1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
5 /* Copyright (c) Tom Kistner <tom@duncanthrax.net> 2004, 2015 */
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(const uschar **listptr)
289 const 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;
374 mime_get_header(FILE *f, uschar *header)
378 int header_value_mode = 0;
379 int header_open_brackets = 0;
384 if ((c = fgetc(f)) == EOF) break;
386 /* always skip CRs */
387 if (c == '\r') continue;
393 /* look if next char is '\t' or ' ' */
394 if ((c = fgetc(f)) == EOF) break;
395 if ( (c == '\t') || (c == ' ') ) continue;
398 /* end of the header, terminate with ';' */
403 /* skip control characters */
404 if (c < 32) continue;
406 if (header_value_mode)
408 /* --------- value mode ----------- */
409 /* skip leading whitespace */
410 if ( ((c == '\t') || (c == ' ')) && (header_value_mode == 1) )
413 /* we have hit a non-whitespace char, start copying value data */
414 header_value_mode = 2;
416 if (c == '"') /* flip "quoted" mode */
417 header_value_mode = header_value_mode==2 ? 3 : 2;
419 /* leave value mode on unquoted ';' */
420 if (header_value_mode == 2 && c == ';') {
421 header_value_mode = 0;
423 /* -------------------------------- */
427 /* -------- non-value mode -------- */
428 /* skip whitespace + tabs */
429 if ( (c == ' ') || (c == '\t') )
433 /* quote next char. can be used
434 to escape brackets. */
435 if ((c = fgetc(f)) == EOF) break;
439 header_open_brackets++;
442 else if ((c == ')') && header_open_brackets)
444 header_open_brackets--;
447 else if ( (c == '=') && !header_open_brackets ) /* enter value mode */
448 header_value_mode = 1;
450 /* skip chars while we are in a comment */
451 if (header_open_brackets > 0)
453 /* -------------------------------- */
456 /* copy the char to the buffer */
457 header[num_copied++] = (uschar)c;
459 /* break if header buffer is full */
460 if (num_copied > MIME_MAX_HEADER_SIZE-1)
464 if ((num_copied > 0) && (header[num_copied-1] != ';'))
465 header[num_copied-1] = ';';
468 header[num_copied] = '\0';
470 /* return 0 for EOF or empty line */
471 if ((c == EOF) || (num_copied == 1))
479 mime_vars_reset(void)
481 mime_anomaly_level = 0;
482 mime_anomaly_text = NULL;
483 mime_boundary = NULL;
485 mime_decoded_filename = NULL;
486 mime_filename = NULL;
487 mime_content_description = NULL;
488 mime_content_disposition = NULL;
489 mime_content_id = NULL;
490 mime_content_transfer_encoding = NULL;
491 mime_content_type = NULL;
492 mime_is_multipart = 0;
493 mime_content_size = 0;
497 /* Grab a parameter value, dealing with quoting.
500 str Input string. Updated on return to point to terminating ; or NUL
503 Allocated string with parameter value
506 mime_param_val(uschar ** sp)
510 int size = 0, ptr = 0;
512 /* debug_printf(" considering paramval '%s'\n", s); */
514 while (*s && *s != ';') /* ; terminates */
517 s++; /* skip opening " */
518 while (*s && *s != '"') /* " protects ; */
519 val = string_cat(val, &size, &ptr, s++, 1);
520 if (*s) s++; /* skip closing " */
523 val = string_cat(val, &size, &ptr, s++, 1);
524 if (val) val[ptr] = '\0';
530 mime_next_semicolon(uschar * s)
532 while (*s && *s != ';') /* ; terminates */
535 s++; /* skip opening " */
536 while (*s && *s != '"') /* " protects ; */
538 if (*s) s++; /* skip closing " */
547 rfc2231_to_2047(const uschar * fname, const uschar * charset, int * len)
549 int size = 0, ptr = 0;
550 uschar * val = string_cat(NULL, &size, &ptr, US"=?", 2);
554 val = string_cat(val, &size, &ptr, charset, Ustrlen(charset));
555 val = string_cat(val, &size, &ptr, US"?Q?", 3);
558 if (c == '%' && isxdigit(fname[1]) && isxdigit(fname[2]))
560 val = string_cat(val, &size, &ptr, US"=", 1);
561 val = string_cat(val, &size, &ptr, ++fname, 2);
565 val = string_cat(val, &size, &ptr, fname++, 1);
567 val = string_cat(val, &size, &ptr, US"?=", 2);
568 val[*len = ptr] = '\0';
574 mime_acl_check(uschar *acl, FILE *f, struct mime_boundary_context *context,
575 uschar **user_msgptr, uschar **log_msgptr)
578 uschar * header = NULL;
579 struct mime_boundary_context nested_context;
581 /* reserve a line buffer to work in */
582 header = store_get(MIME_MAX_HEADER_SIZE+1);
584 /* Not actually used at the moment, but will be vital to fixing
585 * some RFC 2046 nonconformance later... */
586 nested_context.parent = context;
588 /* loop through parts */
591 /* reset all per-part mime variables */
594 /* If boundary is null, we assume that *f is positioned on the start of
595 headers (for example, at the very beginning of a message. If a boundary is
596 given, we must first advance to it to reach the start of the next header
599 /* NOTE -- there's an error here -- RFC2046 specifically says to
600 * check for outer boundaries. This code doesn't do that, and
601 * I haven't fixed this.
603 * (I have moved partway towards adding support, however, by adding
604 * a "parent" field to my new boundary-context structure.)
606 if (context) for (;;)
608 if (!fgets(CS header, MIME_MAX_HEADER_SIZE, f))
610 /* Hit EOF or read error. Ugh. */
611 DEBUG(D_acl) debug_printf("MIME: Hit EOF ...\n");
615 /* boundary line must start with 2 dashes */
616 if ( Ustrncmp(header, "--", 2) == 0
617 && Ustrncmp(header+2, context->boundary, Ustrlen(context->boundary)) == 0
619 { /* found boundary */
620 if (Ustrncmp((header+2+Ustrlen(context->boundary)), "--", 2) == 0)
622 /* END boundary found */
623 DEBUG(D_acl) debug_printf("MIME: End boundary found %s\n",
628 DEBUG(D_acl) debug_printf("MIME: Next part with boundary %s\n",
634 /* parse headers, set up expansion variables */
635 while (mime_get_header(f, header))
637 struct mime_header * mh;
639 /* look for interesting headers */
640 for (mh = mime_header_list;
641 mh < mime_header_list + mime_header_list_size;
642 mh++) if (strncmpic(mh->name, header, mh->namelen) == 0)
644 uschar * p = header + mh->namelen;
647 /* grab the value (normalize to lower case)
648 and copy to its corresponding expansion variable */
650 for (q = p; *q != ';' && *q; q++) ;
651 *mh->value = string_copynlc(p, q-p);
652 DEBUG(D_acl) debug_printf("MIME: found %s header, value is '%s'\n",
653 mh->name, *mh->value);
655 if (*(p = q)) p++; /* jump past the ; */
658 uschar * mime_fname = NULL;
659 uschar * mime_fname_rfc2231 = NULL;
660 uschar * mime_filename_charset = NULL;
661 BOOL decoding_failed = FALSE;
663 /* grab all param=value tags on the remaining line,
664 check if they are interesting */
670 DEBUG(D_acl) debug_printf("MIME: considering paramlist '%s'\n", p);
673 && strncmpic(CUS"content-disposition:", header, 20) == 0
674 && strncmpic(CUS"filename*", p, 9) == 0
676 { /* RFC 2231 filename */
679 /* find value of the filename */
681 while(*p != '=' && *p) p++;
682 if (*p) p++; /* p is filename or NUL */
683 q = mime_param_val(&p); /* p now trailing ; or NUL */
687 uschar * temp_string, * err_msg;
690 /* build up an un-decoded filename over successive
691 filename*= parameters (for use when 2047 decode fails) */
693 mime_fname_rfc2231 = string_sprintf("%#s%s",
694 mime_fname_rfc2231, q);
696 if (!decoding_failed)
699 if (!mime_filename_charset)
703 /* look for a ' in the "filename" */
704 while(*s != '\'' && *s) s++; /* s is 1st ' or NUL */
706 if ((size = s-q) > 0)
707 mime_filename_charset = string_copyn(q, size);
710 while(*p == '\'') p++; /* p is after 2nd ' */
715 DEBUG(D_acl) debug_printf("MIME: charset %s fname '%s'\n",
716 mime_filename_charset ? mime_filename_charset : US"<NULL>", p);
718 temp_string = rfc2231_to_2047(p, mime_filename_charset, &slen);
719 DEBUG(D_acl) debug_printf("MIME: 2047-name %s\n", temp_string);
721 temp_string = rfc2047_decode(temp_string, FALSE, NULL, ' ',
723 DEBUG(D_acl) debug_printf("MIME: plain-name %s\n", temp_string);
725 size = Ustrlen(temp_string);
728 decoding_failed = TRUE;
730 /* build up a decoded filename over successive
731 filename*= parameters */
733 mime_filename = mime_fname = mime_fname
734 ? string_sprintf("%s%s", mime_fname, temp_string)
741 /* look for interesting parameters */
742 for (mp = mime_parameter_list;
743 mp < mime_parameter_list + nelem(mime_parameter_list);
745 ) if (strncmpic(mp->name, p, mp->namelen) == 0)
748 uschar * dummy_errstr;
750 /* grab the value and copy to its expansion variable */
752 q = mime_param_val(&p); /* p now trailing ; or NUL */
755 ? rfc2047_decode(q, check_rfc2047_length, NULL, 32, NULL,
758 DEBUG(D_acl) debug_printf(
759 "MIME: found %s parameter in %s header, value '%s'\n",
760 mp->name, mh->name, *mp->value);
762 break; /* done matching param names */
766 /* There is something, but not one of our interesting parameters.
767 Advance past the next semicolon */
768 p = mime_next_semicolon(p);
770 } /* param scan on line */
772 if (strncmpic(CUS"content-disposition:", header, 20) == 0)
774 if (decoding_failed) mime_filename = mime_fname_rfc2231;
776 DEBUG(D_acl) debug_printf(
777 "MIME: found %s parameter in %s header, value is '%s'\n",
778 "filename", mh->name, mime_filename);
784 /* set additional flag variables (easier access) */
785 if ( mime_content_type
786 && Ustrncmp(mime_content_type,"multipart",9) == 0
788 mime_is_multipart = 1;
790 /* Make a copy of the boundary pointer.
791 Required since mime_boundary is global
792 and can be overwritten further down in recursion */
793 nested_context.boundary = mime_boundary;
795 /* raise global counter */
798 /* copy current file handle to global variable */
800 mime_current_boundary = context ? context->boundary : 0;
802 /* Note the context */
803 mime_is_coverletter = !(context && context->context == MBC_ATTACHMENT);
805 /* call ACL handling function */
806 rc = acl_check(ACL_WHERE_MIME, NULL, acl, user_msgptr, log_msgptr);
809 mime_current_boundary = NULL;
813 /* If we have a multipart entity and a boundary, go recursive */
814 if ( (mime_content_type != NULL) &&
815 (nested_context.boundary != NULL) &&
816 (Ustrncmp(mime_content_type,"multipart",9) == 0) )
819 debug_printf("MIME: Entering multipart recursion, boundary '%s'\n",
820 nested_context.boundary);
822 nested_context.context =
823 context && context->context == MBC_ATTACHMENT
825 : Ustrcmp(mime_content_type,"multipart/alternative") == 0
826 || Ustrcmp(mime_content_type,"multipart/related") == 0
827 ? MBC_COVERLETTER_ALL
828 : MBC_COVERLETTER_ONESHOT;
830 rc = mime_acl_check(acl, f, &nested_context, user_msgptr, log_msgptr);
833 else if ( (mime_content_type != NULL) &&
834 (Ustrncmp(mime_content_type,"message/rfc822",14) == 0) )
836 const uschar *rfc822name = NULL;
837 uschar filename[2048];
841 /* must find first free sequential filename */
845 (void)string_format(filename, 2048,
846 "%s/scan/%s/__rfc822_%05u", spool_directory, message_id, file_nr++);
850 result = stat(CS filename,&mystat);
851 } while (result != -1);
853 rfc822name = filename;
855 /* decode RFC822 attachment */
856 mime_decoded_filename = NULL;
858 mime_current_boundary = context ? context->boundary : NULL;
859 mime_decode(&rfc822name);
861 mime_current_boundary = NULL;
862 if (!mime_decoded_filename) /* decoding failed */
864 log_write(0, LOG_MAIN,
865 "mime_regex acl condition warning - could not decode RFC822 MIME part to file.");
869 mime_decoded_filename = NULL;
873 /* If the boundary of this instance is NULL, we are finished here */
876 if (context->context == MBC_COVERLETTER_ONESHOT)
877 context->context = MBC_ATTACHMENT;
885 #endif /*WITH_CONTENT_SCAN*/