Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/namei.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * Some corrections by tytso. | |
9 | */ | |
10 | ||
11 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
12 | * lookup logic. | |
13 | */ | |
14 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
15 | */ | |
16 | ||
17 | #include <linux/init.h> | |
630d9c47 | 18 | #include <linux/export.h> |
44696908 | 19 | #include <linux/kernel.h> |
1da177e4 LT |
20 | #include <linux/slab.h> |
21 | #include <linux/fs.h> | |
22 | #include <linux/namei.h> | |
1da177e4 | 23 | #include <linux/pagemap.h> |
0eeca283 | 24 | #include <linux/fsnotify.h> |
1da177e4 LT |
25 | #include <linux/personality.h> |
26 | #include <linux/security.h> | |
6146f0d5 | 27 | #include <linux/ima.h> |
1da177e4 LT |
28 | #include <linux/syscalls.h> |
29 | #include <linux/mount.h> | |
30 | #include <linux/audit.h> | |
16f7e0fe | 31 | #include <linux/capability.h> |
834f2a4a | 32 | #include <linux/file.h> |
5590ff0d | 33 | #include <linux/fcntl.h> |
08ce5f16 | 34 | #include <linux/device_cgroup.h> |
5ad4e53b | 35 | #include <linux/fs_struct.h> |
e77819e5 | 36 | #include <linux/posix_acl.h> |
1da177e4 LT |
37 | #include <asm/uaccess.h> |
38 | ||
e81e3f4d | 39 | #include "internal.h" |
c7105365 | 40 | #include "mount.h" |
e81e3f4d | 41 | |
1da177e4 LT |
42 | /* [Feb-1997 T. Schoebel-Theuer] |
43 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
44 | * were necessary because of omirr. The reason is that omirr needs | |
45 | * to know the _real_ pathname, not the user-supplied one, in case | |
46 | * of symlinks (and also when transname replacements occur). | |
47 | * | |
48 | * The new code replaces the old recursive symlink resolution with | |
49 | * an iterative one (in case of non-nested symlink chains). It does | |
50 | * this with calls to <fs>_follow_link(). | |
51 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
52 | * replaced with a single function lookup_dentry() that can handle all | |
53 | * the special cases of the former code. | |
54 | * | |
55 | * With the new dcache, the pathname is stored at each inode, at least as | |
56 | * long as the refcount of the inode is positive. As a side effect, the | |
57 | * size of the dcache depends on the inode cache and thus is dynamic. | |
58 | * | |
59 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
60 | * resolution to correspond with current state of the code. | |
61 | * | |
62 | * Note that the symlink resolution is not *completely* iterative. | |
63 | * There is still a significant amount of tail- and mid- recursion in | |
64 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
65 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
66 | * may return different results than <fs>_follow_link(). Many virtual | |
67 | * filesystems (including /proc) exhibit this behavior. | |
68 | */ | |
69 | ||
70 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
71 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
72 | * and the name already exists in form of a symlink, try to create the new | |
73 | * name indicated by the symlink. The old code always complained that the | |
74 | * name already exists, due to not following the symlink even if its target | |
75 | * is nonexistent. The new semantics affects also mknod() and link() when | |
25985edc | 76 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
77 | * |
78 | * I don't know which semantics is the right one, since I have no access | |
79 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
80 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
81 | * "old" one. Personally, I think the new semantics is much more logical. | |
82 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
83 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
84 | * and in the old Linux semantics. | |
85 | */ | |
86 | ||
87 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
88 | * semantics. See the comments in "open_namei" and "do_link" below. | |
89 | * | |
90 | * [10-Sep-98 Alan Modra] Another symlink change. | |
91 | */ | |
92 | ||
93 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
94 | * inside the path - always follow. | |
95 | * in the last component in creation/removal/renaming - never follow. | |
96 | * if LOOKUP_FOLLOW passed - follow. | |
97 | * if the pathname has trailing slashes - follow. | |
98 | * otherwise - don't follow. | |
99 | * (applied in that order). | |
100 | * | |
101 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
102 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
103 | * During the 2.4 we need to fix the userland stuff depending on it - | |
104 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
105 | * XEmacs seems to be relying on it... | |
106 | */ | |
107 | /* | |
108 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 109 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
110 | * any extra contention... |
111 | */ | |
112 | ||
113 | /* In order to reduce some races, while at the same time doing additional | |
114 | * checking and hopefully speeding things up, we copy filenames to the | |
115 | * kernel data space before using them.. | |
116 | * | |
117 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
118 | * PATH_MAX includes the nul terminator --RR. | |
119 | */ | |
1fa1e7f6 | 120 | static char *getname_flags(const char __user *filename, int flags, int *empty) |
1da177e4 | 121 | { |
3f9f0aa6 LT |
122 | char *result = __getname(), *err; |
123 | int len; | |
4043cde8 | 124 | |
3f9f0aa6 | 125 | if (unlikely(!result)) |
4043cde8 EP |
126 | return ERR_PTR(-ENOMEM); |
127 | ||
3f9f0aa6 LT |
128 | len = strncpy_from_user(result, filename, PATH_MAX); |
129 | err = ERR_PTR(len); | |
130 | if (unlikely(len < 0)) | |
131 | goto error; | |
132 | ||
133 | /* The empty path is special. */ | |
134 | if (unlikely(!len)) { | |
135 | if (empty) | |
4043cde8 | 136 | *empty = 1; |
3f9f0aa6 LT |
137 | err = ERR_PTR(-ENOENT); |
138 | if (!(flags & LOOKUP_EMPTY)) | |
139 | goto error; | |
1da177e4 | 140 | } |
3f9f0aa6 LT |
141 | |
142 | err = ERR_PTR(-ENAMETOOLONG); | |
143 | if (likely(len < PATH_MAX)) { | |
144 | audit_getname(result); | |
145 | return result; | |
146 | } | |
147 | ||
148 | error: | |
149 | __putname(result); | |
150 | return err; | |
1da177e4 LT |
151 | } |
152 | ||
f52e0c11 AV |
153 | char *getname(const char __user * filename) |
154 | { | |
f7493e5d | 155 | return getname_flags(filename, 0, NULL); |
f52e0c11 AV |
156 | } |
157 | ||
1da177e4 LT |
158 | #ifdef CONFIG_AUDITSYSCALL |
159 | void putname(const char *name) | |
160 | { | |
5ac3a9c2 | 161 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
162 | audit_putname(name); |
163 | else | |
164 | __putname(name); | |
165 | } | |
166 | EXPORT_SYMBOL(putname); | |
167 | #endif | |
168 | ||
e77819e5 LT |
169 | static int check_acl(struct inode *inode, int mask) |
170 | { | |
84635d68 | 171 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
172 | struct posix_acl *acl; |
173 | ||
e77819e5 | 174 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
175 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
176 | if (!acl) | |
e77819e5 | 177 | return -EAGAIN; |
3567866b AV |
178 | /* no ->get_acl() calls in RCU mode... */ |
179 | if (acl == ACL_NOT_CACHED) | |
180 | return -ECHILD; | |
206b1d09 | 181 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
182 | } |
183 | ||
184 | acl = get_cached_acl(inode, ACL_TYPE_ACCESS); | |
185 | ||
186 | /* | |
4e34e719 CH |
187 | * A filesystem can force a ACL callback by just never filling the |
188 | * ACL cache. But normally you'd fill the cache either at inode | |
189 | * instantiation time, or on the first ->get_acl call. | |
e77819e5 | 190 | * |
4e34e719 CH |
191 | * If the filesystem doesn't have a get_acl() function at all, we'll |
192 | * just create the negative cache entry. | |
e77819e5 LT |
193 | */ |
194 | if (acl == ACL_NOT_CACHED) { | |
4e34e719 CH |
195 | if (inode->i_op->get_acl) { |
196 | acl = inode->i_op->get_acl(inode, ACL_TYPE_ACCESS); | |
197 | if (IS_ERR(acl)) | |
198 | return PTR_ERR(acl); | |
199 | } else { | |
200 | set_cached_acl(inode, ACL_TYPE_ACCESS, NULL); | |
201 | return -EAGAIN; | |
202 | } | |
e77819e5 LT |
203 | } |
204 | ||
205 | if (acl) { | |
206 | int error = posix_acl_permission(inode, acl, mask); | |
207 | posix_acl_release(acl); | |
208 | return error; | |
209 | } | |
84635d68 | 210 | #endif |
e77819e5 LT |
211 | |
212 | return -EAGAIN; | |
213 | } | |
214 | ||
5909ccaa | 215 | /* |
948409c7 | 216 | * This does the basic permission checking |
1da177e4 | 217 | */ |
7e40145e | 218 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 219 | { |
26cf46be | 220 | unsigned int mode = inode->i_mode; |
1da177e4 | 221 | |
8e96e3b7 | 222 | if (likely(uid_eq(current_fsuid(), inode->i_uid))) |
1da177e4 LT |
223 | mode >>= 6; |
224 | else { | |
e77819e5 | 225 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 226 | int error = check_acl(inode, mask); |
b74c79e9 NP |
227 | if (error != -EAGAIN) |
228 | return error; | |
1da177e4 LT |
229 | } |
230 | ||
231 | if (in_group_p(inode->i_gid)) | |
232 | mode >>= 3; | |
233 | } | |
234 | ||
235 | /* | |
236 | * If the DACs are ok we don't need any capability check. | |
237 | */ | |
9c2c7039 | 238 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 239 | return 0; |
5909ccaa LT |
240 | return -EACCES; |
241 | } | |
242 | ||
243 | /** | |
b74c79e9 | 244 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 245 | * @inode: inode to check access rights for |
8fd90c8d | 246 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
247 | * |
248 | * Used to check for read/write/execute permissions on a file. | |
249 | * We use "fsuid" for this, letting us set arbitrary permissions | |
250 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
251 | * are used for other things. |
252 | * | |
253 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
254 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
255 | * It would then be called again in ref-walk mode. | |
5909ccaa | 256 | */ |
2830ba7f | 257 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
258 | { |
259 | int ret; | |
260 | ||
261 | /* | |
948409c7 | 262 | * Do the basic permission checks. |
5909ccaa | 263 | */ |
7e40145e | 264 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
265 | if (ret != -EACCES) |
266 | return ret; | |
1da177e4 | 267 | |
d594e7ec AV |
268 | if (S_ISDIR(inode->i_mode)) { |
269 | /* DACs are overridable for directories */ | |
1a48e2ac | 270 | if (inode_capable(inode, CAP_DAC_OVERRIDE)) |
d594e7ec AV |
271 | return 0; |
272 | if (!(mask & MAY_WRITE)) | |
1a48e2ac | 273 | if (inode_capable(inode, CAP_DAC_READ_SEARCH)) |
d594e7ec AV |
274 | return 0; |
275 | return -EACCES; | |
276 | } | |
1da177e4 LT |
277 | /* |
278 | * Read/write DACs are always overridable. | |
d594e7ec AV |
279 | * Executable DACs are overridable when there is |
280 | * at least one exec bit set. | |
1da177e4 | 281 | */ |
d594e7ec | 282 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) |
1a48e2ac | 283 | if (inode_capable(inode, CAP_DAC_OVERRIDE)) |
1da177e4 LT |
284 | return 0; |
285 | ||
286 | /* | |
287 | * Searching includes executable on directories, else just read. | |
288 | */ | |
7ea66001 | 289 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 290 | if (mask == MAY_READ) |
1a48e2ac | 291 | if (inode_capable(inode, CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
292 | return 0; |
293 | ||
294 | return -EACCES; | |
295 | } | |
296 | ||
3ddcd056 LT |
297 | /* |
298 | * We _really_ want to just do "generic_permission()" without | |
299 | * even looking at the inode->i_op values. So we keep a cache | |
300 | * flag in inode->i_opflags, that says "this has not special | |
301 | * permission function, use the fast case". | |
302 | */ | |
303 | static inline int do_inode_permission(struct inode *inode, int mask) | |
304 | { | |
305 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
306 | if (likely(inode->i_op->permission)) | |
307 | return inode->i_op->permission(inode, mask); | |
308 | ||
309 | /* This gets set once for the inode lifetime */ | |
310 | spin_lock(&inode->i_lock); | |
311 | inode->i_opflags |= IOP_FASTPERM; | |
312 | spin_unlock(&inode->i_lock); | |
313 | } | |
314 | return generic_permission(inode, mask); | |
315 | } | |
316 | ||
cb23beb5 | 317 | /** |
0bdaea90 DH |
318 | * __inode_permission - Check for access rights to a given inode |
319 | * @inode: Inode to check permission on | |
320 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
cb23beb5 | 321 | * |
0bdaea90 | 322 | * Check for read/write/execute permissions on an inode. |
948409c7 AG |
323 | * |
324 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
0bdaea90 DH |
325 | * |
326 | * This does not check for a read-only file system. You probably want | |
327 | * inode_permission(). | |
cb23beb5 | 328 | */ |
0bdaea90 | 329 | int __inode_permission(struct inode *inode, int mask) |
1da177e4 | 330 | { |
e6305c43 | 331 | int retval; |
1da177e4 | 332 | |
3ddcd056 | 333 | if (unlikely(mask & MAY_WRITE)) { |
1da177e4 LT |
334 | /* |
335 | * Nobody gets write access to an immutable file. | |
336 | */ | |
337 | if (IS_IMMUTABLE(inode)) | |
338 | return -EACCES; | |
339 | } | |
340 | ||
3ddcd056 | 341 | retval = do_inode_permission(inode, mask); |
1da177e4 LT |
342 | if (retval) |
343 | return retval; | |
344 | ||
08ce5f16 SH |
345 | retval = devcgroup_inode_permission(inode, mask); |
346 | if (retval) | |
347 | return retval; | |
348 | ||
d09ca739 | 349 | return security_inode_permission(inode, mask); |
1da177e4 LT |
350 | } |
351 | ||
0bdaea90 DH |
352 | /** |
353 | * sb_permission - Check superblock-level permissions | |
354 | * @sb: Superblock of inode to check permission on | |
55852635 | 355 | * @inode: Inode to check permission on |
0bdaea90 DH |
356 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) |
357 | * | |
358 | * Separate out file-system wide checks from inode-specific permission checks. | |
359 | */ | |
360 | static int sb_permission(struct super_block *sb, struct inode *inode, int mask) | |
361 | { | |
362 | if (unlikely(mask & MAY_WRITE)) { | |
363 | umode_t mode = inode->i_mode; | |
364 | ||
365 | /* Nobody gets write access to a read-only fs. */ | |
366 | if ((sb->s_flags & MS_RDONLY) && | |
367 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
368 | return -EROFS; | |
369 | } | |
370 | return 0; | |
371 | } | |
372 | ||
373 | /** | |
374 | * inode_permission - Check for access rights to a given inode | |
375 | * @inode: Inode to check permission on | |
376 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
377 | * | |
378 | * Check for read/write/execute permissions on an inode. We use fs[ug]id for | |
379 | * this, letting us set arbitrary permissions for filesystem access without | |
380 | * changing the "normal" UIDs which are used for other things. | |
381 | * | |
382 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
383 | */ | |
384 | int inode_permission(struct inode *inode, int mask) | |
385 | { | |
386 | int retval; | |
387 | ||
388 | retval = sb_permission(inode->i_sb, inode, mask); | |
389 | if (retval) | |
390 | return retval; | |
391 | return __inode_permission(inode, mask); | |
392 | } | |
393 | ||
5dd784d0 JB |
394 | /** |
395 | * path_get - get a reference to a path | |
396 | * @path: path to get the reference to | |
397 | * | |
398 | * Given a path increment the reference count to the dentry and the vfsmount. | |
399 | */ | |
400 | void path_get(struct path *path) | |
401 | { | |
402 | mntget(path->mnt); | |
403 | dget(path->dentry); | |
404 | } | |
405 | EXPORT_SYMBOL(path_get); | |
406 | ||
1d957f9b JB |
407 | /** |
408 | * path_put - put a reference to a path | |
409 | * @path: path to put the reference to | |
410 | * | |
411 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
412 | */ | |
413 | void path_put(struct path *path) | |
1da177e4 | 414 | { |
1d957f9b JB |
415 | dput(path->dentry); |
416 | mntput(path->mnt); | |
1da177e4 | 417 | } |
1d957f9b | 418 | EXPORT_SYMBOL(path_put); |
1da177e4 | 419 | |
19660af7 | 420 | /* |
31e6b01f | 421 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
422 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
423 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
424 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
425 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
426 | * got stuck, so ref-walk may continue from there. If this is not successful | |
427 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
428 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 429 | */ |
31e6b01f | 430 | |
32a7991b AV |
431 | static inline void lock_rcu_walk(void) |
432 | { | |
433 | br_read_lock(&vfsmount_lock); | |
434 | rcu_read_lock(); | |
435 | } | |
436 | ||
437 | static inline void unlock_rcu_walk(void) | |
438 | { | |
439 | rcu_read_unlock(); | |
440 | br_read_unlock(&vfsmount_lock); | |
441 | } | |
442 | ||
31e6b01f | 443 | /** |
19660af7 AV |
444 | * unlazy_walk - try to switch to ref-walk mode. |
445 | * @nd: nameidata pathwalk data | |
446 | * @dentry: child of nd->path.dentry or NULL | |
39191628 | 447 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 448 | * |
19660af7 AV |
449 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
450 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
451 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 452 | */ |
19660af7 | 453 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry) |
31e6b01f NP |
454 | { |
455 | struct fs_struct *fs = current->fs; | |
456 | struct dentry *parent = nd->path.dentry; | |
5b6ca027 | 457 | int want_root = 0; |
31e6b01f NP |
458 | |
459 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
5b6ca027 AV |
460 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
461 | want_root = 1; | |
31e6b01f NP |
462 | spin_lock(&fs->lock); |
463 | if (nd->root.mnt != fs->root.mnt || | |
464 | nd->root.dentry != fs->root.dentry) | |
465 | goto err_root; | |
466 | } | |
467 | spin_lock(&parent->d_lock); | |
19660af7 AV |
468 | if (!dentry) { |
469 | if (!__d_rcu_to_refcount(parent, nd->seq)) | |
470 | goto err_parent; | |
471 | BUG_ON(nd->inode != parent->d_inode); | |
472 | } else { | |
94c0d4ec AV |
473 | if (dentry->d_parent != parent) |
474 | goto err_parent; | |
19660af7 AV |
475 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); |
476 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
477 | goto err_child; | |
478 | /* | |
479 | * If the sequence check on the child dentry passed, then | |
480 | * the child has not been removed from its parent. This | |
481 | * means the parent dentry must be valid and able to take | |
482 | * a reference at this point. | |
483 | */ | |
484 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
485 | BUG_ON(!parent->d_count); | |
486 | parent->d_count++; | |
487 | spin_unlock(&dentry->d_lock); | |
488 | } | |
31e6b01f | 489 | spin_unlock(&parent->d_lock); |
5b6ca027 | 490 | if (want_root) { |
31e6b01f NP |
491 | path_get(&nd->root); |
492 | spin_unlock(&fs->lock); | |
493 | } | |
494 | mntget(nd->path.mnt); | |
495 | ||
32a7991b | 496 | unlock_rcu_walk(); |
31e6b01f NP |
497 | nd->flags &= ~LOOKUP_RCU; |
498 | return 0; | |
19660af7 AV |
499 | |
500 | err_child: | |
31e6b01f | 501 | spin_unlock(&dentry->d_lock); |
19660af7 | 502 | err_parent: |
31e6b01f NP |
503 | spin_unlock(&parent->d_lock); |
504 | err_root: | |
5b6ca027 | 505 | if (want_root) |
31e6b01f NP |
506 | spin_unlock(&fs->lock); |
507 | return -ECHILD; | |
508 | } | |
509 | ||
4ce16ef3 | 510 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 511 | { |
4ce16ef3 | 512 | return dentry->d_op->d_revalidate(dentry, flags); |
34286d66 NP |
513 | } |
514 | ||
9f1fafee AV |
515 | /** |
516 | * complete_walk - successful completion of path walk | |
517 | * @nd: pointer nameidata | |
39159de2 | 518 | * |
9f1fafee AV |
519 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
520 | * Revalidate the final result, unless we'd already done that during | |
521 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
522 | * success, -error on failure. In case of failure caller does not | |
523 | * need to drop nd->path. | |
39159de2 | 524 | */ |
9f1fafee | 525 | static int complete_walk(struct nameidata *nd) |
39159de2 | 526 | { |
16c2cd71 | 527 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 528 | int status; |
39159de2 | 529 | |
9f1fafee AV |
530 | if (nd->flags & LOOKUP_RCU) { |
531 | nd->flags &= ~LOOKUP_RCU; | |
532 | if (!(nd->flags & LOOKUP_ROOT)) | |
533 | nd->root.mnt = NULL; | |
534 | spin_lock(&dentry->d_lock); | |
535 | if (unlikely(!__d_rcu_to_refcount(dentry, nd->seq))) { | |
536 | spin_unlock(&dentry->d_lock); | |
32a7991b | 537 | unlock_rcu_walk(); |
9f1fafee AV |
538 | return -ECHILD; |
539 | } | |
540 | BUG_ON(nd->inode != dentry->d_inode); | |
541 | spin_unlock(&dentry->d_lock); | |
542 | mntget(nd->path.mnt); | |
32a7991b | 543 | unlock_rcu_walk(); |
9f1fafee AV |
544 | } |
545 | ||
16c2cd71 AV |
546 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
547 | return 0; | |
548 | ||
549 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
39159de2 JL |
550 | return 0; |
551 | ||
16c2cd71 AV |
552 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) |
553 | return 0; | |
554 | ||
555 | /* Note: we do not d_invalidate() */ | |
4ce16ef3 | 556 | status = d_revalidate(dentry, nd->flags); |
39159de2 JL |
557 | if (status > 0) |
558 | return 0; | |
559 | ||
16c2cd71 | 560 | if (!status) |
39159de2 | 561 | status = -ESTALE; |
16c2cd71 | 562 | |
9f1fafee | 563 | path_put(&nd->path); |
39159de2 JL |
564 | return status; |
565 | } | |
566 | ||
2a737871 AV |
567 | static __always_inline void set_root(struct nameidata *nd) |
568 | { | |
f7ad3c6b MS |
569 | if (!nd->root.mnt) |
570 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
571 | } |
572 | ||
6de88d72 AV |
573 | static int link_path_walk(const char *, struct nameidata *); |
574 | ||
31e6b01f NP |
575 | static __always_inline void set_root_rcu(struct nameidata *nd) |
576 | { | |
577 | if (!nd->root.mnt) { | |
578 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
579 | unsigned seq; |
580 | ||
581 | do { | |
582 | seq = read_seqcount_begin(&fs->seq); | |
583 | nd->root = fs->root; | |
c1530019 | 584 | nd->seq = __read_seqcount_begin(&nd->root.dentry->d_seq); |
c28cc364 | 585 | } while (read_seqcount_retry(&fs->seq, seq)); |
31e6b01f NP |
586 | } |
587 | } | |
588 | ||
f1662356 | 589 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 590 | { |
31e6b01f NP |
591 | int ret; |
592 | ||
1da177e4 LT |
593 | if (IS_ERR(link)) |
594 | goto fail; | |
595 | ||
596 | if (*link == '/') { | |
2a737871 | 597 | set_root(nd); |
1d957f9b | 598 | path_put(&nd->path); |
2a737871 AV |
599 | nd->path = nd->root; |
600 | path_get(&nd->root); | |
16c2cd71 | 601 | nd->flags |= LOOKUP_JUMPED; |
1da177e4 | 602 | } |
31e6b01f | 603 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 604 | |
31e6b01f NP |
605 | ret = link_path_walk(link, nd); |
606 | return ret; | |
1da177e4 | 607 | fail: |
1d957f9b | 608 | path_put(&nd->path); |
1da177e4 LT |
609 | return PTR_ERR(link); |
610 | } | |
611 | ||
1d957f9b | 612 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
613 | { |
614 | dput(path->dentry); | |
4ac91378 | 615 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
616 | mntput(path->mnt); |
617 | } | |
618 | ||
7b9337aa NP |
619 | static inline void path_to_nameidata(const struct path *path, |
620 | struct nameidata *nd) | |
051d3812 | 621 | { |
31e6b01f NP |
622 | if (!(nd->flags & LOOKUP_RCU)) { |
623 | dput(nd->path.dentry); | |
624 | if (nd->path.mnt != path->mnt) | |
625 | mntput(nd->path.mnt); | |
9a229683 | 626 | } |
31e6b01f | 627 | nd->path.mnt = path->mnt; |
4ac91378 | 628 | nd->path.dentry = path->dentry; |
051d3812 IK |
629 | } |
630 | ||
b5fb63c1 CH |
631 | /* |
632 | * Helper to directly jump to a known parsed path from ->follow_link, | |
633 | * caller must have taken a reference to path beforehand. | |
634 | */ | |
635 | void nd_jump_link(struct nameidata *nd, struct path *path) | |
636 | { | |
637 | path_put(&nd->path); | |
638 | ||
639 | nd->path = *path; | |
640 | nd->inode = nd->path.dentry->d_inode; | |
641 | nd->flags |= LOOKUP_JUMPED; | |
642 | ||
643 | BUG_ON(nd->inode->i_op->follow_link); | |
644 | } | |
645 | ||
574197e0 AV |
646 | static inline void put_link(struct nameidata *nd, struct path *link, void *cookie) |
647 | { | |
648 | struct inode *inode = link->dentry->d_inode; | |
6d7b5aae | 649 | if (inode->i_op->put_link) |
574197e0 AV |
650 | inode->i_op->put_link(link->dentry, nd, cookie); |
651 | path_put(link); | |
652 | } | |
653 | ||
800179c9 KC |
654 | int sysctl_protected_symlinks __read_mostly = 1; |
655 | int sysctl_protected_hardlinks __read_mostly = 1; | |
656 | ||
657 | /** | |
658 | * may_follow_link - Check symlink following for unsafe situations | |
659 | * @link: The path of the symlink | |
55852635 | 660 | * @nd: nameidata pathwalk data |
800179c9 KC |
661 | * |
662 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
663 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
664 | * in a sticky world-writable directory. This is to protect privileged | |
665 | * processes from failing races against path names that may change out | |
666 | * from under them by way of other users creating malicious symlinks. | |
667 | * It will permit symlinks to be followed only when outside a sticky | |
668 | * world-writable directory, or when the uid of the symlink and follower | |
669 | * match, or when the directory owner matches the symlink's owner. | |
670 | * | |
671 | * Returns 0 if following the symlink is allowed, -ve on error. | |
672 | */ | |
673 | static inline int may_follow_link(struct path *link, struct nameidata *nd) | |
674 | { | |
675 | const struct inode *inode; | |
676 | const struct inode *parent; | |
677 | ||
678 | if (!sysctl_protected_symlinks) | |
679 | return 0; | |
680 | ||
681 | /* Allowed if owner and follower match. */ | |
682 | inode = link->dentry->d_inode; | |
81abe27b | 683 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
684 | return 0; |
685 | ||
686 | /* Allowed if parent directory not sticky and world-writable. */ | |
687 | parent = nd->path.dentry->d_inode; | |
688 | if ((parent->i_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) | |
689 | return 0; | |
690 | ||
691 | /* Allowed if parent directory and link owner match. */ | |
81abe27b | 692 | if (uid_eq(parent->i_uid, inode->i_uid)) |
800179c9 KC |
693 | return 0; |
694 | ||
ffd8d101 | 695 | audit_log_link_denied("follow_link", link); |
800179c9 KC |
696 | path_put_conditional(link, nd); |
697 | path_put(&nd->path); | |
698 | return -EACCES; | |
699 | } | |
700 | ||
701 | /** | |
702 | * safe_hardlink_source - Check for safe hardlink conditions | |
703 | * @inode: the source inode to hardlink from | |
704 | * | |
705 | * Return false if at least one of the following conditions: | |
706 | * - inode is not a regular file | |
707 | * - inode is setuid | |
708 | * - inode is setgid and group-exec | |
709 | * - access failure for read and write | |
710 | * | |
711 | * Otherwise returns true. | |
712 | */ | |
713 | static bool safe_hardlink_source(struct inode *inode) | |
714 | { | |
715 | umode_t mode = inode->i_mode; | |
716 | ||
717 | /* Special files should not get pinned to the filesystem. */ | |
718 | if (!S_ISREG(mode)) | |
719 | return false; | |
720 | ||
721 | /* Setuid files should not get pinned to the filesystem. */ | |
722 | if (mode & S_ISUID) | |
723 | return false; | |
724 | ||
725 | /* Executable setgid files should not get pinned to the filesystem. */ | |
726 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
727 | return false; | |
728 | ||
729 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
730 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
731 | return false; | |
732 | ||
733 | return true; | |
734 | } | |
735 | ||
736 | /** | |
737 | * may_linkat - Check permissions for creating a hardlink | |
738 | * @link: the source to hardlink from | |
739 | * | |
740 | * Block hardlink when all of: | |
741 | * - sysctl_protected_hardlinks enabled | |
742 | * - fsuid does not match inode | |
743 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
744 | * - not CAP_FOWNER | |
745 | * | |
746 | * Returns 0 if successful, -ve on error. | |
747 | */ | |
748 | static int may_linkat(struct path *link) | |
749 | { | |
750 | const struct cred *cred; | |
751 | struct inode *inode; | |
752 | ||
753 | if (!sysctl_protected_hardlinks) | |
754 | return 0; | |
755 | ||
756 | cred = current_cred(); | |
757 | inode = link->dentry->d_inode; | |
758 | ||
759 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, | |
760 | * otherwise, it must be a safe source. | |
761 | */ | |
81abe27b | 762 | if (uid_eq(cred->fsuid, inode->i_uid) || safe_hardlink_source(inode) || |
800179c9 KC |
763 | capable(CAP_FOWNER)) |
764 | return 0; | |
765 | ||
a51d9eaa | 766 | audit_log_link_denied("linkat", link); |
800179c9 KC |
767 | return -EPERM; |
768 | } | |
769 | ||
def4af30 | 770 | static __always_inline int |
574197e0 | 771 | follow_link(struct path *link, struct nameidata *nd, void **p) |
1da177e4 | 772 | { |
7b9337aa | 773 | struct dentry *dentry = link->dentry; |
6d7b5aae AV |
774 | int error; |
775 | char *s; | |
1da177e4 | 776 | |
844a3917 AV |
777 | BUG_ON(nd->flags & LOOKUP_RCU); |
778 | ||
0e794589 AV |
779 | if (link->mnt == nd->path.mnt) |
780 | mntget(link->mnt); | |
781 | ||
6d7b5aae AV |
782 | error = -ELOOP; |
783 | if (unlikely(current->total_link_count >= 40)) | |
784 | goto out_put_nd_path; | |
785 | ||
574197e0 AV |
786 | cond_resched(); |
787 | current->total_link_count++; | |
788 | ||
68ac1234 | 789 | touch_atime(link); |
1da177e4 | 790 | nd_set_link(nd, NULL); |
cd4e91d3 | 791 | |
36f3b4f6 | 792 | error = security_inode_follow_link(link->dentry, nd); |
6d7b5aae AV |
793 | if (error) |
794 | goto out_put_nd_path; | |
36f3b4f6 | 795 | |
86acdca1 | 796 | nd->last_type = LAST_BIND; |
def4af30 AV |
797 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
798 | error = PTR_ERR(*p); | |
6d7b5aae | 799 | if (IS_ERR(*p)) |
408ef013 | 800 | goto out_put_nd_path; |
6d7b5aae AV |
801 | |
802 | error = 0; | |
803 | s = nd_get_link(nd); | |
804 | if (s) { | |
805 | error = __vfs_follow_link(nd, s); | |
b5fb63c1 CH |
806 | if (unlikely(error)) |
807 | put_link(nd, link, *p); | |
1da177e4 | 808 | } |
6d7b5aae AV |
809 | |
810 | return error; | |
811 | ||
812 | out_put_nd_path: | |
98f6ef64 | 813 | *p = NULL; |
6d7b5aae | 814 | path_put(&nd->path); |
6d7b5aae | 815 | path_put(link); |
1da177e4 LT |
816 | return error; |
817 | } | |
818 | ||
31e6b01f NP |
819 | static int follow_up_rcu(struct path *path) |
820 | { | |
0714a533 AV |
821 | struct mount *mnt = real_mount(path->mnt); |
822 | struct mount *parent; | |
31e6b01f NP |
823 | struct dentry *mountpoint; |
824 | ||
0714a533 AV |
825 | parent = mnt->mnt_parent; |
826 | if (&parent->mnt == path->mnt) | |
31e6b01f | 827 | return 0; |
a73324da | 828 | mountpoint = mnt->mnt_mountpoint; |
31e6b01f | 829 | path->dentry = mountpoint; |
0714a533 | 830 | path->mnt = &parent->mnt; |
31e6b01f NP |
831 | return 1; |
832 | } | |
833 | ||
f015f126 DH |
834 | /* |
835 | * follow_up - Find the mountpoint of path's vfsmount | |
836 | * | |
837 | * Given a path, find the mountpoint of its source file system. | |
838 | * Replace @path with the path of the mountpoint in the parent mount. | |
839 | * Up is towards /. | |
840 | * | |
841 | * Return 1 if we went up a level and 0 if we were already at the | |
842 | * root. | |
843 | */ | |
bab77ebf | 844 | int follow_up(struct path *path) |
1da177e4 | 845 | { |
0714a533 AV |
846 | struct mount *mnt = real_mount(path->mnt); |
847 | struct mount *parent; | |
1da177e4 | 848 | struct dentry *mountpoint; |
99b7db7b | 849 | |
962830df | 850 | br_read_lock(&vfsmount_lock); |
0714a533 | 851 | parent = mnt->mnt_parent; |
3c0a6163 | 852 | if (parent == mnt) { |
962830df | 853 | br_read_unlock(&vfsmount_lock); |
1da177e4 LT |
854 | return 0; |
855 | } | |
0714a533 | 856 | mntget(&parent->mnt); |
a73324da | 857 | mountpoint = dget(mnt->mnt_mountpoint); |
962830df | 858 | br_read_unlock(&vfsmount_lock); |
bab77ebf AV |
859 | dput(path->dentry); |
860 | path->dentry = mountpoint; | |
861 | mntput(path->mnt); | |
0714a533 | 862 | path->mnt = &parent->mnt; |
1da177e4 LT |
863 | return 1; |
864 | } | |
865 | ||
b5c84bf6 | 866 | /* |
9875cf80 DH |
867 | * Perform an automount |
868 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
869 | * were called with. | |
1da177e4 | 870 | */ |
9875cf80 DH |
871 | static int follow_automount(struct path *path, unsigned flags, |
872 | bool *need_mntput) | |
31e6b01f | 873 | { |
9875cf80 | 874 | struct vfsmount *mnt; |
ea5b778a | 875 | int err; |
9875cf80 DH |
876 | |
877 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
878 | return -EREMOTE; | |
879 | ||
0ec26fd0 MS |
880 | /* We don't want to mount if someone's just doing a stat - |
881 | * unless they're stat'ing a directory and appended a '/' to | |
882 | * the name. | |
883 | * | |
884 | * We do, however, want to mount if someone wants to open or | |
885 | * create a file of any type under the mountpoint, wants to | |
886 | * traverse through the mountpoint or wants to open the | |
887 | * mounted directory. Also, autofs may mark negative dentries | |
888 | * as being automount points. These will need the attentions | |
889 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 890 | */ |
0ec26fd0 | 891 | if (!(flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
d94c177b | 892 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
0ec26fd0 MS |
893 | path->dentry->d_inode) |
894 | return -EISDIR; | |
895 | ||
9875cf80 DH |
896 | current->total_link_count++; |
897 | if (current->total_link_count >= 40) | |
898 | return -ELOOP; | |
899 | ||
900 | mnt = path->dentry->d_op->d_automount(path); | |
901 | if (IS_ERR(mnt)) { | |
902 | /* | |
903 | * The filesystem is allowed to return -EISDIR here to indicate | |
904 | * it doesn't want to automount. For instance, autofs would do | |
905 | * this so that its userspace daemon can mount on this dentry. | |
906 | * | |
907 | * However, we can only permit this if it's a terminal point in | |
908 | * the path being looked up; if it wasn't then the remainder of | |
909 | * the path is inaccessible and we should say so. | |
910 | */ | |
49084c3b | 911 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_PARENT)) |
9875cf80 DH |
912 | return -EREMOTE; |
913 | return PTR_ERR(mnt); | |
31e6b01f | 914 | } |
ea5b778a | 915 | |
9875cf80 DH |
916 | if (!mnt) /* mount collision */ |
917 | return 0; | |
31e6b01f | 918 | |
8aef1884 AV |
919 | if (!*need_mntput) { |
920 | /* lock_mount() may release path->mnt on error */ | |
921 | mntget(path->mnt); | |
922 | *need_mntput = true; | |
923 | } | |
19a167af | 924 | err = finish_automount(mnt, path); |
9875cf80 | 925 | |
ea5b778a DH |
926 | switch (err) { |
927 | case -EBUSY: | |
928 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 929 | return 0; |
ea5b778a | 930 | case 0: |
8aef1884 | 931 | path_put(path); |
ea5b778a DH |
932 | path->mnt = mnt; |
933 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 934 | return 0; |
19a167af AV |
935 | default: |
936 | return err; | |
ea5b778a | 937 | } |
19a167af | 938 | |
463ffb2e AV |
939 | } |
940 | ||
9875cf80 DH |
941 | /* |
942 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 943 | * - Flagged for transit management (autofs) |
9875cf80 DH |
944 | * - Flagged as mountpoint |
945 | * - Flagged as automount point | |
946 | * | |
947 | * This may only be called in refwalk mode. | |
948 | * | |
949 | * Serialization is taken care of in namespace.c | |
950 | */ | |
951 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 952 | { |
8aef1884 | 953 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
954 | unsigned managed; |
955 | bool need_mntput = false; | |
8aef1884 | 956 | int ret = 0; |
9875cf80 DH |
957 | |
958 | /* Given that we're not holding a lock here, we retain the value in a | |
959 | * local variable for each dentry as we look at it so that we don't see | |
960 | * the components of that value change under us */ | |
961 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
962 | managed &= DCACHE_MANAGED_DENTRY, | |
963 | unlikely(managed != 0)) { | |
cc53ce53 DH |
964 | /* Allow the filesystem to manage the transit without i_mutex |
965 | * being held. */ | |
966 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
967 | BUG_ON(!path->dentry->d_op); | |
968 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 969 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 970 | if (ret < 0) |
8aef1884 | 971 | break; |
cc53ce53 DH |
972 | } |
973 | ||
9875cf80 DH |
974 | /* Transit to a mounted filesystem. */ |
975 | if (managed & DCACHE_MOUNTED) { | |
976 | struct vfsmount *mounted = lookup_mnt(path); | |
977 | if (mounted) { | |
978 | dput(path->dentry); | |
979 | if (need_mntput) | |
980 | mntput(path->mnt); | |
981 | path->mnt = mounted; | |
982 | path->dentry = dget(mounted->mnt_root); | |
983 | need_mntput = true; | |
984 | continue; | |
985 | } | |
986 | ||
987 | /* Something is mounted on this dentry in another | |
988 | * namespace and/or whatever was mounted there in this | |
989 | * namespace got unmounted before we managed to get the | |
990 | * vfsmount_lock */ | |
991 | } | |
992 | ||
993 | /* Handle an automount point */ | |
994 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
995 | ret = follow_automount(path, flags, &need_mntput); | |
996 | if (ret < 0) | |
8aef1884 | 997 | break; |
9875cf80 DH |
998 | continue; |
999 | } | |
1000 | ||
1001 | /* We didn't change the current path point */ | |
1002 | break; | |
1da177e4 | 1003 | } |
8aef1884 AV |
1004 | |
1005 | if (need_mntput && path->mnt == mnt) | |
1006 | mntput(path->mnt); | |
1007 | if (ret == -EISDIR) | |
1008 | ret = 0; | |
a3fbbde7 | 1009 | return ret < 0 ? ret : need_mntput; |
1da177e4 LT |
1010 | } |
1011 | ||
cc53ce53 | 1012 | int follow_down_one(struct path *path) |
1da177e4 LT |
1013 | { |
1014 | struct vfsmount *mounted; | |
1015 | ||
1c755af4 | 1016 | mounted = lookup_mnt(path); |
1da177e4 | 1017 | if (mounted) { |
9393bd07 AV |
1018 | dput(path->dentry); |
1019 | mntput(path->mnt); | |
1020 | path->mnt = mounted; | |
1021 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1022 | return 1; |
1023 | } | |
1024 | return 0; | |
1025 | } | |
1026 | ||
62a7375e IK |
1027 | static inline bool managed_dentry_might_block(struct dentry *dentry) |
1028 | { | |
1029 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT && | |
1030 | dentry->d_op->d_manage(dentry, true) < 0); | |
1031 | } | |
1032 | ||
9875cf80 | 1033 | /* |
287548e4 AV |
1034 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1035 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1036 | */ |
1037 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 1038 | struct inode **inode) |
9875cf80 | 1039 | { |
62a7375e | 1040 | for (;;) { |
c7105365 | 1041 | struct mount *mounted; |
62a7375e IK |
1042 | /* |
1043 | * Don't forget we might have a non-mountpoint managed dentry | |
1044 | * that wants to block transit. | |
1045 | */ | |
287548e4 | 1046 | if (unlikely(managed_dentry_might_block(path->dentry))) |
ab90911f | 1047 | return false; |
62a7375e IK |
1048 | |
1049 | if (!d_mountpoint(path->dentry)) | |
1050 | break; | |
1051 | ||
9875cf80 DH |
1052 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
1053 | if (!mounted) | |
1054 | break; | |
c7105365 AV |
1055 | path->mnt = &mounted->mnt; |
1056 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1057 | nd->flags |= LOOKUP_JUMPED; |
9875cf80 | 1058 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1059 | /* |
1060 | * Update the inode too. We don't need to re-check the | |
1061 | * dentry sequence number here after this d_inode read, | |
1062 | * because a mount-point is always pinned. | |
1063 | */ | |
1064 | *inode = path->dentry->d_inode; | |
9875cf80 | 1065 | } |
9875cf80 DH |
1066 | return true; |
1067 | } | |
1068 | ||
dea39376 | 1069 | static void follow_mount_rcu(struct nameidata *nd) |
287548e4 | 1070 | { |
dea39376 | 1071 | while (d_mountpoint(nd->path.dentry)) { |
c7105365 | 1072 | struct mount *mounted; |
dea39376 | 1073 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry, 1); |
287548e4 AV |
1074 | if (!mounted) |
1075 | break; | |
c7105365 AV |
1076 | nd->path.mnt = &mounted->mnt; |
1077 | nd->path.dentry = mounted->mnt.mnt_root; | |
dea39376 | 1078 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
287548e4 AV |
1079 | } |
1080 | } | |
1081 | ||
31e6b01f NP |
1082 | static int follow_dotdot_rcu(struct nameidata *nd) |
1083 | { | |
31e6b01f NP |
1084 | set_root_rcu(nd); |
1085 | ||
9875cf80 | 1086 | while (1) { |
31e6b01f NP |
1087 | if (nd->path.dentry == nd->root.dentry && |
1088 | nd->path.mnt == nd->root.mnt) { | |
1089 | break; | |
1090 | } | |
1091 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
1092 | struct dentry *old = nd->path.dentry; | |
1093 | struct dentry *parent = old->d_parent; | |
1094 | unsigned seq; | |
1095 | ||
1096 | seq = read_seqcount_begin(&parent->d_seq); | |
1097 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 1098 | goto failed; |
31e6b01f NP |
1099 | nd->path.dentry = parent; |
1100 | nd->seq = seq; | |
1101 | break; | |
1102 | } | |
1103 | if (!follow_up_rcu(&nd->path)) | |
1104 | break; | |
1105 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
31e6b01f | 1106 | } |
dea39376 AV |
1107 | follow_mount_rcu(nd); |
1108 | nd->inode = nd->path.dentry->d_inode; | |
31e6b01f | 1109 | return 0; |
ef7562d5 AV |
1110 | |
1111 | failed: | |
1112 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1113 | if (!(nd->flags & LOOKUP_ROOT)) |
1114 | nd->root.mnt = NULL; | |
32a7991b | 1115 | unlock_rcu_walk(); |
ef7562d5 | 1116 | return -ECHILD; |
31e6b01f NP |
1117 | } |
1118 | ||
cc53ce53 DH |
1119 | /* |
1120 | * Follow down to the covering mount currently visible to userspace. At each | |
1121 | * point, the filesystem owning that dentry may be queried as to whether the | |
1122 | * caller is permitted to proceed or not. | |
cc53ce53 | 1123 | */ |
7cc90cc3 | 1124 | int follow_down(struct path *path) |
cc53ce53 DH |
1125 | { |
1126 | unsigned managed; | |
1127 | int ret; | |
1128 | ||
1129 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1130 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1131 | /* Allow the filesystem to manage the transit without i_mutex | |
1132 | * being held. | |
1133 | * | |
1134 | * We indicate to the filesystem if someone is trying to mount | |
1135 | * something here. This gives autofs the chance to deny anyone | |
1136 | * other than its daemon the right to mount on its | |
1137 | * superstructure. | |
1138 | * | |
1139 | * The filesystem may sleep at this point. | |
1140 | */ | |
1141 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1142 | BUG_ON(!path->dentry->d_op); | |
1143 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 1144 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 1145 | path->dentry, false); |
cc53ce53 DH |
1146 | if (ret < 0) |
1147 | return ret == -EISDIR ? 0 : ret; | |
1148 | } | |
1149 | ||
1150 | /* Transit to a mounted filesystem. */ | |
1151 | if (managed & DCACHE_MOUNTED) { | |
1152 | struct vfsmount *mounted = lookup_mnt(path); | |
1153 | if (!mounted) | |
1154 | break; | |
1155 | dput(path->dentry); | |
1156 | mntput(path->mnt); | |
1157 | path->mnt = mounted; | |
1158 | path->dentry = dget(mounted->mnt_root); | |
1159 | continue; | |
1160 | } | |
1161 | ||
1162 | /* Don't handle automount points here */ | |
1163 | break; | |
1164 | } | |
1165 | return 0; | |
1166 | } | |
1167 | ||
9875cf80 DH |
1168 | /* |
1169 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1170 | */ | |
1171 | static void follow_mount(struct path *path) | |
1172 | { | |
1173 | while (d_mountpoint(path->dentry)) { | |
1174 | struct vfsmount *mounted = lookup_mnt(path); | |
1175 | if (!mounted) | |
1176 | break; | |
1177 | dput(path->dentry); | |
1178 | mntput(path->mnt); | |
1179 | path->mnt = mounted; | |
1180 | path->dentry = dget(mounted->mnt_root); | |
1181 | } | |
1182 | } | |
1183 | ||
31e6b01f | 1184 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1185 | { |
2a737871 | 1186 | set_root(nd); |
e518ddb7 | 1187 | |
1da177e4 | 1188 | while(1) { |
4ac91378 | 1189 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1190 | |
2a737871 AV |
1191 | if (nd->path.dentry == nd->root.dentry && |
1192 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1193 | break; |
1194 | } | |
4ac91378 | 1195 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1196 | /* rare case of legitimate dget_parent()... */ |
1197 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1198 | dput(old); |
1199 | break; | |
1200 | } | |
3088dd70 | 1201 | if (!follow_up(&nd->path)) |
1da177e4 | 1202 | break; |
1da177e4 | 1203 | } |
79ed0226 | 1204 | follow_mount(&nd->path); |
31e6b01f | 1205 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1206 | } |
1207 | ||
baa03890 | 1208 | /* |
bad61189 MS |
1209 | * This looks up the name in dcache, possibly revalidates the old dentry and |
1210 | * allocates a new one if not found or not valid. In the need_lookup argument | |
1211 | * returns whether i_op->lookup is necessary. | |
1212 | * | |
1213 | * dir->d_inode->i_mutex must be held | |
baa03890 | 1214 | */ |
bad61189 | 1215 | static struct dentry *lookup_dcache(struct qstr *name, struct dentry *dir, |
201f956e | 1216 | unsigned int flags, bool *need_lookup) |
baa03890 | 1217 | { |
baa03890 | 1218 | struct dentry *dentry; |
bad61189 | 1219 | int error; |
baa03890 | 1220 | |
bad61189 MS |
1221 | *need_lookup = false; |
1222 | dentry = d_lookup(dir, name); | |
1223 | if (dentry) { | |
1224 | if (d_need_lookup(dentry)) { | |
1225 | *need_lookup = true; | |
1226 | } else if (dentry->d_flags & DCACHE_OP_REVALIDATE) { | |
201f956e | 1227 | error = d_revalidate(dentry, flags); |
bad61189 MS |
1228 | if (unlikely(error <= 0)) { |
1229 | if (error < 0) { | |
1230 | dput(dentry); | |
1231 | return ERR_PTR(error); | |
1232 | } else if (!d_invalidate(dentry)) { | |
1233 | dput(dentry); | |
1234 | dentry = NULL; | |
1235 | } | |
1236 | } | |
1237 | } | |
1238 | } | |
baa03890 | 1239 | |
bad61189 MS |
1240 | if (!dentry) { |
1241 | dentry = d_alloc(dir, name); | |
1242 | if (unlikely(!dentry)) | |
1243 | return ERR_PTR(-ENOMEM); | |
baa03890 | 1244 | |
bad61189 | 1245 | *need_lookup = true; |
baa03890 NP |
1246 | } |
1247 | return dentry; | |
1248 | } | |
1249 | ||
44396f4b | 1250 | /* |
bad61189 MS |
1251 | * Call i_op->lookup on the dentry. The dentry must be negative but may be |
1252 | * hashed if it was pouplated with DCACHE_NEED_LOOKUP. | |
1253 | * | |
1254 | * dir->d_inode->i_mutex must be held | |
44396f4b | 1255 | */ |
bad61189 | 1256 | static struct dentry *lookup_real(struct inode *dir, struct dentry *dentry, |
72bd866a | 1257 | unsigned int flags) |
44396f4b | 1258 | { |
44396f4b JB |
1259 | struct dentry *old; |
1260 | ||
1261 | /* Don't create child dentry for a dead directory. */ | |
bad61189 | 1262 | if (unlikely(IS_DEADDIR(dir))) { |
e188dc02 | 1263 | dput(dentry); |
44396f4b | 1264 | return ERR_PTR(-ENOENT); |
e188dc02 | 1265 | } |
44396f4b | 1266 | |
72bd866a | 1267 | old = dir->i_op->lookup(dir, dentry, flags); |
44396f4b JB |
1268 | if (unlikely(old)) { |
1269 | dput(dentry); | |
1270 | dentry = old; | |
1271 | } | |
1272 | return dentry; | |
1273 | } | |
1274 | ||
a3255546 | 1275 | static struct dentry *__lookup_hash(struct qstr *name, |
72bd866a | 1276 | struct dentry *base, unsigned int flags) |
a3255546 | 1277 | { |
bad61189 | 1278 | bool need_lookup; |
a3255546 AV |
1279 | struct dentry *dentry; |
1280 | ||
72bd866a | 1281 | dentry = lookup_dcache(name, base, flags, &need_lookup); |
bad61189 MS |
1282 | if (!need_lookup) |
1283 | return dentry; | |
a3255546 | 1284 | |
72bd866a | 1285 | return lookup_real(base->d_inode, dentry, flags); |
a3255546 AV |
1286 | } |
1287 | ||
1da177e4 LT |
1288 | /* |
1289 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1290 | * small and for now I'd prefer to have fast path as straight as possible. | |
1291 | * It _is_ time-critical. | |
1292 | */ | |
697f514d MS |
1293 | static int lookup_fast(struct nameidata *nd, struct qstr *name, |
1294 | struct path *path, struct inode **inode) | |
1da177e4 | 1295 | { |
4ac91378 | 1296 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1297 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1298 | int need_reval = 1; |
1299 | int status = 1; | |
9875cf80 DH |
1300 | int err; |
1301 | ||
b04f784e NP |
1302 | /* |
1303 | * Rename seqlock is not required here because in the off chance | |
1304 | * of a false negative due to a concurrent rename, we're going to | |
1305 | * do the non-racy lookup, below. | |
1306 | */ | |
31e6b01f NP |
1307 | if (nd->flags & LOOKUP_RCU) { |
1308 | unsigned seq; | |
12f8ad4b | 1309 | dentry = __d_lookup_rcu(parent, name, &seq, nd->inode); |
5a18fff2 AV |
1310 | if (!dentry) |
1311 | goto unlazy; | |
1312 | ||
12f8ad4b LT |
1313 | /* |
1314 | * This sequence count validates that the inode matches | |
1315 | * the dentry name information from lookup. | |
1316 | */ | |
1317 | *inode = dentry->d_inode; | |
1318 | if (read_seqcount_retry(&dentry->d_seq, seq)) | |
1319 | return -ECHILD; | |
1320 | ||
1321 | /* | |
1322 | * This sequence count validates that the parent had no | |
1323 | * changes while we did the lookup of the dentry above. | |
1324 | * | |
1325 | * The memory barrier in read_seqcount_begin of child is | |
1326 | * enough, we can use __read_seqcount_retry here. | |
1327 | */ | |
31e6b01f NP |
1328 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) |
1329 | return -ECHILD; | |
31e6b01f | 1330 | nd->seq = seq; |
5a18fff2 | 1331 | |
fa4ee159 MS |
1332 | if (unlikely(d_need_lookup(dentry))) |
1333 | goto unlazy; | |
24643087 | 1334 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
4ce16ef3 | 1335 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1336 | if (unlikely(status <= 0)) { |
1337 | if (status != -ECHILD) | |
1338 | need_reval = 0; | |
1339 | goto unlazy; | |
1340 | } | |
24643087 | 1341 | } |
31e6b01f NP |
1342 | path->mnt = mnt; |
1343 | path->dentry = dentry; | |
d6e9bd25 AV |
1344 | if (unlikely(!__follow_mount_rcu(nd, path, inode))) |
1345 | goto unlazy; | |
1346 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1347 | goto unlazy; | |
1348 | return 0; | |
5a18fff2 | 1349 | unlazy: |
19660af7 AV |
1350 | if (unlazy_walk(nd, dentry)) |
1351 | return -ECHILD; | |
5a18fff2 AV |
1352 | } else { |
1353 | dentry = __d_lookup(parent, name); | |
9875cf80 | 1354 | } |
5a18fff2 | 1355 | |
81e6f520 AV |
1356 | if (unlikely(!dentry)) |
1357 | goto need_lookup; | |
1358 | ||
1359 | if (unlikely(d_need_lookup(dentry))) { | |
44396f4b | 1360 | dput(dentry); |
81e6f520 | 1361 | goto need_lookup; |
5a18fff2 | 1362 | } |
81e6f520 | 1363 | |
5a18fff2 | 1364 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) |
4ce16ef3 | 1365 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1366 | if (unlikely(status <= 0)) { |
1367 | if (status < 0) { | |
1368 | dput(dentry); | |
1369 | return status; | |
1370 | } | |
1371 | if (!d_invalidate(dentry)) { | |
1372 | dput(dentry); | |
81e6f520 | 1373 | goto need_lookup; |
5a18fff2 | 1374 | } |
24643087 | 1375 | } |
697f514d | 1376 | |
9875cf80 DH |
1377 | path->mnt = mnt; |
1378 | path->dentry = dentry; | |
1379 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1380 | if (unlikely(err < 0)) { |
1381 | path_put_conditional(path, nd); | |
9875cf80 | 1382 | return err; |
89312214 | 1383 | } |
a3fbbde7 AV |
1384 | if (err) |
1385 | nd->flags |= LOOKUP_JUMPED; | |
9875cf80 | 1386 | *inode = path->dentry->d_inode; |
1da177e4 | 1387 | return 0; |
81e6f520 AV |
1388 | |
1389 | need_lookup: | |
697f514d MS |
1390 | return 1; |
1391 | } | |
1392 | ||
1393 | /* Fast lookup failed, do it the slow way */ | |
1394 | static int lookup_slow(struct nameidata *nd, struct qstr *name, | |
1395 | struct path *path) | |
1396 | { | |
1397 | struct dentry *dentry, *parent; | |
1398 | int err; | |
1399 | ||
1400 | parent = nd->path.dentry; | |
81e6f520 AV |
1401 | BUG_ON(nd->inode != parent->d_inode); |
1402 | ||
1403 | mutex_lock(&parent->d_inode->i_mutex); | |
72bd866a | 1404 | dentry = __lookup_hash(name, parent, nd->flags); |
81e6f520 AV |
1405 | mutex_unlock(&parent->d_inode->i_mutex); |
1406 | if (IS_ERR(dentry)) | |
1407 | return PTR_ERR(dentry); | |
697f514d MS |
1408 | path->mnt = nd->path.mnt; |
1409 | path->dentry = dentry; | |
1410 | err = follow_managed(path, nd->flags); | |
1411 | if (unlikely(err < 0)) { | |
1412 | path_put_conditional(path, nd); | |
1413 | return err; | |
1414 | } | |
1415 | if (err) | |
1416 | nd->flags |= LOOKUP_JUMPED; | |
1417 | return 0; | |
1da177e4 LT |
1418 | } |
1419 | ||
52094c8a AV |
1420 | static inline int may_lookup(struct nameidata *nd) |
1421 | { | |
1422 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1423 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1424 | if (err != -ECHILD) |
1425 | return err; | |
19660af7 | 1426 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1427 | return -ECHILD; |
1428 | } | |
4ad5abb3 | 1429 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1430 | } |
1431 | ||
9856fa1b AV |
1432 | static inline int handle_dots(struct nameidata *nd, int type) |
1433 | { | |
1434 | if (type == LAST_DOTDOT) { | |
1435 | if (nd->flags & LOOKUP_RCU) { | |
1436 | if (follow_dotdot_rcu(nd)) | |
1437 | return -ECHILD; | |
1438 | } else | |
1439 | follow_dotdot(nd); | |
1440 | } | |
1441 | return 0; | |
1442 | } | |
1443 | ||
951361f9 AV |
1444 | static void terminate_walk(struct nameidata *nd) |
1445 | { | |
1446 | if (!(nd->flags & LOOKUP_RCU)) { | |
1447 | path_put(&nd->path); | |
1448 | } else { | |
1449 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1450 | if (!(nd->flags & LOOKUP_ROOT)) |
1451 | nd->root.mnt = NULL; | |
32a7991b | 1452 | unlock_rcu_walk(); |
951361f9 AV |
1453 | } |
1454 | } | |
1455 | ||
3ddcd056 LT |
1456 | /* |
1457 | * Do we need to follow links? We _really_ want to be able | |
1458 | * to do this check without having to look at inode->i_op, | |
1459 | * so we keep a cache of "no, this doesn't need follow_link" | |
1460 | * for the common case. | |
1461 | */ | |
7813b94a | 1462 | static inline int should_follow_link(struct inode *inode, int follow) |
3ddcd056 LT |
1463 | { |
1464 | if (unlikely(!(inode->i_opflags & IOP_NOFOLLOW))) { | |
1465 | if (likely(inode->i_op->follow_link)) | |
1466 | return follow; | |
1467 | ||
1468 | /* This gets set once for the inode lifetime */ | |
1469 | spin_lock(&inode->i_lock); | |
1470 | inode->i_opflags |= IOP_NOFOLLOW; | |
1471 | spin_unlock(&inode->i_lock); | |
1472 | } | |
1473 | return 0; | |
1474 | } | |
1475 | ||
ce57dfc1 AV |
1476 | static inline int walk_component(struct nameidata *nd, struct path *path, |
1477 | struct qstr *name, int type, int follow) | |
1478 | { | |
1479 | struct inode *inode; | |
1480 | int err; | |
1481 | /* | |
1482 | * "." and ".." are special - ".." especially so because it has | |
1483 | * to be able to know about the current root directory and | |
1484 | * parent relationships. | |
1485 | */ | |
1486 | if (unlikely(type != LAST_NORM)) | |
1487 | return handle_dots(nd, type); | |
697f514d | 1488 | err = lookup_fast(nd, name, path, &inode); |
ce57dfc1 | 1489 | if (unlikely(err)) { |
697f514d MS |
1490 | if (err < 0) |
1491 | goto out_err; | |
1492 | ||
1493 | err = lookup_slow(nd, name, path); | |
1494 | if (err < 0) | |
1495 | goto out_err; | |
1496 | ||
1497 | inode = path->dentry->d_inode; | |
ce57dfc1 | 1498 | } |
697f514d MS |
1499 | err = -ENOENT; |
1500 | if (!inode) | |
1501 | goto out_path_put; | |
1502 | ||
7813b94a | 1503 | if (should_follow_link(inode, follow)) { |
19660af7 AV |
1504 | if (nd->flags & LOOKUP_RCU) { |
1505 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
697f514d MS |
1506 | err = -ECHILD; |
1507 | goto out_err; | |
19660af7 AV |
1508 | } |
1509 | } | |
ce57dfc1 AV |
1510 | BUG_ON(inode != path->dentry->d_inode); |
1511 | return 1; | |
1512 | } | |
1513 | path_to_nameidata(path, nd); | |
1514 | nd->inode = inode; | |
1515 | return 0; | |
697f514d MS |
1516 | |
1517 | out_path_put: | |
1518 | path_to_nameidata(path, nd); | |
1519 | out_err: | |
1520 | terminate_walk(nd); | |
1521 | return err; | |
ce57dfc1 AV |
1522 | } |
1523 | ||
b356379a AV |
1524 | /* |
1525 | * This limits recursive symlink follows to 8, while | |
1526 | * limiting consecutive symlinks to 40. | |
1527 | * | |
1528 | * Without that kind of total limit, nasty chains of consecutive | |
1529 | * symlinks can cause almost arbitrarily long lookups. | |
1530 | */ | |
1531 | static inline int nested_symlink(struct path *path, struct nameidata *nd) | |
1532 | { | |
1533 | int res; | |
1534 | ||
b356379a AV |
1535 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1536 | path_put_conditional(path, nd); | |
1537 | path_put(&nd->path); | |
1538 | return -ELOOP; | |
1539 | } | |
1a4022f8 | 1540 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); |
b356379a AV |
1541 | |
1542 | nd->depth++; | |
1543 | current->link_count++; | |
1544 | ||
1545 | do { | |
1546 | struct path link = *path; | |
1547 | void *cookie; | |
574197e0 AV |
1548 | |
1549 | res = follow_link(&link, nd, &cookie); | |
6d7b5aae AV |
1550 | if (res) |
1551 | break; | |
1552 | res = walk_component(nd, path, &nd->last, | |
1553 | nd->last_type, LOOKUP_FOLLOW); | |
574197e0 | 1554 | put_link(nd, &link, cookie); |
b356379a AV |
1555 | } while (res > 0); |
1556 | ||
1557 | current->link_count--; | |
1558 | nd->depth--; | |
1559 | return res; | |
1560 | } | |
1561 | ||
3ddcd056 LT |
1562 | /* |
1563 | * We really don't want to look at inode->i_op->lookup | |
1564 | * when we don't have to. So we keep a cache bit in | |
1565 | * the inode ->i_opflags field that says "yes, we can | |
1566 | * do lookup on this inode". | |
1567 | */ | |
1568 | static inline int can_lookup(struct inode *inode) | |
1569 | { | |
1570 | if (likely(inode->i_opflags & IOP_LOOKUP)) | |
1571 | return 1; | |
1572 | if (likely(!inode->i_op->lookup)) | |
1573 | return 0; | |
1574 | ||
1575 | /* We do this once for the lifetime of the inode */ | |
1576 | spin_lock(&inode->i_lock); | |
1577 | inode->i_opflags |= IOP_LOOKUP; | |
1578 | spin_unlock(&inode->i_lock); | |
1579 | return 1; | |
1580 | } | |
1581 | ||
bfcfaa77 LT |
1582 | /* |
1583 | * We can do the critical dentry name comparison and hashing | |
1584 | * operations one word at a time, but we are limited to: | |
1585 | * | |
1586 | * - Architectures with fast unaligned word accesses. We could | |
1587 | * do a "get_unaligned()" if this helps and is sufficiently | |
1588 | * fast. | |
1589 | * | |
1590 | * - Little-endian machines (so that we can generate the mask | |
1591 | * of low bytes efficiently). Again, we *could* do a byte | |
1592 | * swapping load on big-endian architectures if that is not | |
1593 | * expensive enough to make the optimization worthless. | |
1594 | * | |
1595 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we | |
1596 | * do not trap on the (extremely unlikely) case of a page | |
1597 | * crossing operation. | |
1598 | * | |
1599 | * - Furthermore, we need an efficient 64-bit compile for the | |
1600 | * 64-bit case in order to generate the "number of bytes in | |
1601 | * the final mask". Again, that could be replaced with a | |
1602 | * efficient population count instruction or similar. | |
1603 | */ | |
1604 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1605 | ||
f68e556e | 1606 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1607 | |
f68e556e | 1608 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1609 | |
1610 | static inline unsigned int fold_hash(unsigned long hash) | |
1611 | { | |
1612 | hash += hash >> (8*sizeof(int)); | |
1613 | return hash; | |
1614 | } | |
1615 | ||
1616 | #else /* 32-bit case */ | |
1617 | ||
bfcfaa77 LT |
1618 | #define fold_hash(x) (x) |
1619 | ||
1620 | #endif | |
1621 | ||
1622 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1623 | { | |
1624 | unsigned long a, mask; | |
1625 | unsigned long hash = 0; | |
1626 | ||
1627 | for (;;) { | |
e419b4cc | 1628 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1629 | if (len < sizeof(unsigned long)) |
1630 | break; | |
1631 | hash += a; | |
f132c5be | 1632 | hash *= 9; |
bfcfaa77 LT |
1633 | name += sizeof(unsigned long); |
1634 | len -= sizeof(unsigned long); | |
1635 | if (!len) | |
1636 | goto done; | |
1637 | } | |
1638 | mask = ~(~0ul << len*8); | |
1639 | hash += mask & a; | |
1640 | done: | |
1641 | return fold_hash(hash); | |
1642 | } | |
1643 | EXPORT_SYMBOL(full_name_hash); | |
1644 | ||
bfcfaa77 LT |
1645 | /* |
1646 | * Calculate the length and hash of the path component, and | |
1647 | * return the length of the component; | |
1648 | */ | |
1649 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1650 | { | |
36126f8f LT |
1651 | unsigned long a, b, adata, bdata, mask, hash, len; |
1652 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; | |
bfcfaa77 LT |
1653 | |
1654 | hash = a = 0; | |
1655 | len = -sizeof(unsigned long); | |
1656 | do { | |
1657 | hash = (hash + a) * 9; | |
1658 | len += sizeof(unsigned long); | |
e419b4cc | 1659 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1660 | b = a ^ REPEAT_BYTE('/'); |
1661 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1662 | ||
1663 | adata = prep_zero_mask(a, adata, &constants); | |
1664 | bdata = prep_zero_mask(b, bdata, &constants); | |
1665 | ||
1666 | mask = create_zero_mask(adata | bdata); | |
1667 | ||
1668 | hash += a & zero_bytemask(mask); | |
bfcfaa77 LT |
1669 | *hashp = fold_hash(hash); |
1670 | ||
36126f8f | 1671 | return len + find_zero(mask); |
bfcfaa77 LT |
1672 | } |
1673 | ||
1674 | #else | |
1675 | ||
0145acc2 LT |
1676 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1677 | { | |
1678 | unsigned long hash = init_name_hash(); | |
1679 | while (len--) | |
1680 | hash = partial_name_hash(*name++, hash); | |
1681 | return end_name_hash(hash); | |
1682 | } | |
ae942ae7 | 1683 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1684 | |
200e9ef7 LT |
1685 | /* |
1686 | * We know there's a real path component here of at least | |
1687 | * one character. | |
1688 | */ | |
1689 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1690 | { | |
1691 | unsigned long hash = init_name_hash(); | |
1692 | unsigned long len = 0, c; | |
1693 | ||
1694 | c = (unsigned char)*name; | |
1695 | do { | |
1696 | len++; | |
1697 | hash = partial_name_hash(c, hash); | |
1698 | c = (unsigned char)name[len]; | |
1699 | } while (c && c != '/'); | |
1700 | *hashp = end_name_hash(hash); | |
1701 | return len; | |
1702 | } | |
1703 | ||
bfcfaa77 LT |
1704 | #endif |
1705 | ||
1da177e4 LT |
1706 | /* |
1707 | * Name resolution. | |
ea3834d9 PM |
1708 | * This is the basic name resolution function, turning a pathname into |
1709 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1710 | * |
ea3834d9 PM |
1711 | * Returns 0 and nd will have valid dentry and mnt on success. |
1712 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1713 | */ |
6de88d72 | 1714 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1715 | { |
1716 | struct path next; | |
1da177e4 | 1717 | int err; |
1da177e4 LT |
1718 | |
1719 | while (*name=='/') | |
1720 | name++; | |
1721 | if (!*name) | |
086e183a | 1722 | return 0; |
1da177e4 | 1723 | |
1da177e4 LT |
1724 | /* At this point we know we have a real path component. */ |
1725 | for(;;) { | |
1da177e4 | 1726 | struct qstr this; |
200e9ef7 | 1727 | long len; |
fe479a58 | 1728 | int type; |
1da177e4 | 1729 | |
52094c8a | 1730 | err = may_lookup(nd); |
1da177e4 LT |
1731 | if (err) |
1732 | break; | |
1733 | ||
200e9ef7 | 1734 | len = hash_name(name, &this.hash); |
1da177e4 | 1735 | this.name = name; |
200e9ef7 | 1736 | this.len = len; |
1da177e4 | 1737 | |
fe479a58 | 1738 | type = LAST_NORM; |
200e9ef7 | 1739 | if (name[0] == '.') switch (len) { |
fe479a58 | 1740 | case 2: |
200e9ef7 | 1741 | if (name[1] == '.') { |
fe479a58 | 1742 | type = LAST_DOTDOT; |
16c2cd71 AV |
1743 | nd->flags |= LOOKUP_JUMPED; |
1744 | } | |
fe479a58 AV |
1745 | break; |
1746 | case 1: | |
1747 | type = LAST_DOT; | |
1748 | } | |
5a202bcd AV |
1749 | if (likely(type == LAST_NORM)) { |
1750 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1751 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd AV |
1752 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
1753 | err = parent->d_op->d_hash(parent, nd->inode, | |
1754 | &this); | |
1755 | if (err < 0) | |
1756 | break; | |
1757 | } | |
1758 | } | |
fe479a58 | 1759 | |
200e9ef7 | 1760 | if (!name[len]) |
1da177e4 | 1761 | goto last_component; |
200e9ef7 LT |
1762 | /* |
1763 | * If it wasn't NUL, we know it was '/'. Skip that | |
1764 | * slash, and continue until no more slashes. | |
1765 | */ | |
1766 | do { | |
1767 | len++; | |
1768 | } while (unlikely(name[len] == '/')); | |
1769 | if (!name[len]) | |
b356379a | 1770 | goto last_component; |
200e9ef7 | 1771 | name += len; |
1da177e4 | 1772 | |
ce57dfc1 AV |
1773 | err = walk_component(nd, &next, &this, type, LOOKUP_FOLLOW); |
1774 | if (err < 0) | |
1775 | return err; | |
1da177e4 | 1776 | |
ce57dfc1 | 1777 | if (err) { |
b356379a | 1778 | err = nested_symlink(&next, nd); |
1da177e4 | 1779 | if (err) |
a7472bab | 1780 | return err; |
31e6b01f | 1781 | } |
3ddcd056 LT |
1782 | if (can_lookup(nd->inode)) |
1783 | continue; | |
1da177e4 | 1784 | err = -ENOTDIR; |
3ddcd056 | 1785 | break; |
1da177e4 LT |
1786 | /* here ends the main loop */ |
1787 | ||
1da177e4 | 1788 | last_component: |
b356379a AV |
1789 | nd->last = this; |
1790 | nd->last_type = type; | |
086e183a | 1791 | return 0; |
1da177e4 | 1792 | } |
951361f9 | 1793 | terminate_walk(nd); |
1da177e4 LT |
1794 | return err; |
1795 | } | |
1796 | ||
70e9b357 AV |
1797 | static int path_init(int dfd, const char *name, unsigned int flags, |
1798 | struct nameidata *nd, struct file **fp) | |
31e6b01f NP |
1799 | { |
1800 | int retval = 0; | |
31e6b01f NP |
1801 | |
1802 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
16c2cd71 | 1803 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 1804 | nd->depth = 0; |
5b6ca027 AV |
1805 | if (flags & LOOKUP_ROOT) { |
1806 | struct inode *inode = nd->root.dentry->d_inode; | |
73d049a4 AV |
1807 | if (*name) { |
1808 | if (!inode->i_op->lookup) | |
1809 | return -ENOTDIR; | |
1810 | retval = inode_permission(inode, MAY_EXEC); | |
1811 | if (retval) | |
1812 | return retval; | |
1813 | } | |
5b6ca027 AV |
1814 | nd->path = nd->root; |
1815 | nd->inode = inode; | |
1816 | if (flags & LOOKUP_RCU) { | |
32a7991b | 1817 | lock_rcu_walk(); |
5b6ca027 AV |
1818 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
1819 | } else { | |
1820 | path_get(&nd->path); | |
1821 | } | |
1822 | return 0; | |
1823 | } | |
1824 | ||
31e6b01f | 1825 | nd->root.mnt = NULL; |
31e6b01f NP |
1826 | |
1827 | if (*name=='/') { | |
e41f7d4e | 1828 | if (flags & LOOKUP_RCU) { |
32a7991b | 1829 | lock_rcu_walk(); |
e41f7d4e AV |
1830 | set_root_rcu(nd); |
1831 | } else { | |
1832 | set_root(nd); | |
1833 | path_get(&nd->root); | |
1834 | } | |
1835 | nd->path = nd->root; | |
31e6b01f | 1836 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1837 | if (flags & LOOKUP_RCU) { |
1838 | struct fs_struct *fs = current->fs; | |
1839 | unsigned seq; | |
31e6b01f | 1840 | |
32a7991b | 1841 | lock_rcu_walk(); |
c28cc364 | 1842 | |
e41f7d4e AV |
1843 | do { |
1844 | seq = read_seqcount_begin(&fs->seq); | |
1845 | nd->path = fs->pwd; | |
1846 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1847 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1848 | } else { | |
1849 | get_fs_pwd(current->fs, &nd->path); | |
1850 | } | |
31e6b01f | 1851 | } else { |
2903ff01 | 1852 | struct fd f = fdget_raw(dfd); |
31e6b01f NP |
1853 | struct dentry *dentry; |
1854 | ||
2903ff01 AV |
1855 | if (!f.file) |
1856 | return -EBADF; | |
31e6b01f | 1857 | |
2903ff01 | 1858 | dentry = f.file->f_path.dentry; |
31e6b01f | 1859 | |
f52e0c11 | 1860 | if (*name) { |
2903ff01 AV |
1861 | if (!S_ISDIR(dentry->d_inode->i_mode)) { |
1862 | fdput(f); | |
1863 | return -ENOTDIR; | |
1864 | } | |
31e6b01f | 1865 | |
4ad5abb3 | 1866 | retval = inode_permission(dentry->d_inode, MAY_EXEC); |
2903ff01 AV |
1867 | if (retval) { |
1868 | fdput(f); | |
1869 | return retval; | |
1870 | } | |
f52e0c11 | 1871 | } |
31e6b01f | 1872 | |
2903ff01 | 1873 | nd->path = f.file->f_path; |
e41f7d4e | 1874 | if (flags & LOOKUP_RCU) { |
2903ff01 AV |
1875 | if (f.need_put) |
1876 | *fp = f.file; | |
e41f7d4e | 1877 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
32a7991b | 1878 | lock_rcu_walk(); |
e41f7d4e | 1879 | } else { |
2903ff01 AV |
1880 | path_get(&nd->path); |
1881 | fdput(f); | |
e41f7d4e | 1882 | } |
31e6b01f | 1883 | } |
31e6b01f | 1884 | |
31e6b01f | 1885 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1886 | return 0; |
9b4a9b14 AV |
1887 | } |
1888 | ||
bd92d7fe AV |
1889 | static inline int lookup_last(struct nameidata *nd, struct path *path) |
1890 | { | |
1891 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1892 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1893 | ||
1894 | nd->flags &= ~LOOKUP_PARENT; | |
1895 | return walk_component(nd, path, &nd->last, nd->last_type, | |
1896 | nd->flags & LOOKUP_FOLLOW); | |
1897 | } | |
1898 | ||
9b4a9b14 | 1899 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
ee0827cd | 1900 | static int path_lookupat(int dfd, const char *name, |
9b4a9b14 AV |
1901 | unsigned int flags, struct nameidata *nd) |
1902 | { | |
70e9b357 | 1903 | struct file *base = NULL; |
bd92d7fe AV |
1904 | struct path path; |
1905 | int err; | |
31e6b01f NP |
1906 | |
1907 | /* | |
1908 | * Path walking is largely split up into 2 different synchronisation | |
1909 | * schemes, rcu-walk and ref-walk (explained in | |
1910 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1911 | * path walk code, but some things particularly setup, cleanup, and | |
1912 | * following mounts are sufficiently divergent that functions are | |
1913 | * duplicated. Typically there is a function foo(), and its RCU | |
1914 | * analogue, foo_rcu(). | |
1915 | * | |
1916 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1917 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1918 | * be handled by restarting a traditional ref-walk (which will always | |
1919 | * be able to complete). | |
1920 | */ | |
bd92d7fe | 1921 | err = path_init(dfd, name, flags | LOOKUP_PARENT, nd, &base); |
ee0827cd | 1922 | |
bd92d7fe AV |
1923 | if (unlikely(err)) |
1924 | return err; | |
ee0827cd AV |
1925 | |
1926 | current->total_link_count = 0; | |
bd92d7fe AV |
1927 | err = link_path_walk(name, nd); |
1928 | ||
1929 | if (!err && !(flags & LOOKUP_PARENT)) { | |
bd92d7fe AV |
1930 | err = lookup_last(nd, &path); |
1931 | while (err > 0) { | |
1932 | void *cookie; | |
1933 | struct path link = path; | |
800179c9 KC |
1934 | err = may_follow_link(&link, nd); |
1935 | if (unlikely(err)) | |
1936 | break; | |
bd92d7fe | 1937 | nd->flags |= LOOKUP_PARENT; |
574197e0 | 1938 | err = follow_link(&link, nd, &cookie); |
6d7b5aae AV |
1939 | if (err) |
1940 | break; | |
1941 | err = lookup_last(nd, &path); | |
574197e0 | 1942 | put_link(nd, &link, cookie); |
bd92d7fe AV |
1943 | } |
1944 | } | |
ee0827cd | 1945 | |
9f1fafee AV |
1946 | if (!err) |
1947 | err = complete_walk(nd); | |
bd92d7fe AV |
1948 | |
1949 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
1950 | if (!nd->inode->i_op->lookup) { | |
1951 | path_put(&nd->path); | |
bd23a539 | 1952 | err = -ENOTDIR; |
bd92d7fe AV |
1953 | } |
1954 | } | |
16c2cd71 | 1955 | |
70e9b357 AV |
1956 | if (base) |
1957 | fput(base); | |
ee0827cd | 1958 | |
5b6ca027 | 1959 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
2a737871 AV |
1960 | path_put(&nd->root); |
1961 | nd->root.mnt = NULL; | |
1962 | } | |
bd92d7fe | 1963 | return err; |
ee0827cd | 1964 | } |
31e6b01f | 1965 | |
ee0827cd AV |
1966 | static int do_path_lookup(int dfd, const char *name, |
1967 | unsigned int flags, struct nameidata *nd) | |
1968 | { | |
1969 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); | |
1970 | if (unlikely(retval == -ECHILD)) | |
1971 | retval = path_lookupat(dfd, name, flags, nd); | |
1972 | if (unlikely(retval == -ESTALE)) | |
1973 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); | |
31e6b01f | 1974 | |
f78570dd JL |
1975 | if (likely(!retval)) |
1976 | audit_inode(name, nd->path.dentry); | |
170aa3d0 | 1977 | return retval; |
1da177e4 LT |
1978 | } |
1979 | ||
79714f72 AV |
1980 | /* does lookup, returns the object with parent locked */ |
1981 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 1982 | { |
79714f72 AV |
1983 | struct nameidata nd; |
1984 | struct dentry *d; | |
1985 | int err = do_path_lookup(AT_FDCWD, name, LOOKUP_PARENT, &nd); | |
1986 | if (err) | |
1987 | return ERR_PTR(err); | |
1988 | if (nd.last_type != LAST_NORM) { | |
1989 | path_put(&nd.path); | |
1990 | return ERR_PTR(-EINVAL); | |
1991 | } | |
1992 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); | |
1e0ea001 | 1993 | d = __lookup_hash(&nd.last, nd.path.dentry, 0); |
79714f72 AV |
1994 | if (IS_ERR(d)) { |
1995 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); | |
1996 | path_put(&nd.path); | |
1997 | return d; | |
1998 | } | |
1999 | *path = nd.path; | |
2000 | return d; | |
5590ff0d UD |
2001 | } |
2002 | ||
d1811465 AV |
2003 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2004 | { | |
2005 | struct nameidata nd; | |
2006 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
2007 | if (!res) | |
2008 | *path = nd.path; | |
2009 | return res; | |
2010 | } | |
2011 | ||
16f18200 JJS |
2012 | /** |
2013 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2014 | * @dentry: pointer to dentry of the base directory | |
2015 | * @mnt: pointer to vfs mount of the base directory | |
2016 | * @name: pointer to file name | |
2017 | * @flags: lookup flags | |
e0a01249 | 2018 | * @path: pointer to struct path to fill |
16f18200 JJS |
2019 | */ |
2020 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2021 | const char *name, unsigned int flags, | |
e0a01249 | 2022 | struct path *path) |
16f18200 | 2023 | { |
e0a01249 AV |
2024 | struct nameidata nd; |
2025 | int err; | |
2026 | nd.root.dentry = dentry; | |
2027 | nd.root.mnt = mnt; | |
2028 | BUG_ON(flags & LOOKUP_PARENT); | |
5b6ca027 | 2029 | /* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */ |
e0a01249 AV |
2030 | err = do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, &nd); |
2031 | if (!err) | |
2032 | *path = nd.path; | |
2033 | return err; | |
16f18200 JJS |
2034 | } |
2035 | ||
057f6c01 JM |
2036 | /* |
2037 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
2038 | * needs parent already locked. Doesn't follow mounts. | |
2039 | * SMP-safe. | |
2040 | */ | |
eead1911 | 2041 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 2042 | { |
72bd866a | 2043 | return __lookup_hash(&nd->last, nd->path.dentry, nd->flags); |
1da177e4 LT |
2044 | } |
2045 | ||
eead1911 | 2046 | /** |
a6b91919 | 2047 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
2048 | * @name: pathname component to lookup |
2049 | * @base: base directory to lookup from | |
2050 | * @len: maximum length @len should be interpreted to | |
2051 | * | |
a6b91919 RD |
2052 | * Note that this routine is purely a helper for filesystem usage and should |
2053 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
2054 | * nameidata argument is passed to the filesystem methods and a filesystem |
2055 | * using this helper needs to be prepared for that. | |
2056 | */ | |
057f6c01 JM |
2057 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
2058 | { | |
057f6c01 | 2059 | struct qstr this; |
6a96ba54 | 2060 | unsigned int c; |
cda309de | 2061 | int err; |
057f6c01 | 2062 | |
2f9092e1 DW |
2063 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
2064 | ||
6a96ba54 AV |
2065 | this.name = name; |
2066 | this.len = len; | |
0145acc2 | 2067 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
2068 | if (!len) |
2069 | return ERR_PTR(-EACCES); | |
2070 | ||
6a96ba54 AV |
2071 | while (len--) { |
2072 | c = *(const unsigned char *)name++; | |
2073 | if (c == '/' || c == '\0') | |
2074 | return ERR_PTR(-EACCES); | |
6a96ba54 | 2075 | } |
5a202bcd AV |
2076 | /* |
2077 | * See if the low-level filesystem might want | |
2078 | * to use its own hash.. | |
2079 | */ | |
2080 | if (base->d_flags & DCACHE_OP_HASH) { | |
2081 | int err = base->d_op->d_hash(base, base->d_inode, &this); | |
2082 | if (err < 0) | |
2083 | return ERR_PTR(err); | |
2084 | } | |
eead1911 | 2085 | |
cda309de MS |
2086 | err = inode_permission(base->d_inode, MAY_EXEC); |
2087 | if (err) | |
2088 | return ERR_PTR(err); | |
2089 | ||
72bd866a | 2090 | return __lookup_hash(&this, base, 0); |
057f6c01 JM |
2091 | } |
2092 | ||
1fa1e7f6 AW |
2093 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2094 | struct path *path, int *empty) | |
1da177e4 | 2095 | { |
2d8f3038 | 2096 | struct nameidata nd; |
1fa1e7f6 | 2097 | char *tmp = getname_flags(name, flags, empty); |
1da177e4 | 2098 | int err = PTR_ERR(tmp); |
1da177e4 | 2099 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
2100 | |
2101 | BUG_ON(flags & LOOKUP_PARENT); | |
2102 | ||
2103 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 2104 | putname(tmp); |
2d8f3038 AV |
2105 | if (!err) |
2106 | *path = nd.path; | |
1da177e4 LT |
2107 | } |
2108 | return err; | |
2109 | } | |
2110 | ||
1fa1e7f6 AW |
2111 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
2112 | struct path *path) | |
2113 | { | |
f7493e5d | 2114 | return user_path_at_empty(dfd, name, flags, path, NULL); |
1fa1e7f6 AW |
2115 | } |
2116 | ||
2ad94ae6 AV |
2117 | static int user_path_parent(int dfd, const char __user *path, |
2118 | struct nameidata *nd, char **name) | |
2119 | { | |
2120 | char *s = getname(path); | |
2121 | int error; | |
2122 | ||
2123 | if (IS_ERR(s)) | |
2124 | return PTR_ERR(s); | |
2125 | ||
2126 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
2127 | if (error) | |
2128 | putname(s); | |
2129 | else | |
2130 | *name = s; | |
2131 | ||
2132 | return error; | |
2133 | } | |
2134 | ||
1da177e4 LT |
2135 | /* |
2136 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
2137 | * minimal. | |
2138 | */ | |
2139 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
2140 | { | |
8e96e3b7 | 2141 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2142 | |
1da177e4 LT |
2143 | if (!(dir->i_mode & S_ISVTX)) |
2144 | return 0; | |
8e96e3b7 | 2145 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2146 | return 0; |
8e96e3b7 | 2147 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2148 | return 0; |
1a48e2ac | 2149 | return !inode_capable(inode, CAP_FOWNER); |
1da177e4 LT |
2150 | } |
2151 | ||
2152 | /* | |
2153 | * Check whether we can remove a link victim from directory dir, check | |
2154 | * whether the type of victim is right. | |
2155 | * 1. We can't do it if dir is read-only (done in permission()) | |
2156 | * 2. We should have write and exec permissions on dir | |
2157 | * 3. We can't remove anything from append-only dir | |
2158 | * 4. We can't do anything with immutable dir (done in permission()) | |
2159 | * 5. If the sticky bit on dir is set we should either | |
2160 | * a. be owner of dir, or | |
2161 | * b. be owner of victim, or | |
2162 | * c. have CAP_FOWNER capability | |
2163 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2164 | * links pointing to it. | |
2165 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2166 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2167 | * 9. We can't remove a root or mountpoint. | |
2168 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
2169 | * nfs_async_unlink(). | |
2170 | */ | |
858119e1 | 2171 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
2172 | { |
2173 | int error; | |
2174 | ||
2175 | if (!victim->d_inode) | |
2176 | return -ENOENT; | |
2177 | ||
2178 | BUG_ON(victim->d_parent->d_inode != dir); | |
c43a25ab | 2179 | audit_inode_child(dir, victim); |
1da177e4 | 2180 | |
f419a2e3 | 2181 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2182 | if (error) |
2183 | return error; | |
2184 | if (IS_APPEND(dir)) | |
2185 | return -EPERM; | |
2186 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 2187 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
2188 | return -EPERM; |
2189 | if (isdir) { | |
2190 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
2191 | return -ENOTDIR; | |
2192 | if (IS_ROOT(victim)) | |
2193 | return -EBUSY; | |
2194 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
2195 | return -EISDIR; | |
2196 | if (IS_DEADDIR(dir)) | |
2197 | return -ENOENT; | |
2198 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2199 | return -EBUSY; | |
2200 | return 0; | |
2201 | } | |
2202 | ||
2203 | /* Check whether we can create an object with dentry child in directory | |
2204 | * dir. | |
2205 | * 1. We can't do it if child already exists (open has special treatment for | |
2206 | * this case, but since we are inlined it's OK) | |
2207 | * 2. We can't do it if dir is read-only (done in permission()) | |
2208 | * 3. We should have write and exec permissions on dir | |
2209 | * 4. We can't do it if dir is immutable (done in permission()) | |
2210 | */ | |
a95164d9 | 2211 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
2212 | { |
2213 | if (child->d_inode) | |
2214 | return -EEXIST; | |
2215 | if (IS_DEADDIR(dir)) | |
2216 | return -ENOENT; | |
f419a2e3 | 2217 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2218 | } |
2219 | ||
1da177e4 LT |
2220 | /* |
2221 | * p1 and p2 should be directories on the same fs. | |
2222 | */ | |
2223 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2224 | { | |
2225 | struct dentry *p; | |
2226 | ||
2227 | if (p1 == p2) { | |
f2eace23 | 2228 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2229 | return NULL; |
2230 | } | |
2231 | ||
a11f3a05 | 2232 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2233 | |
e2761a11 OH |
2234 | p = d_ancestor(p2, p1); |
2235 | if (p) { | |
2236 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2237 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2238 | return p; | |
1da177e4 LT |
2239 | } |
2240 | ||
e2761a11 OH |
2241 | p = d_ancestor(p1, p2); |
2242 | if (p) { | |
2243 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2244 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2245 | return p; | |
1da177e4 LT |
2246 | } |
2247 | ||
f2eace23 IM |
2248 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
2249 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
2250 | return NULL; |
2251 | } | |
2252 | ||
2253 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2254 | { | |
1b1dcc1b | 2255 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2256 | if (p1 != p2) { |
1b1dcc1b | 2257 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2258 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2259 | } |
2260 | } | |
2261 | ||
4acdaf27 | 2262 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2263 | bool want_excl) |
1da177e4 | 2264 | { |
a95164d9 | 2265 | int error = may_create(dir, dentry); |
1da177e4 LT |
2266 | if (error) |
2267 | return error; | |
2268 | ||
acfa4380 | 2269 | if (!dir->i_op->create) |
1da177e4 LT |
2270 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2271 | mode &= S_IALLUGO; | |
2272 | mode |= S_IFREG; | |
2273 | error = security_inode_create(dir, dentry, mode); | |
2274 | if (error) | |
2275 | return error; | |
312b63fb | 2276 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2277 | if (!error) |
f38aa942 | 2278 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2279 | return error; |
2280 | } | |
2281 | ||
73d049a4 | 2282 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2283 | { |
3fb64190 | 2284 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2285 | struct inode *inode = dentry->d_inode; |
2286 | int error; | |
2287 | ||
bcda7652 AV |
2288 | /* O_PATH? */ |
2289 | if (!acc_mode) | |
2290 | return 0; | |
2291 | ||
1da177e4 LT |
2292 | if (!inode) |
2293 | return -ENOENT; | |
2294 | ||
c8fe8f30 CH |
2295 | switch (inode->i_mode & S_IFMT) { |
2296 | case S_IFLNK: | |
1da177e4 | 2297 | return -ELOOP; |
c8fe8f30 CH |
2298 | case S_IFDIR: |
2299 | if (acc_mode & MAY_WRITE) | |
2300 | return -EISDIR; | |
2301 | break; | |
2302 | case S_IFBLK: | |
2303 | case S_IFCHR: | |
3fb64190 | 2304 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2305 | return -EACCES; |
c8fe8f30 CH |
2306 | /*FALLTHRU*/ |
2307 | case S_IFIFO: | |
2308 | case S_IFSOCK: | |
1da177e4 | 2309 | flag &= ~O_TRUNC; |
c8fe8f30 | 2310 | break; |
4a3fd211 | 2311 | } |
b41572e9 | 2312 | |
3fb64190 | 2313 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2314 | if (error) |
2315 | return error; | |
6146f0d5 | 2316 | |
1da177e4 LT |
2317 | /* |
2318 | * An append-only file must be opened in append mode for writing. | |
2319 | */ | |
2320 | if (IS_APPEND(inode)) { | |
8737c930 | 2321 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2322 | return -EPERM; |
1da177e4 | 2323 | if (flag & O_TRUNC) |
7715b521 | 2324 | return -EPERM; |
1da177e4 LT |
2325 | } |
2326 | ||
2327 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2328 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2329 | return -EPERM; |
1da177e4 | 2330 | |
f3c7691e | 2331 | return 0; |
7715b521 | 2332 | } |
1da177e4 | 2333 | |
e1181ee6 | 2334 | static int handle_truncate(struct file *filp) |
7715b521 | 2335 | { |
e1181ee6 | 2336 | struct path *path = &filp->f_path; |
7715b521 AV |
2337 | struct inode *inode = path->dentry->d_inode; |
2338 | int error = get_write_access(inode); | |
2339 | if (error) | |
2340 | return error; | |
2341 | /* | |
2342 | * Refuse to truncate files with mandatory locks held on them. | |
2343 | */ | |
2344 | error = locks_verify_locked(inode); | |
2345 | if (!error) | |
ea0d3ab2 | 2346 | error = security_path_truncate(path); |
7715b521 AV |
2347 | if (!error) { |
2348 | error = do_truncate(path->dentry, 0, | |
2349 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2350 | filp); |
7715b521 AV |
2351 | } |
2352 | put_write_access(inode); | |
acd0c935 | 2353 | return error; |
1da177e4 LT |
2354 | } |
2355 | ||
d57999e1 DH |
2356 | static inline int open_to_namei_flags(int flag) |
2357 | { | |
8a5e929d AV |
2358 | if ((flag & O_ACCMODE) == 3) |
2359 | flag--; | |
d57999e1 DH |
2360 | return flag; |
2361 | } | |
2362 | ||
d18e9008 MS |
2363 | static int may_o_create(struct path *dir, struct dentry *dentry, umode_t mode) |
2364 | { | |
2365 | int error = security_path_mknod(dir, dentry, mode, 0); | |
2366 | if (error) | |
2367 | return error; | |
2368 | ||
2369 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); | |
2370 | if (error) | |
2371 | return error; | |
2372 | ||
2373 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2374 | } | |
2375 | ||
1acf0af9 DH |
2376 | /* |
2377 | * Attempt to atomically look up, create and open a file from a negative | |
2378 | * dentry. | |
2379 | * | |
2380 | * Returns 0 if successful. The file will have been created and attached to | |
2381 | * @file by the filesystem calling finish_open(). | |
2382 | * | |
2383 | * Returns 1 if the file was looked up only or didn't need creating. The | |
2384 | * caller will need to perform the open themselves. @path will have been | |
2385 | * updated to point to the new dentry. This may be negative. | |
2386 | * | |
2387 | * Returns an error code otherwise. | |
2388 | */ | |
2675a4eb AV |
2389 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
2390 | struct path *path, struct file *file, | |
2391 | const struct open_flags *op, | |
64894cf8 | 2392 | bool got_write, bool need_lookup, |
2675a4eb | 2393 | int *opened) |
d18e9008 MS |
2394 | { |
2395 | struct inode *dir = nd->path.dentry->d_inode; | |
2396 | unsigned open_flag = open_to_namei_flags(op->open_flag); | |
2397 | umode_t mode; | |
2398 | int error; | |
2399 | int acc_mode; | |
d18e9008 MS |
2400 | int create_error = 0; |
2401 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; | |
2402 | ||
2403 | BUG_ON(dentry->d_inode); | |
2404 | ||
2405 | /* Don't create child dentry for a dead directory. */ | |
2406 | if (unlikely(IS_DEADDIR(dir))) { | |
2675a4eb | 2407 | error = -ENOENT; |
d18e9008 MS |
2408 | goto out; |
2409 | } | |
2410 | ||
62b259d8 | 2411 | mode = op->mode; |
d18e9008 MS |
2412 | if ((open_flag & O_CREAT) && !IS_POSIXACL(dir)) |
2413 | mode &= ~current_umask(); | |
2414 | ||
f8310c59 | 2415 | if ((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT)) { |
d18e9008 | 2416 | open_flag &= ~O_TRUNC; |
47237687 | 2417 | *opened |= FILE_CREATED; |
d18e9008 MS |
2418 | } |
2419 | ||
2420 | /* | |
2421 | * Checking write permission is tricky, bacuse we don't know if we are | |
2422 | * going to actually need it: O_CREAT opens should work as long as the | |
2423 | * file exists. But checking existence breaks atomicity. The trick is | |
2424 | * to check access and if not granted clear O_CREAT from the flags. | |
2425 | * | |
2426 | * Another problem is returing the "right" error value (e.g. for an | |
2427 | * O_EXCL open we want to return EEXIST not EROFS). | |
2428 | */ | |
64894cf8 AV |
2429 | if (((open_flag & (O_CREAT | O_TRUNC)) || |
2430 | (open_flag & O_ACCMODE) != O_RDONLY) && unlikely(!got_write)) { | |
2431 | if (!(open_flag & O_CREAT)) { | |
d18e9008 MS |
2432 | /* |
2433 | * No O_CREATE -> atomicity not a requirement -> fall | |
2434 | * back to lookup + open | |
2435 | */ | |
2436 | goto no_open; | |
2437 | } else if (open_flag & (O_EXCL | O_TRUNC)) { | |
2438 | /* Fall back and fail with the right error */ | |
64894cf8 | 2439 | create_error = -EROFS; |
d18e9008 MS |
2440 | goto no_open; |
2441 | } else { | |
2442 | /* No side effects, safe to clear O_CREAT */ | |
64894cf8 | 2443 | create_error = -EROFS; |
d18e9008 MS |
2444 | open_flag &= ~O_CREAT; |
2445 | } | |
2446 | } | |
2447 | ||
2448 | if (open_flag & O_CREAT) { | |
38227f78 | 2449 | error = may_o_create(&nd->path, dentry, mode); |
d18e9008 MS |
2450 | if (error) { |
2451 | create_error = error; | |
2452 | if (open_flag & O_EXCL) | |
2453 | goto no_open; | |
2454 | open_flag &= ~O_CREAT; | |
2455 | } | |
2456 | } | |
2457 | ||
2458 | if (nd->flags & LOOKUP_DIRECTORY) | |
2459 | open_flag |= O_DIRECTORY; | |
2460 | ||
30d90494 AV |
2461 | file->f_path.dentry = DENTRY_NOT_SET; |
2462 | file->f_path.mnt = nd->path.mnt; | |
2463 | error = dir->i_op->atomic_open(dir, dentry, file, open_flag, mode, | |
47237687 | 2464 | opened); |
d9585277 | 2465 | if (error < 0) { |
d9585277 AV |
2466 | if (create_error && error == -ENOENT) |
2467 | error = create_error; | |
d18e9008 MS |
2468 | goto out; |
2469 | } | |
2470 | ||
2471 | acc_mode = op->acc_mode; | |
47237687 | 2472 | if (*opened & FILE_CREATED) { |
d18e9008 MS |
2473 | fsnotify_create(dir, dentry); |
2474 | acc_mode = MAY_OPEN; | |
2475 | } | |
2476 | ||
d9585277 | 2477 | if (error) { /* returned 1, that is */ |
30d90494 | 2478 | if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 2479 | error = -EIO; |
d18e9008 MS |
2480 | goto out; |
2481 | } | |
30d90494 | 2482 | if (file->f_path.dentry) { |
d18e9008 | 2483 | dput(dentry); |
30d90494 | 2484 | dentry = file->f_path.dentry; |
d18e9008 | 2485 | } |
62b2ce96 SW |
2486 | if (create_error && dentry->d_inode == NULL) { |
2487 | error = create_error; | |
2488 | goto out; | |
2489 | } | |
d18e9008 MS |
2490 | goto looked_up; |
2491 | } | |
2492 | ||
2493 | /* | |
2494 | * We didn't have the inode before the open, so check open permission | |
2495 | * here. | |
2496 | */ | |
2675a4eb AV |
2497 | error = may_open(&file->f_path, acc_mode, open_flag); |
2498 | if (error) | |
2499 | fput(file); | |
d18e9008 MS |
2500 | |
2501 | out: | |
2502 | dput(dentry); | |
2675a4eb | 2503 | return error; |
d18e9008 | 2504 | |
d18e9008 MS |
2505 | no_open: |
2506 | if (need_lookup) { | |
72bd866a | 2507 | dentry = lookup_real(dir, dentry, nd->flags); |
d18e9008 | 2508 | if (IS_ERR(dentry)) |
2675a4eb | 2509 | return PTR_ERR(dentry); |
d18e9008 MS |
2510 | |
2511 | if (create_error) { | |
2512 | int open_flag = op->open_flag; | |
2513 | ||
2675a4eb | 2514 | error = create_error; |
d18e9008 MS |
2515 | if ((open_flag & O_EXCL)) { |
2516 | if (!dentry->d_inode) | |
2517 | goto out; | |
2518 | } else if (!dentry->d_inode) { | |
2519 | goto out; | |
2520 | } else if ((open_flag & O_TRUNC) && | |
2521 | S_ISREG(dentry->d_inode->i_mode)) { | |
2522 | goto out; | |
2523 | } | |
2524 | /* will fail later, go on to get the right error */ | |
2525 | } | |
2526 | } | |
2527 | looked_up: | |
2528 | path->dentry = dentry; | |
2529 | path->mnt = nd->path.mnt; | |
2675a4eb | 2530 | return 1; |
d18e9008 MS |
2531 | } |
2532 | ||
d58ffd35 | 2533 | /* |
1acf0af9 | 2534 | * Look up and maybe create and open the last component. |
d58ffd35 MS |
2535 | * |
2536 | * Must be called with i_mutex held on parent. | |
2537 | * | |
1acf0af9 DH |
2538 | * Returns 0 if the file was successfully atomically created (if necessary) and |
2539 | * opened. In this case the file will be returned attached to @file. | |
2540 | * | |
2541 | * Returns 1 if the file was not completely opened at this time, though lookups | |
2542 | * and creations will have been performed and the dentry returned in @path will | |
2543 | * be positive upon return if O_CREAT was specified. If O_CREAT wasn't | |
2544 | * specified then a negative dentry may be returned. | |
2545 | * | |
2546 | * An error code is returned otherwise. | |
2547 | * | |
2548 | * FILE_CREATE will be set in @*opened if the dentry was created and will be | |
2549 | * cleared otherwise prior to returning. | |
d58ffd35 | 2550 | */ |
2675a4eb AV |
2551 | static int lookup_open(struct nameidata *nd, struct path *path, |
2552 | struct file *file, | |
2553 | const struct open_flags *op, | |
64894cf8 | 2554 | bool got_write, int *opened) |
d58ffd35 MS |
2555 | { |
2556 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 2557 | struct inode *dir_inode = dir->d_inode; |
d58ffd35 MS |
2558 | struct dentry *dentry; |
2559 | int error; | |
54ef4872 | 2560 | bool need_lookup; |
d58ffd35 | 2561 | |
47237687 | 2562 | *opened &= ~FILE_CREATED; |
201f956e | 2563 | dentry = lookup_dcache(&nd->last, dir, nd->flags, &need_lookup); |
d58ffd35 | 2564 | if (IS_ERR(dentry)) |
2675a4eb | 2565 | return PTR_ERR(dentry); |
d58ffd35 | 2566 | |
d18e9008 MS |
2567 | /* Cached positive dentry: will open in f_op->open */ |
2568 | if (!need_lookup && dentry->d_inode) | |
2569 | goto out_no_open; | |
2570 | ||
2571 | if ((nd->flags & LOOKUP_OPEN) && dir_inode->i_op->atomic_open) { | |
64894cf8 | 2572 | return atomic_open(nd, dentry, path, file, op, got_write, |
47237687 | 2573 | need_lookup, opened); |
d18e9008 MS |
2574 | } |
2575 | ||
54ef4872 MS |
2576 | if (need_lookup) { |
2577 | BUG_ON(dentry->d_inode); | |
2578 | ||
72bd866a | 2579 | dentry = lookup_real(dir_inode, dentry, nd->flags); |
54ef4872 | 2580 | if (IS_ERR(dentry)) |
2675a4eb | 2581 | return PTR_ERR(dentry); |
54ef4872 MS |
2582 | } |
2583 | ||
d58ffd35 MS |
2584 | /* Negative dentry, just create the file */ |
2585 | if (!dentry->d_inode && (op->open_flag & O_CREAT)) { | |
2586 | umode_t mode = op->mode; | |
2587 | if (!IS_POSIXACL(dir->d_inode)) | |
2588 | mode &= ~current_umask(); | |
2589 | /* | |
2590 | * This write is needed to ensure that a | |
2591 | * rw->ro transition does not occur between | |
2592 | * the time when the file is created and when | |
2593 | * a permanent write count is taken through | |
015c3bbc | 2594 | * the 'struct file' in finish_open(). |
d58ffd35 | 2595 | */ |
64894cf8 AV |
2596 | if (!got_write) { |
2597 | error = -EROFS; | |
d58ffd35 | 2598 | goto out_dput; |
64894cf8 | 2599 | } |
47237687 | 2600 | *opened |= FILE_CREATED; |
d58ffd35 MS |
2601 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2602 | if (error) | |
2603 | goto out_dput; | |
312b63fb AV |
2604 | error = vfs_create(dir->d_inode, dentry, mode, |
2605 | nd->flags & LOOKUP_EXCL); | |
d58ffd35 MS |
2606 | if (error) |
2607 | goto out_dput; | |
2608 | } | |
d18e9008 | 2609 | out_no_open: |
d58ffd35 MS |
2610 | path->dentry = dentry; |
2611 | path->mnt = nd->path.mnt; | |
2675a4eb | 2612 | return 1; |
d58ffd35 MS |
2613 | |
2614 | out_dput: | |
2615 | dput(dentry); | |
2675a4eb | 2616 | return error; |
d58ffd35 MS |
2617 | } |
2618 | ||
31e6b01f | 2619 | /* |
fe2d35ff | 2620 | * Handle the last step of open() |
31e6b01f | 2621 | */ |
2675a4eb AV |
2622 | static int do_last(struct nameidata *nd, struct path *path, |
2623 | struct file *file, const struct open_flags *op, | |
2624 | int *opened, const char *pathname) | |
fb1cc555 | 2625 | { |
a1e28038 | 2626 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 2627 | int open_flag = op->open_flag; |
77d660a8 | 2628 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 2629 | bool got_write = false; |
bcda7652 | 2630 | int acc_mode = op->acc_mode; |
a1eb3315 | 2631 | struct inode *inode; |
77d660a8 | 2632 | bool symlink_ok = false; |
16b1c1cd MS |
2633 | struct path save_parent = { .dentry = NULL, .mnt = NULL }; |
2634 | bool retried = false; | |
16c2cd71 | 2635 | int error; |
1f36f774 | 2636 | |
c3e380b0 AV |
2637 | nd->flags &= ~LOOKUP_PARENT; |
2638 | nd->flags |= op->intent; | |
2639 | ||
1f36f774 AV |
2640 | switch (nd->last_type) { |
2641 | case LAST_DOTDOT: | |
176306f5 | 2642 | case LAST_DOT: |
fe2d35ff AV |
2643 | error = handle_dots(nd, nd->last_type); |
2644 | if (error) | |
2675a4eb | 2645 | return error; |
1f36f774 | 2646 | /* fallthrough */ |
1f36f774 | 2647 | case LAST_ROOT: |
9f1fafee | 2648 | error = complete_walk(nd); |
16c2cd71 | 2649 | if (error) |
2675a4eb | 2650 | return error; |
fe2d35ff | 2651 | audit_inode(pathname, nd->path.dentry); |
ca344a89 | 2652 | if (open_flag & O_CREAT) { |
fe2d35ff | 2653 | error = -EISDIR; |
2675a4eb | 2654 | goto out; |
fe2d35ff | 2655 | } |
e83db167 | 2656 | goto finish_open; |
1f36f774 | 2657 | case LAST_BIND: |
9f1fafee | 2658 | error = complete_walk(nd); |
16c2cd71 | 2659 | if (error) |
2675a4eb | 2660 | return error; |
1f36f774 | 2661 | audit_inode(pathname, dir); |
e83db167 | 2662 | goto finish_open; |
1f36f774 | 2663 | } |
67ee3ad2 | 2664 | |
ca344a89 | 2665 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
2666 | if (nd->last.name[nd->last.len]) |
2667 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
bcda7652 | 2668 | if (open_flag & O_PATH && !(nd->flags & LOOKUP_FOLLOW)) |
77d660a8 | 2669 | symlink_ok = true; |
fe2d35ff | 2670 | /* we _can_ be in RCU mode here */ |
a1eb3315 | 2671 | error = lookup_fast(nd, &nd->last, path, &inode); |
71574865 MS |
2672 | if (likely(!error)) |
2673 | goto finish_lookup; | |
2674 | ||
2675 | if (error < 0) | |
2675a4eb | 2676 | goto out; |
71574865 MS |
2677 | |
2678 | BUG_ON(nd->inode != dir->d_inode); | |
b6183df7 MS |
2679 | } else { |
2680 | /* create side of things */ | |
2681 | /* | |
2682 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
2683 | * has been cleared when we got to the last component we are | |
2684 | * about to look up | |
2685 | */ | |
2686 | error = complete_walk(nd); | |
2687 | if (error) | |
2675a4eb | 2688 | return error; |
fe2d35ff | 2689 | |
b6183df7 MS |
2690 | audit_inode(pathname, dir); |
2691 | error = -EISDIR; | |
2692 | /* trailing slashes? */ | |
2693 | if (nd->last.name[nd->last.len]) | |
2675a4eb | 2694 | goto out; |
b6183df7 | 2695 | } |
a2c36b45 | 2696 | |
16b1c1cd | 2697 | retry_lookup: |
64894cf8 AV |
2698 | if (op->open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
2699 | error = mnt_want_write(nd->path.mnt); | |
2700 | if (!error) | |
2701 | got_write = true; | |
2702 | /* | |
2703 | * do _not_ fail yet - we might not need that or fail with | |
2704 | * a different error; let lookup_open() decide; we'll be | |
2705 | * dropping this one anyway. | |
2706 | */ | |
2707 | } | |
a1e28038 | 2708 | mutex_lock(&dir->d_inode->i_mutex); |
64894cf8 | 2709 | error = lookup_open(nd, path, file, op, got_write, opened); |
d58ffd35 | 2710 | mutex_unlock(&dir->d_inode->i_mutex); |
a1e28038 | 2711 | |
2675a4eb AV |
2712 | if (error <= 0) { |
2713 | if (error) | |
d18e9008 MS |
2714 | goto out; |
2715 | ||
47237687 | 2716 | if ((*opened & FILE_CREATED) || |
2675a4eb | 2717 | !S_ISREG(file->f_path.dentry->d_inode->i_mode)) |
77d660a8 | 2718 | will_truncate = false; |
d18e9008 | 2719 | |
2675a4eb | 2720 | audit_inode(pathname, file->f_path.dentry); |
d18e9008 MS |
2721 | goto opened; |
2722 | } | |
fb1cc555 | 2723 | |
47237687 | 2724 | if (*opened & FILE_CREATED) { |
9b44f1b3 | 2725 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 2726 | open_flag &= ~O_TRUNC; |
77d660a8 | 2727 | will_truncate = false; |
bcda7652 | 2728 | acc_mode = MAY_OPEN; |
d58ffd35 | 2729 | path_to_nameidata(path, nd); |
e83db167 | 2730 | goto finish_open_created; |
fb1cc555 AV |
2731 | } |
2732 | ||
2733 | /* | |
3134f37e | 2734 | * create/update audit record if it already exists. |
fb1cc555 | 2735 | */ |
3134f37e JL |
2736 | if (path->dentry->d_inode) |
2737 | audit_inode(pathname, path->dentry); | |
fb1cc555 | 2738 | |
d18e9008 MS |
2739 | /* |
2740 | * If atomic_open() acquired write access it is dropped now due to | |
2741 | * possible mount and symlink following (this might be optimized away if | |
2742 | * necessary...) | |
2743 | */ | |
64894cf8 | 2744 | if (got_write) { |
d18e9008 | 2745 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 2746 | got_write = false; |
d18e9008 MS |
2747 | } |
2748 | ||
fb1cc555 | 2749 | error = -EEXIST; |
f8310c59 | 2750 | if ((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT)) |
fb1cc555 AV |
2751 | goto exit_dput; |
2752 | ||
9875cf80 DH |
2753 | error = follow_managed(path, nd->flags); |
2754 | if (error < 0) | |
2755 | goto exit_dput; | |
fb1cc555 | 2756 | |
a3fbbde7 AV |
2757 | if (error) |
2758 | nd->flags |= LOOKUP_JUMPED; | |
2759 | ||
decf3400 MS |
2760 | BUG_ON(nd->flags & LOOKUP_RCU); |
2761 | inode = path->dentry->d_inode; | |
5f5daac1 MS |
2762 | finish_lookup: |
2763 | /* we _can_ be in RCU mode here */ | |
fb1cc555 | 2764 | error = -ENOENT; |
54c33e7f MS |
2765 | if (!inode) { |
2766 | path_to_nameidata(path, nd); | |
2675a4eb | 2767 | goto out; |
54c33e7f | 2768 | } |
9e67f361 | 2769 | |
d45ea867 MS |
2770 | if (should_follow_link(inode, !symlink_ok)) { |
2771 | if (nd->flags & LOOKUP_RCU) { | |
2772 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
2773 | error = -ECHILD; | |
2675a4eb | 2774 | goto out; |
d45ea867 MS |
2775 | } |
2776 | } | |
2777 | BUG_ON(inode != path->dentry->d_inode); | |
2675a4eb | 2778 | return 1; |
d45ea867 | 2779 | } |
fb1cc555 | 2780 | |
16b1c1cd MS |
2781 | if ((nd->flags & LOOKUP_RCU) || nd->path.mnt != path->mnt) { |
2782 | path_to_nameidata(path, nd); | |
2783 | } else { | |
2784 | save_parent.dentry = nd->path.dentry; | |
2785 | save_parent.mnt = mntget(path->mnt); | |
2786 | nd->path.dentry = path->dentry; | |
2787 | ||
2788 | } | |
decf3400 | 2789 | nd->inode = inode; |
a3fbbde7 AV |
2790 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
2791 | error = complete_walk(nd); | |
16b1c1cd MS |
2792 | if (error) { |
2793 | path_put(&save_parent); | |
2675a4eb | 2794 | return error; |
16b1c1cd | 2795 | } |
fb1cc555 | 2796 | error = -EISDIR; |
050ac841 | 2797 | if ((open_flag & O_CREAT) && S_ISDIR(nd->inode->i_mode)) |
2675a4eb | 2798 | goto out; |
af2f5542 MS |
2799 | error = -ENOTDIR; |
2800 | if ((nd->flags & LOOKUP_DIRECTORY) && !nd->inode->i_op->lookup) | |
2675a4eb | 2801 | goto out; |
d7fdd7f6 | 2802 | audit_inode(pathname, nd->path.dentry); |
e83db167 | 2803 | finish_open: |
6c0d46c4 | 2804 | if (!S_ISREG(nd->inode->i_mode)) |
77d660a8 | 2805 | will_truncate = false; |
6c0d46c4 | 2806 | |
0f9d1a10 AV |
2807 | if (will_truncate) { |
2808 | error = mnt_want_write(nd->path.mnt); | |
2809 | if (error) | |
2675a4eb | 2810 | goto out; |
64894cf8 | 2811 | got_write = true; |
0f9d1a10 | 2812 | } |
e83db167 | 2813 | finish_open_created: |
bcda7652 | 2814 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 2815 | if (error) |
2675a4eb | 2816 | goto out; |
30d90494 AV |
2817 | file->f_path.mnt = nd->path.mnt; |
2818 | error = finish_open(file, nd->path.dentry, NULL, opened); | |
2819 | if (error) { | |
30d90494 | 2820 | if (error == -EOPENSTALE) |
f60dc3db | 2821 | goto stale_open; |
015c3bbc | 2822 | goto out; |
f60dc3db | 2823 | } |
a8277b9b | 2824 | opened: |
2675a4eb | 2825 | error = open_check_o_direct(file); |
015c3bbc MS |
2826 | if (error) |
2827 | goto exit_fput; | |
2675a4eb | 2828 | error = ima_file_check(file, op->acc_mode); |
aa4caadb MS |
2829 | if (error) |
2830 | goto exit_fput; | |
2831 | ||
2832 | if (will_truncate) { | |
2675a4eb | 2833 | error = handle_truncate(file); |
aa4caadb MS |
2834 | if (error) |
2835 | goto exit_fput; | |
0f9d1a10 | 2836 | } |
ca344a89 | 2837 | out: |
64894cf8 | 2838 | if (got_write) |
0f9d1a10 | 2839 | mnt_drop_write(nd->path.mnt); |
16b1c1cd | 2840 | path_put(&save_parent); |
e276ae67 | 2841 | terminate_walk(nd); |
2675a4eb | 2842 | return error; |
fb1cc555 | 2843 | |
fb1cc555 AV |
2844 | exit_dput: |
2845 | path_put_conditional(path, nd); | |
ca344a89 | 2846 | goto out; |
015c3bbc | 2847 | exit_fput: |
2675a4eb AV |
2848 | fput(file); |
2849 | goto out; | |
015c3bbc | 2850 | |
f60dc3db MS |
2851 | stale_open: |
2852 | /* If no saved parent or already retried then can't retry */ | |
2853 | if (!save_parent.dentry || retried) | |
2854 | goto out; | |
2855 | ||
2856 | BUG_ON(save_parent.dentry != dir); | |
2857 | path_put(&nd->path); | |
2858 | nd->path = save_parent; | |
2859 | nd->inode = dir->d_inode; | |
2860 | save_parent.mnt = NULL; | |
2861 | save_parent.dentry = NULL; | |
64894cf8 | 2862 | if (got_write) { |
f60dc3db | 2863 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 2864 | got_write = false; |
f60dc3db MS |
2865 | } |
2866 | retried = true; | |
2867 | goto retry_lookup; | |
fb1cc555 AV |
2868 | } |
2869 | ||
13aab428 | 2870 | static struct file *path_openat(int dfd, const char *pathname, |
73d049a4 | 2871 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 2872 | { |
fe2d35ff | 2873 | struct file *base = NULL; |
30d90494 | 2874 | struct file *file; |
9850c056 | 2875 | struct path path; |
47237687 | 2876 | int opened = 0; |
13aab428 | 2877 | int error; |
31e6b01f | 2878 | |
30d90494 AV |
2879 | file = get_empty_filp(); |
2880 | if (!file) | |
31e6b01f NP |
2881 | return ERR_PTR(-ENFILE); |
2882 | ||
30d90494 | 2883 | file->f_flags = op->open_flag; |
31e6b01f | 2884 | |
73d049a4 | 2885 | error = path_init(dfd, pathname, flags | LOOKUP_PARENT, nd, &base); |
31e6b01f | 2886 | if (unlikely(error)) |
2675a4eb | 2887 | goto out; |
31e6b01f | 2888 | |
fe2d35ff | 2889 | current->total_link_count = 0; |
73d049a4 | 2890 | error = link_path_walk(pathname, nd); |
31e6b01f | 2891 | if (unlikely(error)) |
2675a4eb | 2892 | goto out; |
1da177e4 | 2893 | |
2675a4eb AV |
2894 | error = do_last(nd, &path, file, op, &opened, pathname); |
2895 | while (unlikely(error > 0)) { /* trailing symlink */ | |
7b9337aa | 2896 | struct path link = path; |
def4af30 | 2897 | void *cookie; |
574197e0 | 2898 | if (!(nd->flags & LOOKUP_FOLLOW)) { |
73d049a4 AV |
2899 | path_put_conditional(&path, nd); |
2900 | path_put(&nd->path); | |
2675a4eb | 2901 | error = -ELOOP; |
40b39136 AV |
2902 | break; |
2903 | } | |
800179c9 KC |
2904 | error = may_follow_link(&link, nd); |
2905 | if (unlikely(error)) | |
2906 | break; | |
73d049a4 AV |
2907 | nd->flags |= LOOKUP_PARENT; |
2908 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
574197e0 | 2909 | error = follow_link(&link, nd, &cookie); |
c3e380b0 | 2910 | if (unlikely(error)) |
2675a4eb AV |
2911 | break; |
2912 | error = do_last(nd, &path, file, op, &opened, pathname); | |
574197e0 | 2913 | put_link(nd, &link, cookie); |
806b681c | 2914 | } |
10fa8e62 | 2915 | out: |
73d049a4 AV |
2916 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) |
2917 | path_put(&nd->root); | |
fe2d35ff AV |
2918 | if (base) |
2919 | fput(base); | |
2675a4eb AV |
2920 | if (!(opened & FILE_OPENED)) { |
2921 | BUG_ON(!error); | |
30d90494 | 2922 | put_filp(file); |
16b1c1cd | 2923 | } |
2675a4eb AV |
2924 | if (unlikely(error)) { |
2925 | if (error == -EOPENSTALE) { | |
2926 | if (flags & LOOKUP_RCU) | |
2927 | error = -ECHILD; | |
2928 | else | |
2929 | error = -ESTALE; | |
2930 | } | |
2931 | file = ERR_PTR(error); | |
2932 | } | |
2933 | return file; | |
1da177e4 LT |
2934 | } |
2935 | ||
13aab428 AV |
2936 | struct file *do_filp_open(int dfd, const char *pathname, |
2937 | const struct open_flags *op, int flags) | |
2938 | { | |
73d049a4 | 2939 | struct nameidata nd; |
13aab428 AV |
2940 | struct file *filp; |
2941 | ||
73d049a4 | 2942 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 2943 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 2944 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 2945 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 2946 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
2947 | return filp; |
2948 | } | |
2949 | ||
73d049a4 AV |
2950 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
2951 | const char *name, const struct open_flags *op, int flags) | |
2952 | { | |
2953 | struct nameidata nd; | |
2954 | struct file *file; | |
2955 | ||
2956 | nd.root.mnt = mnt; | |
2957 | nd.root.dentry = dentry; | |
2958 | ||
2959 | flags |= LOOKUP_ROOT; | |
2960 | ||
bcda7652 | 2961 | if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
2962 | return ERR_PTR(-ELOOP); |
2963 | ||
2964 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU); | |
2965 | if (unlikely(file == ERR_PTR(-ECHILD))) | |
2966 | file = path_openat(-1, name, &nd, op, flags); | |
2967 | if (unlikely(file == ERR_PTR(-ESTALE))) | |
2968 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL); | |
2969 | return file; | |
2970 | } | |
2971 | ||
ed75e95d | 2972 | struct dentry *kern_path_create(int dfd, const char *pathname, struct path *path, int is_dir) |
1da177e4 | 2973 | { |
c663e5d8 | 2974 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d | 2975 | struct nameidata nd; |
c30dabfe | 2976 | int err2; |
ed75e95d AV |
2977 | int error = do_path_lookup(dfd, pathname, LOOKUP_PARENT, &nd); |
2978 | if (error) | |
2979 | return ERR_PTR(error); | |
1da177e4 | 2980 | |
c663e5d8 CH |
2981 | /* |
2982 | * Yucky last component or no last component at all? | |
2983 | * (foo/., foo/.., /////) | |
2984 | */ | |
ed75e95d AV |
2985 | if (nd.last_type != LAST_NORM) |
2986 | goto out; | |
2987 | nd.flags &= ~LOOKUP_PARENT; | |
2988 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
c663e5d8 | 2989 | |
c30dabfe JK |
2990 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
2991 | err2 = mnt_want_write(nd.path.mnt); | |
c663e5d8 CH |
2992 | /* |
2993 | * Do the final lookup. | |
2994 | */ | |
ed75e95d AV |
2995 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
2996 | dentry = lookup_hash(&nd); | |
1da177e4 | 2997 | if (IS_ERR(dentry)) |
a8104a9f | 2998 | goto unlock; |
c663e5d8 | 2999 | |
a8104a9f | 3000 | error = -EEXIST; |
e9baf6e5 | 3001 | if (dentry->d_inode) |
a8104a9f | 3002 | goto fail; |
c663e5d8 CH |
3003 | /* |
3004 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3005 | * From the vfs_mknod() POV we just have a negative dentry - | |
3006 | * all is fine. Let's be bastards - you had / on the end, you've | |
3007 | * been asking for (non-existent) directory. -ENOENT for you. | |
3008 | */ | |
ed75e95d | 3009 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
a8104a9f | 3010 | error = -ENOENT; |
ed75e95d | 3011 | goto fail; |
e9baf6e5 | 3012 | } |
c30dabfe JK |
3013 | if (unlikely(err2)) { |
3014 | error = err2; | |
a8104a9f | 3015 | goto fail; |
c30dabfe | 3016 | } |
ed75e95d | 3017 | *path = nd.path; |
1da177e4 | 3018 | return dentry; |
1da177e4 | 3019 | fail: |
a8104a9f AV |
3020 | dput(dentry); |
3021 | dentry = ERR_PTR(error); | |
3022 | unlock: | |
ed75e95d | 3023 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
c30dabfe JK |
3024 | if (!err2) |
3025 | mnt_drop_write(nd.path.mnt); | |
ed75e95d AV |
3026 | out: |
3027 | path_put(&nd.path); | |
1da177e4 LT |
3028 | return dentry; |
3029 | } | |
dae6ad8f AV |
3030 | EXPORT_SYMBOL(kern_path_create); |
3031 | ||
921a1650 AV |
3032 | void done_path_create(struct path *path, struct dentry *dentry) |
3033 | { | |
3034 | dput(dentry); | |
3035 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
a8104a9f | 3036 | mnt_drop_write(path->mnt); |
921a1650 AV |
3037 | path_put(path); |
3038 | } | |
3039 | EXPORT_SYMBOL(done_path_create); | |
3040 | ||
dae6ad8f AV |
3041 | struct dentry *user_path_create(int dfd, const char __user *pathname, struct path *path, int is_dir) |
3042 | { | |
3043 | char *tmp = getname(pathname); | |
3044 | struct dentry *res; | |
3045 | if (IS_ERR(tmp)) | |
3046 | return ERR_CAST(tmp); | |
3047 | res = kern_path_create(dfd, tmp, path, is_dir); | |
3048 | putname(tmp); | |
3049 | return res; | |
3050 | } | |
3051 | EXPORT_SYMBOL(user_path_create); | |
3052 | ||
1a67aafb | 3053 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3054 | { |
a95164d9 | 3055 | int error = may_create(dir, dentry); |
1da177e4 LT |
3056 | |
3057 | if (error) | |
3058 | return error; | |
3059 | ||
975d6b39 | 3060 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3061 | return -EPERM; |
3062 | ||
acfa4380 | 3063 | if (!dir->i_op->mknod) |
1da177e4 LT |
3064 | return -EPERM; |
3065 | ||
08ce5f16 SH |
3066 | error = devcgroup_inode_mknod(mode, dev); |
3067 | if (error) | |
3068 | return error; | |
3069 | ||
1da177e4 LT |
3070 | error = security_inode_mknod(dir, dentry, mode, dev); |
3071 | if (error) | |
3072 | return error; | |
3073 | ||
1da177e4 | 3074 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3075 | if (!error) |
f38aa942 | 3076 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3077 | return error; |
3078 | } | |
3079 | ||
f69aac00 | 3080 | static int may_mknod(umode_t mode) |
463c3197 DH |
3081 | { |
3082 | switch (mode & S_IFMT) { | |
3083 | case S_IFREG: | |
3084 | case S_IFCHR: | |
3085 | case S_IFBLK: | |
3086 | case S_IFIFO: | |
3087 | case S_IFSOCK: | |
3088 | case 0: /* zero mode translates to S_IFREG */ | |
3089 | return 0; | |
3090 | case S_IFDIR: | |
3091 | return -EPERM; | |
3092 | default: | |
3093 | return -EINVAL; | |
3094 | } | |
3095 | } | |
3096 | ||
8208a22b | 3097 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 3098 | unsigned, dev) |
1da177e4 | 3099 | { |
2ad94ae6 | 3100 | struct dentry *dentry; |
dae6ad8f AV |
3101 | struct path path; |
3102 | int error; | |
1da177e4 | 3103 | |
8e4bfca1 AV |
3104 | error = may_mknod(mode); |
3105 | if (error) | |
3106 | return error; | |
1da177e4 | 3107 | |
dae6ad8f AV |
3108 | dentry = user_path_create(dfd, filename, &path, 0); |
3109 | if (IS_ERR(dentry)) | |
3110 | return PTR_ERR(dentry); | |
2ad94ae6 | 3111 | |
dae6ad8f | 3112 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3113 | mode &= ~current_umask(); |
dae6ad8f | 3114 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3115 | if (error) |
a8104a9f | 3116 | goto out; |
463c3197 | 3117 | switch (mode & S_IFMT) { |
1da177e4 | 3118 | case 0: case S_IFREG: |
312b63fb | 3119 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
1da177e4 LT |
3120 | break; |
3121 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3122 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3123 | new_decode_dev(dev)); |
3124 | break; | |
3125 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3126 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3127 | break; |
1da177e4 | 3128 | } |
a8104a9f | 3129 | out: |
921a1650 | 3130 | done_path_create(&path, dentry); |
1da177e4 LT |
3131 | return error; |
3132 | } | |
3133 | ||
8208a22b | 3134 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
3135 | { |
3136 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
3137 | } | |
3138 | ||
18bb1db3 | 3139 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3140 | { |
a95164d9 | 3141 | int error = may_create(dir, dentry); |
8de52778 | 3142 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3143 | |
3144 | if (error) | |
3145 | return error; | |
3146 | ||
acfa4380 | 3147 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3148 | return -EPERM; |
3149 | ||
3150 | mode &= (S_IRWXUGO|S_ISVTX); | |
3151 | error = security_inode_mkdir(dir, dentry, mode); | |
3152 | if (error) | |
3153 | return error; | |
3154 | ||
8de52778 AV |
3155 | if (max_links && dir->i_nlink >= max_links) |
3156 | return -EMLINK; | |
3157 | ||
1da177e4 | 3158 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3159 | if (!error) |
f38aa942 | 3160 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3161 | return error; |
3162 | } | |
3163 | ||
a218d0fd | 3164 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 3165 | { |
6902d925 | 3166 | struct dentry *dentry; |
dae6ad8f AV |
3167 | struct path path; |
3168 | int error; | |
1da177e4 | 3169 | |
dae6ad8f | 3170 | dentry = user_path_create(dfd, pathname, &path, 1); |
6902d925 | 3171 | if (IS_ERR(dentry)) |
dae6ad8f | 3172 | return PTR_ERR(dentry); |
1da177e4 | 3173 | |
dae6ad8f | 3174 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3175 | mode &= ~current_umask(); |
dae6ad8f | 3176 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3177 | if (!error) |
3178 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3179 | done_path_create(&path, dentry); |
1da177e4 LT |
3180 | return error; |
3181 | } | |
3182 | ||
a218d0fd | 3183 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
3184 | { |
3185 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
3186 | } | |
3187 | ||
1da177e4 | 3188 | /* |
a71905f0 | 3189 | * The dentry_unhash() helper will try to drop the dentry early: we |
c0d02594 | 3190 | * should have a usage count of 1 if we're the only user of this |
a71905f0 SW |
3191 | * dentry, and if that is true (possibly after pruning the dcache), |
3192 | * then we drop the dentry now. | |
1da177e4 LT |
3193 | * |
3194 | * A low-level filesystem can, if it choses, legally | |
3195 | * do a | |
3196 | * | |
3197 | * if (!d_unhashed(dentry)) | |
3198 | * return -EBUSY; | |
3199 | * | |
3200 | * if it cannot handle the case of removing a directory | |
3201 | * that is still in use by something else.. | |
3202 | */ | |
3203 | void dentry_unhash(struct dentry *dentry) | |
3204 | { | |
dc168427 | 3205 | shrink_dcache_parent(dentry); |
1da177e4 | 3206 | spin_lock(&dentry->d_lock); |
64252c75 | 3207 | if (dentry->d_count == 1) |
1da177e4 LT |
3208 | __d_drop(dentry); |
3209 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
3210 | } |
3211 | ||
3212 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
3213 | { | |
3214 | int error = may_delete(dir, dentry, 1); | |
3215 | ||
3216 | if (error) | |
3217 | return error; | |
3218 | ||
acfa4380 | 3219 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3220 | return -EPERM; |
3221 | ||
1d2ef590 | 3222 | dget(dentry); |
1b1dcc1b | 3223 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
3224 | |
3225 | error = -EBUSY; | |
1da177e4 | 3226 | if (d_mountpoint(dentry)) |
912dbc15 SW |
3227 | goto out; |
3228 | ||
3229 | error = security_inode_rmdir(dir, dentry); | |
3230 | if (error) | |
3231 | goto out; | |
3232 | ||
3cebde24 | 3233 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3234 | error = dir->i_op->rmdir(dir, dentry); |
3235 | if (error) | |
3236 | goto out; | |
3237 | ||
3238 | dentry->d_inode->i_flags |= S_DEAD; | |
3239 | dont_mount(dentry); | |
3240 | ||
3241 | out: | |
1b1dcc1b | 3242 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 3243 | dput(dentry); |
912dbc15 | 3244 | if (!error) |
1da177e4 | 3245 | d_delete(dentry); |
1da177e4 LT |
3246 | return error; |
3247 | } | |
3248 | ||
5590ff0d | 3249 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3250 | { |
3251 | int error = 0; | |
3252 | char * name; | |
3253 | struct dentry *dentry; | |
3254 | struct nameidata nd; | |
3255 | ||
2ad94ae6 | 3256 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 3257 | if (error) |
2ad94ae6 | 3258 | return error; |
1da177e4 LT |
3259 | |
3260 | switch(nd.last_type) { | |
0612d9fb OH |
3261 | case LAST_DOTDOT: |
3262 | error = -ENOTEMPTY; | |
3263 | goto exit1; | |
3264 | case LAST_DOT: | |
3265 | error = -EINVAL; | |
3266 | goto exit1; | |
3267 | case LAST_ROOT: | |
3268 | error = -EBUSY; | |
3269 | goto exit1; | |
1da177e4 | 3270 | } |
0612d9fb OH |
3271 | |
3272 | nd.flags &= ~LOOKUP_PARENT; | |
c30dabfe JK |
3273 | error = mnt_want_write(nd.path.mnt); |
3274 | if (error) | |
3275 | goto exit1; | |
0612d9fb | 3276 | |
4ac91378 | 3277 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 3278 | dentry = lookup_hash(&nd); |
1da177e4 | 3279 | error = PTR_ERR(dentry); |
6902d925 DH |
3280 | if (IS_ERR(dentry)) |
3281 | goto exit2; | |
e6bc45d6 TT |
3282 | if (!dentry->d_inode) { |
3283 | error = -ENOENT; | |
3284 | goto exit3; | |
3285 | } | |
be6d3e56 KT |
3286 | error = security_path_rmdir(&nd.path, dentry); |
3287 | if (error) | |
c30dabfe | 3288 | goto exit3; |
4ac91378 | 3289 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
0622753b | 3290 | exit3: |
6902d925 DH |
3291 | dput(dentry); |
3292 | exit2: | |
4ac91378 | 3293 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
c30dabfe | 3294 | mnt_drop_write(nd.path.mnt); |
1da177e4 | 3295 | exit1: |
1d957f9b | 3296 | path_put(&nd.path); |
1da177e4 LT |
3297 | putname(name); |
3298 | return error; | |
3299 | } | |
3300 | ||
3cdad428 | 3301 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3302 | { |
3303 | return do_rmdir(AT_FDCWD, pathname); | |
3304 | } | |
3305 | ||
1da177e4 LT |
3306 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
3307 | { | |
3308 | int error = may_delete(dir, dentry, 0); | |
3309 | ||
3310 | if (error) | |
3311 | return error; | |
3312 | ||
acfa4380 | 3313 | if (!dir->i_op->unlink) |
1da177e4 LT |
3314 | return -EPERM; |
3315 | ||
1b1dcc1b | 3316 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
3317 | if (d_mountpoint(dentry)) |
3318 | error = -EBUSY; | |
3319 | else { | |
3320 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3321 | if (!error) { |
1da177e4 | 3322 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 3323 | if (!error) |
d83c49f3 | 3324 | dont_mount(dentry); |
bec1052e | 3325 | } |
1da177e4 | 3326 | } |
1b1dcc1b | 3327 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
3328 | |
3329 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3330 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 3331 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 3332 | d_delete(dentry); |
1da177e4 | 3333 | } |
0eeca283 | 3334 | |
1da177e4 LT |
3335 | return error; |
3336 | } | |
3337 | ||
3338 | /* | |
3339 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3340 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3341 | * writeout happening, and we don't want to prevent access to the directory |
3342 | * while waiting on the I/O. | |
3343 | */ | |
5590ff0d | 3344 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 3345 | { |
2ad94ae6 AV |
3346 | int error; |
3347 | char *name; | |
1da177e4 LT |
3348 | struct dentry *dentry; |
3349 | struct nameidata nd; | |
3350 | struct inode *inode = NULL; | |
3351 | ||
2ad94ae6 | 3352 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 3353 | if (error) |
2ad94ae6 AV |
3354 | return error; |
3355 | ||
1da177e4 LT |
3356 | error = -EISDIR; |
3357 | if (nd.last_type != LAST_NORM) | |
3358 | goto exit1; | |
0612d9fb OH |
3359 | |
3360 | nd.flags &= ~LOOKUP_PARENT; | |
c30dabfe JK |
3361 | error = mnt_want_write(nd.path.mnt); |
3362 | if (error) | |
3363 | goto exit1; | |
0612d9fb | 3364 | |
4ac91378 | 3365 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 3366 | dentry = lookup_hash(&nd); |
1da177e4 LT |
3367 | error = PTR_ERR(dentry); |
3368 | if (!IS_ERR(dentry)) { | |
3369 | /* Why not before? Because we want correct error value */ | |
50338b88 TE |
3370 | if (nd.last.name[nd.last.len]) |
3371 | goto slashes; | |
1da177e4 | 3372 | inode = dentry->d_inode; |
50338b88 | 3373 | if (!inode) |
e6bc45d6 TT |
3374 | goto slashes; |
3375 | ihold(inode); | |
be6d3e56 KT |
3376 | error = security_path_unlink(&nd.path, dentry); |
3377 | if (error) | |
c30dabfe | 3378 | goto exit2; |
4ac91378 | 3379 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
c30dabfe | 3380 | exit2: |
1da177e4 LT |
3381 | dput(dentry); |
3382 | } | |
4ac91378 | 3383 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
3384 | if (inode) |
3385 | iput(inode); /* truncate the inode here */ | |
c30dabfe | 3386 | mnt_drop_write(nd.path.mnt); |
1da177e4 | 3387 | exit1: |
1d957f9b | 3388 | path_put(&nd.path); |
1da177e4 LT |
3389 | putname(name); |
3390 | return error; | |
3391 | ||
3392 | slashes: | |
3393 | error = !dentry->d_inode ? -ENOENT : | |
3394 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
3395 | goto exit2; | |
3396 | } | |
3397 | ||
2e4d0924 | 3398 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3399 | { |
3400 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3401 | return -EINVAL; | |
3402 | ||
3403 | if (flag & AT_REMOVEDIR) | |
3404 | return do_rmdir(dfd, pathname); | |
3405 | ||
3406 | return do_unlinkat(dfd, pathname); | |
3407 | } | |
3408 | ||
3480b257 | 3409 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
3410 | { |
3411 | return do_unlinkat(AT_FDCWD, pathname); | |
3412 | } | |
3413 | ||
db2e747b | 3414 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3415 | { |
a95164d9 | 3416 | int error = may_create(dir, dentry); |
1da177e4 LT |
3417 | |
3418 | if (error) | |
3419 | return error; | |
3420 | ||
acfa4380 | 3421 | if (!dir->i_op->symlink) |
1da177e4 LT |
3422 | return -EPERM; |
3423 | ||
3424 | error = security_inode_symlink(dir, dentry, oldname); | |
3425 | if (error) | |
3426 | return error; | |
3427 | ||
1da177e4 | 3428 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3429 | if (!error) |
f38aa942 | 3430 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3431 | return error; |
3432 | } | |
3433 | ||
2e4d0924 HC |
3434 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
3435 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3436 | { |
2ad94ae6 AV |
3437 | int error; |
3438 | char *from; | |
6902d925 | 3439 | struct dentry *dentry; |
dae6ad8f | 3440 | struct path path; |
1da177e4 LT |
3441 | |
3442 | from = getname(oldname); | |
2ad94ae6 | 3443 | if (IS_ERR(from)) |
1da177e4 | 3444 | return PTR_ERR(from); |
1da177e4 | 3445 | |
dae6ad8f | 3446 | dentry = user_path_create(newdfd, newname, &path, 0); |
6902d925 DH |
3447 | error = PTR_ERR(dentry); |
3448 | if (IS_ERR(dentry)) | |
dae6ad8f | 3449 | goto out_putname; |
6902d925 | 3450 | |
dae6ad8f | 3451 | error = security_path_symlink(&path, dentry, from); |
a8104a9f AV |
3452 | if (!error) |
3453 | error = vfs_symlink(path.dentry->d_inode, dentry, from); | |
921a1650 | 3454 | done_path_create(&path, dentry); |
6902d925 | 3455 | out_putname: |
1da177e4 LT |
3456 | putname(from); |
3457 | return error; | |
3458 | } | |
3459 | ||
3480b257 | 3460 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3461 | { |
3462 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
3463 | } | |
3464 | ||
1da177e4 LT |
3465 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
3466 | { | |
3467 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 3468 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3469 | int error; |
3470 | ||
3471 | if (!inode) | |
3472 | return -ENOENT; | |
3473 | ||
a95164d9 | 3474 | error = may_create(dir, new_dentry); |
1da177e4 LT |
3475 | if (error) |
3476 | return error; | |
3477 | ||
3478 | if (dir->i_sb != inode->i_sb) | |
3479 | return -EXDEV; | |
3480 | ||
3481 | /* | |
3482 | * A link to an append-only or immutable file cannot be created. | |
3483 | */ | |
3484 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
3485 | return -EPERM; | |
acfa4380 | 3486 | if (!dir->i_op->link) |
1da177e4 | 3487 | return -EPERM; |
7e79eedb | 3488 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
3489 | return -EPERM; |
3490 | ||
3491 | error = security_inode_link(old_dentry, dir, new_dentry); | |
3492 | if (error) | |
3493 | return error; | |
3494 | ||
7e79eedb | 3495 | mutex_lock(&inode->i_mutex); |
aae8a97d AK |
3496 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
3497 | if (inode->i_nlink == 0) | |
3498 | error = -ENOENT; | |
8de52778 AV |
3499 | else if (max_links && inode->i_nlink >= max_links) |
3500 | error = -EMLINK; | |
aae8a97d AK |
3501 | else |
3502 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 3503 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 3504 | if (!error) |
7e79eedb | 3505 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
3506 | return error; |
3507 | } | |
3508 | ||
3509 | /* | |
3510 | * Hardlinks are often used in delicate situations. We avoid | |
3511 | * security-related surprises by not following symlinks on the | |
3512 | * newname. --KAB | |
3513 | * | |
3514 | * We don't follow them on the oldname either to be compatible | |
3515 | * with linux 2.0, and to avoid hard-linking to directories | |
3516 | * and other special files. --ADM | |
3517 | */ | |
2e4d0924 HC |
3518 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
3519 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
3520 | { |
3521 | struct dentry *new_dentry; | |
dae6ad8f | 3522 | struct path old_path, new_path; |
11a7b371 | 3523 | int how = 0; |
1da177e4 | 3524 | int error; |
1da177e4 | 3525 | |
11a7b371 | 3526 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 3527 | return -EINVAL; |
11a7b371 AK |
3528 | /* |
3529 | * To use null names we require CAP_DAC_READ_SEARCH | |
3530 | * This ensures that not everyone will be able to create | |
3531 | * handlink using the passed filedescriptor. | |
3532 | */ | |
3533 | if (flags & AT_EMPTY_PATH) { | |
3534 | if (!capable(CAP_DAC_READ_SEARCH)) | |
3535 | return -ENOENT; | |
3536 | how = LOOKUP_EMPTY; | |
3537 | } | |
3538 | ||
3539 | if (flags & AT_SYMLINK_FOLLOW) | |
3540 | how |= LOOKUP_FOLLOW; | |
c04030e1 | 3541 | |
11a7b371 | 3542 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 3543 | if (error) |
2ad94ae6 AV |
3544 | return error; |
3545 | ||
dae6ad8f | 3546 | new_dentry = user_path_create(newdfd, newname, &new_path, 0); |
1da177e4 | 3547 | error = PTR_ERR(new_dentry); |
6902d925 | 3548 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
3549 | goto out; |
3550 | ||
3551 | error = -EXDEV; | |
3552 | if (old_path.mnt != new_path.mnt) | |
3553 | goto out_dput; | |
800179c9 KC |
3554 | error = may_linkat(&old_path); |
3555 | if (unlikely(error)) | |
3556 | goto out_dput; | |
dae6ad8f | 3557 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 3558 | if (error) |
a8104a9f | 3559 | goto out_dput; |
dae6ad8f | 3560 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry); |
75c3f29d | 3561 | out_dput: |
921a1650 | 3562 | done_path_create(&new_path, new_dentry); |
1da177e4 | 3563 | out: |
2d8f3038 | 3564 | path_put(&old_path); |
1da177e4 LT |
3565 | |
3566 | return error; | |
3567 | } | |
3568 | ||
3480b257 | 3569 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 3570 | { |
c04030e1 | 3571 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
3572 | } |
3573 | ||
1da177e4 LT |
3574 | /* |
3575 | * The worst of all namespace operations - renaming directory. "Perverted" | |
3576 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
3577 | * Problems: | |
3578 | * a) we can get into loop creation. Check is done in is_subdir(). | |
3579 | * b) race potential - two innocent renames can create a loop together. | |
3580 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 3581 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
3582 | * story. |
3583 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 3584 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
3585 | * whether the target exists). Solution: try to be smart with locking |
3586 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 3587 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
3588 | * move will be locked. Thus we can rank directories by the tree |
3589 | * (ancestors first) and rank all non-directories after them. | |
3590 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 3591 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
3592 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
3593 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
3594 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 3595 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 3596 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 3597 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 3598 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
3599 | * locking]. |
3600 | */ | |
75c96f85 AB |
3601 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
3602 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3603 | { |
3604 | int error = 0; | |
9055cba7 | 3605 | struct inode *target = new_dentry->d_inode; |
8de52778 | 3606 | unsigned max_links = new_dir->i_sb->s_max_links; |
1da177e4 LT |
3607 | |
3608 | /* | |
3609 | * If we are going to change the parent - check write permissions, | |
3610 | * we'll need to flip '..'. | |
3611 | */ | |
3612 | if (new_dir != old_dir) { | |
f419a2e3 | 3613 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3614 | if (error) |
3615 | return error; | |
3616 | } | |
3617 | ||
3618 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3619 | if (error) | |
3620 | return error; | |
3621 | ||
1d2ef590 | 3622 | dget(new_dentry); |
d83c49f3 | 3623 | if (target) |
1b1dcc1b | 3624 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
3625 | |
3626 | error = -EBUSY; | |
3627 | if (d_mountpoint(old_dentry) || d_mountpoint(new_dentry)) | |
3628 | goto out; | |
3629 | ||
8de52778 AV |
3630 | error = -EMLINK; |
3631 | if (max_links && !target && new_dir != old_dir && | |
3632 | new_dir->i_nlink >= max_links) | |
3633 | goto out; | |
3634 | ||
3cebde24 SW |
3635 | if (target) |
3636 | shrink_dcache_parent(new_dentry); | |
9055cba7 SW |
3637 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
3638 | if (error) | |
3639 | goto out; | |
3640 | ||
1da177e4 | 3641 | if (target) { |
9055cba7 SW |
3642 | target->i_flags |= S_DEAD; |
3643 | dont_mount(new_dentry); | |
1da177e4 | 3644 | } |
9055cba7 SW |
3645 | out: |
3646 | if (target) | |
3647 | mutex_unlock(&target->i_mutex); | |
1d2ef590 | 3648 | dput(new_dentry); |
e31e14ec | 3649 | if (!error) |
349457cc MF |
3650 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3651 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3652 | return error; |
3653 | } | |
3654 | ||
75c96f85 AB |
3655 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3656 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 | 3657 | { |
51892bbb | 3658 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3659 | int error; |
3660 | ||
3661 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3662 | if (error) | |
3663 | return error; | |
3664 | ||
3665 | dget(new_dentry); | |
1da177e4 | 3666 | if (target) |
1b1dcc1b | 3667 | mutex_lock(&target->i_mutex); |
51892bbb SW |
3668 | |
3669 | error = -EBUSY; | |
1da177e4 | 3670 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
51892bbb SW |
3671 | goto out; |
3672 | ||
3673 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3674 | if (error) | |
3675 | goto out; | |
3676 | ||
3677 | if (target) | |
3678 | dont_mount(new_dentry); | |
3679 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) | |
3680 | d_move(old_dentry, new_dentry); | |
3681 | out: | |
1da177e4 | 3682 | if (target) |
1b1dcc1b | 3683 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3684 | dput(new_dentry); |
3685 | return error; | |
3686 | } | |
3687 | ||
3688 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3689 | struct inode *new_dir, struct dentry *new_dentry) | |
3690 | { | |
3691 | int error; | |
3692 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3693 | const unsigned char *old_name; |
1da177e4 LT |
3694 | |
3695 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3696 | return 0; | |
3697 | ||
3698 | error = may_delete(old_dir, old_dentry, is_dir); | |
3699 | if (error) | |
3700 | return error; | |
3701 | ||
3702 | if (!new_dentry->d_inode) | |
a95164d9 | 3703 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3704 | else |
3705 | error = may_delete(new_dir, new_dentry, is_dir); | |
3706 | if (error) | |
3707 | return error; | |
3708 | ||
acfa4380 | 3709 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3710 | return -EPERM; |
3711 | ||
0eeca283 RL |
3712 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3713 | ||
1da177e4 LT |
3714 | if (is_dir) |
3715 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3716 | else | |
3717 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3718 | if (!error) |
3719 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3720 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3721 | fsnotify_oldname_free(old_name); |
3722 | ||
1da177e4 LT |
3723 | return error; |
3724 | } | |
3725 | ||
2e4d0924 HC |
3726 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3727 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3728 | { |
2ad94ae6 AV |
3729 | struct dentry *old_dir, *new_dir; |
3730 | struct dentry *old_dentry, *new_dentry; | |
3731 | struct dentry *trap; | |
1da177e4 | 3732 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3733 | char *from; |
3734 | char *to; | |
3735 | int error; | |
1da177e4 | 3736 | |
2ad94ae6 | 3737 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3738 | if (error) |
3739 | goto exit; | |
3740 | ||
2ad94ae6 | 3741 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3742 | if (error) |
3743 | goto exit1; | |
3744 | ||
3745 | error = -EXDEV; | |
4ac91378 | 3746 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3747 | goto exit2; |
3748 | ||
4ac91378 | 3749 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3750 | error = -EBUSY; |
3751 | if (oldnd.last_type != LAST_NORM) | |
3752 | goto exit2; | |
3753 | ||
4ac91378 | 3754 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3755 | if (newnd.last_type != LAST_NORM) |
3756 | goto exit2; | |
3757 | ||
c30dabfe JK |
3758 | error = mnt_want_write(oldnd.path.mnt); |
3759 | if (error) | |
3760 | goto exit2; | |
3761 | ||
0612d9fb OH |
3762 | oldnd.flags &= ~LOOKUP_PARENT; |
3763 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3764 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3765 | |
1da177e4 LT |
3766 | trap = lock_rename(new_dir, old_dir); |
3767 | ||
49705b77 | 3768 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3769 | error = PTR_ERR(old_dentry); |
3770 | if (IS_ERR(old_dentry)) | |
3771 | goto exit3; | |
3772 | /* source must exist */ | |
3773 | error = -ENOENT; | |
3774 | if (!old_dentry->d_inode) | |
3775 | goto exit4; | |
3776 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3777 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3778 | error = -ENOTDIR; | |
3779 | if (oldnd.last.name[oldnd.last.len]) | |
3780 | goto exit4; | |
3781 | if (newnd.last.name[newnd.last.len]) | |
3782 | goto exit4; | |
3783 | } | |
3784 | /* source should not be ancestor of target */ | |
3785 | error = -EINVAL; | |
3786 | if (old_dentry == trap) | |
3787 | goto exit4; | |
49705b77 | 3788 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3789 | error = PTR_ERR(new_dentry); |
3790 | if (IS_ERR(new_dentry)) | |
3791 | goto exit4; | |
3792 | /* target should not be an ancestor of source */ | |
3793 | error = -ENOTEMPTY; | |
3794 | if (new_dentry == trap) | |
3795 | goto exit5; | |
3796 | ||
be6d3e56 KT |
3797 | error = security_path_rename(&oldnd.path, old_dentry, |
3798 | &newnd.path, new_dentry); | |
3799 | if (error) | |
c30dabfe | 3800 | goto exit5; |
1da177e4 LT |
3801 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3802 | new_dir->d_inode, new_dentry); | |
3803 | exit5: | |
3804 | dput(new_dentry); | |
3805 | exit4: | |
3806 | dput(old_dentry); | |
3807 | exit3: | |
3808 | unlock_rename(new_dir, old_dir); | |
c30dabfe | 3809 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 | 3810 | exit2: |
1d957f9b | 3811 | path_put(&newnd.path); |
2ad94ae6 | 3812 | putname(to); |
1da177e4 | 3813 | exit1: |
1d957f9b | 3814 | path_put(&oldnd.path); |
1da177e4 | 3815 | putname(from); |
2ad94ae6 | 3816 | exit: |
1da177e4 LT |
3817 | return error; |
3818 | } | |
3819 | ||
a26eab24 | 3820 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3821 | { |
3822 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3823 | } | |
3824 | ||
1da177e4 LT |
3825 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3826 | { | |
3827 | int len; | |
3828 | ||
3829 | len = PTR_ERR(link); | |
3830 | if (IS_ERR(link)) | |
3831 | goto out; | |
3832 | ||
3833 | len = strlen(link); | |
3834 | if (len > (unsigned) buflen) | |
3835 | len = buflen; | |
3836 | if (copy_to_user(buffer, link, len)) | |
3837 | len = -EFAULT; | |
3838 | out: | |
3839 | return len; | |
3840 | } | |
3841 | ||
3842 | /* | |
3843 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3844 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3845 | * using) it for any given inode is up to filesystem. | |
3846 | */ | |
3847 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3848 | { | |
3849 | struct nameidata nd; | |
cc314eef | 3850 | void *cookie; |
694a1764 | 3851 | int res; |
cc314eef | 3852 | |
1da177e4 | 3853 | nd.depth = 0; |
cc314eef | 3854 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3855 | if (IS_ERR(cookie)) |
3856 | return PTR_ERR(cookie); | |
3857 | ||
3858 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3859 | if (dentry->d_inode->i_op->put_link) | |
3860 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3861 | return res; | |
1da177e4 LT |
3862 | } |
3863 | ||
3864 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3865 | { | |
3866 | return __vfs_follow_link(nd, link); | |
3867 | } | |
3868 | ||
3869 | /* get the link contents into pagecache */ | |
3870 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3871 | { | |
ebd09abb DG |
3872 | char *kaddr; |
3873 | struct page *page; | |
1da177e4 | 3874 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3875 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3876 | if (IS_ERR(page)) |
6fe6900e | 3877 | return (char*)page; |
1da177e4 | 3878 | *ppage = page; |
ebd09abb DG |
3879 | kaddr = kmap(page); |
3880 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3881 | return kaddr; | |
1da177e4 LT |
3882 | } |
3883 | ||
3884 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3885 | { | |
3886 | struct page *page = NULL; | |
3887 | char *s = page_getlink(dentry, &page); | |
3888 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3889 | if (page) { | |
3890 | kunmap(page); | |
3891 | page_cache_release(page); | |
3892 | } | |
3893 | return res; | |
3894 | } | |
3895 | ||
cc314eef | 3896 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3897 | { |
cc314eef | 3898 | struct page *page = NULL; |
1da177e4 | 3899 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3900 | return page; |
1da177e4 LT |
3901 | } |
3902 | ||
cc314eef | 3903 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3904 | { |
cc314eef LT |
3905 | struct page *page = cookie; |
3906 | ||
3907 | if (page) { | |
1da177e4 LT |
3908 | kunmap(page); |
3909 | page_cache_release(page); | |
1da177e4 LT |
3910 | } |
3911 | } | |
3912 | ||
54566b2c NP |
3913 | /* |
3914 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3915 | */ | |
3916 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3917 | { |
3918 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3919 | struct page *page; |
afddba49 | 3920 | void *fsdata; |
beb497ab | 3921 | int err; |
1da177e4 | 3922 | char *kaddr; |
54566b2c NP |
3923 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3924 | if (nofs) | |
3925 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3926 | |
7e53cac4 | 3927 | retry: |
afddba49 | 3928 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3929 | flags, &page, &fsdata); |
1da177e4 | 3930 | if (err) |
afddba49 NP |
3931 | goto fail; |
3932 | ||
e8e3c3d6 | 3933 | kaddr = kmap_atomic(page); |
1da177e4 | 3934 | memcpy(kaddr, symname, len-1); |
e8e3c3d6 | 3935 | kunmap_atomic(kaddr); |
afddba49 NP |
3936 | |
3937 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3938 | page, fsdata); | |
1da177e4 LT |
3939 | if (err < 0) |
3940 | goto fail; | |
afddba49 NP |
3941 | if (err < len-1) |
3942 | goto retry; | |
3943 | ||
1da177e4 LT |
3944 | mark_inode_dirty(inode); |
3945 | return 0; | |
1da177e4 LT |
3946 | fail: |
3947 | return err; | |
3948 | } | |
3949 | ||
0adb25d2 KK |
3950 | int page_symlink(struct inode *inode, const char *symname, int len) |
3951 | { | |
3952 | return __page_symlink(inode, symname, len, | |
54566b2c | 3953 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3954 | } |
3955 | ||
92e1d5be | 3956 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3957 | .readlink = generic_readlink, |
3958 | .follow_link = page_follow_link_light, | |
3959 | .put_link = page_put_link, | |
3960 | }; | |
3961 | ||
2d8f3038 | 3962 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3963 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3964 | EXPORT_SYMBOL(follow_down); |
3965 | EXPORT_SYMBOL(follow_up); | |
f6d2ac5c | 3966 | EXPORT_SYMBOL(get_write_access); /* nfsd */ |
1da177e4 LT |
3967 | EXPORT_SYMBOL(getname); |
3968 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3969 | EXPORT_SYMBOL(lookup_one_len); |
3970 | EXPORT_SYMBOL(page_follow_link_light); | |
3971 | EXPORT_SYMBOL(page_put_link); | |
3972 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3973 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3974 | EXPORT_SYMBOL(page_symlink); |
3975 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
d1811465 | 3976 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3977 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3978 | EXPORT_SYMBOL(inode_permission); |
1da177e4 LT |
3979 | EXPORT_SYMBOL(unlock_rename); |
3980 | EXPORT_SYMBOL(vfs_create); | |
3981 | EXPORT_SYMBOL(vfs_follow_link); | |
3982 | EXPORT_SYMBOL(vfs_link); | |
3983 | EXPORT_SYMBOL(vfs_mkdir); | |
3984 | EXPORT_SYMBOL(vfs_mknod); | |
3985 | EXPORT_SYMBOL(generic_permission); | |
3986 | EXPORT_SYMBOL(vfs_readlink); | |
3987 | EXPORT_SYMBOL(vfs_rename); | |
3988 | EXPORT_SYMBOL(vfs_rmdir); | |
3989 | EXPORT_SYMBOL(vfs_symlink); | |
3990 | EXPORT_SYMBOL(vfs_unlink); | |
3991 | EXPORT_SYMBOL(dentry_unhash); | |
3992 | EXPORT_SYMBOL(generic_readlink); |