More checks on header line length during reception
[exim.git] / src / src / receive.c
CommitLineData
059ec3d9
PH
1/*************************************************
2* Exim - an Internet mail transport agent *
3*************************************************/
4
f9ba5e22 5/* Copyright (c) University of Cambridge 1995 - 2018 */
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 10#include "exim.h"
9723f966 11#include <setjmp.h>
059ec3d9 12
6a8f9482
TK
13#ifdef EXPERIMENTAL_DCC
14extern int dcc_ok;
15#endif
16
4840604e 17#ifdef EXPERIMENTAL_DMARC
c007c974 18# include "dmarc.h"
4840604e
TL
19#endif /* EXPERIMENTAL_DMARC */
20
059ec3d9
PH
21/*************************************************
22* Local static variables *
23*************************************************/
24
059ec3d9 25static int data_fd = -1;
41313d92 26static uschar *spool_name = US"";
059ec3d9 27
cff70eb1 28enum CH_STATE {LF_SEEN, MID_LINE, CR_SEEN};
059ec3d9 29
9723f966
JH
30#ifdef HAVE_LOCAL_SCAN
31jmp_buf local_scan_env; /* error-handling context for local_scan */
32unsigned had_local_scan_crash;
33unsigned had_local_scan_timeout;
34#endif
35
059ec3d9
PH
36
37/*************************************************
38* Non-SMTP character reading functions *
39*************************************************/
40
41/* These are the default functions that are set up in the variables such as
42receive_getc initially. They just call the standard functions, passing stdin as
43the file. (When SMTP input is occurring, different functions are used by
44changing the pointer variables.) */
45
46int
bd8fbe36 47stdin_getc(unsigned lim)
059ec3d9 48{
9723f966
JH
49int c = getc(stdin);
50
51if (had_data_timeout)
52 {
53 fprintf(stderr, "exim: timed out while reading - message abandoned\n");
54 log_write(L_lost_incoming_connection,
55 LOG_MAIN, "timed out while reading local message");
56 receive_bomb_out(US"data-timeout", NULL); /* Does not return */
57 }
58if (had_data_sigint)
59 {
60 if (filter_test == FTEST_NONE)
61 {
62 fprintf(stderr, "\nexim: %s received - message abandoned\n",
63 had_data_sigint == SIGTERM ? "SIGTERM" : "SIGINT");
64 log_write(0, LOG_MAIN, "%s received while reading local message",
65 had_data_sigint == SIGTERM ? "SIGTERM" : "SIGINT");
66 }
67 receive_bomb_out(US"signal-exit", NULL); /* Does not return */
68 }
69return c;
059ec3d9
PH
70}
71
72int
73stdin_ungetc(int c)
74{
75return ungetc(c, stdin);
76}
77
78int
79stdin_feof(void)
80{
81return feof(stdin);
82}
83
84int
85stdin_ferror(void)
86{
87return ferror(stdin);
88}
89
90
91
92
93/*************************************************
94* Check that a set sender is allowed *
95*************************************************/
96
97/* This function is called when a local caller sets an explicit sender address.
98It checks whether this is permitted, which it is for trusted callers.
99Otherwise, it must match the pattern(s) in untrusted_set_sender.
100
101Arguments: the proposed sender address
102Returns: TRUE for a trusted caller
103 TRUE if the address has been set, untrusted_set_sender has been
104 set, and the address matches something in the list
105 FALSE otherwise
106*/
107
108BOOL
109receive_check_set_sender(uschar *newsender)
110{
111uschar *qnewsender;
8768d548 112if (f.trusted_caller) return TRUE;
36d295f1
JH
113if (!newsender || !untrusted_set_sender) return FALSE;
114qnewsender = Ustrchr(newsender, '@')
115 ? newsender : string_sprintf("%s@%s", newsender, qualify_domain_sender);
116return match_address_list_basic(qnewsender, CUSS &untrusted_set_sender, 0) == OK;
059ec3d9
PH
117}
118
119
120
121
122/*************************************************
5cb8cbc6 123* Read space info for a partition *
059ec3d9
PH
124*************************************************/
125
8e669ac1
PH
126/* This function is called by receive_check_fs() below, and also by string
127expansion for variables such as $spool_space. The field names for the statvfs
5cb8cbc6
PH
128structure are macros, because not all OS have F_FAVAIL and it seems tidier to
129have macros for F_BAVAIL and F_FILES as well. Some kinds of file system do not
130have inodes, and they return -1 for the number available.
059ec3d9 131
5cb8cbc6
PH
132Later: It turns out that some file systems that do not have the concept of
133inodes return 0 rather than -1. Such systems should also return 0 for the total
8e669ac1 134number of inodes, so we require that to be greater than zero before returning
5cb8cbc6 135an inode count.
059ec3d9 136
5cb8cbc6
PH
137Arguments:
138 isspool TRUE for spool partition, FALSE for log partition
139 inodeptr address of int to receive inode count; -1 if there isn't one
8e669ac1 140
5cb8cbc6 141Returns: available on-root space, in kilobytes
8e669ac1
PH
142 -1 for log partition if there isn't one
143
144All values are -1 if the STATFS functions are not available.
059ec3d9
PH
145*/
146
a45431fa 147int_eximarith_t
5cb8cbc6 148receive_statvfs(BOOL isspool, int *inodeptr)
059ec3d9
PH
149{
150#ifdef HAVE_STATFS
059ec3d9 151struct STATVFS statbuf;
ddf1b11a 152struct stat dummy;
5cb8cbc6
PH
153uschar *path;
154uschar *name;
155uschar buffer[1024];
059ec3d9 156
5cb8cbc6 157/* The spool directory must always exist. */
059ec3d9 158
5cb8cbc6 159if (isspool)
059ec3d9 160 {
8e669ac1
PH
161 path = spool_directory;
162 name = US"spool";
163 }
164
059ec3d9
PH
165/* Need to cut down the log file path to the directory, and to ignore any
166appearance of "syslog" in it. */
167
5cb8cbc6 168else
059ec3d9 169 {
059ec3d9 170 int sep = ':'; /* Not variable - outside scripts use */
55414b25 171 const uschar *p = log_file_path;
8e669ac1 172 name = US"log";
059ec3d9
PH
173
174 /* An empty log_file_path means "use the default". This is the same as an
175 empty item in a list. */
176
177 if (*p == 0) p = US":";
55414b25
JH
178 while ((path = string_nextinlist(&p, &sep, buffer, sizeof(buffer))))
179 if (Ustrcmp(path, "syslog") != 0)
180 break;
059ec3d9 181
5cb8cbc6
PH
182 if (path == NULL) /* No log files */
183 {
8e669ac1
PH
184 *inodeptr = -1;
185 return -1;
186 }
059ec3d9 187
8e669ac1
PH
188 /* An empty string means use the default, which is in the spool directory.
189 But don't just use the spool directory, as it is possible that the log
5cb8cbc6 190 subdirectory has been symbolically linked elsewhere. */
059ec3d9 191
8e669ac1 192 if (path[0] == 0)
059ec3d9 193 {
5cb8cbc6
PH
194 sprintf(CS buffer, CS"%s/log", CS spool_directory);
195 path = buffer;
8e669ac1
PH
196 }
197 else
059ec3d9 198 {
8e669ac1 199 uschar *cp;
5cb8cbc6 200 if ((cp = Ustrrchr(path, '/')) != NULL) *cp = 0;
8e669ac1 201 }
5cb8cbc6 202 }
8e669ac1 203
8f128379 204/* We now have the path; do the business */
5cb8cbc6
PH
205
206memset(&statbuf, 0, sizeof(statbuf));
207
208if (STATVFS(CS path, &statbuf) != 0)
ddf1b11a
JH
209 if (stat(CS path, &dummy) == -1 && errno == ENOENT)
210 { /* Can happen on first run after installation */
211 *inodeptr = -1;
212 return -1;
213 }
214 else
215 {
216 log_write(0, LOG_MAIN|LOG_PANIC, "cannot accept message: failed to stat "
217 "%s directory %s: %s", name, path, strerror(errno));
218 smtp_closedown(US"spool or log directory problem");
9bfb7e1b 219 exim_exit(EXIT_FAILURE, NULL);
ddf1b11a 220 }
8e669ac1 221
5cb8cbc6
PH
222*inodeptr = (statbuf.F_FILES > 0)? statbuf.F_FAVAIL : -1;
223
224/* Disks are getting huge. Take care with computing the size in kilobytes. */
8e669ac1 225
a45431fa 226return (int_eximarith_t)(((double)statbuf.F_BAVAIL * (double)statbuf.F_FRSIZE)/1024.0);
5cb8cbc6 227
ddf1b11a 228#else
5cb8cbc6
PH
229/* Unable to find partition sizes in this environment. */
230
5cb8cbc6
PH
231*inodeptr = -1;
232return -1;
233#endif
234}
235
059ec3d9 236
059ec3d9 237
5cb8cbc6
PH
238
239/*************************************************
240* Check space on spool and log partitions *
241*************************************************/
242
243/* This function is called before accepting a message; if any thresholds are
244set, it checks them. If a message_size is supplied, it checks that there is
245enough space for that size plus the threshold - i.e. that the message won't
246reduce the space to the threshold. Not all OS have statvfs(); for those that
247don't, this function always returns TRUE. For some OS the old function and
248struct name statfs is used; that is handled by a macro, defined in exim.h.
249
250Arguments:
251 msg_size the (estimated) size of an incoming message
252
253Returns: FALSE if there isn't enough space, or if the information cannot
254 be obtained
255 TRUE if no check was done or there is enough space
256*/
257
258BOOL
259receive_check_fs(int msg_size)
260{
a45431fa
JH
261int_eximarith_t space;
262int inodes;
5cb8cbc6
PH
263
264if (check_spool_space > 0 || msg_size > 0 || check_spool_inodes > 0)
265 {
8e669ac1
PH
266 space = receive_statvfs(TRUE, &inodes);
267
059ec3d9 268 DEBUG(D_receive)
7434882d 269 debug_printf("spool directory space = " PR_EXIM_ARITH "K inodes = %d "
a45431fa 270 "check_space = " PR_EXIM_ARITH "K inodes = %d msg_size = %d\n",
5cb8cbc6 271 space, inodes, check_spool_space, check_spool_inodes, msg_size);
8e669ac1
PH
272
273 if ((space >= 0 && space < check_spool_space) ||
5cb8cbc6 274 (inodes >= 0 && inodes < check_spool_inodes))
8e669ac1 275 {
8c513105
JH
276 log_write(0, LOG_MAIN, "spool directory space check failed: space="
277 PR_EXIM_ARITH " inodes=%d", space, inodes);
059ec3d9
PH
278 return FALSE;
279 }
280 }
281
5cb8cbc6
PH
282if (check_log_space > 0 || check_log_inodes > 0)
283 {
8e669ac1
PH
284 space = receive_statvfs(FALSE, &inodes);
285
5cb8cbc6 286 DEBUG(D_receive)
7434882d 287 debug_printf("log directory space = " PR_EXIM_ARITH "K inodes = %d "
a45431fa 288 "check_space = " PR_EXIM_ARITH "K inodes = %d\n",
5cb8cbc6 289 space, inodes, check_log_space, check_log_inodes);
8e669ac1 290
7434882d
JH
291 if ( space >= 0 && space < check_log_space
292 || inodes >= 0 && inodes < check_log_inodes)
8e669ac1 293 {
7434882d
JH
294 log_write(0, LOG_MAIN, "log directory space check failed: space=" PR_EXIM_ARITH
295 " inodes=%d", space, inodes);
5cb8cbc6
PH
296 return FALSE;
297 }
8e669ac1
PH
298 }
299
059ec3d9
PH
300return TRUE;
301}
302
303
304
305/*************************************************
306* Bomb out while reading a message *
307*************************************************/
308
309/* The common case of wanting to bomb out is if a SIGTERM or SIGINT is
310received, or if there is a timeout. A rarer case might be if the log files are
311screwed up and Exim can't open them to record a message's arrival. Handling
312that case is done by setting a flag to cause the log functions to call this
313function if there is an ultimate disaster. That is why it is globally
314accessible.
315
8f128379
PH
316Arguments:
317 reason text reason to pass to the not-quit ACL
318 msg default SMTP response to give if in an SMTP session
059ec3d9
PH
319Returns: it doesn't
320*/
321
322void
8f128379 323receive_bomb_out(uschar *reason, uschar *msg)
059ec3d9 324{
ead37e6c
PP
325 static BOOL already_bombing_out;
326/* The smtp_notquit_exit() below can call ACLs which can trigger recursive
327timeouts, if someone has something slow in their quit ACL. Since the only
328things we should be doing are to close down cleanly ASAP, on the second
329pass we also close down stuff that might be opened again, before bypassing
330the ACL call and exiting. */
331
059ec3d9
PH
332/* If spool_name is set, it contains the name of the data file that is being
333written. Unlink it before closing so that it cannot be picked up by a delivery
334process. Ensure that any header file is also removed. */
335
ead37e6c 336if (spool_name[0] != '\0')
059ec3d9
PH
337 {
338 Uunlink(spool_name);
339 spool_name[Ustrlen(spool_name) - 1] = 'H';
340 Uunlink(spool_name);
ead37e6c 341 spool_name[0] = '\0';
059ec3d9
PH
342 }
343
344/* Now close the file if it is open, either as a fd or a stream. */
345
1bd642c2 346if (spool_data_file)
ead37e6c 347 {
1bd642c2
JH
348 (void)fclose(spool_data_file);
349 spool_data_file = NULL;
9723f966
JH
350 }
351else if (data_fd >= 0)
352 {
ead37e6c
PP
353 (void)close(data_fd);
354 data_fd = -1;
355 }
059ec3d9 356
8f128379
PH
357/* Attempt to close down an SMTP connection tidily. For non-batched SMTP, call
358smtp_notquit_exit(), which runs the NOTQUIT ACL, if present, and handles the
359SMTP response. */
059ec3d9 360
ead37e6c 361if (!already_bombing_out)
059ec3d9 362 {
ead37e6c
PP
363 already_bombing_out = TRUE;
364 if (smtp_input)
365 {
366 if (smtp_batched_input)
367 moan_smtp_batch(NULL, "421 %s - message abandoned", msg); /* No return */
368 smtp_notquit_exit(reason, US"421", US"%s %s - closing connection.",
369 smtp_active_hostname, msg);
370 }
059ec3d9
PH
371 }
372
373/* Exit from the program (non-BSMTP cases) */
374
9bfb7e1b 375exim_exit(EXIT_FAILURE, NULL);
059ec3d9
PH
376}
377
378
379/*************************************************
380* Data read timeout *
381*************************************************/
382
383/* Handler function for timeouts that occur while reading the data that
384comprises a message.
385
386Argument: the signal number
387Returns: nothing
388*/
389
390static void
391data_timeout_handler(int sig)
392{
9723f966 393had_data_timeout = sig;
059ec3d9
PH
394}
395
396
397
9723f966 398#ifdef HAVE_LOCAL_SCAN
059ec3d9
PH
399/*************************************************
400* local_scan() timeout *
401*************************************************/
402
403/* Handler function for timeouts that occur while running a local_scan()
9723f966
JH
404function. Posix recommends against calling longjmp() from a signal-handler,
405but the GCC manual says you can so we will, and trust that it's better than
406calling probably non-signal-safe funxtions during logging from within the
407handler, even with other compilers.
408
409See also https://cwe.mitre.org/data/definitions/745.html which also lists
410it as unsafe.
411
412This is all because we have no control over what might be written for a
413local-scan function, so cannot sprinkle had-signal checks after each
414call-site. At least with the default "do-nothing" function we won't
415ever get here.
059ec3d9
PH
416
417Argument: the signal number
418Returns: nothing
419*/
420
421static void
422local_scan_timeout_handler(int sig)
423{
9723f966
JH
424had_local_scan_timeout = sig;
425siglongjmp(local_scan_env, 1);
059ec3d9
PH
426}
427
428
429
430/*************************************************
431* local_scan() crashed *
432*************************************************/
433
434/* Handler function for signals that occur while running a local_scan()
435function.
436
437Argument: the signal number
438Returns: nothing
439*/
440
441static void
442local_scan_crash_handler(int sig)
443{
9723f966
JH
444had_local_scan_crash = sig;
445siglongjmp(local_scan_env, 1);
059ec3d9
PH
446}
447
9723f966
JH
448#endif /*HAVE_LOCAL_SCAN*/
449
059ec3d9
PH
450
451/*************************************************
452* SIGTERM or SIGINT received *
453*************************************************/
454
455/* Handler for SIGTERM or SIGINT signals that occur while reading the
456data that comprises a message.
457
458Argument: the signal number
459Returns: nothing
460*/
461
462static void
463data_sigterm_sigint_handler(int sig)
464{
9723f966 465had_data_sigint = sig;
059ec3d9
PH
466}
467
468
469
470/*************************************************
471* Add new recipient to list *
472*************************************************/
473
474/* This function builds a list of recipient addresses in argc/argv
475format.
476
477Arguments:
478 recipient the next address to add to recipients_list
479 pno parent number for fixed aliases; -1 otherwise
480
481Returns: nothing
482*/
483
484void
485receive_add_recipient(uschar *recipient, int pno)
486{
487if (recipients_count >= recipients_list_max)
488 {
489 recipient_item *oldlist = recipients_list;
490 int oldmax = recipients_list_max;
ea97267c 491 recipients_list_max = recipients_list_max ? 2*recipients_list_max : 50;
059ec3d9
PH
492 recipients_list = store_get(recipients_list_max * sizeof(recipient_item));
493 if (oldlist != NULL)
494 memcpy(recipients_list, oldlist, oldmax * sizeof(recipient_item));
495 }
496
497recipients_list[recipients_count].address = recipient;
498recipients_list[recipients_count].pno = pno;
8523533c
TK
499#ifdef EXPERIMENTAL_BRIGHTMAIL
500recipients_list[recipients_count].bmi_optin = bmi_current_optin;
501/* reset optin string pointer for next recipient */
502bmi_current_optin = NULL;
503#endif
6c1c3d1d
WB
504recipients_list[recipients_count].orcpt = NULL;
505recipients_list[recipients_count].dsn_flags = 0;
059ec3d9
PH
506recipients_list[recipients_count++].errors_to = NULL;
507}
508
509
510
511
512/*************************************************
fd98a5c6
JH
513* Send user response message *
514*************************************************/
61147df4 515
fd98a5c6
JH
516/* This function is passed a default response code and a user message. It calls
517smtp_message_code() to check and possibly modify the response code, and then
518calls smtp_respond() to transmit the response. I put this into a function
519just to avoid a lot of repetition.
61147df4
PP
520
521Arguments:
fd98a5c6
JH
522 code the response code
523 user_msg the user message
524
525Returns: nothing
61147df4
PP
526*/
527
8ccd00b1 528#ifndef DISABLE_PRDR
61147df4 529static void
fd98a5c6 530smtp_user_msg(uschar *code, uschar *user_msg)
61147df4 531{
fd98a5c6 532int len = 3;
4f6ae5c3 533smtp_message_code(&code, &len, &user_msg, NULL, TRUE);
fd98a5c6 534smtp_respond(code, len, TRUE, user_msg);
61147df4
PP
535}
536#endif
537
538
539
540
fd98a5c6
JH
541
542/*************************************************
059ec3d9
PH
543* Remove a recipient from the list *
544*************************************************/
545
546/* This function is provided for local_scan() to use.
547
548Argument:
549 recipient address to remove
550
551Returns: TRUE if it did remove something; FALSE otherwise
552*/
553
554BOOL
555receive_remove_recipient(uschar *recipient)
556{
557int count;
558DEBUG(D_receive) debug_printf("receive_remove_recipient(\"%s\") called\n",
559 recipient);
560for (count = 0; count < recipients_count; count++)
561 {
562 if (Ustrcmp(recipients_list[count].address, recipient) == 0)
563 {
564 if ((--recipients_count - count) > 0)
565 memmove(recipients_list + count, recipients_list + count + 1,
54cdb463 566 (recipients_count - count)*sizeof(recipient_item));
059ec3d9
PH
567 return TRUE;
568 }
569 }
570return FALSE;
571}
572
573
574
575
576
577/*************************************************
578* Read data portion of a non-SMTP message *
579*************************************************/
580
581/* This function is called to read the remainder of a message (following the
582header) when the input is not from SMTP - we are receiving a local message on
583a standard input stream. The message is always terminated by EOF, and is also
584terminated by a dot on a line by itself if the flag dot_ends is TRUE. Split the
585two cases for maximum efficiency.
586
587Ensure that the body ends with a newline. This will naturally be the case when
588the termination is "\n.\n" but may not be otherwise. The RFC defines messages
589as "sequences of lines" - this of course strictly applies only to SMTP, but
590deliveries into BSD-type mailbox files also require it. Exim used to have a
591flag for doing this at delivery time, but as it was always set for all
592transports, I decided to simplify things by putting the check here instead.
593
594There is at least one MUA (dtmail) that sends CRLF via this interface, and
595other programs are known to do this as well. Exim used to have a option for
596dealing with this: in July 2003, after much discussion, the code has been
597changed to default to treat any of LF, CRLF, and bare CR as line terminators.
598
599However, for the case when a dot on a line by itself terminates a message, the
600only recognized terminating sequences before and after the dot are LF and CRLF.
601Otherwise, having read EOL . CR, you don't know whether to read another
602character or not.
603
604Internally, in messages stored in Exim's spool files, LF is used as the line
605terminator. Under the new regime, bare CRs will no longer appear in these
606files.
607
608Arguments:
609 fout a FILE to which to write the message
610
611Returns: One of the END_xxx values indicating why it stopped reading
612*/
613
614static int
615read_message_data(FILE *fout)
616{
617int ch_state;
618register int ch;
d677b2f2 619register int linelength = 0;
059ec3d9
PH
620
621/* Handle the case when only EOF terminates the message */
622
8768d548 623if (!f.dot_ends)
059ec3d9
PH
624 {
625 register int last_ch = '\n';
626
bd8fbe36 627 for (; (ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF; last_ch = ch)
059ec3d9
PH
628 {
629 if (ch == 0) body_zerocount++;
630 if (last_ch == '\r' && ch != '\n')
631 {
d677b2f2
PH
632 if (linelength > max_received_linelength)
633 max_received_linelength = linelength;
634 linelength = 0;
059ec3d9
PH
635 if (fputc('\n', fout) == EOF) return END_WERROR;
636 message_size++;
637 body_linecount++;
638 }
639 if (ch == '\r') continue;
640
641 if (fputc(ch, fout) == EOF) return END_WERROR;
d677b2f2
PH
642 if (ch == '\n')
643 {
644 if (linelength > max_received_linelength)
645 max_received_linelength = linelength;
646 linelength = 0;
647 body_linecount++;
648 }
649 else linelength++;
059ec3d9
PH
650 if (++message_size > thismessage_size_limit) return END_SIZE;
651 }
652
653 if (last_ch != '\n')
654 {
d677b2f2
PH
655 if (linelength > max_received_linelength)
656 max_received_linelength = linelength;
059ec3d9
PH
657 if (fputc('\n', fout) == EOF) return END_WERROR;
658 message_size++;
659 body_linecount++;
660 }
661
662 return END_EOF;
663 }
664
665/* Handle the case when a dot on a line on its own, or EOF, terminates. */
666
667ch_state = 1;
668
bd8fbe36 669while ((ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF)
059ec3d9
PH
670 {
671 if (ch == 0) body_zerocount++;
672 switch (ch_state)
673 {
674 case 0: /* Normal state (previous char written) */
675 if (ch == '\n')
d677b2f2
PH
676 {
677 body_linecount++;
678 if (linelength > max_received_linelength)
679 max_received_linelength = linelength;
680 linelength = -1;
681 ch_state = 1;
682 }
059ec3d9
PH
683 else if (ch == '\r')
684 { ch_state = 2; continue; }
685 break;
686
687 case 1: /* After written "\n" */
688 if (ch == '.') { ch_state = 3; continue; }
6eb02f88 689 if (ch == '\r') { ch_state = 2; continue; }
3581f321
JH
690 if (ch == '\n') { body_linecount++; linelength = -1; }
691 else ch_state = 0;
059ec3d9
PH
692 break;
693
694 case 2:
695 body_linecount++; /* After unwritten "\r" */
d677b2f2
PH
696 if (linelength > max_received_linelength)
697 max_received_linelength = linelength;
059ec3d9 698 if (ch == '\n')
d677b2f2
PH
699 {
700 ch_state = 1;
701 linelength = -1;
702 }
059ec3d9
PH
703 else
704 {
705 if (message_size++, fputc('\n', fout) == EOF) return END_WERROR;
706 if (ch == '\r') continue;
707 ch_state = 0;
d677b2f2 708 linelength = 0;
059ec3d9
PH
709 }
710 break;
711
712 case 3: /* After "\n." (\n written, dot not) */
713 if (ch == '\n') return END_DOT;
714 if (ch == '\r') { ch_state = 4; continue; }
715 message_size++;
d677b2f2 716 linelength++;
059ec3d9
PH
717 if (fputc('.', fout) == EOF) return END_WERROR;
718 ch_state = 0;
719 break;
720
721 case 4: /* After "\n.\r" (\n written, rest not) */
722 if (ch == '\n') return END_DOT;
723 message_size += 2;
724 body_linecount++;
725 if (fputs(".\n", fout) == EOF) return END_WERROR;
726 if (ch == '\r') { ch_state = 2; continue; }
727 ch_state = 0;
728 break;
729 }
730
d677b2f2 731 linelength++;
059ec3d9
PH
732 if (fputc(ch, fout) == EOF) return END_WERROR;
733 if (++message_size > thismessage_size_limit) return END_SIZE;
734 }
735
736/* Get here if EOF read. Unless we have just written "\n", we need to ensure
737the message ends with a newline, and we must also write any characters that
738were saved up while testing for an ending dot. */
739
740if (ch_state != 1)
741 {
742 static uschar *ends[] = { US"\n", NULL, US"\n", US".\n", US".\n" };
743 if (fputs(CS ends[ch_state], fout) == EOF) return END_WERROR;
744 message_size += Ustrlen(ends[ch_state]);
745 body_linecount++;
746 }
747
748return END_EOF;
749}
750
751
752
753
754/*************************************************
755* Read data portion of an SMTP message *
756*************************************************/
757
758/* This function is called to read the remainder of an SMTP message (after the
759headers), or to skip over it when an error has occurred. In this case, the
760output file is passed as NULL.
761
762If any line begins with a dot, that character is skipped. The input should only
763be successfully terminated by CR LF . CR LF unless it is local (non-network)
764SMTP, in which case the CRs are optional, but...
765
766FUDGE: It seems that sites on the net send out messages with just LF
767terminators, despite the warnings in the RFCs, and other MTAs handle this. So
768we make the CRs optional in all cases.
769
770July 2003: Bare CRs cause trouble. We now treat them as line terminators as
771well, so that there are no CRs in spooled messages. However, the message
772terminating dot is not recognized between two bare CRs.
773
774Arguments:
775 fout a FILE to which to write the message; NULL if skipping
776
777Returns: One of the END_xxx values indicating why it stopped reading
778*/
779
780static int
781read_message_data_smtp(FILE *fout)
782{
783int ch_state = 0;
e4bdf652 784int ch;
7e3ce68e 785int linelength = 0;
059ec3d9 786
bd8fbe36 787while ((ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF)
059ec3d9
PH
788 {
789 if (ch == 0) body_zerocount++;
790 switch (ch_state)
791 {
792 case 0: /* After LF or CRLF */
793 if (ch == '.')
794 {
795 ch_state = 3;
796 continue; /* Don't ever write . after LF */
797 }
798 ch_state = 1;
799
800 /* Else fall through to handle as normal uschar. */
801
802 case 1: /* Normal state */
803 if (ch == '\n')
804 {
805 ch_state = 0;
806 body_linecount++;
1f5497b2
PH
807 if (linelength > max_received_linelength)
808 max_received_linelength = linelength;
809 linelength = -1;
059ec3d9
PH
810 }
811 else if (ch == '\r')
812 {
813 ch_state = 2;
814 continue;
815 }
816 break;
817
818 case 2: /* After (unwritten) CR */
819 body_linecount++;
1f5497b2
PH
820 if (linelength > max_received_linelength)
821 max_received_linelength = linelength;
822 linelength = -1;
059ec3d9
PH
823 if (ch == '\n')
824 {
825 ch_state = 0;
826 }
827 else
828 {
829 message_size++;
830 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
6851a9c5 831 cutthrough_data_put_nl();
059ec3d9
PH
832 if (ch != '\r') ch_state = 1; else continue;
833 }
834 break;
835
836 case 3: /* After [CR] LF . */
837 if (ch == '\n')
838 return END_DOT;
839 if (ch == '\r')
840 {
841 ch_state = 4;
842 continue;
843 }
1bc460a6
JH
844 /* The dot was removed at state 3. For a doubled dot, here, reinstate
845 it to cutthrough. The current ch, dot or not, is passed both to cutthrough
846 and to file below. */
847 if (ch == '.')
848 {
849 uschar c= ch;
6851a9c5 850 cutthrough_data_puts(&c, 1);
1bc460a6
JH
851 }
852 ch_state = 1;
059ec3d9
PH
853 break;
854
855 case 4: /* After [CR] LF . CR */
856 if (ch == '\n') return END_DOT;
857 message_size++;
858 body_linecount++;
859 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
6851a9c5 860 cutthrough_data_put_nl();
059ec3d9
PH
861 if (ch == '\r')
862 {
863 ch_state = 2;
864 continue;
865 }
866 ch_state = 1;
867 break;
868 }
869
870 /* Add the character to the spool file, unless skipping; then loop for the
871 next. */
872
873 message_size++;
1f5497b2 874 linelength++;
7e3ce68e 875 if (fout)
059ec3d9
PH
876 {
877 if (fputc(ch, fout) == EOF) return END_WERROR;
878 if (message_size > thismessage_size_limit) return END_SIZE;
879 }
e4bdf652 880 if(ch == '\n')
6851a9c5 881 cutthrough_data_put_nl();
e4bdf652
JH
882 else
883 {
7e3ce68e 884 uschar c = ch;
6851a9c5 885 cutthrough_data_puts(&c, 1);
e4bdf652 886 }
059ec3d9
PH
887 }
888
889/* Fall through here if EOF encountered. This indicates some kind of error,
890since a correct message is terminated by [CR] LF . [CR] LF. */
891
892return END_EOF;
893}
894
895
896
897
7e3ce68e 898/* Variant of the above read_message_data_smtp() specialised for RFC 3030
1ebe15c3
JH
899CHUNKING. Accept input lines separated by either CRLF or CR or LF and write
900LF-delimited spoolfile. Until we have wireformat spoolfiles, we need the
901body_linecount accounting for proper re-expansion for the wire, so use
902a cut-down version of the state-machine above; we don't need to do leading-dot
903detection and unstuffing.
7e3ce68e
JH
904
905Arguments:
544dd905
PP
906 fout a FILE to which to write the message; NULL if skipping;
907 must be open for both writing and reading.
7e3ce68e
JH
908
909Returns: One of the END_xxx values indicating why it stopped reading
910*/
911
912static int
913read_message_bdat_smtp(FILE *fout)
914{
cff70eb1
HSHR
915int linelength = 0, ch;
916enum CH_STATE ch_state = LF_SEEN;
d953610f 917BOOL fix_nl = FALSE;
7e3ce68e 918
1ebe15c3 919for(;;)
7e3ce68e 920 {
7d758a6a 921 switch ((ch = bdat_getc(GETC_BUFFER_UNLIMITED)))
1ebe15c3
JH
922 {
923 case EOF: return END_EOF;
1ebe15c3 924 case ERR: return END_PROTOCOL;
d953610f
HSHR
925 case EOD:
926 /* Nothing to get from the sender anymore. We check the last
927 character written to the spool.
928
929 RFC 3030 states, that BDAT chunks are normal text, terminated by CRLF.
930 If we would be strict, we would refuse such broken messages.
931 But we are liberal, so we fix it. It would be easy just to append
932 the "\n" to the spool.
933
934 But there are some more things (line counting, message size calculation and such),
935 that would need to be duplicated here. So we simply do some ungetc
936 trickery.
937 */
59d98039
JH
938 if (fout)
939 {
940 if (fseek(fout, -1, SEEK_CUR) < 0) return END_PROTOCOL;
941 if (fgetc(fout) == '\n') return END_DOT;
942 }
d953610f
HSHR
943
944 if (linelength == -1) /* \r already seen (see below) */
945 {
946 DEBUG(D_receive) debug_printf("Add missing LF\n");
947 bdat_ungetc('\n');
948 continue;
949 }
950 DEBUG(D_receive) debug_printf("Add missing CRLF\n");
951 bdat_ungetc('\r'); /* not even \r was seen */
952 fix_nl = TRUE;
953
954 continue;
1ebe15c3
JH
955 case '\0': body_zerocount++; break;
956 }
957 switch (ch_state)
958 {
cff70eb1
HSHR
959 case LF_SEEN: /* After LF or CRLF */
960 ch_state = MID_LINE;
1ebe15c3 961 /* fall through to handle as normal uschar. */
7e3ce68e 962
cff70eb1 963 case MID_LINE: /* Mid-line state */
1ebe15c3
JH
964 if (ch == '\n')
965 {
cff70eb1 966 ch_state = LF_SEEN;
1ebe15c3
JH
967 body_linecount++;
968 if (linelength > max_received_linelength)
969 max_received_linelength = linelength;
970 linelength = -1;
971 }
972 else if (ch == '\r')
973 {
cff70eb1 974 ch_state = CR_SEEN;
d953610f 975 if (fix_nl) bdat_ungetc('\n');
1ebe15c3
JH
976 continue; /* don't write CR */
977 }
978 break;
7e3ce68e 979
cff70eb1 980 case CR_SEEN: /* After (unwritten) CR */
1ebe15c3
JH
981 body_linecount++;
982 if (linelength > max_received_linelength)
983 max_received_linelength = linelength;
984 linelength = -1;
985 if (ch == '\n')
cff70eb1 986 ch_state = LF_SEEN;
1ebe15c3
JH
987 else
988 {
989 message_size++;
59d98039 990 if (fout && fputc('\n', fout) == EOF) return END_WERROR;
6851a9c5 991 cutthrough_data_put_nl();
1ebe15c3 992 if (ch == '\r') continue; /* don't write CR */
cff70eb1 993 ch_state = MID_LINE;
1ebe15c3
JH
994 }
995 break;
996 }
997
998 /* Add the character to the spool file, unless skipping */
999
1000 message_size++;
1001 linelength++;
1002 if (fout)
1003 {
1004 if (fputc(ch, fout) == EOF) return END_WERROR;
1005 if (message_size > thismessage_size_limit) return END_SIZE;
1006 }
1007 if(ch == '\n')
6851a9c5 1008 cutthrough_data_put_nl();
1ebe15c3
JH
1009 else
1010 {
1011 uschar c = ch;
6851a9c5 1012 cutthrough_data_puts(&c, 1);
1ebe15c3 1013 }
7e3ce68e
JH
1014 }
1015/*NOTREACHED*/
1016}
1017
328c5688
JH
1018static int
1019read_message_bdat_smtp_wire(FILE *fout)
1020{
1021int ch;
1022
1023/* Remember that this message uses wireformat. */
1024
d21bf202
JH
1025DEBUG(D_receive) debug_printf("CHUNKING: %s\n",
1026 fout ? "writing spoolfile in wire format" : "flushing input");
8768d548 1027f.spool_file_wireformat = TRUE;
328c5688 1028
0d81dabc 1029for (;;)
328c5688 1030 {
0d81dabc
JH
1031 if (chunking_data_left > 0)
1032 {
1033 unsigned len = MAX(chunking_data_left, thismessage_size_limit - message_size + 1);
1034 uschar * buf = bdat_getbuf(&len);
328c5688 1035
8b77d27a 1036 if (!buf) return END_EOF;
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:
74f1a423 1185 if ( cutthrough.cctx.sock >= 0 && cutthrough.delivery
57cc2785 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:
c674e7a4
JH
1226 h->next = header_list;
1227 header_list = h;
1228 DEBUG(D_receive|D_acl) debug_printf_indent(" (at top)");
1229 break;
059ec3d9
PH
1230
1231 case htype_add_rec:
c674e7a4
JH
1232 if (!last_received)
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 &&
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;
1243 DEBUG(D_receive|D_acl) debug_printf_indent(" (after Received:)");
1244 break;
059ec3d9 1245
8523533c 1246 case htype_add_rfc:
c674e7a4
JH
1247 /* add header before any header which is NOT Received: or Resent- */
1248 last_received = header_list;
1249 while ( last_received->next &&
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;
1258 DEBUG(D_receive|D_acl) debug_printf_indent(" (before any non-Received: or Resent-*: header)");
1259 break;
8523533c 1260
059ec3d9 1261 default:
c674e7a4
JH
1262 h->next = NULL;
1263 header_last->next = h;
1264 DEBUG(D_receive|D_acl) debug_printf_indent(" ");
1265 break;
059ec3d9
PH
1266 }
1267
c674e7a4 1268 if (!h->next) header_last = h;
059ec3d9
PH
1269
1270 /* Check for one of the known header types (From:, To:, etc.) though in
1271 practice most added headers are going to be "other". Lower case
1272 identification letters are never stored with the header; they are used
1273 for existence tests when messages are received. So discard any lower case
1274 flag values. */
1275
1276 h->type = header_checkname(h, FALSE);
1277 if (h->type >= 'a') h->type = htype_other;
1278
c674e7a4 1279 DEBUG(D_receive|D_acl) debug_printf("%s", h->text);
059ec3d9
PH
1280 }
1281
71fafd95 1282acl_added_headers = NULL;
e1d04f48 1283DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
059ec3d9
PH
1284}
1285
1286
1287
1288/*************************************************
1289* Add host information for log line *
1290*************************************************/
1291
1292/* Called for acceptance and rejecting log lines. This adds information about
1293the calling host to a string that is being built dynamically.
1294
1295Arguments:
1296 s the dynamic string
059ec3d9
PH
1297
1298Returns: the extended string
1299*/
1300
acec9514
JH
1301static gstring *
1302add_host_info_for_log(gstring * g)
059ec3d9 1303{
fc16abb4 1304if (sender_fullhost)
059ec3d9 1305 {
fc16abb4 1306 if (LOGGING(dnssec) && sender_host_dnssec) /*XXX sender_helo_dnssec? */
acec9514
JH
1307 g = string_catn(g, US" DS", 3);
1308 g = string_append(g, 2, US" H=", sender_fullhost);
52f12a7c
JH
1309 if (LOGGING(incoming_interface) && interface_address)
1310 g = string_fmt_append(g, " I=[%s]:%d", interface_address, interface_port);
059ec3d9 1311 }
8768d548 1312if (f.tcp_in_fastopen && !f.tcp_in_fastopen_logged)
a2673768 1313 {
ee8b8090 1314 g = string_catn(g, US" TFO*", f.tcp_in_fastopen_data ? 5 : 4);
8768d548 1315 f.tcp_in_fastopen_logged = TRUE;
a2673768
JH
1316 }
1317if (sender_ident)
acec9514 1318 g = string_append(g, 2, US" U=", sender_ident);
a2673768 1319if (received_protocol)
acec9514 1320 g = string_append(g, 2, US" P=", received_protocol);
8768d548 1321if (LOGGING(pipelining) && f.smtp_in_pipelining_advertised)
ee8b8090
JH
1322 {
1323 g = string_catn(g, US" L", 2);
1324#ifdef EXPERIMENTAL_PIPE_CONNECT
1325 if (f.smtp_in_early_pipe_used)
1326 g = string_catn(g, US"*", 1);
1327 else if (f.smtp_in_early_pipe_advertised)
1328 g = string_catn(g, US".", 1);
1329#endif
1330 if (!f.smtp_in_pipelining_used)
1331 g = string_catn(g, US"-", 1);
1332 }
acec9514 1333return g;
059ec3d9
PH
1334}
1335
1336
1337
63955bf2 1338#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1339
1340/*************************************************
54cdb463
PH
1341* Run the MIME ACL on a message *
1342*************************************************/
1343
1344/* This code is in a subroutine so that it can be used for both SMTP
1345and non-SMTP messages. It is called with a non-NULL ACL pointer.
1346
1347Arguments:
1348 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1349 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1350 smtp_reply_ptr Where SMTP reply is being built
1351 blackholed_by_ptr Where "blackholed by" message is being built
1352
1353Returns: TRUE to carry on; FALSE to abandon the message
1354*/
1355
1356static BOOL
1357run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1358 uschar **blackholed_by_ptr)
1359{
1360FILE *mbox_file;
5b6f7658 1361uschar * rfc822_file_path = NULL;
54cdb463
PH
1362unsigned long mbox_size;
1363header_line *my_headerlist;
1364uschar *user_msg, *log_msg;
1365int mime_part_count_buffer = -1;
040721f2 1366uschar * mbox_filename;
7156b1ef 1367int rc = OK;
54cdb463 1368
54cdb463 1369/* check if it is a MIME message */
040721f2
JH
1370
1371for (my_headerlist = header_list; my_headerlist; my_headerlist = my_headerlist->next)
1372 if ( my_headerlist->type != '*' /* skip deleted headers */
1373 && strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0
1374 )
4e88a19f 1375 {
54cdb463
PH
1376 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1377 goto DO_MIME_ACL;
4e88a19f 1378 }
54cdb463
PH
1379
1380DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1381return TRUE;
1382
1383DO_MIME_ACL:
040721f2 1384
54cdb463 1385/* make sure the eml mbox file is spooled up */
040721f2
JH
1386if (!(mbox_file = spool_mbox(&mbox_size, NULL, &mbox_filename)))
1387 { /* error while spooling */
54cdb463
PH
1388 log_write(0, LOG_MAIN|LOG_PANIC,
1389 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1390 Uunlink(spool_name);
1391 unspool_mbox();
6f0c431a
PP
1392#ifdef EXPERIMENTAL_DCC
1393 dcc_ok = 0;
1394#endif
a5bd321b 1395 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1396 message_id[0] = 0; /* Indicate no message accepted */
1397 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1398 return FALSE; /* Indicate skip to end of receive function */
040721f2 1399 }
54cdb463
PH
1400
1401mime_is_rfc822 = 0;
1402
1403MIME_ACL_CHECK:
1404mime_part_count = -1;
1405rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1406(void)fclose(mbox_file);
54cdb463 1407
5b6f7658 1408if (rfc822_file_path)
4e88a19f 1409 {
54cdb463
PH
1410 mime_part_count = mime_part_count_buffer;
1411
4e88a19f
PH
1412 if (unlink(CS rfc822_file_path) == -1)
1413 {
54cdb463
PH
1414 log_write(0, LOG_PANIC,
1415 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
5b6f7658 1416 goto END_MIME_ACL;
4e88a19f 1417 }
5b6f7658 1418 rfc822_file_path = NULL;
4e88a19f 1419 }
54cdb463
PH
1420
1421/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1422if (rc == OK)
1423 {
5b6f7658
JH
1424 uschar * scandir = string_copyn(mbox_filename,
1425 Ustrrchr(mbox_filename, '/') - mbox_filename);
e8bc7fca
JH
1426 struct dirent * entry;
1427 DIR * tempdir;
54cdb463 1428
5b6f7658 1429 for (tempdir = opendir(CS scandir); entry = readdir(tempdir); )
e8bc7fca 1430 if (strncmpic(US entry->d_name, US"__rfc822_", 9) == 0)
4e88a19f 1431 {
5b6f7658
JH
1432 rfc822_file_path = string_sprintf("%s/%s", scandir, entry->d_name);
1433 DEBUG(D_receive)
1434 debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
1435 rfc822_file_path);
4e88a19f
PH
1436 break;
1437 }
4e88a19f 1438 closedir(tempdir);
54cdb463 1439
5b6f7658 1440 if (rfc822_file_path)
4e88a19f 1441 {
e8bc7fca 1442 if ((mbox_file = Ufopen(rfc822_file_path, "rb")))
4e88a19f 1443 {
e8bc7fca
JH
1444 /* set RFC822 expansion variable */
1445 mime_is_rfc822 = 1;
1446 mime_part_count_buffer = mime_part_count;
1447 goto MIME_ACL_CHECK;
4e88a19f 1448 }
e8bc7fca
JH
1449 log_write(0, LOG_PANIC,
1450 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1451 unlink(CS rfc822_file_path);
4e88a19f
PH
1452 }
1453 }
54cdb463
PH
1454
1455END_MIME_ACL:
578d43dc 1456add_acl_headers(ACL_WHERE_MIME, US"MIME");
54cdb463
PH
1457if (rc == DISCARD)
1458 {
1459 recipients_count = 0;
1460 *blackholed_by_ptr = US"MIME ACL";
1bd642c2 1461 cancel_cutthrough_connection(TRUE, US"mime acl discard");
54cdb463
PH
1462 }
1463else if (rc != OK)
1464 {
1465 Uunlink(spool_name);
1bd642c2 1466 cancel_cutthrough_connection(TRUE, US"mime acl not ok");
54cdb463 1467 unspool_mbox();
6f0c431a
PP
1468#ifdef EXPERIMENTAL_DCC
1469 dcc_ok = 0;
1470#endif
5b6f7658 1471 if (smtp_input)
4f6ae5c3 1472 {
5b6f7658
JH
1473 if (smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1474 *smtp_yield_ptr = FALSE; /* No more messages after dropped connection */
f4c1088b 1475 *smtp_reply_ptr = US""; /* Indicate reply already sent */
4f6ae5c3 1476 }
54cdb463
PH
1477 message_id[0] = 0; /* Indicate no message accepted */
1478 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1479 }
54cdb463
PH
1480
1481return TRUE;
1482}
1483
63955bf2 1484#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1485
1486
e4bdf652
JH
1487
1488void
1489received_header_gen(void)
1490{
1491uschar *received;
1492uschar *timestamp;
1493header_line *received_header= header_list;
1494
1495timestamp = expand_string(US"${tod_full}");
1496if (recipients_count == 1) received_for = recipients_list[0].address;
1497received = expand_string(received_header_text);
1498received_for = NULL;
1499
d4ff61d1 1500if (!received)
e4bdf652
JH
1501 {
1502 if(spool_name[0] != 0)
1503 Uunlink(spool_name); /* Lose the data file */
1504 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1505 "(received_header_text) failed: %s", string_printing(received_header_text),
1506 expand_string_message);
1507 }
1508
1509/* The first element on the header chain is reserved for the Received header,
1510so all we have to do is fill in the text pointer, and set the type. However, if
1511the result of the expansion is an empty string, we leave the header marked as
1512"old" so as to refrain from adding a Received header. */
1513
1514if (received[0] == 0)
1515 {
1516 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1517 received_header->type = htype_old;
1518 }
1519else
1520 {
1521 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1522 received_header->type = htype_received;
1523 }
1524
1525received_header->slen = Ustrlen(received_header->text);
1526
1527DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1528 received_header->type, received_header->text);
1529}
1530
1531
1532
54cdb463 1533/*************************************************
059ec3d9
PH
1534* Receive message *
1535*************************************************/
1536
1537/* Receive a message on the given input, and put it into a pair of spool files.
1538Either a non-null list of recipients, or the extract flag will be true, or
1539both. The flag sender_local is true for locally generated messages. The flag
1540submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1541suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1542suppress_local_fixups" or -G was passed on the command-line.
1543The flag smtp_input is true if the message is to be
8800895a
PH
1544handled using SMTP conventions about termination and lines starting with dots.
1545For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1546
1547If a message was successfully read, message_id[0] will be non-zero.
1548
1549The general actions of this function are:
1550
1551 . Read the headers of the message (if any) into a chain of store
1552 blocks.
1553
1554 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1555 throw it away, unless the caller is trusted, or unless
1556 active_local_sender_retain is set - which can only happen if
1557 active_local_from_check is false.
059ec3d9
PH
1558
1559 . If recipients are to be extracted from the message, build the
1560 recipients list from the headers, removing any that were on the
1561 original recipients list (unless extract_addresses_remove_arguments is
1562 false), and at the same time, remove any bcc header that may be present.
1563
1564 . Get the spool file for the data, sort out its unique name, open
1565 and lock it (but don't give it the name yet).
1566
1567 . Generate a "Message-Id" header if the message doesn't have one, for
1568 locally-originated messages.
1569
1570 . Generate a "Received" header.
1571
1572 . Ensure the recipients list is fully qualified and rewritten if necessary.
1573
1574 . If there are any rewriting rules, apply them to the sender address
1575 and also to the headers.
1576
1577 . If there is no from: header, generate one, for locally-generated messages
1578 and messages in "submission mode" only.
1579
1580 . If the sender is local, check that from: is correct, and if not, generate
1581 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1582 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1583
1584 . If there is no "date" header, generate one, for locally-originated
1585 or submission mode messages only.
1586
1587 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1588 dot_ends mode, to the data file. Leave it open, to hold the lock.
1589
1590 . Write the envelope and the headers to a new file.
1591
1592 . Set the name for the header file; close it.
1593
1594 . Set the name for the data file; close it.
1595
1596Because this function can potentially be called many times in a single
1597SMTP connection, all store should be got by store_get(), so that it will be
1598automatically retrieved after the message is accepted.
1599
1600FUDGE: It seems that sites on the net send out messages with just LF
1601terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1602we make the CRs optional in all cases.
1603
1604July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1605new regime is now in place in which bare CRs in header lines are turned into LF
1606followed by a space, so as not to terminate the header line.
1607
1608February 2004: A bare LF in a header line in a message whose first line was
1609terminated by CRLF is treated in the same way as a bare CR.
1610
1611Arguments:
1612 extract_recip TRUE if recipients are to be extracted from the message's
1613 headers
1614
1615Returns: TRUE there are more messages to be read (SMTP input)
1616 FALSE there are no more messages to be read (non-SMTP input
1617 or SMTP connection collapsed, or other failure)
1618
1619When reading a message for filter testing, the returned value indicates
1620whether the headers (which is all that is read) were terminated by '.' or
1621not. */
1622
1623BOOL
1624receive_msg(BOOL extract_recip)
1625{
7156b1ef
NM
1626int i;
1627int rc = FAIL;
059ec3d9
PH
1628int msg_size = 0;
1629int process_info_len = Ustrlen(process_info);
d342446f
JH
1630int error_rc = error_handling == ERRORS_SENDER
1631 ? errors_sender_rc : EXIT_FAILURE;
059ec3d9 1632int header_size = 256;
acec9514 1633int start, end, domain;
099afc4f 1634int id_resolution = 0;
059ec3d9 1635int had_zero = 0;
d677b2f2 1636int prevlines_length = 0;
059ec3d9 1637
cfbb0d24 1638int ptr = 0;
059ec3d9
PH
1639
1640BOOL contains_resent_headers = FALSE;
1641BOOL extracted_ignored = FALSE;
1642BOOL first_line_ended_crlf = TRUE_UNSET;
1643BOOL smtp_yield = TRUE;
1644BOOL yield = FALSE;
1645
1646BOOL resents_exist = FALSE;
1647uschar *resent_prefix = US"";
1648uschar *blackholed_by = NULL;
04f7d5b9 1649uschar *blackhole_log_msg = US"";
c5430c20 1650enum {NOT_TRIED, TMP_REJ, PERM_REJ, ACCEPTED} cutthrough_done = NOT_TRIED;
059ec3d9
PH
1651
1652flock_t lock_data;
1653error_block *bad_addresses = NULL;
1654
1655uschar *frozen_by = NULL;
1656uschar *queued_by = NULL;
1657
acec9514
JH
1658uschar *errmsg;
1659gstring * g;
059ec3d9
PH
1660struct stat statbuf;
1661
4e88a19f 1662/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1663
1664uschar *smtp_reply = NULL;
4e88a19f 1665uschar *user_msg, *log_msg;
059ec3d9
PH
1666
1667/* Working header pointers */
1668
1669header_line *h, *next;
1670
2cbb4081 1671/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1672
1673BOOL date_header_exists = FALSE;
1674
1675/* Pointers to receive the addresses of headers whose contents we need. */
1676
1677header_line *from_header = NULL;
1678header_line *subject_header = NULL;
1679header_line *msgid_header = NULL;
1680header_line *received_header;
1681
4840604e
TL
1682#ifdef EXPERIMENTAL_DMARC
1683int dmarc_up = 0;
1684#endif /* EXPERIMENTAL_DMARC */
1685
059ec3d9
PH
1686/* Variables for use when building the Received: header. */
1687
059ec3d9
PH
1688uschar *timestamp;
1689int tslen;
1690
9723f966 1691
059ec3d9
PH
1692/* Release any open files that might have been cached while preparing to
1693accept the message - e.g. by verifying addresses - because reading a message
1694might take a fair bit of real time. */
1695
1696search_tidyup();
1697
e4bdf652
JH
1698/* Extracting the recipient list from an input file is incompatible with
1699cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1700to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1701if (extract_recip || !smtp_input)
57cc2785 1702 cancel_cutthrough_connection(TRUE, US"not smtp input");
e4bdf652 1703
059ec3d9
PH
1704/* Initialize the chain of headers by setting up a place-holder for Received:
1705header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1706pointing to the end of the chain to make adding headers simple. */
1707
1708received_header = header_list = header_last = store_get(sizeof(header_line));
1709header_list->next = NULL;
1710header_list->type = htype_old;
1711header_list->text = NULL;
1712header_list->slen = 0;
1713
1714/* Control block for the next header to be read. */
1715
1716next = store_get(sizeof(header_line));
1717next->text = store_get(header_size);
1718
1719/* Initialize message id to be null (indicating no message read), and the
1720header names list to be the normal list. Indicate there is no data file open
1721yet, initialize the size and warning count, and deal with no size limit. */
1722
1723message_id[0] = 0;
1bd642c2 1724spool_data_file = NULL;
059ec3d9 1725data_fd = -1;
41313d92 1726spool_name = US"";
059ec3d9
PH
1727message_size = 0;
1728warning_count = 0;
d677b2f2 1729received_count = 1; /* For the one we will add */
059ec3d9
PH
1730
1731if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1732
2e0c1448 1733/* While reading the message, the following counts are computed. */
059ec3d9 1734
d677b2f2
PH
1735message_linecount = body_linecount = body_zerocount =
1736 max_received_linelength = 0;
059ec3d9 1737
80a47a2c 1738#ifndef DISABLE_DKIM
e983e85a
JH
1739/* Call into DKIM to set up the context. In CHUNKING mode
1740we clear the dot-stuffing flag */
8768d548 1741if (smtp_input && !smtp_batched_input && !f.dkim_disable_verify)
e983e85a 1742 dkim_exim_verify_init(chunking_state <= CHUNKING_OFFERED);
fb2274d4
TK
1743#endif
1744
4840604e
TL
1745#ifdef EXPERIMENTAL_DMARC
1746/* initialize libopendmarc */
1747dmarc_up = dmarc_init();
1748#endif
1749
059ec3d9
PH
1750/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1751ids, and fractions of a second are required. See the comments that precede the
1752message id creation below. */
1753
1754(void)gettimeofday(&message_id_tv, NULL);
1755
1756/* For other uses of the received time we can operate with granularity of one
1757second, and for that we use the global variable received_time. This is for
306c6c77 1758things like ultimate message timeouts. */
059ec3d9 1759
32dfdf8b 1760received_time = message_id_tv;
059ec3d9
PH
1761
1762/* If SMTP input, set the special handler for timeouts. The alarm() calls
1763happen in the smtp_getc() function when it refills its buffer. */
1764
9723f966
JH
1765had_data_timeout = 0;
1766if (smtp_input)
1767 os_non_restarting_signal(SIGALRM, data_timeout_handler);
059ec3d9
PH
1768
1769/* If not SMTP input, timeout happens only if configured, and we just set a
1770single timeout for the whole message. */
1771
1772else if (receive_timeout > 0)
1773 {
1774 os_non_restarting_signal(SIGALRM, data_timeout_handler);
c2a1bba0 1775 ALARM(receive_timeout);
059ec3d9
PH
1776 }
1777
1778/* SIGTERM and SIGINT are caught always. */
1779
9723f966 1780had_data_sigint = 0;
059ec3d9
PH
1781signal(SIGTERM, data_sigterm_sigint_handler);
1782signal(SIGINT, data_sigterm_sigint_handler);
1783
1784/* Header lines in messages are not supposed to be very long, though when
1785unfolded, to: and cc: headers can take up a lot of store. We must also cope
1786with the possibility of junk being thrown at us. Start by getting 256 bytes for
1787storing the header, and extend this as necessary using string_cat().
1788
1789To cope with total lunacies, impose an upper limit on the length of the header
1790section of the message, as otherwise the store will fill up. We must also cope
1791with the possibility of binary zeros in the data. Hence we cannot use fgets().
1792Folded header lines are joined into one string, leaving the '\n' characters
1793inside them, so that writing them out reproduces the input.
1794
1795Loop for each character of each header; the next structure for chaining the
1796header is set up already, with ptr the offset of the next character in
1797next->text. */
1798
1799for (;;)
1800 {
bd8fbe36 1801 int ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1802
1803 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1804 SMTP must have a correct "." terminator. */
1805
1806 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1807 {
1808 smtp_reply = handle_lost_connection(US" (header)");
1809 smtp_yield = FALSE;
1810 goto TIDYUP; /* Skip to end of function */
1811 }
1812
1813 /* See if we are at the current header's size limit - there must be at least
1814 four bytes left. This allows for the new character plus a zero, plus two for
1815 extra insertions when we are playing games with dots and carriage returns. If
1816 we are at the limit, extend the text buffer. This could have been done
1817 automatically using string_cat() but because this is a tightish loop storing
1818 only one character at a time, we choose to do it inline. Normally
1819 store_extend() will be able to extend the block; only at the end of a big
1820 store block will a copy be needed. To handle the case of very long headers
1821 (and sometimes lunatic messages can have ones that are 100s of K long) we
1822 call store_release() for strings that have been copied - if the string is at
1823 the start of a block (and therefore the only thing in it, because we aren't
4e6ae623
JH
1824 doing any other gets), the block gets freed. We can only do this release if
1825 there were no allocations since the once that we want to free. */
059ec3d9
PH
1826
1827 if (ptr >= header_size - 4)
1828 {
1829 int oldsize = header_size;
56ac062a
JH
1830
1831 if (header_size >= INT_MAX/2)
1832 goto OVERSIZE;
059ec3d9 1833 header_size *= 2;
56ac062a 1834
059ec3d9 1835 if (!store_extend(next->text, oldsize, header_size))
459fca58 1836 next->text = store_newblock(next->text, header_size, ptr);
059ec3d9
PH
1837 }
1838
1839 /* Cope with receiving a binary zero. There is dispute about whether
1840 these should be allowed in RFC 822 messages. The middle view is that they
1841 should not be allowed in headers, at least. Exim takes this attitude at
1842 the moment. We can't just stomp on them here, because we don't know that
1843 this line is a header yet. Set a flag to cause scanning later. */
1844
1845 if (ch == 0) had_zero++;
1846
1847 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1848 those from data files use just LF. Treat LF in local SMTP input as a
1849 terminator too. Treat EOF as a line terminator always. */
1850
1851 if (ch == EOF) goto EOL;
1852
1853 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1854 other MTAs accept this. We are therefore forced into this "liberalisation"
1855 too, so we accept LF as a line terminator whatever the source of the message.
1856 However, if the first line of the message ended with a CRLF, we treat a bare
1857 LF specially by inserting a white space after it to ensure that the header
1858 line is not terminated. */
1859
1860 if (ch == '\n')
1861 {
1862 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1863 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1864 goto EOL;
1865 }
1866
1867 /* This is not the end of the line. If this is SMTP input and this is
1868 the first character in the line and it is a "." character, ignore it.
1869 This implements the dot-doubling rule, though header lines starting with
1870 dots aren't exactly common. They are legal in RFC 822, though. If the
1871 following is CRLF or LF, this is the line that that terminates the
1872 entire message. We set message_ended to indicate this has happened (to
1873 prevent further reading), and break out of the loop, having freed the
1874 empty header, and set next = NULL to indicate no data line. */
1875
8768d548 1876 if (ptr == 0 && ch == '.' && f.dot_ends)
059ec3d9 1877 {
bd8fbe36 1878 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1879 if (ch == '\r')
1880 {
bd8fbe36 1881 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1882 if (ch != '\n')
1883 {
80a47a2c 1884 receive_ungetc(ch);
059ec3d9
PH
1885 ch = '\r'; /* Revert to CR */
1886 }
1887 }
1888 if (ch == '\n')
1889 {
1890 message_ended = END_DOT;
1891 store_reset(next);
1892 next = NULL;
1893 break; /* End character-reading loop */
1894 }
1895
1896 /* For non-SMTP input, the dot at the start of the line was really a data
1897 character. What is now in ch is the following character. We guaranteed
1898 enough space for this above. */
1899
1900 if (!smtp_input)
1901 {
1902 next->text[ptr++] = '.';
1903 message_size++;
1904 }
1905 }
1906
1907 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1908 remember this case if this is the first line ending. */
1909
1910 if (ch == '\r')
1911 {
bd8fbe36 1912 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1913 if (ch == '\n')
1914 {
1915 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1916 goto EOL;
1917 }
1918
1919 /* Otherwise, put back the character after CR, and turn the bare CR
1920 into LF SP. */
1921
80a47a2c 1922 ch = (receive_ungetc)(ch);
059ec3d9
PH
1923 next->text[ptr++] = '\n';
1924 message_size++;
1925 ch = ' ';
1926 }
1927
1928 /* We have a data character for the header line. */
1929
1930 next->text[ptr++] = ch; /* Add to buffer */
1931 message_size++; /* Total message size so far */
1932
1933 /* Handle failure due to a humungously long header section. The >= allows
1934 for the terminating \n. Add what we have so far onto the headers list so
1935 that it gets reflected in any error message, and back up the just-read
1936 character. */
1937
1938 if (message_size >= header_maxsize)
1939 {
56ac062a 1940OVERSIZE:
059ec3d9
PH
1941 next->text[ptr] = 0;
1942 next->slen = ptr;
1943 next->type = htype_other;
1944 next->next = NULL;
1945 header_last->next = next;
1946 header_last = next;
1947
1948 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1949 "%s (more than %d characters): message abandoned",
8768d548 1950 f.sender_host_unknown ? sender_ident : sender_fullhost, header_maxsize);
059ec3d9
PH
1951
1952 if (smtp_input)
1953 {
1954 smtp_reply = US"552 Message header is ridiculously long";
1955 receive_swallow_smtp();
1956 goto TIDYUP; /* Skip to end of function */
1957 }
1958
1959 else
1960 {
1961 give_local_error(ERRMESS_VLONGHEADER,
1962 string_sprintf("message header longer than %d characters received: "
1963 "message not accepted", header_maxsize), US"", error_rc, stdin,
1964 header_list->next);
1965 /* Does not return */
1966 }
1967 }
1968
1969 continue; /* With next input character */
1970
1971 /* End of header line reached */
1972
1973 EOL:
2e0c1448
PH
1974
1975 /* Keep track of lines for BSMTP errors and overall message_linecount. */
1976
1977 receive_linecount++;
1978 message_linecount++;
059ec3d9 1979
d677b2f2
PH
1980 /* Keep track of maximum line length */
1981
1982 if (ptr - prevlines_length > max_received_linelength)
1983 max_received_linelength = ptr - prevlines_length;
1984 prevlines_length = ptr + 1;
1985
059ec3d9
PH
1986 /* Now put in the terminating newline. There is always space for
1987 at least two more characters. */
1988
1989 next->text[ptr++] = '\n';
1990 message_size++;
1991
1992 /* A blank line signals the end of the headers; release the unwanted
1993 space and set next to NULL to indicate this. */
1994
1995 if (ptr == 1)
1996 {
1997 store_reset(next);
1998 next = NULL;
1999 break;
2000 }
2001
2002 /* There is data in the line; see if the next input character is a
2003 whitespace character. If it is, we have a continuation of this header line.
2004 There is always space for at least one character at this point. */
2005
2006 if (ch != EOF)
2007 {
bd8fbe36 2008 int nextch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
2009 if (nextch == ' ' || nextch == '\t')
2010 {
2011 next->text[ptr++] = nextch;
56ac062a
JH
2012 if (++message_size >= header_maxsize)
2013 goto OVERSIZE;
059ec3d9
PH
2014 continue; /* Iterate the loop */
2015 }
80a47a2c 2016 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
2017 else ch = EOF; /* Cause main loop to exit at end */
2018 }
2019
2020 /* We have got to the real line end. Terminate the string and release store
2021 beyond it. If it turns out to be a real header, internal binary zeros will
2022 be squashed later. */
2023
2024 next->text[ptr] = 0;
2025 next->slen = ptr;
2026 store_reset(next->text + ptr + 1);
2027
2028 /* Check the running total size against the overall message size limit. We
2029 don't expect to fail here, but if the overall limit is set less than MESSAGE_
2030 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
2031 headers - the code to read the body will then also hit the buffer. */
2032
2033 if (message_size > thismessage_size_limit) break;
2034
2035 /* A line that is not syntactically correct for a header also marks
2036 the end of the headers. In this case, we leave next containing the
2037 first data line. This might actually be several lines because of the
2038 continuation logic applied above, but that doesn't matter.
2039
2040 It turns out that smail, and presumably sendmail, accept leading lines
2041 of the form
2042
2043 From ph10 Fri Jan 5 12:35 GMT 1996
2044
2045 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
2046 find any documentation of this, but for compatibility it had better be
2047 accepted. Exim restricts it to the case of non-smtp messages, and
2048 treats it as an alternative to the -f command line option. Thus it is
2049 ignored except for trusted users or filter testing. Otherwise it is taken
2050 as the sender address, unless -f was used (sendmail compatibility).
2051
2052 It further turns out that some UUCPs generate the From_line in a different
2053 format, e.g.
2054
2055 From ph10 Fri, 7 Jan 97 14:00:00 GMT
2056
2057 The regex for matching these things is now capable of recognizing both
2058 formats (including 2- and 4-digit years in the latter). In fact, the regex
2059 is now configurable, as is the expansion string to fish out the sender.
2060
2061 Even further on it has been discovered that some broken clients send
2062 these lines in SMTP messages. There is now an option to ignore them from
2063 specified hosts or networks. Sigh. */
2064
d21bf202
JH
2065 if ( header_last == header_list
2066 && ( !smtp_input
2067 || ( sender_host_address
2068 && verify_check_host(&ignore_fromline_hosts) == OK
2069 )
2070 || (!sender_host_address && ignore_fromline_local)
2071 )
2072 && regex_match_and_setup(regex_From, next->text, 0, -1)
2073 )
059ec3d9 2074 {
8768d548 2075 if (!f.sender_address_forced)
059ec3d9
PH
2076 {
2077 uschar *uucp_sender = expand_string(uucp_from_sender);
d21bf202 2078 if (!uucp_sender)
059ec3d9
PH
2079 log_write(0, LOG_MAIN|LOG_PANIC,
2080 "expansion of \"%s\" failed after matching "
2081 "\"From \" line: %s", uucp_from_sender, expand_string_message);
059ec3d9
PH
2082 else
2083 {
2084 int start, end, domain;
2085 uschar *errmess;
2086 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
2087 &start, &end, &domain, TRUE);
d21bf202 2088 if (newsender)
059ec3d9
PH
2089 {
2090 if (domain == 0 && newsender[0] != 0)
2091 newsender = rewrite_address_qualify(newsender, FALSE);
2092
f05da2e8 2093 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
2094 {
2095 sender_address = newsender;
2096
8768d548 2097 if (f.trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
2098 {
2099 authenticated_sender = NULL;
2100 originator_name = US"";
8768d548 2101 f.sender_local = FALSE;
059ec3d9
PH
2102 }
2103
f05da2e8 2104 if (filter_test != FTEST_NONE)
059ec3d9
PH
2105 printf("Sender taken from \"From \" line\n");
2106 }
2107 }
2108 }
2109 }
2110 }
2111
2112 /* Not a leading "From " line. Check to see if it is a valid header line.
2113 Header names may contain any non-control characters except space and colon,
2114 amazingly. */
2115
2116 else
2117 {
2118 uschar *p = next->text;
2119
2120 /* If not a valid header line, break from the header reading loop, leaving
2121 next != NULL, indicating that it holds the first line of the body. */
2122
2123 if (isspace(*p)) break;
2124 while (mac_isgraph(*p) && *p != ':') p++;
2125 while (isspace(*p)) p++;
2126 if (*p != ':')
2127 {
2128 body_zerocount = had_zero;
2129 break;
2130 }
2131
2132 /* We have a valid header line. If there were any binary zeroes in
2133 the line, stomp on them here. */
2134
2135 if (had_zero > 0)
2136 for (p = next->text; p < next->text + ptr; p++) if (*p == 0) *p = '?';
2137
2138 /* It is perfectly legal to have an empty continuation line
2139 at the end of a header, but it is confusing to humans
2140 looking at such messages, since it looks like a blank line.
2141 Reduce confusion by removing redundant white space at the
2142 end. We know that there is at least one printing character
2143 (the ':' tested for above) so there is no danger of running
2144 off the end. */
2145
2146 p = next->text + ptr - 2;
2147 for (;;)
2148 {
2149 while (*p == ' ' || *p == '\t') p--;
2150 if (*p != '\n') break;
2151 ptr = (p--) - next->text + 1;
2152 message_size -= next->slen - ptr;
2153 next->text[ptr] = 0;
2154 next->slen = ptr;
2155 }
2156
2157 /* Add the header to the chain */
2158
2159 next->type = htype_other;
2160 next->next = NULL;
2161 header_last->next = next;
2162 header_last = next;
2163
2164 /* Check the limit for individual line lengths. This comes after adding to
2165 the chain so that the failing line is reflected if a bounce is generated
2166 (for a local message). */
2167
2168 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
2169 {
2170 log_write(0, LOG_MAIN, "overlong message header line received from "
2171 "%s (more than %d characters): message abandoned",
8768d548 2172 f.sender_host_unknown ? sender_ident : sender_fullhost,
059ec3d9
PH
2173 header_line_maxsize);
2174
2175 if (smtp_input)
2176 {
2177 smtp_reply = US"552 A message header line is too long";
2178 receive_swallow_smtp();
2179 goto TIDYUP; /* Skip to end of function */
2180 }
2181
2182 else
059ec3d9
PH
2183 give_local_error(ERRMESS_VLONGHDRLINE,
2184 string_sprintf("message header line longer than %d characters "
2185 "received: message not accepted", header_line_maxsize), US"",
2186 error_rc, stdin, header_list->next);
2187 /* Does not return */
059ec3d9
PH
2188 }
2189
2190 /* Note if any resent- fields exist. */
2191
2192 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
2193 {
2194 resents_exist = TRUE;
2195 resent_prefix = US"Resent-";
2196 }
2197 }
2198
1ebe15c3
JH
2199 /* Reject CHUNKING messages that do not CRLF their first header line */
2200
2201 if (!first_line_ended_crlf && chunking_state > CHUNKING_OFFERED)
2202 {
2203 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2204 "Non-CRLF-terminated header, under CHUNKING: message abandoned",
2205 sender_address,
2206 sender_fullhost ? " H=" : "", sender_fullhost ? sender_fullhost : US"",
2207 sender_ident ? " U=" : "", sender_ident ? sender_ident : US"");
925ac8e4 2208 smtp_printf("552 Message header not CRLF terminated\r\n", FALSE);
1ebe15c3
JH
2209 bdat_flush_data();
2210 smtp_reply = US"";
2211 goto TIDYUP; /* Skip to end of function */
2212 }
2213
059ec3d9
PH
2214 /* The line has been handled. If we have hit EOF, break out of the loop,
2215 indicating no pending data line. */
2216
2217 if (ch == EOF) { next = NULL; break; }
2218
2219 /* Set up for the next header */
2220
2221 header_size = 256;
2222 next = store_get(sizeof(header_line));
2223 next->text = store_get(header_size);
2224 ptr = 0;
2225 had_zero = 0;
d677b2f2 2226 prevlines_length = 0;
059ec3d9
PH
2227 } /* Continue, starting to read the next header */
2228
2229/* At this point, we have read all the headers into a data structure in main
2230store. The first header is still the dummy placeholder for the Received: header
2231we are going to generate a bit later on. If next != NULL, it contains the first
2232data line - which terminated the headers before reaching a blank line (not the
2233normal case). */
2234
2235DEBUG(D_receive)
2236 {
2237 debug_printf(">>Headers received:\n");
1ebe15c3 2238 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2239 debug_printf("%s", h->text);
2240 debug_printf("\n");
2241 }
2242
2243/* End of file on any SMTP connection is an error. If an incoming SMTP call
2244is dropped immediately after valid headers, the next thing we will see is EOF.
2245We must test for this specially, as further down the reading of the data is
2246skipped if already at EOF. */
2247
2248if (smtp_input && (receive_feof)())
2249 {
2250 smtp_reply = handle_lost_connection(US" (after header)");
2251 smtp_yield = FALSE;
2252 goto TIDYUP; /* Skip to end of function */
2253 }
2254
2255/* If this is a filter test run and no headers were read, output a warning
2256in case there is a mistake in the test message. */
2257
f05da2e8 2258if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2259 printf("Warning: no message headers read\n");
2260
2261
2262/* Scan the headers to identify them. Some are merely marked for later
2263processing; some are dealt with here. */
2264
1ebe15c3 2265for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2266 {
2267 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2268 if (is_resent) contains_resent_headers = TRUE;
2269
2270 switch (header_checkname(h, is_resent))
2271 {
059ec3d9 2272 case htype_bcc:
c674e7a4
JH
2273 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
2274 break;
059ec3d9 2275
059ec3d9 2276 case htype_cc:
c674e7a4
JH
2277 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
2278 break;
059ec3d9 2279
c674e7a4 2280 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
059ec3d9
PH
2281
2282 case htype_date:
c674e7a4
JH
2283 if (!resents_exist || is_resent) date_header_exists = TRUE;
2284 break;
059ec3d9 2285
c674e7a4 2286 /* Same comments as about Return-Path: below. */
059ec3d9
PH
2287
2288 case htype_delivery_date:
c674e7a4
JH
2289 if (delivery_date_remove) h->type = htype_old;
2290 break;
059ec3d9 2291
c674e7a4 2292 /* Same comments as about Return-Path: below. */
059ec3d9
PH
2293
2294 case htype_envelope_to:
c674e7a4
JH
2295 if (envelope_to_remove) h->type = htype_old;
2296 break;
059ec3d9 2297
c674e7a4
JH
2298 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2299 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2300 header consists of just the login id of the user who called Exim, rewrite
2301 it with the gecos field first. Apply this rule to Resent-From: if there
2302 are resent- fields. */
059ec3d9
PH
2303
2304 case htype_from:
c674e7a4
JH
2305 h->type = htype_from;
2306 if (!resents_exist || is_resent)
2307 {
2308 from_header = h;
2309 if (!smtp_input)
2310 {
2311 int len;
2312 uschar *s = Ustrchr(h->text, ':') + 1;
2313 while (isspace(*s)) s++;
2314 len = h->slen - (s - h->text) - 1;
2315 if (Ustrlen(originator_login) == len &&
2316 strncmpic(s, originator_login, len) == 0)
2317 {
2318 uschar *name = is_resent? US"Resent-From" : US"From";
2319 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2320 originator_login, qualify_domain_sender);
2321 from_header = header_last;
2322 h->type = htype_old;
2323 DEBUG(D_receive|D_rewrite)
2324 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2325 }
2326 }
2327 }
2328 break;
059ec3d9 2329
c674e7a4
JH
2330 /* Identify the Message-id: header for generating "in-reply-to" in the
2331 autoreply transport. For incoming logging, save any resent- value. In both
2332 cases, take just the first of any multiples. */
059ec3d9
PH
2333
2334 case htype_id:
c674e7a4
JH
2335 if (!msgid_header && (!resents_exist || is_resent))
2336 {
2337 msgid_header = h;
2338 h->type = htype_id;
2339 }
2340 break;
059ec3d9 2341
c674e7a4 2342 /* Flag all Received: headers */
059ec3d9
PH
2343
2344 case htype_received:
c674e7a4
JH
2345 h->type = htype_received;
2346 received_count++;
2347 break;
059ec3d9 2348
c674e7a4 2349 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
059ec3d9
PH
2350
2351 case htype_reply_to:
c674e7a4
JH
2352 h->type = htype_reply_to;
2353 break;
059ec3d9 2354
c674e7a4
JH
2355 /* The Return-path: header is supposed to be added to messages when
2356 they leave the SMTP system. We shouldn't receive messages that already
2357 contain Return-path. However, since Exim generates Return-path: on
2358 local delivery, resent messages may well contain it. We therefore
2359 provide an option (which defaults on) to remove any Return-path: headers
2360 on input. Removal actually means flagging as "old", which prevents the
2361 header being transmitted with the message. */
059ec3d9
PH
2362
2363 case htype_return_path:
c674e7a4 2364 if (return_path_remove) h->type = htype_old;
059ec3d9 2365
c674e7a4
JH
2366 /* If we are testing a mail filter file, use the value of the
2367 Return-Path: header to set up the return_path variable, which is not
2368 otherwise set. However, remove any <> that surround the address
2369 because the variable doesn't have these. */
059ec3d9 2370
c674e7a4
JH
2371 if (filter_test != FTEST_NONE)
2372 {
2373 uschar *start = h->text + 12;
2374 uschar *end = start + Ustrlen(start);
2375 while (isspace(*start)) start++;
2376 while (end > start && isspace(end[-1])) end--;
2377 if (*start == '<' && end[-1] == '>')
2378 {
2379 start++;
2380 end--;
2381 }
2382 return_path = string_copyn(start, end - start);
2383 printf("Return-path taken from \"Return-path:\" header line\n");
2384 }
2385 break;
059ec3d9
PH
2386
2387 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2388 and from an untrusted caller and suppress_local_fixups is not set, or if we
2389 are in submission mode for a remote message, mark it "old" so that it will
2390 not be transmitted with the message, unless active_local_sender_retain is
2391 set. (This can only be true if active_local_from_check is false.) If there
2392 are any resent- headers in the message, apply this rule to Resent-Sender:
2393 instead of Sender:. Messages with multiple resent- header sets cannot be
2394 tidily handled. (For this reason, at least one MUA - Pine - turns old
2395 resent- headers into X-resent- headers when resending, leaving just one
2396 set.) */
059ec3d9
PH
2397
2398 case htype_sender:
8768d548
JH
2399 h->type = !f.active_local_sender_retain
2400 && ( f.sender_local && !f.trusted_caller && !f.suppress_local_fixups
2401 || f.submission_mode
c674e7a4
JH
2402 )
2403 && (!resents_exist || is_resent)
2404 ? htype_old : htype_sender;
2405 break;
059ec3d9 2406
c674e7a4 2407 /* Remember the Subject: header for logging. There is no Resent-Subject */
059ec3d9
PH
2408
2409 case htype_subject:
c674e7a4
JH
2410 subject_header = h;
2411 break;
059ec3d9 2412
c674e7a4
JH
2413 /* "To:" gets flagged, and the existence of a recipient header is noted,
2414 whether it's resent- or not. */
059ec3d9
PH
2415
2416 case htype_to:
c674e7a4
JH
2417 h->type = htype_to;
2418 /****
2419 to_or_cc_header_exists = TRUE;
2420 ****/
2421 break;
059ec3d9
PH
2422 }
2423 }
2424
2425/* Extract recipients from the headers if that is required (the -t option).
2426Note that this is documented as being done *before* any address rewriting takes
2427place. There are two possibilities:
2428
2429(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2430recipients already listed are to be REMOVED from the message. Smail 3 works
2431like this. We need to build a non-recipients tree for that list, because in
2432subsequent processing this data is held in a tree and that's what the
2433spool_write_header() function expects. Make sure that non-recipient addresses
2434are fully qualified and rewritten if necessary.
2435
2436(2) According to other sendmail documentation, -t ADDS extracted recipients to
2437those in the command line arguments (and it is rumoured some other MTAs do
2438this). Therefore, there is an option to make Exim behave this way.
2439
2440*** Notes on "Resent-" header lines ***
2441
2442The presence of resent-headers in the message makes -t horribly ambiguous.
2443Experiments with sendmail showed that it uses recipients for all resent-
2444headers, totally ignoring the concept of "sets of resent- headers" as described
2445in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2446with all the addresses in one instance of each header.
2447
2448This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2449error for -t if there were resent- headers in the message. However, after a
2450discussion on the mailing list, I've learned that there are MUAs that use
2451resent- headers with -t, and also that the stuff about sets of resent- headers
2452and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2453message with -t and resent- header lines makes sure that only *its* resent-
2454headers are present; previous ones are often renamed as X-resent- for example.
2455
2456Consequently, Exim has been changed so that, if any resent- header lines are
2457present, the recipients are taken from all of the appropriate resent- lines,
2458and not from the ordinary To:, Cc:, etc. */
2459
2460if (extract_recip)
2461 {
2462 int rcount = 0;
2463 error_block **bnext = &bad_addresses;
2464
2465 if (extract_addresses_remove_arguments)
2466 {
2467 while (recipients_count-- > 0)
2468 {
2469 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2470 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2471 tree_add_nonrecipient(s);
2472 }
2473 recipients_list = NULL;
2474 recipients_count = recipients_list_max = 0;
2475 }
2476
059ec3d9
PH
2477 /* Now scan the headers */
2478
1ebe15c3 2479 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2480 {
2481 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2482 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2483 {
2484 uschar *s = Ustrchr(h->text, ':') + 1;
2485 while (isspace(*s)) s++;
2486
8768d548 2487 f.parse_allow_group = TRUE; /* Allow address group syntax */
1eccaa59 2488
059ec3d9
PH
2489 while (*s != 0)
2490 {
2491 uschar *ss = parse_find_address_end(s, FALSE);
2492 uschar *recipient, *errmess, *p, *pp;
2493 int start, end, domain;
2494
2495 /* Check on maximum */
2496
2497 if (recipients_max > 0 && ++rcount > recipients_max)
059ec3d9
PH
2498 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2499 US"message rejected: ", error_rc, stdin, NULL);
2500 /* Does not return */
059ec3d9
PH
2501
2502 /* Make a copy of the address, and remove any internal newlines. These
2503 may be present as a result of continuations of the header line. The
2504 white space that follows the newline must not be removed - it is part
2505 of the header. */
2506
2507 pp = recipient = store_get(ss - s + 1);
2508 for (p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
2509 *pp = 0;
250b6871 2510
8c5d388a 2511#ifdef SUPPORT_I18N
250b6871
JH
2512 {
2513 BOOL b = allow_utf8_domains;
2514 allow_utf8_domains = TRUE;
2515#endif
059ec3d9
PH
2516 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2517 &domain, FALSE);
2518
8c5d388a 2519#ifdef SUPPORT_I18N
250b6871
JH
2520 if (string_is_utf8(recipient))
2521 message_smtputf8 = TRUE;
2522 else
2523 allow_utf8_domains = b;
2524 }
2525#endif
2526
059ec3d9
PH
2527 /* Keep a list of all the bad addresses so we can send a single
2528 error message at the end. However, an empty address is not an error;
2529 just ignore it. This can come from an empty group list like
2530
2531 To: Recipients of list:;
2532
2533 If there are no recipients at all, an error will occur later. */
2534
2535 if (recipient == NULL && Ustrcmp(errmess, "empty address") != 0)
2536 {
2537 int len = Ustrlen(s);
2538 error_block *b = store_get(sizeof(error_block));
2539 while (len > 0 && isspace(s[len-1])) len--;
2540 b->next = NULL;
2541 b->text1 = string_printing(string_copyn(s, len));
2542 b->text2 = errmess;
2543 *bnext = b;
2544 bnext = &(b->next);
2545 }
2546
2547 /* If the recipient is already in the nonrecipients tree, it must
2548 have appeared on the command line with the option extract_addresses_
2549 remove_arguments set. Do not add it to the recipients, and keep a note
2550 that this has happened, in order to give a better error if there are
2551 no recipients left. */
2552
2553 else if (recipient != NULL)
2554 {
2555 if (tree_search(tree_nonrecipients, recipient) == NULL)
2556 receive_add_recipient(recipient, -1);
2557 else
2558 extracted_ignored = TRUE;
2559 }
2560
2561 /* Move on past this address */
2562
2563 s = ss + (*ss? 1:0);
2564 while (isspace(*s)) s++;
1eccaa59
PH
2565 } /* Next address */
2566
8768d548
JH
2567 f.parse_allow_group = FALSE; /* Reset group syntax flags */
2568 f.parse_found_group = FALSE;
059ec3d9
PH
2569
2570 /* If this was the bcc: header, mark it "old", which means it
2571 will be kept on the spool, but not transmitted as part of the
2572 message. */
2573
2cbb4081 2574 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2575 } /* For appropriate header line */
2576 } /* For each header line */
2577
059ec3d9
PH
2578 }
2579
2580/* Now build the unique message id. This has changed several times over the
2581lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2582Retaining all the history in the comment has become too unwieldy - read
2583previous release sources if you want it.
2584
2585The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2586The first part is the current time, in seconds. The second part is the current
2587pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2588can hold a number in the range 0-3843. It used to be a computed sequence
2589number, but is now the fractional component of the current time in units of
25901/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2591received, Exim ensures that the timer has ticked at the appropriate level
2592before proceeding, to avoid duplication if the pid happened to be re-used
2593within the same time period. It seems likely that most messages will take at
2594least half a millisecond to be received, so no delay will normally be
2595necessary. At least for some time...
2596
2597There is a modification when localhost_number is set. Formerly this was allowed
2598to be as large as 255. Now it is restricted to the range 0-16, and the final
2599component of the message id becomes (localhost_number * 200) + fractional time
2600in units of 1/200 of a second (i.e. a value in the range 0-3399).
2601
2602Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2603Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2604still allows the tttttt field to hold a large enough number to last for some
2605more decades, and the final two-digit field can hold numbers up to 1295, which
2606is enough for milliseconds (instead of 1/2000 of a second).
2607
2608However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2609pid, so it is probably safe because pids have to be positive. The
2610localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2611final field becomes (localhost_number * 100) + fractional time in centiseconds.
2612
2613Note that string_base62() returns its data in a static storage block, so it
2614must be copied before calling string_base62() again. It always returns exactly
26156 characters.
2616
2617There doesn't seem to be anything in the RFC which requires a message id to
2618start with a letter, but Smail was changed to ensure this. The external form of
2619the message id (as supplied by string expansion) therefore starts with an
2620additional leading 'E'. The spool file names do not include this leading
2621letter and it is not used internally.
2622
2623NOTE: If ever the format of message ids is changed, the regular expression for
2624checking that a string is in this format must be updated in a corresponding
2625way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2540f2f8
JH
2626must also be changed to reflect the correct string length. The queue-sort code
2627needs to know the layout. Then, of course, other programs that rely on the
2628message id format will need updating too. */
059ec3d9
PH
2629
2630Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2631message_id[6] = '-';
2632Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2633
2634/* Deal with the case where the host number is set. The value of the number was
2635checked when it was read, to ensure it isn't too big. The timing granularity is
2636left in id_resolution so that an appropriate wait can be done after receiving
2637the message, if necessary (we hope it won't be). */
2638
306c6c77 2639if (host_number_string)
059ec3d9 2640 {
099afc4f 2641 id_resolution = BASE_62 == 62 ? 5000 : 10000;
059ec3d9
PH
2642 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2643 string_base62((long int)(
2644 host_number * (1000000/id_resolution) +
2645 message_id_tv.tv_usec/id_resolution)) + 4);
2646 }
2647
2648/* Host number not set: final field is just the fractional time at an
2649appropriate resolution. */
2650
2651else
2652 {
099afc4f 2653 id_resolution = BASE_62 == 62 ? 500 : 1000;
059ec3d9
PH
2654 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2655 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2656 }
2657
2658/* Add the current message id onto the current process info string if
2659it will fit. */
2660
2661(void)string_format(process_info + process_info_len,
2662 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2663
2664/* If we are using multiple input directories, set up the one for this message
2665to be the least significant base-62 digit of the time of arrival. Otherwise
2666ensure that it is an empty string. */
2667
a2da3176 2668message_subdir[0] = split_spool_directory ? message_id[5] : 0;
059ec3d9
PH
2669
2670/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2671one, but only for local (without suppress_local_fixups) or submission mode
2672messages. This can be user-configured if required, but we had better flatten
2673any illegal characters therein. */
059ec3d9 2674
306c6c77 2675if ( !msgid_header
8768d548 2676 && ((!sender_host_address && !f.suppress_local_fixups) || f.submission_mode))
059ec3d9
PH
2677 {
2678 uschar *p;
2679 uschar *id_text = US"";
2680 uschar *id_domain = primary_hostname;
2681
2682 /* Permit only letters, digits, dots, and hyphens in the domain */
2683
306c6c77 2684 if (message_id_domain)
059ec3d9
PH
2685 {
2686 uschar *new_id_domain = expand_string(message_id_domain);
306c6c77 2687 if (!new_id_domain)
059ec3d9 2688 {
8768d548 2689 if (!f.expand_string_forcedfail)
059ec3d9
PH
2690 log_write(0, LOG_MAIN|LOG_PANIC,
2691 "expansion of \"%s\" (message_id_header_domain) "
2692 "failed: %s", message_id_domain, expand_string_message);
2693 }
306c6c77 2694 else if (*new_id_domain)
059ec3d9
PH
2695 {
2696 id_domain = new_id_domain;
306c6c77 2697 for (p = id_domain; *p; p++)
059ec3d9
PH
2698 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2699 }
2700 }
2701
2702 /* Permit all characters except controls and RFC 2822 specials in the
2703 additional text part. */
2704
306c6c77 2705 if (message_id_text)
059ec3d9
PH
2706 {
2707 uschar *new_id_text = expand_string(message_id_text);
306c6c77 2708 if (!new_id_text)
059ec3d9 2709 {
8768d548 2710 if (!f.expand_string_forcedfail)
059ec3d9
PH
2711 log_write(0, LOG_MAIN|LOG_PANIC,
2712 "expansion of \"%s\" (message_id_header_text) "
2713 "failed: %s", message_id_text, expand_string_message);
2714 }
306c6c77 2715 else if (*new_id_text)
059ec3d9
PH
2716 {
2717 id_text = new_id_text;
306c6c77 2718 for (p = id_text; *p; p++) if (mac_iscntrl_or_special(*p)) *p = '-';
059ec3d9
PH
2719 }
2720 }
2721
e7e680d6
PP
2722 /* Add the header line
2723 * Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2724 * appended, to preserve classical expectations of header ordering. */
059ec3d9 2725
e7e680d6 2726 header_add_at_position(!resents_exist, NULL, FALSE, htype_id,
5eb690a1
NM
2727 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
2728 (*id_text == 0)? "" : ".", id_text, id_domain);
059ec3d9
PH
2729 }
2730
2731/* If we are to log recipients, keep a copy of the raw ones before any possible
2732rewriting. Must copy the count, because later ACLs and the local_scan()
2733function may mess with the real recipients. */
2734
6c6d6e48 2735if (LOGGING(received_recipients))
059ec3d9
PH
2736 {
2737 raw_recipients = store_get(recipients_count * sizeof(uschar *));
2738 for (i = 0; i < recipients_count; i++)
2739 raw_recipients[i] = string_copy(recipients_list[i].address);
2740 raw_recipients_count = recipients_count;
2741 }
2742
2743/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2744recipients will get here only if the conditions were right (allow_unqualified_
2745recipient is TRUE). */
2746
2747for (i = 0; i < recipients_count; i++)
2748 recipients_list[i].address =
2749 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2750 global_rewrite_rules, rewrite_existflags);
2751
8800895a
PH
2752/* If there is no From: header, generate one for local (without
2753suppress_local_fixups) or submission_mode messages. If there is no sender
2754address, but the sender is local or this is a local delivery error, use the
2755originator login. This shouldn't happen for genuine bounces, but might happen
2756for autoreplies. The addition of From: must be done *before* checking for the
2757possible addition of a Sender: header, because untrusted_set_sender allows an
2758untrusted user to set anything in the envelope (which might then get info
2759From:) but we still want to ensure a valid Sender: if it is required. */
2760
306c6c77 2761if ( !from_header
8768d548 2762 && ((!sender_host_address && !f.suppress_local_fixups) || f.submission_mode))
059ec3d9 2763 {
2fe1a124
PH
2764 uschar *oname = US"";
2765
2766 /* Use the originator_name if this is a locally submitted message and the
2767 caller is not trusted. For trusted callers, use it only if -F was used to
2768 force its value or if we have a non-SMTP message for which -f was not used
2769 to set the sender. */
2770
306c6c77 2771 if (!sender_host_address)
2fe1a124 2772 {
8768d548
JH
2773 if (!f.trusted_caller || f.sender_name_forced ||
2774 (!smtp_input && !f.sender_address_forced))
2fe1a124
PH
2775 oname = originator_name;
2776 }
2777
2778 /* For non-locally submitted messages, the only time we use the originator
2779 name is when it was forced by the /name= option on control=submission. */
2780
306c6c77 2781 else if (submission_name) oname = submission_name;
2fe1a124 2782
059ec3d9
PH
2783 /* Envelope sender is empty */
2784
306c6c77 2785 if (!*sender_address)
059ec3d9 2786 {
87ba3f5f
PH
2787 uschar *fromstart, *fromend;
2788
306c6c77
JH
2789 fromstart = string_sprintf("%sFrom: %s%s",
2790 resent_prefix, oname, *oname ? " <" : "");
2791 fromend = *oname ? US">" : US"";
87ba3f5f 2792
8768d548 2793 if (f.sender_local || f.local_error_message)
87ba3f5f
PH
2794 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2795 local_part_quote(originator_login), qualify_domain_sender,
2796 fromend);
306c6c77 2797
8768d548 2798 else if (f.submission_mode && authenticated_id)
059ec3d9 2799 {
306c6c77 2800 if (!submission_domain)
87ba3f5f
PH
2801 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2802 local_part_quote(authenticated_id), qualify_domain_sender,
2803 fromend);
306c6c77
JH
2804
2805 else if (!*submission_domain) /* empty => whole address set */
87ba3f5f
PH
2806 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2807 fromend);
306c6c77 2808
059ec3d9 2809 else
87ba3f5f 2810 header_add(htype_from, "%s%s@%s%s\n", fromstart,
306c6c77
JH
2811 local_part_quote(authenticated_id), submission_domain, fromend);
2812
059ec3d9
PH
2813 from_header = header_last; /* To get it checked for Sender: */
2814 }
2815 }
2816
2817 /* There is a non-null envelope sender. Build the header using the original
2818 sender address, before any rewriting that might have been done while
2819 verifying it. */
2820
2821 else
2822 {
87ba3f5f 2823 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124 2824 oname,
306c6c77
JH
2825 *oname ? " <" : "",
2826 sender_address_unrewritten ? sender_address_unrewritten : sender_address,
2827 *oname ? ">" : "");
059ec3d9
PH
2828
2829 from_header = header_last; /* To get it checked for Sender: */
2830 }
2831 }
2832
2833
8800895a
PH
2834/* If the sender is local (without suppress_local_fixups), or if we are in
2835submission mode and there is an authenticated_id, check that an existing From:
2836is correct, and if not, generate a Sender: header, unless disabled. Any
2837previously-existing Sender: header was removed above. Note that sender_local,
2838as well as being TRUE if the caller of exim is not trusted, is also true if a
2839trusted caller did not supply a -f argument for non-smtp input. To allow
2840trusted callers to forge From: without supplying -f, we have to test explicitly
2841here. If the From: header contains more than one address, then the call to
2842parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9 2843
306c6c77 2844if ( from_header
8768d548
JH
2845 && ( f.active_local_from_check
2846 && ( f.sender_local && !f.trusted_caller && !f.suppress_local_fixups
2847 || f.submission_mode && authenticated_id
306c6c77 2848 ) ) )
059ec3d9
PH
2849 {
2850 BOOL make_sender = TRUE;
2851 int start, end, domain;
2852 uschar *errmess;
2853 uschar *from_address =
2854 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2855 &start, &end, &domain, FALSE);
2856 uschar *generated_sender_address;
2857
8768d548 2858 generated_sender_address = f.submission_mode
306c6c77
JH
2859 ? !submission_domain
2860 ? string_sprintf("%s@%s",
2861 local_part_quote(authenticated_id), qualify_domain_sender)
2862 : !*submission_domain /* empty => full address */
2863 ? string_sprintf("%s", authenticated_id)
2864 : string_sprintf("%s@%s",
2865 local_part_quote(authenticated_id), submission_domain)
2866 : string_sprintf("%s@%s",
2867 local_part_quote(originator_login), qualify_domain_sender);
059ec3d9
PH
2868
2869 /* Remove permitted prefixes and suffixes from the local part of the From:
2870 address before doing the comparison with the generated sender. */
2871
306c6c77 2872 if (from_address)
059ec3d9
PH
2873 {
2874 int slen;
306c6c77 2875 uschar *at = domain ? from_address + domain - 1 : NULL;
059ec3d9 2876
306c6c77 2877 if (at) *at = 0;
059ec3d9
PH
2878 from_address += route_check_prefix(from_address, local_from_prefix);
2879 slen = route_check_suffix(from_address, local_from_suffix);
2880 if (slen > 0)
2881 {
2882 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2883 from_address += slen;
2884 }
306c6c77 2885 if (at) *at = '@';
059ec3d9 2886
306c6c77
JH
2887 if ( strcmpic(generated_sender_address, from_address) == 0
2888 || (!domain && strcmpic(from_address, originator_login) == 0))
059ec3d9
PH
2889 make_sender = FALSE;
2890 }
2891
2892 /* We have to cause the Sender header to be rewritten if there are
2893 appropriate rewriting rules. */
2894
2895 if (make_sender)
8768d548 2896 if (f.submission_mode && !submission_name)
059ec3d9
PH
2897 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2898 generated_sender_address);
2899 else
2900 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124 2901 resent_prefix,
8768d548 2902 f.submission_mode ? submission_name : originator_name,
2fe1a124 2903 generated_sender_address);
87ba3f5f
PH
2904
2905 /* Ensure that a non-null envelope sender address corresponds to the
2906 submission mode sender address. */
2907
8768d548 2908 if (f.submission_mode && *sender_address)
87ba3f5f 2909 {
306c6c77 2910 if (!sender_address_unrewritten)
87ba3f5f
PH
2911 sender_address_unrewritten = sender_address;
2912 sender_address = generated_sender_address;
089793a4
TF
2913 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2914 log_write(L_address_rewrite, LOG_MAIN,
2915 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2916 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2917 }
059ec3d9
PH
2918 }
2919
059ec3d9
PH
2920/* If there are any rewriting rules, apply them to the sender address, unless
2921it has already been rewritten as part of verification for SMTP input. */
2922
306c6c77 2923if (global_rewrite_rules && !sender_address_unrewritten && *sender_address)
059ec3d9
PH
2924 {
2925 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2926 global_rewrite_rules, rewrite_existflags);
2927 DEBUG(D_receive|D_rewrite)
2928 debug_printf("rewritten sender = %s\n", sender_address);
2929 }
2930
2931
2932/* The headers must be run through rewrite_header(), because it ensures that
2933addresses are fully qualified, as well as applying any rewriting rules that may
2934exist.
2935
2936Qualification of header addresses in a message from a remote host happens only
2937if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2938appropriate. For local messages, qualification always happens, unless -bnq is
2939used to explicitly suppress it. No rewriting is done for an unqualified address
2940that is left untouched.
2941
2942We start at the second header, skipping our own Received:. This rewriting is
2943documented as happening *after* recipient addresses are taken from the headers
2944by the -t command line option. An added Sender: gets rewritten here. */
2945
1ebe15c3 2946for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2947 {
2948 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2949 rewrite_existflags, TRUE);
1ebe15c3 2950 if (newh) h = newh;
059ec3d9
PH
2951 }
2952
2953
2954/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2955"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
2956"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
2957exists, this applies to the set of resent- headers rather than the normal set.
2958
2cbb4081
PH
2959The requirement for a recipient header has been removed in RFC 2822. At this
2960point in the code, earlier versions of Exim added a To: header for locally
2961submitted messages, and an empty Bcc: header for others. In the light of the
2962changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 2963
059ec3d9
PH
2964
2965/* If there is no date header, generate one if the message originates locally
8800895a
PH
2966(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
2967submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
2968to be more confusing if Exim adds one to all remotely-originated messages.
2969As per Message-Id, we prepend if resending, else append.
2970*/
059ec3d9 2971
306c6c77 2972if ( !date_header_exists
8768d548 2973 && ((!sender_host_address && !f.suppress_local_fixups) || f.submission_mode))
e7e680d6
PP
2974 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
2975 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
2976
2977search_tidyup(); /* Free any cached resources */
2978
2979/* Show the complete set of headers if debugging. Note that the first one (the
2980new Received:) has not yet been set. */
2981
2982DEBUG(D_receive)
2983 {
2984 debug_printf(">>Headers after rewriting and local additions:\n");
306c6c77 2985 for (h = header_list->next; h; h = h->next)
059ec3d9
PH
2986 debug_printf("%c %s", h->type, h->text);
2987 debug_printf("\n");
2988 }
2989
2990/* The headers are now complete in store. If we are running in filter
2991testing mode, that is all this function does. Return TRUE if the message
2992ended with a dot. */
2993
f05da2e8 2994if (filter_test != FTEST_NONE)
059ec3d9
PH
2995 {
2996 process_info[process_info_len] = 0;
2997 return message_ended == END_DOT;
2998 }
2999
7e3ce68e
JH
3000/*XXX CHUNKING: need to cancel cutthrough under BDAT, for now. In future,
3001think more if it could be handled. Cannot do onward CHUNKING unless
3002inbound is, but inbound chunking ought to be ok with outbound plain.
3003Could we do onward CHUNKING given inbound CHUNKING?
3004*/
3005if (chunking_state > CHUNKING_OFFERED)
57cc2785 3006 cancel_cutthrough_connection(FALSE, US"chunking active");
7e3ce68e 3007
817d9f57 3008/* Cutthrough delivery:
5032d1cf
JH
3009We have to create the Received header now rather than at the end of reception,
3010so the timestamp behaviour is a change to the normal case.
5032d1cf 3011Having created it, send the headers to the destination. */
57cc2785 3012
74f1a423 3013if (cutthrough.cctx.sock >= 0 && cutthrough.delivery)
e4bdf652 3014 {
817d9f57
JH
3015 if (received_count > received_headers_max)
3016 {
57cc2785 3017 cancel_cutthrough_connection(TRUE, US"too many headers");
817d9f57
JH
3018 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
3019 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3020 "Too many \"Received\" headers",
3021 sender_address,
57cc2785
JH
3022 sender_fullhost ? "H=" : "", sender_fullhost ? sender_fullhost : US"",
3023 sender_ident ? "U=" : "", sender_ident ? sender_ident : US"");
817d9f57
JH
3024 message_id[0] = 0; /* Indicate no message accepted */
3025 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
3026 goto TIDYUP; /* Skip to end of function */
3027 }
e4bdf652 3028 received_header_gen();
578d43dc 3029 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652
JH
3030 (void) cutthrough_headers_send();
3031 }
61147df4 3032
e4bdf652 3033
059ec3d9
PH
3034/* Open a new spool file for the data portion of the message. We need
3035to access it both via a file descriptor and a stream. Try to make the
41313d92 3036directory if it isn't there. */
059ec3d9 3037
41313d92 3038spool_name = spool_fname(US"input", message_subdir, message_id, US"-D");
a2da3176
JH
3039DEBUG(D_receive) debug_printf("Data file name: %s\n", spool_name);
3040
3041if ((data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE)) < 0)
059ec3d9
PH
3042 {
3043 if (errno == ENOENT)
3044 {
0971ec06 3045 (void) directory_make(spool_directory,
41313d92
JH
3046 spool_sname(US"input", message_subdir),
3047 INPUT_DIRECTORY_MODE, TRUE);
059ec3d9
PH
3048 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
3049 }
3050 if (data_fd < 0)
3051 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
3052 spool_name, strerror(errno));
3053 }
3054
3055/* Make sure the file's group is the Exim gid, and double-check the mode
3056because the group setting doesn't always get set automatically. */
3057
1ac6b2e7
JH
3058if (fchown(data_fd, exim_uid, exim_gid))
3059 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
3060 "Failed setting ownership on spool file %s: %s",
3061 spool_name, strerror(errno));
ff790e47 3062(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
3063
3064/* We now have data file open. Build a stream for it and lock it. We lock only
3065the first line of the file (containing the message ID) because otherwise there
3066are problems when Exim is run under Cygwin (I'm told). See comments in
3067spool_in.c, where the same locking is done. */
3068
1bd642c2 3069spool_data_file = fdopen(data_fd, "w+");
059ec3d9
PH
3070lock_data.l_type = F_WRLCK;
3071lock_data.l_whence = SEEK_SET;
3072lock_data.l_start = 0;
3073lock_data.l_len = SPOOL_DATA_START_OFFSET;
3074
3075if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
3076 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
3077 errno, strerror(errno));
3078
3079/* We have an open, locked data file. Write the message id to it to make it
3080self-identifying. Then read the remainder of the input of this message and
3081write it to the data file. If the variable next != NULL, it contains the first
3082data line (which was read as a header but then turned out not to have the right
3083format); write it (remembering that it might contain binary zeros). The result
3084of fwrite() isn't inspected; instead we call ferror() below. */
3085
1bd642c2 3086fprintf(spool_data_file, "%s-D\n", message_id);
306c6c77 3087if (next)
059ec3d9
PH
3088 {
3089 uschar *s = next->text;
3090 int len = next->slen;
1bd642c2 3091 if (fwrite(s, 1, len, spool_data_file) == len) /* "if" for compiler quietening */
cfbb0d24 3092 body_linecount++; /* Assumes only 1 line */
059ec3d9
PH
3093 }
3094
3095/* Note that we might already be at end of file, or the logical end of file
3096(indicated by '.'), or might have encountered an error while writing the
3097message id or "next" line. */
3098
1bd642c2 3099if (!ferror(spool_data_file) && !(receive_feof)() && message_ended != END_DOT)
059ec3d9
PH
3100 {
3101 if (smtp_input)
3102 {
328c5688 3103 message_ended = chunking_state <= CHUNKING_OFFERED
1bd642c2 3104 ? read_message_data_smtp(spool_data_file)
328c5688 3105 : spool_wireformat
1bd642c2
JH
3106 ? read_message_bdat_smtp_wire(spool_data_file)
3107 : read_message_bdat_smtp(spool_data_file);
059ec3d9
PH
3108 receive_linecount++; /* The terminating "." line */
3109 }
cfbb0d24 3110 else
1bd642c2 3111 message_ended = read_message_data(spool_data_file);
059ec3d9
PH
3112
3113 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 3114 message_linecount += body_linecount;
059ec3d9 3115
7e3ce68e 3116 switch (message_ended)
059ec3d9 3117 {
7e3ce68e 3118 /* Handle premature termination of SMTP */
059ec3d9 3119
7e3ce68e
JH
3120 case END_EOF:
3121 if (smtp_input)
3122 {
3123 Uunlink(spool_name); /* Lose data file when closed */
57cc2785 3124 cancel_cutthrough_connection(TRUE, US"sender closed connection");
7e3ce68e
JH
3125 message_id[0] = 0; /* Indicate no message accepted */
3126 smtp_reply = handle_lost_connection(US"");
3127 smtp_yield = FALSE;
3128 goto TIDYUP; /* Skip to end of function */
3129 }
3130 break;
059ec3d9 3131
7e3ce68e
JH
3132 /* Handle message that is too big. Don't use host_or_ident() in the log
3133 message; we want to see the ident value even for non-remote messages. */
059ec3d9 3134
7e3ce68e
JH
3135 case END_SIZE:
3136 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3137 cancel_cutthrough_connection(TRUE, US"mail too big");
7e3ce68e 3138 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
059ec3d9 3139
7e3ce68e
JH
3140 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3141 "message too big: read=%d max=%d",
3142 sender_address,
306c6c77
JH
3143 sender_fullhost ? " H=" : "",
3144 sender_fullhost ? sender_fullhost : US"",
3145 sender_ident ? " U=" : "",
3146 sender_ident ? sender_ident : US"",
7e3ce68e
JH
3147 message_size,
3148 thismessage_size_limit);
3149
3150 if (smtp_input)
3151 {
3152 smtp_reply = US"552 Message size exceeds maximum permitted";
3153 message_id[0] = 0; /* Indicate no message accepted */
3154 goto TIDYUP; /* Skip to end of function */
3155 }
3156 else
3157 {
1bd642c2 3158 fseek(spool_data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
7e3ce68e
JH
3159 give_local_error(ERRMESS_TOOBIG,
3160 string_sprintf("message too big (max=%d)", thismessage_size_limit),
1bd642c2 3161 US"message rejected: ", error_rc, spool_data_file, header_list);
7e3ce68e
JH
3162 /* Does not return */
3163 }
3164 break;
3165
3166 /* Handle bad BDAT protocol sequence */
3167
3168 case END_PROTOCOL:
3169 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3170 cancel_cutthrough_connection(TRUE, US"sender protocol error");
7e3ce68e
JH
3171 smtp_reply = US""; /* Response already sent */
3172 message_id[0] = 0; /* Indicate no message accepted */
3173 goto TIDYUP; /* Skip to end of function */
059ec3d9
PH
3174 }
3175 }
3176
3177/* Restore the standard SIGALRM handler for any subsequent processing. (For
3178example, there may be some expansion in an ACL that uses a timer.) */
3179
3180os_non_restarting_signal(SIGALRM, sigalrm_handler);
3181
3182/* The message body has now been read into the data file. Call fflush() to
3183empty the buffers in C, and then call fsync() to get the data written out onto
3184the disk, as fflush() doesn't do this (or at least, it isn't documented as
3185having to do this). If there was an I/O error on either input or output,
3186attempt to send an error message, and unlink the spool file. For non-SMTP input
3187we can then give up. Note that for SMTP input we must swallow the remainder of
3188the input in cases of output errors, since the far end doesn't expect to see
3189anything until the terminating dot line is sent. */
3190
1bd642c2
JH
3191if (fflush(spool_data_file) == EOF || ferror(spool_data_file) ||
3192 EXIMfsync(fileno(spool_data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
3193 {
3194 uschar *msg_errno = US strerror(errno);
3195 BOOL input_error = (receive_ferror)() != 0;
3196 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
3197 input_error? "Input read" : "Spool write",
3198 msg_errno,
306c6c77 3199 sender_fullhost ? sender_fullhost : sender_ident);
059ec3d9
PH
3200
3201 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
3202 Uunlink(spool_name); /* Lose the data file */
57cc2785 3203 cancel_cutthrough_connection(TRUE, US"error writing spoolfile");
059ec3d9
PH
3204
3205 if (smtp_input)
3206 {
3207 if (input_error)
3208 smtp_reply = US"451 Error while reading input data";
3209 else
3210 {
3211 smtp_reply = US"451 Error while writing spool file";
3212 receive_swallow_smtp();
3213 }
3214 message_id[0] = 0; /* Indicate no message accepted */
3215 goto TIDYUP; /* Skip to end of function */
3216 }
3217
3218 else
3219 {
1bd642c2
JH
3220 fseek(spool_data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3221 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, spool_data_file,
059ec3d9
PH
3222 header_list);
3223 /* Does not return */
3224 }
3225 }
3226
3227
3228/* No I/O errors were encountered while writing the data file. */
3229
3230DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
306c6c77 3231if (LOGGING(receive_time)) timesince(&received_time_taken, &received_time);
059ec3d9
PH
3232
3233
3234/* If there were any bad addresses extracted by -t, or there were no recipients
3235left after -t, send a message to the sender of this message, or write it to
3236stderr if the error handling option is set that way. Note that there may
3237legitimately be no recipients for an SMTP message if they have all been removed
3238by "discard".
3239
3240We need to rewind the data file in order to read it. In the case of no
3241recipients or stderr error writing, throw the data file away afterwards, and
3242exit. (This can't be SMTP, which always ensures there's at least one
3243syntactically good recipient address.) */
3244
306c6c77 3245if (extract_recip && (bad_addresses || recipients_count == 0))
059ec3d9
PH
3246 {
3247 DEBUG(D_receive)
3248 {
3249 if (recipients_count == 0) debug_printf("*** No recipients\n");
306c6c77 3250 if (bad_addresses)
059ec3d9 3251 {
569a8b23 3252 error_block * eblock;
059ec3d9 3253 debug_printf("*** Bad address(es)\n");
569a8b23 3254 for (eblock = bad_addresses; eblock; eblock = eblock->next)
059ec3d9 3255 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
059ec3d9
PH
3256 }
3257 }
3258
569a8b23
JH
3259 log_write(0, LOG_MAIN|LOG_PANIC, "%s %s found in headers",
3260 message_id, bad_addresses ? "bad addresses" : "no recipients");
3261
1bd642c2 3262 fseek(spool_data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
059ec3d9
PH
3263
3264 /* If configured to send errors to the sender, but this fails, force
3265 a failure error code. We use a special one for no recipients so that it
3266 can be detected by the autoreply transport. Otherwise error_rc is set to
3267 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
3268 it is EXIT_SUCCESS. */
3269
3270 if (error_handling == ERRORS_SENDER)
3271 {
3272 if (!moan_to_sender(
569a8b23
JH
3273 bad_addresses
3274 ? recipients_list ? ERRMESS_BADADDRESS : ERRMESS_BADNOADDRESS
3275 : extracted_ignored ? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS,
3276 bad_addresses, header_list, spool_data_file, FALSE
3277 ) )
3278 error_rc = bad_addresses ? EXIT_FAILURE : EXIT_NORECIPIENTS;
059ec3d9
PH
3279 }
3280 else
3281 {
306c6c77 3282 if (!bad_addresses)
059ec3d9
PH
3283 if (extracted_ignored)
3284 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3285 else
3286 fprintf(stderr, "exim: no recipients in message\n");
059ec3d9
PH
3287 else
3288 {
3289 fprintf(stderr, "exim: invalid address%s",
cfbb0d24
JH
3290 bad_addresses->next ? "es:\n" : ":");
3291 for ( ; bad_addresses; bad_addresses = bad_addresses->next)
059ec3d9
PH
3292 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3293 bad_addresses->text2);
059ec3d9
PH
3294 }
3295 }
3296
3297 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3298 {
3299 Uunlink(spool_name);
1bd642c2 3300 (void)fclose(spool_data_file);
9bfb7e1b 3301 exim_exit(error_rc, US"receiving");
059ec3d9
PH
3302 }
3303 }
3304
3305/* Data file successfully written. Generate text for the Received: header by
3306expanding the configured string, and adding a timestamp. By leaving this
3307operation till now, we ensure that the timestamp is the time that message
3308reception was completed. However, this is deliberately done before calling the
3309data ACL and local_scan().
3310
3311This Received: header may therefore be inspected by the data ACL and by code in
3312the local_scan() function. When they have run, we update the timestamp to be
3313the final time of reception.
3314
3315If there is just one recipient, set up its value in the $received_for variable
3316for use when we generate the Received: header.
3317
3318Note: the checking for too many Received: headers is handled by the delivery
3319code. */
e4bdf652 3320/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3321
306c6c77 3322if (!received_header->text) /* Non-cutthrough case */
059ec3d9 3323 {
e4bdf652 3324 received_header_gen();
059ec3d9 3325
e4bdf652 3326 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3327
e4bdf652
JH
3328 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3329 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3330
e4bdf652
JH
3331 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3332 now, before running the DATA ACL. */
059ec3d9 3333
578d43dc 3334 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652 3335 }
817d9f57 3336else
e4bdf652
JH
3337 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3338 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3339
3340/* If an ACL is specified for checking things at this stage of reception of a
3341message, run it, unless all the recipients were removed by "discard" in earlier
3342ACLs. That is the only case in which recipients_count can be zero at this
3343stage. Set deliver_datafile to point to the data file so that $message_body and
3344$message_body_end can be extracted if needed. Allow $recipients in expansions.
3345*/
3346
3347deliver_datafile = data_fd;
4e88a19f 3348user_msg = NULL;
059ec3d9 3349
8768d548 3350f.enable_dollar_recipients = TRUE;
0e20aff9 3351
059ec3d9 3352if (recipients_count == 0)
8768d548 3353 blackholed_by = f.recipients_discarded ? US"MAIL ACL" : US"RCPT ACL";
7e3ce68e 3354
059ec3d9
PH
3355else
3356 {
059ec3d9
PH
3357 /* Handle interactive SMTP messages */
3358
3359 if (smtp_input && !smtp_batched_input)
3360 {
8523533c 3361
80a47a2c 3362#ifndef DISABLE_DKIM
8768d548 3363 if (!f.dkim_disable_verify)
80a47a2c 3364 {
cc55f420 3365 /* Finish verification */
80a47a2c
TK
3366 dkim_exim_verify_finish();
3367
3368 /* Check if we must run the DKIM ACL */
7e3ce68e 3369 if (acl_smtp_dkim && dkim_verify_signers && *dkim_verify_signers)
80a47a2c 3370 {
cc55f420 3371 uschar * dkim_verify_signers_expanded =
80a47a2c 3372 expand_string(dkim_verify_signers);
cc55f420
JH
3373 gstring * results = NULL;
3374 int signer_sep = 0;
3375 const uschar * ptr;
3376 uschar * item;
3377 gstring * seen_items = NULL;
3378 int old_pool = store_pool;
3379
3380 store_pool = POOL_PERM; /* Allow created variables to live to data ACL */
3381
3382 if (!(ptr = dkim_verify_signers_expanded))
80a47a2c
TK
3383 log_write(0, LOG_MAIN|LOG_PANIC,
3384 "expansion of dkim_verify_signers option failed: %s",
3385 expand_string_message);
7e3ce68e 3386
cc55f420
JH
3387 /* Default to OK when no items are present */
3388 rc = OK;
3389 while ((item = string_nextinlist(&ptr, &signer_sep, NULL, 0)))
3390 {
3391 /* Prevent running ACL for an empty item */
3392 if (!item || !*item) continue;
3393
3394 /* Only run ACL once for each domain or identity,
3395 no matter how often it appears in the expanded list. */
3396 if (seen_items)
3397 {
3398 uschar * seen_item;
3399 const uschar * seen_items_list = string_from_gstring(seen_items);
3400 int seen_sep = ':';
3401 BOOL seen_this_item = FALSE;
3402
3403 while ((seen_item = string_nextinlist(&seen_items_list, &seen_sep,
3404 NULL, 0)))
3405 if (Ustrcmp(seen_item,item) == 0)
3406 {
3407 seen_this_item = TRUE;
3408 break;
3409 }
3410
3411 if (seen_this_item)
5032d1cf
JH
3412 {
3413 DEBUG(D_receive)
cc55f420
JH
3414 debug_printf("acl_smtp_dkim: skipping signer %s, "
3415 "already seen\n", item);
3416 continue;
5032d1cf 3417 }
cc55f420 3418
b2bcdd35 3419 seen_items = string_catn(seen_items, US":", 1);
cc55f420 3420 }
cc55f420
JH
3421 seen_items = string_cat(seen_items, item);
3422
18067c75 3423 rc = dkim_exim_acl_run(item, &results, &user_msg, &log_msg);
cc55f420
JH
3424 if (rc != OK)
3425 {
3426 DEBUG(D_receive)
3427 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, "
3428 "skipping remaining items\n", rc, item);
3429 cancel_cutthrough_connection(TRUE, US"dkim acl not ok");
3430 break;
3431 }
3432 }
3433 dkim_verify_status = string_from_gstring(results);
3434 store_pool = old_pool;
3435 add_acl_headers(ACL_WHERE_DKIM, US"DKIM");
3436 if (rc == DISCARD)
3437 {
3438 recipients_count = 0;
3439 blackholed_by = US"DKIM ACL";
3440 if (log_msg)
3441 blackhole_log_msg = string_sprintf(": %s", log_msg);
3442 }
3443 else if (rc != OK)
3444 {
3445 Uunlink(spool_name);
3446 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
3447 smtp_yield = FALSE; /* No more messages after dropped connection */
3448 smtp_reply = US""; /* Indicate reply already sent */
3449 message_id[0] = 0; /* Indicate no message accepted */
3450 goto TIDYUP; /* Skip to end of function */
3451 }
80a47a2c 3452 }
a79d8834
JH
3453 else
3454 dkim_exim_verify_log_all();
80a47a2c 3455 }
4a8ce2d8 3456#endif /* DISABLE_DKIM */
fb2274d4 3457
8523533c 3458#ifdef WITH_CONTENT_SCAN
5b6f7658
JH
3459 if ( recipients_count > 0
3460 && acl_smtp_mime
3461 && !run_mime_acl(acl_smtp_mime, &smtp_yield, &smtp_reply, &blackholed_by)
3462 )
54cdb463 3463 goto TIDYUP;
8523533c
TK
3464#endif /* WITH_CONTENT_SCAN */
3465
4840604e
TL
3466#ifdef EXPERIMENTAL_DMARC
3467 dmarc_up = dmarc_store_data(from_header);
3468#endif /* EXPERIMENTAL_DMARC */
3469
8ccd00b1
JH
3470#ifndef DISABLE_PRDR
3471 if (prdr_requested && recipients_count > 1 && acl_smtp_data_prdr)
fd98a5c6
JH
3472 {
3473 unsigned int c;
3474 int all_pass = OK;
3475 int all_fail = FAIL;
3476
925ac8e4 3477 smtp_printf("353 PRDR content analysis beginning\r\n", TRUE);
fd98a5c6
JH
3478 /* Loop through recipients, responses must be in same order received */
3479 for (c = 0; recipients_count > c; c++)
3480 {
3481 uschar * addr= recipients_list[c].address;
3482 uschar * msg= US"PRDR R=<%s> %s";
3483 uschar * code;
3484 DEBUG(D_receive)
3485 debug_printf("PRDR processing recipient %s (%d of %d)\n",
3486 addr, c+1, recipients_count);
3487 rc = acl_check(ACL_WHERE_PRDR, addr,
3488 acl_smtp_data_prdr, &user_msg, &log_msg);
3489
3490 /* If any recipient rejected content, indicate it in final message */
3491 all_pass |= rc;
3492 /* If all recipients rejected, indicate in final message */
3493 all_fail &= rc;
3494
3495 switch (rc)
3496 {
3497 case OK: case DISCARD: code = US"250"; break;
3498 case DEFER: code = US"450"; break;
3499 default: code = US"550"; break;
3500 }
3501 if (user_msg != NULL)
3502 smtp_user_msg(code, user_msg);
3503 else
3504 {
3505 switch (rc)
3506 {
3507 case OK: case DISCARD:
3508 msg = string_sprintf(CS msg, addr, "acceptance"); break;
3509 case DEFER:
3510 msg = string_sprintf(CS msg, addr, "temporary refusal"); break;
3511 default:
3512 msg = string_sprintf(CS msg, addr, "refusal"); break;
3513 }
3514 smtp_user_msg(code, msg);
3515 }
3516 if (log_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, log_msg);
3517 else if (user_msg) log_write(0, LOG_MAIN, "PRDR %s %s", addr, user_msg);
112b6a93 3518 else log_write(0, LOG_MAIN, "%s", CS msg);
fd98a5c6
JH
3519
3520 if (rc != OK) { receive_remove_recipient(addr); c--; }
3521 }
3522 /* Set up final message, used if data acl gives OK */
3523 smtp_reply = string_sprintf("%s id=%s message %s",
3524 all_fail == FAIL ? US"550" : US"250",
3525 message_id,
3526 all_fail == FAIL
3527 ? US"rejected for all recipients"
3528 : all_pass == OK
3529 ? US"accepted"
3530 : US"accepted for some recipients");