Reduce number of places knowing about filename-construction for mbox file-for-scanning
[exim.git] / src / src / receive.c
CommitLineData
059ec3d9
PH
1/*************************************************
2* Exim - an Internet mail transport agent *
3*************************************************/
4
d4e5e70b 5/* Copyright (c) University of Cambridge 1995 - 2017 */
059ec3d9
PH
6/* See the file NOTICE for conditions of use and distribution. */
7
8/* Code for receiving a message and setting up spool files. */
9
059ec3d9
PH
10#include "exim.h"
11
6a8f9482
TK
12#ifdef EXPERIMENTAL_DCC
13extern int dcc_ok;
14#endif
15
4840604e 16#ifdef EXPERIMENTAL_DMARC
c007c974 17# include "dmarc.h"
4840604e
TL
18#endif /* EXPERIMENTAL_DMARC */
19
059ec3d9
PH
20/*************************************************
21* Local static variables *
22*************************************************/
23
24static FILE *data_file = NULL;
25static int data_fd = -1;
41313d92 26static uschar *spool_name = US"";
059ec3d9 27
cff70eb1 28enum CH_STATE {LF_SEEN, MID_LINE, CR_SEEN};
059ec3d9
PH
29
30
31/*************************************************
32* Non-SMTP character reading functions *
33*************************************************/
34
35/* These are the default functions that are set up in the variables such as
36receive_getc initially. They just call the standard functions, passing stdin as
37the file. (When SMTP input is occurring, different functions are used by
38changing the pointer variables.) */
39
40int
bd8fbe36 41stdin_getc(unsigned lim)
059ec3d9
PH
42{
43return getc(stdin);
44}
45
46int
47stdin_ungetc(int c)
48{
49return ungetc(c, stdin);
50}
51
52int
53stdin_feof(void)
54{
55return feof(stdin);
56}
57
58int
59stdin_ferror(void)
60{
61return ferror(stdin);
62}
63
64
65
66
67/*************************************************
68* Check that a set sender is allowed *
69*************************************************/
70
71/* This function is called when a local caller sets an explicit sender address.
72It checks whether this is permitted, which it is for trusted callers.
73Otherwise, it must match the pattern(s) in untrusted_set_sender.
74
75Arguments: the proposed sender address
76Returns: TRUE for a trusted caller
77 TRUE if the address has been set, untrusted_set_sender has been
78 set, and the address matches something in the list
79 FALSE otherwise
80*/
81
82BOOL
83receive_check_set_sender(uschar *newsender)
84{
85uschar *qnewsender;
86if (trusted_caller) return TRUE;
36d295f1
JH
87if (!newsender || !untrusted_set_sender) return FALSE;
88qnewsender = Ustrchr(newsender, '@')
89 ? newsender : string_sprintf("%s@%s", newsender, qualify_domain_sender);
90return match_address_list_basic(qnewsender, CUSS &untrusted_set_sender, 0) == OK;
059ec3d9
PH
91}
92
93
94
95
96/*************************************************
5cb8cbc6 97* Read space info for a partition *
059ec3d9
PH
98*************************************************/
99
8e669ac1
PH
100/* This function is called by receive_check_fs() below, and also by string
101expansion for variables such as $spool_space. The field names for the statvfs
5cb8cbc6
PH
102structure are macros, because not all OS have F_FAVAIL and it seems tidier to
103have macros for F_BAVAIL and F_FILES as well. Some kinds of file system do not
104have inodes, and they return -1 for the number available.
059ec3d9 105
5cb8cbc6
PH
106Later: It turns out that some file systems that do not have the concept of
107inodes return 0 rather than -1. Such systems should also return 0 for the total
8e669ac1 108number of inodes, so we require that to be greater than zero before returning
5cb8cbc6 109an inode count.
059ec3d9 110
5cb8cbc6
PH
111Arguments:
112 isspool TRUE for spool partition, FALSE for log partition
113 inodeptr address of int to receive inode count; -1 if there isn't one
8e669ac1 114
5cb8cbc6 115Returns: available on-root space, in kilobytes
8e669ac1
PH
116 -1 for log partition if there isn't one
117
118All values are -1 if the STATFS functions are not available.
059ec3d9
PH
119*/
120
8e669ac1 121int
5cb8cbc6 122receive_statvfs(BOOL isspool, int *inodeptr)
059ec3d9
PH
123{
124#ifdef HAVE_STATFS
059ec3d9 125struct STATVFS statbuf;
ddf1b11a 126struct stat dummy;
5cb8cbc6
PH
127uschar *path;
128uschar *name;
129uschar buffer[1024];
059ec3d9 130
5cb8cbc6 131/* The spool directory must always exist. */
059ec3d9 132
5cb8cbc6 133if (isspool)
059ec3d9 134 {
8e669ac1
PH
135 path = spool_directory;
136 name = US"spool";
137 }
138
059ec3d9
PH
139/* Need to cut down the log file path to the directory, and to ignore any
140appearance of "syslog" in it. */
141
5cb8cbc6 142else
059ec3d9 143 {
059ec3d9 144 int sep = ':'; /* Not variable - outside scripts use */
55414b25 145 const uschar *p = log_file_path;
8e669ac1 146 name = US"log";
059ec3d9
PH
147
148 /* An empty log_file_path means "use the default". This is the same as an
149 empty item in a list. */
150
151 if (*p == 0) p = US":";
55414b25
JH
152 while ((path = string_nextinlist(&p, &sep, buffer, sizeof(buffer))))
153 if (Ustrcmp(path, "syslog") != 0)
154 break;
059ec3d9 155
5cb8cbc6
PH
156 if (path == NULL) /* No log files */
157 {
8e669ac1
PH
158 *inodeptr = -1;
159 return -1;
160 }
059ec3d9 161
8e669ac1
PH
162 /* An empty string means use the default, which is in the spool directory.
163 But don't just use the spool directory, as it is possible that the log
5cb8cbc6 164 subdirectory has been symbolically linked elsewhere. */
059ec3d9 165
8e669ac1 166 if (path[0] == 0)
059ec3d9 167 {
5cb8cbc6
PH
168 sprintf(CS buffer, CS"%s/log", CS spool_directory);
169 path = buffer;
8e669ac1
PH
170 }
171 else
059ec3d9 172 {
8e669ac1 173 uschar *cp;
5cb8cbc6 174 if ((cp = Ustrrchr(path, '/')) != NULL) *cp = 0;
8e669ac1 175 }
5cb8cbc6 176 }
8e669ac1 177
8f128379 178/* We now have the path; do the business */
5cb8cbc6
PH
179
180memset(&statbuf, 0, sizeof(statbuf));
181
182if (STATVFS(CS path, &statbuf) != 0)
ddf1b11a
JH
183 if (stat(CS path, &dummy) == -1 && errno == ENOENT)
184 { /* Can happen on first run after installation */
185 *inodeptr = -1;
186 return -1;
187 }
188 else
189 {
190 log_write(0, LOG_MAIN|LOG_PANIC, "cannot accept message: failed to stat "
191 "%s directory %s: %s", name, path, strerror(errno));
192 smtp_closedown(US"spool or log directory problem");
193 exim_exit(EXIT_FAILURE);
194 }
8e669ac1 195
5cb8cbc6
PH
196*inodeptr = (statbuf.F_FILES > 0)? statbuf.F_FAVAIL : -1;
197
198/* Disks are getting huge. Take care with computing the size in kilobytes. */
8e669ac1 199
5cb8cbc6
PH
200return (int)(((double)statbuf.F_BAVAIL * (double)statbuf.F_FRSIZE)/1024.0);
201
ddf1b11a 202#else
5cb8cbc6
PH
203/* Unable to find partition sizes in this environment. */
204
5cb8cbc6
PH
205*inodeptr = -1;
206return -1;
207#endif
208}
209
059ec3d9 210
059ec3d9 211
5cb8cbc6
PH
212
213/*************************************************
214* Check space on spool and log partitions *
215*************************************************/
216
217/* This function is called before accepting a message; if any thresholds are
218set, it checks them. If a message_size is supplied, it checks that there is
219enough space for that size plus the threshold - i.e. that the message won't
220reduce the space to the threshold. Not all OS have statvfs(); for those that
221don't, this function always returns TRUE. For some OS the old function and
222struct name statfs is used; that is handled by a macro, defined in exim.h.
223
224Arguments:
225 msg_size the (estimated) size of an incoming message
226
227Returns: FALSE if there isn't enough space, or if the information cannot
228 be obtained
229 TRUE if no check was done or there is enough space
230*/
231
232BOOL
233receive_check_fs(int msg_size)
234{
235int space, inodes;
236
237if (check_spool_space > 0 || msg_size > 0 || check_spool_inodes > 0)
238 {
8e669ac1
PH
239 space = receive_statvfs(TRUE, &inodes);
240
059ec3d9 241 DEBUG(D_receive)
5cb8cbc6
PH
242 debug_printf("spool directory space = %dK inodes = %d "
243 "check_space = %dK inodes = %d msg_size = %d\n",
244 space, inodes, check_spool_space, check_spool_inodes, msg_size);
8e669ac1
PH
245
246 if ((space >= 0 && space < check_spool_space) ||
5cb8cbc6 247 (inodes >= 0 && inodes < check_spool_inodes))
8e669ac1 248 {
5cb8cbc6
PH
249 log_write(0, LOG_MAIN, "spool directory space check failed: space=%d "
250 "inodes=%d", space, inodes);
059ec3d9
PH
251 return FALSE;
252 }
253 }
254
5cb8cbc6
PH
255if (check_log_space > 0 || check_log_inodes > 0)
256 {
8e669ac1
PH
257 space = receive_statvfs(FALSE, &inodes);
258
5cb8cbc6
PH
259 DEBUG(D_receive)
260 debug_printf("log directory space = %dK inodes = %d "
261 "check_space = %dK inodes = %d\n",
262 space, inodes, check_log_space, check_log_inodes);
8e669ac1
PH
263
264 if ((space >= 0 && space < check_log_space) ||
5cb8cbc6 265 (inodes >= 0 && inodes < check_log_inodes))
8e669ac1 266 {
5cb8cbc6
PH
267 log_write(0, LOG_MAIN, "log directory space check failed: space=%d "
268 "inodes=%d", space, inodes);
269 return FALSE;
270 }
8e669ac1
PH
271 }
272
059ec3d9
PH
273return TRUE;
274}
275
276
277
278/*************************************************
279* Bomb out while reading a message *
280*************************************************/
281
282/* The common case of wanting to bomb out is if a SIGTERM or SIGINT is
283received, or if there is a timeout. A rarer case might be if the log files are
284screwed up and Exim can't open them to record a message's arrival. Handling
285that case is done by setting a flag to cause the log functions to call this
286function if there is an ultimate disaster. That is why it is globally
287accessible.
288
8f128379
PH
289Arguments:
290 reason text reason to pass to the not-quit ACL
291 msg default SMTP response to give if in an SMTP session
059ec3d9
PH
292Returns: it doesn't
293*/
294
295void
8f128379 296receive_bomb_out(uschar *reason, uschar *msg)
059ec3d9 297{
ead37e6c
PP
298 static BOOL already_bombing_out;
299/* The smtp_notquit_exit() below can call ACLs which can trigger recursive
300timeouts, if someone has something slow in their quit ACL. Since the only
301things we should be doing are to close down cleanly ASAP, on the second
302pass we also close down stuff that might be opened again, before bypassing
303the ACL call and exiting. */
304
059ec3d9
PH
305/* If spool_name is set, it contains the name of the data file that is being
306written. Unlink it before closing so that it cannot be picked up by a delivery
307process. Ensure that any header file is also removed. */
308
ead37e6c 309if (spool_name[0] != '\0')
059ec3d9
PH
310 {
311 Uunlink(spool_name);
312 spool_name[Ustrlen(spool_name) - 1] = 'H';
313 Uunlink(spool_name);
ead37e6c 314 spool_name[0] = '\0';
059ec3d9
PH
315 }
316
317/* Now close the file if it is open, either as a fd or a stream. */
318
ead37e6c
PP
319if (data_file != NULL)
320 {
321 (void)fclose(data_file);
322 data_file = NULL;
323} else if (data_fd >= 0) {
324 (void)close(data_fd);
325 data_fd = -1;
326 }
059ec3d9 327
8f128379
PH
328/* Attempt to close down an SMTP connection tidily. For non-batched SMTP, call
329smtp_notquit_exit(), which runs the NOTQUIT ACL, if present, and handles the
330SMTP response. */
059ec3d9 331
ead37e6c 332if (!already_bombing_out)
059ec3d9 333 {
ead37e6c
PP
334 already_bombing_out = TRUE;
335 if (smtp_input)
336 {
337 if (smtp_batched_input)
338 moan_smtp_batch(NULL, "421 %s - message abandoned", msg); /* No return */
339 smtp_notquit_exit(reason, US"421", US"%s %s - closing connection.",
340 smtp_active_hostname, msg);
341 }
059ec3d9
PH
342 }
343
344/* Exit from the program (non-BSMTP cases) */
345
346exim_exit(EXIT_FAILURE);
347}
348
349
350/*************************************************
351* Data read timeout *
352*************************************************/
353
354/* Handler function for timeouts that occur while reading the data that
355comprises a message.
356
357Argument: the signal number
358Returns: nothing
359*/
360
361static void
362data_timeout_handler(int sig)
363{
364uschar *msg = NULL;
365
366sig = sig; /* Keep picky compilers happy */
367
368if (smtp_input)
369 {
370 msg = US"SMTP incoming data timeout";
371 log_write(L_lost_incoming_connection,
372 LOG_MAIN, "SMTP data timeout (message abandoned) on connection "
fed77020
PH
373 "from %s F=<%s>",
374 (sender_fullhost != NULL)? sender_fullhost : US"local process",
375 sender_address);
059ec3d9
PH
376 }
377else
378 {
379 fprintf(stderr, "exim: timed out while reading - message abandoned\n");
380 log_write(L_lost_incoming_connection,
381 LOG_MAIN, "timed out while reading local message");
382 }
383
8f128379 384receive_bomb_out(US"data-timeout", msg); /* Does not return */
059ec3d9
PH
385}
386
387
388
389/*************************************************
390* local_scan() timeout *
391*************************************************/
392
393/* Handler function for timeouts that occur while running a local_scan()
394function.
395
396Argument: the signal number
397Returns: nothing
398*/
399
400static void
401local_scan_timeout_handler(int sig)
402{
403sig = sig; /* Keep picky compilers happy */
404log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function timed out - "
405 "message temporarily rejected (size %d)", message_size);
8f128379
PH
406/* Does not return */
407receive_bomb_out(US"local-scan-timeout", US"local verification problem");
059ec3d9
PH
408}
409
410
411
412/*************************************************
413* local_scan() crashed *
414*************************************************/
415
416/* Handler function for signals that occur while running a local_scan()
417function.
418
419Argument: the signal number
420Returns: nothing
421*/
422
423static void
424local_scan_crash_handler(int sig)
425{
426log_write(0, LOG_MAIN|LOG_REJECT, "local_scan() function crashed with "
427 "signal %d - message temporarily rejected (size %d)", sig, message_size);
8f128379
PH
428/* Does not return */
429receive_bomb_out(US"local-scan-error", US"local verification problem");
059ec3d9
PH
430}
431
432
433/*************************************************
434* SIGTERM or SIGINT received *
435*************************************************/
436
437/* Handler for SIGTERM or SIGINT signals that occur while reading the
438data that comprises a message.
439
440Argument: the signal number
441Returns: nothing
442*/
443
444static void
445data_sigterm_sigint_handler(int sig)
446{
447uschar *msg = NULL;
448
449if (smtp_input)
450 {
451 msg = US"Service not available - SIGTERM or SIGINT received";
452 log_write(0, LOG_MAIN, "%s closed after %s", smtp_get_connection_info(),
453 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
454 }
455else
456 {
f05da2e8 457 if (filter_test == FTEST_NONE)
059ec3d9
PH
458 {
459 fprintf(stderr, "\nexim: %s received - message abandoned\n",
460 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
461 log_write(0, LOG_MAIN, "%s received while reading local message",
462 (sig == SIGTERM)? "SIGTERM" : "SIGINT");
463 }
464 }
465
8f128379 466receive_bomb_out(US"signal-exit", msg); /* Does not return */
059ec3d9
PH
467}
468
469
470
471/*************************************************
472* Add new recipient to list *
473*************************************************/
474
475/* This function builds a list of recipient addresses in argc/argv
476format.
477
478Arguments:
479 recipient the next address to add to recipients_list
480 pno parent number for fixed aliases; -1 otherwise
481
482Returns: nothing
483*/
484
485void
486receive_add_recipient(uschar *recipient, int pno)
487{
488if (recipients_count >= recipients_list_max)
489 {
490 recipient_item *oldlist = recipients_list;
491 int oldmax = recipients_list_max;
492 recipients_list_max = recipients_list_max? 2*recipients_list_max : 50;
493 recipients_list = store_get(recipients_list_max * sizeof(recipient_item));
494 if (oldlist != NULL)
495 memcpy(recipients_list, oldlist, oldmax * sizeof(recipient_item));
496 }
497
498recipients_list[recipients_count].address = recipient;
499recipients_list[recipients_count].pno = pno;
8523533c
TK
500#ifdef EXPERIMENTAL_BRIGHTMAIL
501recipients_list[recipients_count].bmi_optin = bmi_current_optin;
502/* reset optin string pointer for next recipient */
503bmi_current_optin = NULL;
504#endif
6c1c3d1d
WB
505recipients_list[recipients_count].orcpt = NULL;
506recipients_list[recipients_count].dsn_flags = 0;
059ec3d9
PH
507recipients_list[recipients_count++].errors_to = NULL;
508}
509
510
511
512
513/*************************************************
fd98a5c6
JH
514* Send user response message *
515*************************************************/
61147df4 516
fd98a5c6
JH
517/* This function is passed a default response code and a user message. It calls
518smtp_message_code() to check and possibly modify the response code, and then
519calls smtp_respond() to transmit the response. I put this into a function
520just to avoid a lot of repetition.
61147df4
PP
521
522Arguments:
fd98a5c6
JH
523 code the response code
524 user_msg the user message
525
526Returns: nothing
61147df4
PP
527*/
528
8ccd00b1 529#ifndef DISABLE_PRDR
61147df4 530static void
fd98a5c6 531smtp_user_msg(uschar *code, uschar *user_msg)
61147df4 532{
fd98a5c6 533int len = 3;
4f6ae5c3 534smtp_message_code(&code, &len, &user_msg, NULL, TRUE);
fd98a5c6 535smtp_respond(code, len, TRUE, user_msg);
61147df4
PP
536}
537#endif
538
539
540
541
fd98a5c6
JH
542
543/*************************************************
059ec3d9
PH
544* Remove a recipient from the list *
545*************************************************/
546
547/* This function is provided for local_scan() to use.
548
549Argument:
550 recipient address to remove
551
552Returns: TRUE if it did remove something; FALSE otherwise
553*/
554
555BOOL
556receive_remove_recipient(uschar *recipient)
557{
558int count;
559DEBUG(D_receive) debug_printf("receive_remove_recipient(\"%s\") called\n",
560 recipient);
561for (count = 0; count < recipients_count; count++)
562 {
563 if (Ustrcmp(recipients_list[count].address, recipient) == 0)
564 {
565 if ((--recipients_count - count) > 0)
566 memmove(recipients_list + count, recipients_list + count + 1,
54cdb463 567 (recipients_count - count)*sizeof(recipient_item));
059ec3d9
PH
568 return TRUE;
569 }
570 }
571return FALSE;
572}
573
574
575
576
577
578/*************************************************
579* Read data portion of a non-SMTP message *
580*************************************************/
581
582/* This function is called to read the remainder of a message (following the
583header) when the input is not from SMTP - we are receiving a local message on
584a standard input stream. The message is always terminated by EOF, and is also
585terminated by a dot on a line by itself if the flag dot_ends is TRUE. Split the
586two cases for maximum efficiency.
587
588Ensure that the body ends with a newline. This will naturally be the case when
589the termination is "\n.\n" but may not be otherwise. The RFC defines messages
590as "sequences of lines" - this of course strictly applies only to SMTP, but
591deliveries into BSD-type mailbox files also require it. Exim used to have a
592flag for doing this at delivery time, but as it was always set for all
593transports, I decided to simplify things by putting the check here instead.
594
595There is at least one MUA (dtmail) that sends CRLF via this interface, and
596other programs are known to do this as well. Exim used to have a option for
597dealing with this: in July 2003, after much discussion, the code has been
598changed to default to treat any of LF, CRLF, and bare CR as line terminators.
599
600However, for the case when a dot on a line by itself terminates a message, the
601only recognized terminating sequences before and after the dot are LF and CRLF.
602Otherwise, having read EOL . CR, you don't know whether to read another
603character or not.
604
605Internally, in messages stored in Exim's spool files, LF is used as the line
606terminator. Under the new regime, bare CRs will no longer appear in these
607files.
608
609Arguments:
610 fout a FILE to which to write the message
611
612Returns: One of the END_xxx values indicating why it stopped reading
613*/
614
615static int
616read_message_data(FILE *fout)
617{
618int ch_state;
619register int ch;
d677b2f2 620register int linelength = 0;
059ec3d9
PH
621
622/* Handle the case when only EOF terminates the message */
623
624if (!dot_ends)
625 {
626 register int last_ch = '\n';
627
bd8fbe36 628 for (; (ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF; last_ch = ch)
059ec3d9
PH
629 {
630 if (ch == 0) body_zerocount++;
631 if (last_ch == '\r' && ch != '\n')
632 {
d677b2f2
PH
633 if (linelength > max_received_linelength)
634 max_received_linelength = linelength;
635 linelength = 0;
059ec3d9
PH
636 if (fputc('\n', fout) == EOF) return END_WERROR;
637 message_size++;
638 body_linecount++;
639 }
640 if (ch == '\r') continue;
641
642 if (fputc(ch, fout) == EOF) return END_WERROR;
d677b2f2
PH
643 if (ch == '\n')
644 {
645 if (linelength > max_received_linelength)
646 max_received_linelength = linelength;
647 linelength = 0;
648 body_linecount++;
649 }
650 else linelength++;
059ec3d9
PH
651 if (++message_size > thismessage_size_limit) return END_SIZE;
652 }
653
654 if (last_ch != '\n')
655 {
d677b2f2
PH
656 if (linelength > max_received_linelength)
657 max_received_linelength = linelength;
059ec3d9
PH
658 if (fputc('\n', fout) == EOF) return END_WERROR;
659 message_size++;
660 body_linecount++;
661 }
662
663 return END_EOF;
664 }
665
666/* Handle the case when a dot on a line on its own, or EOF, terminates. */
667
668ch_state = 1;
669
bd8fbe36 670while ((ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF)
059ec3d9
PH
671 {
672 if (ch == 0) body_zerocount++;
673 switch (ch_state)
674 {
675 case 0: /* Normal state (previous char written) */
676 if (ch == '\n')
d677b2f2
PH
677 {
678 body_linecount++;
679 if (linelength > max_received_linelength)
680 max_received_linelength = linelength;
681 linelength = -1;
682 ch_state = 1;
683 }
059ec3d9
PH
684 else if (ch == '\r')
685 { ch_state = 2; continue; }
686 break;
687
688 case 1: /* After written "\n" */
689 if (ch == '.') { ch_state = 3; continue; }
6eb02f88 690 if (ch == '\r') { ch_state = 2; continue; }
3581f321
JH
691 if (ch == '\n') { body_linecount++; linelength = -1; }
692 else ch_state = 0;
059ec3d9
PH
693 break;
694
695 case 2:
696 body_linecount++; /* After unwritten "\r" */
d677b2f2
PH
697 if (linelength > max_received_linelength)
698 max_received_linelength = linelength;
059ec3d9 699 if (ch == '\n')
d677b2f2
PH
700 {
701 ch_state = 1;
702 linelength = -1;
703 }
059ec3d9
PH
704 else
705 {
706 if (message_size++, fputc('\n', fout) == EOF) return END_WERROR;
707 if (ch == '\r') continue;
708 ch_state = 0;
d677b2f2 709 linelength = 0;
059ec3d9
PH
710 }
711 break;
712
713 case 3: /* After "\n." (\n written, dot not) */
714 if (ch == '\n') return END_DOT;
715 if (ch == '\r') { ch_state = 4; continue; }
716 message_size++;
d677b2f2 717 linelength++;
059ec3d9
PH
718 if (fputc('.', fout) == EOF) return END_WERROR;
719 ch_state = 0;
720 break;
721
722 case 4: /* After "\n.\r" (\n written, rest not) */
723 if (ch == '\n') return END_DOT;
724 message_size += 2;
725 body_linecount++;
726 if (fputs(".\n", fout) == EOF) return END_WERROR;
727 if (ch == '\r') { ch_state = 2; continue; }
728 ch_state = 0;
729 break;
730 }
731
d677b2f2 732 linelength++;
059ec3d9
PH
733 if (fputc(ch, fout) == EOF) return END_WERROR;
734 if (++message_size > thismessage_size_limit) return END_SIZE;
735 }
736
737/* Get here if EOF read. Unless we have just written "\n", we need to ensure
738the message ends with a newline, and we must also write any characters that
739were saved up while testing for an ending dot. */
740
741if (ch_state != 1)
742 {
743 static uschar *ends[] = { US"\n", NULL, US"\n", US".\n", US".\n" };
744 if (fputs(CS ends[ch_state], fout) == EOF) return END_WERROR;
745 message_size += Ustrlen(ends[ch_state]);
746 body_linecount++;
747 }
748
749return END_EOF;
750}
751
752
753
754
755/*************************************************
756* Read data portion of an SMTP message *
757*************************************************/
758
759/* This function is called to read the remainder of an SMTP message (after the
760headers), or to skip over it when an error has occurred. In this case, the
761output file is passed as NULL.
762
763If any line begins with a dot, that character is skipped. The input should only
764be successfully terminated by CR LF . CR LF unless it is local (non-network)
765SMTP, in which case the CRs are optional, but...
766
767FUDGE: It seems that sites on the net send out messages with just LF
768terminators, despite the warnings in the RFCs, and other MTAs handle this. So
769we make the CRs optional in all cases.
770
771July 2003: Bare CRs cause trouble. We now treat them as line terminators as
772well, so that there are no CRs in spooled messages. However, the message
773terminating dot is not recognized between two bare CRs.
774
775Arguments:
776 fout a FILE to which to write the message; NULL if skipping
777
778Returns: One of the END_xxx values indicating why it stopped reading
779*/
780
781static int
782read_message_data_smtp(FILE *fout)
783{
784int ch_state = 0;
e4bdf652 785int ch;
7e3ce68e 786int linelength = 0;
059ec3d9 787
bd8fbe36 788while ((ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF)
059ec3d9
PH
789 {
790 if (ch == 0) body_zerocount++;
791 switch (ch_state)
792 {
793 case 0: /* After LF or CRLF */
794 if (ch == '.')
795 {
796 ch_state = 3;
797 continue; /* Don't ever write . after LF */
798 }
799 ch_state = 1;
800
801 /* Else fall through to handle as normal uschar. */
802
803 case 1: /* Normal state */
804 if (ch == '\n')
805 {
806 ch_state = 0;
807 body_linecount++;
1f5497b2
PH
808 if (linelength > max_received_linelength)
809 max_received_linelength = linelength;
810 linelength = -1;
059ec3d9
PH
811 }
812 else if (ch == '\r')
813 {
814 ch_state = 2;
815 continue;
816 }
817 break;
818
819 case 2: /* After (unwritten) CR */
820 body_linecount++;
1f5497b2
PH
821 if (linelength > max_received_linelength)
822 max_received_linelength = linelength;
823 linelength = -1;
059ec3d9
PH
824 if (ch == '\n')
825 {
826 ch_state = 0;
827 }
828 else
829 {
830 message_size++;
831 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
6851a9c5 832 cutthrough_data_put_nl();
059ec3d9
PH
833 if (ch != '\r') ch_state = 1; else continue;
834 }
835 break;
836
837 case 3: /* After [CR] LF . */
838 if (ch == '\n')
839 return END_DOT;
840 if (ch == '\r')
841 {
842 ch_state = 4;
843 continue;
844 }
1bc460a6
JH
845 /* The dot was removed at state 3. For a doubled dot, here, reinstate
846 it to cutthrough. The current ch, dot or not, is passed both to cutthrough
847 and to file below. */
848 if (ch == '.')
849 {
850 uschar c= ch;
6851a9c5 851 cutthrough_data_puts(&c, 1);
1bc460a6
JH
852 }
853 ch_state = 1;
059ec3d9
PH
854 break;
855
856 case 4: /* After [CR] LF . CR */
857 if (ch == '\n') return END_DOT;
858 message_size++;
859 body_linecount++;
860 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
6851a9c5 861 cutthrough_data_put_nl();
059ec3d9
PH
862 if (ch == '\r')
863 {
864 ch_state = 2;
865 continue;
866 }
867 ch_state = 1;
868 break;
869 }
870
871 /* Add the character to the spool file, unless skipping; then loop for the
872 next. */
873
874 message_size++;
1f5497b2 875 linelength++;
7e3ce68e 876 if (fout)
059ec3d9
PH
877 {
878 if (fputc(ch, fout) == EOF) return END_WERROR;
879 if (message_size > thismessage_size_limit) return END_SIZE;
880 }
e4bdf652 881 if(ch == '\n')
6851a9c5 882 cutthrough_data_put_nl();
e4bdf652
JH
883 else
884 {
7e3ce68e 885 uschar c = ch;
6851a9c5 886 cutthrough_data_puts(&c, 1);
e4bdf652 887 }
059ec3d9
PH
888 }
889
890/* Fall through here if EOF encountered. This indicates some kind of error,
891since a correct message is terminated by [CR] LF . [CR] LF. */
892
893return END_EOF;
894}
895
896
897
898
7e3ce68e 899/* Variant of the above read_message_data_smtp() specialised for RFC 3030
1ebe15c3
JH
900CHUNKING. Accept input lines separated by either CRLF or CR or LF and write
901LF-delimited spoolfile. Until we have wireformat spoolfiles, we need the
902body_linecount accounting for proper re-expansion for the wire, so use
903a cut-down version of the state-machine above; we don't need to do leading-dot
904detection and unstuffing.
7e3ce68e
JH
905
906Arguments:
544dd905
PP
907 fout a FILE to which to write the message; NULL if skipping;
908 must be open for both writing and reading.
7e3ce68e
JH
909
910Returns: One of the END_xxx values indicating why it stopped reading
911*/
912
913static int
914read_message_bdat_smtp(FILE *fout)
915{
cff70eb1
HSHR
916int linelength = 0, ch;
917enum CH_STATE ch_state = LF_SEEN;
d953610f 918BOOL fix_nl = FALSE;
7e3ce68e 919
1ebe15c3 920for(;;)
7e3ce68e 921 {
1ebe15c3
JH
922 switch ((ch = (bdat_getc)(GETC_BUFFER_UNLIMITED)))
923 {
924 case EOF: return END_EOF;
1ebe15c3 925 case ERR: return END_PROTOCOL;
d953610f
HSHR
926 case EOD:
927 /* Nothing to get from the sender anymore. We check the last
928 character written to the spool.
929
930 RFC 3030 states, that BDAT chunks are normal text, terminated by CRLF.
931 If we would be strict, we would refuse such broken messages.
932 But we are liberal, so we fix it. It would be easy just to append
933 the "\n" to the spool.
934
935 But there are some more things (line counting, message size calculation and such),
936 that would need to be duplicated here. So we simply do some ungetc
937 trickery.
938 */
59d98039
JH
939 if (fout)
940 {
941 if (fseek(fout, -1, SEEK_CUR) < 0) return END_PROTOCOL;
942 if (fgetc(fout) == '\n') return END_DOT;
943 }
d953610f
HSHR
944
945 if (linelength == -1) /* \r already seen (see below) */
946 {
947 DEBUG(D_receive) debug_printf("Add missing LF\n");
948 bdat_ungetc('\n');
949 continue;
950 }
951 DEBUG(D_receive) debug_printf("Add missing CRLF\n");
952 bdat_ungetc('\r'); /* not even \r was seen */
953 fix_nl = TRUE;
954
955 continue;
1ebe15c3
JH
956 case '\0': body_zerocount++; break;
957 }
958 switch (ch_state)
959 {
cff70eb1
HSHR
960 case LF_SEEN: /* After LF or CRLF */
961 ch_state = MID_LINE;
1ebe15c3 962 /* fall through to handle as normal uschar. */
7e3ce68e 963
cff70eb1 964 case MID_LINE: /* Mid-line state */
1ebe15c3
JH
965 if (ch == '\n')
966 {
cff70eb1 967 ch_state = LF_SEEN;
1ebe15c3
JH
968 body_linecount++;
969 if (linelength > max_received_linelength)
970 max_received_linelength = linelength;
971 linelength = -1;
972 }
973 else if (ch == '\r')
974 {
cff70eb1 975 ch_state = CR_SEEN;
d953610f 976 if (fix_nl) bdat_ungetc('\n');
1ebe15c3
JH
977 continue; /* don't write CR */
978 }
979 break;
7e3ce68e 980
cff70eb1 981 case CR_SEEN: /* After (unwritten) CR */
1ebe15c3
JH
982 body_linecount++;
983 if (linelength > max_received_linelength)
984 max_received_linelength = linelength;
985 linelength = -1;
986 if (ch == '\n')
cff70eb1 987 ch_state = LF_SEEN;
1ebe15c3
JH
988 else
989 {
990 message_size++;
59d98039 991 if (fout && fputc('\n', fout) == EOF) return END_WERROR;
6851a9c5 992 cutthrough_data_put_nl();
1ebe15c3 993 if (ch == '\r') continue; /* don't write CR */
cff70eb1 994 ch_state = MID_LINE;
1ebe15c3
JH
995 }
996 break;
997 }
998
999 /* Add the character to the spool file, unless skipping */
1000
1001 message_size++;
1002 linelength++;
1003 if (fout)
1004 {
1005 if (fputc(ch, fout) == EOF) return END_WERROR;
1006 if (message_size > thismessage_size_limit) return END_SIZE;
1007 }
1008 if(ch == '\n')
6851a9c5 1009 cutthrough_data_put_nl();
1ebe15c3
JH
1010 else
1011 {
1012 uschar c = ch;
6851a9c5 1013 cutthrough_data_puts(&c, 1);
1ebe15c3 1014 }
7e3ce68e
JH
1015 }
1016/*NOTREACHED*/
1017}
1018
1019
1020
1021
059ec3d9
PH
1022/*************************************************
1023* Swallow SMTP message *
1024*************************************************/
1025
1026/* This function is called when there has been some kind of error while reading
1027an SMTP message, and the remaining data may need to be swallowed. It is global
1028because it is called from smtp_closedown() to shut down an incoming call
1029tidily.
1030
1031Argument: a FILE from which to read the message
1032Returns: nothing
1033*/
1034
1035void
1036receive_swallow_smtp(void)
1037{
7e3ce68e 1038/*XXX CHUNKING: not enough. read chunks until RSET? */
059ec3d9
PH
1039if (message_ended >= END_NOTENDED)
1040 message_ended = read_message_data_smtp(NULL);
1041}
1042
1043
1044
1045/*************************************************
1046* Handle lost SMTP connection *
1047*************************************************/
1048
1049/* This function logs connection loss incidents and generates an appropriate
1050SMTP response.
1051
1052Argument: additional data for the message
1053Returns: the SMTP response
1054*/
1055
1056static uschar *
1057handle_lost_connection(uschar *s)
1058{
1059log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
1060 "%s lost while reading message data%s", smtp_get_connection_info(), s);
eea0defe 1061smtp_notquit_exit(US"connection-lost", NULL, NULL);
059ec3d9
PH
1062return US"421 Lost incoming connection";
1063}
1064
1065
1066
1067
1068/*************************************************
1069* Handle a non-smtp reception error *
1070*************************************************/
1071
1072/* This function is called for various errors during the reception of non-SMTP
1073messages. It either sends a message to the sender of the problem message, or it
1074writes to the standard error stream.
1075
1076Arguments:
1077 errcode code for moan_to_sender(), identifying the error
1078 text1 first message text, passed to moan_to_sender()
1079 text2 second message text, used only for stderrr
1080 error_rc code to pass to exim_exit if no problem
1081 f FILE containing body of message (may be stdin)
1082 hptr pointer to instore headers or NULL
1083
1084Returns: calls exim_exit(), which does not return
1085*/
1086
1087static void
1088give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
1089 FILE *f, header_line *hptr)
1090{
1091if (error_handling == ERRORS_SENDER)
1092 {
1093 error_block eblock;
1094 eblock.next = NULL;
1095 eblock.text1 = text1;
37f3dc43 1096 eblock.text2 = US"";
059ec3d9
PH
1097 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
1098 error_rc = EXIT_FAILURE;
1099 }
37f3dc43
JH
1100else
1101 fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 1102(void)fclose(f);
059ec3d9
PH
1103exim_exit(error_rc);
1104}
1105
1106
1107
1108/*************************************************
1109* Add header lines set up by ACL *
1110*************************************************/
1111
850635b6
PH
1112/* This function is called to add the header lines that were set up by
1113statements in an ACL to the list of headers in memory. It is done in two stages
1114like this, because when the ACL for RCPT is running, the other headers have not
1115yet been received. This function is called twice; once just before running the
1116DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
1117are visible to the DATA ACL.
059ec3d9
PH
1118
1119Originally these header lines were added at the end. Now there is support for
1120three different places: top, bottom, and after the Received: header(s). There
1121will always be at least one Received: header, even if it is marked deleted, and
1122even if something else has been put in front of it.
1123
1124Arguments:
1125 acl_name text to identify which ACL
1126
1127Returns: nothing
1128*/
1129
1130static void
578d43dc 1131add_acl_headers(int where, uschar *acl_name)
059ec3d9
PH
1132{
1133header_line *h, *next;
1134header_line *last_received = NULL;
e7568d51 1135
578d43dc
JH
1136switch(where)
1137 {
1138 case ACL_WHERE_DKIM:
1139 case ACL_WHERE_MIME:
af4a1bca 1140 case ACL_WHERE_DATA:
57cc2785
JH
1141 if ( cutthrough.fd >= 0 && cutthrough.delivery
1142 && (acl_removed_headers || acl_added_headers))
578d43dc
JH
1143 {
1144 log_write(0, LOG_MAIN|LOG_PANIC, "Header modification in data ACLs"
af4a1bca 1145 " will not take effect on cutthrough deliveries");
578d43dc
JH
1146 return;
1147 }
1148 }
1149
57cc2785 1150if (acl_removed_headers)
e7568d51 1151 {
e1d04f48 1152 DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers removed by %s ACL:\n", acl_name);
e7568d51 1153
57cc2785 1154 for (h = header_list; h; h = h->next) if (h->type != htype_old)
e7568d51 1155 {
55414b25 1156 const uschar * list = acl_removed_headers;
e7568d51
TL
1157 int sep = ':'; /* This is specified as a colon-separated list */
1158 uschar *s;
1159 uschar buffer[128];
4a142059
JH
1160
1161 while ((s = string_nextinlist(&list, &sep, buffer, sizeof(buffer))))
1162 if (header_testname(h, s, Ustrlen(s), FALSE))
e7568d51
TL
1163 {
1164 h->type = htype_old;
e1d04f48 1165 DEBUG(D_receive|D_acl) debug_printf_indent(" %s", h->text);
e7568d51 1166 }
e7568d51
TL
1167 }
1168 acl_removed_headers = NULL;
e1d04f48 1169 DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
e7568d51 1170 }
059ec3d9 1171
57cc2785 1172if (!acl_added_headers) return;
e1d04f48 1173DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers added by %s ACL:\n", acl_name);
059ec3d9 1174
57cc2785 1175for (h = acl_added_headers; h; h = next)
059ec3d9
PH
1176 {
1177 next = h->next;
1178
1179 switch(h->type)
1180 {
1181 case htype_add_top:
1182 h->next = header_list;
1183 header_list = h;
e1d04f48 1184 DEBUG(D_receive|D_acl) debug_printf_indent(" (at top)");
059ec3d9
PH
1185 break;
1186
1187 case htype_add_rec:
1188 if (last_received == NULL)
1189 {
1190 last_received = header_list;
1191 while (!header_testname(last_received, US"Received", 8, FALSE))
1192 last_received = last_received->next;
1193 while (last_received->next != NULL &&
1194 header_testname(last_received->next, US"Received", 8, FALSE))
1195 last_received = last_received->next;
1196 }
1197 h->next = last_received->next;
1198 last_received->next = h;
e1d04f48 1199 DEBUG(D_receive|D_acl) debug_printf_indent(" (after Received:)");
059ec3d9
PH
1200 break;
1201
8523533c
TK
1202 case htype_add_rfc:
1203 /* add header before any header which is NOT Received: or Resent- */
1204 last_received = header_list;
1205 while ( (last_received->next != NULL) &&
1206 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
1207 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
1208 last_received = last_received->next;
1209 /* last_received now points to the last Received: or Resent-* header
1210 in an uninterrupted chain of those header types (seen from the beginning
1211 of all headers. Our current header must follow it. */
1212 h->next = last_received->next;
1213 last_received->next = h;
e1d04f48 1214 DEBUG(D_receive|D_acl) debug_printf_indent(" (before any non-Received: or Resent-*: header)");
8523533c
TK
1215 break;
1216
059ec3d9
PH
1217 default:
1218 h->next = NULL;
1219 header_last->next = h;
1220 break;
1221 }
1222
1223 if (h->next == NULL) header_last = h;
1224
1225 /* Check for one of the known header types (From:, To:, etc.) though in
1226 practice most added headers are going to be "other". Lower case
1227 identification letters are never stored with the header; they are used
1228 for existence tests when messages are received. So discard any lower case
1229 flag values. */
1230
1231 h->type = header_checkname(h, FALSE);
1232 if (h->type >= 'a') h->type = htype_other;
1233
e1d04f48 1234 DEBUG(D_receive|D_acl) debug_printf_indent(" %s", header_last->text);
059ec3d9
PH
1235 }
1236
71fafd95 1237acl_added_headers = NULL;
e1d04f48 1238DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
059ec3d9
PH
1239}
1240
1241
1242
1243/*************************************************
1244* Add host information for log line *
1245*************************************************/
1246
1247/* Called for acceptance and rejecting log lines. This adds information about
1248the calling host to a string that is being built dynamically.
1249
1250Arguments:
1251 s the dynamic string
1252 sizeptr points to the size variable
1253 ptrptr points to the pointer variable
1254
1255Returns: the extended string
1256*/
1257
1258static uschar *
fc16abb4 1259add_host_info_for_log(uschar * s, int * sizeptr, int * ptrptr)
059ec3d9 1260{
fc16abb4 1261if (sender_fullhost)
059ec3d9 1262 {
fc16abb4
JH
1263 if (LOGGING(dnssec) && sender_host_dnssec) /*XXX sender_helo_dnssec? */
1264 s = string_cat(s, sizeptr, ptrptr, US" DS");
059ec3d9 1265 s = string_append(s, sizeptr, ptrptr, 2, US" H=", sender_fullhost);
6c6d6e48 1266 if (LOGGING(incoming_interface) && interface_address != NULL)
059ec3d9 1267 {
fc16abb4
JH
1268 s = string_cat(s, sizeptr, ptrptr,
1269 string_sprintf(" I=[%s]:%d", interface_address, interface_port));
059ec3d9
PH
1270 }
1271 }
1272if (sender_ident != NULL)
1273 s = string_append(s, sizeptr, ptrptr, 2, US" U=", sender_ident);
1274if (received_protocol != NULL)
1275 s = string_append(s, sizeptr, ptrptr, 2, US" P=", received_protocol);
1276return s;
1277}
1278
1279
1280
63955bf2 1281#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1282
1283/*************************************************
54cdb463
PH
1284* Run the MIME ACL on a message *
1285*************************************************/
1286
1287/* This code is in a subroutine so that it can be used for both SMTP
1288and non-SMTP messages. It is called with a non-NULL ACL pointer.
1289
1290Arguments:
1291 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1292 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1293 smtp_reply_ptr Where SMTP reply is being built
1294 blackholed_by_ptr Where "blackholed by" message is being built
1295
1296Returns: TRUE to carry on; FALSE to abandon the message
1297*/
1298
1299static BOOL
1300run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1301 uschar **blackholed_by_ptr)
1302{
1303FILE *mbox_file;
1304uschar rfc822_file_path[2048];
1305unsigned long mbox_size;
1306header_line *my_headerlist;
1307uschar *user_msg, *log_msg;
1308int mime_part_count_buffer = -1;
040721f2 1309uschar * mbox_filename;
7156b1ef 1310int rc = OK;
54cdb463
PH
1311
1312memset(CS rfc822_file_path,0,2048);
1313
1314/* check if it is a MIME message */
040721f2
JH
1315
1316for (my_headerlist = header_list; my_headerlist; my_headerlist = my_headerlist->next)
1317 if ( my_headerlist->type != '*' /* skip deleted headers */
1318 && strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0
1319 )
4e88a19f 1320 {
54cdb463
PH
1321 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1322 goto DO_MIME_ACL;
4e88a19f 1323 }
54cdb463
PH
1324
1325DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1326return TRUE;
1327
1328DO_MIME_ACL:
040721f2 1329
54cdb463 1330/* make sure the eml mbox file is spooled up */
040721f2
JH
1331if (!(mbox_file = spool_mbox(&mbox_size, NULL, &mbox_filename)))
1332 { /* error while spooling */
54cdb463
PH
1333 log_write(0, LOG_MAIN|LOG_PANIC,
1334 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1335 Uunlink(spool_name);
1336 unspool_mbox();
6f0c431a
PP
1337#ifdef EXPERIMENTAL_DCC
1338 dcc_ok = 0;
1339#endif
a5bd321b 1340 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1341 message_id[0] = 0; /* Indicate no message accepted */
1342 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1343 return FALSE; /* Indicate skip to end of receive function */
040721f2 1344 }
54cdb463
PH
1345
1346mime_is_rfc822 = 0;
1347
1348MIME_ACL_CHECK:
1349mime_part_count = -1;
1350rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1351(void)fclose(mbox_file);
54cdb463 1352
4e88a19f
PH
1353if (Ustrlen(rfc822_file_path) > 0)
1354 {
54cdb463
PH
1355 mime_part_count = mime_part_count_buffer;
1356
4e88a19f
PH
1357 if (unlink(CS rfc822_file_path) == -1)
1358 {
54cdb463
PH
1359 log_write(0, LOG_PANIC,
1360 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1361 goto END_MIME_ACL;
4e88a19f
PH
1362 }
1363 }
54cdb463
PH
1364
1365/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1366if (rc == OK)
1367 {
040721f2 1368 uschar * scandir;
e8bc7fca
JH
1369 struct dirent * entry;
1370 DIR * tempdir;
54cdb463 1371
040721f2 1372 scandir = string_copyn(mbox_filename, Ustrrchr(mbox_filename, '/') - mbox_filename);
54cdb463 1373
040721f2 1374 tempdir = opendir(CS scandir);
e8bc7fca 1375 for (;;)
4e88a19f 1376 {
e8bc7fca
JH
1377 if (!(entry = readdir(tempdir)))
1378 break;
1379 if (strncmpic(US entry->d_name, US"__rfc822_", 9) == 0)
4e88a19f 1380 {
e8bc7fca 1381 (void) string_format(rfc822_file_path, sizeof(rfc822_file_path),
040721f2 1382 "%s/%s", scandir, entry->d_name);
e1d04f48 1383 DEBUG(D_receive) debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
e8bc7fca 1384 rfc822_file_path);
4e88a19f
PH
1385 break;
1386 }
e8bc7fca 1387 }
4e88a19f 1388 closedir(tempdir);
54cdb463 1389
e8bc7fca 1390 if (entry)
4e88a19f 1391 {
e8bc7fca 1392 if ((mbox_file = Ufopen(rfc822_file_path, "rb")))
4e88a19f 1393 {
e8bc7fca
JH
1394 /* set RFC822 expansion variable */
1395 mime_is_rfc822 = 1;
1396 mime_part_count_buffer = mime_part_count;
1397 goto MIME_ACL_CHECK;
4e88a19f 1398 }
e8bc7fca
JH
1399 log_write(0, LOG_PANIC,
1400 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1401 unlink(CS rfc822_file_path);
4e88a19f
PH
1402 }
1403 }
54cdb463
PH
1404
1405END_MIME_ACL:
578d43dc 1406add_acl_headers(ACL_WHERE_MIME, US"MIME");
54cdb463
PH
1407if (rc == DISCARD)
1408 {
1409 recipients_count = 0;
1410 *blackholed_by_ptr = US"MIME ACL";
1411 }
1412else if (rc != OK)
1413 {
1414 Uunlink(spool_name);
1415 unspool_mbox();
6f0c431a
PP
1416#ifdef EXPERIMENTAL_DCC
1417 dcc_ok = 0;
1418#endif
4f6ae5c3
JH
1419 if ( smtp_input
1420 && smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1421 {
85ffcba6 1422 *smtp_yield_ptr = FALSE; /* No more messages after dropped connection */
f4c1088b 1423 *smtp_reply_ptr = US""; /* Indicate reply already sent */
4f6ae5c3 1424 }
54cdb463
PH
1425 message_id[0] = 0; /* Indicate no message accepted */
1426 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1427 }
54cdb463
PH
1428
1429return TRUE;
1430}
1431
63955bf2 1432#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1433
1434
e4bdf652
JH
1435
1436void
1437received_header_gen(void)
1438{
1439uschar *received;
1440uschar *timestamp;
1441header_line *received_header= header_list;
1442
1443timestamp = expand_string(US"${tod_full}");
1444if (recipients_count == 1) received_for = recipients_list[0].address;
1445received = expand_string(received_header_text);
1446received_for = NULL;
1447
d4ff61d1 1448if (!received)
e4bdf652
JH
1449 {
1450 if(spool_name[0] != 0)
1451 Uunlink(spool_name); /* Lose the data file */
1452 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1453 "(received_header_text) failed: %s", string_printing(received_header_text),
1454 expand_string_message);
1455 }
1456
1457/* The first element on the header chain is reserved for the Received header,
1458so all we have to do is fill in the text pointer, and set the type. However, if
1459the result of the expansion is an empty string, we leave the header marked as
1460"old" so as to refrain from adding a Received header. */
1461
1462if (received[0] == 0)
1463 {
1464 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1465 received_header->type = htype_old;
1466 }
1467else
1468 {
1469 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1470 received_header->type = htype_received;
1471 }
1472
1473received_header->slen = Ustrlen(received_header->text);
1474
1475DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1476 received_header->type, received_header->text);
1477}
1478
1479
1480
54cdb463 1481/*************************************************
059ec3d9
PH
1482* Receive message *
1483*************************************************/
1484
1485/* Receive a message on the given input, and put it into a pair of spool files.
1486Either a non-null list of recipients, or the extract flag will be true, or
1487both. The flag sender_local is true for locally generated messages. The flag
1488submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1489suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1490suppress_local_fixups" or -G was passed on the command-line.
1491The flag smtp_input is true if the message is to be
8800895a
PH
1492handled using SMTP conventions about termination and lines starting with dots.
1493For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1494
1495If a message was successfully read, message_id[0] will be non-zero.
1496
1497The general actions of this function are:
1498
1499 . Read the headers of the message (if any) into a chain of store
1500 blocks.
1501
1502 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1503 throw it away, unless the caller is trusted, or unless
1504 active_local_sender_retain is set - which can only happen if
1505 active_local_from_check is false.
059ec3d9
PH
1506
1507 . If recipients are to be extracted from the message, build the
1508 recipients list from the headers, removing any that were on the
1509 original recipients list (unless extract_addresses_remove_arguments is
1510 false), and at the same time, remove any bcc header that may be present.
1511
1512 . Get the spool file for the data, sort out its unique name, open
1513 and lock it (but don't give it the name yet).
1514
1515 . Generate a "Message-Id" header if the message doesn't have one, for
1516 locally-originated messages.
1517
1518 . Generate a "Received" header.
1519
1520 . Ensure the recipients list is fully qualified and rewritten if necessary.
1521
1522 . If there are any rewriting rules, apply them to the sender address
1523 and also to the headers.
1524
1525 . If there is no from: header, generate one, for locally-generated messages
1526 and messages in "submission mode" only.
1527
1528 . If the sender is local, check that from: is correct, and if not, generate
1529 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1530 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1531
1532 . If there is no "date" header, generate one, for locally-originated
1533 or submission mode messages only.
1534
1535 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1536 dot_ends mode, to the data file. Leave it open, to hold the lock.
1537
1538 . Write the envelope and the headers to a new file.
1539
1540 . Set the name for the header file; close it.
1541
1542 . Set the name for the data file; close it.
1543
1544Because this function can potentially be called many times in a single
1545SMTP connection, all store should be got by store_get(), so that it will be
1546automatically retrieved after the message is accepted.
1547
1548FUDGE: It seems that sites on the net send out messages with just LF
1549terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1550we make the CRs optional in all cases.
1551
1552July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1553new regime is now in place in which bare CRs in header lines are turned into LF
1554followed by a space, so as not to terminate the header line.
1555
1556February 2004: A bare LF in a header line in a message whose first line was
1557terminated by CRLF is treated in the same way as a bare CR.
1558
1559Arguments:
1560 extract_recip TRUE if recipients are to be extracted from the message's
1561 headers
1562
1563Returns: TRUE there are more messages to be read (SMTP input)
1564 FALSE there are no more messages to be read (non-SMTP input
1565 or SMTP connection collapsed, or other failure)
1566
1567When reading a message for filter testing, the returned value indicates
1568whether the headers (which is all that is read) were terminated by '.' or
1569not. */
1570
1571BOOL
1572receive_msg(BOOL extract_recip)
1573{
7156b1ef
NM
1574int i;
1575int rc = FAIL;
059ec3d9
PH
1576int msg_size = 0;
1577int process_info_len = Ustrlen(process_info);
1578int error_rc = (error_handling == ERRORS_SENDER)?
1579 errors_sender_rc : EXIT_FAILURE;
1580int header_size = 256;
1581int start, end, domain, size, sptr;
1582int id_resolution;
1583int had_zero = 0;
d677b2f2 1584int prevlines_length = 0;
059ec3d9
PH
1585
1586register int ptr = 0;
1587
1588BOOL contains_resent_headers = FALSE;
1589BOOL extracted_ignored = FALSE;
1590BOOL first_line_ended_crlf = TRUE_UNSET;
1591BOOL smtp_yield = TRUE;
1592BOOL yield = FALSE;
1593
1594BOOL resents_exist = FALSE;
1595uschar *resent_prefix = US"";
1596uschar *blackholed_by = NULL;
04f7d5b9 1597uschar *blackhole_log_msg = US"";
c5430c20 1598enum {NOT_TRIED, TMP_REJ, PERM_REJ, ACCEPTED} cutthrough_done = NOT_TRIED;
059ec3d9
PH
1599
1600flock_t lock_data;
1601error_block *bad_addresses = NULL;
1602
1603uschar *frozen_by = NULL;
1604uschar *queued_by = NULL;
1605
1606uschar *errmsg, *s;
1607struct stat statbuf;
1608
4e88a19f 1609/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1610
1611uschar *smtp_reply = NULL;
4e88a19f 1612uschar *user_msg, *log_msg;
059ec3d9
PH
1613
1614/* Working header pointers */
1615
1616header_line *h, *next;
1617
2cbb4081 1618/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1619
1620BOOL date_header_exists = FALSE;
1621
1622/* Pointers to receive the addresses of headers whose contents we need. */
1623
1624header_line *from_header = NULL;
1625header_line *subject_header = NULL;
1626header_line *msgid_header = NULL;
1627header_line *received_header;
1628
4840604e
TL
1629#ifdef EXPERIMENTAL_DMARC
1630int dmarc_up = 0;
1631#endif /* EXPERIMENTAL_DMARC */
1632
059ec3d9
PH
1633/* Variables for use when building the Received: header. */
1634
059ec3d9
PH
1635uschar *timestamp;
1636int tslen;
1637
1638/* Release any open files that might have been cached while preparing to
1639accept the message - e.g. by verifying addresses - because reading a message
1640might take a fair bit of real time. */
1641
1642search_tidyup();
1643
e4bdf652
JH
1644/* Extracting the recipient list from an input file is incompatible with
1645cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1646to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1647if (extract_recip || !smtp_input)
57cc2785 1648 cancel_cutthrough_connection(TRUE, US"not smtp input");
e4bdf652 1649
059ec3d9
PH
1650/* Initialize the chain of headers by setting up a place-holder for Received:
1651header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1652pointing to the end of the chain to make adding headers simple. */
1653
1654received_header = header_list = header_last = store_get(sizeof(header_line));
1655header_list->next = NULL;
1656header_list->type = htype_old;
1657header_list->text = NULL;
1658header_list->slen = 0;
1659
1660/* Control block for the next header to be read. */
1661
1662next = store_get(sizeof(header_line));
1663next->text = store_get(header_size);
1664
1665/* Initialize message id to be null (indicating no message read), and the
1666header names list to be the normal list. Indicate there is no data file open
1667yet, initialize the size and warning count, and deal with no size limit. */
1668
1669message_id[0] = 0;
1670data_file = NULL;
1671data_fd = -1;
41313d92 1672spool_name = US"";
059ec3d9
PH
1673message_size = 0;
1674warning_count = 0;
d677b2f2 1675received_count = 1; /* For the one we will add */
059ec3d9
PH
1676
1677if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1678
2e0c1448 1679/* While reading the message, the following counts are computed. */
059ec3d9 1680
d677b2f2
PH
1681message_linecount = body_linecount = body_zerocount =
1682 max_received_linelength = 0;
059ec3d9 1683
80a47a2c 1684#ifndef DISABLE_DKIM
e983e85a
JH
1685/* Call into DKIM to set up the context. In CHUNKING mode
1686we clear the dot-stuffing flag */
1687if (smtp_input && !smtp_batched_input && !dkim_disable_verify)
1688 dkim_exim_verify_init(chunking_state <= CHUNKING_OFFERED);
fb2274d4
TK
1689#endif
1690
4840604e
TL
1691#ifdef EXPERIMENTAL_DMARC
1692/* initialize libopendmarc */
1693dmarc_up = dmarc_init();
1694#endif
1695
059ec3d9
PH
1696/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1697ids, and fractions of a second are required. See the comments that precede the
1698message id creation below. */
1699
1700(void)gettimeofday(&message_id_tv, NULL);
1701
1702/* For other uses of the received time we can operate with granularity of one
1703second, and for that we use the global variable received_time. This is for
1704things like ultimate message timeouts. */
1705
1706received_time = message_id_tv.tv_sec;
1707
1708/* If SMTP input, set the special handler for timeouts. The alarm() calls
1709happen in the smtp_getc() function when it refills its buffer. */
1710
1711if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1712
1713/* If not SMTP input, timeout happens only if configured, and we just set a
1714single timeout for the whole message. */
1715
1716else if (receive_timeout > 0)
1717 {
1718 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1719 alarm(receive_timeout);
1720 }
1721
1722/* SIGTERM and SIGINT are caught always. */
1723
1724signal(SIGTERM, data_sigterm_sigint_handler);
1725signal(SIGINT, data_sigterm_sigint_handler);
1726
1727/* Header lines in messages are not supposed to be very long, though when
1728unfolded, to: and cc: headers can take up a lot of store. We must also cope
1729with the possibility of junk being thrown at us. Start by getting 256 bytes for
1730storing the header, and extend this as necessary using string_cat().
1731
1732To cope with total lunacies, impose an upper limit on the length of the header
1733section of the message, as otherwise the store will fill up. We must also cope
1734with the possibility of binary zeros in the data. Hence we cannot use fgets().
1735Folded header lines are joined into one string, leaving the '\n' characters
1736inside them, so that writing them out reproduces the input.
1737
1738Loop for each character of each header; the next structure for chaining the
1739header is set up already, with ptr the offset of the next character in
1740next->text. */
1741
1742for (;;)
1743 {
bd8fbe36 1744 int ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1745
1746 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1747 SMTP must have a correct "." terminator. */
1748
1749 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1750 {
1751 smtp_reply = handle_lost_connection(US" (header)");
1752 smtp_yield = FALSE;
1753 goto TIDYUP; /* Skip to end of function */
1754 }
1755
1756 /* See if we are at the current header's size limit - there must be at least
1757 four bytes left. This allows for the new character plus a zero, plus two for
1758 extra insertions when we are playing games with dots and carriage returns. If
1759 we are at the limit, extend the text buffer. This could have been done
1760 automatically using string_cat() but because this is a tightish loop storing
1761 only one character at a time, we choose to do it inline. Normally
1762 store_extend() will be able to extend the block; only at the end of a big
1763 store block will a copy be needed. To handle the case of very long headers
1764 (and sometimes lunatic messages can have ones that are 100s of K long) we
1765 call store_release() for strings that have been copied - if the string is at
1766 the start of a block (and therefore the only thing in it, because we aren't
1767 doing any other gets), the block gets freed. We can only do this because we
1768 know there are no other calls to store_get() going on. */
1769
1770 if (ptr >= header_size - 4)
1771 {
1772 int oldsize = header_size;
1773 /* header_size += 256; */
1774 header_size *= 2;
1775 if (!store_extend(next->text, oldsize, header_size))
1776 {
1777 uschar *newtext = store_get(header_size);
1778 memcpy(newtext, next->text, ptr);
1779 store_release(next->text);
1780 next->text = newtext;
1781 }
1782 }
1783
1784 /* Cope with receiving a binary zero. There is dispute about whether
1785 these should be allowed in RFC 822 messages. The middle view is that they
1786 should not be allowed in headers, at least. Exim takes this attitude at
1787 the moment. We can't just stomp on them here, because we don't know that
1788 this line is a header yet. Set a flag to cause scanning later. */
1789
1790 if (ch == 0) had_zero++;
1791
1792 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1793 those from data files use just LF. Treat LF in local SMTP input as a
1794 terminator too. Treat EOF as a line terminator always. */
1795
1796 if (ch == EOF) goto EOL;
1797
1798 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1799 other MTAs accept this. We are therefore forced into this "liberalisation"
1800 too, so we accept LF as a line terminator whatever the source of the message.
1801 However, if the first line of the message ended with a CRLF, we treat a bare
1802 LF specially by inserting a white space after it to ensure that the header
1803 line is not terminated. */
1804
1805 if (ch == '\n')
1806 {
1807 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1808 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1809 goto EOL;
1810 }
1811
1812 /* This is not the end of the line. If this is SMTP input and this is
1813 the first character in the line and it is a "." character, ignore it.
1814 This implements the dot-doubling rule, though header lines starting with
1815 dots aren't exactly common. They are legal in RFC 822, though. If the
1816 following is CRLF or LF, this is the line that that terminates the
1817 entire message. We set message_ended to indicate this has happened (to
1818 prevent further reading), and break out of the loop, having freed the
1819 empty header, and set next = NULL to indicate no data line. */
1820
1821 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1822 {
bd8fbe36 1823 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1824 if (ch == '\r')
1825 {
bd8fbe36 1826 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1827 if (ch != '\n')
1828 {
80a47a2c 1829 receive_ungetc(ch);
059ec3d9
PH
1830 ch = '\r'; /* Revert to CR */
1831 }
1832 }
1833 if (ch == '\n')
1834 {
1835 message_ended = END_DOT;
1836 store_reset(next);
1837 next = NULL;
1838 break; /* End character-reading loop */
1839 }
1840
1841 /* For non-SMTP input, the dot at the start of the line was really a data
1842 character. What is now in ch is the following character. We guaranteed
1843 enough space for this above. */
1844
1845 if (!smtp_input)
1846 {
1847 next->text[ptr++] = '.';
1848 message_size++;
1849 }
1850 }
1851
1852 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1853 remember this case if this is the first line ending. */
1854
1855 if (ch == '\r')
1856 {
bd8fbe36 1857 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1858 if (ch == '\n')
1859 {
1860 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1861 goto EOL;
1862 }
1863
1864 /* Otherwise, put back the character after CR, and turn the bare CR
1865 into LF SP. */
1866
80a47a2c 1867 ch = (receive_ungetc)(ch);
059ec3d9
PH
1868 next->text[ptr++] = '\n';
1869 message_size++;
1870 ch = ' ';
1871 }
1872
1873 /* We have a data character for the header line. */
1874
1875 next->text[ptr++] = ch; /* Add to buffer */
1876 message_size++; /* Total message size so far */
1877
1878 /* Handle failure due to a humungously long header section. The >= allows
1879 for the terminating \n. Add what we have so far onto the headers list so
1880 that it gets reflected in any error message, and back up the just-read
1881 character. */
1882
1883 if (message_size >= header_maxsize)
1884 {
1885 next->text[ptr] = 0;
1886 next->slen = ptr;
1887 next->type = htype_other;
1888 next->next = NULL;
1889 header_last->next = next;
1890 header_last = next;
1891
1892 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1893 "%s (more than %d characters): message abandoned",
1894 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1895
1896 if (smtp_input)
1897 {
1898 smtp_reply = US"552 Message header is ridiculously long";
1899 receive_swallow_smtp();
1900 goto TIDYUP; /* Skip to end of function */
1901 }
1902
1903 else
1904 {
1905 give_local_error(ERRMESS_VLONGHEADER,
1906 string_sprintf("message header longer than %d characters received: "
1907 "message not accepted", header_maxsize), US"", error_rc, stdin,
1908 header_list->next);
1909 /* Does not return */
1910 }
1911 }
1912
1913 continue; /* With next input character */
1914
1915 /* End of header line reached */
1916
1917 EOL:
2e0c1448
PH
1918
1919 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1920
1921 receive_linecount++;
1922 message_linecount++;
059ec3d9 1923
d677b2f2
PH
1924 /* Keep track of maximum line length */
1925
1926 if (ptr - prevlines_length > max_received_linelength)
1927 max_received_linelength = ptr - prevlines_length;
1928 prevlines_length = ptr + 1;
1929
059ec3d9
PH
1930 /* Now put in the terminating newline. There is always space for
1931 at least two more characters. */
1932
1933 next->text[ptr++] = '\n';
1934 message_size++;
1935
1936 /* A blank line signals the end of the headers; release the unwanted
1937 space and set next to NULL to indicate this. */
1938
1939 if (ptr == 1)
1940 {
1941 store_reset(next);
1942 next = NULL;
1943 break;
1944 }
1945
1946 /* There is data in the line; see if the next input character is a
1947 whitespace character. If it is, we have a continuation of this header line.
1948 There is always space for at least one character at this point. */
1949
1950 if (ch != EOF)
1951 {
bd8fbe36 1952 int nextch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1953 if (nextch == ' ' || nextch == '\t')
1954 {
1955 next->text[ptr++] = nextch;
1956 message_size++;
1957 continue; /* Iterate the loop */
1958 }
80a47a2c 1959 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
1960 else ch = EOF; /* Cause main loop to exit at end */
1961 }
1962
1963 /* We have got to the real line end. Terminate the string and release store
1964 beyond it. If it turns out to be a real header, internal binary zeros will
1965 be squashed later. */
1966
1967 next->text[ptr] = 0;
1968 next->slen = ptr;
1969 store_reset(next->text + ptr + 1);
1970
1971 /* Check the running total size against the overall message size limit. We
1972 don't expect to fail here, but if the overall limit is set less than MESSAGE_
1973 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
1974 headers - the code to read the body will then also hit the buffer. */
1975
1976 if (message_size > thismessage_size_limit) break;
1977
1978 /* A line that is not syntactically correct for a header also marks
1979 the end of the headers. In this case, we leave next containing the
1980 first data line. This might actually be several lines because of the
1981 continuation logic applied above, but that doesn't matter.
1982
1983 It turns out that smail, and presumably sendmail, accept leading lines
1984 of the form
1985
1986 From ph10 Fri Jan 5 12:35 GMT 1996
1987
1988 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
1989 find any documentation of this, but for compatibility it had better be
1990 accepted. Exim restricts it to the case of non-smtp messages, and
1991 treats it as an alternative to the -f command line option. Thus it is
1992 ignored except for trusted users or filter testing. Otherwise it is taken
1993 as the sender address, unless -f was used (sendmail compatibility).
1994
1995 It further turns out that some UUCPs generate the From_line in a different
1996 format, e.g.
1997
1998 From ph10 Fri, 7 Jan 97 14:00:00 GMT
1999
2000 The regex for matching these things is now capable of recognizing both
2001 formats (including 2- and 4-digit years in the latter). In fact, the regex
2002 is now configurable, as is the expansion string to fish out the sender.
2003
2004 Even further on it has been discovered that some broken clients send
2005 these lines in SMTP messages. There is now an option to ignore them from
2006 specified hosts or networks. Sigh. */
2007
2008 if (header_last == header_list &&
2009 (!smtp_input
2010 ||
2011 (sender_host_address != NULL &&
2012 verify_check_host(&ignore_fromline_hosts) == OK)
2013 ||
2014 (sender_host_address == NULL && ignore_fromline_local)
2015 ) &&
2016 regex_match_and_setup(regex_From, next->text, 0, -1))
2017 {
2018 if (!sender_address_forced)
2019 {
2020 uschar *uucp_sender = expand_string(uucp_from_sender);
2021 if (uucp_sender == NULL)
2022 {
2023 log_write(0, LOG_MAIN|LOG_PANIC,
2024 "expansion of \"%s\" failed after matching "
2025 "\"From \" line: %s", uucp_from_sender, expand_string_message);
2026 }
2027 else
2028 {
2029 int start, end, domain;
2030 uschar *errmess;
2031 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
2032 &start, &end, &domain, TRUE);
2033 if (newsender != NULL)
2034 {
2035 if (domain == 0 && newsender[0] != 0)
2036 newsender = rewrite_address_qualify(newsender, FALSE);
2037
f05da2e8 2038 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
2039 {
2040 sender_address = newsender;
2041
f05da2e8 2042 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
2043 {
2044 authenticated_sender = NULL;
2045 originator_name = US"";
2046 sender_local = FALSE;
2047 }
2048
f05da2e8 2049 if (filter_test != FTEST_NONE)
059ec3d9
PH
2050 printf("Sender taken from \"From \" line\n");
2051 }
2052 }
2053 }
2054 }
2055 }
2056
2057 /* Not a leading "From " line. Check to see if it is a valid header line.
2058 Header names may contain any non-control characters except space and colon,
2059 amazingly. */
2060
2061 else
2062 {
2063 uschar *p = next->text;
2064
2065 /* If not a valid header line, break from the header reading loop, leaving
2066 next != NULL, indicating that it holds the first line of the body. */
2067
2068 if (isspace(*p)) break;
2069 while (mac_isgraph(*p) && *p != ':') p++;
2070 while (isspace(*p)) p++;
2071 if (*p != ':')
2072 {
2073 body_zerocount = had_zero;
2074 break;
2075 }
2076
2077 /* We have a valid header line. If there were any binary zeroes in
2078 the line, stomp on them here. */
2079
2080 if (had_zero > 0)
2081 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
2082
2083 /* It is perfectly legal to have an empty continuation line
2084 at the end of a header, but it is confusing to humans
2085 looking at such messages, since it looks like a blank line.
2086 Reduce confusion by removing redundant white space at the
2087 end. We know that there is at least one printing character
2088 (the ':' tested for above) so there is no danger of running
2089 off the end. */
2090
2091 p = next->text + ptr - 2;
2092 for (;;)
2093 {
2094 while (*p == ' ' || *p == '\t') p--;
2095 if (*p != '\n') break;
2096 ptr = (p--) - next->text + 1;
2097 message_size -= next->slen - ptr;
2098 next->text[ptr] = 0;
2099 next->slen = ptr;
2100 }
2101
2102 /* Add the header to the chain */
2103
2104 next->type = htype_other;
2105 next->next = NULL;
2106 header_last->next = next;
2107 header_last = next;
2108
2109 /* Check the limit for individual line lengths. This comes after adding to
2110 the chain so that the failing line is reflected if a bounce is generated
2111 (for a local message). */
2112
2113 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
2114 {
2115 log_write(0, LOG_MAIN, "overlong message header line received from "
2116 "%s (more than %d characters): message abandoned",
2117 sender_host_unknown? sender_ident : sender_fullhost,
2118 header_line_maxsize);
2119
2120 if (smtp_input)
2121 {
2122 smtp_reply = US"552 A message header line is too long";
2123 receive_swallow_smtp();
2124 goto TIDYUP; /* Skip to end of function */
2125 }
2126
2127 else
2128 {
2129 give_local_error(ERRMESS_VLONGHDRLINE,
2130 string_sprintf("message header line longer than %d characters "
2131 "received: message not accepted", header_line_maxsize), US"",
2132 error_rc, stdin, header_list->next);
2133 /* Does not return */
2134 }
2135 }
2136
2137 /* Note if any resent- fields exist. */
2138
2139 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
2140 {
2141 resents_exist = TRUE;
2142 resent_prefix = US"Resent-";
2143 }
2144 }
2145
1ebe15c3
JH
2146 /* Reject CHUNKING messages that do not CRLF their first header line */
2147
2148 if (!first_line_ended_crlf && chunking_state > CHUNKING_OFFERED)
2149 {
2150 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2151 "Non-CRLF-terminated header, under CHUNKING: message abandoned",
2152 sender_address,
2153 sender_fullhost ? " H=" : "", sender_fullhost ? sender_fullhost : US"",
2154 sender_ident ? " U=" : "", sender_ident ? sender_ident : US"");
2155 smtp_printf("552 Message header not CRLF terminated\r\n");
2156 bdat_flush_data();
2157 smtp_reply = US"";
2158 goto TIDYUP; /* Skip to end of function */
2159 }
2160
059ec3d9
PH
2161 /* The line has been handled. If we have hit EOF, break out of the loop,
2162 indicating no pending data line. */
2163
2164 if (ch == EOF) { next = NULL; break; }
2165
2166 /* Set up for the next header */
2167
2168 header_size = 256;
2169 next = store_get(sizeof(header_line));
2170 next->text = store_get(header_size);
2171 ptr = 0;
2172 had_zero = 0;
d677b2f2 2173 prevlines_length = 0;
059ec3d9
PH
2174 } /* Continue, starting to read the next header */
2175
2176/* At this point, we have read all the headers into a data structure in main
2177store. The first header is still the dummy placeholder for the Received: header
2178we are going to generate a bit later on. If next != NULL, it contains the first
2179data line - which terminated the headers before reaching a blank line (not the
2180normal case). */
2181
2182DEBUG(D_receive)
2183 {
2184 debug_printf(">>Headers received:\n");
1ebe15c3 2185 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2186 debug_printf("%s", h->text);
2187 debug_printf("\n");
2188 }
2189
2190/* End of file on any SMTP connection is an error. If an incoming SMTP call
2191is dropped immediately after valid headers, the next thing we will see is EOF.
2192We must test for this specially, as further down the reading of the data is
2193skipped if already at EOF. */
2194
2195if (smtp_input && (receive_feof)())
2196 {
2197 smtp_reply = handle_lost_connection(US" (after header)");
2198 smtp_yield = FALSE;
2199 goto TIDYUP; /* Skip to end of function */
2200 }
2201
2202/* If this is a filter test run and no headers were read, output a warning
2203in case there is a mistake in the test message. */
2204
f05da2e8 2205if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2206 printf("Warning: no message headers read\n");
2207
2208
2209/* Scan the headers to identify them. Some are merely marked for later
2210processing; some are dealt with here. */
2211
1ebe15c3 2212for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2213 {
2214 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2215 if (is_resent) contains_resent_headers = TRUE;
2216
2217 switch (header_checkname(h, is_resent))
2218 {
059ec3d9 2219 case htype_bcc:
2cbb4081 2220 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
2221 break;
2222
059ec3d9 2223 case htype_cc:
2cbb4081 2224 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
2225 break;
2226
2227 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
2228
2229 case htype_date:
4c69d561 2230 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
2231 break;
2232
2233 /* Same comments as about Return-Path: below. */
2234
2235 case htype_delivery_date:
2236 if (delivery_date_remove) h->type = htype_old;
2237 break;
2238
2239 /* Same comments as about Return-Path: below. */
2240
2241 case htype_envelope_to:
2242 if (envelope_to_remove) h->type = htype_old;
2243 break;
2244
2245 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2246 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2247 header consists of just the login id of the user who called Exim, rewrite
2248 it with the gecos field first. Apply this rule to Resent-From: if there
2249 are resent- fields. */
2250
2251 case htype_from:
2252 h->type = htype_from;
2253 if (!resents_exist || is_resent)
2254 {
2255 from_header = h;
2256 if (!smtp_input)
2257 {
5de8faa3 2258 int len;
059ec3d9
PH
2259 uschar *s = Ustrchr(h->text, ':') + 1;
2260 while (isspace(*s)) s++;
5de8faa3 2261 len = h->slen - (s - h->text) - 1;
e0fccd1d
TF
2262 if (Ustrlen(originator_login) == len &&
2263 strncmpic(s, originator_login, len) == 0)
059ec3d9
PH
2264 {
2265 uschar *name = is_resent? US"Resent-From" : US"From";
2266 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2267 originator_login, qualify_domain_sender);
2268 from_header = header_last;
2269 h->type = htype_old;
2270 DEBUG(D_receive|D_rewrite)
2271 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2272 }
2273 }
2274 }
2275 break;
2276
2277 /* Identify the Message-id: header for generating "in-reply-to" in the
2278 autoreply transport. For incoming logging, save any resent- value. In both
2279 cases, take just the first of any multiples. */
2280
2281 case htype_id:
2282 if (msgid_header == NULL && (!resents_exist || is_resent))
2283 {
2284 msgid_header = h;
2285 h->type = htype_id;
2286 }
2287 break;
2288
2289 /* Flag all Received: headers */
2290
2291 case htype_received:
2292 h->type = htype_received;
2293 received_count++;
2294 break;
2295
2296 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
2297
2298 case htype_reply_to:
2299 h->type = htype_reply_to;
2300 break;
2301
2302 /* The Return-path: header is supposed to be added to messages when
2303 they leave the SMTP system. We shouldn't receive messages that already
2304 contain Return-path. However, since Exim generates Return-path: on
2305 local delivery, resent messages may well contain it. We therefore
2306 provide an option (which defaults on) to remove any Return-path: headers
2307 on input. Removal actually means flagging as "old", which prevents the
2308 header being transmitted with the message. */
2309
2310 case htype_return_path:
2311 if (return_path_remove) h->type = htype_old;
2312
2313 /* If we are testing a mail filter file, use the value of the
2314 Return-Path: header to set up the return_path variable, which is not
2315 otherwise set. However, remove any <> that surround the address
2316 because the variable doesn't have these. */
2317
f05da2e8 2318 if (filter_test != FTEST_NONE)
059ec3d9
PH
2319 {
2320 uschar *start = h->text + 12;
2321 uschar *end = start + Ustrlen(start);
2322 while (isspace(*start)) start++;
2323 while (end > start && isspace(end[-1])) end--;
2324 if (*start == '<' && end[-1] == '>')
2325 {
2326 start++;
2327 end--;
2328 }
2329 return_path = string_copyn(start, end - start);
2330 printf("Return-path taken from \"Return-path:\" header line\n");
2331 }
2332 break;
2333
2334 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2335 and from an untrusted caller and suppress_local_fixups is not set, or if we
2336 are in submission mode for a remote message, mark it "old" so that it will
2337 not be transmitted with the message, unless active_local_sender_retain is
2338 set. (This can only be true if active_local_from_check is false.) If there
2339 are any resent- headers in the message, apply this rule to Resent-Sender:
2340 instead of Sender:. Messages with multiple resent- header sets cannot be
2341 tidily handled. (For this reason, at least one MUA - Pine - turns old
2342 resent- headers into X-resent- headers when resending, leaving just one
2343 set.) */
059ec3d9
PH
2344
2345 case htype_sender:
69358f02 2346 h->type = ((!active_local_sender_retain &&
8800895a
PH
2347 (
2348 (sender_local && !trusted_caller && !suppress_local_fixups)
2349 || submission_mode
2350 )
059ec3d9
PH
2351 ) &&
2352 (!resents_exist||is_resent))?
2353 htype_old : htype_sender;
2354 break;
2355
2356 /* Remember the Subject: header for logging. There is no Resent-Subject */
2357
2358 case htype_subject:
2359 subject_header = h;
2360 break;
2361
2362 /* "To:" gets flagged, and the existence of a recipient header is noted,
2363 whether it's resent- or not. */
2364
2365 case htype_to:
2366 h->type = htype_to;
2367 /****
2368 to_or_cc_header_exists = TRUE;
2369 ****/
2370 break;
2371 }
2372 }
2373
2374/* Extract recipients from the headers if that is required (the -t option).
2375Note that this is documented as being done *before* any address rewriting takes
2376place. There are two possibilities:
2377
2378(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2379recipients already listed are to be REMOVED from the message. Smail 3 works
2380like this. We need to build a non-recipients tree for that list, because in
2381subsequent processing this data is held in a tree and that's what the
2382spool_write_header() function expects. Make sure that non-recipient addresses
2383are fully qualified and rewritten if necessary.
2384
2385(2) According to other sendmail documentation, -t ADDS extracted recipients to
2386those in the command line arguments (and it is rumoured some other MTAs do
2387this). Therefore, there is an option to make Exim behave this way.
2388
2389*** Notes on "Resent-" header lines ***
2390
2391The presence of resent-headers in the message makes -t horribly ambiguous.
2392Experiments with sendmail showed that it uses recipients for all resent-
2393headers, totally ignoring the concept of "sets of resent- headers" as described
2394in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2395with all the addresses in one instance of each header.
2396
2397This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2398error for -t if there were resent- headers in the message. However, after a
2399discussion on the mailing list, I've learned that there are MUAs that use
2400resent- headers with -t, and also that the stuff about sets of resent- headers
2401and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2402message with -t and resent- header lines makes sure that only *its* resent-
2403headers are present; previous ones are often renamed as X-resent- for example.
2404
2405Consequently, Exim has been changed so that, if any resent- header lines are
2406present, the recipients are taken from all of the appropriate resent- lines,
2407and not from the ordinary To:, Cc:, etc. */
2408
2409if (extract_recip)
2410 {
2411 int rcount = 0;
2412 error_block **bnext = &bad_addresses;
2413
2414 if (extract_addresses_remove_arguments)
2415 {
2416 while (recipients_count-- > 0)
2417 {
2418 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2419 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2420 tree_add_nonrecipient(s);
2421 }
2422 recipients_list = NULL;
2423 recipients_count = recipients_list_max = 0;
2424 }
2425
059ec3d9
PH
2426 /* Now scan the headers */
2427
1ebe15c3 2428 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2429 {
2430 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2431 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2432 {
2433 uschar *s = Ustrchr(h->text, ':') + 1;
2434 while (isspace(*s)) s++;
2435
1eccaa59
PH
2436 parse_allow_group = TRUE; /* Allow address group syntax */
2437
059ec3d9
PH
2438 while (*s != 0)
2439 {
2440 uschar *ss = parse_find_address_end(s, FALSE);
2441 uschar *recipient, *errmess, *p, *pp;
2442 int start, end, domain;
2443
2444 /* Check on maximum */
2445
2446 if (recipients_max > 0 && ++rcount > recipients_max)
2447 {
2448 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2449 US"message rejected: ", error_rc, stdin, NULL);
2450 /* Does not return */
2451 }
2452
2453 /* Make a copy of the address, and remove any internal newlines. These
2454 may be present as a result of continuations of the header line. The
2455 white space that follows the newline must not be removed - it is part
2456 of the header. */
2457
2458 pp = recipient = store_get(ss - s + 1);
2459 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2460 *pp = 0;
250b6871 2461
8c5d388a 2462#ifdef SUPPORT_I18N
250b6871
JH
2463 {
2464 BOOL b = allow_utf8_domains;
2465 allow_utf8_domains = TRUE;
2466#endif
059ec3d9
PH
2467 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2468 &domain, FALSE);
2469
8c5d388a 2470#ifdef SUPPORT_I18N
250b6871
JH
2471 if (string_is_utf8(recipient))
2472 message_smtputf8 = TRUE;
2473 else
2474 allow_utf8_domains = b;
2475 }
2476#endif
2477
059ec3d9
PH
2478 /* Keep a list of all the bad addresses so we can send a single
2479 error message at the end. However, an empty address is not an error;
2480 just ignore it. This can come from an empty group list like
2481
2482 To: Recipients of list:;
2483
2484 If there are no recipients at all, an error will occur later. */
2485
2486 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2487 {
2488 int len = Ustrlen(s);
2489 error_block *b = store_get(sizeof(error_block));
2490 while (len > 0 && isspace(s[len-1])) len--;
2491 b->next = NULL;
2492 b->text1 = string_printing(string_copyn(s, len));
2493 b->text2 = errmess;
2494 *bnext = b;
2495 bnext = &(b->next);
2496 }
2497
2498 /* If the recipient is already in the nonrecipients tree, it must
2499 have appeared on the command line with the option extract_addresses_
2500 remove_arguments set. Do not add it to the recipients, and keep a note
2501 that this has happened, in order to give a better error if there are
2502 no recipients left. */
2503
2504 else if (recipient != NULL)
2505 {
2506 if (tree_search(tree_nonrecipients, recipient) == NULL)
2507 receive_add_recipient(recipient, -1);
2508 else
2509 extracted_ignored = TRUE;
2510 }
2511
2512 /* Move on past this address */
2513
2514 s = ss + (*ss? 1:0);
2515 while (isspace(*s)) s++;
1eccaa59
PH
2516 } /* Next address */
2517
2518 parse_allow_group = FALSE; /* Reset group syntax flags */
2519 parse_found_group = FALSE;
059ec3d9
PH
2520
2521 /* If this was the bcc: header, mark it "old", which means it
2522 will be kept on the spool, but not transmitted as part of the
2523 message. */
2524
2cbb4081 2525 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2526 } /* For appropriate header line */
2527 } /* For each header line */
2528
059ec3d9
PH
2529 }
2530
2531/* Now build the unique message id. This has changed several times over the
2532lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2533Retaining all the history in the comment has become too unwieldy - read
2534previous release sources if you want it.
2535
2536The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2537The first part is the current time, in seconds. The second part is the current
2538pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2539can hold a number in the range 0-3843. It used to be a computed sequence
2540number, but is now the fractional component of the current time in units of
25411/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2542received, Exim ensures that the timer has ticked at the appropriate level
2543before proceeding, to avoid duplication if the pid happened to be re-used
2544within the same time period. It seems likely that most messages will take at
2545least half a millisecond to be received, so no delay will normally be
2546necessary. At least for some time...
2547
2548There is a modification when localhost_number is set. Formerly this was allowed
2549to be as large as 255. Now it is restricted to the range 0-16, and the final
2550component of the message id becomes (localhost_number * 200) + fractional time
2551in units of 1/200 of a second (i.e. a value in the range 0-3399).
2552
2553Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2554Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2555still allows the tttttt field to hold a large enough number to last for some
2556more decades, and the final two-digit field can hold numbers up to 1295, which
2557is enough for milliseconds (instead of 1/2000 of a second).
2558
2559However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2560pid, so it is probably safe because pids have to be positive. The
2561localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2562final field becomes (localhost_number * 100) + fractional time in centiseconds.
2563
2564Note that string_base62() returns its data in a static storage block, so it
2565must be copied before calling string_base62() again. It always returns exactly
25666 characters.
2567
2568There doesn't seem to be anything in the RFC which requires a message id to
2569start with a letter, but Smail was changed to ensure this. The external form of
2570the message id (as supplied by string expansion) therefore starts with an
2571additional leading 'E'. The spool file names do not include this leading
2572letter and it is not used internally.
2573
2574NOTE: If ever the format of message ids is changed, the regular expression for
2575checking that a string is in this format must be updated in a corresponding
2576way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2577must also be changed to reflect the correct string length. Then, of course,
2578other programs that rely on the message id format will need updating too. */
2579
2580Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2581message_id[6] = '-';
2582Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2583
2584/* Deal with the case where the host number is set. The value of the number was
2585checked when it was read, to ensure it isn't too big. The timing granularity is
2586left in id_resolution so that an appropriate wait can be done after receiving
2587the message, if necessary (we hope it won't be). */
2588
2589if (host_number_string != NULL)
2590 {
2591 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2592 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2593 string_base62((long int)(
2594 host_number * (1000000/id_resolution) +
2595 message_id_tv.tv_usec/id_resolution)) + 4);
2596 }
2597
2598/* Host number not set: final field is just the fractional time at an
2599appropriate resolution. */
2600
2601else
2602 {
2603 id_resolution = (BASE_62 == 62)? 500 : 1000;
2604 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2605 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2606 }
2607
2608/* Add the current message id onto the current process info string if
2609it will fit. */
2610
2611(void)string_format(process_info + process_info_len,
2612 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2613
2614/* If we are using multiple input directories, set up the one for this message
2615to be the least significant base-62 digit of the time of arrival. Otherwise
2616ensure that it is an empty string. */
2617
a2da3176 2618message_subdir[0] = split_spool_directory ? message_id[5] : 0;
059ec3d9
PH
2619
2620/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2621one, but only for local (without suppress_local_fixups) or submission mode
2622messages. This can be user-configured if required, but we had better flatten
2623any illegal characters therein. */
059ec3d9 2624
8800895a
PH
2625if (msgid_header == NULL &&
2626 ((sender_host_address == NULL && !suppress_local_fixups)
2627 || submission_mode))
059ec3d9
PH
2628 {
2629 uschar *p;
2630 uschar *id_text = US"";
2631 uschar *id_domain = primary_hostname;
2632
2633 /* Permit only letters, digits, dots, and hyphens in the domain */
2634
2635 if (message_id_domain != NULL)
2636 {
2637 uschar *new_id_domain = expand_string(message_id_domain);
2638 if (new_id_domain == NULL)
2639 {
2640 if (!expand_string_forcedfail)
2641 log_write(0, LOG_MAIN|LOG_PANIC,
2642 "expansion of \"%s\" (message_id_header_domain) "
2643 "failed: %s", message_id_domain, expand_string_message);
2644 }
2645 else if (*new_id_domain != 0)
2646 {
2647 id_domain = new_id_domain;
2648 for (p = id_domain; *p != 0; p++)
2649 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2650 }
2651 }
2652
2653 /* Permit all characters except controls and RFC 2822 specials in the
2654 additional text part. */
2655
2656 if (message_id_text != NULL)
2657 {
2658 uschar *new_id_text = expand_string(message_id_text);
2659 if (new_id_text == NULL)
2660 {
2661 if (!expand_string_forcedfail)
2662 log_write(0, LOG_MAIN|LOG_PANIC,
2663 "expansion of \"%s\" (message_id_header_text) "
2664 "failed: %s", message_id_text, expand_string_message);
2665 }
2666 else if (*new_id_text != 0)
2667 {
2668 id_text = new_id_text;
2669 for (p = id_text; *p != 0; p++)
2670 if (mac_iscntrl_or_special(*p)) *p = '-';
2671 }
2672 }
2673
e7e680d6
PP
2674 /* Add the header line
2675 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2676 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2677
e7e680d6 2678 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2679 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2680 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2681 }
2682
2683/* If we are to log recipients, keep a copy of the raw ones before any possible
2684rewriting. Must copy the count, because later ACLs and the local_scan()
2685function may mess with the real recipients. */
2686
6c6d6e48 2687if (LOGGING(received_recipients))
059ec3d9
PH
2688 {
2689 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2690 for (i = 0; i < recipients_count; i++)
2691 raw_recipients[i] = string_copy(recipients_list[i].address);
2692 raw_recipients_count = recipients_count;
2693 }
2694
2695/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2696recipients will get here only if the conditions were right (allow_unqualified_
2697recipient is TRUE). */
2698
2699for (i = 0; i < recipients_count; i++)
2700 recipients_list[i].address =
2701 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2702 global_rewrite_rules, rewrite_existflags);
2703
8800895a
PH
2704/* If there is no From: header, generate one for local (without
2705suppress_local_fixups) or submission_mode messages. If there is no sender
2706address, but the sender is local or this is a local delivery error, use the
2707originator login. This shouldn't happen for genuine bounces, but might happen
2708for autoreplies. The addition of From: must be done *before* checking for the
2709possible addition of a Sender: header, because untrusted_set_sender allows an
2710untrusted user to set anything in the envelope (which might then get info
2711From:) but we still want to ensure a valid Sender: if it is required. */
2712
2713if (from_header == NULL &&
2714 ((sender_host_address == NULL && !suppress_local_fixups)
2715 || submission_mode))
059ec3d9 2716 {
2fe1a124
PH
2717 uschar *oname = US"";
2718
2719 /* Use the originator_name if this is a locally submitted message and the
2720 caller is not trusted. For trusted callers, use it only if -F was used to
2721 force its value or if we have a non-SMTP message for which -f was not used
2722 to set the sender. */
2723
2724 if (sender_host_address == NULL)
2725 {
2726 if (!trusted_caller || sender_name_forced ||
2727 (!smtp_input && !sender_address_forced))
2728 oname = originator_name;
2729 }
2730
2731 /* For non-locally submitted messages, the only time we use the originator
2732 name is when it was forced by the /name= option on control=submission. */
2733
2734 else
2735 {
2736 if (submission_name != NULL) oname = submission_name;
2737 }
2738
059ec3d9
PH
2739 /* Envelope sender is empty */
2740
2741 if (sender_address[0] == 0)
2742 {
87ba3f5f
PH
2743 uschar *fromstart, *fromend;
2744
2745 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2746 oname, (oname[0] == 0)? "" : " <");
2747 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2748
059ec3d9
PH
2749 if (sender_local || local_error_message)
2750 {
87ba3f5f
PH
2751 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2752 local_part_quote(originator_login), qualify_domain_sender,
2753 fromend);
059ec3d9
PH
2754 }
2755 else if (submission_mode && authenticated_id != NULL)
2756 {
2757 if (submission_domain == NULL)
2758 {
87ba3f5f
PH
2759 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2760 local_part_quote(authenticated_id), qualify_domain_sender,
2761 fromend);
059ec3d9
PH
2762 }
2763 else if (submission_domain[0] == 0) /* empty => whole address set */
2764 {
87ba3f5f
PH
2765 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2766 fromend);
059ec3d9
PH
2767 }
2768 else
2769 {
87ba3f5f
PH
2770 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2771 local_part_quote(authenticated_id), submission_domain,
2772 fromend);
059ec3d9
PH
2773 }
2774 from_header = header_last; /* To get it checked for Sender: */
2775 }
2776 }
2777
2778 /* There is a non-null envelope sender. Build the header using the original
2779 sender address, before any rewriting that might have been done while
2780 verifying it. */
2781
2782 else
2783 {
87ba3f5f 2784 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2785 oname,
2786 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2787 (sender_address_unrewritten == NULL)?
2788 sender_address : sender_address_unrewritten,
2fe1a124 2789 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2790
2791 from_header = header_last; /* To get it checked for Sender: */
2792 }
2793 }
2794
2795
8800895a
PH
2796/* If the sender is local (without suppress_local_fixups), or if we are in
2797submission mode and there is an authenticated_id, check that an existing From:
2798is correct, and if not, generate a Sender: header, unless disabled. Any
2799previously-existing Sender: header was removed above. Note that sender_local,
2800as well as being TRUE if the caller of exim is not trusted, is also true if a
2801trusted caller did not supply a -f argument for non-smtp input. To allow
2802trusted callers to forge From: without supplying -f, we have to test explicitly
2803here. If the From: header contains more than one address, then the call to
2804parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2805
2806if (from_header != NULL &&
69358f02 2807 (active_local_from_check &&
8800895a 2808 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2809 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2810 ))
2811 {
2812 BOOL make_sender = TRUE;
2813 int start, end, domain;
2814 uschar *errmess;
2815 uschar *from_address =
2816 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2817 &start, &end, &domain, FALSE);
2818 uschar *generated_sender_address;
2819
2820 if (submission_mode)
2821 {
2822 if (submission_domain == NULL)
2823 {
2824 generated_sender_address = string_sprintf("%s@%s",
2825 local_part_quote(authenticated_id), qualify_domain_sender);
2826 }
2827 else if (submission_domain[0] == 0) /* empty => full address */
2828 {
2829 generated_sender_address = string_sprintf("%s",
2830 authenticated_id);
2831 }
2832 else
2833 {
2834 generated_sender_address = string_sprintf("%s@%s",
2835 local_part_quote(authenticated_id), submission_domain);
2836 }
2837 }
2838 else
2839 generated_sender_address = string_sprintf("%s@%s",
2840 local_part_quote(originator_login), qualify_domain_sender);
2841
2842 /* Remove permitted prefixes and suffixes from the local part of the From:
2843 address before doing the comparison with the generated sender. */
2844
2845 if (from_address != NULL)
2846 {
2847 int slen;
2848 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2849
2850 if (at != NULL) *at = 0;
2851 from_address += route_check_prefix(from_address, local_from_prefix);
2852 slen = route_check_suffix(from_address, local_from_suffix);
2853 if (slen > 0)
2854 {
2855 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2856 from_address += slen;
2857 }
2858 if (at != NULL) *at = '@';
2859
2860 if (strcmpic(generated_sender_address, from_address) == 0 ||
2861 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2862 make_sender = FALSE;
2863 }
2864
2865 /* We have to cause the Sender header to be rewritten if there are
2866 appropriate rewriting rules. */
2867
2868 if (make_sender)
2869 {
2fe1a124 2870 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2871 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2872 generated_sender_address);
2873 else
2874 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2875 resent_prefix,
2876 submission_mode? submission_name : originator_name,
2877 generated_sender_address);
059ec3d9 2878 }
87ba3f5f
PH
2879
2880 /* Ensure that a non-null envelope sender address corresponds to the
2881 submission mode sender address. */
2882
2883 if (submission_mode && sender_address[0] != 0)
2884 {
2885 if (sender_address_unrewritten == NULL)
2886 sender_address_unrewritten = sender_address;
2887 sender_address = generated_sender_address;
089793a4
TF
2888 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2889 log_write(L_address_rewrite, LOG_MAIN,
2890 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2891 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2892 }
059ec3d9
PH
2893 }
2894
059ec3d9
PH
2895/* If there are any rewriting rules, apply them to the sender address, unless
2896it has already been rewritten as part of verification for SMTP input. */
2897
2898if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2899 sender_address[0] != 0)
2900 {
2901 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2902 global_rewrite_rules, rewrite_existflags);
2903 DEBUG(D_receive|D_rewrite)
2904 debug_printf("rewritten sender = %s\n", sender_address);
2905 }
2906
2907
2908/* The headers must be run through rewrite_header(), because it ensures that
2909addresses are fully qualified, as well as applying any rewriting rules that may
2910exist.
2911
2912Qualification of header addresses in a message from a remote host happens only
2913if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2914appropriate. For local messages, qualification always happens, unless -bnq is
2915used to explicitly suppress it. No rewriting is done for an unqualified address
2916that is left untouched.
2917
2918We start at the second header, skipping our own Received:. This rewriting is
2919documented as happening *after* recipient addresses are taken from the headers
2920by the -t command line option. An added Sender: gets rewritten here. */
2921
1ebe15c3 2922for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2923 {
2924 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2925 rewrite_existflags, TRUE);
1ebe15c3 2926 if (newh) h = newh;
059ec3d9
PH
2927 }
2928
2929
2930/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2931"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2932"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2933exists, this applies to the set of resent- headers rather than the normal set.
2934
2cbb4081
PH
2935The requirement for a recipient header has been removed in RFC 2822. At this
2936point in the code, earlier versions of Exim added a To: header for locally
2937submitted messages, and an empty Bcc: header for others. In the light of the
2938changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2939
059ec3d9
PH
2940
2941/* If there is no date header, generate one if the message originates locally
8800895a
PH
2942(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2943submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2944to be more confusing if Exim adds one to all remotely-originated messages.
2945As per Message-Id, we prepend if resending, else append.
2946*/
059ec3d9 2947
8800895a
PH
2948if (!date_header_exists &&
2949 ((sender_host_address == NULL && !suppress_local_fixups)
2950 || submission_mode))
e7e680d6
PP
2951 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2952 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2953
2954search_tidyup(); /* Free any cached resources */
2955
2956/* Show the complete set of headers if debugging. Note that the first one (the
2957new Received:) has not yet been set. */
2958
2959DEBUG(D_receive)
2960 {
2961 debug_printf(">>Headers after rewriting and local additions:\n");
2962 for (h = header_list->next; h != NULL; h = h->next)
2963 debug_printf("%c %s", h->type, h->text);
2964 debug_printf("\n");
2965 }
2966
2967/* The headers are now complete in store. If we are running in filter
2968testing mode, that is all this function does. Return TRUE if the message
2969ended with a dot. */
2970
f05da2e8 2971if (filter_test != FTEST_NONE)
059ec3d9
PH
2972 {
2973 process_info[process_info_len] = 0;
2974 return message_ended == END_DOT;
2975 }
2976
7e3ce68e
JH
2977/*XXX CHUNKING: need to cancel cutthrough under BDAT, for now. In future,
2978think more if it could be handled. Cannot do onward CHUNKING unless
2979inbound is, but inbound chunking ought to be ok with outbound plain.
2980Could we do onward CHUNKING given inbound CHUNKING?
2981*/
2982if (chunking_state > CHUNKING_OFFERED)
57cc2785 2983 cancel_cutthrough_connection(FALSE, US"chunking active");
7e3ce68e 2984
817d9f57 2985/* Cutthrough delivery:
5032d1cf
JH
2986We have to create the Received header now rather than at the end of reception,
2987so the timestamp behaviour is a change to the normal case.
2988XXX Ensure this gets documented XXX.
2989Having created it, send the headers to the destination. */
57cc2785
JH
2990
2991if (cutthrough.fd >= 0 && cutthrough.delivery)
e4bdf652 2992 {
817d9f57
JH
2993 if (received_count > received_headers_max)
2994 {
57cc2785 2995 cancel_cutthrough_connection(TRUE, US"too many headers");
817d9f57
JH
2996 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
2997 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2998 "Too many \"Received\" headers",
2999 sender_address,
57cc2785
JH
3000 sender_fullhost ? "H=" : "", sender_fullhost ? sender_fullhost : US"",
3001 sender_ident ? "U=" : "", sender_ident ? sender_ident : US"");
817d9f57
JH
3002 message_id[0] = 0; /* Indicate no message accepted */
3003 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
3004 goto TIDYUP; /* Skip to end of function */
3005 }
e4bdf652 3006 received_header_gen();
578d43dc 3007 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652
JH
3008 (void) cutthrough_headers_send();
3009 }
61147df4 3010
e4bdf652 3011
059ec3d9
PH
3012/* Open a new spool file for the data portion of the message. We need
3013to access it both via a file descriptor and a stream. Try to make the
41313d92 3014directory if it isn't there. */
059ec3d9 3015
41313d92 3016spool_name = spool_fname(US"input", message_subdir, message_id, US"-D");
a2da3176
JH
3017DEBUG(D_receive) debug_printf("Data file name: %s\n", spool_name);
3018
3019if ((data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE)) < 0)
059ec3d9
PH
3020 {
3021 if (errno == ENOENT)
3022 {
0971ec06 3023 (void) directory_make(spool_directory,
41313d92
JH
3024 spool_sname(US"input", message_subdir),
3025 INPUT_DIRECTORY_MODE, TRUE);
059ec3d9
PH
3026 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
3027 }
3028 if (data_fd < 0)
3029 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
3030 spool_name, strerror(errno));
3031 }
3032
3033/* Make sure the file's group is the Exim gid, and double-check the mode
3034because the group setting doesn't always get set automatically. */
3035
1ac6b2e7
JH
3036if (fchown(data_fd, exim_uid, exim_gid))
3037 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
3038 "Failed setting ownership on spool file %s: %s",
3039 spool_name, strerror(errno));
ff790e47 3040(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
3041
3042/* We now have data file open. Build a stream for it and lock it. We lock only
3043the first line of the file (containing the message ID) because otherwise there
3044are problems when Exim is run under Cygwin (I'm told). See comments in
3045spool_in.c, where the same locking is done. */
3046
3047data_file = fdopen(data_fd, "w+");
3048lock_data.l_type = F_WRLCK;
3049lock_data.l_whence = SEEK_SET;
3050lock_data.l_start = 0;
3051lock_data.l_len = SPOOL_DATA_START_OFFSET;
3052
3053if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
3054 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
3055 errno, strerror(errno));
3056
3057/* We have an open, locked data file. Write the message id to it to make it
3058self-identifying. Then read the remainder of the input of this message and
3059write it to the data file. If the variable next != NULL, it contains the first
3060data line (which was read as a header but then turned out not to have the right
3061format); write it (remembering that it might contain binary zeros). The result
3062of fwrite() isn't inspected; instead we call ferror() below. */
3063
3064fprintf(data_file, "%s-D\n", message_id);
3065if (next != NULL)
3066 {
3067 uschar *s = next->text;
3068 int len = next->slen;
1ac6b2e7 3069 len = fwrite(s, 1, len, data_file); len = len; /* compiler quietening */
059ec3d9
PH
3070 body_linecount++; /* Assumes only 1 line */
3071 }
3072
3073/* Note that we might already be at end of file, or the logical end of file
3074(indicated by '.'), or might have encountered an error while writing the
3075message id or "next" line. */
3076
3077if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
3078 {
3079 if (smtp_input)
3080 {
7e3ce68e
JH
3081 message_ended = chunking_state > CHUNKING_OFFERED
3082 ? read_message_bdat_smtp(data_file)
3083 : read_message_data_smtp(data_file);
059ec3d9
PH
3084 receive_linecount++; /* The terminating "." line */
3085 }
3086 else message_ended = read_message_data(data_file);
3087
3088 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 3089 message_linecount += body_linecount;
059ec3d9 3090
7e3ce68e 3091 switch (message_ended)
059ec3d9 3092 {
7e3ce68e 3093 /* Handle premature termination of SMTP */
059ec3d9 3094
7e3ce68e
JH
3095 case END_EOF:
3096 if (smtp_input)
3097 {
3098 Uunlink(spool_name); /* Lose data file when closed */
57cc2785 3099 cancel_cutthrough_connection(TRUE, US"sender closed connection");
7e3ce68e
JH
3100 message_id[0] = 0; /* Indicate no message accepted */
3101 smtp_reply = handle_lost_connection(US"");
3102 smtp_yield = FALSE;
3103 goto TIDYUP; /* Skip to end of function */
3104 }
3105 break;
059ec3d9 3106
7e3ce68e
JH
3107 /* Handle message that is too big. Don't use host_or_ident() in the log
3108 message; we want to see the ident value even for non-remote messages. */
059ec3d9 3109
7e3ce68e
JH
3110 case END_SIZE:
3111 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3112 cancel_cutthrough_connection(TRUE, US"mail too big");
7e3ce68e 3113 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
059ec3d9 3114
7e3ce68e
JH
3115 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3116 "message too big: read=%d max=%d",
3117 sender_address,
3118 (sender_fullhost == NULL)? "" : " H=",
3119 (sender_fullhost == NULL)? US"" : sender_fullhost,
3120 (sender_ident == NULL)? "" : " U=",
3121 (sender_ident == NULL)? US"" : sender_ident,
3122 message_size,
3123 thismessage_size_limit);
3124
3125 if (smtp_input)
3126 {
3127 smtp_reply = US"552 Message size exceeds maximum permitted";
3128 message_id[0] = 0; /* Indicate no message accepted */
3129 goto TIDYUP; /* Skip to end of function */
3130 }
3131 else
3132 {
3133 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3134 give_local_error(ERRMESS_TOOBIG,
3135 string_sprintf("message too big (max=%d)", thismessage_size_limit),
3136 US"message rejected: ", error_rc, data_file, header_list);
3137 /* Does not return */
3138 }
3139 break;
3140
3141 /* Handle bad BDAT protocol sequence */
3142
3143 case END_PROTOCOL:
3144 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3145 cancel_cutthrough_connection(TRUE, US"sender protocol error");
7e3ce68e
JH
3146 smtp_reply = US""; /* Response already sent */
3147 message_id[0] = 0; /* Indicate no message accepted */
3148 goto TIDYUP; /* Skip to end of function */
059ec3d9
PH
3149 }
3150 }
3151
3152/* Restore the standard SIGALRM handler for any subsequent processing. (For
3153example, there may be some expansion in an ACL that uses a timer.) */
3154
3155os_non_restarting_signal(SIGALRM, sigalrm_handler);
3156
3157/* The message body has now been read into the data file. Call fflush() to
3158empty the buffers in C, and then call fsync() to get the data written out onto
3159the disk, as fflush() doesn't do this (or at least, it isn't documented as
3160having to do this). If there was an I/O error on either input or output,
3161attempt to send an error message, and unlink the spool file. For non-SMTP input
3162we can then give up. Note that for SMTP input we must swallow the remainder of
3163the input in cases of output errors, since the far end doesn't expect to see
3164anything until the terminating dot line is sent. */
3165
3166if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 3167 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
3168 {
3169 uschar *msg_errno = US strerror(errno);
3170 BOOL input_error = (receive_ferror)() != 0;
3171 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
3172 input_error? "Input read" : "Spool write",
3173 msg_errno,
3174 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
3175
3176 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
3177 Uunlink(spool_name); /* Lose the data file */
57cc2785 3178 cancel_cutthrough_connection(TRUE, US"error writing spoolfile");
059ec3d9
PH
3179
3180 if (smtp_input)
3181 {
3182 if (input_error)
3183 smtp_reply = US"451 Error while reading input data";
3184 else
3185 {
3186 smtp_reply = US"451 Error while writing spool file";
3187 receive_swallow_smtp();
3188 }
3189 message_id[0] = 0; /* Indicate no message accepted */
3190 goto TIDYUP; /* Skip to end of function */
3191 }
3192
3193 else
3194 {
3195 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3196 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
3197 header_list);
3198 /* Does not return */
3199 }
3200 }
3201
3202
3203/* No I/O errors were encountered while writing the data file. */
3204
3205DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
3206
3207
3208/* If there were any bad addresses extracted by -t, or there were no recipients
3209left after -t, send a message to the sender of this message, or write it to
3210stderr if the error handling option is set that way. Note that there may
3211legitimately be no recipients for an SMTP message if they have all been removed
3212by "discard".
3213
3214We need to rewind the data file in order to read it. In the case of no
3215recipients or stderr error writing, throw the data file away afterwards, and
3216exit. (This can't be SMTP, which always ensures there's at least one
3217syntactically good recipient address.) */
3218
3219if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
3220 {
3221 DEBUG(D_receive)
3222 {
3223 if (recipients_count == 0) debug_printf("*** No recipients\n");
3224 if (bad_addresses != NULL)
3225 {
3226 error_block *eblock = bad_addresses;
3227 debug_printf("*** Bad address(es)\n");
3228 while (eblock != NULL)
3229 {
3230 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
3231 eblock = eblock->next;
3232 }
3233 }
3234 }
3235
3236 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3237
3238 /* If configured to send errors to the sender, but this fails, force
3239 a failure error code. We use a special one for no recipients so that it
3240 can be detected by the autoreply transport. Otherwise error_rc is set to
3241 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
3242 it is EXIT_SUCCESS. */
3243
3244 if (error_handling == ERRORS_SENDER)
3245 {
3246 if (!moan_to_sender(
3247 (bad_addresses == NULL)?
3248 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
3249 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
3250 bad_addresses, header_list, data_file, FALSE))
3251 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
3252 }
3253 else
3254 {
3255 if (bad_addresses == NULL)
3256 {
3257 if (extracted_ignored)
3258 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3259 else
3260 fprintf(stderr, "exim: no recipients in message\n");
3261 }
3262 else
3263 {
3264 fprintf(stderr, "exim: invalid address%s",
3265 (bad_addresses->next == NULL)? ":" : "es:\n");
3266 while (bad_addresses != NULL)
3267 {
3268 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3269 bad_addresses->text2);
3270 bad_addresses = bad_addresses->next;
3271 }
3272 }
3273 }
3274
3275 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3276 {
3277 Uunlink(spool_name);
f1e894f3 3278 (void)fclose(data_file);
059ec3d9
PH
3279 exim_exit(error_rc);
3280 }
3281 }
3282
3283/* Data file successfully written. Generate text for the Received: header by
3284expanding the configured string, and adding a timestamp. By leaving this
3285operation till now, we ensure that the timestamp is the time that message
3286reception was completed. However, this is deliberately done before calling the
3287data ACL and local_scan().
3288
3289This Received: header may therefore be inspected by the data ACL and by code in
3290the local_scan() function. When they have run, we update the timestamp to be
3291the final time of reception.
3292
3293If there is just one recipient, set up its value in the $received_for variable
3294for use when we generate the Received: header.
3295
3296Note: the checking for too many Received: headers is handled by the delivery
3297code. */
e4bdf652 3298/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3299
e4bdf652 3300if (received_header->text == NULL) /* Non-cutthrough case */
059ec3d9 3301 {
e4bdf652 3302 received_header_gen();
059ec3d9 3303
e4bdf652 3304 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3305
e4bdf652
JH
3306 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3307 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3308
e4bdf652
JH
3309 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3310 now, before running the DATA ACL. */
059ec3d9 3311
578d43dc 3312 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652 3313 }
817d9f57 3314else
e4bdf652
JH
3315 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3316 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3317
3318/* If an ACL is specified for checking things at this stage of reception of a
3319message, run it, unless all the recipients were removed by "discard" in earlier
3320ACLs. That is the only case in which recipients_count can be zero at this
3321stage. Set deliver_datafile to point to the data file so that $message_body and
3322$message_body_end can be extracted if needed. Allow $recipients in expansions.
3323*/
3324
3325deliver_datafile = data_fd;
4e88a19f 3326user_msg = NULL;
059ec3d9 3327
0e20aff9
MH
3328enable_dollar_recipients = TRUE;
3329
059ec3d9 3330if (recipients_count == 0)
7e3ce68e
JH
3331 blackholed_by = recipients_discarded ? US"MAIL ACL" : US"RCPT ACL";
3332
059ec3d9
PH
3333else
3334 {
059ec3d9
PH
3335 /* Handle interactive SMTP messages */
3336
3337 if (smtp_input && !smtp_batched_input)
3338 {
8523533c 3339
80a47a2c
TK
3340#ifndef DISABLE_DKIM
3341 if (!dkim_disable_verify)
3342 {
3343 /* Finish verification, this will log individual signature results to
3344 the mainlog */
3345 dkim_exim_verify_finish();
3346
3347 /* Check if we must run the DKIM ACL */
7e3ce68e 3348 if (acl_smtp_dkim && dkim_verify_signers && *dkim_verify_signers)
80a47a2c
TK
3349 {
3350 uschar *dkim_verify_signers_expanded =
3351 expand_string(dkim_verify_signers);
7e3ce68e 3352 if (!dkim_verify_signers_expanded)
80a47a2c
TK
3353 log_write(0, LOG_MAIN|LOG_PANIC,
3354 "expansion of dkim_verify_signers option failed: %s",
3355 expand_string_message);
7e3ce68e 3356
80a47a2c
TK
3357 else
3358 {
3359 int sep = 0;
55414b25 3360 const uschar *ptr = dkim_verify_signers_expanded;
80a47a2c 3361 uschar *item = NULL;
9e5d6b55
TK
3362 uschar *seen_items = NULL;
3363 int seen_items_size = 0;
3364 int seen_items_offset = 0;
9122af94
TK
3365 /* Default to OK when no items are present */
3366 rc = OK;
7e3ce68e 3367 while ((item = string_nextinlist(&ptr, &sep, NULL, 0)))
80a47a2c 3368 {
6119d1ea 3369 /* Prevent running ACL for an empty item */
7e3ce68e 3370 if (!item || !*item) continue;
5032d1cf
JH
3371
3372 /* Only run ACL once for each domain or identity,
3373 no matter how often it appears in the expanded list. */
3374 if (seen_items)
6119d1ea 3375 {
ae9094bf 3376 uschar *seen_item = NULL;
55414b25 3377 const uschar *seen_items_list = seen_items;
5032d1cf 3378 BOOL seen_this_item = FALSE;
61147df4 3379
ae9094bf 3380 while ((seen_item = string_nextinlist(&seen_items_list, &sep,
7e3ce68e 3381 NULL, 0)))
5032d1cf
JH
3382 if (Ustrcmp(seen_item,item) == 0)
3383 {
3384 seen_this_item = TRUE;
3385 break;
3386 }
3387
3388 if (seen_this_item)
6119d1ea
TK
3389 {
3390 DEBUG(D_receive)
5032d1cf
JH
3391 debug_printf("acl_smtp_dkim: skipping signer %s, "
3392 "already seen\n", item);
6119d1ea
TK
3393 continue;
3394 }
61147df4 3395
5032d1cf
JH
3396 seen_items = string_append(seen_items, &seen_items_size,
3397 &seen_items_offset, 1, ":");
6119d1ea
TK
3398 }
3399
5032d1cf
JH
3400 seen_items = string_append(seen_items, &seen_items_size,
3401 &seen_items_offset, 1, item);
4a73449b 3402 seen_items[seen_items_offset] = '\0';
6119d1ea
TK
3403
3404 DEBUG(D_receive)
5032d1cf
JH
3405 debug_printf("calling acl_smtp_dkim for dkim_cur_signer=%s\n",
3406 item);
6119d1ea 3407
80a47a2c 3408 dkim_exim_acl_setup(item);
5032d1cf
JH
3409 rc = acl_check(ACL_WHERE_DKIM, NULL, acl_smtp_dkim,
3410 &user_msg, &log_msg);
6119d1ea
TK
3411
3412 if (rc != OK)
5032d1cf
JH
3413 {
3414 DEBUG(D_receive)
3415 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, "
3416 "skipping remaining items\n", rc, item);
57cc2785 3417 cancel_cutthrough_connection(TRUE, US"dkim acl not ok");
5032d1cf
JH
3418 break;
3419 }
80a47a2c 3420 }
578d43dc 3421 add_acl_headers(ACL_WHERE_DKIM, US"DKIM");
80a47a2c
TK
3422 if (rc == DISCARD)
3423 {
3424 recipients_count = 0;
3425 blackholed_by = US"DKIM ACL";
3426 if (log_msg != NULL)
3427 blackhole_log_msg = string_sprintf(": %s", log_msg);
3428 }
3429 else if (rc != OK)
3430 {
3431 Uunlink(spool_name);
3432 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
85ffcba6 3433 smtp_yield = FALSE; /* No more messages after dropped connection */
80a47a2c
TK
3434 smtp_reply = US""; /* Indicate reply already sent */
3435 message_id[0] = 0; /* Indicate no message accepted */
3436 goto TIDYUP; /* Skip to end of function */
3437 }
3438 }
3439 }
3440 }
4a8ce2d8 3441#endif /* DISABLE_DKIM */
fb2274d4 3442
8523533c 3443#ifdef WITH_CONTENT_SCAN
80a47a2c
TK
3444 if (recipients_count > 0 &&
3445 acl_smtp_mime != NULL &&
54cdb463
PH
3446 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
3447 goto TIDYUP;
8523533c
TK
3448#endif /* WITH_CONTENT_SCAN */
3449
4840604e
TL
3450#ifdef EXPERIMENTAL_DMARC
3451 dmarc_up = dmarc_store_data(from_header);
3452#endif /* EXPERIMENTAL_DMARC */
3453
8ccd00b1
JH
3454#ifndef DISABLE_PRDR
3455 if (prdr_requested && recipients_count > 1 && acl_smtp_data_prdr)
fd98a5c6
JH
3456 {
3457 unsigned int c;
3458 int all_pass = OK;
3459 int all_fail = FAIL;
3460
3461 smtp_printf("353 PRDR content analysis beginning\r\n");
3462 /* Loop through recipients, responses must be in same order received */
3463 for (c = 0; recipients_count > c; c++)
3464 {
3465 uschar * addr= recipients_list[c].address;
3466 uschar * msg= US"PRDR R=<%s> %s";
3467 uschar * code;
3468 DEBUG(D_receive)
3469 debug_printf("PRDR processing recipient %s (%d of %d)\n",
3470 addr, c+1, recipients_count);
3471 rc = acl_check(ACL_WHERE_PRDR, addr,
3472 acl_smtp_data_prdr, &user_msg, &log_msg);
3473
3474 /* If any recipient rejected content, indicate it in final message */
3475 all_pass |= rc;
3476 /* If all recipients rejected, indicate in final message */
3477 all_fail &= rc;
3478
3479 switch (rc)
3480 {
3481 case OK: case DISCARD: code = US"250"; break;
3482 case DEFER: code = US"450"; break;
3483 default: code = US"550"; break;
3484 }
3485 if (user_msg != NULL)
3486 smtp_user_msg(code, user_msg);
3487 else
3488 {
3489 switch (rc)
3490 {
3491 case OK: case DISCARD:
3492 msg = string_sprintf(CS msg, addr, "acceptance"); break;
3493 case DEFER:
3494 msg = string_sprintf(CS msg, addr, "temporary refusal"); break;
3495 default:
3496 msg = string_sprintf(CS msg, addr, "refusal"); break;
3497 }
3498 smtp_user_msg(code, msg);
3499 }
3500 if (log_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, log_msg);
3501 else if (user_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, user_msg);
112b6a93 3502 else log_write(0, LOG_MAIN, "%s", CS msg);
fd98a5c6
JH
3503
3504 if (rc != OK) { receive_remove_recipient(addr); c--; }
3505 }
3506 /* Set up final message, used if data acl gives OK */
3507 smtp_reply = string_sprintf("%s id=%s message %s",
3508 all_fail == FAIL ? US"550" : US"250",
3509 message_id,
3510 all_fail == FAIL
3511 ? US"rejected for all recipients"
3512 : all_pass == OK
3513 ? US"accepted"
3514 : US"accepted for some recipients");
3515 if (recipients_count == 0)
3516 {
3517 message_id[0] = 0; /* Indicate no message accepted */
3518 goto TIDYUP;
3519 }
3520 }
3521 else
3522 prdr_requested = FALSE;
8ccd00b1 3523#endif /* !DISABLE_PRDR */
fd98a5c6 3524
54cdb463
PH
3525 /* Check the recipients count again, as the MIME ACL might have changed
3526 them. */
8523533c 3527
059ec3d9
PH
3528 if (acl_smtp_data != NULL && recipients_count > 0)
3529 {
059ec3d9 3530 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
578d43dc 3531 add_acl_headers(ACL_WHERE_DATA, US"DATA");
059ec3d9
PH
3532 if (rc == DISCARD)
3533 {
3534 recipients_count = 0;
3535 blackholed_by = US"DATA ACL";
57cc2785 3536 if (log_msg)
8e669ac1 3537 blackhole_log_msg = string_sprintf(": %s", log_msg);
57cc2785 3538 cancel_cutthrough_connection(TRUE, US"data acl discard");
059ec3d9
PH
3539 }
3540 else if (rc != OK)
3541 {
3542 Uunlink(spool_name);
57cc2785 3543 cancel_cutthrough_connection(TRUE, US"data acl not ok");
8523533c
TK
3544#ifdef WITH_CONTENT_SCAN
3545 unspool_mbox();
3546#endif
6f0c431a
PP
3547#ifdef EXPERIMENTAL_DCC
3548 dcc_ok = 0;
3549#endif
059ec3d9 3550 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
85ffcba6 3551 smtp_yield = FALSE; /* No more messages after dropped connection */
059ec3d9
PH
3552 smtp_reply = US""; /* Indicate reply already sent */
3553 message_id[0] = 0; /* Indicate no message accepted */
3554 goto TIDYUP; /* Skip to end of function */
3555 }
3556 }
3557 }
3558
3559 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
3560 we cannot take different actions for permanent and temporary rejections. */
3561
54cdb463 3562 else
059ec3d9 3563 {
54cdb463
PH
3564
3565#ifdef WITH_CONTENT_SCAN
3566 if (acl_not_smtp_mime != NULL &&
3567 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
3568 &blackholed_by))
3569 goto TIDYUP;
3570#endif /* WITH_CONTENT_SCAN */
3571
3572 if (acl_not_smtp != NULL)
059ec3d9 3573 {
54cdb463
PH
3574 uschar *user_msg, *log_msg;
3575 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
3576 if (rc == DISCARD)
059ec3d9 3577 {
54cdb463
PH
3578 recipients_count = 0;
3579 blackholed_by = US"non-SMTP ACL";
3580 if (log_msg != NULL)
3581 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 3582 }
54cdb463 3583 else if (rc != OK)
059ec3d9 3584 {
54cdb463
PH
3585 Uunlink(spool_name);
3586#ifdef WITH_CONTENT_SCAN
3587 unspool_mbox();
3588#endif
6f0c431a
PP
3589#ifdef EXPERIMENTAL_DCC
3590 dcc_ok = 0;
3591#endif
6ea85e9a
PH
3592 /* The ACL can specify where rejections are to be logged, possibly
3593 nowhere. The default is main and reject logs. */
3594
3595 if (log_reject_target != 0)
3596 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3597 sender_address, log_msg);
3598
54cdb463
PH
3599 if (user_msg == NULL) user_msg = US"local configuration problem";
3600 if (smtp_batched_input)
3601 {
3602 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3603 /* Does not return */
3604 }
3605 else
3606 {
3607 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3608 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3609 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3610 header_list);
3611 /* Does not return */
3612 }
059ec3d9 3613 }
578d43dc 3614 add_acl_headers(ACL_WHERE_NOTSMTP, US"non-SMTP");
059ec3d9 3615 }
059ec3d9
PH
3616 }
3617
54cdb463
PH
3618 /* The applicable ACLs have been run */
3619
059ec3d9
PH
3620 if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */
3621 if (queue_only_policy) queued_by = US"ACL";
059ec3d9
PH
3622 }
3623
8523533c
TK
3624#ifdef WITH_CONTENT_SCAN
3625unspool_mbox();
3626#endif
3627
6a8f9482
TK
3628#ifdef EXPERIMENTAL_DCC
3629dcc_ok = 0;
3630#endif
3631
3632
059ec3d9
PH
3633/* The final check on the message is to run the scan_local() function. The
3634version supplied with Exim always accepts, but this is a hook for sysadmins to
3635supply their own checking code. The local_scan() function is run even when all
3636the recipients have been discarded. */
3637
3638lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3639
3640/* Arrange to catch crashes in local_scan(), so that the -D file gets
3641deleted, and the incident gets logged. */
3642
3643os_non_restarting_signal(SIGSEGV, local_scan_crash_handler);
3644os_non_restarting_signal(SIGFPE, local_scan_crash_handler);
3645os_non_restarting_signal(SIGILL, local_scan_crash_handler);
3646os_non_restarting_signal(SIGBUS, local_scan_crash_handler);
3647
3648DEBUG(D_receive) debug_printf("calling local_scan(); timeout=%d\n",
3649 local_scan_timeout);
3650local_scan_data = NULL;
3651
3652os_non_restarting_signal(SIGALRM, local_scan_timeout_handler);
3653if (local_scan_timeout > 0) alarm(local_scan_timeout);
3654rc = local_scan(data_fd, &local_scan_data);
3655alarm(0);
3656os_non_restarting_signal(SIGALRM, sigalrm_handler);
3657
0e20aff9
MH
3658enable_dollar_recipients = FALSE;
3659
059ec3d9
PH
3660store_pool = POOL_MAIN; /* In case changed */
3661DEBUG(D_receive) debug_printf("local_scan() returned %d %s\n", rc,
3662 local_scan_data);
3663
3664os_non_restarting_signal(SIGSEGV, SIG_DFL);
3665os_non_restarting_signal(SIGFPE, SIG_DFL);
3666os_non_restarting_signal(SIGILL, SIG_DFL);
3667os_non_restarting_signal(SIGBUS, SIG_DFL);
3668
3669/* The length check is paranoia against some runaway code, and also because
3670(for a success return) lines in the spool file are read into big_buffer. */
3671
3672if (local_scan_data != NULL)
3673 {
3674 int len = Ustrlen(local_scan_data);
3675 if (len > LOCAL_SCAN_MAX_RETURN) len = LOCAL_SCAN_MAX_RETURN;
3676 local_scan_data = string_copyn(local_scan_data, len);
3677 }
3678