CHUNKING: flush input stream after message-fatal error detection. Bug 2201
[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
d21bf202
JH
1026DEBUG(D_receive) debug_printf("CHUNKING: %s\n",
1027 fout ? "writing spoolfile in wire format" : "flushing input");
328c5688
JH
1028spool_file_wireformat = TRUE;
1029
0d81dabc 1030for (;;)
328c5688 1031 {
0d81dabc
JH
1032 if (chunking_data_left > 0)
1033 {
1034 unsigned len = MAX(chunking_data_left, thismessage_size_limit - message_size + 1);
1035 uschar * buf = bdat_getbuf(&len);
328c5688 1036
0d81dabc
JH
1037 message_size += len;
1038 if (fout && fwrite(buf, len, 1, fout) != 1) return END_WERROR;
1039 }
1040 else switch (ch = bdat_getc(GETC_BUFFER_UNLIMITED))
1041 {
1042 case EOF: return END_EOF;
1043 case EOD: return END_DOT;
1044 case ERR: return END_PROTOCOL;
1045
1046 default:
1047 message_size++;
1048 /*XXX not done:
1049 linelength
1050 max_received_linelength
1051 body_linecount
1052 body_zerocount
1053 */
1054 if (fout && fputc(ch, fout) == EOF) return END_WERROR;
1055 break;
1056 }
1057 if (message_size > thismessage_size_limit) return END_SIZE;
328c5688
JH
1058 }
1059/*NOTREACHED*/
1060}
1061
7e3ce68e
JH
1062
1063
1064
059ec3d9
PH
1065/*************************************************
1066* Swallow SMTP message *
1067*************************************************/
1068
1069/* This function is called when there has been some kind of error while reading
1070an SMTP message, and the remaining data may need to be swallowed. It is global
1071because it is called from smtp_closedown() to shut down an incoming call
1072tidily.
1073
1074Argument: a FILE from which to read the message
1075Returns: nothing
1076*/
1077
1078void
1079receive_swallow_smtp(void)
1080{
1081if (message_ended >= END_NOTENDED)
d21bf202
JH
1082 message_ended = chunking_state <= CHUNKING_OFFERED
1083 ? read_message_data_smtp(NULL)
1084 : read_message_bdat_smtp_wire(NULL);
059ec3d9
PH
1085}
1086
1087
1088
1089/*************************************************
1090* Handle lost SMTP connection *
1091*************************************************/
1092
1093/* This function logs connection loss incidents and generates an appropriate
1094SMTP response.
1095
1096Argument: additional data for the message
1097Returns: the SMTP response
1098*/
1099
1100static uschar *
1101handle_lost_connection(uschar *s)
1102{
1103log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
1104 "%s lost while reading message data%s", smtp_get_connection_info(), s);
eea0defe 1105smtp_notquit_exit(US"connection-lost", NULL, NULL);
059ec3d9
PH
1106return US"421 Lost incoming connection";
1107}
1108
1109
1110
1111
1112/*************************************************
1113* Handle a non-smtp reception error *
1114*************************************************/
1115
1116/* This function is called for various errors during the reception of non-SMTP
1117messages. It either sends a message to the sender of the problem message, or it
1118writes to the standard error stream.
1119
1120Arguments:
1121 errcode code for moan_to_sender(), identifying the error
1122 text1 first message text, passed to moan_to_sender()
1123 text2 second message text, used only for stderrr
1124 error_rc code to pass to exim_exit if no problem
1125 f FILE containing body of message (may be stdin)
1126 hptr pointer to instore headers or NULL
1127
1128Returns: calls exim_exit(), which does not return
1129*/
1130
1131static void
1132give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
1133 FILE *f, header_line *hptr)
1134{
1135if (error_handling == ERRORS_SENDER)
1136 {
1137 error_block eblock;
1138 eblock.next = NULL;
1139 eblock.text1 = text1;
37f3dc43 1140 eblock.text2 = US"";
059ec3d9
PH
1141 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
1142 error_rc = EXIT_FAILURE;
1143 }
37f3dc43
JH
1144else
1145 fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 1146(void)fclose(f);
9bfb7e1b 1147exim_exit(error_rc, US"");
059ec3d9
PH
1148}
1149
1150
1151
1152/*************************************************
1153* Add header lines set up by ACL *
1154*************************************************/
1155
850635b6
PH
1156/* This function is called to add the header lines that were set up by
1157statements in an ACL to the list of headers in memory. It is done in two stages
1158like this, because when the ACL for RCPT is running, the other headers have not
1159yet been received. This function is called twice; once just before running the
1160DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
1161are visible to the DATA ACL.
059ec3d9
PH
1162
1163Originally these header lines were added at the end. Now there is support for
1164three different places: top, bottom, and after the Received: header(s). There
1165will always be at least one Received: header, even if it is marked deleted, and
1166even if something else has been put in front of it.
1167
1168Arguments:
1169 acl_name text to identify which ACL
1170
1171Returns: nothing
1172*/
1173
1174static void
578d43dc 1175add_acl_headers(int where, uschar *acl_name)
059ec3d9
PH
1176{
1177header_line *h, *next;
1178header_line *last_received = NULL;
e7568d51 1179
578d43dc
JH
1180switch(where)
1181 {
1182 case ACL_WHERE_DKIM:
1183 case ACL_WHERE_MIME:
af4a1bca 1184 case ACL_WHERE_DATA:
57cc2785
JH
1185 if ( cutthrough.fd >= 0 && cutthrough.delivery
1186 && (acl_removed_headers || acl_added_headers))
578d43dc
JH
1187 {
1188 log_write(0, LOG_MAIN|LOG_PANIC, "Header modification in data ACLs"
af4a1bca 1189 " will not take effect on cutthrough deliveries");
578d43dc
JH
1190 return;
1191 }
1192 }
1193
57cc2785 1194if (acl_removed_headers)
e7568d51 1195 {
e1d04f48 1196 DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers removed by %s ACL:\n", acl_name);
e7568d51 1197
57cc2785 1198 for (h = header_list; h; h = h->next) if (h->type != htype_old)
e7568d51 1199 {
55414b25 1200 const uschar * list = acl_removed_headers;
e7568d51
TL
1201 int sep = ':'; /* This is specified as a colon-separated list */
1202 uschar *s;
1203 uschar buffer[128];
4a142059
JH
1204
1205 while ((s = string_nextinlist(&list, &sep, buffer, sizeof(buffer))))
1206 if (header_testname(h, s, Ustrlen(s), FALSE))
e7568d51
TL
1207 {
1208 h->type = htype_old;
e1d04f48 1209 DEBUG(D_receive|D_acl) debug_printf_indent(" %s", h->text);
e7568d51 1210 }
e7568d51
TL
1211 }
1212 acl_removed_headers = NULL;
e1d04f48 1213 DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
e7568d51 1214 }
059ec3d9 1215
57cc2785 1216if (!acl_added_headers) return;
e1d04f48 1217DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers added by %s ACL:\n", acl_name);
059ec3d9 1218
57cc2785 1219for (h = acl_added_headers; h; h = next)
059ec3d9
PH
1220 {
1221 next = h->next;
1222
1223 switch(h->type)
1224 {
1225 case htype_add_top:
1226 h->next = header_list;
1227 header_list = h;
e1d04f48 1228 DEBUG(D_receive|D_acl) debug_printf_indent(" (at top)");
059ec3d9
PH
1229 break;
1230
1231 case htype_add_rec:
1232 if (last_received == NULL)
1233 {
1234 last_received = header_list;
1235 while (!header_testname(last_received, US"Received", 8, FALSE))
1236 last_received = last_received->next;
1237 while (last_received->next != NULL &&
1238 header_testname(last_received->next, US"Received", 8, FALSE))
1239 last_received = last_received->next;
1240 }
1241 h->next = last_received->next;
1242 last_received->next = h;
e1d04f48 1243 DEBUG(D_receive|D_acl) debug_printf_indent(" (after Received:)");
059ec3d9
PH
1244 break;
1245
8523533c
TK
1246 case htype_add_rfc:
1247 /* add header before any header which is NOT Received: or Resent- */
1248 last_received = header_list;
1249 while ( (last_received->next != NULL) &&
1250 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
1251 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
1252 last_received = last_received->next;
1253 /* last_received now points to the last Received: or Resent-* header
1254 in an uninterrupted chain of those header types (seen from the beginning
1255 of all headers. Our current header must follow it. */
1256 h->next = last_received->next;
1257 last_received->next = h;
e1d04f48 1258 DEBUG(D_receive|D_acl) debug_printf_indent(" (before any non-Received: or Resent-*: header)");
8523533c
TK
1259 break;
1260
059ec3d9
PH
1261 default:
1262 h->next = NULL;
1263 header_last->next = h;
1264 break;
1265 }
1266
1267 if (h->next == NULL) header_last = h;
1268
1269 /* Check for one of the known header types (From:, To:, etc.) though in
1270 practice most added headers are going to be "other". Lower case
1271 identification letters are never stored with the header; they are used
1272 for existence tests when messages are received. So discard any lower case
1273 flag values. */
1274
1275 h->type = header_checkname(h, FALSE);
1276 if (h->type >= 'a') h->type = htype_other;
1277
e1d04f48 1278 DEBUG(D_receive|D_acl) debug_printf_indent(" %s", header_last->text);
059ec3d9
PH
1279 }
1280
71fafd95 1281acl_added_headers = NULL;
e1d04f48 1282DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
059ec3d9
PH
1283}
1284
1285
1286
1287/*************************************************
1288* Add host information for log line *
1289*************************************************/
1290
1291/* Called for acceptance and rejecting log lines. This adds information about
1292the calling host to a string that is being built dynamically.
1293
1294Arguments:
1295 s the dynamic string
059ec3d9
PH
1296
1297Returns: the extended string
1298*/
1299
acec9514
JH
1300static gstring *
1301add_host_info_for_log(gstring * g)
059ec3d9 1302{
fc16abb4 1303if (sender_fullhost)
059ec3d9 1304 {
fc16abb4 1305 if (LOGGING(dnssec) && sender_host_dnssec) /*XXX sender_helo_dnssec? */
acec9514
JH
1306 g = string_catn(g, US" DS", 3);
1307 g = string_append(g, 2, US" H=", sender_fullhost);
6c6d6e48 1308 if (LOGGING(incoming_interface) && interface_address != NULL)
059ec3d9 1309 {
acec9514 1310 g = string_cat(g,
fc16abb4 1311 string_sprintf(" I=[%s]:%d", interface_address, interface_port));
059ec3d9
PH
1312 }
1313 }
a2673768
JH
1314if (tcp_in_fastopen && !tcp_in_fastopen_logged)
1315 {
acec9514 1316 g = string_catn(g, US" TFO", 4);
a2673768
JH
1317 tcp_in_fastopen_logged = TRUE;
1318 }
1319if (sender_ident)
acec9514 1320 g = string_append(g, 2, US" U=", sender_ident);
a2673768 1321if (received_protocol)
acec9514
JH
1322 g = string_append(g, 2, US" P=", received_protocol);
1323return g;
059ec3d9
PH
1324}
1325
1326
1327
63955bf2 1328#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1329
1330/*************************************************
54cdb463
PH
1331* Run the MIME ACL on a message *
1332*************************************************/
1333
1334/* This code is in a subroutine so that it can be used for both SMTP
1335and non-SMTP messages. It is called with a non-NULL ACL pointer.
1336
1337Arguments:
1338 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1339 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1340 smtp_reply_ptr Where SMTP reply is being built
1341 blackholed_by_ptr Where "blackholed by" message is being built
1342
1343Returns: TRUE to carry on; FALSE to abandon the message
1344*/
1345
1346static BOOL
1347run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1348 uschar **blackholed_by_ptr)
1349{
1350FILE *mbox_file;
1351uschar rfc822_file_path[2048];
1352unsigned long mbox_size;
1353header_line *my_headerlist;
1354uschar *user_msg, *log_msg;
1355int mime_part_count_buffer = -1;
040721f2 1356uschar * mbox_filename;
7156b1ef 1357int rc = OK;
54cdb463
PH
1358
1359memset(CS rfc822_file_path,0,2048);
1360
1361/* check if it is a MIME message */
040721f2
JH
1362
1363for (my_headerlist = header_list; my_headerlist; my_headerlist = my_headerlist->next)
1364 if ( my_headerlist->type != '*' /* skip deleted headers */
1365 && strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0
1366 )
4e88a19f 1367 {
54cdb463
PH
1368 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1369 goto DO_MIME_ACL;
4e88a19f 1370 }
54cdb463
PH
1371
1372DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1373return TRUE;
1374
1375DO_MIME_ACL:
040721f2 1376
54cdb463 1377/* make sure the eml mbox file is spooled up */
040721f2
JH
1378if (!(mbox_file = spool_mbox(&mbox_size, NULL, &mbox_filename)))
1379 { /* error while spooling */
54cdb463
PH
1380 log_write(0, LOG_MAIN|LOG_PANIC,
1381 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1382 Uunlink(spool_name);
1383 unspool_mbox();
6f0c431a
PP
1384#ifdef EXPERIMENTAL_DCC
1385 dcc_ok = 0;
1386#endif
a5bd321b 1387 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1388 message_id[0] = 0; /* Indicate no message accepted */
1389 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1390 return FALSE; /* Indicate skip to end of receive function */
040721f2 1391 }
54cdb463
PH
1392
1393mime_is_rfc822 = 0;
1394
1395MIME_ACL_CHECK:
1396mime_part_count = -1;
1397rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1398(void)fclose(mbox_file);
54cdb463 1399
4e88a19f
PH
1400if (Ustrlen(rfc822_file_path) > 0)
1401 {
54cdb463
PH
1402 mime_part_count = mime_part_count_buffer;
1403
4e88a19f
PH
1404 if (unlink(CS rfc822_file_path) == -1)
1405 {
54cdb463
PH
1406 log_write(0, LOG_PANIC,
1407 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
1408 goto END_MIME_ACL;
4e88a19f
PH
1409 }
1410 }
54cdb463
PH
1411
1412/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1413if (rc == OK)
1414 {
040721f2 1415 uschar * scandir;
e8bc7fca
JH
1416 struct dirent * entry;
1417 DIR * tempdir;
54cdb463 1418
040721f2 1419 scandir = string_copyn(mbox_filename, Ustrrchr(mbox_filename, '/') - mbox_filename);
54cdb463 1420
040721f2 1421 tempdir = opendir(CS scandir);
e8bc7fca 1422 for (;;)
4e88a19f 1423 {
e8bc7fca
JH
1424 if (!(entry = readdir(tempdir)))
1425 break;
1426 if (strncmpic(US entry->d_name, US"__rfc822_", 9) == 0)
4e88a19f 1427 {
e8bc7fca 1428 (void) string_format(rfc822_file_path, sizeof(rfc822_file_path),
040721f2 1429 "%s/%s", scandir, entry->d_name);
e1d04f48 1430 DEBUG(D_receive) debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
e8bc7fca 1431 rfc822_file_path);
4e88a19f
PH
1432 break;
1433 }
e8bc7fca 1434 }
4e88a19f 1435 closedir(tempdir);
54cdb463 1436
e8bc7fca 1437 if (entry)
4e88a19f 1438 {
e8bc7fca 1439 if ((mbox_file = Ufopen(rfc822_file_path, "rb")))
4e88a19f 1440 {
e8bc7fca
JH
1441 /* set RFC822 expansion variable */
1442 mime_is_rfc822 = 1;
1443 mime_part_count_buffer = mime_part_count;
1444 goto MIME_ACL_CHECK;
4e88a19f 1445 }
e8bc7fca
JH
1446 log_write(0, LOG_PANIC,
1447 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1448 unlink(CS rfc822_file_path);
4e88a19f
PH
1449 }
1450 }
54cdb463
PH
1451
1452END_MIME_ACL:
578d43dc 1453add_acl_headers(ACL_WHERE_MIME, US"MIME");
54cdb463
PH
1454if (rc == DISCARD)
1455 {
1456 recipients_count = 0;
1457 *blackholed_by_ptr = US"MIME ACL";
1458 }
1459else if (rc != OK)
1460 {
1461 Uunlink(spool_name);
1462 unspool_mbox();
6f0c431a
PP
1463#ifdef EXPERIMENTAL_DCC
1464 dcc_ok = 0;
1465#endif
4f6ae5c3
JH
1466 if ( smtp_input
1467 && smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1468 {
85ffcba6 1469 *smtp_yield_ptr = FALSE; /* No more messages after dropped connection */
f4c1088b 1470 *smtp_reply_ptr = US""; /* Indicate reply already sent */
4f6ae5c3 1471 }
54cdb463
PH
1472 message_id[0] = 0; /* Indicate no message accepted */
1473 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1474 }
54cdb463
PH
1475
1476return TRUE;
1477}
1478
63955bf2 1479#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1480
1481
e4bdf652
JH
1482
1483void
1484received_header_gen(void)
1485{
1486uschar *received;
1487uschar *timestamp;
1488header_line *received_header= header_list;
1489
1490timestamp = expand_string(US"${tod_full}");
1491if (recipients_count == 1) received_for = recipients_list[0].address;
1492received = expand_string(received_header_text);
1493received_for = NULL;
1494
d4ff61d1 1495if (!received)
e4bdf652
JH
1496 {
1497 if(spool_name[0] != 0)
1498 Uunlink(spool_name); /* Lose the data file */
1499 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1500 "(received_header_text) failed: %s", string_printing(received_header_text),
1501 expand_string_message);
1502 }
1503
1504/* The first element on the header chain is reserved for the Received header,
1505so all we have to do is fill in the text pointer, and set the type. However, if
1506the result of the expansion is an empty string, we leave the header marked as
1507"old" so as to refrain from adding a Received header. */
1508
1509if (received[0] == 0)
1510 {
1511 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1512 received_header->type = htype_old;
1513 }
1514else
1515 {
1516 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1517 received_header->type = htype_received;
1518 }
1519
1520received_header->slen = Ustrlen(received_header->text);
1521
1522DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1523 received_header->type, received_header->text);
1524}
1525
1526
1527
54cdb463 1528/*************************************************
059ec3d9
PH
1529* Receive message *
1530*************************************************/
1531
1532/* Receive a message on the given input, and put it into a pair of spool files.
1533Either a non-null list of recipients, or the extract flag will be true, or
1534both. The flag sender_local is true for locally generated messages. The flag
1535submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1536suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1537suppress_local_fixups" or -G was passed on the command-line.
1538The flag smtp_input is true if the message is to be
8800895a
PH
1539handled using SMTP conventions about termination and lines starting with dots.
1540For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1541
1542If a message was successfully read, message_id[0] will be non-zero.
1543
1544The general actions of this function are:
1545
1546 . Read the headers of the message (if any) into a chain of store
1547 blocks.
1548
1549 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1550 throw it away, unless the caller is trusted, or unless
1551 active_local_sender_retain is set - which can only happen if
1552 active_local_from_check is false.
059ec3d9
PH
1553
1554 . If recipients are to be extracted from the message, build the
1555 recipients list from the headers, removing any that were on the
1556 original recipients list (unless extract_addresses_remove_arguments is
1557 false), and at the same time, remove any bcc header that may be present.
1558
1559 . Get the spool file for the data, sort out its unique name, open
1560 and lock it (but don't give it the name yet).
1561
1562 . Generate a "Message-Id" header if the message doesn't have one, for
1563 locally-originated messages.
1564
1565 . Generate a "Received" header.
1566
1567 . Ensure the recipients list is fully qualified and rewritten if necessary.
1568
1569 . If there are any rewriting rules, apply them to the sender address
1570 and also to the headers.
1571
1572 . If there is no from: header, generate one, for locally-generated messages
1573 and messages in "submission mode" only.
1574
1575 . If the sender is local, check that from: is correct, and if not, generate
1576 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1577 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1578
1579 . If there is no "date" header, generate one, for locally-originated
1580 or submission mode messages only.
1581
1582 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1583 dot_ends mode, to the data file. Leave it open, to hold the lock.
1584
1585 . Write the envelope and the headers to a new file.
1586
1587 . Set the name for the header file; close it.
1588
1589 . Set the name for the data file; close it.
1590
1591Because this function can potentially be called many times in a single
1592SMTP connection, all store should be got by store_get(), so that it will be
1593automatically retrieved after the message is accepted.
1594
1595FUDGE: It seems that sites on the net send out messages with just LF
1596terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1597we make the CRs optional in all cases.
1598
1599July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1600new regime is now in place in which bare CRs in header lines are turned into LF
1601followed by a space, so as not to terminate the header line.
1602
1603February 2004: A bare LF in a header line in a message whose first line was
1604terminated by CRLF is treated in the same way as a bare CR.
1605
1606Arguments:
1607 extract_recip TRUE if recipients are to be extracted from the message's
1608 headers
1609
1610Returns: TRUE there are more messages to be read (SMTP input)
1611 FALSE there are no more messages to be read (non-SMTP input
1612 or SMTP connection collapsed, or other failure)
1613
1614When reading a message for filter testing, the returned value indicates
1615whether the headers (which is all that is read) were terminated by '.' or
1616not. */
1617
1618BOOL
1619receive_msg(BOOL extract_recip)
1620{
7156b1ef
NM
1621int i;
1622int rc = FAIL;
059ec3d9
PH
1623int msg_size = 0;
1624int process_info_len = Ustrlen(process_info);
1625int error_rc = (error_handling == ERRORS_SENDER)?
1626 errors_sender_rc : EXIT_FAILURE;
1627int header_size = 256;
acec9514 1628int start, end, domain;
059ec3d9
PH
1629int id_resolution;
1630int had_zero = 0;
d677b2f2 1631int prevlines_length = 0;
059ec3d9
PH
1632
1633register int ptr = 0;
1634
1635BOOL contains_resent_headers = FALSE;
1636BOOL extracted_ignored = FALSE;
1637BOOL first_line_ended_crlf = TRUE_UNSET;
1638BOOL smtp_yield = TRUE;
1639BOOL yield = FALSE;
1640
1641BOOL resents_exist = FALSE;
1642uschar *resent_prefix = US"";
1643uschar *blackholed_by = NULL;
04f7d5b9 1644uschar *blackhole_log_msg = US"";
c5430c20 1645enum {NOT_TRIED, TMP_REJ, PERM_REJ, ACCEPTED} cutthrough_done = NOT_TRIED;
059ec3d9
PH
1646
1647flock_t lock_data;
1648error_block *bad_addresses = NULL;
1649
1650uschar *frozen_by = NULL;
1651uschar *queued_by = NULL;
1652
acec9514
JH
1653uschar *errmsg;
1654gstring * g;
059ec3d9
PH
1655struct stat statbuf;
1656
4e88a19f 1657/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1658
1659uschar *smtp_reply = NULL;
4e88a19f 1660uschar *user_msg, *log_msg;
059ec3d9
PH
1661
1662/* Working header pointers */
1663
1664header_line *h, *next;
1665
2cbb4081 1666/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1667
1668BOOL date_header_exists = FALSE;
1669
1670/* Pointers to receive the addresses of headers whose contents we need. */
1671
1672header_line *from_header = NULL;
1673header_line *subject_header = NULL;
1674header_line *msgid_header = NULL;
1675header_line *received_header;
1676
4840604e
TL
1677#ifdef EXPERIMENTAL_DMARC
1678int dmarc_up = 0;
1679#endif /* EXPERIMENTAL_DMARC */
1680
059ec3d9
PH
1681/* Variables for use when building the Received: header. */
1682
059ec3d9
PH
1683uschar *timestamp;
1684int tslen;
1685
1686/* Release any open files that might have been cached while preparing to
1687accept the message - e.g. by verifying addresses - because reading a message
1688might take a fair bit of real time. */
1689
1690search_tidyup();
1691
e4bdf652
JH
1692/* Extracting the recipient list from an input file is incompatible with
1693cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1694to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1695if (extract_recip || !smtp_input)
57cc2785 1696 cancel_cutthrough_connection(TRUE, US"not smtp input");
e4bdf652 1697
059ec3d9
PH
1698/* Initialize the chain of headers by setting up a place-holder for Received:
1699header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1700pointing to the end of the chain to make adding headers simple. */
1701
1702received_header = header_list = header_last = store_get(sizeof(header_line));
1703header_list->next = NULL;
1704header_list->type = htype_old;
1705header_list->text = NULL;
1706header_list->slen = 0;
1707
1708/* Control block for the next header to be read. */
1709
1710next = store_get(sizeof(header_line));
1711next->text = store_get(header_size);
1712
1713/* Initialize message id to be null (indicating no message read), and the
1714header names list to be the normal list. Indicate there is no data file open
1715yet, initialize the size and warning count, and deal with no size limit. */
1716
1717message_id[0] = 0;
1718data_file = NULL;
1719data_fd = -1;
41313d92 1720spool_name = US"";
059ec3d9
PH
1721message_size = 0;
1722warning_count = 0;
d677b2f2 1723received_count = 1; /* For the one we will add */
059ec3d9
PH
1724
1725if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1726
2e0c1448 1727/* While reading the message, the following counts are computed. */
059ec3d9 1728
d677b2f2
PH
1729message_linecount = body_linecount = body_zerocount =
1730 max_received_linelength = 0;
059ec3d9 1731
80a47a2c 1732#ifndef DISABLE_DKIM
e983e85a
JH
1733/* Call into DKIM to set up the context. In CHUNKING mode
1734we clear the dot-stuffing flag */
1735if (smtp_input && !smtp_batched_input && !dkim_disable_verify)
1736 dkim_exim_verify_init(chunking_state <= CHUNKING_OFFERED);
fb2274d4
TK
1737#endif
1738
4840604e
TL
1739#ifdef EXPERIMENTAL_DMARC
1740/* initialize libopendmarc */
1741dmarc_up = dmarc_init();
1742#endif
1743
059ec3d9
PH
1744/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1745ids, and fractions of a second are required. See the comments that precede the
1746message id creation below. */
1747
1748(void)gettimeofday(&message_id_tv, NULL);
1749
1750/* For other uses of the received time we can operate with granularity of one
1751second, and for that we use the global variable received_time. This is for
32dfdf8b 1752things like ultimate message timeouts.XXX */
059ec3d9 1753
32dfdf8b 1754received_time = message_id_tv;
059ec3d9
PH
1755
1756/* If SMTP input, set the special handler for timeouts. The alarm() calls
1757happen in the smtp_getc() function when it refills its buffer. */
1758
1759if (smtp_input) os_non_restarting_signal(SIGALRM, data_timeout_handler);
1760
1761/* If not SMTP input, timeout happens only if configured, and we just set a
1762single timeout for the whole message. */
1763
1764else if (receive_timeout > 0)
1765 {
1766 os_non_restarting_signal(SIGALRM, data_timeout_handler);
1767 alarm(receive_timeout);
1768 }
1769
1770/* SIGTERM and SIGINT are caught always. */
1771
1772signal(SIGTERM, data_sigterm_sigint_handler);
1773signal(SIGINT, data_sigterm_sigint_handler);
1774
1775/* Header lines in messages are not supposed to be very long, though when
1776unfolded, to: and cc: headers can take up a lot of store. We must also cope
1777with the possibility of junk being thrown at us. Start by getting 256 bytes for
1778storing the header, and extend this as necessary using string_cat().
1779
1780To cope with total lunacies, impose an upper limit on the length of the header
1781section of the message, as otherwise the store will fill up. We must also cope
1782with the possibility of binary zeros in the data. Hence we cannot use fgets().
1783Folded header lines are joined into one string, leaving the '\n' characters
1784inside them, so that writing them out reproduces the input.
1785
1786Loop for each character of each header; the next structure for chaining the
1787header is set up already, with ptr the offset of the next character in
1788next->text. */
1789
1790for (;;)
1791 {
bd8fbe36 1792 int ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1793
1794 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1795 SMTP must have a correct "." terminator. */
1796
1797 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1798 {
1799 smtp_reply = handle_lost_connection(US" (header)");
1800 smtp_yield = FALSE;
1801 goto TIDYUP; /* Skip to end of function */
1802 }
1803
1804 /* See if we are at the current header's size limit - there must be at least
1805 four bytes left. This allows for the new character plus a zero, plus two for
1806 extra insertions when we are playing games with dots and carriage returns. If
1807 we are at the limit, extend the text buffer. This could have been done
1808 automatically using string_cat() but because this is a tightish loop storing
1809 only one character at a time, we choose to do it inline. Normally
1810 store_extend() will be able to extend the block; only at the end of a big
1811 store block will a copy be needed. To handle the case of very long headers
1812 (and sometimes lunatic messages can have ones that are 100s of K long) we
1813 call store_release() for strings that have been copied - if the string is at
1814 the start of a block (and therefore the only thing in it, because we aren't
4e6ae623
JH
1815 doing any other gets), the block gets freed. We can only do this release if
1816 there were no allocations since the once that we want to free. */
059ec3d9
PH
1817
1818 if (ptr >= header_size - 4)
1819 {
1820 int oldsize = header_size;
1821 /* header_size += 256; */
1822 header_size *= 2;
1823 if (!store_extend(next->text, oldsize, header_size))
1824 {
4e6ae623 1825 BOOL release_ok = store_last_get[store_pool] == next->text;
059ec3d9
PH
1826 uschar *newtext = store_get(header_size);
1827 memcpy(newtext, next->text, ptr);
4e6ae623 1828 if (release_ok) store_release(next->text);
059ec3d9
PH
1829 next->text = newtext;
1830 }
1831 }
1832
1833 /* Cope with receiving a binary zero. There is dispute about whether
1834 these should be allowed in RFC 822 messages. The middle view is that they
1835 should not be allowed in headers, at least. Exim takes this attitude at
1836 the moment. We can't just stomp on them here, because we don't know that
1837 this line is a header yet. Set a flag to cause scanning later. */
1838
1839 if (ch == 0) had_zero++;
1840
1841 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1842 those from data files use just LF. Treat LF in local SMTP input as a
1843 terminator too. Treat EOF as a line terminator always. */
1844
1845 if (ch == EOF) goto EOL;
1846
1847 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1848 other MTAs accept this. We are therefore forced into this "liberalisation"
1849 too, so we accept LF as a line terminator whatever the source of the message.
1850 However, if the first line of the message ended with a CRLF, we treat a bare
1851 LF specially by inserting a white space after it to ensure that the header
1852 line is not terminated. */
1853
1854 if (ch == '\n')
1855 {
1856 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1857 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1858 goto EOL;
1859 }
1860
1861 /* This is not the end of the line. If this is SMTP input and this is
1862 the first character in the line and it is a "." character, ignore it.
1863 This implements the dot-doubling rule, though header lines starting with
1864 dots aren't exactly common. They are legal in RFC 822, though. If the
1865 following is CRLF or LF, this is the line that that terminates the
1866 entire message. We set message_ended to indicate this has happened (to
1867 prevent further reading), and break out of the loop, having freed the
1868 empty header, and set next = NULL to indicate no data line. */
1869
178ecb70 1870 if (ptr == 0 && ch == '.' && dot_ends)
059ec3d9 1871 {
bd8fbe36 1872 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1873 if (ch == '\r')
1874 {
bd8fbe36 1875 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1876 if (ch != '\n')
1877 {
80a47a2c 1878 receive_ungetc(ch);
059ec3d9
PH
1879 ch = '\r'; /* Revert to CR */
1880 }
1881 }
1882 if (ch == '\n')
1883 {
1884 message_ended = END_DOT;
1885 store_reset(next);
1886 next = NULL;
1887 break; /* End character-reading loop */
1888 }
1889
1890 /* For non-SMTP input, the dot at the start of the line was really a data
1891 character. What is now in ch is the following character. We guaranteed
1892 enough space for this above. */
1893
1894 if (!smtp_input)
1895 {
1896 next->text[ptr++] = '.';
1897 message_size++;
1898 }
1899 }
1900
1901 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1902 remember this case if this is the first line ending. */
1903
1904 if (ch == '\r')
1905 {
bd8fbe36 1906 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1907 if (ch == '\n')
1908 {
1909 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1910 goto EOL;
1911 }
1912
1913 /* Otherwise, put back the character after CR, and turn the bare CR
1914 into LF SP. */
1915
80a47a2c 1916 ch = (receive_ungetc)(ch);
059ec3d9
PH
1917 next->text[ptr++] = '\n';
1918 message_size++;
1919 ch = ' ';
1920 }
1921
1922 /* We have a data character for the header line. */
1923
1924 next->text[ptr++] = ch; /* Add to buffer */
1925 message_size++; /* Total message size so far */
1926
1927 /* Handle failure due to a humungously long header section. The >= allows
1928 for the terminating \n. Add what we have so far onto the headers list so
1929 that it gets reflected in any error message, and back up the just-read
1930 character. */
1931
1932 if (message_size >= header_maxsize)
1933 {
1934 next->text[ptr] = 0;
1935 next->slen = ptr;
1936 next->type = htype_other;
1937 next->next = NULL;
1938 header_last->next = next;
1939 header_last = next;
1940
1941 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1942 "%s (more than %d characters): message abandoned",
1943 sender_host_unknown? sender_ident : sender_fullhost, header_maxsize);
1944
1945 if (smtp_input)
1946 {
1947 smtp_reply = US"552 Message header is ridiculously long";
1948 receive_swallow_smtp();
1949 goto TIDYUP; /* Skip to end of function */
1950 }
1951
1952 else
1953 {
1954 give_local_error(ERRMESS_VLONGHEADER,
1955 string_sprintf("message header longer than %d characters received: "
1956 "message not accepted", header_maxsize), US"", error_rc, stdin,
1957 header_list->next);
1958 /* Does not return */
1959 }
1960 }
1961
1962 continue; /* With next input character */
1963
1964 /* End of header line reached */
1965
1966 EOL:
2e0c1448
PH
1967
1968 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1969
1970 receive_linecount++;
1971 message_linecount++;
059ec3d9 1972
d677b2f2
PH
1973 /* Keep track of maximum line length */
1974
1975 if (ptr - prevlines_length > max_received_linelength)
1976 max_received_linelength = ptr - prevlines_length;
1977 prevlines_length = ptr + 1;
1978
059ec3d9
PH
1979 /* Now put in the terminating newline. There is always space for
1980 at least two more characters. */
1981
1982 next->text[ptr++] = '\n';
1983 message_size++;
1984
1985 /* A blank line signals the end of the headers; release the unwanted
1986 space and set next to NULL to indicate this. */
1987
1988 if (ptr == 1)
1989 {
1990 store_reset(next);
1991 next = NULL;
1992 break;
1993 }
1994
1995 /* There is data in the line; see if the next input character is a
1996 whitespace character. If it is, we have a continuation of this header line.
1997 There is always space for at least one character at this point. */
1998
1999 if (ch != EOF)
2000 {
bd8fbe36 2001 int nextch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
2002 if (nextch == ' ' || nextch == '\t')
2003 {
2004 next->text[ptr++] = nextch;
2005 message_size++;
2006 continue; /* Iterate the loop */
2007 }
80a47a2c 2008 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
2009 else ch = EOF; /* Cause main loop to exit at end */
2010 }
2011
2012 /* We have got to the real line end. Terminate the string and release store
2013 beyond it. If it turns out to be a real header, internal binary zeros will
2014 be squashed later. */
2015
2016 next->text[ptr] = 0;
2017 next->slen = ptr;
2018 store_reset(next->text + ptr + 1);
2019
2020 /* Check the running total size against the overall message size limit. We
2021 don't expect to fail here, but if the overall limit is set less than MESSAGE_
2022 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
2023 headers - the code to read the body will then also hit the buffer. */
2024
2025 if (message_size > thismessage_size_limit) break;
2026
2027 /* A line that is not syntactically correct for a header also marks
2028 the end of the headers. In this case, we leave next containing the
2029 first data line. This might actually be several lines because of the
2030 continuation logic applied above, but that doesn't matter.
2031
2032 It turns out that smail, and presumably sendmail, accept leading lines
2033 of the form
2034
2035 From ph10 Fri Jan 5 12:35 GMT 1996
2036
2037 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
2038 find any documentation of this, but for compatibility it had better be
2039 accepted. Exim restricts it to the case of non-smtp messages, and
2040 treats it as an alternative to the -f command line option. Thus it is
2041 ignored except for trusted users or filter testing. Otherwise it is taken
2042 as the sender address, unless -f was used (sendmail compatibility).
2043
2044 It further turns out that some UUCPs generate the From_line in a different
2045 format, e.g.
2046
2047 From ph10 Fri, 7 Jan 97 14:00:00 GMT
2048
2049 The regex for matching these things is now capable of recognizing both
2050 formats (including 2- and 4-digit years in the latter). In fact, the regex
2051 is now configurable, as is the expansion string to fish out the sender.
2052
2053 Even further on it has been discovered that some broken clients send
2054 these lines in SMTP messages. There is now an option to ignore them from
2055 specified hosts or networks. Sigh. */
2056
d21bf202
JH
2057 if ( header_last == header_list
2058 && ( !smtp_input
2059 || ( sender_host_address
2060 && verify_check_host(&ignore_fromline_hosts) == OK
2061 )
2062 || (!sender_host_address && ignore_fromline_local)
2063 )
2064 && regex_match_and_setup(regex_From, next->text, 0, -1)
2065 )
059ec3d9
PH
2066 {
2067 if (!sender_address_forced)
2068 {
2069 uschar *uucp_sender = expand_string(uucp_from_sender);
d21bf202 2070 if (!uucp_sender)
059ec3d9
PH
2071 log_write(0, LOG_MAIN|LOG_PANIC,
2072 "expansion of \"%s\" failed after matching "
2073 "\"From \" line: %s", uucp_from_sender, expand_string_message);
059ec3d9
PH
2074 else
2075 {
2076 int start, end, domain;
2077 uschar *errmess;
2078 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
2079 &start, &end, &domain, TRUE);
d21bf202 2080 if (newsender)
059ec3d9
PH
2081 {
2082 if (domain == 0 && newsender[0] != 0)
2083 newsender = rewrite_address_qualify(newsender, FALSE);
2084
f05da2e8 2085 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
2086 {
2087 sender_address = newsender;
2088
f05da2e8 2089 if (trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
2090 {
2091 authenticated_sender = NULL;
2092 originator_name = US"";
2093 sender_local = FALSE;
2094 }
2095
f05da2e8 2096 if (filter_test != FTEST_NONE)
059ec3d9
PH
2097 printf("Sender taken from \"From \" line\n");
2098 }
2099 }
2100 }
2101 }
2102 }
2103
2104 /* Not a leading "From " line. Check to see if it is a valid header line.
2105 Header names may contain any non-control characters except space and colon,
2106 amazingly. */
2107
2108 else
2109 {
2110 uschar *p = next->text;
2111
2112 /* If not a valid header line, break from the header reading loop, leaving
2113 next != NULL, indicating that it holds the first line of the body. */
2114
2115 if (isspace(*p)) break;
2116 while (mac_isgraph(*p) && *p != ':') p++;
2117 while (isspace(*p)) p++;
2118 if (*p != ':')
2119 {
2120 body_zerocount = had_zero;
2121 break;
2122 }
2123
2124 /* We have a valid header line. If there were any binary zeroes in
2125 the line, stomp on them here. */
2126
2127 if (had_zero > 0)
2128 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
2129
2130 /* It is perfectly legal to have an empty continuation line
2131 at the end of a header, but it is confusing to humans
2132 looking at such messages, since it looks like a blank line.
2133 Reduce confusion by removing redundant white space at the
2134 end. We know that there is at least one printing character
2135 (the ':' tested for above) so there is no danger of running
2136 off the end. */
2137
2138 p = next->text + ptr - 2;
2139 for (;;)
2140 {
2141 while (*p == ' ' || *p == '\t') p--;
2142 if (*p != '\n') break;
2143 ptr = (p--) - next->text + 1;
2144 message_size -= next->slen - ptr;
2145 next->text[ptr] = 0;
2146 next->slen = ptr;
2147 }
2148
2149 /* Add the header to the chain */
2150
2151 next->type = htype_other;
2152 next->next = NULL;
2153 header_last->next = next;
2154 header_last = next;
2155
2156 /* Check the limit for individual line lengths. This comes after adding to
2157 the chain so that the failing line is reflected if a bounce is generated
2158 (for a local message). */
2159
2160 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
2161 {
2162 log_write(0, LOG_MAIN, "overlong message header line received from "
2163 "%s (more than %d characters): message abandoned",
2164 sender_host_unknown? sender_ident : sender_fullhost,
2165 header_line_maxsize);
2166
2167 if (smtp_input)
2168 {
2169 smtp_reply = US"552 A message header line is too long";
2170 receive_swallow_smtp();
2171 goto TIDYUP; /* Skip to end of function */
2172 }
2173
2174 else
059ec3d9
PH
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 */
059ec3d9
PH
2180 }
2181
2182 /* Note if any resent- fields exist. */
2183
2184 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
2185 {
2186 resents_exist = TRUE;
2187 resent_prefix = US"Resent-";
2188 }
2189 }
2190
1ebe15c3
JH
2191 /* Reject CHUNKING messages that do not CRLF their first header line */
2192
2193 if (!first_line_ended_crlf && chunking_state > CHUNKING_OFFERED)
2194 {
2195 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2196 "Non-CRLF-terminated header, under CHUNKING: message abandoned",
2197 sender_address,
2198 sender_fullhost ? " H=" : "", sender_fullhost ? sender_fullhost : US"",
2199 sender_ident ? " U=" : "", sender_ident ? sender_ident : US"");
925ac8e4 2200 smtp_printf("552 Message header not CRLF terminated\r\n", FALSE);
1ebe15c3
JH
2201 bdat_flush_data();
2202 smtp_reply = US"";
2203 goto TIDYUP; /* Skip to end of function */
2204 }
2205
059ec3d9
PH
2206 /* The line has been handled. If we have hit EOF, break out of the loop,
2207 indicating no pending data line. */
2208
2209 if (ch == EOF) { next = NULL; break; }
2210
2211 /* Set up for the next header */
2212
2213 header_size = 256;
2214 next = store_get(sizeof(header_line));
2215 next->text = store_get(header_size);
2216 ptr = 0;
2217 had_zero = 0;
d677b2f2 2218 prevlines_length = 0;
059ec3d9
PH
2219 } /* Continue, starting to read the next header */
2220
2221/* At this point, we have read all the headers into a data structure in main
2222store. The first header is still the dummy placeholder for the Received: header
2223we are going to generate a bit later on. If next != NULL, it contains the first
2224data line - which terminated the headers before reaching a blank line (not the
2225normal case). */
2226
2227DEBUG(D_receive)
2228 {
2229 debug_printf(">>Headers received:\n");
1ebe15c3 2230 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2231 debug_printf("%s", h->text);
2232 debug_printf("\n");
2233 }
2234
2235/* End of file on any SMTP connection is an error. If an incoming SMTP call
2236is dropped immediately after valid headers, the next thing we will see is EOF.
2237We must test for this specially, as further down the reading of the data is
2238skipped if already at EOF. */
2239
2240if (smtp_input && (receive_feof)())
2241 {
2242 smtp_reply = handle_lost_connection(US" (after header)");
2243 smtp_yield = FALSE;
2244 goto TIDYUP; /* Skip to end of function */
2245 }
2246
2247/* If this is a filter test run and no headers were read, output a warning
2248in case there is a mistake in the test message. */
2249
f05da2e8 2250if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2251 printf("Warning: no message headers read\n");
2252
2253
2254/* Scan the headers to identify them. Some are merely marked for later
2255processing; some are dealt with here. */
2256
1ebe15c3 2257for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2258 {
2259 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2260 if (is_resent) contains_resent_headers = TRUE;
2261
2262 switch (header_checkname(h, is_resent))
2263 {
059ec3d9 2264 case htype_bcc:
2cbb4081 2265 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
059ec3d9
PH
2266 break;
2267
059ec3d9 2268 case htype_cc:
2cbb4081 2269 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
059ec3d9
PH
2270 break;
2271
2272 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
2273
2274 case htype_date:
4c69d561 2275 if (!resents_exist || is_resent) date_header_exists = TRUE;
059ec3d9
PH
2276 break;
2277
2278 /* Same comments as about Return-Path: below. */
2279
2280 case htype_delivery_date:
2281 if (delivery_date_remove) h->type = htype_old;
2282 break;
2283
2284 /* Same comments as about Return-Path: below. */
2285
2286 case htype_envelope_to:
2287 if (envelope_to_remove) h->type = htype_old;
2288 break;
2289
2290 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2291 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2292 header consists of just the login id of the user who called Exim, rewrite
2293 it with the gecos field first. Apply this rule to Resent-From: if there
2294 are resent- fields. */
2295
2296 case htype_from:
2297 h->type = htype_from;
2298 if (!resents_exist || is_resent)
2299 {
2300 from_header = h;
2301 if (!smtp_input)
2302 {
5de8faa3 2303 int len;
059ec3d9
PH
2304 uschar *s = Ustrchr(h->text, ':') + 1;
2305 while (isspace(*s)) s++;
5de8faa3 2306 len = h->slen - (s - h->text) - 1;
e0fccd1d
TF
2307 if (Ustrlen(originator_login) == len &&
2308 strncmpic(s, originator_login, len) == 0)
059ec3d9
PH
2309 {
2310 uschar *name = is_resent? US"Resent-From" : US"From";
2311 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2312 originator_login, qualify_domain_sender);
2313 from_header = header_last;
2314 h->type = htype_old;
2315 DEBUG(D_receive|D_rewrite)
2316 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2317 }
2318 }
2319 }
2320 break;
2321
2322 /* Identify the Message-id: header for generating "in-reply-to" in the
2323 autoreply transport. For incoming logging, save any resent- value. In both
2324 cases, take just the first of any multiples. */
2325
2326 case htype_id:
2327 if (msgid_header == NULL && (!resents_exist || is_resent))
2328 {
2329 msgid_header = h;
2330 h->type = htype_id;
2331 }
2332 break;
2333
2334 /* Flag all Received: headers */
2335
2336 case htype_received:
2337 h->type = htype_received;
2338 received_count++;
2339 break;
2340
2341 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
2342
2343 case htype_reply_to:
2344 h->type = htype_reply_to;
2345 break;
2346
2347 /* The Return-path: header is supposed to be added to messages when
2348 they leave the SMTP system. We shouldn't receive messages that already
2349 contain Return-path. However, since Exim generates Return-path: on
2350 local delivery, resent messages may well contain it. We therefore
2351 provide an option (which defaults on) to remove any Return-path: headers
2352 on input. Removal actually means flagging as "old", which prevents the
2353 header being transmitted with the message. */
2354
2355 case htype_return_path:
2356 if (return_path_remove) h->type = htype_old;
2357
2358 /* If we are testing a mail filter file, use the value of the
2359 Return-Path: header to set up the return_path variable, which is not
2360 otherwise set. However, remove any <> that surround the address
2361 because the variable doesn't have these. */
2362
f05da2e8 2363 if (filter_test != FTEST_NONE)
059ec3d9
PH
2364 {
2365 uschar *start = h->text + 12;
2366 uschar *end = start + Ustrlen(start);
2367 while (isspace(*start)) start++;
2368 while (end > start && isspace(end[-1])) end--;
2369 if (*start == '<' && end[-1] == '>')
2370 {
2371 start++;
2372 end--;
2373 }
2374 return_path = string_copyn(start, end - start);
2375 printf("Return-path taken from \"Return-path:\" header line\n");
2376 }
2377 break;
2378
2379 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2380 and from an untrusted caller and suppress_local_fixups is not set, or if we
2381 are in submission mode for a remote message, mark it "old" so that it will
2382 not be transmitted with the message, unless active_local_sender_retain is
2383 set. (This can only be true if active_local_from_check is false.) If there
2384 are any resent- headers in the message, apply this rule to Resent-Sender:
2385 instead of Sender:. Messages with multiple resent- header sets cannot be
2386 tidily handled. (For this reason, at least one MUA - Pine - turns old
2387 resent- headers into X-resent- headers when resending, leaving just one
2388 set.) */
059ec3d9
PH
2389
2390 case htype_sender:
69358f02 2391 h->type = ((!active_local_sender_retain &&
8800895a
PH
2392 (
2393 (sender_local && !trusted_caller && !suppress_local_fixups)
2394 || submission_mode
2395 )
059ec3d9
PH
2396 ) &&
2397 (!resents_exist||is_resent))?
2398 htype_old : htype_sender;
2399 break;
2400
2401 /* Remember the Subject: header for logging. There is no Resent-Subject */
2402
2403 case htype_subject:
2404 subject_header = h;
2405 break;
2406
2407 /* "To:" gets flagged, and the existence of a recipient header is noted,
2408 whether it's resent- or not. */
2409
2410 case htype_to:
2411 h->type = htype_to;
2412 /****
2413 to_or_cc_header_exists = TRUE;
2414 ****/
2415 break;
2416 }
2417 }
2418
2419/* Extract recipients from the headers if that is required (the -t option).
2420Note that this is documented as being done *before* any address rewriting takes
2421place. There are two possibilities:
2422
2423(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2424recipients already listed are to be REMOVED from the message. Smail 3 works
2425like this. We need to build a non-recipients tree for that list, because in
2426subsequent processing this data is held in a tree and that's what the
2427spool_write_header() function expects. Make sure that non-recipient addresses
2428are fully qualified and rewritten if necessary.
2429
2430(2) According to other sendmail documentation, -t ADDS extracted recipients to
2431those in the command line arguments (and it is rumoured some other MTAs do
2432this). Therefore, there is an option to make Exim behave this way.
2433
2434*** Notes on "Resent-" header lines ***
2435
2436The presence of resent-headers in the message makes -t horribly ambiguous.
2437Experiments with sendmail showed that it uses recipients for all resent-
2438headers, totally ignoring the concept of "sets of resent- headers" as described
2439in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2440with all the addresses in one instance of each header.
2441
2442This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2443error for -t if there were resent- headers in the message. However, after a
2444discussion on the mailing list, I've learned that there are MUAs that use
2445resent- headers with -t, and also that the stuff about sets of resent- headers
2446and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2447message with -t and resent- header lines makes sure that only *its* resent-
2448headers are present; previous ones are often renamed as X-resent- for example.
2449
2450Consequently, Exim has been changed so that, if any resent- header lines are
2451present, the recipients are taken from all of the appropriate resent- lines,
2452and not from the ordinary To:, Cc:, etc. */
2453
2454if (extract_recip)
2455 {
2456 int rcount = 0;
2457 error_block **bnext = &bad_addresses;
2458
2459 if (extract_addresses_remove_arguments)
2460 {
2461 while (recipients_count-- > 0)
2462 {
2463 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2464 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2465 tree_add_nonrecipient(s);
2466 }
2467 recipients_list = NULL;
2468 recipients_count = recipients_list_max = 0;
2469 }
2470
059ec3d9
PH
2471 /* Now scan the headers */
2472
1ebe15c3 2473 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2474 {
2475 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2476 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2477 {
2478 uschar *s = Ustrchr(h->text, ':') + 1;
2479 while (isspace(*s)) s++;
2480
1eccaa59
PH
2481 parse_allow_group = TRUE; /* Allow address group syntax */
2482
059ec3d9
PH
2483 while (*s != 0)
2484 {
2485 uschar *ss = parse_find_address_end(s, FALSE);
2486 uschar *recipient, *errmess, *p, *pp;
2487 int start, end, domain;
2488
2489 /* Check on maximum */
2490
2491 if (recipients_max > 0 && ++rcount > recipients_max)
2492 {
2493 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2494 US"message rejected: ", error_rc, stdin, NULL);
2495 /* Does not return */
2496 }
2497
2498 /* Make a copy of the address, and remove any internal newlines. These
2499 may be present as a result of continuations of the header line. The
2500 white space that follows the newline must not be removed - it is part
2501 of the header. */
2502
2503 pp = recipient = store_get(ss - s + 1);
2504 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2505 *pp = 0;
250b6871 2506
8c5d388a 2507#ifdef SUPPORT_I18N
250b6871
JH
2508 {
2509 BOOL b = allow_utf8_domains;
2510 allow_utf8_domains = TRUE;
2511#endif
059ec3d9
PH
2512 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2513 &domain, FALSE);
2514
8c5d388a 2515#ifdef SUPPORT_I18N
250b6871
JH
2516 if (string_is_utf8(recipient))
2517 message_smtputf8 = TRUE;
2518 else
2519 allow_utf8_domains = b;
2520 }
2521#endif
2522
059ec3d9
PH
2523 /* Keep a list of all the bad addresses so we can send a single
2524 error message at the end. However, an empty address is not an error;
2525 just ignore it. This can come from an empty group list like
2526
2527 To: Recipients of list:;
2528
2529 If there are no recipients at all, an error will occur later. */
2530
2531 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2532 {
2533 int len = Ustrlen(s);
2534 error_block *b = store_get(sizeof(error_block));
2535 while (len > 0 && isspace(s[len-1])) len--;
2536 b->next = NULL;
2537 b->text1 = string_printing(string_copyn(s, len));
2538 b->text2 = errmess;
2539 *bnext = b;
2540 bnext = &(b->next);
2541 }
2542
2543 /* If the recipient is already in the nonrecipients tree, it must
2544 have appeared on the command line with the option extract_addresses_
2545 remove_arguments set. Do not add it to the recipients, and keep a note
2546 that this has happened, in order to give a better error if there are
2547 no recipients left. */
2548
2549 else if (recipient != NULL)
2550 {
2551 if (tree_search(tree_nonrecipients, recipient) == NULL)
2552 receive_add_recipient(recipient, -1);
2553 else
2554 extracted_ignored = TRUE;
2555 }
2556
2557 /* Move on past this address */
2558
2559 s = ss + (*ss? 1:0);
2560 while (isspace(*s)) s++;
1eccaa59
PH
2561 } /* Next address */
2562
2563 parse_allow_group = FALSE; /* Reset group syntax flags */
2564 parse_found_group = FALSE;
059ec3d9
PH
2565
2566 /* If this was the bcc: header, mark it "old", which means it
2567 will be kept on the spool, but not transmitted as part of the
2568 message. */
2569
2cbb4081 2570 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2571 } /* For appropriate header line */
2572 } /* For each header line */
2573
059ec3d9
PH
2574 }
2575
2576/* Now build the unique message id. This has changed several times over the
2577lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2578Retaining all the history in the comment has become too unwieldy - read
2579previous release sources if you want it.
2580
2581The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2582The first part is the current time, in seconds. The second part is the current
2583pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2584can hold a number in the range 0-3843. It used to be a computed sequence
2585number, but is now the fractional component of the current time in units of
25861/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2587received, Exim ensures that the timer has ticked at the appropriate level
2588before proceeding, to avoid duplication if the pid happened to be re-used
2589within the same time period. It seems likely that most messages will take at
2590least half a millisecond to be received, so no delay will normally be
2591necessary. At least for some time...
2592
2593There is a modification when localhost_number is set. Formerly this was allowed
2594to be as large as 255. Now it is restricted to the range 0-16, and the final
2595component of the message id becomes (localhost_number * 200) + fractional time
2596in units of 1/200 of a second (i.e. a value in the range 0-3399).
2597
2598Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2599Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2600still allows the tttttt field to hold a large enough number to last for some
2601more decades, and the final two-digit field can hold numbers up to 1295, which
2602is enough for milliseconds (instead of 1/2000 of a second).
2603
2604However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2605pid, so it is probably safe because pids have to be positive. The
2606localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2607final field becomes (localhost_number * 100) + fractional time in centiseconds.
2608
2609Note that string_base62() returns its data in a static storage block, so it
2610must be copied before calling string_base62() again. It always returns exactly
26116 characters.
2612
2613There doesn't seem to be anything in the RFC which requires a message id to
2614start with a letter, but Smail was changed to ensure this. The external form of
2615the message id (as supplied by string expansion) therefore starts with an
2616additional leading 'E'. The spool file names do not include this leading
2617letter and it is not used internally.
2618
2619NOTE: If ever the format of message ids is changed, the regular expression for
2620checking that a string is in this format must be updated in a corresponding
2621way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2540f2f8
JH
2622must also be changed to reflect the correct string length. The queue-sort code
2623needs to know the layout. Then, of course, other programs that rely on the
2624message id format will need updating too. */
059ec3d9
PH
2625
2626Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2627message_id[6] = '-';
2628Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2629
2630/* Deal with the case where the host number is set. The value of the number was
2631checked when it was read, to ensure it isn't too big. The timing granularity is
2632left in id_resolution so that an appropriate wait can be done after receiving
2633the message, if necessary (we hope it won't be). */
2634
2635if (host_number_string != NULL)
2636 {
2637 id_resolution = (BASE_62 == 62)? 5000 : 10000;
2638 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2639 string_base62((long int)(
2640 host_number * (1000000/id_resolution) +
2641 message_id_tv.tv_usec/id_resolution)) + 4);
2642 }
2643
2644/* Host number not set: final field is just the fractional time at an
2645appropriate resolution. */
2646
2647else
2648 {
2649 id_resolution = (BASE_62 == 62)? 500 : 1000;
2650 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2651 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2652 }
2653
2654/* Add the current message id onto the current process info string if
2655it will fit. */
2656
2657(void)string_format(process_info + process_info_len,
2658 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2659
2660/* If we are using multiple input directories, set up the one for this message
2661to be the least significant base-62 digit of the time of arrival. Otherwise
2662ensure that it is an empty string. */
2663
a2da3176 2664message_subdir[0] = split_spool_directory ? message_id[5] : 0;
059ec3d9
PH
2665
2666/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2667one, but only for local (without suppress_local_fixups) or submission mode
2668messages. This can be user-configured if required, but we had better flatten
2669any illegal characters therein. */
059ec3d9 2670
8800895a
PH
2671if (msgid_header == NULL &&
2672 ((sender_host_address == NULL && !suppress_local_fixups)
2673 || submission_mode))
059ec3d9
PH
2674 {
2675 uschar *p;
2676 uschar *id_text = US"";
2677 uschar *id_domain = primary_hostname;
2678
2679 /* Permit only letters, digits, dots, and hyphens in the domain */
2680
2681 if (message_id_domain != NULL)
2682 {
2683 uschar *new_id_domain = expand_string(message_id_domain);
2684 if (new_id_domain == NULL)
2685 {
2686 if (!expand_string_forcedfail)
2687 log_write(0, LOG_MAIN|LOG_PANIC,
2688 "expansion of \"%s\" (message_id_header_domain) "
2689 "failed: %s", message_id_domain, expand_string_message);
2690 }
2691 else if (*new_id_domain != 0)
2692 {
2693 id_domain = new_id_domain;
2694 for (p = id_domain; *p != 0; p++)
2695 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2696 }
2697 }
2698
2699 /* Permit all characters except controls and RFC 2822 specials in the
2700 additional text part. */
2701
2702 if (message_id_text != NULL)
2703 {
2704 uschar *new_id_text = expand_string(message_id_text);
2705 if (new_id_text == NULL)
2706 {
2707 if (!expand_string_forcedfail)
2708 log_write(0, LOG_MAIN|LOG_PANIC,
2709 "expansion of \"%s\" (message_id_header_text) "
2710 "failed: %s", message_id_text, expand_string_message);
2711 }
2712 else if (*new_id_text != 0)
2713 {
2714 id_text = new_id_text;
2715 for (p = id_text; *p != 0; p++)
2716 if (mac_iscntrl_or_special(*p)) *p = '-';
2717 }
2718 }
2719
e7e680d6
PP
2720 /* Add the header line
2721 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2722 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2723
e7e680d6 2724 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2725 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2726 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2727 }
2728
2729/* If we are to log recipients, keep a copy of the raw ones before any possible
2730rewriting. Must copy the count, because later ACLs and the local_scan()
2731function may mess with the real recipients. */
2732
6c6d6e48 2733if (LOGGING(received_recipients))
059ec3d9
PH
2734 {
2735 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2736 for (i = 0; i < recipients_count; i++)
2737 raw_recipients[i] = string_copy(recipients_list[i].address);
2738 raw_recipients_count = recipients_count;
2739 }
2740
2741/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2742recipients will get here only if the conditions were right (allow_unqualified_
2743recipient is TRUE). */
2744
2745for (i = 0; i < recipients_count; i++)
2746 recipients_list[i].address =
2747 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2748 global_rewrite_rules, rewrite_existflags);
2749
8800895a
PH
2750/* If there is no From: header, generate one for local (without
2751suppress_local_fixups) or submission_mode messages. If there is no sender
2752address, but the sender is local or this is a local delivery error, use the
2753originator login. This shouldn't happen for genuine bounces, but might happen
2754for autoreplies. The addition of From: must be done *before* checking for the
2755possible addition of a Sender: header, because untrusted_set_sender allows an
2756untrusted user to set anything in the envelope (which might then get info
2757From:) but we still want to ensure a valid Sender: if it is required. */
2758
2759if (from_header == NULL &&
2760 ((sender_host_address == NULL && !suppress_local_fixups)
2761 || submission_mode))
059ec3d9 2762 {
2fe1a124
PH
2763 uschar *oname = US"";
2764
2765 /* Use the originator_name if this is a locally submitted message and the
2766 caller is not trusted. For trusted callers, use it only if -F was used to
2767 force its value or if we have a non-SMTP message for which -f was not used
2768 to set the sender. */
2769
2770 if (sender_host_address == NULL)
2771 {
2772 if (!trusted_caller || sender_name_forced ||
2773 (!smtp_input && !sender_address_forced))
2774 oname = originator_name;
2775 }
2776
2777 /* For non-locally submitted messages, the only time we use the originator
2778 name is when it was forced by the /name= option on control=submission. */
2779
2780 else
2781 {
2782 if (submission_name != NULL) oname = submission_name;
2783 }
2784
059ec3d9
PH
2785 /* Envelope sender is empty */
2786
2787 if (sender_address[0] == 0)
2788 {
87ba3f5f
PH
2789 uschar *fromstart, *fromend;
2790
2791 fromstart = string_sprintf("%sFrom: %s%s", resent_prefix,
2fe1a124
PH
2792 oname, (oname[0] == 0)? "" : " <");
2793 fromend = (oname[0] == 0)? US"" : US">";
87ba3f5f 2794
059ec3d9
PH
2795 if (sender_local || local_error_message)
2796 {
87ba3f5f
PH
2797 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2798 local_part_quote(originator_login), qualify_domain_sender,
2799 fromend);
059ec3d9
PH
2800 }
2801 else if (submission_mode && authenticated_id != NULL)
2802 {
2803 if (submission_domain == NULL)
2804 {
87ba3f5f
PH
2805 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2806 local_part_quote(authenticated_id), qualify_domain_sender,
2807 fromend);
059ec3d9
PH
2808 }
2809 else if (submission_domain[0] == 0) /* empty => whole address set */
2810 {
87ba3f5f
PH
2811 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2812 fromend);
059ec3d9
PH
2813 }
2814 else
2815 {
87ba3f5f
PH
2816 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2817 local_part_quote(authenticated_id), submission_domain,
2818 fromend);
059ec3d9
PH
2819 }
2820 from_header = header_last; /* To get it checked for Sender: */
2821 }
2822 }
2823
2824 /* There is a non-null envelope sender. Build the header using the original
2825 sender address, before any rewriting that might have been done while
2826 verifying it. */
2827
2828 else
2829 {
87ba3f5f 2830 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124
PH
2831 oname,
2832 (oname[0] == 0)? "" : " <",
87ba3f5f
PH
2833 (sender_address_unrewritten == NULL)?
2834 sender_address : sender_address_unrewritten,
2fe1a124 2835 (oname[0] == 0)? "" : ">");
059ec3d9
PH
2836
2837 from_header = header_last; /* To get it checked for Sender: */
2838 }
2839 }
2840
2841
8800895a
PH
2842/* If the sender is local (without suppress_local_fixups), or if we are in
2843submission mode and there is an authenticated_id, check that an existing From:
2844is correct, and if not, generate a Sender: header, unless disabled. Any
2845previously-existing Sender: header was removed above. Note that sender_local,
2846as well as being TRUE if the caller of exim is not trusted, is also true if a
2847trusted caller did not supply a -f argument for non-smtp input. To allow
2848trusted callers to forge From: without supplying -f, we have to test explicitly
2849here. If the From: header contains more than one address, then the call to
2850parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9
PH
2851
2852if (from_header != NULL &&
69358f02 2853 (active_local_from_check &&
8800895a 2854 ((sender_local && !trusted_caller && !suppress_local_fixups) ||
69358f02 2855 (submission_mode && authenticated_id != NULL))
059ec3d9
PH
2856 ))
2857 {
2858 BOOL make_sender = TRUE;
2859 int start, end, domain;
2860 uschar *errmess;
2861 uschar *from_address =
2862 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2863 &start, &end, &domain, FALSE);
2864 uschar *generated_sender_address;
2865
2866 if (submission_mode)
2867 {
2868 if (submission_domain == NULL)
2869 {
2870 generated_sender_address = string_sprintf("%s@%s",
2871 local_part_quote(authenticated_id), qualify_domain_sender);
2872 }
2873 else if (submission_domain[0] == 0) /* empty => full address */
2874 {
2875 generated_sender_address = string_sprintf("%s",
2876 authenticated_id);
2877 }
2878 else
2879 {
2880 generated_sender_address = string_sprintf("%s@%s",
2881 local_part_quote(authenticated_id), submission_domain);
2882 }
2883 }
2884 else
2885 generated_sender_address = string_sprintf("%s@%s",
2886 local_part_quote(originator_login), qualify_domain_sender);
2887
2888 /* Remove permitted prefixes and suffixes from the local part of the From:
2889 address before doing the comparison with the generated sender. */
2890
2891 if (from_address != NULL)
2892 {
2893 int slen;
2894 uschar *at = (domain == 0)? NULL : from_address + domain - 1;
2895
2896 if (at != NULL) *at = 0;
2897 from_address += route_check_prefix(from_address, local_from_prefix);
2898 slen = route_check_suffix(from_address, local_from_suffix);
2899 if (slen > 0)
2900 {
2901 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2902 from_address += slen;
2903 }
2904 if (at != NULL) *at = '@';
2905
2906 if (strcmpic(generated_sender_address, from_address) == 0 ||
2907 (domain == 0 && strcmpic(from_address, originator_login) == 0))
2908 make_sender = FALSE;
2909 }
2910
2911 /* We have to cause the Sender header to be rewritten if there are
2912 appropriate rewriting rules. */
2913
2914 if (make_sender)
2915 {
2fe1a124 2916 if (submission_mode && submission_name == NULL)
059ec3d9
PH
2917 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2918 generated_sender_address);
2919 else
2920 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124
PH
2921 resent_prefix,
2922 submission_mode? submission_name : originator_name,
2923 generated_sender_address);
059ec3d9 2924 }
87ba3f5f
PH
2925
2926 /* Ensure that a non-null envelope sender address corresponds to the
2927 submission mode sender address. */
2928
2929 if (submission_mode && sender_address[0] != 0)
2930 {
2931 if (sender_address_unrewritten == NULL)
2932 sender_address_unrewritten = sender_address;
2933 sender_address = generated_sender_address;
089793a4
TF
2934 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2935 log_write(L_address_rewrite, LOG_MAIN,
2936 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2937 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2938 }
059ec3d9
PH
2939 }
2940
059ec3d9
PH
2941/* If there are any rewriting rules, apply them to the sender address, unless
2942it has already been rewritten as part of verification for SMTP input. */
2943
2944if (global_rewrite_rules != NULL && sender_address_unrewritten == NULL &&
2945 sender_address[0] != 0)
2946 {
2947 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2948 global_rewrite_rules, rewrite_existflags);
2949 DEBUG(D_receive|D_rewrite)
2950 debug_printf("rewritten sender = %s\n", sender_address);
2951 }
2952
2953
2954/* The headers must be run through rewrite_header(), because it ensures that
2955addresses are fully qualified, as well as applying any rewriting rules that may
2956exist.
2957
2958Qualification of header addresses in a message from a remote host happens only
2959if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2960appropriate. For local messages, qualification always happens, unless -bnq is
2961used to explicitly suppress it. No rewriting is done for an unqualified address
2962that is left untouched.
2963
2964We start at the second header, skipping our own Received:. This rewriting is
2965documented as happening *after* recipient addresses are taken from the headers
2966by the -t command line option. An added Sender: gets rewritten here. */
2967
1ebe15c3 2968for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2969 {
2970 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2971 rewrite_existflags, TRUE);
1ebe15c3 2972 if (newh) h = newh;
059ec3d9
PH
2973 }
2974
2975
2976/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2977"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2978"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2979exists, this applies to the set of resent- headers rather than the normal set.
2980
2cbb4081
PH
2981The requirement for a recipient header has been removed in RFC 2822. At this
2982point in the code, earlier versions of Exim added a To: header for locally
2983submitted messages, and an empty Bcc: header for others. In the light of the
2984changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2985
059ec3d9
PH
2986
2987/* If there is no date header, generate one if the message originates locally
8800895a
PH
2988(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2989submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2990to be more confusing if Exim adds one to all remotely-originated messages.
2991As per Message-Id, we prepend if resending, else append.
2992*/
059ec3d9 2993
8800895a
PH
2994if (!date_header_exists &&
2995 ((sender_host_address == NULL && !suppress_local_fixups)
2996 || submission_mode))
e7e680d6
PP
2997 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2998 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2999
3000search_tidyup(); /* Free any cached resources */
3001
3002/* Show the complete set of headers if debugging. Note that the first one (the
3003new Received:) has not yet been set. */
3004
3005DEBUG(D_receive)
3006 {
3007 debug_printf(">>Headers after rewriting and local additions:\n");
3008 for (h = header_list->next; h != NULL; h = h->next)
3009 debug_printf("%c %s", h->type, h->text);
3010 debug_printf("\n");
3011 }
3012
3013/* The headers are now complete in store. If we are running in filter
3014testing mode, that is all this function does. Return TRUE if the message
3015ended with a dot. */
3016
f05da2e8 3017if (filter_test != FTEST_NONE)
059ec3d9
PH
3018 {
3019 process_info[process_info_len] = 0;
3020 return message_ended == END_DOT;
3021 }
3022
7e3ce68e
JH
3023/*XXX CHUNKING: need to cancel cutthrough under BDAT, for now. In future,
3024think more if it could be handled. Cannot do onward CHUNKING unless
3025inbound is, but inbound chunking ought to be ok with outbound plain.
3026Could we do onward CHUNKING given inbound CHUNKING?
3027*/
3028if (chunking_state > CHUNKING_OFFERED)
57cc2785 3029 cancel_cutthrough_connection(FALSE, US"chunking active");
7e3ce68e 3030
817d9f57 3031/* Cutthrough delivery:
5032d1cf
JH
3032We have to create the Received header now rather than at the end of reception,
3033so the timestamp behaviour is a change to the normal case.
5032d1cf 3034Having created it, send the headers to the destination. */
57cc2785
JH
3035
3036if (cutthrough.fd >= 0 && cutthrough.delivery)
e4bdf652 3037 {
817d9f57
JH
3038 if (received_count > received_headers_max)
3039 {
57cc2785 3040 cancel_cutthrough_connection(TRUE, US"too many headers");
817d9f57
JH
3041 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
3042 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3043 "Too many \"Received\" headers",
3044 sender_address,
57cc2785
JH
3045 sender_fullhost ? "H=" : "", sender_fullhost ? sender_fullhost : US"",
3046 sender_ident ? "U=" : "", sender_ident ? sender_ident : US"");
817d9f57
JH
3047 message_id[0] = 0; /* Indicate no message accepted */
3048 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
3049 goto TIDYUP; /* Skip to end of function */
3050 }
e4bdf652 3051 received_header_gen();
578d43dc 3052 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652
JH
3053 (void) cutthrough_headers_send();
3054 }
61147df4 3055
e4bdf652 3056
059ec3d9
PH
3057/* Open a new spool file for the data portion of the message. We need
3058to access it both via a file descriptor and a stream. Try to make the
41313d92 3059directory if it isn't there. */
059ec3d9 3060
41313d92 3061spool_name = spool_fname(US"input", message_subdir, message_id, US"-D");
a2da3176
JH
3062DEBUG(D_receive) debug_printf("Data file name: %s\n", spool_name);
3063
3064if ((data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE)) < 0)
059ec3d9
PH
3065 {
3066 if (errno == ENOENT)
3067 {
0971ec06 3068 (void) directory_make(spool_directory,
41313d92
JH
3069 spool_sname(US"input", message_subdir),
3070 INPUT_DIRECTORY_MODE, TRUE);
059ec3d9
PH
3071 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
3072 }
3073 if (data_fd < 0)
3074 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
3075 spool_name, strerror(errno));
3076 }
3077
3078/* Make sure the file's group is the Exim gid, and double-check the mode
3079because the group setting doesn't always get set automatically. */
3080
1ac6b2e7
JH
3081if (fchown(data_fd, exim_uid, exim_gid))
3082 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
3083 "Failed setting ownership on spool file %s: %s",
3084 spool_name, strerror(errno));
ff790e47 3085(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
3086
3087/* We now have data file open. Build a stream for it and lock it. We lock only
3088the first line of the file (containing the message ID) because otherwise there
3089are problems when Exim is run under Cygwin (I'm told). See comments in
3090spool_in.c, where the same locking is done. */
3091
3092data_file = fdopen(data_fd, "w+");
3093lock_data.l_type = F_WRLCK;
3094lock_data.l_whence = SEEK_SET;
3095lock_data.l_start = 0;
3096lock_data.l_len = SPOOL_DATA_START_OFFSET;
3097
3098if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
3099 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
3100 errno, strerror(errno));
3101
3102/* We have an open, locked data file. Write the message id to it to make it
3103self-identifying. Then read the remainder of the input of this message and
3104write it to the data file. If the variable next != NULL, it contains the first
3105data line (which was read as a header but then turned out not to have the right
3106format); write it (remembering that it might contain binary zeros). The result
3107of fwrite() isn't inspected; instead we call ferror() below. */
3108
3109fprintf(data_file, "%s-D\n", message_id);
3110if (next != NULL)
3111 {
3112 uschar *s = next->text;
3113 int len = next->slen;
1ac6b2e7 3114 len = fwrite(s, 1, len, data_file); len = len; /* compiler quietening */
059ec3d9
PH
3115 body_linecount++; /* Assumes only 1 line */
3116 }
3117
3118/* Note that we might already be at end of file, or the logical end of file
3119(indicated by '.'), or might have encountered an error while writing the
3120message id or "next" line. */
3121
3122if (!ferror(data_file) && !(receive_feof)() && message_ended != END_DOT)
3123 {
3124 if (smtp_input)
3125 {
328c5688
JH
3126 message_ended = chunking_state <= CHUNKING_OFFERED
3127 ? read_message_data_smtp(data_file)
3128 : spool_wireformat
3129 ? read_message_bdat_smtp_wire(data_file)
3130 : read_message_bdat_smtp(data_file);
059ec3d9
PH
3131 receive_linecount++; /* The terminating "." line */
3132 }
3133 else message_ended = read_message_data(data_file);
3134
3135 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 3136 message_linecount += body_linecount;
059ec3d9 3137
7e3ce68e 3138 switch (message_ended)
059ec3d9 3139 {
7e3ce68e 3140 /* Handle premature termination of SMTP */
059ec3d9 3141
7e3ce68e
JH
3142 case END_EOF:
3143 if (smtp_input)
3144 {
3145 Uunlink(spool_name); /* Lose data file when closed */
57cc2785 3146 cancel_cutthrough_connection(TRUE, US"sender closed connection");
7e3ce68e
JH
3147 message_id[0] = 0; /* Indicate no message accepted */
3148 smtp_reply = handle_lost_connection(US"");
3149 smtp_yield = FALSE;
3150 goto TIDYUP; /* Skip to end of function */
3151 }
3152 break;
059ec3d9 3153
7e3ce68e
JH
3154 /* Handle message that is too big. Don't use host_or_ident() in the log
3155 message; we want to see the ident value even for non-remote messages. */
059ec3d9 3156
7e3ce68e
JH
3157 case END_SIZE:
3158 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3159 cancel_cutthrough_connection(TRUE, US"mail too big");
7e3ce68e 3160 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
059ec3d9 3161
7e3ce68e
JH
3162 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3163 "message too big: read=%d max=%d",
3164 sender_address,
3165 (sender_fullhost == NULL)? "" : " H=",
3166 (sender_fullhost == NULL)? US"" : sender_fullhost,
3167 (sender_ident == NULL)? "" : " U=",
3168 (sender_ident == NULL)? US"" : sender_ident,
3169 message_size,
3170 thismessage_size_limit);
3171
3172 if (smtp_input)
3173 {
3174 smtp_reply = US"552 Message size exceeds maximum permitted";
3175 message_id[0] = 0; /* Indicate no message accepted */
3176 goto TIDYUP; /* Skip to end of function */
3177 }
3178 else
3179 {
3180 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3181 give_local_error(ERRMESS_TOOBIG,
3182 string_sprintf("message too big (max=%d)", thismessage_size_limit),
3183 US"message rejected: ", error_rc, data_file, header_list);
3184 /* Does not return */
3185 }
3186 break;
3187
3188 /* Handle bad BDAT protocol sequence */
3189
3190 case END_PROTOCOL:
3191 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3192 cancel_cutthrough_connection(TRUE, US"sender protocol error");
7e3ce68e
JH
3193 smtp_reply = US""; /* Response already sent */
3194 message_id[0] = 0; /* Indicate no message accepted */
3195 goto TIDYUP; /* Skip to end of function */
059ec3d9
PH
3196 }
3197 }
3198
3199/* Restore the standard SIGALRM handler for any subsequent processing. (For
3200example, there may be some expansion in an ACL that uses a timer.) */
3201
3202os_non_restarting_signal(SIGALRM, sigalrm_handler);
3203
3204/* The message body has now been read into the data file. Call fflush() to
3205empty the buffers in C, and then call fsync() to get the data written out onto
3206the disk, as fflush() doesn't do this (or at least, it isn't documented as
3207having to do this). If there was an I/O error on either input or output,
3208attempt to send an error message, and unlink the spool file. For non-SMTP input
3209we can then give up. Note that for SMTP input we must swallow the remainder of
3210the input in cases of output errors, since the far end doesn't expect to see
3211anything until the terminating dot line is sent. */
3212
3213if (fflush(data_file) == EOF || ferror(data_file) ||
54fc8428 3214 EXIMfsync(fileno(data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
3215 {
3216 uschar *msg_errno = US strerror(errno);
3217 BOOL input_error = (receive_ferror)() != 0;
3218 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
3219 input_error? "Input read" : "Spool write",
3220 msg_errno,
3221 (sender_fullhost != NULL)? sender_fullhost : sender_ident);
3222
3223 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
3224 Uunlink(spool_name); /* Lose the data file */
57cc2785 3225 cancel_cutthrough_connection(TRUE, US"error writing spoolfile");
059ec3d9
PH
3226
3227 if (smtp_input)
3228 {
3229 if (input_error)
3230 smtp_reply = US"451 Error while reading input data";
3231 else
3232 {
3233 smtp_reply = US"451 Error while writing spool file";
3234 receive_swallow_smtp();
3235 }
3236 message_id[0] = 0; /* Indicate no message accepted */
3237 goto TIDYUP; /* Skip to end of function */
3238 }
3239
3240 else
3241 {
3242 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3243 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, data_file,
3244 header_list);
3245 /* Does not return */
3246 }
3247 }
3248
3249
3250/* No I/O errors were encountered while writing the data file. */
3251
3252DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
3253
3254
3255/* If there were any bad addresses extracted by -t, or there were no recipients
3256left after -t, send a message to the sender of this message, or write it to
3257stderr if the error handling option is set that way. Note that there may
3258legitimately be no recipients for an SMTP message if they have all been removed
3259by "discard".
3260
3261We need to rewind the data file in order to read it. In the case of no
3262recipients or stderr error writing, throw the data file away afterwards, and
3263exit. (This can't be SMTP, which always ensures there's at least one
3264syntactically good recipient address.) */
3265
3266if (extract_recip && (bad_addresses != NULL || recipients_count == 0))
3267 {
3268 DEBUG(D_receive)
3269 {
3270 if (recipients_count == 0) debug_printf("*** No recipients\n");
3271 if (bad_addresses != NULL)
3272 {
3273 error_block *eblock = bad_addresses;
3274 debug_printf("*** Bad address(es)\n");
3275 while (eblock != NULL)
3276 {
3277 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
3278 eblock = eblock->next;
3279 }
3280 }
3281 }
3282
3283 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3284
3285 /* If configured to send errors to the sender, but this fails, force
3286 a failure error code. We use a special one for no recipients so that it
3287 can be detected by the autoreply transport. Otherwise error_rc is set to
3288 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
3289 it is EXIT_SUCCESS. */
3290
3291 if (error_handling == ERRORS_SENDER)
3292 {
3293 if (!moan_to_sender(
3294 (bad_addresses == NULL)?
3295 (extracted_ignored? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS) :
3296 (recipients_list == NULL)? ERRMESS_BADNOADDRESS : ERRMESS_BADADDRESS,
3297 bad_addresses, header_list, data_file, FALSE))
3298 error_rc = (bad_addresses == NULL)? EXIT_NORECIPIENTS : EXIT_FAILURE;
3299 }
3300 else
3301 {
3302 if (bad_addresses == NULL)
3303 {
3304 if (extracted_ignored)
3305 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3306 else
3307 fprintf(stderr, "exim: no recipients in message\n");
3308 }
3309 else
3310 {
3311 fprintf(stderr, "exim: invalid address%s",
3312 (bad_addresses->next == NULL)? ":" : "es:\n");
3313 while (bad_addresses != NULL)
3314 {
3315 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3316 bad_addresses->text2);
3317 bad_addresses = bad_addresses->next;
3318 }
3319 }
3320 }
3321
3322 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3323 {
3324 Uunlink(spool_name);
f1e894f3 3325 (void)fclose(data_file);
9bfb7e1b 3326 exim_exit(error_rc, US"receiving");
059ec3d9
PH
3327 }
3328 }
3329
3330/* Data file successfully written. Generate text for the Received: header by
3331expanding the configured string, and adding a timestamp. By leaving this
3332operation till now, we ensure that the timestamp is the time that message
3333reception was completed. However, this is deliberately done before calling the
3334data ACL and local_scan().
3335
3336This Received: header may therefore be inspected by the data ACL and by code in
3337the local_scan() function. When they have run, we update the timestamp to be
3338the final time of reception.
3339
3340If there is just one recipient, set up its value in the $received_for variable
3341for use when we generate the Received: header.
3342
3343Note: the checking for too many Received: headers is handled by the delivery
3344code. */
e4bdf652 3345/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3346
e4bdf652 3347if (received_header->text == NULL) /* Non-cutthrough case */
059ec3d9 3348 {
e4bdf652 3349 received_header_gen();
059ec3d9 3350
e4bdf652 3351 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3352
e4bdf652
JH
3353 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3354 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3355
e4bdf652
JH
3356 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3357 now, before running the DATA ACL. */
059ec3d9 3358
578d43dc 3359 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652 3360 }
817d9f57 3361else
e4bdf652
JH
3362 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3363 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3364
3365/* If an ACL is specified for checking things at this stage of reception of a
3366message, run it, unless all the recipients were removed by "discard" in earlier
3367ACLs. That is the only case in which recipients_count can be zero at this
3368stage. Set deliver_datafile to point to the data file so that $message_body and
3369$message_body_end can be extracted if needed. Allow $recipients in expansions.
3370*/
3371
3372deliver_datafile = data_fd;
4e88a19f 3373user_msg = NULL;
059ec3d9 3374
0e20aff9
MH
3375enable_dollar_recipients = TRUE;
3376
059ec3d9 3377if (recipients_count == 0)
7e3ce68e
JH
3378 blackholed_by = recipients_discarded ? US"MAIL ACL" : US"RCPT ACL";
3379
059ec3d9
PH
3380else
3381 {
059ec3d9
PH
3382 /* Handle interactive SMTP messages */
3383
3384 if (smtp_input && !smtp_batched_input)
3385 {
8523533c 3386
80a47a2c
TK
3387#ifndef DISABLE_DKIM
3388 if (!dkim_disable_verify)
3389 {
cc55f420 3390 /* Finish verification */
80a47a2c
TK
3391 dkim_exim_verify_finish();
3392
3393 /* Check if we must run the DKIM ACL */
7e3ce68e 3394 if (acl_smtp_dkim && dkim_verify_signers && *dkim_verify_signers)
80a47a2c 3395 {
cc55f420 3396 uschar * dkim_verify_signers_expanded =
80a47a2c 3397 expand_string(dkim_verify_signers);
cc55f420
JH
3398 gstring * results = NULL;
3399 int signer_sep = 0;
3400 const uschar * ptr;
3401 uschar * item;
3402 gstring * seen_items = NULL;
3403 int old_pool = store_pool;
3404
3405 store_pool = POOL_PERM; /* Allow created variables to live to data ACL */
3406
3407 if (!(ptr = dkim_verify_signers_expanded))
80a47a2c
TK
3408 log_write(0, LOG_MAIN|LOG_PANIC,
3409 "expansion of dkim_verify_signers option failed: %s",
3410 expand_string_message);
7e3ce68e 3411
cc55f420
JH
3412 /* Default to OK when no items are present */
3413 rc = OK;
3414 while ((item = string_nextinlist(&ptr, &signer_sep, NULL, 0)))
3415 {
3416 /* Prevent running ACL for an empty item */
3417 if (!item || !*item) continue;
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)
3422 {
3423 uschar * seen_item;
3424 const uschar * seen_items_list = string_from_gstring(seen_items);
3425 int seen_sep = ':';
3426 BOOL seen_this_item = FALSE;
3427
3428 while ((seen_item = string_nextinlist(&seen_items_list, &seen_sep,
3429 NULL, 0)))
3430 if (Ustrcmp(seen_item,item) == 0)
3431 {
3432 seen_this_item = TRUE;
3433 break;
3434 }
3435
3436 if (seen_this_item)
5032d1cf
JH
3437 {
3438 DEBUG(D_receive)
cc55f420
JH
3439 debug_printf("acl_smtp_dkim: skipping signer %s, "
3440 "already seen\n", item);
3441 continue;
5032d1cf 3442 }
cc55f420 3443
b2bcdd35 3444 seen_items = string_catn(seen_items, US":", 1);
cc55f420 3445 }
cc55f420
JH
3446 seen_items = string_cat(seen_items, item);
3447
18067c75 3448 rc = dkim_exim_acl_run(item, &results, &user_msg, &log_msg);
cc55f420
JH
3449 if (rc != OK)
3450 {
3451 DEBUG(D_receive)
3452 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, "
3453 "skipping remaining items\n", rc, item);
3454 cancel_cutthrough_connection(TRUE, US"dkim acl not ok");
3455 break;
3456 }
3457 }
3458 dkim_verify_status = string_from_gstring(results);
3459 store_pool = old_pool;
3460 add_acl_headers(ACL_WHERE_DKIM, US"DKIM");
3461 if (rc == DISCARD)
3462 {
3463 recipients_count = 0;
3464 blackholed_by = US"DKIM ACL";
3465 if (log_msg)
3466 blackhole_log_msg = string_sprintf(": %s", log_msg);
3467 }
3468 else if (rc != OK)
3469 {
3470 Uunlink(spool_name);
3471 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
3472 smtp_yield = FALSE; /* No more messages after dropped connection */
3473 smtp_reply = US""; /* Indicate reply already sent */
3474 message_id[0] = 0; /* Indicate no message accepted */
3475 goto TIDYUP; /* Skip to end of function */
3476 }
80a47a2c 3477 }
a79d8834
JH
3478 else
3479 dkim_exim_verify_log_all();
80a47a2c 3480 }
4a8ce2d8 3481#endif /* DISABLE_DKIM */
fb2274d4 3482
8523533c 3483#ifdef WITH_CONTENT_SCAN
80a47a2c
TK
3484 if (recipients_count > 0 &&
3485 acl_smtp_mime != NULL &&
54cdb463
PH
3486 !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by))
3487 goto TIDYUP;
8523533c
TK
3488#endif /* WITH_CONTENT_SCAN */
3489
4840604e
TL
3490#ifdef EXPERIMENTAL_DMARC
3491 dmarc_up = dmarc_store_data(from_header);
3492#endif /* EXPERIMENTAL_DMARC */
3493
8ccd00b1
JH
3494#ifndef DISABLE_PRDR
3495 if (prdr_requested && recipients_count > 1 && acl_smtp_data_prdr)
fd98a5c6
JH
3496 {
3497 unsigned int c;
3498 int all_pass = OK;
3499 int all_fail = FAIL;
3500
925ac8e4 3501 smtp_printf("353 PRDR content analysis beginning\r\n", TRUE);
fd98a5c6
JH
3502 /* Loop through recipients, responses must be in same order received */
3503 for (c = 0; recipients_count > c; c++)
3504 {
3505 uschar * addr= recipients_list[c].address;
3506 uschar * msg= US"PRDR R=<%s> %s";
3507 uschar * code;
3508 DEBUG(D_receive)
3509 debug_printf("PRDR processing recipient %s (%d of %d)\n",
3510 addr, c+1, recipients_count);
3511 rc = acl_check(ACL_WHERE_PRDR, addr,
3512 acl_smtp_data_prdr, &user_msg, &log_msg);
3513
3514 /* If any recipient rejected content, indicate it in final message */
3515 all_pass |= rc;
3516 /* If all recipients rejected, indicate in final message */
3517 all_fail &= rc;
3518
3519 switch (rc)
3520 {
3521 case OK: case DISCARD: code = US"250"; break;
3522 case DEFER: code = US"450"; break;
3523 default: code = US"550"; break;
3524 }
3525 if (user_msg != NULL)
3526 smtp_user_msg(code, user_msg);
3527 else
3528 {
3529 switch (rc)
3530 {
3531 case OK: case DISCARD:
3532 msg = string_sprintf(CS msg, addr, "acceptance"); break;
3533 case DEFER:
3534 msg = string_sprintf(CS msg, addr, "temporary refusal"); break;
3535 default:
3536 msg = string_sprintf(CS msg, addr, "refusal"); break;
3537 }
3538 smtp_user_msg(code, msg);
3539 }
3540 if (log_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, log_msg);
3541 else if (user_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, user_msg);
112b6a93 3542 else log_write(0, LOG_MAIN, "%s", CS msg);
fd98a5c6
JH
3543
3544 if (rc != OK) { receive_remove_recipient(addr); c--; }
3545 }
3546 /* Set up final message, used if data acl gives OK */
3547 smtp_reply = string_sprintf("%s id=%s message %s",
3548 all_fail == FAIL ? US"550" : US"250",
3549 message_id,
3550 all_fail == FAIL
3551 ? US"rejected for all recipients"
3552 : all_pass == OK
3553 ? US"accepted"
3554 : US"accepted for some recipients");
3555 if (recipients_count == 0)
3556 {
3557 message_id[0] = 0; /* Indicate no message accepted */
3558 goto TIDYUP;
3559 }
3560 }
3561 else
3562 prdr_requested = FALSE;
8ccd00b1 3563#endif /* !DISABLE_PRDR */
fd98a5c6 3564
54cdb463
PH
3565 /* Check the recipients count again, as the MIME ACL might have changed
3566 them. */
8523533c 3567
059ec3d9
PH
3568 if (acl_smtp_data != NULL && recipients_count > 0)
3569 {
059ec3d9 3570 rc = acl_check(ACL_WHERE_DATA, NULL, acl_smtp_data, &user_msg, &log_msg);
578d43dc 3571 add_acl_headers(ACL_WHERE_DATA, US"DATA");
059ec3d9
PH
3572 if (rc == DISCARD)
3573 {
3574 recipients_count = 0;
3575 blackholed_by = US"DATA ACL";
57cc2785 3576 if (log_msg)
8e669ac1 3577 blackhole_log_msg = string_sprintf(": %s", log_msg);
57cc2785 3578 cancel_cutthrough_connection(TRUE, US"data acl discard");
059ec3d9
PH
3579 }
3580 else if (rc != OK)
3581 {
3582 Uunlink(spool_name);
57cc2785 3583 cancel_cutthrough_connection(TRUE, US"data acl not ok");
8523533c
TK
3584#ifdef WITH_CONTENT_SCAN
3585 unspool_mbox();
3586#endif
6f0c431a
PP
3587#ifdef EXPERIMENTAL_DCC
3588 dcc_ok = 0;
3589#endif
059ec3d9 3590 if (smtp_handle_acl_fail(ACL_WHERE_DATA, rc, user_msg, log_msg) != 0)
85ffcba6 3591 smtp_yield = FALSE; /* No more messages after dropped connection */
059ec3d9
PH
3592 smtp_reply = US""; /* Indicate reply already sent */
3593 message_id[0] = 0; /* Indicate no message accepted */
3594 goto TIDYUP; /* Skip to end of function */
3595 }
3596 }
3597 }
3598
3599 /* Handle non-SMTP and batch SMTP (i.e. non-interactive) messages. Note that
3600 we cannot take different actions for permanent and temporary rejections. */
3601
54cdb463 3602 else
059ec3d9 3603 {
54cdb463
PH
3604
3605#ifdef WITH_CONTENT_SCAN
3606 if (acl_not_smtp_mime != NULL &&
3607 !run_mime_acl(acl_not_smtp_mime, &smtp_yield, &smtp_reply,
3608 &blackholed_by))
3609 goto TIDYUP;
3610#endif /* WITH_CONTENT_SCAN */
3611
3612 if (acl_not_smtp != NULL)
059ec3d9 3613 {
54cdb463
PH
3614 uschar *user_msg, *log_msg;
3615 rc = acl_check(ACL_WHERE_NOTSMTP, NULL, acl_not_smtp, &user_msg, &log_msg);
3616 if (rc == DISCARD)
059ec3d9 3617 {
54cdb463
PH
3618 recipients_count = 0;
3619 blackholed_by = US"non-SMTP ACL";
3620 if (log_msg != NULL)
3621 blackhole_log_msg = string_sprintf(": %s", log_msg);
059ec3d9 3622 }
54cdb463 3623 else if (rc != OK)
059ec3d9 3624 {
54cdb463
PH
3625 Uunlink(spool_name);
3626#ifdef WITH_CONTENT_SCAN
3627 unspool_mbox();
3628#endif
6f0c431a
PP
3629#ifdef EXPERIMENTAL_DCC
3630 dcc_ok = 0;
3631#endif
6ea85e9a
PH
3632 /* The ACL can specify where rejections are to be logged, possibly
3633 nowhere. The default is main and reject logs. */
3634
3635 if (log_reject_target != 0)
3636 log_write(0, log_reject_target, "F=<%s> rejected by non-SMTP ACL: %s",
3637 sender_address, log_msg);
3638
54cdb463
PH
3639 if (user_msg == NULL) user_msg = US"local configuration problem";
3640 if (smtp_batched_input)
3641 {
3642 moan_smtp_batch(NULL, "%d %s", 550, user_msg);
3643 /* Does not return */
3644 }
3645 else
3646 {
3647 fseek(data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3648 give_local_error(ERRMESS_LOCAL_ACL, user_msg,
3649 US"message rejected by non-SMTP ACL: ", error_rc, data_file,
3650 header_list);
3651 /* Does not return */
3652 }
059ec3d9 3653 }
578d43dc 3654 add_acl_headers(ACL_WHERE_NOTSMTP, US"non-SMTP");
059ec3d9 3655 }
059ec3d9
PH
3656 }
3657
54cdb463
PH
3658 /* The applicable ACLs have been run */
3659
059ec3d9
PH
3660 if (deliver_freeze) frozen_by = US"ACL"; /* for later logging */
3661 if (queue_only_policy) queued_by = US"ACL";
059ec3d9
PH
3662