1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
5 /* Copyright (c) Tom Kistner <tom@duncanthrax.net> 2004 - 2015
7 * Copyright (c) The Exim Maintainers 2015 - 2018
11 #ifdef WITH_CONTENT_SCAN /* entire file */
15 FILE *mime_stream
= NULL
;
16 uschar
*mime_current_boundary
= NULL
;
18 static mime_header mime_header_list
[] = {
19 /* name namelen value */
20 { US
"content-type:", 13, &mime_content_type
},
21 { US
"content-disposition:", 20, &mime_content_disposition
},
22 { US
"content-transfer-encoding:", 26, &mime_content_transfer_encoding
},
23 { US
"content-id:", 11, &mime_content_id
},
24 { US
"content-description:", 20, &mime_content_description
}
27 static int mime_header_list_size
= nelem(mime_header_list
);
29 static mime_parameter mime_parameter_list
[] = {
30 /* name namelen value */
31 { US
"name=", 5, &mime_filename
},
32 { US
"filename=", 9, &mime_filename
},
33 { US
"charset=", 8, &mime_charset
},
34 { US
"boundary=", 9, &mime_boundary
}
38 /*************************************************
39 * set MIME anomaly level + text *
40 *************************************************/
42 /* Small wrapper to set the two expandables which
43 give info on detected "problems" in MIME
44 encodings. Indexes are defined in mime.h. */
47 mime_set_anomaly(int idx
)
52 } anom
[] = { {1, CUS
"Broken Quoted-Printable encoding detected"},
53 {2, CUS
"Broken BASE64 encoding detected"} };
55 mime_anomaly_level
= anom
[idx
].level
;
56 mime_anomaly_text
= anom
[idx
].text
;
60 /*************************************************
61 * decode quoted-printable chars *
62 *************************************************/
64 /* gets called when we hit a =
65 returns: new pointer position
68 -1 - soft line break, no char
73 mime_decode_qp_char(uschar
*qp_p
, int *c
)
75 uschar
*initial_pos
= qp_p
;
77 /* advance one char */
80 /* Check for two hex digits and decode them */
81 if (isxdigit(*qp_p
) && isxdigit(qp_p
[1]))
83 /* Do hex conversion */
84 *c
= (isdigit(*qp_p
) ? *qp_p
- '0' : toupper(*qp_p
) - 'A' + 10) <<4;
86 *c
|= isdigit(*qp_p
) ? *qp_p
- '0' : toupper(*qp_p
) - 'A' + 10;
90 /* tab or whitespace may follow just ignore it if it precedes \n */
91 while (*qp_p
== '\t' || *qp_p
== ' ' || *qp_p
== '\r')
94 if (*qp_p
== '\n') /* hit soft line break */
100 /* illegal char here */
106 /* just dump MIME part without any decoding */
108 mime_decode_asis(FILE* in
, FILE* out
, uschar
* boundary
)
110 ssize_t len
, size
= 0;
111 uschar buffer
[MIME_MAX_LINE_LENGTH
];
113 while(fgets(CS buffer
, MIME_MAX_LINE_LENGTH
, mime_stream
) != NULL
)
116 && Ustrncmp(buffer
, "--", 2) == 0
117 && Ustrncmp((buffer
+2), boundary
, Ustrlen(boundary
)) == 0
121 len
= Ustrlen(buffer
);
122 if (fwrite(buffer
, 1, (size_t)len
, out
) < len
)
131 /* decode quoted-printable MIME part */
133 mime_decode_qp(FILE* in
, FILE* out
, uschar
* boundary
)
135 uschar ibuf
[MIME_MAX_LINE_LENGTH
], obuf
[MIME_MAX_LINE_LENGTH
];
137 ssize_t len
, size
= 0;
139 while (fgets(CS ibuf
, MIME_MAX_LINE_LENGTH
, in
) != NULL
)
142 && Ustrncmp(ibuf
, "--", 2) == 0
143 && Ustrncmp((ibuf
+2), boundary
, Ustrlen(boundary
)) == 0
145 break; /* todo: check for missing boundary */
154 int decode_qp_result
;
156 ipos
= mime_decode_qp_char(ipos
, &decode_qp_result
);
158 if (decode_qp_result
== -2)
160 /* Error from decoder. ipos is unchanged. */
161 mime_set_anomaly(MIME_ANOMALY_BROKEN_QP
);
165 else if (decode_qp_result
== -1)
167 else if (decode_qp_result
>= 0)
168 *opos
++ = decode_qp_result
;
173 /* something to write? */
177 if (fwrite(obuf
, 1, len
, out
) != len
) return -1; /* error */
186 * Return open filehandle for combo of path and file.
187 * Side-effect: set mime_decoded_filename, to copy in allocated mem
190 mime_get_decode_file(uschar
*pname
, uschar
*fname
)
193 mime_decoded_filename
= string_sprintf("%s/%s", pname
, fname
);
195 mime_decoded_filename
= string_copy(fname
);
201 /* must find first free sequential filename */
205 mime_decoded_filename
= string_sprintf("%s/%s-%05u", pname
, message_id
, file_nr
++);
209 result
= stat(CS mime_decoded_filename
, &mystat
);
210 } while(result
!= -1);
213 return modefopen(mime_decoded_filename
, "wb+", SPOOL_MODE
);
218 mime_decode(const uschar
**listptr
)
221 const uschar
*list
= *listptr
;
223 uschar
* decode_path
;
224 FILE *decode_file
= NULL
;
226 ssize_t size_counter
= 0;
227 ssize_t (*decode_function
)(FILE*, FILE*, uschar
*);
229 if (!mime_stream
|| (f_pos
= ftell(mime_stream
)) < 0)
232 /* build default decode path (will exist since MBOX must be spooled up) */
233 decode_path
= string_sprintf("%s/scan/%s", spool_directory
, message_id
);
235 /* try to find 1st option */
236 if ((option
= string_nextinlist(&list
, &sep
, NULL
, 0)))
238 /* parse 1st option */
239 if ((Ustrcmp(option
,"false") == 0) || (Ustrcmp(option
,"0") == 0))
240 /* explicitly no decoding */
243 if (Ustrcmp(option
,"default") == 0)
244 /* explicit default path + file names */
247 if (option
[0] == '/')
251 memset(&statbuf
,0,sizeof(statbuf
));
253 /* assume either path or path+file name */
254 if ( (stat(CS option
, &statbuf
) == 0) && S_ISDIR(statbuf
.st_mode
) )
255 /* is directory, use it as decode_path */
256 decode_file
= mime_get_decode_file(option
, NULL
);
258 /* does not exist or is a file, use as full file name */
259 decode_file
= mime_get_decode_file(NULL
, option
);
262 /* assume file name only, use default path */
263 decode_file
= mime_get_decode_file(decode_path
, option
);
267 /* no option? patch default path */
269 decode_file
= mime_get_decode_file(decode_path
, NULL
);
275 /* decode according to mime type */
277 !mime_content_transfer_encoding
278 ? mime_decode_asis
/* no encoding, dump as-is */
279 : Ustrcmp(mime_content_transfer_encoding
, "base64") == 0
281 : Ustrcmp(mime_content_transfer_encoding
, "quoted-printable") == 0
283 : mime_decode_asis
; /* unknown encoding type, just dump as-is */
285 size_counter
= decode_function(mime_stream
, decode_file
, mime_current_boundary
);
287 clearerr(mime_stream
);
288 if (fseek(mime_stream
, f_pos
, SEEK_SET
))
291 if (fclose(decode_file
) != 0 || size_counter
< 0)
294 /* round up to the next KiB */
295 mime_content_size
= (size_counter
+ 1023) / 1024;
302 mime_get_header(FILE *f
, uschar
*header
)
306 int header_value_mode
= 0;
307 int header_open_brackets
= 0;
312 if ((c
= fgetc(f
)) == EOF
) break;
314 /* always skip CRs */
315 if (c
== '\r') continue;
321 /* look if next char is '\t' or ' ' */
322 if ((c
= fgetc(f
)) == EOF
) break;
323 if ( (c
== '\t') || (c
== ' ') ) continue;
326 /* end of the header, terminate with ';' */
331 /* skip control characters */
332 if (c
< 32) continue;
334 if (header_value_mode
)
336 /* --------- value mode ----------- */
337 /* skip leading whitespace */
338 if ( ((c
== '\t') || (c
== ' ')) && (header_value_mode
== 1) )
341 /* we have hit a non-whitespace char, start copying value data */
342 header_value_mode
= 2;
344 if (c
== '"') /* flip "quoted" mode */
345 header_value_mode
= header_value_mode
==2 ? 3 : 2;
347 /* leave value mode on unquoted ';' */
348 if (header_value_mode
== 2 && c
== ';')
349 header_value_mode
= 0;
350 /* -------------------------------- */
354 /* -------- non-value mode -------- */
355 /* skip whitespace + tabs */
356 if ( (c
== ' ') || (c
== '\t') )
360 /* quote next char. can be used
361 to escape brackets. */
362 if ((c
= fgetc(f
)) == EOF
) break;
366 header_open_brackets
++;
369 else if ((c
== ')') && header_open_brackets
)
371 header_open_brackets
--;
374 else if ( (c
== '=') && !header_open_brackets
) /* enter value mode */
375 header_value_mode
= 1;
377 /* skip chars while we are in a comment */
378 if (header_open_brackets
> 0)
380 /* -------------------------------- */
383 /* copy the char to the buffer */
384 header
[num_copied
++] = (uschar
)c
;
386 /* break if header buffer is full */
387 if (num_copied
> MIME_MAX_HEADER_SIZE
-1)
391 if ((num_copied
> 0) && (header
[num_copied
-1] != ';'))
392 header
[num_copied
-1] = ';';
395 header
[num_copied
] = '\0';
397 /* return 0 for EOF or empty line */
398 return c
== EOF
|| num_copied
== 1 ? 0 : 1;
402 /* reset all per-part mime variables */
404 mime_vars_reset(void)
406 mime_anomaly_level
= 0;
407 mime_anomaly_text
= NULL
;
408 mime_boundary
= NULL
;
410 mime_decoded_filename
= NULL
;
411 mime_filename
= NULL
;
412 mime_content_description
= NULL
;
413 mime_content_disposition
= NULL
;
414 mime_content_id
= NULL
;
415 mime_content_transfer_encoding
= NULL
;
416 mime_content_type
= NULL
;
417 mime_is_multipart
= 0;
418 mime_content_size
= 0;
422 /* Grab a parameter value, dealing with quoting.
425 str Input string. Updated on return to point to terminating ; or NUL
428 Allocated string with parameter value
431 mime_param_val(uschar
** sp
)
434 gstring
* val
= NULL
;
436 /* debug_printf_indent(" considering paramval '%s'\n", s); */
438 while (*s
&& *s
!= ';') /* ; terminates */
441 s
++; /* skip opening " */
442 while (*s
&& *s
!= '"') /* " protects ; */
443 val
= string_catn(val
, s
++, 1);
444 if (*s
) s
++; /* skip closing " */
447 val
= string_catn(val
, s
++, 1);
449 return string_from_gstring(val
);
453 mime_next_semicolon(uschar
* s
)
455 while (*s
&& *s
!= ';') /* ; terminates */
458 s
++; /* skip opening " */
459 while (*s
&& *s
!= '"') /* " protects ; */
461 if (*s
) s
++; /* skip closing " */
470 rfc2231_to_2047(const uschar
* fname
, const uschar
* charset
, int * len
)
472 gstring
* val
= string_catn(NULL
, US
"=?", 2);
476 val
= string_cat(val
, charset
);
477 val
= string_catn(val
, US
"?Q?", 3);
480 if (c
== '%' && isxdigit(fname
[1]) && isxdigit(fname
[2]))
482 val
= string_catn(val
, US
"=", 1);
483 val
= string_catn(val
, ++fname
, 2);
487 val
= string_catn(val
, fname
++, 1);
489 val
= string_catn(val
, US
"?=", 2);
491 return string_from_gstring(val
);
496 mime_acl_check(uschar
*acl
, FILE *f
, struct mime_boundary_context
*context
,
497 uschar
**user_msgptr
, uschar
**log_msgptr
)
500 uschar
* header
= NULL
;
501 struct mime_boundary_context nested_context
;
503 /* reserve a line buffer to work in. Assume tainted data. */
504 header
= store_get(MIME_MAX_HEADER_SIZE
+1, TRUE
);
506 /* Not actually used at the moment, but will be vital to fixing
507 * some RFC 2046 nonconformance later... */
508 nested_context
.parent
= context
;
510 /* loop through parts */
513 /* reset all per-part mime variables */
516 /* If boundary is null, we assume that *f is positioned on the start of
517 headers (for example, at the very beginning of a message. If a boundary is
518 given, we must first advance to it to reach the start of the next header
521 /* NOTE -- there's an error here -- RFC2046 specifically says to
522 * check for outer boundaries. This code doesn't do that, and
523 * I haven't fixed this.
525 * (I have moved partway towards adding support, however, by adding
526 * a "parent" field to my new boundary-context structure.)
528 if (context
) for (;;)
530 if (!fgets(CS header
, MIME_MAX_HEADER_SIZE
, f
))
532 /* Hit EOF or read error. Ugh. */
533 DEBUG(D_acl
) debug_printf_indent("MIME: Hit EOF ...\n");
537 /* boundary line must start with 2 dashes */
538 if ( Ustrncmp(header
, "--", 2) == 0
539 && Ustrncmp(header
+2, context
->boundary
, Ustrlen(context
->boundary
)) == 0
541 { /* found boundary */
542 if (Ustrncmp((header
+2+Ustrlen(context
->boundary
)), "--", 2) == 0)
544 /* END boundary found */
545 DEBUG(D_acl
) debug_printf_indent("MIME: End boundary found %s\n",
550 DEBUG(D_acl
) debug_printf_indent("MIME: Next part with boundary %s\n",
556 /* parse headers, set up expansion variables */
557 while (mime_get_header(f
, header
))
559 /* look for interesting headers */
560 for (struct mime_header
* mh
= mime_header_list
;
561 mh
< mime_header_list
+ mime_header_list_size
;
562 mh
++) if (strncmpic(mh
->name
, header
, mh
->namelen
) == 0)
564 uschar
* p
= header
+ mh
->namelen
;
567 /* grab the value (normalize to lower case)
568 and copy to its corresponding expansion variable */
570 for (q
= p
; *q
!= ';' && *q
; q
++) ;
571 *mh
->value
= string_copynlc(p
, q
-p
);
572 DEBUG(D_acl
) debug_printf_indent("MIME: found %s header, value is '%s'\n",
573 mh
->name
, *mh
->value
);
575 if (*(p
= q
)) p
++; /* jump past the ; */
578 uschar
* mime_fname
= NULL
;
579 uschar
* mime_fname_rfc2231
= NULL
;
580 uschar
* mime_filename_charset
= NULL
;
581 BOOL decoding_failed
= FALSE
;
583 /* grab all param=value tags on the remaining line,
584 check if they are interesting */
588 DEBUG(D_acl
) debug_printf_indent("MIME: considering paramlist '%s'\n", p
);
591 && strncmpic(CUS
"content-disposition:", header
, 20) == 0
592 && strncmpic(CUS
"filename*", p
, 9) == 0
594 { /* RFC 2231 filename */
597 /* find value of the filename */
599 while(*p
!= '=' && *p
) p
++;
600 if (*p
) p
++; /* p is filename or NUL */
601 q
= mime_param_val(&p
); /* p now trailing ; or NUL */
605 uschar
* temp_string
, * err_msg
;
608 /* build up an un-decoded filename over successive
609 filename*= parameters (for use when 2047 decode fails) */
611 mime_fname_rfc2231
= string_sprintf("%#s%s",
612 mime_fname_rfc2231
, q
);
614 if (!decoding_failed
)
617 if (!mime_filename_charset
)
621 /* look for a ' in the "filename" */
622 while(*s
!= '\'' && *s
) s
++; /* s is 1st ' or NUL */
624 if ((size
= s
-q
) > 0)
625 mime_filename_charset
= string_copyn(q
, size
);
628 while(*p
== '\'') p
++; /* p is after 2nd ' */
633 DEBUG(D_acl
) debug_printf_indent("MIME: charset %s fname '%s'\n",
634 mime_filename_charset
? mime_filename_charset
: US
"<NULL>", p
);
636 temp_string
= rfc2231_to_2047(p
, mime_filename_charset
, &slen
);
637 DEBUG(D_acl
) debug_printf_indent("MIME: 2047-name %s\n", temp_string
);
639 temp_string
= rfc2047_decode(temp_string
, FALSE
, NULL
, ' ',
641 DEBUG(D_acl
) debug_printf_indent("MIME: plain-name %s\n", temp_string
);
643 if (!temp_string
|| (size
= Ustrlen(temp_string
)) == slen
)
644 decoding_failed
= TRUE
;
646 /* build up a decoded filename over successive
647 filename*= parameters */
649 mime_filename
= mime_fname
= mime_fname
650 ? string_sprintf("%s%s", mime_fname
, temp_string
)
657 /* look for interesting parameters */
658 for (mime_parameter
* mp
= mime_parameter_list
;
659 mp
< mime_parameter_list
+ nelem(mime_parameter_list
);
661 ) if (strncmpic(mp
->name
, p
, mp
->namelen
) == 0)
664 uschar
* dummy_errstr
;
666 /* grab the value and copy to its expansion variable */
668 q
= mime_param_val(&p
); /* p now trailing ; or NUL */
671 ? rfc2047_decode(q
, check_rfc2047_length
, NULL
, 32, NULL
,
674 DEBUG(D_acl
) debug_printf_indent(
675 "MIME: found %s parameter in %s header, value '%s'\n",
676 mp
->name
, mh
->name
, *mp
->value
);
678 break; /* done matching param names */
682 /* There is something, but not one of our interesting parameters.
683 Advance past the next semicolon */
684 p
= mime_next_semicolon(p
);
686 } /* param scan on line */
688 if (strncmpic(CUS
"content-disposition:", header
, 20) == 0)
690 if (decoding_failed
) mime_filename
= mime_fname_rfc2231
;
692 DEBUG(D_acl
) debug_printf_indent(
693 "MIME: found %s parameter in %s header, value is '%s'\n",
694 "filename", mh
->name
, mime_filename
);
699 /* set additional flag variables (easier access) */
700 if ( mime_content_type
701 && Ustrncmp(mime_content_type
,"multipart",9) == 0
703 mime_is_multipart
= 1;
705 /* Make a copy of the boundary pointer.
706 Required since mime_boundary is global
707 and can be overwritten further down in recursion */
708 nested_context
.boundary
= mime_boundary
;
710 /* raise global counter */
713 /* copy current file handle to global variable */
715 mime_current_boundary
= context
? context
->boundary
: 0;
717 /* Note the context */
718 mime_is_coverletter
= !(context
&& context
->context
== MBC_ATTACHMENT
);
720 /* call ACL handling function */
721 rc
= acl_check(ACL_WHERE_MIME
, NULL
, acl
, user_msgptr
, log_msgptr
);
724 mime_current_boundary
= NULL
;
728 /* If we have a multipart entity and a boundary, go recursive */
729 if ( mime_content_type
&& nested_context
.boundary
730 && Ustrncmp(mime_content_type
,"multipart",9) == 0)
733 debug_printf_indent("MIME: Entering multipart recursion, boundary '%s'\n",
734 nested_context
.boundary
);
736 nested_context
.context
=
737 context
&& context
->context
== MBC_ATTACHMENT
739 : Ustrcmp(mime_content_type
,"multipart/alternative") == 0
740 || Ustrcmp(mime_content_type
,"multipart/related") == 0
741 ? MBC_COVERLETTER_ALL
742 : MBC_COVERLETTER_ONESHOT
;
744 rc
= mime_acl_check(acl
, f
, &nested_context
, user_msgptr
, log_msgptr
);
747 else if ( mime_content_type
748 && Ustrncmp(mime_content_type
,"message/rfc822",14) == 0)
750 const uschar
* rfc822name
= NULL
;
755 /* must find first free sequential filename */
756 for (gstring
* g
= string_get(64); result
!= -1; g
->ptr
= 0)
759 g
= string_fmt_append(g
,
760 "%s/scan/%s/__rfc822_%05u", spool_directory
, message_id
, file_nr
++);
764 result
= stat(CS (filename
= string_from_gstring(g
)), &mystat
);
767 rfc822name
= filename
;
769 /* decode RFC822 attachment */
770 mime_decoded_filename
= NULL
;
772 mime_current_boundary
= context
? context
->boundary
: NULL
;
773 mime_decode(&rfc822name
);
775 mime_current_boundary
= NULL
;
776 if (!mime_decoded_filename
) /* decoding failed */
778 log_write(0, LOG_MAIN
,
779 "MIME acl condition warning - could not decode RFC822 MIME part to file.");
783 mime_decoded_filename
= NULL
;
787 /* If the boundary of this instance is NULL, we are finished here */
790 if (context
->context
== MBC_COVERLETTER_ONESHOT
)
791 context
->context
= MBC_ATTACHMENT
;
799 #endif /*WITH_CONTENT_SCAN*/