xref: /freebsd/contrib/sendmail/src/collect.c (revision d39bd2c1)
1 /*
2  * Copyright (c) 1998-2006, 2008, 2023, 2024 Proofpoint, Inc. and its suppliers.
3  *	All rights reserved.
4  * Copyright (c) 1983, 1995-1997 Eric P. Allman.  All rights reserved.
5  * Copyright (c) 1988, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * By using this file, you agree to the terms and conditions set
9  * forth in the LICENSE file which can be found at the top level of
10  * the sendmail distribution.
11  *
12  */
13 
14 #include <sendmail.h>
15 
16 SM_RCSID("@(#)$Id: collect.c,v 8.287 2013-11-22 20:51:55 ca Exp $")
17 
18 #include <sm/sendmail.h>
19 
20 static void	eatfrom __P((char *volatile, ENVELOPE *));
21 static void	collect_doheader __P((ENVELOPE *));
22 static SM_FILE_T *collect_dfopen __P((ENVELOPE *));
23 static SM_FILE_T *collect_eoh __P((ENVELOPE *, int, int));
24 
25 /*
26 **  COLLECT_EOH -- end-of-header processing in collect()
27 **
28 **	Called by collect() when it encounters the blank line
29 **	separating the header from the message body, or when it
30 **	encounters EOF in a message that contains only a header.
31 **
32 **	Parameters:
33 **		e -- envelope
34 **		numhdrs -- number of headers
35 **		hdrslen -- length of headers
36 **
37 **	Returns:
38 **		NULL, or handle to open data file
39 **
40 **	Side Effects:
41 **		end-of-header check ruleset is invoked.
42 **		envelope state is updated.
43 **		headers may be added and deleted.
44 **		selects the queue.
45 **		opens the data file.
46 */
47 
48 static SM_FILE_T *
collect_eoh(e,numhdrs,hdrslen)49 collect_eoh(e, numhdrs, hdrslen)
50 	ENVELOPE *e;
51 	int numhdrs;
52 	int hdrslen;
53 {
54 	char hnum[16];
55 	char hsize[16];
56 
57 	/* call the end-of-header check ruleset */
58 	(void) sm_snprintf(hnum, sizeof(hnum), "%d", numhdrs);
59 	(void) sm_snprintf(hsize, sizeof(hsize), "%d", hdrslen);
60 	if (tTd(30, 10))
61 		sm_dprintf("collect: rscheck(\"check_eoh\", \"%s $| %s\")\n",
62 			   hnum, hsize);
63 	(void) rscheck("check_eoh", hnum, hsize, e, RSF_UNSTRUCTURED|RSF_COUNT,
64 			3, NULL, e->e_id, NULL, NULL);
65 
66 	/*
67 	**  Process the header,
68 	**  select the queue, open the data file.
69 	*/
70 
71 	collect_doheader(e);
72 	return collect_dfopen(e);
73 }
74 
75 /*
76 **  COLLECT_DOHEADER -- process header in collect()
77 **
78 **	Called by collect() after it has finished parsing the header,
79 **	but before it selects the queue and creates the data file.
80 **	The results of processing the header will affect queue selection.
81 **
82 **	Parameters:
83 **		e -- envelope
84 **
85 **	Returns:
86 **		none.
87 **
88 **	Side Effects:
89 **		envelope state is updated.
90 **		headers may be added and deleted.
91 */
92 
93 static void
collect_doheader(e)94 collect_doheader(e)
95 	ENVELOPE *e;
96 {
97 	/*
98 	**  Find out some information from the headers.
99 	**	Examples are who is the from person & the date.
100 	*/
101 
102 	eatheader(e, true, false);
103 
104 	if (GrabTo && e->e_sendqueue == NULL)
105 		usrerr("No recipient addresses found in header");
106 
107 	/*
108 	**  If we have a Return-Receipt-To:, turn it into a DSN.
109 	*/
110 
111 	if (RrtImpliesDsn && hvalue("return-receipt-to", e->e_header) != NULL)
112 	{
113 		ADDRESS *q;
114 
115 		for (q = e->e_sendqueue; q != NULL; q = q->q_next)
116 			if (!bitset(QHASNOTIFY, q->q_flags))
117 				q->q_flags |= QHASNOTIFY|QPINGONSUCCESS;
118 	}
119 
120 	/*
121 	**  Add an appropriate recipient line if we have none.
122 	*/
123 
124 	if (hvalue("to", e->e_header) != NULL ||
125 	    hvalue("cc", e->e_header) != NULL ||
126 	    hvalue("apparently-to", e->e_header) != NULL)
127 	{
128 		/* have a valid recipient header -- delete Bcc: headers */
129 		e->e_flags |= EF_DELETE_BCC;
130 	}
131 	else if (hvalue("bcc", e->e_header) == NULL)
132 	{
133 		/* no valid recipient headers */
134 		register ADDRESS *q;
135 		char *hdr = NULL;
136 
137 		/* create a recipient field */
138 		switch (NoRecipientAction)
139 		{
140 		  case NRA_ADD_APPARENTLY_TO:
141 			hdr = "Apparently-To";
142 			break;
143 
144 		  case NRA_ADD_TO:
145 			hdr = "To";
146 			break;
147 
148 		  case NRA_ADD_BCC:
149 			addheader("Bcc", " ", 0, e, true);
150 			break;
151 
152 		  case NRA_ADD_TO_UNDISCLOSED:
153 			addheader("To", "undisclosed-recipients:;", 0, e, true);
154 			break;
155 		}
156 
157 		if (hdr != NULL)
158 		{
159 			for (q = e->e_sendqueue; q != NULL; q = q->q_next)
160 			{
161 				if (q->q_alias != NULL)
162 					continue;
163 				if (tTd(30, 3))
164 					sm_dprintf("Adding %s: %s\n",
165 						hdr, q->q_paddr);
166 				addheader(hdr, q->q_paddr, 0, e, true);
167 			}
168 		}
169 	}
170 }
171 
172 /*
173 **  COLLECT_DFOPEN -- open the message data file
174 **
175 **	Called by collect() after it has finished processing the header.
176 **	Queue selection occurs at this point, possibly based on the
177 **	envelope's recipient list and on header information.
178 **
179 **	Parameters:
180 **		e -- envelope
181 **
182 **	Returns:
183 **		NULL, or a pointer to an open data file,
184 **		into which the message body will be written by collect().
185 **
186 **	Side Effects:
187 **		Calls syserr, sets EF_FATALERRS and returns NULL
188 **		if there is insufficient disk space.
189 **		Aborts process if data file could not be opened.
190 **		Otherwise, the queue is selected,
191 **		e->e_{dfino,dfdev,msgsize,flags} are updated,
192 **		and a pointer to an open data file is returned.
193 */
194 
195 static SM_FILE_T *
collect_dfopen(e)196 collect_dfopen(e)
197 	ENVELOPE *e;
198 {
199 	MODE_T oldumask = 0;
200 	int dfd;
201 	struct stat stbuf;
202 	SM_FILE_T *df;
203 	char *dfname;
204 
205 	if (!setnewqueue(e))
206 		return NULL;
207 
208 	dfname = queuename(e, DATAFL_LETTER);
209 	if (bitset(S_IWGRP, QueueFileMode))
210 		oldumask = umask(002);
211 	df = bfopen(dfname, QueueFileMode, DataFileBufferSize,
212 		    SFF_OPENASROOT);
213 	if (bitset(S_IWGRP, QueueFileMode))
214 		(void) umask(oldumask);
215 	if (df == NULL)
216 	{
217 		syserr("@Cannot create %s", dfname);
218 		e->e_flags |= EF_NO_BODY_RETN;
219 		flush_errors(true);
220 		finis(false, true, ExitStat);
221 		/* NOTREACHED */
222 	}
223 	dfd = sm_io_getinfo(df, SM_IO_WHAT_FD, NULL);
224 	if (dfd < 0 || fstat(dfd, &stbuf) < 0)
225 		e->e_dfino = -1;
226 	else
227 	{
228 		e->e_dfdev = stbuf.st_dev;
229 		e->e_dfino = stbuf.st_ino;
230 	}
231 	e->e_flags |= EF_HAS_DF;
232 	return df;
233 }
234 
235 /*
236 **  INCBUFLEN -- increase buflen for the header buffer in collect()
237 **
238 **	Parameters:
239 **		buflen -- current size of buffer
240 **
241 **	Returns:
242 **		new buflen
243 */
244 
245 static int incbuflen __P((int));
246 static int
incbuflen(buflen)247 incbuflen(buflen)
248 	int buflen;
249 {
250 	int newlen;
251 
252 	/* this also handles the case of MaxMessageSize == 0 */
253 	if (MaxMessageSize <= MEMCHUNKSIZE)
254 	{
255 		if (buflen < MEMCHUNKSIZE)
256 			return buflen * 2;
257 		else
258 			return buflen + MEMCHUNKSIZE;
259 	}
260 
261 	/* MaxMessageSize > MEMCHUNKSIZE */
262 	newlen = buflen * 2;
263 	if (newlen > 0 && newlen < MaxMessageSize)
264 		return newlen;
265 	else
266 		return MaxMessageSize;
267 }
268 
269 #if _FFR_TESTS
270 /* just for testing/debug output */
271 static const char *
makeprint(c)272 makeprint(c)
273 	char c;
274 {
275 	static char prt[6];
276 
277 	prt[1] = '\0';
278 	prt[2] = '\0';
279 	if (isprint((unsigned char)c))
280 		prt[0] = c;
281 	else if ('\n' == c)
282 	{
283 		prt[0] = 'L';
284 		prt[1] = 'F';
285 	}
286 	else if ('\r' == c)
287 	{
288 		prt[0] = 'C';
289 		prt[1] = 'R';
290 	}
291 	else
292 		snprintf(prt, sizeof(prt), "%o", c);
293 	return prt;
294 }
295 #else /* _FFR_TESTS */
296 # define makeprint(c)	"X"
297 #endif /* _FFR_TESTS */
298 
299 /*
300 **  COLLECT -- read & parse message header & make temp file.
301 **
302 **	Creates a temporary file name and copies the standard
303 **	input to that file.  Leading UNIX-style "From" lines are
304 **	stripped off (after important information is extracted).
305 **
306 **	Parameters:
307 **		fp -- file to read.
308 **		smtpmode -- if >= SMTPMODE_LAX we are running SMTP:
309 **			give an RFC821 style message to say we are
310 **			ready to collect input, and never ignore
311 **			a single dot to mean end of message.
312 **		hdrp -- the location to stash the header.
313 **		e -- the current envelope.
314 **		rsetsize -- reset e_msgsize?
315 **
316 **	Returns:
317 **		none.
318 **
319 **	Side Effects:
320 **		If successful,
321 **		- Data file is created and filled, and e->e_dfp is set.
322 **		- The from person may be set.
323 **		If the "enough disk space" check fails,
324 **		- syserr is called.
325 **		- e->e_dfp is NULL.
326 **		- e->e_flags & EF_FATALERRS is set.
327 **		- collect() returns.
328 **		If data file cannot be created, the process is terminated.
329 */
330 
331 /* values for input state machine */
332 #define IS_NORM		0	/* middle of line */
333 #define IS_BOL		1	/* beginning of line */
334 #define IS_DOT		2	/* read "." at beginning of line */
335 #define IS_DOTCR	3	/* read ".\r" at beginning of line */
336 #define IS_CR		4	/* read "\r" */
337 
338 /* hack to enhance readability of debug output */
339 static const char *istates[] = { "NORM", "BOL", "DOT", "DOTCR", "CR" };
340 #define ISTATE istates[istate]
341 
342 /* values for message state machine */
343 #define MS_UFROM	0	/* reading Unix from line */
344 #define MS_HEADER	1	/* reading message header */
345 #define MS_BODY		2	/* reading message body */
346 #define MS_DISCARD	3	/* discarding rest of message */
347 #define BARE_LF_MSG "Bare linefeed (LF) not allowed"
348 #define BARE_CR_MSG "Bare carriage return (CR) not allowed"
349 
350 void
collect(fp,smtpmode,hdrp,e,rsetsize)351 collect(fp, smtpmode, hdrp, e, rsetsize)
352 	SM_FILE_T *fp;
353 	int smtpmode;
354 	HDR **hdrp;
355 	register ENVELOPE *e;
356 	bool rsetsize;
357 {
358 	register SM_FILE_T *df;
359 	bool ignrdot;
360 	int dbto;
361 	register char *bp;
362 	int c;
363 	bool inputerr;
364 	bool headeronly;
365 	char *buf;
366 	int buflen;
367 	int istate;
368 	int mstate;
369 	int hdrslen;
370 	int numhdrs;
371 	int afd;
372 	int old_rd_tmo;
373 	unsigned char *pbp;
374 	unsigned char peekbuf[8];
375 	char bufbuf[MAXLINE];
376 #if _FFR_REJECT_NUL_BYTE
377 	bool hasNUL;		/* has at least one NUL input byte */
378 #endif
379 	int bare_lf, bare_cr;
380 
381 #define SMTPMODE	(smtpmode >= SMTPMODE_LAX)
382 #define SMTPMODE_STRICT	((smtpmode & SMTPMODE_CRLF) != 0)
383 #define BARE_LF_421	((smtpmode & SMTPMODE_LF_421) != 0)
384 #define BARE_CR_421	((smtpmode & SMTPMODE_CR_421) != 0)
385 #define BARE_LF_SP	((smtpmode & SMTPMODE_LF_SP) != 0)
386 #define BARE_CR_SP	((smtpmode & SMTPMODE_CR_SP) != 0)
387 
388 /* for bare_{lf,cr} */
389 #define BARE_IN_HDR	0x01
390 #define BARE_IN_BDY	0x02
391 #define BARE_WHERE	((MS_BODY == mstate) ? BARE_IN_BDY : BARE_IN_HDR)
392 
393 	df = NULL;
394 	ignrdot = SMTPMODE ? false : IgnrDot;
395 	bare_lf = bare_cr = 0;
396 
397 	/* timeout for I/O functions is in milliseconds */
398 	dbto = SMTPMODE ? ((int) TimeOuts.to_datablock * 1000)
399 			: SM_TIME_FOREVER;
400 	sm_io_setinfo(fp, SM_IO_WHAT_TIMEOUT, &dbto);
401 	old_rd_tmo = set_tls_rd_tmo(TimeOuts.to_datablock);
402 	c = SM_IO_EOF;
403 	inputerr = false;
404 	headeronly = hdrp != NULL;
405 	hdrslen = 0;
406 	numhdrs = 0;
407 	HasEightBits = false;
408 #if _FFR_REJECT_NUL_BYTE
409 	hasNUL = false;
410 #endif
411 	buf = bp = bufbuf;
412 	buflen = sizeof(bufbuf);
413 	pbp = peekbuf;
414 	istate = IS_BOL;
415 	mstate = SaveFrom ? MS_HEADER : MS_UFROM;
416 
417 	/*
418 	**  Tell ARPANET to go ahead.
419 	*/
420 
421 	if (SMTPMODE)
422 		message("354 End data with <CR><LF>.<CR><LF>");
423 
424 	/* simulate an I/O timeout when used as sink */
425 	if (tTd(83, 101))
426 		sleep(319);
427 
428 	if (tTd(30, 2))
429 		sm_dprintf("collect, smtpmode=%#x\n", smtpmode);
430 
431 	/*
432 	**  Read the message.
433 	**
434 	**	This is done using two interleaved state machines.
435 	**	The input state machine is looking for things like
436 	**	hidden dots; the message state machine is handling
437 	**	the larger picture (e.g., header versus body).
438 	*/
439 
440 	if (rsetsize)
441 		e->e_msgsize = 0;
442 	for (;;)
443 	{
444 		if (tTd(30, 35))
445 			sm_dprintf("top, istate=%s, mstate=%d\n", ISTATE,
446 				   mstate);
447 		for (;;)
448 		{
449 			if (pbp > peekbuf)
450 				c = *--pbp;
451 			else
452 			{
453 				while (!sm_io_eof(fp) && !sm_io_error(fp))
454 				{
455 					errno = 0;
456 					c = sm_io_getc(fp, SM_TIME_DEFAULT);
457 					if (c == SM_IO_EOF && errno == EINTR)
458 					{
459 						/* Interrupted, retry */
460 						sm_io_clearerr(fp);
461 						continue;
462 					}
463 
464 					/* timeout? */
465 					if (c == SM_IO_EOF && errno == EAGAIN
466 					    && SMTPMODE)
467 					{
468 						/*
469 						**  Override e_message in
470 						**  usrerr() as this is the
471 						**  reason for failure that
472 						**  should be logged for
473 						**  undelivered recipients.
474 						*/
475 
476 						e->e_message = NULL;
477 						errno = 0;
478 						inputerr = true;
479 						goto readabort;
480 					}
481 					break;
482 				}
483 				if (TrafficLogFile != NULL && !headeronly)
484 				{
485 					if (istate == IS_BOL)
486 						(void) sm_io_fprintf(TrafficLogFile,
487 							SM_TIME_DEFAULT,
488 							"%05d <<< ",
489 							(int) CurrentPid);
490 					if (c == SM_IO_EOF)
491 						(void) sm_io_fprintf(TrafficLogFile,
492 							SM_TIME_DEFAULT,
493 							"[EOF]\n");
494 					else
495 						(void) sm_io_putc(TrafficLogFile,
496 							SM_TIME_DEFAULT,
497 							c);
498 				}
499 #if _FFR_REJECT_NUL_BYTE
500 				if (c == '\0')
501 					hasNUL = true;
502 #endif
503 				if (c == SM_IO_EOF)
504 					goto readdone;
505 				if (SevenBitInput ||
506 				    bitset(EF_7BITBODY, e->e_flags))
507 					c &= 0x7f;
508 				else
509 					HasEightBits |= bitset(0x80, c);
510 			}
511 			if (tTd(30, 94))
512 				sm_dprintf("istate=%s, c=%s (0x%x)\n",
513 					ISTATE, makeprint((char) c), c);
514 			if ('\n' == c && SMTPMODE &&
515 			    !(IS_CR == istate || IS_DOTCR == istate))
516 			{
517 				bare_lf |= BARE_WHERE;
518 				if (BARE_LF_421)
519 				{
520 					inputerr = true;
521 					goto readabort;
522 				}
523 				if (BARE_LF_SP)
524 				{
525 					if (TTD(30, 64))
526 						sm_dprintf("LF: c=%s %#x\n", makeprint((char) c), c);
527 					c = ' ';
528 				}
529 			}
530 			switch (istate)
531 			{
532 			  case IS_BOL:
533 				if (c == '.')
534 				{
535 					istate = IS_DOT;
536 					continue;
537 				}
538 				break;
539 
540 			  case IS_DOT:
541 				if (c == '\n' && !ignrdot && !SMTPMODE_STRICT)
542 					goto readdone;
543 				else if (c == '\r')
544 				{
545 					istate = IS_DOTCR;
546 					continue;
547 				}
548 				else if (ignrdot ||
549 					 (c != '.' &&
550 					  OpMode != MD_SMTP &&
551 					  OpMode != MD_DAEMON &&
552 					  OpMode != MD_ARPAFTP))
553 
554 				{
555 					SM_ASSERT(pbp < peekbuf +
556 							sizeof(peekbuf));
557 					*pbp++ = c;
558 					c = '.';
559 				}
560 				break;
561 
562 			  case IS_DOTCR:
563 				if (c == '\n' && !ignrdot)
564 					goto readdone;
565 				else
566 				{
567 					/* push back the ".\rx" */
568 					SM_ASSERT(pbp < peekbuf +
569 							sizeof(peekbuf));
570 					*pbp++ = c;
571 					if (OpMode != MD_SMTP &&
572 					    OpMode != MD_DAEMON &&
573 					    OpMode != MD_ARPAFTP)
574 					{
575 						SM_ASSERT(pbp < peekbuf +
576 							 sizeof(peekbuf));
577 						*pbp++ = '\r';
578 						c = '.';
579 					}
580 					else
581 						c = '\r';
582 				}
583 				break;
584 
585 			  case IS_CR:
586 				if (c == '\n')
587 				{
588 					if (TTD(30, 64))
589 						sm_dprintf("state=CR, c=%s %#x -> BOL\n", makeprint((char) c), c);
590 					istate = IS_BOL;
591 				}
592 				else
593 				{
594 					if (TTD(30, 64))
595 						sm_dprintf("state=CR, c=%s %#x -> NORM\n", makeprint((char) c), c);
596 					if (SMTPMODE)
597 					{
598 						bare_cr |= BARE_WHERE;
599 						if (BARE_CR_421)
600 						{
601 							inputerr = true;
602 							goto readabort;
603 						}
604 					}
605 					(void) sm_io_ungetc(fp, SM_TIME_DEFAULT,
606 							    c);
607 					if (BARE_CR_SP)
608 						c = ' ';
609 					else
610 						c = '\r';
611 					istate = IS_NORM;
612 				}
613 				goto bufferchar;
614 			}
615 
616 			if (c == '\r')
617 			{
618 				istate = IS_CR;
619 				continue;
620 			}
621 			else if (c == '\n' && !SMTPMODE_STRICT)
622 				istate = IS_BOL;
623 			else
624 				istate = IS_NORM;
625 
626 bufferchar:
627 			if (!headeronly)
628 			{
629 				/* no overflow? */
630 				if (e->e_msgsize >= 0)
631 				{
632 					e->e_msgsize++;
633 					if (MaxMessageSize > 0 &&
634 					    !bitset(EF_TOOBIG, e->e_flags) &&
635 					    e->e_msgsize > MaxMessageSize)
636 						 e->e_flags |= EF_TOOBIG;
637 				}
638 			}
639 			switch (mstate)
640 			{
641 			  case MS_BODY:
642 				/* just put the character out */
643 				if (!bitset(EF_TOOBIG, e->e_flags))
644 					(void) sm_io_putc(df, SM_TIME_DEFAULT,
645 							  c);
646 				if (TTD(30, 64))
647 					sm_dprintf("state=%s, put=%s %#x\n", ISTATE, makeprint((char) c), c);
648 				/* FALLTHROUGH */
649 
650 			  case MS_DISCARD:
651 				continue;
652 			}
653 
654 			SM_ASSERT(mstate == MS_UFROM || mstate == MS_HEADER);
655 
656 			/* header -- buffer up */
657 			if (bp >= &buf[buflen - 2])
658 			{
659 				char *obuf;
660 
661 				/* out of space for header */
662 				obuf = buf;
663 				buflen = incbuflen(buflen);
664 				if (tTd(30, 32))
665 					sm_dprintf("buflen=%d, hdrslen=%d\n", buflen, hdrslen);
666 				if (buflen <= 0)
667 				{
668 					sm_syslog(LOG_NOTICE, e->e_id,
669 						  "header overflow from %s during message collect",
670 						  CURHOSTNAME);
671 					errno = 0;
672 					e->e_flags |= EF_CLRQUEUE;
673 					e->e_status = "5.6.0";
674 					usrerrenh(e->e_status,
675 						  "552 Headers too large");
676 					goto discard;
677 				}
678 				buf = xalloc(buflen);
679 				memmove(buf, obuf, bp - obuf);
680 				bp = &buf[bp - obuf];
681 				if (obuf != bufbuf)
682 					sm_free(obuf);  /* XXX */
683 			}
684 
685 			if (c != '\0')
686 			{
687 				*bp++ = c;
688 				++hdrslen;
689 				if (!headeronly &&
690 				    MaxHeadersLength > 0 &&
691 				    hdrslen > MaxHeadersLength)
692 				{
693 					sm_syslog(LOG_NOTICE, e->e_id,
694 						  "headers too large (%d max) from %s during message collect",
695 						  MaxHeadersLength,
696 						  CURHOSTNAME);
697 					errno = 0;
698 					e->e_flags |= EF_CLRQUEUE;
699 					e->e_status = "5.6.0";
700 					usrerrenh(e->e_status,
701 						  "552 Headers too large (%d max)",
702 						  MaxHeadersLength);
703   discard:
704 					mstate = MS_DISCARD;
705 				}
706 			}
707 			if (istate == IS_BOL)
708 				break;
709 		}
710 		*bp = '\0';
711 
712 nextstate:
713 		if (tTd(30, 35))
714 			sm_dprintf("nextstate, istate=%s, mstate=%d, line=\"%s\"\n",
715 				ISTATE, mstate, buf);
716 		switch (mstate)
717 		{
718 		  case MS_UFROM:
719 			mstate = MS_HEADER;
720 #ifndef NOTUNIX
721 			if (strncmp(buf, "From ", 5) == 0)
722 			{
723 				bp = buf;
724 				eatfrom(buf, e);
725 				continue;
726 			}
727 #endif /* ! NOTUNIX */
728 			/* FALLTHROUGH */
729 
730 		  case MS_HEADER:
731 			if (!isheader(buf))
732 			{
733 				mstate = MS_BODY;
734 				goto nextstate;
735 			}
736 
737 			/* check for possible continuation line */
738 			do
739 			{
740 				sm_io_clearerr(fp);
741 				errno = 0;
742 				c = sm_io_getc(fp, SM_TIME_DEFAULT);
743 
744 				/* timeout? */
745 				if (c == SM_IO_EOF && errno == EAGAIN
746 				    && SMTPMODE)
747 				{
748 					/*
749 					**  Override e_message in
750 					**  usrerr() as this is the
751 					**  reason for failure that
752 					**  should be logged for
753 					**  undelivered recipients.
754 					*/
755 
756 					e->e_message = NULL;
757 					errno = 0;
758 					inputerr = true;
759 					goto readabort;
760 				}
761 			} while (c == SM_IO_EOF && errno == EINTR);
762 			if (c != SM_IO_EOF)
763 				(void) sm_io_ungetc(fp, SM_TIME_DEFAULT, c);
764 			if (c == ' ' || c == '\t')
765 			{
766 				/* yep -- defer this */
767 				continue;
768 			}
769 
770 			SM_ASSERT(bp > buf);
771 
772 			/* guaranteed by isheader(buf) */
773 			SM_ASSERT(*(bp - 1) != '\n' || bp > buf + 1);
774 
775 			/* trim off trailing CRLF or LF */
776 			if (*--bp != '\n' || *--bp != '\r')
777 				bp++;
778 			*bp = '\0';
779 
780 			if (bitset(H_EOH, chompheader(buf,
781 						      CHHDR_CHECK | CHHDR_USER,
782 						      hdrp, e)))
783 			{
784 				mstate = MS_BODY;
785 				goto nextstate;
786 			}
787 			numhdrs++;
788 			break;
789 
790 		  case MS_BODY:
791 			if (tTd(30, 1))
792 				sm_dprintf("EOH\n");
793 
794 			if (headeronly)
795 				goto readdone;
796 
797 			df = collect_eoh(e, numhdrs, hdrslen);
798 			if (df == NULL)
799 				e->e_flags |= EF_TOOBIG;
800 
801 			bp = buf;
802 
803 			/* toss blank line */
804 			if ((bp[0] == '\r' && bp[1] == '\n') ||
805 			    (bp[0] == '\n'))
806 			{
807 				break;
808 			}
809 
810 			/* if not a blank separator, write it out */
811 			if (!bitset(EF_TOOBIG, e->e_flags))
812 			{
813 				while (*bp != '\0')
814 					(void) sm_io_putc(df, SM_TIME_DEFAULT,
815 							  *bp++);
816 			}
817 			break;
818 		}
819 		bp = buf;
820 	}
821 
822 readdone:
823 	if ((sm_io_eof(fp) && SMTPMODE) || sm_io_error(fp))
824 	{
825 		const char *errmsg;
826 
827 		if (sm_io_eof(fp))
828 			errmsg = "unexpected close";
829 		else
830 			errmsg = sm_errstring(errno);
831 		if (tTd(30, 1))
832 			sm_dprintf("collect: premature EOM: %s\n", errmsg);
833 		if (LogLevel > 1)
834 			sm_syslog(LOG_WARNING, e->e_id,
835 				"collect: premature EOM: %s", errmsg);
836 		inputerr = true;
837 	}
838 
839 	if (headeronly)
840 		goto end;
841 
842 	if (mstate != MS_BODY)
843 	{
844 		/* no body or discard, so we never opened the data file */
845 		SM_ASSERT(df == NULL);
846 		df = collect_eoh(e, numhdrs, hdrslen);
847 	}
848 
849 	if (df == NULL)
850 	{
851 		/* skip next few clauses */
852 		/* EMPTY */
853 	}
854 	else if (sm_io_flush(df, SM_TIME_DEFAULT) != 0 || sm_io_error(df))
855 	{
856 		dferror(df, "sm_io_flush||sm_io_error", e);
857 		flush_errors(true);
858 		finis(true, true, ExitStat);
859 		/* NOTREACHED */
860 	}
861 	else if (SuperSafe == SAFE_NO ||
862 		 SuperSafe == SAFE_INTERACTIVE ||
863 		 (SuperSafe == SAFE_REALLY_POSTMILTER && SMTPMODE))
864 	{
865 		/* skip next few clauses */
866 		/* EMPTY */
867 		/* Note: updfs() is not called in this case! */
868 	}
869 	else if (sm_io_setinfo(df, SM_BF_COMMIT, NULL) < 0 && errno != EINVAL)
870 	{
871 		int save_errno = errno;
872 
873 		if (save_errno == EEXIST)
874 		{
875 			char *dfile;
876 			struct stat st;
877 			int dfd;
878 
879 			dfile = queuename(e, DATAFL_LETTER);
880 			if (stat(dfile, &st) < 0)
881 				st.st_size = -1;
882 			errno = EEXIST;
883 			syserr("@collect: bfcommit(%s): already on disk, size=%ld",
884 			       dfile, (long) st.st_size);
885 			dfd = sm_io_getinfo(df, SM_IO_WHAT_FD, NULL);
886 			if (dfd >= 0)
887 				dumpfd(dfd, true, true);
888 		}
889 		errno = save_errno;
890 		dferror(df, "bfcommit", e);
891 		flush_errors(true);
892 		finis(save_errno != EEXIST, true, ExitStat);
893 	}
894 	else if ((afd = sm_io_getinfo(df, SM_IO_WHAT_FD, NULL)) < 0)
895 	{
896 		dferror(df, "sm_io_getinfo", e);
897 		flush_errors(true);
898 		finis(true, true, ExitStat);
899 		/* NOTREACHED */
900 	}
901 	else if (fsync(afd) < 0)
902 	{
903 		dferror(df, "fsync", e);
904 		flush_errors(true);
905 		finis(true, true, ExitStat);
906 		/* NOTREACHED */
907 	}
908 	else if (sm_io_close(df, SM_TIME_DEFAULT) < 0)
909 	{
910 		dferror(df, "sm_io_close", e);
911 		flush_errors(true);
912 		finis(true, true, ExitStat);
913 		/* NOTREACHED */
914 	}
915 	else
916 	{
917 		/* everything is happily flushed to disk */
918 		df = NULL;
919 
920 		/* remove from available space in filesystem */
921 		updfs(e, 0, 1, "collect");
922 	}
923 
924 	/* An EOF when running SMTP is an error */
925   readabort:
926 	if (inputerr && (OpMode == MD_SMTP || OpMode == MD_DAEMON))
927 	{
928 		char *problem;
929 		ADDRESS *q;
930 
931 		if (sm_io_eof(fp))
932 			problem = "unexpected close";
933 		else if (sm_io_error(fp))
934 			problem = "I/O error";
935 		else if (0 != bare_lf)
936 			problem = BARE_LF_MSG;
937 		else if (0 != bare_cr)
938 			problem = BARE_CR_MSG;
939 		else
940 			problem = "read timeout";
941 
942 #define LOG_CLT ((NULL != RealHostName) ? RealHostName: "localhost")
943 #define CONN_ERR_TXT	"collect: relay=%s, from=%s, info=%s%s%s%s"
944 #define CONN_ERR_CODE	"421 4.4.1 "
945 #define CONN_LOG_FROM	shortenstring(e->e_from.q_paddr, MAXSHORTSTR)
946 #define CONN_ERR_BARE (0 != bare_lf) ? BARE_LF_MSG : ((0 != bare_cr) ? BARE_CR_MSG : "")
947 #define CONN_ERR_WHERE(bare_xy) (BARE_IN_HDR==(bare_xy) ? "header" : \
948 	(BARE_IN_BDY==(bare_xy) ? "body" : "header+body"))
949 
950 #define HAS_BARE_XY (0 != (bare_lf | bare_cr))
951 #define CONN_ERR_ARGS LOG_CLT, CONN_LOG_FROM, problem, \
952 	HAS_BARE_XY ? ", where=" : "", \
953 	HAS_BARE_XY ? CONN_ERR_WHERE(bare_lf|bare_cr) : "", \
954 	HAS_BARE_XY ? ", status=tempfail" : ""
955 
956 		if (LogLevel > 0 && (sm_io_eof(fp) || (0 != (bare_lf | bare_cr))))
957 			sm_syslog(LOG_NOTICE, e->e_id,
958 				CONN_ERR_TXT, CONN_ERR_ARGS);
959 		if (0 != (bare_lf | bare_cr))
960 			usrerr("421 4.5.0 %s", CONN_ERR_BARE);
961 		else if (sm_io_eof(fp))
962 			usrerr(CONN_ERR_CODE CONN_ERR_TXT, CONN_ERR_ARGS);
963 		else
964 			syserr(CONN_ERR_CODE CONN_ERR_TXT, CONN_ERR_ARGS);
965 		flush_errors(true);
966 
967 		/* don't return an error indication */
968 		e->e_to = NULL;
969 		e->e_flags &= ~EF_FATALERRS;
970 		e->e_flags |= EF_CLRQUEUE;
971 
972 		/* Don't send any message notification to sender */
973 		for (q = e->e_sendqueue; q != NULL; q = q->q_next)
974 		{
975 			if (QS_IS_DEAD(q->q_state))
976 				continue;
977 			q->q_state = QS_FATALERR;
978 		}
979 
980 		SM_CLOSE_FP(df);
981 		finis(true, true, ExitStat);
982 		/* NOTREACHED */
983 	}
984 
985 	/* Log collection information. */
986 	if (tTd(92, 2))
987 		sm_dprintf("collect: e_id=%s, EF_LOGSENDER=%d, LogLevel=%d\n",
988 			e->e_id, bitset(EF_LOGSENDER, e->e_flags), LogLevel);
989 	if (bitset(EF_LOGSENDER, e->e_flags) && LogLevel > 4)
990 	{
991 		logsender(e, e->e_msgid);
992 		e->e_flags &= ~EF_LOGSENDER;
993 	}
994 
995 #define LOG_BARE_XY(bare_xy, bare_xy_sp, bare_xy_msg)	\
996 	do	\
997 	{	\
998 		if ((0 != bare_xy) && LogLevel > 8)	\
999 			sm_syslog(LOG_NOTICE, e->e_id, \
1000 				"collect: relay=%s, from=%s, info=%s, where=%s%s" \
1001 				, LOG_CLT, CONN_LOG_FROM, bare_xy_msg	\
1002 				, CONN_ERR_WHERE(bare_xy)	\
1003 				, bare_xy_sp ? ", status=replaced" : ""	\
1004 				);	\
1005 	} while (0)
1006 
1007 	LOG_BARE_XY(bare_lf, BARE_LF_SP, BARE_LF_MSG);
1008 	LOG_BARE_XY(bare_cr, BARE_CR_SP, BARE_CR_MSG);
1009 
1010 	/* check for message too large */
1011 	if (bitset(EF_TOOBIG, e->e_flags))
1012 	{
1013 		e->e_flags |= EF_NO_BODY_RETN|EF_CLRQUEUE;
1014 		if (!bitset(EF_FATALERRS, e->e_flags))
1015 		{
1016 			e->e_status = "5.2.3";
1017 			usrerrenh(e->e_status,
1018 				"552 Message exceeds maximum fixed size (%ld)",
1019 				MaxMessageSize);
1020 			if (LogLevel > 6)
1021 				sm_syslog(LOG_NOTICE, e->e_id,
1022 					"message size (%ld) exceeds maximum (%ld)",
1023 					PRT_NONNEGL(e->e_msgsize),
1024 					MaxMessageSize);
1025 		}
1026 	}
1027 
1028 	/* check for illegal 8-bit data */
1029 	if (HasEightBits)
1030 	{
1031 		e->e_flags |= EF_HAS8BIT;
1032 		if (!bitset(MM_PASS8BIT|MM_MIME8BIT, MimeMode) &&
1033 		    !bitset(EF_IS_MIME, e->e_flags))
1034 		{
1035 			e->e_status = "5.6.1";
1036 			usrerrenh(e->e_status, "554 Eight bit data not allowed");
1037 		}
1038 	}
1039 	else
1040 	{
1041 		/* if it claimed to be 8 bits, well, it lied.... */
1042 		if (e->e_bodytype != NULL &&
1043 		    SM_STRCASEEQ(e->e_bodytype, "8bitmime"))
1044 			e->e_bodytype = "7BIT";
1045 	}
1046 
1047 #if _FFR_REJECT_NUL_BYTE
1048 	if (hasNUL && RejectNUL)
1049 	{
1050 		e->e_status = "5.6.1";
1051 		usrerrenh(e->e_status, "554 NUL byte not allowed");
1052 	}
1053 #endif /* _FFR_REJECT_NUL_BYTE */
1054 
1055 	if (SuperSafe == SAFE_REALLY && !bitset(EF_FATALERRS, e->e_flags))
1056 	{
1057 		char *dfname = queuename(e, DATAFL_LETTER);
1058 		if ((e->e_dfp = sm_io_open(SmFtStdio, SM_TIME_DEFAULT, dfname,
1059 					   SM_IO_RDONLY_B, NULL)) == NULL)
1060 		{
1061 			/* we haven't acked receipt yet, so just chuck this */
1062 			syserr("@Cannot reopen %s", dfname);
1063 			finis(true, true, ExitStat);
1064 			/* NOTREACHED */
1065 		}
1066 	}
1067 	else
1068 		e->e_dfp = df;
1069 
1070 	/* collect statistics */
1071 	if (OpMode != MD_VERIFY)
1072 	{
1073 		/*
1074 		**  Recalculate e_msgpriority, it is done at in eatheader()
1075 		**  which is called (in 8.12) after the header is collected,
1076 		**  hence e_msgsize is (most likely) incorrect.
1077 		*/
1078 
1079 		e->e_msgpriority = e->e_msgsize
1080 				 - e->e_class * WkClassFact
1081 				 + e->e_nrcpts * WkRecipFact;
1082 		markstats(e, (ADDRESS *) NULL, STATS_NORMAL);
1083 	}
1084 
1085   end:
1086 	(void) set_tls_rd_tmo(old_rd_tmo);
1087 	if (buf != bufbuf)
1088 		SM_FREE(buf);
1089 }
1090 
1091 /*
1092 **  DFERROR -- signal error on writing the data file.
1093 **
1094 **	Called by collect().  collect() always terminates the process
1095 **	immediately after calling dferror(), which means that the SMTP
1096 **	session will be terminated, which means that any error message
1097 **	issued by dferror must be a 421 error, as per RFC 821.
1098 **
1099 **	Parameters:
1100 **		df -- the file pointer for the data file.
1101 **		msg -- detailed message.
1102 **		e -- the current envelope.
1103 **
1104 **	Returns:
1105 **		none.
1106 **
1107 **	Side Effects:
1108 **		Gives an error message.
1109 **		Arranges for following output to go elsewhere.
1110 */
1111 
1112 void
dferror(df,msg,e)1113 dferror(df, msg, e)
1114 	SM_FILE_T *volatile df;
1115 	char *msg;
1116 	register ENVELOPE *e;
1117 {
1118 	char *dfname;
1119 
1120 	dfname = queuename(e, DATAFL_LETTER);
1121 	setstat(EX_IOERR);
1122 	if (errno == ENOSPC)
1123 	{
1124 #if STAT64 > 0
1125 		struct stat64 st;
1126 #else
1127 		struct stat st;
1128 #endif
1129 		long avail;
1130 		long bsize;
1131 
1132 		e->e_flags |= EF_NO_BODY_RETN;
1133 
1134 		if (
1135 #if STAT64 > 0
1136 		    fstat64(sm_io_getinfo(df, SM_IO_WHAT_FD, NULL), &st)
1137 #else
1138 		    fstat(sm_io_getinfo(df, SM_IO_WHAT_FD, NULL), &st)
1139 #endif
1140 		    < 0)
1141 		  st.st_size = 0;
1142 		(void) sm_io_reopen(SmFtStdio, SM_TIME_DEFAULT, dfname,
1143 				    SM_IO_WRONLY_B, NULL, df);
1144 		if (st.st_size <= 0)
1145 			(void) sm_io_fprintf(df, SM_TIME_DEFAULT,
1146 				"\n*** Mail could not be accepted");
1147 		else
1148 			(void) sm_io_fprintf(df, SM_TIME_DEFAULT,
1149 				"\n*** Mail of at least %llu bytes could not be accepted\n",
1150 				(ULONGLONG_T) st.st_size);
1151 		(void) sm_io_fprintf(df, SM_TIME_DEFAULT,
1152 			"*** at %s due to lack of disk space for temp file.\n",
1153 			MyHostName);
1154 		avail = freediskspace(qid_printqueue(e->e_qgrp, e->e_qdir),
1155 				      &bsize);
1156 		if (avail > 0)
1157 		{
1158 			if (bsize > 1024)
1159 				avail *= bsize / 1024;
1160 			else if (bsize < 1024)
1161 				avail /= 1024 / bsize;
1162 			(void) sm_io_fprintf(df, SM_TIME_DEFAULT,
1163 				"*** Currently, %ld kilobytes are available for mail temp files.\n",
1164 				avail);
1165 		}
1166 #if 0
1167 		/* Wrong response code; should be 421. */
1168 		e->e_status = "4.3.1";
1169 		usrerrenh(e->e_status, "452 Out of disk space for temp file");
1170 #else /* 0 */
1171 		syserr("421 4.3.1 Out of disk space for temp file");
1172 #endif /* 0 */
1173 	}
1174 	else
1175 		syserr("421 4.3.0 collect: Cannot write %s (%s, uid=%ld, gid=%ld)",
1176 			dfname, msg, (long) geteuid(), (long) getegid());
1177 	if (sm_io_reopen(SmFtStdio, SM_TIME_DEFAULT, SM_PATH_DEVNULL,
1178 			 SM_IO_WRONLY, NULL, df) == NULL)
1179 		sm_syslog(LOG_ERR, e->e_id,
1180 			  "dferror: sm_io_reopen(\"/dev/null\") failed: %s",
1181 			  sm_errstring(errno));
1182 }
1183 /*
1184 **  EATFROM -- chew up a UNIX style from line and process
1185 **
1186 **	This does indeed make some assumptions about the format
1187 **	of UNIX messages.
1188 **
1189 **	Parameters:
1190 **		fm -- the from line.
1191 **		e -- envelope
1192 **
1193 **	Returns:
1194 **		none.
1195 **
1196 **	Side Effects:
1197 **		extracts what information it can from the header,
1198 **		such as the date.
1199 */
1200 
1201 #ifndef NOTUNIX
1202 
1203 static char	*DowList[] =
1204 {
1205 	"Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat", NULL
1206 };
1207 
1208 static char	*MonthList[] =
1209 {
1210 	"Jan", "Feb", "Mar", "Apr", "May", "Jun",
1211 	"Jul", "Aug", "Sep", "Oct", "Nov", "Dec",
1212 	NULL
1213 };
1214 
1215 static void
eatfrom(fm,e)1216 eatfrom(fm, e)
1217 	char *volatile fm;
1218 	register ENVELOPE *e;
1219 {
1220 	register char *p;
1221 	register char **dt;
1222 
1223 	if (tTd(30, 2))
1224 		sm_dprintf("eatfrom(%s)\n", fm);
1225 
1226 	/* find the date part */
1227 	p = fm;
1228 	while (*p != '\0')
1229 	{
1230 		/* skip a word */
1231 		while (*p != '\0' && *p != ' ')
1232 			p++;
1233 		while (*p == ' ')
1234 			p++;
1235 		if (strlen(p) < 17)
1236 		{
1237 			/* no room for the date */
1238 			return;
1239 		}
1240 		if (!(isascii(*p) && isupper(*p)) ||
1241 		    p[3] != ' ' || p[13] != ':' || p[16] != ':')
1242 			continue;
1243 
1244 		/* we have a possible date */
1245 		for (dt = DowList; *dt != NULL; dt++)
1246 			if (strncmp(*dt, p, 3) == 0)
1247 				break;
1248 		if (*dt == NULL)
1249 			continue;
1250 
1251 		for (dt = MonthList; *dt != NULL; dt++)
1252 		{
1253 			if (strncmp(*dt, &p[4], 3) == 0)
1254 				break;
1255 		}
1256 		if (*dt != NULL)
1257 			break;
1258 	}
1259 
1260 	if (*p != '\0')
1261 	{
1262 		char *q, buf[25];
1263 
1264 		/* we have found a date */
1265 		(void) sm_strlcpy(buf, p, sizeof(buf));
1266 		q = arpadate(buf);
1267 		macdefine(&e->e_macro, A_TEMP, 'a', q);
1268 	}
1269 }
1270 #endif /* ! NOTUNIX */
1271