]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/namei.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * Some corrections by tytso. | |
9 | */ | |
10 | ||
11 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
12 | * lookup logic. | |
13 | */ | |
14 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
15 | */ | |
16 | ||
17 | #include <linux/init.h> | |
18 | #include <linux/module.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/fs.h> | |
21 | #include <linux/namei.h> | |
1da177e4 | 22 | #include <linux/pagemap.h> |
0eeca283 | 23 | #include <linux/fsnotify.h> |
1da177e4 LT |
24 | #include <linux/personality.h> |
25 | #include <linux/security.h> | |
6146f0d5 | 26 | #include <linux/ima.h> |
1da177e4 LT |
27 | #include <linux/syscalls.h> |
28 | #include <linux/mount.h> | |
29 | #include <linux/audit.h> | |
16f7e0fe | 30 | #include <linux/capability.h> |
834f2a4a | 31 | #include <linux/file.h> |
5590ff0d | 32 | #include <linux/fcntl.h> |
08ce5f16 | 33 | #include <linux/device_cgroup.h> |
5ad4e53b | 34 | #include <linux/fs_struct.h> |
1da177e4 LT |
35 | #include <asm/uaccess.h> |
36 | ||
e81e3f4d EP |
37 | #include "internal.h" |
38 | ||
1da177e4 LT |
39 | /* [Feb-1997 T. Schoebel-Theuer] |
40 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
41 | * were necessary because of omirr. The reason is that omirr needs | |
42 | * to know the _real_ pathname, not the user-supplied one, in case | |
43 | * of symlinks (and also when transname replacements occur). | |
44 | * | |
45 | * The new code replaces the old recursive symlink resolution with | |
46 | * an iterative one (in case of non-nested symlink chains). It does | |
47 | * this with calls to <fs>_follow_link(). | |
48 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
49 | * replaced with a single function lookup_dentry() that can handle all | |
50 | * the special cases of the former code. | |
51 | * | |
52 | * With the new dcache, the pathname is stored at each inode, at least as | |
53 | * long as the refcount of the inode is positive. As a side effect, the | |
54 | * size of the dcache depends on the inode cache and thus is dynamic. | |
55 | * | |
56 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
57 | * resolution to correspond with current state of the code. | |
58 | * | |
59 | * Note that the symlink resolution is not *completely* iterative. | |
60 | * There is still a significant amount of tail- and mid- recursion in | |
61 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
62 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
63 | * may return different results than <fs>_follow_link(). Many virtual | |
64 | * filesystems (including /proc) exhibit this behavior. | |
65 | */ | |
66 | ||
67 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
68 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
69 | * and the name already exists in form of a symlink, try to create the new | |
70 | * name indicated by the symlink. The old code always complained that the | |
71 | * name already exists, due to not following the symlink even if its target | |
72 | * is nonexistent. The new semantics affects also mknod() and link() when | |
73 | * the name is a symlink pointing to a non-existant name. | |
74 | * | |
75 | * I don't know which semantics is the right one, since I have no access | |
76 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
77 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
78 | * "old" one. Personally, I think the new semantics is much more logical. | |
79 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
80 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
81 | * and in the old Linux semantics. | |
82 | */ | |
83 | ||
84 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
85 | * semantics. See the comments in "open_namei" and "do_link" below. | |
86 | * | |
87 | * [10-Sep-98 Alan Modra] Another symlink change. | |
88 | */ | |
89 | ||
90 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
91 | * inside the path - always follow. | |
92 | * in the last component in creation/removal/renaming - never follow. | |
93 | * if LOOKUP_FOLLOW passed - follow. | |
94 | * if the pathname has trailing slashes - follow. | |
95 | * otherwise - don't follow. | |
96 | * (applied in that order). | |
97 | * | |
98 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
99 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
100 | * During the 2.4 we need to fix the userland stuff depending on it - | |
101 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
102 | * XEmacs seems to be relying on it... | |
103 | */ | |
104 | /* | |
105 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 106 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
107 | * any extra contention... |
108 | */ | |
109 | ||
110 | /* In order to reduce some races, while at the same time doing additional | |
111 | * checking and hopefully speeding things up, we copy filenames to the | |
112 | * kernel data space before using them.. | |
113 | * | |
114 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
115 | * PATH_MAX includes the nul terminator --RR. | |
116 | */ | |
858119e1 | 117 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
118 | { |
119 | int retval; | |
120 | unsigned long len = PATH_MAX; | |
121 | ||
122 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
123 | if ((unsigned long) filename >= TASK_SIZE) | |
124 | return -EFAULT; | |
125 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
126 | len = TASK_SIZE - (unsigned long) filename; | |
127 | } | |
128 | ||
129 | retval = strncpy_from_user(page, filename, len); | |
130 | if (retval > 0) { | |
131 | if (retval < len) | |
132 | return 0; | |
133 | return -ENAMETOOLONG; | |
134 | } else if (!retval) | |
135 | retval = -ENOENT; | |
136 | return retval; | |
137 | } | |
138 | ||
139 | char * getname(const char __user * filename) | |
140 | { | |
141 | char *tmp, *result; | |
142 | ||
143 | result = ERR_PTR(-ENOMEM); | |
144 | tmp = __getname(); | |
145 | if (tmp) { | |
146 | int retval = do_getname(filename, tmp); | |
147 | ||
148 | result = tmp; | |
149 | if (retval < 0) { | |
150 | __putname(tmp); | |
151 | result = ERR_PTR(retval); | |
152 | } | |
153 | } | |
154 | audit_getname(result); | |
155 | return result; | |
156 | } | |
157 | ||
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 | ||
5909ccaa LT |
169 | /* |
170 | * This does basic POSIX ACL permission checking | |
1da177e4 | 171 | */ |
b74c79e9 NP |
172 | static int acl_permission_check(struct inode *inode, int mask, unsigned int flags, |
173 | int (*check_acl)(struct inode *inode, int mask, unsigned int flags)) | |
1da177e4 LT |
174 | { |
175 | umode_t mode = inode->i_mode; | |
176 | ||
e6305c43 AV |
177 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
178 | ||
da9592ed | 179 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
180 | mode >>= 6; |
181 | else { | |
182 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
b74c79e9 NP |
183 | int error = check_acl(inode, mask, flags); |
184 | if (error != -EAGAIN) | |
185 | return error; | |
1da177e4 LT |
186 | } |
187 | ||
188 | if (in_group_p(inode->i_gid)) | |
189 | mode >>= 3; | |
190 | } | |
191 | ||
192 | /* | |
193 | * If the DACs are ok we don't need any capability check. | |
194 | */ | |
e6305c43 | 195 | if ((mask & ~mode) == 0) |
1da177e4 | 196 | return 0; |
5909ccaa LT |
197 | return -EACCES; |
198 | } | |
199 | ||
200 | /** | |
b74c79e9 | 201 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa LT |
202 | * @inode: inode to check access rights for |
203 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
204 | * @check_acl: optional callback to check for Posix ACLs | |
39191628 | 205 | * @flags: IPERM_FLAG_ flags. |
5909ccaa LT |
206 | * |
207 | * Used to check for read/write/execute permissions on a file. | |
208 | * We use "fsuid" for this, letting us set arbitrary permissions | |
209 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
210 | * are used for other things. |
211 | * | |
212 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
213 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
214 | * It would then be called again in ref-walk mode. | |
5909ccaa | 215 | */ |
b74c79e9 NP |
216 | int generic_permission(struct inode *inode, int mask, unsigned int flags, |
217 | int (*check_acl)(struct inode *inode, int mask, unsigned int flags)) | |
5909ccaa LT |
218 | { |
219 | int ret; | |
220 | ||
221 | /* | |
222 | * Do the basic POSIX ACL permission checks. | |
223 | */ | |
b74c79e9 | 224 | ret = acl_permission_check(inode, mask, flags, check_acl); |
5909ccaa LT |
225 | if (ret != -EACCES) |
226 | return ret; | |
1da177e4 | 227 | |
1da177e4 LT |
228 | /* |
229 | * Read/write DACs are always overridable. | |
230 | * Executable DACs are overridable if at least one exec bit is set. | |
231 | */ | |
f696a365 | 232 | if (!(mask & MAY_EXEC) || execute_ok(inode)) |
1da177e4 LT |
233 | if (capable(CAP_DAC_OVERRIDE)) |
234 | return 0; | |
235 | ||
236 | /* | |
237 | * Searching includes executable on directories, else just read. | |
238 | */ | |
7ea66001 | 239 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
1da177e4 LT |
240 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) |
241 | if (capable(CAP_DAC_READ_SEARCH)) | |
242 | return 0; | |
243 | ||
244 | return -EACCES; | |
245 | } | |
246 | ||
cb23beb5 CH |
247 | /** |
248 | * inode_permission - check for access rights to a given inode | |
249 | * @inode: inode to check permission on | |
250 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
251 | * | |
252 | * Used to check for read/write/execute permissions on an inode. | |
253 | * We use "fsuid" for this, letting us set arbitrary permissions | |
254 | * for filesystem access without changing the "normal" uids which | |
255 | * are used for other things. | |
256 | */ | |
f419a2e3 | 257 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 258 | { |
e6305c43 | 259 | int retval; |
1da177e4 LT |
260 | |
261 | if (mask & MAY_WRITE) { | |
22590e41 | 262 | umode_t mode = inode->i_mode; |
1da177e4 LT |
263 | |
264 | /* | |
265 | * Nobody gets write access to a read-only fs. | |
266 | */ | |
267 | if (IS_RDONLY(inode) && | |
268 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
269 | return -EROFS; | |
270 | ||
271 | /* | |
272 | * Nobody gets write access to an immutable file. | |
273 | */ | |
274 | if (IS_IMMUTABLE(inode)) | |
275 | return -EACCES; | |
276 | } | |
277 | ||
acfa4380 | 278 | if (inode->i_op->permission) |
b74c79e9 | 279 | retval = inode->i_op->permission(inode, mask, 0); |
f696a365 | 280 | else |
b74c79e9 NP |
281 | retval = generic_permission(inode, mask, 0, |
282 | inode->i_op->check_acl); | |
f696a365 | 283 | |
1da177e4 LT |
284 | if (retval) |
285 | return retval; | |
286 | ||
08ce5f16 SH |
287 | retval = devcgroup_inode_permission(inode, mask); |
288 | if (retval) | |
289 | return retval; | |
290 | ||
d09ca739 | 291 | return security_inode_permission(inode, mask); |
1da177e4 LT |
292 | } |
293 | ||
8c744fb8 CH |
294 | /** |
295 | * file_permission - check for additional access rights to a given file | |
296 | * @file: file to check access rights for | |
297 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
298 | * | |
299 | * Used to check for read/write/execute permissions on an already opened | |
300 | * file. | |
301 | * | |
302 | * Note: | |
303 | * Do not use this function in new code. All access checks should | |
cb23beb5 | 304 | * be done using inode_permission(). |
8c744fb8 CH |
305 | */ |
306 | int file_permission(struct file *file, int mask) | |
307 | { | |
f419a2e3 | 308 | return inode_permission(file->f_path.dentry->d_inode, mask); |
8c744fb8 CH |
309 | } |
310 | ||
1da177e4 LT |
311 | /* |
312 | * get_write_access() gets write permission for a file. | |
313 | * put_write_access() releases this write permission. | |
314 | * This is used for regular files. | |
315 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
316 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
317 | * can have the following values: | |
318 | * 0: no writers, no VM_DENYWRITE mappings | |
319 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
320 | * > 0: (i_writecount) users are writing to the file. | |
321 | * | |
322 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
323 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
324 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
325 | * to do the change if sign is wrong. Exclusion between them is provided by | |
326 | * the inode->i_lock spinlock. | |
327 | */ | |
328 | ||
329 | int get_write_access(struct inode * inode) | |
330 | { | |
331 | spin_lock(&inode->i_lock); | |
332 | if (atomic_read(&inode->i_writecount) < 0) { | |
333 | spin_unlock(&inode->i_lock); | |
334 | return -ETXTBSY; | |
335 | } | |
336 | atomic_inc(&inode->i_writecount); | |
337 | spin_unlock(&inode->i_lock); | |
338 | ||
339 | return 0; | |
340 | } | |
341 | ||
342 | int deny_write_access(struct file * file) | |
343 | { | |
0f7fc9e4 | 344 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
345 | |
346 | spin_lock(&inode->i_lock); | |
347 | if (atomic_read(&inode->i_writecount) > 0) { | |
348 | spin_unlock(&inode->i_lock); | |
349 | return -ETXTBSY; | |
350 | } | |
351 | atomic_dec(&inode->i_writecount); | |
352 | spin_unlock(&inode->i_lock); | |
353 | ||
354 | return 0; | |
355 | } | |
356 | ||
5dd784d0 JB |
357 | /** |
358 | * path_get - get a reference to a path | |
359 | * @path: path to get the reference to | |
360 | * | |
361 | * Given a path increment the reference count to the dentry and the vfsmount. | |
362 | */ | |
363 | void path_get(struct path *path) | |
364 | { | |
365 | mntget(path->mnt); | |
366 | dget(path->dentry); | |
367 | } | |
368 | EXPORT_SYMBOL(path_get); | |
369 | ||
1d957f9b JB |
370 | /** |
371 | * path_put - put a reference to a path | |
372 | * @path: path to put the reference to | |
373 | * | |
374 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
375 | */ | |
376 | void path_put(struct path *path) | |
1da177e4 | 377 | { |
1d957f9b JB |
378 | dput(path->dentry); |
379 | mntput(path->mnt); | |
1da177e4 | 380 | } |
1d957f9b | 381 | EXPORT_SYMBOL(path_put); |
1da177e4 | 382 | |
31e6b01f NP |
383 | /** |
384 | * nameidata_drop_rcu - drop this nameidata out of rcu-walk | |
385 | * @nd: nameidata pathwalk data to drop | |
39191628 | 386 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f NP |
387 | * |
388 | * Path walking has 2 modes, rcu-walk and ref-walk (see | |
389 | * Documentation/filesystems/path-lookup.txt). __drop_rcu* functions attempt | |
390 | * to drop out of rcu-walk mode and take normal reference counts on dentries | |
391 | * and vfsmounts to transition to rcu-walk mode. __drop_rcu* functions take | |
392 | * refcounts at the last known good point before rcu-walk got stuck, so | |
393 | * ref-walk may continue from there. If this is not successful (eg. a seqcount | |
394 | * has changed), then failure is returned and path walk restarts from the | |
395 | * beginning in ref-walk mode. | |
396 | * | |
397 | * nameidata_drop_rcu attempts to drop the current nd->path and nd->root into | |
398 | * ref-walk. Must be called from rcu-walk context. | |
399 | */ | |
400 | static int nameidata_drop_rcu(struct nameidata *nd) | |
401 | { | |
402 | struct fs_struct *fs = current->fs; | |
403 | struct dentry *dentry = nd->path.dentry; | |
404 | ||
405 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
406 | if (nd->root.mnt) { | |
407 | spin_lock(&fs->lock); | |
408 | if (nd->root.mnt != fs->root.mnt || | |
409 | nd->root.dentry != fs->root.dentry) | |
410 | goto err_root; | |
411 | } | |
412 | spin_lock(&dentry->d_lock); | |
413 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
414 | goto err; | |
415 | BUG_ON(nd->inode != dentry->d_inode); | |
416 | spin_unlock(&dentry->d_lock); | |
417 | if (nd->root.mnt) { | |
418 | path_get(&nd->root); | |
419 | spin_unlock(&fs->lock); | |
420 | } | |
421 | mntget(nd->path.mnt); | |
422 | ||
423 | rcu_read_unlock(); | |
424 | br_read_unlock(vfsmount_lock); | |
425 | nd->flags &= ~LOOKUP_RCU; | |
426 | return 0; | |
427 | err: | |
428 | spin_unlock(&dentry->d_lock); | |
429 | err_root: | |
430 | if (nd->root.mnt) | |
431 | spin_unlock(&fs->lock); | |
432 | return -ECHILD; | |
433 | } | |
434 | ||
435 | /* Try to drop out of rcu-walk mode if we were in it, otherwise do nothing. */ | |
436 | static inline int nameidata_drop_rcu_maybe(struct nameidata *nd) | |
437 | { | |
438 | if (nd->flags & LOOKUP_RCU) | |
439 | return nameidata_drop_rcu(nd); | |
440 | return 0; | |
441 | } | |
442 | ||
443 | /** | |
444 | * nameidata_dentry_drop_rcu - drop nameidata and dentry out of rcu-walk | |
445 | * @nd: nameidata pathwalk data to drop | |
446 | * @dentry: dentry to drop | |
39191628 | 447 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f NP |
448 | * |
449 | * nameidata_dentry_drop_rcu attempts to drop the current nd->path and nd->root, | |
450 | * and dentry into ref-walk. @dentry must be a path found by a do_lookup call on | |
451 | * @nd. Must be called from rcu-walk context. | |
452 | */ | |
453 | static int nameidata_dentry_drop_rcu(struct nameidata *nd, struct dentry *dentry) | |
454 | { | |
455 | struct fs_struct *fs = current->fs; | |
456 | struct dentry *parent = nd->path.dentry; | |
457 | ||
90dbb77b NP |
458 | /* |
459 | * It can be possible to revalidate the dentry that we started | |
460 | * the path walk with. force_reval_path may also revalidate the | |
461 | * dentry already committed to the nameidata. | |
462 | */ | |
463 | if (unlikely(parent == dentry)) | |
464 | return nameidata_drop_rcu(nd); | |
465 | ||
31e6b01f NP |
466 | BUG_ON(!(nd->flags & LOOKUP_RCU)); |
467 | if (nd->root.mnt) { | |
468 | spin_lock(&fs->lock); | |
469 | if (nd->root.mnt != fs->root.mnt || | |
470 | nd->root.dentry != fs->root.dentry) | |
471 | goto err_root; | |
472 | } | |
473 | spin_lock(&parent->d_lock); | |
474 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); | |
475 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
476 | goto err; | |
477 | /* | |
478 | * If the sequence check on the child dentry passed, then the child has | |
479 | * not been removed from its parent. This means the parent dentry must | |
480 | * be valid and able to take a reference at this point. | |
481 | */ | |
482 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
483 | BUG_ON(!parent->d_count); | |
484 | parent->d_count++; | |
485 | spin_unlock(&dentry->d_lock); | |
486 | spin_unlock(&parent->d_lock); | |
487 | if (nd->root.mnt) { | |
488 | path_get(&nd->root); | |
489 | spin_unlock(&fs->lock); | |
490 | } | |
491 | mntget(nd->path.mnt); | |
492 | ||
493 | rcu_read_unlock(); | |
494 | br_read_unlock(vfsmount_lock); | |
495 | nd->flags &= ~LOOKUP_RCU; | |
496 | return 0; | |
497 | err: | |
498 | spin_unlock(&dentry->d_lock); | |
499 | spin_unlock(&parent->d_lock); | |
500 | err_root: | |
501 | if (nd->root.mnt) | |
502 | spin_unlock(&fs->lock); | |
503 | return -ECHILD; | |
504 | } | |
505 | ||
506 | /* Try to drop out of rcu-walk mode if we were in it, otherwise do nothing. */ | |
507 | static inline int nameidata_dentry_drop_rcu_maybe(struct nameidata *nd, struct dentry *dentry) | |
508 | { | |
509 | if (nd->flags & LOOKUP_RCU) | |
510 | return nameidata_dentry_drop_rcu(nd, dentry); | |
511 | return 0; | |
512 | } | |
513 | ||
514 | /** | |
515 | * nameidata_drop_rcu_last - drop nameidata ending path walk out of rcu-walk | |
516 | * @nd: nameidata pathwalk data to drop | |
39191628 | 517 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f NP |
518 | * |
519 | * nameidata_drop_rcu_last attempts to drop the current nd->path into ref-walk. | |
520 | * nd->path should be the final element of the lookup, so nd->root is discarded. | |
521 | * Must be called from rcu-walk context. | |
522 | */ | |
523 | static int nameidata_drop_rcu_last(struct nameidata *nd) | |
524 | { | |
525 | struct dentry *dentry = nd->path.dentry; | |
526 | ||
527 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
528 | nd->flags &= ~LOOKUP_RCU; | |
529 | nd->root.mnt = NULL; | |
530 | spin_lock(&dentry->d_lock); | |
531 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
532 | goto err_unlock; | |
533 | BUG_ON(nd->inode != dentry->d_inode); | |
534 | spin_unlock(&dentry->d_lock); | |
535 | ||
536 | mntget(nd->path.mnt); | |
537 | ||
538 | rcu_read_unlock(); | |
539 | br_read_unlock(vfsmount_lock); | |
540 | ||
541 | return 0; | |
542 | ||
543 | err_unlock: | |
544 | spin_unlock(&dentry->d_lock); | |
545 | rcu_read_unlock(); | |
546 | br_read_unlock(vfsmount_lock); | |
547 | return -ECHILD; | |
548 | } | |
549 | ||
550 | /* Try to drop out of rcu-walk mode if we were in it, otherwise do nothing. */ | |
551 | static inline int nameidata_drop_rcu_last_maybe(struct nameidata *nd) | |
552 | { | |
553 | if (likely(nd->flags & LOOKUP_RCU)) | |
554 | return nameidata_drop_rcu_last(nd); | |
555 | return 0; | |
556 | } | |
557 | ||
834f2a4a TM |
558 | /** |
559 | * release_open_intent - free up open intent resources | |
560 | * @nd: pointer to nameidata | |
561 | */ | |
562 | void release_open_intent(struct nameidata *nd) | |
563 | { | |
2dab5974 LT |
564 | struct file *file = nd->intent.open.file; |
565 | ||
566 | if (file && !IS_ERR(file)) { | |
567 | if (file->f_path.dentry == NULL) | |
568 | put_filp(file); | |
569 | else | |
570 | fput(file); | |
571 | } | |
834f2a4a TM |
572 | } |
573 | ||
bb20c18d NP |
574 | /* |
575 | * Call d_revalidate and handle filesystems that request rcu-walk | |
576 | * to be dropped. This may be called and return in rcu-walk mode, | |
577 | * regardless of success or error. If -ECHILD is returned, the caller | |
578 | * must return -ECHILD back up the path walk stack so path walk may | |
579 | * be restarted in ref-walk mode. | |
580 | */ | |
34286d66 NP |
581 | static int d_revalidate(struct dentry *dentry, struct nameidata *nd) |
582 | { | |
583 | int status; | |
584 | ||
585 | status = dentry->d_op->d_revalidate(dentry, nd); | |
586 | if (status == -ECHILD) { | |
587 | if (nameidata_dentry_drop_rcu(nd, dentry)) | |
588 | return status; | |
589 | status = dentry->d_op->d_revalidate(dentry, nd); | |
590 | } | |
591 | ||
592 | return status; | |
593 | } | |
594 | ||
f5e1c1c1 | 595 | static struct dentry * |
bcdc5e01 IK |
596 | do_revalidate(struct dentry *dentry, struct nameidata *nd) |
597 | { | |
f5e1c1c1 | 598 | int status = d_revalidate(dentry, nd); |
bcdc5e01 IK |
599 | if (unlikely(status <= 0)) { |
600 | /* | |
601 | * The dentry failed validation. | |
602 | * If d_revalidate returned 0 attempt to invalidate | |
603 | * the dentry otherwise d_revalidate is asking us | |
604 | * to return a fail status. | |
605 | */ | |
34286d66 | 606 | if (status < 0) { |
f5e1c1c1 | 607 | dput(dentry); |
34286d66 | 608 | dentry = ERR_PTR(status); |
f5e1c1c1 AV |
609 | } else if (!d_invalidate(dentry)) { |
610 | dput(dentry); | |
611 | dentry = NULL; | |
bcdc5e01 IK |
612 | } |
613 | } | |
614 | return dentry; | |
615 | } | |
616 | ||
f5e1c1c1 AV |
617 | static inline struct dentry * |
618 | do_revalidate_rcu(struct dentry *dentry, struct nameidata *nd) | |
619 | { | |
620 | int status = dentry->d_op->d_revalidate(dentry, nd); | |
621 | if (likely(status > 0)) | |
622 | return dentry; | |
623 | if (status == -ECHILD) { | |
624 | if (nameidata_dentry_drop_rcu(nd, dentry)) | |
625 | return ERR_PTR(-ECHILD); | |
626 | return do_revalidate(dentry, nd); | |
627 | } | |
628 | if (status < 0) | |
629 | return ERR_PTR(status); | |
630 | /* Don't d_invalidate in rcu-walk mode */ | |
631 | if (nameidata_dentry_drop_rcu(nd, dentry)) | |
632 | return ERR_PTR(-ECHILD); | |
633 | if (!d_invalidate(dentry)) { | |
634 | dput(dentry); | |
635 | dentry = NULL; | |
636 | } | |
637 | return dentry; | |
638 | } | |
639 | ||
fb045adb NP |
640 | static inline int need_reval_dot(struct dentry *dentry) |
641 | { | |
642 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
643 | return 0; | |
644 | ||
645 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) | |
646 | return 0; | |
647 | ||
648 | return 1; | |
649 | } | |
650 | ||
39159de2 JL |
651 | /* |
652 | * force_reval_path - force revalidation of a dentry | |
653 | * | |
654 | * In some situations the path walking code will trust dentries without | |
655 | * revalidating them. This causes problems for filesystems that depend on | |
656 | * d_revalidate to handle file opens (e.g. NFSv4). When FS_REVAL_DOT is set | |
657 | * (which indicates that it's possible for the dentry to go stale), force | |
658 | * a d_revalidate call before proceeding. | |
659 | * | |
660 | * Returns 0 if the revalidation was successful. If the revalidation fails, | |
661 | * either return the error returned by d_revalidate or -ESTALE if the | |
662 | * revalidation it just returned 0. If d_revalidate returns 0, we attempt to | |
663 | * invalidate the dentry. It's up to the caller to handle putting references | |
664 | * to the path if necessary. | |
665 | */ | |
666 | static int | |
667 | force_reval_path(struct path *path, struct nameidata *nd) | |
668 | { | |
669 | int status; | |
670 | struct dentry *dentry = path->dentry; | |
671 | ||
672 | /* | |
673 | * only check on filesystems where it's possible for the dentry to | |
fb045adb | 674 | * become stale. |
39159de2 | 675 | */ |
fb045adb | 676 | if (!need_reval_dot(dentry)) |
39159de2 JL |
677 | return 0; |
678 | ||
34286d66 | 679 | status = d_revalidate(dentry, nd); |
39159de2 JL |
680 | if (status > 0) |
681 | return 0; | |
682 | ||
683 | if (!status) { | |
684 | d_invalidate(dentry); | |
685 | status = -ESTALE; | |
686 | } | |
687 | return status; | |
688 | } | |
689 | ||
1da177e4 | 690 | /* |
b75b5086 AV |
691 | * Short-cut version of permission(), for calling on directories |
692 | * during pathname resolution. Combines parts of permission() | |
693 | * and generic_permission(), and tests ONLY for MAY_EXEC permission. | |
1da177e4 LT |
694 | * |
695 | * If appropriate, check DAC only. If not appropriate, or | |
b75b5086 | 696 | * short-cut DAC fails, then call ->permission() to do more |
1da177e4 LT |
697 | * complete permission check. |
698 | */ | |
b74c79e9 | 699 | static inline int exec_permission(struct inode *inode, unsigned int flags) |
1da177e4 | 700 | { |
5909ccaa | 701 | int ret; |
1da177e4 | 702 | |
cb9179ea | 703 | if (inode->i_op->permission) { |
b74c79e9 NP |
704 | ret = inode->i_op->permission(inode, MAY_EXEC, flags); |
705 | } else { | |
706 | ret = acl_permission_check(inode, MAY_EXEC, flags, | |
707 | inode->i_op->check_acl); | |
cb9179ea | 708 | } |
b74c79e9 | 709 | if (likely(!ret)) |
1da177e4 | 710 | goto ok; |
b74c79e9 | 711 | if (ret == -ECHILD) |
31e6b01f | 712 | return ret; |
1da177e4 | 713 | |
f1ac9f6b | 714 | if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
715 | goto ok; |
716 | ||
5909ccaa | 717 | return ret; |
1da177e4 | 718 | ok: |
b74c79e9 | 719 | return security_inode_exec_permission(inode, flags); |
1da177e4 LT |
720 | } |
721 | ||
2a737871 AV |
722 | static __always_inline void set_root(struct nameidata *nd) |
723 | { | |
f7ad3c6b MS |
724 | if (!nd->root.mnt) |
725 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
726 | } |
727 | ||
6de88d72 AV |
728 | static int link_path_walk(const char *, struct nameidata *); |
729 | ||
31e6b01f NP |
730 | static __always_inline void set_root_rcu(struct nameidata *nd) |
731 | { | |
732 | if (!nd->root.mnt) { | |
733 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
734 | unsigned seq; |
735 | ||
736 | do { | |
737 | seq = read_seqcount_begin(&fs->seq); | |
738 | nd->root = fs->root; | |
739 | } while (read_seqcount_retry(&fs->seq, seq)); | |
31e6b01f NP |
740 | } |
741 | } | |
742 | ||
f1662356 | 743 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 744 | { |
31e6b01f NP |
745 | int ret; |
746 | ||
1da177e4 LT |
747 | if (IS_ERR(link)) |
748 | goto fail; | |
749 | ||
750 | if (*link == '/') { | |
2a737871 | 751 | set_root(nd); |
1d957f9b | 752 | path_put(&nd->path); |
2a737871 AV |
753 | nd->path = nd->root; |
754 | path_get(&nd->root); | |
1da177e4 | 755 | } |
31e6b01f | 756 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 757 | |
31e6b01f NP |
758 | ret = link_path_walk(link, nd); |
759 | return ret; | |
1da177e4 | 760 | fail: |
1d957f9b | 761 | path_put(&nd->path); |
1da177e4 LT |
762 | return PTR_ERR(link); |
763 | } | |
764 | ||
1d957f9b | 765 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
766 | { |
767 | dput(path->dentry); | |
4ac91378 | 768 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
769 | mntput(path->mnt); |
770 | } | |
771 | ||
7b9337aa NP |
772 | static inline void path_to_nameidata(const struct path *path, |
773 | struct nameidata *nd) | |
051d3812 | 774 | { |
31e6b01f NP |
775 | if (!(nd->flags & LOOKUP_RCU)) { |
776 | dput(nd->path.dentry); | |
777 | if (nd->path.mnt != path->mnt) | |
778 | mntput(nd->path.mnt); | |
9a229683 | 779 | } |
31e6b01f | 780 | nd->path.mnt = path->mnt; |
4ac91378 | 781 | nd->path.dentry = path->dentry; |
051d3812 IK |
782 | } |
783 | ||
def4af30 | 784 | static __always_inline int |
7b9337aa | 785 | __do_follow_link(const struct path *link, struct nameidata *nd, void **p) |
1da177e4 LT |
786 | { |
787 | int error; | |
7b9337aa | 788 | struct dentry *dentry = link->dentry; |
1da177e4 | 789 | |
844a3917 AV |
790 | BUG_ON(nd->flags & LOOKUP_RCU); |
791 | ||
7b9337aa | 792 | touch_atime(link->mnt, dentry); |
1da177e4 | 793 | nd_set_link(nd, NULL); |
cd4e91d3 | 794 | |
87556ef1 DH |
795 | if (link->mnt == nd->path.mnt) |
796 | mntget(link->mnt); | |
31e6b01f | 797 | |
86acdca1 | 798 | nd->last_type = LAST_BIND; |
def4af30 AV |
799 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
800 | error = PTR_ERR(*p); | |
801 | if (!IS_ERR(*p)) { | |
1da177e4 | 802 | char *s = nd_get_link(nd); |
cc314eef | 803 | error = 0; |
1da177e4 LT |
804 | if (s) |
805 | error = __vfs_follow_link(nd, s); | |
39159de2 JL |
806 | else if (nd->last_type == LAST_BIND) { |
807 | error = force_reval_path(&nd->path, nd); | |
808 | if (error) | |
809 | path_put(&nd->path); | |
810 | } | |
1da177e4 | 811 | } |
1da177e4 LT |
812 | return error; |
813 | } | |
814 | ||
815 | /* | |
816 | * This limits recursive symlink follows to 8, while | |
817 | * limiting consecutive symlinks to 40. | |
818 | * | |
819 | * Without that kind of total limit, nasty chains of consecutive | |
820 | * symlinks can cause almost arbitrarily long lookups. | |
821 | */ | |
90ebe565 | 822 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 | 823 | { |
def4af30 | 824 | void *cookie; |
1da177e4 | 825 | int err = -ELOOP; |
844a3917 AV |
826 | |
827 | /* We drop rcu-walk here */ | |
828 | if (nameidata_dentry_drop_rcu_maybe(nd, path->dentry)) | |
829 | return -ECHILD; | |
830 | ||
1da177e4 LT |
831 | if (current->link_count >= MAX_NESTED_LINKS) |
832 | goto loop; | |
833 | if (current->total_link_count >= 40) | |
834 | goto loop; | |
835 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
836 | cond_resched(); | |
90ebe565 | 837 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
838 | if (err) |
839 | goto loop; | |
840 | current->link_count++; | |
841 | current->total_link_count++; | |
842 | nd->depth++; | |
def4af30 AV |
843 | err = __do_follow_link(path, nd, &cookie); |
844 | if (!IS_ERR(cookie) && path->dentry->d_inode->i_op->put_link) | |
845 | path->dentry->d_inode->i_op->put_link(path->dentry, nd, cookie); | |
258fa999 | 846 | path_put(path); |
839d9f93 AV |
847 | current->link_count--; |
848 | nd->depth--; | |
1da177e4 LT |
849 | return err; |
850 | loop: | |
1d957f9b JB |
851 | path_put_conditional(path, nd); |
852 | path_put(&nd->path); | |
1da177e4 LT |
853 | return err; |
854 | } | |
855 | ||
31e6b01f NP |
856 | static int follow_up_rcu(struct path *path) |
857 | { | |
858 | struct vfsmount *parent; | |
859 | struct dentry *mountpoint; | |
860 | ||
861 | parent = path->mnt->mnt_parent; | |
862 | if (parent == path->mnt) | |
863 | return 0; | |
864 | mountpoint = path->mnt->mnt_mountpoint; | |
865 | path->dentry = mountpoint; | |
866 | path->mnt = parent; | |
867 | return 1; | |
868 | } | |
869 | ||
bab77ebf | 870 | int follow_up(struct path *path) |
1da177e4 LT |
871 | { |
872 | struct vfsmount *parent; | |
873 | struct dentry *mountpoint; | |
99b7db7b NP |
874 | |
875 | br_read_lock(vfsmount_lock); | |
bab77ebf AV |
876 | parent = path->mnt->mnt_parent; |
877 | if (parent == path->mnt) { | |
99b7db7b | 878 | br_read_unlock(vfsmount_lock); |
1da177e4 LT |
879 | return 0; |
880 | } | |
881 | mntget(parent); | |
bab77ebf | 882 | mountpoint = dget(path->mnt->mnt_mountpoint); |
99b7db7b | 883 | br_read_unlock(vfsmount_lock); |
bab77ebf AV |
884 | dput(path->dentry); |
885 | path->dentry = mountpoint; | |
886 | mntput(path->mnt); | |
887 | path->mnt = parent; | |
1da177e4 LT |
888 | return 1; |
889 | } | |
890 | ||
b5c84bf6 | 891 | /* |
9875cf80 DH |
892 | * Perform an automount |
893 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
894 | * were called with. | |
1da177e4 | 895 | */ |
9875cf80 DH |
896 | static int follow_automount(struct path *path, unsigned flags, |
897 | bool *need_mntput) | |
31e6b01f | 898 | { |
9875cf80 | 899 | struct vfsmount *mnt; |
ea5b778a | 900 | int err; |
9875cf80 DH |
901 | |
902 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
903 | return -EREMOTE; | |
904 | ||
6f45b656 DH |
905 | /* We don't want to mount if someone supplied AT_NO_AUTOMOUNT |
906 | * and this is the terminal part of the path. | |
907 | */ | |
908 | if ((flags & LOOKUP_NO_AUTOMOUNT) && !(flags & LOOKUP_CONTINUE)) | |
909 | return -EISDIR; /* we actually want to stop here */ | |
910 | ||
9875cf80 DH |
911 | /* We want to mount if someone is trying to open/create a file of any |
912 | * type under the mountpoint, wants to traverse through the mountpoint | |
913 | * or wants to open the mounted directory. | |
914 | * | |
915 | * We don't want to mount if someone's just doing a stat and they've | |
916 | * set AT_SYMLINK_NOFOLLOW - unless they're stat'ing a directory and | |
917 | * appended a '/' to the name. | |
918 | */ | |
919 | if (!(flags & LOOKUP_FOLLOW) && | |
920 | !(flags & (LOOKUP_CONTINUE | LOOKUP_DIRECTORY | | |
921 | LOOKUP_OPEN | LOOKUP_CREATE))) | |
922 | return -EISDIR; | |
923 | ||
924 | current->total_link_count++; | |
925 | if (current->total_link_count >= 40) | |
926 | return -ELOOP; | |
927 | ||
928 | mnt = path->dentry->d_op->d_automount(path); | |
929 | if (IS_ERR(mnt)) { | |
930 | /* | |
931 | * The filesystem is allowed to return -EISDIR here to indicate | |
932 | * it doesn't want to automount. For instance, autofs would do | |
933 | * this so that its userspace daemon can mount on this dentry. | |
934 | * | |
935 | * However, we can only permit this if it's a terminal point in | |
936 | * the path being looked up; if it wasn't then the remainder of | |
937 | * the path is inaccessible and we should say so. | |
938 | */ | |
939 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_CONTINUE)) | |
940 | return -EREMOTE; | |
941 | return PTR_ERR(mnt); | |
31e6b01f | 942 | } |
ea5b778a | 943 | |
9875cf80 DH |
944 | if (!mnt) /* mount collision */ |
945 | return 0; | |
31e6b01f | 946 | |
19a167af | 947 | err = finish_automount(mnt, path); |
9875cf80 | 948 | |
ea5b778a DH |
949 | switch (err) { |
950 | case -EBUSY: | |
951 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 952 | return 0; |
ea5b778a | 953 | case 0: |
ea5b778a DH |
954 | dput(path->dentry); |
955 | if (*need_mntput) | |
956 | mntput(path->mnt); | |
957 | path->mnt = mnt; | |
958 | path->dentry = dget(mnt->mnt_root); | |
959 | *need_mntput = true; | |
960 | return 0; | |
19a167af AV |
961 | default: |
962 | return err; | |
ea5b778a | 963 | } |
19a167af | 964 | |
463ffb2e AV |
965 | } |
966 | ||
9875cf80 DH |
967 | /* |
968 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 969 | * - Flagged for transit management (autofs) |
9875cf80 DH |
970 | * - Flagged as mountpoint |
971 | * - Flagged as automount point | |
972 | * | |
973 | * This may only be called in refwalk mode. | |
974 | * | |
975 | * Serialization is taken care of in namespace.c | |
976 | */ | |
977 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 978 | { |
9875cf80 DH |
979 | unsigned managed; |
980 | bool need_mntput = false; | |
981 | int ret; | |
982 | ||
983 | /* Given that we're not holding a lock here, we retain the value in a | |
984 | * local variable for each dentry as we look at it so that we don't see | |
985 | * the components of that value change under us */ | |
986 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
987 | managed &= DCACHE_MANAGED_DENTRY, | |
988 | unlikely(managed != 0)) { | |
cc53ce53 DH |
989 | /* Allow the filesystem to manage the transit without i_mutex |
990 | * being held. */ | |
991 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
992 | BUG_ON(!path->dentry->d_op); | |
993 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f DH |
994 | ret = path->dentry->d_op->d_manage(path->dentry, |
995 | false, false); | |
cc53ce53 DH |
996 | if (ret < 0) |
997 | return ret == -EISDIR ? 0 : ret; | |
998 | } | |
999 | ||
9875cf80 DH |
1000 | /* Transit to a mounted filesystem. */ |
1001 | if (managed & DCACHE_MOUNTED) { | |
1002 | struct vfsmount *mounted = lookup_mnt(path); | |
1003 | if (mounted) { | |
1004 | dput(path->dentry); | |
1005 | if (need_mntput) | |
1006 | mntput(path->mnt); | |
1007 | path->mnt = mounted; | |
1008 | path->dentry = dget(mounted->mnt_root); | |
1009 | need_mntput = true; | |
1010 | continue; | |
1011 | } | |
1012 | ||
1013 | /* Something is mounted on this dentry in another | |
1014 | * namespace and/or whatever was mounted there in this | |
1015 | * namespace got unmounted before we managed to get the | |
1016 | * vfsmount_lock */ | |
1017 | } | |
1018 | ||
1019 | /* Handle an automount point */ | |
1020 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
1021 | ret = follow_automount(path, flags, &need_mntput); | |
1022 | if (ret < 0) | |
1023 | return ret == -EISDIR ? 0 : ret; | |
1024 | continue; | |
1025 | } | |
1026 | ||
1027 | /* We didn't change the current path point */ | |
1028 | break; | |
1da177e4 | 1029 | } |
9875cf80 | 1030 | return 0; |
1da177e4 LT |
1031 | } |
1032 | ||
cc53ce53 | 1033 | int follow_down_one(struct path *path) |
1da177e4 LT |
1034 | { |
1035 | struct vfsmount *mounted; | |
1036 | ||
1c755af4 | 1037 | mounted = lookup_mnt(path); |
1da177e4 | 1038 | if (mounted) { |
9393bd07 AV |
1039 | dput(path->dentry); |
1040 | mntput(path->mnt); | |
1041 | path->mnt = mounted; | |
1042 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1043 | return 1; |
1044 | } | |
1045 | return 0; | |
1046 | } | |
1047 | ||
9875cf80 DH |
1048 | /* |
1049 | * Skip to top of mountpoint pile in rcuwalk mode. We abort the rcu-walk if we | |
cc53ce53 | 1050 | * meet a managed dentry and we're not walking to "..". True is returned to |
9875cf80 DH |
1051 | * continue, false to abort. |
1052 | */ | |
1053 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
1054 | struct inode **inode, bool reverse_transit) | |
1055 | { | |
1056 | while (d_mountpoint(path->dentry)) { | |
1057 | struct vfsmount *mounted; | |
ab90911f DH |
1058 | if (unlikely(path->dentry->d_flags & DCACHE_MANAGE_TRANSIT) && |
1059 | !reverse_transit && | |
1060 | path->dentry->d_op->d_manage(path->dentry, false, true) < 0) | |
1061 | return false; | |
9875cf80 DH |
1062 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
1063 | if (!mounted) | |
1064 | break; | |
1065 | path->mnt = mounted; | |
1066 | path->dentry = mounted->mnt_root; | |
1067 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); | |
1068 | *inode = path->dentry->d_inode; | |
1069 | } | |
1070 | ||
1071 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1072 | return reverse_transit; | |
1073 | return true; | |
1074 | } | |
1075 | ||
31e6b01f NP |
1076 | static int follow_dotdot_rcu(struct nameidata *nd) |
1077 | { | |
1078 | struct inode *inode = nd->inode; | |
1079 | ||
1080 | set_root_rcu(nd); | |
1081 | ||
9875cf80 | 1082 | while (1) { |
31e6b01f NP |
1083 | if (nd->path.dentry == nd->root.dentry && |
1084 | nd->path.mnt == nd->root.mnt) { | |
1085 | break; | |
1086 | } | |
1087 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
1088 | struct dentry *old = nd->path.dentry; | |
1089 | struct dentry *parent = old->d_parent; | |
1090 | unsigned seq; | |
1091 | ||
1092 | seq = read_seqcount_begin(&parent->d_seq); | |
1093 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
1094 | return -ECHILD; | |
1095 | inode = parent->d_inode; | |
1096 | nd->path.dentry = parent; | |
1097 | nd->seq = seq; | |
1098 | break; | |
1099 | } | |
1100 | if (!follow_up_rcu(&nd->path)) | |
1101 | break; | |
1102 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
1103 | inode = nd->path.dentry->d_inode; | |
1104 | } | |
9875cf80 | 1105 | __follow_mount_rcu(nd, &nd->path, &inode, true); |
31e6b01f NP |
1106 | nd->inode = inode; |
1107 | ||
1108 | return 0; | |
1109 | } | |
1110 | ||
cc53ce53 DH |
1111 | /* |
1112 | * Follow down to the covering mount currently visible to userspace. At each | |
1113 | * point, the filesystem owning that dentry may be queried as to whether the | |
1114 | * caller is permitted to proceed or not. | |
1115 | * | |
1116 | * Care must be taken as namespace_sem may be held (indicated by mounting_here | |
1117 | * being true). | |
1118 | */ | |
1119 | int follow_down(struct path *path, bool mounting_here) | |
1120 | { | |
1121 | unsigned managed; | |
1122 | int ret; | |
1123 | ||
1124 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1125 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1126 | /* Allow the filesystem to manage the transit without i_mutex | |
1127 | * being held. | |
1128 | * | |
1129 | * We indicate to the filesystem if someone is trying to mount | |
1130 | * something here. This gives autofs the chance to deny anyone | |
1131 | * other than its daemon the right to mount on its | |
1132 | * superstructure. | |
1133 | * | |
1134 | * The filesystem may sleep at this point. | |
1135 | */ | |
1136 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1137 | BUG_ON(!path->dentry->d_op); | |
1138 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f DH |
1139 | ret = path->dentry->d_op->d_manage( |
1140 | path->dentry, mounting_here, false); | |
cc53ce53 DH |
1141 | if (ret < 0) |
1142 | return ret == -EISDIR ? 0 : ret; | |
1143 | } | |
1144 | ||
1145 | /* Transit to a mounted filesystem. */ | |
1146 | if (managed & DCACHE_MOUNTED) { | |
1147 | struct vfsmount *mounted = lookup_mnt(path); | |
1148 | if (!mounted) | |
1149 | break; | |
1150 | dput(path->dentry); | |
1151 | mntput(path->mnt); | |
1152 | path->mnt = mounted; | |
1153 | path->dentry = dget(mounted->mnt_root); | |
1154 | continue; | |
1155 | } | |
1156 | ||
1157 | /* Don't handle automount points here */ | |
1158 | break; | |
1159 | } | |
1160 | return 0; | |
1161 | } | |
1162 | ||
9875cf80 DH |
1163 | /* |
1164 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1165 | */ | |
1166 | static void follow_mount(struct path *path) | |
1167 | { | |
1168 | while (d_mountpoint(path->dentry)) { | |
1169 | struct vfsmount *mounted = lookup_mnt(path); | |
1170 | if (!mounted) | |
1171 | break; | |
1172 | dput(path->dentry); | |
1173 | mntput(path->mnt); | |
1174 | path->mnt = mounted; | |
1175 | path->dentry = dget(mounted->mnt_root); | |
1176 | } | |
1177 | } | |
1178 | ||
31e6b01f | 1179 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1180 | { |
2a737871 | 1181 | set_root(nd); |
e518ddb7 | 1182 | |
1da177e4 | 1183 | while(1) { |
4ac91378 | 1184 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1185 | |
2a737871 AV |
1186 | if (nd->path.dentry == nd->root.dentry && |
1187 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1188 | break; |
1189 | } | |
4ac91378 | 1190 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1191 | /* rare case of legitimate dget_parent()... */ |
1192 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1193 | dput(old); |
1194 | break; | |
1195 | } | |
3088dd70 | 1196 | if (!follow_up(&nd->path)) |
1da177e4 | 1197 | break; |
1da177e4 | 1198 | } |
79ed0226 | 1199 | follow_mount(&nd->path); |
31e6b01f | 1200 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1201 | } |
1202 | ||
baa03890 NP |
1203 | /* |
1204 | * Allocate a dentry with name and parent, and perform a parent | |
1205 | * directory ->lookup on it. Returns the new dentry, or ERR_PTR | |
1206 | * on error. parent->d_inode->i_mutex must be held. d_lookup must | |
1207 | * have verified that no child exists while under i_mutex. | |
1208 | */ | |
1209 | static struct dentry *d_alloc_and_lookup(struct dentry *parent, | |
1210 | struct qstr *name, struct nameidata *nd) | |
1211 | { | |
1212 | struct inode *inode = parent->d_inode; | |
1213 | struct dentry *dentry; | |
1214 | struct dentry *old; | |
1215 | ||
1216 | /* Don't create child dentry for a dead directory. */ | |
1217 | if (unlikely(IS_DEADDIR(inode))) | |
1218 | return ERR_PTR(-ENOENT); | |
1219 | ||
1220 | dentry = d_alloc(parent, name); | |
1221 | if (unlikely(!dentry)) | |
1222 | return ERR_PTR(-ENOMEM); | |
1223 | ||
1224 | old = inode->i_op->lookup(inode, dentry, nd); | |
1225 | if (unlikely(old)) { | |
1226 | dput(dentry); | |
1227 | dentry = old; | |
1228 | } | |
1229 | return dentry; | |
1230 | } | |
1231 | ||
1da177e4 LT |
1232 | /* |
1233 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1234 | * small and for now I'd prefer to have fast path as straight as possible. | |
1235 | * It _is_ time-critical. | |
1236 | */ | |
1237 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
31e6b01f | 1238 | struct path *path, struct inode **inode) |
1da177e4 | 1239 | { |
4ac91378 | 1240 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1241 | struct dentry *dentry, *parent = nd->path.dentry; |
6e6b1bd1 | 1242 | struct inode *dir; |
9875cf80 DH |
1243 | int err; |
1244 | ||
3cac260a AV |
1245 | /* |
1246 | * See if the low-level filesystem might want | |
1247 | * to use its own hash.. | |
1248 | */ | |
fb045adb | 1249 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
9875cf80 | 1250 | err = parent->d_op->d_hash(parent, nd->inode, name); |
3cac260a AV |
1251 | if (err < 0) |
1252 | return err; | |
1253 | } | |
1da177e4 | 1254 | |
b04f784e NP |
1255 | /* |
1256 | * Rename seqlock is not required here because in the off chance | |
1257 | * of a false negative due to a concurrent rename, we're going to | |
1258 | * do the non-racy lookup, below. | |
1259 | */ | |
31e6b01f NP |
1260 | if (nd->flags & LOOKUP_RCU) { |
1261 | unsigned seq; | |
1262 | ||
1263 | *inode = nd->inode; | |
1264 | dentry = __d_lookup_rcu(parent, name, &seq, inode); | |
1265 | if (!dentry) { | |
1266 | if (nameidata_drop_rcu(nd)) | |
1267 | return -ECHILD; | |
1268 | goto need_lookup; | |
1269 | } | |
1270 | /* Memory barrier in read_seqcount_begin of child is enough */ | |
1271 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) | |
1272 | return -ECHILD; | |
1273 | ||
1274 | nd->seq = seq; | |
24643087 | 1275 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
f5e1c1c1 | 1276 | dentry = do_revalidate_rcu(dentry, nd); |
24643087 AV |
1277 | if (!dentry) |
1278 | goto need_lookup; | |
1279 | if (IS_ERR(dentry)) | |
1280 | goto fail; | |
1281 | if (!(nd->flags & LOOKUP_RCU)) | |
1282 | goto done; | |
1283 | } | |
31e6b01f NP |
1284 | path->mnt = mnt; |
1285 | path->dentry = dentry; | |
9875cf80 DH |
1286 | if (likely(__follow_mount_rcu(nd, path, inode, false))) |
1287 | return 0; | |
1288 | if (nameidata_drop_rcu(nd)) | |
1289 | return -ECHILD; | |
1290 | /* fallthru */ | |
1291 | } | |
1292 | dentry = __d_lookup(parent, name); | |
1293 | if (!dentry) | |
1294 | goto need_lookup; | |
2e2e88ea | 1295 | found: |
24643087 AV |
1296 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
1297 | dentry = do_revalidate(dentry, nd); | |
1298 | if (!dentry) | |
1299 | goto need_lookup; | |
1300 | if (IS_ERR(dentry)) | |
1301 | goto fail; | |
1302 | } | |
1da177e4 | 1303 | done: |
9875cf80 DH |
1304 | path->mnt = mnt; |
1305 | path->dentry = dentry; | |
1306 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1307 | if (unlikely(err < 0)) { |
1308 | path_put_conditional(path, nd); | |
9875cf80 | 1309 | return err; |
89312214 | 1310 | } |
9875cf80 | 1311 | *inode = path->dentry->d_inode; |
1da177e4 LT |
1312 | return 0; |
1313 | ||
1314 | need_lookup: | |
6e6b1bd1 | 1315 | dir = parent->d_inode; |
31e6b01f | 1316 | BUG_ON(nd->inode != dir); |
6e6b1bd1 AV |
1317 | |
1318 | mutex_lock(&dir->i_mutex); | |
1319 | /* | |
1320 | * First re-do the cached lookup just in case it was created | |
b04f784e NP |
1321 | * while we waited for the directory semaphore, or the first |
1322 | * lookup failed due to an unrelated rename. | |
6e6b1bd1 | 1323 | * |
b04f784e NP |
1324 | * This could use version numbering or similar to avoid unnecessary |
1325 | * cache lookups, but then we'd have to do the first lookup in the | |
1326 | * non-racy way. However in the common case here, everything should | |
1327 | * be hot in cache, so would it be a big win? | |
6e6b1bd1 AV |
1328 | */ |
1329 | dentry = d_lookup(parent, name); | |
baa03890 NP |
1330 | if (likely(!dentry)) { |
1331 | dentry = d_alloc_and_lookup(parent, name, nd); | |
6e6b1bd1 AV |
1332 | mutex_unlock(&dir->i_mutex); |
1333 | if (IS_ERR(dentry)) | |
1334 | goto fail; | |
1335 | goto done; | |
1336 | } | |
6e6b1bd1 AV |
1337 | /* |
1338 | * Uhhuh! Nasty case: the cache was re-populated while | |
1339 | * we waited on the semaphore. Need to revalidate. | |
1340 | */ | |
1341 | mutex_unlock(&dir->i_mutex); | |
2e2e88ea | 1342 | goto found; |
1da177e4 | 1343 | |
1da177e4 LT |
1344 | fail: |
1345 | return PTR_ERR(dentry); | |
1346 | } | |
1347 | ||
1348 | /* | |
1349 | * Name resolution. | |
ea3834d9 PM |
1350 | * This is the basic name resolution function, turning a pathname into |
1351 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1352 | * |
ea3834d9 PM |
1353 | * Returns 0 and nd will have valid dentry and mnt on success. |
1354 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1355 | */ |
6de88d72 | 1356 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1357 | { |
1358 | struct path next; | |
1da177e4 LT |
1359 | int err; |
1360 | unsigned int lookup_flags = nd->flags; | |
1361 | ||
1362 | while (*name=='/') | |
1363 | name++; | |
1364 | if (!*name) | |
1365 | goto return_reval; | |
1366 | ||
1da177e4 | 1367 | if (nd->depth) |
f55eab82 | 1368 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
1369 | |
1370 | /* At this point we know we have a real path component. */ | |
1371 | for(;;) { | |
31e6b01f | 1372 | struct inode *inode; |
1da177e4 LT |
1373 | unsigned long hash; |
1374 | struct qstr this; | |
1375 | unsigned int c; | |
1376 | ||
cdce5d6b | 1377 | nd->flags |= LOOKUP_CONTINUE; |
31e6b01f | 1378 | if (nd->flags & LOOKUP_RCU) { |
b74c79e9 | 1379 | err = exec_permission(nd->inode, IPERM_FLAG_RCU); |
31e6b01f NP |
1380 | if (err == -ECHILD) { |
1381 | if (nameidata_drop_rcu(nd)) | |
1382 | return -ECHILD; | |
1383 | goto exec_again; | |
1384 | } | |
1385 | } else { | |
1386 | exec_again: | |
b74c79e9 | 1387 | err = exec_permission(nd->inode, 0); |
31e6b01f | 1388 | } |
1da177e4 LT |
1389 | if (err) |
1390 | break; | |
1391 | ||
1392 | this.name = name; | |
1393 | c = *(const unsigned char *)name; | |
1394 | ||
1395 | hash = init_name_hash(); | |
1396 | do { | |
1397 | name++; | |
1398 | hash = partial_name_hash(c, hash); | |
1399 | c = *(const unsigned char *)name; | |
1400 | } while (c && (c != '/')); | |
1401 | this.len = name - (const char *) this.name; | |
1402 | this.hash = end_name_hash(hash); | |
1403 | ||
1404 | /* remove trailing slashes? */ | |
1405 | if (!c) | |
1406 | goto last_component; | |
1407 | while (*++name == '/'); | |
1408 | if (!*name) | |
1409 | goto last_with_slashes; | |
1410 | ||
1411 | /* | |
1412 | * "." and ".." are special - ".." especially so because it has | |
1413 | * to be able to know about the current root directory and | |
1414 | * parent relationships. | |
1415 | */ | |
1416 | if (this.name[0] == '.') switch (this.len) { | |
1417 | default: | |
1418 | break; | |
31e6b01f | 1419 | case 2: |
1da177e4 LT |
1420 | if (this.name[1] != '.') |
1421 | break; | |
31e6b01f NP |
1422 | if (nd->flags & LOOKUP_RCU) { |
1423 | if (follow_dotdot_rcu(nd)) | |
1424 | return -ECHILD; | |
1425 | } else | |
1426 | follow_dotdot(nd); | |
1da177e4 LT |
1427 | /* fallthrough */ |
1428 | case 1: | |
1429 | continue; | |
1430 | } | |
1da177e4 | 1431 | /* This does the actual lookups.. */ |
31e6b01f | 1432 | err = do_lookup(nd, &this, &next, &inode); |
1da177e4 LT |
1433 | if (err) |
1434 | break; | |
1da177e4 | 1435 | err = -ENOENT; |
1da177e4 LT |
1436 | if (!inode) |
1437 | goto out_dput; | |
1da177e4 LT |
1438 | |
1439 | if (inode->i_op->follow_link) { | |
31e6b01f | 1440 | BUG_ON(inode != next.dentry->d_inode); |
90ebe565 | 1441 | err = do_follow_link(&next, nd); |
1da177e4 LT |
1442 | if (err) |
1443 | goto return_err; | |
31e6b01f | 1444 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 | 1445 | err = -ENOENT; |
31e6b01f | 1446 | if (!nd->inode) |
1da177e4 | 1447 | break; |
31e6b01f | 1448 | } else { |
09dd17d3 | 1449 | path_to_nameidata(&next, nd); |
31e6b01f NP |
1450 | nd->inode = inode; |
1451 | } | |
1da177e4 | 1452 | err = -ENOTDIR; |
31e6b01f | 1453 | if (!nd->inode->i_op->lookup) |
1da177e4 LT |
1454 | break; |
1455 | continue; | |
1456 | /* here ends the main loop */ | |
1457 | ||
1458 | last_with_slashes: | |
1459 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1460 | last_component: | |
f55eab82 TM |
1461 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
1462 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
1463 | if (lookup_flags & LOOKUP_PARENT) |
1464 | goto lookup_parent; | |
1465 | if (this.name[0] == '.') switch (this.len) { | |
1466 | default: | |
1467 | break; | |
31e6b01f | 1468 | case 2: |
1da177e4 LT |
1469 | if (this.name[1] != '.') |
1470 | break; | |
31e6b01f NP |
1471 | if (nd->flags & LOOKUP_RCU) { |
1472 | if (follow_dotdot_rcu(nd)) | |
1473 | return -ECHILD; | |
1474 | } else | |
1475 | follow_dotdot(nd); | |
1da177e4 LT |
1476 | /* fallthrough */ |
1477 | case 1: | |
1478 | goto return_reval; | |
1479 | } | |
31e6b01f | 1480 | err = do_lookup(nd, &this, &next, &inode); |
1da177e4 LT |
1481 | if (err) |
1482 | break; | |
db372915 DH |
1483 | if (inode && unlikely(inode->i_op->follow_link) && |
1484 | (lookup_flags & LOOKUP_FOLLOW)) { | |
31e6b01f | 1485 | BUG_ON(inode != next.dentry->d_inode); |
90ebe565 | 1486 | err = do_follow_link(&next, nd); |
1da177e4 LT |
1487 | if (err) |
1488 | goto return_err; | |
31e6b01f NP |
1489 | nd->inode = nd->path.dentry->d_inode; |
1490 | } else { | |
09dd17d3 | 1491 | path_to_nameidata(&next, nd); |
31e6b01f NP |
1492 | nd->inode = inode; |
1493 | } | |
1da177e4 | 1494 | err = -ENOENT; |
31e6b01f | 1495 | if (!nd->inode) |
1da177e4 LT |
1496 | break; |
1497 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
1498 | err = -ENOTDIR; | |
31e6b01f | 1499 | if (!nd->inode->i_op->lookup) |
1da177e4 LT |
1500 | break; |
1501 | } | |
1502 | goto return_base; | |
1503 | lookup_parent: | |
1504 | nd->last = this; | |
1505 | nd->last_type = LAST_NORM; | |
1506 | if (this.name[0] != '.') | |
1507 | goto return_base; | |
1508 | if (this.len == 1) | |
1509 | nd->last_type = LAST_DOT; | |
1510 | else if (this.len == 2 && this.name[1] == '.') | |
1511 | nd->last_type = LAST_DOTDOT; | |
1512 | else | |
1513 | goto return_base; | |
1514 | return_reval: | |
1515 | /* | |
1516 | * We bypassed the ordinary revalidation routines. | |
1517 | * We may need to check the cached dentry for staleness. | |
1518 | */ | |
fb045adb | 1519 | if (need_reval_dot(nd->path.dentry)) { |
1da177e4 | 1520 | /* Note: we do not d_invalidate() */ |
34286d66 NP |
1521 | err = d_revalidate(nd->path.dentry, nd); |
1522 | if (!err) | |
1523 | err = -ESTALE; | |
1524 | if (err < 0) | |
1da177e4 LT |
1525 | break; |
1526 | } | |
1527 | return_base: | |
31e6b01f NP |
1528 | if (nameidata_drop_rcu_last_maybe(nd)) |
1529 | return -ECHILD; | |
1da177e4 LT |
1530 | return 0; |
1531 | out_dput: | |
31e6b01f NP |
1532 | if (!(nd->flags & LOOKUP_RCU)) |
1533 | path_put_conditional(&next, nd); | |
1da177e4 LT |
1534 | break; |
1535 | } | |
31e6b01f NP |
1536 | if (!(nd->flags & LOOKUP_RCU)) |
1537 | path_put(&nd->path); | |
1da177e4 LT |
1538 | return_err: |
1539 | return err; | |
1540 | } | |
1541 | ||
31e6b01f NP |
1542 | static inline int path_walk_rcu(const char *name, struct nameidata *nd) |
1543 | { | |
1544 | current->total_link_count = 0; | |
1545 | ||
1546 | return link_path_walk(name, nd); | |
1547 | } | |
1548 | ||
1549 | static inline int path_walk_simple(const char *name, struct nameidata *nd) | |
1550 | { | |
1551 | current->total_link_count = 0; | |
1552 | ||
1553 | return link_path_walk(name, nd); | |
1554 | } | |
1555 | ||
fc9b52cd | 1556 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 1557 | { |
6de88d72 AV |
1558 | struct path save = nd->path; |
1559 | int result; | |
1560 | ||
1da177e4 | 1561 | current->total_link_count = 0; |
6de88d72 AV |
1562 | |
1563 | /* make sure the stuff we saved doesn't go away */ | |
1564 | path_get(&save); | |
1565 | ||
1566 | result = link_path_walk(name, nd); | |
1567 | if (result == -ESTALE) { | |
1568 | /* nd->path had been dropped */ | |
1569 | current->total_link_count = 0; | |
1570 | nd->path = save; | |
1571 | path_get(&nd->path); | |
1572 | nd->flags |= LOOKUP_REVAL; | |
1573 | result = link_path_walk(name, nd); | |
1574 | } | |
1575 | ||
1576 | path_put(&save); | |
1577 | ||
1578 | return result; | |
1da177e4 LT |
1579 | } |
1580 | ||
31e6b01f NP |
1581 | static void path_finish_rcu(struct nameidata *nd) |
1582 | { | |
1583 | if (nd->flags & LOOKUP_RCU) { | |
1584 | /* RCU dangling. Cancel it. */ | |
1585 | nd->flags &= ~LOOKUP_RCU; | |
1586 | nd->root.mnt = NULL; | |
1587 | rcu_read_unlock(); | |
1588 | br_read_unlock(vfsmount_lock); | |
1589 | } | |
1590 | if (nd->file) | |
1591 | fput(nd->file); | |
1592 | } | |
1593 | ||
1594 | static int path_init_rcu(int dfd, const char *name, unsigned int flags, struct nameidata *nd) | |
1595 | { | |
1596 | int retval = 0; | |
1597 | int fput_needed; | |
1598 | struct file *file; | |
1599 | ||
1600 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1601 | nd->flags = flags | LOOKUP_RCU; | |
1602 | nd->depth = 0; | |
1603 | nd->root.mnt = NULL; | |
1604 | nd->file = NULL; | |
1605 | ||
1606 | if (*name=='/') { | |
1607 | struct fs_struct *fs = current->fs; | |
c28cc364 | 1608 | unsigned seq; |
31e6b01f NP |
1609 | |
1610 | br_read_lock(vfsmount_lock); | |
1611 | rcu_read_lock(); | |
1612 | ||
c28cc364 NP |
1613 | do { |
1614 | seq = read_seqcount_begin(&fs->seq); | |
1615 | nd->root = fs->root; | |
1616 | nd->path = nd->root; | |
1617 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1618 | } while (read_seqcount_retry(&fs->seq, seq)); | |
31e6b01f NP |
1619 | |
1620 | } else if (dfd == AT_FDCWD) { | |
1621 | struct fs_struct *fs = current->fs; | |
c28cc364 | 1622 | unsigned seq; |
31e6b01f NP |
1623 | |
1624 | br_read_lock(vfsmount_lock); | |
1625 | rcu_read_lock(); | |
1626 | ||
c28cc364 NP |
1627 | do { |
1628 | seq = read_seqcount_begin(&fs->seq); | |
1629 | nd->path = fs->pwd; | |
1630 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1631 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1632 | ||
31e6b01f NP |
1633 | } else { |
1634 | struct dentry *dentry; | |
1635 | ||
1636 | file = fget_light(dfd, &fput_needed); | |
1637 | retval = -EBADF; | |
1638 | if (!file) | |
1639 | goto out_fail; | |
1640 | ||
1641 | dentry = file->f_path.dentry; | |
1642 | ||
1643 | retval = -ENOTDIR; | |
1644 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
1645 | goto fput_fail; | |
1646 | ||
1647 | retval = file_permission(file, MAY_EXEC); | |
1648 | if (retval) | |
1649 | goto fput_fail; | |
1650 | ||
1651 | nd->path = file->f_path; | |
1652 | if (fput_needed) | |
1653 | nd->file = file; | |
1654 | ||
c28cc364 | 1655 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
31e6b01f NP |
1656 | br_read_lock(vfsmount_lock); |
1657 | rcu_read_lock(); | |
1658 | } | |
1659 | nd->inode = nd->path.dentry->d_inode; | |
1660 | return 0; | |
1661 | ||
1662 | fput_fail: | |
1663 | fput_light(file, fput_needed); | |
1664 | out_fail: | |
1665 | return retval; | |
1666 | } | |
1667 | ||
9b4a9b14 | 1668 | static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) |
1da177e4 | 1669 | { |
ea3834d9 | 1670 | int retval = 0; |
170aa3d0 UD |
1671 | int fput_needed; |
1672 | struct file *file; | |
1da177e4 LT |
1673 | |
1674 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1675 | nd->flags = flags; | |
1676 | nd->depth = 0; | |
2a737871 | 1677 | nd->root.mnt = NULL; |
1da177e4 | 1678 | |
1da177e4 | 1679 | if (*name=='/') { |
2a737871 AV |
1680 | set_root(nd); |
1681 | nd->path = nd->root; | |
1682 | path_get(&nd->root); | |
5590ff0d | 1683 | } else if (dfd == AT_FDCWD) { |
f7ad3c6b | 1684 | get_fs_pwd(current->fs, &nd->path); |
5590ff0d | 1685 | } else { |
5590ff0d UD |
1686 | struct dentry *dentry; |
1687 | ||
1688 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1689 | retval = -EBADF; |
1690 | if (!file) | |
6d09bb62 | 1691 | goto out_fail; |
5590ff0d | 1692 | |
0f7fc9e4 | 1693 | dentry = file->f_path.dentry; |
5590ff0d | 1694 | |
170aa3d0 UD |
1695 | retval = -ENOTDIR; |
1696 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1697 | goto fput_fail; |
5590ff0d UD |
1698 | |
1699 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1700 | if (retval) |
6d09bb62 | 1701 | goto fput_fail; |
5590ff0d | 1702 | |
5dd784d0 JB |
1703 | nd->path = file->f_path; |
1704 | path_get(&file->f_path); | |
5590ff0d UD |
1705 | |
1706 | fput_light(file, fput_needed); | |
1da177e4 | 1707 | } |
31e6b01f | 1708 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1709 | return 0; |
2dfdd266 | 1710 | |
9b4a9b14 AV |
1711 | fput_fail: |
1712 | fput_light(file, fput_needed); | |
1713 | out_fail: | |
1714 | return retval; | |
1715 | } | |
1716 | ||
1717 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ | |
1718 | static int do_path_lookup(int dfd, const char *name, | |
1719 | unsigned int flags, struct nameidata *nd) | |
1720 | { | |
31e6b01f NP |
1721 | int retval; |
1722 | ||
1723 | /* | |
1724 | * Path walking is largely split up into 2 different synchronisation | |
1725 | * schemes, rcu-walk and ref-walk (explained in | |
1726 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1727 | * path walk code, but some things particularly setup, cleanup, and | |
1728 | * following mounts are sufficiently divergent that functions are | |
1729 | * duplicated. Typically there is a function foo(), and its RCU | |
1730 | * analogue, foo_rcu(). | |
1731 | * | |
1732 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1733 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1734 | * be handled by restarting a traditional ref-walk (which will always | |
1735 | * be able to complete). | |
1736 | */ | |
1737 | retval = path_init_rcu(dfd, name, flags, nd); | |
1738 | if (unlikely(retval)) | |
1739 | return retval; | |
1740 | retval = path_walk_rcu(name, nd); | |
1741 | path_finish_rcu(nd); | |
2a737871 AV |
1742 | if (nd->root.mnt) { |
1743 | path_put(&nd->root); | |
1744 | nd->root.mnt = NULL; | |
1745 | } | |
31e6b01f NP |
1746 | |
1747 | if (unlikely(retval == -ECHILD || retval == -ESTALE)) { | |
1748 | /* slower, locked walk */ | |
1749 | if (retval == -ESTALE) | |
1750 | flags |= LOOKUP_REVAL; | |
1751 | retval = path_init(dfd, name, flags, nd); | |
1752 | if (unlikely(retval)) | |
1753 | return retval; | |
1754 | retval = path_walk(name, nd); | |
1755 | if (nd->root.mnt) { | |
1756 | path_put(&nd->root); | |
1757 | nd->root.mnt = NULL; | |
1758 | } | |
1759 | } | |
1760 | ||
1761 | if (likely(!retval)) { | |
1762 | if (unlikely(!audit_dummy_context())) { | |
1763 | if (nd->path.dentry && nd->inode) | |
1764 | audit_inode(name, nd->path.dentry); | |
1765 | } | |
1766 | } | |
1767 | ||
170aa3d0 | 1768 | return retval; |
1da177e4 LT |
1769 | } |
1770 | ||
fc9b52cd | 1771 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1772 | struct nameidata *nd) |
1773 | { | |
1774 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1775 | } | |
1776 | ||
d1811465 AV |
1777 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1778 | { | |
1779 | struct nameidata nd; | |
1780 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1781 | if (!res) | |
1782 | *path = nd.path; | |
1783 | return res; | |
1784 | } | |
1785 | ||
16f18200 JJS |
1786 | /** |
1787 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1788 | * @dentry: pointer to dentry of the base directory | |
1789 | * @mnt: pointer to vfs mount of the base directory | |
1790 | * @name: pointer to file name | |
1791 | * @flags: lookup flags | |
1792 | * @nd: pointer to nameidata | |
1793 | */ | |
1794 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1795 | const char *name, unsigned int flags, | |
1796 | struct nameidata *nd) | |
1797 | { | |
1798 | int retval; | |
1799 | ||
1800 | /* same as do_path_lookup */ | |
1801 | nd->last_type = LAST_ROOT; | |
1802 | nd->flags = flags; | |
1803 | nd->depth = 0; | |
1804 | ||
c8e7f449 JB |
1805 | nd->path.dentry = dentry; |
1806 | nd->path.mnt = mnt; | |
1807 | path_get(&nd->path); | |
5b857119 AV |
1808 | nd->root = nd->path; |
1809 | path_get(&nd->root); | |
31e6b01f | 1810 | nd->inode = nd->path.dentry->d_inode; |
16f18200 JJS |
1811 | |
1812 | retval = path_walk(name, nd); | |
4ac91378 | 1813 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
31e6b01f | 1814 | nd->inode)) |
4ac91378 | 1815 | audit_inode(name, nd->path.dentry); |
16f18200 | 1816 | |
5b857119 AV |
1817 | path_put(&nd->root); |
1818 | nd->root.mnt = NULL; | |
16f18200 | 1819 | |
2a737871 | 1820 | return retval; |
16f18200 JJS |
1821 | } |
1822 | ||
eead1911 CH |
1823 | static struct dentry *__lookup_hash(struct qstr *name, |
1824 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1825 | { |
81fca444 | 1826 | struct inode *inode = base->d_inode; |
057f6c01 | 1827 | struct dentry *dentry; |
1da177e4 LT |
1828 | int err; |
1829 | ||
b74c79e9 | 1830 | err = exec_permission(inode, 0); |
81fca444 CH |
1831 | if (err) |
1832 | return ERR_PTR(err); | |
1da177e4 LT |
1833 | |
1834 | /* | |
1835 | * See if the low-level filesystem might want | |
1836 | * to use its own hash.. | |
1837 | */ | |
fb045adb | 1838 | if (base->d_flags & DCACHE_OP_HASH) { |
b1e6a015 | 1839 | err = base->d_op->d_hash(base, inode, name); |
1da177e4 LT |
1840 | dentry = ERR_PTR(err); |
1841 | if (err < 0) | |
1842 | goto out; | |
1843 | } | |
1844 | ||
b04f784e NP |
1845 | /* |
1846 | * Don't bother with __d_lookup: callers are for creat as | |
1847 | * well as unlink, so a lot of the time it would cost | |
1848 | * a double lookup. | |
6e6b1bd1 | 1849 | */ |
b04f784e | 1850 | dentry = d_lookup(base, name); |
6e6b1bd1 | 1851 | |
fb045adb | 1852 | if (dentry && (dentry->d_flags & DCACHE_OP_REVALIDATE)) |
6e6b1bd1 AV |
1853 | dentry = do_revalidate(dentry, nd); |
1854 | ||
baa03890 NP |
1855 | if (!dentry) |
1856 | dentry = d_alloc_and_lookup(base, name, nd); | |
1da177e4 LT |
1857 | out: |
1858 | return dentry; | |
1859 | } | |
1860 | ||
057f6c01 JM |
1861 | /* |
1862 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1863 | * needs parent already locked. Doesn't follow mounts. | |
1864 | * SMP-safe. | |
1865 | */ | |
eead1911 | 1866 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1867 | { |
4ac91378 | 1868 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1869 | } |
1870 | ||
eead1911 CH |
1871 | static int __lookup_one_len(const char *name, struct qstr *this, |
1872 | struct dentry *base, int len) | |
1da177e4 LT |
1873 | { |
1874 | unsigned long hash; | |
1da177e4 LT |
1875 | unsigned int c; |
1876 | ||
057f6c01 JM |
1877 | this->name = name; |
1878 | this->len = len; | |
1da177e4 | 1879 | if (!len) |
057f6c01 | 1880 | return -EACCES; |
1da177e4 LT |
1881 | |
1882 | hash = init_name_hash(); | |
1883 | while (len--) { | |
1884 | c = *(const unsigned char *)name++; | |
1885 | if (c == '/' || c == '\0') | |
057f6c01 | 1886 | return -EACCES; |
1da177e4 LT |
1887 | hash = partial_name_hash(c, hash); |
1888 | } | |
057f6c01 JM |
1889 | this->hash = end_name_hash(hash); |
1890 | return 0; | |
1891 | } | |
1da177e4 | 1892 | |
eead1911 | 1893 | /** |
a6b91919 | 1894 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1895 | * @name: pathname component to lookup |
1896 | * @base: base directory to lookup from | |
1897 | * @len: maximum length @len should be interpreted to | |
1898 | * | |
a6b91919 RD |
1899 | * Note that this routine is purely a helper for filesystem usage and should |
1900 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1901 | * nameidata argument is passed to the filesystem methods and a filesystem |
1902 | * using this helper needs to be prepared for that. | |
1903 | */ | |
057f6c01 JM |
1904 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1905 | { | |
1906 | int err; | |
1907 | struct qstr this; | |
1908 | ||
2f9092e1 DW |
1909 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1910 | ||
057f6c01 | 1911 | err = __lookup_one_len(name, &this, base, len); |
eead1911 CH |
1912 | if (err) |
1913 | return ERR_PTR(err); | |
1914 | ||
49705b77 | 1915 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1916 | } |
1917 | ||
2d8f3038 AV |
1918 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1919 | struct path *path) | |
1da177e4 | 1920 | { |
2d8f3038 | 1921 | struct nameidata nd; |
1da177e4 LT |
1922 | char *tmp = getname(name); |
1923 | int err = PTR_ERR(tmp); | |
1da177e4 | 1924 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1925 | |
1926 | BUG_ON(flags & LOOKUP_PARENT); | |
1927 | ||
1928 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1929 | putname(tmp); |
2d8f3038 AV |
1930 | if (!err) |
1931 | *path = nd.path; | |
1da177e4 LT |
1932 | } |
1933 | return err; | |
1934 | } | |
1935 | ||
2ad94ae6 AV |
1936 | static int user_path_parent(int dfd, const char __user *path, |
1937 | struct nameidata *nd, char **name) | |
1938 | { | |
1939 | char *s = getname(path); | |
1940 | int error; | |
1941 | ||
1942 | if (IS_ERR(s)) | |
1943 | return PTR_ERR(s); | |
1944 | ||
1945 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1946 | if (error) | |
1947 | putname(s); | |
1948 | else | |
1949 | *name = s; | |
1950 | ||
1951 | return error; | |
1952 | } | |
1953 | ||
1da177e4 LT |
1954 | /* |
1955 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1956 | * minimal. | |
1957 | */ | |
1958 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1959 | { | |
da9592ed DH |
1960 | uid_t fsuid = current_fsuid(); |
1961 | ||
1da177e4 LT |
1962 | if (!(dir->i_mode & S_ISVTX)) |
1963 | return 0; | |
da9592ed | 1964 | if (inode->i_uid == fsuid) |
1da177e4 | 1965 | return 0; |
da9592ed | 1966 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1967 | return 0; |
1968 | return !capable(CAP_FOWNER); | |
1969 | } | |
1970 | ||
1971 | /* | |
1972 | * Check whether we can remove a link victim from directory dir, check | |
1973 | * whether the type of victim is right. | |
1974 | * 1. We can't do it if dir is read-only (done in permission()) | |
1975 | * 2. We should have write and exec permissions on dir | |
1976 | * 3. We can't remove anything from append-only dir | |
1977 | * 4. We can't do anything with immutable dir (done in permission()) | |
1978 | * 5. If the sticky bit on dir is set we should either | |
1979 | * a. be owner of dir, or | |
1980 | * b. be owner of victim, or | |
1981 | * c. have CAP_FOWNER capability | |
1982 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1983 | * links pointing to it. | |
1984 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1985 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1986 | * 9. We can't remove a root or mountpoint. | |
1987 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1988 | * nfs_async_unlink(). | |
1989 | */ | |
858119e1 | 1990 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1991 | { |
1992 | int error; | |
1993 | ||
1994 | if (!victim->d_inode) | |
1995 | return -ENOENT; | |
1996 | ||
1997 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 1998 | audit_inode_child(victim, dir); |
1da177e4 | 1999 | |
f419a2e3 | 2000 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2001 | if (error) |
2002 | return error; | |
2003 | if (IS_APPEND(dir)) | |
2004 | return -EPERM; | |
2005 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 2006 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
2007 | return -EPERM; |
2008 | if (isdir) { | |
2009 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
2010 | return -ENOTDIR; | |
2011 | if (IS_ROOT(victim)) | |
2012 | return -EBUSY; | |
2013 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
2014 | return -EISDIR; | |
2015 | if (IS_DEADDIR(dir)) | |
2016 | return -ENOENT; | |
2017 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2018 | return -EBUSY; | |
2019 | return 0; | |
2020 | } | |
2021 | ||
2022 | /* Check whether we can create an object with dentry child in directory | |
2023 | * dir. | |
2024 | * 1. We can't do it if child already exists (open has special treatment for | |
2025 | * this case, but since we are inlined it's OK) | |
2026 | * 2. We can't do it if dir is read-only (done in permission()) | |
2027 | * 3. We should have write and exec permissions on dir | |
2028 | * 4. We can't do it if dir is immutable (done in permission()) | |
2029 | */ | |
a95164d9 | 2030 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
2031 | { |
2032 | if (child->d_inode) | |
2033 | return -EEXIST; | |
2034 | if (IS_DEADDIR(dir)) | |
2035 | return -ENOENT; | |
f419a2e3 | 2036 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2037 | } |
2038 | ||
1da177e4 LT |
2039 | /* |
2040 | * p1 and p2 should be directories on the same fs. | |
2041 | */ | |
2042 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2043 | { | |
2044 | struct dentry *p; | |
2045 | ||
2046 | if (p1 == p2) { | |
f2eace23 | 2047 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2048 | return NULL; |
2049 | } | |
2050 | ||
a11f3a05 | 2051 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2052 | |
e2761a11 OH |
2053 | p = d_ancestor(p2, p1); |
2054 | if (p) { | |
2055 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2056 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2057 | return p; | |
1da177e4 LT |
2058 | } |
2059 | ||
e2761a11 OH |
2060 | p = d_ancestor(p1, p2); |
2061 | if (p) { | |
2062 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2063 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2064 | return p; | |
1da177e4 LT |
2065 | } |
2066 | ||
f2eace23 IM |
2067 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
2068 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
2069 | return NULL; |
2070 | } | |
2071 | ||
2072 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2073 | { | |
1b1dcc1b | 2074 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2075 | if (p1 != p2) { |
1b1dcc1b | 2076 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2077 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2078 | } |
2079 | } | |
2080 | ||
2081 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
2082 | struct nameidata *nd) | |
2083 | { | |
a95164d9 | 2084 | int error = may_create(dir, dentry); |
1da177e4 LT |
2085 | |
2086 | if (error) | |
2087 | return error; | |
2088 | ||
acfa4380 | 2089 | if (!dir->i_op->create) |
1da177e4 LT |
2090 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2091 | mode &= S_IALLUGO; | |
2092 | mode |= S_IFREG; | |
2093 | error = security_inode_create(dir, dentry, mode); | |
2094 | if (error) | |
2095 | return error; | |
1da177e4 | 2096 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 2097 | if (!error) |
f38aa942 | 2098 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2099 | return error; |
2100 | } | |
2101 | ||
3fb64190 | 2102 | int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2103 | { |
3fb64190 | 2104 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2105 | struct inode *inode = dentry->d_inode; |
2106 | int error; | |
2107 | ||
2108 | if (!inode) | |
2109 | return -ENOENT; | |
2110 | ||
c8fe8f30 CH |
2111 | switch (inode->i_mode & S_IFMT) { |
2112 | case S_IFLNK: | |
1da177e4 | 2113 | return -ELOOP; |
c8fe8f30 CH |
2114 | case S_IFDIR: |
2115 | if (acc_mode & MAY_WRITE) | |
2116 | return -EISDIR; | |
2117 | break; | |
2118 | case S_IFBLK: | |
2119 | case S_IFCHR: | |
3fb64190 | 2120 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2121 | return -EACCES; |
c8fe8f30 CH |
2122 | /*FALLTHRU*/ |
2123 | case S_IFIFO: | |
2124 | case S_IFSOCK: | |
1da177e4 | 2125 | flag &= ~O_TRUNC; |
c8fe8f30 | 2126 | break; |
4a3fd211 | 2127 | } |
b41572e9 | 2128 | |
3fb64190 | 2129 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2130 | if (error) |
2131 | return error; | |
6146f0d5 | 2132 | |
1da177e4 LT |
2133 | /* |
2134 | * An append-only file must be opened in append mode for writing. | |
2135 | */ | |
2136 | if (IS_APPEND(inode)) { | |
8737c930 | 2137 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2138 | return -EPERM; |
1da177e4 | 2139 | if (flag & O_TRUNC) |
7715b521 | 2140 | return -EPERM; |
1da177e4 LT |
2141 | } |
2142 | ||
2143 | /* O_NOATIME can only be set by the owner or superuser */ | |
7715b521 AV |
2144 | if (flag & O_NOATIME && !is_owner_or_cap(inode)) |
2145 | return -EPERM; | |
1da177e4 LT |
2146 | |
2147 | /* | |
2148 | * Ensure there are no outstanding leases on the file. | |
2149 | */ | |
b65a9cfc | 2150 | return break_lease(inode, flag); |
7715b521 | 2151 | } |
1da177e4 | 2152 | |
e1181ee6 | 2153 | static int handle_truncate(struct file *filp) |
7715b521 | 2154 | { |
e1181ee6 | 2155 | struct path *path = &filp->f_path; |
7715b521 AV |
2156 | struct inode *inode = path->dentry->d_inode; |
2157 | int error = get_write_access(inode); | |
2158 | if (error) | |
2159 | return error; | |
2160 | /* | |
2161 | * Refuse to truncate files with mandatory locks held on them. | |
2162 | */ | |
2163 | error = locks_verify_locked(inode); | |
2164 | if (!error) | |
ea0d3ab2 | 2165 | error = security_path_truncate(path); |
7715b521 AV |
2166 | if (!error) { |
2167 | error = do_truncate(path->dentry, 0, | |
2168 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2169 | filp); |
7715b521 AV |
2170 | } |
2171 | put_write_access(inode); | |
acd0c935 | 2172 | return error; |
1da177e4 LT |
2173 | } |
2174 | ||
d57999e1 DH |
2175 | /* |
2176 | * Be careful about ever adding any more callers of this | |
2177 | * function. Its flags must be in the namei format, not | |
2178 | * what get passed to sys_open(). | |
2179 | */ | |
2180 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
8737c930 | 2181 | int open_flag, int mode) |
aab520e2 DH |
2182 | { |
2183 | int error; | |
4ac91378 | 2184 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
2185 | |
2186 | if (!IS_POSIXACL(dir->d_inode)) | |
ce3b0f8d | 2187 | mode &= ~current_umask(); |
be6d3e56 KT |
2188 | error = security_path_mknod(&nd->path, path->dentry, mode, 0); |
2189 | if (error) | |
2190 | goto out_unlock; | |
aab520e2 | 2191 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); |
be6d3e56 | 2192 | out_unlock: |
aab520e2 | 2193 | mutex_unlock(&dir->d_inode->i_mutex); |
4ac91378 JB |
2194 | dput(nd->path.dentry); |
2195 | nd->path.dentry = path->dentry; | |
31e6b01f | 2196 | |
aab520e2 DH |
2197 | if (error) |
2198 | return error; | |
2199 | /* Don't check for write permission, don't truncate */ | |
8737c930 | 2200 | return may_open(&nd->path, 0, open_flag & ~O_TRUNC); |
aab520e2 DH |
2201 | } |
2202 | ||
d57999e1 DH |
2203 | /* |
2204 | * Note that while the flag value (low two bits) for sys_open means: | |
2205 | * 00 - read-only | |
2206 | * 01 - write-only | |
2207 | * 10 - read-write | |
2208 | * 11 - special | |
2209 | * it is changed into | |
2210 | * 00 - no permissions needed | |
2211 | * 01 - read-permission | |
2212 | * 10 - write-permission | |
2213 | * 11 - read-write | |
2214 | * for the internal routines (ie open_namei()/follow_link() etc) | |
2215 | * This is more logical, and also allows the 00 "no perm needed" | |
2216 | * to be used for symlinks (where the permissions are checked | |
2217 | * later). | |
2218 | * | |
2219 | */ | |
2220 | static inline int open_to_namei_flags(int flag) | |
2221 | { | |
2222 | if ((flag+1) & O_ACCMODE) | |
2223 | flag++; | |
2224 | return flag; | |
2225 | } | |
2226 | ||
7715b521 | 2227 | static int open_will_truncate(int flag, struct inode *inode) |
4a3fd211 DH |
2228 | { |
2229 | /* | |
2230 | * We'll never write to the fs underlying | |
2231 | * a device file. | |
2232 | */ | |
2233 | if (special_file(inode->i_mode)) | |
2234 | return 0; | |
2235 | return (flag & O_TRUNC); | |
2236 | } | |
2237 | ||
648fa861 | 2238 | static struct file *finish_open(struct nameidata *nd, |
9a66179e | 2239 | int open_flag, int acc_mode) |
648fa861 AV |
2240 | { |
2241 | struct file *filp; | |
2242 | int will_truncate; | |
2243 | int error; | |
2244 | ||
9a66179e | 2245 | will_truncate = open_will_truncate(open_flag, nd->path.dentry->d_inode); |
648fa861 AV |
2246 | if (will_truncate) { |
2247 | error = mnt_want_write(nd->path.mnt); | |
2248 | if (error) | |
2249 | goto exit; | |
2250 | } | |
2251 | error = may_open(&nd->path, acc_mode, open_flag); | |
2252 | if (error) { | |
2253 | if (will_truncate) | |
2254 | mnt_drop_write(nd->path.mnt); | |
2255 | goto exit; | |
2256 | } | |
2257 | filp = nameidata_to_filp(nd); | |
2258 | if (!IS_ERR(filp)) { | |
2259 | error = ima_file_check(filp, acc_mode); | |
2260 | if (error) { | |
2261 | fput(filp); | |
2262 | filp = ERR_PTR(error); | |
2263 | } | |
2264 | } | |
2265 | if (!IS_ERR(filp)) { | |
648fa861 | 2266 | if (will_truncate) { |
e1181ee6 | 2267 | error = handle_truncate(filp); |
648fa861 AV |
2268 | if (error) { |
2269 | fput(filp); | |
2270 | filp = ERR_PTR(error); | |
2271 | } | |
2272 | } | |
2273 | } | |
2274 | /* | |
2275 | * It is now safe to drop the mnt write | |
2276 | * because the filp has had a write taken | |
2277 | * on its behalf. | |
2278 | */ | |
2279 | if (will_truncate) | |
2280 | mnt_drop_write(nd->path.mnt); | |
d893f1bc | 2281 | path_put(&nd->path); |
648fa861 AV |
2282 | return filp; |
2283 | ||
2284 | exit: | |
648fa861 AV |
2285 | path_put(&nd->path); |
2286 | return ERR_PTR(error); | |
2287 | } | |
2288 | ||
31e6b01f NP |
2289 | /* |
2290 | * Handle O_CREAT case for do_filp_open | |
2291 | */ | |
fb1cc555 | 2292 | static struct file *do_last(struct nameidata *nd, struct path *path, |
5b369df8 | 2293 | int open_flag, int acc_mode, |
3e297b61 | 2294 | int mode, const char *pathname) |
fb1cc555 | 2295 | { |
a1e28038 | 2296 | struct dentry *dir = nd->path.dentry; |
fb1cc555 | 2297 | struct file *filp; |
1f36f774 AV |
2298 | int error = -EISDIR; |
2299 | ||
2300 | switch (nd->last_type) { | |
2301 | case LAST_DOTDOT: | |
2302 | follow_dotdot(nd); | |
2303 | dir = nd->path.dentry; | |
176306f5 | 2304 | case LAST_DOT: |
fb045adb | 2305 | if (need_reval_dot(dir)) { |
f20877d9 O |
2306 | int status = d_revalidate(nd->path.dentry, nd); |
2307 | if (!status) | |
2308 | status = -ESTALE; | |
2309 | if (status < 0) { | |
2310 | error = status; | |
1f36f774 | 2311 | goto exit; |
f20877d9 | 2312 | } |
1f36f774 AV |
2313 | } |
2314 | /* fallthrough */ | |
1f36f774 | 2315 | case LAST_ROOT: |
31e6b01f | 2316 | goto exit; |
1f36f774 AV |
2317 | case LAST_BIND: |
2318 | audit_inode(pathname, dir); | |
67ee3ad2 | 2319 | goto ok; |
1f36f774 | 2320 | } |
67ee3ad2 | 2321 | |
1f36f774 | 2322 | /* trailing slashes? */ |
31e6b01f NP |
2323 | if (nd->last.name[nd->last.len]) |
2324 | goto exit; | |
a2c36b45 | 2325 | |
a1e28038 AV |
2326 | mutex_lock(&dir->d_inode->i_mutex); |
2327 | ||
2328 | path->dentry = lookup_hash(nd); | |
2329 | path->mnt = nd->path.mnt; | |
2330 | ||
fb1cc555 AV |
2331 | error = PTR_ERR(path->dentry); |
2332 | if (IS_ERR(path->dentry)) { | |
2333 | mutex_unlock(&dir->d_inode->i_mutex); | |
2334 | goto exit; | |
2335 | } | |
2336 | ||
2337 | if (IS_ERR(nd->intent.open.file)) { | |
2338 | error = PTR_ERR(nd->intent.open.file); | |
2339 | goto exit_mutex_unlock; | |
2340 | } | |
2341 | ||
2342 | /* Negative dentry, just create the file */ | |
2343 | if (!path->dentry->d_inode) { | |
2344 | /* | |
2345 | * This write is needed to ensure that a | |
2346 | * ro->rw transition does not occur between | |
2347 | * the time when the file is created and when | |
2348 | * a permanent write count is taken through | |
2349 | * the 'struct file' in nameidata_to_filp(). | |
2350 | */ | |
2351 | error = mnt_want_write(nd->path.mnt); | |
2352 | if (error) | |
2353 | goto exit_mutex_unlock; | |
2354 | error = __open_namei_create(nd, path, open_flag, mode); | |
2355 | if (error) { | |
2356 | mnt_drop_write(nd->path.mnt); | |
2357 | goto exit; | |
2358 | } | |
2359 | filp = nameidata_to_filp(nd); | |
2360 | mnt_drop_write(nd->path.mnt); | |
d893f1bc | 2361 | path_put(&nd->path); |
fb1cc555 AV |
2362 | if (!IS_ERR(filp)) { |
2363 | error = ima_file_check(filp, acc_mode); | |
2364 | if (error) { | |
2365 | fput(filp); | |
2366 | filp = ERR_PTR(error); | |
2367 | } | |
2368 | } | |
2369 | return filp; | |
2370 | } | |
2371 | ||
2372 | /* | |
2373 | * It already exists. | |
2374 | */ | |
2375 | mutex_unlock(&dir->d_inode->i_mutex); | |
2376 | audit_inode(pathname, path->dentry); | |
2377 | ||
2378 | error = -EEXIST; | |
5b369df8 | 2379 | if (open_flag & O_EXCL) |
fb1cc555 AV |
2380 | goto exit_dput; |
2381 | ||
9875cf80 DH |
2382 | error = follow_managed(path, nd->flags); |
2383 | if (error < 0) | |
2384 | goto exit_dput; | |
fb1cc555 AV |
2385 | |
2386 | error = -ENOENT; | |
2387 | if (!path->dentry->d_inode) | |
2388 | goto exit_dput; | |
9e67f361 AV |
2389 | |
2390 | if (path->dentry->d_inode->i_op->follow_link) | |
fb1cc555 | 2391 | return NULL; |
fb1cc555 AV |
2392 | |
2393 | path_to_nameidata(path, nd); | |
31e6b01f | 2394 | nd->inode = path->dentry->d_inode; |
fb1cc555 | 2395 | error = -EISDIR; |
31e6b01f | 2396 | if (S_ISDIR(nd->inode->i_mode)) |
fb1cc555 | 2397 | goto exit; |
67ee3ad2 | 2398 | ok: |
9a66179e | 2399 | filp = finish_open(nd, open_flag, acc_mode); |
fb1cc555 AV |
2400 | return filp; |
2401 | ||
2402 | exit_mutex_unlock: | |
2403 | mutex_unlock(&dir->d_inode->i_mutex); | |
2404 | exit_dput: | |
2405 | path_put_conditional(path, nd); | |
2406 | exit: | |
fb1cc555 AV |
2407 | path_put(&nd->path); |
2408 | return ERR_PTR(error); | |
2409 | } | |
2410 | ||
1da177e4 | 2411 | /* |
4a3fd211 DH |
2412 | * Note that the low bits of the passed in "open_flag" |
2413 | * are not the same as in the local variable "flag". See | |
2414 | * open_to_namei_flags() for more details. | |
1da177e4 | 2415 | */ |
a70e65df | 2416 | struct file *do_filp_open(int dfd, const char *pathname, |
6e8341a1 | 2417 | int open_flag, int mode, int acc_mode) |
1da177e4 | 2418 | { |
4a3fd211 | 2419 | struct file *filp; |
a70e65df | 2420 | struct nameidata nd; |
6e8341a1 | 2421 | int error; |
9850c056 | 2422 | struct path path; |
1da177e4 | 2423 | int count = 0; |
d57999e1 | 2424 | int flag = open_to_namei_flags(open_flag); |
31e6b01f | 2425 | int flags; |
1f36f774 AV |
2426 | |
2427 | if (!(open_flag & O_CREAT)) | |
2428 | mode = 0; | |
1da177e4 | 2429 | |
b1085ba8 LS |
2430 | /* Must never be set by userspace */ |
2431 | open_flag &= ~FMODE_NONOTIFY; | |
2432 | ||
6b2f3d1f CH |
2433 | /* |
2434 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
2435 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
2436 | * always set instead of having to deal with possibly weird behaviour | |
2437 | * for malicious applications setting only __O_SYNC. | |
2438 | */ | |
2439 | if (open_flag & __O_SYNC) | |
2440 | open_flag |= O_DSYNC; | |
2441 | ||
6e8341a1 | 2442 | if (!acc_mode) |
6d125529 | 2443 | acc_mode = MAY_OPEN | ACC_MODE(open_flag); |
1da177e4 | 2444 | |
834f2a4a | 2445 | /* O_TRUNC implies we need access checks for write permissions */ |
4296e2cb | 2446 | if (open_flag & O_TRUNC) |
834f2a4a TM |
2447 | acc_mode |= MAY_WRITE; |
2448 | ||
1da177e4 LT |
2449 | /* Allow the LSM permission hook to distinguish append |
2450 | access from general write access. */ | |
4296e2cb | 2451 | if (open_flag & O_APPEND) |
1da177e4 LT |
2452 | acc_mode |= MAY_APPEND; |
2453 | ||
31e6b01f NP |
2454 | flags = LOOKUP_OPEN; |
2455 | if (open_flag & O_CREAT) { | |
2456 | flags |= LOOKUP_CREATE; | |
2457 | if (open_flag & O_EXCL) | |
2458 | flags |= LOOKUP_EXCL; | |
2459 | } | |
2460 | if (open_flag & O_DIRECTORY) | |
2461 | flags |= LOOKUP_DIRECTORY; | |
2462 | if (!(open_flag & O_NOFOLLOW)) | |
2463 | flags |= LOOKUP_FOLLOW; | |
2464 | ||
2465 | filp = get_empty_filp(); | |
2466 | if (!filp) | |
2467 | return ERR_PTR(-ENFILE); | |
2468 | ||
2469 | filp->f_flags = open_flag; | |
2470 | nd.intent.open.file = filp; | |
2471 | nd.intent.open.flags = flag; | |
2472 | nd.intent.open.create_mode = mode; | |
2473 | ||
2474 | if (open_flag & O_CREAT) | |
2475 | goto creat; | |
2476 | ||
2477 | /* !O_CREAT, simple open */ | |
2478 | error = do_path_lookup(dfd, pathname, flags, &nd); | |
2479 | if (unlikely(error)) | |
2480 | goto out_filp; | |
2481 | error = -ELOOP; | |
2482 | if (!(nd.flags & LOOKUP_FOLLOW)) { | |
2483 | if (nd.inode->i_op->follow_link) | |
2484 | goto out_path; | |
2485 | } | |
2486 | error = -ENOTDIR; | |
2487 | if (nd.flags & LOOKUP_DIRECTORY) { | |
2488 | if (!nd.inode->i_op->lookup) | |
2489 | goto out_path; | |
2490 | } | |
2491 | audit_inode(pathname, nd.path.dentry); | |
2492 | filp = finish_open(&nd, open_flag, acc_mode); | |
2dab5974 | 2493 | release_open_intent(&nd); |
31e6b01f NP |
2494 | return filp; |
2495 | ||
2496 | creat: | |
2497 | /* OK, have to create the file. Find the parent. */ | |
2498 | error = path_init_rcu(dfd, pathname, | |
2499 | LOOKUP_PARENT | (flags & LOOKUP_REVAL), &nd); | |
1da177e4 | 2500 | if (error) |
31e6b01f NP |
2501 | goto out_filp; |
2502 | error = path_walk_rcu(pathname, &nd); | |
2503 | path_finish_rcu(&nd); | |
2504 | if (unlikely(error == -ECHILD || error == -ESTALE)) { | |
2505 | /* slower, locked walk */ | |
2506 | if (error == -ESTALE) { | |
2507 | reval: | |
2508 | flags |= LOOKUP_REVAL; | |
2509 | } | |
2510 | error = path_init(dfd, pathname, | |
2511 | LOOKUP_PARENT | (flags & LOOKUP_REVAL), &nd); | |
2512 | if (error) | |
2513 | goto out_filp; | |
3866248e | 2514 | |
31e6b01f | 2515 | error = path_walk_simple(pathname, &nd); |
654f562c | 2516 | } |
31e6b01f NP |
2517 | if (unlikely(error)) |
2518 | goto out_filp; | |
2519 | if (unlikely(!audit_dummy_context())) | |
9b4a9b14 | 2520 | audit_inode(pathname, nd.path.dentry); |
1da177e4 LT |
2521 | |
2522 | /* | |
a2c36b45 | 2523 | * We have the parent and last component. |
1da177e4 | 2524 | */ |
31e6b01f | 2525 | nd.flags = flags; |
3e297b61 | 2526 | filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); |
806b681c | 2527 | while (unlikely(!filp)) { /* trailing symlink */ |
7b9337aa NP |
2528 | struct path link = path; |
2529 | struct inode *linki = link.dentry->d_inode; | |
def4af30 | 2530 | void *cookie; |
806b681c | 2531 | error = -ELOOP; |
db372915 | 2532 | if (!(nd.flags & LOOKUP_FOLLOW)) |
1f36f774 AV |
2533 | goto exit_dput; |
2534 | if (count++ == 32) | |
806b681c AV |
2535 | goto exit_dput; |
2536 | /* | |
2537 | * This is subtle. Instead of calling do_follow_link() we do | |
2538 | * the thing by hands. The reason is that this way we have zero | |
2539 | * link_count and path_walk() (called from ->follow_link) | |
2540 | * honoring LOOKUP_PARENT. After that we have the parent and | |
2541 | * last component, i.e. we are in the same situation as after | |
2542 | * the first path_walk(). Well, almost - if the last component | |
2543 | * is normal we get its copy stored in nd->last.name and we will | |
2544 | * have to putname() it when we are done. Procfs-like symlinks | |
2545 | * just set LAST_BIND. | |
2546 | */ | |
2547 | nd.flags |= LOOKUP_PARENT; | |
7b9337aa | 2548 | error = security_inode_follow_link(link.dentry, &nd); |
806b681c AV |
2549 | if (error) |
2550 | goto exit_dput; | |
7b9337aa | 2551 | error = __do_follow_link(&link, &nd, &cookie); |
def4af30 | 2552 | if (unlikely(error)) { |
7b9337aa NP |
2553 | if (!IS_ERR(cookie) && linki->i_op->put_link) |
2554 | linki->i_op->put_link(link.dentry, &nd, cookie); | |
806b681c | 2555 | /* nd.path had been dropped */ |
7b9337aa | 2556 | nd.path = link; |
31e6b01f | 2557 | goto out_path; |
806b681c AV |
2558 | } |
2559 | nd.flags &= ~LOOKUP_PARENT; | |
3e297b61 | 2560 | filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); |
7b9337aa NP |
2561 | if (linki->i_op->put_link) |
2562 | linki->i_op->put_link(link.dentry, &nd, cookie); | |
2563 | path_put(&link); | |
806b681c | 2564 | } |
10fa8e62 | 2565 | out: |
2a737871 AV |
2566 | if (nd.root.mnt) |
2567 | path_put(&nd.root); | |
31e6b01f | 2568 | if (filp == ERR_PTR(-ESTALE) && !(flags & LOOKUP_REVAL)) |
10fa8e62 | 2569 | goto reval; |
2dab5974 | 2570 | release_open_intent(&nd); |
10fa8e62 | 2571 | return filp; |
1da177e4 | 2572 | |
806b681c AV |
2573 | exit_dput: |
2574 | path_put_conditional(&path, &nd); | |
31e6b01f NP |
2575 | out_path: |
2576 | path_put(&nd.path); | |
2577 | out_filp: | |
806b681c | 2578 | filp = ERR_PTR(error); |
10fa8e62 | 2579 | goto out; |
1da177e4 LT |
2580 | } |
2581 | ||
a70e65df CH |
2582 | /** |
2583 | * filp_open - open file and return file pointer | |
2584 | * | |
2585 | * @filename: path to open | |
2586 | * @flags: open flags as per the open(2) second argument | |
2587 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
2588 | * | |
2589 | * This is the helper to open a file from kernelspace if you really | |
2590 | * have to. But in generally you should not do this, so please move | |
2591 | * along, nothing to see here.. | |
2592 | */ | |
2593 | struct file *filp_open(const char *filename, int flags, int mode) | |
2594 | { | |
6e8341a1 | 2595 | return do_filp_open(AT_FDCWD, filename, flags, mode, 0); |
a70e65df CH |
2596 | } |
2597 | EXPORT_SYMBOL(filp_open); | |
2598 | ||
1da177e4 LT |
2599 | /** |
2600 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
2601 | * @nd: nameidata info | |
2602 | * @is_dir: directory flag | |
2603 | * | |
2604 | * Simple function to lookup and return a dentry and create it | |
2605 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 2606 | * |
4ac91378 | 2607 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
2608 | */ |
2609 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
2610 | { | |
c663e5d8 | 2611 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 2612 | |
4ac91378 | 2613 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
2614 | /* |
2615 | * Yucky last component or no last component at all? | |
2616 | * (foo/., foo/.., /////) | |
2617 | */ | |
1da177e4 LT |
2618 | if (nd->last_type != LAST_NORM) |
2619 | goto fail; | |
2620 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 2621 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 2622 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
2623 | |
2624 | /* | |
2625 | * Do the final lookup. | |
2626 | */ | |
49705b77 | 2627 | dentry = lookup_hash(nd); |
1da177e4 LT |
2628 | if (IS_ERR(dentry)) |
2629 | goto fail; | |
c663e5d8 | 2630 | |
e9baf6e5 AV |
2631 | if (dentry->d_inode) |
2632 | goto eexist; | |
c663e5d8 CH |
2633 | /* |
2634 | * Special case - lookup gave negative, but... we had foo/bar/ | |
2635 | * From the vfs_mknod() POV we just have a negative dentry - | |
2636 | * all is fine. Let's be bastards - you had / on the end, you've | |
2637 | * been asking for (non-existent) directory. -ENOENT for you. | |
2638 | */ | |
e9baf6e5 AV |
2639 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
2640 | dput(dentry); | |
2641 | dentry = ERR_PTR(-ENOENT); | |
2642 | } | |
1da177e4 | 2643 | return dentry; |
e9baf6e5 | 2644 | eexist: |
1da177e4 | 2645 | dput(dentry); |
e9baf6e5 | 2646 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
2647 | fail: |
2648 | return dentry; | |
2649 | } | |
f81a0bff | 2650 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
2651 | |
2652 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
2653 | { | |
a95164d9 | 2654 | int error = may_create(dir, dentry); |
1da177e4 LT |
2655 | |
2656 | if (error) | |
2657 | return error; | |
2658 | ||
2659 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
2660 | return -EPERM; | |
2661 | ||
acfa4380 | 2662 | if (!dir->i_op->mknod) |
1da177e4 LT |
2663 | return -EPERM; |
2664 | ||
08ce5f16 SH |
2665 | error = devcgroup_inode_mknod(mode, dev); |
2666 | if (error) | |
2667 | return error; | |
2668 | ||
1da177e4 LT |
2669 | error = security_inode_mknod(dir, dentry, mode, dev); |
2670 | if (error) | |
2671 | return error; | |
2672 | ||
1da177e4 | 2673 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 2674 | if (!error) |
f38aa942 | 2675 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2676 | return error; |
2677 | } | |
2678 | ||
463c3197 DH |
2679 | static int may_mknod(mode_t mode) |
2680 | { | |
2681 | switch (mode & S_IFMT) { | |
2682 | case S_IFREG: | |
2683 | case S_IFCHR: | |
2684 | case S_IFBLK: | |
2685 | case S_IFIFO: | |
2686 | case S_IFSOCK: | |
2687 | case 0: /* zero mode translates to S_IFREG */ | |
2688 | return 0; | |
2689 | case S_IFDIR: | |
2690 | return -EPERM; | |
2691 | default: | |
2692 | return -EINVAL; | |
2693 | } | |
2694 | } | |
2695 | ||
2e4d0924 HC |
2696 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
2697 | unsigned, dev) | |
1da177e4 | 2698 | { |
2ad94ae6 AV |
2699 | int error; |
2700 | char *tmp; | |
2701 | struct dentry *dentry; | |
1da177e4 LT |
2702 | struct nameidata nd; |
2703 | ||
2704 | if (S_ISDIR(mode)) | |
2705 | return -EPERM; | |
1da177e4 | 2706 | |
2ad94ae6 | 2707 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 2708 | if (error) |
2ad94ae6 AV |
2709 | return error; |
2710 | ||
1da177e4 | 2711 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2712 | if (IS_ERR(dentry)) { |
2713 | error = PTR_ERR(dentry); | |
2714 | goto out_unlock; | |
2715 | } | |
4ac91378 | 2716 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2717 | mode &= ~current_umask(); |
463c3197 DH |
2718 | error = may_mknod(mode); |
2719 | if (error) | |
2720 | goto out_dput; | |
2721 | error = mnt_want_write(nd.path.mnt); | |
2722 | if (error) | |
2723 | goto out_dput; | |
be6d3e56 KT |
2724 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2725 | if (error) | |
2726 | goto out_drop_write; | |
463c3197 | 2727 | switch (mode & S_IFMT) { |
1da177e4 | 2728 | case 0: case S_IFREG: |
4ac91378 | 2729 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2730 | break; |
2731 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2732 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2733 | new_decode_dev(dev)); |
2734 | break; | |
2735 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2736 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2737 | break; |
1da177e4 | 2738 | } |
be6d3e56 | 2739 | out_drop_write: |
463c3197 DH |
2740 | mnt_drop_write(nd.path.mnt); |
2741 | out_dput: | |
2742 | dput(dentry); | |
2743 | out_unlock: | |
4ac91378 | 2744 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2745 | path_put(&nd.path); |
1da177e4 LT |
2746 | putname(tmp); |
2747 | ||
2748 | return error; | |
2749 | } | |
2750 | ||
3480b257 | 2751 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2752 | { |
2753 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2754 | } | |
2755 | ||
1da177e4 LT |
2756 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2757 | { | |
a95164d9 | 2758 | int error = may_create(dir, dentry); |
1da177e4 LT |
2759 | |
2760 | if (error) | |
2761 | return error; | |
2762 | ||
acfa4380 | 2763 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2764 | return -EPERM; |
2765 | ||
2766 | mode &= (S_IRWXUGO|S_ISVTX); | |
2767 | error = security_inode_mkdir(dir, dentry, mode); | |
2768 | if (error) | |
2769 | return error; | |
2770 | ||
1da177e4 | 2771 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2772 | if (!error) |
f38aa942 | 2773 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2774 | return error; |
2775 | } | |
2776 | ||
2e4d0924 | 2777 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2778 | { |
2779 | int error = 0; | |
2780 | char * tmp; | |
6902d925 DH |
2781 | struct dentry *dentry; |
2782 | struct nameidata nd; | |
1da177e4 | 2783 | |
2ad94ae6 AV |
2784 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2785 | if (error) | |
6902d925 | 2786 | goto out_err; |
1da177e4 | 2787 | |
6902d925 DH |
2788 | dentry = lookup_create(&nd, 1); |
2789 | error = PTR_ERR(dentry); | |
2790 | if (IS_ERR(dentry)) | |
2791 | goto out_unlock; | |
1da177e4 | 2792 | |
4ac91378 | 2793 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2794 | mode &= ~current_umask(); |
463c3197 DH |
2795 | error = mnt_want_write(nd.path.mnt); |
2796 | if (error) | |
2797 | goto out_dput; | |
be6d3e56 KT |
2798 | error = security_path_mkdir(&nd.path, dentry, mode); |
2799 | if (error) | |
2800 | goto out_drop_write; | |
4ac91378 | 2801 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2802 | out_drop_write: |
463c3197 DH |
2803 | mnt_drop_write(nd.path.mnt); |
2804 | out_dput: | |
6902d925 DH |
2805 | dput(dentry); |
2806 | out_unlock: | |
4ac91378 | 2807 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2808 | path_put(&nd.path); |
6902d925 DH |
2809 | putname(tmp); |
2810 | out_err: | |
1da177e4 LT |
2811 | return error; |
2812 | } | |
2813 | ||
3cdad428 | 2814 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2815 | { |
2816 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2817 | } | |
2818 | ||
1da177e4 LT |
2819 | /* |
2820 | * We try to drop the dentry early: we should have | |
2821 | * a usage count of 2 if we're the only user of this | |
2822 | * dentry, and if that is true (possibly after pruning | |
2823 | * the dcache), then we drop the dentry now. | |
2824 | * | |
2825 | * A low-level filesystem can, if it choses, legally | |
2826 | * do a | |
2827 | * | |
2828 | * if (!d_unhashed(dentry)) | |
2829 | * return -EBUSY; | |
2830 | * | |
2831 | * if it cannot handle the case of removing a directory | |
2832 | * that is still in use by something else.. | |
2833 | */ | |
2834 | void dentry_unhash(struct dentry *dentry) | |
2835 | { | |
2836 | dget(dentry); | |
dc168427 | 2837 | shrink_dcache_parent(dentry); |
1da177e4 | 2838 | spin_lock(&dentry->d_lock); |
b7ab39f6 | 2839 | if (dentry->d_count == 2) |
1da177e4 LT |
2840 | __d_drop(dentry); |
2841 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
2842 | } |
2843 | ||
2844 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2845 | { | |
2846 | int error = may_delete(dir, dentry, 1); | |
2847 | ||
2848 | if (error) | |
2849 | return error; | |
2850 | ||
acfa4380 | 2851 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2852 | return -EPERM; |
2853 | ||
1b1dcc1b | 2854 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2855 | dentry_unhash(dentry); |
2856 | if (d_mountpoint(dentry)) | |
2857 | error = -EBUSY; | |
2858 | else { | |
2859 | error = security_inode_rmdir(dir, dentry); | |
2860 | if (!error) { | |
2861 | error = dir->i_op->rmdir(dir, dentry); | |
d83c49f3 | 2862 | if (!error) { |
1da177e4 | 2863 | dentry->d_inode->i_flags |= S_DEAD; |
d83c49f3 AV |
2864 | dont_mount(dentry); |
2865 | } | |
1da177e4 LT |
2866 | } |
2867 | } | |
1b1dcc1b | 2868 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2869 | if (!error) { |
1da177e4 LT |
2870 | d_delete(dentry); |
2871 | } | |
2872 | dput(dentry); | |
2873 | ||
2874 | return error; | |
2875 | } | |
2876 | ||
5590ff0d | 2877 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2878 | { |
2879 | int error = 0; | |
2880 | char * name; | |
2881 | struct dentry *dentry; | |
2882 | struct nameidata nd; | |
2883 | ||
2ad94ae6 | 2884 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2885 | if (error) |
2ad94ae6 | 2886 | return error; |
1da177e4 LT |
2887 | |
2888 | switch(nd.last_type) { | |
0612d9fb OH |
2889 | case LAST_DOTDOT: |
2890 | error = -ENOTEMPTY; | |
2891 | goto exit1; | |
2892 | case LAST_DOT: | |
2893 | error = -EINVAL; | |
2894 | goto exit1; | |
2895 | case LAST_ROOT: | |
2896 | error = -EBUSY; | |
2897 | goto exit1; | |
1da177e4 | 2898 | } |
0612d9fb OH |
2899 | |
2900 | nd.flags &= ~LOOKUP_PARENT; | |
2901 | ||
4ac91378 | 2902 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2903 | dentry = lookup_hash(&nd); |
1da177e4 | 2904 | error = PTR_ERR(dentry); |
6902d925 DH |
2905 | if (IS_ERR(dentry)) |
2906 | goto exit2; | |
0622753b DH |
2907 | error = mnt_want_write(nd.path.mnt); |
2908 | if (error) | |
2909 | goto exit3; | |
be6d3e56 KT |
2910 | error = security_path_rmdir(&nd.path, dentry); |
2911 | if (error) | |
2912 | goto exit4; | |
4ac91378 | 2913 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2914 | exit4: |
0622753b DH |
2915 | mnt_drop_write(nd.path.mnt); |
2916 | exit3: | |
6902d925 DH |
2917 | dput(dentry); |
2918 | exit2: | |
4ac91378 | 2919 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2920 | exit1: |
1d957f9b | 2921 | path_put(&nd.path); |
1da177e4 LT |
2922 | putname(name); |
2923 | return error; | |
2924 | } | |
2925 | ||
3cdad428 | 2926 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2927 | { |
2928 | return do_rmdir(AT_FDCWD, pathname); | |
2929 | } | |
2930 | ||
1da177e4 LT |
2931 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2932 | { | |
2933 | int error = may_delete(dir, dentry, 0); | |
2934 | ||
2935 | if (error) | |
2936 | return error; | |
2937 | ||
acfa4380 | 2938 | if (!dir->i_op->unlink) |
1da177e4 LT |
2939 | return -EPERM; |
2940 | ||
1b1dcc1b | 2941 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2942 | if (d_mountpoint(dentry)) |
2943 | error = -EBUSY; | |
2944 | else { | |
2945 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 2946 | if (!error) { |
1da177e4 | 2947 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 2948 | if (!error) |
d83c49f3 | 2949 | dont_mount(dentry); |
bec1052e | 2950 | } |
1da177e4 | 2951 | } |
1b1dcc1b | 2952 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2953 | |
2954 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2955 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2956 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2957 | d_delete(dentry); |
1da177e4 | 2958 | } |
0eeca283 | 2959 | |
1da177e4 LT |
2960 | return error; |
2961 | } | |
2962 | ||
2963 | /* | |
2964 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2965 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2966 | * writeout happening, and we don't want to prevent access to the directory |
2967 | * while waiting on the I/O. | |
2968 | */ | |
5590ff0d | 2969 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2970 | { |
2ad94ae6 AV |
2971 | int error; |
2972 | char *name; | |
1da177e4 LT |
2973 | struct dentry *dentry; |
2974 | struct nameidata nd; | |
2975 | struct inode *inode = NULL; | |
2976 | ||
2ad94ae6 | 2977 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2978 | if (error) |
2ad94ae6 AV |
2979 | return error; |
2980 | ||
1da177e4 LT |
2981 | error = -EISDIR; |
2982 | if (nd.last_type != LAST_NORM) | |
2983 | goto exit1; | |
0612d9fb OH |
2984 | |
2985 | nd.flags &= ~LOOKUP_PARENT; | |
2986 | ||
4ac91378 | 2987 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2988 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2989 | error = PTR_ERR(dentry); |
2990 | if (!IS_ERR(dentry)) { | |
2991 | /* Why not before? Because we want correct error value */ | |
2992 | if (nd.last.name[nd.last.len]) | |
2993 | goto slashes; | |
2994 | inode = dentry->d_inode; | |
2995 | if (inode) | |
7de9c6ee | 2996 | ihold(inode); |
0622753b DH |
2997 | error = mnt_want_write(nd.path.mnt); |
2998 | if (error) | |
2999 | goto exit2; | |
be6d3e56 KT |
3000 | error = security_path_unlink(&nd.path, dentry); |
3001 | if (error) | |
3002 | goto exit3; | |
4ac91378 | 3003 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 3004 | exit3: |
0622753b | 3005 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
3006 | exit2: |
3007 | dput(dentry); | |
3008 | } | |
4ac91378 | 3009 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
3010 | if (inode) |
3011 | iput(inode); /* truncate the inode here */ | |
3012 | exit1: | |
1d957f9b | 3013 | path_put(&nd.path); |
1da177e4 LT |
3014 | putname(name); |
3015 | return error; | |
3016 | ||
3017 | slashes: | |
3018 | error = !dentry->d_inode ? -ENOENT : | |
3019 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
3020 | goto exit2; | |
3021 | } | |
3022 | ||
2e4d0924 | 3023 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3024 | { |
3025 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3026 | return -EINVAL; | |
3027 | ||
3028 | if (flag & AT_REMOVEDIR) | |
3029 | return do_rmdir(dfd, pathname); | |
3030 | ||
3031 | return do_unlinkat(dfd, pathname); | |
3032 | } | |
3033 | ||
3480b257 | 3034 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
3035 | { |
3036 | return do_unlinkat(AT_FDCWD, pathname); | |
3037 | } | |
3038 | ||
db2e747b | 3039 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3040 | { |
a95164d9 | 3041 | int error = may_create(dir, dentry); |
1da177e4 LT |
3042 | |
3043 | if (error) | |
3044 | return error; | |
3045 | ||
acfa4380 | 3046 | if (!dir->i_op->symlink) |
1da177e4 LT |
3047 | return -EPERM; |
3048 | ||
3049 | error = security_inode_symlink(dir, dentry, oldname); | |
3050 | if (error) | |
3051 | return error; | |
3052 | ||
1da177e4 | 3053 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3054 | if (!error) |
f38aa942 | 3055 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3056 | return error; |
3057 | } | |
3058 | ||
2e4d0924 HC |
3059 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
3060 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3061 | { |
2ad94ae6 AV |
3062 | int error; |
3063 | char *from; | |
3064 | char *to; | |
6902d925 DH |
3065 | struct dentry *dentry; |
3066 | struct nameidata nd; | |
1da177e4 LT |
3067 | |
3068 | from = getname(oldname); | |
2ad94ae6 | 3069 | if (IS_ERR(from)) |
1da177e4 | 3070 | return PTR_ERR(from); |
1da177e4 | 3071 | |
2ad94ae6 | 3072 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 3073 | if (error) |
2ad94ae6 AV |
3074 | goto out_putname; |
3075 | ||
6902d925 DH |
3076 | dentry = lookup_create(&nd, 0); |
3077 | error = PTR_ERR(dentry); | |
3078 | if (IS_ERR(dentry)) | |
3079 | goto out_unlock; | |
3080 | ||
75c3f29d DH |
3081 | error = mnt_want_write(nd.path.mnt); |
3082 | if (error) | |
3083 | goto out_dput; | |
be6d3e56 KT |
3084 | error = security_path_symlink(&nd.path, dentry, from); |
3085 | if (error) | |
3086 | goto out_drop_write; | |
db2e747b | 3087 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 3088 | out_drop_write: |
75c3f29d DH |
3089 | mnt_drop_write(nd.path.mnt); |
3090 | out_dput: | |
6902d925 DH |
3091 | dput(dentry); |
3092 | out_unlock: | |
4ac91378 | 3093 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 3094 | path_put(&nd.path); |
6902d925 DH |
3095 | putname(to); |
3096 | out_putname: | |
1da177e4 LT |
3097 | putname(from); |
3098 | return error; | |
3099 | } | |
3100 | ||
3480b257 | 3101 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3102 | { |
3103 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
3104 | } | |
3105 | ||
1da177e4 LT |
3106 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
3107 | { | |
3108 | struct inode *inode = old_dentry->d_inode; | |
3109 | int error; | |
3110 | ||
3111 | if (!inode) | |
3112 | return -ENOENT; | |
3113 | ||
a95164d9 | 3114 | error = may_create(dir, new_dentry); |
1da177e4 LT |
3115 | if (error) |
3116 | return error; | |
3117 | ||
3118 | if (dir->i_sb != inode->i_sb) | |
3119 | return -EXDEV; | |
3120 | ||
3121 | /* | |
3122 | * A link to an append-only or immutable file cannot be created. | |
3123 | */ | |
3124 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
3125 | return -EPERM; | |
acfa4380 | 3126 | if (!dir->i_op->link) |
1da177e4 | 3127 | return -EPERM; |
7e79eedb | 3128 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
3129 | return -EPERM; |
3130 | ||
3131 | error = security_inode_link(old_dentry, dir, new_dentry); | |
3132 | if (error) | |
3133 | return error; | |
3134 | ||
7e79eedb | 3135 | mutex_lock(&inode->i_mutex); |
1da177e4 | 3136 | error = dir->i_op->link(old_dentry, dir, new_dentry); |
7e79eedb | 3137 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 3138 | if (!error) |
7e79eedb | 3139 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
3140 | return error; |
3141 | } | |
3142 | ||
3143 | /* | |
3144 | * Hardlinks are often used in delicate situations. We avoid | |
3145 | * security-related surprises by not following symlinks on the | |
3146 | * newname. --KAB | |
3147 | * | |
3148 | * We don't follow them on the oldname either to be compatible | |
3149 | * with linux 2.0, and to avoid hard-linking to directories | |
3150 | * and other special files. --ADM | |
3151 | */ | |
2e4d0924 HC |
3152 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
3153 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
3154 | { |
3155 | struct dentry *new_dentry; | |
2d8f3038 AV |
3156 | struct nameidata nd; |
3157 | struct path old_path; | |
1da177e4 | 3158 | int error; |
2ad94ae6 | 3159 | char *to; |
1da177e4 | 3160 | |
45c9b11a | 3161 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
3162 | return -EINVAL; |
3163 | ||
2d8f3038 AV |
3164 | error = user_path_at(olddfd, oldname, |
3165 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
3166 | &old_path); | |
1da177e4 | 3167 | if (error) |
2ad94ae6 AV |
3168 | return error; |
3169 | ||
3170 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
3171 | if (error) |
3172 | goto out; | |
3173 | error = -EXDEV; | |
2d8f3038 | 3174 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
3175 | goto out_release; |
3176 | new_dentry = lookup_create(&nd, 0); | |
3177 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
3178 | if (IS_ERR(new_dentry)) |
3179 | goto out_unlock; | |
75c3f29d DH |
3180 | error = mnt_want_write(nd.path.mnt); |
3181 | if (error) | |
3182 | goto out_dput; | |
be6d3e56 KT |
3183 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
3184 | if (error) | |
3185 | goto out_drop_write; | |
2d8f3038 | 3186 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 3187 | out_drop_write: |
75c3f29d DH |
3188 | mnt_drop_write(nd.path.mnt); |
3189 | out_dput: | |
6902d925 DH |
3190 | dput(new_dentry); |
3191 | out_unlock: | |
4ac91378 | 3192 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 3193 | out_release: |
1d957f9b | 3194 | path_put(&nd.path); |
2ad94ae6 | 3195 | putname(to); |
1da177e4 | 3196 | out: |
2d8f3038 | 3197 | path_put(&old_path); |
1da177e4 LT |
3198 | |
3199 | return error; | |
3200 | } | |
3201 | ||
3480b257 | 3202 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 3203 | { |
c04030e1 | 3204 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
3205 | } |
3206 | ||
1da177e4 LT |
3207 | /* |
3208 | * The worst of all namespace operations - renaming directory. "Perverted" | |
3209 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
3210 | * Problems: | |
3211 | * a) we can get into loop creation. Check is done in is_subdir(). | |
3212 | * b) race potential - two innocent renames can create a loop together. | |
3213 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 3214 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
3215 | * story. |
3216 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 3217 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
3218 | * whether the target exists). Solution: try to be smart with locking |
3219 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 3220 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
3221 | * move will be locked. Thus we can rank directories by the tree |
3222 | * (ancestors first) and rank all non-directories after them. | |
3223 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 3224 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
3225 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
3226 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
3227 | * we'd better make sure that there's no link(2) for them. | |
3228 | * d) some filesystems don't support opened-but-unlinked directories, | |
3229 | * either because of layout or because they are not ready to deal with | |
3230 | * all cases correctly. The latter will be fixed (taking this sort of | |
3231 | * stuff into VFS), but the former is not going away. Solution: the same | |
3232 | * trick as in rmdir(). | |
3233 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 3234 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 3235 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 3236 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
3237 | * locking]. |
3238 | */ | |
75c96f85 AB |
3239 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
3240 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3241 | { |
3242 | int error = 0; | |
3243 | struct inode *target; | |
3244 | ||
3245 | /* | |
3246 | * If we are going to change the parent - check write permissions, | |
3247 | * we'll need to flip '..'. | |
3248 | */ | |
3249 | if (new_dir != old_dir) { | |
f419a2e3 | 3250 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3251 | if (error) |
3252 | return error; | |
3253 | } | |
3254 | ||
3255 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3256 | if (error) | |
3257 | return error; | |
3258 | ||
3259 | target = new_dentry->d_inode; | |
d83c49f3 | 3260 | if (target) |
1b1dcc1b | 3261 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
3262 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
3263 | error = -EBUSY; | |
d83c49f3 AV |
3264 | else { |
3265 | if (target) | |
3266 | dentry_unhash(new_dentry); | |
1da177e4 | 3267 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
d83c49f3 | 3268 | } |
1da177e4 | 3269 | if (target) { |
d83c49f3 | 3270 | if (!error) { |
1da177e4 | 3271 | target->i_flags |= S_DEAD; |
d83c49f3 AV |
3272 | dont_mount(new_dentry); |
3273 | } | |
1b1dcc1b | 3274 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3275 | if (d_unhashed(new_dentry)) |
3276 | d_rehash(new_dentry); | |
3277 | dput(new_dentry); | |
3278 | } | |
e31e14ec | 3279 | if (!error) |
349457cc MF |
3280 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3281 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3282 | return error; |
3283 | } | |
3284 | ||
75c96f85 AB |
3285 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3286 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3287 | { |
3288 | struct inode *target; | |
3289 | int error; | |
3290 | ||
3291 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3292 | if (error) | |
3293 | return error; | |
3294 | ||
3295 | dget(new_dentry); | |
3296 | target = new_dentry->d_inode; | |
3297 | if (target) | |
1b1dcc1b | 3298 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
3299 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
3300 | error = -EBUSY; | |
3301 | else | |
3302 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3303 | if (!error) { | |
bec1052e | 3304 | if (target) |
d83c49f3 | 3305 | dont_mount(new_dentry); |
349457cc | 3306 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 3307 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
3308 | } |
3309 | if (target) | |
1b1dcc1b | 3310 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3311 | dput(new_dentry); |
3312 | return error; | |
3313 | } | |
3314 | ||
3315 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3316 | struct inode *new_dir, struct dentry *new_dentry) | |
3317 | { | |
3318 | int error; | |
3319 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3320 | const unsigned char *old_name; |
1da177e4 LT |
3321 | |
3322 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3323 | return 0; | |
3324 | ||
3325 | error = may_delete(old_dir, old_dentry, is_dir); | |
3326 | if (error) | |
3327 | return error; | |
3328 | ||
3329 | if (!new_dentry->d_inode) | |
a95164d9 | 3330 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3331 | else |
3332 | error = may_delete(new_dir, new_dentry, is_dir); | |
3333 | if (error) | |
3334 | return error; | |
3335 | ||
acfa4380 | 3336 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3337 | return -EPERM; |
3338 | ||
0eeca283 RL |
3339 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3340 | ||
1da177e4 LT |
3341 | if (is_dir) |
3342 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3343 | else | |
3344 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3345 | if (!error) |
3346 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3347 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3348 | fsnotify_oldname_free(old_name); |
3349 | ||
1da177e4 LT |
3350 | return error; |
3351 | } | |
3352 | ||
2e4d0924 HC |
3353 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3354 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3355 | { |
2ad94ae6 AV |
3356 | struct dentry *old_dir, *new_dir; |
3357 | struct dentry *old_dentry, *new_dentry; | |
3358 | struct dentry *trap; | |
1da177e4 | 3359 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3360 | char *from; |
3361 | char *to; | |
3362 | int error; | |
1da177e4 | 3363 | |
2ad94ae6 | 3364 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3365 | if (error) |
3366 | goto exit; | |
3367 | ||
2ad94ae6 | 3368 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3369 | if (error) |
3370 | goto exit1; | |
3371 | ||
3372 | error = -EXDEV; | |
4ac91378 | 3373 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3374 | goto exit2; |
3375 | ||
4ac91378 | 3376 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3377 | error = -EBUSY; |
3378 | if (oldnd.last_type != LAST_NORM) | |
3379 | goto exit2; | |
3380 | ||
4ac91378 | 3381 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3382 | if (newnd.last_type != LAST_NORM) |
3383 | goto exit2; | |
3384 | ||
0612d9fb OH |
3385 | oldnd.flags &= ~LOOKUP_PARENT; |
3386 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3387 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3388 | |
1da177e4 LT |
3389 | trap = lock_rename(new_dir, old_dir); |
3390 | ||
49705b77 | 3391 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3392 | error = PTR_ERR(old_dentry); |
3393 | if (IS_ERR(old_dentry)) | |
3394 | goto exit3; | |
3395 | /* source must exist */ | |
3396 | error = -ENOENT; | |
3397 | if (!old_dentry->d_inode) | |
3398 | goto exit4; | |
3399 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3400 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3401 | error = -ENOTDIR; | |
3402 | if (oldnd.last.name[oldnd.last.len]) | |
3403 | goto exit4; | |
3404 | if (newnd.last.name[newnd.last.len]) | |
3405 | goto exit4; | |
3406 | } | |
3407 | /* source should not be ancestor of target */ | |
3408 | error = -EINVAL; | |
3409 | if (old_dentry == trap) | |
3410 | goto exit4; | |
49705b77 | 3411 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3412 | error = PTR_ERR(new_dentry); |
3413 | if (IS_ERR(new_dentry)) | |
3414 | goto exit4; | |
3415 | /* target should not be an ancestor of source */ | |
3416 | error = -ENOTEMPTY; | |
3417 | if (new_dentry == trap) | |
3418 | goto exit5; | |
3419 | ||
9079b1eb DH |
3420 | error = mnt_want_write(oldnd.path.mnt); |
3421 | if (error) | |
3422 | goto exit5; | |
be6d3e56 KT |
3423 | error = security_path_rename(&oldnd.path, old_dentry, |
3424 | &newnd.path, new_dentry); | |
3425 | if (error) | |
3426 | goto exit6; | |
1da177e4 LT |
3427 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3428 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 3429 | exit6: |
9079b1eb | 3430 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
3431 | exit5: |
3432 | dput(new_dentry); | |
3433 | exit4: | |
3434 | dput(old_dentry); | |
3435 | exit3: | |
3436 | unlock_rename(new_dir, old_dir); | |
3437 | exit2: | |
1d957f9b | 3438 | path_put(&newnd.path); |
2ad94ae6 | 3439 | putname(to); |
1da177e4 | 3440 | exit1: |
1d957f9b | 3441 | path_put(&oldnd.path); |
1da177e4 | 3442 | putname(from); |
2ad94ae6 | 3443 | exit: |
1da177e4 LT |
3444 | return error; |
3445 | } | |
3446 | ||
a26eab24 | 3447 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3448 | { |
3449 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3450 | } | |
3451 | ||
1da177e4 LT |
3452 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3453 | { | |
3454 | int len; | |
3455 | ||
3456 | len = PTR_ERR(link); | |
3457 | if (IS_ERR(link)) | |
3458 | goto out; | |
3459 | ||
3460 | len = strlen(link); | |
3461 | if (len > (unsigned) buflen) | |
3462 | len = buflen; | |
3463 | if (copy_to_user(buffer, link, len)) | |
3464 | len = -EFAULT; | |
3465 | out: | |
3466 | return len; | |
3467 | } | |
3468 | ||
3469 | /* | |
3470 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3471 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3472 | * using) it for any given inode is up to filesystem. | |
3473 | */ | |
3474 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3475 | { | |
3476 | struct nameidata nd; | |
cc314eef | 3477 | void *cookie; |
694a1764 | 3478 | int res; |
cc314eef | 3479 | |
1da177e4 | 3480 | nd.depth = 0; |
cc314eef | 3481 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3482 | if (IS_ERR(cookie)) |
3483 | return PTR_ERR(cookie); | |
3484 | ||
3485 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3486 | if (dentry->d_inode->i_op->put_link) | |
3487 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3488 | return res; | |
1da177e4 LT |
3489 | } |
3490 | ||
3491 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3492 | { | |
3493 | return __vfs_follow_link(nd, link); | |
3494 | } | |
3495 | ||
3496 | /* get the link contents into pagecache */ | |
3497 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3498 | { | |
ebd09abb DG |
3499 | char *kaddr; |
3500 | struct page *page; | |
1da177e4 | 3501 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3502 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3503 | if (IS_ERR(page)) |
6fe6900e | 3504 | return (char*)page; |
1da177e4 | 3505 | *ppage = page; |
ebd09abb DG |
3506 | kaddr = kmap(page); |
3507 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3508 | return kaddr; | |
1da177e4 LT |
3509 | } |
3510 | ||
3511 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3512 | { | |
3513 | struct page *page = NULL; | |
3514 | char *s = page_getlink(dentry, &page); | |
3515 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3516 | if (page) { | |
3517 | kunmap(page); | |
3518 | page_cache_release(page); | |
3519 | } | |
3520 | return res; | |
3521 | } | |
3522 | ||
cc314eef | 3523 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3524 | { |
cc314eef | 3525 | struct page *page = NULL; |
1da177e4 | 3526 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3527 | return page; |
1da177e4 LT |
3528 | } |
3529 | ||
cc314eef | 3530 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3531 | { |
cc314eef LT |
3532 | struct page *page = cookie; |
3533 | ||
3534 | if (page) { | |
1da177e4 LT |
3535 | kunmap(page); |
3536 | page_cache_release(page); | |
1da177e4 LT |
3537 | } |
3538 | } | |
3539 | ||
54566b2c NP |
3540 | /* |
3541 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3542 | */ | |
3543 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3544 | { |
3545 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3546 | struct page *page; |
afddba49 | 3547 | void *fsdata; |
beb497ab | 3548 | int err; |
1da177e4 | 3549 | char *kaddr; |
54566b2c NP |
3550 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3551 | if (nofs) | |
3552 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3553 | |
7e53cac4 | 3554 | retry: |
afddba49 | 3555 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3556 | flags, &page, &fsdata); |
1da177e4 | 3557 | if (err) |
afddba49 NP |
3558 | goto fail; |
3559 | ||
1da177e4 LT |
3560 | kaddr = kmap_atomic(page, KM_USER0); |
3561 | memcpy(kaddr, symname, len-1); | |
3562 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
3563 | |
3564 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3565 | page, fsdata); | |
1da177e4 LT |
3566 | if (err < 0) |
3567 | goto fail; | |
afddba49 NP |
3568 | if (err < len-1) |
3569 | goto retry; | |
3570 | ||
1da177e4 LT |
3571 | mark_inode_dirty(inode); |
3572 | return 0; | |
1da177e4 LT |
3573 | fail: |
3574 | return err; | |
3575 | } | |
3576 | ||
0adb25d2 KK |
3577 | int page_symlink(struct inode *inode, const char *symname, int len) |
3578 | { | |
3579 | return __page_symlink(inode, symname, len, | |
54566b2c | 3580 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3581 | } |
3582 | ||
92e1d5be | 3583 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3584 | .readlink = generic_readlink, |
3585 | .follow_link = page_follow_link_light, | |
3586 | .put_link = page_put_link, | |
3587 | }; | |
3588 | ||
2d8f3038 | 3589 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3590 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3591 | EXPORT_SYMBOL(follow_down); |
3592 | EXPORT_SYMBOL(follow_up); | |
3593 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
3594 | EXPORT_SYMBOL(getname); | |
3595 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3596 | EXPORT_SYMBOL(lookup_one_len); |
3597 | EXPORT_SYMBOL(page_follow_link_light); | |
3598 | EXPORT_SYMBOL(page_put_link); | |
3599 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3600 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3601 | EXPORT_SYMBOL(page_symlink); |
3602 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
3603 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 3604 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3605 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3606 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 3607 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
3608 | EXPORT_SYMBOL(unlock_rename); |
3609 | EXPORT_SYMBOL(vfs_create); | |
3610 | EXPORT_SYMBOL(vfs_follow_link); | |
3611 | EXPORT_SYMBOL(vfs_link); | |
3612 | EXPORT_SYMBOL(vfs_mkdir); | |
3613 | EXPORT_SYMBOL(vfs_mknod); | |
3614 | EXPORT_SYMBOL(generic_permission); | |
3615 | EXPORT_SYMBOL(vfs_readlink); | |
3616 | EXPORT_SYMBOL(vfs_rename); | |
3617 | EXPORT_SYMBOL(vfs_rmdir); | |
3618 | EXPORT_SYMBOL(vfs_symlink); | |
3619 | EXPORT_SYMBOL(vfs_unlink); | |
3620 | EXPORT_SYMBOL(dentry_unhash); | |
3621 | EXPORT_SYMBOL(generic_readlink); |