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