1 /*
2  * Copyright (c) 1994, 2010, Oracle and/or its affiliates. All rights reserved.
3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4  *
5  * This code is free software; you can redistribute it and/or modify it
6  * under the terms of the GNU General Public License version 2 only, as
7  * published by the Free Software Foundation.  Oracle designates this
8  * particular file as subject to the "Classpath" exception as provided
9  * by Oracle in the LICENSE file that accompanied this code.
10  *
11  * This code is distributed in the hope that it will be useful, but WITHOUT
12  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
13  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
14  * version 2 for more details (a copy is included in the LICENSE file that
15  * accompanied this code).
16  *
17  * You should have received a copy of the GNU General Public License version
18  * 2 along with this work; if not, write to the Free Software Foundation,
19  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
20  *
21  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
22  * or visit www.oracle.com if you need additional information or have any
23  * questions.
24  */
25 
26 /*
27  * Pathname canonicalization for Unix file systems
28  */
29 
30 #include <stdio.h>
31 #include <stdlib.h>
32 #include <string.h>
33 #include <sys/stat.h>
34 #include <errno.h>
35 #include <limits.h>
36 #if !defined(_ALLBSD_SOURCE)
37 #include <alloca.h>
38 #endif
39 
40 
41 /* Note: The comments in this file use the terminology
42          defined in the java.io.File class */
43 
44 
45 /* Check the given name sequence to see if it can be further collapsed.
46    Return zero if not, otherwise return the number of names in the sequence. */
47 
48 static int
collapsible(char * names)49 collapsible(char *names)
50 {
51     char *p = names;
52     int dots = 0, n = 0;
53 
54     while (*p) {
55         if ((p[0] == '.') && ((p[1] == '\0')
56                               || (p[1] == '/')
57                               || ((p[1] == '.') && ((p[2] == '\0')
58                                                     || (p[2] == '/'))))) {
59             dots = 1;
60         }
61         n++;
62         while (*p) {
63             if (*p == '/') {
64                 p++;
65                 break;
66             }
67             p++;
68         }
69     }
70     return (dots ? n : 0);
71 }
72 
73 
74 /* Split the names in the given name sequence,
75    replacing slashes with nulls and filling in the given index array */
76 
77 static void
splitNames(char * names,char ** ix)78 splitNames(char *names, char **ix)
79 {
80     char *p = names;
81     int i = 0;
82 
83     while (*p) {
84         ix[i++] = p++;
85         while (*p) {
86             if (*p == '/') {
87                 *p++ = '\0';
88                 break;
89             }
90             p++;
91         }
92     }
93 }
94 
95 
96 /* Join the names in the given name sequence, ignoring names whose index
97    entries have been cleared and replacing nulls with slashes as needed */
98 
99 static void
joinNames(char * names,int nc,char ** ix)100 joinNames(char *names, int nc, char **ix)
101 {
102     int i;
103     char *p;
104 
105     for (i = 0, p = names; i < nc; i++) {
106         if (!ix[i]) continue;
107         if (i > 0) {
108             p[-1] = '/';
109         }
110         if (p == ix[i]) {
111             p += strlen(p) + 1;
112         } else {
113             char *q = ix[i];
114             while ((*p++ = *q++));
115         }
116     }
117     *p = '\0';
118 }
119 
120 
121 /* Collapse "." and ".." names in the given path wherever possible.
122    A "." name may always be eliminated; a ".." name may be eliminated if it
123    follows a name that is neither "." nor "..".  This is a syntactic operation
124    that performs no filesystem queries, so it should only be used to cleanup
125    after invoking the realpath() procedure. */
126 
127 static void
collapse(char * path)128 collapse(char *path)
129 {
130     char *names = (path[0] == '/') ? path + 1 : path; /* Preserve first '/' */
131     int nc;
132     char **ix;
133     int i, j;
134     char *p, *q;
135 
136     nc = collapsible(names);
137     if (nc < 2) return;         /* Nothing to do */
138     ix = (char **)alloca(nc * sizeof(char *));
139     splitNames(names, ix);
140 
141     for (i = 0; i < nc; i++) {
142         int dots = 0;
143 
144         /* Find next occurrence of "." or ".." */
145         do {
146             char *p = ix[i];
147             if (p[0] == '.') {
148                 if (p[1] == '\0') {
149                     dots = 1;
150                     break;
151                 }
152                 if ((p[1] == '.') && (p[2] == '\0')) {
153                     dots = 2;
154                     break;
155                 }
156             }
157             i++;
158         } while (i < nc);
159         if (i >= nc) break;
160 
161         /* At this point i is the index of either a "." or a "..", so take the
162            appropriate action and then continue the outer loop */
163         if (dots == 1) {
164             /* Remove this instance of "." */
165             ix[i] = 0;
166         }
167         else {
168             /* If there is a preceding name, remove both that name and this
169                instance of ".."; otherwise, leave the ".." as is */
170             for (j = i - 1; j >= 0; j--) {
171                 if (ix[j]) break;
172             }
173             if (j < 0) continue;
174             ix[j] = 0;
175             ix[i] = 0;
176         }
177         /* i will be incremented at the top of the loop */
178     }
179 
180     joinNames(names, nc, ix);
181 }
182 
183 
184 /* Convert a pathname to canonical form.  The input path is assumed to contain
185    no duplicate slashes.  On Solaris we can use realpath() to do most of the
186    work, though once that's done we still must collapse any remaining "." and
187    ".." names by hand. */
188 
189 int
canonicalize(char * original,char * resolved,int len)190 canonicalize(char *original, char *resolved, int len)
191 {
192     if (len < PATH_MAX) {
193         errno = EINVAL;
194         return -1;
195     }
196 
197     if (strlen(original) > PATH_MAX) {
198         errno = ENAMETOOLONG;
199         return -1;
200     }
201 
202     /* First try realpath() on the entire path */
203     if (realpath(original, resolved)) {
204         /* That worked, so return it */
205         collapse(resolved);
206         return 0;
207     }
208     else {
209         /* Something's bogus in the original path, so remove names from the end
210            until either some subpath works or we run out of names */
211         char *p, *end, *r = NULL;
212         char path[PATH_MAX + 1];
213 
214         strncpy(path, original, sizeof(path));
215         if (path[PATH_MAX] != '\0') {
216             errno = ENAMETOOLONG;
217             return -1;
218         }
219         end = path + strlen(path);
220 
221         for (p = end; p > path;) {
222 
223             /* Skip last element */
224             while ((--p > path) && (*p != '/'));
225             if (p == path) break;
226 
227             /* Try realpath() on this subpath */
228             *p = '\0';
229             r = realpath(path, resolved);
230             *p = (p == end) ? '\0' : '/';
231 
232             if (r != NULL) {
233                 /* The subpath has a canonical path */
234                 break;
235             }
236             else if (errno == ENOENT || errno == ENOTDIR || errno == EACCES || errno == ENOTCONN) {
237                 /* If the lookup of a particular subpath fails because the file
238                    does not exist, because it is of the wrong type, or because
239                    access is denied, then remove its last name and try again.
240                    Other I/O problems cause an error return. */
241 
242                 /* NOTE: ENOTCONN seems like an odd errno to expect, but this is
243                    the behaviour on linux for fuse filesystems when the fuse device
244                    associated with the FS is closed but the filesystem is not
245                    unmounted. */
246                 continue;
247             }
248             else {
249                 return -1;
250             }
251         }
252 
253         if (r != NULL) {
254             /* Append unresolved subpath to resolved subpath */
255             int rn = strlen(r);
256             if (rn + (int)strlen(p) >= len) {
257                 /* Buffer overflow */
258                 errno = ENAMETOOLONG;
259                 return -1;
260             }
261             if ((rn > 0) && (r[rn - 1] == '/') && (*p == '/')) {
262                 /* Avoid duplicate slashes */
263                 p++;
264             }
265             strcpy(r + rn, p);
266             collapse(r);
267             return 0;
268         }
269         else {
270             /* Nothing resolved, so just return the original path */
271             strcpy(resolved, path);
272             collapse(resolved);
273             return 0;
274         }
275     }
276 
277 }
278