Alex Miller's patch for LDAP_RES_SEARCH_REFERENCE.
[exim.git] / src / src / mime.c
1 /* $Cambridge: exim/src/src/mime.c,v 1.3 2004/12/17 14:52:44 ph10 Exp $ */
2
3 /*************************************************
4 * Exim - an Internet mail transport agent *
5 *************************************************/
6
7 /* Copyright (c) Tom Kistner <tom@duncanthrax.net> 2004 */
8 /* License: GPL */
9
10 #include "exim.h"
11 #ifdef WITH_CONTENT_SCAN
12 #include "mime.h"
13 #include <sys/stat.h>
14
15 FILE *mime_stream = NULL;
16 uschar *mime_current_boundary = NULL;
17
18 /*************************************************
19 * set MIME anomaly level + text *
20 *************************************************/
21
22 /* Small wrapper to set the two expandables which
23 give info on detected "problems" in MIME
24 encodings. Those are defined in mime.h. */
25
26 void mime_set_anomaly(int level, char *text) {
27 mime_anomaly_level = level;
28 mime_anomaly_text = US text;
29 }
30
31
32 /*************************************************
33 * decode quoted-printable chars *
34 *************************************************/
35
36 /* gets called when we hit a =
37 returns: new pointer position
38 result code in c:
39 -2 - decode error
40 -1 - soft line break, no char
41 0-255 - char to write
42 */
43
44 unsigned int mime_qp_hstr_i(uschar *cptr) {
45 unsigned int i, j = 0;
46 while (cptr && *cptr && isxdigit(*cptr)) {
47 i = *cptr++ - '0';
48 if (9 < i) i -= 7;
49 j <<= 4;
50 j |= (i & 0x0f);
51 }
52 return(j);
53 }
54
55 uschar *mime_decode_qp_char(uschar *qp_p,int *c) {
56 uschar hex[] = {0,0,0};
57 int nan = 0;
58 uschar *initial_pos = qp_p;
59
60 /* advance one char */
61 qp_p++;
62
63 REPEAT_FIRST:
64 if ( (*qp_p == '\t') || (*qp_p == ' ') || (*qp_p == '\r') ) {
65 /* tab or whitespace may follow
66 just ignore it, but remember
67 that this is not a valid hex
68 encoding any more */
69 nan = 1;
70 qp_p++;
71 goto REPEAT_FIRST;
72 }
73 else if ( (('0' <= *qp_p) && (*qp_p <= '9')) || (('A' <= *qp_p) && (*qp_p <= 'F')) || (('a' <= *qp_p) && (*qp_p <= 'f')) ) {
74 /* this is a valid hex char, if nan is unset */
75 if (nan) {
76 /* this is illegal */
77 *c = -2;
78 return initial_pos;
79 }
80 else {
81 hex[0] = *qp_p;
82 qp_p++;
83 };
84 }
85 else if (*qp_p == '\n') {
86 /* hit soft line break already, continue */
87 *c = -1;
88 return qp_p;
89 }
90 else {
91 /* illegal char here */
92 *c = -2;
93 return initial_pos;
94 };
95
96 if ( (('0' <= *qp_p) && (*qp_p <= '9')) || (('A' <= *qp_p) && (*qp_p <= 'F')) || (('a' <= *qp_p) && (*qp_p <= 'f')) ) {
97 if (hex[0] > 0) {
98 hex[1] = *qp_p;
99 /* do hex conversion */
100 *c = mime_qp_hstr_i(hex);
101 qp_p++;
102 return qp_p;
103 }
104 else {
105 /* huh ? */
106 *c = -2;
107 return initial_pos;
108 };
109 }
110 else {
111 /* illegal char */
112 *c = -2;
113 return initial_pos;
114 };
115 }
116
117
118 uschar *mime_parse_line(uschar *buffer, uschar *encoding, int *num_decoded) {
119 uschar *data = NULL;
120
121 data = (uschar *)malloc(Ustrlen(buffer)+2);
122
123 if (encoding == NULL) {
124 /* no encoding type at all */
125 NO_DECODING:
126 memcpy(data, buffer, Ustrlen(buffer));
127 data[(Ustrlen(buffer))] = 0;
128 *num_decoded = Ustrlen(data);
129 return data;
130 }
131 else if (Ustrcmp(encoding,"base64") == 0) {
132 uschar *p = buffer;
133 int offset = 0;
134
135 /* ----- BASE64 ---------------------------------------------------- */
136 /* NULL out '\r' and '\n' chars */
137 while (Ustrrchr(p,'\r') != NULL) {
138 *(Ustrrchr(p,'\r')) = '\0';
139 };
140 while (Ustrrchr(p,'\n') != NULL) {
141 *(Ustrrchr(p,'\n')) = '\0';
142 };
143
144 while (*(p+offset) != '\0') {
145 /* hit illegal char ? */
146 if (mime_b64[*(p+offset)] == 128) {
147 mime_set_anomaly(MIME_ANOMALY_BROKEN_BASE64);
148 offset++;
149 }
150 else {
151 *p = mime_b64[*(p+offset)];
152 p++;
153 };
154 };
155 *p = 255;
156
157 /* line is translated, start bit shifting */
158 p = buffer;
159 *num_decoded = 0;
160 while(*p != 255) {
161 uschar tmp_c;
162
163 /* byte 0 ---------------------- */
164 if (*(p+1) == 255) {
165 mime_set_anomaly(MIME_ANOMALY_BROKEN_BASE64);
166 break;
167 }
168 data[(*num_decoded)] = *p;
169 data[(*num_decoded)] <<= 2;
170 tmp_c = *(p+1);
171 tmp_c >>= 4;
172 data[(*num_decoded)] |= tmp_c;
173 (*num_decoded)++;
174 p++;
175 /* byte 1 ---------------------- */
176 if (*(p+1) == 255) {
177 mime_set_anomaly(MIME_ANOMALY_BROKEN_BASE64);
178 break;
179 }
180 data[(*num_decoded)] = *p;
181 data[(*num_decoded)] <<= 4;
182 tmp_c = *(p+1);
183 tmp_c >>= 2;
184 data[(*num_decoded)] |= tmp_c;
185 (*num_decoded)++;
186 p++;
187 /* byte 2 ---------------------- */
188 if (*(p+1) == 255) {
189 mime_set_anomaly(MIME_ANOMALY_BROKEN_BASE64);
190 break;
191 }
192 data[(*num_decoded)] = *p;
193 data[(*num_decoded)] <<= 6;
194 data[(*num_decoded)] |= *(p+1);
195 (*num_decoded)++;
196 p+=2;
197
198 };
199 return data;
200 /* ----------------------------------------------------------------- */
201 }
202 else if (Ustrcmp(encoding,"quoted-printable") == 0) {
203 uschar *p = buffer;
204
205 /* ----- QP -------------------------------------------------------- */
206 *num_decoded = 0;
207 while (*p != 0) {
208 if (*p == '=') {
209 int decode_qp_result;
210
211 p = mime_decode_qp_char(p,&decode_qp_result);
212
213 if (decode_qp_result == -2) {
214 /* Error from decoder. p is unchanged. */
215 mime_set_anomaly(MIME_ANOMALY_BROKEN_QP);
216 data[(*num_decoded)] = '=';
217 (*num_decoded)++;
218 p++;
219 }
220 else if (decode_qp_result == -1) {
221 break;
222 }
223 else if (decode_qp_result >= 0) {
224 data[(*num_decoded)] = decode_qp_result;
225 (*num_decoded)++;
226 };
227 }
228 else {
229 data[(*num_decoded)] = *p;
230 (*num_decoded)++;
231 p++;
232 };
233 };
234 return data;
235 /* ----------------------------------------------------------------- */
236 }
237 /* unknown encoding type, just dump as-is */
238 else goto NO_DECODING;
239 }
240
241
242 FILE *mime_get_decode_file(uschar *pname, uschar *fname) {
243 FILE *f = NULL;
244 uschar *filename;
245
246 filename = (uschar *)malloc(2048);
247
248 if ((pname != NULL) && (fname != NULL)) {
249 snprintf(CS filename, 2048, "%s/%s", pname, fname);
250 f = fopen(CS filename,"w+");
251 }
252 else if (pname == NULL) {
253 f = fopen(CS fname,"w+");
254 }
255 else if (fname == NULL) {
256 int file_nr = 0;
257 int result = 0;
258
259 /* must find first free sequential filename */
260 do {
261 struct stat mystat;
262 snprintf(CS filename,2048,"%s/%s-%05u", pname, message_id, file_nr);
263 file_nr++;
264 /* security break */
265 if (file_nr >= 1024)
266 break;
267 result = stat(CS filename,&mystat);
268 }
269 while(result != -1);
270 f = fopen(CS filename,"w+");
271 };
272
273 /* set expansion variable */
274 mime_decoded_filename = filename;
275
276 return f;
277 }
278
279
280 int mime_decode(uschar **listptr) {
281 int sep = 0;
282 uschar *list = *listptr;
283 uschar *option;
284 uschar option_buffer[1024];
285 uschar decode_path[1024];
286 FILE *decode_file = NULL;
287 uschar *buffer = NULL;
288 long f_pos = 0;
289 unsigned int size_counter = 0;
290
291 if (mime_stream == NULL)
292 return FAIL;
293
294 f_pos = ftell(mime_stream);
295
296 /* build default decode path (will exist since MBOX must be spooled up) */
297 snprintf(CS decode_path,1024,"%s/scan/%s",spool_directory,message_id);
298
299 /* reserve a line buffer to work in */
300 buffer = (uschar *)malloc(MIME_MAX_LINE_LENGTH+1);
301 if (buffer == NULL) {
302 log_write(0, LOG_PANIC,
303 "decode ACL condition: can't allocate %d bytes of memory.", MIME_MAX_LINE_LENGTH+1);
304 return DEFER;
305 };
306
307 /* try to find 1st option */
308 if ((option = string_nextinlist(&list, &sep,
309 option_buffer,
310 sizeof(option_buffer))) != NULL) {
311
312 /* parse 1st option */
313 if ( (Ustrcmp(option,"false") == 0) || (Ustrcmp(option,"0") == 0) ) {
314 /* explicitly no decoding */
315 return FAIL;
316 };
317
318 if (Ustrcmp(option,"default") == 0) {
319 /* explicit default path + file names */
320 goto DEFAULT_PATH;
321 };
322
323 if (option[0] == '/') {
324 struct stat statbuf;
325
326 memset(&statbuf,0,sizeof(statbuf));
327
328 /* assume either path or path+file name */
329 if ( (stat(CS option, &statbuf) == 0) && S_ISDIR(statbuf.st_mode) )
330 /* is directory, use it as decode_path */
331 decode_file = mime_get_decode_file(option, NULL);
332 else
333 /* does not exist or is a file, use as full file name */
334 decode_file = mime_get_decode_file(NULL, option);
335 }
336 else
337 /* assume file name only, use default path */
338 decode_file = mime_get_decode_file(decode_path, option);
339 }
340 else
341 /* no option? patch default path */
342 DEFAULT_PATH: decode_file = mime_get_decode_file(decode_path, NULL);
343
344 if (decode_file == NULL)
345 return DEFER;
346
347 /* read data linewise and dump it to the file,
348 while looking for the current boundary */
349 while(fgets(CS buffer, MIME_MAX_LINE_LENGTH, mime_stream) != NULL) {
350 uschar *decoded_line = NULL;
351 int decoded_line_length = 0;
352
353 if (mime_current_boundary != NULL) {
354 /* boundary line must start with 2 dashes */
355 if (Ustrncmp(buffer,"--",2) == 0) {
356 if (Ustrncmp((buffer+2),mime_current_boundary,Ustrlen(mime_current_boundary)) == 0)
357 break;
358 };
359 };
360
361 decoded_line = mime_parse_line(buffer, mime_content_transfer_encoding, &decoded_line_length);
362 /* write line to decode file */
363 if (fwrite(decoded_line, 1, decoded_line_length, decode_file) < decoded_line_length) {
364 /* error/short write */
365 clearerr(mime_stream);
366 fseek(mime_stream,f_pos,SEEK_SET);
367 return DEFER;
368 };
369 size_counter += decoded_line_length;
370
371 if (size_counter > 1023) {
372 if ((mime_content_size + (size_counter / 1024)) < 65535)
373 mime_content_size += (size_counter / 1024);
374 else
375 mime_content_size = 65535;
376 size_counter = (size_counter % 1024);
377 };
378
379 free(decoded_line);
380 }
381
382 fclose(decode_file);
383
384 clearerr(mime_stream);
385 fseek(mime_stream,f_pos,SEEK_SET);
386
387 /* round up remaining size bytes to one k */
388 if (size_counter) {
389 mime_content_size++;
390 };
391
392 return OK;
393 }
394
395 int mime_get_header(FILE *f, uschar *header) {
396 int c = EOF;
397 int done = 0;
398 int header_value_mode = 0;
399 int header_open_brackets = 0;
400 int num_copied = 0;
401
402 while(!done) {
403
404 c = fgetc(f);
405 if (c == EOF) break;
406
407 /* always skip CRs */
408 if (c == '\r') continue;
409
410 if (c == '\n') {
411 if (num_copied > 0) {
412 /* look if next char is '\t' or ' ' */
413 c = fgetc(f);
414 if (c == EOF) break;
415 if ( (c == '\t') || (c == ' ') ) continue;
416 ungetc(c,f);
417 };
418 /* end of the header, terminate with ';' */
419 c = ';';
420 done = 1;
421 };
422
423 /* skip control characters */
424 if (c < 32) continue;
425
426 if (header_value_mode) {
427 /* --------- value mode ----------- */
428 /* skip leading whitespace */
429 if ( ((c == '\t') || (c == ' ')) && (header_value_mode == 1) )
430 continue;
431
432 /* we have hit a non-whitespace char, start copying value data */
433 header_value_mode = 2;
434
435 /* skip quotes */
436 if (c == '"') continue;
437
438 /* leave value mode on ';' */
439 if (c == ';') {
440 header_value_mode = 0;
441 };
442 /* -------------------------------- */
443 }
444 else {
445 /* -------- non-value mode -------- */
446 /* skip whitespace + tabs */
447 if ( (c == ' ') || (c == '\t') )
448 continue;
449 if (c == '\\') {
450 /* quote next char. can be used
451 to escape brackets. */
452 c = fgetc(f);
453 if (c == EOF) break;
454 }
455 else if (c == '(') {
456 header_open_brackets++;
457 continue;
458 }
459 else if ((c == ')') && header_open_brackets) {
460 header_open_brackets--;
461 continue;
462 }
463 else if ( (c == '=') && !header_open_brackets ) {
464 /* enter value mode */
465 header_value_mode = 1;
466 };
467
468 /* skip chars while we are in a comment */
469 if (header_open_brackets > 0)
470 continue;
471 /* -------------------------------- */
472 };
473
474 /* copy the char to the buffer */
475 header[num_copied] = (uschar)c;
476 /* raise counter */
477 num_copied++;
478
479 /* break if header buffer is full */
480 if (num_copied > MIME_MAX_HEADER_SIZE-1) {
481 done = 1;
482 };
483 };
484
485 if (header[num_copied-1] != ';') {
486 header[num_copied-1] = ';';
487 };
488
489 /* 0-terminate */
490 header[num_copied] = '\0';
491
492 /* return 0 for EOF or empty line */
493 if ((c == EOF) || (num_copied == 1))
494 return 0;
495 else
496 return 1;
497 }
498
499
500 int mime_acl_check(FILE *f, struct mime_boundary_context *context, uschar
501 **user_msgptr, uschar **log_msgptr) {
502 int rc = OK;
503 uschar *header = NULL;
504 struct mime_boundary_context nested_context;
505
506 /* reserve a line buffer to work in */
507 header = (uschar *)malloc(MIME_MAX_HEADER_SIZE+1);
508 if (header == NULL) {
509 log_write(0, LOG_PANIC,
510 "acl_smtp_mime: can't allocate %d bytes of memory.", MIME_MAX_HEADER_SIZE+1);
511 return DEFER;
512 };
513
514 /* Not actually used at the moment, but will be vital to fixing
515 * some RFC 2046 nonconformance later... */
516 nested_context.parent = context;
517
518 /* loop through parts */
519 while(1) {
520
521 /* reset all per-part mime variables */
522 mime_anomaly_level = 0;
523 mime_anomaly_text = NULL;
524 mime_boundary = NULL;
525 mime_charset = NULL;
526 mime_decoded_filename = NULL;
527 mime_filename = NULL;
528 mime_content_description = NULL;
529 mime_content_disposition = NULL;
530 mime_content_id = NULL;
531 mime_content_transfer_encoding = NULL;
532 mime_content_type = NULL;
533 mime_is_multipart = 0;
534 mime_content_size = 0;
535
536 /*
537 If boundary is null, we assume that *f is positioned on the start of headers (for example,
538 at the very beginning of a message.
539 If a boundary is given, we must first advance to it to reach the start of the next header
540 block.
541 */
542
543 /* NOTE -- there's an error here -- RFC2046 specifically says to
544 * check for outer boundaries. This code doesn't do that, and
545 * I haven't fixed this.
546 *
547 * (I have moved partway towards adding support, however, by adding
548 * a "parent" field to my new boundary-context structure.)
549 */
550 if (context != NULL) {
551 while(fgets(CS header, MIME_MAX_HEADER_SIZE, f) != NULL) {
552 /* boundary line must start with 2 dashes */
553 if (Ustrncmp(header,"--",2) == 0) {
554 if (Ustrncmp((header+2),context->boundary,Ustrlen(context->boundary)) == 0) {
555 /* found boundary */
556 if (Ustrncmp((header+2+Ustrlen(context->boundary)),"--",2) == 0) {
557 /* END boundary found */
558 debug_printf("End boundary found %s\n", context->boundary);
559 return rc;
560 }
561 else {
562 debug_printf("Next part with boundary %s\n", context->boundary);
563 };
564 /* can't use break here */
565 goto DECODE_HEADERS;
566 }
567 };
568 }
569 /* Hit EOF or read error. Ugh. */
570 debug_printf("Hit EOF ...\n");
571 return rc;
572 };
573
574 DECODE_HEADERS:
575 /* parse headers, set up expansion variables */
576 while(mime_get_header(f,header)) {
577 int i;
578 /* loop through header list */
579 for (i = 0; i < mime_header_list_size; i++) {
580 uschar *header_value = NULL;
581 int header_value_len = 0;
582
583 /* found an interesting header? */
584 if (strncmpic(mime_header_list[i].name,header,mime_header_list[i].namelen) == 0) {
585 uschar *p = header + mime_header_list[i].namelen;
586 /* yes, grab the value (normalize to lower case)
587 and copy to its corresponding expansion variable */
588 while(*p != ';') {
589 *p = tolower(*p);
590 p++;
591 };
592 header_value_len = (p - (header + mime_header_list[i].namelen));
593 header_value = (uschar *)malloc(header_value_len+1);
594 memset(header_value,0,header_value_len+1);
595 p = header + mime_header_list[i].namelen;
596 Ustrncpy(header_value, p, header_value_len);
597 debug_printf("Found %s MIME header, value is '%s'\n", mime_header_list[i].name, header_value);
598 *((uschar **)(mime_header_list[i].value)) = header_value;
599
600 /* make p point to the next character after the closing ';' */
601 p += (header_value_len+1);
602
603 /* grab all param=value tags on the remaining line, check if they are interesting */
604 NEXT_PARAM_SEARCH: while (*p != 0) {
605 int j;
606 for (j = 0; j < mime_parameter_list_size; j++) {
607 uschar *param_value = NULL;
608 int param_value_len = 0;
609
610 /* found an interesting parameter? */
611 if (strncmpic(mime_parameter_list[j].name,p,mime_parameter_list[j].namelen) == 0) {
612 uschar *q = p + mime_parameter_list[j].namelen;
613 /* yes, grab the value and copy to its corresponding expansion variable */
614 while(*q != ';') q++;
615 param_value_len = (q - (p + mime_parameter_list[j].namelen));
616 param_value = (uschar *)malloc(param_value_len+1);
617 memset(param_value,0,param_value_len+1);
618 q = p + mime_parameter_list[j].namelen;
619 Ustrncpy(param_value, q, param_value_len);
620 param_value = rfc2047_decode(param_value, TRUE, NULL, 32, &param_value_len, &q);
621 debug_printf("Found %s MIME parameter in %s header, value is '%s'\n", mime_parameter_list[j].name, mime_header_list[i].name, param_value);
622 *((uschar **)(mime_parameter_list[j].value)) = param_value;
623 p += (mime_parameter_list[j].namelen + param_value_len + 1);
624 goto NEXT_PARAM_SEARCH;
625 };
626 }
627 /* There is something, but not one of our interesting parameters.
628 Advance to the next semicolon */
629 while(*p != ';') p++;
630 p++;
631 };
632 };
633 };
634 };
635
636 /* set additional flag variables (easier access) */
637 if ( (mime_content_type != NULL) &&
638 (Ustrncmp(mime_content_type,"multipart",9) == 0) )
639 mime_is_multipart = 1;
640
641 /* Make a copy of the boundary pointer.
642 Required since mime_boundary is global
643 and can be overwritten further down in recursion */
644 nested_context.boundary = mime_boundary;
645
646 /* raise global counter */
647 mime_part_count++;
648
649 /* copy current file handle to global variable */
650 mime_stream = f;
651 mime_current_boundary = context ? context->boundary : 0;
652
653 /* Note the context */
654 mime_is_coverletter = !(context && context->context == MBC_ATTACHMENT);
655
656 /* call ACL handling function */
657 rc = acl_check(ACL_WHERE_MIME, NULL, acl_smtp_mime, user_msgptr, log_msgptr);
658
659 mime_stream = NULL;
660 mime_current_boundary = NULL;
661
662 if (rc != OK) break;
663
664 /* If we have a multipart entity and a boundary, go recursive */
665 if ( (mime_content_type != NULL) &&
666 (nested_context.boundary != NULL) &&
667 (Ustrncmp(mime_content_type,"multipart",9) == 0) ) {
668 debug_printf("Entering multipart recursion, boundary '%s'\n", nested_context.boundary);
669
670 if (context && context->context == MBC_ATTACHMENT)
671 nested_context.context = MBC_ATTACHMENT;
672 else if (!Ustrcmp(mime_content_type,"multipart/alternative")
673 || !Ustrcmp(mime_content_type,"multipart/related"))
674 nested_context.context = MBC_COVERLETTER_ALL;
675 else
676 nested_context.context = MBC_COVERLETTER_ONESHOT;
677
678 rc = mime_acl_check(f, &nested_context, user_msgptr, log_msgptr);
679 if (rc != OK) break;
680 }
681 else if ( (mime_content_type != NULL) &&
682 (Ustrncmp(mime_content_type,"message/rfc822",14) == 0) ) {
683 uschar *rfc822name = NULL;
684 uschar filename[2048];
685 int file_nr = 0;
686 int result = 0;
687
688 /* must find first free sequential filename */
689 do {
690 struct stat mystat;
691 snprintf(CS filename,2048,"%s/scan/%s/__rfc822_%05u", spool_directory, message_id, file_nr);
692 file_nr++;
693 /* security break */
694 if (file_nr >= 128)
695 goto NO_RFC822;
696 result = stat(CS filename,&mystat);
697 }
698 while(result != -1);
699
700 rfc822name = filename;
701
702 /* decode RFC822 attachment */
703 mime_decoded_filename = NULL;
704 mime_stream = f;
705 mime_current_boundary = context ? context->boundary : NULL;
706 mime_decode(&rfc822name);
707 mime_stream = NULL;
708 mime_current_boundary = NULL;
709 if (mime_decoded_filename == NULL) {
710 /* decoding failed */
711 log_write(0, LOG_MAIN,
712 "mime_regex acl condition warning - could not decode RFC822 MIME part to file.");
713 return DEFER;
714 };
715 mime_decoded_filename = NULL;
716 };
717
718 NO_RFC822:
719 /* If the boundary of this instance is NULL, we are finished here */
720 if (context == NULL) break;
721
722 if (context->context == MBC_COVERLETTER_ONESHOT)
723 context->context = MBC_ATTACHMENT;
724
725 };
726
727 return rc;
728 }
729
730 #endif