Update all copyright messages to cover 1995 - 2009. Remove tab from exim_checkaccess.src
[exim.git] / src / src / mime.c
1 /* $Cambridge: exim/src/src/mime.c,v 1.19 2009/11/11 10:08:02 nm4 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 uschar *mime_decode_qp_char(uschar *qp_p, int *c) {
45 uschar *initial_pos = qp_p;
46
47 /* advance one char */
48 qp_p++;
49
50 /* Check for two hex digits and decode them */
51 if (isxdigit(*qp_p) && isxdigit(qp_p[1])) {
52 /* Do hex conversion */
53 if (isdigit(*qp_p)) {*c = *qp_p - '0';}
54 else {*c = toupper(*qp_p) - 'A' + 10;};
55 *c <<= 4;
56 if (isdigit(qp_p[1])) {*c |= qp_p[1] - '0';}
57 else {*c |= toupper(qp_p[1]) - 'A' + 10;};
58 return qp_p + 2;
59 };
60
61 /* tab or whitespace may follow just ignore it if it precedes \n */
62 while (*qp_p == '\t' || *qp_p == ' ' || *qp_p == '\r')
63 qp_p++;
64
65 if (*qp_p == '\n') {
66 /* hit soft line break */
67 *c = -1;
68 return qp_p;
69 };
70
71 /* illegal char here */
72 *c = -2;
73 return initial_pos;
74 }
75
76
77 /* just dump MIME part without any decoding */
78 static int mime_decode_asis(FILE* in, FILE* out, uschar* boundary)
79 {
80 int len, size = 0;
81 uschar buffer[MIME_MAX_LINE_LENGTH];
82
83 while(fgets(CS buffer, MIME_MAX_LINE_LENGTH, mime_stream) != NULL) {
84 if (boundary != NULL
85 && Ustrncmp(buffer, "--", 2) == 0
86 && Ustrncmp((buffer+2), boundary, Ustrlen(boundary)) == 0
87 )
88 break;
89
90 len = Ustrlen(buffer);
91 if (fwrite(buffer, 1, (size_t)len, out) < len)
92 return -1;
93 size += len;
94 } /* while */
95 return size;
96 }
97
98
99 /* decode base64 MIME part */
100 static int mime_decode_base64(FILE* in, FILE* out, uschar* boundary)
101 {
102 uschar ibuf[MIME_MAX_LINE_LENGTH], obuf[MIME_MAX_LINE_LENGTH];
103 uschar *ipos, *opos;
104 size_t len, size = 0;
105 int bytestate = 0;
106
107 opos = obuf;
108
109 while (Ufgets(ibuf, MIME_MAX_LINE_LENGTH, in) != NULL)
110 {
111 if (boundary != NULL
112 && Ustrncmp(ibuf, "--", 2) == 0
113 && Ustrncmp((ibuf+2), boundary, Ustrlen(boundary)) == 0
114 )
115 break;
116
117 for (ipos = ibuf ; *ipos != '\r' && *ipos != '\n' && *ipos != 0; ++ipos) {
118 /* skip padding */
119 if (*ipos == '=') {
120 ++bytestate;
121 continue;
122 }
123 /* skip bad characters */
124 if (mime_b64[*ipos] == 128) {
125 mime_set_anomaly(MIME_ANOMALY_BROKEN_BASE64);
126 continue;
127 }
128 /* simple state-machine */
129 switch((bytestate++) & 3) {
130 case 0:
131 *opos = mime_b64[*ipos] << 2;
132 break;
133 case 1:
134 *opos |= mime_b64[*ipos] >> 4;
135 ++opos;
136 *opos = mime_b64[*ipos] << 4;
137 break;
138 case 2:
139 *opos |= mime_b64[*ipos] >> 2;
140 ++opos;
141 *opos = mime_b64[*ipos] << 6;
142 break;
143 case 3:
144 *opos |= mime_b64[*ipos];
145 ++opos;
146 break;
147 } /* switch */
148 } /* for */
149 /* something to write? */
150 len = opos - obuf;
151 if (len > 0) {
152 if (fwrite(obuf, 1, len, out) != len)
153 return -1; /* error */
154 size += len;
155 /* copy incomplete last byte to start of obuf, where we continue */
156 if ((bytestate & 3) != 0)
157 *obuf = *opos;
158 opos = obuf;
159 }
160 } /* while */
161
162 /* write out last byte if it was incomplete */
163 if (bytestate & 3) {
164 if (fwrite(obuf, 1, 1, out) != 1)
165 return -1;
166 ++size;
167 }
168
169 return size;
170 }
171
172
173 /* decode quoted-printable MIME part */
174 static int mime_decode_qp(FILE* in, FILE* out, uschar* boundary)
175 {
176 uschar ibuf[MIME_MAX_LINE_LENGTH], obuf[MIME_MAX_LINE_LENGTH];
177 uschar *ipos, *opos;
178 size_t len, size = 0;
179
180 while (fgets(CS ibuf, MIME_MAX_LINE_LENGTH, in) != NULL)
181 {
182 if (boundary != NULL
183 && Ustrncmp(ibuf, "--", 2) == 0
184 && Ustrncmp((ibuf+2), boundary, Ustrlen(boundary)) == 0
185 )
186 break; /* todo: check for missing boundary */
187
188 ipos = ibuf;
189 opos = obuf;
190
191 while (*ipos != 0) {
192 if (*ipos == '=') {
193 int decode_qp_result;
194
195 ipos = mime_decode_qp_char(ipos, &decode_qp_result);
196
197 if (decode_qp_result == -2) {
198 /* Error from decoder. ipos is unchanged. */
199 mime_set_anomaly(MIME_ANOMALY_BROKEN_QP);
200 *opos = '=';
201 ++opos;
202 ++ipos;
203 }
204 else if (decode_qp_result == -1) {
205 break;
206 }
207 else if (decode_qp_result >= 0) {
208 *opos = decode_qp_result;
209 ++opos;
210 }
211 }
212 else {
213 *opos = *ipos;
214 ++opos;
215 ++ipos;
216 }
217 }
218 /* something to write? */
219 len = opos - obuf;
220 if (len > 0) {
221 if (fwrite(obuf, 1, len, out) != len)
222 return -1; /* error */
223 size += len;
224 }
225 }
226 return size;
227 }
228
229
230 FILE *mime_get_decode_file(uschar *pname, uschar *fname) {
231 FILE *f = NULL;
232 uschar *filename;
233
234 filename = (uschar *)malloc(2048);
235
236 if ((pname != NULL) && (fname != NULL)) {
237 (void)string_format(filename, 2048, "%s/%s", pname, fname);
238 f = modefopen(filename,"wb+",SPOOL_MODE);
239 }
240 else if (pname == NULL) {
241 f = modefopen(fname,"wb+",SPOOL_MODE);
242 }
243 else if (fname == NULL) {
244 int file_nr = 0;
245 int result = 0;
246
247 /* must find first free sequential filename */
248 do {
249 struct stat mystat;
250 (void)string_format(filename,2048,"%s/%s-%05u", pname, message_id, file_nr);
251 file_nr++;
252 /* security break */
253 if (file_nr >= 1024)
254 break;
255 result = stat(CS filename,&mystat);
256 }
257 while(result != -1);
258 f = modefopen(filename,"wb+",SPOOL_MODE);
259 };
260
261 /* set expansion variable */
262 mime_decoded_filename = filename;
263
264 return f;
265 }
266
267
268 int mime_decode(uschar **listptr) {
269 int sep = 0;
270 uschar *list = *listptr;
271 uschar *option;
272 uschar option_buffer[1024];
273 uschar decode_path[1024];
274 FILE *decode_file = NULL;
275 long f_pos = 0;
276 unsigned int size_counter = 0;
277 int (*decode_function)(FILE*, FILE*, uschar*);
278
279 if (mime_stream == NULL)
280 return FAIL;
281
282 f_pos = ftell(mime_stream);
283
284 /* build default decode path (will exist since MBOX must be spooled up) */
285 (void)string_format(decode_path,1024,"%s/scan/%s",spool_directory,message_id);
286
287 /* try to find 1st option */
288 if ((option = string_nextinlist(&list, &sep,
289 option_buffer,
290 sizeof(option_buffer))) != NULL) {
291
292 /* parse 1st option */
293 if ( (Ustrcmp(option,"false") == 0) || (Ustrcmp(option,"0") == 0) ) {
294 /* explicitly no decoding */
295 return FAIL;
296 };
297
298 if (Ustrcmp(option,"default") == 0) {
299 /* explicit default path + file names */
300 goto DEFAULT_PATH;
301 };
302
303 if (option[0] == '/') {
304 struct stat statbuf;
305
306 memset(&statbuf,0,sizeof(statbuf));
307
308 /* assume either path or path+file name */
309 if ( (stat(CS option, &statbuf) == 0) && S_ISDIR(statbuf.st_mode) )
310 /* is directory, use it as decode_path */
311 decode_file = mime_get_decode_file(option, NULL);
312 else
313 /* does not exist or is a file, use as full file name */
314 decode_file = mime_get_decode_file(NULL, option);
315 }
316 else
317 /* assume file name only, use default path */
318 decode_file = mime_get_decode_file(decode_path, option);
319 }
320 else
321 /* no option? patch default path */
322 DEFAULT_PATH: decode_file = mime_get_decode_file(decode_path, NULL);
323
324 if (decode_file == NULL)
325 return DEFER;
326
327 /* decode according to mime type */
328 if (mime_content_transfer_encoding == NULL)
329 /* no encoding, dump as-is */
330 decode_function = mime_decode_asis;
331 else if (Ustrcmp(mime_content_transfer_encoding, "base64") == 0)
332 decode_function = mime_decode_base64;
333 else if (Ustrcmp(mime_content_transfer_encoding, "quoted-printable") == 0)
334 decode_function = mime_decode_qp;
335 else
336 /* unknown encoding type, just dump as-is */
337 decode_function = mime_decode_asis;
338
339 size_counter = decode_function(mime_stream, decode_file, mime_current_boundary);
340
341 clearerr(mime_stream);
342 fseek(mime_stream, f_pos, SEEK_SET);
343
344 if (fclose(decode_file) != 0 || size_counter < 0)
345 return DEFER;
346
347 /* round up to the next KiB */
348 mime_content_size = (size_counter + 1023) / 1024;
349
350 return OK;
351 }
352
353 int mime_get_header(FILE *f, uschar *header) {
354 int c = EOF;
355 int done = 0;
356 int header_value_mode = 0;
357 int header_open_brackets = 0;
358 int num_copied = 0;
359
360 while(!done) {
361
362 c = fgetc(f);
363 if (c == EOF) break;
364
365 /* always skip CRs */
366 if (c == '\r') continue;
367
368 if (c == '\n') {
369 if (num_copied > 0) {
370 /* look if next char is '\t' or ' ' */
371 c = fgetc(f);
372 if (c == EOF) break;
373 if ( (c == '\t') || (c == ' ') ) continue;
374 (void)ungetc(c,f);
375 };
376 /* end of the header, terminate with ';' */
377 c = ';';
378 done = 1;
379 };
380
381 /* skip control characters */
382 if (c < 32) continue;
383
384 if (header_value_mode) {
385 /* --------- value mode ----------- */
386 /* skip leading whitespace */
387 if ( ((c == '\t') || (c == ' ')) && (header_value_mode == 1) )
388 continue;
389
390 /* we have hit a non-whitespace char, start copying value data */
391 header_value_mode = 2;
392
393 /* skip quotes */
394 if (c == '"') continue;
395
396 /* leave value mode on ';' */
397 if (c == ';') {
398 header_value_mode = 0;
399 };
400 /* -------------------------------- */
401 }
402 else {
403 /* -------- non-value mode -------- */
404 /* skip whitespace + tabs */
405 if ( (c == ' ') || (c == '\t') )
406 continue;
407 if (c == '\\') {
408 /* quote next char. can be used
409 to escape brackets. */
410 c = fgetc(f);
411 if (c == EOF) break;
412 }
413 else if (c == '(') {
414 header_open_brackets++;
415 continue;
416 }
417 else if ((c == ')') && header_open_brackets) {
418 header_open_brackets--;
419 continue;
420 }
421 else if ( (c == '=') && !header_open_brackets ) {
422 /* enter value mode */
423 header_value_mode = 1;
424 };
425
426 /* skip chars while we are in a comment */
427 if (header_open_brackets > 0)
428 continue;
429 /* -------------------------------- */
430 };
431
432 /* copy the char to the buffer */
433 header[num_copied] = (uschar)c;
434 /* raise counter */
435 num_copied++;
436
437 /* break if header buffer is full */
438 if (num_copied > MIME_MAX_HEADER_SIZE-1) {
439 done = 1;
440 };
441 };
442
443 if ((num_copied > 0) && (header[num_copied-1] != ';')) {
444 header[num_copied-1] = ';';
445 };
446
447 /* 0-terminate */
448 header[num_copied] = '\0';
449
450 /* return 0 for EOF or empty line */
451 if ((c == EOF) || (num_copied == 1))
452 return 0;
453 else
454 return 1;
455 }
456
457
458 int mime_acl_check(uschar *acl, FILE *f, struct mime_boundary_context *context,
459 uschar **user_msgptr, uschar **log_msgptr) {
460 int rc = OK;
461 uschar *header = NULL;
462 struct mime_boundary_context nested_context;
463
464 /* reserve a line buffer to work in */
465 header = (uschar *)malloc(MIME_MAX_HEADER_SIZE+1);
466 if (header == NULL) {
467 log_write(0, LOG_PANIC,
468 "MIME ACL: can't allocate %d bytes of memory.", MIME_MAX_HEADER_SIZE+1);
469 return DEFER;
470 };
471
472 /* Not actually used at the moment, but will be vital to fixing
473 * some RFC 2046 nonconformance later... */
474 nested_context.parent = context;
475
476 /* loop through parts */
477 while(1) {
478
479 /* reset all per-part mime variables */
480 mime_anomaly_level = 0;
481 mime_anomaly_text = NULL;
482 mime_boundary = NULL;
483 mime_charset = NULL;
484 mime_decoded_filename = NULL;
485 mime_filename = NULL;
486 mime_content_description = NULL;
487 mime_content_disposition = NULL;
488 mime_content_id = NULL;
489 mime_content_transfer_encoding = NULL;
490 mime_content_type = NULL;
491 mime_is_multipart = 0;
492 mime_content_size = 0;
493
494 /*
495 If boundary is null, we assume that *f is positioned on the start of headers (for example,
496 at the very beginning of a message.
497 If a boundary is given, we must first advance to it to reach the start of the next header
498 block.
499 */
500
501 /* NOTE -- there's an error here -- RFC2046 specifically says to
502 * check for outer boundaries. This code doesn't do that, and
503 * I haven't fixed this.
504 *
505 * (I have moved partway towards adding support, however, by adding
506 * a "parent" field to my new boundary-context structure.)
507 */
508 if (context != NULL) {
509 while(fgets(CS header, MIME_MAX_HEADER_SIZE, f) != NULL) {
510 /* boundary line must start with 2 dashes */
511 if (Ustrncmp(header,"--",2) == 0) {
512 if (Ustrncmp((header+2),context->boundary,Ustrlen(context->boundary)) == 0) {
513 /* found boundary */
514 if (Ustrncmp((header+2+Ustrlen(context->boundary)),"--",2) == 0) {
515 /* END boundary found */
516 debug_printf("End boundary found %s\n", context->boundary);
517 return rc;
518 }
519 else {
520 debug_printf("Next part with boundary %s\n", context->boundary);
521 };
522 /* can't use break here */
523 goto DECODE_HEADERS;
524 }
525 };
526 }
527 /* Hit EOF or read error. Ugh. */
528 debug_printf("Hit EOF ...\n");
529 return rc;
530 };
531
532 DECODE_HEADERS:
533 /* parse headers, set up expansion variables */
534 while(mime_get_header(f,header)) {
535 int i;
536 /* loop through header list */
537 for (i = 0; i < mime_header_list_size; i++) {
538 uschar *header_value = NULL;
539 int header_value_len = 0;
540
541 /* found an interesting header? */
542 if (strncmpic(mime_header_list[i].name,header,mime_header_list[i].namelen) == 0) {
543 uschar *p = header + mime_header_list[i].namelen;
544 /* yes, grab the value (normalize to lower case)
545 and copy to its corresponding expansion variable */
546 while(*p != ';') {
547 *p = tolower(*p);
548 p++;
549 };
550 header_value_len = (p - (header + mime_header_list[i].namelen));
551 header_value = (uschar *)malloc(header_value_len+1);
552 memset(header_value,0,header_value_len+1);
553 p = header + mime_header_list[i].namelen;
554 Ustrncpy(header_value, p, header_value_len);
555 debug_printf("Found %s MIME header, value is '%s'\n", mime_header_list[i].name, header_value);
556 *((uschar **)(mime_header_list[i].value)) = header_value;
557
558 /* make p point to the next character after the closing ';' */
559 p += (header_value_len+1);
560
561 /* grab all param=value tags on the remaining line, check if they are interesting */
562 NEXT_PARAM_SEARCH: while (*p != 0) {
563 int j;
564 for (j = 0; j < mime_parameter_list_size; j++) {
565 uschar *param_value = NULL;
566 int param_value_len = 0;
567
568 /* found an interesting parameter? */
569 if (strncmpic(mime_parameter_list[j].name,p,mime_parameter_list[j].namelen) == 0) {
570 uschar *q = p + mime_parameter_list[j].namelen;
571 /* yes, grab the value and copy to its corresponding expansion variable */
572 while(*q != ';') q++;
573 param_value_len = (q - (p + mime_parameter_list[j].namelen));
574 param_value = (uschar *)malloc(param_value_len+1);
575 memset(param_value,0,param_value_len+1);
576 q = p + mime_parameter_list[j].namelen;
577 Ustrncpy(param_value, q, param_value_len);
578 param_value = rfc2047_decode(param_value, check_rfc2047_length, NULL, 32, &param_value_len, &q);
579 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);
580 *((uschar **)(mime_parameter_list[j].value)) = param_value;
581 p += (mime_parameter_list[j].namelen + param_value_len + 1);
582 goto NEXT_PARAM_SEARCH;
583 };
584 }
585 /* There is something, but not one of our interesting parameters.
586 Advance to the next semicolon */
587 while(*p != ';') p++;
588 p++;
589 };
590 };
591 };
592 };
593
594 /* set additional flag variables (easier access) */
595 if ( (mime_content_type != NULL) &&
596 (Ustrncmp(mime_content_type,"multipart",9) == 0) )
597 mime_is_multipart = 1;
598
599 /* Make a copy of the boundary pointer.
600 Required since mime_boundary is global
601 and can be overwritten further down in recursion */
602 nested_context.boundary = mime_boundary;
603
604 /* raise global counter */
605 mime_part_count++;
606
607 /* copy current file handle to global variable */
608 mime_stream = f;
609 mime_current_boundary = context ? context->boundary : 0;
610
611 /* Note the context */
612 mime_is_coverletter = !(context && context->context == MBC_ATTACHMENT);
613
614 /* call ACL handling function */
615 rc = acl_check(ACL_WHERE_MIME, NULL, acl, user_msgptr, log_msgptr);
616
617 mime_stream = NULL;
618 mime_current_boundary = NULL;
619
620 if (rc != OK) break;
621
622 /* If we have a multipart entity and a boundary, go recursive */
623 if ( (mime_content_type != NULL) &&
624 (nested_context.boundary != NULL) &&
625 (Ustrncmp(mime_content_type,"multipart",9) == 0) ) {
626 debug_printf("Entering multipart recursion, boundary '%s'\n", nested_context.boundary);
627
628 if (context && context->context == MBC_ATTACHMENT)
629 nested_context.context = MBC_ATTACHMENT;
630 else if (!Ustrcmp(mime_content_type,"multipart/alternative")
631 || !Ustrcmp(mime_content_type,"multipart/related"))
632 nested_context.context = MBC_COVERLETTER_ALL;
633 else
634 nested_context.context = MBC_COVERLETTER_ONESHOT;
635
636 rc = mime_acl_check(acl, f, &nested_context, user_msgptr, log_msgptr);
637 if (rc != OK) break;
638 }
639 else if ( (mime_content_type != NULL) &&
640 (Ustrncmp(mime_content_type,"message/rfc822",14) == 0) ) {
641 uschar *rfc822name = NULL;
642 uschar filename[2048];
643 int file_nr = 0;
644 int result = 0;
645
646 /* must find first free sequential filename */
647 do {
648 struct stat mystat;
649 (void)string_format(filename,2048,"%s/scan/%s/__rfc822_%05u", spool_directory, message_id, file_nr);
650 file_nr++;
651 /* security break */
652 if (file_nr >= 128)
653 goto NO_RFC822;
654 result = stat(CS filename,&mystat);
655 }
656 while(result != -1);
657
658 rfc822name = filename;
659
660 /* decode RFC822 attachment */
661 mime_decoded_filename = NULL;
662 mime_stream = f;
663 mime_current_boundary = context ? context->boundary : NULL;
664 mime_decode(&rfc822name);
665 mime_stream = NULL;
666 mime_current_boundary = NULL;
667 if (mime_decoded_filename == NULL) {
668 /* decoding failed */
669 log_write(0, LOG_MAIN,
670 "mime_regex acl condition warning - could not decode RFC822 MIME part to file.");
671 return DEFER;
672 };
673 mime_decoded_filename = NULL;
674 };
675
676 NO_RFC822:
677 /* If the boundary of this instance is NULL, we are finished here */
678 if (context == NULL) break;
679
680 if (context->context == MBC_COVERLETTER_ONESHOT)
681 context->context = MBC_ATTACHMENT;
682
683 };
684
685 return rc;
686 }
687
688 #endif