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; | |
683 | if (current_cred()->fsuid == inode->i_uid) | |
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. */ | |
692 | if (parent->i_uid == inode->i_uid) | |
693 | return 0; | |
694 | ||
695 | path_put_conditional(link, nd); | |
696 | path_put(&nd->path); | |
a51d9eaa | 697 | audit_log_link_denied("follow_link", link); |
800179c9 KC |
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 | */ | |
762 | if (cred->fsuid == inode->i_uid || safe_hardlink_source(inode) || | |
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: | |
813 | path_put(&nd->path); | |
6d7b5aae | 814 | path_put(link); |
1da177e4 LT |
815 | return error; |
816 | } | |
817 | ||
31e6b01f NP |
818 | static int follow_up_rcu(struct path *path) |
819 | { | |
0714a533 AV |
820 | struct mount *mnt = real_mount(path->mnt); |
821 | struct mount *parent; | |
31e6b01f NP |
822 | struct dentry *mountpoint; |
823 | ||
0714a533 AV |
824 | parent = mnt->mnt_parent; |
825 | if (&parent->mnt == path->mnt) | |
31e6b01f | 826 | return 0; |
a73324da | 827 | mountpoint = mnt->mnt_mountpoint; |
31e6b01f | 828 | path->dentry = mountpoint; |
0714a533 | 829 | path->mnt = &parent->mnt; |
31e6b01f NP |
830 | return 1; |
831 | } | |
832 | ||
f015f126 DH |
833 | /* |
834 | * follow_up - Find the mountpoint of path's vfsmount | |
835 | * | |
836 | * Given a path, find the mountpoint of its source file system. | |
837 | * Replace @path with the path of the mountpoint in the parent mount. | |
838 | * Up is towards /. | |
839 | * | |
840 | * Return 1 if we went up a level and 0 if we were already at the | |
841 | * root. | |
842 | */ | |
bab77ebf | 843 | int follow_up(struct path *path) |
1da177e4 | 844 | { |
0714a533 AV |
845 | struct mount *mnt = real_mount(path->mnt); |
846 | struct mount *parent; | |
1da177e4 | 847 | struct dentry *mountpoint; |
99b7db7b | 848 | |
962830df | 849 | br_read_lock(&vfsmount_lock); |
0714a533 | 850 | parent = mnt->mnt_parent; |
3c0a6163 | 851 | if (parent == mnt) { |
962830df | 852 | br_read_unlock(&vfsmount_lock); |
1da177e4 LT |
853 | return 0; |
854 | } | |
0714a533 | 855 | mntget(&parent->mnt); |
a73324da | 856 | mountpoint = dget(mnt->mnt_mountpoint); |
962830df | 857 | br_read_unlock(&vfsmount_lock); |
bab77ebf AV |
858 | dput(path->dentry); |
859 | path->dentry = mountpoint; | |
860 | mntput(path->mnt); | |
0714a533 | 861 | path->mnt = &parent->mnt; |
1da177e4 LT |
862 | return 1; |
863 | } | |
864 | ||
b5c84bf6 | 865 | /* |
9875cf80 DH |
866 | * Perform an automount |
867 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
868 | * were called with. | |
1da177e4 | 869 | */ |
9875cf80 DH |
870 | static int follow_automount(struct path *path, unsigned flags, |
871 | bool *need_mntput) | |
31e6b01f | 872 | { |
9875cf80 | 873 | struct vfsmount *mnt; |
ea5b778a | 874 | int err; |
9875cf80 DH |
875 | |
876 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
877 | return -EREMOTE; | |
878 | ||
0ec26fd0 MS |
879 | /* We don't want to mount if someone's just doing a stat - |
880 | * unless they're stat'ing a directory and appended a '/' to | |
881 | * the name. | |
882 | * | |
883 | * We do, however, want to mount if someone wants to open or | |
884 | * create a file of any type under the mountpoint, wants to | |
885 | * traverse through the mountpoint or wants to open the | |
886 | * mounted directory. Also, autofs may mark negative dentries | |
887 | * as being automount points. These will need the attentions | |
888 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 889 | */ |
0ec26fd0 | 890 | if (!(flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
d94c177b | 891 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
0ec26fd0 MS |
892 | path->dentry->d_inode) |
893 | return -EISDIR; | |
894 | ||
9875cf80 DH |
895 | current->total_link_count++; |
896 | if (current->total_link_count >= 40) | |
897 | return -ELOOP; | |
898 | ||
899 | mnt = path->dentry->d_op->d_automount(path); | |
900 | if (IS_ERR(mnt)) { | |
901 | /* | |
902 | * The filesystem is allowed to return -EISDIR here to indicate | |
903 | * it doesn't want to automount. For instance, autofs would do | |
904 | * this so that its userspace daemon can mount on this dentry. | |
905 | * | |
906 | * However, we can only permit this if it's a terminal point in | |
907 | * the path being looked up; if it wasn't then the remainder of | |
908 | * the path is inaccessible and we should say so. | |
909 | */ | |
49084c3b | 910 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_PARENT)) |
9875cf80 DH |
911 | return -EREMOTE; |
912 | return PTR_ERR(mnt); | |
31e6b01f | 913 | } |
ea5b778a | 914 | |
9875cf80 DH |
915 | if (!mnt) /* mount collision */ |
916 | return 0; | |
31e6b01f | 917 | |
8aef1884 AV |
918 | if (!*need_mntput) { |
919 | /* lock_mount() may release path->mnt on error */ | |
920 | mntget(path->mnt); | |
921 | *need_mntput = true; | |
922 | } | |
19a167af | 923 | err = finish_automount(mnt, path); |
9875cf80 | 924 | |
ea5b778a DH |
925 | switch (err) { |
926 | case -EBUSY: | |
927 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 928 | return 0; |
ea5b778a | 929 | case 0: |
8aef1884 | 930 | path_put(path); |
ea5b778a DH |
931 | path->mnt = mnt; |
932 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 933 | return 0; |
19a167af AV |
934 | default: |
935 | return err; | |
ea5b778a | 936 | } |
19a167af | 937 | |
463ffb2e AV |
938 | } |
939 | ||
9875cf80 DH |
940 | /* |
941 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 942 | * - Flagged for transit management (autofs) |
9875cf80 DH |
943 | * - Flagged as mountpoint |
944 | * - Flagged as automount point | |
945 | * | |
946 | * This may only be called in refwalk mode. | |
947 | * | |
948 | * Serialization is taken care of in namespace.c | |
949 | */ | |
950 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 951 | { |
8aef1884 | 952 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
953 | unsigned managed; |
954 | bool need_mntput = false; | |
8aef1884 | 955 | int ret = 0; |
9875cf80 DH |
956 | |
957 | /* Given that we're not holding a lock here, we retain the value in a | |
958 | * local variable for each dentry as we look at it so that we don't see | |
959 | * the components of that value change under us */ | |
960 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
961 | managed &= DCACHE_MANAGED_DENTRY, | |
962 | unlikely(managed != 0)) { | |
cc53ce53 DH |
963 | /* Allow the filesystem to manage the transit without i_mutex |
964 | * being held. */ | |
965 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
966 | BUG_ON(!path->dentry->d_op); | |
967 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 968 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 969 | if (ret < 0) |
8aef1884 | 970 | break; |
cc53ce53 DH |
971 | } |
972 | ||
9875cf80 DH |
973 | /* Transit to a mounted filesystem. */ |
974 | if (managed & DCACHE_MOUNTED) { | |
975 | struct vfsmount *mounted = lookup_mnt(path); | |
976 | if (mounted) { | |
977 | dput(path->dentry); | |
978 | if (need_mntput) | |
979 | mntput(path->mnt); | |
980 | path->mnt = mounted; | |
981 | path->dentry = dget(mounted->mnt_root); | |
982 | need_mntput = true; | |
983 | continue; | |
984 | } | |
985 | ||
986 | /* Something is mounted on this dentry in another | |
987 | * namespace and/or whatever was mounted there in this | |
988 | * namespace got unmounted before we managed to get the | |
989 | * vfsmount_lock */ | |
990 | } | |
991 | ||
992 | /* Handle an automount point */ | |
993 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
994 | ret = follow_automount(path, flags, &need_mntput); | |
995 | if (ret < 0) | |
8aef1884 | 996 | break; |
9875cf80 DH |
997 | continue; |
998 | } | |
999 | ||
1000 | /* We didn't change the current path point */ | |
1001 | break; | |
1da177e4 | 1002 | } |
8aef1884 AV |
1003 | |
1004 | if (need_mntput && path->mnt == mnt) | |
1005 | mntput(path->mnt); | |
1006 | if (ret == -EISDIR) | |
1007 | ret = 0; | |
a3fbbde7 | 1008 | return ret < 0 ? ret : need_mntput; |
1da177e4 LT |
1009 | } |
1010 | ||
cc53ce53 | 1011 | int follow_down_one(struct path *path) |
1da177e4 LT |
1012 | { |
1013 | struct vfsmount *mounted; | |
1014 | ||
1c755af4 | 1015 | mounted = lookup_mnt(path); |
1da177e4 | 1016 | if (mounted) { |
9393bd07 AV |
1017 | dput(path->dentry); |
1018 | mntput(path->mnt); | |
1019 | path->mnt = mounted; | |
1020 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1021 | return 1; |
1022 | } | |
1023 | return 0; | |
1024 | } | |
1025 | ||
62a7375e IK |
1026 | static inline bool managed_dentry_might_block(struct dentry *dentry) |
1027 | { | |
1028 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT && | |
1029 | dentry->d_op->d_manage(dentry, true) < 0); | |
1030 | } | |
1031 | ||
9875cf80 | 1032 | /* |
287548e4 AV |
1033 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1034 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1035 | */ |
1036 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 1037 | struct inode **inode) |
9875cf80 | 1038 | { |
62a7375e | 1039 | for (;;) { |
c7105365 | 1040 | struct mount *mounted; |
62a7375e IK |
1041 | /* |
1042 | * Don't forget we might have a non-mountpoint managed dentry | |
1043 | * that wants to block transit. | |
1044 | */ | |
287548e4 | 1045 | if (unlikely(managed_dentry_might_block(path->dentry))) |
ab90911f | 1046 | return false; |
62a7375e IK |
1047 | |
1048 | if (!d_mountpoint(path->dentry)) | |
1049 | break; | |
1050 | ||
9875cf80 DH |
1051 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
1052 | if (!mounted) | |
1053 | break; | |
c7105365 AV |
1054 | path->mnt = &mounted->mnt; |
1055 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1056 | nd->flags |= LOOKUP_JUMPED; |
9875cf80 | 1057 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1058 | /* |
1059 | * Update the inode too. We don't need to re-check the | |
1060 | * dentry sequence number here after this d_inode read, | |
1061 | * because a mount-point is always pinned. | |
1062 | */ | |
1063 | *inode = path->dentry->d_inode; | |
9875cf80 | 1064 | } |
9875cf80 DH |
1065 | return true; |
1066 | } | |
1067 | ||
dea39376 | 1068 | static void follow_mount_rcu(struct nameidata *nd) |
287548e4 | 1069 | { |
dea39376 | 1070 | while (d_mountpoint(nd->path.dentry)) { |
c7105365 | 1071 | struct mount *mounted; |
dea39376 | 1072 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry, 1); |
287548e4 AV |
1073 | if (!mounted) |
1074 | break; | |
c7105365 AV |
1075 | nd->path.mnt = &mounted->mnt; |
1076 | nd->path.dentry = mounted->mnt.mnt_root; | |
dea39376 | 1077 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
287548e4 AV |
1078 | } |
1079 | } | |
1080 | ||
31e6b01f NP |
1081 | static int follow_dotdot_rcu(struct nameidata *nd) |
1082 | { | |
31e6b01f NP |
1083 | set_root_rcu(nd); |
1084 | ||
9875cf80 | 1085 | while (1) { |
31e6b01f NP |
1086 | if (nd->path.dentry == nd->root.dentry && |
1087 | nd->path.mnt == nd->root.mnt) { | |
1088 | break; | |
1089 | } | |
1090 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
1091 | struct dentry *old = nd->path.dentry; | |
1092 | struct dentry *parent = old->d_parent; | |
1093 | unsigned seq; | |
1094 | ||
1095 | seq = read_seqcount_begin(&parent->d_seq); | |
1096 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 1097 | goto failed; |
31e6b01f NP |
1098 | nd->path.dentry = parent; |
1099 | nd->seq = seq; | |
1100 | break; | |
1101 | } | |
1102 | if (!follow_up_rcu(&nd->path)) | |
1103 | break; | |
1104 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
31e6b01f | 1105 | } |
dea39376 AV |
1106 | follow_mount_rcu(nd); |
1107 | nd->inode = nd->path.dentry->d_inode; | |
31e6b01f | 1108 | return 0; |
ef7562d5 AV |
1109 | |
1110 | failed: | |
1111 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1112 | if (!(nd->flags & LOOKUP_ROOT)) |
1113 | nd->root.mnt = NULL; | |
32a7991b | 1114 | unlock_rcu_walk(); |
ef7562d5 | 1115 | return -ECHILD; |
31e6b01f NP |
1116 | } |
1117 | ||
cc53ce53 DH |
1118 | /* |
1119 | * Follow down to the covering mount currently visible to userspace. At each | |
1120 | * point, the filesystem owning that dentry may be queried as to whether the | |
1121 | * caller is permitted to proceed or not. | |
cc53ce53 | 1122 | */ |
7cc90cc3 | 1123 | int follow_down(struct path *path) |
cc53ce53 DH |
1124 | { |
1125 | unsigned managed; | |
1126 | int ret; | |
1127 | ||
1128 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1129 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1130 | /* Allow the filesystem to manage the transit without i_mutex | |
1131 | * being held. | |
1132 | * | |
1133 | * We indicate to the filesystem if someone is trying to mount | |
1134 | * something here. This gives autofs the chance to deny anyone | |
1135 | * other than its daemon the right to mount on its | |
1136 | * superstructure. | |
1137 | * | |
1138 | * The filesystem may sleep at this point. | |
1139 | */ | |
1140 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1141 | BUG_ON(!path->dentry->d_op); | |
1142 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 1143 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 1144 | path->dentry, false); |
cc53ce53 DH |
1145 | if (ret < 0) |
1146 | return ret == -EISDIR ? 0 : ret; | |
1147 | } | |
1148 | ||
1149 | /* Transit to a mounted filesystem. */ | |
1150 | if (managed & DCACHE_MOUNTED) { | |
1151 | struct vfsmount *mounted = lookup_mnt(path); | |
1152 | if (!mounted) | |
1153 | break; | |
1154 | dput(path->dentry); | |
1155 | mntput(path->mnt); | |
1156 | path->mnt = mounted; | |
1157 | path->dentry = dget(mounted->mnt_root); | |
1158 | continue; | |
1159 | } | |
1160 | ||
1161 | /* Don't handle automount points here */ | |
1162 | break; | |
1163 | } | |
1164 | return 0; | |
1165 | } | |
1166 | ||
9875cf80 DH |
1167 | /* |
1168 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1169 | */ | |
1170 | static void follow_mount(struct path *path) | |
1171 | { | |
1172 | while (d_mountpoint(path->dentry)) { | |
1173 | struct vfsmount *mounted = lookup_mnt(path); | |
1174 | if (!mounted) | |
1175 | break; | |
1176 | dput(path->dentry); | |
1177 | mntput(path->mnt); | |
1178 | path->mnt = mounted; | |
1179 | path->dentry = dget(mounted->mnt_root); | |
1180 | } | |
1181 | } | |
1182 | ||
31e6b01f | 1183 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1184 | { |
2a737871 | 1185 | set_root(nd); |
e518ddb7 | 1186 | |
1da177e4 | 1187 | while(1) { |
4ac91378 | 1188 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1189 | |
2a737871 AV |
1190 | if (nd->path.dentry == nd->root.dentry && |
1191 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1192 | break; |
1193 | } | |
4ac91378 | 1194 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1195 | /* rare case of legitimate dget_parent()... */ |
1196 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1197 | dput(old); |
1198 | break; | |
1199 | } | |
3088dd70 | 1200 | if (!follow_up(&nd->path)) |
1da177e4 | 1201 | break; |
1da177e4 | 1202 | } |
79ed0226 | 1203 | follow_mount(&nd->path); |
31e6b01f | 1204 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1205 | } |
1206 | ||
baa03890 | 1207 | /* |
bad61189 MS |
1208 | * This looks up the name in dcache, possibly revalidates the old dentry and |
1209 | * allocates a new one if not found or not valid. In the need_lookup argument | |
1210 | * returns whether i_op->lookup is necessary. | |
1211 | * | |
1212 | * dir->d_inode->i_mutex must be held | |
baa03890 | 1213 | */ |
bad61189 | 1214 | static struct dentry *lookup_dcache(struct qstr *name, struct dentry *dir, |
201f956e | 1215 | unsigned int flags, bool *need_lookup) |
baa03890 | 1216 | { |
baa03890 | 1217 | struct dentry *dentry; |
bad61189 | 1218 | int error; |
baa03890 | 1219 | |
bad61189 MS |
1220 | *need_lookup = false; |
1221 | dentry = d_lookup(dir, name); | |
1222 | if (dentry) { | |
1223 | if (d_need_lookup(dentry)) { | |
1224 | *need_lookup = true; | |
1225 | } else if (dentry->d_flags & DCACHE_OP_REVALIDATE) { | |
201f956e | 1226 | error = d_revalidate(dentry, flags); |
bad61189 MS |
1227 | if (unlikely(error <= 0)) { |
1228 | if (error < 0) { | |
1229 | dput(dentry); | |
1230 | return ERR_PTR(error); | |
1231 | } else if (!d_invalidate(dentry)) { | |
1232 | dput(dentry); | |
1233 | dentry = NULL; | |
1234 | } | |
1235 | } | |
1236 | } | |
1237 | } | |
baa03890 | 1238 | |
bad61189 MS |
1239 | if (!dentry) { |
1240 | dentry = d_alloc(dir, name); | |
1241 | if (unlikely(!dentry)) | |
1242 | return ERR_PTR(-ENOMEM); | |
baa03890 | 1243 | |
bad61189 | 1244 | *need_lookup = true; |
baa03890 NP |
1245 | } |
1246 | return dentry; | |
1247 | } | |
1248 | ||
44396f4b | 1249 | /* |
bad61189 MS |
1250 | * Call i_op->lookup on the dentry. The dentry must be negative but may be |
1251 | * hashed if it was pouplated with DCACHE_NEED_LOOKUP. | |
1252 | * | |
1253 | * dir->d_inode->i_mutex must be held | |
44396f4b | 1254 | */ |
bad61189 | 1255 | static struct dentry *lookup_real(struct inode *dir, struct dentry *dentry, |
72bd866a | 1256 | unsigned int flags) |
44396f4b | 1257 | { |
44396f4b JB |
1258 | struct dentry *old; |
1259 | ||
1260 | /* Don't create child dentry for a dead directory. */ | |
bad61189 | 1261 | if (unlikely(IS_DEADDIR(dir))) { |
e188dc02 | 1262 | dput(dentry); |
44396f4b | 1263 | return ERR_PTR(-ENOENT); |
e188dc02 | 1264 | } |
44396f4b | 1265 | |
72bd866a | 1266 | old = dir->i_op->lookup(dir, dentry, flags); |
44396f4b JB |
1267 | if (unlikely(old)) { |
1268 | dput(dentry); | |
1269 | dentry = old; | |
1270 | } | |
1271 | return dentry; | |
1272 | } | |
1273 | ||
a3255546 | 1274 | static struct dentry *__lookup_hash(struct qstr *name, |
72bd866a | 1275 | struct dentry *base, unsigned int flags) |
a3255546 | 1276 | { |
bad61189 | 1277 | bool need_lookup; |
a3255546 AV |
1278 | struct dentry *dentry; |
1279 | ||
72bd866a | 1280 | dentry = lookup_dcache(name, base, flags, &need_lookup); |
bad61189 MS |
1281 | if (!need_lookup) |
1282 | return dentry; | |
a3255546 | 1283 | |
72bd866a | 1284 | return lookup_real(base->d_inode, dentry, flags); |
a3255546 AV |
1285 | } |
1286 | ||
1da177e4 LT |
1287 | /* |
1288 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1289 | * small and for now I'd prefer to have fast path as straight as possible. | |
1290 | * It _is_ time-critical. | |
1291 | */ | |
697f514d MS |
1292 | static int lookup_fast(struct nameidata *nd, struct qstr *name, |
1293 | struct path *path, struct inode **inode) | |
1da177e4 | 1294 | { |
4ac91378 | 1295 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1296 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1297 | int need_reval = 1; |
1298 | int status = 1; | |
9875cf80 DH |
1299 | int err; |
1300 | ||
b04f784e NP |
1301 | /* |
1302 | * Rename seqlock is not required here because in the off chance | |
1303 | * of a false negative due to a concurrent rename, we're going to | |
1304 | * do the non-racy lookup, below. | |
1305 | */ | |
31e6b01f NP |
1306 | if (nd->flags & LOOKUP_RCU) { |
1307 | unsigned seq; | |
12f8ad4b | 1308 | dentry = __d_lookup_rcu(parent, name, &seq, nd->inode); |
5a18fff2 AV |
1309 | if (!dentry) |
1310 | goto unlazy; | |
1311 | ||
12f8ad4b LT |
1312 | /* |
1313 | * This sequence count validates that the inode matches | |
1314 | * the dentry name information from lookup. | |
1315 | */ | |
1316 | *inode = dentry->d_inode; | |
1317 | if (read_seqcount_retry(&dentry->d_seq, seq)) | |
1318 | return -ECHILD; | |
1319 | ||
1320 | /* | |
1321 | * This sequence count validates that the parent had no | |
1322 | * changes while we did the lookup of the dentry above. | |
1323 | * | |
1324 | * The memory barrier in read_seqcount_begin of child is | |
1325 | * enough, we can use __read_seqcount_retry here. | |
1326 | */ | |
31e6b01f NP |
1327 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) |
1328 | return -ECHILD; | |
31e6b01f | 1329 | nd->seq = seq; |
5a18fff2 | 1330 | |
fa4ee159 MS |
1331 | if (unlikely(d_need_lookup(dentry))) |
1332 | goto unlazy; | |
24643087 | 1333 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
4ce16ef3 | 1334 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1335 | if (unlikely(status <= 0)) { |
1336 | if (status != -ECHILD) | |
1337 | need_reval = 0; | |
1338 | goto unlazy; | |
1339 | } | |
24643087 | 1340 | } |
31e6b01f NP |
1341 | path->mnt = mnt; |
1342 | path->dentry = dentry; | |
d6e9bd25 AV |
1343 | if (unlikely(!__follow_mount_rcu(nd, path, inode))) |
1344 | goto unlazy; | |
1345 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1346 | goto unlazy; | |
1347 | return 0; | |
5a18fff2 | 1348 | unlazy: |
19660af7 AV |
1349 | if (unlazy_walk(nd, dentry)) |
1350 | return -ECHILD; | |
5a18fff2 AV |
1351 | } else { |
1352 | dentry = __d_lookup(parent, name); | |
9875cf80 | 1353 | } |
5a18fff2 | 1354 | |
81e6f520 AV |
1355 | if (unlikely(!dentry)) |
1356 | goto need_lookup; | |
1357 | ||
1358 | if (unlikely(d_need_lookup(dentry))) { | |
44396f4b | 1359 | dput(dentry); |
81e6f520 | 1360 | goto need_lookup; |
5a18fff2 | 1361 | } |
81e6f520 | 1362 | |
5a18fff2 | 1363 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) |
4ce16ef3 | 1364 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1365 | if (unlikely(status <= 0)) { |
1366 | if (status < 0) { | |
1367 | dput(dentry); | |
1368 | return status; | |
1369 | } | |
1370 | if (!d_invalidate(dentry)) { | |
1371 | dput(dentry); | |
81e6f520 | 1372 | goto need_lookup; |
5a18fff2 | 1373 | } |
24643087 | 1374 | } |
697f514d | 1375 | |
9875cf80 DH |
1376 | path->mnt = mnt; |
1377 | path->dentry = dentry; | |
1378 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1379 | if (unlikely(err < 0)) { |
1380 | path_put_conditional(path, nd); | |
9875cf80 | 1381 | return err; |
89312214 | 1382 | } |
a3fbbde7 AV |
1383 | if (err) |
1384 | nd->flags |= LOOKUP_JUMPED; | |
9875cf80 | 1385 | *inode = path->dentry->d_inode; |
1da177e4 | 1386 | return 0; |
81e6f520 AV |
1387 | |
1388 | need_lookup: | |
697f514d MS |
1389 | return 1; |
1390 | } | |
1391 | ||
1392 | /* Fast lookup failed, do it the slow way */ | |
1393 | static int lookup_slow(struct nameidata *nd, struct qstr *name, | |
1394 | struct path *path) | |
1395 | { | |
1396 | struct dentry *dentry, *parent; | |
1397 | int err; | |
1398 | ||
1399 | parent = nd->path.dentry; | |
81e6f520 AV |
1400 | BUG_ON(nd->inode != parent->d_inode); |
1401 | ||
1402 | mutex_lock(&parent->d_inode->i_mutex); | |
72bd866a | 1403 | dentry = __lookup_hash(name, parent, nd->flags); |
81e6f520 AV |
1404 | mutex_unlock(&parent->d_inode->i_mutex); |
1405 | if (IS_ERR(dentry)) | |
1406 | return PTR_ERR(dentry); | |
697f514d MS |
1407 | path->mnt = nd->path.mnt; |
1408 | path->dentry = dentry; | |
1409 | err = follow_managed(path, nd->flags); | |
1410 | if (unlikely(err < 0)) { | |
1411 | path_put_conditional(path, nd); | |
1412 | return err; | |
1413 | } | |
1414 | if (err) | |
1415 | nd->flags |= LOOKUP_JUMPED; | |
1416 | return 0; | |
1da177e4 LT |
1417 | } |
1418 | ||
52094c8a AV |
1419 | static inline int may_lookup(struct nameidata *nd) |
1420 | { | |
1421 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1422 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1423 | if (err != -ECHILD) |
1424 | return err; | |
19660af7 | 1425 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1426 | return -ECHILD; |
1427 | } | |
4ad5abb3 | 1428 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1429 | } |
1430 | ||
9856fa1b AV |
1431 | static inline int handle_dots(struct nameidata *nd, int type) |
1432 | { | |
1433 | if (type == LAST_DOTDOT) { | |
1434 | if (nd->flags & LOOKUP_RCU) { | |
1435 | if (follow_dotdot_rcu(nd)) | |
1436 | return -ECHILD; | |
1437 | } else | |
1438 | follow_dotdot(nd); | |
1439 | } | |
1440 | return 0; | |
1441 | } | |
1442 | ||
951361f9 AV |
1443 | static void terminate_walk(struct nameidata *nd) |
1444 | { | |
1445 | if (!(nd->flags & LOOKUP_RCU)) { | |
1446 | path_put(&nd->path); | |
1447 | } else { | |
1448 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1449 | if (!(nd->flags & LOOKUP_ROOT)) |
1450 | nd->root.mnt = NULL; | |
32a7991b | 1451 | unlock_rcu_walk(); |
951361f9 AV |
1452 | } |
1453 | } | |
1454 | ||
3ddcd056 LT |
1455 | /* |
1456 | * Do we need to follow links? We _really_ want to be able | |
1457 | * to do this check without having to look at inode->i_op, | |
1458 | * so we keep a cache of "no, this doesn't need follow_link" | |
1459 | * for the common case. | |
1460 | */ | |
7813b94a | 1461 | static inline int should_follow_link(struct inode *inode, int follow) |
3ddcd056 LT |
1462 | { |
1463 | if (unlikely(!(inode->i_opflags & IOP_NOFOLLOW))) { | |
1464 | if (likely(inode->i_op->follow_link)) | |
1465 | return follow; | |
1466 | ||
1467 | /* This gets set once for the inode lifetime */ | |
1468 | spin_lock(&inode->i_lock); | |
1469 | inode->i_opflags |= IOP_NOFOLLOW; | |
1470 | spin_unlock(&inode->i_lock); | |
1471 | } | |
1472 | return 0; | |
1473 | } | |
1474 | ||
ce57dfc1 AV |
1475 | static inline int walk_component(struct nameidata *nd, struct path *path, |
1476 | struct qstr *name, int type, int follow) | |
1477 | { | |
1478 | struct inode *inode; | |
1479 | int err; | |
1480 | /* | |
1481 | * "." and ".." are special - ".." especially so because it has | |
1482 | * to be able to know about the current root directory and | |
1483 | * parent relationships. | |
1484 | */ | |
1485 | if (unlikely(type != LAST_NORM)) | |
1486 | return handle_dots(nd, type); | |
697f514d | 1487 | err = lookup_fast(nd, name, path, &inode); |
ce57dfc1 | 1488 | if (unlikely(err)) { |
697f514d MS |
1489 | if (err < 0) |
1490 | goto out_err; | |
1491 | ||
1492 | err = lookup_slow(nd, name, path); | |
1493 | if (err < 0) | |
1494 | goto out_err; | |
1495 | ||
1496 | inode = path->dentry->d_inode; | |
ce57dfc1 | 1497 | } |
697f514d MS |
1498 | err = -ENOENT; |
1499 | if (!inode) | |
1500 | goto out_path_put; | |
1501 | ||
7813b94a | 1502 | if (should_follow_link(inode, follow)) { |
19660af7 AV |
1503 | if (nd->flags & LOOKUP_RCU) { |
1504 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
697f514d MS |
1505 | err = -ECHILD; |
1506 | goto out_err; | |
19660af7 AV |
1507 | } |
1508 | } | |
ce57dfc1 AV |
1509 | BUG_ON(inode != path->dentry->d_inode); |
1510 | return 1; | |
1511 | } | |
1512 | path_to_nameidata(path, nd); | |
1513 | nd->inode = inode; | |
1514 | return 0; | |
697f514d MS |
1515 | |
1516 | out_path_put: | |
1517 | path_to_nameidata(path, nd); | |
1518 | out_err: | |
1519 | terminate_walk(nd); | |
1520 | return err; | |
ce57dfc1 AV |
1521 | } |
1522 | ||
b356379a AV |
1523 | /* |
1524 | * This limits recursive symlink follows to 8, while | |
1525 | * limiting consecutive symlinks to 40. | |
1526 | * | |
1527 | * Without that kind of total limit, nasty chains of consecutive | |
1528 | * symlinks can cause almost arbitrarily long lookups. | |
1529 | */ | |
1530 | static inline int nested_symlink(struct path *path, struct nameidata *nd) | |
1531 | { | |
1532 | int res; | |
1533 | ||
b356379a AV |
1534 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1535 | path_put_conditional(path, nd); | |
1536 | path_put(&nd->path); | |
1537 | return -ELOOP; | |
1538 | } | |
1a4022f8 | 1539 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); |
b356379a AV |
1540 | |
1541 | nd->depth++; | |
1542 | current->link_count++; | |
1543 | ||
1544 | do { | |
1545 | struct path link = *path; | |
1546 | void *cookie; | |
574197e0 AV |
1547 | |
1548 | res = follow_link(&link, nd, &cookie); | |
6d7b5aae AV |
1549 | if (res) |
1550 | break; | |
1551 | res = walk_component(nd, path, &nd->last, | |
1552 | nd->last_type, LOOKUP_FOLLOW); | |
574197e0 | 1553 | put_link(nd, &link, cookie); |
b356379a AV |
1554 | } while (res > 0); |
1555 | ||
1556 | current->link_count--; | |
1557 | nd->depth--; | |
1558 | return res; | |
1559 | } | |
1560 | ||
3ddcd056 LT |
1561 | /* |
1562 | * We really don't want to look at inode->i_op->lookup | |
1563 | * when we don't have to. So we keep a cache bit in | |
1564 | * the inode ->i_opflags field that says "yes, we can | |
1565 | * do lookup on this inode". | |
1566 | */ | |
1567 | static inline int can_lookup(struct inode *inode) | |
1568 | { | |
1569 | if (likely(inode->i_opflags & IOP_LOOKUP)) | |
1570 | return 1; | |
1571 | if (likely(!inode->i_op->lookup)) | |
1572 | return 0; | |
1573 | ||
1574 | /* We do this once for the lifetime of the inode */ | |
1575 | spin_lock(&inode->i_lock); | |
1576 | inode->i_opflags |= IOP_LOOKUP; | |
1577 | spin_unlock(&inode->i_lock); | |
1578 | return 1; | |
1579 | } | |
1580 | ||
bfcfaa77 LT |
1581 | /* |
1582 | * We can do the critical dentry name comparison and hashing | |
1583 | * operations one word at a time, but we are limited to: | |
1584 | * | |
1585 | * - Architectures with fast unaligned word accesses. We could | |
1586 | * do a "get_unaligned()" if this helps and is sufficiently | |
1587 | * fast. | |
1588 | * | |
1589 | * - Little-endian machines (so that we can generate the mask | |
1590 | * of low bytes efficiently). Again, we *could* do a byte | |
1591 | * swapping load on big-endian architectures if that is not | |
1592 | * expensive enough to make the optimization worthless. | |
1593 | * | |
1594 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we | |
1595 | * do not trap on the (extremely unlikely) case of a page | |
1596 | * crossing operation. | |
1597 | * | |
1598 | * - Furthermore, we need an efficient 64-bit compile for the | |
1599 | * 64-bit case in order to generate the "number of bytes in | |
1600 | * the final mask". Again, that could be replaced with a | |
1601 | * efficient population count instruction or similar. | |
1602 | */ | |
1603 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1604 | ||
f68e556e | 1605 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1606 | |
f68e556e | 1607 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1608 | |
1609 | static inline unsigned int fold_hash(unsigned long hash) | |
1610 | { | |
1611 | hash += hash >> (8*sizeof(int)); | |
1612 | return hash; | |
1613 | } | |
1614 | ||
1615 | #else /* 32-bit case */ | |
1616 | ||
bfcfaa77 LT |
1617 | #define fold_hash(x) (x) |
1618 | ||
1619 | #endif | |
1620 | ||
1621 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1622 | { | |
1623 | unsigned long a, mask; | |
1624 | unsigned long hash = 0; | |
1625 | ||
1626 | for (;;) { | |
e419b4cc | 1627 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1628 | if (len < sizeof(unsigned long)) |
1629 | break; | |
1630 | hash += a; | |
f132c5be | 1631 | hash *= 9; |
bfcfaa77 LT |
1632 | name += sizeof(unsigned long); |
1633 | len -= sizeof(unsigned long); | |
1634 | if (!len) | |
1635 | goto done; | |
1636 | } | |
1637 | mask = ~(~0ul << len*8); | |
1638 | hash += mask & a; | |
1639 | done: | |
1640 | return fold_hash(hash); | |
1641 | } | |
1642 | EXPORT_SYMBOL(full_name_hash); | |
1643 | ||
bfcfaa77 LT |
1644 | /* |
1645 | * Calculate the length and hash of the path component, and | |
1646 | * return the length of the component; | |
1647 | */ | |
1648 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1649 | { | |
36126f8f LT |
1650 | unsigned long a, b, adata, bdata, mask, hash, len; |
1651 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; | |
bfcfaa77 LT |
1652 | |
1653 | hash = a = 0; | |
1654 | len = -sizeof(unsigned long); | |
1655 | do { | |
1656 | hash = (hash + a) * 9; | |
1657 | len += sizeof(unsigned long); | |
e419b4cc | 1658 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1659 | b = a ^ REPEAT_BYTE('/'); |
1660 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1661 | ||
1662 | adata = prep_zero_mask(a, adata, &constants); | |
1663 | bdata = prep_zero_mask(b, bdata, &constants); | |
1664 | ||
1665 | mask = create_zero_mask(adata | bdata); | |
1666 | ||
1667 | hash += a & zero_bytemask(mask); | |
bfcfaa77 LT |
1668 | *hashp = fold_hash(hash); |
1669 | ||
36126f8f | 1670 | return len + find_zero(mask); |
bfcfaa77 LT |
1671 | } |
1672 | ||
1673 | #else | |
1674 | ||
0145acc2 LT |
1675 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1676 | { | |
1677 | unsigned long hash = init_name_hash(); | |
1678 | while (len--) | |
1679 | hash = partial_name_hash(*name++, hash); | |
1680 | return end_name_hash(hash); | |
1681 | } | |
ae942ae7 | 1682 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1683 | |
200e9ef7 LT |
1684 | /* |
1685 | * We know there's a real path component here of at least | |
1686 | * one character. | |
1687 | */ | |
1688 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1689 | { | |
1690 | unsigned long hash = init_name_hash(); | |
1691 | unsigned long len = 0, c; | |
1692 | ||
1693 | c = (unsigned char)*name; | |
1694 | do { | |
1695 | len++; | |
1696 | hash = partial_name_hash(c, hash); | |
1697 | c = (unsigned char)name[len]; | |
1698 | } while (c && c != '/'); | |
1699 | *hashp = end_name_hash(hash); | |
1700 | return len; | |
1701 | } | |
1702 | ||
bfcfaa77 LT |
1703 | #endif |
1704 | ||
1da177e4 LT |
1705 | /* |
1706 | * Name resolution. | |
ea3834d9 PM |
1707 | * This is the basic name resolution function, turning a pathname into |
1708 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1709 | * |
ea3834d9 PM |
1710 | * Returns 0 and nd will have valid dentry and mnt on success. |
1711 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1712 | */ |
6de88d72 | 1713 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1714 | { |
1715 | struct path next; | |
1da177e4 | 1716 | int err; |
1da177e4 LT |
1717 | |
1718 | while (*name=='/') | |
1719 | name++; | |
1720 | if (!*name) | |
086e183a | 1721 | return 0; |
1da177e4 | 1722 | |
1da177e4 LT |
1723 | /* At this point we know we have a real path component. */ |
1724 | for(;;) { | |
1da177e4 | 1725 | struct qstr this; |
200e9ef7 | 1726 | long len; |
fe479a58 | 1727 | int type; |
1da177e4 | 1728 | |
52094c8a | 1729 | err = may_lookup(nd); |
1da177e4 LT |
1730 | if (err) |
1731 | break; | |
1732 | ||
200e9ef7 | 1733 | len = hash_name(name, &this.hash); |
1da177e4 | 1734 | this.name = name; |
200e9ef7 | 1735 | this.len = len; |
1da177e4 | 1736 | |
fe479a58 | 1737 | type = LAST_NORM; |
200e9ef7 | 1738 | if (name[0] == '.') switch (len) { |
fe479a58 | 1739 | case 2: |
200e9ef7 | 1740 | if (name[1] == '.') { |
fe479a58 | 1741 | type = LAST_DOTDOT; |
16c2cd71 AV |
1742 | nd->flags |= LOOKUP_JUMPED; |
1743 | } | |
fe479a58 AV |
1744 | break; |
1745 | case 1: | |
1746 | type = LAST_DOT; | |
1747 | } | |
5a202bcd AV |
1748 | if (likely(type == LAST_NORM)) { |
1749 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1750 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd AV |
1751 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
1752 | err = parent->d_op->d_hash(parent, nd->inode, | |
1753 | &this); | |
1754 | if (err < 0) | |
1755 | break; | |
1756 | } | |
1757 | } | |
fe479a58 | 1758 | |
200e9ef7 | 1759 | if (!name[len]) |
1da177e4 | 1760 | goto last_component; |
200e9ef7 LT |
1761 | /* |
1762 | * If it wasn't NUL, we know it was '/'. Skip that | |
1763 | * slash, and continue until no more slashes. | |
1764 | */ | |
1765 | do { | |
1766 | len++; | |
1767 | } while (unlikely(name[len] == '/')); | |
1768 | if (!name[len]) | |
b356379a | 1769 | goto last_component; |
200e9ef7 | 1770 | name += len; |
1da177e4 | 1771 | |
ce57dfc1 AV |
1772 | err = walk_component(nd, &next, &this, type, LOOKUP_FOLLOW); |
1773 | if (err < 0) | |
1774 | return err; | |
1da177e4 | 1775 | |
ce57dfc1 | 1776 | if (err) { |
b356379a | 1777 | err = nested_symlink(&next, nd); |
1da177e4 | 1778 | if (err) |
a7472bab | 1779 | return err; |
31e6b01f | 1780 | } |
3ddcd056 LT |
1781 | if (can_lookup(nd->inode)) |
1782 | continue; | |
1da177e4 | 1783 | err = -ENOTDIR; |
3ddcd056 | 1784 | break; |
1da177e4 LT |
1785 | /* here ends the main loop */ |
1786 | ||
1da177e4 | 1787 | last_component: |
b356379a AV |
1788 | nd->last = this; |
1789 | nd->last_type = type; | |
086e183a | 1790 | return 0; |
1da177e4 | 1791 | } |
951361f9 | 1792 | terminate_walk(nd); |
1da177e4 LT |
1793 | return err; |
1794 | } | |
1795 | ||
70e9b357 AV |
1796 | static int path_init(int dfd, const char *name, unsigned int flags, |
1797 | struct nameidata *nd, struct file **fp) | |
31e6b01f NP |
1798 | { |
1799 | int retval = 0; | |
1800 | int fput_needed; | |
1801 | struct file *file; | |
1802 | ||
1803 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
16c2cd71 | 1804 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 1805 | nd->depth = 0; |
5b6ca027 AV |
1806 | if (flags & LOOKUP_ROOT) { |
1807 | struct inode *inode = nd->root.dentry->d_inode; | |
73d049a4 AV |
1808 | if (*name) { |
1809 | if (!inode->i_op->lookup) | |
1810 | return -ENOTDIR; | |
1811 | retval = inode_permission(inode, MAY_EXEC); | |
1812 | if (retval) | |
1813 | return retval; | |
1814 | } | |
5b6ca027 AV |
1815 | nd->path = nd->root; |
1816 | nd->inode = inode; | |
1817 | if (flags & LOOKUP_RCU) { | |
32a7991b | 1818 | lock_rcu_walk(); |
5b6ca027 AV |
1819 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
1820 | } else { | |
1821 | path_get(&nd->path); | |
1822 | } | |
1823 | return 0; | |
1824 | } | |
1825 | ||
31e6b01f | 1826 | nd->root.mnt = NULL; |
31e6b01f NP |
1827 | |
1828 | if (*name=='/') { | |
e41f7d4e | 1829 | if (flags & LOOKUP_RCU) { |
32a7991b | 1830 | lock_rcu_walk(); |
e41f7d4e AV |
1831 | set_root_rcu(nd); |
1832 | } else { | |
1833 | set_root(nd); | |
1834 | path_get(&nd->root); | |
1835 | } | |
1836 | nd->path = nd->root; | |
31e6b01f | 1837 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1838 | if (flags & LOOKUP_RCU) { |
1839 | struct fs_struct *fs = current->fs; | |
1840 | unsigned seq; | |
31e6b01f | 1841 | |
32a7991b | 1842 | lock_rcu_walk(); |
c28cc364 | 1843 | |
e41f7d4e AV |
1844 | do { |
1845 | seq = read_seqcount_begin(&fs->seq); | |
1846 | nd->path = fs->pwd; | |
1847 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1848 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1849 | } else { | |
1850 | get_fs_pwd(current->fs, &nd->path); | |
1851 | } | |
31e6b01f NP |
1852 | } else { |
1853 | struct dentry *dentry; | |
1854 | ||
1abf0c71 | 1855 | file = fget_raw_light(dfd, &fput_needed); |
31e6b01f NP |
1856 | retval = -EBADF; |
1857 | if (!file) | |
1858 | goto out_fail; | |
1859 | ||
1860 | dentry = file->f_path.dentry; | |
1861 | ||
f52e0c11 AV |
1862 | if (*name) { |
1863 | retval = -ENOTDIR; | |
1864 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
1865 | goto fput_fail; | |
31e6b01f | 1866 | |
4ad5abb3 | 1867 | retval = inode_permission(dentry->d_inode, MAY_EXEC); |
f52e0c11 AV |
1868 | if (retval) |
1869 | goto fput_fail; | |
1870 | } | |
31e6b01f NP |
1871 | |
1872 | nd->path = file->f_path; | |
e41f7d4e AV |
1873 | if (flags & LOOKUP_RCU) { |
1874 | if (fput_needed) | |
70e9b357 | 1875 | *fp = file; |
e41f7d4e | 1876 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
32a7991b | 1877 | lock_rcu_walk(); |
e41f7d4e AV |
1878 | } else { |
1879 | path_get(&file->f_path); | |
1880 | fput_light(file, fput_needed); | |
1881 | } | |
31e6b01f | 1882 | } |
31e6b01f | 1883 | |
31e6b01f | 1884 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1885 | return 0; |
2dfdd266 | 1886 | |
9b4a9b14 AV |
1887 | fput_fail: |
1888 | fput_light(file, fput_needed); | |
1889 | out_fail: | |
1890 | return retval; | |
1891 | } | |
1892 | ||
bd92d7fe AV |
1893 | static inline int lookup_last(struct nameidata *nd, struct path *path) |
1894 | { | |
1895 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1896 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1897 | ||
1898 | nd->flags &= ~LOOKUP_PARENT; | |
1899 | return walk_component(nd, path, &nd->last, nd->last_type, | |
1900 | nd->flags & LOOKUP_FOLLOW); | |
1901 | } | |
1902 | ||
9b4a9b14 | 1903 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
ee0827cd | 1904 | static int path_lookupat(int dfd, const char *name, |
9b4a9b14 AV |
1905 | unsigned int flags, struct nameidata *nd) |
1906 | { | |
70e9b357 | 1907 | struct file *base = NULL; |
bd92d7fe AV |
1908 | struct path path; |
1909 | int err; | |
31e6b01f NP |
1910 | |
1911 | /* | |
1912 | * Path walking is largely split up into 2 different synchronisation | |
1913 | * schemes, rcu-walk and ref-walk (explained in | |
1914 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1915 | * path walk code, but some things particularly setup, cleanup, and | |
1916 | * following mounts are sufficiently divergent that functions are | |
1917 | * duplicated. Typically there is a function foo(), and its RCU | |
1918 | * analogue, foo_rcu(). | |
1919 | * | |
1920 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1921 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1922 | * be handled by restarting a traditional ref-walk (which will always | |
1923 | * be able to complete). | |
1924 | */ | |
bd92d7fe | 1925 | err = path_init(dfd, name, flags | LOOKUP_PARENT, nd, &base); |
ee0827cd | 1926 | |
bd92d7fe AV |
1927 | if (unlikely(err)) |
1928 | return err; | |
ee0827cd AV |
1929 | |
1930 | current->total_link_count = 0; | |
bd92d7fe AV |
1931 | err = link_path_walk(name, nd); |
1932 | ||
1933 | if (!err && !(flags & LOOKUP_PARENT)) { | |
bd92d7fe AV |
1934 | err = lookup_last(nd, &path); |
1935 | while (err > 0) { | |
1936 | void *cookie; | |
1937 | struct path link = path; | |
800179c9 KC |
1938 | err = may_follow_link(&link, nd); |
1939 | if (unlikely(err)) | |
1940 | break; | |
bd92d7fe | 1941 | nd->flags |= LOOKUP_PARENT; |
574197e0 | 1942 | err = follow_link(&link, nd, &cookie); |
6d7b5aae AV |
1943 | if (err) |
1944 | break; | |
1945 | err = lookup_last(nd, &path); | |
574197e0 | 1946 | put_link(nd, &link, cookie); |
bd92d7fe AV |
1947 | } |
1948 | } | |
ee0827cd | 1949 | |
9f1fafee AV |
1950 | if (!err) |
1951 | err = complete_walk(nd); | |
bd92d7fe AV |
1952 | |
1953 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
1954 | if (!nd->inode->i_op->lookup) { | |
1955 | path_put(&nd->path); | |
bd23a539 | 1956 | err = -ENOTDIR; |
bd92d7fe AV |
1957 | } |
1958 | } | |
16c2cd71 | 1959 | |
70e9b357 AV |
1960 | if (base) |
1961 | fput(base); | |
ee0827cd | 1962 | |
5b6ca027 | 1963 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
2a737871 AV |
1964 | path_put(&nd->root); |
1965 | nd->root.mnt = NULL; | |
1966 | } | |
bd92d7fe | 1967 | return err; |
ee0827cd | 1968 | } |
31e6b01f | 1969 | |
ee0827cd AV |
1970 | static int do_path_lookup(int dfd, const char *name, |
1971 | unsigned int flags, struct nameidata *nd) | |
1972 | { | |
1973 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); | |
1974 | if (unlikely(retval == -ECHILD)) | |
1975 | retval = path_lookupat(dfd, name, flags, nd); | |
1976 | if (unlikely(retval == -ESTALE)) | |
1977 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); | |
31e6b01f NP |
1978 | |
1979 | if (likely(!retval)) { | |
1980 | if (unlikely(!audit_dummy_context())) { | |
1981 | if (nd->path.dentry && nd->inode) | |
1982 | audit_inode(name, nd->path.dentry); | |
1983 | } | |
1984 | } | |
170aa3d0 | 1985 | return retval; |
1da177e4 LT |
1986 | } |
1987 | ||
79714f72 AV |
1988 | /* does lookup, returns the object with parent locked */ |
1989 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 1990 | { |
79714f72 AV |
1991 | struct nameidata nd; |
1992 | struct dentry *d; | |
1993 | int err = do_path_lookup(AT_FDCWD, name, LOOKUP_PARENT, &nd); | |
1994 | if (err) | |
1995 | return ERR_PTR(err); | |
1996 | if (nd.last_type != LAST_NORM) { | |
1997 | path_put(&nd.path); | |
1998 | return ERR_PTR(-EINVAL); | |
1999 | } | |
2000 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); | |
1e0ea001 | 2001 | d = __lookup_hash(&nd.last, nd.path.dentry, 0); |
79714f72 AV |
2002 | if (IS_ERR(d)) { |
2003 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); | |
2004 | path_put(&nd.path); | |
2005 | return d; | |
2006 | } | |
2007 | *path = nd.path; | |
2008 | return d; | |
5590ff0d UD |
2009 | } |
2010 | ||
d1811465 AV |
2011 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2012 | { | |
2013 | struct nameidata nd; | |
2014 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
2015 | if (!res) | |
2016 | *path = nd.path; | |
2017 | return res; | |
2018 | } | |
2019 | ||
16f18200 JJS |
2020 | /** |
2021 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2022 | * @dentry: pointer to dentry of the base directory | |
2023 | * @mnt: pointer to vfs mount of the base directory | |
2024 | * @name: pointer to file name | |
2025 | * @flags: lookup flags | |
e0a01249 | 2026 | * @path: pointer to struct path to fill |
16f18200 JJS |
2027 | */ |
2028 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2029 | const char *name, unsigned int flags, | |
e0a01249 | 2030 | struct path *path) |
16f18200 | 2031 | { |
e0a01249 AV |
2032 | struct nameidata nd; |
2033 | int err; | |
2034 | nd.root.dentry = dentry; | |
2035 | nd.root.mnt = mnt; | |
2036 | BUG_ON(flags & LOOKUP_PARENT); | |
5b6ca027 | 2037 | /* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */ |
e0a01249 AV |
2038 | err = do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, &nd); |
2039 | if (!err) | |
2040 | *path = nd.path; | |
2041 | return err; | |
16f18200 JJS |
2042 | } |
2043 | ||
057f6c01 JM |
2044 | /* |
2045 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
2046 | * needs parent already locked. Doesn't follow mounts. | |
2047 | * SMP-safe. | |
2048 | */ | |
eead1911 | 2049 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 2050 | { |
72bd866a | 2051 | return __lookup_hash(&nd->last, nd->path.dentry, nd->flags); |
1da177e4 LT |
2052 | } |
2053 | ||
eead1911 | 2054 | /** |
a6b91919 | 2055 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
2056 | * @name: pathname component to lookup |
2057 | * @base: base directory to lookup from | |
2058 | * @len: maximum length @len should be interpreted to | |
2059 | * | |
a6b91919 RD |
2060 | * Note that this routine is purely a helper for filesystem usage and should |
2061 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
2062 | * nameidata argument is passed to the filesystem methods and a filesystem |
2063 | * using this helper needs to be prepared for that. | |
2064 | */ | |
057f6c01 JM |
2065 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
2066 | { | |
057f6c01 | 2067 | struct qstr this; |
6a96ba54 | 2068 | unsigned int c; |
cda309de | 2069 | int err; |
057f6c01 | 2070 | |
2f9092e1 DW |
2071 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
2072 | ||
6a96ba54 AV |
2073 | this.name = name; |
2074 | this.len = len; | |
0145acc2 | 2075 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
2076 | if (!len) |
2077 | return ERR_PTR(-EACCES); | |
2078 | ||
6a96ba54 AV |
2079 | while (len--) { |
2080 | c = *(const unsigned char *)name++; | |
2081 | if (c == '/' || c == '\0') | |
2082 | return ERR_PTR(-EACCES); | |
6a96ba54 | 2083 | } |
5a202bcd AV |
2084 | /* |
2085 | * See if the low-level filesystem might want | |
2086 | * to use its own hash.. | |
2087 | */ | |
2088 | if (base->d_flags & DCACHE_OP_HASH) { | |
2089 | int err = base->d_op->d_hash(base, base->d_inode, &this); | |
2090 | if (err < 0) | |
2091 | return ERR_PTR(err); | |
2092 | } | |
eead1911 | 2093 | |
cda309de MS |
2094 | err = inode_permission(base->d_inode, MAY_EXEC); |
2095 | if (err) | |
2096 | return ERR_PTR(err); | |
2097 | ||
72bd866a | 2098 | return __lookup_hash(&this, base, 0); |
057f6c01 JM |
2099 | } |
2100 | ||
1fa1e7f6 AW |
2101 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2102 | struct path *path, int *empty) | |
1da177e4 | 2103 | { |
2d8f3038 | 2104 | struct nameidata nd; |
1fa1e7f6 | 2105 | char *tmp = getname_flags(name, flags, empty); |
1da177e4 | 2106 | int err = PTR_ERR(tmp); |
1da177e4 | 2107 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
2108 | |
2109 | BUG_ON(flags & LOOKUP_PARENT); | |
2110 | ||
2111 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 2112 | putname(tmp); |
2d8f3038 AV |
2113 | if (!err) |
2114 | *path = nd.path; | |
1da177e4 LT |
2115 | } |
2116 | return err; | |
2117 | } | |
2118 | ||
1fa1e7f6 AW |
2119 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
2120 | struct path *path) | |
2121 | { | |
f7493e5d | 2122 | return user_path_at_empty(dfd, name, flags, path, NULL); |
1fa1e7f6 AW |
2123 | } |
2124 | ||
2ad94ae6 AV |
2125 | static int user_path_parent(int dfd, const char __user *path, |
2126 | struct nameidata *nd, char **name) | |
2127 | { | |
2128 | char *s = getname(path); | |
2129 | int error; | |
2130 | ||
2131 | if (IS_ERR(s)) | |
2132 | return PTR_ERR(s); | |
2133 | ||
2134 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
2135 | if (error) | |
2136 | putname(s); | |
2137 | else | |
2138 | *name = s; | |
2139 | ||
2140 | return error; | |
2141 | } | |
2142 | ||
1da177e4 LT |
2143 | /* |
2144 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
2145 | * minimal. | |
2146 | */ | |
2147 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
2148 | { | |
8e96e3b7 | 2149 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2150 | |
1da177e4 LT |
2151 | if (!(dir->i_mode & S_ISVTX)) |
2152 | return 0; | |
8e96e3b7 | 2153 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2154 | return 0; |
8e96e3b7 | 2155 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2156 | return 0; |
1a48e2ac | 2157 | return !inode_capable(inode, CAP_FOWNER); |
1da177e4 LT |
2158 | } |
2159 | ||
2160 | /* | |
2161 | * Check whether we can remove a link victim from directory dir, check | |
2162 | * whether the type of victim is right. | |
2163 | * 1. We can't do it if dir is read-only (done in permission()) | |
2164 | * 2. We should have write and exec permissions on dir | |
2165 | * 3. We can't remove anything from append-only dir | |
2166 | * 4. We can't do anything with immutable dir (done in permission()) | |
2167 | * 5. If the sticky bit on dir is set we should either | |
2168 | * a. be owner of dir, or | |
2169 | * b. be owner of victim, or | |
2170 | * c. have CAP_FOWNER capability | |
2171 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2172 | * links pointing to it. | |
2173 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2174 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2175 | * 9. We can't remove a root or mountpoint. | |
2176 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
2177 | * nfs_async_unlink(). | |
2178 | */ | |
858119e1 | 2179 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
2180 | { |
2181 | int error; | |
2182 | ||
2183 | if (!victim->d_inode) | |
2184 | return -ENOENT; | |
2185 | ||
2186 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 2187 | audit_inode_child(victim, dir); |
1da177e4 | 2188 | |
f419a2e3 | 2189 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2190 | if (error) |
2191 | return error; | |
2192 | if (IS_APPEND(dir)) | |
2193 | return -EPERM; | |
2194 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 2195 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
2196 | return -EPERM; |
2197 | if (isdir) { | |
2198 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
2199 | return -ENOTDIR; | |
2200 | if (IS_ROOT(victim)) | |
2201 | return -EBUSY; | |
2202 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
2203 | return -EISDIR; | |
2204 | if (IS_DEADDIR(dir)) | |
2205 | return -ENOENT; | |
2206 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2207 | return -EBUSY; | |
2208 | return 0; | |
2209 | } | |
2210 | ||
2211 | /* Check whether we can create an object with dentry child in directory | |
2212 | * dir. | |
2213 | * 1. We can't do it if child already exists (open has special treatment for | |
2214 | * this case, but since we are inlined it's OK) | |
2215 | * 2. We can't do it if dir is read-only (done in permission()) | |
2216 | * 3. We should have write and exec permissions on dir | |
2217 | * 4. We can't do it if dir is immutable (done in permission()) | |
2218 | */ | |
a95164d9 | 2219 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
2220 | { |
2221 | if (child->d_inode) | |
2222 | return -EEXIST; | |
2223 | if (IS_DEADDIR(dir)) | |
2224 | return -ENOENT; | |
f419a2e3 | 2225 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2226 | } |
2227 | ||
1da177e4 LT |
2228 | /* |
2229 | * p1 and p2 should be directories on the same fs. | |
2230 | */ | |
2231 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2232 | { | |
2233 | struct dentry *p; | |
2234 | ||
2235 | if (p1 == p2) { | |
f2eace23 | 2236 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2237 | return NULL; |
2238 | } | |
2239 | ||
a11f3a05 | 2240 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2241 | |
e2761a11 OH |
2242 | p = d_ancestor(p2, p1); |
2243 | if (p) { | |
2244 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2245 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2246 | return p; | |
1da177e4 LT |
2247 | } |
2248 | ||
e2761a11 OH |
2249 | p = d_ancestor(p1, p2); |
2250 | if (p) { | |
2251 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2252 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2253 | return p; | |
1da177e4 LT |
2254 | } |
2255 | ||
f2eace23 IM |
2256 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
2257 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
2258 | return NULL; |
2259 | } | |
2260 | ||
2261 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2262 | { | |
1b1dcc1b | 2263 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2264 | if (p1 != p2) { |
1b1dcc1b | 2265 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2266 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2267 | } |
2268 | } | |
2269 | ||
4acdaf27 | 2270 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2271 | bool want_excl) |
1da177e4 | 2272 | { |
a95164d9 | 2273 | int error = may_create(dir, dentry); |
1da177e4 LT |
2274 | if (error) |
2275 | return error; | |
2276 | ||
acfa4380 | 2277 | if (!dir->i_op->create) |
1da177e4 LT |
2278 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2279 | mode &= S_IALLUGO; | |
2280 | mode |= S_IFREG; | |
2281 | error = security_inode_create(dir, dentry, mode); | |
2282 | if (error) | |
2283 | return error; | |
312b63fb | 2284 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2285 | if (!error) |
f38aa942 | 2286 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2287 | return error; |
2288 | } | |
2289 | ||
73d049a4 | 2290 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2291 | { |
3fb64190 | 2292 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2293 | struct inode *inode = dentry->d_inode; |
2294 | int error; | |
2295 | ||
bcda7652 AV |
2296 | /* O_PATH? */ |
2297 | if (!acc_mode) | |
2298 | return 0; | |
2299 | ||
1da177e4 LT |
2300 | if (!inode) |
2301 | return -ENOENT; | |
2302 | ||
c8fe8f30 CH |
2303 | switch (inode->i_mode & S_IFMT) { |
2304 | case S_IFLNK: | |
1da177e4 | 2305 | return -ELOOP; |
c8fe8f30 CH |
2306 | case S_IFDIR: |
2307 | if (acc_mode & MAY_WRITE) | |
2308 | return -EISDIR; | |
2309 | break; | |
2310 | case S_IFBLK: | |
2311 | case S_IFCHR: | |
3fb64190 | 2312 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2313 | return -EACCES; |
c8fe8f30 CH |
2314 | /*FALLTHRU*/ |
2315 | case S_IFIFO: | |
2316 | case S_IFSOCK: | |
1da177e4 | 2317 | flag &= ~O_TRUNC; |
c8fe8f30 | 2318 | break; |
4a3fd211 | 2319 | } |
b41572e9 | 2320 | |
3fb64190 | 2321 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2322 | if (error) |
2323 | return error; | |
6146f0d5 | 2324 | |
1da177e4 LT |
2325 | /* |
2326 | * An append-only file must be opened in append mode for writing. | |
2327 | */ | |
2328 | if (IS_APPEND(inode)) { | |
8737c930 | 2329 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2330 | return -EPERM; |
1da177e4 | 2331 | if (flag & O_TRUNC) |
7715b521 | 2332 | return -EPERM; |
1da177e4 LT |
2333 | } |
2334 | ||
2335 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2336 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2337 | return -EPERM; |
1da177e4 | 2338 | |
f3c7691e | 2339 | return 0; |
7715b521 | 2340 | } |
1da177e4 | 2341 | |
e1181ee6 | 2342 | static int handle_truncate(struct file *filp) |
7715b521 | 2343 | { |
e1181ee6 | 2344 | struct path *path = &filp->f_path; |
7715b521 AV |
2345 | struct inode *inode = path->dentry->d_inode; |
2346 | int error = get_write_access(inode); | |
2347 | if (error) | |
2348 | return error; | |
2349 | /* | |
2350 | * Refuse to truncate files with mandatory locks held on them. | |
2351 | */ | |
2352 | error = locks_verify_locked(inode); | |
2353 | if (!error) | |
ea0d3ab2 | 2354 | error = security_path_truncate(path); |
7715b521 AV |
2355 | if (!error) { |
2356 | error = do_truncate(path->dentry, 0, | |
2357 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2358 | filp); |
7715b521 AV |
2359 | } |
2360 | put_write_access(inode); | |
acd0c935 | 2361 | return error; |
1da177e4 LT |
2362 | } |
2363 | ||
d57999e1 DH |
2364 | static inline int open_to_namei_flags(int flag) |
2365 | { | |
8a5e929d AV |
2366 | if ((flag & O_ACCMODE) == 3) |
2367 | flag--; | |
d57999e1 DH |
2368 | return flag; |
2369 | } | |
2370 | ||
d18e9008 MS |
2371 | static int may_o_create(struct path *dir, struct dentry *dentry, umode_t mode) |
2372 | { | |
2373 | int error = security_path_mknod(dir, dentry, mode, 0); | |
2374 | if (error) | |
2375 | return error; | |
2376 | ||
2377 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); | |
2378 | if (error) | |
2379 | return error; | |
2380 | ||
2381 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2382 | } | |
2383 | ||
1acf0af9 DH |
2384 | /* |
2385 | * Attempt to atomically look up, create and open a file from a negative | |
2386 | * dentry. | |
2387 | * | |
2388 | * Returns 0 if successful. The file will have been created and attached to | |
2389 | * @file by the filesystem calling finish_open(). | |
2390 | * | |
2391 | * Returns 1 if the file was looked up only or didn't need creating. The | |
2392 | * caller will need to perform the open themselves. @path will have been | |
2393 | * updated to point to the new dentry. This may be negative. | |
2394 | * | |
2395 | * Returns an error code otherwise. | |
2396 | */ | |
2675a4eb AV |
2397 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
2398 | struct path *path, struct file *file, | |
2399 | const struct open_flags *op, | |
64894cf8 | 2400 | bool got_write, bool need_lookup, |
2675a4eb | 2401 | int *opened) |
d18e9008 MS |
2402 | { |
2403 | struct inode *dir = nd->path.dentry->d_inode; | |
2404 | unsigned open_flag = open_to_namei_flags(op->open_flag); | |
2405 | umode_t mode; | |
2406 | int error; | |
2407 | int acc_mode; | |
d18e9008 MS |
2408 | int create_error = 0; |
2409 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; | |
2410 | ||
2411 | BUG_ON(dentry->d_inode); | |
2412 | ||
2413 | /* Don't create child dentry for a dead directory. */ | |
2414 | if (unlikely(IS_DEADDIR(dir))) { | |
2675a4eb | 2415 | error = -ENOENT; |
d18e9008 MS |
2416 | goto out; |
2417 | } | |
2418 | ||
62b259d8 | 2419 | mode = op->mode; |
d18e9008 MS |
2420 | if ((open_flag & O_CREAT) && !IS_POSIXACL(dir)) |
2421 | mode &= ~current_umask(); | |
2422 | ||
f8310c59 | 2423 | if ((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT)) { |
d18e9008 | 2424 | open_flag &= ~O_TRUNC; |
47237687 | 2425 | *opened |= FILE_CREATED; |
d18e9008 MS |
2426 | } |
2427 | ||
2428 | /* | |
2429 | * Checking write permission is tricky, bacuse we don't know if we are | |
2430 | * going to actually need it: O_CREAT opens should work as long as the | |
2431 | * file exists. But checking existence breaks atomicity. The trick is | |
2432 | * to check access and if not granted clear O_CREAT from the flags. | |
2433 | * | |
2434 | * Another problem is returing the "right" error value (e.g. for an | |
2435 | * O_EXCL open we want to return EEXIST not EROFS). | |
2436 | */ | |
64894cf8 AV |
2437 | if (((open_flag & (O_CREAT | O_TRUNC)) || |
2438 | (open_flag & O_ACCMODE) != O_RDONLY) && unlikely(!got_write)) { | |
2439 | if (!(open_flag & O_CREAT)) { | |
d18e9008 MS |
2440 | /* |
2441 | * No O_CREATE -> atomicity not a requirement -> fall | |
2442 | * back to lookup + open | |
2443 | */ | |
2444 | goto no_open; | |
2445 | } else if (open_flag & (O_EXCL | O_TRUNC)) { | |
2446 | /* Fall back and fail with the right error */ | |
64894cf8 | 2447 | create_error = -EROFS; |
d18e9008 MS |
2448 | goto no_open; |
2449 | } else { | |
2450 | /* No side effects, safe to clear O_CREAT */ | |
64894cf8 | 2451 | create_error = -EROFS; |
d18e9008 MS |
2452 | open_flag &= ~O_CREAT; |
2453 | } | |
2454 | } | |
2455 | ||
2456 | if (open_flag & O_CREAT) { | |
38227f78 | 2457 | error = may_o_create(&nd->path, dentry, mode); |
d18e9008 MS |
2458 | if (error) { |
2459 | create_error = error; | |
2460 | if (open_flag & O_EXCL) | |
2461 | goto no_open; | |
2462 | open_flag &= ~O_CREAT; | |
2463 | } | |
2464 | } | |
2465 | ||
2466 | if (nd->flags & LOOKUP_DIRECTORY) | |
2467 | open_flag |= O_DIRECTORY; | |
2468 | ||
30d90494 AV |
2469 | file->f_path.dentry = DENTRY_NOT_SET; |
2470 | file->f_path.mnt = nd->path.mnt; | |
2471 | error = dir->i_op->atomic_open(dir, dentry, file, open_flag, mode, | |
47237687 | 2472 | opened); |
d9585277 | 2473 | if (error < 0) { |
d9585277 AV |
2474 | if (create_error && error == -ENOENT) |
2475 | error = create_error; | |
d18e9008 MS |
2476 | goto out; |
2477 | } | |
2478 | ||
2479 | acc_mode = op->acc_mode; | |
47237687 | 2480 | if (*opened & FILE_CREATED) { |
d18e9008 MS |
2481 | fsnotify_create(dir, dentry); |
2482 | acc_mode = MAY_OPEN; | |
2483 | } | |
2484 | ||
d9585277 | 2485 | if (error) { /* returned 1, that is */ |
30d90494 | 2486 | if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 2487 | error = -EIO; |
d18e9008 MS |
2488 | goto out; |
2489 | } | |
30d90494 | 2490 | if (file->f_path.dentry) { |
d18e9008 | 2491 | dput(dentry); |
30d90494 | 2492 | dentry = file->f_path.dentry; |
d18e9008 | 2493 | } |
62b2ce96 SW |
2494 | if (create_error && dentry->d_inode == NULL) { |
2495 | error = create_error; | |
2496 | goto out; | |
2497 | } | |
d18e9008 MS |
2498 | goto looked_up; |
2499 | } | |
2500 | ||
2501 | /* | |
2502 | * We didn't have the inode before the open, so check open permission | |
2503 | * here. | |
2504 | */ | |
2675a4eb AV |
2505 | error = may_open(&file->f_path, acc_mode, open_flag); |
2506 | if (error) | |
2507 | fput(file); | |
d18e9008 MS |
2508 | |
2509 | out: | |
2510 | dput(dentry); | |
2675a4eb | 2511 | return error; |
d18e9008 | 2512 | |
d18e9008 MS |
2513 | no_open: |
2514 | if (need_lookup) { | |
72bd866a | 2515 | dentry = lookup_real(dir, dentry, nd->flags); |
d18e9008 | 2516 | if (IS_ERR(dentry)) |
2675a4eb | 2517 | return PTR_ERR(dentry); |
d18e9008 MS |
2518 | |
2519 | if (create_error) { | |
2520 | int open_flag = op->open_flag; | |
2521 | ||
2675a4eb | 2522 | error = create_error; |
d18e9008 MS |
2523 | if ((open_flag & O_EXCL)) { |
2524 | if (!dentry->d_inode) | |
2525 | goto out; | |
2526 | } else if (!dentry->d_inode) { | |
2527 | goto out; | |
2528 | } else if ((open_flag & O_TRUNC) && | |
2529 | S_ISREG(dentry->d_inode->i_mode)) { | |
2530 | goto out; | |
2531 | } | |
2532 | /* will fail later, go on to get the right error */ | |
2533 | } | |
2534 | } | |
2535 | looked_up: | |
2536 | path->dentry = dentry; | |
2537 | path->mnt = nd->path.mnt; | |
2675a4eb | 2538 | return 1; |
d18e9008 MS |
2539 | } |
2540 | ||
d58ffd35 | 2541 | /* |
1acf0af9 | 2542 | * Look up and maybe create and open the last component. |
d58ffd35 MS |
2543 | * |
2544 | * Must be called with i_mutex held on parent. | |
2545 | * | |
1acf0af9 DH |
2546 | * Returns 0 if the file was successfully atomically created (if necessary) and |
2547 | * opened. In this case the file will be returned attached to @file. | |
2548 | * | |
2549 | * Returns 1 if the file was not completely opened at this time, though lookups | |
2550 | * and creations will have been performed and the dentry returned in @path will | |
2551 | * be positive upon return if O_CREAT was specified. If O_CREAT wasn't | |
2552 | * specified then a negative dentry may be returned. | |
2553 | * | |
2554 | * An error code is returned otherwise. | |
2555 | * | |
2556 | * FILE_CREATE will be set in @*opened if the dentry was created and will be | |
2557 | * cleared otherwise prior to returning. | |
d58ffd35 | 2558 | */ |
2675a4eb AV |
2559 | static int lookup_open(struct nameidata *nd, struct path *path, |
2560 | struct file *file, | |
2561 | const struct open_flags *op, | |
64894cf8 | 2562 | bool got_write, int *opened) |
d58ffd35 MS |
2563 | { |
2564 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 2565 | struct inode *dir_inode = dir->d_inode; |
d58ffd35 MS |
2566 | struct dentry *dentry; |
2567 | int error; | |
54ef4872 | 2568 | bool need_lookup; |
d58ffd35 | 2569 | |
47237687 | 2570 | *opened &= ~FILE_CREATED; |
201f956e | 2571 | dentry = lookup_dcache(&nd->last, dir, nd->flags, &need_lookup); |
d58ffd35 | 2572 | if (IS_ERR(dentry)) |
2675a4eb | 2573 | return PTR_ERR(dentry); |
d58ffd35 | 2574 | |
d18e9008 MS |
2575 | /* Cached positive dentry: will open in f_op->open */ |
2576 | if (!need_lookup && dentry->d_inode) | |
2577 | goto out_no_open; | |
2578 | ||
2579 | if ((nd->flags & LOOKUP_OPEN) && dir_inode->i_op->atomic_open) { | |
64894cf8 | 2580 | return atomic_open(nd, dentry, path, file, op, got_write, |
47237687 | 2581 | need_lookup, opened); |
d18e9008 MS |
2582 | } |
2583 | ||
54ef4872 MS |
2584 | if (need_lookup) { |
2585 | BUG_ON(dentry->d_inode); | |
2586 | ||
72bd866a | 2587 | dentry = lookup_real(dir_inode, dentry, nd->flags); |
54ef4872 | 2588 | if (IS_ERR(dentry)) |
2675a4eb | 2589 | return PTR_ERR(dentry); |
54ef4872 MS |
2590 | } |
2591 | ||
d58ffd35 MS |
2592 | /* Negative dentry, just create the file */ |
2593 | if (!dentry->d_inode && (op->open_flag & O_CREAT)) { | |
2594 | umode_t mode = op->mode; | |
2595 | if (!IS_POSIXACL(dir->d_inode)) | |
2596 | mode &= ~current_umask(); | |
2597 | /* | |
2598 | * This write is needed to ensure that a | |
2599 | * rw->ro transition does not occur between | |
2600 | * the time when the file is created and when | |
2601 | * a permanent write count is taken through | |
015c3bbc | 2602 | * the 'struct file' in finish_open(). |
d58ffd35 | 2603 | */ |
64894cf8 AV |
2604 | if (!got_write) { |
2605 | error = -EROFS; | |
d58ffd35 | 2606 | goto out_dput; |
64894cf8 | 2607 | } |
47237687 | 2608 | *opened |= FILE_CREATED; |
d58ffd35 MS |
2609 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2610 | if (error) | |
2611 | goto out_dput; | |
312b63fb AV |
2612 | error = vfs_create(dir->d_inode, dentry, mode, |
2613 | nd->flags & LOOKUP_EXCL); | |
d58ffd35 MS |
2614 | if (error) |
2615 | goto out_dput; | |
2616 | } | |
d18e9008 | 2617 | out_no_open: |
d58ffd35 MS |
2618 | path->dentry = dentry; |
2619 | path->mnt = nd->path.mnt; | |
2675a4eb | 2620 | return 1; |
d58ffd35 MS |
2621 | |
2622 | out_dput: | |
2623 | dput(dentry); | |
2675a4eb | 2624 | return error; |
d58ffd35 MS |
2625 | } |
2626 | ||
31e6b01f | 2627 | /* |
fe2d35ff | 2628 | * Handle the last step of open() |
31e6b01f | 2629 | */ |
2675a4eb AV |
2630 | static int do_last(struct nameidata *nd, struct path *path, |
2631 | struct file *file, const struct open_flags *op, | |
2632 | int *opened, const char *pathname) | |
fb1cc555 | 2633 | { |
a1e28038 | 2634 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 2635 | int open_flag = op->open_flag; |
77d660a8 | 2636 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 2637 | bool got_write = false; |
bcda7652 | 2638 | int acc_mode = op->acc_mode; |
a1eb3315 | 2639 | struct inode *inode; |
77d660a8 | 2640 | bool symlink_ok = false; |
16b1c1cd MS |
2641 | struct path save_parent = { .dentry = NULL, .mnt = NULL }; |
2642 | bool retried = false; | |
16c2cd71 | 2643 | int error; |
1f36f774 | 2644 | |
c3e380b0 AV |
2645 | nd->flags &= ~LOOKUP_PARENT; |
2646 | nd->flags |= op->intent; | |
2647 | ||
1f36f774 AV |
2648 | switch (nd->last_type) { |
2649 | case LAST_DOTDOT: | |
176306f5 | 2650 | case LAST_DOT: |
fe2d35ff AV |
2651 | error = handle_dots(nd, nd->last_type); |
2652 | if (error) | |
2675a4eb | 2653 | return error; |
1f36f774 | 2654 | /* fallthrough */ |
1f36f774 | 2655 | case LAST_ROOT: |
9f1fafee | 2656 | error = complete_walk(nd); |
16c2cd71 | 2657 | if (error) |
2675a4eb | 2658 | return error; |
fe2d35ff | 2659 | audit_inode(pathname, nd->path.dentry); |
ca344a89 | 2660 | if (open_flag & O_CREAT) { |
fe2d35ff | 2661 | error = -EISDIR; |
2675a4eb | 2662 | goto out; |
fe2d35ff | 2663 | } |
e83db167 | 2664 | goto finish_open; |
1f36f774 | 2665 | case LAST_BIND: |
9f1fafee | 2666 | error = complete_walk(nd); |
16c2cd71 | 2667 | if (error) |
2675a4eb | 2668 | return error; |
1f36f774 | 2669 | audit_inode(pathname, dir); |
e83db167 | 2670 | goto finish_open; |
1f36f774 | 2671 | } |
67ee3ad2 | 2672 | |
ca344a89 | 2673 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
2674 | if (nd->last.name[nd->last.len]) |
2675 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
bcda7652 | 2676 | if (open_flag & O_PATH && !(nd->flags & LOOKUP_FOLLOW)) |
77d660a8 | 2677 | symlink_ok = true; |
fe2d35ff | 2678 | /* we _can_ be in RCU mode here */ |
a1eb3315 | 2679 | error = lookup_fast(nd, &nd->last, path, &inode); |
71574865 MS |
2680 | if (likely(!error)) |
2681 | goto finish_lookup; | |
2682 | ||
2683 | if (error < 0) | |
2675a4eb | 2684 | goto out; |
71574865 MS |
2685 | |
2686 | BUG_ON(nd->inode != dir->d_inode); | |
b6183df7 MS |
2687 | } else { |
2688 | /* create side of things */ | |
2689 | /* | |
2690 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
2691 | * has been cleared when we got to the last component we are | |
2692 | * about to look up | |
2693 | */ | |
2694 | error = complete_walk(nd); | |
2695 | if (error) | |
2675a4eb | 2696 | return error; |
fe2d35ff | 2697 | |
b6183df7 MS |
2698 | audit_inode(pathname, dir); |
2699 | error = -EISDIR; | |
2700 | /* trailing slashes? */ | |
2701 | if (nd->last.name[nd->last.len]) | |
2675a4eb | 2702 | goto out; |
b6183df7 | 2703 | } |
a2c36b45 | 2704 | |
16b1c1cd | 2705 | retry_lookup: |
64894cf8 AV |
2706 | if (op->open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
2707 | error = mnt_want_write(nd->path.mnt); | |
2708 | if (!error) | |
2709 | got_write = true; | |
2710 | /* | |
2711 | * do _not_ fail yet - we might not need that or fail with | |
2712 | * a different error; let lookup_open() decide; we'll be | |
2713 | * dropping this one anyway. | |
2714 | */ | |
2715 | } | |
a1e28038 | 2716 | mutex_lock(&dir->d_inode->i_mutex); |
64894cf8 | 2717 | error = lookup_open(nd, path, file, op, got_write, opened); |
d58ffd35 | 2718 | mutex_unlock(&dir->d_inode->i_mutex); |
a1e28038 | 2719 | |
2675a4eb AV |
2720 | if (error <= 0) { |
2721 | if (error) | |
d18e9008 MS |
2722 | goto out; |
2723 | ||
47237687 | 2724 | if ((*opened & FILE_CREATED) || |
2675a4eb | 2725 | !S_ISREG(file->f_path.dentry->d_inode->i_mode)) |
77d660a8 | 2726 | will_truncate = false; |
d18e9008 | 2727 | |
2675a4eb | 2728 | audit_inode(pathname, file->f_path.dentry); |
d18e9008 MS |
2729 | goto opened; |
2730 | } | |
fb1cc555 | 2731 | |
47237687 | 2732 | if (*opened & FILE_CREATED) { |
9b44f1b3 | 2733 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 2734 | open_flag &= ~O_TRUNC; |
77d660a8 | 2735 | will_truncate = false; |
bcda7652 | 2736 | acc_mode = MAY_OPEN; |
d58ffd35 | 2737 | path_to_nameidata(path, nd); |
e83db167 | 2738 | goto finish_open_created; |
fb1cc555 AV |
2739 | } |
2740 | ||
2741 | /* | |
3134f37e | 2742 | * create/update audit record if it already exists. |
fb1cc555 | 2743 | */ |
3134f37e JL |
2744 | if (path->dentry->d_inode) |
2745 | audit_inode(pathname, path->dentry); | |
fb1cc555 | 2746 | |
d18e9008 MS |
2747 | /* |
2748 | * If atomic_open() acquired write access it is dropped now due to | |
2749 | * possible mount and symlink following (this might be optimized away if | |
2750 | * necessary...) | |
2751 | */ | |
64894cf8 | 2752 | if (got_write) { |
d18e9008 | 2753 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 2754 | got_write = false; |
d18e9008 MS |
2755 | } |
2756 | ||
fb1cc555 | 2757 | error = -EEXIST; |
f8310c59 | 2758 | if ((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT)) |
fb1cc555 AV |
2759 | goto exit_dput; |
2760 | ||
9875cf80 DH |
2761 | error = follow_managed(path, nd->flags); |
2762 | if (error < 0) | |
2763 | goto exit_dput; | |
fb1cc555 | 2764 | |
a3fbbde7 AV |
2765 | if (error) |
2766 | nd->flags |= LOOKUP_JUMPED; | |
2767 | ||
decf3400 MS |
2768 | BUG_ON(nd->flags & LOOKUP_RCU); |
2769 | inode = path->dentry->d_inode; | |
5f5daac1 MS |
2770 | finish_lookup: |
2771 | /* we _can_ be in RCU mode here */ | |
fb1cc555 | 2772 | error = -ENOENT; |
54c33e7f MS |
2773 | if (!inode) { |
2774 | path_to_nameidata(path, nd); | |
2675a4eb | 2775 | goto out; |
54c33e7f | 2776 | } |
9e67f361 | 2777 | |
d45ea867 MS |
2778 | if (should_follow_link(inode, !symlink_ok)) { |
2779 | if (nd->flags & LOOKUP_RCU) { | |
2780 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
2781 | error = -ECHILD; | |
2675a4eb | 2782 | goto out; |
d45ea867 MS |
2783 | } |
2784 | } | |
2785 | BUG_ON(inode != path->dentry->d_inode); | |
2675a4eb | 2786 | return 1; |
d45ea867 | 2787 | } |
fb1cc555 | 2788 | |
16b1c1cd MS |
2789 | if ((nd->flags & LOOKUP_RCU) || nd->path.mnt != path->mnt) { |
2790 | path_to_nameidata(path, nd); | |
2791 | } else { | |
2792 | save_parent.dentry = nd->path.dentry; | |
2793 | save_parent.mnt = mntget(path->mnt); | |
2794 | nd->path.dentry = path->dentry; | |
2795 | ||
2796 | } | |
decf3400 | 2797 | nd->inode = inode; |
a3fbbde7 AV |
2798 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
2799 | error = complete_walk(nd); | |
16b1c1cd MS |
2800 | if (error) { |
2801 | path_put(&save_parent); | |
2675a4eb | 2802 | return error; |
16b1c1cd | 2803 | } |
fb1cc555 | 2804 | error = -EISDIR; |
050ac841 | 2805 | if ((open_flag & O_CREAT) && S_ISDIR(nd->inode->i_mode)) |
2675a4eb | 2806 | goto out; |
af2f5542 MS |
2807 | error = -ENOTDIR; |
2808 | if ((nd->flags & LOOKUP_DIRECTORY) && !nd->inode->i_op->lookup) | |
2675a4eb | 2809 | goto out; |
d7fdd7f6 | 2810 | audit_inode(pathname, nd->path.dentry); |
e83db167 | 2811 | finish_open: |
6c0d46c4 | 2812 | if (!S_ISREG(nd->inode->i_mode)) |
77d660a8 | 2813 | will_truncate = false; |
6c0d46c4 | 2814 | |
0f9d1a10 AV |
2815 | if (will_truncate) { |
2816 | error = mnt_want_write(nd->path.mnt); | |
2817 | if (error) | |
2675a4eb | 2818 | goto out; |
64894cf8 | 2819 | got_write = true; |
0f9d1a10 | 2820 | } |
e83db167 | 2821 | finish_open_created: |
bcda7652 | 2822 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 2823 | if (error) |
2675a4eb | 2824 | goto out; |
30d90494 AV |
2825 | file->f_path.mnt = nd->path.mnt; |
2826 | error = finish_open(file, nd->path.dentry, NULL, opened); | |
2827 | if (error) { | |
30d90494 | 2828 | if (error == -EOPENSTALE) |
f60dc3db | 2829 | goto stale_open; |
015c3bbc | 2830 | goto out; |
f60dc3db | 2831 | } |
a8277b9b | 2832 | opened: |
2675a4eb | 2833 | error = open_check_o_direct(file); |
015c3bbc MS |
2834 | if (error) |
2835 | goto exit_fput; | |
2675a4eb | 2836 | error = ima_file_check(file, op->acc_mode); |
aa4caadb MS |
2837 | if (error) |
2838 | goto exit_fput; | |
2839 | ||
2840 | if (will_truncate) { | |
2675a4eb | 2841 | error = handle_truncate(file); |
aa4caadb MS |
2842 | if (error) |
2843 | goto exit_fput; | |
0f9d1a10 | 2844 | } |
ca344a89 | 2845 | out: |
64894cf8 | 2846 | if (got_write) |
0f9d1a10 | 2847 | mnt_drop_write(nd->path.mnt); |
16b1c1cd | 2848 | path_put(&save_parent); |
e276ae67 | 2849 | terminate_walk(nd); |
2675a4eb | 2850 | return error; |
fb1cc555 | 2851 | |
fb1cc555 AV |
2852 | exit_dput: |
2853 | path_put_conditional(path, nd); | |
ca344a89 | 2854 | goto out; |
015c3bbc | 2855 | exit_fput: |
2675a4eb AV |
2856 | fput(file); |
2857 | goto out; | |
015c3bbc | 2858 | |
f60dc3db MS |
2859 | stale_open: |
2860 | /* If no saved parent or already retried then can't retry */ | |
2861 | if (!save_parent.dentry || retried) | |
2862 | goto out; | |
2863 | ||
2864 | BUG_ON(save_parent.dentry != dir); | |
2865 | path_put(&nd->path); | |
2866 | nd->path = save_parent; | |
2867 | nd->inode = dir->d_inode; | |
2868 | save_parent.mnt = NULL; | |
2869 | save_parent.dentry = NULL; | |
64894cf8 | 2870 | if (got_write) { |
f60dc3db | 2871 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 2872 | got_write = false; |
f60dc3db MS |
2873 | } |
2874 | retried = true; | |
2875 | goto retry_lookup; | |
fb1cc555 AV |
2876 | } |
2877 | ||
13aab428 | 2878 | static struct file *path_openat(int dfd, const char *pathname, |
73d049a4 | 2879 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 2880 | { |
fe2d35ff | 2881 | struct file *base = NULL; |
30d90494 | 2882 | struct file *file; |
9850c056 | 2883 | struct path path; |
47237687 | 2884 | int opened = 0; |
13aab428 | 2885 | int error; |
31e6b01f | 2886 | |
30d90494 AV |
2887 | file = get_empty_filp(); |
2888 | if (!file) | |
31e6b01f NP |
2889 | return ERR_PTR(-ENFILE); |
2890 | ||
30d90494 | 2891 | file->f_flags = op->open_flag; |
31e6b01f | 2892 | |
73d049a4 | 2893 | error = path_init(dfd, pathname, flags | LOOKUP_PARENT, nd, &base); |
31e6b01f | 2894 | if (unlikely(error)) |
2675a4eb | 2895 | goto out; |
31e6b01f | 2896 | |
fe2d35ff | 2897 | current->total_link_count = 0; |
73d049a4 | 2898 | error = link_path_walk(pathname, nd); |
31e6b01f | 2899 | if (unlikely(error)) |
2675a4eb | 2900 | goto out; |
1da177e4 | 2901 | |
2675a4eb AV |
2902 | error = do_last(nd, &path, file, op, &opened, pathname); |
2903 | while (unlikely(error > 0)) { /* trailing symlink */ | |
7b9337aa | 2904 | struct path link = path; |
def4af30 | 2905 | void *cookie; |
574197e0 | 2906 | if (!(nd->flags & LOOKUP_FOLLOW)) { |
73d049a4 AV |
2907 | path_put_conditional(&path, nd); |
2908 | path_put(&nd->path); | |
2675a4eb | 2909 | error = -ELOOP; |
40b39136 AV |
2910 | break; |
2911 | } | |
800179c9 KC |
2912 | error = may_follow_link(&link, nd); |
2913 | if (unlikely(error)) | |
2914 | break; | |
73d049a4 AV |
2915 | nd->flags |= LOOKUP_PARENT; |
2916 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
574197e0 | 2917 | error = follow_link(&link, nd, &cookie); |
c3e380b0 | 2918 | if (unlikely(error)) |
2675a4eb AV |
2919 | break; |
2920 | error = do_last(nd, &path, file, op, &opened, pathname); | |
574197e0 | 2921 | put_link(nd, &link, cookie); |
806b681c | 2922 | } |
10fa8e62 | 2923 | out: |
73d049a4 AV |
2924 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) |
2925 | path_put(&nd->root); | |
fe2d35ff AV |
2926 | if (base) |
2927 | fput(base); | |
2675a4eb AV |
2928 | if (!(opened & FILE_OPENED)) { |
2929 | BUG_ON(!error); | |
30d90494 | 2930 | put_filp(file); |
16b1c1cd | 2931 | } |
2675a4eb AV |
2932 | if (unlikely(error)) { |
2933 | if (error == -EOPENSTALE) { | |
2934 | if (flags & LOOKUP_RCU) | |
2935 | error = -ECHILD; | |
2936 | else | |
2937 | error = -ESTALE; | |
2938 | } | |
2939 | file = ERR_PTR(error); | |
2940 | } | |
2941 | return file; | |
1da177e4 LT |
2942 | } |
2943 | ||
13aab428 AV |
2944 | struct file *do_filp_open(int dfd, const char *pathname, |
2945 | const struct open_flags *op, int flags) | |
2946 | { | |
73d049a4 | 2947 | struct nameidata nd; |
13aab428 AV |
2948 | struct file *filp; |
2949 | ||
73d049a4 | 2950 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 2951 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 2952 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 2953 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 2954 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
2955 | return filp; |
2956 | } | |
2957 | ||
73d049a4 AV |
2958 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
2959 | const char *name, const struct open_flags *op, int flags) | |
2960 | { | |
2961 | struct nameidata nd; | |
2962 | struct file *file; | |
2963 | ||
2964 | nd.root.mnt = mnt; | |
2965 | nd.root.dentry = dentry; | |
2966 | ||
2967 | flags |= LOOKUP_ROOT; | |
2968 | ||
bcda7652 | 2969 | if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
2970 | return ERR_PTR(-ELOOP); |
2971 | ||
2972 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU); | |
2973 | if (unlikely(file == ERR_PTR(-ECHILD))) | |
2974 | file = path_openat(-1, name, &nd, op, flags); | |
2975 | if (unlikely(file == ERR_PTR(-ESTALE))) | |
2976 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL); | |
2977 | return file; | |
2978 | } | |
2979 | ||
ed75e95d | 2980 | struct dentry *kern_path_create(int dfd, const char *pathname, struct path *path, int is_dir) |
1da177e4 | 2981 | { |
c663e5d8 | 2982 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d | 2983 | struct nameidata nd; |
c30dabfe | 2984 | int err2; |
ed75e95d AV |
2985 | int error = do_path_lookup(dfd, pathname, LOOKUP_PARENT, &nd); |
2986 | if (error) | |
2987 | return ERR_PTR(error); | |
1da177e4 | 2988 | |
c663e5d8 CH |
2989 | /* |
2990 | * Yucky last component or no last component at all? | |
2991 | * (foo/., foo/.., /////) | |
2992 | */ | |
ed75e95d AV |
2993 | if (nd.last_type != LAST_NORM) |
2994 | goto out; | |
2995 | nd.flags &= ~LOOKUP_PARENT; | |
2996 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
c663e5d8 | 2997 | |
c30dabfe JK |
2998 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
2999 | err2 = mnt_want_write(nd.path.mnt); | |
c663e5d8 CH |
3000 | /* |
3001 | * Do the final lookup. | |
3002 | */ | |
ed75e95d AV |
3003 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
3004 | dentry = lookup_hash(&nd); | |
1da177e4 | 3005 | if (IS_ERR(dentry)) |
a8104a9f | 3006 | goto unlock; |
c663e5d8 | 3007 | |
a8104a9f | 3008 | error = -EEXIST; |
e9baf6e5 | 3009 | if (dentry->d_inode) |
a8104a9f | 3010 | goto fail; |
c663e5d8 CH |
3011 | /* |
3012 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3013 | * From the vfs_mknod() POV we just have a negative dentry - | |
3014 | * all is fine. Let's be bastards - you had / on the end, you've | |
3015 | * been asking for (non-existent) directory. -ENOENT for you. | |
3016 | */ | |
ed75e95d | 3017 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
a8104a9f | 3018 | error = -ENOENT; |
ed75e95d | 3019 | goto fail; |
e9baf6e5 | 3020 | } |
c30dabfe JK |
3021 | if (unlikely(err2)) { |
3022 | error = err2; | |
a8104a9f | 3023 | goto fail; |
c30dabfe | 3024 | } |
ed75e95d | 3025 | *path = nd.path; |
1da177e4 | 3026 | return dentry; |
1da177e4 | 3027 | fail: |
a8104a9f AV |
3028 | dput(dentry); |
3029 | dentry = ERR_PTR(error); | |
3030 | unlock: | |
ed75e95d | 3031 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
c30dabfe JK |
3032 | if (!err2) |
3033 | mnt_drop_write(nd.path.mnt); | |
ed75e95d AV |
3034 | out: |
3035 | path_put(&nd.path); | |
1da177e4 LT |
3036 | return dentry; |
3037 | } | |
dae6ad8f AV |
3038 | EXPORT_SYMBOL(kern_path_create); |
3039 | ||
921a1650 AV |
3040 | void done_path_create(struct path *path, struct dentry *dentry) |
3041 | { | |
3042 | dput(dentry); | |
3043 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
a8104a9f | 3044 | mnt_drop_write(path->mnt); |
921a1650 AV |
3045 | path_put(path); |
3046 | } | |
3047 | EXPORT_SYMBOL(done_path_create); | |
3048 | ||
dae6ad8f AV |
3049 | struct dentry *user_path_create(int dfd, const char __user *pathname, struct path *path, int is_dir) |
3050 | { | |
3051 | char *tmp = getname(pathname); | |
3052 | struct dentry *res; | |
3053 | if (IS_ERR(tmp)) | |
3054 | return ERR_CAST(tmp); | |
3055 | res = kern_path_create(dfd, tmp, path, is_dir); | |
3056 | putname(tmp); | |
3057 | return res; | |
3058 | } | |
3059 | EXPORT_SYMBOL(user_path_create); | |
3060 | ||
1a67aafb | 3061 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3062 | { |
a95164d9 | 3063 | int error = may_create(dir, dentry); |
1da177e4 LT |
3064 | |
3065 | if (error) | |
3066 | return error; | |
3067 | ||
975d6b39 | 3068 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3069 | return -EPERM; |
3070 | ||
acfa4380 | 3071 | if (!dir->i_op->mknod) |
1da177e4 LT |
3072 | return -EPERM; |
3073 | ||
08ce5f16 SH |
3074 | error = devcgroup_inode_mknod(mode, dev); |
3075 | if (error) | |
3076 | return error; | |
3077 | ||
1da177e4 LT |
3078 | error = security_inode_mknod(dir, dentry, mode, dev); |
3079 | if (error) | |
3080 | return error; | |
3081 | ||
1da177e4 | 3082 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3083 | if (!error) |
f38aa942 | 3084 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3085 | return error; |
3086 | } | |
3087 | ||
f69aac00 | 3088 | static int may_mknod(umode_t mode) |
463c3197 DH |
3089 | { |
3090 | switch (mode & S_IFMT) { | |
3091 | case S_IFREG: | |
3092 | case S_IFCHR: | |
3093 | case S_IFBLK: | |
3094 | case S_IFIFO: | |
3095 | case S_IFSOCK: | |
3096 | case 0: /* zero mode translates to S_IFREG */ | |
3097 | return 0; | |
3098 | case S_IFDIR: | |
3099 | return -EPERM; | |
3100 | default: | |
3101 | return -EINVAL; | |
3102 | } | |
3103 | } | |
3104 | ||
8208a22b | 3105 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 3106 | unsigned, dev) |
1da177e4 | 3107 | { |
2ad94ae6 | 3108 | struct dentry *dentry; |
dae6ad8f AV |
3109 | struct path path; |
3110 | int error; | |
1da177e4 | 3111 | |
8e4bfca1 AV |
3112 | error = may_mknod(mode); |
3113 | if (error) | |
3114 | return error; | |
1da177e4 | 3115 | |
dae6ad8f AV |
3116 | dentry = user_path_create(dfd, filename, &path, 0); |
3117 | if (IS_ERR(dentry)) | |
3118 | return PTR_ERR(dentry); | |
2ad94ae6 | 3119 | |
dae6ad8f | 3120 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3121 | mode &= ~current_umask(); |
dae6ad8f | 3122 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3123 | if (error) |
a8104a9f | 3124 | goto out; |
463c3197 | 3125 | switch (mode & S_IFMT) { |
1da177e4 | 3126 | case 0: case S_IFREG: |
312b63fb | 3127 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
1da177e4 LT |
3128 | break; |
3129 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3130 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3131 | new_decode_dev(dev)); |
3132 | break; | |
3133 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3134 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3135 | break; |
1da177e4 | 3136 | } |
a8104a9f | 3137 | out: |
921a1650 | 3138 | done_path_create(&path, dentry); |
1da177e4 LT |
3139 | return error; |
3140 | } | |
3141 | ||
8208a22b | 3142 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
3143 | { |
3144 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
3145 | } | |
3146 | ||
18bb1db3 | 3147 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3148 | { |
a95164d9 | 3149 | int error = may_create(dir, dentry); |
8de52778 | 3150 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3151 | |
3152 | if (error) | |
3153 | return error; | |
3154 | ||
acfa4380 | 3155 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3156 | return -EPERM; |
3157 | ||
3158 | mode &= (S_IRWXUGO|S_ISVTX); | |
3159 | error = security_inode_mkdir(dir, dentry, mode); | |
3160 | if (error) | |
3161 | return error; | |
3162 | ||
8de52778 AV |
3163 | if (max_links && dir->i_nlink >= max_links) |
3164 | return -EMLINK; | |
3165 | ||
1da177e4 | 3166 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3167 | if (!error) |
f38aa942 | 3168 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3169 | return error; |
3170 | } | |
3171 | ||
a218d0fd | 3172 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 3173 | { |
6902d925 | 3174 | struct dentry *dentry; |
dae6ad8f AV |
3175 | struct path path; |
3176 | int error; | |
1da177e4 | 3177 | |
dae6ad8f | 3178 | dentry = user_path_create(dfd, pathname, &path, 1); |
6902d925 | 3179 | if (IS_ERR(dentry)) |
dae6ad8f | 3180 | return PTR_ERR(dentry); |
1da177e4 | 3181 | |
dae6ad8f | 3182 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3183 | mode &= ~current_umask(); |
dae6ad8f | 3184 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3185 | if (!error) |
3186 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3187 | done_path_create(&path, dentry); |
1da177e4 LT |
3188 | return error; |
3189 | } | |
3190 | ||
a218d0fd | 3191 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
3192 | { |
3193 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
3194 | } | |
3195 | ||
1da177e4 | 3196 | /* |
a71905f0 | 3197 | * The dentry_unhash() helper will try to drop the dentry early: we |
c0d02594 | 3198 | * should have a usage count of 1 if we're the only user of this |
a71905f0 SW |
3199 | * dentry, and if that is true (possibly after pruning the dcache), |
3200 | * then we drop the dentry now. | |
1da177e4 LT |
3201 | * |
3202 | * A low-level filesystem can, if it choses, legally | |
3203 | * do a | |
3204 | * | |
3205 | * if (!d_unhashed(dentry)) | |
3206 | * return -EBUSY; | |
3207 | * | |
3208 | * if it cannot handle the case of removing a directory | |
3209 | * that is still in use by something else.. | |
3210 | */ | |
3211 | void dentry_unhash(struct dentry *dentry) | |
3212 | { | |
dc168427 | 3213 | shrink_dcache_parent(dentry); |
1da177e4 | 3214 | spin_lock(&dentry->d_lock); |
64252c75 | 3215 | if (dentry->d_count == 1) |
1da177e4 LT |
3216 | __d_drop(dentry); |
3217 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
3218 | } |
3219 | ||
3220 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
3221 | { | |
3222 | int error = may_delete(dir, dentry, 1); | |
3223 | ||
3224 | if (error) | |
3225 | return error; | |
3226 | ||
acfa4380 | 3227 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3228 | return -EPERM; |
3229 | ||
1d2ef590 | 3230 | dget(dentry); |
1b1dcc1b | 3231 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
3232 | |
3233 | error = -EBUSY; | |
1da177e4 | 3234 | if (d_mountpoint(dentry)) |
912dbc15 SW |
3235 | goto out; |
3236 | ||
3237 | error = security_inode_rmdir(dir, dentry); | |
3238 | if (error) | |
3239 | goto out; | |
3240 | ||
3cebde24 | 3241 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3242 | error = dir->i_op->rmdir(dir, dentry); |
3243 | if (error) | |
3244 | goto out; | |
3245 | ||
3246 | dentry->d_inode->i_flags |= S_DEAD; | |
3247 | dont_mount(dentry); | |
3248 | ||
3249 | out: | |
1b1dcc1b | 3250 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 3251 | dput(dentry); |
912dbc15 | 3252 | if (!error) |
1da177e4 | 3253 | d_delete(dentry); |
1da177e4 LT |
3254 | return error; |
3255 | } | |
3256 | ||
5590ff0d | 3257 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3258 | { |
3259 | int error = 0; | |
3260 | char * name; | |
3261 | struct dentry *dentry; | |
3262 | struct nameidata nd; | |
3263 | ||
2ad94ae6 | 3264 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 3265 | if (error) |
2ad94ae6 | 3266 | return error; |
1da177e4 LT |
3267 | |
3268 | switch(nd.last_type) { | |
0612d9fb OH |
3269 | case LAST_DOTDOT: |
3270 | error = -ENOTEMPTY; | |
3271 | goto exit1; | |
3272 | case LAST_DOT: | |
3273 | error = -EINVAL; | |
3274 | goto exit1; | |
3275 | case LAST_ROOT: | |
3276 | error = -EBUSY; | |
3277 | goto exit1; | |
1da177e4 | 3278 | } |
0612d9fb OH |
3279 | |
3280 | nd.flags &= ~LOOKUP_PARENT; | |
c30dabfe JK |
3281 | error = mnt_want_write(nd.path.mnt); |
3282 | if (error) | |
3283 | goto exit1; | |
0612d9fb | 3284 | |
4ac91378 | 3285 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 3286 | dentry = lookup_hash(&nd); |
1da177e4 | 3287 | error = PTR_ERR(dentry); |
6902d925 DH |
3288 | if (IS_ERR(dentry)) |
3289 | goto exit2; | |
e6bc45d6 TT |
3290 | if (!dentry->d_inode) { |
3291 | error = -ENOENT; | |
3292 | goto exit3; | |
3293 | } | |
be6d3e56 KT |
3294 | error = security_path_rmdir(&nd.path, dentry); |
3295 | if (error) | |
c30dabfe | 3296 | goto exit3; |
4ac91378 | 3297 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
0622753b | 3298 | exit3: |
6902d925 DH |
3299 | dput(dentry); |
3300 | exit2: | |
4ac91378 | 3301 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
c30dabfe | 3302 | mnt_drop_write(nd.path.mnt); |
1da177e4 | 3303 | exit1: |
1d957f9b | 3304 | path_put(&nd.path); |
1da177e4 LT |
3305 | putname(name); |
3306 | return error; | |
3307 | } | |
3308 | ||
3cdad428 | 3309 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3310 | { |
3311 | return do_rmdir(AT_FDCWD, pathname); | |
3312 | } | |
3313 | ||
1da177e4 LT |
3314 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
3315 | { | |
3316 | int error = may_delete(dir, dentry, 0); | |
3317 | ||
3318 | if (error) | |
3319 | return error; | |
3320 | ||
acfa4380 | 3321 | if (!dir->i_op->unlink) |
1da177e4 LT |
3322 | return -EPERM; |
3323 | ||
1b1dcc1b | 3324 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
3325 | if (d_mountpoint(dentry)) |
3326 | error = -EBUSY; | |
3327 | else { | |
3328 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3329 | if (!error) { |
1da177e4 | 3330 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 3331 | if (!error) |
d83c49f3 | 3332 | dont_mount(dentry); |
bec1052e | 3333 | } |
1da177e4 | 3334 | } |
1b1dcc1b | 3335 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
3336 | |
3337 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3338 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 3339 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 3340 | d_delete(dentry); |
1da177e4 | 3341 | } |
0eeca283 | 3342 | |
1da177e4 LT |
3343 | return error; |
3344 | } | |
3345 | ||
3346 | /* | |
3347 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3348 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3349 | * writeout happening, and we don't want to prevent access to the directory |
3350 | * while waiting on the I/O. | |
3351 | */ | |
5590ff0d | 3352 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 3353 | { |
2ad94ae6 AV |
3354 | int error; |
3355 | char *name; | |
1da177e4 LT |
3356 | struct dentry *dentry; |
3357 | struct nameidata nd; | |
3358 | struct inode *inode = NULL; | |
3359 | ||
2ad94ae6 | 3360 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 3361 | if (error) |
2ad94ae6 AV |
3362 | return error; |
3363 | ||
1da177e4 LT |
3364 | error = -EISDIR; |
3365 | if (nd.last_type != LAST_NORM) | |
3366 | goto exit1; | |
0612d9fb OH |
3367 | |
3368 | nd.flags &= ~LOOKUP_PARENT; | |
c30dabfe JK |
3369 | error = mnt_want_write(nd.path.mnt); |
3370 | if (error) | |
3371 | goto exit1; | |
0612d9fb | 3372 | |
4ac91378 | 3373 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 3374 | dentry = lookup_hash(&nd); |
1da177e4 LT |
3375 | error = PTR_ERR(dentry); |
3376 | if (!IS_ERR(dentry)) { | |
3377 | /* Why not before? Because we want correct error value */ | |
50338b88 TE |
3378 | if (nd.last.name[nd.last.len]) |
3379 | goto slashes; | |
1da177e4 | 3380 | inode = dentry->d_inode; |
50338b88 | 3381 | if (!inode) |
e6bc45d6 TT |
3382 | goto slashes; |
3383 | ihold(inode); | |
be6d3e56 KT |
3384 | error = security_path_unlink(&nd.path, dentry); |
3385 | if (error) | |
c30dabfe | 3386 | goto exit2; |
4ac91378 | 3387 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
c30dabfe | 3388 | exit2: |
1da177e4 LT |
3389 | dput(dentry); |
3390 | } | |
4ac91378 | 3391 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
3392 | if (inode) |
3393 | iput(inode); /* truncate the inode here */ | |
c30dabfe | 3394 | mnt_drop_write(nd.path.mnt); |
1da177e4 | 3395 | exit1: |
1d957f9b | 3396 | path_put(&nd.path); |
1da177e4 LT |
3397 | putname(name); |
3398 | return error; | |
3399 | ||
3400 | slashes: | |
3401 | error = !dentry->d_inode ? -ENOENT : | |
3402 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
3403 | goto exit2; | |
3404 | } | |
3405 | ||
2e4d0924 | 3406 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3407 | { |
3408 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3409 | return -EINVAL; | |
3410 | ||
3411 | if (flag & AT_REMOVEDIR) | |
3412 | return do_rmdir(dfd, pathname); | |
3413 | ||
3414 | return do_unlinkat(dfd, pathname); | |
3415 | } | |
3416 | ||
3480b257 | 3417 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
3418 | { |
3419 | return do_unlinkat(AT_FDCWD, pathname); | |
3420 | } | |
3421 | ||
db2e747b | 3422 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3423 | { |
a95164d9 | 3424 | int error = may_create(dir, dentry); |
1da177e4 LT |
3425 | |
3426 | if (error) | |
3427 | return error; | |
3428 | ||
acfa4380 | 3429 | if (!dir->i_op->symlink) |
1da177e4 LT |
3430 | return -EPERM; |
3431 | ||
3432 | error = security_inode_symlink(dir, dentry, oldname); | |
3433 | if (error) | |
3434 | return error; | |
3435 | ||
1da177e4 | 3436 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3437 | if (!error) |
f38aa942 | 3438 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3439 | return error; |
3440 | } | |
3441 | ||
2e4d0924 HC |
3442 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
3443 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3444 | { |
2ad94ae6 AV |
3445 | int error; |
3446 | char *from; | |
6902d925 | 3447 | struct dentry *dentry; |
dae6ad8f | 3448 | struct path path; |
1da177e4 LT |
3449 | |
3450 | from = getname(oldname); | |
2ad94ae6 | 3451 | if (IS_ERR(from)) |
1da177e4 | 3452 | return PTR_ERR(from); |
1da177e4 | 3453 | |
dae6ad8f | 3454 | dentry = user_path_create(newdfd, newname, &path, 0); |
6902d925 DH |
3455 | error = PTR_ERR(dentry); |
3456 | if (IS_ERR(dentry)) | |
dae6ad8f | 3457 | goto out_putname; |
6902d925 | 3458 | |
dae6ad8f | 3459 | error = security_path_symlink(&path, dentry, from); |
a8104a9f AV |
3460 | if (!error) |
3461 | error = vfs_symlink(path.dentry->d_inode, dentry, from); | |
921a1650 | 3462 | done_path_create(&path, dentry); |
6902d925 | 3463 | out_putname: |
1da177e4 LT |
3464 | putname(from); |
3465 | return error; | |
3466 | } | |
3467 | ||
3480b257 | 3468 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3469 | { |
3470 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
3471 | } | |
3472 | ||
1da177e4 LT |
3473 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
3474 | { | |
3475 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 3476 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3477 | int error; |
3478 | ||
3479 | if (!inode) | |
3480 | return -ENOENT; | |
3481 | ||
a95164d9 | 3482 | error = may_create(dir, new_dentry); |
1da177e4 LT |
3483 | if (error) |
3484 | return error; | |
3485 | ||
3486 | if (dir->i_sb != inode->i_sb) | |
3487 | return -EXDEV; | |
3488 | ||
3489 | /* | |
3490 | * A link to an append-only or immutable file cannot be created. | |
3491 | */ | |
3492 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
3493 | return -EPERM; | |
acfa4380 | 3494 | if (!dir->i_op->link) |
1da177e4 | 3495 | return -EPERM; |
7e79eedb | 3496 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
3497 | return -EPERM; |
3498 | ||
3499 | error = security_inode_link(old_dentry, dir, new_dentry); | |
3500 | if (error) | |
3501 | return error; | |
3502 | ||
7e79eedb | 3503 | mutex_lock(&inode->i_mutex); |
aae8a97d AK |
3504 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
3505 | if (inode->i_nlink == 0) | |
3506 | error = -ENOENT; | |
8de52778 AV |
3507 | else if (max_links && inode->i_nlink >= max_links) |
3508 | error = -EMLINK; | |
aae8a97d AK |
3509 | else |
3510 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 3511 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 3512 | if (!error) |
7e79eedb | 3513 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
3514 | return error; |
3515 | } | |
3516 | ||
3517 | /* | |
3518 | * Hardlinks are often used in delicate situations. We avoid | |
3519 | * security-related surprises by not following symlinks on the | |
3520 | * newname. --KAB | |
3521 | * | |
3522 | * We don't follow them on the oldname either to be compatible | |
3523 | * with linux 2.0, and to avoid hard-linking to directories | |
3524 | * and other special files. --ADM | |
3525 | */ | |
2e4d0924 HC |
3526 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
3527 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
3528 | { |
3529 | struct dentry *new_dentry; | |
dae6ad8f | 3530 | struct path old_path, new_path; |
11a7b371 | 3531 | int how = 0; |
1da177e4 | 3532 | int error; |
1da177e4 | 3533 | |
11a7b371 | 3534 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 3535 | return -EINVAL; |
11a7b371 AK |
3536 | /* |
3537 | * To use null names we require CAP_DAC_READ_SEARCH | |
3538 | * This ensures that not everyone will be able to create | |
3539 | * handlink using the passed filedescriptor. | |
3540 | */ | |
3541 | if (flags & AT_EMPTY_PATH) { | |
3542 | if (!capable(CAP_DAC_READ_SEARCH)) | |
3543 | return -ENOENT; | |
3544 | how = LOOKUP_EMPTY; | |
3545 | } | |
3546 | ||
3547 | if (flags & AT_SYMLINK_FOLLOW) | |
3548 | how |= LOOKUP_FOLLOW; | |
c04030e1 | 3549 | |
11a7b371 | 3550 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 3551 | if (error) |
2ad94ae6 AV |
3552 | return error; |
3553 | ||
dae6ad8f | 3554 | new_dentry = user_path_create(newdfd, newname, &new_path, 0); |
1da177e4 | 3555 | error = PTR_ERR(new_dentry); |
6902d925 | 3556 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
3557 | goto out; |
3558 | ||
3559 | error = -EXDEV; | |
3560 | if (old_path.mnt != new_path.mnt) | |
3561 | goto out_dput; | |
800179c9 KC |
3562 | error = may_linkat(&old_path); |
3563 | if (unlikely(error)) | |
3564 | goto out_dput; | |
dae6ad8f | 3565 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 3566 | if (error) |
a8104a9f | 3567 | goto out_dput; |
dae6ad8f | 3568 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry); |
75c3f29d | 3569 | out_dput: |
921a1650 | 3570 | done_path_create(&new_path, new_dentry); |
1da177e4 | 3571 | out: |
2d8f3038 | 3572 | path_put(&old_path); |
1da177e4 LT |
3573 | |
3574 | return error; | |
3575 | } | |
3576 | ||
3480b257 | 3577 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 3578 | { |
c04030e1 | 3579 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
3580 | } |
3581 | ||
1da177e4 LT |
3582 | /* |
3583 | * The worst of all namespace operations - renaming directory. "Perverted" | |
3584 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
3585 | * Problems: | |
3586 | * a) we can get into loop creation. Check is done in is_subdir(). | |
3587 | * b) race potential - two innocent renames can create a loop together. | |
3588 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 3589 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
3590 | * story. |
3591 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 3592 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
3593 | * whether the target exists). Solution: try to be smart with locking |
3594 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 3595 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
3596 | * move will be locked. Thus we can rank directories by the tree |
3597 | * (ancestors first) and rank all non-directories after them. | |
3598 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 3599 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
3600 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
3601 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
3602 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 3603 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 3604 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 3605 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 3606 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
3607 | * locking]. |
3608 | */ | |
75c96f85 AB |
3609 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
3610 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3611 | { |
3612 | int error = 0; | |
9055cba7 | 3613 | struct inode *target = new_dentry->d_inode; |
8de52778 | 3614 | unsigned max_links = new_dir->i_sb->s_max_links; |
1da177e4 LT |
3615 | |
3616 | /* | |
3617 | * If we are going to change the parent - check write permissions, | |
3618 | * we'll need to flip '..'. | |
3619 | */ | |
3620 | if (new_dir != old_dir) { | |
f419a2e3 | 3621 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3622 | if (error) |
3623 | return error; | |
3624 | } | |
3625 | ||
3626 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3627 | if (error) | |
3628 | return error; | |
3629 | ||
1d2ef590 | 3630 | dget(new_dentry); |
d83c49f3 | 3631 | if (target) |
1b1dcc1b | 3632 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
3633 | |
3634 | error = -EBUSY; | |
3635 | if (d_mountpoint(old_dentry) || d_mountpoint(new_dentry)) | |
3636 | goto out; | |
3637 | ||
8de52778 AV |
3638 | error = -EMLINK; |
3639 | if (max_links && !target && new_dir != old_dir && | |
3640 | new_dir->i_nlink >= max_links) | |
3641 | goto out; | |
3642 | ||
3cebde24 SW |
3643 | if (target) |
3644 | shrink_dcache_parent(new_dentry); | |
9055cba7 SW |
3645 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
3646 | if (error) | |
3647 | goto out; | |
3648 | ||
1da177e4 | 3649 | if (target) { |
9055cba7 SW |
3650 | target->i_flags |= S_DEAD; |
3651 | dont_mount(new_dentry); | |
1da177e4 | 3652 | } |
9055cba7 SW |
3653 | out: |
3654 | if (target) | |
3655 | mutex_unlock(&target->i_mutex); | |
1d2ef590 | 3656 | dput(new_dentry); |
e31e14ec | 3657 | if (!error) |
349457cc MF |
3658 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3659 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3660 | return error; |
3661 | } | |
3662 | ||
75c96f85 AB |
3663 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3664 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 | 3665 | { |
51892bbb | 3666 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3667 | int error; |
3668 | ||
3669 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3670 | if (error) | |
3671 | return error; | |
3672 | ||
3673 | dget(new_dentry); | |
1da177e4 | 3674 | if (target) |
1b1dcc1b | 3675 | mutex_lock(&target->i_mutex); |
51892bbb SW |
3676 | |
3677 | error = -EBUSY; | |
1da177e4 | 3678 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
51892bbb SW |
3679 | goto out; |
3680 | ||
3681 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3682 | if (error) | |
3683 | goto out; | |
3684 | ||
3685 | if (target) | |
3686 | dont_mount(new_dentry); | |
3687 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) | |
3688 | d_move(old_dentry, new_dentry); | |
3689 | out: | |
1da177e4 | 3690 | if (target) |
1b1dcc1b | 3691 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3692 | dput(new_dentry); |
3693 | return error; | |
3694 | } | |
3695 | ||
3696 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3697 | struct inode *new_dir, struct dentry *new_dentry) | |
3698 | { | |
3699 | int error; | |
3700 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3701 | const unsigned char *old_name; |
1da177e4 LT |
3702 | |
3703 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3704 | return 0; | |
3705 | ||
3706 | error = may_delete(old_dir, old_dentry, is_dir); | |
3707 | if (error) | |
3708 | return error; | |
3709 | ||
3710 | if (!new_dentry->d_inode) | |
a95164d9 | 3711 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3712 | else |
3713 | error = may_delete(new_dir, new_dentry, is_dir); | |
3714 | if (error) | |
3715 | return error; | |
3716 | ||
acfa4380 | 3717 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3718 | return -EPERM; |
3719 | ||
0eeca283 RL |
3720 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3721 | ||
1da177e4 LT |
3722 | if (is_dir) |
3723 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3724 | else | |
3725 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3726 | if (!error) |
3727 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3728 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3729 | fsnotify_oldname_free(old_name); |
3730 | ||
1da177e4 LT |
3731 | return error; |
3732 | } | |
3733 | ||
2e4d0924 HC |
3734 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3735 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3736 | { |
2ad94ae6 AV |
3737 | struct dentry *old_dir, *new_dir; |
3738 | struct dentry *old_dentry, *new_dentry; | |
3739 | struct dentry *trap; | |
1da177e4 | 3740 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3741 | char *from; |
3742 | char *to; | |
3743 | int error; | |
1da177e4 | 3744 | |
2ad94ae6 | 3745 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3746 | if (error) |
3747 | goto exit; | |
3748 | ||
2ad94ae6 | 3749 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3750 | if (error) |
3751 | goto exit1; | |
3752 | ||
3753 | error = -EXDEV; | |
4ac91378 | 3754 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3755 | goto exit2; |
3756 | ||
4ac91378 | 3757 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3758 | error = -EBUSY; |
3759 | if (oldnd.last_type != LAST_NORM) | |
3760 | goto exit2; | |
3761 | ||
4ac91378 | 3762 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3763 | if (newnd.last_type != LAST_NORM) |
3764 | goto exit2; | |
3765 | ||
c30dabfe JK |
3766 | error = mnt_want_write(oldnd.path.mnt); |
3767 | if (error) | |
3768 | goto exit2; | |
3769 | ||
0612d9fb OH |
3770 | oldnd.flags &= ~LOOKUP_PARENT; |
3771 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3772 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3773 | |
1da177e4 LT |
3774 | trap = lock_rename(new_dir, old_dir); |
3775 | ||
49705b77 | 3776 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3777 | error = PTR_ERR(old_dentry); |
3778 | if (IS_ERR(old_dentry)) | |
3779 | goto exit3; | |
3780 | /* source must exist */ | |
3781 | error = -ENOENT; | |
3782 | if (!old_dentry->d_inode) | |
3783 | goto exit4; | |
3784 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3785 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3786 | error = -ENOTDIR; | |
3787 | if (oldnd.last.name[oldnd.last.len]) | |
3788 | goto exit4; | |
3789 | if (newnd.last.name[newnd.last.len]) | |
3790 | goto exit4; | |
3791 | } | |
3792 | /* source should not be ancestor of target */ | |
3793 | error = -EINVAL; | |
3794 | if (old_dentry == trap) | |
3795 | goto exit4; | |
49705b77 | 3796 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3797 | error = PTR_ERR(new_dentry); |
3798 | if (IS_ERR(new_dentry)) | |
3799 | goto exit4; | |
3800 | /* target should not be an ancestor of source */ | |
3801 | error = -ENOTEMPTY; | |
3802 | if (new_dentry == trap) | |
3803 | goto exit5; | |
3804 | ||
be6d3e56 KT |
3805 | error = security_path_rename(&oldnd.path, old_dentry, |
3806 | &newnd.path, new_dentry); | |
3807 | if (error) | |
c30dabfe | 3808 | goto exit5; |
1da177e4 LT |
3809 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3810 | new_dir->d_inode, new_dentry); | |
3811 | exit5: | |
3812 | dput(new_dentry); | |
3813 | exit4: | |
3814 | dput(old_dentry); | |
3815 | exit3: | |
3816 | unlock_rename(new_dir, old_dir); | |
c30dabfe | 3817 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 | 3818 | exit2: |
1d957f9b | 3819 | path_put(&newnd.path); |
2ad94ae6 | 3820 | putname(to); |
1da177e4 | 3821 | exit1: |
1d957f9b | 3822 | path_put(&oldnd.path); |
1da177e4 | 3823 | putname(from); |
2ad94ae6 | 3824 | exit: |
1da177e4 LT |
3825 | return error; |
3826 | } | |
3827 | ||
a26eab24 | 3828 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3829 | { |
3830 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3831 | } | |
3832 | ||
1da177e4 LT |
3833 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3834 | { | |
3835 | int len; | |
3836 | ||
3837 | len = PTR_ERR(link); | |
3838 | if (IS_ERR(link)) | |
3839 | goto out; | |
3840 | ||
3841 | len = strlen(link); | |
3842 | if (len > (unsigned) buflen) | |
3843 | len = buflen; | |
3844 | if (copy_to_user(buffer, link, len)) | |
3845 | len = -EFAULT; | |
3846 | out: | |
3847 | return len; | |
3848 | } | |
3849 | ||
3850 | /* | |
3851 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3852 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3853 | * using) it for any given inode is up to filesystem. | |
3854 | */ | |
3855 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3856 | { | |
3857 | struct nameidata nd; | |
cc314eef | 3858 | void *cookie; |
694a1764 | 3859 | int res; |
cc314eef | 3860 | |
1da177e4 | 3861 | nd.depth = 0; |
cc314eef | 3862 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3863 | if (IS_ERR(cookie)) |
3864 | return PTR_ERR(cookie); | |
3865 | ||
3866 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3867 | if (dentry->d_inode->i_op->put_link) | |
3868 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3869 | return res; | |
1da177e4 LT |
3870 | } |
3871 | ||
3872 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3873 | { | |
3874 | return __vfs_follow_link(nd, link); | |
3875 | } | |
3876 | ||
3877 | /* get the link contents into pagecache */ | |
3878 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3879 | { | |
ebd09abb DG |
3880 | char *kaddr; |
3881 | struct page *page; | |
1da177e4 | 3882 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3883 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3884 | if (IS_ERR(page)) |
6fe6900e | 3885 | return (char*)page; |
1da177e4 | 3886 | *ppage = page; |
ebd09abb DG |
3887 | kaddr = kmap(page); |
3888 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3889 | return kaddr; | |
1da177e4 LT |
3890 | } |
3891 | ||
3892 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3893 | { | |
3894 | struct page *page = NULL; | |
3895 | char *s = page_getlink(dentry, &page); | |
3896 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3897 | if (page) { | |
3898 | kunmap(page); | |
3899 | page_cache_release(page); | |
3900 | } | |
3901 | return res; | |
3902 | } | |
3903 | ||
cc314eef | 3904 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3905 | { |
cc314eef | 3906 | struct page *page = NULL; |
1da177e4 | 3907 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3908 | return page; |
1da177e4 LT |
3909 | } |
3910 | ||
cc314eef | 3911 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3912 | { |
cc314eef LT |
3913 | struct page *page = cookie; |
3914 | ||
3915 | if (page) { | |
1da177e4 LT |
3916 | kunmap(page); |
3917 | page_cache_release(page); | |
1da177e4 LT |
3918 | } |
3919 | } | |
3920 | ||
54566b2c NP |
3921 | /* |
3922 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3923 | */ | |
3924 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3925 | { |
3926 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3927 | struct page *page; |
afddba49 | 3928 | void *fsdata; |
beb497ab | 3929 | int err; |
1da177e4 | 3930 | char *kaddr; |
54566b2c NP |
3931 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3932 | if (nofs) | |
3933 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3934 | |
7e53cac4 | 3935 | retry: |
afddba49 | 3936 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3937 | flags, &page, &fsdata); |
1da177e4 | 3938 | if (err) |
afddba49 NP |
3939 | goto fail; |
3940 | ||
e8e3c3d6 | 3941 | kaddr = kmap_atomic(page); |
1da177e4 | 3942 | memcpy(kaddr, symname, len-1); |
e8e3c3d6 | 3943 | kunmap_atomic(kaddr); |
afddba49 NP |
3944 | |
3945 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3946 | page, fsdata); | |
1da177e4 LT |
3947 | if (err < 0) |
3948 | goto fail; | |
afddba49 NP |
3949 | if (err < len-1) |
3950 | goto retry; | |
3951 | ||
1da177e4 LT |
3952 | mark_inode_dirty(inode); |
3953 | return 0; | |
1da177e4 LT |
3954 | fail: |
3955 | return err; | |
3956 | } | |
3957 | ||
0adb25d2 KK |
3958 | int page_symlink(struct inode *inode, const char *symname, int len) |
3959 | { | |
3960 | return __page_symlink(inode, symname, len, | |
54566b2c | 3961 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3962 | } |
3963 | ||
92e1d5be | 3964 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3965 | .readlink = generic_readlink, |
3966 | .follow_link = page_follow_link_light, | |
3967 | .put_link = page_put_link, | |
3968 | }; | |
3969 | ||
2d8f3038 | 3970 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3971 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3972 | EXPORT_SYMBOL(follow_down); |
3973 | EXPORT_SYMBOL(follow_up); | |
3974 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
3975 | EXPORT_SYMBOL(getname); | |
3976 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3977 | EXPORT_SYMBOL(lookup_one_len); |
3978 | EXPORT_SYMBOL(page_follow_link_light); | |
3979 | EXPORT_SYMBOL(page_put_link); | |
3980 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3981 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3982 | EXPORT_SYMBOL(page_symlink); |
3983 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
d1811465 | 3984 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3985 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3986 | EXPORT_SYMBOL(inode_permission); |
1da177e4 LT |
3987 | EXPORT_SYMBOL(unlock_rename); |
3988 | EXPORT_SYMBOL(vfs_create); | |
3989 | EXPORT_SYMBOL(vfs_follow_link); | |
3990 | EXPORT_SYMBOL(vfs_link); | |
3991 | EXPORT_SYMBOL(vfs_mkdir); | |
3992 | EXPORT_SYMBOL(vfs_mknod); | |
3993 | EXPORT_SYMBOL(generic_permission); | |
3994 | EXPORT_SYMBOL(vfs_readlink); | |
3995 | EXPORT_SYMBOL(vfs_rename); | |
3996 | EXPORT_SYMBOL(vfs_rmdir); | |
3997 | EXPORT_SYMBOL(vfs_symlink); | |
3998 | EXPORT_SYMBOL(vfs_unlink); | |
3999 | EXPORT_SYMBOL(dentry_unhash); | |
4000 | EXPORT_SYMBOL(generic_readlink); |