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
main(int argc,char * argv[])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 errc(1, ENOTDIR, "%s", 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 errc(1, ENOENT, "%s", to.p_path);
242 else
243 errc(1, ENOTDIR, "%s", to.p_path);
244 }
245 } else {
246 /*
247 * Case (2). Target is a directory.
248 */
249 type = FILE_TO_DIR;
250 }
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 static int
copy(char * argv[],enum op type,int fts_options,struct stat * root_stat)264 copy(char *argv[], enum op type, int fts_options, struct stat *root_stat)
265 {
266 char rootname[NAME_MAX];
267 struct stat created_root_stat, to_stat;
268 FTS *ftsp;
269 FTSENT *curr;
270 int base = 0, dne, badcp, rval;
271 size_t nlen;
272 char *p, *recurse_path, *target_mid;
273 mode_t mask, mode;
274
275 /*
276 * Keep an inverted copy of the umask, for use in correcting
277 * permissions on created directories when not using -p.
278 */
279 mask = ~umask(0777);
280 umask(~mask);
281
282 recurse_path = NULL;
283 if ((ftsp = fts_open(argv, fts_options, NULL)) == NULL)
284 err(1, "fts_open");
285 for (badcp = rval = 0; (curr = fts_read(ftsp)) != NULL; badcp = 0) {
286 switch (curr->fts_info) {
287 case FTS_NS:
288 case FTS_DNR:
289 case FTS_ERR:
290 warnc(curr->fts_errno, "%s", curr->fts_path);
291 badcp = rval = 1;
292 continue;
293 case FTS_DC: /* Warn, continue. */
294 warnx("%s: directory causes a cycle", curr->fts_path);
295 badcp = rval = 1;
296 continue;
297 default:
298 ;
299 }
300
301 /*
302 * Stash the root basename off for detecting recursion later.
303 *
304 * This will be essential if the root is a symlink and we're
305 * rolling with -L or -H. The later bits will need this bit in
306 * particular.
307 */
308 if (curr->fts_level == FTS_ROOTLEVEL) {
309 strlcpy(rootname, curr->fts_name, sizeof(rootname));
310 }
311
312 /*
313 * If we are in case (2) or (3) above, we need to append the
314 * source name to the target name.
315 */
316 if (type != FILE_TO_FILE) {
317 /*
318 * Need to remember the roots of traversals to create
319 * correct pathnames. If there's a directory being
320 * copied to a non-existent directory, e.g.
321 * cp -R a/dir noexist
322 * the resulting path name should be noexist/foo, not
323 * noexist/dir/foo (where foo is a file in dir), which
324 * is the case where the target exists.
325 *
326 * Also, check for "..". This is for correct path
327 * concatenation for paths ending in "..", e.g.
328 * cp -R .. /tmp
329 * Paths ending in ".." are changed to ".". This is
330 * tricky, but seems the easiest way to fix the problem.
331 *
332 * XXX
333 * Since the first level MUST be FTS_ROOTLEVEL, base
334 * is always initialized.
335 */
336 if (curr->fts_level == FTS_ROOTLEVEL) {
337 if (type != DIR_TO_DNE) {
338 p = strrchr(curr->fts_path, '/');
339 base = (p == NULL) ? 0 :
340 (int)(p - curr->fts_path + 1);
341
342 if (strcmp(curr->fts_path + base, "..")
343 == 0)
344 base += 1;
345 } else
346 base = curr->fts_pathlen;
347 }
348
349 p = &curr->fts_path[base];
350 nlen = curr->fts_pathlen - base;
351 target_mid = to.target_end;
352 if (*p != '/' && target_mid[-1] != '/')
353 *target_mid++ = '/';
354 *target_mid = 0;
355 if (target_mid - to.p_path + nlen >= PATH_MAX) {
356 warnx("%s%s: name too long (not copied)",
357 to.p_path, p);
358 badcp = rval = 1;
359 continue;
360 }
361 (void)strncat(target_mid, p, nlen);
362 to.p_end = target_mid + nlen;
363 *to.p_end = 0;
364 STRIP_TRAILING_SLASH(to);
365
366 /*
367 * We're on the verge of recursing on ourselves. Either
368 * we need to stop right here (we knowingly just created
369 * it), or we will in an immediate descendant. Record
370 * the path of the immediate descendant to make our
371 * lives a little less complicated looking.
372 */
373 if (curr->fts_info == FTS_D && root_stat != NULL &&
374 root_stat->st_dev == curr->fts_statp->st_dev &&
375 root_stat->st_ino == curr->fts_statp->st_ino) {
376 assert(recurse_path == NULL);
377
378 if (root_stat == &created_root_stat) {
379 /*
380 * This directory didn't exist when we
381 * started, we created it as part of
382 * traversal. Stop right here before we
383 * do something silly.
384 */
385 fts_set(ftsp, curr, FTS_SKIP);
386 continue;
387 }
388
389 if (asprintf(&recurse_path, "%s/%s", to.p_path,
390 rootname) == -1)
391 err(1, "asprintf");
392 }
393
394 if (recurse_path != NULL &&
395 strcmp(to.p_path, recurse_path) == 0) {
396 fts_set(ftsp, curr, FTS_SKIP);
397 continue;
398 }
399 }
400
401 if (curr->fts_info == FTS_DP) {
402 /*
403 * We are nearly finished with this directory. If we
404 * didn't actually copy it, or otherwise don't need to
405 * change its attributes, then we are done.
406 */
407 if (!curr->fts_number)
408 continue;
409 /*
410 * If -p is in effect, set all the attributes.
411 * Otherwise, set the correct permissions, limited
412 * by the umask. Optimise by avoiding a chmod()
413 * if possible (which is usually the case if we
414 * made the directory). Note that mkdir() does not
415 * honour setuid, setgid and sticky bits, but we
416 * normally want to preserve them on directories.
417 */
418 if (pflag) {
419 if (setfile(curr->fts_statp, -1))
420 rval = 1;
421 if (preserve_dir_acls(curr->fts_statp,
422 curr->fts_accpath, to.p_path) != 0)
423 rval = 1;
424 } else {
425 mode = curr->fts_statp->st_mode;
426 if ((mode & (S_ISUID | S_ISGID | S_ISTXT)) ||
427 ((mode | S_IRWXU) & mask) != (mode & mask))
428 if (chmod(to.p_path, mode & mask) !=
429 0) {
430 warn("chmod: %s", to.p_path);
431 rval = 1;
432 }
433 }
434 continue;
435 }
436
437 /* Check if source and destination are identical. */
438 if (stat(to.p_path, &to_stat) == 0 &&
439 to_stat.st_dev == curr->fts_statp->st_dev &&
440 to_stat.st_ino == curr->fts_statp->st_ino) {
441 warnx("%s and %s are identical (not copied).",
442 to.p_path, curr->fts_path);
443 badcp = rval = 1;
444 if (S_ISDIR(curr->fts_statp->st_mode))
445 (void)fts_set(ftsp, curr, FTS_SKIP);
446 continue;
447 }
448
449 /* Not an error but need to remember it happened. */
450 dne = lstat(to.p_path, &to_stat) != 0;
451
452 switch (curr->fts_statp->st_mode & S_IFMT) {
453 case S_IFLNK:
454 if ((fts_options & FTS_LOGICAL) ||
455 ((fts_options & FTS_COMFOLLOW) &&
456 curr->fts_level == 0)) {
457 /*
458 * We asked FTS to follow links but got
459 * here anyway, which means the target is
460 * nonexistent or inaccessible. Let
461 * copy_file() deal with the error.
462 */
463 if (copy_file(curr, dne))
464 badcp = rval = 1;
465 } else {
466 /* Copy the link. */
467 if (copy_link(curr, !dne))
468 badcp = rval = 1;
469 }
470 break;
471 case S_IFDIR:
472 if (!Rflag) {
473 warnx("%s is a directory (not copied).",
474 curr->fts_path);
475 (void)fts_set(ftsp, curr, FTS_SKIP);
476 badcp = rval = 1;
477 break;
478 }
479 /*
480 * If the directory doesn't exist, create the new
481 * one with the from file mode plus owner RWX bits,
482 * modified by the umask. Trade-off between being
483 * able to write the directory (if from directory is
484 * 555) and not causing a permissions race. If the
485 * umask blocks owner writes, we fail.
486 */
487 if (dne) {
488 mode = curr->fts_statp->st_mode | S_IRWXU;
489 if (mkdir(to.p_path, mode) != 0) {
490 warn("%s", to.p_path);
491 (void)fts_set(ftsp, curr, FTS_SKIP);
492 badcp = rval = 1;
493 break;
494 }
495 /*
496 * First DNE with a NULL root_stat is the root
497 * path, so set root_stat. We can't really
498 * tell in all cases if the target path is
499 * within the src path, so we just stat() the
500 * first directory we created and use that.
501 */
502 if (root_stat == NULL &&
503 stat(to.p_path, &created_root_stat) != 0) {
504 warn("%s", to.p_path);
505 (void)fts_set(ftsp, curr, FTS_SKIP);
506 badcp = rval = 1;
507 break;
508 }
509 if (root_stat == NULL)
510 root_stat = &created_root_stat;
511 } else if (!S_ISDIR(to_stat.st_mode)) {
512 warnc(ENOTDIR, "%s", to.p_path);
513 (void)fts_set(ftsp, curr, FTS_SKIP);
514 badcp = rval = 1;
515 break;
516 }
517 /*
518 * Arrange to correct directory attributes later
519 * (in the post-order phase) if this is a new
520 * directory, or if the -p flag is in effect.
521 */
522 curr->fts_number = pflag || dne;
523 break;
524 case S_IFBLK:
525 case S_IFCHR:
526 if (Rflag && !sflag) {
527 if (copy_special(curr->fts_statp, !dne))
528 badcp = rval = 1;
529 } else {
530 if (copy_file(curr, dne))
531 badcp = rval = 1;
532 }
533 break;
534 case S_IFSOCK:
535 warnx("%s is a socket (not copied).",
536 curr->fts_path);
537 break;
538 case S_IFIFO:
539 if (Rflag && !sflag) {
540 if (copy_fifo(curr->fts_statp, !dne))
541 badcp = rval = 1;
542 } else {
543 if (copy_file(curr, dne))
544 badcp = rval = 1;
545 }
546 break;
547 default:
548 if (copy_file(curr, dne))
549 badcp = rval = 1;
550 break;
551 }
552 if (vflag && !badcp)
553 (void)printf("%s -> %s\n", curr->fts_path, to.p_path);
554 }
555 if (errno)
556 err(1, "fts_read");
557 fts_close(ftsp);
558 free(recurse_path);
559 return (rval);
560 }
561
562 static void
siginfo(int sig __unused)563 siginfo(int sig __unused)
564 {
565
566 info = 1;
567 }
568