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 * decode_path;
222 FILE *decode_file = NULL;
224 ssize_t size_counter = 0;
225 ssize_t (*decode_function)(FILE*, FILE*, uschar*);
227 if (!mime_stream || (f_pos = ftell(mime_stream)) < 0)
230 /* build default decode path (will exist since MBOX must be spooled up) */
231 decode_path = string_sprintf("%s/scan/%s", spool_directory, message_id);
233 /* try to find 1st option */
234 if ((option = string_nextinlist(&list, &sep, NULL, 0)))
236 /* parse 1st option */
237 if ((Ustrcmp(option,"false") == 0) || (Ustrcmp(option,"0") == 0))
238 /* explicitly no decoding */
241 if (Ustrcmp(option,"default") == 0)
242 /* explicit default path + file names */
245 if (option[0] == '/')
249 memset(&statbuf,0,sizeof(statbuf));
251 /* assume either path or path+file name */
252 if ( (stat(CS option, &statbuf) == 0) && S_ISDIR(statbuf.st_mode) )
253 /* is directory, use it as decode_path */
254 decode_file = mime_get_decode_file(option, NULL);
256 /* does not exist or is a file, use as full file name */
257 decode_file = mime_get_decode_file(NULL, option);
260 /* assume file name only, use default path */
261 decode_file = mime_get_decode_file(decode_path, option);
265 /* no option? patch default path */
267 decode_file = mime_get_decode_file(decode_path, NULL);
273 /* decode according to mime type */
275 !mime_content_transfer_encoding
276 ? mime_decode_asis /* no encoding, dump as-is */
277 : Ustrcmp(mime_content_transfer_encoding, "base64") == 0
279 : Ustrcmp(mime_content_transfer_encoding, "quoted-printable") == 0
281 : mime_decode_asis; /* unknown encoding type, just dump as-is */
283 size_counter = decode_function(mime_stream, decode_file, mime_current_boundary);
285 clearerr(mime_stream);
286 if (fseek(mime_stream, f_pos, SEEK_SET))
289 if (fclose(decode_file) != 0 || size_counter < 0)
292 /* round up to the next KiB */
293 mime_content_size = (size_counter + 1023) / 1024;
300 mime_get_header(FILE *f, uschar *header)
304 int header_value_mode = 0;
305 int header_open_brackets = 0;
310 if ((c = fgetc(f)) == EOF) break;
312 /* always skip CRs */
313 if (c == '\r') continue;
319 /* look if next char is '\t' or ' ' */
320 if ((c = fgetc(f)) == EOF) break;
321 if ( (c == '\t') || (c == ' ') ) continue;
324 /* end of the header, terminate with ';' */
329 /* skip control characters */
330 if (c < 32) continue;
332 if (header_value_mode)
334 /* --------- value mode ----------- */
335 /* skip leading whitespace */
336 if ( ((c == '\t') || (c == ' ')) && (header_value_mode == 1) )
339 /* we have hit a non-whitespace char, start copying value data */
340 header_value_mode = 2;
342 if (c == '"') /* flip "quoted" mode */
343 header_value_mode = header_value_mode==2 ? 3 : 2;
345 /* leave value mode on unquoted ';' */
346 if (header_value_mode == 2 && c == ';') {
347 header_value_mode = 0;
349 /* -------------------------------- */
353 /* -------- non-value mode -------- */
354 /* skip whitespace + tabs */
355 if ( (c == ' ') || (c == '\t') )
359 /* quote next char. can be used
360 to escape brackets. */
361 if ((c = fgetc(f)) == EOF) break;
365 header_open_brackets++;
368 else if ((c == ')') && header_open_brackets)
370 header_open_brackets--;
373 else if ( (c == '=') && !header_open_brackets ) /* enter value mode */
374 header_value_mode = 1;
376 /* skip chars while we are in a comment */
377 if (header_open_brackets > 0)
379 /* -------------------------------- */
382 /* copy the char to the buffer */
383 header[num_copied++] = (uschar)c;
385 /* break if header buffer is full */
386 if (num_copied > MIME_MAX_HEADER_SIZE-1)
390 if ((num_copied > 0) && (header[num_copied-1] != ';'))
391 header[num_copied-1] = ';';
394 header[num_copied] = '\0';
396 /* return 0 for EOF or empty line */
397 if ((c == EOF) || (num_copied == 1))
405 mime_vars_reset(void)
407 mime_anomaly_level = 0;
408 mime_anomaly_text = NULL;
409 mime_boundary = NULL;
411 mime_decoded_filename = NULL;
412 mime_filename = NULL;
413 mime_content_description = NULL;
414 mime_content_disposition = NULL;
415 mime_content_id = NULL;
416 mime_content_transfer_encoding = NULL;
417 mime_content_type = NULL;
418 mime_is_multipart = 0;
419 mime_content_size = 0;
423 /* Grab a parameter value, dealing with quoting.
426 str Input string. Updated on return to point to terminating ; or NUL
429 Allocated string with parameter value
432 mime_param_val(uschar ** sp)
436 int size = 0, ptr = 0;
438 /* debug_printf_indent(" considering paramval '%s'\n", s); */
440 while (*s && *s != ';') /* ; terminates */
443 s++; /* skip opening " */
444 while (*s && *s != '"') /* " protects ; */
445 val = string_catn(val, &size, &ptr, s++, 1);
446 if (*s) s++; /* skip closing " */
449 val = string_catn(val, &size, &ptr, s++, 1);
450 if (val) val[ptr] = '\0';
456 mime_next_semicolon(uschar * s)
458 while (*s && *s != ';') /* ; terminates */
461 s++; /* skip opening " */
462 while (*s && *s != '"') /* " protects ; */
464 if (*s) s++; /* skip closing " */
473 rfc2231_to_2047(const uschar * fname, const uschar * charset, int * len)
475 int size = 0, ptr = 0;
476 uschar * val = string_catn(NULL, &size, &ptr, US"=?", 2);
480 val = string_cat(val, &size, &ptr, charset);
481 val = string_catn(val, &size, &ptr, US"?Q?", 3);
484 if (c == '%' && isxdigit(fname[1]) && isxdigit(fname[2]))
486 val = string_catn(val, &size, &ptr, US"=", 1);
487 val = string_catn(val, &size, &ptr, ++fname, 2);
491 val = string_catn(val, &size, &ptr, fname++, 1);
493 val = string_catn(val, &size, &ptr, US"?=", 2);
494 val[*len = ptr] = '\0';
500 mime_acl_check(uschar *acl, FILE *f, struct mime_boundary_context *context,
501 uschar **user_msgptr, uschar **log_msgptr)
504 uschar * header = NULL;
505 struct mime_boundary_context nested_context;
507 /* reserve a line buffer to work in */
508 header = store_get(MIME_MAX_HEADER_SIZE+1);
510 /* Not actually used at the moment, but will be vital to fixing
511 * some RFC 2046 nonconformance later... */
512 nested_context.parent = context;
514 /* loop through parts */
517 /* reset all per-part mime variables */
520 /* If boundary is null, we assume that *f is positioned on the start of
521 headers (for example, at the very beginning of a message. If a boundary is
522 given, we must first advance to it to reach the start of the next header
525 /* NOTE -- there's an error here -- RFC2046 specifically says to
526 * check for outer boundaries. This code doesn't do that, and
527 * I haven't fixed this.
529 * (I have moved partway towards adding support, however, by adding
530 * a "parent" field to my new boundary-context structure.)
532 if (context) for (;;)
534 if (!fgets(CS header, MIME_MAX_HEADER_SIZE, f))
536 /* Hit EOF or read error. Ugh. */
537 DEBUG(D_acl) debug_printf_indent("MIME: Hit EOF ...\n");
541 /* boundary line must start with 2 dashes */
542 if ( Ustrncmp(header, "--", 2) == 0
543 && Ustrncmp(header+2, context->boundary, Ustrlen(context->boundary)) == 0
545 { /* found boundary */
546 if (Ustrncmp((header+2+Ustrlen(context->boundary)), "--", 2) == 0)
548 /* END boundary found */
549 DEBUG(D_acl) debug_printf_indent("MIME: End boundary found %s\n",
554 DEBUG(D_acl) debug_printf_indent("MIME: Next part with boundary %s\n",
560 /* parse headers, set up expansion variables */
561 while (mime_get_header(f, header))
563 struct mime_header * mh;
565 /* look for interesting headers */
566 for (mh = mime_header_list;
567 mh < mime_header_list + mime_header_list_size;
568 mh++) if (strncmpic(mh->name, header, mh->namelen) == 0)
570 uschar * p = header + mh->namelen;
573 /* grab the value (normalize to lower case)
574 and copy to its corresponding expansion variable */
576 for (q = p; *q != ';' && *q; q++) ;
577 *mh->value = string_copynlc(p, q-p);
578 DEBUG(D_acl) debug_printf_indent("MIME: found %s header, value is '%s'\n",
579 mh->name, *mh->value);
581 if (*(p = q)) p++; /* jump past the ; */
584 uschar * mime_fname = NULL;
585 uschar * mime_fname_rfc2231 = NULL;
586 uschar * mime_filename_charset = NULL;
587 BOOL decoding_failed = FALSE;
589 /* grab all param=value tags on the remaining line,
590 check if they are interesting */
596 DEBUG(D_acl) debug_printf_indent("MIME: considering paramlist '%s'\n", p);
599 && strncmpic(CUS"content-disposition:", header, 20) == 0
600 && strncmpic(CUS"filename*", p, 9) == 0
602 { /* RFC 2231 filename */
605 /* find value of the filename */
607 while(*p != '=' && *p) p++;
608 if (*p) p++; /* p is filename or NUL */
609 q = mime_param_val(&p); /* p now trailing ; or NUL */
613 uschar * temp_string, * err_msg;
616 /* build up an un-decoded filename over successive
617 filename*= parameters (for use when 2047 decode fails) */
619 mime_fname_rfc2231 = string_sprintf("%#s%s",
620 mime_fname_rfc2231, q);
622 if (!decoding_failed)
625 if (!mime_filename_charset)
629 /* look for a ' in the "filename" */
630 while(*s != '\'' && *s) s++; /* s is 1st ' or NUL */
632 if ((size = s-q) > 0)
633 mime_filename_charset = string_copyn(q, size);
636 while(*p == '\'') p++; /* p is after 2nd ' */
641 DEBUG(D_acl) debug_printf_indent("MIME: charset %s fname '%s'\n",
642 mime_filename_charset ? mime_filename_charset : US"<NULL>", p);
644 temp_string = rfc2231_to_2047(p, mime_filename_charset, &slen);
645 DEBUG(D_acl) debug_printf_indent("MIME: 2047-name %s\n", temp_string);
647 temp_string = rfc2047_decode(temp_string, FALSE, NULL, ' ',
649 DEBUG(D_acl) debug_printf_indent("MIME: plain-name %s\n", temp_string);
651 size = Ustrlen(temp_string);
654 decoding_failed = TRUE;
656 /* build up a decoded filename over successive
657 filename*= parameters */
659 mime_filename = mime_fname = mime_fname
660 ? string_sprintf("%s%s", mime_fname, temp_string)
667 /* look for interesting parameters */
668 for (mp = mime_parameter_list;
669 mp < mime_parameter_list + nelem(mime_parameter_list);
671 ) if (strncmpic(mp->name, p, mp->namelen) == 0)
674 uschar * dummy_errstr;
676 /* grab the value and copy to its expansion variable */
678 q = mime_param_val(&p); /* p now trailing ; or NUL */
681 ? rfc2047_decode(q, check_rfc2047_length, NULL, 32, NULL,
684 DEBUG(D_acl) debug_printf_indent(
685 "MIME: found %s parameter in %s header, value '%s'\n",
686 mp->name, mh->name, *mp->value);
688 break; /* done matching param names */
692 /* There is something, but not one of our interesting parameters.
693 Advance past the next semicolon */
694 p = mime_next_semicolon(p);
696 } /* param scan on line */
698 if (strncmpic(CUS"content-disposition:", header, 20) == 0)
700 if (decoding_failed) mime_filename = mime_fname_rfc2231;
702 DEBUG(D_acl) debug_printf_indent(
703 "MIME: found %s parameter in %s header, value is '%s'\n",
704 "filename", mh->name, mime_filename);
710 /* set additional flag variables (easier access) */
711 if ( mime_content_type
712 && Ustrncmp(mime_content_type,"multipart",9) == 0
714 mime_is_multipart = 1;
716 /* Make a copy of the boundary pointer.
717 Required since mime_boundary is global
718 and can be overwritten further down in recursion */
719 nested_context.boundary = mime_boundary;
721 /* raise global counter */
724 /* copy current file handle to global variable */
726 mime_current_boundary = context ? context->boundary : 0;
728 /* Note the context */
729 mime_is_coverletter = !(context && context->context == MBC_ATTACHMENT);
731 /* call ACL handling function */
732 rc = acl_check(ACL_WHERE_MIME, NULL, acl, user_msgptr, log_msgptr);
735 mime_current_boundary = NULL;
739 /* If we have a multipart entity and a boundary, go recursive */
740 if ( (mime_content_type != NULL) &&
741 (nested_context.boundary != NULL) &&
742 (Ustrncmp(mime_content_type,"multipart",9) == 0) )
745 debug_printf_indent("MIME: Entering multipart recursion, boundary '%s'\n",
746 nested_context.boundary);
748 nested_context.context =
749 context && context->context == MBC_ATTACHMENT
751 : Ustrcmp(mime_content_type,"multipart/alternative") == 0
752 || Ustrcmp(mime_content_type,"multipart/related") == 0
753 ? MBC_COVERLETTER_ALL
754 : MBC_COVERLETTER_ONESHOT;
756 rc = mime_acl_check(acl, f, &nested_context, user_msgptr, log_msgptr);
759 else if ( (mime_content_type != NULL) &&
760 (Ustrncmp(mime_content_type,"message/rfc822",14) == 0) )
762 const uschar *rfc822name = NULL;
763 uschar filename[2048];
767 /* must find first free sequential filename */
771 (void)string_format(filename, 2048,
772 "%s/scan/%s/__rfc822_%05u", spool_directory, message_id, file_nr++);
776 result = stat(CS filename,&mystat);
777 } while (result != -1);
779 rfc822name = filename;
781 /* decode RFC822 attachment */
782 mime_decoded_filename = NULL;
784 mime_current_boundary = context ? context->boundary : NULL;
785 mime_decode(&rfc822name);
787 mime_current_boundary = NULL;
788 if (!mime_decoded_filename) /* decoding failed */
790 log_write(0, LOG_MAIN,
791 "MIME acl condition warning - could not decode RFC822 MIME part to file.");
795 mime_decoded_filename = NULL;
799 /* If the boundary of this instance is NULL, we are finished here */
802 if (context->context == MBC_COVERLETTER_ONESHOT)
803 context->context = MBC_ATTACHMENT;
811 #endif /*WITH_CONTENT_SCAN*/