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> |
e77819e5 | 35 | #include <linux/posix_acl.h> |
1da177e4 LT |
36 | #include <asm/uaccess.h> |
37 | ||
e81e3f4d EP |
38 | #include "internal.h" |
39 | ||
1da177e4 LT |
40 | /* [Feb-1997 T. Schoebel-Theuer] |
41 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
42 | * were necessary because of omirr. The reason is that omirr needs | |
43 | * to know the _real_ pathname, not the user-supplied one, in case | |
44 | * of symlinks (and also when transname replacements occur). | |
45 | * | |
46 | * The new code replaces the old recursive symlink resolution with | |
47 | * an iterative one (in case of non-nested symlink chains). It does | |
48 | * this with calls to <fs>_follow_link(). | |
49 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
50 | * replaced with a single function lookup_dentry() that can handle all | |
51 | * the special cases of the former code. | |
52 | * | |
53 | * With the new dcache, the pathname is stored at each inode, at least as | |
54 | * long as the refcount of the inode is positive. As a side effect, the | |
55 | * size of the dcache depends on the inode cache and thus is dynamic. | |
56 | * | |
57 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
58 | * resolution to correspond with current state of the code. | |
59 | * | |
60 | * Note that the symlink resolution is not *completely* iterative. | |
61 | * There is still a significant amount of tail- and mid- recursion in | |
62 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
63 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
64 | * may return different results than <fs>_follow_link(). Many virtual | |
65 | * filesystems (including /proc) exhibit this behavior. | |
66 | */ | |
67 | ||
68 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
69 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
70 | * and the name already exists in form of a symlink, try to create the new | |
71 | * name indicated by the symlink. The old code always complained that the | |
72 | * name already exists, due to not following the symlink even if its target | |
73 | * is nonexistent. The new semantics affects also mknod() and link() when | |
25985edc | 74 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
75 | * |
76 | * I don't know which semantics is the right one, since I have no access | |
77 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
78 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
79 | * "old" one. Personally, I think the new semantics is much more logical. | |
80 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
81 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
82 | * and in the old Linux semantics. | |
83 | */ | |
84 | ||
85 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
86 | * semantics. See the comments in "open_namei" and "do_link" below. | |
87 | * | |
88 | * [10-Sep-98 Alan Modra] Another symlink change. | |
89 | */ | |
90 | ||
91 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
92 | * inside the path - always follow. | |
93 | * in the last component in creation/removal/renaming - never follow. | |
94 | * if LOOKUP_FOLLOW passed - follow. | |
95 | * if the pathname has trailing slashes - follow. | |
96 | * otherwise - don't follow. | |
97 | * (applied in that order). | |
98 | * | |
99 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
100 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
101 | * During the 2.4 we need to fix the userland stuff depending on it - | |
102 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
103 | * XEmacs seems to be relying on it... | |
104 | */ | |
105 | /* | |
106 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 107 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
108 | * any extra contention... |
109 | */ | |
110 | ||
111 | /* In order to reduce some races, while at the same time doing additional | |
112 | * checking and hopefully speeding things up, we copy filenames to the | |
113 | * kernel data space before using them.. | |
114 | * | |
115 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
116 | * PATH_MAX includes the nul terminator --RR. | |
117 | */ | |
858119e1 | 118 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
119 | { |
120 | int retval; | |
121 | unsigned long len = PATH_MAX; | |
122 | ||
123 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
124 | if ((unsigned long) filename >= TASK_SIZE) | |
125 | return -EFAULT; | |
126 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
127 | len = TASK_SIZE - (unsigned long) filename; | |
128 | } | |
129 | ||
130 | retval = strncpy_from_user(page, filename, len); | |
131 | if (retval > 0) { | |
132 | if (retval < len) | |
133 | return 0; | |
134 | return -ENAMETOOLONG; | |
135 | } else if (!retval) | |
136 | retval = -ENOENT; | |
137 | return retval; | |
138 | } | |
139 | ||
1fa1e7f6 | 140 | static char *getname_flags(const char __user *filename, int flags, int *empty) |
1da177e4 LT |
141 | { |
142 | char *tmp, *result; | |
143 | ||
144 | result = ERR_PTR(-ENOMEM); | |
145 | tmp = __getname(); | |
146 | if (tmp) { | |
147 | int retval = do_getname(filename, tmp); | |
148 | ||
149 | result = tmp; | |
150 | if (retval < 0) { | |
1fa1e7f6 AW |
151 | if (retval == -ENOENT && empty) |
152 | *empty = 1; | |
f52e0c11 AV |
153 | if (retval != -ENOENT || !(flags & LOOKUP_EMPTY)) { |
154 | __putname(tmp); | |
155 | result = ERR_PTR(retval); | |
156 | } | |
1da177e4 LT |
157 | } |
158 | } | |
159 | audit_getname(result); | |
160 | return result; | |
161 | } | |
162 | ||
f52e0c11 AV |
163 | char *getname(const char __user * filename) |
164 | { | |
1fa1e7f6 | 165 | return getname_flags(filename, 0, 0); |
f52e0c11 AV |
166 | } |
167 | ||
1da177e4 LT |
168 | #ifdef CONFIG_AUDITSYSCALL |
169 | void putname(const char *name) | |
170 | { | |
5ac3a9c2 | 171 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
172 | audit_putname(name); |
173 | else | |
174 | __putname(name); | |
175 | } | |
176 | EXPORT_SYMBOL(putname); | |
177 | #endif | |
178 | ||
e77819e5 LT |
179 | static int check_acl(struct inode *inode, int mask) |
180 | { | |
84635d68 | 181 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
182 | struct posix_acl *acl; |
183 | ||
e77819e5 | 184 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
185 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
186 | if (!acl) | |
e77819e5 | 187 | return -EAGAIN; |
3567866b AV |
188 | /* no ->get_acl() calls in RCU mode... */ |
189 | if (acl == ACL_NOT_CACHED) | |
190 | return -ECHILD; | |
206b1d09 | 191 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
192 | } |
193 | ||
194 | acl = get_cached_acl(inode, ACL_TYPE_ACCESS); | |
195 | ||
196 | /* | |
4e34e719 CH |
197 | * A filesystem can force a ACL callback by just never filling the |
198 | * ACL cache. But normally you'd fill the cache either at inode | |
199 | * instantiation time, or on the first ->get_acl call. | |
e77819e5 | 200 | * |
4e34e719 CH |
201 | * If the filesystem doesn't have a get_acl() function at all, we'll |
202 | * just create the negative cache entry. | |
e77819e5 LT |
203 | */ |
204 | if (acl == ACL_NOT_CACHED) { | |
4e34e719 CH |
205 | if (inode->i_op->get_acl) { |
206 | acl = inode->i_op->get_acl(inode, ACL_TYPE_ACCESS); | |
207 | if (IS_ERR(acl)) | |
208 | return PTR_ERR(acl); | |
209 | } else { | |
210 | set_cached_acl(inode, ACL_TYPE_ACCESS, NULL); | |
211 | return -EAGAIN; | |
212 | } | |
e77819e5 LT |
213 | } |
214 | ||
215 | if (acl) { | |
216 | int error = posix_acl_permission(inode, acl, mask); | |
217 | posix_acl_release(acl); | |
218 | return error; | |
219 | } | |
84635d68 | 220 | #endif |
e77819e5 LT |
221 | |
222 | return -EAGAIN; | |
223 | } | |
224 | ||
5909ccaa | 225 | /* |
948409c7 | 226 | * This does the basic permission checking |
1da177e4 | 227 | */ |
7e40145e | 228 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 229 | { |
26cf46be | 230 | unsigned int mode = inode->i_mode; |
1da177e4 | 231 | |
e795b717 SH |
232 | if (current_user_ns() != inode_userns(inode)) |
233 | goto other_perms; | |
234 | ||
14067ff5 | 235 | if (likely(current_fsuid() == inode->i_uid)) |
1da177e4 LT |
236 | mode >>= 6; |
237 | else { | |
e77819e5 | 238 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 239 | int error = check_acl(inode, mask); |
b74c79e9 NP |
240 | if (error != -EAGAIN) |
241 | return error; | |
1da177e4 LT |
242 | } |
243 | ||
244 | if (in_group_p(inode->i_gid)) | |
245 | mode >>= 3; | |
246 | } | |
247 | ||
e795b717 | 248 | other_perms: |
1da177e4 LT |
249 | /* |
250 | * If the DACs are ok we don't need any capability check. | |
251 | */ | |
9c2c7039 | 252 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 253 | return 0; |
5909ccaa LT |
254 | return -EACCES; |
255 | } | |
256 | ||
257 | /** | |
b74c79e9 | 258 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 259 | * @inode: inode to check access rights for |
8fd90c8d | 260 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
261 | * |
262 | * Used to check for read/write/execute permissions on a file. | |
263 | * We use "fsuid" for this, letting us set arbitrary permissions | |
264 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
265 | * are used for other things. |
266 | * | |
267 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
268 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
269 | * It would then be called again in ref-walk mode. | |
5909ccaa | 270 | */ |
2830ba7f | 271 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
272 | { |
273 | int ret; | |
274 | ||
275 | /* | |
948409c7 | 276 | * Do the basic permission checks. |
5909ccaa | 277 | */ |
7e40145e | 278 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
279 | if (ret != -EACCES) |
280 | return ret; | |
1da177e4 | 281 | |
d594e7ec AV |
282 | if (S_ISDIR(inode->i_mode)) { |
283 | /* DACs are overridable for directories */ | |
284 | if (ns_capable(inode_userns(inode), CAP_DAC_OVERRIDE)) | |
285 | return 0; | |
286 | if (!(mask & MAY_WRITE)) | |
287 | if (ns_capable(inode_userns(inode), CAP_DAC_READ_SEARCH)) | |
288 | return 0; | |
289 | return -EACCES; | |
290 | } | |
1da177e4 LT |
291 | /* |
292 | * Read/write DACs are always overridable. | |
d594e7ec AV |
293 | * Executable DACs are overridable when there is |
294 | * at least one exec bit set. | |
1da177e4 | 295 | */ |
d594e7ec | 296 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) |
e795b717 | 297 | if (ns_capable(inode_userns(inode), CAP_DAC_OVERRIDE)) |
1da177e4 LT |
298 | return 0; |
299 | ||
300 | /* | |
301 | * Searching includes executable on directories, else just read. | |
302 | */ | |
7ea66001 | 303 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 304 | if (mask == MAY_READ) |
e795b717 | 305 | if (ns_capable(inode_userns(inode), CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
306 | return 0; |
307 | ||
308 | return -EACCES; | |
309 | } | |
310 | ||
3ddcd056 LT |
311 | /* |
312 | * We _really_ want to just do "generic_permission()" without | |
313 | * even looking at the inode->i_op values. So we keep a cache | |
314 | * flag in inode->i_opflags, that says "this has not special | |
315 | * permission function, use the fast case". | |
316 | */ | |
317 | static inline int do_inode_permission(struct inode *inode, int mask) | |
318 | { | |
319 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
320 | if (likely(inode->i_op->permission)) | |
321 | return inode->i_op->permission(inode, mask); | |
322 | ||
323 | /* This gets set once for the inode lifetime */ | |
324 | spin_lock(&inode->i_lock); | |
325 | inode->i_opflags |= IOP_FASTPERM; | |
326 | spin_unlock(&inode->i_lock); | |
327 | } | |
328 | return generic_permission(inode, mask); | |
329 | } | |
330 | ||
cb23beb5 CH |
331 | /** |
332 | * inode_permission - check for access rights to a given inode | |
333 | * @inode: inode to check permission on | |
8fd90c8d | 334 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
cb23beb5 CH |
335 | * |
336 | * Used to check for read/write/execute permissions on an inode. | |
337 | * We use "fsuid" for this, letting us set arbitrary permissions | |
338 | * for filesystem access without changing the "normal" uids which | |
339 | * are used for other things. | |
948409c7 AG |
340 | * |
341 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
cb23beb5 | 342 | */ |
f419a2e3 | 343 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 344 | { |
e6305c43 | 345 | int retval; |
1da177e4 | 346 | |
3ddcd056 | 347 | if (unlikely(mask & MAY_WRITE)) { |
22590e41 | 348 | umode_t mode = inode->i_mode; |
1da177e4 LT |
349 | |
350 | /* | |
351 | * Nobody gets write access to a read-only fs. | |
352 | */ | |
353 | if (IS_RDONLY(inode) && | |
354 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
355 | return -EROFS; | |
356 | ||
357 | /* | |
358 | * Nobody gets write access to an immutable file. | |
359 | */ | |
360 | if (IS_IMMUTABLE(inode)) | |
361 | return -EACCES; | |
362 | } | |
363 | ||
3ddcd056 | 364 | retval = do_inode_permission(inode, mask); |
1da177e4 LT |
365 | if (retval) |
366 | return retval; | |
367 | ||
08ce5f16 SH |
368 | retval = devcgroup_inode_permission(inode, mask); |
369 | if (retval) | |
370 | return retval; | |
371 | ||
d09ca739 | 372 | return security_inode_permission(inode, mask); |
1da177e4 LT |
373 | } |
374 | ||
5dd784d0 JB |
375 | /** |
376 | * path_get - get a reference to a path | |
377 | * @path: path to get the reference to | |
378 | * | |
379 | * Given a path increment the reference count to the dentry and the vfsmount. | |
380 | */ | |
381 | void path_get(struct path *path) | |
382 | { | |
383 | mntget(path->mnt); | |
384 | dget(path->dentry); | |
385 | } | |
386 | EXPORT_SYMBOL(path_get); | |
387 | ||
1d957f9b JB |
388 | /** |
389 | * path_put - put a reference to a path | |
390 | * @path: path to put the reference to | |
391 | * | |
392 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
393 | */ | |
394 | void path_put(struct path *path) | |
1da177e4 | 395 | { |
1d957f9b JB |
396 | dput(path->dentry); |
397 | mntput(path->mnt); | |
1da177e4 | 398 | } |
1d957f9b | 399 | EXPORT_SYMBOL(path_put); |
1da177e4 | 400 | |
19660af7 | 401 | /* |
31e6b01f | 402 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
403 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
404 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
405 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
406 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
407 | * got stuck, so ref-walk may continue from there. If this is not successful | |
408 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
409 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 410 | */ |
31e6b01f NP |
411 | |
412 | /** | |
19660af7 AV |
413 | * unlazy_walk - try to switch to ref-walk mode. |
414 | * @nd: nameidata pathwalk data | |
415 | * @dentry: child of nd->path.dentry or NULL | |
39191628 | 416 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 417 | * |
19660af7 AV |
418 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
419 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
420 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 421 | */ |
19660af7 | 422 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry) |
31e6b01f NP |
423 | { |
424 | struct fs_struct *fs = current->fs; | |
425 | struct dentry *parent = nd->path.dentry; | |
5b6ca027 | 426 | int want_root = 0; |
31e6b01f NP |
427 | |
428 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
5b6ca027 AV |
429 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
430 | want_root = 1; | |
31e6b01f NP |
431 | spin_lock(&fs->lock); |
432 | if (nd->root.mnt != fs->root.mnt || | |
433 | nd->root.dentry != fs->root.dentry) | |
434 | goto err_root; | |
435 | } | |
436 | spin_lock(&parent->d_lock); | |
19660af7 AV |
437 | if (!dentry) { |
438 | if (!__d_rcu_to_refcount(parent, nd->seq)) | |
439 | goto err_parent; | |
440 | BUG_ON(nd->inode != parent->d_inode); | |
441 | } else { | |
94c0d4ec AV |
442 | if (dentry->d_parent != parent) |
443 | goto err_parent; | |
19660af7 AV |
444 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); |
445 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
446 | goto err_child; | |
447 | /* | |
448 | * If the sequence check on the child dentry passed, then | |
449 | * the child has not been removed from its parent. This | |
450 | * means the parent dentry must be valid and able to take | |
451 | * a reference at this point. | |
452 | */ | |
453 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
454 | BUG_ON(!parent->d_count); | |
455 | parent->d_count++; | |
456 | spin_unlock(&dentry->d_lock); | |
457 | } | |
31e6b01f | 458 | spin_unlock(&parent->d_lock); |
5b6ca027 | 459 | if (want_root) { |
31e6b01f NP |
460 | path_get(&nd->root); |
461 | spin_unlock(&fs->lock); | |
462 | } | |
463 | mntget(nd->path.mnt); | |
464 | ||
465 | rcu_read_unlock(); | |
466 | br_read_unlock(vfsmount_lock); | |
467 | nd->flags &= ~LOOKUP_RCU; | |
468 | return 0; | |
19660af7 AV |
469 | |
470 | err_child: | |
31e6b01f | 471 | spin_unlock(&dentry->d_lock); |
19660af7 | 472 | err_parent: |
31e6b01f NP |
473 | spin_unlock(&parent->d_lock); |
474 | err_root: | |
5b6ca027 | 475 | if (want_root) |
31e6b01f NP |
476 | spin_unlock(&fs->lock); |
477 | return -ECHILD; | |
478 | } | |
479 | ||
834f2a4a TM |
480 | /** |
481 | * release_open_intent - free up open intent resources | |
482 | * @nd: pointer to nameidata | |
483 | */ | |
484 | void release_open_intent(struct nameidata *nd) | |
485 | { | |
2dab5974 LT |
486 | struct file *file = nd->intent.open.file; |
487 | ||
488 | if (file && !IS_ERR(file)) { | |
489 | if (file->f_path.dentry == NULL) | |
490 | put_filp(file); | |
491 | else | |
492 | fput(file); | |
493 | } | |
834f2a4a TM |
494 | } |
495 | ||
f60aef7e | 496 | static inline int d_revalidate(struct dentry *dentry, struct nameidata *nd) |
34286d66 | 497 | { |
f60aef7e | 498 | return dentry->d_op->d_revalidate(dentry, nd); |
34286d66 NP |
499 | } |
500 | ||
9f1fafee AV |
501 | /** |
502 | * complete_walk - successful completion of path walk | |
503 | * @nd: pointer nameidata | |
39159de2 | 504 | * |
9f1fafee AV |
505 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
506 | * Revalidate the final result, unless we'd already done that during | |
507 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
508 | * success, -error on failure. In case of failure caller does not | |
509 | * need to drop nd->path. | |
39159de2 | 510 | */ |
9f1fafee | 511 | static int complete_walk(struct nameidata *nd) |
39159de2 | 512 | { |
16c2cd71 | 513 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 514 | int status; |
39159de2 | 515 | |
9f1fafee AV |
516 | if (nd->flags & LOOKUP_RCU) { |
517 | nd->flags &= ~LOOKUP_RCU; | |
518 | if (!(nd->flags & LOOKUP_ROOT)) | |
519 | nd->root.mnt = NULL; | |
520 | spin_lock(&dentry->d_lock); | |
521 | if (unlikely(!__d_rcu_to_refcount(dentry, nd->seq))) { | |
522 | spin_unlock(&dentry->d_lock); | |
523 | rcu_read_unlock(); | |
524 | br_read_unlock(vfsmount_lock); | |
525 | return -ECHILD; | |
526 | } | |
527 | BUG_ON(nd->inode != dentry->d_inode); | |
528 | spin_unlock(&dentry->d_lock); | |
529 | mntget(nd->path.mnt); | |
530 | rcu_read_unlock(); | |
531 | br_read_unlock(vfsmount_lock); | |
532 | } | |
533 | ||
16c2cd71 AV |
534 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
535 | return 0; | |
536 | ||
537 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
39159de2 JL |
538 | return 0; |
539 | ||
16c2cd71 AV |
540 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) |
541 | return 0; | |
542 | ||
543 | /* Note: we do not d_invalidate() */ | |
34286d66 | 544 | status = d_revalidate(dentry, nd); |
39159de2 JL |
545 | if (status > 0) |
546 | return 0; | |
547 | ||
16c2cd71 | 548 | if (!status) |
39159de2 | 549 | status = -ESTALE; |
16c2cd71 | 550 | |
9f1fafee | 551 | path_put(&nd->path); |
39159de2 JL |
552 | return status; |
553 | } | |
554 | ||
2a737871 AV |
555 | static __always_inline void set_root(struct nameidata *nd) |
556 | { | |
f7ad3c6b MS |
557 | if (!nd->root.mnt) |
558 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
559 | } |
560 | ||
6de88d72 AV |
561 | static int link_path_walk(const char *, struct nameidata *); |
562 | ||
31e6b01f NP |
563 | static __always_inline void set_root_rcu(struct nameidata *nd) |
564 | { | |
565 | if (!nd->root.mnt) { | |
566 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
567 | unsigned seq; |
568 | ||
569 | do { | |
570 | seq = read_seqcount_begin(&fs->seq); | |
571 | nd->root = fs->root; | |
c1530019 | 572 | nd->seq = __read_seqcount_begin(&nd->root.dentry->d_seq); |
c28cc364 | 573 | } while (read_seqcount_retry(&fs->seq, seq)); |
31e6b01f NP |
574 | } |
575 | } | |
576 | ||
f1662356 | 577 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 578 | { |
31e6b01f NP |
579 | int ret; |
580 | ||
1da177e4 LT |
581 | if (IS_ERR(link)) |
582 | goto fail; | |
583 | ||
584 | if (*link == '/') { | |
2a737871 | 585 | set_root(nd); |
1d957f9b | 586 | path_put(&nd->path); |
2a737871 AV |
587 | nd->path = nd->root; |
588 | path_get(&nd->root); | |
16c2cd71 | 589 | nd->flags |= LOOKUP_JUMPED; |
1da177e4 | 590 | } |
31e6b01f | 591 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 592 | |
31e6b01f NP |
593 | ret = link_path_walk(link, nd); |
594 | return ret; | |
1da177e4 | 595 | fail: |
1d957f9b | 596 | path_put(&nd->path); |
1da177e4 LT |
597 | return PTR_ERR(link); |
598 | } | |
599 | ||
1d957f9b | 600 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
601 | { |
602 | dput(path->dentry); | |
4ac91378 | 603 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
604 | mntput(path->mnt); |
605 | } | |
606 | ||
7b9337aa NP |
607 | static inline void path_to_nameidata(const struct path *path, |
608 | struct nameidata *nd) | |
051d3812 | 609 | { |
31e6b01f NP |
610 | if (!(nd->flags & LOOKUP_RCU)) { |
611 | dput(nd->path.dentry); | |
612 | if (nd->path.mnt != path->mnt) | |
613 | mntput(nd->path.mnt); | |
9a229683 | 614 | } |
31e6b01f | 615 | nd->path.mnt = path->mnt; |
4ac91378 | 616 | nd->path.dentry = path->dentry; |
051d3812 IK |
617 | } |
618 | ||
574197e0 AV |
619 | static inline void put_link(struct nameidata *nd, struct path *link, void *cookie) |
620 | { | |
621 | struct inode *inode = link->dentry->d_inode; | |
622 | if (!IS_ERR(cookie) && inode->i_op->put_link) | |
623 | inode->i_op->put_link(link->dentry, nd, cookie); | |
624 | path_put(link); | |
625 | } | |
626 | ||
def4af30 | 627 | static __always_inline int |
574197e0 | 628 | follow_link(struct path *link, struct nameidata *nd, void **p) |
1da177e4 LT |
629 | { |
630 | int error; | |
7b9337aa | 631 | struct dentry *dentry = link->dentry; |
1da177e4 | 632 | |
844a3917 AV |
633 | BUG_ON(nd->flags & LOOKUP_RCU); |
634 | ||
0e794589 AV |
635 | if (link->mnt == nd->path.mnt) |
636 | mntget(link->mnt); | |
637 | ||
574197e0 AV |
638 | if (unlikely(current->total_link_count >= 40)) { |
639 | *p = ERR_PTR(-ELOOP); /* no ->put_link(), please */ | |
574197e0 AV |
640 | path_put(&nd->path); |
641 | return -ELOOP; | |
642 | } | |
643 | cond_resched(); | |
644 | current->total_link_count++; | |
645 | ||
7b9337aa | 646 | touch_atime(link->mnt, dentry); |
1da177e4 | 647 | nd_set_link(nd, NULL); |
cd4e91d3 | 648 | |
36f3b4f6 AV |
649 | error = security_inode_follow_link(link->dentry, nd); |
650 | if (error) { | |
651 | *p = ERR_PTR(error); /* no ->put_link(), please */ | |
652 | path_put(&nd->path); | |
653 | return error; | |
654 | } | |
655 | ||
86acdca1 | 656 | nd->last_type = LAST_BIND; |
def4af30 AV |
657 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
658 | error = PTR_ERR(*p); | |
659 | if (!IS_ERR(*p)) { | |
1da177e4 | 660 | char *s = nd_get_link(nd); |
cc314eef | 661 | error = 0; |
1da177e4 LT |
662 | if (s) |
663 | error = __vfs_follow_link(nd, s); | |
bcda7652 | 664 | else if (nd->last_type == LAST_BIND) { |
16c2cd71 | 665 | nd->flags |= LOOKUP_JUMPED; |
b21041d0 AV |
666 | nd->inode = nd->path.dentry->d_inode; |
667 | if (nd->inode->i_op->follow_link) { | |
bcda7652 AV |
668 | /* stepped on a _really_ weird one */ |
669 | path_put(&nd->path); | |
670 | error = -ELOOP; | |
671 | } | |
672 | } | |
1da177e4 | 673 | } |
1da177e4 LT |
674 | return error; |
675 | } | |
676 | ||
31e6b01f NP |
677 | static int follow_up_rcu(struct path *path) |
678 | { | |
679 | struct vfsmount *parent; | |
680 | struct dentry *mountpoint; | |
681 | ||
682 | parent = path->mnt->mnt_parent; | |
683 | if (parent == path->mnt) | |
684 | return 0; | |
685 | mountpoint = path->mnt->mnt_mountpoint; | |
686 | path->dentry = mountpoint; | |
687 | path->mnt = parent; | |
688 | return 1; | |
689 | } | |
690 | ||
bab77ebf | 691 | int follow_up(struct path *path) |
1da177e4 LT |
692 | { |
693 | struct vfsmount *parent; | |
694 | struct dentry *mountpoint; | |
99b7db7b NP |
695 | |
696 | br_read_lock(vfsmount_lock); | |
bab77ebf AV |
697 | parent = path->mnt->mnt_parent; |
698 | if (parent == path->mnt) { | |
99b7db7b | 699 | br_read_unlock(vfsmount_lock); |
1da177e4 LT |
700 | return 0; |
701 | } | |
702 | mntget(parent); | |
bab77ebf | 703 | mountpoint = dget(path->mnt->mnt_mountpoint); |
99b7db7b | 704 | br_read_unlock(vfsmount_lock); |
bab77ebf AV |
705 | dput(path->dentry); |
706 | path->dentry = mountpoint; | |
707 | mntput(path->mnt); | |
708 | path->mnt = parent; | |
1da177e4 LT |
709 | return 1; |
710 | } | |
711 | ||
b5c84bf6 | 712 | /* |
9875cf80 DH |
713 | * Perform an automount |
714 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
715 | * were called with. | |
1da177e4 | 716 | */ |
9875cf80 DH |
717 | static int follow_automount(struct path *path, unsigned flags, |
718 | bool *need_mntput) | |
31e6b01f | 719 | { |
9875cf80 | 720 | struct vfsmount *mnt; |
ea5b778a | 721 | int err; |
9875cf80 DH |
722 | |
723 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
724 | return -EREMOTE; | |
725 | ||
0ec26fd0 MS |
726 | /* We don't want to mount if someone's just doing a stat - |
727 | * unless they're stat'ing a directory and appended a '/' to | |
728 | * the name. | |
729 | * | |
730 | * We do, however, want to mount if someone wants to open or | |
731 | * create a file of any type under the mountpoint, wants to | |
732 | * traverse through the mountpoint or wants to open the | |
733 | * mounted directory. Also, autofs may mark negative dentries | |
734 | * as being automount points. These will need the attentions | |
735 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 736 | */ |
0ec26fd0 | 737 | if (!(flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
d94c177b | 738 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
0ec26fd0 MS |
739 | path->dentry->d_inode) |
740 | return -EISDIR; | |
741 | ||
9875cf80 DH |
742 | current->total_link_count++; |
743 | if (current->total_link_count >= 40) | |
744 | return -ELOOP; | |
745 | ||
746 | mnt = path->dentry->d_op->d_automount(path); | |
747 | if (IS_ERR(mnt)) { | |
748 | /* | |
749 | * The filesystem is allowed to return -EISDIR here to indicate | |
750 | * it doesn't want to automount. For instance, autofs would do | |
751 | * this so that its userspace daemon can mount on this dentry. | |
752 | * | |
753 | * However, we can only permit this if it's a terminal point in | |
754 | * the path being looked up; if it wasn't then the remainder of | |
755 | * the path is inaccessible and we should say so. | |
756 | */ | |
49084c3b | 757 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_PARENT)) |
9875cf80 DH |
758 | return -EREMOTE; |
759 | return PTR_ERR(mnt); | |
31e6b01f | 760 | } |
ea5b778a | 761 | |
9875cf80 DH |
762 | if (!mnt) /* mount collision */ |
763 | return 0; | |
31e6b01f | 764 | |
8aef1884 AV |
765 | if (!*need_mntput) { |
766 | /* lock_mount() may release path->mnt on error */ | |
767 | mntget(path->mnt); | |
768 | *need_mntput = true; | |
769 | } | |
19a167af | 770 | err = finish_automount(mnt, path); |
9875cf80 | 771 | |
ea5b778a DH |
772 | switch (err) { |
773 | case -EBUSY: | |
774 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 775 | return 0; |
ea5b778a | 776 | case 0: |
8aef1884 | 777 | path_put(path); |
ea5b778a DH |
778 | path->mnt = mnt; |
779 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 780 | return 0; |
19a167af AV |
781 | default: |
782 | return err; | |
ea5b778a | 783 | } |
19a167af | 784 | |
463ffb2e AV |
785 | } |
786 | ||
9875cf80 DH |
787 | /* |
788 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 789 | * - Flagged for transit management (autofs) |
9875cf80 DH |
790 | * - Flagged as mountpoint |
791 | * - Flagged as automount point | |
792 | * | |
793 | * This may only be called in refwalk mode. | |
794 | * | |
795 | * Serialization is taken care of in namespace.c | |
796 | */ | |
797 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 798 | { |
8aef1884 | 799 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
800 | unsigned managed; |
801 | bool need_mntput = false; | |
8aef1884 | 802 | int ret = 0; |
9875cf80 DH |
803 | |
804 | /* Given that we're not holding a lock here, we retain the value in a | |
805 | * local variable for each dentry as we look at it so that we don't see | |
806 | * the components of that value change under us */ | |
807 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
808 | managed &= DCACHE_MANAGED_DENTRY, | |
809 | unlikely(managed != 0)) { | |
cc53ce53 DH |
810 | /* Allow the filesystem to manage the transit without i_mutex |
811 | * being held. */ | |
812 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
813 | BUG_ON(!path->dentry->d_op); | |
814 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 815 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 816 | if (ret < 0) |
8aef1884 | 817 | break; |
cc53ce53 DH |
818 | } |
819 | ||
9875cf80 DH |
820 | /* Transit to a mounted filesystem. */ |
821 | if (managed & DCACHE_MOUNTED) { | |
822 | struct vfsmount *mounted = lookup_mnt(path); | |
823 | if (mounted) { | |
824 | dput(path->dentry); | |
825 | if (need_mntput) | |
826 | mntput(path->mnt); | |
827 | path->mnt = mounted; | |
828 | path->dentry = dget(mounted->mnt_root); | |
829 | need_mntput = true; | |
830 | continue; | |
831 | } | |
832 | ||
833 | /* Something is mounted on this dentry in another | |
834 | * namespace and/or whatever was mounted there in this | |
835 | * namespace got unmounted before we managed to get the | |
836 | * vfsmount_lock */ | |
837 | } | |
838 | ||
839 | /* Handle an automount point */ | |
840 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
841 | ret = follow_automount(path, flags, &need_mntput); | |
842 | if (ret < 0) | |
8aef1884 | 843 | break; |
9875cf80 DH |
844 | continue; |
845 | } | |
846 | ||
847 | /* We didn't change the current path point */ | |
848 | break; | |
1da177e4 | 849 | } |
8aef1884 AV |
850 | |
851 | if (need_mntput && path->mnt == mnt) | |
852 | mntput(path->mnt); | |
853 | if (ret == -EISDIR) | |
854 | ret = 0; | |
855 | return ret; | |
1da177e4 LT |
856 | } |
857 | ||
cc53ce53 | 858 | int follow_down_one(struct path *path) |
1da177e4 LT |
859 | { |
860 | struct vfsmount *mounted; | |
861 | ||
1c755af4 | 862 | mounted = lookup_mnt(path); |
1da177e4 | 863 | if (mounted) { |
9393bd07 AV |
864 | dput(path->dentry); |
865 | mntput(path->mnt); | |
866 | path->mnt = mounted; | |
867 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
868 | return 1; |
869 | } | |
870 | return 0; | |
871 | } | |
872 | ||
62a7375e IK |
873 | static inline bool managed_dentry_might_block(struct dentry *dentry) |
874 | { | |
875 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT && | |
876 | dentry->d_op->d_manage(dentry, true) < 0); | |
877 | } | |
878 | ||
9875cf80 | 879 | /* |
287548e4 AV |
880 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
881 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
882 | */ |
883 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 884 | struct inode **inode) |
9875cf80 | 885 | { |
62a7375e | 886 | for (;;) { |
9875cf80 | 887 | struct vfsmount *mounted; |
62a7375e IK |
888 | /* |
889 | * Don't forget we might have a non-mountpoint managed dentry | |
890 | * that wants to block transit. | |
891 | */ | |
287548e4 | 892 | if (unlikely(managed_dentry_might_block(path->dentry))) |
ab90911f | 893 | return false; |
62a7375e IK |
894 | |
895 | if (!d_mountpoint(path->dentry)) | |
896 | break; | |
897 | ||
9875cf80 DH |
898 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
899 | if (!mounted) | |
900 | break; | |
901 | path->mnt = mounted; | |
902 | path->dentry = mounted->mnt_root; | |
903 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); | |
59430262 LT |
904 | /* |
905 | * Update the inode too. We don't need to re-check the | |
906 | * dentry sequence number here after this d_inode read, | |
907 | * because a mount-point is always pinned. | |
908 | */ | |
909 | *inode = path->dentry->d_inode; | |
9875cf80 | 910 | } |
9875cf80 DH |
911 | return true; |
912 | } | |
913 | ||
dea39376 | 914 | static void follow_mount_rcu(struct nameidata *nd) |
287548e4 | 915 | { |
dea39376 | 916 | while (d_mountpoint(nd->path.dentry)) { |
287548e4 | 917 | struct vfsmount *mounted; |
dea39376 | 918 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry, 1); |
287548e4 AV |
919 | if (!mounted) |
920 | break; | |
dea39376 AV |
921 | nd->path.mnt = mounted; |
922 | nd->path.dentry = mounted->mnt_root; | |
923 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
287548e4 AV |
924 | } |
925 | } | |
926 | ||
31e6b01f NP |
927 | static int follow_dotdot_rcu(struct nameidata *nd) |
928 | { | |
31e6b01f NP |
929 | set_root_rcu(nd); |
930 | ||
9875cf80 | 931 | while (1) { |
31e6b01f NP |
932 | if (nd->path.dentry == nd->root.dentry && |
933 | nd->path.mnt == nd->root.mnt) { | |
934 | break; | |
935 | } | |
936 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
937 | struct dentry *old = nd->path.dentry; | |
938 | struct dentry *parent = old->d_parent; | |
939 | unsigned seq; | |
940 | ||
941 | seq = read_seqcount_begin(&parent->d_seq); | |
942 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 943 | goto failed; |
31e6b01f NP |
944 | nd->path.dentry = parent; |
945 | nd->seq = seq; | |
946 | break; | |
947 | } | |
948 | if (!follow_up_rcu(&nd->path)) | |
949 | break; | |
950 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
31e6b01f | 951 | } |
dea39376 AV |
952 | follow_mount_rcu(nd); |
953 | nd->inode = nd->path.dentry->d_inode; | |
31e6b01f | 954 | return 0; |
ef7562d5 AV |
955 | |
956 | failed: | |
957 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
958 | if (!(nd->flags & LOOKUP_ROOT)) |
959 | nd->root.mnt = NULL; | |
ef7562d5 AV |
960 | rcu_read_unlock(); |
961 | br_read_unlock(vfsmount_lock); | |
962 | return -ECHILD; | |
31e6b01f NP |
963 | } |
964 | ||
cc53ce53 DH |
965 | /* |
966 | * Follow down to the covering mount currently visible to userspace. At each | |
967 | * point, the filesystem owning that dentry may be queried as to whether the | |
968 | * caller is permitted to proceed or not. | |
cc53ce53 | 969 | */ |
7cc90cc3 | 970 | int follow_down(struct path *path) |
cc53ce53 DH |
971 | { |
972 | unsigned managed; | |
973 | int ret; | |
974 | ||
975 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
976 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
977 | /* Allow the filesystem to manage the transit without i_mutex | |
978 | * being held. | |
979 | * | |
980 | * We indicate to the filesystem if someone is trying to mount | |
981 | * something here. This gives autofs the chance to deny anyone | |
982 | * other than its daemon the right to mount on its | |
983 | * superstructure. | |
984 | * | |
985 | * The filesystem may sleep at this point. | |
986 | */ | |
987 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
988 | BUG_ON(!path->dentry->d_op); | |
989 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 990 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 991 | path->dentry, false); |
cc53ce53 DH |
992 | if (ret < 0) |
993 | return ret == -EISDIR ? 0 : ret; | |
994 | } | |
995 | ||
996 | /* Transit to a mounted filesystem. */ | |
997 | if (managed & DCACHE_MOUNTED) { | |
998 | struct vfsmount *mounted = lookup_mnt(path); | |
999 | if (!mounted) | |
1000 | break; | |
1001 | dput(path->dentry); | |
1002 | mntput(path->mnt); | |
1003 | path->mnt = mounted; | |
1004 | path->dentry = dget(mounted->mnt_root); | |
1005 | continue; | |
1006 | } | |
1007 | ||
1008 | /* Don't handle automount points here */ | |
1009 | break; | |
1010 | } | |
1011 | return 0; | |
1012 | } | |
1013 | ||
9875cf80 DH |
1014 | /* |
1015 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1016 | */ | |
1017 | static void follow_mount(struct path *path) | |
1018 | { | |
1019 | while (d_mountpoint(path->dentry)) { | |
1020 | struct vfsmount *mounted = lookup_mnt(path); | |
1021 | if (!mounted) | |
1022 | break; | |
1023 | dput(path->dentry); | |
1024 | mntput(path->mnt); | |
1025 | path->mnt = mounted; | |
1026 | path->dentry = dget(mounted->mnt_root); | |
1027 | } | |
1028 | } | |
1029 | ||
31e6b01f | 1030 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1031 | { |
2a737871 | 1032 | set_root(nd); |
e518ddb7 | 1033 | |
1da177e4 | 1034 | while(1) { |
4ac91378 | 1035 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1036 | |
2a737871 AV |
1037 | if (nd->path.dentry == nd->root.dentry && |
1038 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1039 | break; |
1040 | } | |
4ac91378 | 1041 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1042 | /* rare case of legitimate dget_parent()... */ |
1043 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1044 | dput(old); |
1045 | break; | |
1046 | } | |
3088dd70 | 1047 | if (!follow_up(&nd->path)) |
1da177e4 | 1048 | break; |
1da177e4 | 1049 | } |
79ed0226 | 1050 | follow_mount(&nd->path); |
31e6b01f | 1051 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1052 | } |
1053 | ||
baa03890 NP |
1054 | /* |
1055 | * Allocate a dentry with name and parent, and perform a parent | |
1056 | * directory ->lookup on it. Returns the new dentry, or ERR_PTR | |
1057 | * on error. parent->d_inode->i_mutex must be held. d_lookup must | |
1058 | * have verified that no child exists while under i_mutex. | |
1059 | */ | |
1060 | static struct dentry *d_alloc_and_lookup(struct dentry *parent, | |
1061 | struct qstr *name, struct nameidata *nd) | |
1062 | { | |
1063 | struct inode *inode = parent->d_inode; | |
1064 | struct dentry *dentry; | |
1065 | struct dentry *old; | |
1066 | ||
1067 | /* Don't create child dentry for a dead directory. */ | |
1068 | if (unlikely(IS_DEADDIR(inode))) | |
1069 | return ERR_PTR(-ENOENT); | |
1070 | ||
1071 | dentry = d_alloc(parent, name); | |
1072 | if (unlikely(!dentry)) | |
1073 | return ERR_PTR(-ENOMEM); | |
1074 | ||
1075 | old = inode->i_op->lookup(inode, dentry, nd); | |
1076 | if (unlikely(old)) { | |
1077 | dput(dentry); | |
1078 | dentry = old; | |
1079 | } | |
1080 | return dentry; | |
1081 | } | |
1082 | ||
44396f4b JB |
1083 | /* |
1084 | * We already have a dentry, but require a lookup to be performed on the parent | |
1085 | * directory to fill in d_inode. Returns the new dentry, or ERR_PTR on error. | |
1086 | * parent->d_inode->i_mutex must be held. d_lookup must have verified that no | |
1087 | * child exists while under i_mutex. | |
1088 | */ | |
1089 | static struct dentry *d_inode_lookup(struct dentry *parent, struct dentry *dentry, | |
1090 | struct nameidata *nd) | |
1091 | { | |
1092 | struct inode *inode = parent->d_inode; | |
1093 | struct dentry *old; | |
1094 | ||
1095 | /* Don't create child dentry for a dead directory. */ | |
1096 | if (unlikely(IS_DEADDIR(inode))) | |
1097 | return ERR_PTR(-ENOENT); | |
1098 | ||
1099 | old = inode->i_op->lookup(inode, dentry, nd); | |
1100 | if (unlikely(old)) { | |
1101 | dput(dentry); | |
1102 | dentry = old; | |
1103 | } | |
1104 | return dentry; | |
1105 | } | |
1106 | ||
1da177e4 LT |
1107 | /* |
1108 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1109 | * small and for now I'd prefer to have fast path as straight as possible. | |
1110 | * It _is_ time-critical. | |
1111 | */ | |
1112 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
31e6b01f | 1113 | struct path *path, struct inode **inode) |
1da177e4 | 1114 | { |
4ac91378 | 1115 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1116 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1117 | int need_reval = 1; |
1118 | int status = 1; | |
9875cf80 DH |
1119 | int err; |
1120 | ||
b04f784e NP |
1121 | /* |
1122 | * Rename seqlock is not required here because in the off chance | |
1123 | * of a false negative due to a concurrent rename, we're going to | |
1124 | * do the non-racy lookup, below. | |
1125 | */ | |
31e6b01f NP |
1126 | if (nd->flags & LOOKUP_RCU) { |
1127 | unsigned seq; | |
31e6b01f NP |
1128 | *inode = nd->inode; |
1129 | dentry = __d_lookup_rcu(parent, name, &seq, inode); | |
5a18fff2 AV |
1130 | if (!dentry) |
1131 | goto unlazy; | |
1132 | ||
31e6b01f NP |
1133 | /* Memory barrier in read_seqcount_begin of child is enough */ |
1134 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) | |
1135 | return -ECHILD; | |
31e6b01f | 1136 | nd->seq = seq; |
5a18fff2 | 1137 | |
24643087 | 1138 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
5a18fff2 AV |
1139 | status = d_revalidate(dentry, nd); |
1140 | if (unlikely(status <= 0)) { | |
1141 | if (status != -ECHILD) | |
1142 | need_reval = 0; | |
1143 | goto unlazy; | |
1144 | } | |
24643087 | 1145 | } |
44396f4b JB |
1146 | if (unlikely(d_need_lookup(dentry))) |
1147 | goto unlazy; | |
31e6b01f NP |
1148 | path->mnt = mnt; |
1149 | path->dentry = dentry; | |
d6e9bd25 AV |
1150 | if (unlikely(!__follow_mount_rcu(nd, path, inode))) |
1151 | goto unlazy; | |
1152 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1153 | goto unlazy; | |
1154 | return 0; | |
5a18fff2 | 1155 | unlazy: |
19660af7 AV |
1156 | if (unlazy_walk(nd, dentry)) |
1157 | return -ECHILD; | |
5a18fff2 AV |
1158 | } else { |
1159 | dentry = __d_lookup(parent, name); | |
9875cf80 | 1160 | } |
5a18fff2 | 1161 | |
44396f4b JB |
1162 | if (dentry && unlikely(d_need_lookup(dentry))) { |
1163 | dput(dentry); | |
1164 | dentry = NULL; | |
1165 | } | |
5a18fff2 AV |
1166 | retry: |
1167 | if (unlikely(!dentry)) { | |
1168 | struct inode *dir = parent->d_inode; | |
1169 | BUG_ON(nd->inode != dir); | |
1170 | ||
1171 | mutex_lock(&dir->i_mutex); | |
1172 | dentry = d_lookup(parent, name); | |
1173 | if (likely(!dentry)) { | |
1174 | dentry = d_alloc_and_lookup(parent, name, nd); | |
1175 | if (IS_ERR(dentry)) { | |
1176 | mutex_unlock(&dir->i_mutex); | |
1177 | return PTR_ERR(dentry); | |
1178 | } | |
1179 | /* known good */ | |
1180 | need_reval = 0; | |
1181 | status = 1; | |
44396f4b JB |
1182 | } else if (unlikely(d_need_lookup(dentry))) { |
1183 | dentry = d_inode_lookup(parent, dentry, nd); | |
1184 | if (IS_ERR(dentry)) { | |
1185 | mutex_unlock(&dir->i_mutex); | |
1186 | return PTR_ERR(dentry); | |
1187 | } | |
1188 | /* known good */ | |
1189 | need_reval = 0; | |
1190 | status = 1; | |
5a18fff2 AV |
1191 | } |
1192 | mutex_unlock(&dir->i_mutex); | |
1193 | } | |
1194 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) | |
1195 | status = d_revalidate(dentry, nd); | |
1196 | if (unlikely(status <= 0)) { | |
1197 | if (status < 0) { | |
1198 | dput(dentry); | |
1199 | return status; | |
1200 | } | |
1201 | if (!d_invalidate(dentry)) { | |
1202 | dput(dentry); | |
1203 | dentry = NULL; | |
1204 | need_reval = 1; | |
1205 | goto retry; | |
1206 | } | |
24643087 | 1207 | } |
5a18fff2 | 1208 | |
9875cf80 DH |
1209 | path->mnt = mnt; |
1210 | path->dentry = dentry; | |
1211 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1212 | if (unlikely(err < 0)) { |
1213 | path_put_conditional(path, nd); | |
9875cf80 | 1214 | return err; |
89312214 | 1215 | } |
9875cf80 | 1216 | *inode = path->dentry->d_inode; |
1da177e4 | 1217 | return 0; |
1da177e4 LT |
1218 | } |
1219 | ||
52094c8a AV |
1220 | static inline int may_lookup(struct nameidata *nd) |
1221 | { | |
1222 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1223 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1224 | if (err != -ECHILD) |
1225 | return err; | |
19660af7 | 1226 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1227 | return -ECHILD; |
1228 | } | |
4ad5abb3 | 1229 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1230 | } |
1231 | ||
9856fa1b AV |
1232 | static inline int handle_dots(struct nameidata *nd, int type) |
1233 | { | |
1234 | if (type == LAST_DOTDOT) { | |
1235 | if (nd->flags & LOOKUP_RCU) { | |
1236 | if (follow_dotdot_rcu(nd)) | |
1237 | return -ECHILD; | |
1238 | } else | |
1239 | follow_dotdot(nd); | |
1240 | } | |
1241 | return 0; | |
1242 | } | |
1243 | ||
951361f9 AV |
1244 | static void terminate_walk(struct nameidata *nd) |
1245 | { | |
1246 | if (!(nd->flags & LOOKUP_RCU)) { | |
1247 | path_put(&nd->path); | |
1248 | } else { | |
1249 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1250 | if (!(nd->flags & LOOKUP_ROOT)) |
1251 | nd->root.mnt = NULL; | |
951361f9 AV |
1252 | rcu_read_unlock(); |
1253 | br_read_unlock(vfsmount_lock); | |
1254 | } | |
1255 | } | |
1256 | ||
3ddcd056 LT |
1257 | /* |
1258 | * Do we need to follow links? We _really_ want to be able | |
1259 | * to do this check without having to look at inode->i_op, | |
1260 | * so we keep a cache of "no, this doesn't need follow_link" | |
1261 | * for the common case. | |
1262 | */ | |
7813b94a | 1263 | static inline int should_follow_link(struct inode *inode, int follow) |
3ddcd056 LT |
1264 | { |
1265 | if (unlikely(!(inode->i_opflags & IOP_NOFOLLOW))) { | |
1266 | if (likely(inode->i_op->follow_link)) | |
1267 | return follow; | |
1268 | ||
1269 | /* This gets set once for the inode lifetime */ | |
1270 | spin_lock(&inode->i_lock); | |
1271 | inode->i_opflags |= IOP_NOFOLLOW; | |
1272 | spin_unlock(&inode->i_lock); | |
1273 | } | |
1274 | return 0; | |
1275 | } | |
1276 | ||
ce57dfc1 AV |
1277 | static inline int walk_component(struct nameidata *nd, struct path *path, |
1278 | struct qstr *name, int type, int follow) | |
1279 | { | |
1280 | struct inode *inode; | |
1281 | int err; | |
1282 | /* | |
1283 | * "." and ".." are special - ".." especially so because it has | |
1284 | * to be able to know about the current root directory and | |
1285 | * parent relationships. | |
1286 | */ | |
1287 | if (unlikely(type != LAST_NORM)) | |
1288 | return handle_dots(nd, type); | |
1289 | err = do_lookup(nd, name, path, &inode); | |
1290 | if (unlikely(err)) { | |
1291 | terminate_walk(nd); | |
1292 | return err; | |
1293 | } | |
1294 | if (!inode) { | |
1295 | path_to_nameidata(path, nd); | |
1296 | terminate_walk(nd); | |
1297 | return -ENOENT; | |
1298 | } | |
7813b94a | 1299 | if (should_follow_link(inode, follow)) { |
19660af7 AV |
1300 | if (nd->flags & LOOKUP_RCU) { |
1301 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
1302 | terminate_walk(nd); | |
1303 | return -ECHILD; | |
1304 | } | |
1305 | } | |
ce57dfc1 AV |
1306 | BUG_ON(inode != path->dentry->d_inode); |
1307 | return 1; | |
1308 | } | |
1309 | path_to_nameidata(path, nd); | |
1310 | nd->inode = inode; | |
1311 | return 0; | |
1312 | } | |
1313 | ||
b356379a AV |
1314 | /* |
1315 | * This limits recursive symlink follows to 8, while | |
1316 | * limiting consecutive symlinks to 40. | |
1317 | * | |
1318 | * Without that kind of total limit, nasty chains of consecutive | |
1319 | * symlinks can cause almost arbitrarily long lookups. | |
1320 | */ | |
1321 | static inline int nested_symlink(struct path *path, struct nameidata *nd) | |
1322 | { | |
1323 | int res; | |
1324 | ||
b356379a AV |
1325 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1326 | path_put_conditional(path, nd); | |
1327 | path_put(&nd->path); | |
1328 | return -ELOOP; | |
1329 | } | |
1a4022f8 | 1330 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); |
b356379a AV |
1331 | |
1332 | nd->depth++; | |
1333 | current->link_count++; | |
1334 | ||
1335 | do { | |
1336 | struct path link = *path; | |
1337 | void *cookie; | |
574197e0 AV |
1338 | |
1339 | res = follow_link(&link, nd, &cookie); | |
b356379a AV |
1340 | if (!res) |
1341 | res = walk_component(nd, path, &nd->last, | |
1342 | nd->last_type, LOOKUP_FOLLOW); | |
574197e0 | 1343 | put_link(nd, &link, cookie); |
b356379a AV |
1344 | } while (res > 0); |
1345 | ||
1346 | current->link_count--; | |
1347 | nd->depth--; | |
1348 | return res; | |
1349 | } | |
1350 | ||
3ddcd056 LT |
1351 | /* |
1352 | * We really don't want to look at inode->i_op->lookup | |
1353 | * when we don't have to. So we keep a cache bit in | |
1354 | * the inode ->i_opflags field that says "yes, we can | |
1355 | * do lookup on this inode". | |
1356 | */ | |
1357 | static inline int can_lookup(struct inode *inode) | |
1358 | { | |
1359 | if (likely(inode->i_opflags & IOP_LOOKUP)) | |
1360 | return 1; | |
1361 | if (likely(!inode->i_op->lookup)) | |
1362 | return 0; | |
1363 | ||
1364 | /* We do this once for the lifetime of the inode */ | |
1365 | spin_lock(&inode->i_lock); | |
1366 | inode->i_opflags |= IOP_LOOKUP; | |
1367 | spin_unlock(&inode->i_lock); | |
1368 | return 1; | |
1369 | } | |
1370 | ||
1da177e4 LT |
1371 | /* |
1372 | * Name resolution. | |
ea3834d9 PM |
1373 | * This is the basic name resolution function, turning a pathname into |
1374 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1375 | * |
ea3834d9 PM |
1376 | * Returns 0 and nd will have valid dentry and mnt on success. |
1377 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1378 | */ |
6de88d72 | 1379 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1380 | { |
1381 | struct path next; | |
1da177e4 | 1382 | int err; |
1da177e4 LT |
1383 | |
1384 | while (*name=='/') | |
1385 | name++; | |
1386 | if (!*name) | |
086e183a | 1387 | return 0; |
1da177e4 | 1388 | |
1da177e4 LT |
1389 | /* At this point we know we have a real path component. */ |
1390 | for(;;) { | |
1391 | unsigned long hash; | |
1392 | struct qstr this; | |
1393 | unsigned int c; | |
fe479a58 | 1394 | int type; |
1da177e4 | 1395 | |
52094c8a | 1396 | err = may_lookup(nd); |
1da177e4 LT |
1397 | if (err) |
1398 | break; | |
1399 | ||
1400 | this.name = name; | |
1401 | c = *(const unsigned char *)name; | |
1402 | ||
1403 | hash = init_name_hash(); | |
1404 | do { | |
1405 | name++; | |
1406 | hash = partial_name_hash(c, hash); | |
1407 | c = *(const unsigned char *)name; | |
1408 | } while (c && (c != '/')); | |
1409 | this.len = name - (const char *) this.name; | |
1410 | this.hash = end_name_hash(hash); | |
1411 | ||
fe479a58 AV |
1412 | type = LAST_NORM; |
1413 | if (this.name[0] == '.') switch (this.len) { | |
1414 | case 2: | |
16c2cd71 | 1415 | if (this.name[1] == '.') { |
fe479a58 | 1416 | type = LAST_DOTDOT; |
16c2cd71 AV |
1417 | nd->flags |= LOOKUP_JUMPED; |
1418 | } | |
fe479a58 AV |
1419 | break; |
1420 | case 1: | |
1421 | type = LAST_DOT; | |
1422 | } | |
5a202bcd AV |
1423 | if (likely(type == LAST_NORM)) { |
1424 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1425 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd AV |
1426 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
1427 | err = parent->d_op->d_hash(parent, nd->inode, | |
1428 | &this); | |
1429 | if (err < 0) | |
1430 | break; | |
1431 | } | |
1432 | } | |
fe479a58 | 1433 | |
1da177e4 LT |
1434 | /* remove trailing slashes? */ |
1435 | if (!c) | |
1436 | goto last_component; | |
1437 | while (*++name == '/'); | |
1438 | if (!*name) | |
b356379a | 1439 | goto last_component; |
1da177e4 | 1440 | |
ce57dfc1 AV |
1441 | err = walk_component(nd, &next, &this, type, LOOKUP_FOLLOW); |
1442 | if (err < 0) | |
1443 | return err; | |
1da177e4 | 1444 | |
ce57dfc1 | 1445 | if (err) { |
b356379a | 1446 | err = nested_symlink(&next, nd); |
1da177e4 | 1447 | if (err) |
a7472bab | 1448 | return err; |
31e6b01f | 1449 | } |
3ddcd056 LT |
1450 | if (can_lookup(nd->inode)) |
1451 | continue; | |
1da177e4 | 1452 | err = -ENOTDIR; |
3ddcd056 | 1453 | break; |
1da177e4 LT |
1454 | /* here ends the main loop */ |
1455 | ||
1da177e4 | 1456 | last_component: |
b356379a AV |
1457 | nd->last = this; |
1458 | nd->last_type = type; | |
086e183a | 1459 | return 0; |
1da177e4 | 1460 | } |
951361f9 | 1461 | terminate_walk(nd); |
1da177e4 LT |
1462 | return err; |
1463 | } | |
1464 | ||
70e9b357 AV |
1465 | static int path_init(int dfd, const char *name, unsigned int flags, |
1466 | struct nameidata *nd, struct file **fp) | |
31e6b01f NP |
1467 | { |
1468 | int retval = 0; | |
1469 | int fput_needed; | |
1470 | struct file *file; | |
1471 | ||
1472 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
16c2cd71 | 1473 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 1474 | nd->depth = 0; |
5b6ca027 AV |
1475 | if (flags & LOOKUP_ROOT) { |
1476 | struct inode *inode = nd->root.dentry->d_inode; | |
73d049a4 AV |
1477 | if (*name) { |
1478 | if (!inode->i_op->lookup) | |
1479 | return -ENOTDIR; | |
1480 | retval = inode_permission(inode, MAY_EXEC); | |
1481 | if (retval) | |
1482 | return retval; | |
1483 | } | |
5b6ca027 AV |
1484 | nd->path = nd->root; |
1485 | nd->inode = inode; | |
1486 | if (flags & LOOKUP_RCU) { | |
1487 | br_read_lock(vfsmount_lock); | |
1488 | rcu_read_lock(); | |
1489 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1490 | } else { | |
1491 | path_get(&nd->path); | |
1492 | } | |
1493 | return 0; | |
1494 | } | |
1495 | ||
31e6b01f | 1496 | nd->root.mnt = NULL; |
31e6b01f NP |
1497 | |
1498 | if (*name=='/') { | |
e41f7d4e AV |
1499 | if (flags & LOOKUP_RCU) { |
1500 | br_read_lock(vfsmount_lock); | |
1501 | rcu_read_lock(); | |
1502 | set_root_rcu(nd); | |
1503 | } else { | |
1504 | set_root(nd); | |
1505 | path_get(&nd->root); | |
1506 | } | |
1507 | nd->path = nd->root; | |
31e6b01f | 1508 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1509 | if (flags & LOOKUP_RCU) { |
1510 | struct fs_struct *fs = current->fs; | |
1511 | unsigned seq; | |
31e6b01f | 1512 | |
e41f7d4e AV |
1513 | br_read_lock(vfsmount_lock); |
1514 | rcu_read_lock(); | |
c28cc364 | 1515 | |
e41f7d4e AV |
1516 | do { |
1517 | seq = read_seqcount_begin(&fs->seq); | |
1518 | nd->path = fs->pwd; | |
1519 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1520 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1521 | } else { | |
1522 | get_fs_pwd(current->fs, &nd->path); | |
1523 | } | |
31e6b01f NP |
1524 | } else { |
1525 | struct dentry *dentry; | |
1526 | ||
1abf0c71 | 1527 | file = fget_raw_light(dfd, &fput_needed); |
31e6b01f NP |
1528 | retval = -EBADF; |
1529 | if (!file) | |
1530 | goto out_fail; | |
1531 | ||
1532 | dentry = file->f_path.dentry; | |
1533 | ||
f52e0c11 AV |
1534 | if (*name) { |
1535 | retval = -ENOTDIR; | |
1536 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
1537 | goto fput_fail; | |
31e6b01f | 1538 | |
4ad5abb3 | 1539 | retval = inode_permission(dentry->d_inode, MAY_EXEC); |
f52e0c11 AV |
1540 | if (retval) |
1541 | goto fput_fail; | |
1542 | } | |
31e6b01f NP |
1543 | |
1544 | nd->path = file->f_path; | |
e41f7d4e AV |
1545 | if (flags & LOOKUP_RCU) { |
1546 | if (fput_needed) | |
70e9b357 | 1547 | *fp = file; |
e41f7d4e AV |
1548 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
1549 | br_read_lock(vfsmount_lock); | |
1550 | rcu_read_lock(); | |
1551 | } else { | |
1552 | path_get(&file->f_path); | |
1553 | fput_light(file, fput_needed); | |
1554 | } | |
31e6b01f | 1555 | } |
31e6b01f | 1556 | |
31e6b01f | 1557 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1558 | return 0; |
2dfdd266 | 1559 | |
9b4a9b14 AV |
1560 | fput_fail: |
1561 | fput_light(file, fput_needed); | |
1562 | out_fail: | |
1563 | return retval; | |
1564 | } | |
1565 | ||
bd92d7fe AV |
1566 | static inline int lookup_last(struct nameidata *nd, struct path *path) |
1567 | { | |
1568 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1569 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1570 | ||
1571 | nd->flags &= ~LOOKUP_PARENT; | |
1572 | return walk_component(nd, path, &nd->last, nd->last_type, | |
1573 | nd->flags & LOOKUP_FOLLOW); | |
1574 | } | |
1575 | ||
9b4a9b14 | 1576 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
ee0827cd | 1577 | static int path_lookupat(int dfd, const char *name, |
9b4a9b14 AV |
1578 | unsigned int flags, struct nameidata *nd) |
1579 | { | |
70e9b357 | 1580 | struct file *base = NULL; |
bd92d7fe AV |
1581 | struct path path; |
1582 | int err; | |
31e6b01f NP |
1583 | |
1584 | /* | |
1585 | * Path walking is largely split up into 2 different synchronisation | |
1586 | * schemes, rcu-walk and ref-walk (explained in | |
1587 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1588 | * path walk code, but some things particularly setup, cleanup, and | |
1589 | * following mounts are sufficiently divergent that functions are | |
1590 | * duplicated. Typically there is a function foo(), and its RCU | |
1591 | * analogue, foo_rcu(). | |
1592 | * | |
1593 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1594 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1595 | * be handled by restarting a traditional ref-walk (which will always | |
1596 | * be able to complete). | |
1597 | */ | |
bd92d7fe | 1598 | err = path_init(dfd, name, flags | LOOKUP_PARENT, nd, &base); |
ee0827cd | 1599 | |
bd92d7fe AV |
1600 | if (unlikely(err)) |
1601 | return err; | |
ee0827cd AV |
1602 | |
1603 | current->total_link_count = 0; | |
bd92d7fe AV |
1604 | err = link_path_walk(name, nd); |
1605 | ||
1606 | if (!err && !(flags & LOOKUP_PARENT)) { | |
bd92d7fe AV |
1607 | err = lookup_last(nd, &path); |
1608 | while (err > 0) { | |
1609 | void *cookie; | |
1610 | struct path link = path; | |
bd92d7fe | 1611 | nd->flags |= LOOKUP_PARENT; |
574197e0 | 1612 | err = follow_link(&link, nd, &cookie); |
bd92d7fe AV |
1613 | if (!err) |
1614 | err = lookup_last(nd, &path); | |
574197e0 | 1615 | put_link(nd, &link, cookie); |
bd92d7fe AV |
1616 | } |
1617 | } | |
ee0827cd | 1618 | |
9f1fafee AV |
1619 | if (!err) |
1620 | err = complete_walk(nd); | |
bd92d7fe AV |
1621 | |
1622 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
1623 | if (!nd->inode->i_op->lookup) { | |
1624 | path_put(&nd->path); | |
bd23a539 | 1625 | err = -ENOTDIR; |
bd92d7fe AV |
1626 | } |
1627 | } | |
16c2cd71 | 1628 | |
70e9b357 AV |
1629 | if (base) |
1630 | fput(base); | |
ee0827cd | 1631 | |
5b6ca027 | 1632 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
2a737871 AV |
1633 | path_put(&nd->root); |
1634 | nd->root.mnt = NULL; | |
1635 | } | |
bd92d7fe | 1636 | return err; |
ee0827cd | 1637 | } |
31e6b01f | 1638 | |
ee0827cd AV |
1639 | static int do_path_lookup(int dfd, const char *name, |
1640 | unsigned int flags, struct nameidata *nd) | |
1641 | { | |
1642 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); | |
1643 | if (unlikely(retval == -ECHILD)) | |
1644 | retval = path_lookupat(dfd, name, flags, nd); | |
1645 | if (unlikely(retval == -ESTALE)) | |
1646 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); | |
31e6b01f NP |
1647 | |
1648 | if (likely(!retval)) { | |
1649 | if (unlikely(!audit_dummy_context())) { | |
1650 | if (nd->path.dentry && nd->inode) | |
1651 | audit_inode(name, nd->path.dentry); | |
1652 | } | |
1653 | } | |
170aa3d0 | 1654 | return retval; |
1da177e4 LT |
1655 | } |
1656 | ||
c9c6cac0 | 1657 | int kern_path_parent(const char *name, struct nameidata *nd) |
5590ff0d | 1658 | { |
c9c6cac0 | 1659 | return do_path_lookup(AT_FDCWD, name, LOOKUP_PARENT, nd); |
5590ff0d UD |
1660 | } |
1661 | ||
d1811465 AV |
1662 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1663 | { | |
1664 | struct nameidata nd; | |
1665 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1666 | if (!res) | |
1667 | *path = nd.path; | |
1668 | return res; | |
1669 | } | |
1670 | ||
16f18200 JJS |
1671 | /** |
1672 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1673 | * @dentry: pointer to dentry of the base directory | |
1674 | * @mnt: pointer to vfs mount of the base directory | |
1675 | * @name: pointer to file name | |
1676 | * @flags: lookup flags | |
e0a01249 | 1677 | * @path: pointer to struct path to fill |
16f18200 JJS |
1678 | */ |
1679 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1680 | const char *name, unsigned int flags, | |
e0a01249 | 1681 | struct path *path) |
16f18200 | 1682 | { |
e0a01249 AV |
1683 | struct nameidata nd; |
1684 | int err; | |
1685 | nd.root.dentry = dentry; | |
1686 | nd.root.mnt = mnt; | |
1687 | BUG_ON(flags & LOOKUP_PARENT); | |
5b6ca027 | 1688 | /* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */ |
e0a01249 AV |
1689 | err = do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, &nd); |
1690 | if (!err) | |
1691 | *path = nd.path; | |
1692 | return err; | |
16f18200 JJS |
1693 | } |
1694 | ||
eead1911 CH |
1695 | static struct dentry *__lookup_hash(struct qstr *name, |
1696 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1697 | { |
81fca444 | 1698 | struct inode *inode = base->d_inode; |
057f6c01 | 1699 | struct dentry *dentry; |
1da177e4 LT |
1700 | int err; |
1701 | ||
4ad5abb3 | 1702 | err = inode_permission(inode, MAY_EXEC); |
81fca444 CH |
1703 | if (err) |
1704 | return ERR_PTR(err); | |
1da177e4 | 1705 | |
b04f784e NP |
1706 | /* |
1707 | * Don't bother with __d_lookup: callers are for creat as | |
1708 | * well as unlink, so a lot of the time it would cost | |
1709 | * a double lookup. | |
6e6b1bd1 | 1710 | */ |
b04f784e | 1711 | dentry = d_lookup(base, name); |
6e6b1bd1 | 1712 | |
44396f4b JB |
1713 | if (dentry && d_need_lookup(dentry)) { |
1714 | /* | |
1715 | * __lookup_hash is called with the parent dir's i_mutex already | |
1716 | * held, so we are good to go here. | |
1717 | */ | |
1718 | dentry = d_inode_lookup(base, dentry, nd); | |
1719 | if (IS_ERR(dentry)) | |
1720 | return dentry; | |
1721 | } | |
1722 | ||
d2d9e9fb AV |
1723 | if (dentry && (dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
1724 | int status = d_revalidate(dentry, nd); | |
1725 | if (unlikely(status <= 0)) { | |
1726 | /* | |
1727 | * The dentry failed validation. | |
1728 | * If d_revalidate returned 0 attempt to invalidate | |
1729 | * the dentry otherwise d_revalidate is asking us | |
1730 | * to return a fail status. | |
1731 | */ | |
1732 | if (status < 0) { | |
1733 | dput(dentry); | |
1734 | return ERR_PTR(status); | |
1735 | } else if (!d_invalidate(dentry)) { | |
1736 | dput(dentry); | |
1737 | dentry = NULL; | |
1738 | } | |
1739 | } | |
1740 | } | |
6e6b1bd1 | 1741 | |
baa03890 NP |
1742 | if (!dentry) |
1743 | dentry = d_alloc_and_lookup(base, name, nd); | |
5a202bcd | 1744 | |
1da177e4 LT |
1745 | return dentry; |
1746 | } | |
1747 | ||
057f6c01 JM |
1748 | /* |
1749 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1750 | * needs parent already locked. Doesn't follow mounts. | |
1751 | * SMP-safe. | |
1752 | */ | |
eead1911 | 1753 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1754 | { |
4ac91378 | 1755 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1756 | } |
1757 | ||
eead1911 | 1758 | /** |
a6b91919 | 1759 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1760 | * @name: pathname component to lookup |
1761 | * @base: base directory to lookup from | |
1762 | * @len: maximum length @len should be interpreted to | |
1763 | * | |
a6b91919 RD |
1764 | * Note that this routine is purely a helper for filesystem usage and should |
1765 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1766 | * nameidata argument is passed to the filesystem methods and a filesystem |
1767 | * using this helper needs to be prepared for that. | |
1768 | */ | |
057f6c01 JM |
1769 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1770 | { | |
057f6c01 | 1771 | struct qstr this; |
6a96ba54 AV |
1772 | unsigned long hash; |
1773 | unsigned int c; | |
057f6c01 | 1774 | |
2f9092e1 DW |
1775 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1776 | ||
6a96ba54 AV |
1777 | this.name = name; |
1778 | this.len = len; | |
1779 | if (!len) | |
1780 | return ERR_PTR(-EACCES); | |
1781 | ||
1782 | hash = init_name_hash(); | |
1783 | while (len--) { | |
1784 | c = *(const unsigned char *)name++; | |
1785 | if (c == '/' || c == '\0') | |
1786 | return ERR_PTR(-EACCES); | |
1787 | hash = partial_name_hash(c, hash); | |
1788 | } | |
1789 | this.hash = end_name_hash(hash); | |
5a202bcd AV |
1790 | /* |
1791 | * See if the low-level filesystem might want | |
1792 | * to use its own hash.. | |
1793 | */ | |
1794 | if (base->d_flags & DCACHE_OP_HASH) { | |
1795 | int err = base->d_op->d_hash(base, base->d_inode, &this); | |
1796 | if (err < 0) | |
1797 | return ERR_PTR(err); | |
1798 | } | |
eead1911 | 1799 | |
49705b77 | 1800 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1801 | } |
1802 | ||
1fa1e7f6 AW |
1803 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
1804 | struct path *path, int *empty) | |
1da177e4 | 1805 | { |
2d8f3038 | 1806 | struct nameidata nd; |
1fa1e7f6 | 1807 | char *tmp = getname_flags(name, flags, empty); |
1da177e4 | 1808 | int err = PTR_ERR(tmp); |
1da177e4 | 1809 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1810 | |
1811 | BUG_ON(flags & LOOKUP_PARENT); | |
1812 | ||
1813 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1814 | putname(tmp); |
2d8f3038 AV |
1815 | if (!err) |
1816 | *path = nd.path; | |
1da177e4 LT |
1817 | } |
1818 | return err; | |
1819 | } | |
1820 | ||
1fa1e7f6 AW |
1821 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1822 | struct path *path) | |
1823 | { | |
1824 | return user_path_at_empty(dfd, name, flags, path, 0); | |
1825 | } | |
1826 | ||
2ad94ae6 AV |
1827 | static int user_path_parent(int dfd, const char __user *path, |
1828 | struct nameidata *nd, char **name) | |
1829 | { | |
1830 | char *s = getname(path); | |
1831 | int error; | |
1832 | ||
1833 | if (IS_ERR(s)) | |
1834 | return PTR_ERR(s); | |
1835 | ||
1836 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1837 | if (error) | |
1838 | putname(s); | |
1839 | else | |
1840 | *name = s; | |
1841 | ||
1842 | return error; | |
1843 | } | |
1844 | ||
1da177e4 LT |
1845 | /* |
1846 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1847 | * minimal. | |
1848 | */ | |
1849 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1850 | { | |
da9592ed DH |
1851 | uid_t fsuid = current_fsuid(); |
1852 | ||
1da177e4 LT |
1853 | if (!(dir->i_mode & S_ISVTX)) |
1854 | return 0; | |
e795b717 SH |
1855 | if (current_user_ns() != inode_userns(inode)) |
1856 | goto other_userns; | |
da9592ed | 1857 | if (inode->i_uid == fsuid) |
1da177e4 | 1858 | return 0; |
da9592ed | 1859 | if (dir->i_uid == fsuid) |
1da177e4 | 1860 | return 0; |
e795b717 SH |
1861 | |
1862 | other_userns: | |
1863 | return !ns_capable(inode_userns(inode), CAP_FOWNER); | |
1da177e4 LT |
1864 | } |
1865 | ||
1866 | /* | |
1867 | * Check whether we can remove a link victim from directory dir, check | |
1868 | * whether the type of victim is right. | |
1869 | * 1. We can't do it if dir is read-only (done in permission()) | |
1870 | * 2. We should have write and exec permissions on dir | |
1871 | * 3. We can't remove anything from append-only dir | |
1872 | * 4. We can't do anything with immutable dir (done in permission()) | |
1873 | * 5. If the sticky bit on dir is set we should either | |
1874 | * a. be owner of dir, or | |
1875 | * b. be owner of victim, or | |
1876 | * c. have CAP_FOWNER capability | |
1877 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1878 | * links pointing to it. | |
1879 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1880 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1881 | * 9. We can't remove a root or mountpoint. | |
1882 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1883 | * nfs_async_unlink(). | |
1884 | */ | |
858119e1 | 1885 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1886 | { |
1887 | int error; | |
1888 | ||
1889 | if (!victim->d_inode) | |
1890 | return -ENOENT; | |
1891 | ||
1892 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 1893 | audit_inode_child(victim, dir); |
1da177e4 | 1894 | |
f419a2e3 | 1895 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1896 | if (error) |
1897 | return error; | |
1898 | if (IS_APPEND(dir)) | |
1899 | return -EPERM; | |
1900 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1901 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1902 | return -EPERM; |
1903 | if (isdir) { | |
1904 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1905 | return -ENOTDIR; | |
1906 | if (IS_ROOT(victim)) | |
1907 | return -EBUSY; | |
1908 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1909 | return -EISDIR; | |
1910 | if (IS_DEADDIR(dir)) | |
1911 | return -ENOENT; | |
1912 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1913 | return -EBUSY; | |
1914 | return 0; | |
1915 | } | |
1916 | ||
1917 | /* Check whether we can create an object with dentry child in directory | |
1918 | * dir. | |
1919 | * 1. We can't do it if child already exists (open has special treatment for | |
1920 | * this case, but since we are inlined it's OK) | |
1921 | * 2. We can't do it if dir is read-only (done in permission()) | |
1922 | * 3. We should have write and exec permissions on dir | |
1923 | * 4. We can't do it if dir is immutable (done in permission()) | |
1924 | */ | |
a95164d9 | 1925 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1926 | { |
1927 | if (child->d_inode) | |
1928 | return -EEXIST; | |
1929 | if (IS_DEADDIR(dir)) | |
1930 | return -ENOENT; | |
f419a2e3 | 1931 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1932 | } |
1933 | ||
1da177e4 LT |
1934 | /* |
1935 | * p1 and p2 should be directories on the same fs. | |
1936 | */ | |
1937 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1938 | { | |
1939 | struct dentry *p; | |
1940 | ||
1941 | if (p1 == p2) { | |
f2eace23 | 1942 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1943 | return NULL; |
1944 | } | |
1945 | ||
a11f3a05 | 1946 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1947 | |
e2761a11 OH |
1948 | p = d_ancestor(p2, p1); |
1949 | if (p) { | |
1950 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1951 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1952 | return p; | |
1da177e4 LT |
1953 | } |
1954 | ||
e2761a11 OH |
1955 | p = d_ancestor(p1, p2); |
1956 | if (p) { | |
1957 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1958 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1959 | return p; | |
1da177e4 LT |
1960 | } |
1961 | ||
f2eace23 IM |
1962 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1963 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1964 | return NULL; |
1965 | } | |
1966 | ||
1967 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1968 | { | |
1b1dcc1b | 1969 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1970 | if (p1 != p2) { |
1b1dcc1b | 1971 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1972 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1973 | } |
1974 | } | |
1975 | ||
1976 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1977 | struct nameidata *nd) | |
1978 | { | |
a95164d9 | 1979 | int error = may_create(dir, dentry); |
1da177e4 LT |
1980 | |
1981 | if (error) | |
1982 | return error; | |
1983 | ||
acfa4380 | 1984 | if (!dir->i_op->create) |
1da177e4 LT |
1985 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1986 | mode &= S_IALLUGO; | |
1987 | mode |= S_IFREG; | |
1988 | error = security_inode_create(dir, dentry, mode); | |
1989 | if (error) | |
1990 | return error; | |
1da177e4 | 1991 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1992 | if (!error) |
f38aa942 | 1993 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1994 | return error; |
1995 | } | |
1996 | ||
73d049a4 | 1997 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1998 | { |
3fb64190 | 1999 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2000 | struct inode *inode = dentry->d_inode; |
2001 | int error; | |
2002 | ||
bcda7652 AV |
2003 | /* O_PATH? */ |
2004 | if (!acc_mode) | |
2005 | return 0; | |
2006 | ||
1da177e4 LT |
2007 | if (!inode) |
2008 | return -ENOENT; | |
2009 | ||
c8fe8f30 CH |
2010 | switch (inode->i_mode & S_IFMT) { |
2011 | case S_IFLNK: | |
1da177e4 | 2012 | return -ELOOP; |
c8fe8f30 CH |
2013 | case S_IFDIR: |
2014 | if (acc_mode & MAY_WRITE) | |
2015 | return -EISDIR; | |
2016 | break; | |
2017 | case S_IFBLK: | |
2018 | case S_IFCHR: | |
3fb64190 | 2019 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2020 | return -EACCES; |
c8fe8f30 CH |
2021 | /*FALLTHRU*/ |
2022 | case S_IFIFO: | |
2023 | case S_IFSOCK: | |
1da177e4 | 2024 | flag &= ~O_TRUNC; |
c8fe8f30 | 2025 | break; |
4a3fd211 | 2026 | } |
b41572e9 | 2027 | |
3fb64190 | 2028 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2029 | if (error) |
2030 | return error; | |
6146f0d5 | 2031 | |
1da177e4 LT |
2032 | /* |
2033 | * An append-only file must be opened in append mode for writing. | |
2034 | */ | |
2035 | if (IS_APPEND(inode)) { | |
8737c930 | 2036 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2037 | return -EPERM; |
1da177e4 | 2038 | if (flag & O_TRUNC) |
7715b521 | 2039 | return -EPERM; |
1da177e4 LT |
2040 | } |
2041 | ||
2042 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2043 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2044 | return -EPERM; |
1da177e4 | 2045 | |
f3c7691e | 2046 | return 0; |
7715b521 | 2047 | } |
1da177e4 | 2048 | |
e1181ee6 | 2049 | static int handle_truncate(struct file *filp) |
7715b521 | 2050 | { |
e1181ee6 | 2051 | struct path *path = &filp->f_path; |
7715b521 AV |
2052 | struct inode *inode = path->dentry->d_inode; |
2053 | int error = get_write_access(inode); | |
2054 | if (error) | |
2055 | return error; | |
2056 | /* | |
2057 | * Refuse to truncate files with mandatory locks held on them. | |
2058 | */ | |
2059 | error = locks_verify_locked(inode); | |
2060 | if (!error) | |
ea0d3ab2 | 2061 | error = security_path_truncate(path); |
7715b521 AV |
2062 | if (!error) { |
2063 | error = do_truncate(path->dentry, 0, | |
2064 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2065 | filp); |
7715b521 AV |
2066 | } |
2067 | put_write_access(inode); | |
acd0c935 | 2068 | return error; |
1da177e4 LT |
2069 | } |
2070 | ||
d57999e1 DH |
2071 | static inline int open_to_namei_flags(int flag) |
2072 | { | |
8a5e929d AV |
2073 | if ((flag & O_ACCMODE) == 3) |
2074 | flag--; | |
d57999e1 DH |
2075 | return flag; |
2076 | } | |
2077 | ||
31e6b01f | 2078 | /* |
fe2d35ff | 2079 | * Handle the last step of open() |
31e6b01f | 2080 | */ |
fb1cc555 | 2081 | static struct file *do_last(struct nameidata *nd, struct path *path, |
c3e380b0 | 2082 | const struct open_flags *op, const char *pathname) |
fb1cc555 | 2083 | { |
a1e28038 | 2084 | struct dentry *dir = nd->path.dentry; |
6c0d46c4 | 2085 | struct dentry *dentry; |
ca344a89 | 2086 | int open_flag = op->open_flag; |
6c0d46c4 | 2087 | int will_truncate = open_flag & O_TRUNC; |
ca344a89 | 2088 | int want_write = 0; |
bcda7652 | 2089 | int acc_mode = op->acc_mode; |
fb1cc555 | 2090 | struct file *filp; |
16c2cd71 | 2091 | int error; |
1f36f774 | 2092 | |
c3e380b0 AV |
2093 | nd->flags &= ~LOOKUP_PARENT; |
2094 | nd->flags |= op->intent; | |
2095 | ||
1f36f774 AV |
2096 | switch (nd->last_type) { |
2097 | case LAST_DOTDOT: | |
176306f5 | 2098 | case LAST_DOT: |
fe2d35ff AV |
2099 | error = handle_dots(nd, nd->last_type); |
2100 | if (error) | |
2101 | return ERR_PTR(error); | |
1f36f774 | 2102 | /* fallthrough */ |
1f36f774 | 2103 | case LAST_ROOT: |
9f1fafee | 2104 | error = complete_walk(nd); |
16c2cd71 | 2105 | if (error) |
9f1fafee | 2106 | return ERR_PTR(error); |
fe2d35ff | 2107 | audit_inode(pathname, nd->path.dentry); |
ca344a89 | 2108 | if (open_flag & O_CREAT) { |
fe2d35ff AV |
2109 | error = -EISDIR; |
2110 | goto exit; | |
2111 | } | |
2112 | goto ok; | |
1f36f774 | 2113 | case LAST_BIND: |
9f1fafee | 2114 | error = complete_walk(nd); |
16c2cd71 | 2115 | if (error) |
9f1fafee | 2116 | return ERR_PTR(error); |
1f36f774 | 2117 | audit_inode(pathname, dir); |
67ee3ad2 | 2118 | goto ok; |
1f36f774 | 2119 | } |
67ee3ad2 | 2120 | |
ca344a89 | 2121 | if (!(open_flag & O_CREAT)) { |
bcda7652 | 2122 | int symlink_ok = 0; |
fe2d35ff AV |
2123 | if (nd->last.name[nd->last.len]) |
2124 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
bcda7652 AV |
2125 | if (open_flag & O_PATH && !(nd->flags & LOOKUP_FOLLOW)) |
2126 | symlink_ok = 1; | |
fe2d35ff | 2127 | /* we _can_ be in RCU mode here */ |
ce57dfc1 AV |
2128 | error = walk_component(nd, path, &nd->last, LAST_NORM, |
2129 | !symlink_ok); | |
2130 | if (error < 0) | |
fe2d35ff | 2131 | return ERR_PTR(error); |
ce57dfc1 | 2132 | if (error) /* symlink */ |
fe2d35ff | 2133 | return NULL; |
fe2d35ff | 2134 | /* sayonara */ |
9f1fafee AV |
2135 | error = complete_walk(nd); |
2136 | if (error) | |
2137 | return ERR_PTR(-ECHILD); | |
fe2d35ff AV |
2138 | |
2139 | error = -ENOTDIR; | |
2140 | if (nd->flags & LOOKUP_DIRECTORY) { | |
ce57dfc1 | 2141 | if (!nd->inode->i_op->lookup) |
fe2d35ff AV |
2142 | goto exit; |
2143 | } | |
2144 | audit_inode(pathname, nd->path.dentry); | |
2145 | goto ok; | |
2146 | } | |
2147 | ||
2148 | /* create side of things */ | |
9f1fafee AV |
2149 | error = complete_walk(nd); |
2150 | if (error) | |
2151 | return ERR_PTR(error); | |
fe2d35ff AV |
2152 | |
2153 | audit_inode(pathname, dir); | |
16c2cd71 | 2154 | error = -EISDIR; |
1f36f774 | 2155 | /* trailing slashes? */ |
31e6b01f NP |
2156 | if (nd->last.name[nd->last.len]) |
2157 | goto exit; | |
a2c36b45 | 2158 | |
a1e28038 AV |
2159 | mutex_lock(&dir->d_inode->i_mutex); |
2160 | ||
6c0d46c4 AV |
2161 | dentry = lookup_hash(nd); |
2162 | error = PTR_ERR(dentry); | |
2163 | if (IS_ERR(dentry)) { | |
fb1cc555 AV |
2164 | mutex_unlock(&dir->d_inode->i_mutex); |
2165 | goto exit; | |
2166 | } | |
2167 | ||
6c0d46c4 AV |
2168 | path->dentry = dentry; |
2169 | path->mnt = nd->path.mnt; | |
2170 | ||
fb1cc555 | 2171 | /* Negative dentry, just create the file */ |
6c0d46c4 AV |
2172 | if (!dentry->d_inode) { |
2173 | int mode = op->mode; | |
2174 | if (!IS_POSIXACL(dir->d_inode)) | |
2175 | mode &= ~current_umask(); | |
fb1cc555 AV |
2176 | /* |
2177 | * This write is needed to ensure that a | |
6c0d46c4 | 2178 | * rw->ro transition does not occur between |
fb1cc555 AV |
2179 | * the time when the file is created and when |
2180 | * a permanent write count is taken through | |
2181 | * the 'struct file' in nameidata_to_filp(). | |
2182 | */ | |
2183 | error = mnt_want_write(nd->path.mnt); | |
2184 | if (error) | |
2185 | goto exit_mutex_unlock; | |
ca344a89 | 2186 | want_write = 1; |
9b44f1b3 | 2187 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 2188 | open_flag &= ~O_TRUNC; |
6c0d46c4 | 2189 | will_truncate = 0; |
bcda7652 | 2190 | acc_mode = MAY_OPEN; |
6c0d46c4 AV |
2191 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2192 | if (error) | |
2193 | goto exit_mutex_unlock; | |
2194 | error = vfs_create(dir->d_inode, dentry, mode, nd); | |
2195 | if (error) | |
2196 | goto exit_mutex_unlock; | |
2197 | mutex_unlock(&dir->d_inode->i_mutex); | |
2198 | dput(nd->path.dentry); | |
2199 | nd->path.dentry = dentry; | |
ca344a89 | 2200 | goto common; |
fb1cc555 AV |
2201 | } |
2202 | ||
2203 | /* | |
2204 | * It already exists. | |
2205 | */ | |
2206 | mutex_unlock(&dir->d_inode->i_mutex); | |
2207 | audit_inode(pathname, path->dentry); | |
2208 | ||
2209 | error = -EEXIST; | |
ca344a89 | 2210 | if (open_flag & O_EXCL) |
fb1cc555 AV |
2211 | goto exit_dput; |
2212 | ||
9875cf80 DH |
2213 | error = follow_managed(path, nd->flags); |
2214 | if (error < 0) | |
2215 | goto exit_dput; | |
fb1cc555 AV |
2216 | |
2217 | error = -ENOENT; | |
2218 | if (!path->dentry->d_inode) | |
2219 | goto exit_dput; | |
9e67f361 AV |
2220 | |
2221 | if (path->dentry->d_inode->i_op->follow_link) | |
fb1cc555 | 2222 | return NULL; |
fb1cc555 AV |
2223 | |
2224 | path_to_nameidata(path, nd); | |
31e6b01f | 2225 | nd->inode = path->dentry->d_inode; |
fb1cc555 | 2226 | error = -EISDIR; |
31e6b01f | 2227 | if (S_ISDIR(nd->inode->i_mode)) |
fb1cc555 | 2228 | goto exit; |
67ee3ad2 | 2229 | ok: |
6c0d46c4 AV |
2230 | if (!S_ISREG(nd->inode->i_mode)) |
2231 | will_truncate = 0; | |
2232 | ||
0f9d1a10 AV |
2233 | if (will_truncate) { |
2234 | error = mnt_want_write(nd->path.mnt); | |
2235 | if (error) | |
2236 | goto exit; | |
ca344a89 | 2237 | want_write = 1; |
0f9d1a10 | 2238 | } |
ca344a89 | 2239 | common: |
bcda7652 | 2240 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 2241 | if (error) |
0f9d1a10 | 2242 | goto exit; |
0f9d1a10 AV |
2243 | filp = nameidata_to_filp(nd); |
2244 | if (!IS_ERR(filp)) { | |
2245 | error = ima_file_check(filp, op->acc_mode); | |
2246 | if (error) { | |
2247 | fput(filp); | |
2248 | filp = ERR_PTR(error); | |
2249 | } | |
2250 | } | |
2251 | if (!IS_ERR(filp)) { | |
2252 | if (will_truncate) { | |
2253 | error = handle_truncate(filp); | |
2254 | if (error) { | |
2255 | fput(filp); | |
2256 | filp = ERR_PTR(error); | |
2257 | } | |
2258 | } | |
2259 | } | |
ca344a89 AV |
2260 | out: |
2261 | if (want_write) | |
0f9d1a10 AV |
2262 | mnt_drop_write(nd->path.mnt); |
2263 | path_put(&nd->path); | |
fb1cc555 AV |
2264 | return filp; |
2265 | ||
2266 | exit_mutex_unlock: | |
2267 | mutex_unlock(&dir->d_inode->i_mutex); | |
2268 | exit_dput: | |
2269 | path_put_conditional(path, nd); | |
2270 | exit: | |
ca344a89 AV |
2271 | filp = ERR_PTR(error); |
2272 | goto out; | |
fb1cc555 AV |
2273 | } |
2274 | ||
13aab428 | 2275 | static struct file *path_openat(int dfd, const char *pathname, |
73d049a4 | 2276 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 2277 | { |
fe2d35ff | 2278 | struct file *base = NULL; |
4a3fd211 | 2279 | struct file *filp; |
9850c056 | 2280 | struct path path; |
13aab428 | 2281 | int error; |
31e6b01f NP |
2282 | |
2283 | filp = get_empty_filp(); | |
2284 | if (!filp) | |
2285 | return ERR_PTR(-ENFILE); | |
2286 | ||
47c805dc | 2287 | filp->f_flags = op->open_flag; |
73d049a4 AV |
2288 | nd->intent.open.file = filp; |
2289 | nd->intent.open.flags = open_to_namei_flags(op->open_flag); | |
2290 | nd->intent.open.create_mode = op->mode; | |
31e6b01f | 2291 | |
73d049a4 | 2292 | error = path_init(dfd, pathname, flags | LOOKUP_PARENT, nd, &base); |
31e6b01f | 2293 | if (unlikely(error)) |
13aab428 | 2294 | goto out_filp; |
31e6b01f | 2295 | |
fe2d35ff | 2296 | current->total_link_count = 0; |
73d049a4 | 2297 | error = link_path_walk(pathname, nd); |
31e6b01f NP |
2298 | if (unlikely(error)) |
2299 | goto out_filp; | |
1da177e4 | 2300 | |
73d049a4 | 2301 | filp = do_last(nd, &path, op, pathname); |
806b681c | 2302 | while (unlikely(!filp)) { /* trailing symlink */ |
7b9337aa | 2303 | struct path link = path; |
def4af30 | 2304 | void *cookie; |
574197e0 | 2305 | if (!(nd->flags & LOOKUP_FOLLOW)) { |
73d049a4 AV |
2306 | path_put_conditional(&path, nd); |
2307 | path_put(&nd->path); | |
40b39136 AV |
2308 | filp = ERR_PTR(-ELOOP); |
2309 | break; | |
2310 | } | |
73d049a4 AV |
2311 | nd->flags |= LOOKUP_PARENT; |
2312 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
574197e0 | 2313 | error = follow_link(&link, nd, &cookie); |
c3e380b0 | 2314 | if (unlikely(error)) |
f1afe9ef | 2315 | filp = ERR_PTR(error); |
c3e380b0 | 2316 | else |
73d049a4 | 2317 | filp = do_last(nd, &path, op, pathname); |
574197e0 | 2318 | put_link(nd, &link, cookie); |
806b681c | 2319 | } |
10fa8e62 | 2320 | out: |
73d049a4 AV |
2321 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) |
2322 | path_put(&nd->root); | |
fe2d35ff AV |
2323 | if (base) |
2324 | fput(base); | |
73d049a4 | 2325 | release_open_intent(nd); |
10fa8e62 | 2326 | return filp; |
1da177e4 | 2327 | |
31e6b01f | 2328 | out_filp: |
806b681c | 2329 | filp = ERR_PTR(error); |
10fa8e62 | 2330 | goto out; |
1da177e4 LT |
2331 | } |
2332 | ||
13aab428 AV |
2333 | struct file *do_filp_open(int dfd, const char *pathname, |
2334 | const struct open_flags *op, int flags) | |
2335 | { | |
73d049a4 | 2336 | struct nameidata nd; |
13aab428 AV |
2337 | struct file *filp; |
2338 | ||
73d049a4 | 2339 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 2340 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 2341 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 2342 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 2343 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
2344 | return filp; |
2345 | } | |
2346 | ||
73d049a4 AV |
2347 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
2348 | const char *name, const struct open_flags *op, int flags) | |
2349 | { | |
2350 | struct nameidata nd; | |
2351 | struct file *file; | |
2352 | ||
2353 | nd.root.mnt = mnt; | |
2354 | nd.root.dentry = dentry; | |
2355 | ||
2356 | flags |= LOOKUP_ROOT; | |
2357 | ||
bcda7652 | 2358 | if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
2359 | return ERR_PTR(-ELOOP); |
2360 | ||
2361 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU); | |
2362 | if (unlikely(file == ERR_PTR(-ECHILD))) | |
2363 | file = path_openat(-1, name, &nd, op, flags); | |
2364 | if (unlikely(file == ERR_PTR(-ESTALE))) | |
2365 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL); | |
2366 | return file; | |
2367 | } | |
2368 | ||
ed75e95d | 2369 | struct dentry *kern_path_create(int dfd, const char *pathname, struct path *path, int is_dir) |
1da177e4 | 2370 | { |
c663e5d8 | 2371 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d AV |
2372 | struct nameidata nd; |
2373 | int error = do_path_lookup(dfd, pathname, LOOKUP_PARENT, &nd); | |
2374 | if (error) | |
2375 | return ERR_PTR(error); | |
1da177e4 | 2376 | |
c663e5d8 CH |
2377 | /* |
2378 | * Yucky last component or no last component at all? | |
2379 | * (foo/., foo/.., /////) | |
2380 | */ | |
ed75e95d AV |
2381 | if (nd.last_type != LAST_NORM) |
2382 | goto out; | |
2383 | nd.flags &= ~LOOKUP_PARENT; | |
2384 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
2385 | nd.intent.open.flags = O_EXCL; | |
c663e5d8 CH |
2386 | |
2387 | /* | |
2388 | * Do the final lookup. | |
2389 | */ | |
ed75e95d AV |
2390 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
2391 | dentry = lookup_hash(&nd); | |
1da177e4 LT |
2392 | if (IS_ERR(dentry)) |
2393 | goto fail; | |
c663e5d8 | 2394 | |
e9baf6e5 AV |
2395 | if (dentry->d_inode) |
2396 | goto eexist; | |
c663e5d8 CH |
2397 | /* |
2398 | * Special case - lookup gave negative, but... we had foo/bar/ | |
2399 | * From the vfs_mknod() POV we just have a negative dentry - | |
2400 | * all is fine. Let's be bastards - you had / on the end, you've | |
2401 | * been asking for (non-existent) directory. -ENOENT for you. | |
2402 | */ | |
ed75e95d | 2403 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
e9baf6e5 AV |
2404 | dput(dentry); |
2405 | dentry = ERR_PTR(-ENOENT); | |
ed75e95d | 2406 | goto fail; |
e9baf6e5 | 2407 | } |
ed75e95d | 2408 | *path = nd.path; |
1da177e4 | 2409 | return dentry; |
e9baf6e5 | 2410 | eexist: |
1da177e4 | 2411 | dput(dentry); |
e9baf6e5 | 2412 | dentry = ERR_PTR(-EEXIST); |
1da177e4 | 2413 | fail: |
ed75e95d AV |
2414 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
2415 | out: | |
2416 | path_put(&nd.path); | |
1da177e4 LT |
2417 | return dentry; |
2418 | } | |
dae6ad8f AV |
2419 | EXPORT_SYMBOL(kern_path_create); |
2420 | ||
2421 | struct dentry *user_path_create(int dfd, const char __user *pathname, struct path *path, int is_dir) | |
2422 | { | |
2423 | char *tmp = getname(pathname); | |
2424 | struct dentry *res; | |
2425 | if (IS_ERR(tmp)) | |
2426 | return ERR_CAST(tmp); | |
2427 | res = kern_path_create(dfd, tmp, path, is_dir); | |
2428 | putname(tmp); | |
2429 | return res; | |
2430 | } | |
2431 | EXPORT_SYMBOL(user_path_create); | |
2432 | ||
1da177e4 LT |
2433 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) |
2434 | { | |
a95164d9 | 2435 | int error = may_create(dir, dentry); |
1da177e4 LT |
2436 | |
2437 | if (error) | |
2438 | return error; | |
2439 | ||
e795b717 SH |
2440 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && |
2441 | !ns_capable(inode_userns(dir), CAP_MKNOD)) | |
1da177e4 LT |
2442 | return -EPERM; |
2443 | ||
acfa4380 | 2444 | if (!dir->i_op->mknod) |
1da177e4 LT |
2445 | return -EPERM; |
2446 | ||
08ce5f16 SH |
2447 | error = devcgroup_inode_mknod(mode, dev); |
2448 | if (error) | |
2449 | return error; | |
2450 | ||
1da177e4 LT |
2451 | error = security_inode_mknod(dir, dentry, mode, dev); |
2452 | if (error) | |
2453 | return error; | |
2454 | ||
1da177e4 | 2455 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 2456 | if (!error) |
f38aa942 | 2457 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2458 | return error; |
2459 | } | |
2460 | ||
463c3197 DH |
2461 | static int may_mknod(mode_t mode) |
2462 | { | |
2463 | switch (mode & S_IFMT) { | |
2464 | case S_IFREG: | |
2465 | case S_IFCHR: | |
2466 | case S_IFBLK: | |
2467 | case S_IFIFO: | |
2468 | case S_IFSOCK: | |
2469 | case 0: /* zero mode translates to S_IFREG */ | |
2470 | return 0; | |
2471 | case S_IFDIR: | |
2472 | return -EPERM; | |
2473 | default: | |
2474 | return -EINVAL; | |
2475 | } | |
2476 | } | |
2477 | ||
2e4d0924 HC |
2478 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
2479 | unsigned, dev) | |
1da177e4 | 2480 | { |
2ad94ae6 | 2481 | struct dentry *dentry; |
dae6ad8f AV |
2482 | struct path path; |
2483 | int error; | |
1da177e4 LT |
2484 | |
2485 | if (S_ISDIR(mode)) | |
2486 | return -EPERM; | |
1da177e4 | 2487 | |
dae6ad8f AV |
2488 | dentry = user_path_create(dfd, filename, &path, 0); |
2489 | if (IS_ERR(dentry)) | |
2490 | return PTR_ERR(dentry); | |
2ad94ae6 | 2491 | |
dae6ad8f | 2492 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 2493 | mode &= ~current_umask(); |
463c3197 DH |
2494 | error = may_mknod(mode); |
2495 | if (error) | |
2496 | goto out_dput; | |
dae6ad8f | 2497 | error = mnt_want_write(path.mnt); |
463c3197 DH |
2498 | if (error) |
2499 | goto out_dput; | |
dae6ad8f | 2500 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 KT |
2501 | if (error) |
2502 | goto out_drop_write; | |
463c3197 | 2503 | switch (mode & S_IFMT) { |
1da177e4 | 2504 | case 0: case S_IFREG: |
dae6ad8f | 2505 | error = vfs_create(path.dentry->d_inode,dentry,mode,NULL); |
1da177e4 LT |
2506 | break; |
2507 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 2508 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2509 | new_decode_dev(dev)); |
2510 | break; | |
2511 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 2512 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2513 | break; |
1da177e4 | 2514 | } |
be6d3e56 | 2515 | out_drop_write: |
dae6ad8f | 2516 | mnt_drop_write(path.mnt); |
463c3197 DH |
2517 | out_dput: |
2518 | dput(dentry); | |
dae6ad8f AV |
2519 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
2520 | path_put(&path); | |
1da177e4 LT |
2521 | |
2522 | return error; | |
2523 | } | |
2524 | ||
3480b257 | 2525 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2526 | { |
2527 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2528 | } | |
2529 | ||
1da177e4 LT |
2530 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2531 | { | |
a95164d9 | 2532 | int error = may_create(dir, dentry); |
1da177e4 LT |
2533 | |
2534 | if (error) | |
2535 | return error; | |
2536 | ||
acfa4380 | 2537 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2538 | return -EPERM; |
2539 | ||
2540 | mode &= (S_IRWXUGO|S_ISVTX); | |
2541 | error = security_inode_mkdir(dir, dentry, mode); | |
2542 | if (error) | |
2543 | return error; | |
2544 | ||
1da177e4 | 2545 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2546 | if (!error) |
f38aa942 | 2547 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2548 | return error; |
2549 | } | |
2550 | ||
2e4d0924 | 2551 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 | 2552 | { |
6902d925 | 2553 | struct dentry *dentry; |
dae6ad8f AV |
2554 | struct path path; |
2555 | int error; | |
1da177e4 | 2556 | |
dae6ad8f | 2557 | dentry = user_path_create(dfd, pathname, &path, 1); |
6902d925 | 2558 | if (IS_ERR(dentry)) |
dae6ad8f | 2559 | return PTR_ERR(dentry); |
1da177e4 | 2560 | |
dae6ad8f | 2561 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 2562 | mode &= ~current_umask(); |
dae6ad8f | 2563 | error = mnt_want_write(path.mnt); |
463c3197 DH |
2564 | if (error) |
2565 | goto out_dput; | |
dae6ad8f | 2566 | error = security_path_mkdir(&path, dentry, mode); |
be6d3e56 KT |
2567 | if (error) |
2568 | goto out_drop_write; | |
dae6ad8f | 2569 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2570 | out_drop_write: |
dae6ad8f | 2571 | mnt_drop_write(path.mnt); |
463c3197 | 2572 | out_dput: |
6902d925 | 2573 | dput(dentry); |
dae6ad8f AV |
2574 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
2575 | path_put(&path); | |
1da177e4 LT |
2576 | return error; |
2577 | } | |
2578 | ||
3cdad428 | 2579 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2580 | { |
2581 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2582 | } | |
2583 | ||
1da177e4 | 2584 | /* |
a71905f0 SW |
2585 | * The dentry_unhash() helper will try to drop the dentry early: we |
2586 | * should have a usage count of 2 if we're the only user of this | |
2587 | * dentry, and if that is true (possibly after pruning the dcache), | |
2588 | * then we drop the dentry now. | |
1da177e4 LT |
2589 | * |
2590 | * A low-level filesystem can, if it choses, legally | |
2591 | * do a | |
2592 | * | |
2593 | * if (!d_unhashed(dentry)) | |
2594 | * return -EBUSY; | |
2595 | * | |
2596 | * if it cannot handle the case of removing a directory | |
2597 | * that is still in use by something else.. | |
2598 | */ | |
2599 | void dentry_unhash(struct dentry *dentry) | |
2600 | { | |
dc168427 | 2601 | shrink_dcache_parent(dentry); |
1da177e4 | 2602 | spin_lock(&dentry->d_lock); |
64252c75 | 2603 | if (dentry->d_count == 1) |
1da177e4 LT |
2604 | __d_drop(dentry); |
2605 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
2606 | } |
2607 | ||
2608 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2609 | { | |
2610 | int error = may_delete(dir, dentry, 1); | |
2611 | ||
2612 | if (error) | |
2613 | return error; | |
2614 | ||
acfa4380 | 2615 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2616 | return -EPERM; |
2617 | ||
1d2ef590 | 2618 | dget(dentry); |
1b1dcc1b | 2619 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
2620 | |
2621 | error = -EBUSY; | |
1da177e4 | 2622 | if (d_mountpoint(dentry)) |
912dbc15 SW |
2623 | goto out; |
2624 | ||
2625 | error = security_inode_rmdir(dir, dentry); | |
2626 | if (error) | |
2627 | goto out; | |
2628 | ||
3cebde24 | 2629 | shrink_dcache_parent(dentry); |
912dbc15 SW |
2630 | error = dir->i_op->rmdir(dir, dentry); |
2631 | if (error) | |
2632 | goto out; | |
2633 | ||
2634 | dentry->d_inode->i_flags |= S_DEAD; | |
2635 | dont_mount(dentry); | |
2636 | ||
2637 | out: | |
1b1dcc1b | 2638 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 2639 | dput(dentry); |
912dbc15 | 2640 | if (!error) |
1da177e4 | 2641 | d_delete(dentry); |
1da177e4 LT |
2642 | return error; |
2643 | } | |
2644 | ||
5590ff0d | 2645 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2646 | { |
2647 | int error = 0; | |
2648 | char * name; | |
2649 | struct dentry *dentry; | |
2650 | struct nameidata nd; | |
2651 | ||
2ad94ae6 | 2652 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2653 | if (error) |
2ad94ae6 | 2654 | return error; |
1da177e4 LT |
2655 | |
2656 | switch(nd.last_type) { | |
0612d9fb OH |
2657 | case LAST_DOTDOT: |
2658 | error = -ENOTEMPTY; | |
2659 | goto exit1; | |
2660 | case LAST_DOT: | |
2661 | error = -EINVAL; | |
2662 | goto exit1; | |
2663 | case LAST_ROOT: | |
2664 | error = -EBUSY; | |
2665 | goto exit1; | |
1da177e4 | 2666 | } |
0612d9fb OH |
2667 | |
2668 | nd.flags &= ~LOOKUP_PARENT; | |
2669 | ||
4ac91378 | 2670 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2671 | dentry = lookup_hash(&nd); |
1da177e4 | 2672 | error = PTR_ERR(dentry); |
6902d925 DH |
2673 | if (IS_ERR(dentry)) |
2674 | goto exit2; | |
e6bc45d6 TT |
2675 | if (!dentry->d_inode) { |
2676 | error = -ENOENT; | |
2677 | goto exit3; | |
2678 | } | |
0622753b DH |
2679 | error = mnt_want_write(nd.path.mnt); |
2680 | if (error) | |
2681 | goto exit3; | |
be6d3e56 KT |
2682 | error = security_path_rmdir(&nd.path, dentry); |
2683 | if (error) | |
2684 | goto exit4; | |
4ac91378 | 2685 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2686 | exit4: |
0622753b DH |
2687 | mnt_drop_write(nd.path.mnt); |
2688 | exit3: | |
6902d925 DH |
2689 | dput(dentry); |
2690 | exit2: | |
4ac91378 | 2691 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2692 | exit1: |
1d957f9b | 2693 | path_put(&nd.path); |
1da177e4 LT |
2694 | putname(name); |
2695 | return error; | |
2696 | } | |
2697 | ||
3cdad428 | 2698 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2699 | { |
2700 | return do_rmdir(AT_FDCWD, pathname); | |
2701 | } | |
2702 | ||
1da177e4 LT |
2703 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2704 | { | |
2705 | int error = may_delete(dir, dentry, 0); | |
2706 | ||
2707 | if (error) | |
2708 | return error; | |
2709 | ||
acfa4380 | 2710 | if (!dir->i_op->unlink) |
1da177e4 LT |
2711 | return -EPERM; |
2712 | ||
1b1dcc1b | 2713 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2714 | if (d_mountpoint(dentry)) |
2715 | error = -EBUSY; | |
2716 | else { | |
2717 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 2718 | if (!error) { |
1da177e4 | 2719 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 2720 | if (!error) |
d83c49f3 | 2721 | dont_mount(dentry); |
bec1052e | 2722 | } |
1da177e4 | 2723 | } |
1b1dcc1b | 2724 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2725 | |
2726 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2727 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2728 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2729 | d_delete(dentry); |
1da177e4 | 2730 | } |
0eeca283 | 2731 | |
1da177e4 LT |
2732 | return error; |
2733 | } | |
2734 | ||
2735 | /* | |
2736 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2737 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2738 | * writeout happening, and we don't want to prevent access to the directory |
2739 | * while waiting on the I/O. | |
2740 | */ | |
5590ff0d | 2741 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2742 | { |
2ad94ae6 AV |
2743 | int error; |
2744 | char *name; | |
1da177e4 LT |
2745 | struct dentry *dentry; |
2746 | struct nameidata nd; | |
2747 | struct inode *inode = NULL; | |
2748 | ||
2ad94ae6 | 2749 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2750 | if (error) |
2ad94ae6 AV |
2751 | return error; |
2752 | ||
1da177e4 LT |
2753 | error = -EISDIR; |
2754 | if (nd.last_type != LAST_NORM) | |
2755 | goto exit1; | |
0612d9fb OH |
2756 | |
2757 | nd.flags &= ~LOOKUP_PARENT; | |
2758 | ||
4ac91378 | 2759 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2760 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2761 | error = PTR_ERR(dentry); |
2762 | if (!IS_ERR(dentry)) { | |
2763 | /* Why not before? Because we want correct error value */ | |
50338b88 TE |
2764 | if (nd.last.name[nd.last.len]) |
2765 | goto slashes; | |
1da177e4 | 2766 | inode = dentry->d_inode; |
50338b88 | 2767 | if (!inode) |
e6bc45d6 TT |
2768 | goto slashes; |
2769 | ihold(inode); | |
0622753b DH |
2770 | error = mnt_want_write(nd.path.mnt); |
2771 | if (error) | |
2772 | goto exit2; | |
be6d3e56 KT |
2773 | error = security_path_unlink(&nd.path, dentry); |
2774 | if (error) | |
2775 | goto exit3; | |
4ac91378 | 2776 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2777 | exit3: |
0622753b | 2778 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2779 | exit2: |
2780 | dput(dentry); | |
2781 | } | |
4ac91378 | 2782 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2783 | if (inode) |
2784 | iput(inode); /* truncate the inode here */ | |
2785 | exit1: | |
1d957f9b | 2786 | path_put(&nd.path); |
1da177e4 LT |
2787 | putname(name); |
2788 | return error; | |
2789 | ||
2790 | slashes: | |
2791 | error = !dentry->d_inode ? -ENOENT : | |
2792 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2793 | goto exit2; | |
2794 | } | |
2795 | ||
2e4d0924 | 2796 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2797 | { |
2798 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2799 | return -EINVAL; | |
2800 | ||
2801 | if (flag & AT_REMOVEDIR) | |
2802 | return do_rmdir(dfd, pathname); | |
2803 | ||
2804 | return do_unlinkat(dfd, pathname); | |
2805 | } | |
2806 | ||
3480b257 | 2807 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2808 | { |
2809 | return do_unlinkat(AT_FDCWD, pathname); | |
2810 | } | |
2811 | ||
db2e747b | 2812 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2813 | { |
a95164d9 | 2814 | int error = may_create(dir, dentry); |
1da177e4 LT |
2815 | |
2816 | if (error) | |
2817 | return error; | |
2818 | ||
acfa4380 | 2819 | if (!dir->i_op->symlink) |
1da177e4 LT |
2820 | return -EPERM; |
2821 | ||
2822 | error = security_inode_symlink(dir, dentry, oldname); | |
2823 | if (error) | |
2824 | return error; | |
2825 | ||
1da177e4 | 2826 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2827 | if (!error) |
f38aa942 | 2828 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2829 | return error; |
2830 | } | |
2831 | ||
2e4d0924 HC |
2832 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2833 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2834 | { |
2ad94ae6 AV |
2835 | int error; |
2836 | char *from; | |
6902d925 | 2837 | struct dentry *dentry; |
dae6ad8f | 2838 | struct path path; |
1da177e4 LT |
2839 | |
2840 | from = getname(oldname); | |
2ad94ae6 | 2841 | if (IS_ERR(from)) |
1da177e4 | 2842 | return PTR_ERR(from); |
1da177e4 | 2843 | |
dae6ad8f | 2844 | dentry = user_path_create(newdfd, newname, &path, 0); |
6902d925 DH |
2845 | error = PTR_ERR(dentry); |
2846 | if (IS_ERR(dentry)) | |
dae6ad8f | 2847 | goto out_putname; |
6902d925 | 2848 | |
dae6ad8f | 2849 | error = mnt_want_write(path.mnt); |
75c3f29d DH |
2850 | if (error) |
2851 | goto out_dput; | |
dae6ad8f | 2852 | error = security_path_symlink(&path, dentry, from); |
be6d3e56 KT |
2853 | if (error) |
2854 | goto out_drop_write; | |
dae6ad8f | 2855 | error = vfs_symlink(path.dentry->d_inode, dentry, from); |
be6d3e56 | 2856 | out_drop_write: |
dae6ad8f | 2857 | mnt_drop_write(path.mnt); |
75c3f29d | 2858 | out_dput: |
6902d925 | 2859 | dput(dentry); |
dae6ad8f AV |
2860 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
2861 | path_put(&path); | |
6902d925 | 2862 | out_putname: |
1da177e4 LT |
2863 | putname(from); |
2864 | return error; | |
2865 | } | |
2866 | ||
3480b257 | 2867 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2868 | { |
2869 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2870 | } | |
2871 | ||
1da177e4 LT |
2872 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2873 | { | |
2874 | struct inode *inode = old_dentry->d_inode; | |
2875 | int error; | |
2876 | ||
2877 | if (!inode) | |
2878 | return -ENOENT; | |
2879 | ||
a95164d9 | 2880 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2881 | if (error) |
2882 | return error; | |
2883 | ||
2884 | if (dir->i_sb != inode->i_sb) | |
2885 | return -EXDEV; | |
2886 | ||
2887 | /* | |
2888 | * A link to an append-only or immutable file cannot be created. | |
2889 | */ | |
2890 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2891 | return -EPERM; | |
acfa4380 | 2892 | if (!dir->i_op->link) |
1da177e4 | 2893 | return -EPERM; |
7e79eedb | 2894 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2895 | return -EPERM; |
2896 | ||
2897 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2898 | if (error) | |
2899 | return error; | |
2900 | ||
7e79eedb | 2901 | mutex_lock(&inode->i_mutex); |
aae8a97d AK |
2902 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
2903 | if (inode->i_nlink == 0) | |
2904 | error = -ENOENT; | |
2905 | else | |
2906 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 2907 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2908 | if (!error) |
7e79eedb | 2909 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2910 | return error; |
2911 | } | |
2912 | ||
2913 | /* | |
2914 | * Hardlinks are often used in delicate situations. We avoid | |
2915 | * security-related surprises by not following symlinks on the | |
2916 | * newname. --KAB | |
2917 | * | |
2918 | * We don't follow them on the oldname either to be compatible | |
2919 | * with linux 2.0, and to avoid hard-linking to directories | |
2920 | * and other special files. --ADM | |
2921 | */ | |
2e4d0924 HC |
2922 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2923 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2924 | { |
2925 | struct dentry *new_dentry; | |
dae6ad8f | 2926 | struct path old_path, new_path; |
11a7b371 | 2927 | int how = 0; |
1da177e4 | 2928 | int error; |
1da177e4 | 2929 | |
11a7b371 | 2930 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 2931 | return -EINVAL; |
11a7b371 AK |
2932 | /* |
2933 | * To use null names we require CAP_DAC_READ_SEARCH | |
2934 | * This ensures that not everyone will be able to create | |
2935 | * handlink using the passed filedescriptor. | |
2936 | */ | |
2937 | if (flags & AT_EMPTY_PATH) { | |
2938 | if (!capable(CAP_DAC_READ_SEARCH)) | |
2939 | return -ENOENT; | |
2940 | how = LOOKUP_EMPTY; | |
2941 | } | |
2942 | ||
2943 | if (flags & AT_SYMLINK_FOLLOW) | |
2944 | how |= LOOKUP_FOLLOW; | |
c04030e1 | 2945 | |
11a7b371 | 2946 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 2947 | if (error) |
2ad94ae6 AV |
2948 | return error; |
2949 | ||
dae6ad8f | 2950 | new_dentry = user_path_create(newdfd, newname, &new_path, 0); |
1da177e4 | 2951 | error = PTR_ERR(new_dentry); |
6902d925 | 2952 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
2953 | goto out; |
2954 | ||
2955 | error = -EXDEV; | |
2956 | if (old_path.mnt != new_path.mnt) | |
2957 | goto out_dput; | |
2958 | error = mnt_want_write(new_path.mnt); | |
75c3f29d DH |
2959 | if (error) |
2960 | goto out_dput; | |
dae6ad8f | 2961 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 KT |
2962 | if (error) |
2963 | goto out_drop_write; | |
dae6ad8f | 2964 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry); |
be6d3e56 | 2965 | out_drop_write: |
dae6ad8f | 2966 | mnt_drop_write(new_path.mnt); |
75c3f29d | 2967 | out_dput: |
6902d925 | 2968 | dput(new_dentry); |
dae6ad8f AV |
2969 | mutex_unlock(&new_path.dentry->d_inode->i_mutex); |
2970 | path_put(&new_path); | |
1da177e4 | 2971 | out: |
2d8f3038 | 2972 | path_put(&old_path); |
1da177e4 LT |
2973 | |
2974 | return error; | |
2975 | } | |
2976 | ||
3480b257 | 2977 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2978 | { |
c04030e1 | 2979 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2980 | } |
2981 | ||
1da177e4 LT |
2982 | /* |
2983 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2984 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2985 | * Problems: | |
2986 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2987 | * b) race potential - two innocent renames can create a loop together. | |
2988 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2989 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2990 | * story. |
2991 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2992 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2993 | * whether the target exists). Solution: try to be smart with locking |
2994 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2995 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2996 | * move will be locked. Thus we can rank directories by the tree |
2997 | * (ancestors first) and rank all non-directories after them. | |
2998 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2999 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
3000 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
3001 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
3002 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 3003 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 3004 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 3005 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 3006 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
3007 | * locking]. |
3008 | */ | |
75c96f85 AB |
3009 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
3010 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3011 | { |
3012 | int error = 0; | |
9055cba7 | 3013 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3014 | |
3015 | /* | |
3016 | * If we are going to change the parent - check write permissions, | |
3017 | * we'll need to flip '..'. | |
3018 | */ | |
3019 | if (new_dir != old_dir) { | |
f419a2e3 | 3020 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3021 | if (error) |
3022 | return error; | |
3023 | } | |
3024 | ||
3025 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3026 | if (error) | |
3027 | return error; | |
3028 | ||
1d2ef590 | 3029 | dget(new_dentry); |
d83c49f3 | 3030 | if (target) |
1b1dcc1b | 3031 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
3032 | |
3033 | error = -EBUSY; | |
3034 | if (d_mountpoint(old_dentry) || d_mountpoint(new_dentry)) | |
3035 | goto out; | |
3036 | ||
3cebde24 SW |
3037 | if (target) |
3038 | shrink_dcache_parent(new_dentry); | |
9055cba7 SW |
3039 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
3040 | if (error) | |
3041 | goto out; | |
3042 | ||
1da177e4 | 3043 | if (target) { |
9055cba7 SW |
3044 | target->i_flags |= S_DEAD; |
3045 | dont_mount(new_dentry); | |
1da177e4 | 3046 | } |
9055cba7 SW |
3047 | out: |
3048 | if (target) | |
3049 | mutex_unlock(&target->i_mutex); | |
1d2ef590 | 3050 | dput(new_dentry); |
e31e14ec | 3051 | if (!error) |
349457cc MF |
3052 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3053 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3054 | return error; |
3055 | } | |
3056 | ||
75c96f85 AB |
3057 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3058 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 | 3059 | { |
51892bbb | 3060 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3061 | int error; |
3062 | ||
3063 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3064 | if (error) | |
3065 | return error; | |
3066 | ||
3067 | dget(new_dentry); | |
1da177e4 | 3068 | if (target) |
1b1dcc1b | 3069 | mutex_lock(&target->i_mutex); |
51892bbb SW |
3070 | |
3071 | error = -EBUSY; | |
1da177e4 | 3072 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
51892bbb SW |
3073 | goto out; |
3074 | ||
3075 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3076 | if (error) | |
3077 | goto out; | |
3078 | ||
3079 | if (target) | |
3080 | dont_mount(new_dentry); | |
3081 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) | |
3082 | d_move(old_dentry, new_dentry); | |
3083 | out: | |
1da177e4 | 3084 | if (target) |
1b1dcc1b | 3085 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3086 | dput(new_dentry); |
3087 | return error; | |
3088 | } | |
3089 | ||
3090 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3091 | struct inode *new_dir, struct dentry *new_dentry) | |
3092 | { | |
3093 | int error; | |
3094 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3095 | const unsigned char *old_name; |
1da177e4 LT |
3096 | |
3097 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3098 | return 0; | |
3099 | ||
3100 | error = may_delete(old_dir, old_dentry, is_dir); | |
3101 | if (error) | |
3102 | return error; | |
3103 | ||
3104 | if (!new_dentry->d_inode) | |
a95164d9 | 3105 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3106 | else |
3107 | error = may_delete(new_dir, new_dentry, is_dir); | |
3108 | if (error) | |
3109 | return error; | |
3110 | ||
acfa4380 | 3111 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3112 | return -EPERM; |
3113 | ||
0eeca283 RL |
3114 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3115 | ||
1da177e4 LT |
3116 | if (is_dir) |
3117 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3118 | else | |
3119 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3120 | if (!error) |
3121 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3122 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3123 | fsnotify_oldname_free(old_name); |
3124 | ||
1da177e4 LT |
3125 | return error; |
3126 | } | |
3127 | ||
2e4d0924 HC |
3128 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3129 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3130 | { |
2ad94ae6 AV |
3131 | struct dentry *old_dir, *new_dir; |
3132 | struct dentry *old_dentry, *new_dentry; | |
3133 | struct dentry *trap; | |
1da177e4 | 3134 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3135 | char *from; |
3136 | char *to; | |
3137 | int error; | |
1da177e4 | 3138 | |
2ad94ae6 | 3139 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3140 | if (error) |
3141 | goto exit; | |
3142 | ||
2ad94ae6 | 3143 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3144 | if (error) |
3145 | goto exit1; | |
3146 | ||
3147 | error = -EXDEV; | |
4ac91378 | 3148 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3149 | goto exit2; |
3150 | ||
4ac91378 | 3151 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3152 | error = -EBUSY; |
3153 | if (oldnd.last_type != LAST_NORM) | |
3154 | goto exit2; | |
3155 | ||
4ac91378 | 3156 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3157 | if (newnd.last_type != LAST_NORM) |
3158 | goto exit2; | |
3159 | ||
0612d9fb OH |
3160 | oldnd.flags &= ~LOOKUP_PARENT; |
3161 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3162 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3163 | |
1da177e4 LT |
3164 | trap = lock_rename(new_dir, old_dir); |
3165 | ||
49705b77 | 3166 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3167 | error = PTR_ERR(old_dentry); |
3168 | if (IS_ERR(old_dentry)) | |
3169 | goto exit3; | |
3170 | /* source must exist */ | |
3171 | error = -ENOENT; | |
3172 | if (!old_dentry->d_inode) | |
3173 | goto exit4; | |
3174 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3175 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3176 | error = -ENOTDIR; | |
3177 | if (oldnd.last.name[oldnd.last.len]) | |
3178 | goto exit4; | |
3179 | if (newnd.last.name[newnd.last.len]) | |
3180 | goto exit4; | |
3181 | } | |
3182 | /* source should not be ancestor of target */ | |
3183 | error = -EINVAL; | |
3184 | if (old_dentry == trap) | |
3185 | goto exit4; | |
49705b77 | 3186 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3187 | error = PTR_ERR(new_dentry); |
3188 | if (IS_ERR(new_dentry)) | |
3189 | goto exit4; | |
3190 | /* target should not be an ancestor of source */ | |
3191 | error = -ENOTEMPTY; | |
3192 | if (new_dentry == trap) | |
3193 | goto exit5; | |
3194 | ||
9079b1eb DH |
3195 | error = mnt_want_write(oldnd.path.mnt); |
3196 | if (error) | |
3197 | goto exit5; | |
be6d3e56 KT |
3198 | error = security_path_rename(&oldnd.path, old_dentry, |
3199 | &newnd.path, new_dentry); | |
3200 | if (error) | |
3201 | goto exit6; | |
1da177e4 LT |
3202 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3203 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 3204 | exit6: |
9079b1eb | 3205 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
3206 | exit5: |
3207 | dput(new_dentry); | |
3208 | exit4: | |
3209 | dput(old_dentry); | |
3210 | exit3: | |
3211 | unlock_rename(new_dir, old_dir); | |
3212 | exit2: | |
1d957f9b | 3213 | path_put(&newnd.path); |
2ad94ae6 | 3214 | putname(to); |
1da177e4 | 3215 | exit1: |
1d957f9b | 3216 | path_put(&oldnd.path); |
1da177e4 | 3217 | putname(from); |
2ad94ae6 | 3218 | exit: |
1da177e4 LT |
3219 | return error; |
3220 | } | |
3221 | ||
a26eab24 | 3222 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3223 | { |
3224 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3225 | } | |
3226 | ||
1da177e4 LT |
3227 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3228 | { | |
3229 | int len; | |
3230 | ||
3231 | len = PTR_ERR(link); | |
3232 | if (IS_ERR(link)) | |
3233 | goto out; | |
3234 | ||
3235 | len = strlen(link); | |
3236 | if (len > (unsigned) buflen) | |
3237 | len = buflen; | |
3238 | if (copy_to_user(buffer, link, len)) | |
3239 | len = -EFAULT; | |
3240 | out: | |
3241 | return len; | |
3242 | } | |
3243 | ||
3244 | /* | |
3245 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3246 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3247 | * using) it for any given inode is up to filesystem. | |
3248 | */ | |
3249 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3250 | { | |
3251 | struct nameidata nd; | |
cc314eef | 3252 | void *cookie; |
694a1764 | 3253 | int res; |
cc314eef | 3254 | |
1da177e4 | 3255 | nd.depth = 0; |
cc314eef | 3256 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3257 | if (IS_ERR(cookie)) |
3258 | return PTR_ERR(cookie); | |
3259 | ||
3260 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3261 | if (dentry->d_inode->i_op->put_link) | |
3262 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3263 | return res; | |
1da177e4 LT |
3264 | } |
3265 | ||
3266 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3267 | { | |
3268 | return __vfs_follow_link(nd, link); | |
3269 | } | |
3270 | ||
3271 | /* get the link contents into pagecache */ | |
3272 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3273 | { | |
ebd09abb DG |
3274 | char *kaddr; |
3275 | struct page *page; | |
1da177e4 | 3276 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3277 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3278 | if (IS_ERR(page)) |
6fe6900e | 3279 | return (char*)page; |
1da177e4 | 3280 | *ppage = page; |
ebd09abb DG |
3281 | kaddr = kmap(page); |
3282 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3283 | return kaddr; | |
1da177e4 LT |
3284 | } |
3285 | ||
3286 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3287 | { | |
3288 | struct page *page = NULL; | |
3289 | char *s = page_getlink(dentry, &page); | |
3290 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3291 | if (page) { | |
3292 | kunmap(page); | |
3293 | page_cache_release(page); | |
3294 | } | |
3295 | return res; | |
3296 | } | |
3297 | ||
cc314eef | 3298 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3299 | { |
cc314eef | 3300 | struct page *page = NULL; |
1da177e4 | 3301 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3302 | return page; |
1da177e4 LT |
3303 | } |
3304 | ||
cc314eef | 3305 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3306 | { |
cc314eef LT |
3307 | struct page *page = cookie; |
3308 | ||
3309 | if (page) { | |
1da177e4 LT |
3310 | kunmap(page); |
3311 | page_cache_release(page); | |
1da177e4 LT |
3312 | } |
3313 | } | |
3314 | ||
54566b2c NP |
3315 | /* |
3316 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3317 | */ | |
3318 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3319 | { |
3320 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3321 | struct page *page; |
afddba49 | 3322 | void *fsdata; |
beb497ab | 3323 | int err; |
1da177e4 | 3324 | char *kaddr; |
54566b2c NP |
3325 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3326 | if (nofs) | |
3327 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3328 | |
7e53cac4 | 3329 | retry: |
afddba49 | 3330 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3331 | flags, &page, &fsdata); |
1da177e4 | 3332 | if (err) |
afddba49 NP |
3333 | goto fail; |
3334 | ||
1da177e4 LT |
3335 | kaddr = kmap_atomic(page, KM_USER0); |
3336 | memcpy(kaddr, symname, len-1); | |
3337 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
3338 | |
3339 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3340 | page, fsdata); | |
1da177e4 LT |
3341 | if (err < 0) |
3342 | goto fail; | |
afddba49 NP |
3343 | if (err < len-1) |
3344 | goto retry; | |
3345 | ||
1da177e4 LT |
3346 | mark_inode_dirty(inode); |
3347 | return 0; | |
1da177e4 LT |
3348 | fail: |
3349 | return err; | |
3350 | } | |
3351 | ||
0adb25d2 KK |
3352 | int page_symlink(struct inode *inode, const char *symname, int len) |
3353 | { | |
3354 | return __page_symlink(inode, symname, len, | |
54566b2c | 3355 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3356 | } |
3357 | ||
92e1d5be | 3358 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3359 | .readlink = generic_readlink, |
3360 | .follow_link = page_follow_link_light, | |
3361 | .put_link = page_put_link, | |
3362 | }; | |
3363 | ||
2d8f3038 | 3364 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3365 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3366 | EXPORT_SYMBOL(follow_down); |
3367 | EXPORT_SYMBOL(follow_up); | |
3368 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
3369 | EXPORT_SYMBOL(getname); | |
3370 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3371 | EXPORT_SYMBOL(lookup_one_len); |
3372 | EXPORT_SYMBOL(page_follow_link_light); | |
3373 | EXPORT_SYMBOL(page_put_link); | |
3374 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3375 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3376 | EXPORT_SYMBOL(page_symlink); |
3377 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
d1811465 | 3378 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3379 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3380 | EXPORT_SYMBOL(inode_permission); |
1da177e4 LT |
3381 | EXPORT_SYMBOL(unlock_rename); |
3382 | EXPORT_SYMBOL(vfs_create); | |
3383 | EXPORT_SYMBOL(vfs_follow_link); | |
3384 | EXPORT_SYMBOL(vfs_link); | |
3385 | EXPORT_SYMBOL(vfs_mkdir); | |
3386 | EXPORT_SYMBOL(vfs_mknod); | |
3387 | EXPORT_SYMBOL(generic_permission); | |
3388 | EXPORT_SYMBOL(vfs_readlink); | |
3389 | EXPORT_SYMBOL(vfs_rename); | |
3390 | EXPORT_SYMBOL(vfs_rmdir); | |
3391 | EXPORT_SYMBOL(vfs_symlink); | |
3392 | EXPORT_SYMBOL(vfs_unlink); | |
3393 | EXPORT_SYMBOL(dentry_unhash); | |
3394 | EXPORT_SYMBOL(generic_readlink); |