xref: /netbsd-src/bin/cp/cp.c (revision f75f5aae154fcd0572e8889e4fea2a51d67bbf08)
1 /* $NetBSD: cp.c,v 1.53 2009/10/08 20:36:41 pooka 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\
39  The Regents of the University of California.  All rights reserved.");
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 1.53 2009/10/08 20:36:41 pooka 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 <assert.h>
69 #include <err.h>
70 #include <errno.h>
71 #include <fts.h>
72 #include <locale.h>
73 #include <stdlib.h>
74 #include <stdio.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 empty[] = "";
86 PATH_T to = { .p_end = to.p_path, .target_end = empty  };
87 
88 uid_t myuid;
89 int Hflag, Lflag, Rflag, Pflag, fflag, iflag, pflag, rflag, vflag, Nflag;
90 mode_t myumask;
91 
92 enum op { FILE_TO_FILE, FILE_TO_DIR, DIR_TO_DNE };
93 
94 int 	main(int, char *[]);
95 int 	copy(char *[], enum op, int);
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 	setprogname(argv[0]);
106 	(void)setlocale(LC_ALL, "");
107 
108 	Hflag = Lflag = Pflag = Rflag = 0;
109 	while ((ch = getopt(argc, argv, "HLNPRfiprv")) != -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 'N':
120 			Nflag = 1;
121 			break;
122 		case 'P':
123 			Pflag = 1;
124 			Hflag = Lflag = 0;
125 			break;
126 		case 'R':
127 			Rflag = 1;
128 			break;
129 		case 'f':
130 			fflag = 1;
131 			iflag = 0;
132 			break;
133 		case 'i':
134 			iflag = isatty(fileno(stdin));
135 			fflag = 0;
136 			break;
137 		case 'p':
138 			pflag = 1;
139 			break;
140 		case 'r':
141 			rflag = 1;
142 			break;
143 		case 'v':
144 			vflag = 1;
145 			break;
146 		case '?':
147 		default:
148 			usage();
149 			/* NOTREACHED */
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(EXIT_FAILURE,
161 		    "the -R and -r options may not be specified together.");
162 			/* NOTREACHED */
163 		}
164 		if (Hflag || Lflag || Pflag) {
165 			errx(EXIT_FAILURE,
166 	"the -H, -L, and -P options may not be specified with the -r option.");
167 			/* NOTREACHED */
168 		}
169 		fts_options &= ~FTS_PHYSICAL;
170 		fts_options |= FTS_LOGICAL;
171 	}
172 
173 	if (Rflag) {
174 		if (Hflag)
175 			fts_options |= FTS_COMFOLLOW;
176 		if (Lflag) {
177 			fts_options &= ~FTS_PHYSICAL;
178 			fts_options |= FTS_LOGICAL;
179 		}
180 	} else if (!Pflag) {
181 		fts_options &= ~FTS_PHYSICAL;
182 		fts_options |= FTS_LOGICAL | FTS_COMFOLLOW;
183 	}
184 
185 	myuid = getuid();
186 
187 	/* Copy the umask for explicit mode setting. */
188 	myumask = umask(0);
189 	(void)umask(myumask);
190 
191 	/* Save the target base in "to". */
192 	target = argv[--argc];
193 	if (strlcpy(to.p_path, target, sizeof(to.p_path)) >= sizeof(to.p_path))
194 		errx(EXIT_FAILURE, "%s: name too long", target);
195 	to.p_end = to.p_path + strlen(to.p_path);
196 	have_trailing_slash = (to.p_end[-1] == '/');
197 	if (have_trailing_slash)
198 		STRIP_TRAILING_SLASH(to);
199 	to.target_end = to.p_end;
200 
201 	/* Set end of argument list for fts(3). */
202 	argv[argc] = NULL;
203 
204 	/*
205 	 * Cp has two distinct cases:
206 	 *
207 	 * cp [-R] source target
208 	 * cp [-R] source1 ... sourceN directory
209 	 *
210 	 * In both cases, source can be either a file or a directory.
211 	 *
212 	 * In (1), the target becomes a copy of the source. That is, if the
213 	 * source is a file, the target will be a file, and likewise for
214 	 * directories.
215 	 *
216 	 * In (2), the real target is not directory, but "directory/source".
217 	 */
218 	if (Pflag)
219 		r = lstat(to.p_path, &to_stat);
220 	else
221 		r = stat(to.p_path, &to_stat);
222 	if (r == -1 && errno != ENOENT) {
223 		err(EXIT_FAILURE, "%s", to.p_path);
224 		/* NOTREACHED */
225 	}
226 	if (r == -1 || !S_ISDIR(to_stat.st_mode)) {
227 		/*
228 		 * Case (1).  Target is not a directory.
229 		 */
230 		if (argc > 1)
231 			usage();
232 		/*
233 		 * Need to detect the case:
234 		 *	cp -R dir foo
235 		 * Where dir is a directory and foo does not exist, where
236 		 * we want pathname concatenations turned on but not for
237 		 * the initial mkdir().
238 		 */
239 		if (r == -1) {
240 			if (rflag || (Rflag && (Lflag || Hflag)))
241 				r = stat(*argv, &tmp_stat);
242 			else
243 				r = lstat(*argv, &tmp_stat);
244 			if (r == -1) {
245 				err(EXIT_FAILURE, "%s", *argv);
246 				/* NOTREACHED */
247 			}
248 
249 			if (S_ISDIR(tmp_stat.st_mode) && (Rflag || rflag))
250 				type = DIR_TO_DNE;
251 			else
252 				type = FILE_TO_FILE;
253 		} else
254 			type = FILE_TO_FILE;
255 
256 		if (have_trailing_slash && type == FILE_TO_FILE) {
257 			if (r == -1)
258 				errx(1, "directory %s does not exist",
259 				     to.p_path);
260 			else
261 				errx(1, "%s is not a directory", to.p_path);
262 		}
263 	} else {
264 		/*
265 		 * Case (2).  Target is a directory.
266 		 */
267 		type = FILE_TO_DIR;
268 	}
269 
270 	/*
271 	 * make "cp -rp src/ dst" behave like "cp -rp src dst" not
272 	 * like "cp -rp src/. dst"
273 	 */
274 	for (src = argv; *src; src++) {
275 		size_t len = strlen(*src);
276 		while (len-- > 1 && (*src)[len] == '/')
277 			(*src)[len] = '\0';
278 	}
279 
280 	exit(copy(argv, type, fts_options));
281 	/* NOTREACHED */
282 }
283 
284 static int dnestack[MAXPATHLEN]; /* unlikely we'll have more nested dirs */
285 static ssize_t dnesp;
286 static void
287 pushdne(int dne)
288 {
289 
290 	dnestack[dnesp++] = dne;
291 	assert(dnesp < MAXPATHLEN);
292 }
293 
294 static int
295 popdne(void)
296 {
297 	int rv;
298 
299 	rv = dnestack[--dnesp];
300 	assert(dnesp >= 0);
301 	return rv;
302 }
303 
304 int
305 copy(char *argv[], enum op type, int fts_options)
306 {
307 	struct stat to_stat;
308 	FTS *ftsp;
309 	FTSENT *curr;
310 	int base, dne, sval;
311 	int this_failed, any_failed;
312 	size_t nlen;
313 	char *p, *target_mid;
314 
315 	base = 0;	/* XXX gcc -Wuninitialized (see comment below) */
316 
317 	if ((ftsp = fts_open(argv, fts_options, NULL)) == NULL)
318 		err(EXIT_FAILURE, "%s", argv[0]);
319 		/* NOTREACHED */
320 	for (any_failed = 0; (curr = fts_read(ftsp)) != NULL;) {
321 		this_failed = 0;
322 		switch (curr->fts_info) {
323 		case FTS_NS:
324 		case FTS_DNR:
325 		case FTS_ERR:
326 			warnx("%s: %s", curr->fts_path,
327 					strerror(curr->fts_errno));
328 			this_failed = any_failed = 1;
329 			continue;
330 		case FTS_DC:			/* Warn, continue. */
331 			warnx("%s: directory causes a cycle", curr->fts_path);
332 			this_failed = any_failed = 1;
333 			continue;
334 		}
335 
336 		/*
337 		 * If we are in case (2) or (3) above, we need to append the
338                  * source name to the target name.
339                  */
340 		if (type != FILE_TO_FILE) {
341 			if ((curr->fts_namelen +
342 			    to.target_end - to.p_path + 1) > MAXPATHLEN) {
343 				warnx("%s/%s: name too long (not copied)",
344 						to.p_path, curr->fts_name);
345 				this_failed = any_failed = 1;
346 				continue;
347 			}
348 
349 			/*
350 			 * Need to remember the roots of traversals to create
351 			 * correct pathnames.  If there's a directory being
352 			 * copied to a non-existent directory, e.g.
353 			 *	cp -R a/dir noexist
354 			 * the resulting path name should be noexist/foo, not
355 			 * noexist/dir/foo (where foo is a file in dir), which
356 			 * is the case where the target exists.
357 			 *
358 			 * Also, check for "..".  This is for correct path
359 			 * concatentation for paths ending in "..", e.g.
360 			 *	cp -R .. /tmp
361 			 * Paths ending in ".." are changed to ".".  This is
362 			 * tricky, but seems the easiest way to fix the problem.
363 			 *
364 			 * XXX
365 			 * Since the first level MUST be FTS_ROOTLEVEL, base
366 			 * is always initialized.
367 			 */
368 			if (curr->fts_level == FTS_ROOTLEVEL) {
369 				if (type != DIR_TO_DNE) {
370 					p = strrchr(curr->fts_path, '/');
371 					base = (p == NULL) ? 0 :
372 					    (int)(p - curr->fts_path + 1);
373 
374 					if (!strcmp(&curr->fts_path[base],
375 					    ".."))
376 						base += 1;
377 				} else
378 					base = curr->fts_pathlen;
379 			}
380 
381 			p = &curr->fts_path[base];
382 			nlen = curr->fts_pathlen - base;
383 			target_mid = to.target_end;
384 			if (*p != '/' && target_mid[-1] != '/')
385 				*target_mid++ = '/';
386 			*target_mid = 0;
387 
388 			if (target_mid - to.p_path + nlen >= PATH_MAX) {
389 				warnx("%s%s: name too long (not copied)",
390 				    to.p_path, p);
391 				this_failed = any_failed = 1;
392 				continue;
393 			}
394 			(void)strncat(target_mid, p, nlen);
395 			to.p_end = target_mid + nlen;
396 			*to.p_end = 0;
397 			STRIP_TRAILING_SLASH(to);
398 		}
399 
400 		sval = Pflag ? lstat(to.p_path, &to_stat) : stat(to.p_path, &to_stat);
401 		/* Not an error but need to remember it happened */
402 		if (sval == -1)
403 			dne = 1;
404 		else {
405 			if (to_stat.st_dev == curr->fts_statp->st_dev &&
406 			    to_stat.st_ino == curr->fts_statp->st_ino) {
407 				warnx("%s and %s are identical (not copied).",
408 				    to.p_path, curr->fts_path);
409 				this_failed = any_failed = 1;
410 				if (S_ISDIR(curr->fts_statp->st_mode))
411 					(void)fts_set(ftsp, curr, FTS_SKIP);
412 				continue;
413 			}
414 			if (!S_ISDIR(curr->fts_statp->st_mode) &&
415 			    S_ISDIR(to_stat.st_mode)) {
416 		warnx("cannot overwrite directory %s with non-directory %s",
417 				    to.p_path, curr->fts_path);
418 				this_failed = any_failed = 1;
419 				continue;
420 			}
421 			dne = 0;
422 		}
423 
424 		switch (curr->fts_statp->st_mode & S_IFMT) {
425 		case S_IFLNK:
426 			/* Catch special case of a non dangling symlink */
427 			if((fts_options & FTS_LOGICAL) ||
428 			   ((fts_options & FTS_COMFOLLOW) && curr->fts_level == 0)) {
429 				if (copy_file(curr, dne))
430 					this_failed = any_failed = 1;
431 			} else {
432 				if (copy_link(curr, !dne))
433 					this_failed = any_failed = 1;
434 			}
435 			break;
436 		case S_IFDIR:
437 			if (!Rflag && !rflag) {
438 				if (curr->fts_info == FTS_D)
439 					warnx("%s is a directory (not copied).",
440 					    curr->fts_path);
441 				(void)fts_set(ftsp, curr, FTS_SKIP);
442 				this_failed = any_failed = 1;
443 				break;
444 			}
445 
446                         /*
447                          * Directories get noticed twice:
448                          *  In the first pass, create it if needed.
449                          *  In the second pass, after the children have been copied, set the permissions.
450                          */
451 			if (curr->fts_info == FTS_D) /* First pass */
452 			{
453 				/*
454 				 * If the directory doesn't exist, create the new
455 				 * one with the from file mode plus owner RWX bits,
456 				 * modified by the umask.  Trade-off between being
457 				 * able to write the directory (if from directory is
458 				 * 555) and not causing a permissions race.  If the
459 				 * umask blocks owner writes, we fail..
460 				 */
461 				pushdne(dne);
462 				if (dne) {
463 					if (mkdir(to.p_path,
464 					    curr->fts_statp->st_mode | S_IRWXU) < 0)
465 						err(EXIT_FAILURE, "%s",
466 						    to.p_path);
467 						/* NOTREACHED */
468 				} else if (!S_ISDIR(to_stat.st_mode)) {
469 					errno = ENOTDIR;
470 					err(EXIT_FAILURE, "%s",
471 						to.p_path);
472 					/* NOTREACHED */
473 				}
474 			}
475 			else if (curr->fts_info == FTS_DP) /* Second pass */
476 			{
477 	                        /*
478 				 * If not -p and directory didn't exist, set it to be
479 				 * the same as the from directory, umodified by the
480                         	 * umask; arguably wrong, but it's been that way
481                         	 * forever.
482 				 */
483 				if (pflag && setfile(curr->fts_statp, 0))
484 					this_failed = any_failed = 1;
485 				else if ((dne = popdne()))
486 					(void)chmod(to.p_path,
487 					    curr->fts_statp->st_mode);
488 			}
489 			else
490 			{
491 				warnx("directory %s encountered when not expected.",
492 				    curr->fts_path);
493 				this_failed = any_failed = 1;
494 				break;
495 			}
496 
497 			break;
498 		case S_IFBLK:
499 		case S_IFCHR:
500 			if (Rflag) {
501 				if (copy_special(curr->fts_statp, !dne))
502 					this_failed = any_failed = 1;
503 			} else
504 				if (copy_file(curr, dne))
505 					this_failed = any_failed = 1;
506 			break;
507 		case S_IFIFO:
508 			if (Rflag) {
509 				if (copy_fifo(curr->fts_statp, !dne))
510 					this_failed = any_failed = 1;
511 			} else
512 				if (copy_file(curr, dne))
513 					this_failed = any_failed = 1;
514 			break;
515 		default:
516 			if (copy_file(curr, dne))
517 				this_failed = any_failed = 1;
518 			break;
519 		}
520 		if (vflag && !this_failed)
521 			(void)printf("%s -> %s\n", curr->fts_path, to.p_path);
522 	}
523 	if (errno) {
524 		err(EXIT_FAILURE, "fts_read");
525 		/* NOTREACHED */
526 	}
527 	(void)fts_close(ftsp);
528 	return (any_failed);
529 }
530