xref: /dragonfly/bin/cp/cp.c (revision 0ca59c34)
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. 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 /*
34  * Cp copies source files to target files.
35  *
36  * The global PATH_T structure "to" always contains the path to the
37  * current target file.  Since fts(3) does not change directories,
38  * this path can be either absolute or dot-relative.
39  *
40  * The basic algorithm is to initialize "to" and use fts(3) to traverse
41  * the file hierarchy rooted in the argument list.  A trivial case is the
42  * case of 'cp file1 file2'.  The more interesting case is the case of
43  * 'cp file1 file2 ... fileN dir' where the hierarchy is traversed and the
44  * path (relative to the root of the traversal) is appended to dir (stored
45  * in "to") to form the final target path.
46  */
47 
48 #include <sys/types.h>
49 #include <sys/stat.h>
50 
51 #include <err.h>
52 #include <errno.h>
53 #include <fts.h>
54 #include <limits.h>
55 #include <signal.h>
56 #include <stdio.h>
57 #include <stdlib.h>
58 #include <string.h>
59 #include <unistd.h>
60 
61 #include "extern.h"
62 
63 #define	STRIP_TRAILING_SLASH(p) {					\
64         while ((p).p_end > (p).p_path + 1 && (p).p_end[-1] == '/')	\
65                 *--(p).p_end = 0;					\
66 }
67 
68 static char emptystring[] = "";
69 
70 PATH_T to = { to.p_path, emptystring, "" };
71 
72 int fflag, iflag, lflag, nflag, pflag, vflag;
73 static int Rflag, rflag;
74 volatile sig_atomic_t info;
75 
76 enum op { FILE_TO_FILE, FILE_TO_DIR, DIR_TO_DNE };
77 
78 static int copy(char *[], enum op, int);
79 static int mastercmp (const FTSENT * const *, const FTSENT * const *);
80 static void siginfo (int);
81 
82 int
83 main(int argc, char *argv[])
84 {
85 	struct stat to_stat, tmp_stat;
86 	enum op type;
87 	int Hflag, Lflag, Pflag, ch, fts_options, r, have_trailing_slash;
88 	char *target;
89 
90 	fts_options = FTS_NOCHDIR | FTS_PHYSICAL;
91 	Hflag = Lflag = Pflag = 0;
92 	while ((ch = getopt(argc, argv, "HLPRafilnprvx")) != -1)
93 		switch (ch) {
94 		case 'H':
95 			Hflag = 1;
96 			Lflag = Pflag = 0;
97 			break;
98 		case 'L':
99 			Lflag = 1;
100 			Hflag = Pflag = 0;
101 			break;
102 		case 'P':
103 			Pflag = 1;
104 			Hflag = Lflag = 0;
105 			break;
106 		case 'R':
107 			Rflag = 1;
108 			break;
109 		case 'a':
110 			Pflag = 1;
111 			pflag = 1;
112 			Rflag = 1;
113 			Hflag = Lflag = 0;
114 			break;
115 		case 'f':
116 			fflag = 1;
117 			iflag = nflag = 0;
118 			break;
119 		case 'i':
120 			iflag = 1;
121 			fflag = nflag = 0;
122 			break;
123 		case 'l':
124 			lflag = 1;
125 			break;
126 		case 'n':
127 			nflag = 1;
128 			fflag = iflag = 0;
129 			break;
130 		case 'p':
131 			pflag = 1;
132 			break;
133 		case 'r':
134 			rflag = Lflag = 1;
135 			Hflag = Pflag = 0;
136 			break;
137 		case 'v':
138 			vflag = 1;
139 			break;
140 		case 'x':
141 			fts_options |= FTS_XDEV;
142 			break;
143 		default:
144 			usage();
145 			break;
146 		}
147 	argc -= optind;
148 	argv += optind;
149 
150 	if (argc < 2)
151 		usage();
152 
153 	if (Rflag && rflag)
154 		errx(1, "the -R and -r options may not be specified together");
155 	if (rflag)
156 		Rflag = 1;
157 	if (Rflag) {
158 		if (Hflag)
159 			fts_options |= FTS_COMFOLLOW;
160 		if (Lflag) {
161 			fts_options &= ~FTS_PHYSICAL;
162 			fts_options |= FTS_LOGICAL;
163 		}
164 	} else {
165 		fts_options &= ~FTS_PHYSICAL;
166 		fts_options |= FTS_LOGICAL | FTS_COMFOLLOW;
167 	}
168 	signal(SIGINFO, siginfo);
169 
170 	/* Save the target base in "to". */
171 	target = argv[--argc];
172 	if (strlcpy(to.p_path, target, sizeof(to.p_path)) >= sizeof(to.p_path))
173 		errx(1, "%s: name too long", target);
174 	to.p_end = to.p_path + strlen(to.p_path);
175 	if (to.p_path == to.p_end) {
176 		*to.p_end++ = '.';
177 		*to.p_end = 0;
178 	}
179 	have_trailing_slash = (to.p_end[-1] == '/');
180 	if (have_trailing_slash)
181 		STRIP_TRAILING_SLASH(to);
182 	to.target_end = to.p_end;
183 
184 	/* Set end of argument list for fts(3). */
185 	argv[argc] = NULL;
186 
187 	/*
188 	 * Cp has two distinct cases:
189 	 *
190 	 * cp [-R] source target
191 	 * cp [-R] source1 ... sourceN directory
192 	 *
193 	 * In both cases, source can be either a file or a directory.
194 	 *
195 	 * In (1), the target becomes a copy of the source. That is, if the
196 	 * source is a file, the target will be a file, and likewise for
197 	 * directories.
198 	 *
199 	 * In (2), the real target is not directory, but "directory/source".
200 	 */
201 	r = stat(to.p_path, &to_stat);
202 	if (r == -1 && errno != ENOENT)
203 		err(1, "%s", to.p_path);
204 	if (r == -1 || !S_ISDIR(to_stat.st_mode)) {
205 		/*
206 		 * Case (1).  Target is not a directory.
207 		 */
208 		if (argc > 1)
209 			errx(1, "%s is not a directory", to.p_path);
210 
211 		/*
212 		 * Need to detect the case:
213 		 *	cp -R dir foo
214 		 * Where dir is a directory and foo does not exist, where
215 		 * we want pathname concatenations turned on but not for
216 		 * the initial mkdir().
217 		 */
218 		if (r == -1) {
219 			if (Rflag && (Lflag || Hflag))
220 				stat(*argv, &tmp_stat);
221 			else
222 				lstat(*argv, &tmp_stat);
223 
224 			if (S_ISDIR(tmp_stat.st_mode) && Rflag)
225 				type = DIR_TO_DNE;
226 			else
227 				type = FILE_TO_FILE;
228 		} else
229 			type = FILE_TO_FILE;
230 
231 		if (have_trailing_slash && type == FILE_TO_FILE) {
232 			if (r == -1)
233 				errx(1, "directory %s does not exist",
234 				     to.p_path);
235 			else
236 				errx(1, "%s is not a directory", to.p_path);
237 		}
238 	} else
239 		/*
240 		 * Case (2).  Target is a directory.
241 		 */
242 		type = FILE_TO_DIR;
243 
244 	exit (copy(argv, type, fts_options));
245 }
246 
247 static int
248 copy(char *argv[], enum op type, int fts_options)
249 {
250 	struct stat to_stat;
251 	FTS *ftsp;
252 	FTSENT *curr;
253 	int base = 0, dne, badcp, rval;
254 	size_t nlen;
255 	char *p, *target_mid;
256 	mode_t mask, mode;
257 
258 	/*
259 	 * Keep an inverted copy of the umask, for use in correcting
260 	 * permissions on created directories when not using -p.
261 	 */
262 	mask = ~umask(0777);
263 	umask(~mask);
264 
265 	if ((ftsp = fts_open(argv, fts_options, mastercmp)) == NULL)
266 		err(1, "fts_open");
267 	for (badcp = rval = 0; (curr = fts_read(ftsp)) != NULL; badcp = 0) {
268 		switch (curr->fts_info) {
269 		case FTS_NS:
270 		case FTS_DNR:
271 		case FTS_ERR:
272 			warnx("%s: %s",
273 			    curr->fts_path, strerror(curr->fts_errno));
274 			rval = 1;
275 			continue;
276 		case FTS_DC:			/* Warn, continue. */
277 			warnx("%s: directory causes a cycle", curr->fts_path);
278 			rval = 1;
279 			continue;
280 		default:
281 			;
282 		}
283 
284 		/*
285 		 * If we are in case (2) or (3) above, we need to append the
286 		 * source name to the target name.
287 		 */
288 		if (type != FILE_TO_FILE) {
289 			/*
290 			 * Need to remember the roots of traversals to create
291 			 * correct pathnames.  If there's a directory being
292 			 * copied to a non-existent directory, e.g.
293 			 *	cp -R a/dir noexist
294 			 * the resulting path name should be noexist/foo, not
295 			 * noexist/dir/foo (where foo is a file in dir), which
296 			 * is the case where the target exists.
297 			 *
298 			 * Also, check for "..".  This is for correct path
299 			 * concatenation for paths ending in "..", e.g.
300 			 *	cp -R .. /tmp
301 			 * Paths ending in ".." are changed to ".".  This is
302 			 * tricky, but seems the easiest way to fix the problem.
303 			 *
304 			 * XXX
305 			 * Since the first level MUST be FTS_ROOTLEVEL, base
306 			 * is always initialized.
307 			 */
308 			if (curr->fts_level == FTS_ROOTLEVEL) {
309 				if (type != DIR_TO_DNE) {
310 					p = strrchr(curr->fts_path, '/');
311 					base = (p == NULL) ? 0 :
312 					    (int)(p - curr->fts_path + 1);
313 
314 					if (!strcmp(&curr->fts_path[base],
315 					    ".."))
316 						base += 1;
317 				} else
318 					base = curr->fts_pathlen;
319 			}
320 
321 			p = &curr->fts_path[base];
322 			nlen = curr->fts_pathlen - base;
323 			target_mid = to.target_end;
324 			if (*p != '/' && target_mid[-1] != '/')
325 				*target_mid++ = '/';
326 			*target_mid = 0;
327 			if (target_mid - to.p_path + nlen >= PATH_MAX) {
328 				warnx("%s%s: name too long (not copied)",
329 				    to.p_path, p);
330 				rval = 1;
331 				continue;
332 			}
333 			strncat(target_mid, p, nlen);
334 			to.p_end = target_mid + nlen;
335 			*to.p_end = 0;
336 			STRIP_TRAILING_SLASH(to);
337 		}
338 
339 		if (curr->fts_info == FTS_DP) {
340 			/*
341 			 * We are nearly finished with this directory.  If we
342 			 * didn't actually copy it, or otherwise don't need to
343 			 * change its attributes, then we are done.
344 			 */
345 			if (!curr->fts_number)
346 				continue;
347 			/*
348 			 * If -p is in effect, set all the attributes.
349 			 * Otherwise, set the correct permissions, limited
350 			 * by the umask.  Optimise by avoiding a chmod()
351 			 * if possible (which is usually the case if we
352 			 * made the directory).  Note that mkdir() does not
353 			 * honour setuid, setgid and sticky bits, but we
354 			 * normally want to preserve them on directories.
355 			 */
356 			if (pflag) {
357 				if (setfile(curr->fts_statp, -1))
358 				    rval = 1;
359 			} else {
360 				mode = curr->fts_statp->st_mode;
361 				if ((mode & (S_ISUID | S_ISGID | S_ISTXT)) ||
362 				    ((mode | S_IRWXU) & mask) != (mode & mask))
363 					if (chmod(to.p_path, mode & mask) != 0){
364 						warn("chmod: %s", to.p_path);
365 						rval = 1;
366 					}
367 			}
368 			continue;
369 		}
370 
371 		/* Not an error but need to remember it happened */
372 		if (stat(to.p_path, &to_stat) == -1)
373 			dne = 1;
374 		else {
375 			if (to_stat.st_dev == curr->fts_statp->st_dev &&
376 			    to_stat.st_ino == curr->fts_statp->st_ino) {
377 				warnx("%s and %s are identical (not copied).",
378 				    to.p_path, curr->fts_path);
379 				rval = 1;
380 				if (S_ISDIR(curr->fts_statp->st_mode))
381 					fts_set(ftsp, curr, FTS_SKIP);
382 				continue;
383 			}
384 			if (!S_ISDIR(curr->fts_statp->st_mode) &&
385 			    S_ISDIR(to_stat.st_mode)) {
386 				warnx("cannot overwrite directory %s with "
387 				    "non-directory %s",
388 				    to.p_path, curr->fts_path);
389 				rval = 1;
390 				continue;
391 			}
392 			dne = 0;
393 		}
394 
395 		switch (curr->fts_statp->st_mode & S_IFMT) {
396 		case S_IFLNK:
397 			/* Catch special case of a non-dangling symlink */
398 			if ((fts_options & FTS_LOGICAL) ||
399 			    ((fts_options & FTS_COMFOLLOW) &&
400 			    curr->fts_level == 0)) {
401 				if (copy_file(curr, dne))
402 					badcp = rval = 1;
403 			} else {
404 				if (copy_link(curr, !dne))
405 					badcp = rval = 1;
406 			}
407 			break;
408 		case S_IFDIR:
409 			if (!Rflag) {
410 				warnx("%s is a directory (not copied).",
411 				    curr->fts_path);
412 				fts_set(ftsp, curr, FTS_SKIP);
413 				badcp = rval = 1;
414 				break;
415 			}
416 			/*
417 			 * If the directory doesn't exist, create the new
418 			 * one with the from file mode plus owner RWX bits,
419 			 * modified by the umask.  Trade-off between being
420 			 * able to write the directory (if from directory is
421 			 * 555) and not causing a permissions race.  If the
422 			 * umask blocks owner writes, we fail..
423 			 */
424 			if (dne) {
425 				if (mkdir(to.p_path,
426 				    curr->fts_statp->st_mode | S_IRWXU) < 0)
427 					err(1, "%s", to.p_path);
428 			} else if (!S_ISDIR(to_stat.st_mode)) {
429 				errno = ENOTDIR;
430 				err(1, "%s", to.p_path);
431 			}
432 			/*
433 			 * Arrange to correct directory attributes later
434 			 * (in the post-order phase) if this is a new
435 			 * directory, or if the -p flag is in effect.
436 			 */
437 			curr->fts_number = pflag || dne;
438 			break;
439 		case S_IFBLK:
440 		case S_IFCHR:
441 			if (Rflag) {
442 				if (copy_special(curr->fts_statp, !dne))
443 					badcp = rval = 1;
444 			} else {
445 				if (copy_file(curr, dne))
446 					badcp = rval = 1;
447 			}
448 			break;
449 		case S_IFSOCK:
450 			warnx("%s is a socket (not copied).",
451 				    curr->fts_path);
452 			break;
453 		case S_IFIFO:
454 			if (Rflag) {
455 				if (copy_fifo(curr->fts_statp, !dne))
456 					badcp = rval = 1;
457 			} else {
458 				if (copy_file(curr, dne))
459 					badcp = rval = 1;
460 			}
461 			break;
462 		default:
463 			if (copy_file(curr, dne))
464 				badcp = rval = 1;
465 			break;
466 		}
467 		if (vflag && !badcp)
468 			printf("%s -> %s\n", curr->fts_path, to.p_path);
469 	}
470 	if (errno)
471 		err(1, "fts_read");
472 	fts_close(ftsp);
473 	return (rval);
474 }
475 
476 /*
477  * mastercmp --
478  *	The comparison function for the copy order.  The order is to copy
479  *	non-directory files before directory files.  The reason for this
480  *	is because files tend to be in the same cylinder group as their
481  *	parent directory, whereas directories tend not to be.  Copying the
482  *	files first reduces seeking.
483  */
484 static int
485 mastercmp(const FTSENT * const *a, const FTSENT * const *b)
486 {
487 	int a_info, b_info;
488 
489 	a_info = (*a)->fts_info;
490 	if (a_info == FTS_ERR || a_info == FTS_NS || a_info == FTS_DNR)
491 		return (0);
492 	b_info = (*b)->fts_info;
493 	if (b_info == FTS_ERR || b_info == FTS_NS || b_info == FTS_DNR)
494 		return (0);
495 	if (a_info == FTS_D)
496 		return (-1);
497 	if (b_info == FTS_D)
498 		return (1);
499 	return (0);
500 }
501 
502 static void
503 siginfo(int sig __unused)
504 {
505 
506 	info = 1;
507 }
508