]>
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 | |
25985edc | 73 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
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 | ||
f52e0c11 | 139 | static char *getname_flags(const char __user * filename, int flags) |
1da177e4 LT |
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) { | |
f52e0c11 AV |
150 | if (retval != -ENOENT || !(flags & LOOKUP_EMPTY)) { |
151 | __putname(tmp); | |
152 | result = ERR_PTR(retval); | |
153 | } | |
1da177e4 LT |
154 | } |
155 | } | |
156 | audit_getname(result); | |
157 | return result; | |
158 | } | |
159 | ||
f52e0c11 AV |
160 | char *getname(const char __user * filename) |
161 | { | |
162 | return getname_flags(filename, 0); | |
163 | } | |
164 | ||
1da177e4 LT |
165 | #ifdef CONFIG_AUDITSYSCALL |
166 | void putname(const char *name) | |
167 | { | |
5ac3a9c2 | 168 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
169 | audit_putname(name); |
170 | else | |
171 | __putname(name); | |
172 | } | |
173 | EXPORT_SYMBOL(putname); | |
174 | #endif | |
175 | ||
5909ccaa LT |
176 | /* |
177 | * This does basic POSIX ACL permission checking | |
1da177e4 | 178 | */ |
b74c79e9 NP |
179 | static int acl_permission_check(struct inode *inode, int mask, unsigned int flags, |
180 | int (*check_acl)(struct inode *inode, int mask, unsigned int flags)) | |
1da177e4 | 181 | { |
26cf46be | 182 | unsigned int mode = inode->i_mode; |
1da177e4 | 183 | |
e6305c43 AV |
184 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
185 | ||
e795b717 SH |
186 | if (current_user_ns() != inode_userns(inode)) |
187 | goto other_perms; | |
188 | ||
da9592ed | 189 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
190 | mode >>= 6; |
191 | else { | |
192 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { | |
b74c79e9 NP |
193 | int error = check_acl(inode, mask, flags); |
194 | if (error != -EAGAIN) | |
195 | return error; | |
1da177e4 LT |
196 | } |
197 | ||
198 | if (in_group_p(inode->i_gid)) | |
199 | mode >>= 3; | |
200 | } | |
201 | ||
e795b717 | 202 | other_perms: |
1da177e4 LT |
203 | /* |
204 | * If the DACs are ok we don't need any capability check. | |
205 | */ | |
e6305c43 | 206 | if ((mask & ~mode) == 0) |
1da177e4 | 207 | return 0; |
5909ccaa LT |
208 | return -EACCES; |
209 | } | |
210 | ||
211 | /** | |
b74c79e9 | 212 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa LT |
213 | * @inode: inode to check access rights for |
214 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
215 | * @check_acl: optional callback to check for Posix ACLs | |
39191628 | 216 | * @flags: IPERM_FLAG_ flags. |
5909ccaa LT |
217 | * |
218 | * Used to check for read/write/execute permissions on a file. | |
219 | * We use "fsuid" for this, letting us set arbitrary permissions | |
220 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
221 | * are used for other things. |
222 | * | |
223 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
224 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
225 | * It would then be called again in ref-walk mode. | |
5909ccaa | 226 | */ |
b74c79e9 NP |
227 | int generic_permission(struct inode *inode, int mask, unsigned int flags, |
228 | int (*check_acl)(struct inode *inode, int mask, unsigned int flags)) | |
5909ccaa LT |
229 | { |
230 | int ret; | |
231 | ||
232 | /* | |
233 | * Do the basic POSIX ACL permission checks. | |
234 | */ | |
b74c79e9 | 235 | ret = acl_permission_check(inode, mask, flags, check_acl); |
5909ccaa LT |
236 | if (ret != -EACCES) |
237 | return ret; | |
1da177e4 | 238 | |
1da177e4 LT |
239 | /* |
240 | * Read/write DACs are always overridable. | |
8e833fd2 AV |
241 | * Executable DACs are overridable for all directories and |
242 | * for non-directories that have least one exec bit set. | |
1da177e4 | 243 | */ |
f696a365 | 244 | if (!(mask & MAY_EXEC) || execute_ok(inode)) |
e795b717 | 245 | if (ns_capable(inode_userns(inode), CAP_DAC_OVERRIDE)) |
1da177e4 LT |
246 | return 0; |
247 | ||
248 | /* | |
249 | * Searching includes executable on directories, else just read. | |
250 | */ | |
7ea66001 | 251 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
1da177e4 | 252 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) |
e795b717 | 253 | if (ns_capable(inode_userns(inode), CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
254 | return 0; |
255 | ||
256 | return -EACCES; | |
257 | } | |
258 | ||
cb23beb5 CH |
259 | /** |
260 | * inode_permission - check for access rights to a given inode | |
261 | * @inode: inode to check permission on | |
262 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
263 | * | |
264 | * Used to check for read/write/execute permissions on an inode. | |
265 | * We use "fsuid" for this, letting us set arbitrary permissions | |
266 | * for filesystem access without changing the "normal" uids which | |
267 | * are used for other things. | |
268 | */ | |
f419a2e3 | 269 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 270 | { |
e6305c43 | 271 | int retval; |
1da177e4 LT |
272 | |
273 | if (mask & MAY_WRITE) { | |
22590e41 | 274 | umode_t mode = inode->i_mode; |
1da177e4 LT |
275 | |
276 | /* | |
277 | * Nobody gets write access to a read-only fs. | |
278 | */ | |
279 | if (IS_RDONLY(inode) && | |
280 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
281 | return -EROFS; | |
282 | ||
283 | /* | |
284 | * Nobody gets write access to an immutable file. | |
285 | */ | |
286 | if (IS_IMMUTABLE(inode)) | |
287 | return -EACCES; | |
288 | } | |
289 | ||
acfa4380 | 290 | if (inode->i_op->permission) |
b74c79e9 | 291 | retval = inode->i_op->permission(inode, mask, 0); |
f696a365 | 292 | else |
b74c79e9 NP |
293 | retval = generic_permission(inode, mask, 0, |
294 | inode->i_op->check_acl); | |
f696a365 | 295 | |
1da177e4 LT |
296 | if (retval) |
297 | return retval; | |
298 | ||
08ce5f16 SH |
299 | retval = devcgroup_inode_permission(inode, mask); |
300 | if (retval) | |
301 | return retval; | |
302 | ||
d09ca739 | 303 | return security_inode_permission(inode, mask); |
1da177e4 LT |
304 | } |
305 | ||
8c744fb8 CH |
306 | /** |
307 | * file_permission - check for additional access rights to a given file | |
308 | * @file: file to check access rights for | |
309 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
310 | * | |
311 | * Used to check for read/write/execute permissions on an already opened | |
312 | * file. | |
313 | * | |
314 | * Note: | |
315 | * Do not use this function in new code. All access checks should | |
cb23beb5 | 316 | * be done using inode_permission(). |
8c744fb8 CH |
317 | */ |
318 | int file_permission(struct file *file, int mask) | |
319 | { | |
f419a2e3 | 320 | return inode_permission(file->f_path.dentry->d_inode, mask); |
8c744fb8 CH |
321 | } |
322 | ||
1da177e4 LT |
323 | /* |
324 | * get_write_access() gets write permission for a file. | |
325 | * put_write_access() releases this write permission. | |
326 | * This is used for regular files. | |
327 | * We cannot support write (and maybe mmap read-write shared) accesses and | |
328 | * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode | |
329 | * can have the following values: | |
330 | * 0: no writers, no VM_DENYWRITE mappings | |
331 | * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist | |
332 | * > 0: (i_writecount) users are writing to the file. | |
333 | * | |
334 | * Normally we operate on that counter with atomic_{inc,dec} and it's safe | |
335 | * except for the cases where we don't hold i_writecount yet. Then we need to | |
336 | * use {get,deny}_write_access() - these functions check the sign and refuse | |
337 | * to do the change if sign is wrong. Exclusion between them is provided by | |
338 | * the inode->i_lock spinlock. | |
339 | */ | |
340 | ||
341 | int get_write_access(struct inode * inode) | |
342 | { | |
343 | spin_lock(&inode->i_lock); | |
344 | if (atomic_read(&inode->i_writecount) < 0) { | |
345 | spin_unlock(&inode->i_lock); | |
346 | return -ETXTBSY; | |
347 | } | |
348 | atomic_inc(&inode->i_writecount); | |
349 | spin_unlock(&inode->i_lock); | |
350 | ||
351 | return 0; | |
352 | } | |
353 | ||
354 | int deny_write_access(struct file * file) | |
355 | { | |
0f7fc9e4 | 356 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
357 | |
358 | spin_lock(&inode->i_lock); | |
359 | if (atomic_read(&inode->i_writecount) > 0) { | |
360 | spin_unlock(&inode->i_lock); | |
361 | return -ETXTBSY; | |
362 | } | |
363 | atomic_dec(&inode->i_writecount); | |
364 | spin_unlock(&inode->i_lock); | |
365 | ||
366 | return 0; | |
367 | } | |
368 | ||
5dd784d0 JB |
369 | /** |
370 | * path_get - get a reference to a path | |
371 | * @path: path to get the reference to | |
372 | * | |
373 | * Given a path increment the reference count to the dentry and the vfsmount. | |
374 | */ | |
375 | void path_get(struct path *path) | |
376 | { | |
377 | mntget(path->mnt); | |
378 | dget(path->dentry); | |
379 | } | |
380 | EXPORT_SYMBOL(path_get); | |
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 | |
19660af7 | 395 | /* |
31e6b01f | 396 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
397 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
398 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
399 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
400 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
401 | * got stuck, so ref-walk may continue from there. If this is not successful | |
402 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
403 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 404 | */ |
31e6b01f NP |
405 | |
406 | /** | |
19660af7 AV |
407 | * unlazy_walk - try to switch to ref-walk mode. |
408 | * @nd: nameidata pathwalk data | |
409 | * @dentry: child of nd->path.dentry or NULL | |
39191628 | 410 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 411 | * |
19660af7 AV |
412 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
413 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
414 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 415 | */ |
19660af7 | 416 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry) |
31e6b01f NP |
417 | { |
418 | struct fs_struct *fs = current->fs; | |
419 | struct dentry *parent = nd->path.dentry; | |
5b6ca027 | 420 | int want_root = 0; |
31e6b01f NP |
421 | |
422 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
5b6ca027 AV |
423 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
424 | want_root = 1; | |
31e6b01f NP |
425 | spin_lock(&fs->lock); |
426 | if (nd->root.mnt != fs->root.mnt || | |
427 | nd->root.dentry != fs->root.dentry) | |
428 | goto err_root; | |
429 | } | |
430 | spin_lock(&parent->d_lock); | |
19660af7 AV |
431 | if (!dentry) { |
432 | if (!__d_rcu_to_refcount(parent, nd->seq)) | |
433 | goto err_parent; | |
434 | BUG_ON(nd->inode != parent->d_inode); | |
435 | } else { | |
94c0d4ec AV |
436 | if (dentry->d_parent != parent) |
437 | goto err_parent; | |
19660af7 AV |
438 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); |
439 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
440 | goto err_child; | |
441 | /* | |
442 | * If the sequence check on the child dentry passed, then | |
443 | * the child has not been removed from its parent. This | |
444 | * means the parent dentry must be valid and able to take | |
445 | * a reference at this point. | |
446 | */ | |
447 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
448 | BUG_ON(!parent->d_count); | |
449 | parent->d_count++; | |
450 | spin_unlock(&dentry->d_lock); | |
451 | } | |
31e6b01f | 452 | spin_unlock(&parent->d_lock); |
5b6ca027 | 453 | if (want_root) { |
31e6b01f NP |
454 | path_get(&nd->root); |
455 | spin_unlock(&fs->lock); | |
456 | } | |
457 | mntget(nd->path.mnt); | |
458 | ||
459 | rcu_read_unlock(); | |
460 | br_read_unlock(vfsmount_lock); | |
461 | nd->flags &= ~LOOKUP_RCU; | |
462 | return 0; | |
19660af7 AV |
463 | |
464 | err_child: | |
31e6b01f | 465 | spin_unlock(&dentry->d_lock); |
19660af7 | 466 | err_parent: |
31e6b01f NP |
467 | spin_unlock(&parent->d_lock); |
468 | err_root: | |
5b6ca027 | 469 | if (want_root) |
31e6b01f NP |
470 | spin_unlock(&fs->lock); |
471 | return -ECHILD; | |
472 | } | |
473 | ||
834f2a4a TM |
474 | /** |
475 | * release_open_intent - free up open intent resources | |
476 | * @nd: pointer to nameidata | |
477 | */ | |
478 | void release_open_intent(struct nameidata *nd) | |
479 | { | |
2dab5974 LT |
480 | struct file *file = nd->intent.open.file; |
481 | ||
482 | if (file && !IS_ERR(file)) { | |
483 | if (file->f_path.dentry == NULL) | |
484 | put_filp(file); | |
485 | else | |
486 | fput(file); | |
487 | } | |
834f2a4a TM |
488 | } |
489 | ||
f60aef7e | 490 | static inline int d_revalidate(struct dentry *dentry, struct nameidata *nd) |
34286d66 | 491 | { |
f60aef7e | 492 | return dentry->d_op->d_revalidate(dentry, nd); |
34286d66 NP |
493 | } |
494 | ||
f5e1c1c1 | 495 | static struct dentry * |
bcdc5e01 IK |
496 | do_revalidate(struct dentry *dentry, struct nameidata *nd) |
497 | { | |
f5e1c1c1 | 498 | int status = d_revalidate(dentry, nd); |
bcdc5e01 IK |
499 | if (unlikely(status <= 0)) { |
500 | /* | |
501 | * The dentry failed validation. | |
502 | * If d_revalidate returned 0 attempt to invalidate | |
503 | * the dentry otherwise d_revalidate is asking us | |
504 | * to return a fail status. | |
505 | */ | |
34286d66 | 506 | if (status < 0) { |
f5e1c1c1 | 507 | dput(dentry); |
34286d66 | 508 | dentry = ERR_PTR(status); |
f5e1c1c1 AV |
509 | } else if (!d_invalidate(dentry)) { |
510 | dput(dentry); | |
511 | dentry = NULL; | |
bcdc5e01 IK |
512 | } |
513 | } | |
514 | return dentry; | |
515 | } | |
516 | ||
9f1fafee AV |
517 | /** |
518 | * complete_walk - successful completion of path walk | |
519 | * @nd: pointer nameidata | |
39159de2 | 520 | * |
9f1fafee AV |
521 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
522 | * Revalidate the final result, unless we'd already done that during | |
523 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
524 | * success, -error on failure. In case of failure caller does not | |
525 | * need to drop nd->path. | |
39159de2 | 526 | */ |
9f1fafee | 527 | static int complete_walk(struct nameidata *nd) |
39159de2 | 528 | { |
16c2cd71 | 529 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 530 | int status; |
39159de2 | 531 | |
9f1fafee AV |
532 | if (nd->flags & LOOKUP_RCU) { |
533 | nd->flags &= ~LOOKUP_RCU; | |
534 | if (!(nd->flags & LOOKUP_ROOT)) | |
535 | nd->root.mnt = NULL; | |
536 | spin_lock(&dentry->d_lock); | |
537 | if (unlikely(!__d_rcu_to_refcount(dentry, nd->seq))) { | |
538 | spin_unlock(&dentry->d_lock); | |
539 | rcu_read_unlock(); | |
540 | br_read_unlock(vfsmount_lock); | |
541 | return -ECHILD; | |
542 | } | |
543 | BUG_ON(nd->inode != dentry->d_inode); | |
544 | spin_unlock(&dentry->d_lock); | |
545 | mntget(nd->path.mnt); | |
546 | rcu_read_unlock(); | |
547 | br_read_unlock(vfsmount_lock); | |
548 | } | |
549 | ||
16c2cd71 AV |
550 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
551 | return 0; | |
552 | ||
553 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
39159de2 JL |
554 | return 0; |
555 | ||
16c2cd71 AV |
556 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) |
557 | return 0; | |
558 | ||
559 | /* Note: we do not d_invalidate() */ | |
34286d66 | 560 | status = d_revalidate(dentry, nd); |
39159de2 JL |
561 | if (status > 0) |
562 | return 0; | |
563 | ||
16c2cd71 | 564 | if (!status) |
39159de2 | 565 | status = -ESTALE; |
16c2cd71 | 566 | |
9f1fafee | 567 | path_put(&nd->path); |
39159de2 JL |
568 | return status; |
569 | } | |
570 | ||
1da177e4 | 571 | /* |
b75b5086 AV |
572 | * Short-cut version of permission(), for calling on directories |
573 | * during pathname resolution. Combines parts of permission() | |
574 | * and generic_permission(), and tests ONLY for MAY_EXEC permission. | |
1da177e4 LT |
575 | * |
576 | * If appropriate, check DAC only. If not appropriate, or | |
b75b5086 | 577 | * short-cut DAC fails, then call ->permission() to do more |
1da177e4 LT |
578 | * complete permission check. |
579 | */ | |
b74c79e9 | 580 | static inline int exec_permission(struct inode *inode, unsigned int flags) |
1da177e4 | 581 | { |
5909ccaa | 582 | int ret; |
e795b717 | 583 | struct user_namespace *ns = inode_userns(inode); |
1da177e4 | 584 | |
cb9179ea | 585 | if (inode->i_op->permission) { |
b74c79e9 NP |
586 | ret = inode->i_op->permission(inode, MAY_EXEC, flags); |
587 | } else { | |
588 | ret = acl_permission_check(inode, MAY_EXEC, flags, | |
589 | inode->i_op->check_acl); | |
cb9179ea | 590 | } |
b74c79e9 | 591 | if (likely(!ret)) |
1da177e4 | 592 | goto ok; |
b74c79e9 | 593 | if (ret == -ECHILD) |
31e6b01f | 594 | return ret; |
1da177e4 | 595 | |
e795b717 SH |
596 | if (ns_capable(ns, CAP_DAC_OVERRIDE) || |
597 | ns_capable(ns, CAP_DAC_READ_SEARCH)) | |
1da177e4 LT |
598 | goto ok; |
599 | ||
5909ccaa | 600 | return ret; |
1da177e4 | 601 | ok: |
b74c79e9 | 602 | return security_inode_exec_permission(inode, flags); |
1da177e4 LT |
603 | } |
604 | ||
2a737871 AV |
605 | static __always_inline void set_root(struct nameidata *nd) |
606 | { | |
f7ad3c6b MS |
607 | if (!nd->root.mnt) |
608 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
609 | } |
610 | ||
6de88d72 AV |
611 | static int link_path_walk(const char *, struct nameidata *); |
612 | ||
31e6b01f NP |
613 | static __always_inline void set_root_rcu(struct nameidata *nd) |
614 | { | |
615 | if (!nd->root.mnt) { | |
616 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
617 | unsigned seq; |
618 | ||
619 | do { | |
620 | seq = read_seqcount_begin(&fs->seq); | |
621 | nd->root = fs->root; | |
c1530019 | 622 | nd->seq = __read_seqcount_begin(&nd->root.dentry->d_seq); |
c28cc364 | 623 | } while (read_seqcount_retry(&fs->seq, seq)); |
31e6b01f NP |
624 | } |
625 | } | |
626 | ||
f1662356 | 627 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 628 | { |
31e6b01f NP |
629 | int ret; |
630 | ||
1da177e4 LT |
631 | if (IS_ERR(link)) |
632 | goto fail; | |
633 | ||
634 | if (*link == '/') { | |
2a737871 | 635 | set_root(nd); |
1d957f9b | 636 | path_put(&nd->path); |
2a737871 AV |
637 | nd->path = nd->root; |
638 | path_get(&nd->root); | |
16c2cd71 | 639 | nd->flags |= LOOKUP_JUMPED; |
1da177e4 | 640 | } |
31e6b01f | 641 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 642 | |
31e6b01f NP |
643 | ret = link_path_walk(link, nd); |
644 | return ret; | |
1da177e4 | 645 | fail: |
1d957f9b | 646 | path_put(&nd->path); |
1da177e4 LT |
647 | return PTR_ERR(link); |
648 | } | |
649 | ||
1d957f9b | 650 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
651 | { |
652 | dput(path->dentry); | |
4ac91378 | 653 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
654 | mntput(path->mnt); |
655 | } | |
656 | ||
7b9337aa NP |
657 | static inline void path_to_nameidata(const struct path *path, |
658 | struct nameidata *nd) | |
051d3812 | 659 | { |
31e6b01f NP |
660 | if (!(nd->flags & LOOKUP_RCU)) { |
661 | dput(nd->path.dentry); | |
662 | if (nd->path.mnt != path->mnt) | |
663 | mntput(nd->path.mnt); | |
9a229683 | 664 | } |
31e6b01f | 665 | nd->path.mnt = path->mnt; |
4ac91378 | 666 | nd->path.dentry = path->dentry; |
051d3812 IK |
667 | } |
668 | ||
574197e0 AV |
669 | static inline void put_link(struct nameidata *nd, struct path *link, void *cookie) |
670 | { | |
671 | struct inode *inode = link->dentry->d_inode; | |
672 | if (!IS_ERR(cookie) && inode->i_op->put_link) | |
673 | inode->i_op->put_link(link->dentry, nd, cookie); | |
674 | path_put(link); | |
675 | } | |
676 | ||
def4af30 | 677 | static __always_inline int |
574197e0 | 678 | follow_link(struct path *link, struct nameidata *nd, void **p) |
1da177e4 LT |
679 | { |
680 | int error; | |
7b9337aa | 681 | struct dentry *dentry = link->dentry; |
1da177e4 | 682 | |
844a3917 AV |
683 | BUG_ON(nd->flags & LOOKUP_RCU); |
684 | ||
0e794589 AV |
685 | if (link->mnt == nd->path.mnt) |
686 | mntget(link->mnt); | |
687 | ||
574197e0 AV |
688 | if (unlikely(current->total_link_count >= 40)) { |
689 | *p = ERR_PTR(-ELOOP); /* no ->put_link(), please */ | |
574197e0 AV |
690 | path_put(&nd->path); |
691 | return -ELOOP; | |
692 | } | |
693 | cond_resched(); | |
694 | current->total_link_count++; | |
695 | ||
7b9337aa | 696 | touch_atime(link->mnt, dentry); |
1da177e4 | 697 | nd_set_link(nd, NULL); |
cd4e91d3 | 698 | |
36f3b4f6 AV |
699 | error = security_inode_follow_link(link->dentry, nd); |
700 | if (error) { | |
701 | *p = ERR_PTR(error); /* no ->put_link(), please */ | |
702 | path_put(&nd->path); | |
703 | return error; | |
704 | } | |
705 | ||
86acdca1 | 706 | nd->last_type = LAST_BIND; |
def4af30 AV |
707 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
708 | error = PTR_ERR(*p); | |
709 | if (!IS_ERR(*p)) { | |
1da177e4 | 710 | char *s = nd_get_link(nd); |
cc314eef | 711 | error = 0; |
1da177e4 LT |
712 | if (s) |
713 | error = __vfs_follow_link(nd, s); | |
bcda7652 | 714 | else if (nd->last_type == LAST_BIND) { |
16c2cd71 | 715 | nd->flags |= LOOKUP_JUMPED; |
b21041d0 AV |
716 | nd->inode = nd->path.dentry->d_inode; |
717 | if (nd->inode->i_op->follow_link) { | |
bcda7652 AV |
718 | /* stepped on a _really_ weird one */ |
719 | path_put(&nd->path); | |
720 | error = -ELOOP; | |
721 | } | |
722 | } | |
1da177e4 | 723 | } |
1da177e4 LT |
724 | return error; |
725 | } | |
726 | ||
31e6b01f NP |
727 | static int follow_up_rcu(struct path *path) |
728 | { | |
729 | struct vfsmount *parent; | |
730 | struct dentry *mountpoint; | |
731 | ||
732 | parent = path->mnt->mnt_parent; | |
733 | if (parent == path->mnt) | |
734 | return 0; | |
735 | mountpoint = path->mnt->mnt_mountpoint; | |
736 | path->dentry = mountpoint; | |
737 | path->mnt = parent; | |
738 | return 1; | |
739 | } | |
740 | ||
bab77ebf | 741 | int follow_up(struct path *path) |
1da177e4 LT |
742 | { |
743 | struct vfsmount *parent; | |
744 | struct dentry *mountpoint; | |
99b7db7b NP |
745 | |
746 | br_read_lock(vfsmount_lock); | |
bab77ebf AV |
747 | parent = path->mnt->mnt_parent; |
748 | if (parent == path->mnt) { | |
99b7db7b | 749 | br_read_unlock(vfsmount_lock); |
1da177e4 LT |
750 | return 0; |
751 | } | |
752 | mntget(parent); | |
bab77ebf | 753 | mountpoint = dget(path->mnt->mnt_mountpoint); |
99b7db7b | 754 | br_read_unlock(vfsmount_lock); |
bab77ebf AV |
755 | dput(path->dentry); |
756 | path->dentry = mountpoint; | |
757 | mntput(path->mnt); | |
758 | path->mnt = parent; | |
1da177e4 LT |
759 | return 1; |
760 | } | |
761 | ||
b5c84bf6 | 762 | /* |
9875cf80 DH |
763 | * Perform an automount |
764 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
765 | * were called with. | |
1da177e4 | 766 | */ |
9875cf80 DH |
767 | static int follow_automount(struct path *path, unsigned flags, |
768 | bool *need_mntput) | |
31e6b01f | 769 | { |
9875cf80 | 770 | struct vfsmount *mnt; |
ea5b778a | 771 | int err; |
9875cf80 DH |
772 | |
773 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
774 | return -EREMOTE; | |
775 | ||
6f45b656 DH |
776 | /* We don't want to mount if someone supplied AT_NO_AUTOMOUNT |
777 | * and this is the terminal part of the path. | |
778 | */ | |
779 | if ((flags & LOOKUP_NO_AUTOMOUNT) && !(flags & LOOKUP_CONTINUE)) | |
780 | return -EISDIR; /* we actually want to stop here */ | |
781 | ||
9875cf80 DH |
782 | /* We want to mount if someone is trying to open/create a file of any |
783 | * type under the mountpoint, wants to traverse through the mountpoint | |
784 | * or wants to open the mounted directory. | |
785 | * | |
786 | * We don't want to mount if someone's just doing a stat and they've | |
787 | * set AT_SYMLINK_NOFOLLOW - unless they're stat'ing a directory and | |
788 | * appended a '/' to the name. | |
789 | */ | |
790 | if (!(flags & LOOKUP_FOLLOW) && | |
791 | !(flags & (LOOKUP_CONTINUE | LOOKUP_DIRECTORY | | |
792 | LOOKUP_OPEN | LOOKUP_CREATE))) | |
793 | return -EISDIR; | |
794 | ||
795 | current->total_link_count++; | |
796 | if (current->total_link_count >= 40) | |
797 | return -ELOOP; | |
798 | ||
799 | mnt = path->dentry->d_op->d_automount(path); | |
800 | if (IS_ERR(mnt)) { | |
801 | /* | |
802 | * The filesystem is allowed to return -EISDIR here to indicate | |
803 | * it doesn't want to automount. For instance, autofs would do | |
804 | * this so that its userspace daemon can mount on this dentry. | |
805 | * | |
806 | * However, we can only permit this if it's a terminal point in | |
807 | * the path being looked up; if it wasn't then the remainder of | |
808 | * the path is inaccessible and we should say so. | |
809 | */ | |
810 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_CONTINUE)) | |
811 | return -EREMOTE; | |
812 | return PTR_ERR(mnt); | |
31e6b01f | 813 | } |
ea5b778a | 814 | |
9875cf80 DH |
815 | if (!mnt) /* mount collision */ |
816 | return 0; | |
31e6b01f | 817 | |
8aef1884 AV |
818 | if (!*need_mntput) { |
819 | /* lock_mount() may release path->mnt on error */ | |
820 | mntget(path->mnt); | |
821 | *need_mntput = true; | |
822 | } | |
19a167af | 823 | err = finish_automount(mnt, path); |
9875cf80 | 824 | |
ea5b778a DH |
825 | switch (err) { |
826 | case -EBUSY: | |
827 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 828 | return 0; |
ea5b778a | 829 | case 0: |
8aef1884 | 830 | path_put(path); |
ea5b778a DH |
831 | path->mnt = mnt; |
832 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 833 | return 0; |
19a167af AV |
834 | default: |
835 | return err; | |
ea5b778a | 836 | } |
19a167af | 837 | |
463ffb2e AV |
838 | } |
839 | ||
9875cf80 DH |
840 | /* |
841 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 842 | * - Flagged for transit management (autofs) |
9875cf80 DH |
843 | * - Flagged as mountpoint |
844 | * - Flagged as automount point | |
845 | * | |
846 | * This may only be called in refwalk mode. | |
847 | * | |
848 | * Serialization is taken care of in namespace.c | |
849 | */ | |
850 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 851 | { |
8aef1884 | 852 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
853 | unsigned managed; |
854 | bool need_mntput = false; | |
8aef1884 | 855 | int ret = 0; |
9875cf80 DH |
856 | |
857 | /* Given that we're not holding a lock here, we retain the value in a | |
858 | * local variable for each dentry as we look at it so that we don't see | |
859 | * the components of that value change under us */ | |
860 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
861 | managed &= DCACHE_MANAGED_DENTRY, | |
862 | unlikely(managed != 0)) { | |
cc53ce53 DH |
863 | /* Allow the filesystem to manage the transit without i_mutex |
864 | * being held. */ | |
865 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
866 | BUG_ON(!path->dentry->d_op); | |
867 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 868 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 869 | if (ret < 0) |
8aef1884 | 870 | break; |
cc53ce53 DH |
871 | } |
872 | ||
9875cf80 DH |
873 | /* Transit to a mounted filesystem. */ |
874 | if (managed & DCACHE_MOUNTED) { | |
875 | struct vfsmount *mounted = lookup_mnt(path); | |
876 | if (mounted) { | |
877 | dput(path->dentry); | |
878 | if (need_mntput) | |
879 | mntput(path->mnt); | |
880 | path->mnt = mounted; | |
881 | path->dentry = dget(mounted->mnt_root); | |
882 | need_mntput = true; | |
883 | continue; | |
884 | } | |
885 | ||
886 | /* Something is mounted on this dentry in another | |
887 | * namespace and/or whatever was mounted there in this | |
888 | * namespace got unmounted before we managed to get the | |
889 | * vfsmount_lock */ | |
890 | } | |
891 | ||
892 | /* Handle an automount point */ | |
893 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
894 | ret = follow_automount(path, flags, &need_mntput); | |
895 | if (ret < 0) | |
8aef1884 | 896 | break; |
9875cf80 DH |
897 | continue; |
898 | } | |
899 | ||
900 | /* We didn't change the current path point */ | |
901 | break; | |
1da177e4 | 902 | } |
8aef1884 AV |
903 | |
904 | if (need_mntput && path->mnt == mnt) | |
905 | mntput(path->mnt); | |
906 | if (ret == -EISDIR) | |
907 | ret = 0; | |
908 | return ret; | |
1da177e4 LT |
909 | } |
910 | ||
cc53ce53 | 911 | int follow_down_one(struct path *path) |
1da177e4 LT |
912 | { |
913 | struct vfsmount *mounted; | |
914 | ||
1c755af4 | 915 | mounted = lookup_mnt(path); |
1da177e4 | 916 | if (mounted) { |
9393bd07 AV |
917 | dput(path->dentry); |
918 | mntput(path->mnt); | |
919 | path->mnt = mounted; | |
920 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
921 | return 1; |
922 | } | |
923 | return 0; | |
924 | } | |
925 | ||
62a7375e IK |
926 | static inline bool managed_dentry_might_block(struct dentry *dentry) |
927 | { | |
928 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT && | |
929 | dentry->d_op->d_manage(dentry, true) < 0); | |
930 | } | |
931 | ||
9875cf80 | 932 | /* |
287548e4 AV |
933 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
934 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
935 | */ |
936 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 937 | struct inode **inode) |
9875cf80 | 938 | { |
62a7375e | 939 | for (;;) { |
9875cf80 | 940 | struct vfsmount *mounted; |
62a7375e IK |
941 | /* |
942 | * Don't forget we might have a non-mountpoint managed dentry | |
943 | * that wants to block transit. | |
944 | */ | |
287548e4 | 945 | if (unlikely(managed_dentry_might_block(path->dentry))) |
ab90911f | 946 | return false; |
62a7375e IK |
947 | |
948 | if (!d_mountpoint(path->dentry)) | |
949 | break; | |
950 | ||
9875cf80 DH |
951 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
952 | if (!mounted) | |
953 | break; | |
954 | path->mnt = mounted; | |
955 | path->dentry = mounted->mnt_root; | |
956 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); | |
59430262 LT |
957 | /* |
958 | * Update the inode too. We don't need to re-check the | |
959 | * dentry sequence number here after this d_inode read, | |
960 | * because a mount-point is always pinned. | |
961 | */ | |
962 | *inode = path->dentry->d_inode; | |
9875cf80 | 963 | } |
9875cf80 DH |
964 | return true; |
965 | } | |
966 | ||
dea39376 | 967 | static void follow_mount_rcu(struct nameidata *nd) |
287548e4 | 968 | { |
dea39376 | 969 | while (d_mountpoint(nd->path.dentry)) { |
287548e4 | 970 | struct vfsmount *mounted; |
dea39376 | 971 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry, 1); |
287548e4 AV |
972 | if (!mounted) |
973 | break; | |
dea39376 AV |
974 | nd->path.mnt = mounted; |
975 | nd->path.dentry = mounted->mnt_root; | |
976 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
287548e4 AV |
977 | } |
978 | } | |
979 | ||
31e6b01f NP |
980 | static int follow_dotdot_rcu(struct nameidata *nd) |
981 | { | |
31e6b01f NP |
982 | set_root_rcu(nd); |
983 | ||
9875cf80 | 984 | while (1) { |
31e6b01f NP |
985 | if (nd->path.dentry == nd->root.dentry && |
986 | nd->path.mnt == nd->root.mnt) { | |
987 | break; | |
988 | } | |
989 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
990 | struct dentry *old = nd->path.dentry; | |
991 | struct dentry *parent = old->d_parent; | |
992 | unsigned seq; | |
993 | ||
994 | seq = read_seqcount_begin(&parent->d_seq); | |
995 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 996 | goto failed; |
31e6b01f NP |
997 | nd->path.dentry = parent; |
998 | nd->seq = seq; | |
999 | break; | |
1000 | } | |
1001 | if (!follow_up_rcu(&nd->path)) | |
1002 | break; | |
1003 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
31e6b01f | 1004 | } |
dea39376 AV |
1005 | follow_mount_rcu(nd); |
1006 | nd->inode = nd->path.dentry->d_inode; | |
31e6b01f | 1007 | return 0; |
ef7562d5 AV |
1008 | |
1009 | failed: | |
1010 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1011 | if (!(nd->flags & LOOKUP_ROOT)) |
1012 | nd->root.mnt = NULL; | |
ef7562d5 AV |
1013 | rcu_read_unlock(); |
1014 | br_read_unlock(vfsmount_lock); | |
1015 | return -ECHILD; | |
31e6b01f NP |
1016 | } |
1017 | ||
cc53ce53 DH |
1018 | /* |
1019 | * Follow down to the covering mount currently visible to userspace. At each | |
1020 | * point, the filesystem owning that dentry may be queried as to whether the | |
1021 | * caller is permitted to proceed or not. | |
cc53ce53 | 1022 | */ |
7cc90cc3 | 1023 | int follow_down(struct path *path) |
cc53ce53 DH |
1024 | { |
1025 | unsigned managed; | |
1026 | int ret; | |
1027 | ||
1028 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1029 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1030 | /* Allow the filesystem to manage the transit without i_mutex | |
1031 | * being held. | |
1032 | * | |
1033 | * We indicate to the filesystem if someone is trying to mount | |
1034 | * something here. This gives autofs the chance to deny anyone | |
1035 | * other than its daemon the right to mount on its | |
1036 | * superstructure. | |
1037 | * | |
1038 | * The filesystem may sleep at this point. | |
1039 | */ | |
1040 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1041 | BUG_ON(!path->dentry->d_op); | |
1042 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 1043 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 1044 | path->dentry, false); |
cc53ce53 DH |
1045 | if (ret < 0) |
1046 | return ret == -EISDIR ? 0 : ret; | |
1047 | } | |
1048 | ||
1049 | /* Transit to a mounted filesystem. */ | |
1050 | if (managed & DCACHE_MOUNTED) { | |
1051 | struct vfsmount *mounted = lookup_mnt(path); | |
1052 | if (!mounted) | |
1053 | break; | |
1054 | dput(path->dentry); | |
1055 | mntput(path->mnt); | |
1056 | path->mnt = mounted; | |
1057 | path->dentry = dget(mounted->mnt_root); | |
1058 | continue; | |
1059 | } | |
1060 | ||
1061 | /* Don't handle automount points here */ | |
1062 | break; | |
1063 | } | |
1064 | return 0; | |
1065 | } | |
1066 | ||
9875cf80 DH |
1067 | /* |
1068 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1069 | */ | |
1070 | static void follow_mount(struct path *path) | |
1071 | { | |
1072 | while (d_mountpoint(path->dentry)) { | |
1073 | struct vfsmount *mounted = lookup_mnt(path); | |
1074 | if (!mounted) | |
1075 | break; | |
1076 | dput(path->dentry); | |
1077 | mntput(path->mnt); | |
1078 | path->mnt = mounted; | |
1079 | path->dentry = dget(mounted->mnt_root); | |
1080 | } | |
1081 | } | |
1082 | ||
31e6b01f | 1083 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1084 | { |
2a737871 | 1085 | set_root(nd); |
e518ddb7 | 1086 | |
1da177e4 | 1087 | while(1) { |
4ac91378 | 1088 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1089 | |
2a737871 AV |
1090 | if (nd->path.dentry == nd->root.dentry && |
1091 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1092 | break; |
1093 | } | |
4ac91378 | 1094 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1095 | /* rare case of legitimate dget_parent()... */ |
1096 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1097 | dput(old); |
1098 | break; | |
1099 | } | |
3088dd70 | 1100 | if (!follow_up(&nd->path)) |
1da177e4 | 1101 | break; |
1da177e4 | 1102 | } |
79ed0226 | 1103 | follow_mount(&nd->path); |
31e6b01f | 1104 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1105 | } |
1106 | ||
baa03890 NP |
1107 | /* |
1108 | * Allocate a dentry with name and parent, and perform a parent | |
1109 | * directory ->lookup on it. Returns the new dentry, or ERR_PTR | |
1110 | * on error. parent->d_inode->i_mutex must be held. d_lookup must | |
1111 | * have verified that no child exists while under i_mutex. | |
1112 | */ | |
1113 | static struct dentry *d_alloc_and_lookup(struct dentry *parent, | |
1114 | struct qstr *name, struct nameidata *nd) | |
1115 | { | |
1116 | struct inode *inode = parent->d_inode; | |
1117 | struct dentry *dentry; | |
1118 | struct dentry *old; | |
1119 | ||
1120 | /* Don't create child dentry for a dead directory. */ | |
1121 | if (unlikely(IS_DEADDIR(inode))) | |
1122 | return ERR_PTR(-ENOENT); | |
1123 | ||
1124 | dentry = d_alloc(parent, name); | |
1125 | if (unlikely(!dentry)) | |
1126 | return ERR_PTR(-ENOMEM); | |
1127 | ||
1128 | old = inode->i_op->lookup(inode, dentry, nd); | |
1129 | if (unlikely(old)) { | |
1130 | dput(dentry); | |
1131 | dentry = old; | |
1132 | } | |
1133 | return dentry; | |
1134 | } | |
1135 | ||
1da177e4 LT |
1136 | /* |
1137 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1138 | * small and for now I'd prefer to have fast path as straight as possible. | |
1139 | * It _is_ time-critical. | |
1140 | */ | |
1141 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
31e6b01f | 1142 | struct path *path, struct inode **inode) |
1da177e4 | 1143 | { |
4ac91378 | 1144 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1145 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1146 | int need_reval = 1; |
1147 | int status = 1; | |
9875cf80 DH |
1148 | int err; |
1149 | ||
b04f784e NP |
1150 | /* |
1151 | * Rename seqlock is not required here because in the off chance | |
1152 | * of a false negative due to a concurrent rename, we're going to | |
1153 | * do the non-racy lookup, below. | |
1154 | */ | |
31e6b01f NP |
1155 | if (nd->flags & LOOKUP_RCU) { |
1156 | unsigned seq; | |
31e6b01f NP |
1157 | *inode = nd->inode; |
1158 | dentry = __d_lookup_rcu(parent, name, &seq, inode); | |
5a18fff2 AV |
1159 | if (!dentry) |
1160 | goto unlazy; | |
1161 | ||
31e6b01f NP |
1162 | /* Memory barrier in read_seqcount_begin of child is enough */ |
1163 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) | |
1164 | return -ECHILD; | |
31e6b01f | 1165 | nd->seq = seq; |
5a18fff2 | 1166 | |
24643087 | 1167 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
5a18fff2 AV |
1168 | status = d_revalidate(dentry, nd); |
1169 | if (unlikely(status <= 0)) { | |
1170 | if (status != -ECHILD) | |
1171 | need_reval = 0; | |
1172 | goto unlazy; | |
1173 | } | |
24643087 | 1174 | } |
31e6b01f NP |
1175 | path->mnt = mnt; |
1176 | path->dentry = dentry; | |
d6e9bd25 AV |
1177 | if (unlikely(!__follow_mount_rcu(nd, path, inode))) |
1178 | goto unlazy; | |
1179 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1180 | goto unlazy; | |
1181 | return 0; | |
5a18fff2 | 1182 | unlazy: |
19660af7 AV |
1183 | if (unlazy_walk(nd, dentry)) |
1184 | return -ECHILD; | |
5a18fff2 AV |
1185 | } else { |
1186 | dentry = __d_lookup(parent, name); | |
9875cf80 | 1187 | } |
5a18fff2 AV |
1188 | |
1189 | retry: | |
1190 | if (unlikely(!dentry)) { | |
1191 | struct inode *dir = parent->d_inode; | |
1192 | BUG_ON(nd->inode != dir); | |
1193 | ||
1194 | mutex_lock(&dir->i_mutex); | |
1195 | dentry = d_lookup(parent, name); | |
1196 | if (likely(!dentry)) { | |
1197 | dentry = d_alloc_and_lookup(parent, name, nd); | |
1198 | if (IS_ERR(dentry)) { | |
1199 | mutex_unlock(&dir->i_mutex); | |
1200 | return PTR_ERR(dentry); | |
1201 | } | |
1202 | /* known good */ | |
1203 | need_reval = 0; | |
1204 | status = 1; | |
1205 | } | |
1206 | mutex_unlock(&dir->i_mutex); | |
1207 | } | |
1208 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) | |
1209 | status = d_revalidate(dentry, nd); | |
1210 | if (unlikely(status <= 0)) { | |
1211 | if (status < 0) { | |
1212 | dput(dentry); | |
1213 | return status; | |
1214 | } | |
1215 | if (!d_invalidate(dentry)) { | |
1216 | dput(dentry); | |
1217 | dentry = NULL; | |
1218 | need_reval = 1; | |
1219 | goto retry; | |
1220 | } | |
24643087 | 1221 | } |
5a18fff2 | 1222 | |
9875cf80 DH |
1223 | path->mnt = mnt; |
1224 | path->dentry = dentry; | |
1225 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1226 | if (unlikely(err < 0)) { |
1227 | path_put_conditional(path, nd); | |
9875cf80 | 1228 | return err; |
89312214 | 1229 | } |
9875cf80 | 1230 | *inode = path->dentry->d_inode; |
1da177e4 | 1231 | return 0; |
1da177e4 LT |
1232 | } |
1233 | ||
52094c8a AV |
1234 | static inline int may_lookup(struct nameidata *nd) |
1235 | { | |
1236 | if (nd->flags & LOOKUP_RCU) { | |
1237 | int err = exec_permission(nd->inode, IPERM_FLAG_RCU); | |
1238 | if (err != -ECHILD) | |
1239 | return err; | |
19660af7 | 1240 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1241 | return -ECHILD; |
1242 | } | |
1243 | return exec_permission(nd->inode, 0); | |
1244 | } | |
1245 | ||
9856fa1b AV |
1246 | static inline int handle_dots(struct nameidata *nd, int type) |
1247 | { | |
1248 | if (type == LAST_DOTDOT) { | |
1249 | if (nd->flags & LOOKUP_RCU) { | |
1250 | if (follow_dotdot_rcu(nd)) | |
1251 | return -ECHILD; | |
1252 | } else | |
1253 | follow_dotdot(nd); | |
1254 | } | |
1255 | return 0; | |
1256 | } | |
1257 | ||
951361f9 AV |
1258 | static void terminate_walk(struct nameidata *nd) |
1259 | { | |
1260 | if (!(nd->flags & LOOKUP_RCU)) { | |
1261 | path_put(&nd->path); | |
1262 | } else { | |
1263 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1264 | if (!(nd->flags & LOOKUP_ROOT)) |
1265 | nd->root.mnt = NULL; | |
951361f9 AV |
1266 | rcu_read_unlock(); |
1267 | br_read_unlock(vfsmount_lock); | |
1268 | } | |
1269 | } | |
1270 | ||
ce57dfc1 AV |
1271 | static inline int walk_component(struct nameidata *nd, struct path *path, |
1272 | struct qstr *name, int type, int follow) | |
1273 | { | |
1274 | struct inode *inode; | |
1275 | int err; | |
1276 | /* | |
1277 | * "." and ".." are special - ".." especially so because it has | |
1278 | * to be able to know about the current root directory and | |
1279 | * parent relationships. | |
1280 | */ | |
1281 | if (unlikely(type != LAST_NORM)) | |
1282 | return handle_dots(nd, type); | |
1283 | err = do_lookup(nd, name, path, &inode); | |
1284 | if (unlikely(err)) { | |
1285 | terminate_walk(nd); | |
1286 | return err; | |
1287 | } | |
1288 | if (!inode) { | |
1289 | path_to_nameidata(path, nd); | |
1290 | terminate_walk(nd); | |
1291 | return -ENOENT; | |
1292 | } | |
1293 | if (unlikely(inode->i_op->follow_link) && follow) { | |
19660af7 AV |
1294 | if (nd->flags & LOOKUP_RCU) { |
1295 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
1296 | terminate_walk(nd); | |
1297 | return -ECHILD; | |
1298 | } | |
1299 | } | |
ce57dfc1 AV |
1300 | BUG_ON(inode != path->dentry->d_inode); |
1301 | return 1; | |
1302 | } | |
1303 | path_to_nameidata(path, nd); | |
1304 | nd->inode = inode; | |
1305 | return 0; | |
1306 | } | |
1307 | ||
b356379a AV |
1308 | /* |
1309 | * This limits recursive symlink follows to 8, while | |
1310 | * limiting consecutive symlinks to 40. | |
1311 | * | |
1312 | * Without that kind of total limit, nasty chains of consecutive | |
1313 | * symlinks can cause almost arbitrarily long lookups. | |
1314 | */ | |
1315 | static inline int nested_symlink(struct path *path, struct nameidata *nd) | |
1316 | { | |
1317 | int res; | |
1318 | ||
b356379a AV |
1319 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1320 | path_put_conditional(path, nd); | |
1321 | path_put(&nd->path); | |
1322 | return -ELOOP; | |
1323 | } | |
1a4022f8 | 1324 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); |
b356379a AV |
1325 | |
1326 | nd->depth++; | |
1327 | current->link_count++; | |
1328 | ||
1329 | do { | |
1330 | struct path link = *path; | |
1331 | void *cookie; | |
574197e0 AV |
1332 | |
1333 | res = follow_link(&link, nd, &cookie); | |
b356379a AV |
1334 | if (!res) |
1335 | res = walk_component(nd, path, &nd->last, | |
1336 | nd->last_type, LOOKUP_FOLLOW); | |
574197e0 | 1337 | put_link(nd, &link, cookie); |
b356379a AV |
1338 | } while (res > 0); |
1339 | ||
1340 | current->link_count--; | |
1341 | nd->depth--; | |
1342 | return res; | |
1343 | } | |
1344 | ||
1da177e4 LT |
1345 | /* |
1346 | * Name resolution. | |
ea3834d9 PM |
1347 | * This is the basic name resolution function, turning a pathname into |
1348 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1349 | * |
ea3834d9 PM |
1350 | * Returns 0 and nd will have valid dentry and mnt on success. |
1351 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1352 | */ |
6de88d72 | 1353 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1354 | { |
1355 | struct path next; | |
1da177e4 LT |
1356 | int err; |
1357 | unsigned int lookup_flags = nd->flags; | |
1358 | ||
1359 | while (*name=='/') | |
1360 | name++; | |
1361 | if (!*name) | |
086e183a | 1362 | return 0; |
1da177e4 | 1363 | |
1da177e4 LT |
1364 | /* At this point we know we have a real path component. */ |
1365 | for(;;) { | |
1366 | unsigned long hash; | |
1367 | struct qstr this; | |
1368 | unsigned int c; | |
fe479a58 | 1369 | int type; |
1da177e4 | 1370 | |
cdce5d6b | 1371 | nd->flags |= LOOKUP_CONTINUE; |
52094c8a AV |
1372 | |
1373 | err = may_lookup(nd); | |
1da177e4 LT |
1374 | if (err) |
1375 | break; | |
1376 | ||
1377 | this.name = name; | |
1378 | c = *(const unsigned char *)name; | |
1379 | ||
1380 | hash = init_name_hash(); | |
1381 | do { | |
1382 | name++; | |
1383 | hash = partial_name_hash(c, hash); | |
1384 | c = *(const unsigned char *)name; | |
1385 | } while (c && (c != '/')); | |
1386 | this.len = name - (const char *) this.name; | |
1387 | this.hash = end_name_hash(hash); | |
1388 | ||
fe479a58 AV |
1389 | type = LAST_NORM; |
1390 | if (this.name[0] == '.') switch (this.len) { | |
1391 | case 2: | |
16c2cd71 | 1392 | if (this.name[1] == '.') { |
fe479a58 | 1393 | type = LAST_DOTDOT; |
16c2cd71 AV |
1394 | nd->flags |= LOOKUP_JUMPED; |
1395 | } | |
fe479a58 AV |
1396 | break; |
1397 | case 1: | |
1398 | type = LAST_DOT; | |
1399 | } | |
5a202bcd AV |
1400 | if (likely(type == LAST_NORM)) { |
1401 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1402 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd AV |
1403 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
1404 | err = parent->d_op->d_hash(parent, nd->inode, | |
1405 | &this); | |
1406 | if (err < 0) | |
1407 | break; | |
1408 | } | |
1409 | } | |
fe479a58 | 1410 | |
1da177e4 LT |
1411 | /* remove trailing slashes? */ |
1412 | if (!c) | |
1413 | goto last_component; | |
1414 | while (*++name == '/'); | |
1415 | if (!*name) | |
b356379a | 1416 | goto last_component; |
1da177e4 | 1417 | |
ce57dfc1 AV |
1418 | err = walk_component(nd, &next, &this, type, LOOKUP_FOLLOW); |
1419 | if (err < 0) | |
1420 | return err; | |
1da177e4 | 1421 | |
ce57dfc1 | 1422 | if (err) { |
b356379a | 1423 | err = nested_symlink(&next, nd); |
1da177e4 | 1424 | if (err) |
a7472bab | 1425 | return err; |
31e6b01f | 1426 | } |
1da177e4 | 1427 | err = -ENOTDIR; |
31e6b01f | 1428 | if (!nd->inode->i_op->lookup) |
1da177e4 LT |
1429 | break; |
1430 | continue; | |
1431 | /* here ends the main loop */ | |
1432 | ||
1da177e4 | 1433 | last_component: |
f55eab82 TM |
1434 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
1435 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
b356379a AV |
1436 | nd->last = this; |
1437 | nd->last_type = type; | |
086e183a | 1438 | return 0; |
1da177e4 | 1439 | } |
951361f9 | 1440 | terminate_walk(nd); |
1da177e4 LT |
1441 | return err; |
1442 | } | |
1443 | ||
70e9b357 AV |
1444 | static int path_init(int dfd, const char *name, unsigned int flags, |
1445 | struct nameidata *nd, struct file **fp) | |
31e6b01f NP |
1446 | { |
1447 | int retval = 0; | |
1448 | int fput_needed; | |
1449 | struct file *file; | |
1450 | ||
1451 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
16c2cd71 | 1452 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 1453 | nd->depth = 0; |
5b6ca027 AV |
1454 | if (flags & LOOKUP_ROOT) { |
1455 | struct inode *inode = nd->root.dentry->d_inode; | |
73d049a4 AV |
1456 | if (*name) { |
1457 | if (!inode->i_op->lookup) | |
1458 | return -ENOTDIR; | |
1459 | retval = inode_permission(inode, MAY_EXEC); | |
1460 | if (retval) | |
1461 | return retval; | |
1462 | } | |
5b6ca027 AV |
1463 | nd->path = nd->root; |
1464 | nd->inode = inode; | |
1465 | if (flags & LOOKUP_RCU) { | |
1466 | br_read_lock(vfsmount_lock); | |
1467 | rcu_read_lock(); | |
1468 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1469 | } else { | |
1470 | path_get(&nd->path); | |
1471 | } | |
1472 | return 0; | |
1473 | } | |
1474 | ||
31e6b01f | 1475 | nd->root.mnt = NULL; |
31e6b01f NP |
1476 | |
1477 | if (*name=='/') { | |
e41f7d4e AV |
1478 | if (flags & LOOKUP_RCU) { |
1479 | br_read_lock(vfsmount_lock); | |
1480 | rcu_read_lock(); | |
1481 | set_root_rcu(nd); | |
1482 | } else { | |
1483 | set_root(nd); | |
1484 | path_get(&nd->root); | |
1485 | } | |
1486 | nd->path = nd->root; | |
31e6b01f | 1487 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1488 | if (flags & LOOKUP_RCU) { |
1489 | struct fs_struct *fs = current->fs; | |
1490 | unsigned seq; | |
31e6b01f | 1491 | |
e41f7d4e AV |
1492 | br_read_lock(vfsmount_lock); |
1493 | rcu_read_lock(); | |
c28cc364 | 1494 | |
e41f7d4e AV |
1495 | do { |
1496 | seq = read_seqcount_begin(&fs->seq); | |
1497 | nd->path = fs->pwd; | |
1498 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1499 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1500 | } else { | |
1501 | get_fs_pwd(current->fs, &nd->path); | |
1502 | } | |
31e6b01f NP |
1503 | } else { |
1504 | struct dentry *dentry; | |
1505 | ||
1abf0c71 | 1506 | file = fget_raw_light(dfd, &fput_needed); |
31e6b01f NP |
1507 | retval = -EBADF; |
1508 | if (!file) | |
1509 | goto out_fail; | |
1510 | ||
1511 | dentry = file->f_path.dentry; | |
1512 | ||
f52e0c11 AV |
1513 | if (*name) { |
1514 | retval = -ENOTDIR; | |
1515 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
1516 | goto fput_fail; | |
31e6b01f | 1517 | |
f52e0c11 AV |
1518 | retval = file_permission(file, MAY_EXEC); |
1519 | if (retval) | |
1520 | goto fput_fail; | |
1521 | } | |
31e6b01f NP |
1522 | |
1523 | nd->path = file->f_path; | |
e41f7d4e AV |
1524 | if (flags & LOOKUP_RCU) { |
1525 | if (fput_needed) | |
70e9b357 | 1526 | *fp = file; |
e41f7d4e AV |
1527 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
1528 | br_read_lock(vfsmount_lock); | |
1529 | rcu_read_lock(); | |
1530 | } else { | |
1531 | path_get(&file->f_path); | |
1532 | fput_light(file, fput_needed); | |
1533 | } | |
31e6b01f | 1534 | } |
31e6b01f | 1535 | |
31e6b01f | 1536 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1537 | return 0; |
2dfdd266 | 1538 | |
9b4a9b14 AV |
1539 | fput_fail: |
1540 | fput_light(file, fput_needed); | |
1541 | out_fail: | |
1542 | return retval; | |
1543 | } | |
1544 | ||
bd92d7fe AV |
1545 | static inline int lookup_last(struct nameidata *nd, struct path *path) |
1546 | { | |
1547 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1548 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1549 | ||
1550 | nd->flags &= ~LOOKUP_PARENT; | |
1551 | return walk_component(nd, path, &nd->last, nd->last_type, | |
1552 | nd->flags & LOOKUP_FOLLOW); | |
1553 | } | |
1554 | ||
9b4a9b14 | 1555 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
ee0827cd | 1556 | static int path_lookupat(int dfd, const char *name, |
9b4a9b14 AV |
1557 | unsigned int flags, struct nameidata *nd) |
1558 | { | |
70e9b357 | 1559 | struct file *base = NULL; |
bd92d7fe AV |
1560 | struct path path; |
1561 | int err; | |
31e6b01f NP |
1562 | |
1563 | /* | |
1564 | * Path walking is largely split up into 2 different synchronisation | |
1565 | * schemes, rcu-walk and ref-walk (explained in | |
1566 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1567 | * path walk code, but some things particularly setup, cleanup, and | |
1568 | * following mounts are sufficiently divergent that functions are | |
1569 | * duplicated. Typically there is a function foo(), and its RCU | |
1570 | * analogue, foo_rcu(). | |
1571 | * | |
1572 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1573 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1574 | * be handled by restarting a traditional ref-walk (which will always | |
1575 | * be able to complete). | |
1576 | */ | |
bd92d7fe | 1577 | err = path_init(dfd, name, flags | LOOKUP_PARENT, nd, &base); |
ee0827cd | 1578 | |
bd92d7fe AV |
1579 | if (unlikely(err)) |
1580 | return err; | |
ee0827cd AV |
1581 | |
1582 | current->total_link_count = 0; | |
bd92d7fe AV |
1583 | err = link_path_walk(name, nd); |
1584 | ||
1585 | if (!err && !(flags & LOOKUP_PARENT)) { | |
bd92d7fe AV |
1586 | err = lookup_last(nd, &path); |
1587 | while (err > 0) { | |
1588 | void *cookie; | |
1589 | struct path link = path; | |
bd92d7fe | 1590 | nd->flags |= LOOKUP_PARENT; |
574197e0 | 1591 | err = follow_link(&link, nd, &cookie); |
bd92d7fe AV |
1592 | if (!err) |
1593 | err = lookup_last(nd, &path); | |
574197e0 | 1594 | put_link(nd, &link, cookie); |
bd92d7fe AV |
1595 | } |
1596 | } | |
ee0827cd | 1597 | |
9f1fafee AV |
1598 | if (!err) |
1599 | err = complete_walk(nd); | |
bd92d7fe AV |
1600 | |
1601 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
1602 | if (!nd->inode->i_op->lookup) { | |
1603 | path_put(&nd->path); | |
bd23a539 | 1604 | err = -ENOTDIR; |
bd92d7fe AV |
1605 | } |
1606 | } | |
16c2cd71 | 1607 | |
70e9b357 AV |
1608 | if (base) |
1609 | fput(base); | |
ee0827cd | 1610 | |
5b6ca027 | 1611 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
2a737871 AV |
1612 | path_put(&nd->root); |
1613 | nd->root.mnt = NULL; | |
1614 | } | |
bd92d7fe | 1615 | return err; |
ee0827cd | 1616 | } |
31e6b01f | 1617 | |
ee0827cd AV |
1618 | static int do_path_lookup(int dfd, const char *name, |
1619 | unsigned int flags, struct nameidata *nd) | |
1620 | { | |
1621 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); | |
1622 | if (unlikely(retval == -ECHILD)) | |
1623 | retval = path_lookupat(dfd, name, flags, nd); | |
1624 | if (unlikely(retval == -ESTALE)) | |
1625 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); | |
31e6b01f NP |
1626 | |
1627 | if (likely(!retval)) { | |
1628 | if (unlikely(!audit_dummy_context())) { | |
1629 | if (nd->path.dentry && nd->inode) | |
1630 | audit_inode(name, nd->path.dentry); | |
1631 | } | |
1632 | } | |
170aa3d0 | 1633 | return retval; |
1da177e4 LT |
1634 | } |
1635 | ||
c9c6cac0 | 1636 | int kern_path_parent(const char *name, struct nameidata *nd) |
5590ff0d | 1637 | { |
c9c6cac0 | 1638 | return do_path_lookup(AT_FDCWD, name, LOOKUP_PARENT, nd); |
5590ff0d UD |
1639 | } |
1640 | ||
d1811465 AV |
1641 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1642 | { | |
1643 | struct nameidata nd; | |
1644 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1645 | if (!res) | |
1646 | *path = nd.path; | |
1647 | return res; | |
1648 | } | |
1649 | ||
16f18200 JJS |
1650 | /** |
1651 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1652 | * @dentry: pointer to dentry of the base directory | |
1653 | * @mnt: pointer to vfs mount of the base directory | |
1654 | * @name: pointer to file name | |
1655 | * @flags: lookup flags | |
1656 | * @nd: pointer to nameidata | |
1657 | */ | |
1658 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1659 | const char *name, unsigned int flags, | |
1660 | struct nameidata *nd) | |
1661 | { | |
5b6ca027 AV |
1662 | nd->root.dentry = dentry; |
1663 | nd->root.mnt = mnt; | |
1664 | /* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */ | |
1665 | return do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, nd); | |
16f18200 JJS |
1666 | } |
1667 | ||
eead1911 CH |
1668 | static struct dentry *__lookup_hash(struct qstr *name, |
1669 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1670 | { |
81fca444 | 1671 | struct inode *inode = base->d_inode; |
057f6c01 | 1672 | struct dentry *dentry; |
1da177e4 LT |
1673 | int err; |
1674 | ||
b74c79e9 | 1675 | err = exec_permission(inode, 0); |
81fca444 CH |
1676 | if (err) |
1677 | return ERR_PTR(err); | |
1da177e4 | 1678 | |
b04f784e NP |
1679 | /* |
1680 | * Don't bother with __d_lookup: callers are for creat as | |
1681 | * well as unlink, so a lot of the time it would cost | |
1682 | * a double lookup. | |
6e6b1bd1 | 1683 | */ |
b04f784e | 1684 | dentry = d_lookup(base, name); |
6e6b1bd1 | 1685 | |
fb045adb | 1686 | if (dentry && (dentry->d_flags & DCACHE_OP_REVALIDATE)) |
6e6b1bd1 AV |
1687 | dentry = do_revalidate(dentry, nd); |
1688 | ||
baa03890 NP |
1689 | if (!dentry) |
1690 | dentry = d_alloc_and_lookup(base, name, nd); | |
5a202bcd | 1691 | |
1da177e4 LT |
1692 | return dentry; |
1693 | } | |
1694 | ||
057f6c01 JM |
1695 | /* |
1696 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1697 | * needs parent already locked. Doesn't follow mounts. | |
1698 | * SMP-safe. | |
1699 | */ | |
eead1911 | 1700 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1701 | { |
4ac91378 | 1702 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1703 | } |
1704 | ||
eead1911 | 1705 | /** |
a6b91919 | 1706 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1707 | * @name: pathname component to lookup |
1708 | * @base: base directory to lookup from | |
1709 | * @len: maximum length @len should be interpreted to | |
1710 | * | |
a6b91919 RD |
1711 | * Note that this routine is purely a helper for filesystem usage and should |
1712 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1713 | * nameidata argument is passed to the filesystem methods and a filesystem |
1714 | * using this helper needs to be prepared for that. | |
1715 | */ | |
057f6c01 JM |
1716 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1717 | { | |
057f6c01 | 1718 | struct qstr this; |
6a96ba54 AV |
1719 | unsigned long hash; |
1720 | unsigned int c; | |
057f6c01 | 1721 | |
2f9092e1 DW |
1722 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1723 | ||
6a96ba54 AV |
1724 | this.name = name; |
1725 | this.len = len; | |
1726 | if (!len) | |
1727 | return ERR_PTR(-EACCES); | |
1728 | ||
1729 | hash = init_name_hash(); | |
1730 | while (len--) { | |
1731 | c = *(const unsigned char *)name++; | |
1732 | if (c == '/' || c == '\0') | |
1733 | return ERR_PTR(-EACCES); | |
1734 | hash = partial_name_hash(c, hash); | |
1735 | } | |
1736 | this.hash = end_name_hash(hash); | |
5a202bcd AV |
1737 | /* |
1738 | * See if the low-level filesystem might want | |
1739 | * to use its own hash.. | |
1740 | */ | |
1741 | if (base->d_flags & DCACHE_OP_HASH) { | |
1742 | int err = base->d_op->d_hash(base, base->d_inode, &this); | |
1743 | if (err < 0) | |
1744 | return ERR_PTR(err); | |
1745 | } | |
eead1911 | 1746 | |
49705b77 | 1747 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1748 | } |
1749 | ||
2d8f3038 AV |
1750 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1751 | struct path *path) | |
1da177e4 | 1752 | { |
2d8f3038 | 1753 | struct nameidata nd; |
f52e0c11 | 1754 | char *tmp = getname_flags(name, flags); |
1da177e4 | 1755 | int err = PTR_ERR(tmp); |
1da177e4 | 1756 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1757 | |
1758 | BUG_ON(flags & LOOKUP_PARENT); | |
1759 | ||
1760 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1761 | putname(tmp); |
2d8f3038 AV |
1762 | if (!err) |
1763 | *path = nd.path; | |
1da177e4 LT |
1764 | } |
1765 | return err; | |
1766 | } | |
1767 | ||
2ad94ae6 AV |
1768 | static int user_path_parent(int dfd, const char __user *path, |
1769 | struct nameidata *nd, char **name) | |
1770 | { | |
1771 | char *s = getname(path); | |
1772 | int error; | |
1773 | ||
1774 | if (IS_ERR(s)) | |
1775 | return PTR_ERR(s); | |
1776 | ||
1777 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1778 | if (error) | |
1779 | putname(s); | |
1780 | else | |
1781 | *name = s; | |
1782 | ||
1783 | return error; | |
1784 | } | |
1785 | ||
1da177e4 LT |
1786 | /* |
1787 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1788 | * minimal. | |
1789 | */ | |
1790 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1791 | { | |
da9592ed DH |
1792 | uid_t fsuid = current_fsuid(); |
1793 | ||
1da177e4 LT |
1794 | if (!(dir->i_mode & S_ISVTX)) |
1795 | return 0; | |
e795b717 SH |
1796 | if (current_user_ns() != inode_userns(inode)) |
1797 | goto other_userns; | |
da9592ed | 1798 | if (inode->i_uid == fsuid) |
1da177e4 | 1799 | return 0; |
da9592ed | 1800 | if (dir->i_uid == fsuid) |
1da177e4 | 1801 | return 0; |
e795b717 SH |
1802 | |
1803 | other_userns: | |
1804 | return !ns_capable(inode_userns(inode), CAP_FOWNER); | |
1da177e4 LT |
1805 | } |
1806 | ||
1807 | /* | |
1808 | * Check whether we can remove a link victim from directory dir, check | |
1809 | * whether the type of victim is right. | |
1810 | * 1. We can't do it if dir is read-only (done in permission()) | |
1811 | * 2. We should have write and exec permissions on dir | |
1812 | * 3. We can't remove anything from append-only dir | |
1813 | * 4. We can't do anything with immutable dir (done in permission()) | |
1814 | * 5. If the sticky bit on dir is set we should either | |
1815 | * a. be owner of dir, or | |
1816 | * b. be owner of victim, or | |
1817 | * c. have CAP_FOWNER capability | |
1818 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1819 | * links pointing to it. | |
1820 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1821 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1822 | * 9. We can't remove a root or mountpoint. | |
1823 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1824 | * nfs_async_unlink(). | |
1825 | */ | |
858119e1 | 1826 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1827 | { |
1828 | int error; | |
1829 | ||
1830 | if (!victim->d_inode) | |
1831 | return -ENOENT; | |
1832 | ||
1833 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 1834 | audit_inode_child(victim, dir); |
1da177e4 | 1835 | |
f419a2e3 | 1836 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1837 | if (error) |
1838 | return error; | |
1839 | if (IS_APPEND(dir)) | |
1840 | return -EPERM; | |
1841 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1842 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1843 | return -EPERM; |
1844 | if (isdir) { | |
1845 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1846 | return -ENOTDIR; | |
1847 | if (IS_ROOT(victim)) | |
1848 | return -EBUSY; | |
1849 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1850 | return -EISDIR; | |
1851 | if (IS_DEADDIR(dir)) | |
1852 | return -ENOENT; | |
1853 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1854 | return -EBUSY; | |
1855 | return 0; | |
1856 | } | |
1857 | ||
1858 | /* Check whether we can create an object with dentry child in directory | |
1859 | * dir. | |
1860 | * 1. We can't do it if child already exists (open has special treatment for | |
1861 | * this case, but since we are inlined it's OK) | |
1862 | * 2. We can't do it if dir is read-only (done in permission()) | |
1863 | * 3. We should have write and exec permissions on dir | |
1864 | * 4. We can't do it if dir is immutable (done in permission()) | |
1865 | */ | |
a95164d9 | 1866 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1867 | { |
1868 | if (child->d_inode) | |
1869 | return -EEXIST; | |
1870 | if (IS_DEADDIR(dir)) | |
1871 | return -ENOENT; | |
f419a2e3 | 1872 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1873 | } |
1874 | ||
1da177e4 LT |
1875 | /* |
1876 | * p1 and p2 should be directories on the same fs. | |
1877 | */ | |
1878 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1879 | { | |
1880 | struct dentry *p; | |
1881 | ||
1882 | if (p1 == p2) { | |
f2eace23 | 1883 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1884 | return NULL; |
1885 | } | |
1886 | ||
a11f3a05 | 1887 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1888 | |
e2761a11 OH |
1889 | p = d_ancestor(p2, p1); |
1890 | if (p) { | |
1891 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1892 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1893 | return p; | |
1da177e4 LT |
1894 | } |
1895 | ||
e2761a11 OH |
1896 | p = d_ancestor(p1, p2); |
1897 | if (p) { | |
1898 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1899 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1900 | return p; | |
1da177e4 LT |
1901 | } |
1902 | ||
f2eace23 IM |
1903 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1904 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1905 | return NULL; |
1906 | } | |
1907 | ||
1908 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1909 | { | |
1b1dcc1b | 1910 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1911 | if (p1 != p2) { |
1b1dcc1b | 1912 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1913 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1914 | } |
1915 | } | |
1916 | ||
1917 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1918 | struct nameidata *nd) | |
1919 | { | |
a95164d9 | 1920 | int error = may_create(dir, dentry); |
1da177e4 LT |
1921 | |
1922 | if (error) | |
1923 | return error; | |
1924 | ||
acfa4380 | 1925 | if (!dir->i_op->create) |
1da177e4 LT |
1926 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1927 | mode &= S_IALLUGO; | |
1928 | mode |= S_IFREG; | |
1929 | error = security_inode_create(dir, dentry, mode); | |
1930 | if (error) | |
1931 | return error; | |
1da177e4 | 1932 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1933 | if (!error) |
f38aa942 | 1934 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1935 | return error; |
1936 | } | |
1937 | ||
73d049a4 | 1938 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1939 | { |
3fb64190 | 1940 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1941 | struct inode *inode = dentry->d_inode; |
1942 | int error; | |
1943 | ||
bcda7652 AV |
1944 | /* O_PATH? */ |
1945 | if (!acc_mode) | |
1946 | return 0; | |
1947 | ||
1da177e4 LT |
1948 | if (!inode) |
1949 | return -ENOENT; | |
1950 | ||
c8fe8f30 CH |
1951 | switch (inode->i_mode & S_IFMT) { |
1952 | case S_IFLNK: | |
1da177e4 | 1953 | return -ELOOP; |
c8fe8f30 CH |
1954 | case S_IFDIR: |
1955 | if (acc_mode & MAY_WRITE) | |
1956 | return -EISDIR; | |
1957 | break; | |
1958 | case S_IFBLK: | |
1959 | case S_IFCHR: | |
3fb64190 | 1960 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1961 | return -EACCES; |
c8fe8f30 CH |
1962 | /*FALLTHRU*/ |
1963 | case S_IFIFO: | |
1964 | case S_IFSOCK: | |
1da177e4 | 1965 | flag &= ~O_TRUNC; |
c8fe8f30 | 1966 | break; |
4a3fd211 | 1967 | } |
b41572e9 | 1968 | |
3fb64190 | 1969 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1970 | if (error) |
1971 | return error; | |
6146f0d5 | 1972 | |
1da177e4 LT |
1973 | /* |
1974 | * An append-only file must be opened in append mode for writing. | |
1975 | */ | |
1976 | if (IS_APPEND(inode)) { | |
8737c930 | 1977 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 1978 | return -EPERM; |
1da177e4 | 1979 | if (flag & O_TRUNC) |
7715b521 | 1980 | return -EPERM; |
1da177e4 LT |
1981 | } |
1982 | ||
1983 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 1984 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 1985 | return -EPERM; |
1da177e4 LT |
1986 | |
1987 | /* | |
1988 | * Ensure there are no outstanding leases on the file. | |
1989 | */ | |
b65a9cfc | 1990 | return break_lease(inode, flag); |
7715b521 | 1991 | } |
1da177e4 | 1992 | |
e1181ee6 | 1993 | static int handle_truncate(struct file *filp) |
7715b521 | 1994 | { |
e1181ee6 | 1995 | struct path *path = &filp->f_path; |
7715b521 AV |
1996 | struct inode *inode = path->dentry->d_inode; |
1997 | int error = get_write_access(inode); | |
1998 | if (error) | |
1999 | return error; | |
2000 | /* | |
2001 | * Refuse to truncate files with mandatory locks held on them. | |
2002 | */ | |
2003 | error = locks_verify_locked(inode); | |
2004 | if (!error) | |
ea0d3ab2 | 2005 | error = security_path_truncate(path); |
7715b521 AV |
2006 | if (!error) { |
2007 | error = do_truncate(path->dentry, 0, | |
2008 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2009 | filp); |
7715b521 AV |
2010 | } |
2011 | put_write_access(inode); | |
acd0c935 | 2012 | return error; |
1da177e4 LT |
2013 | } |
2014 | ||
d57999e1 DH |
2015 | /* |
2016 | * Note that while the flag value (low two bits) for sys_open means: | |
2017 | * 00 - read-only | |
2018 | * 01 - write-only | |
2019 | * 10 - read-write | |
2020 | * 11 - special | |
2021 | * it is changed into | |
2022 | * 00 - no permissions needed | |
2023 | * 01 - read-permission | |
2024 | * 10 - write-permission | |
2025 | * 11 - read-write | |
2026 | * for the internal routines (ie open_namei()/follow_link() etc) | |
2027 | * This is more logical, and also allows the 00 "no perm needed" | |
2028 | * to be used for symlinks (where the permissions are checked | |
2029 | * later). | |
2030 | * | |
2031 | */ | |
2032 | static inline int open_to_namei_flags(int flag) | |
2033 | { | |
2034 | if ((flag+1) & O_ACCMODE) | |
2035 | flag++; | |
2036 | return flag; | |
2037 | } | |
2038 | ||
31e6b01f | 2039 | /* |
fe2d35ff | 2040 | * Handle the last step of open() |
31e6b01f | 2041 | */ |
fb1cc555 | 2042 | static struct file *do_last(struct nameidata *nd, struct path *path, |
c3e380b0 | 2043 | const struct open_flags *op, const char *pathname) |
fb1cc555 | 2044 | { |
a1e28038 | 2045 | struct dentry *dir = nd->path.dentry; |
6c0d46c4 | 2046 | struct dentry *dentry; |
ca344a89 | 2047 | int open_flag = op->open_flag; |
6c0d46c4 | 2048 | int will_truncate = open_flag & O_TRUNC; |
ca344a89 | 2049 | int want_write = 0; |
bcda7652 | 2050 | int acc_mode = op->acc_mode; |
fb1cc555 | 2051 | struct file *filp; |
16c2cd71 | 2052 | int error; |
1f36f774 | 2053 | |
c3e380b0 AV |
2054 | nd->flags &= ~LOOKUP_PARENT; |
2055 | nd->flags |= op->intent; | |
2056 | ||
1f36f774 AV |
2057 | switch (nd->last_type) { |
2058 | case LAST_DOTDOT: | |
176306f5 | 2059 | case LAST_DOT: |
fe2d35ff AV |
2060 | error = handle_dots(nd, nd->last_type); |
2061 | if (error) | |
2062 | return ERR_PTR(error); | |
1f36f774 | 2063 | /* fallthrough */ |
1f36f774 | 2064 | case LAST_ROOT: |
9f1fafee | 2065 | error = complete_walk(nd); |
16c2cd71 | 2066 | if (error) |
9f1fafee | 2067 | return ERR_PTR(error); |
fe2d35ff | 2068 | audit_inode(pathname, nd->path.dentry); |
ca344a89 | 2069 | if (open_flag & O_CREAT) { |
fe2d35ff AV |
2070 | error = -EISDIR; |
2071 | goto exit; | |
2072 | } | |
2073 | goto ok; | |
1f36f774 | 2074 | case LAST_BIND: |
9f1fafee | 2075 | error = complete_walk(nd); |
16c2cd71 | 2076 | if (error) |
9f1fafee | 2077 | return ERR_PTR(error); |
1f36f774 | 2078 | audit_inode(pathname, dir); |
67ee3ad2 | 2079 | goto ok; |
1f36f774 | 2080 | } |
67ee3ad2 | 2081 | |
ca344a89 | 2082 | if (!(open_flag & O_CREAT)) { |
bcda7652 | 2083 | int symlink_ok = 0; |
fe2d35ff AV |
2084 | if (nd->last.name[nd->last.len]) |
2085 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
bcda7652 AV |
2086 | if (open_flag & O_PATH && !(nd->flags & LOOKUP_FOLLOW)) |
2087 | symlink_ok = 1; | |
fe2d35ff | 2088 | /* we _can_ be in RCU mode here */ |
ce57dfc1 AV |
2089 | error = walk_component(nd, path, &nd->last, LAST_NORM, |
2090 | !symlink_ok); | |
2091 | if (error < 0) | |
fe2d35ff | 2092 | return ERR_PTR(error); |
ce57dfc1 | 2093 | if (error) /* symlink */ |
fe2d35ff | 2094 | return NULL; |
fe2d35ff | 2095 | /* sayonara */ |
9f1fafee AV |
2096 | error = complete_walk(nd); |
2097 | if (error) | |
2098 | return ERR_PTR(-ECHILD); | |
fe2d35ff AV |
2099 | |
2100 | error = -ENOTDIR; | |
2101 | if (nd->flags & LOOKUP_DIRECTORY) { | |
ce57dfc1 | 2102 | if (!nd->inode->i_op->lookup) |
fe2d35ff AV |
2103 | goto exit; |
2104 | } | |
2105 | audit_inode(pathname, nd->path.dentry); | |
2106 | goto ok; | |
2107 | } | |
2108 | ||
2109 | /* create side of things */ | |
9f1fafee AV |
2110 | error = complete_walk(nd); |
2111 | if (error) | |
2112 | return ERR_PTR(error); | |
fe2d35ff AV |
2113 | |
2114 | audit_inode(pathname, dir); | |
16c2cd71 | 2115 | error = -EISDIR; |
1f36f774 | 2116 | /* trailing slashes? */ |
31e6b01f NP |
2117 | if (nd->last.name[nd->last.len]) |
2118 | goto exit; | |
a2c36b45 | 2119 | |
a1e28038 AV |
2120 | mutex_lock(&dir->d_inode->i_mutex); |
2121 | ||
6c0d46c4 AV |
2122 | dentry = lookup_hash(nd); |
2123 | error = PTR_ERR(dentry); | |
2124 | if (IS_ERR(dentry)) { | |
fb1cc555 AV |
2125 | mutex_unlock(&dir->d_inode->i_mutex); |
2126 | goto exit; | |
2127 | } | |
2128 | ||
6c0d46c4 AV |
2129 | path->dentry = dentry; |
2130 | path->mnt = nd->path.mnt; | |
2131 | ||
fb1cc555 | 2132 | /* Negative dentry, just create the file */ |
6c0d46c4 AV |
2133 | if (!dentry->d_inode) { |
2134 | int mode = op->mode; | |
2135 | if (!IS_POSIXACL(dir->d_inode)) | |
2136 | mode &= ~current_umask(); | |
fb1cc555 AV |
2137 | /* |
2138 | * This write is needed to ensure that a | |
6c0d46c4 | 2139 | * rw->ro transition does not occur between |
fb1cc555 AV |
2140 | * the time when the file is created and when |
2141 | * a permanent write count is taken through | |
2142 | * the 'struct file' in nameidata_to_filp(). | |
2143 | */ | |
2144 | error = mnt_want_write(nd->path.mnt); | |
2145 | if (error) | |
2146 | goto exit_mutex_unlock; | |
ca344a89 | 2147 | want_write = 1; |
9b44f1b3 | 2148 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 2149 | open_flag &= ~O_TRUNC; |
6c0d46c4 | 2150 | will_truncate = 0; |
bcda7652 | 2151 | acc_mode = MAY_OPEN; |
6c0d46c4 AV |
2152 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2153 | if (error) | |
2154 | goto exit_mutex_unlock; | |
2155 | error = vfs_create(dir->d_inode, dentry, mode, nd); | |
2156 | if (error) | |
2157 | goto exit_mutex_unlock; | |
2158 | mutex_unlock(&dir->d_inode->i_mutex); | |
2159 | dput(nd->path.dentry); | |
2160 | nd->path.dentry = dentry; | |
ca344a89 | 2161 | goto common; |
fb1cc555 AV |
2162 | } |
2163 | ||
2164 | /* | |
2165 | * It already exists. | |
2166 | */ | |
2167 | mutex_unlock(&dir->d_inode->i_mutex); | |
2168 | audit_inode(pathname, path->dentry); | |
2169 | ||
2170 | error = -EEXIST; | |
ca344a89 | 2171 | if (open_flag & O_EXCL) |
fb1cc555 AV |
2172 | goto exit_dput; |
2173 | ||
9875cf80 DH |
2174 | error = follow_managed(path, nd->flags); |
2175 | if (error < 0) | |
2176 | goto exit_dput; | |
fb1cc555 AV |
2177 | |
2178 | error = -ENOENT; | |
2179 | if (!path->dentry->d_inode) | |
2180 | goto exit_dput; | |
9e67f361 AV |
2181 | |
2182 | if (path->dentry->d_inode->i_op->follow_link) | |
fb1cc555 | 2183 | return NULL; |
fb1cc555 AV |
2184 | |
2185 | path_to_nameidata(path, nd); | |
31e6b01f | 2186 | nd->inode = path->dentry->d_inode; |
fb1cc555 | 2187 | error = -EISDIR; |
31e6b01f | 2188 | if (S_ISDIR(nd->inode->i_mode)) |
fb1cc555 | 2189 | goto exit; |
67ee3ad2 | 2190 | ok: |
6c0d46c4 AV |
2191 | if (!S_ISREG(nd->inode->i_mode)) |
2192 | will_truncate = 0; | |
2193 | ||
0f9d1a10 AV |
2194 | if (will_truncate) { |
2195 | error = mnt_want_write(nd->path.mnt); | |
2196 | if (error) | |
2197 | goto exit; | |
ca344a89 | 2198 | want_write = 1; |
0f9d1a10 | 2199 | } |
ca344a89 | 2200 | common: |
bcda7652 | 2201 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 2202 | if (error) |
0f9d1a10 | 2203 | goto exit; |
0f9d1a10 AV |
2204 | filp = nameidata_to_filp(nd); |
2205 | if (!IS_ERR(filp)) { | |
2206 | error = ima_file_check(filp, op->acc_mode); | |
2207 | if (error) { | |
2208 | fput(filp); | |
2209 | filp = ERR_PTR(error); | |
2210 | } | |
2211 | } | |
2212 | if (!IS_ERR(filp)) { | |
2213 | if (will_truncate) { | |
2214 | error = handle_truncate(filp); | |
2215 | if (error) { | |
2216 | fput(filp); | |
2217 | filp = ERR_PTR(error); | |
2218 | } | |
2219 | } | |
2220 | } | |
ca344a89 AV |
2221 | out: |
2222 | if (want_write) | |
0f9d1a10 AV |
2223 | mnt_drop_write(nd->path.mnt); |
2224 | path_put(&nd->path); | |
fb1cc555 AV |
2225 | return filp; |
2226 | ||
2227 | exit_mutex_unlock: | |
2228 | mutex_unlock(&dir->d_inode->i_mutex); | |
2229 | exit_dput: | |
2230 | path_put_conditional(path, nd); | |
2231 | exit: | |
ca344a89 AV |
2232 | filp = ERR_PTR(error); |
2233 | goto out; | |
fb1cc555 AV |
2234 | } |
2235 | ||
13aab428 | 2236 | static struct file *path_openat(int dfd, const char *pathname, |
73d049a4 | 2237 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 2238 | { |
fe2d35ff | 2239 | struct file *base = NULL; |
4a3fd211 | 2240 | struct file *filp; |
9850c056 | 2241 | struct path path; |
13aab428 | 2242 | int error; |
31e6b01f NP |
2243 | |
2244 | filp = get_empty_filp(); | |
2245 | if (!filp) | |
2246 | return ERR_PTR(-ENFILE); | |
2247 | ||
47c805dc | 2248 | filp->f_flags = op->open_flag; |
73d049a4 AV |
2249 | nd->intent.open.file = filp; |
2250 | nd->intent.open.flags = open_to_namei_flags(op->open_flag); | |
2251 | nd->intent.open.create_mode = op->mode; | |
31e6b01f | 2252 | |
73d049a4 | 2253 | error = path_init(dfd, pathname, flags | LOOKUP_PARENT, nd, &base); |
31e6b01f | 2254 | if (unlikely(error)) |
13aab428 | 2255 | goto out_filp; |
31e6b01f | 2256 | |
fe2d35ff | 2257 | current->total_link_count = 0; |
73d049a4 | 2258 | error = link_path_walk(pathname, nd); |
31e6b01f NP |
2259 | if (unlikely(error)) |
2260 | goto out_filp; | |
1da177e4 | 2261 | |
73d049a4 | 2262 | filp = do_last(nd, &path, op, pathname); |
806b681c | 2263 | while (unlikely(!filp)) { /* trailing symlink */ |
7b9337aa | 2264 | struct path link = path; |
def4af30 | 2265 | void *cookie; |
574197e0 | 2266 | if (!(nd->flags & LOOKUP_FOLLOW)) { |
73d049a4 AV |
2267 | path_put_conditional(&path, nd); |
2268 | path_put(&nd->path); | |
40b39136 AV |
2269 | filp = ERR_PTR(-ELOOP); |
2270 | break; | |
2271 | } | |
73d049a4 AV |
2272 | nd->flags |= LOOKUP_PARENT; |
2273 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
574197e0 | 2274 | error = follow_link(&link, nd, &cookie); |
c3e380b0 | 2275 | if (unlikely(error)) |
f1afe9ef | 2276 | filp = ERR_PTR(error); |
c3e380b0 | 2277 | else |
73d049a4 | 2278 | filp = do_last(nd, &path, op, pathname); |
574197e0 | 2279 | put_link(nd, &link, cookie); |
806b681c | 2280 | } |
10fa8e62 | 2281 | out: |
73d049a4 AV |
2282 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) |
2283 | path_put(&nd->root); | |
fe2d35ff AV |
2284 | if (base) |
2285 | fput(base); | |
73d049a4 | 2286 | release_open_intent(nd); |
10fa8e62 | 2287 | return filp; |
1da177e4 | 2288 | |
31e6b01f | 2289 | out_filp: |
806b681c | 2290 | filp = ERR_PTR(error); |
10fa8e62 | 2291 | goto out; |
1da177e4 LT |
2292 | } |
2293 | ||
13aab428 AV |
2294 | struct file *do_filp_open(int dfd, const char *pathname, |
2295 | const struct open_flags *op, int flags) | |
2296 | { | |
73d049a4 | 2297 | struct nameidata nd; |
13aab428 AV |
2298 | struct file *filp; |
2299 | ||
73d049a4 | 2300 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 2301 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 2302 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 2303 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 2304 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
2305 | return filp; |
2306 | } | |
2307 | ||
73d049a4 AV |
2308 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
2309 | const char *name, const struct open_flags *op, int flags) | |
2310 | { | |
2311 | struct nameidata nd; | |
2312 | struct file *file; | |
2313 | ||
2314 | nd.root.mnt = mnt; | |
2315 | nd.root.dentry = dentry; | |
2316 | ||
2317 | flags |= LOOKUP_ROOT; | |
2318 | ||
bcda7652 | 2319 | if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
2320 | return ERR_PTR(-ELOOP); |
2321 | ||
2322 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU); | |
2323 | if (unlikely(file == ERR_PTR(-ECHILD))) | |
2324 | file = path_openat(-1, name, &nd, op, flags); | |
2325 | if (unlikely(file == ERR_PTR(-ESTALE))) | |
2326 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL); | |
2327 | return file; | |
2328 | } | |
2329 | ||
1da177e4 LT |
2330 | /** |
2331 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
2332 | * @nd: nameidata info | |
2333 | * @is_dir: directory flag | |
2334 | * | |
2335 | * Simple function to lookup and return a dentry and create it | |
2336 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 2337 | * |
4ac91378 | 2338 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
2339 | */ |
2340 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
2341 | { | |
c663e5d8 | 2342 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 2343 | |
4ac91378 | 2344 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
2345 | /* |
2346 | * Yucky last component or no last component at all? | |
2347 | * (foo/., foo/.., /////) | |
2348 | */ | |
1da177e4 LT |
2349 | if (nd->last_type != LAST_NORM) |
2350 | goto fail; | |
2351 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 2352 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 2353 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
2354 | |
2355 | /* | |
2356 | * Do the final lookup. | |
2357 | */ | |
49705b77 | 2358 | dentry = lookup_hash(nd); |
1da177e4 LT |
2359 | if (IS_ERR(dentry)) |
2360 | goto fail; | |
c663e5d8 | 2361 | |
e9baf6e5 AV |
2362 | if (dentry->d_inode) |
2363 | goto eexist; | |
c663e5d8 CH |
2364 | /* |
2365 | * Special case - lookup gave negative, but... we had foo/bar/ | |
2366 | * From the vfs_mknod() POV we just have a negative dentry - | |
2367 | * all is fine. Let's be bastards - you had / on the end, you've | |
2368 | * been asking for (non-existent) directory. -ENOENT for you. | |
2369 | */ | |
e9baf6e5 AV |
2370 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
2371 | dput(dentry); | |
2372 | dentry = ERR_PTR(-ENOENT); | |
2373 | } | |
1da177e4 | 2374 | return dentry; |
e9baf6e5 | 2375 | eexist: |
1da177e4 | 2376 | dput(dentry); |
e9baf6e5 | 2377 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
2378 | fail: |
2379 | return dentry; | |
2380 | } | |
f81a0bff | 2381 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
2382 | |
2383 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
2384 | { | |
a95164d9 | 2385 | int error = may_create(dir, dentry); |
1da177e4 LT |
2386 | |
2387 | if (error) | |
2388 | return error; | |
2389 | ||
e795b717 SH |
2390 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && |
2391 | !ns_capable(inode_userns(dir), CAP_MKNOD)) | |
1da177e4 LT |
2392 | return -EPERM; |
2393 | ||
acfa4380 | 2394 | if (!dir->i_op->mknod) |
1da177e4 LT |
2395 | return -EPERM; |
2396 | ||
08ce5f16 SH |
2397 | error = devcgroup_inode_mknod(mode, dev); |
2398 | if (error) | |
2399 | return error; | |
2400 | ||
1da177e4 LT |
2401 | error = security_inode_mknod(dir, dentry, mode, dev); |
2402 | if (error) | |
2403 | return error; | |
2404 | ||
1da177e4 | 2405 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 2406 | if (!error) |
f38aa942 | 2407 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2408 | return error; |
2409 | } | |
2410 | ||
463c3197 DH |
2411 | static int may_mknod(mode_t mode) |
2412 | { | |
2413 | switch (mode & S_IFMT) { | |
2414 | case S_IFREG: | |
2415 | case S_IFCHR: | |
2416 | case S_IFBLK: | |
2417 | case S_IFIFO: | |
2418 | case S_IFSOCK: | |
2419 | case 0: /* zero mode translates to S_IFREG */ | |
2420 | return 0; | |
2421 | case S_IFDIR: | |
2422 | return -EPERM; | |
2423 | default: | |
2424 | return -EINVAL; | |
2425 | } | |
2426 | } | |
2427 | ||
2e4d0924 HC |
2428 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
2429 | unsigned, dev) | |
1da177e4 | 2430 | { |
2ad94ae6 AV |
2431 | int error; |
2432 | char *tmp; | |
2433 | struct dentry *dentry; | |
1da177e4 LT |
2434 | struct nameidata nd; |
2435 | ||
2436 | if (S_ISDIR(mode)) | |
2437 | return -EPERM; | |
1da177e4 | 2438 | |
2ad94ae6 | 2439 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 2440 | if (error) |
2ad94ae6 AV |
2441 | return error; |
2442 | ||
1da177e4 | 2443 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2444 | if (IS_ERR(dentry)) { |
2445 | error = PTR_ERR(dentry); | |
2446 | goto out_unlock; | |
2447 | } | |
4ac91378 | 2448 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2449 | mode &= ~current_umask(); |
463c3197 DH |
2450 | error = may_mknod(mode); |
2451 | if (error) | |
2452 | goto out_dput; | |
2453 | error = mnt_want_write(nd.path.mnt); | |
2454 | if (error) | |
2455 | goto out_dput; | |
be6d3e56 KT |
2456 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2457 | if (error) | |
2458 | goto out_drop_write; | |
463c3197 | 2459 | switch (mode & S_IFMT) { |
1da177e4 | 2460 | case 0: case S_IFREG: |
4ac91378 | 2461 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2462 | break; |
2463 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2464 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2465 | new_decode_dev(dev)); |
2466 | break; | |
2467 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2468 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2469 | break; |
1da177e4 | 2470 | } |
be6d3e56 | 2471 | out_drop_write: |
463c3197 DH |
2472 | mnt_drop_write(nd.path.mnt); |
2473 | out_dput: | |
2474 | dput(dentry); | |
2475 | out_unlock: | |
4ac91378 | 2476 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2477 | path_put(&nd.path); |
1da177e4 LT |
2478 | putname(tmp); |
2479 | ||
2480 | return error; | |
2481 | } | |
2482 | ||
3480b257 | 2483 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2484 | { |
2485 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2486 | } | |
2487 | ||
1da177e4 LT |
2488 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2489 | { | |
a95164d9 | 2490 | int error = may_create(dir, dentry); |
1da177e4 LT |
2491 | |
2492 | if (error) | |
2493 | return error; | |
2494 | ||
acfa4380 | 2495 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2496 | return -EPERM; |
2497 | ||
2498 | mode &= (S_IRWXUGO|S_ISVTX); | |
2499 | error = security_inode_mkdir(dir, dentry, mode); | |
2500 | if (error) | |
2501 | return error; | |
2502 | ||
1da177e4 | 2503 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2504 | if (!error) |
f38aa942 | 2505 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2506 | return error; |
2507 | } | |
2508 | ||
2e4d0924 | 2509 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2510 | { |
2511 | int error = 0; | |
2512 | char * tmp; | |
6902d925 DH |
2513 | struct dentry *dentry; |
2514 | struct nameidata nd; | |
1da177e4 | 2515 | |
2ad94ae6 AV |
2516 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2517 | if (error) | |
6902d925 | 2518 | goto out_err; |
1da177e4 | 2519 | |
6902d925 DH |
2520 | dentry = lookup_create(&nd, 1); |
2521 | error = PTR_ERR(dentry); | |
2522 | if (IS_ERR(dentry)) | |
2523 | goto out_unlock; | |
1da177e4 | 2524 | |
4ac91378 | 2525 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2526 | mode &= ~current_umask(); |
463c3197 DH |
2527 | error = mnt_want_write(nd.path.mnt); |
2528 | if (error) | |
2529 | goto out_dput; | |
be6d3e56 KT |
2530 | error = security_path_mkdir(&nd.path, dentry, mode); |
2531 | if (error) | |
2532 | goto out_drop_write; | |
4ac91378 | 2533 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2534 | out_drop_write: |
463c3197 DH |
2535 | mnt_drop_write(nd.path.mnt); |
2536 | out_dput: | |
6902d925 DH |
2537 | dput(dentry); |
2538 | out_unlock: | |
4ac91378 | 2539 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2540 | path_put(&nd.path); |
6902d925 DH |
2541 | putname(tmp); |
2542 | out_err: | |
1da177e4 LT |
2543 | return error; |
2544 | } | |
2545 | ||
3cdad428 | 2546 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2547 | { |
2548 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2549 | } | |
2550 | ||
1da177e4 | 2551 | /* |
a71905f0 SW |
2552 | * The dentry_unhash() helper will try to drop the dentry early: we |
2553 | * should have a usage count of 2 if we're the only user of this | |
2554 | * dentry, and if that is true (possibly after pruning the dcache), | |
2555 | * then we drop the dentry now. | |
1da177e4 LT |
2556 | * |
2557 | * A low-level filesystem can, if it choses, legally | |
2558 | * do a | |
2559 | * | |
2560 | * if (!d_unhashed(dentry)) | |
2561 | * return -EBUSY; | |
2562 | * | |
2563 | * if it cannot handle the case of removing a directory | |
2564 | * that is still in use by something else.. | |
2565 | */ | |
2566 | void dentry_unhash(struct dentry *dentry) | |
2567 | { | |
dc168427 | 2568 | shrink_dcache_parent(dentry); |
1da177e4 | 2569 | spin_lock(&dentry->d_lock); |
64252c75 | 2570 | if (dentry->d_count == 1) |
1da177e4 LT |
2571 | __d_drop(dentry); |
2572 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
2573 | } |
2574 | ||
2575 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2576 | { | |
2577 | int error = may_delete(dir, dentry, 1); | |
2578 | ||
2579 | if (error) | |
2580 | return error; | |
2581 | ||
acfa4380 | 2582 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2583 | return -EPERM; |
2584 | ||
1b1dcc1b | 2585 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
2586 | |
2587 | error = -EBUSY; | |
1da177e4 | 2588 | if (d_mountpoint(dentry)) |
912dbc15 SW |
2589 | goto out; |
2590 | ||
2591 | error = security_inode_rmdir(dir, dentry); | |
2592 | if (error) | |
2593 | goto out; | |
2594 | ||
3cebde24 | 2595 | shrink_dcache_parent(dentry); |
912dbc15 SW |
2596 | error = dir->i_op->rmdir(dir, dentry); |
2597 | if (error) | |
2598 | goto out; | |
2599 | ||
2600 | dentry->d_inode->i_flags |= S_DEAD; | |
2601 | dont_mount(dentry); | |
2602 | ||
2603 | out: | |
1b1dcc1b | 2604 | mutex_unlock(&dentry->d_inode->i_mutex); |
912dbc15 | 2605 | if (!error) |
1da177e4 | 2606 | d_delete(dentry); |
1da177e4 LT |
2607 | return error; |
2608 | } | |
2609 | ||
5590ff0d | 2610 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2611 | { |
2612 | int error = 0; | |
2613 | char * name; | |
2614 | struct dentry *dentry; | |
2615 | struct nameidata nd; | |
2616 | ||
2ad94ae6 | 2617 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2618 | if (error) |
2ad94ae6 | 2619 | return error; |
1da177e4 LT |
2620 | |
2621 | switch(nd.last_type) { | |
0612d9fb OH |
2622 | case LAST_DOTDOT: |
2623 | error = -ENOTEMPTY; | |
2624 | goto exit1; | |
2625 | case LAST_DOT: | |
2626 | error = -EINVAL; | |
2627 | goto exit1; | |
2628 | case LAST_ROOT: | |
2629 | error = -EBUSY; | |
2630 | goto exit1; | |
1da177e4 | 2631 | } |
0612d9fb OH |
2632 | |
2633 | nd.flags &= ~LOOKUP_PARENT; | |
2634 | ||
4ac91378 | 2635 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2636 | dentry = lookup_hash(&nd); |
1da177e4 | 2637 | error = PTR_ERR(dentry); |
6902d925 DH |
2638 | if (IS_ERR(dentry)) |
2639 | goto exit2; | |
e6bc45d6 TT |
2640 | if (!dentry->d_inode) { |
2641 | error = -ENOENT; | |
2642 | goto exit3; | |
2643 | } | |
0622753b DH |
2644 | error = mnt_want_write(nd.path.mnt); |
2645 | if (error) | |
2646 | goto exit3; | |
be6d3e56 KT |
2647 | error = security_path_rmdir(&nd.path, dentry); |
2648 | if (error) | |
2649 | goto exit4; | |
4ac91378 | 2650 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2651 | exit4: |
0622753b DH |
2652 | mnt_drop_write(nd.path.mnt); |
2653 | exit3: | |
6902d925 DH |
2654 | dput(dentry); |
2655 | exit2: | |
4ac91378 | 2656 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2657 | exit1: |
1d957f9b | 2658 | path_put(&nd.path); |
1da177e4 LT |
2659 | putname(name); |
2660 | return error; | |
2661 | } | |
2662 | ||
3cdad428 | 2663 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2664 | { |
2665 | return do_rmdir(AT_FDCWD, pathname); | |
2666 | } | |
2667 | ||
1da177e4 LT |
2668 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2669 | { | |
2670 | int error = may_delete(dir, dentry, 0); | |
2671 | ||
2672 | if (error) | |
2673 | return error; | |
2674 | ||
acfa4380 | 2675 | if (!dir->i_op->unlink) |
1da177e4 LT |
2676 | return -EPERM; |
2677 | ||
1b1dcc1b | 2678 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2679 | if (d_mountpoint(dentry)) |
2680 | error = -EBUSY; | |
2681 | else { | |
2682 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 2683 | if (!error) { |
1da177e4 | 2684 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 2685 | if (!error) |
d83c49f3 | 2686 | dont_mount(dentry); |
bec1052e | 2687 | } |
1da177e4 | 2688 | } |
1b1dcc1b | 2689 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2690 | |
2691 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2692 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2693 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2694 | d_delete(dentry); |
1da177e4 | 2695 | } |
0eeca283 | 2696 | |
1da177e4 LT |
2697 | return error; |
2698 | } | |
2699 | ||
2700 | /* | |
2701 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2702 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2703 | * writeout happening, and we don't want to prevent access to the directory |
2704 | * while waiting on the I/O. | |
2705 | */ | |
5590ff0d | 2706 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2707 | { |
2ad94ae6 AV |
2708 | int error; |
2709 | char *name; | |
1da177e4 LT |
2710 | struct dentry *dentry; |
2711 | struct nameidata nd; | |
2712 | struct inode *inode = NULL; | |
2713 | ||
2ad94ae6 | 2714 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2715 | if (error) |
2ad94ae6 AV |
2716 | return error; |
2717 | ||
1da177e4 LT |
2718 | error = -EISDIR; |
2719 | if (nd.last_type != LAST_NORM) | |
2720 | goto exit1; | |
0612d9fb OH |
2721 | |
2722 | nd.flags &= ~LOOKUP_PARENT; | |
2723 | ||
4ac91378 | 2724 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2725 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2726 | error = PTR_ERR(dentry); |
2727 | if (!IS_ERR(dentry)) { | |
2728 | /* Why not before? Because we want correct error value */ | |
50338b88 TE |
2729 | if (nd.last.name[nd.last.len]) |
2730 | goto slashes; | |
1da177e4 | 2731 | inode = dentry->d_inode; |
50338b88 | 2732 | if (!inode) |
e6bc45d6 TT |
2733 | goto slashes; |
2734 | ihold(inode); | |
0622753b DH |
2735 | error = mnt_want_write(nd.path.mnt); |
2736 | if (error) | |
2737 | goto exit2; | |
be6d3e56 KT |
2738 | error = security_path_unlink(&nd.path, dentry); |
2739 | if (error) | |
2740 | goto exit3; | |
4ac91378 | 2741 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2742 | exit3: |
0622753b | 2743 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2744 | exit2: |
2745 | dput(dentry); | |
2746 | } | |
4ac91378 | 2747 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2748 | if (inode) |
2749 | iput(inode); /* truncate the inode here */ | |
2750 | exit1: | |
1d957f9b | 2751 | path_put(&nd.path); |
1da177e4 LT |
2752 | putname(name); |
2753 | return error; | |
2754 | ||
2755 | slashes: | |
2756 | error = !dentry->d_inode ? -ENOENT : | |
2757 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2758 | goto exit2; | |
2759 | } | |
2760 | ||
2e4d0924 | 2761 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2762 | { |
2763 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2764 | return -EINVAL; | |
2765 | ||
2766 | if (flag & AT_REMOVEDIR) | |
2767 | return do_rmdir(dfd, pathname); | |
2768 | ||
2769 | return do_unlinkat(dfd, pathname); | |
2770 | } | |
2771 | ||
3480b257 | 2772 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2773 | { |
2774 | return do_unlinkat(AT_FDCWD, pathname); | |
2775 | } | |
2776 | ||
db2e747b | 2777 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2778 | { |
a95164d9 | 2779 | int error = may_create(dir, dentry); |
1da177e4 LT |
2780 | |
2781 | if (error) | |
2782 | return error; | |
2783 | ||
acfa4380 | 2784 | if (!dir->i_op->symlink) |
1da177e4 LT |
2785 | return -EPERM; |
2786 | ||
2787 | error = security_inode_symlink(dir, dentry, oldname); | |
2788 | if (error) | |
2789 | return error; | |
2790 | ||
1da177e4 | 2791 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2792 | if (!error) |
f38aa942 | 2793 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2794 | return error; |
2795 | } | |
2796 | ||
2e4d0924 HC |
2797 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2798 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2799 | { |
2ad94ae6 AV |
2800 | int error; |
2801 | char *from; | |
2802 | char *to; | |
6902d925 DH |
2803 | struct dentry *dentry; |
2804 | struct nameidata nd; | |
1da177e4 LT |
2805 | |
2806 | from = getname(oldname); | |
2ad94ae6 | 2807 | if (IS_ERR(from)) |
1da177e4 | 2808 | return PTR_ERR(from); |
1da177e4 | 2809 | |
2ad94ae6 | 2810 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2811 | if (error) |
2ad94ae6 AV |
2812 | goto out_putname; |
2813 | ||
6902d925 DH |
2814 | dentry = lookup_create(&nd, 0); |
2815 | error = PTR_ERR(dentry); | |
2816 | if (IS_ERR(dentry)) | |
2817 | goto out_unlock; | |
2818 | ||
75c3f29d DH |
2819 | error = mnt_want_write(nd.path.mnt); |
2820 | if (error) | |
2821 | goto out_dput; | |
be6d3e56 KT |
2822 | error = security_path_symlink(&nd.path, dentry, from); |
2823 | if (error) | |
2824 | goto out_drop_write; | |
db2e747b | 2825 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2826 | out_drop_write: |
75c3f29d DH |
2827 | mnt_drop_write(nd.path.mnt); |
2828 | out_dput: | |
6902d925 DH |
2829 | dput(dentry); |
2830 | out_unlock: | |
4ac91378 | 2831 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2832 | path_put(&nd.path); |
6902d925 DH |
2833 | putname(to); |
2834 | out_putname: | |
1da177e4 LT |
2835 | putname(from); |
2836 | return error; | |
2837 | } | |
2838 | ||
3480b257 | 2839 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2840 | { |
2841 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2842 | } | |
2843 | ||
1da177e4 LT |
2844 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2845 | { | |
2846 | struct inode *inode = old_dentry->d_inode; | |
2847 | int error; | |
2848 | ||
2849 | if (!inode) | |
2850 | return -ENOENT; | |
2851 | ||
a95164d9 | 2852 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2853 | if (error) |
2854 | return error; | |
2855 | ||
2856 | if (dir->i_sb != inode->i_sb) | |
2857 | return -EXDEV; | |
2858 | ||
2859 | /* | |
2860 | * A link to an append-only or immutable file cannot be created. | |
2861 | */ | |
2862 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2863 | return -EPERM; | |
acfa4380 | 2864 | if (!dir->i_op->link) |
1da177e4 | 2865 | return -EPERM; |
7e79eedb | 2866 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2867 | return -EPERM; |
2868 | ||
2869 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2870 | if (error) | |
2871 | return error; | |
2872 | ||
7e79eedb | 2873 | mutex_lock(&inode->i_mutex); |
aae8a97d AK |
2874 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
2875 | if (inode->i_nlink == 0) | |
2876 | error = -ENOENT; | |
2877 | else | |
2878 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 2879 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2880 | if (!error) |
7e79eedb | 2881 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2882 | return error; |
2883 | } | |
2884 | ||
2885 | /* | |
2886 | * Hardlinks are often used in delicate situations. We avoid | |
2887 | * security-related surprises by not following symlinks on the | |
2888 | * newname. --KAB | |
2889 | * | |
2890 | * We don't follow them on the oldname either to be compatible | |
2891 | * with linux 2.0, and to avoid hard-linking to directories | |
2892 | * and other special files. --ADM | |
2893 | */ | |
2e4d0924 HC |
2894 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2895 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2896 | { |
2897 | struct dentry *new_dentry; | |
2d8f3038 AV |
2898 | struct nameidata nd; |
2899 | struct path old_path; | |
11a7b371 | 2900 | int how = 0; |
1da177e4 | 2901 | int error; |
2ad94ae6 | 2902 | char *to; |
1da177e4 | 2903 | |
11a7b371 | 2904 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 2905 | return -EINVAL; |
11a7b371 AK |
2906 | /* |
2907 | * To use null names we require CAP_DAC_READ_SEARCH | |
2908 | * This ensures that not everyone will be able to create | |
2909 | * handlink using the passed filedescriptor. | |
2910 | */ | |
2911 | if (flags & AT_EMPTY_PATH) { | |
2912 | if (!capable(CAP_DAC_READ_SEARCH)) | |
2913 | return -ENOENT; | |
2914 | how = LOOKUP_EMPTY; | |
2915 | } | |
2916 | ||
2917 | if (flags & AT_SYMLINK_FOLLOW) | |
2918 | how |= LOOKUP_FOLLOW; | |
c04030e1 | 2919 | |
11a7b371 | 2920 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 2921 | if (error) |
2ad94ae6 AV |
2922 | return error; |
2923 | ||
2924 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2925 | if (error) |
2926 | goto out; | |
2927 | error = -EXDEV; | |
2d8f3038 | 2928 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2929 | goto out_release; |
2930 | new_dentry = lookup_create(&nd, 0); | |
2931 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2932 | if (IS_ERR(new_dentry)) |
2933 | goto out_unlock; | |
75c3f29d DH |
2934 | error = mnt_want_write(nd.path.mnt); |
2935 | if (error) | |
2936 | goto out_dput; | |
be6d3e56 KT |
2937 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2938 | if (error) | |
2939 | goto out_drop_write; | |
2d8f3038 | 2940 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2941 | out_drop_write: |
75c3f29d DH |
2942 | mnt_drop_write(nd.path.mnt); |
2943 | out_dput: | |
6902d925 DH |
2944 | dput(new_dentry); |
2945 | out_unlock: | |
4ac91378 | 2946 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2947 | out_release: |
1d957f9b | 2948 | path_put(&nd.path); |
2ad94ae6 | 2949 | putname(to); |
1da177e4 | 2950 | out: |
2d8f3038 | 2951 | path_put(&old_path); |
1da177e4 LT |
2952 | |
2953 | return error; | |
2954 | } | |
2955 | ||
3480b257 | 2956 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2957 | { |
c04030e1 | 2958 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2959 | } |
2960 | ||
1da177e4 LT |
2961 | /* |
2962 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2963 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2964 | * Problems: | |
2965 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2966 | * b) race potential - two innocent renames can create a loop together. | |
2967 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2968 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2969 | * story. |
2970 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2971 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2972 | * whether the target exists). Solution: try to be smart with locking |
2973 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2974 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2975 | * move will be locked. Thus we can rank directories by the tree |
2976 | * (ancestors first) and rank all non-directories after them. | |
2977 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2978 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2979 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2980 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2981 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 2982 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 2983 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2984 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 2985 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
2986 | * locking]. |
2987 | */ | |
75c96f85 AB |
2988 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2989 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2990 | { |
2991 | int error = 0; | |
9055cba7 | 2992 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
2993 | |
2994 | /* | |
2995 | * If we are going to change the parent - check write permissions, | |
2996 | * we'll need to flip '..'. | |
2997 | */ | |
2998 | if (new_dir != old_dir) { | |
f419a2e3 | 2999 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3000 | if (error) |
3001 | return error; | |
3002 | } | |
3003 | ||
3004 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3005 | if (error) | |
3006 | return error; | |
3007 | ||
d83c49f3 | 3008 | if (target) |
1b1dcc1b | 3009 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
3010 | |
3011 | error = -EBUSY; | |
3012 | if (d_mountpoint(old_dentry) || d_mountpoint(new_dentry)) | |
3013 | goto out; | |
3014 | ||
3cebde24 SW |
3015 | if (target) |
3016 | shrink_dcache_parent(new_dentry); | |
9055cba7 SW |
3017 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
3018 | if (error) | |
3019 | goto out; | |
3020 | ||
1da177e4 | 3021 | if (target) { |
9055cba7 SW |
3022 | target->i_flags |= S_DEAD; |
3023 | dont_mount(new_dentry); | |
1da177e4 | 3024 | } |
9055cba7 SW |
3025 | out: |
3026 | if (target) | |
3027 | mutex_unlock(&target->i_mutex); | |
e31e14ec | 3028 | if (!error) |
349457cc MF |
3029 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3030 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3031 | return error; |
3032 | } | |
3033 | ||
75c96f85 AB |
3034 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3035 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 | 3036 | { |
51892bbb | 3037 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3038 | int error; |
3039 | ||
3040 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3041 | if (error) | |
3042 | return error; | |
3043 | ||
3044 | dget(new_dentry); | |
1da177e4 | 3045 | if (target) |
1b1dcc1b | 3046 | mutex_lock(&target->i_mutex); |
51892bbb SW |
3047 | |
3048 | error = -EBUSY; | |
1da177e4 | 3049 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
51892bbb SW |
3050 | goto out; |
3051 | ||
3052 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3053 | if (error) | |
3054 | goto out; | |
3055 | ||
3056 | if (target) | |
3057 | dont_mount(new_dentry); | |
3058 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) | |
3059 | d_move(old_dentry, new_dentry); | |
3060 | out: | |
1da177e4 | 3061 | if (target) |
1b1dcc1b | 3062 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3063 | dput(new_dentry); |
3064 | return error; | |
3065 | } | |
3066 | ||
3067 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3068 | struct inode *new_dir, struct dentry *new_dentry) | |
3069 | { | |
3070 | int error; | |
3071 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3072 | const unsigned char *old_name; |
1da177e4 LT |
3073 | |
3074 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3075 | return 0; | |
3076 | ||
3077 | error = may_delete(old_dir, old_dentry, is_dir); | |
3078 | if (error) | |
3079 | return error; | |
3080 | ||
3081 | if (!new_dentry->d_inode) | |
a95164d9 | 3082 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3083 | else |
3084 | error = may_delete(new_dir, new_dentry, is_dir); | |
3085 | if (error) | |
3086 | return error; | |
3087 | ||
acfa4380 | 3088 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3089 | return -EPERM; |
3090 | ||
0eeca283 RL |
3091 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3092 | ||
1da177e4 LT |
3093 | if (is_dir) |
3094 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3095 | else | |
3096 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3097 | if (!error) |
3098 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3099 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3100 | fsnotify_oldname_free(old_name); |
3101 | ||
1da177e4 LT |
3102 | return error; |
3103 | } | |
3104 | ||
2e4d0924 HC |
3105 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3106 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3107 | { |
2ad94ae6 AV |
3108 | struct dentry *old_dir, *new_dir; |
3109 | struct dentry *old_dentry, *new_dentry; | |
3110 | struct dentry *trap; | |
1da177e4 | 3111 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3112 | char *from; |
3113 | char *to; | |
3114 | int error; | |
1da177e4 | 3115 | |
2ad94ae6 | 3116 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3117 | if (error) |
3118 | goto exit; | |
3119 | ||
2ad94ae6 | 3120 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3121 | if (error) |
3122 | goto exit1; | |
3123 | ||
3124 | error = -EXDEV; | |
4ac91378 | 3125 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3126 | goto exit2; |
3127 | ||
4ac91378 | 3128 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3129 | error = -EBUSY; |
3130 | if (oldnd.last_type != LAST_NORM) | |
3131 | goto exit2; | |
3132 | ||
4ac91378 | 3133 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3134 | if (newnd.last_type != LAST_NORM) |
3135 | goto exit2; | |
3136 | ||
0612d9fb OH |
3137 | oldnd.flags &= ~LOOKUP_PARENT; |
3138 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3139 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3140 | |
1da177e4 LT |
3141 | trap = lock_rename(new_dir, old_dir); |
3142 | ||
49705b77 | 3143 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3144 | error = PTR_ERR(old_dentry); |
3145 | if (IS_ERR(old_dentry)) | |
3146 | goto exit3; | |
3147 | /* source must exist */ | |
3148 | error = -ENOENT; | |
3149 | if (!old_dentry->d_inode) | |
3150 | goto exit4; | |
3151 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3152 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3153 | error = -ENOTDIR; | |
3154 | if (oldnd.last.name[oldnd.last.len]) | |
3155 | goto exit4; | |
3156 | if (newnd.last.name[newnd.last.len]) | |
3157 | goto exit4; | |
3158 | } | |
3159 | /* source should not be ancestor of target */ | |
3160 | error = -EINVAL; | |
3161 | if (old_dentry == trap) | |
3162 | goto exit4; | |
49705b77 | 3163 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3164 | error = PTR_ERR(new_dentry); |
3165 | if (IS_ERR(new_dentry)) | |
3166 | goto exit4; | |
3167 | /* target should not be an ancestor of source */ | |
3168 | error = -ENOTEMPTY; | |
3169 | if (new_dentry == trap) | |
3170 | goto exit5; | |
3171 | ||
9079b1eb DH |
3172 | error = mnt_want_write(oldnd.path.mnt); |
3173 | if (error) | |
3174 | goto exit5; | |
be6d3e56 KT |
3175 | error = security_path_rename(&oldnd.path, old_dentry, |
3176 | &newnd.path, new_dentry); | |
3177 | if (error) | |
3178 | goto exit6; | |
1da177e4 LT |
3179 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3180 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 3181 | exit6: |
9079b1eb | 3182 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
3183 | exit5: |
3184 | dput(new_dentry); | |
3185 | exit4: | |
3186 | dput(old_dentry); | |
3187 | exit3: | |
3188 | unlock_rename(new_dir, old_dir); | |
3189 | exit2: | |
1d957f9b | 3190 | path_put(&newnd.path); |
2ad94ae6 | 3191 | putname(to); |
1da177e4 | 3192 | exit1: |
1d957f9b | 3193 | path_put(&oldnd.path); |
1da177e4 | 3194 | putname(from); |
2ad94ae6 | 3195 | exit: |
1da177e4 LT |
3196 | return error; |
3197 | } | |
3198 | ||
a26eab24 | 3199 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3200 | { |
3201 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3202 | } | |
3203 | ||
1da177e4 LT |
3204 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3205 | { | |
3206 | int len; | |
3207 | ||
3208 | len = PTR_ERR(link); | |
3209 | if (IS_ERR(link)) | |
3210 | goto out; | |
3211 | ||
3212 | len = strlen(link); | |
3213 | if (len > (unsigned) buflen) | |
3214 | len = buflen; | |
3215 | if (copy_to_user(buffer, link, len)) | |
3216 | len = -EFAULT; | |
3217 | out: | |
3218 | return len; | |
3219 | } | |
3220 | ||
3221 | /* | |
3222 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3223 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3224 | * using) it for any given inode is up to filesystem. | |
3225 | */ | |
3226 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3227 | { | |
3228 | struct nameidata nd; | |
cc314eef | 3229 | void *cookie; |
694a1764 | 3230 | int res; |
cc314eef | 3231 | |
1da177e4 | 3232 | nd.depth = 0; |
cc314eef | 3233 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3234 | if (IS_ERR(cookie)) |
3235 | return PTR_ERR(cookie); | |
3236 | ||
3237 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3238 | if (dentry->d_inode->i_op->put_link) | |
3239 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3240 | return res; | |
1da177e4 LT |
3241 | } |
3242 | ||
3243 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3244 | { | |
3245 | return __vfs_follow_link(nd, link); | |
3246 | } | |
3247 | ||
3248 | /* get the link contents into pagecache */ | |
3249 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3250 | { | |
ebd09abb DG |
3251 | char *kaddr; |
3252 | struct page *page; | |
1da177e4 | 3253 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3254 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3255 | if (IS_ERR(page)) |
6fe6900e | 3256 | return (char*)page; |
1da177e4 | 3257 | *ppage = page; |
ebd09abb DG |
3258 | kaddr = kmap(page); |
3259 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3260 | return kaddr; | |
1da177e4 LT |
3261 | } |
3262 | ||
3263 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3264 | { | |
3265 | struct page *page = NULL; | |
3266 | char *s = page_getlink(dentry, &page); | |
3267 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3268 | if (page) { | |
3269 | kunmap(page); | |
3270 | page_cache_release(page); | |
3271 | } | |
3272 | return res; | |
3273 | } | |
3274 | ||
cc314eef | 3275 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3276 | { |
cc314eef | 3277 | struct page *page = NULL; |
1da177e4 | 3278 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3279 | return page; |
1da177e4 LT |
3280 | } |
3281 | ||
cc314eef | 3282 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3283 | { |
cc314eef LT |
3284 | struct page *page = cookie; |
3285 | ||
3286 | if (page) { | |
1da177e4 LT |
3287 | kunmap(page); |
3288 | page_cache_release(page); | |
1da177e4 LT |
3289 | } |
3290 | } | |
3291 | ||
54566b2c NP |
3292 | /* |
3293 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3294 | */ | |
3295 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3296 | { |
3297 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3298 | struct page *page; |
afddba49 | 3299 | void *fsdata; |
beb497ab | 3300 | int err; |
1da177e4 | 3301 | char *kaddr; |
54566b2c NP |
3302 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3303 | if (nofs) | |
3304 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3305 | |
7e53cac4 | 3306 | retry: |
afddba49 | 3307 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3308 | flags, &page, &fsdata); |
1da177e4 | 3309 | if (err) |
afddba49 NP |
3310 | goto fail; |
3311 | ||
1da177e4 LT |
3312 | kaddr = kmap_atomic(page, KM_USER0); |
3313 | memcpy(kaddr, symname, len-1); | |
3314 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
3315 | |
3316 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3317 | page, fsdata); | |
1da177e4 LT |
3318 | if (err < 0) |
3319 | goto fail; | |
afddba49 NP |
3320 | if (err < len-1) |
3321 | goto retry; | |
3322 | ||
1da177e4 LT |
3323 | mark_inode_dirty(inode); |
3324 | return 0; | |
1da177e4 LT |
3325 | fail: |
3326 | return err; | |
3327 | } | |
3328 | ||
0adb25d2 KK |
3329 | int page_symlink(struct inode *inode, const char *symname, int len) |
3330 | { | |
3331 | return __page_symlink(inode, symname, len, | |
54566b2c | 3332 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3333 | } |
3334 | ||
92e1d5be | 3335 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3336 | .readlink = generic_readlink, |
3337 | .follow_link = page_follow_link_light, | |
3338 | .put_link = page_put_link, | |
3339 | }; | |
3340 | ||
2d8f3038 | 3341 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3342 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3343 | EXPORT_SYMBOL(follow_down); |
3344 | EXPORT_SYMBOL(follow_up); | |
3345 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
3346 | EXPORT_SYMBOL(getname); | |
3347 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3348 | EXPORT_SYMBOL(lookup_one_len); |
3349 | EXPORT_SYMBOL(page_follow_link_light); | |
3350 | EXPORT_SYMBOL(page_put_link); | |
3351 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3352 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3353 | EXPORT_SYMBOL(page_symlink); |
3354 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
c9c6cac0 | 3355 | EXPORT_SYMBOL(kern_path_parent); |
d1811465 | 3356 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3357 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3358 | EXPORT_SYMBOL(inode_permission); |
8c744fb8 | 3359 | EXPORT_SYMBOL(file_permission); |
1da177e4 LT |
3360 | EXPORT_SYMBOL(unlock_rename); |
3361 | EXPORT_SYMBOL(vfs_create); | |
3362 | EXPORT_SYMBOL(vfs_follow_link); | |
3363 | EXPORT_SYMBOL(vfs_link); | |
3364 | EXPORT_SYMBOL(vfs_mkdir); | |
3365 | EXPORT_SYMBOL(vfs_mknod); | |
3366 | EXPORT_SYMBOL(generic_permission); | |
3367 | EXPORT_SYMBOL(vfs_readlink); | |
3368 | EXPORT_SYMBOL(vfs_rename); | |
3369 | EXPORT_SYMBOL(vfs_rmdir); | |
3370 | EXPORT_SYMBOL(vfs_symlink); | |
3371 | EXPORT_SYMBOL(vfs_unlink); | |
3372 | EXPORT_SYMBOL(dentry_unhash); | |
3373 | EXPORT_SYMBOL(generic_readlink); |