1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
5 /* Copyright (c) Tom Kistner <tom@duncanthrax.net> 2004 */
9 #ifdef WITH_CONTENT_SCAN /* entire file */
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. */
25 mime_set_anomaly(int level
, const char *text
)
27 mime_anomaly_level
= level
;
28 mime_anomaly_text
= CUS text
;
32 /*************************************************
33 * decode quoted-printable chars *
34 *************************************************/
36 /* gets called when we hit a =
37 returns: new pointer position
40 -1 - soft line break, no char
45 mime_decode_qp_char(uschar
*qp_p
, int *c
)
47 uschar
*initial_pos
= qp_p
;
49 /* advance one char */
52 /* Check for two hex digits and decode them */
53 if (isxdigit(*qp_p
) && isxdigit(qp_p
[1]))
55 /* Do hex conversion */
56 *c
= (isdigit(*qp_p
) ? *qp_p
- '0' : toupper(*qp_p
) - 'A' + 10) <<4;
58 *c
|= isdigit(*qp_p
) ? *qp_p
- '0' : toupper(*qp_p
) - 'A' + 10;
62 /* tab or whitespace may follow just ignore it if it precedes \n */
63 while (*qp_p
== '\t' || *qp_p
== ' ' || *qp_p
== '\r')
66 if (*qp_p
== '\n') /* hit soft line break */
72 /* illegal char here */
78 /* just dump MIME part without any decoding */
80 mime_decode_asis(FILE* in
, FILE* out
, uschar
* boundary
)
82 ssize_t len
, size
= 0;
83 uschar buffer
[MIME_MAX_LINE_LENGTH
];
85 while(fgets(CS buffer
, MIME_MAX_LINE_LENGTH
, mime_stream
) != NULL
)
88 && Ustrncmp(buffer
, "--", 2) == 0
89 && Ustrncmp((buffer
+2), boundary
, Ustrlen(boundary
)) == 0
93 len
= Ustrlen(buffer
);
94 if (fwrite(buffer
, 1, (size_t)len
, out
) < len
)
102 /* decode base64 MIME part */
104 mime_decode_base64(FILE* in
, FILE* out
, uschar
* boundary
)
106 uschar ibuf
[MIME_MAX_LINE_LENGTH
], obuf
[MIME_MAX_LINE_LENGTH
];
108 ssize_t len
, size
= 0;
113 while (Ufgets(ibuf
, MIME_MAX_LINE_LENGTH
, in
) != NULL
)
116 && Ustrncmp(ibuf
, "--", 2) == 0
117 && Ustrncmp((ibuf
+2), boundary
, Ustrlen(boundary
)) == 0
121 for (ipos
= ibuf
; *ipos
!= '\r' && *ipos
!= '\n' && *ipos
!= 0; ++ipos
)
123 if (*ipos
== '=') /* skip padding */
128 if (mime_b64
[*ipos
] == 128) /* skip bad characters */
130 mime_set_anomaly(MIME_ANOMALY_BROKEN_BASE64
);
134 /* simple state-machine */
135 switch((bytestate
++) & 3)
138 *opos
= mime_b64
[*ipos
] << 2;
141 *opos
|= mime_b64
[*ipos
] >> 4;
143 *opos
= mime_b64
[*ipos
] << 4;
146 *opos
|= mime_b64
[*ipos
] >> 2;
148 *opos
= mime_b64
[*ipos
] << 6;
151 *opos
|= mime_b64
[*ipos
];
157 /* something to write? */
161 if (fwrite(obuf
, 1, len
, out
) != len
) return -1; /* error */
163 /* copy incomplete last byte to start of obuf, where we continue */
164 if ((bytestate
& 3) != 0)
170 /* write out last byte if it was incomplete */
173 if (fwrite(obuf
, 1, 1, out
) != 1) return -1;
181 /* decode quoted-printable MIME part */
183 mime_decode_qp(FILE* in
, FILE* out
, uschar
* boundary
)
185 uschar ibuf
[MIME_MAX_LINE_LENGTH
], obuf
[MIME_MAX_LINE_LENGTH
];
187 ssize_t len
, size
= 0;
189 while (fgets(CS ibuf
, MIME_MAX_LINE_LENGTH
, in
) != NULL
)
192 && Ustrncmp(ibuf
, "--", 2) == 0
193 && Ustrncmp((ibuf
+2), boundary
, Ustrlen(boundary
)) == 0
195 break; /* todo: check for missing boundary */
204 int decode_qp_result
;
206 ipos
= mime_decode_qp_char(ipos
, &decode_qp_result
);
208 if (decode_qp_result
== -2)
210 /* Error from decoder. ipos is unchanged. */
211 mime_set_anomaly(MIME_ANOMALY_BROKEN_QP
);
216 else if (decode_qp_result
== -1)
218 else if (decode_qp_result
>= 0)
220 *opos
= decode_qp_result
;
231 /* something to write? */
235 if (fwrite(obuf
, 1, len
, out
) != len
) return -1; /* error */
244 mime_get_decode_file(uschar
*pname
, uschar
*fname
)
249 filename
= (uschar
*)malloc(2048);
253 (void)string_format(filename
, 2048, "%s/%s", pname
, fname
);
254 f
= modefopen(filename
,"wb+",SPOOL_MODE
);
257 f
= modefopen(fname
,"wb+",SPOOL_MODE
);
263 /* must find first free sequential filename */
267 (void)string_format(filename
, 2048,
268 "%s/%s-%05u", pname
, message_id
, file_nr
++);
272 result
= stat(CS filename
, &mystat
);
273 } while(result
!= -1);
275 f
= modefopen(filename
, "wb+", SPOOL_MODE
);
278 /* set expansion variable */
279 mime_decoded_filename
= filename
;
286 mime_decode(uschar
**listptr
)
289 uschar
*list
= *listptr
;
291 uschar option_buffer
[1024];
292 uschar decode_path
[1024];
293 FILE *decode_file
= NULL
;
295 ssize_t size_counter
= 0;
296 ssize_t (*decode_function
)(FILE*, FILE*, uschar
*);
298 if (mime_stream
== NULL
)
301 f_pos
= ftell(mime_stream
);
303 /* build default decode path (will exist since MBOX must be spooled up) */
304 (void)string_format(decode_path
,1024,"%s/scan/%s",spool_directory
,message_id
);
306 /* try to find 1st option */
307 if ((option
= string_nextinlist(&list
, &sep
,
309 sizeof(option_buffer
))) != NULL
)
311 /* parse 1st option */
312 if ( (Ustrcmp(option
,"false") == 0) || (Ustrcmp(option
,"0") == 0) )
313 /* explicitly no decoding */
316 if (Ustrcmp(option
,"default") == 0)
317 /* explicit default path + file names */
320 if (option
[0] == '/')
324 memset(&statbuf
,0,sizeof(statbuf
));
326 /* assume either path or path+file name */
327 if ( (stat(CS option
, &statbuf
) == 0) && S_ISDIR(statbuf
.st_mode
) )
328 /* is directory, use it as decode_path */
329 decode_file
= mime_get_decode_file(option
, NULL
);
331 /* does not exist or is a file, use as full file name */
332 decode_file
= mime_get_decode_file(NULL
, option
);
335 /* assume file name only, use default path */
336 decode_file
= mime_get_decode_file(decode_path
, option
);
340 /* no option? patch default path */
342 decode_file
= mime_get_decode_file(decode_path
, NULL
);
348 /* decode according to mime type */
350 !mime_content_transfer_encoding
351 ? mime_decode_asis
/* no encoding, dump as-is */
352 : Ustrcmp(mime_content_transfer_encoding
, "base64") == 0
354 : Ustrcmp(mime_content_transfer_encoding
, "quoted-printable") == 0
356 : mime_decode_asis
; /* unknown encoding type, just dump as-is */
358 size_counter
= decode_function(mime_stream
, decode_file
, mime_current_boundary
);
360 clearerr(mime_stream
);
361 fseek(mime_stream
, f_pos
, SEEK_SET
);
363 if (fclose(decode_file
) != 0 || size_counter
< 0)
366 /* round up to the next KiB */
367 mime_content_size
= (size_counter
+ 1023) / 1024;
373 mime_get_header(FILE *f
, uschar
*header
)
377 int header_value_mode
= 0;
378 int header_open_brackets
= 0;
383 if ((c
= fgetc(f
)) == EOF
) break;
385 /* always skip CRs */
386 if (c
== '\r') continue;
392 /* look if next char is '\t' or ' ' */
393 if ((c
= fgetc(f
)) == EOF
) break;
394 if ( (c
== '\t') || (c
== ' ') ) continue;
397 /* end of the header, terminate with ';' */
402 /* skip control characters */
403 if (c
< 32) continue;
405 if (header_value_mode
)
407 /* --------- value mode ----------- */
408 /* skip leading whitespace */
409 if ( ((c
== '\t') || (c
== ' ')) && (header_value_mode
== 1) )
412 /* we have hit a non-whitespace char, start copying value data */
413 header_value_mode
= 2;
415 if (c
== '"') /* flip "quoted" mode */
416 header_value_mode
= header_value_mode
==2 ? 3 : 2;
418 /* leave value mode on unquoted ';' */
419 if (header_value_mode
== 2 && c
== ';') {
420 header_value_mode
= 0;
422 /* -------------------------------- */
426 /* -------- non-value mode -------- */
427 /* skip whitespace + tabs */
428 if ( (c
== ' ') || (c
== '\t') )
432 /* quote next char. can be used
433 to escape brackets. */
434 if ((c
= fgetc(f
)) == EOF
) break;
438 header_open_brackets
++;
441 else if ((c
== ')') && header_open_brackets
)
443 header_open_brackets
--;
446 else if ( (c
== '=') && !header_open_brackets
) /* enter value mode */
447 header_value_mode
= 1;
449 /* skip chars while we are in a comment */
450 if (header_open_brackets
> 0)
452 /* -------------------------------- */
455 /* copy the char to the buffer */
456 header
[num_copied
++] = (uschar
)c
;
458 /* break if header buffer is full */
459 if (num_copied
> MIME_MAX_HEADER_SIZE
-1)
463 if ((num_copied
> 0) && (header
[num_copied
-1] != ';'))
464 header
[num_copied
-1] = ';';
467 header
[num_copied
] = '\0';
469 /* return 0 for EOF or empty line */
470 if ((c
== EOF
) || (num_copied
== 1))
478 mime_acl_check(uschar
*acl
, FILE *f
, struct mime_boundary_context
*context
,
479 uschar
**user_msgptr
, uschar
**log_msgptr
)
482 uschar
*header
= NULL
;
483 struct mime_boundary_context nested_context
;
485 /* reserve a line buffer to work in */
486 if (!(header
= (uschar
*)malloc(MIME_MAX_HEADER_SIZE
+1)))
488 log_write(0, LOG_PANIC
,
489 "MIME ACL: can't allocate %d bytes of memory.", MIME_MAX_HEADER_SIZE
+1);
493 /* Not actually used at the moment, but will be vital to fixing
494 * some RFC 2046 nonconformance later... */
495 nested_context
.parent
= context
;
497 /* loop through parts */
500 /* reset all per-part mime variables */
501 mime_anomaly_level
= 0;
502 mime_anomaly_text
= NULL
;
503 mime_boundary
= NULL
;
505 mime_decoded_filename
= NULL
;
506 mime_filename
= NULL
;
507 mime_content_description
= NULL
;
508 mime_content_disposition
= NULL
;
509 mime_content_id
= NULL
;
510 mime_content_transfer_encoding
= NULL
;
511 mime_content_type
= NULL
;
512 mime_is_multipart
= 0;
513 mime_content_size
= 0;
516 If boundary is null, we assume that *f is positioned on the start of headers (for example,
517 at the very beginning of a message.
518 If a boundary is given, we must first advance to it to reach the start of the next header
522 /* NOTE -- there's an error here -- RFC2046 specifically says to
523 * check for outer boundaries. This code doesn't do that, and
524 * I haven't fixed this.
526 * (I have moved partway towards adding support, however, by adding
527 * a "parent" field to my new boundary-context structure.)
531 while(fgets(CS header
, MIME_MAX_HEADER_SIZE
, f
))
533 /* boundary line must start with 2 dashes */
534 if ( Ustrncmp(header
, "--", 2) == 0
535 && Ustrncmp(header
+2, context
->boundary
, Ustrlen(context
->boundary
)) == 0)
538 if (Ustrncmp((header
+2+Ustrlen(context
->boundary
)), "--", 2) == 0)
540 /* END boundary found */
541 debug_printf("End boundary found %s\n", context
->boundary
);
545 debug_printf("Next part with boundary %s\n", context
->boundary
);
547 /* can't use break here */
551 /* Hit EOF or read error. Ugh. */
552 debug_printf("Hit EOF ...\n");
557 /* parse headers, set up expansion variables */
558 while (mime_get_header(f
, header
))
561 /* loop through header list */
562 for (i
= 0; i
< mime_header_list_size
; i
++)
563 if (strncmpic(mime_header_list
[i
].name
,
564 header
, mime_header_list
[i
].namelen
) == 0)
565 { /* found an interesting header */
566 uschar
* header_value
;
567 int header_value_len
;
568 uschar
* p
= header
+ mime_header_list
[i
].namelen
;
570 /* grab the value (normalize to lower case)
571 and copy to its corresponding expansion variable */
577 header_value_len
= p
- (header
+ mime_header_list
[i
].namelen
);
578 p
= header
+ mime_header_list
[i
].namelen
;
579 header_value
= string_copyn(p
, header_value_len
);
580 debug_printf("Found %s MIME header, value is '%s'\n",
581 mime_header_list
[i
].name
, header_value
);
582 *((uschar
**)(mime_header_list
[i
].value
)) = header_value
;
584 /* make p point to the next character after the closing ';' */
585 p
+= header_value_len
+1;
587 /* grab all param=value tags on the remaining line,
588 check if they are interesting */
592 /* debug_printf(" considering paramlist '%s'\n", p); */
594 for (mp
= mime_parameter_list
;
595 mp
< &mime_parameter_list
[mime_parameter_list_size
];
598 uschar
* param_value
= NULL
;
600 /* found an interesting parameter? */
601 if (strncmpic(mp
->name
, p
, mp
->namelen
) == 0)
606 /* yes, grab the value and copy to its corresponding expansion variable */
608 while(*p
&& *p
!= ';') /* ; terminates */
611 p
++; /* skip leading " */
612 while(*p
&& *p
!= '"') /* " protects ; */
613 param_value
= string_cat(param_value
, &size
, &ptr
, p
++, 1);
614 if (*p
) p
++; /* skip trailing " */
617 param_value
= string_cat(param_value
, &size
, &ptr
, p
++, 1);
618 if (*p
) p
++; /* skip trailing ; */
623 param_value
[ptr
++] = '\0';
625 param_value
= rfc2047_decode(param_value
,
626 check_rfc2047_length
, NULL
, 32, NULL
, &dummy
);
627 debug_printf(" Found %s MIME parameter in %s header, "
628 "value is '%s'\n", mp
->name
, mime_header_list
[i
].name
,
631 *mp
->value
= param_value
;
632 goto NEXT_PARAM_SEARCH
;
635 /* There is something, but not one of our interesting parameters.
636 Advance to the next unquoted semicolon */
637 while(*p
&& *p
!= ';')
640 while(*++p
&& *p
!= '"') ;
650 /* set additional flag variables (easier access) */
651 if ( (mime_content_type
!= NULL
) &&
652 (Ustrncmp(mime_content_type
,"multipart",9) == 0) )
653 mime_is_multipart
= 1;
655 /* Make a copy of the boundary pointer.
656 Required since mime_boundary is global
657 and can be overwritten further down in recursion */
658 nested_context
.boundary
= mime_boundary
;
660 /* raise global counter */
663 /* copy current file handle to global variable */
665 mime_current_boundary
= context
? context
->boundary
: 0;
667 /* Note the context */
668 mime_is_coverletter
= !(context
&& context
->context
== MBC_ATTACHMENT
);
670 /* call ACL handling function */
671 rc
= acl_check(ACL_WHERE_MIME
, NULL
, acl
, user_msgptr
, log_msgptr
);
674 mime_current_boundary
= NULL
;
678 /* If we have a multipart entity and a boundary, go recursive */
679 if ( (mime_content_type
!= NULL
) &&
680 (nested_context
.boundary
!= NULL
) &&
681 (Ustrncmp(mime_content_type
,"multipart",9) == 0) )
683 debug_printf("Entering multipart recursion, boundary '%s'\n", nested_context
.boundary
);
685 nested_context
.context
=
686 context
&& context
->context
== MBC_ATTACHMENT
688 : Ustrcmp(mime_content_type
,"multipart/alternative") == 0
689 || Ustrcmp(mime_content_type
,"multipart/related") == 0
690 ? MBC_COVERLETTER_ALL
691 : MBC_COVERLETTER_ONESHOT
;
693 rc
= mime_acl_check(acl
, f
, &nested_context
, user_msgptr
, log_msgptr
);
696 else if ( (mime_content_type
!= NULL
) &&
697 (Ustrncmp(mime_content_type
,"message/rfc822",14) == 0) )
699 uschar
*rfc822name
= NULL
;
700 uschar filename
[2048];
704 /* must find first free sequential filename */
708 (void)string_format(filename
, 2048,
709 "%s/scan/%s/__rfc822_%05u", spool_directory
, message_id
, file_nr
++);
713 result
= stat(CS filename
,&mystat
);
714 } while (result
!= -1);
716 rfc822name
= filename
;
718 /* decode RFC822 attachment */
719 mime_decoded_filename
= NULL
;
721 mime_current_boundary
= context
? context
->boundary
: NULL
;
722 mime_decode(&rfc822name
);
724 mime_current_boundary
= NULL
;
725 if (!mime_decoded_filename
) /* decoding failed */
727 log_write(0, LOG_MAIN
,
728 "mime_regex acl condition warning - could not decode RFC822 MIME part to file.");
731 mime_decoded_filename
= NULL
;
735 /* If the boundary of this instance is NULL, we are finished here */
736 if (context
== NULL
) break;
738 if (context
->context
== MBC_COVERLETTER_ONESHOT
)
739 context
->context
= MBC_ATTACHMENT
;
745 #endif /*WITH_CONTENT_SCAN*/