Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/namei.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * Some corrections by tytso. | |
9 | */ | |
10 | ||
11 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
12 | * lookup logic. | |
13 | */ | |
14 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
15 | */ | |
16 | ||
17 | #include <linux/init.h> | |
18 | #include <linux/module.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/fs.h> | |
21 | #include <linux/namei.h> | |
1da177e4 | 22 | #include <linux/pagemap.h> |
0eeca283 | 23 | #include <linux/fsnotify.h> |
1da177e4 LT |
24 | #include <linux/personality.h> |
25 | #include <linux/security.h> | |
6146f0d5 | 26 | #include <linux/ima.h> |
1da177e4 LT |
27 | #include <linux/syscalls.h> |
28 | #include <linux/mount.h> | |
29 | #include <linux/audit.h> | |
16f7e0fe | 30 | #include <linux/capability.h> |
834f2a4a | 31 | #include <linux/file.h> |
5590ff0d | 32 | #include <linux/fcntl.h> |
08ce5f16 | 33 | #include <linux/device_cgroup.h> |
5ad4e53b | 34 | #include <linux/fs_struct.h> |
1da177e4 LT |
35 | #include <asm/uaccess.h> |
36 | ||
e81e3f4d EP |
37 | #include "internal.h" |
38 | ||
1da177e4 LT |
39 | /* [Feb-1997 T. Schoebel-Theuer] |
40 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
41 | * were necessary because of omirr. The reason is that omirr needs | |
42 | * to know the _real_ pathname, not the user-supplied one, in case | |
43 | * of symlinks (and also when transname replacements occur). | |
44 | * | |
45 | * The new code replaces the old recursive symlink resolution with | |
46 | * an iterative one (in case of non-nested symlink chains). It does | |
47 | * this with calls to <fs>_follow_link(). | |
48 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
49 | * replaced with a single function lookup_dentry() that can handle all | |
50 | * the special cases of the former code. | |
51 | * | |
52 | * With the new dcache, the pathname is stored at each inode, at least as | |
53 | * long as the refcount of the inode is positive. As a side effect, the | |
54 | * size of the dcache depends on the inode cache and thus is dynamic. | |
55 | * | |
56 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
57 | * resolution to correspond with current state of the code. | |
58 | * | |
59 | * Note that the symlink resolution is not *completely* iterative. | |
60 | * There is still a significant amount of tail- and mid- recursion in | |
61 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
62 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
63 | * may return different results than <fs>_follow_link(). Many virtual | |
64 | * filesystems (including /proc) exhibit this behavior. | |
65 | */ | |
66 | ||
67 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
68 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
69 | * and the name already exists in form of a symlink, try to create the new | |
70 | * name indicated by the symlink. The old code always complained that the | |
71 | * name already exists, due to not following the symlink even if its target | |
72 | * is nonexistent. The new semantics affects also mknod() and link() when | |
25985edc | 73 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
74 | * |
75 | * I don't know which semantics is the right one, since I have no access | |
76 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
77 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
78 | * "old" one. Personally, I think the new semantics is much more logical. | |
79 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
80 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
81 | * and in the old Linux semantics. | |
82 | */ | |
83 | ||
84 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
85 | * semantics. See the comments in "open_namei" and "do_link" below. | |
86 | * | |
87 | * [10-Sep-98 Alan Modra] Another symlink change. | |
88 | */ | |
89 | ||
90 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
91 | * inside the path - always follow. | |
92 | * in the last component in creation/removal/renaming - never follow. | |
93 | * if LOOKUP_FOLLOW passed - follow. | |
94 | * if the pathname has trailing slashes - follow. | |
95 | * otherwise - don't follow. | |
96 | * (applied in that order). | |
97 | * | |
98 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
99 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
100 | * During the 2.4 we need to fix the userland stuff depending on it - | |
101 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
102 | * XEmacs seems to be relying on it... | |
103 | */ | |
104 | /* | |
105 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 106 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
107 | * any extra contention... |
108 | */ | |
109 | ||
110 | /* In order to reduce some races, while at the same time doing additional | |
111 | * checking and hopefully speeding things up, we copy filenames to the | |
112 | * kernel data space before using them.. | |
113 | * | |
114 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
115 | * PATH_MAX includes the nul terminator --RR. | |
116 | */ | |
858119e1 | 117 | static int do_getname(const char __user *filename, char *page) |
1da177e4 LT |
118 | { |
119 | int retval; | |
120 | unsigned long len = PATH_MAX; | |
121 | ||
122 | if (!segment_eq(get_fs(), KERNEL_DS)) { | |
123 | if ((unsigned long) filename >= TASK_SIZE) | |
124 | return -EFAULT; | |
125 | if (TASK_SIZE - (unsigned long) filename < PATH_MAX) | |
126 | len = TASK_SIZE - (unsigned long) filename; | |
127 | } | |
128 | ||
129 | retval = strncpy_from_user(page, filename, len); | |
130 | if (retval > 0) { | |
131 | if (retval < len) | |
132 | return 0; | |
133 | return -ENAMETOOLONG; | |
134 | } else if (!retval) | |
135 | retval = -ENOENT; | |
136 | return retval; | |
137 | } | |
138 | ||
f52e0c11 | 139 | static char *getname_flags(const char __user * filename, int flags) |
1da177e4 LT |
140 | { |
141 | char *tmp, *result; | |
142 | ||
143 | result = ERR_PTR(-ENOMEM); | |
144 | tmp = __getname(); | |
145 | if (tmp) { | |
146 | int retval = do_getname(filename, tmp); | |
147 | ||
148 | result = tmp; | |
149 | if (retval < 0) { | |
f52e0c11 AV |
150 | if (retval != -ENOENT || !(flags & LOOKUP_EMPTY)) { |
151 | __putname(tmp); | |
152 | result = ERR_PTR(retval); | |
153 | } | |
1da177e4 LT |
154 | } |
155 | } | |
156 | audit_getname(result); | |
157 | return result; | |
158 | } | |
159 | ||
f52e0c11 AV |
160 | char *getname(const char __user * filename) |
161 | { | |
162 | return getname_flags(filename, 0); | |
163 | } | |
164 | ||
1da177e4 LT |
165 | #ifdef CONFIG_AUDITSYSCALL |
166 | void putname(const char *name) | |
167 | { | |
5ac3a9c2 | 168 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
169 | audit_putname(name); |
170 | else | |
171 | __putname(name); | |
172 | } | |
173 | EXPORT_SYMBOL(putname); | |
174 | #endif | |
175 | ||
5909ccaa LT |
176 | /* |
177 | * This does basic POSIX ACL permission checking | |
1da177e4 | 178 | */ |
178ea735 | 179 | static int acl_permission_check(struct inode *inode, int mask, unsigned int flags) |
1da177e4 | 180 | { |
178ea735 | 181 | int (*check_acl)(struct inode *inode, int mask, unsigned int flags); |
26cf46be | 182 | unsigned int mode = inode->i_mode; |
1da177e4 | 183 | |
e6305c43 AV |
184 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
185 | ||
e795b717 SH |
186 | if (current_user_ns() != inode_userns(inode)) |
187 | goto other_perms; | |
188 | ||
da9592ed | 189 | if (current_fsuid() == inode->i_uid) |
1da177e4 LT |
190 | mode >>= 6; |
191 | else { | |
178ea735 | 192 | check_acl = inode->i_op->check_acl; |
1da177e4 | 193 | if (IS_POSIXACL(inode) && (mode & S_IRWXG) && check_acl) { |
b74c79e9 NP |
194 | int error = check_acl(inode, mask, flags); |
195 | if (error != -EAGAIN) | |
196 | return error; | |
1da177e4 LT |
197 | } |
198 | ||
199 | if (in_group_p(inode->i_gid)) | |
200 | mode >>= 3; | |
201 | } | |
202 | ||
e795b717 | 203 | other_perms: |
1da177e4 LT |
204 | /* |
205 | * If the DACs are ok we don't need any capability check. | |
206 | */ | |
e6305c43 | 207 | if ((mask & ~mode) == 0) |
1da177e4 | 208 | return 0; |
5909ccaa LT |
209 | return -EACCES; |
210 | } | |
211 | ||
212 | /** | |
b74c79e9 | 213 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa LT |
214 | * @inode: inode to check access rights for |
215 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
39191628 | 216 | * @flags: IPERM_FLAG_ flags. |
5909ccaa LT |
217 | * |
218 | * Used to check for read/write/execute permissions on a file. | |
219 | * We use "fsuid" for this, letting us set arbitrary permissions | |
220 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
221 | * are used for other things. |
222 | * | |
223 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
224 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
225 | * It would then be called again in ref-walk mode. | |
5909ccaa | 226 | */ |
178ea735 | 227 | int generic_permission(struct inode *inode, int mask, unsigned int flags) |
5909ccaa LT |
228 | { |
229 | int ret; | |
230 | ||
231 | /* | |
232 | * Do the basic POSIX ACL permission checks. | |
233 | */ | |
178ea735 | 234 | ret = acl_permission_check(inode, mask, flags); |
5909ccaa LT |
235 | if (ret != -EACCES) |
236 | return ret; | |
1da177e4 | 237 | |
1da177e4 LT |
238 | /* |
239 | * Read/write DACs are always overridable. | |
8e833fd2 AV |
240 | * Executable DACs are overridable for all directories and |
241 | * for non-directories that have least one exec bit set. | |
1da177e4 | 242 | */ |
f696a365 | 243 | if (!(mask & MAY_EXEC) || execute_ok(inode)) |
e795b717 | 244 | if (ns_capable(inode_userns(inode), CAP_DAC_OVERRIDE)) |
1da177e4 LT |
245 | return 0; |
246 | ||
247 | /* | |
248 | * Searching includes executable on directories, else just read. | |
249 | */ | |
7ea66001 | 250 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
1da177e4 | 251 | if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE))) |
e795b717 | 252 | if (ns_capable(inode_userns(inode), CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
253 | return 0; |
254 | ||
255 | return -EACCES; | |
256 | } | |
257 | ||
cb23beb5 CH |
258 | /** |
259 | * inode_permission - check for access rights to a given inode | |
260 | * @inode: inode to check permission on | |
261 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
262 | * | |
263 | * Used to check for read/write/execute permissions on an inode. | |
264 | * We use "fsuid" for this, letting us set arbitrary permissions | |
265 | * for filesystem access without changing the "normal" uids which | |
266 | * are used for other things. | |
267 | */ | |
f419a2e3 | 268 | int inode_permission(struct inode *inode, int mask) |
1da177e4 | 269 | { |
e6305c43 | 270 | int retval; |
1da177e4 LT |
271 | |
272 | if (mask & MAY_WRITE) { | |
22590e41 | 273 | umode_t mode = inode->i_mode; |
1da177e4 LT |
274 | |
275 | /* | |
276 | * Nobody gets write access to a read-only fs. | |
277 | */ | |
278 | if (IS_RDONLY(inode) && | |
279 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
280 | return -EROFS; | |
281 | ||
282 | /* | |
283 | * Nobody gets write access to an immutable file. | |
284 | */ | |
285 | if (IS_IMMUTABLE(inode)) | |
286 | return -EACCES; | |
287 | } | |
288 | ||
acfa4380 | 289 | if (inode->i_op->permission) |
b74c79e9 | 290 | retval = inode->i_op->permission(inode, mask, 0); |
f696a365 | 291 | else |
178ea735 | 292 | retval = generic_permission(inode, mask, 0); |
f696a365 | 293 | |
1da177e4 LT |
294 | if (retval) |
295 | return retval; | |
296 | ||
08ce5f16 SH |
297 | retval = devcgroup_inode_permission(inode, mask); |
298 | if (retval) | |
299 | return retval; | |
300 | ||
d09ca739 | 301 | return security_inode_permission(inode, mask); |
1da177e4 LT |
302 | } |
303 | ||
f4d6ff89 AV |
304 | /** |
305 | * exec_permission - check for right to do lookups in a given directory | |
306 | * @inode: inode to check permission on | |
307 | * @flags: IPERM_FLAG_ flags. | |
308 | * | |
309 | * Short-cut version of inode_permission(), for calling on directories | |
310 | * during pathname resolution. Combines parts of inode_permission() | |
311 | * and generic_permission(), and tests ONLY for MAY_EXEC permission. | |
312 | * | |
313 | * If appropriate, check DAC only. If not appropriate, or | |
314 | * short-cut DAC fails, then call ->permission() to do more | |
315 | * complete permission check. | |
316 | */ | |
317 | static inline int exec_permission(struct inode *inode, unsigned int flags) | |
318 | { | |
319 | int ret; | |
320 | struct user_namespace *ns = inode_userns(inode); | |
1fc0f78c AV |
321 | int mask = MAY_EXEC; |
322 | if (flags & IPERM_FLAG_RCU) | |
323 | mask |= MAY_NOT_BLOCK; | |
f4d6ff89 AV |
324 | |
325 | if (inode->i_op->permission) { | |
1fc0f78c | 326 | ret = inode->i_op->permission(inode, mask, flags); |
f4d6ff89 AV |
327 | if (likely(!ret)) |
328 | goto ok; | |
329 | } else { | |
1fc0f78c | 330 | ret = acl_permission_check(inode, mask, flags); |
f4d6ff89 AV |
331 | if (likely(!ret)) |
332 | goto ok; | |
333 | if (ret != -EACCES) | |
334 | return ret; | |
335 | if (ns_capable(ns, CAP_DAC_OVERRIDE) || | |
336 | ns_capable(ns, CAP_DAC_READ_SEARCH)) | |
337 | goto ok; | |
338 | } | |
339 | return ret; | |
340 | ok: | |
341 | return security_inode_exec_permission(inode, flags); | |
342 | } | |
343 | ||
5dd784d0 JB |
344 | /** |
345 | * path_get - get a reference to a path | |
346 | * @path: path to get the reference to | |
347 | * | |
348 | * Given a path increment the reference count to the dentry and the vfsmount. | |
349 | */ | |
350 | void path_get(struct path *path) | |
351 | { | |
352 | mntget(path->mnt); | |
353 | dget(path->dentry); | |
354 | } | |
355 | EXPORT_SYMBOL(path_get); | |
356 | ||
1d957f9b JB |
357 | /** |
358 | * path_put - put a reference to a path | |
359 | * @path: path to put the reference to | |
360 | * | |
361 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
362 | */ | |
363 | void path_put(struct path *path) | |
1da177e4 | 364 | { |
1d957f9b JB |
365 | dput(path->dentry); |
366 | mntput(path->mnt); | |
1da177e4 | 367 | } |
1d957f9b | 368 | EXPORT_SYMBOL(path_put); |
1da177e4 | 369 | |
19660af7 | 370 | /* |
31e6b01f | 371 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
372 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
373 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
374 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
375 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
376 | * got stuck, so ref-walk may continue from there. If this is not successful | |
377 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
378 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 379 | */ |
31e6b01f NP |
380 | |
381 | /** | |
19660af7 AV |
382 | * unlazy_walk - try to switch to ref-walk mode. |
383 | * @nd: nameidata pathwalk data | |
384 | * @dentry: child of nd->path.dentry or NULL | |
39191628 | 385 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 386 | * |
19660af7 AV |
387 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
388 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
389 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 390 | */ |
19660af7 | 391 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry) |
31e6b01f NP |
392 | { |
393 | struct fs_struct *fs = current->fs; | |
394 | struct dentry *parent = nd->path.dentry; | |
5b6ca027 | 395 | int want_root = 0; |
31e6b01f NP |
396 | |
397 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
5b6ca027 AV |
398 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
399 | want_root = 1; | |
31e6b01f NP |
400 | spin_lock(&fs->lock); |
401 | if (nd->root.mnt != fs->root.mnt || | |
402 | nd->root.dentry != fs->root.dentry) | |
403 | goto err_root; | |
404 | } | |
405 | spin_lock(&parent->d_lock); | |
19660af7 AV |
406 | if (!dentry) { |
407 | if (!__d_rcu_to_refcount(parent, nd->seq)) | |
408 | goto err_parent; | |
409 | BUG_ON(nd->inode != parent->d_inode); | |
410 | } else { | |
94c0d4ec AV |
411 | if (dentry->d_parent != parent) |
412 | goto err_parent; | |
19660af7 AV |
413 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); |
414 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
415 | goto err_child; | |
416 | /* | |
417 | * If the sequence check on the child dentry passed, then | |
418 | * the child has not been removed from its parent. This | |
419 | * means the parent dentry must be valid and able to take | |
420 | * a reference at this point. | |
421 | */ | |
422 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
423 | BUG_ON(!parent->d_count); | |
424 | parent->d_count++; | |
425 | spin_unlock(&dentry->d_lock); | |
426 | } | |
31e6b01f | 427 | spin_unlock(&parent->d_lock); |
5b6ca027 | 428 | if (want_root) { |
31e6b01f NP |
429 | path_get(&nd->root); |
430 | spin_unlock(&fs->lock); | |
431 | } | |
432 | mntget(nd->path.mnt); | |
433 | ||
434 | rcu_read_unlock(); | |
435 | br_read_unlock(vfsmount_lock); | |
436 | nd->flags &= ~LOOKUP_RCU; | |
437 | return 0; | |
19660af7 AV |
438 | |
439 | err_child: | |
31e6b01f | 440 | spin_unlock(&dentry->d_lock); |
19660af7 | 441 | err_parent: |
31e6b01f NP |
442 | spin_unlock(&parent->d_lock); |
443 | err_root: | |
5b6ca027 | 444 | if (want_root) |
31e6b01f NP |
445 | spin_unlock(&fs->lock); |
446 | return -ECHILD; | |
447 | } | |
448 | ||
834f2a4a TM |
449 | /** |
450 | * release_open_intent - free up open intent resources | |
451 | * @nd: pointer to nameidata | |
452 | */ | |
453 | void release_open_intent(struct nameidata *nd) | |
454 | { | |
2dab5974 LT |
455 | struct file *file = nd->intent.open.file; |
456 | ||
457 | if (file && !IS_ERR(file)) { | |
458 | if (file->f_path.dentry == NULL) | |
459 | put_filp(file); | |
460 | else | |
461 | fput(file); | |
462 | } | |
834f2a4a TM |
463 | } |
464 | ||
f60aef7e | 465 | static inline int d_revalidate(struct dentry *dentry, struct nameidata *nd) |
34286d66 | 466 | { |
f60aef7e | 467 | return dentry->d_op->d_revalidate(dentry, nd); |
34286d66 NP |
468 | } |
469 | ||
f5e1c1c1 | 470 | static struct dentry * |
bcdc5e01 IK |
471 | do_revalidate(struct dentry *dentry, struct nameidata *nd) |
472 | { | |
f5e1c1c1 | 473 | int status = d_revalidate(dentry, nd); |
bcdc5e01 IK |
474 | if (unlikely(status <= 0)) { |
475 | /* | |
476 | * The dentry failed validation. | |
477 | * If d_revalidate returned 0 attempt to invalidate | |
478 | * the dentry otherwise d_revalidate is asking us | |
479 | * to return a fail status. | |
480 | */ | |
34286d66 | 481 | if (status < 0) { |
f5e1c1c1 | 482 | dput(dentry); |
34286d66 | 483 | dentry = ERR_PTR(status); |
f5e1c1c1 AV |
484 | } else if (!d_invalidate(dentry)) { |
485 | dput(dentry); | |
486 | dentry = NULL; | |
bcdc5e01 IK |
487 | } |
488 | } | |
489 | return dentry; | |
490 | } | |
491 | ||
9f1fafee AV |
492 | /** |
493 | * complete_walk - successful completion of path walk | |
494 | * @nd: pointer nameidata | |
39159de2 | 495 | * |
9f1fafee AV |
496 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
497 | * Revalidate the final result, unless we'd already done that during | |
498 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
499 | * success, -error on failure. In case of failure caller does not | |
500 | * need to drop nd->path. | |
39159de2 | 501 | */ |
9f1fafee | 502 | static int complete_walk(struct nameidata *nd) |
39159de2 | 503 | { |
16c2cd71 | 504 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 505 | int status; |
39159de2 | 506 | |
9f1fafee AV |
507 | if (nd->flags & LOOKUP_RCU) { |
508 | nd->flags &= ~LOOKUP_RCU; | |
509 | if (!(nd->flags & LOOKUP_ROOT)) | |
510 | nd->root.mnt = NULL; | |
511 | spin_lock(&dentry->d_lock); | |
512 | if (unlikely(!__d_rcu_to_refcount(dentry, nd->seq))) { | |
513 | spin_unlock(&dentry->d_lock); | |
514 | rcu_read_unlock(); | |
515 | br_read_unlock(vfsmount_lock); | |
516 | return -ECHILD; | |
517 | } | |
518 | BUG_ON(nd->inode != dentry->d_inode); | |
519 | spin_unlock(&dentry->d_lock); | |
520 | mntget(nd->path.mnt); | |
521 | rcu_read_unlock(); | |
522 | br_read_unlock(vfsmount_lock); | |
523 | } | |
524 | ||
16c2cd71 AV |
525 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
526 | return 0; | |
527 | ||
528 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
39159de2 JL |
529 | return 0; |
530 | ||
16c2cd71 AV |
531 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) |
532 | return 0; | |
533 | ||
534 | /* Note: we do not d_invalidate() */ | |
34286d66 | 535 | status = d_revalidate(dentry, nd); |
39159de2 JL |
536 | if (status > 0) |
537 | return 0; | |
538 | ||
16c2cd71 | 539 | if (!status) |
39159de2 | 540 | status = -ESTALE; |
16c2cd71 | 541 | |
9f1fafee | 542 | path_put(&nd->path); |
39159de2 JL |
543 | return status; |
544 | } | |
545 | ||
2a737871 AV |
546 | static __always_inline void set_root(struct nameidata *nd) |
547 | { | |
f7ad3c6b MS |
548 | if (!nd->root.mnt) |
549 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
550 | } |
551 | ||
6de88d72 AV |
552 | static int link_path_walk(const char *, struct nameidata *); |
553 | ||
31e6b01f NP |
554 | static __always_inline void set_root_rcu(struct nameidata *nd) |
555 | { | |
556 | if (!nd->root.mnt) { | |
557 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
558 | unsigned seq; |
559 | ||
560 | do { | |
561 | seq = read_seqcount_begin(&fs->seq); | |
562 | nd->root = fs->root; | |
c1530019 | 563 | nd->seq = __read_seqcount_begin(&nd->root.dentry->d_seq); |
c28cc364 | 564 | } while (read_seqcount_retry(&fs->seq, seq)); |
31e6b01f NP |
565 | } |
566 | } | |
567 | ||
f1662356 | 568 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 569 | { |
31e6b01f NP |
570 | int ret; |
571 | ||
1da177e4 LT |
572 | if (IS_ERR(link)) |
573 | goto fail; | |
574 | ||
575 | if (*link == '/') { | |
2a737871 | 576 | set_root(nd); |
1d957f9b | 577 | path_put(&nd->path); |
2a737871 AV |
578 | nd->path = nd->root; |
579 | path_get(&nd->root); | |
16c2cd71 | 580 | nd->flags |= LOOKUP_JUMPED; |
1da177e4 | 581 | } |
31e6b01f | 582 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 583 | |
31e6b01f NP |
584 | ret = link_path_walk(link, nd); |
585 | return ret; | |
1da177e4 | 586 | fail: |
1d957f9b | 587 | path_put(&nd->path); |
1da177e4 LT |
588 | return PTR_ERR(link); |
589 | } | |
590 | ||
1d957f9b | 591 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
592 | { |
593 | dput(path->dentry); | |
4ac91378 | 594 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
595 | mntput(path->mnt); |
596 | } | |
597 | ||
7b9337aa NP |
598 | static inline void path_to_nameidata(const struct path *path, |
599 | struct nameidata *nd) | |
051d3812 | 600 | { |
31e6b01f NP |
601 | if (!(nd->flags & LOOKUP_RCU)) { |
602 | dput(nd->path.dentry); | |
603 | if (nd->path.mnt != path->mnt) | |
604 | mntput(nd->path.mnt); | |
9a229683 | 605 | } |
31e6b01f | 606 | nd->path.mnt = path->mnt; |
4ac91378 | 607 | nd->path.dentry = path->dentry; |
051d3812 IK |
608 | } |
609 | ||
574197e0 AV |
610 | static inline void put_link(struct nameidata *nd, struct path *link, void *cookie) |
611 | { | |
612 | struct inode *inode = link->dentry->d_inode; | |
613 | if (!IS_ERR(cookie) && inode->i_op->put_link) | |
614 | inode->i_op->put_link(link->dentry, nd, cookie); | |
615 | path_put(link); | |
616 | } | |
617 | ||
def4af30 | 618 | static __always_inline int |
574197e0 | 619 | follow_link(struct path *link, struct nameidata *nd, void **p) |
1da177e4 LT |
620 | { |
621 | int error; | |
7b9337aa | 622 | struct dentry *dentry = link->dentry; |
1da177e4 | 623 | |
844a3917 AV |
624 | BUG_ON(nd->flags & LOOKUP_RCU); |
625 | ||
0e794589 AV |
626 | if (link->mnt == nd->path.mnt) |
627 | mntget(link->mnt); | |
628 | ||
574197e0 AV |
629 | if (unlikely(current->total_link_count >= 40)) { |
630 | *p = ERR_PTR(-ELOOP); /* no ->put_link(), please */ | |
574197e0 AV |
631 | path_put(&nd->path); |
632 | return -ELOOP; | |
633 | } | |
634 | cond_resched(); | |
635 | current->total_link_count++; | |
636 | ||
7b9337aa | 637 | touch_atime(link->mnt, dentry); |
1da177e4 | 638 | nd_set_link(nd, NULL); |
cd4e91d3 | 639 | |
36f3b4f6 AV |
640 | error = security_inode_follow_link(link->dentry, nd); |
641 | if (error) { | |
642 | *p = ERR_PTR(error); /* no ->put_link(), please */ | |
643 | path_put(&nd->path); | |
644 | return error; | |
645 | } | |
646 | ||
86acdca1 | 647 | nd->last_type = LAST_BIND; |
def4af30 AV |
648 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
649 | error = PTR_ERR(*p); | |
650 | if (!IS_ERR(*p)) { | |
1da177e4 | 651 | char *s = nd_get_link(nd); |
cc314eef | 652 | error = 0; |
1da177e4 LT |
653 | if (s) |
654 | error = __vfs_follow_link(nd, s); | |
bcda7652 | 655 | else if (nd->last_type == LAST_BIND) { |
16c2cd71 | 656 | nd->flags |= LOOKUP_JUMPED; |
b21041d0 AV |
657 | nd->inode = nd->path.dentry->d_inode; |
658 | if (nd->inode->i_op->follow_link) { | |
bcda7652 AV |
659 | /* stepped on a _really_ weird one */ |
660 | path_put(&nd->path); | |
661 | error = -ELOOP; | |
662 | } | |
663 | } | |
1da177e4 | 664 | } |
1da177e4 LT |
665 | return error; |
666 | } | |
667 | ||
31e6b01f NP |
668 | static int follow_up_rcu(struct path *path) |
669 | { | |
670 | struct vfsmount *parent; | |
671 | struct dentry *mountpoint; | |
672 | ||
673 | parent = path->mnt->mnt_parent; | |
674 | if (parent == path->mnt) | |
675 | return 0; | |
676 | mountpoint = path->mnt->mnt_mountpoint; | |
677 | path->dentry = mountpoint; | |
678 | path->mnt = parent; | |
679 | return 1; | |
680 | } | |
681 | ||
bab77ebf | 682 | int follow_up(struct path *path) |
1da177e4 LT |
683 | { |
684 | struct vfsmount *parent; | |
685 | struct dentry *mountpoint; | |
99b7db7b NP |
686 | |
687 | br_read_lock(vfsmount_lock); | |
bab77ebf AV |
688 | parent = path->mnt->mnt_parent; |
689 | if (parent == path->mnt) { | |
99b7db7b | 690 | br_read_unlock(vfsmount_lock); |
1da177e4 LT |
691 | return 0; |
692 | } | |
693 | mntget(parent); | |
bab77ebf | 694 | mountpoint = dget(path->mnt->mnt_mountpoint); |
99b7db7b | 695 | br_read_unlock(vfsmount_lock); |
bab77ebf AV |
696 | dput(path->dentry); |
697 | path->dentry = mountpoint; | |
698 | mntput(path->mnt); | |
699 | path->mnt = parent; | |
1da177e4 LT |
700 | return 1; |
701 | } | |
702 | ||
b5c84bf6 | 703 | /* |
9875cf80 DH |
704 | * Perform an automount |
705 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
706 | * were called with. | |
1da177e4 | 707 | */ |
9875cf80 DH |
708 | static int follow_automount(struct path *path, unsigned flags, |
709 | bool *need_mntput) | |
31e6b01f | 710 | { |
9875cf80 | 711 | struct vfsmount *mnt; |
ea5b778a | 712 | int err; |
9875cf80 DH |
713 | |
714 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
715 | return -EREMOTE; | |
716 | ||
6f45b656 DH |
717 | /* We don't want to mount if someone supplied AT_NO_AUTOMOUNT |
718 | * and this is the terminal part of the path. | |
719 | */ | |
720 | if ((flags & LOOKUP_NO_AUTOMOUNT) && !(flags & LOOKUP_CONTINUE)) | |
721 | return -EISDIR; /* we actually want to stop here */ | |
722 | ||
9875cf80 DH |
723 | /* We want to mount if someone is trying to open/create a file of any |
724 | * type under the mountpoint, wants to traverse through the mountpoint | |
725 | * or wants to open the mounted directory. | |
726 | * | |
727 | * We don't want to mount if someone's just doing a stat and they've | |
728 | * set AT_SYMLINK_NOFOLLOW - unless they're stat'ing a directory and | |
729 | * appended a '/' to the name. | |
730 | */ | |
731 | if (!(flags & LOOKUP_FOLLOW) && | |
732 | !(flags & (LOOKUP_CONTINUE | LOOKUP_DIRECTORY | | |
733 | LOOKUP_OPEN | LOOKUP_CREATE))) | |
734 | return -EISDIR; | |
735 | ||
736 | current->total_link_count++; | |
737 | if (current->total_link_count >= 40) | |
738 | return -ELOOP; | |
739 | ||
740 | mnt = path->dentry->d_op->d_automount(path); | |
741 | if (IS_ERR(mnt)) { | |
742 | /* | |
743 | * The filesystem is allowed to return -EISDIR here to indicate | |
744 | * it doesn't want to automount. For instance, autofs would do | |
745 | * this so that its userspace daemon can mount on this dentry. | |
746 | * | |
747 | * However, we can only permit this if it's a terminal point in | |
748 | * the path being looked up; if it wasn't then the remainder of | |
749 | * the path is inaccessible and we should say so. | |
750 | */ | |
751 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_CONTINUE)) | |
752 | return -EREMOTE; | |
753 | return PTR_ERR(mnt); | |
31e6b01f | 754 | } |
ea5b778a | 755 | |
9875cf80 DH |
756 | if (!mnt) /* mount collision */ |
757 | return 0; | |
31e6b01f | 758 | |
8aef1884 AV |
759 | if (!*need_mntput) { |
760 | /* lock_mount() may release path->mnt on error */ | |
761 | mntget(path->mnt); | |
762 | *need_mntput = true; | |
763 | } | |
19a167af | 764 | err = finish_automount(mnt, path); |
9875cf80 | 765 | |
ea5b778a DH |
766 | switch (err) { |
767 | case -EBUSY: | |
768 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 769 | return 0; |
ea5b778a | 770 | case 0: |
8aef1884 | 771 | path_put(path); |
ea5b778a DH |
772 | path->mnt = mnt; |
773 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 774 | return 0; |
19a167af AV |
775 | default: |
776 | return err; | |
ea5b778a | 777 | } |
19a167af | 778 | |
463ffb2e AV |
779 | } |
780 | ||
9875cf80 DH |
781 | /* |
782 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 783 | * - Flagged for transit management (autofs) |
9875cf80 DH |
784 | * - Flagged as mountpoint |
785 | * - Flagged as automount point | |
786 | * | |
787 | * This may only be called in refwalk mode. | |
788 | * | |
789 | * Serialization is taken care of in namespace.c | |
790 | */ | |
791 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 792 | { |
8aef1884 | 793 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
794 | unsigned managed; |
795 | bool need_mntput = false; | |
8aef1884 | 796 | int ret = 0; |
9875cf80 DH |
797 | |
798 | /* Given that we're not holding a lock here, we retain the value in a | |
799 | * local variable for each dentry as we look at it so that we don't see | |
800 | * the components of that value change under us */ | |
801 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
802 | managed &= DCACHE_MANAGED_DENTRY, | |
803 | unlikely(managed != 0)) { | |
cc53ce53 DH |
804 | /* Allow the filesystem to manage the transit without i_mutex |
805 | * being held. */ | |
806 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
807 | BUG_ON(!path->dentry->d_op); | |
808 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 809 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 810 | if (ret < 0) |
8aef1884 | 811 | break; |
cc53ce53 DH |
812 | } |
813 | ||
9875cf80 DH |
814 | /* Transit to a mounted filesystem. */ |
815 | if (managed & DCACHE_MOUNTED) { | |
816 | struct vfsmount *mounted = lookup_mnt(path); | |
817 | if (mounted) { | |
818 | dput(path->dentry); | |
819 | if (need_mntput) | |
820 | mntput(path->mnt); | |
821 | path->mnt = mounted; | |
822 | path->dentry = dget(mounted->mnt_root); | |
823 | need_mntput = true; | |
824 | continue; | |
825 | } | |
826 | ||
827 | /* Something is mounted on this dentry in another | |
828 | * namespace and/or whatever was mounted there in this | |
829 | * namespace got unmounted before we managed to get the | |
830 | * vfsmount_lock */ | |
831 | } | |
832 | ||
833 | /* Handle an automount point */ | |
834 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
835 | ret = follow_automount(path, flags, &need_mntput); | |
836 | if (ret < 0) | |
8aef1884 | 837 | break; |
9875cf80 DH |
838 | continue; |
839 | } | |
840 | ||
841 | /* We didn't change the current path point */ | |
842 | break; | |
1da177e4 | 843 | } |
8aef1884 AV |
844 | |
845 | if (need_mntput && path->mnt == mnt) | |
846 | mntput(path->mnt); | |
847 | if (ret == -EISDIR) | |
848 | ret = 0; | |
849 | return ret; | |
1da177e4 LT |
850 | } |
851 | ||
cc53ce53 | 852 | int follow_down_one(struct path *path) |
1da177e4 LT |
853 | { |
854 | struct vfsmount *mounted; | |
855 | ||
1c755af4 | 856 | mounted = lookup_mnt(path); |
1da177e4 | 857 | if (mounted) { |
9393bd07 AV |
858 | dput(path->dentry); |
859 | mntput(path->mnt); | |
860 | path->mnt = mounted; | |
861 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
862 | return 1; |
863 | } | |
864 | return 0; | |
865 | } | |
866 | ||
62a7375e IK |
867 | static inline bool managed_dentry_might_block(struct dentry *dentry) |
868 | { | |
869 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT && | |
870 | dentry->d_op->d_manage(dentry, true) < 0); | |
871 | } | |
872 | ||
9875cf80 | 873 | /* |
287548e4 AV |
874 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
875 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
876 | */ |
877 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 878 | struct inode **inode) |
9875cf80 | 879 | { |
62a7375e | 880 | for (;;) { |
9875cf80 | 881 | struct vfsmount *mounted; |
62a7375e IK |
882 | /* |
883 | * Don't forget we might have a non-mountpoint managed dentry | |
884 | * that wants to block transit. | |
885 | */ | |
287548e4 | 886 | if (unlikely(managed_dentry_might_block(path->dentry))) |
ab90911f | 887 | return false; |
62a7375e IK |
888 | |
889 | if (!d_mountpoint(path->dentry)) | |
890 | break; | |
891 | ||
9875cf80 DH |
892 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
893 | if (!mounted) | |
894 | break; | |
895 | path->mnt = mounted; | |
896 | path->dentry = mounted->mnt_root; | |
897 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); | |
59430262 LT |
898 | /* |
899 | * Update the inode too. We don't need to re-check the | |
900 | * dentry sequence number here after this d_inode read, | |
901 | * because a mount-point is always pinned. | |
902 | */ | |
903 | *inode = path->dentry->d_inode; | |
9875cf80 | 904 | } |
9875cf80 DH |
905 | return true; |
906 | } | |
907 | ||
dea39376 | 908 | static void follow_mount_rcu(struct nameidata *nd) |
287548e4 | 909 | { |
dea39376 | 910 | while (d_mountpoint(nd->path.dentry)) { |
287548e4 | 911 | struct vfsmount *mounted; |
dea39376 | 912 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry, 1); |
287548e4 AV |
913 | if (!mounted) |
914 | break; | |
dea39376 AV |
915 | nd->path.mnt = mounted; |
916 | nd->path.dentry = mounted->mnt_root; | |
917 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
287548e4 AV |
918 | } |
919 | } | |
920 | ||
31e6b01f NP |
921 | static int follow_dotdot_rcu(struct nameidata *nd) |
922 | { | |
31e6b01f NP |
923 | set_root_rcu(nd); |
924 | ||
9875cf80 | 925 | while (1) { |
31e6b01f NP |
926 | if (nd->path.dentry == nd->root.dentry && |
927 | nd->path.mnt == nd->root.mnt) { | |
928 | break; | |
929 | } | |
930 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
931 | struct dentry *old = nd->path.dentry; | |
932 | struct dentry *parent = old->d_parent; | |
933 | unsigned seq; | |
934 | ||
935 | seq = read_seqcount_begin(&parent->d_seq); | |
936 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 937 | goto failed; |
31e6b01f NP |
938 | nd->path.dentry = parent; |
939 | nd->seq = seq; | |
940 | break; | |
941 | } | |
942 | if (!follow_up_rcu(&nd->path)) | |
943 | break; | |
944 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
31e6b01f | 945 | } |
dea39376 AV |
946 | follow_mount_rcu(nd); |
947 | nd->inode = nd->path.dentry->d_inode; | |
31e6b01f | 948 | return 0; |
ef7562d5 AV |
949 | |
950 | failed: | |
951 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
952 | if (!(nd->flags & LOOKUP_ROOT)) |
953 | nd->root.mnt = NULL; | |
ef7562d5 AV |
954 | rcu_read_unlock(); |
955 | br_read_unlock(vfsmount_lock); | |
956 | return -ECHILD; | |
31e6b01f NP |
957 | } |
958 | ||
cc53ce53 DH |
959 | /* |
960 | * Follow down to the covering mount currently visible to userspace. At each | |
961 | * point, the filesystem owning that dentry may be queried as to whether the | |
962 | * caller is permitted to proceed or not. | |
cc53ce53 | 963 | */ |
7cc90cc3 | 964 | int follow_down(struct path *path) |
cc53ce53 DH |
965 | { |
966 | unsigned managed; | |
967 | int ret; | |
968 | ||
969 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
970 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
971 | /* Allow the filesystem to manage the transit without i_mutex | |
972 | * being held. | |
973 | * | |
974 | * We indicate to the filesystem if someone is trying to mount | |
975 | * something here. This gives autofs the chance to deny anyone | |
976 | * other than its daemon the right to mount on its | |
977 | * superstructure. | |
978 | * | |
979 | * The filesystem may sleep at this point. | |
980 | */ | |
981 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
982 | BUG_ON(!path->dentry->d_op); | |
983 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 984 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 985 | path->dentry, false); |
cc53ce53 DH |
986 | if (ret < 0) |
987 | return ret == -EISDIR ? 0 : ret; | |
988 | } | |
989 | ||
990 | /* Transit to a mounted filesystem. */ | |
991 | if (managed & DCACHE_MOUNTED) { | |
992 | struct vfsmount *mounted = lookup_mnt(path); | |
993 | if (!mounted) | |
994 | break; | |
995 | dput(path->dentry); | |
996 | mntput(path->mnt); | |
997 | path->mnt = mounted; | |
998 | path->dentry = dget(mounted->mnt_root); | |
999 | continue; | |
1000 | } | |
1001 | ||
1002 | /* Don't handle automount points here */ | |
1003 | break; | |
1004 | } | |
1005 | return 0; | |
1006 | } | |
1007 | ||
9875cf80 DH |
1008 | /* |
1009 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1010 | */ | |
1011 | static void follow_mount(struct path *path) | |
1012 | { | |
1013 | while (d_mountpoint(path->dentry)) { | |
1014 | struct vfsmount *mounted = lookup_mnt(path); | |
1015 | if (!mounted) | |
1016 | break; | |
1017 | dput(path->dentry); | |
1018 | mntput(path->mnt); | |
1019 | path->mnt = mounted; | |
1020 | path->dentry = dget(mounted->mnt_root); | |
1021 | } | |
1022 | } | |
1023 | ||
31e6b01f | 1024 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1025 | { |
2a737871 | 1026 | set_root(nd); |
e518ddb7 | 1027 | |
1da177e4 | 1028 | while(1) { |
4ac91378 | 1029 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1030 | |
2a737871 AV |
1031 | if (nd->path.dentry == nd->root.dentry && |
1032 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1033 | break; |
1034 | } | |
4ac91378 | 1035 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1036 | /* rare case of legitimate dget_parent()... */ |
1037 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1038 | dput(old); |
1039 | break; | |
1040 | } | |
3088dd70 | 1041 | if (!follow_up(&nd->path)) |
1da177e4 | 1042 | break; |
1da177e4 | 1043 | } |
79ed0226 | 1044 | follow_mount(&nd->path); |
31e6b01f | 1045 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1046 | } |
1047 | ||
baa03890 NP |
1048 | /* |
1049 | * Allocate a dentry with name and parent, and perform a parent | |
1050 | * directory ->lookup on it. Returns the new dentry, or ERR_PTR | |
1051 | * on error. parent->d_inode->i_mutex must be held. d_lookup must | |
1052 | * have verified that no child exists while under i_mutex. | |
1053 | */ | |
1054 | static struct dentry *d_alloc_and_lookup(struct dentry *parent, | |
1055 | struct qstr *name, struct nameidata *nd) | |
1056 | { | |
1057 | struct inode *inode = parent->d_inode; | |
1058 | struct dentry *dentry; | |
1059 | struct dentry *old; | |
1060 | ||
1061 | /* Don't create child dentry for a dead directory. */ | |
1062 | if (unlikely(IS_DEADDIR(inode))) | |
1063 | return ERR_PTR(-ENOENT); | |
1064 | ||
1065 | dentry = d_alloc(parent, name); | |
1066 | if (unlikely(!dentry)) | |
1067 | return ERR_PTR(-ENOMEM); | |
1068 | ||
1069 | old = inode->i_op->lookup(inode, dentry, nd); | |
1070 | if (unlikely(old)) { | |
1071 | dput(dentry); | |
1072 | dentry = old; | |
1073 | } | |
1074 | return dentry; | |
1075 | } | |
1076 | ||
44396f4b JB |
1077 | /* |
1078 | * We already have a dentry, but require a lookup to be performed on the parent | |
1079 | * directory to fill in d_inode. Returns the new dentry, or ERR_PTR on error. | |
1080 | * parent->d_inode->i_mutex must be held. d_lookup must have verified that no | |
1081 | * child exists while under i_mutex. | |
1082 | */ | |
1083 | static struct dentry *d_inode_lookup(struct dentry *parent, struct dentry *dentry, | |
1084 | struct nameidata *nd) | |
1085 | { | |
1086 | struct inode *inode = parent->d_inode; | |
1087 | struct dentry *old; | |
1088 | ||
1089 | /* Don't create child dentry for a dead directory. */ | |
1090 | if (unlikely(IS_DEADDIR(inode))) | |
1091 | return ERR_PTR(-ENOENT); | |
1092 | ||
1093 | old = inode->i_op->lookup(inode, dentry, nd); | |
1094 | if (unlikely(old)) { | |
1095 | dput(dentry); | |
1096 | dentry = old; | |
1097 | } | |
1098 | return dentry; | |
1099 | } | |
1100 | ||
1da177e4 LT |
1101 | /* |
1102 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1103 | * small and for now I'd prefer to have fast path as straight as possible. | |
1104 | * It _is_ time-critical. | |
1105 | */ | |
1106 | static int do_lookup(struct nameidata *nd, struct qstr *name, | |
31e6b01f | 1107 | struct path *path, struct inode **inode) |
1da177e4 | 1108 | { |
4ac91378 | 1109 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1110 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1111 | int need_reval = 1; |
1112 | int status = 1; | |
9875cf80 DH |
1113 | int err; |
1114 | ||
b04f784e NP |
1115 | /* |
1116 | * Rename seqlock is not required here because in the off chance | |
1117 | * of a false negative due to a concurrent rename, we're going to | |
1118 | * do the non-racy lookup, below. | |
1119 | */ | |
31e6b01f NP |
1120 | if (nd->flags & LOOKUP_RCU) { |
1121 | unsigned seq; | |
31e6b01f NP |
1122 | *inode = nd->inode; |
1123 | dentry = __d_lookup_rcu(parent, name, &seq, inode); | |
5a18fff2 AV |
1124 | if (!dentry) |
1125 | goto unlazy; | |
1126 | ||
31e6b01f NP |
1127 | /* Memory barrier in read_seqcount_begin of child is enough */ |
1128 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) | |
1129 | return -ECHILD; | |
31e6b01f | 1130 | nd->seq = seq; |
5a18fff2 | 1131 | |
24643087 | 1132 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
5a18fff2 AV |
1133 | status = d_revalidate(dentry, nd); |
1134 | if (unlikely(status <= 0)) { | |
1135 | if (status != -ECHILD) | |
1136 | need_reval = 0; | |
1137 | goto unlazy; | |
1138 | } | |
24643087 | 1139 | } |
44396f4b JB |
1140 | if (unlikely(d_need_lookup(dentry))) |
1141 | goto unlazy; | |
31e6b01f NP |
1142 | path->mnt = mnt; |
1143 | path->dentry = dentry; | |
d6e9bd25 AV |
1144 | if (unlikely(!__follow_mount_rcu(nd, path, inode))) |
1145 | goto unlazy; | |
1146 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1147 | goto unlazy; | |
1148 | return 0; | |
5a18fff2 | 1149 | unlazy: |
19660af7 AV |
1150 | if (unlazy_walk(nd, dentry)) |
1151 | return -ECHILD; | |
5a18fff2 AV |
1152 | } else { |
1153 | dentry = __d_lookup(parent, name); | |
9875cf80 | 1154 | } |
5a18fff2 | 1155 | |
44396f4b JB |
1156 | if (dentry && unlikely(d_need_lookup(dentry))) { |
1157 | dput(dentry); | |
1158 | dentry = NULL; | |
1159 | } | |
5a18fff2 AV |
1160 | retry: |
1161 | if (unlikely(!dentry)) { | |
1162 | struct inode *dir = parent->d_inode; | |
1163 | BUG_ON(nd->inode != dir); | |
1164 | ||
1165 | mutex_lock(&dir->i_mutex); | |
1166 | dentry = d_lookup(parent, name); | |
1167 | if (likely(!dentry)) { | |
1168 | dentry = d_alloc_and_lookup(parent, name, nd); | |
1169 | if (IS_ERR(dentry)) { | |
1170 | mutex_unlock(&dir->i_mutex); | |
1171 | return PTR_ERR(dentry); | |
1172 | } | |
1173 | /* known good */ | |
1174 | need_reval = 0; | |
1175 | status = 1; | |
44396f4b JB |
1176 | } else if (unlikely(d_need_lookup(dentry))) { |
1177 | dentry = d_inode_lookup(parent, dentry, nd); | |
1178 | if (IS_ERR(dentry)) { | |
1179 | mutex_unlock(&dir->i_mutex); | |
1180 | return PTR_ERR(dentry); | |
1181 | } | |
1182 | /* known good */ | |
1183 | need_reval = 0; | |
1184 | status = 1; | |
5a18fff2 AV |
1185 | } |
1186 | mutex_unlock(&dir->i_mutex); | |
1187 | } | |
1188 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) | |
1189 | status = d_revalidate(dentry, nd); | |
1190 | if (unlikely(status <= 0)) { | |
1191 | if (status < 0) { | |
1192 | dput(dentry); | |
1193 | return status; | |
1194 | } | |
1195 | if (!d_invalidate(dentry)) { | |
1196 | dput(dentry); | |
1197 | dentry = NULL; | |
1198 | need_reval = 1; | |
1199 | goto retry; | |
1200 | } | |
24643087 | 1201 | } |
5a18fff2 | 1202 | |
9875cf80 DH |
1203 | path->mnt = mnt; |
1204 | path->dentry = dentry; | |
1205 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1206 | if (unlikely(err < 0)) { |
1207 | path_put_conditional(path, nd); | |
9875cf80 | 1208 | return err; |
89312214 | 1209 | } |
9875cf80 | 1210 | *inode = path->dentry->d_inode; |
1da177e4 | 1211 | return 0; |
1da177e4 LT |
1212 | } |
1213 | ||
52094c8a AV |
1214 | static inline int may_lookup(struct nameidata *nd) |
1215 | { | |
1216 | if (nd->flags & LOOKUP_RCU) { | |
1217 | int err = exec_permission(nd->inode, IPERM_FLAG_RCU); | |
1218 | if (err != -ECHILD) | |
1219 | return err; | |
19660af7 | 1220 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1221 | return -ECHILD; |
1222 | } | |
1223 | return exec_permission(nd->inode, 0); | |
1224 | } | |
1225 | ||
9856fa1b AV |
1226 | static inline int handle_dots(struct nameidata *nd, int type) |
1227 | { | |
1228 | if (type == LAST_DOTDOT) { | |
1229 | if (nd->flags & LOOKUP_RCU) { | |
1230 | if (follow_dotdot_rcu(nd)) | |
1231 | return -ECHILD; | |
1232 | } else | |
1233 | follow_dotdot(nd); | |
1234 | } | |
1235 | return 0; | |
1236 | } | |
1237 | ||
951361f9 AV |
1238 | static void terminate_walk(struct nameidata *nd) |
1239 | { | |
1240 | if (!(nd->flags & LOOKUP_RCU)) { | |
1241 | path_put(&nd->path); | |
1242 | } else { | |
1243 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1244 | if (!(nd->flags & LOOKUP_ROOT)) |
1245 | nd->root.mnt = NULL; | |
951361f9 AV |
1246 | rcu_read_unlock(); |
1247 | br_read_unlock(vfsmount_lock); | |
1248 | } | |
1249 | } | |
1250 | ||
ce57dfc1 AV |
1251 | static inline int walk_component(struct nameidata *nd, struct path *path, |
1252 | struct qstr *name, int type, int follow) | |
1253 | { | |
1254 | struct inode *inode; | |
1255 | int err; | |
1256 | /* | |
1257 | * "." and ".." are special - ".." especially so because it has | |
1258 | * to be able to know about the current root directory and | |
1259 | * parent relationships. | |
1260 | */ | |
1261 | if (unlikely(type != LAST_NORM)) | |
1262 | return handle_dots(nd, type); | |
1263 | err = do_lookup(nd, name, path, &inode); | |
1264 | if (unlikely(err)) { | |
1265 | terminate_walk(nd); | |
1266 | return err; | |
1267 | } | |
1268 | if (!inode) { | |
1269 | path_to_nameidata(path, nd); | |
1270 | terminate_walk(nd); | |
1271 | return -ENOENT; | |
1272 | } | |
1273 | if (unlikely(inode->i_op->follow_link) && follow) { | |
19660af7 AV |
1274 | if (nd->flags & LOOKUP_RCU) { |
1275 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
1276 | terminate_walk(nd); | |
1277 | return -ECHILD; | |
1278 | } | |
1279 | } | |
ce57dfc1 AV |
1280 | BUG_ON(inode != path->dentry->d_inode); |
1281 | return 1; | |
1282 | } | |
1283 | path_to_nameidata(path, nd); | |
1284 | nd->inode = inode; | |
1285 | return 0; | |
1286 | } | |
1287 | ||
b356379a AV |
1288 | /* |
1289 | * This limits recursive symlink follows to 8, while | |
1290 | * limiting consecutive symlinks to 40. | |
1291 | * | |
1292 | * Without that kind of total limit, nasty chains of consecutive | |
1293 | * symlinks can cause almost arbitrarily long lookups. | |
1294 | */ | |
1295 | static inline int nested_symlink(struct path *path, struct nameidata *nd) | |
1296 | { | |
1297 | int res; | |
1298 | ||
b356379a AV |
1299 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1300 | path_put_conditional(path, nd); | |
1301 | path_put(&nd->path); | |
1302 | return -ELOOP; | |
1303 | } | |
1a4022f8 | 1304 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); |
b356379a AV |
1305 | |
1306 | nd->depth++; | |
1307 | current->link_count++; | |
1308 | ||
1309 | do { | |
1310 | struct path link = *path; | |
1311 | void *cookie; | |
574197e0 AV |
1312 | |
1313 | res = follow_link(&link, nd, &cookie); | |
b356379a AV |
1314 | if (!res) |
1315 | res = walk_component(nd, path, &nd->last, | |
1316 | nd->last_type, LOOKUP_FOLLOW); | |
574197e0 | 1317 | put_link(nd, &link, cookie); |
b356379a AV |
1318 | } while (res > 0); |
1319 | ||
1320 | current->link_count--; | |
1321 | nd->depth--; | |
1322 | return res; | |
1323 | } | |
1324 | ||
1da177e4 LT |
1325 | /* |
1326 | * Name resolution. | |
ea3834d9 PM |
1327 | * This is the basic name resolution function, turning a pathname into |
1328 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1329 | * |
ea3834d9 PM |
1330 | * Returns 0 and nd will have valid dentry and mnt on success. |
1331 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1332 | */ |
6de88d72 | 1333 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1334 | { |
1335 | struct path next; | |
1da177e4 LT |
1336 | int err; |
1337 | unsigned int lookup_flags = nd->flags; | |
1338 | ||
1339 | while (*name=='/') | |
1340 | name++; | |
1341 | if (!*name) | |
086e183a | 1342 | return 0; |
1da177e4 | 1343 | |
1da177e4 LT |
1344 | /* At this point we know we have a real path component. */ |
1345 | for(;;) { | |
1346 | unsigned long hash; | |
1347 | struct qstr this; | |
1348 | unsigned int c; | |
fe479a58 | 1349 | int type; |
1da177e4 | 1350 | |
cdce5d6b | 1351 | nd->flags |= LOOKUP_CONTINUE; |
52094c8a AV |
1352 | |
1353 | err = may_lookup(nd); | |
1da177e4 LT |
1354 | if (err) |
1355 | break; | |
1356 | ||
1357 | this.name = name; | |
1358 | c = *(const unsigned char *)name; | |
1359 | ||
1360 | hash = init_name_hash(); | |
1361 | do { | |
1362 | name++; | |
1363 | hash = partial_name_hash(c, hash); | |
1364 | c = *(const unsigned char *)name; | |
1365 | } while (c && (c != '/')); | |
1366 | this.len = name - (const char *) this.name; | |
1367 | this.hash = end_name_hash(hash); | |
1368 | ||
fe479a58 AV |
1369 | type = LAST_NORM; |
1370 | if (this.name[0] == '.') switch (this.len) { | |
1371 | case 2: | |
16c2cd71 | 1372 | if (this.name[1] == '.') { |
fe479a58 | 1373 | type = LAST_DOTDOT; |
16c2cd71 AV |
1374 | nd->flags |= LOOKUP_JUMPED; |
1375 | } | |
fe479a58 AV |
1376 | break; |
1377 | case 1: | |
1378 | type = LAST_DOT; | |
1379 | } | |
5a202bcd AV |
1380 | if (likely(type == LAST_NORM)) { |
1381 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1382 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd AV |
1383 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
1384 | err = parent->d_op->d_hash(parent, nd->inode, | |
1385 | &this); | |
1386 | if (err < 0) | |
1387 | break; | |
1388 | } | |
1389 | } | |
fe479a58 | 1390 | |
1da177e4 LT |
1391 | /* remove trailing slashes? */ |
1392 | if (!c) | |
1393 | goto last_component; | |
1394 | while (*++name == '/'); | |
1395 | if (!*name) | |
b356379a | 1396 | goto last_component; |
1da177e4 | 1397 | |
ce57dfc1 AV |
1398 | err = walk_component(nd, &next, &this, type, LOOKUP_FOLLOW); |
1399 | if (err < 0) | |
1400 | return err; | |
1da177e4 | 1401 | |
ce57dfc1 | 1402 | if (err) { |
b356379a | 1403 | err = nested_symlink(&next, nd); |
1da177e4 | 1404 | if (err) |
a7472bab | 1405 | return err; |
31e6b01f | 1406 | } |
1da177e4 | 1407 | err = -ENOTDIR; |
31e6b01f | 1408 | if (!nd->inode->i_op->lookup) |
1da177e4 LT |
1409 | break; |
1410 | continue; | |
1411 | /* here ends the main loop */ | |
1412 | ||
1da177e4 | 1413 | last_component: |
f55eab82 TM |
1414 | /* Clear LOOKUP_CONTINUE iff it was previously unset */ |
1415 | nd->flags &= lookup_flags | ~LOOKUP_CONTINUE; | |
b356379a AV |
1416 | nd->last = this; |
1417 | nd->last_type = type; | |
086e183a | 1418 | return 0; |
1da177e4 | 1419 | } |
951361f9 | 1420 | terminate_walk(nd); |
1da177e4 LT |
1421 | return err; |
1422 | } | |
1423 | ||
70e9b357 AV |
1424 | static int path_init(int dfd, const char *name, unsigned int flags, |
1425 | struct nameidata *nd, struct file **fp) | |
31e6b01f NP |
1426 | { |
1427 | int retval = 0; | |
1428 | int fput_needed; | |
1429 | struct file *file; | |
1430 | ||
1431 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
16c2cd71 | 1432 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 1433 | nd->depth = 0; |
5b6ca027 AV |
1434 | if (flags & LOOKUP_ROOT) { |
1435 | struct inode *inode = nd->root.dentry->d_inode; | |
73d049a4 AV |
1436 | if (*name) { |
1437 | if (!inode->i_op->lookup) | |
1438 | return -ENOTDIR; | |
1439 | retval = inode_permission(inode, MAY_EXEC); | |
1440 | if (retval) | |
1441 | return retval; | |
1442 | } | |
5b6ca027 AV |
1443 | nd->path = nd->root; |
1444 | nd->inode = inode; | |
1445 | if (flags & LOOKUP_RCU) { | |
1446 | br_read_lock(vfsmount_lock); | |
1447 | rcu_read_lock(); | |
1448 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1449 | } else { | |
1450 | path_get(&nd->path); | |
1451 | } | |
1452 | return 0; | |
1453 | } | |
1454 | ||
31e6b01f | 1455 | nd->root.mnt = NULL; |
31e6b01f NP |
1456 | |
1457 | if (*name=='/') { | |
e41f7d4e AV |
1458 | if (flags & LOOKUP_RCU) { |
1459 | br_read_lock(vfsmount_lock); | |
1460 | rcu_read_lock(); | |
1461 | set_root_rcu(nd); | |
1462 | } else { | |
1463 | set_root(nd); | |
1464 | path_get(&nd->root); | |
1465 | } | |
1466 | nd->path = nd->root; | |
31e6b01f | 1467 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1468 | if (flags & LOOKUP_RCU) { |
1469 | struct fs_struct *fs = current->fs; | |
1470 | unsigned seq; | |
31e6b01f | 1471 | |
e41f7d4e AV |
1472 | br_read_lock(vfsmount_lock); |
1473 | rcu_read_lock(); | |
c28cc364 | 1474 | |
e41f7d4e AV |
1475 | do { |
1476 | seq = read_seqcount_begin(&fs->seq); | |
1477 | nd->path = fs->pwd; | |
1478 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1479 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1480 | } else { | |
1481 | get_fs_pwd(current->fs, &nd->path); | |
1482 | } | |
31e6b01f NP |
1483 | } else { |
1484 | struct dentry *dentry; | |
1485 | ||
1abf0c71 | 1486 | file = fget_raw_light(dfd, &fput_needed); |
31e6b01f NP |
1487 | retval = -EBADF; |
1488 | if (!file) | |
1489 | goto out_fail; | |
1490 | ||
1491 | dentry = file->f_path.dentry; | |
1492 | ||
f52e0c11 AV |
1493 | if (*name) { |
1494 | retval = -ENOTDIR; | |
1495 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
1496 | goto fput_fail; | |
31e6b01f | 1497 | |
78f32a9b | 1498 | retval = exec_permission(dentry->d_inode, 0); |
f52e0c11 AV |
1499 | if (retval) |
1500 | goto fput_fail; | |
1501 | } | |
31e6b01f NP |
1502 | |
1503 | nd->path = file->f_path; | |
e41f7d4e AV |
1504 | if (flags & LOOKUP_RCU) { |
1505 | if (fput_needed) | |
70e9b357 | 1506 | *fp = file; |
e41f7d4e AV |
1507 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
1508 | br_read_lock(vfsmount_lock); | |
1509 | rcu_read_lock(); | |
1510 | } else { | |
1511 | path_get(&file->f_path); | |
1512 | fput_light(file, fput_needed); | |
1513 | } | |
31e6b01f | 1514 | } |
31e6b01f | 1515 | |
31e6b01f | 1516 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1517 | return 0; |
2dfdd266 | 1518 | |
9b4a9b14 AV |
1519 | fput_fail: |
1520 | fput_light(file, fput_needed); | |
1521 | out_fail: | |
1522 | return retval; | |
1523 | } | |
1524 | ||
bd92d7fe AV |
1525 | static inline int lookup_last(struct nameidata *nd, struct path *path) |
1526 | { | |
1527 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1528 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1529 | ||
1530 | nd->flags &= ~LOOKUP_PARENT; | |
1531 | return walk_component(nd, path, &nd->last, nd->last_type, | |
1532 | nd->flags & LOOKUP_FOLLOW); | |
1533 | } | |
1534 | ||
9b4a9b14 | 1535 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
ee0827cd | 1536 | static int path_lookupat(int dfd, const char *name, |
9b4a9b14 AV |
1537 | unsigned int flags, struct nameidata *nd) |
1538 | { | |
70e9b357 | 1539 | struct file *base = NULL; |
bd92d7fe AV |
1540 | struct path path; |
1541 | int err; | |
31e6b01f NP |
1542 | |
1543 | /* | |
1544 | * Path walking is largely split up into 2 different synchronisation | |
1545 | * schemes, rcu-walk and ref-walk (explained in | |
1546 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1547 | * path walk code, but some things particularly setup, cleanup, and | |
1548 | * following mounts are sufficiently divergent that functions are | |
1549 | * duplicated. Typically there is a function foo(), and its RCU | |
1550 | * analogue, foo_rcu(). | |
1551 | * | |
1552 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1553 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1554 | * be handled by restarting a traditional ref-walk (which will always | |
1555 | * be able to complete). | |
1556 | */ | |
bd92d7fe | 1557 | err = path_init(dfd, name, flags | LOOKUP_PARENT, nd, &base); |
ee0827cd | 1558 | |
bd92d7fe AV |
1559 | if (unlikely(err)) |
1560 | return err; | |
ee0827cd AV |
1561 | |
1562 | current->total_link_count = 0; | |
bd92d7fe AV |
1563 | err = link_path_walk(name, nd); |
1564 | ||
1565 | if (!err && !(flags & LOOKUP_PARENT)) { | |
bd92d7fe AV |
1566 | err = lookup_last(nd, &path); |
1567 | while (err > 0) { | |
1568 | void *cookie; | |
1569 | struct path link = path; | |
bd92d7fe | 1570 | nd->flags |= LOOKUP_PARENT; |
574197e0 | 1571 | err = follow_link(&link, nd, &cookie); |
bd92d7fe AV |
1572 | if (!err) |
1573 | err = lookup_last(nd, &path); | |
574197e0 | 1574 | put_link(nd, &link, cookie); |
bd92d7fe AV |
1575 | } |
1576 | } | |
ee0827cd | 1577 | |
9f1fafee AV |
1578 | if (!err) |
1579 | err = complete_walk(nd); | |
bd92d7fe AV |
1580 | |
1581 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
1582 | if (!nd->inode->i_op->lookup) { | |
1583 | path_put(&nd->path); | |
bd23a539 | 1584 | err = -ENOTDIR; |
bd92d7fe AV |
1585 | } |
1586 | } | |
16c2cd71 | 1587 | |
70e9b357 AV |
1588 | if (base) |
1589 | fput(base); | |
ee0827cd | 1590 | |
5b6ca027 | 1591 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
2a737871 AV |
1592 | path_put(&nd->root); |
1593 | nd->root.mnt = NULL; | |
1594 | } | |
bd92d7fe | 1595 | return err; |
ee0827cd | 1596 | } |
31e6b01f | 1597 | |
ee0827cd AV |
1598 | static int do_path_lookup(int dfd, const char *name, |
1599 | unsigned int flags, struct nameidata *nd) | |
1600 | { | |
1601 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); | |
1602 | if (unlikely(retval == -ECHILD)) | |
1603 | retval = path_lookupat(dfd, name, flags, nd); | |
1604 | if (unlikely(retval == -ESTALE)) | |
1605 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); | |
31e6b01f NP |
1606 | |
1607 | if (likely(!retval)) { | |
1608 | if (unlikely(!audit_dummy_context())) { | |
1609 | if (nd->path.dentry && nd->inode) | |
1610 | audit_inode(name, nd->path.dentry); | |
1611 | } | |
1612 | } | |
170aa3d0 | 1613 | return retval; |
1da177e4 LT |
1614 | } |
1615 | ||
c9c6cac0 | 1616 | int kern_path_parent(const char *name, struct nameidata *nd) |
5590ff0d | 1617 | { |
c9c6cac0 | 1618 | return do_path_lookup(AT_FDCWD, name, LOOKUP_PARENT, nd); |
5590ff0d UD |
1619 | } |
1620 | ||
d1811465 AV |
1621 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1622 | { | |
1623 | struct nameidata nd; | |
1624 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1625 | if (!res) | |
1626 | *path = nd.path; | |
1627 | return res; | |
1628 | } | |
1629 | ||
16f18200 JJS |
1630 | /** |
1631 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1632 | * @dentry: pointer to dentry of the base directory | |
1633 | * @mnt: pointer to vfs mount of the base directory | |
1634 | * @name: pointer to file name | |
1635 | * @flags: lookup flags | |
1636 | * @nd: pointer to nameidata | |
1637 | */ | |
1638 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1639 | const char *name, unsigned int flags, | |
1640 | struct nameidata *nd) | |
1641 | { | |
5b6ca027 AV |
1642 | nd->root.dentry = dentry; |
1643 | nd->root.mnt = mnt; | |
1644 | /* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */ | |
1645 | return do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, nd); | |
16f18200 JJS |
1646 | } |
1647 | ||
eead1911 CH |
1648 | static struct dentry *__lookup_hash(struct qstr *name, |
1649 | struct dentry *base, struct nameidata *nd) | |
1da177e4 | 1650 | { |
81fca444 | 1651 | struct inode *inode = base->d_inode; |
057f6c01 | 1652 | struct dentry *dentry; |
1da177e4 LT |
1653 | int err; |
1654 | ||
b74c79e9 | 1655 | err = exec_permission(inode, 0); |
81fca444 CH |
1656 | if (err) |
1657 | return ERR_PTR(err); | |
1da177e4 | 1658 | |
b04f784e NP |
1659 | /* |
1660 | * Don't bother with __d_lookup: callers are for creat as | |
1661 | * well as unlink, so a lot of the time it would cost | |
1662 | * a double lookup. | |
6e6b1bd1 | 1663 | */ |
b04f784e | 1664 | dentry = d_lookup(base, name); |
6e6b1bd1 | 1665 | |
44396f4b JB |
1666 | if (dentry && d_need_lookup(dentry)) { |
1667 | /* | |
1668 | * __lookup_hash is called with the parent dir's i_mutex already | |
1669 | * held, so we are good to go here. | |
1670 | */ | |
1671 | dentry = d_inode_lookup(base, dentry, nd); | |
1672 | if (IS_ERR(dentry)) | |
1673 | return dentry; | |
1674 | } | |
1675 | ||
fb045adb | 1676 | if (dentry && (dentry->d_flags & DCACHE_OP_REVALIDATE)) |
6e6b1bd1 AV |
1677 | dentry = do_revalidate(dentry, nd); |
1678 | ||
baa03890 NP |
1679 | if (!dentry) |
1680 | dentry = d_alloc_and_lookup(base, name, nd); | |
5a202bcd | 1681 | |
1da177e4 LT |
1682 | return dentry; |
1683 | } | |
1684 | ||
057f6c01 JM |
1685 | /* |
1686 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1687 | * needs parent already locked. Doesn't follow mounts. | |
1688 | * SMP-safe. | |
1689 | */ | |
eead1911 | 1690 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1691 | { |
4ac91378 | 1692 | return __lookup_hash(&nd->last, nd->path.dentry, nd); |
1da177e4 LT |
1693 | } |
1694 | ||
eead1911 | 1695 | /** |
a6b91919 | 1696 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1697 | * @name: pathname component to lookup |
1698 | * @base: base directory to lookup from | |
1699 | * @len: maximum length @len should be interpreted to | |
1700 | * | |
a6b91919 RD |
1701 | * Note that this routine is purely a helper for filesystem usage and should |
1702 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1703 | * nameidata argument is passed to the filesystem methods and a filesystem |
1704 | * using this helper needs to be prepared for that. | |
1705 | */ | |
057f6c01 JM |
1706 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1707 | { | |
057f6c01 | 1708 | struct qstr this; |
6a96ba54 AV |
1709 | unsigned long hash; |
1710 | unsigned int c; | |
057f6c01 | 1711 | |
2f9092e1 DW |
1712 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1713 | ||
6a96ba54 AV |
1714 | this.name = name; |
1715 | this.len = len; | |
1716 | if (!len) | |
1717 | return ERR_PTR(-EACCES); | |
1718 | ||
1719 | hash = init_name_hash(); | |
1720 | while (len--) { | |
1721 | c = *(const unsigned char *)name++; | |
1722 | if (c == '/' || c == '\0') | |
1723 | return ERR_PTR(-EACCES); | |
1724 | hash = partial_name_hash(c, hash); | |
1725 | } | |
1726 | this.hash = end_name_hash(hash); | |
5a202bcd AV |
1727 | /* |
1728 | * See if the low-level filesystem might want | |
1729 | * to use its own hash.. | |
1730 | */ | |
1731 | if (base->d_flags & DCACHE_OP_HASH) { | |
1732 | int err = base->d_op->d_hash(base, base->d_inode, &this); | |
1733 | if (err < 0) | |
1734 | return ERR_PTR(err); | |
1735 | } | |
eead1911 | 1736 | |
49705b77 | 1737 | return __lookup_hash(&this, base, NULL); |
057f6c01 JM |
1738 | } |
1739 | ||
2d8f3038 AV |
1740 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1741 | struct path *path) | |
1da177e4 | 1742 | { |
2d8f3038 | 1743 | struct nameidata nd; |
f52e0c11 | 1744 | char *tmp = getname_flags(name, flags); |
1da177e4 | 1745 | int err = PTR_ERR(tmp); |
1da177e4 | 1746 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1747 | |
1748 | BUG_ON(flags & LOOKUP_PARENT); | |
1749 | ||
1750 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1751 | putname(tmp); |
2d8f3038 AV |
1752 | if (!err) |
1753 | *path = nd.path; | |
1da177e4 LT |
1754 | } |
1755 | return err; | |
1756 | } | |
1757 | ||
2ad94ae6 AV |
1758 | static int user_path_parent(int dfd, const char __user *path, |
1759 | struct nameidata *nd, char **name) | |
1760 | { | |
1761 | char *s = getname(path); | |
1762 | int error; | |
1763 | ||
1764 | if (IS_ERR(s)) | |
1765 | return PTR_ERR(s); | |
1766 | ||
1767 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
1768 | if (error) | |
1769 | putname(s); | |
1770 | else | |
1771 | *name = s; | |
1772 | ||
1773 | return error; | |
1774 | } | |
1775 | ||
1da177e4 LT |
1776 | /* |
1777 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
1778 | * minimal. | |
1779 | */ | |
1780 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
1781 | { | |
da9592ed DH |
1782 | uid_t fsuid = current_fsuid(); |
1783 | ||
1da177e4 LT |
1784 | if (!(dir->i_mode & S_ISVTX)) |
1785 | return 0; | |
e795b717 SH |
1786 | if (current_user_ns() != inode_userns(inode)) |
1787 | goto other_userns; | |
da9592ed | 1788 | if (inode->i_uid == fsuid) |
1da177e4 | 1789 | return 0; |
da9592ed | 1790 | if (dir->i_uid == fsuid) |
1da177e4 | 1791 | return 0; |
e795b717 SH |
1792 | |
1793 | other_userns: | |
1794 | return !ns_capable(inode_userns(inode), CAP_FOWNER); | |
1da177e4 LT |
1795 | } |
1796 | ||
1797 | /* | |
1798 | * Check whether we can remove a link victim from directory dir, check | |
1799 | * whether the type of victim is right. | |
1800 | * 1. We can't do it if dir is read-only (done in permission()) | |
1801 | * 2. We should have write and exec permissions on dir | |
1802 | * 3. We can't remove anything from append-only dir | |
1803 | * 4. We can't do anything with immutable dir (done in permission()) | |
1804 | * 5. If the sticky bit on dir is set we should either | |
1805 | * a. be owner of dir, or | |
1806 | * b. be owner of victim, or | |
1807 | * c. have CAP_FOWNER capability | |
1808 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
1809 | * links pointing to it. | |
1810 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
1811 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
1812 | * 9. We can't remove a root or mountpoint. | |
1813 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1814 | * nfs_async_unlink(). | |
1815 | */ | |
858119e1 | 1816 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
1817 | { |
1818 | int error; | |
1819 | ||
1820 | if (!victim->d_inode) | |
1821 | return -ENOENT; | |
1822 | ||
1823 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 1824 | audit_inode_child(victim, dir); |
1da177e4 | 1825 | |
f419a2e3 | 1826 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1827 | if (error) |
1828 | return error; | |
1829 | if (IS_APPEND(dir)) | |
1830 | return -EPERM; | |
1831 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 1832 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
1833 | return -EPERM; |
1834 | if (isdir) { | |
1835 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
1836 | return -ENOTDIR; | |
1837 | if (IS_ROOT(victim)) | |
1838 | return -EBUSY; | |
1839 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
1840 | return -EISDIR; | |
1841 | if (IS_DEADDIR(dir)) | |
1842 | return -ENOENT; | |
1843 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
1844 | return -EBUSY; | |
1845 | return 0; | |
1846 | } | |
1847 | ||
1848 | /* Check whether we can create an object with dentry child in directory | |
1849 | * dir. | |
1850 | * 1. We can't do it if child already exists (open has special treatment for | |
1851 | * this case, but since we are inlined it's OK) | |
1852 | * 2. We can't do it if dir is read-only (done in permission()) | |
1853 | * 3. We should have write and exec permissions on dir | |
1854 | * 4. We can't do it if dir is immutable (done in permission()) | |
1855 | */ | |
a95164d9 | 1856 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
1857 | { |
1858 | if (child->d_inode) | |
1859 | return -EEXIST; | |
1860 | if (IS_DEADDIR(dir)) | |
1861 | return -ENOENT; | |
f419a2e3 | 1862 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
1863 | } |
1864 | ||
1da177e4 LT |
1865 | /* |
1866 | * p1 and p2 should be directories on the same fs. | |
1867 | */ | |
1868 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
1869 | { | |
1870 | struct dentry *p; | |
1871 | ||
1872 | if (p1 == p2) { | |
f2eace23 | 1873 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
1874 | return NULL; |
1875 | } | |
1876 | ||
a11f3a05 | 1877 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 1878 | |
e2761a11 OH |
1879 | p = d_ancestor(p2, p1); |
1880 | if (p) { | |
1881 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
1882 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
1883 | return p; | |
1da177e4 LT |
1884 | } |
1885 | ||
e2761a11 OH |
1886 | p = d_ancestor(p1, p2); |
1887 | if (p) { | |
1888 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
1889 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1890 | return p; | |
1da177e4 LT |
1891 | } |
1892 | ||
f2eace23 IM |
1893 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1894 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
1895 | return NULL; |
1896 | } | |
1897 | ||
1898 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
1899 | { | |
1b1dcc1b | 1900 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 1901 | if (p1 != p2) { |
1b1dcc1b | 1902 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 1903 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
1904 | } |
1905 | } | |
1906 | ||
1907 | int vfs_create(struct inode *dir, struct dentry *dentry, int mode, | |
1908 | struct nameidata *nd) | |
1909 | { | |
a95164d9 | 1910 | int error = may_create(dir, dentry); |
1da177e4 LT |
1911 | |
1912 | if (error) | |
1913 | return error; | |
1914 | ||
acfa4380 | 1915 | if (!dir->i_op->create) |
1da177e4 LT |
1916 | return -EACCES; /* shouldn't it be ENOSYS? */ |
1917 | mode &= S_IALLUGO; | |
1918 | mode |= S_IFREG; | |
1919 | error = security_inode_create(dir, dentry, mode); | |
1920 | if (error) | |
1921 | return error; | |
1da177e4 | 1922 | error = dir->i_op->create(dir, dentry, mode, nd); |
a74574aa | 1923 | if (!error) |
f38aa942 | 1924 | fsnotify_create(dir, dentry); |
1da177e4 LT |
1925 | return error; |
1926 | } | |
1927 | ||
73d049a4 | 1928 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 1929 | { |
3fb64190 | 1930 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
1931 | struct inode *inode = dentry->d_inode; |
1932 | int error; | |
1933 | ||
bcda7652 AV |
1934 | /* O_PATH? */ |
1935 | if (!acc_mode) | |
1936 | return 0; | |
1937 | ||
1da177e4 LT |
1938 | if (!inode) |
1939 | return -ENOENT; | |
1940 | ||
c8fe8f30 CH |
1941 | switch (inode->i_mode & S_IFMT) { |
1942 | case S_IFLNK: | |
1da177e4 | 1943 | return -ELOOP; |
c8fe8f30 CH |
1944 | case S_IFDIR: |
1945 | if (acc_mode & MAY_WRITE) | |
1946 | return -EISDIR; | |
1947 | break; | |
1948 | case S_IFBLK: | |
1949 | case S_IFCHR: | |
3fb64190 | 1950 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 1951 | return -EACCES; |
c8fe8f30 CH |
1952 | /*FALLTHRU*/ |
1953 | case S_IFIFO: | |
1954 | case S_IFSOCK: | |
1da177e4 | 1955 | flag &= ~O_TRUNC; |
c8fe8f30 | 1956 | break; |
4a3fd211 | 1957 | } |
b41572e9 | 1958 | |
3fb64190 | 1959 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
1960 | if (error) |
1961 | return error; | |
6146f0d5 | 1962 | |
1da177e4 LT |
1963 | /* |
1964 | * An append-only file must be opened in append mode for writing. | |
1965 | */ | |
1966 | if (IS_APPEND(inode)) { | |
8737c930 | 1967 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 1968 | return -EPERM; |
1da177e4 | 1969 | if (flag & O_TRUNC) |
7715b521 | 1970 | return -EPERM; |
1da177e4 LT |
1971 | } |
1972 | ||
1973 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 1974 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 1975 | return -EPERM; |
1da177e4 LT |
1976 | |
1977 | /* | |
1978 | * Ensure there are no outstanding leases on the file. | |
1979 | */ | |
b65a9cfc | 1980 | return break_lease(inode, flag); |
7715b521 | 1981 | } |
1da177e4 | 1982 | |
e1181ee6 | 1983 | static int handle_truncate(struct file *filp) |
7715b521 | 1984 | { |
e1181ee6 | 1985 | struct path *path = &filp->f_path; |
7715b521 AV |
1986 | struct inode *inode = path->dentry->d_inode; |
1987 | int error = get_write_access(inode); | |
1988 | if (error) | |
1989 | return error; | |
1990 | /* | |
1991 | * Refuse to truncate files with mandatory locks held on them. | |
1992 | */ | |
1993 | error = locks_verify_locked(inode); | |
1994 | if (!error) | |
ea0d3ab2 | 1995 | error = security_path_truncate(path); |
7715b521 AV |
1996 | if (!error) { |
1997 | error = do_truncate(path->dentry, 0, | |
1998 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 1999 | filp); |
7715b521 AV |
2000 | } |
2001 | put_write_access(inode); | |
acd0c935 | 2002 | return error; |
1da177e4 LT |
2003 | } |
2004 | ||
d57999e1 DH |
2005 | /* |
2006 | * Note that while the flag value (low two bits) for sys_open means: | |
2007 | * 00 - read-only | |
2008 | * 01 - write-only | |
2009 | * 10 - read-write | |
2010 | * 11 - special | |
2011 | * it is changed into | |
2012 | * 00 - no permissions needed | |
2013 | * 01 - read-permission | |
2014 | * 10 - write-permission | |
2015 | * 11 - read-write | |
2016 | * for the internal routines (ie open_namei()/follow_link() etc) | |
2017 | * This is more logical, and also allows the 00 "no perm needed" | |
2018 | * to be used for symlinks (where the permissions are checked | |
2019 | * later). | |
2020 | * | |
2021 | */ | |
2022 | static inline int open_to_namei_flags(int flag) | |
2023 | { | |
2024 | if ((flag+1) & O_ACCMODE) | |
2025 | flag++; | |
2026 | return flag; | |
2027 | } | |
2028 | ||
31e6b01f | 2029 | /* |
fe2d35ff | 2030 | * Handle the last step of open() |
31e6b01f | 2031 | */ |
fb1cc555 | 2032 | static struct file *do_last(struct nameidata *nd, struct path *path, |
c3e380b0 | 2033 | const struct open_flags *op, const char *pathname) |
fb1cc555 | 2034 | { |
a1e28038 | 2035 | struct dentry *dir = nd->path.dentry; |
6c0d46c4 | 2036 | struct dentry *dentry; |
ca344a89 | 2037 | int open_flag = op->open_flag; |
6c0d46c4 | 2038 | int will_truncate = open_flag & O_TRUNC; |
ca344a89 | 2039 | int want_write = 0; |
bcda7652 | 2040 | int acc_mode = op->acc_mode; |
fb1cc555 | 2041 | struct file *filp; |
16c2cd71 | 2042 | int error; |
1f36f774 | 2043 | |
c3e380b0 AV |
2044 | nd->flags &= ~LOOKUP_PARENT; |
2045 | nd->flags |= op->intent; | |
2046 | ||
1f36f774 AV |
2047 | switch (nd->last_type) { |
2048 | case LAST_DOTDOT: | |
176306f5 | 2049 | case LAST_DOT: |
fe2d35ff AV |
2050 | error = handle_dots(nd, nd->last_type); |
2051 | if (error) | |
2052 | return ERR_PTR(error); | |
1f36f774 | 2053 | /* fallthrough */ |
1f36f774 | 2054 | case LAST_ROOT: |
9f1fafee | 2055 | error = complete_walk(nd); |
16c2cd71 | 2056 | if (error) |
9f1fafee | 2057 | return ERR_PTR(error); |
fe2d35ff | 2058 | audit_inode(pathname, nd->path.dentry); |
ca344a89 | 2059 | if (open_flag & O_CREAT) { |
fe2d35ff AV |
2060 | error = -EISDIR; |
2061 | goto exit; | |
2062 | } | |
2063 | goto ok; | |
1f36f774 | 2064 | case LAST_BIND: |
9f1fafee | 2065 | error = complete_walk(nd); |
16c2cd71 | 2066 | if (error) |
9f1fafee | 2067 | return ERR_PTR(error); |
1f36f774 | 2068 | audit_inode(pathname, dir); |
67ee3ad2 | 2069 | goto ok; |
1f36f774 | 2070 | } |
67ee3ad2 | 2071 | |
ca344a89 | 2072 | if (!(open_flag & O_CREAT)) { |
bcda7652 | 2073 | int symlink_ok = 0; |
fe2d35ff AV |
2074 | if (nd->last.name[nd->last.len]) |
2075 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
bcda7652 AV |
2076 | if (open_flag & O_PATH && !(nd->flags & LOOKUP_FOLLOW)) |
2077 | symlink_ok = 1; | |
fe2d35ff | 2078 | /* we _can_ be in RCU mode here */ |
ce57dfc1 AV |
2079 | error = walk_component(nd, path, &nd->last, LAST_NORM, |
2080 | !symlink_ok); | |
2081 | if (error < 0) | |
fe2d35ff | 2082 | return ERR_PTR(error); |
ce57dfc1 | 2083 | if (error) /* symlink */ |
fe2d35ff | 2084 | return NULL; |
fe2d35ff | 2085 | /* sayonara */ |
9f1fafee AV |
2086 | error = complete_walk(nd); |
2087 | if (error) | |
2088 | return ERR_PTR(-ECHILD); | |
fe2d35ff AV |
2089 | |
2090 | error = -ENOTDIR; | |
2091 | if (nd->flags & LOOKUP_DIRECTORY) { | |
ce57dfc1 | 2092 | if (!nd->inode->i_op->lookup) |
fe2d35ff AV |
2093 | goto exit; |
2094 | } | |
2095 | audit_inode(pathname, nd->path.dentry); | |
2096 | goto ok; | |
2097 | } | |
2098 | ||
2099 | /* create side of things */ | |
9f1fafee AV |
2100 | error = complete_walk(nd); |
2101 | if (error) | |
2102 | return ERR_PTR(error); | |
fe2d35ff AV |
2103 | |
2104 | audit_inode(pathname, dir); | |
16c2cd71 | 2105 | error = -EISDIR; |
1f36f774 | 2106 | /* trailing slashes? */ |
31e6b01f NP |
2107 | if (nd->last.name[nd->last.len]) |
2108 | goto exit; | |
a2c36b45 | 2109 | |
a1e28038 AV |
2110 | mutex_lock(&dir->d_inode->i_mutex); |
2111 | ||
6c0d46c4 AV |
2112 | dentry = lookup_hash(nd); |
2113 | error = PTR_ERR(dentry); | |
2114 | if (IS_ERR(dentry)) { | |
fb1cc555 AV |
2115 | mutex_unlock(&dir->d_inode->i_mutex); |
2116 | goto exit; | |
2117 | } | |
2118 | ||
6c0d46c4 AV |
2119 | path->dentry = dentry; |
2120 | path->mnt = nd->path.mnt; | |
2121 | ||
fb1cc555 | 2122 | /* Negative dentry, just create the file */ |
6c0d46c4 AV |
2123 | if (!dentry->d_inode) { |
2124 | int mode = op->mode; | |
2125 | if (!IS_POSIXACL(dir->d_inode)) | |
2126 | mode &= ~current_umask(); | |
fb1cc555 AV |
2127 | /* |
2128 | * This write is needed to ensure that a | |
6c0d46c4 | 2129 | * rw->ro transition does not occur between |
fb1cc555 AV |
2130 | * the time when the file is created and when |
2131 | * a permanent write count is taken through | |
2132 | * the 'struct file' in nameidata_to_filp(). | |
2133 | */ | |
2134 | error = mnt_want_write(nd->path.mnt); | |
2135 | if (error) | |
2136 | goto exit_mutex_unlock; | |
ca344a89 | 2137 | want_write = 1; |
9b44f1b3 | 2138 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 2139 | open_flag &= ~O_TRUNC; |
6c0d46c4 | 2140 | will_truncate = 0; |
bcda7652 | 2141 | acc_mode = MAY_OPEN; |
6c0d46c4 AV |
2142 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2143 | if (error) | |
2144 | goto exit_mutex_unlock; | |
2145 | error = vfs_create(dir->d_inode, dentry, mode, nd); | |
2146 | if (error) | |
2147 | goto exit_mutex_unlock; | |
2148 | mutex_unlock(&dir->d_inode->i_mutex); | |
2149 | dput(nd->path.dentry); | |
2150 | nd->path.dentry = dentry; | |
ca344a89 | 2151 | goto common; |
fb1cc555 AV |
2152 | } |
2153 | ||
2154 | /* | |
2155 | * It already exists. | |
2156 | */ | |
2157 | mutex_unlock(&dir->d_inode->i_mutex); | |
2158 | audit_inode(pathname, path->dentry); | |
2159 | ||
2160 | error = -EEXIST; | |
ca344a89 | 2161 | if (open_flag & O_EXCL) |
fb1cc555 AV |
2162 | goto exit_dput; |
2163 | ||
9875cf80 DH |
2164 | error = follow_managed(path, nd->flags); |
2165 | if (error < 0) | |
2166 | goto exit_dput; | |
fb1cc555 AV |
2167 | |
2168 | error = -ENOENT; | |
2169 | if (!path->dentry->d_inode) | |
2170 | goto exit_dput; | |
9e67f361 AV |
2171 | |
2172 | if (path->dentry->d_inode->i_op->follow_link) | |
fb1cc555 | 2173 | return NULL; |
fb1cc555 AV |
2174 | |
2175 | path_to_nameidata(path, nd); | |
31e6b01f | 2176 | nd->inode = path->dentry->d_inode; |
fb1cc555 | 2177 | error = -EISDIR; |
31e6b01f | 2178 | if (S_ISDIR(nd->inode->i_mode)) |
fb1cc555 | 2179 | goto exit; |
67ee3ad2 | 2180 | ok: |
6c0d46c4 AV |
2181 | if (!S_ISREG(nd->inode->i_mode)) |
2182 | will_truncate = 0; | |
2183 | ||
0f9d1a10 AV |
2184 | if (will_truncate) { |
2185 | error = mnt_want_write(nd->path.mnt); | |
2186 | if (error) | |
2187 | goto exit; | |
ca344a89 | 2188 | want_write = 1; |
0f9d1a10 | 2189 | } |
ca344a89 | 2190 | common: |
bcda7652 | 2191 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 2192 | if (error) |
0f9d1a10 | 2193 | goto exit; |
0f9d1a10 AV |
2194 | filp = nameidata_to_filp(nd); |
2195 | if (!IS_ERR(filp)) { | |
2196 | error = ima_file_check(filp, op->acc_mode); | |
2197 | if (error) { | |
2198 | fput(filp); | |
2199 | filp = ERR_PTR(error); | |
2200 | } | |
2201 | } | |
2202 | if (!IS_ERR(filp)) { | |
2203 | if (will_truncate) { | |
2204 | error = handle_truncate(filp); | |
2205 | if (error) { | |
2206 | fput(filp); | |
2207 | filp = ERR_PTR(error); | |
2208 | } | |
2209 | } | |
2210 | } | |
ca344a89 AV |
2211 | out: |
2212 | if (want_write) | |
0f9d1a10 AV |
2213 | mnt_drop_write(nd->path.mnt); |
2214 | path_put(&nd->path); | |
fb1cc555 AV |
2215 | return filp; |
2216 | ||
2217 | exit_mutex_unlock: | |
2218 | mutex_unlock(&dir->d_inode->i_mutex); | |
2219 | exit_dput: | |
2220 | path_put_conditional(path, nd); | |
2221 | exit: | |
ca344a89 AV |
2222 | filp = ERR_PTR(error); |
2223 | goto out; | |
fb1cc555 AV |
2224 | } |
2225 | ||
13aab428 | 2226 | static struct file *path_openat(int dfd, const char *pathname, |
73d049a4 | 2227 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 2228 | { |
fe2d35ff | 2229 | struct file *base = NULL; |
4a3fd211 | 2230 | struct file *filp; |
9850c056 | 2231 | struct path path; |
13aab428 | 2232 | int error; |
31e6b01f NP |
2233 | |
2234 | filp = get_empty_filp(); | |
2235 | if (!filp) | |
2236 | return ERR_PTR(-ENFILE); | |
2237 | ||
47c805dc | 2238 | filp->f_flags = op->open_flag; |
73d049a4 AV |
2239 | nd->intent.open.file = filp; |
2240 | nd->intent.open.flags = open_to_namei_flags(op->open_flag); | |
2241 | nd->intent.open.create_mode = op->mode; | |
31e6b01f | 2242 | |
73d049a4 | 2243 | error = path_init(dfd, pathname, flags | LOOKUP_PARENT, nd, &base); |
31e6b01f | 2244 | if (unlikely(error)) |
13aab428 | 2245 | goto out_filp; |
31e6b01f | 2246 | |
fe2d35ff | 2247 | current->total_link_count = 0; |
73d049a4 | 2248 | error = link_path_walk(pathname, nd); |
31e6b01f NP |
2249 | if (unlikely(error)) |
2250 | goto out_filp; | |
1da177e4 | 2251 | |
73d049a4 | 2252 | filp = do_last(nd, &path, op, pathname); |
806b681c | 2253 | while (unlikely(!filp)) { /* trailing symlink */ |
7b9337aa | 2254 | struct path link = path; |
def4af30 | 2255 | void *cookie; |
574197e0 | 2256 | if (!(nd->flags & LOOKUP_FOLLOW)) { |
73d049a4 AV |
2257 | path_put_conditional(&path, nd); |
2258 | path_put(&nd->path); | |
40b39136 AV |
2259 | filp = ERR_PTR(-ELOOP); |
2260 | break; | |
2261 | } | |
73d049a4 AV |
2262 | nd->flags |= LOOKUP_PARENT; |
2263 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
574197e0 | 2264 | error = follow_link(&link, nd, &cookie); |
c3e380b0 | 2265 | if (unlikely(error)) |
f1afe9ef | 2266 | filp = ERR_PTR(error); |
c3e380b0 | 2267 | else |
73d049a4 | 2268 | filp = do_last(nd, &path, op, pathname); |
574197e0 | 2269 | put_link(nd, &link, cookie); |
806b681c | 2270 | } |
10fa8e62 | 2271 | out: |
73d049a4 AV |
2272 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) |
2273 | path_put(&nd->root); | |
fe2d35ff AV |
2274 | if (base) |
2275 | fput(base); | |
73d049a4 | 2276 | release_open_intent(nd); |
10fa8e62 | 2277 | return filp; |
1da177e4 | 2278 | |
31e6b01f | 2279 | out_filp: |
806b681c | 2280 | filp = ERR_PTR(error); |
10fa8e62 | 2281 | goto out; |
1da177e4 LT |
2282 | } |
2283 | ||
13aab428 AV |
2284 | struct file *do_filp_open(int dfd, const char *pathname, |
2285 | const struct open_flags *op, int flags) | |
2286 | { | |
73d049a4 | 2287 | struct nameidata nd; |
13aab428 AV |
2288 | struct file *filp; |
2289 | ||
73d049a4 | 2290 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 2291 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 2292 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 2293 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 2294 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
2295 | return filp; |
2296 | } | |
2297 | ||
73d049a4 AV |
2298 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
2299 | const char *name, const struct open_flags *op, int flags) | |
2300 | { | |
2301 | struct nameidata nd; | |
2302 | struct file *file; | |
2303 | ||
2304 | nd.root.mnt = mnt; | |
2305 | nd.root.dentry = dentry; | |
2306 | ||
2307 | flags |= LOOKUP_ROOT; | |
2308 | ||
bcda7652 | 2309 | if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
2310 | return ERR_PTR(-ELOOP); |
2311 | ||
2312 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU); | |
2313 | if (unlikely(file == ERR_PTR(-ECHILD))) | |
2314 | file = path_openat(-1, name, &nd, op, flags); | |
2315 | if (unlikely(file == ERR_PTR(-ESTALE))) | |
2316 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL); | |
2317 | return file; | |
2318 | } | |
2319 | ||
1da177e4 LT |
2320 | /** |
2321 | * lookup_create - lookup a dentry, creating it if it doesn't exist | |
2322 | * @nd: nameidata info | |
2323 | * @is_dir: directory flag | |
2324 | * | |
2325 | * Simple function to lookup and return a dentry and create it | |
2326 | * if it doesn't exist. Is SMP-safe. | |
c663e5d8 | 2327 | * |
4ac91378 | 2328 | * Returns with nd->path.dentry->d_inode->i_mutex locked. |
1da177e4 LT |
2329 | */ |
2330 | struct dentry *lookup_create(struct nameidata *nd, int is_dir) | |
2331 | { | |
c663e5d8 | 2332 | struct dentry *dentry = ERR_PTR(-EEXIST); |
1da177e4 | 2333 | |
4ac91378 | 2334 | mutex_lock_nested(&nd->path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
c663e5d8 CH |
2335 | /* |
2336 | * Yucky last component or no last component at all? | |
2337 | * (foo/., foo/.., /////) | |
2338 | */ | |
1da177e4 LT |
2339 | if (nd->last_type != LAST_NORM) |
2340 | goto fail; | |
2341 | nd->flags &= ~LOOKUP_PARENT; | |
3516586a | 2342 | nd->flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
a634904a | 2343 | nd->intent.open.flags = O_EXCL; |
c663e5d8 CH |
2344 | |
2345 | /* | |
2346 | * Do the final lookup. | |
2347 | */ | |
49705b77 | 2348 | dentry = lookup_hash(nd); |
1da177e4 LT |
2349 | if (IS_ERR(dentry)) |
2350 | goto fail; | |
c663e5d8 | 2351 | |
e9baf6e5 AV |
2352 | if (dentry->d_inode) |
2353 | goto eexist; | |
c663e5d8 CH |
2354 | /* |
2355 | * Special case - lookup gave negative, but... we had foo/bar/ | |
2356 | * From the vfs_mknod() POV we just have a negative dentry - | |
2357 | * all is fine. Let's be bastards - you had / on the end, you've | |
2358 | * been asking for (non-existent) directory. -ENOENT for you. | |
2359 | */ | |
e9baf6e5 AV |
2360 | if (unlikely(!is_dir && nd->last.name[nd->last.len])) { |
2361 | dput(dentry); | |
2362 | dentry = ERR_PTR(-ENOENT); | |
2363 | } | |
1da177e4 | 2364 | return dentry; |
e9baf6e5 | 2365 | eexist: |
1da177e4 | 2366 | dput(dentry); |
e9baf6e5 | 2367 | dentry = ERR_PTR(-EEXIST); |
1da177e4 LT |
2368 | fail: |
2369 | return dentry; | |
2370 | } | |
f81a0bff | 2371 | EXPORT_SYMBOL_GPL(lookup_create); |
1da177e4 LT |
2372 | |
2373 | int vfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) | |
2374 | { | |
a95164d9 | 2375 | int error = may_create(dir, dentry); |
1da177e4 LT |
2376 | |
2377 | if (error) | |
2378 | return error; | |
2379 | ||
e795b717 SH |
2380 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && |
2381 | !ns_capable(inode_userns(dir), CAP_MKNOD)) | |
1da177e4 LT |
2382 | return -EPERM; |
2383 | ||
acfa4380 | 2384 | if (!dir->i_op->mknod) |
1da177e4 LT |
2385 | return -EPERM; |
2386 | ||
08ce5f16 SH |
2387 | error = devcgroup_inode_mknod(mode, dev); |
2388 | if (error) | |
2389 | return error; | |
2390 | ||
1da177e4 LT |
2391 | error = security_inode_mknod(dir, dentry, mode, dev); |
2392 | if (error) | |
2393 | return error; | |
2394 | ||
1da177e4 | 2395 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 2396 | if (!error) |
f38aa942 | 2397 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2398 | return error; |
2399 | } | |
2400 | ||
463c3197 DH |
2401 | static int may_mknod(mode_t mode) |
2402 | { | |
2403 | switch (mode & S_IFMT) { | |
2404 | case S_IFREG: | |
2405 | case S_IFCHR: | |
2406 | case S_IFBLK: | |
2407 | case S_IFIFO: | |
2408 | case S_IFSOCK: | |
2409 | case 0: /* zero mode translates to S_IFREG */ | |
2410 | return 0; | |
2411 | case S_IFDIR: | |
2412 | return -EPERM; | |
2413 | default: | |
2414 | return -EINVAL; | |
2415 | } | |
2416 | } | |
2417 | ||
2e4d0924 HC |
2418 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, int, mode, |
2419 | unsigned, dev) | |
1da177e4 | 2420 | { |
2ad94ae6 AV |
2421 | int error; |
2422 | char *tmp; | |
2423 | struct dentry *dentry; | |
1da177e4 LT |
2424 | struct nameidata nd; |
2425 | ||
2426 | if (S_ISDIR(mode)) | |
2427 | return -EPERM; | |
1da177e4 | 2428 | |
2ad94ae6 | 2429 | error = user_path_parent(dfd, filename, &nd, &tmp); |
1da177e4 | 2430 | if (error) |
2ad94ae6 AV |
2431 | return error; |
2432 | ||
1da177e4 | 2433 | dentry = lookup_create(&nd, 0); |
463c3197 DH |
2434 | if (IS_ERR(dentry)) { |
2435 | error = PTR_ERR(dentry); | |
2436 | goto out_unlock; | |
2437 | } | |
4ac91378 | 2438 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2439 | mode &= ~current_umask(); |
463c3197 DH |
2440 | error = may_mknod(mode); |
2441 | if (error) | |
2442 | goto out_dput; | |
2443 | error = mnt_want_write(nd.path.mnt); | |
2444 | if (error) | |
2445 | goto out_dput; | |
be6d3e56 KT |
2446 | error = security_path_mknod(&nd.path, dentry, mode, dev); |
2447 | if (error) | |
2448 | goto out_drop_write; | |
463c3197 | 2449 | switch (mode & S_IFMT) { |
1da177e4 | 2450 | case 0: case S_IFREG: |
4ac91378 | 2451 | error = vfs_create(nd.path.dentry->d_inode,dentry,mode,&nd); |
1da177e4 LT |
2452 | break; |
2453 | case S_IFCHR: case S_IFBLK: | |
4ac91378 | 2454 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2455 | new_decode_dev(dev)); |
2456 | break; | |
2457 | case S_IFIFO: case S_IFSOCK: | |
4ac91378 | 2458 | error = vfs_mknod(nd.path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2459 | break; |
1da177e4 | 2460 | } |
be6d3e56 | 2461 | out_drop_write: |
463c3197 DH |
2462 | mnt_drop_write(nd.path.mnt); |
2463 | out_dput: | |
2464 | dput(dentry); | |
2465 | out_unlock: | |
4ac91378 | 2466 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2467 | path_put(&nd.path); |
1da177e4 LT |
2468 | putname(tmp); |
2469 | ||
2470 | return error; | |
2471 | } | |
2472 | ||
3480b257 | 2473 | SYSCALL_DEFINE3(mknod, const char __user *, filename, int, mode, unsigned, dev) |
5590ff0d UD |
2474 | { |
2475 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2476 | } | |
2477 | ||
1da177e4 LT |
2478 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2479 | { | |
a95164d9 | 2480 | int error = may_create(dir, dentry); |
1da177e4 LT |
2481 | |
2482 | if (error) | |
2483 | return error; | |
2484 | ||
acfa4380 | 2485 | if (!dir->i_op->mkdir) |
1da177e4 LT |
2486 | return -EPERM; |
2487 | ||
2488 | mode &= (S_IRWXUGO|S_ISVTX); | |
2489 | error = security_inode_mkdir(dir, dentry, mode); | |
2490 | if (error) | |
2491 | return error; | |
2492 | ||
1da177e4 | 2493 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 2494 | if (!error) |
f38aa942 | 2495 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
2496 | return error; |
2497 | } | |
2498 | ||
2e4d0924 | 2499 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, int, mode) |
1da177e4 LT |
2500 | { |
2501 | int error = 0; | |
2502 | char * tmp; | |
6902d925 DH |
2503 | struct dentry *dentry; |
2504 | struct nameidata nd; | |
1da177e4 | 2505 | |
2ad94ae6 AV |
2506 | error = user_path_parent(dfd, pathname, &nd, &tmp); |
2507 | if (error) | |
6902d925 | 2508 | goto out_err; |
1da177e4 | 2509 | |
6902d925 DH |
2510 | dentry = lookup_create(&nd, 1); |
2511 | error = PTR_ERR(dentry); | |
2512 | if (IS_ERR(dentry)) | |
2513 | goto out_unlock; | |
1da177e4 | 2514 | |
4ac91378 | 2515 | if (!IS_POSIXACL(nd.path.dentry->d_inode)) |
ce3b0f8d | 2516 | mode &= ~current_umask(); |
463c3197 DH |
2517 | error = mnt_want_write(nd.path.mnt); |
2518 | if (error) | |
2519 | goto out_dput; | |
be6d3e56 KT |
2520 | error = security_path_mkdir(&nd.path, dentry, mode); |
2521 | if (error) | |
2522 | goto out_drop_write; | |
4ac91378 | 2523 | error = vfs_mkdir(nd.path.dentry->d_inode, dentry, mode); |
be6d3e56 | 2524 | out_drop_write: |
463c3197 DH |
2525 | mnt_drop_write(nd.path.mnt); |
2526 | out_dput: | |
6902d925 DH |
2527 | dput(dentry); |
2528 | out_unlock: | |
4ac91378 | 2529 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2530 | path_put(&nd.path); |
6902d925 DH |
2531 | putname(tmp); |
2532 | out_err: | |
1da177e4 LT |
2533 | return error; |
2534 | } | |
2535 | ||
3cdad428 | 2536 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, int, mode) |
5590ff0d UD |
2537 | { |
2538 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
2539 | } | |
2540 | ||
1da177e4 | 2541 | /* |
a71905f0 SW |
2542 | * The dentry_unhash() helper will try to drop the dentry early: we |
2543 | * should have a usage count of 2 if we're the only user of this | |
2544 | * dentry, and if that is true (possibly after pruning the dcache), | |
2545 | * then we drop the dentry now. | |
1da177e4 LT |
2546 | * |
2547 | * A low-level filesystem can, if it choses, legally | |
2548 | * do a | |
2549 | * | |
2550 | * if (!d_unhashed(dentry)) | |
2551 | * return -EBUSY; | |
2552 | * | |
2553 | * if it cannot handle the case of removing a directory | |
2554 | * that is still in use by something else.. | |
2555 | */ | |
2556 | void dentry_unhash(struct dentry *dentry) | |
2557 | { | |
dc168427 | 2558 | shrink_dcache_parent(dentry); |
1da177e4 | 2559 | spin_lock(&dentry->d_lock); |
64252c75 | 2560 | if (dentry->d_count == 1) |
1da177e4 LT |
2561 | __d_drop(dentry); |
2562 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
2563 | } |
2564 | ||
2565 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
2566 | { | |
2567 | int error = may_delete(dir, dentry, 1); | |
2568 | ||
2569 | if (error) | |
2570 | return error; | |
2571 | ||
acfa4380 | 2572 | if (!dir->i_op->rmdir) |
1da177e4 LT |
2573 | return -EPERM; |
2574 | ||
1b1dcc1b | 2575 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
2576 | |
2577 | error = -EBUSY; | |
1da177e4 | 2578 | if (d_mountpoint(dentry)) |
912dbc15 SW |
2579 | goto out; |
2580 | ||
2581 | error = security_inode_rmdir(dir, dentry); | |
2582 | if (error) | |
2583 | goto out; | |
2584 | ||
3cebde24 | 2585 | shrink_dcache_parent(dentry); |
912dbc15 SW |
2586 | error = dir->i_op->rmdir(dir, dentry); |
2587 | if (error) | |
2588 | goto out; | |
2589 | ||
2590 | dentry->d_inode->i_flags |= S_DEAD; | |
2591 | dont_mount(dentry); | |
2592 | ||
2593 | out: | |
1b1dcc1b | 2594 | mutex_unlock(&dentry->d_inode->i_mutex); |
912dbc15 | 2595 | if (!error) |
1da177e4 | 2596 | d_delete(dentry); |
1da177e4 LT |
2597 | return error; |
2598 | } | |
2599 | ||
5590ff0d | 2600 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
2601 | { |
2602 | int error = 0; | |
2603 | char * name; | |
2604 | struct dentry *dentry; | |
2605 | struct nameidata nd; | |
2606 | ||
2ad94ae6 | 2607 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2608 | if (error) |
2ad94ae6 | 2609 | return error; |
1da177e4 LT |
2610 | |
2611 | switch(nd.last_type) { | |
0612d9fb OH |
2612 | case LAST_DOTDOT: |
2613 | error = -ENOTEMPTY; | |
2614 | goto exit1; | |
2615 | case LAST_DOT: | |
2616 | error = -EINVAL; | |
2617 | goto exit1; | |
2618 | case LAST_ROOT: | |
2619 | error = -EBUSY; | |
2620 | goto exit1; | |
1da177e4 | 2621 | } |
0612d9fb OH |
2622 | |
2623 | nd.flags &= ~LOOKUP_PARENT; | |
2624 | ||
4ac91378 | 2625 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2626 | dentry = lookup_hash(&nd); |
1da177e4 | 2627 | error = PTR_ERR(dentry); |
6902d925 DH |
2628 | if (IS_ERR(dentry)) |
2629 | goto exit2; | |
e6bc45d6 TT |
2630 | if (!dentry->d_inode) { |
2631 | error = -ENOENT; | |
2632 | goto exit3; | |
2633 | } | |
0622753b DH |
2634 | error = mnt_want_write(nd.path.mnt); |
2635 | if (error) | |
2636 | goto exit3; | |
be6d3e56 KT |
2637 | error = security_path_rmdir(&nd.path, dentry); |
2638 | if (error) | |
2639 | goto exit4; | |
4ac91378 | 2640 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2641 | exit4: |
0622753b DH |
2642 | mnt_drop_write(nd.path.mnt); |
2643 | exit3: | |
6902d925 DH |
2644 | dput(dentry); |
2645 | exit2: | |
4ac91378 | 2646 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2647 | exit1: |
1d957f9b | 2648 | path_put(&nd.path); |
1da177e4 LT |
2649 | putname(name); |
2650 | return error; | |
2651 | } | |
2652 | ||
3cdad428 | 2653 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
2654 | { |
2655 | return do_rmdir(AT_FDCWD, pathname); | |
2656 | } | |
2657 | ||
1da177e4 LT |
2658 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
2659 | { | |
2660 | int error = may_delete(dir, dentry, 0); | |
2661 | ||
2662 | if (error) | |
2663 | return error; | |
2664 | ||
acfa4380 | 2665 | if (!dir->i_op->unlink) |
1da177e4 LT |
2666 | return -EPERM; |
2667 | ||
1b1dcc1b | 2668 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2669 | if (d_mountpoint(dentry)) |
2670 | error = -EBUSY; | |
2671 | else { | |
2672 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 2673 | if (!error) { |
1da177e4 | 2674 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 2675 | if (!error) |
d83c49f3 | 2676 | dont_mount(dentry); |
bec1052e | 2677 | } |
1da177e4 | 2678 | } |
1b1dcc1b | 2679 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
2680 | |
2681 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
2682 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 2683 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 2684 | d_delete(dentry); |
1da177e4 | 2685 | } |
0eeca283 | 2686 | |
1da177e4 LT |
2687 | return error; |
2688 | } | |
2689 | ||
2690 | /* | |
2691 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 2692 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
2693 | * writeout happening, and we don't want to prevent access to the directory |
2694 | * while waiting on the I/O. | |
2695 | */ | |
5590ff0d | 2696 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 2697 | { |
2ad94ae6 AV |
2698 | int error; |
2699 | char *name; | |
1da177e4 LT |
2700 | struct dentry *dentry; |
2701 | struct nameidata nd; | |
2702 | struct inode *inode = NULL; | |
2703 | ||
2ad94ae6 | 2704 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 2705 | if (error) |
2ad94ae6 AV |
2706 | return error; |
2707 | ||
1da177e4 LT |
2708 | error = -EISDIR; |
2709 | if (nd.last_type != LAST_NORM) | |
2710 | goto exit1; | |
0612d9fb OH |
2711 | |
2712 | nd.flags &= ~LOOKUP_PARENT; | |
2713 | ||
4ac91378 | 2714 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 2715 | dentry = lookup_hash(&nd); |
1da177e4 LT |
2716 | error = PTR_ERR(dentry); |
2717 | if (!IS_ERR(dentry)) { | |
2718 | /* Why not before? Because we want correct error value */ | |
50338b88 TE |
2719 | if (nd.last.name[nd.last.len]) |
2720 | goto slashes; | |
1da177e4 | 2721 | inode = dentry->d_inode; |
50338b88 | 2722 | if (!inode) |
e6bc45d6 TT |
2723 | goto slashes; |
2724 | ihold(inode); | |
0622753b DH |
2725 | error = mnt_want_write(nd.path.mnt); |
2726 | if (error) | |
2727 | goto exit2; | |
be6d3e56 KT |
2728 | error = security_path_unlink(&nd.path, dentry); |
2729 | if (error) | |
2730 | goto exit3; | |
4ac91378 | 2731 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 2732 | exit3: |
0622753b | 2733 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
2734 | exit2: |
2735 | dput(dentry); | |
2736 | } | |
4ac91378 | 2737 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
2738 | if (inode) |
2739 | iput(inode); /* truncate the inode here */ | |
2740 | exit1: | |
1d957f9b | 2741 | path_put(&nd.path); |
1da177e4 LT |
2742 | putname(name); |
2743 | return error; | |
2744 | ||
2745 | slashes: | |
2746 | error = !dentry->d_inode ? -ENOENT : | |
2747 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
2748 | goto exit2; | |
2749 | } | |
2750 | ||
2e4d0924 | 2751 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
2752 | { |
2753 | if ((flag & ~AT_REMOVEDIR) != 0) | |
2754 | return -EINVAL; | |
2755 | ||
2756 | if (flag & AT_REMOVEDIR) | |
2757 | return do_rmdir(dfd, pathname); | |
2758 | ||
2759 | return do_unlinkat(dfd, pathname); | |
2760 | } | |
2761 | ||
3480b257 | 2762 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
2763 | { |
2764 | return do_unlinkat(AT_FDCWD, pathname); | |
2765 | } | |
2766 | ||
db2e747b | 2767 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 2768 | { |
a95164d9 | 2769 | int error = may_create(dir, dentry); |
1da177e4 LT |
2770 | |
2771 | if (error) | |
2772 | return error; | |
2773 | ||
acfa4380 | 2774 | if (!dir->i_op->symlink) |
1da177e4 LT |
2775 | return -EPERM; |
2776 | ||
2777 | error = security_inode_symlink(dir, dentry, oldname); | |
2778 | if (error) | |
2779 | return error; | |
2780 | ||
1da177e4 | 2781 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 2782 | if (!error) |
f38aa942 | 2783 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2784 | return error; |
2785 | } | |
2786 | ||
2e4d0924 HC |
2787 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
2788 | int, newdfd, const char __user *, newname) | |
1da177e4 | 2789 | { |
2ad94ae6 AV |
2790 | int error; |
2791 | char *from; | |
2792 | char *to; | |
6902d925 DH |
2793 | struct dentry *dentry; |
2794 | struct nameidata nd; | |
1da177e4 LT |
2795 | |
2796 | from = getname(oldname); | |
2ad94ae6 | 2797 | if (IS_ERR(from)) |
1da177e4 | 2798 | return PTR_ERR(from); |
1da177e4 | 2799 | |
2ad94ae6 | 2800 | error = user_path_parent(newdfd, newname, &nd, &to); |
6902d925 | 2801 | if (error) |
2ad94ae6 AV |
2802 | goto out_putname; |
2803 | ||
6902d925 DH |
2804 | dentry = lookup_create(&nd, 0); |
2805 | error = PTR_ERR(dentry); | |
2806 | if (IS_ERR(dentry)) | |
2807 | goto out_unlock; | |
2808 | ||
75c3f29d DH |
2809 | error = mnt_want_write(nd.path.mnt); |
2810 | if (error) | |
2811 | goto out_dput; | |
be6d3e56 KT |
2812 | error = security_path_symlink(&nd.path, dentry, from); |
2813 | if (error) | |
2814 | goto out_drop_write; | |
db2e747b | 2815 | error = vfs_symlink(nd.path.dentry->d_inode, dentry, from); |
be6d3e56 | 2816 | out_drop_write: |
75c3f29d DH |
2817 | mnt_drop_write(nd.path.mnt); |
2818 | out_dput: | |
6902d925 DH |
2819 | dput(dentry); |
2820 | out_unlock: | |
4ac91378 | 2821 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1d957f9b | 2822 | path_put(&nd.path); |
6902d925 DH |
2823 | putname(to); |
2824 | out_putname: | |
1da177e4 LT |
2825 | putname(from); |
2826 | return error; | |
2827 | } | |
2828 | ||
3480b257 | 2829 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
2830 | { |
2831 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
2832 | } | |
2833 | ||
1da177e4 LT |
2834 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
2835 | { | |
2836 | struct inode *inode = old_dentry->d_inode; | |
2837 | int error; | |
2838 | ||
2839 | if (!inode) | |
2840 | return -ENOENT; | |
2841 | ||
a95164d9 | 2842 | error = may_create(dir, new_dentry); |
1da177e4 LT |
2843 | if (error) |
2844 | return error; | |
2845 | ||
2846 | if (dir->i_sb != inode->i_sb) | |
2847 | return -EXDEV; | |
2848 | ||
2849 | /* | |
2850 | * A link to an append-only or immutable file cannot be created. | |
2851 | */ | |
2852 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
2853 | return -EPERM; | |
acfa4380 | 2854 | if (!dir->i_op->link) |
1da177e4 | 2855 | return -EPERM; |
7e79eedb | 2856 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
2857 | return -EPERM; |
2858 | ||
2859 | error = security_inode_link(old_dentry, dir, new_dentry); | |
2860 | if (error) | |
2861 | return error; | |
2862 | ||
7e79eedb | 2863 | mutex_lock(&inode->i_mutex); |
aae8a97d AK |
2864 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
2865 | if (inode->i_nlink == 0) | |
2866 | error = -ENOENT; | |
2867 | else | |
2868 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 2869 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 2870 | if (!error) |
7e79eedb | 2871 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
2872 | return error; |
2873 | } | |
2874 | ||
2875 | /* | |
2876 | * Hardlinks are often used in delicate situations. We avoid | |
2877 | * security-related surprises by not following symlinks on the | |
2878 | * newname. --KAB | |
2879 | * | |
2880 | * We don't follow them on the oldname either to be compatible | |
2881 | * with linux 2.0, and to avoid hard-linking to directories | |
2882 | * and other special files. --ADM | |
2883 | */ | |
2e4d0924 HC |
2884 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
2885 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
2886 | { |
2887 | struct dentry *new_dentry; | |
2d8f3038 AV |
2888 | struct nameidata nd; |
2889 | struct path old_path; | |
11a7b371 | 2890 | int how = 0; |
1da177e4 | 2891 | int error; |
2ad94ae6 | 2892 | char *to; |
1da177e4 | 2893 | |
11a7b371 | 2894 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 2895 | return -EINVAL; |
11a7b371 AK |
2896 | /* |
2897 | * To use null names we require CAP_DAC_READ_SEARCH | |
2898 | * This ensures that not everyone will be able to create | |
2899 | * handlink using the passed filedescriptor. | |
2900 | */ | |
2901 | if (flags & AT_EMPTY_PATH) { | |
2902 | if (!capable(CAP_DAC_READ_SEARCH)) | |
2903 | return -ENOENT; | |
2904 | how = LOOKUP_EMPTY; | |
2905 | } | |
2906 | ||
2907 | if (flags & AT_SYMLINK_FOLLOW) | |
2908 | how |= LOOKUP_FOLLOW; | |
c04030e1 | 2909 | |
11a7b371 | 2910 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 2911 | if (error) |
2ad94ae6 AV |
2912 | return error; |
2913 | ||
2914 | error = user_path_parent(newdfd, newname, &nd, &to); | |
1da177e4 LT |
2915 | if (error) |
2916 | goto out; | |
2917 | error = -EXDEV; | |
2d8f3038 | 2918 | if (old_path.mnt != nd.path.mnt) |
1da177e4 LT |
2919 | goto out_release; |
2920 | new_dentry = lookup_create(&nd, 0); | |
2921 | error = PTR_ERR(new_dentry); | |
6902d925 DH |
2922 | if (IS_ERR(new_dentry)) |
2923 | goto out_unlock; | |
75c3f29d DH |
2924 | error = mnt_want_write(nd.path.mnt); |
2925 | if (error) | |
2926 | goto out_dput; | |
be6d3e56 KT |
2927 | error = security_path_link(old_path.dentry, &nd.path, new_dentry); |
2928 | if (error) | |
2929 | goto out_drop_write; | |
2d8f3038 | 2930 | error = vfs_link(old_path.dentry, nd.path.dentry->d_inode, new_dentry); |
be6d3e56 | 2931 | out_drop_write: |
75c3f29d DH |
2932 | mnt_drop_write(nd.path.mnt); |
2933 | out_dput: | |
6902d925 DH |
2934 | dput(new_dentry); |
2935 | out_unlock: | |
4ac91378 | 2936 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 2937 | out_release: |
1d957f9b | 2938 | path_put(&nd.path); |
2ad94ae6 | 2939 | putname(to); |
1da177e4 | 2940 | out: |
2d8f3038 | 2941 | path_put(&old_path); |
1da177e4 LT |
2942 | |
2943 | return error; | |
2944 | } | |
2945 | ||
3480b257 | 2946 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 2947 | { |
c04030e1 | 2948 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
2949 | } |
2950 | ||
1da177e4 LT |
2951 | /* |
2952 | * The worst of all namespace operations - renaming directory. "Perverted" | |
2953 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
2954 | * Problems: | |
2955 | * a) we can get into loop creation. Check is done in is_subdir(). | |
2956 | * b) race potential - two innocent renames can create a loop together. | |
2957 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 2958 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
2959 | * story. |
2960 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 2961 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
2962 | * whether the target exists). Solution: try to be smart with locking |
2963 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 2964 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
2965 | * move will be locked. Thus we can rank directories by the tree |
2966 | * (ancestors first) and rank all non-directories after them. | |
2967 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 2968 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
2969 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
2970 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
2971 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 2972 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 2973 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 2974 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 2975 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
2976 | * locking]. |
2977 | */ | |
75c96f85 AB |
2978 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
2979 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
2980 | { |
2981 | int error = 0; | |
9055cba7 | 2982 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
2983 | |
2984 | /* | |
2985 | * If we are going to change the parent - check write permissions, | |
2986 | * we'll need to flip '..'. | |
2987 | */ | |
2988 | if (new_dir != old_dir) { | |
f419a2e3 | 2989 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
2990 | if (error) |
2991 | return error; | |
2992 | } | |
2993 | ||
2994 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
2995 | if (error) | |
2996 | return error; | |
2997 | ||
d83c49f3 | 2998 | if (target) |
1b1dcc1b | 2999 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
3000 | |
3001 | error = -EBUSY; | |
3002 | if (d_mountpoint(old_dentry) || d_mountpoint(new_dentry)) | |
3003 | goto out; | |
3004 | ||
3cebde24 SW |
3005 | if (target) |
3006 | shrink_dcache_parent(new_dentry); | |
9055cba7 SW |
3007 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
3008 | if (error) | |
3009 | goto out; | |
3010 | ||
1da177e4 | 3011 | if (target) { |
9055cba7 SW |
3012 | target->i_flags |= S_DEAD; |
3013 | dont_mount(new_dentry); | |
1da177e4 | 3014 | } |
9055cba7 SW |
3015 | out: |
3016 | if (target) | |
3017 | mutex_unlock(&target->i_mutex); | |
e31e14ec | 3018 | if (!error) |
349457cc MF |
3019 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3020 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3021 | return error; |
3022 | } | |
3023 | ||
75c96f85 AB |
3024 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3025 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 | 3026 | { |
51892bbb | 3027 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3028 | int error; |
3029 | ||
3030 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3031 | if (error) | |
3032 | return error; | |
3033 | ||
3034 | dget(new_dentry); | |
1da177e4 | 3035 | if (target) |
1b1dcc1b | 3036 | mutex_lock(&target->i_mutex); |
51892bbb SW |
3037 | |
3038 | error = -EBUSY; | |
1da177e4 | 3039 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
51892bbb SW |
3040 | goto out; |
3041 | ||
3042 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3043 | if (error) | |
3044 | goto out; | |
3045 | ||
3046 | if (target) | |
3047 | dont_mount(new_dentry); | |
3048 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) | |
3049 | d_move(old_dentry, new_dentry); | |
3050 | out: | |
1da177e4 | 3051 | if (target) |
1b1dcc1b | 3052 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3053 | dput(new_dentry); |
3054 | return error; | |
3055 | } | |
3056 | ||
3057 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3058 | struct inode *new_dir, struct dentry *new_dentry) | |
3059 | { | |
3060 | int error; | |
3061 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3062 | const unsigned char *old_name; |
1da177e4 LT |
3063 | |
3064 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3065 | return 0; | |
3066 | ||
3067 | error = may_delete(old_dir, old_dentry, is_dir); | |
3068 | if (error) | |
3069 | return error; | |
3070 | ||
3071 | if (!new_dentry->d_inode) | |
a95164d9 | 3072 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3073 | else |
3074 | error = may_delete(new_dir, new_dentry, is_dir); | |
3075 | if (error) | |
3076 | return error; | |
3077 | ||
acfa4380 | 3078 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3079 | return -EPERM; |
3080 | ||
0eeca283 RL |
3081 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3082 | ||
1da177e4 LT |
3083 | if (is_dir) |
3084 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3085 | else | |
3086 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3087 | if (!error) |
3088 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3089 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3090 | fsnotify_oldname_free(old_name); |
3091 | ||
1da177e4 LT |
3092 | return error; |
3093 | } | |
3094 | ||
2e4d0924 HC |
3095 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3096 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3097 | { |
2ad94ae6 AV |
3098 | struct dentry *old_dir, *new_dir; |
3099 | struct dentry *old_dentry, *new_dentry; | |
3100 | struct dentry *trap; | |
1da177e4 | 3101 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3102 | char *from; |
3103 | char *to; | |
3104 | int error; | |
1da177e4 | 3105 | |
2ad94ae6 | 3106 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3107 | if (error) |
3108 | goto exit; | |
3109 | ||
2ad94ae6 | 3110 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3111 | if (error) |
3112 | goto exit1; | |
3113 | ||
3114 | error = -EXDEV; | |
4ac91378 | 3115 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3116 | goto exit2; |
3117 | ||
4ac91378 | 3118 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3119 | error = -EBUSY; |
3120 | if (oldnd.last_type != LAST_NORM) | |
3121 | goto exit2; | |
3122 | ||
4ac91378 | 3123 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3124 | if (newnd.last_type != LAST_NORM) |
3125 | goto exit2; | |
3126 | ||
0612d9fb OH |
3127 | oldnd.flags &= ~LOOKUP_PARENT; |
3128 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3129 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3130 | |
1da177e4 LT |
3131 | trap = lock_rename(new_dir, old_dir); |
3132 | ||
49705b77 | 3133 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3134 | error = PTR_ERR(old_dentry); |
3135 | if (IS_ERR(old_dentry)) | |
3136 | goto exit3; | |
3137 | /* source must exist */ | |
3138 | error = -ENOENT; | |
3139 | if (!old_dentry->d_inode) | |
3140 | goto exit4; | |
3141 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3142 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3143 | error = -ENOTDIR; | |
3144 | if (oldnd.last.name[oldnd.last.len]) | |
3145 | goto exit4; | |
3146 | if (newnd.last.name[newnd.last.len]) | |
3147 | goto exit4; | |
3148 | } | |
3149 | /* source should not be ancestor of target */ | |
3150 | error = -EINVAL; | |
3151 | if (old_dentry == trap) | |
3152 | goto exit4; | |
49705b77 | 3153 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3154 | error = PTR_ERR(new_dentry); |
3155 | if (IS_ERR(new_dentry)) | |
3156 | goto exit4; | |
3157 | /* target should not be an ancestor of source */ | |
3158 | error = -ENOTEMPTY; | |
3159 | if (new_dentry == trap) | |
3160 | goto exit5; | |
3161 | ||
9079b1eb DH |
3162 | error = mnt_want_write(oldnd.path.mnt); |
3163 | if (error) | |
3164 | goto exit5; | |
be6d3e56 KT |
3165 | error = security_path_rename(&oldnd.path, old_dentry, |
3166 | &newnd.path, new_dentry); | |
3167 | if (error) | |
3168 | goto exit6; | |
1da177e4 LT |
3169 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3170 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 3171 | exit6: |
9079b1eb | 3172 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
3173 | exit5: |
3174 | dput(new_dentry); | |
3175 | exit4: | |
3176 | dput(old_dentry); | |
3177 | exit3: | |
3178 | unlock_rename(new_dir, old_dir); | |
3179 | exit2: | |
1d957f9b | 3180 | path_put(&newnd.path); |
2ad94ae6 | 3181 | putname(to); |
1da177e4 | 3182 | exit1: |
1d957f9b | 3183 | path_put(&oldnd.path); |
1da177e4 | 3184 | putname(from); |
2ad94ae6 | 3185 | exit: |
1da177e4 LT |
3186 | return error; |
3187 | } | |
3188 | ||
a26eab24 | 3189 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3190 | { |
3191 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3192 | } | |
3193 | ||
1da177e4 LT |
3194 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3195 | { | |
3196 | int len; | |
3197 | ||
3198 | len = PTR_ERR(link); | |
3199 | if (IS_ERR(link)) | |
3200 | goto out; | |
3201 | ||
3202 | len = strlen(link); | |
3203 | if (len > (unsigned) buflen) | |
3204 | len = buflen; | |
3205 | if (copy_to_user(buffer, link, len)) | |
3206 | len = -EFAULT; | |
3207 | out: | |
3208 | return len; | |
3209 | } | |
3210 | ||
3211 | /* | |
3212 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3213 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3214 | * using) it for any given inode is up to filesystem. | |
3215 | */ | |
3216 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3217 | { | |
3218 | struct nameidata nd; | |
cc314eef | 3219 | void *cookie; |
694a1764 | 3220 | int res; |
cc314eef | 3221 | |
1da177e4 | 3222 | nd.depth = 0; |
cc314eef | 3223 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3224 | if (IS_ERR(cookie)) |
3225 | return PTR_ERR(cookie); | |
3226 | ||
3227 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3228 | if (dentry->d_inode->i_op->put_link) | |
3229 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3230 | return res; | |
1da177e4 LT |
3231 | } |
3232 | ||
3233 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3234 | { | |
3235 | return __vfs_follow_link(nd, link); | |
3236 | } | |
3237 | ||
3238 | /* get the link contents into pagecache */ | |
3239 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3240 | { | |
ebd09abb DG |
3241 | char *kaddr; |
3242 | struct page *page; | |
1da177e4 | 3243 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3244 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3245 | if (IS_ERR(page)) |
6fe6900e | 3246 | return (char*)page; |
1da177e4 | 3247 | *ppage = page; |
ebd09abb DG |
3248 | kaddr = kmap(page); |
3249 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3250 | return kaddr; | |
1da177e4 LT |
3251 | } |
3252 | ||
3253 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3254 | { | |
3255 | struct page *page = NULL; | |
3256 | char *s = page_getlink(dentry, &page); | |
3257 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3258 | if (page) { | |
3259 | kunmap(page); | |
3260 | page_cache_release(page); | |
3261 | } | |
3262 | return res; | |
3263 | } | |
3264 | ||
cc314eef | 3265 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3266 | { |
cc314eef | 3267 | struct page *page = NULL; |
1da177e4 | 3268 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3269 | return page; |
1da177e4 LT |
3270 | } |
3271 | ||
cc314eef | 3272 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3273 | { |
cc314eef LT |
3274 | struct page *page = cookie; |
3275 | ||
3276 | if (page) { | |
1da177e4 LT |
3277 | kunmap(page); |
3278 | page_cache_release(page); | |
1da177e4 LT |
3279 | } |
3280 | } | |
3281 | ||
54566b2c NP |
3282 | /* |
3283 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3284 | */ | |
3285 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3286 | { |
3287 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3288 | struct page *page; |
afddba49 | 3289 | void *fsdata; |
beb497ab | 3290 | int err; |
1da177e4 | 3291 | char *kaddr; |
54566b2c NP |
3292 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3293 | if (nofs) | |
3294 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3295 | |
7e53cac4 | 3296 | retry: |
afddba49 | 3297 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3298 | flags, &page, &fsdata); |
1da177e4 | 3299 | if (err) |
afddba49 NP |
3300 | goto fail; |
3301 | ||
1da177e4 LT |
3302 | kaddr = kmap_atomic(page, KM_USER0); |
3303 | memcpy(kaddr, symname, len-1); | |
3304 | kunmap_atomic(kaddr, KM_USER0); | |
afddba49 NP |
3305 | |
3306 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3307 | page, fsdata); | |
1da177e4 LT |
3308 | if (err < 0) |
3309 | goto fail; | |
afddba49 NP |
3310 | if (err < len-1) |
3311 | goto retry; | |
3312 | ||
1da177e4 LT |
3313 | mark_inode_dirty(inode); |
3314 | return 0; | |
1da177e4 LT |
3315 | fail: |
3316 | return err; | |
3317 | } | |
3318 | ||
0adb25d2 KK |
3319 | int page_symlink(struct inode *inode, const char *symname, int len) |
3320 | { | |
3321 | return __page_symlink(inode, symname, len, | |
54566b2c | 3322 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3323 | } |
3324 | ||
92e1d5be | 3325 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3326 | .readlink = generic_readlink, |
3327 | .follow_link = page_follow_link_light, | |
3328 | .put_link = page_put_link, | |
3329 | }; | |
3330 | ||
2d8f3038 | 3331 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3332 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3333 | EXPORT_SYMBOL(follow_down); |
3334 | EXPORT_SYMBOL(follow_up); | |
3335 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
3336 | EXPORT_SYMBOL(getname); | |
3337 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3338 | EXPORT_SYMBOL(lookup_one_len); |
3339 | EXPORT_SYMBOL(page_follow_link_light); | |
3340 | EXPORT_SYMBOL(page_put_link); | |
3341 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3342 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3343 | EXPORT_SYMBOL(page_symlink); |
3344 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
c9c6cac0 | 3345 | EXPORT_SYMBOL(kern_path_parent); |
d1811465 | 3346 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3347 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3348 | EXPORT_SYMBOL(inode_permission); |
1da177e4 LT |
3349 | EXPORT_SYMBOL(unlock_rename); |
3350 | EXPORT_SYMBOL(vfs_create); | |
3351 | EXPORT_SYMBOL(vfs_follow_link); | |
3352 | EXPORT_SYMBOL(vfs_link); | |
3353 | EXPORT_SYMBOL(vfs_mkdir); | |
3354 | EXPORT_SYMBOL(vfs_mknod); | |
3355 | EXPORT_SYMBOL(generic_permission); | |
3356 | EXPORT_SYMBOL(vfs_readlink); | |
3357 | EXPORT_SYMBOL(vfs_rename); | |
3358 | EXPORT_SYMBOL(vfs_rmdir); | |
3359 | EXPORT_SYMBOL(vfs_symlink); | |
3360 | EXPORT_SYMBOL(vfs_unlink); | |
3361 | EXPORT_SYMBOL(dentry_unhash); | |
3362 | EXPORT_SYMBOL(generic_readlink); |