1 /* $MirOS: src/bin/rm/rm.c,v 1.8 2010/09/21 21:24:02 tg Exp $ */
2 /* $NetBSD: rm.c,v 1.46 2007/06/24 17:59:31 christos Exp $ */
3 /* $OpenBSD: rm.c,v 1.18 2005/06/14 19:15:35 millert Exp $ */
4 
5 /*-
6  * Copyright (c) 1990, 1993, 1994, 2003
7  *	The Regents of the University of California.  All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 #include <sys/cdefs.h>
35 __COPYRIGHT("@(#) Copyright (c) 1990, 1993, 1994\n\
36 	The Regents of the University of California.  All rights reserved.\n");
37 __SCCSID("@(#)rm.c	8.8 (Berkeley) 4/27/95");
38 __RCSID("$NetBSD: rm.c,v 1.46 2007/06/24 17:59:31 christos Exp $");
39 __RCSID("$MirOS: src/bin/rm/rm.c,v 1.8 2010/09/21 21:24:02 tg Exp $");
40 
41 #include <sys/param.h>
42 #include <sys/stat.h>
43 #include <sys/mount.h>
44 
45 #include <err.h>
46 #include <errno.h>
47 #include <fcntl.h>
48 #include <fts.h>
49 #include <grp.h>
50 #include <libgen.h>
51 #include <locale.h>
52 #include <pwd.h>
53 #include <stdio.h>
54 #include <stdlib.h>
55 #include <string.h>
56 #include <unistd.h>
57 #include <pwd.h>
58 #include <grp.h>
59 
60 extern const char *__progname;
61 
62 int dflag, eval, fflag, iflag, Pflag, stdin_ok, vflag;
63 
64 int	check(char *, char *, struct stat *);
65 void	checkdot(char **);
66 void	rm_file(char **);
67 int	rm_overwrite(char *, struct stat *);
68 void	rm_tree(char **);
69 void	usage(void) __dead;
70 
71 /*
72  * For the sake of the `-f' flag, check whether an error number indicates the
73  * failure of an operation due to an non-existent file, either per se (ENOENT)
74  * or because its filename argument was illegal (ENAMETOOLONG, ENOTDIR).
75  */
76 #define NONEXISTENT(x) \
77     ((x) == ENOENT || (x) == ENAMETOOLONG || (x) == ENOTDIR)
78 
79 /*
80  * rm --
81  *	This rm is different from historic rms, but is expected to match
82  *	POSIX 1003.2 behavior.  The most visible difference is that -f
83  *	has two specific effects now, ignore non-existent files and force
84  *	file removal.
85  */
86 int
main(int argc,char * argv[])87 main(int argc, char *argv[])
88 {
89 	int ch, rflag;
90 
91 #ifndef __MirBSD__	/* irrelevant, as there is only one locale */
92 	setlocale(LC_ALL, "");
93 #endif
94 
95 	Pflag = rflag = 0;
96 	while ((ch = getopt(argc, argv, "dfiPRrv")) != -1)
97 		switch (ch) {
98 		case 'd':
99 			dflag = 1;
100 			break;
101 		case 'f':
102 			fflag = 1;
103 			iflag = 0;
104 			break;
105 		case 'i':
106 			fflag = 0;
107 			iflag = 1;
108 			break;
109 		case 'P':
110 			Pflag = 1;
111 			break;
112 		case 'R':
113 		case 'r':			/* Compatibility. */
114 			rflag = 1;
115 			break;
116 		case 'v':
117 			vflag = 1;
118 			break;
119 		default:
120 			usage();
121 		}
122 	argc -= optind;
123 	argv += optind;
124 
125 	if (argc < 1 && fflag == 0)
126 		usage();
127 
128 	checkdot(argv);
129 
130 	if (*argv) {
131 		stdin_ok = isatty(STDIN_FILENO);
132 
133 		if (rflag)
134 			rm_tree(argv);
135 		else
136 			rm_file(argv);
137 	}
138 
139 	exit(eval);
140 	/* NOTREACHED */
141 }
142 
143 #define TRYRENAME(fn, func, rv) do {				\
144 	char dname[MAXPATHLEN];					\
145 	size_t dcount = strlen(fn);				\
146 	uint64_t dhash = dcount * (intptr_t)(fn);		\
147 								\
148 	/* for fun: push back hash of original pathname */	\
149 	while (dcount--)					\
150 		dhash += (fn)[dcount] * dcount;			\
151 	arc4random_pushb_fast(&dhash, sizeof(dhash));		\
152 								\
153 	/* try to rename entry randomly before removal */	\
154 	do {							\
155 		if ((size_t)snprintf(dname, sizeof (dname),	\
156 		    "%s/rm.%08X", dirname(fn),			\
157 		    arc4random()) >= (sizeof(dname) + 5)) {	\
158 			/* resulting path would be too long */	\
159 			memcpy(dname, (fn), strlen(fn) + 1);	\
160 			break;					\
161 		}						\
162 	} while (rename((fn), dname));				\
163 	rv = func(dname);					\
164 } while (/* CONSTCOND */ 0)
165 
166 void
rm_tree(char ** argv)167 rm_tree(char **argv)
168 {
169 	FTS *fts;
170 	FTSENT *p;
171 	int flags, needstat, rval;
172 
173 	/*
174 	 * Remove a file hierarchy.  If forcing removal (-f), or interactive
175 	 * (-i) or can't ask anyway (stdin_ok), don't stat the file.
176 	 */
177 	needstat = !fflag && !iflag && stdin_ok;
178 
179 	/*
180 	 * If the -i option is specified, the user can skip on the pre-order
181 	 * visit.  The fts_number field flags skipped directories.
182 	 */
183 #define	SKIPPED	1
184 
185 	flags = FTS_PHYSICAL;
186 	if (!needstat)
187 		flags |= FTS_NOSTAT;
188 	if (!(fts = fts_open(argv, flags,
189 	    (int (*)(const FTSENT **, const FTSENT **))NULL)))
190 		err(1, NULL);
191 	while ((p = fts_read(fts)) != NULL) {
192 
193 		switch (p->fts_info) {
194 		case FTS_DNR:
195 			if (!fflag || p->fts_errno != ENOENT) {
196 				warnx("%s: %s", p->fts_path,
197 				    strerror(p->fts_errno));
198 				eval = 1;
199 			}
200 			continue;
201 		case FTS_ERR:
202 			errx(EXIT_FAILURE, "%s: %s", p->fts_path,
203 			    strerror(p->fts_errno));
204 			/* NOTREACHED */
205 		case FTS_NS:
206 			/*
207 			 * FTS_NS: assume that if can't stat the file, it
208 			 * can't be unlinked.
209 			 */
210 			if (fflag && NONEXISTENT(p->fts_errno))
211 				continue;
212 			if (needstat) {
213 				warnx("%s: %s", p->fts_path,
214 				    strerror(p->fts_errno));
215 				eval = 1;
216 				continue;
217 			}
218 			break;
219 		case FTS_D:
220 			/* Pre-order: give user chance to skip. */
221 			if (!fflag && !check(p->fts_path, p->fts_accpath,
222 			    p->fts_statp)) {
223 				(void)fts_set(fts, p, FTS_SKIP);
224 				p->fts_number = SKIPPED;
225 			}
226 			continue;
227 		case FTS_DP:
228 			/* Post-order: see if user skipped. */
229 			if (p->fts_number == SKIPPED)
230 				continue;
231 			break;
232 		default:
233 			if (!fflag &&
234 			    !check(p->fts_path, p->fts_accpath, p->fts_statp))
235 				continue;
236 		}
237 
238 		rval = 0;
239 		/*
240 		 * If we can't read or search the directory, may still be
241 		 * able to remove it.  Don't print out the un{read,search}able
242 		 * message unless the remove fails.
243 		 */
244 		switch (p->fts_info) {
245 		case FTS_DP:
246 		case FTS_DNR:
247 			if (Pflag)
248 				TRYRENAME(p->fts_accpath, rmdir, rval);
249 			else
250 				rval = rmdir(p->fts_accpath);
251 			if (rval != 0 && fflag && errno == ENOENT)
252 				continue;
253 			break;
254 
255 		default:
256 			if (Pflag) {
257 				if (rm_overwrite(p->fts_accpath, NULL))
258 					continue;
259 				TRYRENAME(p->fts_accpath, unlink, rval);
260 			} else
261 				rval = unlink(p->fts_accpath);
262 			if (rval != 0 && fflag && NONEXISTENT(errno))
263 				continue;
264 			break;
265 		}
266 		if (rval != 0) {
267 			warn("%s", p->fts_path);
268 			eval = 1;
269 		} else if (vflag)
270 			(void)printf("%s\n", p->fts_path);
271 	}
272 	if (errno)
273 		err(1, "fts_read");
274 	fts_close(fts);
275 }
276 
277 void
rm_file(char ** argv)278 rm_file(char **argv)
279 {
280 	struct stat sb;
281 	int rval;
282 	char *f;
283 
284 	/*
285 	 * Remove a file.  POSIX 1003.2 states that, by default, attempting
286 	 * to remove a directory is an error, so must always stat the file.
287 	 */
288 	while ((f = *argv++) != NULL) {
289 		/* Assume if can't stat the file, can't unlink it. */
290 		if (lstat(f, &sb)) {
291 			if (!fflag || !NONEXISTENT(errno)) {
292 				warn("%s", f);
293 				eval = 1;
294 			}
295 			continue;
296 		}
297 
298 		if (S_ISDIR(sb.st_mode) && !dflag) {
299 			warnx("%s: is a directory", f);
300 			eval = 1;
301 			continue;
302 		}
303 		if (!fflag && !check(f, f, &sb))
304 			continue;
305 		else if (S_ISDIR(sb.st_mode)) {
306 			if (Pflag)
307 				TRYRENAME(f, rmdir, rval);
308 			else
309 				rval = rmdir(f);
310 		} else {
311 			if (Pflag) {
312 				if (rm_overwrite(f, &sb))
313 					continue;
314 				TRYRENAME(f, unlink, rval);
315 			} else
316 				rval = unlink(f);
317 		}
318 		if (rval && (!fflag || !NONEXISTENT(errno))) {
319 			warn("%s", f);
320 			eval = 1;
321 		}
322 		if (vflag && rval == 0)
323 			(void)printf("%s\n", f);
324 	}
325 }
326 
327 /*
328  * rm_overwrite --
329  *	Overwrite the file 3 times with varying bit patterns.
330  *
331  * This is an expensive way to keep people from recovering files from your
332  * non-snapshotted FFS filesystems using fsdb(8).  Really.  No more.  Only
333  * regular files are deleted, directories will remain.
334  * However, names are no longer recoverable as any entries deleted with -P
335  * are renamed to entries with the basename “rm.XXXXXXXX” (where ‘X’en are
336  * generated randomly) in the same parent directory first if the length of
337  * the entire pathname (including that “/rm.XXXXXXXX”) is smaller than the
338  * maximum allowed pathname length, i.e. 1024 on MirBSD.
339  * Also, this assumes a fixed-block file system (like FFS, or a V7 or a
340  * System V file system).  In a logging file system, you'll have to have
341  * kernel support.
342  *
343  * A note on standards:  U.S. DoD 5220.22-M "National Industrial Security
344  * Program Operating Manual" ("NISPOM") is often cited as a reference
345  * for clearing and sanitizing magnetic media.  In fact, a matrix of
346  * "clearing" and "sanitization" methods for various media was given in
347  * Chapter 8 of the original 1995 version of NISPOM.  However, that
348  * matrix was *removed from the document* when Chapter 8 was rewritten
349  * in Change 2 to the document in 2001.  Recently, the Defense Security
350  * Service has made a revised clearing and sanitization matrix available
351  * in Microsoft Word format on the DSS web site.  The standardization
352  * status of this matrix is unclear.  Furthermore, one must be very
353  * careful when referring to this matrix: it is intended for the "clearing"
354  * prior to reuse or "sanitization" prior to disposal of *entire media*,
355  * not individual files and the only non-physically-destructive method of
356  * "sanitization" that is permitted for magnetic disks of any kind is
357  * specifically noted to be prohibited for media that have contained
358  * Top Secret data.
359  *
360  * It is impossible to actually conform to the exact procedure given in
361  * the matrix if one is overwriting a file, not an entire disk, because
362  * the procedure requires examination and comparison of the disk's defect
363  * lists.  Any program that claims to securely erase *files* while
364  * conforming to the standard, then, is not correct.  We do as much of
365  * what the standard requires as can actually be done when erasing a
366  * file, rather than an entire disk; but that does not make us conformant.
367  *
368  * Furthermore, the presence of track caches, disk and controller write
369  * caches, and so forth make it extremely difficult to ensure that data
370  * have actually been written to the disk, particularly when one tries
371  * to repeatedly overwrite the same sectors in quick succession.  We call
372  * fsync(), but controllers with nonvolatile cache, as well as IDE disks
373  * that just plain lie about the stable storage of data, will defeat this.
374  *
375  * Finally, widely respected research suggests that the given procedure
376  * is nowhere near sufficient to prevent the recovery of data using special
377  * forensic equipment and techniques that are well-known.  This is
378  * presumably one reason that the matrix requires physical media destruction,
379  * rather than any technique of the sort attempted here, for secret data.
380  *
381  * Caveat Emptor.
382  *
383  * rm_overwrite will return 0 on success.
384  */
385 int
rm_overwrite(char * file,struct stat * sbp)386 rm_overwrite(char *file, struct stat *sbp)
387 {
388 	struct statfs fsb;
389 	size_t bsize;
390 	char *buf = NULL;
391 
392 	struct stat sb;
393 	int fd;
394 	char randchar;
395 
396 	fd = -1;
397 	if (sbp == NULL) {
398 		if (lstat(file, &sb))
399 			goto err;
400 		sbp = &sb;
401 	}
402 	if (!S_ISREG(sbp->st_mode))
403 		return (0);
404 	if (sbp->st_nlink > 1) {
405 		warnx("%s (inode %u): not overwritten due to multiple links",
406 		    file, sbp->st_ino);
407 		/* if -f return success else failure */
408 		return (fflag ? 0 : 1);
409 	}
410 
411 	/* flags to try to defeat hidden caching by forcing seeks */
412 	if ((fd = open(file, O_RDWR|O_SYNC|O_RSYNC, 0)) == -1)
413 		goto err;
414 
415 	/* find out optimal transfer size */
416 	if (fstatfs(fd, &fsb) == -1)
417 		goto err;
418 	bsize = MAX(fsb.f_iosize, 131072U);
419 	if ((buf = malloc(bsize)) == NULL)
420 		err(1, "%s: malloc", file);
421 
422 #define RAND_BYTES	1
423 #define THIS_BYTE	0
424 
425 #define	WRITE_PASS(mode, byte, buf, bufsz) do {				\
426 	off_t len;							\
427 	size_t wlen, i;							\
428 	u_int32_t *qbuf = (u_int32_t *)buf;				\
429 									\
430 	if (fsync(fd) || lseek(fd, (off_t)0, SEEK_SET))			\
431 		goto err;						\
432 									\
433 	if (mode == THIS_BYTE)						\
434 		memset(buf, byte, bufsz);				\
435 	for (len = sbp->st_size; len > 0; len -= wlen) {		\
436 		wlen = MIN(len, bufsz);					\
437 		if (mode == RAND_BYTES)					\
438 			arc4random_buf(buf, wlen);			\
439 		if ((size_t)write(fd, buf, wlen) != wlen)		\
440 			goto err;					\
441 	}								\
442 	sync();		/* another poke at hidden caches */		\
443 } while (/* CONSTCOND */ 0)
444 
445 #define READ_PASS(byte, buf, bufsz) do {				\
446 	off_t len;							\
447 	size_t rlen;							\
448 	char pattern[bufsz];						\
449 									\
450 	if (fsync(fd) || lseek(fd, (off_t)0, SEEK_SET))			\
451 		goto err;						\
452 									\
453 	memset(pattern, byte, bufsz);					\
454 	for (len = sbp->st_size; len > 0; len -= rlen) {		\
455 		rlen = MIN(len, bufsz);					\
456 		if ((size_t)read(fd, buf, rlen) != rlen)		\
457 			goto err;					\
458 		if (memcmp(buf, pattern, rlen))				\
459 			goto err;					\
460 	}								\
461 	sync();		/* another poke at hidden caches */		\
462 } while (/* CONSTCOND */ 0)
463 
464 	/*
465 	 * DSS sanitization matrix "clear" for magnetic disks:
466 	 * option 'c' "Overwrite all addressable locations with a single
467 	 * character."
468 	 */
469 	randchar = (char)arc4random();
470 	WRITE_PASS(THIS_BYTE, randchar, buf, bsize);
471 
472 	/*
473 	 * DSS sanitization matrix "sanitize" for magnetic disks:
474 	 * option 'd', sub 2 "Overwrite all addressable locations with a
475 	 * character, then its complement.  Verify "complement" character
476 	 * was written successfully to all addressable locations, then
477 	 * overwrite all addressable locations with random characters; or
478 	 * verify third overwrite of random characters."  The rest of the
479 	 * text in d-sub-2 specifies requirements for overwriting spared
480 	 * sectors; we cannot conform to it when erasing only a file, thus
481 	 * we do not conform to the standard.
482 	 */
483 
484 	/* 1. "a character" */
485 	WRITE_PASS(THIS_BYTE, 0xAA, buf, bsize);
486 
487 	/* 2. "its complement" */
488 	WRITE_PASS(THIS_BYTE, 0x55, buf, bsize);
489 
490 	/* 3. "Verify 'complement' character" */
491 	READ_PASS(0x55, buf, bsize);
492 
493 	/* 4. "overwrite all addressable locations with random characters" */
494 
495 	WRITE_PASS(RAND_BYTES, 0, buf, bsize);
496 
497 	/*
498 	 * As the file might be huge, and we note that this revision of
499 	 * the matrix says "random characters", not "a random character"
500 	 * as the original did, we do not verify the random-character
501 	 * write; the "or" in the standard allows this.
502 	 */
503 
504 	if (buf != NULL) {
505 		free(buf);
506 		buf = NULL;
507 	}
508 
509 	if (close(fd) == -1) {
510 		fd = -1;
511 		goto err;
512 	}
513 
514 	return (0);
515 
516 err:	eval = 1;
517 	if (buf != NULL)
518 		free(buf);
519 	warn("%s", file);
520 	if (fd != -1)
521 		close(fd);
522 	return (1);
523 }
524 
525 int
check(char * path,char * name,struct stat * sp)526 check(char *path, char *name, struct stat *sp)
527 {
528 	int ch, first;
529 	char modep[15];
530 
531 	/* Check -i first. */
532 	if (iflag)
533 		(void)fprintf(stderr, "remove '%s'? ", path);
534 	else {
535 		/*
536 		 * If it's not a symbolic link and it's unwritable and we're
537 		 * talking to a terminal, ask.  Symbolic links are excluded
538 		 * because their permissions are meaningless.  Check stdin_ok
539 		 * first because we may not have stat'ed the file.
540 		 */
541 		if (!stdin_ok || S_ISLNK(sp->st_mode) ||
542 		    !(access(name, W_OK) && (errno != ETXTBSY)))
543 			return (1);
544 		strmode(sp->st_mode, modep);
545 		if (Pflag) {
546 			warnx(
547 			    "%s: -P was specified but file could not"
548 			    " be overwritten", path);
549 			return 0;
550 		}
551 		(void)fprintf(stderr, "override %s%s%s:%s for '%s'? ",
552 		    modep + 1, modep[9] == ' ' ? "" : " ",
553 		    user_from_uid(sp->st_uid, 0),
554 		    group_from_gid(sp->st_gid, 0), path);
555 	}
556 	(void)fflush(stderr);
557 
558 	first = ch = getchar();
559 	while (ch != '\n' && ch != EOF)
560 		ch = getchar();
561 	return (first == 'y' || first == 'Y');
562 }
563 
564 /*
565  * POSIX.2 requires that if "." or ".." are specified as the basename
566  * portion of an operand, a diagnostic message be written to standard
567  * error and nothing more be done with such operands.
568  *
569  * Since POSIX.2 defines basename as the final portion of a path after
570  * trailing slashes have been removed, we'll remove them here.
571  */
572 #define ISDOT(a)	((a)[0] == '.' && (!(a)[1] || ((a)[1] == '.' && !(a)[2])))
573 void
checkdot(char ** argv)574 checkdot(char **argv)
575 {
576 	char *p, **save, **t;
577 	int complained;
578 
579 	complained = 0;
580 	for (t = argv; *t;) {
581 		/* strip trailing slashes */
582 		p = strrchr(*t, '\0');
583 		while (--p > *t && *p == '/')
584 			*p = '\0';
585 
586 		/* extract basename */
587 		if ((p = strrchr(*t, '/')) != NULL)
588 			++p;
589 		else
590 			p = *t;
591 
592 		if (ISDOT(p)) {
593 			if (!complained++)
594 				warnx("\".\" and \"..\" may not be removed");
595 			eval = 1;
596 			for (save = t; (t[0] = t[1]) != NULL; ++t)
597 				continue;
598 			t = save;
599 		} else
600 			++t;
601 	}
602 }
603 
604 void
usage(void)605 usage(void)
606 {
607 	(void)fprintf(stderr, "usage: %s [-f|-i] [-dPRrv] file ...\n",
608 	    __progname);
609 	exit(1);
610 	/* NOTREACHED */
611 }
612