1 /* $NetBSD: apprentice.c,v 1.14 2015/01/02 21:15:32 christos Exp $ */
2
3 /*
4 * Copyright (c) Ian F. Darwin 1986-1995.
5 * Software written by Ian F. Darwin and others;
6 * maintained 1995-present by Christos Zoulas and others.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice immediately at the beginning of the file, without modification,
13 * this list of conditions, and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
22 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * SUCH DAMAGE.
29 */
30 /*
31 * apprentice - make one pass through /etc/magic, learning its secrets.
32 */
33
34 #include "file.h"
35
36 #ifndef lint
37 #if 0
38 FILE_RCSID("@(#)$File: apprentice.c,v 1.229 2015/01/01 17:07:34 christos Exp $")
39 #else
40 __RCSID("$NetBSD: apprentice.c,v 1.14 2015/01/02 21:15:32 christos Exp $");
41 #endif
42 #endif /* lint */
43
44 #include "magic.h"
45 #include <stdlib.h>
46 #ifdef HAVE_UNISTD_H
47 #include <unistd.h>
48 #endif
49 #ifdef HAVE_STDDEF_H
50 #include <stddef.h>
51 #endif
52 #include <string.h>
53 #include <assert.h>
54 #include <ctype.h>
55 #include <fcntl.h>
56 #ifdef QUICK
57 #include <sys/mman.h>
58 #endif
59 #include <dirent.h>
60 #if defined(HAVE_LIMITS_H)
61 #include <limits.h>
62 #endif
63
64 #ifndef SSIZE_MAX
65 #define MAXMAGIC_SIZE ((ssize_t)0x7fffffff)
66 #else
67 #define MAXMAGIC_SIZE SSIZE_MAX
68 #endif
69
70 #define EATAB {while (isascii((unsigned char) *l) && \
71 isspace((unsigned char) *l)) ++l;}
72 #define LOWCASE(l) (isupper((unsigned char) (l)) ? \
73 tolower((unsigned char) (l)) : (l))
74 /*
75 * Work around a bug in headers on Digital Unix.
76 * At least confirmed for: OSF1 V4.0 878
77 */
78 #if defined(__osf__) && defined(__DECC)
79 #ifdef MAP_FAILED
80 #undef MAP_FAILED
81 #endif
82 #endif
83
84 #ifndef MAP_FAILED
85 #define MAP_FAILED (void *) -1
86 #endif
87
88 #ifndef MAP_FILE
89 #define MAP_FILE 0
90 #endif
91
92 #define ALLOC_CHUNK (size_t)10
93 #define ALLOC_INCR (size_t)200
94
95 #define MAP_TYPE_MMAP 0
96 #define MAP_TYPE_MALLOC 1
97 #define MAP_TYPE_USER 2
98
99 struct magic_entry {
100 struct magic *mp;
101 uint32_t cont_count;
102 uint32_t max_count;
103 };
104
105 struct magic_entry_set {
106 struct magic_entry *me;
107 uint32_t count;
108 uint32_t max;
109 };
110
111 struct magic_map {
112 void *p;
113 size_t len;
114 int type;
115 struct magic *magic[MAGIC_SETS];
116 uint32_t nmagic[MAGIC_SETS];
117 };
118
119 int file_formats[FILE_NAMES_SIZE];
120 const size_t file_nformats = FILE_NAMES_SIZE;
121 const char *file_names[FILE_NAMES_SIZE];
122 const size_t file_nnames = FILE_NAMES_SIZE;
123
124 private int getvalue(struct magic_set *ms, struct magic *, const char **, int);
125 private int hextoint(int);
126 private const char *getstr(struct magic_set *, struct magic *, const char *,
127 int);
128 private int parse(struct magic_set *, struct magic_entry *, const char *,
129 size_t, int);
130 private void eatsize(const char **);
131 private int apprentice_1(struct magic_set *, const char *, int);
132 private size_t apprentice_magic_strength(const struct magic *);
133 private int apprentice_sort(const void *, const void *);
134 private void apprentice_list(struct mlist *, int );
135 private struct magic_map *apprentice_load(struct magic_set *,
136 const char *, int);
137 private struct mlist *mlist_alloc(void);
138 private void mlist_free(struct mlist *);
139 private void byteswap(struct magic *, uint32_t);
140 private void bs1(struct magic *);
141 private uint16_t swap2(uint16_t);
142 private uint32_t swap4(uint32_t);
143 private uint64_t swap8(uint64_t);
144 private char *mkdbname(struct magic_set *, const char *, int);
145 private struct magic_map *apprentice_buf(struct magic_set *, struct magic *,
146 size_t);
147 private struct magic_map *apprentice_map(struct magic_set *, const char *);
148 private int check_buffer(struct magic_set *, struct magic_map *, const char *);
149 private void apprentice_unmap(struct magic_map *);
150 private int apprentice_compile(struct magic_set *, struct magic_map *,
151 const char *);
152 private int check_format_type(const char *, int);
153 private int check_format(struct magic_set *, struct magic *);
154 private int get_op(char);
155 private int parse_mime(struct magic_set *, struct magic_entry *, const char *);
156 private int parse_strength(struct magic_set *, struct magic_entry *, const char *);
157 private int parse_apple(struct magic_set *, struct magic_entry *, const char *);
158
159
160 private size_t magicsize = sizeof(struct magic);
161
162 private const char usg_hdr[] = "cont\toffset\ttype\topcode\tmask\tvalue\tdesc";
163
164 private struct {
165 const char *name;
166 size_t len;
167 int (*fun)(struct magic_set *, struct magic_entry *, const char *);
168 } bang[] = {
169 #define DECLARE_FIELD(name) { # name, sizeof(# name) - 1, parse_ ## name }
170 DECLARE_FIELD(mime),
171 DECLARE_FIELD(apple),
172 DECLARE_FIELD(strength),
173 #undef DECLARE_FIELD
174 { NULL, 0, NULL }
175 };
176
177 #ifdef COMPILE_ONLY
178
179 int main(int, char *[]);
180
181 int
main(int argc,char * argv[])182 main(int argc, char *argv[])
183 {
184 int ret;
185 struct magic_set *ms;
186 char *progname;
187
188 if ((progname = strrchr(argv[0], '/')) != NULL)
189 progname++;
190 else
191 progname = argv[0];
192
193 if (argc != 2) {
194 (void)fprintf(stderr, "Usage: %s file\n", progname);
195 return 1;
196 }
197
198 if ((ms = magic_open(MAGIC_CHECK)) == NULL) {
199 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno));
200 return 1;
201 }
202 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0;
203 if (ret == 1)
204 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms));
205 magic_close(ms);
206 return ret;
207 }
208 #endif /* COMPILE_ONLY */
209
210 struct type_tbl_s {
211 const char name[16];
212 const size_t len;
213 const int type;
214 const int format;
215 };
216
217 /*
218 * XXX - the actual Single UNIX Specification says that "long" means "long",
219 * as in the C data type, but we treat it as meaning "4-byte integer".
220 * Given that the OS X version of file 5.04 did the same, I guess that passes
221 * the actual test; having "long" be dependent on how big a "long" is on
222 * the machine running "file" is silly.
223 */
224 static const struct type_tbl_s type_tbl[] = {
225 # define XX(s) s, (sizeof(s) - 1)
226 # define XX_NULL "", 0
227 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE },
228 { XX("byte"), FILE_BYTE, FILE_FMT_NUM },
229 { XX("short"), FILE_SHORT, FILE_FMT_NUM },
230 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE },
231 { XX("long"), FILE_LONG, FILE_FMT_NUM },
232 { XX("string"), FILE_STRING, FILE_FMT_STR },
233 { XX("date"), FILE_DATE, FILE_FMT_STR },
234 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM },
235 { XX("belong"), FILE_BELONG, FILE_FMT_NUM },
236 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR },
237 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM },
238 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM },
239 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR },
240 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR },
241 { XX("ldate"), FILE_LDATE, FILE_FMT_STR },
242 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR },
243 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR },
244 { XX("regex"), FILE_REGEX, FILE_FMT_STR },
245 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR },
246 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR },
247 { XX("search"), FILE_SEARCH, FILE_FMT_STR },
248 { XX("medate"), FILE_MEDATE, FILE_FMT_STR },
249 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR },
250 { XX("melong"), FILE_MELONG, FILE_FMT_NUM },
251 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD },
252 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD },
253 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD },
254 { XX("qdate"), FILE_QDATE, FILE_FMT_STR },
255 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR },
256 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR },
257 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR },
258 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR },
259 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR },
260 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT },
261 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT },
262 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT },
263 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE },
264 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE },
265 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE },
266 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM },
267 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM },
268 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM },
269 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR },
270 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR },
271 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR },
272 { XX("name"), FILE_NAME, FILE_FMT_NONE },
273 { XX("use"), FILE_USE, FILE_FMT_NONE },
274 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE },
275 { XX_NULL, FILE_INVALID, FILE_FMT_NONE },
276 };
277
278 /*
279 * These are not types, and cannot be preceded by "u" to make them
280 * unsigned.
281 */
282 static const struct type_tbl_s special_tbl[] = {
283 { XX("name"), FILE_NAME, FILE_FMT_STR },
284 { XX("use"), FILE_USE, FILE_FMT_STR },
285 { XX_NULL, FILE_INVALID, FILE_FMT_NONE },
286 };
287 # undef XX
288 # undef XX_NULL
289
290 private int
get_type(const struct type_tbl_s * tbl,const char * l,const char ** t)291 get_type(const struct type_tbl_s *tbl, const char *l, const char **t)
292 {
293 const struct type_tbl_s *p;
294
295 for (p = tbl; p->len; p++) {
296 if (strncmp(l, p->name, p->len) == 0) {
297 if (t)
298 *t = l + p->len;
299 break;
300 }
301 }
302 return p->type;
303 }
304
305 private int
get_standard_integer_type(const char * l,const char ** t)306 get_standard_integer_type(const char *l, const char **t)
307 {
308 int type;
309
310 if (isalpha((unsigned char)l[1])) {
311 switch (l[1]) {
312 case 'C':
313 /* "dC" and "uC" */
314 type = FILE_BYTE;
315 break;
316 case 'S':
317 /* "dS" and "uS" */
318 type = FILE_SHORT;
319 break;
320 case 'I':
321 case 'L':
322 /*
323 * "dI", "dL", "uI", and "uL".
324 *
325 * XXX - the actual Single UNIX Specification says
326 * that "L" means "long", as in the C data type,
327 * but we treat it as meaning "4-byte integer".
328 * Given that the OS X version of file 5.04 did
329 * the same, I guess that passes the actual SUS
330 * validation suite; having "dL" be dependent on
331 * how big a "long" is on the machine running
332 * "file" is silly.
333 */
334 type = FILE_LONG;
335 break;
336 case 'Q':
337 /* "dQ" and "uQ" */
338 type = FILE_QUAD;
339 break;
340 default:
341 /* "d{anything else}", "u{anything else}" */
342 return FILE_INVALID;
343 }
344 l += 2;
345 } else if (isdigit((unsigned char)l[1])) {
346 /*
347 * "d{num}" and "u{num}"; we only support {num} values
348 * of 1, 2, 4, and 8 - the Single UNIX Specification
349 * doesn't say anything about whether arbitrary
350 * values should be supported, but both the Solaris 10
351 * and OS X Mountain Lion versions of file passed the
352 * Single UNIX Specification validation suite, and
353 * neither of them support values bigger than 8 or
354 * non-power-of-2 values.
355 */
356 if (isdigit((unsigned char)l[2])) {
357 /* Multi-digit, so > 9 */
358 return FILE_INVALID;
359 }
360 switch (l[1]) {
361 case '1':
362 type = FILE_BYTE;
363 break;
364 case '2':
365 type = FILE_SHORT;
366 break;
367 case '4':
368 type = FILE_LONG;
369 break;
370 case '8':
371 type = FILE_QUAD;
372 break;
373 default:
374 /* XXX - what about 3, 5, 6, or 7? */
375 return FILE_INVALID;
376 }
377 l += 2;
378 } else {
379 /*
380 * "d" or "u" by itself.
381 */
382 type = FILE_LONG;
383 ++l;
384 }
385 if (t)
386 *t = l;
387 return type;
388 }
389
390 private void
init_file_tables(void)391 init_file_tables(void)
392 {
393 static int done = 0;
394 const struct type_tbl_s *p;
395
396 if (done)
397 return;
398 done++;
399
400 for (p = type_tbl; p->len; p++) {
401 assert(p->type < FILE_NAMES_SIZE);
402 file_names[p->type] = p->name;
403 file_formats[p->type] = p->format;
404 }
405 assert(p - type_tbl == FILE_NAMES_SIZE);
406 }
407
408 private int
add_mlist(struct mlist * mlp,struct magic_map * map,size_t idx)409 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx)
410 {
411 struct mlist *ml;
412
413 mlp->map = idx == 0 ? map : NULL;
414 if ((ml = CAST(struct mlist *, malloc(sizeof(*ml)))) == NULL)
415 return -1;
416
417 ml->map = NULL;
418 ml->magic = map->magic[idx];
419 ml->nmagic = map->nmagic[idx];
420
421 mlp->prev->next = ml;
422 ml->prev = mlp->prev;
423 ml->next = mlp;
424 mlp->prev = ml;
425 return 0;
426 }
427
428 /*
429 * Handle one file or directory.
430 */
431 private int
apprentice_1(struct magic_set * ms,const char * fn,int action)432 apprentice_1(struct magic_set *ms, const char *fn, int action)
433 {
434 struct magic_map *map;
435 #ifndef COMPILE_ONLY
436 struct mlist *ml;
437 size_t i;
438 #endif
439
440 if (magicsize != FILE_MAGICSIZE) {
441 file_error(ms, 0, "magic element size %lu != %lu",
442 (unsigned long)sizeof(*map->magic[0]),
443 (unsigned long)FILE_MAGICSIZE);
444 return -1;
445 }
446
447 if (action == FILE_COMPILE) {
448 map = apprentice_load(ms, fn, action);
449 if (map == NULL)
450 return -1;
451 return apprentice_compile(ms, map, fn);
452 }
453
454 #ifndef COMPILE_ONLY
455 map = apprentice_map(ms, fn);
456 if (map == NULL) {
457 if (ms->flags & MAGIC_CHECK)
458 file_magwarn(ms, "using regular magic file `%s'", fn);
459 map = apprentice_load(ms, fn, action);
460 if (map == NULL)
461 return -1;
462 }
463
464 for (i = 0; i < MAGIC_SETS; i++) {
465 if (add_mlist(ms->mlist[i], map, i) == -1) {
466 file_oomem(ms, sizeof(*ml));
467 goto fail;
468 }
469 }
470
471 if (action == FILE_LIST) {
472 for (i = 0; i < MAGIC_SETS; i++) {
473 printf("Set %" SIZE_T_FORMAT "u:\nBinary patterns:\n",
474 i);
475 apprentice_list(ms->mlist[i], BINTEST);
476 printf("Text patterns:\n");
477 apprentice_list(ms->mlist[i], TEXTTEST);
478 }
479 }
480 return 0;
481 fail:
482 for (i = 0; i < MAGIC_SETS; i++) {
483 mlist_free(ms->mlist[i]);
484 ms->mlist[i] = NULL;
485 }
486 return -1;
487 #else
488 return 0;
489 #endif /* COMPILE_ONLY */
490 }
491
492 protected void
file_ms_free(struct magic_set * ms)493 file_ms_free(struct magic_set *ms)
494 {
495 size_t i;
496 if (ms == NULL)
497 return;
498 for (i = 0; i < MAGIC_SETS; i++)
499 mlist_free(ms->mlist[i]);
500 free(ms->o.pbuf);
501 free(ms->o.buf);
502 free(ms->c.li);
503 free(ms);
504 }
505
506 protected struct magic_set *
file_ms_alloc(int flags)507 file_ms_alloc(int flags)
508 {
509 struct magic_set *ms;
510 size_t i, len;
511
512 if ((ms = CAST(struct magic_set *, calloc((size_t)1,
513 sizeof(struct magic_set)))) == NULL)
514 return NULL;
515
516 if (magic_setflags(ms, flags) == -1) {
517 errno = EINVAL;
518 goto free;
519 }
520
521 ms->o.buf = ms->o.pbuf = NULL;
522 len = (ms->c.len = 10) * sizeof(*ms->c.li);
523
524 if ((ms->c.li = CAST(struct level_info *, malloc(len))) == NULL)
525 goto free;
526
527 ms->event_flags = 0;
528 ms->error = -1;
529 for (i = 0; i < MAGIC_SETS; i++)
530 ms->mlist[i] = NULL;
531 ms->file = "unknown";
532 ms->line = 0;
533 ms->indir_max = FILE_INDIR_MAX;
534 ms->name_max = FILE_NAME_MAX;
535 ms->elf_shnum_max = FILE_ELF_SHNUM_MAX;
536 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX;
537 ms->elf_notes_max = FILE_ELF_NOTES_MAX;
538 return ms;
539 free:
540 free(ms);
541 return NULL;
542 }
543
544 private void
apprentice_unmap(struct magic_map * map)545 apprentice_unmap(struct magic_map *map)
546 {
547 if (map == NULL)
548 return;
549
550 switch (map->type) {
551 #ifdef QUICK
552 case MAP_TYPE_MMAP:
553 if (map->p)
554 (void)munmap(map->p, map->len);
555 break;
556 #endif
557 case MAP_TYPE_MALLOC:
558 free(map->p);
559 break;
560 case MAP_TYPE_USER:
561 break;
562 default:
563 abort();
564 }
565 free(map);
566 }
567
568 private struct mlist *
mlist_alloc(void)569 mlist_alloc(void)
570 {
571 struct mlist *mlist;
572 if ((mlist = CAST(struct mlist *, calloc(1, sizeof(*mlist)))) == NULL) {
573 return NULL;
574 }
575 mlist->next = mlist->prev = mlist;
576 return mlist;
577 }
578
579 private void
mlist_free(struct mlist * mlist)580 mlist_free(struct mlist *mlist)
581 {
582 struct mlist *ml, *next;
583
584 if (mlist == NULL)
585 return;
586
587 ml = mlist->next;
588 for (ml = mlist->next; (next = ml->next) != NULL; ml = next) {
589 if (ml->map)
590 apprentice_unmap(ml->map);
591 free(ml);
592 if (ml == mlist)
593 break;
594 }
595 }
596
597 #ifndef COMPILE_ONLY
598 /* void **bufs: an array of compiled magic files */
599 protected int
buffer_apprentice(struct magic_set * ms,struct magic ** bufs,size_t * sizes,size_t nbufs)600 buffer_apprentice(struct magic_set *ms, struct magic **bufs,
601 size_t *sizes, size_t nbufs)
602 {
603 size_t i, j;
604 struct mlist *ml;
605 struct magic_map *map;
606
607 if (nbufs == 0)
608 return -1;
609
610 if (ms->mlist[0] != NULL)
611 file_reset(ms);
612
613 init_file_tables();
614
615 for (i = 0; i < MAGIC_SETS; i++) {
616 mlist_free(ms->mlist[i]);
617 if ((ms->mlist[i] = mlist_alloc()) == NULL) {
618 file_oomem(ms, sizeof(*ms->mlist[i]));
619 goto fail;
620 }
621 }
622
623 for (i = 0; i < nbufs; i++) {
624 map = apprentice_buf(ms, bufs[i], sizes[i]);
625 if (map == NULL)
626 goto fail;
627
628 for (j = 0; j < MAGIC_SETS; j++) {
629 if (add_mlist(ms->mlist[j], map, j) == -1) {
630 file_oomem(ms, sizeof(*ml));
631 goto fail;
632 }
633 }
634 }
635
636 return 0;
637 fail:
638 for (i = 0; i < MAGIC_SETS; i++) {
639 mlist_free(ms->mlist[i]);
640 ms->mlist[i] = NULL;
641 }
642 return -1;
643 }
644 #endif
645
646 /* const char *fn: list of magic files and directories */
647 protected int
file_apprentice(struct magic_set * ms,const char * fn,int action)648 file_apprentice(struct magic_set *ms, const char *fn, int action)
649 {
650 char *p, *mfn;
651 int file_err, errs = -1;
652 size_t i;
653
654 if (ms->mlist[0] != NULL)
655 file_reset(ms);
656
657 if ((fn = magic_getpath(fn, action)) == NULL)
658 return -1;
659
660 init_file_tables();
661
662 if ((mfn = strdup(fn)) == NULL) {
663 file_oomem(ms, strlen(fn));
664 return -1;
665 }
666
667 for (i = 0; i < MAGIC_SETS; i++) {
668 mlist_free(ms->mlist[i]);
669 if ((ms->mlist[i] = mlist_alloc()) == NULL) {
670 file_oomem(ms, sizeof(*ms->mlist[i]));
671 while (i-- > 0) {
672 mlist_free(ms->mlist[i]);
673 ms->mlist[i] = NULL;
674 }
675 free(mfn);
676 return -1;
677 }
678 }
679 fn = mfn;
680
681 while (fn) {
682 p = strchr(fn, PATHSEP);
683 if (p)
684 *p++ = '\0';
685 if (*fn == '\0')
686 break;
687 file_err = apprentice_1(ms, fn, action);
688 errs = MAX(errs, file_err);
689 fn = p;
690 }
691
692 free(mfn);
693
694 if (errs == -1) {
695 for (i = 0; i < MAGIC_SETS; i++) {
696 mlist_free(ms->mlist[i]);
697 ms->mlist[i] = NULL;
698 }
699 file_error(ms, 0, "could not find any valid magic files!");
700 return -1;
701 }
702
703 #if 0
704 /*
705 * Always leave the database loaded
706 */
707 if (action == FILE_LOAD)
708 return 0;
709
710 for (i = 0; i < MAGIC_SETS; i++) {
711 mlist_free(ms->mlist[i]);
712 ms->mlist[i] = NULL;
713 }
714 #endif
715
716 switch (action) {
717 case FILE_LOAD:
718 case FILE_COMPILE:
719 case FILE_CHECK:
720 case FILE_LIST:
721 return 0;
722 default:
723 file_error(ms, 0, "Invalid action %d", action);
724 return -1;
725 }
726 }
727
728 /*
729 * Compute the real length of a magic expression, for the purposes
730 * of determining how "strong" a magic expression is (approximating
731 * how specific its matches are):
732 * - magic characters count 0 unless escaped.
733 * - [] expressions count 1
734 * - {} expressions count 0
735 * - regular characters or escaped magic characters count 1
736 * - 0 length expressions count as one
737 */
738 private size_t
nonmagic(const char * str)739 nonmagic(const char *str)
740 {
741 const char *p;
742 size_t rv = 0;
743
744 for (p = str; *p; p++)
745 switch (*p) {
746 case '\\': /* Escaped anything counts 1 */
747 if (!*++p)
748 p--;
749 rv++;
750 continue;
751 case '?': /* Magic characters count 0 */
752 case '*':
753 case '.':
754 case '+':
755 case '^':
756 case '$':
757 continue;
758 case '[': /* Bracketed expressions count 1 the ']' */
759 while (*p && *p != ']')
760 p++;
761 p--;
762 continue;
763 case '{': /* Braced expressions count 0 */
764 while (*p && *p != '}')
765 p++;
766 if (!*p)
767 p--;
768 continue;
769 default: /* Anything else counts 1 */
770 rv++;
771 continue;
772 }
773
774 return rv == 0 ? 1 : rv; /* Return at least 1 */
775 }
776
777 /*
778 * Get weight of this magic entry, for sorting purposes.
779 */
780 private size_t
apprentice_magic_strength(const struct magic * m)781 apprentice_magic_strength(const struct magic *m)
782 {
783 #define MULT 10
784 size_t v, val = 2 * MULT; /* baseline strength */
785
786 switch (m->type) {
787 case FILE_DEFAULT: /* make sure this sorts last */
788 if (m->factor_op != FILE_FACTOR_OP_NONE)
789 abort();
790 return 0;
791
792 case FILE_BYTE:
793 val += 1 * MULT;
794 break;
795
796 case FILE_SHORT:
797 case FILE_LESHORT:
798 case FILE_BESHORT:
799 val += 2 * MULT;
800 break;
801
802 case FILE_LONG:
803 case FILE_LELONG:
804 case FILE_BELONG:
805 case FILE_MELONG:
806 val += 4 * MULT;
807 break;
808
809 case FILE_PSTRING:
810 case FILE_STRING:
811 val += m->vallen * MULT;
812 break;
813
814 case FILE_BESTRING16:
815 case FILE_LESTRING16:
816 val += m->vallen * MULT / 2;
817 break;
818
819 case FILE_SEARCH:
820 val += m->vallen * MAX(MULT / m->vallen, 1);
821 break;
822
823 case FILE_REGEX:
824 v = nonmagic(m->value.s);
825 val += v * MAX(MULT / v, 1);
826 break;
827
828 case FILE_DATE:
829 case FILE_LEDATE:
830 case FILE_BEDATE:
831 case FILE_MEDATE:
832 case FILE_LDATE:
833 case FILE_LELDATE:
834 case FILE_BELDATE:
835 case FILE_MELDATE:
836 case FILE_FLOAT:
837 case FILE_BEFLOAT:
838 case FILE_LEFLOAT:
839 val += 4 * MULT;
840 break;
841
842 case FILE_QUAD:
843 case FILE_BEQUAD:
844 case FILE_LEQUAD:
845 case FILE_QDATE:
846 case FILE_LEQDATE:
847 case FILE_BEQDATE:
848 case FILE_QLDATE:
849 case FILE_LEQLDATE:
850 case FILE_BEQLDATE:
851 case FILE_QWDATE:
852 case FILE_LEQWDATE:
853 case FILE_BEQWDATE:
854 case FILE_DOUBLE:
855 case FILE_BEDOUBLE:
856 case FILE_LEDOUBLE:
857 val += 8 * MULT;
858 break;
859
860 case FILE_INDIRECT:
861 case FILE_NAME:
862 case FILE_USE:
863 break;
864
865 default:
866 (void)fprintf(stderr, "Bad type %d\n", m->type);
867 abort();
868 }
869
870 switch (m->reln) {
871 case 'x': /* matches anything penalize */
872 case '!': /* matches almost anything penalize */
873 val = 0;
874 break;
875
876 case '=': /* Exact match, prefer */
877 val += MULT;
878 break;
879
880 case '>':
881 case '<': /* comparison match reduce strength */
882 val -= 2 * MULT;
883 break;
884
885 case '^':
886 case '&': /* masking bits, we could count them too */
887 val -= MULT;
888 break;
889
890 default:
891 (void)fprintf(stderr, "Bad relation %c\n", m->reln);
892 abort();
893 }
894
895 if (val == 0) /* ensure we only return 0 for FILE_DEFAULT */
896 val = 1;
897
898 switch (m->factor_op) {
899 case FILE_FACTOR_OP_NONE:
900 break;
901 case FILE_FACTOR_OP_PLUS:
902 val += m->factor;
903 break;
904 case FILE_FACTOR_OP_MINUS:
905 val -= m->factor;
906 break;
907 case FILE_FACTOR_OP_TIMES:
908 val *= m->factor;
909 break;
910 case FILE_FACTOR_OP_DIV:
911 val /= m->factor;
912 break;
913 default:
914 abort();
915 }
916
917 /*
918 * Magic entries with no description get a bonus because they depend
919 * on subsequent magic entries to print something.
920 */
921 if (m->desc[0] == '\0')
922 val++;
923 return val;
924 }
925
926 /*
927 * Sort callback for sorting entries by "strength" (basically length)
928 */
929 private int
apprentice_sort(const void * a,const void * b)930 apprentice_sort(const void *a, const void *b)
931 {
932 const struct magic_entry *ma = CAST(const struct magic_entry *, a);
933 const struct magic_entry *mb = CAST(const struct magic_entry *, b);
934 size_t sa = apprentice_magic_strength(ma->mp);
935 size_t sb = apprentice_magic_strength(mb->mp);
936 if (sa == sb)
937 return 0;
938 else if (sa > sb)
939 return -1;
940 else
941 return 1;
942 }
943
944 /*
945 * Shows sorted patterns list in the order which is used for the matching
946 */
947 private void
apprentice_list(struct mlist * mlist,int mode)948 apprentice_list(struct mlist *mlist, int mode)
949 {
950 uint32_t magindex = 0;
951 struct mlist *ml;
952 for (ml = mlist->next; ml != mlist; ml = ml->next) {
953 for (magindex = 0; magindex < ml->nmagic; magindex++) {
954 struct magic *m = &ml->magic[magindex];
955 if ((m->flag & mode) != mode) {
956 /* Skip sub-tests */
957 while (magindex + 1 < ml->nmagic &&
958 ml->magic[magindex + 1].cont_level != 0)
959 ++magindex;
960 continue; /* Skip to next top-level test*/
961 }
962
963 /*
964 * Try to iterate over the tree until we find item with
965 * description/mimetype.
966 */
967 while (magindex + 1 < ml->nmagic &&
968 ml->magic[magindex + 1].cont_level != 0 &&
969 *ml->magic[magindex].desc == '\0' &&
970 *ml->magic[magindex].mimetype == '\0')
971 magindex++;
972
973 printf("Strength = %3" SIZE_T_FORMAT "u : %s [%s]\n",
974 apprentice_magic_strength(m),
975 ml->magic[magindex].desc,
976 ml->magic[magindex].mimetype);
977 }
978 }
979 }
980
981 private void
set_test_type(struct magic * mstart,struct magic * m)982 set_test_type(struct magic *mstart, struct magic *m)
983 {
984 switch (m->type) {
985 case FILE_BYTE:
986 case FILE_SHORT:
987 case FILE_LONG:
988 case FILE_DATE:
989 case FILE_BESHORT:
990 case FILE_BELONG:
991 case FILE_BEDATE:
992 case FILE_LESHORT:
993 case FILE_LELONG:
994 case FILE_LEDATE:
995 case FILE_LDATE:
996 case FILE_BELDATE:
997 case FILE_LELDATE:
998 case FILE_MEDATE:
999 case FILE_MELDATE:
1000 case FILE_MELONG:
1001 case FILE_QUAD:
1002 case FILE_LEQUAD:
1003 case FILE_BEQUAD:
1004 case FILE_QDATE:
1005 case FILE_LEQDATE:
1006 case FILE_BEQDATE:
1007 case FILE_QLDATE:
1008 case FILE_LEQLDATE:
1009 case FILE_BEQLDATE:
1010 case FILE_QWDATE:
1011 case FILE_LEQWDATE:
1012 case FILE_BEQWDATE:
1013 case FILE_FLOAT:
1014 case FILE_BEFLOAT:
1015 case FILE_LEFLOAT:
1016 case FILE_DOUBLE:
1017 case FILE_BEDOUBLE:
1018 case FILE_LEDOUBLE:
1019 mstart->flag |= BINTEST;
1020 break;
1021 case FILE_STRING:
1022 case FILE_PSTRING:
1023 case FILE_BESTRING16:
1024 case FILE_LESTRING16:
1025 /* Allow text overrides */
1026 if (mstart->str_flags & STRING_TEXTTEST)
1027 mstart->flag |= TEXTTEST;
1028 else
1029 mstart->flag |= BINTEST;
1030 break;
1031 case FILE_REGEX:
1032 case FILE_SEARCH:
1033 /* Check for override */
1034 if (mstart->str_flags & STRING_BINTEST)
1035 mstart->flag |= BINTEST;
1036 if (mstart->str_flags & STRING_TEXTTEST)
1037 mstart->flag |= TEXTTEST;
1038
1039 if (mstart->flag & (TEXTTEST|BINTEST))
1040 break;
1041
1042 /* binary test if pattern is not text */
1043 if (file_looks_utf8(m->value.us, (size_t)m->vallen, NULL,
1044 NULL) <= 0)
1045 mstart->flag |= BINTEST;
1046 else
1047 mstart->flag |= TEXTTEST;
1048 break;
1049 case FILE_DEFAULT:
1050 /* can't deduce anything; we shouldn't see this at the
1051 top level anyway */
1052 break;
1053 case FILE_INVALID:
1054 default:
1055 /* invalid search type, but no need to complain here */
1056 break;
1057 }
1058 }
1059
1060 private int
addentry(struct magic_set * ms,struct magic_entry * me,struct magic_entry_set * mset)1061 addentry(struct magic_set *ms, struct magic_entry *me,
1062 struct magic_entry_set *mset)
1063 {
1064 size_t i = me->mp->type == FILE_NAME ? 1 : 0;
1065 if (mset[i].count == mset[i].max) {
1066 struct magic_entry *mp;
1067
1068 mset[i].max += ALLOC_INCR;
1069 if ((mp = CAST(struct magic_entry *,
1070 realloc(mset[i].me, sizeof(*mp) * mset[i].max))) ==
1071 NULL) {
1072 file_oomem(ms, sizeof(*mp) * mset[i].max);
1073 return -1;
1074 }
1075 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) *
1076 ALLOC_INCR);
1077 mset[i].me = mp;
1078 }
1079 mset[i].me[mset[i].count++] = *me;
1080 memset(me, 0, sizeof(*me));
1081 return 0;
1082 }
1083
1084 /*
1085 * Load and parse one file.
1086 */
1087 private void
load_1(struct magic_set * ms,int action,const char * fn,int * errs,struct magic_entry_set * mset)1088 load_1(struct magic_set *ms, int action, const char *fn, int *errs,
1089 struct magic_entry_set *mset)
1090 {
1091 size_t lineno = 0, llen = 0;
1092 char *line = NULL;
1093 ssize_t len;
1094 struct magic_entry me;
1095
1096 FILE *f = fopen(ms->file = fn, "r");
1097 if (f == NULL) {
1098 if (errno != ENOENT)
1099 file_error(ms, errno, "cannot read magic file `%s'",
1100 fn);
1101 (*errs)++;
1102 return;
1103 }
1104
1105 memset(&me, 0, sizeof(me));
1106 /* read and parse this file */
1107 for (ms->line = 1; (len = getline(&line, &llen, f)) != -1;
1108 ms->line++) {
1109 if (len == 0) /* null line, garbage, etc */
1110 continue;
1111 if (line[len - 1] == '\n') {
1112 lineno++;
1113 line[len - 1] = '\0'; /* delete newline */
1114 }
1115 switch (line[0]) {
1116 case '\0': /* empty, do not parse */
1117 case '#': /* comment, do not parse */
1118 continue;
1119 case '!':
1120 if (line[1] == ':') {
1121 size_t i;
1122
1123 for (i = 0; bang[i].name != NULL; i++) {
1124 if ((size_t)(len - 2) > bang[i].len &&
1125 memcmp(bang[i].name, line + 2,
1126 bang[i].len) == 0)
1127 break;
1128 }
1129 if (bang[i].name == NULL) {
1130 file_error(ms, 0,
1131 "Unknown !: entry `%s'", line);
1132 (*errs)++;
1133 continue;
1134 }
1135 if (me.mp == NULL) {
1136 file_error(ms, 0,
1137 "No current entry for :!%s type",
1138 bang[i].name);
1139 (*errs)++;
1140 continue;
1141 }
1142 if ((*bang[i].fun)(ms, &me,
1143 line + bang[i].len + 2) != 0) {
1144 (*errs)++;
1145 continue;
1146 }
1147 continue;
1148 }
1149 /*FALLTHROUGH*/
1150 default:
1151 again:
1152 switch (parse(ms, &me, line, lineno, action)) {
1153 case 0:
1154 continue;
1155 case 1:
1156 (void)addentry(ms, &me, mset);
1157 goto again;
1158 default:
1159 (*errs)++;
1160 break;
1161 }
1162 }
1163 }
1164 if (me.mp)
1165 (void)addentry(ms, &me, mset);
1166 free(line);
1167 (void)fclose(f);
1168 }
1169
1170 /*
1171 * parse a file or directory of files
1172 * const char *fn: name of magic file or directory
1173 */
1174 private int
cmpstrp(const void * p1,const void * p2)1175 cmpstrp(const void *p1, const void *p2)
1176 {
1177 return strcmp(*(char *const *)p1, *(char *const *)p2);
1178 }
1179
1180
1181 private uint32_t
set_text_binary(struct magic_set * ms,struct magic_entry * me,uint32_t nme,uint32_t starttest)1182 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme,
1183 uint32_t starttest)
1184 {
1185 static const char text[] = "text";
1186 static const char binary[] = "binary";
1187 static const size_t len = sizeof(text);
1188
1189 uint32_t i = starttest;
1190
1191 do {
1192 set_test_type(me[starttest].mp, me[i].mp);
1193 if ((ms->flags & MAGIC_DEBUG) == 0)
1194 continue;
1195 (void)fprintf(stderr, "%s%s%s: %s\n",
1196 me[i].mp->mimetype,
1197 me[i].mp->mimetype[0] == '\0' ? "" : "; ",
1198 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)",
1199 me[i].mp->flag & BINTEST ? binary : text);
1200 if (me[i].mp->flag & BINTEST) {
1201 char *p = strstr(me[i].mp->desc, text);
1202 if (p && (p == me[i].mp->desc ||
1203 isspace((unsigned char)p[-1])) &&
1204 (p + len - me[i].mp->desc == MAXstring
1205 || (p[len] == '\0' ||
1206 isspace((unsigned char)p[len]))))
1207 (void)fprintf(stderr, "*** Possible "
1208 "binary test for text type\n");
1209 }
1210 } while (++i < nme && me[i].mp->cont_level != 0);
1211 return i;
1212 }
1213
1214 private void
set_last_default(struct magic_set * ms,struct magic_entry * me,uint32_t nme)1215 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme)
1216 {
1217 uint32_t i;
1218 for (i = 0; i < nme; i++) {
1219 if (me[i].mp->cont_level == 0 &&
1220 me[i].mp->type == FILE_DEFAULT) {
1221 while (++i < nme)
1222 if (me[i].mp->cont_level == 0)
1223 break;
1224 if (i != nme) {
1225 /* XXX - Ugh! */
1226 ms->line = me[i].mp->lineno;
1227 file_magwarn(ms,
1228 "level 0 \"default\" did not sort last");
1229 }
1230 return;
1231 }
1232 }
1233 }
1234
1235 private int
coalesce_entries(struct magic_set * ms,struct magic_entry * me,uint32_t nme,struct magic ** ma,uint32_t * nma)1236 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme,
1237 struct magic **ma, uint32_t *nma)
1238 {
1239 uint32_t i, mentrycount = 0;
1240 size_t slen;
1241
1242 for (i = 0; i < nme; i++)
1243 mentrycount += me[i].cont_count;
1244
1245 slen = sizeof(**ma) * mentrycount;
1246 if ((*ma = CAST(struct magic *, malloc(slen))) == NULL) {
1247 file_oomem(ms, slen);
1248 return -1;
1249 }
1250
1251 mentrycount = 0;
1252 for (i = 0; i < nme; i++) {
1253 (void)memcpy(*ma + mentrycount, me[i].mp,
1254 me[i].cont_count * sizeof(**ma));
1255 mentrycount += me[i].cont_count;
1256 }
1257 *nma = mentrycount;
1258 return 0;
1259 }
1260
1261 private void
magic_entry_free(struct magic_entry * me,uint32_t nme)1262 magic_entry_free(struct magic_entry *me, uint32_t nme)
1263 {
1264 uint32_t i;
1265 if (me == NULL)
1266 return;
1267 for (i = 0; i < nme; i++)
1268 free(me[i].mp);
1269 free(me);
1270 }
1271
1272 private struct magic_map *
apprentice_load(struct magic_set * ms,const char * fn,int action)1273 apprentice_load(struct magic_set *ms, const char *fn, int action)
1274 {
1275 int errs = 0;
1276 uint32_t i, j;
1277 size_t files = 0, maxfiles = 0;
1278 char **filearr = NULL, *mfn;
1279 struct stat st;
1280 struct magic_map *map;
1281 struct magic_entry_set mset[MAGIC_SETS];
1282 DIR *dir;
1283 struct dirent *d;
1284
1285 memset(mset, 0, sizeof(mset));
1286 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */
1287
1288
1289 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL)
1290 {
1291 file_oomem(ms, sizeof(*map));
1292 return NULL;
1293 }
1294
1295 /* print silly verbose header for USG compat. */
1296 if (action == FILE_CHECK)
1297 (void)fprintf(stderr, "%s\n", usg_hdr);
1298
1299 /* load directory or file */
1300 if (stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) {
1301 dir = opendir(fn);
1302 if (!dir) {
1303 errs++;
1304 goto out;
1305 }
1306 while ((d = readdir(dir)) != NULL) {
1307 if (asprintf(&mfn, "%s/%s", fn, d->d_name) < 0) {
1308 file_oomem(ms,
1309 strlen(fn) + strlen(d->d_name) + 2);
1310 errs++;
1311 closedir(dir);
1312 goto out;
1313 }
1314 if (stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) {
1315 free(mfn);
1316 continue;
1317 }
1318 if (files >= maxfiles) {
1319 size_t mlen;
1320 maxfiles = (maxfiles + 1) * 2;
1321 mlen = maxfiles * sizeof(*filearr);
1322 if ((filearr = CAST(char **,
1323 realloc(filearr, mlen))) == NULL) {
1324 file_oomem(ms, mlen);
1325 free(mfn);
1326 closedir(dir);
1327 errs++;
1328 goto out;
1329 }
1330 }
1331 filearr[files++] = mfn;
1332 }
1333 closedir(dir);
1334 qsort(filearr, files, sizeof(*filearr), cmpstrp);
1335 for (i = 0; i < files; i++) {
1336 load_1(ms, action, filearr[i], &errs, mset);
1337 free(filearr[i]);
1338 }
1339 free(filearr);
1340 } else
1341 load_1(ms, action, fn, &errs, mset);
1342 if (errs)
1343 goto out;
1344
1345 for (j = 0; j < MAGIC_SETS; j++) {
1346 /* Set types of tests */
1347 for (i = 0; i < mset[j].count; ) {
1348 if (mset[j].me[i].mp->cont_level != 0) {
1349 i++;
1350 continue;
1351 }
1352 i = set_text_binary(ms, mset[j].me, mset[j].count, i);
1353 }
1354 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me),
1355 apprentice_sort);
1356
1357 /*
1358 * Make sure that any level 0 "default" line is last
1359 * (if one exists).
1360 */
1361 set_last_default(ms, mset[j].me, mset[j].count);
1362
1363 /* coalesce per file arrays into a single one */
1364 if (coalesce_entries(ms, mset[j].me, mset[j].count,
1365 &map->magic[j], &map->nmagic[j]) == -1) {
1366 errs++;
1367 goto out;
1368 }
1369 }
1370
1371 out:
1372 for (j = 0; j < MAGIC_SETS; j++)
1373 magic_entry_free(mset[j].me, mset[j].count);
1374
1375 if (errs) {
1376 apprentice_unmap(map);
1377 return NULL;
1378 }
1379 return map;
1380 }
1381
1382 /*
1383 * extend the sign bit if the comparison is to be signed
1384 */
1385 protected uint64_t
file_signextend(struct magic_set * ms,struct magic * m,uint64_t v)1386 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v)
1387 {
1388 if (!(m->flag & UNSIGNED)) {
1389 switch(m->type) {
1390 /*
1391 * Do not remove the casts below. They are
1392 * vital. When later compared with the data,
1393 * the sign extension must have happened.
1394 */
1395 case FILE_BYTE:
1396 v = (signed char) v;
1397 break;
1398 case FILE_SHORT:
1399 case FILE_BESHORT:
1400 case FILE_LESHORT:
1401 v = (short) v;
1402 break;
1403 case FILE_DATE:
1404 case FILE_BEDATE:
1405 case FILE_LEDATE:
1406 case FILE_MEDATE:
1407 case FILE_LDATE:
1408 case FILE_BELDATE:
1409 case FILE_LELDATE:
1410 case FILE_MELDATE:
1411 case FILE_LONG:
1412 case FILE_BELONG:
1413 case FILE_LELONG:
1414 case FILE_MELONG:
1415 case FILE_FLOAT:
1416 case FILE_BEFLOAT:
1417 case FILE_LEFLOAT:
1418 v = (int32_t) v;
1419 break;
1420 case FILE_QUAD:
1421 case FILE_BEQUAD:
1422 case FILE_LEQUAD:
1423 case FILE_QDATE:
1424 case FILE_QLDATE:
1425 case FILE_QWDATE:
1426 case FILE_BEQDATE:
1427 case FILE_BEQLDATE:
1428 case FILE_BEQWDATE:
1429 case FILE_LEQDATE:
1430 case FILE_LEQLDATE:
1431 case FILE_LEQWDATE:
1432 case FILE_DOUBLE:
1433 case FILE_BEDOUBLE:
1434 case FILE_LEDOUBLE:
1435 v = (int64_t) v;
1436 break;
1437 case FILE_STRING:
1438 case FILE_PSTRING:
1439 case FILE_BESTRING16:
1440 case FILE_LESTRING16:
1441 case FILE_REGEX:
1442 case FILE_SEARCH:
1443 case FILE_DEFAULT:
1444 case FILE_INDIRECT:
1445 case FILE_NAME:
1446 case FILE_USE:
1447 case FILE_CLEAR:
1448 break;
1449 default:
1450 if (ms->flags & MAGIC_CHECK)
1451 file_magwarn(ms, "cannot happen: m->type=%d\n",
1452 m->type);
1453 return ~0U;
1454 }
1455 }
1456 return v;
1457 }
1458
1459 private int
string_modifier_check(struct magic_set * ms,struct magic * m)1460 string_modifier_check(struct magic_set *ms, struct magic *m)
1461 {
1462 if ((ms->flags & MAGIC_CHECK) == 0)
1463 return 0;
1464
1465 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) &&
1466 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) {
1467 file_magwarn(ms,
1468 "'/BHhLl' modifiers are only allowed for pascal strings\n");
1469 return -1;
1470 }
1471 switch (m->type) {
1472 case FILE_BESTRING16:
1473 case FILE_LESTRING16:
1474 if (m->str_flags != 0) {
1475 file_magwarn(ms,
1476 "no modifiers allowed for 16-bit strings\n");
1477 return -1;
1478 }
1479 break;
1480 case FILE_STRING:
1481 case FILE_PSTRING:
1482 if ((m->str_flags & REGEX_OFFSET_START) != 0) {
1483 file_magwarn(ms,
1484 "'/%c' only allowed on regex and search\n",
1485 CHAR_REGEX_OFFSET_START);
1486 return -1;
1487 }
1488 break;
1489 case FILE_SEARCH:
1490 if (m->str_range == 0) {
1491 file_magwarn(ms,
1492 "missing range; defaulting to %d\n",
1493 STRING_DEFAULT_RANGE);
1494 m->str_range = STRING_DEFAULT_RANGE;
1495 return -1;
1496 }
1497 break;
1498 case FILE_REGEX:
1499 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) {
1500 file_magwarn(ms, "'/%c' not allowed on regex\n",
1501 CHAR_COMPACT_WHITESPACE);
1502 return -1;
1503 }
1504 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) {
1505 file_magwarn(ms, "'/%c' not allowed on regex\n",
1506 CHAR_COMPACT_OPTIONAL_WHITESPACE);
1507 return -1;
1508 }
1509 break;
1510 default:
1511 file_magwarn(ms, "coding error: m->type=%d\n",
1512 m->type);
1513 return -1;
1514 }
1515 return 0;
1516 }
1517
1518 private int
get_op(char c)1519 get_op(char c)
1520 {
1521 switch (c) {
1522 case '&':
1523 return FILE_OPAND;
1524 case '|':
1525 return FILE_OPOR;
1526 case '^':
1527 return FILE_OPXOR;
1528 case '+':
1529 return FILE_OPADD;
1530 case '-':
1531 return FILE_OPMINUS;
1532 case '*':
1533 return FILE_OPMULTIPLY;
1534 case '/':
1535 return FILE_OPDIVIDE;
1536 case '%':
1537 return FILE_OPMODULO;
1538 default:
1539 return -1;
1540 }
1541 }
1542
1543 #ifdef ENABLE_CONDITIONALS
1544 private int
get_cond(const char * l,const char ** t)1545 get_cond(const char *l, const char **t)
1546 {
1547 static const struct cond_tbl_s {
1548 char name[8];
1549 size_t len;
1550 int cond;
1551 } cond_tbl[] = {
1552 { "if", 2, COND_IF },
1553 { "elif", 4, COND_ELIF },
1554 { "else", 4, COND_ELSE },
1555 { "", 0, COND_NONE },
1556 };
1557 const struct cond_tbl_s *p;
1558
1559 for (p = cond_tbl; p->len; p++) {
1560 if (strncmp(l, p->name, p->len) == 0 &&
1561 isspace((unsigned char)l[p->len])) {
1562 if (t)
1563 *t = l + p->len;
1564 break;
1565 }
1566 }
1567 return p->cond;
1568 }
1569
1570 private int
check_cond(struct magic_set * ms,int cond,uint32_t cont_level)1571 check_cond(struct magic_set *ms, int cond, uint32_t cont_level)
1572 {
1573 int last_cond;
1574 last_cond = ms->c.li[cont_level].last_cond;
1575
1576 switch (cond) {
1577 case COND_IF:
1578 if (last_cond != COND_NONE && last_cond != COND_ELIF) {
1579 if (ms->flags & MAGIC_CHECK)
1580 file_magwarn(ms, "syntax error: `if'");
1581 return -1;
1582 }
1583 last_cond = COND_IF;
1584 break;
1585
1586 case COND_ELIF:
1587 if (last_cond != COND_IF && last_cond != COND_ELIF) {
1588 if (ms->flags & MAGIC_CHECK)
1589 file_magwarn(ms, "syntax error: `elif'");
1590 return -1;
1591 }
1592 last_cond = COND_ELIF;
1593 break;
1594
1595 case COND_ELSE:
1596 if (last_cond != COND_IF && last_cond != COND_ELIF) {
1597 if (ms->flags & MAGIC_CHECK)
1598 file_magwarn(ms, "syntax error: `else'");
1599 return -1;
1600 }
1601 last_cond = COND_NONE;
1602 break;
1603
1604 case COND_NONE:
1605 last_cond = COND_NONE;
1606 break;
1607 }
1608
1609 ms->c.li[cont_level].last_cond = last_cond;
1610 return 0;
1611 }
1612 #endif /* ENABLE_CONDITIONALS */
1613
1614 private int
parse_indirect_modifier(struct magic_set * ms,struct magic * m,const char ** lp)1615 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp)
1616 {
1617 const char *l = *lp;
1618
1619 while (!isspace((unsigned char)*++l))
1620 switch (*l) {
1621 case CHAR_INDIRECT_RELATIVE:
1622 m->str_flags |= INDIRECT_RELATIVE;
1623 break;
1624 default:
1625 if (ms->flags & MAGIC_CHECK)
1626 file_magwarn(ms, "indirect modifier `%c' "
1627 "invalid", *l);
1628 *lp = l;
1629 return -1;
1630 }
1631 *lp = l;
1632 return 0;
1633 }
1634
1635 private void
parse_op_modifier(struct magic_set * ms,struct magic * m,const char ** lp,int op)1636 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp,
1637 int op)
1638 {
1639 const char *l = *lp;
1640 char *t;
1641 uint64_t val;
1642
1643 ++l;
1644 m->mask_op |= op;
1645 val = (uint64_t)strtoull(l, &t, 0);
1646 l = t;
1647 m->num_mask = file_signextend(ms, m, val);
1648 eatsize(&l);
1649 *lp = l;
1650 }
1651
1652 private int
parse_string_modifier(struct magic_set * ms,struct magic * m,const char ** lp)1653 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp)
1654 {
1655 const char *l = *lp;
1656 char *t;
1657 int have_range = 0;
1658
1659 while (!isspace((unsigned char)*++l)) {
1660 switch (*l) {
1661 case '0': case '1': case '2':
1662 case '3': case '4': case '5':
1663 case '6': case '7': case '8':
1664 case '9':
1665 if (have_range && (ms->flags & MAGIC_CHECK))
1666 file_magwarn(ms, "multiple ranges");
1667 have_range = 1;
1668 m->str_range = CAST(uint32_t, strtoul(l, &t, 0));
1669 if (m->str_range == 0)
1670 file_magwarn(ms, "zero range");
1671 l = t - 1;
1672 break;
1673 case CHAR_COMPACT_WHITESPACE:
1674 m->str_flags |= STRING_COMPACT_WHITESPACE;
1675 break;
1676 case CHAR_COMPACT_OPTIONAL_WHITESPACE:
1677 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE;
1678 break;
1679 case CHAR_IGNORE_LOWERCASE:
1680 m->str_flags |= STRING_IGNORE_LOWERCASE;
1681 break;
1682 case CHAR_IGNORE_UPPERCASE:
1683 m->str_flags |= STRING_IGNORE_UPPERCASE;
1684 break;
1685 case CHAR_REGEX_OFFSET_START:
1686 m->str_flags |= REGEX_OFFSET_START;
1687 break;
1688 case CHAR_BINTEST:
1689 m->str_flags |= STRING_BINTEST;
1690 break;
1691 case CHAR_TEXTTEST:
1692 m->str_flags |= STRING_TEXTTEST;
1693 break;
1694 case CHAR_TRIM:
1695 m->str_flags |= STRING_TRIM;
1696 break;
1697 case CHAR_PSTRING_1_LE:
1698 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a)
1699 if (m->type != FILE_PSTRING)
1700 goto bad;
1701 SET_LENGTH(PSTRING_1_LE);
1702 break;
1703 case CHAR_PSTRING_2_BE:
1704 if (m->type != FILE_PSTRING)
1705 goto bad;
1706 SET_LENGTH(PSTRING_2_BE);
1707 break;
1708 case CHAR_PSTRING_2_LE:
1709 if (m->type != FILE_PSTRING)
1710 goto bad;
1711 SET_LENGTH(PSTRING_2_LE);
1712 break;
1713 case CHAR_PSTRING_4_BE:
1714 if (m->type != FILE_PSTRING)
1715 goto bad;
1716 SET_LENGTH(PSTRING_4_BE);
1717 break;
1718 case CHAR_PSTRING_4_LE:
1719 switch (m->type) {
1720 case FILE_PSTRING:
1721 case FILE_REGEX:
1722 break;
1723 default:
1724 goto bad;
1725 }
1726 SET_LENGTH(PSTRING_4_LE);
1727 break;
1728 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF:
1729 if (m->type != FILE_PSTRING)
1730 goto bad;
1731 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF;
1732 break;
1733 default:
1734 bad:
1735 if (ms->flags & MAGIC_CHECK)
1736 file_magwarn(ms, "string modifier `%c' "
1737 "invalid", *l);
1738 goto out;
1739 }
1740 /* allow multiple '/' for readability */
1741 if (l[1] == '/' && !isspace((unsigned char)l[2]))
1742 l++;
1743 }
1744 if (string_modifier_check(ms, m) == -1)
1745 goto out;
1746 *lp = l;
1747 return 0;
1748 out:
1749 *lp = l;
1750 return -1;
1751 }
1752
1753 /*
1754 * parse one line from magic file, put into magic[index++] if valid
1755 */
1756 private int
parse(struct magic_set * ms,struct magic_entry * me,const char * line,size_t lineno,int action)1757 parse(struct magic_set *ms, struct magic_entry *me, const char *line,
1758 size_t lineno, int action)
1759 {
1760 #ifdef ENABLE_CONDITIONALS
1761 static uint32_t last_cont_level = 0;
1762 #endif
1763 size_t i;
1764 struct magic *m;
1765 const char *l = line;
1766 char *t;
1767 int op;
1768 uint32_t cont_level;
1769 int32_t diff;
1770
1771 cont_level = 0;
1772
1773 /*
1774 * Parse the offset.
1775 */
1776 while (*l == '>') {
1777 ++l; /* step over */
1778 cont_level++;
1779 }
1780 #ifdef ENABLE_CONDITIONALS
1781 if (cont_level == 0 || cont_level > last_cont_level)
1782 if (file_check_mem(ms, cont_level) == -1)
1783 return -1;
1784 last_cont_level = cont_level;
1785 #endif
1786 if (cont_level != 0) {
1787 if (me->mp == NULL) {
1788 file_magerror(ms, "No current entry for continuation");
1789 return -1;
1790 }
1791 if (me->cont_count == 0) {
1792 file_magerror(ms, "Continuations present with 0 count");
1793 return -1;
1794 }
1795 m = &me->mp[me->cont_count - 1];
1796 diff = (int32_t)cont_level - (int32_t)m->cont_level;
1797 if (diff > 1)
1798 file_magwarn(ms, "New continuation level %u is more "
1799 "than one larger than current level %u", cont_level,
1800 m->cont_level);
1801 if (me->cont_count == me->max_count) {
1802 struct magic *nm;
1803 size_t cnt = me->max_count + ALLOC_CHUNK;
1804 if ((nm = CAST(struct magic *, realloc(me->mp,
1805 sizeof(*nm) * cnt))) == NULL) {
1806 file_oomem(ms, sizeof(*nm) * cnt);
1807 return -1;
1808 }
1809 me->mp = m = nm;
1810 me->max_count = CAST(uint32_t, cnt);
1811 }
1812 m = &me->mp[me->cont_count++];
1813 (void)memset(m, 0, sizeof(*m));
1814 m->cont_level = cont_level;
1815 } else {
1816 static const size_t len = sizeof(*m) * ALLOC_CHUNK;
1817 if (me->mp != NULL)
1818 return 1;
1819 if ((m = CAST(struct magic *, malloc(len))) == NULL) {
1820 file_oomem(ms, len);
1821 return -1;
1822 }
1823 me->mp = m;
1824 me->max_count = ALLOC_CHUNK;
1825 (void)memset(m, 0, sizeof(*m));
1826 m->factor_op = FILE_FACTOR_OP_NONE;
1827 m->cont_level = 0;
1828 me->cont_count = 1;
1829 }
1830 m->lineno = CAST(uint32_t, lineno);
1831
1832 if (*l == '&') { /* m->cont_level == 0 checked below. */
1833 ++l; /* step over */
1834 m->flag |= OFFADD;
1835 }
1836 if (*l == '(') {
1837 ++l; /* step over */
1838 m->flag |= INDIR;
1839 if (m->flag & OFFADD)
1840 m->flag = (m->flag & ~OFFADD) | INDIROFFADD;
1841
1842 if (*l == '&') { /* m->cont_level == 0 checked below */
1843 ++l; /* step over */
1844 m->flag |= OFFADD;
1845 }
1846 }
1847 /* Indirect offsets are not valid at level 0. */
1848 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD)))
1849 if (ms->flags & MAGIC_CHECK)
1850 file_magwarn(ms, "relative offset at level 0");
1851
1852 /* get offset, then skip over it */
1853 m->offset = (uint32_t)strtoul(l, &t, 0);
1854 if (l == t)
1855 if (ms->flags & MAGIC_CHECK)
1856 file_magwarn(ms, "offset `%s' invalid", l);
1857 l = t;
1858
1859 if (m->flag & INDIR) {
1860 m->in_type = FILE_LONG;
1861 m->in_offset = 0;
1862 /*
1863 * read [.lbs][+-]nnnnn)
1864 */
1865 if (*l == '.') {
1866 l++;
1867 switch (*l) {
1868 case 'l':
1869 m->in_type = FILE_LELONG;
1870 break;
1871 case 'L':
1872 m->in_type = FILE_BELONG;
1873 break;
1874 case 'm':
1875 m->in_type = FILE_MELONG;
1876 break;
1877 case 'h':
1878 case 's':
1879 m->in_type = FILE_LESHORT;
1880 break;
1881 case 'H':
1882 case 'S':
1883 m->in_type = FILE_BESHORT;
1884 break;
1885 case 'c':
1886 case 'b':
1887 case 'C':
1888 case 'B':
1889 m->in_type = FILE_BYTE;
1890 break;
1891 case 'e':
1892 case 'f':
1893 case 'g':
1894 m->in_type = FILE_LEDOUBLE;
1895 break;
1896 case 'E':
1897 case 'F':
1898 case 'G':
1899 m->in_type = FILE_BEDOUBLE;
1900 break;
1901 case 'i':
1902 m->in_type = FILE_LEID3;
1903 break;
1904 case 'I':
1905 m->in_type = FILE_BEID3;
1906 break;
1907 default:
1908 if (ms->flags & MAGIC_CHECK)
1909 file_magwarn(ms,
1910 "indirect offset type `%c' invalid",
1911 *l);
1912 break;
1913 }
1914 l++;
1915 }
1916
1917 m->in_op = 0;
1918 if (*l == '~') {
1919 m->in_op |= FILE_OPINVERSE;
1920 l++;
1921 }
1922 if ((op = get_op(*l)) != -1) {
1923 m->in_op |= op;
1924 l++;
1925 }
1926 if (*l == '(') {
1927 m->in_op |= FILE_OPINDIRECT;
1928 l++;
1929 }
1930 if (isdigit((unsigned char)*l) || *l == '-') {
1931 m->in_offset = (int32_t)strtol(l, &t, 0);
1932 if (l == t)
1933 if (ms->flags & MAGIC_CHECK)
1934 file_magwarn(ms,
1935 "in_offset `%s' invalid", l);
1936 l = t;
1937 }
1938 if (*l++ != ')' ||
1939 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')'))
1940 if (ms->flags & MAGIC_CHECK)
1941 file_magwarn(ms,
1942 "missing ')' in indirect offset");
1943 }
1944 EATAB;
1945
1946 #ifdef ENABLE_CONDITIONALS
1947 m->cond = get_cond(l, &l);
1948 if (check_cond(ms, m->cond, cont_level) == -1)
1949 return -1;
1950
1951 EATAB;
1952 #endif
1953
1954 /*
1955 * Parse the type.
1956 */
1957 if (*l == 'u') {
1958 /*
1959 * Try it as a keyword type prefixed by "u"; match what
1960 * follows the "u". If that fails, try it as an SUS
1961 * integer type.
1962 */
1963 m->type = get_type(type_tbl, l + 1, &l);
1964 if (m->type == FILE_INVALID) {
1965 /*
1966 * Not a keyword type; parse it as an SUS type,
1967 * 'u' possibly followed by a number or C/S/L.
1968 */
1969 m->type = get_standard_integer_type(l, &l);
1970 }
1971 /* It's unsigned. */
1972 if (m->type != FILE_INVALID)
1973 m->flag |= UNSIGNED;
1974 } else {
1975 /*
1976 * Try it as a keyword type. If that fails, try it as
1977 * an SUS integer type if it begins with "d" or as an
1978 * SUS string type if it begins with "s". In any case,
1979 * it's not unsigned.
1980 */
1981 m->type = get_type(type_tbl, l, &l);
1982 if (m->type == FILE_INVALID) {
1983 /*
1984 * Not a keyword type; parse it as an SUS type,
1985 * either 'd' possibly followed by a number or
1986 * C/S/L, or just 's'.
1987 */
1988 if (*l == 'd')
1989 m->type = get_standard_integer_type(l, &l);
1990 else if (*l == 's' && !isalpha((unsigned char)l[1])) {
1991 m->type = FILE_STRING;
1992 ++l;
1993 }
1994 }
1995 }
1996
1997 if (m->type == FILE_INVALID) {
1998 /* Not found - try it as a special keyword. */
1999 m->type = get_type(special_tbl, l, &l);
2000 }
2001
2002 if (m->type == FILE_INVALID) {
2003 if (ms->flags & MAGIC_CHECK)
2004 file_magwarn(ms, "type `%s' invalid", l);
2005 return -1;
2006 }
2007
2008 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */
2009 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */
2010
2011 m->mask_op = 0;
2012 if (*l == '~') {
2013 if (!IS_STRING(m->type))
2014 m->mask_op |= FILE_OPINVERSE;
2015 else if (ms->flags & MAGIC_CHECK)
2016 file_magwarn(ms, "'~' invalid for string types");
2017 ++l;
2018 }
2019 m->str_range = 0;
2020 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0;
2021 if ((op = get_op(*l)) != -1) {
2022 if (IS_STRING(m->type)) {
2023 int r;
2024
2025 if (op != FILE_OPDIVIDE) {
2026 if (ms->flags & MAGIC_CHECK)
2027 file_magwarn(ms,
2028 "invalid string/indirect op: "
2029 "`%c'", *t);
2030 return -1;
2031 }
2032
2033 if (m->type == FILE_INDIRECT)
2034 r = parse_indirect_modifier(ms, m, &l);
2035 else
2036 r = parse_string_modifier(ms, m, &l);
2037 if (r == -1)
2038 return -1;
2039 } else
2040 parse_op_modifier(ms, m, &l, op);
2041 }
2042
2043 /*
2044 * We used to set mask to all 1's here, instead let's just not do
2045 * anything if mask = 0 (unless you have a better idea)
2046 */
2047 EATAB;
2048
2049 switch (*l) {
2050 case '>':
2051 case '<':
2052 m->reln = *l;
2053 ++l;
2054 if (*l == '=') {
2055 if (ms->flags & MAGIC_CHECK) {
2056 file_magwarn(ms, "%c= not supported",
2057 m->reln);
2058 return -1;
2059 }
2060 ++l;
2061 }
2062 break;
2063 /* Old-style anding: "0 byte &0x80 dynamically linked" */
2064 case '&':
2065 case '^':
2066 case '=':
2067 m->reln = *l;
2068 ++l;
2069 if (*l == '=') {
2070 /* HP compat: ignore &= etc. */
2071 ++l;
2072 }
2073 break;
2074 case '!':
2075 m->reln = *l;
2076 ++l;
2077 break;
2078 default:
2079 m->reln = '='; /* the default relation */
2080 if (*l == 'x' && ((isascii((unsigned char)l[1]) &&
2081 isspace((unsigned char)l[1])) || !l[1])) {
2082 m->reln = *l;
2083 ++l;
2084 }
2085 break;
2086 }
2087 /*
2088 * Grab the value part, except for an 'x' reln.
2089 */
2090 if (m->reln != 'x' && getvalue(ms, m, &l, action))
2091 return -1;
2092
2093 /*
2094 * TODO finish this macro and start using it!
2095 * #define offsetcheck {if (offset > HOWMANY-1)
2096 * magwarn("offset too big"); }
2097 */
2098
2099 /*
2100 * Now get last part - the description
2101 */
2102 EATAB;
2103 if (l[0] == '\b') {
2104 ++l;
2105 m->flag |= NOSPACE;
2106 } else if ((l[0] == '\\') && (l[1] == 'b')) {
2107 ++l;
2108 ++l;
2109 m->flag |= NOSPACE;
2110 }
2111 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); )
2112 continue;
2113 if (i == sizeof(m->desc)) {
2114 m->desc[sizeof(m->desc) - 1] = '\0';
2115 if (ms->flags & MAGIC_CHECK)
2116 file_magwarn(ms, "description `%s' truncated", m->desc);
2117 }
2118
2119 /*
2120 * We only do this check while compiling, or if any of the magic
2121 * files were not compiled.
2122 */
2123 if (ms->flags & MAGIC_CHECK) {
2124 if (check_format(ms, m) == -1)
2125 return -1;
2126 }
2127 #ifndef COMPILE_ONLY
2128 if (action == FILE_CHECK) {
2129 file_mdump(m);
2130 }
2131 #endif
2132 m->mimetype[0] = '\0'; /* initialise MIME type to none */
2133 return 0;
2134 }
2135
2136 /*
2137 * parse a STRENGTH annotation line from magic file, put into magic[index - 1]
2138 * if valid
2139 */
2140 private int
parse_strength(struct magic_set * ms,struct magic_entry * me,const char * line)2141 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line)
2142 {
2143 const char *l = line;
2144 char *el;
2145 unsigned long factor;
2146 struct magic *m = &me->mp[0];
2147
2148 if (m->factor_op != FILE_FACTOR_OP_NONE) {
2149 file_magwarn(ms,
2150 "Current entry already has a strength type: %c %d",
2151 m->factor_op, m->factor);
2152 return -1;
2153 }
2154 if (m->type == FILE_NAME) {
2155 file_magwarn(ms, "%s: Strength setting is not supported in "
2156 "\"name\" magic entries", m->value.s);
2157 return -1;
2158 }
2159 EATAB;
2160 switch (*l) {
2161 case FILE_FACTOR_OP_NONE:
2162 case FILE_FACTOR_OP_PLUS:
2163 case FILE_FACTOR_OP_MINUS:
2164 case FILE_FACTOR_OP_TIMES:
2165 case FILE_FACTOR_OP_DIV:
2166 m->factor_op = *l++;
2167 break;
2168 default:
2169 file_magwarn(ms, "Unknown factor op `%c'", *l);
2170 return -1;
2171 }
2172 EATAB;
2173 factor = strtoul(l, &el, 0);
2174 if (factor > 255) {
2175 file_magwarn(ms, "Too large factor `%lu'", factor);
2176 goto out;
2177 }
2178 if (*el && !isspace((unsigned char)*el)) {
2179 file_magwarn(ms, "Bad factor `%s'", l);
2180 goto out;
2181 }
2182 m->factor = (uint8_t)factor;
2183 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) {
2184 file_magwarn(ms, "Cannot have factor op `%c' and factor %u",
2185 m->factor_op, m->factor);
2186 goto out;
2187 }
2188 return 0;
2189 out:
2190 m->factor_op = FILE_FACTOR_OP_NONE;
2191 m->factor = 0;
2192 return -1;
2193 }
2194
2195 private int
goodchar(unsigned char x,const char * extra)2196 goodchar(unsigned char x, const char *extra)
2197 {
2198 return (isascii(x) && isalnum(x)) || strchr(extra, x);
2199 }
2200
2201 private int
parse_extra(struct magic_set * ms,struct magic_entry * me,const char * line,off_t off,size_t len,const char * name,const char * extra,int nt)2202 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line,
2203 off_t off, size_t len, const char *name, const char *extra, int nt)
2204 {
2205 size_t i;
2206 const char *l = line;
2207 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1];
2208 char *buf = CAST(char *, CAST(void *, m)) + off;
2209
2210 if (buf[0] != '\0') {
2211 len = nt ? strlen(buf) : len;
2212 file_magwarn(ms, "Current entry already has a %s type "
2213 "`%.*s', new type `%s'", name, (int)len, buf, l);
2214 return -1;
2215 }
2216
2217 if (*m->desc == '\0') {
2218 file_magwarn(ms, "Current entry does not yet have a "
2219 "description for adding a %s type", name);
2220 return -1;
2221 }
2222
2223 EATAB;
2224 for (i = 0; *l && i < len && goodchar(*l, extra); buf[i++] = *l++)
2225 continue;
2226
2227 if (i == len && *l) {
2228 if (nt)
2229 buf[len - 1] = '\0';
2230 if (ms->flags & MAGIC_CHECK)
2231 file_magwarn(ms, "%s type `%s' truncated %"
2232 SIZE_T_FORMAT "u", name, line, i);
2233 } else {
2234 if (!isspace((unsigned char)*l) && !goodchar(*l, extra))
2235 file_magwarn(ms, "%s type `%s' has bad char '%c'",
2236 name, line, *l);
2237 if (nt)
2238 buf[i] = '\0';
2239 }
2240
2241 if (i > 0)
2242 return 0;
2243
2244 file_magerror(ms, "Bad magic entry '%s'", line);
2245 return -1;
2246 }
2247
2248 /*
2249 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into
2250 * magic[index - 1]
2251 */
2252 private int
parse_apple(struct magic_set * ms,struct magic_entry * me,const char * line)2253 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line)
2254 {
2255 struct magic *m = &me->mp[0];
2256
2257 return parse_extra(ms, me, line,
2258 CAST(off_t, offsetof(struct magic, apple)),
2259 sizeof(m->apple), "APPLE", "!+-./", 0);
2260 }
2261
2262 /*
2263 * parse a MIME annotation line from magic file, put into magic[index - 1]
2264 * if valid
2265 */
2266 private int
parse_mime(struct magic_set * ms,struct magic_entry * me,const char * line)2267 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line)
2268 {
2269 struct magic *m = &me->mp[0];
2270
2271 return parse_extra(ms, me, line,
2272 CAST(off_t, offsetof(struct magic, mimetype)),
2273 sizeof(m->mimetype), "MIME", "+-/.", 1);
2274 }
2275
2276 private int
check_format_type(const char * ptr,int type)2277 check_format_type(const char *ptr, int type)
2278 {
2279 int quad = 0, h;
2280 if (*ptr == '\0') {
2281 /* Missing format string; bad */
2282 return -1;
2283 }
2284
2285 switch (file_formats[type]) {
2286 case FILE_FMT_QUAD:
2287 quad = 1;
2288 /*FALLTHROUGH*/
2289 case FILE_FMT_NUM:
2290 if (quad == 0) {
2291 switch (type) {
2292 case FILE_BYTE:
2293 h = 2;
2294 break;
2295 case FILE_SHORT:
2296 case FILE_BESHORT:
2297 case FILE_LESHORT:
2298 h = 1;
2299 break;
2300 case FILE_LONG:
2301 case FILE_BELONG:
2302 case FILE_LELONG:
2303 case FILE_MELONG:
2304 case FILE_LEID3:
2305 case FILE_BEID3:
2306 case FILE_INDIRECT:
2307 h = 0;
2308 break;
2309 default:
2310 abort();
2311 }
2312 } else
2313 h = 0;
2314 if (*ptr == '-')
2315 ptr++;
2316 if (*ptr == '.')
2317 ptr++;
2318 while (isdigit((unsigned char)*ptr)) ptr++;
2319 if (*ptr == '.')
2320 ptr++;
2321 while (isdigit((unsigned char)*ptr)) ptr++;
2322 if (quad) {
2323 if (*ptr++ != 'l')
2324 return -1;
2325 if (*ptr++ != 'l')
2326 return -1;
2327 }
2328
2329 switch (*ptr++) {
2330 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */
2331 /* so don't accept the 'l' modifier */
2332 case 'l':
2333 switch (*ptr++) {
2334 case 'i':
2335 case 'd':
2336 case 'u':
2337 case 'o':
2338 case 'x':
2339 case 'X':
2340 return h != 0 ? -1 : 0;
2341 default:
2342 return -1;
2343 }
2344
2345 /*
2346 * Don't accept h and hh modifiers. They make writing
2347 * magic entries more complicated, for very little benefit
2348 */
2349 case 'h':
2350 if (h-- <= 0)
2351 return -1;
2352 switch (*ptr++) {
2353 case 'h':
2354 if (h-- <= 0)
2355 return -1;
2356 switch (*ptr++) {
2357 case 'i':
2358 case 'd':
2359 case 'u':
2360 case 'o':
2361 case 'x':
2362 case 'X':
2363 return 0;
2364 default:
2365 return -1;
2366 }
2367 case 'i':
2368 case 'd':
2369 case 'u':
2370 case 'o':
2371 case 'x':
2372 case 'X':
2373 return h != 0 ? -1 : 0;
2374 default:
2375 return -1;
2376 }
2377 #endif
2378 case 'c':
2379 return h != 2 ? -1 : 0;
2380 case 'i':
2381 case 'd':
2382 case 'u':
2383 case 'o':
2384 case 'x':
2385 case 'X':
2386 #ifdef STRICT_FORMAT
2387 return h != 0 ? -1 : 0;
2388 #else
2389 return 0;
2390 #endif
2391 default:
2392 return -1;
2393 }
2394
2395 case FILE_FMT_FLOAT:
2396 case FILE_FMT_DOUBLE:
2397 if (*ptr == '-')
2398 ptr++;
2399 if (*ptr == '.')
2400 ptr++;
2401 while (isdigit((unsigned char)*ptr)) ptr++;
2402 if (*ptr == '.')
2403 ptr++;
2404 while (isdigit((unsigned char)*ptr)) ptr++;
2405
2406 switch (*ptr++) {
2407 case 'e':
2408 case 'E':
2409 case 'f':
2410 case 'F':
2411 case 'g':
2412 case 'G':
2413 return 0;
2414
2415 default:
2416 return -1;
2417 }
2418
2419
2420 case FILE_FMT_STR:
2421 if (*ptr == '-')
2422 ptr++;
2423 while (isdigit((unsigned char )*ptr))
2424 ptr++;
2425 if (*ptr == '.') {
2426 ptr++;
2427 while (isdigit((unsigned char )*ptr))
2428 ptr++;
2429 }
2430
2431 switch (*ptr++) {
2432 case 's':
2433 return 0;
2434 default:
2435 return -1;
2436 }
2437
2438 default:
2439 /* internal error */
2440 abort();
2441 }
2442 /*NOTREACHED*/
2443 return -1;
2444 }
2445
2446 /*
2447 * Check that the optional printf format in description matches
2448 * the type of the magic.
2449 */
2450 private int
check_format(struct magic_set * ms,struct magic * m)2451 check_format(struct magic_set *ms, struct magic *m)
2452 {
2453 char *ptr;
2454
2455 for (ptr = m->desc; *ptr; ptr++)
2456 if (*ptr == '%')
2457 break;
2458 if (*ptr == '\0') {
2459 /* No format string; ok */
2460 return 1;
2461 }
2462
2463 assert(file_nformats == file_nnames);
2464
2465 if (m->type >= file_nformats) {
2466 file_magwarn(ms, "Internal error inconsistency between "
2467 "m->type and format strings");
2468 return -1;
2469 }
2470 if (file_formats[m->type] == FILE_FMT_NONE) {
2471 file_magwarn(ms, "No format string for `%s' with description "
2472 "`%s'", m->desc, file_names[m->type]);
2473 return -1;
2474 }
2475
2476 ptr++;
2477 if (check_format_type(ptr, m->type) == -1) {
2478 /*
2479 * TODO: this error message is unhelpful if the format
2480 * string is not one character long
2481 */
2482 file_magwarn(ms, "Printf format `%c' is not valid for type "
2483 "`%s' in description `%s'", *ptr ? *ptr : '?',
2484 file_names[m->type], m->desc);
2485 return -1;
2486 }
2487
2488 for (; *ptr; ptr++) {
2489 if (*ptr == '%') {
2490 file_magwarn(ms,
2491 "Too many format strings (should have at most one) "
2492 "for `%s' with description `%s'",
2493 file_names[m->type], m->desc);
2494 return -1;
2495 }
2496 }
2497 return 0;
2498 }
2499
2500 /*
2501 * Read a numeric value from a pointer, into the value union of a magic
2502 * pointer, according to the magic type. Update the string pointer to point
2503 * just after the number read. Return 0 for success, non-zero for failure.
2504 */
2505 private int
getvalue(struct magic_set * ms,struct magic * m,const char ** p,int action)2506 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action)
2507 {
2508 switch (m->type) {
2509 case FILE_BESTRING16:
2510 case FILE_LESTRING16:
2511 case FILE_STRING:
2512 case FILE_PSTRING:
2513 case FILE_REGEX:
2514 case FILE_SEARCH:
2515 case FILE_NAME:
2516 case FILE_USE:
2517 *p = getstr(ms, m, *p, action == FILE_COMPILE);
2518 if (*p == NULL) {
2519 if (ms->flags & MAGIC_CHECK)
2520 file_magwarn(ms, "cannot get string from `%s'",
2521 m->value.s);
2522 return -1;
2523 }
2524 if (m->type == FILE_REGEX) {
2525 file_regex_t rx;
2526 int rc = file_regcomp(&rx, m->value.s, REG_EXTENDED);
2527 if (rc) {
2528 if (ms->flags & MAGIC_CHECK)
2529 file_regerror(&rx, rc, ms);
2530 }
2531 file_regfree(&rx);
2532 return rc ? -1 : 0;
2533 }
2534 return 0;
2535 case FILE_FLOAT:
2536 case FILE_BEFLOAT:
2537 case FILE_LEFLOAT:
2538 if (m->reln != 'x') {
2539 char *ep;
2540 #ifdef HAVE_STRTOF
2541 m->value.f = strtof(*p, &ep);
2542 #else
2543 m->value.f = (float)strtod(*p, &ep);
2544 #endif
2545 *p = ep;
2546 }
2547 return 0;
2548 case FILE_DOUBLE:
2549 case FILE_BEDOUBLE:
2550 case FILE_LEDOUBLE:
2551 if (m->reln != 'x') {
2552 char *ep;
2553 m->value.d = strtod(*p, &ep);
2554 *p = ep;
2555 }
2556 return 0;
2557 default:
2558 if (m->reln != 'x') {
2559 char *ep;
2560 m->value.q = file_signextend(ms, m,
2561 (uint64_t)strtoull(*p, &ep, 0));
2562 *p = ep;
2563 eatsize(p);
2564 }
2565 return 0;
2566 }
2567 }
2568
2569 /*
2570 * Convert a string containing C character escapes. Stop at an unescaped
2571 * space or tab.
2572 * Copy the converted version to "m->value.s", and the length in m->vallen.
2573 * Return updated scan pointer as function result. Warn if set.
2574 */
2575 private const char *
getstr(struct magic_set * ms,struct magic * m,const char * s,int warn)2576 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn)
2577 {
2578 const char *origs = s;
2579 char *p = m->value.s;
2580 size_t plen = sizeof(m->value.s);
2581 char *origp = p;
2582 char *pmax = p + plen - 1;
2583 int c;
2584 int val;
2585
2586 while ((c = *s++) != '\0') {
2587 if (isspace((unsigned char) c))
2588 break;
2589 if (p >= pmax) {
2590 file_error(ms, 0, "string too long: `%s'", origs);
2591 return NULL;
2592 }
2593 if (c == '\\') {
2594 switch(c = *s++) {
2595
2596 case '\0':
2597 if (warn)
2598 file_magwarn(ms, "incomplete escape");
2599 goto out;
2600
2601 case '\t':
2602 if (warn) {
2603 file_magwarn(ms,
2604 "escaped tab found, use \\t instead");
2605 warn = 0; /* already did */
2606 }
2607 /*FALLTHROUGH*/
2608 default:
2609 if (warn) {
2610 if (isprint((unsigned char)c)) {
2611 /* Allow escaping of
2612 * ``relations'' */
2613 if (strchr("<>&^=!", c) == NULL
2614 && (m->type != FILE_REGEX ||
2615 strchr("[]().*?^$|{}", c)
2616 == NULL)) {
2617 file_magwarn(ms, "no "
2618 "need to escape "
2619 "`%c'", c);
2620 }
2621 } else {
2622 file_magwarn(ms,
2623 "unknown escape sequence: "
2624 "\\%03o", c);
2625 }
2626 }
2627 /*FALLTHROUGH*/
2628 /* space, perhaps force people to use \040? */
2629 case ' ':
2630 #if 0
2631 /*
2632 * Other things people escape, but shouldn't need to,
2633 * so we disallow them
2634 */
2635 case '\'':
2636 case '"':
2637 case '?':
2638 #endif
2639 /* Relations */
2640 case '>':
2641 case '<':
2642 case '&':
2643 case '^':
2644 case '=':
2645 case '!':
2646 /* and baskslash itself */
2647 case '\\':
2648 *p++ = (char) c;
2649 break;
2650
2651 case 'a':
2652 *p++ = '\a';
2653 break;
2654
2655 case 'b':
2656 *p++ = '\b';
2657 break;
2658
2659 case 'f':
2660 *p++ = '\f';
2661 break;
2662
2663 case 'n':
2664 *p++ = '\n';
2665 break;
2666
2667 case 'r':
2668 *p++ = '\r';
2669 break;
2670
2671 case 't':
2672 *p++ = '\t';
2673 break;
2674
2675 case 'v':
2676 *p++ = '\v';
2677 break;
2678
2679 /* \ and up to 3 octal digits */
2680 case '0':
2681 case '1':
2682 case '2':
2683 case '3':
2684 case '4':
2685 case '5':
2686 case '6':
2687 case '7':
2688 val = c - '0';
2689 c = *s++; /* try for 2 */
2690 if (c >= '0' && c <= '7') {
2691 val = (val << 3) | (c - '0');
2692 c = *s++; /* try for 3 */
2693 if (c >= '0' && c <= '7')
2694 val = (val << 3) | (c-'0');
2695 else
2696 --s;
2697 }
2698 else
2699 --s;
2700 *p++ = (char)val;
2701 break;
2702
2703 /* \x and up to 2 hex digits */
2704 case 'x':
2705 val = 'x'; /* Default if no digits */
2706 c = hextoint(*s++); /* Get next char */
2707 if (c >= 0) {
2708 val = c;
2709 c = hextoint(*s++);
2710 if (c >= 0)
2711 val = (val << 4) + c;
2712 else
2713 --s;
2714 } else
2715 --s;
2716 *p++ = (char)val;
2717 break;
2718 }
2719 } else
2720 *p++ = (char)c;
2721 }
2722 out:
2723 *p = '\0';
2724 m->vallen = CAST(unsigned char, (p - origp));
2725 if (m->type == FILE_PSTRING)
2726 m->vallen += (unsigned char)file_pstring_length_size(m);
2727 return s;
2728 }
2729
2730
2731 /* Single hex char to int; -1 if not a hex char. */
2732 private int
hextoint(int c)2733 hextoint(int c)
2734 {
2735 if (!isascii((unsigned char) c))
2736 return -1;
2737 if (isdigit((unsigned char) c))
2738 return c - '0';
2739 if ((c >= 'a') && (c <= 'f'))
2740 return c + 10 - 'a';
2741 if (( c>= 'A') && (c <= 'F'))
2742 return c + 10 - 'A';
2743 return -1;
2744 }
2745
2746
2747 /*
2748 * Print a string containing C character escapes.
2749 */
2750 protected void
file_showstr(FILE * fp,const char * s,size_t len)2751 file_showstr(FILE *fp, const char *s, size_t len)
2752 {
2753 char c;
2754
2755 for (;;) {
2756 if (len == ~0U) {
2757 c = *s++;
2758 if (c == '\0')
2759 break;
2760 }
2761 else {
2762 if (len-- == 0)
2763 break;
2764 c = *s++;
2765 }
2766 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */
2767 (void) fputc(c, fp);
2768 else {
2769 (void) fputc('\\', fp);
2770 switch (c) {
2771 case '\a':
2772 (void) fputc('a', fp);
2773 break;
2774
2775 case '\b':
2776 (void) fputc('b', fp);
2777 break;
2778
2779 case '\f':
2780 (void) fputc('f', fp);
2781 break;
2782
2783 case '\n':
2784 (void) fputc('n', fp);
2785 break;
2786
2787 case '\r':
2788 (void) fputc('r', fp);
2789 break;
2790
2791 case '\t':
2792 (void) fputc('t', fp);
2793 break;
2794
2795 case '\v':
2796 (void) fputc('v', fp);
2797 break;
2798
2799 default:
2800 (void) fprintf(fp, "%.3o", c & 0377);
2801 break;
2802 }
2803 }
2804 }
2805 }
2806
2807 /*
2808 * eatsize(): Eat the size spec from a number [eg. 10UL]
2809 */
2810 private void
eatsize(const char ** p)2811 eatsize(const char **p)
2812 {
2813 const char *l = *p;
2814
2815 if (LOWCASE(*l) == 'u')
2816 l++;
2817
2818 switch (LOWCASE(*l)) {
2819 case 'l': /* long */
2820 case 's': /* short */
2821 case 'h': /* short */
2822 case 'b': /* char/byte */
2823 case 'c': /* char/byte */
2824 l++;
2825 /*FALLTHROUGH*/
2826 default:
2827 break;
2828 }
2829
2830 *p = l;
2831 }
2832
2833 /*
2834 * handle a buffer containing a compiled file.
2835 */
2836 private struct magic_map *
apprentice_buf(struct magic_set * ms,struct magic * buf,size_t len)2837 apprentice_buf(struct magic_set *ms, struct magic *buf, size_t len)
2838 {
2839 struct magic_map *map;
2840
2841 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) {
2842 file_oomem(ms, sizeof(*map));
2843 return NULL;
2844 }
2845 map->len = len;
2846 map->p = buf;
2847 map->type = MAP_TYPE_USER;
2848 if (check_buffer(ms, map, "buffer") != 0) {
2849 apprentice_unmap(map);
2850 return NULL;
2851 }
2852 return map;
2853 }
2854
2855 /*
2856 * handle a compiled file.
2857 */
2858
2859 private struct magic_map *
apprentice_map(struct magic_set * ms,const char * fn)2860 apprentice_map(struct magic_set *ms, const char *fn)
2861 {
2862 int fd;
2863 struct stat st;
2864 char *dbname = NULL;
2865 struct magic_map *map;
2866
2867 fd = -1;
2868 if ((map = CAST(struct magic_map *, calloc(1, sizeof(*map)))) == NULL) {
2869 file_oomem(ms, sizeof(*map));
2870 goto error;
2871 }
2872
2873 dbname = mkdbname(ms, fn, 0);
2874 if (dbname == NULL)
2875 goto error;
2876
2877 if ((fd = open(dbname, O_RDONLY|O_BINARY)) == -1)
2878 goto error;
2879
2880 if (fstat(fd, &st) == -1) {
2881 file_error(ms, errno, "cannot stat `%s'", dbname);
2882 goto error;
2883 }
2884 if (st.st_size < 8 || st.st_size > MAXMAGIC_SIZE) {
2885 file_error(ms, 0, "file `%s' is too %s", dbname,
2886 st.st_size < 8 ? "small" : "large");
2887 goto error;
2888 }
2889
2890 map->len = (size_t)st.st_size;
2891 #ifdef QUICK
2892 if ((map->p = mmap(0, (size_t)st.st_size, PROT_READ|PROT_WRITE,
2893 MAP_PRIVATE|MAP_FILE, fd, (off_t)0)) == MAP_FAILED) {
2894 file_error(ms, errno, "cannot map `%s'", dbname);
2895 goto error;
2896 }
2897 map->type = MAP_TYPE_MMAP;
2898 #else
2899 if ((map->p = CAST(void *, malloc(map->len))) == NULL) {
2900 file_oomem(ms, map->len);
2901 goto error;
2902 }
2903 if (read(fd, map->p, map->len) != (ssize_t)map->len) {
2904 file_badread(ms);
2905 goto error;
2906 }
2907 map->type = MAP_TYPE_MALLOC;
2908 #define RET 1
2909 #endif
2910 (void)close(fd);
2911 fd = -1;
2912
2913 if (check_buffer(ms, map, dbname) != 0)
2914 goto error;
2915
2916 free(dbname);
2917 return map;
2918
2919 error:
2920 if (fd != -1)
2921 (void)close(fd);
2922 apprentice_unmap(map);
2923 free(dbname);
2924 return NULL;
2925 }
2926
2927 private int
check_buffer(struct magic_set * ms,struct magic_map * map,const char * dbname)2928 check_buffer(struct magic_set *ms, struct magic_map *map, const char *dbname)
2929 {
2930 uint32_t *ptr;
2931 uint32_t entries, nentries;
2932 uint32_t version;
2933 int i, needsbyteswap;
2934
2935 ptr = CAST(uint32_t *, map->p);
2936 if (*ptr != MAGICNO) {
2937 if (swap4(*ptr) != MAGICNO) {
2938 file_error(ms, 0, "bad magic in `%s'", dbname);
2939 return -1;
2940 }
2941 needsbyteswap = 1;
2942 } else
2943 needsbyteswap = 0;
2944 if (needsbyteswap)
2945 version = swap4(ptr[1]);
2946 else
2947 version = ptr[1];
2948 if (version != VERSIONNO) {
2949 file_error(ms, 0, "File %s supports only version %d magic "
2950 "files. `%s' is version %d", VERSION,
2951 VERSIONNO, dbname, version);
2952 return -1;
2953 }
2954 entries = (uint32_t)(map->len / sizeof(struct magic));
2955 if ((entries * sizeof(struct magic)) != map->len) {
2956 file_error(ms, 0, "Size of `%s' %" SIZE_T_FORMAT "u is not "
2957 "a multiple of %" SIZE_T_FORMAT "u",
2958 dbname, map->len, sizeof(struct magic));
2959 return -1;
2960 }
2961 map->magic[0] = CAST(struct magic *, map->p) + 1;
2962 nentries = 0;
2963 for (i = 0; i < MAGIC_SETS; i++) {
2964 if (needsbyteswap)
2965 map->nmagic[i] = swap4(ptr[i + 2]);
2966 else
2967 map->nmagic[i] = ptr[i + 2];
2968 if (i != MAGIC_SETS - 1)
2969 map->magic[i + 1] = map->magic[i] + map->nmagic[i];
2970 nentries += map->nmagic[i];
2971 }
2972 if (entries != nentries + 1) {
2973 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u",
2974 dbname, entries, nentries + 1);
2975 return -1;
2976 }
2977 if (needsbyteswap)
2978 for (i = 0; i < MAGIC_SETS; i++)
2979 byteswap(map->magic[i], map->nmagic[i]);
2980 return 0;
2981 }
2982
2983 /*
2984 * handle an mmaped file.
2985 */
2986 private int
apprentice_compile(struct magic_set * ms,struct magic_map * map,const char * fn)2987 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn)
2988 {
2989 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS;
2990 static const size_t m = sizeof(**map->magic);
2991 int fd = -1;
2992 size_t len;
2993 char *dbname;
2994 int rv = -1;
2995 uint32_t i;
2996 union {
2997 struct magic m;
2998 uint32_t h[2 + MAGIC_SETS];
2999 } hdr;
3000
3001 dbname = mkdbname(ms, fn, 1);
3002
3003 if (dbname == NULL)
3004 goto out;
3005
3006 if ((fd = open(dbname, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0644)) == -1)
3007 {
3008 file_error(ms, errno, "cannot open `%s'", dbname);
3009 goto out;
3010 }
3011 memset(&hdr, 0, sizeof(hdr));
3012 hdr.h[0] = MAGICNO;
3013 hdr.h[1] = VERSIONNO;
3014 memcpy(hdr.h + 2, map->nmagic, nm);
3015
3016 if (write(fd, &hdr, sizeof(hdr)) != (ssize_t)sizeof(hdr)) {
3017 file_error(ms, errno, "error writing `%s'", dbname);
3018 goto out;
3019 }
3020
3021 for (i = 0; i < MAGIC_SETS; i++) {
3022 len = m * map->nmagic[i];
3023 if (write(fd, map->magic[i], len) != (ssize_t)len) {
3024 file_error(ms, errno, "error writing `%s'", dbname);
3025 goto out;
3026 }
3027 }
3028
3029 if (fd != -1)
3030 (void)close(fd);
3031 rv = 0;
3032 out:
3033 free(dbname);
3034 return rv;
3035 }
3036
3037 private const char ext[] = ".mgc";
3038 /*
3039 * make a dbname
3040 */
3041 private char *
mkdbname(struct magic_set * ms,const char * fn,int strip)3042 mkdbname(struct magic_set *ms, const char *fn, int strip)
3043 {
3044 const char *p, *q;
3045 char *buf;
3046
3047 if (strip) {
3048 if ((p = strrchr(fn, '/')) != NULL)
3049 fn = ++p;
3050 }
3051
3052 for (q = fn; *q; q++)
3053 continue;
3054 /* Look for .mgc */
3055 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--)
3056 if (*p != *q)
3057 break;
3058
3059 /* Did not find .mgc, restore q */
3060 if (p >= ext)
3061 while (*q)
3062 q++;
3063
3064 q++;
3065 /* Compatibility with old code that looked in .mime */
3066 if (ms->flags & MAGIC_MIME) {
3067 if (asprintf(&buf, "%.*s.mime%s", (int)(q - fn), fn, ext) < 0)
3068 return NULL;
3069 if (access(buf, R_OK) != -1) {
3070 ms->flags &= MAGIC_MIME_TYPE;
3071 return buf;
3072 }
3073 free(buf);
3074 }
3075 if (asprintf(&buf, "%.*s%s", (int)(q - fn), fn, ext) < 0)
3076 return NULL;
3077
3078 /* Compatibility with old code that looked in .mime */
3079 if (strstr(p, ".mime") != NULL)
3080 ms->flags &= MAGIC_MIME_TYPE;
3081 return buf;
3082 }
3083
3084 /*
3085 * Byteswap an mmap'ed file if needed
3086 */
3087 private void
byteswap(struct magic * magic,uint32_t nmagic)3088 byteswap(struct magic *magic, uint32_t nmagic)
3089 {
3090 uint32_t i;
3091 for (i = 0; i < nmagic; i++)
3092 bs1(&magic[i]);
3093 }
3094
3095 /*
3096 * swap a short
3097 */
3098 private uint16_t
swap2(uint16_t sv)3099 swap2(uint16_t sv)
3100 {
3101 uint16_t rv;
3102 uint8_t *s = (uint8_t *)(void *)&sv;
3103 uint8_t *d = (uint8_t *)(void *)&rv;
3104 d[0] = s[1];
3105 d[1] = s[0];
3106 return rv;
3107 }
3108
3109 /*
3110 * swap an int
3111 */
3112 private uint32_t
swap4(uint32_t sv)3113 swap4(uint32_t sv)
3114 {
3115 uint32_t rv;
3116 uint8_t *s = (uint8_t *)(void *)&sv;
3117 uint8_t *d = (uint8_t *)(void *)&rv;
3118 d[0] = s[3];
3119 d[1] = s[2];
3120 d[2] = s[1];
3121 d[3] = s[0];
3122 return rv;
3123 }
3124
3125 /*
3126 * swap a quad
3127 */
3128 private uint64_t
swap8(uint64_t sv)3129 swap8(uint64_t sv)
3130 {
3131 uint64_t rv;
3132 uint8_t *s = (uint8_t *)(void *)&sv;
3133 uint8_t *d = (uint8_t *)(void *)&rv;
3134 #if 0
3135 d[0] = s[3];
3136 d[1] = s[2];
3137 d[2] = s[1];
3138 d[3] = s[0];
3139 d[4] = s[7];
3140 d[5] = s[6];
3141 d[6] = s[5];
3142 d[7] = s[4];
3143 #else
3144 d[0] = s[7];
3145 d[1] = s[6];
3146 d[2] = s[5];
3147 d[3] = s[4];
3148 d[4] = s[3];
3149 d[5] = s[2];
3150 d[6] = s[1];
3151 d[7] = s[0];
3152 #endif
3153 return rv;
3154 }
3155
3156 /*
3157 * byteswap a single magic entry
3158 */
3159 private void
bs1(struct magic * m)3160 bs1(struct magic *m)
3161 {
3162 m->cont_level = swap2(m->cont_level);
3163 m->offset = swap4((uint32_t)m->offset);
3164 m->in_offset = swap4((uint32_t)m->in_offset);
3165 m->lineno = swap4((uint32_t)m->lineno);
3166 if (IS_STRING(m->type)) {
3167 m->str_range = swap4(m->str_range);
3168 m->str_flags = swap4(m->str_flags);
3169 }
3170 else {
3171 m->value.q = swap8(m->value.q);
3172 m->num_mask = swap8(m->num_mask);
3173 }
3174 }
3175
3176 protected size_t
file_pstring_length_size(const struct magic * m)3177 file_pstring_length_size(const struct magic *m)
3178 {
3179 switch (m->str_flags & PSTRING_LEN) {
3180 case PSTRING_1_LE:
3181 return 1;
3182 case PSTRING_2_LE:
3183 case PSTRING_2_BE:
3184 return 2;
3185 case PSTRING_4_LE:
3186 case PSTRING_4_BE:
3187 return 4;
3188 default:
3189 abort(); /* Impossible */
3190 return 1;
3191 }
3192 }
3193 protected size_t
file_pstring_get_length(const struct magic * m,const char * s)3194 file_pstring_get_length(const struct magic *m, const char *s)
3195 {
3196 size_t len = 0;
3197
3198 switch (m->str_flags & PSTRING_LEN) {
3199 case PSTRING_1_LE:
3200 len = *s;
3201 break;
3202 case PSTRING_2_LE:
3203 len = (s[1] << 8) | s[0];
3204 break;
3205 case PSTRING_2_BE:
3206 len = (s[0] << 8) | s[1];
3207 break;
3208 case PSTRING_4_LE:
3209 len = (s[3] << 24) | (s[2] << 16) | (s[1] << 8) | s[0];
3210 break;
3211 case PSTRING_4_BE:
3212 len = (s[0] << 24) | (s[1] << 16) | (s[2] << 8) | s[3];
3213 break;
3214 default:
3215 abort(); /* Impossible */
3216 }
3217
3218 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF)
3219 len -= file_pstring_length_size(m);
3220
3221 return len;
3222 }
3223
3224 protected int
file_magicfind(struct magic_set * ms,const char * name,struct mlist * v)3225 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v)
3226 {
3227 uint32_t i, j;
3228 struct mlist *mlist, *ml;
3229
3230 mlist = ms->mlist[1];
3231
3232 for (ml = mlist->next; ml != mlist; ml = ml->next) {
3233 struct magic *ma = ml->magic;
3234 uint32_t nma = ml->nmagic;
3235 for (i = 0; i < nma; i++) {
3236 if (ma[i].type != FILE_NAME)
3237 continue;
3238 if (strcmp(ma[i].value.s, name) == 0) {
3239 v->magic = &ma[i];
3240 for (j = i + 1; j < nma; j++)
3241 if (ma[j].cont_level == 0)
3242 break;
3243 v->nmagic = j - i;
3244 return 0;
3245 }
3246 }
3247 }
3248 return -1;
3249 }
3250