xref: /freebsd/lib/libc/gen/opendir.c (revision 480093f4440d54b30b3025afeac24b48f2ba7a2e)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1983, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 __SCCSID("@(#)opendir.c	8.8 (Berkeley) 5/1/95");
34 __FBSDID("$FreeBSD$");
35 
36 #include "namespace.h"
37 #include <sys/param.h>
38 #include <sys/mount.h>
39 #include <sys/stat.h>
40 
41 #include <dirent.h>
42 #include <errno.h>
43 #include <fcntl.h>
44 #include <stdlib.h>
45 #include <string.h>
46 #include <unistd.h>
47 #include "un-namespace.h"
48 
49 #include "gen-private.h"
50 #include "telldir.h"
51 
52 static DIR * __opendir_common(int, int, bool);
53 
54 /*
55  * Open a directory.
56  */
57 DIR *
58 opendir(const char *name)
59 {
60 
61 	return (__opendir2(name, DTF_HIDEW|DTF_NODUP));
62 }
63 
64 /*
65  * Open a directory with existing file descriptor.
66  */
67 DIR *
68 fdopendir(int fd)
69 {
70 
71 	if (_fcntl(fd, F_SETFD, FD_CLOEXEC) == -1)
72 		return (NULL);
73 	return (__opendir_common(fd, DTF_HIDEW|DTF_NODUP, true));
74 }
75 
76 DIR *
77 __opendir2(const char *name, int flags)
78 {
79 	int fd;
80 	DIR *dir;
81 	int saved_errno;
82 
83 	if ((flags & (__DTF_READALL | __DTF_SKIPREAD)) != 0)
84 		return (NULL);
85 	if ((fd = _open(name,
86 	    O_RDONLY | O_NONBLOCK | O_DIRECTORY | O_CLOEXEC)) == -1)
87 		return (NULL);
88 
89 	dir = __opendir_common(fd, flags, false);
90 	if (dir == NULL) {
91 		saved_errno = errno;
92 		_close(fd);
93 		errno = saved_errno;
94 	}
95 	return (dir);
96 }
97 
98 static int
99 opendir_compar(const void *p1, const void *p2)
100 {
101 
102 	return (strcmp((*(const struct dirent * const *)p1)->d_name,
103 	    (*(const struct dirent * const *)p2)->d_name));
104 }
105 
106 /*
107  * For a directory at the top of a unionfs stack, the entire directory's
108  * contents are read and cached locally until the next call to rewinddir().
109  * For the fdopendir() case, the initial seek position must be preserved.
110  * For rewinddir(), the full directory should always be re-read from the
111  * beginning.
112  *
113  * If an error occurs, the existing buffer and state of 'dirp' is left
114  * unchanged.
115  */
116 bool
117 _filldir(DIR *dirp, bool use_current_pos)
118 {
119 	struct dirent **dpv;
120 	char *buf, *ddptr, *ddeptr;
121 	off_t pos;
122 	int fd2, incr, len, n, saved_errno, space;
123 
124 	len = 0;
125 	space = 0;
126 	buf = NULL;
127 	ddptr = NULL;
128 
129 	/*
130 	 * Use the system page size if that is a multiple of DIRBLKSIZ.
131 	 * Hopefully this can be a big win someday by allowing page
132 	 * trades to user space to be done by _getdirentries().
133 	 */
134 	incr = getpagesize();
135 	if ((incr % DIRBLKSIZ) != 0)
136 		incr = DIRBLKSIZ;
137 
138 	/*
139 	 * The strategy here is to read all the directory
140 	 * entries into a buffer, sort the buffer, and
141 	 * remove duplicate entries by setting the inode
142 	 * number to zero.
143 	 *
144 	 * We reopen the directory because _getdirentries()
145 	 * on a MNT_UNION mount modifies the open directory,
146 	 * making it refer to the lower directory after the
147 	 * upper directory's entries are exhausted.
148 	 * This would otherwise break software that uses
149 	 * the directory descriptor for fchdir or *at
150 	 * functions, such as fts.c.
151 	 */
152 	if ((fd2 = _openat(dirp->dd_fd, ".", O_RDONLY | O_CLOEXEC)) == -1)
153 		return (false);
154 
155 	if (use_current_pos) {
156 		pos = lseek(dirp->dd_fd, 0, SEEK_CUR);
157 		if (pos == -1 || lseek(fd2, pos, SEEK_SET) == -1) {
158 			saved_errno = errno;
159 			_close(fd2);
160 			errno = saved_errno;
161 			return (false);
162 		}
163 	}
164 
165 	do {
166 		/*
167 		 * Always make at least DIRBLKSIZ bytes
168 		 * available to _getdirentries
169 		 */
170 		if (space < DIRBLKSIZ) {
171 			space += incr;
172 			len += incr;
173 			buf = reallocf(buf, len);
174 			if (buf == NULL) {
175 				saved_errno = errno;
176 				_close(fd2);
177 				errno = saved_errno;
178 				return (false);
179 			}
180 			ddptr = buf + (len - space);
181 		}
182 
183 		n = _getdirentries(fd2, ddptr, space, &dirp->dd_seek);
184 		if (n > 0) {
185 			ddptr += n;
186 			space -= n;
187 		}
188 		if (n < 0) {
189 			saved_errno = errno;
190 			_close(fd2);
191 			errno = saved_errno;
192 			return (false);
193 		}
194 	} while (n > 0);
195 	_close(fd2);
196 
197 	ddeptr = ddptr;
198 
199 	/*
200 	 * There is now a buffer full of (possibly) duplicate
201 	 * names.
202 	 */
203 	dirp->dd_buf = buf;
204 
205 	/*
206 	 * Go round this loop twice...
207 	 *
208 	 * Scan through the buffer, counting entries.
209 	 * On the second pass, save pointers to each one.
210 	 * Then sort the pointers and remove duplicate names.
211 	 */
212 	for (dpv = NULL;;) {
213 		n = 0;
214 		ddptr = buf;
215 		while (ddptr < ddeptr) {
216 			struct dirent *dp;
217 
218 			dp = (struct dirent *) ddptr;
219 			if ((long)dp & 03L)
220 				break;
221 			if ((dp->d_reclen <= 0) ||
222 			    (dp->d_reclen > (ddeptr + 1 - ddptr)))
223 				break;
224 			ddptr += dp->d_reclen;
225 			if (dp->d_fileno) {
226 				if (dpv)
227 					dpv[n] = dp;
228 				n++;
229 			}
230 		}
231 
232 		if (dpv) {
233 			struct dirent *xp;
234 
235 			/*
236 			 * This sort must be stable.
237 			 */
238 			mergesort(dpv, n, sizeof(*dpv), opendir_compar);
239 
240 			dpv[n] = NULL;
241 			xp = NULL;
242 
243 			/*
244 			 * Scan through the buffer in sort order,
245 			 * zapping the inode number of any
246 			 * duplicate names.
247 			 */
248 			for (n = 0; dpv[n]; n++) {
249 				struct dirent *dp = dpv[n];
250 
251 				if ((xp == NULL) ||
252 				    strcmp(dp->d_name, xp->d_name)) {
253 					xp = dp;
254 				} else {
255 					dp->d_fileno = 0;
256 				}
257 				if (dp->d_type == DT_WHT &&
258 				    (dirp->dd_flags & DTF_HIDEW))
259 					dp->d_fileno = 0;
260 			}
261 
262 			free(dpv);
263 			break;
264 		} else {
265 			dpv = malloc((n+1) * sizeof(struct dirent *));
266 			if (dpv == NULL)
267 				break;
268 		}
269 	}
270 
271 	dirp->dd_len = len;
272 	dirp->dd_size = ddptr - dirp->dd_buf;
273 	return (true);
274 }
275 
276 static bool
277 is_unionstack(int fd)
278 {
279 	struct statfs sfb;
280 	int unionstack;
281 
282 	unionstack = _fcntl(fd, F_ISUNIONSTACK, 0);
283 	if (unionstack != -1)
284 		return (unionstack);
285 
286 	/*
287 	 * Temporary compat for kernels which don't provide F_ISUNIONSTACK.
288 	 */
289 	if (_fstatfs(fd, &sfb) < 0)
290 		return (true);
291 	return (strcmp(sfb.f_fstypename, "unionfs") == 0 ||
292 	    (sfb.f_flags & MNT_UNION));
293 }
294 
295 /*
296  * Common routine for opendir(3), __opendir2(3) and fdopendir(3).
297  */
298 static DIR *
299 __opendir_common(int fd, int flags, bool use_current_pos)
300 {
301 	DIR *dirp;
302 	int incr;
303 	int saved_errno;
304 	bool unionstack;
305 
306 	if ((dirp = malloc(sizeof(DIR) + sizeof(struct _telldir))) == NULL)
307 		return (NULL);
308 
309 	dirp->dd_buf = NULL;
310 	dirp->dd_fd = fd;
311 	dirp->dd_flags = flags;
312 	dirp->dd_loc = 0;
313 	dirp->dd_lock = NULL;
314 	dirp->dd_td = (struct _telldir *)((char *)dirp + sizeof(DIR));
315 	LIST_INIT(&dirp->dd_td->td_locq);
316 	dirp->dd_td->td_loccnt = 0;
317 	dirp->dd_compat_de = NULL;
318 
319 	/*
320 	 * Use the system page size if that is a multiple of DIRBLKSIZ.
321 	 * Hopefully this can be a big win someday by allowing page
322 	 * trades to user space to be done by _getdirentries().
323 	 */
324 	incr = getpagesize();
325 	if ((incr % DIRBLKSIZ) != 0)
326 		incr = DIRBLKSIZ;
327 
328 	/*
329 	 * Determine whether this directory is the top of a union stack.
330 	 */
331 	unionstack = false;
332 	if (flags & DTF_NODUP) {
333 		unionstack = is_unionstack(fd);
334 	}
335 
336 	if (unionstack) {
337 		if (!_filldir(dirp, use_current_pos))
338 			goto fail;
339 		dirp->dd_flags |= __DTF_READALL;
340 	} else {
341 		dirp->dd_len = incr;
342 		dirp->dd_buf = malloc(dirp->dd_len);
343 		if (dirp->dd_buf == NULL)
344 			goto fail;
345 		if (use_current_pos) {
346 			/*
347 			 * Read the first batch of directory entries
348 			 * to prime dd_seek.  This also checks if the
349 			 * fd passed to fdopendir() is a directory.
350 			 */
351 			dirp->dd_size = _getdirentries(dirp->dd_fd,
352 			    dirp->dd_buf, dirp->dd_len, &dirp->dd_seek);
353 			if (dirp->dd_size < 0) {
354 				if (errno == EINVAL)
355 					errno = ENOTDIR;
356 				goto fail;
357 			}
358 			dirp->dd_flags |= __DTF_SKIPREAD;
359 		} else {
360 			dirp->dd_size = 0;
361 			dirp->dd_seek = 0;
362 		}
363 	}
364 
365 	return (dirp);
366 
367 fail:
368 	saved_errno = errno;
369 	free(dirp->dd_buf);
370 	free(dirp);
371 	errno = saved_errno;
372 	return (NULL);
373 }
374