xref: /openbsd-src/lib/libc/gen/opendir.c (revision b2ea75c1b17e1a9a339660e7ed45cd24946b230e)
1 /*
2  * Copyright (c) 1983, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 #if defined(LIBC_SCCS) && !defined(lint)
35 static char rcsid[] = "$OpenBSD: opendir.c,v 1.6 1998/08/15 08:10:14 deraadt Exp $";
36 #endif /* LIBC_SCCS and not lint */
37 
38 #include <sys/param.h>
39 #include <sys/mount.h>
40 #include <sys/stat.h>
41 
42 #include <dirent.h>
43 #include <errno.h>
44 #include <fcntl.h>
45 #include <stdlib.h>
46 #include <string.h>
47 #include <unistd.h>
48 
49 /*
50  * Open a directory.
51  */
52 DIR *
53 opendir(name)
54 	const char *name;
55 {
56 
57 	return (__opendir2(name, DTF_HIDEW|DTF_NODUP));
58 }
59 
60 DIR *
61 __opendir2(name, flags)
62 	const char *name;
63 	int flags;
64 {
65 	DIR *dirp;
66 	int fd;
67 	struct stat sb;
68 	int pagesz;
69 	int incr;
70 	int unionstack;
71 
72 	if ((fd = open(name, O_RDONLY | O_NONBLOCK)) == -1)
73 		return (NULL);
74 	if (fstat(fd, &sb) || !S_ISDIR(sb.st_mode)) {
75 		errno = ENOTDIR;
76 		close(fd);
77 		return (NULL);
78 	}
79 	if (fcntl(fd, F_SETFD, FD_CLOEXEC) == -1 ||
80 	    (dirp = (DIR *)malloc(sizeof(DIR))) == NULL) {
81 		close(fd);
82 		return (NULL);
83 	}
84 
85 	/*
86 	 * If the machine's page size is an exact multiple of DIRBLKSIZ,
87 	 * use a buffer that is cluster boundary aligned.
88 	 * Hopefully this can be a big win someday by allowing page trades
89 	 * to user space to be done by getdirentries()
90 	 */
91 	if (((pagesz = getpagesize()) % DIRBLKSIZ) == 0)
92 		incr = pagesz;
93 	else
94 		incr = DIRBLKSIZ;
95 
96 	/*
97 	 * Determine whether this directory is the top of a union stack.
98 	 */
99 	if (flags & DTF_NODUP) {
100 		struct statfs sfb;
101 
102 		if (fstatfs(fd, &sfb) < 0) {
103 			free(dirp);
104 			close(fd);
105 			return (NULL);
106 		}
107 		unionstack = strncmp(sfb.f_fstypename, MOUNT_UNION, MFSNAMELEN) == 0 ||
108 			     (sfb.f_flags & MNT_UNION);
109 	} else {
110 		unionstack = 0;
111 	}
112 
113 	if (unionstack) {
114 		int len = 0;
115 		int space = 0;
116 		char *buf = 0;
117 		char *ddptr = 0;
118 		char *ddeptr;
119 		int n;
120 		struct dirent **dpv;
121 
122 		/*
123 		 * The strategy here is to read all the directory
124 		 * entries into a buffer, sort the buffer, and
125 		 * remove duplicate entries by setting the inode
126 		 * number to zero.
127 		 */
128 
129 		do {
130 			/*
131 			 * Always make at least DIRBLKSIZ bytes
132 			 * available to getdirentries
133 			 */
134 			if (space < DIRBLKSIZ) {
135 				char *nbuf;
136 
137 				space += incr;
138 				len += incr;
139 				nbuf = realloc(buf, len);
140 				if (nbuf == NULL) {
141 					if (buf)
142 						free(buf);
143 					free(dirp);
144 					close(fd);
145 					return (NULL);
146 				}
147 				buf = nbuf;
148 				ddptr = buf + (len - space);
149 			}
150 
151 			n = getdirentries(fd, ddptr, space, &dirp->dd_seek);
152 			if (n > 0) {
153 				ddptr += n;
154 				space -= n;
155 			}
156 		} while (n > 0);
157 
158 		ddeptr = ddptr;
159 		flags |= __DTF_READALL;
160 
161 		/*
162 		 * Re-open the directory.
163 		 * This has the effect of rewinding back to the
164 		 * top of the union stack and is needed by
165 		 * programs which plan to fchdir to a descriptor
166 		 * which has also been read -- see fts.c.
167 		 */
168 		if (flags & DTF_REWIND) {
169 			(void) close(fd);
170 			if ((fd = open(name, O_RDONLY)) == -1) {
171 				free(buf);
172 				free(dirp);
173 				return (NULL);
174 			}
175 		}
176 
177 		/*
178 		 * There is now a buffer full of (possibly) duplicate
179 		 * names.
180 		 */
181 		dirp->dd_buf = buf;
182 
183 		/*
184 		 * Go round this loop twice...
185 		 *
186 		 * Scan through the buffer, counting entries.
187 		 * On the second pass, save pointers to each one.
188 		 * Then sort the pointers and remove duplicate names.
189 		 */
190 		for (dpv = 0;;) {
191 			for (n = 0, ddptr = buf; ddptr < ddeptr;) {
192 				struct dirent *dp;
193 
194 				dp = (struct dirent *) ddptr;
195 				if ((long)dp & 03)
196 					break;
197 				if ((dp->d_reclen <= 0) ||
198 				    (dp->d_reclen > (ddeptr + 1 - ddptr)))
199 					break;
200 				ddptr += dp->d_reclen;
201 				if (dp->d_fileno) {
202 					if (dpv)
203 						dpv[n] = dp;
204 					n++;
205 				}
206 			}
207 
208 			if (dpv) {
209 				struct dirent *xp;
210 
211 				/*
212 				 * This sort must be stable.
213 				 */
214 				mergesort(dpv, n, sizeof(*dpv), alphasort);
215 
216 				dpv[n] = NULL;
217 				xp = NULL;
218 
219 				/*
220 				 * Scan through the buffer in sort order,
221 				 * zapping the inode number of any
222 				 * duplicate names.
223 				 */
224 				for (n = 0; dpv[n]; n++) {
225 					struct dirent *dp = dpv[n];
226 
227 					if ((xp == NULL) ||
228 					    strcmp(dp->d_name, xp->d_name))
229 						xp = dp;
230 					else
231 						dp->d_fileno = 0;
232 					if (dp->d_type == DT_WHT &&
233 					    (flags & DTF_HIDEW))
234 						dp->d_fileno = 0;
235 				}
236 
237 				free(dpv);
238 				break;
239 			} else {
240 				dpv = malloc((n+1) * sizeof(struct dirent *));
241 				if (dpv == NULL)
242 					break;
243 			}
244 		}
245 
246 		dirp->dd_len = len;
247 		dirp->dd_size = ddptr - dirp->dd_buf;
248 	} else {
249 		dirp->dd_len = incr;
250 		dirp->dd_buf = malloc(dirp->dd_len);
251 		if (dirp->dd_buf == NULL) {
252 			free(dirp);
253 			close (fd);
254 			return (NULL);
255 		}
256 		dirp->dd_seek = 0;
257 		flags &= ~DTF_REWIND;
258 	}
259 
260 	dirp->dd_loc = 0;
261 	dirp->dd_fd = fd;
262 	dirp->dd_flags = flags;
263 
264 	/*
265 	 * Set up seek point for rewinddir.
266 	 */
267 	dirp->dd_rewind = telldir(dirp);
268 
269 	return (dirp);
270 }
271