xref: /openbsd-src/usr.bin/file/file.c (revision f37ffad5a4aea53505e78b1619ed19b924c626d4)
1 /* $OpenBSD: file.c,v 1.55 2015/11/13 08:32:10 nicm Exp $ */
2 
3 /*
4  * Copyright (c) 2015 Nicholas Marriott <nicm@openbsd.org>
5  *
6  * Permission to use, copy, modify, and distribute this software for any
7  * purpose with or without fee is hereby granted, provided that the above
8  * copyright notice and this permission notice appear in all copies.
9  *
10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14  * WHATSOEVER RESULTING FROM LOSS OF MIND, USE, DATA OR PROFITS, WHETHER
15  * IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING
16  * OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17  */
18 
19 #include <sys/types.h>
20 #include <sys/ioctl.h>
21 #include <sys/mman.h>
22 #include <sys/socket.h>
23 #include <sys/queue.h>
24 #include <sys/uio.h>
25 #include <sys/wait.h>
26 
27 #include <errno.h>
28 #include <imsg.h>
29 #include <libgen.h>
30 #include <getopt.h>
31 #include <fcntl.h>
32 #include <pwd.h>
33 #include <stdlib.h>
34 #include <time.h>
35 #include <unistd.h>
36 #include <limits.h>
37 
38 #include "file.h"
39 #include "magic.h"
40 #include "xmalloc.h"
41 
42 struct input_msg
43 {
44 	int		idx;
45 
46 	struct stat	sb;
47 	int		error;
48 
49 	char		link_path[PATH_MAX];
50 	int		link_error;
51 	int		link_target;
52 };
53 
54 struct input_ack
55 {
56 	int		idx;
57 };
58 
59 struct input_file
60 {
61 	struct magic		*m;
62 	struct input_msg	*msg;
63 
64 	const char		*path;
65 	int			 fd;
66 
67 	void			*base;
68 	size_t			 size;
69 	int			 mapped;
70 	char			*result;
71 };
72 
73 extern char	*__progname;
74 
75 __dead void	 usage(void);
76 
77 static int	 prepare_message(struct input_msg *, int, const char *);
78 static void	 send_message(struct imsgbuf *, void *, size_t, int);
79 static int	 read_message(struct imsgbuf *, struct imsg *, pid_t);
80 
81 static void	 read_link(struct input_msg *, const char *);
82 
83 static __dead void child(int, pid_t, int, char **);
84 
85 static void	 test_file(struct input_file *, size_t);
86 
87 static int	 try_stat(struct input_file *);
88 static int	 try_empty(struct input_file *);
89 static int	 try_access(struct input_file *);
90 static int	 try_text(struct input_file *);
91 static int	 try_magic(struct input_file *);
92 static int	 try_unknown(struct input_file *);
93 
94 static int	 bflag;
95 static int	 cflag;
96 static int	 iflag;
97 static int	 Lflag;
98 static int	 sflag;
99 static int	 Wflag;
100 
101 static char	*magicpath;
102 static FILE	*magicfp;
103 
104 static struct option longopts[] = {
105 	{ "mime",      no_argument, NULL, 'i' },
106 	{ "mime-type", no_argument, NULL, 'i' },
107 	{ NULL,        0,           NULL, 0   }
108 };
109 
110 __dead void
111 usage(void)
112 {
113 	fprintf(stderr, "usage: %s [-bchiLsW] file ...\n", __progname);
114 	exit(1);
115 }
116 
117 int
118 main(int argc, char **argv)
119 {
120 	int			 opt, pair[2], fd, idx;
121 	char			*home;
122 	struct passwd		*pw;
123 	struct imsgbuf		 ibuf;
124 	struct imsg		 imsg;
125 	struct input_msg	 msg;
126 	struct input_ack	*ack;
127 	pid_t			 pid, parent;
128 
129 	tzset();
130 
131 	for (;;) {
132 		opt = getopt_long(argc, argv, "bchiLsW", longopts, NULL);
133 		if (opt == -1)
134 			break;
135 		switch (opt) {
136 		case 'b':
137 			bflag = 1;
138 			break;
139 		case 'c':
140 			cflag = 1;
141 			break;
142 		case 'h':
143 			Lflag = 0;
144 			break;
145 		case 'i':
146 			iflag = 1;
147 			break;
148 		case 'L':
149 			Lflag = 1;
150 			break;
151 		case 's':
152 			sflag = 1;
153 			break;
154 		case 'W':
155 			Wflag = 1;
156 			break;
157 		default:
158 			usage();
159 		}
160 	}
161 	argc -= optind;
162 	argv += optind;
163 	if (cflag) {
164 		if (argc != 0)
165 			usage();
166 	} else if (argc == 0)
167 		usage();
168 
169 	magicfp = NULL;
170 	if (geteuid() != 0 && !issetugid()) {
171 		home = getenv("HOME");
172 		if (home == NULL || *home == '\0') {
173 			pw = getpwuid(getuid());
174 			if (pw != NULL)
175 				home = pw->pw_dir;
176 			else
177 				home = NULL;
178 		}
179 		if (home != NULL) {
180 			xasprintf(&magicpath, "%s/.magic", home);
181 			magicfp = fopen(magicpath, "r");
182 			if (magicfp == NULL)
183 				free(magicpath);
184 		}
185 	}
186 	if (magicfp == NULL) {
187 		magicpath = xstrdup("/etc/magic");
188 		magicfp = fopen(magicpath, "r");
189 	}
190 	if (magicfp == NULL)
191 		err(1, "%s", magicpath);
192 
193 	parent = getpid();
194 	if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, pair) != 0)
195 		err(1, "socketpair");
196 	switch (pid = fork()) {
197 	case -1:
198 		err(1, "fork");
199 	case 0:
200 		close(pair[0]);
201 		child(pair[1], parent, argc, argv);
202 	}
203 	close(pair[1]);
204 
205 	fclose(magicfp);
206 	magicfp = NULL;
207 
208 	if (cflag)
209 		goto wait_for_child;
210 
211 	imsg_init(&ibuf, pair[0]);
212 	for (idx = 0; idx < argc; idx++) {
213 		fd = prepare_message(&msg, idx, argv[idx]);
214 		send_message(&ibuf, &msg, sizeof msg, fd);
215 
216 		if (read_message(&ibuf, &imsg, pid) == 0)
217 			break;
218 		if (imsg.hdr.len != IMSG_HEADER_SIZE + sizeof *ack)
219 			errx(1, "message too small");
220 		ack = imsg.data;
221 		if (ack->idx != idx)
222 			errx(1, "index not expected");
223 		imsg_free(&imsg);
224 	}
225 
226 wait_for_child:
227 	close(pair[0]);
228 	while (wait(NULL) == -1 && errno != ECHILD) {
229 		if (errno != EINTR)
230 			err(1, "wait");
231 	}
232 	_exit(0); /* let the child flush */
233 }
234 
235 static int
236 prepare_message(struct input_msg *msg, int idx, const char *path)
237 {
238 	int	fd, mode, error;
239 
240 	memset(msg, 0, sizeof *msg);
241 	msg->idx = idx;
242 
243 	if (strcmp(path, "-") == 0) {
244 		if (fstat(STDIN_FILENO, &msg->sb) == -1) {
245 			msg->error = errno;
246 			return (-1);
247 		}
248 		return (STDIN_FILENO);
249 	}
250 
251 	if (Lflag)
252 		error = stat(path, &msg->sb);
253 	else
254 		error = lstat(path, &msg->sb);
255 	if (error == -1) {
256 		msg->error = errno;
257 		return (-1);
258 	}
259 
260 	/*
261 	 * pledge(2) doesn't let us pass directory file descriptors around -
262 	 * but in fact we don't need them, so just don't open directories or
263 	 * symlinks (which could be to directories).
264 	 */
265 	mode = msg->sb.st_mode;
266 	if (!S_ISDIR(mode) && !S_ISLNK(mode)) {
267 		fd = open(path, O_RDONLY|O_NONBLOCK);
268 		if (fd == -1 && (errno == ENFILE || errno == EMFILE))
269 			err(1, "open");
270 	} else
271 		fd = -1;
272 	if (S_ISLNK(mode))
273 		read_link(msg, path);
274 	return (fd);
275 
276 }
277 
278 static void
279 send_message(struct imsgbuf *ibuf, void *msg, size_t msglen, int fd)
280 {
281 	if (imsg_compose(ibuf, -1, -1, 0, fd, msg, msglen) != 1)
282 		err(1, "imsg_compose");
283 	if (imsg_flush(ibuf) != 0)
284 		err(1, "imsg_flush");
285 }
286 
287 static int
288 read_message(struct imsgbuf *ibuf, struct imsg *imsg, pid_t from)
289 {
290 	int	n;
291 
292 	if ((n = imsg_read(ibuf)) == -1)
293 		err(1, "imsg_read");
294 	if (n == 0)
295 		return (0);
296 
297 	if ((n = imsg_get(ibuf, imsg)) == -1)
298 		err(1, "imsg_get");
299 	if (n == 0)
300 		return (0);
301 
302 	if ((pid_t)imsg->hdr.pid != from)
303 		errx(1, "PIDs don't match");
304 
305 	return (n);
306 
307 }
308 
309 static void
310 read_link(struct input_msg *msg, const char *path)
311 {
312 	struct stat	 sb;
313 	char		 lpath[PATH_MAX];
314 	char		*copy, *root;
315 	int		 used;
316 	ssize_t		 size;
317 
318 	size = readlink(path, lpath, sizeof lpath - 1);
319 	if (size == -1) {
320 		msg->link_error = errno;
321 		return;
322 	}
323 	lpath[size] = '\0';
324 
325 	if (*lpath == '/')
326 		strlcpy(msg->link_path, lpath, sizeof msg->link_path);
327 	else {
328 		copy = xstrdup(path);
329 
330 		root = dirname(copy);
331 		if (*root == '\0' || strcmp(root, ".") == 0 ||
332 		    strcmp (root, "/") == 0)
333 			strlcpy(msg->link_path, lpath, sizeof msg->link_path);
334 		else {
335 			used = snprintf(msg->link_path, sizeof msg->link_path,
336 			    "%s/%s", root, lpath);
337 			if (used < 0 || (size_t)used >= sizeof msg->link_path) {
338 				msg->link_error = ENAMETOOLONG;
339 				free(copy);
340 				return;
341 			}
342 		}
343 
344 		free(copy);
345 	}
346 
347 	if (!Lflag && stat(path, &sb) == -1)
348 		msg->link_target = errno;
349 }
350 
351 static __dead void
352 child(int fd, pid_t parent, int argc, char **argv)
353 {
354 	struct passwd		*pw;
355 	struct magic		*m;
356 	struct imsgbuf		 ibuf;
357 	struct imsg		 imsg;
358 	struct input_msg	*msg;
359 	struct input_ack	 ack;
360 	struct input_file	 inf;
361 	int			 i, idx;
362 	size_t			 len, width = 0;
363 
364 	if (pledge("stdio getpw recvfd id", NULL) == -1)
365 		err(1, "pledge");
366 
367 	if (geteuid() == 0) {
368 		pw = getpwnam(FILE_USER);
369 		if (pw == NULL)
370 			errx(1, "unknown user %s", FILE_USER);
371 		if (setgroups(1, &pw->pw_gid) != 0)
372 			err(1, "setgroups");
373 		if (setresgid(pw->pw_gid, pw->pw_gid, pw->pw_gid) != 0)
374 			err(1, "setresgid");
375 		if (setresuid(pw->pw_uid, pw->pw_uid, pw->pw_uid) != 0)
376 			err(1, "setresuid");
377 	}
378 
379 	if (pledge("stdio recvfd", NULL) == -1)
380 		err(1, "pledge");
381 
382 	m = magic_load(magicfp, magicpath, cflag || Wflag);
383 	if (cflag) {
384 		magic_dump(m);
385 		exit(0);
386 	}
387 
388 	for (i = 0; i < argc; i++) {
389 		len = strlen(argv[i]) + 1;
390 		if (len > width)
391 			width = len;
392 	}
393 
394 	imsg_init(&ibuf, fd);
395 	for (;;) {
396 		if (read_message(&ibuf, &imsg, parent) == 0)
397 			break;
398 		if (imsg.hdr.len != IMSG_HEADER_SIZE + sizeof *msg)
399 			errx(1, "message too small");
400 		msg = imsg.data;
401 
402 		idx = msg->idx;
403 		if (idx < 0 || idx >= argc)
404 			errx(1, "index out of range");
405 
406 		memset(&inf, 0, sizeof inf);
407 		inf.m = m;
408 		inf.msg = msg;
409 
410 		inf.path = argv[idx];
411 		inf.fd = imsg.fd;
412 
413 		test_file(&inf, width);
414 
415 		if (imsg.fd != -1)
416 			close(imsg.fd);
417 		imsg_free(&imsg);
418 
419 		ack.idx = idx;
420 		send_message(&ibuf, &ack, sizeof ack, -1);
421 	}
422 	exit(0);
423 }
424 
425 static void *
426 fill_buffer(int fd, size_t size, size_t *used)
427 {
428 	static void	*buffer;
429 	ssize_t		 got;
430 	size_t		 left;
431 	void		*next;
432 
433 	if (buffer == NULL)
434 		buffer = xmalloc(FILE_READ_SIZE);
435 
436 	next = buffer;
437 	left = size;
438 	while (left != 0) {
439 		got = read(fd, next, left);
440 		if (got == -1) {
441 			if (errno == EINTR)
442 				continue;
443 			return NULL;
444 		}
445 		if (got == 0)
446 			break;
447 		next = (char *)next + got;
448 		left -= got;
449 	}
450 	*used = size - left;
451 	return buffer;
452 }
453 
454 static int
455 load_file(struct input_file *inf)
456 {
457 	size_t	used;
458 
459 	if (inf->msg->sb.st_size == 0 && S_ISREG(inf->msg->sb.st_mode))
460 		return (0); /* empty file */
461 	if (inf->msg->sb.st_size == 0 || inf->msg->sb.st_size > FILE_READ_SIZE)
462 		inf->size = FILE_READ_SIZE;
463 	else
464 		inf->size = inf->msg->sb.st_size;
465 
466 	if (!S_ISREG(inf->msg->sb.st_mode))
467 		goto try_read;
468 
469 	inf->base = mmap(NULL, inf->size, PROT_READ, MAP_PRIVATE, inf->fd, 0);
470 	if (inf->base == MAP_FAILED)
471 		goto try_read;
472 	inf->mapped = 1;
473 	return (0);
474 
475 try_read:
476 	inf->base = fill_buffer(inf->fd, inf->size, &used);
477 	if (inf->base == NULL) {
478 		xasprintf(&inf->result, "cannot read '%s' (%s)", inf->path,
479 		    strerror(errno));
480 		return (1);
481 	}
482 	inf->size = used;
483 	return (0);
484 }
485 
486 static int
487 try_stat(struct input_file *inf)
488 {
489 	if (inf->msg->error != 0) {
490 		xasprintf(&inf->result, "cannot stat '%s' (%s)", inf->path,
491 		    strerror(inf->msg->error));
492 		return (1);
493 	}
494 	if (sflag || strcmp(inf->path, "-") == 0) {
495 		switch (inf->msg->sb.st_mode & S_IFMT) {
496 		case S_IFIFO:
497 			if (strcmp(inf->path, "-") != 0)
498 				break;
499 		case S_IFBLK:
500 		case S_IFCHR:
501 		case S_IFREG:
502 			return (0);
503 		}
504 	}
505 
506 	if (iflag && (inf->msg->sb.st_mode & S_IFMT) != S_IFREG) {
507 		xasprintf(&inf->result, "application/x-not-regular-file");
508 		return (1);
509 	}
510 
511 	switch (inf->msg->sb.st_mode & S_IFMT) {
512 	case S_IFDIR:
513 		xasprintf(&inf->result, "directory");
514 		return (1);
515 	case S_IFLNK:
516 		if (inf->msg->link_error != 0) {
517 			xasprintf(&inf->result, "unreadable symlink '%s' (%s)",
518 			    inf->path, strerror(inf->msg->link_error));
519 			return (1);
520 		}
521 		if (inf->msg->link_target == ELOOP)
522 			xasprintf(&inf->result, "symbolic link in a loop");
523 		else if (inf->msg->link_target != 0) {
524 			xasprintf(&inf->result, "broken symbolic link to '%s'",
525 			    inf->msg->link_path);
526 		} else {
527 			xasprintf(&inf->result, "symbolic link to '%s'",
528 			    inf->msg->link_path);
529 		}
530 		return (1);
531 	case S_IFSOCK:
532 		xasprintf(&inf->result, "socket");
533 		return (1);
534 	case S_IFBLK:
535 		xasprintf(&inf->result, "block special (%ld/%ld)",
536 		    (long)major(inf->msg->sb.st_rdev),
537 		    (long)minor(inf->msg->sb.st_rdev));
538 		return (1);
539 	case S_IFCHR:
540 		xasprintf(&inf->result, "character special (%ld/%ld)",
541 		    (long)major(inf->msg->sb.st_rdev),
542 		    (long)minor(inf->msg->sb.st_rdev));
543 		return (1);
544 	case S_IFIFO:
545 		xasprintf(&inf->result, "fifo (named pipe)");
546 		return (1);
547 	}
548 	return (0);
549 }
550 
551 static int
552 try_empty(struct input_file *inf)
553 {
554 	if (inf->size != 0)
555 		return (0);
556 
557 	if (iflag)
558 		xasprintf(&inf->result, "application/x-empty");
559 	else
560 		xasprintf(&inf->result, "empty");
561 	return (1);
562 }
563 
564 static int
565 try_access(struct input_file *inf)
566 {
567 	char tmp[256] = "";
568 
569 	if (inf->msg->sb.st_size == 0 && S_ISREG(inf->msg->sb.st_mode))
570 		return (0); /* empty file */
571 	if (inf->fd != -1)
572 		return (0);
573 
574 	if (inf->msg->sb.st_mode & (S_IWUSR|S_IWGRP|S_IWOTH))
575 		strlcat(tmp, "writable, ", sizeof tmp);
576 	if (inf->msg->sb.st_mode & (S_IXUSR|S_IXGRP|S_IXOTH))
577 		strlcat(tmp, "executable, ", sizeof tmp);
578 	if (S_ISREG(inf->msg->sb.st_mode))
579 		strlcat(tmp, "regular file, ", sizeof tmp);
580 	strlcat(tmp, "no read permission", sizeof tmp);
581 
582 	inf->result = xstrdup(tmp);
583 	return (1);
584 }
585 
586 static int
587 try_text(struct input_file *inf)
588 {
589 	const char	*type, *s;
590 	int		 flags;
591 
592 	flags = MAGIC_TEST_TEXT;
593 	if (iflag)
594 		flags |= MAGIC_TEST_MIME;
595 
596 	type = text_get_type(inf->base, inf->size);
597 	if (type == NULL)
598 		return (0);
599 
600 	s = magic_test(inf->m, inf->base, inf->size, flags);
601 	if (s != NULL) {
602 		inf->result = xstrdup(s);
603 		return (1);
604 	}
605 
606 	s = text_try_words(inf->base, inf->size, flags);
607 	if (s != NULL) {
608 		if (iflag)
609 			inf->result = xstrdup(s);
610 		else
611 			xasprintf(&inf->result, "%s %s text", type, s);
612 		return (1);
613 	}
614 
615 	if (iflag)
616 		inf->result = xstrdup("text/plain");
617 	else
618 		xasprintf(&inf->result, "%s text", type);
619 	return (1);
620 }
621 
622 static int
623 try_magic(struct input_file *inf)
624 {
625 	const char	*s;
626 	int		 flags;
627 
628 	flags = 0;
629 	if (iflag)
630 		flags |= MAGIC_TEST_MIME;
631 
632 	s = magic_test(inf->m, inf->base, inf->size, flags);
633 	if (s != NULL) {
634 		inf->result = xstrdup(s);
635 		return (1);
636 	}
637 	return (0);
638 }
639 
640 static int
641 try_unknown(struct input_file *inf)
642 {
643 	if (iflag)
644 		xasprintf(&inf->result, "application/x-not-regular-file");
645 	else
646 		xasprintf(&inf->result, "data");
647 	return (1);
648 }
649 
650 static void
651 test_file(struct input_file *inf, size_t width)
652 {
653 	char	*label;
654 	int	 stop;
655 
656 	stop = 0;
657 	if (!stop)
658 		stop = try_stat(inf);
659 	if (!stop)
660 		stop = try_access(inf);
661 	if (!stop)
662 		stop = load_file(inf);
663 	if (!stop)
664 		stop = try_empty(inf);
665 	if (!stop)
666 		stop = try_magic(inf);
667 	if (!stop)
668 		stop = try_text(inf);
669 	if (!stop)
670 		stop = try_unknown(inf);
671 
672 	if (bflag)
673 		printf("%s\n", inf->result);
674 	else {
675 		if (strcmp(inf->path, "-") == 0)
676 			xasprintf(&label, "/dev/stdin:");
677 		else
678 			xasprintf(&label, "%s:", inf->path);
679 		printf("%-*s %s\n", (int)width, label, inf->result);
680 		free(label);
681 	}
682 	free(inf->result);
683 
684 	if (inf->mapped && inf->base != NULL)
685 		munmap(inf->base, inf->size);
686 }
687