Copyright updates:
[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 */
1e1ddfac 6/* Copyright (c) The Exim Maintainers 2020 */
059ec3d9
PH
7/* See the file NOTICE for conditions of use and distribution. */
8
9/* Code for receiving a message and setting up spool files. */
10
059ec3d9 11#include "exim.h"
9723f966 12#include <setjmp.h>
059ec3d9 13
6a8f9482
TK
14#ifdef EXPERIMENTAL_DCC
15extern int dcc_ok;
16#endif
17
1a2e76e1 18#ifdef SUPPORT_DMARC
c007c974 19# include "dmarc.h"
1a2e76e1 20#endif
4840604e 21
059ec3d9
PH
22/*************************************************
23* Local static variables *
24*************************************************/
25
059ec3d9 26static int data_fd = -1;
41313d92 27static uschar *spool_name = US"";
059ec3d9 28
cff70eb1 29enum CH_STATE {LF_SEEN, MID_LINE, CR_SEEN};
059ec3d9 30
9723f966
JH
31#ifdef HAVE_LOCAL_SCAN
32jmp_buf local_scan_env; /* error-handling context for local_scan */
33unsigned had_local_scan_crash;
34unsigned had_local_scan_timeout;
35#endif
36
059ec3d9
PH
37
38/*************************************************
39* Non-SMTP character reading functions *
40*************************************************/
41
42/* These are the default functions that are set up in the variables such as
43receive_getc initially. They just call the standard functions, passing stdin as
44the file. (When SMTP input is occurring, different functions are used by
45changing the pointer variables.) */
46
47int
bd8fbe36 48stdin_getc(unsigned lim)
059ec3d9 49{
9723f966
JH
50int c = getc(stdin);
51
52if (had_data_timeout)
53 {
54 fprintf(stderr, "exim: timed out while reading - message abandoned\n");
55 log_write(L_lost_incoming_connection,
56 LOG_MAIN, "timed out while reading local message");
57 receive_bomb_out(US"data-timeout", NULL); /* Does not return */
58 }
59if (had_data_sigint)
60 {
61 if (filter_test == FTEST_NONE)
62 {
63 fprintf(stderr, "\nexim: %s received - message abandoned\n",
64 had_data_sigint == SIGTERM ? "SIGTERM" : "SIGINT");
65 log_write(0, LOG_MAIN, "%s received while reading local message",
66 had_data_sigint == SIGTERM ? "SIGTERM" : "SIGINT");
67 }
68 receive_bomb_out(US"signal-exit", NULL); /* Does not return */
69 }
70return c;
059ec3d9
PH
71}
72
73int
74stdin_ungetc(int c)
75{
76return ungetc(c, stdin);
77}
78
79int
80stdin_feof(void)
81{
82return feof(stdin);
83}
84
85int
86stdin_ferror(void)
87{
88return ferror(stdin);
89}
90
91
92
93
94/*************************************************
95* Check that a set sender is allowed *
96*************************************************/
97
98/* This function is called when a local caller sets an explicit sender address.
99It checks whether this is permitted, which it is for trusted callers.
100Otherwise, it must match the pattern(s) in untrusted_set_sender.
101
102Arguments: the proposed sender address
103Returns: TRUE for a trusted caller
104 TRUE if the address has been set, untrusted_set_sender has been
105 set, and the address matches something in the list
106 FALSE otherwise
107*/
108
109BOOL
110receive_check_set_sender(uschar *newsender)
111{
112uschar *qnewsender;
8768d548 113if (f.trusted_caller) return TRUE;
36d295f1
JH
114if (!newsender || !untrusted_set_sender) return FALSE;
115qnewsender = Ustrchr(newsender, '@')
116 ? newsender : string_sprintf("%s@%s", newsender, qualify_domain_sender);
117return match_address_list_basic(qnewsender, CUSS &untrusted_set_sender, 0) == OK;
059ec3d9
PH
118}
119
120
121
122
123/*************************************************
5cb8cbc6 124* Read space info for a partition *
059ec3d9
PH
125*************************************************/
126
8e669ac1
PH
127/* This function is called by receive_check_fs() below, and also by string
128expansion for variables such as $spool_space. The field names for the statvfs
5cb8cbc6
PH
129structure are macros, because not all OS have F_FAVAIL and it seems tidier to
130have macros for F_BAVAIL and F_FILES as well. Some kinds of file system do not
131have inodes, and they return -1 for the number available.
059ec3d9 132
5cb8cbc6
PH
133Later: It turns out that some file systems that do not have the concept of
134inodes return 0 rather than -1. Such systems should also return 0 for the total
8e669ac1 135number of inodes, so we require that to be greater than zero before returning
5cb8cbc6 136an inode count.
059ec3d9 137
5cb8cbc6
PH
138Arguments:
139 isspool TRUE for spool partition, FALSE for log partition
140 inodeptr address of int to receive inode count; -1 if there isn't one
8e669ac1 141
5cb8cbc6 142Returns: available on-root space, in kilobytes
8e669ac1
PH
143 -1 for log partition if there isn't one
144
145All values are -1 if the STATFS functions are not available.
059ec3d9
PH
146*/
147
a45431fa 148int_eximarith_t
5cb8cbc6 149receive_statvfs(BOOL isspool, int *inodeptr)
059ec3d9
PH
150{
151#ifdef HAVE_STATFS
059ec3d9 152struct STATVFS statbuf;
ddf1b11a 153struct stat dummy;
5cb8cbc6
PH
154uschar *path;
155uschar *name;
156uschar buffer[1024];
059ec3d9 157
5cb8cbc6 158/* The spool directory must always exist. */
059ec3d9 159
5cb8cbc6 160if (isspool)
059ec3d9 161 {
8e669ac1
PH
162 path = spool_directory;
163 name = US"spool";
164 }
165
059ec3d9
PH
166/* Need to cut down the log file path to the directory, and to ignore any
167appearance of "syslog" in it. */
168
5cb8cbc6 169else
059ec3d9 170 {
059ec3d9 171 int sep = ':'; /* Not variable - outside scripts use */
55414b25 172 const uschar *p = log_file_path;
8e669ac1 173 name = US"log";
059ec3d9
PH
174
175 /* An empty log_file_path means "use the default". This is the same as an
176 empty item in a list. */
177
178 if (*p == 0) p = US":";
55414b25
JH
179 while ((path = string_nextinlist(&p, &sep, buffer, sizeof(buffer))))
180 if (Ustrcmp(path, "syslog") != 0)
181 break;
059ec3d9 182
5cb8cbc6
PH
183 if (path == NULL) /* No log files */
184 {
8e669ac1
PH
185 *inodeptr = -1;
186 return -1;
187 }
059ec3d9 188
8e669ac1
PH
189 /* An empty string means use the default, which is in the spool directory.
190 But don't just use the spool directory, as it is possible that the log
5cb8cbc6 191 subdirectory has been symbolically linked elsewhere. */
059ec3d9 192
8e669ac1 193 if (path[0] == 0)
059ec3d9 194 {
5cb8cbc6
PH
195 sprintf(CS buffer, CS"%s/log", CS spool_directory);
196 path = buffer;
8e669ac1
PH
197 }
198 else
059ec3d9 199 {
8e669ac1 200 uschar *cp;
5cb8cbc6 201 if ((cp = Ustrrchr(path, '/')) != NULL) *cp = 0;
8e669ac1 202 }
5cb8cbc6 203 }
8e669ac1 204
8f128379 205/* We now have the path; do the business */
5cb8cbc6
PH
206
207memset(&statbuf, 0, sizeof(statbuf));
208
209if (STATVFS(CS path, &statbuf) != 0)
ddf1b11a
JH
210 if (stat(CS path, &dummy) == -1 && errno == ENOENT)
211 { /* Can happen on first run after installation */
212 *inodeptr = -1;
213 return -1;
214 }
215 else
216 {
217 log_write(0, LOG_MAIN|LOG_PANIC, "cannot accept message: failed to stat "
218 "%s directory %s: %s", name, path, strerror(errno));
219 smtp_closedown(US"spool or log directory problem");
81022793 220 exim_exit(EXIT_FAILURE);
ddf1b11a 221 }
8e669ac1 222
5cb8cbc6
PH
223*inodeptr = (statbuf.F_FILES > 0)? statbuf.F_FAVAIL : -1;
224
225/* Disks are getting huge. Take care with computing the size in kilobytes. */
8e669ac1 226
a45431fa 227return (int_eximarith_t)(((double)statbuf.F_BAVAIL * (double)statbuf.F_FRSIZE)/1024.0);
5cb8cbc6 228
ddf1b11a 229#else
5cb8cbc6
PH
230/* Unable to find partition sizes in this environment. */
231
5cb8cbc6
PH
232*inodeptr = -1;
233return -1;
234#endif
235}
236
059ec3d9 237
059ec3d9 238
5cb8cbc6
PH
239
240/*************************************************
241* Check space on spool and log partitions *
242*************************************************/
243
244/* This function is called before accepting a message; if any thresholds are
245set, it checks them. If a message_size is supplied, it checks that there is
246enough space for that size plus the threshold - i.e. that the message won't
247reduce the space to the threshold. Not all OS have statvfs(); for those that
248don't, this function always returns TRUE. For some OS the old function and
249struct name statfs is used; that is handled by a macro, defined in exim.h.
250
251Arguments:
252 msg_size the (estimated) size of an incoming message
253
254Returns: FALSE if there isn't enough space, or if the information cannot
255 be obtained
256 TRUE if no check was done or there is enough space
257*/
258
259BOOL
260receive_check_fs(int msg_size)
261{
a45431fa
JH
262int_eximarith_t space;
263int inodes;
5cb8cbc6
PH
264
265if (check_spool_space > 0 || msg_size > 0 || check_spool_inodes > 0)
266 {
8e669ac1
PH
267 space = receive_statvfs(TRUE, &inodes);
268
059ec3d9 269 DEBUG(D_receive)
7434882d 270 debug_printf("spool directory space = " PR_EXIM_ARITH "K inodes = %d "
a45431fa 271 "check_space = " PR_EXIM_ARITH "K inodes = %d msg_size = %d\n",
5cb8cbc6 272 space, inodes, check_spool_space, check_spool_inodes, msg_size);
8e669ac1 273
dd6711f0
EL
274 if ( space >= 0 && space + msg_size / 1024 < check_spool_space
275 || inodes >= 0 && inodes < check_spool_inodes)
8e669ac1 276 {
8c513105
JH
277 log_write(0, LOG_MAIN, "spool directory space check failed: space="
278 PR_EXIM_ARITH " inodes=%d", space, inodes);
059ec3d9
PH
279 return FALSE;
280 }
281 }
282
5cb8cbc6
PH
283if (check_log_space > 0 || check_log_inodes > 0)
284 {
8e669ac1
PH
285 space = receive_statvfs(FALSE, &inodes);
286
5cb8cbc6 287 DEBUG(D_receive)
7434882d 288 debug_printf("log directory space = " PR_EXIM_ARITH "K inodes = %d "
a45431fa 289 "check_space = " PR_EXIM_ARITH "K inodes = %d\n",
5cb8cbc6 290 space, inodes, check_log_space, check_log_inodes);
8e669ac1 291
7434882d
JH
292 if ( space >= 0 && space < check_log_space
293 || inodes >= 0 && inodes < check_log_inodes)
8e669ac1 294 {
7434882d
JH
295 log_write(0, LOG_MAIN, "log directory space check failed: space=" PR_EXIM_ARITH
296 " inodes=%d", space, inodes);
5cb8cbc6
PH
297 return FALSE;
298 }
8e669ac1
PH
299 }
300
059ec3d9
PH
301return TRUE;
302}
303
304
305
306/*************************************************
307* Bomb out while reading a message *
308*************************************************/
309
310/* The common case of wanting to bomb out is if a SIGTERM or SIGINT is
311received, or if there is a timeout. A rarer case might be if the log files are
312screwed up and Exim can't open them to record a message's arrival. Handling
313that case is done by setting a flag to cause the log functions to call this
314function if there is an ultimate disaster. That is why it is globally
315accessible.
316
8f128379
PH
317Arguments:
318 reason text reason to pass to the not-quit ACL
319 msg default SMTP response to give if in an SMTP session
059ec3d9
PH
320Returns: it doesn't
321*/
322
323void
8f128379 324receive_bomb_out(uschar *reason, uschar *msg)
059ec3d9 325{
ead37e6c
PP
326 static BOOL already_bombing_out;
327/* The smtp_notquit_exit() below can call ACLs which can trigger recursive
328timeouts, if someone has something slow in their quit ACL. Since the only
329things we should be doing are to close down cleanly ASAP, on the second
330pass we also close down stuff that might be opened again, before bypassing
331the ACL call and exiting. */
332
059ec3d9
PH
333/* If spool_name is set, it contains the name of the data file that is being
334written. Unlink it before closing so that it cannot be picked up by a delivery
335process. Ensure that any header file is also removed. */
336
ead37e6c 337if (spool_name[0] != '\0')
059ec3d9
PH
338 {
339 Uunlink(spool_name);
340 spool_name[Ustrlen(spool_name) - 1] = 'H';
341 Uunlink(spool_name);
ead37e6c 342 spool_name[0] = '\0';
059ec3d9
PH
343 }
344
345/* Now close the file if it is open, either as a fd or a stream. */
346
1bd642c2 347if (spool_data_file)
ead37e6c 348 {
1bd642c2
JH
349 (void)fclose(spool_data_file);
350 spool_data_file = NULL;
9723f966
JH
351 }
352else if (data_fd >= 0)
353 {
ead37e6c
PP
354 (void)close(data_fd);
355 data_fd = -1;
356 }
059ec3d9 357
8f128379
PH
358/* Attempt to close down an SMTP connection tidily. For non-batched SMTP, call
359smtp_notquit_exit(), which runs the NOTQUIT ACL, if present, and handles the
360SMTP response. */
059ec3d9 361
ead37e6c 362if (!already_bombing_out)
059ec3d9 363 {
ead37e6c
PP
364 already_bombing_out = TRUE;
365 if (smtp_input)
366 {
367 if (smtp_batched_input)
368 moan_smtp_batch(NULL, "421 %s - message abandoned", msg); /* No return */
369 smtp_notquit_exit(reason, US"421", US"%s %s - closing connection.",
370 smtp_active_hostname, msg);
371 }
059ec3d9
PH
372 }
373
374/* Exit from the program (non-BSMTP cases) */
375
81022793 376exim_exit(EXIT_FAILURE);
059ec3d9
PH
377}
378
379
380/*************************************************
381* Data read timeout *
382*************************************************/
383
384/* Handler function for timeouts that occur while reading the data that
385comprises a message.
386
387Argument: the signal number
388Returns: nothing
389*/
390
391static void
392data_timeout_handler(int sig)
393{
9723f966 394had_data_timeout = sig;
059ec3d9
PH
395}
396
397
398
9723f966 399#ifdef HAVE_LOCAL_SCAN
059ec3d9
PH
400/*************************************************
401* local_scan() timeout *
402*************************************************/
403
404/* Handler function for timeouts that occur while running a local_scan()
9723f966
JH
405function. Posix recommends against calling longjmp() from a signal-handler,
406but the GCC manual says you can so we will, and trust that it's better than
407calling probably non-signal-safe funxtions during logging from within the
408handler, even with other compilers.
409
410See also https://cwe.mitre.org/data/definitions/745.html which also lists
411it as unsafe.
412
413This is all because we have no control over what might be written for a
414local-scan function, so cannot sprinkle had-signal checks after each
415call-site. At least with the default "do-nothing" function we won't
416ever get here.
059ec3d9
PH
417
418Argument: the signal number
419Returns: nothing
420*/
421
422static void
423local_scan_timeout_handler(int sig)
424{
9723f966
JH
425had_local_scan_timeout = sig;
426siglongjmp(local_scan_env, 1);
059ec3d9
PH
427}
428
429
430
431/*************************************************
432* local_scan() crashed *
433*************************************************/
434
435/* Handler function for signals that occur while running a local_scan()
436function.
437
438Argument: the signal number
439Returns: nothing
440*/
441
442static void
443local_scan_crash_handler(int sig)
444{
9723f966
JH
445had_local_scan_crash = sig;
446siglongjmp(local_scan_env, 1);
059ec3d9
PH
447}
448
9723f966
JH
449#endif /*HAVE_LOCAL_SCAN*/
450
059ec3d9
PH
451
452/*************************************************
453* SIGTERM or SIGINT received *
454*************************************************/
455
456/* Handler for SIGTERM or SIGINT signals that occur while reading the
457data that comprises a message.
458
459Argument: the signal number
460Returns: nothing
461*/
462
463static void
464data_sigterm_sigint_handler(int sig)
465{
9723f966 466had_data_sigint = sig;
059ec3d9
PH
467}
468
469
470
471/*************************************************
472* Add new recipient to list *
473*************************************************/
474
475/* This function builds a list of recipient addresses in argc/argv
476format.
477
478Arguments:
479 recipient the next address to add to recipients_list
480 pno parent number for fixed aliases; -1 otherwise
481
482Returns: nothing
483*/
484
485void
486receive_add_recipient(uschar *recipient, int pno)
487{
488if (recipients_count >= recipients_list_max)
489 {
490 recipient_item *oldlist = recipients_list;
491 int oldmax = recipients_list_max;
ea97267c 492 recipients_list_max = recipients_list_max ? 2*recipients_list_max : 50;
f3ebb786 493 recipients_list = store_get(recipients_list_max * sizeof(recipient_item), FALSE);
059ec3d9
PH
494 if (oldlist != NULL)
495 memcpy(recipients_list, oldlist, oldmax * sizeof(recipient_item));
496 }
497
498recipients_list[recipients_count].address = recipient;
499recipients_list[recipients_count].pno = pno;
8523533c
TK
500#ifdef EXPERIMENTAL_BRIGHTMAIL
501recipients_list[recipients_count].bmi_optin = bmi_current_optin;
502/* reset optin string pointer for next recipient */
503bmi_current_optin = NULL;
504#endif
6c1c3d1d
WB
505recipients_list[recipients_count].orcpt = NULL;
506recipients_list[recipients_count].dsn_flags = 0;
059ec3d9
PH
507recipients_list[recipients_count++].errors_to = NULL;
508}
509
510
511
512
513/*************************************************
fd98a5c6
JH
514* Send user response message *
515*************************************************/
61147df4 516
fd98a5c6
JH
517/* This function is passed a default response code and a user message. It calls
518smtp_message_code() to check and possibly modify the response code, and then
519calls smtp_respond() to transmit the response. I put this into a function
520just to avoid a lot of repetition.
61147df4
PP
521
522Arguments:
fd98a5c6
JH
523 code the response code
524 user_msg the user message
525
526Returns: nothing
61147df4
PP
527*/
528
8ccd00b1 529#ifndef DISABLE_PRDR
61147df4 530static void
fd98a5c6 531smtp_user_msg(uschar *code, uschar *user_msg)
61147df4 532{
fd98a5c6 533int len = 3;
4f6ae5c3 534smtp_message_code(&code, &len, &user_msg, NULL, TRUE);
fd98a5c6 535smtp_respond(code, len, TRUE, user_msg);
61147df4
PP
536}
537#endif
538
539
540
541
fd98a5c6
JH
542
543/*************************************************
059ec3d9
PH
544* Remove a recipient from the list *
545*************************************************/
546
547/* This function is provided for local_scan() to use.
548
549Argument:
550 recipient address to remove
551
552Returns: TRUE if it did remove something; FALSE otherwise
553*/
554
555BOOL
556receive_remove_recipient(uschar *recipient)
557{
059ec3d9
PH
558DEBUG(D_receive) debug_printf("receive_remove_recipient(\"%s\") called\n",
559 recipient);
d7978c0f 560for (int count = 0; count < recipients_count; count++)
059ec3d9
PH
561 if (Ustrcmp(recipients_list[count].address, recipient) == 0)
562 {
563 if ((--recipients_count - count) > 0)
564 memmove(recipients_list + count, recipients_list + count + 1,
54cdb463 565 (recipients_count - count)*sizeof(recipient_item));
059ec3d9
PH
566 return TRUE;
567 }
059ec3d9
PH
568return FALSE;
569}
570
571
572
573
574
3c55eef2
JH
575/* Pause for a while waiting for input. If none received in that time,
576close the logfile, if we had one open; then if we wait for a long-running
577datasource (months, in one use-case) log rotation will not leave us holding
578the file copy. */
579
580static void
581log_close_chk(void)
582{
583if (!receive_timeout)
584 {
585 struct timeval t;
586 timesince(&t, &received_time);
587 if (t.tv_sec > 30*60)
588 mainlog_close();
589 else
590 {
591 fd_set r;
592 FD_ZERO(&r); FD_SET(0, &r);
593 t.tv_sec = 30*60 - t.tv_sec; t.tv_usec = 0;
594 if (select(1, &r, NULL, NULL, &t) == 0) mainlog_close();
595 }
596 }
597}
598
059ec3d9
PH
599/*************************************************
600* Read data portion of a non-SMTP message *
601*************************************************/
602
603/* This function is called to read the remainder of a message (following the
604header) when the input is not from SMTP - we are receiving a local message on
605a standard input stream. The message is always terminated by EOF, and is also
606terminated by a dot on a line by itself if the flag dot_ends is TRUE. Split the
607two cases for maximum efficiency.
608
609Ensure that the body ends with a newline. This will naturally be the case when
610the termination is "\n.\n" but may not be otherwise. The RFC defines messages
611as "sequences of lines" - this of course strictly applies only to SMTP, but
612deliveries into BSD-type mailbox files also require it. Exim used to have a
613flag for doing this at delivery time, but as it was always set for all
614transports, I decided to simplify things by putting the check here instead.
615
616There is at least one MUA (dtmail) that sends CRLF via this interface, and
617other programs are known to do this as well. Exim used to have a option for
618dealing with this: in July 2003, after much discussion, the code has been
619changed to default to treat any of LF, CRLF, and bare CR as line terminators.
620
621However, for the case when a dot on a line by itself terminates a message, the
622only recognized terminating sequences before and after the dot are LF and CRLF.
623Otherwise, having read EOL . CR, you don't know whether to read another
624character or not.
625
626Internally, in messages stored in Exim's spool files, LF is used as the line
627terminator. Under the new regime, bare CRs will no longer appear in these
628files.
629
630Arguments:
631 fout a FILE to which to write the message
632
633Returns: One of the END_xxx values indicating why it stopped reading
634*/
635
636static int
637read_message_data(FILE *fout)
638{
639int ch_state;
640register int ch;
d677b2f2 641register int linelength = 0;
059ec3d9
PH
642
643/* Handle the case when only EOF terminates the message */
644
8768d548 645if (!f.dot_ends)
059ec3d9 646 {
3c55eef2 647 int last_ch = '\n';
059ec3d9 648
305e4fae
HSHR
649/*XXX we do a gettimeofday before checking for every received char,
650which is hardly clever. The function-indirection doesn't help, but
651an additional function to check for nonempty read buffer would help.
652See stdin_getc() / smtp_getc() / tls_getc() / bdat_getc(). */
653
3c55eef2
JH
654 for ( ;
655 log_close_chk(), (ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF;
656 last_ch = ch)
059ec3d9
PH
657 {
658 if (ch == 0) body_zerocount++;
659 if (last_ch == '\r' && ch != '\n')
660 {
d677b2f2
PH
661 if (linelength > max_received_linelength)
662 max_received_linelength = linelength;
663 linelength = 0;
059ec3d9
PH
664 if (fputc('\n', fout) == EOF) return END_WERROR;
665 message_size++;
666 body_linecount++;
667 }
668 if (ch == '\r') continue;
669
670 if (fputc(ch, fout) == EOF) return END_WERROR;
d677b2f2
PH
671 if (ch == '\n')
672 {
673 if (linelength > max_received_linelength)
674 max_received_linelength = linelength;
675 linelength = 0;
676 body_linecount++;
677 }
678 else linelength++;
059ec3d9
PH
679 if (++message_size > thismessage_size_limit) return END_SIZE;
680 }
681
682 if (last_ch != '\n')
683 {
d677b2f2
PH
684 if (linelength > max_received_linelength)
685 max_received_linelength = linelength;
059ec3d9
PH
686 if (fputc('\n', fout) == EOF) return END_WERROR;
687 message_size++;
688 body_linecount++;
689 }
690
691 return END_EOF;
692 }
693
694/* Handle the case when a dot on a line on its own, or EOF, terminates. */
695
696ch_state = 1;
697
3c55eef2 698while (log_close_chk(), (ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF)
059ec3d9
PH
699 {
700 if (ch == 0) body_zerocount++;
701 switch (ch_state)
702 {
703 case 0: /* Normal state (previous char written) */
704 if (ch == '\n')
d677b2f2
PH
705 {
706 body_linecount++;
707 if (linelength > max_received_linelength)
708 max_received_linelength = linelength;
709 linelength = -1;
710 ch_state = 1;
711 }
059ec3d9
PH
712 else if (ch == '\r')
713 { ch_state = 2; continue; }
714 break;
715
716 case 1: /* After written "\n" */
717 if (ch == '.') { ch_state = 3; continue; }
6eb02f88 718 if (ch == '\r') { ch_state = 2; continue; }
3581f321
JH
719 if (ch == '\n') { body_linecount++; linelength = -1; }
720 else ch_state = 0;
059ec3d9
PH
721 break;
722
723 case 2:
724 body_linecount++; /* After unwritten "\r" */
d677b2f2
PH
725 if (linelength > max_received_linelength)
726 max_received_linelength = linelength;
059ec3d9 727 if (ch == '\n')
d677b2f2
PH
728 {
729 ch_state = 1;
730 linelength = -1;
731 }
059ec3d9
PH
732 else
733 {
734 if (message_size++, fputc('\n', fout) == EOF) return END_WERROR;
735 if (ch == '\r') continue;
736 ch_state = 0;
d677b2f2 737 linelength = 0;
059ec3d9
PH
738 }
739 break;
740
741 case 3: /* After "\n." (\n written, dot not) */
742 if (ch == '\n') return END_DOT;
743 if (ch == '\r') { ch_state = 4; continue; }
744 message_size++;
d677b2f2 745 linelength++;
059ec3d9
PH
746 if (fputc('.', fout) == EOF) return END_WERROR;
747 ch_state = 0;
748 break;
749
750 case 4: /* After "\n.\r" (\n written, rest not) */
751 if (ch == '\n') return END_DOT;
752 message_size += 2;
753 body_linecount++;
754 if (fputs(".\n", fout) == EOF) return END_WERROR;
755 if (ch == '\r') { ch_state = 2; continue; }
756 ch_state = 0;
757 break;
758 }
759
d677b2f2 760 linelength++;
059ec3d9
PH
761 if (fputc(ch, fout) == EOF) return END_WERROR;
762 if (++message_size > thismessage_size_limit) return END_SIZE;
763 }
764
765/* Get here if EOF read. Unless we have just written "\n", we need to ensure
766the message ends with a newline, and we must also write any characters that
767were saved up while testing for an ending dot. */
768
769if (ch_state != 1)
770 {
771 static uschar *ends[] = { US"\n", NULL, US"\n", US".\n", US".\n" };
772 if (fputs(CS ends[ch_state], fout) == EOF) return END_WERROR;
773 message_size += Ustrlen(ends[ch_state]);
774 body_linecount++;
775 }
776
777return END_EOF;
778}
779
780
781
782
783/*************************************************
784* Read data portion of an SMTP message *
785*************************************************/
786
787/* This function is called to read the remainder of an SMTP message (after the
788headers), or to skip over it when an error has occurred. In this case, the
789output file is passed as NULL.
790
791If any line begins with a dot, that character is skipped. The input should only
792be successfully terminated by CR LF . CR LF unless it is local (non-network)
793SMTP, in which case the CRs are optional, but...
794
795FUDGE: It seems that sites on the net send out messages with just LF
796terminators, despite the warnings in the RFCs, and other MTAs handle this. So
797we make the CRs optional in all cases.
798
799July 2003: Bare CRs cause trouble. We now treat them as line terminators as
800well, so that there are no CRs in spooled messages. However, the message
801terminating dot is not recognized between two bare CRs.
802
803Arguments:
804 fout a FILE to which to write the message; NULL if skipping
805
806Returns: One of the END_xxx values indicating why it stopped reading
807*/
808
809static int
810read_message_data_smtp(FILE *fout)
811{
812int ch_state = 0;
e4bdf652 813int ch;
7e3ce68e 814int linelength = 0;
059ec3d9 815
bd8fbe36 816while ((ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF)
059ec3d9
PH
817 {
818 if (ch == 0) body_zerocount++;
819 switch (ch_state)
820 {
821 case 0: /* After LF or CRLF */
822 if (ch == '.')
823 {
824 ch_state = 3;
825 continue; /* Don't ever write . after LF */
826 }
827 ch_state = 1;
828
829 /* Else fall through to handle as normal uschar. */
830
831 case 1: /* Normal state */
832 if (ch == '\n')
833 {
834 ch_state = 0;
835 body_linecount++;
1f5497b2
PH
836 if (linelength > max_received_linelength)
837 max_received_linelength = linelength;
838 linelength = -1;
059ec3d9
PH
839 }
840 else if (ch == '\r')
841 {
842 ch_state = 2;
843 continue;
844 }
845 break;
846
847 case 2: /* After (unwritten) CR */
848 body_linecount++;
1f5497b2
PH
849 if (linelength > max_received_linelength)
850 max_received_linelength = linelength;
851 linelength = -1;
059ec3d9
PH
852 if (ch == '\n')
853 {
854 ch_state = 0;
855 }
856 else
857 {
858 message_size++;
859 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
6851a9c5 860 cutthrough_data_put_nl();
059ec3d9
PH
861 if (ch != '\r') ch_state = 1; else continue;
862 }
863 break;
864
865 case 3: /* After [CR] LF . */
866 if (ch == '\n')
867 return END_DOT;
868 if (ch == '\r')
869 {
870 ch_state = 4;
871 continue;
872 }
1bc460a6
JH
873 /* The dot was removed at state 3. For a doubled dot, here, reinstate
874 it to cutthrough. The current ch, dot or not, is passed both to cutthrough
875 and to file below. */
876 if (ch == '.')
877 {
878 uschar c= ch;
6851a9c5 879 cutthrough_data_puts(&c, 1);
1bc460a6
JH
880 }
881 ch_state = 1;
059ec3d9
PH
882 break;
883
884 case 4: /* After [CR] LF . CR */
885 if (ch == '\n') return END_DOT;
886 message_size++;
887 body_linecount++;
888 if (fout != NULL && fputc('\n', fout) == EOF) return END_WERROR;
6851a9c5 889 cutthrough_data_put_nl();
059ec3d9
PH
890 if (ch == '\r')
891 {
892 ch_state = 2;
893 continue;
894 }
895 ch_state = 1;
896 break;
897 }
898
899 /* Add the character to the spool file, unless skipping; then loop for the
900 next. */
901
902 message_size++;
1f5497b2 903 linelength++;
7e3ce68e 904 if (fout)
059ec3d9
PH
905 {
906 if (fputc(ch, fout) == EOF) return END_WERROR;
907 if (message_size > thismessage_size_limit) return END_SIZE;
908 }
e4bdf652 909 if(ch == '\n')
6851a9c5 910 cutthrough_data_put_nl();
e4bdf652
JH
911 else
912 {
7e3ce68e 913 uschar c = ch;
6851a9c5 914 cutthrough_data_puts(&c, 1);
e4bdf652 915 }
059ec3d9
PH
916 }
917
918/* Fall through here if EOF encountered. This indicates some kind of error,
919since a correct message is terminated by [CR] LF . [CR] LF. */
920
921return END_EOF;
922}
923
924
925
926
7e3ce68e 927/* Variant of the above read_message_data_smtp() specialised for RFC 3030
1ebe15c3
JH
928CHUNKING. Accept input lines separated by either CRLF or CR or LF and write
929LF-delimited spoolfile. Until we have wireformat spoolfiles, we need the
930body_linecount accounting for proper re-expansion for the wire, so use
931a cut-down version of the state-machine above; we don't need to do leading-dot
932detection and unstuffing.
7e3ce68e
JH
933
934Arguments:
544dd905
PP
935 fout a FILE to which to write the message; NULL if skipping;
936 must be open for both writing and reading.
7e3ce68e
JH
937
938Returns: One of the END_xxx values indicating why it stopped reading
939*/
940
941static int
942read_message_bdat_smtp(FILE *fout)
943{
cff70eb1
HSHR
944int linelength = 0, ch;
945enum CH_STATE ch_state = LF_SEEN;
d953610f 946BOOL fix_nl = FALSE;
7e3ce68e 947
1ebe15c3 948for(;;)
7e3ce68e 949 {
7d758a6a 950 switch ((ch = bdat_getc(GETC_BUFFER_UNLIMITED)))
1ebe15c3
JH
951 {
952 case EOF: return END_EOF;
1ebe15c3 953 case ERR: return END_PROTOCOL;
d953610f
HSHR
954 case EOD:
955 /* Nothing to get from the sender anymore. We check the last
956 character written to the spool.
957
958 RFC 3030 states, that BDAT chunks are normal text, terminated by CRLF.
959 If we would be strict, we would refuse such broken messages.
960 But we are liberal, so we fix it. It would be easy just to append
961 the "\n" to the spool.
962
963 But there are some more things (line counting, message size calculation and such),
964 that would need to be duplicated here. So we simply do some ungetc
965 trickery.
966 */
59d98039
JH
967 if (fout)
968 {
969 if (fseek(fout, -1, SEEK_CUR) < 0) return END_PROTOCOL;
970 if (fgetc(fout) == '\n') return END_DOT;
971 }
d953610f
HSHR
972
973 if (linelength == -1) /* \r already seen (see below) */
974 {
975 DEBUG(D_receive) debug_printf("Add missing LF\n");
976 bdat_ungetc('\n');
977 continue;
978 }
979 DEBUG(D_receive) debug_printf("Add missing CRLF\n");
980 bdat_ungetc('\r'); /* not even \r was seen */
981 fix_nl = TRUE;
982
983 continue;
1ebe15c3
JH
984 case '\0': body_zerocount++; break;
985 }
986 switch (ch_state)
987 {
cff70eb1
HSHR
988 case LF_SEEN: /* After LF or CRLF */
989 ch_state = MID_LINE;
1ebe15c3 990 /* fall through to handle as normal uschar. */
7e3ce68e 991
cff70eb1 992 case MID_LINE: /* Mid-line state */
1ebe15c3
JH
993 if (ch == '\n')
994 {
cff70eb1 995 ch_state = LF_SEEN;
1ebe15c3
JH
996 body_linecount++;
997 if (linelength > max_received_linelength)
998 max_received_linelength = linelength;
999 linelength = -1;
1000 }
1001 else if (ch == '\r')
1002 {
cff70eb1 1003 ch_state = CR_SEEN;
d953610f 1004 if (fix_nl) bdat_ungetc('\n');
1ebe15c3
JH
1005 continue; /* don't write CR */
1006 }
1007 break;
7e3ce68e 1008
cff70eb1 1009 case CR_SEEN: /* After (unwritten) CR */
1ebe15c3
JH
1010 body_linecount++;
1011 if (linelength > max_received_linelength)
1012 max_received_linelength = linelength;
1013 linelength = -1;
1014 if (ch == '\n')
cff70eb1 1015 ch_state = LF_SEEN;
1ebe15c3
JH
1016 else
1017 {
1018 message_size++;
59d98039 1019 if (fout && fputc('\n', fout) == EOF) return END_WERROR;
6851a9c5 1020 cutthrough_data_put_nl();
1ebe15c3 1021 if (ch == '\r') continue; /* don't write CR */
cff70eb1 1022 ch_state = MID_LINE;
1ebe15c3
JH
1023 }
1024 break;
1025 }
1026
1027 /* Add the character to the spool file, unless skipping */
1028
1029 message_size++;
1030 linelength++;
1031 if (fout)
1032 {
1033 if (fputc(ch, fout) == EOF) return END_WERROR;
1034 if (message_size > thismessage_size_limit) return END_SIZE;
1035 }
1036 if(ch == '\n')
6851a9c5 1037 cutthrough_data_put_nl();
1ebe15c3
JH
1038 else
1039 {
1040 uschar c = ch;
6851a9c5 1041 cutthrough_data_puts(&c, 1);
1ebe15c3 1042 }
7e3ce68e
JH
1043 }
1044/*NOTREACHED*/
1045}
1046
328c5688
JH
1047static int
1048read_message_bdat_smtp_wire(FILE *fout)
1049{
1050int ch;
1051
1052/* Remember that this message uses wireformat. */
1053
d21bf202
JH
1054DEBUG(D_receive) debug_printf("CHUNKING: %s\n",
1055 fout ? "writing spoolfile in wire format" : "flushing input");
8768d548 1056f.spool_file_wireformat = TRUE;
328c5688 1057
0d81dabc 1058for (;;)
328c5688 1059 {
0d81dabc
JH
1060 if (chunking_data_left > 0)
1061 {
1062 unsigned len = MAX(chunking_data_left, thismessage_size_limit - message_size + 1);
1063 uschar * buf = bdat_getbuf(&len);
328c5688 1064
8b77d27a 1065 if (!buf) return END_EOF;
0d81dabc
JH
1066 message_size += len;
1067 if (fout && fwrite(buf, len, 1, fout) != 1) return END_WERROR;
1068 }
1069 else switch (ch = bdat_getc(GETC_BUFFER_UNLIMITED))
1070 {
1071 case EOF: return END_EOF;
1072 case EOD: return END_DOT;
1073 case ERR: return END_PROTOCOL;
1074
1075 default:
1076 message_size++;
1077 /*XXX not done:
1078 linelength
1079 max_received_linelength
1080 body_linecount
1081 body_zerocount
1082 */
1083 if (fout && fputc(ch, fout) == EOF) return END_WERROR;
1084 break;
1085 }
1086 if (message_size > thismessage_size_limit) return END_SIZE;
328c5688
JH
1087 }
1088/*NOTREACHED*/
1089}
1090
7e3ce68e
JH
1091
1092
1093
059ec3d9
PH
1094/*************************************************
1095* Swallow SMTP message *
1096*************************************************/
1097
1098/* This function is called when there has been some kind of error while reading
1099an SMTP message, and the remaining data may need to be swallowed. It is global
1100because it is called from smtp_closedown() to shut down an incoming call
1101tidily.
1102
1103Argument: a FILE from which to read the message
1104Returns: nothing
1105*/
1106
1107void
1108receive_swallow_smtp(void)
1109{
1110if (message_ended >= END_NOTENDED)
d21bf202
JH
1111 message_ended = chunking_state <= CHUNKING_OFFERED
1112 ? read_message_data_smtp(NULL)
1113 : read_message_bdat_smtp_wire(NULL);
059ec3d9
PH
1114}
1115
1116
1117
1118/*************************************************
1119* Handle lost SMTP connection *
1120*************************************************/
1121
1122/* This function logs connection loss incidents and generates an appropriate
1123SMTP response.
1124
1125Argument: additional data for the message
1126Returns: the SMTP response
1127*/
1128
1129static uschar *
1130handle_lost_connection(uschar *s)
1131{
1132log_write(L_lost_incoming_connection | L_smtp_connection, LOG_MAIN,
1133 "%s lost while reading message data%s", smtp_get_connection_info(), s);
eea0defe 1134smtp_notquit_exit(US"connection-lost", NULL, NULL);
059ec3d9
PH
1135return US"421 Lost incoming connection";
1136}
1137
1138
1139
1140
1141/*************************************************
1142* Handle a non-smtp reception error *
1143*************************************************/
1144
1145/* This function is called for various errors during the reception of non-SMTP
1146messages. It either sends a message to the sender of the problem message, or it
1147writes to the standard error stream.
1148
1149Arguments:
1150 errcode code for moan_to_sender(), identifying the error
1151 text1 first message text, passed to moan_to_sender()
1152 text2 second message text, used only for stderrr
1153 error_rc code to pass to exim_exit if no problem
1154 f FILE containing body of message (may be stdin)
1155 hptr pointer to instore headers or NULL
1156
1157Returns: calls exim_exit(), which does not return
1158*/
1159
1160static void
1161give_local_error(int errcode, uschar *text1, uschar *text2, int error_rc,
1162 FILE *f, header_line *hptr)
1163{
1164if (error_handling == ERRORS_SENDER)
1165 {
1166 error_block eblock;
1167 eblock.next = NULL;
1168 eblock.text1 = text1;
37f3dc43 1169 eblock.text2 = US"";
059ec3d9
PH
1170 if (!moan_to_sender(errcode, &eblock, hptr, f, FALSE))
1171 error_rc = EXIT_FAILURE;
1172 }
37f3dc43
JH
1173else
1174 fprintf(stderr, "exim: %s%s\n", text2, text1); /* Sic */
f1e894f3 1175(void)fclose(f);
81022793 1176exim_exit(error_rc);
059ec3d9
PH
1177}
1178
1179
1180
1181/*************************************************
1182* Add header lines set up by ACL *
1183*************************************************/
1184
850635b6
PH
1185/* This function is called to add the header lines that were set up by
1186statements in an ACL to the list of headers in memory. It is done in two stages
1187like this, because when the ACL for RCPT is running, the other headers have not
1188yet been received. This function is called twice; once just before running the
1189DATA ACL, and once after. This is so that header lines added by MAIL or RCPT
1190are visible to the DATA ACL.
059ec3d9
PH
1191
1192Originally these header lines were added at the end. Now there is support for
1193three different places: top, bottom, and after the Received: header(s). There
1194will always be at least one Received: header, even if it is marked deleted, and
1195even if something else has been put in front of it.
1196
1197Arguments:
1198 acl_name text to identify which ACL
1199
1200Returns: nothing
1201*/
1202
1203static void
578d43dc 1204add_acl_headers(int where, uschar *acl_name)
059ec3d9 1205{
059ec3d9 1206header_line *last_received = NULL;
e7568d51 1207
578d43dc
JH
1208switch(where)
1209 {
1210 case ACL_WHERE_DKIM:
1211 case ACL_WHERE_MIME:
af4a1bca 1212 case ACL_WHERE_DATA:
74f1a423 1213 if ( cutthrough.cctx.sock >= 0 && cutthrough.delivery
57cc2785 1214 && (acl_removed_headers || acl_added_headers))
578d43dc
JH
1215 {
1216 log_write(0, LOG_MAIN|LOG_PANIC, "Header modification in data ACLs"
af4a1bca 1217 " will not take effect on cutthrough deliveries");
578d43dc
JH
1218 return;
1219 }
1220 }
1221
57cc2785 1222if (acl_removed_headers)
e7568d51 1223 {
e1d04f48 1224 DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers removed by %s ACL:\n", acl_name);
e7568d51 1225
d7978c0f 1226 for (header_line * h = header_list; h; h = h->next) if (h->type != htype_old)
e7568d51 1227 {
55414b25 1228 const uschar * list = acl_removed_headers;
e7568d51
TL
1229 int sep = ':'; /* This is specified as a colon-separated list */
1230 uschar *s;
1231 uschar buffer[128];
4a142059
JH
1232
1233 while ((s = string_nextinlist(&list, &sep, buffer, sizeof(buffer))))
1234 if (header_testname(h, s, Ustrlen(s), FALSE))
e7568d51
TL
1235 {
1236 h->type = htype_old;
e1d04f48 1237 DEBUG(D_receive|D_acl) debug_printf_indent(" %s", h->text);
e7568d51 1238 }
e7568d51
TL
1239 }
1240 acl_removed_headers = NULL;
e1d04f48 1241 DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
e7568d51 1242 }
059ec3d9 1243
57cc2785 1244if (!acl_added_headers) return;
e1d04f48 1245DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers added by %s ACL:\n", acl_name);
059ec3d9 1246
d7978c0f 1247for (header_line * h = acl_added_headers, * next; h; h = next)
059ec3d9
PH
1248 {
1249 next = h->next;
1250
1251 switch(h->type)
1252 {
1253 case htype_add_top:
c674e7a4
JH
1254 h->next = header_list;
1255 header_list = h;
1256 DEBUG(D_receive|D_acl) debug_printf_indent(" (at top)");
1257 break;
059ec3d9
PH
1258
1259 case htype_add_rec:
c674e7a4
JH
1260 if (!last_received)
1261 {
1262 last_received = header_list;
1263 while (!header_testname(last_received, US"Received", 8, FALSE))
1264 last_received = last_received->next;
1265 while (last_received->next &&
1266 header_testname(last_received->next, US"Received", 8, FALSE))
1267 last_received = last_received->next;
1268 }
1269 h->next = last_received->next;
1270 last_received->next = h;
1271 DEBUG(D_receive|D_acl) debug_printf_indent(" (after Received:)");
1272 break;
059ec3d9 1273
8523533c 1274 case htype_add_rfc:
c674e7a4
JH
1275 /* add header before any header which is NOT Received: or Resent- */
1276 last_received = header_list;
1277 while ( last_received->next &&
1278 ( (header_testname(last_received->next, US"Received", 8, FALSE)) ||
1279 (header_testname_incomplete(last_received->next, US"Resent-", 7, FALSE)) ) )
1280 last_received = last_received->next;
1281 /* last_received now points to the last Received: or Resent-* header
1282 in an uninterrupted chain of those header types (seen from the beginning
1283 of all headers. Our current header must follow it. */
1284 h->next = last_received->next;
1285 last_received->next = h;
1286 DEBUG(D_receive|D_acl) debug_printf_indent(" (before any non-Received: or Resent-*: header)");
1287 break;
8523533c 1288
059ec3d9 1289 default:
c674e7a4
JH
1290 h->next = NULL;
1291 header_last->next = h;
1292 DEBUG(D_receive|D_acl) debug_printf_indent(" ");
1293 break;
059ec3d9
PH
1294 }
1295
c674e7a4 1296 if (!h->next) header_last = h;
059ec3d9
PH
1297
1298 /* Check for one of the known header types (From:, To:, etc.) though in
1299 practice most added headers are going to be "other". Lower case
1300 identification letters are never stored with the header; they are used
1301 for existence tests when messages are received. So discard any lower case
1302 flag values. */
1303
1304 h->type = header_checkname(h, FALSE);
1305 if (h->type >= 'a') h->type = htype_other;
1306
c674e7a4 1307 DEBUG(D_receive|D_acl) debug_printf("%s", h->text);
059ec3d9
PH
1308 }
1309
71fafd95 1310acl_added_headers = NULL;
e1d04f48 1311DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");
059ec3d9
PH
1312}
1313
1314
1315
1316/*************************************************
1317* Add host information for log line *
1318*************************************************/
1319
1320/* Called for acceptance and rejecting log lines. This adds information about
1321the calling host to a string that is being built dynamically.
1322
1323Arguments:
1324 s the dynamic string
059ec3d9
PH
1325
1326Returns: the extended string
1327*/
1328
acec9514
JH
1329static gstring *
1330add_host_info_for_log(gstring * g)
059ec3d9 1331{
fc16abb4 1332if (sender_fullhost)
059ec3d9 1333 {
fc16abb4 1334 if (LOGGING(dnssec) && sender_host_dnssec) /*XXX sender_helo_dnssec? */
acec9514
JH
1335 g = string_catn(g, US" DS", 3);
1336 g = string_append(g, 2, US" H=", sender_fullhost);
52f12a7c
JH
1337 if (LOGGING(incoming_interface) && interface_address)
1338 g = string_fmt_append(g, " I=[%s]:%d", interface_address, interface_port);
059ec3d9 1339 }
8768d548 1340if (f.tcp_in_fastopen && !f.tcp_in_fastopen_logged)
a2673768 1341 {
ee8b8090 1342 g = string_catn(g, US" TFO*", f.tcp_in_fastopen_data ? 5 : 4);
8768d548 1343 f.tcp_in_fastopen_logged = TRUE;
a2673768
JH
1344 }
1345if (sender_ident)
acec9514 1346 g = string_append(g, 2, US" U=", sender_ident);
a2673768 1347if (received_protocol)
acec9514 1348 g = string_append(g, 2, US" P=", received_protocol);
8768d548 1349if (LOGGING(pipelining) && f.smtp_in_pipelining_advertised)
ee8b8090
JH
1350 {
1351 g = string_catn(g, US" L", 2);
81344b40 1352#ifndef DISABLE_PIPE_CONNECT
ee8b8090
JH
1353 if (f.smtp_in_early_pipe_used)
1354 g = string_catn(g, US"*", 1);
1355 else if (f.smtp_in_early_pipe_advertised)
1356 g = string_catn(g, US".", 1);
1357#endif
1358 if (!f.smtp_in_pipelining_used)
1359 g = string_catn(g, US"-", 1);
1360 }
acec9514 1361return g;
059ec3d9
PH
1362}
1363
1364
1365
63955bf2 1366#ifdef WITH_CONTENT_SCAN
059ec3d9
PH
1367
1368/*************************************************
54cdb463
PH
1369* Run the MIME ACL on a message *
1370*************************************************/
1371
1372/* This code is in a subroutine so that it can be used for both SMTP
1373and non-SMTP messages. It is called with a non-NULL ACL pointer.
1374
1375Arguments:
1376 acl The ACL to run (acl_smtp_mime or acl_not_smtp_mime)
1377 smtp_yield_ptr Set FALSE to kill messages after dropped connection
1378 smtp_reply_ptr Where SMTP reply is being built
1379 blackholed_by_ptr Where "blackholed by" message is being built
1380
1381Returns: TRUE to carry on; FALSE to abandon the message
1382*/
1383
1384static BOOL
1385run_mime_acl(uschar *acl, BOOL *smtp_yield_ptr, uschar **smtp_reply_ptr,
1386 uschar **blackholed_by_ptr)
1387{
1388FILE *mbox_file;
5b6f7658 1389uschar * rfc822_file_path = NULL;
54cdb463 1390unsigned long mbox_size;
54cdb463
PH
1391uschar *user_msg, *log_msg;
1392int mime_part_count_buffer = -1;
040721f2 1393uschar * mbox_filename;
7156b1ef 1394int rc = OK;
54cdb463 1395
54cdb463 1396/* check if it is a MIME message */
040721f2 1397
d7978c0f
JH
1398for (header_line * my_headerlist = header_list; my_headerlist;
1399 my_headerlist = my_headerlist->next)
040721f2
JH
1400 if ( my_headerlist->type != '*' /* skip deleted headers */
1401 && strncmpic(my_headerlist->text, US"Content-Type:", 13) == 0
1402 )
4e88a19f 1403 {
54cdb463
PH
1404 DEBUG(D_receive) debug_printf("Found Content-Type: header - executing acl_smtp_mime.\n");
1405 goto DO_MIME_ACL;
4e88a19f 1406 }
54cdb463
PH
1407
1408DEBUG(D_receive) debug_printf("No Content-Type: header - presumably not a MIME message.\n");
1409return TRUE;
1410
1411DO_MIME_ACL:
040721f2 1412
54cdb463 1413/* make sure the eml mbox file is spooled up */
040721f2
JH
1414if (!(mbox_file = spool_mbox(&mbox_size, NULL, &mbox_filename)))
1415 { /* error while spooling */
54cdb463
PH
1416 log_write(0, LOG_MAIN|LOG_PANIC,
1417 "acl_smtp_mime: error while creating mbox spool file, message temporarily rejected.");
1418 Uunlink(spool_name);
1419 unspool_mbox();
6f0c431a
PP
1420#ifdef EXPERIMENTAL_DCC
1421 dcc_ok = 0;
1422#endif
a5bd321b 1423 smtp_respond(US"451", 3, TRUE, US"temporary local problem");
54cdb463
PH
1424 message_id[0] = 0; /* Indicate no message accepted */
1425 *smtp_reply_ptr = US""; /* Indicate reply already sent */
1426 return FALSE; /* Indicate skip to end of receive function */
040721f2 1427 }
54cdb463
PH
1428
1429mime_is_rfc822 = 0;
1430
1431MIME_ACL_CHECK:
1432mime_part_count = -1;
1433rc = mime_acl_check(acl, mbox_file, NULL, &user_msg, &log_msg);
f1e894f3 1434(void)fclose(mbox_file);
54cdb463 1435
5b6f7658 1436if (rfc822_file_path)
4e88a19f 1437 {
54cdb463
PH
1438 mime_part_count = mime_part_count_buffer;
1439
4e88a19f
PH
1440 if (unlink(CS rfc822_file_path) == -1)
1441 {
54cdb463
PH
1442 log_write(0, LOG_PANIC,
1443 "acl_smtp_mime: can't unlink RFC822 spool file, skipping.");
5b6f7658 1444 goto END_MIME_ACL;
4e88a19f 1445 }
5b6f7658 1446 rfc822_file_path = NULL;
4e88a19f 1447 }
54cdb463
PH
1448
1449/* check if we must check any message/rfc822 attachments */
4e88a19f
PH
1450if (rc == OK)
1451 {
5b6f7658
JH
1452 uschar * scandir = string_copyn(mbox_filename,
1453 Ustrrchr(mbox_filename, '/') - mbox_filename);
e8bc7fca
JH
1454 struct dirent * entry;
1455 DIR * tempdir;
54cdb463 1456
54a2a2a9 1457 for (tempdir = exim_opendir(scandir); entry = readdir(tempdir); )
e8bc7fca 1458 if (strncmpic(US entry->d_name, US"__rfc822_", 9) == 0)
4e88a19f 1459 {
5b6f7658
JH
1460 rfc822_file_path = string_sprintf("%s/%s", scandir, entry->d_name);
1461 DEBUG(D_receive)
1462 debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
1463 rfc822_file_path);
4e88a19f
PH
1464 break;
1465 }
4e88a19f 1466 closedir(tempdir);
54cdb463 1467
5b6f7658 1468 if (rfc822_file_path)
4e88a19f 1469 {
e8bc7fca 1470 if ((mbox_file = Ufopen(rfc822_file_path, "rb")))
4e88a19f 1471 {
e8bc7fca
JH
1472 /* set RFC822 expansion variable */
1473 mime_is_rfc822 = 1;
1474 mime_part_count_buffer = mime_part_count;
1475 goto MIME_ACL_CHECK;
4e88a19f 1476 }
e8bc7fca
JH
1477 log_write(0, LOG_PANIC,
1478 "acl_smtp_mime: can't open RFC822 spool file, skipping.");
1479 unlink(CS rfc822_file_path);
4e88a19f
PH
1480 }
1481 }
54cdb463
PH
1482
1483END_MIME_ACL:
578d43dc 1484add_acl_headers(ACL_WHERE_MIME, US"MIME");
54cdb463
PH
1485if (rc == DISCARD)
1486 {
1487 recipients_count = 0;
1488 *blackholed_by_ptr = US"MIME ACL";
1bd642c2 1489 cancel_cutthrough_connection(TRUE, US"mime acl discard");
54cdb463
PH
1490 }
1491else if (rc != OK)
1492 {
1493 Uunlink(spool_name);
1bd642c2 1494 cancel_cutthrough_connection(TRUE, US"mime acl not ok");
54cdb463 1495 unspool_mbox();
6f0c431a
PP
1496#ifdef EXPERIMENTAL_DCC
1497 dcc_ok = 0;
1498#endif
5b6f7658 1499 if (smtp_input)
4f6ae5c3 1500 {
5b6f7658
JH
1501 if (smtp_handle_acl_fail(ACL_WHERE_MIME, rc, user_msg, log_msg) != 0)
1502 *smtp_yield_ptr = FALSE; /* No more messages after dropped connection */
f4c1088b 1503 *smtp_reply_ptr = US""; /* Indicate reply already sent */
4f6ae5c3 1504 }
54cdb463
PH
1505 message_id[0] = 0; /* Indicate no message accepted */
1506 return FALSE; /* Cause skip to end of receive function */
4e88a19f 1507 }
54cdb463
PH
1508
1509return TRUE;
1510}
1511
63955bf2 1512#endif /* WITH_CONTENT_SCAN */
54cdb463
PH
1513
1514
e4bdf652
JH
1515
1516void
1517received_header_gen(void)
1518{
1519uschar *received;
1520uschar *timestamp;
1521header_line *received_header= header_list;
1522
1523timestamp = expand_string(US"${tod_full}");
1524if (recipients_count == 1) received_for = recipients_list[0].address;
1525received = expand_string(received_header_text);
1526received_for = NULL;
1527
d4ff61d1 1528if (!received)
e4bdf652
JH
1529 {
1530 if(spool_name[0] != 0)
1531 Uunlink(spool_name); /* Lose the data file */
1532 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Expansion of \"%s\" "
1533 "(received_header_text) failed: %s", string_printing(received_header_text),
1534 expand_string_message);
1535 }
1536
1537/* The first element on the header chain is reserved for the Received header,
1538so all we have to do is fill in the text pointer, and set the type. However, if
1539the result of the expansion is an empty string, we leave the header marked as
1540"old" so as to refrain from adding a Received header. */
1541
1542if (received[0] == 0)
1543 {
1544 received_header->text = string_sprintf("Received: ; %s\n", timestamp);
1545 received_header->type = htype_old;
1546 }
1547else
1548 {
1549 received_header->text = string_sprintf("%s; %s\n", received, timestamp);
1550 received_header->type = htype_received;
1551 }
1552
1553received_header->slen = Ustrlen(received_header->text);
1554
1555DEBUG(D_receive) debug_printf(">>Generated Received: header line\n%c %s",
1556 received_header->type, received_header->text);
1557}
1558
1559
1560
54cdb463 1561/*************************************************
059ec3d9
PH
1562* Receive message *
1563*************************************************/
1564
1565/* Receive a message on the given input, and put it into a pair of spool files.
1566Either a non-null list of recipients, or the extract flag will be true, or
1567both. The flag sender_local is true for locally generated messages. The flag
1568submission_mode is true if an ACL has obeyed "control = submission". The flag
8800895a 1569suppress_local_fixups is true if an ACL has obeyed "control =
f4ee74ac
PP
1570suppress_local_fixups" or -G was passed on the command-line.
1571The flag smtp_input is true if the message is to be
8800895a
PH
1572handled using SMTP conventions about termination and lines starting with dots.
1573For non-SMTP messages, dot_ends is true for dot-terminated messages.
059ec3d9
PH
1574
1575If a message was successfully read, message_id[0] will be non-zero.
1576
1577The general actions of this function are:
1578
1579 . Read the headers of the message (if any) into a chain of store
1580 blocks.
1581
1582 . If there is a "sender:" header and the message is locally originated,
69358f02
PH
1583 throw it away, unless the caller is trusted, or unless
1584 active_local_sender_retain is set - which can only happen if
1585 active_local_from_check is false.
059ec3d9
PH
1586
1587 . If recipients are to be extracted from the message, build the
1588 recipients list from the headers, removing any that were on the
1589 original recipients list (unless extract_addresses_remove_arguments is
1590 false), and at the same time, remove any bcc header that may be present.
1591
1592 . Get the spool file for the data, sort out its unique name, open
1593 and lock it (but don't give it the name yet).
1594
1595 . Generate a "Message-Id" header if the message doesn't have one, for
1596 locally-originated messages.
1597
1598 . Generate a "Received" header.
1599
1600 . Ensure the recipients list is fully qualified and rewritten if necessary.
1601
1602 . If there are any rewriting rules, apply them to the sender address
1603 and also to the headers.
1604
1605 . If there is no from: header, generate one, for locally-generated messages
1606 and messages in "submission mode" only.
1607
1608 . If the sender is local, check that from: is correct, and if not, generate
1609 a Sender: header, unless message comes from a trusted caller, or this
69358f02 1610 feature is disabled by active_local_from_check being false.
059ec3d9
PH
1611
1612 . If there is no "date" header, generate one, for locally-originated
1613 or submission mode messages only.
1614
1615 . Copy the rest of the input, or up to a terminating "." if in SMTP or
1616 dot_ends mode, to the data file. Leave it open, to hold the lock.
1617
1618 . Write the envelope and the headers to a new file.
1619
1620 . Set the name for the header file; close it.
1621
1622 . Set the name for the data file; close it.
1623
1624Because this function can potentially be called many times in a single
1625SMTP connection, all store should be got by store_get(), so that it will be
1626automatically retrieved after the message is accepted.
1627
1628FUDGE: It seems that sites on the net send out messages with just LF
1629terminators, despite the warnings in the RFCs, and other MTAs handle this. So
1630we make the CRs optional in all cases.
1631
1632July 2003: Bare CRs in messages, especially in header lines, cause trouble. A
1633new regime is now in place in which bare CRs in header lines are turned into LF
1634followed by a space, so as not to terminate the header line.
1635
1636February 2004: A bare LF in a header line in a message whose first line was
1637terminated by CRLF is treated in the same way as a bare CR.
1638
1639Arguments:
1640 extract_recip TRUE if recipients are to be extracted from the message's
1641 headers
1642
1643Returns: TRUE there are more messages to be read (SMTP input)
1644 FALSE there are no more messages to be read (non-SMTP input
1645 or SMTP connection collapsed, or other failure)
1646
1647When reading a message for filter testing, the returned value indicates
1648whether the headers (which is all that is read) were terminated by '.' or
1649not. */
1650
1651BOOL
1652receive_msg(BOOL extract_recip)
1653{
7156b1ef 1654int rc = FAIL;
059ec3d9
PH
1655int msg_size = 0;
1656int process_info_len = Ustrlen(process_info);
d342446f
JH
1657int error_rc = error_handling == ERRORS_SENDER
1658 ? errors_sender_rc : EXIT_FAILURE;
059ec3d9 1659int header_size = 256;
acec9514 1660int start, end, domain;
099afc4f 1661int id_resolution = 0;
059ec3d9 1662int had_zero = 0;
d677b2f2 1663int prevlines_length = 0;
059ec3d9 1664
cfbb0d24 1665int ptr = 0;
059ec3d9
PH
1666
1667BOOL contains_resent_headers = FALSE;
1668BOOL extracted_ignored = FALSE;
1669BOOL first_line_ended_crlf = TRUE_UNSET;
1670BOOL smtp_yield = TRUE;
1671BOOL yield = FALSE;
1672
1673BOOL resents_exist = FALSE;
1674uschar *resent_prefix = US"";
1675uschar *blackholed_by = NULL;
04f7d5b9 1676uschar *blackhole_log_msg = US"";
c5430c20 1677enum {NOT_TRIED, TMP_REJ, PERM_REJ, ACCEPTED} cutthrough_done = NOT_TRIED;
059ec3d9
PH
1678
1679flock_t lock_data;
1680error_block *bad_addresses = NULL;
1681
1682uschar *frozen_by = NULL;
1683uschar *queued_by = NULL;
1684
acec9514 1685uschar *errmsg;
f3ebb786 1686rmark rcvd_log_reset_point;
acec9514 1687gstring * g;
059ec3d9
PH
1688struct stat statbuf;
1689
4e88a19f 1690/* Final message to give to SMTP caller, and messages from ACLs */
059ec3d9
PH
1691
1692uschar *smtp_reply = NULL;
4e88a19f 1693uschar *user_msg, *log_msg;
059ec3d9
PH
1694
1695/* Working header pointers */
1696
f3ebb786 1697rmark reset_point;
d7978c0f 1698header_line *next;
059ec3d9 1699
2cbb4081 1700/* Flags for noting the existence of certain headers (only one left) */
059ec3d9
PH
1701
1702BOOL date_header_exists = FALSE;
1703
1704/* Pointers to receive the addresses of headers whose contents we need. */
1705
1706header_line *from_header = NULL;
1707header_line *subject_header = NULL;
1708header_line *msgid_header = NULL;
1709header_line *received_header;
049782c0 1710BOOL msgid_header_newly_created = FALSE;
059ec3d9
PH
1711
1712/* Variables for use when building the Received: header. */
1713
059ec3d9
PH
1714uschar *timestamp;
1715int tslen;
1716
9723f966 1717
059ec3d9
PH
1718/* Release any open files that might have been cached while preparing to
1719accept the message - e.g. by verifying addresses - because reading a message
1720might take a fair bit of real time. */
1721
1722search_tidyup();
1723
e4bdf652
JH
1724/* Extracting the recipient list from an input file is incompatible with
1725cutthrough delivery with the no-spool option. It shouldn't be possible
817d9f57 1726to set up the combination, but just in case kill any ongoing connection. */
e4bdf652 1727if (extract_recip || !smtp_input)
57cc2785 1728 cancel_cutthrough_connection(TRUE, US"not smtp input");
e4bdf652 1729
059ec3d9
PH
1730/* Initialize the chain of headers by setting up a place-holder for Received:
1731header. Temporarily mark it as "old", i.e. not to be used. We keep header_last
1732pointing to the end of the chain to make adding headers simple. */
1733
f3ebb786 1734received_header = header_list = header_last = store_get(sizeof(header_line), FALSE);
059ec3d9
PH
1735header_list->next = NULL;
1736header_list->type = htype_old;
1737header_list->text = NULL;
1738header_list->slen = 0;
1739
1740/* Control block for the next header to be read. */
1741
f3ebb786
JH
1742reset_point = store_mark();
1743next = store_get(sizeof(header_line), FALSE); /* not tainted */
1744next->text = store_get(header_size, TRUE); /* tainted */
059ec3d9
PH
1745
1746/* Initialize message id to be null (indicating no message read), and the
1747header names list to be the normal list. Indicate there is no data file open
1748yet, initialize the size and warning count, and deal with no size limit. */
1749
1750message_id[0] = 0;
1bd642c2 1751spool_data_file = NULL;
059ec3d9 1752data_fd = -1;
41313d92 1753spool_name = US"";
059ec3d9
PH
1754message_size = 0;
1755warning_count = 0;
d677b2f2 1756received_count = 1; /* For the one we will add */
059ec3d9
PH
1757
1758if (thismessage_size_limit <= 0) thismessage_size_limit = INT_MAX;
1759
2e0c1448 1760/* While reading the message, the following counts are computed. */
059ec3d9 1761
d677b2f2
PH
1762message_linecount = body_linecount = body_zerocount =
1763 max_received_linelength = 0;
b273058b
JH
1764
1765#ifdef WITH_CONTENT_SCAN
1766/* reset non-per-part mime variables */
1767mime_is_coverletter = 0;
1768mime_is_rfc822 = 0;
1769mime_part_count = -1;
1770#endif
059ec3d9 1771
80a47a2c 1772#ifndef DISABLE_DKIM
e983e85a
JH
1773/* Call into DKIM to set up the context. In CHUNKING mode
1774we clear the dot-stuffing flag */
8768d548 1775if (smtp_input && !smtp_batched_input && !f.dkim_disable_verify)
e983e85a 1776 dkim_exim_verify_init(chunking_state <= CHUNKING_OFFERED);
fb2274d4
TK
1777#endif
1778
1a2e76e1 1779#ifdef SUPPORT_DMARC
39fdec3c 1780if (sender_host_address) dmarc_init(); /* initialize libopendmarc */
4840604e
TL
1781#endif
1782
059ec3d9
PH
1783/* Remember the time of reception. Exim uses time+pid for uniqueness of message
1784ids, and fractions of a second are required. See the comments that precede the
1785message id creation below. */
1786
1787(void)gettimeofday(&message_id_tv, NULL);
1788
1789/* For other uses of the received time we can operate with granularity of one
1790second, and for that we use the global variable received_time. This is for
306c6c77 1791things like ultimate message timeouts. */
059ec3d9 1792
32dfdf8b 1793received_time = message_id_tv;
059ec3d9
PH
1794
1795/* If SMTP input, set the special handler for timeouts. The alarm() calls
1796happen in the smtp_getc() function when it refills its buffer. */
1797
9723f966
JH
1798had_data_timeout = 0;
1799if (smtp_input)
1800 os_non_restarting_signal(SIGALRM, data_timeout_handler);
059ec3d9
PH
1801
1802/* If not SMTP input, timeout happens only if configured, and we just set a
1803single timeout for the whole message. */
1804
1805else if (receive_timeout > 0)
1806 {
1807 os_non_restarting_signal(SIGALRM, data_timeout_handler);
c2a1bba0 1808 ALARM(receive_timeout);
059ec3d9
PH
1809 }
1810
1811/* SIGTERM and SIGINT are caught always. */
1812
9723f966 1813had_data_sigint = 0;
059ec3d9
PH
1814signal(SIGTERM, data_sigterm_sigint_handler);
1815signal(SIGINT, data_sigterm_sigint_handler);
1816
1817/* Header lines in messages are not supposed to be very long, though when
1818unfolded, to: and cc: headers can take up a lot of store. We must also cope
1819with the possibility of junk being thrown at us. Start by getting 256 bytes for
1820storing the header, and extend this as necessary using string_cat().
1821
1822To cope with total lunacies, impose an upper limit on the length of the header
1823section of the message, as otherwise the store will fill up. We must also cope
1824with the possibility of binary zeros in the data. Hence we cannot use fgets().
1825Folded header lines are joined into one string, leaving the '\n' characters
1826inside them, so that writing them out reproduces the input.
1827
1828Loop for each character of each header; the next structure for chaining the
1829header is set up already, with ptr the offset of the next character in
1830next->text. */
1831
1832for (;;)
1833 {
bd8fbe36 1834 int ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1835
1836 /* If we hit EOF on a SMTP connection, it's an error, since incoming
1837 SMTP must have a correct "." terminator. */
1838
1839 if (ch == EOF && smtp_input /* && !smtp_batched_input */)
1840 {
1841 smtp_reply = handle_lost_connection(US" (header)");
1842 smtp_yield = FALSE;
1843 goto TIDYUP; /* Skip to end of function */
1844 }
1845
1846 /* See if we are at the current header's size limit - there must be at least
1847 four bytes left. This allows for the new character plus a zero, plus two for
1848 extra insertions when we are playing games with dots and carriage returns. If
1849 we are at the limit, extend the text buffer. This could have been done
1850 automatically using string_cat() but because this is a tightish loop storing
1851 only one character at a time, we choose to do it inline. Normally
1852 store_extend() will be able to extend the block; only at the end of a big
1853 store block will a copy be needed. To handle the case of very long headers
1854 (and sometimes lunatic messages can have ones that are 100s of K long) we
1855 call store_release() for strings that have been copied - if the string is at
1856 the start of a block (and therefore the only thing in it, because we aren't
4e6ae623
JH
1857 doing any other gets), the block gets freed. We can only do this release if
1858 there were no allocations since the once that we want to free. */
059ec3d9
PH
1859
1860 if (ptr >= header_size - 4)
1861 {
1862 int oldsize = header_size;
56ac062a
JH
1863
1864 if (header_size >= INT_MAX/2)
1865 goto OVERSIZE;
059ec3d9 1866 header_size *= 2;
56ac062a 1867
f3ebb786
JH
1868 /* The data came from the message, so is tainted. */
1869
1870 if (!store_extend(next->text, TRUE, oldsize, header_size))
1871 next->text = store_newblock(next->text, TRUE, header_size, ptr);
059ec3d9
PH
1872 }
1873
1874 /* Cope with receiving a binary zero. There is dispute about whether
1875 these should be allowed in RFC 822 messages. The middle view is that they
1876 should not be allowed in headers, at least. Exim takes this attitude at
1877 the moment. We can't just stomp on them here, because we don't know that
1878 this line is a header yet. Set a flag to cause scanning later. */
1879
1880 if (ch == 0) had_zero++;
1881
1882 /* Test for termination. Lines in remote SMTP are terminated by CRLF, while
1883 those from data files use just LF. Treat LF in local SMTP input as a
1884 terminator too. Treat EOF as a line terminator always. */
1885
1886 if (ch == EOF) goto EOL;
1887
1888 /* FUDGE: There are sites out there that don't send CRs before their LFs, and
1889 other MTAs accept this. We are therefore forced into this "liberalisation"
1890 too, so we accept LF as a line terminator whatever the source of the message.
1891 However, if the first line of the message ended with a CRLF, we treat a bare
1892 LF specially by inserting a white space after it to ensure that the header
1893 line is not terminated. */
1894
1895 if (ch == '\n')
1896 {
1897 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = FALSE;
80a47a2c 1898 else if (first_line_ended_crlf) receive_ungetc(' ');
059ec3d9
PH
1899 goto EOL;
1900 }
1901
1902 /* This is not the end of the line. If this is SMTP input and this is
1903 the first character in the line and it is a "." character, ignore it.
1904 This implements the dot-doubling rule, though header lines starting with
1905 dots aren't exactly common. They are legal in RFC 822, though. If the
1906 following is CRLF or LF, this is the line that that terminates the
1907 entire message. We set message_ended to indicate this has happened (to
1908 prevent further reading), and break out of the loop, having freed the
1909 empty header, and set next = NULL to indicate no data line. */
1910
8768d548 1911 if (ptr == 0 && ch == '.' && f.dot_ends)
059ec3d9 1912 {
bd8fbe36 1913 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1914 if (ch == '\r')
1915 {
bd8fbe36 1916 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1917 if (ch != '\n')
1918 {
80a47a2c 1919 receive_ungetc(ch);
059ec3d9
PH
1920 ch = '\r'; /* Revert to CR */
1921 }
1922 }
1923 if (ch == '\n')
1924 {
1925 message_ended = END_DOT;
f3ebb786 1926 reset_point = store_reset(reset_point);
059ec3d9
PH
1927 next = NULL;
1928 break; /* End character-reading loop */
1929 }
1930
1931 /* For non-SMTP input, the dot at the start of the line was really a data
1932 character. What is now in ch is the following character. We guaranteed
1933 enough space for this above. */
1934
1935 if (!smtp_input)
1936 {
1937 next->text[ptr++] = '.';
1938 message_size++;
1939 }
1940 }
1941
1942 /* If CR is immediately followed by LF, end the line, ignoring the CR, and
1943 remember this case if this is the first line ending. */
1944
1945 if (ch == '\r')
1946 {
bd8fbe36 1947 ch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
1948 if (ch == '\n')
1949 {
1950 if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
1951 goto EOL;
1952 }
1953
1954 /* Otherwise, put back the character after CR, and turn the bare CR
1955 into LF SP. */
1956
80a47a2c 1957 ch = (receive_ungetc)(ch);
059ec3d9
PH
1958 next->text[ptr++] = '\n';
1959 message_size++;
1960 ch = ' ';
1961 }
1962
1963 /* We have a data character for the header line. */
1964
1965 next->text[ptr++] = ch; /* Add to buffer */
1966 message_size++; /* Total message size so far */
1967
1968 /* Handle failure due to a humungously long header section. The >= allows
1969 for the terminating \n. Add what we have so far onto the headers list so
1970 that it gets reflected in any error message, and back up the just-read
1971 character. */
1972
1973 if (message_size >= header_maxsize)
1974 {
56ac062a 1975OVERSIZE:
059ec3d9
PH
1976 next->text[ptr] = 0;
1977 next->slen = ptr;
1978 next->type = htype_other;
1979 next->next = NULL;
1980 header_last->next = next;
1981 header_last = next;
1982
1983 log_write(0, LOG_MAIN, "ridiculously long message header received from "
1984 "%s (more than %d characters): message abandoned",
8768d548 1985 f.sender_host_unknown ? sender_ident : sender_fullhost, header_maxsize);
059ec3d9
PH
1986
1987 if (smtp_input)
1988 {
1989 smtp_reply = US"552 Message header is ridiculously long";
1990 receive_swallow_smtp();
1991 goto TIDYUP; /* Skip to end of function */
1992 }
1993
1994 else
1995 {
1996 give_local_error(ERRMESS_VLONGHEADER,
1997 string_sprintf("message header longer than %d characters received: "
1998 "message not accepted", header_maxsize), US"", error_rc, stdin,
1999 header_list->next);
2000 /* Does not return */
2001 }
2002 }
2003
2004 continue; /* With next input character */
2005
2006 /* End of header line reached */
2007
2008 EOL:
2e0c1448
PH
2009
2010 /* Keep track of lines for BSMTP errors and overall message_linecount. */
2011
2012 receive_linecount++;
2013 message_linecount++;
059ec3d9 2014
d677b2f2
PH
2015 /* Keep track of maximum line length */
2016
2017 if (ptr - prevlines_length > max_received_linelength)
2018 max_received_linelength = ptr - prevlines_length;
2019 prevlines_length = ptr + 1;
2020
059ec3d9
PH
2021 /* Now put in the terminating newline. There is always space for
2022 at least two more characters. */
2023
2024 next->text[ptr++] = '\n';
2025 message_size++;
2026
2027 /* A blank line signals the end of the headers; release the unwanted
2028 space and set next to NULL to indicate this. */
2029
2030 if (ptr == 1)
2031 {
f3ebb786 2032 reset_point = store_reset(reset_point);
059ec3d9
PH
2033 next = NULL;
2034 break;
2035 }
2036
2037 /* There is data in the line; see if the next input character is a
2038 whitespace character. If it is, we have a continuation of this header line.
2039 There is always space for at least one character at this point. */
2040
2041 if (ch != EOF)
2042 {
bd8fbe36 2043 int nextch = (receive_getc)(GETC_BUFFER_UNLIMITED);
059ec3d9
PH
2044 if (nextch == ' ' || nextch == '\t')
2045 {
2046 next->text[ptr++] = nextch;
56ac062a
JH
2047 if (++message_size >= header_maxsize)
2048 goto OVERSIZE;
059ec3d9
PH
2049 continue; /* Iterate the loop */
2050 }
80a47a2c 2051 else if (nextch != EOF) (receive_ungetc)(nextch); /* For next time */
059ec3d9
PH
2052 else ch = EOF; /* Cause main loop to exit at end */
2053 }
2054
2055 /* We have got to the real line end. Terminate the string and release store
2056 beyond it. If it turns out to be a real header, internal binary zeros will
2057 be squashed later. */
2058
2059 next->text[ptr] = 0;
2060 next->slen = ptr;
f3ebb786 2061 store_release_above(next->text + ptr + 1);
059ec3d9
PH
2062
2063 /* Check the running total size against the overall message size limit. We
2064 don't expect to fail here, but if the overall limit is set less than MESSAGE_
2065 MAXSIZE and a big header is sent, we want to catch it. Just stop reading
2066 headers - the code to read the body will then also hit the buffer. */
2067
2068 if (message_size > thismessage_size_limit) break;
2069
2070 /* A line that is not syntactically correct for a header also marks
2071 the end of the headers. In this case, we leave next containing the
2072 first data line. This might actually be several lines because of the
2073 continuation logic applied above, but that doesn't matter.
2074
2075 It turns out that smail, and presumably sendmail, accept leading lines
2076 of the form
2077
2078 From ph10 Fri Jan 5 12:35 GMT 1996
2079
2080 in messages. The "mail" command on Solaris 2 sends such lines. I cannot
2081 find any documentation of this, but for compatibility it had better be
2082 accepted. Exim restricts it to the case of non-smtp messages, and
2083 treats it as an alternative to the -f command line option. Thus it is
2084 ignored except for trusted users or filter testing. Otherwise it is taken
2085 as the sender address, unless -f was used (sendmail compatibility).
2086
2087 It further turns out that some UUCPs generate the From_line in a different
2088 format, e.g.
2089
2090 From ph10 Fri, 7 Jan 97 14:00:00 GMT
2091
2092 The regex for matching these things is now capable of recognizing both
2093 formats (including 2- and 4-digit years in the latter). In fact, the regex
2094 is now configurable, as is the expansion string to fish out the sender.
2095
2096 Even further on it has been discovered that some broken clients send
2097 these lines in SMTP messages. There is now an option to ignore them from
2098 specified hosts or networks. Sigh. */
2099
d21bf202
JH
2100 if ( header_last == header_list
2101 && ( !smtp_input
2102 || ( sender_host_address
2103 && verify_check_host(&ignore_fromline_hosts) == OK
2104 )
2105 || (!sender_host_address && ignore_fromline_local)
2106 )
2107 && regex_match_and_setup(regex_From, next->text, 0, -1)
2108 )
059ec3d9 2109 {
8768d548 2110 if (!f.sender_address_forced)
059ec3d9
PH
2111 {
2112 uschar *uucp_sender = expand_string(uucp_from_sender);
d21bf202 2113 if (!uucp_sender)
059ec3d9
PH
2114 log_write(0, LOG_MAIN|LOG_PANIC,
2115 "expansion of \"%s\" failed after matching "
2116 "\"From \" line: %s", uucp_from_sender, expand_string_message);
059ec3d9
PH
2117 else
2118 {
2119 int start, end, domain;
2120 uschar *errmess;
2121 uschar *newsender = parse_extract_address(uucp_sender, &errmess,
2122 &start, &end, &domain, TRUE);
d21bf202 2123 if (newsender)
059ec3d9
PH
2124 {
2125 if (domain == 0 && newsender[0] != 0)
2126 newsender = rewrite_address_qualify(newsender, FALSE);
2127
f05da2e8 2128 if (filter_test != FTEST_NONE || receive_check_set_sender(newsender))
059ec3d9
PH
2129 {
2130 sender_address = newsender;
2131
8768d548 2132 if (f.trusted_caller || filter_test != FTEST_NONE)
059ec3d9
PH
2133 {
2134 authenticated_sender = NULL;
2135 originator_name = US"";
8768d548 2136 f.sender_local = FALSE;
059ec3d9
PH
2137 }
2138
f05da2e8 2139 if (filter_test != FTEST_NONE)
059ec3d9
PH
2140 printf("Sender taken from \"From \" line\n");
2141 }
2142 }
2143 }
2144 }
2145 }
2146
2147 /* Not a leading "From " line. Check to see if it is a valid header line.
2148 Header names may contain any non-control characters except space and colon,
2149 amazingly. */
2150
2151 else
2152 {
2153 uschar *p = next->text;
2154
2155 /* If not a valid header line, break from the header reading loop, leaving
2156 next != NULL, indicating that it holds the first line of the body. */
2157
2158 if (isspace(*p)) break;
2159 while (mac_isgraph(*p) && *p != ':') p++;
2160 while (isspace(*p)) p++;
2161 if (*p != ':')
2162 {
2163 body_zerocount = had_zero;
2164 break;
2165 }
2166
2167 /* We have a valid header line. If there were any binary zeroes in
2168 the line, stomp on them here. */
2169
2170 if (had_zero > 0)
d7978c0f
JH
2171 for (uschar * p = next->text; p < next->text + ptr; p++) if (*p == 0)
2172 *p = '?';
059ec3d9
PH
2173
2174 /* It is perfectly legal to have an empty continuation line
2175 at the end of a header, but it is confusing to humans
2176 looking at such messages, since it looks like a blank line.
2177 Reduce confusion by removing redundant white space at the
2178 end. We know that there is at least one printing character
2179 (the ':' tested for above) so there is no danger of running
2180 off the end. */
2181
2182 p = next->text + ptr - 2;
2183 for (;;)
2184 {
2185 while (*p == ' ' || *p == '\t') p--;
2186 if (*p != '\n') break;
2187 ptr = (p--) - next->text + 1;
2188 message_size -= next->slen - ptr;
2189 next->text[ptr] = 0;
2190 next->slen = ptr;
2191 }
2192
2193 /* Add the header to the chain */
2194
2195 next->type = htype_other;
2196 next->next = NULL;
2197 header_last->next = next;
2198 header_last = next;
2199
2200 /* Check the limit for individual line lengths. This comes after adding to
2201 the chain so that the failing line is reflected if a bounce is generated
2202 (for a local message). */
2203
2204 if (header_line_maxsize > 0 && next->slen > header_line_maxsize)
2205 {
2206 log_write(0, LOG_MAIN, "overlong message header line received from "
2207 "%s (more than %d characters): message abandoned",
8768d548 2208 f.sender_host_unknown ? sender_ident : sender_fullhost,
059ec3d9
PH
2209 header_line_maxsize);
2210
2211 if (smtp_input)
2212 {
2213 smtp_reply = US"552 A message header line is too long";
2214 receive_swallow_smtp();
2215 goto TIDYUP; /* Skip to end of function */
2216 }
2217
2218 else
059ec3d9
PH
2219 give_local_error(ERRMESS_VLONGHDRLINE,
2220 string_sprintf("message header line longer than %d characters "
2221 "received: message not accepted", header_line_maxsize), US"",
2222 error_rc, stdin, header_list->next);
2223 /* Does not return */
059ec3d9
PH
2224 }
2225
2226 /* Note if any resent- fields exist. */
2227
2228 if (!resents_exist && strncmpic(next->text, US"resent-", 7) == 0)
2229 {
2230 resents_exist = TRUE;
2231 resent_prefix = US"Resent-";
2232 }
2233 }
2234
1ebe15c3
JH
2235 /* Reject CHUNKING messages that do not CRLF their first header line */
2236
2237 if (!first_line_ended_crlf && chunking_state > CHUNKING_OFFERED)
2238 {
2239 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
2240 "Non-CRLF-terminated header, under CHUNKING: message abandoned",
2241 sender_address,
2242 sender_fullhost ? " H=" : "", sender_fullhost ? sender_fullhost : US"",
2243 sender_ident ? " U=" : "", sender_ident ? sender_ident : US"");
925ac8e4 2244 smtp_printf("552 Message header not CRLF terminated\r\n", FALSE);
1ebe15c3
JH
2245 bdat_flush_data();
2246 smtp_reply = US"";
2247 goto TIDYUP; /* Skip to end of function */
2248 }
2249
059ec3d9
PH
2250 /* The line has been handled. If we have hit EOF, break out of the loop,
2251 indicating no pending data line. */
2252
2253 if (ch == EOF) { next = NULL; break; }
2254
2255 /* Set up for the next header */
2256
f3ebb786 2257 reset_point = store_mark();
059ec3d9 2258 header_size = 256;
f3ebb786
JH
2259 next = store_get(sizeof(header_line), FALSE);
2260 next->text = store_get(header_size, TRUE);
059ec3d9
PH
2261 ptr = 0;
2262 had_zero = 0;
d677b2f2 2263 prevlines_length = 0;
059ec3d9
PH
2264 } /* Continue, starting to read the next header */
2265
2266/* At this point, we have read all the headers into a data structure in main
2267store. The first header is still the dummy placeholder for the Received: header
2268we are going to generate a bit later on. If next != NULL, it contains the first
2269data line - which terminated the headers before reaching a blank line (not the
2270normal case). */
2271
2272DEBUG(D_receive)
2273 {
2274 debug_printf(">>Headers received:\n");
d7978c0f 2275 for (header_line * h = header_list->next; h; h = h->next)
059ec3d9
PH
2276 debug_printf("%s", h->text);
2277 debug_printf("\n");
2278 }
2279
2280/* End of file on any SMTP connection is an error. If an incoming SMTP call
2281is dropped immediately after valid headers, the next thing we will see is EOF.
2282We must test for this specially, as further down the reading of the data is
2283skipped if already at EOF. */
2284
2285if (smtp_input && (receive_feof)())
2286 {
2287 smtp_reply = handle_lost_connection(US" (after header)");
2288 smtp_yield = FALSE;
2289 goto TIDYUP; /* Skip to end of function */
2290 }
2291
2292/* If this is a filter test run and no headers were read, output a warning
2293in case there is a mistake in the test message. */
2294
f05da2e8 2295if (filter_test != FTEST_NONE && header_list->next == NULL)
059ec3d9
PH
2296 printf("Warning: no message headers read\n");
2297
2298
2299/* Scan the headers to identify them. Some are merely marked for later
2300processing; some are dealt with here. */
2301
d7978c0f 2302for (header_line * h = header_list->next; h; h = h->next)
059ec3d9
PH
2303 {
2304 BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
2305 if (is_resent) contains_resent_headers = TRUE;
2306
2307 switch (header_checkname(h, is_resent))
2308 {
059ec3d9 2309 case htype_bcc:
c674e7a4
JH
2310 h->type = htype_bcc; /* Both Bcc: and Resent-Bcc: */
2311 break;
059ec3d9 2312
059ec3d9 2313 case htype_cc:
c674e7a4
JH
2314 h->type = htype_cc; /* Both Cc: and Resent-Cc: */
2315 break;
059ec3d9 2316
c674e7a4 2317 /* Record whether a Date: or Resent-Date: header exists, as appropriate. */
059ec3d9
PH
2318
2319 case htype_date:
c674e7a4
JH
2320 if (!resents_exist || is_resent) date_header_exists = TRUE;
2321 break;
059ec3d9 2322
c674e7a4 2323 /* Same comments as about Return-Path: below. */
059ec3d9
PH
2324
2325 case htype_delivery_date:
c674e7a4
JH
2326 if (delivery_date_remove) h->type = htype_old;
2327 break;
059ec3d9 2328
c674e7a4 2329 /* Same comments as about Return-Path: below. */
059ec3d9
PH
2330
2331 case htype_envelope_to:
c674e7a4
JH
2332 if (envelope_to_remove) h->type = htype_old;
2333 break;
059ec3d9 2334
c674e7a4
JH
2335 /* Mark all "From:" headers so they get rewritten. Save the one that is to
2336 be used for Sender: checking. For Sendmail compatibility, if the "From:"
2337 header consists of just the login id of the user who called Exim, rewrite
2338 it with the gecos field first. Apply this rule to Resent-From: if there
2339 are resent- fields. */
059ec3d9
PH
2340
2341 case htype_from:
c674e7a4
JH
2342 h->type = htype_from;
2343 if (!resents_exist || is_resent)
2344 {
2345 from_header = h;
2346 if (!smtp_input)
2347 {
2348 int len;
2349 uschar *s = Ustrchr(h->text, ':') + 1;
2350 while (isspace(*s)) s++;
2351 len = h->slen - (s - h->text) - 1;
2352 if (Ustrlen(originator_login) == len &&
2353 strncmpic(s, originator_login, len) == 0)
2354 {
2355 uschar *name = is_resent? US"Resent-From" : US"From";
2356 header_add(htype_from, "%s: %s <%s@%s>\n", name, originator_name,
2357 originator_login, qualify_domain_sender);
2358 from_header = header_last;
2359 h->type = htype_old;
2360 DEBUG(D_receive|D_rewrite)
2361 debug_printf("rewrote \"%s:\" header using gecos\n", name);
2362 }
2363 }
2364 }
2365 break;
059ec3d9 2366
c674e7a4
JH
2367 /* Identify the Message-id: header for generating "in-reply-to" in the
2368 autoreply transport. For incoming logging, save any resent- value. In both
2369 cases, take just the first of any multiples. */
059ec3d9
PH
2370
2371 case htype_id:
c674e7a4
JH
2372 if (!msgid_header && (!resents_exist || is_resent))
2373 {
2374 msgid_header = h;
2375 h->type = htype_id;
2376 }
2377 break;
059ec3d9 2378
c674e7a4 2379 /* Flag all Received: headers */
059ec3d9
PH
2380
2381 case htype_received:
c674e7a4
JH
2382 h->type = htype_received;
2383 received_count++;
2384 break;
059ec3d9 2385
c674e7a4 2386 /* "Reply-to:" is just noted (there is no resent-reply-to field) */
059ec3d9
PH
2387
2388 case htype_reply_to:
c674e7a4
JH
2389 h->type = htype_reply_to;
2390 break;
059ec3d9 2391
c674e7a4
JH
2392 /* The Return-path: header is supposed to be added to messages when
2393 they leave the SMTP system. We shouldn't receive messages that already
2394 contain Return-path. However, since Exim generates Return-path: on
2395 local delivery, resent messages may well contain it. We therefore
2396 provide an option (which defaults on) to remove any Return-path: headers
2397 on input. Removal actually means flagging as "old", which prevents the
2398 header being transmitted with the message. */
059ec3d9
PH
2399
2400 case htype_return_path:
c674e7a4 2401 if (return_path_remove) h->type = htype_old;
059ec3d9 2402
c674e7a4
JH
2403 /* If we are testing a mail filter file, use the value of the
2404 Return-Path: header to set up the return_path variable, which is not
2405 otherwise set. However, remove any <> that surround the address
2406 because the variable doesn't have these. */
059ec3d9 2407
c674e7a4
JH
2408 if (filter_test != FTEST_NONE)
2409 {
2410 uschar *start = h->text + 12;
2411 uschar *end = start + Ustrlen(start);
2412 while (isspace(*start)) start++;
2413 while (end > start && isspace(end[-1])) end--;
2414 if (*start == '<' && end[-1] == '>')
2415 {
2416 start++;
2417 end--;
2418 }
2419 return_path = string_copyn(start, end - start);
2420 printf("Return-path taken from \"Return-path:\" header line\n");
2421 }
2422 break;
059ec3d9
PH
2423
2424 /* If there is a "Sender:" header and the message is locally originated,
8800895a
PH
2425 and from an untrusted caller and suppress_local_fixups is not set, or if we
2426 are in submission mode for a remote message, mark it "old" so that it will
2427 not be transmitted with the message, unless active_local_sender_retain is
2428 set. (This can only be true if active_local_from_check is false.) If there
2429 are any resent- headers in the message, apply this rule to Resent-Sender:
2430 instead of Sender:. Messages with multiple resent- header sets cannot be
2431 tidily handled. (For this reason, at least one MUA - Pine - turns old
2432 resent- headers into X-resent- headers when resending, leaving just one
2433 set.) */
059ec3d9
PH
2434
2435 case htype_sender:
8768d548
JH
2436 h->type = !f.active_local_sender_retain
2437 && ( f.sender_local && !f.trusted_caller && !f.suppress_local_fixups
2438 || f.submission_mode
c674e7a4
JH
2439 )
2440 && (!resents_exist || is_resent)
2441 ? htype_old : htype_sender;
2442 break;
059ec3d9 2443
c674e7a4 2444 /* Remember the Subject: header for logging. There is no Resent-Subject */
059ec3d9
PH
2445
2446 case htype_subject:
c674e7a4
JH
2447 subject_header = h;
2448 break;
059ec3d9 2449
c674e7a4
JH
2450 /* "To:" gets flagged, and the existence of a recipient header is noted,
2451 whether it's resent- or not. */
059ec3d9
PH
2452
2453 case htype_to:
c674e7a4
JH
2454 h->type = htype_to;
2455 /****
2456 to_or_cc_header_exists = TRUE;
2457 ****/
2458 break;
059ec3d9
PH
2459 }
2460 }
2461
2462/* Extract recipients from the headers if that is required (the -t option).
2463Note that this is documented as being done *before* any address rewriting takes
2464place. There are two possibilities:
2465
2466(1) According to sendmail documentation for Solaris, IRIX, and HP-UX, any
2467recipients already listed are to be REMOVED from the message. Smail 3 works
2468like this. We need to build a non-recipients tree for that list, because in
2469subsequent processing this data is held in a tree and that's what the
2470spool_write_header() function expects. Make sure that non-recipient addresses
2471are fully qualified and rewritten if necessary.
2472
2473(2) According to other sendmail documentation, -t ADDS extracted recipients to
2474those in the command line arguments (and it is rumoured some other MTAs do
2475this). Therefore, there is an option to make Exim behave this way.
2476
2477*** Notes on "Resent-" header lines ***
2478
2479The presence of resent-headers in the message makes -t horribly ambiguous.
2480Experiments with sendmail showed that it uses recipients for all resent-
2481headers, totally ignoring the concept of "sets of resent- headers" as described
2482in RFC 2822 section 3.6.6. Sendmail also amalgamates them into a single set
2483with all the addresses in one instance of each header.
2484
2485This seems to me not to be at all sensible. Before release 4.20, Exim 4 gave an
2486error for -t if there were resent- headers in the message. However, after a
2487discussion on the mailing list, I've learned that there are MUAs that use
2488resent- headers with -t, and also that the stuff about sets of resent- headers
2489and their ordering in RFC 2822 is generally ignored. An MUA that submits a
2490message with -t and resent- header lines makes sure that only *its* resent-
2491headers are present; previous ones are often renamed as X-resent- for example.
2492
2493Consequently, Exim has been changed so that, if any resent- header lines are
2494present, the recipients are taken from all of the appropriate resent- lines,
2495and not from the ordinary To:, Cc:, etc. */
2496
2497if (extract_recip)
2498 {
2499 int rcount = 0;
2500 error_block **bnext = &bad_addresses;
2501
2502 if (extract_addresses_remove_arguments)
2503 {
2504 while (recipients_count-- > 0)
2505 {
2506 uschar *s = rewrite_address(recipients_list[recipients_count].address,
2507 TRUE, TRUE, global_rewrite_rules, rewrite_existflags);
2508 tree_add_nonrecipient(s);
2509 }
2510 recipients_list = NULL;
2511 recipients_count = recipients_list_max = 0;
2512 }
2513
059ec3d9
PH
2514 /* Now scan the headers */
2515
d7978c0f 2516 for (header_line * h = header_list->next; h; h = h->next)
059ec3d9
PH
2517 {
2518 if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
2519 (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
2520 {
2521 uschar *s = Ustrchr(h->text, ':') + 1;
2522 while (isspace(*s)) s++;
2523
8768d548 2524 f.parse_allow_group = TRUE; /* Allow address group syntax */
1eccaa59 2525
059ec3d9
PH
2526 while (*s != 0)
2527 {
2528 uschar *ss = parse_find_address_end(s, FALSE);
d7978c0f 2529 uschar *recipient, *errmess, *pp;
059ec3d9
PH
2530 int start, end, domain;
2531
2532 /* Check on maximum */
2533
2534 if (recipients_max > 0 && ++rcount > recipients_max)
059ec3d9
PH
2535 give_local_error(ERRMESS_TOOMANYRECIP, US"too many recipients",
2536 US"message rejected: ", error_rc, stdin, NULL);
2537 /* Does not return */
059ec3d9
PH
2538
2539 /* Make a copy of the address, and remove any internal newlines. These
2540 may be present as a result of continuations of the header line. The
2541 white space that follows the newline must not be removed - it is part
2542 of the header. */
2543
f3ebb786 2544 pp = recipient = store_get(ss - s + 1, is_tainted(s));
d7978c0f 2545 for (uschar * p = s; p < ss; p++) if (*p != '\n') *pp++ = *p;
059ec3d9 2546 *pp = 0;
250b6871 2547
8c5d388a 2548#ifdef SUPPORT_I18N
250b6871
JH
2549 {
2550 BOOL b = allow_utf8_domains;
2551 allow_utf8_domains = TRUE;
2552#endif
059ec3d9
PH
2553 recipient = parse_extract_address(recipient, &errmess, &start, &end,
2554 &domain, FALSE);
2555
8c5d388a 2556#ifdef SUPPORT_I18N
250b6871
JH
2557 if (string_is_utf8(recipient))
2558 message_smtputf8 = TRUE;
2559 else
2560 allow_utf8_domains = b;
2561 }
2562#endif
2563
059ec3d9
PH
2564 /* Keep a list of all the bad addresses so we can send a single
2565 error message at the end. However, an empty address is not an error;
2566 just ignore it. This can come from an empty group list like
2567
2568 To: Recipients of list:;
2569
2570 If there are no recipients at all, an error will occur later. */
2571
5fcc791a 2572 if (!recipient && Ustrcmp(errmess, "empty address") != 0)
059ec3d9
PH
2573 {
2574 int len = Ustrlen(s);
f3ebb786 2575 error_block *b = store_get(sizeof(error_block), FALSE);
059ec3d9
PH
2576 while (len > 0 && isspace(s[len-1])) len--;
2577 b->next = NULL;
2578 b->text1 = string_printing(string_copyn(s, len));
2579 b->text2 = errmess;
2580 *bnext = b;
2581 bnext = &(b->next);
2582 }
2583
2584 /* If the recipient is already in the nonrecipients tree, it must
2585 have appeared on the command line with the option extract_addresses_
2586 remove_arguments set. Do not add it to the recipients, and keep a note
2587 that this has happened, in order to give a better error if there are
2588 no recipients left. */
2589
2590 else if (recipient != NULL)
2591 {
2592 if (tree_search(tree_nonrecipients, recipient) == NULL)
2593 receive_add_recipient(recipient, -1);
2594 else
2595 extracted_ignored = TRUE;
2596 }
2597
2598 /* Move on past this address */
2599
2600 s = ss + (*ss? 1:0);
2601 while (isspace(*s)) s++;
1eccaa59
PH
2602 } /* Next address */
2603
8768d548
JH
2604 f.parse_allow_group = FALSE; /* Reset group syntax flags */
2605 f.parse_found_group = FALSE;
059ec3d9
PH
2606
2607 /* If this was the bcc: header, mark it "old", which means it
2608 will be kept on the spool, but not transmitted as part of the
2609 message. */
2610
2cbb4081 2611 if (h->type == htype_bcc) h->type = htype_old;
059ec3d9
PH
2612 } /* For appropriate header line */
2613 } /* For each header line */
2614
059ec3d9
PH
2615 }
2616
2617/* Now build the unique message id. This has changed several times over the
2618lifetime of Exim. This description was rewritten for Exim 4.14 (February 2003).
2619Retaining all the history in the comment has become too unwieldy - read
2620previous release sources if you want it.
2621
2622The message ID has 3 parts: tttttt-pppppp-ss. Each part is a number in base 62.
2623The first part is the current time, in seconds. The second part is the current
2624pid. Both are large enough to hold 32-bit numbers in base 62. The third part
2625can hold a number in the range 0-3843. It used to be a computed sequence
2626number, but is now the fractional component of the current time in units of
26271/2000 of a second (i.e. a value in the range 0-1999). After a message has been
2628received, Exim ensures that the timer has ticked at the appropriate level
2629before proceeding, to avoid duplication if the pid happened to be re-used
2630within the same time period. It seems likely that most messages will take at
2631least half a millisecond to be received, so no delay will normally be
2632necessary. At least for some time...
2633
2634There is a modification when localhost_number is set. Formerly this was allowed
2635to be as large as 255. Now it is restricted to the range 0-16, and the final
2636component of the message id becomes (localhost_number * 200) + fractional time
2637in units of 1/200 of a second (i.e. a value in the range 0-3399).
2638
2639Some not-really-Unix operating systems use case-insensitive file names (Darwin,
2640Cygwin). For these, we have to use base 36 instead of base 62. Luckily, this
2641still allows the tttttt field to hold a large enough number to last for some
2642more decades, and the final two-digit field can hold numbers up to 1295, which
2643is enough for milliseconds (instead of 1/2000 of a second).
2644
2645However, the pppppp field cannot hold a 32-bit pid, but it can hold a 31-bit
2646pid, so it is probably safe because pids have to be positive. The
2647localhost_number is restricted to 0-10 for these hosts, and when it is set, the
2648final field becomes (localhost_number * 100) + fractional time in centiseconds.
2649
2650Note that string_base62() returns its data in a static storage block, so it
2651must be copied before calling string_base62() again. It always returns exactly
26526 characters.
2653
2654There doesn't seem to be anything in the RFC which requires a message id to
2655start with a letter, but Smail was changed to ensure this. The external form of
2656the message id (as supplied by string expansion) therefore starts with an
2657additional leading 'E'. The spool file names do not include this leading
2658letter and it is not used internally.
2659
2660NOTE: If ever the format of message ids is changed, the regular expression for
2661checking that a string is in this format must be updated in a corresponding
2662way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
2540f2f8
JH
2663must also be changed to reflect the correct string length. The queue-sort code
2664needs to know the layout. Then, of course, other programs that rely on the
2665message id format will need updating too. */
059ec3d9
PH
2666
2667Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
2668message_id[6] = '-';
2669Ustrncpy(message_id + 7, string_base62((long int)getpid()), 6);
2670
2671/* Deal with the case where the host number is set. The value of the number was
2672checked when it was read, to ensure it isn't too big. The timing granularity is
2673left in id_resolution so that an appropriate wait can be done after receiving
2674the message, if necessary (we hope it won't be). */
2675
306c6c77 2676if (host_number_string)
059ec3d9 2677 {
099afc4f 2678 id_resolution = BASE_62 == 62 ? 5000 : 10000;
059ec3d9
PH
2679 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2680 string_base62((long int)(
2681 host_number * (1000000/id_resolution) +
2682 message_id_tv.tv_usec/id_resolution)) + 4);
2683 }
2684
2685/* Host number not set: final field is just the fractional time at an
2686appropriate resolution. */
2687
2688else
2689 {
099afc4f 2690 id_resolution = BASE_62 == 62 ? 500 : 1000;
059ec3d9
PH
2691 sprintf(CS(message_id + MESSAGE_ID_LENGTH - 3), "-%2s",
2692 string_base62((long int)(message_id_tv.tv_usec/id_resolution)) + 4);
2693 }
2694
2695/* Add the current message id onto the current process info string if
2696it will fit. */
2697
2698(void)string_format(process_info + process_info_len,
2699 PROCESS_INFO_SIZE - process_info_len, " id=%s", message_id);
2700
2701/* If we are using multiple input directories, set up the one for this message
2702to be the least significant base-62 digit of the time of arrival. Otherwise
2703ensure that it is an empty string. */
2704
59a93276 2705set_subdir_str(message_subdir, message_id, 0);
059ec3d9
PH
2706
2707/* Now that we have the message-id, if there is no message-id: header, generate
8800895a
PH
2708one, but only for local (without suppress_local_fixups) or submission mode
2709messages. This can be user-configured if required, but we had better flatten
2710any illegal characters therein. */
059ec3d9 2711
306c6c77 2712if ( !msgid_header
8768d548 2713 && ((!sender_host_address && !f.suppress_local_fixups) || f.submission_mode))
059ec3d9 2714 {
059ec3d9
PH
2715 uschar *id_text = US"";
2716 uschar *id_domain = primary_hostname;
049782c0 2717 header_line * h;
059ec3d9
PH
2718
2719 /* Permit only letters, digits, dots, and hyphens in the domain */
2720
306c6c77 2721 if (message_id_domain)
059ec3d9
PH
2722 {
2723 uschar *new_id_domain = expand_string(message_id_domain);
306c6c77 2724 if (!new_id_domain)
059ec3d9 2725 {
8768d548 2726 if (!f.expand_string_forcedfail)
059ec3d9
PH
2727 log_write(0, LOG_MAIN|LOG_PANIC,
2728 "expansion of \"%s\" (message_id_header_domain) "
2729 "failed: %s", message_id_domain, expand_string_message);
2730 }
306c6c77 2731 else if (*new_id_domain)
059ec3d9
PH
2732 {
2733 id_domain = new_id_domain;
d7978c0f 2734 for (uschar * p = id_domain; *p; p++)
059ec3d9
PH
2735 if (!isalnum(*p) && *p != '.') *p = '-'; /* No need to test '-' ! */
2736 }
2737 }
2738
2739 /* Permit all characters except controls and RFC 2822 specials in the
2740 additional text part. */
2741
306c6c77 2742 if (message_id_text)
059ec3d9
PH
2743 {
2744 uschar *new_id_text = expand_string(message_id_text);
306c6c77 2745 if (!new_id_text)
059ec3d9 2746 {
8768d548 2747 if (!f.expand_string_forcedfail)
059ec3d9
PH
2748 log_write(0, LOG_MAIN|LOG_PANIC,
2749 "expansion of \"%s\" (message_id_header_text) "
2750 "failed: %s", message_id_text, expand_string_message);
2751 }
306c6c77 2752 else if (*new_id_text)
059ec3d9
PH
2753 {
2754 id_text = new_id_text;
d7978c0f 2755 for (uschar * p = id_text; *p; p++) if (mac_iscntrl_or_special(*p)) *p = '-';
059ec3d9
PH
2756 }
2757 }
2758
049782c0
JH
2759 /* Add the header line.
2760 Resent-* headers are prepended, per RFC 5322 3.6.6. Non-Resent-* are
2761 appended, to preserve classical expectations of header ordering. */
059ec3d9 2762
049782c0 2763 h = header_add_at_position_internal(!resents_exist, NULL, FALSE, htype_id,
5eb690a1 2764 "%sMessage-Id: <%s%s%s@%s>\n", resent_prefix, message_id_external,
049782c0
JH
2765 *id_text == 0 ? "" : ".", id_text, id_domain);
2766
2767 /* Arrange for newly-created Message-Id to be logged */
2768
2769 if (!resents_exist)
2770 {
2771 msgid_header_newly_created = TRUE;
2772 msgid_header = h;
2773 }
059ec3d9
PH
2774 }
2775
2776/* If we are to log recipients, keep a copy of the raw ones before any possible
2777rewriting. Must copy the count, because later ACLs and the local_scan()
2778function may mess with the real recipients. */
2779
6c6d6e48 2780if (LOGGING(received_recipients))
059ec3d9 2781 {
f3ebb786 2782 raw_recipients = store_get(recipients_count * sizeof(uschar *), FALSE);
d7978c0f 2783 for (int i = 0; i < recipients_count; i++)
059ec3d9
PH
2784 raw_recipients[i] = string_copy(recipients_list[i].address);
2785 raw_recipients_count = recipients_count;
2786 }
2787
2788/* Ensure the recipients list is fully qualified and rewritten. Unqualified
2789recipients will get here only if the conditions were right (allow_unqualified_
2790recipient is TRUE). */
2791
d7978c0f 2792for (int i = 0; i < recipients_count; i++)
059ec3d9
PH
2793 recipients_list[i].address =
2794 rewrite_address(recipients_list[i].address, TRUE, TRUE,
2795 global_rewrite_rules, rewrite_existflags);
2796
8800895a
PH
2797/* If there is no From: header, generate one for local (without
2798suppress_local_fixups) or submission_mode messages. If there is no sender
2799address, but the sender is local or this is a local delivery error, use the
2800originator login. This shouldn't happen for genuine bounces, but might happen
2801for autoreplies. The addition of From: must be done *before* checking for the
2802possible addition of a Sender: header, because untrusted_set_sender allows an
2803untrusted user to set anything in the envelope (which might then get info
2804From:) but we still want to ensure a valid Sender: if it is required. */
2805
306c6c77 2806if ( !from_header
8768d548 2807 && ((!sender_host_address && !f.suppress_local_fixups) || f.submission_mode))
059ec3d9 2808 {
2fe1a124
PH
2809 uschar *oname = US"";
2810
2811 /* Use the originator_name if this is a locally submitted message and the
2812 caller is not trusted. For trusted callers, use it only if -F was used to
2813 force its value or if we have a non-SMTP message for which -f was not used
2814 to set the sender. */
2815
306c6c77 2816 if (!sender_host_address)
2fe1a124 2817 {
8768d548
JH
2818 if (!f.trusted_caller || f.sender_name_forced ||
2819 (!smtp_input && !f.sender_address_forced))
2fe1a124
PH
2820 oname = originator_name;
2821 }
2822
2823 /* For non-locally submitted messages, the only time we use the originator
2824 name is when it was forced by the /name= option on control=submission. */
2825
306c6c77 2826 else if (submission_name) oname = submission_name;
2fe1a124 2827
059ec3d9
PH
2828 /* Envelope sender is empty */
2829
306c6c77 2830 if (!*sender_address)
059ec3d9 2831 {
87ba3f5f
PH
2832 uschar *fromstart, *fromend;
2833
306c6c77
JH
2834 fromstart = string_sprintf("%sFrom: %s%s",
2835 resent_prefix, oname, *oname ? " <" : "");
2836 fromend = *oname ? US">" : US"";
87ba3f5f 2837
8768d548 2838 if (f.sender_local || f.local_error_message)
87ba3f5f
PH
2839 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2840 local_part_quote(originator_login), qualify_domain_sender,
2841 fromend);
306c6c77 2842
8768d548 2843 else if (f.submission_mode && authenticated_id)
059ec3d9 2844 {
306c6c77 2845 if (!submission_domain)
87ba3f5f
PH
2846 header_add(htype_from, "%s%s@%s%s\n", fromstart,
2847 local_part_quote(authenticated_id), qualify_domain_sender,
2848 fromend);
306c6c77
JH
2849
2850 else if (!*submission_domain) /* empty => whole address set */
87ba3f5f
PH
2851 header_add(htype_from, "%s%s%s\n", fromstart, authenticated_id,
2852 fromend);
306c6c77 2853
059ec3d9 2854 else
87ba3f5f 2855 header_add(htype_from, "%s%s@%s%s\n", fromstart,
306c6c77
JH
2856 local_part_quote(authenticated_id), submission_domain, fromend);
2857
059ec3d9
PH
2858 from_header = header_last; /* To get it checked for Sender: */
2859 }
2860 }
2861
2862 /* There is a non-null envelope sender. Build the header using the original
2863 sender address, before any rewriting that might have been done while
2864 verifying it. */
2865
2866 else
2867 {
87ba3f5f 2868 header_add(htype_from, "%sFrom: %s%s%s%s\n", resent_prefix,
2fe1a124 2869 oname,
306c6c77
JH
2870 *oname ? " <" : "",
2871 sender_address_unrewritten ? sender_address_unrewritten : sender_address,
2872 *oname ? ">" : "");
059ec3d9
PH
2873
2874 from_header = header_last; /* To get it checked for Sender: */
2875 }
2876 }
2877
2878
8800895a
PH
2879/* If the sender is local (without suppress_local_fixups), or if we are in
2880submission mode and there is an authenticated_id, check that an existing From:
2881is correct, and if not, generate a Sender: header, unless disabled. Any
2882previously-existing Sender: header was removed above. Note that sender_local,
2883as well as being TRUE if the caller of exim is not trusted, is also true if a
2884trusted caller did not supply a -f argument for non-smtp input. To allow
2885trusted callers to forge From: without supplying -f, we have to test explicitly
2886here. If the From: header contains more than one address, then the call to
2887parse_extract_address fails, and a Sender: header is inserted, as required. */
059ec3d9 2888
306c6c77 2889if ( from_header
8768d548
JH
2890 && ( f.active_local_from_check
2891 && ( f.sender_local && !f.trusted_caller && !f.suppress_local_fixups
2892 || f.submission_mode && authenticated_id
306c6c77 2893 ) ) )
059ec3d9
PH
2894 {
2895 BOOL make_sender = TRUE;
2896 int start, end, domain;
2897 uschar *errmess;
2898 uschar *from_address =
2899 parse_extract_address(Ustrchr(from_header->text, ':') + 1, &errmess,
2900 &start, &end, &domain, FALSE);
2901 uschar *generated_sender_address;
2902
8768d548 2903 generated_sender_address = f.submission_mode
306c6c77
JH
2904 ? !submission_domain
2905 ? string_sprintf("%s@%s",
2906 local_part_quote(authenticated_id), qualify_domain_sender)
2907 : !*submission_domain /* empty => full address */
2908 ? string_sprintf("%s", authenticated_id)
2909 : string_sprintf("%s@%s",
2910 local_part_quote(authenticated_id), submission_domain)
2911 : string_sprintf("%s@%s",
2912 local_part_quote(originator_login), qualify_domain_sender);
059ec3d9
PH
2913
2914 /* Remove permitted prefixes and suffixes from the local part of the From:
2915 address before doing the comparison with the generated sender. */
2916
306c6c77 2917 if (from_address)
059ec3d9
PH
2918 {
2919 int slen;
306c6c77 2920 uschar *at = domain ? from_address + domain - 1 : NULL;
059ec3d9 2921
306c6c77 2922 if (at) *at = 0;
759502e5
JH
2923 from_address += route_check_prefix(from_address, local_from_prefix, NULL);
2924 if ((slen = route_check_suffix(from_address, local_from_suffix, NULL)) > 0)
059ec3d9
PH
2925 {
2926 memmove(from_address+slen, from_address, Ustrlen(from_address)-slen);
2927 from_address += slen;
2928 }
306c6c77 2929 if (at) *at = '@';
059ec3d9 2930
306c6c77
JH
2931 if ( strcmpic(generated_sender_address, from_address) == 0
2932 || (!domain && strcmpic(from_address, originator_login) == 0))
059ec3d9
PH
2933 make_sender = FALSE;
2934 }
2935
2936 /* We have to cause the Sender header to be rewritten if there are
2937 appropriate rewriting rules. */
2938
2939 if (make_sender)
8768d548 2940 if (f.submission_mode && !submission_name)
059ec3d9
PH
2941 header_add(htype_sender, "%sSender: %s\n", resent_prefix,
2942 generated_sender_address);
2943 else
2944 header_add(htype_sender, "%sSender: %s <%s>\n",
2fe1a124 2945 resent_prefix,
8768d548 2946 f.submission_mode ? submission_name : originator_name,
2fe1a124 2947 generated_sender_address);
87ba3f5f
PH
2948
2949 /* Ensure that a non-null envelope sender address corresponds to the
2950 submission mode sender address. */
2951
8768d548 2952 if (f.submission_mode && *sender_address)
87ba3f5f 2953 {
306c6c77 2954 if (!sender_address_unrewritten)
87ba3f5f
PH
2955 sender_address_unrewritten = sender_address;
2956 sender_address = generated_sender_address;
089793a4
TF
2957 if (Ustrcmp(sender_address_unrewritten, generated_sender_address) != 0)
2958 log_write(L_address_rewrite, LOG_MAIN,
2959 "\"%s\" from env-from rewritten as \"%s\" by submission mode",
2960 sender_address_unrewritten, generated_sender_address);
87ba3f5f 2961 }
059ec3d9
PH
2962 }
2963
059ec3d9
PH
2964/* If there are any rewriting rules, apply them to the sender address, unless
2965it has already been rewritten as part of verification for SMTP input. */
2966
306c6c77 2967if (global_rewrite_rules && !sender_address_unrewritten && *sender_address)
059ec3d9
PH
2968 {
2969 sender_address = rewrite_address(sender_address, FALSE, TRUE,
2970 global_rewrite_rules, rewrite_existflags);
2971 DEBUG(D_receive|D_rewrite)
2972 debug_printf("rewritten sender = %s\n", sender_address);
2973 }
2974
2975
2976/* The headers must be run through rewrite_header(), because it ensures that
2977addresses are fully qualified, as well as applying any rewriting rules that may
2978exist.
2979
2980Qualification of header addresses in a message from a remote host happens only
2981if the host is in sender_unqualified_hosts or recipient_unqualified hosts, as
2982appropriate. For local messages, qualification always happens, unless -bnq is
2983used to explicitly suppress it. No rewriting is done for an unqualified address
2984that is left untouched.
2985
2986We start at the second header, skipping our own Received:. This rewriting is
2987documented as happening *after* recipient addresses are taken from the headers
2988by the -t command line option. An added Sender: gets rewritten here. */
2989
d7978c0f 2990for (header_line * h = header_list->next; h; h = h->next)
059ec3d9
PH
2991 {
2992 header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
2993 rewrite_existflags, TRUE);
1ebe15c3 2994 if (newh) h = newh;
059ec3d9
PH
2995 }
2996
2997
2998/* An RFC 822 (sic) message is not legal unless it has at least one of "to",
2cbb4081 2999"cc", or "bcc". Note that although the minimal examples in RFC 822 show just
059ec3d9
PH
3000"to" or "bcc", the full syntax spec allows "cc" as well. If any resent- header
3001exists, this applies to the set of resent- headers rather than the normal set.
3002
2cbb4081
PH
3003The requirement for a recipient header has been removed in RFC 2822. At this
3004point in the code, earlier versions of Exim added a To: header for locally
3005submitted messages, and an empty Bcc: header for others. In the light of the
3006changes in RFC 2822, this was dropped in November 2003. */
059ec3d9 3007
059ec3d9
PH
3008
3009/* If there is no date header, generate one if the message originates locally
8800895a
PH
3010(i.e. not over TCP/IP) and suppress_local_fixups is not set, or if the
3011submission mode flag is set. Messages without Date: are not valid, but it seems
e7e680d6
PP
3012to be more confusing if Exim adds one to all remotely-originated messages.
3013As per Message-Id, we prepend if resending, else append.
3014*/
059ec3d9 3015
306c6c77 3016if ( !date_header_exists
8768d548 3017 && ((!sender_host_address && !f.suppress_local_fixups) || f.submission_mode))
e7e680d6
PP
3018 header_add_at_position(!resents_exist, NULL, FALSE, htype_other,
3019 "%sDate: %s\n", resent_prefix, tod_stamp(tod_full));
059ec3d9
PH
3020
3021search_tidyup(); /* Free any cached resources */
3022
3023/* Show the complete set of headers if debugging. Note that the first one (the
3024new Received:) has not yet been set. */
3025
3026DEBUG(D_receive)
3027 {
3028 debug_printf(">>Headers after rewriting and local additions:\n");
d7978c0f 3029 for (header_line * h = header_list->next; h; h = h->next)
059ec3d9
PH
3030 debug_printf("%c %s", h->type, h->text);
3031 debug_printf("\n");
3032 }
3033
3034/* The headers are now complete in store. If we are running in filter
3035testing mode, that is all this function does. Return TRUE if the message
3036ended with a dot. */
3037
f05da2e8 3038if (filter_test != FTEST_NONE)
059ec3d9
PH
3039 {
3040 process_info[process_info_len] = 0;
3041 return message_ended == END_DOT;
3042 }
3043
7e3ce68e
JH
3044/*XXX CHUNKING: need to cancel cutthrough under BDAT, for now. In future,
3045think more if it could be handled. Cannot do onward CHUNKING unless
3046inbound is, but inbound chunking ought to be ok with outbound plain.
3047Could we do onward CHUNKING given inbound CHUNKING?
3048*/
3049if (chunking_state > CHUNKING_OFFERED)
57cc2785 3050 cancel_cutthrough_connection(FALSE, US"chunking active");
7e3ce68e 3051
817d9f57 3052/* Cutthrough delivery:
5032d1cf
JH
3053We have to create the Received header now rather than at the end of reception,
3054so the timestamp behaviour is a change to the normal case.
5032d1cf 3055Having created it, send the headers to the destination. */
57cc2785 3056
74f1a423 3057if (cutthrough.cctx.sock >= 0 && cutthrough.delivery)
e4bdf652 3058 {
817d9f57
JH
3059 if (received_count > received_headers_max)
3060 {
57cc2785 3061 cancel_cutthrough_connection(TRUE, US"too many headers");
817d9f57
JH
3062 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
3063 log_write(0, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3064 "Too many \"Received\" headers",
3065 sender_address,
57cc2785
JH
3066 sender_fullhost ? "H=" : "", sender_fullhost ? sender_fullhost : US"",
3067 sender_ident ? "U=" : "", sender_ident ? sender_ident : US"");
817d9f57
JH
3068 message_id[0] = 0; /* Indicate no message accepted */
3069 smtp_reply = US"550 Too many \"Received\" headers - suspected mail loop";
3070 goto TIDYUP; /* Skip to end of function */
3071 }
e4bdf652 3072 received_header_gen();
578d43dc 3073 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652
JH
3074 (void) cutthrough_headers_send();
3075 }
61147df4 3076
e4bdf652 3077
059ec3d9
PH
3078/* Open a new spool file for the data portion of the message. We need
3079to access it both via a file descriptor and a stream. Try to make the
41313d92 3080directory if it isn't there. */
059ec3d9 3081
41313d92 3082spool_name = spool_fname(US"input", message_subdir, message_id, US"-D");
a2da3176
JH
3083DEBUG(D_receive) debug_printf("Data file name: %s\n", spool_name);
3084
3085if ((data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE)) < 0)
059ec3d9
PH
3086 {
3087 if (errno == ENOENT)
3088 {
0971ec06 3089 (void) directory_make(spool_directory,
41313d92
JH
3090 spool_sname(US"input", message_subdir),
3091 INPUT_DIRECTORY_MODE, TRUE);
059ec3d9
PH
3092 data_fd = Uopen(spool_name, O_RDWR|O_CREAT|O_EXCL, SPOOL_MODE);
3093 }
3094 if (data_fd < 0)
3095 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Failed to create spool file %s: %s",
3096 spool_name, strerror(errno));
3097 }
3098
3099/* Make sure the file's group is the Exim gid, and double-check the mode
3100because the group setting doesn't always get set automatically. */
3101
b66fecb4 3102if (0 != exim_fchown(data_fd, exim_uid, exim_gid, spool_name))
1ac6b2e7
JH
3103 log_write(0, LOG_MAIN|LOG_PANIC_DIE,
3104 "Failed setting ownership on spool file %s: %s",
3105 spool_name, strerror(errno));
ff790e47 3106(void)fchmod(data_fd, SPOOL_MODE);
059ec3d9
PH
3107
3108/* We now have data file open. Build a stream for it and lock it. We lock only
3109the first line of the file (containing the message ID) because otherwise there
3110are problems when Exim is run under Cygwin (I'm told). See comments in
3111spool_in.c, where the same locking is done. */
3112
1bd642c2 3113spool_data_file = fdopen(data_fd, "w+");
059ec3d9
PH
3114lock_data.l_type = F_WRLCK;
3115lock_data.l_whence = SEEK_SET;
3116lock_data.l_start = 0;
3117lock_data.l_len = SPOOL_DATA_START_OFFSET;
3118
3119if (fcntl(data_fd, F_SETLK, &lock_data) < 0)
3120 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Cannot lock %s (%d): %s", spool_name,
3121 errno, strerror(errno));
3122
3123/* We have an open, locked data file. Write the message id to it to make it
3124self-identifying. Then read the remainder of the input of this message and
3125write it to the data file. If the variable next != NULL, it contains the first
3126data line (which was read as a header but then turned out not to have the right
3127format); write it (remembering that it might contain binary zeros). The result
3128of fwrite() isn't inspected; instead we call ferror() below. */
3129
1bd642c2 3130fprintf(spool_data_file, "%s-D\n", message_id);
306c6c77 3131if (next)
059ec3d9
PH
3132 {
3133 uschar *s = next->text;
3134 int len = next->slen;
1bd642c2 3135 if (fwrite(s, 1, len, spool_data_file) == len) /* "if" for compiler quietening */
cfbb0d24 3136 body_linecount++; /* Assumes only 1 line */
059ec3d9
PH
3137 }
3138
3139/* Note that we might already be at end of file, or the logical end of file
3140(indicated by '.'), or might have encountered an error while writing the
3141message id or "next" line. */
3142
1bd642c2 3143if (!ferror(spool_data_file) && !(receive_feof)() && message_ended != END_DOT)
059ec3d9
PH
3144 {
3145 if (smtp_input)
3146 {
328c5688 3147 message_ended = chunking_state <= CHUNKING_OFFERED
1bd642c2 3148 ? read_message_data_smtp(spool_data_file)
328c5688 3149 : spool_wireformat
1bd642c2
JH
3150 ? read_message_bdat_smtp_wire(spool_data_file)
3151 : read_message_bdat_smtp(spool_data_file);
059ec3d9
PH
3152 receive_linecount++; /* The terminating "." line */
3153 }
cfbb0d24 3154 else
1bd642c2 3155 message_ended = read_message_data(spool_data_file);
059ec3d9
PH
3156
3157 receive_linecount += body_linecount; /* For BSMTP errors mainly */
2e0c1448 3158 message_linecount += body_linecount;
059ec3d9 3159
7e3ce68e 3160 switch (message_ended)
059ec3d9 3161 {
7e3ce68e 3162 /* Handle premature termination of SMTP */
059ec3d9 3163
7e3ce68e
JH
3164 case END_EOF:
3165 if (smtp_input)
3166 {
3167 Uunlink(spool_name); /* Lose data file when closed */
57cc2785 3168 cancel_cutthrough_connection(TRUE, US"sender closed connection");
7e3ce68e
JH
3169 message_id[0] = 0; /* Indicate no message accepted */
3170 smtp_reply = handle_lost_connection(US"");
3171 smtp_yield = FALSE;
3172 goto TIDYUP; /* Skip to end of function */
3173 }
3174 break;
059ec3d9 3175
7e3ce68e
JH
3176 /* Handle message that is too big. Don't use host_or_ident() in the log
3177 message; we want to see the ident value even for non-remote messages. */
059ec3d9 3178
7e3ce68e
JH
3179 case END_SIZE:
3180 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3181 cancel_cutthrough_connection(TRUE, US"mail too big");
7e3ce68e 3182 if (smtp_input) receive_swallow_smtp(); /* Swallow incoming SMTP */
059ec3d9 3183
7e3ce68e
JH
3184 log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
3185 "message too big: read=%d max=%d",
3186 sender_address,
306c6c77
JH
3187 sender_fullhost ? " H=" : "",
3188 sender_fullhost ? sender_fullhost : US"",
3189 sender_ident ? " U=" : "",
3190 sender_ident ? sender_ident : US"",
7e3ce68e
JH
3191 message_size,
3192 thismessage_size_limit);
3193
3194 if (smtp_input)
3195 {
3196 smtp_reply = US"552 Message size exceeds maximum permitted";
3197 message_id[0] = 0; /* Indicate no message accepted */
3198 goto TIDYUP; /* Skip to end of function */
3199 }
3200 else
3201 {
1bd642c2 3202 fseek(spool_data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
7e3ce68e
JH
3203 give_local_error(ERRMESS_TOOBIG,
3204 string_sprintf("message too big (max=%d)", thismessage_size_limit),
1bd642c2 3205 US"message rejected: ", error_rc, spool_data_file, header_list);
7e3ce68e
JH
3206 /* Does not return */
3207 }
3208 break;
3209
3210 /* Handle bad BDAT protocol sequence */
3211
3212 case END_PROTOCOL:
3213 Uunlink(spool_name); /* Lose the data file when closed */
57cc2785 3214 cancel_cutthrough_connection(TRUE, US"sender protocol error");
7e3ce68e
JH
3215 smtp_reply = US""; /* Response already sent */
3216 message_id[0] = 0; /* Indicate no message accepted */
3217 goto TIDYUP; /* Skip to end of function */
059ec3d9
PH
3218 }
3219 }
3220
3221/* Restore the standard SIGALRM handler for any subsequent processing. (For
3222example, there may be some expansion in an ACL that uses a timer.) */
3223
3224os_non_restarting_signal(SIGALRM, sigalrm_handler);
3225
3226/* The message body has now been read into the data file. Call fflush() to
3227empty the buffers in C, and then call fsync() to get the data written out onto
3228the disk, as fflush() doesn't do this (or at least, it isn't documented as
3229having to do this). If there was an I/O error on either input or output,
3230attempt to send an error message, and unlink the spool file. For non-SMTP input
3231we can then give up. Note that for SMTP input we must swallow the remainder of
3232the input in cases of output errors, since the far end doesn't expect to see
3233anything until the terminating dot line is sent. */
3234
1bd642c2
JH
3235if (fflush(spool_data_file) == EOF || ferror(spool_data_file) ||
3236 EXIMfsync(fileno(spool_data_file)) < 0 || (receive_ferror)())
059ec3d9
PH
3237 {
3238 uschar *msg_errno = US strerror(errno);
3239 BOOL input_error = (receive_ferror)() != 0;
3240 uschar *msg = string_sprintf("%s error (%s) while receiving message from %s",
3241 input_error? "Input read" : "Spool write",
3242 msg_errno,
306c6c77 3243 sender_fullhost ? sender_fullhost : sender_ident);
059ec3d9
PH
3244
3245 log_write(0, LOG_MAIN, "Message abandoned: %s", msg);
3246 Uunlink(spool_name); /* Lose the data file */
57cc2785 3247 cancel_cutthrough_connection(TRUE, US"error writing spoolfile");
059ec3d9
PH
3248
3249 if (smtp_input)
3250 {
3251 if (input_error)
3252 smtp_reply = US"451 Error while reading input data";
3253 else
3254 {
3255 smtp_reply = US"451 Error while writing spool file";
3256 receive_swallow_smtp();
3257 }
3258 message_id[0] = 0; /* Indicate no message accepted */
3259 goto TIDYUP; /* Skip to end of function */
3260 }
3261
3262 else
3263 {
1bd642c2
JH
3264 fseek(spool_data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
3265 give_local_error(ERRMESS_IOERR, msg, US"", error_rc, spool_data_file,
059ec3d9
PH
3266 header_list);
3267 /* Does not return */
3268 }
3269 }
3270
3271
3272/* No I/O errors were encountered while writing the data file. */
3273
3274DEBUG(D_receive) debug_printf("Data file written for message %s\n", message_id);
306c6c77 3275if (LOGGING(receive_time)) timesince(&received_time_taken, &received_time);
059ec3d9
PH
3276
3277
3278/* If there were any bad addresses extracted by -t, or there were no recipients
3279left after -t, send a message to the sender of this message, or write it to
3280stderr if the error handling option is set that way. Note that there may
3281legitimately be no recipients for an SMTP message if they have all been removed
3282by "discard".
3283
3284We need to rewind the data file in order to read it. In the case of no
3285recipients or stderr error writing, throw the data file away afterwards, and
3286exit. (This can't be SMTP, which always ensures there's at least one
3287syntactically good recipient address.) */
3288
306c6c77 3289if (extract_recip && (bad_addresses || recipients_count == 0))
059ec3d9
PH
3290 {
3291 DEBUG(D_receive)
3292 {
3293 if (recipients_count == 0) debug_printf("*** No recipients\n");
306c6c77 3294 if (bad_addresses)
059ec3d9 3295 {
059ec3d9 3296 debug_printf("*** Bad address(es)\n");
d7978c0f 3297 for (error_block * eblock = bad_addresses; eblock; eblock = eblock->next)
059ec3d9 3298 debug_printf(" %s: %s\n", eblock->text1, eblock->text2);
059ec3d9
PH
3299 }
3300 }
3301
569a8b23
JH
3302 log_write(0, LOG_MAIN|LOG_PANIC, "%s %s found in headers",
3303 message_id, bad_addresses ? "bad addresses" : "no recipients");
3304
1bd642c2 3305 fseek(spool_data_file, (long int)SPOOL_DATA_START_OFFSET, SEEK_SET);
059ec3d9
PH
3306
3307 /* If configured to send errors to the sender, but this fails, force
3308 a failure error code. We use a special one for no recipients so that it
3309 can be detected by the autoreply transport. Otherwise error_rc is set to
3310 errors_sender_rc, which is EXIT_FAILURE unless -oee was given, in which case
3311 it is EXIT_SUCCESS. */
3312
3313 if (error_handling == ERRORS_SENDER)
3314 {
3315 if (!moan_to_sender(
569a8b23
JH
3316 bad_addresses
3317 ? recipients_list ? ERRMESS_BADADDRESS : ERRMESS_BADNOADDRESS
3318 : extracted_ignored ? ERRMESS_IGADDRESS : ERRMESS_NOADDRESS,
3319 bad_addresses, header_list, spool_data_file, FALSE
3320 ) )
3321 error_rc = bad_addresses ? EXIT_FAILURE : EXIT_NORECIPIENTS;
059ec3d9
PH
3322 }
3323 else
3324 {
306c6c77 3325 if (!bad_addresses)
059ec3d9
PH
3326 if (extracted_ignored)
3327 fprintf(stderr, "exim: all -t recipients overridden by command line\n");
3328 else
3329 fprintf(stderr, "exim: no recipients in message\n");
059ec3d9
PH
3330 else
3331 {
3332 fprintf(stderr, "exim: invalid address%s",
cfbb0d24
JH
3333 bad_addresses->next ? "es:\n" : ":");
3334 for ( ; bad_addresses; bad_addresses = bad_addresses->next)
059ec3d9
PH
3335 fprintf(stderr, " %s: %s\n", bad_addresses->text1,
3336 bad_addresses->text2);
059ec3d9
PH
3337 }
3338 }
3339
3340 if (recipients_count == 0 || error_handling == ERRORS_STDERR)
3341 {
3342 Uunlink(spool_name);
1bd642c2 3343 (void)fclose(spool_data_file);
81022793 3344 exim_exit(error_rc);
059ec3d9
PH
3345 }
3346 }
3347
3348/* Data file successfully written. Generate text for the Received: header by
3349expanding the configured string, and adding a timestamp. By leaving this
3350operation till now, we ensure that the timestamp is the time that message
3351reception was completed. However, this is deliberately done before calling the
3352data ACL and local_scan().
3353
3354This Received: header may therefore be inspected by the data ACL and by code in
3355the local_scan() function. When they have run, we update the timestamp to be
3356the final time of reception.
3357
3358If there is just one recipient, set up its value in the $received_for variable
3359for use when we generate the Received: header.
3360
3361Note: the checking for too many Received: headers is handled by the delivery
3362code. */
e4bdf652 3363/*XXX eventually add excess Received: check for cutthrough case back when classifying them */
059ec3d9 3364
306c6c77 3365if (!received_header->text) /* Non-cutthrough case */
059ec3d9 3366 {
e4bdf652 3367 received_header_gen();
059ec3d9 3368
e4bdf652 3369 /* Set the value of message_body_size for the DATA ACL and for local_scan() */
059ec3d9 3370
e4bdf652
JH
3371 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3372 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9 3373
e4bdf652
JH
3374 /* If an ACL from any RCPT commands set up any warning headers to add, do so
3375 now, before running the DATA ACL. */
059ec3d9 3376
578d43dc 3377 add_acl_headers(ACL_WHERE_RCPT, US"MAIL or RCPT");
e4bdf652 3378 }
817d9f57 3379else
e4bdf652
JH
3380 message_body_size = (fstat(data_fd, &statbuf) == 0)?
3381 statbuf.st_size - SPOOL_DATA_START_OFFSET : -1;
059ec3d9
PH
3382
3383/* If an ACL is specified for checking things at this stage of reception of a
3384message, run it, unless all the recipients were removed by "discard" in earlier
3385ACLs. That is the only case in which recipients_count can be zero at this
3386stage. Set deliver_datafile to point to the data file so that $message_body and
3387$message_body_end can be extracted if needed. Allow $recipients in expansions.
3388*/
3389
3390deliver_datafile = data_fd;
4e88a19f 3391user_msg = NULL;
059ec3d9 3392
8768d548 3393f.enable_dollar_recipients = TRUE;
0e20aff9 3394
059ec3d9 3395if (recipients_count == 0)
8768d548 3396 blackholed_by = f.recipients_discarded ? US"MAIL ACL" : US"RCPT ACL";
7e3ce68e 3397
059ec3d9
PH
3398else
3399 {
059ec3d9
PH
3400 /* Handle interactive SMTP messages */
3401
3402 if (smtp_input && !smtp_batched_input)
3403 {
8523533c 3404
80a47a2c 3405#ifndef DISABLE_DKIM
8768d548 3406 if (!f.dkim_disable_verify)
80a47a2c 3407 {
cc55f420 3408 /* Finish verification */
80a47a2c
TK
3409 dkim_exim_verify_finish();
3410
3411 /* Check if we must run the DKIM ACL */
7e3ce68e 3412 if (acl_smtp_dkim && dkim_verify_signers && *dkim_verify_signers)
80a47a2c 3413 {
cc55f420 3414 uschar * dkim_verify_signers_expanded =
80a47a2c 3415 expand_string(dkim_verify_signers);
cc55f420
JH
3416 gstring * results = NULL;
3417 int signer_sep = 0;
3418 const uschar * ptr;
3419 uschar * item;
3420 gstring * seen_items = NULL;
3421 int old_pool = store_pool;
3422
3423 store_pool = POOL_PERM; /* Allow created variables to live to data ACL */
3424
3425 if (!(ptr = dkim_verify_signers_expanded))
80a47a2c
TK
3426 log_write(0, LOG_MAIN|LOG_PANIC,
3427 "expansion of dkim_verify_signers option failed: %s",
3428 expand_string_message);
7e3ce68e 3429
cc55f420
JH
3430 /* Default to OK when no items are present */
3431 rc = OK;
3432 while ((item = string_nextinlist(&ptr, &signer_sep, NULL, 0)))
3433 {
3434 /* Prevent running ACL for an empty item */
3435 if (!item || !*item) continue;
3436
3437 /* Only run ACL once for each domain or identity,
3438 no matter how often it appears in the expanded list. */
3439 if (seen_items)
3440 {
3441 uschar * seen_item;
3442 const uschar * seen_items_list = string_from_gstring(seen_items);
3443 int seen_sep = ':';
3444 BOOL seen_this_item = FALSE;
3445
3446 while ((seen_item = string_nextinlist(&seen_items_list, &seen_sep,
3447 NULL, 0)))
3448 if (Ustrcmp(seen_item,item) == 0)
3449 {
3450 seen_this_item = TRUE;
3451 break;
3452 }
3453
3454 if (seen_this_item)
5032d1cf
JH
3455 {
3456 DEBUG(D_receive)
cc55f420
JH
3457 debug_printf("acl_smtp_dkim: skipping signer %s, "
3458 "already seen\n", item);
3459 continue;
5032d1cf 3460 }
cc55f420 3461
b2bcdd35 3462 seen_items = string_catn(seen_items, US":", 1);
cc55f420 3463 }
cc55f420
JH
3464 seen_items = string_cat(seen_items, item);
3465
18067c75 3466 rc = dkim_exim_acl_run(item, &results, &user_msg, &log_msg);
cc55f420
JH
3467 if (rc != OK)
3468 {
3469 DEBUG(D_receive)
3470 debug_printf("acl_smtp_dkim: acl_check returned %d on %s, "
3471 "skipping remaining items\n", rc, item);
3472 cancel_cutthrough_connection(TRUE, US"dkim acl not ok");
3473 break;
3474 }
3475 }
3476 dkim_verify_status = string_from_gstring(results);
3477 store_pool = old_pool;
3478 add_acl_headers(ACL_WHERE_DKIM, US"DKIM");
3479 if (rc == DISCARD)
3480 {
3481 recipients_count = 0;
3482 blackholed_by = US"DKIM ACL";
3483 if (log_msg)
3484 blackhole_log_msg = string_sprintf(": %s", log_msg);
3485 }
3486 else if (rc != OK)
3487 {
3488 Uunlink(spool_name);
3489 if (smtp_handle_acl_fail(ACL_WHERE_DKIM, rc, user_msg, log_msg) != 0)
3490 smtp_yield = FALSE; /* No more messages after dropped connection */
3491 smtp_reply = US""; /* Indicate reply already sent */
3492 message_id[0] = 0; /* Indicate no message accepted */
3493 goto TIDYUP; /* Skip to end of function */
3494 }
80a47a2c 3495 }
a79d8834
JH
3496 else