cp.c revision 1.38 1 /* $NetBSD: cp.c,v 1.38 2005/10/15 18:22:18 christos 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 1.38 2005/10/15 18:22:18 christos 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 = { to.p_path, empty };
86
87 uid_t myuid;
88 int Rflag, 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 Hflag, Lflag, Pflag, ch, fts_options, r;
103 char *target;
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 if (Rflag) {
172 if (Hflag)
173 fts_options |= FTS_COMFOLLOW;
174 if (Lflag) {
175 fts_options &= ~FTS_PHYSICAL;
176 fts_options |= FTS_LOGICAL;
177 }
178 } else {
179 fts_options &= ~FTS_PHYSICAL;
180 fts_options |= FTS_LOGICAL | FTS_COMFOLLOW;
181 }
182
183 myuid = getuid();
184
185 /* Copy the umask for explicit mode setting. */
186 myumask = umask(0);
187 (void)umask(myumask);
188
189 /* Save the target base in "to". */
190 target = argv[--argc];
191 if (strlen(target) > MAXPATHLEN) {
192 errx(EXIT_FAILURE, "%s: name too long", target);
193 /* NOTREACHED */
194 }
195 (void)strcpy(to.p_path, target);
196 to.p_end = to.p_path + strlen(to.p_path);
197 if (to.p_path == to.p_end) {
198 *to.p_end++ = '.';
199 *to.p_end = 0;
200 }
201 STRIP_TRAILING_SLASH(to);
202 to.target_end = to.p_end;
203
204 /* Set end of argument list for fts(3). */
205 argv[argc] = NULL;
206
207 /*
208 * Cp has two distinct cases:
209 *
210 * cp [-R] source target
211 * cp [-R] source1 ... sourceN directory
212 *
213 * In both cases, source can be either a file or a directory.
214 *
215 * In (1), the target becomes a copy of the source. That is, if the
216 * source is a file, the target will be a file, and likewise for
217 * directories.
218 *
219 * In (2), the real target is not directory, but "directory/source".
220 */
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 } else {
256 /*
257 * Case (2). Target is a directory.
258 */
259 type = FILE_TO_DIR;
260 }
261
262 exit(copy(argv, type, fts_options));
263 /* NOTREACHED */
264 }
265
266 int
267 copy(char *argv[], enum op type, int fts_options)
268 {
269 struct stat to_stat;
270 FTS *ftsp;
271 FTSENT *curr;
272 int base, dne, rval;
273 size_t nlen;
274 char *p, *tmp;
275
276 base = 0; /* XXX gcc -Wuninitialized (see comment below) */
277
278 if ((ftsp = fts_open(argv, fts_options, mastercmp)) == NULL)
279 err(EXIT_FAILURE, "%s", argv[0]);
280 /* NOTREACHED */
281 for (rval = 0; (curr = fts_read(ftsp)) != NULL;) {
282 switch (curr->fts_info) {
283 case FTS_NS:
284 case FTS_DNR:
285 case FTS_ERR:
286 warnx("%s: %s", curr->fts_path,
287 strerror(curr->fts_errno));
288 rval = 1;
289 continue;
290 case FTS_DC: /* Warn, continue. */
291 warnx("%s: directory causes a cycle", curr->fts_path);
292 rval = 1;
293 continue;
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 if ((curr->fts_namelen +
302 to.target_end - to.p_path + 1) > MAXPATHLEN) {
303 warnx("%s/%s: name too long (not copied)",
304 to.p_path, curr->fts_name);
305 rval = 1;
306 continue;
307 }
308
309 /*
310 * Need to remember the roots of traversals to create
311 * correct pathnames. If there's a directory being
312 * copied to a non-existent directory, e.g.
313 * cp -R a/dir noexist
314 * the resulting path name should be noexist/foo, not
315 * noexist/dir/foo (where foo is a file in dir), which
316 * is the case where the target exists.
317 *
318 * Also, check for "..". This is for correct path
319 * concatentation for paths ending in "..", e.g.
320 * cp -R .. /tmp
321 * Paths ending in ".." are changed to ".". This is
322 * tricky, but seems the easiest way to fix the problem.
323 *
324 * XXX
325 * Since the first level MUST be FTS_ROOTLEVEL, base
326 * is always initialized.
327 */
328 if (curr->fts_level == FTS_ROOTLEVEL) {
329 if (type != DIR_TO_DNE) {
330 p = strrchr(curr->fts_path, '/');
331 base = (p == NULL) ? 0 :
332 (int)(p - curr->fts_path + 1);
333
334 if (!strcmp(&curr->fts_path[base],
335 ".."))
336 base += 1;
337 } else
338 base = curr->fts_pathlen;
339 }
340
341 p = &curr->fts_path[base];
342 nlen = curr->fts_pathlen - base;
343
344 tmp = to.target_end;
345 if (*p != '/' && *(tmp - 1) != '/')
346 *tmp++ = '/';
347 *tmp = 0;
348
349 (void)strncat(tmp, p, nlen);
350 to.p_end = tmp + nlen;
351 *to.p_end = 0;
352 STRIP_TRAILING_SLASH(to);
353 }
354
355 /* Not an error but need to remember it happened */
356 if (stat(to.p_path, &to_stat) == -1)
357 dne = 1;
358 else {
359 if (to_stat.st_dev == curr->fts_statp->st_dev &&
360 to_stat.st_ino == curr->fts_statp->st_ino) {
361 warnx("%s and %s are identical (not copied).",
362 to.p_path, curr->fts_path);
363 rval = 1;
364 if (S_ISDIR(curr->fts_statp->st_mode))
365 (void)fts_set(ftsp, curr, FTS_SKIP);
366 continue;
367 }
368 if (!S_ISDIR(curr->fts_statp->st_mode) &&
369 S_ISDIR(to_stat.st_mode)) {
370 warnx("cannot overwrite directory %s with non-directory %s",
371 to.p_path, curr->fts_path);
372 rval = 1;
373 continue;
374 }
375 dne = 0;
376 }
377
378 switch (curr->fts_statp->st_mode & S_IFMT) {
379 case S_IFLNK:
380 /* Catch special case of a non dangling symlink */
381 if((fts_options & FTS_LOGICAL) ||
382 ((fts_options & FTS_COMFOLLOW) && curr->fts_level == 0)) {
383 if (copy_file(curr, dne))
384 rval = 1;
385 } else {
386 if (copy_link(curr, !dne))
387 rval = 1;
388 }
389 break;
390 case S_IFDIR:
391 if (!Rflag && !rflag) {
392 if (curr->fts_info == FTS_D)
393 warnx("%s is a directory (not copied).",
394 curr->fts_path);
395 (void)fts_set(ftsp, curr, FTS_SKIP);
396 rval = 1;
397 break;
398 }
399
400 /*
401 * Directories get noticed twice:
402 * In the first pass, create it if needed.
403 * In the second pass, after the children have been copied, set the permissions.
404 */
405 if (curr->fts_info == FTS_D) /* First pass */
406 {
407 /*
408 * If the directory doesn't exist, create the new
409 * one with the from file mode plus owner RWX bits,
410 * modified by the umask. Trade-off between being
411 * able to write the directory (if from directory is
412 * 555) and not causing a permissions race. If the
413 * umask blocks owner writes, we fail..
414 */
415 if (dne) {
416 if (mkdir(to.p_path,
417 curr->fts_statp->st_mode | S_IRWXU) < 0)
418 err(EXIT_FAILURE, "%s",
419 to.p_path);
420 /* NOTREACHED */
421 } else if (!S_ISDIR(to_stat.st_mode)) {
422 errno = ENOTDIR;
423 err(EXIT_FAILURE, "%s",
424 to.p_path);
425 /* NOTREACHED */
426 }
427 }
428 else if (curr->fts_info == FTS_DP) /* Second pass */
429 {
430 /*
431 * If not -p and directory didn't exist, set it to be
432 * the same as the from directory, umodified by the
433 * umask; arguably wrong, but it's been that way
434 * forever.
435 */
436 if (pflag && setfile(curr->fts_statp, 0))
437 rval = 1;
438 else if (dne)
439 (void)chmod(to.p_path,
440 curr->fts_statp->st_mode);
441 }
442 else
443 {
444 warnx("directory %s encountered when not expected.",
445 curr->fts_path);
446 rval = 1;
447 break;
448 }
449
450 break;
451 case S_IFBLK:
452 case S_IFCHR:
453 if (Rflag) {
454 if (copy_special(curr->fts_statp, !dne))
455 rval = 1;
456 } else
457 if (copy_file(curr, dne))
458 rval = 1;
459 break;
460 case S_IFIFO:
461 if (Rflag) {
462 if (copy_fifo(curr->fts_statp, !dne))
463 rval = 1;
464 } else
465 if (copy_file(curr, dne))
466 rval = 1;
467 break;
468 default:
469 if (copy_file(curr, dne))
470 rval = 1;
471 break;
472 }
473 if (vflag && !rval)
474 (void)printf("%s -> %s\n", curr->fts_path, to.p_path);
475 }
476 if (errno) {
477 err(EXIT_FAILURE, "fts_read");
478 /* NOTREACHED */
479 }
480 return (rval);
481 }
482
483 /*
484 * mastercmp --
485 * The comparison function for the copy order. The order is to copy
486 * non-directory files before directory files. The reason for this
487 * is because files tend to be in the same cylinder group as their
488 * parent directory, whereas directories tend not to be. Copying the
489 * files first reduces seeking.
490 */
491 int
492 mastercmp(const FTSENT **a, const FTSENT **b)
493 {
494 int a_info, b_info;
495
496 a_info = (*a)->fts_info;
497 if (a_info == FTS_ERR || a_info == FTS_NS || a_info == FTS_DNR)
498 return (0);
499 b_info = (*b)->fts_info;
500 if (b_info == FTS_ERR || b_info == FTS_NS || b_info == FTS_DNR)
501 return (0);
502 if (a_info == FTS_D)
503 return (-1);
504 if (b_info == FTS_D)
505 return (1);
506 return (0);
507 }
508