xref: /freebsd/bin/cp/cp.c (revision 730cecb05aaf016ac52ef7cfc691ccec3a0408cd)
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, ch, fts_options, r, have_trailing_slash;
102 	char *target;
103 
104 	fts_options = FTS_NOCHDIR | FTS_PHYSICAL;
105 	Hflag = Lflag = 0;
106 	while ((ch = getopt(argc, argv, "HLPRafilnprvx")) != -1)
107 		switch (ch) {
108 		case 'H':
109 			Hflag = 1;
110 			Lflag = 0;
111 			break;
112 		case 'L':
113 			Lflag = 1;
114 			Hflag = 0;
115 			break;
116 		case 'P':
117 			Hflag = Lflag = 0;
118 			break;
119 		case 'R':
120 			Rflag = 1;
121 			break;
122 		case 'a':
123 			pflag = 1;
124 			Rflag = 1;
125 			Hflag = Lflag = 0;
126 			break;
127 		case 'f':
128 			fflag = 1;
129 			iflag = nflag = 0;
130 			break;
131 		case 'i':
132 			iflag = 1;
133 			fflag = nflag = 0;
134 			break;
135 		case 'l':
136 			lflag = 1;
137 			break;
138 		case 'n':
139 			nflag = 1;
140 			fflag = iflag = 0;
141 			break;
142 		case 'p':
143 			pflag = 1;
144 			break;
145 		case 'r':
146 			rflag = Lflag = 1;
147 			Hflag = 0;
148 			break;
149 		case 'v':
150 			vflag = 1;
151 			break;
152 		case 'x':
153 			fts_options |= FTS_XDEV;
154 			break;
155 		default:
156 			usage();
157 			break;
158 		}
159 	argc -= optind;
160 	argv += optind;
161 
162 	if (argc < 2)
163 		usage();
164 
165 	if (Rflag && rflag)
166 		errx(1, "the -R and -r options may not be specified together");
167 	if (rflag)
168 		Rflag = 1;
169 	if (Rflag) {
170 		if (Hflag)
171 			fts_options |= FTS_COMFOLLOW;
172 		if (Lflag) {
173 			fts_options &= ~FTS_PHYSICAL;
174 			fts_options |= FTS_LOGICAL;
175 		}
176 	} else {
177 		fts_options &= ~FTS_PHYSICAL;
178 		fts_options |= FTS_LOGICAL | FTS_COMFOLLOW;
179 	}
180 	(void)signal(SIGINFO, siginfo);
181 
182 	/* Save the target base in "to". */
183 	target = argv[--argc];
184 	if (strlcpy(to.p_path, target, sizeof(to.p_path)) >= sizeof(to.p_path))
185 		errx(1, "%s: name too long", target);
186 	to.p_end = to.p_path + strlen(to.p_path);
187         if (to.p_path == to.p_end) {
188 		*to.p_end++ = '.';
189 		*to.p_end = 0;
190 	}
191 	have_trailing_slash = (to.p_end[-1] == '/');
192 	if (have_trailing_slash)
193 		STRIP_TRAILING_SLASH(to);
194 	to.target_end = to.p_end;
195 
196 	/* Set end of argument list for fts(3). */
197 	argv[argc] = NULL;
198 
199 	/*
200 	 * Cp has two distinct cases:
201 	 *
202 	 * cp [-R] source target
203 	 * cp [-R] source1 ... sourceN directory
204 	 *
205 	 * In both cases, source can be either a file or a directory.
206 	 *
207 	 * In (1), the target becomes a copy of the source. That is, if the
208 	 * source is a file, the target will be a file, and likewise for
209 	 * directories.
210 	 *
211 	 * In (2), the real target is not directory, but "directory/source".
212 	 */
213 	r = stat(to.p_path, &to_stat);
214 	if (r == -1 && errno != ENOENT)
215 		err(1, "%s", to.p_path);
216 	if (r == -1 || !S_ISDIR(to_stat.st_mode)) {
217 		/*
218 		 * Case (1).  Target is not a directory.
219 		 */
220 		if (argc > 1)
221 			errx(1, "%s is not a directory", to.p_path);
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 && (Lflag || Hflag))
232 				stat(*argv, &tmp_stat);
233 			else
234 				lstat(*argv, &tmp_stat);
235 
236 			if (S_ISDIR(tmp_stat.st_mode) && 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 				if (preserve_dir_acls(curr->fts_statp,
372 				    curr->fts_accpath, to.p_path) != 0)
373 					rval = 1;
374 			} else {
375 				mode = curr->fts_statp->st_mode;
376 				if ((mode & (S_ISUID | S_ISGID | S_ISTXT)) ||
377 				    ((mode | S_IRWXU) & mask) != (mode & mask))
378 					if (chmod(to.p_path, mode & mask) != 0){
379 						warn("chmod: %s", to.p_path);
380 						rval = 1;
381 					}
382 			}
383 			continue;
384 		}
385 
386 		/* Not an error but need to remember it happened */
387 		if (stat(to.p_path, &to_stat) == -1)
388 			dne = 1;
389 		else {
390 			if (to_stat.st_dev == curr->fts_statp->st_dev &&
391 			    to_stat.st_ino == curr->fts_statp->st_ino) {
392 				warnx("%s and %s are identical (not copied).",
393 				    to.p_path, curr->fts_path);
394 				badcp = rval = 1;
395 				if (S_ISDIR(curr->fts_statp->st_mode))
396 					(void)fts_set(ftsp, curr, FTS_SKIP);
397 				continue;
398 			}
399 			if (!S_ISDIR(curr->fts_statp->st_mode) &&
400 			    S_ISDIR(to_stat.st_mode)) {
401 				warnx("cannot overwrite directory %s with "
402 				    "non-directory %s",
403 				    to.p_path, curr->fts_path);
404 				badcp = rval = 1;
405 				continue;
406 			}
407 			dne = 0;
408 		}
409 
410 		switch (curr->fts_statp->st_mode & S_IFMT) {
411 		case S_IFLNK:
412 			/* Catch special case of a non-dangling symlink */
413 			if ((fts_options & FTS_LOGICAL) ||
414 			    ((fts_options & FTS_COMFOLLOW) &&
415 			    curr->fts_level == 0)) {
416 				if (copy_file(curr, dne))
417 					badcp = rval = 1;
418 			} else {
419 				if (copy_link(curr, !dne))
420 					badcp = rval = 1;
421 			}
422 			break;
423 		case S_IFDIR:
424 			if (!Rflag) {
425 				warnx("%s is a directory (not copied).",
426 				    curr->fts_path);
427 				(void)fts_set(ftsp, curr, FTS_SKIP);
428 				badcp = rval = 1;
429 				break;
430 			}
431 			/*
432 			 * If the directory doesn't exist, create the new
433 			 * one with the from file mode plus owner RWX bits,
434 			 * modified by the umask.  Trade-off between being
435 			 * able to write the directory (if from directory is
436 			 * 555) and not causing a permissions race.  If the
437 			 * umask blocks owner writes, we fail..
438 			 */
439 			if (dne) {
440 				if (mkdir(to.p_path,
441 				    curr->fts_statp->st_mode | S_IRWXU) < 0)
442 					err(1, "%s", to.p_path);
443 			} else if (!S_ISDIR(to_stat.st_mode)) {
444 				errno = ENOTDIR;
445 				err(1, "%s", to.p_path);
446 			}
447 			/*
448 			 * Arrange to correct directory attributes later
449 			 * (in the post-order phase) if this is a new
450 			 * directory, or if the -p flag is in effect.
451 			 */
452 			curr->fts_number = pflag || dne;
453 			break;
454 		case S_IFBLK:
455 		case S_IFCHR:
456 			if (Rflag) {
457 				if (copy_special(curr->fts_statp, !dne))
458 					badcp = rval = 1;
459 			} else {
460 				if (copy_file(curr, dne))
461 					badcp = rval = 1;
462 			}
463 			break;
464 		case S_IFSOCK:
465 			warnx("%s is a socket (not copied).",
466 				    curr->fts_path);
467 			break;
468 		case S_IFIFO:
469 			if (Rflag) {
470 				if (copy_fifo(curr->fts_statp, !dne))
471 					badcp = rval = 1;
472 			} else {
473 				if (copy_file(curr, dne))
474 					badcp = rval = 1;
475 			}
476 			break;
477 		default:
478 			if (copy_file(curr, dne))
479 				badcp = rval = 1;
480 			break;
481 		}
482 		if (vflag && !badcp)
483 			(void)printf("%s -> %s\n", curr->fts_path, to.p_path);
484 	}
485 	if (errno)
486 		err(1, "fts_read");
487 	fts_close(ftsp);
488 	return (rval);
489 }
490 
491 /*
492  * mastercmp --
493  *	The comparison function for the copy order.  The order is to copy
494  *	non-directory files before directory files.  The reason for this
495  *	is because files tend to be in the same cylinder group as their
496  *	parent directory, whereas directories tend not to be.  Copying the
497  *	files first reduces seeking.
498  */
499 static int
500 mastercmp(const FTSENT * const *a, const FTSENT * const *b)
501 {
502 	int a_info, b_info;
503 
504 	a_info = (*a)->fts_info;
505 	if (a_info == FTS_ERR || a_info == FTS_NS || a_info == FTS_DNR)
506 		return (0);
507 	b_info = (*b)->fts_info;
508 	if (b_info == FTS_ERR || b_info == FTS_NS || b_info == FTS_DNR)
509 		return (0);
510 	if (a_info == FTS_D)
511 		return (-1);
512 	if (b_info == FTS_D)
513 		return (1);
514 	return (0);
515 }
516 
517 static void
518 siginfo(int sig __unused)
519 {
520 
521 	info = 1;
522 }
523