1 /* $NetBSD: rm.c,v 1.43 2006/03/17 23:55:42 peter Exp $ */ 2 3 /*- 4 * Copyright (c) 1990, 1993, 1994, 2003 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, 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 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 #ifndef lint 34 __COPYRIGHT("@(#) Copyright (c) 1990, 1993, 1994\n\ 35 The Regents of the University of California. All rights reserved.\n"); 36 #endif /* not lint */ 37 38 #ifndef lint 39 #if 0 40 static char sccsid[] = "@(#)rm.c 8.8 (Berkeley) 4/27/95"; 41 #else 42 __RCSID("$NetBSD: rm.c,v 1.43 2006/03/17 23:55:42 peter Exp $"); 43 #endif 44 #endif /* not lint */ 45 46 #include <sys/param.h> 47 #include <sys/stat.h> 48 #include <sys/types.h> 49 50 #include <err.h> 51 #include <errno.h> 52 #include <fcntl.h> 53 #include <fts.h> 54 #include <grp.h> 55 #include <locale.h> 56 #include <pwd.h> 57 #include <stdio.h> 58 #include <stdlib.h> 59 #include <string.h> 60 #include <unistd.h> 61 62 int dflag, eval, fflag, iflag, Pflag, stdin_ok, vflag, Wflag; 63 64 int check(char *, char *, struct stat *); 65 void checkdot(char **); 66 void rm_file(char **); 67 void rm_overwrite(char *, struct stat *); 68 void rm_tree(char **); 69 void usage(void); 70 int main(int, char *[]); 71 72 /* 73 * For the sake of the `-f' flag, check whether an error number indicates the 74 * failure of an operation due to an non-existent file, either per se (ENOENT) 75 * or because its filename argument was illegal (ENAMETOOLONG, ENOTDIR). 76 */ 77 #define NONEXISTENT(x) \ 78 ((x) == ENOENT || (x) == ENAMETOOLONG || (x) == ENOTDIR) 79 80 /* 81 * rm -- 82 * This rm is different from historic rm's, but is expected to match 83 * POSIX 1003.2 behavior. The most visible difference is that -f 84 * has two specific effects now, ignore non-existent files and force 85 * file removal. 86 */ 87 int 88 main(int argc, char *argv[]) 89 { 90 int ch, rflag; 91 92 setprogname(argv[0]); 93 (void)setlocale(LC_ALL, ""); 94 95 Pflag = rflag = 0; 96 while ((ch = getopt(argc, argv, "dfiPRrvW")) != -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 case 'W': 120 Wflag = 1; 121 break; 122 case '?': 123 default: 124 usage(); 125 } 126 argc -= optind; 127 argv += optind; 128 129 if (argc < 1) 130 usage(); 131 132 checkdot(argv); 133 134 if (*argv) { 135 stdin_ok = isatty(STDIN_FILENO); 136 137 if (rflag) 138 rm_tree(argv); 139 else 140 rm_file(argv); 141 } 142 143 exit(eval); 144 /* NOTREACHED */ 145 } 146 147 void 148 rm_tree(char **argv) 149 { 150 FTS *fts; 151 FTSENT *p; 152 int flags, needstat, rval; 153 154 /* 155 * Remove a file hierarchy. If forcing removal (-f), or interactive 156 * (-i) or can't ask anyway (stdin_ok), don't stat the file. 157 */ 158 needstat = !fflag && !iflag && stdin_ok; 159 160 /* 161 * If the -i option is specified, the user can skip on the pre-order 162 * visit. The fts_number field flags skipped directories. 163 */ 164 #define SKIPPED 1 165 166 flags = FTS_PHYSICAL; 167 if (!needstat) 168 flags |= FTS_NOSTAT; 169 if (Wflag) 170 flags |= FTS_WHITEOUT; 171 if (!(fts = fts_open(argv, flags, 172 (int (*)(const FTSENT **, const FTSENT **))NULL))) 173 err(1, NULL); 174 while ((p = fts_read(fts)) != NULL) { 175 176 switch (p->fts_info) { 177 case FTS_DNR: 178 if (!fflag || p->fts_errno != ENOENT) { 179 warnx("%s: %s", p->fts_path, 180 strerror(p->fts_errno)); 181 eval = 1; 182 } 183 continue; 184 case FTS_ERR: 185 errx(EXIT_FAILURE, "%s: %s", p->fts_path, 186 strerror(p->fts_errno)); 187 /* NOTREACHED */ 188 case FTS_NS: 189 /* 190 * FTS_NS: assume that if can't stat the file, it 191 * can't be unlinked. 192 */ 193 if (fflag && NONEXISTENT(p->fts_errno)) 194 continue; 195 if (needstat) { 196 warnx("%s: %s", p->fts_path, 197 strerror(p->fts_errno)); 198 eval = 1; 199 continue; 200 } 201 break; 202 case FTS_D: 203 /* Pre-order: give user chance to skip. */ 204 if (!fflag && !check(p->fts_path, p->fts_accpath, 205 p->fts_statp)) { 206 (void)fts_set(fts, p, FTS_SKIP); 207 p->fts_number = SKIPPED; 208 } 209 continue; 210 case FTS_DP: 211 /* Post-order: see if user skipped. */ 212 if (p->fts_number == SKIPPED) 213 continue; 214 break; 215 default: 216 if (!fflag && 217 !check(p->fts_path, p->fts_accpath, p->fts_statp)) 218 continue; 219 } 220 221 rval = 0; 222 /* 223 * If we can't read or search the directory, may still be 224 * able to remove it. Don't print out the un{read,search}able 225 * message unless the remove fails. 226 */ 227 switch (p->fts_info) { 228 case FTS_DP: 229 case FTS_DNR: 230 rval = rmdir(p->fts_accpath); 231 if (rval != 0 && fflag && errno == ENOENT) 232 continue; 233 break; 234 235 case FTS_W: 236 rval = undelete(p->fts_accpath); 237 if (rval != 0 && fflag && errno == ENOENT) 238 continue; 239 break; 240 241 default: 242 if (Pflag) 243 rm_overwrite(p->fts_accpath, NULL); 244 rval = unlink(p->fts_accpath); 245 if (rval != 0 && fflag && NONEXISTENT(errno)) 246 continue; 247 break; 248 } 249 if (rval != 0) { 250 warn("%s", p->fts_path); 251 eval = 1; 252 } else if (vflag) 253 (void)printf("%s\n", p->fts_path); 254 } 255 if (errno) 256 err(1, "fts_read"); 257 fts_close(fts); 258 } 259 260 void 261 rm_file(char **argv) 262 { 263 struct stat sb; 264 int rval; 265 char *f; 266 267 /* 268 * Remove a file. POSIX 1003.2 states that, by default, attempting 269 * to remove a directory is an error, so must always stat the file. 270 */ 271 while ((f = *argv++) != NULL) { 272 /* Assume if can't stat the file, can't unlink it. */ 273 if (lstat(f, &sb)) { 274 if (Wflag) { 275 sb.st_mode = S_IFWHT|S_IWUSR|S_IRUSR; 276 } else { 277 if (!fflag || !NONEXISTENT(errno)) { 278 warn("%s", f); 279 eval = 1; 280 } 281 continue; 282 } 283 } else if (Wflag) { 284 warnx("%s: %s", f, strerror(EEXIST)); 285 eval = 1; 286 continue; 287 } 288 289 if (S_ISDIR(sb.st_mode) && !dflag) { 290 warnx("%s: is a directory", f); 291 eval = 1; 292 continue; 293 } 294 if (!fflag && !S_ISWHT(sb.st_mode) && !check(f, f, &sb)) 295 continue; 296 if (S_ISWHT(sb.st_mode)) 297 rval = undelete(f); 298 else if (S_ISDIR(sb.st_mode)) 299 rval = rmdir(f); 300 else { 301 if (Pflag) 302 rm_overwrite(f, &sb); 303 rval = unlink(f); 304 } 305 if (rval && (!fflag || !NONEXISTENT(errno))) { 306 warn("%s", f); 307 eval = 1; 308 } 309 if (vflag && rval == 0) 310 (void)printf("%s\n", f); 311 } 312 } 313 314 /* 315 * rm_overwrite -- 316 * Overwrite the file 3 times with varying bit patterns. 317 * 318 * This is an expensive way to keep people from recovering files from your 319 * non-snapshotted FFS filesystems using fsdb(8). Really. No more. Only 320 * regular files are deleted, directories (and therefore names) will remain. 321 * Also, this assumes a fixed-block file system (like FFS, or a V7 or a 322 * System V file system). In a logging file system, you'll have to have 323 * kernel support. 324 * 325 * A note on standards: U.S. DoD 5220.22-M "National Industrial Security 326 * Program Operating Manual" ("NISPOM") is often cited as a reference 327 * for clearing and sanitizing magnetic media. In fact, a matrix of 328 * "clearing" and "sanitization" methods for various media was given in 329 * Chapter 8 of the original 1995 version of NISPOM. However, that 330 * matrix was *removed from the document* when Chapter 8 was rewritten 331 * in Change 2 to the document in 2001. Recently, the Defense Security 332 * Service has made a revised clearing and sanitization matrix available 333 * in Microsoft Word format on the DSS web site. The standardization 334 * status of this matrix is unclear. Furthermore, one must be very 335 * careful when referring to this matrix: it is intended for the "clearing" 336 * prior to reuse or "sanitization" prior to disposal of *entire media*, 337 * not individual files and the only non-physically-destructive method of 338 * "sanitization" that is permitted for magnetic disks of any kind is 339 * specifically noted to be prohibited for media that have contained 340 * Top Secret data. 341 * 342 * It is impossible to actually conform to the exact procedure given in 343 * the matrix if one is overwriting a file, not an entire disk, because 344 * the procedure requires examination and comparison of the disk's defect 345 * lists. Any program that claims to securely erase *files* while 346 * conforming to the standard, then, is not correct. We do everything 347 * 348 * Furthermore, the presence of track caches, disk and controller write 349 * caches, and so forth make it extremely difficult to ensure that data 350 * have actually been written to the disk, particularly when one tries 351 * to repeatedly overwrite the same sectors in quick succession. We call 352 * fsync(), but controllers with nonvolatile cache, as well as IDE disks 353 * that just plain lie about the stable storage of data, will defeat this. 354 * 355 * Finally, widely respected research suggests that the given procedure 356 * is nowhere near sufficient to prevent the recovery of data using special 357 * forensic equipment and techniques that are well-known. This is 358 * presumably one reason that the matrix requires physical media destruction, 359 * rather than any technique of the sort attempted here, for secret data. 360 * 361 * Caveat Emptor. 362 */ 363 364 void 365 rm_overwrite(char *file, struct stat *sbp) 366 { 367 struct stat sb; 368 int fd, randint; 369 char randchar; 370 371 fd = -1; 372 if (sbp == NULL) { 373 if (lstat(file, &sb)) 374 goto err; 375 sbp = &sb; 376 } 377 if (!S_ISREG(sbp->st_mode)) 378 return; 379 380 /* flags to try to defeat hidden caching by forcing seeks */ 381 if ((fd = open(file, O_RDWR|O_SYNC|O_RSYNC, 0)) == -1) 382 goto err; 383 384 #define RAND_BYTES 1 385 #define THIS_BYTE 0 386 387 #define WRITE_PASS(mode, byte) do { \ 388 off_t len; \ 389 int wlen, i; \ 390 char buf[8 * 1024]; \ 391 \ 392 if (fsync(fd) || lseek(fd, (off_t)0, SEEK_SET)) \ 393 goto err; \ 394 \ 395 if (mode == THIS_BYTE) \ 396 memset(buf, byte, sizeof(buf)); \ 397 for (len = sbp->st_size; len > 0; len -= wlen) { \ 398 if (mode == RAND_BYTES) { \ 399 for (i = 0; i < sizeof(buf); \ 400 i+= sizeof(u_int32_t)) \ 401 *(int *)(buf + i) = arc4random(); \ 402 } \ 403 wlen = len < sizeof(buf) ? len : sizeof(buf); \ 404 if (write(fd, buf, wlen) != wlen) \ 405 goto err; \ 406 } \ 407 sync(); /* another poke at hidden caches */ \ 408 } while (/* CONSTCOND */ 0) 409 410 #define READ_PASS(byte) do { \ 411 off_t len; \ 412 int rlen; \ 413 char pattern[8 * 1024]; \ 414 char buf[8 * 1024]; \ 415 \ 416 if (fsync(fd) || lseek(fd, (off_t)0, SEEK_SET)) \ 417 goto err; \ 418 \ 419 memset(pattern, byte, sizeof(pattern)); \ 420 for(len = sbp->st_size; len > 0; len -= rlen) { \ 421 rlen = len < sizeof(buf) ? len : sizeof(buf); \ 422 if(read(fd, buf, rlen) != rlen) \ 423 goto err; \ 424 if(memcmp(buf, pattern, rlen)) \ 425 goto err; \ 426 } \ 427 sync(); /* another poke at hidden caches */ \ 428 } while (/* CONSTCOND */ 0) 429 430 /* 431 * DSS sanitization matrix "clear" for magnetic disks: 432 * option 'c' "Overwrite all addressable locations with a single 433 * character." 434 */ 435 randint = arc4random(); 436 randchar = *(char *)&randint; 437 WRITE_PASS(THIS_BYTE, randchar); 438 439 /* 440 * DSS sanitization matrix "sanitize" for magnetic disks: 441 * option 'd', sub 2 "Overwrite all addressable locations with a 442 * character, then its complement. Verify "complement" character 443 * was written successfully to all addressable locations, then 444 * overwrite all addressable locations with random characters; or 445 * verify third overwrite of random characters." The rest of the 446 * text in d-sub-2 specifies requirements for overwriting spared 447 * sectors; we cannot conform to it when erasing only a file, thus 448 * we do not conform to the standard. 449 */ 450 451 /* 1. "a character" */ 452 WRITE_PASS(THIS_BYTE, 0xff); 453 454 /* 2. "its complement" */ 455 WRITE_PASS(THIS_BYTE, 0x00); 456 457 /* 3. "Verify 'complement' character" */ 458 READ_PASS(0x00); 459 460 /* 4. "overwrite all addressable locations with random characters" */ 461 462 WRITE_PASS(RAND_BYTES, 0x00); 463 464 /* 465 * As the file might be huge, and we note that this revision of 466 * the matrix says "random characters", not "a random character" 467 * as the original did, we do not verify the random-character 468 * write; the "or" in the standard allows this. 469 */ 470 471 if (!close(fd)) 472 return; 473 474 err: eval = 1; 475 warn("%s", file); 476 } 477 478 int 479 check(char *path, char *name, struct stat *sp) 480 { 481 int ch, first; 482 char modep[15]; 483 484 /* Check -i first. */ 485 if (iflag) 486 (void)fprintf(stderr, "remove '%s'? ", path); 487 else { 488 /* 489 * If it's not a symbolic link and it's unwritable and we're 490 * talking to a terminal, ask. Symbolic links are excluded 491 * because their permissions are meaningless. Check stdin_ok 492 * first because we may not have stat'ed the file. 493 */ 494 if (!stdin_ok || S_ISLNK(sp->st_mode) || 495 !(access(name, W_OK) && (errno != ETXTBSY))) 496 return (1); 497 strmode(sp->st_mode, modep); 498 (void)fprintf(stderr, "override %s%s%s/%s for '%s'? ", 499 modep + 1, modep[9] == ' ' ? "" : " ", 500 user_from_uid(sp->st_uid, 0), 501 group_from_gid(sp->st_gid, 0), path); 502 } 503 (void)fflush(stderr); 504 505 first = ch = getchar(); 506 while (ch != '\n' && ch != EOF) 507 ch = getchar(); 508 return (first == 'y' || first == 'Y'); 509 } 510 511 /* 512 * POSIX.2 requires that if "." or ".." are specified as the basename 513 * portion of an operand, a diagnostic message be written to standard 514 * error and nothing more be done with such operands. 515 * 516 * Since POSIX.2 defines basename as the final portion of a path after 517 * trailing slashes have been removed, we'll remove them here. 518 */ 519 #define ISDOT(a) ((a)[0] == '.' && (!(a)[1] || ((a)[1] == '.' && !(a)[2]))) 520 void 521 checkdot(char **argv) 522 { 523 char *p, **save, **t; 524 int complained; 525 526 complained = 0; 527 for (t = argv; *t;) { 528 /* strip trailing slashes */ 529 p = strrchr(*t, '\0'); 530 while (--p > *t && *p == '/') 531 *p = '\0'; 532 533 /* extract basename */ 534 if ((p = strrchr(*t, '/')) != NULL) 535 ++p; 536 else 537 p = *t; 538 539 if (ISDOT(p)) { 540 if (!complained++) 541 warnx("\".\" and \"..\" may not be removed"); 542 eval = 1; 543 for (save = t; (t[0] = t[1]) != NULL; ++t) 544 continue; 545 t = save; 546 } else 547 ++t; 548 } 549 } 550 551 void 552 usage(void) 553 { 554 555 (void)fprintf(stderr, "usage: %s [-f|-i] [-dPRrvW] file ...\n", 556 getprogname()); 557 exit(1); 558 /* NOTREACHED */ 559 } 560