xref: /freebsd/bin/cp/cp.c (revision 6a66acb565dde6d5b0ee52eef2a631fbb09df153)
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  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  */
36 
37 #ifndef lint
38 static char const copyright[] =
39 "@(#) Copyright (c) 1988, 1993, 1994\n\
40 	The Regents of the University of California.  All rights reserved.\n";
41 #endif /* not lint */
42 
43 #ifndef lint
44 #if 0
45 static char sccsid[] = "@(#)cp.c	8.2 (Berkeley) 4/1/94";
46 #endif
47 #endif /* not lint */
48 #include <sys/cdefs.h>
49 __FBSDID("$FreeBSD$");
50 
51 /*
52  * Cp copies source files to target files.
53  *
54  * The global PATH_T structure "to" always contains the path to the
55  * current target file.  Since fts(3) does not change directories,
56  * this path can be either absolute or dot-relative.
57  *
58  * The basic algorithm is to initialize "to" and use fts(3) to traverse
59  * the file hierarchy rooted in the argument list.  A trivial case is the
60  * case of 'cp file1 file2'.  The more interesting case is the case of
61  * 'cp file1 file2 ... fileN dir' where the hierarchy is traversed and the
62  * path (relative to the root of the traversal) is appended to dir (stored
63  * in "to") to form the final target path.
64  */
65 
66 #include <sys/param.h>
67 #include <sys/stat.h>
68 
69 #include <err.h>
70 #include <errno.h>
71 #include <fts.h>
72 #include <limits.h>
73 #include <stdio.h>
74 #include <stdlib.h>
75 #include <string.h>
76 #include <unistd.h>
77 
78 #include "extern.h"
79 
80 #define	STRIP_TRAILING_SLASH(p) {					\
81         while ((p).p_end > (p).p_path + 1 && (p).p_end[-1] == '/')	\
82                 *--(p).p_end = 0;					\
83 }
84 
85 static char emptystring[] = "";
86 
87 PATH_T to = { to.p_path, emptystring, "" };
88 
89 int fflag, iflag, nflag, pflag, vflag;
90 static int Rflag, rflag;
91 
92 enum op { FILE_TO_FILE, FILE_TO_DIR, DIR_TO_DNE };
93 
94 static int copy(char *[], enum op, int);
95 static int mastercmp(const FTSENT * const *, const FTSENT * const *);
96 
97 int
98 main(int argc, char *argv[])
99 {
100 	struct stat to_stat, tmp_stat;
101 	enum op type;
102 	int Hflag, Lflag, Pflag, ch, fts_options, r, have_trailing_slash;
103 	char *target;
104 
105 	Hflag = Lflag = Pflag = 0;
106 	while ((ch = getopt(argc, argv, "HLPRfinprv")) != -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 'f':
124 			fflag = 1;
125 			iflag = nflag = 0;
126 			break;
127 		case 'i':
128 			iflag = 1;
129 			fflag = nflag = 0;
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 = 1;
140 			break;
141 		case 'v':
142 			vflag = 1;
143 			break;
144 		default:
145 			usage();
146 			break;
147 		}
148 	argc -= optind;
149 	argv += optind;
150 
151 	if (argc < 2)
152 		usage();
153 
154 	fts_options = FTS_NOCHDIR | FTS_PHYSICAL;
155 	if (rflag) {
156 		if (Rflag)
157 			errx(1,
158 		    "the -R and -r options may not be specified together.");
159 		if (Hflag || Lflag || Pflag)
160 			errx(1,
161 	"the -H, -L, and -P options may not be specified with the -r option.");
162 		fts_options &= ~FTS_PHYSICAL;
163 		fts_options |= FTS_LOGICAL;
164 	}
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 {
173 		fts_options &= ~FTS_PHYSICAL;
174 		fts_options |= FTS_LOGICAL | FTS_COMFOLLOW;
175 	}
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 			usage();
217 			exit(1);
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 || (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 || 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 	exit (copy(argv, type, fts_options));
253 }
254 
255 static int
256 copy(char *argv[], enum op type, int fts_options)
257 {
258 	struct stat to_stat;
259 	FTS *ftsp;
260 	FTSENT *curr;
261 	int base = 0, dne, badcp, rval;
262 	size_t nlen;
263 	char *p, *target_mid;
264 	mode_t mask, mode;
265 
266 	/*
267 	 * Keep an inverted copy of the umask, for use in correcting
268 	 * permissions on created directories when not using -p.
269 	 */
270 	mask = ~umask(0777);
271 	umask(~mask);
272 
273 	if ((ftsp = fts_open(argv, fts_options, mastercmp)) == NULL)
274 		err(1, "fts_open");
275 	for (badcp = rval = 0; (curr = fts_read(ftsp)) != NULL; badcp = 0) {
276 		switch (curr->fts_info) {
277 		case FTS_NS:
278 		case FTS_DNR:
279 		case FTS_ERR:
280 			warnx("%s: %s",
281 			    curr->fts_path, strerror(curr->fts_errno));
282 			badcp = rval = 1;
283 			continue;
284 		case FTS_DC:			/* Warn, continue. */
285 			warnx("%s: directory causes a cycle", curr->fts_path);
286 			badcp = rval = 1;
287 			continue;
288 		default:
289 			;
290 		}
291 
292 		/*
293 		 * If we are in case (2) or (3) above, we need to append the
294                  * source name to the target name.
295                  */
296 		if (type != FILE_TO_FILE) {
297 			/*
298 			 * Need to remember the roots of traversals to create
299 			 * correct pathnames.  If there's a directory being
300 			 * copied to a non-existent directory, e.g.
301 			 *	cp -R a/dir noexist
302 			 * the resulting path name should be noexist/foo, not
303 			 * noexist/dir/foo (where foo is a file in dir), which
304 			 * is the case where the target exists.
305 			 *
306 			 * Also, check for "..".  This is for correct path
307 			 * concatenation for paths ending in "..", e.g.
308 			 *	cp -R .. /tmp
309 			 * Paths ending in ".." are changed to ".".  This is
310 			 * tricky, but seems the easiest way to fix the problem.
311 			 *
312 			 * XXX
313 			 * Since the first level MUST be FTS_ROOTLEVEL, base
314 			 * is always initialized.
315 			 */
316 			if (curr->fts_level == FTS_ROOTLEVEL) {
317 				if (type != DIR_TO_DNE) {
318 					p = strrchr(curr->fts_path, '/');
319 					base = (p == NULL) ? 0 :
320 					    (int)(p - curr->fts_path + 1);
321 
322 					if (!strcmp(&curr->fts_path[base],
323 					    ".."))
324 						base += 1;
325 				} else
326 					base = curr->fts_pathlen;
327 			}
328 
329 			p = &curr->fts_path[base];
330 			nlen = curr->fts_pathlen - base;
331 			target_mid = to.target_end;
332 			if (*p != '/' && target_mid[-1] != '/')
333 				*target_mid++ = '/';
334 			*target_mid = 0;
335 			if (target_mid - to.p_path + nlen >= PATH_MAX) {
336 				warnx("%s%s: name too long (not copied)",
337 				    to.p_path, p);
338 				badcp = rval = 1;
339 				continue;
340 			}
341 			(void)strncat(target_mid, p, nlen);
342 			to.p_end = target_mid + nlen;
343 			*to.p_end = 0;
344 			STRIP_TRAILING_SLASH(to);
345 		}
346 
347 		if (curr->fts_info == FTS_DP) {
348 			/*
349 			 * We are nearly finished with this directory.  If we
350 			 * didn't actually copy it, or otherwise don't need to
351 			 * change its attributes, then we are done.
352 			 */
353 			if (!curr->fts_number)
354 				continue;
355 			/*
356 			 * If -p is in effect, set all the attributes.
357 			 * Otherwise, set the correct permissions, limited
358 			 * by the umask.  Optimise by avoiding a chmod()
359 			 * if possible (which is usually the case if we
360 			 * made the directory).  Note that mkdir() does not
361 			 * honour setuid, setgid and sticky bits, but we
362 			 * normally want to preserve them on directories.
363 			 */
364 			if (pflag) {
365 				if (setfile(curr->fts_statp, 0))
366 				    rval = 1;
367 			} else {
368 				mode = curr->fts_statp->st_mode;
369 				if ((mode & (S_ISUID | S_ISGID | S_ISTXT)) ||
370 				    ((mode | S_IRWXU) & mask) != (mode & mask))
371 					if (chmod(to.p_path, mode & mask) != 0){
372 						warn("chmod: %s", to.p_path);
373 						rval = 1;
374 					}
375 			}
376 			continue;
377 		}
378 
379 		/* Not an error but need to remember it happened */
380 		if (stat(to.p_path, &to_stat) == -1)
381 			dne = 1;
382 		else {
383 			if (to_stat.st_dev == curr->fts_statp->st_dev &&
384 			    to_stat.st_ino == curr->fts_statp->st_ino) {
385 				warnx("%s and %s are identical (not copied).",
386 				    to.p_path, curr->fts_path);
387 				badcp = rval = 1;
388 				if (S_ISDIR(curr->fts_statp->st_mode))
389 					(void)fts_set(ftsp, curr, FTS_SKIP);
390 				continue;
391 			}
392 			if (!S_ISDIR(curr->fts_statp->st_mode) &&
393 			    S_ISDIR(to_stat.st_mode)) {
394 				warnx("cannot overwrite directory %s with "
395 				    "non-directory %s",
396 				    to.p_path, curr->fts_path);
397 				badcp = rval = 1;
398 				continue;
399 			}
400 			dne = 0;
401 		}
402 
403 		switch (curr->fts_statp->st_mode & S_IFMT) {
404 		case S_IFLNK:
405 			/* Catch special case of a non-dangling symlink */
406 			if ((fts_options & FTS_LOGICAL) ||
407 			    ((fts_options & FTS_COMFOLLOW) &&
408 			    curr->fts_level == 0)) {
409 				if (copy_file(curr, dne))
410 					badcp = rval = 1;
411 			} else {
412 				if (copy_link(curr, !dne))
413 					badcp = rval = 1;
414 			}
415 			break;
416 		case S_IFDIR:
417 			if (!Rflag && !rflag) {
418 				warnx("%s is a directory (not copied).",
419 				    curr->fts_path);
420 				(void)fts_set(ftsp, curr, FTS_SKIP);
421 				badcp = rval = 1;
422 				break;
423 			}
424 			/*
425 			 * If the directory doesn't exist, create the new
426 			 * one with the from file mode plus owner RWX bits,
427 			 * modified by the umask.  Trade-off between being
428 			 * able to write the directory (if from directory is
429 			 * 555) and not causing a permissions race.  If the
430 			 * umask blocks owner writes, we fail..
431 			 */
432 			if (dne) {
433 				if (mkdir(to.p_path,
434 				    curr->fts_statp->st_mode | S_IRWXU) < 0)
435 					err(1, "%s", to.p_path);
436 			} else if (!S_ISDIR(to_stat.st_mode)) {
437 				errno = ENOTDIR;
438 				err(1, "%s", to.p_path);
439 			}
440 			/*
441 			 * Arrange to correct directory attributes later
442 			 * (in the post-order phase) if this is a new
443 			 * directory, or if the -p flag is in effect.
444 			 */
445 			curr->fts_number = pflag || dne;
446 			break;
447 		case S_IFBLK:
448 		case S_IFCHR:
449 			if (Rflag) {
450 				if (copy_special(curr->fts_statp, !dne))
451 					badcp = rval = 1;
452 			} else {
453 				if (copy_file(curr, dne))
454 					badcp = rval = 1;
455 			}
456 			break;
457 		case S_IFIFO:
458 			if (Rflag) {
459 				if (copy_fifo(curr->fts_statp, !dne))
460 					badcp = rval = 1;
461 			} else {
462 				if (copy_file(curr, dne))
463 					badcp = rval = 1;
464 			}
465 			break;
466 		default:
467 			if (copy_file(curr, dne))
468 				badcp = rval = 1;
469 			break;
470 		}
471 		if (vflag && !badcp)
472 			(void)printf("%s -> %s\n", curr->fts_path, to.p_path);
473 	}
474 	if (errno)
475 		err(1, "fts_read");
476 	return (rval);
477 }
478 
479 /*
480  * mastercmp --
481  *	The comparison function for the copy order.  The order is to copy
482  *	non-directory files before directory files.  The reason for this
483  *	is because files tend to be in the same cylinder group as their
484  *	parent directory, whereas directories tend not to be.  Copying the
485  *	files first reduces seeking.
486  */
487 static int
488 mastercmp(const FTSENT * const *a, const FTSENT * const *b)
489 {
490 	int a_info, b_info;
491 
492 	a_info = (*a)->fts_info;
493 	if (a_info == FTS_ERR || a_info == FTS_NS || a_info == FTS_DNR)
494 		return (0);
495 	b_info = (*b)->fts_info;
496 	if (b_info == FTS_ERR || b_info == FTS_NS || b_info == FTS_DNR)
497 		return (0);
498 	if (a_info == FTS_D)
499 		return (-1);
500 	if (b_info == FTS_D)
501 		return (1);
502 	return (0);
503 }
504