2 * Copyright (c) 1998-2006, 2008 Proofpoint, Inc. and its suppliers.
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.
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.
16 SM_RCSID("@(#)$Id: collect.c,v 8.287 2013-11-22 20:51:55 ca Exp $")
18 static void eatfrom __P((char *volatile, ENVELOPE *));
19 static void collect_doheader __P((ENVELOPE *));
20 static SM_FILE_T *collect_dfopen __P((ENVELOPE *));
21 static SM_FILE_T *collect_eoh __P((ENVELOPE *, int, int));
24 ** COLLECT_EOH -- end-of-header processing in collect()
26 ** Called by collect() when it encounters the blank line
27 ** separating the header from the message body, or when it
28 ** encounters EOF in a message that contains only a header.
32 ** numhdrs -- number of headers
33 ** hdrslen -- length of headers
36 ** NULL, or handle to open data file
39 ** end-of-header check ruleset is invoked.
40 ** envelope state is updated.
41 ** headers may be added and deleted.
43 ** opens the data file.
47 collect_eoh(e, numhdrs, hdrslen)
55 /* call the end-of-header check ruleset */
56 (void) sm_snprintf(hnum, sizeof(hnum), "%d", numhdrs);
57 (void) sm_snprintf(hsize, sizeof(hsize), "%d", hdrslen);
59 sm_dprintf("collect: rscheck(\"check_eoh\", \"%s $| %s\")\n",
61 (void) rscheck("check_eoh", hnum, hsize, e, RSF_UNSTRUCTURED|RSF_COUNT,
62 3, NULL, e->e_id, NULL, NULL);
65 ** Process the header,
66 ** select the queue, open the data file.
70 return collect_dfopen(e);
74 ** COLLECT_DOHEADER -- process header in collect()
76 ** Called by collect() after it has finished parsing the header,
77 ** but before it selects the queue and creates the data file.
78 ** The results of processing the header will affect queue selection.
87 ** envelope state is updated.
88 ** headers may be added and deleted.
96 ** Find out some information from the headers.
97 ** Examples are who is the from person & the date.
100 eatheader(e, true, false);
102 if (GrabTo && e->e_sendqueue == NULL)
103 usrerr("No recipient addresses found in header");
106 ** If we have a Return-Receipt-To:, turn it into a DSN.
109 if (RrtImpliesDsn && hvalue("return-receipt-to", e->e_header) != NULL)
113 for (q = e->e_sendqueue; q != NULL; q = q->q_next)
114 if (!bitset(QHASNOTIFY, q->q_flags))
115 q->q_flags |= QHASNOTIFY|QPINGONSUCCESS;
119 ** Add an appropriate recipient line if we have none.
122 if (hvalue("to", e->e_header) != NULL ||
123 hvalue("cc", e->e_header) != NULL ||
124 hvalue("apparently-to", e->e_header) != NULL)
126 /* have a valid recipient header -- delete Bcc: headers */
127 e->e_flags |= EF_DELETE_BCC;
129 else if (hvalue("bcc", e->e_header) == NULL)
131 /* no valid recipient headers */
135 /* create a recipient field */
136 switch (NoRecipientAction)
138 case NRA_ADD_APPARENTLY_TO:
139 hdr = "Apparently-To";
147 addheader("Bcc", " ", 0, e, true);
150 case NRA_ADD_TO_UNDISCLOSED:
151 addheader("To", "undisclosed-recipients:;", 0, e, true);
157 for (q = e->e_sendqueue; q != NULL; q = q->q_next)
159 if (q->q_alias != NULL)
162 sm_dprintf("Adding %s: %s\n",
164 addheader(hdr, q->q_paddr, 0, e, true);
171 ** COLLECT_DFOPEN -- open the message data file
173 ** Called by collect() after it has finished processing the header.
174 ** Queue selection occurs at this point, possibly based on the
175 ** envelope's recipient list and on header information.
181 ** NULL, or a pointer to an open data file,
182 ** into which the message body will be written by collect().
185 ** Calls syserr, sets EF_FATALERRS and returns NULL
186 ** if there is insufficient disk space.
187 ** Aborts process if data file could not be opened.
188 ** Otherwise, the queue is selected,
189 ** e->e_{dfino,dfdev,msgsize,flags} are updated,
190 ** and a pointer to an open data file is returned.
206 dfname = queuename(e, DATAFL_LETTER);
207 if (bitset(S_IWGRP, QueueFileMode))
208 oldumask = umask(002);
209 df = bfopen(dfname, QueueFileMode, DataFileBufferSize,
211 if (bitset(S_IWGRP, QueueFileMode))
212 (void) umask(oldumask);
215 syserr("@Cannot create %s", dfname);
216 e->e_flags |= EF_NO_BODY_RETN;
218 finis(false, true, ExitStat);
221 dfd = sm_io_getinfo(df, SM_IO_WHAT_FD, NULL);
222 if (dfd < 0 || fstat(dfd, &stbuf) < 0)
226 e->e_dfdev = stbuf.st_dev;
227 e->e_dfino = stbuf.st_ino;
229 e->e_flags |= EF_HAS_DF;
234 ** COLLECT -- read & parse message header & make temp file.
236 ** Creates a temporary file name and copies the standard
237 ** input to that file. Leading UNIX-style "From" lines are
238 ** stripped off (after important information is extracted).
241 ** fp -- file to read.
242 ** smtpmode -- if set, we are running SMTP: give an RFC821
243 ** style message to say we are ready to collect
244 ** input, and never ignore a single dot to mean
246 ** hdrp -- the location to stash the header.
247 ** e -- the current envelope.
248 ** rsetsize -- reset e_msgsize?
255 ** - Data file is created and filled, and e->e_dfp is set.
256 ** - The from person may be set.
257 ** If the "enough disk space" check fails,
258 ** - syserr is called.
259 ** - e->e_dfp is NULL.
260 ** - e->e_flags & EF_FATALERRS is set.
261 ** - collect() returns.
262 ** If data file cannot be created, the process is terminated.
265 /* values for input state machine */
266 #define IS_NORM 0 /* middle of line */
267 #define IS_BOL 1 /* beginning of line */
268 #define IS_DOT 2 /* read a dot at beginning of line */
269 #define IS_DOTCR 3 /* read ".\r" at beginning of line */
270 #define IS_CR 4 /* read a carriage return */
272 /* values for message state machine */
273 #define MS_UFROM 0 /* reading Unix from line */
274 #define MS_HEADER 1 /* reading message header */
275 #define MS_BODY 2 /* reading message body */
276 #define MS_DISCARD 3 /* discarding rest of message */
279 collect(fp, smtpmode, hdrp, e, rsetsize)
283 register ENVELOPE *e;
286 register SM_FILE_T *df;
302 unsigned char peekbuf[8];
303 char bufbuf[MAXLINE];
304 #if _FFR_REJECT_NUL_BYTE
305 bool hasNUL; /* has at least one NUL input byte */
309 ignrdot = smtpmode ? false : IgnrDot;
311 /* timeout for I/O functions is in milliseconds */
312 dbto = smtpmode ? ((int) TimeOuts.to_datablock * 1000)
314 sm_io_setinfo(fp, SM_IO_WHAT_TIMEOUT, &dbto);
315 old_rd_tmo = set_tls_rd_tmo(TimeOuts.to_datablock);
318 headeronly = hdrp != NULL;
321 HasEightBits = false;
322 #if _FFR_REJECT_NUL_BYTE
326 buflen = sizeof(bufbuf);
329 mstate = SaveFrom ? MS_HEADER : MS_UFROM;
332 ** Tell ARPANET to go ahead.
336 message("354 Enter mail, end with \".\" on a line by itself");
338 /* simulate an I/O timeout when used as sink */
343 sm_dprintf("collect\n");
348 ** This is done using two interleaved state machines.
349 ** The input state machine is looking for things like
350 ** hidden dots; the message state machine is handling
351 ** the larger picture (e.g., header versus body).
359 sm_dprintf("top, istate=%d, mstate=%d\n", istate,
367 while (!sm_io_eof(fp) && !sm_io_error(fp))
370 c = sm_io_getc(fp, SM_TIME_DEFAULT);
371 if (c == SM_IO_EOF && errno == EINTR)
373 /* Interrupted, retry */
379 if (c == SM_IO_EOF && errno == EAGAIN
383 ** Override e_message in
384 ** usrerr() as this is the
385 ** reason for failure that
386 ** should be logged for
387 ** undelivered recipients.
397 if (TrafficLogFile != NULL && !headeronly)
399 if (istate == IS_BOL)
400 (void) sm_io_fprintf(TrafficLogFile,
405 (void) sm_io_fprintf(TrafficLogFile,
409 (void) sm_io_putc(TrafficLogFile,
413 #if _FFR_REJECT_NUL_BYTE
422 HasEightBits |= bitset(0x80, c);
425 sm_dprintf("istate=%d, c=%c (0x%x)\n",
426 istate, (char) c, c);
438 if (c == '\n' && !ignrdot)
448 OpMode != MD_DAEMON &&
449 OpMode != MD_ARPAFTP))
452 SM_ASSERT(pbp < peekbuf +
460 if (c == '\n' && !ignrdot)
464 /* push back the ".\rx" */
465 SM_ASSERT(pbp < peekbuf +
468 if (OpMode != MD_SMTP &&
469 OpMode != MD_DAEMON &&
470 OpMode != MD_ARPAFTP)
472 SM_ASSERT(pbp < peekbuf +
487 (void) sm_io_ungetc(fp, SM_TIME_DEFAULT,
509 if (e->e_msgsize >= 0)
512 if (MaxMessageSize > 0 &&
513 !bitset(EF_TOOBIG, e->e_flags) &&
514 e->e_msgsize > MaxMessageSize)
515 e->e_flags |= EF_TOOBIG;
521 /* just put the character out */
522 if (!bitset(EF_TOOBIG, e->e_flags))
523 (void) sm_io_putc(df, SM_TIME_DEFAULT,
532 SM_ASSERT(mstate == MS_UFROM || mstate == MS_HEADER);
534 /* header -- buffer up */
535 if (bp >= &buf[buflen - 2])
539 /* out of space for header */
541 if (buflen < MEMCHUNKSIZE)
544 buflen += MEMCHUNKSIZE;
547 sm_syslog(LOG_NOTICE, e->e_id,
548 "header overflow from %s during message collect",
551 e->e_flags |= EF_CLRQUEUE;
552 e->e_status = "5.6.0";
553 usrerrenh(e->e_status,
554 "552 Headers too large");
557 buf = xalloc(buflen);
558 memmove(buf, obuf, bp - obuf);
559 bp = &buf[bp - obuf];
561 sm_free(obuf); /* XXX */
569 MaxHeadersLength > 0 &&
570 hdrslen > MaxHeadersLength)
572 sm_syslog(LOG_NOTICE, e->e_id,
573 "headers too large (%d max) from %s during message collect",
577 e->e_flags |= EF_CLRQUEUE;
578 e->e_status = "5.6.0";
579 usrerrenh(e->e_status,
580 "552 Headers too large (%d max)",
586 if (istate == IS_BOL)
593 sm_dprintf("nextstate, istate=%d, mstate=%d, line=\"%s\"\n",
594 istate, mstate, buf);
600 if (strncmp(buf, "From ", 5) == 0)
606 #endif /* ! NOTUNIX */
616 /* check for possible continuation line */
621 c = sm_io_getc(fp, SM_TIME_DEFAULT);
624 if (c == SM_IO_EOF && errno == EAGAIN
628 ** Override e_message in
629 ** usrerr() as this is the
630 ** reason for failure that
631 ** should be logged for
632 ** undelivered recipients.
640 } while (c == SM_IO_EOF && errno == EINTR);
642 (void) sm_io_ungetc(fp, SM_TIME_DEFAULT, c);
643 if (c == ' ' || c == '\t')
645 /* yep -- defer this */
651 /* guaranteed by isheader(buf) */
652 SM_ASSERT(*(bp - 1) != '\n' || bp > buf + 1);
654 /* trim off trailing CRLF or NL */
655 if (*--bp != '\n' || *--bp != '\r')
659 if (bitset(H_EOH, chompheader(buf,
660 CHHDR_CHECK | CHHDR_USER,
676 df = collect_eoh(e, numhdrs, hdrslen);
678 e->e_flags |= EF_TOOBIG;
682 /* toss blank line */
683 if ((bp[0] == '\r' && bp[1] == '\n') ||
689 /* if not a blank separator, write it out */
690 if (!bitset(EF_TOOBIG, e->e_flags))
693 (void) sm_io_putc(df, SM_TIME_DEFAULT,
702 if ((sm_io_eof(fp) && smtpmode) || sm_io_error(fp))
707 errmsg = "unexpected close";
709 errmsg = sm_errstring(errno);
711 sm_dprintf("collect: premature EOM: %s\n", errmsg);
713 sm_syslog(LOG_WARNING, e->e_id,
714 "collect: premature EOM: %s", errmsg);
721 if (mstate != MS_BODY)
723 /* no body or discard, so we never opened the data file */
724 SM_ASSERT(df == NULL);
725 df = collect_eoh(e, numhdrs, hdrslen);
730 /* skip next few clauses */
733 else if (sm_io_flush(df, SM_TIME_DEFAULT) != 0 || sm_io_error(df))
735 dferror(df, "sm_io_flush||sm_io_error", e);
737 finis(true, true, ExitStat);
740 else if (SuperSafe == SAFE_NO ||
741 SuperSafe == SAFE_INTERACTIVE ||
742 (SuperSafe == SAFE_REALLY_POSTMILTER && smtpmode))
744 /* skip next few clauses */
746 /* Note: updfs() is not called in this case! */
748 else if (sm_io_setinfo(df, SM_BF_COMMIT, NULL) < 0 && errno != EINVAL)
750 int save_errno = errno;
752 if (save_errno == EEXIST)
758 dfile = queuename(e, DATAFL_LETTER);
759 if (stat(dfile, &st) < 0)
762 syserr("@collect: bfcommit(%s): already on disk, size=%ld",
763 dfile, (long) st.st_size);
764 dfd = sm_io_getinfo(df, SM_IO_WHAT_FD, NULL);
766 dumpfd(dfd, true, true);
769 dferror(df, "bfcommit", e);
771 finis(save_errno != EEXIST, true, ExitStat);
773 else if ((afd = sm_io_getinfo(df, SM_IO_WHAT_FD, NULL)) < 0)
775 dferror(df, "sm_io_getinfo", e);
777 finis(true, true, ExitStat);
780 else if (fsync(afd) < 0)
782 dferror(df, "fsync", e);
784 finis(true, true, ExitStat);
787 else if (sm_io_close(df, SM_TIME_DEFAULT) < 0)
789 dferror(df, "sm_io_close", e);
791 finis(true, true, ExitStat);
796 /* everything is happily flushed to disk */
799 /* remove from available space in filesystem */
800 updfs(e, 0, 1, "collect");
803 /* An EOF when running SMTP is an error */
805 if (inputerr && (OpMode == MD_SMTP || OpMode == MD_DAEMON))
816 problem = "unexpected close";
817 else if (sm_io_error(fp))
818 problem = "I/O error";
820 problem = "read timeout";
821 if (LogLevel > 0 && sm_io_eof(fp))
822 sm_syslog(LOG_NOTICE, e->e_id,
823 "collect: %s on connection from %.100s, sender=%s",
825 shortenstring(e->e_from.q_paddr, MAXSHORTSTR));
827 usrerr("421 4.4.1 collect: %s on connection from %s, from=%s",
829 shortenstring(e->e_from.q_paddr, MAXSHORTSTR));
831 syserr("421 4.4.1 collect: %s on connection from %s, from=%s",
833 shortenstring(e->e_from.q_paddr, MAXSHORTSTR));
836 /* don't return an error indication */
838 e->e_flags &= ~EF_FATALERRS;
839 e->e_flags |= EF_CLRQUEUE;
841 /* Don't send any message notification to sender */
842 for (q = e->e_sendqueue; q != NULL; q = q->q_next)
844 if (QS_IS_DEAD(q->q_state))
846 q->q_state = QS_FATALERR;
849 (void) sm_io_close(df, SM_TIME_DEFAULT);
851 finis(true, true, ExitStat);
855 /* Log collection information. */
857 sm_dprintf("collect: e_id=%s, EF_LOGSENDER=%d, LogLevel=%d\n",
858 e->e_id, bitset(EF_LOGSENDER, e->e_flags), LogLevel);
859 if (bitset(EF_LOGSENDER, e->e_flags) && LogLevel > 4)
861 logsender(e, e->e_msgid);
862 e->e_flags &= ~EF_LOGSENDER;
865 /* check for message too large */
866 if (bitset(EF_TOOBIG, e->e_flags))
868 e->e_flags |= EF_NO_BODY_RETN|EF_CLRQUEUE;
869 if (!bitset(EF_FATALERRS, e->e_flags))
871 e->e_status = "5.2.3";
872 usrerrenh(e->e_status,
873 "552 Message exceeds maximum fixed size (%ld)",
876 sm_syslog(LOG_NOTICE, e->e_id,
877 "message size (%ld) exceeds maximum (%ld)",
878 PRT_NONNEGL(e->e_msgsize),
883 /* check for illegal 8-bit data */
886 e->e_flags |= EF_HAS8BIT;
887 if (!bitset(MM_PASS8BIT|MM_MIME8BIT, MimeMode) &&
888 !bitset(EF_IS_MIME, e->e_flags))
890 e->e_status = "5.6.1";
891 usrerrenh(e->e_status, "554 Eight bit data not allowed");
896 /* if it claimed to be 8 bits, well, it lied.... */
897 if (e->e_bodytype != NULL &&
898 sm_strcasecmp(e->e_bodytype, "8BITMIME") == 0)
899 e->e_bodytype = "7BIT";
902 #if _FFR_REJECT_NUL_BYTE
903 if (hasNUL && RejectNUL)
905 e->e_status = "5.6.1";
906 usrerrenh(e->e_status, "554 NUL byte not allowed");
908 #endif /* _FFR_REJECT_NUL_BYTE */
910 if (SuperSafe == SAFE_REALLY && !bitset(EF_FATALERRS, e->e_flags))
912 char *dfname = queuename(e, DATAFL_LETTER);
913 if ((e->e_dfp = sm_io_open(SmFtStdio, SM_TIME_DEFAULT, dfname,
914 SM_IO_RDONLY_B, NULL)) == NULL)
916 /* we haven't acked receipt yet, so just chuck this */
917 syserr("@Cannot reopen %s", dfname);
918 finis(true, true, ExitStat);
925 /* collect statistics */
926 if (OpMode != MD_VERIFY)
929 ** Recalculate e_msgpriority, it is done at in eatheader()
930 ** which is called (in 8.12) after the header is collected,
931 ** hence e_msgsize is (most likely) incorrect.
934 e->e_msgpriority = e->e_msgsize
935 - e->e_class * WkClassFact
936 + e->e_nrcpts * WkRecipFact;
937 markstats(e, (ADDRESS *) NULL, STATS_NORMAL);
941 (void) set_tls_rd_tmo(old_rd_tmo);
945 ** DFERROR -- signal error on writing the data file.
947 ** Called by collect(). Collect() always terminates the process
948 ** immediately after calling dferror(), which means that the SMTP
949 ** session will be terminated, which means that any error message
950 ** issued by dferror must be a 421 error, as per RFC 821.
953 ** df -- the file pointer for the data file.
954 ** msg -- detailed message.
955 ** e -- the current envelope.
961 ** Gives an error message.
962 ** Arranges for following output to go elsewhere.
967 SM_FILE_T *volatile df;
969 register ENVELOPE *e;
973 dfname = queuename(e, DATAFL_LETTER);
985 e->e_flags |= EF_NO_BODY_RETN;
989 fstat64(sm_io_getinfo(df, SM_IO_WHAT_FD, NULL), &st)
991 fstat(sm_io_getinfo(df, SM_IO_WHAT_FD, NULL), &st)
995 (void) sm_io_reopen(SmFtStdio, SM_TIME_DEFAULT, dfname,
996 SM_IO_WRONLY_B, NULL, df);
998 (void) sm_io_fprintf(df, SM_TIME_DEFAULT,
999 "\n*** Mail could not be accepted");
1001 (void) sm_io_fprintf(df, SM_TIME_DEFAULT,
1002 "\n*** Mail of at least %llu bytes could not be accepted\n",
1003 (ULONGLONG_T) st.st_size);
1004 (void) sm_io_fprintf(df, SM_TIME_DEFAULT,
1005 "*** at %s due to lack of disk space for temp file.\n",
1007 avail = freediskspace(qid_printqueue(e->e_qgrp, e->e_qdir),
1012 avail *= bsize / 1024;
1013 else if (bsize < 1024)
1014 avail /= 1024 / bsize;
1015 (void) sm_io_fprintf(df, SM_TIME_DEFAULT,
1016 "*** Currently, %ld kilobytes are available for mail temp files.\n",
1020 /* Wrong response code; should be 421. */
1021 e->e_status = "4.3.1";
1022 usrerrenh(e->e_status, "452 Out of disk space for temp file");
1024 syserr("421 4.3.1 Out of disk space for temp file");
1028 syserr("421 4.3.0 collect: Cannot write %s (%s, uid=%ld, gid=%ld)",
1029 dfname, msg, (long) geteuid(), (long) getegid());
1030 if (sm_io_reopen(SmFtStdio, SM_TIME_DEFAULT, SM_PATH_DEVNULL,
1031 SM_IO_WRONLY, NULL, df) == NULL)
1032 sm_syslog(LOG_ERR, e->e_id,
1033 "dferror: sm_io_reopen(\"/dev/null\") failed: %s",
1034 sm_errstring(errno));
1037 ** EATFROM -- chew up a UNIX style from line and process
1039 ** This does indeed make some assumptions about the format
1040 ** of UNIX messages.
1043 ** fm -- the from line.
1050 ** extracts what information it can from the header,
1051 ** such as the date.
1056 static char *DowList[] =
1058 "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat", NULL
1061 static char *MonthList[] =
1063 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
1064 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec",
1071 register ENVELOPE *e;
1077 sm_dprintf("eatfrom(%s)\n", fm);
1079 /* find the date part */
1084 while (*p != '\0' && *p != ' ')
1090 /* no room for the date */
1093 if (!(isascii(*p) && isupper(*p)) ||
1094 p[3] != ' ' || p[13] != ':' || p[16] != ':')
1097 /* we have a possible date */
1098 for (dt = DowList; *dt != NULL; dt++)
1099 if (strncmp(*dt, p, 3) == 0)
1104 for (dt = MonthList; *dt != NULL; dt++)
1106 if (strncmp(*dt, &p[4], 3) == 0)
1117 /* we have found a date */
1118 (void) sm_strlcpy(buf, p, sizeof(buf));
1120 macdefine(&e->e_macro, A_TEMP, 'a', q);
1123 #endif /* ! NOTUNIX */