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> | |
18 | #include <linux/module.h> | |
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 | */ | |
858119e1 | 119 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
120 | { |
121 | int retval; | |
122 | unsigned long len = PATH_MAX; | |
123 | ||
124 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
125 | if ((unsigned long) filename >= TASK_SIZE) | |
126 | return -EFAULT; | |
127 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
128 | len = TASK_SIZE - (unsigned long) filename; | |
129 | } | |
130 | ||
131 | retval = strncpy_from_user(page, filename, len); | |
132 | if (retval > 0) { | |
133 | if (retval < len) | |
134 | return 0; | |
135 | return -ENAMETOOLONG; | |
136 | } else if (!retval) | |
137 | retval = -ENOENT; | |
138 | return retval; | |
139 | } | |
140 | ||
1fa1e7f6 | 141 | static char *getname_flags(const char __user *filename, int flags, int *empty) |
1da177e4 | 142 | { |
4043cde8 EP |
143 | char *result = __getname(); |
144 | int retval; | |
145 | ||
146 | if (!result) | |
147 | return ERR_PTR(-ENOMEM); | |
148 | ||
149 | retval = do_getname(filename, result); | |
150 | if (retval < 0) { | |
151 | if (retval == -ENOENT && empty) | |
152 | *empty = 1; | |
153 | if (retval != -ENOENT || !(flags & LOOKUP_EMPTY)) { | |
154 | __putname(result); | |
155 | return ERR_PTR(retval); | |
1da177e4 LT |
156 | } |
157 | } | |
158 | audit_getname(result); | |
159 | return result; | |
160 | } | |
161 | ||
f52e0c11 AV |
162 | char *getname(const char __user * filename) |
163 | { | |
1fa1e7f6 | 164 | return getname_flags(filename, 0, 0); |
f52e0c11 AV |
165 | } |
166 | ||
1da177e4 LT |
167 | #ifdef CONFIG_AUDITSYSCALL |
168 | void putname(const char *name) | |
169 | { | |
5ac3a9c2 | 170 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
171 | audit_putname(name); |
172 | else | |
173 | __putname(name); | |
174 | } | |
175 | EXPORT_SYMBOL(putname); | |
176 | #endif | |
177 | ||
e77819e5 LT |
178 | static int check_acl(struct inode *inode, int mask) |
179 | { | |
84635d68 | 180 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
181 | struct posix_acl *acl; |
182 | ||
e77819e5 | 183 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
184 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
185 | if (!acl) | |
e77819e5 | 186 | return -EAGAIN; |
3567866b AV |
187 | /* no ->get_acl() calls in RCU mode... */ |
188 | if (acl == ACL_NOT_CACHED) | |
189 | return -ECHILD; | |
206b1d09 | 190 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
191 | } |
192 | ||
193 | acl = get_cached_acl(inode, ACL_TYPE_ACCESS); | |
194 | ||
195 | /* | |
4e34e719 CH |
196 | * A filesystem can force a ACL callback by just never filling the |
197 | * ACL cache. But normally you'd fill the cache either at inode | |
198 | * instantiation time, or on the first ->get_acl call. | |
e77819e5 | 199 | * |
4e34e719 CH |
200 | * If the filesystem doesn't have a get_acl() function at all, we'll |
201 | * just create the negative cache entry. | |
e77819e5 LT |
202 | */ |
203 | if (acl == ACL_NOT_CACHED) { | |
4e34e719 CH |
204 | if (inode->i_op->get_acl) { |
205 | acl = inode->i_op->get_acl(inode, ACL_TYPE_ACCESS); | |
206 | if (IS_ERR(acl)) | |
207 | return PTR_ERR(acl); | |
208 | } else { | |
209 | set_cached_acl(inode, ACL_TYPE_ACCESS, NULL); | |
210 | return -EAGAIN; | |
211 | } | |
e77819e5 LT |
212 | } |
213 | ||
214 | if (acl) { | |
215 | int error = posix_acl_permission(inode, acl, mask); | |
216 | posix_acl_release(acl); | |
217 | return error; | |
218 | } | |
84635d68 | 219 | #endif |
e77819e5 LT |
220 | |
221 | return -EAGAIN; | |
222 | } | |
223 | ||
5909ccaa | 224 | /* |
948409c7 | 225 | * This does the basic permission checking |
1da177e4 | 226 | */ |
7e40145e | 227 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 228 | { |
26cf46be | 229 | unsigned int mode = inode->i_mode; |
1da177e4 | 230 | |
e795b717 SH |
231 | if (current_user_ns() != inode_userns(inode)) |
232 | goto other_perms; | |
233 | ||
14067ff5 | 234 | if (likely(current_fsuid() == inode->i_uid)) |
1da177e4 LT |
235 | mode >>= 6; |
236 | else { | |
e77819e5 | 237 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 238 | int error = check_acl(inode, mask); |
b74c79e9 NP |
239 | if (error != -EAGAIN) |
240 | return error; | |
1da177e4 LT |
241 | } |
242 | ||
243 | if (in_group_p(inode->i_gid)) | |
244 | mode >>= 3; | |
245 | } | |
246 | ||
e795b717 | 247 | other_perms: |
1da177e4 LT |
248 | /* |
249 | * If the DACs are ok we don't need any capability check. | |
250 | */ | |
9c2c7039 | 251 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 252 | return 0; |
5909ccaa LT |
253 | return -EACCES; |
254 | } | |
255 | ||
256 | /** | |
b74c79e9 | 257 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 258 | * @inode: inode to check access rights for |
8fd90c8d | 259 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
260 | * |
261 | * Used to check for read/write/execute permissions on a file. | |
262 | * We use "fsuid" for this, letting us set arbitrary permissions | |
263 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
264 | * are used for other things. |
265 | * | |
266 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
267 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
268 | * It would then be called again in ref-walk mode. | |
5909ccaa | 269 | */ |
2830ba7f | 270 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
271 | { |
272 | int ret; | |
273 | ||
274 | /* | |
948409c7 | 275 | * Do the basic permission checks. |
5909ccaa | 276 | */ |
7e40145e | 277 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
278 | if (ret != -EACCES) |
279 | return ret; | |
1da177e4 | 280 | |
d594e7ec AV |
281 | if (S_ISDIR(inode->i_mode)) { |
282 | /* DACs are overridable for directories */ | |
283 | if (ns_capable(inode_userns(inode), CAP_DAC_OVERRIDE)) | |
284 | return 0; | |
285 | if (!(mask & MAY_WRITE)) | |
286 | if (ns_capable(inode_userns(inode), CAP_DAC_READ_SEARCH)) | |
287 | return 0; | |
288 | return -EACCES; | |
289 | } | |
1da177e4 LT |
290 | /* |
291 | * Read/write DACs are always overridable. | |
d594e7ec AV |
292 | * Executable DACs are overridable when there is |
293 | * at least one exec bit set. | |
1da177e4 | 294 | */ |
d594e7ec | 295 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) |
e795b717 | 296 | if (ns_capable(inode_userns(inode), CAP_DAC_OVERRIDE)) |
1da177e4 LT |
297 | return 0; |
298 | ||
299 | /* | |
300 | * Searching includes executable on directories, else just read. | |
301 | */ | |
7ea66001 | 302 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 303 | if (mask == MAY_READ) |
e795b717 | 304 | if (ns_capable(inode_userns(inode), CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
305 | return 0; |
306 | ||
307 | return -EACCES; | |
308 | } | |
309 | ||
3ddcd056 LT |
310 | /* |
311 | * We _really_ want to just do "generic_permission()" without | |
312 | * even looking at the inode->i_op values. So we keep a cache | |
313 | * flag in inode->i_opflags, that says "this has not special | |
314 | * permission function, use the fast case". | |
315 | */ | |
316 | static inline int do_inode_permission(struct inode *inode, int mask) | |
317 | { | |
318 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
319 | if (likely(inode->i_op->permission)) | |
320 | return inode->i_op->permission(inode, mask); | |
321 | ||
322 | /* This gets set once for the inode lifetime */ | |
323 | spin_lock(&inode->i_lock); | |
324 | inode->i_opflags |= IOP_FASTPERM; | |
325 | spin_unlock(&inode->i_lock); | |
326 | } | |
327 | return generic_permission(inode, mask); | |
328 | } | |
329 | ||
cb23beb5 CH |
330 | /** |
331 | * inode_permission - check for access rights to a given inode | |
332 | * @inode: inode to check permission on | |
8fd90c8d | 333 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
cb23beb5 CH |
334 | * |
335 | * Used to check for read/write/execute permissions on an inode. | |
336 | * We use "fsuid" for this, letting us set arbitrary permissions | |
337 | * for filesystem access without changing the "normal" uids which | |
338 | * are used for other things. | |
948409c7 AG |
339 | * |
340 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
cb23beb5 | 341 | */ |
f419a2e3 | 342 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 343 | { |
e6305c43 | 344 | int retval; |
1da177e4 | 345 | |
3ddcd056 | 346 | if (unlikely(mask & MAY_WRITE)) { |
22590e41 | 347 | umode_t mode = inode->i_mode; |
1da177e4 LT |
348 | |
349 | /* | |
350 | * Nobody gets write access to a read-only fs. | |
351 | */ | |
352 | if (IS_RDONLY(inode) && | |
353 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
354 | return -EROFS; | |
355 | ||
356 | /* | |
357 | * Nobody gets write access to an immutable file. | |
358 | */ | |
359 | if (IS_IMMUTABLE(inode)) | |
360 | return -EACCES; | |
361 | } | |
362 | ||
3ddcd056 | 363 | retval = do_inode_permission(inode, mask); |
1da177e4 LT |
364 | if (retval) |
365 | return retval; | |
366 | ||
08ce5f16 SH |
367 | retval = devcgroup_inode_permission(inode, mask); |
368 | if (retval) | |
369 | return retval; | |
370 | ||
d09ca739 | 371 | return security_inode_permission(inode, mask); |
1da177e4 LT |
372 | } |
373 | ||
5dd784d0 JB |
374 | /** |
375 | * path_get - get a reference to a path | |
376 | * @path: path to get the reference to | |
377 | * | |
378 | * Given a path increment the reference count to the dentry and the vfsmount. | |
379 | */ | |
380 | void path_get(struct path *path) | |
381 | { | |
382 | mntget(path->mnt); | |
383 | dget(path->dentry); | |
384 | } | |
385 | EXPORT_SYMBOL(path_get); | |
386 | ||
1d957f9b JB |
387 | /** |
388 | * path_put - put a reference to a path | |
389 | * @path: path to put the reference to | |
390 | * | |
391 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
392 | */ | |
393 | void path_put(struct path *path) | |
1da177e4 | 394 | { |
1d957f9b JB |
395 | dput(path->dentry); |
396 | mntput(path->mnt); | |
1da177e4 | 397 | } |
1d957f9b | 398 | EXPORT_SYMBOL(path_put); |
1da177e4 | 399 | |
19660af7 | 400 | /* |
31e6b01f | 401 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
402 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
403 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
404 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
405 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
406 | * got stuck, so ref-walk may continue from there. If this is not successful | |
407 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
408 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 409 | */ |
31e6b01f NP |
410 | |
411 | /** | |
19660af7 AV |
412 | * unlazy_walk - try to switch to ref-walk mode. |
413 | * @nd: nameidata pathwalk data | |
414 | * @dentry: child of nd->path.dentry or NULL | |
39191628 | 415 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 416 | * |
19660af7 AV |
417 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
418 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
419 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 420 | */ |
19660af7 | 421 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry) |
31e6b01f NP |
422 | { |
423 | struct fs_struct *fs = current->fs; | |
424 | struct dentry *parent = nd->path.dentry; | |
5b6ca027 | 425 | int want_root = 0; |
31e6b01f NP |
426 | |
427 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
5b6ca027 AV |
428 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
429 | want_root = 1; | |
31e6b01f NP |
430 | spin_lock(&fs->lock); |
431 | if (nd->root.mnt != fs->root.mnt || | |
432 | nd->root.dentry != fs->root.dentry) | |
433 | goto err_root; | |
434 | } | |
435 | spin_lock(&parent->d_lock); | |
19660af7 AV |
436 | if (!dentry) { |
437 | if (!__d_rcu_to_refcount(parent, nd->seq)) | |
438 | goto err_parent; | |
439 | BUG_ON(nd->inode != parent->d_inode); | |
440 | } else { | |
94c0d4ec AV |
441 | if (dentry->d_parent != parent) |
442 | goto err_parent; | |
19660af7 AV |
443 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); |
444 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
445 | goto err_child; | |
446 | /* | |
447 | * If the sequence check on the child dentry passed, then | |
448 | * the child has not been removed from its parent. This | |
449 | * means the parent dentry must be valid and able to take | |
450 | * a reference at this point. | |
451 | */ | |
452 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
453 | BUG_ON(!parent->d_count); | |
454 | parent->d_count++; | |
455 | spin_unlock(&dentry->d_lock); | |
456 | } | |
31e6b01f | 457 | spin_unlock(&parent->d_lock); |
5b6ca027 | 458 | if (want_root) { |
31e6b01f NP |
459 | path_get(&nd->root); |
460 | spin_unlock(&fs->lock); | |
461 | } | |
462 | mntget(nd->path.mnt); | |
463 | ||
464 | rcu_read_unlock(); | |
465 | br_read_unlock(vfsmount_lock); | |
466 | nd->flags &= ~LOOKUP_RCU; | |
467 | return 0; | |
19660af7 AV |
468 | |
469 | err_child: | |
31e6b01f | 470 | spin_unlock(&dentry->d_lock); |
19660af7 | 471 | err_parent: |
31e6b01f NP |
472 | spin_unlock(&parent->d_lock); |
473 | err_root: | |
5b6ca027 | 474 | if (want_root) |
31e6b01f NP |
475 | spin_unlock(&fs->lock); |
476 | return -ECHILD; | |
477 | } | |
478 | ||
834f2a4a TM |
479 | /** |
480 | * release_open_intent - free up open intent resources | |
481 | * @nd: pointer to nameidata | |
482 | */ | |
483 | void release_open_intent(struct nameidata *nd) | |
484 | { | |
2dab5974 LT |
485 | struct file *file = nd->intent.open.file; |
486 | ||
487 | if (file && !IS_ERR(file)) { | |
488 | if (file->f_path.dentry == NULL) | |
489 | put_filp(file); | |
490 | else | |
491 | fput(file); | |
492 | } | |
834f2a4a TM |
493 | } |
494 | ||
f60aef7e | 495 | static inline int d_revalidate(struct dentry *dentry, struct nameidata *nd) |
34286d66 | 496 | { |
f60aef7e | 497 | return dentry->d_op->d_revalidate(dentry, nd); |
34286d66 NP |
498 | } |
499 | ||
9f1fafee AV |
500 | /** |
501 | * complete_walk - successful completion of path walk | |
502 | * @nd: pointer nameidata | |
39159de2 | 503 | * |
9f1fafee AV |
504 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
505 | * Revalidate the final result, unless we'd already done that during | |
506 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
507 | * success, -error on failure. In case of failure caller does not | |
508 | * need to drop nd->path. | |
39159de2 | 509 | */ |
9f1fafee | 510 | static int complete_walk(struct nameidata *nd) |
39159de2 | 511 | { |
16c2cd71 | 512 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 513 | int status; |
39159de2 | 514 | |
9f1fafee AV |
515 | if (nd->flags & LOOKUP_RCU) { |
516 | nd->flags &= ~LOOKUP_RCU; | |
517 | if (!(nd->flags & LOOKUP_ROOT)) | |
518 | nd->root.mnt = NULL; | |
519 | spin_lock(&dentry->d_lock); | |
520 | if (unlikely(!__d_rcu_to_refcount(dentry, nd->seq))) { | |
521 | spin_unlock(&dentry->d_lock); | |
522 | rcu_read_unlock(); | |
523 | br_read_unlock(vfsmount_lock); | |
524 | return -ECHILD; | |
525 | } | |
526 | BUG_ON(nd->inode != dentry->d_inode); | |
527 | spin_unlock(&dentry->d_lock); | |
528 | mntget(nd->path.mnt); | |
529 | rcu_read_unlock(); | |
530 | br_read_unlock(vfsmount_lock); | |
531 | } | |
532 | ||
16c2cd71 AV |
533 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
534 | return 0; | |
535 | ||
536 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
39159de2 JL |
537 | return 0; |
538 | ||
16c2cd71 AV |
539 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) |
540 | return 0; | |
541 | ||
542 | /* Note: we do not d_invalidate() */ | |
34286d66 | 543 | status = d_revalidate(dentry, nd); |
39159de2 JL |
544 | if (status > 0) |
545 | return 0; | |
546 | ||
16c2cd71 | 547 | if (!status) |
39159de2 | 548 | status = -ESTALE; |
16c2cd71 | 549 | |
9f1fafee | 550 | path_put(&nd->path); |
39159de2 JL |
551 | return status; |
552 | } | |
553 | ||
2a737871 AV |
554 | static __always_inline void set_root(struct nameidata *nd) |
555 | { | |
f7ad3c6b MS |
556 | if (!nd->root.mnt) |
557 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
558 | } |
559 | ||
6de88d72 AV |
560 | static int link_path_walk(const char *, struct nameidata *); |
561 | ||
31e6b01f NP |
562 | static __always_inline void set_root_rcu(struct nameidata *nd) |
563 | { | |
564 | if (!nd->root.mnt) { | |
565 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
566 | unsigned seq; |
567 | ||
568 | do { | |
569 | seq = read_seqcount_begin(&fs->seq); | |
570 | nd->root = fs->root; | |
c1530019 | 571 | nd->seq = __read_seqcount_begin(&nd->root.dentry->d_seq); |
c28cc364 | 572 | } while (read_seqcount_retry(&fs->seq, seq)); |
31e6b01f NP |
573 | } |
574 | } | |
575 | ||
f1662356 | 576 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 577 | { |
31e6b01f NP |
578 | int ret; |
579 | ||
1da177e4 LT |
580 | if (IS_ERR(link)) |
581 | goto fail; | |
582 | ||
583 | if (*link == '/') { | |
2a737871 | 584 | set_root(nd); |
1d957f9b | 585 | path_put(&nd->path); |
2a737871 AV |
586 | nd->path = nd->root; |
587 | path_get(&nd->root); | |
16c2cd71 | 588 | nd->flags |= LOOKUP_JUMPED; |
1da177e4 | 589 | } |
31e6b01f | 590 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 591 | |
31e6b01f NP |
592 | ret = link_path_walk(link, nd); |
593 | return ret; | |
1da177e4 | 594 | fail: |
1d957f9b | 595 | path_put(&nd->path); |
1da177e4 LT |
596 | return PTR_ERR(link); |
597 | } | |
598 | ||
1d957f9b | 599 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
600 | { |
601 | dput(path->dentry); | |
4ac91378 | 602 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
603 | mntput(path->mnt); |
604 | } | |
605 | ||
7b9337aa NP |
606 | static inline void path_to_nameidata(const struct path *path, |
607 | struct nameidata *nd) | |
051d3812 | 608 | { |
31e6b01f NP |
609 | if (!(nd->flags & LOOKUP_RCU)) { |
610 | dput(nd->path.dentry); | |
611 | if (nd->path.mnt != path->mnt) | |
612 | mntput(nd->path.mnt); | |
9a229683 | 613 | } |
31e6b01f | 614 | nd->path.mnt = path->mnt; |
4ac91378 | 615 | nd->path.dentry = path->dentry; |
051d3812 IK |
616 | } |
617 | ||
574197e0 AV |
618 | static inline void put_link(struct nameidata *nd, struct path *link, void *cookie) |
619 | { | |
620 | struct inode *inode = link->dentry->d_inode; | |
621 | if (!IS_ERR(cookie) && inode->i_op->put_link) | |
622 | inode->i_op->put_link(link->dentry, nd, cookie); | |
623 | path_put(link); | |
624 | } | |
625 | ||
def4af30 | 626 | static __always_inline int |
574197e0 | 627 | follow_link(struct path *link, struct nameidata *nd, void **p) |
1da177e4 LT |
628 | { |
629 | int error; | |
7b9337aa | 630 | struct dentry *dentry = link->dentry; |
1da177e4 | 631 | |
844a3917 AV |
632 | BUG_ON(nd->flags & LOOKUP_RCU); |
633 | ||
0e794589 AV |
634 | if (link->mnt == nd->path.mnt) |
635 | mntget(link->mnt); | |
636 | ||
574197e0 AV |
637 | if (unlikely(current->total_link_count >= 40)) { |
638 | *p = ERR_PTR(-ELOOP); /* no ->put_link(), please */ | |
574197e0 AV |
639 | path_put(&nd->path); |
640 | return -ELOOP; | |
641 | } | |
642 | cond_resched(); | |
643 | current->total_link_count++; | |
644 | ||
7b9337aa | 645 | touch_atime(link->mnt, dentry); |
1da177e4 | 646 | nd_set_link(nd, NULL); |
cd4e91d3 | 647 | |
36f3b4f6 AV |
648 | error = security_inode_follow_link(link->dentry, nd); |
649 | if (error) { | |
650 | *p = ERR_PTR(error); /* no ->put_link(), please */ | |
651 | path_put(&nd->path); | |
652 | return error; | |
653 | } | |
654 | ||
86acdca1 | 655 | nd->last_type = LAST_BIND; |
def4af30 AV |
656 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
657 | error = PTR_ERR(*p); | |
658 | if (!IS_ERR(*p)) { | |
1da177e4 | 659 | char *s = nd_get_link(nd); |
cc314eef | 660 | error = 0; |
1da177e4 LT |
661 | if (s) |
662 | error = __vfs_follow_link(nd, s); | |
bcda7652 | 663 | else if (nd->last_type == LAST_BIND) { |
16c2cd71 | 664 | nd->flags |= LOOKUP_JUMPED; |
b21041d0 AV |
665 | nd->inode = nd->path.dentry->d_inode; |
666 | if (nd->inode->i_op->follow_link) { | |
bcda7652 AV |
667 | /* stepped on a _really_ weird one */ |
668 | path_put(&nd->path); | |
669 | error = -ELOOP; | |
670 | } | |
671 | } | |
1da177e4 | 672 | } |
1da177e4 LT |
673 | return error; |
674 | } | |
675 | ||
31e6b01f NP |
676 | static int follow_up_rcu(struct path *path) |
677 | { | |
0714a533 AV |
678 | struct mount *mnt = real_mount(path->mnt); |
679 | struct mount *parent; | |
31e6b01f NP |
680 | struct dentry *mountpoint; |
681 | ||
0714a533 AV |
682 | parent = mnt->mnt_parent; |
683 | if (&parent->mnt == path->mnt) | |
31e6b01f | 684 | return 0; |
a73324da | 685 | mountpoint = mnt->mnt_mountpoint; |
31e6b01f | 686 | path->dentry = mountpoint; |
0714a533 | 687 | path->mnt = &parent->mnt; |
31e6b01f NP |
688 | return 1; |
689 | } | |
690 | ||
bab77ebf | 691 | int follow_up(struct path *path) |
1da177e4 | 692 | { |
0714a533 AV |
693 | struct mount *mnt = real_mount(path->mnt); |
694 | struct mount *parent; | |
1da177e4 | 695 | struct dentry *mountpoint; |
99b7db7b NP |
696 | |
697 | br_read_lock(vfsmount_lock); | |
0714a533 AV |
698 | parent = mnt->mnt_parent; |
699 | if (&parent->mnt == path->mnt) { | |
99b7db7b | 700 | br_read_unlock(vfsmount_lock); |
1da177e4 LT |
701 | return 0; |
702 | } | |
0714a533 | 703 | mntget(&parent->mnt); |
a73324da | 704 | mountpoint = dget(mnt->mnt_mountpoint); |
99b7db7b | 705 | br_read_unlock(vfsmount_lock); |
bab77ebf AV |
706 | dput(path->dentry); |
707 | path->dentry = mountpoint; | |
708 | mntput(path->mnt); | |
0714a533 | 709 | path->mnt = &parent->mnt; |
1da177e4 LT |
710 | return 1; |
711 | } | |
712 | ||
b5c84bf6 | 713 | /* |
9875cf80 DH |
714 | * Perform an automount |
715 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
716 | * were called with. | |
1da177e4 | 717 | */ |
9875cf80 DH |
718 | static int follow_automount(struct path *path, unsigned flags, |
719 | bool *need_mntput) | |
31e6b01f | 720 | { |
9875cf80 | 721 | struct vfsmount *mnt; |
ea5b778a | 722 | int err; |
9875cf80 DH |
723 | |
724 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
725 | return -EREMOTE; | |
726 | ||
0ec26fd0 MS |
727 | /* We don't want to mount if someone's just doing a stat - |
728 | * unless they're stat'ing a directory and appended a '/' to | |
729 | * the name. | |
730 | * | |
731 | * We do, however, want to mount if someone wants to open or | |
732 | * create a file of any type under the mountpoint, wants to | |
733 | * traverse through the mountpoint or wants to open the | |
734 | * mounted directory. Also, autofs may mark negative dentries | |
735 | * as being automount points. These will need the attentions | |
736 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 737 | */ |
0ec26fd0 | 738 | if (!(flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
d94c177b | 739 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
0ec26fd0 MS |
740 | path->dentry->d_inode) |
741 | return -EISDIR; | |
742 | ||
9875cf80 DH |
743 | current->total_link_count++; |
744 | if (current->total_link_count >= 40) | |
745 | return -ELOOP; | |
746 | ||
747 | mnt = path->dentry->d_op->d_automount(path); | |
748 | if (IS_ERR(mnt)) { | |
749 | /* | |
750 | * The filesystem is allowed to return -EISDIR here to indicate | |
751 | * it doesn't want to automount. For instance, autofs would do | |
752 | * this so that its userspace daemon can mount on this dentry. | |
753 | * | |
754 | * However, we can only permit this if it's a terminal point in | |
755 | * the path being looked up; if it wasn't then the remainder of | |
756 | * the path is inaccessible and we should say so. | |
757 | */ | |
49084c3b | 758 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_PARENT)) |
9875cf80 DH |
759 | return -EREMOTE; |
760 | return PTR_ERR(mnt); | |
31e6b01f | 761 | } |
ea5b778a | 762 | |
9875cf80 DH |
763 | if (!mnt) /* mount collision */ |
764 | return 0; | |
31e6b01f | 765 | |
8aef1884 AV |
766 | if (!*need_mntput) { |
767 | /* lock_mount() may release path->mnt on error */ | |
768 | mntget(path->mnt); | |
769 | *need_mntput = true; | |
770 | } | |
19a167af | 771 | err = finish_automount(mnt, path); |
9875cf80 | 772 | |
ea5b778a DH |
773 | switch (err) { |
774 | case -EBUSY: | |
775 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 776 | return 0; |
ea5b778a | 777 | case 0: |
8aef1884 | 778 | path_put(path); |
ea5b778a DH |
779 | path->mnt = mnt; |
780 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 781 | return 0; |
19a167af AV |
782 | default: |
783 | return err; | |
ea5b778a | 784 | } |
19a167af | 785 | |
463ffb2e AV |
786 | } |
787 | ||
9875cf80 DH |
788 | /* |
789 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 790 | * - Flagged for transit management (autofs) |
9875cf80 DH |
791 | * - Flagged as mountpoint |
792 | * - Flagged as automount point | |
793 | * | |
794 | * This may only be called in refwalk mode. | |
795 | * | |
796 | * Serialization is taken care of in namespace.c | |
797 | */ | |
798 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 799 | { |
8aef1884 | 800 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
801 | unsigned managed; |
802 | bool need_mntput = false; | |
8aef1884 | 803 | int ret = 0; |
9875cf80 DH |
804 | |
805 | /* Given that we're not holding a lock here, we retain the value in a | |
806 | * local variable for each dentry as we look at it so that we don't see | |
807 | * the components of that value change under us */ | |
808 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
809 | managed &= DCACHE_MANAGED_DENTRY, | |
810 | unlikely(managed != 0)) { | |
cc53ce53 DH |
811 | /* Allow the filesystem to manage the transit without i_mutex |
812 | * being held. */ | |
813 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
814 | BUG_ON(!path->dentry->d_op); | |
815 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 816 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 817 | if (ret < 0) |
8aef1884 | 818 | break; |
cc53ce53 DH |
819 | } |
820 | ||
9875cf80 DH |
821 | /* Transit to a mounted filesystem. */ |
822 | if (managed & DCACHE_MOUNTED) { | |
823 | struct vfsmount *mounted = lookup_mnt(path); | |
824 | if (mounted) { | |
825 | dput(path->dentry); | |
826 | if (need_mntput) | |
827 | mntput(path->mnt); | |
828 | path->mnt = mounted; | |
829 | path->dentry = dget(mounted->mnt_root); | |
830 | need_mntput = true; | |
831 | continue; | |
832 | } | |
833 | ||
834 | /* Something is mounted on this dentry in another | |
835 | * namespace and/or whatever was mounted there in this | |
836 | * namespace got unmounted before we managed to get the | |
837 | * vfsmount_lock */ | |
838 | } | |
839 | ||
840 | /* Handle an automount point */ | |
841 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
842 | ret = follow_automount(path, flags, &need_mntput); | |
843 | if (ret < 0) | |
8aef1884 | 844 | break; |
9875cf80 DH |
845 | continue; |
846 | } | |
847 | ||
848 | /* We didn't change the current path point */ | |
849 | break; | |
1da177e4 | 850 | } |
8aef1884 AV |
851 | |
852 | if (need_mntput && path->mnt == mnt) | |
853 | mntput(path->mnt); | |
854 | if (ret == -EISDIR) | |
855 | ret = 0; | |
a3fbbde7 | 856 | return ret < 0 ? ret : need_mntput; |
1da177e4 LT |
857 | } |
858 | ||
cc53ce53 | 859 | int follow_down_one(struct path *path) |
1da177e4 LT |
860 | { |
861 | struct vfsmount *mounted; | |
862 | ||
1c755af4 | 863 | mounted = lookup_mnt(path); |
1da177e4 | 864 | if (mounted) { |
9393bd07 AV |
865 | dput(path->dentry); |
866 | mntput(path->mnt); | |
867 | path->mnt = mounted; | |
868 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
869 | return 1; |
870 | } | |
871 | return 0; | |
872 | } | |
873 | ||
62a7375e IK |
874 | static inline bool managed_dentry_might_block(struct dentry *dentry) |
875 | { | |
876 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT && | |
877 | dentry->d_op->d_manage(dentry, true) < 0); | |
878 | } | |
879 | ||
9875cf80 | 880 | /* |
287548e4 AV |
881 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
882 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
883 | */ |
884 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 885 | struct inode **inode) |
9875cf80 | 886 | { |
62a7375e | 887 | for (;;) { |
c7105365 | 888 | struct mount *mounted; |
62a7375e IK |
889 | /* |
890 | * Don't forget we might have a non-mountpoint managed dentry | |
891 | * that wants to block transit. | |
892 | */ | |
287548e4 | 893 | if (unlikely(managed_dentry_might_block(path->dentry))) |
ab90911f | 894 | return false; |
62a7375e IK |
895 | |
896 | if (!d_mountpoint(path->dentry)) | |
897 | break; | |
898 | ||
9875cf80 DH |
899 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
900 | if (!mounted) | |
901 | break; | |
c7105365 AV |
902 | path->mnt = &mounted->mnt; |
903 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 904 | nd->flags |= LOOKUP_JUMPED; |
9875cf80 | 905 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
906 | /* |
907 | * Update the inode too. We don't need to re-check the | |
908 | * dentry sequence number here after this d_inode read, | |
909 | * because a mount-point is always pinned. | |
910 | */ | |
911 | *inode = path->dentry->d_inode; | |
9875cf80 | 912 | } |
9875cf80 DH |
913 | return true; |
914 | } | |
915 | ||
dea39376 | 916 | static void follow_mount_rcu(struct nameidata *nd) |
287548e4 | 917 | { |
dea39376 | 918 | while (d_mountpoint(nd->path.dentry)) { |
c7105365 | 919 | struct mount *mounted; |
dea39376 | 920 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry, 1); |
287548e4 AV |
921 | if (!mounted) |
922 | break; | |
c7105365 AV |
923 | nd->path.mnt = &mounted->mnt; |
924 | nd->path.dentry = mounted->mnt.mnt_root; | |
dea39376 | 925 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
287548e4 AV |
926 | } |
927 | } | |
928 | ||
31e6b01f NP |
929 | static int follow_dotdot_rcu(struct nameidata *nd) |
930 | { | |
31e6b01f NP |
931 | set_root_rcu(nd); |
932 | ||
9875cf80 | 933 | while (1) { |
31e6b01f NP |
934 | if (nd->path.dentry == nd->root.dentry && |
935 | nd->path.mnt == nd->root.mnt) { | |
936 | break; | |
937 | } | |
938 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
939 | struct dentry *old = nd->path.dentry; | |
940 | struct dentry *parent = old->d_parent; | |
941 | unsigned seq; | |
942 | ||
943 | seq = read_seqcount_begin(&parent->d_seq); | |
944 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 945 | goto failed; |
31e6b01f NP |
946 | nd->path.dentry = parent; |
947 | nd->seq = seq; | |
948 | break; | |
949 | } | |
950 | if (!follow_up_rcu(&nd->path)) | |
951 | break; | |
952 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
31e6b01f | 953 | } |
dea39376 AV |
954 | follow_mount_rcu(nd); |
955 | nd->inode = nd->path.dentry->d_inode; | |
31e6b01f | 956 | return 0; |
ef7562d5 AV |
957 | |
958 | failed: | |
959 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
960 | if (!(nd->flags & LOOKUP_ROOT)) |
961 | nd->root.mnt = NULL; | |
ef7562d5 AV |
962 | rcu_read_unlock(); |
963 | br_read_unlock(vfsmount_lock); | |
964 | return -ECHILD; | |
31e6b01f NP |
965 | } |
966 | ||
cc53ce53 DH |
967 | /* |
968 | * Follow down to the covering mount currently visible to userspace. At each | |
969 | * point, the filesystem owning that dentry may be queried as to whether the | |
970 | * caller is permitted to proceed or not. | |
cc53ce53 | 971 | */ |
7cc90cc3 | 972 | int follow_down(struct path *path) |
cc53ce53 DH |
973 | { |
974 | unsigned managed; | |
975 | int ret; | |
976 | ||
977 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
978 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
979 | /* Allow the filesystem to manage the transit without i_mutex | |
980 | * being held. | |
981 | * | |
982 | * We indicate to the filesystem if someone is trying to mount | |
983 | * something here. This gives autofs the chance to deny anyone | |
984 | * other than its daemon the right to mount on its | |
985 | * superstructure. | |
986 | * | |
987 | * The filesystem may sleep at this point. | |
988 | */ | |
989 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
990 | BUG_ON(!path->dentry->d_op); | |
991 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 992 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 993 | path->dentry, false); |
cc53ce53 DH |
994 | if (ret < 0) |
995 | return ret == -EISDIR ? 0 : ret; | |
996 | } | |
997 | ||
998 | /* Transit to a mounted filesystem. */ | |
999 | if (managed & DCACHE_MOUNTED) { | |
1000 | struct vfsmount *mounted = lookup_mnt(path); | |
1001 | if (!mounted) | |
1002 | break; | |
1003 | dput(path->dentry); | |
1004 | mntput(path->mnt); | |
1005 | path->mnt = mounted; | |
1006 | path->dentry = dget(mounted->mnt_root); | |
1007 | continue; | |
1008 | } | |
1009 | ||
1010 | /* Don't handle automount points here */ | |
1011 | break; | |
1012 | } | |
1013 | return 0; | |
1014 | } | |
1015 | ||
9875cf80 DH |
1016 | /* |
1017 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1018 | */ | |
1019 | static void follow_mount(struct path *path) | |
1020 | { | |
1021 | while (d_mountpoint(path->dentry)) { | |
1022 | struct vfsmount *mounted = lookup_mnt(path); | |
1023 | if (!mounted) | |
1024 | break; | |
1025 | dput(path->dentry); | |
1026 | mntput(path->mnt); | |
1027 | path->mnt = mounted; | |
1028 | path->dentry = dget(mounted->mnt_root); | |
1029 | } | |
1030 | } | |
1031 | ||
31e6b01f | 1032 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1033 | { |
2a737871 | 1034 | set_root(nd); |
e518ddb7 | 1035 | |
1da177e4 | 1036 | while(1) { |
4ac91378 | 1037 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1038 | |
2a737871 AV |
1039 | if (nd->path.dentry == nd->root.dentry && |
1040 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1041 | break; |
1042 | } | |
4ac91378 | 1043 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1044 | /* rare case of legitimate dget_parent()... */ |
1045 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1046 | dput(old); |
1047 | break; | |
1048 | } | |
3088dd70 | 1049 | if (!follow_up(&nd->path)) |
1da177e4 | 1050 | break; |
1da177e4 | 1051 | } |
79ed0226 | 1052 | follow_mount(&nd->path); |
31e6b01f | 1053 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1054 | } |
1055 | ||
baa03890 NP |
1056 | /* |
1057 | * Allocate a dentry with name and parent, and perform a parent | |
1058 | * directory ->lookup on it. Returns the new dentry, or ERR_PTR | |
1059 | * on error. parent->d_inode->i_mutex must be held. d_lookup must | |
1060 | * have verified that no child exists while under i_mutex. | |
1061 | */ | |
1062 | static struct dentry *d_alloc_and_lookup(struct dentry *parent, | |
1063 | struct qstr *name, struct nameidata *nd) | |
1064 | { | |
1065 | struct inode *inode = parent->d_inode; | |
1066 | struct dentry *dentry; | |
1067 | struct dentry *old; | |
1068 | ||
1069 | /* Don't create child dentry for a dead directory. */ | |
1070 | if (unlikely(IS_DEADDIR(inode))) | |
1071 | return ERR_PTR(-ENOENT); | |
1072 | ||
1073 | dentry = d_alloc(parent, name); | |
1074 | if (unlikely(!dentry)) | |
1075 | return ERR_PTR(-ENOMEM); | |
1076 | ||
1077 | old = inode->i_op->lookup(inode, dentry, nd); | |
1078 | if (unlikely(old)) { | |
1079 | dput(dentry); | |
1080 | dentry = old; | |
1081 | } | |
1082 | return dentry; | |
1083 | } | |
1084 | ||
44396f4b JB |
1085 | /* |
1086 | * We already have a dentry, but require a lookup to be performed on the parent | |
1087 | * directory to fill in d_inode. Returns the new dentry, or ERR_PTR on error. | |
1088 | * parent->d_inode->i_mutex must be held. d_lookup must have verified that no | |
1089 | * child exists while under i_mutex. | |
1090 | */ | |
1091 | static struct dentry *d_inode_lookup(struct dentry *parent, struct dentry *dentry, | |
1092 | struct nameidata *nd) | |
1093 | { | |
1094 | struct inode *inode = parent->d_inode; | |
1095 | struct dentry *old; | |
1096 | ||
1097 | /* Don't create child dentry for a dead directory. */ | |
e188dc02 MS |
1098 | if (unlikely(IS_DEADDIR(inode))) { |
1099 | dput(dentry); | |
44396f4b | 1100 | return ERR_PTR(-ENOENT); |
e188dc02 | 1101 | } |
44396f4b JB |
1102 | |
1103 | old = inode->i_op->lookup(inode, dentry, nd); | |
1104 | if (unlikely(old)) { | |
1105 | dput(dentry); | |
1106 | dentry = old; | |
1107 | } | |
1108 | return dentry; | |
1109 | } | |
1110 | ||
1da177e4 LT |
1111 | /* |
1112 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1113 | * small and for now I'd prefer to have fast path as straight as possible. | |
1114 | * It _is_ time-critical. | |
1115 | */ | |
1116 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
31e6b01f | 1117 | struct path *path, struct inode **inode) |
1da177e4 | 1118 | { |
4ac91378 | 1119 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1120 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1121 | int need_reval = 1; |
1122 | int status = 1; | |
9875cf80 DH |
1123 | int err; |
1124 | ||
b04f784e NP |
1125 | /* |
1126 | * Rename seqlock is not required here because in the off chance | |
1127 | * of a false negative due to a concurrent rename, we're going to | |
1128 | * do the non-racy lookup, below. | |
1129 | */ | |
31e6b01f NP |
1130 | if (nd->flags & LOOKUP_RCU) { |
1131 | unsigned seq; | |
31e6b01f NP |
1132 | *inode = nd->inode; |
1133 | dentry = __d_lookup_rcu(parent, name, &seq, inode); | |
5a18fff2 AV |
1134 | if (!dentry) |
1135 | goto unlazy; | |
1136 | ||
31e6b01f NP |
1137 | /* Memory barrier in read_seqcount_begin of child is enough */ |
1138 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) | |
1139 | return -ECHILD; | |
31e6b01f | 1140 | nd->seq = seq; |
5a18fff2 | 1141 | |
24643087 | 1142 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
5a18fff2 AV |
1143 | status = d_revalidate(dentry, nd); |
1144 | if (unlikely(status <= 0)) { | |
1145 | if (status != -ECHILD) | |
1146 | need_reval = 0; | |
1147 | goto unlazy; | |
1148 | } | |
24643087 | 1149 | } |
44396f4b JB |
1150 | if (unlikely(d_need_lookup(dentry))) |
1151 | goto unlazy; | |
31e6b01f NP |
1152 | path->mnt = mnt; |
1153 | path->dentry = dentry; | |
d6e9bd25 AV |
1154 | if (unlikely(!__follow_mount_rcu(nd, path, inode))) |
1155 | goto unlazy; | |
1156 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1157 | goto unlazy; | |
1158 | return 0; | |
5a18fff2 | 1159 | unlazy: |
19660af7 AV |
1160 | if (unlazy_walk(nd, dentry)) |
1161 | return -ECHILD; | |
5a18fff2 AV |
1162 | } else { |
1163 | dentry = __d_lookup(parent, name); | |
9875cf80 | 1164 | } |
5a18fff2 | 1165 | |
44396f4b JB |
1166 | if (dentry && unlikely(d_need_lookup(dentry))) { |
1167 | dput(dentry); | |
1168 | dentry = NULL; | |
1169 | } | |
5a18fff2 AV |
1170 | retry: |
1171 | if (unlikely(!dentry)) { | |
1172 | struct inode *dir = parent->d_inode; | |
1173 | BUG_ON(nd->inode != dir); | |
1174 | ||
1175 | mutex_lock(&dir->i_mutex); | |
1176 | dentry = d_lookup(parent, name); | |
1177 | if (likely(!dentry)) { | |
1178 | dentry = d_alloc_and_lookup(parent, name, nd); | |
1179 | if (IS_ERR(dentry)) { | |
1180 | mutex_unlock(&dir->i_mutex); | |
1181 | return PTR_ERR(dentry); | |
1182 | } | |
1183 | /* known good */ | |
1184 | need_reval = 0; | |
1185 | status = 1; | |
44396f4b JB |
1186 | } else if (unlikely(d_need_lookup(dentry))) { |
1187 | dentry = d_inode_lookup(parent, dentry, nd); | |
1188 | if (IS_ERR(dentry)) { | |
1189 | mutex_unlock(&dir->i_mutex); | |
1190 | return PTR_ERR(dentry); | |
1191 | } | |
1192 | /* known good */ | |
1193 | need_reval = 0; | |
1194 | status = 1; | |
5a18fff2 AV |
1195 | } |
1196 | mutex_unlock(&dir->i_mutex); | |
1197 | } | |
1198 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) | |
1199 | status = d_revalidate(dentry, nd); | |
1200 | if (unlikely(status <= 0)) { | |
1201 | if (status < 0) { | |
1202 | dput(dentry); | |
1203 | return status; | |
1204 | } | |
1205 | if (!d_invalidate(dentry)) { | |
1206 | dput(dentry); | |
1207 | dentry = NULL; | |
1208 | need_reval = 1; | |
1209 | goto retry; | |
1210 | } | |
24643087 | 1211 | } |
5a18fff2 | 1212 | |
9875cf80 DH |
1213 | path->mnt = mnt; |
1214 | path->dentry = dentry; | |
1215 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1216 | if (unlikely(err < 0)) { |
1217 | path_put_conditional(path, nd); | |
9875cf80 | 1218 | return err; |
89312214 | 1219 | } |
a3fbbde7 AV |
1220 | if (err) |
1221 | nd->flags |= LOOKUP_JUMPED; | |
9875cf80 | 1222 | *inode = path->dentry->d_inode; |
1da177e4 | 1223 | return 0; |
1da177e4 LT |
1224 | } |
1225 | ||
52094c8a AV |
1226 | static inline int may_lookup(struct nameidata *nd) |
1227 | { | |
1228 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1229 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1230 | if (err != -ECHILD) |
1231 | return err; | |
19660af7 | 1232 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1233 | return -ECHILD; |
1234 | } | |
4ad5abb3 | 1235 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1236 | } |
1237 | ||
9856fa1b AV |
1238 | static inline int handle_dots(struct nameidata *nd, int type) |
1239 | { | |
1240 | if (type == LAST_DOTDOT) { | |
1241 | if (nd->flags & LOOKUP_RCU) { | |
1242 | if (follow_dotdot_rcu(nd)) | |
1243 | return -ECHILD; | |
1244 | } else | |
1245 | follow_dotdot(nd); | |
1246 | } | |
1247 | return 0; | |
1248 | } | |
1249 | ||
951361f9 AV |
1250 | static void terminate_walk(struct nameidata *nd) |
1251 | { | |
1252 | if (!(nd->flags & LOOKUP_RCU)) { | |
1253 | path_put(&nd->path); | |
1254 | } else { | |
1255 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1256 | if (!(nd->flags & LOOKUP_ROOT)) |
1257 | nd->root.mnt = NULL; | |
951361f9 AV |
1258 | rcu_read_unlock(); |
1259 | br_read_unlock(vfsmount_lock); | |
1260 | } | |
1261 | } | |
1262 | ||
3ddcd056 LT |
1263 | /* |
1264 | * Do we need to follow links? We _really_ want to be able | |
1265 | * to do this check without having to look at inode->i_op, | |
1266 | * so we keep a cache of "no, this doesn't need follow_link" | |
1267 | * for the common case. | |
1268 | */ | |
7813b94a | 1269 | static inline int should_follow_link(struct inode *inode, int follow) |
3ddcd056 LT |
1270 | { |
1271 | if (unlikely(!(inode->i_opflags & IOP_NOFOLLOW))) { | |
1272 | if (likely(inode->i_op->follow_link)) | |
1273 | return follow; | |
1274 | ||
1275 | /* This gets set once for the inode lifetime */ | |
1276 | spin_lock(&inode->i_lock); | |
1277 | inode->i_opflags |= IOP_NOFOLLOW; | |
1278 | spin_unlock(&inode->i_lock); | |
1279 | } | |
1280 | return 0; | |
1281 | } | |
1282 | ||
ce57dfc1 AV |
1283 | static inline int walk_component(struct nameidata *nd, struct path *path, |
1284 | struct qstr *name, int type, int follow) | |
1285 | { | |
1286 | struct inode *inode; | |
1287 | int err; | |
1288 | /* | |
1289 | * "." and ".." are special - ".." especially so because it has | |
1290 | * to be able to know about the current root directory and | |
1291 | * parent relationships. | |
1292 | */ | |
1293 | if (unlikely(type != LAST_NORM)) | |
1294 | return handle_dots(nd, type); | |
1295 | err = do_lookup(nd, name, path, &inode); | |
1296 | if (unlikely(err)) { | |
1297 | terminate_walk(nd); | |
1298 | return err; | |
1299 | } | |
1300 | if (!inode) { | |
1301 | path_to_nameidata(path, nd); | |
1302 | terminate_walk(nd); | |
1303 | return -ENOENT; | |
1304 | } | |
7813b94a | 1305 | if (should_follow_link(inode, follow)) { |
19660af7 AV |
1306 | if (nd->flags & LOOKUP_RCU) { |
1307 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
1308 | terminate_walk(nd); | |
1309 | return -ECHILD; | |
1310 | } | |
1311 | } | |
ce57dfc1 AV |
1312 | BUG_ON(inode != path->dentry->d_inode); |
1313 | return 1; | |
1314 | } | |
1315 | path_to_nameidata(path, nd); | |
1316 | nd->inode = inode; | |
1317 | return 0; | |
1318 | } | |
1319 | ||
b356379a AV |
1320 | /* |
1321 | * This limits recursive symlink follows to 8, while | |
1322 | * limiting consecutive symlinks to 40. | |
1323 | * | |
1324 | * Without that kind of total limit, nasty chains of consecutive | |
1325 | * symlinks can cause almost arbitrarily long lookups. | |
1326 | */ | |
1327 | static inline int nested_symlink(struct path *path, struct nameidata *nd) | |
1328 | { | |
1329 | int res; | |
1330 | ||
b356379a AV |
1331 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1332 | path_put_conditional(path, nd); | |
1333 | path_put(&nd->path); | |
1334 | return -ELOOP; | |
1335 | } | |
1a4022f8 | 1336 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); |
b356379a AV |
1337 | |
1338 | nd->depth++; | |
1339 | current->link_count++; | |
1340 | ||
1341 | do { | |
1342 | struct path link = *path; | |
1343 | void *cookie; | |
574197e0 AV |
1344 | |
1345 | res = follow_link(&link, nd, &cookie); | |
b356379a AV |
1346 | if (!res) |
1347 | res = walk_component(nd, path, &nd->last, | |
1348 | nd->last_type, LOOKUP_FOLLOW); | |
574197e0 | 1349 | put_link(nd, &link, cookie); |
b356379a AV |
1350 | } while (res > 0); |
1351 | ||
1352 | current->link_count--; | |
1353 | nd->depth--; | |
1354 | return res; | |
1355 | } | |
1356 | ||
3ddcd056 LT |
1357 | /* |
1358 | * We really don't want to look at inode->i_op->lookup | |
1359 | * when we don't have to. So we keep a cache bit in | |
1360 | * the inode ->i_opflags field that says "yes, we can | |
1361 | * do lookup on this inode". | |
1362 | */ | |
1363 | static inline int can_lookup(struct inode *inode) | |
1364 | { | |
1365 | if (likely(inode->i_opflags & IOP_LOOKUP)) | |
1366 | return 1; | |
1367 | if (likely(!inode->i_op->lookup)) | |
1368 | return 0; | |
1369 | ||
1370 | /* We do this once for the lifetime of the inode */ | |
1371 | spin_lock(&inode->i_lock); | |
1372 | inode->i_opflags |= IOP_LOOKUP; | |
1373 | spin_unlock(&inode->i_lock); | |
1374 | return 1; | |
1375 | } | |
1376 | ||
bfcfaa77 LT |
1377 | /* |
1378 | * We can do the critical dentry name comparison and hashing | |
1379 | * operations one word at a time, but we are limited to: | |
1380 | * | |
1381 | * - Architectures with fast unaligned word accesses. We could | |
1382 | * do a "get_unaligned()" if this helps and is sufficiently | |
1383 | * fast. | |
1384 | * | |
1385 | * - Little-endian machines (so that we can generate the mask | |
1386 | * of low bytes efficiently). Again, we *could* do a byte | |
1387 | * swapping load on big-endian architectures if that is not | |
1388 | * expensive enough to make the optimization worthless. | |
1389 | * | |
1390 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we | |
1391 | * do not trap on the (extremely unlikely) case of a page | |
1392 | * crossing operation. | |
1393 | * | |
1394 | * - Furthermore, we need an efficient 64-bit compile for the | |
1395 | * 64-bit case in order to generate the "number of bytes in | |
1396 | * the final mask". Again, that could be replaced with a | |
1397 | * efficient population count instruction or similar. | |
1398 | */ | |
1399 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1400 | ||
1401 | #ifdef CONFIG_64BIT | |
1402 | ||
1403 | /* | |
1404 | * Jan Achrenius on G+: microoptimized version of | |
1405 | * the simpler "(mask & ONEBYTES) * ONEBYTES >> 56" | |
1406 | * that works for the bytemasks without having to | |
1407 | * mask them first. | |
1408 | */ | |
1409 | static inline long count_masked_bytes(unsigned long mask) | |
1410 | { | |
1411 | return mask*0x0001020304050608 >> 56; | |
1412 | } | |
1413 | ||
1414 | static inline unsigned int fold_hash(unsigned long hash) | |
1415 | { | |
1416 | hash += hash >> (8*sizeof(int)); | |
1417 | return hash; | |
1418 | } | |
1419 | ||
1420 | #else /* 32-bit case */ | |
1421 | ||
1422 | /* Carl Chatfield / Jan Achrenius G+ version for 32-bit */ | |
1423 | static inline long count_masked_bytes(long mask) | |
1424 | { | |
1425 | /* (000000 0000ff 00ffff ffffff) -> ( 1 1 2 3 ) */ | |
1426 | long a = (0x0ff0001+mask) >> 23; | |
1427 | /* Fix the 1 for 00 case */ | |
1428 | return a & mask; | |
1429 | } | |
1430 | ||
1431 | #define fold_hash(x) (x) | |
1432 | ||
1433 | #endif | |
1434 | ||
1435 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1436 | { | |
1437 | unsigned long a, mask; | |
1438 | unsigned long hash = 0; | |
1439 | ||
1440 | for (;;) { | |
1441 | a = *(unsigned long *)name; | |
1442 | hash *= 9; | |
1443 | if (len < sizeof(unsigned long)) | |
1444 | break; | |
1445 | hash += a; | |
1446 | name += sizeof(unsigned long); | |
1447 | len -= sizeof(unsigned long); | |
1448 | if (!len) | |
1449 | goto done; | |
1450 | } | |
1451 | mask = ~(~0ul << len*8); | |
1452 | hash += mask & a; | |
1453 | done: | |
1454 | return fold_hash(hash); | |
1455 | } | |
1456 | EXPORT_SYMBOL(full_name_hash); | |
1457 | ||
1de5b41c | 1458 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1459 | #define ONEBYTES 0x0101010101010101ul |
1460 | #define SLASHBYTES 0x2f2f2f2f2f2f2f2ful | |
1461 | #define HIGHBITS 0x8080808080808080ul | |
1de5b41c AM |
1462 | #else |
1463 | #define ONEBYTES 0x01010101ul | |
1464 | #define SLASHBYTES 0x2f2f2f2ful | |
1465 | #define HIGHBITS 0x80808080ul | |
1466 | #endif | |
bfcfaa77 LT |
1467 | |
1468 | /* Return the high bit set in the first byte that is a zero */ | |
1469 | static inline unsigned long has_zero(unsigned long a) | |
1470 | { | |
1471 | return ((a - ONEBYTES) & ~a) & HIGHBITS; | |
1472 | } | |
1473 | ||
1474 | /* | |
1475 | * Calculate the length and hash of the path component, and | |
1476 | * return the length of the component; | |
1477 | */ | |
1478 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1479 | { | |
1480 | unsigned long a, mask, hash, len; | |
1481 | ||
1482 | hash = a = 0; | |
1483 | len = -sizeof(unsigned long); | |
1484 | do { | |
1485 | hash = (hash + a) * 9; | |
1486 | len += sizeof(unsigned long); | |
1487 | a = *(unsigned long *)(name+len); | |
1488 | /* Do we have any NUL or '/' bytes in this word? */ | |
1489 | mask = has_zero(a) | has_zero(a ^ SLASHBYTES); | |
1490 | } while (!mask); | |
1491 | ||
1492 | /* The mask *below* the first high bit set */ | |
1493 | mask = (mask - 1) & ~mask; | |
1494 | mask >>= 7; | |
1495 | hash += a & mask; | |
1496 | *hashp = fold_hash(hash); | |
1497 | ||
1498 | return len + count_masked_bytes(mask); | |
1499 | } | |
1500 | ||
1501 | #else | |
1502 | ||
0145acc2 LT |
1503 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1504 | { | |
1505 | unsigned long hash = init_name_hash(); | |
1506 | while (len--) | |
1507 | hash = partial_name_hash(*name++, hash); | |
1508 | return end_name_hash(hash); | |
1509 | } | |
ae942ae7 | 1510 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1511 | |
200e9ef7 LT |
1512 | /* |
1513 | * We know there's a real path component here of at least | |
1514 | * one character. | |
1515 | */ | |
1516 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1517 | { | |
1518 | unsigned long hash = init_name_hash(); | |
1519 | unsigned long len = 0, c; | |
1520 | ||
1521 | c = (unsigned char)*name; | |
1522 | do { | |
1523 | len++; | |
1524 | hash = partial_name_hash(c, hash); | |
1525 | c = (unsigned char)name[len]; | |
1526 | } while (c && c != '/'); | |
1527 | *hashp = end_name_hash(hash); | |
1528 | return len; | |
1529 | } | |
1530 | ||
bfcfaa77 LT |
1531 | #endif |
1532 | ||
1da177e4 LT |
1533 | /* |
1534 | * Name resolution. | |
ea3834d9 PM |
1535 | * This is the basic name resolution function, turning a pathname into |
1536 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1537 | * |
ea3834d9 PM |
1538 | * Returns 0 and nd will have valid dentry and mnt on success. |
1539 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1540 | */ |
6de88d72 | 1541 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1542 | { |
1543 | struct path next; | |
1da177e4 | 1544 | int err; |
1da177e4 LT |
1545 | |
1546 | while (*name=='/') | |
1547 | name++; | |
1548 | if (!*name) | |
086e183a | 1549 | return 0; |
1da177e4 | 1550 | |
1da177e4 LT |
1551 | /* At this point we know we have a real path component. */ |
1552 | for(;;) { | |
1da177e4 | 1553 | struct qstr this; |
200e9ef7 | 1554 | long len; |
fe479a58 | 1555 | int type; |
1da177e4 | 1556 | |
52094c8a | 1557 | err = may_lookup(nd); |
1da177e4 LT |
1558 | if (err) |
1559 | break; | |
1560 | ||
200e9ef7 | 1561 | len = hash_name(name, &this.hash); |
1da177e4 | 1562 | this.name = name; |
200e9ef7 | 1563 | this.len = len; |
1da177e4 | 1564 | |
fe479a58 | 1565 | type = LAST_NORM; |
200e9ef7 | 1566 | if (name[0] == '.') switch (len) { |
fe479a58 | 1567 | case 2: |
200e9ef7 | 1568 | if (name[1] == '.') { |
fe479a58 | 1569 | type = LAST_DOTDOT; |
16c2cd71 AV |
1570 | nd->flags |= LOOKUP_JUMPED; |
1571 | } | |
fe479a58 AV |
1572 | break; |
1573 | case 1: | |
1574 | type = LAST_DOT; | |
1575 | } | |
5a202bcd AV |
1576 | if (likely(type == LAST_NORM)) { |
1577 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1578 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd AV |
1579 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
1580 | err = parent->d_op->d_hash(parent, nd->inode, | |
1581 | &this); | |
1582 | if (err < 0) | |
1583 | break; | |
1584 | } | |
1585 | } | |
fe479a58 | 1586 | |
200e9ef7 | 1587 | if (!name[len]) |
1da177e4 | 1588 | goto last_component; |
200e9ef7 LT |
1589 | /* |
1590 | * If it wasn't NUL, we know it was '/'. Skip that | |
1591 | * slash, and continue until no more slashes. | |
1592 | */ | |
1593 | do { | |
1594 | len++; | |
1595 | } while (unlikely(name[len] == '/')); | |
1596 | if (!name[len]) | |
b356379a | 1597 | goto last_component; |
200e9ef7 | 1598 | name += len; |
1da177e4 | 1599 | |
ce57dfc1 AV |
1600 | err = walk_component(nd, &next, &this, type, LOOKUP_FOLLOW); |
1601 | if (err < 0) | |
1602 | return err; | |
1da177e4 | 1603 | |
ce57dfc1 | 1604 | if (err) { |
b356379a | 1605 | err = nested_symlink(&next, nd); |
1da177e4 | 1606 | if (err) |
a7472bab | 1607 | return err; |
31e6b01f | 1608 | } |
3ddcd056 LT |
1609 | if (can_lookup(nd->inode)) |
1610 | continue; | |
1da177e4 | 1611 | err = -ENOTDIR; |
3ddcd056 | 1612 | break; |
1da177e4 LT |
1613 | /* here ends the main loop */ |
1614 | ||
1da177e4 | 1615 | last_component: |
b356379a AV |
1616 | nd->last = this; |
1617 | nd->last_type = type; | |
086e183a | 1618 | return 0; |
1da177e4 | 1619 | } |
951361f9 | 1620 | terminate_walk(nd); |
1da177e4 LT |
1621 | return err; |
1622 | } | |
1623 | ||
70e9b357 AV |
1624 | static int path_init(int dfd, const char *name, unsigned int flags, |
1625 | struct nameidata *nd, struct file **fp) | |
31e6b01f NP |
1626 | { |
1627 | int retval = 0; | |
1628 | int fput_needed; | |
1629 | struct file *file; | |
1630 | ||
1631 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
16c2cd71 | 1632 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 1633 | nd->depth = 0; |
5b6ca027 AV |
1634 | if (flags & LOOKUP_ROOT) { |
1635 | struct inode *inode = nd->root.dentry->d_inode; | |
73d049a4 AV |
1636 | if (*name) { |
1637 | if (!inode->i_op->lookup) | |
1638 | return -ENOTDIR; | |
1639 | retval = inode_permission(inode, MAY_EXEC); | |
1640 | if (retval) | |
1641 | return retval; | |
1642 | } | |
5b6ca027 AV |
1643 | nd->path = nd->root; |
1644 | nd->inode = inode; | |
1645 | if (flags & LOOKUP_RCU) { | |
1646 | br_read_lock(vfsmount_lock); | |
1647 | rcu_read_lock(); | |
1648 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1649 | } else { | |
1650 | path_get(&nd->path); | |
1651 | } | |
1652 | return 0; | |
1653 | } | |
1654 | ||
31e6b01f | 1655 | nd->root.mnt = NULL; |
31e6b01f NP |
1656 | |
1657 | if (*name=='/') { | |
e41f7d4e AV |
1658 | if (flags & LOOKUP_RCU) { |
1659 | br_read_lock(vfsmount_lock); | |
1660 | rcu_read_lock(); | |
1661 | set_root_rcu(nd); | |
1662 | } else { | |
1663 | set_root(nd); | |
1664 | path_get(&nd->root); | |
1665 | } | |
1666 | nd->path = nd->root; | |
31e6b01f | 1667 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1668 | if (flags & LOOKUP_RCU) { |
1669 | struct fs_struct *fs = current->fs; | |
1670 | unsigned seq; | |
31e6b01f | 1671 | |
e41f7d4e AV |
1672 | br_read_lock(vfsmount_lock); |
1673 | rcu_read_lock(); | |
c28cc364 | 1674 | |
e41f7d4e AV |
1675 | do { |
1676 | seq = read_seqcount_begin(&fs->seq); | |
1677 | nd->path = fs->pwd; | |
1678 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1679 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1680 | } else { | |
1681 | get_fs_pwd(current->fs, &nd->path); | |
1682 | } | |
31e6b01f NP |
1683 | } else { |
1684 | struct dentry *dentry; | |
1685 | ||
1abf0c71 | 1686 | file = fget_raw_light(dfd, &fput_needed); |
31e6b01f NP |
1687 | retval = -EBADF; |
1688 | if (!file) | |
1689 | goto out_fail; | |
1690 | ||
1691 | dentry = file->f_path.dentry; | |
1692 | ||
f52e0c11 AV |
1693 | if (*name) { |
1694 | retval = -ENOTDIR; | |
1695 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
1696 | goto fput_fail; | |
31e6b01f | 1697 | |
4ad5abb3 | 1698 | retval = inode_permission(dentry->d_inode, MAY_EXEC); |
f52e0c11 AV |
1699 | if (retval) |
1700 | goto fput_fail; | |
1701 | } | |
31e6b01f NP |
1702 | |
1703 | nd->path = file->f_path; | |
e41f7d4e AV |
1704 | if (flags & LOOKUP_RCU) { |
1705 | if (fput_needed) | |
70e9b357 | 1706 | *fp = file; |
e41f7d4e AV |
1707 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
1708 | br_read_lock(vfsmount_lock); | |
1709 | rcu_read_lock(); | |
1710 | } else { | |
1711 | path_get(&file->f_path); | |
1712 | fput_light(file, fput_needed); | |
1713 | } | |
31e6b01f | 1714 | } |
31e6b01f | 1715 | |
31e6b01f | 1716 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1717 | return 0; |
2dfdd266 | 1718 | |
9b4a9b14 AV |
1719 | fput_fail: |
1720 | fput_light(file, fput_needed); | |
1721 | out_fail: | |
1722 | return retval; | |
1723 | } | |
1724 | ||
bd92d7fe AV |
1725 | static inline int lookup_last(struct nameidata *nd, struct path *path) |
1726 | { | |
1727 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1728 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1729 | ||
1730 | nd->flags &= ~LOOKUP_PARENT; | |
1731 | return walk_component(nd, path, &nd->last, nd->last_type, | |
1732 | nd->flags & LOOKUP_FOLLOW); | |
1733 | } | |
1734 | ||
9b4a9b14 | 1735 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
ee0827cd | 1736 | static int path_lookupat(int dfd, const char *name, |
9b4a9b14 AV |
1737 | unsigned int flags, struct nameidata *nd) |
1738 | { | |
70e9b357 | 1739 | struct file *base = NULL; |
bd92d7fe AV |
1740 | struct path path; |
1741 | int err; | |
31e6b01f NP |
1742 | |
1743 | /* | |
1744 | * Path walking is largely split up into 2 different synchronisation | |
1745 | * schemes, rcu-walk and ref-walk (explained in | |
1746 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1747 | * path walk code, but some things particularly setup, cleanup, and | |
1748 | * following mounts are sufficiently divergent that functions are | |
1749 | * duplicated. Typically there is a function foo(), and its RCU | |
1750 | * analogue, foo_rcu(). | |
1751 | * | |
1752 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1753 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1754 | * be handled by restarting a traditional ref-walk (which will always | |
1755 | * be able to complete). | |
1756 | */ | |
bd92d7fe | 1757 | err = path_init(dfd, name, flags | LOOKUP_PARENT, nd, &base); |
ee0827cd | 1758 | |
bd92d7fe AV |
1759 | if (unlikely(err)) |
1760 | return err; | |
ee0827cd AV |
1761 | |
1762 | current->total_link_count = 0; | |
bd92d7fe AV |
1763 | err = link_path_walk(name, nd); |
1764 | ||
1765 | if (!err && !(flags & LOOKUP_PARENT)) { | |
bd92d7fe AV |
1766 | err = lookup_last(nd, &path); |
1767 | while (err > 0) { | |
1768 | void *cookie; | |
1769 | struct path link = path; | |
bd92d7fe | 1770 | nd->flags |= LOOKUP_PARENT; |
574197e0 | 1771 | err = follow_link(&link, nd, &cookie); |
bd92d7fe AV |
1772 | if (!err) |
1773 | err = lookup_last(nd, &path); | |
574197e0 | 1774 | put_link(nd, &link, cookie); |
bd92d7fe AV |
1775 | } |
1776 | } | |
ee0827cd | 1777 | |
9f1fafee AV |
1778 | if (!err) |
1779 | err = complete_walk(nd); | |
bd92d7fe AV |
1780 | |
1781 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
1782 | if (!nd->inode->i_op->lookup) { | |
1783 | path_put(&nd->path); | |
bd23a539 | 1784 | err = -ENOTDIR; |
bd92d7fe AV |
1785 | } |
1786 | } | |
16c2cd71 | 1787 | |
70e9b357 AV |
1788 | if (base) |
1789 | fput(base); | |
ee0827cd | 1790 | |
5b6ca027 | 1791 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
2a737871 AV |
1792 | path_put(&nd->root); |
1793 | nd->root.mnt = NULL; | |
1794 | } | |
bd92d7fe | 1795 | return err; |
ee0827cd | 1796 | } |
31e6b01f | 1797 | |
ee0827cd AV |
1798 | static int do_path_lookup(int dfd, const char *name, |
1799 | unsigned int flags, struct nameidata *nd) | |
1800 | { | |
1801 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); | |
1802 | if (unlikely(retval == -ECHILD)) | |
1803 | retval = path_lookupat(dfd, name, flags, nd); | |
1804 | if (unlikely(retval == -ESTALE)) | |
1805 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); | |
31e6b01f NP |
1806 | |
1807 | if (likely(!retval)) { | |
1808 | if (unlikely(!audit_dummy_context())) { | |
1809 | if (nd->path.dentry && nd->inode) | |
1810 | audit_inode(name, nd->path.dentry); | |
1811 | } | |
1812 | } | |
170aa3d0 | 1813 | return retval; |
1da177e4 LT |
1814 | } |
1815 | ||
c9c6cac0 | 1816 | int kern_path_parent(const char *name, struct nameidata *nd) |
5590ff0d | 1817 | { |
c9c6cac0 | 1818 | return do_path_lookup(AT_FDCWD, name, LOOKUP_PARENT, nd); |
5590ff0d UD |
1819 | } |
1820 | ||
d1811465 AV |
1821 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1822 | { | |
1823 | struct nameidata nd; | |
1824 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1825 | if (!res) | |
1826 | *path = nd.path; | |
1827 | return res; | |
1828 | } | |
1829 | ||
16f18200 JJS |
1830 | /** |
1831 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1832 | * @dentry: pointer to dentry of the base directory | |
1833 | * @mnt: pointer to vfs mount of the base directory | |
1834 | * @name: pointer to file name | |
1835 | * @flags: lookup flags | |
e0a01249 | 1836 | * @path: pointer to struct path to fill |
16f18200 JJS |
1837 | */ |
1838 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1839 | const char *name, unsigned int flags, | |
e0a01249 | 1840 | struct path *path) |
16f18200 | 1841 | { |
e0a01249 AV |
1842 | struct nameidata nd; |
1843 | int err; | |
1844 | nd.root.dentry = dentry; | |
1845 | nd.root.mnt = mnt; | |
1846 | BUG_ON(flags & LOOKUP_PARENT); | |
5b6ca027 | 1847 | /* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */ |
e0a01249 AV |
1848 | err = do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, &nd); |
1849 | if (!err) | |
1850 | *path = nd.path; | |
1851 | return err; | |
16f18200 JJS |
1852 | } |
1853 | ||
eead1911 CH |
1854 | static struct dentry *__lookup_hash(struct qstr *name, |
1855 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1856 | { |
81fca444 | 1857 | struct inode *inode = base->d_inode; |
057f6c01 | 1858 | struct dentry *dentry; |
1da177e4 LT |
1859 | int err; |
1860 | ||
4ad5abb3 | 1861 | err = inode_permission(inode, MAY_EXEC); |
81fca444 CH |
1862 | if (err) |
1863 | return ERR_PTR(err); | |
1da177e4 | 1864 | |
b04f784e NP |
1865 | /* |
1866 | * Don't bother with __d_lookup: callers are for creat as | |
1867 | * well as unlink, so a lot of the time it would cost | |
1868 | * a double lookup. | |
6e6b1bd1 | 1869 | */ |
b04f784e | 1870 | dentry = d_lookup(base, name); |
6e6b1bd1 | 1871 | |
44396f4b JB |
1872 | if (dentry && d_need_lookup(dentry)) { |
1873 | /* | |
1874 | * __lookup_hash is called with the parent dir's i_mutex already | |
1875 | * held, so we are good to go here. | |
1876 | */ | |
1877 | dentry = d_inode_lookup(base, dentry, nd); | |
1878 | if (IS_ERR(dentry)) | |
1879 | return dentry; | |
1880 | } | |
1881 | ||
d2d9e9fb AV |
1882 | if (dentry && (dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
1883 | int status = d_revalidate(dentry, nd); | |
1884 | if (unlikely(status <= 0)) { | |
1885 | /* | |
1886 | * The dentry failed validation. | |
1887 | * If d_revalidate returned 0 attempt to invalidate | |
1888 | * the dentry otherwise d_revalidate is asking us | |
1889 | * to return a fail status. | |
1890 | */ | |
1891 | if (status < 0) { | |
1892 | dput(dentry); | |
1893 | return ERR_PTR(status); | |
1894 | } else if (!d_invalidate(dentry)) { | |
1895 | dput(dentry); | |
1896 | dentry = NULL; | |
1897 | } | |
1898 | } | |
1899 | } | |
6e6b1bd1 | 1900 | |
baa03890 NP |
1901 | if (!dentry) |
1902 | dentry = d_alloc_and_lookup(base, name, nd); | |
5a202bcd | 1903 | |
1da177e4 LT |
1904 | return dentry; |
1905 | } | |
1906 | ||
057f6c01 JM |
1907 | /* |
1908 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1909 | * needs parent already locked. Doesn't follow mounts. | |
1910 | * SMP-safe. | |
1911 | */ | |
eead1911 | 1912 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1913 | { |
4ac91378 | 1914 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1915 | } |
1916 | ||
eead1911 | 1917 | /** |
a6b91919 | 1918 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1919 | * @name: pathname component to lookup |
1920 | * @base: base directory to lookup from | |
1921 | * @len: maximum length @len should be interpreted to | |
1922 | * | |
a6b91919 RD |
1923 | * Note that this routine is purely a helper for filesystem usage and should |
1924 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1925 | * nameidata argument is passed to the filesystem methods and a filesystem |
1926 | * using this helper needs to be prepared for that. | |
1927 | */ | |
057f6c01 JM |
1928 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1929 | { | |
057f6c01 | 1930 | struct qstr this; |
6a96ba54 | 1931 | unsigned int c; |
057f6c01 | 1932 | |
2f9092e1 DW |
1933 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1934 | ||
6a96ba54 AV |
1935 | this.name = name; |
1936 | this.len = len; | |
0145acc2 | 1937 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
1938 | if (!len) |
1939 | return ERR_PTR(-EACCES); | |
1940 | ||
6a96ba54 AV |
1941 | while (len--) { |
1942 | c = *(const unsigned char *)name++; | |
1943 | if (c == '/' || c == '\0') | |
1944 | return ERR_PTR(-EACCES); | |
6a96ba54 | 1945 | } |
5a202bcd AV |
1946 | /* |
1947 | * See if the low-level filesystem might want | |
1948 | * to use its own hash.. | |
1949 | */ | |
1950 | if (base->d_flags & DCACHE_OP_HASH) { | |
1951 | int err = base->d_op->d_hash(base, base->d_inode, &this); | |
1952 | if (err < 0) | |
1953 | return ERR_PTR(err); | |
1954 | } | |
eead1911 | 1955 | |
49705b77 | 1956 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1957 | } |
1958 | ||
1fa1e7f6 AW |
1959 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
1960 | struct path *path, int *empty) | |
1da177e4 | 1961 | { |
2d8f3038 | 1962 | struct nameidata nd; |
1fa1e7f6 | 1963 | char *tmp = getname_flags(name, flags, empty); |
1da177e4 | 1964 | int err = PTR_ERR(tmp); |
1da177e4 | 1965 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1966 | |
1967 | BUG_ON(flags & LOOKUP_PARENT); | |
1968 | ||
1969 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1970 | putname(tmp); |
2d8f3038 AV |
1971 | if (!err) |
1972 | *path = nd.path; | |
1da177e4 LT |
1973 | } |
1974 | return err; | |
1975 | } | |
1976 | ||
1fa1e7f6 AW |
1977 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1978 | struct path *path) | |
1979 | { | |
1980 | return user_path_at_empty(dfd, name, flags, path, 0); | |
1981 | } | |
1982 | ||
2ad94ae6 AV |
1983 | static int user_path_parent(int dfd, const char __user *path, |
1984 | struct nameidata *nd, char **name) | |
1985 | { | |
1986 | char *s = getname(path); | |
1987 | int error; | |
1988 | ||
1989 | if (IS_ERR(s)) | |
1990 | return PTR_ERR(s); | |
1991 | ||
1992 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1993 | if (error) | |
1994 | putname(s); | |
1995 | else | |
1996 | *name = s; | |
1997 | ||
1998 | return error; | |
1999 | } | |
2000 | ||
1da177e4 LT |
2001 | /* |
2002 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
2003 | * minimal. | |
2004 | */ | |
2005 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
2006 | { | |
da9592ed DH |
2007 | uid_t fsuid = current_fsuid(); |
2008 | ||
1da177e4 LT |
2009 | if (!(dir->i_mode & S_ISVTX)) |
2010 | return 0; | |
e795b717 SH |
2011 | if (current_user_ns() != inode_userns(inode)) |
2012 | goto other_userns; | |
da9592ed | 2013 | if (inode->i_uid == fsuid) |
1da177e4 | 2014 | return 0; |
da9592ed | 2015 | if (dir->i_uid == fsuid) |
1da177e4 | 2016 | return 0; |
e795b717 SH |
2017 | |
2018 | other_userns: | |
2019 | return !ns_capable(inode_userns(inode), CAP_FOWNER); | |
1da177e4 LT |
2020 | } |
2021 | ||
2022 | /* | |
2023 | * Check whether we can remove a link victim from directory dir, check | |
2024 | * whether the type of victim is right. | |
2025 | * 1. We can't do it if dir is read-only (done in permission()) | |
2026 | * 2. We should have write and exec permissions on dir | |
2027 | * 3. We can't remove anything from append-only dir | |
2028 | * 4. We can't do anything with immutable dir (done in permission()) | |
2029 | * 5. If the sticky bit on dir is set we should either | |
2030 | * a. be owner of dir, or | |
2031 | * b. be owner of victim, or | |
2032 | * c. have CAP_FOWNER capability | |
2033 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2034 | * links pointing to it. | |
2035 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2036 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2037 | * 9. We can't remove a root or mountpoint. | |
2038 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
2039 | * nfs_async_unlink(). | |
2040 | */ | |
858119e1 | 2041 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
2042 | { |
2043 | int error; | |
2044 | ||
2045 | if (!victim->d_inode) | |
2046 | return -ENOENT; | |
2047 | ||
2048 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 2049 | audit_inode_child(victim, dir); |
1da177e4 | 2050 | |
f419a2e3 | 2051 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2052 | if (error) |
2053 | return error; | |
2054 | if (IS_APPEND(dir)) | |
2055 | return -EPERM; | |
2056 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 2057 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
2058 | return -EPERM; |
2059 | if (isdir) { | |
2060 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
2061 | return -ENOTDIR; | |
2062 | if (IS_ROOT(victim)) | |
2063 | return -EBUSY; | |
2064 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
2065 | return -EISDIR; | |
2066 | if (IS_DEADDIR(dir)) | |
2067 | return -ENOENT; | |
2068 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2069 | return -EBUSY; | |
2070 | return 0; | |
2071 | } | |
2072 | ||
2073 | /* Check whether we can create an object with dentry child in directory | |
2074 | * dir. | |
2075 | * 1. We can't do it if child already exists (open has special treatment for | |
2076 | * this case, but since we are inlined it's OK) | |
2077 | * 2. We can't do it if dir is read-only (done in permission()) | |
2078 | * 3. We should have write and exec permissions on dir | |
2079 | * 4. We can't do it if dir is immutable (done in permission()) | |
2080 | */ | |
a95164d9 | 2081 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
2082 | { |
2083 | if (child->d_inode) | |
2084 | return -EEXIST; | |
2085 | if (IS_DEADDIR(dir)) | |
2086 | return -ENOENT; | |
f419a2e3 | 2087 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2088 | } |
2089 | ||
1da177e4 LT |
2090 | /* |
2091 | * p1 and p2 should be directories on the same fs. | |
2092 | */ | |
2093 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2094 | { | |
2095 | struct dentry *p; | |
2096 | ||
2097 | if (p1 == p2) { | |
f2eace23 | 2098 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2099 | return NULL; |
2100 | } | |
2101 | ||
a11f3a05 | 2102 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2103 | |
e2761a11 OH |
2104 | p = d_ancestor(p2, p1); |
2105 | if (p) { | |
2106 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2107 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2108 | return p; | |
1da177e4 LT |
2109 | } |
2110 | ||
e2761a11 OH |
2111 | p = d_ancestor(p1, p2); |
2112 | if (p) { | |
2113 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2114 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2115 | return p; | |
1da177e4 LT |
2116 | } |
2117 | ||
f2eace23 IM |
2118 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
2119 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
2120 | return NULL; |
2121 | } | |
2122 | ||
2123 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2124 | { | |
1b1dcc1b | 2125 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2126 | if (p1 != p2) { |
1b1dcc1b | 2127 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2128 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2129 | } |
2130 | } | |
2131 | ||
4acdaf27 | 2132 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
1da177e4 LT |
2133 | struct nameidata *nd) |
2134 | { | |
a95164d9 | 2135 | int error = may_create(dir, dentry); |
1da177e4 LT |
2136 | |
2137 | if (error) | |
2138 | return error; | |
2139 | ||
acfa4380 | 2140 | if (!dir->i_op->create) |
1da177e4 LT |
2141 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2142 | mode &= S_IALLUGO; | |
2143 | mode |= S_IFREG; | |
2144 | error = security_inode_create(dir, dentry, mode); | |
2145 | if (error) | |
2146 | return error; | |
1da177e4 | 2147 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 2148 | if (!error) |
f38aa942 | 2149 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2150 | return error; |
2151 | } | |
2152 | ||
73d049a4 | 2153 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2154 | { |
3fb64190 | 2155 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2156 | struct inode *inode = dentry->d_inode; |
2157 | int error; | |
2158 | ||
bcda7652 AV |
2159 | /* O_PATH? */ |
2160 | if (!acc_mode) | |
2161 | return 0; | |
2162 | ||
1da177e4 LT |
2163 | if (!inode) |
2164 | return -ENOENT; | |
2165 | ||
c8fe8f30 CH |
2166 | switch (inode->i_mode & S_IFMT) { |
2167 | case S_IFLNK: | |
1da177e4 | 2168 | return -ELOOP; |
c8fe8f30 CH |
2169 | case S_IFDIR: |
2170 | if (acc_mode & MAY_WRITE) | |
2171 | return -EISDIR; | |
2172 | break; | |
2173 | case S_IFBLK: | |
2174 | case S_IFCHR: | |
3fb64190 | 2175 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2176 | return -EACCES; |
c8fe8f30 CH |
2177 | /*FALLTHRU*/ |
2178 | case S_IFIFO: | |
2179 | case S_IFSOCK: | |
1da177e4 | 2180 | flag &= ~O_TRUNC; |
c8fe8f30 | 2181 | break; |
4a3fd211 | 2182 | } |
b41572e9 | 2183 | |
3fb64190 | 2184 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2185 | if (error) |
2186 | return error; | |
6146f0d5 | 2187 | |
1da177e4 LT |
2188 | /* |
2189 | * An append-only file must be opened in append mode for writing. | |
2190 | */ | |
2191 | if (IS_APPEND(inode)) { | |
8737c930 | 2192 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2193 | return -EPERM; |
1da177e4 | 2194 | if (flag & O_TRUNC) |
7715b521 | 2195 | return -EPERM; |
1da177e4 LT |
2196 | } |
2197 | ||
2198 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2199 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2200 | return -EPERM; |
1da177e4 | 2201 | |
f3c7691e | 2202 | return 0; |
7715b521 | 2203 | } |
1da177e4 | 2204 | |
e1181ee6 | 2205 | static int handle_truncate(struct file *filp) |
7715b521 | 2206 | { |
e1181ee6 | 2207 | struct path *path = &filp->f_path; |
7715b521 AV |
2208 | struct inode *inode = path->dentry->d_inode; |
2209 | int error = get_write_access(inode); | |
2210 | if (error) | |
2211 | return error; | |
2212 | /* | |
2213 | * Refuse to truncate files with mandatory locks held on them. | |
2214 | */ | |
2215 | error = locks_verify_locked(inode); | |
2216 | if (!error) | |
ea0d3ab2 | 2217 | error = security_path_truncate(path); |
7715b521 AV |
2218 | if (!error) { |
2219 | error = do_truncate(path->dentry, 0, | |
2220 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2221 | filp); |
7715b521 AV |
2222 | } |
2223 | put_write_access(inode); | |
acd0c935 | 2224 | return error; |
1da177e4 LT |
2225 | } |
2226 | ||
d57999e1 DH |
2227 | static inline int open_to_namei_flags(int flag) |
2228 | { | |
8a5e929d AV |
2229 | if ((flag & O_ACCMODE) == 3) |
2230 | flag--; | |
d57999e1 DH |
2231 | return flag; |
2232 | } | |
2233 | ||
31e6b01f | 2234 | /* |
fe2d35ff | 2235 | * Handle the last step of open() |
31e6b01f | 2236 | */ |
fb1cc555 | 2237 | static struct file *do_last(struct nameidata *nd, struct path *path, |
c3e380b0 | 2238 | const struct open_flags *op, const char *pathname) |
fb1cc555 | 2239 | { |
a1e28038 | 2240 | struct dentry *dir = nd->path.dentry; |
6c0d46c4 | 2241 | struct dentry *dentry; |
ca344a89 | 2242 | int open_flag = op->open_flag; |
6c0d46c4 | 2243 | int will_truncate = open_flag & O_TRUNC; |
ca344a89 | 2244 | int want_write = 0; |
bcda7652 | 2245 | int acc_mode = op->acc_mode; |
fb1cc555 | 2246 | struct file *filp; |
16c2cd71 | 2247 | int error; |
1f36f774 | 2248 | |
c3e380b0 AV |
2249 | nd->flags &= ~LOOKUP_PARENT; |
2250 | nd->flags |= op->intent; | |
2251 | ||
1f36f774 AV |
2252 | switch (nd->last_type) { |
2253 | case LAST_DOTDOT: | |
176306f5 | 2254 | case LAST_DOT: |
fe2d35ff AV |
2255 | error = handle_dots(nd, nd->last_type); |
2256 | if (error) | |
2257 | return ERR_PTR(error); | |
1f36f774 | 2258 | /* fallthrough */ |
1f36f774 | 2259 | case LAST_ROOT: |
9f1fafee | 2260 | error = complete_walk(nd); |
16c2cd71 | 2261 | if (error) |
9f1fafee | 2262 | return ERR_PTR(error); |
fe2d35ff | 2263 | audit_inode(pathname, nd->path.dentry); |
ca344a89 | 2264 | if (open_flag & O_CREAT) { |
fe2d35ff AV |
2265 | error = -EISDIR; |
2266 | goto exit; | |
2267 | } | |
2268 | goto ok; | |
1f36f774 | 2269 | case LAST_BIND: |
9f1fafee | 2270 | error = complete_walk(nd); |
16c2cd71 | 2271 | if (error) |
9f1fafee | 2272 | return ERR_PTR(error); |
1f36f774 | 2273 | audit_inode(pathname, dir); |
67ee3ad2 | 2274 | goto ok; |
1f36f774 | 2275 | } |
67ee3ad2 | 2276 | |
ca344a89 | 2277 | if (!(open_flag & O_CREAT)) { |
bcda7652 | 2278 | int symlink_ok = 0; |
fe2d35ff AV |
2279 | if (nd->last.name[nd->last.len]) |
2280 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
bcda7652 AV |
2281 | if (open_flag & O_PATH && !(nd->flags & LOOKUP_FOLLOW)) |
2282 | symlink_ok = 1; | |
fe2d35ff | 2283 | /* we _can_ be in RCU mode here */ |
ce57dfc1 AV |
2284 | error = walk_component(nd, path, &nd->last, LAST_NORM, |
2285 | !symlink_ok); | |
2286 | if (error < 0) | |
fe2d35ff | 2287 | return ERR_PTR(error); |
ce57dfc1 | 2288 | if (error) /* symlink */ |
fe2d35ff | 2289 | return NULL; |
fe2d35ff | 2290 | /* sayonara */ |
9f1fafee AV |
2291 | error = complete_walk(nd); |
2292 | if (error) | |
7f6c7e62 | 2293 | return ERR_PTR(error); |
fe2d35ff AV |
2294 | |
2295 | error = -ENOTDIR; | |
2296 | if (nd->flags & LOOKUP_DIRECTORY) { | |
ce57dfc1 | 2297 | if (!nd->inode->i_op->lookup) |
fe2d35ff AV |
2298 | goto exit; |
2299 | } | |
2300 | audit_inode(pathname, nd->path.dentry); | |
2301 | goto ok; | |
2302 | } | |
2303 | ||
2304 | /* create side of things */ | |
a3fbbde7 AV |
2305 | /* |
2306 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED has been | |
2307 | * cleared when we got to the last component we are about to look up | |
2308 | */ | |
9f1fafee AV |
2309 | error = complete_walk(nd); |
2310 | if (error) | |
2311 | return ERR_PTR(error); | |
fe2d35ff AV |
2312 | |
2313 | audit_inode(pathname, dir); | |
16c2cd71 | 2314 | error = -EISDIR; |
1f36f774 | 2315 | /* trailing slashes? */ |
31e6b01f NP |
2316 | if (nd->last.name[nd->last.len]) |
2317 | goto exit; | |
a2c36b45 | 2318 | |
a1e28038 AV |
2319 | mutex_lock(&dir->d_inode->i_mutex); |
2320 | ||
6c0d46c4 AV |
2321 | dentry = lookup_hash(nd); |
2322 | error = PTR_ERR(dentry); | |
2323 | if (IS_ERR(dentry)) { | |
fb1cc555 AV |
2324 | mutex_unlock(&dir->d_inode->i_mutex); |
2325 | goto exit; | |
2326 | } | |
2327 | ||
6c0d46c4 AV |
2328 | path->dentry = dentry; |
2329 | path->mnt = nd->path.mnt; | |
2330 | ||
fb1cc555 | 2331 | /* Negative dentry, just create the file */ |
6c0d46c4 | 2332 | if (!dentry->d_inode) { |
a218d0fd | 2333 | umode_t mode = op->mode; |
6c0d46c4 AV |
2334 | if (!IS_POSIXACL(dir->d_inode)) |
2335 | mode &= ~current_umask(); | |
fb1cc555 AV |
2336 | /* |
2337 | * This write is needed to ensure that a | |
6c0d46c4 | 2338 | * rw->ro transition does not occur between |
fb1cc555 AV |
2339 | * the time when the file is created and when |
2340 | * a permanent write count is taken through | |
2341 | * the 'struct file' in nameidata_to_filp(). | |
2342 | */ | |
2343 | error = mnt_want_write(nd->path.mnt); | |
2344 | if (error) | |
2345 | goto exit_mutex_unlock; | |
ca344a89 | 2346 | want_write = 1; |
9b44f1b3 | 2347 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 2348 | open_flag &= ~O_TRUNC; |
6c0d46c4 | 2349 | will_truncate = 0; |
bcda7652 | 2350 | acc_mode = MAY_OPEN; |
6c0d46c4 AV |
2351 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2352 | if (error) | |
2353 | goto exit_mutex_unlock; | |
2354 | error = vfs_create(dir->d_inode, dentry, mode, nd); | |
2355 | if (error) | |
2356 | goto exit_mutex_unlock; | |
2357 | mutex_unlock(&dir->d_inode->i_mutex); | |
2358 | dput(nd->path.dentry); | |
2359 | nd->path.dentry = dentry; | |
ca344a89 | 2360 | goto common; |
fb1cc555 AV |
2361 | } |
2362 | ||
2363 | /* | |
2364 | * It already exists. | |
2365 | */ | |
2366 | mutex_unlock(&dir->d_inode->i_mutex); | |
2367 | audit_inode(pathname, path->dentry); | |
2368 | ||
2369 | error = -EEXIST; | |
ca344a89 | 2370 | if (open_flag & O_EXCL) |
fb1cc555 AV |
2371 | goto exit_dput; |
2372 | ||
9875cf80 DH |
2373 | error = follow_managed(path, nd->flags); |
2374 | if (error < 0) | |
2375 | goto exit_dput; | |
fb1cc555 | 2376 | |
a3fbbde7 AV |
2377 | if (error) |
2378 | nd->flags |= LOOKUP_JUMPED; | |
2379 | ||
fb1cc555 AV |
2380 | error = -ENOENT; |
2381 | if (!path->dentry->d_inode) | |
2382 | goto exit_dput; | |
9e67f361 AV |
2383 | |
2384 | if (path->dentry->d_inode->i_op->follow_link) | |
fb1cc555 | 2385 | return NULL; |
fb1cc555 AV |
2386 | |
2387 | path_to_nameidata(path, nd); | |
31e6b01f | 2388 | nd->inode = path->dentry->d_inode; |
a3fbbde7 AV |
2389 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
2390 | error = complete_walk(nd); | |
2391 | if (error) | |
097b180c | 2392 | return ERR_PTR(error); |
fb1cc555 | 2393 | error = -EISDIR; |
31e6b01f | 2394 | if (S_ISDIR(nd->inode->i_mode)) |
fb1cc555 | 2395 | goto exit; |
67ee3ad2 | 2396 | ok: |
6c0d46c4 AV |
2397 | if (!S_ISREG(nd->inode->i_mode)) |
2398 | will_truncate = 0; | |
2399 | ||
0f9d1a10 AV |
2400 | if (will_truncate) { |
2401 | error = mnt_want_write(nd->path.mnt); | |
2402 | if (error) | |
2403 | goto exit; | |
ca344a89 | 2404 | want_write = 1; |
0f9d1a10 | 2405 | } |
ca344a89 | 2406 | common: |
bcda7652 | 2407 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 2408 | if (error) |
0f9d1a10 | 2409 | goto exit; |
0f9d1a10 AV |
2410 | filp = nameidata_to_filp(nd); |
2411 | if (!IS_ERR(filp)) { | |
2412 | error = ima_file_check(filp, op->acc_mode); | |
2413 | if (error) { | |
2414 | fput(filp); | |
2415 | filp = ERR_PTR(error); | |
2416 | } | |
2417 | } | |
2418 | if (!IS_ERR(filp)) { | |
2419 | if (will_truncate) { | |
2420 | error = handle_truncate(filp); | |
2421 | if (error) { | |
2422 | fput(filp); | |
2423 | filp = ERR_PTR(error); | |
2424 | } | |
2425 | } | |
2426 | } | |
ca344a89 AV |
2427 | out: |
2428 | if (want_write) | |
0f9d1a10 AV |
2429 | mnt_drop_write(nd->path.mnt); |
2430 | path_put(&nd->path); | |
fb1cc555 AV |
2431 | return filp; |
2432 | ||
2433 | exit_mutex_unlock: | |
2434 | mutex_unlock(&dir->d_inode->i_mutex); | |
2435 | exit_dput: | |
2436 | path_put_conditional(path, nd); | |
2437 | exit: | |
ca344a89 AV |
2438 | filp = ERR_PTR(error); |
2439 | goto out; | |
fb1cc555 AV |
2440 | } |
2441 | ||
13aab428 | 2442 | static struct file *path_openat(int dfd, const char *pathname, |
73d049a4 | 2443 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 2444 | { |
fe2d35ff | 2445 | struct file *base = NULL; |
4a3fd211 | 2446 | struct file *filp; |
9850c056 | 2447 | struct path path; |
13aab428 | 2448 | int error; |
31e6b01f NP |
2449 | |
2450 | filp = get_empty_filp(); | |
2451 | if (!filp) | |
2452 | return ERR_PTR(-ENFILE); | |
2453 | ||
47c805dc | 2454 | filp->f_flags = op->open_flag; |
73d049a4 AV |
2455 | nd->intent.open.file = filp; |
2456 | nd->intent.open.flags = open_to_namei_flags(op->open_flag); | |
2457 | nd->intent.open.create_mode = op->mode; | |
31e6b01f | 2458 | |
73d049a4 | 2459 | error = path_init(dfd, pathname, flags | LOOKUP_PARENT, nd, &base); |
31e6b01f | 2460 | if (unlikely(error)) |
13aab428 | 2461 | goto out_filp; |
31e6b01f | 2462 | |
fe2d35ff | 2463 | current->total_link_count = 0; |
73d049a4 | 2464 | error = link_path_walk(pathname, nd); |
31e6b01f NP |
2465 | if (unlikely(error)) |
2466 | goto out_filp; | |
1da177e4 | 2467 | |
73d049a4 | 2468 | filp = do_last(nd, &path, op, pathname); |
806b681c | 2469 | while (unlikely(!filp)) { /* trailing symlink */ |
7b9337aa | 2470 | struct path link = path; |
def4af30 | 2471 | void *cookie; |
574197e0 | 2472 | if (!(nd->flags & LOOKUP_FOLLOW)) { |
73d049a4 AV |
2473 | path_put_conditional(&path, nd); |
2474 | path_put(&nd->path); | |
40b39136 AV |
2475 | filp = ERR_PTR(-ELOOP); |
2476 | break; | |
2477 | } | |
73d049a4 AV |
2478 | nd->flags |= LOOKUP_PARENT; |
2479 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
574197e0 | 2480 | error = follow_link(&link, nd, &cookie); |
c3e380b0 | 2481 | if (unlikely(error)) |
f1afe9ef | 2482 | filp = ERR_PTR(error); |
c3e380b0 | 2483 | else |
73d049a4 | 2484 | filp = do_last(nd, &path, op, pathname); |
574197e0 | 2485 | put_link(nd, &link, cookie); |
806b681c | 2486 | } |
10fa8e62 | 2487 | out: |
73d049a4 AV |
2488 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) |
2489 | path_put(&nd->root); | |
fe2d35ff AV |
2490 | if (base) |
2491 | fput(base); | |
73d049a4 | 2492 | release_open_intent(nd); |
10fa8e62 | 2493 | return filp; |
1da177e4 | 2494 | |
31e6b01f | 2495 | out_filp: |
806b681c | 2496 | filp = ERR_PTR(error); |
10fa8e62 | 2497 | goto out; |
1da177e4 LT |
2498 | } |
2499 | ||
13aab428 AV |
2500 | struct file *do_filp_open(int dfd, const char *pathname, |
2501 | const struct open_flags *op, int flags) | |
2502 | { | |
73d049a4 | 2503 | struct nameidata nd; |
13aab428 AV |
2504 | struct file *filp; |
2505 | ||
73d049a4 | 2506 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 2507 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 2508 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 2509 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 2510 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
2511 | return filp; |
2512 | } | |
2513 | ||
73d049a4 AV |
2514 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
2515 | const char *name, const struct open_flags *op, int flags) | |
2516 | { | |
2517 | struct nameidata nd; | |
2518 | struct file *file; | |
2519 | ||
2520 | nd.root.mnt = mnt; | |
2521 | nd.root.dentry = dentry; | |
2522 | ||
2523 | flags |= LOOKUP_ROOT; | |
2524 | ||
bcda7652 | 2525 | if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
2526 | return ERR_PTR(-ELOOP); |
2527 | ||
2528 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU); | |
2529 | if (unlikely(file == ERR_PTR(-ECHILD))) | |
2530 | file = path_openat(-1, name, &nd, op, flags); | |
2531 | if (unlikely(file == ERR_PTR(-ESTALE))) | |
2532 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL); | |
2533 | return file; | |
2534 | } | |
2535 | ||
ed75e95d | 2536 | struct dentry *kern_path_create(int dfd, const char *pathname, struct path *path, int is_dir) |
1da177e4 | 2537 | { |
c663e5d8 | 2538 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d AV |
2539 | struct nameidata nd; |
2540 | int error = do_path_lookup(dfd, pathname, LOOKUP_PARENT, &nd); | |
2541 | if (error) | |
2542 | return ERR_PTR(error); | |
1da177e4 | 2543 | |
c663e5d8 CH |
2544 | /* |
2545 | * Yucky last component or no last component at all? | |
2546 | * (foo/., foo/.., /////) | |
2547 | */ | |
ed75e95d AV |
2548 | if (nd.last_type != LAST_NORM) |
2549 | goto out; | |
2550 | nd.flags &= ~LOOKUP_PARENT; | |
2551 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
2552 | nd.intent.open.flags = O_EXCL; | |
c663e5d8 CH |
2553 | |
2554 | /* | |
2555 | * Do the final lookup. | |
2556 | */ | |
ed75e95d AV |
2557 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
2558 | dentry = lookup_hash(&nd); | |
1da177e4 LT |
2559 | if (IS_ERR(dentry)) |
2560 | goto fail; | |
c663e5d8 | 2561 | |
e9baf6e5 AV |
2562 | if (dentry->d_inode) |
2563 | goto eexist; | |
c663e5d8 CH |
2564 | /* |
2565 | * Special case - lookup gave negative, but... we had foo/bar/ | |
2566 | * From the vfs_mknod() POV we just have a negative dentry - | |
2567 | * all is fine. Let's be bastards - you had / on the end, you've | |
2568 | * been asking for (non-existent) directory. -ENOENT for you. | |
2569 | */ | |
ed75e95d | 2570 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
e9baf6e5 AV |
2571 | dput(dentry); |
2572 | dentry = ERR_PTR(-ENOENT); | |
ed75e95d | 2573 | goto fail; |
e9baf6e5 | 2574 | } |
ed75e95d | 2575 | *path = nd.path; |
1da177e4 | 2576 | return dentry; |
e9baf6e5 | 2577 | eexist: |
1da177e4 | 2578 | dput(dentry); |
e9baf6e5 | 2579 | dentry = ERR_PTR(-EEXIST); |
1da177e4 | 2580 | fail: |
ed75e95d AV |
2581 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
2582 | out: | |
2583 | path_put(&nd.path); | |
1da177e4 LT |
2584 | return dentry; |
2585 | } | |
dae6ad8f AV |
2586 | EXPORT_SYMBOL(kern_path_create); |
2587 | ||
2588 | struct dentry *user_path_create(int dfd, const char __user *pathname, struct path *path, int is_dir) | |
2589 | { | |
2590 | char *tmp = getname(pathname); | |
2591 | struct dentry *res; | |
2592 | if (IS_ERR(tmp)) | |
2593 | return ERR_CAST(tmp); | |
2594 | res = kern_path_create(dfd, tmp, path, is_dir); | |
2595 | putname(tmp); | |
2596 | return res; | |
2597 | } | |
2598 | EXPORT_SYMBOL(user_path_create); | |
2599 | ||
1a67aafb | 2600 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 2601 | { |
a95164d9 | 2602 | int error = may_create(dir, dentry); |
1da177e4 LT |
2603 | |
2604 | if (error) | |
2605 | return error; | |
2606 | ||
e795b717 SH |
2607 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && |
2608 | !ns_capable(inode_userns(dir), CAP_MKNOD)) | |
1da177e4 LT |
2609 | return -EPERM; |
2610 | ||
acfa4380 | 2611 | if (!dir->i_op->mknod) |
1da177e4 LT |
2612 | return -EPERM; |
2613 | ||
08ce5f16 SH |
2614 | error = devcgroup_inode_mknod(mode, dev); |
2615 | if (error) | |
2616 | return error; | |
2617 | ||
1da177e4 LT |
2618 | error = security_inode_mknod(dir, dentry, mode, dev); |
2619 | if (error) | |
2620 | return error; | |
2621 | ||
1da177e4 | 2622 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 2623 | if (!error) |
f38aa942 | 2624 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2625 | return error; |
2626 | } | |
2627 | ||
f69aac00 | 2628 | static int may_mknod(umode_t mode) |
463c3197 DH |
2629 | { |
2630 | switch (mode & S_IFMT) { | |
2631 | case S_IFREG: | |
2632 | case S_IFCHR: | |
2633 | case S_IFBLK: | |
2634 | case S_IFIFO: | |
2635 | case S_IFSOCK: | |
2636 | case 0: /* zero mode translates to S_IFREG */ | |
2637 | return 0; | |
2638 | case S_IFDIR: | |
2639 | return -EPERM; | |
2640 | default: | |
2641 | return -EINVAL; | |
2642 | } | |
2643 | } | |
2644 | ||
8208a22b | 2645 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 2646 | unsigned, dev) |
1da177e4 | 2647 | { |
2ad94ae6 | 2648 | struct dentry *dentry; |
dae6ad8f AV |
2649 | struct path path; |
2650 | int error; | |
1da177e4 LT |
2651 | |
2652 | if (S_ISDIR(mode)) | |
2653 | return -EPERM; | |
1da177e4 | 2654 | |
dae6ad8f AV |
2655 | dentry = user_path_create(dfd, filename, &path, 0); |
2656 | if (IS_ERR(dentry)) | |
2657 | return PTR_ERR(dentry); | |
2ad94ae6 | 2658 | |
dae6ad8f | 2659 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 2660 | mode &= ~current_umask(); |
463c3197 DH |
2661 | error = may_mknod(mode); |
2662 | if (error) | |
2663 | goto out_dput; | |
dae6ad8f | 2664 | error = mnt_want_write(path.mnt); |
463c3197 DH |
2665 | if (error) |
2666 | goto out_dput; | |
dae6ad8f | 2667 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 KT |
2668 | if (error) |
2669 | goto out_drop_write; | |
463c3197 | 2670 | switch (mode & S_IFMT) { |
1da177e4 | 2671 | case 0: case S_IFREG: |
dae6ad8f | 2672 | error = vfs_create(path.dentry->d_inode,dentry,mode,NULL); |
1da177e4 LT |
2673 | break; |
2674 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 2675 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2676 | new_decode_dev(dev)); |
2677 | break; | |
2678 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 2679 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2680 | break; |
1da177e4 | 2681 | } |
be6d3e56 | 2682 | out_drop_write: |
dae6ad8f | 2683 | mnt_drop_write(path.mnt); |
463c3197 DH |
2684 | out_dput: |
2685 | dput(dentry); | |
dae6ad8f AV |
2686 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
2687 | path_put(&path); | |
1da177e4 LT |
2688 | |
2689 | return error; | |
2690 | } | |
2691 | ||
8208a22b | 2692 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
2693 | { |
2694 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2695 | } | |
2696 | ||
18bb1db3 | 2697 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 2698 | { |
a95164d9 | 2699 | int error = may_create(dir, dentry); |
1da177e4 LT |
2700 | |
2701 | if (error) | |
2702 | return error; | |
2703 | ||
acfa4380 | 2704 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2705 | return -EPERM; |
2706 | ||
2707 | mode &= (S_IRWXUGO|S_ISVTX); | |
2708 | error = security_inode_mkdir(dir, dentry, mode); | |
2709 | if (error) | |
2710 | return error; | |
2711 | ||
1da177e4 | 2712 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2713 | if (!error) |
f38aa942 | 2714 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2715 | return error; |
2716 | } | |
2717 | ||
a218d0fd | 2718 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 2719 | { |
6902d925 | 2720 | struct dentry *dentry; |
dae6ad8f AV |
2721 | struct path path; |
2722 | int error; | |
1da177e4 | 2723 | |
dae6ad8f | 2724 | dentry = user_path_create(dfd, pathname, &path, 1); |
6902d925 | 2725 | if (IS_ERR(dentry)) |
dae6ad8f | 2726 | return PTR_ERR(dentry); |
1da177e4 | 2727 | |
dae6ad8f | 2728 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 2729 | mode &= ~current_umask(); |
dae6ad8f | 2730 | error = mnt_want_write(path.mnt); |
463c3197 DH |
2731 | if (error) |
2732 | goto out_dput; | |
dae6ad8f | 2733 | error = security_path_mkdir(&path, dentry, mode); |
be6d3e56 KT |
2734 | if (error) |
2735 | goto out_drop_write; | |
dae6ad8f | 2736 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2737 | out_drop_write: |
dae6ad8f | 2738 | mnt_drop_write(path.mnt); |
463c3197 | 2739 | out_dput: |
6902d925 | 2740 | dput(dentry); |
dae6ad8f AV |
2741 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
2742 | path_put(&path); | |
1da177e4 LT |
2743 | return error; |
2744 | } | |
2745 | ||
a218d0fd | 2746 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
2747 | { |
2748 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2749 | } | |
2750 | ||
1da177e4 | 2751 | /* |
a71905f0 SW |
2752 | * The dentry_unhash() helper will try to drop the dentry early: we |
2753 | * should have a usage count of 2 if we're the only user of this | |
2754 | * dentry, and if that is true (possibly after pruning the dcache), | |
2755 | * then we drop the dentry now. | |
1da177e4 LT |
2756 | * |
2757 | * A low-level filesystem can, if it choses, legally | |
2758 | * do a | |
2759 | * | |
2760 | * if (!d_unhashed(dentry)) | |
2761 | * return -EBUSY; | |
2762 | * | |
2763 | * if it cannot handle the case of removing a directory | |
2764 | * that is still in use by something else.. | |
2765 | */ | |
2766 | void dentry_unhash(struct dentry *dentry) | |
2767 | { | |
dc168427 | 2768 | shrink_dcache_parent(dentry); |
1da177e4 | 2769 | spin_lock(&dentry->d_lock); |
64252c75 | 2770 | if (dentry->d_count == 1) |
1da177e4 LT |
2771 | __d_drop(dentry); |
2772 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
2773 | } |
2774 | ||
2775 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2776 | { | |
2777 | int error = may_delete(dir, dentry, 1); | |
2778 | ||
2779 | if (error) | |
2780 | return error; | |
2781 | ||
acfa4380 | 2782 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2783 | return -EPERM; |
2784 | ||
1d2ef590 | 2785 | dget(dentry); |
1b1dcc1b | 2786 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
2787 | |
2788 | error = -EBUSY; | |
1da177e4 | 2789 | if (d_mountpoint(dentry)) |
912dbc15 SW |
2790 | goto out; |
2791 | ||
2792 | error = security_inode_rmdir(dir, dentry); | |
2793 | if (error) | |
2794 | goto out; | |
2795 | ||
3cebde24 | 2796 | shrink_dcache_parent(dentry); |
912dbc15 SW |
2797 | error = dir->i_op->rmdir(dir, dentry); |
2798 | if (error) | |
2799 | goto out; | |
2800 | ||
2801 | dentry->d_inode->i_flags |= S_DEAD; | |
2802 | dont_mount(dentry); | |
2803 | ||
2804 | out: | |
1b1dcc1b | 2805 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 2806 | dput(dentry); |
912dbc15 | 2807 | if (!error) |
1da177e4 | 2808 | d_delete(dentry); |
1da177e4 LT |
2809 | return error; |
2810 | } | |
2811 | ||
5590ff0d | 2812 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2813 | { |
2814 | int error = 0; | |
2815 | char * name; | |
2816 | struct dentry *dentry; | |
2817 | struct nameidata nd; | |
2818 | ||
2ad94ae6 | 2819 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2820 | if (error) |
2ad94ae6 | 2821 | return error; |
1da177e4 LT |
2822 | |
2823 | switch(nd.last_type) { | |
0612d9fb OH |
2824 | case LAST_DOTDOT: |
2825 | error = -ENOTEMPTY; | |
2826 | goto exit1; | |
2827 | case LAST_DOT: | |
2828 | error = -EINVAL; | |
2829 | goto exit1; | |
2830 | case LAST_ROOT: | |
2831 | error = -EBUSY; | |
2832 | goto exit1; | |
1da177e4 | 2833 | } |
0612d9fb OH |
2834 | |
2835 | nd.flags &= ~LOOKUP_PARENT; | |
2836 | ||
4ac91378 | 2837 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2838 | dentry = lookup_hash(&nd); |
1da177e4 | 2839 | error = PTR_ERR(dentry); |
6902d925 DH |
2840 | if (IS_ERR(dentry)) |
2841 | goto exit2; | |
e6bc45d6 TT |
2842 | if (!dentry->d_inode) { |
2843 | error = -ENOENT; | |
2844 | goto exit3; | |
2845 | } | |
0622753b DH |
2846 | error = mnt_want_write(nd.path.mnt); |
2847 | if (error) | |
2848 | goto exit3; | |
be6d3e56 KT |
2849 | error = security_path_rmdir(&nd.path, dentry); |
2850 | if (error) | |
2851 | goto exit4; | |
4ac91378 | 2852 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2853 | exit4: |
0622753b DH |
2854 | mnt_drop_write(nd.path.mnt); |
2855 | exit3: | |
6902d925 DH |
2856 | dput(dentry); |
2857 | exit2: | |
4ac91378 | 2858 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2859 | exit1: |
1d957f9b | 2860 | path_put(&nd.path); |
1da177e4 LT |
2861 | putname(name); |
2862 | return error; | |
2863 | } | |
2864 | ||
3cdad428 | 2865 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2866 | { |
2867 | return do_rmdir(AT_FDCWD, pathname); | |
2868 | } | |
2869 | ||
1da177e4 LT |
2870 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2871 | { | |
2872 | int error = may_delete(dir, dentry, 0); | |
2873 | ||
2874 | if (error) | |
2875 | return error; | |
2876 | ||
acfa4380 | 2877 | if (!dir->i_op->unlink) |
1da177e4 LT |
2878 | return -EPERM; |
2879 | ||
1b1dcc1b | 2880 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2881 | if (d_mountpoint(dentry)) |
2882 | error = -EBUSY; | |
2883 | else { | |
2884 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 2885 | if (!error) { |
1da177e4 | 2886 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 2887 | if (!error) |
d83c49f3 | 2888 | dont_mount(dentry); |
bec1052e | 2889 | } |
1da177e4 | 2890 | } |
1b1dcc1b | 2891 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2892 | |
2893 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2894 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2895 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2896 | d_delete(dentry); |
1da177e4 | 2897 | } |
0eeca283 | 2898 | |
1da177e4 LT |
2899 | return error; |
2900 | } | |
2901 | ||
2902 | /* | |
2903 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2904 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2905 | * writeout happening, and we don't want to prevent access to the directory |
2906 | * while waiting on the I/O. | |
2907 | */ | |
5590ff0d | 2908 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2909 | { |
2ad94ae6 AV |
2910 | int error; |
2911 | char *name; | |
1da177e4 LT |
2912 | struct dentry *dentry; |
2913 | struct nameidata nd; | |
2914 | struct inode *inode = NULL; | |
2915 | ||
2ad94ae6 | 2916 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2917 | if (error) |
2ad94ae6 AV |
2918 | return error; |
2919 | ||
1da177e4 LT |
2920 | error = -EISDIR; |
2921 | if (nd.last_type != LAST_NORM) | |
2922 | goto exit1; | |
0612d9fb OH |
2923 | |
2924 | nd.flags &= ~LOOKUP_PARENT; | |
2925 | ||
4ac91378 | 2926 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2927 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2928 | error = PTR_ERR(dentry); |
2929 | if (!IS_ERR(dentry)) { | |
2930 | /* Why not before? Because we want correct error value */ | |
50338b88 TE |
2931 | if (nd.last.name[nd.last.len]) |
2932 | goto slashes; | |
1da177e4 | 2933 | inode = dentry->d_inode; |
50338b88 | 2934 | if (!inode) |
e6bc45d6 TT |
2935 | goto slashes; |
2936 | ihold(inode); | |
0622753b DH |
2937 | error = mnt_want_write(nd.path.mnt); |
2938 | if (error) | |
2939 | goto exit2; | |
be6d3e56 KT |
2940 | error = security_path_unlink(&nd.path, dentry); |
2941 | if (error) | |
2942 | goto exit3; | |
4ac91378 | 2943 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2944 | exit3: |
0622753b | 2945 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2946 | exit2: |
2947 | dput(dentry); | |
2948 | } | |
4ac91378 | 2949 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2950 | if (inode) |
2951 | iput(inode); /* truncate the inode here */ | |
2952 | exit1: | |
1d957f9b | 2953 | path_put(&nd.path); |
1da177e4 LT |
2954 | putname(name); |
2955 | return error; | |
2956 | ||
2957 | slashes: | |
2958 | error = !dentry->d_inode ? -ENOENT : | |
2959 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2960 | goto exit2; | |
2961 | } | |
2962 | ||
2e4d0924 | 2963 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2964 | { |
2965 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2966 | return -EINVAL; | |
2967 | ||
2968 | if (flag & AT_REMOVEDIR) | |
2969 | return do_rmdir(dfd, pathname); | |
2970 | ||
2971 | return do_unlinkat(dfd, pathname); | |
2972 | } | |
2973 | ||
3480b257 | 2974 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2975 | { |
2976 | return do_unlinkat(AT_FDCWD, pathname); | |
2977 | } | |
2978 | ||
db2e747b | 2979 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2980 | { |
a95164d9 | 2981 | int error = may_create(dir, dentry); |
1da177e4 LT |
2982 | |
2983 | if (error) | |
2984 | return error; | |
2985 | ||
acfa4380 | 2986 | if (!dir->i_op->symlink) |
1da177e4 LT |
2987 | return -EPERM; |
2988 | ||
2989 | error = security_inode_symlink(dir, dentry, oldname); | |
2990 | if (error) | |
2991 | return error; | |
2992 | ||
1da177e4 | 2993 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2994 | if (!error) |
f38aa942 | 2995 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2996 | return error; |
2997 | } | |
2998 | ||
2e4d0924 HC |
2999 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
3000 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3001 | { |
2ad94ae6 AV |
3002 | int error; |
3003 | char *from; | |
6902d925 | 3004 | struct dentry *dentry; |
dae6ad8f | 3005 | struct path path; |
1da177e4 LT |
3006 | |
3007 | from = getname(oldname); | |
2ad94ae6 | 3008 | if (IS_ERR(from)) |
1da177e4 | 3009 | return PTR_ERR(from); |
1da177e4 | 3010 | |
dae6ad8f | 3011 | dentry = user_path_create(newdfd, newname, &path, 0); |
6902d925 DH |
3012 | error = PTR_ERR(dentry); |
3013 | if (IS_ERR(dentry)) | |
dae6ad8f | 3014 | goto out_putname; |
6902d925 | 3015 | |
dae6ad8f | 3016 | error = mnt_want_write(path.mnt); |
75c3f29d DH |
3017 | if (error) |
3018 | goto out_dput; | |
dae6ad8f | 3019 | error = security_path_symlink(&path, dentry, from); |
be6d3e56 KT |
3020 | if (error) |
3021 | goto out_drop_write; | |
dae6ad8f | 3022 | error = vfs_symlink(path.dentry->d_inode, dentry, from); |
be6d3e56 | 3023 | out_drop_write: |
dae6ad8f | 3024 | mnt_drop_write(path.mnt); |
75c3f29d | 3025 | out_dput: |
6902d925 | 3026 | dput(dentry); |
dae6ad8f AV |
3027 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
3028 | path_put(&path); | |
6902d925 | 3029 | out_putname: |
1da177e4 LT |
3030 | putname(from); |
3031 | return error; | |
3032 | } | |
3033 | ||
3480b257 | 3034 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3035 | { |
3036 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
3037 | } | |
3038 | ||
1da177e4 LT |
3039 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
3040 | { | |
3041 | struct inode *inode = old_dentry->d_inode; | |
3042 | int error; | |
3043 | ||
3044 | if (!inode) | |
3045 | return -ENOENT; | |
3046 | ||
a95164d9 | 3047 | error = may_create(dir, new_dentry); |
1da177e4 LT |
3048 | if (error) |
3049 | return error; | |
3050 | ||
3051 | if (dir->i_sb != inode->i_sb) | |
3052 | return -EXDEV; | |
3053 | ||
3054 | /* | |
3055 | * A link to an append-only or immutable file cannot be created. | |
3056 | */ | |
3057 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
3058 | return -EPERM; | |
acfa4380 | 3059 | if (!dir->i_op->link) |
1da177e4 | 3060 | return -EPERM; |
7e79eedb | 3061 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
3062 | return -EPERM; |
3063 | ||
3064 | error = security_inode_link(old_dentry, dir, new_dentry); | |
3065 | if (error) | |
3066 | return error; | |
3067 | ||
7e79eedb | 3068 | mutex_lock(&inode->i_mutex); |
aae8a97d AK |
3069 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
3070 | if (inode->i_nlink == 0) | |
3071 | error = -ENOENT; | |
3072 | else | |
3073 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 3074 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 3075 | if (!error) |
7e79eedb | 3076 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
3077 | return error; |
3078 | } | |
3079 | ||
3080 | /* | |
3081 | * Hardlinks are often used in delicate situations. We avoid | |
3082 | * security-related surprises by not following symlinks on the | |
3083 | * newname. --KAB | |
3084 | * | |
3085 | * We don't follow them on the oldname either to be compatible | |
3086 | * with linux 2.0, and to avoid hard-linking to directories | |
3087 | * and other special files. --ADM | |
3088 | */ | |
2e4d0924 HC |
3089 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
3090 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
3091 | { |
3092 | struct dentry *new_dentry; | |
dae6ad8f | 3093 | struct path old_path, new_path; |
11a7b371 | 3094 | int how = 0; |
1da177e4 | 3095 | int error; |
1da177e4 | 3096 | |
11a7b371 | 3097 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 3098 | return -EINVAL; |
11a7b371 AK |
3099 | /* |
3100 | * To use null names we require CAP_DAC_READ_SEARCH | |
3101 | * This ensures that not everyone will be able to create | |
3102 | * handlink using the passed filedescriptor. | |
3103 | */ | |
3104 | if (flags & AT_EMPTY_PATH) { | |
3105 | if (!capable(CAP_DAC_READ_SEARCH)) | |
3106 | return -ENOENT; | |
3107 | how = LOOKUP_EMPTY; | |
3108 | } | |
3109 | ||
3110 | if (flags & AT_SYMLINK_FOLLOW) | |
3111 | how |= LOOKUP_FOLLOW; | |
c04030e1 | 3112 | |
11a7b371 | 3113 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 3114 | if (error) |
2ad94ae6 AV |
3115 | return error; |
3116 | ||
dae6ad8f | 3117 | new_dentry = user_path_create(newdfd, newname, &new_path, 0); |
1da177e4 | 3118 | error = PTR_ERR(new_dentry); |
6902d925 | 3119 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
3120 | goto out; |
3121 | ||
3122 | error = -EXDEV; | |
3123 | if (old_path.mnt != new_path.mnt) | |
3124 | goto out_dput; | |
3125 | error = mnt_want_write(new_path.mnt); | |
75c3f29d DH |
3126 | if (error) |
3127 | goto out_dput; | |
dae6ad8f | 3128 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 KT |
3129 | if (error) |
3130 | goto out_drop_write; | |
dae6ad8f | 3131 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry); |
be6d3e56 | 3132 | out_drop_write: |
dae6ad8f | 3133 | mnt_drop_write(new_path.mnt); |
75c3f29d | 3134 | out_dput: |
6902d925 | 3135 | dput(new_dentry); |
dae6ad8f AV |
3136 | mutex_unlock(&new_path.dentry->d_inode->i_mutex); |
3137 | path_put(&new_path); | |
1da177e4 | 3138 | out: |
2d8f3038 | 3139 | path_put(&old_path); |
1da177e4 LT |
3140 | |
3141 | return error; | |
3142 | } | |
3143 | ||
3480b257 | 3144 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 3145 | { |
c04030e1 | 3146 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
3147 | } |
3148 | ||
1da177e4 LT |
3149 | /* |
3150 | * The worst of all namespace operations - renaming directory. "Perverted" | |
3151 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
3152 | * Problems: | |
3153 | * a) we can get into loop creation. Check is done in is_subdir(). | |
3154 | * b) race potential - two innocent renames can create a loop together. | |
3155 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 3156 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
3157 | * story. |
3158 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 3159 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
3160 | * whether the target exists). Solution: try to be smart with locking |
3161 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 3162 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
3163 | * move will be locked. Thus we can rank directories by the tree |
3164 | * (ancestors first) and rank all non-directories after them. | |
3165 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 3166 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
3167 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
3168 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
3169 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 3170 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 3171 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 3172 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 3173 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
3174 | * locking]. |
3175 | */ | |
75c96f85 AB |
3176 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
3177 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3178 | { |
3179 | int error = 0; | |
9055cba7 | 3180 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3181 | |
3182 | /* | |
3183 | * If we are going to change the parent - check write permissions, | |
3184 | * we'll need to flip '..'. | |
3185 | */ | |
3186 | if (new_dir != old_dir) { | |
f419a2e3 | 3187 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3188 | if (error) |
3189 | return error; | |
3190 | } | |
3191 | ||
3192 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3193 | if (error) | |
3194 | return error; | |
3195 | ||
1d2ef590 | 3196 | dget(new_dentry); |
d83c49f3 | 3197 | if (target) |
1b1dcc1b | 3198 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
3199 | |
3200 | error = -EBUSY; | |
3201 | if (d_mountpoint(old_dentry) || d_mountpoint(new_dentry)) | |
3202 | goto out; | |
3203 | ||
3cebde24 SW |
3204 | if (target) |
3205 | shrink_dcache_parent(new_dentry); | |
9055cba7 SW |
3206 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
3207 | if (error) | |
3208 | goto out; | |
3209 | ||
1da177e4 | 3210 | if (target) { |
9055cba7 SW |
3211 | target->i_flags |= S_DEAD; |
3212 | dont_mount(new_dentry); | |
1da177e4 | 3213 | } |
9055cba7 SW |
3214 | out: |
3215 | if (target) | |
3216 | mutex_unlock(&target->i_mutex); | |
1d2ef590 | 3217 | dput(new_dentry); |
e31e14ec | 3218 | if (!error) |
349457cc MF |
3219 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3220 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3221 | return error; |
3222 | } | |
3223 | ||
75c96f85 AB |
3224 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3225 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 | 3226 | { |
51892bbb | 3227 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3228 | int error; |
3229 | ||
3230 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3231 | if (error) | |
3232 | return error; | |
3233 | ||
3234 | dget(new_dentry); | |
1da177e4 | 3235 | if (target) |
1b1dcc1b | 3236 | mutex_lock(&target->i_mutex); |
51892bbb SW |
3237 | |
3238 | error = -EBUSY; | |
1da177e4 | 3239 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
51892bbb SW |
3240 | goto out; |
3241 | ||
3242 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3243 | if (error) | |
3244 | goto out; | |
3245 | ||
3246 | if (target) | |
3247 | dont_mount(new_dentry); | |
3248 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) | |
3249 | d_move(old_dentry, new_dentry); | |
3250 | out: | |
1da177e4 | 3251 | if (target) |
1b1dcc1b | 3252 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3253 | dput(new_dentry); |
3254 | return error; | |
3255 | } | |
3256 | ||
3257 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3258 | struct inode *new_dir, struct dentry *new_dentry) | |
3259 | { | |
3260 | int error; | |
3261 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3262 | const unsigned char *old_name; |
1da177e4 LT |
3263 | |
3264 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3265 | return 0; | |
3266 | ||
3267 | error = may_delete(old_dir, old_dentry, is_dir); | |
3268 | if (error) | |
3269 | return error; | |
3270 | ||
3271 | if (!new_dentry->d_inode) | |
a95164d9 | 3272 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3273 | else |
3274 | error = may_delete(new_dir, new_dentry, is_dir); | |
3275 | if (error) | |
3276 | return error; | |
3277 | ||
acfa4380 | 3278 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3279 | return -EPERM; |
3280 | ||
0eeca283 RL |
3281 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3282 | ||
1da177e4 LT |
3283 | if (is_dir) |
3284 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3285 | else | |
3286 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3287 | if (!error) |
3288 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3289 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3290 | fsnotify_oldname_free(old_name); |
3291 | ||
1da177e4 LT |
3292 | return error; |
3293 | } | |
3294 | ||
2e4d0924 HC |
3295 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3296 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3297 | { |
2ad94ae6 AV |
3298 | struct dentry *old_dir, *new_dir; |
3299 | struct dentry *old_dentry, *new_dentry; | |
3300 | struct dentry *trap; | |
1da177e4 | 3301 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3302 | char *from; |
3303 | char *to; | |
3304 | int error; | |
1da177e4 | 3305 | |
2ad94ae6 | 3306 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3307 | if (error) |
3308 | goto exit; | |
3309 | ||
2ad94ae6 | 3310 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3311 | if (error) |
3312 | goto exit1; | |
3313 | ||
3314 | error = -EXDEV; | |
4ac91378 | 3315 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3316 | goto exit2; |
3317 | ||
4ac91378 | 3318 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3319 | error = -EBUSY; |
3320 | if (oldnd.last_type != LAST_NORM) | |
3321 | goto exit2; | |
3322 | ||
4ac91378 | 3323 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3324 | if (newnd.last_type != LAST_NORM) |
3325 | goto exit2; | |
3326 | ||
0612d9fb OH |
3327 | oldnd.flags &= ~LOOKUP_PARENT; |
3328 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3329 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3330 | |
1da177e4 LT |
3331 | trap = lock_rename(new_dir, old_dir); |
3332 | ||
49705b77 | 3333 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3334 | error = PTR_ERR(old_dentry); |
3335 | if (IS_ERR(old_dentry)) | |
3336 | goto exit3; | |
3337 | /* source must exist */ | |
3338 | error = -ENOENT; | |
3339 | if (!old_dentry->d_inode) | |
3340 | goto exit4; | |
3341 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3342 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3343 | error = -ENOTDIR; | |
3344 | if (oldnd.last.name[oldnd.last.len]) | |
3345 | goto exit4; | |
3346 | if (newnd.last.name[newnd.last.len]) | |
3347 | goto exit4; | |
3348 | } | |
3349 | /* source should not be ancestor of target */ | |
3350 | error = -EINVAL; | |
3351 | if (old_dentry == trap) | |
3352 | goto exit4; | |
49705b77 | 3353 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3354 | error = PTR_ERR(new_dentry); |
3355 | if (IS_ERR(new_dentry)) | |
3356 | goto exit4; | |
3357 | /* target should not be an ancestor of source */ | |
3358 | error = -ENOTEMPTY; | |
3359 | if (new_dentry == trap) | |
3360 | goto exit5; | |
3361 | ||
9079b1eb DH |
3362 | error = mnt_want_write(oldnd.path.mnt); |
3363 | if (error) | |
3364 | goto exit5; | |
be6d3e56 KT |
3365 | error = security_path_rename(&oldnd.path, old_dentry, |
3366 | &newnd.path, new_dentry); | |
3367 | if (error) | |
3368 | goto exit6; | |
1da177e4 LT |
3369 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3370 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 3371 | exit6: |
9079b1eb | 3372 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
3373 | exit5: |
3374 | dput(new_dentry); | |
3375 | exit4: | |
3376 | dput(old_dentry); | |
3377 | exit3: | |
3378 | unlock_rename(new_dir, old_dir); | |
3379 | exit2: | |
1d957f9b | 3380 | path_put(&newnd.path); |
2ad94ae6 | 3381 | putname(to); |
1da177e4 | 3382 | exit1: |
1d957f9b | 3383 | path_put(&oldnd.path); |
1da177e4 | 3384 | putname(from); |
2ad94ae6 | 3385 | exit: |
1da177e4 LT |
3386 | return error; |
3387 | } | |
3388 | ||
a26eab24 | 3389 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3390 | { |
3391 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3392 | } | |
3393 | ||
1da177e4 LT |
3394 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3395 | { | |
3396 | int len; | |
3397 | ||
3398 | len = PTR_ERR(link); | |
3399 | if (IS_ERR(link)) | |
3400 | goto out; | |
3401 | ||
3402 | len = strlen(link); | |
3403 | if (len > (unsigned) buflen) | |
3404 | len = buflen; | |
3405 | if (copy_to_user(buffer, link, len)) | |
3406 | len = -EFAULT; | |
3407 | out: | |
3408 | return len; | |
3409 | } | |
3410 | ||
3411 | /* | |
3412 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3413 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3414 | * using) it for any given inode is up to filesystem. | |
3415 | */ | |
3416 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3417 | { | |
3418 | struct nameidata nd; | |
cc314eef | 3419 | void *cookie; |
694a1764 | 3420 | int res; |
cc314eef | 3421 | |
1da177e4 | 3422 | nd.depth = 0; |
cc314eef | 3423 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3424 | if (IS_ERR(cookie)) |
3425 | return PTR_ERR(cookie); | |
3426 | ||
3427 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3428 | if (dentry->d_inode->i_op->put_link) | |
3429 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3430 | return res; | |
1da177e4 LT |
3431 | } |
3432 | ||
3433 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3434 | { | |
3435 | return __vfs_follow_link(nd, link); | |
3436 | } | |
3437 | ||
3438 | /* get the link contents into pagecache */ | |
3439 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3440 | { | |
ebd09abb DG |
3441 | char *kaddr; |
3442 | struct page *page; | |
1da177e4 | 3443 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3444 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3445 | if (IS_ERR(page)) |
6fe6900e | 3446 | return (char*)page; |
1da177e4 | 3447 | *ppage = page; |
ebd09abb DG |
3448 | kaddr = kmap(page); |
3449 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3450 | return kaddr; | |
1da177e4 LT |
3451 | } |
3452 | ||
3453 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3454 | { | |
3455 | struct page *page = NULL; | |
3456 | char *s = page_getlink(dentry, &page); | |
3457 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3458 | if (page) { | |
3459 | kunmap(page); | |
3460 | page_cache_release(page); | |
3461 | } | |
3462 | return res; | |
3463 | } | |
3464 | ||
cc314eef | 3465 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3466 | { |
cc314eef | 3467 | struct page *page = NULL; |
1da177e4 | 3468 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3469 | return page; |
1da177e4 LT |
3470 | } |
3471 | ||
cc314eef | 3472 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3473 | { |
cc314eef LT |
3474 | struct page *page = cookie; |
3475 | ||
3476 | if (page) { | |
1da177e4 LT |
3477 | kunmap(page); |
3478 | page_cache_release(page); | |
1da177e4 LT |
3479 | } |
3480 | } | |
3481 | ||
54566b2c NP |
3482 | /* |
3483 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3484 | */ | |
3485 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3486 | { |
3487 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3488 | struct page *page; |
afddba49 | 3489 | void *fsdata; |
beb497ab | 3490 | int err; |
1da177e4 | 3491 | char *kaddr; |
54566b2c NP |
3492 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3493 | if (nofs) | |
3494 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3495 | |
7e53cac4 | 3496 | retry: |
afddba49 | 3497 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3498 | flags, &page, &fsdata); |
1da177e4 | 3499 | if (err) |
afddba49 NP |
3500 | goto fail; |
3501 | ||
e8e3c3d6 | 3502 | kaddr = kmap_atomic(page); |
1da177e4 | 3503 | memcpy(kaddr, symname, len-1); |
e8e3c3d6 | 3504 | kunmap_atomic(kaddr); |
afddba49 NP |
3505 | |
3506 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3507 | page, fsdata); | |
1da177e4 LT |
3508 | if (err < 0) |
3509 | goto fail; | |
afddba49 NP |
3510 | if (err < len-1) |
3511 | goto retry; | |
3512 | ||
1da177e4 LT |
3513 | mark_inode_dirty(inode); |
3514 | return 0; | |
1da177e4 LT |
3515 | fail: |
3516 | return err; | |
3517 | } | |
3518 | ||
0adb25d2 KK |
3519 | int page_symlink(struct inode *inode, const char *symname, int len) |
3520 | { | |
3521 | return __page_symlink(inode, symname, len, | |
54566b2c | 3522 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3523 | } |
3524 | ||
92e1d5be | 3525 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3526 | .readlink = generic_readlink, |
3527 | .follow_link = page_follow_link_light, | |
3528 | .put_link = page_put_link, | |
3529 | }; | |
3530 | ||
2d8f3038 | 3531 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3532 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3533 | EXPORT_SYMBOL(follow_down); |
3534 | EXPORT_SYMBOL(follow_up); | |
3535 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
3536 | EXPORT_SYMBOL(getname); | |
3537 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3538 | EXPORT_SYMBOL(lookup_one_len); |
3539 | EXPORT_SYMBOL(page_follow_link_light); | |
3540 | EXPORT_SYMBOL(page_put_link); | |
3541 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3542 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3543 | EXPORT_SYMBOL(page_symlink); |
3544 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
d1811465 | 3545 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3546 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3547 | EXPORT_SYMBOL(inode_permission); |
1da177e4 LT |
3548 | EXPORT_SYMBOL(unlock_rename); |
3549 | EXPORT_SYMBOL(vfs_create); | |
3550 | EXPORT_SYMBOL(vfs_follow_link); | |
3551 | EXPORT_SYMBOL(vfs_link); | |
3552 | EXPORT_SYMBOL(vfs_mkdir); | |
3553 | EXPORT_SYMBOL(vfs_mknod); | |
3554 | EXPORT_SYMBOL(generic_permission); | |
3555 | EXPORT_SYMBOL(vfs_readlink); | |
3556 | EXPORT_SYMBOL(vfs_rename); | |
3557 | EXPORT_SYMBOL(vfs_rmdir); | |
3558 | EXPORT_SYMBOL(vfs_symlink); | |
3559 | EXPORT_SYMBOL(vfs_unlink); | |
3560 | EXPORT_SYMBOL(dentry_unhash); | |
3561 | EXPORT_SYMBOL(generic_readlink); |