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, 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, "HLNPRafilnprsvx")) != -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 'N': 105 Nflag = 1; 106 break; 107 case 'P': 108 Pflag = 1; 109 Hflag = Lflag = 0; 110 break; 111 case 'R': 112 Rflag = 1; 113 break; 114 case 'a': 115 pflag = 1; 116 Rflag = 1; 117 Pflag = 1; 118 Hflag = Lflag = 0; 119 break; 120 case 'f': 121 fflag = 1; 122 iflag = nflag = 0; 123 break; 124 case 'i': 125 iflag = 1; 126 fflag = nflag = 0; 127 break; 128 case 'l': 129 lflag = 1; 130 break; 131 case 'n': 132 nflag = 1; 133 fflag = iflag = 0; 134 break; 135 case 'p': 136 pflag = 1; 137 break; 138 case 'r': 139 rflag = Lflag = 1; 140 Hflag = Pflag = 0; 141 break; 142 case 's': 143 sflag = 1; 144 break; 145 case 'v': 146 vflag = 1; 147 break; 148 case 'x': 149 fts_options |= FTS_XDEV; 150 break; 151 default: 152 usage(); 153 break; 154 } 155 argc -= optind; 156 argv += optind; 157 158 if (argc < 2) 159 usage(); 160 161 if (Rflag && rflag) 162 errx(1, "the -R and -r options may not be specified together"); 163 if (lflag && sflag) 164 errx(1, "the -l and -s options may not be specified together"); 165 if (rflag) 166 Rflag = 1; 167 if (Rflag) { 168 if (Hflag) 169 fts_options |= FTS_COMFOLLOW; 170 if (Lflag) { 171 fts_options &= ~FTS_PHYSICAL; 172 fts_options |= FTS_LOGICAL; 173 } 174 } else if (!Pflag) { 175 fts_options &= ~FTS_PHYSICAL; 176 fts_options |= FTS_LOGICAL | FTS_COMFOLLOW; 177 } 178 (void)signal(SIGINFO, siginfo); 179 180 /* Save the target base in "to". */ 181 target = argv[--argc]; 182 if (strlcpy(to.p_path, target, sizeof(to.p_path)) >= sizeof(to.p_path)) 183 errx(1, "%s: name too long", target); 184 to.p_end = to.p_path + strlen(to.p_path); 185 if (to.p_path == to.p_end) { 186 *to.p_end++ = '.'; 187 *to.p_end = 0; 188 } 189 have_trailing_slash = (to.p_end[-1] == '/'); 190 if (have_trailing_slash) 191 STRIP_TRAILING_SLASH(to); 192 to.target_end = to.p_end; 193 194 /* Set end of argument list for fts(3). */ 195 argv[argc] = NULL; 196 197 /* 198 * Cp has two distinct cases: 199 * 200 * cp [-R] source target 201 * cp [-R] source1 ... sourceN directory 202 * 203 * In both cases, source can be either a file or a directory. 204 * 205 * In (1), the target becomes a copy of the source. That is, if the 206 * source is a file, the target will be a file, and likewise for 207 * directories. 208 * 209 * In (2), the real target is not directory, but "directory/source". 210 */ 211 r = stat(to.p_path, &to_stat); 212 if (r == -1 && errno != ENOENT) 213 err(1, "%s", to.p_path); 214 if (r == -1 || !S_ISDIR(to_stat.st_mode)) { 215 /* 216 * Case (1). Target is not a directory. 217 */ 218 if (argc > 1) 219 errx(1, "%s is not a directory", to.p_path); 220 221 /* 222 * Need to detect the case: 223 * cp -R dir foo 224 * Where dir is a directory and foo does not exist, where 225 * we want pathname concatenations turned on but not for 226 * the initial mkdir(). 227 */ 228 if (r == -1) { 229 if (Rflag && (Lflag || Hflag)) 230 stat(*argv, &tmp_stat); 231 else 232 lstat(*argv, &tmp_stat); 233 234 if (S_ISDIR(tmp_stat.st_mode) && Rflag) 235 type = DIR_TO_DNE; 236 else 237 type = FILE_TO_FILE; 238 } else 239 type = FILE_TO_FILE; 240 241 if (have_trailing_slash && type == FILE_TO_FILE) { 242 if (r == -1) { 243 errx(1, "directory %s does not exist", 244 to.p_path); 245 } else 246 errx(1, "%s is not a directory", to.p_path); 247 } 248 } else 249 /* 250 * Case (2). Target is a directory. 251 */ 252 type = FILE_TO_DIR; 253 254 /* 255 * For DIR_TO_DNE, we could provide copy() with the to_stat we've 256 * already allocated on the stack here that isn't being used for 257 * anything. Not doing so, though, simplifies later logic a little bit 258 * as we need to skip checking root_stat on the first iteration and 259 * ensure that we set it with the first mkdir(). 260 */ 261 exit (copy(argv, type, fts_options, (type == DIR_TO_DNE ? NULL : 262 &to_stat))); 263 } 264 265 /* Does the right thing based on -R + -H/-L/-P */ 266 static int 267 copy_stat(const char *path, struct stat *sb) 268 { 269 270 /* 271 * For -R -H/-P, we need to lstat() instead; copy() cares about the link 272 * itself rather than the target if we're not following links during the 273 * traversal. 274 */ 275 if (!Rflag || Lflag) 276 return (stat(path, sb)); 277 return (lstat(path, sb)); 278 } 279 280 281 static int 282 copy(char *argv[], enum op type, int fts_options, struct stat *root_stat) 283 { 284 char rootname[NAME_MAX]; 285 struct stat created_root_stat, to_stat; 286 FTS *ftsp; 287 FTSENT *curr; 288 int base = 0, dne, badcp, rval; 289 size_t nlen; 290 char *p, *recurse_path, *target_mid; 291 mode_t mask, mode; 292 293 /* 294 * Keep an inverted copy of the umask, for use in correcting 295 * permissions on created directories when not using -p. 296 */ 297 mask = ~umask(0777); 298 umask(~mask); 299 300 recurse_path = NULL; 301 if ((ftsp = fts_open(argv, fts_options, NULL)) == NULL) 302 err(1, "fts_open"); 303 for (badcp = rval = 0; (curr = fts_read(ftsp)) != NULL; badcp = 0) { 304 switch (curr->fts_info) { 305 case FTS_NS: 306 case FTS_DNR: 307 case FTS_ERR: 308 warnx("%s: %s", 309 curr->fts_path, strerror(curr->fts_errno)); 310 badcp = rval = 1; 311 continue; 312 case FTS_DC: /* Warn, continue. */ 313 warnx("%s: directory causes a cycle", curr->fts_path); 314 badcp = rval = 1; 315 continue; 316 default: 317 ; 318 } 319 320 /* 321 * Stash the root basename off for detecting recursion later. 322 * 323 * This will be essential if the root is a symlink and we're 324 * rolling with -L or -H. The later bits will need this bit in 325 * particular. 326 */ 327 if (curr->fts_level == FTS_ROOTLEVEL) { 328 strlcpy(rootname, curr->fts_name, sizeof(rootname)); 329 } 330 331 /* 332 * If we are in case (2) or (3) above, we need to append the 333 * source name to the target name. 334 */ 335 if (type != FILE_TO_FILE) { 336 /* 337 * Need to remember the roots of traversals to create 338 * correct pathnames. If there's a directory being 339 * copied to a non-existent directory, e.g. 340 * cp -R a/dir noexist 341 * the resulting path name should be noexist/foo, not 342 * noexist/dir/foo (where foo is a file in dir), which 343 * is the case where the target exists. 344 * 345 * Also, check for "..". This is for correct path 346 * concatenation for paths ending in "..", e.g. 347 * cp -R .. /tmp 348 * Paths ending in ".." are changed to ".". This is 349 * tricky, but seems the easiest way to fix the problem. 350 * 351 * XXX 352 * Since the first level MUST be FTS_ROOTLEVEL, base 353 * is always initialized. 354 */ 355 if (curr->fts_level == FTS_ROOTLEVEL) { 356 if (type != DIR_TO_DNE) { 357 p = strrchr(curr->fts_path, '/'); 358 base = (p == NULL) ? 0 : 359 (int)(p - curr->fts_path + 1); 360 361 if (!strcmp(&curr->fts_path[base], 362 "..")) 363 base += 1; 364 } else 365 base = curr->fts_pathlen; 366 } 367 368 p = &curr->fts_path[base]; 369 nlen = curr->fts_pathlen - base; 370 target_mid = to.target_end; 371 if (*p != '/' && target_mid[-1] != '/') 372 *target_mid++ = '/'; 373 *target_mid = 0; 374 if (target_mid - to.p_path + nlen >= PATH_MAX) { 375 warnx("%s%s: name too long (not copied)", 376 to.p_path, p); 377 badcp = rval = 1; 378 continue; 379 } 380 (void)strncat(target_mid, p, nlen); 381 to.p_end = target_mid + nlen; 382 *to.p_end = 0; 383 STRIP_TRAILING_SLASH(to); 384 385 /* 386 * We're on the verge of recursing on ourselves. Either 387 * we need to stop right here (we knowingly just created 388 * it), or we will in an immediate descendant. Record 389 * the path of the immediate descendant to make our 390 * lives a little less complicated looking. 391 */ 392 if (curr->fts_info == FTS_D && root_stat != NULL && 393 root_stat->st_dev == curr->fts_statp->st_dev && 394 root_stat->st_ino == curr->fts_statp->st_ino) { 395 assert(recurse_path == NULL); 396 397 if (root_stat == &created_root_stat) { 398 /* 399 * This directory didn't exist when we 400 * started, we created it as part of 401 * traversal. Stop right here before we 402 * do something silly. 403 */ 404 fts_set(ftsp, curr, FTS_SKIP); 405 continue; 406 } 407 408 409 if (asprintf(&recurse_path, "%s/%s", to.p_path, 410 rootname) == -1) 411 err(1, "asprintf"); 412 } 413 414 if (recurse_path != NULL && 415 strcmp(to.p_path, recurse_path) == 0) { 416 fts_set(ftsp, curr, FTS_SKIP); 417 continue; 418 } 419 } 420 421 if (curr->fts_info == FTS_DP) { 422 /* 423 * We are nearly finished with this directory. If we 424 * didn't actually copy it, or otherwise don't need to 425 * change its attributes, then we are done. 426 */ 427 if (!curr->fts_number) 428 continue; 429 /* 430 * If -p is in effect, set all the attributes. 431 * Otherwise, set the correct permissions, limited 432 * by the umask. Optimise by avoiding a chmod() 433 * if possible (which is usually the case if we 434 * made the directory). Note that mkdir() does not 435 * honour setuid, setgid and sticky bits, but we 436 * normally want to preserve them on directories. 437 */ 438 if (pflag) { 439 if (setfile(curr->fts_statp, -1)) 440 rval = 1; 441 if (preserve_dir_acls(curr->fts_statp, 442 curr->fts_accpath, to.p_path) != 0) 443 rval = 1; 444 } else { 445 mode = curr->fts_statp->st_mode; 446 if ((mode & (S_ISUID | S_ISGID | S_ISTXT)) || 447 ((mode | S_IRWXU) & mask) != (mode & mask)) 448 if (chmod(to.p_path, mode & mask) != 449 0) { 450 warn("chmod: %s", to.p_path); 451 rval = 1; 452 } 453 } 454 continue; 455 } 456 457 /* Not an error but need to remember it happened. */ 458 if (copy_stat(to.p_path, &to_stat) == -1) 459 dne = 1; 460 else { 461 if (to_stat.st_dev == curr->fts_statp->st_dev && 462 to_stat.st_ino == curr->fts_statp->st_ino) { 463 warnx("%s and %s are identical (not copied).", 464 to.p_path, curr->fts_path); 465 badcp = rval = 1; 466 if (S_ISDIR(curr->fts_statp->st_mode)) 467 (void)fts_set(ftsp, curr, FTS_SKIP); 468 continue; 469 } 470 if (!S_ISDIR(curr->fts_statp->st_mode) && 471 S_ISDIR(to_stat.st_mode)) { 472 warnx("cannot overwrite directory %s with " 473 "non-directory %s", 474 to.p_path, curr->fts_path); 475 badcp = rval = 1; 476 continue; 477 } 478 dne = 0; 479 } 480 481 switch (curr->fts_statp->st_mode & S_IFMT) { 482 case S_IFLNK: 483 /* Catch special case of a non-dangling symlink. */ 484 if ((fts_options & FTS_LOGICAL) || 485 ((fts_options & FTS_COMFOLLOW) && 486 curr->fts_level == 0)) { 487 if (copy_file(curr, dne)) 488 badcp = rval = 1; 489 } else { 490 if (copy_link(curr, !dne)) 491 badcp = rval = 1; 492 } 493 break; 494 case S_IFDIR: 495 if (!Rflag) { 496 warnx("%s is a directory (not copied).", 497 curr->fts_path); 498 (void)fts_set(ftsp, curr, FTS_SKIP); 499 badcp = rval = 1; 500 break; 501 } 502 /* 503 * If the directory doesn't exist, create the new 504 * one with the from file mode plus owner RWX bits, 505 * modified by the umask. Trade-off between being 506 * able to write the directory (if from directory is 507 * 555) and not causing a permissions race. If the 508 * umask blocks owner writes, we fail. 509 */ 510 if (dne) { 511 if (mkdir(to.p_path, 512 curr->fts_statp->st_mode | S_IRWXU) < 0) 513 err(1, "%s", to.p_path); 514 /* 515 * First DNE with a NULL root_stat is the root 516 * path, so set root_stat. We can't really 517 * tell in all cases if the target path is 518 * within the src path, so we just stat() the 519 * first directory we created and use that. 520 */ 521 if (root_stat == NULL && 522 stat(to.p_path, &created_root_stat) == -1) { 523 err(1, "stat"); 524 } else if (root_stat == NULL) { 525 root_stat = &created_root_stat; 526 } 527 } else if (!S_ISDIR(to_stat.st_mode)) { 528 errno = ENOTDIR; 529 err(1, "%s", to.p_path); 530 } 531 /* 532 * Arrange to correct directory attributes later 533 * (in the post-order phase) if this is a new 534 * directory, or if the -p flag is in effect. 535 */ 536 curr->fts_number = pflag || dne; 537 break; 538 case S_IFBLK: 539 case S_IFCHR: 540 if (Rflag && !sflag) { 541 if (copy_special(curr->fts_statp, !dne)) 542 badcp = rval = 1; 543 } else { 544 if (copy_file(curr, dne)) 545 badcp = rval = 1; 546 } 547 break; 548 case S_IFSOCK: 549 warnx("%s is a socket (not copied).", 550 curr->fts_path); 551 break; 552 case S_IFIFO: 553 if (Rflag && !sflag) { 554 if (copy_fifo(curr->fts_statp, !dne)) 555 badcp = rval = 1; 556 } else { 557 if (copy_file(curr, dne)) 558 badcp = rval = 1; 559 } 560 break; 561 default: 562 if (copy_file(curr, dne)) 563 badcp = rval = 1; 564 break; 565 } 566 if (vflag && !badcp) 567 (void)printf("%s -> %s\n", curr->fts_path, to.p_path); 568 } 569 if (errno) 570 err(1, "fts_read"); 571 fts_close(ftsp); 572 free(recurse_path); 573 return (rval); 574 } 575 576 static void 577 siginfo(int sig __unused) 578 { 579 580 info = 1; 581 } 582