1 /*
2 * Copyright (c) Ian F. Darwin 1986-1995.
3 * Software written by Ian F. Darwin and others;
4 * maintained 1995-present by Christos Zoulas and others.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice immediately at the beginning of the file, without modification,
11 * this list of conditions, and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
20 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 */
28 /*
29 * apprentice - make one pass through /etc/magic, learning its secrets.
30 */
31
32 #include "file.h"
33
34 #ifndef lint
35 FILE_RCSID("@(#)$File: apprentice.c,v 1.326 2022/09/13 18:46:07 christos Exp $")
36 #endif /* lint */
37
38 #include "magic.h"
39 #include <stdlib.h>
40 #ifdef HAVE_UNISTD_H
41 #include <unistd.h>
42 #endif
43 #include <stddef.h>
44 #include <string.h>
45 #include <assert.h>
46 #include <ctype.h>
47 #include <fcntl.h>
48 #ifdef QUICK
49 #include <sys/mman.h>
50 #endif
51 #ifdef HAVE_DIRENT_H
52 #include <dirent.h>
53 #endif
54 #include <limits.h>
55 #ifdef HAVE_BYTESWAP_H
56 #include <byteswap.h>
57 #endif
58 #ifdef HAVE_SYS_BSWAP_H
59 #include <sys/bswap.h>
60 #endif
61
62
63 #define EATAB {while (isascii(CAST(unsigned char, *l)) && \
64 isspace(CAST(unsigned char, *l))) ++l;}
65 #define LOWCASE(l) (isupper(CAST(unsigned char, l)) ? \
66 tolower(CAST(unsigned char, l)) : (l))
67 /*
68 * Work around a bug in headers on Digital Unix.
69 * At least confirmed for: OSF1 V4.0 878
70 */
71 #if defined(__osf__) && defined(__DECC)
72 #ifdef MAP_FAILED
73 #undef MAP_FAILED
74 #endif
75 #endif
76
77 #ifndef MAP_FAILED
78 #define MAP_FAILED (void *) -1
79 #endif
80
81 #ifndef MAP_FILE
82 #define MAP_FILE 0
83 #endif
84
85 #define ALLOC_CHUNK CAST(size_t, 10)
86 #define ALLOC_INCR CAST(size_t, 200)
87
88 #define MAP_TYPE_USER 0
89 #define MAP_TYPE_MALLOC 1
90 #define MAP_TYPE_MMAP 2
91
92 struct magic_entry {
93 struct magic *mp;
94 uint32_t cont_count;
95 uint32_t max_count;
96 };
97
98 struct magic_entry_set {
99 struct magic_entry *me;
100 uint32_t count;
101 uint32_t max;
102 };
103
104 struct magic_map {
105 void *p;
106 size_t len;
107 int type;
108 struct magic *magic[MAGIC_SETS];
109 uint32_t nmagic[MAGIC_SETS];
110 };
111
112 int file_formats[FILE_NAMES_SIZE];
113 const size_t file_nformats = FILE_NAMES_SIZE;
114 const char *file_names[FILE_NAMES_SIZE];
115 const size_t file_nnames = FILE_NAMES_SIZE;
116
117 private int getvalue(struct magic_set *ms, struct magic *, const char **, int);
118 private int hextoint(int);
119 private const char *getstr(struct magic_set *, struct magic *, const char *,
120 int);
121 private int parse(struct magic_set *, struct magic_entry *, const char *,
122 size_t, int);
123 private void eatsize(const char **);
124 private int apprentice_1(struct magic_set *, const char *, int);
125 private ssize_t apprentice_magic_strength_1(const struct magic *);
126 private size_t apprentice_magic_strength(const struct magic *, size_t);
127 private int apprentice_sort(const void *, const void *);
128 private void apprentice_list(struct mlist *, int );
129 private struct magic_map *apprentice_load(struct magic_set *,
130 const char *, int);
131 private struct mlist *mlist_alloc(void);
132 private void mlist_free_all(struct magic_set *);
133 private void mlist_free(struct mlist *);
134 private void byteswap(struct magic *, uint32_t);
135 private void bs1(struct magic *);
136
137 #if defined(HAVE_BYTESWAP_H)
138 #define swap2(x) bswap_16(x)
139 #define swap4(x) bswap_32(x)
140 #define swap8(x) bswap_64(x)
141 #elif defined(HAVE_SYS_BSWAP_H)
142 #define swap2(x) bswap16(x)
143 #define swap4(x) bswap32(x)
144 #define swap8(x) bswap64(x)
145 #else
146 private uint16_t swap2(uint16_t);
147 private uint32_t swap4(uint32_t);
148 private uint64_t swap8(uint64_t);
149 #endif
150
151 private char *mkdbname(struct magic_set *, const char *, int);
152 private struct magic_map *apprentice_map(struct magic_set *, const char *);
153 private void apprentice_unmap(struct magic_map *);
154 private int apprentice_compile(struct magic_set *, struct magic_map *,
155 const char *);
156 private int check_format_type(const char *, int, const char **);
157 private int check_format(struct magic_set *, struct magic *);
158 private int get_op(char);
159 private int parse_mime(struct magic_set *, struct magic_entry *, const char *,
160 size_t);
161 private int parse_strength(struct magic_set *, struct magic_entry *,
162 const char *, size_t);
163 private int parse_apple(struct magic_set *, struct magic_entry *, const char *,
164 size_t);
165 private int parse_ext(struct magic_set *, struct magic_entry *, const char *,
166 size_t);
167
168
169 private size_t magicsize = sizeof(struct magic);
170
171 private const char usg_hdr[] = "cont\toffset\ttype\topcode\tmask\tvalue\tdesc";
172
173 private struct {
174 const char *name;
175 size_t len;
176 int (*fun)(struct magic_set *, struct magic_entry *, const char *,
177 size_t);
178 } bang[] = {
179 #define DECLARE_FIELD(name) { # name, sizeof(# name) - 1, parse_ ## name }
180 DECLARE_FIELD(mime),
181 DECLARE_FIELD(apple),
182 DECLARE_FIELD(ext),
183 DECLARE_FIELD(strength),
184 #undef DECLARE_FIELD
185 { NULL, 0, NULL }
186 };
187
188 #include "../data_file.c"
189
190 #ifdef COMPILE_ONLY
191
192 int main(int, char *[]);
193
194 int
main(int argc,char * argv[])195 main(int argc, char *argv[])
196 {
197 int ret;
198 struct magic_set *ms;
199 char *progname;
200
201 if ((progname = strrchr(argv[0], '/')) != NULL)
202 progname++;
203 else
204 progname = argv[0];
205
206 if (argc != 2) {
207 (void)fprintf(stderr, "Usage: %s file\n", progname);
208 return 1;
209 }
210
211 if ((ms = magic_open(MAGIC_CHECK)) == NULL) {
212 (void)fprintf(stderr, "%s: %s\n", progname, strerror(errno));
213 return 1;
214 }
215 ret = magic_compile(ms, argv[1]) == -1 ? 1 : 0;
216 if (ret == 1)
217 (void)fprintf(stderr, "%s: %s\n", progname, magic_error(ms));
218 magic_close(ms);
219 return ret;
220 }
221 #endif /* COMPILE_ONLY */
222
223 struct type_tbl_s {
224 const char name[16];
225 const size_t len;
226 const int type;
227 const int format;
228 };
229
230 /*
231 * XXX - the actual Single UNIX Specification says that "long" means "long",
232 * as in the C data type, but we treat it as meaning "4-byte integer".
233 * Given that the OS X version of file 5.04 did the same, I guess that passes
234 * the actual test; having "long" be dependent on how big a "long" is on
235 * the machine running "file" is silly.
236 */
237 static const struct type_tbl_s type_tbl[] = {
238 # define XX(s) s, (sizeof(s) - 1)
239 # define XX_NULL "", 0
240 { XX("invalid"), FILE_INVALID, FILE_FMT_NONE },
241 { XX("byte"), FILE_BYTE, FILE_FMT_NUM },
242 { XX("short"), FILE_SHORT, FILE_FMT_NUM },
243 { XX("default"), FILE_DEFAULT, FILE_FMT_NONE },
244 { XX("long"), FILE_LONG, FILE_FMT_NUM },
245 { XX("string"), FILE_STRING, FILE_FMT_STR },
246 { XX("date"), FILE_DATE, FILE_FMT_STR },
247 { XX("beshort"), FILE_BESHORT, FILE_FMT_NUM },
248 { XX("belong"), FILE_BELONG, FILE_FMT_NUM },
249 { XX("bedate"), FILE_BEDATE, FILE_FMT_STR },
250 { XX("leshort"), FILE_LESHORT, FILE_FMT_NUM },
251 { XX("lelong"), FILE_LELONG, FILE_FMT_NUM },
252 { XX("ledate"), FILE_LEDATE, FILE_FMT_STR },
253 { XX("pstring"), FILE_PSTRING, FILE_FMT_STR },
254 { XX("ldate"), FILE_LDATE, FILE_FMT_STR },
255 { XX("beldate"), FILE_BELDATE, FILE_FMT_STR },
256 { XX("leldate"), FILE_LELDATE, FILE_FMT_STR },
257 { XX("regex"), FILE_REGEX, FILE_FMT_STR },
258 { XX("bestring16"), FILE_BESTRING16, FILE_FMT_STR },
259 { XX("lestring16"), FILE_LESTRING16, FILE_FMT_STR },
260 { XX("search"), FILE_SEARCH, FILE_FMT_STR },
261 { XX("medate"), FILE_MEDATE, FILE_FMT_STR },
262 { XX("meldate"), FILE_MELDATE, FILE_FMT_STR },
263 { XX("melong"), FILE_MELONG, FILE_FMT_NUM },
264 { XX("quad"), FILE_QUAD, FILE_FMT_QUAD },
265 { XX("lequad"), FILE_LEQUAD, FILE_FMT_QUAD },
266 { XX("bequad"), FILE_BEQUAD, FILE_FMT_QUAD },
267 { XX("qdate"), FILE_QDATE, FILE_FMT_STR },
268 { XX("leqdate"), FILE_LEQDATE, FILE_FMT_STR },
269 { XX("beqdate"), FILE_BEQDATE, FILE_FMT_STR },
270 { XX("qldate"), FILE_QLDATE, FILE_FMT_STR },
271 { XX("leqldate"), FILE_LEQLDATE, FILE_FMT_STR },
272 { XX("beqldate"), FILE_BEQLDATE, FILE_FMT_STR },
273 { XX("float"), FILE_FLOAT, FILE_FMT_FLOAT },
274 { XX("befloat"), FILE_BEFLOAT, FILE_FMT_FLOAT },
275 { XX("lefloat"), FILE_LEFLOAT, FILE_FMT_FLOAT },
276 { XX("double"), FILE_DOUBLE, FILE_FMT_DOUBLE },
277 { XX("bedouble"), FILE_BEDOUBLE, FILE_FMT_DOUBLE },
278 { XX("ledouble"), FILE_LEDOUBLE, FILE_FMT_DOUBLE },
279 { XX("leid3"), FILE_LEID3, FILE_FMT_NUM },
280 { XX("beid3"), FILE_BEID3, FILE_FMT_NUM },
281 { XX("indirect"), FILE_INDIRECT, FILE_FMT_NUM },
282 { XX("qwdate"), FILE_QWDATE, FILE_FMT_STR },
283 { XX("leqwdate"), FILE_LEQWDATE, FILE_FMT_STR },
284 { XX("beqwdate"), FILE_BEQWDATE, FILE_FMT_STR },
285 { XX("name"), FILE_NAME, FILE_FMT_NONE },
286 { XX("use"), FILE_USE, FILE_FMT_NONE },
287 { XX("clear"), FILE_CLEAR, FILE_FMT_NONE },
288 { XX("der"), FILE_DER, FILE_FMT_STR },
289 { XX("guid"), FILE_GUID, FILE_FMT_STR },
290 { XX("offset"), FILE_OFFSET, FILE_FMT_QUAD },
291 { XX("bevarint"), FILE_BEVARINT, FILE_FMT_STR },
292 { XX("levarint"), FILE_LEVARINT, FILE_FMT_STR },
293 { XX("msdosdate"), FILE_MSDOSDATE, FILE_FMT_STR },
294 { XX("lemsdosdate"), FILE_LEMSDOSDATE, FILE_FMT_STR },
295 { XX("bemsdosdate"), FILE_BEMSDOSDATE, FILE_FMT_STR },
296 { XX("msdostime"), FILE_MSDOSTIME, FILE_FMT_STR },
297 { XX("lemsdostime"), FILE_LEMSDOSTIME, FILE_FMT_STR },
298 { XX("bemsdostime"), FILE_BEMSDOSTIME, FILE_FMT_STR },
299 { XX("octal"), FILE_OCTAL, FILE_FMT_STR },
300 { XX_NULL, FILE_INVALID, FILE_FMT_NONE },
301 };
302
303 /*
304 * These are not types, and cannot be preceded by "u" to make them
305 * unsigned.
306 */
307 static const struct type_tbl_s special_tbl[] = {
308 { XX("der"), FILE_DER, FILE_FMT_STR },
309 { XX("name"), FILE_NAME, FILE_FMT_STR },
310 { XX("use"), FILE_USE, FILE_FMT_STR },
311 { XX("octal"), FILE_OCTAL, FILE_FMT_STR },
312 { XX_NULL, FILE_INVALID, FILE_FMT_NONE },
313 };
314 # undef XX
315 # undef XX_NULL
316
317 private int
get_type(const struct type_tbl_s * tbl,const char * l,const char ** t)318 get_type(const struct type_tbl_s *tbl, const char *l, const char **t)
319 {
320 const struct type_tbl_s *p;
321
322 for (p = tbl; p->len; p++) {
323 if (strncmp(l, p->name, p->len) == 0) {
324 if (t)
325 *t = l + p->len;
326 break;
327 }
328 }
329 return p->type;
330 }
331
332 private off_t
maxoff_t(void)333 maxoff_t(void) {
334 if (/*CONSTCOND*/sizeof(off_t) == sizeof(int))
335 return CAST(off_t, INT_MAX);
336 if (/*CONSTCOND*/sizeof(off_t) == sizeof(long))
337 return CAST(off_t, LONG_MAX);
338 return 0x7fffffff;
339 }
340
341 private int
get_standard_integer_type(const char * l,const char ** t)342 get_standard_integer_type(const char *l, const char **t)
343 {
344 int type;
345
346 if (isalpha(CAST(unsigned char, l[1]))) {
347 switch (l[1]) {
348 case 'C':
349 /* "dC" and "uC" */
350 type = FILE_BYTE;
351 break;
352 case 'S':
353 /* "dS" and "uS" */
354 type = FILE_SHORT;
355 break;
356 case 'I':
357 case 'L':
358 /*
359 * "dI", "dL", "uI", and "uL".
360 *
361 * XXX - the actual Single UNIX Specification says
362 * that "L" means "long", as in the C data type,
363 * but we treat it as meaning "4-byte integer".
364 * Given that the OS X version of file 5.04 did
365 * the same, I guess that passes the actual SUS
366 * validation suite; having "dL" be dependent on
367 * how big a "long" is on the machine running
368 * "file" is silly.
369 */
370 type = FILE_LONG;
371 break;
372 case 'Q':
373 /* "dQ" and "uQ" */
374 type = FILE_QUAD;
375 break;
376 default:
377 /* "d{anything else}", "u{anything else}" */
378 return FILE_INVALID;
379 }
380 l += 2;
381 } else if (isdigit(CAST(unsigned char, l[1]))) {
382 /*
383 * "d{num}" and "u{num}"; we only support {num} values
384 * of 1, 2, 4, and 8 - the Single UNIX Specification
385 * doesn't say anything about whether arbitrary
386 * values should be supported, but both the Solaris 10
387 * and OS X Mountain Lion versions of file passed the
388 * Single UNIX Specification validation suite, and
389 * neither of them support values bigger than 8 or
390 * non-power-of-2 values.
391 */
392 if (isdigit(CAST(unsigned char, l[2]))) {
393 /* Multi-digit, so > 9 */
394 return FILE_INVALID;
395 }
396 switch (l[1]) {
397 case '1':
398 type = FILE_BYTE;
399 break;
400 case '2':
401 type = FILE_SHORT;
402 break;
403 case '4':
404 type = FILE_LONG;
405 break;
406 case '8':
407 type = FILE_QUAD;
408 break;
409 default:
410 /* XXX - what about 3, 5, 6, or 7? */
411 return FILE_INVALID;
412 }
413 l += 2;
414 } else {
415 /*
416 * "d" or "u" by itself.
417 */
418 type = FILE_LONG;
419 ++l;
420 }
421 if (t)
422 *t = l;
423 return type;
424 }
425
426 private void
init_file_tables(void)427 init_file_tables(void)
428 {
429 static int done = 0;
430 const struct type_tbl_s *p;
431
432 if (done)
433 return;
434 done++;
435
436 for (p = type_tbl; p->len; p++) {
437 assert(p->type < FILE_NAMES_SIZE);
438 file_names[p->type] = p->name;
439 file_formats[p->type] = p->format;
440 }
441 assert(p - type_tbl == FILE_NAMES_SIZE);
442 }
443
444 private int
add_mlist(struct mlist * mlp,struct magic_map * map,size_t idx)445 add_mlist(struct mlist *mlp, struct magic_map *map, size_t idx)
446 {
447 struct mlist *ml;
448
449 mlp->map = NULL;
450 if ((ml = CAST(struct mlist *, emalloc(sizeof(*ml)))) == NULL)
451 return -1;
452
453 ml->map = idx == 0 ? map : NULL;
454 ml->magic = map->magic[idx];
455 ml->nmagic = map->nmagic[idx];
456
457 mlp->prev->next = ml;
458 ml->prev = mlp->prev;
459 ml->next = mlp;
460 mlp->prev = ml;
461 return 0;
462 }
463
464 /*
465 * Handle one file or directory.
466 */
467 private int
apprentice_1(struct magic_set * ms,const char * fn,int action)468 apprentice_1(struct magic_set *ms, const char *fn, int action)
469 {
470 struct magic_map *map;
471 #ifndef COMPILE_ONLY
472 struct mlist *ml;
473 size_t i;
474 #endif
475
476 if (magicsize != FILE_MAGICSIZE) {
477 file_error(ms, 0, "magic element size %lu != %lu",
478 CAST(unsigned long, sizeof(*map->magic[0])),
479 CAST(unsigned long, FILE_MAGICSIZE));
480 return -1;
481 }
482
483 if (action == FILE_COMPILE) {
484 map = apprentice_load(ms, fn, action);
485 if (map == NULL)
486 return -1;
487 return apprentice_compile(ms, map, fn);
488 }
489
490 #ifndef COMPILE_ONLY
491 map = apprentice_map(ms, fn);
492 if (map == NULL) {
493 if (ms->flags & MAGIC_CHECK)
494 file_magwarn(ms, "using regular magic file `%s'", fn);
495 map = apprentice_load(ms, fn, action);
496 if (map == NULL)
497 return -1;
498 }
499
500 for (i = 0; i < MAGIC_SETS; i++) {
501 if (add_mlist(ms->mlist[i], map, i) == -1) {
502 /* failed to add to any list, free explicitly */
503 if (i == 0)
504 apprentice_unmap(map);
505 else
506 mlist_free_all(ms);
507 file_oomem(ms, sizeof(*ml));
508 return -1;
509 }
510 }
511
512 if (action == FILE_LIST) {
513 for (i = 0; i < MAGIC_SETS; i++) {
514 printf("Set %" SIZE_T_FORMAT "u:\nBinary patterns:\n",
515 i);
516 apprentice_list(ms->mlist[i], BINTEST);
517 printf("Text patterns:\n");
518 apprentice_list(ms->mlist[i], TEXTTEST);
519 }
520 }
521 return 0;
522 #else
523 return 0;
524 #endif /* COMPILE_ONLY */
525 }
526
527 protected void
file_ms_free(struct magic_set * ms)528 file_ms_free(struct magic_set *ms)
529 {
530 size_t i;
531 if (ms == NULL)
532 return;
533 for (i = 0; i < MAGIC_SETS; i++)
534 mlist_free(ms->mlist[i]);
535 if (ms->o.pbuf) {
536 efree(ms->o.pbuf);
537 }
538 if (ms->o.buf) {
539 efree(ms->o.buf);
540 }
541 if (ms->c.li) {
542 efree(ms->c.li);
543 }
544 #ifdef USE_C_LOCALE
545 freelocale(ms->c_lc_ctype);
546 #endif
547 efree(ms);
548 }
549
550 protected struct magic_set *
file_ms_alloc(int flags)551 file_ms_alloc(int flags)
552 {
553 struct magic_set *ms;
554 size_t i, len;
555
556 if ((ms = CAST(struct magic_set *, ecalloc(CAST(size_t, 1u),
557 sizeof(struct magic_set)))) == NULL)
558 return NULL;
559
560 if (magic_setflags(ms, flags) == -1) {
561 errno = EINVAL;
562 goto free;
563 }
564
565 ms->o.buf = ms->o.pbuf = NULL;
566 ms->o.blen = 0;
567 len = (ms->c.len = 10) * sizeof(*ms->c.li);
568
569 if ((ms->c.li = CAST(struct level_info *, emalloc(len))) == NULL)
570 goto free;
571
572 ms->event_flags = 0;
573 ms->error = -1;
574 for (i = 0; i < MAGIC_SETS; i++)
575 ms->mlist[i] = NULL;
576 ms->file = "unknown";
577 ms->line = 0;
578 ms->indir_max = FILE_INDIR_MAX;
579 ms->name_max = FILE_NAME_MAX;
580 ms->elf_shnum_max = FILE_ELF_SHNUM_MAX;
581 ms->elf_phnum_max = FILE_ELF_PHNUM_MAX;
582 ms->elf_notes_max = FILE_ELF_NOTES_MAX;
583 ms->regex_max = FILE_REGEX_MAX;
584 ms->bytes_max = FILE_BYTES_MAX;
585 ms->encoding_max = FILE_ENCODING_MAX;
586 #ifdef USE_C_LOCALE
587 ms->c_lc_ctype = newlocale(LC_CTYPE_MASK, "C", 0);
588 assert(ms->c_lc_ctype != NULL);
589 #endif
590 return ms;
591 free:
592 free(ms);
593 return NULL;
594 }
595
596 private void
apprentice_unmap(struct magic_map * map)597 apprentice_unmap(struct magic_map *map)
598 {
599 if (map == NULL)
600 return;
601 if (map->p != php_magic_database) {
602 if (map->p == NULL) {
603 int j;
604 for (j = 0; j < MAGIC_SETS; j++) {
605 if (map->magic[j]) {
606 efree(map->magic[j]);
607 }
608 }
609 } else {
610 efree(map->p);
611 }
612 }
613 efree(map);
614 }
615
616 private struct mlist *
mlist_alloc(void)617 mlist_alloc(void)
618 {
619 struct mlist *mlist;
620 if ((mlist = CAST(struct mlist *, ecalloc(1, sizeof(*mlist)))) == NULL) {
621 return NULL;
622 }
623 mlist->next = mlist->prev = mlist;
624 return mlist;
625 }
626
627 private void
mlist_free_all(struct magic_set * ms)628 mlist_free_all(struct magic_set *ms)
629 {
630 size_t i;
631
632 for (i = 0; i < MAGIC_SETS; i++) {
633 mlist_free(ms->mlist[i]);
634 ms->mlist[i] = NULL;
635 }
636 }
637
638 private void
mlist_free_one(struct mlist * ml)639 mlist_free_one(struct mlist *ml)
640 {
641 if (ml->map)
642 apprentice_unmap(CAST(struct magic_map *, ml->map));
643 efree(ml);
644 }
645
646 private void
mlist_free(struct mlist * mlist)647 mlist_free(struct mlist *mlist)
648 {
649 struct mlist *ml, *next;
650
651 if (mlist == NULL)
652 return;
653
654 for (ml = mlist->next; ml != mlist;) {
655 next = ml->next;
656 mlist_free_one(ml);
657 ml = next;
658 }
659 mlist_free_one(mlist);
660 }
661
662 /* const char *fn: list of magic files and directories */
663 protected int
file_apprentice(struct magic_set * ms,const char * fn,int action)664 file_apprentice(struct magic_set *ms, const char *fn, int action)
665 {
666 char *p, *mfn;
667 int fileerr, errs = -1;
668 size_t i, j;
669
670 (void)file_reset(ms, 0);
671
672 /* XXX disabling default magic loading so the compiled in data is used */
673 #if 0
674 if ((fn = magic_getpath(fn, action)) == NULL)
675 return -1;
676 #endif
677
678 init_file_tables();
679
680 if (fn == NULL)
681 fn = getenv("MAGIC");
682 if (fn == NULL) {
683 for (i = 0; i < MAGIC_SETS; i++) {
684 mlist_free(ms->mlist[i]);
685 if ((ms->mlist[i] = mlist_alloc()) == NULL) {
686 file_oomem(ms, sizeof(*ms->mlist[i]));
687 return -1;
688 }
689 }
690 return apprentice_1(ms, fn, action);
691 }
692
693 if ((mfn = estrdup(fn)) == NULL) {
694 file_oomem(ms, strlen(fn));
695 return -1;
696 }
697
698 for (i = 0; i < MAGIC_SETS; i++) {
699 mlist_free(ms->mlist[i]);
700 if ((ms->mlist[i] = mlist_alloc()) == NULL) {
701 file_oomem(ms, sizeof(*ms->mlist[i]));
702 for (j = 0; j < i; j++) {
703 mlist_free(ms->mlist[j]);
704 ms->mlist[j] = NULL;
705 }
706 efree(mfn);
707 return -1;
708 }
709 }
710 fn = mfn;
711
712 while (fn) {
713 p = CCAST(char *, strchr(fn, PATHSEP));
714 if (p)
715 *p++ = '\0';
716 if (*fn == '\0')
717 break;
718 fileerr = apprentice_1(ms, fn, action);
719 errs = MAX(errs, fileerr);
720 fn = p;
721 }
722
723 efree(mfn);
724
725 if (errs == -1) {
726 for (i = 0; i < MAGIC_SETS; i++) {
727 mlist_free(ms->mlist[i]);
728 ms->mlist[i] = NULL;
729 }
730 file_error(ms, 0, "could not find any valid magic files!");
731 return -1;
732 }
733
734 #if 0
735 /*
736 * Always leave the database loaded
737 */
738 if (action == FILE_LOAD)
739 return 0;
740
741 for (i = 0; i < MAGIC_SETS; i++) {
742 mlist_free(ms->mlist[i]);
743 ms->mlist[i] = NULL;
744 }
745 #endif
746
747 switch (action) {
748 case FILE_LOAD:
749 case FILE_COMPILE:
750 case FILE_CHECK:
751 case FILE_LIST:
752 return 0;
753 default:
754 file_error(ms, 0, "Invalid action %d", action);
755 return -1;
756 }
757 }
758
759 /*
760 * Compute the real length of a magic expression, for the purposes
761 * of determining how "strong" a magic expression is (approximating
762 * how specific its matches are):
763 * - magic characters count 0 unless escaped.
764 * - [] expressions count 1
765 * - {} expressions count 0
766 * - regular characters or escaped magic characters count 1
767 * - 0 length expressions count as one
768 */
769 private size_t
nonmagic(const char * str)770 nonmagic(const char *str)
771 {
772 const char *p;
773 size_t rv = 0;
774
775 for (p = str; *p; p++)
776 switch (*p) {
777 case '\\': /* Escaped anything counts 1 */
778 if (!*++p)
779 p--;
780 rv++;
781 continue;
782 case '?': /* Magic characters count 0 */
783 case '*':
784 case '.':
785 case '+':
786 case '^':
787 case '$':
788 continue;
789 case '[': /* Bracketed expressions count 1 the ']' */
790 while (*p && *p != ']')
791 p++;
792 p--;
793 continue;
794 case '{': /* Braced expressions count 0 */
795 while (*p && *p != '}')
796 p++;
797 if (!*p)
798 p--;
799 continue;
800 default: /* Anything else counts 1 */
801 rv++;
802 continue;
803 }
804
805 return rv == 0 ? 1 : rv; /* Return at least 1 */
806 }
807
808
809 private size_t
typesize(int type)810 typesize(int type)
811 {
812 switch (type) {
813 case FILE_BYTE:
814 return 1;
815
816 case FILE_SHORT:
817 case FILE_LESHORT:
818 case FILE_BESHORT:
819 case FILE_MSDOSDATE:
820 case FILE_BEMSDOSDATE:
821 case FILE_LEMSDOSDATE:
822 case FILE_MSDOSTIME:
823 case FILE_BEMSDOSTIME:
824 case FILE_LEMSDOSTIME:
825 return 2;
826
827 case FILE_LONG:
828 case FILE_LELONG:
829 case FILE_BELONG:
830 case FILE_MELONG:
831 return 4;
832
833 case FILE_DATE:
834 case FILE_LEDATE:
835 case FILE_BEDATE:
836 case FILE_MEDATE:
837 case FILE_LDATE:
838 case FILE_LELDATE:
839 case FILE_BELDATE:
840 case FILE_MELDATE:
841 case FILE_FLOAT:
842 case FILE_BEFLOAT:
843 case FILE_LEFLOAT:
844 case FILE_BEID3:
845 case FILE_LEID3:
846 return 4;
847
848 case FILE_QUAD:
849 case FILE_BEQUAD:
850 case FILE_LEQUAD:
851 case FILE_QDATE:
852 case FILE_LEQDATE:
853 case FILE_BEQDATE:
854 case FILE_QLDATE:
855 case FILE_LEQLDATE:
856 case FILE_BEQLDATE:
857 case FILE_QWDATE:
858 case FILE_LEQWDATE:
859 case FILE_BEQWDATE:
860 case FILE_DOUBLE:
861 case FILE_BEDOUBLE:
862 case FILE_LEDOUBLE:
863 case FILE_OFFSET:
864 case FILE_BEVARINT:
865 case FILE_LEVARINT:
866 return 8;
867
868 case FILE_GUID:
869 return 16;
870
871 default:
872 return FILE_BADSIZE;
873 }
874 }
875
876 /*
877 * Get weight of this magic entry, for sorting purposes.
878 */
879 private ssize_t
apprentice_magic_strength_1(const struct magic * m)880 apprentice_magic_strength_1(const struct magic *m)
881 {
882 #define MULT 10U
883 size_t ts, v;
884 ssize_t val = 2 * MULT; /* baseline strength */
885
886 switch (m->type) {
887 case FILE_DEFAULT: /* make sure this sorts last */
888 if (m->factor_op != FILE_FACTOR_OP_NONE) {
889 fprintf(stderr, "Bad factor_op %d", m->factor_op);
890 abort();
891 }
892 return 0;
893
894 case FILE_BYTE:
895 case FILE_SHORT:
896 case FILE_LESHORT:
897 case FILE_BESHORT:
898 case FILE_LONG:
899 case FILE_LELONG:
900 case FILE_BELONG:
901 case FILE_MELONG:
902 case FILE_DATE:
903 case FILE_LEDATE:
904 case FILE_BEDATE:
905 case FILE_MEDATE:
906 case FILE_LDATE:
907 case FILE_LELDATE:
908 case FILE_BELDATE:
909 case FILE_MELDATE:
910 case FILE_FLOAT:
911 case FILE_BEFLOAT:
912 case FILE_LEFLOAT:
913 case FILE_QUAD:
914 case FILE_BEQUAD:
915 case FILE_LEQUAD:
916 case FILE_QDATE:
917 case FILE_LEQDATE:
918 case FILE_BEQDATE:
919 case FILE_QLDATE:
920 case FILE_LEQLDATE:
921 case FILE_BEQLDATE:
922 case FILE_QWDATE:
923 case FILE_LEQWDATE:
924 case FILE_BEQWDATE:
925 case FILE_DOUBLE:
926 case FILE_BEDOUBLE:
927 case FILE_LEDOUBLE:
928 case FILE_BEVARINT:
929 case FILE_LEVARINT:
930 case FILE_GUID:
931 case FILE_BEID3:
932 case FILE_LEID3:
933 case FILE_OFFSET:
934 case FILE_MSDOSDATE:
935 case FILE_BEMSDOSDATE:
936 case FILE_LEMSDOSDATE:
937 case FILE_MSDOSTIME:
938 case FILE_BEMSDOSTIME:
939 case FILE_LEMSDOSTIME:
940 ts = typesize(m->type);
941 if (ts == FILE_BADSIZE) {
942 (void)fprintf(stderr, "Bad size for type %d\n",
943 m->type);
944 abort();
945 }
946 val += ts * MULT;
947 break;
948
949 case FILE_PSTRING:
950 case FILE_STRING:
951 case FILE_OCTAL:
952 val += m->vallen * MULT;
953 break;
954
955 case FILE_BESTRING16:
956 case FILE_LESTRING16:
957 val += m->vallen * MULT / 2;
958 break;
959
960 case FILE_SEARCH:
961 if (m->vallen == 0)
962 break;
963 val += m->vallen * MAX(MULT / m->vallen, 1);
964 break;
965
966 case FILE_REGEX:
967 v = nonmagic(m->value.s);
968 val += v * MAX(MULT / v, 1);
969 break;
970
971 case FILE_INDIRECT:
972 case FILE_NAME:
973 case FILE_USE:
974 case FILE_CLEAR:
975 break;
976
977 case FILE_DER:
978 val += MULT;
979 break;
980
981 default:
982 (void)fprintf(stderr, "Bad type %d\n", m->type);
983 abort();
984 }
985
986 switch (m->reln) {
987 case 'x': /* matches anything penalize */
988 case '!': /* matches almost anything penalize */
989 val = 0;
990 break;
991
992 case '=': /* Exact match, prefer */
993 val += MULT;
994 break;
995
996 case '>':
997 case '<': /* comparison match reduce strength */
998 val -= 2 * MULT;
999 break;
1000
1001 case '^':
1002 case '&': /* masking bits, we could count them too */
1003 val -= MULT;
1004 break;
1005
1006 default:
1007 (void)fprintf(stderr, "Bad relation %c\n", m->reln);
1008 abort();
1009 }
1010
1011 return val;
1012 }
1013
1014
1015 /*ARGSUSED*/
1016 private size_t
apprentice_magic_strength(const struct magic * m,size_t nmagic)1017 apprentice_magic_strength(const struct magic *m,
1018 size_t nmagic __attribute__((__unused__)))
1019 {
1020 ssize_t val = apprentice_magic_strength_1(m);
1021
1022 #ifdef notyet
1023 if (m->desc[0] == '\0') {
1024 size_t i;
1025 /*
1026 * Magic entries with no description get their continuations
1027 * added
1028 */
1029 for (i = 1; m[i].cont_level != 0 && i < MIN(nmagic, 3); i++) {
1030 ssize_t v = apprentice_magic_strength_1(&m[i]) >>
1031 (i + 1);
1032 val += v;
1033 if (m[i].desc[0] != '\0')
1034 break;
1035 }
1036 }
1037 #endif
1038
1039 switch (m->factor_op) {
1040 case FILE_FACTOR_OP_NONE:
1041 break;
1042 case FILE_FACTOR_OP_PLUS:
1043 val += m->factor;
1044 break;
1045 case FILE_FACTOR_OP_MINUS:
1046 val -= m->factor;
1047 break;
1048 case FILE_FACTOR_OP_TIMES:
1049 val *= m->factor;
1050 break;
1051 case FILE_FACTOR_OP_DIV:
1052 val /= m->factor;
1053 break;
1054 default:
1055 (void)fprintf(stderr, "Bad factor_op %u\n", m->factor_op);
1056 abort();
1057 }
1058
1059 if (val <= 0) /* ensure we only return 0 for FILE_DEFAULT */
1060 val = 1;
1061
1062 #ifndef notyet
1063 /*
1064 * Magic entries with no description get a bonus because they depend
1065 * on subsequent magic entries to print something.
1066 */
1067 if (m->desc[0] == '\0')
1068 val++;
1069 #endif
1070
1071 return val;
1072 }
1073
1074 /*
1075 * Sort callback for sorting entries by "strength" (basically length)
1076 */
1077 private int
apprentice_sort(const void * a,const void * b)1078 apprentice_sort(const void *a, const void *b)
1079 {
1080 const struct magic_entry *ma = CAST(const struct magic_entry *, a);
1081 const struct magic_entry *mb = CAST(const struct magic_entry *, b);
1082 size_t sa = apprentice_magic_strength(ma->mp, ma->cont_count);
1083 size_t sb = apprentice_magic_strength(mb->mp, mb->cont_count);
1084 if (sa == sb)
1085 return 0;
1086 else if (sa > sb)
1087 return -1;
1088 else
1089 return 1;
1090 }
1091
1092 /*
1093 * Shows sorted patterns list in the order which is used for the matching
1094 */
1095 private void
apprentice_list(struct mlist * mlist,int mode)1096 apprentice_list(struct mlist *mlist, int mode)
1097 {
1098 uint32_t magindex, descindex, mimeindex, lineindex;
1099 struct mlist *ml;
1100 for (ml = mlist->next; ml != mlist; ml = ml->next) {
1101 for (magindex = 0; magindex < ml->nmagic; magindex++) {
1102 struct magic *m = &ml->magic[magindex];
1103 if ((m->flag & mode) != mode) {
1104 /* Skip sub-tests */
1105 while (magindex + 1 < ml->nmagic &&
1106 ml->magic[magindex + 1].cont_level != 0)
1107 ++magindex;
1108 continue; /* Skip to next top-level test*/
1109 }
1110
1111 /*
1112 * Try to iterate over the tree until we find item with
1113 * description/mimetype.
1114 */
1115 lineindex = descindex = mimeindex = magindex;
1116 for (magindex++; magindex < ml->nmagic &&
1117 ml->magic[magindex].cont_level != 0; magindex++) {
1118 if (*ml->magic[descindex].desc == '\0'
1119 && *ml->magic[magindex].desc)
1120 descindex = magindex;
1121 if (*ml->magic[mimeindex].mimetype == '\0'
1122 && *ml->magic[magindex].mimetype)
1123 mimeindex = magindex;
1124 }
1125
1126 printf("Strength = %3" SIZE_T_FORMAT "u@%u: %s [%s]\n",
1127 apprentice_magic_strength(m, ml->nmagic - magindex),
1128 ml->magic[lineindex].lineno,
1129 ml->magic[descindex].desc,
1130 ml->magic[mimeindex].mimetype);
1131 }
1132 }
1133 }
1134
1135 private void
set_test_type(struct magic * mstart,struct magic * m)1136 set_test_type(struct magic *mstart, struct magic *m)
1137 {
1138 switch (m->type) {
1139 case FILE_BYTE:
1140 case FILE_SHORT:
1141 case FILE_LONG:
1142 case FILE_DATE:
1143 case FILE_BESHORT:
1144 case FILE_BELONG:
1145 case FILE_BEDATE:
1146 case FILE_LESHORT:
1147 case FILE_LELONG:
1148 case FILE_LEDATE:
1149 case FILE_LDATE:
1150 case FILE_BELDATE:
1151 case FILE_LELDATE:
1152 case FILE_MEDATE:
1153 case FILE_MELDATE:
1154 case FILE_MELONG:
1155 case FILE_QUAD:
1156 case FILE_LEQUAD:
1157 case FILE_BEQUAD:
1158 case FILE_QDATE:
1159 case FILE_LEQDATE:
1160 case FILE_BEQDATE:
1161 case FILE_QLDATE:
1162 case FILE_LEQLDATE:
1163 case FILE_BEQLDATE:
1164 case FILE_QWDATE:
1165 case FILE_LEQWDATE:
1166 case FILE_BEQWDATE:
1167 case FILE_FLOAT:
1168 case FILE_BEFLOAT:
1169 case FILE_LEFLOAT:
1170 case FILE_DOUBLE:
1171 case FILE_BEDOUBLE:
1172 case FILE_LEDOUBLE:
1173 case FILE_BEVARINT:
1174 case FILE_LEVARINT:
1175 case FILE_DER:
1176 case FILE_GUID:
1177 case FILE_OFFSET:
1178 case FILE_MSDOSDATE:
1179 case FILE_BEMSDOSDATE:
1180 case FILE_LEMSDOSDATE:
1181 case FILE_MSDOSTIME:
1182 case FILE_BEMSDOSTIME:
1183 case FILE_LEMSDOSTIME:
1184 case FILE_OCTAL:
1185 mstart->flag |= BINTEST;
1186 break;
1187 case FILE_STRING:
1188 case FILE_PSTRING:
1189 case FILE_BESTRING16:
1190 case FILE_LESTRING16:
1191 /* Allow text overrides */
1192 if (mstart->str_flags & STRING_TEXTTEST)
1193 mstart->flag |= TEXTTEST;
1194 else
1195 mstart->flag |= BINTEST;
1196 break;
1197 case FILE_REGEX:
1198 case FILE_SEARCH:
1199 /* Check for override */
1200 if (mstart->str_flags & STRING_BINTEST)
1201 mstart->flag |= BINTEST;
1202 if (mstart->str_flags & STRING_TEXTTEST)
1203 mstart->flag |= TEXTTEST;
1204
1205 if (mstart->flag & (TEXTTEST|BINTEST))
1206 break;
1207
1208 /* binary test if pattern is not text */
1209 if (file_looks_utf8(m->value.us, CAST(size_t, m->vallen), NULL,
1210 NULL) <= 0)
1211 mstart->flag |= BINTEST;
1212 else
1213 mstart->flag |= TEXTTEST;
1214 break;
1215 case FILE_DEFAULT:
1216 /* can't deduce anything; we shouldn't see this at the
1217 top level anyway */
1218 break;
1219 case FILE_INVALID:
1220 default:
1221 /* invalid search type, but no need to complain here */
1222 break;
1223 }
1224 }
1225
1226 private int
addentry(struct magic_set * ms,struct magic_entry * me,struct magic_entry_set * mset)1227 addentry(struct magic_set *ms, struct magic_entry *me,
1228 struct magic_entry_set *mset)
1229 {
1230 size_t i = me->mp->type == FILE_NAME ? 1 : 0;
1231 if (mset[i].me == NULL || mset[i].count == mset[i].max) {
1232 struct magic_entry *mp;
1233
1234 size_t incr = mset[i].max + ALLOC_INCR;
1235 if ((mp = CAST(struct magic_entry *,
1236 erealloc(mset[i].me, sizeof(*mp) * incr))) ==
1237 NULL) {
1238 file_oomem(ms, sizeof(*mp) * incr);
1239 return -1;
1240 }
1241 (void)memset(&mp[mset[i].count], 0, sizeof(*mp) *
1242 ALLOC_INCR);
1243 mset[i].me = mp;
1244 mset[i].max = CAST(uint32_t, incr);
1245 assert(mset[i].max == incr);
1246 }
1247 mset[i].me[mset[i].count++] = *me;
1248 memset(me, 0, sizeof(*me));
1249 return 0;
1250 }
1251
1252 /*
1253 * Load and parse one file.
1254 */
1255 private void
load_1(struct magic_set * ms,int action,const char * fn,int * errs,struct magic_entry_set * mset)1256 load_1(struct magic_set *ms, int action, const char *fn, int *errs,
1257 struct magic_entry_set *mset)
1258 {
1259 char buffer[BUFSIZ + 1];
1260 char *line = NULL;
1261 size_t len;
1262 size_t lineno = 0;
1263 struct magic_entry me;
1264
1265 php_stream *stream;
1266
1267
1268 ms->file = fn;
1269 stream = php_stream_open_wrapper((char *)fn, "rb", REPORT_ERRORS, NULL);
1270
1271 if (stream == NULL) {
1272 if (errno != ENOENT)
1273 file_error(ms, errno, "cannot read magic file `%s'",
1274 fn);
1275 (*errs)++;
1276 return;
1277 }
1278
1279 memset(&me, 0, sizeof(me));
1280 /* read and parse this file */
1281 for (ms->line = 1; (line = php_stream_get_line(stream, buffer , BUFSIZ, &len)) != NULL; ms->line++) {
1282 if (len == 0) /* null line, garbage, etc */
1283 continue;
1284 if (line[len - 1] == '\n') {
1285 lineno++;
1286 line[len - 1] = '\0'; /* delete newline */
1287 }
1288 switch (line[0]) {
1289 case '\0': /* empty, do not parse */
1290 case '#': /* comment, do not parse */
1291 continue;
1292 case '!':
1293 if (line[1] == ':') {
1294 size_t i;
1295
1296 for (i = 0; bang[i].name != NULL; i++) {
1297 if (CAST(size_t, len - 2) > bang[i].len &&
1298 memcmp(bang[i].name, line + 2,
1299 bang[i].len) == 0)
1300 break;
1301 }
1302 if (bang[i].name == NULL) {
1303 file_error(ms, 0,
1304 "Unknown !: entry `%s'", line);
1305 (*errs)++;
1306 continue;
1307 }
1308 if (me.mp == NULL) {
1309 file_error(ms, 0,
1310 "No current entry for :!%s type",
1311 bang[i].name);
1312 (*errs)++;
1313 continue;
1314 }
1315 if ((*bang[i].fun)(ms, &me,
1316 line + bang[i].len + 2,
1317 len - bang[i].len - 2) != 0) {
1318 (*errs)++;
1319 continue;
1320 }
1321 continue;
1322 }
1323 /*FALLTHROUGH*/
1324 default:
1325 again:
1326 switch (parse(ms, &me, line, lineno, action)) {
1327 case 0:
1328 continue;
1329 case 1:
1330 (void)addentry(ms, &me, mset);
1331 goto again;
1332 default:
1333 (*errs)++;
1334 break;
1335 }
1336 }
1337 }
1338 if (me.mp)
1339 (void)addentry(ms, &me, mset);
1340 efree(line);
1341 php_stream_close(stream);
1342 }
1343
1344 /*
1345 * parse a file or directory of files
1346 * const char *fn: name of magic file or directory
1347 */
1348 private int
cmpstrp(const void * p1,const void * p2)1349 cmpstrp(const void *p1, const void *p2)
1350 {
1351 return strcmp(*RCAST(char *const *, p1), *RCAST(char *const *, p2));
1352 }
1353
1354
1355 private uint32_t
set_text_binary(struct magic_set * ms,struct magic_entry * me,uint32_t nme,uint32_t starttest)1356 set_text_binary(struct magic_set *ms, struct magic_entry *me, uint32_t nme,
1357 uint32_t starttest)
1358 {
1359 static const char text[] = "text";
1360 static const char binary[] = "binary";
1361 static const size_t len = sizeof(text);
1362
1363 uint32_t i = starttest;
1364
1365 do {
1366 set_test_type(me[starttest].mp, me[i].mp);
1367 if ((ms->flags & MAGIC_DEBUG) == 0)
1368 continue;
1369 (void)fprintf(stderr, "%s%s%s: %s\n",
1370 me[i].mp->mimetype,
1371 me[i].mp->mimetype[0] == '\0' ? "" : "; ",
1372 me[i].mp->desc[0] ? me[i].mp->desc : "(no description)",
1373 me[i].mp->flag & BINTEST ? binary : text);
1374 if (me[i].mp->flag & BINTEST) {
1375 char *p = strstr(me[i].mp->desc, text);
1376 if (p && (p == me[i].mp->desc ||
1377 isspace(CAST(unsigned char, p[-1]))) &&
1378 (p + len - me[i].mp->desc == MAXstring
1379 || (p[len] == '\0' ||
1380 isspace(CAST(unsigned char, p[len])))))
1381 (void)fprintf(stderr, "*** Possible "
1382 "binary test for text type\n");
1383 }
1384 } while (++i < nme && me[i].mp->cont_level != 0);
1385 return i;
1386 }
1387
1388 private void
set_last_default(struct magic_set * ms,struct magic_entry * me,uint32_t nme)1389 set_last_default(struct magic_set *ms, struct magic_entry *me, uint32_t nme)
1390 {
1391 uint32_t i;
1392 for (i = 0; i < nme; i++) {
1393 if (me[i].mp->cont_level == 0 &&
1394 me[i].mp->type == FILE_DEFAULT) {
1395 while (++i < nme)
1396 if (me[i].mp->cont_level == 0)
1397 break;
1398 if (i != nme) {
1399 /* XXX - Ugh! */
1400 ms->line = me[i].mp->lineno;
1401 file_magwarn(ms,
1402 "level 0 \"default\" did not sort last");
1403 }
1404 return;
1405 }
1406 }
1407 }
1408
1409 private int
coalesce_entries(struct magic_set * ms,struct magic_entry * me,uint32_t nme,struct magic ** ma,uint32_t * nma)1410 coalesce_entries(struct magic_set *ms, struct magic_entry *me, uint32_t nme,
1411 struct magic **ma, uint32_t *nma)
1412 {
1413 uint32_t i, mentrycount = 0;
1414 size_t slen;
1415
1416 for (i = 0; i < nme; i++)
1417 mentrycount += me[i].cont_count;
1418
1419 if (mentrycount == 0) {
1420 *ma = NULL;
1421 *nma = 0;
1422 return 0;
1423 }
1424
1425 slen = sizeof(**ma) * mentrycount;
1426 if ((*ma = CAST(struct magic *, emalloc(slen))) == NULL) {
1427 file_oomem(ms, slen);
1428 return -1;
1429 }
1430
1431 mentrycount = 0;
1432 for (i = 0; i < nme; i++) {
1433 (void)memcpy(*ma + mentrycount, me[i].mp,
1434 me[i].cont_count * sizeof(**ma));
1435 mentrycount += me[i].cont_count;
1436 }
1437 *nma = mentrycount;
1438 return 0;
1439 }
1440
1441 private void
magic_entry_free(struct magic_entry * me,uint32_t nme)1442 magic_entry_free(struct magic_entry *me, uint32_t nme)
1443 {
1444 uint32_t i;
1445 if (me == NULL)
1446 return;
1447 for (i = 0; i < nme; i++)
1448 efree(me[i].mp);
1449 efree(me);
1450 }
1451
1452 private struct magic_map *
apprentice_load(struct magic_set * ms,const char * fn,int action)1453 apprentice_load(struct magic_set *ms, const char *fn, int action)
1454 {
1455 int errs = 0;
1456 uint32_t i, j;
1457 size_t files = 0, maxfiles = 0;
1458 char **filearr = NULL;
1459 zend_stat_t st = {0};
1460 struct magic_map *map;
1461 struct magic_entry_set mset[MAGIC_SETS];
1462 php_stream *dir;
1463 php_stream_dirent d;
1464
1465
1466 memset(mset, 0, sizeof(mset));
1467 ms->flags |= MAGIC_CHECK; /* Enable checks for parsed files */
1468
1469
1470 if ((map = CAST(struct magic_map *, ecalloc(1, sizeof(*map)))) == NULL)
1471 {
1472 file_oomem(ms, sizeof(*map));
1473 return NULL;
1474 }
1475 map->type = MAP_TYPE_MALLOC;
1476
1477 /* print silly verbose header for USG compat. */
1478 if (action == FILE_CHECK)
1479 (void)fprintf(stderr, "%s\n", usg_hdr);
1480
1481 /* load directory or file */
1482 /* FIXME: Read file names and sort them to prevent
1483 non-determinism. See Debian bug #488562. */
1484 if (php_sys_stat(fn, &st) == 0 && S_ISDIR(st.st_mode)) {
1485 int mflen;
1486 char mfn[MAXPATHLEN];
1487
1488 dir = php_stream_opendir((char *)fn, REPORT_ERRORS, NULL);
1489 if (!dir) {
1490 errs++;
1491 goto out;
1492 }
1493 while (php_stream_readdir(dir, &d)) {
1494 if ((mflen = snprintf(mfn, sizeof(mfn), "%s/%s", fn, d.d_name)) < 0) {
1495 file_oomem(ms,
1496 strlen(fn) + strlen(d.d_name) + 2);
1497 errs++;
1498 php_stream_closedir(dir);
1499 goto out;
1500 }
1501 if (zend_stat(mfn, &st) == -1 || !S_ISREG(st.st_mode)) {
1502 continue;
1503 }
1504 if (files >= maxfiles) {
1505 size_t mlen;
1506 maxfiles = (maxfiles + 1) * 2;
1507 mlen = maxfiles * sizeof(*filearr);
1508 if ((filearr = CAST(char **,
1509 erealloc(filearr, mlen))) == NULL) {
1510 file_oomem(ms, mlen);
1511 php_stream_closedir(dir);
1512 errs++;
1513 goto out;
1514 }
1515 }
1516 filearr[files++] = estrndup(mfn, (mflen > sizeof(mfn) - 1)? sizeof(mfn) - 1: mflen);
1517 }
1518 php_stream_closedir(dir);
1519 if (filearr) {
1520 qsort(filearr, files, sizeof(*filearr), cmpstrp);
1521 for (i = 0; i < files; i++) {
1522 load_1(ms, action, filearr[i], &errs, mset);
1523 efree(filearr[i]);
1524 }
1525 efree(filearr);
1526 }
1527 } else
1528 load_1(ms, action, fn, &errs, mset);
1529 if (errs)
1530 goto out;
1531
1532 for (j = 0; j < MAGIC_SETS; j++) {
1533 /* Set types of tests */
1534 for (i = 0; i < mset[j].count; ) {
1535 if (mset[j].me[i].mp->cont_level != 0) {
1536 i++;
1537 continue;
1538 }
1539 i = set_text_binary(ms, mset[j].me, mset[j].count, i);
1540 }
1541 if (mset[j].me)
1542 qsort(mset[j].me, mset[j].count, sizeof(*mset[j].me),
1543 apprentice_sort);
1544
1545 /*
1546 * Make sure that any level 0 "default" line is last
1547 * (if one exists).
1548 */
1549 set_last_default(ms, mset[j].me, mset[j].count);
1550
1551 /* coalesce per file arrays into a single one, if needed */
1552 if (mset[j].count == 0)
1553 continue;
1554
1555 if (coalesce_entries(ms, mset[j].me, mset[j].count,
1556 &map->magic[j], &map->nmagic[j]) == -1) {
1557 errs++;
1558 goto out;
1559 }
1560 }
1561
1562 out:
1563 for (j = 0; j < MAGIC_SETS; j++)
1564 magic_entry_free(mset[j].me, mset[j].count);
1565
1566 if (errs) {
1567 apprentice_unmap(map);
1568 return NULL;
1569 }
1570 return map;
1571 }
1572
1573 /*
1574 * extend the sign bit if the comparison is to be signed
1575 */
1576 protected uint64_t
file_signextend(struct magic_set * ms,struct magic * m,uint64_t v)1577 file_signextend(struct magic_set *ms, struct magic *m, uint64_t v)
1578 {
1579 if (!(m->flag & UNSIGNED)) {
1580 switch(m->type) {
1581 /*
1582 * Do not remove the casts below. They are
1583 * vital. When later compared with the data,
1584 * the sign extension must have happened.
1585 */
1586 case FILE_BYTE:
1587 v = CAST(signed char, v);
1588 break;
1589 case FILE_SHORT:
1590 case FILE_BESHORT:
1591 case FILE_LESHORT:
1592 v = CAST(short, v);
1593 break;
1594 case FILE_DATE:
1595 case FILE_BEDATE:
1596 case FILE_LEDATE:
1597 case FILE_MEDATE:
1598 case FILE_LDATE:
1599 case FILE_BELDATE:
1600 case FILE_LELDATE:
1601 case FILE_MELDATE:
1602 case FILE_LONG:
1603 case FILE_BELONG:
1604 case FILE_LELONG:
1605 case FILE_MELONG:
1606 case FILE_FLOAT:
1607 case FILE_BEFLOAT:
1608 case FILE_LEFLOAT:
1609 case FILE_MSDOSDATE:
1610 case FILE_BEMSDOSDATE:
1611 case FILE_LEMSDOSDATE:
1612 case FILE_MSDOSTIME:
1613 case FILE_BEMSDOSTIME:
1614 case FILE_LEMSDOSTIME:
1615 v = CAST(int32_t, v);
1616 break;
1617 case FILE_QUAD:
1618 case FILE_BEQUAD:
1619 case FILE_LEQUAD:
1620 case FILE_QDATE:
1621 case FILE_QLDATE:
1622 case FILE_QWDATE:
1623 case FILE_BEQDATE:
1624 case FILE_BEQLDATE:
1625 case FILE_BEQWDATE:
1626 case FILE_LEQDATE:
1627 case FILE_LEQLDATE:
1628 case FILE_LEQWDATE:
1629 case FILE_DOUBLE:
1630 case FILE_BEDOUBLE:
1631 case FILE_LEDOUBLE:
1632 case FILE_OFFSET:
1633 case FILE_BEVARINT:
1634 case FILE_LEVARINT:
1635 v = CAST(int64_t, v);
1636 break;
1637 case FILE_STRING:
1638 case FILE_PSTRING:
1639 case FILE_BESTRING16:
1640 case FILE_LESTRING16:
1641 case FILE_REGEX:
1642 case FILE_SEARCH:
1643 case FILE_DEFAULT:
1644 case FILE_INDIRECT:
1645 case FILE_NAME:
1646 case FILE_USE:
1647 case FILE_CLEAR:
1648 case FILE_DER:
1649 case FILE_GUID:
1650 case FILE_OCTAL:
1651 break;
1652 default:
1653 if (ms->flags & MAGIC_CHECK)
1654 file_magwarn(ms, "cannot happen: m->type=%d\n",
1655 m->type);
1656 return FILE_BADSIZE;
1657 }
1658 }
1659 return v;
1660 }
1661
1662 private int
string_modifier_check(struct magic_set * ms,struct magic * m)1663 string_modifier_check(struct magic_set *ms, struct magic *m)
1664 {
1665 if ((ms->flags & MAGIC_CHECK) == 0)
1666 return 0;
1667
1668 if ((m->type != FILE_REGEX || (m->str_flags & REGEX_LINE_COUNT) == 0) &&
1669 (m->type != FILE_PSTRING && (m->str_flags & PSTRING_LEN) != 0)) {
1670 file_magwarn(ms,
1671 "'/BHhLl' modifiers are only allowed for pascal strings\n");
1672 return -1;
1673 }
1674 switch (m->type) {
1675 case FILE_BESTRING16:
1676 case FILE_LESTRING16:
1677 if (m->str_flags != 0) {
1678 file_magwarn(ms,
1679 "no modifiers allowed for 16-bit strings\n");
1680 return -1;
1681 }
1682 break;
1683 case FILE_STRING:
1684 case FILE_PSTRING:
1685 if ((m->str_flags & REGEX_OFFSET_START) != 0) {
1686 file_magwarn(ms,
1687 "'/%c' only allowed on regex and search\n",
1688 CHAR_REGEX_OFFSET_START);
1689 return -1;
1690 }
1691 break;
1692 case FILE_SEARCH:
1693 if (m->str_range == 0) {
1694 file_magwarn(ms,
1695 "missing range; defaulting to %d\n",
1696 STRING_DEFAULT_RANGE);
1697 m->str_range = STRING_DEFAULT_RANGE;
1698 return -1;
1699 }
1700 break;
1701 case FILE_REGEX:
1702 if ((m->str_flags & STRING_COMPACT_WHITESPACE) != 0) {
1703 file_magwarn(ms, "'/%c' not allowed on regex\n",
1704 CHAR_COMPACT_WHITESPACE);
1705 return -1;
1706 }
1707 if ((m->str_flags & STRING_COMPACT_OPTIONAL_WHITESPACE) != 0) {
1708 file_magwarn(ms, "'/%c' not allowed on regex\n",
1709 CHAR_COMPACT_OPTIONAL_WHITESPACE);
1710 return -1;
1711 }
1712 break;
1713 default:
1714 file_magwarn(ms, "coding error: m->type=%d\n",
1715 m->type);
1716 return -1;
1717 }
1718 return 0;
1719 }
1720
1721 private int
get_op(char c)1722 get_op(char c)
1723 {
1724 switch (c) {
1725 case '&':
1726 return FILE_OPAND;
1727 case '|':
1728 return FILE_OPOR;
1729 case '^':
1730 return FILE_OPXOR;
1731 case '+':
1732 return FILE_OPADD;
1733 case '-':
1734 return FILE_OPMINUS;
1735 case '*':
1736 return FILE_OPMULTIPLY;
1737 case '/':
1738 return FILE_OPDIVIDE;
1739 case '%':
1740 return FILE_OPMODULO;
1741 default:
1742 return -1;
1743 }
1744 }
1745
1746 #ifdef ENABLE_CONDITIONALS
1747 private int
get_cond(const char * l,const char ** t)1748 get_cond(const char *l, const char **t)
1749 {
1750 static const struct cond_tbl_s {
1751 char name[8];
1752 size_t len;
1753 int cond;
1754 } cond_tbl[] = {
1755 { "if", 2, COND_IF },
1756 { "elif", 4, COND_ELIF },
1757 { "else", 4, COND_ELSE },
1758 { "", 0, COND_NONE },
1759 };
1760 const struct cond_tbl_s *p;
1761
1762 for (p = cond_tbl; p->len; p++) {
1763 if (strncmp(l, p->name, p->len) == 0 &&
1764 isspace(CAST(unsigned char, l[p->len]))) {
1765 if (t)
1766 *t = l + p->len;
1767 break;
1768 }
1769 }
1770 return p->cond;
1771 }
1772
1773 private int
check_cond(struct magic_set * ms,int cond,uint32_t cont_level)1774 check_cond(struct magic_set *ms, int cond, uint32_t cont_level)
1775 {
1776 int last_cond;
1777 last_cond = ms->c.li[cont_level].last_cond;
1778
1779 switch (cond) {
1780 case COND_IF:
1781 if (last_cond != COND_NONE && last_cond != COND_ELIF) {
1782 if (ms->flags & MAGIC_CHECK)
1783 file_magwarn(ms, "syntax error: `if'");
1784 return -1;
1785 }
1786 last_cond = COND_IF;
1787 break;
1788
1789 case COND_ELIF:
1790 if (last_cond != COND_IF && last_cond != COND_ELIF) {
1791 if (ms->flags & MAGIC_CHECK)
1792 file_magwarn(ms, "syntax error: `elif'");
1793 return -1;
1794 }
1795 last_cond = COND_ELIF;
1796 break;
1797
1798 case COND_ELSE:
1799 if (last_cond != COND_IF && last_cond != COND_ELIF) {
1800 if (ms->flags & MAGIC_CHECK)
1801 file_magwarn(ms, "syntax error: `else'");
1802 return -1;
1803 }
1804 last_cond = COND_NONE;
1805 break;
1806
1807 case COND_NONE:
1808 last_cond = COND_NONE;
1809 break;
1810 }
1811
1812 ms->c.li[cont_level].last_cond = last_cond;
1813 return 0;
1814 }
1815 #endif /* ENABLE_CONDITIONALS */
1816
1817 private int
parse_indirect_modifier(struct magic_set * ms,struct magic * m,const char ** lp)1818 parse_indirect_modifier(struct magic_set *ms, struct magic *m, const char **lp)
1819 {
1820 const char *l = *lp;
1821
1822 while (!isspace(CAST(unsigned char, *++l)))
1823 switch (*l) {
1824 case CHAR_INDIRECT_RELATIVE:
1825 m->str_flags |= INDIRECT_RELATIVE;
1826 break;
1827 default:
1828 if (ms->flags & MAGIC_CHECK)
1829 file_magwarn(ms, "indirect modifier `%c' "
1830 "invalid", *l);
1831 *lp = l;
1832 return -1;
1833 }
1834 *lp = l;
1835 return 0;
1836 }
1837
1838 private void
parse_op_modifier(struct magic_set * ms,struct magic * m,const char ** lp,int op)1839 parse_op_modifier(struct magic_set *ms, struct magic *m, const char **lp,
1840 int op)
1841 {
1842 const char *l = *lp;
1843 char *t;
1844 uint64_t val;
1845
1846 ++l;
1847 m->mask_op |= op;
1848 val = CAST(uint64_t, strtoull(l, &t, 0));
1849 l = t;
1850 m->num_mask = file_signextend(ms, m, val);
1851 eatsize(&l);
1852 *lp = l;
1853 }
1854
1855 private int
parse_string_modifier(struct magic_set * ms,struct magic * m,const char ** lp)1856 parse_string_modifier(struct magic_set *ms, struct magic *m, const char **lp)
1857 {
1858 const char *l = *lp;
1859 char *t;
1860 int have_range = 0;
1861
1862 while (!isspace(CAST(unsigned char, *++l))) {
1863 switch (*l) {
1864 case '0': case '1': case '2':
1865 case '3': case '4': case '5':
1866 case '6': case '7': case '8':
1867 case '9':
1868 if (have_range && (ms->flags & MAGIC_CHECK))
1869 file_magwarn(ms, "multiple ranges");
1870 have_range = 1;
1871 m->str_range = CAST(uint32_t, strtoul(l, &t, 0));
1872 if (m->str_range == 0)
1873 file_magwarn(ms, "zero range");
1874 l = t - 1;
1875 break;
1876 case CHAR_COMPACT_WHITESPACE:
1877 m->str_flags |= STRING_COMPACT_WHITESPACE;
1878 break;
1879 case CHAR_COMPACT_OPTIONAL_WHITESPACE:
1880 m->str_flags |= STRING_COMPACT_OPTIONAL_WHITESPACE;
1881 break;
1882 case CHAR_IGNORE_LOWERCASE:
1883 m->str_flags |= STRING_IGNORE_LOWERCASE;
1884 break;
1885 case CHAR_IGNORE_UPPERCASE:
1886 m->str_flags |= STRING_IGNORE_UPPERCASE;
1887 break;
1888 case CHAR_REGEX_OFFSET_START:
1889 m->str_flags |= REGEX_OFFSET_START;
1890 break;
1891 case CHAR_BINTEST:
1892 m->str_flags |= STRING_BINTEST;
1893 break;
1894 case CHAR_TEXTTEST:
1895 m->str_flags |= STRING_TEXTTEST;
1896 break;
1897 case CHAR_TRIM:
1898 m->str_flags |= STRING_TRIM;
1899 break;
1900 case CHAR_FULL_WORD:
1901 m->str_flags |= STRING_FULL_WORD;
1902 break;
1903 case CHAR_PSTRING_1_LE:
1904 #define SET_LENGTH(a) m->str_flags = (m->str_flags & ~PSTRING_LEN) | (a)
1905 if (m->type != FILE_PSTRING)
1906 goto bad;
1907 SET_LENGTH(PSTRING_1_LE);
1908 break;
1909 case CHAR_PSTRING_2_BE:
1910 if (m->type != FILE_PSTRING)
1911 goto bad;
1912 SET_LENGTH(PSTRING_2_BE);
1913 break;
1914 case CHAR_PSTRING_2_LE:
1915 if (m->type != FILE_PSTRING)
1916 goto bad;
1917 SET_LENGTH(PSTRING_2_LE);
1918 break;
1919 case CHAR_PSTRING_4_BE:
1920 if (m->type != FILE_PSTRING)
1921 goto bad;
1922 SET_LENGTH(PSTRING_4_BE);
1923 break;
1924 case CHAR_PSTRING_4_LE:
1925 switch (m->type) {
1926 case FILE_PSTRING:
1927 case FILE_REGEX:
1928 break;
1929 default:
1930 goto bad;
1931 }
1932 SET_LENGTH(PSTRING_4_LE);
1933 break;
1934 case CHAR_PSTRING_LENGTH_INCLUDES_ITSELF:
1935 if (m->type != FILE_PSTRING)
1936 goto bad;
1937 m->str_flags |= PSTRING_LENGTH_INCLUDES_ITSELF;
1938 break;
1939 default:
1940 bad:
1941 if (ms->flags & MAGIC_CHECK)
1942 file_magwarn(ms, "string modifier `%c' "
1943 "invalid", *l);
1944 goto out;
1945 }
1946 /* allow multiple '/' for readability */
1947 if (l[1] == '/' && !isspace(CAST(unsigned char, l[2])))
1948 l++;
1949 }
1950 if (string_modifier_check(ms, m) == -1)
1951 goto out;
1952 *lp = l;
1953 return 0;
1954 out:
1955 *lp = l;
1956 return -1;
1957 }
1958
1959 /*
1960 * parse one line from magic file, put into magic[index++] if valid
1961 */
1962 private int
parse(struct magic_set * ms,struct magic_entry * me,const char * line,size_t lineno,int action)1963 parse(struct magic_set *ms, struct magic_entry *me, const char *line,
1964 size_t lineno, int action)
1965 {
1966 #ifdef ENABLE_CONDITIONALS
1967 static uint32_t last_cont_level = 0;
1968 #endif
1969 size_t i;
1970 struct magic *m;
1971 const char *l = line;
1972 char *t;
1973 int op;
1974 uint32_t cont_level;
1975 int32_t diff;
1976
1977 cont_level = 0;
1978
1979 /*
1980 * Parse the offset.
1981 */
1982 while (*l == '>') {
1983 ++l; /* step over */
1984 cont_level++;
1985 }
1986 #ifdef ENABLE_CONDITIONALS
1987 if (cont_level == 0 || cont_level > last_cont_level)
1988 if (file_check_mem(ms, cont_level) == -1)
1989 return -1;
1990 last_cont_level = cont_level;
1991 #endif
1992 if (cont_level != 0) {
1993 if (me->mp == NULL) {
1994 file_magerror(ms, "No current entry for continuation");
1995 return -1;
1996 }
1997 if (me->cont_count == 0) {
1998 file_magerror(ms, "Continuations present with 0 count");
1999 return -1;
2000 }
2001 m = &me->mp[me->cont_count - 1];
2002 diff = CAST(int32_t, cont_level) - CAST(int32_t, m->cont_level);
2003 if (diff > 1)
2004 file_magwarn(ms, "New continuation level %u is more "
2005 "than one larger than current level %u", cont_level,
2006 m->cont_level);
2007 if (me->cont_count == me->max_count) {
2008 struct magic *nm;
2009 size_t cnt = me->max_count + ALLOC_CHUNK;
2010 if ((nm = CAST(struct magic *, erealloc(me->mp,
2011 sizeof(*nm) * cnt))) == NULL) {
2012 file_oomem(ms, sizeof(*nm) * cnt);
2013 return -1;
2014 }
2015 me->mp = nm;
2016 me->max_count = CAST(uint32_t, cnt);
2017 }
2018 m = &me->mp[me->cont_count++];
2019 (void)memset(m, 0, sizeof(*m));
2020 m->cont_level = cont_level;
2021 } else {
2022 static const size_t len = sizeof(*m) * ALLOC_CHUNK;
2023 if (me->mp != NULL)
2024 return 1;
2025 if ((m = CAST(struct magic *, emalloc(len))) == NULL) {
2026 file_oomem(ms, len);
2027 return -1;
2028 }
2029 me->mp = m;
2030 me->max_count = ALLOC_CHUNK;
2031 (void)memset(m, 0, sizeof(*m));
2032 m->factor_op = FILE_FACTOR_OP_NONE;
2033 m->cont_level = 0;
2034 me->cont_count = 1;
2035 }
2036 m->lineno = CAST(uint32_t, lineno);
2037
2038 if (*l == '&') { /* m->cont_level == 0 checked below. */
2039 ++l; /* step over */
2040 m->flag |= OFFADD;
2041 }
2042 if (*l == '(') {
2043 ++l; /* step over */
2044 m->flag |= INDIR;
2045 if (m->flag & OFFADD)
2046 m->flag = (m->flag & ~OFFADD) | INDIROFFADD;
2047
2048 if (*l == '&') { /* m->cont_level == 0 checked below */
2049 ++l; /* step over */
2050 m->flag |= OFFADD;
2051 }
2052 }
2053 /* Indirect offsets are not valid at level 0. */
2054 if (m->cont_level == 0 && (m->flag & (OFFADD | INDIROFFADD))) {
2055 if (ms->flags & MAGIC_CHECK)
2056 file_magwarn(ms, "relative offset at level 0");
2057 return -1;
2058 }
2059
2060 /* get offset, then skip over it */
2061 if (*l == '-') {
2062 ++l; /* step over */
2063 m->flag |= OFFNEGATIVE;
2064 }
2065 m->offset = CAST(int32_t, strtol(l, &t, 0));
2066 if (l == t) {
2067 if (ms->flags & MAGIC_CHECK)
2068 file_magwarn(ms, "offset `%s' invalid", l);
2069 return -1;
2070 }
2071
2072 l = t;
2073
2074 if (m->flag & INDIR) {
2075 m->in_type = FILE_LONG;
2076 m->in_offset = 0;
2077 m->in_op = 0;
2078 /*
2079 * read [.,lbs][+-]nnnnn)
2080 */
2081 if (*l == '.' || *l == ',') {
2082 if (*l == ',')
2083 m->in_op |= FILE_OPSIGNED;
2084 l++;
2085 switch (*l) {
2086 case 'l':
2087 m->in_type = FILE_LELONG;
2088 break;
2089 case 'L':
2090 m->in_type = FILE_BELONG;
2091 break;
2092 case 'm':
2093 m->in_type = FILE_MELONG;
2094 break;
2095 case 'h':
2096 case 's':
2097 m->in_type = FILE_LESHORT;
2098 break;
2099 case 'H':
2100 case 'S':
2101 m->in_type = FILE_BESHORT;
2102 break;
2103 case 'c':
2104 case 'b':
2105 case 'C':
2106 case 'B':
2107 m->in_type = FILE_BYTE;
2108 break;
2109 case 'e':
2110 case 'f':
2111 case 'g':
2112 m->in_type = FILE_LEDOUBLE;
2113 break;
2114 case 'E':
2115 case 'F':
2116 case 'G':
2117 m->in_type = FILE_BEDOUBLE;
2118 break;
2119 case 'i':
2120 m->in_type = FILE_LEID3;
2121 break;
2122 case 'I':
2123 m->in_type = FILE_BEID3;
2124 break;
2125 case 'o':
2126 m->in_type = FILE_OCTAL;
2127 break;
2128 case 'q':
2129 m->in_type = FILE_LEQUAD;
2130 break;
2131 case 'Q':
2132 m->in_type = FILE_BEQUAD;
2133 break;
2134 default:
2135 if (ms->flags & MAGIC_CHECK)
2136 file_magwarn(ms,
2137 "indirect offset type `%c' invalid",
2138 *l);
2139 return -1;
2140 }
2141 l++;
2142 }
2143
2144 if (*l == '~') {
2145 m->in_op |= FILE_OPINVERSE;
2146 l++;
2147 }
2148 if ((op = get_op(*l)) != -1) {
2149 m->in_op |= op;
2150 l++;
2151 }
2152 if (*l == '(') {
2153 m->in_op |= FILE_OPINDIRECT;
2154 l++;
2155 }
2156 if (isdigit(CAST(unsigned char, *l)) || *l == '-') {
2157 m->in_offset = CAST(int32_t, strtol(l, &t, 0));
2158 if (l == t) {
2159 if (ms->flags & MAGIC_CHECK)
2160 file_magwarn(ms,
2161 "in_offset `%s' invalid", l);
2162 return -1;
2163 }
2164 l = t;
2165 }
2166 if (*l++ != ')' ||
2167 ((m->in_op & FILE_OPINDIRECT) && *l++ != ')')) {
2168 if (ms->flags & MAGIC_CHECK)
2169 file_magwarn(ms,
2170 "missing ')' in indirect offset");
2171 return -1;
2172 }
2173 }
2174 EATAB;
2175
2176 #ifdef ENABLE_CONDITIONALS
2177 m->cond = get_cond(l, &l);
2178 if (check_cond(ms, m->cond, cont_level) == -1)
2179 return -1;
2180
2181 EATAB;
2182 #endif
2183
2184 /*
2185 * Parse the type.
2186 */
2187 if (*l == 'u') {
2188 /*
2189 * Try it as a keyword type prefixed by "u"; match what
2190 * follows the "u". If that fails, try it as an SUS
2191 * integer type.
2192 */
2193 m->type = get_type(type_tbl, l + 1, &l);
2194 if (m->type == FILE_INVALID) {
2195 /*
2196 * Not a keyword type; parse it as an SUS type,
2197 * 'u' possibly followed by a number or C/S/L.
2198 */
2199 m->type = get_standard_integer_type(l, &l);
2200 }
2201 /* It's unsigned. */
2202 if (m->type != FILE_INVALID)
2203 m->flag |= UNSIGNED;
2204 } else {
2205 /*
2206 * Try it as a keyword type. If that fails, try it as
2207 * an SUS integer type if it begins with "d" or as an
2208 * SUS string type if it begins with "s". In any case,
2209 * it's not unsigned.
2210 */
2211 m->type = get_type(type_tbl, l, &l);
2212 if (m->type == FILE_INVALID) {
2213 /*
2214 * Not a keyword type; parse it as an SUS type,
2215 * either 'd' possibly followed by a number or
2216 * C/S/L, or just 's'.
2217 */
2218 if (*l == 'd')
2219 m->type = get_standard_integer_type(l, &l);
2220 else if (*l == 's'
2221 && !isalpha(CAST(unsigned char, l[1]))) {
2222 m->type = FILE_STRING;
2223 ++l;
2224 }
2225 }
2226 }
2227
2228 if (m->type == FILE_INVALID) {
2229 /* Not found - try it as a special keyword. */
2230 m->type = get_type(special_tbl, l, &l);
2231 }
2232
2233 if (m->type == FILE_INVALID) {
2234 if (ms->flags & MAGIC_CHECK)
2235 file_magwarn(ms, "type `%s' invalid", l);
2236 return -1;
2237 }
2238
2239 if (m->type == FILE_NAME && cont_level != 0) {
2240 if (ms->flags & MAGIC_CHECK)
2241 file_magwarn(ms, "`name%s' entries can only be "
2242 "declared at top level", l);
2243 return -1;
2244 }
2245
2246 /* New-style anding: "0 byte&0x80 =0x80 dynamically linked" */
2247 /* New and improved: ~ & | ^ + - * / % -- exciting, isn't it? */
2248
2249 m->mask_op = 0;
2250 if (*l == '~') {
2251 if (!IS_LIBMAGIC_STRING(m->type))
2252 m->mask_op |= FILE_OPINVERSE;
2253 else if (ms->flags & MAGIC_CHECK)
2254 file_magwarn(ms, "'~' invalid for string types");
2255 ++l;
2256 }
2257 m->str_range = 0;
2258 m->str_flags = m->type == FILE_PSTRING ? PSTRING_1_LE : 0;
2259 if ((op = get_op(*l)) != -1) {
2260 if (IS_LIBMAGIC_STRING(m->type)) {
2261 int r;
2262
2263 if (op != FILE_OPDIVIDE) {
2264 if (ms->flags & MAGIC_CHECK)
2265 file_magwarn(ms,
2266 "invalid string/indirect op: "
2267 "`%c'", *t);
2268 return -1;
2269 }
2270
2271 if (m->type == FILE_INDIRECT)
2272 r = parse_indirect_modifier(ms, m, &l);
2273 else
2274 r = parse_string_modifier(ms, m, &l);
2275 if (r == -1)
2276 return -1;
2277 } else
2278 parse_op_modifier(ms, m, &l, op);
2279 }
2280
2281 /*
2282 * We used to set mask to all 1's here, instead let's just not do
2283 * anything if mask = 0 (unless you have a better idea)
2284 */
2285 EATAB;
2286
2287 switch (*l) {
2288 case '>':
2289 case '<':
2290 m->reln = *l;
2291 ++l;
2292 if (*l == '=') {
2293 if (ms->flags & MAGIC_CHECK) {
2294 file_magwarn(ms, "%c= not supported",
2295 m->reln);
2296 return -1;
2297 }
2298 ++l;
2299 }
2300 break;
2301 /* Old-style anding: "0 byte &0x80 dynamically linked" */
2302 case '&':
2303 case '^':
2304 case '=':
2305 m->reln = *l;
2306 ++l;
2307 if (*l == '=') {
2308 /* HP compat: ignore &= etc. */
2309 ++l;
2310 }
2311 break;
2312 case '!':
2313 m->reln = *l;
2314 ++l;
2315 break;
2316 default:
2317 m->reln = '='; /* the default relation */
2318 if (*l == 'x' && ((isascii(CAST(unsigned char, l[1])) &&
2319 isspace(CAST(unsigned char, l[1]))) || !l[1])) {
2320 m->reln = *l;
2321 ++l;
2322 }
2323 break;
2324 }
2325 /*
2326 * Grab the value part, except for an 'x' reln.
2327 */
2328 if (m->reln != 'x' && getvalue(ms, m, &l, action))
2329 return -1;
2330
2331 /*
2332 * TODO finish this macro and start using it!
2333 * #define offsetcheck {if (offset > ms->bytes_max -1)
2334 * magwarn("offset too big"); }
2335 */
2336
2337 /*
2338 * Now get last part - the description
2339 */
2340 EATAB;
2341 if (l[0] == '\b') {
2342 ++l;
2343 m->flag |= NOSPACE;
2344 } else if ((l[0] == '\\') && (l[1] == 'b')) {
2345 ++l;
2346 ++l;
2347 m->flag |= NOSPACE;
2348 }
2349 for (i = 0; (m->desc[i++] = *l++) != '\0' && i < sizeof(m->desc); )
2350 continue;
2351 if (i == sizeof(m->desc)) {
2352 m->desc[sizeof(m->desc) - 1] = '\0';
2353 if (ms->flags & MAGIC_CHECK)
2354 file_magwarn(ms, "description `%s' truncated", m->desc);
2355 }
2356
2357 /*
2358 * We only do this check while compiling, or if any of the magic
2359 * files were not compiled.
2360 */
2361 if (ms->flags & MAGIC_CHECK) {
2362 if (check_format(ms, m) == -1)
2363 return -1;
2364 }
2365 #ifndef COMPILE_ONLY
2366 if (action == FILE_CHECK) {
2367 file_mdump(m);
2368 }
2369 #endif
2370 m->mimetype[0] = '\0'; /* initialise MIME type to none */
2371 return 0;
2372 }
2373
2374 /*
2375 * parse a STRENGTH annotation line from magic file, put into magic[index - 1]
2376 * if valid
2377 */
2378 /*ARGSUSED*/
2379 private int
parse_strength(struct magic_set * ms,struct magic_entry * me,const char * line,size_t len)2380 parse_strength(struct magic_set *ms, struct magic_entry *me, const char *line,
2381 size_t len __attribute__((__unused__)))
2382 {
2383 const char *l = line;
2384 char *el;
2385 unsigned long factor;
2386 struct magic *m = &me->mp[0];
2387
2388 if (m->factor_op != FILE_FACTOR_OP_NONE) {
2389 file_magwarn(ms,
2390 "Current entry already has a strength type: %c %d",
2391 m->factor_op, m->factor);
2392 return -1;
2393 }
2394 if (m->type == FILE_NAME) {
2395 file_magwarn(ms, "%s: Strength setting is not supported in "
2396 "\"name\" magic entries", m->value.s);
2397 return -1;
2398 }
2399 EATAB;
2400 switch (*l) {
2401 case FILE_FACTOR_OP_NONE:
2402 case FILE_FACTOR_OP_PLUS:
2403 case FILE_FACTOR_OP_MINUS:
2404 case FILE_FACTOR_OP_TIMES:
2405 case FILE_FACTOR_OP_DIV:
2406 m->factor_op = *l++;
2407 break;
2408 default:
2409 file_magwarn(ms, "Unknown factor op `%c'", *l);
2410 return -1;
2411 }
2412 EATAB;
2413 factor = strtoul(l, &el, 0);
2414 if (factor > 255) {
2415 file_magwarn(ms, "Too large factor `%lu'", factor);
2416 goto out;
2417 }
2418 if (*el && !isspace(CAST(unsigned char, *el))) {
2419 file_magwarn(ms, "Bad factor `%s'", l);
2420 goto out;
2421 }
2422 m->factor = CAST(uint8_t, factor);
2423 if (m->factor == 0 && m->factor_op == FILE_FACTOR_OP_DIV) {
2424 file_magwarn(ms, "Cannot have factor op `%c' and factor %u",
2425 m->factor_op, m->factor);
2426 goto out;
2427 }
2428 return 0;
2429 out:
2430 m->factor_op = FILE_FACTOR_OP_NONE;
2431 m->factor = 0;
2432 return -1;
2433 }
2434
2435 private int
goodchar(unsigned char x,const char * extra)2436 goodchar(unsigned char x, const char *extra)
2437 {
2438 return (isascii(x) && isalnum(x)) || strchr(extra, x);
2439 }
2440
2441 private int
parse_extra(struct magic_set * ms,struct magic_entry * me,const char * line,size_t llen,zend_off_t off,size_t len,const char * name,const char * extra,int nt)2442 parse_extra(struct magic_set *ms, struct magic_entry *me, const char *line,
2443 size_t llen, zend_off_t off, size_t len, const char *name, const char *extra, int nt)
2444 {
2445 size_t i;
2446 const char *l = line;
2447 struct magic *m = &me->mp[me->cont_count == 0 ? 0 : me->cont_count - 1];
2448 char *buf = CAST(char *, CAST(void *, m)) + off;
2449
2450 if (buf[0] != '\0') {
2451 len = nt ? strlen(buf) : len;
2452 file_magwarn(ms, "Current entry already has a %s type "
2453 "`%.*s', new type `%s'", name, CAST(int, len), buf, l);
2454 return -1;
2455 }
2456
2457 if (*m->desc == '\0') {
2458 file_magwarn(ms, "Current entry does not yet have a "
2459 "description for adding a %s type", name);
2460 return -1;
2461 }
2462
2463 EATAB;
2464 for (i = 0; *l && i < llen && i < len && goodchar(*l, extra);
2465 buf[i++] = *l++)
2466 continue;
2467
2468 if (i == len && *l) {
2469 if (nt)
2470 buf[len - 1] = '\0';
2471 if (ms->flags & MAGIC_CHECK)
2472 file_magwarn(ms, "%s type `%s' truncated %"
2473 SIZE_T_FORMAT "u", name, line, i);
2474 } else {
2475 if (!isspace(CAST(unsigned char, *l)) && !goodchar(*l, extra))
2476 file_magwarn(ms, "%s type `%s' has bad char '%c'",
2477 name, line, *l);
2478 if (nt)
2479 buf[i] = '\0';
2480 }
2481
2482 if (i > 0)
2483 return 0;
2484
2485 file_magerror(ms, "Bad magic entry '%s'", line);
2486 return -1;
2487 }
2488
2489 /*
2490 * Parse an Apple CREATOR/TYPE annotation from magic file and put it into
2491 * magic[index - 1]
2492 */
2493 private int
parse_apple(struct magic_set * ms,struct magic_entry * me,const char * line,size_t len)2494 parse_apple(struct magic_set *ms, struct magic_entry *me, const char *line,
2495 size_t len)
2496 {
2497 return parse_extra(ms, me, line, len,
2498 CAST(off_t, offsetof(struct magic, apple)),
2499 sizeof(me->mp[0].apple), "APPLE", "!+-./?", 0);
2500 }
2501
2502 /*
2503 * Parse a comma-separated list of extensions
2504 */
2505 private int
parse_ext(struct magic_set * ms,struct magic_entry * me,const char * line,size_t len)2506 parse_ext(struct magic_set *ms, struct magic_entry *me, const char *line,
2507 size_t len)
2508 {
2509 return parse_extra(ms, me, line, len,
2510 CAST(off_t, offsetof(struct magic, ext)),
2511 sizeof(me->mp[0].ext), "EXTENSION", ",!+-/@?_$&", 0); /* & for b&w */
2512 }
2513
2514 /*
2515 * parse a MIME annotation line from magic file, put into magic[index - 1]
2516 * if valid
2517 */
2518 private int
parse_mime(struct magic_set * ms,struct magic_entry * me,const char * line,size_t len)2519 parse_mime(struct magic_set *ms, struct magic_entry *me, const char *line,
2520 size_t len)
2521 {
2522 return parse_extra(ms, me, line, len,
2523 CAST(off_t, offsetof(struct magic, mimetype)),
2524 sizeof(me->mp[0].mimetype), "MIME", "+-/.$?:{}", 1);
2525 }
2526
2527 private int
check_format_type(const char * ptr,int type,const char ** estr)2528 check_format_type(const char *ptr, int type, const char **estr)
2529 {
2530 int quad = 0, h;
2531 size_t len, cnt;
2532 if (*ptr == '\0') {
2533 /* Missing format string; bad */
2534 *estr = "missing format spec";
2535 return -1;
2536 }
2537
2538 switch (file_formats[type]) {
2539 case FILE_FMT_QUAD:
2540 quad = 1;
2541 /*FALLTHROUGH*/
2542 case FILE_FMT_NUM:
2543 if (quad == 0) {
2544 switch (type) {
2545 case FILE_BYTE:
2546 h = 2;
2547 break;
2548 case FILE_SHORT:
2549 case FILE_BESHORT:
2550 case FILE_LESHORT:
2551 h = 1;
2552 break;
2553 case FILE_LONG:
2554 case FILE_BELONG:
2555 case FILE_LELONG:
2556 case FILE_MELONG:
2557 case FILE_LEID3:
2558 case FILE_BEID3:
2559 case FILE_INDIRECT:
2560 h = 0;
2561 break;
2562 default:
2563 fprintf(stderr, "Bad number format %d", type);
2564 abort();
2565 }
2566 } else
2567 h = 0;
2568 while (*ptr && strchr("-.#", *ptr) != NULL)
2569 ptr++;
2570 #define CHECKLEN() do { \
2571 for (len = cnt = 0; isdigit(CAST(unsigned char, *ptr)); ptr++, cnt++) \
2572 len = len * 10 + (*ptr - '0'); \
2573 if (cnt > 5 || len > 1024) \
2574 goto toolong; \
2575 } while (/*CONSTCOND*/0)
2576
2577 CHECKLEN();
2578 if (*ptr == '.')
2579 ptr++;
2580 CHECKLEN();
2581 if (quad) {
2582 if (*ptr++ != 'l')
2583 goto invalid;
2584 if (*ptr++ != 'l')
2585 goto invalid;
2586 }
2587
2588 switch (*ptr++) {
2589 #ifdef STRICT_FORMAT /* "long" formats are int formats for us */
2590 /* so don't accept the 'l' modifier */
2591 case 'l':
2592 switch (*ptr++) {
2593 case 'i':
2594 case 'd':
2595 case 'u':
2596 case 'o':
2597 case 'x':
2598 case 'X':
2599 if (h == 0)
2600 return 0;
2601 /*FALLTHROUGH*/
2602 default:
2603 goto invalid;
2604 }
2605
2606 /*
2607 * Don't accept h and hh modifiers. They make writing
2608 * magic entries more complicated, for very little benefit
2609 */
2610 case 'h':
2611 if (h-- <= 0)
2612 goto invalid;
2613 switch (*ptr++) {
2614 case 'h':
2615 if (h-- <= 0)
2616 goto invalid;
2617 switch (*ptr++) {
2618 case 'i':
2619 case 'd':
2620 case 'u':
2621 case 'o':
2622 case 'x':
2623 case 'X':
2624 return 0;
2625 default:
2626 goto invalid;
2627 }
2628 case 'i':
2629 case 'd':
2630 case 'u':
2631 case 'o':
2632 case 'x':
2633 case 'X':
2634 if (h == 0)
2635 return 0;
2636 /*FALLTHROUGH*/
2637 default:
2638 goto invalid;
2639 }
2640 #endif
2641 case 'c':
2642 if (h == 2)
2643 return 0;
2644 goto invalid;
2645 case 'i':
2646 case 'd':
2647 case 'u':
2648 case 'o':
2649 case 'x':
2650 case 'X':
2651 #ifdef STRICT_FORMAT
2652 if (h == 0)
2653 return 0;
2654 /*FALLTHROUGH*/
2655 #else
2656 return 0;
2657 #endif
2658 default:
2659 goto invalid;
2660 }
2661
2662 case FILE_FMT_FLOAT:
2663 case FILE_FMT_DOUBLE:
2664 if (*ptr == '-')
2665 ptr++;
2666 if (*ptr == '.')
2667 ptr++;
2668 CHECKLEN();
2669 if (*ptr == '.')
2670 ptr++;
2671 CHECKLEN();
2672 switch (*ptr++) {
2673 case 'e':
2674 case 'E':
2675 case 'f':
2676 case 'F':
2677 case 'g':
2678 case 'G':
2679 return 0;
2680
2681 default:
2682 goto invalid;
2683 }
2684
2685
2686 case FILE_FMT_STR:
2687 if (*ptr == '-')
2688 ptr++;
2689 while (isdigit(CAST(unsigned char, *ptr)))
2690 ptr++;
2691 if (*ptr == '.') {
2692 ptr++;
2693 while (isdigit(CAST(unsigned char , *ptr)))
2694 ptr++;
2695 }
2696
2697 switch (*ptr++) {
2698 case 's':
2699 return 0;
2700 default:
2701 goto invalid;
2702 }
2703
2704 default:
2705 /* internal error */
2706 fprintf(stderr, "Bad file format %d", type);
2707 abort();
2708 }
2709 invalid:
2710 *estr = "not valid";
2711 toolong:
2712 *estr = "too long";
2713 return -1;
2714 }
2715
2716 /*
2717 * Check that the optional printf format in description matches
2718 * the type of the magic.
2719 */
2720 private int
check_format(struct magic_set * ms,struct magic * m)2721 check_format(struct magic_set *ms, struct magic *m)
2722 {
2723 char *ptr;
2724 const char *estr;
2725
2726 for (ptr = m->desc; *ptr; ptr++)
2727 if (*ptr == '%')
2728 break;
2729 if (*ptr == '\0') {
2730 /* No format string; ok */
2731 return 1;
2732 }
2733
2734 assert(file_nformats == file_nnames);
2735
2736 if (m->type >= file_nformats) {
2737 file_magwarn(ms, "Internal error inconsistency between "
2738 "m->type and format strings");
2739 return -1;
2740 }
2741 if (file_formats[m->type] == FILE_FMT_NONE) {
2742 file_magwarn(ms, "No format string for `%s' with description "
2743 "`%s'", m->desc, file_names[m->type]);
2744 return -1;
2745 }
2746
2747 ptr++;
2748 if (check_format_type(ptr, m->type, &estr) == -1) {
2749 /*
2750 * TODO: this error message is unhelpful if the format
2751 * string is not one character long
2752 */
2753 file_magwarn(ms, "Printf format is %s for type "
2754 "`%s' in description `%s'", estr,
2755 file_names[m->type], m->desc);
2756 return -1;
2757 }
2758
2759 for (; *ptr; ptr++) {
2760 if (*ptr == '%') {
2761 file_magwarn(ms,
2762 "Too many format strings (should have at most one) "
2763 "for `%s' with description `%s'",
2764 file_names[m->type], m->desc);
2765 return -1;
2766 }
2767 }
2768 return 0;
2769 }
2770
2771 /*
2772 * Read a numeric value from a pointer, into the value union of a magic
2773 * pointer, according to the magic type. Update the string pointer to point
2774 * just after the number read. Return 0 for success, non-zero for failure.
2775 */
2776 private int
getvalue(struct magic_set * ms,struct magic * m,const char ** p,int action)2777 getvalue(struct magic_set *ms, struct magic *m, const char **p, int action)
2778 {
2779 char *ep;
2780 uint64_t ull;
2781
2782 switch (m->type) {
2783 case FILE_BESTRING16:
2784 case FILE_LESTRING16:
2785 case FILE_STRING:
2786 case FILE_PSTRING:
2787 case FILE_REGEX:
2788 case FILE_SEARCH:
2789 case FILE_NAME:
2790 case FILE_USE:
2791 case FILE_DER:
2792 case FILE_OCTAL:
2793 *p = getstr(ms, m, *p, action == FILE_COMPILE);
2794 if (*p == NULL) {
2795 if (ms->flags & MAGIC_CHECK)
2796 file_magwarn(ms, "cannot get string from `%s'",
2797 m->value.s);
2798 return -1;
2799 }
2800 if (m->type == FILE_REGEX) {
2801 zend_string *pattern;
2802 int options = 0;
2803 pcre_cache_entry *pce;
2804
2805 pattern = convert_libmagic_pattern(m->value.s, strlen(m->value.s), options);
2806
2807 if ((pce = pcre_get_compiled_regex_cache(pattern)) == NULL) {
2808 zend_string_release(pattern);
2809 return -1;
2810 }
2811 zend_string_release(pattern);
2812
2813 return 0;
2814 }
2815 return 0;
2816 default:
2817 if (m->reln == 'x')
2818 return 0;
2819 break;
2820 }
2821
2822 switch (m->type) {
2823 case FILE_FLOAT:
2824 case FILE_BEFLOAT:
2825 case FILE_LEFLOAT:
2826 errno = 0;
2827 #ifdef HAVE_STRTOF
2828 m->value.f = strtof(*p, &ep);
2829 #else
2830 m->value.f = (float)strtod(*p, &ep);
2831 #endif
2832 if (errno == 0)
2833 *p = ep;
2834 return 0;
2835 case FILE_DOUBLE:
2836 case FILE_BEDOUBLE:
2837 case FILE_LEDOUBLE:
2838 errno = 0;
2839 m->value.d = strtod(*p, &ep);
2840 if (errno == 0)
2841 *p = ep;
2842 return 0;
2843 case FILE_GUID:
2844 if (file_parse_guid(*p, m->value.guid) == -1)
2845 return -1;
2846 *p += FILE_GUID_SIZE - 1;
2847 return 0;
2848 default:
2849 errno = 0;
2850 ull = CAST(uint64_t, strtoull(*p, &ep, 0));
2851 m->value.q = file_signextend(ms, m, ull);
2852 if (*p == ep) {
2853 file_magwarn(ms, "Unparsable number `%s'", *p);
2854 } else {
2855 size_t ts = typesize(m->type);
2856 uint64_t x;
2857 const char *q;
2858
2859 if (ts == FILE_BADSIZE) {
2860 file_magwarn(ms,
2861 "Expected numeric type got `%s'",
2862 type_tbl[m->type].name);
2863 }
2864 for (q = *p; isspace(CAST(unsigned char, *q)); q++)
2865 continue;
2866 if (*q == '-')
2867 ull = -CAST(int64_t, ull);
2868 switch (ts) {
2869 case 1:
2870 x = CAST(uint64_t, ull & ~0xffULL);
2871 break;
2872 case 2:
2873 x = CAST(uint64_t, ull & ~0xffffULL);
2874 break;
2875 case 4:
2876 x = CAST(uint64_t, ull & ~0xffffffffULL);
2877 break;
2878 case 8:
2879 x = 0;
2880 break;
2881 default:
2882 fprintf(stderr, "Bad width %zu", ts);
2883 abort();
2884 }
2885 if (x) {
2886 file_magwarn(ms, "Overflow for numeric"
2887 " type `%s' value %#" PRIx64,
2888 type_tbl[m->type].name, ull);
2889 }
2890 }
2891 if (errno == 0) {
2892 *p = ep;
2893 eatsize(p);
2894 }
2895 return 0;
2896 }
2897 }
2898
2899 /*
2900 * Convert a string containing C character escapes. Stop at an unescaped
2901 * space or tab.
2902 * Copy the converted version to "m->value.s", and the length in m->vallen.
2903 * Return updated scan pointer as function result. Warn if set.
2904 */
2905 private const char *
getstr(struct magic_set * ms,struct magic * m,const char * s,int warn)2906 getstr(struct magic_set *ms, struct magic *m, const char *s, int warn)
2907 {
2908 const char *origs = s;
2909 char *p = m->value.s;
2910 size_t plen = sizeof(m->value.s);
2911 char *origp = p;
2912 char *pmax = p + plen - 1;
2913 int c;
2914 int val;
2915 size_t bracket_nesting = 0;
2916
2917 while ((c = *s++) != '\0') {
2918 if (isspace(CAST(unsigned char, c)))
2919 break;
2920 if (p >= pmax) {
2921 file_error(ms, 0, "string too long: `%s'", origs);
2922 return NULL;
2923 }
2924 if (c != '\\') {
2925 if (c == '[') {
2926 bracket_nesting++;
2927 }
2928 if (c == ']' && bracket_nesting > 0) {
2929 bracket_nesting--;
2930 }
2931 *p++ = CAST(char, c);
2932 continue;
2933 }
2934 switch(c = *s++) {
2935
2936 case '\0':
2937 if (warn)
2938 file_magwarn(ms, "incomplete escape");
2939 s--;
2940 goto out;
2941 case '.':
2942 if (m->type == FILE_REGEX &&
2943 bracket_nesting == 0 && warn) {
2944 file_magwarn(ms, "escaped dot ('.') found, "
2945 "use \\\\. instead");
2946 }
2947 warn = 0; /* already did */
2948 /*FALLTHROUGH*/
2949 case '\t':
2950 if (warn) {
2951 file_magwarn(ms,
2952 "escaped tab found, use \\\\t instead");
2953 warn = 0; /* already did */
2954 }
2955 /*FALLTHROUGH*/
2956 default:
2957 if (warn) {
2958 if (isprint(CAST(unsigned char, c))) {
2959 /* Allow escaping of
2960 * ``relations'' */
2961 if (strchr("<>&^=!", c) == NULL
2962 && (m->type != FILE_REGEX ||
2963 strchr("[]().*?^$|{}", c)
2964 == NULL)) {
2965 file_magwarn(ms, "no "
2966 "need to escape "
2967 "`%c'", c);
2968 }
2969 } else {
2970 file_magwarn(ms,
2971 "unknown escape sequence: "
2972 "\\%03o", c);
2973 }
2974 }
2975 /*FALLTHROUGH*/
2976 /* space, perhaps force people to use \040? */
2977 case ' ':
2978 #if 0
2979 /*
2980 * Other things people escape, but shouldn't need to,
2981 * so we disallow them
2982 */
2983 case '\'':
2984 case '"':
2985 case '?':
2986 #endif
2987 /* Relations */
2988 case '>':
2989 case '<':
2990 case '&':
2991 case '^':
2992 case '=':
2993 case '!':
2994 /* and backslash itself */
2995 case '\\':
2996 *p++ = CAST(char, c);
2997 break;
2998
2999 case 'a':
3000 *p++ = '\a';
3001 break;
3002
3003 case 'b':
3004 *p++ = '\b';
3005 break;
3006
3007 case 'f':
3008 *p++ = '\f';
3009 break;
3010
3011 case 'n':
3012 *p++ = '\n';
3013 break;
3014
3015 case 'r':
3016 *p++ = '\r';
3017 break;
3018
3019 case 't':
3020 *p++ = '\t';
3021 break;
3022
3023 case 'v':
3024 *p++ = '\v';
3025 break;
3026
3027 /* \ and up to 3 octal digits */
3028 case '0':
3029 case '1':
3030 case '2':
3031 case '3':
3032 case '4':
3033 case '5':
3034 case '6':
3035 case '7':
3036 val = c - '0';
3037 c = *s++; /* try for 2 */
3038 if (c >= '0' && c <= '7') {
3039 val = (val << 3) | (c - '0');
3040 c = *s++; /* try for 3 */
3041 if (c >= '0' && c <= '7')
3042 val = (val << 3) | (c-'0');
3043 else
3044 --s;
3045 }
3046 else
3047 --s;
3048 *p++ = CAST(char, val);
3049 break;
3050
3051 /* \x and up to 2 hex digits */
3052 case 'x':
3053 val = 'x'; /* Default if no digits */
3054 c = hextoint(*s++); /* Get next char */
3055 if (c >= 0) {
3056 val = c;
3057 c = hextoint(*s++);
3058 if (c >= 0)
3059 val = (val << 4) + c;
3060 else
3061 --s;
3062 } else
3063 --s;
3064 *p++ = CAST(char, val);
3065 break;
3066 }
3067 }
3068 --s;
3069 out:
3070 *p = '\0';
3071 m->vallen = CAST(unsigned char, (p - origp));
3072 if (m->type == FILE_PSTRING) {
3073 size_t l = file_pstring_length_size(ms, m);
3074 if (l == FILE_BADSIZE)
3075 return NULL;
3076 m->vallen += CAST(unsigned char, l);
3077 }
3078 return s;
3079 }
3080
3081
3082 /* Single hex char to int; -1 if not a hex char. */
3083 private int
hextoint(int c)3084 hextoint(int c)
3085 {
3086 if (!isascii(CAST(unsigned char, c)))
3087 return -1;
3088 if (isdigit(CAST(unsigned char, c)))
3089 return c - '0';
3090 if ((c >= 'a') && (c <= 'f'))
3091 return c + 10 - 'a';
3092 if (( c>= 'A') && (c <= 'F'))
3093 return c + 10 - 'A';
3094 return -1;
3095 }
3096
3097
3098 /*
3099 * Print a string containing C character escapes.
3100 */
3101 protected void
file_showstr(FILE * fp,const char * s,size_t len)3102 file_showstr(FILE *fp, const char *s, size_t len)
3103 {
3104 char c;
3105
3106 for (;;) {
3107 if (len == FILE_BADSIZE) {
3108 c = *s++;
3109 if (c == '\0')
3110 break;
3111 }
3112 else {
3113 if (len-- == 0)
3114 break;
3115 c = *s++;
3116 }
3117 if (c >= 040 && c <= 0176) /* TODO isprint && !iscntrl */
3118 (void) fputc(c, fp);
3119 else {
3120 (void) fputc('\\', fp);
3121 switch (c) {
3122 case '\a':
3123 (void) fputc('a', fp);
3124 break;
3125
3126 case '\b':
3127 (void) fputc('b', fp);
3128 break;
3129
3130 case '\f':
3131 (void) fputc('f', fp);
3132 break;
3133
3134 case '\n':
3135 (void) fputc('n', fp);
3136 break;
3137
3138 case '\r':
3139 (void) fputc('r', fp);
3140 break;
3141
3142 case '\t':
3143 (void) fputc('t', fp);
3144 break;
3145
3146 case '\v':
3147 (void) fputc('v', fp);
3148 break;
3149
3150 default:
3151 (void) fprintf(fp, "%.3o", c & 0377);
3152 break;
3153 }
3154 }
3155 }
3156 }
3157
3158 /*
3159 * eatsize(): Eat the size spec from a number [eg. 10UL]
3160 */
3161 private void
eatsize(const char ** p)3162 eatsize(const char **p)
3163 {
3164 const char *l = *p;
3165
3166 if (LOWCASE(*l) == 'u')
3167 l++;
3168
3169 switch (LOWCASE(*l)) {
3170 case 'l': /* long */
3171 case 's': /* short */
3172 case 'h': /* short */
3173 case 'b': /* char/byte */
3174 case 'c': /* char/byte */
3175 l++;
3176 /*FALLTHROUGH*/
3177 default:
3178 break;
3179 }
3180
3181 *p = l;
3182 }
3183
3184 /*
3185 * handle a compiled file.
3186 */
3187
3188 private struct magic_map *
apprentice_map(struct magic_set * ms,const char * fn)3189 apprentice_map(struct magic_set *ms, const char *fn)
3190 {
3191 uint32_t *ptr;
3192 uint32_t version, entries = 0, nentries;
3193 int needsbyteswap;
3194 char *dbname = NULL;
3195 struct magic_map *map;
3196 size_t i;
3197 php_stream *stream = NULL;
3198 php_stream_statbuf st;
3199
3200
3201
3202 if ((map = CAST(struct magic_map *, ecalloc(1, sizeof(*map)))) == NULL) {
3203 file_oomem(ms, sizeof(*map));
3204 return NULL;
3205 }
3206
3207 if (fn == NULL) {
3208 map->p = (void *)&php_magic_database;
3209 goto internal_loaded;
3210 }
3211
3212 #ifdef PHP_WIN32
3213 /* Don't bother on windows with php_stream_open_wrapper,
3214 return to give apprentice_load() a chance. */
3215 if (php_stream_stat_path_ex((char *)fn, 0, &st, NULL) == SUCCESS) {
3216 if (st.sb.st_mode & S_IFDIR) {
3217 goto error;
3218 }
3219 }
3220 #endif
3221
3222 dbname = mkdbname(ms, fn, 0);
3223 if (dbname == NULL)
3224 goto error;
3225
3226 stream = php_stream_open_wrapper((char *)fn, "rb", REPORT_ERRORS, NULL);
3227
3228 if (!stream) {
3229 goto error;
3230 }
3231
3232 #ifndef PHP_WIN32
3233 if (php_stream_stat(stream, &st) < 0) {
3234 file_error(ms, errno, "cannot stat `%s'", dbname);
3235 goto error;
3236 }
3237 #endif
3238 if (st.sb.st_size < 8 || st.sb.st_size > maxoff_t()) {
3239 file_error(ms, 0, "file `%s' is too %s", dbname,
3240 st.sb.st_size < 8 ? "small" : "large");
3241 goto error;
3242 }
3243
3244 map->type = MAP_TYPE_MALLOC;
3245 map->len = CAST(size_t, st.sb.st_size);
3246 map->p = CAST(void *, emalloc(map->len));
3247
3248 if (php_stream_read(stream, map->p, (size_t)st.sb.st_size) != (size_t)st.sb.st_size) {
3249 file_badread(ms);
3250 goto error;
3251 }
3252
3253 php_stream_close(stream);
3254 stream = NULL;
3255
3256 internal_loaded:
3257 ptr = (uint32_t *)(void *)map->p;
3258 if (*ptr != MAGICNO) {
3259 if (swap4(*ptr) != MAGICNO) {
3260 file_error(ms, 0, "bad magic in `%s'", dbname);
3261 goto error;
3262 }
3263 needsbyteswap = 1;
3264 } else
3265 needsbyteswap = 0;
3266 if (needsbyteswap)
3267 version = swap4(ptr[1]);
3268 else
3269 version = ptr[1];
3270 if (version != VERSIONNO) {
3271 file_error(ms, 0, "File %d supports only version %d magic "
3272 "files. `%s' is version %d", MAGIC_VERSION,
3273 VERSIONNO, dbname, version);
3274 goto error;
3275 }
3276
3277 /* php_magic_database is a const, performing writes will segfault. This is for big-endian
3278 machines only, PPC and Sparc specifically. Consider static variable or MINIT in
3279 future. */
3280 if (needsbyteswap && fn == NULL) {
3281 map->p = emalloc(sizeof(php_magic_database));
3282 map->p = memcpy(map->p, php_magic_database, sizeof(php_magic_database));
3283 }
3284
3285 if (NULL != fn) {
3286 nentries = (uint32_t)(st.sb.st_size / sizeof(struct magic));
3287 entries = (uint32_t)(st.sb.st_size / sizeof(struct magic));
3288 if ((zend_off_t)(entries * sizeof(struct magic)) != st.sb.st_size) {
3289 file_error(ms, 0, "Size of `%s' %llu is not a multiple of %zu",
3290 dbname, (unsigned long long)st.sb.st_size,
3291 sizeof(struct magic));
3292 goto error;
3293 }
3294 }
3295 map->magic[0] = CAST(struct magic *, map->p) + 1;
3296 nentries = 0;
3297 for (i = 0; i < MAGIC_SETS; i++) {
3298 if (needsbyteswap)
3299 map->nmagic[i] = swap4(ptr[i + 2]);
3300 else
3301 map->nmagic[i] = ptr[i + 2];
3302 if (i != MAGIC_SETS - 1)
3303 map->magic[i + 1] = map->magic[i] + map->nmagic[i];
3304 nentries += map->nmagic[i];
3305 }
3306 if (NULL != fn && entries != nentries + 1) {
3307 file_error(ms, 0, "Inconsistent entries in `%s' %u != %u",
3308 dbname, entries, nentries + 1);
3309 goto error;
3310 }
3311 if (needsbyteswap)
3312 for (i = 0; i < MAGIC_SETS; i++)
3313 byteswap(map->magic[i], map->nmagic[i]);
3314
3315 if (dbname) {
3316 efree(dbname);
3317 }
3318 return map;
3319
3320 error:
3321 if (stream) {
3322 php_stream_close(stream);
3323 }
3324 apprentice_unmap(map);
3325 if (dbname) {
3326 efree(dbname);
3327 }
3328 return NULL;
3329 }
3330
3331 /*
3332 * handle an mmaped file.
3333 */
3334 private int
apprentice_compile(struct magic_set * ms,struct magic_map * map,const char * fn)3335 apprentice_compile(struct magic_set *ms, struct magic_map *map, const char *fn)
3336 {
3337 static const size_t nm = sizeof(*map->nmagic) * MAGIC_SETS;
3338 static const size_t m = sizeof(**map->magic);
3339 php_stream *stream;
3340 size_t len;
3341 char *dbname;
3342 int rv = -1;
3343 uint32_t i;
3344 union {
3345 struct magic m;
3346 uint32_t h[2 + MAGIC_SETS];
3347 } hdr;
3348
3349 dbname = mkdbname(ms, fn, 1);
3350
3351 if (dbname == NULL)
3352 goto out;
3353
3354 /* wb+ == O_WRONLY|O_CREAT|O_TRUNC|O_BINARY */
3355 stream = php_stream_open_wrapper((char *)fn, "wb+", REPORT_ERRORS, NULL);
3356
3357 if (!stream) {
3358 file_error(ms, errno, "cannot open `%s'", dbname);
3359 goto out;
3360 }
3361 memset(&hdr, 0, sizeof(hdr));
3362 hdr.h[0] = MAGICNO;
3363 hdr.h[1] = VERSIONNO;
3364 memcpy(hdr.h + 2, map->nmagic, nm);
3365
3366 if (php_stream_write(stream,(const char *)&hdr, sizeof(hdr)) != (ssize_t)sizeof(hdr)) {
3367 file_error(ms, errno, "error writing `%s'", dbname);
3368 goto out;
3369 }
3370
3371 for (i = 0; i < MAGIC_SETS; i++) {
3372 len = m * map->nmagic[i];
3373 if (php_stream_write(stream, (const char *)map->magic[i], len) != (ssize_t)len) {
3374 file_error(ms, errno, "error writing `%s'", dbname);
3375 goto out;
3376 }
3377 }
3378
3379 rv = 0;
3380 if (stream) {
3381 php_stream_close(stream);
3382 }
3383 out:
3384 efree(dbname);
3385 return rv;
3386 }
3387
3388 private const char ext[] = ".mgc";
3389 /*
3390 * make a dbname
3391 */
3392 private char *
mkdbname(struct magic_set * ms,const char * fn,int strip)3393 mkdbname(struct magic_set *ms, const char *fn, int strip)
3394 {
3395 const char *p, *q;
3396 char *buf;
3397
3398 if (strip) {
3399 if ((p = strrchr(fn, '/')) != NULL)
3400 fn = ++p;
3401 }
3402
3403 for (q = fn; *q; q++)
3404 continue;
3405 /* Look for .mgc */
3406 for (p = ext + sizeof(ext) - 1; p >= ext && q >= fn; p--, q--)
3407 if (*p != *q)
3408 break;
3409
3410 /* Did not find .mgc, restore q */
3411 if (p >= ext)
3412 while (*q)
3413 q++;
3414
3415 q++;
3416 /* Compatibility with old code that looked in .mime */
3417 if (ms->flags & MAGIC_MIME) {
3418 spprintf(&buf, MAXPATHLEN, "%.*s.mime%s", CAST(int, q - fn), fn, ext);
3419 #ifdef PHP_WIN32
3420 if (VCWD_ACCESS(buf, R_OK) == 0) {
3421 #else
3422 if (VCWD_ACCESS(buf, R_OK) != -1) {
3423 #endif
3424 ms->flags &= MAGIC_MIME_TYPE;
3425 return buf;
3426 }
3427 efree(buf);
3428 }
3429 spprintf(&buf, MAXPATHLEN, "%.*s%s", CAST(int, q - fn), fn, ext);
3430
3431 /* Compatibility with old code that looked in .mime */
3432 if (strstr(fn, ".mime") != NULL)
3433 ms->flags &= MAGIC_MIME_TYPE;
3434 return buf;
3435 }
3436
3437 /*
3438 * Byteswap an mmap'ed file if needed
3439 */
3440 private void
3441 byteswap(struct magic *magic, uint32_t nmagic)
3442 {
3443 uint32_t i;
3444 for (i = 0; i < nmagic; i++)
3445 bs1(&magic[i]);
3446 }
3447
3448 #if !defined(HAVE_BYTESWAP_H) && !defined(HAVE_SYS_BSWAP_H)
3449 /*
3450 * swap a short
3451 */
3452 private uint16_t
3453 swap2(uint16_t sv)
3454 {
3455 uint16_t rv;
3456 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv));
3457 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv));
3458 d[0] = s[1];
3459 d[1] = s[0];
3460 return rv;
3461 }
3462
3463 /*
3464 * swap an int
3465 */
3466 private uint32_t
3467 swap4(uint32_t sv)
3468 {
3469 uint32_t rv;
3470 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv));
3471 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv));
3472 d[0] = s[3];
3473 d[1] = s[2];
3474 d[2] = s[1];
3475 d[3] = s[0];
3476 return rv;
3477 }
3478
3479 /*
3480 * swap a quad
3481 */
3482 private uint64_t
3483 swap8(uint64_t sv)
3484 {
3485 uint64_t rv;
3486 uint8_t *s = RCAST(uint8_t *, RCAST(void *, &sv));
3487 uint8_t *d = RCAST(uint8_t *, RCAST(void *, &rv));
3488 # if 0
3489 d[0] = s[3];
3490 d[1] = s[2];
3491 d[2] = s[1];
3492 d[3] = s[0];
3493 d[4] = s[7];
3494 d[5] = s[6];
3495 d[6] = s[5];
3496 d[7] = s[4];
3497 # else
3498 d[0] = s[7];
3499 d[1] = s[6];
3500 d[2] = s[5];
3501 d[3] = s[4];
3502 d[4] = s[3];
3503 d[5] = s[2];
3504 d[6] = s[1];
3505 d[7] = s[0];
3506 # endif
3507 return rv;
3508 }
3509 #endif
3510
3511 protected uintmax_t
3512 file_varint2uintmax_t(const unsigned char *us, int t, size_t *l)
3513 {
3514 uintmax_t x = 0;
3515 const unsigned char *c;
3516 if (t == FILE_LEVARINT) {
3517 for (c = us; *c; c++) {
3518 if ((*c & 0x80) == 0)
3519 break;
3520 }
3521 if (l)
3522 *l = c - us + 1;
3523 for (; c >= us; c--) {
3524 x |= *c & 0x7f;
3525 x <<= 7;
3526 }
3527 } else {
3528 for (c = us; *c; c++) {
3529 x |= *c & 0x7f;
3530 if ((*c & 0x80) == 0)
3531 break;
3532 x <<= 7;
3533 }
3534 if (l)
3535 *l = c - us + 1;
3536 }
3537 return x;
3538 }
3539
3540
3541 /*
3542 * byteswap a single magic entry
3543 */
3544 private void
3545 bs1(struct magic *m)
3546 {
3547 m->cont_level = swap2(m->cont_level);
3548 m->offset = swap4(CAST(uint32_t, m->offset));
3549 m->in_offset = swap4(CAST(uint32_t, m->in_offset));
3550 m->lineno = swap4(CAST(uint32_t, m->lineno));
3551 if (IS_LIBMAGIC_STRING(m->type)) {
3552 m->str_range = swap4(m->str_range);
3553 m->str_flags = swap4(m->str_flags);
3554 }
3555 else {
3556 m->value.q = swap8(m->value.q);
3557 m->num_mask = swap8(m->num_mask);
3558 }
3559 }
3560
3561 protected size_t
3562 file_pstring_length_size(struct magic_set *ms, const struct magic *m)
3563 {
3564 switch (m->str_flags & PSTRING_LEN) {
3565 case PSTRING_1_LE:
3566 return 1;
3567 case PSTRING_2_LE:
3568 case PSTRING_2_BE:
3569 return 2;
3570 case PSTRING_4_LE:
3571 case PSTRING_4_BE:
3572 return 4;
3573 default:
3574 file_error(ms, 0, "corrupt magic file "
3575 "(bad pascal string length %d)",
3576 m->str_flags & PSTRING_LEN);
3577 return FILE_BADSIZE;
3578 }
3579 }
3580 protected size_t
3581 file_pstring_get_length(struct magic_set *ms, const struct magic *m,
3582 const char *ss)
3583 {
3584 size_t len = 0;
3585 const unsigned char *s = RCAST(const unsigned char *, ss);
3586 unsigned int s3, s2, s1, s0;
3587
3588 switch (m->str_flags & PSTRING_LEN) {
3589 case PSTRING_1_LE:
3590 len = *s;
3591 break;
3592 case PSTRING_2_LE:
3593 s0 = s[0];
3594 s1 = s[1];
3595 len = (s1 << 8) | s0;
3596 break;
3597 case PSTRING_2_BE:
3598 s0 = s[0];
3599 s1 = s[1];
3600 len = (s0 << 8) | s1;
3601 break;
3602 case PSTRING_4_LE:
3603 s0 = s[0];
3604 s1 = s[1];
3605 s2 = s[2];
3606 s3 = s[3];
3607 len = (s3 << 24) | (s2 << 16) | (s1 << 8) | s0;
3608 break;
3609 case PSTRING_4_BE:
3610 s0 = s[0];
3611 s1 = s[1];
3612 s2 = s[2];
3613 s3 = s[3];
3614 len = (s0 << 24) | (s1 << 16) | (s2 << 8) | s3;
3615 break;
3616 default:
3617 file_error(ms, 0, "corrupt magic file "
3618 "(bad pascal string length %d)",
3619 m->str_flags & PSTRING_LEN);
3620 return FILE_BADSIZE;
3621 }
3622
3623 if (m->str_flags & PSTRING_LENGTH_INCLUDES_ITSELF) {
3624 size_t l = file_pstring_length_size(ms, m);
3625 if (l == FILE_BADSIZE)
3626 return l;
3627 len -= l;
3628 }
3629
3630 return len;
3631 }
3632
3633 protected int
3634 file_magicfind(struct magic_set *ms, const char *name, struct mlist *v)
3635 {
3636 uint32_t i, j;
3637 struct mlist *mlist, *ml;
3638
3639 mlist = ms->mlist[1];
3640
3641 for (ml = mlist->next; ml != mlist; ml = ml->next) {
3642 struct magic *ma = ml->magic;
3643 for (i = 0; i < ml->nmagic; i++) {
3644 if (ma[i].type != FILE_NAME)
3645 continue;
3646 if (strcmp(ma[i].value.s, name) == 0) {
3647 v->magic = &ma[i];
3648 for (j = i + 1; j < ml->nmagic; j++)
3649 if (ma[j].cont_level == 0)
3650 break;
3651 v->nmagic = j - i;
3652 return 0;
3653 }
3654 }
3655 }
3656 return -1;
3657 }
3658
