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