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);
164 else if (decode_qp_result == -1)
166 else if (decode_qp_result >= 0)
168 *opos = decode_qp_result;
179 /* something to write? */
183 if (fwrite(obuf, 1, len, out) != len) return -1; /* error */
192 * Return open filehandle for combo of path and file.
193 * Side-effect: set mime_decoded_filename, to copy in allocated mem
196 mime_get_decode_file(uschar *pname, uschar *fname)
199 mime_decoded_filename = string_sprintf("%s/%s", pname, fname);
201 mime_decoded_filename = string_copy(fname);
207 /* must find first free sequential filename */
211 mime_decoded_filename = string_sprintf("%s/%s-%05u", pname, message_id, file_nr++);
215 result = stat(CS mime_decoded_filename, &mystat);
216 } while(result != -1);
219 return modefopen(mime_decoded_filename, "wb+", SPOOL_MODE);
224 mime_decode(const uschar **listptr)
227 const uschar *list = *listptr;
229 uschar option_buffer[1024];
230 uschar decode_path[1024];
231 FILE *decode_file = NULL;
233 ssize_t size_counter = 0;
234 ssize_t (*decode_function)(FILE*, FILE*, uschar*);
236 if (mime_stream == NULL)
239 f_pos = ftell(mime_stream);
241 /* build default decode path (will exist since MBOX must be spooled up) */
242 (void)string_format(decode_path,1024,"%s/scan/%s",spool_directory,message_id);
244 /* try to find 1st option */
245 if ((option = string_nextinlist(&list, &sep,
247 sizeof(option_buffer))) != NULL)
249 /* parse 1st option */
250 if ( (Ustrcmp(option,"false") == 0) || (Ustrcmp(option,"0") == 0) )
251 /* explicitly no decoding */
254 if (Ustrcmp(option,"default") == 0)
255 /* explicit default path + file names */
258 if (option[0] == '/')
262 memset(&statbuf,0,sizeof(statbuf));
264 /* assume either path or path+file name */
265 if ( (stat(CS option, &statbuf) == 0) && S_ISDIR(statbuf.st_mode) )
266 /* is directory, use it as decode_path */
267 decode_file = mime_get_decode_file(option, NULL);
269 /* does not exist or is a file, use as full file name */
270 decode_file = mime_get_decode_file(NULL, option);
273 /* assume file name only, use default path */
274 decode_file = mime_get_decode_file(decode_path, option);
278 /* no option? patch default path */
280 decode_file = mime_get_decode_file(decode_path, NULL);
286 /* decode according to mime type */
288 !mime_content_transfer_encoding
289 ? mime_decode_asis /* no encoding, dump as-is */
290 : Ustrcmp(mime_content_transfer_encoding, "base64") == 0
292 : Ustrcmp(mime_content_transfer_encoding, "quoted-printable") == 0
294 : mime_decode_asis; /* unknown encoding type, just dump as-is */
296 size_counter = decode_function(mime_stream, decode_file, mime_current_boundary);
298 clearerr(mime_stream);
299 if (fseek(mime_stream, f_pos, SEEK_SET))
302 if (fclose(decode_file) != 0 || size_counter < 0)
305 /* round up to the next KiB */
306 mime_content_size = (size_counter + 1023) / 1024;
313 mime_get_header(FILE *f, uschar *header)
317 int header_value_mode = 0;
318 int header_open_brackets = 0;
323 if ((c = fgetc(f)) == EOF) break;
325 /* always skip CRs */
326 if (c == '\r') continue;
332 /* look if next char is '\t' or ' ' */
333 if ((c = fgetc(f)) == EOF) break;
334 if ( (c == '\t') || (c == ' ') ) continue;
337 /* end of the header, terminate with ';' */
342 /* skip control characters */
343 if (c < 32) continue;
345 if (header_value_mode)
347 /* --------- value mode ----------- */
348 /* skip leading whitespace */
349 if ( ((c == '\t') || (c == ' ')) && (header_value_mode == 1) )
352 /* we have hit a non-whitespace char, start copying value data */
353 header_value_mode = 2;
355 if (c == '"') /* flip "quoted" mode */
356 header_value_mode = header_value_mode==2 ? 3 : 2;
358 /* leave value mode on unquoted ';' */
359 if (header_value_mode == 2 && c == ';') {
360 header_value_mode = 0;
362 /* -------------------------------- */
366 /* -------- non-value mode -------- */
367 /* skip whitespace + tabs */
368 if ( (c == ' ') || (c == '\t') )
372 /* quote next char. can be used
373 to escape brackets. */
374 if ((c = fgetc(f)) == EOF) break;
378 header_open_brackets++;
381 else if ((c == ')') && header_open_brackets)
383 header_open_brackets--;
386 else if ( (c == '=') && !header_open_brackets ) /* enter value mode */
387 header_value_mode = 1;
389 /* skip chars while we are in a comment */
390 if (header_open_brackets > 0)
392 /* -------------------------------- */
395 /* copy the char to the buffer */
396 header[num_copied++] = (uschar)c;
398 /* break if header buffer is full */
399 if (num_copied > MIME_MAX_HEADER_SIZE-1)
403 if ((num_copied > 0) && (header[num_copied-1] != ';'))
404 header[num_copied-1] = ';';
407 header[num_copied] = '\0';
409 /* return 0 for EOF or empty line */
410 if ((c == EOF) || (num_copied == 1))
418 mime_vars_reset(void)
420 mime_anomaly_level = 0;
421 mime_anomaly_text = NULL;
422 mime_boundary = NULL;
424 mime_decoded_filename = NULL;
425 mime_filename = NULL;
426 mime_content_description = NULL;
427 mime_content_disposition = NULL;
428 mime_content_id = NULL;
429 mime_content_transfer_encoding = NULL;
430 mime_content_type = NULL;
431 mime_is_multipart = 0;
432 mime_content_size = 0;
436 /* Grab a parameter value, dealing with quoting.
439 str Input string. Updated on return to point to terminating ; or NUL
442 Allocated string with parameter value
445 mime_param_val(uschar ** sp)
449 int size = 0, ptr = 0;
451 /* debug_printf(" considering paramval '%s'\n", s); */
453 while (*s && *s != ';') /* ; terminates */
456 s++; /* skip opening " */
457 while (*s && *s != '"') /* " protects ; */
458 val = string_catn(val, &size, &ptr, s++, 1);
459 if (*s) s++; /* skip closing " */
462 val = string_catn(val, &size, &ptr, s++, 1);
463 if (val) val[ptr] = '\0';
469 mime_next_semicolon(uschar * s)
471 while (*s && *s != ';') /* ; terminates */
474 s++; /* skip opening " */
475 while (*s && *s != '"') /* " protects ; */
477 if (*s) s++; /* skip closing " */
486 rfc2231_to_2047(const uschar * fname, const uschar * charset, int * len)
488 int size = 0, ptr = 0;
489 uschar * val = string_catn(NULL, &size, &ptr, US"=?", 2);
493 val = string_cat(val, &size, &ptr, charset);
494 val = string_catn(val, &size, &ptr, US"?Q?", 3);
497 if (c == '%' && isxdigit(fname[1]) && isxdigit(fname[2]))
499 val = string_catn(val, &size, &ptr, US"=", 1);
500 val = string_catn(val, &size, &ptr, ++fname, 2);
504 val = string_catn(val, &size, &ptr, fname++, 1);
506 val = string_catn(val, &size, &ptr, US"?=", 2);
507 val[*len = ptr] = '\0';
513 mime_acl_check(uschar *acl, FILE *f, struct mime_boundary_context *context,
514 uschar **user_msgptr, uschar **log_msgptr)
517 uschar * header = NULL;
518 struct mime_boundary_context nested_context;
520 /* reserve a line buffer to work in */
521 header = store_get(MIME_MAX_HEADER_SIZE+1);
523 /* Not actually used at the moment, but will be vital to fixing
524 * some RFC 2046 nonconformance later... */
525 nested_context.parent = context;
527 /* loop through parts */
530 /* reset all per-part mime variables */
533 /* If boundary is null, we assume that *f is positioned on the start of
534 headers (for example, at the very beginning of a message. If a boundary is
535 given, we must first advance to it to reach the start of the next header
538 /* NOTE -- there's an error here -- RFC2046 specifically says to
539 * check for outer boundaries. This code doesn't do that, and
540 * I haven't fixed this.
542 * (I have moved partway towards adding support, however, by adding
543 * a "parent" field to my new boundary-context structure.)
545 if (context) for (;;)
547 if (!fgets(CS header, MIME_MAX_HEADER_SIZE, f))
549 /* Hit EOF or read error. Ugh. */
550 DEBUG(D_acl) debug_printf("MIME: Hit EOF ...\n");
554 /* boundary line must start with 2 dashes */
555 if ( Ustrncmp(header, "--", 2) == 0
556 && Ustrncmp(header+2, context->boundary, Ustrlen(context->boundary)) == 0
558 { /* found boundary */
559 if (Ustrncmp((header+2+Ustrlen(context->boundary)), "--", 2) == 0)
561 /* END boundary found */
562 DEBUG(D_acl) debug_printf("MIME: End boundary found %s\n",
567 DEBUG(D_acl) debug_printf("MIME: Next part with boundary %s\n",
573 /* parse headers, set up expansion variables */
574 while (mime_get_header(f, header))
576 struct mime_header * mh;
578 /* look for interesting headers */
579 for (mh = mime_header_list;
580 mh < mime_header_list + mime_header_list_size;
581 mh++) if (strncmpic(mh->name, header, mh->namelen) == 0)
583 uschar * p = header + mh->namelen;
586 /* grab the value (normalize to lower case)
587 and copy to its corresponding expansion variable */
589 for (q = p; *q != ';' && *q; q++) ;
590 *mh->value = string_copynlc(p, q-p);
591 DEBUG(D_acl) debug_printf("MIME: found %s header, value is '%s'\n",
592 mh->name, *mh->value);
594 if (*(p = q)) p++; /* jump past the ; */
597 uschar * mime_fname = NULL;
598 uschar * mime_fname_rfc2231 = NULL;
599 uschar * mime_filename_charset = NULL;
600 BOOL decoding_failed = FALSE;
602 /* grab all param=value tags on the remaining line,
603 check if they are interesting */
609 DEBUG(D_acl) debug_printf("MIME: considering paramlist '%s'\n", p);
612 && strncmpic(CUS"content-disposition:", header, 20) == 0
613 && strncmpic(CUS"filename*", p, 9) == 0
615 { /* RFC 2231 filename */
618 /* find value of the filename */
620 while(*p != '=' && *p) p++;
621 if (*p) p++; /* p is filename or NUL */
622 q = mime_param_val(&p); /* p now trailing ; or NUL */
626 uschar * temp_string, * err_msg;
629 /* build up an un-decoded filename over successive
630 filename*= parameters (for use when 2047 decode fails) */
632 mime_fname_rfc2231 = string_sprintf("%#s%s",
633 mime_fname_rfc2231, q);
635 if (!decoding_failed)
638 if (!mime_filename_charset)
642 /* look for a ' in the "filename" */
643 while(*s != '\'' && *s) s++; /* s is 1st ' or NUL */
645 if ((size = s-q) > 0)
646 mime_filename_charset = string_copyn(q, size);
649 while(*p == '\'') p++; /* p is after 2nd ' */
654 DEBUG(D_acl) debug_printf("MIME: charset %s fname '%s'\n",
655 mime_filename_charset ? mime_filename_charset : US"<NULL>", p);
657 temp_string = rfc2231_to_2047(p, mime_filename_charset, &slen);
658 DEBUG(D_acl) debug_printf("MIME: 2047-name %s\n", temp_string);
660 temp_string = rfc2047_decode(temp_string, FALSE, NULL, ' ',
662 DEBUG(D_acl) debug_printf("MIME: plain-name %s\n", temp_string);
664 size = Ustrlen(temp_string);
667 decoding_failed = TRUE;
669 /* build up a decoded filename over successive
670 filename*= parameters */
672 mime_filename = mime_fname = mime_fname
673 ? string_sprintf("%s%s", mime_fname, temp_string)
680 /* look for interesting parameters */
681 for (mp = mime_parameter_list;
682 mp < mime_parameter_list + nelem(mime_parameter_list);
684 ) if (strncmpic(mp->name, p, mp->namelen) == 0)
687 uschar * dummy_errstr;
689 /* grab the value and copy to its expansion variable */
691 q = mime_param_val(&p); /* p now trailing ; or NUL */
694 ? rfc2047_decode(q, check_rfc2047_length, NULL, 32, NULL,
697 DEBUG(D_acl) debug_printf(
698 "MIME: found %s parameter in %s header, value '%s'\n",
699 mp->name, mh->name, *mp->value);
701 break; /* done matching param names */
705 /* There is something, but not one of our interesting parameters.
706 Advance past the next semicolon */
707 p = mime_next_semicolon(p);
709 } /* param scan on line */
711 if (strncmpic(CUS"content-disposition:", header, 20) == 0)
713 if (decoding_failed) mime_filename = mime_fname_rfc2231;
715 DEBUG(D_acl) debug_printf(
716 "MIME: found %s parameter in %s header, value is '%s'\n",
717 "filename", mh->name, mime_filename);
723 /* set additional flag variables (easier access) */
724 if ( mime_content_type
725 && Ustrncmp(mime_content_type,"multipart",9) == 0
727 mime_is_multipart = 1;
729 /* Make a copy of the boundary pointer.
730 Required since mime_boundary is global
731 and can be overwritten further down in recursion */
732 nested_context.boundary = mime_boundary;
734 /* raise global counter */
737 /* copy current file handle to global variable */
739 mime_current_boundary = context ? context->boundary : 0;
741 /* Note the context */
742 mime_is_coverletter = !(context && context->context == MBC_ATTACHMENT);
744 /* call ACL handling function */
745 rc = acl_check(ACL_WHERE_MIME, NULL, acl, user_msgptr, log_msgptr);
748 mime_current_boundary = NULL;
752 /* If we have a multipart entity and a boundary, go recursive */
753 if ( (mime_content_type != NULL) &&
754 (nested_context.boundary != NULL) &&
755 (Ustrncmp(mime_content_type,"multipart",9) == 0) )
758 debug_printf("MIME: Entering multipart recursion, boundary '%s'\n",
759 nested_context.boundary);
761 nested_context.context =
762 context && context->context == MBC_ATTACHMENT
764 : Ustrcmp(mime_content_type,"multipart/alternative") == 0
765 || Ustrcmp(mime_content_type,"multipart/related") == 0
766 ? MBC_COVERLETTER_ALL
767 : MBC_COVERLETTER_ONESHOT;
769 rc = mime_acl_check(acl, f, &nested_context, user_msgptr, log_msgptr);
772 else if ( (mime_content_type != NULL) &&
773 (Ustrncmp(mime_content_type,"message/rfc822",14) == 0) )
775 const uschar *rfc822name = NULL;
776 uschar filename[2048];
780 /* must find first free sequential filename */
784 (void)string_format(filename, 2048,
785 "%s/scan/%s/__rfc822_%05u", spool_directory, message_id, file_nr++);
789 result = stat(CS filename,&mystat);
790 } while (result != -1);
792 rfc822name = filename;
794 /* decode RFC822 attachment */
795 mime_decoded_filename = NULL;
797 mime_current_boundary = context ? context->boundary : NULL;
798 mime_decode(&rfc822name);
800 mime_current_boundary = NULL;
801 if (!mime_decoded_filename) /* decoding failed */
803 log_write(0, LOG_MAIN,
804 "MIME acl condition warning - could not decode RFC822 MIME part to file.");
808 mime_decoded_filename = NULL;
812 /* If the boundary of this instance is NULL, we are finished here */
815 if (context->context == MBC_COVERLETTER_ONESHOT)
816 context->context = MBC_ATTACHMENT;
824 #endif /*WITH_CONTENT_SCAN*/