1 /*-
2  * Copyright (c) 2003-2007 Tim Kientzle
3  * 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  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
15  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
16  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
17  * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
18  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
19  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
20  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
21  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
22  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
23  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
24  */
25 
26 #include "archive_platform.h"
27 __FBSDID("$FreeBSD: src/lib/libarchive/archive_write_disk_set_standard_lookup.c,v 1.4 2007/05/29 01:00:19 kientzle Exp $");
28 
29 #ifdef HAVE_SYS_TYPES_H
30 #include <sys/types.h>
31 #endif
32 #ifdef HAVE_ERRNO_H
33 #include <errno.h>
34 #endif
35 #ifdef HAVE_GRP_H
36 #include <grp.h>
37 #endif
38 #ifdef HAVE_PWD_H
39 #include <pwd.h>
40 #endif
41 #ifdef HAVE_STDLIB_H
42 #include <stdlib.h>
43 #endif
44 #ifdef HAVE_STRING_H
45 #include <string.h>
46 #endif
47 
48 #include "archive.h"
49 #include "archive_private.h"
50 #include "archive_read_private.h"
51 #include "archive_write_disk_private.h"
52 
53 struct bucket {
54 	char	*name;
55 	int	 hash;
56 	id_t	 id;
57 };
58 
59 static const size_t cache_size = 127;
60 static unsigned int	hash(const char *);
61 static gid_t	lookup_gid(void *, const char *uname, gid_t);
62 static uid_t	lookup_uid(void *, const char *uname, uid_t);
63 static void	cleanup(void *);
64 
65 /*
66  * Installs functions that use getpwnam()/getgrnam()---along with
67  * a simple cache to accelerate such lookups---into the archive_write_disk
68  * object.  This is in a separate file because getpwnam()/getgrnam()
69  * can pull in a LOT of library code (including NIS/LDAP functions, which
70  * pull in DNS resolveers, etc).  This can easily top 500kB, which makes
71  * it inappropriate for some space-constrained applications.
72  *
73  * Applications that are size-sensitive may want to just use the
74  * real default functions (defined in archive_write_disk.c) that just
75  * use the uid/gid without the lookup.  Or define your own custom functions
76  * if you prefer.
77  *
78  * TODO: Replace these hash tables with simpler move-to-front LRU
79  * lists with a bounded size (128 items?).  The hash is a bit faster,
80  * but has a bad pathology in which it thrashes a single bucket.  Even
81  * walking a list of 128 items is a lot faster than calling
82  * getpwnam()!
83  */
84 int
85 archive_write_disk_set_standard_lookup(struct archive *a)
86 {
87 	struct bucket *ucache = malloc(cache_size * sizeof(struct bucket));
88 	struct bucket *gcache = malloc(cache_size * sizeof(struct bucket));
89 	memset(ucache, 0, cache_size * sizeof(struct bucket));
90 	memset(gcache, 0, cache_size * sizeof(struct bucket));
91 	archive_write_disk_set_group_lookup(a, gcache, lookup_gid, cleanup);
92 	archive_write_disk_set_user_lookup(a, ucache, lookup_uid, cleanup);
93 	return (ARCHIVE_OK);
94 }
95 
96 static gid_t
97 lookup_gid(void *private_data, const char *gname, gid_t gid)
98 {
99 	int h;
100 	struct bucket *b;
101 	struct bucket *gcache = (struct bucket *)private_data;
102 
103 	/* If no gname, just use the gid provided. */
104 	if (gname == NULL || *gname == '\0')
105 		return (gid);
106 
107 	/* Try to find gname in the cache. */
108 	h = hash(gname);
109 	b = &gcache[h % cache_size ];
110 	if (b->name != NULL && b->hash == h && strcmp(gname, b->name) == 0)
111 		return ((gid_t)b->id);
112 
113 	/* Free the cache slot for a new entry. */
114 	if (b->name != NULL)
115 		free(b->name);
116 	b->name = strdup(gname);
117 	/* Note: If strdup fails, that's okay; we just won't cache. */
118 	b->hash = h;
119 #if HAVE_GRP_H
120 	{
121 		char _buffer[128];
122 		size_t bufsize = 128;
123 		char *buffer = _buffer;
124 		struct group	grent, *result;
125 		int r;
126 
127 		for (;;) {
128 			r = getgrnam_r(gname, &grent, buffer, bufsize, &result);
129 			if (r == 0)
130 				break;
131 			if (r != ERANGE)
132 				break;
133 			bufsize *= 2;
134 			if (buffer != _buffer)
135 				free(buffer);
136 			buffer = malloc(bufsize);
137 			if (buffer == NULL)
138 				break;
139 		}
140 		if (result != NULL)
141 			gid = result->gr_gid;
142 		if (buffer != _buffer)
143 			free(buffer);
144 	}
145 #elif defined(_WIN32) && !defined(__CYGWIN__)
146 	/* TODO: do a gname->gid lookup for Windows. */
147 #else
148 	#error No way to perform gid lookups on this platform
149 #endif
150 	b->id = gid;
151 
152 	return (gid);
153 }
154 
155 static uid_t
156 lookup_uid(void *private_data, const char *uname, uid_t uid)
157 {
158 	int h;
159 	struct bucket *b;
160 	struct bucket *ucache = (struct bucket *)private_data;
161 
162 	/* If no uname, just use the uid provided. */
163 	if (uname == NULL || *uname == '\0')
164 		return (uid);
165 
166 	/* Try to find uname in the cache. */
167 	h = hash(uname);
168 	b = &ucache[h % cache_size ];
169 	if (b->name != NULL && b->hash == h && strcmp(uname, b->name) == 0)
170 		return ((uid_t)b->id);
171 
172 	/* Free the cache slot for a new entry. */
173 	if (b->name != NULL)
174 		free(b->name);
175 	b->name = strdup(uname);
176 	/* Note: If strdup fails, that's okay; we just won't cache. */
177 	b->hash = h;
178 #if HAVE_PWD_H
179 	{
180 		char _buffer[128];
181 		size_t bufsize = 128;
182 		char *buffer = _buffer;
183 		struct passwd	pwent, *result;
184 		int r;
185 
186 		for (;;) {
187 			r = getpwnam_r(uname, &pwent, buffer, bufsize, &result);
188 			if (r == 0)
189 				break;
190 			if (r != ERANGE)
191 				break;
192 			bufsize *= 2;
193 			if (buffer != _buffer)
194 				free(buffer);
195 			buffer = malloc(bufsize);
196 			if (buffer == NULL)
197 				break;
198 		}
199 		if (result != NULL)
200 			uid = result->pw_uid;
201 		if (buffer != _buffer)
202 			free(buffer);
203 	}
204 #elif defined(_WIN32) && !defined(__CYGWIN__)
205 	/* TODO: do a uname->uid lookup for Windows. */
206 #else
207 	#error No way to look up uids on this platform
208 #endif
209 	b->id = uid;
210 
211 	return (uid);
212 }
213 
214 static void
215 cleanup(void *private)
216 {
217 	size_t i;
218 	struct bucket *cache = (struct bucket *)private;
219 
220 	for (i = 0; i < cache_size; i++)
221 		free(cache[i].name);
222 	free(cache);
223 }
224 
225 
226 static unsigned int
227 hash(const char *p)
228 {
229 	/* A 32-bit version of Peter Weinberger's (PJW) hash algorithm,
230 	   as used by ELF for hashing function names. */
231 	unsigned g, h = 0;
232 	while (*p != '\0') {
233 		h = ( h << 4 ) + *p++;
234 		if (( g = h & 0xF0000000 )) {
235 			h ^= g >> 24;
236 			h &= 0x0FFFFFFF;
237 		}
238 	}
239 	return h;
240 }
241