1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
5 /* Copyright (c) Tom Kistner <tom@duncanthrax.net> 2004 */
9 #ifdef WITH_CONTENT_SCAN
13 FILE *mime_stream
= NULL
;
14 uschar
*mime_current_boundary
= NULL
;
16 /*************************************************
17 * set MIME anomaly level + text *
18 *************************************************/
20 /* Small wrapper to set the two expandables which
21 give info on detected "problems" in MIME
22 encodings. Those are defined in mime.h. */
24 void mime_set_anomaly(int level
, const char *text
) {
25 mime_anomaly_level
= level
;
26 mime_anomaly_text
= CUS text
;
30 /*************************************************
31 * decode quoted-printable chars *
32 *************************************************/
34 /* gets called when we hit a =
35 returns: new pointer position
38 -1 - soft line break, no char
42 uschar
*mime_decode_qp_char(uschar
*qp_p
, int *c
) {
43 uschar
*initial_pos
= qp_p
;
45 /* advance one char */
48 /* Check for two hex digits and decode them */
49 if (isxdigit(*qp_p
) && isxdigit(qp_p
[1])) {
50 /* Do hex conversion */
51 if (isdigit(*qp_p
)) {*c
= *qp_p
- '0';}
52 else {*c
= toupper(*qp_p
) - 'A' + 10;};
54 if (isdigit(qp_p
[1])) {*c
|= qp_p
[1] - '0';}
55 else {*c
|= toupper(qp_p
[1]) - 'A' + 10;};
59 /* tab or whitespace may follow just ignore it if it precedes \n */
60 while (*qp_p
== '\t' || *qp_p
== ' ' || *qp_p
== '\r')
64 /* hit soft line break */
69 /* illegal char here */
75 /* just dump MIME part without any decoding */
76 static int mime_decode_asis(FILE* in
, FILE* out
, uschar
* boundary
)
79 uschar buffer
[MIME_MAX_LINE_LENGTH
];
81 while(fgets(CS buffer
, MIME_MAX_LINE_LENGTH
, mime_stream
) != NULL
) {
83 && Ustrncmp(buffer
, "--", 2) == 0
84 && Ustrncmp((buffer
+2), boundary
, Ustrlen(boundary
)) == 0
88 len
= Ustrlen(buffer
);
89 if (fwrite(buffer
, 1, (size_t)len
, out
) < len
)
97 /* decode base64 MIME part */
98 static int mime_decode_base64(FILE* in
, FILE* out
, uschar
* boundary
)
100 uschar ibuf
[MIME_MAX_LINE_LENGTH
], obuf
[MIME_MAX_LINE_LENGTH
];
102 size_t len
, size
= 0;
107 while (Ufgets(ibuf
, MIME_MAX_LINE_LENGTH
, in
) != NULL
)
110 && Ustrncmp(ibuf
, "--", 2) == 0
111 && Ustrncmp((ibuf
+2), boundary
, Ustrlen(boundary
)) == 0
115 for (ipos
= ibuf
; *ipos
!= '\r' && *ipos
!= '\n' && *ipos
!= 0; ++ipos
) {
121 /* skip bad characters */
122 if (mime_b64
[*ipos
] == 128) {
123 mime_set_anomaly(MIME_ANOMALY_BROKEN_BASE64
);
126 /* simple state-machine */
127 switch((bytestate
++) & 3) {
129 *opos
= mime_b64
[*ipos
] << 2;
132 *opos
|= mime_b64
[*ipos
] >> 4;
134 *opos
= mime_b64
[*ipos
] << 4;
137 *opos
|= mime_b64
[*ipos
] >> 2;
139 *opos
= mime_b64
[*ipos
] << 6;
142 *opos
|= mime_b64
[*ipos
];
147 /* something to write? */
150 if (fwrite(obuf
, 1, len
, out
) != len
)
151 return -1; /* error */
153 /* copy incomplete last byte to start of obuf, where we continue */
154 if ((bytestate
& 3) != 0)
160 /* write out last byte if it was incomplete */
162 if (fwrite(obuf
, 1, 1, out
) != 1)
171 /* decode quoted-printable MIME part */
172 static int mime_decode_qp(FILE* in
, FILE* out
, uschar
* boundary
)
174 uschar ibuf
[MIME_MAX_LINE_LENGTH
], obuf
[MIME_MAX_LINE_LENGTH
];
176 size_t len
, size
= 0;
178 while (fgets(CS ibuf
, MIME_MAX_LINE_LENGTH
, in
) != NULL
)
181 && Ustrncmp(ibuf
, "--", 2) == 0
182 && Ustrncmp((ibuf
+2), boundary
, Ustrlen(boundary
)) == 0
184 break; /* todo: check for missing boundary */
191 int decode_qp_result
;
193 ipos
= mime_decode_qp_char(ipos
, &decode_qp_result
);
195 if (decode_qp_result
== -2) {
196 /* Error from decoder. ipos is unchanged. */
197 mime_set_anomaly(MIME_ANOMALY_BROKEN_QP
);
202 else if (decode_qp_result
== -1) {
205 else if (decode_qp_result
>= 0) {
206 *opos
= decode_qp_result
;
216 /* something to write? */
219 if (fwrite(obuf
, 1, len
, out
) != len
)
220 return -1; /* error */
228 FILE *mime_get_decode_file(uschar
*pname
, uschar
*fname
) {
232 filename
= (uschar
*)malloc(2048);
234 if ((pname
!= NULL
) && (fname
!= NULL
)) {
235 (void)string_format(filename
, 2048, "%s/%s", pname
, fname
);
236 f
= modefopen(filename
,"wb+",SPOOL_MODE
);
238 else if (pname
== NULL
) {
239 f
= modefopen(fname
,"wb+",SPOOL_MODE
);
241 else if (fname
== NULL
) {
245 /* must find first free sequential filename */
248 (void)string_format(filename
,2048,"%s/%s-%05u", pname
, message_id
, file_nr
);
253 result
= stat(CS filename
,&mystat
);
256 f
= modefopen(filename
,"wb+",SPOOL_MODE
);
259 /* set expansion variable */
260 mime_decoded_filename
= filename
;
266 int mime_decode(uschar
**listptr
) {
268 uschar
*list
= *listptr
;
270 uschar option_buffer
[1024];
271 uschar decode_path
[1024];
272 FILE *decode_file
= NULL
;
274 unsigned int size_counter
= 0;
275 int (*decode_function
)(FILE*, FILE*, uschar
*);
277 if (mime_stream
== NULL
)
280 f_pos
= ftell(mime_stream
);
282 /* build default decode path (will exist since MBOX must be spooled up) */
283 (void)string_format(decode_path
,1024,"%s/scan/%s",spool_directory
,message_id
);
285 /* try to find 1st option */
286 if ((option
= string_nextinlist(&list
, &sep
,
288 sizeof(option_buffer
))) != NULL
) {
290 /* parse 1st option */
291 if ( (Ustrcmp(option
,"false") == 0) || (Ustrcmp(option
,"0") == 0) ) {
292 /* explicitly no decoding */
296 if (Ustrcmp(option
,"default") == 0) {
297 /* explicit default path + file names */
301 if (option
[0] == '/') {
304 memset(&statbuf
,0,sizeof(statbuf
));
306 /* assume either path or path+file name */
307 if ( (stat(CS option
, &statbuf
) == 0) && S_ISDIR(statbuf
.st_mode
) )
308 /* is directory, use it as decode_path */
309 decode_file
= mime_get_decode_file(option
, NULL
);
311 /* does not exist or is a file, use as full file name */
312 decode_file
= mime_get_decode_file(NULL
, option
);
315 /* assume file name only, use default path */
316 decode_file
= mime_get_decode_file(decode_path
, option
);
319 /* no option? patch default path */
320 DEFAULT_PATH
: decode_file
= mime_get_decode_file(decode_path
, NULL
);
322 if (decode_file
== NULL
)
325 /* decode according to mime type */
326 if (mime_content_transfer_encoding
== NULL
)
327 /* no encoding, dump as-is */
328 decode_function
= mime_decode_asis
;
329 else if (Ustrcmp(mime_content_transfer_encoding
, "base64") == 0)
330 decode_function
= mime_decode_base64
;
331 else if (Ustrcmp(mime_content_transfer_encoding
, "quoted-printable") == 0)
332 decode_function
= mime_decode_qp
;
334 /* unknown encoding type, just dump as-is */
335 decode_function
= mime_decode_asis
;
337 size_counter
= decode_function(mime_stream
, decode_file
, mime_current_boundary
);
339 clearerr(mime_stream
);
340 fseek(mime_stream
, f_pos
, SEEK_SET
);
342 if (fclose(decode_file
) != 0 || size_counter
< 0)
345 /* round up to the next KiB */
346 mime_content_size
= (size_counter
+ 1023) / 1024;
351 int mime_get_header(FILE *f
, uschar
*header
) {
354 int header_value_mode
= 0;
355 int header_open_brackets
= 0;
363 /* always skip CRs */
364 if (c
== '\r') continue;
367 if (num_copied
> 0) {
368 /* look if next char is '\t' or ' ' */
371 if ( (c
== '\t') || (c
== ' ') ) continue;
374 /* end of the header, terminate with ';' */
379 /* skip control characters */
380 if (c
< 32) continue;
382 if (header_value_mode
) {
383 /* --------- value mode ----------- */
384 /* skip leading whitespace */
385 if ( ((c
== '\t') || (c
== ' ')) && (header_value_mode
== 1) )
388 /* we have hit a non-whitespace char, start copying value data */
389 header_value_mode
= 2;
392 if (c
== '"') continue;
394 /* leave value mode on ';' */
396 header_value_mode
= 0;
398 /* -------------------------------- */
401 /* -------- non-value mode -------- */
402 /* skip whitespace + tabs */
403 if ( (c
== ' ') || (c
== '\t') )
406 /* quote next char. can be used
407 to escape brackets. */
412 header_open_brackets
++;
415 else if ((c
== ')') && header_open_brackets
) {
416 header_open_brackets
--;
419 else if ( (c
== '=') && !header_open_brackets
) {
420 /* enter value mode */
421 header_value_mode
= 1;
424 /* skip chars while we are in a comment */
425 if (header_open_brackets
> 0)
427 /* -------------------------------- */
430 /* copy the char to the buffer */
431 header
[num_copied
] = (uschar
)c
;
435 /* break if header buffer is full */
436 if (num_copied
> MIME_MAX_HEADER_SIZE
-1) {
441 if ((num_copied
> 0) && (header
[num_copied
-1] != ';')) {
442 header
[num_copied
-1] = ';';
446 header
[num_copied
] = '\0';
448 /* return 0 for EOF or empty line */
449 if ((c
== EOF
) || (num_copied
== 1))
456 int mime_acl_check(uschar
*acl
, FILE *f
, struct mime_boundary_context
*context
,
457 uschar
**user_msgptr
, uschar
**log_msgptr
) {
459 uschar
*header
= NULL
;
460 struct mime_boundary_context nested_context
;
462 /* reserve a line buffer to work in */
463 header
= (uschar
*)malloc(MIME_MAX_HEADER_SIZE
+1);
464 if (header
== NULL
) {
465 log_write(0, LOG_PANIC
,
466 "MIME ACL: can't allocate %d bytes of memory.", MIME_MAX_HEADER_SIZE
+1);
470 /* Not actually used at the moment, but will be vital to fixing
471 * some RFC 2046 nonconformance later... */
472 nested_context
.parent
= context
;
474 /* loop through parts */
477 /* reset all per-part mime variables */
478 mime_anomaly_level
= 0;
479 mime_anomaly_text
= NULL
;
480 mime_boundary
= NULL
;
482 mime_decoded_filename
= NULL
;
483 mime_filename
= NULL
;
484 mime_content_description
= NULL
;
485 mime_content_disposition
= NULL
;
486 mime_content_id
= NULL
;
487 mime_content_transfer_encoding
= NULL
;
488 mime_content_type
= NULL
;
489 mime_is_multipart
= 0;
490 mime_content_size
= 0;
493 If boundary is null, we assume that *f is positioned on the start of headers (for example,
494 at the very beginning of a message.
495 If a boundary is given, we must first advance to it to reach the start of the next header
499 /* NOTE -- there's an error here -- RFC2046 specifically says to
500 * check for outer boundaries. This code doesn't do that, and
501 * I haven't fixed this.
503 * (I have moved partway towards adding support, however, by adding
504 * a "parent" field to my new boundary-context structure.)
506 if (context
!= NULL
) {
507 while(fgets(CS header
, MIME_MAX_HEADER_SIZE
, f
) != NULL
) {
508 /* boundary line must start with 2 dashes */
509 if (Ustrncmp(header
,"--",2) == 0) {
510 if (Ustrncmp((header
+2),context
->boundary
,Ustrlen(context
->boundary
)) == 0) {
512 if (Ustrncmp((header
+2+Ustrlen(context
->boundary
)),"--",2) == 0) {
513 /* END boundary found */
514 debug_printf("End boundary found %s\n", context
->boundary
);
518 debug_printf("Next part with boundary %s\n", context
->boundary
);
520 /* can't use break here */
525 /* Hit EOF or read error. Ugh. */
526 debug_printf("Hit EOF ...\n");
531 /* parse headers, set up expansion variables */
532 while(mime_get_header(f
,header
)) {
534 /* loop through header list */
535 for (i
= 0; i
< mime_header_list_size
; i
++) {
536 uschar
*header_value
= NULL
;
537 int header_value_len
= 0;
539 /* found an interesting header? */
540 if (strncmpic(mime_header_list
[i
].name
,header
,mime_header_list
[i
].namelen
) == 0) {
541 uschar
*p
= header
+ mime_header_list
[i
].namelen
;
542 /* yes, grab the value (normalize to lower case)
543 and copy to its corresponding expansion variable */
548 header_value_len
= (p
- (header
+ mime_header_list
[i
].namelen
));
549 header_value
= (uschar
*)malloc(header_value_len
+1);
550 memset(header_value
,0,header_value_len
+1);
551 p
= header
+ mime_header_list
[i
].namelen
;
552 Ustrncpy(header_value
, p
, header_value_len
);
553 debug_printf("Found %s MIME header, value is '%s'\n", mime_header_list
[i
].name
, header_value
);
554 *((uschar
**)(mime_header_list
[i
].value
)) = header_value
;
556 /* make p point to the next character after the closing ';' */
557 p
+= (header_value_len
+1);
559 /* grab all param=value tags on the remaining line, check if they are interesting */
560 NEXT_PARAM_SEARCH
: while (*p
!= 0) {
562 for (j
= 0; j
< mime_parameter_list_size
; j
++) {
563 uschar
*param_value
= NULL
;
564 int param_value_len
= 0;
566 /* found an interesting parameter? */
567 if (strncmpic(mime_parameter_list
[j
].name
,p
,mime_parameter_list
[j
].namelen
) == 0) {
568 uschar
*q
= p
+ mime_parameter_list
[j
].namelen
;
569 /* yes, grab the value and copy to its corresponding expansion variable */
570 while(*q
!= ';') q
++;
571 param_value_len
= (q
- (p
+ mime_parameter_list
[j
].namelen
));
572 param_value
= (uschar
*)malloc(param_value_len
+1);
573 memset(param_value
,0,param_value_len
+1);
574 q
= p
+ mime_parameter_list
[j
].namelen
;
575 Ustrncpy(param_value
, q
, param_value_len
);
576 param_value
= rfc2047_decode(param_value
, check_rfc2047_length
, NULL
, 32, ¶m_value_len
, &q
);
577 debug_printf("Found %s MIME parameter in %s header, value is '%s'\n", mime_parameter_list
[j
].name
, mime_header_list
[i
].name
, param_value
);
578 *((uschar
**)(mime_parameter_list
[j
].value
)) = param_value
;
579 p
+= (mime_parameter_list
[j
].namelen
+ param_value_len
+ 1);
580 goto NEXT_PARAM_SEARCH
;
583 /* There is something, but not one of our interesting parameters.
584 Advance to the next semicolon */
585 while(*p
!= ';') p
++;
592 /* set additional flag variables (easier access) */
593 if ( (mime_content_type
!= NULL
) &&
594 (Ustrncmp(mime_content_type
,"multipart",9) == 0) )
595 mime_is_multipart
= 1;
597 /* Make a copy of the boundary pointer.
598 Required since mime_boundary is global
599 and can be overwritten further down in recursion */
600 nested_context
.boundary
= mime_boundary
;
602 /* raise global counter */
605 /* copy current file handle to global variable */
607 mime_current_boundary
= context
? context
->boundary
: 0;
609 /* Note the context */
610 mime_is_coverletter
= !(context
&& context
->context
== MBC_ATTACHMENT
);
612 /* call ACL handling function */
613 rc
= acl_check(ACL_WHERE_MIME
, NULL
, acl
, user_msgptr
, log_msgptr
);
616 mime_current_boundary
= NULL
;
620 /* If we have a multipart entity and a boundary, go recursive */
621 if ( (mime_content_type
!= NULL
) &&
622 (nested_context
.boundary
!= NULL
) &&
623 (Ustrncmp(mime_content_type
,"multipart",9) == 0) ) {
624 debug_printf("Entering multipart recursion, boundary '%s'\n", nested_context
.boundary
);
626 if (context
&& context
->context
== MBC_ATTACHMENT
)
627 nested_context
.context
= MBC_ATTACHMENT
;
628 else if (!Ustrcmp(mime_content_type
,"multipart/alternative")
629 || !Ustrcmp(mime_content_type
,"multipart/related"))
630 nested_context
.context
= MBC_COVERLETTER_ALL
;
632 nested_context
.context
= MBC_COVERLETTER_ONESHOT
;
634 rc
= mime_acl_check(acl
, f
, &nested_context
, user_msgptr
, log_msgptr
);
637 else if ( (mime_content_type
!= NULL
) &&
638 (Ustrncmp(mime_content_type
,"message/rfc822",14) == 0) ) {
639 uschar
*rfc822name
= NULL
;
640 uschar filename
[2048];
644 /* must find first free sequential filename */
647 (void)string_format(filename
,2048,"%s/scan/%s/__rfc822_%05u", spool_directory
, message_id
, file_nr
);
652 result
= stat(CS filename
,&mystat
);
656 rfc822name
= filename
;
658 /* decode RFC822 attachment */
659 mime_decoded_filename
= NULL
;
661 mime_current_boundary
= context
? context
->boundary
: NULL
;
662 mime_decode(&rfc822name
);
664 mime_current_boundary
= NULL
;
665 if (mime_decoded_filename
== NULL
) {
666 /* decoding failed */
667 log_write(0, LOG_MAIN
,
668 "mime_regex acl condition warning - could not decode RFC822 MIME part to file.");
671 mime_decoded_filename
= NULL
;
675 /* If the boundary of this instance is NULL, we are finished here */
676 if (context
== NULL
) break;
678 if (context
->context
== MBC_COVERLETTER_ONESHOT
)
679 context
->context
= MBC_ATTACHMENT
;