1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1988, 1993, 1994 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * David Hitz of Auspex Systems Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 */ 34 35 #if 0 36 #ifndef lint 37 static char const copyright[] = 38 "@(#) Copyright (c) 1988, 1993, 1994\n\ 39 The Regents of the University of California. All rights reserved.\n"; 40 #endif /* not lint */ 41 42 #endif 43 #include <sys/cdefs.h> 44 /* 45 * Cp copies source files to target files. 46 * 47 * The global PATH_T structure "to" always contains the path to the 48 * current target file. Since fts(3) does not change directories, 49 * this path can be either absolute or dot-relative. 50 * 51 * The basic algorithm is to initialize "to" and use fts(3) to traverse 52 * the file hierarchy rooted in the argument list. A trivial case is the 53 * case of 'cp file1 file2'. The more interesting case is the case of 54 * 'cp file1 file2 ... fileN dir' where the hierarchy is traversed and the 55 * path (relative to the root of the traversal) is appended to dir (stored 56 * in "to") to form the final target path. 57 */ 58 59 #include <sys/types.h> 60 #include <sys/stat.h> 61 62 #include <assert.h> 63 #include <err.h> 64 #include <errno.h> 65 #include <fts.h> 66 #include <limits.h> 67 #include <signal.h> 68 #include <stdio.h> 69 #include <stdlib.h> 70 #include <string.h> 71 #include <unistd.h> 72 73 #include "extern.h" 74 75 #define STRIP_TRAILING_SLASH(p) { \ 76 while ((p).p_end > (p).p_path + 1 && (p).p_end[-1] == '/') \ 77 *--(p).p_end = 0; \ 78 } 79 80 static char emptystring[] = ""; 81 82 PATH_T to = { to.p_path, emptystring, "" }; 83 84 int fflag, iflag, lflag, nflag, pflag, sflag, vflag; 85 static int Hflag, Lflag, Rflag, rflag; 86 volatile sig_atomic_t info; 87 88 enum op { FILE_TO_FILE, FILE_TO_DIR, DIR_TO_DNE }; 89 90 static int copy(char *[], enum op, int, struct stat *); 91 static void siginfo(int __unused); 92 93 int 94 main(int argc, char *argv[]) 95 { 96 struct stat to_stat, tmp_stat; 97 enum op type; 98 int Pflag, ch, fts_options, r, have_trailing_slash; 99 char *target; 100 101 fts_options = FTS_NOCHDIR | FTS_PHYSICAL; 102 Pflag = 0; 103 while ((ch = getopt(argc, argv, "HLPRafilnprsvx")) != -1) 104 switch (ch) { 105 case 'H': 106 Hflag = 1; 107 Lflag = Pflag = 0; 108 break; 109 case 'L': 110 Lflag = 1; 111 Hflag = Pflag = 0; 112 break; 113 case 'P': 114 Pflag = 1; 115 Hflag = Lflag = 0; 116 break; 117 case 'R': 118 Rflag = 1; 119 break; 120 case 'a': 121 pflag = 1; 122 Rflag = 1; 123 Pflag = 1; 124 Hflag = Lflag = 0; 125 break; 126 case 'f': 127 fflag = 1; 128 iflag = nflag = 0; 129 break; 130 case 'i': 131 iflag = 1; 132 fflag = nflag = 0; 133 break; 134 case 'l': 135 lflag = 1; 136 break; 137 case 'n': 138 nflag = 1; 139 fflag = iflag = 0; 140 break; 141 case 'p': 142 pflag = 1; 143 break; 144 case 'r': 145 rflag = Lflag = 1; 146 Hflag = Pflag = 0; 147 break; 148 case 's': 149 sflag = 1; 150 break; 151 case 'v': 152 vflag = 1; 153 break; 154 case 'x': 155 fts_options |= FTS_XDEV; 156 break; 157 default: 158 usage(); 159 break; 160 } 161 argc -= optind; 162 argv += optind; 163 164 if (argc < 2) 165 usage(); 166 167 if (Rflag && rflag) 168 errx(1, "the -R and -r options may not be specified together"); 169 if (lflag && sflag) 170 errx(1, "the -l and -s options may not be specified together"); 171 if (rflag) 172 Rflag = 1; 173 if (Rflag) { 174 if (Hflag) 175 fts_options |= FTS_COMFOLLOW; 176 if (Lflag) { 177 fts_options &= ~FTS_PHYSICAL; 178 fts_options |= FTS_LOGICAL; 179 } 180 } else if (!Pflag) { 181 fts_options &= ~FTS_PHYSICAL; 182 fts_options |= FTS_LOGICAL | FTS_COMFOLLOW; 183 } 184 (void)signal(SIGINFO, siginfo); 185 186 /* Save the target base in "to". */ 187 target = argv[--argc]; 188 if (strlcpy(to.p_path, target, sizeof(to.p_path)) >= sizeof(to.p_path)) 189 errx(1, "%s: name too long", target); 190 to.p_end = to.p_path + strlen(to.p_path); 191 if (to.p_path == to.p_end) { 192 *to.p_end++ = '.'; 193 *to.p_end = 0; 194 } 195 have_trailing_slash = (to.p_end[-1] == '/'); 196 if (have_trailing_slash) 197 STRIP_TRAILING_SLASH(to); 198 to.target_end = to.p_end; 199 200 /* Set end of argument list for fts(3). */ 201 argv[argc] = NULL; 202 203 /* 204 * Cp has two distinct cases: 205 * 206 * cp [-R] source target 207 * cp [-R] source1 ... sourceN directory 208 * 209 * In both cases, source can be either a file or a directory. 210 * 211 * In (1), the target becomes a copy of the source. That is, if the 212 * source is a file, the target will be a file, and likewise for 213 * directories. 214 * 215 * In (2), the real target is not directory, but "directory/source". 216 */ 217 r = stat(to.p_path, &to_stat); 218 if (r == -1 && errno != ENOENT) 219 err(1, "%s", to.p_path); 220 if (r == -1 || !S_ISDIR(to_stat.st_mode)) { 221 /* 222 * Case (1). Target is not a directory. 223 */ 224 if (argc > 1) 225 errx(1, "%s is not a directory", to.p_path); 226 227 /* 228 * Need to detect the case: 229 * cp -R dir foo 230 * Where dir is a directory and foo does not exist, where 231 * we want pathname concatenations turned on but not for 232 * the initial mkdir(). 233 */ 234 if (r == -1) { 235 if (Rflag && (Lflag || Hflag)) 236 stat(*argv, &tmp_stat); 237 else 238 lstat(*argv, &tmp_stat); 239 240 if (S_ISDIR(tmp_stat.st_mode) && Rflag) 241 type = DIR_TO_DNE; 242 else 243 type = FILE_TO_FILE; 244 } else 245 type = FILE_TO_FILE; 246 247 if (have_trailing_slash && type == FILE_TO_FILE) { 248 if (r == -1) { 249 errx(1, "directory %s does not exist", 250 to.p_path); 251 } else 252 errx(1, "%s is not a directory", to.p_path); 253 } 254 } else 255 /* 256 * Case (2). Target is a directory. 257 */ 258 type = FILE_TO_DIR; 259 260 /* 261 * For DIR_TO_DNE, we could provide copy() with the to_stat we've 262 * already allocated on the stack here that isn't being used for 263 * anything. Not doing so, though, simplifies later logic a little bit 264 * as we need to skip checking root_stat on the first iteration and 265 * ensure that we set it with the first mkdir(). 266 */ 267 exit (copy(argv, type, fts_options, (type == DIR_TO_DNE ? NULL : 268 &to_stat))); 269 } 270 271 /* Does the right thing based on -R + -H/-L/-P */ 272 static int 273 copy_stat(const char *path, struct stat *sb) 274 { 275 276 /* 277 * For -R -H/-P, we need to lstat() instead; copy() cares about the link 278 * itself rather than the target if we're not following links during the 279 * traversal. 280 */ 281 if (!Rflag || Lflag) 282 return (stat(path, sb)); 283 return (lstat(path, sb)); 284 } 285 286 287 static int 288 copy(char *argv[], enum op type, int fts_options, struct stat *root_stat) 289 { 290 char rootname[NAME_MAX]; 291 struct stat created_root_stat, to_stat; 292 FTS *ftsp; 293 FTSENT *curr; 294 int base = 0, dne, badcp, rval; 295 size_t nlen; 296 char *p, *recurse_path, *target_mid; 297 mode_t mask, mode; 298 299 /* 300 * Keep an inverted copy of the umask, for use in correcting 301 * permissions on created directories when not using -p. 302 */ 303 mask = ~umask(0777); 304 umask(~mask); 305 306 recurse_path = NULL; 307 if ((ftsp = fts_open(argv, fts_options, NULL)) == NULL) 308 err(1, "fts_open"); 309 for (badcp = rval = 0; (curr = fts_read(ftsp)) != NULL; badcp = 0) { 310 switch (curr->fts_info) { 311 case FTS_NS: 312 case FTS_DNR: 313 case FTS_ERR: 314 warnx("%s: %s", 315 curr->fts_path, strerror(curr->fts_errno)); 316 badcp = rval = 1; 317 continue; 318 case FTS_DC: /* Warn, continue. */ 319 warnx("%s: directory causes a cycle", curr->fts_path); 320 badcp = rval = 1; 321 continue; 322 default: 323 ; 324 } 325 326 /* 327 * Stash the root basename off for detecting recursion later. 328 * 329 * This will be essential if the root is a symlink and we're 330 * rolling with -L or -H. The later bits will need this bit in 331 * particular. 332 */ 333 if (curr->fts_level == FTS_ROOTLEVEL) { 334 strlcpy(rootname, curr->fts_name, sizeof(rootname)); 335 } 336 337 /* 338 * If we are in case (2) or (3) above, we need to append the 339 * source name to the target name. 340 */ 341 if (type != FILE_TO_FILE) { 342 /* 343 * Need to remember the roots of traversals to create 344 * correct pathnames. If there's a directory being 345 * copied to a non-existent directory, e.g. 346 * cp -R a/dir noexist 347 * the resulting path name should be noexist/foo, not 348 * noexist/dir/foo (where foo is a file in dir), which 349 * is the case where the target exists. 350 * 351 * Also, check for "..". This is for correct path 352 * concatenation for paths ending in "..", e.g. 353 * cp -R .. /tmp 354 * Paths ending in ".." are changed to ".". This is 355 * tricky, but seems the easiest way to fix the problem. 356 * 357 * XXX 358 * Since the first level MUST be FTS_ROOTLEVEL, base 359 * is always initialized. 360 */ 361 if (curr->fts_level == FTS_ROOTLEVEL) { 362 if (type != DIR_TO_DNE) { 363 p = strrchr(curr->fts_path, '/'); 364 base = (p == NULL) ? 0 : 365 (int)(p - curr->fts_path + 1); 366 367 if (!strcmp(&curr->fts_path[base], 368 "..")) 369 base += 1; 370 } else 371 base = curr->fts_pathlen; 372 } 373 374 p = &curr->fts_path[base]; 375 nlen = curr->fts_pathlen - base; 376 target_mid = to.target_end; 377 if (*p != '/' && target_mid[-1] != '/') 378 *target_mid++ = '/'; 379 *target_mid = 0; 380 if (target_mid - to.p_path + nlen >= PATH_MAX) { 381 warnx("%s%s: name too long (not copied)", 382 to.p_path, p); 383 badcp = rval = 1; 384 continue; 385 } 386 (void)strncat(target_mid, p, nlen); 387 to.p_end = target_mid + nlen; 388 *to.p_end = 0; 389 STRIP_TRAILING_SLASH(to); 390 391 /* 392 * We're on the verge of recursing on ourselves. Either 393 * we need to stop right here (we knowingly just created 394 * it), or we will in an immediate descendant. Record 395 * the path of the immediate descendant to make our 396 * lives a little less complicated looking. 397 */ 398 if (curr->fts_info == FTS_D && root_stat != NULL && 399 root_stat->st_dev == curr->fts_statp->st_dev && 400 root_stat->st_ino == curr->fts_statp->st_ino) { 401 assert(recurse_path == NULL); 402 403 if (root_stat == &created_root_stat) { 404 /* 405 * This directory didn't exist when we 406 * started, we created it as part of 407 * traversal. Stop right here before we 408 * do something silly. 409 */ 410 fts_set(ftsp, curr, FTS_SKIP); 411 continue; 412 } 413 414 415 if (asprintf(&recurse_path, "%s/%s", to.p_path, 416 rootname) == -1) 417 err(1, "asprintf"); 418 } 419 420 if (recurse_path != NULL && 421 strcmp(to.p_path, recurse_path) == 0) { 422 fts_set(ftsp, curr, FTS_SKIP); 423 continue; 424 } 425 } 426 427 if (curr->fts_info == FTS_DP) { 428 /* 429 * We are nearly finished with this directory. If we 430 * didn't actually copy it, or otherwise don't need to 431 * change its attributes, then we are done. 432 */ 433 if (!curr->fts_number) 434 continue; 435 /* 436 * If -p is in effect, set all the attributes. 437 * Otherwise, set the correct permissions, limited 438 * by the umask. Optimise by avoiding a chmod() 439 * if possible (which is usually the case if we 440 * made the directory). Note that mkdir() does not 441 * honour setuid, setgid and sticky bits, but we 442 * normally want to preserve them on directories. 443 */ 444 if (pflag) { 445 if (setfile(curr->fts_statp, -1)) 446 rval = 1; 447 if (preserve_dir_acls(curr->fts_statp, 448 curr->fts_accpath, to.p_path) != 0) 449 rval = 1; 450 } else { 451 mode = curr->fts_statp->st_mode; 452 if ((mode & (S_ISUID | S_ISGID | S_ISTXT)) || 453 ((mode | S_IRWXU) & mask) != (mode & mask)) 454 if (chmod(to.p_path, mode & mask) != 455 0) { 456 warn("chmod: %s", to.p_path); 457 rval = 1; 458 } 459 } 460 continue; 461 } 462 463 /* Not an error but need to remember it happened. */ 464 if (copy_stat(to.p_path, &to_stat) == -1) 465 dne = 1; 466 else { 467 if (to_stat.st_dev == curr->fts_statp->st_dev && 468 to_stat.st_ino == curr->fts_statp->st_ino) { 469 warnx("%s and %s are identical (not copied).", 470 to.p_path, curr->fts_path); 471 badcp = rval = 1; 472 if (S_ISDIR(curr->fts_statp->st_mode)) 473 (void)fts_set(ftsp, curr, FTS_SKIP); 474 continue; 475 } 476 if (!S_ISDIR(curr->fts_statp->st_mode) && 477 S_ISDIR(to_stat.st_mode)) { 478 warnx("cannot overwrite directory %s with " 479 "non-directory %s", 480 to.p_path, curr->fts_path); 481 badcp = rval = 1; 482 continue; 483 } 484 dne = 0; 485 } 486 487 switch (curr->fts_statp->st_mode & S_IFMT) { 488 case S_IFLNK: 489 /* Catch special case of a non-dangling symlink. */ 490 if ((fts_options & FTS_LOGICAL) || 491 ((fts_options & FTS_COMFOLLOW) && 492 curr->fts_level == 0)) { 493 if (copy_file(curr, dne)) 494 badcp = rval = 1; 495 } else { 496 if (copy_link(curr, !dne)) 497 badcp = rval = 1; 498 } 499 break; 500 case S_IFDIR: 501 if (!Rflag) { 502 warnx("%s is a directory (not copied).", 503 curr->fts_path); 504 (void)fts_set(ftsp, curr, FTS_SKIP); 505 badcp = rval = 1; 506 break; 507 } 508 /* 509 * If the directory doesn't exist, create the new 510 * one with the from file mode plus owner RWX bits, 511 * modified by the umask. Trade-off between being 512 * able to write the directory (if from directory is 513 * 555) and not causing a permissions race. If the 514 * umask blocks owner writes, we fail. 515 */ 516 if (dne) { 517 if (mkdir(to.p_path, 518 curr->fts_statp->st_mode | S_IRWXU) < 0) 519 err(1, "%s", to.p_path); 520 /* 521 * First DNE with a NULL root_stat is the root 522 * path, so set root_stat. We can't really 523 * tell in all cases if the target path is 524 * within the src path, so we just stat() the 525 * first directory we created and use that. 526 */ 527 if (root_stat == NULL && 528 stat(to.p_path, &created_root_stat) == -1) { 529 err(1, "stat"); 530 } else if (root_stat == NULL) { 531 root_stat = &created_root_stat; 532 } 533 } else if (!S_ISDIR(to_stat.st_mode)) { 534 errno = ENOTDIR; 535 err(1, "%s", to.p_path); 536 } 537 /* 538 * Arrange to correct directory attributes later 539 * (in the post-order phase) if this is a new 540 * directory, or if the -p flag is in effect. 541 */ 542 curr->fts_number = pflag || dne; 543 break; 544 case S_IFBLK: 545 case S_IFCHR: 546 if (Rflag && !sflag) { 547 if (copy_special(curr->fts_statp, !dne)) 548 badcp = rval = 1; 549 } else { 550 if (copy_file(curr, dne)) 551 badcp = rval = 1; 552 } 553 break; 554 case S_IFSOCK: 555 warnx("%s is a socket (not copied).", 556 curr->fts_path); 557 break; 558 case S_IFIFO: 559 if (Rflag && !sflag) { 560 if (copy_fifo(curr->fts_statp, !dne)) 561 badcp = rval = 1; 562 } else { 563 if (copy_file(curr, dne)) 564 badcp = rval = 1; 565 } 566 break; 567 default: 568 if (copy_file(curr, dne)) 569 badcp = rval = 1; 570 break; 571 } 572 if (vflag && !badcp) 573 (void)printf("%s -> %s\n", curr->fts_path, to.p_path); 574 } 575 if (errno) 576 err(1, "fts_read"); 577 fts_close(ftsp); 578 free(recurse_path); 579 return (rval); 580 } 581 582 static void 583 siginfo(int sig __unused) 584 { 585 586 info = 1; 587 } 588