xref: /freebsd/bin/cp/cp.c (revision 884a2a699669ec61e2366e3e358342dbc94be24a)
1 /*-
2  * Copyright (c) 1988, 1993, 1994
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * David Hitz of Auspex Systems Inc.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #if 0
34 #ifndef lint
35 static char const copyright[] =
36 "@(#) Copyright (c) 1988, 1993, 1994\n\
37 	The Regents of the University of California.  All rights reserved.\n";
38 #endif /* not lint */
39 
40 #ifndef lint
41 static char sccsid[] = "@(#)cp.c	8.2 (Berkeley) 4/1/94";
42 #endif /* not lint */
43 #endif
44 #include <sys/cdefs.h>
45 __FBSDID("$FreeBSD$");
46 
47 /*
48  * Cp copies source files to target files.
49  *
50  * The global PATH_T structure "to" always contains the path to the
51  * current target file.  Since fts(3) does not change directories,
52  * this path can be either absolute or dot-relative.
53  *
54  * The basic algorithm is to initialize "to" and use fts(3) to traverse
55  * the file hierarchy rooted in the argument list.  A trivial case is the
56  * case of 'cp file1 file2'.  The more interesting case is the case of
57  * 'cp file1 file2 ... fileN dir' where the hierarchy is traversed and the
58  * path (relative to the root of the traversal) is appended to dir (stored
59  * in "to") to form the final target path.
60  */
61 
62 #include <sys/types.h>
63 #include <sys/stat.h>
64 
65 #include <err.h>
66 #include <errno.h>
67 #include <fts.h>
68 #include <limits.h>
69 #include <signal.h>
70 #include <stdio.h>
71 #include <stdlib.h>
72 #include <string.h>
73 #include <unistd.h>
74 
75 #include "extern.h"
76 
77 #define	STRIP_TRAILING_SLASH(p) {					\
78         while ((p).p_end > (p).p_path + 1 && (p).p_end[-1] == '/')	\
79                 *--(p).p_end = 0;					\
80 }
81 
82 static char emptystring[] = "";
83 
84 PATH_T to = { to.p_path, emptystring, "" };
85 
86 int fflag, iflag, lflag, nflag, pflag, vflag;
87 static int Rflag, rflag;
88 volatile sig_atomic_t info;
89 
90 enum op { FILE_TO_FILE, FILE_TO_DIR, DIR_TO_DNE };
91 
92 static int copy(char *[], enum op, int);
93 static int mastercmp(const FTSENT * const *, const FTSENT * const *);
94 static void siginfo(int __unused);
95 
96 int
97 main(int argc, char *argv[])
98 {
99 	struct stat to_stat, tmp_stat;
100 	enum op type;
101 	int Hflag, Lflag, Pflag, ch, fts_options, r, have_trailing_slash;
102 	char *target;
103 
104 	fts_options = FTS_NOCHDIR | FTS_PHYSICAL;
105 	Hflag = Lflag = Pflag = 0;
106 	while ((ch = getopt(argc, argv, "HLPRafilnprvx")) != -1)
107 		switch (ch) {
108 		case 'H':
109 			Hflag = 1;
110 			Lflag = Pflag = 0;
111 			break;
112 		case 'L':
113 			Lflag = 1;
114 			Hflag = Pflag = 0;
115 			break;
116 		case 'P':
117 			Pflag = 1;
118 			Hflag = Lflag = 0;
119 			break;
120 		case 'R':
121 			Rflag = 1;
122 			break;
123 		case 'a':
124 			Pflag = 1;
125 			pflag = 1;
126 			Rflag = 1;
127 			Hflag = Lflag = 0;
128 			break;
129 		case 'f':
130 			fflag = 1;
131 			iflag = nflag = 0;
132 			break;
133 		case 'i':
134 			iflag = 1;
135 			fflag = nflag = 0;
136 			break;
137 		case 'l':
138 			lflag = 1;
139 			break;
140 		case 'n':
141 			nflag = 1;
142 			fflag = iflag = 0;
143 			break;
144 		case 'p':
145 			pflag = 1;
146 			break;
147 		case 'r':
148 			rflag = Lflag = 1;
149 			Hflag = Pflag = 0;
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 (rflag)
170 		Rflag = 1;
171 	if (Rflag) {
172 		if (Hflag)
173 			fts_options |= FTS_COMFOLLOW;
174 		if (Lflag) {
175 			fts_options &= ~FTS_PHYSICAL;
176 			fts_options |= FTS_LOGICAL;
177 		}
178 	} else {
179 		fts_options &= ~FTS_PHYSICAL;
180 		fts_options |= FTS_LOGICAL | FTS_COMFOLLOW;
181 	}
182 	(void)signal(SIGINFO, siginfo);
183 
184 	/* Save the target base in "to". */
185 	target = argv[--argc];
186 	if (strlcpy(to.p_path, target, sizeof(to.p_path)) >= sizeof(to.p_path))
187 		errx(1, "%s: name too long", target);
188 	to.p_end = to.p_path + strlen(to.p_path);
189         if (to.p_path == to.p_end) {
190 		*to.p_end++ = '.';
191 		*to.p_end = 0;
192 	}
193 	have_trailing_slash = (to.p_end[-1] == '/');
194 	if (have_trailing_slash)
195 		STRIP_TRAILING_SLASH(to);
196 	to.target_end = to.p_end;
197 
198 	/* Set end of argument list for fts(3). */
199 	argv[argc] = NULL;
200 
201 	/*
202 	 * Cp has two distinct cases:
203 	 *
204 	 * cp [-R] source target
205 	 * cp [-R] source1 ... sourceN directory
206 	 *
207 	 * In both cases, source can be either a file or a directory.
208 	 *
209 	 * In (1), the target becomes a copy of the source. That is, if the
210 	 * source is a file, the target will be a file, and likewise for
211 	 * directories.
212 	 *
213 	 * In (2), the real target is not directory, but "directory/source".
214 	 */
215 	r = stat(to.p_path, &to_stat);
216 	if (r == -1 && errno != ENOENT)
217 		err(1, "%s", to.p_path);
218 	if (r == -1 || !S_ISDIR(to_stat.st_mode)) {
219 		/*
220 		 * Case (1).  Target is not a directory.
221 		 */
222 		if (argc > 1)
223 			errx(1, "%s is not a directory", to.p_path);
224 
225 		/*
226 		 * Need to detect the case:
227 		 *	cp -R dir foo
228 		 * Where dir is a directory and foo does not exist, where
229 		 * we want pathname concatenations turned on but not for
230 		 * the initial mkdir().
231 		 */
232 		if (r == -1) {
233 			if (Rflag && (Lflag || Hflag))
234 				stat(*argv, &tmp_stat);
235 			else
236 				lstat(*argv, &tmp_stat);
237 
238 			if (S_ISDIR(tmp_stat.st_mode) && Rflag)
239 				type = DIR_TO_DNE;
240 			else
241 				type = FILE_TO_FILE;
242 		} else
243 			type = FILE_TO_FILE;
244 
245 		if (have_trailing_slash && type == FILE_TO_FILE) {
246 			if (r == -1)
247 				errx(1, "directory %s does not exist",
248 				     to.p_path);
249 			else
250 				errx(1, "%s is not a directory", to.p_path);
251 		}
252 	} else
253 		/*
254 		 * Case (2).  Target is a directory.
255 		 */
256 		type = FILE_TO_DIR;
257 
258 	exit (copy(argv, type, fts_options));
259 }
260 
261 static int
262 copy(char *argv[], enum op type, int fts_options)
263 {
264 	struct stat to_stat;
265 	FTS *ftsp;
266 	FTSENT *curr;
267 	int base = 0, dne, badcp, rval;
268 	size_t nlen;
269 	char *p, *target_mid;
270 	mode_t mask, mode;
271 
272 	/*
273 	 * Keep an inverted copy of the umask, for use in correcting
274 	 * permissions on created directories when not using -p.
275 	 */
276 	mask = ~umask(0777);
277 	umask(~mask);
278 
279 	if ((ftsp = fts_open(argv, fts_options, mastercmp)) == NULL)
280 		err(1, "fts_open");
281 	for (badcp = rval = 0; (curr = fts_read(ftsp)) != NULL; badcp = 0) {
282 		switch (curr->fts_info) {
283 		case FTS_NS:
284 		case FTS_DNR:
285 		case FTS_ERR:
286 			warnx("%s: %s",
287 			    curr->fts_path, strerror(curr->fts_errno));
288 			badcp = rval = 1;
289 			continue;
290 		case FTS_DC:			/* Warn, continue. */
291 			warnx("%s: directory causes a cycle", curr->fts_path);
292 			badcp = rval = 1;
293 			continue;
294 		default:
295 			;
296 		}
297 
298 		/*
299 		 * If we are in case (2) or (3) above, we need to append the
300                  * source name to the target name.
301                  */
302 		if (type != FILE_TO_FILE) {
303 			/*
304 			 * Need to remember the roots of traversals to create
305 			 * correct pathnames.  If there's a directory being
306 			 * copied to a non-existent directory, e.g.
307 			 *	cp -R a/dir noexist
308 			 * the resulting path name should be noexist/foo, not
309 			 * noexist/dir/foo (where foo is a file in dir), which
310 			 * is the case where the target exists.
311 			 *
312 			 * Also, check for "..".  This is for correct path
313 			 * concatenation for paths ending in "..", e.g.
314 			 *	cp -R .. /tmp
315 			 * Paths ending in ".." are changed to ".".  This is
316 			 * tricky, but seems the easiest way to fix the problem.
317 			 *
318 			 * XXX
319 			 * Since the first level MUST be FTS_ROOTLEVEL, base
320 			 * is always initialized.
321 			 */
322 			if (curr->fts_level == FTS_ROOTLEVEL) {
323 				if (type != DIR_TO_DNE) {
324 					p = strrchr(curr->fts_path, '/');
325 					base = (p == NULL) ? 0 :
326 					    (int)(p - curr->fts_path + 1);
327 
328 					if (!strcmp(&curr->fts_path[base],
329 					    ".."))
330 						base += 1;
331 				} else
332 					base = curr->fts_pathlen;
333 			}
334 
335 			p = &curr->fts_path[base];
336 			nlen = curr->fts_pathlen - base;
337 			target_mid = to.target_end;
338 			if (*p != '/' && target_mid[-1] != '/')
339 				*target_mid++ = '/';
340 			*target_mid = 0;
341 			if (target_mid - to.p_path + nlen >= PATH_MAX) {
342 				warnx("%s%s: name too long (not copied)",
343 				    to.p_path, p);
344 				badcp = rval = 1;
345 				continue;
346 			}
347 			(void)strncat(target_mid, p, nlen);
348 			to.p_end = target_mid + nlen;
349 			*to.p_end = 0;
350 			STRIP_TRAILING_SLASH(to);
351 		}
352 
353 		if (curr->fts_info == FTS_DP) {
354 			/*
355 			 * We are nearly finished with this directory.  If we
356 			 * didn't actually copy it, or otherwise don't need to
357 			 * change its attributes, then we are done.
358 			 */
359 			if (!curr->fts_number)
360 				continue;
361 			/*
362 			 * If -p is in effect, set all the attributes.
363 			 * Otherwise, set the correct permissions, limited
364 			 * by the umask.  Optimise by avoiding a chmod()
365 			 * if possible (which is usually the case if we
366 			 * made the directory).  Note that mkdir() does not
367 			 * honour setuid, setgid and sticky bits, but we
368 			 * normally want to preserve them on directories.
369 			 */
370 			if (pflag) {
371 				if (setfile(curr->fts_statp, -1))
372 					rval = 1;
373 				if (preserve_dir_acls(curr->fts_statp,
374 				    curr->fts_accpath, to.p_path) != 0)
375 					rval = 1;
376 			} else {
377 				mode = curr->fts_statp->st_mode;
378 				if ((mode & (S_ISUID | S_ISGID | S_ISTXT)) ||
379 				    ((mode | S_IRWXU) & mask) != (mode & mask))
380 					if (chmod(to.p_path, mode & mask) != 0){
381 						warn("chmod: %s", to.p_path);
382 						rval = 1;
383 					}
384 			}
385 			continue;
386 		}
387 
388 		/* Not an error but need to remember it happened */
389 		if (stat(to.p_path, &to_stat) == -1)
390 			dne = 1;
391 		else {
392 			if (to_stat.st_dev == curr->fts_statp->st_dev &&
393 			    to_stat.st_ino == curr->fts_statp->st_ino) {
394 				warnx("%s and %s are identical (not copied).",
395 				    to.p_path, curr->fts_path);
396 				badcp = rval = 1;
397 				if (S_ISDIR(curr->fts_statp->st_mode))
398 					(void)fts_set(ftsp, curr, FTS_SKIP);
399 				continue;
400 			}
401 			if (!S_ISDIR(curr->fts_statp->st_mode) &&
402 			    S_ISDIR(to_stat.st_mode)) {
403 				warnx("cannot overwrite directory %s with "
404 				    "non-directory %s",
405 				    to.p_path, curr->fts_path);
406 				badcp = rval = 1;
407 				continue;
408 			}
409 			dne = 0;
410 		}
411 
412 		switch (curr->fts_statp->st_mode & S_IFMT) {
413 		case S_IFLNK:
414 			/* Catch special case of a non-dangling symlink */
415 			if ((fts_options & FTS_LOGICAL) ||
416 			    ((fts_options & FTS_COMFOLLOW) &&
417 			    curr->fts_level == 0)) {
418 				if (copy_file(curr, dne))
419 					badcp = rval = 1;
420 			} else {
421 				if (copy_link(curr, !dne))
422 					badcp = rval = 1;
423 			}
424 			break;
425 		case S_IFDIR:
426 			if (!Rflag) {
427 				warnx("%s is a directory (not copied).",
428 				    curr->fts_path);
429 				(void)fts_set(ftsp, curr, FTS_SKIP);
430 				badcp = rval = 1;
431 				break;
432 			}
433 			/*
434 			 * If the directory doesn't exist, create the new
435 			 * one with the from file mode plus owner RWX bits,
436 			 * modified by the umask.  Trade-off between being
437 			 * able to write the directory (if from directory is
438 			 * 555) and not causing a permissions race.  If the
439 			 * umask blocks owner writes, we fail..
440 			 */
441 			if (dne) {
442 				if (mkdir(to.p_path,
443 				    curr->fts_statp->st_mode | S_IRWXU) < 0)
444 					err(1, "%s", to.p_path);
445 			} else if (!S_ISDIR(to_stat.st_mode)) {
446 				errno = ENOTDIR;
447 				err(1, "%s", to.p_path);
448 			}
449 			/*
450 			 * Arrange to correct directory attributes later
451 			 * (in the post-order phase) if this is a new
452 			 * directory, or if the -p flag is in effect.
453 			 */
454 			curr->fts_number = pflag || dne;
455 			break;
456 		case S_IFBLK:
457 		case S_IFCHR:
458 			if (Rflag) {
459 				if (copy_special(curr->fts_statp, !dne))
460 					badcp = rval = 1;
461 			} else {
462 				if (copy_file(curr, dne))
463 					badcp = rval = 1;
464 			}
465 			break;
466 		case S_IFSOCK:
467 			warnx("%s is a socket (not copied).",
468 				    curr->fts_path);
469 			break;
470 		case S_IFIFO:
471 			if (Rflag) {
472 				if (copy_fifo(curr->fts_statp, !dne))
473 					badcp = rval = 1;
474 			} else {
475 				if (copy_file(curr, dne))
476 					badcp = rval = 1;
477 			}
478 			break;
479 		default:
480 			if (copy_file(curr, dne))
481 				badcp = rval = 1;
482 			break;
483 		}
484 		if (vflag && !badcp)
485 			(void)printf("%s -> %s\n", curr->fts_path, to.p_path);
486 	}
487 	if (errno)
488 		err(1, "fts_read");
489 	fts_close(ftsp);
490 	return (rval);
491 }
492 
493 /*
494  * mastercmp --
495  *	The comparison function for the copy order.  The order is to copy
496  *	non-directory files before directory files.  The reason for this
497  *	is because files tend to be in the same cylinder group as their
498  *	parent directory, whereas directories tend not to be.  Copying the
499  *	files first reduces seeking.
500  */
501 static int
502 mastercmp(const FTSENT * const *a, const FTSENT * const *b)
503 {
504 	int a_info, b_info;
505 
506 	a_info = (*a)->fts_info;
507 	if (a_info == FTS_ERR || a_info == FTS_NS || a_info == FTS_DNR)
508 		return (0);
509 	b_info = (*b)->fts_info;
510 	if (b_info == FTS_ERR || b_info == FTS_NS || b_info == FTS_DNR)
511 		return (0);
512 	if (a_info == FTS_D)
513 		return (-1);
514 	if (b_info == FTS_D)
515 		return (1);
516 	return (0);
517 }
518 
519 static void
520 siginfo(int sig __unused)
521 {
522 
523 	info = 1;
524 }
525