2007 while (end > start && isspace(end[-1])) end--;
2008 if (*start == '<' && end[-1] == '>')
2009 {
2010 start++;
2011 end--;
2012 }
2013 return_path = string_copyn(start, end - start);
2014 printf("Return-path taken from \"Return-path:\" header line\n");
2015 }
2016 break;
2017
2018 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2019 and from an untrusted caller and suppress_local_fixups is not set, or if we
2020 are in submission mode for a remote message, mark it "old" so that it will
2021 not be transmitted with the message, unless active_local_sender_retain is
2022 set. (This can only be true if active_local_from_check is false.) If there
2023 are any resent- headers in the message, apply this rule to Resent-Sender:
2024 instead of Sender:. Messages with multiple resent- header sets cannot be
2025 tidily handled. (For this reason, at least one MUA - Pine - turns old
2026 resent- headers into X-resent- headers when resending, leaving just one
2027 set.) */
059ec3d9
PH
2028
2029 case htype_sender:
69358f02 2030 h->type = ((!active_local_sender_retain &&
8800895a
PH
2031 (
2032 (sender_local && !trusted_caller && !suppress_local_fixups)
2033 || submission_mode
2034 )
059ec3d9
PH
2035 ) &&
2036 (!resents_exist||is_resent))?
2037 htype_old : htype_sender;
2038 break;
2039
2040 /* Remember the Subject: header for logging. There is no Resent-Subject */
2041
2042 case htype_subject:
2043 subject_header = h;
2044 break;
2045
2046 /* "To:" gets flagged, and the existence of a recipient header is noted,
2047 whether it's resent- or not. */
2048
2049 case htype_to:
2050 h->type = htype_to;
2051 /****
2052 to_or_cc_header_exists = TRUE;
2053 ****/
2054 break;
2055 }
2056 }
2057
2058/* Extract recipients from the headers if that is required (the -t option).
2059Note that this is documented as being done *before* any address rewriting takes
2060place. There are two possibilities:
2061
2062(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2063recipients already listed are to be REMOVED from the message. Smail 3 works
2064like this. We need to build a non-recipients tree for that list, because in
2065subsequent processing this data is held in a tree and that's what the
2066spool_write_header() function expects. Make sure that non-recipient addresses
2067are fully qualified and rewritten if necessary.
2068
2069(2) According to other sendmail documentation, -t ADDS extracted recipients to
2070those in the command line arguments (and it is rumoured some other MTAs do
2071this). Therefore, there is an option to make Exim behave this way.
2072
2073*** Notes on "Resent-" header lines ***
2074
2075The presence of resent-headers in the message makes -t horribly ambiguous.
2076Experiments with sendmail showed that it uses recipients for all resent-
2077headers, totally ignoring the concept of "sets of resent- headers" as described
2078in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2079with all the addresses in one instance of each header.
2080
2081This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2082error for -t if there were resent- headers in the message. However, after a
2083discussion on the mailing list, I've learned that there are MUAs that use
2084resent- headers with -t, and also that the stuff about sets of resent- headers
2085and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2086message with -t and resent- header lines makes sure that only *its* resent-
2087headers are present; previous ones are often renamed as X-resent- for example.
2088
2089Consequently, Exim has been changed so that, if any resent- header lines are
2090present, the recipients are taken from all of the appropriate resent- lines,
2091and not from the ordinary To:, Cc:, etc. */
2092
2093if (extract_recip)
2094 {
2095 int rcount = 0;
2096 error_block **bnext = &bad_addresses;
2097
2098 if (extract_addresses_remove_arguments)
2099 {
2100 while (recipients_count-- > 0)
2101 {
2102 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2103 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2104 tree_add_nonrecipient(s);
2105 }
2106 recipients_list = NULL;
2107 recipients_count = recipients_list_max = 0;
2108 }
2109
059ec3d9
PH
2110 /* Now scan the headers */
2111
2112 for (h = header_list->next; h != NULL; h = h->next)
2113 {
2114 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2115 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2116 {
2117 uschar *s = Ustrchr(h->text, ':') + 1;
2118 while (isspace(*s)) s++;
2119
1eccaa59
PH
2120 parse_allow_group = TRUE; /* Allow address group syntax */
2121
059ec3d9
PH
2122 while (*s != 0)
2123 {
2124 uschar *ss = parse_find_address_end(s, FALSE);
2125 uschar *recipient, *errmess, *p, *pp;
2126 int start, end, domain;
2127
2128 /* Check on maximum */
2129
2130 if (recipients_max > 0 && ++rcount > recipients_max)
2131 {
2132 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2133 US"message rejected: ", error_rc, stdin, NULL);
2134 /* Does not return */
2135 }
2136
2137 /* Make a copy of the address, and remove any internal newlines. These
2138 may be present as a result of continuations of the header line. The
2139 white space that follows the newline must not be removed - it is part
2140 of the header. */
2141
2142 pp = recipient = store_get(ss - s + 1);
2143 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2144 *pp = 0;
2145 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2146 &domain, FALSE);
2147
2148 /* Keep a list of all the bad addresses so we can send a single
2149 error message at the end. However, an empty address is not an error;
2150 just ignore it. This can come from an empty group list like
2151
2152 To: Recipients of list:;
2153
2154 If there are no recipients at all, an error will occur later. */
2155
2156 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2157 {
2158 int len = Ustrlen(s);
2159 error_block *b = store_get(sizeof(error_block));
2160 while (len > 0 && isspace(s[len-1])) len--;
2161 b->next = NULL;
2162 b->text1 = string_printing(string_copyn(s, len));
2163 b->text2 = errmess;
2164 *bnext = b;
2165 bnext = &(b->next);
2166 }
2167
2168 /* If the recipient is already in the nonrecipients tree, it must
2169 have appeared on the command line with the option extract_addresses_
2170 remove_arguments set. Do not add it to the recipients, and keep a note
2171 that this has happened, in order to give a better error if there are
2172 no recipients left. */
2173
2174 else if (recipient != NULL)
2175 {
2176 if (tree_search(tree_nonrecipients, recipient) == NULL)
2177 receive_add_recipient(recipient, -1);
2178 else
2179 extracted_ignored = TRUE;
2180 }
2181
2182 /* Move on past this address */
2183
2184 s = ss + (*ss? 1:0);
2185 while (isspace(*s)) s++;
1eccaa59
PH
2186 } /* Next address */
2187
2188 parse_allow_group = FALSE; /* Reset group syntax flags */
2189 parse_found_group = FALSE;
059ec3d9
PH
2190
2191 /* If this was the bcc: header, mark it "old", which means it
2192 will be kept on the spool, but not transmitted as part of the
2193 message. */
2194
2cbb4081 2195 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2196 } /* For appropriate header line */
2197 } /* For each header line */
2198
059ec3d9
PH
2199 }
2200
2201/* Now build the unique message id. This has changed several times over the
2202lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2203Retaining all the history in the comment has become too unwieldy - read
2204previous release sources if you want it.
2205
2206The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2207The first part is the current time, in seconds. The second part is the current
2208pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2209can hold a number in the range 0-3843. It used to be a computed sequence
2210number, but is now the fractional component of the current time in units of
22111/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2212received, Exim ensures that the timer has ticked at the appropriate level
2213before proceeding, to avoid duplication if the pid happened to be re-used
2214within the same time period. It seems likely that most messages will take at
2215least half a millisecond to be received, so no delay will normally be
2216necessary. At least for some time...
2217
2218There is a modification when localhost_number is set. Formerly this was allowed
2219to be as large as 255. Now it is restricted to the range 0-16, and the final
2220component of the message id becomes (localhost_number * 200) + fractional time
2221in units of 1/200 of a second (i.e. a value in the range 0-3399).
2222
2223Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2224Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2225still allows the tttttt field to hold a large enough number to last for some
2226more decades, and the final two-digit field can hold numbers up to 1295, which
2227is enough for milliseconds (instead of 1/2000 of a second).
2228
2229However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2230pid, so it is probably safe because pids have to be positive. The
2231localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2232final field becomes (localhost_number * 100) + fractional time in centiseconds.
2233
2234Note that string_base62() returns its data in a static storage block, so it
2235must be copied before calling string_base62() again. It always returns exactly
22366 characters.
2237
2238There doesn't seem to be anything in the RFC which requires a message id to
2239start with a letter, but Smail was changed to ensure this. The external form of
2240the message id (as supplied by string expansion) therefore starts with an
2241additional leading 'E'. The spool file names do not include this leading
2242letter and it is not used internally.
2243
2244NOTE: If ever the format of message ids is changed, the regular expression for
2245checking that a string is in this format must be updated in a corresponding
2246way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2247must also be changed to reflect the correct string length. Then, of course,
2248other programs that rely on the message id format will need updating too. */
2249
2250Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2251message_id[6] = '-';
2252Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2253
2254/* Deal with the case where the host number is set. The value of the number was
2255checked when it was read, to ensure it isn't too big. The timing granularity is
2256left in id_resolution so that an appropriate wait can be done after receiving
2257the message, if necessary (we hope it won't be). */
2258
2259if (host_number_string != NULL)
2260 {
2261 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2262 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2263 string_base62((long int)(
2264 host_number * (1000000/id_resolution) +
2265 message_id_tv.tv_usec/id_resolution)) + 4);
2266 }
2267
2268/* Host number not set: final field is just the fractional time at an
2269appropriate resolution. */
2270
2271else
2272 {
2273 id_resolution = (BASE_62 == 62)? 500 : 1000;
2274 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2275 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2276 }
2277
2278/* Add the current message id onto the current process info string if
2279it will fit. */
2280
2281(void)string_format(process_info + process_info_len,
2282 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2283
2284/* If we are using multiple input directories, set up the one for this message
2285to be the least significant base-62 digit of the time of arrival. Otherwise
2286ensure that it is an empty string. */
2287
2288message_subdir[0] = split_spool_directory? message_id[5] : 0;
2289
2290/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2291one, but only for local (without suppress_local_fixups) or submission mode
2292messages. This can be user-configured if required, but we had better flatten
2293any illegal characters therein. */
059ec3d9 2294
8800895a
PH
2295if (msgid_header == NULL &&
2296 ((sender_host_address == NULL && !suppress_local_fixups)
2297 || submission_mode))
059ec3d9
PH
2298 {
2299 uschar *p;
2300 uschar *id_text = US"";
2301 uschar *id_domain = primary_hostname;
2302
2303 /* Permit only letters, digits, dots, and hyphens in the domain */
2304
2305 if (message_id_domain != NULL)
2306 {
2307 uschar *new_id_domain = expand_string(message_id_domain);
2308 if (new_id_domain == NULL)
2309 {
2310 if (!expand_string_forcedfail)
2311 log_write(0, LOG_MAIN|LOG_PANIC,
2312 "expansion of \"%s\" (message_id_header_domain) "
2313 "failed: %s", message_id_domain, expand_string_message);
2314 }
2315 else if (*new_id_domain != 0)
2316 {
2317 id_domain = new_id_domain;
2318 for (p = id_domain; *p != 0; p++)
2319 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2320 }
2321 }
2322
2323 /* Permit all characters except controls and RFC 2822 specials in the
2324 additional text part. */
2325
2326 if (message_id_text != NULL)
2327 {
2328 uschar *new_id_text = expand_string(message_id_text);
2329 if (new_id_text == NULL)
2330 {
2331 if (!expand_string_forcedfail)
2332 log_write(0, LOG_MAIN|LOG_PANIC,
2333 "expansion of \"%s\" (message_id_header_text) "
2334 "failed: %s", message_id_text, expand_string_message);
2335 }
2336 else if (*new_id_text != 0)
2337 {
2338 id_text = new_id_text;
2339 for (p = id_text; *p != 0; p++)
2340 if (mac_iscntrl_or_special(*p)) *p = '-';
2341 }
2342 }
2343
e7e680d6
PP
2344 /* Add the header line
2345 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2346 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2347
e7e680d6 2348 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2349 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2350 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2351 }
2352
2353/* If we are to log recipients, keep a copy of the raw ones before any possible
2354rewriting. Must copy the count, because later ACLs and the local_scan()
2355function may mess with the real recipients. */
2356
2357if ((log_extra_selector & LX_received_recipients) != 0)
2358 {
2359 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2360 for (i = 0; i < recipients_count; i++)
2361 raw_recipients[i] = string_copy(recipients_list[i].address);
2362 raw_recipients_count = recipients_count;
2363 }
2364
2365/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2366recipients will get here only if the conditions were right (allow_unqualified_
2367recipient is TRUE). */
2368
2369for (i = 0; i < recipients_count; i++)
2370 recipients_list[i].address =
2371 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2372 global_rewrite_rules, rewrite_existflags);
2373
8800895a
PH
2374/* If there is no From: header, generate one for local (without
2375suppress_local_fixups) or submission_mode messages. If there is no sender
2376address, but the sender is local or this is a local delivery error, use the
2377originator login. This shouldn't happen for genuine bounces, but might happen
2378for autoreplies. The addition of From: must be done *before* checking for the
2379possible addition of a Sender: header, because untrusted_set_sender allows an
2380untrusted user to set anything in the envelope (which might then get info
2381From:) but we still want to ensure a valid Sender: if it is required. */
2382
2383if (from_header == NULL &&
2384 ((sender_host_address == NULL && !suppress_local_fixups)
2385 || submission_mode))
059ec3d9 2386 {
2fe1a124
PH
2387 uschar *oname = US"";
2388
2389 /* Use the originator_name if this is a locally submitted message and the
2390 caller is not trusted. For trusted callers, use it only if -F was used to
2391 force its value or if we have a non-SMTP message for which -f was not used
2392 to set the sender. */
2393
2394 if (sender_host_address == NULL)
2395 {
2396 if (!trusted_caller || sender_name_forced ||
2397 (!smtp_input && !sender_address_forced))
2398 oname = originator_name;
2399 }
2400
2401 /* For non-locally submitted messages, the only time we use the originator
2402 name is when it was forced by the /name= option on control=submission. */
2403
2404 else
2405 {
2406 if (submission_name != NULL) oname = submission_name;
2407 }
2408
059ec3d9
PH
2409 /* Envelope sender is empty */
2410
2411 if (sender_address[0] == 0)
2412 {
87ba3f5f
PH
2413 uschar *fromstart, *fromend;
2414
2415 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2416 oname, (oname[0] == 0)? "" : " <");
2417 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2418
059ec3d9
PH
2419 if (sender_local || local_error_message)
2420 {
87ba3f5f
PH
2421 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2422 local_part_quote(originator_login), qualify_domain_sender,
2423 fromend);
059ec3d9
PH
2424 }
2425 else if (submission_mode && authenticated_id != NULL)
2426 {
2427 if (submission_domain == NULL)
2428 {
87ba3f5f
PH
2429 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2430 local_part_quote(authenticated_id), qualify_domain_sender,
2431 fromend);
059ec3d9
PH
2432 }
2433 else if (submission_domain[0] == 0) /* empty => whole address set */
2434 {
87ba3f5f
PH
2435 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2436 fromend);
059ec3d9
PH
2437 }
2438 else
2439 {
87ba3f5f
PH
2440 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2441 local_part_quote(authenticated_id), submission_domain,
2442 fromend);
059ec3d9
PH
2443 }
2444 from_header = header_last; /* To get it checked for Sender: */
2445 }
2446 }
2447
2448 /* There is a non-null envelope sender. Build the header using the original
2449 sender address, before any rewriting that might have been done while
2450 verifying it. */
2451
2452 else
2453 {
87ba3f5f 2454 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2455 oname,
2456 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2457 (sender_address_unrewritten == NULL)?
2458 sender_address : sender_address_unrewritten,
2fe1a124 2459 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2460
2461 from_header = header_last; /* To get it checked for Sender: */
2462 }
2463 }
2464
2465
8800895a
PH
2466/* If the sender is local (without suppress_local_fixups), or if we are in
2467submission mode and there is an authenticated_id, check that an existing From:
2468is correct, and if not, generate a Sender: header, unless disabled. Any
2469previously-existing Sender: header was removed above. Note that sender_local,
2470as well as being TRUE if the caller of exim is not trusted, is also true if a
2471trusted caller did not supply a -f argument for non-smtp input. To allow
2472trusted callers to forge From: without supplying -f, we have to test explicitly
2473here. If the From: header contains more than one address, then the call to
2474parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2475
2476if (from_header != NULL &&
69358f02 2477 (active_local_from_check &&
8800895a 2478 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2479 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2480 ))
2481 {
2482 BOOL make_sender = TRUE;
2483 int start, end, domain;
2484 uschar *errmess;
2485 uschar *from_address =
2486 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2487 &start, &end, &domain, FALSE);
2488 uschar *generated_sender_address;
2489
2490 if (submission_mode)
2491 {
2492 if (submission_domain == NULL)
2493 {
2494 generated_sender_address = string_sprintf("%s@%s",
2495 local_part_quote(authenticated_id), qualify_domain_sender);
2496 }
2497 else if (submission_domain[0] == 0) /* empty => full address */
2498 {
2499 generated_sender_address = string_sprintf("%s",
2500 authenticated_id);
2501 }
2502 else
2503 {
2504 generated_sender_address = string_sprintf("%s@%s",
2505 local_part_quote(authenticated_id), submission_domain);
2506 }
2507 }
2508 else
2509 generated_sender_address = string_sprintf("%s@%s",
2510 local_part_quote(originator_login), qualify_domain_sender);
2511
2512 /* Remove permitted prefixes and suffixes from the local part of the From:
2513 address before doing the comparison with the generated sender. */
2514
2515 if (from_address != NULL)
2516 {
2517 int slen;
2518 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2519
2520 if (at != NULL) *at = 0;
2521 from_address += route_check_prefix(from_address, local_from_prefix);
2522 slen = route_check_suffix(from_address, local_from_suffix);
2523 if (slen > 0)
2524 {
2525 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2526 from_address += slen;
2527 }
2528 if (at != NULL) *at = '@';
2529
2530 if (strcmpic(generated_sender_address, from_address) == 0 ||
2531 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2532 make_sender = FALSE;
2533 }
2534
2535 /* We have to cause the Sender header to be rewritten if there are
2536 appropriate rewriting rules. */
2537
2538 if (make_sender)
2539 {
2fe1a124 2540 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2541 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2542 generated_sender_address);
2543 else
2544 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2545 resent_prefix,
2546 submission_mode? submission_name : originator_name,
2547 generated_sender_address);
059ec3d9 2548 }
87ba3f5f
PH
2549
2550 /* Ensure that a non-null envelope sender address corresponds to the
2551 submission mode sender address. */
2552
2553 if (submission_mode && sender_address[0] != 0)
2554 {
2555 if (sender_address_unrewritten == NULL)
2556 sender_address_unrewritten = sender_address;
2557 sender_address = generated_sender_address;
089793a4
TF
2558 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2559 log_write(L_address_rewrite, LOG_MAIN,
2560 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2561 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2562 }
059ec3d9
PH
2563 }
2564
2565
2566/* If there are any rewriting rules, apply them to the sender address, unless
2567it has already been rewritten as part of verification for SMTP input. */
2568
2569if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2570 sender_address[0] != 0)
2571 {
2572 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2573 global_rewrite_rules, rewrite_existflags);
2574 DEBUG(D_receive|D_rewrite)
2575 debug_printf("rewritten sender = %s\n", sender_address);
2576 }
2577
2578
2579/* The headers must be run through rewrite_header(), because it ensures that
2580addresses are fully qualified, as well as applying any rewriting rules that may
2581exist.
2582
2583Qualification of header addresses in a message from a remote host happens only
2584if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2585appropriate. For local messages, qualification always happens, unless -bnq is
2586used to explicitly suppress it. No rewriting is done for an unqualified address
2587that is left untouched.
2588
2589We start at the second header, skipping our own Received:. This rewriting is
2590documented as happening *after* recipient addresses are taken from the headers
2591by the -t command line option. An added Sender: gets rewritten here. */
2592
2593for (h = header_list->next; h != NULL; h = h->next)
2594 {
2595 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2596 rewrite_existflags, TRUE);
2597 if (newh != NULL) h = newh;
2598 }
2599
2600
2601/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2602"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2603"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2604exists, this applies to the set of resent- headers rather than the normal set.
2605
2cbb4081
PH
2606The requirement for a recipient header has been removed in RFC 2822. At this
2607point in the code, earlier versions of Exim added a To: header for locally
2608submitted messages, and an empty Bcc: header for others. In the light of the
2609changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2610
059ec3d9
PH
2611
2612/* If there is no date header, generate one if the message originates locally
8800895a
PH
2613(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2614submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2615to be more confusing if Exim adds one to all remotely-originated messages.
2616As per Message-Id, we prepend if resending, else append.
2617*/
059ec3d9 2618
8800895a
PH
2619if (!date_header_exists &&
2620 ((sender_host_address == NULL && !suppress_local_fixups)
2621 || submission_mode))
e7e680d6
PP
2622 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2623 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2624
2625search_tidyup(); /* Free any cached resources */
2626
2627/* Show the complete set of headers if debugging. Note that the first one (the
2628new Received:) has not yet been set. */
2629
2630DEBUG(D_receive)
2631 {
2632 debug_printf(">>Headers after rewriting and local additions:\n");
2633 for (h = header_list->next; h != NULL; h = h->next)
2634 debug_printf("%c %s", h->type, h->text);
2635 debug_printf("\n");
2636 }
2637
2638/* The headers are now complete in store. If we are running in filter
2639testing mode, that is all this function does. Return TRUE if the message
2640ended with a dot. */
2641
f05da2e8 2642if (filter_test != FTEST_NONE)
059ec3d9
PH
2643 {
2644 process_info[process_info_len] = 0;
2645 return message_ended == END_DOT;
2646 }
2647
2648/* Open a new spool file for the data portion of the message. We need
2649to access it both via a file descriptor and a stream. Try to make the
2650directory if it isn't there. Note re use of sprintf: spool_directory
2651is checked on input to be < 200 characters long. */
2652
2653sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory, message_subdir,
2654 message_id);
2655data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2656if (data_fd < 0)
2657 {
2658 if (errno == ENOENT)
2659 {
2660 uschar temp[16];
2661 sprintf(CS temp, "input/%s", message_subdir);
2662 if (message_subdir[0] == 0) temp[5] = 0;
2663 (void)directory_make(spool_directory, temp, INPUT_DIRECTORY_MODE, TRUE);
2664 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
2665 }
2666 if (data_fd < 0)
2667 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
2668 spool_name, strerror(errno));
2669 }
2670
2671/* Make sure the file's group is the Exim gid, and double-check the mode
2672because the group setting doesn't always get set automatically. */
2673
ff790e47
PH
2674(void)fchown(data_fd, exim_uid, exim_gid);
2675(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
2676
2677/* We now have data file open. Build a stream for it and lock it. We lock only
2678the first line of the file (containing the message ID) because otherwise there
2679are problems when Exim is run under Cygwin (I'm told). See comments in
2680spool_in.c, where the same locking is done. */
2681
2682data_file = fdopen(data_fd, "w+");
2683lock_data.l_type = F_WRLCK;
2684lock_data.l_whence = SEEK_SET;
2685lock_data.l_start = 0;
2686lock_data.l_len = SPOOL_DATA_START_OFFSET;
2687
2688if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
2689 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
2690 errno, strerror(errno));
2691
2692/* We have an open, locked data file. Write the message id to it to make it
2693self-identifying. Then read the remainder of the input of this message and
2694write it to the data file. If the variable next != NULL, it contains the first
2695data line (which was read as a header but then turned out not to have the right
2696format); write it (remembering that it might contain binary zeros). The result
2697of fwrite() isn't inspected; instead we call ferror() below. */
2698
2699fprintf(data_file, "%s-D\n", message_id);
2700if (next != NULL)
2701 {
2702 uschar *s = next->text;
2703 int len = next->slen;
f1e894f3 2704 (void)fwrite(s, 1, len, data_file);
059ec3d9
PH
2705 body_linecount++; /* Assumes only 1 line */
2706 }
2707
2708/* Note that we might already be at end of file, or the logical end of file
2709(indicated by '.'), or might have encountered an error while writing the
2710message id or "next" line. */
2711
2712if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
2713 {
2714 if (smtp_input)
2715 {
2716 message_ended = read_message_data_smtp(data_file);
2717 receive_linecount++; /* The terminating "." line */
2718 }
2719 else message_ended = read_message_data(data_file);
2720
2721 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 2722 message_linecount += body_linecount;
059ec3d9
PH
2723
2724 /* Handle premature termination of SMTP */
2725
2726 if (smtp_input && message_ended == END_EOF)
2727 {
2728 Uunlink(spool_name); /* Lose data file when closed */
2729 message_id[0] = 0; /* Indicate no message accepted */
2730 smtp_reply = handle_lost_connection(US"");
2731 smtp_yield = FALSE;
2732 goto TIDYUP; /* Skip to end of function */
2733 }
2734
2735 /* Handle message that is too big. Don't use host_or_ident() in the log
2736 message; we want to see the ident value even for non-remote messages. */
2737
2738 if (message_ended == END_SIZE)
2739 {
2740 Uunlink(spool_name); /* Lose the data file when closed */
2741 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2742
2743 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2744 "message too big: read=%d max=%d",
2745 sender_address,
2746 (sender_fullhost == NULL)? "" : " H=",
2747 (sender_fullhost == NULL)? US"" : sender_fullhost,
2748 (sender_ident == NULL)? "" : " U=",
2749 (sender_ident == NULL)? US"" : sender_ident,
2750 message_size,
2751 thismessage_size_limit);
2752
2753 if (smtp_input)
2754 {
2755 smtp_reply = US"552 Message size exceeds maximum permitted";
2756 message_id[0] = 0; /* Indicate no message accepted */
2757 goto TIDYUP; /* Skip to end of function */
2758 }
2759 else
2760 {
2761 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2762 give_local_error(ERRMESS_TOOBIG,
2763 string_sprintf("message too big (max=%d)", thismessage_size_limit),
2764 US"message rejected: ", error_rc, data_file, header_list);
2765 /* Does not return */
2766 }
2767 }
2768 }
2769
2770/* Restore the standard SIGALRM handler for any subsequent processing. (For
2771example, there may be some expansion in an ACL that uses a timer.) */
2772
2773os_non_restarting_signal(SIGALRM, sigalrm_handler);
2774
2775/* The message body has now been read into the data file. Call fflush() to
2776empty the buffers in C, and then call fsync() to get the data written out onto
2777the disk, as fflush() doesn't do this (or at least, it isn't documented as
2778having to do this). If there was an I/O error on either input or output,
2779attempt to send an error message, and unlink the spool file. For non-SMTP input
2780we can then give up. Note that for SMTP input we must swallow the remainder of
2781the input in cases of output errors, since the far end doesn't expect to see
2782anything until the terminating dot line is sent. */
2783
2784if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 2785 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
2786 {
2787 uschar *msg_errno = US strerror(errno);
2788 BOOL input_error = (receive_ferror)() != 0;
2789 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
2790 input_error? "Input read" : "Spool write",
2791 msg_errno,
2792 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
2793
2794 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
2795 Uunlink(spool_name); /* Lose the data file */
2796
2797 if (smtp_input)
2798 {
2799 if (input_error)
2800 smtp_reply = US"451 Error while reading input data";
2801 else
2802 {
2803 smtp_reply = US"451 Error while writing spool file";
2804 receive_swallow_smtp();
2805 }
2806 message_id[0] = 0; /* Indicate no message accepted */
2807 goto TIDYUP; /* Skip to end of function */
2808 }
2809
2810 else
2811 {
2812 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2813 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
2814 header_list);
2815 /* Does not return */
2816 }
2817 }
2818
2819
2820/* No I/O errors were encountered while writing the data file. */
2821
2822DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
2823
2824
2825/* If there were any bad addresses extracted by -t, or there were no recipients
2826left after -t, send a message to the sender of this message, or write it to
2827stderr if the error handling option is set that way. Note that there may
2828legitimately be no recipients for an SMTP message if they have all been removed
2829by "discard".
2830
2831We need to rewind the data file in order to read it. In the case of no
2832recipients or stderr error writing, throw the data file away afterwards, and
2833exit. (This can't be SMTP, which always ensures there's at least one
2834syntactically good recipient address.) */
2835
2836if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
2837 {
2838 DEBUG(D_receive)
2839 {
2840 if (recipients_count == 0) debug_printf("*** No recipients\n");
2841 if (bad_addresses != NULL)
2842 {
2843 error_block *eblock = bad_addresses;
2844 debug_printf("*** Bad address(es)\n");
2845 while (eblock != NULL)
2846 {
2847 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
2848 eblock = eblock->next;
2849 }
2850 }
2851 }
2852
2853 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
2854
2855 /* If configured to send errors to the sender, but this fails, force
2856 a failure error code. We use a special one for no recipients so that it
2857 can be detected by the autoreply transport. Otherwise error_rc is set to
2858 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
2859 it is EXIT_SUCCESS. */
2860
2861 if (error_handling == ERRORS_SENDER)
2862 {
2863 if (!moan_to_sender(
2864 (bad_addresses == NULL)?
2865 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
2866 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
2867 bad_addresses, header_list, data_file, FALSE))
2868 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
2869 }
2870 else
2871 {
2872 if (bad_addresses == NULL)
2873 {
2874 if (extracted_ignored)
2875 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
2876 else
2877 fprintf(stderr, "exim: no recipients in message\n");
2878 }
2879 else
2880 {
2881 fprintf(stderr, "exim: invalid address%s",
2882 (bad_addresses->next == NULL)? ":" : "es:\n");
2883 while (bad_addresses != NULL)
2884 {
2885 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
2886 bad_addresses->text2);
2887 bad_addresses = bad_addresses->next;
2888 }
2889 }
2890 }
2891
2892 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
2893 {
2894 Uunlink(spool_name);
f1e894f3 2895 (void)fclose(data_file);
059ec3d9
PH
2896 exim_exit(error_rc);
2897 }
2898 }
2899
2900/* Data file successfully written. Generate text for the Received: header by
2901expanding the configured string, and adding a timestamp. By leaving this
2902operation till now, we ensure that the timestamp is the time that message
2903reception was completed. However, this is deliberately done before calling the
2904data ACL and local_scan().
2905
2906This Received: header may therefore be inspected by the data ACL and by code in
2907the local_scan() function. When they have run, we update the timestamp to be
2908the final time of reception.
2909
2910If there is just one recipient, set up its value in the $received_for variable
2911for use when we generate the Received: header.
2912
2913Note: the checking for too many Received: headers is handled by the delivery
2914code. */
2915
2916timestamp = expand_string(US"${tod_full}");
2917if (recipients_count == 1) received_for = recipients_list[0].address;
2918received = expand_string(received_header_text);
2919received_for = NULL;
2920
2921if (received == NULL)
2922 {
2923 Uunlink(spool_name); /* Lose the data file */
2924 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
2925 "(received_header_text) failed: %s", string_printing(received_header_text),
2926 expand_string_message);
2927 }
2928
2929/* The first element on the header chain is reserved for the Received header,
2930so all we have to do is fill in the text pointer, and set the type. However, if
2931the result of the expansion is an empty string, we leave the header marked as
2932"old" so as to refrain from adding a Received header. */
2933
2934if (received[0] == 0)
2935 {
2936 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
2937 received_header->type = htype_old;
2938 }
2939else
2940 {
2941 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
2942 received_header->type = htype_received;
2943 }
2944
2945received_header->slen = Ustrlen(received_header->text);
2946
2947DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
2948 received_header->type, received_header->text);
2949
2950/* Set the value of message_body_size for the DATA ACL and for local_scan() */
2951
2952message_body_size = (fstat(data_fd, &statbuf) == 0)?
2953 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
2954
2955/* If an ACL from any RCPT commands set up any warning headers to add, do so
2956now, before running the DATA ACL. */
2957
2958add_acl_headers(US"MAIL or RCPT");
2959
2960/* If an ACL is specified for checking things at this stage of reception of a
2961message, run it, unless all the recipients were removed by "discard" in earlier
2962ACLs. That is the only case in which recipients_count can be zero at this
2963stage. Set deliver_datafile to point to the data file so that $message_body and
2964$message_body_end can be extracted if needed. Allow $recipients in expansions.
2965*/
2966
2967deliver_datafile = data_fd;
4e88a19f 2968user_msg = NULL;
059ec3d9 2969
0e20aff9
MH
2970enable_dollar_recipients = TRUE;
2971
059ec3d9
PH
2972if (recipients_count == 0)
2973 {
2974 blackholed_by = recipients_discarded? US"MAIL ACL" : US"RCPT ACL";
2975 }
2976else
2977 {
059ec3d9
PH
2978 /* Handle interactive SMTP messages */
2979
2980 if (smtp_input && !smtp_batched_input)
2981 {
8523533c 2982
80a47a2c
TK
2983#ifndef DISABLE_DKIM
2984 if (!dkim_disable_verify)
2985 {
2986 /* Finish verification, this will log individual signature results to
2987 the mainlog */
2988 dkim_exim_verify_finish();
2989
2990 /* Check if we must run the DKIM ACL */
2991 if ((acl_smtp_dkim != NULL) &&
2992 (dkim_verify_signers != NULL) &&
2993 (dkim_verify_signers[0] != '\0'))
2994 {
2995 uschar *dkim_verify_signers_expanded =
2996 expand_string(dkim_verify_signers);
2997 if (dkim_verify_signers_expanded == NULL)
2998 {
2999 log_write(0, LOG_MAIN|LOG_PANIC,
3000 "expansion of dkim_verify_signers option failed: %s",
3001 expand_string_message);
3002 }
3003 else
3004 {
3005 int sep = 0;
3006 uschar *ptr = dkim_verify_signers_expanded;
3007 uschar *item = NULL;
9e5d6b55
TK
3008 uschar *seen_items = NULL;
3009 int seen_items_size = 0;
3010 int seen_items_offset = 0;
80a47a2c 3011 uschar itembuf[256];
9122af94
TK
3012 /* Default to OK when no items are present */
3013 rc = OK;
80a47a2c
TK
3014 while ((item = string_nextinlist(&ptr, &sep,
3015 itembuf,
3016 sizeof(itembuf))) != NULL)
3017 {
6119d1ea
TK
3018 /* Prevent running ACL for an empty item */
3019 if (!item || (item[0] == '\0')) continue;
9e5d6b55
TK
3020 /* Only run ACL once for each domain or identity, no matter how often it
3021 appears in the expanded list. */
6119d1ea
TK
3022 if (seen_items != NULL)
3023 {
ae9094bf
TK
3024 uschar *seen_item = NULL;
3025 uschar seen_item_buf[256];
4a73449b 3026 uschar *seen_items_list = seen_items;
ae9094bf
TK
3027 int seen_this_item = 0;
3028
3029 while ((seen_item = string_nextinlist(&seen_items_list, &sep,
3030 seen_item_buf,
3031 sizeof(seen_item_buf))) != NULL)
3032 {
3033 if (Ustrcmp(seen_item,item) == 0)
3034 {
3035 seen_this_item = 1;
3036 break;
3037 }
3038 }
3039
3040 if (seen_this_item > 0)
6119d1ea
TK
3041 {
3042 DEBUG(D_receive)
3043 debug_printf("acl_smtp_dkim: skipping signer %s, already seen\n", item);
3044 continue;
3045 }
ae9094bf 3046
4a73449b 3047 seen_items = string_append(seen_items,&seen_items_size,&seen_items_offset,1,":");
6119d1ea
TK
3048 }
3049
4a73449b
TK
3050 seen_items = string_append(seen_items,&seen_items_size,&seen_items_offset,1,item);
3051 seen_items[seen_items_offset] = '\0';
6119d1ea
TK
3052
3053 DEBUG(D_receive)
3054 debug_printf("calling acl_smtp_dkim for dkim_cur_signer=%s\n", item);
3055
80a47a2c
TK
3056 dkim_exim_acl_setup(item);
3057 rc = acl_check(ACL_WHERE_DKIM, NULL, acl_smtp_dkim, &user_msg, &log_msg);
6119d1ea
TK
3058
3059 if (rc != OK)
3060 {
3061 DEBUG(D_receive)
3062 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, skipping remaining items\n", rc, item);
3063 break;
3064 }
80a47a2c
TK
3065 }
3066 add_acl_headers(US"DKIM");
3067 if (rc == DISCARD)
3068 {
3069 recipients_count = 0;
3070 blackholed_by = US"DKIM ACL";
3071 if (log_msg != NULL)
3072 blackhole_log_msg = string_sprintf(": %s", log_msg);
3073 }
3074 else if (rc != OK)
3075 {
3076 Uunlink(spool_name);
3077 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
3078 smtp_yield = FALSE; /* No more messsages after dropped connection */
3079 smtp_reply = US""; /* Indicate reply already sent */
3080 message_id[0] = 0; /* Indicate no message accepted */
3081 goto TIDYUP; /* Skip to end of function */
3082 }
3083 }
3084 }
3085 }
3086#endif /* DISABLE_DKIM */
fb2274d4 3087
8523533c 3088#ifdef WITH_CONTENT_SCAN
80a47a2c
TK
3089 if (recipients_count > 0 &&
3090 acl_smtp_mime != NULL &&
54cdb463
PH
3091 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
3092 goto TIDYUP;
8523533c
TK
3093#endif /* WITH_CONTENT_SCAN */
3094
54cdb463
PH
3095 /* Check the recipients count again, as the MIME ACL might have changed
3096 them. */
8523533c 3097
059ec3d9
PH
3098 if (acl_smtp_data != NULL && recipients_count > 0)
3099 {
059ec3d9
PH
3100 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
3101 add_acl_headers(US"DATA");
3102 if (rc == DISCARD)
3103 {
3104 recipients_count = 0;
3105 blackholed_by = US"DATA ACL";
8e669ac1
PH
3106 if (log_msg != NULL)
3107 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9
PH
3108 }
3109 else if (rc != OK)
3110 {
3111 Uunlink(spool_name);
8523533c
TK
3112#ifdef WITH_CONTENT_SCAN
3113 unspool_mbox();
3114#endif
6f0c431a
PP
3115#ifdef EXPERIMENTAL_DCC
3116 dcc_ok = 0;
3117#endif
059ec3d9
PH
3118 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
3119 smtp_yield = FALSE; /* No more messsages after dropped connection */
3120 smtp_reply = US""; /* Indicate reply already sent */
3121 message_id[0] = 0; /* Indicate no message accepted */
3122 goto TIDYUP; /* Skip to end of function */
3123 }
3124 }
3125 }
3126
3127 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
3128 we cannot take different actions for permanent and temporary rejections. */
3129
54cdb463 3130 else
059ec3d9 3131 {
54cdb463
PH
3132
3133#ifdef WITH_CONTENT_SCAN
3134 if (acl_not_smtp_mime != NULL &&
3135 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
3136 &blackholed_by))
3137 goto TIDYUP;
3138#endif /* WITH_CONTENT_SCAN */
3139
3140 if (acl_not_smtp != NULL)
059ec3d9 3141 {
54cdb463
PH
3142 uschar *user_msg, *log_msg;
3143 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
3144 if (rc == DISCARD)
059ec3d9 3145 {
54cdb463
PH
3146 recipients_count = 0;
3147 blackholed_by = US"non-SMTP ACL";
3148 if (log_msg != NULL)
3149 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 3150 }
54cdb463 3151 else if (rc != OK)
059ec3d9 3152 {
54cdb463
PH
3153 Uunlink(spool_name);
3154#ifdef WITH_CONTENT_SCAN
3155 unspool_mbox();
3156#endif
6f0c431a
PP
3157#ifdef EXPERIMENTAL_DCC
3158 dcc_ok = 0;
3159#endif
6ea85e9a
PH
3160 /* The ACL can specify where rejections are to be logged, possibly
3161 nowhere. The default is main and reject logs. */
3162
3163 if (log_reject_target != 0)
3164 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3165 sender_address, log_msg);
3166
54cdb463
PH
3167 if (user_msg == NULL) user_msg = US"local configuration problem";
3168 if (smtp_batched_input)
3169 {
3170 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3171 /* Does not return */
3172 }
3173 else
3174 {
3175 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3176 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3177 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3178 header_list);
3179 /* Does not return */
3180 }
059ec3d9 3181 }
54cdb463 3182 add_acl_headers(US"non-SMTP");
059ec3d9 3183 }
059ec3d9
PH
3184 }
3185
54cdb463
PH
3186 /* The applicable ACLs have been run */
3187
059ec3d9
PH
3188 if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */
3189 if (queue_only_policy) queued_by = US"ACL";
059ec3d9
PH
3190 }
3191
8523533c
TK
3192#ifdef WITH_CONTENT_SCAN
3193unspool_mbox();
3194#endif
3195
6a8f9482
TK
3196#ifdef EXPERIMENTAL_DCC
3197dcc_ok = 0;
3198#endif
3199
3200
059ec3d9
PH
3201/* The final check on the message is to run the scan_local() function. The
3202version supplied with Exim always accepts, but this is a hook for sysadmins to
3203supply their own checking code. The local_scan() function is run even when all
3204the recipients have been discarded. */
3205
3206lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3207
3208/* Arrange to catch crashes in local_scan(), so that the -D file gets
3209deleted, and the incident gets logged. */
3210
3211os_non_restarting_signal(SIGSEGV, local_scan_crash_handler);
3212os_non_restarting_signal(SIGFPE, local_scan_crash_handler);
3213os_non_restarting_signal(SIGILL, local_scan_crash_handler);
3214os_non_restarting_signal(SIGBUS, local_scan_crash_handler);
3215
3216DEBUG(D_receive) debug_printf("calling local_scan(); timeout=%d\n",
3217 local_scan_timeout);
3218local_scan_data = NULL;
3219
3220os_non_restarting_signal(SIGALRM, local_scan_timeout_handler);
3221if (local_scan_timeout > 0) alarm(local_scan_timeout);
3222rc = local_scan(data_fd, &local_scan_data);
3223alarm(0);
3224os_non_restarting_signal(SIGALRM, sigalrm_handler);
3225
0e20aff9
MH
3226enable_dollar_recipients = FALSE;
3227
059ec3d9
PH
3228store_pool = POOL_MAIN; /* In case changed */
3229DEBUG(D_receive) debug_printf("local_scan() returned %d %s\n", rc,
3230 local_scan_data);
3231
3232os_non_restarting_signal(SIGSEGV, SIG_DFL);
3233os_non_restarting_signal(SIGFPE, SIG_DFL);
3234os_non_restarting_signal(SIGILL, SIG_DFL);
3235os_non_restarting_signal(SIGBUS, SIG_DFL);
3236
3237/* The length check is paranoia against some runaway code, and also because
3238(for a success return) lines in the spool file are read into big_buffer. */
3239
3240if (local_scan_data != NULL)
3241 {
3242 int len = Ustrlen(local_scan_data);
3243 if (len > LOCAL_SCAN_MAX_RETURN) len = LOCAL_SCAN_MAX_RETURN;
3244 local_scan_data = string_copyn(local_scan_data, len);
3245 }
3246
3247if (rc == LOCAL_SCAN_ACCEPT_FREEZE)
3248 {
58eb016e 3249 if (!deliver_freeze) /* ACL might have already frozen */
059ec3d9
PH
3250 {
3251 deliver_freeze = TRUE;
3252 deliver_frozen_at = time(NULL);
3253 frozen_by = US"local_scan()";
3254 }
3255 rc = LOCAL_SCAN_ACCEPT;
3256 }
3257else if (rc == LOCAL_SCAN_ACCEPT_QUEUE)
3258 {
3259 if (!queue_only_policy) /* ACL might have already queued */
3260 {
3261 queue_only_policy = TRUE;
3262 queued_by = US"local_scan()";
3263 }
3264 rc = LOCAL_SCAN_ACCEPT;
3265 }
3266
3267/* Message accepted: remove newlines in local_scan_data because otherwise
3268the spool file gets corrupted. Ensure that all recipients are qualified. */
3269
3270if (rc == LOCAL_SCAN_ACCEPT)
3271 {
3272 if (local_scan_data != NULL)
3273 {
3274 uschar *s;
3275 for (s = local_scan_data; *s != 0; s++) if (*s == '\n') *s = ' ';
3276 }
3277 for (i = 0; i < recipients_count; i++)
3278 {
3279 recipient_item *r = recipients_list + i;
3280 r->address = rewrite_address_qualify(r->address, TRUE);
3281 if (r->errors_to != NULL)
3282 r->errors_to = rewrite_address_qualify(r->errors_to, TRUE);
3283 }
3284 if (recipients_count == 0 && blackholed_by == NULL)
3285 blackholed_by = US"local_scan";
3286 }
3287
3288/* Message rejected: newlines permitted in local_scan_data to generate
3289multiline SMTP responses. */
3290
3291else
3292 {
3293 uschar *istemp = US"";
3294 uschar *s = NULL;
a5bd321b 3295 uschar *smtp_code;
059ec3d9
PH
3296 int size = 0;
3297 int sptr = 0;
059ec3d9
PH
3298
3299 errmsg = local_scan_data;
3300
3301 Uunlink(spool_name); /* Cancel this message */
3302 switch(rc)
3303 {
3304 default:
3305 log_write(0, LOG_MAIN, "invalid return %d from local_scan(). Temporary "
3306 "rejection given", rc);
3307 goto TEMPREJECT;
3308
3309 case LOCAL_SCAN_REJECT_NOLOGHDR:
3310 log_extra_selector &= ~LX_rejected_header;
3311 /* Fall through */
3312
3313 case LOCAL_SCAN_REJECT:
a5bd321b 3314 smtp_code = US"550";
059ec3d9
PH
3315 if (errmsg == NULL) errmsg = US"Administrative prohibition";
3316 break;
3317
3318 case LOCAL_SCAN_TEMPREJECT_NOLOGHDR:
3319 log_extra_selector &= ~LX_rejected_header;
3320 /* Fall through */
3321
3322 case LOCAL_SCAN_TEMPREJECT:
3323 TEMPREJECT:
a5bd321b 3324 smtp_code = US"451";
059ec3d9
PH
3325 if (errmsg == NULL) errmsg = US"Temporary local problem";
3326 istemp = US"temporarily ";
3327 break;
3328 }
3329
3330 s = string_append(s, &size, &sptr, 2, US"F=",
3331 (sender_address[0] == 0)? US"<>" : sender_address);
3332 s = add_host_info_for_log(s, &size, &sptr);
3333 s[sptr] = 0;
3334
3335 log_write(0, LOG_MAIN|LOG_REJECT, "%s %srejected by local_scan(): %.256s",
3336 s, istemp, string_printing(errmsg));
3337
3338 if (smtp_input)
3339 {
3340 if (!smtp_batched_input)
3341 {
a5bd321b 3342 smtp_respond(smtp_code, 3, TRUE, errmsg);
059ec3d9
PH
3343 message_id[0] = 0; /* Indicate no message accepted */
3344 smtp_reply = US""; /* Indicate reply already sent */
3345 goto TIDYUP; /* Skip to end of function */
3346 }
3347 else
3348 {
a5bd321b 3349 moan_smtp_batch(NULL, "%s %s", smtp_code, errmsg);
059ec3d9
PH
3350 /* Does not return */
3351 }
3352 }
3353 else
3354 {
3355 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3356 give_local_error(ERRMESS_LOCAL_SCAN, errmsg,
3357 US"message rejected by local scan code: ", error_rc, data_file,
3358 header_list);
3359 /* Does not return */
3360 }
3361 }
3362
3363/* Reset signal handlers to ignore signals that previously would have caused
3364the message to be abandoned. */
3365
3366signal(SIGTERM, SIG_IGN);
3367signal(SIGINT, SIG_IGN);
3368
3369/* Ensure the first time flag is set in the newly-received message. */
3370
3371deliver_firsttime = TRUE;
3372
8523533c
TK
3373#ifdef EXPERIMENTAL_BRIGHTMAIL
3374if (bmi_run == 1) {
3375 /* rewind data file */
3376 lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3377 bmi_verdicts = bmi_process_message(header_list, data_fd);
3378};
3379#endif
3380
059ec3d9
PH
3381/* Update the timstamp in our Received: header to account for any time taken by
3382an ACL or by local_scan(). The new time is the time that all reception
3383processing is complete. */
3384
3385timestamp = expand_string(US"${tod_full}");
3386tslen = Ustrlen(timestamp);
3387
3388memcpy(received_header->text + received_header->slen - tslen - 1,
3389 timestamp, tslen);
3390
3391/* In MUA wrapper mode, ignore queueing actions set by ACL or local_scan() */
3392
3393if (mua_wrapper)
3394 {
3395 deliver_freeze = FALSE;
3396 queue_only_policy = FALSE;
3397 }
3398
3399/* Keep the data file open until we have written the header file, in order to
3400hold onto the lock. In a -bh run, or if the message is to be blackholed, we
3401don't write the header file, and we unlink the data file. If writing the header
3402file fails, we have failed to accept this message. */
3403
3404if (host_checking || blackholed_by != NULL)
3405 {
3406 header_line *h;
3407 Uunlink(spool_name);
3408 msg_size = 0; /* Compute size for log line */
3409 for (h = header_list; h != NULL; h = h->next)
3410 if (h->type != '*') msg_size += h->slen;
3411 }
3412
3413/* Write the -H file */
3414
3415else
3416 {
3417 if ((msg_size = spool_write_header(message_id, SW_RECEIVING, &errmsg)) < 0)
3418 {
3419 log_write(0, LOG_MAIN, "Message abandoned: %s", errmsg);
3420 Uunlink(spool_name); /* Lose the data file */
3421
3422 if (smtp_input)
3423 {
3424 smtp_reply = US"451 Error in writing spool file";
3425 message_id[0] = 0; /* Indicate no message accepted */
3426 goto TIDYUP;
3427 }
3428 else
3429 {
3430 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3431 give_local_error(ERRMESS_IOERR, errmsg, US"", error_rc, data_file,
3432 header_list);
3433 /* Does not return */
3434 }
3435 }
3436 }
3437
3438
3439/* The message has now been successfully received. */
3440
3441receive_messagecount++;
3442
3443/* In SMTP sessions we may receive several in one connection. After each one,
3444we wait for the clock to tick at the level of message-id granularity. This is
3445so that the combination of time+pid is unique, even on systems where the pid
3446can be re-used within our time interval. We can't shorten the interval without
3447re-designing the message-id. See comments above where the message id is
3448created. This is Something For The Future. */
3449
3450message_id_tv.tv_usec = (message_id_tv.tv_usec/id_resolution) * id_resolution;
3451exim_wait_tick(&message_id_tv, id_resolution);
3452
3453/* Add data size to written header size. We do not count the initial file name
3454that is in the file, but we do add one extra for the notional blank line that
3455precedes the data. This total differs from message_size in that it include the
3456added Received: header and any other headers that got created locally. */
3457
3458fflush(data_file);
3459fstat(data_fd, &statbuf);
3460
3461msg_size += statbuf.st_size - SPOOL_DATA_START_OFFSET + 1;
3462
3463/* Generate a "message received" log entry. We do this by building up a dynamic
3464string as required. Since we commonly want to add two items at a time, use a
3465macro to simplify the coding. We log the arrival of a new message while the
3466file is still locked, just in case the machine is *really* fast, and delivers
3467it first! Include any message id that is in the message - since the syntax of a
3468message id is actually an addr-spec, we can use the parse routine to canonicize
3469it. */
3470
3471size = 256;
3472sptr = 0;
3473s = store_get(size);
3474
3475s = string_append(s, &size, &sptr, 2, US"<= ",
3476 (sender_address[0] == 0)? US"<>" : sender_address);
3477if (message_reference != NULL)
3478 s = string_append(s, &size, &sptr, 2, US" R=", message_reference);
3479
3480s = add_host_info_for_log(s, &size, &sptr);
3481
3482#ifdef SUPPORT_TLS
3483if ((log_extra_selector & LX_tls_cipher) != 0 && tls_cipher != NULL)
3484 s = string_append(s, &size, &sptr, 2, US" X=", tls_cipher);
3485if ((log_extra_selector & LX_tls_certificate_verified) != 0 &&
3486 tls_cipher != NULL)
3487 s = string_append(s, &size, &sptr, 2, US" CV=",
3488 tls_certificate_verified? "yes":"no");
3489if ((log_extra_selector & LX_tls_peerdn) != 0 && tls_peerdn != NULL)
48ed62d9
PH
3490 s = string_append(s, &size, &sptr, 3, US" DN=\"",
3491 string_printing(tls_peerdn), US"\"");
3f0945ff
PP
3492if ((log_extra_selector & LX_tls_sni) != 0 && tls_sni != NULL)
3493 s = string_append(s, &size, &sptr, 3, US" SNI=\"",
3494 string_printing(tls_sni), US"\"");
3495#endif
059ec3d9
PH
3496
3497if (sender_host_authenticated != NULL)
3498 {
3499 s = string_append(s, &size, &sptr, 2, US" A=", sender_host_authenticated);
3500 if (authenticated_id != NULL)
3501 s = string_append(s, &size, &sptr, 2, US":", authenticated_id);
3502 }
3503
3504sprintf(CS big_buffer, "%d", msg_size);
3505s = string_append(s, &size, &sptr, 2, US" S=", big_buffer);
3506
3507/* If an addr-spec in a message-id contains a quoted string, it can contain
3508any characters except " \ and CR and so in particular it can contain NL!
3509Therefore, make sure we use a printing-characters only version for the log.
3510Also, allow for domain literals in the message id. */
3511
3512if (msgid_header != NULL)
3513 {
3514 uschar *old_id;
3515 BOOL save_allow_domain_literals = allow_domain_literals;
3516 allow_domain_literals = TRUE;
3517 old_id = parse_extract_address(Ustrchr(msgid_header->text, ':') + 1,
3518 &errmsg, &start, &end, &domain, FALSE);
3519 allow_domain_literals = save_allow_domain_literals;
3520 if (old_id != NULL)
3521 s = string_append(s, &size, &sptr, 2, US" id=", string_printing(old_id));
3522 }
3523
3524/* If subject logging is turned on, create suitable printing-character
3525text. By expanding $h_subject: we make use of the MIME decoding. */
3526
3527if ((log_extra_selector & LX_subject) != 0 && subject_header != NULL)
3528 {
3529 int i;
3530 uschar *p = big_buffer;
3531 uschar *ss = expand_string(US"$h_subject:");
3532
3533 /* Backslash-quote any double quotes or backslashes so as to make a
3534 a C-like string, and turn any non-printers into escape sequences. */
3535
3536 *p++ = '\"';
3537 if (*ss != 0) for (i = 0; i < 100 && ss[i] != 0; i++)
3538 {
3539 if (ss[i] == '\"' || ss[i] == '\\') *p++ = '\\';
3540 *p++ = ss[i];
3541 }
3542 *p++ = '\"';
3543 *p = 0;
3544 s = string_append(s, &size, &sptr, 2, US" T=", string_printing(big_buffer));
3545 }
3546
3547/* Terminate the string: string_cat() and string_append() leave room, but do
3548not put the zero in. */
3549
3550s[sptr] = 0;
3551
059ec3d9
PH
3552/* Create a message log file if message logs are being used and this message is
3553not blackholed. Write the reception stuff to it. We used to leave message log
3554creation until the first delivery, but this has proved confusing for somep
3555people. */
3556
3557if (message_logs && blackholed_by == NULL)
3558 {
3559 int fd;
3560
3561 sprintf(CS spool_name, "%s/msglog/%s/%s", spool_directory, message_subdir,
3562 message_id);
3563 fd = Uopen(spool_name, O_WRONLY|O_APPEND|O_CREAT, SPOOL_MODE);
3564
3565 if (fd < 0 && errno == ENOENT)
3566 {
3567 uschar temp[16];
3568 sprintf(CS temp, "msglog/%s", message_subdir);
3569 if (message_subdir[0] == 0) temp[6] = 0;
3570 (void)directory_make(spool_directory, temp, MSGLOG_DIRECTORY_MODE, TRUE);
3571 fd = Uopen(spool_name, O_WRONLY|O_APPEND|O_CREAT, SPOOL_MODE);
3572 }
3573
3574 if (fd < 0)
3575 {
3576 log_write(0, LOG_MAIN|LOG_PANIC, "Couldn't open message log %s: %s",
3577 spool_name, strerror(errno));
3578 }
3579
3580 else
3581 {
3582 FILE *message_log = fdopen(fd, "a");
3583 if (message_log == NULL)
3584 {
3585 log_write(0, LOG_MAIN|LOG_PANIC, "Couldn't fdopen message log %s: %s",
3586 spool_name, strerror(errno));
f1e894f3 3587 (void)close(fd);
059ec3d9
PH
3588 }
3589 else
3590 {
3591 uschar *now = tod_stamp(tod_log);
3592 fprintf(message_log, "%s Received from %s\n", now, s+3);
3593 if (deliver_freeze) fprintf(message_log, "%s frozen by %s\n", now,
3594 frozen_by);
3595 if (queue_only_policy) fprintf(message_log,
3596 "%s no immediate delivery: queued by %s\n", now, queued_by);
f1e894f3 3597 (void)fclose(message_log);
059ec3d9
PH
3598 }
3599 }
3600 }
3601
58eb016e
PH
3602/* Everything has now been done for a successful message except logging its
3603arrival, and outputting an SMTP response. While writing to the log, set a flag
3604to cause a call to receive_bomb_out() if the log cannot be opened. */
3605
3606receive_call_bombout = TRUE;
3607
563b63fa
PH
3608/* Before sending an SMTP response in a TCP/IP session, we check to see if the
3609connection has gone away. This can only be done if there is no unconsumed input
3610waiting in the local input buffer. We can test for this by calling
3611receive_smtp_buffered(). RFC 2920 (pipelining) explicitly allows for additional
3612input to be sent following the final dot, so the presence of following input is
3613not an error.
58eb016e 3614
563b63fa
PH
3615If the connection is still present, but there is no unread input for the
3616socket, the result of a select() call will be zero. If, however, the connection
3617has gone away, or if there is pending input, the result of select() will be
3618non-zero. The two cases can be distinguished by trying to read the next input
3619character. If we succeed, we can unread it so that it remains in the local
3620buffer for handling later. If not, the connection has been lost.
58eb016e 3621
563b63fa
PH
3622Of course, since TCP/IP is asynchronous, there is always a chance that the
3623connection will vanish between the time of this test and the sending of the
3624response, but the chance of this happening should be small. */
3625
3626if (smtp_input && sender_host_address != NULL && !sender_host_notsocket &&
3627 !receive_smtp_buffered())
58eb016e
PH
3628 {
3629 struct timeval tv;
3630 fd_set select_check;
3631 FD_ZERO(&select_check);
3632 FD_SET(fileno(smtp_in), &select_check);
3633 tv.tv_sec = 0;
3634 tv.tv_usec = 0;
3635
563b63fa 3636 if (select(fileno(smtp_in) + 1, &select_check, NULL, NULL, &tv) != 0)
58eb016e 3637 {
80a47a2c
TK
3638 int c = (receive_getc)();
3639 if (c != EOF) (receive_ungetc)(c); else
58eb016e 3640 {
563b63fa
PH
3641 uschar *msg = US"SMTP connection lost after final dot";
3642 smtp_reply = US""; /* No attempt to send a response */
3643 smtp_yield = FALSE; /* Nothing more on this connection */
58eb016e 3644
563b63fa 3645 /* Re-use the log line workspace */
58eb016e 3646
563b63fa
PH
3647 sptr = 0;
3648 s = string_cat(s, &size, &sptr, msg, Ustrlen(msg));
3649 s = add_host_info_for_log(s, &size, &sptr);
3650 s[sptr] = 0;
3651 log_write(0, LOG_MAIN, "%s", s);
58eb016e 3652
563b63fa 3653 /* Delete the files for this aborted message. */
58eb016e 3654
563b63fa
PH
3655 sprintf(CS spool_name, "%s/input/%s/%s-D", spool_directory,
3656 message_subdir, message_id);
3657 Uunlink(spool_name);
58eb016e 3658
563b63fa
PH
3659 sprintf(CS spool_name, "%s/input/%s/%s-H", spool_directory,
3660 message_subdir, message_id);
3661 Uunlink(spool_name);
58eb016e 3662
563b63fa
PH
3663 sprintf(CS spool_name, "%s/msglog/%s/%s", spool_directory,
3664 message_subdir, message_id);
3665 Uunlink(spool_name);
58eb016e 3666
563b63fa
PH
3667 goto TIDYUP;
3668 }
58eb016e
PH
3669 }
3670 }
3671
3672/* The connection has not gone away; we really are going to take responsibility
3673for this message. */
3674
3675log_write(0, LOG_MAIN |
3676 (((log_extra_selector & LX_received_recipients) != 0)? LOG_RECIPIENTS : 0) |
3677 (((log_extra_selector & LX_received_sender) != 0)? LOG_SENDER : 0),
3678 "%s", s);
3679receive_call_bombout = FALSE;
3680
3681/* Log any control actions taken by an ACL or local_scan(). */
3682
3683if (deliver_freeze) log_write(0, LOG_MAIN, "frozen by %s", frozen_by);
3684if (queue_only_policy) log_write(L_delay_delivery, LOG_MAIN,
3685 "no immediate delivery: queued by %s", queued_by);
3686
059ec3d9
PH
3687store_reset(s); /* The store for the main log message can be reused */
3688
3689/* If the message is frozen, and freeze_tell is set, do the telling. */
3690
3691if (deliver_freeze && freeze_tell != NULL && freeze_tell[0] != 0)
3692 {
3693 moan_tell_someone(freeze_tell, NULL, US"Message frozen on arrival",
3694 "Message %s was frozen on arrival by %s.\nThe sender is <%s>.\n",
3695 message_id, frozen_by, sender_address);
3696 }
3697
3698
3699/* Either a message has been successfully received and written to the two spool
3700files, or an error in writing the spool has occurred for an SMTP message, or
58eb016e
PH
3701an SMTP message has been rejected for policy reasons. (For a non-SMTP message
3702we will have already given up because there's no point in carrying on!) In
3703either event, we must now close (and thereby unlock) the data file. In the
059ec3d9
PH
3704successful case, this leaves the message on the spool, ready for delivery. In
3705the error case, the spool file will be deleted. Then tidy up store, interact
3706with an SMTP call if necessary, and return.
3707
3708A fflush() was done earlier in the expectation that any write errors on the
3709data file will be flushed(!) out thereby. Nevertheless, it is theoretically
3710possible for fclose() to fail - but what to do? What has happened to the lock
3711if this happens? */
3712
3713TIDYUP:
f1e894f3
PH
3714process_info[process_info_len] = 0; /* Remove message id */
3715if (data_file != NULL) (void)fclose(data_file); /* Frees the lock */
059ec3d9
PH
3716
3717/* Now reset signal handlers to their defaults */
3718
3719signal(SIGTERM, SIG_DFL);
3720signal(SIGINT, SIG_DFL);
3721
3722/* Tell an SMTP caller the state of play, and arrange to return the SMTP return
3723value, which defaults TRUE - meaning there may be more incoming messages from
3724this connection. For non-SMTP callers (where there is only ever one message),
3725the default is FALSE. */
3726
3727if (smtp_input)
3728 {
3729 yield = smtp_yield;
3730
3731 /* Handle interactive SMTP callers. After several kinds of error, smtp_reply
58eb016e
PH
3732 is set to the response that should be sent. When it is NULL, we generate
3733 default responses. After an ACL error or local_scan() error, the response has
3734 already been sent, and smtp_reply is an empty string to indicate this. */
059ec3d9
PH
3735
3736 if (!smtp_batched_input)
3737 {
3738 if (smtp_reply == NULL)
3739 {
29aba418 3740 if (fake_response != OK)
a5bd321b
PH
3741 smtp_respond((fake_response == DEFER)? US"450" : US"550", 3, TRUE,
3742 fake_response_text);
4e88a19f
PH
3743
3744 /* An OK response is required; use "message" text if present. */
3745
3746 else if (user_msg != NULL)
3747 {
3748 uschar *code = US"250";
3749 int len = 3;
3750 smtp_message_code(&code, &len, &user_msg, NULL);
3751 smtp_respond(code, len, TRUE, user_msg);
3752 }
3753
3754 /* Default OK response */
3755
8e669ac1
PH
3756 else
3757 smtp_printf("250 OK id=%s\r\n", message_id);
059ec3d9
PH
3758 if (host_checking)
3759 fprintf(stdout,
3760 "\n**** SMTP testing: that is not a real message id!\n\n");
3761 }
4e88a19f 3762
58eb016e 3763 /* smtp_reply is set non-empty */
4e88a19f 3764
8523533c
TK
3765 else if (smtp_reply[0] != 0)
3766 {
29aba418 3767 if (fake_response != OK && (smtp_reply[0] == '2'))
a5bd321b
PH
3768 smtp_respond((fake_response == DEFER)? US"450" : US"550", 3, TRUE,
3769 fake_response_text);
8e669ac1 3770 else
6951ac6c
PH
3771 smtp_printf("%.1024s\r\n", smtp_reply);
3772 }
059ec3d9
PH
3773 }
3774
3775 /* For batched SMTP, generate an error message on failure, and do
3776 nothing on success. The function moan_smtp_batch() does not return -
3777 it exits from the program with a non-zero return code. */
3778
3779 else if (smtp_reply != NULL) moan_smtp_batch(NULL, "%s", smtp_reply);
3780 }
3781
3782
3783/* If blackholing, we can immediately log this message's sad fate. The data
3784file has already been unlinked, and the header file was never written to disk.
3785We must now indicate that nothing was received, to prevent a delivery from
3786starting. */
3787
3788if (blackholed_by != NULL)
3789 {
3790 uschar *detail = (local_scan_data != NULL)?
3791 string_printing(local_scan_data) :
3792 string_sprintf("(%s discarded recipients)", blackholed_by);
04f7d5b9 3793 log_write(0, LOG_MAIN, "=> blackhole %s%s", detail, blackhole_log_msg);
059ec3d9
PH
3794 log_write(0, LOG_MAIN, "Completed");
3795 message_id[0] = 0;
3796 }
3797
3798/* Reset headers so that logging of rejects for a subsequent message doesn't
3799include them. It is also important to set header_last = NULL before exiting
3800from this function, as this prevents certain rewrites that might happen during
3801subsequent verifying (of another incoming message) from trying to add headers
3802when they shouldn't. */
3803
3804header_list = header_last = NULL;
3805
3806return yield; /* TRUE if more messages (SMTP only) */
3807}
3808
3809/* End of receive.c */