tidying
[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;
7156b1ef 1309int rc = OK;
54cdb463
PH
1310
1311memset(CS rfc822_file_path,0,2048);
1312
1313/* check if it is a MIME message */
1314my_headerlist = header_list;
4e88a19f
PH
1315while (my_headerlist != NULL)
1316 {
54cdb463 1317 /* skip deleted headers */
4e88a19f
PH
1318 if (my_headerlist->type == '*')
1319 {
54cdb463
PH
1320 my_headerlist = my_headerlist->next;
1321 continue;
4e88a19f
PH
1322 }
1323 if (strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0)
1324 {
54cdb463
PH
1325 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1326 goto DO_MIME_ACL;
4e88a19f 1327 }
54cdb463 1328 my_headerlist = my_headerlist->next;
4e88a19f 1329 }
54cdb463
PH
1330
1331DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1332return TRUE;
1333
1334DO_MIME_ACL:
1335/* make sure the eml mbox file is spooled up */
8544e77a 1336mbox_file = spool_mbox(&mbox_size, NULL);
54cdb463
PH
1337if (mbox_file == NULL) {
1338 /* error while spooling */
1339 log_write(0, LOG_MAIN|LOG_PANIC,
1340 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1341 Uunlink(spool_name);
1342 unspool_mbox();
6f0c431a
PP
1343#ifdef EXPERIMENTAL_DCC
1344 dcc_ok = 0;
1345#endif
a5bd321b 1346 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1347 message_id[0] = 0; /* Indicate no message accepted */
1348 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1349 return FALSE; /* Indicate skip to end of receive function */
1350};
1351
1352mime_is_rfc822 = 0;
1353
1354MIME_ACL_CHECK:
1355mime_part_count = -1;
1356rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1357(void)fclose(mbox_file);
54cdb463 1358
4e88a19f
PH
1359if (Ustrlen(rfc822_file_path) > 0)
1360 {
54cdb463
PH
1361 mime_part_count = mime_part_count_buffer;
1362
4e88a19f
PH
1363 if (unlink(CS rfc822_file_path) == -1)
1364 {
54cdb463
PH
1365 log_write(0, LOG_PANIC,
1366 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1367 goto END_MIME_ACL;
4e88a19f
PH
1368 }
1369 }
54cdb463
PH
1370
1371/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1372if (rc == OK)
1373 {
54cdb463 1374 uschar temp_path[1024];
e8bc7fca
JH
1375 struct dirent * entry;
1376 DIR * tempdir;
54cdb463 1377
e8bc7fca
JH
1378 (void) string_format(temp_path, sizeof(temp_path), "%s/scan/%s",
1379 spool_directory, message_id);
54cdb463 1380
4e88a19f 1381 tempdir = opendir(CS temp_path);
e8bc7fca 1382 for (;;)
4e88a19f 1383 {
e8bc7fca
JH
1384 if (!(entry = readdir(tempdir)))
1385 break;
1386 if (strncmpic(US entry->d_name, US"__rfc822_", 9) == 0)
4e88a19f 1387 {
e8bc7fca
JH
1388 (void) string_format(rfc822_file_path, sizeof(rfc822_file_path),
1389 "%s/scan/%s/%s", spool_directory, message_id, entry->d_name);
e1d04f48 1390 DEBUG(D_receive) debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
e8bc7fca 1391 rfc822_file_path);
4e88a19f
PH
1392 break;
1393 }
e8bc7fca 1394 }
4e88a19f 1395 closedir(tempdir);
54cdb463 1396
e8bc7fca 1397 if (entry)
4e88a19f 1398 {
e8bc7fca 1399 if ((mbox_file = Ufopen(rfc822_file_path, "rb")))
4e88a19f 1400 {
e8bc7fca
JH
1401 /* set RFC822 expansion variable */
1402 mime_is_rfc822 = 1;
1403 mime_part_count_buffer = mime_part_count;
1404 goto MIME_ACL_CHECK;
4e88a19f 1405 }
e8bc7fca
JH
1406 log_write(0, LOG_PANIC,
1407 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1408 unlink(CS rfc822_file_path);
4e88a19f
PH
1409 }
1410 }
54cdb463
PH
1411
1412END_MIME_ACL:
578d43dc 1413add_acl_headers(ACL_WHERE_MIME, US"MIME");
54cdb463
PH
1414if (rc == DISCARD)
1415 {
1416 recipients_count = 0;
1417 *blackholed_by_ptr = US"MIME ACL";
1418 }
1419else if (rc != OK)
1420 {
1421 Uunlink(spool_name);
1422 unspool_mbox();
6f0c431a
PP
1423#ifdef EXPERIMENTAL_DCC
1424 dcc_ok = 0;
1425#endif
4f6ae5c3
JH
1426 if ( smtp_input
1427 && smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1428 {
85ffcba6 1429 *smtp_yield_ptr = FALSE; /* No more messages after dropped connection */
f4c1088b 1430 *smtp_reply_ptr = US""; /* Indicate reply already sent */
4f6ae5c3 1431 }
54cdb463
PH
1432 message_id[0] = 0; /* Indicate no message accepted */
1433 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1434 }
54cdb463
PH
1435
1436return TRUE;
1437}
1438
63955bf2 1439#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1440
1441
e4bdf652
JH
1442
1443void
1444received_header_gen(void)
1445{
1446uschar *received;
1447uschar *timestamp;
1448header_line *received_header= header_list;
1449
1450timestamp = expand_string(US"${tod_full}");
1451if (recipients_count == 1) received_for = recipients_list[0].address;
1452received = expand_string(received_header_text);
1453received_for = NULL;
1454
d4ff61d1 1455if (!received)
e4bdf652
JH
1456 {
1457 if(spool_name[0] != 0)
1458 Uunlink(spool_name); /* Lose the data file */
1459 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1460 "(received_header_text) failed: %s", string_printing(received_header_text),
1461 expand_string_message);
1462 }
1463
1464/* The first element on the header chain is reserved for the Received header,
1465so all we have to do is fill in the text pointer, and set the type. However, if
1466the result of the expansion is an empty string, we leave the header marked as
1467"old" so as to refrain from adding a Received header. */
1468
1469if (received[0] == 0)
1470 {
1471 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1472 received_header->type = htype_old;
1473 }
1474else
1475 {
1476 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1477 received_header->type = htype_received;
1478 }
1479
1480received_header->slen = Ustrlen(received_header->text);
1481
1482DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1483 received_header->type, received_header->text);
1484}
1485
1486
1487
54cdb463 1488/*************************************************
059ec3d9
PH
1489* Receive message *
1490*************************************************/
1491
1492/* Receive a message on the given input, and put it into a pair of spool files.
1493Either a non-null list of recipients, or the extract flag will be true, or
1494both. The flag sender_local is true for locally generated messages. The flag
1495submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1496suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1497suppress_local_fixups" or -G was passed on the command-line.
1498The flag smtp_input is true if the message is to be
8800895a
PH
1499handled using SMTP conventions about termination and lines starting with dots.
1500For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1501
1502If a message was successfully read, message_id[0] will be non-zero.
1503
1504The general actions of this function are:
1505
1506 . Read the headers of the message (if any) into a chain of store
1507 blocks.
1508
1509 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1510 throw it away, unless the caller is trusted, or unless
1511 active_local_sender_retain is set - which can only happen if
1512 active_local_from_check is false.
059ec3d9
PH
1513
1514 . If recipients are to be extracted from the message, build the
1515 recipients list from the headers, removing any that were on the
1516 original recipients list (unless extract_addresses_remove_arguments is
1517 false), and at the same time, remove any bcc header that may be present.
1518
1519 . Get the spool file for the data, sort out its unique name, open
1520 and lock it (but don't give it the name yet).
1521
1522 . Generate a "Message-Id" header if the message doesn't have one, for
1523 locally-originated messages.
1524
1525 . Generate a "Received" header.
1526
1527 . Ensure the recipients list is fully qualified and rewritten if necessary.
1528
1529 . If there are any rewriting rules, apply them to the sender address
1530 and also to the headers.
1531
1532 . If there is no from: header, generate one, for locally-generated messages
1533 and messages in "submission mode" only.
1534
1535 . If the sender is local, check that from: is correct, and if not, generate
1536 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1537 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1538
1539 . If there is no "date" header, generate one, for locally-originated
1540 or submission mode messages only.
1541
1542 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1543 dot_ends mode, to the data file. Leave it open, to hold the lock.
1544
1545 . Write the envelope and the headers to a new file.
1546
1547 . Set the name for the header file; close it.
1548
1549 . Set the name for the data file; close it.
1550
1551Because this function can potentially be called many times in a single
1552SMTP connection, all store should be got by store_get(), so that it will be
1553automatically retrieved after the message is accepted.
1554
1555FUDGE: It seems that sites on the net send out messages with just LF
1556terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1557we make the CRs optional in all cases.
1558
1559July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1560new regime is now in place in which bare CRs in header lines are turned into LF
1561followed by a space, so as not to terminate the header line.
1562
1563February 2004: A bare LF in a header line in a message whose first line was
1564terminated by CRLF is treated in the same way as a bare CR.
1565
1566Arguments:
1567 extract_recip TRUE if recipients are to be extracted from the message's
1568 headers
1569
1570Returns: TRUE there are more messages to be read (SMTP input)
1571 FALSE there are no more messages to be read (non-SMTP input
1572 or SMTP connection collapsed, or other failure)
1573
1574When reading a message for filter testing, the returned value indicates
1575whether the headers (which is all that is read) were terminated by '.' or
1576not. */
1577
1578BOOL
1579receive_msg(BOOL extract_recip)
1580{
7156b1ef
NM
1581int i;
1582int rc = FAIL;
059ec3d9
PH
1583int msg_size = 0;
1584int process_info_len = Ustrlen(process_info);
1585int error_rc = (error_handling == ERRORS_SENDER)?
1586 errors_sender_rc : EXIT_FAILURE;
1587int header_size = 256;
1588int start, end, domain, size, sptr;
1589int id_resolution;
1590int had_zero = 0;
d677b2f2 1591int prevlines_length = 0;
059ec3d9
PH
1592
1593register int ptr = 0;
1594
1595BOOL contains_resent_headers = FALSE;
1596BOOL extracted_ignored = FALSE;
1597BOOL first_line_ended_crlf = TRUE_UNSET;
1598BOOL smtp_yield = TRUE;
1599BOOL yield = FALSE;
1600
1601BOOL resents_exist = FALSE;
1602uschar *resent_prefix = US"";
1603uschar *blackholed_by = NULL;
04f7d5b9 1604uschar *blackhole_log_msg = US"";
c5430c20 1605enum {NOT_TRIED, TMP_REJ, PERM_REJ, ACCEPTED} cutthrough_done = NOT_TRIED;
059ec3d9
PH
1606
1607flock_t lock_data;
1608error_block *bad_addresses = NULL;
1609
1610uschar *frozen_by = NULL;
1611uschar *queued_by = NULL;
1612
1613uschar *errmsg, *s;
1614struct stat statbuf;
1615
4e88a19f 1616/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1617
1618uschar *smtp_reply = NULL;
4e88a19f 1619uschar *user_msg, *log_msg;
059ec3d9
PH
1620
1621/* Working header pointers */
1622
1623header_line *h, *next;
1624
2cbb4081 1625/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1626
1627BOOL date_header_exists = FALSE;
1628
1629/* Pointers to receive the addresses of headers whose contents we need. */
1630
1631header_line *from_header = NULL;
1632header_line *subject_header = NULL;
1633header_line *msgid_header = NULL;
1634header_line *received_header;
1635
4840604e
TL
1636#ifdef EXPERIMENTAL_DMARC
1637int dmarc_up = 0;
1638#endif /* EXPERIMENTAL_DMARC */
1639
059ec3d9
PH
1640/* Variables for use when building the Received: header. */
1641
059ec3d9
PH
1642uschar *timestamp;
1643int tslen;
1644
1645/* Release any open files that might have been cached while preparing to
1646accept the message - e.g. by verifying addresses - because reading a message
1647might take a fair bit of real time. */
1648
1649search_tidyup();
1650
e4bdf652
JH
1651/* Extracting the recipient list from an input file is incompatible with
1652cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1653to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1654if (extract_recip || !smtp_input)
57cc2785 1655 cancel_cutthrough_connection(TRUE, US"not smtp input");
e4bdf652 1656
059ec3d9
PH
1657/* Initialize the chain of headers by setting up a place-holder for Received:
1658header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1659pointing to the end of the chain to make adding headers simple. */
1660
1661received_header = header_list = header_last = store_get(sizeof(header_line));
1662header_list->next = NULL;
1663header_list->type = htype_old;
1664header_list->text = NULL;
1665header_list->slen = 0;
1666
1667/* Control block for the next header to be read. */
1668
1669next = store_get(sizeof(header_line));
1670next->text = store_get(header_size);
1671
1672/* Initialize message id to be null (indicating no message read), and the
1673header names list to be the normal list. Indicate there is no data file open
1674yet, initialize the size and warning count, and deal with no size limit. */
1675
1676message_id[0] = 0;
1677data_file = NULL;
1678data_fd = -1;
41313d92 1679spool_name = US"";
059ec3d9
PH
1680message_size = 0;
1681warning_count = 0;
d677b2f2 1682received_count = 1; /* For the one we will add */
059ec3d9
PH
1683
1684if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1685
2e0c1448 1686/* While reading the message, the following counts are computed. */
059ec3d9 1687
d677b2f2
PH
1688message_linecount = body_linecount = body_zerocount =
1689 max_received_linelength = 0;
059ec3d9 1690
80a47a2c 1691#ifndef DISABLE_DKIM
e983e85a
JH
1692/* Call into DKIM to set up the context. In CHUNKING mode
1693we clear the dot-stuffing flag */
1694if (smtp_input && !smtp_batched_input && !dkim_disable_verify)
1695 dkim_exim_verify_init(chunking_state <= CHUNKING_OFFERED);
fb2274d4
TK
1696#endif
1697
4840604e
TL
1698#ifdef EXPERIMENTAL_DMARC
1699/* initialize libopendmarc */
1700dmarc_up = dmarc_init();
1701#endif
1702
059ec3d9
PH
1703/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1704ids, and fractions of a second are required. See the comments that precede the
1705message id creation below. */
1706
1707(void)gettimeofday(&message_id_tv, NULL);
1708
1709/* For other uses of the received time we can operate with granularity of one
1710second, and for that we use the global variable received_time. This is for
1711things like ultimate message timeouts. */
1712
1713received_time = message_id_tv.tv_sec;
1714
1715/* If SMTP input, set the special handler for timeouts. The alarm() calls
1716happen in the smtp_getc() function when it refills its buffer. */
1717
1718if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1719
1720/* If not SMTP input, timeout happens only if configured, and we just set a
1721single timeout for the whole message. */
1722
1723else if (receive_timeout > 0)
1724 {
1725 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1726 alarm(receive_timeout);
1727 }
1728
1729/* SIGTERM and SIGINT are caught always. */
1730
1731signal(SIGTERM, data_sigterm_sigint_handler);
1732signal(SIGINT, data_sigterm_sigint_handler);
1733
1734/* Header lines in messages are not supposed to be very long, though when
1735unfolded, to: and cc: headers can take up a lot of store. We must also cope
1736with the possibility of junk being thrown at us. Start by getting 256 bytes for
1737storing the header, and extend this as necessary using string_cat().
1738
1739To cope with total lunacies, impose an upper limit on the length of the header
1740section of the message, as otherwise the store will fill up. We must also cope
1741with the possibility of binary zeros in the data. Hence we cannot use fgets().
1742Folded header lines are joined into one string, leaving the '\n' characters
1743inside them, so that writing them out reproduces the input.
1744
1745Loop for each character of each header; the next structure for chaining the
1746header is set up already, with ptr the offset of the next character in
1747next->text. */
1748
1749for (;;)
1750 {
bd8fbe36 1751 int ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1752
1753 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1754 SMTP must have a correct "." terminator. */
1755
1756 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1757 {
1758 smtp_reply = handle_lost_connection(US" (header)");
1759 smtp_yield = FALSE;
1760 goto TIDYUP; /* Skip to end of function */
1761 }
1762
1763 /* See if we are at the current header's size limit - there must be at least
1764 four bytes left. This allows for the new character plus a zero, plus two for
1765 extra insertions when we are playing games with dots and carriage returns. If
1766 we are at the limit, extend the text buffer. This could have been done
1767 automatically using string_cat() but because this is a tightish loop storing
1768 only one character at a time, we choose to do it inline. Normally
1769 store_extend() will be able to extend the block; only at the end of a big
1770 store block will a copy be needed. To handle the case of very long headers
1771 (and sometimes lunatic messages can have ones that are 100s of K long) we
1772 call store_release() for strings that have been copied - if the string is at
1773 the start of a block (and therefore the only thing in it, because we aren't
1774 doing any other gets), the block gets freed. We can only do this because we
1775 know there are no other calls to store_get() going on. */
1776
1777 if (ptr >= header_size - 4)
1778 {
1779 int oldsize = header_size;
1780 /* header_size += 256; */
1781 header_size *= 2;
1782 if (!store_extend(next->text, oldsize, header_size))
1783 {
1784 uschar *newtext = store_get(header_size);
1785 memcpy(newtext, next->text, ptr);
1786 store_release(next->text);
1787 next->text = newtext;
1788 }
1789 }
1790
1791 /* Cope with receiving a binary zero. There is dispute about whether
1792 these should be allowed in RFC 822 messages. The middle view is that they
1793 should not be allowed in headers, at least. Exim takes this attitude at
1794 the moment. We can't just stomp on them here, because we don't know that
1795 this line is a header yet. Set a flag to cause scanning later. */
1796
1797 if (ch == 0) had_zero++;
1798
1799 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1800 those from data files use just LF. Treat LF in local SMTP input as a
1801 terminator too. Treat EOF as a line terminator always. */
1802
1803 if (ch == EOF) goto EOL;
1804
1805 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1806 other MTAs accept this. We are therefore forced into this "liberalisation"
1807 too, so we accept LF as a line terminator whatever the source of the message.
1808 However, if the first line of the message ended with a CRLF, we treat a bare
1809 LF specially by inserting a white space after it to ensure that the header
1810 line is not terminated. */
1811
1812 if (ch == '\n')
1813 {
1814 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1815 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1816 goto EOL;
1817 }
1818
1819 /* This is not the end of the line. If this is SMTP input and this is
1820 the first character in the line and it is a "." character, ignore it.
1821 This implements the dot-doubling rule, though header lines starting with
1822 dots aren't exactly common. They are legal in RFC 822, though. If the
1823 following is CRLF or LF, this is the line that that terminates the
1824 entire message. We set message_ended to indicate this has happened (to
1825 prevent further reading), and break out of the loop, having freed the
1826 empty header, and set next = NULL to indicate no data line. */
1827
1828 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1829 {
bd8fbe36 1830 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1831 if (ch == '\r')
1832 {
bd8fbe36 1833 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1834 if (ch != '\n')
1835 {
80a47a2c 1836 receive_ungetc(ch);
059ec3d9
PH
1837 ch = '\r'; /* Revert to CR */
1838 }
1839 }
1840 if (ch == '\n')
1841 {
1842 message_ended = END_DOT;
1843 store_reset(next);
1844 next = NULL;
1845 break; /* End character-reading loop */
1846 }
1847
1848 /* For non-SMTP input, the dot at the start of the line was really a data
1849 character. What is now in ch is the following character. We guaranteed
1850 enough space for this above. */
1851
1852 if (!smtp_input)
1853 {
1854 next->text[ptr++] = '.';
1855 message_size++;
1856 }
1857 }
1858
1859 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1860 remember this case if this is the first line ending. */
1861
1862 if (ch == '\r')
1863 {
bd8fbe36 1864 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1865 if (ch == '\n')
1866 {
1867 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1868 goto EOL;
1869 }
1870
1871 /* Otherwise, put back the character after CR, and turn the bare CR
1872 into LF SP. */
1873
80a47a2c 1874 ch = (receive_ungetc)(ch);
059ec3d9
PH
1875 next->text[ptr++] = '\n';
1876 message_size++;
1877 ch = ' ';
1878 }
1879
1880 /* We have a data character for the header line. */
1881
1882 next->text[ptr++] = ch; /* Add to buffer */
1883 message_size++; /* Total message size so far */
1884
1885 /* Handle failure due to a humungously long header section. The >= allows
1886 for the terminating \n. Add what we have so far onto the headers list so
1887 that it gets reflected in any error message, and back up the just-read
1888 character. */
1889
1890 if (message_size >= header_maxsize)
1891 {
1892 next->text[ptr] = 0;
1893 next->slen = ptr;
1894 next->type = htype_other;
1895 next->next = NULL;
1896 header_last->next = next;
1897 header_last = next;
1898
1899 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1900 "%s (more than %d characters): message abandoned",
1901 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1902
1903 if (smtp_input)
1904 {
1905 smtp_reply = US"552 Message header is ridiculously long";
1906 receive_swallow_smtp();
1907 goto TIDYUP; /* Skip to end of function */
1908 }
1909
1910 else
1911 {
1912 give_local_error(ERRMESS_VLONGHEADER,
1913 string_sprintf("message header longer than %d characters received: "
1914 "message not accepted", header_maxsize), US"", error_rc, stdin,
1915 header_list->next);
1916 /* Does not return */
1917 }
1918 }
1919
1920 continue; /* With next input character */
1921
1922 /* End of header line reached */
1923
1924 EOL:
2e0c1448
PH
1925
1926 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1927
1928 receive_linecount++;
1929 message_linecount++;
059ec3d9 1930
d677b2f2
PH
1931 /* Keep track of maximum line length */
1932
1933 if (ptr - prevlines_length > max_received_linelength)
1934 max_received_linelength = ptr - prevlines_length;
1935 prevlines_length = ptr + 1;
1936
059ec3d9
PH
1937 /* Now put in the terminating newline. There is always space for
1938 at least two more characters. */
1939
1940 next->text[ptr++] = '\n';
1941 message_size++;
1942
1943 /* A blank line signals the end of the headers; release the unwanted
1944 space and set next to NULL to indicate this. */
1945
1946 if (ptr == 1)
1947 {
1948 store_reset(next);
1949 next = NULL;
1950 break;
1951 }
1952
1953 /* There is data in the line; see if the next input character is a
1954 whitespace character. If it is, we have a continuation of this header line.
1955 There is always space for at least one character at this point. */
1956
1957 if (ch != EOF)
1958 {
bd8fbe36 1959 int nextch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1960 if (nextch == ' ' || nextch == '\t')
1961 {
1962 next->text[ptr++] = nextch;
1963 message_size++;
1964 continue; /* Iterate the loop */
1965 }
80a47a2c 1966 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
1967 else ch = EOF; /* Cause main loop to exit at end */
1968 }
1969
1970 /* We have got to the real line end. Terminate the string and release store
1971 beyond it. If it turns out to be a real header, internal binary zeros will
1972 be squashed later. */
1973
1974 next->text[ptr] = 0;
1975 next->slen = ptr;
1976 store_reset(next->text + ptr + 1);
1977
1978 /* Check the running total size against the overall message size limit. We
1979 don't expect to fail here, but if the overall limit is set less than MESSAGE_
1980 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
1981 headers - the code to read the body will then also hit the buffer. */
1982
1983 if (message_size > thismessage_size_limit) break;
1984
1985 /* A line that is not syntactically correct for a header also marks
1986 the end of the headers. In this case, we leave next containing the
1987 first data line. This might actually be several lines because of the
1988 continuation logic applied above, but that doesn't matter.
1989
1990 It turns out that smail, and presumably sendmail, accept leading lines
1991 of the form
1992
1993 From ph10 Fri Jan 5 12:35 GMT 1996
1994
1995 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
1996 find any documentation of this, but for compatibility it had better be
1997 accepted. Exim restricts it to the case of non-smtp messages, and
1998 treats it as an alternative to the -f command line option. Thus it is
1999 ignored except for trusted users or filter testing. Otherwise it is taken
2000 as the sender address, unless -f was used (sendmail compatibility).
2001
2002 It further turns out that some UUCPs generate the From_line in a different
2003 format, e.g.
2004
2005 From ph10 Fri, 7 Jan 97 14:00:00 GMT
2006
2007 The regex for matching these things is now capable of recognizing both
2008 formats (including 2- and 4-digit years in the latter). In fact, the regex
2009 is now configurable, as is the expansion string to fish out the sender.
2010
2011 Even further on it has been discovered that some broken clients send
2012 these lines in SMTP messages. There is now an option to ignore them from
2013 specified hosts or networks. Sigh. */
2014
2015 if (header_last == header_list &&
2016 (!smtp_input
2017 ||
2018 (sender_host_address != NULL &&
2019 verify_check_host(&ignore_fromline_hosts) == OK)
2020 ||
2021 (sender_host_address == NULL && ignore_fromline_local)
2022 ) &&
2023 regex_match_and_setup(regex_From, next->text, 0, -1))
2024 {
2025 if (!sender_address_forced)
2026 {
2027 uschar *uucp_sender = expand_string(uucp_from_sender);
2028 if (uucp_sender == NULL)
2029 {
2030 log_write(0, LOG_MAIN|LOG_PANIC,
2031 "expansion of \"%s\" failed after matching "
2032 "\"From \" line: %s", uucp_from_sender, expand_string_message);
2033 }
2034 else
2035 {
2036 int start, end, domain;
2037 uschar *errmess;
2038 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
2039 &start, &end, &domain, TRUE);
2040 if (newsender != NULL)
2041 {
2042 if (domain == 0 && newsender[0] != 0)
2043 newsender = rewrite_address_qualify(newsender, FALSE);
2044
f05da2e8 2045 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
2046 {
2047 sender_address = newsender;
2048
f05da2e8 2049 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
2050 {
2051 authenticated_sender = NULL;
2052 originator_name = US"";
2053 sender_local = FALSE;
2054 }
2055
f05da2e8 2056 if (filter_test != FTEST_NONE)
059ec3d9
PH
2057 printf("Sender taken from \"From \" line\n");
2058 }
2059 }
2060 }
2061 }
2062 }
2063
2064 /* Not a leading "From " line. Check to see if it is a valid header line.
2065 Header names may contain any non-control characters except space and colon,
2066 amazingly. */
2067
2068 else
2069 {
2070 uschar *p = next->text;
2071
2072 /* If not a valid header line, break from the header reading loop, leaving
2073 next != NULL, indicating that it holds the first line of the body. */
2074
2075 if (isspace(*p)) break;
2076 while (mac_isgraph(*p) && *p != ':') p++;
2077 while (isspace(*p)) p++;
2078 if (*p != ':')
2079 {
2080 body_zerocount = had_zero;
2081 break;
2082 }
2083
2084 /* We have a valid header line. If there were any binary zeroes in
2085 the line, stomp on them here. */
2086
2087 if (had_zero > 0)
2088 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
2089
2090 /* It is perfectly legal to have an empty continuation line
2091 at the end of a header, but it is confusing to humans
2092 looking at such messages, since it looks like a blank line.
2093 Reduce confusion by removing redundant white space at the
2094 end. We know that there is at least one printing character
2095 (the ':' tested for above) so there is no danger of running
2096 off the end. */
2097
2098 p = next->text + ptr - 2;
2099 for (;;)
2100 {
2101 while (*p == ' ' || *p == '\t') p--;
2102 if (*p != '\n') break;
2103 ptr = (p--) - next->text + 1;
2104 message_size -= next->slen - ptr;
2105 next->text[ptr] = 0;
2106 next->slen = ptr;
2107 }
2108
2109 /* Add the header to the chain */
2110
2111 next->type = htype_other;
2112 next->next = NULL;
2113 header_last->next = next;
2114 header_last = next;
2115
2116 /* Check the limit for individual line lengths. This comes after adding to
2117 the chain so that the failing line is reflected if a bounce is generated
2118 (for a local message). */
2119
2120 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
2121 {
2122 log_write(0, LOG_MAIN, "overlong message header line received from "
2123 "%s (more than %d characters): message abandoned",
2124 sender_host_unknown? sender_ident : sender_fullhost,
2125 header_line_maxsize);
2126
2127 if (smtp_input)
2128 {
2129 smtp_reply = US"552 A message header line is too long";
2130 receive_swallow_smtp();
2131 goto TIDYUP; /* Skip to end of function */
2132 }
2133
2134 else
2135 {
2136 give_local_error(ERRMESS_VLONGHDRLINE,
2137 string_sprintf("message header line longer than %d characters "
2138 "received: message not accepted", header_line_maxsize), US"",
2139 error_rc, stdin, header_list->next);
2140 /* Does not return */
2141 }
2142 }
2143
2144 /* Note if any resent- fields exist. */
2145
2146 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
2147 {
2148 resents_exist = TRUE;
2149 resent_prefix = US"Resent-";
2150 }
2151 }
2152
1ebe15c3
JH
2153 /* Reject CHUNKING messages that do not CRLF their first header line */
2154
2155 if (!first_line_ended_crlf && chunking_state > CHUNKING_OFFERED)
2156 {
2157 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2158 "Non-CRLF-terminated header, under CHUNKING: message abandoned",
2159 sender_address,
2160 sender_fullhost ? " H=" : "", sender_fullhost ? sender_fullhost : US"",
2161 sender_ident ? " U=" : "", sender_ident ? sender_ident : US"");
2162 smtp_printf("552 Message header not CRLF terminated\r\n");
2163 bdat_flush_data();
2164 smtp_reply = US"";
2165 goto TIDYUP; /* Skip to end of function */
2166 }
2167
059ec3d9
PH
2168 /* The line has been handled. If we have hit EOF, break out of the loop,
2169 indicating no pending data line. */
2170
2171 if (ch == EOF) { next = NULL; break; }
2172
2173 /* Set up for the next header */
2174
2175 header_size = 256;
2176 next = store_get(sizeof(header_line));
2177 next->text = store_get(header_size);
2178 ptr = 0;
2179 had_zero = 0;
d677b2f2 2180 prevlines_length = 0;
059ec3d9
PH
2181 } /* Continue, starting to read the next header */
2182
2183/* At this point, we have read all the headers into a data structure in main
2184store. The first header is still the dummy placeholder for the Received: header
2185we are going to generate a bit later on. If next != NULL, it contains the first
2186data line - which terminated the headers before reaching a blank line (not the
2187normal case). */
2188
2189DEBUG(D_receive)
2190 {
2191 debug_printf(">>Headers received:\n");
1ebe15c3 2192 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2193 debug_printf("%s", h->text);
2194 debug_printf("\n");
2195 }
2196
2197/* End of file on any SMTP connection is an error. If an incoming SMTP call
2198is dropped immediately after valid headers, the next thing we will see is EOF.
2199We must test for this specially, as further down the reading of the data is
2200skipped if already at EOF. */
2201
2202if (smtp_input && (receive_feof)())
2203 {
2204 smtp_reply = handle_lost_connection(US" (after header)");
2205 smtp_yield = FALSE;
2206 goto TIDYUP; /* Skip to end of function */
2207 }
2208
2209/* If this is a filter test run and no headers were read, output a warning
2210in case there is a mistake in the test message. */
2211
f05da2e8 2212if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2213 printf("Warning: no message headers read\n");
2214
2215
2216/* Scan the headers to identify them. Some are merely marked for later
2217processing; some are dealt with here. */
2218
1ebe15c3 2219for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2220 {
2221 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2222 if (is_resent) contains_resent_headers = TRUE;
2223
2224 switch (header_checkname(h, is_resent))
2225 {
059ec3d9 2226 case htype_bcc:
2cbb4081 2227 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
2228 break;
2229
059ec3d9 2230 case htype_cc:
2cbb4081 2231 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
2232 break;
2233
2234 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
2235
2236 case htype_date:
4c69d561 2237 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
2238 break;
2239
2240 /* Same comments as about Return-Path: below. */
2241
2242 case htype_delivery_date:
2243 if (delivery_date_remove) h->type = htype_old;
2244 break;
2245
2246 /* Same comments as about Return-Path: below. */
2247
2248 case htype_envelope_to:
2249 if (envelope_to_remove) h->type = htype_old;
2250 break;
2251
2252 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2253 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2254 header consists of just the login id of the user who called Exim, rewrite
2255 it with the gecos field first. Apply this rule to Resent-From: if there
2256 are resent- fields. */
2257
2258 case htype_from:
2259 h->type = htype_from;
2260 if (!resents_exist || is_resent)
2261 {
2262 from_header = h;
2263 if (!smtp_input)
2264 {
5de8faa3 2265 int len;
059ec3d9
PH
2266 uschar *s = Ustrchr(h->text, ':') + 1;
2267 while (isspace(*s)) s++;
5de8faa3 2268 len = h->slen - (s - h->text) - 1;
e0fccd1d
TF
2269 if (Ustrlen(originator_login) == len &&
2270 strncmpic(s, originator_login, len) == 0)
059ec3d9
PH
2271 {
2272 uschar *name = is_resent? US"Resent-From" : US"From";
2273 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2274 originator_login, qualify_domain_sender);
2275 from_header = header_last;
2276 h->type = htype_old;
2277 DEBUG(D_receive|D_rewrite)
2278 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2279 }
2280 }
2281 }
2282 break;
2283
2284 /* Identify the Message-id: header for generating "in-reply-to" in the
2285 autoreply transport. For incoming logging, save any resent- value. In both
2286 cases, take just the first of any multiples. */
2287
2288 case htype_id:
2289 if (msgid_header == NULL && (!resents_exist || is_resent))
2290 {
2291 msgid_header = h;
2292 h->type = htype_id;
2293 }
2294 break;
2295
2296 /* Flag all Received: headers */
2297
2298 case htype_received:
2299 h->type = htype_received;
2300 received_count++;
2301 break;
2302
2303 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
2304
2305 case htype_reply_to:
2306 h->type = htype_reply_to;
2307 break;
2308
2309 /* The Return-path: header is supposed to be added to messages when
2310 they leave the SMTP system. We shouldn't receive messages that already
2311 contain Return-path. However, since Exim generates Return-path: on
2312 local delivery, resent messages may well contain it. We therefore
2313 provide an option (which defaults on) to remove any Return-path: headers
2314 on input. Removal actually means flagging as "old", which prevents the
2315 header being transmitted with the message. */
2316
2317 case htype_return_path:
2318 if (return_path_remove) h->type = htype_old;
2319
2320 /* If we are testing a mail filter file, use the value of the
2321 Return-Path: header to set up the return_path variable, which is not
2322 otherwise set. However, remove any <> that surround the address
2323 because the variable doesn't have these. */
2324
f05da2e8 2325 if (filter_test != FTEST_NONE)
059ec3d9
PH
2326 {
2327 uschar *start = h->text + 12;
2328 uschar *end = start + Ustrlen(start);
2329 while (isspace(*start)) start++;
2330 while (end > start && isspace(end[-1])) end--;
2331 if (*start == '<' && end[-1] == '>')
2332 {
2333 start++;
2334 end--;
2335 }
2336 return_path = string_copyn(start, end - start);
2337 printf("Return-path taken from \"Return-path:\" header line\n");
2338 }
2339 break;
2340
2341 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2342 and from an untrusted caller and suppress_local_fixups is not set, or if we
2343 are in submission mode for a remote message, mark it "old" so that it will
2344 not be transmitted with the message, unless active_local_sender_retain is
2345 set. (This can only be true if active_local_from_check is false.) If there
2346 are any resent- headers in the message, apply this rule to Resent-Sender:
2347 instead of Sender:. Messages with multiple resent- header sets cannot be
2348 tidily handled. (For this reason, at least one MUA - Pine - turns old
2349 resent- headers into X-resent- headers when resending, leaving just one
2350 set.) */
059ec3d9
PH
2351
2352 case htype_sender:
69358f02 2353 h->type = ((!active_local_sender_retain &&
8800895a
PH
2354 (
2355 (sender_local && !trusted_caller && !suppress_local_fixups)
2356 || submission_mode
2357 )
059ec3d9
PH
2358 ) &&
2359 (!resents_exist||is_resent))?
2360 htype_old : htype_sender;
2361 break;
2362
2363 /* Remember the Subject: header for logging. There is no Resent-Subject */
2364
2365 case htype_subject:
2366 subject_header = h;
2367 break;
2368
2369 /* "To:" gets flagged, and the existence of a recipient header is noted,
2370 whether it's resent- or not. */
2371
2372 case htype_to:
2373 h->type = htype_to;
2374 /****
2375 to_or_cc_header_exists = TRUE;
2376 ****/
2377 break;
2378 }
2379 }
2380
2381/* Extract recipients from the headers if that is required (the -t option).
2382Note that this is documented as being done *before* any address rewriting takes
2383place. There are two possibilities:
2384
2385(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2386recipients already listed are to be REMOVED from the message. Smail 3 works
2387like this. We need to build a non-recipients tree for that list, because in
2388subsequent processing this data is held in a tree and that's what the
2389spool_write_header() function expects. Make sure that non-recipient addresses
2390are fully qualified and rewritten if necessary.
2391
2392(2) According to other sendmail documentation, -t ADDS extracted recipients to
2393those in the command line arguments (and it is rumoured some other MTAs do
2394this). Therefore, there is an option to make Exim behave this way.
2395
2396*** Notes on "Resent-" header lines ***
2397
2398The presence of resent-headers in the message makes -t horribly ambiguous.
2399Experiments with sendmail showed that it uses recipients for all resent-
2400headers, totally ignoring the concept of "sets of resent- headers" as described
2401in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2402with all the addresses in one instance of each header.
2403
2404This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2405error for -t if there were resent- headers in the message. However, after a
2406discussion on the mailing list, I've learned that there are MUAs that use
2407resent- headers with -t, and also that the stuff about sets of resent- headers
2408and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2409message with -t and resent- header lines makes sure that only *its* resent-
2410headers are present; previous ones are often renamed as X-resent- for example.
2411
2412Consequently, Exim has been changed so that, if any resent- header lines are
2413present, the recipients are taken from all of the appropriate resent- lines,
2414and not from the ordinary To:, Cc:, etc. */
2415
2416if (extract_recip)
2417 {
2418 int rcount = 0;
2419 error_block **bnext = &bad_addresses;
2420
2421 if (extract_addresses_remove_arguments)
2422 {
2423 while (recipients_count-- > 0)
2424 {
2425 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2426 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2427 tree_add_nonrecipient(s);
2428 }
2429 recipients_list = NULL;
2430 recipients_count = recipients_list_max = 0;
2431 }
2432
059ec3d9
PH
2433 /* Now scan the headers */
2434
1ebe15c3 2435 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2436 {
2437 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2438 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2439 {
2440 uschar *s = Ustrchr(h->text, ':') + 1;
2441 while (isspace(*s)) s++;
2442
1eccaa59
PH
2443 parse_allow_group = TRUE; /* Allow address group syntax */
2444
059ec3d9
PH
2445 while (*s != 0)
2446 {
2447 uschar *ss = parse_find_address_end(s, FALSE);
2448 uschar *recipient, *errmess, *p, *pp;
2449 int start, end, domain;
2450
2451 /* Check on maximum */
2452
2453 if (recipients_max > 0 && ++rcount > recipients_max)
2454 {
2455 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2456 US"message rejected: ", error_rc, stdin, NULL);
2457 /* Does not return */
2458 }
2459
2460 /* Make a copy of the address, and remove any internal newlines. These
2461 may be present as a result of continuations of the header line. The
2462 white space that follows the newline must not be removed - it is part
2463 of the header. */
2464
2465 pp = recipient = store_get(ss - s + 1);
2466 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2467 *pp = 0;
250b6871 2468
8c5d388a 2469#ifdef SUPPORT_I18N
250b6871
JH
2470 {
2471 BOOL b = allow_utf8_domains;
2472 allow_utf8_domains = TRUE;
2473#endif
059ec3d9
PH
2474 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2475 &domain, FALSE);
2476
8c5d388a 2477#ifdef SUPPORT_I18N
250b6871
JH
2478 if (string_is_utf8(recipient))
2479 message_smtputf8 = TRUE;
2480 else
2481 allow_utf8_domains = b;
2482 }
2483#endif
2484
059ec3d9
PH
2485 /* Keep a list of all the bad addresses so we can send a single
2486 error message at the end. However, an empty address is not an error;
2487 just ignore it. This can come from an empty group list like
2488
2489 To: Recipients of list:;
2490
2491 If there are no recipients at all, an error will occur later. */
2492
2493 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2494 {
2495 int len = Ustrlen(s);
2496 error_block *b = store_get(sizeof(error_block));
2497 while (len > 0 && isspace(s[len-1])) len--;
2498 b->next = NULL;
2499 b->text1 = string_printing(string_copyn(s, len));
2500 b->text2 = errmess;
2501 *bnext = b;
2502 bnext = &(b->next);
2503 }
2504
2505 /* If the recipient is already in the nonrecipients tree, it must
2506 have appeared on the command line with the option extract_addresses_
2507 remove_arguments set. Do not add it to the recipients, and keep a note
2508 that this has happened, in order to give a better error if there are
2509 no recipients left. */
2510
2511 else if (recipient != NULL)
2512 {
2513 if (tree_search(tree_nonrecipients, recipient) == NULL)
2514 receive_add_recipient(recipient, -1);
2515 else
2516 extracted_ignored = TRUE;
2517 }
2518
2519 /* Move on past this address */
2520
2521 s = ss + (*ss? 1:0);
2522 while (isspace(*s)) s++;
1eccaa59
PH
2523 } /* Next address */
2524
2525 parse_allow_group = FALSE; /* Reset group syntax flags */
2526 parse_found_group = FALSE;
059ec3d9
PH
2527
2528 /* If this was the bcc: header, mark it "old", which means it
2529 will be kept on the spool, but not transmitted as part of the
2530 message. */
2531
2cbb4081 2532 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2533 } /* For appropriate header line */
2534 } /* For each header line */
2535
059ec3d9
PH
2536 }
2537
2538/* Now build the unique message id. This has changed several times over the
2539lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2540Retaining all the history in the comment has become too unwieldy - read
2541previous release sources if you want it.
2542
2543The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2544The first part is the current time, in seconds. The second part is the current
2545pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2546can hold a number in the range 0-3843. It used to be a computed sequence
2547number, but is now the fractional component of the current time in units of
25481/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2549received, Exim ensures that the timer has ticked at the appropriate level
2550before proceeding, to avoid duplication if the pid happened to be re-used
2551within the same time period. It seems likely that most messages will take at
2552least half a millisecond to be received, so no delay will normally be
2553necessary. At least for some time...
2554
2555There is a modification when localhost_number is set. Formerly this was allowed
2556to be as large as 255. Now it is restricted to the range 0-16, and the final
2557component of the message id becomes (localhost_number * 200) + fractional time
2558in units of 1/200 of a second (i.e. a value in the range 0-3399).
2559
2560Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2561Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2562still allows the tttttt field to hold a large enough number to last for some
2563more decades, and the final two-digit field can hold numbers up to 1295, which
2564is enough for milliseconds (instead of 1/2000 of a second).
2565
2566However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2567pid, so it is probably safe because pids have to be positive. The
2568localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2569final field becomes (localhost_number * 100) + fractional time in centiseconds.
2570
2571Note that string_base62() returns its data in a static storage block, so it
2572must be copied before calling string_base62() again. It always returns exactly
25736 characters.
2574
2575There doesn't seem to be anything in the RFC which requires a message id to
2576start with a letter, but Smail was changed to ensure this. The external form of
2577the message id (as supplied by string expansion) therefore starts with an
2578additional leading 'E'. The spool file names do not include this leading
2579letter and it is not used internally.
2580
2581NOTE: If ever the format of message ids is changed, the regular expression for
2582checking that a string is in this format must be updated in a corresponding
2583way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2584must also be changed to reflect the correct string length. Then, of course,
2585other programs that rely on the message id format will need updating too. */
2586
2587Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2588message_id[6] = '-';
2589Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2590
2591/* Deal with the case where the host number is set. The value of the number was
2592checked when it was read, to ensure it isn't too big. The timing granularity is
2593left in id_resolution so that an appropriate wait can be done after receiving
2594the message, if necessary (we hope it won't be). */
2595
2596if (host_number_string != NULL)
2597 {
2598 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2599 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2600 string_base62((long int)(
2601 host_number * (1000000/id_resolution) +
2602 message_id_tv.tv_usec/id_resolution)) + 4);
2603 }
2604
2605/* Host number not set: final field is just the fractional time at an
2606appropriate resolution. */
2607
2608else
2609 {
2610 id_resolution = (BASE_62 == 62)? 500 : 1000;
2611 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2612 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2613 }
2614
2615/* Add the current message id onto the current process info string if
2616it will fit. */
2617
2618(void)string_format(process_info + process_info_len,
2619 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2620
2621/* If we are using multiple input directories, set up the one for this message
2622to be the least significant base-62 digit of the time of arrival. Otherwise
2623ensure that it is an empty string. */
2624
a2da3176 2625message_subdir[0] = split_spool_directory ? message_id[5] : 0;
059ec3d9
PH
2626
2627/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2628one, but only for local (without suppress_local_fixups) or submission mode
2629messages. This can be user-configured if required, but we had better flatten
2630any illegal characters therein. */
059ec3d9 2631
8800895a
PH
2632if (msgid_header == NULL &&
2633 ((sender_host_address == NULL && !suppress_local_fixups)
2634 || submission_mode))
059ec3d9
PH
2635 {
2636 uschar *p;
2637 uschar *id_text = US"";
2638 uschar *id_domain = primary_hostname;
2639
2640 /* Permit only letters, digits, dots, and hyphens in the domain */
2641
2642 if (message_id_domain != NULL)
2643 {
2644 uschar *new_id_domain = expand_string(message_id_domain);
2645 if (new_id_domain == NULL)
2646 {
2647 if (!expand_string_forcedfail)
2648 log_write(0, LOG_MAIN|LOG_PANIC,
2649 "expansion of \"%s\" (message_id_header_domain) "
2650 "failed: %s", message_id_domain, expand_string_message);
2651 }
2652 else if (*new_id_domain != 0)
2653 {
2654 id_domain = new_id_domain;
2655 for (p = id_domain; *p != 0; p++)
2656 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2657 }
2658 }
2659
2660 /* Permit all characters except controls and RFC 2822 specials in the
2661 additional text part. */
2662
2663 if (message_id_text != NULL)
2664 {
2665 uschar *new_id_text = expand_string(message_id_text);
2666 if (new_id_text == NULL)
2667 {
2668 if (!expand_string_forcedfail)
2669 log_write(0, LOG_MAIN|LOG_PANIC,
2670 "expansion of \"%s\" (message_id_header_text) "
2671 "failed: %s", message_id_text, expand_string_message);
2672 }
2673 else if (*new_id_text != 0)
2674 {
2675 id_text = new_id_text;
2676 for (p = id_text; *p != 0; p++)
2677 if (mac_iscntrl_or_special(*p)) *p = '-';
2678 }
2679 }
2680
e7e680d6
PP
2681 /* Add the header line
2682 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2683 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2684
e7e680d6 2685 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2686 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2687 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2688 }
2689
2690/* If we are to log recipients, keep a copy of the raw ones before any possible
2691rewriting. Must copy the count, because later ACLs and the local_scan()
2692function may mess with the real recipients. */
2693
6c6d6e48 2694if (LOGGING(received_recipients))
059ec3d9
PH
2695 {
2696 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2697 for (i = 0; i < recipients_count; i++)
2698 raw_recipients[i] = string_copy(recipients_list[i].address);
2699 raw_recipients_count = recipients_count;
2700 }
2701
2702/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2703recipients will get here only if the conditions were right (allow_unqualified_
2704recipient is TRUE). */
2705
2706for (i = 0; i < recipients_count; i++)
2707 recipients_list[i].address =
2708 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2709 global_rewrite_rules, rewrite_existflags);
2710
8800895a
PH
2711/* If there is no From: header, generate one for local (without
2712suppress_local_fixups) or submission_mode messages. If there is no sender
2713address, but the sender is local or this is a local delivery error, use the
2714originator login. This shouldn't happen for genuine bounces, but might happen
2715for autoreplies. The addition of From: must be done *before* checking for the
2716possible addition of a Sender: header, because untrusted_set_sender allows an
2717untrusted user to set anything in the envelope (which might then get info
2718From:) but we still want to ensure a valid Sender: if it is required. */
2719
2720if (from_header == NULL &&
2721 ((sender_host_address == NULL && !suppress_local_fixups)
2722 || submission_mode))
059ec3d9 2723 {
2fe1a124
PH
2724 uschar *oname = US"";
2725
2726 /* Use the originator_name if this is a locally submitted message and the
2727 caller is not trusted. For trusted callers, use it only if -F was used to
2728 force its value or if we have a non-SMTP message for which -f was not used
2729 to set the sender. */
2730
2731 if (sender_host_address == NULL)
2732 {
2733 if (!trusted_caller || sender_name_forced ||
2734 (!smtp_input && !sender_address_forced))
2735 oname = originator_name;
2736 }
2737
2738 /* For non-locally submitted messages, the only time we use the originator
2739 name is when it was forced by the /name= option on control=submission. */
2740
2741 else
2742 {
2743 if (submission_name != NULL) oname = submission_name;
2744 }
2745
059ec3d9
PH
2746 /* Envelope sender is empty */
2747
2748 if (sender_address[0] == 0)
2749 {
87ba3f5f
PH
2750 uschar *fromstart, *fromend;
2751
2752 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2753 oname, (oname[0] == 0)? "" : " <");
2754 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2755
059ec3d9
PH
2756 if (sender_local || local_error_message)
2757 {
87ba3f5f
PH
2758 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2759 local_part_quote(originator_login), qualify_domain_sender,
2760 fromend);
059ec3d9
PH
2761 }
2762 else if (submission_mode && authenticated_id != NULL)
2763 {
2764 if (submission_domain == NULL)
2765 {
87ba3f5f
PH
2766 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2767 local_part_quote(authenticated_id), qualify_domain_sender,
2768 fromend);
059ec3d9
PH
2769 }
2770 else if (submission_domain[0] == 0) /* empty => whole address set */
2771 {
87ba3f5f
PH
2772 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2773 fromend);
059ec3d9
PH
2774 }
2775 else
2776 {
87ba3f5f
PH
2777 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2778 local_part_quote(authenticated_id), submission_domain,
2779 fromend);
059ec3d9
PH
2780 }
2781 from_header = header_last; /* To get it checked for Sender: */
2782 }
2783 }
2784
2785 /* There is a non-null envelope sender. Build the header using the original
2786 sender address, before any rewriting that might have been done while
2787 verifying it. */
2788
2789 else
2790 {
87ba3f5f 2791 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2792 oname,
2793 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2794 (sender_address_unrewritten == NULL)?
2795 sender_address : sender_address_unrewritten,
2fe1a124 2796 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2797
2798 from_header = header_last; /* To get it checked for Sender: */
2799 }
2800 }
2801
2802
8800895a
PH
2803/* If the sender is local (without suppress_local_fixups), or if we are in
2804submission mode and there is an authenticated_id, check that an existing From:
2805is correct, and if not, generate a Sender: header, unless disabled. Any
2806previously-existing Sender: header was removed above. Note that sender_local,
2807as well as being TRUE if the caller of exim is not trusted, is also true if a
2808trusted caller did not supply a -f argument for non-smtp input. To allow
2809trusted callers to forge From: without supplying -f, we have to test explicitly
2810here. If the From: header contains more than one address, then the call to
2811parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2812
2813if (from_header != NULL &&
69358f02 2814 (active_local_from_check &&
8800895a 2815 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2816 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2817 ))
2818 {
2819 BOOL make_sender = TRUE;
2820 int start, end, domain;
2821 uschar *errmess;
2822 uschar *from_address =
2823 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2824 &start, &end, &domain, FALSE);
2825 uschar *generated_sender_address;
2826
2827 if (submission_mode)
2828 {
2829 if (submission_domain == NULL)
2830 {
2831 generated_sender_address = string_sprintf("%s@%s",
2832 local_part_quote(authenticated_id), qualify_domain_sender);
2833 }
2834 else if (submission_domain[0] == 0) /* empty => full address */
2835 {
2836 generated_sender_address = string_sprintf("%s",
2837 authenticated_id);
2838 }
2839 else
2840 {
2841 generated_sender_address = string_sprintf("%s@%s",
2842 local_part_quote(authenticated_id), submission_domain);
2843 }
2844 }
2845 else
2846 generated_sender_address = string_sprintf("%s@%s",
2847 local_part_quote(originator_login), qualify_domain_sender);
2848
2849 /* Remove permitted prefixes and suffixes from the local part of the From:
2850 address before doing the comparison with the generated sender. */
2851
2852 if (from_address != NULL)
2853 {
2854 int slen;
2855 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2856
2857 if (at != NULL) *at = 0;
2858 from_address += route_check_prefix(from_address, local_from_prefix);
2859 slen = route_check_suffix(from_address, local_from_suffix);
2860 if (slen > 0)
2861 {
2862 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2863 from_address += slen;
2864 }
2865 if (at != NULL) *at = '@';
2866
2867 if (strcmpic(generated_sender_address, from_address) == 0 ||
2868 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2869 make_sender = FALSE;
2870 }
2871
2872 /* We have to cause the Sender header to be rewritten if there are
2873 appropriate rewriting rules. */
2874
2875 if (make_sender)
2876 {
2fe1a124 2877 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2878 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2879 generated_sender_address);
2880 else
2881 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2882 resent_prefix,
2883 submission_mode? submission_name : originator_name,
2884 generated_sender_address);
059ec3d9 2885 }
87ba3f5f
PH
2886
2887 /* Ensure that a non-null envelope sender address corresponds to the
2888 submission mode sender address. */
2889
2890 if (submission_mode && sender_address[0] != 0)
2891 {
2892 if (sender_address_unrewritten == NULL)
2893 sender_address_unrewritten = sender_address;
2894 sender_address = generated_sender_address;
089793a4
TF
2895 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2896 log_write(L_address_rewrite, LOG_MAIN,
2897 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2898 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2899 }
059ec3d9
PH
2900 }
2901
059ec3d9
PH
2902/* If there are any rewriting rules, apply them to the sender address, unless
2903it has already been rewritten as part of verification for SMTP input. */
2904
2905if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2906 sender_address[0] != 0)
2907 {
2908 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2909 global_rewrite_rules, rewrite_existflags);
2910 DEBUG(D_receive|D_rewrite)
2911 debug_printf("rewritten sender = %s\n", sender_address);
2912 }
2913
2914
2915/* The headers must be run through rewrite_header(), because it ensures that
2916addresses are fully qualified, as well as applying any rewriting rules that may
2917exist.
2918
2919Qualification of header addresses in a message from a remote host happens only
2920if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2921appropriate. For local messages, qualification always happens, unless -bnq is
2922used to explicitly suppress it. No rewriting is done for an unqualified address
2923that is left untouched.
2924
2925We start at the second header, skipping our own Received:. This rewriting is
2926documented as happening *after* recipient addresses are taken from the headers
2927by the -t command line option. An added Sender: gets rewritten here. */
2928
1ebe15c3 2929for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2930 {
2931 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2932 rewrite_existflags, TRUE);
1ebe15c3 2933 if (newh) h = newh;
059ec3d9
PH
2934 }
2935
2936
2937/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2938"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2939"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2940exists, this applies to the set of resent- headers rather than the normal set.
2941
2cbb4081
PH
2942The requirement for a recipient header has been removed in RFC 2822. At this
2943point in the code, earlier versions of Exim added a To: header for locally
2944submitted messages, and an empty Bcc: header for others. In the light of the
2945changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2946
059ec3d9
PH
2947
2948/* If there is no date header, generate one if the message originates locally
8800895a
PH
2949(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2950submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2951to be more confusing if Exim adds one to all remotely-originated messages.
2952As per Message-Id, we prepend if resending, else append.
2953*/
059ec3d9 2954
8800895a
PH
2955if (!date_header_exists &&
2956 ((sender_host_address == NULL && !suppress_local_fixups)
2957 || submission_mode))
e7e680d6
PP
2958 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2959 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2960
2961search_tidyup(); /* Free any cached resources */
2962
2963/* Show the complete set of headers if debugging. Note that the first one (the
2964new Received:) has not yet been set. */
2965
2966DEBUG(D_receive)
2967 {
2968 debug_printf(">>Headers after rewriting and local additions:\n");
2969 for (h = header_list->next; h != NULL; h = h->next)
2970 debug_printf("%c %s", h->type, h->text);
2971 debug_printf("\n");
2972 }
2973
2974/* The headers are now complete in store. If we are running in filter
2975testing mode, that is all this function does. Return TRUE if the message
2976ended with a dot. */
2977
f05da2e8 2978if (filter_test != FTEST_NONE)
059ec3d9
PH
2979 {
2980 process_info[process_info_len] = 0;
2981 return message_ended == END_DOT;
2982 }
2983
7e3ce68e
JH
2984/*XXX CHUNKING: need to cancel cutthrough under BDAT, for now. In future,
2985think more if it could be handled. Cannot do onward CHUNKING unless
2986inbound is, but inbound chunking ought to be ok with outbound plain.
2987Could we do onward CHUNKING given inbound CHUNKING?
2988*/
2989if (chunking_state > CHUNKING_OFFERED)
57cc2785 2990 cancel_cutthrough_connection(FALSE, US"chunking active");
7e3ce68e 2991
817d9f57 2992/* Cutthrough delivery:
5032d1cf
JH
2993We have to create the Received header now rather than at the end of reception,
2994so the timestamp behaviour is a change to the normal case.
2995XXX Ensure this gets documented XXX.
2996Having created it, send the headers to the destination. */
57cc2785
JH
2997
2998if (cutthrough.fd >= 0 && cutthrough.delivery)
e4bdf652 2999 {
817d9f57
JH
3000 if (received_count > received_headers_max)
3001 {
57cc2785 3002 cancel_cutthrough_connection(TRUE, US"too many headers");
817d9f57
JH
3003 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
3004 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3005 "Too many \"Received\" headers",
3006 sender_address,
57cc2785
JH
3007 sender_fullhost ? "H=" : "", sender_fullhost ? sender_fullhost : US"",
3008 sender_ident ? "U=" : "", sender_ident ? sender_ident : US"");
817d9f57
JH
3009 message_id[0] = 0; /* Indicate no message accepted */
3010 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
3011 goto TIDYUP; /* Skip to end of function */
3012 }
e4bdf652 3013 received_header_gen();
578d43dc 3014 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652
JH
3015 (void) cutthrough_headers_send();
3016 }
61147df4 3017
e4bdf652 3018
059ec3d9
PH
3019/* Open a new spool file for the data portion of the message. We need
3020to access it both via a file descriptor and a stream. Try to make the
41313d92 3021directory if it isn't there. */
059ec3d9 3022
41313d92 3023spool_name = spool_fname(US"input", message_subdir, message_id, US"-D");
a2da3176
JH
3024DEBUG(D_receive) debug_printf("Data file name: %s\n", spool_name);
3025
3026if ((data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE)) < 0)
059ec3d9
PH
3027 {
3028 if (errno == ENOENT)
3029 {
0971ec06 3030 (void) directory_make(spool_directory,
41313d92
JH
3031 spool_sname(US"input", message_subdir),
3032 INPUT_DIRECTORY_MODE, TRUE);
059ec3d9
PH
3033 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
3034 }
3035 if (data_fd < 0)
3036 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
3037 spool_name, strerror(errno));
3038 }
3039
3040/* Make sure the file's group is the Exim gid, and double-check the mode
3041because the group setting doesn't always get set automatically. */
3042
1ac6b2e7
JH
3043if (fchown(data_fd, exim_uid, exim_gid))
3044 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
3045 "Failed setting ownership on spool file %s: %s",
3046 spool_name, strerror(errno));
ff790e47 3047(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
3048
3049/* We now have data file open. Build a stream for it and lock it. We lock only
3050the first line of the file (containing the message ID) because otherwise there
3051are problems when Exim is run under Cygwin (I'm told). See comments in
3052spool_in.c, where the same locking is done. */
3053
3054data_file = fdopen(data_fd, "w+");
3055lock_data.l_type = F_WRLCK;
3056lock_data.l_whence = SEEK_SET;
3057lock_data.l_start = 0;
3058lock_data.l_len = SPOOL_DATA_START_OFFSET;
3059
3060if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
3061 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
3062 errno, strerror(errno));
3063
3064/* We have an open, locked data file. Write the message id to it to make it
3065self-identifying. Then read the remainder of the input of this message and
3066write it to the data file. If the variable next != NULL, it contains the first
3067data line (which was read as a header but then turned out not to have the right
3068format); write it (remembering that it might contain binary zeros). The result
3069of fwrite() isn't inspected; instead we call ferror() below. */
3070
3071fprintf(data_file, "%s-D\n", message_id);
3072if (next != NULL)
3073 {
3074 uschar *s = next->text;
3075 int len = next->slen;
1ac6b2e7 3076 len = fwrite(s, 1, len, data_file); len = len; /* compiler quietening */
059ec3d9
PH
3077 body_linecount++; /* Assumes only 1 line */
3078 }
3079
3080/* Note that we might already be at end of file, or the logical end of file
3081(indicated by '.'), or might have encountered an error while writing the
3082message id or "next" line. */
3083
3084if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
3085 {
3086 if (smtp_input)
3087 {
7e3ce68e
JH
3088 message_ended = chunking_state > CHUNKING_OFFERED
3089 ? read_message_bdat_smtp(data_file)
3090 : read_message_data_smtp(data_file);
059ec3d9
PH
3091 receive_linecount++; /* The terminating "." line */
3092 }
3093 else message_ended = read_message_data(data_file);
3094
3095 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 3096 message_linecount += body_linecount;
059ec3d9 3097
7e3ce68e 3098 switch (message_ended)
059ec3d9 3099 {
7e3ce68e 3100 /* Handle premature termination of SMTP */
059ec3d9 3101
7e3ce68e
JH
3102 case END_EOF:
3103 if (smtp_input)
3104 {
3105 Uunlink(spool_name); /* Lose data file when closed */
57cc2785 3106 cancel_cutthrough_connection(TRUE, US"sender closed connection");
7e3ce68e
JH
3107 message_id[0] = 0; /* Indicate no message accepted */
3108 smtp_reply = handle_lost_connection(US"");
3109 smtp_yield = FALSE;
3110 goto TIDYUP; /* Skip to end of function */
3111 }
3112 break;
059ec3d9 3113
7e3ce68e
JH
3114 /* Handle message that is too big. Don't use host_or_ident() in the log
3115 message; we want to see the ident value even for non-remote messages. */
059ec3d9 3116
7e3ce68e
JH
3117 case END_SIZE:
3118 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3119 cancel_cutthrough_connection(TRUE, US"mail too big");
7e3ce68e 3120 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
059ec3d9 3121
7e3ce68e
JH
3122 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3123 "message too big: read=%d max=%d",
3124 sender_address,
3125 (sender_fullhost == NULL)? "" : " H=",
3126 (sender_fullhost == NULL)? US"" : sender_fullhost,
3127 (sender_ident == NULL)? "" : " U=",
3128 (sender_ident == NULL)? US"" : sender_ident,
3129 message_size,
3130 thismessage_size_limit);
3131
3132 if (smtp_input)
3133 {
3134 smtp_reply = US"552 Message size exceeds maximum permitted";
3135 message_id[0] = 0; /* Indicate no message accepted */
3136 goto TIDYUP; /* Skip to end of function */
3137 }
3138 else
3139 {
3140 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3141 give_local_error(ERRMESS_TOOBIG,
3142 string_sprintf("message too big (max=%d)", thismessage_size_limit),
3143 US"message rejected: ", error_rc, data_file, header_list);
3144 /* Does not return */
3145 }
3146 break;
3147
3148 /* Handle bad BDAT protocol sequence */
3149
3150 case END_PROTOCOL:
3151 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3152 cancel_cutthrough_connection(TRUE, US"sender protocol error");
7e3ce68e
JH
3153 smtp_reply = US""; /* Response already sent */
3154 message_id[0] = 0; /* Indicate no message accepted */
3155 goto TIDYUP; /* Skip to end of function */
059ec3d9
PH
3156 }
3157 }
3158
3159/* Restore the standard SIGALRM handler for any subsequent processing. (For
3160example, there may be some expansion in an ACL that uses a timer.) */
3161
3162os_non_restarting_signal(SIGALRM, sigalrm_handler);
3163
3164/* The message body has now been read into the data file. Call fflush() to
3165empty the buffers in C, and then call fsync() to get the data written out onto
3166the disk, as fflush() doesn't do this (or at least, it isn't documented as
3167having to do this). If there was an I/O error on either input or output,
3168attempt to send an error message, and unlink the spool file. For non-SMTP input
3169we can then give up. Note that for SMTP input we must swallow the remainder of
3170the input in cases of output errors, since the far end doesn't expect to see
3171anything until the terminating dot line is sent. */
3172
3173if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 3174 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
3175 {
3176 uschar *msg_errno = US strerror(errno);
3177 BOOL input_error = (receive_ferror)() != 0;
3178 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
3179 input_error? "Input read" : "Spool write",
3180 msg_errno,
3181 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
3182
3183 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
3184 Uunlink(spool_name); /* Lose the data file */
57cc2785 3185 cancel_cutthrough_connection(TRUE, US"error writing spoolfile");
059ec3d9
PH
3186
3187 if (smtp_input)
3188 {
3189 if (input_error)
3190 smtp_reply = US"451 Error while reading input data";
3191 else
3192 {
3193 smtp_reply = US"451 Error while writing spool file";
3194 receive_swallow_smtp();
3195 }
3196 message_id[0] = 0; /* Indicate no message accepted */
3197 goto TIDYUP; /* Skip to end of function */
3198 }
3199
3200 else
3201 {
3202 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3203 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
3204 header_list);
3205 /* Does not return */
3206 }
3207 }
3208
3209
3210/* No I/O errors were encountered while writing the data file. */
3211
3212DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
3213
3214
3215/* If there were any bad addresses extracted by -t, or there were no recipients
3216left after -t, send a message to the sender of this message, or write it to
3217stderr if the error handling option is set that way. Note that there may
3218legitimately be no recipients for an SMTP message if they have all been removed
3219by "discard".
3220
3221We need to rewind the data file in order to read it. In the case of no
3222recipients or stderr error writing, throw the data file away afterwards, and
3223exit. (This can't be SMTP, which always ensures there's at least one
3224syntactically good recipient address.) */
3225
3226if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
3227 {
3228 DEBUG(D_receive)
3229 {
3230 if (recipients_count == 0) debug_printf("*** No recipients\n");
3231 if (bad_addresses != NULL)
3232 {
3233 error_block *eblock = bad_addresses;
3234 debug_printf("*** Bad address(es)\n");
3235 while (eblock != NULL)
3236 {
3237 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
3238 eblock = eblock->next;
3239 }
3240 }
3241 }
3242
3243 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3244
3245 /* If configured to send errors to the sender, but this fails, force
3246 a failure error code. We use a special one for no recipients so that it
3247 can be detected by the autoreply transport. Otherwise error_rc is set to
3248 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
3249 it is EXIT_SUCCESS. */
3250
3251 if (error_handling == ERRORS_SENDER)
3252 {
3253 if (!moan_to_sender(
3254 (bad_addresses == NULL)?
3255 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
3256 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
3257 bad_addresses, header_list, data_file, FALSE))
3258 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
3259 }
3260 else
3261 {
3262 if (bad_addresses == NULL)
3263 {
3264 if (extracted_ignored)
3265 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3266 else
3267 fprintf(stderr, "exim: no recipients in message\n");
3268 }
3269 else
3270 {
3271 fprintf(stderr, "exim: invalid address%s",
3272 (bad_addresses->next == NULL)? ":" : "es:\n");
3273 while (bad_addresses != NULL)
3274 {
3275 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3276 bad_addresses->text2);
3277 bad_addresses = bad_addresses->next;
3278 }
3279 }
3280 }
3281
3282 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3283 {
3284 Uunlink(spool_name);
f1e894f3 3285 (void)fclose(data_file);
059ec3d9
PH
3286 exim_exit(error_rc);
3287 }
3288 }
3289
3290/* Data file successfully written. Generate text for the Received: header by
3291expanding the configured string, and adding a timestamp. By leaving this
3292operation till now, we ensure that the timestamp is the time that message
3293reception was completed. However, this is deliberately done before calling the
3294data ACL and local_scan().
3295
3296This Received: header may therefore be inspected by the data ACL and by code in
3297the local_scan() function. When they have run, we update the timestamp to be
3298the final time of reception.
3299
3300If there is just one recipient, set up its value in the $received_for variable
3301for use when we generate the Received: header.
3302
3303Note: the checking for too many Received: headers is handled by the delivery
3304code. */
e4bdf652 3305/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3306
e4bdf652 3307if (received_header->text == NULL) /* Non-cutthrough case */
059ec3d9 3308 {
e4bdf652 3309 received_header_gen();
059ec3d9 3310
e4bdf652 3311 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3312
e4bdf652
JH
3313 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3314 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3315
e4bdf652
JH
3316 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3317 now, before running the DATA ACL. */
059ec3d9 3318
578d43dc 3319 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652 3320 }
817d9f57 3321else
e4bdf652
JH
3322 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3323 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3324
3325/* If an ACL is specified for checking things at this stage of reception of a
3326message, run it, unless all the recipients were removed by "discard" in earlier
3327ACLs. That is the only case in which recipients_count can be zero at this
3328stage. Set deliver_datafile to point to the data file so that $message_body and
3329$message_body_end can be extracted if needed. Allow $recipients in expansions.
3330*/
3331
3332deliver_datafile = data_fd;
4e88a19f 3333user_msg = NULL;
059ec3d9 3334
0e20aff9
MH
3335enable_dollar_recipients = TRUE;
3336
059ec3d9 3337if (recipients_count == 0)
7e3ce68e
JH
3338 blackholed_by = recipients_discarded ? US"MAIL ACL" : US"RCPT ACL";
3339
059ec3d9
PH
3340else
3341 {
059ec3d9
PH
3342 /* Handle interactive SMTP messages */
3343
3344 if (smtp_input && !smtp_batched_input)
3345 {
8523533c 3346
80a47a2c
TK
3347#ifndef DISABLE_DKIM
3348 if (!dkim_disable_verify)
3349 {
3350 /* Finish verification, this will log individual signature results to
3351 the mainlog */
3352 dkim_exim_verify_finish();
3353
3354 /* Check if we must run the DKIM ACL */
7e3ce68e 3355 if (acl_smtp_dkim && dkim_verify_signers && *dkim_verify_signers)
80a47a2c
TK
3356 {
3357 uschar *dkim_verify_signers_expanded =
3358 expand_string(dkim_verify_signers);
7e3ce68e 3359 if (!dkim_verify_signers_expanded)
80a47a2c
TK
3360 log_write(0, LOG_MAIN|LOG_PANIC,
3361 "expansion of dkim_verify_signers option failed: %s",
3362 expand_string_message);
7e3ce68e 3363
80a47a2c
TK
3364 else
3365 {
3366 int sep = 0;
55414b25 3367 const uschar *ptr = dkim_verify_signers_expanded;
80a47a2c 3368 uschar *item = NULL;
9e5d6b55
TK
3369 uschar *seen_items = NULL;
3370 int seen_items_size = 0;
3371 int seen_items_offset = 0;
9122af94
TK
3372 /* Default to OK when no items are present */
3373 rc = OK;
7e3ce68e 3374 while ((item = string_nextinlist(&ptr, &sep, NULL, 0)))
80a47a2c 3375 {
6119d1ea 3376 /* Prevent running ACL for an empty item */
7e3ce68e 3377 if (!item || !*item) continue;
5032d1cf
JH
3378
3379 /* Only run ACL once for each domain or identity,
3380 no matter how often it appears in the expanded list. */
3381 if (seen_items)
6119d1ea 3382 {
ae9094bf 3383 uschar *seen_item = NULL;
55414b25 3384 const uschar *seen_items_list = seen_items;
5032d1cf 3385 BOOL seen_this_item = FALSE;
61147df4 3386
ae9094bf 3387 while ((seen_item = string_nextinlist(&seen_items_list, &sep,
7e3ce68e 3388 NULL, 0)))
5032d1cf
JH
3389 if (Ustrcmp(seen_item,item) == 0)
3390 {
3391 seen_this_item = TRUE;
3392 break;
3393 }
3394
3395 if (seen_this_item)
6119d1ea
TK
3396 {
3397 DEBUG(D_receive)
5032d1cf
JH
3398 debug_printf("acl_smtp_dkim: skipping signer %s, "
3399 "already seen\n", item);
6119d1ea
TK
3400 continue;
3401 }
61147df4 3402
5032d1cf
JH
3403 seen_items = string_append(seen_items, &seen_items_size,
3404 &seen_items_offset, 1, ":");
6119d1ea
TK
3405 }
3406
5032d1cf
JH
3407 seen_items = string_append(seen_items, &seen_items_size,
3408 &seen_items_offset, 1, item);
4a73449b 3409 seen_items[seen_items_offset] = '\0';
6119d1ea
TK
3410
3411 DEBUG(D_receive)
5032d1cf
JH
3412 debug_printf("calling acl_smtp_dkim for dkim_cur_signer=%s\n",
3413 item);
6119d1ea 3414
80a47a2c 3415 dkim_exim_acl_setup(item);
5032d1cf
JH
3416 rc = acl_check(ACL_WHERE_DKIM, NULL, acl_smtp_dkim,
3417 &user_msg, &log_msg);
6119d1ea
TK
3418
3419 if (rc != OK)
5032d1cf
JH
3420 {
3421 DEBUG(D_receive)
3422 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, "
3423 "skipping remaining items\n", rc, item);
57cc2785 3424 cancel_cutthrough_connection(TRUE, US"dkim acl not ok");
5032d1cf
JH
3425 break;
3426 }
80a47a2c 3427 }
578d43dc 3428 add_acl_headers(ACL_WHERE_DKIM, US"DKIM");
80a47a2c
TK
3429 if (rc == DISCARD)
3430 {
3431 recipients_count = 0;
3432 blackholed_by = US"DKIM ACL";
3433 if (log_msg != NULL)
3434 blackhole_log_msg = string_sprintf(": %s", log_msg);
3435 }
3436 else if (rc != OK)
3437 {
3438 Uunlink(spool_name);
3439 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
85ffcba6 3440 smtp_yield = FALSE; /* No more messages after dropped connection */
80a47a2c
TK
3441 smtp_reply = US""; /* Indicate reply already sent */
3442 message_id[0] = 0; /* Indicate no message accepted */
3443 goto TIDYUP; /* Skip to end of function */
3444 }
3445 }
3446 }
3447 }
4a8ce2d8 3448#endif /* DISABLE_DKIM */
fb2274d4 3449
8523533c 3450#ifdef WITH_CONTENT_SCAN
80a47a2c
TK
3451 if (recipients_count > 0 &&
3452 acl_smtp_mime != NULL &&
54cdb463
PH
3453 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
3454 goto TIDYUP;
8523533c
TK
3455#endif /* WITH_CONTENT_SCAN */
3456
4840604e
TL
3457#ifdef EXPERIMENTAL_DMARC
3458 dmarc_up = dmarc_store_data(from_header);
3459#endif /* EXPERIMENTAL_DMARC */
3460
8ccd00b1
JH
3461#ifndef DISABLE_PRDR
3462 if (prdr_requested && recipients_count > 1 && acl_smtp_data_prdr)
fd98a5c6
JH
3463 {
3464 unsigned int c;
3465 int all_pass = OK;
3466 int all_fail = FAIL;
3467
3468 smtp_printf("353 PRDR content analysis beginning\r\n");
3469 /* Loop through recipients, responses must be in same order received */
3470 for (c = 0; recipients_count > c; c++)
3471 {
3472 uschar * addr= recipients_list[c].address;
3473 uschar * msg= US"PRDR R=<%s> %s";
3474 uschar * code;
3475 DEBUG(D_receive)
3476 debug_printf("PRDR processing recipient %s (%d of %d)\n",
3477 addr, c+1, recipients_count);
3478 rc = acl_check(ACL_WHERE_PRDR, addr,
3479 acl_smtp_data_prdr, &user_msg, &log_msg);
3480
3481 /* If any recipient rejected content, indicate it in final message */
3482 all_pass |= rc;
3483 /* If all recipients rejected, indicate in final message */
3484 all_fail &= rc;
3485
3486 switch (rc)
3487 {
3488 case OK: case DISCARD: code = US"250"; break;
3489 case DEFER: code = US"450"; break;
3490 default: code = US"550"; break;
3491 }
3492 if (user_msg != NULL)
3493 smtp_user_msg(code, user_msg);
3494 else
3495 {
3496 switch (rc)
3497 {
3498 case OK: case DISCARD:
3499 msg = string_sprintf(CS msg, addr, "acceptance"); break;
3500 case DEFER:
3501 msg = string_sprintf(CS msg, addr, "temporary refusal"); break;
3502 default:
3503 msg = string_sprintf(CS msg, addr, "refusal"); break;
3504 }
3505 smtp_user_msg(code, msg);
3506 }
3507 if (log_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, log_msg);
3508 else if (user_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, user_msg);
112b6a93 3509 else log_write(0, LOG_MAIN, "%s", CS msg);
fd98a5c6
JH
3510
3511 if (rc != OK) { receive_remove_recipient(addr); c--; }
3512 }
3513 /* Set up final message, used if data acl gives OK */
3514 smtp_reply = string_sprintf("%s id=%s message %s",
3515 all_fail == FAIL ? US"550" : US"250",
3516 message_id,
3517 all_fail == FAIL
3518 ? US"rejected for all recipients"
3519 : all_pass == OK
3520 ? US"accepted"
3521 : US"accepted for some recipients");
3522 if (recipients_count == 0)
3523 {
3524 message_id[0] = 0; /* Indicate no message accepted */
3525 goto TIDYUP;
3526 }
3527 }
3528 else
3529 prdr_requested = FALSE;
8ccd00b1 3530#endif /* !DISABLE_PRDR */
fd98a5c6 3531
54cdb463
PH
3532 /* Check the recipients count again, as the MIME ACL might have changed
3533 them. */
8523533c 3534
059ec3d9
PH
3535 if (acl_smtp_data != NULL && recipients_count > 0)
3536 {
059ec3d9 3537 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
578d43dc 3538 add_acl_headers(ACL_WHERE_DATA, US"DATA");
059ec3d9
PH
3539 if (rc == DISCARD)
3540 {
3541 recipients_count = 0;
3542 blackholed_by = US"DATA ACL";
57cc2785 3543 if (log_msg)
8e669ac1 3544 blackhole_log_msg = string_sprintf(": %s", log_msg);
57cc2785 3545 cancel_cutthrough_connection(TRUE, US"data acl discard");
059ec3d9
PH
3546 }
3547 else if (rc != OK)
3548 {
3549 Uunlink(spool_name);
57cc2785 3550 cancel_cutthrough_connection(TRUE, US"data acl not ok");
8523533c
TK
3551#ifdef WITH_CONTENT_SCAN
3552 unspool_mbox();
3553#endif
6f0c431a
PP
3554#ifdef EXPERIMENTAL_DCC
3555 dcc_ok = 0;
3556#endif
059ec3d9 3557 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
85ffcba6 3558 smtp_yield = FALSE; /* No more messages after dropped connection */
059ec3d9
PH
3559 smtp_reply = US""; /* Indicate reply already sent */
3560 message_id[0] = 0; /* Indicate no message accepted */
3561 goto TIDYUP; /* Skip to end of function */
3562 }
3563 }
3564 }
3565
3566 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
3567 we cannot take different actions for permanent and temporary rejections. */
3568
54cdb463 3569 else
059ec3d9 3570 {
54cdb463
PH
3571
3572#ifdef WITH_CONTENT_SCAN
3573 if (acl_not_smtp_mime != NULL &&
3574 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
3575 &blackholed_by))
3576 goto TIDYUP;
3577#endif /* WITH_CONTENT_SCAN */
3578
3579 if (acl_not_smtp != NULL)
059ec3d9 3580 {
54cdb463
PH
3581 uschar *user_msg, *log_msg;
3582 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
3583 if (rc == DISCARD)
059ec3d9 3584 {
54cdb463
PH
3585 recipients_count = 0;
3586 blackholed_by = US"non-SMTP ACL";
3587 if (log_msg != NULL)
3588 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 3589 }
54cdb463 3590 else if (rc != OK)
059ec3d9 3591 {
54cdb463
PH
3592 Uunlink(spool_name);
3593#ifdef WITH_CONTENT_SCAN
3594 unspool_mbox();
3595#endif
6f0c431a
PP
3596#ifdef EXPERIMENTAL_DCC
3597 dcc_ok = 0;
3598#endif
6ea85e9a
PH
3599 /* The ACL can specify where rejections are to be logged, possibly
3600 nowhere. The default is main and reject logs. */
3601
3602 if (log_reject_target != 0)
3603 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3604 sender_address, log_msg);
3605
54cdb463
PH
3606 if (user_msg == NULL) user_msg = US"local configuration problem";
3607 if (smtp_batched_input)
3608 {
3609 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3610 /* Does not return */
3611 }
3612 else
3613 {
3614 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3615 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3616 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3617 header_list);
3618 /* Does not return */
3619 }
059ec3d9 3620 }
578d43dc 3621 add_acl_headers(ACL_WHERE_NOTSMTP, US"non-SMTP");
059ec3d9 3622 }
059ec3d9
PH
3623 }
3624
54cdb463
PH
3625 /* The applicable ACLs have been run */
3626
059ec3d9
PH
3627 if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */
3628 if (queue_only_policy) queued_by = US"ACL";
059ec3d9
PH
3629 }
3630
8523533c
TK
3631#ifdef WITH_CONTENT_SCAN
3632unspool_mbox();
3633#endif
3634
6a8f9482
TK
3635#ifdef EXPERIMENTAL_DCC
3636dcc_ok = 0;
3637#endif
3638
3639
059ec3d9
PH
3640/* The final check on the message is to run the scan_local() function. The
3641version supplied with Exim always accepts, but this is a hook for sysadmins to
3642supply their own checking code. The local_scan() function is run even when all
3643the recipients have been discarded. */
3644
3645lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3646
3647/* Arrange to catch crashes in local_scan(), so that the -D file gets
3648deleted, and the incident gets logged. */
3649
3650os_non_restarting_signal(SIGSEGV, local_scan_crash_handler);
3651os_non_restarting_signal(SIGFPE, local_scan_crash_handler);
3652os_non_restarting_signal(SIGILL, local_scan_crash_handler);
3653os_non_restarting_signal(SIGBUS, local_scan_crash_handler);
3654
3655DEBUG(D_receive) debug_printf("calling local_scan(); timeout=%d\n",
3656 local_scan_timeout);
3657local_scan_data = NULL;
3658
3659os_non_restarting_signal(SIGALRM, local_scan_timeout_handler);
3660if (local_scan_timeout > 0) alarm(local_scan_timeout);
3661rc = local_scan(data_fd, &local_scan_data);
3662alarm(0);
3663os_non_restarting_signal(SIGALRM, sigalrm_handler);
3664
0e20aff9
MH
3665enable_dollar_recipients = FALSE;
3666
059ec3d9
PH
3667store_pool = POOL_MAIN; /* In case changed */
3668DEBUG(D_receive) debug_printf("local_scan() returned %d %s\n", rc,
3669 local_scan_data);
3670
3671os_non_restarting_signal(SIGSEGV, SIG_DFL);
3672os_non_restarting_signal(SIGFPE, SIG_DFL);
3673os_non_restarting_signal(SIGILL, SIG_DFL);
3674os_non_restarting_signal(SIGBUS, SIG_DFL);
3675
3676/* The length check is paranoia against some runaway code, and also because
3677(for a success return) lines in the spool file are read into big_buffer. */
3678