xref: /netbsd/bin/cp/cp.c (revision bf9ec67e)
1 /* $NetBSD: cp.c,v 1.31 2001/09/13 09:53:59 wiz 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. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the University of
21  *	California, Berkeley and its contributors.
22  * 4. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  */
38 
39 #include <sys/cdefs.h>
40 #ifndef lint
41 __COPYRIGHT(
42 "@(#) Copyright (c) 1988, 1993, 1994\n\
43 	The Regents of the University of California.  All rights reserved.\n");
44 #endif /* not lint */
45 
46 #ifndef lint
47 #if 0
48 static char sccsid[] = "@(#)cp.c	8.5 (Berkeley) 4/29/95";
49 #else
50 __RCSID("$NetBSD: cp.c,v 1.31 2001/09/13 09:53:59 wiz Exp $");
51 #endif
52 #endif /* not lint */
53 
54 /*
55  * Cp copies source files to target files.
56  *
57  * The global PATH_T structure "to" always contains the path to the
58  * current target file.  Since fts(3) does not change directories,
59  * this path can be either absolute or dot-relative.
60  *
61  * The basic algorithm is to initialize "to" and use fts(3) to traverse
62  * the file hierarchy rooted in the argument list.  A trivial case is the
63  * case of 'cp file1 file2'.  The more interesting case is the case of
64  * 'cp file1 file2 ... fileN dir' where the hierarchy is traversed and the
65  * path (relative to the root of the traversal) is appended to dir (stored
66  * in "to") to form the final target path.
67  */
68 
69 #include <sys/param.h>
70 #include <sys/stat.h>
71 
72 #include <err.h>
73 #include <errno.h>
74 #include <fts.h>
75 #include <locale.h>
76 #include <stdlib.h>
77 #include <stdio.h>
78 #include <string.h>
79 #include <unistd.h>
80 
81 #include "extern.h"
82 
83 #define	STRIP_TRAILING_SLASH(p) {					\
84         while ((p).p_end > (p).p_path + 1 && (p).p_end[-1] == '/')	\
85                 *--(p).p_end = '\0';					\
86 }
87 
88 PATH_T to = { to.p_path, "" };
89 
90 uid_t myuid;
91 int Rflag, fflag, iflag, pflag, rflag;
92 mode_t myumask;
93 
94 enum op { FILE_TO_FILE, FILE_TO_DIR, DIR_TO_DNE };
95 
96 int main(int, char *[]);
97 int copy(char *[], enum op, int);
98 int mastercmp(const FTSENT **, const FTSENT **);
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;
106 	char *target;
107 
108 	(void)setlocale(LC_ALL, "");
109 
110 	Hflag = Lflag = Pflag = Rflag = 0;
111 	while ((ch = getopt(argc, argv, "HLPRfipr")) != -1)
112 		switch (ch) {
113 		case 'H':
114 			Hflag = 1;
115 			Lflag = Pflag = 0;
116 			break;
117 		case 'L':
118 			Lflag = 1;
119 			Hflag = Pflag = 0;
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 '?':
143 		default:
144 			usage();
145 			break;
146 		}
147 	argc -= optind;
148 	argv += optind;
149 
150 	if (argc < 2)
151 		usage();
152 
153 	fts_options = FTS_NOCHDIR | FTS_PHYSICAL;
154 	if (rflag) {
155 		if (Rflag)
156 			errx(1,
157 		    "the -R and -r options may not be specified together.");
158 		if (Hflag || Lflag || Pflag)
159 			errx(1,
160 	"the -H, -L, and -P options may not be specified with the -r option.");
161 		fts_options &= ~FTS_PHYSICAL;
162 		fts_options |= FTS_LOGICAL;
163 	}
164 	if (Rflag) {
165 		if (Hflag)
166 			fts_options |= FTS_COMFOLLOW;
167 		if (Lflag) {
168 			fts_options &= ~FTS_PHYSICAL;
169 			fts_options |= FTS_LOGICAL;
170 		}
171 	} else {
172 		fts_options &= ~FTS_PHYSICAL;
173 		fts_options |= FTS_LOGICAL | FTS_COMFOLLOW;
174 	}
175 
176 	myuid = getuid();
177 
178 	/* Copy the umask for explicit mode setting. */
179 	myumask = umask(0);
180 	(void)umask(myumask);
181 
182 	/* Save the target base in "to". */
183 	target = argv[--argc];
184 	if (strlen(target) > MAXPATHLEN)
185 		errx(1, "%s: name too long", target);
186 	(void)strcpy(to.p_path, target);
187 	to.p_end = to.p_path + strlen(to.p_path);
188         if (to.p_path == to.p_end) {
189 		*to.p_end++ = '.';
190 		*to.p_end = 0;
191 	}
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 				r = stat(*argv, &tmp_stat);
233 			else
234 				r = lstat(*argv, &tmp_stat);
235 			if (r == -1)
236 				err(1, "%s", *argv);
237 
238 			if (S_ISDIR(tmp_stat.st_mode) && (Rflag || rflag))
239 				type = DIR_TO_DNE;
240 			else
241 				type = FILE_TO_FILE;
242 		} else
243 			type = FILE_TO_FILE;
244 	} else {
245 		/*
246 		 * Case (2).  Target is a directory.
247 		 */
248 		type = FILE_TO_DIR;
249 	}
250 
251 	exit(copy(argv, type, fts_options));
252 	/* NOTREACHED */
253 }
254 
255 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, dne, nlen, rval;
262 	char *p, *tmp;
263 
264 	base = 0;	/* XXX gcc -Wuninitialized (see comment below) */
265 
266 	if ((ftsp = fts_open(argv, fts_options, mastercmp)) == NULL)
267 		err(1, "%s", argv[0]);
268 	for (rval = 0; (curr = fts_read(ftsp)) != NULL;) {
269 		switch (curr->fts_info) {
270 		case FTS_NS:
271 		case FTS_DNR:
272 		case FTS_ERR:
273 			warnx("%s: %s",
274 			    curr->fts_path, strerror(curr->fts_errno));
275 			rval = 1;
276 			continue;
277 		case FTS_DC:			/* Warn, continue. */
278 			warnx("%s: directory causes a cycle", curr->fts_path);
279 			rval = 1;
280 			continue;
281 		}
282 
283 		/*
284 		 * If we are in case (2) or (3) above, we need to append the
285                  * source name to the target name.
286                  */
287 		if (type != FILE_TO_FILE) {
288 			if ((curr->fts_namelen +
289 			    to.target_end - to.p_path + 1) > MAXPATHLEN) {
290 				warnx("%s/%s: name too long (not copied)",
291 				    to.p_path, curr->fts_name);
292 				rval = 1;
293 				continue;
294 			}
295 
296 			/*
297 			 * Need to remember the roots of traversals to create
298 			 * correct pathnames.  If there's a directory being
299 			 * copied to a non-existent directory, e.g.
300 			 *	cp -R a/dir noexist
301 			 * the resulting path name should be noexist/foo, not
302 			 * noexist/dir/foo (where foo is a file in dir), which
303 			 * is the case where the target exists.
304 			 *
305 			 * Also, check for "..".  This is for correct path
306 			 * concatentation for paths ending in "..", e.g.
307 			 *	cp -R .. /tmp
308 			 * Paths ending in ".." are changed to ".".  This is
309 			 * tricky, but seems the easiest way to fix the problem.
310 			 *
311 			 * XXX
312 			 * Since the first level MUST be FTS_ROOTLEVEL, base
313 			 * is always initialized.
314 			 */
315 			if (curr->fts_level == FTS_ROOTLEVEL) {
316 				if (type != DIR_TO_DNE) {
317 					p = strrchr(curr->fts_path, '/');
318 					base = (p == NULL) ? 0 :
319 					    (int)(p - curr->fts_path + 1);
320 
321 					if (!strcmp(&curr->fts_path[base],
322 					    ".."))
323 						base += 1;
324 				} else
325 					base = curr->fts_pathlen;
326 			}
327 
328 			p = &curr->fts_path[base];
329 			nlen = curr->fts_pathlen - base;
330 
331 			tmp = to.target_end;
332 			if (*p != '/' && *(tmp - 1) != '/')
333 				*tmp++ = '/';
334 			*tmp = 0;
335 
336 			(void)strncat(tmp, p, nlen);
337 			to.p_end = tmp + nlen;
338 			*to.p_end = 0;
339 			STRIP_TRAILING_SLASH(to);
340 		}
341 
342 		/* Not an error but need to remember it happened */
343 		if (stat(to.p_path, &to_stat) == -1)
344 			dne = 1;
345 		else {
346 			if (to_stat.st_dev == curr->fts_statp->st_dev &&
347 			    to_stat.st_ino == curr->fts_statp->st_ino) {
348 				warnx("%s and %s are identical (not copied).",
349 				    to.p_path, curr->fts_path);
350 				rval = 1;
351 				if (S_ISDIR(curr->fts_statp->st_mode))
352 					(void)fts_set(ftsp, curr, FTS_SKIP);
353 				continue;
354 			}
355 			if (!S_ISDIR(curr->fts_statp->st_mode) &&
356 			    S_ISDIR(to_stat.st_mode)) {
357 		warnx("cannot overwrite directory %s with non-directory %s",
358 				    to.p_path, curr->fts_path);
359 				rval = 1;
360 				continue;
361 			}
362 			dne = 0;
363 		}
364 
365 		switch (curr->fts_statp->st_mode & S_IFMT) {
366 		case S_IFLNK:
367 			/* Catch special case of a non dangling symlink */
368 			if((fts_options & FTS_LOGICAL) ||
369 			   ((fts_options & FTS_COMFOLLOW) && curr->fts_level == 0)) {
370 				if (copy_file(curr, dne))
371 					rval = 1;
372 			} else {
373 				if (copy_link(curr, !dne))
374 					rval = 1;
375 			}
376 			break;
377 		case S_IFDIR:
378 			if (!Rflag && !rflag) {
379 				if (curr->fts_info == FTS_D)
380 					warnx("%s is a directory (not copied).",
381 					    curr->fts_path);
382 				(void)fts_set(ftsp, curr, FTS_SKIP);
383 				rval = 1;
384 				break;
385 			}
386 
387                         /*
388                          * Directories get noticed twice:
389                          *  In the first pass, create it if needed.
390                          *  In the second pass, after the children have been copied, set the permissions.
391                          */
392 			if (curr->fts_info == FTS_D) /* First pass */
393 			{
394 				/*
395 				 * If the directory doesn't exist, create the new
396 				 * one with the from file mode plus owner RWX bits,
397 				 * modified by the umask.  Trade-off between being
398 				 * able to write the directory (if from directory is
399 				 * 555) and not causing a permissions race.  If the
400 				 * umask blocks owner writes, we fail..
401 				 */
402 				if (dne) {
403 					if (mkdir(to.p_path,
404 					    curr->fts_statp->st_mode | S_IRWXU) < 0)
405 						err(1, "%s", to.p_path);
406 				} else if (!S_ISDIR(to_stat.st_mode)) {
407 					errno = ENOTDIR;
408 					err(1, "%s", to.p_path);
409 				}
410 			}
411 			else if (curr->fts_info == FTS_DP) /* Second pass */
412 			{
413 	                        /*
414 				 * If not -p and directory didn't exist, set it to be
415 				 * the same as the from directory, umodified by the
416                         	 * umask; arguably wrong, but it's been that way
417                         	 * forever.
418 				 */
419 				if (pflag && setfile(curr->fts_statp, 0))
420 					rval = 1;
421 				else if (dne)
422 					(void)chmod(to.p_path,
423 					    curr->fts_statp->st_mode);
424 			}
425 			else
426                         {
427                         	warnx("directory %s encountered when not expected.", curr->fts_path);
428                         	rval = 1;
429                                 break;
430                         }
431 
432 			break;
433 		case S_IFBLK:
434 		case S_IFCHR:
435 			if (Rflag) {
436 				if (copy_special(curr->fts_statp, !dne))
437 					rval = 1;
438 			} else
439 				if (copy_file(curr, dne))
440 					rval = 1;
441 			break;
442 		case S_IFIFO:
443 			if (Rflag) {
444 				if (copy_fifo(curr->fts_statp, !dne))
445 					rval = 1;
446 			} else
447 				if (copy_file(curr, dne))
448 					rval = 1;
449 			break;
450 		default:
451 			if (copy_file(curr, dne))
452 				rval = 1;
453 			break;
454 		}
455 	}
456 	if (errno)
457 		err(1, "fts_read");
458 	return (rval);
459 }
460 
461 /*
462  * mastercmp --
463  *	The comparison function for the copy order.  The order is to copy
464  *	non-directory files before directory files.  The reason for this
465  *	is because files tend to be in the same cylinder group as their
466  *	parent directory, whereas directories tend not to be.  Copying the
467  *	files first reduces seeking.
468  */
469 int
470 mastercmp(const FTSENT **a, const FTSENT **b)
471 {
472 	int a_info, b_info;
473 
474 	a_info = (*a)->fts_info;
475 	if (a_info == FTS_ERR || a_info == FTS_NS || a_info == FTS_DNR)
476 		return (0);
477 	b_info = (*b)->fts_info;
478 	if (b_info == FTS_ERR || b_info == FTS_NS || b_info == FTS_DNR)
479 		return (0);
480 	if (a_info == FTS_D)
481 		return (-1);
482 	if (b_info == FTS_D)
483 		return (1);
484 	return (0);
485 }
486