]>
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 | |
b74c79e9 | 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 | ||
b3e19d92 NP |
370 | /** |
371 | * path_get_long - get a long reference to a path | |
372 | * @path: path to get the reference to | |
373 | * | |
374 | * Given a path increment the reference count to the dentry and the vfsmount. | |
375 | */ | |
376 | void path_get_long(struct path *path) | |
377 | { | |
378 | mntget_long(path->mnt); | |
379 | dget(path->dentry); | |
380 | } | |
381 | ||
1d957f9b JB |
382 | /** |
383 | * path_put - put a reference to a path | |
384 | * @path: path to put the reference to | |
385 | * | |
386 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
387 | */ | |
388 | void path_put(struct path *path) | |
1da177e4 | 389 | { |
1d957f9b JB |
390 | dput(path->dentry); |
391 | mntput(path->mnt); | |
1da177e4 | 392 | } |
1d957f9b | 393 | EXPORT_SYMBOL(path_put); |
1da177e4 | 394 | |
b3e19d92 NP |
395 | /** |
396 | * path_put_long - put a long reference to a path | |
397 | * @path: path to put the reference to | |
398 | * | |
399 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
400 | */ | |
401 | void path_put_long(struct path *path) | |
402 | { | |
403 | dput(path->dentry); | |
404 | mntput_long(path->mnt); | |
405 | } | |
406 | ||
31e6b01f NP |
407 | /** |
408 | * nameidata_drop_rcu - drop this nameidata out of rcu-walk | |
409 | * @nd: nameidata pathwalk data to drop | |
410 | * @Returns: 0 on success, -ECHLID on failure | |
411 | * | |
412 | * Path walking has 2 modes, rcu-walk and ref-walk (see | |
413 | * Documentation/filesystems/path-lookup.txt). __drop_rcu* functions attempt | |
414 | * to drop out of rcu-walk mode and take normal reference counts on dentries | |
415 | * and vfsmounts to transition to rcu-walk mode. __drop_rcu* functions take | |
416 | * refcounts at the last known good point before rcu-walk got stuck, so | |
417 | * ref-walk may continue from there. If this is not successful (eg. a seqcount | |
418 | * has changed), then failure is returned and path walk restarts from the | |
419 | * beginning in ref-walk mode. | |
420 | * | |
421 | * nameidata_drop_rcu attempts to drop the current nd->path and nd->root into | |
422 | * ref-walk. Must be called from rcu-walk context. | |
423 | */ | |
424 | static int nameidata_drop_rcu(struct nameidata *nd) | |
425 | { | |
426 | struct fs_struct *fs = current->fs; | |
427 | struct dentry *dentry = nd->path.dentry; | |
428 | ||
429 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
430 | if (nd->root.mnt) { | |
431 | spin_lock(&fs->lock); | |
432 | if (nd->root.mnt != fs->root.mnt || | |
433 | nd->root.dentry != fs->root.dentry) | |
434 | goto err_root; | |
435 | } | |
436 | spin_lock(&dentry->d_lock); | |
437 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
438 | goto err; | |
439 | BUG_ON(nd->inode != dentry->d_inode); | |
440 | spin_unlock(&dentry->d_lock); | |
441 | if (nd->root.mnt) { | |
442 | path_get(&nd->root); | |
443 | spin_unlock(&fs->lock); | |
444 | } | |
445 | mntget(nd->path.mnt); | |
446 | ||
447 | rcu_read_unlock(); | |
448 | br_read_unlock(vfsmount_lock); | |
449 | nd->flags &= ~LOOKUP_RCU; | |
450 | return 0; | |
451 | err: | |
452 | spin_unlock(&dentry->d_lock); | |
453 | err_root: | |
454 | if (nd->root.mnt) | |
455 | spin_unlock(&fs->lock); | |
456 | return -ECHILD; | |
457 | } | |
458 | ||
459 | /* Try to drop out of rcu-walk mode if we were in it, otherwise do nothing. */ | |
460 | static inline int nameidata_drop_rcu_maybe(struct nameidata *nd) | |
461 | { | |
462 | if (nd->flags & LOOKUP_RCU) | |
463 | return nameidata_drop_rcu(nd); | |
464 | return 0; | |
465 | } | |
466 | ||
467 | /** | |
468 | * nameidata_dentry_drop_rcu - drop nameidata and dentry out of rcu-walk | |
469 | * @nd: nameidata pathwalk data to drop | |
470 | * @dentry: dentry to drop | |
471 | * @Returns: 0 on success, -ECHLID on failure | |
472 | * | |
473 | * nameidata_dentry_drop_rcu attempts to drop the current nd->path and nd->root, | |
474 | * and dentry into ref-walk. @dentry must be a path found by a do_lookup call on | |
475 | * @nd. Must be called from rcu-walk context. | |
476 | */ | |
477 | static int nameidata_dentry_drop_rcu(struct nameidata *nd, struct dentry *dentry) | |
478 | { | |
479 | struct fs_struct *fs = current->fs; | |
480 | struct dentry *parent = nd->path.dentry; | |
481 | ||
482 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
483 | if (nd->root.mnt) { | |
484 | spin_lock(&fs->lock); | |
485 | if (nd->root.mnt != fs->root.mnt || | |
486 | nd->root.dentry != fs->root.dentry) | |
487 | goto err_root; | |
488 | } | |
489 | spin_lock(&parent->d_lock); | |
490 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); | |
491 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
492 | goto err; | |
493 | /* | |
494 | * If the sequence check on the child dentry passed, then the child has | |
495 | * not been removed from its parent. This means the parent dentry must | |
496 | * be valid and able to take a reference at this point. | |
497 | */ | |
498 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
499 | BUG_ON(!parent->d_count); | |
500 | parent->d_count++; | |
501 | spin_unlock(&dentry->d_lock); | |
502 | spin_unlock(&parent->d_lock); | |
503 | if (nd->root.mnt) { | |
504 | path_get(&nd->root); | |
505 | spin_unlock(&fs->lock); | |
506 | } | |
507 | mntget(nd->path.mnt); | |
508 | ||
509 | rcu_read_unlock(); | |
510 | br_read_unlock(vfsmount_lock); | |
511 | nd->flags &= ~LOOKUP_RCU; | |
512 | return 0; | |
513 | err: | |
514 | spin_unlock(&dentry->d_lock); | |
515 | spin_unlock(&parent->d_lock); | |
516 | err_root: | |
517 | if (nd->root.mnt) | |
518 | spin_unlock(&fs->lock); | |
519 | return -ECHILD; | |
520 | } | |
521 | ||
522 | /* Try to drop out of rcu-walk mode if we were in it, otherwise do nothing. */ | |
523 | static inline int nameidata_dentry_drop_rcu_maybe(struct nameidata *nd, struct dentry *dentry) | |
524 | { | |
525 | if (nd->flags & LOOKUP_RCU) | |
526 | return nameidata_dentry_drop_rcu(nd, dentry); | |
527 | return 0; | |
528 | } | |
529 | ||
530 | /** | |
531 | * nameidata_drop_rcu_last - drop nameidata ending path walk out of rcu-walk | |
532 | * @nd: nameidata pathwalk data to drop | |
533 | * @Returns: 0 on success, -ECHLID on failure | |
534 | * | |
535 | * nameidata_drop_rcu_last attempts to drop the current nd->path into ref-walk. | |
536 | * nd->path should be the final element of the lookup, so nd->root is discarded. | |
537 | * Must be called from rcu-walk context. | |
538 | */ | |
539 | static int nameidata_drop_rcu_last(struct nameidata *nd) | |
540 | { | |
541 | struct dentry *dentry = nd->path.dentry; | |
542 | ||
543 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
544 | nd->flags &= ~LOOKUP_RCU; | |
545 | nd->root.mnt = NULL; | |
546 | spin_lock(&dentry->d_lock); | |
547 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
548 | goto err_unlock; | |
549 | BUG_ON(nd->inode != dentry->d_inode); | |
550 | spin_unlock(&dentry->d_lock); | |
551 | ||
552 | mntget(nd->path.mnt); | |
553 | ||
554 | rcu_read_unlock(); | |
555 | br_read_unlock(vfsmount_lock); | |
556 | ||
557 | return 0; | |
558 | ||
559 | err_unlock: | |
560 | spin_unlock(&dentry->d_lock); | |
561 | rcu_read_unlock(); | |
562 | br_read_unlock(vfsmount_lock); | |
563 | return -ECHILD; | |
564 | } | |
565 | ||
566 | /* Try to drop out of rcu-walk mode if we were in it, otherwise do nothing. */ | |
567 | static inline int nameidata_drop_rcu_last_maybe(struct nameidata *nd) | |
568 | { | |
569 | if (likely(nd->flags & LOOKUP_RCU)) | |
570 | return nameidata_drop_rcu_last(nd); | |
571 | return 0; | |
572 | } | |
573 | ||
834f2a4a TM |
574 | /** |
575 | * release_open_intent - free up open intent resources | |
576 | * @nd: pointer to nameidata | |
577 | */ | |
578 | void release_open_intent(struct nameidata *nd) | |
579 | { | |
0f7fc9e4 | 580 | if (nd->intent.open.file->f_path.dentry == NULL) |
834f2a4a TM |
581 | put_filp(nd->intent.open.file); |
582 | else | |
583 | fput(nd->intent.open.file); | |
584 | } | |
585 | ||
34286d66 NP |
586 | static int d_revalidate(struct dentry *dentry, struct nameidata *nd) |
587 | { | |
588 | int status; | |
589 | ||
590 | status = dentry->d_op->d_revalidate(dentry, nd); | |
591 | if (status == -ECHILD) { | |
592 | if (nameidata_dentry_drop_rcu(nd, dentry)) | |
593 | return status; | |
594 | status = dentry->d_op->d_revalidate(dentry, nd); | |
595 | } | |
596 | ||
597 | return status; | |
598 | } | |
599 | ||
bcdc5e01 IK |
600 | static inline struct dentry * |
601 | do_revalidate(struct dentry *dentry, struct nameidata *nd) | |
602 | { | |
34286d66 NP |
603 | int status; |
604 | ||
605 | status = d_revalidate(dentry, nd); | |
bcdc5e01 IK |
606 | if (unlikely(status <= 0)) { |
607 | /* | |
608 | * The dentry failed validation. | |
609 | * If d_revalidate returned 0 attempt to invalidate | |
610 | * the dentry otherwise d_revalidate is asking us | |
611 | * to return a fail status. | |
612 | */ | |
34286d66 NP |
613 | if (status < 0) { |
614 | /* If we're in rcu-walk, we don't have a ref */ | |
615 | if (!(nd->flags & LOOKUP_RCU)) | |
616 | dput(dentry); | |
617 | dentry = ERR_PTR(status); | |
618 | ||
619 | } else { | |
620 | /* Don't d_invalidate in rcu-walk mode */ | |
621 | if (nameidata_dentry_drop_rcu_maybe(nd, dentry)) | |
622 | return ERR_PTR(-ECHILD); | |
bcdc5e01 IK |
623 | if (!d_invalidate(dentry)) { |
624 | dput(dentry); | |
625 | dentry = NULL; | |
626 | } | |
bcdc5e01 IK |
627 | } |
628 | } | |
629 | return dentry; | |
630 | } | |
631 | ||
fb045adb NP |
632 | static inline int need_reval_dot(struct dentry *dentry) |
633 | { | |
634 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
635 | return 0; | |
636 | ||
637 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) | |
638 | return 0; | |
639 | ||
640 | return 1; | |
641 | } | |
642 | ||
39159de2 JL |
643 | /* |
644 | * force_reval_path - force revalidation of a dentry | |
645 | * | |
646 | * In some situations the path walking code will trust dentries without | |
647 | * revalidating them. This causes problems for filesystems that depend on | |
648 | * d_revalidate to handle file opens (e.g. NFSv4). When FS_REVAL_DOT is set | |
649 | * (which indicates that it's possible for the dentry to go stale), force | |
650 | * a d_revalidate call before proceeding. | |
651 | * | |
652 | * Returns 0 if the revalidation was successful. If the revalidation fails, | |
653 | * either return the error returned by d_revalidate or -ESTALE if the | |
654 | * revalidation it just returned 0. If d_revalidate returns 0, we attempt to | |
655 | * invalidate the dentry. It's up to the caller to handle putting references | |
656 | * to the path if necessary. | |
657 | */ | |
658 | static int | |
659 | force_reval_path(struct path *path, struct nameidata *nd) | |
660 | { | |
661 | int status; | |
662 | struct dentry *dentry = path->dentry; | |
663 | ||
664 | /* | |
665 | * only check on filesystems where it's possible for the dentry to | |
fb045adb | 666 | * become stale. |
39159de2 | 667 | */ |
fb045adb | 668 | if (!need_reval_dot(dentry)) |
39159de2 JL |
669 | return 0; |
670 | ||
34286d66 | 671 | status = d_revalidate(dentry, nd); |
39159de2 JL |
672 | if (status > 0) |
673 | return 0; | |
674 | ||
675 | if (!status) { | |
676 | d_invalidate(dentry); | |
677 | status = -ESTALE; | |
678 | } | |
679 | return status; | |
680 | } | |
681 | ||
1da177e4 | 682 | /* |
b75b5086 AV |
683 | * Short-cut version of permission(), for calling on directories |
684 | * during pathname resolution. Combines parts of permission() | |
685 | * and generic_permission(), and tests ONLY for MAY_EXEC permission. | |
1da177e4 LT |
686 | * |
687 | * If appropriate, check DAC only. If not appropriate, or | |
b75b5086 | 688 | * short-cut DAC fails, then call ->permission() to do more |
1da177e4 LT |
689 | * complete permission check. |
690 | */ | |
b74c79e9 | 691 | static inline int exec_permission(struct inode *inode, unsigned int flags) |
1da177e4 | 692 | { |
5909ccaa | 693 | int ret; |
1da177e4 | 694 | |
cb9179ea | 695 | if (inode->i_op->permission) { |
b74c79e9 NP |
696 | ret = inode->i_op->permission(inode, MAY_EXEC, flags); |
697 | } else { | |
698 | ret = acl_permission_check(inode, MAY_EXEC, flags, | |
699 | inode->i_op->check_acl); | |
cb9179ea | 700 | } |
b74c79e9 | 701 | if (likely(!ret)) |
1da177e4 | 702 | goto ok; |
b74c79e9 | 703 | if (ret == -ECHILD) |
31e6b01f | 704 | return ret; |
1da177e4 | 705 | |
f1ac9f6b | 706 | if (capable(CAP_DAC_OVERRIDE) || capable(CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
707 | goto ok; |
708 | ||
5909ccaa | 709 | return ret; |
1da177e4 | 710 | ok: |
b74c79e9 | 711 | return security_inode_exec_permission(inode, flags); |
1da177e4 LT |
712 | } |
713 | ||
2a737871 AV |
714 | static __always_inline void set_root(struct nameidata *nd) |
715 | { | |
f7ad3c6b MS |
716 | if (!nd->root.mnt) |
717 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
718 | } |
719 | ||
6de88d72 AV |
720 | static int link_path_walk(const char *, struct nameidata *); |
721 | ||
31e6b01f NP |
722 | static __always_inline void set_root_rcu(struct nameidata *nd) |
723 | { | |
724 | if (!nd->root.mnt) { | |
725 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
726 | unsigned seq; |
727 | ||
728 | do { | |
729 | seq = read_seqcount_begin(&fs->seq); | |
730 | nd->root = fs->root; | |
731 | } while (read_seqcount_retry(&fs->seq, seq)); | |
31e6b01f NP |
732 | } |
733 | } | |
734 | ||
f1662356 | 735 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 736 | { |
31e6b01f NP |
737 | int ret; |
738 | ||
1da177e4 LT |
739 | if (IS_ERR(link)) |
740 | goto fail; | |
741 | ||
742 | if (*link == '/') { | |
2a737871 | 743 | set_root(nd); |
1d957f9b | 744 | path_put(&nd->path); |
2a737871 AV |
745 | nd->path = nd->root; |
746 | path_get(&nd->root); | |
1da177e4 | 747 | } |
31e6b01f | 748 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 749 | |
31e6b01f NP |
750 | ret = link_path_walk(link, nd); |
751 | return ret; | |
1da177e4 | 752 | fail: |
1d957f9b | 753 | path_put(&nd->path); |
1da177e4 LT |
754 | return PTR_ERR(link); |
755 | } | |
756 | ||
1d957f9b | 757 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
758 | { |
759 | dput(path->dentry); | |
4ac91378 | 760 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
761 | mntput(path->mnt); |
762 | } | |
763 | ||
764 | static inline void path_to_nameidata(struct path *path, struct nameidata *nd) | |
765 | { | |
31e6b01f NP |
766 | if (!(nd->flags & LOOKUP_RCU)) { |
767 | dput(nd->path.dentry); | |
768 | if (nd->path.mnt != path->mnt) | |
769 | mntput(nd->path.mnt); | |
9a229683 | 770 | } |
31e6b01f | 771 | nd->path.mnt = path->mnt; |
4ac91378 | 772 | nd->path.dentry = path->dentry; |
051d3812 IK |
773 | } |
774 | ||
def4af30 AV |
775 | static __always_inline int |
776 | __do_follow_link(struct path *path, struct nameidata *nd, void **p) | |
1da177e4 LT |
777 | { |
778 | int error; | |
cd4e91d3 | 779 | struct dentry *dentry = path->dentry; |
1da177e4 | 780 | |
d671a1cb | 781 | touch_atime(path->mnt, dentry); |
1da177e4 | 782 | nd_set_link(nd, NULL); |
cd4e91d3 | 783 | |
4ac91378 | 784 | if (path->mnt != nd->path.mnt) { |
051d3812 | 785 | path_to_nameidata(path, nd); |
31e6b01f | 786 | nd->inode = nd->path.dentry->d_inode; |
051d3812 IK |
787 | dget(dentry); |
788 | } | |
789 | mntget(path->mnt); | |
31e6b01f | 790 | |
86acdca1 | 791 | nd->last_type = LAST_BIND; |
def4af30 AV |
792 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
793 | error = PTR_ERR(*p); | |
794 | if (!IS_ERR(*p)) { | |
1da177e4 | 795 | char *s = nd_get_link(nd); |
cc314eef | 796 | error = 0; |
1da177e4 LT |
797 | if (s) |
798 | error = __vfs_follow_link(nd, s); | |
39159de2 JL |
799 | else if (nd->last_type == LAST_BIND) { |
800 | error = force_reval_path(&nd->path, nd); | |
801 | if (error) | |
802 | path_put(&nd->path); | |
803 | } | |
1da177e4 | 804 | } |
1da177e4 LT |
805 | return error; |
806 | } | |
807 | ||
808 | /* | |
809 | * This limits recursive symlink follows to 8, while | |
810 | * limiting consecutive symlinks to 40. | |
811 | * | |
812 | * Without that kind of total limit, nasty chains of consecutive | |
813 | * symlinks can cause almost arbitrarily long lookups. | |
814 | */ | |
90ebe565 | 815 | static inline int do_follow_link(struct path *path, struct nameidata *nd) |
1da177e4 | 816 | { |
def4af30 | 817 | void *cookie; |
1da177e4 LT |
818 | int err = -ELOOP; |
819 | if (current->link_count >= MAX_NESTED_LINKS) | |
820 | goto loop; | |
821 | if (current->total_link_count >= 40) | |
822 | goto loop; | |
823 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); | |
824 | cond_resched(); | |
90ebe565 | 825 | err = security_inode_follow_link(path->dentry, nd); |
1da177e4 LT |
826 | if (err) |
827 | goto loop; | |
828 | current->link_count++; | |
829 | current->total_link_count++; | |
830 | nd->depth++; | |
def4af30 AV |
831 | err = __do_follow_link(path, nd, &cookie); |
832 | if (!IS_ERR(cookie) && path->dentry->d_inode->i_op->put_link) | |
833 | path->dentry->d_inode->i_op->put_link(path->dentry, nd, cookie); | |
258fa999 | 834 | path_put(path); |
839d9f93 AV |
835 | current->link_count--; |
836 | nd->depth--; | |
1da177e4 LT |
837 | return err; |
838 | loop: | |
1d957f9b JB |
839 | path_put_conditional(path, nd); |
840 | path_put(&nd->path); | |
1da177e4 LT |
841 | return err; |
842 | } | |
843 | ||
31e6b01f NP |
844 | static int follow_up_rcu(struct path *path) |
845 | { | |
846 | struct vfsmount *parent; | |
847 | struct dentry *mountpoint; | |
848 | ||
849 | parent = path->mnt->mnt_parent; | |
850 | if (parent == path->mnt) | |
851 | return 0; | |
852 | mountpoint = path->mnt->mnt_mountpoint; | |
853 | path->dentry = mountpoint; | |
854 | path->mnt = parent; | |
855 | return 1; | |
856 | } | |
857 | ||
bab77ebf | 858 | int follow_up(struct path *path) |
1da177e4 LT |
859 | { |
860 | struct vfsmount *parent; | |
861 | struct dentry *mountpoint; | |
99b7db7b NP |
862 | |
863 | br_read_lock(vfsmount_lock); | |
bab77ebf AV |
864 | parent = path->mnt->mnt_parent; |
865 | if (parent == path->mnt) { | |
99b7db7b | 866 | br_read_unlock(vfsmount_lock); |
1da177e4 LT |
867 | return 0; |
868 | } | |
869 | mntget(parent); | |
bab77ebf | 870 | mountpoint = dget(path->mnt->mnt_mountpoint); |
99b7db7b | 871 | br_read_unlock(vfsmount_lock); |
bab77ebf AV |
872 | dput(path->dentry); |
873 | path->dentry = mountpoint; | |
874 | mntput(path->mnt); | |
875 | path->mnt = parent; | |
1da177e4 LT |
876 | return 1; |
877 | } | |
878 | ||
b5c84bf6 NP |
879 | /* |
880 | * serialization is taken care of in namespace.c | |
1da177e4 | 881 | */ |
31e6b01f NP |
882 | static void __follow_mount_rcu(struct nameidata *nd, struct path *path, |
883 | struct inode **inode) | |
884 | { | |
885 | while (d_mountpoint(path->dentry)) { | |
886 | struct vfsmount *mounted; | |
887 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); | |
888 | if (!mounted) | |
889 | return; | |
890 | path->mnt = mounted; | |
891 | path->dentry = mounted->mnt_root; | |
892 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); | |
893 | *inode = path->dentry->d_inode; | |
894 | } | |
895 | } | |
896 | ||
463ffb2e AV |
897 | static int __follow_mount(struct path *path) |
898 | { | |
899 | int res = 0; | |
900 | while (d_mountpoint(path->dentry)) { | |
1c755af4 | 901 | struct vfsmount *mounted = lookup_mnt(path); |
463ffb2e AV |
902 | if (!mounted) |
903 | break; | |
904 | dput(path->dentry); | |
905 | if (res) | |
906 | mntput(path->mnt); | |
907 | path->mnt = mounted; | |
908 | path->dentry = dget(mounted->mnt_root); | |
909 | res = 1; | |
910 | } | |
911 | return res; | |
912 | } | |
913 | ||
79ed0226 | 914 | static void follow_mount(struct path *path) |
1da177e4 | 915 | { |
79ed0226 | 916 | while (d_mountpoint(path->dentry)) { |
1c755af4 | 917 | struct vfsmount *mounted = lookup_mnt(path); |
1da177e4 LT |
918 | if (!mounted) |
919 | break; | |
79ed0226 AV |
920 | dput(path->dentry); |
921 | mntput(path->mnt); | |
922 | path->mnt = mounted; | |
923 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 | 924 | } |
1da177e4 LT |
925 | } |
926 | ||
9393bd07 | 927 | int follow_down(struct path *path) |
1da177e4 LT |
928 | { |
929 | struct vfsmount *mounted; | |
930 | ||
1c755af4 | 931 | mounted = lookup_mnt(path); |
1da177e4 | 932 | if (mounted) { |
9393bd07 AV |
933 | dput(path->dentry); |
934 | mntput(path->mnt); | |
935 | path->mnt = mounted; | |
936 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
937 | return 1; |
938 | } | |
939 | return 0; | |
940 | } | |
941 | ||
31e6b01f NP |
942 | static int follow_dotdot_rcu(struct nameidata *nd) |
943 | { | |
944 | struct inode *inode = nd->inode; | |
945 | ||
946 | set_root_rcu(nd); | |
947 | ||
948 | while(1) { | |
949 | if (nd->path.dentry == nd->root.dentry && | |
950 | nd->path.mnt == nd->root.mnt) { | |
951 | break; | |
952 | } | |
953 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
954 | struct dentry *old = nd->path.dentry; | |
955 | struct dentry *parent = old->d_parent; | |
956 | unsigned seq; | |
957 | ||
958 | seq = read_seqcount_begin(&parent->d_seq); | |
959 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
960 | return -ECHILD; | |
961 | inode = parent->d_inode; | |
962 | nd->path.dentry = parent; | |
963 | nd->seq = seq; | |
964 | break; | |
965 | } | |
966 | if (!follow_up_rcu(&nd->path)) | |
967 | break; | |
968 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
969 | inode = nd->path.dentry->d_inode; | |
970 | } | |
971 | __follow_mount_rcu(nd, &nd->path, &inode); | |
972 | nd->inode = inode; | |
973 | ||
974 | return 0; | |
975 | } | |
976 | ||
977 | static void follow_dotdot(struct nameidata *nd) | |
1da177e4 | 978 | { |
2a737871 | 979 | set_root(nd); |
e518ddb7 | 980 | |
1da177e4 | 981 | while(1) { |
4ac91378 | 982 | struct dentry *old = nd->path.dentry; |
1da177e4 | 983 | |
2a737871 AV |
984 | if (nd->path.dentry == nd->root.dentry && |
985 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
986 | break; |
987 | } | |
4ac91378 | 988 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
989 | /* rare case of legitimate dget_parent()... */ |
990 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
991 | dput(old); |
992 | break; | |
993 | } | |
3088dd70 | 994 | if (!follow_up(&nd->path)) |
1da177e4 | 995 | break; |
1da177e4 | 996 | } |
79ed0226 | 997 | follow_mount(&nd->path); |
31e6b01f | 998 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
999 | } |
1000 | ||
baa03890 NP |
1001 | /* |
1002 | * Allocate a dentry with name and parent, and perform a parent | |
1003 | * directory ->lookup on it. Returns the new dentry, or ERR_PTR | |
1004 | * on error. parent->d_inode->i_mutex must be held. d_lookup must | |
1005 | * have verified that no child exists while under i_mutex. | |
1006 | */ | |
1007 | static struct dentry *d_alloc_and_lookup(struct dentry *parent, | |
1008 | struct qstr *name, struct nameidata *nd) | |
1009 | { | |
1010 | struct inode *inode = parent->d_inode; | |
1011 | struct dentry *dentry; | |
1012 | struct dentry *old; | |
1013 | ||
1014 | /* Don't create child dentry for a dead directory. */ | |
1015 | if (unlikely(IS_DEADDIR(inode))) | |
1016 | return ERR_PTR(-ENOENT); | |
1017 | ||
1018 | dentry = d_alloc(parent, name); | |
1019 | if (unlikely(!dentry)) | |
1020 | return ERR_PTR(-ENOMEM); | |
1021 | ||
1022 | old = inode->i_op->lookup(inode, dentry, nd); | |
1023 | if (unlikely(old)) { | |
1024 | dput(dentry); | |
1025 | dentry = old; | |
1026 | } | |
1027 | return dentry; | |
1028 | } | |
1029 | ||
1da177e4 LT |
1030 | /* |
1031 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1032 | * small and for now I'd prefer to have fast path as straight as possible. | |
1033 | * It _is_ time-critical. | |
1034 | */ | |
1035 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
31e6b01f | 1036 | struct path *path, struct inode **inode) |
1da177e4 | 1037 | { |
4ac91378 | 1038 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1039 | struct dentry *dentry, *parent = nd->path.dentry; |
6e6b1bd1 | 1040 | struct inode *dir; |
3cac260a AV |
1041 | /* |
1042 | * See if the low-level filesystem might want | |
1043 | * to use its own hash.. | |
1044 | */ | |
fb045adb | 1045 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
31e6b01f | 1046 | int err = parent->d_op->d_hash(parent, nd->inode, name); |
3cac260a AV |
1047 | if (err < 0) |
1048 | return err; | |
1049 | } | |
1da177e4 | 1050 | |
b04f784e NP |
1051 | /* |
1052 | * Rename seqlock is not required here because in the off chance | |
1053 | * of a false negative due to a concurrent rename, we're going to | |
1054 | * do the non-racy lookup, below. | |
1055 | */ | |
31e6b01f NP |
1056 | if (nd->flags & LOOKUP_RCU) { |
1057 | unsigned seq; | |
1058 | ||
1059 | *inode = nd->inode; | |
1060 | dentry = __d_lookup_rcu(parent, name, &seq, inode); | |
1061 | if (!dentry) { | |
1062 | if (nameidata_drop_rcu(nd)) | |
1063 | return -ECHILD; | |
1064 | goto need_lookup; | |
1065 | } | |
1066 | /* Memory barrier in read_seqcount_begin of child is enough */ | |
1067 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) | |
1068 | return -ECHILD; | |
1069 | ||
1070 | nd->seq = seq; | |
34286d66 | 1071 | if (dentry->d_flags & DCACHE_OP_REVALIDATE) |
31e6b01f | 1072 | goto need_revalidate; |
31e6b01f NP |
1073 | path->mnt = mnt; |
1074 | path->dentry = dentry; | |
1075 | __follow_mount_rcu(nd, path, inode); | |
1076 | } else { | |
1077 | dentry = __d_lookup(parent, name); | |
1078 | if (!dentry) | |
1079 | goto need_lookup; | |
2e2e88ea | 1080 | found: |
fb045adb | 1081 | if (dentry->d_flags & DCACHE_OP_REVALIDATE) |
31e6b01f | 1082 | goto need_revalidate; |
1da177e4 | 1083 | done: |
31e6b01f NP |
1084 | path->mnt = mnt; |
1085 | path->dentry = dentry; | |
1086 | __follow_mount(path); | |
1087 | *inode = path->dentry->d_inode; | |
1088 | } | |
1da177e4 LT |
1089 | return 0; |
1090 | ||
1091 | need_lookup: | |
6e6b1bd1 | 1092 | dir = parent->d_inode; |
31e6b01f | 1093 | BUG_ON(nd->inode != dir); |
6e6b1bd1 AV |
1094 | |
1095 | mutex_lock(&dir->i_mutex); | |
1096 | /* | |
1097 | * First re-do the cached lookup just in case it was created | |
b04f784e NP |
1098 | * while we waited for the directory semaphore, or the first |
1099 | * lookup failed due to an unrelated rename. | |
6e6b1bd1 | 1100 | * |
b04f784e NP |
1101 | * This could use version numbering or similar to avoid unnecessary |
1102 | * cache lookups, but then we'd have to do the first lookup in the | |
1103 | * non-racy way. However in the common case here, everything should | |
1104 | * be hot in cache, so would it be a big win? | |
6e6b1bd1 AV |
1105 | */ |
1106 | dentry = d_lookup(parent, name); | |
baa03890 NP |
1107 | if (likely(!dentry)) { |
1108 | dentry = d_alloc_and_lookup(parent, name, nd); | |
6e6b1bd1 AV |
1109 | mutex_unlock(&dir->i_mutex); |
1110 | if (IS_ERR(dentry)) | |
1111 | goto fail; | |
1112 | goto done; | |
1113 | } | |
6e6b1bd1 AV |
1114 | /* |
1115 | * Uhhuh! Nasty case: the cache was re-populated while | |
1116 | * we waited on the semaphore. Need to revalidate. | |
1117 | */ | |
1118 | mutex_unlock(&dir->i_mutex); | |
2e2e88ea | 1119 | goto found; |
1da177e4 LT |
1120 | |
1121 | need_revalidate: | |
bcdc5e01 IK |
1122 | dentry = do_revalidate(dentry, nd); |
1123 | if (!dentry) | |
1124 | goto need_lookup; | |
1125 | if (IS_ERR(dentry)) | |
1126 | goto fail; | |
1127 | goto done; | |
1da177e4 LT |
1128 | |
1129 | fail: | |
1130 | return PTR_ERR(dentry); | |
1131 | } | |
1132 | ||
ac278a9c AV |
1133 | /* |
1134 | * This is a temporary kludge to deal with "automount" symlinks; proper | |
1135 | * solution is to trigger them on follow_mount(), so that do_lookup() | |
1136 | * would DTRT. To be killed before 2.6.34-final. | |
1137 | */ | |
1138 | static inline int follow_on_final(struct inode *inode, unsigned lookup_flags) | |
1139 | { | |
1140 | return inode && unlikely(inode->i_op->follow_link) && | |
1141 | ((lookup_flags & LOOKUP_FOLLOW) || S_ISDIR(inode->i_mode)); | |
1142 | } | |
1143 | ||
1da177e4 LT |
1144 | /* |
1145 | * Name resolution. | |
ea3834d9 PM |
1146 | * This is the basic name resolution function, turning a pathname into |
1147 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1148 | * |
ea3834d9 PM |
1149 | * Returns 0 and nd will have valid dentry and mnt on success. |
1150 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1151 | */ |
6de88d72 | 1152 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1153 | { |
1154 | struct path next; | |
1da177e4 LT |
1155 | int err; |
1156 | unsigned int lookup_flags = nd->flags; | |
1157 | ||
1158 | while (*name=='/') | |
1159 | name++; | |
1160 | if (!*name) | |
1161 | goto return_reval; | |
1162 | ||
1da177e4 | 1163 | if (nd->depth) |
f55eab82 | 1164 | lookup_flags = LOOKUP_FOLLOW | (nd->flags & LOOKUP_CONTINUE); |
1da177e4 LT |
1165 | |
1166 | /* At this point we know we have a real path component. */ | |
1167 | for(;;) { | |
31e6b01f | 1168 | struct inode *inode; |
1da177e4 LT |
1169 | unsigned long hash; |
1170 | struct qstr this; | |
1171 | unsigned int c; | |
1172 | ||
cdce5d6b | 1173 | nd->flags |= LOOKUP_CONTINUE; |
31e6b01f | 1174 | if (nd->flags & LOOKUP_RCU) { |
b74c79e9 | 1175 | err = exec_permission(nd->inode, IPERM_FLAG_RCU); |
31e6b01f NP |
1176 | if (err == -ECHILD) { |
1177 | if (nameidata_drop_rcu(nd)) | |
1178 | return -ECHILD; | |
1179 | goto exec_again; | |
1180 | } | |
1181 | } else { | |
1182 | exec_again: | |
b74c79e9 | 1183 | err = exec_permission(nd->inode, 0); |
31e6b01f | 1184 | } |
1da177e4 LT |
1185 | if (err) |
1186 | break; | |
1187 | ||
1188 | this.name = name; | |
1189 | c = *(const unsigned char *)name; | |
1190 | ||
1191 | hash = init_name_hash(); | |
1192 | do { | |
1193 | name++; | |
1194 | hash = partial_name_hash(c, hash); | |
1195 | c = *(const unsigned char *)name; | |
1196 | } while (c && (c != '/')); | |
1197 | this.len = name - (const char *) this.name; | |
1198 | this.hash = end_name_hash(hash); | |
1199 | ||
1200 | /* remove trailing slashes? */ | |
1201 | if (!c) | |
1202 | goto last_component; | |
1203 | while (*++name == '/'); | |
1204 | if (!*name) | |
1205 | goto last_with_slashes; | |
1206 | ||
1207 | /* | |
1208 | * "." and ".." are special - ".." especially so because it has | |
1209 | * to be able to know about the current root directory and | |
1210 | * parent relationships. | |
1211 | */ | |
1212 | if (this.name[0] == '.') switch (this.len) { | |
1213 | default: | |
1214 | break; | |
31e6b01f | 1215 | case 2: |
1da177e4 LT |
1216 | if (this.name[1] != '.') |
1217 | break; | |
31e6b01f NP |
1218 | if (nd->flags & LOOKUP_RCU) { |
1219 | if (follow_dotdot_rcu(nd)) | |
1220 | return -ECHILD; | |
1221 | } else | |
1222 | follow_dotdot(nd); | |
1da177e4 LT |
1223 | /* fallthrough */ |
1224 | case 1: | |
1225 | continue; | |
1226 | } | |
1da177e4 | 1227 | /* This does the actual lookups.. */ |
31e6b01f | 1228 | err = do_lookup(nd, &this, &next, &inode); |
1da177e4 LT |
1229 | if (err) |
1230 | break; | |
1da177e4 | 1231 | err = -ENOENT; |
1da177e4 LT |
1232 | if (!inode) |
1233 | goto out_dput; | |
1da177e4 LT |
1234 | |
1235 | if (inode->i_op->follow_link) { | |
31e6b01f NP |
1236 | /* We commonly drop rcu-walk here */ |
1237 | if (nameidata_dentry_drop_rcu_maybe(nd, next.dentry)) | |
1238 | return -ECHILD; | |
1239 | BUG_ON(inode != next.dentry->d_inode); | |
90ebe565 | 1240 | err = do_follow_link(&next, nd); |
1da177e4 LT |
1241 | if (err) |
1242 | goto return_err; | |
31e6b01f | 1243 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 | 1244 | err = -ENOENT; |
31e6b01f | 1245 | if (!nd->inode) |
1da177e4 | 1246 | break; |
31e6b01f | 1247 | } else { |
09dd17d3 | 1248 | path_to_nameidata(&next, nd); |
31e6b01f NP |
1249 | nd->inode = inode; |
1250 | } | |
1da177e4 | 1251 | err = -ENOTDIR; |
31e6b01f | 1252 | if (!nd->inode->i_op->lookup) |
1da177e4 LT |
1253 | break; |
1254 | continue; | |
1255 | /* here ends the main loop */ | |
1256 | ||
1257 | last_with_slashes: | |
1258 | lookup_flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1259 | last_component: | |
f55eab82 TM |
1260 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
1261 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
1da177e4 LT |
1262 | if (lookup_flags & LOOKUP_PARENT) |
1263 | goto lookup_parent; | |
1264 | if (this.name[0] == '.') switch (this.len) { | |
1265 | default: | |
1266 | break; | |
31e6b01f | 1267 | case 2: |
1da177e4 LT |
1268 | if (this.name[1] != '.') |
1269 | break; | |
31e6b01f NP |
1270 | if (nd->flags & LOOKUP_RCU) { |
1271 | if (follow_dotdot_rcu(nd)) | |
1272 | return -ECHILD; | |
1273 | } else | |
1274 | follow_dotdot(nd); | |
1da177e4 LT |
1275 | /* fallthrough */ |
1276 | case 1: | |
1277 | goto return_reval; | |
1278 | } | |
31e6b01f | 1279 | err = do_lookup(nd, &this, &next, &inode); |
1da177e4 LT |
1280 | if (err) |
1281 | break; | |
ac278a9c | 1282 | if (follow_on_final(inode, lookup_flags)) { |
31e6b01f NP |
1283 | if (nameidata_dentry_drop_rcu_maybe(nd, next.dentry)) |
1284 | return -ECHILD; | |
1285 | BUG_ON(inode != next.dentry->d_inode); | |
90ebe565 | 1286 | err = do_follow_link(&next, nd); |
1da177e4 LT |
1287 | if (err) |
1288 | goto return_err; | |
31e6b01f NP |
1289 | nd->inode = nd->path.dentry->d_inode; |
1290 | } else { | |
09dd17d3 | 1291 | path_to_nameidata(&next, nd); |
31e6b01f NP |
1292 | nd->inode = inode; |
1293 | } | |
1da177e4 | 1294 | err = -ENOENT; |
31e6b01f | 1295 | if (!nd->inode) |
1da177e4 LT |
1296 | break; |
1297 | if (lookup_flags & LOOKUP_DIRECTORY) { | |
1298 | err = -ENOTDIR; | |
31e6b01f | 1299 | if (!nd->inode->i_op->lookup) |
1da177e4 LT |
1300 | break; |
1301 | } | |
1302 | goto return_base; | |
1303 | lookup_parent: | |
1304 | nd->last = this; | |
1305 | nd->last_type = LAST_NORM; | |
1306 | if (this.name[0] != '.') | |
1307 | goto return_base; | |
1308 | if (this.len == 1) | |
1309 | nd->last_type = LAST_DOT; | |
1310 | else if (this.len == 2 && this.name[1] == '.') | |
1311 | nd->last_type = LAST_DOTDOT; | |
1312 | else | |
1313 | goto return_base; | |
1314 | return_reval: | |
1315 | /* | |
1316 | * We bypassed the ordinary revalidation routines. | |
1317 | * We may need to check the cached dentry for staleness. | |
1318 | */ | |
fb045adb | 1319 | if (need_reval_dot(nd->path.dentry)) { |
1da177e4 | 1320 | /* Note: we do not d_invalidate() */ |
34286d66 NP |
1321 | err = d_revalidate(nd->path.dentry, nd); |
1322 | if (!err) | |
1323 | err = -ESTALE; | |
1324 | if (err < 0) | |
1da177e4 LT |
1325 | break; |
1326 | } | |
1327 | return_base: | |
31e6b01f NP |
1328 | if (nameidata_drop_rcu_last_maybe(nd)) |
1329 | return -ECHILD; | |
1da177e4 LT |
1330 | return 0; |
1331 | out_dput: | |
31e6b01f NP |
1332 | if (!(nd->flags & LOOKUP_RCU)) |
1333 | path_put_conditional(&next, nd); | |
1da177e4 LT |
1334 | break; |
1335 | } | |
31e6b01f NP |
1336 | if (!(nd->flags & LOOKUP_RCU)) |
1337 | path_put(&nd->path); | |
1da177e4 LT |
1338 | return_err: |
1339 | return err; | |
1340 | } | |
1341 | ||
31e6b01f NP |
1342 | static inline int path_walk_rcu(const char *name, struct nameidata *nd) |
1343 | { | |
1344 | current->total_link_count = 0; | |
1345 | ||
1346 | return link_path_walk(name, nd); | |
1347 | } | |
1348 | ||
1349 | static inline int path_walk_simple(const char *name, struct nameidata *nd) | |
1350 | { | |
1351 | current->total_link_count = 0; | |
1352 | ||
1353 | return link_path_walk(name, nd); | |
1354 | } | |
1355 | ||
fc9b52cd | 1356 | static int path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 1357 | { |
6de88d72 AV |
1358 | struct path save = nd->path; |
1359 | int result; | |
1360 | ||
1da177e4 | 1361 | current->total_link_count = 0; |
6de88d72 AV |
1362 | |
1363 | /* make sure the stuff we saved doesn't go away */ | |
1364 | path_get(&save); | |
1365 | ||
1366 | result = link_path_walk(name, nd); | |
1367 | if (result == -ESTALE) { | |
1368 | /* nd->path had been dropped */ | |
1369 | current->total_link_count = 0; | |
1370 | nd->path = save; | |
1371 | path_get(&nd->path); | |
1372 | nd->flags |= LOOKUP_REVAL; | |
1373 | result = link_path_walk(name, nd); | |
1374 | } | |
1375 | ||
1376 | path_put(&save); | |
1377 | ||
1378 | return result; | |
1da177e4 LT |
1379 | } |
1380 | ||
31e6b01f NP |
1381 | static void path_finish_rcu(struct nameidata *nd) |
1382 | { | |
1383 | if (nd->flags & LOOKUP_RCU) { | |
1384 | /* RCU dangling. Cancel it. */ | |
1385 | nd->flags &= ~LOOKUP_RCU; | |
1386 | nd->root.mnt = NULL; | |
1387 | rcu_read_unlock(); | |
1388 | br_read_unlock(vfsmount_lock); | |
1389 | } | |
1390 | if (nd->file) | |
1391 | fput(nd->file); | |
1392 | } | |
1393 | ||
1394 | static int path_init_rcu(int dfd, const char *name, unsigned int flags, struct nameidata *nd) | |
1395 | { | |
1396 | int retval = 0; | |
1397 | int fput_needed; | |
1398 | struct file *file; | |
1399 | ||
1400 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1401 | nd->flags = flags | LOOKUP_RCU; | |
1402 | nd->depth = 0; | |
1403 | nd->root.mnt = NULL; | |
1404 | nd->file = NULL; | |
1405 | ||
1406 | if (*name=='/') { | |
1407 | struct fs_struct *fs = current->fs; | |
c28cc364 | 1408 | unsigned seq; |
31e6b01f NP |
1409 | |
1410 | br_read_lock(vfsmount_lock); | |
1411 | rcu_read_lock(); | |
1412 | ||
c28cc364 NP |
1413 | do { |
1414 | seq = read_seqcount_begin(&fs->seq); | |
1415 | nd->root = fs->root; | |
1416 | nd->path = nd->root; | |
1417 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1418 | } while (read_seqcount_retry(&fs->seq, seq)); | |
31e6b01f NP |
1419 | |
1420 | } else if (dfd == AT_FDCWD) { | |
1421 | struct fs_struct *fs = current->fs; | |
c28cc364 | 1422 | unsigned seq; |
31e6b01f NP |
1423 | |
1424 | br_read_lock(vfsmount_lock); | |
1425 | rcu_read_lock(); | |
1426 | ||
c28cc364 NP |
1427 | do { |
1428 | seq = read_seqcount_begin(&fs->seq); | |
1429 | nd->path = fs->pwd; | |
1430 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1431 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1432 | ||
31e6b01f NP |
1433 | } else { |
1434 | struct dentry *dentry; | |
1435 | ||
1436 | file = fget_light(dfd, &fput_needed); | |
1437 | retval = -EBADF; | |
1438 | if (!file) | |
1439 | goto out_fail; | |
1440 | ||
1441 | dentry = file->f_path.dentry; | |
1442 | ||
1443 | retval = -ENOTDIR; | |
1444 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
1445 | goto fput_fail; | |
1446 | ||
1447 | retval = file_permission(file, MAY_EXEC); | |
1448 | if (retval) | |
1449 | goto fput_fail; | |
1450 | ||
1451 | nd->path = file->f_path; | |
1452 | if (fput_needed) | |
1453 | nd->file = file; | |
1454 | ||
c28cc364 | 1455 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
31e6b01f NP |
1456 | br_read_lock(vfsmount_lock); |
1457 | rcu_read_lock(); | |
1458 | } | |
1459 | nd->inode = nd->path.dentry->d_inode; | |
1460 | return 0; | |
1461 | ||
1462 | fput_fail: | |
1463 | fput_light(file, fput_needed); | |
1464 | out_fail: | |
1465 | return retval; | |
1466 | } | |
1467 | ||
9b4a9b14 | 1468 | static int path_init(int dfd, const char *name, unsigned int flags, struct nameidata *nd) |
1da177e4 | 1469 | { |
ea3834d9 | 1470 | int retval = 0; |
170aa3d0 UD |
1471 | int fput_needed; |
1472 | struct file *file; | |
1da177e4 LT |
1473 | |
1474 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
1475 | nd->flags = flags; | |
1476 | nd->depth = 0; | |
2a737871 | 1477 | nd->root.mnt = NULL; |
1da177e4 | 1478 | |
1da177e4 | 1479 | if (*name=='/') { |
2a737871 AV |
1480 | set_root(nd); |
1481 | nd->path = nd->root; | |
1482 | path_get(&nd->root); | |
5590ff0d | 1483 | } else if (dfd == AT_FDCWD) { |
f7ad3c6b | 1484 | get_fs_pwd(current->fs, &nd->path); |
5590ff0d | 1485 | } else { |
5590ff0d UD |
1486 | struct dentry *dentry; |
1487 | ||
1488 | file = fget_light(dfd, &fput_needed); | |
170aa3d0 UD |
1489 | retval = -EBADF; |
1490 | if (!file) | |
6d09bb62 | 1491 | goto out_fail; |
5590ff0d | 1492 | |
0f7fc9e4 | 1493 | dentry = file->f_path.dentry; |
5590ff0d | 1494 | |
170aa3d0 UD |
1495 | retval = -ENOTDIR; |
1496 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
6d09bb62 | 1497 | goto fput_fail; |
5590ff0d UD |
1498 | |
1499 | retval = file_permission(file, MAY_EXEC); | |
170aa3d0 | 1500 | if (retval) |
6d09bb62 | 1501 | goto fput_fail; |
5590ff0d | 1502 | |
5dd784d0 JB |
1503 | nd->path = file->f_path; |
1504 | path_get(&file->f_path); | |
5590ff0d UD |
1505 | |
1506 | fput_light(file, fput_needed); | |
1da177e4 | 1507 | } |
31e6b01f | 1508 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1509 | return 0; |
2dfdd266 | 1510 | |
9b4a9b14 AV |
1511 | fput_fail: |
1512 | fput_light(file, fput_needed); | |
1513 | out_fail: | |
1514 | return retval; | |
1515 | } | |
1516 | ||
1517 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ | |
1518 | static int do_path_lookup(int dfd, const char *name, | |
1519 | unsigned int flags, struct nameidata *nd) | |
1520 | { | |
31e6b01f NP |
1521 | int retval; |
1522 | ||
1523 | /* | |
1524 | * Path walking is largely split up into 2 different synchronisation | |
1525 | * schemes, rcu-walk and ref-walk (explained in | |
1526 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1527 | * path walk code, but some things particularly setup, cleanup, and | |
1528 | * following mounts are sufficiently divergent that functions are | |
1529 | * duplicated. Typically there is a function foo(), and its RCU | |
1530 | * analogue, foo_rcu(). | |
1531 | * | |
1532 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1533 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1534 | * be handled by restarting a traditional ref-walk (which will always | |
1535 | * be able to complete). | |
1536 | */ | |
1537 | retval = path_init_rcu(dfd, name, flags, nd); | |
1538 | if (unlikely(retval)) | |
1539 | return retval; | |
1540 | retval = path_walk_rcu(name, nd); | |
1541 | path_finish_rcu(nd); | |
2a737871 AV |
1542 | if (nd->root.mnt) { |
1543 | path_put(&nd->root); | |
1544 | nd->root.mnt = NULL; | |
1545 | } | |
31e6b01f NP |
1546 | |
1547 | if (unlikely(retval == -ECHILD || retval == -ESTALE)) { | |
1548 | /* slower, locked walk */ | |
1549 | if (retval == -ESTALE) | |
1550 | flags |= LOOKUP_REVAL; | |
1551 | retval = path_init(dfd, name, flags, nd); | |
1552 | if (unlikely(retval)) | |
1553 | return retval; | |
1554 | retval = path_walk(name, nd); | |
1555 | if (nd->root.mnt) { | |
1556 | path_put(&nd->root); | |
1557 | nd->root.mnt = NULL; | |
1558 | } | |
1559 | } | |
1560 | ||
1561 | if (likely(!retval)) { | |
1562 | if (unlikely(!audit_dummy_context())) { | |
1563 | if (nd->path.dentry && nd->inode) | |
1564 | audit_inode(name, nd->path.dentry); | |
1565 | } | |
1566 | } | |
1567 | ||
170aa3d0 | 1568 | return retval; |
1da177e4 LT |
1569 | } |
1570 | ||
fc9b52cd | 1571 | int path_lookup(const char *name, unsigned int flags, |
5590ff0d UD |
1572 | struct nameidata *nd) |
1573 | { | |
1574 | return do_path_lookup(AT_FDCWD, name, flags, nd); | |
1575 | } | |
1576 | ||
d1811465 AV |
1577 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1578 | { | |
1579 | struct nameidata nd; | |
1580 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1581 | if (!res) | |
1582 | *path = nd.path; | |
1583 | return res; | |
1584 | } | |
1585 | ||
16f18200 JJS |
1586 | /** |
1587 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1588 | * @dentry: pointer to dentry of the base directory | |
1589 | * @mnt: pointer to vfs mount of the base directory | |
1590 | * @name: pointer to file name | |
1591 | * @flags: lookup flags | |
1592 | * @nd: pointer to nameidata | |
1593 | */ | |
1594 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1595 | const char *name, unsigned int flags, | |
1596 | struct nameidata *nd) | |
1597 | { | |
1598 | int retval; | |
1599 | ||
1600 | /* same as do_path_lookup */ | |
1601 | nd->last_type = LAST_ROOT; | |
1602 | nd->flags = flags; | |
1603 | nd->depth = 0; | |
1604 | ||
c8e7f449 JB |
1605 | nd->path.dentry = dentry; |
1606 | nd->path.mnt = mnt; | |
1607 | path_get(&nd->path); | |
5b857119 AV |
1608 | nd->root = nd->path; |
1609 | path_get(&nd->root); | |
31e6b01f | 1610 | nd->inode = nd->path.dentry->d_inode; |
16f18200 JJS |
1611 | |
1612 | retval = path_walk(name, nd); | |
4ac91378 | 1613 | if (unlikely(!retval && !audit_dummy_context() && nd->path.dentry && |
31e6b01f | 1614 | nd->inode)) |
4ac91378 | 1615 | audit_inode(name, nd->path.dentry); |
16f18200 | 1616 | |
5b857119 AV |
1617 | path_put(&nd->root); |
1618 | nd->root.mnt = NULL; | |
16f18200 | 1619 | |
2a737871 | 1620 | return retval; |
16f18200 JJS |
1621 | } |
1622 | ||
eead1911 CH |
1623 | static struct dentry *__lookup_hash(struct qstr *name, |
1624 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1625 | { |
81fca444 | 1626 | struct inode *inode = base->d_inode; |
057f6c01 | 1627 | struct dentry *dentry; |
1da177e4 LT |
1628 | int err; |
1629 | ||
b74c79e9 | 1630 | err = exec_permission(inode, 0); |
81fca444 CH |
1631 | if (err) |
1632 | return ERR_PTR(err); | |
1da177e4 LT |
1633 | |
1634 | /* | |
1635 | * See if the low-level filesystem might want | |
1636 | * to use its own hash.. | |
1637 | */ | |
fb045adb | 1638 | if (base->d_flags & DCACHE_OP_HASH) { |
b1e6a015 | 1639 | err = base->d_op->d_hash(base, inode, name); |
1da177e4 LT |
1640 | dentry = ERR_PTR(err); |
1641 | if (err < 0) | |
1642 | goto out; | |
1643 | } | |
1644 | ||
b04f784e NP |
1645 | /* |
1646 | * Don't bother with __d_lookup: callers are for creat as | |
1647 | * well as unlink, so a lot of the time it would cost | |
1648 | * a double lookup. | |
6e6b1bd1 | 1649 | */ |
b04f784e | 1650 | dentry = d_lookup(base, name); |
6e6b1bd1 | 1651 | |
fb045adb | 1652 | if (dentry && (dentry->d_flags & DCACHE_OP_REVALIDATE)) |
6e6b1bd1 AV |
1653 | dentry = do_revalidate(dentry, nd); |
1654 | ||
baa03890 NP |
1655 | if (!dentry) |
1656 | dentry = d_alloc_and_lookup(base, name, nd); | |
1da177e4 LT |
1657 | out: |
1658 | return dentry; | |
1659 | } | |
1660 | ||
057f6c01 JM |
1661 | /* |
1662 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1663 | * needs parent already locked. Doesn't follow mounts. | |
1664 | * SMP-safe. | |
1665 | */ | |
eead1911 | 1666 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1667 | { |
4ac91378 | 1668 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1669 | } |
1670 | ||
eead1911 CH |
1671 | static int __lookup_one_len(const char *name, struct qstr *this, |
1672 | struct dentry *base, int len) | |
1da177e4 LT |
1673 | { |
1674 | unsigned long hash; | |
1da177e4 LT |
1675 | unsigned int c; |
1676 | ||
057f6c01 JM |
1677 | this->name = name; |
1678 | this->len = len; | |
1da177e4 | 1679 | if (!len) |
057f6c01 | 1680 | return -EACCES; |
1da177e4 LT |
1681 | |
1682 | hash = init_name_hash(); | |
1683 | while (len--) { | |
1684 | c = *(const unsigned char *)name++; | |
1685 | if (c == '/' || c == '\0') | |
057f6c01 | 1686 | return -EACCES; |
1da177e4 LT |
1687 | hash = partial_name_hash(c, hash); |
1688 | } | |
057f6c01 JM |
1689 | this->hash = end_name_hash(hash); |
1690 | return 0; | |
1691 | } | |
1da177e4 | 1692 | |
eead1911 | 1693 | /** |
a6b91919 | 1694 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1695 | * @name: pathname component to lookup |
1696 | * @base: base directory to lookup from | |
1697 | * @len: maximum length @len should be interpreted to | |
1698 | * | |
a6b91919 RD |
1699 | * Note that this routine is purely a helper for filesystem usage and should |
1700 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1701 | * nameidata argument is passed to the filesystem methods and a filesystem |
1702 | * using this helper needs to be prepared for that. | |
1703 | */ | |
057f6c01 JM |
1704 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1705 | { | |
1706 | int err; | |
1707 | struct qstr this; | |
1708 | ||
2f9092e1 DW |
1709 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1710 | ||
057f6c01 | 1711 | err = __lookup_one_len(name, &this, base, len); |
eead1911 CH |
1712 | if (err) |
1713 | return ERR_PTR(err); | |
1714 | ||
49705b77 | 1715 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1716 | } |
1717 | ||
2d8f3038 AV |
1718 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1719 | struct path *path) | |
1da177e4 | 1720 | { |
2d8f3038 | 1721 | struct nameidata nd; |
1da177e4 LT |
1722 | char *tmp = getname(name); |
1723 | int err = PTR_ERR(tmp); | |
1da177e4 | 1724 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1725 | |
1726 | BUG_ON(flags & LOOKUP_PARENT); | |
1727 | ||
1728 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1729 | putname(tmp); |
2d8f3038 AV |
1730 | if (!err) |
1731 | *path = nd.path; | |
1da177e4 LT |
1732 | } |
1733 | return err; | |
1734 | } | |
1735 | ||
2ad94ae6 AV |
1736 | static int user_path_parent(int dfd, const char __user *path, |
1737 | struct nameidata *nd, char **name) | |
1738 | { | |
1739 | char *s = getname(path); | |
1740 | int error; | |
1741 | ||
1742 | if (IS_ERR(s)) | |
1743 | return PTR_ERR(s); | |
1744 | ||
1745 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1746 | if (error) | |
1747 | putname(s); | |
1748 | else | |
1749 | *name = s; | |
1750 | ||
1751 | return error; | |
1752 | } | |
1753 | ||
1da177e4 LT |
1754 | /* |
1755 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1756 | * minimal. | |
1757 | */ | |
1758 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1759 | { | |
da9592ed DH |
1760 | uid_t fsuid = current_fsuid(); |
1761 | ||
1da177e4 LT |
1762 | if (!(dir->i_mode & S_ISVTX)) |
1763 | return 0; | |
da9592ed | 1764 | if (inode->i_uid == fsuid) |
1da177e4 | 1765 | return 0; |
da9592ed | 1766 | if (dir->i_uid == fsuid) |
1da177e4 LT |
1767 | return 0; |
1768 | return !capable(CAP_FOWNER); | |
1769 | } | |
1770 | ||
1771 | /* | |
1772 | * Check whether we can remove a link victim from directory dir, check | |
1773 | * whether the type of victim is right. | |
1774 | * 1. We can't do it if dir is read-only (done in permission()) | |
1775 | * 2. We should have write and exec permissions on dir | |
1776 | * 3. We can't remove anything from append-only dir | |
1777 | * 4. We can't do anything with immutable dir (done in permission()) | |
1778 | * 5. If the sticky bit on dir is set we should either | |
1779 | * a. be owner of dir, or | |
1780 | * b. be owner of victim, or | |
1781 | * c. have CAP_FOWNER capability | |
1782 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1783 | * links pointing to it. | |
1784 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1785 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1786 | * 9. We can't remove a root or mountpoint. | |
1787 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1788 | * nfs_async_unlink(). | |
1789 | */ | |
858119e1 | 1790 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1791 | { |
1792 | int error; | |
1793 | ||
1794 | if (!victim->d_inode) | |
1795 | return -ENOENT; | |
1796 | ||
1797 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 1798 | audit_inode_child(victim, dir); |
1da177e4 | 1799 | |
f419a2e3 | 1800 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1801 | if (error) |
1802 | return error; | |
1803 | if (IS_APPEND(dir)) | |
1804 | return -EPERM; | |
1805 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1806 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1807 | return -EPERM; |
1808 | if (isdir) { | |
1809 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1810 | return -ENOTDIR; | |
1811 | if (IS_ROOT(victim)) | |
1812 | return -EBUSY; | |
1813 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1814 | return -EISDIR; | |
1815 | if (IS_DEADDIR(dir)) | |
1816 | return -ENOENT; | |
1817 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1818 | return -EBUSY; | |
1819 | return 0; | |
1820 | } | |
1821 | ||
1822 | /* Check whether we can create an object with dentry child in directory | |
1823 | * dir. | |
1824 | * 1. We can't do it if child already exists (open has special treatment for | |
1825 | * this case, but since we are inlined it's OK) | |
1826 | * 2. We can't do it if dir is read-only (done in permission()) | |
1827 | * 3. We should have write and exec permissions on dir | |
1828 | * 4. We can't do it if dir is immutable (done in permission()) | |
1829 | */ | |
a95164d9 | 1830 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1831 | { |
1832 | if (child->d_inode) | |
1833 | return -EEXIST; | |
1834 | if (IS_DEADDIR(dir)) | |
1835 | return -ENOENT; | |
f419a2e3 | 1836 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1837 | } |
1838 | ||
1da177e4 LT |
1839 | /* |
1840 | * p1 and p2 should be directories on the same fs. | |
1841 | */ | |
1842 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1843 | { | |
1844 | struct dentry *p; | |
1845 | ||
1846 | if (p1 == p2) { | |
f2eace23 | 1847 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1848 | return NULL; |
1849 | } | |
1850 | ||
a11f3a05 | 1851 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1852 | |
e2761a11 OH |
1853 | p = d_ancestor(p2, p1); |
1854 | if (p) { | |
1855 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1856 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1857 | return p; | |
1da177e4 LT |
1858 | } |
1859 | ||
e2761a11 OH |
1860 | p = d_ancestor(p1, p2); |
1861 | if (p) { | |
1862 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1863 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1864 | return p; | |
1da177e4 LT |
1865 | } |
1866 | ||
f2eace23 IM |
1867 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1868 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1869 | return NULL; |
1870 | } | |
1871 | ||
1872 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1873 | { | |
1b1dcc1b | 1874 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1875 | if (p1 != p2) { |
1b1dcc1b | 1876 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1877 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1878 | } |
1879 | } | |
1880 | ||
1881 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1882 | struct nameidata *nd) | |
1883 | { | |
a95164d9 | 1884 | int error = may_create(dir, dentry); |
1da177e4 LT |
1885 | |
1886 | if (error) | |
1887 | return error; | |
1888 | ||
acfa4380 | 1889 | if (!dir->i_op->create) |
1da177e4 LT |
1890 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1891 | mode &= S_IALLUGO; | |
1892 | mode |= S_IFREG; | |
1893 | error = security_inode_create(dir, dentry, mode); | |
1894 | if (error) | |
1895 | return error; | |
1da177e4 | 1896 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1897 | if (!error) |
f38aa942 | 1898 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1899 | return error; |
1900 | } | |
1901 | ||
3fb64190 | 1902 | int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1903 | { |
3fb64190 | 1904 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1905 | struct inode *inode = dentry->d_inode; |
1906 | int error; | |
1907 | ||
1908 | if (!inode) | |
1909 | return -ENOENT; | |
1910 | ||
c8fe8f30 CH |
1911 | switch (inode->i_mode & S_IFMT) { |
1912 | case S_IFLNK: | |
1da177e4 | 1913 | return -ELOOP; |
c8fe8f30 CH |
1914 | case S_IFDIR: |
1915 | if (acc_mode & MAY_WRITE) | |
1916 | return -EISDIR; | |
1917 | break; | |
1918 | case S_IFBLK: | |
1919 | case S_IFCHR: | |
3fb64190 | 1920 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1921 | return -EACCES; |
c8fe8f30 CH |
1922 | /*FALLTHRU*/ |
1923 | case S_IFIFO: | |
1924 | case S_IFSOCK: | |
1da177e4 | 1925 | flag &= ~O_TRUNC; |
c8fe8f30 | 1926 | break; |
4a3fd211 | 1927 | } |
b41572e9 | 1928 | |
3fb64190 | 1929 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1930 | if (error) |
1931 | return error; | |
6146f0d5 | 1932 | |
1da177e4 LT |
1933 | /* |
1934 | * An append-only file must be opened in append mode for writing. | |
1935 | */ | |
1936 | if (IS_APPEND(inode)) { | |
8737c930 | 1937 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 1938 | return -EPERM; |
1da177e4 | 1939 | if (flag & O_TRUNC) |
7715b521 | 1940 | return -EPERM; |
1da177e4 LT |
1941 | } |
1942 | ||
1943 | /* O_NOATIME can only be set by the owner or superuser */ | |
7715b521 AV |
1944 | if (flag & O_NOATIME && !is_owner_or_cap(inode)) |
1945 | return -EPERM; | |
1da177e4 LT |
1946 | |
1947 | /* | |
1948 | * Ensure there are no outstanding leases on the file. | |
1949 | */ | |
b65a9cfc | 1950 | return break_lease(inode, flag); |
7715b521 | 1951 | } |
1da177e4 | 1952 | |
7715b521 AV |
1953 | static int handle_truncate(struct path *path) |
1954 | { | |
1955 | struct inode *inode = path->dentry->d_inode; | |
1956 | int error = get_write_access(inode); | |
1957 | if (error) | |
1958 | return error; | |
1959 | /* | |
1960 | * Refuse to truncate files with mandatory locks held on them. | |
1961 | */ | |
1962 | error = locks_verify_locked(inode); | |
1963 | if (!error) | |
ea0d3ab2 | 1964 | error = security_path_truncate(path); |
7715b521 AV |
1965 | if (!error) { |
1966 | error = do_truncate(path->dentry, 0, | |
1967 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
1968 | NULL); | |
1969 | } | |
1970 | put_write_access(inode); | |
acd0c935 | 1971 | return error; |
1da177e4 LT |
1972 | } |
1973 | ||
d57999e1 DH |
1974 | /* |
1975 | * Be careful about ever adding any more callers of this | |
1976 | * function. Its flags must be in the namei format, not | |
1977 | * what get passed to sys_open(). | |
1978 | */ | |
1979 | static int __open_namei_create(struct nameidata *nd, struct path *path, | |
8737c930 | 1980 | int open_flag, int mode) |
aab520e2 DH |
1981 | { |
1982 | int error; | |
4ac91378 | 1983 | struct dentry *dir = nd->path.dentry; |
aab520e2 DH |
1984 | |
1985 | if (!IS_POSIXACL(dir->d_inode)) | |
ce3b0f8d | 1986 | mode &= ~current_umask(); |
be6d3e56 KT |
1987 | error = security_path_mknod(&nd->path, path->dentry, mode, 0); |
1988 | if (error) | |
1989 | goto out_unlock; | |
aab520e2 | 1990 | error = vfs_create(dir->d_inode, path->dentry, mode, nd); |
be6d3e56 | 1991 | out_unlock: |
aab520e2 | 1992 | mutex_unlock(&dir->d_inode->i_mutex); |
4ac91378 JB |
1993 | dput(nd->path.dentry); |
1994 | nd->path.dentry = path->dentry; | |
31e6b01f | 1995 | |
aab520e2 DH |
1996 | if (error) |
1997 | return error; | |
1998 | /* Don't check for write permission, don't truncate */ | |
8737c930 | 1999 | return may_open(&nd->path, 0, open_flag & ~O_TRUNC); |
aab520e2 DH |
2000 | } |
2001 | ||
d57999e1 DH |
2002 | /* |
2003 | * Note that while the flag value (low two bits) for sys_open means: | |
2004 | * 00 - read-only | |
2005 | * 01 - write-only | |
2006 | * 10 - read-write | |
2007 | * 11 - special | |
2008 | * it is changed into | |
2009 | * 00 - no permissions needed | |
2010 | * 01 - read-permission | |
2011 | * 10 - write-permission | |
2012 | * 11 - read-write | |
2013 | * for the internal routines (ie open_namei()/follow_link() etc) | |
2014 | * This is more logical, and also allows the 00 "no perm needed" | |
2015 | * to be used for symlinks (where the permissions are checked | |
2016 | * later). | |
2017 | * | |
2018 | */ | |
2019 | static inline int open_to_namei_flags(int flag) | |
2020 | { | |
2021 | if ((flag+1) & O_ACCMODE) | |
2022 | flag++; | |
2023 | return flag; | |
2024 | } | |
2025 | ||
7715b521 | 2026 | static int open_will_truncate(int flag, struct inode *inode) |
4a3fd211 DH |
2027 | { |
2028 | /* | |
2029 | * We'll never write to the fs underlying | |
2030 | * a device file. | |
2031 | */ | |
2032 | if (special_file(inode->i_mode)) | |
2033 | return 0; | |
2034 | return (flag & O_TRUNC); | |
2035 | } | |
2036 | ||
648fa861 | 2037 | static struct file *finish_open(struct nameidata *nd, |
9a66179e | 2038 | int open_flag, int acc_mode) |
648fa861 AV |
2039 | { |
2040 | struct file *filp; | |
2041 | int will_truncate; | |
2042 | int error; | |
2043 | ||
9a66179e | 2044 | will_truncate = open_will_truncate(open_flag, nd->path.dentry->d_inode); |
648fa861 AV |
2045 | if (will_truncate) { |
2046 | error = mnt_want_write(nd->path.mnt); | |
2047 | if (error) | |
2048 | goto exit; | |
2049 | } | |
2050 | error = may_open(&nd->path, acc_mode, open_flag); | |
2051 | if (error) { | |
2052 | if (will_truncate) | |
2053 | mnt_drop_write(nd->path.mnt); | |
2054 | goto exit; | |
2055 | } | |
2056 | filp = nameidata_to_filp(nd); | |
2057 | if (!IS_ERR(filp)) { | |
2058 | error = ima_file_check(filp, acc_mode); | |
2059 | if (error) { | |
2060 | fput(filp); | |
2061 | filp = ERR_PTR(error); | |
2062 | } | |
2063 | } | |
2064 | if (!IS_ERR(filp)) { | |
648fa861 AV |
2065 | if (will_truncate) { |
2066 | error = handle_truncate(&nd->path); | |
2067 | if (error) { | |
2068 | fput(filp); | |
2069 | filp = ERR_PTR(error); | |
2070 | } | |
2071 | } | |
2072 | } | |
2073 | /* | |
2074 | * It is now safe to drop the mnt write | |
2075 | * because the filp has had a write taken | |
2076 | * on its behalf. | |
2077 | */ | |
2078 | if (will_truncate) | |
2079 | mnt_drop_write(nd->path.mnt); | |
d893f1bc | 2080 | path_put(&nd->path); |
648fa861 AV |
2081 | return filp; |
2082 | ||
2083 | exit: | |
2084 | if (!IS_ERR(nd->intent.open.file)) | |
2085 | release_open_intent(nd); | |
2086 | path_put(&nd->path); | |
2087 | return ERR_PTR(error); | |
2088 | } | |
2089 | ||
31e6b01f NP |
2090 | /* |
2091 | * Handle O_CREAT case for do_filp_open | |
2092 | */ | |
fb1cc555 | 2093 | static struct file *do_last(struct nameidata *nd, struct path *path, |
5b369df8 | 2094 | int open_flag, int acc_mode, |
3e297b61 | 2095 | int mode, const char *pathname) |
fb1cc555 | 2096 | { |
a1e28038 | 2097 | struct dentry *dir = nd->path.dentry; |
fb1cc555 | 2098 | struct file *filp; |
1f36f774 AV |
2099 | int error = -EISDIR; |
2100 | ||
2101 | switch (nd->last_type) { | |
2102 | case LAST_DOTDOT: | |
2103 | follow_dotdot(nd); | |
2104 | dir = nd->path.dentry; | |
176306f5 | 2105 | case LAST_DOT: |
fb045adb | 2106 | if (need_reval_dot(dir)) { |
34286d66 NP |
2107 | error = d_revalidate(nd->path.dentry, nd); |
2108 | if (!error) | |
1f36f774 | 2109 | error = -ESTALE; |
34286d66 | 2110 | if (error < 0) |
1f36f774 | 2111 | goto exit; |
1f36f774 AV |
2112 | } |
2113 | /* fallthrough */ | |
1f36f774 | 2114 | case LAST_ROOT: |
31e6b01f | 2115 | goto exit; |
1f36f774 AV |
2116 | case LAST_BIND: |
2117 | audit_inode(pathname, dir); | |
67ee3ad2 | 2118 | goto ok; |
1f36f774 | 2119 | } |
67ee3ad2 | 2120 | |
1f36f774 | 2121 | /* trailing slashes? */ |
31e6b01f NP |
2122 | if (nd->last.name[nd->last.len]) |
2123 | goto exit; | |
a2c36b45 | 2124 | |
a1e28038 AV |
2125 | mutex_lock(&dir->d_inode->i_mutex); |
2126 | ||
2127 | path->dentry = lookup_hash(nd); | |
2128 | path->mnt = nd->path.mnt; | |
2129 | ||
fb1cc555 AV |
2130 | error = PTR_ERR(path->dentry); |
2131 | if (IS_ERR(path->dentry)) { | |
2132 | mutex_unlock(&dir->d_inode->i_mutex); | |
2133 | goto exit; | |
2134 | } | |
2135 | ||
2136 | if (IS_ERR(nd->intent.open.file)) { | |
2137 | error = PTR_ERR(nd->intent.open.file); | |
2138 | goto exit_mutex_unlock; | |
2139 | } | |
2140 | ||
2141 | /* Negative dentry, just create the file */ | |
2142 | if (!path->dentry->d_inode) { | |
2143 | /* | |
2144 | * This write is needed to ensure that a | |
2145 | * ro->rw transition does not occur between | |
2146 | * the time when the file is created and when | |
2147 | * a permanent write count is taken through | |
2148 | * the 'struct file' in nameidata_to_filp(). | |
2149 | */ | |
2150 | error = mnt_want_write(nd->path.mnt); | |
2151 | if (error) | |
2152 | goto exit_mutex_unlock; | |
2153 | error = __open_namei_create(nd, path, open_flag, mode); | |
2154 | if (error) { | |
2155 | mnt_drop_write(nd->path.mnt); | |
2156 | goto exit; | |
2157 | } | |
2158 | filp = nameidata_to_filp(nd); | |
2159 | mnt_drop_write(nd->path.mnt); | |
d893f1bc | 2160 | path_put(&nd->path); |
fb1cc555 AV |
2161 | if (!IS_ERR(filp)) { |
2162 | error = ima_file_check(filp, acc_mode); | |
2163 | if (error) { | |
2164 | fput(filp); | |
2165 | filp = ERR_PTR(error); | |
2166 | } | |
2167 | } | |
2168 | return filp; | |
2169 | } | |
2170 | ||
2171 | /* | |
2172 | * It already exists. | |
2173 | */ | |
2174 | mutex_unlock(&dir->d_inode->i_mutex); | |
2175 | audit_inode(pathname, path->dentry); | |
2176 | ||
2177 | error = -EEXIST; | |
5b369df8 | 2178 | if (open_flag & O_EXCL) |
fb1cc555 AV |
2179 | goto exit_dput; |
2180 | ||
2181 | if (__follow_mount(path)) { | |
2182 | error = -ELOOP; | |
5b369df8 | 2183 | if (open_flag & O_NOFOLLOW) |
fb1cc555 AV |
2184 | goto exit_dput; |
2185 | } | |
2186 | ||
2187 | error = -ENOENT; | |
2188 | if (!path->dentry->d_inode) | |
2189 | goto exit_dput; | |
9e67f361 AV |
2190 | |
2191 | if (path->dentry->d_inode->i_op->follow_link) | |
fb1cc555 | 2192 | return NULL; |
fb1cc555 AV |
2193 | |
2194 | path_to_nameidata(path, nd); | |
31e6b01f | 2195 | nd->inode = path->dentry->d_inode; |
fb1cc555 | 2196 | error = -EISDIR; |
31e6b01f | 2197 | if (S_ISDIR(nd->inode->i_mode)) |
fb1cc555 | 2198 | goto exit; |
67ee3ad2 | 2199 | ok: |
9a66179e | 2200 | filp = finish_open(nd, open_flag, acc_mode); |
fb1cc555 AV |
2201 | return filp; |
2202 | ||
2203 | exit_mutex_unlock: | |
2204 | mutex_unlock(&dir->d_inode->i_mutex); | |
2205 | exit_dput: | |
2206 | path_put_conditional(path, nd); | |
2207 | exit: | |
2208 | if (!IS_ERR(nd->intent.open.file)) | |
2209 | release_open_intent(nd); | |
fb1cc555 AV |
2210 | path_put(&nd->path); |
2211 | return ERR_PTR(error); | |
2212 | } | |
2213 | ||
1da177e4 | 2214 | /* |
4a3fd211 DH |
2215 | * Note that the low bits of the passed in "open_flag" |
2216 | * are not the same as in the local variable "flag". See | |
2217 | * open_to_namei_flags() for more details. | |
1da177e4 | 2218 | */ |
a70e65df | 2219 | struct file *do_filp_open(int dfd, const char *pathname, |
6e8341a1 | 2220 | int open_flag, int mode, int acc_mode) |
1da177e4 | 2221 | { |
4a3fd211 | 2222 | struct file *filp; |
a70e65df | 2223 | struct nameidata nd; |
6e8341a1 | 2224 | int error; |
9850c056 | 2225 | struct path path; |
1da177e4 | 2226 | int count = 0; |
d57999e1 | 2227 | int flag = open_to_namei_flags(open_flag); |
31e6b01f | 2228 | int flags; |
1f36f774 AV |
2229 | |
2230 | if (!(open_flag & O_CREAT)) | |
2231 | mode = 0; | |
1da177e4 | 2232 | |
b1085ba8 LS |
2233 | /* Must never be set by userspace */ |
2234 | open_flag &= ~FMODE_NONOTIFY; | |
2235 | ||
6b2f3d1f CH |
2236 | /* |
2237 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
2238 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
2239 | * always set instead of having to deal with possibly weird behaviour | |
2240 | * for malicious applications setting only __O_SYNC. | |
2241 | */ | |
2242 | if (open_flag & __O_SYNC) | |
2243 | open_flag |= O_DSYNC; | |
2244 | ||
6e8341a1 | 2245 | if (!acc_mode) |
6d125529 | 2246 | acc_mode = MAY_OPEN | ACC_MODE(open_flag); |
1da177e4 | 2247 | |
834f2a4a | 2248 | /* O_TRUNC implies we need access checks for write permissions */ |
4296e2cb | 2249 | if (open_flag & O_TRUNC) |
834f2a4a TM |
2250 | acc_mode |= MAY_WRITE; |
2251 | ||
1da177e4 LT |
2252 | /* Allow the LSM permission hook to distinguish append |
2253 | access from general write access. */ | |
4296e2cb | 2254 | if (open_flag & O_APPEND) |
1da177e4 LT |
2255 | acc_mode |= MAY_APPEND; |
2256 | ||
31e6b01f NP |
2257 | flags = LOOKUP_OPEN; |
2258 | if (open_flag & O_CREAT) { | |
2259 | flags |= LOOKUP_CREATE; | |
2260 | if (open_flag & O_EXCL) | |
2261 | flags |= LOOKUP_EXCL; | |
2262 | } | |
2263 | if (open_flag & O_DIRECTORY) | |
2264 | flags |= LOOKUP_DIRECTORY; | |
2265 | if (!(open_flag & O_NOFOLLOW)) | |
2266 | flags |= LOOKUP_FOLLOW; | |
2267 | ||
2268 | filp = get_empty_filp(); | |
2269 | if (!filp) | |
2270 | return ERR_PTR(-ENFILE); | |
2271 | ||
2272 | filp->f_flags = open_flag; | |
2273 | nd.intent.open.file = filp; | |
2274 | nd.intent.open.flags = flag; | |
2275 | nd.intent.open.create_mode = mode; | |
2276 | ||
2277 | if (open_flag & O_CREAT) | |
2278 | goto creat; | |
2279 | ||
2280 | /* !O_CREAT, simple open */ | |
2281 | error = do_path_lookup(dfd, pathname, flags, &nd); | |
2282 | if (unlikely(error)) | |
2283 | goto out_filp; | |
2284 | error = -ELOOP; | |
2285 | if (!(nd.flags & LOOKUP_FOLLOW)) { | |
2286 | if (nd.inode->i_op->follow_link) | |
2287 | goto out_path; | |
2288 | } | |
2289 | error = -ENOTDIR; | |
2290 | if (nd.flags & LOOKUP_DIRECTORY) { | |
2291 | if (!nd.inode->i_op->lookup) | |
2292 | goto out_path; | |
2293 | } | |
2294 | audit_inode(pathname, nd.path.dentry); | |
2295 | filp = finish_open(&nd, open_flag, acc_mode); | |
2296 | return filp; | |
2297 | ||
2298 | creat: | |
2299 | /* OK, have to create the file. Find the parent. */ | |
2300 | error = path_init_rcu(dfd, pathname, | |
2301 | LOOKUP_PARENT | (flags & LOOKUP_REVAL), &nd); | |
1da177e4 | 2302 | if (error) |
31e6b01f NP |
2303 | goto out_filp; |
2304 | error = path_walk_rcu(pathname, &nd); | |
2305 | path_finish_rcu(&nd); | |
2306 | if (unlikely(error == -ECHILD || error == -ESTALE)) { | |
2307 | /* slower, locked walk */ | |
2308 | if (error == -ESTALE) { | |
2309 | reval: | |
2310 | flags |= LOOKUP_REVAL; | |
2311 | } | |
2312 | error = path_init(dfd, pathname, | |
2313 | LOOKUP_PARENT | (flags & LOOKUP_REVAL), &nd); | |
2314 | if (error) | |
2315 | goto out_filp; | |
3866248e | 2316 | |
31e6b01f | 2317 | error = path_walk_simple(pathname, &nd); |
654f562c | 2318 | } |
31e6b01f NP |
2319 | if (unlikely(error)) |
2320 | goto out_filp; | |
2321 | if (unlikely(!audit_dummy_context())) | |
9b4a9b14 | 2322 | audit_inode(pathname, nd.path.dentry); |
1da177e4 LT |
2323 | |
2324 | /* | |
a2c36b45 | 2325 | * We have the parent and last component. |
1da177e4 | 2326 | */ |
31e6b01f | 2327 | nd.flags = flags; |
3e297b61 | 2328 | filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); |
806b681c | 2329 | while (unlikely(!filp)) { /* trailing symlink */ |
def4af30 | 2330 | struct path holder; |
def4af30 | 2331 | void *cookie; |
806b681c | 2332 | error = -ELOOP; |
1f36f774 | 2333 | /* S_ISDIR part is a temporary automount kludge */ |
31e6b01f | 2334 | if (!(nd.flags & LOOKUP_FOLLOW) && !S_ISDIR(nd.inode->i_mode)) |
1f36f774 AV |
2335 | goto exit_dput; |
2336 | if (count++ == 32) | |
806b681c AV |
2337 | goto exit_dput; |
2338 | /* | |
2339 | * This is subtle. Instead of calling do_follow_link() we do | |
2340 | * the thing by hands. The reason is that this way we have zero | |
2341 | * link_count and path_walk() (called from ->follow_link) | |
2342 | * honoring LOOKUP_PARENT. After that we have the parent and | |
2343 | * last component, i.e. we are in the same situation as after | |
2344 | * the first path_walk(). Well, almost - if the last component | |
2345 | * is normal we get its copy stored in nd->last.name and we will | |
2346 | * have to putname() it when we are done. Procfs-like symlinks | |
2347 | * just set LAST_BIND. | |
2348 | */ | |
2349 | nd.flags |= LOOKUP_PARENT; | |
2350 | error = security_inode_follow_link(path.dentry, &nd); | |
2351 | if (error) | |
2352 | goto exit_dput; | |
def4af30 AV |
2353 | error = __do_follow_link(&path, &nd, &cookie); |
2354 | if (unlikely(error)) { | |
31e6b01f NP |
2355 | if (!IS_ERR(cookie) && nd.inode->i_op->put_link) |
2356 | nd.inode->i_op->put_link(path.dentry, &nd, cookie); | |
806b681c | 2357 | /* nd.path had been dropped */ |
31e6b01f NP |
2358 | nd.path = path; |
2359 | goto out_path; | |
806b681c | 2360 | } |
def4af30 | 2361 | holder = path; |
806b681c | 2362 | nd.flags &= ~LOOKUP_PARENT; |
3e297b61 | 2363 | filp = do_last(&nd, &path, open_flag, acc_mode, mode, pathname); |
31e6b01f NP |
2364 | if (nd.inode->i_op->put_link) |
2365 | nd.inode->i_op->put_link(holder.dentry, &nd, cookie); | |
def4af30 | 2366 | path_put(&holder); |
806b681c | 2367 | } |
10fa8e62 | 2368 | out: |
2a737871 AV |
2369 | if (nd.root.mnt) |
2370 | path_put(&nd.root); | |
31e6b01f | 2371 | if (filp == ERR_PTR(-ESTALE) && !(flags & LOOKUP_REVAL)) |
10fa8e62 | 2372 | goto reval; |
10fa8e62 | 2373 | return filp; |
1da177e4 | 2374 | |
806b681c AV |
2375 | exit_dput: |
2376 | path_put_conditional(&path, &nd); | |
31e6b01f NP |
2377 | out_path: |
2378 | path_put(&nd.path); | |
2379 | out_filp: | |
806b681c | 2380 | if (!IS_ERR(nd.intent.open.file)) |
a70e65df | 2381 | release_open_intent(&nd); |
806b681c | 2382 | filp = ERR_PTR(error); |
10fa8e62 | 2383 | goto out; |
1da177e4 LT |
2384 | } |
2385 | ||
a70e65df CH |
2386 | /** |
2387 | * filp_open - open file and return file pointer | |
2388 | * | |
2389 | * @filename: path to open | |
2390 | * @flags: open flags as per the open(2) second argument | |
2391 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
2392 | * | |
2393 | * This is the helper to open a file from kernelspace if you really | |
2394 | * have to. But in generally you should not do this, so please move | |
2395 | * along, nothing to see here.. | |
2396 | */ | |
2397 | struct file *filp_open(const char *filename, int flags, int mode) | |
2398 | { | |
6e8341a1 | 2399 | return do_filp_open(AT_FDCWD, filename, flags, mode, 0); |
a70e65df CH |
2400 | } |
2401 | EXPORT_SYMBOL(filp_open); | |
2402 | ||
1da177e4 LT |
2403 | /** |
2404 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
2405 | * @nd: nameidata info | |
2406 | * @is_dir: directory flag | |
2407 | * | |
2408 | * Simple function to lookup and return a dentry and create it | |
2409 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 2410 | * |
4ac91378 | 2411 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
2412 | */ |
2413 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
2414 | { | |
c663e5d8 | 2415 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 2416 | |
4ac91378 | 2417 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
2418 | /* |
2419 | * Yucky last component or no last component at all? | |
2420 | * (foo/., foo/.., /////) | |
2421 | */ | |
1da177e4 LT |
2422 | if (nd->last_type != LAST_NORM) |
2423 | goto fail; | |
2424 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 2425 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 2426 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
2427 | |
2428 | /* | |
2429 | * Do the final lookup. | |
2430 | */ | |
49705b77 | 2431 | dentry = lookup_hash(nd); |
1da177e4 LT |
2432 | if (IS_ERR(dentry)) |
2433 | goto fail; | |
c663e5d8 | 2434 | |
e9baf6e5 AV |
2435 | if (dentry->d_inode) |
2436 | goto eexist; | |
c663e5d8 CH |
2437 | /* |
2438 | * Special case - lookup gave negative, but... we had foo/bar/ | |
2439 | * From the vfs_mknod() POV we just have a negative dentry - | |
2440 | * all is fine. Let's be bastards - you had / on the end, you've | |
2441 | * been asking for (non-existent) directory. -ENOENT for you. | |
2442 | */ | |
e9baf6e5 AV |
2443 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
2444 | dput(dentry); | |
2445 | dentry = ERR_PTR(-ENOENT); | |
2446 | } | |
1da177e4 | 2447 | return dentry; |
e9baf6e5 | 2448 | eexist: |
1da177e4 | 2449 | dput(dentry); |
e9baf6e5 | 2450 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
2451 | fail: |
2452 | return dentry; | |
2453 | } | |
f81a0bff | 2454 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
2455 | |
2456 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
2457 | { | |
a95164d9 | 2458 | int error = may_create(dir, dentry); |
1da177e4 LT |
2459 | |
2460 | if (error) | |
2461 | return error; | |
2462 | ||
2463 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) | |
2464 | return -EPERM; | |
2465 | ||
acfa4380 | 2466 | if (!dir->i_op->mknod) |
1da177e4 LT |
2467 | return -EPERM; |
2468 | ||
08ce5f16 SH |
2469 | error = devcgroup_inode_mknod(mode, dev); |
2470 | if (error) | |
2471 | return error; | |
2472 | ||
1da177e4 LT |
2473 | error = security_inode_mknod(dir, dentry, mode, dev); |
2474 | if (error) | |
2475 | return error; | |
2476 | ||
1da177e4 | 2477 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 2478 | if (!error) |
f38aa942 | 2479 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2480 | return error; |
2481 | } | |
2482 | ||
463c3197 DH |
2483 | static int may_mknod(mode_t mode) |
2484 | { | |
2485 | switch (mode & S_IFMT) { | |
2486 | case S_IFREG: | |
2487 | case S_IFCHR: | |
2488 | case S_IFBLK: | |
2489 | case S_IFIFO: | |
2490 | case S_IFSOCK: | |
2491 | case 0: /* zero mode translates to S_IFREG */ | |
2492 | return 0; | |
2493 | case S_IFDIR: | |
2494 | return -EPERM; | |
2495 | default: | |
2496 | return -EINVAL; | |
2497 | } | |
2498 | } | |
2499 | ||
2e4d0924 HC |
2500 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
2501 | unsigned, dev) | |
1da177e4 | 2502 | { |
2ad94ae6 AV |
2503 | int error; |
2504 | char *tmp; | |
2505 | struct dentry *dentry; | |
1da177e4 LT |
2506 | struct nameidata nd; |
2507 | ||
2508 | if (S_ISDIR(mode)) | |
2509 | return -EPERM; | |
1da177e4 | 2510 | |
2ad94ae6 | 2511 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 2512 | if (error) |
2ad94ae6 AV |
2513 | return error; |
2514 | ||
1da177e4 | 2515 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2516 | if (IS_ERR(dentry)) { |
2517 | error = PTR_ERR(dentry); | |
2518 | goto out_unlock; | |
2519 | } | |
4ac91378 | 2520 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2521 | mode &= ~current_umask(); |
463c3197 DH |
2522 | error = may_mknod(mode); |
2523 | if (error) | |
2524 | goto out_dput; | |
2525 | error = mnt_want_write(nd.path.mnt); | |
2526 | if (error) | |
2527 | goto out_dput; | |
be6d3e56 KT |
2528 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2529 | if (error) | |
2530 | goto out_drop_write; | |
463c3197 | 2531 | switch (mode & S_IFMT) { |
1da177e4 | 2532 | case 0: case S_IFREG: |
4ac91378 | 2533 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2534 | break; |
2535 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2536 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2537 | new_decode_dev(dev)); |
2538 | break; | |
2539 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2540 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2541 | break; |
1da177e4 | 2542 | } |
be6d3e56 | 2543 | out_drop_write: |
463c3197 DH |
2544 | mnt_drop_write(nd.path.mnt); |
2545 | out_dput: | |
2546 | dput(dentry); | |
2547 | out_unlock: | |
4ac91378 | 2548 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2549 | path_put(&nd.path); |
1da177e4 LT |
2550 | putname(tmp); |
2551 | ||
2552 | return error; | |
2553 | } | |
2554 | ||
3480b257 | 2555 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2556 | { |
2557 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2558 | } | |
2559 | ||
1da177e4 LT |
2560 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2561 | { | |
a95164d9 | 2562 | int error = may_create(dir, dentry); |
1da177e4 LT |
2563 | |
2564 | if (error) | |
2565 | return error; | |
2566 | ||
acfa4380 | 2567 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2568 | return -EPERM; |
2569 | ||
2570 | mode &= (S_IRWXUGO|S_ISVTX); | |
2571 | error = security_inode_mkdir(dir, dentry, mode); | |
2572 | if (error) | |
2573 | return error; | |
2574 | ||
1da177e4 | 2575 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2576 | if (!error) |
f38aa942 | 2577 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2578 | return error; |
2579 | } | |
2580 | ||
2e4d0924 | 2581 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2582 | { |
2583 | int error = 0; | |
2584 | char * tmp; | |
6902d925 DH |
2585 | struct dentry *dentry; |
2586 | struct nameidata nd; | |
1da177e4 | 2587 | |
2ad94ae6 AV |
2588 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2589 | if (error) | |
6902d925 | 2590 | goto out_err; |
1da177e4 | 2591 | |
6902d925 DH |
2592 | dentry = lookup_create(&nd, 1); |
2593 | error = PTR_ERR(dentry); | |
2594 | if (IS_ERR(dentry)) | |
2595 | goto out_unlock; | |
1da177e4 | 2596 | |
4ac91378 | 2597 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2598 | mode &= ~current_umask(); |
463c3197 DH |
2599 | error = mnt_want_write(nd.path.mnt); |
2600 | if (error) | |
2601 | goto out_dput; | |
be6d3e56 KT |
2602 | error = security_path_mkdir(&nd.path, dentry, mode); |
2603 | if (error) | |
2604 | goto out_drop_write; | |
4ac91378 | 2605 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2606 | out_drop_write: |
463c3197 DH |
2607 | mnt_drop_write(nd.path.mnt); |
2608 | out_dput: | |
6902d925 DH |
2609 | dput(dentry); |
2610 | out_unlock: | |
4ac91378 | 2611 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2612 | path_put(&nd.path); |
6902d925 DH |
2613 | putname(tmp); |
2614 | out_err: | |
1da177e4 LT |
2615 | return error; |
2616 | } | |
2617 | ||
3cdad428 | 2618 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2619 | { |
2620 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2621 | } | |
2622 | ||
1da177e4 LT |
2623 | /* |
2624 | * We try to drop the dentry early: we should have | |
2625 | * a usage count of 2 if we're the only user of this | |
2626 | * dentry, and if that is true (possibly after pruning | |
2627 | * the dcache), then we drop the dentry now. | |
2628 | * | |
2629 | * A low-level filesystem can, if it choses, legally | |
2630 | * do a | |
2631 | * | |
2632 | * if (!d_unhashed(dentry)) | |
2633 | * return -EBUSY; | |
2634 | * | |
2635 | * if it cannot handle the case of removing a directory | |
2636 | * that is still in use by something else.. | |
2637 | */ | |
2638 | void dentry_unhash(struct dentry *dentry) | |
2639 | { | |
2640 | dget(dentry); | |
dc168427 | 2641 | shrink_dcache_parent(dentry); |
1da177e4 | 2642 | spin_lock(&dentry->d_lock); |
b7ab39f6 | 2643 | if (dentry->d_count == 2) |
1da177e4 LT |
2644 | __d_drop(dentry); |
2645 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
2646 | } |
2647 | ||
2648 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2649 | { | |
2650 | int error = may_delete(dir, dentry, 1); | |
2651 | ||
2652 | if (error) | |
2653 | return error; | |
2654 | ||
acfa4380 | 2655 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2656 | return -EPERM; |
2657 | ||
1b1dcc1b | 2658 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2659 | dentry_unhash(dentry); |
2660 | if (d_mountpoint(dentry)) | |
2661 | error = -EBUSY; | |
2662 | else { | |
2663 | error = security_inode_rmdir(dir, dentry); | |
2664 | if (!error) { | |
2665 | error = dir->i_op->rmdir(dir, dentry); | |
d83c49f3 | 2666 | if (!error) { |
1da177e4 | 2667 | dentry->d_inode->i_flags |= S_DEAD; |
d83c49f3 AV |
2668 | dont_mount(dentry); |
2669 | } | |
1da177e4 LT |
2670 | } |
2671 | } | |
1b1dcc1b | 2672 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 | 2673 | if (!error) { |
1da177e4 LT |
2674 | d_delete(dentry); |
2675 | } | |
2676 | dput(dentry); | |
2677 | ||
2678 | return error; | |
2679 | } | |
2680 | ||
5590ff0d | 2681 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2682 | { |
2683 | int error = 0; | |
2684 | char * name; | |
2685 | struct dentry *dentry; | |
2686 | struct nameidata nd; | |
2687 | ||
2ad94ae6 | 2688 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2689 | if (error) |
2ad94ae6 | 2690 | return error; |
1da177e4 LT |
2691 | |
2692 | switch(nd.last_type) { | |
0612d9fb OH |
2693 | case LAST_DOTDOT: |
2694 | error = -ENOTEMPTY; | |
2695 | goto exit1; | |
2696 | case LAST_DOT: | |
2697 | error = -EINVAL; | |
2698 | goto exit1; | |
2699 | case LAST_ROOT: | |
2700 | error = -EBUSY; | |
2701 | goto exit1; | |
1da177e4 | 2702 | } |
0612d9fb OH |
2703 | |
2704 | nd.flags &= ~LOOKUP_PARENT; | |
2705 | ||
4ac91378 | 2706 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2707 | dentry = lookup_hash(&nd); |
1da177e4 | 2708 | error = PTR_ERR(dentry); |
6902d925 DH |
2709 | if (IS_ERR(dentry)) |
2710 | goto exit2; | |
0622753b DH |
2711 | error = mnt_want_write(nd.path.mnt); |
2712 | if (error) | |
2713 | goto exit3; | |
be6d3e56 KT |
2714 | error = security_path_rmdir(&nd.path, dentry); |
2715 | if (error) | |
2716 | goto exit4; | |
4ac91378 | 2717 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2718 | exit4: |
0622753b DH |
2719 | mnt_drop_write(nd.path.mnt); |
2720 | exit3: | |
6902d925 DH |
2721 | dput(dentry); |
2722 | exit2: | |
4ac91378 | 2723 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2724 | exit1: |
1d957f9b | 2725 | path_put(&nd.path); |
1da177e4 LT |
2726 | putname(name); |
2727 | return error; | |
2728 | } | |
2729 | ||
3cdad428 | 2730 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2731 | { |
2732 | return do_rmdir(AT_FDCWD, pathname); | |
2733 | } | |
2734 | ||
1da177e4 LT |
2735 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2736 | { | |
2737 | int error = may_delete(dir, dentry, 0); | |
2738 | ||
2739 | if (error) | |
2740 | return error; | |
2741 | ||
acfa4380 | 2742 | if (!dir->i_op->unlink) |
1da177e4 LT |
2743 | return -EPERM; |
2744 | ||
1b1dcc1b | 2745 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2746 | if (d_mountpoint(dentry)) |
2747 | error = -EBUSY; | |
2748 | else { | |
2749 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 2750 | if (!error) { |
1da177e4 | 2751 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 2752 | if (!error) |
d83c49f3 | 2753 | dont_mount(dentry); |
bec1052e | 2754 | } |
1da177e4 | 2755 | } |
1b1dcc1b | 2756 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2757 | |
2758 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2759 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2760 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2761 | d_delete(dentry); |
1da177e4 | 2762 | } |
0eeca283 | 2763 | |
1da177e4 LT |
2764 | return error; |
2765 | } | |
2766 | ||
2767 | /* | |
2768 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2769 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2770 | * writeout happening, and we don't want to prevent access to the directory |
2771 | * while waiting on the I/O. | |
2772 | */ | |
5590ff0d | 2773 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2774 | { |
2ad94ae6 AV |
2775 | int error; |
2776 | char *name; | |
1da177e4 LT |
2777 | struct dentry *dentry; |
2778 | struct nameidata nd; | |
2779 | struct inode *inode = NULL; | |
2780 | ||
2ad94ae6 | 2781 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2782 | if (error) |
2ad94ae6 AV |
2783 | return error; |
2784 | ||
1da177e4 LT |
2785 | error = -EISDIR; |
2786 | if (nd.last_type != LAST_NORM) | |
2787 | goto exit1; | |
0612d9fb OH |
2788 | |
2789 | nd.flags &= ~LOOKUP_PARENT; | |
2790 | ||
4ac91378 | 2791 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2792 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2793 | error = PTR_ERR(dentry); |
2794 | if (!IS_ERR(dentry)) { | |
2795 | /* Why not before? Because we want correct error value */ | |
2796 | if (nd.last.name[nd.last.len]) | |
2797 | goto slashes; | |
2798 | inode = dentry->d_inode; | |
2799 | if (inode) | |
7de9c6ee | 2800 | ihold(inode); |
0622753b DH |
2801 | error = mnt_want_write(nd.path.mnt); |
2802 | if (error) | |
2803 | goto exit2; | |
be6d3e56 KT |
2804 | error = security_path_unlink(&nd.path, dentry); |
2805 | if (error) | |
2806 | goto exit3; | |
4ac91378 | 2807 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2808 | exit3: |
0622753b | 2809 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2810 | exit2: |
2811 | dput(dentry); | |
2812 | } | |
4ac91378 | 2813 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2814 | if (inode) |
2815 | iput(inode); /* truncate the inode here */ | |
2816 | exit1: | |
1d957f9b | 2817 | path_put(&nd.path); |
1da177e4 LT |
2818 | putname(name); |
2819 | return error; | |
2820 | ||
2821 | slashes: | |
2822 | error = !dentry->d_inode ? -ENOENT : | |
2823 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2824 | goto exit2; | |
2825 | } | |
2826 | ||
2e4d0924 | 2827 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2828 | { |
2829 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2830 | return -EINVAL; | |
2831 | ||
2832 | if (flag & AT_REMOVEDIR) | |
2833 | return do_rmdir(dfd, pathname); | |
2834 | ||
2835 | return do_unlinkat(dfd, pathname); | |
2836 | } | |
2837 | ||
3480b257 | 2838 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2839 | { |
2840 | return do_unlinkat(AT_FDCWD, pathname); | |
2841 | } | |
2842 | ||
db2e747b | 2843 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2844 | { |
a95164d9 | 2845 | int error = may_create(dir, dentry); |
1da177e4 LT |
2846 | |
2847 | if (error) | |
2848 | return error; | |
2849 | ||
acfa4380 | 2850 | if (!dir->i_op->symlink) |
1da177e4 LT |
2851 | return -EPERM; |
2852 | ||
2853 | error = security_inode_symlink(dir, dentry, oldname); | |
2854 | if (error) | |
2855 | return error; | |
2856 | ||
1da177e4 | 2857 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2858 | if (!error) |
f38aa942 | 2859 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2860 | return error; |
2861 | } | |
2862 | ||
2e4d0924 HC |
2863 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2864 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2865 | { |
2ad94ae6 AV |
2866 | int error; |
2867 | char *from; | |
2868 | char *to; | |
6902d925 DH |
2869 | struct dentry *dentry; |
2870 | struct nameidata nd; | |
1da177e4 LT |
2871 | |
2872 | from = getname(oldname); | |
2ad94ae6 | 2873 | if (IS_ERR(from)) |
1da177e4 | 2874 | return PTR_ERR(from); |
1da177e4 | 2875 | |
2ad94ae6 | 2876 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2877 | if (error) |
2ad94ae6 AV |
2878 | goto out_putname; |
2879 | ||
6902d925 DH |
2880 | dentry = lookup_create(&nd, 0); |
2881 | error = PTR_ERR(dentry); | |
2882 | if (IS_ERR(dentry)) | |
2883 | goto out_unlock; | |
2884 | ||
75c3f29d DH |
2885 | error = mnt_want_write(nd.path.mnt); |
2886 | if (error) | |
2887 | goto out_dput; | |
be6d3e56 KT |
2888 | error = security_path_symlink(&nd.path, dentry, from); |
2889 | if (error) | |
2890 | goto out_drop_write; | |
db2e747b | 2891 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2892 | out_drop_write: |
75c3f29d DH |
2893 | mnt_drop_write(nd.path.mnt); |
2894 | out_dput: | |
6902d925 DH |
2895 | dput(dentry); |
2896 | out_unlock: | |
4ac91378 | 2897 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2898 | path_put(&nd.path); |
6902d925 DH |
2899 | putname(to); |
2900 | out_putname: | |
1da177e4 LT |
2901 | putname(from); |
2902 | return error; | |
2903 | } | |
2904 | ||
3480b257 | 2905 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2906 | { |
2907 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2908 | } | |
2909 | ||
1da177e4 LT |
2910 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2911 | { | |
2912 | struct inode *inode = old_dentry->d_inode; | |
2913 | int error; | |
2914 | ||
2915 | if (!inode) | |
2916 | return -ENOENT; | |
2917 | ||
a95164d9 | 2918 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2919 | if (error) |
2920 | return error; | |
2921 | ||
2922 | if (dir->i_sb != inode->i_sb) | |
2923 | return -EXDEV; | |
2924 | ||
2925 | /* | |
2926 | * A link to an append-only or immutable file cannot be created. | |
2927 | */ | |
2928 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2929 | return -EPERM; | |
acfa4380 | 2930 | if (!dir->i_op->link) |
1da177e4 | 2931 | return -EPERM; |
7e79eedb | 2932 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2933 | return -EPERM; |
2934 | ||
2935 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2936 | if (error) | |
2937 | return error; | |
2938 | ||
7e79eedb | 2939 | mutex_lock(&inode->i_mutex); |
1da177e4 | 2940 | error = dir->i_op->link(old_dentry, dir, new_dentry); |
7e79eedb | 2941 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2942 | if (!error) |
7e79eedb | 2943 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2944 | return error; |
2945 | } | |
2946 | ||
2947 | /* | |
2948 | * Hardlinks are often used in delicate situations. We avoid | |
2949 | * security-related surprises by not following symlinks on the | |
2950 | * newname. --KAB | |
2951 | * | |
2952 | * We don't follow them on the oldname either to be compatible | |
2953 | * with linux 2.0, and to avoid hard-linking to directories | |
2954 | * and other special files. --ADM | |
2955 | */ | |
2e4d0924 HC |
2956 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2957 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2958 | { |
2959 | struct dentry *new_dentry; | |
2d8f3038 AV |
2960 | struct nameidata nd; |
2961 | struct path old_path; | |
1da177e4 | 2962 | int error; |
2ad94ae6 | 2963 | char *to; |
1da177e4 | 2964 | |
45c9b11a | 2965 | if ((flags & ~AT_SYMLINK_FOLLOW) != 0) |
c04030e1 UD |
2966 | return -EINVAL; |
2967 | ||
2d8f3038 AV |
2968 | error = user_path_at(olddfd, oldname, |
2969 | flags & AT_SYMLINK_FOLLOW ? LOOKUP_FOLLOW : 0, | |
2970 | &old_path); | |
1da177e4 | 2971 | if (error) |
2ad94ae6 AV |
2972 | return error; |
2973 | ||
2974 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2975 | if (error) |
2976 | goto out; | |
2977 | error = -EXDEV; | |
2d8f3038 | 2978 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2979 | goto out_release; |
2980 | new_dentry = lookup_create(&nd, 0); | |
2981 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2982 | if (IS_ERR(new_dentry)) |
2983 | goto out_unlock; | |
75c3f29d DH |
2984 | error = mnt_want_write(nd.path.mnt); |
2985 | if (error) | |
2986 | goto out_dput; | |
be6d3e56 KT |
2987 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2988 | if (error) | |
2989 | goto out_drop_write; | |
2d8f3038 | 2990 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2991 | out_drop_write: |
75c3f29d DH |
2992 | mnt_drop_write(nd.path.mnt); |
2993 | out_dput: | |
6902d925 DH |
2994 | dput(new_dentry); |
2995 | out_unlock: | |
4ac91378 | 2996 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2997 | out_release: |
1d957f9b | 2998 | path_put(&nd.path); |
2ad94ae6 | 2999 | putname(to); |
1da177e4 | 3000 | out: |
2d8f3038 | 3001 | path_put(&old_path); |
1da177e4 LT |
3002 | |
3003 | return error; | |
3004 | } | |
3005 | ||
3480b257 | 3006 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 3007 | { |
c04030e1 | 3008 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
3009 | } |
3010 | ||
1da177e4 LT |
3011 | /* |
3012 | * The worst of all namespace operations - renaming directory. "Perverted" | |
3013 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
3014 | * Problems: | |
3015 | * a) we can get into loop creation. Check is done in is_subdir(). | |
3016 | * b) race potential - two innocent renames can create a loop together. | |
3017 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 3018 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
3019 | * story. |
3020 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 3021 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
3022 | * whether the target exists). Solution: try to be smart with locking |
3023 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 3024 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
3025 | * move will be locked. Thus we can rank directories by the tree |
3026 | * (ancestors first) and rank all non-directories after them. | |
3027 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 3028 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
3029 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
3030 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
3031 | * we'd better make sure that there's no link(2) for them. | |
3032 | * d) some filesystems don't support opened-but-unlinked directories, | |
3033 | * either because of layout or because they are not ready to deal with | |
3034 | * all cases correctly. The latter will be fixed (taking this sort of | |
3035 | * stuff into VFS), but the former is not going away. Solution: the same | |
3036 | * trick as in rmdir(). | |
3037 | * e) conversion from fhandle to dentry may come in the wrong moment - when | |
1b1dcc1b | 3038 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 3039 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 3040 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
3041 | * locking]. |
3042 | */ | |
75c96f85 AB |
3043 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
3044 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3045 | { |
3046 | int error = 0; | |
3047 | struct inode *target; | |
3048 | ||
3049 | /* | |
3050 | * If we are going to change the parent - check write permissions, | |
3051 | * we'll need to flip '..'. | |
3052 | */ | |
3053 | if (new_dir != old_dir) { | |
f419a2e3 | 3054 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3055 | if (error) |
3056 | return error; | |
3057 | } | |
3058 | ||
3059 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3060 | if (error) | |
3061 | return error; | |
3062 | ||
3063 | target = new_dentry->d_inode; | |
d83c49f3 | 3064 | if (target) |
1b1dcc1b | 3065 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
3066 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
3067 | error = -EBUSY; | |
d83c49f3 AV |
3068 | else { |
3069 | if (target) | |
3070 | dentry_unhash(new_dentry); | |
1da177e4 | 3071 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
d83c49f3 | 3072 | } |
1da177e4 | 3073 | if (target) { |
d83c49f3 | 3074 | if (!error) { |
1da177e4 | 3075 | target->i_flags |= S_DEAD; |
d83c49f3 AV |
3076 | dont_mount(new_dentry); |
3077 | } | |
1b1dcc1b | 3078 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3079 | if (d_unhashed(new_dentry)) |
3080 | d_rehash(new_dentry); | |
3081 | dput(new_dentry); | |
3082 | } | |
e31e14ec | 3083 | if (!error) |
349457cc MF |
3084 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3085 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3086 | return error; |
3087 | } | |
3088 | ||
75c96f85 AB |
3089 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3090 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3091 | { |
3092 | struct inode *target; | |
3093 | int error; | |
3094 | ||
3095 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3096 | if (error) | |
3097 | return error; | |
3098 | ||
3099 | dget(new_dentry); | |
3100 | target = new_dentry->d_inode; | |
3101 | if (target) | |
1b1dcc1b | 3102 | mutex_lock(&target->i_mutex); |
1da177e4 LT |
3103 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
3104 | error = -EBUSY; | |
3105 | else | |
3106 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3107 | if (!error) { | |
bec1052e | 3108 | if (target) |
d83c49f3 | 3109 | dont_mount(new_dentry); |
349457cc | 3110 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
1da177e4 | 3111 | d_move(old_dentry, new_dentry); |
1da177e4 LT |
3112 | } |
3113 | if (target) | |
1b1dcc1b | 3114 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3115 | dput(new_dentry); |
3116 | return error; | |
3117 | } | |
3118 | ||
3119 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3120 | struct inode *new_dir, struct dentry *new_dentry) | |
3121 | { | |
3122 | int error; | |
3123 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3124 | const unsigned char *old_name; |
1da177e4 LT |
3125 | |
3126 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3127 | return 0; | |
3128 | ||
3129 | error = may_delete(old_dir, old_dentry, is_dir); | |
3130 | if (error) | |
3131 | return error; | |
3132 | ||
3133 | if (!new_dentry->d_inode) | |
a95164d9 | 3134 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3135 | else |
3136 | error = may_delete(new_dir, new_dentry, is_dir); | |
3137 | if (error) | |
3138 | return error; | |
3139 | ||
acfa4380 | 3140 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3141 | return -EPERM; |
3142 | ||
0eeca283 RL |
3143 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3144 | ||
1da177e4 LT |
3145 | if (is_dir) |
3146 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3147 | else | |
3148 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3149 | if (!error) |
3150 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3151 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3152 | fsnotify_oldname_free(old_name); |
3153 | ||
1da177e4 LT |
3154 | return error; |
3155 | } | |
3156 | ||
2e4d0924 HC |
3157 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3158 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3159 | { |
2ad94ae6 AV |
3160 | struct dentry *old_dir, *new_dir; |
3161 | struct dentry *old_dentry, *new_dentry; | |
3162 | struct dentry *trap; | |
1da177e4 | 3163 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3164 | char *from; |
3165 | char *to; | |
3166 | int error; | |
1da177e4 | 3167 | |
2ad94ae6 | 3168 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3169 | if (error) |
3170 | goto exit; | |
3171 | ||
2ad94ae6 | 3172 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3173 | if (error) |
3174 | goto exit1; | |
3175 | ||
3176 | error = -EXDEV; | |
4ac91378 | 3177 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3178 | goto exit2; |
3179 | ||
4ac91378 | 3180 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3181 | error = -EBUSY; |
3182 | if (oldnd.last_type != LAST_NORM) | |
3183 | goto exit2; | |
3184 | ||
4ac91378 | 3185 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3186 | if (newnd.last_type != LAST_NORM) |
3187 | goto exit2; | |
3188 | ||
0612d9fb OH |
3189 | oldnd.flags &= ~LOOKUP_PARENT; |
3190 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3191 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3192 | |
1da177e4 LT |
3193 | trap = lock_rename(new_dir, old_dir); |
3194 | ||
49705b77 | 3195 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3196 | error = PTR_ERR(old_dentry); |
3197 | if (IS_ERR(old_dentry)) | |
3198 | goto exit3; | |
3199 | /* source must exist */ | |
3200 | error = -ENOENT; | |
3201 | if (!old_dentry->d_inode) | |
3202 | goto exit4; | |
3203 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3204 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3205 | error = -ENOTDIR; | |
3206 | if (oldnd.last.name[oldnd.last.len]) | |
3207 | goto exit4; | |
3208 | if (newnd.last.name[newnd.last.len]) | |
3209 | goto exit4; | |
3210 | } | |
3211 | /* source should not be ancestor of target */ | |
3212 | error = -EINVAL; | |
3213 | if (old_dentry == trap) | |
3214 | goto exit4; | |
49705b77 | 3215 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3216 | error = PTR_ERR(new_dentry); |
3217 | if (IS_ERR(new_dentry)) | |
3218 | goto exit4; | |
3219 | /* target should not be an ancestor of source */ | |
3220 | error = -ENOTEMPTY; | |
3221 | if (new_dentry == trap) | |
3222 | goto exit5; | |
3223 | ||
9079b1eb DH |
3224 | error = mnt_want_write(oldnd.path.mnt); |
3225 | if (error) | |
3226 | goto exit5; | |
be6d3e56 KT |
3227 | error = security_path_rename(&oldnd.path, old_dentry, |
3228 | &newnd.path, new_dentry); | |
3229 | if (error) | |
3230 | goto exit6; | |
1da177e4 LT |
3231 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3232 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 3233 | exit6: |
9079b1eb | 3234 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
3235 | exit5: |
3236 | dput(new_dentry); | |
3237 | exit4: | |
3238 | dput(old_dentry); | |
3239 | exit3: | |
3240 | unlock_rename(new_dir, old_dir); | |
3241 | exit2: | |
1d957f9b | 3242 | path_put(&newnd.path); |
2ad94ae6 | 3243 | putname(to); |
1da177e4 | 3244 | exit1: |
1d957f9b | 3245 | path_put(&oldnd.path); |
1da177e4 | 3246 | putname(from); |
2ad94ae6 | 3247 | exit: |
1da177e4 LT |
3248 | return error; |
3249 | } | |
3250 | ||
a26eab24 | 3251 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3252 | { |
3253 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3254 | } | |
3255 | ||
1da177e4 LT |
3256 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3257 | { | |
3258 | int len; | |
3259 | ||
3260 | len = PTR_ERR(link); | |
3261 | if (IS_ERR(link)) | |
3262 | goto out; | |
3263 | ||
3264 | len = strlen(link); | |
3265 | if (len > (unsigned) buflen) | |
3266 | len = buflen; | |
3267 | if (copy_to_user(buffer, link, len)) | |
3268 | len = -EFAULT; | |
3269 | out: | |
3270 | return len; | |
3271 | } | |
3272 | ||
3273 | /* | |
3274 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3275 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3276 | * using) it for any given inode is up to filesystem. | |
3277 | */ | |
3278 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3279 | { | |
3280 | struct nameidata nd; | |
cc314eef | 3281 | void *cookie; |
694a1764 | 3282 | int res; |
cc314eef | 3283 | |
1da177e4 | 3284 | nd.depth = 0; |
cc314eef | 3285 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3286 | if (IS_ERR(cookie)) |
3287 | return PTR_ERR(cookie); | |
3288 | ||
3289 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3290 | if (dentry->d_inode->i_op->put_link) | |
3291 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3292 | return res; | |
1da177e4 LT |
3293 | } |
3294 | ||
3295 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3296 | { | |
3297 | return __vfs_follow_link(nd, link); | |
3298 | } | |
3299 | ||
3300 | /* get the link contents into pagecache */ | |
3301 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3302 | { | |
ebd09abb DG |
3303 | char *kaddr; |
3304 | struct page *page; | |
1da177e4 | 3305 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3306 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3307 | if (IS_ERR(page)) |
6fe6900e | 3308 | return (char*)page; |
1da177e4 | 3309 | *ppage = page; |
ebd09abb DG |
3310 | kaddr = kmap(page); |
3311 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3312 | return kaddr; | |
1da177e4 LT |
3313 | } |
3314 | ||
3315 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3316 | { | |
3317 | struct page *page = NULL; | |
3318 | char *s = page_getlink(dentry, &page); | |
3319 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3320 | if (page) { | |
3321 | kunmap(page); | |
3322 | page_cache_release(page); | |
3323 | } | |
3324 | return res; | |
3325 | } | |
3326 | ||
cc314eef | 3327 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3328 | { |
cc314eef | 3329 | struct page *page = NULL; |
1da177e4 | 3330 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3331 | return page; |
1da177e4 LT |
3332 | } |
3333 | ||
cc314eef | 3334 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3335 | { |
cc314eef LT |
3336 | struct page *page = cookie; |
3337 | ||
3338 | if (page) { | |
1da177e4 LT |
3339 | kunmap(page); |
3340 | page_cache_release(page); | |
1da177e4 LT |
3341 | } |
3342 | } | |
3343 | ||
54566b2c NP |
3344 | /* |
3345 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3346 | */ | |
3347 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3348 | { |
3349 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3350 | struct page *page; |
afddba49 | 3351 | void *fsdata; |
beb497ab | 3352 | int err; |
1da177e4 | 3353 | char *kaddr; |
54566b2c NP |
3354 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3355 | if (nofs) | |
3356 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3357 | |
7e53cac4 | 3358 | retry: |
afddba49 | 3359 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3360 | flags, &page, &fsdata); |
1da177e4 | 3361 | if (err) |
afddba49 NP |
3362 | goto fail; |
3363 | ||
1da177e4 LT |
3364 | kaddr = kmap_atomic(page, KM_USER0); |
3365 | memcpy(kaddr, symname, len-1); | |
3366 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
3367 | |
3368 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3369 | page, fsdata); | |
1da177e4 LT |
3370 | if (err < 0) |
3371 | goto fail; | |
afddba49 NP |
3372 | if (err < len-1) |
3373 | goto retry; | |
3374 | ||
1da177e4 LT |
3375 | mark_inode_dirty(inode); |
3376 | return 0; | |
1da177e4 LT |
3377 | fail: |
3378 | return err; | |
3379 | } | |
3380 | ||
0adb25d2 KK |
3381 | int page_symlink(struct inode *inode, const char *symname, int len) |
3382 | { | |
3383 | return __page_symlink(inode, symname, len, | |
54566b2c | 3384 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3385 | } |
3386 | ||
92e1d5be | 3387 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3388 | .readlink = generic_readlink, |
3389 | .follow_link = page_follow_link_light, | |
3390 | .put_link = page_put_link, | |
3391 | }; | |
3392 | ||
2d8f3038 | 3393 | EXPORT_SYMBOL(user_path_at); |
1da177e4 LT |
3394 | EXPORT_SYMBOL(follow_down); |
3395 | EXPORT_SYMBOL(follow_up); | |
3396 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
3397 | EXPORT_SYMBOL(getname); | |
3398 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3399 | EXPORT_SYMBOL(lookup_one_len); |
3400 | EXPORT_SYMBOL(page_follow_link_light); | |
3401 | EXPORT_SYMBOL(page_put_link); | |
3402 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3403 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3404 | EXPORT_SYMBOL(page_symlink); |
3405 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
3406 | EXPORT_SYMBOL(path_lookup); | |
d1811465 | 3407 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3408 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3409 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 3410 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
3411 | EXPORT_SYMBOL(unlock_rename); |
3412 | EXPORT_SYMBOL(vfs_create); | |
3413 | EXPORT_SYMBOL(vfs_follow_link); | |
3414 | EXPORT_SYMBOL(vfs_link); | |
3415 | EXPORT_SYMBOL(vfs_mkdir); | |
3416 | EXPORT_SYMBOL(vfs_mknod); | |
3417 | EXPORT_SYMBOL(generic_permission); | |
3418 | EXPORT_SYMBOL(vfs_readlink); | |
3419 | EXPORT_SYMBOL(vfs_rename); | |
3420 | EXPORT_SYMBOL(vfs_rmdir); | |
3421 | EXPORT_SYMBOL(vfs_symlink); | |
3422 | EXPORT_SYMBOL(vfs_unlink); | |
3423 | EXPORT_SYMBOL(dentry_unhash); | |
3424 | EXPORT_SYMBOL(generic_readlink); |