DKIM: Allow the DKIM ACL to override verification results. Bug 2186
[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");
9bfb7e1b 193 exim_exit(EXIT_FAILURE, NULL);
ddf1b11a 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
9bfb7e1b 346exim_exit(EXIT_FAILURE, NULL);
059ec3d9
PH
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 {
7d758a6a 922 switch ((ch = bdat_getc(GETC_BUFFER_UNLIMITED)))
1ebe15c3
JH
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
0d81dabc 1029for (;;)
328c5688 1030 {
0d81dabc
JH
1031 if (chunking_data_left > 0)
1032 {
1033 unsigned len = MAX(chunking_data_left, thismessage_size_limit - message_size + 1);
1034 uschar * buf = bdat_getbuf(&len);
328c5688 1035
0d81dabc
JH
1036 message_size += len;
1037 if (fout && fwrite(buf, len, 1, fout) != 1) return END_WERROR;
1038 }
1039 else switch (ch = bdat_getc(GETC_BUFFER_UNLIMITED))
1040 {
1041 case EOF: return END_EOF;
1042 case EOD: return END_DOT;
1043 case ERR: return END_PROTOCOL;
1044
1045 default:
1046 message_size++;
1047 /*XXX not done:
1048 linelength
1049 max_received_linelength
1050 body_linecount
1051 body_zerocount
1052 */
1053 if (fout && fputc(ch, fout) == EOF) return END_WERROR;
1054 break;
1055 }
1056 if (message_size > thismessage_size_limit) return END_SIZE;
328c5688
JH
1057 }
1058/*NOTREACHED*/
1059}
1060
7e3ce68e
JH
1061
1062
1063
059ec3d9
PH
1064/*************************************************
1065* Swallow SMTP message *
1066*************************************************/
1067
1068/* This function is called when there has been some kind of error while reading
1069an SMTP message, and the remaining data may need to be swallowed. It is global
1070because it is called from smtp_closedown() to shut down an incoming call
1071tidily.
1072
1073Argument: a FILE from which to read the message
1074Returns: nothing
1075*/
1076
1077void
1078receive_swallow_smtp(void)
1079{
7e3ce68e 1080/*XXX CHUNKING: not enough. read chunks until RSET? */
059ec3d9
PH
1081if (message_ended >= END_NOTENDED)
1082 message_ended = read_message_data_smtp(NULL);
1083}
1084
1085
1086
1087/*************************************************
1088* Handle lost SMTP connection *
1089*************************************************/
1090
1091/* This function logs connection loss incidents and generates an appropriate
1092SMTP response.
1093
1094Argument: additional data for the message
1095Returns: the SMTP response
1096*/
1097
1098static uschar *
1099handle_lost_connection(uschar *s)
1100{
1101log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
1102 "%s lost while reading message data%s", smtp_get_connection_info(), s);
eea0defe 1103smtp_notquit_exit(US"connection-lost", NULL, NULL);
059ec3d9
PH
1104return US"421 Lost incoming connection";
1105}
1106
1107
1108
1109
1110/*************************************************
1111* Handle a non-smtp reception error *
1112*************************************************/
1113
1114/* This function is called for various errors during the reception of non-SMTP
1115messages. It either sends a message to the sender of the problem message, or it
1116writes to the standard error stream.
1117
1118Arguments:
1119 errcode code for moan_to_sender(), identifying the error
1120 text1 first message text, passed to moan_to_sender()
1121 text2 second message text, used only for stderrr
1122 error_rc code to pass to exim_exit if no problem
1123 f FILE containing body of message (may be stdin)
1124 hptr pointer to instore headers or NULL
1125
1126Returns: calls exim_exit(), which does not return
1127*/
1128
1129static void
1130give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
1131 FILE *f, header_line *hptr)
1132{
1133if (error_handling == ERRORS_SENDER)
1134 {
1135 error_block eblock;
1136 eblock.next = NULL;
1137 eblock.text1 = text1;
37f3dc43 1138 eblock.text2 = US"";
059ec3d9
PH
1139 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
1140 error_rc = EXIT_FAILURE;
1141 }
37f3dc43
JH
1142else
1143 fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 1144(void)fclose(f);
9bfb7e1b 1145exim_exit(error_rc, US"");
059ec3d9
PH
1146}
1147
1148
1149
1150/*************************************************
1151* Add header lines set up by ACL *
1152*************************************************/
1153
850635b6
PH
1154/* This function is called to add the header lines that were set up by
1155statements in an ACL to the list of headers in memory. It is done in two stages
1156like this, because when the ACL for RCPT is running, the other headers have not
1157yet been received. This function is called twice; once just before running the
1158DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
1159are visible to the DATA ACL.
059ec3d9
PH
1160
1161Originally these header lines were added at the end. Now there is support for
1162three different places: top, bottom, and after the Received: header(s). There
1163will always be at least one Received: header, even if it is marked deleted, and
1164even if something else has been put in front of it.
1165
1166Arguments:
1167 acl_name text to identify which ACL
1168
1169Returns: nothing
1170*/
1171
1172static void
578d43dc 1173add_acl_headers(int where, uschar *acl_name)
059ec3d9
PH
1174{
1175header_line *h, *next;
1176header_line *last_received = NULL;
e7568d51 1177
578d43dc
JH
1178switch(where)
1179 {
1180 case ACL_WHERE_DKIM:
1181 case ACL_WHERE_MIME:
af4a1bca 1182 case ACL_WHERE_DATA:
57cc2785
JH
1183 if ( cutthrough.fd >= 0 && cutthrough.delivery
1184 && (acl_removed_headers || acl_added_headers))
578d43dc
JH
1185 {
1186 log_write(0, LOG_MAIN|LOG_PANIC, "Header modification in data ACLs"
af4a1bca 1187 " will not take effect on cutthrough deliveries");
578d43dc
JH
1188 return;
1189 }
1190 }
1191
57cc2785 1192if (acl_removed_headers)
e7568d51 1193 {
e1d04f48 1194 DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers removed by %s ACL:\n", acl_name);
e7568d51 1195
57cc2785 1196 for (h = header_list; h; h = h->next) if (h->type != htype_old)
e7568d51 1197 {
55414b25 1198 const uschar * list = acl_removed_headers;
e7568d51
TL
1199 int sep = ':'; /* This is specified as a colon-separated list */
1200 uschar *s;
1201 uschar buffer[128];
4a142059
JH
1202
1203 while ((s = string_nextinlist(&list, &sep, buffer, sizeof(buffer))))
1204 if (header_testname(h, s, Ustrlen(s), FALSE))
e7568d51
TL
1205 {
1206 h->type = htype_old;
e1d04f48 1207 DEBUG(D_receive|D_acl) debug_printf_indent(" %s", h->text);
e7568d51 1208 }
e7568d51
TL
1209 }
1210 acl_removed_headers = NULL;
e1d04f48 1211 DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
e7568d51 1212 }
059ec3d9 1213
57cc2785 1214if (!acl_added_headers) return;
e1d04f48 1215DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers added by %s ACL:\n", acl_name);
059ec3d9 1216
57cc2785 1217for (h = acl_added_headers; h; h = next)
059ec3d9
PH
1218 {
1219 next = h->next;
1220
1221 switch(h->type)
1222 {
1223 case htype_add_top:
1224 h->next = header_list;
1225 header_list = h;
e1d04f48 1226 DEBUG(D_receive|D_acl) debug_printf_indent(" (at top)");
059ec3d9
PH
1227 break;
1228
1229 case htype_add_rec:
1230 if (last_received == NULL)
1231 {
1232 last_received = header_list;
1233 while (!header_testname(last_received, US"Received", 8, FALSE))
1234 last_received = last_received->next;
1235 while (last_received->next != NULL &&
1236 header_testname(last_received->next, US"Received", 8, FALSE))
1237 last_received = last_received->next;
1238 }
1239 h->next = last_received->next;
1240 last_received->next = h;
e1d04f48 1241 DEBUG(D_receive|D_acl) debug_printf_indent(" (after Received:)");
059ec3d9
PH
1242 break;
1243
8523533c
TK
1244 case htype_add_rfc:
1245 /* add header before any header which is NOT Received: or Resent- */
1246 last_received = header_list;
1247 while ( (last_received->next != NULL) &&
1248 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
1249 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
1250 last_received = last_received->next;
1251 /* last_received now points to the last Received: or Resent-* header
1252 in an uninterrupted chain of those header types (seen from the beginning
1253 of all headers. Our current header must follow it. */
1254 h->next = last_received->next;
1255 last_received->next = h;
e1d04f48 1256 DEBUG(D_receive|D_acl) debug_printf_indent(" (before any non-Received: or Resent-*: header)");
8523533c
TK
1257 break;
1258
059ec3d9
PH
1259 default:
1260 h->next = NULL;
1261 header_last->next = h;
1262 break;
1263 }
1264
1265 if (h->next == NULL) header_last = h;
1266
1267 /* Check for one of the known header types (From:, To:, etc.) though in
1268 practice most added headers are going to be "other". Lower case
1269 identification letters are never stored with the header; they are used
1270 for existence tests when messages are received. So discard any lower case
1271 flag values. */
1272
1273 h->type = header_checkname(h, FALSE);
1274 if (h->type >= 'a') h->type = htype_other;
1275
e1d04f48 1276 DEBUG(D_receive|D_acl) debug_printf_indent(" %s", header_last->text);
059ec3d9
PH
1277 }
1278
71fafd95 1279acl_added_headers = NULL;
e1d04f48 1280DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
059ec3d9
PH
1281}
1282
1283
1284
1285/*************************************************
1286* Add host information for log line *
1287*************************************************/
1288
1289/* Called for acceptance and rejecting log lines. This adds information about
1290the calling host to a string that is being built dynamically.
1291
1292Arguments:
1293 s the dynamic string
059ec3d9
PH
1294
1295Returns: the extended string
1296*/
1297
acec9514
JH
1298static gstring *
1299add_host_info_for_log(gstring * g)
059ec3d9 1300{
fc16abb4 1301if (sender_fullhost)
059ec3d9 1302 {
fc16abb4 1303 if (LOGGING(dnssec) && sender_host_dnssec) /*XXX sender_helo_dnssec? */
acec9514
JH
1304 g = string_catn(g, US" DS", 3);
1305 g = string_append(g, 2, US" H=", sender_fullhost);
6c6d6e48 1306 if (LOGGING(incoming_interface) && interface_address != NULL)
059ec3d9 1307 {
acec9514 1308 g = string_cat(g,
fc16abb4 1309 string_sprintf(" I=[%s]:%d", interface_address, interface_port));
059ec3d9
PH
1310 }
1311 }
a2673768
JH
1312if (tcp_in_fastopen && !tcp_in_fastopen_logged)
1313 {
acec9514 1314 g = string_catn(g, US" TFO", 4);
a2673768
JH
1315 tcp_in_fastopen_logged = TRUE;
1316 }
1317if (sender_ident)
acec9514 1318 g = string_append(g, 2, US" U=", sender_ident);
a2673768 1319if (received_protocol)
acec9514
JH
1320 g = string_append(g, 2, US" P=", received_protocol);
1321return g;
059ec3d9
PH
1322}
1323
1324
1325
63955bf2 1326#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1327
1328/*************************************************
54cdb463
PH
1329* Run the MIME ACL on a message *
1330*************************************************/
1331
1332/* This code is in a subroutine so that it can be used for both SMTP
1333and non-SMTP messages. It is called with a non-NULL ACL pointer.
1334
1335Arguments:
1336 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1337 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1338 smtp_reply_ptr Where SMTP reply is being built
1339 blackholed_by_ptr Where "blackholed by" message is being built
1340
1341Returns: TRUE to carry on; FALSE to abandon the message
1342*/
1343
1344static BOOL
1345run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1346 uschar **blackholed_by_ptr)
1347{
1348FILE *mbox_file;
1349uschar rfc822_file_path[2048];
1350unsigned long mbox_size;
1351header_line *my_headerlist;
1352uschar *user_msg, *log_msg;
1353int mime_part_count_buffer = -1;
040721f2 1354uschar * mbox_filename;
7156b1ef 1355int rc = OK;
54cdb463
PH
1356
1357memset(CS rfc822_file_path,0,2048);
1358
1359/* check if it is a MIME message */
040721f2
JH
1360
1361for (my_headerlist = header_list; my_headerlist; my_headerlist = my_headerlist->next)
1362 if ( my_headerlist->type != '*' /* skip deleted headers */
1363 && strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0
1364 )
4e88a19f 1365 {
54cdb463
PH
1366 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1367 goto DO_MIME_ACL;
4e88a19f 1368 }
54cdb463
PH
1369
1370DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1371return TRUE;
1372
1373DO_MIME_ACL:
040721f2 1374
54cdb463 1375/* make sure the eml mbox file is spooled up */
040721f2
JH
1376if (!(mbox_file = spool_mbox(&mbox_size, NULL, &mbox_filename)))
1377 { /* error while spooling */
54cdb463
PH
1378 log_write(0, LOG_MAIN|LOG_PANIC,
1379 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1380 Uunlink(spool_name);
1381 unspool_mbox();
6f0c431a
PP
1382#ifdef EXPERIMENTAL_DCC
1383 dcc_ok = 0;
1384#endif
a5bd321b 1385 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1386 message_id[0] = 0; /* Indicate no message accepted */
1387 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1388 return FALSE; /* Indicate skip to end of receive function */
040721f2 1389 }
54cdb463
PH
1390
1391mime_is_rfc822 = 0;
1392
1393MIME_ACL_CHECK:
1394mime_part_count = -1;
1395rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1396(void)fclose(mbox_file);
54cdb463 1397
4e88a19f
PH
1398if (Ustrlen(rfc822_file_path) > 0)
1399 {
54cdb463
PH
1400 mime_part_count = mime_part_count_buffer;
1401
4e88a19f
PH
1402 if (unlink(CS rfc822_file_path) == -1)
1403 {
54cdb463
PH
1404 log_write(0, LOG_PANIC,
1405 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1406 goto END_MIME_ACL;
4e88a19f
PH
1407 }
1408 }
54cdb463
PH
1409
1410/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1411if (rc == OK)
1412 {
040721f2 1413 uschar * scandir;
e8bc7fca
JH
1414 struct dirent * entry;
1415 DIR * tempdir;
54cdb463 1416
040721f2 1417 scandir = string_copyn(mbox_filename, Ustrrchr(mbox_filename, '/') - mbox_filename);
54cdb463 1418
040721f2 1419 tempdir = opendir(CS scandir);
e8bc7fca 1420 for (;;)
4e88a19f 1421 {
e8bc7fca
JH
1422 if (!(entry = readdir(tempdir)))
1423 break;
1424 if (strncmpic(US entry->d_name, US"__rfc822_", 9) == 0)
4e88a19f 1425 {
e8bc7fca 1426 (void) string_format(rfc822_file_path, sizeof(rfc822_file_path),
040721f2 1427 "%s/%s", scandir, entry->d_name);
e1d04f48 1428 DEBUG(D_receive) debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
e8bc7fca 1429 rfc822_file_path);
4e88a19f
PH
1430 break;
1431 }
e8bc7fca 1432 }
4e88a19f 1433 closedir(tempdir);
54cdb463 1434
e8bc7fca 1435 if (entry)
4e88a19f 1436 {
e8bc7fca 1437 if ((mbox_file = Ufopen(rfc822_file_path, "rb")))
4e88a19f 1438 {
e8bc7fca
JH
1439 /* set RFC822 expansion variable */
1440 mime_is_rfc822 = 1;
1441 mime_part_count_buffer = mime_part_count;
1442 goto MIME_ACL_CHECK;
4e88a19f 1443 }
e8bc7fca
JH
1444 log_write(0, LOG_PANIC,
1445 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1446 unlink(CS rfc822_file_path);
4e88a19f
PH
1447 }
1448 }
54cdb463
PH
1449
1450END_MIME_ACL:
578d43dc 1451add_acl_headers(ACL_WHERE_MIME, US"MIME");
54cdb463
PH
1452if (rc == DISCARD)
1453 {
1454 recipients_count = 0;
1455 *blackholed_by_ptr = US"MIME ACL";
1456 }
1457else if (rc != OK)
1458 {
1459 Uunlink(spool_name);
1460 unspool_mbox();
6f0c431a
PP
1461#ifdef EXPERIMENTAL_DCC
1462 dcc_ok = 0;
1463#endif
4f6ae5c3
JH
1464 if ( smtp_input
1465 && smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1466 {
85ffcba6 1467 *smtp_yield_ptr = FALSE; /* No more messages after dropped connection */
f4c1088b 1468 *smtp_reply_ptr = US""; /* Indicate reply already sent */
4f6ae5c3 1469 }
54cdb463
PH
1470 message_id[0] = 0; /* Indicate no message accepted */
1471 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1472 }
54cdb463
PH
1473
1474return TRUE;
1475}
1476
63955bf2 1477#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1478
1479
e4bdf652
JH
1480
1481void
1482received_header_gen(void)
1483{
1484uschar *received;
1485uschar *timestamp;
1486header_line *received_header= header_list;
1487
1488timestamp = expand_string(US"${tod_full}");
1489if (recipients_count == 1) received_for = recipients_list[0].address;
1490received = expand_string(received_header_text);
1491received_for = NULL;
1492
d4ff61d1 1493if (!received)
e4bdf652
JH
1494 {
1495 if(spool_name[0] != 0)
1496 Uunlink(spool_name); /* Lose the data file */
1497 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1498 "(received_header_text) failed: %s", string_printing(received_header_text),
1499 expand_string_message);
1500 }
1501
1502/* The first element on the header chain is reserved for the Received header,
1503so all we have to do is fill in the text pointer, and set the type. However, if
1504the result of the expansion is an empty string, we leave the header marked as
1505"old" so as to refrain from adding a Received header. */
1506
1507if (received[0] == 0)
1508 {
1509 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1510 received_header->type = htype_old;
1511 }
1512else
1513 {
1514 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1515 received_header->type = htype_received;
1516 }
1517
1518received_header->slen = Ustrlen(received_header->text);
1519
1520DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1521 received_header->type, received_header->text);
1522}
1523
1524
1525
54cdb463 1526/*************************************************
059ec3d9
PH
1527* Receive message *
1528*************************************************/
1529
1530/* Receive a message on the given input, and put it into a pair of spool files.
1531Either a non-null list of recipients, or the extract flag will be true, or
1532both. The flag sender_local is true for locally generated messages. The flag
1533submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1534suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1535suppress_local_fixups" or -G was passed on the command-line.
1536The flag smtp_input is true if the message is to be
8800895a
PH
1537handled using SMTP conventions about termination and lines starting with dots.
1538For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1539
1540If a message was successfully read, message_id[0] will be non-zero.
1541
1542The general actions of this function are:
1543
1544 . Read the headers of the message (if any) into a chain of store
1545 blocks.
1546
1547 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1548 throw it away, unless the caller is trusted, or unless
1549 active_local_sender_retain is set - which can only happen if
1550 active_local_from_check is false.
059ec3d9
PH
1551
1552 . If recipients are to be extracted from the message, build the
1553 recipients list from the headers, removing any that were on the
1554 original recipients list (unless extract_addresses_remove_arguments is
1555 false), and at the same time, remove any bcc header that may be present.
1556
1557 . Get the spool file for the data, sort out its unique name, open
1558 and lock it (but don't give it the name yet).
1559
1560 . Generate a "Message-Id" header if the message doesn't have one, for
1561 locally-originated messages.
1562
1563 . Generate a "Received" header.
1564
1565 . Ensure the recipients list is fully qualified and rewritten if necessary.
1566
1567 . If there are any rewriting rules, apply them to the sender address
1568 and also to the headers.
1569
1570 . If there is no from: header, generate one, for locally-generated messages
1571 and messages in "submission mode" only.
1572
1573 . If the sender is local, check that from: is correct, and if not, generate
1574 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1575 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1576
1577 . If there is no "date" header, generate one, for locally-originated
1578 or submission mode messages only.
1579
1580 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1581 dot_ends mode, to the data file. Leave it open, to hold the lock.
1582
1583 . Write the envelope and the headers to a new file.
1584
1585 . Set the name for the header file; close it.
1586
1587 . Set the name for the data file; close it.
1588
1589Because this function can potentially be called many times in a single
1590SMTP connection, all store should be got by store_get(), so that it will be
1591automatically retrieved after the message is accepted.
1592
1593FUDGE: It seems that sites on the net send out messages with just LF
1594terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1595we make the CRs optional in all cases.
1596
1597July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1598new regime is now in place in which bare CRs in header lines are turned into LF
1599followed by a space, so as not to terminate the header line.
1600
1601February 2004: A bare LF in a header line in a message whose first line was
1602terminated by CRLF is treated in the same way as a bare CR.
1603
1604Arguments:
1605 extract_recip TRUE if recipients are to be extracted from the message's
1606 headers
1607
1608Returns: TRUE there are more messages to be read (SMTP input)
1609 FALSE there are no more messages to be read (non-SMTP input
1610 or SMTP connection collapsed, or other failure)
1611
1612When reading a message for filter testing, the returned value indicates
1613whether the headers (which is all that is read) were terminated by '.' or
1614not. */
1615
1616BOOL
1617receive_msg(BOOL extract_recip)
1618{
7156b1ef
NM
1619int i;
1620int rc = FAIL;
059ec3d9
PH
1621int msg_size = 0;
1622int process_info_len = Ustrlen(process_info);
1623int error_rc = (error_handling == ERRORS_SENDER)?
1624 errors_sender_rc : EXIT_FAILURE;
1625int header_size = 256;
acec9514 1626int start, end, domain;
059ec3d9
PH
1627int id_resolution;
1628int had_zero = 0;
d677b2f2 1629int prevlines_length = 0;
059ec3d9
PH
1630
1631register int ptr = 0;
1632
1633BOOL contains_resent_headers = FALSE;
1634BOOL extracted_ignored = FALSE;
1635BOOL first_line_ended_crlf = TRUE_UNSET;
1636BOOL smtp_yield = TRUE;
1637BOOL yield = FALSE;
1638
1639BOOL resents_exist = FALSE;
1640uschar *resent_prefix = US"";
1641uschar *blackholed_by = NULL;
04f7d5b9 1642uschar *blackhole_log_msg = US"";
c5430c20 1643enum {NOT_TRIED, TMP_REJ, PERM_REJ, ACCEPTED} cutthrough_done = NOT_TRIED;
059ec3d9
PH
1644
1645flock_t lock_data;
1646error_block *bad_addresses = NULL;
1647
1648uschar *frozen_by = NULL;
1649uschar *queued_by = NULL;
1650
acec9514
JH
1651uschar *errmsg;
1652gstring * g;
059ec3d9
PH
1653struct stat statbuf;
1654
4e88a19f 1655/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1656
1657uschar *smtp_reply = NULL;
4e88a19f 1658uschar *user_msg, *log_msg;
059ec3d9
PH
1659
1660/* Working header pointers */
1661
1662header_line *h, *next;
1663
2cbb4081 1664/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1665
1666BOOL date_header_exists = FALSE;
1667
1668/* Pointers to receive the addresses of headers whose contents we need. */
1669
1670header_line *from_header = NULL;
1671header_line *subject_header = NULL;
1672header_line *msgid_header = NULL;
1673header_line *received_header;
1674
4840604e
TL
1675#ifdef EXPERIMENTAL_DMARC
1676int dmarc_up = 0;
1677#endif /* EXPERIMENTAL_DMARC */
1678
059ec3d9
PH
1679/* Variables for use when building the Received: header. */
1680
059ec3d9
PH
1681uschar *timestamp;
1682int tslen;
1683
1684/* Release any open files that might have been cached while preparing to
1685accept the message - e.g. by verifying addresses - because reading a message
1686might take a fair bit of real time. */
1687
1688search_tidyup();
1689
e4bdf652
JH
1690/* Extracting the recipient list from an input file is incompatible with
1691cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1692to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1693if (extract_recip || !smtp_input)
57cc2785 1694 cancel_cutthrough_connection(TRUE, US"not smtp input");
e4bdf652 1695
059ec3d9
PH
1696/* Initialize the chain of headers by setting up a place-holder for Received:
1697header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1698pointing to the end of the chain to make adding headers simple. */
1699
1700received_header = header_list = header_last = store_get(sizeof(header_line));
1701header_list->next = NULL;
1702header_list->type = htype_old;
1703header_list->text = NULL;
1704header_list->slen = 0;
1705
1706/* Control block for the next header to be read. */
1707
1708next = store_get(sizeof(header_line));
1709next->text = store_get(header_size);
1710
1711/* Initialize message id to be null (indicating no message read), and the
1712header names list to be the normal list. Indicate there is no data file open
1713yet, initialize the size and warning count, and deal with no size limit. */
1714
1715message_id[0] = 0;
1716data_file = NULL;
1717data_fd = -1;
41313d92 1718spool_name = US"";
059ec3d9
PH
1719message_size = 0;
1720warning_count = 0;
d677b2f2 1721received_count = 1; /* For the one we will add */
059ec3d9
PH
1722
1723if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1724
2e0c1448 1725/* While reading the message, the following counts are computed. */
059ec3d9 1726
d677b2f2
PH
1727message_linecount = body_linecount = body_zerocount =
1728 max_received_linelength = 0;
059ec3d9 1729
80a47a2c 1730#ifndef DISABLE_DKIM
e983e85a
JH
1731/* Call into DKIM to set up the context. In CHUNKING mode
1732we clear the dot-stuffing flag */
1733if (smtp_input && !smtp_batched_input && !dkim_disable_verify)
1734 dkim_exim_verify_init(chunking_state <= CHUNKING_OFFERED);
fb2274d4
TK
1735#endif
1736
4840604e
TL
1737#ifdef EXPERIMENTAL_DMARC
1738/* initialize libopendmarc */
1739dmarc_up = dmarc_init();
1740#endif
1741
059ec3d9
PH
1742/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1743ids, and fractions of a second are required. See the comments that precede the
1744message id creation below. */
1745
1746(void)gettimeofday(&message_id_tv, NULL);
1747
1748/* For other uses of the received time we can operate with granularity of one
1749second, and for that we use the global variable received_time. This is for
32dfdf8b 1750things like ultimate message timeouts.XXX */
059ec3d9 1751
32dfdf8b 1752received_time = message_id_tv;
059ec3d9
PH
1753
1754/* If SMTP input, set the special handler for timeouts. The alarm() calls
1755happen in the smtp_getc() function when it refills its buffer. */
1756
1757if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1758
1759/* If not SMTP input, timeout happens only if configured, and we just set a
1760single timeout for the whole message. */
1761
1762else if (receive_timeout > 0)
1763 {
1764 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1765 alarm(receive_timeout);
1766 }
1767
1768/* SIGTERM and SIGINT are caught always. */
1769
1770signal(SIGTERM, data_sigterm_sigint_handler);
1771signal(SIGINT, data_sigterm_sigint_handler);
1772
1773/* Header lines in messages are not supposed to be very long, though when
1774unfolded, to: and cc: headers can take up a lot of store. We must also cope
1775with the possibility of junk being thrown at us. Start by getting 256 bytes for
1776storing the header, and extend this as necessary using string_cat().
1777
1778To cope with total lunacies, impose an upper limit on the length of the header
1779section of the message, as otherwise the store will fill up. We must also cope
1780with the possibility of binary zeros in the data. Hence we cannot use fgets().
1781Folded header lines are joined into one string, leaving the '\n' characters
1782inside them, so that writing them out reproduces the input.
1783
1784Loop for each character of each header; the next structure for chaining the
1785header is set up already, with ptr the offset of the next character in
1786next->text. */
1787
1788for (;;)
1789 {
bd8fbe36 1790 int ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1791
1792 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1793 SMTP must have a correct "." terminator. */
1794
1795 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1796 {
1797 smtp_reply = handle_lost_connection(US" (header)");
1798 smtp_yield = FALSE;
1799 goto TIDYUP; /* Skip to end of function */
1800 }
1801
1802 /* See if we are at the current header's size limit - there must be at least
1803 four bytes left. This allows for the new character plus a zero, plus two for
1804 extra insertions when we are playing games with dots and carriage returns. If
1805 we are at the limit, extend the text buffer. This could have been done
1806 automatically using string_cat() but because this is a tightish loop storing
1807 only one character at a time, we choose to do it inline. Normally
1808 store_extend() will be able to extend the block; only at the end of a big
1809 store block will a copy be needed. To handle the case of very long headers
1810 (and sometimes lunatic messages can have ones that are 100s of K long) we
1811 call store_release() for strings that have been copied - if the string is at
1812 the start of a block (and therefore the only thing in it, because we aren't
1813 doing any other gets), the block gets freed. We can only do this because we
1814 know there are no other calls to store_get() going on. */
1815
1816 if (ptr >= header_size - 4)
1817 {
1818 int oldsize = header_size;
1819 /* header_size += 256; */
1820 header_size *= 2;
1821 if (!store_extend(next->text, oldsize, header_size))
1822 {
1823 uschar *newtext = store_get(header_size);
1824 memcpy(newtext, next->text, ptr);
1825 store_release(next->text);
1826 next->text = newtext;
1827 }
1828 }
1829
1830 /* Cope with receiving a binary zero. There is dispute about whether
1831 these should be allowed in RFC 822 messages. The middle view is that they
1832 should not be allowed in headers, at least. Exim takes this attitude at
1833 the moment. We can't just stomp on them here, because we don't know that
1834 this line is a header yet. Set a flag to cause scanning later. */
1835
1836 if (ch == 0) had_zero++;
1837
1838 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1839 those from data files use just LF. Treat LF in local SMTP input as a
1840 terminator too. Treat EOF as a line terminator always. */
1841
1842 if (ch == EOF) goto EOL;
1843
1844 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1845 other MTAs accept this. We are therefore forced into this "liberalisation"
1846 too, so we accept LF as a line terminator whatever the source of the message.
1847 However, if the first line of the message ended with a CRLF, we treat a bare
1848 LF specially by inserting a white space after it to ensure that the header
1849 line is not terminated. */
1850
1851 if (ch == '\n')
1852 {
1853 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1854 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1855 goto EOL;
1856 }
1857
1858 /* This is not the end of the line. If this is SMTP input and this is
1859 the first character in the line and it is a "." character, ignore it.
1860 This implements the dot-doubling rule, though header lines starting with
1861 dots aren't exactly common. They are legal in RFC 822, though. If the
1862 following is CRLF or LF, this is the line that that terminates the
1863 entire message. We set message_ended to indicate this has happened (to
1864 prevent further reading), and break out of the loop, having freed the
1865 empty header, and set next = NULL to indicate no data line. */
1866
1867 if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
1868 {
bd8fbe36 1869 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1870 if (ch == '\r')
1871 {
bd8fbe36 1872 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1873 if (ch != '\n')
1874 {
80a47a2c 1875 receive_ungetc(ch);
059ec3d9
PH
1876 ch = '\r'; /* Revert to CR */
1877 }
1878 }
1879 if (ch == '\n')
1880 {
1881 message_ended = END_DOT;
1882 store_reset(next);
1883 next = NULL;
1884 break; /* End character-reading loop */
1885 }
1886
1887 /* For non-SMTP input, the dot at the start of the line was really a data
1888 character. What is now in ch is the following character. We guaranteed
1889 enough space for this above. */
1890
1891 if (!smtp_input)
1892 {
1893 next->text[ptr++] = '.';
1894 message_size++;
1895 }
1896 }
1897
1898 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1899 remember this case if this is the first line ending. */
1900
1901 if (ch == '\r')
1902 {
bd8fbe36 1903 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1904 if (ch == '\n')
1905 {
1906 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1907 goto EOL;
1908 }
1909
1910 /* Otherwise, put back the character after CR, and turn the bare CR
1911 into LF SP. */
1912
80a47a2c 1913 ch = (receive_ungetc)(ch);
059ec3d9
PH
1914 next->text[ptr++] = '\n';
1915 message_size++;
1916 ch = ' ';
1917 }
1918
1919 /* We have a data character for the header line. */
1920
1921 next->text[ptr++] = ch; /* Add to buffer */
1922 message_size++; /* Total message size so far */
1923
1924 /* Handle failure due to a humungously long header section. The >= allows
1925 for the terminating \n. Add what we have so far onto the headers list so
1926 that it gets reflected in any error message, and back up the just-read
1927 character. */
1928
1929 if (message_size >= header_maxsize)
1930 {
1931 next->text[ptr] = 0;
1932 next->slen = ptr;
1933 next->type = htype_other;
1934 next->next = NULL;
1935 header_last->next = next;
1936 header_last = next;
1937
1938 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1939 "%s (more than %d characters): message abandoned",
1940 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1941
1942 if (smtp_input)
1943 {
1944 smtp_reply = US"552 Message header is ridiculously long";
1945 receive_swallow_smtp();
1946 goto TIDYUP; /* Skip to end of function */
1947 }
1948
1949 else
1950 {
1951 give_local_error(ERRMESS_VLONGHEADER,
1952 string_sprintf("message header longer than %d characters received: "
1953 "message not accepted", header_maxsize), US"", error_rc, stdin,
1954 header_list->next);
1955 /* Does not return */
1956 }
1957 }
1958
1959 continue; /* With next input character */
1960
1961 /* End of header line reached */
1962
1963 EOL:
2e0c1448
PH
1964
1965 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1966
1967 receive_linecount++;
1968 message_linecount++;
059ec3d9 1969
d677b2f2
PH
1970 /* Keep track of maximum line length */
1971
1972 if (ptr - prevlines_length > max_received_linelength)
1973 max_received_linelength = ptr - prevlines_length;
1974 prevlines_length = ptr + 1;
1975
059ec3d9
PH
1976 /* Now put in the terminating newline. There is always space for
1977 at least two more characters. */
1978
1979 next->text[ptr++] = '\n';
1980 message_size++;
1981
1982 /* A blank line signals the end of the headers; release the unwanted
1983 space and set next to NULL to indicate this. */
1984
1985 if (ptr == 1)
1986 {
1987 store_reset(next);
1988 next = NULL;
1989 break;
1990 }
1991
1992 /* There is data in the line; see if the next input character is a
1993 whitespace character. If it is, we have a continuation of this header line.
1994 There is always space for at least one character at this point. */
1995
1996 if (ch != EOF)
1997 {
bd8fbe36 1998 int nextch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1999 if (nextch == ' ' || nextch == '\t')
2000 {
2001 next->text[ptr++] = nextch;
2002 message_size++;
2003 continue; /* Iterate the loop */
2004 }
80a47a2c 2005 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
2006 else ch = EOF; /* Cause main loop to exit at end */
2007 }
2008
2009 /* We have got to the real line end. Terminate the string and release store
2010 beyond it. If it turns out to be a real header, internal binary zeros will
2011 be squashed later. */
2012
2013 next->text[ptr] = 0;
2014 next->slen = ptr;
2015 store_reset(next->text + ptr + 1);
2016
2017 /* Check the running total size against the overall message size limit. We
2018 don't expect to fail here, but if the overall limit is set less than MESSAGE_
2019 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
2020 headers - the code to read the body will then also hit the buffer. */
2021
2022 if (message_size > thismessage_size_limit) break;
2023
2024 /* A line that is not syntactically correct for a header also marks
2025 the end of the headers. In this case, we leave next containing the
2026 first data line. This might actually be several lines because of the
2027 continuation logic applied above, but that doesn't matter.
2028
2029 It turns out that smail, and presumably sendmail, accept leading lines
2030 of the form
2031
2032 From ph10 Fri Jan 5 12:35 GMT 1996
2033
2034 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
2035 find any documentation of this, but for compatibility it had better be
2036 accepted. Exim restricts it to the case of non-smtp messages, and
2037 treats it as an alternative to the -f command line option. Thus it is
2038 ignored except for trusted users or filter testing. Otherwise it is taken
2039 as the sender address, unless -f was used (sendmail compatibility).
2040
2041 It further turns out that some UUCPs generate the From_line in a different
2042 format, e.g.
2043
2044 From ph10 Fri, 7 Jan 97 14:00:00 GMT
2045
2046 The regex for matching these things is now capable of recognizing both
2047 formats (including 2- and 4-digit years in the latter). In fact, the regex
2048 is now configurable, as is the expansion string to fish out the sender.
2049
2050 Even further on it has been discovered that some broken clients send
2051 these lines in SMTP messages. There is now an option to ignore them from
2052 specified hosts or networks. Sigh. */
2053
2054 if (header_last == header_list &&
2055 (!smtp_input
2056 ||
2057 (sender_host_address != NULL &&
2058 verify_check_host(&ignore_fromline_hosts) == OK)
2059 ||
2060 (sender_host_address == NULL && ignore_fromline_local)
2061 ) &&
2062 regex_match_and_setup(regex_From, next->text, 0, -1))
2063 {
2064 if (!sender_address_forced)
2065 {
2066 uschar *uucp_sender = expand_string(uucp_from_sender);
2067 if (uucp_sender == NULL)
2068 {
2069 log_write(0, LOG_MAIN|LOG_PANIC,
2070 "expansion of \"%s\" failed after matching "
2071 "\"From \" line: %s", uucp_from_sender, expand_string_message);
2072 }
2073 else
2074 {
2075 int start, end, domain;
2076 uschar *errmess;
2077 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
2078 &start, &end, &domain, TRUE);
2079 if (newsender != NULL)
2080 {
2081 if (domain == 0 && newsender[0] != 0)
2082 newsender = rewrite_address_qualify(newsender, FALSE);
2083
f05da2e8 2084 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
2085 {
2086 sender_address = newsender;
2087
f05da2e8 2088 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
2089 {
2090 authenticated_sender = NULL;
2091 originator_name = US"";
2092 sender_local = FALSE;
2093 }
2094
f05da2e8 2095 if (filter_test != FTEST_NONE)
059ec3d9
PH
2096 printf("Sender taken from \"From \" line\n");
2097 }
2098 }
2099 }
2100 }
2101 }
2102
2103 /* Not a leading "From " line. Check to see if it is a valid header line.
2104 Header names may contain any non-control characters except space and colon,
2105 amazingly. */
2106
2107 else
2108 {
2109 uschar *p = next->text;
2110
2111 /* If not a valid header line, break from the header reading loop, leaving
2112 next != NULL, indicating that it holds the first line of the body. */
2113
2114 if (isspace(*p)) break;
2115 while (mac_isgraph(*p) && *p != ':') p++;
2116 while (isspace(*p)) p++;
2117 if (*p != ':')
2118 {
2119 body_zerocount = had_zero;
2120 break;
2121 }
2122
2123 /* We have a valid header line. If there were any binary zeroes in
2124 the line, stomp on them here. */
2125
2126 if (had_zero > 0)
2127 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
2128
2129 /* It is perfectly legal to have an empty continuation line
2130 at the end of a header, but it is confusing to humans
2131 looking at such messages, since it looks like a blank line.
2132 Reduce confusion by removing redundant white space at the
2133 end. We know that there is at least one printing character
2134 (the ':' tested for above) so there is no danger of running
2135 off the end. */
2136
2137 p = next->text + ptr - 2;
2138 for (;;)
2139 {
2140 while (*p == ' ' || *p == '\t') p--;
2141 if (*p != '\n') break;
2142 ptr = (p--) - next->text + 1;
2143 message_size -= next->slen - ptr;
2144 next->text[ptr] = 0;
2145 next->slen = ptr;
2146 }
2147
2148 /* Add the header to the chain */
2149
2150 next->type = htype_other;
2151 next->next = NULL;
2152 header_last->next = next;
2153 header_last = next;
2154
2155 /* Check the limit for individual line lengths. This comes after adding to
2156 the chain so that the failing line is reflected if a bounce is generated
2157 (for a local message). */
2158
2159 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
2160 {
2161 log_write(0, LOG_MAIN, "overlong message header line received from "
2162 "%s (more than %d characters): message abandoned",
2163 sender_host_unknown? sender_ident : sender_fullhost,
2164 header_line_maxsize);
2165
2166 if (smtp_input)
2167 {
2168 smtp_reply = US"552 A message header line is too long";
2169 receive_swallow_smtp();
2170 goto TIDYUP; /* Skip to end of function */
2171 }
2172
2173 else
2174 {
2175 give_local_error(ERRMESS_VLONGHDRLINE,
2176 string_sprintf("message header line longer than %d characters "
2177 "received: message not accepted", header_line_maxsize), US"",
2178 error_rc, stdin, header_list->next);
2179 /* Does not return */
2180 }
2181 }
2182
2183 /* Note if any resent- fields exist. */
2184
2185 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
2186 {
2187 resents_exist = TRUE;
2188 resent_prefix = US"Resent-";
2189 }
2190 }
2191
1ebe15c3
JH
2192 /* Reject CHUNKING messages that do not CRLF their first header line */
2193
2194 if (!first_line_ended_crlf && chunking_state > CHUNKING_OFFERED)
2195 {
2196 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2197 "Non-CRLF-terminated header, under CHUNKING: message abandoned",
2198 sender_address,
2199 sender_fullhost ? " H=" : "", sender_fullhost ? sender_fullhost : US"",
2200 sender_ident ? " U=" : "", sender_ident ? sender_ident : US"");
925ac8e4 2201 smtp_printf("552 Message header not CRLF terminated\r\n", FALSE);
1ebe15c3
JH
2202 bdat_flush_data();
2203 smtp_reply = US"";
2204 goto TIDYUP; /* Skip to end of function */
2205 }
2206
059ec3d9
PH
2207 /* The line has been handled. If we have hit EOF, break out of the loop,
2208 indicating no pending data line. */
2209
2210 if (ch == EOF) { next = NULL; break; }
2211
2212 /* Set up for the next header */
2213
2214 header_size = 256;
2215 next = store_get(sizeof(header_line));
2216 next->text = store_get(header_size);
2217 ptr = 0;
2218 had_zero = 0;
d677b2f2 2219 prevlines_length = 0;
059ec3d9
PH
2220 } /* Continue, starting to read the next header */
2221
2222/* At this point, we have read all the headers into a data structure in main
2223store. The first header is still the dummy placeholder for the Received: header
2224we are going to generate a bit later on. If next != NULL, it contains the first
2225data line - which terminated the headers before reaching a blank line (not the
2226normal case). */
2227
2228DEBUG(D_receive)
2229 {
2230 debug_printf(">>Headers received:\n");
1ebe15c3 2231 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2232 debug_printf("%s", h->text);
2233 debug_printf("\n");
2234 }
2235
2236/* End of file on any SMTP connection is an error. If an incoming SMTP call
2237is dropped immediately after valid headers, the next thing we will see is EOF.
2238We must test for this specially, as further down the reading of the data is
2239skipped if already at EOF. */
2240
2241if (smtp_input && (receive_feof)())
2242 {
2243 smtp_reply = handle_lost_connection(US" (after header)");
2244 smtp_yield = FALSE;
2245 goto TIDYUP; /* Skip to end of function */
2246 }
2247
2248/* If this is a filter test run and no headers were read, output a warning
2249in case there is a mistake in the test message. */
2250
f05da2e8 2251if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2252 printf("Warning: no message headers read\n");
2253
2254
2255/* Scan the headers to identify them. Some are merely marked for later
2256processing; some are dealt with here. */
2257
1ebe15c3 2258for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2259 {
2260 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2261 if (is_resent) contains_resent_headers = TRUE;
2262
2263 switch (header_checkname(h, is_resent))
2264 {
059ec3d9 2265 case htype_bcc:
2cbb4081 2266 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
2267 break;
2268
059ec3d9 2269 case htype_cc:
2cbb4081 2270 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
2271 break;
2272
2273 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
2274
2275 case htype_date:
4c69d561 2276 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
2277 break;
2278
2279 /* Same comments as about Return-Path: below. */
2280
2281 case htype_delivery_date:
2282 if (delivery_date_remove) h->type = htype_old;
2283 break;
2284
2285 /* Same comments as about Return-Path: below. */
2286
2287 case htype_envelope_to:
2288 if (envelope_to_remove) h->type = htype_old;
2289 break;
2290
2291 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2292 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2293 header consists of just the login id of the user who called Exim, rewrite
2294 it with the gecos field first. Apply this rule to Resent-From: if there
2295 are resent- fields. */
2296
2297 case htype_from:
2298 h->type = htype_from;
2299 if (!resents_exist || is_resent)
2300 {
2301 from_header = h;
2302 if (!smtp_input)
2303 {
5de8faa3 2304 int len;
059ec3d9
PH
2305 uschar *s = Ustrchr(h->text, ':') + 1;
2306 while (isspace(*s)) s++;
5de8faa3 2307 len = h->slen - (s - h->text) - 1;
e0fccd1d
TF
2308 if (Ustrlen(originator_login) == len &&
2309 strncmpic(s, originator_login, len) == 0)
059ec3d9
PH
2310 {
2311 uschar *name = is_resent? US"Resent-From" : US"From";
2312 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2313 originator_login, qualify_domain_sender);
2314 from_header = header_last;
2315 h->type = htype_old;
2316 DEBUG(D_receive|D_rewrite)
2317 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2318 }
2319 }
2320 }
2321 break;
2322
2323 /* Identify the Message-id: header for generating "in-reply-to" in the
2324 autoreply transport. For incoming logging, save any resent- value. In both
2325 cases, take just the first of any multiples. */
2326
2327 case htype_id:
2328 if (msgid_header == NULL && (!resents_exist || is_resent))
2329 {
2330 msgid_header = h;
2331 h->type = htype_id;
2332 }
2333 break;
2334
2335 /* Flag all Received: headers */
2336
2337 case htype_received:
2338 h->type = htype_received;
2339 received_count++;
2340 break;
2341
2342 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
2343
2344 case htype_reply_to:
2345 h->type = htype_reply_to;
2346 break;
2347
2348 /* The Return-path: header is supposed to be added to messages when
2349 they leave the SMTP system. We shouldn't receive messages that already
2350 contain Return-path. However, since Exim generates Return-path: on
2351 local delivery, resent messages may well contain it. We therefore
2352 provide an option (which defaults on) to remove any Return-path: headers
2353 on input. Removal actually means flagging as "old", which prevents the
2354 header being transmitted with the message. */
2355
2356 case htype_return_path:
2357 if (return_path_remove) h->type = htype_old;
2358
2359 /* If we are testing a mail filter file, use the value of the
2360 Return-Path: header to set up the return_path variable, which is not
2361 otherwise set. However, remove any <> that surround the address
2362 because the variable doesn't have these. */
2363
f05da2e8 2364 if (filter_test != FTEST_NONE)
059ec3d9
PH
2365 {
2366 uschar *start = h->text + 12;
2367 uschar *end = start + Ustrlen(start);
2368 while (isspace(*start)) start++;
2369 while (end > start && isspace(end[-1])) end--;
2370 if (*start == '<' && end[-1] == '>')
2371 {
2372 start++;
2373 end--;
2374 }
2375 return_path = string_copyn(start, end - start);
2376 printf("Return-path taken from \"Return-path:\" header line\n");
2377 }
2378 break;
2379
2380 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2381 and from an untrusted caller and suppress_local_fixups is not set, or if we
2382 are in submission mode for a remote message, mark it "old" so that it will
2383 not be transmitted with the message, unless active_local_sender_retain is
2384 set. (This can only be true if active_local_from_check is false.) If there
2385 are any resent- headers in the message, apply this rule to Resent-Sender:
2386 instead of Sender:. Messages with multiple resent- header sets cannot be
2387 tidily handled. (For this reason, at least one MUA - Pine - turns old
2388 resent- headers into X-resent- headers when resending, leaving just one
2389 set.) */
059ec3d9
PH
2390
2391 case htype_sender:
69358f02 2392 h->type = ((!active_local_sender_retain &&
8800895a
PH
2393 (
2394 (sender_local && !trusted_caller && !suppress_local_fixups)
2395 || submission_mode
2396 )
059ec3d9
PH
2397 ) &&
2398 (!resents_exist||is_resent))?
2399 htype_old : htype_sender;
2400 break;
2401
2402 /* Remember the Subject: header for logging. There is no Resent-Subject */
2403
2404 case htype_subject:
2405 subject_header = h;
2406 break;
2407
2408 /* "To:" gets flagged, and the existence of a recipient header is noted,
2409 whether it's resent- or not. */
2410
2411 case htype_to:
2412 h->type = htype_to;
2413 /****
2414 to_or_cc_header_exists = TRUE;
2415 ****/
2416 break;
2417 }
2418 }
2419
2420/* Extract recipients from the headers if that is required (the -t option).
2421Note that this is documented as being done *before* any address rewriting takes
2422place. There are two possibilities:
2423
2424(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2425recipients already listed are to be REMOVED from the message. Smail 3 works
2426like this. We need to build a non-recipients tree for that list, because in
2427subsequent processing this data is held in a tree and that's what the
2428spool_write_header() function expects. Make sure that non-recipient addresses
2429are fully qualified and rewritten if necessary.
2430
2431(2) According to other sendmail documentation, -t ADDS extracted recipients to
2432those in the command line arguments (and it is rumoured some other MTAs do
2433this). Therefore, there is an option to make Exim behave this way.
2434
2435*** Notes on "Resent-" header lines ***
2436
2437The presence of resent-headers in the message makes -t horribly ambiguous.
2438Experiments with sendmail showed that it uses recipients for all resent-
2439headers, totally ignoring the concept of "sets of resent- headers" as described
2440in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2441with all the addresses in one instance of each header.
2442
2443This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2444error for -t if there were resent- headers in the message. However, after a
2445discussion on the mailing list, I've learned that there are MUAs that use
2446resent- headers with -t, and also that the stuff about sets of resent- headers
2447and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2448message with -t and resent- header lines makes sure that only *its* resent-
2449headers are present; previous ones are often renamed as X-resent- for example.
2450
2451Consequently, Exim has been changed so that, if any resent- header lines are
2452present, the recipients are taken from all of the appropriate resent- lines,
2453and not from the ordinary To:, Cc:, etc. */
2454
2455if (extract_recip)
2456 {
2457 int rcount = 0;
2458 error_block **bnext = &bad_addresses;
2459
2460 if (extract_addresses_remove_arguments)
2461 {
2462 while (recipients_count-- > 0)
2463 {
2464 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2465 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2466 tree_add_nonrecipient(s);
2467 }
2468 recipients_list = NULL;
2469 recipients_count = recipients_list_max = 0;
2470 }
2471
059ec3d9
PH
2472 /* Now scan the headers */
2473
1ebe15c3 2474 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2475 {
2476 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2477 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2478 {
2479 uschar *s = Ustrchr(h->text, ':') + 1;
2480 while (isspace(*s)) s++;
2481
1eccaa59
PH
2482 parse_allow_group = TRUE; /* Allow address group syntax */
2483
059ec3d9
PH
2484 while (*s != 0)
2485 {
2486 uschar *ss = parse_find_address_end(s, FALSE);
2487 uschar *recipient, *errmess, *p, *pp;
2488 int start, end, domain;
2489
2490 /* Check on maximum */
2491
2492 if (recipients_max > 0 && ++rcount > recipients_max)
2493 {
2494 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2495 US"message rejected: ", error_rc, stdin, NULL);
2496 /* Does not return */
2497 }
2498
2499 /* Make a copy of the address, and remove any internal newlines. These
2500 may be present as a result of continuations of the header line. The
2501 white space that follows the newline must not be removed - it is part
2502 of the header. */
2503
2504 pp = recipient = store_get(ss - s + 1);
2505 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2506 *pp = 0;
250b6871 2507
8c5d388a 2508#ifdef SUPPORT_I18N
250b6871
JH
2509 {
2510 BOOL b = allow_utf8_domains;
2511 allow_utf8_domains = TRUE;
2512#endif
059ec3d9
PH
2513 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2514 &domain, FALSE);
2515
8c5d388a 2516#ifdef SUPPORT_I18N
250b6871
JH
2517 if (string_is_utf8(recipient))
2518 message_smtputf8 = TRUE;
2519 else
2520 allow_utf8_domains = b;
2521 }
2522#endif
2523
059ec3d9
PH
2524 /* Keep a list of all the bad addresses so we can send a single
2525 error message at the end. However, an empty address is not an error;
2526 just ignore it. This can come from an empty group list like
2527
2528 To: Recipients of list:;
2529
2530 If there are no recipients at all, an error will occur later. */
2531
2532 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2533 {
2534 int len = Ustrlen(s);
2535 error_block *b = store_get(sizeof(error_block));
2536 while (len > 0 && isspace(s[len-1])) len--;
2537 b->next = NULL;
2538 b->text1 = string_printing(string_copyn(s, len));
2539 b->text2 = errmess;
2540 *bnext = b;
2541 bnext = &(b->next);
2542 }
2543
2544 /* If the recipient is already in the nonrecipients tree, it must
2545 have appeared on the command line with the option extract_addresses_
2546 remove_arguments set. Do not add it to the recipients, and keep a note
2547 that this has happened, in order to give a better error if there are
2548 no recipients left. */
2549
2550 else if (recipient != NULL)
2551 {
2552 if (tree_search(tree_nonrecipients, recipient) == NULL)
2553 receive_add_recipient(recipient, -1);
2554 else
2555 extracted_ignored = TRUE;
2556 }
2557
2558 /* Move on past this address */
2559
2560 s = ss + (*ss? 1:0);
2561 while (isspace(*s)) s++;
1eccaa59
PH
2562 } /* Next address */
2563
2564 parse_allow_group = FALSE; /* Reset group syntax flags */
2565 parse_found_group = FALSE;
059ec3d9
PH
2566
2567 /* If this was the bcc: header, mark it "old", which means it
2568 will be kept on the spool, but not transmitted as part of the
2569 message. */
2570
2cbb4081 2571 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2572 } /* For appropriate header line */
2573 } /* For each header line */
2574
059ec3d9
PH
2575 }
2576
2577/* Now build the unique message id. This has changed several times over the
2578lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2579Retaining all the history in the comment has become too unwieldy - read
2580previous release sources if you want it.
2581
2582The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2583The first part is the current time, in seconds. The second part is the current
2584pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2585can hold a number in the range 0-3843. It used to be a computed sequence
2586number, but is now the fractional component of the current time in units of
25871/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2588received, Exim ensures that the timer has ticked at the appropriate level
2589before proceeding, to avoid duplication if the pid happened to be re-used
2590within the same time period. It seems likely that most messages will take at
2591least half a millisecond to be received, so no delay will normally be
2592necessary. At least for some time...
2593
2594There is a modification when localhost_number is set. Formerly this was allowed
2595to be as large as 255. Now it is restricted to the range 0-16, and the final
2596component of the message id becomes (localhost_number * 200) + fractional time
2597in units of 1/200 of a second (i.e. a value in the range 0-3399).
2598
2599Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2600Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2601still allows the tttttt field to hold a large enough number to last for some
2602more decades, and the final two-digit field can hold numbers up to 1295, which
2603is enough for milliseconds (instead of 1/2000 of a second).
2604
2605However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2606pid, so it is probably safe because pids have to be positive. The
2607localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2608final field becomes (localhost_number * 100) + fractional time in centiseconds.
2609
2610Note that string_base62() returns its data in a static storage block, so it
2611must be copied before calling string_base62() again. It always returns exactly
26126 characters.
2613
2614There doesn't seem to be anything in the RFC which requires a message id to
2615start with a letter, but Smail was changed to ensure this. The external form of
2616the message id (as supplied by string expansion) therefore starts with an
2617additional leading 'E'. The spool file names do not include this leading
2618letter and it is not used internally.
2619
2620NOTE: If ever the format of message ids is changed, the regular expression for
2621checking that a string is in this format must be updated in a corresponding
2622way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2540f2f8
JH
2623must also be changed to reflect the correct string length. The queue-sort code
2624needs to know the layout. Then, of course, other programs that rely on the
2625message id format will need updating too. */
059ec3d9
PH
2626
2627Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2628message_id[6] = '-';
2629Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2630
2631/* Deal with the case where the host number is set. The value of the number was
2632checked when it was read, to ensure it isn't too big. The timing granularity is
2633left in id_resolution so that an appropriate wait can be done after receiving
2634the message, if necessary (we hope it won't be). */
2635
2636if (host_number_string != NULL)
2637 {
2638 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2639 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2640 string_base62((long int)(
2641 host_number * (1000000/id_resolution) +
2642 message_id_tv.tv_usec/id_resolution)) + 4);
2643 }
2644
2645/* Host number not set: final field is just the fractional time at an
2646appropriate resolution. */
2647
2648else
2649 {
2650 id_resolution = (BASE_62 == 62)? 500 : 1000;
2651 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2652 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2653 }
2654
2655/* Add the current message id onto the current process info string if
2656it will fit. */
2657
2658(void)string_format(process_info + process_info_len,
2659 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2660
2661/* If we are using multiple input directories, set up the one for this message
2662to be the least significant base-62 digit of the time of arrival. Otherwise
2663ensure that it is an empty string. */
2664
a2da3176 2665message_subdir[0] = split_spool_directory ? message_id[5] : 0;
059ec3d9
PH
2666
2667/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2668one, but only for local (without suppress_local_fixups) or submission mode
2669messages. This can be user-configured if required, but we had better flatten
2670any illegal characters therein. */
059ec3d9 2671
8800895a
PH
2672if (msgid_header == NULL &&
2673 ((sender_host_address == NULL && !suppress_local_fixups)
2674 || submission_mode))
059ec3d9
PH
2675 {
2676 uschar *p;
2677 uschar *id_text = US"";
2678 uschar *id_domain = primary_hostname;
2679
2680 /* Permit only letters, digits, dots, and hyphens in the domain */
2681
2682 if (message_id_domain != NULL)
2683 {
2684 uschar *new_id_domain = expand_string(message_id_domain);
2685 if (new_id_domain == NULL)
2686 {
2687 if (!expand_string_forcedfail)
2688 log_write(0, LOG_MAIN|LOG_PANIC,
2689 "expansion of \"%s\" (message_id_header_domain) "
2690 "failed: %s", message_id_domain, expand_string_message);
2691 }
2692 else if (*new_id_domain != 0)
2693 {
2694 id_domain = new_id_domain;
2695 for (p = id_domain; *p != 0; p++)
2696 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2697 }
2698 }
2699
2700 /* Permit all characters except controls and RFC 2822 specials in the
2701 additional text part. */
2702
2703 if (message_id_text != NULL)
2704 {
2705 uschar *new_id_text = expand_string(message_id_text);
2706 if (new_id_text == NULL)
2707 {
2708 if (!expand_string_forcedfail)
2709 log_write(0, LOG_MAIN|LOG_PANIC,
2710 "expansion of \"%s\" (message_id_header_text) "
2711 "failed: %s", message_id_text, expand_string_message);
2712 }
2713 else if (*new_id_text != 0)
2714 {
2715 id_text = new_id_text;
2716 for (p = id_text; *p != 0; p++)
2717 if (mac_iscntrl_or_special(*p)) *p = '-';
2718 }
2719 }
2720
e7e680d6
PP
2721 /* Add the header line
2722 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2723 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2724
e7e680d6 2725 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2726 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2727 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2728 }
2729
2730/* If we are to log recipients, keep a copy of the raw ones before any possible
2731rewriting. Must copy the count, because later ACLs and the local_scan()
2732function may mess with the real recipients. */
2733
6c6d6e48 2734if (LOGGING(received_recipients))
059ec3d9
PH
2735 {
2736 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2737 for (i = 0; i < recipients_count; i++)
2738 raw_recipients[i] = string_copy(recipients_list[i].address);
2739 raw_recipients_count = recipients_count;
2740 }
2741
2742/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2743recipients will get here only if the conditions were right (allow_unqualified_
2744recipient is TRUE). */
2745
2746for (i = 0; i < recipients_count; i++)
2747 recipients_list[i].address =
2748 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2749 global_rewrite_rules, rewrite_existflags);
2750
8800895a
PH
2751/* If there is no From: header, generate one for local (without
2752suppress_local_fixups) or submission_mode messages. If there is no sender
2753address, but the sender is local or this is a local delivery error, use the
2754originator login. This shouldn't happen for genuine bounces, but might happen
2755for autoreplies. The addition of From: must be done *before* checking for the
2756possible addition of a Sender: header, because untrusted_set_sender allows an
2757untrusted user to set anything in the envelope (which might then get info
2758From:) but we still want to ensure a valid Sender: if it is required. */
2759
2760if (from_header == NULL &&
2761 ((sender_host_address == NULL && !suppress_local_fixups)
2762 || submission_mode))
059ec3d9 2763 {
2fe1a124
PH
2764 uschar *oname = US"";
2765
2766 /* Use the originator_name if this is a locally submitted message and the
2767 caller is not trusted. For trusted callers, use it only if -F was used to
2768 force its value or if we have a non-SMTP message for which -f was not used
2769 to set the sender. */
2770
2771 if (sender_host_address == NULL)
2772 {
2773 if (!trusted_caller || sender_name_forced ||
2774 (!smtp_input && !sender_address_forced))
2775 oname = originator_name;
2776 }
2777
2778 /* For non-locally submitted messages, the only time we use the originator
2779 name is when it was forced by the /name= option on control=submission. */
2780
2781 else
2782 {
2783 if (submission_name != NULL) oname = submission_name;
2784 }
2785
059ec3d9
PH
2786 /* Envelope sender is empty */
2787
2788 if (sender_address[0] == 0)
2789 {
87ba3f5f
PH
2790 uschar *fromstart, *fromend;
2791
2792 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2793 oname, (oname[0] == 0)? "" : " <");
2794 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2795
059ec3d9
PH
2796 if (sender_local || local_error_message)
2797 {
87ba3f5f
PH
2798 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2799 local_part_quote(originator_login), qualify_domain_sender,
2800 fromend);
059ec3d9
PH
2801 }
2802 else if (submission_mode && authenticated_id != NULL)
2803 {
2804 if (submission_domain == NULL)
2805 {
87ba3f5f
PH
2806 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2807 local_part_quote(authenticated_id), qualify_domain_sender,
2808 fromend);
059ec3d9
PH
2809 }
2810 else if (submission_domain[0] == 0) /* empty => whole address set */
2811 {
87ba3f5f
PH
2812 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2813 fromend);
059ec3d9
PH
2814 }
2815 else
2816 {
87ba3f5f
PH
2817 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2818 local_part_quote(authenticated_id), submission_domain,
2819 fromend);
059ec3d9
PH
2820 }
2821 from_header = header_last; /* To get it checked for Sender: */
2822 }
2823 }
2824
2825 /* There is a non-null envelope sender. Build the header using the original
2826 sender address, before any rewriting that might have been done while
2827 verifying it. */
2828
2829 else
2830 {
87ba3f5f 2831 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2832 oname,
2833 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2834 (sender_address_unrewritten == NULL)?
2835 sender_address : sender_address_unrewritten,
2fe1a124 2836 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2837
2838 from_header = header_last; /* To get it checked for Sender: */
2839 }
2840 }
2841
2842
8800895a
PH
2843/* If the sender is local (without suppress_local_fixups), or if we are in
2844submission mode and there is an authenticated_id, check that an existing From:
2845is correct, and if not, generate a Sender: header, unless disabled. Any
2846previously-existing Sender: header was removed above. Note that sender_local,
2847as well as being TRUE if the caller of exim is not trusted, is also true if a
2848trusted caller did not supply a -f argument for non-smtp input. To allow
2849trusted callers to forge From: without supplying -f, we have to test explicitly
2850here. If the From: header contains more than one address, then the call to
2851parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2852
2853if (from_header != NULL &&
69358f02 2854 (active_local_from_check &&
8800895a 2855 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2856 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2857 ))
2858 {
2859 BOOL make_sender = TRUE;
2860 int start, end, domain;
2861 uschar *errmess;
2862 uschar *from_address =
2863 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2864 &start, &end, &domain, FALSE);
2865 uschar *generated_sender_address;
2866
2867 if (submission_mode)
2868 {
2869 if (submission_domain == NULL)
2870 {
2871 generated_sender_address = string_sprintf("%s@%s",
2872 local_part_quote(authenticated_id), qualify_domain_sender);
2873 }
2874 else if (submission_domain[0] == 0) /* empty => full address */
2875 {
2876 generated_sender_address = string_sprintf("%s",
2877 authenticated_id);
2878 }
2879 else
2880 {
2881 generated_sender_address = string_sprintf("%s@%s",
2882 local_part_quote(authenticated_id), submission_domain);
2883 }
2884 }
2885 else
2886 generated_sender_address = string_sprintf("%s@%s",
2887 local_part_quote(originator_login), qualify_domain_sender);
2888
2889 /* Remove permitted prefixes and suffixes from the local part of the From:
2890 address before doing the comparison with the generated sender. */
2891
2892 if (from_address != NULL)
2893 {
2894 int slen;
2895 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2896
2897 if (at != NULL) *at = 0;
2898 from_address += route_check_prefix(from_address, local_from_prefix);
2899 slen = route_check_suffix(from_address, local_from_suffix);
2900 if (slen > 0)
2901 {
2902 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2903 from_address += slen;
2904 }
2905 if (at != NULL) *at = '@';
2906
2907 if (strcmpic(generated_sender_address, from_address) == 0 ||
2908 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2909 make_sender = FALSE;
2910 }
2911
2912 /* We have to cause the Sender header to be rewritten if there are
2913 appropriate rewriting rules. */
2914
2915 if (make_sender)
2916 {
2fe1a124 2917 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2918 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2919 generated_sender_address);
2920 else
2921 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2922 resent_prefix,
2923 submission_mode? submission_name : originator_name,
2924 generated_sender_address);
059ec3d9 2925 }
87ba3f5f
PH
2926
2927 /* Ensure that a non-null envelope sender address corresponds to the
2928 submission mode sender address. */
2929
2930 if (submission_mode && sender_address[0] != 0)
2931 {
2932 if (sender_address_unrewritten == NULL)
2933 sender_address_unrewritten = sender_address;
2934 sender_address = generated_sender_address;
089793a4
TF
2935 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2936 log_write(L_address_rewrite, LOG_MAIN,
2937 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2938 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2939 }
059ec3d9
PH
2940 }
2941
059ec3d9
PH
2942/* If there are any rewriting rules, apply them to the sender address, unless
2943it has already been rewritten as part of verification for SMTP input. */
2944
2945if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2946 sender_address[0] != 0)
2947 {
2948 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2949 global_rewrite_rules, rewrite_existflags);
2950 DEBUG(D_receive|D_rewrite)
2951 debug_printf("rewritten sender = %s\n", sender_address);
2952 }
2953
2954
2955/* The headers must be run through rewrite_header(), because it ensures that
2956addresses are fully qualified, as well as applying any rewriting rules that may
2957exist.
2958
2959Qualification of header addresses in a message from a remote host happens only
2960if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2961appropriate. For local messages, qualification always happens, unless -bnq is
2962used to explicitly suppress it. No rewriting is done for an unqualified address
2963that is left untouched.
2964
2965We start at the second header, skipping our own Received:. This rewriting is
2966documented as happening *after* recipient addresses are taken from the headers
2967by the -t command line option. An added Sender: gets rewritten here. */
2968
1ebe15c3 2969for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2970 {
2971 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2972 rewrite_existflags, TRUE);
1ebe15c3 2973 if (newh) h = newh;
059ec3d9
PH
2974 }
2975
2976
2977/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2978"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2979"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2980exists, this applies to the set of resent- headers rather than the normal set.
2981
2cbb4081
PH
2982The requirement for a recipient header has been removed in RFC 2822. At this
2983point in the code, earlier versions of Exim added a To: header for locally
2984submitted messages, and an empty Bcc: header for others. In the light of the
2985changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2986
059ec3d9
PH
2987
2988/* If there is no date header, generate one if the message originates locally
8800895a
PH
2989(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2990submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2991to be more confusing if Exim adds one to all remotely-originated messages.
2992As per Message-Id, we prepend if resending, else append.
2993*/
059ec3d9 2994
8800895a
PH
2995if (!date_header_exists &&
2996 ((sender_host_address == NULL && !suppress_local_fixups)
2997 || submission_mode))
e7e680d6
PP
2998 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2999 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
3000
3001search_tidyup(); /* Free any cached resources */
3002
3003/* Show the complete set of headers if debugging. Note that the first one (the
3004new Received:) has not yet been set. */
3005
3006DEBUG(D_receive)
3007 {
3008 debug_printf(">>Headers after rewriting and local additions:\n");
3009 for (h = header_list->next; h != NULL; h = h->next)
3010 debug_printf("%c %s", h->type, h->text);
3011 debug_printf("\n");
3012 }
3013
3014/* The headers are now complete in store. If we are running in filter
3015testing mode, that is all this function does. Return TRUE if the message
3016ended with a dot. */
3017
f05da2e8 3018if (filter_test != FTEST_NONE)
059ec3d9
PH
3019 {
3020 process_info[process_info_len] = 0;
3021 return message_ended == END_DOT;
3022 }
3023
7e3ce68e
JH
3024/*XXX CHUNKING: need to cancel cutthrough under BDAT, for now. In future,
3025think more if it could be handled. Cannot do onward CHUNKING unless
3026inbound is, but inbound chunking ought to be ok with outbound plain.
3027Could we do onward CHUNKING given inbound CHUNKING?
3028*/
3029if (chunking_state > CHUNKING_OFFERED)
57cc2785 3030 cancel_cutthrough_connection(FALSE, US"chunking active");
7e3ce68e 3031
817d9f57 3032/* Cutthrough delivery:
5032d1cf
JH
3033We have to create the Received header now rather than at the end of reception,
3034so the timestamp behaviour is a change to the normal case.
5032d1cf 3035Having created it, send the headers to the destination. */
57cc2785
JH
3036
3037if (cutthrough.fd >= 0 && cutthrough.delivery)
e4bdf652 3038 {
817d9f57
JH
3039 if (received_count > received_headers_max)
3040 {
57cc2785 3041 cancel_cutthrough_connection(TRUE, US"too many headers");
817d9f57
JH
3042 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
3043 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3044 "Too many \"Received\" headers",
3045 sender_address,
57cc2785
JH
3046 sender_fullhost ? "H=" : "", sender_fullhost ? sender_fullhost : US"",
3047 sender_ident ? "U=" : "", sender_ident ? sender_ident : US"");
817d9f57
JH
3048 message_id[0] = 0; /* Indicate no message accepted */
3049 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
3050 goto TIDYUP; /* Skip to end of function */
3051 }
e4bdf652 3052 received_header_gen();
578d43dc 3053 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652
JH
3054 (void) cutthrough_headers_send();
3055 }
61147df4 3056
e4bdf652 3057
059ec3d9
PH
3058/* Open a new spool file for the data portion of the message. We need
3059to access it both via a file descriptor and a stream. Try to make the
41313d92 3060directory if it isn't there. */
059ec3d9 3061
41313d92 3062spool_name = spool_fname(US"input", message_subdir, message_id, US"-D");
a2da3176
JH
3063DEBUG(D_receive) debug_printf("Data file name: %s\n", spool_name);
3064
3065if ((data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE)) < 0)
059ec3d9
PH
3066 {
3067 if (errno == ENOENT)
3068 {
0971ec06 3069 (void) directory_make(spool_directory,
41313d92
JH
3070 spool_sname(US"input", message_subdir),
3071 INPUT_DIRECTORY_MODE, TRUE);
059ec3d9
PH
3072 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
3073 }
3074 if (data_fd < 0)
3075 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
3076 spool_name, strerror(errno));
3077 }
3078
3079/* Make sure the file's group is the Exim gid, and double-check the mode
3080because the group setting doesn't always get set automatically. */
3081
1ac6b2e7
JH
3082if (fchown(data_fd, exim_uid, exim_gid))
3083 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
3084 "Failed setting ownership on spool file %s: %s",
3085 spool_name, strerror(errno));
ff790e47 3086(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
3087
3088/* We now have data file open. Build a stream for it and lock it. We lock only
3089the first line of the file (containing the message ID) because otherwise there
3090are problems when Exim is run under Cygwin (I'm told). See comments in
3091spool_in.c, where the same locking is done. */
3092
3093data_file = fdopen(data_fd, "w+");
3094lock_data.l_type = F_WRLCK;
3095lock_data.l_whence = SEEK_SET;
3096lock_data.l_start = 0;
3097lock_data.l_len = SPOOL_DATA_START_OFFSET;
3098
3099if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
3100 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
3101 errno, strerror(errno));
3102
3103/* We have an open, locked data file. Write the message id to it to make it
3104self-identifying. Then read the remainder of the input of this message and
3105write it to the data file. If the variable next != NULL, it contains the first
3106data line (which was read as a header but then turned out not to have the right
3107format); write it (remembering that it might contain binary zeros). The result
3108of fwrite() isn't inspected; instead we call ferror() below. */
3109
3110fprintf(data_file, "%s-D\n", message_id);
3111if (next != NULL)
3112 {
3113 uschar *s = next->text;
3114 int len = next->slen;
1ac6b2e7 3115 len = fwrite(s, 1, len, data_file); len = len; /* compiler quietening */
059ec3d9
PH
3116 body_linecount++; /* Assumes only 1 line */
3117 }
3118
3119/* Note that we might already be at end of file, or the logical end of file
3120(indicated by '.'), or might have encountered an error while writing the
3121message id or "next" line. */
3122
3123if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
3124 {
3125 if (smtp_input)
3126 {
328c5688
JH
3127 message_ended = chunking_state <= CHUNKING_OFFERED
3128 ? read_message_data_smtp(data_file)
3129 : spool_wireformat
3130 ? read_message_bdat_smtp_wire(data_file)
3131 : read_message_bdat_smtp(data_file);
059ec3d9
PH
3132 receive_linecount++; /* The terminating "." line */
3133 }
3134 else message_ended = read_message_data(data_file);
3135
3136 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 3137 message_linecount += body_linecount;
059ec3d9 3138
7e3ce68e 3139 switch (message_ended)
059ec3d9 3140 {
7e3ce68e 3141 /* Handle premature termination of SMTP */
059ec3d9 3142
7e3ce68e
JH
3143 case END_EOF:
3144 if (smtp_input)
3145 {
3146 Uunlink(spool_name); /* Lose data file when closed */
57cc2785 3147 cancel_cutthrough_connection(TRUE, US"sender closed connection");
7e3ce68e
JH
3148 message_id[0] = 0; /* Indicate no message accepted */
3149 smtp_reply = handle_lost_connection(US"");
3150 smtp_yield = FALSE;
3151 goto TIDYUP; /* Skip to end of function */
3152 }
3153 break;
059ec3d9 3154
7e3ce68e
JH
3155 /* Handle message that is too big. Don't use host_or_ident() in the log
3156 message; we want to see the ident value even for non-remote messages. */
059ec3d9 3157
7e3ce68e
JH
3158 case END_SIZE:
3159 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3160 cancel_cutthrough_connection(TRUE, US"mail too big");
7e3ce68e 3161 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
059ec3d9 3162
7e3ce68e
JH
3163 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3164 "message too big: read=%d max=%d",
3165 sender_address,
3166 (sender_fullhost == NULL)? "" : " H=",
3167 (sender_fullhost == NULL)? US"" : sender_fullhost,
3168 (sender_ident == NULL)? "" : " U=",
3169 (sender_ident == NULL)? US"" : sender_ident,
3170 message_size,
3171 thismessage_size_limit);
3172
3173 if (smtp_input)
3174 {
3175 smtp_reply = US"552 Message size exceeds maximum permitted";
3176 message_id[0] = 0; /* Indicate no message accepted */
3177 goto TIDYUP; /* Skip to end of function */
3178 }
3179 else
3180 {
3181 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3182 give_local_error(ERRMESS_TOOBIG,
3183 string_sprintf("message too big (max=%d)", thismessage_size_limit),
3184 US"message rejected: ", error_rc, data_file, header_list);
3185 /* Does not return */
3186 }
3187 break;
3188
3189 /* Handle bad BDAT protocol sequence */
3190
3191 case END_PROTOCOL:
3192 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3193 cancel_cutthrough_connection(TRUE, US"sender protocol error");
7e3ce68e
JH
3194 smtp_reply = US""; /* Response already sent */
3195 message_id[0] = 0; /* Indicate no message accepted */
3196 goto TIDYUP; /* Skip to end of function */
059ec3d9
PH
3197 }
3198 }
3199
3200/* Restore the standard SIGALRM handler for any subsequent processing. (For
3201example, there may be some expansion in an ACL that uses a timer.) */
3202
3203os_non_restarting_signal(SIGALRM, sigalrm_handler);
3204
3205/* The message body has now been read into the data file. Call fflush() to
3206empty the buffers in C, and then call fsync() to get the data written out onto
3207the disk, as fflush() doesn't do this (or at least, it isn't documented as
3208having to do this). If there was an I/O error on either input or output,
3209attempt to send an error message, and unlink the spool file. For non-SMTP input
3210we can then give up. Note that for SMTP input we must swallow the remainder of
3211the input in cases of output errors, since the far end doesn't expect to see
3212anything until the terminating dot line is sent. */
3213
3214if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 3215 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
3216 {
3217 uschar *msg_errno = US strerror(errno);
3218 BOOL input_error = (receive_ferror)() != 0;
3219 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
3220 input_error? "Input read" : "Spool write",
3221 msg_errno,
3222 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
3223
3224 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
3225 Uunlink(spool_name); /* Lose the data file */
57cc2785 3226 cancel_cutthrough_connection(TRUE, US"error writing spoolfile");
059ec3d9
PH
3227
3228 if (smtp_input)
3229 {
3230 if (input_error)
3231 smtp_reply = US"451 Error while reading input data";
3232 else
3233 {
3234 smtp_reply = US"451 Error while writing spool file";
3235 receive_swallow_smtp();
3236 }
3237 message_id[0] = 0; /* Indicate no message accepted */
3238 goto TIDYUP; /* Skip to end of function */
3239 }
3240
3241 else
3242 {
3243 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3244 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
3245 header_list);
3246 /* Does not return */
3247 }
3248 }
3249
3250
3251/* No I/O errors were encountered while writing the data file. */
3252
3253DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
3254
3255
3256/* If there were any bad addresses extracted by -t, or there were no recipients
3257left after -t, send a message to the sender of this message, or write it to
3258stderr if the error handling option is set that way. Note that there may
3259legitimately be no recipients for an SMTP message if they have all been removed
3260by "discard".
3261
3262We need to rewind the data file in order to read it. In the case of no
3263recipients or stderr error writing, throw the data file away afterwards, and
3264exit. (This can't be SMTP, which always ensures there's at least one
3265syntactically good recipient address.) */
3266
3267if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
3268 {
3269 DEBUG(D_receive)
3270 {
3271 if (recipients_count == 0) debug_printf("*** No recipients\n");
3272 if (bad_addresses != NULL)
3273 {
3274 error_block *eblock = bad_addresses;
3275 debug_printf("*** Bad address(es)\n");
3276 while (eblock != NULL)
3277 {
3278 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
3279 eblock = eblock->next;
3280 }
3281 }
3282 }
3283
3284 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3285
3286 /* If configured to send errors to the sender, but this fails, force
3287 a failure error code. We use a special one for no recipients so that it
3288 can be detected by the autoreply transport. Otherwise error_rc is set to
3289 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
3290 it is EXIT_SUCCESS. */
3291
3292 if (error_handling == ERRORS_SENDER)
3293 {
3294 if (!moan_to_sender(
3295 (bad_addresses == NULL)?
3296 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
3297 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
3298 bad_addresses, header_list, data_file, FALSE))
3299 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
3300 }
3301 else
3302 {
3303 if (bad_addresses == NULL)
3304 {
3305 if (extracted_ignored)
3306 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3307 else
3308 fprintf(stderr, "exim: no recipients in message\n");
3309 }
3310 else
3311 {
3312 fprintf(stderr, "exim: invalid address%s",
3313 (bad_addresses->next == NULL)? ":" : "es:\n");
3314 while (bad_addresses != NULL)
3315 {
3316 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3317 bad_addresses->text2);
3318 bad_addresses = bad_addresses->next;
3319 }
3320 }
3321 }
3322
3323 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3324 {
3325 Uunlink(spool_name);
f1e894f3 3326 (void)fclose(data_file);
9bfb7e1b 3327 exim_exit(error_rc, US"receiving");
059ec3d9
PH
3328 }
3329 }
3330
3331/* Data file successfully written. Generate text for the Received: header by
3332expanding the configured string, and adding a timestamp. By leaving this
3333operation till now, we ensure that the timestamp is the time that message
3334reception was completed. However, this is deliberately done before calling the
3335data ACL and local_scan().
3336
3337This Received: header may therefore be inspected by the data ACL and by code in
3338the local_scan() function. When they have run, we update the timestamp to be
3339the final time of reception.
3340
3341If there is just one recipient, set up its value in the $received_for variable
3342for use when we generate the Received: header.
3343
3344Note: the checking for too many Received: headers is handled by the delivery
3345code. */
e4bdf652 3346/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3347
e4bdf652 3348if (received_header->text == NULL) /* Non-cutthrough case */
059ec3d9 3349 {
e4bdf652 3350 received_header_gen();
059ec3d9 3351
e4bdf652 3352 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3353
e4bdf652
JH
3354 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3355 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3356
e4bdf652
JH
3357 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3358 now, before running the DATA ACL. */
059ec3d9 3359
578d43dc 3360 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652 3361 }
817d9f57 3362else
e4bdf652
JH
3363 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3364 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3365
3366/* If an ACL is specified for checking things at this stage of reception of a
3367message, run it, unless all the recipients were removed by "discard" in earlier
3368ACLs. That is the only case in which recipients_count can be zero at this
3369stage. Set deliver_datafile to point to the data file so that $message_body and
3370$message_body_end can be extracted if needed. Allow $recipients in expansions.
3371*/
3372
3373deliver_datafile = data_fd;
4e88a19f 3374user_msg = NULL;
059ec3d9 3375
0e20aff9
MH
3376enable_dollar_recipients = TRUE;
3377
059ec3d9 3378if (recipients_count == 0)
7e3ce68e
JH
3379 blackholed_by = recipients_discarded ? US"MAIL ACL" : US"RCPT ACL";
3380
059ec3d9
PH
3381else
3382 {
059ec3d9
PH
3383 /* Handle interactive SMTP messages */
3384
3385 if (smtp_input && !smtp_batched_input)
3386 {
8523533c 3387
80a47a2c
TK
3388#ifndef DISABLE_DKIM
3389 if (!dkim_disable_verify)
3390 {
3391 /* Finish verification, this will log individual signature results to
3392 the mainlog */
3393 dkim_exim_verify_finish();
3394
3395 /* Check if we must run the DKIM ACL */
7e3ce68e 3396 if (acl_smtp_dkim && dkim_verify_signers && *dkim_verify_signers)
80a47a2c
TK
3397 {
3398 uschar *dkim_verify_signers_expanded =
3399 expand_string(dkim_verify_signers);
7e3ce68e 3400 if (!dkim_verify_signers_expanded)
80a47a2c
TK
3401 log_write(0, LOG_MAIN|LOG_PANIC,
3402 "expansion of dkim_verify_signers option failed: %s",
3403 expand_string_message);
7e3ce68e 3404
80a47a2c
TK
3405 else
3406 {
3407 int sep = 0;
a79d8834
JH
3408 const uschar * ptr = dkim_verify_signers_expanded;
3409 uschar * item = NULL;
acec9514
JH
3410 gstring * seen_items = NULL;
3411
9122af94
TK
3412 /* Default to OK when no items are present */
3413 rc = OK;
7e3ce68e 3414 while ((item = string_nextinlist(&ptr, &sep, NULL, 0)))
80a47a2c 3415 {
6119d1ea 3416 /* Prevent running ACL for an empty item */
7e3ce68e 3417 if (!item || !*item) continue;
5032d1cf
JH
3418
3419 /* Only run ACL once for each domain or identity,
3420 no matter how often it appears in the expanded list. */
3421 if (seen_items)
6119d1ea 3422 {
acec9514
JH
3423 uschar *seen_item;
3424 const uschar *seen_items_list = string_from_gstring(seen_items);
5032d1cf 3425 BOOL seen_this_item = FALSE;
61147df4 3426
ae9094bf 3427 while ((seen_item = string_nextinlist(&seen_items_list, &sep,
7e3ce68e 3428 NULL, 0)))
5032d1cf
JH
3429 if (Ustrcmp(seen_item,item) == 0)
3430 {
3431 seen_this_item = TRUE;
3432 break;
3433 }
3434
3435 if (seen_this_item)
6119d1ea
TK
3436 {
3437 DEBUG(D_receive)
5032d1cf
JH
3438 debug_printf("acl_smtp_dkim: skipping signer %s, "
3439 "already seen\n", item);
6119d1ea
TK
3440 continue;
3441 }
61147df4 3442
acec9514 3443 seen_items = string_cat(seen_items, ":");
6119d1ea
TK
3444 }
3445
acec9514 3446 seen_items = string_cat(seen_items, item);
6119d1ea
TK
3447
3448 DEBUG(D_receive)
5032d1cf
JH
3449 debug_printf("calling acl_smtp_dkim for dkim_cur_signer=%s\n",
3450 item);
6119d1ea 3451
80a47a2c 3452 dkim_exim_acl_setup(item);
5032d1cf
JH
3453 rc = acl_check(ACL_WHERE_DKIM, NULL, acl_smtp_dkim,
3454 &user_msg, &log_msg);
a79d8834 3455 dkim_exim_verify_log_item();
6119d1ea
TK
3456
3457 if (rc != OK)
5032d1cf
JH
3458 {
3459 DEBUG(D_receive)
3460 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, "
3461 "skipping remaining items\n", rc, item);
57cc2785 3462 cancel_cutthrough_connection(TRUE, US"dkim acl not ok");
5032d1cf
JH
3463 break;
3464 }
80a47a2c 3465 }
578d43dc 3466 add_acl_headers(ACL_WHERE_DKIM, US"DKIM");
80a47a2c
TK
3467 if (rc == DISCARD)
3468 {
3469 recipients_count = 0;
3470 blackholed_by = US"DKIM ACL";
a79d8834 3471 if (log_msg)
80a47a2c
TK
3472 blackhole_log_msg = string_sprintf(": %s", log_msg);
3473 }
3474 else if (rc != OK)
3475 {
3476 Uunlink(spool_name);
3477 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
85ffcba6 3478 smtp_yield = FALSE; /* No more messages after dropped connection */
80a47a2c
TK
3479 smtp_reply = US""; /* Indicate reply already sent */
3480 message_id[0] = 0; /* Indicate no message accepted */
3481 goto TIDYUP; /* Skip to end of function */
3482 }
3483 }
3484 }
a79d8834
JH
3485 else
3486 dkim_exim_verify_log_all();
80a47a2c 3487 }
4a8ce2d8 3488#endif /* DISABLE_DKIM */
fb2274d4 3489
8523533c 3490#ifdef WITH_CONTENT_SCAN
80a47a2c
TK
3491 if (recipients_count > 0 &&
3492 acl_smtp_mime != NULL &&
54cdb463
PH
3493 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
3494 goto TIDYUP;
8523533c
TK
3495#endif /* WITH_CONTENT_SCAN */
3496
4840604e
TL
3497#ifdef EXPERIMENTAL_DMARC
3498 dmarc_up = dmarc_store_data(from_header);
3499#endif /* EXPERIMENTAL_DMARC */
3500
8ccd00b1
JH
3501#ifndef DISABLE_PRDR
3502 if (prdr_requested && recipients_count > 1 && acl_smtp_data_prdr)
fd98a5c6
JH
3503 {
3504 unsigned int c;
3505 int all_pass = OK;
3506 int all_fail = FAIL;
3507
925ac8e4 3508 smtp_printf("353 PRDR content analysis beginning\r\n", TRUE);
fd98a5c6
JH
3509 /* Loop through recipients, responses must be in same order received */
3510 for (c = 0; recipients_count > c; c++)
3511 {
3512 uschar * addr= recipients_list[c].address;
3513 uschar * msg= US"PRDR R=<%s> %s";
3514 uschar * code;
3515 DEBUG(D_receive)
3516 debug_printf("PRDR processing recipient %s (%d of %d)\n",
3517 addr, c+1, recipients_count);
3518 rc = acl_check(ACL_WHERE_PRDR, addr,
3519 acl_smtp_data_prdr, &user_msg, &log_msg);
3520
3521 /* If any recipient rejected content, indicate it in final message */
3522 all_pass |= rc;
3523 /* If all recipients rejected, indicate in final message */
3524 all_fail &= rc;
3525
3526 switch (rc)
3527 {
3528 case OK: case DISCARD: code = US"250"; break;
3529 case DEFER: code = US"450"; break;
3530 default: code = US"550"; break;
3531 }
3532 if (user_msg != NULL)
3533 smtp_user_msg(code, user_msg);
3534 else
3535 {
3536 switch (rc)
3537 {
3538 case OK: case DISCARD:
3539 msg = string_sprintf(CS msg, addr, "acceptance"); break;
3540 case DEFER:
3541 msg = string_sprintf(CS msg, addr, "temporary refusal"); break;
3542 default:
3543 msg = string_sprintf(CS msg, addr, "refusal"); break;
3544 }
3545 smtp_user_msg(code, msg);
3546 }
3547 if (log_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, log_msg);
3548 else if (user_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, user_msg);
112b6a93 3549 else log_write(0, LOG_MAIN, "%s", CS msg);
fd98a5c6
JH
3550
3551 if (rc != OK) { receive_remove_recipient(addr); c--; }
3552 }
3553 /* Set up final message, used if data acl gives OK */
3554 smtp_reply = string_sprintf("%s id=%s message %s",
3555 all_fail == FAIL ? US"550" : US"250",
3556 message_id,
3557 all_fail == FAIL
3558 ? US"rejected for all recipients"
3559 : all_pass == OK
3560 ? US"accepted"
3561 : US"accepted for some recipients");
3562 if (recipients_count == 0)
3563 {
3564 message_id[0] = 0; /* Indicate no message accepted */
3565 goto TIDYUP;
3566 }
3567 }
3568 else
3569 prdr_requested = FALSE;
8ccd00b1 3570#endif /* !DISABLE_PRDR */
fd98a5c6 3571
54cdb463
PH
3572 /* Check the recipients count again, as the MIME ACL might have changed
3573 them. */
8523533c 3574
059ec3d9
PH
3575 if (acl_smtp_data != NULL && recipients_count > 0)
3576 {
059ec3d9 3577 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
578d43dc 3578 add_acl_headers(ACL_WHERE_DATA, US"DATA");
059ec3d9
PH
3579 if (rc == DISCARD)
3580 {
3581 recipients_count = 0;
3582 blackholed_by = US"DATA ACL";
57cc2785 3583 if (log_msg)
8e669ac1 3584 blackhole_log_msg = string_sprintf(": %s", log_msg);
57cc2785 3585 cancel_cutthrough_connection(TRUE, US"data acl discard");
059ec3d9
PH
3586 }
3587 else if (rc != OK)
3588 {
3589 Uunlink(spool_name);
57cc2785 3590 cancel_cutthrough_connection(TRUE, US"data acl not ok");
8523533c
TK
3591#ifdef WITH_CONTENT_SCAN
3592 unspool_mbox();
3593#endif
6f0c431a
PP
3594#ifdef EXPERIMENTAL_DCC
3595 dcc_ok = 0;
3596#endif
059ec3d9 3597 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
85ffcba6 3598 smtp_yield = FALSE; /* No more messages after dropped connection */
059ec3d9
PH
3599 smtp_reply = US""; /* Indicate reply already sent */
3600 message_id[0] = 0; /* Indicate no message accepted */
3601 goto TIDYUP; /* Skip to end of function */
3602 }
3603 }
3604 }
3605
3606 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
3607 we cannot take different actions for permanent and temporary rejections. */
3608
54cdb463 3609 else
059ec3d9 3610 {
54cdb463
PH
3611
3612#ifdef WITH_CONTENT_SCAN
3613 if (acl_not_smtp_mime != NULL &&
3614 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
3615 &blackholed_by))
3616 goto TIDYUP;
3617#endif /* WITH_CONTENT_SCAN */
3618
3619 if (acl_not_smtp != NULL)
059ec3d9 3620 {
54cdb463
PH
3621 uschar *user_msg, *log_msg;
3622 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
3623 if (rc == DISCARD)
059ec3d9 3624 {
54cdb463
PH
3625 recipients_count = 0;
3626 blackholed_by = US"non-SMTP ACL";
3627 if (log_msg != NULL)
3628 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 3629 }
54cdb463 3630 else if (rc != OK)
059ec3d9 3631 {
54cdb463
PH
3632 Uunlink(spool_name);
3633#ifdef WITH_CONTENT_SCAN
3634 unspool_mbox();
3635#endif
6f0c431a
PP
3636#ifdef EXPERIMENTAL_DCC
3637 dcc_ok = 0;
3638#endif
6ea85e9a
PH
3639 /* The ACL can specify where rejections are to be logged, possibly
3640 nowhere. The default is main and reject logs. */
3641
3642 if (log_reject_target != 0)
3643 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3644 sender_address, log_msg);
3645
54cdb463
PH
3646 if (user_msg == NULL) user_msg = US"local configuration problem";
3647 if (smtp_batched_input)
3648 {
3649 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3650 /* Does not return */
3651 }
3652 else
3653 {
3654 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3655 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3656 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3657 header_list);
3658 /* Does not return */
3659 }
059ec3d9 3660