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 2016
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 { US"content-type:", 13, &mime_content_type },
20 { US"content-disposition:", 20, &mime_content_disposition },
21 { US"content-transfer-encoding:", 26, &mime_content_transfer_encoding },
22 { US"content-id:", 11, &mime_content_id },
23 { US"content-description:", 20, &mime_content_description }
26 static int mime_header_list_size = nelem(mime_header_list);
28 static mime_parameter mime_parameter_list[] = {
29 { US"name=", 5, &mime_filename },
30 { US"filename=", 9, &mime_filename },
31 { US"charset=", 8, &mime_charset },
32 { US"boundary=", 9, &mime_boundary }
36 /*************************************************
37 * set MIME anomaly level + text *
38 *************************************************/
40 /* Small wrapper to set the two expandables which
41 give info on detected "problems" in MIME
42 encodings. Indexes are defined in mime.h. */
45 mime_set_anomaly(int idx)
50 } anom[] = { {1, CUS"Broken Quoted-Printable encoding detected"},
51 {2, CUS"Broken BASE64 encoding detected"} };
53 mime_anomaly_level = anom[idx].level;
54 mime_anomaly_text = anom[idx].text;
58 /*************************************************
59 * decode quoted-printable chars *
60 *************************************************/
62 /* gets called when we hit a =
63 returns: new pointer position
66 -1 - soft line break, no char
71 mime_decode_qp_char(uschar *qp_p, int *c)
73 uschar *initial_pos = qp_p;
75 /* advance one char */
78 /* Check for two hex digits and decode them */
79 if (isxdigit(*qp_p) && isxdigit(qp_p[1]))
81 /* Do hex conversion */
82 *c = (isdigit(*qp_p) ? *qp_p - '0' : toupper(*qp_p) - 'A' + 10) <<4;
84 *c |= isdigit(*qp_p) ? *qp_p - '0' : toupper(*qp_p) - 'A' + 10;
88 /* tab or whitespace may follow just ignore it if it precedes \n */
89 while (*qp_p == '\t' || *qp_p == ' ' || *qp_p == '\r')
92 if (*qp_p == '\n') /* hit soft line break */
98 /* illegal char here */
104 /* just dump MIME part without any decoding */
106 mime_decode_asis(FILE* in, FILE* out, uschar* boundary)
108 ssize_t len, size = 0;
109 uschar buffer[MIME_MAX_LINE_LENGTH];
111 while(fgets(CS buffer, MIME_MAX_LINE_LENGTH, mime_stream) != NULL)
114 && Ustrncmp(buffer, "--", 2) == 0
115 && Ustrncmp((buffer+2), boundary, Ustrlen(boundary)) == 0
119 len = Ustrlen(buffer);
120 if (fwrite(buffer, 1, (size_t)len, out) < len)
129 /* decode quoted-printable MIME part */
131 mime_decode_qp(FILE* in, FILE* out, uschar* boundary)
133 uschar ibuf[MIME_MAX_LINE_LENGTH], obuf[MIME_MAX_LINE_LENGTH];
135 ssize_t len, size = 0;
137 while (fgets(CS ibuf, MIME_MAX_LINE_LENGTH, in) != NULL)
140 && Ustrncmp(ibuf, "--", 2) == 0
141 && Ustrncmp((ibuf+2), boundary, Ustrlen(boundary)) == 0
143 break; /* todo: check for missing boundary */
152 int decode_qp_result;
154 ipos = mime_decode_qp_char(ipos, &decode_qp_result);
156 if (decode_qp_result == -2)
158 /* Error from decoder. ipos is unchanged. */
159 mime_set_anomaly(MIME_ANOMALY_BROKEN_QP);
163 else if (decode_qp_result == -1)
165 else if (decode_qp_result >= 0)
166 *opos++ = decode_qp_result;
171 /* something to write? */
175 if (fwrite(obuf, 1, len, out) != len) return -1; /* error */
184 * Return open filehandle for combo of path and file.
185 * Side-effect: set mime_decoded_filename, to copy in allocated mem
188 mime_get_decode_file(uschar *pname, uschar *fname)
191 mime_decoded_filename = string_sprintf("%s/%s", pname, fname);
193 mime_decoded_filename = string_copy(fname);
199 /* must find first free sequential filename */
203 mime_decoded_filename = string_sprintf("%s/%s-%05u", pname, message_id, file_nr++);
207 result = stat(CS mime_decoded_filename, &mystat);
208 } while(result != -1);
211 return modefopen(mime_decoded_filename, "wb+", SPOOL_MODE);
216 mime_decode(const uschar **listptr)
219 const uschar *list = *listptr;
221 uschar option_buffer[1024];
222 uschar decode_path[1024];
223 FILE *decode_file = NULL;
225 ssize_t size_counter = 0;
226 ssize_t (*decode_function)(FILE*, FILE*, uschar*);
228 if (!mime_stream || (f_pos = ftell(mime_stream)) < 0)
231 /* build default decode path (will exist since MBOX must be spooled up) */
232 (void)string_format(decode_path,1024,"%s/scan/%s",spool_directory,message_id);
234 /* try to find 1st option */
235 if ((option = string_nextinlist(&list, &sep,
237 sizeof(option_buffer))) != NULL)
239 /* parse 1st option */
240 if ((Ustrcmp(option,"false") == 0) || (Ustrcmp(option,"0") == 0))
241 /* explicitly no decoding */
244 if (Ustrcmp(option,"default") == 0)
245 /* explicit default path + file names */
248 if (option[0] == '/')
252 memset(&statbuf,0,sizeof(statbuf));
254 /* assume either path or path+file name */
255 if ( (stat(CS option, &statbuf) == 0) && S_ISDIR(statbuf.st_mode) )
256 /* is directory, use it as decode_path */
257 decode_file = mime_get_decode_file(option, NULL);
259 /* does not exist or is a file, use as full file name */
260 decode_file = mime_get_decode_file(NULL, option);
263 /* assume file name only, use default path */
264 decode_file = mime_get_decode_file(decode_path, option);
268 /* no option? patch default path */
270 decode_file = mime_get_decode_file(decode_path, NULL);
276 /* decode according to mime type */
278 !mime_content_transfer_encoding
279 ? mime_decode_asis /* no encoding, dump as-is */
280 : Ustrcmp(mime_content_transfer_encoding, "base64") == 0
282 : Ustrcmp(mime_content_transfer_encoding, "quoted-printable") == 0
284 : mime_decode_asis; /* unknown encoding type, just dump as-is */
286 size_counter = decode_function(mime_stream, decode_file, mime_current_boundary);
288 clearerr(mime_stream);
289 if (fseek(mime_stream, f_pos, SEEK_SET))
292 if (fclose(decode_file) != 0 || size_counter < 0)
295 /* round up to the next KiB */
296 mime_content_size = (size_counter + 1023) / 1024;
303 mime_get_header(FILE *f, uschar *header)
307 int header_value_mode = 0;
308 int header_open_brackets = 0;
313 if ((c = fgetc(f)) == EOF) break;
315 /* always skip CRs */
316 if (c == '\r') continue;
322 /* look if next char is '\t' or ' ' */
323 if ((c = fgetc(f)) == EOF) break;
324 if ( (c == '\t') || (c == ' ') ) continue;
327 /* end of the header, terminate with ';' */
332 /* skip control characters */
333 if (c < 32) continue;
335 if (header_value_mode)
337 /* --------- value mode ----------- */
338 /* skip leading whitespace */
339 if ( ((c == '\t') || (c == ' ')) && (header_value_mode == 1) )
342 /* we have hit a non-whitespace char, start copying value data */
343 header_value_mode = 2;
345 if (c == '"') /* flip "quoted" mode */
346 header_value_mode = header_value_mode==2 ? 3 : 2;
348 /* leave value mode on unquoted ';' */
349 if (header_value_mode == 2 && c == ';') {
350 header_value_mode = 0;
352 /* -------------------------------- */
356 /* -------- non-value mode -------- */
357 /* skip whitespace + tabs */
358 if ( (c == ' ') || (c == '\t') )
362 /* quote next char. can be used
363 to escape brackets. */
364 if ((c = fgetc(f)) == EOF) break;
368 header_open_brackets++;
371 else if ((c == ')') && header_open_brackets)
373 header_open_brackets--;
376 else if ( (c == '=') && !header_open_brackets ) /* enter value mode */
377 header_value_mode = 1;
379 /* skip chars while we are in a comment */
380 if (header_open_brackets > 0)
382 /* -------------------------------- */
385 /* copy the char to the buffer */
386 header[num_copied++] = (uschar)c;
388 /* break if header buffer is full */
389 if (num_copied > MIME_MAX_HEADER_SIZE-1)
393 if ((num_copied > 0) && (header[num_copied-1] != ';'))
394 header[num_copied-1] = ';';
397 header[num_copied] = '\0';
399 /* return 0 for EOF or empty line */
400 if ((c == EOF) || (num_copied == 1))
408 mime_vars_reset(void)
410 mime_anomaly_level = 0;
411 mime_anomaly_text = NULL;
412 mime_boundary = NULL;
414 mime_decoded_filename = NULL;
415 mime_filename = NULL;
416 mime_content_description = NULL;
417 mime_content_disposition = NULL;
418 mime_content_id = NULL;
419 mime_content_transfer_encoding = NULL;
420 mime_content_type = NULL;
421 mime_is_multipart = 0;
422 mime_content_size = 0;
426 /* Grab a parameter value, dealing with quoting.
429 str Input string. Updated on return to point to terminating ; or NUL
432 Allocated string with parameter value
435 mime_param_val(uschar ** sp)
439 int size = 0, ptr = 0;
441 /* debug_printf_indent(" considering paramval '%s'\n", s); */
443 while (*s && *s != ';') /* ; terminates */
446 s++; /* skip opening " */
447 while (*s && *s != '"') /* " protects ; */
448 val = string_catn(val, &size, &ptr, s++, 1);
449 if (*s) s++; /* skip closing " */
452 val = string_catn(val, &size, &ptr, s++, 1);
453 if (val) val[ptr] = '\0';
459 mime_next_semicolon(uschar * s)
461 while (*s && *s != ';') /* ; terminates */
464 s++; /* skip opening " */
465 while (*s && *s != '"') /* " protects ; */
467 if (*s) s++; /* skip closing " */
476 rfc2231_to_2047(const uschar * fname, const uschar * charset, int * len)
478 int size = 0, ptr = 0;
479 uschar * val = string_catn(NULL, &size, &ptr, US"=?", 2);
483 val = string_cat(val, &size, &ptr, charset);
484 val = string_catn(val, &size, &ptr, US"?Q?", 3);
487 if (c == '%' && isxdigit(fname[1]) && isxdigit(fname[2]))
489 val = string_catn(val, &size, &ptr, US"=", 1);
490 val = string_catn(val, &size, &ptr, ++fname, 2);
494 val = string_catn(val, &size, &ptr, fname++, 1);
496 val = string_catn(val, &size, &ptr, US"?=", 2);
497 val[*len = ptr] = '\0';
503 mime_acl_check(uschar *acl, FILE *f, struct mime_boundary_context *context,
504 uschar **user_msgptr, uschar **log_msgptr)
507 uschar * header = NULL;
508 struct mime_boundary_context nested_context;
510 /* reserve a line buffer to work in */
511 header = store_get(MIME_MAX_HEADER_SIZE+1);
513 /* Not actually used at the moment, but will be vital to fixing
514 * some RFC 2046 nonconformance later... */
515 nested_context.parent = context;
517 /* loop through parts */
520 /* reset all per-part mime variables */
523 /* If boundary is null, we assume that *f is positioned on the start of
524 headers (for example, at the very beginning of a message. If a boundary is
525 given, we must first advance to it to reach the start of the next header
528 /* NOTE -- there's an error here -- RFC2046 specifically says to
529 * check for outer boundaries. This code doesn't do that, and
530 * I haven't fixed this.
532 * (I have moved partway towards adding support, however, by adding
533 * a "parent" field to my new boundary-context structure.)
535 if (context) for (;;)
537 if (!fgets(CS header, MIME_MAX_HEADER_SIZE, f))
539 /* Hit EOF or read error. Ugh. */
540 DEBUG(D_acl) debug_printf_indent("MIME: Hit EOF ...\n");
544 /* boundary line must start with 2 dashes */
545 if ( Ustrncmp(header, "--", 2) == 0
546 && Ustrncmp(header+2, context->boundary, Ustrlen(context->boundary)) == 0
548 { /* found boundary */
549 if (Ustrncmp((header+2+Ustrlen(context->boundary)), "--", 2) == 0)
551 /* END boundary found */
552 DEBUG(D_acl) debug_printf_indent("MIME: End boundary found %s\n",
557 DEBUG(D_acl) debug_printf_indent("MIME: Next part with boundary %s\n",
563 /* parse headers, set up expansion variables */
564 while (mime_get_header(f, header))
566 struct mime_header * mh;
568 /* look for interesting headers */
569 for (mh = mime_header_list;
570 mh < mime_header_list + mime_header_list_size;
571 mh++) if (strncmpic(mh->name, header, mh->namelen) == 0)
573 uschar * p = header + mh->namelen;
576 /* grab the value (normalize to lower case)
577 and copy to its corresponding expansion variable */
579 for (q = p; *q != ';' && *q; q++) ;
580 *mh->value = string_copynlc(p, q-p);
581 DEBUG(D_acl) debug_printf_indent("MIME: found %s header, value is '%s'\n",
582 mh->name, *mh->value);
584 if (*(p = q)) p++; /* jump past the ; */
587 uschar * mime_fname = NULL;
588 uschar * mime_fname_rfc2231 = NULL;
589 uschar * mime_filename_charset = NULL;
590 BOOL decoding_failed = FALSE;
592 /* grab all param=value tags on the remaining line,
593 check if they are interesting */
599 DEBUG(D_acl) debug_printf_indent("MIME: considering paramlist '%s'\n", p);
602 && strncmpic(CUS"content-disposition:", header, 20) == 0
603 && strncmpic(CUS"filename*", p, 9) == 0
605 { /* RFC 2231 filename */
608 /* find value of the filename */
610 while(*p != '=' && *p) p++;
611 if (*p) p++; /* p is filename or NUL */
612 q = mime_param_val(&p); /* p now trailing ; or NUL */
616 uschar * temp_string, * err_msg;
619 /* build up an un-decoded filename over successive
620 filename*= parameters (for use when 2047 decode fails) */
622 mime_fname_rfc2231 = string_sprintf("%#s%s",
623 mime_fname_rfc2231, q);
625 if (!decoding_failed)
628 if (!mime_filename_charset)
632 /* look for a ' in the "filename" */
633 while(*s != '\'' && *s) s++; /* s is 1st ' or NUL */
635 if ((size = s-q) > 0)
636 mime_filename_charset = string_copyn(q, size);
639 while(*p == '\'') p++; /* p is after 2nd ' */
644 DEBUG(D_acl) debug_printf_indent("MIME: charset %s fname '%s'\n",
645 mime_filename_charset ? mime_filename_charset : US"<NULL>", p);
647 temp_string = rfc2231_to_2047(p, mime_filename_charset, &slen);
648 DEBUG(D_acl) debug_printf_indent("MIME: 2047-name %s\n", temp_string);
650 temp_string = rfc2047_decode(temp_string, FALSE, NULL, ' ',
652 DEBUG(D_acl) debug_printf_indent("MIME: plain-name %s\n", temp_string);
654 size = Ustrlen(temp_string);
657 decoding_failed = TRUE;
659 /* build up a decoded filename over successive
660 filename*= parameters */
662 mime_filename = mime_fname = mime_fname
663 ? string_sprintf("%s%s", mime_fname, temp_string)
670 /* look for interesting parameters */
671 for (mp = mime_parameter_list;
672 mp < mime_parameter_list + nelem(mime_parameter_list);
674 ) if (strncmpic(mp->name, p, mp->namelen) == 0)
677 uschar * dummy_errstr;
679 /* grab the value and copy to its expansion variable */
681 q = mime_param_val(&p); /* p now trailing ; or NUL */
684 ? rfc2047_decode(q, check_rfc2047_length, NULL, 32, NULL,
687 DEBUG(D_acl) debug_printf_indent(
688 "MIME: found %s parameter in %s header, value '%s'\n",
689 mp->name, mh->name, *mp->value);
691 break; /* done matching param names */
695 /* There is something, but not one of our interesting parameters.
696 Advance past the next semicolon */
697 p = mime_next_semicolon(p);
699 } /* param scan on line */
701 if (strncmpic(CUS"content-disposition:", header, 20) == 0)
703 if (decoding_failed) mime_filename = mime_fname_rfc2231;
705 DEBUG(D_acl) debug_printf_indent(
706 "MIME: found %s parameter in %s header, value is '%s'\n",
707 "filename", mh->name, mime_filename);
713 /* set additional flag variables (easier access) */
714 if ( mime_content_type
715 && Ustrncmp(mime_content_type,"multipart",9) == 0
717 mime_is_multipart = 1;
719 /* Make a copy of the boundary pointer.
720 Required since mime_boundary is global
721 and can be overwritten further down in recursion */
722 nested_context.boundary = mime_boundary;
724 /* raise global counter */
727 /* copy current file handle to global variable */
729 mime_current_boundary = context ? context->boundary : 0;
731 /* Note the context */
732 mime_is_coverletter = !(context && context->context == MBC_ATTACHMENT);
734 /* call ACL handling function */
735 rc = acl_check(ACL_WHERE_MIME, NULL, acl, user_msgptr, log_msgptr);
738 mime_current_boundary = NULL;
742 /* If we have a multipart entity and a boundary, go recursive */
743 if ( (mime_content_type != NULL) &&
744 (nested_context.boundary != NULL) &&
745 (Ustrncmp(mime_content_type,"multipart",9) == 0) )
748 debug_printf_indent("MIME: Entering multipart recursion, boundary '%s'\n",
749 nested_context.boundary);
751 nested_context.context =
752 context && context->context == MBC_ATTACHMENT
754 : Ustrcmp(mime_content_type,"multipart/alternative") == 0
755 || Ustrcmp(mime_content_type,"multipart/related") == 0
756 ? MBC_COVERLETTER_ALL
757 : MBC_COVERLETTER_ONESHOT;
759 rc = mime_acl_check(acl, f, &nested_context, user_msgptr, log_msgptr);
762 else if ( (mime_content_type != NULL) &&
763 (Ustrncmp(mime_content_type,"message/rfc822",14) == 0) )
765 const uschar *rfc822name = NULL;
766 uschar filename[2048];
770 /* must find first free sequential filename */
774 (void)string_format(filename, 2048,
775 "%s/scan/%s/__rfc822_%05u", spool_directory, message_id, file_nr++);
779 result = stat(CS filename,&mystat);
780 } while (result != -1);
782 rfc822name = filename;
784 /* decode RFC822 attachment */
785 mime_decoded_filename = NULL;
787 mime_current_boundary = context ? context->boundary : NULL;
788 mime_decode(&rfc822name);
790 mime_current_boundary = NULL;
791 if (!mime_decoded_filename) /* decoding failed */
793 log_write(0, LOG_MAIN,
794 "MIME acl condition warning - could not decode RFC822 MIME part to file.");
798 mime_decoded_filename = NULL;
802 /* If the boundary of this instance is NULL, we are finished here */
805 if (context->context == MBC_COVERLETTER_ONESHOT)
806 context->context = MBC_ATTACHMENT;
814 #endif /*WITH_CONTENT_SCAN*/