2 * Copyright (c) 1998-2006 Sendmail, 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 8.273 2006/03/31 18:51:47 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,
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);
150 case NRA_ADD_TO_UNDISCLOSED:
151 addheader("To", "undisclosed-recipients:;", 0, e);
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);
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;
301 unsigned char peekbuf[8];
302 char bufbuf[MAXLINE];
305 ignrdot = smtpmode ? false : IgnrDot;
307 /* timeout for I/O functions is in milliseconds */
308 dbto = smtpmode ? ((int) TimeOuts.to_datablock * 1000)
310 sm_io_setinfo(fp, SM_IO_WHAT_TIMEOUT, &dbto);
313 headeronly = hdrp != NULL;
316 HasEightBits = false;
318 buflen = sizeof bufbuf;
321 mstate = SaveFrom ? MS_HEADER : MS_UFROM;
324 ** Tell ARPANET to go ahead.
328 message("354 Enter mail, end with \".\" on a line by itself");
330 /* simulate an I/O timeout when used as sink */
335 sm_dprintf("collect\n");
340 ** This is done using two interleaved state machines.
341 ** The input state machine is looking for things like
342 ** hidden dots; the message state machine is handling
343 ** the larger picture (e.g., header versus body).
351 sm_dprintf("top, istate=%d, mstate=%d\n", istate,
359 while (!sm_io_eof(fp) && !sm_io_error(fp))
362 c = sm_io_getc(fp, SM_TIME_DEFAULT);
363 if (c == SM_IO_EOF && errno == EINTR)
365 /* Interrupted, retry */
371 if (c == SM_IO_EOF && errno == EAGAIN
375 ** Override e_message in
376 ** usrerr() as this is the
377 ** reason for failure that
378 ** should be logged for
379 ** undelivered recipients.
389 if (TrafficLogFile != NULL && !headeronly)
391 if (istate == IS_BOL)
392 (void) sm_io_fprintf(TrafficLogFile,
397 (void) sm_io_fprintf(TrafficLogFile,
401 (void) sm_io_putc(TrafficLogFile,
410 HasEightBits |= bitset(0x80, c);
413 sm_dprintf("istate=%d, c=%c (0x%x)\n",
414 istate, (char) c, c);
426 if (c == '\n' && !ignrdot &&
427 !bitset(EF_NL_NOT_EOL, e->e_flags))
429 else if (c == '\r' &&
430 !bitset(EF_CRLF_NOT_EOL, e->e_flags))
438 OpMode != MD_DAEMON &&
439 OpMode != MD_ARPAFTP))
442 SM_ASSERT(pbp < peekbuf +
450 if (c == '\n' && !ignrdot)
454 /* push back the ".\rx" */
455 SM_ASSERT(pbp < peekbuf +
458 if (OpMode != MD_SMTP &&
459 OpMode != MD_DAEMON &&
460 OpMode != MD_ARPAFTP)
462 SM_ASSERT(pbp < peekbuf +
477 (void) sm_io_ungetc(fp, SM_TIME_DEFAULT,
485 if (c == '\r' && !bitset(EF_CRLF_NOT_EOL, e->e_flags))
490 else if (c == '\n' && !bitset(EF_NL_NOT_EOL,
500 if (e->e_msgsize >= 0)
503 if (MaxMessageSize > 0 &&
504 !bitset(EF_TOOBIG, e->e_flags) &&
505 e->e_msgsize > MaxMessageSize)
506 e->e_flags |= EF_TOOBIG;
512 /* just put the character out */
513 if (!bitset(EF_TOOBIG, e->e_flags))
514 (void) sm_io_putc(df, SM_TIME_DEFAULT,
523 SM_ASSERT(mstate == MS_UFROM || mstate == MS_HEADER);
525 /* header -- buffer up */
526 if (bp >= &buf[buflen - 2])
530 /* out of space for header */
532 if (buflen < MEMCHUNKSIZE)
535 buflen += MEMCHUNKSIZE;
538 sm_syslog(LOG_NOTICE, e->e_id,
539 "header overflow from %s during message collect",
542 e->e_flags |= EF_CLRQUEUE;
543 e->e_status = "5.6.0";
544 usrerrenh(e->e_status,
545 "552 Headers too large");
548 buf = xalloc(buflen);
549 memmove(buf, obuf, bp - obuf);
550 bp = &buf[bp - obuf];
552 sm_free(obuf); /* XXX */
556 ** XXX Notice: the logic here is broken.
557 ** An input to sendmail that doesn't contain a
558 ** header but starts immediately with the body whose
559 ** first line contain characters which match the
560 ** following "if" will cause problems: those
561 ** characters will NOT appear in the output...
565 if (c >= 0200 && c <= 0237)
567 #if 0 /* causes complaints -- figure out something for 8.n+1 */
568 usrerr("Illegal character 0x%x in header", c);
578 MaxHeadersLength > 0 &&
579 hdrslen > MaxHeadersLength)
581 sm_syslog(LOG_NOTICE, e->e_id,
582 "headers too large (%d max) from %s during message collect",
586 e->e_flags |= EF_CLRQUEUE;
587 e->e_status = "5.6.0";
588 usrerrenh(e->e_status,
589 "552 Headers too large (%d max)",
595 if (istate == IS_BOL)
602 sm_dprintf("nextstate, istate=%d, mstate=%d, line = \"%s\"\n",
603 istate, mstate, buf);
609 if (strncmp(buf, "From ", 5) == 0)
615 #endif /* ! NOTUNIX */
625 /* check for possible continuation line */
630 c = sm_io_getc(fp, SM_TIME_DEFAULT);
633 if (c == SM_IO_EOF && errno == EAGAIN
637 ** Override e_message in
638 ** usrerr() as this is the
639 ** reason for failure that
640 ** should be logged for
641 ** undelivered recipients.
649 } while (c == SM_IO_EOF && errno == EINTR);
651 (void) sm_io_ungetc(fp, SM_TIME_DEFAULT, c);
652 if (c == ' ' || c == '\t')
654 /* yep -- defer this */
660 /* guaranteed by isheader(buf) */
661 SM_ASSERT(*(bp - 1) != '\n' || bp > buf + 1);
663 /* trim off trailing CRLF or NL */
664 if (*--bp != '\n' || *--bp != '\r')
668 if (bitset(H_EOH, chompheader(buf,
669 CHHDR_CHECK | CHHDR_USER,
685 df = collect_eoh(e, numhdrs, hdrslen);
687 e->e_flags |= EF_TOOBIG;
691 /* toss blank line */
692 if ((!bitset(EF_CRLF_NOT_EOL, e->e_flags) &&
693 bp[0] == '\r' && bp[1] == '\n') ||
694 (!bitset(EF_NL_NOT_EOL, e->e_flags) &&
700 /* if not a blank separator, write it out */
701 if (!bitset(EF_TOOBIG, e->e_flags))
704 (void) sm_io_putc(df, SM_TIME_DEFAULT,
713 if ((sm_io_eof(fp) && smtpmode) || sm_io_error(fp))
718 errmsg = "unexpected close";
720 errmsg = sm_errstring(errno);
722 sm_dprintf("collect: premature EOM: %s\n", errmsg);
724 sm_syslog(LOG_WARNING, e->e_id,
725 "collect: premature EOM: %s", errmsg);
732 if (mstate != MS_BODY)
734 /* no body or discard, so we never opened the data file */
735 SM_ASSERT(df == NULL);
736 df = collect_eoh(e, numhdrs, hdrslen);
741 /* skip next few clauses */
744 else if (sm_io_flush(df, SM_TIME_DEFAULT) != 0 || sm_io_error(df))
746 dferror(df, "sm_io_flush||sm_io_error", e);
748 finis(true, true, ExitStat);
751 else if (SuperSafe == SAFE_NO ||
752 SuperSafe == SAFE_INTERACTIVE ||
753 (SuperSafe == SAFE_REALLY_POSTMILTER && smtpmode))
755 /* skip next few clauses */
757 /* Note: updfs() is not called in this case! */
759 else if (sm_io_setinfo(df, SM_BF_COMMIT, NULL) < 0 && errno != EINVAL)
761 int save_errno = errno;
763 if (save_errno == EEXIST)
769 dfile = queuename(e, DATAFL_LETTER);
770 if (stat(dfile, &st) < 0)
773 syserr("@collect: bfcommit(%s): already on disk, size=%ld",
774 dfile, (long) st.st_size);
775 dfd = sm_io_getinfo(df, SM_IO_WHAT_FD, NULL);
777 dumpfd(dfd, true, true);
780 dferror(df, "bfcommit", e);
782 finis(save_errno != EEXIST, true, ExitStat);
784 else if ((afd = sm_io_getinfo(df, SM_IO_WHAT_FD, NULL)) < 0)
786 dferror(df, "sm_io_getinfo", e);
788 finis(true, true, ExitStat);
791 else if (fsync(afd) < 0)
793 dferror(df, "fsync", e);
795 finis(true, true, ExitStat);
798 else if (sm_io_close(df, SM_TIME_DEFAULT) < 0)
800 dferror(df, "sm_io_close", e);
802 finis(true, true, ExitStat);
807 /* everything is happily flushed to disk */
810 /* remove from available space in filesystem */
811 updfs(e, 0, 1, "collect");
814 /* An EOF when running SMTP is an error */
816 if (inputerr && (OpMode == MD_SMTP || OpMode == MD_DAEMON))
827 problem = "unexpected close";
828 else if (sm_io_error(fp))
829 problem = "I/O error";
831 problem = "read timeout";
832 if (LogLevel > 0 && sm_io_eof(fp))
833 sm_syslog(LOG_NOTICE, e->e_id,
834 "collect: %s on connection from %.100s, sender=%s",
836 shortenstring(e->e_from.q_paddr, MAXSHORTSTR));
838 usrerr("421 4.4.1 collect: %s on connection from %s, from=%s",
840 shortenstring(e->e_from.q_paddr, MAXSHORTSTR));
842 syserr("421 4.4.1 collect: %s on connection from %s, from=%s",
844 shortenstring(e->e_from.q_paddr, MAXSHORTSTR));
847 /* don't return an error indication */
849 e->e_flags &= ~EF_FATALERRS;
850 e->e_flags |= EF_CLRQUEUE;
852 /* Don't send any message notification to sender */
853 for (q = e->e_sendqueue; q != NULL; q = q->q_next)
855 if (QS_IS_DEAD(q->q_state))
857 q->q_state = QS_FATALERR;
860 (void) sm_io_close(df, SM_TIME_DEFAULT);
862 finis(true, true, ExitStat);
866 /* Log collection information. */
867 if (bitset(EF_LOGSENDER, e->e_flags) && LogLevel > 4)
869 logsender(e, e->e_msgid);
870 e->e_flags &= ~EF_LOGSENDER;
873 /* check for message too large */
874 if (bitset(EF_TOOBIG, e->e_flags))
876 e->e_flags |= EF_NO_BODY_RETN|EF_CLRQUEUE;
877 if (!bitset(EF_FATALERRS, e->e_flags))
879 e->e_status = "5.2.3";
880 usrerrenh(e->e_status,
881 "552 Message exceeds maximum fixed size (%ld)",
884 sm_syslog(LOG_NOTICE, e->e_id,
885 "message size (%ld) exceeds maximum (%ld)",
886 e->e_msgsize, MaxMessageSize);
890 /* check for illegal 8-bit data */
893 e->e_flags |= EF_HAS8BIT;
894 if (!bitset(MM_PASS8BIT|MM_MIME8BIT, MimeMode) &&
895 !bitset(EF_IS_MIME, e->e_flags))
897 e->e_status = "5.6.1";
898 usrerrenh(e->e_status, "554 Eight bit data not allowed");
903 /* if it claimed to be 8 bits, well, it lied.... */
904 if (e->e_bodytype != NULL &&
905 sm_strcasecmp(e->e_bodytype, "8BITMIME") == 0)
906 e->e_bodytype = "7BIT";
909 if (SuperSafe == SAFE_REALLY && !bitset(EF_FATALERRS, e->e_flags))
911 char *dfname = queuename(e, DATAFL_LETTER);
912 if ((e->e_dfp = sm_io_open(SmFtStdio, SM_TIME_DEFAULT, dfname,
913 SM_IO_RDONLY_B, NULL)) == NULL)
915 /* we haven't acked receipt yet, so just chuck this */
916 syserr("@Cannot reopen %s", dfname);
917 finis(true, true, ExitStat);
924 /* collect statistics */
925 if (OpMode != MD_VERIFY)
928 ** Recalculate e_msgpriority, it is done at in eatheader()
929 ** which is called (in 8.12) after the header is collected,
930 ** hence e_msgsize is (most likely) incorrect.
933 e->e_msgpriority = e->e_msgsize
934 - e->e_class * WkClassFact
935 + e->e_nrcpts * WkRecipFact;
936 markstats(e, (ADDRESS *) NULL, STATS_NORMAL);
941 ** DFERROR -- signal error on writing the data file.
943 ** Called by collect(). Collect() always terminates the process
944 ** immediately after calling dferror(), which means that the SMTP
945 ** session will be terminated, which means that any error message
946 ** issued by dferror must be a 421 error, as per RFC 821.
949 ** df -- the file pointer for the data file.
950 ** msg -- detailed message.
951 ** e -- the current envelope.
957 ** Gives an error message.
958 ** Arranges for following output to go elsewhere.
963 SM_FILE_T *volatile df;
965 register ENVELOPE *e;
969 dfname = queuename(e, DATAFL_LETTER);
975 #else /* STAT64 > 0 */
977 #endif /* STAT64 > 0 */
981 e->e_flags |= EF_NO_BODY_RETN;
985 fstat64(sm_io_getinfo(df, SM_IO_WHAT_FD, NULL), &st)
986 #else /* STAT64 > 0 */
987 fstat(sm_io_getinfo(df, SM_IO_WHAT_FD, NULL), &st)
988 #endif /* STAT64 > 0 */
991 (void) sm_io_reopen(SmFtStdio, SM_TIME_DEFAULT, dfname,
992 SM_IO_WRONLY_B, NULL, df);
994 (void) sm_io_fprintf(df, SM_TIME_DEFAULT,
995 "\n*** Mail could not be accepted");
997 (void) sm_io_fprintf(df, SM_TIME_DEFAULT,
998 "\n*** Mail of at least %llu bytes could not be accepted\n",
999 (ULONGLONG_T) st.st_size);
1000 (void) sm_io_fprintf(df, SM_TIME_DEFAULT,
1001 "*** at %s due to lack of disk space for temp file.\n",
1003 avail = freediskspace(qid_printqueue(e->e_qgrp, e->e_qdir),
1008 avail *= bsize / 1024;
1009 else if (bsize < 1024)
1010 avail /= 1024 / bsize;
1011 (void) sm_io_fprintf(df, SM_TIME_DEFAULT,
1012 "*** Currently, %ld kilobytes are available for mail temp files.\n",
1016 /* Wrong response code; should be 421. */
1017 e->e_status = "4.3.1";
1018 usrerrenh(e->e_status, "452 Out of disk space for temp file");
1020 syserr("421 4.3.1 Out of disk space for temp file");
1024 syserr("421 4.3.0 collect: Cannot write %s (%s, uid=%d, gid=%d)",
1025 dfname, msg, (int) geteuid(), (int) getegid());
1026 if (sm_io_reopen(SmFtStdio, SM_TIME_DEFAULT, SM_PATH_DEVNULL,
1027 SM_IO_WRONLY, NULL, df) == NULL)
1028 sm_syslog(LOG_ERR, e->e_id,
1029 "dferror: sm_io_reopen(\"/dev/null\") failed: %s",
1030 sm_errstring(errno));
1033 ** EATFROM -- chew up a UNIX style from line and process
1035 ** This does indeed make some assumptions about the format
1036 ** of UNIX messages.
1039 ** fm -- the from line.
1046 ** extracts what information it can from the header,
1047 ** such as the date.
1052 static char *DowList[] =
1054 "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat", NULL
1057 static char *MonthList[] =
1059 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
1060 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec",
1067 register ENVELOPE *e;
1073 sm_dprintf("eatfrom(%s)\n", fm);
1075 /* find the date part */
1080 while (*p != '\0' && *p != ' ')
1086 /* no room for the date */
1089 if (!(isascii(*p) && isupper(*p)) ||
1090 p[3] != ' ' || p[13] != ':' || p[16] != ':')
1093 /* we have a possible date */
1094 for (dt = DowList; *dt != NULL; dt++)
1095 if (strncmp(*dt, p, 3) == 0)
1100 for (dt = MonthList; *dt != NULL; dt++)
1102 if (strncmp(*dt, &p[4], 3) == 0)
1113 /* we have found a date */
1114 (void) sm_strlcpy(buf, p, sizeof(buf));
1116 macdefine(&e->e_macro, A_TEMP, 'a', q);
1119 #endif /* ! NOTUNIX */