]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/namei.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * Some corrections by tytso. | |
9 | */ | |
10 | ||
11 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
12 | * lookup logic. | |
13 | */ | |
14 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
15 | */ | |
16 | ||
17 | #include <linux/init.h> | |
630d9c47 | 18 | #include <linux/export.h> |
1da177e4 LT |
19 | #include <linux/slab.h> |
20 | #include <linux/fs.h> | |
21 | #include <linux/namei.h> | |
1da177e4 | 22 | #include <linux/pagemap.h> |
0eeca283 | 23 | #include <linux/fsnotify.h> |
1da177e4 LT |
24 | #include <linux/personality.h> |
25 | #include <linux/security.h> | |
6146f0d5 | 26 | #include <linux/ima.h> |
1da177e4 LT |
27 | #include <linux/syscalls.h> |
28 | #include <linux/mount.h> | |
29 | #include <linux/audit.h> | |
16f7e0fe | 30 | #include <linux/capability.h> |
834f2a4a | 31 | #include <linux/file.h> |
5590ff0d | 32 | #include <linux/fcntl.h> |
08ce5f16 | 33 | #include <linux/device_cgroup.h> |
5ad4e53b | 34 | #include <linux/fs_struct.h> |
e77819e5 | 35 | #include <linux/posix_acl.h> |
1da177e4 LT |
36 | #include <asm/uaccess.h> |
37 | ||
e81e3f4d | 38 | #include "internal.h" |
c7105365 | 39 | #include "mount.h" |
e81e3f4d | 40 | |
1da177e4 LT |
41 | /* [Feb-1997 T. Schoebel-Theuer] |
42 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
43 | * were necessary because of omirr. The reason is that omirr needs | |
44 | * to know the _real_ pathname, not the user-supplied one, in case | |
45 | * of symlinks (and also when transname replacements occur). | |
46 | * | |
47 | * The new code replaces the old recursive symlink resolution with | |
48 | * an iterative one (in case of non-nested symlink chains). It does | |
49 | * this with calls to <fs>_follow_link(). | |
50 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
51 | * replaced with a single function lookup_dentry() that can handle all | |
52 | * the special cases of the former code. | |
53 | * | |
54 | * With the new dcache, the pathname is stored at each inode, at least as | |
55 | * long as the refcount of the inode is positive. As a side effect, the | |
56 | * size of the dcache depends on the inode cache and thus is dynamic. | |
57 | * | |
58 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
59 | * resolution to correspond with current state of the code. | |
60 | * | |
61 | * Note that the symlink resolution is not *completely* iterative. | |
62 | * There is still a significant amount of tail- and mid- recursion in | |
63 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
64 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
65 | * may return different results than <fs>_follow_link(). Many virtual | |
66 | * filesystems (including /proc) exhibit this behavior. | |
67 | */ | |
68 | ||
69 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
70 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
71 | * and the name already exists in form of a symlink, try to create the new | |
72 | * name indicated by the symlink. The old code always complained that the | |
73 | * name already exists, due to not following the symlink even if its target | |
74 | * is nonexistent. The new semantics affects also mknod() and link() when | |
25985edc | 75 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
76 | * |
77 | * I don't know which semantics is the right one, since I have no access | |
78 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
79 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
80 | * "old" one. Personally, I think the new semantics is much more logical. | |
81 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
82 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
83 | * and in the old Linux semantics. | |
84 | */ | |
85 | ||
86 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
87 | * semantics. See the comments in "open_namei" and "do_link" below. | |
88 | * | |
89 | * [10-Sep-98 Alan Modra] Another symlink change. | |
90 | */ | |
91 | ||
92 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
93 | * inside the path - always follow. | |
94 | * in the last component in creation/removal/renaming - never follow. | |
95 | * if LOOKUP_FOLLOW passed - follow. | |
96 | * if the pathname has trailing slashes - follow. | |
97 | * otherwise - don't follow. | |
98 | * (applied in that order). | |
99 | * | |
100 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
101 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
102 | * During the 2.4 we need to fix the userland stuff depending on it - | |
103 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
104 | * XEmacs seems to be relying on it... | |
105 | */ | |
106 | /* | |
107 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 108 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
109 | * any extra contention... |
110 | */ | |
111 | ||
112 | /* In order to reduce some races, while at the same time doing additional | |
113 | * checking and hopefully speeding things up, we copy filenames to the | |
114 | * kernel data space before using them.. | |
115 | * | |
116 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
117 | * PATH_MAX includes the nul terminator --RR. | |
118 | */ | |
1fa1e7f6 | 119 | static char *getname_flags(const char __user *filename, int flags, int *empty) |
1da177e4 | 120 | { |
3f9f0aa6 LT |
121 | char *result = __getname(), *err; |
122 | int len; | |
4043cde8 | 123 | |
3f9f0aa6 | 124 | if (unlikely(!result)) |
4043cde8 EP |
125 | return ERR_PTR(-ENOMEM); |
126 | ||
3f9f0aa6 LT |
127 | len = strncpy_from_user(result, filename, PATH_MAX); |
128 | err = ERR_PTR(len); | |
129 | if (unlikely(len < 0)) | |
130 | goto error; | |
131 | ||
132 | /* The empty path is special. */ | |
133 | if (unlikely(!len)) { | |
134 | if (empty) | |
4043cde8 | 135 | *empty = 1; |
3f9f0aa6 LT |
136 | err = ERR_PTR(-ENOENT); |
137 | if (!(flags & LOOKUP_EMPTY)) | |
138 | goto error; | |
1da177e4 | 139 | } |
3f9f0aa6 LT |
140 | |
141 | err = ERR_PTR(-ENAMETOOLONG); | |
142 | if (likely(len < PATH_MAX)) { | |
143 | audit_getname(result); | |
144 | return result; | |
145 | } | |
146 | ||
147 | error: | |
148 | __putname(result); | |
149 | return err; | |
1da177e4 LT |
150 | } |
151 | ||
f52e0c11 AV |
152 | char *getname(const char __user * filename) |
153 | { | |
f7493e5d | 154 | return getname_flags(filename, 0, NULL); |
f52e0c11 AV |
155 | } |
156 | ||
1da177e4 LT |
157 | #ifdef CONFIG_AUDITSYSCALL |
158 | void putname(const char *name) | |
159 | { | |
5ac3a9c2 | 160 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
161 | audit_putname(name); |
162 | else | |
163 | __putname(name); | |
164 | } | |
165 | EXPORT_SYMBOL(putname); | |
166 | #endif | |
167 | ||
e77819e5 LT |
168 | static int check_acl(struct inode *inode, int mask) |
169 | { | |
84635d68 | 170 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
171 | struct posix_acl *acl; |
172 | ||
e77819e5 | 173 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
174 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
175 | if (!acl) | |
e77819e5 | 176 | return -EAGAIN; |
3567866b AV |
177 | /* no ->get_acl() calls in RCU mode... */ |
178 | if (acl == ACL_NOT_CACHED) | |
179 | return -ECHILD; | |
206b1d09 | 180 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
181 | } |
182 | ||
183 | acl = get_cached_acl(inode, ACL_TYPE_ACCESS); | |
184 | ||
185 | /* | |
4e34e719 CH |
186 | * A filesystem can force a ACL callback by just never filling the |
187 | * ACL cache. But normally you'd fill the cache either at inode | |
188 | * instantiation time, or on the first ->get_acl call. | |
e77819e5 | 189 | * |
4e34e719 CH |
190 | * If the filesystem doesn't have a get_acl() function at all, we'll |
191 | * just create the negative cache entry. | |
e77819e5 LT |
192 | */ |
193 | if (acl == ACL_NOT_CACHED) { | |
4e34e719 CH |
194 | if (inode->i_op->get_acl) { |
195 | acl = inode->i_op->get_acl(inode, ACL_TYPE_ACCESS); | |
196 | if (IS_ERR(acl)) | |
197 | return PTR_ERR(acl); | |
198 | } else { | |
199 | set_cached_acl(inode, ACL_TYPE_ACCESS, NULL); | |
200 | return -EAGAIN; | |
201 | } | |
e77819e5 LT |
202 | } |
203 | ||
204 | if (acl) { | |
205 | int error = posix_acl_permission(inode, acl, mask); | |
206 | posix_acl_release(acl); | |
207 | return error; | |
208 | } | |
84635d68 | 209 | #endif |
e77819e5 LT |
210 | |
211 | return -EAGAIN; | |
212 | } | |
213 | ||
5909ccaa | 214 | /* |
948409c7 | 215 | * This does the basic permission checking |
1da177e4 | 216 | */ |
7e40145e | 217 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 218 | { |
26cf46be | 219 | unsigned int mode = inode->i_mode; |
1da177e4 | 220 | |
e795b717 SH |
221 | if (current_user_ns() != inode_userns(inode)) |
222 | goto other_perms; | |
223 | ||
14067ff5 | 224 | if (likely(current_fsuid() == inode->i_uid)) |
1da177e4 LT |
225 | mode >>= 6; |
226 | else { | |
e77819e5 | 227 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 228 | int error = check_acl(inode, mask); |
b74c79e9 NP |
229 | if (error != -EAGAIN) |
230 | return error; | |
1da177e4 LT |
231 | } |
232 | ||
233 | if (in_group_p(inode->i_gid)) | |
234 | mode >>= 3; | |
235 | } | |
236 | ||
e795b717 | 237 | other_perms: |
1da177e4 LT |
238 | /* |
239 | * If the DACs are ok we don't need any capability check. | |
240 | */ | |
9c2c7039 | 241 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 242 | return 0; |
5909ccaa LT |
243 | return -EACCES; |
244 | } | |
245 | ||
246 | /** | |
b74c79e9 | 247 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 248 | * @inode: inode to check access rights for |
8fd90c8d | 249 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
250 | * |
251 | * Used to check for read/write/execute permissions on a file. | |
252 | * We use "fsuid" for this, letting us set arbitrary permissions | |
253 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
254 | * are used for other things. |
255 | * | |
256 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
257 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
258 | * It would then be called again in ref-walk mode. | |
5909ccaa | 259 | */ |
2830ba7f | 260 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
261 | { |
262 | int ret; | |
263 | ||
264 | /* | |
948409c7 | 265 | * Do the basic permission checks. |
5909ccaa | 266 | */ |
7e40145e | 267 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
268 | if (ret != -EACCES) |
269 | return ret; | |
1da177e4 | 270 | |
d594e7ec AV |
271 | if (S_ISDIR(inode->i_mode)) { |
272 | /* DACs are overridable for directories */ | |
273 | if (ns_capable(inode_userns(inode), CAP_DAC_OVERRIDE)) | |
274 | return 0; | |
275 | if (!(mask & MAY_WRITE)) | |
276 | if (ns_capable(inode_userns(inode), CAP_DAC_READ_SEARCH)) | |
277 | return 0; | |
278 | return -EACCES; | |
279 | } | |
1da177e4 LT |
280 | /* |
281 | * Read/write DACs are always overridable. | |
d594e7ec AV |
282 | * Executable DACs are overridable when there is |
283 | * at least one exec bit set. | |
1da177e4 | 284 | */ |
d594e7ec | 285 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) |
e795b717 | 286 | if (ns_capable(inode_userns(inode), CAP_DAC_OVERRIDE)) |
1da177e4 LT |
287 | return 0; |
288 | ||
289 | /* | |
290 | * Searching includes executable on directories, else just read. | |
291 | */ | |
7ea66001 | 292 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 293 | if (mask == MAY_READ) |
e795b717 | 294 | if (ns_capable(inode_userns(inode), CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
295 | return 0; |
296 | ||
297 | return -EACCES; | |
298 | } | |
299 | ||
3ddcd056 LT |
300 | /* |
301 | * We _really_ want to just do "generic_permission()" without | |
302 | * even looking at the inode->i_op values. So we keep a cache | |
303 | * flag in inode->i_opflags, that says "this has not special | |
304 | * permission function, use the fast case". | |
305 | */ | |
306 | static inline int do_inode_permission(struct inode *inode, int mask) | |
307 | { | |
308 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
309 | if (likely(inode->i_op->permission)) | |
310 | return inode->i_op->permission(inode, mask); | |
311 | ||
312 | /* This gets set once for the inode lifetime */ | |
313 | spin_lock(&inode->i_lock); | |
314 | inode->i_opflags |= IOP_FASTPERM; | |
315 | spin_unlock(&inode->i_lock); | |
316 | } | |
317 | return generic_permission(inode, mask); | |
318 | } | |
319 | ||
cb23beb5 CH |
320 | /** |
321 | * inode_permission - check for access rights to a given inode | |
322 | * @inode: inode to check permission on | |
8fd90c8d | 323 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
cb23beb5 CH |
324 | * |
325 | * Used to check for read/write/execute permissions on an inode. | |
326 | * We use "fsuid" for this, letting us set arbitrary permissions | |
327 | * for filesystem access without changing the "normal" uids which | |
328 | * are used for other things. | |
948409c7 AG |
329 | * |
330 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
cb23beb5 | 331 | */ |
f419a2e3 | 332 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 333 | { |
e6305c43 | 334 | int retval; |
1da177e4 | 335 | |
3ddcd056 | 336 | if (unlikely(mask & MAY_WRITE)) { |
22590e41 | 337 | umode_t mode = inode->i_mode; |
1da177e4 LT |
338 | |
339 | /* | |
340 | * Nobody gets write access to a read-only fs. | |
341 | */ | |
342 | if (IS_RDONLY(inode) && | |
343 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
344 | return -EROFS; | |
345 | ||
346 | /* | |
347 | * Nobody gets write access to an immutable file. | |
348 | */ | |
349 | if (IS_IMMUTABLE(inode)) | |
350 | return -EACCES; | |
351 | } | |
352 | ||
3ddcd056 | 353 | retval = do_inode_permission(inode, mask); |
1da177e4 LT |
354 | if (retval) |
355 | return retval; | |
356 | ||
08ce5f16 SH |
357 | retval = devcgroup_inode_permission(inode, mask); |
358 | if (retval) | |
359 | return retval; | |
360 | ||
d09ca739 | 361 | return security_inode_permission(inode, mask); |
1da177e4 LT |
362 | } |
363 | ||
5dd784d0 JB |
364 | /** |
365 | * path_get - get a reference to a path | |
366 | * @path: path to get the reference to | |
367 | * | |
368 | * Given a path increment the reference count to the dentry and the vfsmount. | |
369 | */ | |
370 | void path_get(struct path *path) | |
371 | { | |
372 | mntget(path->mnt); | |
373 | dget(path->dentry); | |
374 | } | |
375 | EXPORT_SYMBOL(path_get); | |
376 | ||
1d957f9b JB |
377 | /** |
378 | * path_put - put a reference to a path | |
379 | * @path: path to put the reference to | |
380 | * | |
381 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
382 | */ | |
383 | void path_put(struct path *path) | |
1da177e4 | 384 | { |
1d957f9b JB |
385 | dput(path->dentry); |
386 | mntput(path->mnt); | |
1da177e4 | 387 | } |
1d957f9b | 388 | EXPORT_SYMBOL(path_put); |
1da177e4 | 389 | |
19660af7 | 390 | /* |
31e6b01f | 391 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
392 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
393 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
394 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
395 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
396 | * got stuck, so ref-walk may continue from there. If this is not successful | |
397 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
398 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 399 | */ |
31e6b01f NP |
400 | |
401 | /** | |
19660af7 AV |
402 | * unlazy_walk - try to switch to ref-walk mode. |
403 | * @nd: nameidata pathwalk data | |
404 | * @dentry: child of nd->path.dentry or NULL | |
39191628 | 405 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 406 | * |
19660af7 AV |
407 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
408 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
409 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 410 | */ |
19660af7 | 411 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry) |
31e6b01f NP |
412 | { |
413 | struct fs_struct *fs = current->fs; | |
414 | struct dentry *parent = nd->path.dentry; | |
5b6ca027 | 415 | int want_root = 0; |
31e6b01f NP |
416 | |
417 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
5b6ca027 AV |
418 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
419 | want_root = 1; | |
31e6b01f NP |
420 | spin_lock(&fs->lock); |
421 | if (nd->root.mnt != fs->root.mnt || | |
422 | nd->root.dentry != fs->root.dentry) | |
423 | goto err_root; | |
424 | } | |
425 | spin_lock(&parent->d_lock); | |
19660af7 AV |
426 | if (!dentry) { |
427 | if (!__d_rcu_to_refcount(parent, nd->seq)) | |
428 | goto err_parent; | |
429 | BUG_ON(nd->inode != parent->d_inode); | |
430 | } else { | |
94c0d4ec AV |
431 | if (dentry->d_parent != parent) |
432 | goto err_parent; | |
19660af7 AV |
433 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); |
434 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
435 | goto err_child; | |
436 | /* | |
437 | * If the sequence check on the child dentry passed, then | |
438 | * the child has not been removed from its parent. This | |
439 | * means the parent dentry must be valid and able to take | |
440 | * a reference at this point. | |
441 | */ | |
442 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
443 | BUG_ON(!parent->d_count); | |
444 | parent->d_count++; | |
445 | spin_unlock(&dentry->d_lock); | |
446 | } | |
31e6b01f | 447 | spin_unlock(&parent->d_lock); |
5b6ca027 | 448 | if (want_root) { |
31e6b01f NP |
449 | path_get(&nd->root); |
450 | spin_unlock(&fs->lock); | |
451 | } | |
452 | mntget(nd->path.mnt); | |
453 | ||
454 | rcu_read_unlock(); | |
455 | br_read_unlock(vfsmount_lock); | |
456 | nd->flags &= ~LOOKUP_RCU; | |
457 | return 0; | |
19660af7 AV |
458 | |
459 | err_child: | |
31e6b01f | 460 | spin_unlock(&dentry->d_lock); |
19660af7 | 461 | err_parent: |
31e6b01f NP |
462 | spin_unlock(&parent->d_lock); |
463 | err_root: | |
5b6ca027 | 464 | if (want_root) |
31e6b01f NP |
465 | spin_unlock(&fs->lock); |
466 | return -ECHILD; | |
467 | } | |
468 | ||
834f2a4a TM |
469 | /** |
470 | * release_open_intent - free up open intent resources | |
471 | * @nd: pointer to nameidata | |
472 | */ | |
473 | void release_open_intent(struct nameidata *nd) | |
474 | { | |
2dab5974 LT |
475 | struct file *file = nd->intent.open.file; |
476 | ||
477 | if (file && !IS_ERR(file)) { | |
478 | if (file->f_path.dentry == NULL) | |
479 | put_filp(file); | |
480 | else | |
481 | fput(file); | |
482 | } | |
834f2a4a TM |
483 | } |
484 | ||
f60aef7e | 485 | static inline int d_revalidate(struct dentry *dentry, struct nameidata *nd) |
34286d66 | 486 | { |
f60aef7e | 487 | return dentry->d_op->d_revalidate(dentry, nd); |
34286d66 NP |
488 | } |
489 | ||
9f1fafee AV |
490 | /** |
491 | * complete_walk - successful completion of path walk | |
492 | * @nd: pointer nameidata | |
39159de2 | 493 | * |
9f1fafee AV |
494 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
495 | * Revalidate the final result, unless we'd already done that during | |
496 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
497 | * success, -error on failure. In case of failure caller does not | |
498 | * need to drop nd->path. | |
39159de2 | 499 | */ |
9f1fafee | 500 | static int complete_walk(struct nameidata *nd) |
39159de2 | 501 | { |
16c2cd71 | 502 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 503 | int status; |
39159de2 | 504 | |
9f1fafee AV |
505 | if (nd->flags & LOOKUP_RCU) { |
506 | nd->flags &= ~LOOKUP_RCU; | |
507 | if (!(nd->flags & LOOKUP_ROOT)) | |
508 | nd->root.mnt = NULL; | |
509 | spin_lock(&dentry->d_lock); | |
510 | if (unlikely(!__d_rcu_to_refcount(dentry, nd->seq))) { | |
511 | spin_unlock(&dentry->d_lock); | |
512 | rcu_read_unlock(); | |
513 | br_read_unlock(vfsmount_lock); | |
514 | return -ECHILD; | |
515 | } | |
516 | BUG_ON(nd->inode != dentry->d_inode); | |
517 | spin_unlock(&dentry->d_lock); | |
518 | mntget(nd->path.mnt); | |
519 | rcu_read_unlock(); | |
520 | br_read_unlock(vfsmount_lock); | |
521 | } | |
522 | ||
16c2cd71 AV |
523 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
524 | return 0; | |
525 | ||
526 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
39159de2 JL |
527 | return 0; |
528 | ||
16c2cd71 AV |
529 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) |
530 | return 0; | |
531 | ||
532 | /* Note: we do not d_invalidate() */ | |
34286d66 | 533 | status = d_revalidate(dentry, nd); |
39159de2 JL |
534 | if (status > 0) |
535 | return 0; | |
536 | ||
16c2cd71 | 537 | if (!status) |
39159de2 | 538 | status = -ESTALE; |
16c2cd71 | 539 | |
9f1fafee | 540 | path_put(&nd->path); |
39159de2 JL |
541 | return status; |
542 | } | |
543 | ||
2a737871 AV |
544 | static __always_inline void set_root(struct nameidata *nd) |
545 | { | |
f7ad3c6b MS |
546 | if (!nd->root.mnt) |
547 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
548 | } |
549 | ||
6de88d72 AV |
550 | static int link_path_walk(const char *, struct nameidata *); |
551 | ||
31e6b01f NP |
552 | static __always_inline void set_root_rcu(struct nameidata *nd) |
553 | { | |
554 | if (!nd->root.mnt) { | |
555 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
556 | unsigned seq; |
557 | ||
558 | do { | |
559 | seq = read_seqcount_begin(&fs->seq); | |
560 | nd->root = fs->root; | |
c1530019 | 561 | nd->seq = __read_seqcount_begin(&nd->root.dentry->d_seq); |
c28cc364 | 562 | } while (read_seqcount_retry(&fs->seq, seq)); |
31e6b01f NP |
563 | } |
564 | } | |
565 | ||
f1662356 | 566 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 567 | { |
31e6b01f NP |
568 | int ret; |
569 | ||
1da177e4 LT |
570 | if (IS_ERR(link)) |
571 | goto fail; | |
572 | ||
573 | if (*link == '/') { | |
2a737871 | 574 | set_root(nd); |
1d957f9b | 575 | path_put(&nd->path); |
2a737871 AV |
576 | nd->path = nd->root; |
577 | path_get(&nd->root); | |
16c2cd71 | 578 | nd->flags |= LOOKUP_JUMPED; |
1da177e4 | 579 | } |
31e6b01f | 580 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 581 | |
31e6b01f NP |
582 | ret = link_path_walk(link, nd); |
583 | return ret; | |
1da177e4 | 584 | fail: |
1d957f9b | 585 | path_put(&nd->path); |
1da177e4 LT |
586 | return PTR_ERR(link); |
587 | } | |
588 | ||
1d957f9b | 589 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
590 | { |
591 | dput(path->dentry); | |
4ac91378 | 592 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
593 | mntput(path->mnt); |
594 | } | |
595 | ||
7b9337aa NP |
596 | static inline void path_to_nameidata(const struct path *path, |
597 | struct nameidata *nd) | |
051d3812 | 598 | { |
31e6b01f NP |
599 | if (!(nd->flags & LOOKUP_RCU)) { |
600 | dput(nd->path.dentry); | |
601 | if (nd->path.mnt != path->mnt) | |
602 | mntput(nd->path.mnt); | |
9a229683 | 603 | } |
31e6b01f | 604 | nd->path.mnt = path->mnt; |
4ac91378 | 605 | nd->path.dentry = path->dentry; |
051d3812 IK |
606 | } |
607 | ||
574197e0 AV |
608 | static inline void put_link(struct nameidata *nd, struct path *link, void *cookie) |
609 | { | |
610 | struct inode *inode = link->dentry->d_inode; | |
611 | if (!IS_ERR(cookie) && inode->i_op->put_link) | |
612 | inode->i_op->put_link(link->dentry, nd, cookie); | |
613 | path_put(link); | |
614 | } | |
615 | ||
def4af30 | 616 | static __always_inline int |
574197e0 | 617 | follow_link(struct path *link, struct nameidata *nd, void **p) |
1da177e4 LT |
618 | { |
619 | int error; | |
7b9337aa | 620 | struct dentry *dentry = link->dentry; |
1da177e4 | 621 | |
844a3917 AV |
622 | BUG_ON(nd->flags & LOOKUP_RCU); |
623 | ||
0e794589 AV |
624 | if (link->mnt == nd->path.mnt) |
625 | mntget(link->mnt); | |
626 | ||
574197e0 AV |
627 | if (unlikely(current->total_link_count >= 40)) { |
628 | *p = ERR_PTR(-ELOOP); /* no ->put_link(), please */ | |
574197e0 AV |
629 | path_put(&nd->path); |
630 | return -ELOOP; | |
631 | } | |
632 | cond_resched(); | |
633 | current->total_link_count++; | |
634 | ||
68ac1234 | 635 | touch_atime(link); |
1da177e4 | 636 | nd_set_link(nd, NULL); |
cd4e91d3 | 637 | |
36f3b4f6 AV |
638 | error = security_inode_follow_link(link->dentry, nd); |
639 | if (error) { | |
640 | *p = ERR_PTR(error); /* no ->put_link(), please */ | |
641 | path_put(&nd->path); | |
642 | return error; | |
643 | } | |
644 | ||
86acdca1 | 645 | nd->last_type = LAST_BIND; |
def4af30 AV |
646 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
647 | error = PTR_ERR(*p); | |
648 | if (!IS_ERR(*p)) { | |
1da177e4 | 649 | char *s = nd_get_link(nd); |
cc314eef | 650 | error = 0; |
1da177e4 LT |
651 | if (s) |
652 | error = __vfs_follow_link(nd, s); | |
bcda7652 | 653 | else if (nd->last_type == LAST_BIND) { |
16c2cd71 | 654 | nd->flags |= LOOKUP_JUMPED; |
b21041d0 AV |
655 | nd->inode = nd->path.dentry->d_inode; |
656 | if (nd->inode->i_op->follow_link) { | |
bcda7652 AV |
657 | /* stepped on a _really_ weird one */ |
658 | path_put(&nd->path); | |
659 | error = -ELOOP; | |
660 | } | |
661 | } | |
1da177e4 | 662 | } |
1da177e4 LT |
663 | return error; |
664 | } | |
665 | ||
31e6b01f NP |
666 | static int follow_up_rcu(struct path *path) |
667 | { | |
0714a533 AV |
668 | struct mount *mnt = real_mount(path->mnt); |
669 | struct mount *parent; | |
31e6b01f NP |
670 | struct dentry *mountpoint; |
671 | ||
0714a533 AV |
672 | parent = mnt->mnt_parent; |
673 | if (&parent->mnt == path->mnt) | |
31e6b01f | 674 | return 0; |
a73324da | 675 | mountpoint = mnt->mnt_mountpoint; |
31e6b01f | 676 | path->dentry = mountpoint; |
0714a533 | 677 | path->mnt = &parent->mnt; |
31e6b01f NP |
678 | return 1; |
679 | } | |
680 | ||
bab77ebf | 681 | int follow_up(struct path *path) |
1da177e4 | 682 | { |
0714a533 AV |
683 | struct mount *mnt = real_mount(path->mnt); |
684 | struct mount *parent; | |
1da177e4 | 685 | struct dentry *mountpoint; |
99b7db7b NP |
686 | |
687 | br_read_lock(vfsmount_lock); | |
0714a533 AV |
688 | parent = mnt->mnt_parent; |
689 | if (&parent->mnt == path->mnt) { | |
99b7db7b | 690 | br_read_unlock(vfsmount_lock); |
1da177e4 LT |
691 | return 0; |
692 | } | |
0714a533 | 693 | mntget(&parent->mnt); |
a73324da | 694 | mountpoint = dget(mnt->mnt_mountpoint); |
99b7db7b | 695 | br_read_unlock(vfsmount_lock); |
bab77ebf AV |
696 | dput(path->dentry); |
697 | path->dentry = mountpoint; | |
698 | mntput(path->mnt); | |
0714a533 | 699 | path->mnt = &parent->mnt; |
1da177e4 LT |
700 | return 1; |
701 | } | |
702 | ||
b5c84bf6 | 703 | /* |
9875cf80 DH |
704 | * Perform an automount |
705 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
706 | * were called with. | |
1da177e4 | 707 | */ |
9875cf80 DH |
708 | static int follow_automount(struct path *path, unsigned flags, |
709 | bool *need_mntput) | |
31e6b01f | 710 | { |
9875cf80 | 711 | struct vfsmount *mnt; |
ea5b778a | 712 | int err; |
9875cf80 DH |
713 | |
714 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
715 | return -EREMOTE; | |
716 | ||
0ec26fd0 MS |
717 | /* We don't want to mount if someone's just doing a stat - |
718 | * unless they're stat'ing a directory and appended a '/' to | |
719 | * the name. | |
720 | * | |
721 | * We do, however, want to mount if someone wants to open or | |
722 | * create a file of any type under the mountpoint, wants to | |
723 | * traverse through the mountpoint or wants to open the | |
724 | * mounted directory. Also, autofs may mark negative dentries | |
725 | * as being automount points. These will need the attentions | |
726 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 727 | */ |
0ec26fd0 | 728 | if (!(flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
d94c177b | 729 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
0ec26fd0 MS |
730 | path->dentry->d_inode) |
731 | return -EISDIR; | |
732 | ||
9875cf80 DH |
733 | current->total_link_count++; |
734 | if (current->total_link_count >= 40) | |
735 | return -ELOOP; | |
736 | ||
737 | mnt = path->dentry->d_op->d_automount(path); | |
738 | if (IS_ERR(mnt)) { | |
739 | /* | |
740 | * The filesystem is allowed to return -EISDIR here to indicate | |
741 | * it doesn't want to automount. For instance, autofs would do | |
742 | * this so that its userspace daemon can mount on this dentry. | |
743 | * | |
744 | * However, we can only permit this if it's a terminal point in | |
745 | * the path being looked up; if it wasn't then the remainder of | |
746 | * the path is inaccessible and we should say so. | |
747 | */ | |
49084c3b | 748 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_PARENT)) |
9875cf80 DH |
749 | return -EREMOTE; |
750 | return PTR_ERR(mnt); | |
31e6b01f | 751 | } |
ea5b778a | 752 | |
9875cf80 DH |
753 | if (!mnt) /* mount collision */ |
754 | return 0; | |
31e6b01f | 755 | |
8aef1884 AV |
756 | if (!*need_mntput) { |
757 | /* lock_mount() may release path->mnt on error */ | |
758 | mntget(path->mnt); | |
759 | *need_mntput = true; | |
760 | } | |
19a167af | 761 | err = finish_automount(mnt, path); |
9875cf80 | 762 | |
ea5b778a DH |
763 | switch (err) { |
764 | case -EBUSY: | |
765 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 766 | return 0; |
ea5b778a | 767 | case 0: |
8aef1884 | 768 | path_put(path); |
ea5b778a DH |
769 | path->mnt = mnt; |
770 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 771 | return 0; |
19a167af AV |
772 | default: |
773 | return err; | |
ea5b778a | 774 | } |
19a167af | 775 | |
463ffb2e AV |
776 | } |
777 | ||
9875cf80 DH |
778 | /* |
779 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 780 | * - Flagged for transit management (autofs) |
9875cf80 DH |
781 | * - Flagged as mountpoint |
782 | * - Flagged as automount point | |
783 | * | |
784 | * This may only be called in refwalk mode. | |
785 | * | |
786 | * Serialization is taken care of in namespace.c | |
787 | */ | |
788 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 789 | { |
8aef1884 | 790 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
791 | unsigned managed; |
792 | bool need_mntput = false; | |
8aef1884 | 793 | int ret = 0; |
9875cf80 DH |
794 | |
795 | /* Given that we're not holding a lock here, we retain the value in a | |
796 | * local variable for each dentry as we look at it so that we don't see | |
797 | * the components of that value change under us */ | |
798 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
799 | managed &= DCACHE_MANAGED_DENTRY, | |
800 | unlikely(managed != 0)) { | |
cc53ce53 DH |
801 | /* Allow the filesystem to manage the transit without i_mutex |
802 | * being held. */ | |
803 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
804 | BUG_ON(!path->dentry->d_op); | |
805 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 806 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 807 | if (ret < 0) |
8aef1884 | 808 | break; |
cc53ce53 DH |
809 | } |
810 | ||
9875cf80 DH |
811 | /* Transit to a mounted filesystem. */ |
812 | if (managed & DCACHE_MOUNTED) { | |
813 | struct vfsmount *mounted = lookup_mnt(path); | |
814 | if (mounted) { | |
815 | dput(path->dentry); | |
816 | if (need_mntput) | |
817 | mntput(path->mnt); | |
818 | path->mnt = mounted; | |
819 | path->dentry = dget(mounted->mnt_root); | |
820 | need_mntput = true; | |
821 | continue; | |
822 | } | |
823 | ||
824 | /* Something is mounted on this dentry in another | |
825 | * namespace and/or whatever was mounted there in this | |
826 | * namespace got unmounted before we managed to get the | |
827 | * vfsmount_lock */ | |
828 | } | |
829 | ||
830 | /* Handle an automount point */ | |
831 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
832 | ret = follow_automount(path, flags, &need_mntput); | |
833 | if (ret < 0) | |
8aef1884 | 834 | break; |
9875cf80 DH |
835 | continue; |
836 | } | |
837 | ||
838 | /* We didn't change the current path point */ | |
839 | break; | |
1da177e4 | 840 | } |
8aef1884 AV |
841 | |
842 | if (need_mntput && path->mnt == mnt) | |
843 | mntput(path->mnt); | |
844 | if (ret == -EISDIR) | |
845 | ret = 0; | |
a3fbbde7 | 846 | return ret < 0 ? ret : need_mntput; |
1da177e4 LT |
847 | } |
848 | ||
cc53ce53 | 849 | int follow_down_one(struct path *path) |
1da177e4 LT |
850 | { |
851 | struct vfsmount *mounted; | |
852 | ||
1c755af4 | 853 | mounted = lookup_mnt(path); |
1da177e4 | 854 | if (mounted) { |
9393bd07 AV |
855 | dput(path->dentry); |
856 | mntput(path->mnt); | |
857 | path->mnt = mounted; | |
858 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
859 | return 1; |
860 | } | |
861 | return 0; | |
862 | } | |
863 | ||
62a7375e IK |
864 | static inline bool managed_dentry_might_block(struct dentry *dentry) |
865 | { | |
866 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT && | |
867 | dentry->d_op->d_manage(dentry, true) < 0); | |
868 | } | |
869 | ||
9875cf80 | 870 | /* |
287548e4 AV |
871 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
872 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
873 | */ |
874 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 875 | struct inode **inode) |
9875cf80 | 876 | { |
62a7375e | 877 | for (;;) { |
c7105365 | 878 | struct mount *mounted; |
62a7375e IK |
879 | /* |
880 | * Don't forget we might have a non-mountpoint managed dentry | |
881 | * that wants to block transit. | |
882 | */ | |
287548e4 | 883 | if (unlikely(managed_dentry_might_block(path->dentry))) |
ab90911f | 884 | return false; |
62a7375e IK |
885 | |
886 | if (!d_mountpoint(path->dentry)) | |
887 | break; | |
888 | ||
9875cf80 DH |
889 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
890 | if (!mounted) | |
891 | break; | |
c7105365 AV |
892 | path->mnt = &mounted->mnt; |
893 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 894 | nd->flags |= LOOKUP_JUMPED; |
9875cf80 | 895 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
896 | /* |
897 | * Update the inode too. We don't need to re-check the | |
898 | * dentry sequence number here after this d_inode read, | |
899 | * because a mount-point is always pinned. | |
900 | */ | |
901 | *inode = path->dentry->d_inode; | |
9875cf80 | 902 | } |
9875cf80 DH |
903 | return true; |
904 | } | |
905 | ||
dea39376 | 906 | static void follow_mount_rcu(struct nameidata *nd) |
287548e4 | 907 | { |
dea39376 | 908 | while (d_mountpoint(nd->path.dentry)) { |
c7105365 | 909 | struct mount *mounted; |
dea39376 | 910 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry, 1); |
287548e4 AV |
911 | if (!mounted) |
912 | break; | |
c7105365 AV |
913 | nd->path.mnt = &mounted->mnt; |
914 | nd->path.dentry = mounted->mnt.mnt_root; | |
dea39376 | 915 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
287548e4 AV |
916 | } |
917 | } | |
918 | ||
31e6b01f NP |
919 | static int follow_dotdot_rcu(struct nameidata *nd) |
920 | { | |
31e6b01f NP |
921 | set_root_rcu(nd); |
922 | ||
9875cf80 | 923 | while (1) { |
31e6b01f NP |
924 | if (nd->path.dentry == nd->root.dentry && |
925 | nd->path.mnt == nd->root.mnt) { | |
926 | break; | |
927 | } | |
928 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
929 | struct dentry *old = nd->path.dentry; | |
930 | struct dentry *parent = old->d_parent; | |
931 | unsigned seq; | |
932 | ||
933 | seq = read_seqcount_begin(&parent->d_seq); | |
934 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 935 | goto failed; |
31e6b01f NP |
936 | nd->path.dentry = parent; |
937 | nd->seq = seq; | |
938 | break; | |
939 | } | |
940 | if (!follow_up_rcu(&nd->path)) | |
941 | break; | |
942 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
31e6b01f | 943 | } |
dea39376 AV |
944 | follow_mount_rcu(nd); |
945 | nd->inode = nd->path.dentry->d_inode; | |
31e6b01f | 946 | return 0; |
ef7562d5 AV |
947 | |
948 | failed: | |
949 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
950 | if (!(nd->flags & LOOKUP_ROOT)) |
951 | nd->root.mnt = NULL; | |
ef7562d5 AV |
952 | rcu_read_unlock(); |
953 | br_read_unlock(vfsmount_lock); | |
954 | return -ECHILD; | |
31e6b01f NP |
955 | } |
956 | ||
cc53ce53 DH |
957 | /* |
958 | * Follow down to the covering mount currently visible to userspace. At each | |
959 | * point, the filesystem owning that dentry may be queried as to whether the | |
960 | * caller is permitted to proceed or not. | |
cc53ce53 | 961 | */ |
7cc90cc3 | 962 | int follow_down(struct path *path) |
cc53ce53 DH |
963 | { |
964 | unsigned managed; | |
965 | int ret; | |
966 | ||
967 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
968 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
969 | /* Allow the filesystem to manage the transit without i_mutex | |
970 | * being held. | |
971 | * | |
972 | * We indicate to the filesystem if someone is trying to mount | |
973 | * something here. This gives autofs the chance to deny anyone | |
974 | * other than its daemon the right to mount on its | |
975 | * superstructure. | |
976 | * | |
977 | * The filesystem may sleep at this point. | |
978 | */ | |
979 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
980 | BUG_ON(!path->dentry->d_op); | |
981 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 982 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 983 | path->dentry, false); |
cc53ce53 DH |
984 | if (ret < 0) |
985 | return ret == -EISDIR ? 0 : ret; | |
986 | } | |
987 | ||
988 | /* Transit to a mounted filesystem. */ | |
989 | if (managed & DCACHE_MOUNTED) { | |
990 | struct vfsmount *mounted = lookup_mnt(path); | |
991 | if (!mounted) | |
992 | break; | |
993 | dput(path->dentry); | |
994 | mntput(path->mnt); | |
995 | path->mnt = mounted; | |
996 | path->dentry = dget(mounted->mnt_root); | |
997 | continue; | |
998 | } | |
999 | ||
1000 | /* Don't handle automount points here */ | |
1001 | break; | |
1002 | } | |
1003 | return 0; | |
1004 | } | |
1005 | ||
9875cf80 DH |
1006 | /* |
1007 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1008 | */ | |
1009 | static void follow_mount(struct path *path) | |
1010 | { | |
1011 | while (d_mountpoint(path->dentry)) { | |
1012 | struct vfsmount *mounted = lookup_mnt(path); | |
1013 | if (!mounted) | |
1014 | break; | |
1015 | dput(path->dentry); | |
1016 | mntput(path->mnt); | |
1017 | path->mnt = mounted; | |
1018 | path->dentry = dget(mounted->mnt_root); | |
1019 | } | |
1020 | } | |
1021 | ||
31e6b01f | 1022 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1023 | { |
2a737871 | 1024 | set_root(nd); |
e518ddb7 | 1025 | |
1da177e4 | 1026 | while(1) { |
4ac91378 | 1027 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1028 | |
2a737871 AV |
1029 | if (nd->path.dentry == nd->root.dentry && |
1030 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1031 | break; |
1032 | } | |
4ac91378 | 1033 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1034 | /* rare case of legitimate dget_parent()... */ |
1035 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1036 | dput(old); |
1037 | break; | |
1038 | } | |
3088dd70 | 1039 | if (!follow_up(&nd->path)) |
1da177e4 | 1040 | break; |
1da177e4 | 1041 | } |
79ed0226 | 1042 | follow_mount(&nd->path); |
31e6b01f | 1043 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1044 | } |
1045 | ||
baa03890 | 1046 | /* |
bad61189 MS |
1047 | * This looks up the name in dcache, possibly revalidates the old dentry and |
1048 | * allocates a new one if not found or not valid. In the need_lookup argument | |
1049 | * returns whether i_op->lookup is necessary. | |
1050 | * | |
1051 | * dir->d_inode->i_mutex must be held | |
baa03890 | 1052 | */ |
bad61189 MS |
1053 | static struct dentry *lookup_dcache(struct qstr *name, struct dentry *dir, |
1054 | struct nameidata *nd, bool *need_lookup) | |
baa03890 | 1055 | { |
baa03890 | 1056 | struct dentry *dentry; |
bad61189 | 1057 | int error; |
baa03890 | 1058 | |
bad61189 MS |
1059 | *need_lookup = false; |
1060 | dentry = d_lookup(dir, name); | |
1061 | if (dentry) { | |
1062 | if (d_need_lookup(dentry)) { | |
1063 | *need_lookup = true; | |
1064 | } else if (dentry->d_flags & DCACHE_OP_REVALIDATE) { | |
1065 | error = d_revalidate(dentry, nd); | |
1066 | if (unlikely(error <= 0)) { | |
1067 | if (error < 0) { | |
1068 | dput(dentry); | |
1069 | return ERR_PTR(error); | |
1070 | } else if (!d_invalidate(dentry)) { | |
1071 | dput(dentry); | |
1072 | dentry = NULL; | |
1073 | } | |
1074 | } | |
1075 | } | |
1076 | } | |
baa03890 | 1077 | |
bad61189 MS |
1078 | if (!dentry) { |
1079 | dentry = d_alloc(dir, name); | |
1080 | if (unlikely(!dentry)) | |
1081 | return ERR_PTR(-ENOMEM); | |
baa03890 | 1082 | |
bad61189 | 1083 | *need_lookup = true; |
baa03890 NP |
1084 | } |
1085 | return dentry; | |
1086 | } | |
1087 | ||
44396f4b | 1088 | /* |
bad61189 MS |
1089 | * Call i_op->lookup on the dentry. The dentry must be negative but may be |
1090 | * hashed if it was pouplated with DCACHE_NEED_LOOKUP. | |
1091 | * | |
1092 | * dir->d_inode->i_mutex must be held | |
44396f4b | 1093 | */ |
bad61189 MS |
1094 | static struct dentry *lookup_real(struct inode *dir, struct dentry *dentry, |
1095 | struct nameidata *nd) | |
44396f4b | 1096 | { |
44396f4b JB |
1097 | struct dentry *old; |
1098 | ||
1099 | /* Don't create child dentry for a dead directory. */ | |
bad61189 | 1100 | if (unlikely(IS_DEADDIR(dir))) { |
e188dc02 | 1101 | dput(dentry); |
44396f4b | 1102 | return ERR_PTR(-ENOENT); |
e188dc02 | 1103 | } |
44396f4b | 1104 | |
bad61189 | 1105 | old = dir->i_op->lookup(dir, dentry, nd); |
44396f4b JB |
1106 | if (unlikely(old)) { |
1107 | dput(dentry); | |
1108 | dentry = old; | |
1109 | } | |
1110 | return dentry; | |
1111 | } | |
1112 | ||
a3255546 AV |
1113 | static struct dentry *__lookup_hash(struct qstr *name, |
1114 | struct dentry *base, struct nameidata *nd) | |
1115 | { | |
bad61189 | 1116 | bool need_lookup; |
a3255546 AV |
1117 | struct dentry *dentry; |
1118 | ||
bad61189 MS |
1119 | dentry = lookup_dcache(name, base, nd, &need_lookup); |
1120 | if (!need_lookup) | |
1121 | return dentry; | |
a3255546 | 1122 | |
bad61189 | 1123 | return lookup_real(base->d_inode, dentry, nd); |
a3255546 AV |
1124 | } |
1125 | ||
1da177e4 LT |
1126 | /* |
1127 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1128 | * small and for now I'd prefer to have fast path as straight as possible. | |
1129 | * It _is_ time-critical. | |
1130 | */ | |
1131 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
31e6b01f | 1132 | struct path *path, struct inode **inode) |
1da177e4 | 1133 | { |
4ac91378 | 1134 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1135 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1136 | int need_reval = 1; |
1137 | int status = 1; | |
9875cf80 DH |
1138 | int err; |
1139 | ||
b04f784e NP |
1140 | /* |
1141 | * Rename seqlock is not required here because in the off chance | |
1142 | * of a false negative due to a concurrent rename, we're going to | |
1143 | * do the non-racy lookup, below. | |
1144 | */ | |
31e6b01f NP |
1145 | if (nd->flags & LOOKUP_RCU) { |
1146 | unsigned seq; | |
12f8ad4b | 1147 | dentry = __d_lookup_rcu(parent, name, &seq, nd->inode); |
5a18fff2 AV |
1148 | if (!dentry) |
1149 | goto unlazy; | |
1150 | ||
12f8ad4b LT |
1151 | /* |
1152 | * This sequence count validates that the inode matches | |
1153 | * the dentry name information from lookup. | |
1154 | */ | |
1155 | *inode = dentry->d_inode; | |
1156 | if (read_seqcount_retry(&dentry->d_seq, seq)) | |
1157 | return -ECHILD; | |
1158 | ||
1159 | /* | |
1160 | * This sequence count validates that the parent had no | |
1161 | * changes while we did the lookup of the dentry above. | |
1162 | * | |
1163 | * The memory barrier in read_seqcount_begin of child is | |
1164 | * enough, we can use __read_seqcount_retry here. | |
1165 | */ | |
31e6b01f NP |
1166 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) |
1167 | return -ECHILD; | |
31e6b01f | 1168 | nd->seq = seq; |
5a18fff2 | 1169 | |
fa4ee159 MS |
1170 | if (unlikely(d_need_lookup(dentry))) |
1171 | goto unlazy; | |
24643087 | 1172 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
5a18fff2 AV |
1173 | status = d_revalidate(dentry, nd); |
1174 | if (unlikely(status <= 0)) { | |
1175 | if (status != -ECHILD) | |
1176 | need_reval = 0; | |
1177 | goto unlazy; | |
1178 | } | |
24643087 | 1179 | } |
31e6b01f NP |
1180 | path->mnt = mnt; |
1181 | path->dentry = dentry; | |
d6e9bd25 AV |
1182 | if (unlikely(!__follow_mount_rcu(nd, path, inode))) |
1183 | goto unlazy; | |
1184 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1185 | goto unlazy; | |
1186 | return 0; | |
5a18fff2 | 1187 | unlazy: |
19660af7 AV |
1188 | if (unlazy_walk(nd, dentry)) |
1189 | return -ECHILD; | |
5a18fff2 AV |
1190 | } else { |
1191 | dentry = __d_lookup(parent, name); | |
9875cf80 | 1192 | } |
5a18fff2 | 1193 | |
81e6f520 AV |
1194 | if (unlikely(!dentry)) |
1195 | goto need_lookup; | |
1196 | ||
1197 | if (unlikely(d_need_lookup(dentry))) { | |
44396f4b | 1198 | dput(dentry); |
81e6f520 | 1199 | goto need_lookup; |
5a18fff2 | 1200 | } |
81e6f520 | 1201 | |
5a18fff2 AV |
1202 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) |
1203 | status = d_revalidate(dentry, nd); | |
1204 | if (unlikely(status <= 0)) { | |
1205 | if (status < 0) { | |
1206 | dput(dentry); | |
1207 | return status; | |
1208 | } | |
1209 | if (!d_invalidate(dentry)) { | |
1210 | dput(dentry); | |
81e6f520 | 1211 | goto need_lookup; |
5a18fff2 | 1212 | } |
24643087 | 1213 | } |
3f6c7c71 | 1214 | done: |
9875cf80 DH |
1215 | path->mnt = mnt; |
1216 | path->dentry = dentry; | |
1217 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1218 | if (unlikely(err < 0)) { |
1219 | path_put_conditional(path, nd); | |
9875cf80 | 1220 | return err; |
89312214 | 1221 | } |
a3fbbde7 AV |
1222 | if (err) |
1223 | nd->flags |= LOOKUP_JUMPED; | |
9875cf80 | 1224 | *inode = path->dentry->d_inode; |
1da177e4 | 1225 | return 0; |
81e6f520 AV |
1226 | |
1227 | need_lookup: | |
1228 | BUG_ON(nd->inode != parent->d_inode); | |
1229 | ||
1230 | mutex_lock(&parent->d_inode->i_mutex); | |
1231 | dentry = __lookup_hash(name, parent, nd); | |
1232 | mutex_unlock(&parent->d_inode->i_mutex); | |
1233 | if (IS_ERR(dentry)) | |
1234 | return PTR_ERR(dentry); | |
1235 | goto done; | |
1da177e4 LT |
1236 | } |
1237 | ||
52094c8a AV |
1238 | static inline int may_lookup(struct nameidata *nd) |
1239 | { | |
1240 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1241 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1242 | if (err != -ECHILD) |
1243 | return err; | |
19660af7 | 1244 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1245 | return -ECHILD; |
1246 | } | |
4ad5abb3 | 1247 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1248 | } |
1249 | ||
9856fa1b AV |
1250 | static inline int handle_dots(struct nameidata *nd, int type) |
1251 | { | |
1252 | if (type == LAST_DOTDOT) { | |
1253 | if (nd->flags & LOOKUP_RCU) { | |
1254 | if (follow_dotdot_rcu(nd)) | |
1255 | return -ECHILD; | |
1256 | } else | |
1257 | follow_dotdot(nd); | |
1258 | } | |
1259 | return 0; | |
1260 | } | |
1261 | ||
951361f9 AV |
1262 | static void terminate_walk(struct nameidata *nd) |
1263 | { | |
1264 | if (!(nd->flags & LOOKUP_RCU)) { | |
1265 | path_put(&nd->path); | |
1266 | } else { | |
1267 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1268 | if (!(nd->flags & LOOKUP_ROOT)) |
1269 | nd->root.mnt = NULL; | |
951361f9 AV |
1270 | rcu_read_unlock(); |
1271 | br_read_unlock(vfsmount_lock); | |
1272 | } | |
1273 | } | |
1274 | ||
3ddcd056 LT |
1275 | /* |
1276 | * Do we need to follow links? We _really_ want to be able | |
1277 | * to do this check without having to look at inode->i_op, | |
1278 | * so we keep a cache of "no, this doesn't need follow_link" | |
1279 | * for the common case. | |
1280 | */ | |
7813b94a | 1281 | static inline int should_follow_link(struct inode *inode, int follow) |
3ddcd056 LT |
1282 | { |
1283 | if (unlikely(!(inode->i_opflags & IOP_NOFOLLOW))) { | |
1284 | if (likely(inode->i_op->follow_link)) | |
1285 | return follow; | |
1286 | ||
1287 | /* This gets set once for the inode lifetime */ | |
1288 | spin_lock(&inode->i_lock); | |
1289 | inode->i_opflags |= IOP_NOFOLLOW; | |
1290 | spin_unlock(&inode->i_lock); | |
1291 | } | |
1292 | return 0; | |
1293 | } | |
1294 | ||
ce57dfc1 AV |
1295 | static inline int walk_component(struct nameidata *nd, struct path *path, |
1296 | struct qstr *name, int type, int follow) | |
1297 | { | |
1298 | struct inode *inode; | |
1299 | int err; | |
1300 | /* | |
1301 | * "." and ".." are special - ".." especially so because it has | |
1302 | * to be able to know about the current root directory and | |
1303 | * parent relationships. | |
1304 | */ | |
1305 | if (unlikely(type != LAST_NORM)) | |
1306 | return handle_dots(nd, type); | |
1307 | err = do_lookup(nd, name, path, &inode); | |
1308 | if (unlikely(err)) { | |
1309 | terminate_walk(nd); | |
1310 | return err; | |
1311 | } | |
1312 | if (!inode) { | |
1313 | path_to_nameidata(path, nd); | |
1314 | terminate_walk(nd); | |
1315 | return -ENOENT; | |
1316 | } | |
7813b94a | 1317 | if (should_follow_link(inode, follow)) { |
19660af7 AV |
1318 | if (nd->flags & LOOKUP_RCU) { |
1319 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
1320 | terminate_walk(nd); | |
1321 | return -ECHILD; | |
1322 | } | |
1323 | } | |
ce57dfc1 AV |
1324 | BUG_ON(inode != path->dentry->d_inode); |
1325 | return 1; | |
1326 | } | |
1327 | path_to_nameidata(path, nd); | |
1328 | nd->inode = inode; | |
1329 | return 0; | |
1330 | } | |
1331 | ||
b356379a AV |
1332 | /* |
1333 | * This limits recursive symlink follows to 8, while | |
1334 | * limiting consecutive symlinks to 40. | |
1335 | * | |
1336 | * Without that kind of total limit, nasty chains of consecutive | |
1337 | * symlinks can cause almost arbitrarily long lookups. | |
1338 | */ | |
1339 | static inline int nested_symlink(struct path *path, struct nameidata *nd) | |
1340 | { | |
1341 | int res; | |
1342 | ||
b356379a AV |
1343 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1344 | path_put_conditional(path, nd); | |
1345 | path_put(&nd->path); | |
1346 | return -ELOOP; | |
1347 | } | |
1a4022f8 | 1348 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); |
b356379a AV |
1349 | |
1350 | nd->depth++; | |
1351 | current->link_count++; | |
1352 | ||
1353 | do { | |
1354 | struct path link = *path; | |
1355 | void *cookie; | |
574197e0 AV |
1356 | |
1357 | res = follow_link(&link, nd, &cookie); | |
b356379a AV |
1358 | if (!res) |
1359 | res = walk_component(nd, path, &nd->last, | |
1360 | nd->last_type, LOOKUP_FOLLOW); | |
574197e0 | 1361 | put_link(nd, &link, cookie); |
b356379a AV |
1362 | } while (res > 0); |
1363 | ||
1364 | current->link_count--; | |
1365 | nd->depth--; | |
1366 | return res; | |
1367 | } | |
1368 | ||
3ddcd056 LT |
1369 | /* |
1370 | * We really don't want to look at inode->i_op->lookup | |
1371 | * when we don't have to. So we keep a cache bit in | |
1372 | * the inode ->i_opflags field that says "yes, we can | |
1373 | * do lookup on this inode". | |
1374 | */ | |
1375 | static inline int can_lookup(struct inode *inode) | |
1376 | { | |
1377 | if (likely(inode->i_opflags & IOP_LOOKUP)) | |
1378 | return 1; | |
1379 | if (likely(!inode->i_op->lookup)) | |
1380 | return 0; | |
1381 | ||
1382 | /* We do this once for the lifetime of the inode */ | |
1383 | spin_lock(&inode->i_lock); | |
1384 | inode->i_opflags |= IOP_LOOKUP; | |
1385 | spin_unlock(&inode->i_lock); | |
1386 | return 1; | |
1387 | } | |
1388 | ||
bfcfaa77 LT |
1389 | /* |
1390 | * We can do the critical dentry name comparison and hashing | |
1391 | * operations one word at a time, but we are limited to: | |
1392 | * | |
1393 | * - Architectures with fast unaligned word accesses. We could | |
1394 | * do a "get_unaligned()" if this helps and is sufficiently | |
1395 | * fast. | |
1396 | * | |
1397 | * - Little-endian machines (so that we can generate the mask | |
1398 | * of low bytes efficiently). Again, we *could* do a byte | |
1399 | * swapping load on big-endian architectures if that is not | |
1400 | * expensive enough to make the optimization worthless. | |
1401 | * | |
1402 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we | |
1403 | * do not trap on the (extremely unlikely) case of a page | |
1404 | * crossing operation. | |
1405 | * | |
1406 | * - Furthermore, we need an efficient 64-bit compile for the | |
1407 | * 64-bit case in order to generate the "number of bytes in | |
1408 | * the final mask". Again, that could be replaced with a | |
1409 | * efficient population count instruction or similar. | |
1410 | */ | |
1411 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1412 | ||
f68e556e | 1413 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1414 | |
f68e556e | 1415 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1416 | |
1417 | static inline unsigned int fold_hash(unsigned long hash) | |
1418 | { | |
1419 | hash += hash >> (8*sizeof(int)); | |
1420 | return hash; | |
1421 | } | |
1422 | ||
1423 | #else /* 32-bit case */ | |
1424 | ||
bfcfaa77 LT |
1425 | #define fold_hash(x) (x) |
1426 | ||
1427 | #endif | |
1428 | ||
1429 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1430 | { | |
1431 | unsigned long a, mask; | |
1432 | unsigned long hash = 0; | |
1433 | ||
1434 | for (;;) { | |
e419b4cc | 1435 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1436 | if (len < sizeof(unsigned long)) |
1437 | break; | |
1438 | hash += a; | |
f132c5be | 1439 | hash *= 9; |
bfcfaa77 LT |
1440 | name += sizeof(unsigned long); |
1441 | len -= sizeof(unsigned long); | |
1442 | if (!len) | |
1443 | goto done; | |
1444 | } | |
1445 | mask = ~(~0ul << len*8); | |
1446 | hash += mask & a; | |
1447 | done: | |
1448 | return fold_hash(hash); | |
1449 | } | |
1450 | EXPORT_SYMBOL(full_name_hash); | |
1451 | ||
bfcfaa77 LT |
1452 | /* |
1453 | * Calculate the length and hash of the path component, and | |
1454 | * return the length of the component; | |
1455 | */ | |
1456 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1457 | { | |
1458 | unsigned long a, mask, hash, len; | |
1459 | ||
1460 | hash = a = 0; | |
1461 | len = -sizeof(unsigned long); | |
1462 | do { | |
1463 | hash = (hash + a) * 9; | |
1464 | len += sizeof(unsigned long); | |
e419b4cc | 1465 | a = load_unaligned_zeropad(name+len); |
bfcfaa77 | 1466 | /* Do we have any NUL or '/' bytes in this word? */ |
f68e556e | 1467 | mask = has_zero(a) | has_zero(a ^ REPEAT_BYTE('/')); |
bfcfaa77 LT |
1468 | } while (!mask); |
1469 | ||
1470 | /* The mask *below* the first high bit set */ | |
1471 | mask = (mask - 1) & ~mask; | |
1472 | mask >>= 7; | |
1473 | hash += a & mask; | |
1474 | *hashp = fold_hash(hash); | |
1475 | ||
1476 | return len + count_masked_bytes(mask); | |
1477 | } | |
1478 | ||
1479 | #else | |
1480 | ||
0145acc2 LT |
1481 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1482 | { | |
1483 | unsigned long hash = init_name_hash(); | |
1484 | while (len--) | |
1485 | hash = partial_name_hash(*name++, hash); | |
1486 | return end_name_hash(hash); | |
1487 | } | |
ae942ae7 | 1488 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1489 | |
200e9ef7 LT |
1490 | /* |
1491 | * We know there's a real path component here of at least | |
1492 | * one character. | |
1493 | */ | |
1494 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1495 | { | |
1496 | unsigned long hash = init_name_hash(); | |
1497 | unsigned long len = 0, c; | |
1498 | ||
1499 | c = (unsigned char)*name; | |
1500 | do { | |
1501 | len++; | |
1502 | hash = partial_name_hash(c, hash); | |
1503 | c = (unsigned char)name[len]; | |
1504 | } while (c && c != '/'); | |
1505 | *hashp = end_name_hash(hash); | |
1506 | return len; | |
1507 | } | |
1508 | ||
bfcfaa77 LT |
1509 | #endif |
1510 | ||
1da177e4 LT |
1511 | /* |
1512 | * Name resolution. | |
ea3834d9 PM |
1513 | * This is the basic name resolution function, turning a pathname into |
1514 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1515 | * |
ea3834d9 PM |
1516 | * Returns 0 and nd will have valid dentry and mnt on success. |
1517 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1518 | */ |
6de88d72 | 1519 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1520 | { |
1521 | struct path next; | |
1da177e4 | 1522 | int err; |
1da177e4 LT |
1523 | |
1524 | while (*name=='/') | |
1525 | name++; | |
1526 | if (!*name) | |
086e183a | 1527 | return 0; |
1da177e4 | 1528 | |
1da177e4 LT |
1529 | /* At this point we know we have a real path component. */ |
1530 | for(;;) { | |
1da177e4 | 1531 | struct qstr this; |
200e9ef7 | 1532 | long len; |
fe479a58 | 1533 | int type; |
1da177e4 | 1534 | |
52094c8a | 1535 | err = may_lookup(nd); |
1da177e4 LT |
1536 | if (err) |
1537 | break; | |
1538 | ||
200e9ef7 | 1539 | len = hash_name(name, &this.hash); |
1da177e4 | 1540 | this.name = name; |
200e9ef7 | 1541 | this.len = len; |
1da177e4 | 1542 | |
fe479a58 | 1543 | type = LAST_NORM; |
200e9ef7 | 1544 | if (name[0] == '.') switch (len) { |
fe479a58 | 1545 | case 2: |
200e9ef7 | 1546 | if (name[1] == '.') { |
fe479a58 | 1547 | type = LAST_DOTDOT; |
16c2cd71 AV |
1548 | nd->flags |= LOOKUP_JUMPED; |
1549 | } | |
fe479a58 AV |
1550 | break; |
1551 | case 1: | |
1552 | type = LAST_DOT; | |
1553 | } | |
5a202bcd AV |
1554 | if (likely(type == LAST_NORM)) { |
1555 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1556 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd AV |
1557 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
1558 | err = parent->d_op->d_hash(parent, nd->inode, | |
1559 | &this); | |
1560 | if (err < 0) | |
1561 | break; | |
1562 | } | |
1563 | } | |
fe479a58 | 1564 | |
200e9ef7 | 1565 | if (!name[len]) |
1da177e4 | 1566 | goto last_component; |
200e9ef7 LT |
1567 | /* |
1568 | * If it wasn't NUL, we know it was '/'. Skip that | |
1569 | * slash, and continue until no more slashes. | |
1570 | */ | |
1571 | do { | |
1572 | len++; | |
1573 | } while (unlikely(name[len] == '/')); | |
1574 | if (!name[len]) | |
b356379a | 1575 | goto last_component; |
200e9ef7 | 1576 | name += len; |
1da177e4 | 1577 | |
ce57dfc1 AV |
1578 | err = walk_component(nd, &next, &this, type, LOOKUP_FOLLOW); |
1579 | if (err < 0) | |
1580 | return err; | |
1da177e4 | 1581 | |
ce57dfc1 | 1582 | if (err) { |
b356379a | 1583 | err = nested_symlink(&next, nd); |
1da177e4 | 1584 | if (err) |
a7472bab | 1585 | return err; |
31e6b01f | 1586 | } |
3ddcd056 LT |
1587 | if (can_lookup(nd->inode)) |
1588 | continue; | |
1da177e4 | 1589 | err = -ENOTDIR; |
3ddcd056 | 1590 | break; |
1da177e4 LT |
1591 | /* here ends the main loop */ |
1592 | ||
1da177e4 | 1593 | last_component: |
b356379a AV |
1594 | nd->last = this; |
1595 | nd->last_type = type; | |
086e183a | 1596 | return 0; |
1da177e4 | 1597 | } |
951361f9 | 1598 | terminate_walk(nd); |
1da177e4 LT |
1599 | return err; |
1600 | } | |
1601 | ||
70e9b357 AV |
1602 | static int path_init(int dfd, const char *name, unsigned int flags, |
1603 | struct nameidata *nd, struct file **fp) | |
31e6b01f NP |
1604 | { |
1605 | int retval = 0; | |
1606 | int fput_needed; | |
1607 | struct file *file; | |
1608 | ||
1609 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
16c2cd71 | 1610 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 1611 | nd->depth = 0; |
5b6ca027 AV |
1612 | if (flags & LOOKUP_ROOT) { |
1613 | struct inode *inode = nd->root.dentry->d_inode; | |
73d049a4 AV |
1614 | if (*name) { |
1615 | if (!inode->i_op->lookup) | |
1616 | return -ENOTDIR; | |
1617 | retval = inode_permission(inode, MAY_EXEC); | |
1618 | if (retval) | |
1619 | return retval; | |
1620 | } | |
5b6ca027 AV |
1621 | nd->path = nd->root; |
1622 | nd->inode = inode; | |
1623 | if (flags & LOOKUP_RCU) { | |
1624 | br_read_lock(vfsmount_lock); | |
1625 | rcu_read_lock(); | |
1626 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1627 | } else { | |
1628 | path_get(&nd->path); | |
1629 | } | |
1630 | return 0; | |
1631 | } | |
1632 | ||
31e6b01f | 1633 | nd->root.mnt = NULL; |
31e6b01f NP |
1634 | |
1635 | if (*name=='/') { | |
e41f7d4e AV |
1636 | if (flags & LOOKUP_RCU) { |
1637 | br_read_lock(vfsmount_lock); | |
1638 | rcu_read_lock(); | |
1639 | set_root_rcu(nd); | |
1640 | } else { | |
1641 | set_root(nd); | |
1642 | path_get(&nd->root); | |
1643 | } | |
1644 | nd->path = nd->root; | |
31e6b01f | 1645 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1646 | if (flags & LOOKUP_RCU) { |
1647 | struct fs_struct *fs = current->fs; | |
1648 | unsigned seq; | |
31e6b01f | 1649 | |
e41f7d4e AV |
1650 | br_read_lock(vfsmount_lock); |
1651 | rcu_read_lock(); | |
c28cc364 | 1652 | |
e41f7d4e AV |
1653 | do { |
1654 | seq = read_seqcount_begin(&fs->seq); | |
1655 | nd->path = fs->pwd; | |
1656 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1657 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1658 | } else { | |
1659 | get_fs_pwd(current->fs, &nd->path); | |
1660 | } | |
31e6b01f NP |
1661 | } else { |
1662 | struct dentry *dentry; | |
1663 | ||
1abf0c71 | 1664 | file = fget_raw_light(dfd, &fput_needed); |
31e6b01f NP |
1665 | retval = -EBADF; |
1666 | if (!file) | |
1667 | goto out_fail; | |
1668 | ||
1669 | dentry = file->f_path.dentry; | |
1670 | ||
f52e0c11 AV |
1671 | if (*name) { |
1672 | retval = -ENOTDIR; | |
1673 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
1674 | goto fput_fail; | |
31e6b01f | 1675 | |
4ad5abb3 | 1676 | retval = inode_permission(dentry->d_inode, MAY_EXEC); |
f52e0c11 AV |
1677 | if (retval) |
1678 | goto fput_fail; | |
1679 | } | |
31e6b01f NP |
1680 | |
1681 | nd->path = file->f_path; | |
e41f7d4e AV |
1682 | if (flags & LOOKUP_RCU) { |
1683 | if (fput_needed) | |
70e9b357 | 1684 | *fp = file; |
e41f7d4e AV |
1685 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
1686 | br_read_lock(vfsmount_lock); | |
1687 | rcu_read_lock(); | |
1688 | } else { | |
1689 | path_get(&file->f_path); | |
1690 | fput_light(file, fput_needed); | |
1691 | } | |
31e6b01f | 1692 | } |
31e6b01f | 1693 | |
31e6b01f | 1694 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1695 | return 0; |
2dfdd266 | 1696 | |
9b4a9b14 AV |
1697 | fput_fail: |
1698 | fput_light(file, fput_needed); | |
1699 | out_fail: | |
1700 | return retval; | |
1701 | } | |
1702 | ||
bd92d7fe AV |
1703 | static inline int lookup_last(struct nameidata *nd, struct path *path) |
1704 | { | |
1705 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1706 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1707 | ||
1708 | nd->flags &= ~LOOKUP_PARENT; | |
1709 | return walk_component(nd, path, &nd->last, nd->last_type, | |
1710 | nd->flags & LOOKUP_FOLLOW); | |
1711 | } | |
1712 | ||
9b4a9b14 | 1713 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
ee0827cd | 1714 | static int path_lookupat(int dfd, const char *name, |
9b4a9b14 AV |
1715 | unsigned int flags, struct nameidata *nd) |
1716 | { | |
70e9b357 | 1717 | struct file *base = NULL; |
bd92d7fe AV |
1718 | struct path path; |
1719 | int err; | |
31e6b01f NP |
1720 | |
1721 | /* | |
1722 | * Path walking is largely split up into 2 different synchronisation | |
1723 | * schemes, rcu-walk and ref-walk (explained in | |
1724 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1725 | * path walk code, but some things particularly setup, cleanup, and | |
1726 | * following mounts are sufficiently divergent that functions are | |
1727 | * duplicated. Typically there is a function foo(), and its RCU | |
1728 | * analogue, foo_rcu(). | |
1729 | * | |
1730 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1731 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1732 | * be handled by restarting a traditional ref-walk (which will always | |
1733 | * be able to complete). | |
1734 | */ | |
bd92d7fe | 1735 | err = path_init(dfd, name, flags | LOOKUP_PARENT, nd, &base); |
ee0827cd | 1736 | |
bd92d7fe AV |
1737 | if (unlikely(err)) |
1738 | return err; | |
ee0827cd AV |
1739 | |
1740 | current->total_link_count = 0; | |
bd92d7fe AV |
1741 | err = link_path_walk(name, nd); |
1742 | ||
1743 | if (!err && !(flags & LOOKUP_PARENT)) { | |
bd92d7fe AV |
1744 | err = lookup_last(nd, &path); |
1745 | while (err > 0) { | |
1746 | void *cookie; | |
1747 | struct path link = path; | |
bd92d7fe | 1748 | nd->flags |= LOOKUP_PARENT; |
574197e0 | 1749 | err = follow_link(&link, nd, &cookie); |
bd92d7fe AV |
1750 | if (!err) |
1751 | err = lookup_last(nd, &path); | |
574197e0 | 1752 | put_link(nd, &link, cookie); |
bd92d7fe AV |
1753 | } |
1754 | } | |
ee0827cd | 1755 | |
9f1fafee AV |
1756 | if (!err) |
1757 | err = complete_walk(nd); | |
bd92d7fe AV |
1758 | |
1759 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
1760 | if (!nd->inode->i_op->lookup) { | |
1761 | path_put(&nd->path); | |
bd23a539 | 1762 | err = -ENOTDIR; |
bd92d7fe AV |
1763 | } |
1764 | } | |
16c2cd71 | 1765 | |
70e9b357 AV |
1766 | if (base) |
1767 | fput(base); | |
ee0827cd | 1768 | |
5b6ca027 | 1769 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
2a737871 AV |
1770 | path_put(&nd->root); |
1771 | nd->root.mnt = NULL; | |
1772 | } | |
bd92d7fe | 1773 | return err; |
ee0827cd | 1774 | } |
31e6b01f | 1775 | |
ee0827cd AV |
1776 | static int do_path_lookup(int dfd, const char *name, |
1777 | unsigned int flags, struct nameidata *nd) | |
1778 | { | |
1779 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); | |
1780 | if (unlikely(retval == -ECHILD)) | |
1781 | retval = path_lookupat(dfd, name, flags, nd); | |
1782 | if (unlikely(retval == -ESTALE)) | |
1783 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); | |
31e6b01f NP |
1784 | |
1785 | if (likely(!retval)) { | |
1786 | if (unlikely(!audit_dummy_context())) { | |
1787 | if (nd->path.dentry && nd->inode) | |
1788 | audit_inode(name, nd->path.dentry); | |
1789 | } | |
1790 | } | |
170aa3d0 | 1791 | return retval; |
1da177e4 LT |
1792 | } |
1793 | ||
c9c6cac0 | 1794 | int kern_path_parent(const char *name, struct nameidata *nd) |
5590ff0d | 1795 | { |
c9c6cac0 | 1796 | return do_path_lookup(AT_FDCWD, name, LOOKUP_PARENT, nd); |
5590ff0d UD |
1797 | } |
1798 | ||
d1811465 AV |
1799 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1800 | { | |
1801 | struct nameidata nd; | |
1802 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1803 | if (!res) | |
1804 | *path = nd.path; | |
1805 | return res; | |
1806 | } | |
1807 | ||
16f18200 JJS |
1808 | /** |
1809 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1810 | * @dentry: pointer to dentry of the base directory | |
1811 | * @mnt: pointer to vfs mount of the base directory | |
1812 | * @name: pointer to file name | |
1813 | * @flags: lookup flags | |
e0a01249 | 1814 | * @path: pointer to struct path to fill |
16f18200 JJS |
1815 | */ |
1816 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1817 | const char *name, unsigned int flags, | |
e0a01249 | 1818 | struct path *path) |
16f18200 | 1819 | { |
e0a01249 AV |
1820 | struct nameidata nd; |
1821 | int err; | |
1822 | nd.root.dentry = dentry; | |
1823 | nd.root.mnt = mnt; | |
1824 | BUG_ON(flags & LOOKUP_PARENT); | |
5b6ca027 | 1825 | /* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */ |
e0a01249 AV |
1826 | err = do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, &nd); |
1827 | if (!err) | |
1828 | *path = nd.path; | |
1829 | return err; | |
16f18200 JJS |
1830 | } |
1831 | ||
057f6c01 JM |
1832 | /* |
1833 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1834 | * needs parent already locked. Doesn't follow mounts. | |
1835 | * SMP-safe. | |
1836 | */ | |
eead1911 | 1837 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1838 | { |
4ac91378 | 1839 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1840 | } |
1841 | ||
eead1911 | 1842 | /** |
a6b91919 | 1843 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1844 | * @name: pathname component to lookup |
1845 | * @base: base directory to lookup from | |
1846 | * @len: maximum length @len should be interpreted to | |
1847 | * | |
a6b91919 RD |
1848 | * Note that this routine is purely a helper for filesystem usage and should |
1849 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1850 | * nameidata argument is passed to the filesystem methods and a filesystem |
1851 | * using this helper needs to be prepared for that. | |
1852 | */ | |
057f6c01 JM |
1853 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1854 | { | |
057f6c01 | 1855 | struct qstr this; |
6a96ba54 | 1856 | unsigned int c; |
cda309de | 1857 | int err; |
057f6c01 | 1858 | |
2f9092e1 DW |
1859 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1860 | ||
6a96ba54 AV |
1861 | this.name = name; |
1862 | this.len = len; | |
0145acc2 | 1863 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
1864 | if (!len) |
1865 | return ERR_PTR(-EACCES); | |
1866 | ||
6a96ba54 AV |
1867 | while (len--) { |
1868 | c = *(const unsigned char *)name++; | |
1869 | if (c == '/' || c == '\0') | |
1870 | return ERR_PTR(-EACCES); | |
6a96ba54 | 1871 | } |
5a202bcd AV |
1872 | /* |
1873 | * See if the low-level filesystem might want | |
1874 | * to use its own hash.. | |
1875 | */ | |
1876 | if (base->d_flags & DCACHE_OP_HASH) { | |
1877 | int err = base->d_op->d_hash(base, base->d_inode, &this); | |
1878 | if (err < 0) | |
1879 | return ERR_PTR(err); | |
1880 | } | |
eead1911 | 1881 | |
cda309de MS |
1882 | err = inode_permission(base->d_inode, MAY_EXEC); |
1883 | if (err) | |
1884 | return ERR_PTR(err); | |
1885 | ||
49705b77 | 1886 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1887 | } |
1888 | ||
1fa1e7f6 AW |
1889 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
1890 | struct path *path, int *empty) | |
1da177e4 | 1891 | { |
2d8f3038 | 1892 | struct nameidata nd; |
1fa1e7f6 | 1893 | char *tmp = getname_flags(name, flags, empty); |
1da177e4 | 1894 | int err = PTR_ERR(tmp); |
1da177e4 | 1895 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1896 | |
1897 | BUG_ON(flags & LOOKUP_PARENT); | |
1898 | ||
1899 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1900 | putname(tmp); |
2d8f3038 AV |
1901 | if (!err) |
1902 | *path = nd.path; | |
1da177e4 LT |
1903 | } |
1904 | return err; | |
1905 | } | |
1906 | ||
1fa1e7f6 AW |
1907 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1908 | struct path *path) | |
1909 | { | |
f7493e5d | 1910 | return user_path_at_empty(dfd, name, flags, path, NULL); |
1fa1e7f6 AW |
1911 | } |
1912 | ||
2ad94ae6 AV |
1913 | static int user_path_parent(int dfd, const char __user *path, |
1914 | struct nameidata *nd, char **name) | |
1915 | { | |
1916 | char *s = getname(path); | |
1917 | int error; | |
1918 | ||
1919 | if (IS_ERR(s)) | |
1920 | return PTR_ERR(s); | |
1921 | ||
1922 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1923 | if (error) | |
1924 | putname(s); | |
1925 | else | |
1926 | *name = s; | |
1927 | ||
1928 | return error; | |
1929 | } | |
1930 | ||
1da177e4 LT |
1931 | /* |
1932 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1933 | * minimal. | |
1934 | */ | |
1935 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1936 | { | |
da9592ed DH |
1937 | uid_t fsuid = current_fsuid(); |
1938 | ||
1da177e4 LT |
1939 | if (!(dir->i_mode & S_ISVTX)) |
1940 | return 0; | |
e795b717 SH |
1941 | if (current_user_ns() != inode_userns(inode)) |
1942 | goto other_userns; | |
da9592ed | 1943 | if (inode->i_uid == fsuid) |
1da177e4 | 1944 | return 0; |
da9592ed | 1945 | if (dir->i_uid == fsuid) |
1da177e4 | 1946 | return 0; |
e795b717 SH |
1947 | |
1948 | other_userns: | |
1949 | return !ns_capable(inode_userns(inode), CAP_FOWNER); | |
1da177e4 LT |
1950 | } |
1951 | ||
1952 | /* | |
1953 | * Check whether we can remove a link victim from directory dir, check | |
1954 | * whether the type of victim is right. | |
1955 | * 1. We can't do it if dir is read-only (done in permission()) | |
1956 | * 2. We should have write and exec permissions on dir | |
1957 | * 3. We can't remove anything from append-only dir | |
1958 | * 4. We can't do anything with immutable dir (done in permission()) | |
1959 | * 5. If the sticky bit on dir is set we should either | |
1960 | * a. be owner of dir, or | |
1961 | * b. be owner of victim, or | |
1962 | * c. have CAP_FOWNER capability | |
1963 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1964 | * links pointing to it. | |
1965 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1966 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1967 | * 9. We can't remove a root or mountpoint. | |
1968 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1969 | * nfs_async_unlink(). | |
1970 | */ | |
858119e1 | 1971 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1972 | { |
1973 | int error; | |
1974 | ||
1975 | if (!victim->d_inode) | |
1976 | return -ENOENT; | |
1977 | ||
1978 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 1979 | audit_inode_child(victim, dir); |
1da177e4 | 1980 | |
f419a2e3 | 1981 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1982 | if (error) |
1983 | return error; | |
1984 | if (IS_APPEND(dir)) | |
1985 | return -EPERM; | |
1986 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1987 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1988 | return -EPERM; |
1989 | if (isdir) { | |
1990 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1991 | return -ENOTDIR; | |
1992 | if (IS_ROOT(victim)) | |
1993 | return -EBUSY; | |
1994 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1995 | return -EISDIR; | |
1996 | if (IS_DEADDIR(dir)) | |
1997 | return -ENOENT; | |
1998 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1999 | return -EBUSY; | |
2000 | return 0; | |
2001 | } | |
2002 | ||
2003 | /* Check whether we can create an object with dentry child in directory | |
2004 | * dir. | |
2005 | * 1. We can't do it if child already exists (open has special treatment for | |
2006 | * this case, but since we are inlined it's OK) | |
2007 | * 2. We can't do it if dir is read-only (done in permission()) | |
2008 | * 3. We should have write and exec permissions on dir | |
2009 | * 4. We can't do it if dir is immutable (done in permission()) | |
2010 | */ | |
a95164d9 | 2011 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
2012 | { |
2013 | if (child->d_inode) | |
2014 | return -EEXIST; | |
2015 | if (IS_DEADDIR(dir)) | |
2016 | return -ENOENT; | |
f419a2e3 | 2017 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2018 | } |
2019 | ||
1da177e4 LT |
2020 | /* |
2021 | * p1 and p2 should be directories on the same fs. | |
2022 | */ | |
2023 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2024 | { | |
2025 | struct dentry *p; | |
2026 | ||
2027 | if (p1 == p2) { | |
f2eace23 | 2028 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2029 | return NULL; |
2030 | } | |
2031 | ||
a11f3a05 | 2032 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2033 | |
e2761a11 OH |
2034 | p = d_ancestor(p2, p1); |
2035 | if (p) { | |
2036 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2037 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2038 | return p; | |
1da177e4 LT |
2039 | } |
2040 | ||
e2761a11 OH |
2041 | p = d_ancestor(p1, p2); |
2042 | if (p) { | |
2043 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2044 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2045 | return p; | |
1da177e4 LT |
2046 | } |
2047 | ||
f2eace23 IM |
2048 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
2049 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
2050 | return NULL; |
2051 | } | |
2052 | ||
2053 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2054 | { | |
1b1dcc1b | 2055 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2056 | if (p1 != p2) { |
1b1dcc1b | 2057 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2058 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2059 | } |
2060 | } | |
2061 | ||
4acdaf27 | 2062 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
1da177e4 LT |
2063 | struct nameidata *nd) |
2064 | { | |
a95164d9 | 2065 | int error = may_create(dir, dentry); |
1da177e4 LT |
2066 | |
2067 | if (error) | |
2068 | return error; | |
2069 | ||
acfa4380 | 2070 | if (!dir->i_op->create) |
1da177e4 LT |
2071 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2072 | mode &= S_IALLUGO; | |
2073 | mode |= S_IFREG; | |
2074 | error = security_inode_create(dir, dentry, mode); | |
2075 | if (error) | |
2076 | return error; | |
1da177e4 | 2077 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 2078 | if (!error) |
f38aa942 | 2079 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2080 | return error; |
2081 | } | |
2082 | ||
73d049a4 | 2083 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2084 | { |
3fb64190 | 2085 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2086 | struct inode *inode = dentry->d_inode; |
2087 | int error; | |
2088 | ||
bcda7652 AV |
2089 | /* O_PATH? */ |
2090 | if (!acc_mode) | |
2091 | return 0; | |
2092 | ||
1da177e4 LT |
2093 | if (!inode) |
2094 | return -ENOENT; | |
2095 | ||
c8fe8f30 CH |
2096 | switch (inode->i_mode & S_IFMT) { |
2097 | case S_IFLNK: | |
1da177e4 | 2098 | return -ELOOP; |
c8fe8f30 CH |
2099 | case S_IFDIR: |
2100 | if (acc_mode & MAY_WRITE) | |
2101 | return -EISDIR; | |
2102 | break; | |
2103 | case S_IFBLK: | |
2104 | case S_IFCHR: | |
3fb64190 | 2105 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2106 | return -EACCES; |
c8fe8f30 CH |
2107 | /*FALLTHRU*/ |
2108 | case S_IFIFO: | |
2109 | case S_IFSOCK: | |
1da177e4 | 2110 | flag &= ~O_TRUNC; |
c8fe8f30 | 2111 | break; |
4a3fd211 | 2112 | } |
b41572e9 | 2113 | |
3fb64190 | 2114 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2115 | if (error) |
2116 | return error; | |
6146f0d5 | 2117 | |
1da177e4 LT |
2118 | /* |
2119 | * An append-only file must be opened in append mode for writing. | |
2120 | */ | |
2121 | if (IS_APPEND(inode)) { | |
8737c930 | 2122 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2123 | return -EPERM; |
1da177e4 | 2124 | if (flag & O_TRUNC) |
7715b521 | 2125 | return -EPERM; |
1da177e4 LT |
2126 | } |
2127 | ||
2128 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2129 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2130 | return -EPERM; |
1da177e4 | 2131 | |
f3c7691e | 2132 | return 0; |
7715b521 | 2133 | } |
1da177e4 | 2134 | |
e1181ee6 | 2135 | static int handle_truncate(struct file *filp) |
7715b521 | 2136 | { |
e1181ee6 | 2137 | struct path *path = &filp->f_path; |
7715b521 AV |
2138 | struct inode *inode = path->dentry->d_inode; |
2139 | int error = get_write_access(inode); | |
2140 | if (error) | |
2141 | return error; | |
2142 | /* | |
2143 | * Refuse to truncate files with mandatory locks held on them. | |
2144 | */ | |
2145 | error = locks_verify_locked(inode); | |
2146 | if (!error) | |
ea0d3ab2 | 2147 | error = security_path_truncate(path); |
7715b521 AV |
2148 | if (!error) { |
2149 | error = do_truncate(path->dentry, 0, | |
2150 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2151 | filp); |
7715b521 AV |
2152 | } |
2153 | put_write_access(inode); | |
acd0c935 | 2154 | return error; |
1da177e4 LT |
2155 | } |
2156 | ||
d57999e1 DH |
2157 | static inline int open_to_namei_flags(int flag) |
2158 | { | |
8a5e929d AV |
2159 | if ((flag & O_ACCMODE) == 3) |
2160 | flag--; | |
d57999e1 DH |
2161 | return flag; |
2162 | } | |
2163 | ||
31e6b01f | 2164 | /* |
fe2d35ff | 2165 | * Handle the last step of open() |
31e6b01f | 2166 | */ |
fb1cc555 | 2167 | static struct file *do_last(struct nameidata *nd, struct path *path, |
c3e380b0 | 2168 | const struct open_flags *op, const char *pathname) |
fb1cc555 | 2169 | { |
a1e28038 | 2170 | struct dentry *dir = nd->path.dentry; |
6c0d46c4 | 2171 | struct dentry *dentry; |
ca344a89 | 2172 | int open_flag = op->open_flag; |
6c0d46c4 | 2173 | int will_truncate = open_flag & O_TRUNC; |
ca344a89 | 2174 | int want_write = 0; |
bcda7652 | 2175 | int acc_mode = op->acc_mode; |
fb1cc555 | 2176 | struct file *filp; |
16c2cd71 | 2177 | int error; |
1f36f774 | 2178 | |
c3e380b0 AV |
2179 | nd->flags &= ~LOOKUP_PARENT; |
2180 | nd->flags |= op->intent; | |
2181 | ||
1f36f774 AV |
2182 | switch (nd->last_type) { |
2183 | case LAST_DOTDOT: | |
176306f5 | 2184 | case LAST_DOT: |
fe2d35ff AV |
2185 | error = handle_dots(nd, nd->last_type); |
2186 | if (error) | |
2187 | return ERR_PTR(error); | |
1f36f774 | 2188 | /* fallthrough */ |
1f36f774 | 2189 | case LAST_ROOT: |
9f1fafee | 2190 | error = complete_walk(nd); |
16c2cd71 | 2191 | if (error) |
9f1fafee | 2192 | return ERR_PTR(error); |
fe2d35ff | 2193 | audit_inode(pathname, nd->path.dentry); |
ca344a89 | 2194 | if (open_flag & O_CREAT) { |
fe2d35ff AV |
2195 | error = -EISDIR; |
2196 | goto exit; | |
2197 | } | |
2198 | goto ok; | |
1f36f774 | 2199 | case LAST_BIND: |
9f1fafee | 2200 | error = complete_walk(nd); |
16c2cd71 | 2201 | if (error) |
9f1fafee | 2202 | return ERR_PTR(error); |
1f36f774 | 2203 | audit_inode(pathname, dir); |
67ee3ad2 | 2204 | goto ok; |
1f36f774 | 2205 | } |
67ee3ad2 | 2206 | |
ca344a89 | 2207 | if (!(open_flag & O_CREAT)) { |
bcda7652 | 2208 | int symlink_ok = 0; |
fe2d35ff AV |
2209 | if (nd->last.name[nd->last.len]) |
2210 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
bcda7652 AV |
2211 | if (open_flag & O_PATH && !(nd->flags & LOOKUP_FOLLOW)) |
2212 | symlink_ok = 1; | |
fe2d35ff | 2213 | /* we _can_ be in RCU mode here */ |
ce57dfc1 AV |
2214 | error = walk_component(nd, path, &nd->last, LAST_NORM, |
2215 | !symlink_ok); | |
2216 | if (error < 0) | |
fe2d35ff | 2217 | return ERR_PTR(error); |
ce57dfc1 | 2218 | if (error) /* symlink */ |
fe2d35ff | 2219 | return NULL; |
fe2d35ff | 2220 | /* sayonara */ |
9f1fafee AV |
2221 | error = complete_walk(nd); |
2222 | if (error) | |
7f6c7e62 | 2223 | return ERR_PTR(error); |
fe2d35ff AV |
2224 | |
2225 | error = -ENOTDIR; | |
2226 | if (nd->flags & LOOKUP_DIRECTORY) { | |
ce57dfc1 | 2227 | if (!nd->inode->i_op->lookup) |
fe2d35ff AV |
2228 | goto exit; |
2229 | } | |
2230 | audit_inode(pathname, nd->path.dentry); | |
2231 | goto ok; | |
2232 | } | |
2233 | ||
2234 | /* create side of things */ | |
a3fbbde7 AV |
2235 | /* |
2236 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED has been | |
2237 | * cleared when we got to the last component we are about to look up | |
2238 | */ | |
9f1fafee AV |
2239 | error = complete_walk(nd); |
2240 | if (error) | |
2241 | return ERR_PTR(error); | |
fe2d35ff AV |
2242 | |
2243 | audit_inode(pathname, dir); | |
16c2cd71 | 2244 | error = -EISDIR; |
1f36f774 | 2245 | /* trailing slashes? */ |
31e6b01f NP |
2246 | if (nd->last.name[nd->last.len]) |
2247 | goto exit; | |
a2c36b45 | 2248 | |
a1e28038 AV |
2249 | mutex_lock(&dir->d_inode->i_mutex); |
2250 | ||
6c0d46c4 AV |
2251 | dentry = lookup_hash(nd); |
2252 | error = PTR_ERR(dentry); | |
2253 | if (IS_ERR(dentry)) { | |
fb1cc555 AV |
2254 | mutex_unlock(&dir->d_inode->i_mutex); |
2255 | goto exit; | |
2256 | } | |
2257 | ||
6c0d46c4 AV |
2258 | path->dentry = dentry; |
2259 | path->mnt = nd->path.mnt; | |
2260 | ||
fb1cc555 | 2261 | /* Negative dentry, just create the file */ |
6c0d46c4 | 2262 | if (!dentry->d_inode) { |
a218d0fd | 2263 | umode_t mode = op->mode; |
6c0d46c4 AV |
2264 | if (!IS_POSIXACL(dir->d_inode)) |
2265 | mode &= ~current_umask(); | |
fb1cc555 AV |
2266 | /* |
2267 | * This write is needed to ensure that a | |
6c0d46c4 | 2268 | * rw->ro transition does not occur between |
fb1cc555 AV |
2269 | * the time when the file is created and when |
2270 | * a permanent write count is taken through | |
2271 | * the 'struct file' in nameidata_to_filp(). | |
2272 | */ | |
2273 | error = mnt_want_write(nd->path.mnt); | |
2274 | if (error) | |
2275 | goto exit_mutex_unlock; | |
ca344a89 | 2276 | want_write = 1; |
9b44f1b3 | 2277 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 2278 | open_flag &= ~O_TRUNC; |
6c0d46c4 | 2279 | will_truncate = 0; |
bcda7652 | 2280 | acc_mode = MAY_OPEN; |
6c0d46c4 AV |
2281 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2282 | if (error) | |
2283 | goto exit_mutex_unlock; | |
2284 | error = vfs_create(dir->d_inode, dentry, mode, nd); | |
2285 | if (error) | |
2286 | goto exit_mutex_unlock; | |
2287 | mutex_unlock(&dir->d_inode->i_mutex); | |
2288 | dput(nd->path.dentry); | |
2289 | nd->path.dentry = dentry; | |
ca344a89 | 2290 | goto common; |
fb1cc555 AV |
2291 | } |
2292 | ||
2293 | /* | |
2294 | * It already exists. | |
2295 | */ | |
2296 | mutex_unlock(&dir->d_inode->i_mutex); | |
2297 | audit_inode(pathname, path->dentry); | |
2298 | ||
2299 | error = -EEXIST; | |
ca344a89 | 2300 | if (open_flag & O_EXCL) |
fb1cc555 AV |
2301 | goto exit_dput; |
2302 | ||
9875cf80 DH |
2303 | error = follow_managed(path, nd->flags); |
2304 | if (error < 0) | |
2305 | goto exit_dput; | |
fb1cc555 | 2306 | |
a3fbbde7 AV |
2307 | if (error) |
2308 | nd->flags |= LOOKUP_JUMPED; | |
2309 | ||
fb1cc555 AV |
2310 | error = -ENOENT; |
2311 | if (!path->dentry->d_inode) | |
2312 | goto exit_dput; | |
9e67f361 AV |
2313 | |
2314 | if (path->dentry->d_inode->i_op->follow_link) | |
fb1cc555 | 2315 | return NULL; |
fb1cc555 AV |
2316 | |
2317 | path_to_nameidata(path, nd); | |
31e6b01f | 2318 | nd->inode = path->dentry->d_inode; |
a3fbbde7 AV |
2319 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
2320 | error = complete_walk(nd); | |
2321 | if (error) | |
097b180c | 2322 | return ERR_PTR(error); |
fb1cc555 | 2323 | error = -EISDIR; |
31e6b01f | 2324 | if (S_ISDIR(nd->inode->i_mode)) |
fb1cc555 | 2325 | goto exit; |
67ee3ad2 | 2326 | ok: |
6c0d46c4 AV |
2327 | if (!S_ISREG(nd->inode->i_mode)) |
2328 | will_truncate = 0; | |
2329 | ||
0f9d1a10 AV |
2330 | if (will_truncate) { |
2331 | error = mnt_want_write(nd->path.mnt); | |
2332 | if (error) | |
2333 | goto exit; | |
ca344a89 | 2334 | want_write = 1; |
0f9d1a10 | 2335 | } |
ca344a89 | 2336 | common: |
bcda7652 | 2337 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 2338 | if (error) |
0f9d1a10 | 2339 | goto exit; |
0f9d1a10 AV |
2340 | filp = nameidata_to_filp(nd); |
2341 | if (!IS_ERR(filp)) { | |
2342 | error = ima_file_check(filp, op->acc_mode); | |
2343 | if (error) { | |
2344 | fput(filp); | |
2345 | filp = ERR_PTR(error); | |
2346 | } | |
2347 | } | |
2348 | if (!IS_ERR(filp)) { | |
2349 | if (will_truncate) { | |
2350 | error = handle_truncate(filp); | |
2351 | if (error) { | |
2352 | fput(filp); | |
2353 | filp = ERR_PTR(error); | |
2354 | } | |
2355 | } | |
2356 | } | |
ca344a89 AV |
2357 | out: |
2358 | if (want_write) | |
0f9d1a10 AV |
2359 | mnt_drop_write(nd->path.mnt); |
2360 | path_put(&nd->path); | |
fb1cc555 AV |
2361 | return filp; |
2362 | ||
2363 | exit_mutex_unlock: | |
2364 | mutex_unlock(&dir->d_inode->i_mutex); | |
2365 | exit_dput: | |
2366 | path_put_conditional(path, nd); | |
2367 | exit: | |
ca344a89 AV |
2368 | filp = ERR_PTR(error); |
2369 | goto out; | |
fb1cc555 AV |
2370 | } |
2371 | ||
13aab428 | 2372 | static struct file *path_openat(int dfd, const char *pathname, |
73d049a4 | 2373 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 2374 | { |
fe2d35ff | 2375 | struct file *base = NULL; |
4a3fd211 | 2376 | struct file *filp; |
9850c056 | 2377 | struct path path; |
13aab428 | 2378 | int error; |
31e6b01f NP |
2379 | |
2380 | filp = get_empty_filp(); | |
2381 | if (!filp) | |
2382 | return ERR_PTR(-ENFILE); | |
2383 | ||
47c805dc | 2384 | filp->f_flags = op->open_flag; |
73d049a4 AV |
2385 | nd->intent.open.file = filp; |
2386 | nd->intent.open.flags = open_to_namei_flags(op->open_flag); | |
2387 | nd->intent.open.create_mode = op->mode; | |
31e6b01f | 2388 | |
73d049a4 | 2389 | error = path_init(dfd, pathname, flags | LOOKUP_PARENT, nd, &base); |
31e6b01f | 2390 | if (unlikely(error)) |
13aab428 | 2391 | goto out_filp; |
31e6b01f | 2392 | |
fe2d35ff | 2393 | current->total_link_count = 0; |
73d049a4 | 2394 | error = link_path_walk(pathname, nd); |
31e6b01f NP |
2395 | if (unlikely(error)) |
2396 | goto out_filp; | |
1da177e4 | 2397 | |
73d049a4 | 2398 | filp = do_last(nd, &path, op, pathname); |
806b681c | 2399 | while (unlikely(!filp)) { /* trailing symlink */ |
7b9337aa | 2400 | struct path link = path; |
def4af30 | 2401 | void *cookie; |
574197e0 | 2402 | if (!(nd->flags & LOOKUP_FOLLOW)) { |
73d049a4 AV |
2403 | path_put_conditional(&path, nd); |
2404 | path_put(&nd->path); | |
40b39136 AV |
2405 | filp = ERR_PTR(-ELOOP); |
2406 | break; | |
2407 | } | |
73d049a4 AV |
2408 | nd->flags |= LOOKUP_PARENT; |
2409 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
574197e0 | 2410 | error = follow_link(&link, nd, &cookie); |
c3e380b0 | 2411 | if (unlikely(error)) |
f1afe9ef | 2412 | filp = ERR_PTR(error); |
c3e380b0 | 2413 | else |
73d049a4 | 2414 | filp = do_last(nd, &path, op, pathname); |
574197e0 | 2415 | put_link(nd, &link, cookie); |
806b681c | 2416 | } |
10fa8e62 | 2417 | out: |
73d049a4 AV |
2418 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) |
2419 | path_put(&nd->root); | |
fe2d35ff AV |
2420 | if (base) |
2421 | fput(base); | |
73d049a4 | 2422 | release_open_intent(nd); |
10fa8e62 | 2423 | return filp; |
1da177e4 | 2424 | |
31e6b01f | 2425 | out_filp: |
806b681c | 2426 | filp = ERR_PTR(error); |
10fa8e62 | 2427 | goto out; |
1da177e4 LT |
2428 | } |
2429 | ||
13aab428 AV |
2430 | struct file *do_filp_open(int dfd, const char *pathname, |
2431 | const struct open_flags *op, int flags) | |
2432 | { | |
73d049a4 | 2433 | struct nameidata nd; |
13aab428 AV |
2434 | struct file *filp; |
2435 | ||
73d049a4 | 2436 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 2437 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 2438 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 2439 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 2440 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
2441 | return filp; |
2442 | } | |
2443 | ||
73d049a4 AV |
2444 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
2445 | const char *name, const struct open_flags *op, int flags) | |
2446 | { | |
2447 | struct nameidata nd; | |
2448 | struct file *file; | |
2449 | ||
2450 | nd.root.mnt = mnt; | |
2451 | nd.root.dentry = dentry; | |
2452 | ||
2453 | flags |= LOOKUP_ROOT; | |
2454 | ||
bcda7652 | 2455 | if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
2456 | return ERR_PTR(-ELOOP); |
2457 | ||
2458 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU); | |
2459 | if (unlikely(file == ERR_PTR(-ECHILD))) | |
2460 | file = path_openat(-1, name, &nd, op, flags); | |
2461 | if (unlikely(file == ERR_PTR(-ESTALE))) | |
2462 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL); | |
2463 | return file; | |
2464 | } | |
2465 | ||
ed75e95d | 2466 | struct dentry *kern_path_create(int dfd, const char *pathname, struct path *path, int is_dir) |
1da177e4 | 2467 | { |
c663e5d8 | 2468 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d AV |
2469 | struct nameidata nd; |
2470 | int error = do_path_lookup(dfd, pathname, LOOKUP_PARENT, &nd); | |
2471 | if (error) | |
2472 | return ERR_PTR(error); | |
1da177e4 | 2473 | |
c663e5d8 CH |
2474 | /* |
2475 | * Yucky last component or no last component at all? | |
2476 | * (foo/., foo/.., /////) | |
2477 | */ | |
ed75e95d AV |
2478 | if (nd.last_type != LAST_NORM) |
2479 | goto out; | |
2480 | nd.flags &= ~LOOKUP_PARENT; | |
2481 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
2482 | nd.intent.open.flags = O_EXCL; | |
c663e5d8 CH |
2483 | |
2484 | /* | |
2485 | * Do the final lookup. | |
2486 | */ | |
ed75e95d AV |
2487 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
2488 | dentry = lookup_hash(&nd); | |
1da177e4 LT |
2489 | if (IS_ERR(dentry)) |
2490 | goto fail; | |
c663e5d8 | 2491 | |
e9baf6e5 AV |
2492 | if (dentry->d_inode) |
2493 | goto eexist; | |
c663e5d8 CH |
2494 | /* |
2495 | * Special case - lookup gave negative, but... we had foo/bar/ | |
2496 | * From the vfs_mknod() POV we just have a negative dentry - | |
2497 | * all is fine. Let's be bastards - you had / on the end, you've | |
2498 | * been asking for (non-existent) directory. -ENOENT for you. | |
2499 | */ | |
ed75e95d | 2500 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
e9baf6e5 AV |
2501 | dput(dentry); |
2502 | dentry = ERR_PTR(-ENOENT); | |
ed75e95d | 2503 | goto fail; |
e9baf6e5 | 2504 | } |
ed75e95d | 2505 | *path = nd.path; |
1da177e4 | 2506 | return dentry; |
e9baf6e5 | 2507 | eexist: |
1da177e4 | 2508 | dput(dentry); |
e9baf6e5 | 2509 | dentry = ERR_PTR(-EEXIST); |
1da177e4 | 2510 | fail: |
ed75e95d AV |
2511 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
2512 | out: | |
2513 | path_put(&nd.path); | |
1da177e4 LT |
2514 | return dentry; |
2515 | } | |
dae6ad8f AV |
2516 | EXPORT_SYMBOL(kern_path_create); |
2517 | ||
2518 | struct dentry *user_path_create(int dfd, const char __user *pathname, struct path *path, int is_dir) | |
2519 | { | |
2520 | char *tmp = getname(pathname); | |
2521 | struct dentry *res; | |
2522 | if (IS_ERR(tmp)) | |
2523 | return ERR_CAST(tmp); | |
2524 | res = kern_path_create(dfd, tmp, path, is_dir); | |
2525 | putname(tmp); | |
2526 | return res; | |
2527 | } | |
2528 | EXPORT_SYMBOL(user_path_create); | |
2529 | ||
1a67aafb | 2530 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 2531 | { |
a95164d9 | 2532 | int error = may_create(dir, dentry); |
1da177e4 LT |
2533 | |
2534 | if (error) | |
2535 | return error; | |
2536 | ||
e795b717 SH |
2537 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && |
2538 | !ns_capable(inode_userns(dir), CAP_MKNOD)) | |
1da177e4 LT |
2539 | return -EPERM; |
2540 | ||
acfa4380 | 2541 | if (!dir->i_op->mknod) |
1da177e4 LT |
2542 | return -EPERM; |
2543 | ||
08ce5f16 SH |
2544 | error = devcgroup_inode_mknod(mode, dev); |
2545 | if (error) | |
2546 | return error; | |
2547 | ||
1da177e4 LT |
2548 | error = security_inode_mknod(dir, dentry, mode, dev); |
2549 | if (error) | |
2550 | return error; | |
2551 | ||
1da177e4 | 2552 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 2553 | if (!error) |
f38aa942 | 2554 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2555 | return error; |
2556 | } | |
2557 | ||
f69aac00 | 2558 | static int may_mknod(umode_t mode) |
463c3197 DH |
2559 | { |
2560 | switch (mode & S_IFMT) { | |
2561 | case S_IFREG: | |
2562 | case S_IFCHR: | |
2563 | case S_IFBLK: | |
2564 | case S_IFIFO: | |
2565 | case S_IFSOCK: | |
2566 | case 0: /* zero mode translates to S_IFREG */ | |
2567 | return 0; | |
2568 | case S_IFDIR: | |
2569 | return -EPERM; | |
2570 | default: | |
2571 | return -EINVAL; | |
2572 | } | |
2573 | } | |
2574 | ||
8208a22b | 2575 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 2576 | unsigned, dev) |
1da177e4 | 2577 | { |
2ad94ae6 | 2578 | struct dentry *dentry; |
dae6ad8f AV |
2579 | struct path path; |
2580 | int error; | |
1da177e4 LT |
2581 | |
2582 | if (S_ISDIR(mode)) | |
2583 | return -EPERM; | |
1da177e4 | 2584 | |
dae6ad8f AV |
2585 | dentry = user_path_create(dfd, filename, &path, 0); |
2586 | if (IS_ERR(dentry)) | |
2587 | return PTR_ERR(dentry); | |
2ad94ae6 | 2588 | |
dae6ad8f | 2589 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 2590 | mode &= ~current_umask(); |
463c3197 DH |
2591 | error = may_mknod(mode); |
2592 | if (error) | |
2593 | goto out_dput; | |
dae6ad8f | 2594 | error = mnt_want_write(path.mnt); |
463c3197 DH |
2595 | if (error) |
2596 | goto out_dput; | |
dae6ad8f | 2597 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 KT |
2598 | if (error) |
2599 | goto out_drop_write; | |
463c3197 | 2600 | switch (mode & S_IFMT) { |
1da177e4 | 2601 | case 0: case S_IFREG: |
dae6ad8f | 2602 | error = vfs_create(path.dentry->d_inode,dentry,mode,NULL); |
1da177e4 LT |
2603 | break; |
2604 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 2605 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2606 | new_decode_dev(dev)); |
2607 | break; | |
2608 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 2609 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2610 | break; |
1da177e4 | 2611 | } |
be6d3e56 | 2612 | out_drop_write: |
dae6ad8f | 2613 | mnt_drop_write(path.mnt); |
463c3197 DH |
2614 | out_dput: |
2615 | dput(dentry); | |
dae6ad8f AV |
2616 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
2617 | path_put(&path); | |
1da177e4 LT |
2618 | |
2619 | return error; | |
2620 | } | |
2621 | ||
8208a22b | 2622 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
2623 | { |
2624 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2625 | } | |
2626 | ||
18bb1db3 | 2627 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 2628 | { |
a95164d9 | 2629 | int error = may_create(dir, dentry); |
8de52778 | 2630 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
2631 | |
2632 | if (error) | |
2633 | return error; | |
2634 | ||
acfa4380 | 2635 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2636 | return -EPERM; |
2637 | ||
2638 | mode &= (S_IRWXUGO|S_ISVTX); | |
2639 | error = security_inode_mkdir(dir, dentry, mode); | |
2640 | if (error) | |
2641 | return error; | |
2642 | ||
8de52778 AV |
2643 | if (max_links && dir->i_nlink >= max_links) |
2644 | return -EMLINK; | |
2645 | ||
1da177e4 | 2646 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2647 | if (!error) |
f38aa942 | 2648 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2649 | return error; |
2650 | } | |
2651 | ||
a218d0fd | 2652 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 2653 | { |
6902d925 | 2654 | struct dentry *dentry; |
dae6ad8f AV |
2655 | struct path path; |
2656 | int error; | |
1da177e4 | 2657 | |
dae6ad8f | 2658 | dentry = user_path_create(dfd, pathname, &path, 1); |
6902d925 | 2659 | if (IS_ERR(dentry)) |
dae6ad8f | 2660 | return PTR_ERR(dentry); |
1da177e4 | 2661 | |
dae6ad8f | 2662 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 2663 | mode &= ~current_umask(); |
dae6ad8f | 2664 | error = mnt_want_write(path.mnt); |
463c3197 DH |
2665 | if (error) |
2666 | goto out_dput; | |
dae6ad8f | 2667 | error = security_path_mkdir(&path, dentry, mode); |
be6d3e56 KT |
2668 | if (error) |
2669 | goto out_drop_write; | |
dae6ad8f | 2670 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2671 | out_drop_write: |
dae6ad8f | 2672 | mnt_drop_write(path.mnt); |
463c3197 | 2673 | out_dput: |
6902d925 | 2674 | dput(dentry); |
dae6ad8f AV |
2675 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
2676 | path_put(&path); | |
1da177e4 LT |
2677 | return error; |
2678 | } | |
2679 | ||
a218d0fd | 2680 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
2681 | { |
2682 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2683 | } | |
2684 | ||
1da177e4 | 2685 | /* |
a71905f0 | 2686 | * The dentry_unhash() helper will try to drop the dentry early: we |
c0d02594 | 2687 | * should have a usage count of 1 if we're the only user of this |
a71905f0 SW |
2688 | * dentry, and if that is true (possibly after pruning the dcache), |
2689 | * then we drop the dentry now. | |
1da177e4 LT |
2690 | * |
2691 | * A low-level filesystem can, if it choses, legally | |
2692 | * do a | |
2693 | * | |
2694 | * if (!d_unhashed(dentry)) | |
2695 | * return -EBUSY; | |
2696 | * | |
2697 | * if it cannot handle the case of removing a directory | |
2698 | * that is still in use by something else.. | |
2699 | */ | |
2700 | void dentry_unhash(struct dentry *dentry) | |
2701 | { | |
dc168427 | 2702 | shrink_dcache_parent(dentry); |
1da177e4 | 2703 | spin_lock(&dentry->d_lock); |
64252c75 | 2704 | if (dentry->d_count == 1) |
1da177e4 LT |
2705 | __d_drop(dentry); |
2706 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
2707 | } |
2708 | ||
2709 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2710 | { | |
2711 | int error = may_delete(dir, dentry, 1); | |
2712 | ||
2713 | if (error) | |
2714 | return error; | |
2715 | ||
acfa4380 | 2716 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2717 | return -EPERM; |
2718 | ||
1d2ef590 | 2719 | dget(dentry); |
1b1dcc1b | 2720 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
2721 | |
2722 | error = -EBUSY; | |
1da177e4 | 2723 | if (d_mountpoint(dentry)) |
912dbc15 SW |
2724 | goto out; |
2725 | ||
2726 | error = security_inode_rmdir(dir, dentry); | |
2727 | if (error) | |
2728 | goto out; | |
2729 | ||
3cebde24 | 2730 | shrink_dcache_parent(dentry); |
912dbc15 SW |
2731 | error = dir->i_op->rmdir(dir, dentry); |
2732 | if (error) | |
2733 | goto out; | |
2734 | ||
2735 | dentry->d_inode->i_flags |= S_DEAD; | |
2736 | dont_mount(dentry); | |
2737 | ||
2738 | out: | |
1b1dcc1b | 2739 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 2740 | dput(dentry); |
912dbc15 | 2741 | if (!error) |
1da177e4 | 2742 | d_delete(dentry); |
1da177e4 LT |
2743 | return error; |
2744 | } | |
2745 | ||
5590ff0d | 2746 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2747 | { |
2748 | int error = 0; | |
2749 | char * name; | |
2750 | struct dentry *dentry; | |
2751 | struct nameidata nd; | |
2752 | ||
2ad94ae6 | 2753 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2754 | if (error) |
2ad94ae6 | 2755 | return error; |
1da177e4 LT |
2756 | |
2757 | switch(nd.last_type) { | |
0612d9fb OH |
2758 | case LAST_DOTDOT: |
2759 | error = -ENOTEMPTY; | |
2760 | goto exit1; | |
2761 | case LAST_DOT: | |
2762 | error = -EINVAL; | |
2763 | goto exit1; | |
2764 | case LAST_ROOT: | |
2765 | error = -EBUSY; | |
2766 | goto exit1; | |
1da177e4 | 2767 | } |
0612d9fb OH |
2768 | |
2769 | nd.flags &= ~LOOKUP_PARENT; | |
2770 | ||
4ac91378 | 2771 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2772 | dentry = lookup_hash(&nd); |
1da177e4 | 2773 | error = PTR_ERR(dentry); |
6902d925 DH |
2774 | if (IS_ERR(dentry)) |
2775 | goto exit2; | |
e6bc45d6 TT |
2776 | if (!dentry->d_inode) { |
2777 | error = -ENOENT; | |
2778 | goto exit3; | |
2779 | } | |
0622753b DH |
2780 | error = mnt_want_write(nd.path.mnt); |
2781 | if (error) | |
2782 | goto exit3; | |
be6d3e56 KT |
2783 | error = security_path_rmdir(&nd.path, dentry); |
2784 | if (error) | |
2785 | goto exit4; | |
4ac91378 | 2786 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2787 | exit4: |
0622753b DH |
2788 | mnt_drop_write(nd.path.mnt); |
2789 | exit3: | |
6902d925 DH |
2790 | dput(dentry); |
2791 | exit2: | |
4ac91378 | 2792 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2793 | exit1: |
1d957f9b | 2794 | path_put(&nd.path); |
1da177e4 LT |
2795 | putname(name); |
2796 | return error; | |
2797 | } | |
2798 | ||
3cdad428 | 2799 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2800 | { |
2801 | return do_rmdir(AT_FDCWD, pathname); | |
2802 | } | |
2803 | ||
1da177e4 LT |
2804 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2805 | { | |
2806 | int error = may_delete(dir, dentry, 0); | |
2807 | ||
2808 | if (error) | |
2809 | return error; | |
2810 | ||
acfa4380 | 2811 | if (!dir->i_op->unlink) |
1da177e4 LT |
2812 | return -EPERM; |
2813 | ||
1b1dcc1b | 2814 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2815 | if (d_mountpoint(dentry)) |
2816 | error = -EBUSY; | |
2817 | else { | |
2818 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 2819 | if (!error) { |
1da177e4 | 2820 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 2821 | if (!error) |
d83c49f3 | 2822 | dont_mount(dentry); |
bec1052e | 2823 | } |
1da177e4 | 2824 | } |
1b1dcc1b | 2825 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2826 | |
2827 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2828 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2829 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2830 | d_delete(dentry); |
1da177e4 | 2831 | } |
0eeca283 | 2832 | |
1da177e4 LT |
2833 | return error; |
2834 | } | |
2835 | ||
2836 | /* | |
2837 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2838 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2839 | * writeout happening, and we don't want to prevent access to the directory |
2840 | * while waiting on the I/O. | |
2841 | */ | |
5590ff0d | 2842 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2843 | { |
2ad94ae6 AV |
2844 | int error; |
2845 | char *name; | |
1da177e4 LT |
2846 | struct dentry *dentry; |
2847 | struct nameidata nd; | |
2848 | struct inode *inode = NULL; | |
2849 | ||
2ad94ae6 | 2850 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2851 | if (error) |
2ad94ae6 AV |
2852 | return error; |
2853 | ||
1da177e4 LT |
2854 | error = -EISDIR; |
2855 | if (nd.last_type != LAST_NORM) | |
2856 | goto exit1; | |
0612d9fb OH |
2857 | |
2858 | nd.flags &= ~LOOKUP_PARENT; | |
2859 | ||
4ac91378 | 2860 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2861 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2862 | error = PTR_ERR(dentry); |
2863 | if (!IS_ERR(dentry)) { | |
2864 | /* Why not before? Because we want correct error value */ | |
50338b88 TE |
2865 | if (nd.last.name[nd.last.len]) |
2866 | goto slashes; | |
1da177e4 | 2867 | inode = dentry->d_inode; |
50338b88 | 2868 | if (!inode) |
e6bc45d6 TT |
2869 | goto slashes; |
2870 | ihold(inode); | |
0622753b DH |
2871 | error = mnt_want_write(nd.path.mnt); |
2872 | if (error) | |
2873 | goto exit2; | |
be6d3e56 KT |
2874 | error = security_path_unlink(&nd.path, dentry); |
2875 | if (error) | |
2876 | goto exit3; | |
4ac91378 | 2877 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2878 | exit3: |
0622753b | 2879 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2880 | exit2: |
2881 | dput(dentry); | |
2882 | } | |
4ac91378 | 2883 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2884 | if (inode) |
2885 | iput(inode); /* truncate the inode here */ | |
2886 | exit1: | |
1d957f9b | 2887 | path_put(&nd.path); |
1da177e4 LT |
2888 | putname(name); |
2889 | return error; | |
2890 | ||
2891 | slashes: | |
2892 | error = !dentry->d_inode ? -ENOENT : | |
2893 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2894 | goto exit2; | |
2895 | } | |
2896 | ||
2e4d0924 | 2897 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2898 | { |
2899 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2900 | return -EINVAL; | |
2901 | ||
2902 | if (flag & AT_REMOVEDIR) | |
2903 | return do_rmdir(dfd, pathname); | |
2904 | ||
2905 | return do_unlinkat(dfd, pathname); | |
2906 | } | |
2907 | ||
3480b257 | 2908 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2909 | { |
2910 | return do_unlinkat(AT_FDCWD, pathname); | |
2911 | } | |
2912 | ||
db2e747b | 2913 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2914 | { |
a95164d9 | 2915 | int error = may_create(dir, dentry); |
1da177e4 LT |
2916 | |
2917 | if (error) | |
2918 | return error; | |
2919 | ||
acfa4380 | 2920 | if (!dir->i_op->symlink) |
1da177e4 LT |
2921 | return -EPERM; |
2922 | ||
2923 | error = security_inode_symlink(dir, dentry, oldname); | |
2924 | if (error) | |
2925 | return error; | |
2926 | ||
1da177e4 | 2927 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2928 | if (!error) |
f38aa942 | 2929 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2930 | return error; |
2931 | } | |
2932 | ||
2e4d0924 HC |
2933 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2934 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2935 | { |
2ad94ae6 AV |
2936 | int error; |
2937 | char *from; | |
6902d925 | 2938 | struct dentry *dentry; |
dae6ad8f | 2939 | struct path path; |
1da177e4 LT |
2940 | |
2941 | from = getname(oldname); | |
2ad94ae6 | 2942 | if (IS_ERR(from)) |
1da177e4 | 2943 | return PTR_ERR(from); |
1da177e4 | 2944 | |
dae6ad8f | 2945 | dentry = user_path_create(newdfd, newname, &path, 0); |
6902d925 DH |
2946 | error = PTR_ERR(dentry); |
2947 | if (IS_ERR(dentry)) | |
dae6ad8f | 2948 | goto out_putname; |
6902d925 | 2949 | |
dae6ad8f | 2950 | error = mnt_want_write(path.mnt); |
75c3f29d DH |
2951 | if (error) |
2952 | goto out_dput; | |
dae6ad8f | 2953 | error = security_path_symlink(&path, dentry, from); |
be6d3e56 KT |
2954 | if (error) |
2955 | goto out_drop_write; | |
dae6ad8f | 2956 | error = vfs_symlink(path.dentry->d_inode, dentry, from); |
be6d3e56 | 2957 | out_drop_write: |
dae6ad8f | 2958 | mnt_drop_write(path.mnt); |
75c3f29d | 2959 | out_dput: |
6902d925 | 2960 | dput(dentry); |
dae6ad8f AV |
2961 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
2962 | path_put(&path); | |
6902d925 | 2963 | out_putname: |
1da177e4 LT |
2964 | putname(from); |
2965 | return error; | |
2966 | } | |
2967 | ||
3480b257 | 2968 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2969 | { |
2970 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2971 | } | |
2972 | ||
1da177e4 LT |
2973 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2974 | { | |
2975 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 2976 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
2977 | int error; |
2978 | ||
2979 | if (!inode) | |
2980 | return -ENOENT; | |
2981 | ||
a95164d9 | 2982 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2983 | if (error) |
2984 | return error; | |
2985 | ||
2986 | if (dir->i_sb != inode->i_sb) | |
2987 | return -EXDEV; | |
2988 | ||
2989 | /* | |
2990 | * A link to an append-only or immutable file cannot be created. | |
2991 | */ | |
2992 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2993 | return -EPERM; | |
acfa4380 | 2994 | if (!dir->i_op->link) |
1da177e4 | 2995 | return -EPERM; |
7e79eedb | 2996 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2997 | return -EPERM; |
2998 | ||
2999 | error = security_inode_link(old_dentry, dir, new_dentry); | |
3000 | if (error) | |
3001 | return error; | |
3002 | ||
7e79eedb | 3003 | mutex_lock(&inode->i_mutex); |
aae8a97d AK |
3004 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
3005 | if (inode->i_nlink == 0) | |
3006 | error = -ENOENT; | |
8de52778 AV |
3007 | else if (max_links && inode->i_nlink >= max_links) |
3008 | error = -EMLINK; | |
aae8a97d AK |
3009 | else |
3010 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 3011 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 3012 | if (!error) |
7e79eedb | 3013 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
3014 | return error; |
3015 | } | |
3016 | ||
3017 | /* | |
3018 | * Hardlinks are often used in delicate situations. We avoid | |
3019 | * security-related surprises by not following symlinks on the | |
3020 | * newname. --KAB | |
3021 | * | |
3022 | * We don't follow them on the oldname either to be compatible | |
3023 | * with linux 2.0, and to avoid hard-linking to directories | |
3024 | * and other special files. --ADM | |
3025 | */ | |
2e4d0924 HC |
3026 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
3027 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
3028 | { |
3029 | struct dentry *new_dentry; | |
dae6ad8f | 3030 | struct path old_path, new_path; |
11a7b371 | 3031 | int how = 0; |
1da177e4 | 3032 | int error; |
1da177e4 | 3033 | |
11a7b371 | 3034 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 3035 | return -EINVAL; |
11a7b371 AK |
3036 | /* |
3037 | * To use null names we require CAP_DAC_READ_SEARCH | |
3038 | * This ensures that not everyone will be able to create | |
3039 | * handlink using the passed filedescriptor. | |
3040 | */ | |
3041 | if (flags & AT_EMPTY_PATH) { | |
3042 | if (!capable(CAP_DAC_READ_SEARCH)) | |
3043 | return -ENOENT; | |
3044 | how = LOOKUP_EMPTY; | |
3045 | } | |
3046 | ||
3047 | if (flags & AT_SYMLINK_FOLLOW) | |
3048 | how |= LOOKUP_FOLLOW; | |
c04030e1 | 3049 | |
11a7b371 | 3050 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 3051 | if (error) |
2ad94ae6 AV |
3052 | return error; |
3053 | ||
dae6ad8f | 3054 | new_dentry = user_path_create(newdfd, newname, &new_path, 0); |
1da177e4 | 3055 | error = PTR_ERR(new_dentry); |
6902d925 | 3056 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
3057 | goto out; |
3058 | ||
3059 | error = -EXDEV; | |
3060 | if (old_path.mnt != new_path.mnt) | |
3061 | goto out_dput; | |
3062 | error = mnt_want_write(new_path.mnt); | |
75c3f29d DH |
3063 | if (error) |
3064 | goto out_dput; | |
dae6ad8f | 3065 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 KT |
3066 | if (error) |
3067 | goto out_drop_write; | |
dae6ad8f | 3068 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry); |
be6d3e56 | 3069 | out_drop_write: |
dae6ad8f | 3070 | mnt_drop_write(new_path.mnt); |
75c3f29d | 3071 | out_dput: |
6902d925 | 3072 | dput(new_dentry); |
dae6ad8f AV |
3073 | mutex_unlock(&new_path.dentry->d_inode->i_mutex); |
3074 | path_put(&new_path); | |
1da177e4 | 3075 | out: |
2d8f3038 | 3076 | path_put(&old_path); |
1da177e4 LT |
3077 | |
3078 | return error; | |
3079 | } | |
3080 | ||
3480b257 | 3081 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 3082 | { |
c04030e1 | 3083 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
3084 | } |
3085 | ||
1da177e4 LT |
3086 | /* |
3087 | * The worst of all namespace operations - renaming directory. "Perverted" | |
3088 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
3089 | * Problems: | |
3090 | * a) we can get into loop creation. Check is done in is_subdir(). | |
3091 | * b) race potential - two innocent renames can create a loop together. | |
3092 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 3093 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
3094 | * story. |
3095 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 3096 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
3097 | * whether the target exists). Solution: try to be smart with locking |
3098 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 3099 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
3100 | * move will be locked. Thus we can rank directories by the tree |
3101 | * (ancestors first) and rank all non-directories after them. | |
3102 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 3103 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
3104 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
3105 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
3106 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 3107 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 3108 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 3109 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 3110 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
3111 | * locking]. |
3112 | */ | |
75c96f85 AB |
3113 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
3114 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3115 | { |
3116 | int error = 0; | |
9055cba7 | 3117 | struct inode *target = new_dentry->d_inode; |
8de52778 | 3118 | unsigned max_links = new_dir->i_sb->s_max_links; |
1da177e4 LT |
3119 | |
3120 | /* | |
3121 | * If we are going to change the parent - check write permissions, | |
3122 | * we'll need to flip '..'. | |
3123 | */ | |
3124 | if (new_dir != old_dir) { | |
f419a2e3 | 3125 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3126 | if (error) |
3127 | return error; | |
3128 | } | |
3129 | ||
3130 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3131 | if (error) | |
3132 | return error; | |
3133 | ||
1d2ef590 | 3134 | dget(new_dentry); |
d83c49f3 | 3135 | if (target) |
1b1dcc1b | 3136 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
3137 | |
3138 | error = -EBUSY; | |
3139 | if (d_mountpoint(old_dentry) || d_mountpoint(new_dentry)) | |
3140 | goto out; | |
3141 | ||
8de52778 AV |
3142 | error = -EMLINK; |
3143 | if (max_links && !target && new_dir != old_dir && | |
3144 | new_dir->i_nlink >= max_links) | |
3145 | goto out; | |
3146 | ||
3cebde24 SW |
3147 | if (target) |
3148 | shrink_dcache_parent(new_dentry); | |
9055cba7 SW |
3149 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
3150 | if (error) | |
3151 | goto out; | |
3152 | ||
1da177e4 | 3153 | if (target) { |
9055cba7 SW |
3154 | target->i_flags |= S_DEAD; |
3155 | dont_mount(new_dentry); | |
1da177e4 | 3156 | } |
9055cba7 SW |
3157 | out: |
3158 | if (target) | |
3159 | mutex_unlock(&target->i_mutex); | |
1d2ef590 | 3160 | dput(new_dentry); |
e31e14ec | 3161 | if (!error) |
349457cc MF |
3162 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3163 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3164 | return error; |
3165 | } | |
3166 | ||
75c96f85 AB |
3167 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3168 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 | 3169 | { |
51892bbb | 3170 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3171 | int error; |
3172 | ||
3173 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3174 | if (error) | |
3175 | return error; | |
3176 | ||
3177 | dget(new_dentry); | |
1da177e4 | 3178 | if (target) |
1b1dcc1b | 3179 | mutex_lock(&target->i_mutex); |
51892bbb SW |
3180 | |
3181 | error = -EBUSY; | |
1da177e4 | 3182 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
51892bbb SW |
3183 | goto out; |
3184 | ||
3185 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3186 | if (error) | |
3187 | goto out; | |
3188 | ||
3189 | if (target) | |
3190 | dont_mount(new_dentry); | |
3191 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) | |
3192 | d_move(old_dentry, new_dentry); | |
3193 | out: | |
1da177e4 | 3194 | if (target) |
1b1dcc1b | 3195 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3196 | dput(new_dentry); |
3197 | return error; | |
3198 | } | |
3199 | ||
3200 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3201 | struct inode *new_dir, struct dentry *new_dentry) | |
3202 | { | |
3203 | int error; | |
3204 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3205 | const unsigned char *old_name; |
1da177e4 LT |
3206 | |
3207 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3208 | return 0; | |
3209 | ||
3210 | error = may_delete(old_dir, old_dentry, is_dir); | |
3211 | if (error) | |
3212 | return error; | |
3213 | ||
3214 | if (!new_dentry->d_inode) | |
a95164d9 | 3215 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3216 | else |
3217 | error = may_delete(new_dir, new_dentry, is_dir); | |
3218 | if (error) | |
3219 | return error; | |
3220 | ||
acfa4380 | 3221 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3222 | return -EPERM; |
3223 | ||
0eeca283 RL |
3224 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3225 | ||
1da177e4 LT |
3226 | if (is_dir) |
3227 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3228 | else | |
3229 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3230 | if (!error) |
3231 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3232 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3233 | fsnotify_oldname_free(old_name); |
3234 | ||
1da177e4 LT |
3235 | return error; |
3236 | } | |
3237 | ||
2e4d0924 HC |
3238 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3239 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3240 | { |
2ad94ae6 AV |
3241 | struct dentry *old_dir, *new_dir; |
3242 | struct dentry *old_dentry, *new_dentry; | |
3243 | struct dentry *trap; | |
1da177e4 | 3244 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3245 | char *from; |
3246 | char *to; | |
3247 | int error; | |
1da177e4 | 3248 | |
2ad94ae6 | 3249 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3250 | if (error) |
3251 | goto exit; | |
3252 | ||
2ad94ae6 | 3253 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3254 | if (error) |
3255 | goto exit1; | |
3256 | ||
3257 | error = -EXDEV; | |
4ac91378 | 3258 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3259 | goto exit2; |
3260 | ||
4ac91378 | 3261 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3262 | error = -EBUSY; |
3263 | if (oldnd.last_type != LAST_NORM) | |
3264 | goto exit2; | |
3265 | ||
4ac91378 | 3266 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3267 | if (newnd.last_type != LAST_NORM) |
3268 | goto exit2; | |
3269 | ||
0612d9fb OH |
3270 | oldnd.flags &= ~LOOKUP_PARENT; |
3271 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3272 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3273 | |
1da177e4 LT |
3274 | trap = lock_rename(new_dir, old_dir); |
3275 | ||
49705b77 | 3276 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3277 | error = PTR_ERR(old_dentry); |
3278 | if (IS_ERR(old_dentry)) | |
3279 | goto exit3; | |
3280 | /* source must exist */ | |
3281 | error = -ENOENT; | |
3282 | if (!old_dentry->d_inode) | |
3283 | goto exit4; | |
3284 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3285 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3286 | error = -ENOTDIR; | |
3287 | if (oldnd.last.name[oldnd.last.len]) | |
3288 | goto exit4; | |
3289 | if (newnd.last.name[newnd.last.len]) | |
3290 | goto exit4; | |
3291 | } | |
3292 | /* source should not be ancestor of target */ | |
3293 | error = -EINVAL; | |
3294 | if (old_dentry == trap) | |
3295 | goto exit4; | |
49705b77 | 3296 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3297 | error = PTR_ERR(new_dentry); |
3298 | if (IS_ERR(new_dentry)) | |
3299 | goto exit4; | |
3300 | /* target should not be an ancestor of source */ | |
3301 | error = -ENOTEMPTY; | |
3302 | if (new_dentry == trap) | |
3303 | goto exit5; | |
3304 | ||
9079b1eb DH |
3305 | error = mnt_want_write(oldnd.path.mnt); |
3306 | if (error) | |
3307 | goto exit5; | |
be6d3e56 KT |
3308 | error = security_path_rename(&oldnd.path, old_dentry, |
3309 | &newnd.path, new_dentry); | |
3310 | if (error) | |
3311 | goto exit6; | |
1da177e4 LT |
3312 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3313 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 3314 | exit6: |
9079b1eb | 3315 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
3316 | exit5: |
3317 | dput(new_dentry); | |
3318 | exit4: | |
3319 | dput(old_dentry); | |
3320 | exit3: | |
3321 | unlock_rename(new_dir, old_dir); | |
3322 | exit2: | |
1d957f9b | 3323 | path_put(&newnd.path); |
2ad94ae6 | 3324 | putname(to); |
1da177e4 | 3325 | exit1: |
1d957f9b | 3326 | path_put(&oldnd.path); |
1da177e4 | 3327 | putname(from); |
2ad94ae6 | 3328 | exit: |
1da177e4 LT |
3329 | return error; |
3330 | } | |
3331 | ||
a26eab24 | 3332 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3333 | { |
3334 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3335 | } | |
3336 | ||
1da177e4 LT |
3337 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3338 | { | |
3339 | int len; | |
3340 | ||
3341 | len = PTR_ERR(link); | |
3342 | if (IS_ERR(link)) | |
3343 | goto out; | |
3344 | ||
3345 | len = strlen(link); | |
3346 | if (len > (unsigned) buflen) | |
3347 | len = buflen; | |
3348 | if (copy_to_user(buffer, link, len)) | |
3349 | len = -EFAULT; | |
3350 | out: | |
3351 | return len; | |
3352 | } | |
3353 | ||
3354 | /* | |
3355 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3356 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3357 | * using) it for any given inode is up to filesystem. | |
3358 | */ | |
3359 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3360 | { | |
3361 | struct nameidata nd; | |
cc314eef | 3362 | void *cookie; |
694a1764 | 3363 | int res; |
cc314eef | 3364 | |
1da177e4 | 3365 | nd.depth = 0; |
cc314eef | 3366 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3367 | if (IS_ERR(cookie)) |
3368 | return PTR_ERR(cookie); | |
3369 | ||
3370 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3371 | if (dentry->d_inode->i_op->put_link) | |
3372 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3373 | return res; | |
1da177e4 LT |
3374 | } |
3375 | ||
3376 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3377 | { | |
3378 | return __vfs_follow_link(nd, link); | |
3379 | } | |
3380 | ||
3381 | /* get the link contents into pagecache */ | |
3382 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3383 | { | |
ebd09abb DG |
3384 | char *kaddr; |
3385 | struct page *page; | |
1da177e4 | 3386 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3387 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3388 | if (IS_ERR(page)) |
6fe6900e | 3389 | return (char*)page; |
1da177e4 | 3390 | *ppage = page; |
ebd09abb DG |
3391 | kaddr = kmap(page); |
3392 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3393 | return kaddr; | |
1da177e4 LT |
3394 | } |
3395 | ||
3396 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3397 | { | |
3398 | struct page *page = NULL; | |
3399 | char *s = page_getlink(dentry, &page); | |
3400 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3401 | if (page) { | |
3402 | kunmap(page); | |
3403 | page_cache_release(page); | |
3404 | } | |
3405 | return res; | |
3406 | } | |
3407 | ||
cc314eef | 3408 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3409 | { |
cc314eef | 3410 | struct page *page = NULL; |
1da177e4 | 3411 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3412 | return page; |
1da177e4 LT |
3413 | } |
3414 | ||
cc314eef | 3415 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3416 | { |
cc314eef LT |
3417 | struct page *page = cookie; |
3418 | ||
3419 | if (page) { | |
1da177e4 LT |
3420 | kunmap(page); |
3421 | page_cache_release(page); | |
1da177e4 LT |
3422 | } |
3423 | } | |
3424 | ||
54566b2c NP |
3425 | /* |
3426 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3427 | */ | |
3428 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3429 | { |
3430 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3431 | struct page *page; |
afddba49 | 3432 | void *fsdata; |
beb497ab | 3433 | int err; |
1da177e4 | 3434 | char *kaddr; |
54566b2c NP |
3435 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3436 | if (nofs) | |
3437 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3438 | |
7e53cac4 | 3439 | retry: |
afddba49 | 3440 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3441 | flags, &page, &fsdata); |
1da177e4 | 3442 | if (err) |
afddba49 NP |
3443 | goto fail; |
3444 | ||
e8e3c3d6 | 3445 | kaddr = kmap_atomic(page); |
1da177e4 | 3446 | memcpy(kaddr, symname, len-1); |
e8e3c3d6 | 3447 | kunmap_atomic(kaddr); |
afddba49 NP |
3448 | |
3449 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3450 | page, fsdata); | |
1da177e4 LT |
3451 | if (err < 0) |
3452 | goto fail; | |
afddba49 NP |
3453 | if (err < len-1) |
3454 | goto retry; | |
3455 | ||
1da177e4 LT |
3456 | mark_inode_dirty(inode); |
3457 | return 0; | |
1da177e4 LT |
3458 | fail: |
3459 | return err; | |
3460 | } | |
3461 | ||
0adb25d2 KK |
3462 | int page_symlink(struct inode *inode, const char *symname, int len) |
3463 | { | |
3464 | return __page_symlink(inode, symname, len, | |
54566b2c | 3465 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3466 | } |
3467 | ||
92e1d5be | 3468 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3469 | .readlink = generic_readlink, |
3470 | .follow_link = page_follow_link_light, | |
3471 | .put_link = page_put_link, | |
3472 | }; | |
3473 | ||
2d8f3038 | 3474 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3475 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3476 | EXPORT_SYMBOL(follow_down); |
3477 | EXPORT_SYMBOL(follow_up); | |
3478 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
3479 | EXPORT_SYMBOL(getname); | |
3480 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3481 | EXPORT_SYMBOL(lookup_one_len); |
3482 | EXPORT_SYMBOL(page_follow_link_light); | |
3483 | EXPORT_SYMBOL(page_put_link); | |
3484 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3485 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3486 | EXPORT_SYMBOL(page_symlink); |
3487 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
d1811465 | 3488 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3489 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3490 | EXPORT_SYMBOL(inode_permission); |
1da177e4 LT |
3491 | EXPORT_SYMBOL(unlock_rename); |
3492 | EXPORT_SYMBOL(vfs_create); | |
3493 | EXPORT_SYMBOL(vfs_follow_link); | |
3494 | EXPORT_SYMBOL(vfs_link); | |
3495 | EXPORT_SYMBOL(vfs_mkdir); | |
3496 | EXPORT_SYMBOL(vfs_mknod); | |
3497 | EXPORT_SYMBOL(generic_permission); | |
3498 | EXPORT_SYMBOL(vfs_readlink); | |
3499 | EXPORT_SYMBOL(vfs_rename); | |
3500 | EXPORT_SYMBOL(vfs_rmdir); | |
3501 | EXPORT_SYMBOL(vfs_symlink); | |
3502 | EXPORT_SYMBOL(vfs_unlink); | |
3503 | EXPORT_SYMBOL(dentry_unhash); | |
3504 | EXPORT_SYMBOL(generic_readlink); |