Account properly for quoted or 2047-encoded MIME parameters while walking headers...
[exim.git] / src / src / mime.c
1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
4
5 /* Copyright (c) Tom Kistner <tom@duncanthrax.net> 2004 */
6 /* License: GPL */
7
8 #include "exim.h"
9 #ifdef WITH_CONTENT_SCAN /* entire file */
10 #include "mime.h"
11 #include <sys/stat.h>
12
13 FILE *mime_stream = NULL;
14 uschar *mime_current_boundary = NULL;
15
16 /*************************************************
17 * set MIME anomaly level + text *
18 *************************************************/
19
20 /* Small wrapper to set the two expandables which
21 give info on detected "problems" in MIME
22 encodings. Those are defined in mime.h. */
23
24 void
25 mime_set_anomaly(int level, const char *text)
26 {
27 mime_anomaly_level = level;
28 mime_anomaly_text = CUS 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 *
45 mime_decode_qp_char(uschar *qp_p, int *c)
46 {
47 uschar *initial_pos = qp_p;
48
49 /* advance one char */
50 qp_p++;
51
52 /* Check for two hex digits and decode them */
53 if (isxdigit(*qp_p) && isxdigit(qp_p[1]))
54 {
55 /* Do hex conversion */
56 *c = (isdigit(*qp_p) ? *qp_p - '0' : toupper(*qp_p) - 'A' + 10) <<4;
57 qp_p++;
58 *c |= isdigit(*qp_p) ? *qp_p - '0' : toupper(*qp_p) - 'A' + 10;
59 return qp_p + 1;
60 }
61
62 /* tab or whitespace may follow just ignore it if it precedes \n */
63 while (*qp_p == '\t' || *qp_p == ' ' || *qp_p == '\r')
64 qp_p++;
65
66 if (*qp_p == '\n') /* hit soft line break */
67 {
68 *c = -1;
69 return qp_p;
70 }
71
72 /* illegal char here */
73 *c = -2;
74 return initial_pos;
75 }
76
77
78 /* just dump MIME part without any decoding */
79 static ssize_t
80 mime_decode_asis(FILE* in, FILE* out, uschar* boundary)
81 {
82 ssize_t len, size = 0;
83 uschar buffer[MIME_MAX_LINE_LENGTH];
84
85 while(fgets(CS buffer, MIME_MAX_LINE_LENGTH, mime_stream) != NULL)
86 {
87 if (boundary != NULL
88 && Ustrncmp(buffer, "--", 2) == 0
89 && Ustrncmp((buffer+2), boundary, Ustrlen(boundary)) == 0
90 )
91 break;
92
93 len = Ustrlen(buffer);
94 if (fwrite(buffer, 1, (size_t)len, out) < len)
95 return -1;
96 size += len;
97 } /* while */
98 return size;
99 }
100
101
102 /* decode base64 MIME part */
103 static ssize_t
104 mime_decode_base64(FILE* in, FILE* out, uschar* boundary)
105 {
106 uschar ibuf[MIME_MAX_LINE_LENGTH], obuf[MIME_MAX_LINE_LENGTH];
107 uschar *ipos, *opos;
108 ssize_t len, size = 0;
109 int bytestate = 0;
110
111 opos = obuf;
112
113 while (Ufgets(ibuf, MIME_MAX_LINE_LENGTH, in) != NULL)
114 {
115 if (boundary != NULL
116 && Ustrncmp(ibuf, "--", 2) == 0
117 && Ustrncmp((ibuf+2), boundary, Ustrlen(boundary)) == 0
118 )
119 break;
120
121 for (ipos = ibuf ; *ipos != '\r' && *ipos != '\n' && *ipos != 0; ++ipos)
122 {
123 if (*ipos == '=') /* skip padding */
124 {
125 ++bytestate;
126 continue;
127 }
128 if (mime_b64[*ipos] == 128) /* skip bad characters */
129 {
130 mime_set_anomaly(MIME_ANOMALY_BROKEN_BASE64);
131 continue;
132 }
133
134 /* simple state-machine */
135 switch((bytestate++) & 3)
136 {
137 case 0:
138 *opos = mime_b64[*ipos] << 2;
139 break;
140 case 1:
141 *opos |= mime_b64[*ipos] >> 4;
142 ++opos;
143 *opos = mime_b64[*ipos] << 4;
144 break;
145 case 2:
146 *opos |= mime_b64[*ipos] >> 2;
147 ++opos;
148 *opos = mime_b64[*ipos] << 6;
149 break;
150 case 3:
151 *opos |= mime_b64[*ipos];
152 ++opos;
153 break;
154 } /* switch */
155 } /* for */
156
157 /* something to write? */
158 len = opos - obuf;
159 if (len > 0)
160 {
161 if (fwrite(obuf, 1, len, out) != len) return -1; /* error */
162 size += len;
163 /* copy incomplete last byte to start of obuf, where we continue */
164 if ((bytestate & 3) != 0)
165 *obuf = *opos;
166 opos = obuf;
167 }
168 } /* while */
169
170 /* write out last byte if it was incomplete */
171 if (bytestate & 3)
172 {
173 if (fwrite(obuf, 1, 1, out) != 1) return -1;
174 ++size;
175 }
176
177 return size;
178 }
179
180
181 /* decode quoted-printable MIME part */
182 static ssize_t
183 mime_decode_qp(FILE* in, FILE* out, uschar* boundary)
184 {
185 uschar ibuf[MIME_MAX_LINE_LENGTH], obuf[MIME_MAX_LINE_LENGTH];
186 uschar *ipos, *opos;
187 ssize_t len, size = 0;
188
189 while (fgets(CS ibuf, MIME_MAX_LINE_LENGTH, in) != NULL)
190 {
191 if (boundary != NULL
192 && Ustrncmp(ibuf, "--", 2) == 0
193 && Ustrncmp((ibuf+2), boundary, Ustrlen(boundary)) == 0
194 )
195 break; /* todo: check for missing boundary */
196
197 ipos = ibuf;
198 opos = obuf;
199
200 while (*ipos != 0)
201 {
202 if (*ipos == '=')
203 {
204 int decode_qp_result;
205
206 ipos = mime_decode_qp_char(ipos, &decode_qp_result);
207
208 if (decode_qp_result == -2)
209 {
210 /* Error from decoder. ipos is unchanged. */
211 mime_set_anomaly(MIME_ANOMALY_BROKEN_QP);
212 *opos = '=';
213 ++opos;
214 ++ipos;
215 }
216 else if (decode_qp_result == -1)
217 break;
218 else if (decode_qp_result >= 0)
219 {
220 *opos = decode_qp_result;
221 ++opos;
222 }
223 }
224 else
225 {
226 *opos = *ipos;
227 ++opos;
228 ++ipos;
229 }
230 }
231 /* something to write? */
232 len = opos - obuf;
233 if (len > 0)
234 {
235 if (fwrite(obuf, 1, len, out) != len) return -1; /* error */
236 size += len;
237 }
238 }
239 return size;
240 }
241
242
243 FILE *
244 mime_get_decode_file(uschar *pname, uschar *fname)
245 {
246 FILE *f = NULL;
247 uschar *filename;
248
249 filename = (uschar *)malloc(2048);
250
251 if (pname && fname)
252 {
253 (void)string_format(filename, 2048, "%s/%s", pname, fname);
254 f = modefopen(filename,"wb+",SPOOL_MODE);
255 }
256 else if (!pname)
257 f = modefopen(fname,"wb+",SPOOL_MODE);
258 else if (!fname)
259 {
260 int file_nr = 0;
261 int result = 0;
262
263 /* must find first free sequential filename */
264 do
265 {
266 struct stat mystat;
267 (void)string_format(filename, 2048,
268 "%s/%s-%05u", pname, message_id, file_nr++);
269 /* security break */
270 if (file_nr >= 1024)
271 break;
272 result = stat(CS filename, &mystat);
273 } while(result != -1);
274
275 f = modefopen(filename, "wb+", SPOOL_MODE);
276 }
277
278 /* set expansion variable */
279 mime_decoded_filename = filename;
280
281 return f;
282 }
283
284
285 int
286 mime_decode(uschar **listptr)
287 {
288 int sep = 0;
289 uschar *list = *listptr;
290 uschar *option;
291 uschar option_buffer[1024];
292 uschar decode_path[1024];
293 FILE *decode_file = NULL;
294 long f_pos = 0;
295 ssize_t size_counter = 0;
296 ssize_t (*decode_function)(FILE*, FILE*, uschar*);
297
298 if (mime_stream == NULL)
299 return FAIL;
300
301 f_pos = ftell(mime_stream);
302
303 /* build default decode path (will exist since MBOX must be spooled up) */
304 (void)string_format(decode_path,1024,"%s/scan/%s",spool_directory,message_id);
305
306 /* try to find 1st option */
307 if ((option = string_nextinlist(&list, &sep,
308 option_buffer,
309 sizeof(option_buffer))) != NULL)
310 {
311 /* parse 1st option */
312 if ( (Ustrcmp(option,"false") == 0) || (Ustrcmp(option,"0") == 0) )
313 /* explicitly no decoding */
314 return FAIL;
315
316 if (Ustrcmp(option,"default") == 0)
317 /* explicit default path + file names */
318 goto DEFAULT_PATH;
319
320 if (option[0] == '/')
321 {
322 struct stat statbuf;
323
324 memset(&statbuf,0,sizeof(statbuf));
325
326 /* assume either path or path+file name */
327 if ( (stat(CS option, &statbuf) == 0) && S_ISDIR(statbuf.st_mode) )
328 /* is directory, use it as decode_path */
329 decode_file = mime_get_decode_file(option, NULL);
330 else
331 /* does not exist or is a file, use as full file name */
332 decode_file = mime_get_decode_file(NULL, option);
333 }
334 else
335 /* assume file name only, use default path */
336 decode_file = mime_get_decode_file(decode_path, option);
337 }
338 else
339 {
340 /* no option? patch default path */
341 DEFAULT_PATH:
342 decode_file = mime_get_decode_file(decode_path, NULL);
343 }
344
345 if (!decode_file)
346 return DEFER;
347
348 /* decode according to mime type */
349 decode_function =
350 !mime_content_transfer_encoding
351 ? mime_decode_asis /* no encoding, dump as-is */
352 : Ustrcmp(mime_content_transfer_encoding, "base64") == 0
353 ? mime_decode_base64
354 : Ustrcmp(mime_content_transfer_encoding, "quoted-printable") == 0
355 ? mime_decode_qp
356 : mime_decode_asis; /* unknown encoding type, just dump as-is */
357
358 size_counter = decode_function(mime_stream, decode_file, mime_current_boundary);
359
360 clearerr(mime_stream);
361 fseek(mime_stream, f_pos, SEEK_SET);
362
363 if (fclose(decode_file) != 0 || size_counter < 0)
364 return DEFER;
365
366 /* round up to the next KiB */
367 mime_content_size = (size_counter + 1023) / 1024;
368
369 return OK;
370 }
371
372 int
373 mime_get_header(FILE *f, uschar *header)
374 {
375 int c = EOF;
376 int done = 0;
377 int header_value_mode = 0;
378 int header_open_brackets = 0;
379 int num_copied = 0;
380
381 while(!done)
382 {
383 if ((c = fgetc(f)) == EOF) break;
384
385 /* always skip CRs */
386 if (c == '\r') continue;
387
388 if (c == '\n')
389 {
390 if (num_copied > 0)
391 {
392 /* look if next char is '\t' or ' ' */
393 if ((c = fgetc(f)) == EOF) break;
394 if ( (c == '\t') || (c == ' ') ) continue;
395 (void)ungetc(c,f);
396 }
397 /* end of the header, terminate with ';' */
398 c = ';';
399 done = 1;
400 }
401
402 /* skip control characters */
403 if (c < 32) continue;
404
405 if (header_value_mode)
406 {
407 /* --------- value mode ----------- */
408 /* skip leading whitespace */
409 if ( ((c == '\t') || (c == ' ')) && (header_value_mode == 1) )
410 continue;
411
412 /* we have hit a non-whitespace char, start copying value data */
413 header_value_mode = 2;
414
415 if (c == '"') /* flip "quoted" mode */
416 header_value_mode = header_value_mode==2 ? 3 : 2;
417
418 /* leave value mode on unquoted ';' */
419 if (header_value_mode == 2 && c == ';') {
420 header_value_mode = 0;
421 };
422 /* -------------------------------- */
423 }
424 else
425 {
426 /* -------- non-value mode -------- */
427 /* skip whitespace + tabs */
428 if ( (c == ' ') || (c == '\t') )
429 continue;
430 if (c == '\\')
431 {
432 /* quote next char. can be used
433 to escape brackets. */
434 if ((c = fgetc(f)) == EOF) break;
435 }
436 else if (c == '(')
437 {
438 header_open_brackets++;
439 continue;
440 }
441 else if ((c == ')') && header_open_brackets)
442 {
443 header_open_brackets--;
444 continue;
445 }
446 else if ( (c == '=') && !header_open_brackets ) /* enter value mode */
447 header_value_mode = 1;
448
449 /* skip chars while we are in a comment */
450 if (header_open_brackets > 0)
451 continue;
452 /* -------------------------------- */
453 }
454
455 /* copy the char to the buffer */
456 header[num_copied++] = (uschar)c;
457
458 /* break if header buffer is full */
459 if (num_copied > MIME_MAX_HEADER_SIZE-1)
460 done = 1;
461 }
462
463 if ((num_copied > 0) && (header[num_copied-1] != ';'))
464 header[num_copied-1] = ';';
465
466 /* 0-terminate */
467 header[num_copied] = '\0';
468
469 /* return 0 for EOF or empty line */
470 if ((c == EOF) || (num_copied == 1))
471 return 0;
472 else
473 return 1;
474 }
475
476
477 int
478 mime_acl_check(uschar *acl, FILE *f, struct mime_boundary_context *context,
479 uschar **user_msgptr, uschar **log_msgptr)
480 {
481 int rc = OK;
482 uschar *header = NULL;
483 struct mime_boundary_context nested_context;
484
485 /* reserve a line buffer to work in */
486 if (!(header = (uschar *)malloc(MIME_MAX_HEADER_SIZE+1)))
487 {
488 log_write(0, LOG_PANIC,
489 "MIME ACL: can't allocate %d bytes of memory.", MIME_MAX_HEADER_SIZE+1);
490 return DEFER;
491 }
492
493 /* Not actually used at the moment, but will be vital to fixing
494 * some RFC 2046 nonconformance later... */
495 nested_context.parent = context;
496
497 /* loop through parts */
498 while(1)
499 {
500 /* reset all per-part mime variables */
501 mime_anomaly_level = 0;
502 mime_anomaly_text = NULL;
503 mime_boundary = NULL;
504 mime_charset = NULL;
505 mime_decoded_filename = NULL;
506 mime_filename = NULL;
507 mime_content_description = NULL;
508 mime_content_disposition = NULL;
509 mime_content_id = NULL;
510 mime_content_transfer_encoding = NULL;
511 mime_content_type = NULL;
512 mime_is_multipart = 0;
513 mime_content_size = 0;
514
515 /*
516 If boundary is null, we assume that *f is positioned on the start of headers (for example,
517 at the very beginning of a message.
518 If a boundary is given, we must first advance to it to reach the start of the next header
519 block.
520 */
521
522 /* NOTE -- there's an error here -- RFC2046 specifically says to
523 * check for outer boundaries. This code doesn't do that, and
524 * I haven't fixed this.
525 *
526 * (I have moved partway towards adding support, however, by adding
527 * a "parent" field to my new boundary-context structure.)
528 */
529 if (context != NULL)
530 {
531 while(fgets(CS header, MIME_MAX_HEADER_SIZE, f))
532 {
533 /* boundary line must start with 2 dashes */
534 if ( Ustrncmp(header, "--", 2) == 0
535 && Ustrncmp(header+2, context->boundary, Ustrlen(context->boundary)) == 0)
536 {
537 /* found boundary */
538 if (Ustrncmp((header+2+Ustrlen(context->boundary)), "--", 2) == 0)
539 {
540 /* END boundary found */
541 debug_printf("End boundary found %s\n", context->boundary);
542 return rc;
543 }
544 else
545 debug_printf("Next part with boundary %s\n", context->boundary);
546
547 /* can't use break here */
548 goto DECODE_HEADERS;
549 }
550 }
551 /* Hit EOF or read error. Ugh. */
552 debug_printf("Hit EOF ...\n");
553 return rc;
554 }
555
556 DECODE_HEADERS:
557 /* parse headers, set up expansion variables */
558 while (mime_get_header(f, header))
559 {
560 int i;
561 /* loop through header list */
562 for (i = 0; i < mime_header_list_size; i++)
563 if (strncmpic(mime_header_list[i].name,
564 header, mime_header_list[i].namelen) == 0)
565 { /* found an interesting header */
566 uschar * header_value;
567 int header_value_len;
568 uschar * p = header + mime_header_list[i].namelen;
569
570 /* grab the value (normalize to lower case)
571 and copy to its corresponding expansion variable */
572 while(*p != ';')
573 {
574 *p = tolower(*p);
575 p++;
576 }
577 header_value_len = p - (header + mime_header_list[i].namelen);
578 p = header + mime_header_list[i].namelen;
579 header_value = string_copyn(p, header_value_len);
580 debug_printf("Found %s MIME header, value is '%s'\n",
581 mime_header_list[i].name, header_value);
582 *((uschar **)(mime_header_list[i].value)) = header_value;
583
584 /* make p point to the next character after the closing ';' */
585 p += header_value_len+1;
586
587 /* grab all param=value tags on the remaining line,
588 check if they are interesting */
589 NEXT_PARAM_SEARCH:
590 while (*p)
591 {
592 mime_parameter * mp;
593 for (mp = mime_parameter_list;
594 mp < &mime_parameter_list[mime_parameter_list_size];
595 mp++)
596 {
597 uschar * param_value = NULL;
598
599 /* found an interesting parameter? */
600 if (strncmpic(mp->name, p, mp->namelen) == 0)
601 {
602 uschar * q = p + mp->namelen;
603 int plen = 0;
604 int size = 0;
605 int ptr = 0;
606
607 /* yes, grab the value and copy to its corresponding expansion variable */
608 while(*q && *q != ';') /* ; terminates */
609 if (*q == '"')
610 {
611 q++; /* skip leading " */
612 plen++; /* and account for the skip */
613 while(*q && *q != '"') /* " protects ; */
614 {
615 param_value = string_cat(param_value, &size, &ptr, q++, 1);
616 plen++;
617 }
618 if (*q)
619 {
620 q++; /* skip trailing " */
621 plen++;
622 }
623 }
624 else
625 {
626 param_value = string_cat(param_value, &size, &ptr, q++, 1);
627 plen++;
628 }
629
630 if (param_value)
631 {
632 param_value[ptr++] = '\0';
633
634 param_value = rfc2047_decode(param_value,
635 check_rfc2047_length, NULL, 32, NULL, &q);
636 debug_printf("Found %s MIME parameter in %s header, "
637 "value is '%s'\n", mp->name, mime_header_list[i].name,
638 param_value);
639 }
640 *mp->value = param_value;
641 p += mp->namelen + plen + 1; /* name=, content, ; */
642 goto NEXT_PARAM_SEARCH;
643 }
644 }
645 /* There is something, but not one of our interesting parameters.
646 Advance to the next semicolon */
647 while(*p != ';')
648 {
649 if (*p == '"') while(*++p && *p != '"') ;
650 p++;
651 }
652 p++;
653 }
654 }
655 }
656
657 /* set additional flag variables (easier access) */
658 if ( (mime_content_type != NULL) &&
659 (Ustrncmp(mime_content_type,"multipart",9) == 0) )
660 mime_is_multipart = 1;
661
662 /* Make a copy of the boundary pointer.
663 Required since mime_boundary is global
664 and can be overwritten further down in recursion */
665 nested_context.boundary = mime_boundary;
666
667 /* raise global counter */
668 mime_part_count++;
669
670 /* copy current file handle to global variable */
671 mime_stream = f;
672 mime_current_boundary = context ? context->boundary : 0;
673
674 /* Note the context */
675 mime_is_coverletter = !(context && context->context == MBC_ATTACHMENT);
676
677 /* call ACL handling function */
678 rc = acl_check(ACL_WHERE_MIME, NULL, acl, user_msgptr, log_msgptr);
679
680 mime_stream = NULL;
681 mime_current_boundary = NULL;
682
683 if (rc != OK) break;
684
685 /* If we have a multipart entity and a boundary, go recursive */
686 if ( (mime_content_type != NULL) &&
687 (nested_context.boundary != NULL) &&
688 (Ustrncmp(mime_content_type,"multipart",9) == 0) )
689 {
690 debug_printf("Entering multipart recursion, boundary '%s'\n", nested_context.boundary);
691
692 nested_context.context =
693 context && context->context == MBC_ATTACHMENT
694 ? MBC_ATTACHMENT
695 : Ustrcmp(mime_content_type,"multipart/alternative") == 0
696 || Ustrcmp(mime_content_type,"multipart/related") == 0
697 ? MBC_COVERLETTER_ALL
698 : MBC_COVERLETTER_ONESHOT;
699
700 rc = mime_acl_check(acl, f, &nested_context, user_msgptr, log_msgptr);
701 if (rc != OK) break;
702 }
703 else if ( (mime_content_type != NULL) &&
704 (Ustrncmp(mime_content_type,"message/rfc822",14) == 0) )
705 {
706 uschar *rfc822name = NULL;
707 uschar filename[2048];
708 int file_nr = 0;
709 int result = 0;
710
711 /* must find first free sequential filename */
712 do
713 {
714 struct stat mystat;
715 (void)string_format(filename, 2048,
716 "%s/scan/%s/__rfc822_%05u", spool_directory, message_id, file_nr++);
717 /* security break */
718 if (file_nr >= 128)
719 goto NO_RFC822;
720 result = stat(CS filename,&mystat);
721 } while (result != -1);
722
723 rfc822name = filename;
724
725 /* decode RFC822 attachment */
726 mime_decoded_filename = NULL;
727 mime_stream = f;
728 mime_current_boundary = context ? context->boundary : NULL;
729 mime_decode(&rfc822name);
730 mime_stream = NULL;
731 mime_current_boundary = NULL;
732 if (!mime_decoded_filename) /* decoding failed */
733 {
734 log_write(0, LOG_MAIN,
735 "mime_regex acl condition warning - could not decode RFC822 MIME part to file.");
736 return DEFER;
737 }
738 mime_decoded_filename = NULL;
739 }
740
741 NO_RFC822:
742 /* If the boundary of this instance is NULL, we are finished here */
743 if (context == NULL) break;
744
745 if (context->context == MBC_COVERLETTER_ONESHOT)
746 context->context = MBC_ATTACHMENT;
747 }
748
749 return rc;
750 }
751
752 #endif /*WITH_CONTENT_SCAN*/
753
754 /* vi: sw ai sw=2
755 */