xref: /freebsd/bin/cp/cp.c (revision 7660b554bc59a07be0431c17e0e33815818baa69)
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 #if 0
38 #ifndef lint
39 static char const copyright[] =
40 "@(#) Copyright (c) 1988, 1993, 1994\n\
41 	The Regents of the University of California.  All rights reserved.\n";
42 #endif /* not lint */
43 
44 #ifndef lint
45 static char sccsid[] = "@(#)cp.c	8.2 (Berkeley) 4/1/94";
46 #endif /* not lint */
47 #endif
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/types.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 <signal.h>
74 #include <stdio.h>
75 #include <stdlib.h>
76 #include <string.h>
77 #include <unistd.h>
78 
79 #include "extern.h"
80 
81 #define	STRIP_TRAILING_SLASH(p) {					\
82         while ((p).p_end > (p).p_path + 1 && (p).p_end[-1] == '/')	\
83                 *--(p).p_end = 0;					\
84 }
85 
86 static char emptystring[] = "";
87 
88 PATH_T to = { to.p_path, emptystring, "" };
89 
90 int fflag, iflag, nflag, pflag, vflag;
91 static int Rflag, rflag;
92 volatile sig_atomic_t info;
93 
94 enum op { FILE_TO_FILE, FILE_TO_DIR, DIR_TO_DNE };
95 
96 static int copy(char *[], enum op, int);
97 static int mastercmp(const FTSENT * const *, const FTSENT * const *);
98 static void siginfo(int __unused);
99 
100 int
101 main(int argc, char *argv[])
102 {
103 	struct stat to_stat, tmp_stat;
104 	enum op type;
105 	int Hflag, Lflag, Pflag, ch, fts_options, r, have_trailing_slash;
106 	char *target;
107 
108 	Hflag = Lflag = Pflag = 0;
109 	while ((ch = getopt(argc, argv, "HLPRfinprv")) != -1)
110 		switch (ch) {
111 		case 'H':
112 			Hflag = 1;
113 			Lflag = Pflag = 0;
114 			break;
115 		case 'L':
116 			Lflag = 1;
117 			Hflag = Pflag = 0;
118 			break;
119 		case 'P':
120 			Pflag = 1;
121 			Hflag = Lflag = 0;
122 			break;
123 		case 'R':
124 			Rflag = 1;
125 			break;
126 		case 'f':
127 			fflag = 1;
128 			iflag = nflag = 0;
129 			break;
130 		case 'i':
131 			iflag = 1;
132 			fflag = nflag = 0;
133 			break;
134 		case 'n':
135 			nflag = 1;
136 			fflag = iflag = 0;
137 			break;
138 		case 'p':
139 			pflag = 1;
140 			break;
141 		case 'r':
142 			rflag = 1;
143 			break;
144 		case 'v':
145 			vflag = 1;
146 			break;
147 		default:
148 			usage();
149 			break;
150 		}
151 	argc -= optind;
152 	argv += optind;
153 
154 	if (argc < 2)
155 		usage();
156 
157 	fts_options = FTS_NOCHDIR | FTS_PHYSICAL;
158 	if (rflag) {
159 		if (Rflag)
160 			errx(1,
161 		    "the -R and -r options may not be specified together.");
162 		if (Hflag || Lflag || Pflag)
163 			errx(1,
164 	"the -H, -L, and -P options may not be specified with the -r option.");
165 		fts_options &= ~FTS_PHYSICAL;
166 		fts_options |= FTS_LOGICAL;
167 	}
168 	if (Rflag) {
169 		if (Hflag)
170 			fts_options |= FTS_COMFOLLOW;
171 		if (Lflag) {
172 			fts_options &= ~FTS_PHYSICAL;
173 			fts_options |= FTS_LOGICAL;
174 		}
175 	} else {
176 		fts_options &= ~FTS_PHYSICAL;
177 		fts_options |= FTS_LOGICAL | FTS_COMFOLLOW;
178 	}
179 	(void)signal(SIGINFO, siginfo);
180 
181 	/* Save the target base in "to". */
182 	target = argv[--argc];
183 	if (strlcpy(to.p_path, target, sizeof(to.p_path)) >= sizeof(to.p_path))
184 		errx(1, "%s: name too long", target);
185 	to.p_end = to.p_path + strlen(to.p_path);
186         if (to.p_path == to.p_end) {
187 		*to.p_end++ = '.';
188 		*to.p_end = 0;
189 	}
190 	have_trailing_slash = (to.p_end[-1] == '/');
191 	if (have_trailing_slash)
192 		STRIP_TRAILING_SLASH(to);
193 	to.target_end = to.p_end;
194 
195 	/* Set end of argument list for fts(3). */
196 	argv[argc] = NULL;
197 
198 	/*
199 	 * Cp has two distinct cases:
200 	 *
201 	 * cp [-R] source target
202 	 * cp [-R] source1 ... sourceN directory
203 	 *
204 	 * In both cases, source can be either a file or a directory.
205 	 *
206 	 * In (1), the target becomes a copy of the source. That is, if the
207 	 * source is a file, the target will be a file, and likewise for
208 	 * directories.
209 	 *
210 	 * In (2), the real target is not directory, but "directory/source".
211 	 */
212 	r = stat(to.p_path, &to_stat);
213 	if (r == -1 && errno != ENOENT)
214 		err(1, "%s", to.p_path);
215 	if (r == -1 || !S_ISDIR(to_stat.st_mode)) {
216 		/*
217 		 * Case (1).  Target is not a directory.
218 		 */
219 		if (argc > 1) {
220 			usage();
221 			exit(1);
222 		}
223 		/*
224 		 * Need to detect the case:
225 		 *	cp -R dir foo
226 		 * Where dir is a directory and foo does not exist, where
227 		 * we want pathname concatenations turned on but not for
228 		 * the initial mkdir().
229 		 */
230 		if (r == -1) {
231 			if (rflag || (Rflag && (Lflag || Hflag)))
232 				stat(*argv, &tmp_stat);
233 			else
234 				lstat(*argv, &tmp_stat);
235 
236 			if (S_ISDIR(tmp_stat.st_mode) && (Rflag || rflag))
237 				type = DIR_TO_DNE;
238 			else
239 				type = FILE_TO_FILE;
240 		} else
241 			type = FILE_TO_FILE;
242 
243 		if (have_trailing_slash && type == FILE_TO_FILE) {
244 			if (r == -1)
245 				errx(1, "directory %s does not exist",
246 				     to.p_path);
247 			else
248 				errx(1, "%s is not a directory", to.p_path);
249 		}
250 	} else
251 		/*
252 		 * Case (2).  Target is a directory.
253 		 */
254 		type = FILE_TO_DIR;
255 
256 	exit (copy(argv, type, fts_options));
257 }
258 
259 static int
260 copy(char *argv[], enum op type, int fts_options)
261 {
262 	struct stat to_stat;
263 	FTS *ftsp;
264 	FTSENT *curr;
265 	int base = 0, dne, badcp, rval;
266 	size_t nlen;
267 	char *p, *target_mid;
268 	mode_t mask, mode;
269 
270 	/*
271 	 * Keep an inverted copy of the umask, for use in correcting
272 	 * permissions on created directories when not using -p.
273 	 */
274 	mask = ~umask(0777);
275 	umask(~mask);
276 
277 	if ((ftsp = fts_open(argv, fts_options, mastercmp)) == NULL)
278 		err(1, "fts_open");
279 	for (badcp = rval = 0; (curr = fts_read(ftsp)) != NULL; badcp = 0) {
280 		switch (curr->fts_info) {
281 		case FTS_NS:
282 		case FTS_DNR:
283 		case FTS_ERR:
284 			warnx("%s: %s",
285 			    curr->fts_path, strerror(curr->fts_errno));
286 			badcp = rval = 1;
287 			continue;
288 		case FTS_DC:			/* Warn, continue. */
289 			warnx("%s: directory causes a cycle", curr->fts_path);
290 			badcp = rval = 1;
291 			continue;
292 		default:
293 			;
294 		}
295 
296 		/*
297 		 * If we are in case (2) or (3) above, we need to append the
298                  * source name to the target name.
299                  */
300 		if (type != FILE_TO_FILE) {
301 			/*
302 			 * Need to remember the roots of traversals to create
303 			 * correct pathnames.  If there's a directory being
304 			 * copied to a non-existent directory, e.g.
305 			 *	cp -R a/dir noexist
306 			 * the resulting path name should be noexist/foo, not
307 			 * noexist/dir/foo (where foo is a file in dir), which
308 			 * is the case where the target exists.
309 			 *
310 			 * Also, check for "..".  This is for correct path
311 			 * concatenation for paths ending in "..", e.g.
312 			 *	cp -R .. /tmp
313 			 * Paths ending in ".." are changed to ".".  This is
314 			 * tricky, but seems the easiest way to fix the problem.
315 			 *
316 			 * XXX
317 			 * Since the first level MUST be FTS_ROOTLEVEL, base
318 			 * is always initialized.
319 			 */
320 			if (curr->fts_level == FTS_ROOTLEVEL) {
321 				if (type != DIR_TO_DNE) {
322 					p = strrchr(curr->fts_path, '/');
323 					base = (p == NULL) ? 0 :
324 					    (int)(p - curr->fts_path + 1);
325 
326 					if (!strcmp(&curr->fts_path[base],
327 					    ".."))
328 						base += 1;
329 				} else
330 					base = curr->fts_pathlen;
331 			}
332 
333 			p = &curr->fts_path[base];
334 			nlen = curr->fts_pathlen - base;
335 			target_mid = to.target_end;
336 			if (*p != '/' && target_mid[-1] != '/')
337 				*target_mid++ = '/';
338 			*target_mid = 0;
339 			if (target_mid - to.p_path + nlen >= PATH_MAX) {
340 				warnx("%s%s: name too long (not copied)",
341 				    to.p_path, p);
342 				badcp = rval = 1;
343 				continue;
344 			}
345 			(void)strncat(target_mid, p, nlen);
346 			to.p_end = target_mid + nlen;
347 			*to.p_end = 0;
348 			STRIP_TRAILING_SLASH(to);
349 		}
350 
351 		if (curr->fts_info == FTS_DP) {
352 			/*
353 			 * We are nearly finished with this directory.  If we
354 			 * didn't actually copy it, or otherwise don't need to
355 			 * change its attributes, then we are done.
356 			 */
357 			if (!curr->fts_number)
358 				continue;
359 			/*
360 			 * If -p is in effect, set all the attributes.
361 			 * Otherwise, set the correct permissions, limited
362 			 * by the umask.  Optimise by avoiding a chmod()
363 			 * if possible (which is usually the case if we
364 			 * made the directory).  Note that mkdir() does not
365 			 * honour setuid, setgid and sticky bits, but we
366 			 * normally want to preserve them on directories.
367 			 */
368 			if (pflag) {
369 				if (setfile(curr->fts_statp, -1))
370 				    rval = 1;
371 			} else {
372 				mode = curr->fts_statp->st_mode;
373 				if ((mode & (S_ISUID | S_ISGID | S_ISTXT)) ||
374 				    ((mode | S_IRWXU) & mask) != (mode & mask))
375 					if (chmod(to.p_path, mode & mask) != 0){
376 						warn("chmod: %s", to.p_path);
377 						rval = 1;
378 					}
379 			}
380 			continue;
381 		}
382 
383 		/* Not an error but need to remember it happened */
384 		if (stat(to.p_path, &to_stat) == -1)
385 			dne = 1;
386 		else {
387 			if (to_stat.st_dev == curr->fts_statp->st_dev &&
388 			    to_stat.st_ino == curr->fts_statp->st_ino) {
389 				warnx("%s and %s are identical (not copied).",
390 				    to.p_path, curr->fts_path);
391 				badcp = rval = 1;
392 				if (S_ISDIR(curr->fts_statp->st_mode))
393 					(void)fts_set(ftsp, curr, FTS_SKIP);
394 				continue;
395 			}
396 			if (!S_ISDIR(curr->fts_statp->st_mode) &&
397 			    S_ISDIR(to_stat.st_mode)) {
398 				warnx("cannot overwrite directory %s with "
399 				    "non-directory %s",
400 				    to.p_path, curr->fts_path);
401 				badcp = rval = 1;
402 				continue;
403 			}
404 			dne = 0;
405 		}
406 
407 		switch (curr->fts_statp->st_mode & S_IFMT) {
408 		case S_IFLNK:
409 			/* Catch special case of a non-dangling symlink */
410 			if ((fts_options & FTS_LOGICAL) ||
411 			    ((fts_options & FTS_COMFOLLOW) &&
412 			    curr->fts_level == 0)) {
413 				if (copy_file(curr, dne))
414 					badcp = rval = 1;
415 			} else {
416 				if (copy_link(curr, !dne))
417 					badcp = rval = 1;
418 			}
419 			break;
420 		case S_IFDIR:
421 			if (!Rflag && !rflag) {
422 				warnx("%s is a directory (not copied).",
423 				    curr->fts_path);
424 				(void)fts_set(ftsp, curr, FTS_SKIP);
425 				badcp = rval = 1;
426 				break;
427 			}
428 			/*
429 			 * If the directory doesn't exist, create the new
430 			 * one with the from file mode plus owner RWX bits,
431 			 * modified by the umask.  Trade-off between being
432 			 * able to write the directory (if from directory is
433 			 * 555) and not causing a permissions race.  If the
434 			 * umask blocks owner writes, we fail..
435 			 */
436 			if (dne) {
437 				if (mkdir(to.p_path,
438 				    curr->fts_statp->st_mode | S_IRWXU) < 0)
439 					err(1, "%s", to.p_path);
440 			} else if (!S_ISDIR(to_stat.st_mode)) {
441 				errno = ENOTDIR;
442 				err(1, "%s", to.p_path);
443 			}
444 			/*
445 			 * Arrange to correct directory attributes later
446 			 * (in the post-order phase) if this is a new
447 			 * directory, or if the -p flag is in effect.
448 			 */
449 			curr->fts_number = pflag || dne;
450 			break;
451 		case S_IFBLK:
452 		case S_IFCHR:
453 			if (Rflag) {
454 				if (copy_special(curr->fts_statp, !dne))
455 					badcp = rval = 1;
456 			} else {
457 				if (copy_file(curr, dne))
458 					badcp = rval = 1;
459 			}
460 			break;
461 		case S_IFIFO:
462 			if (Rflag) {
463 				if (copy_fifo(curr->fts_statp, !dne))
464 					badcp = rval = 1;
465 			} else {
466 				if (copy_file(curr, dne))
467 					badcp = rval = 1;
468 			}
469 			break;
470 		default:
471 			if (copy_file(curr, dne))
472 				badcp = rval = 1;
473 			break;
474 		}
475 		if (vflag && !badcp)
476 			(void)printf("%s -> %s\n", curr->fts_path, to.p_path);
477 	}
478 	if (errno)
479 		err(1, "fts_read");
480 	return (rval);
481 }
482 
483 /*
484  * mastercmp --
485  *	The comparison function for the copy order.  The order is to copy
486  *	non-directory files before directory files.  The reason for this
487  *	is because files tend to be in the same cylinder group as their
488  *	parent directory, whereas directories tend not to be.  Copying the
489  *	files first reduces seeking.
490  */
491 static int
492 mastercmp(const FTSENT * const *a, const FTSENT * const *b)
493 {
494 	int a_info, b_info;
495 
496 	a_info = (*a)->fts_info;
497 	if (a_info == FTS_ERR || a_info == FTS_NS || a_info == FTS_DNR)
498 		return (0);
499 	b_info = (*b)->fts_info;
500 	if (b_info == FTS_ERR || b_info == FTS_NS || b_info == FTS_DNR)
501 		return (0);
502 	if (a_info == FTS_D)
503 		return (-1);
504 	if (b_info == FTS_D)
505 		return (1);
506 	return (0);
507 }
508 
509 static void
510 siginfo(int sig __unused)
511 {
512 
513 	info = 1;
514 }
515