xref: /netbsd-src/bin/cp/cp.c (revision 7fa608457b817eca6e0977b37f758ae064f3c99c)
1 /* $NetBSD: cp.c,v 1.48 2006/12/26 00:13:24 alc Exp $ */
2 
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 #include <sys/cdefs.h>
36 #ifndef lint
37 __COPYRIGHT(
38 "@(#) Copyright (c) 1988, 1993, 1994\n\
39 	The Regents of the University of California.  All rights reserved.\n");
40 #endif /* not lint */
41 
42 #ifndef lint
43 #if 0
44 static char sccsid[] = "@(#)cp.c	8.5 (Berkeley) 4/29/95";
45 #else
46 __RCSID("$NetBSD: cp.c,v 1.48 2006/12/26 00:13:24 alc Exp $");
47 #endif
48 #endif /* not lint */
49 
50 /*
51  * Cp copies source files to target files.
52  *
53  * The global PATH_T structure "to" always contains the path to the
54  * current target file.  Since fts(3) does not change directories,
55  * this path can be either absolute or dot-relative.
56  *
57  * The basic algorithm is to initialize "to" and use fts(3) to traverse
58  * the file hierarchy rooted in the argument list.  A trivial case is the
59  * case of 'cp file1 file2'.  The more interesting case is the case of
60  * 'cp file1 file2 ... fileN dir' where the hierarchy is traversed and the
61  * path (relative to the root of the traversal) is appended to dir (stored
62  * in "to") to form the final target path.
63  */
64 
65 #include <sys/param.h>
66 #include <sys/stat.h>
67 
68 #include <err.h>
69 #include <errno.h>
70 #include <fts.h>
71 #include <locale.h>
72 #include <stdlib.h>
73 #include <stdio.h>
74 #include <string.h>
75 #include <unistd.h>
76 
77 #include "extern.h"
78 
79 #define	STRIP_TRAILING_SLASH(p) {					\
80         while ((p).p_end > (p).p_path + 1 && (p).p_end[-1] == '/')	\
81                 *--(p).p_end = '\0';					\
82 }
83 
84 static char empty[] = "";
85 PATH_T to = { .p_end = to.p_path, .target_end = empty  };
86 
87 uid_t myuid;
88 int Hflag, Lflag, Rflag, Pflag, fflag, iflag, pflag, rflag, vflag, Nflag;
89 mode_t myumask;
90 
91 enum op { FILE_TO_FILE, FILE_TO_DIR, DIR_TO_DNE };
92 
93 int 	main(int, char *[]);
94 int 	copy(char *[], enum op, int);
95 int 	mastercmp(const FTSENT **, const FTSENT **);
96 
97 int
98 main(int argc, char *argv[])
99 {
100 	struct stat to_stat, tmp_stat;
101 	enum op type;
102 	int ch, fts_options, r, have_trailing_slash;
103 	char *target, **src;
104 
105 	(void)setlocale(LC_ALL, "");
106 
107 	Hflag = Lflag = Pflag = Rflag = 0;
108 	while ((ch = getopt(argc, argv, "HLNPRfiprv")) != -1)
109 		switch (ch) {
110 		case 'H':
111 			Hflag = 1;
112 			Lflag = Pflag = 0;
113 			break;
114 		case 'L':
115 			Lflag = 1;
116 			Hflag = Pflag = 0;
117 			break;
118 		case 'N':
119 			Nflag = 1;
120 			break;
121 		case 'P':
122 			Pflag = 1;
123 			Hflag = Lflag = 0;
124 			break;
125 		case 'R':
126 			Rflag = 1;
127 			break;
128 		case 'f':
129 			fflag = 1;
130 			iflag = 0;
131 			break;
132 		case 'i':
133 			iflag = isatty(fileno(stdin));
134 			fflag = 0;
135 			break;
136 		case 'p':
137 			pflag = 1;
138 			break;
139 		case 'r':
140 			rflag = 1;
141 			break;
142 		case 'v':
143 			vflag = 1;
144 			break;
145 		case '?':
146 		default:
147 			usage();
148 			/* NOTREACHED */
149 		}
150 	argc -= optind;
151 	argv += optind;
152 
153 	if (argc < 2)
154 		usage();
155 
156 	fts_options = FTS_NOCHDIR | FTS_PHYSICAL;
157 	if (rflag) {
158 		if (Rflag) {
159 			errx(EXIT_FAILURE,
160 		    "the -R and -r options may not be specified together.");
161 			/* NOTREACHED */
162 		}
163 		if (Hflag || Lflag || Pflag) {
164 			errx(EXIT_FAILURE,
165 	"the -H, -L, and -P options may not be specified with the -r option.");
166 			/* NOTREACHED */
167 		}
168 		fts_options &= ~FTS_PHYSICAL;
169 		fts_options |= FTS_LOGICAL;
170 	}
171 
172 	if (Rflag) {
173 		if (Hflag)
174 			fts_options |= FTS_COMFOLLOW;
175 		if (Lflag) {
176 			fts_options &= ~FTS_PHYSICAL;
177 			fts_options |= FTS_LOGICAL;
178 		}
179 	} else if (!Pflag) {
180 		fts_options &= ~FTS_PHYSICAL;
181 		fts_options |= FTS_LOGICAL | FTS_COMFOLLOW;
182 	}
183 
184 	myuid = getuid();
185 
186 	/* Copy the umask for explicit mode setting. */
187 	myumask = umask(0);
188 	(void)umask(myumask);
189 
190 	/* Save the target base in "to". */
191 	target = argv[--argc];
192 	if (strlcpy(to.p_path, target, sizeof(to.p_path)) >= sizeof(to.p_path))
193 		errx(EXIT_FAILURE, "%s: name too long", target);
194 	to.p_end = to.p_path + strlen(to.p_path);
195 	have_trailing_slash = (to.p_end[-1] == '/');
196 	if (have_trailing_slash)
197 		STRIP_TRAILING_SLASH(to);
198 	to.target_end = to.p_end;
199 
200 	/* Set end of argument list for fts(3). */
201 	argv[argc] = NULL;
202 
203 	/*
204 	 * Cp has two distinct cases:
205 	 *
206 	 * cp [-R] source target
207 	 * cp [-R] source1 ... sourceN directory
208 	 *
209 	 * In both cases, source can be either a file or a directory.
210 	 *
211 	 * In (1), the target becomes a copy of the source. That is, if the
212 	 * source is a file, the target will be a file, and likewise for
213 	 * directories.
214 	 *
215 	 * In (2), the real target is not directory, but "directory/source".
216 	 */
217 	if (Pflag)
218 		r = lstat(to.p_path, &to_stat);
219 	else
220 		r = stat(to.p_path, &to_stat);
221 	if (r == -1 && errno != ENOENT) {
222 		err(EXIT_FAILURE, "%s", to.p_path);
223 		/* NOTREACHED */
224 	}
225 	if (r == -1 || !S_ISDIR(to_stat.st_mode)) {
226 		/*
227 		 * Case (1).  Target is not a directory.
228 		 */
229 		if (argc > 1)
230 			usage();
231 		/*
232 		 * Need to detect the case:
233 		 *	cp -R dir foo
234 		 * Where dir is a directory and foo does not exist, where
235 		 * we want pathname concatenations turned on but not for
236 		 * the initial mkdir().
237 		 */
238 		if (r == -1) {
239 			if (rflag || (Rflag && (Lflag || Hflag)))
240 				r = stat(*argv, &tmp_stat);
241 			else
242 				r = lstat(*argv, &tmp_stat);
243 			if (r == -1) {
244 				err(EXIT_FAILURE, "%s", *argv);
245 				/* NOTREACHED */
246 			}
247 
248 			if (S_ISDIR(tmp_stat.st_mode) && (Rflag || rflag))
249 				type = DIR_TO_DNE;
250 			else
251 				type = FILE_TO_FILE;
252 		} else
253 			type = FILE_TO_FILE;
254 
255 		if (have_trailing_slash && type == FILE_TO_FILE) {
256 			if (r == -1)
257 				errx(1, "directory %s does not exist",
258 				     to.p_path);
259 			else
260 				errx(1, "%s is not a directory", to.p_path);
261 		}
262 	} else {
263 		/*
264 		 * Case (2).  Target is a directory.
265 		 */
266 		type = FILE_TO_DIR;
267 	}
268 
269 	/*
270 	 * make "cp -rp src/ dst" behave like "cp -rp src dst" not
271 	 * like "cp -rp src/. dst"
272 	 */
273 	for (src = argv; *src; src++) {
274 		size_t len = strlen(*src);
275 		while (len-- > 1 && (*src)[len] == '/')
276 			(*src)[len] = '\0';
277 	}
278 
279 	exit(copy(argv, type, fts_options));
280 	/* NOTREACHED */
281 }
282 
283 int
284 copy(char *argv[], enum op type, int fts_options)
285 {
286 	struct stat to_stat;
287 	FTS *ftsp;
288 	FTSENT *curr;
289 	int base, dne, rval, sval;
290 	size_t nlen;
291 	char *p, *target_mid;
292 
293 	dne = 0;
294 	base = 0;	/* XXX gcc -Wuninitialized (see comment below) */
295 
296 	if ((ftsp = fts_open(argv, fts_options, mastercmp)) == NULL)
297 		err(EXIT_FAILURE, "%s", argv[0]);
298 		/* NOTREACHED */
299 	for (rval = 0; (curr = fts_read(ftsp)) != NULL;) {
300 		switch (curr->fts_info) {
301 		case FTS_NS:
302 		case FTS_DNR:
303 		case FTS_ERR:
304 			warnx("%s: %s", curr->fts_path,
305 					strerror(curr->fts_errno));
306 			rval = 1;
307 			continue;
308 		case FTS_DC:			/* Warn, continue. */
309 			warnx("%s: directory causes a cycle", curr->fts_path);
310 			rval = 1;
311 			continue;
312 		}
313 
314 		/*
315 		 * If we are in case (2) or (3) above, we need to append the
316                  * source name to the target name.
317                  */
318 		if (type != FILE_TO_FILE) {
319 			if ((curr->fts_namelen +
320 			    to.target_end - to.p_path + 1) > MAXPATHLEN) {
321 				warnx("%s/%s: name too long (not copied)",
322 						to.p_path, curr->fts_name);
323 				rval = 1;
324 				continue;
325 			}
326 
327 			/*
328 			 * Need to remember the roots of traversals to create
329 			 * correct pathnames.  If there's a directory being
330 			 * copied to a non-existent directory, e.g.
331 			 *	cp -R a/dir noexist
332 			 * the resulting path name should be noexist/foo, not
333 			 * noexist/dir/foo (where foo is a file in dir), which
334 			 * is the case where the target exists.
335 			 *
336 			 * Also, check for "..".  This is for correct path
337 			 * concatentation for paths ending in "..", e.g.
338 			 *	cp -R .. /tmp
339 			 * Paths ending in ".." are changed to ".".  This is
340 			 * tricky, but seems the easiest way to fix the problem.
341 			 *
342 			 * XXX
343 			 * Since the first level MUST be FTS_ROOTLEVEL, base
344 			 * is always initialized.
345 			 */
346 			if (curr->fts_level == FTS_ROOTLEVEL) {
347 				if (type != DIR_TO_DNE) {
348 					p = strrchr(curr->fts_path, '/');
349 					base = (p == NULL) ? 0 :
350 					    (int)(p - curr->fts_path + 1);
351 
352 					if (!strcmp(&curr->fts_path[base],
353 					    ".."))
354 						base += 1;
355 				} else
356 					base = curr->fts_pathlen;
357 			}
358 
359 			p = &curr->fts_path[base];
360 			nlen = curr->fts_pathlen - base;
361 			target_mid = to.target_end;
362 			if (*p != '/' && target_mid[-1] != '/')
363 				*target_mid++ = '/';
364 			*target_mid = 0;
365 
366 			if (target_mid - to.p_path + nlen >= PATH_MAX) {
367 				warnx("%s%s: name too long (not copied)",
368 				    to.p_path, p);
369 				rval = 1;
370 				continue;
371 			}
372 			(void)strncat(target_mid, p, nlen);
373 			to.p_end = target_mid + nlen;
374 			*to.p_end = 0;
375 			STRIP_TRAILING_SLASH(to);
376 		}
377 
378 		sval = Pflag ? lstat(to.p_path, &to_stat) : stat(to.p_path, &to_stat);
379 		/* Not an error but need to remember it happened */
380 		if (sval == -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 				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 non-directory %s",
395 				    to.p_path, curr->fts_path);
396 				rval = 1;
397 				continue;
398 			}
399 			if (!S_ISDIR(curr->fts_statp->st_mode))
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) && curr->fts_level == 0)) {
408 				if (copy_file(curr, dne))
409 					rval = 1;
410 			} else {
411 				if (copy_link(curr, !dne))
412 					rval = 1;
413 			}
414 			break;
415 		case S_IFDIR:
416 			if (!Rflag && !rflag) {
417 				if (curr->fts_info == FTS_D)
418 					warnx("%s is a directory (not copied).",
419 					    curr->fts_path);
420 				(void)fts_set(ftsp, curr, FTS_SKIP);
421 				rval = 1;
422 				break;
423 			}
424 
425                         /*
426                          * Directories get noticed twice:
427                          *  In the first pass, create it if needed.
428                          *  In the second pass, after the children have been copied, set the permissions.
429                          */
430 			if (curr->fts_info == FTS_D) /* First pass */
431 			{
432 				/*
433 				 * If the directory doesn't exist, create the new
434 				 * one with the from file mode plus owner RWX bits,
435 				 * modified by the umask.  Trade-off between being
436 				 * able to write the directory (if from directory is
437 				 * 555) and not causing a permissions race.  If the
438 				 * umask blocks owner writes, we fail..
439 				 */
440 				if (dne) {
441 					if (mkdir(to.p_path,
442 					    curr->fts_statp->st_mode | S_IRWXU) < 0)
443 						err(EXIT_FAILURE, "%s",
444 						    to.p_path);
445 						/* NOTREACHED */
446 				} else if (!S_ISDIR(to_stat.st_mode)) {
447 					errno = ENOTDIR;
448 					err(EXIT_FAILURE, "%s",
449 						to.p_path);
450 					/* NOTREACHED */
451 				}
452 			}
453 			else if (curr->fts_info == FTS_DP) /* Second pass */
454 			{
455 	                        /*
456 				 * If not -p and directory didn't exist, set it to be
457 				 * the same as the from directory, umodified by the
458                         	 * umask; arguably wrong, but it's been that way
459                         	 * forever.
460 				 */
461 				if (pflag && setfile(curr->fts_statp, 0))
462 					rval = 1;
463 				else if (dne)
464 					(void)chmod(to.p_path,
465 					    curr->fts_statp->st_mode);
466 
467 				/*
468 				 * Since this is the second pass, we already
469 				 * noted (and acted on) the existence of the
470 				 * directory.
471 				 */
472 				dne = 0;
473 			}
474 			else
475 			{
476 				warnx("directory %s encountered when not expected.",
477 				    curr->fts_path);
478 				rval = 1;
479 				break;
480 			}
481 
482 			break;
483 		case S_IFBLK:
484 		case S_IFCHR:
485 			if (Rflag) {
486 				if (copy_special(curr->fts_statp, !dne))
487 					rval = 1;
488 			} else
489 				if (copy_file(curr, dne))
490 					rval = 1;
491 			break;
492 		case S_IFIFO:
493 			if (Rflag) {
494 				if (copy_fifo(curr->fts_statp, !dne))
495 					rval = 1;
496 			} else
497 				if (copy_file(curr, dne))
498 					rval = 1;
499 			break;
500 		default:
501 			if (copy_file(curr, dne))
502 				rval = 1;
503 			break;
504 		}
505 		if (vflag && !rval)
506 			(void)printf("%s -> %s\n", curr->fts_path, to.p_path);
507 	}
508 	if (errno) {
509 		err(EXIT_FAILURE, "fts_read");
510 		/* NOTREACHED */
511 	}
512 	(void)fts_close(ftsp);
513 	return (rval);
514 }
515 
516 /*
517  * mastercmp --
518  *	The comparison function for the copy order.  The order is to copy
519  *	non-directory files before directory files.  The reason for this
520  *	is because files tend to be in the same cylinder group as their
521  *	parent directory, whereas directories tend not to be.  Copying the
522  *	files first reduces seeking.
523  */
524 int
525 mastercmp(const FTSENT **a, const FTSENT **b)
526 {
527 	int a_info, b_info;
528 
529 	a_info = (*a)->fts_info;
530 	if (a_info == FTS_ERR || a_info == FTS_NS || a_info == FTS_DNR)
531 		return (0);
532 	b_info = (*b)->fts_info;
533 	if (b_info == FTS_ERR || b_info == FTS_NS || b_info == FTS_DNR)
534 		return (0);
535 	if (a_info == FTS_D)
536 		return (-1);
537 	if (b_info == FTS_D)
538 		return (1);
539 	return (0);
540 }
541