testsuite output changes
[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
328c5688
JH
1019static int
1020read_message_bdat_smtp_wire(FILE *fout)
1021{
1022int ch;
1023
1024/* Remember that this message uses wireformat. */
1025
1026DEBUG(D_receive) debug_printf("CHUNKING: writing spoolfile in wire format\n");
1027spool_file_wireformat = TRUE;
1028
1029/* Unfortunately cannot use sendfile() even if not TLS
1030as that requires (on linux) mmap-like operations on the input fd.
1031
1032XXX but worthwhile doing a block interface to the bdat_getc buffer
1033in the future */
1034
1035for (;;) switch (ch = bdat_getc(GETC_BUFFER_UNLIMITED))
1036 {
1037 case EOF: return END_EOF;
1038 case EOD: return END_DOT;
1039 case ERR: return END_PROTOCOL;
1040
1041 default:
1042 message_size++;
1043/*XXX not done:
1044linelength
1045max_received_linelength
1046body_linecount
1047body_zerocount
1048*/
1049 if (fout)
1050 {
1051 if (fputc(ch, fout) == EOF) return END_WERROR;
1052 if (message_size > thismessage_size_limit) return END_SIZE;
1053 }
1054 break;
1055 }
1056/*NOTREACHED*/
1057}
1058
7e3ce68e
JH
1059
1060
1061
059ec3d9
PH
1062/*************************************************
1063* Swallow SMTP message *
1064*************************************************/
1065
1066/* This function is called when there has been some kind of error while reading
1067an SMTP message, and the remaining data may need to be swallowed. It is global
1068because it is called from smtp_closedown() to shut down an incoming call
1069tidily.
1070
1071Argument: a FILE from which to read the message
1072Returns: nothing
1073*/
1074
1075void
1076receive_swallow_smtp(void)
1077{
7e3ce68e 1078/*XXX CHUNKING: not enough. read chunks until RSET? */
059ec3d9
PH
1079if (message_ended >= END_NOTENDED)
1080 message_ended = read_message_data_smtp(NULL);
1081}
1082
1083
1084
1085/*************************************************
1086* Handle lost SMTP connection *
1087*************************************************/
1088
1089/* This function logs connection loss incidents and generates an appropriate
1090SMTP response.
1091
1092Argument: additional data for the message
1093Returns: the SMTP response
1094*/
1095
1096static uschar *
1097handle_lost_connection(uschar *s)
1098{
1099log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
1100 "%s lost while reading message data%s", smtp_get_connection_info(), s);
eea0defe 1101smtp_notquit_exit(US"connection-lost", NULL, NULL);
059ec3d9
PH
1102return US"421 Lost incoming connection";
1103}
1104
1105
1106
1107
1108/*************************************************
1109* Handle a non-smtp reception error *
1110*************************************************/
1111
1112/* This function is called for various errors during the reception of non-SMTP
1113messages. It either sends a message to the sender of the problem message, or it
1114writes to the standard error stream.
1115
1116Arguments:
1117 errcode code for moan_to_sender(), identifying the error
1118 text1 first message text, passed to moan_to_sender()
1119 text2 second message text, used only for stderrr
1120 error_rc code to pass to exim_exit if no problem
1121 f FILE containing body of message (may be stdin)
1122 hptr pointer to instore headers or NULL
1123
1124Returns: calls exim_exit(), which does not return
1125*/
1126
1127static void
1128give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
1129 FILE *f, header_line *hptr)
1130{
1131if (error_handling == ERRORS_SENDER)
1132 {
1133 error_block eblock;
1134 eblock.next = NULL;
1135 eblock.text1 = text1;
37f3dc43 1136 eblock.text2 = US"";
059ec3d9
PH
1137 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
1138 error_rc = EXIT_FAILURE;
1139 }
37f3dc43
JH
1140else
1141 fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 1142(void)fclose(f);
059ec3d9
PH
1143exim_exit(error_rc);
1144}
1145
1146
1147
1148/*************************************************
1149* Add header lines set up by ACL *
1150*************************************************/
1151
850635b6
PH
1152/* This function is called to add the header lines that were set up by
1153statements in an ACL to the list of headers in memory. It is done in two stages
1154like this, because when the ACL for RCPT is running, the other headers have not
1155yet been received. This function is called twice; once just before running the
1156DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
1157are visible to the DATA ACL.
059ec3d9
PH
1158
1159Originally these header lines were added at the end. Now there is support for
1160three different places: top, bottom, and after the Received: header(s). There
1161will always be at least one Received: header, even if it is marked deleted, and
1162even if something else has been put in front of it.
1163
1164Arguments:
1165 acl_name text to identify which ACL
1166
1167Returns: nothing
1168*/
1169
1170static void
578d43dc 1171add_acl_headers(int where, uschar *acl_name)
059ec3d9
PH
1172{
1173header_line *h, *next;
1174header_line *last_received = NULL;
e7568d51 1175
578d43dc
JH
1176switch(where)
1177 {
1178 case ACL_WHERE_DKIM:
1179 case ACL_WHERE_MIME:
af4a1bca 1180 case ACL_WHERE_DATA:
57cc2785
JH
1181 if ( cutthrough.fd >= 0 && cutthrough.delivery
1182 && (acl_removed_headers || acl_added_headers))
578d43dc
JH
1183 {
1184 log_write(0, LOG_MAIN|LOG_PANIC, "Header modification in data ACLs"
af4a1bca 1185 " will not take effect on cutthrough deliveries");
578d43dc
JH
1186 return;
1187 }
1188 }
1189
57cc2785 1190if (acl_removed_headers)
e7568d51 1191 {
e1d04f48 1192 DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers removed by %s ACL:\n", acl_name);
e7568d51 1193
57cc2785 1194 for (h = header_list; h; h = h->next) if (h->type != htype_old)
e7568d51 1195 {
55414b25 1196 const uschar * list = acl_removed_headers;
e7568d51
TL
1197 int sep = ':'; /* This is specified as a colon-separated list */
1198 uschar *s;
1199 uschar buffer[128];
4a142059
JH
1200
1201 while ((s = string_nextinlist(&list, &sep, buffer, sizeof(buffer))))
1202 if (header_testname(h, s, Ustrlen(s), FALSE))
e7568d51
TL
1203 {
1204 h->type = htype_old;
e1d04f48 1205 DEBUG(D_receive|D_acl) debug_printf_indent(" %s", h->text);
e7568d51 1206 }
e7568d51
TL
1207 }
1208 acl_removed_headers = NULL;
e1d04f48 1209 DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
e7568d51 1210 }
059ec3d9 1211
57cc2785 1212if (!acl_added_headers) return;
e1d04f48 1213DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers added by %s ACL:\n", acl_name);
059ec3d9 1214
57cc2785 1215for (h = acl_added_headers; h; h = next)
059ec3d9
PH
1216 {
1217 next = h->next;
1218
1219 switch(h->type)
1220 {
1221 case htype_add_top:
1222 h->next = header_list;
1223 header_list = h;
e1d04f48 1224 DEBUG(D_receive|D_acl) debug_printf_indent(" (at top)");
059ec3d9
PH
1225 break;
1226
1227 case htype_add_rec:
1228 if (last_received == NULL)
1229 {
1230 last_received = header_list;
1231 while (!header_testname(last_received, US"Received", 8, FALSE))
1232 last_received = last_received->next;
1233 while (last_received->next != NULL &&
1234 header_testname(last_received->next, US"Received", 8, FALSE))
1235 last_received = last_received->next;
1236 }
1237 h->next = last_received->next;
1238 last_received->next = h;
e1d04f48 1239 DEBUG(D_receive|D_acl) debug_printf_indent(" (after Received:)");
059ec3d9
PH
1240 break;
1241
8523533c
TK
1242 case htype_add_rfc:
1243 /* add header before any header which is NOT Received: or Resent- */
1244 last_received = header_list;
1245 while ( (last_received->next != NULL) &&
1246 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
1247 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
1248 last_received = last_received->next;
1249 /* last_received now points to the last Received: or Resent-* header
1250 in an uninterrupted chain of those header types (seen from the beginning
1251 of all headers. Our current header must follow it. */
1252 h->next = last_received->next;
1253 last_received->next = h;
e1d04f48 1254 DEBUG(D_receive|D_acl) debug_printf_indent(" (before any non-Received: or Resent-*: header)");
8523533c
TK
1255 break;
1256
059ec3d9
PH
1257 default:
1258 h->next = NULL;
1259 header_last->next = h;
1260 break;
1261 }
1262
1263 if (h->next == NULL) header_last = h;
1264
1265 /* Check for one of the known header types (From:, To:, etc.) though in
1266 practice most added headers are going to be "other". Lower case
1267 identification letters are never stored with the header; they are used
1268 for existence tests when messages are received. So discard any lower case
1269 flag values. */
1270
1271 h->type = header_checkname(h, FALSE);
1272 if (h->type >= 'a') h->type = htype_other;
1273
e1d04f48 1274 DEBUG(D_receive|D_acl) debug_printf_indent(" %s", header_last->text);
059ec3d9
PH
1275 }
1276
71fafd95 1277acl_added_headers = NULL;
e1d04f48 1278DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
059ec3d9
PH
1279}
1280
1281
1282
1283/*************************************************
1284* Add host information for log line *
1285*************************************************/
1286
1287/* Called for acceptance and rejecting log lines. This adds information about
1288the calling host to a string that is being built dynamically.
1289
1290Arguments:
1291 s the dynamic string
1292 sizeptr points to the size variable
1293 ptrptr points to the pointer variable
1294
1295Returns: the extended string
1296*/
1297
1298static uschar *
fc16abb4 1299add_host_info_for_log(uschar * s, int * sizeptr, int * ptrptr)
059ec3d9 1300{
fc16abb4 1301if (sender_fullhost)
059ec3d9 1302 {
fc16abb4
JH
1303 if (LOGGING(dnssec) && sender_host_dnssec) /*XXX sender_helo_dnssec? */
1304 s = string_cat(s, sizeptr, ptrptr, US" DS");
059ec3d9 1305 s = string_append(s, sizeptr, ptrptr, 2, US" H=", sender_fullhost);
6c6d6e48 1306 if (LOGGING(incoming_interface) && interface_address != NULL)
059ec3d9 1307 {
fc16abb4
JH
1308 s = string_cat(s, sizeptr, ptrptr,
1309 string_sprintf(" I=[%s]:%d", interface_address, interface_port));
059ec3d9
PH
1310 }
1311 }
1312if (sender_ident != NULL)
1313 s = string_append(s, sizeptr, ptrptr, 2, US" U=", sender_ident);
1314if (received_protocol != NULL)
1315 s = string_append(s, sizeptr, ptrptr, 2, US" P=", received_protocol);
1316return s;
1317}
1318
1319
1320
63955bf2 1321#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1322
1323/*************************************************
54cdb463
PH
1324* Run the MIME ACL on a message *
1325*************************************************/
1326
1327/* This code is in a subroutine so that it can be used for both SMTP
1328and non-SMTP messages. It is called with a non-NULL ACL pointer.
1329
1330Arguments:
1331 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1332 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1333 smtp_reply_ptr Where SMTP reply is being built
1334 blackholed_by_ptr Where "blackholed by" message is being built
1335
1336Returns: TRUE to carry on; FALSE to abandon the message
1337*/
1338
1339static BOOL
1340run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1341 uschar **blackholed_by_ptr)
1342{
1343FILE *mbox_file;
1344uschar rfc822_file_path[2048];
1345unsigned long mbox_size;
1346header_line *my_headerlist;
1347uschar *user_msg, *log_msg;
1348int mime_part_count_buffer = -1;
040721f2 1349uschar * mbox_filename;
7156b1ef 1350int rc = OK;
54cdb463
PH
1351
1352memset(CS rfc822_file_path,0,2048);
1353
1354/* check if it is a MIME message */
040721f2
JH
1355
1356for (my_headerlist = header_list; my_headerlist; my_headerlist = my_headerlist->next)
1357 if ( my_headerlist->type != '*' /* skip deleted headers */
1358 && strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0
1359 )
4e88a19f 1360 {
54cdb463
PH
1361 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1362 goto DO_MIME_ACL;
4e88a19f 1363 }
54cdb463
PH
1364
1365DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1366return TRUE;
1367
1368DO_MIME_ACL:
040721f2 1369
54cdb463 1370/* make sure the eml mbox file is spooled up */
040721f2
JH
1371if (!(mbox_file = spool_mbox(&mbox_size, NULL, &mbox_filename)))
1372 { /* error while spooling */
54cdb463
PH
1373 log_write(0, LOG_MAIN|LOG_PANIC,
1374 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1375 Uunlink(spool_name);
1376 unspool_mbox();
6f0c431a
PP
1377#ifdef EXPERIMENTAL_DCC
1378 dcc_ok = 0;
1379#endif
a5bd321b 1380 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1381 message_id[0] = 0; /* Indicate no message accepted */
1382 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1383 return FALSE; /* Indicate skip to end of receive function */
040721f2 1384 }
54cdb463
PH
1385
1386mime_is_rfc822 = 0;
1387
1388MIME_ACL_CHECK:
1389mime_part_count = -1;
1390rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1391(void)fclose(mbox_file);
54cdb463 1392
4e88a19f
PH
1393if (Ustrlen(rfc822_file_path) > 0)
1394 {
54cdb463
PH
1395 mime_part_count = mime_part_count_buffer;
1396
4e88a19f
PH
1397 if (unlink(CS rfc822_file_path) == -1)
1398 {
54cdb463
PH
1399 log_write(0, LOG_PANIC,
1400 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1401 goto END_MIME_ACL;
4e88a19f
PH
1402 }
1403 }
54cdb463
PH
1404
1405/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1406if (rc == OK)
1407 {
040721f2 1408 uschar * scandir;
e8bc7fca
JH
1409 struct dirent * entry;
1410 DIR * tempdir;
54cdb463 1411
040721f2 1412 scandir = string_copyn(mbox_filename, Ustrrchr(mbox_filename, '/') - mbox_filename);
54cdb463 1413
040721f2 1414 tempdir = opendir(CS scandir);
e8bc7fca 1415 for (;;)
4e88a19f 1416 {
e8bc7fca
JH
1417 if (!(entry = readdir(tempdir)))
1418 break;
1419 if (strncmpic(US entry->d_name, US"__rfc822_", 9) == 0)
4e88a19f 1420 {
e8bc7fca 1421 (void) string_format(rfc822_file_path, sizeof(rfc822_file_path),
040721f2 1422 "%s/%s", scandir, entry->d_name);
e1d04f48 1423 DEBUG(D_receive) debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
e8bc7fca 1424 rfc822_file_path);
4e88a19f
PH
1425 break;
1426 }
e8bc7fca 1427 }
4e88a19f 1428 closedir(tempdir);
54cdb463 1429
e8bc7fca 1430 if (entry)
4e88a19f 1431 {
e8bc7fca 1432 if ((mbox_file = Ufopen(rfc822_file_path, "rb")))
4e88a19f 1433 {
e8bc7fca
JH
1434 /* set RFC822 expansion variable */
1435 mime_is_rfc822 = 1;
1436 mime_part_count_buffer = mime_part_count;
1437 goto MIME_ACL_CHECK;
4e88a19f 1438 }
e8bc7fca
JH
1439 log_write(0, LOG_PANIC,
1440 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1441 unlink(CS rfc822_file_path);
4e88a19f
PH
1442 }
1443 }
54cdb463
PH
1444
1445END_MIME_ACL:
578d43dc 1446add_acl_headers(ACL_WHERE_MIME, US"MIME");
54cdb463
PH
1447if (rc == DISCARD)
1448 {
1449 recipients_count = 0;
1450 *blackholed_by_ptr = US"MIME ACL";
1451 }
1452else if (rc != OK)
1453 {
1454 Uunlink(spool_name);
1455 unspool_mbox();
6f0c431a
PP
1456#ifdef EXPERIMENTAL_DCC
1457 dcc_ok = 0;
1458#endif
4f6ae5c3
JH
1459 if ( smtp_input
1460 && smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1461 {
85ffcba6 1462 *smtp_yield_ptr = FALSE; /* No more messages after dropped connection */
f4c1088b 1463 *smtp_reply_ptr = US""; /* Indicate reply already sent */
4f6ae5c3 1464 }
54cdb463
PH
1465 message_id[0] = 0; /* Indicate no message accepted */
1466 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1467 }
54cdb463
PH
1468
1469return TRUE;
1470}
1471
63955bf2 1472#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1473
1474
e4bdf652
JH
1475
1476void
1477received_header_gen(void)
1478{
1479uschar *received;
1480uschar *timestamp;
1481header_line *received_header= header_list;
1482
1483timestamp = expand_string(US"${tod_full}");
1484if (recipients_count == 1) received_for = recipients_list[0].address;
1485received = expand_string(received_header_text);
1486received_for = NULL;
1487
d4ff61d1 1488if (!received)
e4bdf652
JH
1489 {
1490 if(spool_name[0] != 0)
1491 Uunlink(spool_name); /* Lose the data file */
1492 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1493 "(received_header_text) failed: %s", string_printing(received_header_text),
1494 expand_string_message);
1495 }
1496
1497/* The first element on the header chain is reserved for the Received header,
1498so all we have to do is fill in the text pointer, and set the type. However, if
1499the result of the expansion is an empty string, we leave the header marked as
1500"old" so as to refrain from adding a Received header. */
1501
1502if (received[0] == 0)
1503 {
1504 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1505 received_header->type = htype_old;
1506 }
1507else
1508 {
1509 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1510 received_header->type = htype_received;
1511 }
1512
1513received_header->slen = Ustrlen(received_header->text);
1514
1515DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1516 received_header->type, received_header->text);
1517}
1518
1519
1520
54cdb463 1521/*************************************************
059ec3d9
PH
1522* Receive message *
1523*************************************************/
1524
1525/* Receive a message on the given input, and put it into a pair of spool files.
1526Either a non-null list of recipients, or the extract flag will be true, or
1527both. The flag sender_local is true for locally generated messages. The flag
1528submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1529suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1530suppress_local_fixups" or -G was passed on the command-line.
1531The flag smtp_input is true if the message is to be
8800895a
PH
1532handled using SMTP conventions about termination and lines starting with dots.
1533For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1534
1535If a message was successfully read, message_id[0] will be non-zero.
1536
1537The general actions of this function are:
1538
1539 . Read the headers of the message (if any) into a chain of store
1540 blocks.
1541
1542 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1543 throw it away, unless the caller is trusted, or unless
1544 active_local_sender_retain is set - which can only happen if
1545 active_local_from_check is false.
059ec3d9
PH
1546
1547 . If recipients are to be extracted from the message, build the
1548 recipients list from the headers, removing any that were on the
1549 original recipients list (unless extract_addresses_remove_arguments is
1550 false), and at the same time, remove any bcc header that may be present.
1551
1552 . Get the spool file for the data, sort out its unique name, open
1553 and lock it (but don't give it the name yet).
1554
1555 . Generate a "Message-Id" header if the message doesn't have one, for
1556 locally-originated messages.
1557
1558 . Generate a "Received" header.
1559
1560 . Ensure the recipients list is fully qualified and rewritten if necessary.
1561
1562 . If there are any rewriting rules, apply them to the sender address
1563 and also to the headers.
1564
1565 . If there is no from: header, generate one, for locally-generated messages
1566 and messages in "submission mode" only.
1567
1568 . If the sender is local, check that from: is correct, and if not, generate
1569 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1570 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1571
1572 . If there is no "date" header, generate one, for locally-originated
1573 or submission mode messages only.
1574
1575 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1576 dot_ends mode, to the data file. Leave it open, to hold the lock.
1577
1578 . Write the envelope and the headers to a new file.
1579
1580 . Set the name for the header file; close it.
1581
1582 . Set the name for the data file; close it.
1583
1584Because this function can potentially be called many times in a single
1585SMTP connection, all store should be got by store_get(), so that it will be
1586automatically retrieved after the message is accepted.
1587
1588FUDGE: It seems that sites on the net send out messages with just LF
1589terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1590we make the CRs optional in all cases.
1591
1592July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1593new regime is now in place in which bare CRs in header lines are turned into LF
1594followed by a space, so as not to terminate the header line.
1595
1596February 2004: A bare LF in a header line in a message whose first line was
1597terminated by CRLF is treated in the same way as a bare CR.
1598
1599Arguments:
1600 extract_recip TRUE if recipients are to be extracted from the message's
1601 headers
1602
1603Returns: TRUE there are more messages to be read (SMTP input)
1604 FALSE there are no more messages to be read (non-SMTP input
1605 or SMTP connection collapsed, or other failure)
1606
1607When reading a message for filter testing, the returned value indicates
1608whether the headers (which is all that is read) were terminated by '.' or
1609not. */
1610
1611BOOL
1612receive_msg(BOOL extract_recip)
1613{
7156b1ef
NM
1614int i;
1615int rc = FAIL;
059ec3d9
PH
1616int msg_size = 0;
1617int process_info_len = Ustrlen(process_info);
1618int error_rc = (error_handling == ERRORS_SENDER)?
1619 errors_sender_rc : EXIT_FAILURE;
1620int header_size = 256;
1621int start, end, domain, size, sptr;
1622int id_resolution;
1623int had_zero = 0;
d677b2f2 1624int prevlines_length = 0;
059ec3d9
PH
1625
1626register int ptr = 0;
1627
1628BOOL contains_resent_headers = FALSE;
1629BOOL extracted_ignored = FALSE;
1630BOOL first_line_ended_crlf = TRUE_UNSET;
1631BOOL smtp_yield = TRUE;
1632BOOL yield = FALSE;
1633
1634BOOL resents_exist = FALSE;
1635uschar *resent_prefix = US"";
1636uschar *blackholed_by = NULL;
04f7d5b9 1637uschar *blackhole_log_msg = US"";
c5430c20 1638enum {NOT_TRIED, TMP_REJ, PERM_REJ, ACCEPTED} cutthrough_done = NOT_TRIED;
059ec3d9
PH
1639
1640flock_t lock_data;
1641error_block *bad_addresses = NULL;
1642
1643uschar *frozen_by = NULL;
1644uschar *queued_by = NULL;
1645
1646uschar *errmsg, *s;
1647struct stat statbuf;
1648
4e88a19f 1649/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1650
1651uschar *smtp_reply = NULL;
4e88a19f 1652uschar *user_msg, *log_msg;
059ec3d9
PH
1653
1654/* Working header pointers */
1655
1656header_line *h, *next;
1657
2cbb4081 1658/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1659
1660BOOL date_header_exists = FALSE;
1661
1662/* Pointers to receive the addresses of headers whose contents we need. */
1663
1664header_line *from_header = NULL;
1665header_line *subject_header = NULL;
1666header_line *msgid_header = NULL;
1667header_line *received_header;
1668
4840604e
TL
1669#ifdef EXPERIMENTAL_DMARC
1670int dmarc_up = 0;
1671#endif /* EXPERIMENTAL_DMARC */
1672
059ec3d9
PH
1673/* Variables for use when building the Received: header. */
1674
059ec3d9
PH
1675uschar *timestamp;
1676int tslen;
1677
1678/* Release any open files that might have been cached while preparing to
1679accept the message - e.g. by verifying addresses - because reading a message
1680might take a fair bit of real time. */
1681
1682search_tidyup();
1683
e4bdf652
JH
1684/* Extracting the recipient list from an input file is incompatible with
1685cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1686to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1687if (extract_recip || !smtp_input)
57cc2785 1688 cancel_cutthrough_connection(TRUE, US"not smtp input");
e4bdf652 1689
059ec3d9
PH
1690/* Initialize the chain of headers by setting up a place-holder for Received:
1691header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1692pointing to the end of the chain to make adding headers simple. */
1693
1694received_header = header_list = header_last = store_get(sizeof(header_line));
1695header_list->next = NULL;
1696header_list->type = htype_old;
1697header_list->text = NULL;
1698header_list->slen = 0;
1699
1700/* Control block for the next header to be read. */
1701
1702next = store_get(sizeof(header_line));
1703next->text = store_get(header_size);
1704
1705/* Initialize message id to be null (indicating no message read), and the
1706header names list to be the normal list. Indicate there is no data file open
1707yet, initialize the size and warning count, and deal with no size limit. */
1708
1709message_id[0] = 0;
1710data_file = NULL;
1711data_fd = -1;
41313d92 1712spool_name = US"";
059ec3d9
PH
1713message_size = 0;
1714warning_count = 0;
d677b2f2 1715received_count = 1; /* For the one we will add */
059ec3d9
PH
1716
1717if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1718
2e0c1448 1719/* While reading the message, the following counts are computed. */
059ec3d9 1720
d677b2f2
PH
1721message_linecount = body_linecount = body_zerocount =
1722 max_received_linelength = 0;
059ec3d9 1723
80a47a2c 1724#ifndef DISABLE_DKIM
e983e85a
JH
1725/* Call into DKIM to set up the context. In CHUNKING mode
1726we clear the dot-stuffing flag */
1727if (smtp_input && !smtp_batched_input && !dkim_disable_verify)
1728 dkim_exim_verify_init(chunking_state <= CHUNKING_OFFERED);
fb2274d4
TK
1729#endif
1730
4840604e
TL
1731#ifdef EXPERIMENTAL_DMARC
1732/* initialize libopendmarc */
1733dmarc_up = dmarc_init();
1734#endif
1735
059ec3d9
PH
1736/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1737ids, and fractions of a second are required. See the comments that precede the
1738message id creation below. */
1739
1740(void)gettimeofday(&message_id_tv, NULL);
1741
1742/* For other uses of the received time we can operate with granularity of one
1743second, and for that we use the global variable received_time. This is for
1744things like ultimate message timeouts. */
1745
1746received_time = message_id_tv.tv_sec;
1747
1748/* If SMTP input, set the special handler for timeouts. The alarm() calls
1749happen in the smtp_getc() function when it refills its buffer. */
1750
1751if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1752
1753/* If not SMTP input, timeout happens only if configured, and we just set a
1754single timeout for the whole message. */
1755
1756else if (receive_timeout > 0)
1757 {
1758 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1759 alarm(receive_timeout);
1760 }
1761
1762/* SIGTERM and SIGINT are caught always. */
1763
1764signal(SIGTERM, data_sigterm_sigint_handler);
1765signal(SIGINT, data_sigterm_sigint_handler);
1766
1767/* Header lines in messages are not supposed to be very long, though when
1768unfolded, to: and cc: headers can take up a lot of store. We must also cope
1769with the possibility of junk being thrown at us. Start by getting 256 bytes for
1770storing the header, and extend this as necessary using string_cat().
1771
1772To cope with total lunacies, impose an upper limit on the length of the header
1773section of the message, as otherwise the store will fill up. We must also cope
1774with the possibility of binary zeros in the data. Hence we cannot use fgets().
1775Folded header lines are joined into one string, leaving the '\n' characters
1776inside them, so that writing them out reproduces the input.
1777
1778Loop for each character of each header; the next structure for chaining the
1779header is set up already, with ptr the offset of the next character in
1780next->text. */
1781
1782for (;;)
1783 {
bd8fbe36 1784 int ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1785
1786 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1787 SMTP must have a correct "." terminator. */
1788
1789 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1790 {
1791 smtp_reply = handle_lost_connection(US" (header)");
1792 smtp_yield = FALSE;
1793 goto TIDYUP; /* Skip to end of function */
1794 }
1795
1796 /* See if we are at the current header's size limit - there must be at least
1797 four bytes left. This allows for the new character plus a zero, plus two for
1798 extra insertions when we are playing games with dots and carriage returns. If
1799 we are at the limit, extend the text buffer. This could have been done
1800 automatically using string_cat() but because this is a tightish loop storing
1801 only one character at a time, we choose to do it inline. Normally
1802 store_extend() will be able to extend the block; only at the end of a big
1803 store block will a copy be needed. To handle the case of very long headers
1804 (and sometimes lunatic messages can have ones that are 100s of K long) we
1805 call store_release() for strings that have been copied - if the string is at
1806 the start of a block (and therefore the only thing in it, because we aren't
1807 doing any other gets), the block gets freed. We can only do this because we
1808 know there are no other calls to store_get() going on. */
1809
1810 if (ptr >= header_size - 4)
1811 {
1812 int oldsize = header_size;
1813 /* header_size += 256; */
1814 header_size *= 2;
1815 if (!store_extend(next->text, oldsize, header_size))
1816 {
1817 uschar *newtext = store_get(header_size);
1818 memcpy(newtext, next->text, ptr);
1819 store_release(next->text);
1820 next->text = newtext;
1821 }
1822 }
1823
1824 /* Cope with receiving a binary zero. There is dispute about whether
1825 these should be allowed in RFC 822 messages. The middle view is that they
1826 should not be allowed in headers, at least. Exim takes this attitude at
1827 the moment. We can't just stomp on them here, because we don't know that
1828 this line is a header yet. Set a flag to cause scanning later. */
1829
1830 if (ch == 0) had_zero++;
1831
1832 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1833 those from data files use just LF. Treat LF in local SMTP input as a
1834 terminator too. Treat EOF as a line terminator always. */
1835
1836 if (ch == EOF) goto EOL;
1837
1838 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1839 other MTAs accept this. We are therefore forced into this "liberalisation"
1840 too, so we accept LF as a line terminator whatever the source of the message.
1841 However, if the first line of the message ended with a CRLF, we treat a bare
1842 LF specially by inserting a white space after it to ensure that the header
1843 line is not terminated. */
1844
1845 if (ch == '\n')
1846 {
1847 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1848 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1849 goto EOL;
1850 }
1851
1852 /* This is not the end of the line. If this is SMTP input and this is
1853 the first character in the line and it is a "." character, ignore it.
1854 This implements the dot-doubling rule, though header lines starting with
1855 dots aren't exactly common. They are legal in RFC 822, though. If the
1856 following is CRLF or LF, this is the line that that terminates the
1857 entire message. We set message_ended to indicate this has happened (to
1858 prevent further reading), and break out of the loop, having freed the
1859 empty header, and set next = NULL to indicate no data line. */
1860
1861 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1862 {
bd8fbe36 1863 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1864 if (ch == '\r')
1865 {
bd8fbe36 1866 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1867 if (ch != '\n')
1868 {
80a47a2c 1869 receive_ungetc(ch);
059ec3d9
PH
1870 ch = '\r'; /* Revert to CR */
1871 }
1872 }
1873 if (ch == '\n')
1874 {
1875 message_ended = END_DOT;
1876 store_reset(next);
1877 next = NULL;
1878 break; /* End character-reading loop */
1879 }
1880
1881 /* For non-SMTP input, the dot at the start of the line was really a data
1882 character. What is now in ch is the following character. We guaranteed
1883 enough space for this above. */
1884
1885 if (!smtp_input)
1886 {
1887 next->text[ptr++] = '.';
1888 message_size++;
1889 }
1890 }
1891
1892 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1893 remember this case if this is the first line ending. */
1894
1895 if (ch == '\r')
1896 {
bd8fbe36 1897 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1898 if (ch == '\n')
1899 {
1900 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1901 goto EOL;
1902 }
1903
1904 /* Otherwise, put back the character after CR, and turn the bare CR
1905 into LF SP. */
1906
80a47a2c 1907 ch = (receive_ungetc)(ch);
059ec3d9
PH
1908 next->text[ptr++] = '\n';
1909 message_size++;
1910 ch = ' ';
1911 }
1912
1913 /* We have a data character for the header line. */
1914
1915 next->text[ptr++] = ch; /* Add to buffer */
1916 message_size++; /* Total message size so far */
1917
1918 /* Handle failure due to a humungously long header section. The >= allows
1919 for the terminating \n. Add what we have so far onto the headers list so
1920 that it gets reflected in any error message, and back up the just-read
1921 character. */
1922
1923 if (message_size >= header_maxsize)
1924 {
1925 next->text[ptr] = 0;
1926 next->slen = ptr;
1927 next->type = htype_other;
1928 next->next = NULL;
1929 header_last->next = next;
1930 header_last = next;
1931
1932 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1933 "%s (more than %d characters): message abandoned",
1934 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1935
1936 if (smtp_input)
1937 {
1938 smtp_reply = US"552 Message header is ridiculously long";
1939 receive_swallow_smtp();
1940 goto TIDYUP; /* Skip to end of function */
1941 }
1942
1943 else
1944 {
1945 give_local_error(ERRMESS_VLONGHEADER,
1946 string_sprintf("message header longer than %d characters received: "
1947 "message not accepted", header_maxsize), US"", error_rc, stdin,
1948 header_list->next);
1949 /* Does not return */
1950 }
1951 }
1952
1953 continue; /* With next input character */
1954
1955 /* End of header line reached */
1956
1957 EOL:
2e0c1448
PH
1958
1959 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1960
1961 receive_linecount++;
1962 message_linecount++;
059ec3d9 1963
d677b2f2
PH
1964 /* Keep track of maximum line length */
1965
1966 if (ptr - prevlines_length > max_received_linelength)
1967 max_received_linelength = ptr - prevlines_length;
1968 prevlines_length = ptr + 1;
1969
059ec3d9
PH
1970 /* Now put in the terminating newline. There is always space for
1971 at least two more characters. */
1972
1973 next->text[ptr++] = '\n';
1974 message_size++;
1975
1976 /* A blank line signals the end of the headers; release the unwanted
1977 space and set next to NULL to indicate this. */
1978
1979 if (ptr == 1)
1980 {
1981 store_reset(next);
1982 next = NULL;
1983 break;
1984 }
1985
1986 /* There is data in the line; see if the next input character is a
1987 whitespace character. If it is, we have a continuation of this header line.
1988 There is always space for at least one character at this point. */
1989
1990 if (ch != EOF)
1991 {
bd8fbe36 1992 int nextch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1993 if (nextch == ' ' || nextch == '\t')
1994 {
1995 next->text[ptr++] = nextch;
1996 message_size++;
1997 continue; /* Iterate the loop */
1998 }
80a47a2c 1999 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
2000 else ch = EOF; /* Cause main loop to exit at end */
2001 }
2002
2003 /* We have got to the real line end. Terminate the string and release store
2004 beyond it. If it turns out to be a real header, internal binary zeros will
2005 be squashed later. */
2006
2007 next->text[ptr] = 0;
2008 next->slen = ptr;
2009 store_reset(next->text + ptr + 1);
2010
2011 /* Check the running total size against the overall message size limit. We
2012 don't expect to fail here, but if the overall limit is set less than MESSAGE_
2013 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
2014 headers - the code to read the body will then also hit the buffer. */
2015
2016 if (message_size > thismessage_size_limit) break;
2017
2018 /* A line that is not syntactically correct for a header also marks
2019 the end of the headers. In this case, we leave next containing the
2020 first data line. This might actually be several lines because of the
2021 continuation logic applied above, but that doesn't matter.
2022
2023 It turns out that smail, and presumably sendmail, accept leading lines
2024 of the form
2025
2026 From ph10 Fri Jan 5 12:35 GMT 1996
2027
2028 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
2029 find any documentation of this, but for compatibility it had better be
2030 accepted. Exim restricts it to the case of non-smtp messages, and
2031 treats it as an alternative to the -f command line option. Thus it is
2032 ignored except for trusted users or filter testing. Otherwise it is taken
2033 as the sender address, unless -f was used (sendmail compatibility).
2034
2035 It further turns out that some UUCPs generate the From_line in a different
2036 format, e.g.
2037
2038 From ph10 Fri, 7 Jan 97 14:00:00 GMT
2039
2040 The regex for matching these things is now capable of recognizing both
2041 formats (including 2- and 4-digit years in the latter). In fact, the regex
2042 is now configurable, as is the expansion string to fish out the sender.
2043
2044 Even further on it has been discovered that some broken clients send
2045 these lines in SMTP messages. There is now an option to ignore them from
2046 specified hosts or networks. Sigh. */
2047
2048 if (header_last == header_list &&
2049 (!smtp_input
2050 ||
2051 (sender_host_address != NULL &&
2052 verify_check_host(&ignore_fromline_hosts) == OK)
2053 ||
2054 (sender_host_address == NULL && ignore_fromline_local)
2055 ) &&
2056 regex_match_and_setup(regex_From, next->text, 0, -1))
2057 {
2058 if (!sender_address_forced)
2059 {
2060 uschar *uucp_sender = expand_string(uucp_from_sender);
2061 if (uucp_sender == NULL)
2062 {
2063 log_write(0, LOG_MAIN|LOG_PANIC,
2064 "expansion of \"%s\" failed after matching "
2065 "\"From \" line: %s", uucp_from_sender, expand_string_message);
2066 }
2067 else
2068 {
2069 int start, end, domain;
2070 uschar *errmess;
2071 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
2072 &start, &end, &domain, TRUE);
2073 if (newsender != NULL)
2074 {
2075 if (domain == 0 && newsender[0] != 0)
2076 newsender = rewrite_address_qualify(newsender, FALSE);
2077
f05da2e8 2078 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
2079 {
2080 sender_address = newsender;
2081
f05da2e8 2082 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
2083 {
2084 authenticated_sender = NULL;
2085 originator_name = US"";
2086 sender_local = FALSE;
2087 }
2088
f05da2e8 2089 if (filter_test != FTEST_NONE)
059ec3d9
PH
2090 printf("Sender taken from \"From \" line\n");
2091 }
2092 }
2093 }
2094 }
2095 }
2096
2097 /* Not a leading "From " line. Check to see if it is a valid header line.
2098 Header names may contain any non-control characters except space and colon,
2099 amazingly. */
2100
2101 else
2102 {
2103 uschar *p = next->text;
2104
2105 /* If not a valid header line, break from the header reading loop, leaving
2106 next != NULL, indicating that it holds the first line of the body. */
2107
2108 if (isspace(*p)) break;
2109 while (mac_isgraph(*p) && *p != ':') p++;
2110 while (isspace(*p)) p++;
2111 if (*p != ':')
2112 {
2113 body_zerocount = had_zero;
2114 break;
2115 }
2116
2117 /* We have a valid header line. If there were any binary zeroes in
2118 the line, stomp on them here. */
2119
2120 if (had_zero > 0)
2121 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
2122
2123 /* It is perfectly legal to have an empty continuation line
2124 at the end of a header, but it is confusing to humans
2125 looking at such messages, since it looks like a blank line.
2126 Reduce confusion by removing redundant white space at the
2127 end. We know that there is at least one printing character
2128 (the ':' tested for above) so there is no danger of running
2129 off the end. */
2130
2131 p = next->text + ptr - 2;
2132 for (;;)
2133 {
2134 while (*p == ' ' || *p == '\t') p--;
2135 if (*p != '\n') break;
2136 ptr = (p--) - next->text + 1;
2137 message_size -= next->slen - ptr;
2138 next->text[ptr] = 0;
2139 next->slen = ptr;
2140 }
2141
2142 /* Add the header to the chain */
2143
2144 next->type = htype_other;
2145 next->next = NULL;
2146 header_last->next = next;
2147 header_last = next;
2148
2149 /* Check the limit for individual line lengths. This comes after adding to
2150 the chain so that the failing line is reflected if a bounce is generated
2151 (for a local message). */
2152
2153 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
2154 {
2155 log_write(0, LOG_MAIN, "overlong message header line received from "
2156 "%s (more than %d characters): message abandoned",
2157 sender_host_unknown? sender_ident : sender_fullhost,
2158 header_line_maxsize);
2159
2160 if (smtp_input)
2161 {
2162 smtp_reply = US"552 A message header line is too long";
2163 receive_swallow_smtp();
2164 goto TIDYUP; /* Skip to end of function */
2165 }
2166
2167 else
2168 {
2169 give_local_error(ERRMESS_VLONGHDRLINE,
2170 string_sprintf("message header line longer than %d characters "
2171 "received: message not accepted", header_line_maxsize), US"",
2172 error_rc, stdin, header_list->next);
2173 /* Does not return */
2174 }
2175 }
2176
2177 /* Note if any resent- fields exist. */
2178
2179 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
2180 {
2181 resents_exist = TRUE;
2182 resent_prefix = US"Resent-";
2183 }
2184 }
2185
1ebe15c3
JH
2186 /* Reject CHUNKING messages that do not CRLF their first header line */
2187
2188 if (!first_line_ended_crlf && chunking_state > CHUNKING_OFFERED)
2189 {
2190 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2191 "Non-CRLF-terminated header, under CHUNKING: message abandoned",
2192 sender_address,
2193 sender_fullhost ? " H=" : "", sender_fullhost ? sender_fullhost : US"",
2194 sender_ident ? " U=" : "", sender_ident ? sender_ident : US"");
2195 smtp_printf("552 Message header not CRLF terminated\r\n");
2196 bdat_flush_data();
2197 smtp_reply = US"";
2198 goto TIDYUP; /* Skip to end of function */
2199 }
2200
059ec3d9
PH
2201 /* The line has been handled. If we have hit EOF, break out of the loop,
2202 indicating no pending data line. */
2203
2204 if (ch == EOF) { next = NULL; break; }
2205
2206 /* Set up for the next header */
2207
2208 header_size = 256;
2209 next = store_get(sizeof(header_line));
2210 next->text = store_get(header_size);
2211 ptr = 0;
2212 had_zero = 0;
d677b2f2 2213 prevlines_length = 0;
059ec3d9
PH
2214 } /* Continue, starting to read the next header */
2215
2216/* At this point, we have read all the headers into a data structure in main
2217store. The first header is still the dummy placeholder for the Received: header
2218we are going to generate a bit later on. If next != NULL, it contains the first
2219data line - which terminated the headers before reaching a blank line (not the
2220normal case). */
2221
2222DEBUG(D_receive)
2223 {
2224 debug_printf(">>Headers received:\n");
1ebe15c3 2225 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2226 debug_printf("%s", h->text);
2227 debug_printf("\n");
2228 }
2229
2230/* End of file on any SMTP connection is an error. If an incoming SMTP call
2231is dropped immediately after valid headers, the next thing we will see is EOF.
2232We must test for this specially, as further down the reading of the data is
2233skipped if already at EOF. */
2234
2235if (smtp_input && (receive_feof)())
2236 {
2237 smtp_reply = handle_lost_connection(US" (after header)");
2238 smtp_yield = FALSE;
2239 goto TIDYUP; /* Skip to end of function */
2240 }
2241
2242/* If this is a filter test run and no headers were read, output a warning
2243in case there is a mistake in the test message. */
2244
f05da2e8 2245if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2246 printf("Warning: no message headers read\n");
2247
2248
2249/* Scan the headers to identify them. Some are merely marked for later
2250processing; some are dealt with here. */
2251
1ebe15c3 2252for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2253 {
2254 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2255 if (is_resent) contains_resent_headers = TRUE;
2256
2257 switch (header_checkname(h, is_resent))
2258 {
059ec3d9 2259 case htype_bcc:
2cbb4081 2260 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
2261 break;
2262
059ec3d9 2263 case htype_cc:
2cbb4081 2264 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
2265 break;
2266
2267 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
2268
2269 case htype_date:
4c69d561 2270 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
2271 break;
2272
2273 /* Same comments as about Return-Path: below. */
2274
2275 case htype_delivery_date:
2276 if (delivery_date_remove) h->type = htype_old;
2277 break;
2278
2279 /* Same comments as about Return-Path: below. */
2280
2281 case htype_envelope_to:
2282 if (envelope_to_remove) h->type = htype_old;
2283 break;
2284
2285 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2286 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2287 header consists of just the login id of the user who called Exim, rewrite
2288 it with the gecos field first. Apply this rule to Resent-From: if there
2289 are resent- fields. */
2290
2291 case htype_from:
2292 h->type = htype_from;
2293 if (!resents_exist || is_resent)
2294 {
2295 from_header = h;
2296 if (!smtp_input)
2297 {
5de8faa3 2298 int len;
059ec3d9
PH
2299 uschar *s = Ustrchr(h->text, ':') + 1;
2300 while (isspace(*s)) s++;
5de8faa3 2301 len = h->slen - (s - h->text) - 1;
e0fccd1d
TF
2302 if (Ustrlen(originator_login) == len &&
2303 strncmpic(s, originator_login, len) == 0)
059ec3d9
PH
2304 {
2305 uschar *name = is_resent? US"Resent-From" : US"From";
2306 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2307 originator_login, qualify_domain_sender);
2308 from_header = header_last;
2309 h->type = htype_old;
2310 DEBUG(D_receive|D_rewrite)
2311 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2312 }
2313 }
2314 }
2315 break;
2316
2317 /* Identify the Message-id: header for generating "in-reply-to" in the
2318 autoreply transport. For incoming logging, save any resent- value. In both
2319 cases, take just the first of any multiples. */
2320
2321 case htype_id:
2322 if (msgid_header == NULL && (!resents_exist || is_resent))
2323 {
2324 msgid_header = h;
2325 h->type = htype_id;
2326 }
2327 break;
2328
2329 /* Flag all Received: headers */
2330
2331 case htype_received:
2332 h->type = htype_received;
2333 received_count++;
2334 break;
2335
2336 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
2337
2338 case htype_reply_to:
2339 h->type = htype_reply_to;
2340 break;
2341
2342 /* The Return-path: header is supposed to be added to messages when
2343 they leave the SMTP system. We shouldn't receive messages that already
2344 contain Return-path. However, since Exim generates Return-path: on
2345 local delivery, resent messages may well contain it. We therefore
2346 provide an option (which defaults on) to remove any Return-path: headers
2347 on input. Removal actually means flagging as "old", which prevents the
2348 header being transmitted with the message. */
2349
2350 case htype_return_path:
2351 if (return_path_remove) h->type = htype_old;
2352
2353 /* If we are testing a mail filter file, use the value of the
2354 Return-Path: header to set up the return_path variable, which is not
2355 otherwise set. However, remove any <> that surround the address
2356 because the variable doesn't have these. */
2357
f05da2e8 2358 if (filter_test != FTEST_NONE)
059ec3d9
PH
2359 {
2360 uschar *start = h->text + 12;
2361 uschar *end = start + Ustrlen(start);
2362 while (isspace(*start)) start++;
2363 while (end > start && isspace(end[-1])) end--;
2364 if (*start == '<' && end[-1] == '>')
2365 {
2366 start++;
2367 end--;
2368 }
2369 return_path = string_copyn(start, end - start);
2370 printf("Return-path taken from \"Return-path:\" header line\n");
2371 }
2372 break;
2373
2374 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2375 and from an untrusted caller and suppress_local_fixups is not set, or if we
2376 are in submission mode for a remote message, mark it "old" so that it will
2377 not be transmitted with the message, unless active_local_sender_retain is
2378 set. (This can only be true if active_local_from_check is false.) If there
2379 are any resent- headers in the message, apply this rule to Resent-Sender:
2380 instead of Sender:. Messages with multiple resent- header sets cannot be
2381 tidily handled. (For this reason, at least one MUA - Pine - turns old
2382 resent- headers into X-resent- headers when resending, leaving just one
2383 set.) */
059ec3d9
PH
2384
2385 case htype_sender:
69358f02 2386 h->type = ((!active_local_sender_retain &&
8800895a
PH
2387 (
2388 (sender_local && !trusted_caller && !suppress_local_fixups)
2389 || submission_mode
2390 )
059ec3d9
PH
2391 ) &&
2392 (!resents_exist||is_resent))?
2393 htype_old : htype_sender;
2394 break;
2395
2396 /* Remember the Subject: header for logging. There is no Resent-Subject */
2397
2398 case htype_subject:
2399 subject_header = h;
2400 break;
2401
2402 /* "To:" gets flagged, and the existence of a recipient header is noted,
2403 whether it's resent- or not. */
2404
2405 case htype_to:
2406 h->type = htype_to;
2407 /****
2408 to_or_cc_header_exists = TRUE;
2409 ****/
2410 break;
2411 }
2412 }
2413
2414/* Extract recipients from the headers if that is required (the -t option).
2415Note that this is documented as being done *before* any address rewriting takes
2416place. There are two possibilities:
2417
2418(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2419recipients already listed are to be REMOVED from the message. Smail 3 works
2420like this. We need to build a non-recipients tree for that list, because in
2421subsequent processing this data is held in a tree and that's what the
2422spool_write_header() function expects. Make sure that non-recipient addresses
2423are fully qualified and rewritten if necessary.
2424
2425(2) According to other sendmail documentation, -t ADDS extracted recipients to
2426those in the command line arguments (and it is rumoured some other MTAs do
2427this). Therefore, there is an option to make Exim behave this way.
2428
2429*** Notes on "Resent-" header lines ***
2430
2431The presence of resent-headers in the message makes -t horribly ambiguous.
2432Experiments with sendmail showed that it uses recipients for all resent-
2433headers, totally ignoring the concept of "sets of resent- headers" as described
2434in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2435with all the addresses in one instance of each header.
2436
2437This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2438error for -t if there were resent- headers in the message. However, after a
2439discussion on the mailing list, I've learned that there are MUAs that use
2440resent- headers with -t, and also that the stuff about sets of resent- headers
2441and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2442message with -t and resent- header lines makes sure that only *its* resent-
2443headers are present; previous ones are often renamed as X-resent- for example.
2444
2445Consequently, Exim has been changed so that, if any resent- header lines are
2446present, the recipients are taken from all of the appropriate resent- lines,
2447and not from the ordinary To:, Cc:, etc. */
2448
2449if (extract_recip)
2450 {
2451 int rcount = 0;
2452 error_block **bnext = &bad_addresses;
2453
2454 if (extract_addresses_remove_arguments)
2455 {
2456 while (recipients_count-- > 0)
2457 {
2458 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2459 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2460 tree_add_nonrecipient(s);
2461 }
2462 recipients_list = NULL;
2463 recipients_count = recipients_list_max = 0;
2464 }
2465
059ec3d9
PH
2466 /* Now scan the headers */
2467
1ebe15c3 2468 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2469 {
2470 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2471 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2472 {
2473 uschar *s = Ustrchr(h->text, ':') + 1;
2474 while (isspace(*s)) s++;
2475
1eccaa59
PH
2476 parse_allow_group = TRUE; /* Allow address group syntax */
2477
059ec3d9
PH
2478 while (*s != 0)
2479 {
2480 uschar *ss = parse_find_address_end(s, FALSE);
2481 uschar *recipient, *errmess, *p, *pp;
2482 int start, end, domain;
2483
2484 /* Check on maximum */
2485
2486 if (recipients_max > 0 && ++rcount > recipients_max)
2487 {
2488 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2489 US"message rejected: ", error_rc, stdin, NULL);
2490 /* Does not return */
2491 }
2492
2493 /* Make a copy of the address, and remove any internal newlines. These
2494 may be present as a result of continuations of the header line. The
2495 white space that follows the newline must not be removed - it is part
2496 of the header. */
2497
2498 pp = recipient = store_get(ss - s + 1);
2499 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2500 *pp = 0;
250b6871 2501
8c5d388a 2502#ifdef SUPPORT_I18N
250b6871
JH
2503 {
2504 BOOL b = allow_utf8_domains;
2505 allow_utf8_domains = TRUE;
2506#endif
059ec3d9
PH
2507 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2508 &domain, FALSE);
2509
8c5d388a 2510#ifdef SUPPORT_I18N
250b6871
JH
2511 if (string_is_utf8(recipient))
2512 message_smtputf8 = TRUE;
2513 else
2514 allow_utf8_domains = b;
2515 }
2516#endif
2517
059ec3d9
PH
2518 /* Keep a list of all the bad addresses so we can send a single
2519 error message at the end. However, an empty address is not an error;
2520 just ignore it. This can come from an empty group list like
2521
2522 To: Recipients of list:;
2523
2524 If there are no recipients at all, an error will occur later. */
2525
2526 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2527 {
2528 int len = Ustrlen(s);
2529 error_block *b = store_get(sizeof(error_block));
2530 while (len > 0 && isspace(s[len-1])) len--;
2531 b->next = NULL;
2532 b->text1 = string_printing(string_copyn(s, len));
2533 b->text2 = errmess;
2534 *bnext = b;
2535 bnext = &(b->next);
2536 }
2537
2538 /* If the recipient is already in the nonrecipients tree, it must
2539 have appeared on the command line with the option extract_addresses_
2540 remove_arguments set. Do not add it to the recipients, and keep a note
2541 that this has happened, in order to give a better error if there are
2542 no recipients left. */
2543
2544 else if (recipient != NULL)
2545 {
2546 if (tree_search(tree_nonrecipients, recipient) == NULL)
2547 receive_add_recipient(recipient, -1);
2548 else
2549 extracted_ignored = TRUE;
2550 }
2551
2552 /* Move on past this address */
2553
2554 s = ss + (*ss? 1:0);
2555 while (isspace(*s)) s++;
1eccaa59
PH
2556 } /* Next address */
2557
2558 parse_allow_group = FALSE; /* Reset group syntax flags */
2559 parse_found_group = FALSE;
059ec3d9
PH
2560
2561 /* If this was the bcc: header, mark it "old", which means it
2562 will be kept on the spool, but not transmitted as part of the
2563 message. */
2564
2cbb4081 2565 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2566 } /* For appropriate header line */
2567 } /* For each header line */
2568
059ec3d9
PH
2569 }
2570
2571/* Now build the unique message id. This has changed several times over the
2572lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2573Retaining all the history in the comment has become too unwieldy - read
2574previous release sources if you want it.
2575
2576The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2577The first part is the current time, in seconds. The second part is the current
2578pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2579can hold a number in the range 0-3843. It used to be a computed sequence
2580number, but is now the fractional component of the current time in units of
25811/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2582received, Exim ensures that the timer has ticked at the appropriate level
2583before proceeding, to avoid duplication if the pid happened to be re-used
2584within the same time period. It seems likely that most messages will take at
2585least half a millisecond to be received, so no delay will normally be
2586necessary. At least for some time...
2587
2588There is a modification when localhost_number is set. Formerly this was allowed
2589to be as large as 255. Now it is restricted to the range 0-16, and the final
2590component of the message id becomes (localhost_number * 200) + fractional time
2591in units of 1/200 of a second (i.e. a value in the range 0-3399).
2592
2593Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2594Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2595still allows the tttttt field to hold a large enough number to last for some
2596more decades, and the final two-digit field can hold numbers up to 1295, which
2597is enough for milliseconds (instead of 1/2000 of a second).
2598
2599However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2600pid, so it is probably safe because pids have to be positive. The
2601localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2602final field becomes (localhost_number * 100) + fractional time in centiseconds.
2603
2604Note that string_base62() returns its data in a static storage block, so it
2605must be copied before calling string_base62() again. It always returns exactly
26066 characters.
2607
2608There doesn't seem to be anything in the RFC which requires a message id to
2609start with a letter, but Smail was changed to ensure this. The external form of
2610the message id (as supplied by string expansion) therefore starts with an
2611additional leading 'E'. The spool file names do not include this leading
2612letter and it is not used internally.
2613
2614NOTE: If ever the format of message ids is changed, the regular expression for
2615checking that a string is in this format must be updated in a corresponding
2616way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2617must also be changed to reflect the correct string length. Then, of course,
2618other programs that rely on the message id format will need updating too. */
2619
2620Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2621message_id[6] = '-';
2622Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2623
2624/* Deal with the case where the host number is set. The value of the number was
2625checked when it was read, to ensure it isn't too big. The timing granularity is
2626left in id_resolution so that an appropriate wait can be done after receiving
2627the message, if necessary (we hope it won't be). */
2628
2629if (host_number_string != NULL)
2630 {
2631 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2632 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2633 string_base62((long int)(
2634 host_number * (1000000/id_resolution) +
2635 message_id_tv.tv_usec/id_resolution)) + 4);
2636 }
2637
2638/* Host number not set: final field is just the fractional time at an
2639appropriate resolution. */
2640
2641else
2642 {
2643 id_resolution = (BASE_62 == 62)? 500 : 1000;
2644 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2645 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2646 }
2647
2648/* Add the current message id onto the current process info string if
2649it will fit. */
2650
2651(void)string_format(process_info + process_info_len,
2652 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2653
2654/* If we are using multiple input directories, set up the one for this message
2655to be the least significant base-62 digit of the time of arrival. Otherwise
2656ensure that it is an empty string. */
2657
a2da3176 2658message_subdir[0] = split_spool_directory ? message_id[5] : 0;
059ec3d9
PH
2659
2660/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2661one, but only for local (without suppress_local_fixups) or submission mode
2662messages. This can be user-configured if required, but we had better flatten
2663any illegal characters therein. */
059ec3d9 2664
8800895a
PH
2665if (msgid_header == NULL &&
2666 ((sender_host_address == NULL && !suppress_local_fixups)
2667 || submission_mode))
059ec3d9
PH
2668 {
2669 uschar *p;
2670 uschar *id_text = US"";
2671 uschar *id_domain = primary_hostname;
2672
2673 /* Permit only letters, digits, dots, and hyphens in the domain */
2674
2675 if (message_id_domain != NULL)
2676 {
2677 uschar *new_id_domain = expand_string(message_id_domain);
2678 if (new_id_domain == NULL)
2679 {
2680 if (!expand_string_forcedfail)
2681 log_write(0, LOG_MAIN|LOG_PANIC,
2682 "expansion of \"%s\" (message_id_header_domain) "
2683 "failed: %s", message_id_domain, expand_string_message);
2684 }
2685 else if (*new_id_domain != 0)
2686 {
2687 id_domain = new_id_domain;
2688 for (p = id_domain; *p != 0; p++)
2689 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2690 }
2691 }
2692
2693 /* Permit all characters except controls and RFC 2822 specials in the
2694 additional text part. */
2695
2696 if (message_id_text != NULL)
2697 {
2698 uschar *new_id_text = expand_string(message_id_text);
2699 if (new_id_text == NULL)
2700 {
2701 if (!expand_string_forcedfail)
2702 log_write(0, LOG_MAIN|LOG_PANIC,
2703 "expansion of \"%s\" (message_id_header_text) "
2704 "failed: %s", message_id_text, expand_string_message);
2705 }
2706 else if (*new_id_text != 0)
2707 {
2708 id_text = new_id_text;
2709 for (p = id_text; *p != 0; p++)
2710 if (mac_iscntrl_or_special(*p)) *p = '-';
2711 }
2712 }
2713
e7e680d6
PP
2714 /* Add the header line
2715 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2716 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2717
e7e680d6 2718 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2719 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2720 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2721 }
2722
2723/* If we are to log recipients, keep a copy of the raw ones before any possible
2724rewriting. Must copy the count, because later ACLs and the local_scan()
2725function may mess with the real recipients. */
2726
6c6d6e48 2727if (LOGGING(received_recipients))
059ec3d9
PH
2728 {
2729 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2730 for (i = 0; i < recipients_count; i++)
2731 raw_recipients[i] = string_copy(recipients_list[i].address);
2732 raw_recipients_count = recipients_count;
2733 }
2734
2735/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2736recipients will get here only if the conditions were right (allow_unqualified_
2737recipient is TRUE). */
2738
2739for (i = 0; i < recipients_count; i++)
2740 recipients_list[i].address =
2741 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2742 global_rewrite_rules, rewrite_existflags);
2743
8800895a
PH
2744/* If there is no From: header, generate one for local (without
2745suppress_local_fixups) or submission_mode messages. If there is no sender
2746address, but the sender is local or this is a local delivery error, use the
2747originator login. This shouldn't happen for genuine bounces, but might happen
2748for autoreplies. The addition of From: must be done *before* checking for the
2749possible addition of a Sender: header, because untrusted_set_sender allows an
2750untrusted user to set anything in the envelope (which might then get info
2751From:) but we still want to ensure a valid Sender: if it is required. */
2752
2753if (from_header == NULL &&
2754 ((sender_host_address == NULL && !suppress_local_fixups)
2755 || submission_mode))
059ec3d9 2756 {
2fe1a124
PH
2757 uschar *oname = US"";
2758
2759 /* Use the originator_name if this is a locally submitted message and the
2760 caller is not trusted. For trusted callers, use it only if -F was used to
2761 force its value or if we have a non-SMTP message for which -f was not used
2762 to set the sender. */
2763
2764 if (sender_host_address == NULL)
2765 {
2766 if (!trusted_caller || sender_name_forced ||
2767 (!smtp_input && !sender_address_forced))
2768 oname = originator_name;
2769 }
2770
2771 /* For non-locally submitted messages, the only time we use the originator
2772 name is when it was forced by the /name= option on control=submission. */
2773
2774 else
2775 {
2776 if (submission_name != NULL) oname = submission_name;
2777 }
2778
059ec3d9
PH
2779 /* Envelope sender is empty */
2780
2781 if (sender_address[0] == 0)
2782 {
87ba3f5f
PH
2783 uschar *fromstart, *fromend;
2784
2785 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2786 oname, (oname[0] == 0)? "" : " <");
2787 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2788
059ec3d9
PH
2789 if (sender_local || local_error_message)
2790 {
87ba3f5f
PH
2791 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2792 local_part_quote(originator_login), qualify_domain_sender,
2793 fromend);
059ec3d9
PH
2794 }
2795 else if (submission_mode && authenticated_id != NULL)
2796 {
2797 if (submission_domain == NULL)
2798 {
87ba3f5f
PH
2799 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2800 local_part_quote(authenticated_id), qualify_domain_sender,
2801 fromend);
059ec3d9
PH
2802 }
2803 else if (submission_domain[0] == 0) /* empty => whole address set */
2804 {
87ba3f5f
PH
2805 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2806 fromend);
059ec3d9
PH
2807 }
2808 else
2809 {
87ba3f5f
PH
2810 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2811 local_part_quote(authenticated_id), submission_domain,
2812 fromend);
059ec3d9
PH
2813 }
2814 from_header = header_last; /* To get it checked for Sender: */
2815 }
2816 }
2817
2818 /* There is a non-null envelope sender. Build the header using the original
2819 sender address, before any rewriting that might have been done while
2820 verifying it. */
2821
2822 else
2823 {
87ba3f5f 2824 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2825 oname,
2826 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2827 (sender_address_unrewritten == NULL)?
2828 sender_address : sender_address_unrewritten,
2fe1a124 2829 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2830
2831 from_header = header_last; /* To get it checked for Sender: */
2832 }
2833 }
2834
2835
8800895a
PH
2836/* If the sender is local (without suppress_local_fixups), or if we are in
2837submission mode and there is an authenticated_id, check that an existing From:
2838is correct, and if not, generate a Sender: header, unless disabled. Any
2839previously-existing Sender: header was removed above. Note that sender_local,
2840as well as being TRUE if the caller of exim is not trusted, is also true if a
2841trusted caller did not supply a -f argument for non-smtp input. To allow
2842trusted callers to forge From: without supplying -f, we have to test explicitly
2843here. If the From: header contains more than one address, then the call to
2844parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2845
2846if (from_header != NULL &&
69358f02 2847 (active_local_from_check &&
8800895a 2848 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2849 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2850 ))
2851 {
2852 BOOL make_sender = TRUE;
2853 int start, end, domain;
2854 uschar *errmess;
2855 uschar *from_address =
2856 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2857 &start, &end, &domain, FALSE);
2858 uschar *generated_sender_address;
2859
2860 if (submission_mode)
2861 {
2862 if (submission_domain == NULL)
2863 {
2864 generated_sender_address = string_sprintf("%s@%s",
2865 local_part_quote(authenticated_id), qualify_domain_sender);
2866 }
2867 else if (submission_domain[0] == 0) /* empty => full address */
2868 {
2869 generated_sender_address = string_sprintf("%s",
2870 authenticated_id);
2871 }
2872 else
2873 {
2874 generated_sender_address = string_sprintf("%s@%s",
2875 local_part_quote(authenticated_id), submission_domain);
2876 }
2877 }
2878 else
2879 generated_sender_address = string_sprintf("%s@%s",
2880 local_part_quote(originator_login), qualify_domain_sender);
2881
2882 /* Remove permitted prefixes and suffixes from the local part of the From:
2883 address before doing the comparison with the generated sender. */
2884
2885 if (from_address != NULL)
2886 {
2887 int slen;
2888 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2889
2890 if (at != NULL) *at = 0;
2891 from_address += route_check_prefix(from_address, local_from_prefix);
2892 slen = route_check_suffix(from_address, local_from_suffix);
2893 if (slen > 0)
2894 {
2895 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2896 from_address += slen;
2897 }
2898 if (at != NULL) *at = '@';
2899
2900 if (strcmpic(generated_sender_address, from_address) == 0 ||
2901 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2902 make_sender = FALSE;
2903 }
2904
2905 /* We have to cause the Sender header to be rewritten if there are
2906 appropriate rewriting rules. */
2907
2908 if (make_sender)
2909 {
2fe1a124 2910 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2911 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2912 generated_sender_address);
2913 else
2914 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2915 resent_prefix,
2916 submission_mode? submission_name : originator_name,
2917 generated_sender_address);
059ec3d9 2918 }
87ba3f5f
PH
2919
2920 /* Ensure that a non-null envelope sender address corresponds to the
2921 submission mode sender address. */
2922
2923 if (submission_mode && sender_address[0] != 0)
2924 {
2925 if (sender_address_unrewritten == NULL)
2926 sender_address_unrewritten = sender_address;
2927 sender_address = generated_sender_address;
089793a4
TF
2928 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2929 log_write(L_address_rewrite, LOG_MAIN,
2930 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2931 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2932 }
059ec3d9
PH
2933 }
2934
059ec3d9
PH
2935/* If there are any rewriting rules, apply them to the sender address, unless
2936it has already been rewritten as part of verification for SMTP input. */
2937
2938if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2939 sender_address[0] != 0)
2940 {
2941 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2942 global_rewrite_rules, rewrite_existflags);
2943 DEBUG(D_receive|D_rewrite)
2944 debug_printf("rewritten sender = %s\n", sender_address);
2945 }
2946
2947
2948/* The headers must be run through rewrite_header(), because it ensures that
2949addresses are fully qualified, as well as applying any rewriting rules that may
2950exist.
2951
2952Qualification of header addresses in a message from a remote host happens only
2953if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2954appropriate. For local messages, qualification always happens, unless -bnq is
2955used to explicitly suppress it. No rewriting is done for an unqualified address
2956that is left untouched.
2957
2958We start at the second header, skipping our own Received:. This rewriting is
2959documented as happening *after* recipient addresses are taken from the headers
2960by the -t command line option. An added Sender: gets rewritten here. */
2961
1ebe15c3 2962for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2963 {
2964 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2965 rewrite_existflags, TRUE);
1ebe15c3 2966 if (newh) h = newh;
059ec3d9
PH
2967 }
2968
2969
2970/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2971"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2972"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2973exists, this applies to the set of resent- headers rather than the normal set.
2974
2cbb4081
PH
2975The requirement for a recipient header has been removed in RFC 2822. At this
2976point in the code, earlier versions of Exim added a To: header for locally
2977submitted messages, and an empty Bcc: header for others. In the light of the
2978changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2979
059ec3d9
PH
2980
2981/* If there is no date header, generate one if the message originates locally
8800895a
PH
2982(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2983submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2984to be more confusing if Exim adds one to all remotely-originated messages.
2985As per Message-Id, we prepend if resending, else append.
2986*/
059ec3d9 2987
8800895a
PH
2988if (!date_header_exists &&
2989 ((sender_host_address == NULL && !suppress_local_fixups)
2990 || submission_mode))
e7e680d6
PP
2991 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2992 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2993
2994search_tidyup(); /* Free any cached resources */
2995
2996/* Show the complete set of headers if debugging. Note that the first one (the
2997new Received:) has not yet been set. */
2998
2999DEBUG(D_receive)
3000 {
3001 debug_printf(">>Headers after rewriting and local additions:\n");
3002 for (h = header_list->next; h != NULL; h = h->next)
3003 debug_printf("%c %s", h->type, h->text);
3004 debug_printf("\n");
3005 }
3006
3007/* The headers are now complete in store. If we are running in filter
3008testing mode, that is all this function does. Return TRUE if the message
3009ended with a dot. */
3010
f05da2e8 3011if (filter_test != FTEST_NONE)
059ec3d9
PH
3012 {
3013 process_info[process_info_len] = 0;
3014 return message_ended == END_DOT;
3015 }
3016
7e3ce68e
JH
3017/*XXX CHUNKING: need to cancel cutthrough under BDAT, for now. In future,
3018think more if it could be handled. Cannot do onward CHUNKING unless
3019inbound is, but inbound chunking ought to be ok with outbound plain.
3020Could we do onward CHUNKING given inbound CHUNKING?
3021*/
3022if (chunking_state > CHUNKING_OFFERED)
57cc2785 3023 cancel_cutthrough_connection(FALSE, US"chunking active");
7e3ce68e 3024
817d9f57 3025/* Cutthrough delivery:
5032d1cf
JH
3026We have to create the Received header now rather than at the end of reception,
3027so the timestamp behaviour is a change to the normal case.
3028XXX Ensure this gets documented XXX.
3029Having created it, send the headers to the destination. */
57cc2785
JH
3030
3031if (cutthrough.fd >= 0 && cutthrough.delivery)
e4bdf652 3032 {
817d9f57
JH
3033 if (received_count > received_headers_max)
3034 {
57cc2785 3035 cancel_cutthrough_connection(TRUE, US"too many headers");
817d9f57
JH
3036 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
3037 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3038 "Too many \"Received\" headers",
3039 sender_address,
57cc2785
JH
3040 sender_fullhost ? "H=" : "", sender_fullhost ? sender_fullhost : US"",
3041 sender_ident ? "U=" : "", sender_ident ? sender_ident : US"");
817d9f57
JH
3042 message_id[0] = 0; /* Indicate no message accepted */
3043 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
3044 goto TIDYUP; /* Skip to end of function */
3045 }
e4bdf652 3046 received_header_gen();
578d43dc 3047 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652
JH
3048 (void) cutthrough_headers_send();
3049 }
61147df4 3050
e4bdf652 3051
059ec3d9
PH
3052/* Open a new spool file for the data portion of the message. We need
3053to access it both via a file descriptor and a stream. Try to make the
41313d92 3054directory if it isn't there. */
059ec3d9 3055
41313d92 3056spool_name = spool_fname(US"input", message_subdir, message_id, US"-D");
a2da3176
JH
3057DEBUG(D_receive) debug_printf("Data file name: %s\n", spool_name);
3058
3059if ((data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE)) < 0)
059ec3d9
PH
3060 {
3061 if (errno == ENOENT)
3062 {
0971ec06 3063 (void) directory_make(spool_directory,
41313d92
JH
3064 spool_sname(US"input", message_subdir),
3065 INPUT_DIRECTORY_MODE, TRUE);
059ec3d9
PH
3066 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
3067 }
3068 if (data_fd < 0)
3069 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
3070 spool_name, strerror(errno));
3071 }
3072
3073/* Make sure the file's group is the Exim gid, and double-check the mode
3074because the group setting doesn't always get set automatically. */
3075
1ac6b2e7
JH
3076if (fchown(data_fd, exim_uid, exim_gid))
3077 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
3078 "Failed setting ownership on spool file %s: %s",
3079 spool_name, strerror(errno));
ff790e47 3080(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
3081
3082/* We now have data file open. Build a stream for it and lock it. We lock only
3083the first line of the file (containing the message ID) because otherwise there
3084are problems when Exim is run under Cygwin (I'm told). See comments in
3085spool_in.c, where the same locking is done. */
3086
3087data_file = fdopen(data_fd, "w+");
3088lock_data.l_type = F_WRLCK;
3089lock_data.l_whence = SEEK_SET;
3090lock_data.l_start = 0;
3091lock_data.l_len = SPOOL_DATA_START_OFFSET;
3092
3093if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
3094 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
3095 errno, strerror(errno));
3096
3097/* We have an open, locked data file. Write the message id to it to make it
3098self-identifying. Then read the remainder of the input of this message and
3099write it to the data file. If the variable next != NULL, it contains the first
3100data line (which was read as a header but then turned out not to have the right
3101format); write it (remembering that it might contain binary zeros). The result
3102of fwrite() isn't inspected; instead we call ferror() below. */
3103
3104fprintf(data_file, "%s-D\n", message_id);
3105if (next != NULL)
3106 {
3107 uschar *s = next->text;
3108 int len = next->slen;
1ac6b2e7 3109 len = fwrite(s, 1, len, data_file); len = len; /* compiler quietening */
059ec3d9
PH
3110 body_linecount++; /* Assumes only 1 line */
3111 }
3112
3113/* Note that we might already be at end of file, or the logical end of file
3114(indicated by '.'), or might have encountered an error while writing the
3115message id or "next" line. */
3116
3117if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
3118 {
3119 if (smtp_input)
3120 {
328c5688
JH
3121 message_ended = chunking_state <= CHUNKING_OFFERED
3122 ? read_message_data_smtp(data_file)
3123 : spool_wireformat
3124 ? read_message_bdat_smtp_wire(data_file)
3125 : read_message_bdat_smtp(data_file);
059ec3d9
PH
3126 receive_linecount++; /* The terminating "." line */
3127 }
3128 else message_ended = read_message_data(data_file);
3129
3130 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 3131 message_linecount += body_linecount;
059ec3d9 3132
7e3ce68e 3133 switch (message_ended)
059ec3d9 3134 {
7e3ce68e 3135 /* Handle premature termination of SMTP */
059ec3d9 3136
7e3ce68e
JH
3137 case END_EOF:
3138 if (smtp_input)
3139 {
3140 Uunlink(spool_name); /* Lose data file when closed */
57cc2785 3141 cancel_cutthrough_connection(TRUE, US"sender closed connection");
7e3ce68e
JH
3142 message_id[0] = 0; /* Indicate no message accepted */
3143 smtp_reply = handle_lost_connection(US"");
3144 smtp_yield = FALSE;
3145 goto TIDYUP; /* Skip to end of function */
3146 }
3147 break;
059ec3d9 3148
7e3ce68e
JH
3149 /* Handle message that is too big. Don't use host_or_ident() in the log
3150 message; we want to see the ident value even for non-remote messages. */
059ec3d9 3151
7e3ce68e
JH
3152 case END_SIZE:
3153 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3154 cancel_cutthrough_connection(TRUE, US"mail too big");
7e3ce68e 3155 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
059ec3d9 3156
7e3ce68e
JH
3157 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3158 "message too big: read=%d max=%d",
3159 sender_address,
3160 (sender_fullhost == NULL)? "" : " H=",
3161 (sender_fullhost == NULL)? US"" : sender_fullhost,
3162 (sender_ident == NULL)? "" : " U=",
3163 (sender_ident == NULL)? US"" : sender_ident,
3164 message_size,
3165 thismessage_size_limit);
3166
3167 if (smtp_input)
3168 {
3169 smtp_reply = US"552 Message size exceeds maximum permitted";
3170 message_id[0] = 0; /* Indicate no message accepted */
3171 goto TIDYUP; /* Skip to end of function */
3172 }
3173 else
3174 {
3175 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3176 give_local_error(ERRMESS_TOOBIG,
3177 string_sprintf("message too big (max=%d)", thismessage_size_limit),
3178 US"message rejected: ", error_rc, data_file, header_list);
3179 /* Does not return */
3180 }
3181 break;
3182
3183 /* Handle bad BDAT protocol sequence */
3184
3185 case END_PROTOCOL:
3186 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3187 cancel_cutthrough_connection(TRUE, US"sender protocol error");
7e3ce68e
JH
3188 smtp_reply = US""; /* Response already sent */
3189 message_id[0] = 0; /* Indicate no message accepted */
3190 goto TIDYUP; /* Skip to end of function */
059ec3d9
PH
3191 }
3192 }
3193
3194/* Restore the standard SIGALRM handler for any subsequent processing. (For
3195example, there may be some expansion in an ACL that uses a timer.) */
3196
3197os_non_restarting_signal(SIGALRM, sigalrm_handler);
3198
3199/* The message body has now been read into the data file. Call fflush() to
3200empty the buffers in C, and then call fsync() to get the data written out onto
3201the disk, as fflush() doesn't do this (or at least, it isn't documented as
3202having to do this). If there was an I/O error on either input or output,
3203attempt to send an error message, and unlink the spool file. For non-SMTP input
3204we can then give up. Note that for SMTP input we must swallow the remainder of
3205the input in cases of output errors, since the far end doesn't expect to see
3206anything until the terminating dot line is sent. */
3207
3208if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 3209 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
3210 {
3211 uschar *msg_errno = US strerror(errno);
3212 BOOL input_error = (receive_ferror)() != 0;
3213 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
3214 input_error? "Input read" : "Spool write",
3215 msg_errno,
3216 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
3217
3218 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
3219 Uunlink(spool_name); /* Lose the data file */
57cc2785 3220 cancel_cutthrough_connection(TRUE, US"error writing spoolfile");
059ec3d9
PH
3221
3222 if (smtp_input)
3223 {
3224 if (input_error)
3225 smtp_reply = US"451 Error while reading input data";
3226 else
3227 {
3228 smtp_reply = US"451 Error while writing spool file";
3229 receive_swallow_smtp();
3230 }
3231 message_id[0] = 0; /* Indicate no message accepted */
3232 goto TIDYUP; /* Skip to end of function */
3233 }
3234
3235 else
3236 {
3237 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3238 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
3239 header_list);
3240 /* Does not return */
3241 }
3242 }
3243
3244
3245/* No I/O errors were encountered while writing the data file. */
3246
3247DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
3248
3249
3250/* If there were any bad addresses extracted by -t, or there were no recipients
3251left after -t, send a message to the sender of this message, or write it to
3252stderr if the error handling option is set that way. Note that there may
3253legitimately be no recipients for an SMTP message if they have all been removed
3254by "discard".
3255
3256We need to rewind the data file in order to read it. In the case of no
3257recipients or stderr error writing, throw the data file away afterwards, and
3258exit. (This can't be SMTP, which always ensures there's at least one
3259syntactically good recipient address.) */
3260
3261if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
3262 {
3263 DEBUG(D_receive)
3264 {
3265 if (recipients_count == 0) debug_printf("*** No recipients\n");
3266 if (bad_addresses != NULL)
3267 {
3268 error_block *eblock = bad_addresses;
3269 debug_printf("*** Bad address(es)\n");
3270 while (eblock != NULL)
3271 {
3272 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
3273 eblock = eblock->next;
3274 }
3275 }
3276 }
3277
3278 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3279
3280 /* If configured to send errors to the sender, but this fails, force
3281 a failure error code. We use a special one for no recipients so that it
3282 can be detected by the autoreply transport. Otherwise error_rc is set to
3283 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
3284 it is EXIT_SUCCESS. */
3285
3286 if (error_handling == ERRORS_SENDER)
3287 {
3288 if (!moan_to_sender(
3289 (bad_addresses == NULL)?
3290 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
3291 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
3292 bad_addresses, header_list, data_file, FALSE))
3293 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
3294 }
3295 else
3296 {
3297 if (bad_addresses == NULL)
3298 {
3299 if (extracted_ignored)
3300 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3301 else
3302 fprintf(stderr, "exim: no recipients in message\n");
3303 }
3304 else
3305 {
3306 fprintf(stderr, "exim: invalid address%s",
3307 (bad_addresses->next == NULL)? ":" : "es:\n");
3308 while (bad_addresses != NULL)
3309 {
3310 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3311 bad_addresses->text2);
3312 bad_addresses = bad_addresses->next;
3313 }
3314 }
3315 }
3316
3317 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3318 {
3319 Uunlink(spool_name);
f1e894f3 3320 (void)fclose(data_file);
059ec3d9
PH
3321 exim_exit(error_rc);
3322 }
3323 }
3324
3325/* Data file successfully written. Generate text for the Received: header by
3326expanding the configured string, and adding a timestamp. By leaving this
3327operation till now, we ensure that the timestamp is the time that message
3328reception was completed. However, this is deliberately done before calling the
3329data ACL and local_scan().
3330
3331This Received: header may therefore be inspected by the data ACL and by code in
3332the local_scan() function. When they have run, we update the timestamp to be
3333the final time of reception.
3334
3335If there is just one recipient, set up its value in the $received_for variable
3336for use when we generate the Received: header.
3337
3338Note: the checking for too many Received: headers is handled by the delivery
3339code. */
e4bdf652 3340/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3341
e4bdf652 3342if (received_header->text == NULL) /* Non-cutthrough case */
059ec3d9 3343 {
e4bdf652 3344 received_header_gen();
059ec3d9 3345
e4bdf652 3346 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3347
e4bdf652
JH
3348 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3349 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3350
e4bdf652
JH
3351 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3352 now, before running the DATA ACL. */
059ec3d9 3353
578d43dc 3354 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652 3355 }
817d9f57 3356else
e4bdf652
JH
3357 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3358 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3359
3360/* If an ACL is specified for checking things at this stage of reception of a
3361message, run it, unless all the recipients were removed by "discard" in earlier
3362ACLs. That is the only case in which recipients_count can be zero at this
3363stage. Set deliver_datafile to point to the data file so that $message_body and
3364$message_body_end can be extracted if needed. Allow $recipients in expansions.
3365*/
3366
3367deliver_datafile = data_fd;
4e88a19f 3368user_msg = NULL;
059ec3d9 3369
0e20aff9
MH
3370enable_dollar_recipients = TRUE;
3371
059ec3d9 3372if (recipients_count == 0)
7e3ce68e
JH
3373 blackholed_by = recipients_discarded ? US"MAIL ACL" : US"RCPT ACL";
3374
059ec3d9
PH
3375else
3376 {
059ec3d9
PH
3377 /* Handle interactive SMTP messages */
3378
3379 if (smtp_input && !smtp_batched_input)
3380 {
8523533c 3381
80a47a2c
TK
3382#ifndef DISABLE_DKIM
3383 if (!dkim_disable_verify)
3384 {
3385 /* Finish verification, this will log individual signature results to
3386 the mainlog */
3387 dkim_exim_verify_finish();
3388
3389 /* Check if we must run the DKIM ACL */
7e3ce68e 3390 if (acl_smtp_dkim && dkim_verify_signers && *dkim_verify_signers)
80a47a2c
TK
3391 {
3392 uschar *dkim_verify_signers_expanded =
3393 expand_string(dkim_verify_signers);
7e3ce68e 3394 if (!dkim_verify_signers_expanded)
80a47a2c
TK
3395 log_write(0, LOG_MAIN|LOG_PANIC,
3396 "expansion of dkim_verify_signers option failed: %s",
3397 expand_string_message);
7e3ce68e 3398
80a47a2c
TK
3399 else
3400 {
3401 int sep = 0;
55414b25 3402 const uschar *ptr = dkim_verify_signers_expanded;
80a47a2c 3403 uschar *item = NULL;
9e5d6b55
TK
3404 uschar *seen_items = NULL;
3405 int seen_items_size = 0;
3406 int seen_items_offset = 0;
9122af94
TK
3407 /* Default to OK when no items are present */
3408 rc = OK;
7e3ce68e 3409 while ((item = string_nextinlist(&ptr, &sep, NULL, 0)))
80a47a2c 3410 {
6119d1ea 3411 /* Prevent running ACL for an empty item */
7e3ce68e 3412 if (!item || !*item) continue;
5032d1cf
JH
3413
3414 /* Only run ACL once for each domain or identity,
3415 no matter how often it appears in the expanded list. */
3416 if (seen_items)
6119d1ea 3417 {
ae9094bf 3418 uschar *seen_item = NULL;
55414b25 3419 const uschar *seen_items_list = seen_items;
5032d1cf 3420 BOOL seen_this_item = FALSE;
61147df4 3421
ae9094bf 3422 while ((seen_item = string_nextinlist(&seen_items_list, &sep,
7e3ce68e 3423 NULL, 0)))
5032d1cf
JH
3424 if (Ustrcmp(seen_item,item) == 0)
3425 {
3426 seen_this_item = TRUE;
3427 break;
3428 }
3429
3430 if (seen_this_item)
6119d1ea
TK
3431 {
3432 DEBUG(D_receive)
5032d1cf
JH
3433 debug_printf("acl_smtp_dkim: skipping signer %s, "
3434 "already seen\n", item);
6119d1ea
TK
3435 continue;
3436 }
61147df4 3437
5032d1cf
JH
3438 seen_items = string_append(seen_items, &seen_items_size,
3439 &seen_items_offset, 1, ":");
6119d1ea
TK
3440 }
3441
5032d1cf
JH
3442 seen_items = string_append(seen_items, &seen_items_size,
3443 &seen_items_offset, 1, item);
4a73449b 3444 seen_items[seen_items_offset] = '\0';
6119d1ea
TK
3445
3446 DEBUG(D_receive)
5032d1cf
JH
3447 debug_printf("calling acl_smtp_dkim for dkim_cur_signer=%s\n",
3448 item);
6119d1ea 3449
80a47a2c 3450 dkim_exim_acl_setup(item);
5032d1cf
JH
3451 rc = acl_check(ACL_WHERE_DKIM, NULL, acl_smtp_dkim,
3452 &user_msg, &log_msg);
6119d1ea
TK
3453
3454 if (rc != OK)
5032d1cf
JH
3455 {
3456 DEBUG(D_receive)
3457 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, "
3458 "skipping remaining items\n", rc, item);
57cc2785 3459 cancel_cutthrough_connection(TRUE, US"dkim acl not ok");
5032d1cf
JH
3460 break;
3461 }
80a47a2c 3462 }
578d43dc 3463 add_acl_headers(ACL_WHERE_DKIM, US"DKIM");
80a47a2c
TK
3464 if (rc == DISCARD)
3465 {
3466 recipients_count = 0;
3467 blackholed_by = US"DKIM ACL";
3468 if (log_msg != NULL)
3469 blackhole_log_msg = string_sprintf(": %s", log_msg);
3470 }
3471 else if (rc != OK)
3472 {
3473 Uunlink(spool_name);
3474 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
85ffcba6 3475 smtp_yield = FALSE; /* No more messages after dropped connection */
80a47a2c
TK
3476 smtp_reply = US""; /* Indicate reply already sent */
3477 message_id[0] = 0; /* Indicate no message accepted */
3478 goto TIDYUP; /* Skip to end of function */
3479 }
3480 }
3481 }
3482 }
4a8ce2d8 3483#endif /* DISABLE_DKIM */
fb2274d4 3484
8523533c 3485#ifdef WITH_CONTENT_SCAN
80a47a2c
TK
3486 if (recipients_count > 0 &&
3487 acl_smtp_mime != NULL &&
54cdb463
PH
3488 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
3489 goto TIDYUP;
8523533c
TK
3490#endif /* WITH_CONTENT_SCAN */
3491
4840604e
TL
3492#ifdef EXPERIMENTAL_DMARC
3493 dmarc_up = dmarc_store_data(from_header);
3494#endif /* EXPERIMENTAL_DMARC */
3495
8ccd00b1
JH
3496#ifndef DISABLE_PRDR
3497 if (prdr_requested && recipients_count > 1 && acl_smtp_data_prdr)
fd98a5c6
JH
3498 {
3499 unsigned int c;
3500 int all_pass = OK;
3501 int all_fail = FAIL;
3502
3503 smtp_printf("353 PRDR content analysis beginning\r\n");
3504 /* Loop through recipients, responses must be in same order received */
3505 for (c = 0; recipients_count > c; c++)
3506 {
3507 uschar * addr= recipients_list[c].address;
3508 uschar * msg= US"PRDR R=<%s> %s";
3509 uschar * code;
3510 DEBUG(D_receive)
3511 debug_printf("PRDR processing recipient %s (%d of %d)\n",
3512 addr, c+1, recipients_count);
3513 rc = acl_check(ACL_WHERE_PRDR, addr,
3514 acl_smtp_data_prdr, &user_msg, &log_msg);
3515
3516 /* If any recipient rejected content, indicate it in final message */
3517 all_pass |= rc;
3518 /* If all recipients rejected, indicate in final message */
3519 all_fail &= rc;
3520
3521 switch (rc)
3522 {
3523 case OK: case DISCARD: code = US"250"; break;
3524 case DEFER: code = US"450"; break;
3525 default: code = US"550"; break;
3526 }
3527 if (user_msg != NULL)
3528 smtp_user_msg(code, user_msg);
3529 else
3530 {
3531 switch (rc)
3532 {
3533 case OK: case DISCARD:
3534 msg = string_sprintf(CS msg, addr, "acceptance"); break;
3535 case DEFER:
3536 msg = string_sprintf(CS msg, addr, "temporary refusal"); break;
3537 default:
3538 msg = string_sprintf(CS msg, addr, "refusal"); break;
3539 }
3540 smtp_user_msg(code, msg);
3541 }
3542 if (log_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, log_msg);
3543 else if (user_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, user_msg);
112b6a93 3544 else log_write(0, LOG_MAIN, "%s", CS msg);
fd98a5c6
JH
3545
3546 if (rc != OK) { receive_remove_recipient(addr); c--; }
3547 }
3548 /* Set up final message, used if data acl gives OK */
3549 smtp_reply = string_sprintf("%s id=%s message %s",
3550 all_fail == FAIL ? US"550" : US"250",
3551 message_id,
3552 all_fail == FAIL
3553 ? US"rejected for all recipients"
3554 : all_pass == OK
3555 ? US"accepted"
3556 : US"accepted for some recipients");
3557 if (recipients_count == 0)
3558 {
3559 message_id[0] = 0; /* Indicate no message accepted */
3560 goto TIDYUP;
3561 }
3562 }
3563 else
3564 prdr_requested = FALSE;
8ccd00b1 3565#endif /* !DISABLE_PRDR */
fd98a5c6 3566
54cdb463
PH
3567 /* Check the recipients count again, as the MIME ACL might have changed
3568 them. */
8523533c 3569
059ec3d9
PH
3570 if (acl_smtp_data != NULL && recipients_count > 0)
3571 {
059ec3d9 3572 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
578d43dc 3573 add_acl_headers(ACL_WHERE_DATA, US"DATA");
059ec3d9
PH
3574 if (rc == DISCARD)
3575 {
3576 recipients_count = 0;
3577 blackholed_by = US"DATA ACL";
57cc2785 3578 if (log_msg)
8e669ac1 3579 blackhole_log_msg = string_sprintf(": %s", log_msg);
57cc2785 3580 cancel_cutthrough_connection(TRUE, US"data acl discard");
059ec3d9
PH
3581 }
3582 else if (rc != OK)
3583 {
3584 Uunlink(spool_name);
57cc2785 3585 cancel_cutthrough_connection(TRUE, US"data acl not ok");
8523533c
TK
3586#ifdef WITH_CONTENT_SCAN
3587 unspool_mbox();
3588#endif
6f0c431a
PP
3589#ifdef EXPERIMENTAL_DCC
3590 dcc_ok = 0;
3591#endif
059ec3d9 3592 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
85ffcba6 3593 smtp_yield = FALSE; /* No more messages after dropped connection */
059ec3d9
PH
3594 smtp_reply = US""; /* Indicate reply already sent */
3595 message_id[0] = 0; /* Indicate no message accepted */
3596 goto TIDYUP; /* Skip to end of function */
3597 }
3598 }
3599 }
3600
3601 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
3602 we cannot take different actions for permanent and temporary rejections. */
3603
54cdb463 3604 else
059ec3d9 3605 {
54cdb463
PH
3606
3607#ifdef WITH_CONTENT_SCAN
3608 if (acl_not_smtp_mime != NULL &&
3609 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
3610 &blackholed_by))
3611 goto TIDYUP;
3612#endif /* WITH_CONTENT_SCAN */
3613
3614 if (acl_not_smtp != NULL)
059ec3d9 3615 {
54cdb463
PH
3616 uschar *user_msg, *log_msg;
3617 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
3618 if (rc == DISCARD)
059ec3d9 3619 {
54cdb463
PH
3620 recipients_count = 0;
3621 blackholed_by = US"non-SMTP ACL";
3622 if (log_msg != NULL)
3623 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 3624 }
54cdb463 3625 else if (rc != OK)
059ec3d9 3626 {
54cdb463
PH
3627 Uunlink(spool_name);
3628#ifdef WITH_CONTENT_SCAN
3629 unspool_mbox();
3630#endif
6f0c431a
PP
3631#ifdef EXPERIMENTAL_DCC
3632 dcc_ok = 0;
3633#endif
6ea85e9a
PH
3634 /* The ACL can specify where rejections are to be logged, possibly
3635 nowhere. The default is main and reject logs. */
3636
3637 if (log_reject_target != 0)
3638 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3639 sender_address, log_msg);
3640
54cdb463
PH
3641 if (user_msg == NULL) user_msg = US"local configuration problem";
3642 if (smtp_batched_input)
3643 {
3644 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3645 /* Does not return */
3646 }
3647 else
3648 {
3649 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3650 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3651 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3652 header_list);
3653 /* Does not return */
3654 }
059ec3d9 3655 }
578d43dc 3656 add_acl_headers(ACL_WHERE_NOTSMTP, US"non-SMTP");
059ec3d9 3657 }
059ec3d9
PH
3658 }
3659
54cdb463
PH
3660 /* The applicable ACLs have been run */
3661
059ec3d9
PH
3662 if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */
3663 if (queue_only_policy) queued_by = US"ACL";
059ec3d9
PH
3664 }
3665
8523533c
TK
3666#ifdef WITH_CONTENT_SCAN
3667unspool_mbox();
3668#endif
3669
6a8f9482
TK
3670#ifdef EXPERIMENTAL_DCC
3671dcc_ok = 0;
3672#endif
3673
3674
059ec3d9
PH
3675/* The final check on the message is to run the scan_local() function. The
3676version supplied with Exim always accepts, but this is a hook for sysadmins to
3677supply their own checking code. The local_scan() function is run even when all
3678the recipients have been discarded. */
d315eda1 3679/*XXS could we avoid this for the standard case, given that few people will use it? */
059ec3d9
PH
3680
3681lseek(data_fd, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);