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> | |
630d9c47 | 18 | #include <linux/export.h> |
44696908 | 19 | #include <linux/kernel.h> |
1da177e4 LT |
20 | #include <linux/slab.h> |
21 | #include <linux/fs.h> | |
22 | #include <linux/namei.h> | |
1da177e4 | 23 | #include <linux/pagemap.h> |
0eeca283 | 24 | #include <linux/fsnotify.h> |
1da177e4 LT |
25 | #include <linux/personality.h> |
26 | #include <linux/security.h> | |
6146f0d5 | 27 | #include <linux/ima.h> |
1da177e4 LT |
28 | #include <linux/syscalls.h> |
29 | #include <linux/mount.h> | |
30 | #include <linux/audit.h> | |
16f7e0fe | 31 | #include <linux/capability.h> |
834f2a4a | 32 | #include <linux/file.h> |
5590ff0d | 33 | #include <linux/fcntl.h> |
08ce5f16 | 34 | #include <linux/device_cgroup.h> |
5ad4e53b | 35 | #include <linux/fs_struct.h> |
e77819e5 | 36 | #include <linux/posix_acl.h> |
1da177e4 LT |
37 | #include <asm/uaccess.h> |
38 | ||
e81e3f4d | 39 | #include "internal.h" |
c7105365 | 40 | #include "mount.h" |
e81e3f4d | 41 | |
1da177e4 LT |
42 | /* [Feb-1997 T. Schoebel-Theuer] |
43 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
44 | * were necessary because of omirr. The reason is that omirr needs | |
45 | * to know the _real_ pathname, not the user-supplied one, in case | |
46 | * of symlinks (and also when transname replacements occur). | |
47 | * | |
48 | * The new code replaces the old recursive symlink resolution with | |
49 | * an iterative one (in case of non-nested symlink chains). It does | |
50 | * this with calls to <fs>_follow_link(). | |
51 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
52 | * replaced with a single function lookup_dentry() that can handle all | |
53 | * the special cases of the former code. | |
54 | * | |
55 | * With the new dcache, the pathname is stored at each inode, at least as | |
56 | * long as the refcount of the inode is positive. As a side effect, the | |
57 | * size of the dcache depends on the inode cache and thus is dynamic. | |
58 | * | |
59 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
60 | * resolution to correspond with current state of the code. | |
61 | * | |
62 | * Note that the symlink resolution is not *completely* iterative. | |
63 | * There is still a significant amount of tail- and mid- recursion in | |
64 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
65 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
66 | * may return different results than <fs>_follow_link(). Many virtual | |
67 | * filesystems (including /proc) exhibit this behavior. | |
68 | */ | |
69 | ||
70 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
71 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
72 | * and the name already exists in form of a symlink, try to create the new | |
73 | * name indicated by the symlink. The old code always complained that the | |
74 | * name already exists, due to not following the symlink even if its target | |
75 | * is nonexistent. The new semantics affects also mknod() and link() when | |
25985edc | 76 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
77 | * |
78 | * I don't know which semantics is the right one, since I have no access | |
79 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
80 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
81 | * "old" one. Personally, I think the new semantics is much more logical. | |
82 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
83 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
84 | * and in the old Linux semantics. | |
85 | */ | |
86 | ||
87 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
88 | * semantics. See the comments in "open_namei" and "do_link" below. | |
89 | * | |
90 | * [10-Sep-98 Alan Modra] Another symlink change. | |
91 | */ | |
92 | ||
93 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
94 | * inside the path - always follow. | |
95 | * in the last component in creation/removal/renaming - never follow. | |
96 | * if LOOKUP_FOLLOW passed - follow. | |
97 | * if the pathname has trailing slashes - follow. | |
98 | * otherwise - don't follow. | |
99 | * (applied in that order). | |
100 | * | |
101 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
102 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
103 | * During the 2.4 we need to fix the userland stuff depending on it - | |
104 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
105 | * XEmacs seems to be relying on it... | |
106 | */ | |
107 | /* | |
108 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 109 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
110 | * any extra contention... |
111 | */ | |
112 | ||
113 | /* In order to reduce some races, while at the same time doing additional | |
114 | * checking and hopefully speeding things up, we copy filenames to the | |
115 | * kernel data space before using them.. | |
116 | * | |
117 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
118 | * PATH_MAX includes the nul terminator --RR. | |
119 | */ | |
1fa1e7f6 | 120 | static char *getname_flags(const char __user *filename, int flags, int *empty) |
1da177e4 | 121 | { |
3f9f0aa6 LT |
122 | char *result = __getname(), *err; |
123 | int len; | |
4043cde8 | 124 | |
3f9f0aa6 | 125 | if (unlikely(!result)) |
4043cde8 EP |
126 | return ERR_PTR(-ENOMEM); |
127 | ||
3f9f0aa6 LT |
128 | len = strncpy_from_user(result, filename, PATH_MAX); |
129 | err = ERR_PTR(len); | |
130 | if (unlikely(len < 0)) | |
131 | goto error; | |
132 | ||
133 | /* The empty path is special. */ | |
134 | if (unlikely(!len)) { | |
135 | if (empty) | |
4043cde8 | 136 | *empty = 1; |
3f9f0aa6 LT |
137 | err = ERR_PTR(-ENOENT); |
138 | if (!(flags & LOOKUP_EMPTY)) | |
139 | goto error; | |
1da177e4 | 140 | } |
3f9f0aa6 LT |
141 | |
142 | err = ERR_PTR(-ENAMETOOLONG); | |
143 | if (likely(len < PATH_MAX)) { | |
144 | audit_getname(result); | |
145 | return result; | |
146 | } | |
147 | ||
148 | error: | |
149 | __putname(result); | |
150 | return err; | |
1da177e4 LT |
151 | } |
152 | ||
f52e0c11 AV |
153 | char *getname(const char __user * filename) |
154 | { | |
f7493e5d | 155 | return getname_flags(filename, 0, NULL); |
f52e0c11 AV |
156 | } |
157 | ||
1da177e4 LT |
158 | #ifdef CONFIG_AUDITSYSCALL |
159 | void putname(const char *name) | |
160 | { | |
5ac3a9c2 | 161 | if (unlikely(!audit_dummy_context())) |
1da177e4 LT |
162 | audit_putname(name); |
163 | else | |
164 | __putname(name); | |
165 | } | |
166 | EXPORT_SYMBOL(putname); | |
167 | #endif | |
168 | ||
e77819e5 LT |
169 | static int check_acl(struct inode *inode, int mask) |
170 | { | |
84635d68 | 171 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
172 | struct posix_acl *acl; |
173 | ||
e77819e5 | 174 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
175 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
176 | if (!acl) | |
e77819e5 | 177 | return -EAGAIN; |
3567866b AV |
178 | /* no ->get_acl() calls in RCU mode... */ |
179 | if (acl == ACL_NOT_CACHED) | |
180 | return -ECHILD; | |
206b1d09 | 181 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
182 | } |
183 | ||
184 | acl = get_cached_acl(inode, ACL_TYPE_ACCESS); | |
185 | ||
186 | /* | |
4e34e719 CH |
187 | * A filesystem can force a ACL callback by just never filling the |
188 | * ACL cache. But normally you'd fill the cache either at inode | |
189 | * instantiation time, or on the first ->get_acl call. | |
e77819e5 | 190 | * |
4e34e719 CH |
191 | * If the filesystem doesn't have a get_acl() function at all, we'll |
192 | * just create the negative cache entry. | |
e77819e5 LT |
193 | */ |
194 | if (acl == ACL_NOT_CACHED) { | |
4e34e719 CH |
195 | if (inode->i_op->get_acl) { |
196 | acl = inode->i_op->get_acl(inode, ACL_TYPE_ACCESS); | |
197 | if (IS_ERR(acl)) | |
198 | return PTR_ERR(acl); | |
199 | } else { | |
200 | set_cached_acl(inode, ACL_TYPE_ACCESS, NULL); | |
201 | return -EAGAIN; | |
202 | } | |
e77819e5 LT |
203 | } |
204 | ||
205 | if (acl) { | |
206 | int error = posix_acl_permission(inode, acl, mask); | |
207 | posix_acl_release(acl); | |
208 | return error; | |
209 | } | |
84635d68 | 210 | #endif |
e77819e5 LT |
211 | |
212 | return -EAGAIN; | |
213 | } | |
214 | ||
5909ccaa | 215 | /* |
948409c7 | 216 | * This does the basic permission checking |
1da177e4 | 217 | */ |
7e40145e | 218 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 219 | { |
26cf46be | 220 | unsigned int mode = inode->i_mode; |
1da177e4 | 221 | |
8e96e3b7 | 222 | if (likely(uid_eq(current_fsuid(), inode->i_uid))) |
1da177e4 LT |
223 | mode >>= 6; |
224 | else { | |
e77819e5 | 225 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 226 | int error = check_acl(inode, mask); |
b74c79e9 NP |
227 | if (error != -EAGAIN) |
228 | return error; | |
1da177e4 LT |
229 | } |
230 | ||
231 | if (in_group_p(inode->i_gid)) | |
232 | mode >>= 3; | |
233 | } | |
234 | ||
235 | /* | |
236 | * If the DACs are ok we don't need any capability check. | |
237 | */ | |
9c2c7039 | 238 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 239 | return 0; |
5909ccaa LT |
240 | return -EACCES; |
241 | } | |
242 | ||
243 | /** | |
b74c79e9 | 244 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 245 | * @inode: inode to check access rights for |
8fd90c8d | 246 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
247 | * |
248 | * Used to check for read/write/execute permissions on a file. | |
249 | * We use "fsuid" for this, letting us set arbitrary permissions | |
250 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
251 | * are used for other things. |
252 | * | |
253 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
254 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
255 | * It would then be called again in ref-walk mode. | |
5909ccaa | 256 | */ |
2830ba7f | 257 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
258 | { |
259 | int ret; | |
260 | ||
261 | /* | |
948409c7 | 262 | * Do the basic permission checks. |
5909ccaa | 263 | */ |
7e40145e | 264 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
265 | if (ret != -EACCES) |
266 | return ret; | |
1da177e4 | 267 | |
d594e7ec AV |
268 | if (S_ISDIR(inode->i_mode)) { |
269 | /* DACs are overridable for directories */ | |
1a48e2ac | 270 | if (inode_capable(inode, CAP_DAC_OVERRIDE)) |
d594e7ec AV |
271 | return 0; |
272 | if (!(mask & MAY_WRITE)) | |
1a48e2ac | 273 | if (inode_capable(inode, CAP_DAC_READ_SEARCH)) |
d594e7ec AV |
274 | return 0; |
275 | return -EACCES; | |
276 | } | |
1da177e4 LT |
277 | /* |
278 | * Read/write DACs are always overridable. | |
d594e7ec AV |
279 | * Executable DACs are overridable when there is |
280 | * at least one exec bit set. | |
1da177e4 | 281 | */ |
d594e7ec | 282 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) |
1a48e2ac | 283 | if (inode_capable(inode, CAP_DAC_OVERRIDE)) |
1da177e4 LT |
284 | return 0; |
285 | ||
286 | /* | |
287 | * Searching includes executable on directories, else just read. | |
288 | */ | |
7ea66001 | 289 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 290 | if (mask == MAY_READ) |
1a48e2ac | 291 | if (inode_capable(inode, CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
292 | return 0; |
293 | ||
294 | return -EACCES; | |
295 | } | |
296 | ||
3ddcd056 LT |
297 | /* |
298 | * We _really_ want to just do "generic_permission()" without | |
299 | * even looking at the inode->i_op values. So we keep a cache | |
300 | * flag in inode->i_opflags, that says "this has not special | |
301 | * permission function, use the fast case". | |
302 | */ | |
303 | static inline int do_inode_permission(struct inode *inode, int mask) | |
304 | { | |
305 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
306 | if (likely(inode->i_op->permission)) | |
307 | return inode->i_op->permission(inode, mask); | |
308 | ||
309 | /* This gets set once for the inode lifetime */ | |
310 | spin_lock(&inode->i_lock); | |
311 | inode->i_opflags |= IOP_FASTPERM; | |
312 | spin_unlock(&inode->i_lock); | |
313 | } | |
314 | return generic_permission(inode, mask); | |
315 | } | |
316 | ||
cb23beb5 | 317 | /** |
0bdaea90 DH |
318 | * __inode_permission - Check for access rights to a given inode |
319 | * @inode: Inode to check permission on | |
320 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
cb23beb5 | 321 | * |
0bdaea90 | 322 | * Check for read/write/execute permissions on an inode. |
948409c7 AG |
323 | * |
324 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
0bdaea90 DH |
325 | * |
326 | * This does not check for a read-only file system. You probably want | |
327 | * inode_permission(). | |
cb23beb5 | 328 | */ |
0bdaea90 | 329 | int __inode_permission(struct inode *inode, int mask) |
1da177e4 | 330 | { |
e6305c43 | 331 | int retval; |
1da177e4 | 332 | |
3ddcd056 | 333 | if (unlikely(mask & MAY_WRITE)) { |
1da177e4 LT |
334 | /* |
335 | * Nobody gets write access to an immutable file. | |
336 | */ | |
337 | if (IS_IMMUTABLE(inode)) | |
338 | return -EACCES; | |
339 | } | |
340 | ||
3ddcd056 | 341 | retval = do_inode_permission(inode, mask); |
1da177e4 LT |
342 | if (retval) |
343 | return retval; | |
344 | ||
08ce5f16 SH |
345 | retval = devcgroup_inode_permission(inode, mask); |
346 | if (retval) | |
347 | return retval; | |
348 | ||
d09ca739 | 349 | return security_inode_permission(inode, mask); |
1da177e4 LT |
350 | } |
351 | ||
0bdaea90 DH |
352 | /** |
353 | * sb_permission - Check superblock-level permissions | |
354 | * @sb: Superblock of inode to check permission on | |
355 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
356 | * | |
357 | * Separate out file-system wide checks from inode-specific permission checks. | |
358 | */ | |
359 | static int sb_permission(struct super_block *sb, struct inode *inode, int mask) | |
360 | { | |
361 | if (unlikely(mask & MAY_WRITE)) { | |
362 | umode_t mode = inode->i_mode; | |
363 | ||
364 | /* Nobody gets write access to a read-only fs. */ | |
365 | if ((sb->s_flags & MS_RDONLY) && | |
366 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
367 | return -EROFS; | |
368 | } | |
369 | return 0; | |
370 | } | |
371 | ||
372 | /** | |
373 | * inode_permission - Check for access rights to a given inode | |
374 | * @inode: Inode to check permission on | |
375 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
376 | * | |
377 | * Check for read/write/execute permissions on an inode. We use fs[ug]id for | |
378 | * this, letting us set arbitrary permissions for filesystem access without | |
379 | * changing the "normal" UIDs which are used for other things. | |
380 | * | |
381 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
382 | */ | |
383 | int inode_permission(struct inode *inode, int mask) | |
384 | { | |
385 | int retval; | |
386 | ||
387 | retval = sb_permission(inode->i_sb, inode, mask); | |
388 | if (retval) | |
389 | return retval; | |
390 | return __inode_permission(inode, mask); | |
391 | } | |
392 | ||
5dd784d0 JB |
393 | /** |
394 | * path_get - get a reference to a path | |
395 | * @path: path to get the reference to | |
396 | * | |
397 | * Given a path increment the reference count to the dentry and the vfsmount. | |
398 | */ | |
399 | void path_get(struct path *path) | |
400 | { | |
401 | mntget(path->mnt); | |
402 | dget(path->dentry); | |
403 | } | |
404 | EXPORT_SYMBOL(path_get); | |
405 | ||
1d957f9b JB |
406 | /** |
407 | * path_put - put a reference to a path | |
408 | * @path: path to put the reference to | |
409 | * | |
410 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
411 | */ | |
412 | void path_put(struct path *path) | |
1da177e4 | 413 | { |
1d957f9b JB |
414 | dput(path->dentry); |
415 | mntput(path->mnt); | |
1da177e4 | 416 | } |
1d957f9b | 417 | EXPORT_SYMBOL(path_put); |
1da177e4 | 418 | |
19660af7 | 419 | /* |
31e6b01f | 420 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
421 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
422 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
423 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
424 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
425 | * got stuck, so ref-walk may continue from there. If this is not successful | |
426 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
427 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 428 | */ |
31e6b01f NP |
429 | |
430 | /** | |
19660af7 AV |
431 | * unlazy_walk - try to switch to ref-walk mode. |
432 | * @nd: nameidata pathwalk data | |
433 | * @dentry: child of nd->path.dentry or NULL | |
39191628 | 434 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 435 | * |
19660af7 AV |
436 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
437 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
438 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 439 | */ |
19660af7 | 440 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry) |
31e6b01f NP |
441 | { |
442 | struct fs_struct *fs = current->fs; | |
443 | struct dentry *parent = nd->path.dentry; | |
5b6ca027 | 444 | int want_root = 0; |
31e6b01f NP |
445 | |
446 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
5b6ca027 AV |
447 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
448 | want_root = 1; | |
31e6b01f NP |
449 | spin_lock(&fs->lock); |
450 | if (nd->root.mnt != fs->root.mnt || | |
451 | nd->root.dentry != fs->root.dentry) | |
452 | goto err_root; | |
453 | } | |
454 | spin_lock(&parent->d_lock); | |
19660af7 AV |
455 | if (!dentry) { |
456 | if (!__d_rcu_to_refcount(parent, nd->seq)) | |
457 | goto err_parent; | |
458 | BUG_ON(nd->inode != parent->d_inode); | |
459 | } else { | |
94c0d4ec AV |
460 | if (dentry->d_parent != parent) |
461 | goto err_parent; | |
19660af7 AV |
462 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); |
463 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
464 | goto err_child; | |
465 | /* | |
466 | * If the sequence check on the child dentry passed, then | |
467 | * the child has not been removed from its parent. This | |
468 | * means the parent dentry must be valid and able to take | |
469 | * a reference at this point. | |
470 | */ | |
471 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
472 | BUG_ON(!parent->d_count); | |
473 | parent->d_count++; | |
474 | spin_unlock(&dentry->d_lock); | |
475 | } | |
31e6b01f | 476 | spin_unlock(&parent->d_lock); |
5b6ca027 | 477 | if (want_root) { |
31e6b01f NP |
478 | path_get(&nd->root); |
479 | spin_unlock(&fs->lock); | |
480 | } | |
481 | mntget(nd->path.mnt); | |
482 | ||
483 | rcu_read_unlock(); | |
962830df | 484 | br_read_unlock(&vfsmount_lock); |
31e6b01f NP |
485 | nd->flags &= ~LOOKUP_RCU; |
486 | return 0; | |
19660af7 AV |
487 | |
488 | err_child: | |
31e6b01f | 489 | spin_unlock(&dentry->d_lock); |
19660af7 | 490 | err_parent: |
31e6b01f NP |
491 | spin_unlock(&parent->d_lock); |
492 | err_root: | |
5b6ca027 | 493 | if (want_root) |
31e6b01f NP |
494 | spin_unlock(&fs->lock); |
495 | return -ECHILD; | |
496 | } | |
497 | ||
4ce16ef3 | 498 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 499 | { |
4ce16ef3 | 500 | return dentry->d_op->d_revalidate(dentry, flags); |
34286d66 NP |
501 | } |
502 | ||
9f1fafee AV |
503 | /** |
504 | * complete_walk - successful completion of path walk | |
505 | * @nd: pointer nameidata | |
39159de2 | 506 | * |
9f1fafee AV |
507 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
508 | * Revalidate the final result, unless we'd already done that during | |
509 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
510 | * success, -error on failure. In case of failure caller does not | |
511 | * need to drop nd->path. | |
39159de2 | 512 | */ |
9f1fafee | 513 | static int complete_walk(struct nameidata *nd) |
39159de2 | 514 | { |
16c2cd71 | 515 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 516 | int status; |
39159de2 | 517 | |
9f1fafee AV |
518 | if (nd->flags & LOOKUP_RCU) { |
519 | nd->flags &= ~LOOKUP_RCU; | |
520 | if (!(nd->flags & LOOKUP_ROOT)) | |
521 | nd->root.mnt = NULL; | |
522 | spin_lock(&dentry->d_lock); | |
523 | if (unlikely(!__d_rcu_to_refcount(dentry, nd->seq))) { | |
524 | spin_unlock(&dentry->d_lock); | |
525 | rcu_read_unlock(); | |
962830df | 526 | br_read_unlock(&vfsmount_lock); |
9f1fafee AV |
527 | return -ECHILD; |
528 | } | |
529 | BUG_ON(nd->inode != dentry->d_inode); | |
530 | spin_unlock(&dentry->d_lock); | |
531 | mntget(nd->path.mnt); | |
532 | rcu_read_unlock(); | |
962830df | 533 | br_read_unlock(&vfsmount_lock); |
9f1fafee AV |
534 | } |
535 | ||
16c2cd71 AV |
536 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
537 | return 0; | |
538 | ||
539 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
39159de2 JL |
540 | return 0; |
541 | ||
16c2cd71 AV |
542 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) |
543 | return 0; | |
544 | ||
545 | /* Note: we do not d_invalidate() */ | |
4ce16ef3 | 546 | status = d_revalidate(dentry, nd->flags); |
39159de2 JL |
547 | if (status > 0) |
548 | return 0; | |
549 | ||
16c2cd71 | 550 | if (!status) |
39159de2 | 551 | status = -ESTALE; |
16c2cd71 | 552 | |
9f1fafee | 553 | path_put(&nd->path); |
39159de2 JL |
554 | return status; |
555 | } | |
556 | ||
2a737871 AV |
557 | static __always_inline void set_root(struct nameidata *nd) |
558 | { | |
f7ad3c6b MS |
559 | if (!nd->root.mnt) |
560 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
561 | } |
562 | ||
6de88d72 AV |
563 | static int link_path_walk(const char *, struct nameidata *); |
564 | ||
31e6b01f NP |
565 | static __always_inline void set_root_rcu(struct nameidata *nd) |
566 | { | |
567 | if (!nd->root.mnt) { | |
568 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
569 | unsigned seq; |
570 | ||
571 | do { | |
572 | seq = read_seqcount_begin(&fs->seq); | |
573 | nd->root = fs->root; | |
c1530019 | 574 | nd->seq = __read_seqcount_begin(&nd->root.dentry->d_seq); |
c28cc364 | 575 | } while (read_seqcount_retry(&fs->seq, seq)); |
31e6b01f NP |
576 | } |
577 | } | |
578 | ||
f1662356 | 579 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 580 | { |
31e6b01f NP |
581 | int ret; |
582 | ||
1da177e4 LT |
583 | if (IS_ERR(link)) |
584 | goto fail; | |
585 | ||
586 | if (*link == '/') { | |
2a737871 | 587 | set_root(nd); |
1d957f9b | 588 | path_put(&nd->path); |
2a737871 AV |
589 | nd->path = nd->root; |
590 | path_get(&nd->root); | |
16c2cd71 | 591 | nd->flags |= LOOKUP_JUMPED; |
1da177e4 | 592 | } |
31e6b01f | 593 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 594 | |
31e6b01f NP |
595 | ret = link_path_walk(link, nd); |
596 | return ret; | |
1da177e4 | 597 | fail: |
1d957f9b | 598 | path_put(&nd->path); |
1da177e4 LT |
599 | return PTR_ERR(link); |
600 | } | |
601 | ||
1d957f9b | 602 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
603 | { |
604 | dput(path->dentry); | |
4ac91378 | 605 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
606 | mntput(path->mnt); |
607 | } | |
608 | ||
7b9337aa NP |
609 | static inline void path_to_nameidata(const struct path *path, |
610 | struct nameidata *nd) | |
051d3812 | 611 | { |
31e6b01f NP |
612 | if (!(nd->flags & LOOKUP_RCU)) { |
613 | dput(nd->path.dentry); | |
614 | if (nd->path.mnt != path->mnt) | |
615 | mntput(nd->path.mnt); | |
9a229683 | 616 | } |
31e6b01f | 617 | nd->path.mnt = path->mnt; |
4ac91378 | 618 | nd->path.dentry = path->dentry; |
051d3812 IK |
619 | } |
620 | ||
b5fb63c1 CH |
621 | /* |
622 | * Helper to directly jump to a known parsed path from ->follow_link, | |
623 | * caller must have taken a reference to path beforehand. | |
624 | */ | |
625 | void nd_jump_link(struct nameidata *nd, struct path *path) | |
626 | { | |
627 | path_put(&nd->path); | |
628 | ||
629 | nd->path = *path; | |
630 | nd->inode = nd->path.dentry->d_inode; | |
631 | nd->flags |= LOOKUP_JUMPED; | |
632 | ||
633 | BUG_ON(nd->inode->i_op->follow_link); | |
634 | } | |
635 | ||
574197e0 AV |
636 | static inline void put_link(struct nameidata *nd, struct path *link, void *cookie) |
637 | { | |
638 | struct inode *inode = link->dentry->d_inode; | |
6d7b5aae | 639 | if (inode->i_op->put_link) |
574197e0 AV |
640 | inode->i_op->put_link(link->dentry, nd, cookie); |
641 | path_put(link); | |
642 | } | |
643 | ||
def4af30 | 644 | static __always_inline int |
574197e0 | 645 | follow_link(struct path *link, struct nameidata *nd, void **p) |
1da177e4 | 646 | { |
7b9337aa | 647 | struct dentry *dentry = link->dentry; |
6d7b5aae AV |
648 | int error; |
649 | char *s; | |
1da177e4 | 650 | |
844a3917 AV |
651 | BUG_ON(nd->flags & LOOKUP_RCU); |
652 | ||
0e794589 AV |
653 | if (link->mnt == nd->path.mnt) |
654 | mntget(link->mnt); | |
655 | ||
6d7b5aae AV |
656 | error = -ELOOP; |
657 | if (unlikely(current->total_link_count >= 40)) | |
658 | goto out_put_nd_path; | |
659 | ||
574197e0 AV |
660 | cond_resched(); |
661 | current->total_link_count++; | |
662 | ||
68ac1234 | 663 | touch_atime(link); |
1da177e4 | 664 | nd_set_link(nd, NULL); |
cd4e91d3 | 665 | |
36f3b4f6 | 666 | error = security_inode_follow_link(link->dentry, nd); |
6d7b5aae AV |
667 | if (error) |
668 | goto out_put_nd_path; | |
36f3b4f6 | 669 | |
86acdca1 | 670 | nd->last_type = LAST_BIND; |
def4af30 AV |
671 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
672 | error = PTR_ERR(*p); | |
6d7b5aae | 673 | if (IS_ERR(*p)) |
408ef013 | 674 | goto out_put_nd_path; |
6d7b5aae AV |
675 | |
676 | error = 0; | |
677 | s = nd_get_link(nd); | |
678 | if (s) { | |
679 | error = __vfs_follow_link(nd, s); | |
b5fb63c1 CH |
680 | if (unlikely(error)) |
681 | put_link(nd, link, *p); | |
1da177e4 | 682 | } |
6d7b5aae AV |
683 | |
684 | return error; | |
685 | ||
686 | out_put_nd_path: | |
687 | path_put(&nd->path); | |
6d7b5aae | 688 | path_put(link); |
1da177e4 LT |
689 | return error; |
690 | } | |
691 | ||
31e6b01f NP |
692 | static int follow_up_rcu(struct path *path) |
693 | { | |
0714a533 AV |
694 | struct mount *mnt = real_mount(path->mnt); |
695 | struct mount *parent; | |
31e6b01f NP |
696 | struct dentry *mountpoint; |
697 | ||
0714a533 AV |
698 | parent = mnt->mnt_parent; |
699 | if (&parent->mnt == path->mnt) | |
31e6b01f | 700 | return 0; |
a73324da | 701 | mountpoint = mnt->mnt_mountpoint; |
31e6b01f | 702 | path->dentry = mountpoint; |
0714a533 | 703 | path->mnt = &parent->mnt; |
31e6b01f NP |
704 | return 1; |
705 | } | |
706 | ||
f015f126 DH |
707 | /* |
708 | * follow_up - Find the mountpoint of path's vfsmount | |
709 | * | |
710 | * Given a path, find the mountpoint of its source file system. | |
711 | * Replace @path with the path of the mountpoint in the parent mount. | |
712 | * Up is towards /. | |
713 | * | |
714 | * Return 1 if we went up a level and 0 if we were already at the | |
715 | * root. | |
716 | */ | |
bab77ebf | 717 | int follow_up(struct path *path) |
1da177e4 | 718 | { |
0714a533 AV |
719 | struct mount *mnt = real_mount(path->mnt); |
720 | struct mount *parent; | |
1da177e4 | 721 | struct dentry *mountpoint; |
99b7db7b | 722 | |
962830df | 723 | br_read_lock(&vfsmount_lock); |
0714a533 AV |
724 | parent = mnt->mnt_parent; |
725 | if (&parent->mnt == path->mnt) { | |
962830df | 726 | br_read_unlock(&vfsmount_lock); |
1da177e4 LT |
727 | return 0; |
728 | } | |
0714a533 | 729 | mntget(&parent->mnt); |
a73324da | 730 | mountpoint = dget(mnt->mnt_mountpoint); |
962830df | 731 | br_read_unlock(&vfsmount_lock); |
bab77ebf AV |
732 | dput(path->dentry); |
733 | path->dentry = mountpoint; | |
734 | mntput(path->mnt); | |
0714a533 | 735 | path->mnt = &parent->mnt; |
1da177e4 LT |
736 | return 1; |
737 | } | |
738 | ||
b5c84bf6 | 739 | /* |
9875cf80 DH |
740 | * Perform an automount |
741 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
742 | * were called with. | |
1da177e4 | 743 | */ |
9875cf80 DH |
744 | static int follow_automount(struct path *path, unsigned flags, |
745 | bool *need_mntput) | |
31e6b01f | 746 | { |
9875cf80 | 747 | struct vfsmount *mnt; |
ea5b778a | 748 | int err; |
9875cf80 DH |
749 | |
750 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
751 | return -EREMOTE; | |
752 | ||
0ec26fd0 MS |
753 | /* We don't want to mount if someone's just doing a stat - |
754 | * unless they're stat'ing a directory and appended a '/' to | |
755 | * the name. | |
756 | * | |
757 | * We do, however, want to mount if someone wants to open or | |
758 | * create a file of any type under the mountpoint, wants to | |
759 | * traverse through the mountpoint or wants to open the | |
760 | * mounted directory. Also, autofs may mark negative dentries | |
761 | * as being automount points. These will need the attentions | |
762 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 763 | */ |
0ec26fd0 | 764 | if (!(flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
d94c177b | 765 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
0ec26fd0 MS |
766 | path->dentry->d_inode) |
767 | return -EISDIR; | |
768 | ||
9875cf80 DH |
769 | current->total_link_count++; |
770 | if (current->total_link_count >= 40) | |
771 | return -ELOOP; | |
772 | ||
773 | mnt = path->dentry->d_op->d_automount(path); | |
774 | if (IS_ERR(mnt)) { | |
775 | /* | |
776 | * The filesystem is allowed to return -EISDIR here to indicate | |
777 | * it doesn't want to automount. For instance, autofs would do | |
778 | * this so that its userspace daemon can mount on this dentry. | |
779 | * | |
780 | * However, we can only permit this if it's a terminal point in | |
781 | * the path being looked up; if it wasn't then the remainder of | |
782 | * the path is inaccessible and we should say so. | |
783 | */ | |
49084c3b | 784 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_PARENT)) |
9875cf80 DH |
785 | return -EREMOTE; |
786 | return PTR_ERR(mnt); | |
31e6b01f | 787 | } |
ea5b778a | 788 | |
9875cf80 DH |
789 | if (!mnt) /* mount collision */ |
790 | return 0; | |
31e6b01f | 791 | |
8aef1884 AV |
792 | if (!*need_mntput) { |
793 | /* lock_mount() may release path->mnt on error */ | |
794 | mntget(path->mnt); | |
795 | *need_mntput = true; | |
796 | } | |
19a167af | 797 | err = finish_automount(mnt, path); |
9875cf80 | 798 | |
ea5b778a DH |
799 | switch (err) { |
800 | case -EBUSY: | |
801 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 802 | return 0; |
ea5b778a | 803 | case 0: |
8aef1884 | 804 | path_put(path); |
ea5b778a DH |
805 | path->mnt = mnt; |
806 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 807 | return 0; |
19a167af AV |
808 | default: |
809 | return err; | |
ea5b778a | 810 | } |
19a167af | 811 | |
463ffb2e AV |
812 | } |
813 | ||
9875cf80 DH |
814 | /* |
815 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 816 | * - Flagged for transit management (autofs) |
9875cf80 DH |
817 | * - Flagged as mountpoint |
818 | * - Flagged as automount point | |
819 | * | |
820 | * This may only be called in refwalk mode. | |
821 | * | |
822 | * Serialization is taken care of in namespace.c | |
823 | */ | |
824 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 825 | { |
8aef1884 | 826 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
827 | unsigned managed; |
828 | bool need_mntput = false; | |
8aef1884 | 829 | int ret = 0; |
9875cf80 DH |
830 | |
831 | /* Given that we're not holding a lock here, we retain the value in a | |
832 | * local variable for each dentry as we look at it so that we don't see | |
833 | * the components of that value change under us */ | |
834 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
835 | managed &= DCACHE_MANAGED_DENTRY, | |
836 | unlikely(managed != 0)) { | |
cc53ce53 DH |
837 | /* Allow the filesystem to manage the transit without i_mutex |
838 | * being held. */ | |
839 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
840 | BUG_ON(!path->dentry->d_op); | |
841 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 842 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 843 | if (ret < 0) |
8aef1884 | 844 | break; |
cc53ce53 DH |
845 | } |
846 | ||
9875cf80 DH |
847 | /* Transit to a mounted filesystem. */ |
848 | if (managed & DCACHE_MOUNTED) { | |
849 | struct vfsmount *mounted = lookup_mnt(path); | |
850 | if (mounted) { | |
851 | dput(path->dentry); | |
852 | if (need_mntput) | |
853 | mntput(path->mnt); | |
854 | path->mnt = mounted; | |
855 | path->dentry = dget(mounted->mnt_root); | |
856 | need_mntput = true; | |
857 | continue; | |
858 | } | |
859 | ||
860 | /* Something is mounted on this dentry in another | |
861 | * namespace and/or whatever was mounted there in this | |
862 | * namespace got unmounted before we managed to get the | |
863 | * vfsmount_lock */ | |
864 | } | |
865 | ||
866 | /* Handle an automount point */ | |
867 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
868 | ret = follow_automount(path, flags, &need_mntput); | |
869 | if (ret < 0) | |
8aef1884 | 870 | break; |
9875cf80 DH |
871 | continue; |
872 | } | |
873 | ||
874 | /* We didn't change the current path point */ | |
875 | break; | |
1da177e4 | 876 | } |
8aef1884 AV |
877 | |
878 | if (need_mntput && path->mnt == mnt) | |
879 | mntput(path->mnt); | |
880 | if (ret == -EISDIR) | |
881 | ret = 0; | |
a3fbbde7 | 882 | return ret < 0 ? ret : need_mntput; |
1da177e4 LT |
883 | } |
884 | ||
cc53ce53 | 885 | int follow_down_one(struct path *path) |
1da177e4 LT |
886 | { |
887 | struct vfsmount *mounted; | |
888 | ||
1c755af4 | 889 | mounted = lookup_mnt(path); |
1da177e4 | 890 | if (mounted) { |
9393bd07 AV |
891 | dput(path->dentry); |
892 | mntput(path->mnt); | |
893 | path->mnt = mounted; | |
894 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
895 | return 1; |
896 | } | |
897 | return 0; | |
898 | } | |
899 | ||
62a7375e IK |
900 | static inline bool managed_dentry_might_block(struct dentry *dentry) |
901 | { | |
902 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT && | |
903 | dentry->d_op->d_manage(dentry, true) < 0); | |
904 | } | |
905 | ||
9875cf80 | 906 | /* |
287548e4 AV |
907 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
908 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
909 | */ |
910 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 911 | struct inode **inode) |
9875cf80 | 912 | { |
62a7375e | 913 | for (;;) { |
c7105365 | 914 | struct mount *mounted; |
62a7375e IK |
915 | /* |
916 | * Don't forget we might have a non-mountpoint managed dentry | |
917 | * that wants to block transit. | |
918 | */ | |
287548e4 | 919 | if (unlikely(managed_dentry_might_block(path->dentry))) |
ab90911f | 920 | return false; |
62a7375e IK |
921 | |
922 | if (!d_mountpoint(path->dentry)) | |
923 | break; | |
924 | ||
9875cf80 DH |
925 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
926 | if (!mounted) | |
927 | break; | |
c7105365 AV |
928 | path->mnt = &mounted->mnt; |
929 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 930 | nd->flags |= LOOKUP_JUMPED; |
9875cf80 | 931 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
932 | /* |
933 | * Update the inode too. We don't need to re-check the | |
934 | * dentry sequence number here after this d_inode read, | |
935 | * because a mount-point is always pinned. | |
936 | */ | |
937 | *inode = path->dentry->d_inode; | |
9875cf80 | 938 | } |
9875cf80 DH |
939 | return true; |
940 | } | |
941 | ||
dea39376 | 942 | static void follow_mount_rcu(struct nameidata *nd) |
287548e4 | 943 | { |
dea39376 | 944 | while (d_mountpoint(nd->path.dentry)) { |
c7105365 | 945 | struct mount *mounted; |
dea39376 | 946 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry, 1); |
287548e4 AV |
947 | if (!mounted) |
948 | break; | |
c7105365 AV |
949 | nd->path.mnt = &mounted->mnt; |
950 | nd->path.dentry = mounted->mnt.mnt_root; | |
dea39376 | 951 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
287548e4 AV |
952 | } |
953 | } | |
954 | ||
31e6b01f NP |
955 | static int follow_dotdot_rcu(struct nameidata *nd) |
956 | { | |
31e6b01f NP |
957 | set_root_rcu(nd); |
958 | ||
9875cf80 | 959 | while (1) { |
31e6b01f NP |
960 | if (nd->path.dentry == nd->root.dentry && |
961 | nd->path.mnt == nd->root.mnt) { | |
962 | break; | |
963 | } | |
964 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
965 | struct dentry *old = nd->path.dentry; | |
966 | struct dentry *parent = old->d_parent; | |
967 | unsigned seq; | |
968 | ||
969 | seq = read_seqcount_begin(&parent->d_seq); | |
970 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 971 | goto failed; |
31e6b01f NP |
972 | nd->path.dentry = parent; |
973 | nd->seq = seq; | |
974 | break; | |
975 | } | |
976 | if (!follow_up_rcu(&nd->path)) | |
977 | break; | |
978 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
31e6b01f | 979 | } |
dea39376 AV |
980 | follow_mount_rcu(nd); |
981 | nd->inode = nd->path.dentry->d_inode; | |
31e6b01f | 982 | return 0; |
ef7562d5 AV |
983 | |
984 | failed: | |
985 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
986 | if (!(nd->flags & LOOKUP_ROOT)) |
987 | nd->root.mnt = NULL; | |
ef7562d5 | 988 | rcu_read_unlock(); |
962830df | 989 | br_read_unlock(&vfsmount_lock); |
ef7562d5 | 990 | return -ECHILD; |
31e6b01f NP |
991 | } |
992 | ||
cc53ce53 DH |
993 | /* |
994 | * Follow down to the covering mount currently visible to userspace. At each | |
995 | * point, the filesystem owning that dentry may be queried as to whether the | |
996 | * caller is permitted to proceed or not. | |
cc53ce53 | 997 | */ |
7cc90cc3 | 998 | int follow_down(struct path *path) |
cc53ce53 DH |
999 | { |
1000 | unsigned managed; | |
1001 | int ret; | |
1002 | ||
1003 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1004 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1005 | /* Allow the filesystem to manage the transit without i_mutex | |
1006 | * being held. | |
1007 | * | |
1008 | * We indicate to the filesystem if someone is trying to mount | |
1009 | * something here. This gives autofs the chance to deny anyone | |
1010 | * other than its daemon the right to mount on its | |
1011 | * superstructure. | |
1012 | * | |
1013 | * The filesystem may sleep at this point. | |
1014 | */ | |
1015 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1016 | BUG_ON(!path->dentry->d_op); | |
1017 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 1018 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 1019 | path->dentry, false); |
cc53ce53 DH |
1020 | if (ret < 0) |
1021 | return ret == -EISDIR ? 0 : ret; | |
1022 | } | |
1023 | ||
1024 | /* Transit to a mounted filesystem. */ | |
1025 | if (managed & DCACHE_MOUNTED) { | |
1026 | struct vfsmount *mounted = lookup_mnt(path); | |
1027 | if (!mounted) | |
1028 | break; | |
1029 | dput(path->dentry); | |
1030 | mntput(path->mnt); | |
1031 | path->mnt = mounted; | |
1032 | path->dentry = dget(mounted->mnt_root); | |
1033 | continue; | |
1034 | } | |
1035 | ||
1036 | /* Don't handle automount points here */ | |
1037 | break; | |
1038 | } | |
1039 | return 0; | |
1040 | } | |
1041 | ||
9875cf80 DH |
1042 | /* |
1043 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1044 | */ | |
1045 | static void follow_mount(struct path *path) | |
1046 | { | |
1047 | while (d_mountpoint(path->dentry)) { | |
1048 | struct vfsmount *mounted = lookup_mnt(path); | |
1049 | if (!mounted) | |
1050 | break; | |
1051 | dput(path->dentry); | |
1052 | mntput(path->mnt); | |
1053 | path->mnt = mounted; | |
1054 | path->dentry = dget(mounted->mnt_root); | |
1055 | } | |
1056 | } | |
1057 | ||
31e6b01f | 1058 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1059 | { |
2a737871 | 1060 | set_root(nd); |
e518ddb7 | 1061 | |
1da177e4 | 1062 | while(1) { |
4ac91378 | 1063 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1064 | |
2a737871 AV |
1065 | if (nd->path.dentry == nd->root.dentry && |
1066 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1067 | break; |
1068 | } | |
4ac91378 | 1069 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1070 | /* rare case of legitimate dget_parent()... */ |
1071 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1072 | dput(old); |
1073 | break; | |
1074 | } | |
3088dd70 | 1075 | if (!follow_up(&nd->path)) |
1da177e4 | 1076 | break; |
1da177e4 | 1077 | } |
79ed0226 | 1078 | follow_mount(&nd->path); |
31e6b01f | 1079 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1080 | } |
1081 | ||
baa03890 | 1082 | /* |
bad61189 MS |
1083 | * This looks up the name in dcache, possibly revalidates the old dentry and |
1084 | * allocates a new one if not found or not valid. In the need_lookup argument | |
1085 | * returns whether i_op->lookup is necessary. | |
1086 | * | |
1087 | * dir->d_inode->i_mutex must be held | |
baa03890 | 1088 | */ |
bad61189 | 1089 | static struct dentry *lookup_dcache(struct qstr *name, struct dentry *dir, |
201f956e | 1090 | unsigned int flags, bool *need_lookup) |
baa03890 | 1091 | { |
baa03890 | 1092 | struct dentry *dentry; |
bad61189 | 1093 | int error; |
baa03890 | 1094 | |
bad61189 MS |
1095 | *need_lookup = false; |
1096 | dentry = d_lookup(dir, name); | |
1097 | if (dentry) { | |
1098 | if (d_need_lookup(dentry)) { | |
1099 | *need_lookup = true; | |
1100 | } else if (dentry->d_flags & DCACHE_OP_REVALIDATE) { | |
201f956e | 1101 | error = d_revalidate(dentry, flags); |
bad61189 MS |
1102 | if (unlikely(error <= 0)) { |
1103 | if (error < 0) { | |
1104 | dput(dentry); | |
1105 | return ERR_PTR(error); | |
1106 | } else if (!d_invalidate(dentry)) { | |
1107 | dput(dentry); | |
1108 | dentry = NULL; | |
1109 | } | |
1110 | } | |
1111 | } | |
1112 | } | |
baa03890 | 1113 | |
bad61189 MS |
1114 | if (!dentry) { |
1115 | dentry = d_alloc(dir, name); | |
1116 | if (unlikely(!dentry)) | |
1117 | return ERR_PTR(-ENOMEM); | |
baa03890 | 1118 | |
bad61189 | 1119 | *need_lookup = true; |
baa03890 NP |
1120 | } |
1121 | return dentry; | |
1122 | } | |
1123 | ||
44396f4b | 1124 | /* |
bad61189 MS |
1125 | * Call i_op->lookup on the dentry. The dentry must be negative but may be |
1126 | * hashed if it was pouplated with DCACHE_NEED_LOOKUP. | |
1127 | * | |
1128 | * dir->d_inode->i_mutex must be held | |
44396f4b | 1129 | */ |
bad61189 | 1130 | static struct dentry *lookup_real(struct inode *dir, struct dentry *dentry, |
72bd866a | 1131 | unsigned int flags) |
44396f4b | 1132 | { |
44396f4b JB |
1133 | struct dentry *old; |
1134 | ||
1135 | /* Don't create child dentry for a dead directory. */ | |
bad61189 | 1136 | if (unlikely(IS_DEADDIR(dir))) { |
e188dc02 | 1137 | dput(dentry); |
44396f4b | 1138 | return ERR_PTR(-ENOENT); |
e188dc02 | 1139 | } |
44396f4b | 1140 | |
72bd866a | 1141 | old = dir->i_op->lookup(dir, dentry, flags); |
44396f4b JB |
1142 | if (unlikely(old)) { |
1143 | dput(dentry); | |
1144 | dentry = old; | |
1145 | } | |
1146 | return dentry; | |
1147 | } | |
1148 | ||
a3255546 | 1149 | static struct dentry *__lookup_hash(struct qstr *name, |
72bd866a | 1150 | struct dentry *base, unsigned int flags) |
a3255546 | 1151 | { |
bad61189 | 1152 | bool need_lookup; |
a3255546 AV |
1153 | struct dentry *dentry; |
1154 | ||
72bd866a | 1155 | dentry = lookup_dcache(name, base, flags, &need_lookup); |
bad61189 MS |
1156 | if (!need_lookup) |
1157 | return dentry; | |
a3255546 | 1158 | |
72bd866a | 1159 | return lookup_real(base->d_inode, dentry, flags); |
a3255546 AV |
1160 | } |
1161 | ||
1da177e4 LT |
1162 | /* |
1163 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1164 | * small and for now I'd prefer to have fast path as straight as possible. | |
1165 | * It _is_ time-critical. | |
1166 | */ | |
697f514d MS |
1167 | static int lookup_fast(struct nameidata *nd, struct qstr *name, |
1168 | struct path *path, struct inode **inode) | |
1da177e4 | 1169 | { |
4ac91378 | 1170 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1171 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1172 | int need_reval = 1; |
1173 | int status = 1; | |
9875cf80 DH |
1174 | int err; |
1175 | ||
b04f784e NP |
1176 | /* |
1177 | * Rename seqlock is not required here because in the off chance | |
1178 | * of a false negative due to a concurrent rename, we're going to | |
1179 | * do the non-racy lookup, below. | |
1180 | */ | |
31e6b01f NP |
1181 | if (nd->flags & LOOKUP_RCU) { |
1182 | unsigned seq; | |
12f8ad4b | 1183 | dentry = __d_lookup_rcu(parent, name, &seq, nd->inode); |
5a18fff2 AV |
1184 | if (!dentry) |
1185 | goto unlazy; | |
1186 | ||
12f8ad4b LT |
1187 | /* |
1188 | * This sequence count validates that the inode matches | |
1189 | * the dentry name information from lookup. | |
1190 | */ | |
1191 | *inode = dentry->d_inode; | |
1192 | if (read_seqcount_retry(&dentry->d_seq, seq)) | |
1193 | return -ECHILD; | |
1194 | ||
1195 | /* | |
1196 | * This sequence count validates that the parent had no | |
1197 | * changes while we did the lookup of the dentry above. | |
1198 | * | |
1199 | * The memory barrier in read_seqcount_begin of child is | |
1200 | * enough, we can use __read_seqcount_retry here. | |
1201 | */ | |
31e6b01f NP |
1202 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) |
1203 | return -ECHILD; | |
31e6b01f | 1204 | nd->seq = seq; |
5a18fff2 | 1205 | |
fa4ee159 MS |
1206 | if (unlikely(d_need_lookup(dentry))) |
1207 | goto unlazy; | |
24643087 | 1208 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
4ce16ef3 | 1209 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1210 | if (unlikely(status <= 0)) { |
1211 | if (status != -ECHILD) | |
1212 | need_reval = 0; | |
1213 | goto unlazy; | |
1214 | } | |
24643087 | 1215 | } |
31e6b01f NP |
1216 | path->mnt = mnt; |
1217 | path->dentry = dentry; | |
d6e9bd25 AV |
1218 | if (unlikely(!__follow_mount_rcu(nd, path, inode))) |
1219 | goto unlazy; | |
1220 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1221 | goto unlazy; | |
1222 | return 0; | |
5a18fff2 | 1223 | unlazy: |
19660af7 AV |
1224 | if (unlazy_walk(nd, dentry)) |
1225 | return -ECHILD; | |
5a18fff2 AV |
1226 | } else { |
1227 | dentry = __d_lookup(parent, name); | |
9875cf80 | 1228 | } |
5a18fff2 | 1229 | |
81e6f520 AV |
1230 | if (unlikely(!dentry)) |
1231 | goto need_lookup; | |
1232 | ||
1233 | if (unlikely(d_need_lookup(dentry))) { | |
44396f4b | 1234 | dput(dentry); |
81e6f520 | 1235 | goto need_lookup; |
5a18fff2 | 1236 | } |
81e6f520 | 1237 | |
5a18fff2 | 1238 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) |
4ce16ef3 | 1239 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1240 | if (unlikely(status <= 0)) { |
1241 | if (status < 0) { | |
1242 | dput(dentry); | |
1243 | return status; | |
1244 | } | |
1245 | if (!d_invalidate(dentry)) { | |
1246 | dput(dentry); | |
81e6f520 | 1247 | goto need_lookup; |
5a18fff2 | 1248 | } |
24643087 | 1249 | } |
697f514d | 1250 | |
9875cf80 DH |
1251 | path->mnt = mnt; |
1252 | path->dentry = dentry; | |
1253 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1254 | if (unlikely(err < 0)) { |
1255 | path_put_conditional(path, nd); | |
9875cf80 | 1256 | return err; |
89312214 | 1257 | } |
a3fbbde7 AV |
1258 | if (err) |
1259 | nd->flags |= LOOKUP_JUMPED; | |
9875cf80 | 1260 | *inode = path->dentry->d_inode; |
1da177e4 | 1261 | return 0; |
81e6f520 AV |
1262 | |
1263 | need_lookup: | |
697f514d MS |
1264 | return 1; |
1265 | } | |
1266 | ||
1267 | /* Fast lookup failed, do it the slow way */ | |
1268 | static int lookup_slow(struct nameidata *nd, struct qstr *name, | |
1269 | struct path *path) | |
1270 | { | |
1271 | struct dentry *dentry, *parent; | |
1272 | int err; | |
1273 | ||
1274 | parent = nd->path.dentry; | |
81e6f520 AV |
1275 | BUG_ON(nd->inode != parent->d_inode); |
1276 | ||
1277 | mutex_lock(&parent->d_inode->i_mutex); | |
72bd866a | 1278 | dentry = __lookup_hash(name, parent, nd->flags); |
81e6f520 AV |
1279 | mutex_unlock(&parent->d_inode->i_mutex); |
1280 | if (IS_ERR(dentry)) | |
1281 | return PTR_ERR(dentry); | |
697f514d MS |
1282 | path->mnt = nd->path.mnt; |
1283 | path->dentry = dentry; | |
1284 | err = follow_managed(path, nd->flags); | |
1285 | if (unlikely(err < 0)) { | |
1286 | path_put_conditional(path, nd); | |
1287 | return err; | |
1288 | } | |
1289 | if (err) | |
1290 | nd->flags |= LOOKUP_JUMPED; | |
1291 | return 0; | |
1da177e4 LT |
1292 | } |
1293 | ||
52094c8a AV |
1294 | static inline int may_lookup(struct nameidata *nd) |
1295 | { | |
1296 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1297 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1298 | if (err != -ECHILD) |
1299 | return err; | |
19660af7 | 1300 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1301 | return -ECHILD; |
1302 | } | |
4ad5abb3 | 1303 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1304 | } |
1305 | ||
9856fa1b AV |
1306 | static inline int handle_dots(struct nameidata *nd, int type) |
1307 | { | |
1308 | if (type == LAST_DOTDOT) { | |
1309 | if (nd->flags & LOOKUP_RCU) { | |
1310 | if (follow_dotdot_rcu(nd)) | |
1311 | return -ECHILD; | |
1312 | } else | |
1313 | follow_dotdot(nd); | |
1314 | } | |
1315 | return 0; | |
1316 | } | |
1317 | ||
951361f9 AV |
1318 | static void terminate_walk(struct nameidata *nd) |
1319 | { | |
1320 | if (!(nd->flags & LOOKUP_RCU)) { | |
1321 | path_put(&nd->path); | |
1322 | } else { | |
1323 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1324 | if (!(nd->flags & LOOKUP_ROOT)) |
1325 | nd->root.mnt = NULL; | |
951361f9 | 1326 | rcu_read_unlock(); |
962830df | 1327 | br_read_unlock(&vfsmount_lock); |
951361f9 AV |
1328 | } |
1329 | } | |
1330 | ||
3ddcd056 LT |
1331 | /* |
1332 | * Do we need to follow links? We _really_ want to be able | |
1333 | * to do this check without having to look at inode->i_op, | |
1334 | * so we keep a cache of "no, this doesn't need follow_link" | |
1335 | * for the common case. | |
1336 | */ | |
7813b94a | 1337 | static inline int should_follow_link(struct inode *inode, int follow) |
3ddcd056 LT |
1338 | { |
1339 | if (unlikely(!(inode->i_opflags & IOP_NOFOLLOW))) { | |
1340 | if (likely(inode->i_op->follow_link)) | |
1341 | return follow; | |
1342 | ||
1343 | /* This gets set once for the inode lifetime */ | |
1344 | spin_lock(&inode->i_lock); | |
1345 | inode->i_opflags |= IOP_NOFOLLOW; | |
1346 | spin_unlock(&inode->i_lock); | |
1347 | } | |
1348 | return 0; | |
1349 | } | |
1350 | ||
ce57dfc1 AV |
1351 | static inline int walk_component(struct nameidata *nd, struct path *path, |
1352 | struct qstr *name, int type, int follow) | |
1353 | { | |
1354 | struct inode *inode; | |
1355 | int err; | |
1356 | /* | |
1357 | * "." and ".." are special - ".." especially so because it has | |
1358 | * to be able to know about the current root directory and | |
1359 | * parent relationships. | |
1360 | */ | |
1361 | if (unlikely(type != LAST_NORM)) | |
1362 | return handle_dots(nd, type); | |
697f514d | 1363 | err = lookup_fast(nd, name, path, &inode); |
ce57dfc1 | 1364 | if (unlikely(err)) { |
697f514d MS |
1365 | if (err < 0) |
1366 | goto out_err; | |
1367 | ||
1368 | err = lookup_slow(nd, name, path); | |
1369 | if (err < 0) | |
1370 | goto out_err; | |
1371 | ||
1372 | inode = path->dentry->d_inode; | |
ce57dfc1 | 1373 | } |
697f514d MS |
1374 | err = -ENOENT; |
1375 | if (!inode) | |
1376 | goto out_path_put; | |
1377 | ||
7813b94a | 1378 | if (should_follow_link(inode, follow)) { |
19660af7 AV |
1379 | if (nd->flags & LOOKUP_RCU) { |
1380 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
697f514d MS |
1381 | err = -ECHILD; |
1382 | goto out_err; | |
19660af7 AV |
1383 | } |
1384 | } | |
ce57dfc1 AV |
1385 | BUG_ON(inode != path->dentry->d_inode); |
1386 | return 1; | |
1387 | } | |
1388 | path_to_nameidata(path, nd); | |
1389 | nd->inode = inode; | |
1390 | return 0; | |
697f514d MS |
1391 | |
1392 | out_path_put: | |
1393 | path_to_nameidata(path, nd); | |
1394 | out_err: | |
1395 | terminate_walk(nd); | |
1396 | return err; | |
ce57dfc1 AV |
1397 | } |
1398 | ||
b356379a AV |
1399 | /* |
1400 | * This limits recursive symlink follows to 8, while | |
1401 | * limiting consecutive symlinks to 40. | |
1402 | * | |
1403 | * Without that kind of total limit, nasty chains of consecutive | |
1404 | * symlinks can cause almost arbitrarily long lookups. | |
1405 | */ | |
1406 | static inline int nested_symlink(struct path *path, struct nameidata *nd) | |
1407 | { | |
1408 | int res; | |
1409 | ||
b356379a AV |
1410 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1411 | path_put_conditional(path, nd); | |
1412 | path_put(&nd->path); | |
1413 | return -ELOOP; | |
1414 | } | |
1a4022f8 | 1415 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); |
b356379a AV |
1416 | |
1417 | nd->depth++; | |
1418 | current->link_count++; | |
1419 | ||
1420 | do { | |
1421 | struct path link = *path; | |
1422 | void *cookie; | |
574197e0 AV |
1423 | |
1424 | res = follow_link(&link, nd, &cookie); | |
6d7b5aae AV |
1425 | if (res) |
1426 | break; | |
1427 | res = walk_component(nd, path, &nd->last, | |
1428 | nd->last_type, LOOKUP_FOLLOW); | |
574197e0 | 1429 | put_link(nd, &link, cookie); |
b356379a AV |
1430 | } while (res > 0); |
1431 | ||
1432 | current->link_count--; | |
1433 | nd->depth--; | |
1434 | return res; | |
1435 | } | |
1436 | ||
3ddcd056 LT |
1437 | /* |
1438 | * We really don't want to look at inode->i_op->lookup | |
1439 | * when we don't have to. So we keep a cache bit in | |
1440 | * the inode ->i_opflags field that says "yes, we can | |
1441 | * do lookup on this inode". | |
1442 | */ | |
1443 | static inline int can_lookup(struct inode *inode) | |
1444 | { | |
1445 | if (likely(inode->i_opflags & IOP_LOOKUP)) | |
1446 | return 1; | |
1447 | if (likely(!inode->i_op->lookup)) | |
1448 | return 0; | |
1449 | ||
1450 | /* We do this once for the lifetime of the inode */ | |
1451 | spin_lock(&inode->i_lock); | |
1452 | inode->i_opflags |= IOP_LOOKUP; | |
1453 | spin_unlock(&inode->i_lock); | |
1454 | return 1; | |
1455 | } | |
1456 | ||
bfcfaa77 LT |
1457 | /* |
1458 | * We can do the critical dentry name comparison and hashing | |
1459 | * operations one word at a time, but we are limited to: | |
1460 | * | |
1461 | * - Architectures with fast unaligned word accesses. We could | |
1462 | * do a "get_unaligned()" if this helps and is sufficiently | |
1463 | * fast. | |
1464 | * | |
1465 | * - Little-endian machines (so that we can generate the mask | |
1466 | * of low bytes efficiently). Again, we *could* do a byte | |
1467 | * swapping load on big-endian architectures if that is not | |
1468 | * expensive enough to make the optimization worthless. | |
1469 | * | |
1470 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we | |
1471 | * do not trap on the (extremely unlikely) case of a page | |
1472 | * crossing operation. | |
1473 | * | |
1474 | * - Furthermore, we need an efficient 64-bit compile for the | |
1475 | * 64-bit case in order to generate the "number of bytes in | |
1476 | * the final mask". Again, that could be replaced with a | |
1477 | * efficient population count instruction or similar. | |
1478 | */ | |
1479 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1480 | ||
f68e556e | 1481 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1482 | |
f68e556e | 1483 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1484 | |
1485 | static inline unsigned int fold_hash(unsigned long hash) | |
1486 | { | |
1487 | hash += hash >> (8*sizeof(int)); | |
1488 | return hash; | |
1489 | } | |
1490 | ||
1491 | #else /* 32-bit case */ | |
1492 | ||
bfcfaa77 LT |
1493 | #define fold_hash(x) (x) |
1494 | ||
1495 | #endif | |
1496 | ||
1497 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1498 | { | |
1499 | unsigned long a, mask; | |
1500 | unsigned long hash = 0; | |
1501 | ||
1502 | for (;;) { | |
e419b4cc | 1503 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1504 | if (len < sizeof(unsigned long)) |
1505 | break; | |
1506 | hash += a; | |
f132c5be | 1507 | hash *= 9; |
bfcfaa77 LT |
1508 | name += sizeof(unsigned long); |
1509 | len -= sizeof(unsigned long); | |
1510 | if (!len) | |
1511 | goto done; | |
1512 | } | |
1513 | mask = ~(~0ul << len*8); | |
1514 | hash += mask & a; | |
1515 | done: | |
1516 | return fold_hash(hash); | |
1517 | } | |
1518 | EXPORT_SYMBOL(full_name_hash); | |
1519 | ||
bfcfaa77 LT |
1520 | /* |
1521 | * Calculate the length and hash of the path component, and | |
1522 | * return the length of the component; | |
1523 | */ | |
1524 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1525 | { | |
36126f8f LT |
1526 | unsigned long a, b, adata, bdata, mask, hash, len; |
1527 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; | |
bfcfaa77 LT |
1528 | |
1529 | hash = a = 0; | |
1530 | len = -sizeof(unsigned long); | |
1531 | do { | |
1532 | hash = (hash + a) * 9; | |
1533 | len += sizeof(unsigned long); | |
e419b4cc | 1534 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1535 | b = a ^ REPEAT_BYTE('/'); |
1536 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1537 | ||
1538 | adata = prep_zero_mask(a, adata, &constants); | |
1539 | bdata = prep_zero_mask(b, bdata, &constants); | |
1540 | ||
1541 | mask = create_zero_mask(adata | bdata); | |
1542 | ||
1543 | hash += a & zero_bytemask(mask); | |
bfcfaa77 LT |
1544 | *hashp = fold_hash(hash); |
1545 | ||
36126f8f | 1546 | return len + find_zero(mask); |
bfcfaa77 LT |
1547 | } |
1548 | ||
1549 | #else | |
1550 | ||
0145acc2 LT |
1551 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1552 | { | |
1553 | unsigned long hash = init_name_hash(); | |
1554 | while (len--) | |
1555 | hash = partial_name_hash(*name++, hash); | |
1556 | return end_name_hash(hash); | |
1557 | } | |
ae942ae7 | 1558 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1559 | |
200e9ef7 LT |
1560 | /* |
1561 | * We know there's a real path component here of at least | |
1562 | * one character. | |
1563 | */ | |
1564 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1565 | { | |
1566 | unsigned long hash = init_name_hash(); | |
1567 | unsigned long len = 0, c; | |
1568 | ||
1569 | c = (unsigned char)*name; | |
1570 | do { | |
1571 | len++; | |
1572 | hash = partial_name_hash(c, hash); | |
1573 | c = (unsigned char)name[len]; | |
1574 | } while (c && c != '/'); | |
1575 | *hashp = end_name_hash(hash); | |
1576 | return len; | |
1577 | } | |
1578 | ||
bfcfaa77 LT |
1579 | #endif |
1580 | ||
1da177e4 LT |
1581 | /* |
1582 | * Name resolution. | |
ea3834d9 PM |
1583 | * This is the basic name resolution function, turning a pathname into |
1584 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1585 | * |
ea3834d9 PM |
1586 | * Returns 0 and nd will have valid dentry and mnt on success. |
1587 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1588 | */ |
6de88d72 | 1589 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1590 | { |
1591 | struct path next; | |
1da177e4 | 1592 | int err; |
1da177e4 LT |
1593 | |
1594 | while (*name=='/') | |
1595 | name++; | |
1596 | if (!*name) | |
086e183a | 1597 | return 0; |
1da177e4 | 1598 | |
1da177e4 LT |
1599 | /* At this point we know we have a real path component. */ |
1600 | for(;;) { | |
1da177e4 | 1601 | struct qstr this; |
200e9ef7 | 1602 | long len; |
fe479a58 | 1603 | int type; |
1da177e4 | 1604 | |
52094c8a | 1605 | err = may_lookup(nd); |
1da177e4 LT |
1606 | if (err) |
1607 | break; | |
1608 | ||
200e9ef7 | 1609 | len = hash_name(name, &this.hash); |
1da177e4 | 1610 | this.name = name; |
200e9ef7 | 1611 | this.len = len; |
1da177e4 | 1612 | |
fe479a58 | 1613 | type = LAST_NORM; |
200e9ef7 | 1614 | if (name[0] == '.') switch (len) { |
fe479a58 | 1615 | case 2: |
200e9ef7 | 1616 | if (name[1] == '.') { |
fe479a58 | 1617 | type = LAST_DOTDOT; |
16c2cd71 AV |
1618 | nd->flags |= LOOKUP_JUMPED; |
1619 | } | |
fe479a58 AV |
1620 | break; |
1621 | case 1: | |
1622 | type = LAST_DOT; | |
1623 | } | |
5a202bcd AV |
1624 | if (likely(type == LAST_NORM)) { |
1625 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1626 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd AV |
1627 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
1628 | err = parent->d_op->d_hash(parent, nd->inode, | |
1629 | &this); | |
1630 | if (err < 0) | |
1631 | break; | |
1632 | } | |
1633 | } | |
fe479a58 | 1634 | |
200e9ef7 | 1635 | if (!name[len]) |
1da177e4 | 1636 | goto last_component; |
200e9ef7 LT |
1637 | /* |
1638 | * If it wasn't NUL, we know it was '/'. Skip that | |
1639 | * slash, and continue until no more slashes. | |
1640 | */ | |
1641 | do { | |
1642 | len++; | |
1643 | } while (unlikely(name[len] == '/')); | |
1644 | if (!name[len]) | |
b356379a | 1645 | goto last_component; |
200e9ef7 | 1646 | name += len; |
1da177e4 | 1647 | |
ce57dfc1 AV |
1648 | err = walk_component(nd, &next, &this, type, LOOKUP_FOLLOW); |
1649 | if (err < 0) | |
1650 | return err; | |
1da177e4 | 1651 | |
ce57dfc1 | 1652 | if (err) { |
b356379a | 1653 | err = nested_symlink(&next, nd); |
1da177e4 | 1654 | if (err) |
a7472bab | 1655 | return err; |
31e6b01f | 1656 | } |
3ddcd056 LT |
1657 | if (can_lookup(nd->inode)) |
1658 | continue; | |
1da177e4 | 1659 | err = -ENOTDIR; |
3ddcd056 | 1660 | break; |
1da177e4 LT |
1661 | /* here ends the main loop */ |
1662 | ||
1da177e4 | 1663 | last_component: |
b356379a AV |
1664 | nd->last = this; |
1665 | nd->last_type = type; | |
086e183a | 1666 | return 0; |
1da177e4 | 1667 | } |
951361f9 | 1668 | terminate_walk(nd); |
1da177e4 LT |
1669 | return err; |
1670 | } | |
1671 | ||
70e9b357 AV |
1672 | static int path_init(int dfd, const char *name, unsigned int flags, |
1673 | struct nameidata *nd, struct file **fp) | |
31e6b01f NP |
1674 | { |
1675 | int retval = 0; | |
1676 | int fput_needed; | |
1677 | struct file *file; | |
1678 | ||
1679 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
16c2cd71 | 1680 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 1681 | nd->depth = 0; |
5b6ca027 AV |
1682 | if (flags & LOOKUP_ROOT) { |
1683 | struct inode *inode = nd->root.dentry->d_inode; | |
73d049a4 AV |
1684 | if (*name) { |
1685 | if (!inode->i_op->lookup) | |
1686 | return -ENOTDIR; | |
1687 | retval = inode_permission(inode, MAY_EXEC); | |
1688 | if (retval) | |
1689 | return retval; | |
1690 | } | |
5b6ca027 AV |
1691 | nd->path = nd->root; |
1692 | nd->inode = inode; | |
1693 | if (flags & LOOKUP_RCU) { | |
962830df | 1694 | br_read_lock(&vfsmount_lock); |
5b6ca027 AV |
1695 | rcu_read_lock(); |
1696 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1697 | } else { | |
1698 | path_get(&nd->path); | |
1699 | } | |
1700 | return 0; | |
1701 | } | |
1702 | ||
31e6b01f | 1703 | nd->root.mnt = NULL; |
31e6b01f NP |
1704 | |
1705 | if (*name=='/') { | |
e41f7d4e | 1706 | if (flags & LOOKUP_RCU) { |
962830df | 1707 | br_read_lock(&vfsmount_lock); |
e41f7d4e AV |
1708 | rcu_read_lock(); |
1709 | set_root_rcu(nd); | |
1710 | } else { | |
1711 | set_root(nd); | |
1712 | path_get(&nd->root); | |
1713 | } | |
1714 | nd->path = nd->root; | |
31e6b01f | 1715 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1716 | if (flags & LOOKUP_RCU) { |
1717 | struct fs_struct *fs = current->fs; | |
1718 | unsigned seq; | |
31e6b01f | 1719 | |
962830df | 1720 | br_read_lock(&vfsmount_lock); |
e41f7d4e | 1721 | rcu_read_lock(); |
c28cc364 | 1722 | |
e41f7d4e AV |
1723 | do { |
1724 | seq = read_seqcount_begin(&fs->seq); | |
1725 | nd->path = fs->pwd; | |
1726 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1727 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1728 | } else { | |
1729 | get_fs_pwd(current->fs, &nd->path); | |
1730 | } | |
31e6b01f NP |
1731 | } else { |
1732 | struct dentry *dentry; | |
1733 | ||
1abf0c71 | 1734 | file = fget_raw_light(dfd, &fput_needed); |
31e6b01f NP |
1735 | retval = -EBADF; |
1736 | if (!file) | |
1737 | goto out_fail; | |
1738 | ||
1739 | dentry = file->f_path.dentry; | |
1740 | ||
f52e0c11 AV |
1741 | if (*name) { |
1742 | retval = -ENOTDIR; | |
1743 | if (!S_ISDIR(dentry->d_inode->i_mode)) | |
1744 | goto fput_fail; | |
31e6b01f | 1745 | |
4ad5abb3 | 1746 | retval = inode_permission(dentry->d_inode, MAY_EXEC); |
f52e0c11 AV |
1747 | if (retval) |
1748 | goto fput_fail; | |
1749 | } | |
31e6b01f NP |
1750 | |
1751 | nd->path = file->f_path; | |
e41f7d4e AV |
1752 | if (flags & LOOKUP_RCU) { |
1753 | if (fput_needed) | |
70e9b357 | 1754 | *fp = file; |
e41f7d4e | 1755 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
962830df | 1756 | br_read_lock(&vfsmount_lock); |
e41f7d4e AV |
1757 | rcu_read_lock(); |
1758 | } else { | |
1759 | path_get(&file->f_path); | |
1760 | fput_light(file, fput_needed); | |
1761 | } | |
31e6b01f | 1762 | } |
31e6b01f | 1763 | |
31e6b01f | 1764 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1765 | return 0; |
2dfdd266 | 1766 | |
9b4a9b14 AV |
1767 | fput_fail: |
1768 | fput_light(file, fput_needed); | |
1769 | out_fail: | |
1770 | return retval; | |
1771 | } | |
1772 | ||
bd92d7fe AV |
1773 | static inline int lookup_last(struct nameidata *nd, struct path *path) |
1774 | { | |
1775 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1776 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1777 | ||
1778 | nd->flags &= ~LOOKUP_PARENT; | |
1779 | return walk_component(nd, path, &nd->last, nd->last_type, | |
1780 | nd->flags & LOOKUP_FOLLOW); | |
1781 | } | |
1782 | ||
9b4a9b14 | 1783 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
ee0827cd | 1784 | static int path_lookupat(int dfd, const char *name, |
9b4a9b14 AV |
1785 | unsigned int flags, struct nameidata *nd) |
1786 | { | |
70e9b357 | 1787 | struct file *base = NULL; |
bd92d7fe AV |
1788 | struct path path; |
1789 | int err; | |
31e6b01f NP |
1790 | |
1791 | /* | |
1792 | * Path walking is largely split up into 2 different synchronisation | |
1793 | * schemes, rcu-walk and ref-walk (explained in | |
1794 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1795 | * path walk code, but some things particularly setup, cleanup, and | |
1796 | * following mounts are sufficiently divergent that functions are | |
1797 | * duplicated. Typically there is a function foo(), and its RCU | |
1798 | * analogue, foo_rcu(). | |
1799 | * | |
1800 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1801 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1802 | * be handled by restarting a traditional ref-walk (which will always | |
1803 | * be able to complete). | |
1804 | */ | |
bd92d7fe | 1805 | err = path_init(dfd, name, flags | LOOKUP_PARENT, nd, &base); |
ee0827cd | 1806 | |
bd92d7fe AV |
1807 | if (unlikely(err)) |
1808 | return err; | |
ee0827cd AV |
1809 | |
1810 | current->total_link_count = 0; | |
bd92d7fe AV |
1811 | err = link_path_walk(name, nd); |
1812 | ||
1813 | if (!err && !(flags & LOOKUP_PARENT)) { | |
bd92d7fe AV |
1814 | err = lookup_last(nd, &path); |
1815 | while (err > 0) { | |
1816 | void *cookie; | |
1817 | struct path link = path; | |
bd92d7fe | 1818 | nd->flags |= LOOKUP_PARENT; |
574197e0 | 1819 | err = follow_link(&link, nd, &cookie); |
6d7b5aae AV |
1820 | if (err) |
1821 | break; | |
1822 | err = lookup_last(nd, &path); | |
574197e0 | 1823 | put_link(nd, &link, cookie); |
bd92d7fe AV |
1824 | } |
1825 | } | |
ee0827cd | 1826 | |
9f1fafee AV |
1827 | if (!err) |
1828 | err = complete_walk(nd); | |
bd92d7fe AV |
1829 | |
1830 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
1831 | if (!nd->inode->i_op->lookup) { | |
1832 | path_put(&nd->path); | |
bd23a539 | 1833 | err = -ENOTDIR; |
bd92d7fe AV |
1834 | } |
1835 | } | |
16c2cd71 | 1836 | |
70e9b357 AV |
1837 | if (base) |
1838 | fput(base); | |
ee0827cd | 1839 | |
5b6ca027 | 1840 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
2a737871 AV |
1841 | path_put(&nd->root); |
1842 | nd->root.mnt = NULL; | |
1843 | } | |
bd92d7fe | 1844 | return err; |
ee0827cd | 1845 | } |
31e6b01f | 1846 | |
ee0827cd AV |
1847 | static int do_path_lookup(int dfd, const char *name, |
1848 | unsigned int flags, struct nameidata *nd) | |
1849 | { | |
1850 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); | |
1851 | if (unlikely(retval == -ECHILD)) | |
1852 | retval = path_lookupat(dfd, name, flags, nd); | |
1853 | if (unlikely(retval == -ESTALE)) | |
1854 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); | |
31e6b01f NP |
1855 | |
1856 | if (likely(!retval)) { | |
1857 | if (unlikely(!audit_dummy_context())) { | |
1858 | if (nd->path.dentry && nd->inode) | |
1859 | audit_inode(name, nd->path.dentry); | |
1860 | } | |
1861 | } | |
170aa3d0 | 1862 | return retval; |
1da177e4 LT |
1863 | } |
1864 | ||
79714f72 AV |
1865 | /* does lookup, returns the object with parent locked */ |
1866 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 1867 | { |
79714f72 AV |
1868 | struct nameidata nd; |
1869 | struct dentry *d; | |
1870 | int err = do_path_lookup(AT_FDCWD, name, LOOKUP_PARENT, &nd); | |
1871 | if (err) | |
1872 | return ERR_PTR(err); | |
1873 | if (nd.last_type != LAST_NORM) { | |
1874 | path_put(&nd.path); | |
1875 | return ERR_PTR(-EINVAL); | |
1876 | } | |
1877 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); | |
1e0ea001 | 1878 | d = __lookup_hash(&nd.last, nd.path.dentry, 0); |
79714f72 AV |
1879 | if (IS_ERR(d)) { |
1880 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); | |
1881 | path_put(&nd.path); | |
1882 | return d; | |
1883 | } | |
1884 | *path = nd.path; | |
1885 | return d; | |
5590ff0d UD |
1886 | } |
1887 | ||
d1811465 AV |
1888 | int kern_path(const char *name, unsigned int flags, struct path *path) |
1889 | { | |
1890 | struct nameidata nd; | |
1891 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
1892 | if (!res) | |
1893 | *path = nd.path; | |
1894 | return res; | |
1895 | } | |
1896 | ||
16f18200 JJS |
1897 | /** |
1898 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
1899 | * @dentry: pointer to dentry of the base directory | |
1900 | * @mnt: pointer to vfs mount of the base directory | |
1901 | * @name: pointer to file name | |
1902 | * @flags: lookup flags | |
e0a01249 | 1903 | * @path: pointer to struct path to fill |
16f18200 JJS |
1904 | */ |
1905 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
1906 | const char *name, unsigned int flags, | |
e0a01249 | 1907 | struct path *path) |
16f18200 | 1908 | { |
e0a01249 AV |
1909 | struct nameidata nd; |
1910 | int err; | |
1911 | nd.root.dentry = dentry; | |
1912 | nd.root.mnt = mnt; | |
1913 | BUG_ON(flags & LOOKUP_PARENT); | |
5b6ca027 | 1914 | /* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */ |
e0a01249 AV |
1915 | err = do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, &nd); |
1916 | if (!err) | |
1917 | *path = nd.path; | |
1918 | return err; | |
16f18200 JJS |
1919 | } |
1920 | ||
057f6c01 JM |
1921 | /* |
1922 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
1923 | * needs parent already locked. Doesn't follow mounts. | |
1924 | * SMP-safe. | |
1925 | */ | |
eead1911 | 1926 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 1927 | { |
72bd866a | 1928 | return __lookup_hash(&nd->last, nd->path.dentry, nd->flags); |
1da177e4 LT |
1929 | } |
1930 | ||
eead1911 | 1931 | /** |
a6b91919 | 1932 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
1933 | * @name: pathname component to lookup |
1934 | * @base: base directory to lookup from | |
1935 | * @len: maximum length @len should be interpreted to | |
1936 | * | |
a6b91919 RD |
1937 | * Note that this routine is purely a helper for filesystem usage and should |
1938 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
1939 | * nameidata argument is passed to the filesystem methods and a filesystem |
1940 | * using this helper needs to be prepared for that. | |
1941 | */ | |
057f6c01 JM |
1942 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
1943 | { | |
057f6c01 | 1944 | struct qstr this; |
6a96ba54 | 1945 | unsigned int c; |
cda309de | 1946 | int err; |
057f6c01 | 1947 | |
2f9092e1 DW |
1948 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
1949 | ||
6a96ba54 AV |
1950 | this.name = name; |
1951 | this.len = len; | |
0145acc2 | 1952 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
1953 | if (!len) |
1954 | return ERR_PTR(-EACCES); | |
1955 | ||
6a96ba54 AV |
1956 | while (len--) { |
1957 | c = *(const unsigned char *)name++; | |
1958 | if (c == '/' || c == '\0') | |
1959 | return ERR_PTR(-EACCES); | |
6a96ba54 | 1960 | } |
5a202bcd AV |
1961 | /* |
1962 | * See if the low-level filesystem might want | |
1963 | * to use its own hash.. | |
1964 | */ | |
1965 | if (base->d_flags & DCACHE_OP_HASH) { | |
1966 | int err = base->d_op->d_hash(base, base->d_inode, &this); | |
1967 | if (err < 0) | |
1968 | return ERR_PTR(err); | |
1969 | } | |
eead1911 | 1970 | |
cda309de MS |
1971 | err = inode_permission(base->d_inode, MAY_EXEC); |
1972 | if (err) | |
1973 | return ERR_PTR(err); | |
1974 | ||
72bd866a | 1975 | return __lookup_hash(&this, base, 0); |
057f6c01 JM |
1976 | } |
1977 | ||
1fa1e7f6 AW |
1978 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
1979 | struct path *path, int *empty) | |
1da177e4 | 1980 | { |
2d8f3038 | 1981 | struct nameidata nd; |
1fa1e7f6 | 1982 | char *tmp = getname_flags(name, flags, empty); |
1da177e4 | 1983 | int err = PTR_ERR(tmp); |
1da177e4 | 1984 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
1985 | |
1986 | BUG_ON(flags & LOOKUP_PARENT); | |
1987 | ||
1988 | err = do_path_lookup(dfd, tmp, flags, &nd); | |
1da177e4 | 1989 | putname(tmp); |
2d8f3038 AV |
1990 | if (!err) |
1991 | *path = nd.path; | |
1da177e4 LT |
1992 | } |
1993 | return err; | |
1994 | } | |
1995 | ||
1fa1e7f6 AW |
1996 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
1997 | struct path *path) | |
1998 | { | |
f7493e5d | 1999 | return user_path_at_empty(dfd, name, flags, path, NULL); |
1fa1e7f6 AW |
2000 | } |
2001 | ||
2ad94ae6 AV |
2002 | static int user_path_parent(int dfd, const char __user *path, |
2003 | struct nameidata *nd, char **name) | |
2004 | { | |
2005 | char *s = getname(path); | |
2006 | int error; | |
2007 | ||
2008 | if (IS_ERR(s)) | |
2009 | return PTR_ERR(s); | |
2010 | ||
2011 | error = do_path_lookup(dfd, s, LOOKUP_PARENT, nd); | |
2012 | if (error) | |
2013 | putname(s); | |
2014 | else | |
2015 | *name = s; | |
2016 | ||
2017 | return error; | |
2018 | } | |
2019 | ||
1da177e4 LT |
2020 | /* |
2021 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
2022 | * minimal. | |
2023 | */ | |
2024 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
2025 | { | |
8e96e3b7 | 2026 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2027 | |
1da177e4 LT |
2028 | if (!(dir->i_mode & S_ISVTX)) |
2029 | return 0; | |
8e96e3b7 | 2030 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2031 | return 0; |
8e96e3b7 | 2032 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2033 | return 0; |
1a48e2ac | 2034 | return !inode_capable(inode, CAP_FOWNER); |
1da177e4 LT |
2035 | } |
2036 | ||
2037 | /* | |
2038 | * Check whether we can remove a link victim from directory dir, check | |
2039 | * whether the type of victim is right. | |
2040 | * 1. We can't do it if dir is read-only (done in permission()) | |
2041 | * 2. We should have write and exec permissions on dir | |
2042 | * 3. We can't remove anything from append-only dir | |
2043 | * 4. We can't do anything with immutable dir (done in permission()) | |
2044 | * 5. If the sticky bit on dir is set we should either | |
2045 | * a. be owner of dir, or | |
2046 | * b. be owner of victim, or | |
2047 | * c. have CAP_FOWNER capability | |
2048 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2049 | * links pointing to it. | |
2050 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2051 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2052 | * 9. We can't remove a root or mountpoint. | |
2053 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
2054 | * nfs_async_unlink(). | |
2055 | */ | |
858119e1 | 2056 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
2057 | { |
2058 | int error; | |
2059 | ||
2060 | if (!victim->d_inode) | |
2061 | return -ENOENT; | |
2062 | ||
2063 | BUG_ON(victim->d_parent->d_inode != dir); | |
cccc6bba | 2064 | audit_inode_child(victim, dir); |
1da177e4 | 2065 | |
f419a2e3 | 2066 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2067 | if (error) |
2068 | return error; | |
2069 | if (IS_APPEND(dir)) | |
2070 | return -EPERM; | |
2071 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 2072 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
2073 | return -EPERM; |
2074 | if (isdir) { | |
2075 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
2076 | return -ENOTDIR; | |
2077 | if (IS_ROOT(victim)) | |
2078 | return -EBUSY; | |
2079 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
2080 | return -EISDIR; | |
2081 | if (IS_DEADDIR(dir)) | |
2082 | return -ENOENT; | |
2083 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2084 | return -EBUSY; | |
2085 | return 0; | |
2086 | } | |
2087 | ||
2088 | /* Check whether we can create an object with dentry child in directory | |
2089 | * dir. | |
2090 | * 1. We can't do it if child already exists (open has special treatment for | |
2091 | * this case, but since we are inlined it's OK) | |
2092 | * 2. We can't do it if dir is read-only (done in permission()) | |
2093 | * 3. We should have write and exec permissions on dir | |
2094 | * 4. We can't do it if dir is immutable (done in permission()) | |
2095 | */ | |
a95164d9 | 2096 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
2097 | { |
2098 | if (child->d_inode) | |
2099 | return -EEXIST; | |
2100 | if (IS_DEADDIR(dir)) | |
2101 | return -ENOENT; | |
f419a2e3 | 2102 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2103 | } |
2104 | ||
1da177e4 LT |
2105 | /* |
2106 | * p1 and p2 should be directories on the same fs. | |
2107 | */ | |
2108 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2109 | { | |
2110 | struct dentry *p; | |
2111 | ||
2112 | if (p1 == p2) { | |
f2eace23 | 2113 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2114 | return NULL; |
2115 | } | |
2116 | ||
a11f3a05 | 2117 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2118 | |
e2761a11 OH |
2119 | p = d_ancestor(p2, p1); |
2120 | if (p) { | |
2121 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2122 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2123 | return p; | |
1da177e4 LT |
2124 | } |
2125 | ||
e2761a11 OH |
2126 | p = d_ancestor(p1, p2); |
2127 | if (p) { | |
2128 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2129 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2130 | return p; | |
1da177e4 LT |
2131 | } |
2132 | ||
f2eace23 IM |
2133 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
2134 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
2135 | return NULL; |
2136 | } | |
2137 | ||
2138 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2139 | { | |
1b1dcc1b | 2140 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2141 | if (p1 != p2) { |
1b1dcc1b | 2142 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2143 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2144 | } |
2145 | } | |
2146 | ||
4acdaf27 | 2147 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2148 | bool want_excl) |
1da177e4 | 2149 | { |
a95164d9 | 2150 | int error = may_create(dir, dentry); |
1da177e4 LT |
2151 | if (error) |
2152 | return error; | |
2153 | ||
acfa4380 | 2154 | if (!dir->i_op->create) |
1da177e4 LT |
2155 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2156 | mode &= S_IALLUGO; | |
2157 | mode |= S_IFREG; | |
2158 | error = security_inode_create(dir, dentry, mode); | |
2159 | if (error) | |
2160 | return error; | |
312b63fb | 2161 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2162 | if (!error) |
f38aa942 | 2163 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2164 | return error; |
2165 | } | |
2166 | ||
73d049a4 | 2167 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2168 | { |
3fb64190 | 2169 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2170 | struct inode *inode = dentry->d_inode; |
2171 | int error; | |
2172 | ||
bcda7652 AV |
2173 | /* O_PATH? */ |
2174 | if (!acc_mode) | |
2175 | return 0; | |
2176 | ||
1da177e4 LT |
2177 | if (!inode) |
2178 | return -ENOENT; | |
2179 | ||
c8fe8f30 CH |
2180 | switch (inode->i_mode & S_IFMT) { |
2181 | case S_IFLNK: | |
1da177e4 | 2182 | return -ELOOP; |
c8fe8f30 CH |
2183 | case S_IFDIR: |
2184 | if (acc_mode & MAY_WRITE) | |
2185 | return -EISDIR; | |
2186 | break; | |
2187 | case S_IFBLK: | |
2188 | case S_IFCHR: | |
3fb64190 | 2189 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2190 | return -EACCES; |
c8fe8f30 CH |
2191 | /*FALLTHRU*/ |
2192 | case S_IFIFO: | |
2193 | case S_IFSOCK: | |
1da177e4 | 2194 | flag &= ~O_TRUNC; |
c8fe8f30 | 2195 | break; |
4a3fd211 | 2196 | } |
b41572e9 | 2197 | |
3fb64190 | 2198 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2199 | if (error) |
2200 | return error; | |
6146f0d5 | 2201 | |
1da177e4 LT |
2202 | /* |
2203 | * An append-only file must be opened in append mode for writing. | |
2204 | */ | |
2205 | if (IS_APPEND(inode)) { | |
8737c930 | 2206 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2207 | return -EPERM; |
1da177e4 | 2208 | if (flag & O_TRUNC) |
7715b521 | 2209 | return -EPERM; |
1da177e4 LT |
2210 | } |
2211 | ||
2212 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2213 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2214 | return -EPERM; |
1da177e4 | 2215 | |
f3c7691e | 2216 | return 0; |
7715b521 | 2217 | } |
1da177e4 | 2218 | |
e1181ee6 | 2219 | static int handle_truncate(struct file *filp) |
7715b521 | 2220 | { |
e1181ee6 | 2221 | struct path *path = &filp->f_path; |
7715b521 AV |
2222 | struct inode *inode = path->dentry->d_inode; |
2223 | int error = get_write_access(inode); | |
2224 | if (error) | |
2225 | return error; | |
2226 | /* | |
2227 | * Refuse to truncate files with mandatory locks held on them. | |
2228 | */ | |
2229 | error = locks_verify_locked(inode); | |
2230 | if (!error) | |
ea0d3ab2 | 2231 | error = security_path_truncate(path); |
7715b521 AV |
2232 | if (!error) { |
2233 | error = do_truncate(path->dentry, 0, | |
2234 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2235 | filp); |
7715b521 AV |
2236 | } |
2237 | put_write_access(inode); | |
acd0c935 | 2238 | return error; |
1da177e4 LT |
2239 | } |
2240 | ||
d57999e1 DH |
2241 | static inline int open_to_namei_flags(int flag) |
2242 | { | |
8a5e929d AV |
2243 | if ((flag & O_ACCMODE) == 3) |
2244 | flag--; | |
d57999e1 DH |
2245 | return flag; |
2246 | } | |
2247 | ||
d18e9008 MS |
2248 | static int may_o_create(struct path *dir, struct dentry *dentry, umode_t mode) |
2249 | { | |
2250 | int error = security_path_mknod(dir, dentry, mode, 0); | |
2251 | if (error) | |
2252 | return error; | |
2253 | ||
2254 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); | |
2255 | if (error) | |
2256 | return error; | |
2257 | ||
2258 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2259 | } | |
2260 | ||
1acf0af9 DH |
2261 | /* |
2262 | * Attempt to atomically look up, create and open a file from a negative | |
2263 | * dentry. | |
2264 | * | |
2265 | * Returns 0 if successful. The file will have been created and attached to | |
2266 | * @file by the filesystem calling finish_open(). | |
2267 | * | |
2268 | * Returns 1 if the file was looked up only or didn't need creating. The | |
2269 | * caller will need to perform the open themselves. @path will have been | |
2270 | * updated to point to the new dentry. This may be negative. | |
2271 | * | |
2272 | * Returns an error code otherwise. | |
2273 | */ | |
2675a4eb AV |
2274 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
2275 | struct path *path, struct file *file, | |
2276 | const struct open_flags *op, | |
2277 | bool *want_write, bool need_lookup, | |
2278 | int *opened) | |
d18e9008 MS |
2279 | { |
2280 | struct inode *dir = nd->path.dentry->d_inode; | |
2281 | unsigned open_flag = open_to_namei_flags(op->open_flag); | |
2282 | umode_t mode; | |
2283 | int error; | |
2284 | int acc_mode; | |
d18e9008 MS |
2285 | int create_error = 0; |
2286 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; | |
2287 | ||
2288 | BUG_ON(dentry->d_inode); | |
2289 | ||
2290 | /* Don't create child dentry for a dead directory. */ | |
2291 | if (unlikely(IS_DEADDIR(dir))) { | |
2675a4eb | 2292 | error = -ENOENT; |
d18e9008 MS |
2293 | goto out; |
2294 | } | |
2295 | ||
2296 | mode = op->mode & S_IALLUGO; | |
2297 | if ((open_flag & O_CREAT) && !IS_POSIXACL(dir)) | |
2298 | mode &= ~current_umask(); | |
2299 | ||
2300 | if (open_flag & O_EXCL) { | |
2301 | open_flag &= ~O_TRUNC; | |
47237687 | 2302 | *opened |= FILE_CREATED; |
d18e9008 MS |
2303 | } |
2304 | ||
2305 | /* | |
2306 | * Checking write permission is tricky, bacuse we don't know if we are | |
2307 | * going to actually need it: O_CREAT opens should work as long as the | |
2308 | * file exists. But checking existence breaks atomicity. The trick is | |
2309 | * to check access and if not granted clear O_CREAT from the flags. | |
2310 | * | |
2311 | * Another problem is returing the "right" error value (e.g. for an | |
2312 | * O_EXCL open we want to return EEXIST not EROFS). | |
2313 | */ | |
2314 | if ((open_flag & (O_CREAT | O_TRUNC)) || | |
2315 | (open_flag & O_ACCMODE) != O_RDONLY) { | |
2316 | error = mnt_want_write(nd->path.mnt); | |
2317 | if (!error) { | |
77d660a8 | 2318 | *want_write = true; |
d18e9008 MS |
2319 | } else if (!(open_flag & O_CREAT)) { |
2320 | /* | |
2321 | * No O_CREATE -> atomicity not a requirement -> fall | |
2322 | * back to lookup + open | |
2323 | */ | |
2324 | goto no_open; | |
2325 | } else if (open_flag & (O_EXCL | O_TRUNC)) { | |
2326 | /* Fall back and fail with the right error */ | |
2327 | create_error = error; | |
2328 | goto no_open; | |
2329 | } else { | |
2330 | /* No side effects, safe to clear O_CREAT */ | |
2331 | create_error = error; | |
2332 | open_flag &= ~O_CREAT; | |
2333 | } | |
2334 | } | |
2335 | ||
2336 | if (open_flag & O_CREAT) { | |
2337 | error = may_o_create(&nd->path, dentry, op->mode); | |
2338 | if (error) { | |
2339 | create_error = error; | |
2340 | if (open_flag & O_EXCL) | |
2341 | goto no_open; | |
2342 | open_flag &= ~O_CREAT; | |
2343 | } | |
2344 | } | |
2345 | ||
2346 | if (nd->flags & LOOKUP_DIRECTORY) | |
2347 | open_flag |= O_DIRECTORY; | |
2348 | ||
30d90494 AV |
2349 | file->f_path.dentry = DENTRY_NOT_SET; |
2350 | file->f_path.mnt = nd->path.mnt; | |
2351 | error = dir->i_op->atomic_open(dir, dentry, file, open_flag, mode, | |
47237687 | 2352 | opened); |
d9585277 | 2353 | if (error < 0) { |
d9585277 AV |
2354 | if (create_error && error == -ENOENT) |
2355 | error = create_error; | |
d18e9008 MS |
2356 | goto out; |
2357 | } | |
2358 | ||
2359 | acc_mode = op->acc_mode; | |
47237687 | 2360 | if (*opened & FILE_CREATED) { |
d18e9008 MS |
2361 | fsnotify_create(dir, dentry); |
2362 | acc_mode = MAY_OPEN; | |
2363 | } | |
2364 | ||
d9585277 | 2365 | if (error) { /* returned 1, that is */ |
30d90494 | 2366 | if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 2367 | error = -EIO; |
d18e9008 MS |
2368 | goto out; |
2369 | } | |
30d90494 | 2370 | if (file->f_path.dentry) { |
d18e9008 | 2371 | dput(dentry); |
30d90494 | 2372 | dentry = file->f_path.dentry; |
d18e9008 MS |
2373 | } |
2374 | goto looked_up; | |
2375 | } | |
2376 | ||
2377 | /* | |
2378 | * We didn't have the inode before the open, so check open permission | |
2379 | * here. | |
2380 | */ | |
2675a4eb AV |
2381 | error = may_open(&file->f_path, acc_mode, open_flag); |
2382 | if (error) | |
2383 | fput(file); | |
d18e9008 MS |
2384 | |
2385 | out: | |
2386 | dput(dentry); | |
2675a4eb | 2387 | return error; |
d18e9008 | 2388 | |
d18e9008 MS |
2389 | no_open: |
2390 | if (need_lookup) { | |
72bd866a | 2391 | dentry = lookup_real(dir, dentry, nd->flags); |
d18e9008 | 2392 | if (IS_ERR(dentry)) |
2675a4eb | 2393 | return PTR_ERR(dentry); |
d18e9008 MS |
2394 | |
2395 | if (create_error) { | |
2396 | int open_flag = op->open_flag; | |
2397 | ||
2675a4eb | 2398 | error = create_error; |
d18e9008 MS |
2399 | if ((open_flag & O_EXCL)) { |
2400 | if (!dentry->d_inode) | |
2401 | goto out; | |
2402 | } else if (!dentry->d_inode) { | |
2403 | goto out; | |
2404 | } else if ((open_flag & O_TRUNC) && | |
2405 | S_ISREG(dentry->d_inode->i_mode)) { | |
2406 | goto out; | |
2407 | } | |
2408 | /* will fail later, go on to get the right error */ | |
2409 | } | |
2410 | } | |
2411 | looked_up: | |
2412 | path->dentry = dentry; | |
2413 | path->mnt = nd->path.mnt; | |
2675a4eb | 2414 | return 1; |
d18e9008 MS |
2415 | } |
2416 | ||
d58ffd35 | 2417 | /* |
1acf0af9 | 2418 | * Look up and maybe create and open the last component. |
d58ffd35 MS |
2419 | * |
2420 | * Must be called with i_mutex held on parent. | |
2421 | * | |
1acf0af9 DH |
2422 | * Returns 0 if the file was successfully atomically created (if necessary) and |
2423 | * opened. In this case the file will be returned attached to @file. | |
2424 | * | |
2425 | * Returns 1 if the file was not completely opened at this time, though lookups | |
2426 | * and creations will have been performed and the dentry returned in @path will | |
2427 | * be positive upon return if O_CREAT was specified. If O_CREAT wasn't | |
2428 | * specified then a negative dentry may be returned. | |
2429 | * | |
2430 | * An error code is returned otherwise. | |
2431 | * | |
2432 | * FILE_CREATE will be set in @*opened if the dentry was created and will be | |
2433 | * cleared otherwise prior to returning. | |
d58ffd35 | 2434 | */ |
2675a4eb AV |
2435 | static int lookup_open(struct nameidata *nd, struct path *path, |
2436 | struct file *file, | |
2437 | const struct open_flags *op, | |
2438 | bool *want_write, int *opened) | |
d58ffd35 MS |
2439 | { |
2440 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 2441 | struct inode *dir_inode = dir->d_inode; |
d58ffd35 MS |
2442 | struct dentry *dentry; |
2443 | int error; | |
54ef4872 | 2444 | bool need_lookup; |
d58ffd35 | 2445 | |
47237687 | 2446 | *opened &= ~FILE_CREATED; |
201f956e | 2447 | dentry = lookup_dcache(&nd->last, dir, nd->flags, &need_lookup); |
d58ffd35 | 2448 | if (IS_ERR(dentry)) |
2675a4eb | 2449 | return PTR_ERR(dentry); |
d58ffd35 | 2450 | |
d18e9008 MS |
2451 | /* Cached positive dentry: will open in f_op->open */ |
2452 | if (!need_lookup && dentry->d_inode) | |
2453 | goto out_no_open; | |
2454 | ||
2455 | if ((nd->flags & LOOKUP_OPEN) && dir_inode->i_op->atomic_open) { | |
30d90494 | 2456 | return atomic_open(nd, dentry, path, file, op, want_write, |
47237687 | 2457 | need_lookup, opened); |
d18e9008 MS |
2458 | } |
2459 | ||
54ef4872 MS |
2460 | if (need_lookup) { |
2461 | BUG_ON(dentry->d_inode); | |
2462 | ||
72bd866a | 2463 | dentry = lookup_real(dir_inode, dentry, nd->flags); |
54ef4872 | 2464 | if (IS_ERR(dentry)) |
2675a4eb | 2465 | return PTR_ERR(dentry); |
54ef4872 MS |
2466 | } |
2467 | ||
d58ffd35 MS |
2468 | /* Negative dentry, just create the file */ |
2469 | if (!dentry->d_inode && (op->open_flag & O_CREAT)) { | |
2470 | umode_t mode = op->mode; | |
2471 | if (!IS_POSIXACL(dir->d_inode)) | |
2472 | mode &= ~current_umask(); | |
2473 | /* | |
2474 | * This write is needed to ensure that a | |
2475 | * rw->ro transition does not occur between | |
2476 | * the time when the file is created and when | |
2477 | * a permanent write count is taken through | |
015c3bbc | 2478 | * the 'struct file' in finish_open(). |
d58ffd35 MS |
2479 | */ |
2480 | error = mnt_want_write(nd->path.mnt); | |
2481 | if (error) | |
2482 | goto out_dput; | |
77d660a8 | 2483 | *want_write = true; |
47237687 | 2484 | *opened |= FILE_CREATED; |
d58ffd35 MS |
2485 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2486 | if (error) | |
2487 | goto out_dput; | |
312b63fb AV |
2488 | error = vfs_create(dir->d_inode, dentry, mode, |
2489 | nd->flags & LOOKUP_EXCL); | |
d58ffd35 MS |
2490 | if (error) |
2491 | goto out_dput; | |
2492 | } | |
d18e9008 | 2493 | out_no_open: |
d58ffd35 MS |
2494 | path->dentry = dentry; |
2495 | path->mnt = nd->path.mnt; | |
2675a4eb | 2496 | return 1; |
d58ffd35 MS |
2497 | |
2498 | out_dput: | |
2499 | dput(dentry); | |
2675a4eb | 2500 | return error; |
d58ffd35 MS |
2501 | } |
2502 | ||
31e6b01f | 2503 | /* |
fe2d35ff | 2504 | * Handle the last step of open() |
31e6b01f | 2505 | */ |
2675a4eb AV |
2506 | static int do_last(struct nameidata *nd, struct path *path, |
2507 | struct file *file, const struct open_flags *op, | |
2508 | int *opened, const char *pathname) | |
fb1cc555 | 2509 | { |
a1e28038 | 2510 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 2511 | int open_flag = op->open_flag; |
77d660a8 MS |
2512 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
2513 | bool want_write = false; | |
bcda7652 | 2514 | int acc_mode = op->acc_mode; |
a1eb3315 | 2515 | struct inode *inode; |
77d660a8 | 2516 | bool symlink_ok = false; |
16b1c1cd MS |
2517 | struct path save_parent = { .dentry = NULL, .mnt = NULL }; |
2518 | bool retried = false; | |
16c2cd71 | 2519 | int error; |
1f36f774 | 2520 | |
c3e380b0 AV |
2521 | nd->flags &= ~LOOKUP_PARENT; |
2522 | nd->flags |= op->intent; | |
2523 | ||
1f36f774 AV |
2524 | switch (nd->last_type) { |
2525 | case LAST_DOTDOT: | |
176306f5 | 2526 | case LAST_DOT: |
fe2d35ff AV |
2527 | error = handle_dots(nd, nd->last_type); |
2528 | if (error) | |
2675a4eb | 2529 | return error; |
1f36f774 | 2530 | /* fallthrough */ |
1f36f774 | 2531 | case LAST_ROOT: |
9f1fafee | 2532 | error = complete_walk(nd); |
16c2cd71 | 2533 | if (error) |
2675a4eb | 2534 | return error; |
fe2d35ff | 2535 | audit_inode(pathname, nd->path.dentry); |
ca344a89 | 2536 | if (open_flag & O_CREAT) { |
fe2d35ff | 2537 | error = -EISDIR; |
2675a4eb | 2538 | goto out; |
fe2d35ff | 2539 | } |
e83db167 | 2540 | goto finish_open; |
1f36f774 | 2541 | case LAST_BIND: |
9f1fafee | 2542 | error = complete_walk(nd); |
16c2cd71 | 2543 | if (error) |
2675a4eb | 2544 | return error; |
1f36f774 | 2545 | audit_inode(pathname, dir); |
e83db167 | 2546 | goto finish_open; |
1f36f774 | 2547 | } |
67ee3ad2 | 2548 | |
ca344a89 | 2549 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
2550 | if (nd->last.name[nd->last.len]) |
2551 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
bcda7652 | 2552 | if (open_flag & O_PATH && !(nd->flags & LOOKUP_FOLLOW)) |
77d660a8 | 2553 | symlink_ok = true; |
fe2d35ff | 2554 | /* we _can_ be in RCU mode here */ |
a1eb3315 | 2555 | error = lookup_fast(nd, &nd->last, path, &inode); |
71574865 MS |
2556 | if (likely(!error)) |
2557 | goto finish_lookup; | |
2558 | ||
2559 | if (error < 0) | |
2675a4eb | 2560 | goto out; |
71574865 MS |
2561 | |
2562 | BUG_ON(nd->inode != dir->d_inode); | |
b6183df7 MS |
2563 | } else { |
2564 | /* create side of things */ | |
2565 | /* | |
2566 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
2567 | * has been cleared when we got to the last component we are | |
2568 | * about to look up | |
2569 | */ | |
2570 | error = complete_walk(nd); | |
2571 | if (error) | |
2675a4eb | 2572 | return error; |
fe2d35ff | 2573 | |
b6183df7 MS |
2574 | audit_inode(pathname, dir); |
2575 | error = -EISDIR; | |
2576 | /* trailing slashes? */ | |
2577 | if (nd->last.name[nd->last.len]) | |
2675a4eb | 2578 | goto out; |
b6183df7 | 2579 | } |
a2c36b45 | 2580 | |
16b1c1cd | 2581 | retry_lookup: |
a1e28038 | 2582 | mutex_lock(&dir->d_inode->i_mutex); |
2675a4eb | 2583 | error = lookup_open(nd, path, file, op, &want_write, opened); |
d58ffd35 | 2584 | mutex_unlock(&dir->d_inode->i_mutex); |
a1e28038 | 2585 | |
2675a4eb AV |
2586 | if (error <= 0) { |
2587 | if (error) | |
d18e9008 MS |
2588 | goto out; |
2589 | ||
47237687 | 2590 | if ((*opened & FILE_CREATED) || |
2675a4eb | 2591 | !S_ISREG(file->f_path.dentry->d_inode->i_mode)) |
77d660a8 | 2592 | will_truncate = false; |
d18e9008 | 2593 | |
2675a4eb | 2594 | audit_inode(pathname, file->f_path.dentry); |
d18e9008 MS |
2595 | goto opened; |
2596 | } | |
fb1cc555 | 2597 | |
47237687 | 2598 | if (*opened & FILE_CREATED) { |
9b44f1b3 | 2599 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 2600 | open_flag &= ~O_TRUNC; |
77d660a8 | 2601 | will_truncate = false; |
bcda7652 | 2602 | acc_mode = MAY_OPEN; |
d58ffd35 | 2603 | path_to_nameidata(path, nd); |
e83db167 | 2604 | goto finish_open_created; |
fb1cc555 AV |
2605 | } |
2606 | ||
2607 | /* | |
2608 | * It already exists. | |
2609 | */ | |
fb1cc555 AV |
2610 | audit_inode(pathname, path->dentry); |
2611 | ||
d18e9008 MS |
2612 | /* |
2613 | * If atomic_open() acquired write access it is dropped now due to | |
2614 | * possible mount and symlink following (this might be optimized away if | |
2615 | * necessary...) | |
2616 | */ | |
2617 | if (want_write) { | |
2618 | mnt_drop_write(nd->path.mnt); | |
77d660a8 | 2619 | want_write = false; |
d18e9008 MS |
2620 | } |
2621 | ||
fb1cc555 | 2622 | error = -EEXIST; |
ca344a89 | 2623 | if (open_flag & O_EXCL) |
fb1cc555 AV |
2624 | goto exit_dput; |
2625 | ||
9875cf80 DH |
2626 | error = follow_managed(path, nd->flags); |
2627 | if (error < 0) | |
2628 | goto exit_dput; | |
fb1cc555 | 2629 | |
a3fbbde7 AV |
2630 | if (error) |
2631 | nd->flags |= LOOKUP_JUMPED; | |
2632 | ||
decf3400 MS |
2633 | BUG_ON(nd->flags & LOOKUP_RCU); |
2634 | inode = path->dentry->d_inode; | |
5f5daac1 MS |
2635 | finish_lookup: |
2636 | /* we _can_ be in RCU mode here */ | |
fb1cc555 | 2637 | error = -ENOENT; |
54c33e7f MS |
2638 | if (!inode) { |
2639 | path_to_nameidata(path, nd); | |
2675a4eb | 2640 | goto out; |
54c33e7f | 2641 | } |
9e67f361 | 2642 | |
d45ea867 MS |
2643 | if (should_follow_link(inode, !symlink_ok)) { |
2644 | if (nd->flags & LOOKUP_RCU) { | |
2645 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
2646 | error = -ECHILD; | |
2675a4eb | 2647 | goto out; |
d45ea867 MS |
2648 | } |
2649 | } | |
2650 | BUG_ON(inode != path->dentry->d_inode); | |
2675a4eb | 2651 | return 1; |
d45ea867 | 2652 | } |
fb1cc555 | 2653 | |
16b1c1cd MS |
2654 | if ((nd->flags & LOOKUP_RCU) || nd->path.mnt != path->mnt) { |
2655 | path_to_nameidata(path, nd); | |
2656 | } else { | |
2657 | save_parent.dentry = nd->path.dentry; | |
2658 | save_parent.mnt = mntget(path->mnt); | |
2659 | nd->path.dentry = path->dentry; | |
2660 | ||
2661 | } | |
decf3400 | 2662 | nd->inode = inode; |
a3fbbde7 AV |
2663 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
2664 | error = complete_walk(nd); | |
16b1c1cd MS |
2665 | if (error) { |
2666 | path_put(&save_parent); | |
2675a4eb | 2667 | return error; |
16b1c1cd | 2668 | } |
fb1cc555 | 2669 | error = -EISDIR; |
050ac841 | 2670 | if ((open_flag & O_CREAT) && S_ISDIR(nd->inode->i_mode)) |
2675a4eb | 2671 | goto out; |
af2f5542 MS |
2672 | error = -ENOTDIR; |
2673 | if ((nd->flags & LOOKUP_DIRECTORY) && !nd->inode->i_op->lookup) | |
2675a4eb | 2674 | goto out; |
d7fdd7f6 | 2675 | audit_inode(pathname, nd->path.dentry); |
e83db167 | 2676 | finish_open: |
6c0d46c4 | 2677 | if (!S_ISREG(nd->inode->i_mode)) |
77d660a8 | 2678 | will_truncate = false; |
6c0d46c4 | 2679 | |
0f9d1a10 AV |
2680 | if (will_truncate) { |
2681 | error = mnt_want_write(nd->path.mnt); | |
2682 | if (error) | |
2675a4eb | 2683 | goto out; |
77d660a8 | 2684 | want_write = true; |
0f9d1a10 | 2685 | } |
e83db167 | 2686 | finish_open_created: |
bcda7652 | 2687 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 2688 | if (error) |
2675a4eb | 2689 | goto out; |
30d90494 AV |
2690 | file->f_path.mnt = nd->path.mnt; |
2691 | error = finish_open(file, nd->path.dentry, NULL, opened); | |
2692 | if (error) { | |
30d90494 | 2693 | if (error == -EOPENSTALE) |
f60dc3db | 2694 | goto stale_open; |
015c3bbc | 2695 | goto out; |
f60dc3db | 2696 | } |
a8277b9b | 2697 | opened: |
2675a4eb | 2698 | error = open_check_o_direct(file); |
015c3bbc MS |
2699 | if (error) |
2700 | goto exit_fput; | |
2675a4eb | 2701 | error = ima_file_check(file, op->acc_mode); |
aa4caadb MS |
2702 | if (error) |
2703 | goto exit_fput; | |
2704 | ||
2705 | if (will_truncate) { | |
2675a4eb | 2706 | error = handle_truncate(file); |
aa4caadb MS |
2707 | if (error) |
2708 | goto exit_fput; | |
0f9d1a10 | 2709 | } |
ca344a89 AV |
2710 | out: |
2711 | if (want_write) | |
0f9d1a10 | 2712 | mnt_drop_write(nd->path.mnt); |
16b1c1cd | 2713 | path_put(&save_parent); |
e276ae67 | 2714 | terminate_walk(nd); |
2675a4eb | 2715 | return error; |
fb1cc555 | 2716 | |
fb1cc555 AV |
2717 | exit_dput: |
2718 | path_put_conditional(path, nd); | |
ca344a89 | 2719 | goto out; |
015c3bbc | 2720 | exit_fput: |
2675a4eb AV |
2721 | fput(file); |
2722 | goto out; | |
015c3bbc | 2723 | |
f60dc3db MS |
2724 | stale_open: |
2725 | /* If no saved parent or already retried then can't retry */ | |
2726 | if (!save_parent.dentry || retried) | |
2727 | goto out; | |
2728 | ||
2729 | BUG_ON(save_parent.dentry != dir); | |
2730 | path_put(&nd->path); | |
2731 | nd->path = save_parent; | |
2732 | nd->inode = dir->d_inode; | |
2733 | save_parent.mnt = NULL; | |
2734 | save_parent.dentry = NULL; | |
2735 | if (want_write) { | |
2736 | mnt_drop_write(nd->path.mnt); | |
2737 | want_write = false; | |
2738 | } | |
2739 | retried = true; | |
2740 | goto retry_lookup; | |
fb1cc555 AV |
2741 | } |
2742 | ||
13aab428 | 2743 | static struct file *path_openat(int dfd, const char *pathname, |
73d049a4 | 2744 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 2745 | { |
fe2d35ff | 2746 | struct file *base = NULL; |
30d90494 | 2747 | struct file *file; |
9850c056 | 2748 | struct path path; |
47237687 | 2749 | int opened = 0; |
13aab428 | 2750 | int error; |
31e6b01f | 2751 | |
30d90494 AV |
2752 | file = get_empty_filp(); |
2753 | if (!file) | |
31e6b01f NP |
2754 | return ERR_PTR(-ENFILE); |
2755 | ||
30d90494 | 2756 | file->f_flags = op->open_flag; |
31e6b01f | 2757 | |
73d049a4 | 2758 | error = path_init(dfd, pathname, flags | LOOKUP_PARENT, nd, &base); |
31e6b01f | 2759 | if (unlikely(error)) |
2675a4eb | 2760 | goto out; |
31e6b01f | 2761 | |
fe2d35ff | 2762 | current->total_link_count = 0; |
73d049a4 | 2763 | error = link_path_walk(pathname, nd); |
31e6b01f | 2764 | if (unlikely(error)) |
2675a4eb | 2765 | goto out; |
1da177e4 | 2766 | |
2675a4eb AV |
2767 | error = do_last(nd, &path, file, op, &opened, pathname); |
2768 | while (unlikely(error > 0)) { /* trailing symlink */ | |
7b9337aa | 2769 | struct path link = path; |
def4af30 | 2770 | void *cookie; |
574197e0 | 2771 | if (!(nd->flags & LOOKUP_FOLLOW)) { |
73d049a4 AV |
2772 | path_put_conditional(&path, nd); |
2773 | path_put(&nd->path); | |
2675a4eb | 2774 | error = -ELOOP; |
40b39136 AV |
2775 | break; |
2776 | } | |
73d049a4 AV |
2777 | nd->flags |= LOOKUP_PARENT; |
2778 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
574197e0 | 2779 | error = follow_link(&link, nd, &cookie); |
c3e380b0 | 2780 | if (unlikely(error)) |
2675a4eb AV |
2781 | break; |
2782 | error = do_last(nd, &path, file, op, &opened, pathname); | |
574197e0 | 2783 | put_link(nd, &link, cookie); |
806b681c | 2784 | } |
10fa8e62 | 2785 | out: |
73d049a4 AV |
2786 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) |
2787 | path_put(&nd->root); | |
fe2d35ff AV |
2788 | if (base) |
2789 | fput(base); | |
2675a4eb AV |
2790 | if (!(opened & FILE_OPENED)) { |
2791 | BUG_ON(!error); | |
30d90494 | 2792 | put_filp(file); |
16b1c1cd | 2793 | } |
2675a4eb AV |
2794 | if (unlikely(error)) { |
2795 | if (error == -EOPENSTALE) { | |
2796 | if (flags & LOOKUP_RCU) | |
2797 | error = -ECHILD; | |
2798 | else | |
2799 | error = -ESTALE; | |
2800 | } | |
2801 | file = ERR_PTR(error); | |
2802 | } | |
2803 | return file; | |
1da177e4 LT |
2804 | } |
2805 | ||
13aab428 AV |
2806 | struct file *do_filp_open(int dfd, const char *pathname, |
2807 | const struct open_flags *op, int flags) | |
2808 | { | |
73d049a4 | 2809 | struct nameidata nd; |
13aab428 AV |
2810 | struct file *filp; |
2811 | ||
73d049a4 | 2812 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 2813 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 2814 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 2815 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 2816 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
2817 | return filp; |
2818 | } | |
2819 | ||
73d049a4 AV |
2820 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
2821 | const char *name, const struct open_flags *op, int flags) | |
2822 | { | |
2823 | struct nameidata nd; | |
2824 | struct file *file; | |
2825 | ||
2826 | nd.root.mnt = mnt; | |
2827 | nd.root.dentry = dentry; | |
2828 | ||
2829 | flags |= LOOKUP_ROOT; | |
2830 | ||
bcda7652 | 2831 | if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
2832 | return ERR_PTR(-ELOOP); |
2833 | ||
2834 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU); | |
2835 | if (unlikely(file == ERR_PTR(-ECHILD))) | |
2836 | file = path_openat(-1, name, &nd, op, flags); | |
2837 | if (unlikely(file == ERR_PTR(-ESTALE))) | |
2838 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL); | |
2839 | return file; | |
2840 | } | |
2841 | ||
ed75e95d | 2842 | struct dentry *kern_path_create(int dfd, const char *pathname, struct path *path, int is_dir) |
1da177e4 | 2843 | { |
c663e5d8 | 2844 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d AV |
2845 | struct nameidata nd; |
2846 | int error = do_path_lookup(dfd, pathname, LOOKUP_PARENT, &nd); | |
2847 | if (error) | |
2848 | return ERR_PTR(error); | |
1da177e4 | 2849 | |
c663e5d8 CH |
2850 | /* |
2851 | * Yucky last component or no last component at all? | |
2852 | * (foo/., foo/.., /////) | |
2853 | */ | |
ed75e95d AV |
2854 | if (nd.last_type != LAST_NORM) |
2855 | goto out; | |
2856 | nd.flags &= ~LOOKUP_PARENT; | |
2857 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
c663e5d8 CH |
2858 | |
2859 | /* | |
2860 | * Do the final lookup. | |
2861 | */ | |
ed75e95d AV |
2862 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
2863 | dentry = lookup_hash(&nd); | |
1da177e4 LT |
2864 | if (IS_ERR(dentry)) |
2865 | goto fail; | |
c663e5d8 | 2866 | |
e9baf6e5 AV |
2867 | if (dentry->d_inode) |
2868 | goto eexist; | |
c663e5d8 CH |
2869 | /* |
2870 | * Special case - lookup gave negative, but... we had foo/bar/ | |
2871 | * From the vfs_mknod() POV we just have a negative dentry - | |
2872 | * all is fine. Let's be bastards - you had / on the end, you've | |
2873 | * been asking for (non-existent) directory. -ENOENT for you. | |
2874 | */ | |
ed75e95d | 2875 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
e9baf6e5 AV |
2876 | dput(dentry); |
2877 | dentry = ERR_PTR(-ENOENT); | |
ed75e95d | 2878 | goto fail; |
e9baf6e5 | 2879 | } |
ed75e95d | 2880 | *path = nd.path; |
1da177e4 | 2881 | return dentry; |
e9baf6e5 | 2882 | eexist: |
1da177e4 | 2883 | dput(dentry); |
e9baf6e5 | 2884 | dentry = ERR_PTR(-EEXIST); |
1da177e4 | 2885 | fail: |
ed75e95d AV |
2886 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
2887 | out: | |
2888 | path_put(&nd.path); | |
1da177e4 LT |
2889 | return dentry; |
2890 | } | |
dae6ad8f AV |
2891 | EXPORT_SYMBOL(kern_path_create); |
2892 | ||
2893 | struct dentry *user_path_create(int dfd, const char __user *pathname, struct path *path, int is_dir) | |
2894 | { | |
2895 | char *tmp = getname(pathname); | |
2896 | struct dentry *res; | |
2897 | if (IS_ERR(tmp)) | |
2898 | return ERR_CAST(tmp); | |
2899 | res = kern_path_create(dfd, tmp, path, is_dir); | |
2900 | putname(tmp); | |
2901 | return res; | |
2902 | } | |
2903 | EXPORT_SYMBOL(user_path_create); | |
2904 | ||
1a67aafb | 2905 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 2906 | { |
a95164d9 | 2907 | int error = may_create(dir, dentry); |
1da177e4 LT |
2908 | |
2909 | if (error) | |
2910 | return error; | |
2911 | ||
975d6b39 | 2912 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
2913 | return -EPERM; |
2914 | ||
acfa4380 | 2915 | if (!dir->i_op->mknod) |
1da177e4 LT |
2916 | return -EPERM; |
2917 | ||
08ce5f16 SH |
2918 | error = devcgroup_inode_mknod(mode, dev); |
2919 | if (error) | |
2920 | return error; | |
2921 | ||
1da177e4 LT |
2922 | error = security_inode_mknod(dir, dentry, mode, dev); |
2923 | if (error) | |
2924 | return error; | |
2925 | ||
1da177e4 | 2926 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 2927 | if (!error) |
f38aa942 | 2928 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2929 | return error; |
2930 | } | |
2931 | ||
f69aac00 | 2932 | static int may_mknod(umode_t mode) |
463c3197 DH |
2933 | { |
2934 | switch (mode & S_IFMT) { | |
2935 | case S_IFREG: | |
2936 | case S_IFCHR: | |
2937 | case S_IFBLK: | |
2938 | case S_IFIFO: | |
2939 | case S_IFSOCK: | |
2940 | case 0: /* zero mode translates to S_IFREG */ | |
2941 | return 0; | |
2942 | case S_IFDIR: | |
2943 | return -EPERM; | |
2944 | default: | |
2945 | return -EINVAL; | |
2946 | } | |
2947 | } | |
2948 | ||
8208a22b | 2949 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 2950 | unsigned, dev) |
1da177e4 | 2951 | { |
2ad94ae6 | 2952 | struct dentry *dentry; |
dae6ad8f AV |
2953 | struct path path; |
2954 | int error; | |
1da177e4 LT |
2955 | |
2956 | if (S_ISDIR(mode)) | |
2957 | return -EPERM; | |
1da177e4 | 2958 | |
dae6ad8f AV |
2959 | dentry = user_path_create(dfd, filename, &path, 0); |
2960 | if (IS_ERR(dentry)) | |
2961 | return PTR_ERR(dentry); | |
2ad94ae6 | 2962 | |
dae6ad8f | 2963 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 2964 | mode &= ~current_umask(); |
463c3197 DH |
2965 | error = may_mknod(mode); |
2966 | if (error) | |
2967 | goto out_dput; | |
dae6ad8f | 2968 | error = mnt_want_write(path.mnt); |
463c3197 DH |
2969 | if (error) |
2970 | goto out_dput; | |
dae6ad8f | 2971 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 KT |
2972 | if (error) |
2973 | goto out_drop_write; | |
463c3197 | 2974 | switch (mode & S_IFMT) { |
1da177e4 | 2975 | case 0: case S_IFREG: |
312b63fb | 2976 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
1da177e4 LT |
2977 | break; |
2978 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 2979 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
2980 | new_decode_dev(dev)); |
2981 | break; | |
2982 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 2983 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 2984 | break; |
1da177e4 | 2985 | } |
be6d3e56 | 2986 | out_drop_write: |
dae6ad8f | 2987 | mnt_drop_write(path.mnt); |
463c3197 DH |
2988 | out_dput: |
2989 | dput(dentry); | |
dae6ad8f AV |
2990 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
2991 | path_put(&path); | |
1da177e4 LT |
2992 | |
2993 | return error; | |
2994 | } | |
2995 | ||
8208a22b | 2996 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
2997 | { |
2998 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
2999 | } | |
3000 | ||
18bb1db3 | 3001 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3002 | { |
a95164d9 | 3003 | int error = may_create(dir, dentry); |
8de52778 | 3004 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3005 | |
3006 | if (error) | |
3007 | return error; | |
3008 | ||
acfa4380 | 3009 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3010 | return -EPERM; |
3011 | ||
3012 | mode &= (S_IRWXUGO|S_ISVTX); | |
3013 | error = security_inode_mkdir(dir, dentry, mode); | |
3014 | if (error) | |
3015 | return error; | |
3016 | ||
8de52778 AV |
3017 | if (max_links && dir->i_nlink >= max_links) |
3018 | return -EMLINK; | |
3019 | ||
1da177e4 | 3020 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3021 | if (!error) |
f38aa942 | 3022 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3023 | return error; |
3024 | } | |
3025 | ||
a218d0fd | 3026 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 3027 | { |
6902d925 | 3028 | struct dentry *dentry; |
dae6ad8f AV |
3029 | struct path path; |
3030 | int error; | |
1da177e4 | 3031 | |
dae6ad8f | 3032 | dentry = user_path_create(dfd, pathname, &path, 1); |
6902d925 | 3033 | if (IS_ERR(dentry)) |
dae6ad8f | 3034 | return PTR_ERR(dentry); |
1da177e4 | 3035 | |
dae6ad8f | 3036 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3037 | mode &= ~current_umask(); |
dae6ad8f | 3038 | error = mnt_want_write(path.mnt); |
463c3197 DH |
3039 | if (error) |
3040 | goto out_dput; | |
dae6ad8f | 3041 | error = security_path_mkdir(&path, dentry, mode); |
be6d3e56 KT |
3042 | if (error) |
3043 | goto out_drop_write; | |
dae6ad8f | 3044 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); |
be6d3e56 | 3045 | out_drop_write: |
dae6ad8f | 3046 | mnt_drop_write(path.mnt); |
463c3197 | 3047 | out_dput: |
6902d925 | 3048 | dput(dentry); |
dae6ad8f AV |
3049 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
3050 | path_put(&path); | |
1da177e4 LT |
3051 | return error; |
3052 | } | |
3053 | ||
a218d0fd | 3054 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
3055 | { |
3056 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
3057 | } | |
3058 | ||
1da177e4 | 3059 | /* |
a71905f0 | 3060 | * The dentry_unhash() helper will try to drop the dentry early: we |
c0d02594 | 3061 | * should have a usage count of 1 if we're the only user of this |
a71905f0 SW |
3062 | * dentry, and if that is true (possibly after pruning the dcache), |
3063 | * then we drop the dentry now. | |
1da177e4 LT |
3064 | * |
3065 | * A low-level filesystem can, if it choses, legally | |
3066 | * do a | |
3067 | * | |
3068 | * if (!d_unhashed(dentry)) | |
3069 | * return -EBUSY; | |
3070 | * | |
3071 | * if it cannot handle the case of removing a directory | |
3072 | * that is still in use by something else.. | |
3073 | */ | |
3074 | void dentry_unhash(struct dentry *dentry) | |
3075 | { | |
dc168427 | 3076 | shrink_dcache_parent(dentry); |
1da177e4 | 3077 | spin_lock(&dentry->d_lock); |
64252c75 | 3078 | if (dentry->d_count == 1) |
1da177e4 LT |
3079 | __d_drop(dentry); |
3080 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
3081 | } |
3082 | ||
3083 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
3084 | { | |
3085 | int error = may_delete(dir, dentry, 1); | |
3086 | ||
3087 | if (error) | |
3088 | return error; | |
3089 | ||
acfa4380 | 3090 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3091 | return -EPERM; |
3092 | ||
1d2ef590 | 3093 | dget(dentry); |
1b1dcc1b | 3094 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
3095 | |
3096 | error = -EBUSY; | |
1da177e4 | 3097 | if (d_mountpoint(dentry)) |
912dbc15 SW |
3098 | goto out; |
3099 | ||
3100 | error = security_inode_rmdir(dir, dentry); | |
3101 | if (error) | |
3102 | goto out; | |
3103 | ||
3cebde24 | 3104 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3105 | error = dir->i_op->rmdir(dir, dentry); |
3106 | if (error) | |
3107 | goto out; | |
3108 | ||
3109 | dentry->d_inode->i_flags |= S_DEAD; | |
3110 | dont_mount(dentry); | |
3111 | ||
3112 | out: | |
1b1dcc1b | 3113 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 3114 | dput(dentry); |
912dbc15 | 3115 | if (!error) |
1da177e4 | 3116 | d_delete(dentry); |
1da177e4 LT |
3117 | return error; |
3118 | } | |
3119 | ||
5590ff0d | 3120 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3121 | { |
3122 | int error = 0; | |
3123 | char * name; | |
3124 | struct dentry *dentry; | |
3125 | struct nameidata nd; | |
3126 | ||
2ad94ae6 | 3127 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 3128 | if (error) |
2ad94ae6 | 3129 | return error; |
1da177e4 LT |
3130 | |
3131 | switch(nd.last_type) { | |
0612d9fb OH |
3132 | case LAST_DOTDOT: |
3133 | error = -ENOTEMPTY; | |
3134 | goto exit1; | |
3135 | case LAST_DOT: | |
3136 | error = -EINVAL; | |
3137 | goto exit1; | |
3138 | case LAST_ROOT: | |
3139 | error = -EBUSY; | |
3140 | goto exit1; | |
1da177e4 | 3141 | } |
0612d9fb OH |
3142 | |
3143 | nd.flags &= ~LOOKUP_PARENT; | |
3144 | ||
4ac91378 | 3145 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 3146 | dentry = lookup_hash(&nd); |
1da177e4 | 3147 | error = PTR_ERR(dentry); |
6902d925 DH |
3148 | if (IS_ERR(dentry)) |
3149 | goto exit2; | |
e6bc45d6 TT |
3150 | if (!dentry->d_inode) { |
3151 | error = -ENOENT; | |
3152 | goto exit3; | |
3153 | } | |
0622753b DH |
3154 | error = mnt_want_write(nd.path.mnt); |
3155 | if (error) | |
3156 | goto exit3; | |
be6d3e56 KT |
3157 | error = security_path_rmdir(&nd.path, dentry); |
3158 | if (error) | |
3159 | goto exit4; | |
4ac91378 | 3160 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 3161 | exit4: |
0622753b DH |
3162 | mnt_drop_write(nd.path.mnt); |
3163 | exit3: | |
6902d925 DH |
3164 | dput(dentry); |
3165 | exit2: | |
4ac91378 | 3166 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 | 3167 | exit1: |
1d957f9b | 3168 | path_put(&nd.path); |
1da177e4 LT |
3169 | putname(name); |
3170 | return error; | |
3171 | } | |
3172 | ||
3cdad428 | 3173 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3174 | { |
3175 | return do_rmdir(AT_FDCWD, pathname); | |
3176 | } | |
3177 | ||
1da177e4 LT |
3178 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
3179 | { | |
3180 | int error = may_delete(dir, dentry, 0); | |
3181 | ||
3182 | if (error) | |
3183 | return error; | |
3184 | ||
acfa4380 | 3185 | if (!dir->i_op->unlink) |
1da177e4 LT |
3186 | return -EPERM; |
3187 | ||
1b1dcc1b | 3188 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
3189 | if (d_mountpoint(dentry)) |
3190 | error = -EBUSY; | |
3191 | else { | |
3192 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3193 | if (!error) { |
1da177e4 | 3194 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 3195 | if (!error) |
d83c49f3 | 3196 | dont_mount(dentry); |
bec1052e | 3197 | } |
1da177e4 | 3198 | } |
1b1dcc1b | 3199 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
3200 | |
3201 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3202 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 3203 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 3204 | d_delete(dentry); |
1da177e4 | 3205 | } |
0eeca283 | 3206 | |
1da177e4 LT |
3207 | return error; |
3208 | } | |
3209 | ||
3210 | /* | |
3211 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3212 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3213 | * writeout happening, and we don't want to prevent access to the directory |
3214 | * while waiting on the I/O. | |
3215 | */ | |
5590ff0d | 3216 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 3217 | { |
2ad94ae6 AV |
3218 | int error; |
3219 | char *name; | |
1da177e4 LT |
3220 | struct dentry *dentry; |
3221 | struct nameidata nd; | |
3222 | struct inode *inode = NULL; | |
3223 | ||
2ad94ae6 | 3224 | error = user_path_parent(dfd, pathname, &nd, &name); |
1da177e4 | 3225 | if (error) |
2ad94ae6 AV |
3226 | return error; |
3227 | ||
1da177e4 LT |
3228 | error = -EISDIR; |
3229 | if (nd.last_type != LAST_NORM) | |
3230 | goto exit1; | |
0612d9fb OH |
3231 | |
3232 | nd.flags &= ~LOOKUP_PARENT; | |
3233 | ||
4ac91378 | 3234 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 3235 | dentry = lookup_hash(&nd); |
1da177e4 LT |
3236 | error = PTR_ERR(dentry); |
3237 | if (!IS_ERR(dentry)) { | |
3238 | /* Why not before? Because we want correct error value */ | |
50338b88 TE |
3239 | if (nd.last.name[nd.last.len]) |
3240 | goto slashes; | |
1da177e4 | 3241 | inode = dentry->d_inode; |
50338b88 | 3242 | if (!inode) |
e6bc45d6 TT |
3243 | goto slashes; |
3244 | ihold(inode); | |
0622753b DH |
3245 | error = mnt_want_write(nd.path.mnt); |
3246 | if (error) | |
3247 | goto exit2; | |
be6d3e56 KT |
3248 | error = security_path_unlink(&nd.path, dentry); |
3249 | if (error) | |
3250 | goto exit3; | |
4ac91378 | 3251 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
be6d3e56 | 3252 | exit3: |
0622753b | 3253 | mnt_drop_write(nd.path.mnt); |
1da177e4 LT |
3254 | exit2: |
3255 | dput(dentry); | |
3256 | } | |
4ac91378 | 3257 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
3258 | if (inode) |
3259 | iput(inode); /* truncate the inode here */ | |
3260 | exit1: | |
1d957f9b | 3261 | path_put(&nd.path); |
1da177e4 LT |
3262 | putname(name); |
3263 | return error; | |
3264 | ||
3265 | slashes: | |
3266 | error = !dentry->d_inode ? -ENOENT : | |
3267 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
3268 | goto exit2; | |
3269 | } | |
3270 | ||
2e4d0924 | 3271 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3272 | { |
3273 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3274 | return -EINVAL; | |
3275 | ||
3276 | if (flag & AT_REMOVEDIR) | |
3277 | return do_rmdir(dfd, pathname); | |
3278 | ||
3279 | return do_unlinkat(dfd, pathname); | |
3280 | } | |
3281 | ||
3480b257 | 3282 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
3283 | { |
3284 | return do_unlinkat(AT_FDCWD, pathname); | |
3285 | } | |
3286 | ||
db2e747b | 3287 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3288 | { |
a95164d9 | 3289 | int error = may_create(dir, dentry); |
1da177e4 LT |
3290 | |
3291 | if (error) | |
3292 | return error; | |
3293 | ||
acfa4380 | 3294 | if (!dir->i_op->symlink) |
1da177e4 LT |
3295 | return -EPERM; |
3296 | ||
3297 | error = security_inode_symlink(dir, dentry, oldname); | |
3298 | if (error) | |
3299 | return error; | |
3300 | ||
1da177e4 | 3301 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3302 | if (!error) |
f38aa942 | 3303 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3304 | return error; |
3305 | } | |
3306 | ||
2e4d0924 HC |
3307 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
3308 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3309 | { |
2ad94ae6 AV |
3310 | int error; |
3311 | char *from; | |
6902d925 | 3312 | struct dentry *dentry; |
dae6ad8f | 3313 | struct path path; |
1da177e4 LT |
3314 | |
3315 | from = getname(oldname); | |
2ad94ae6 | 3316 | if (IS_ERR(from)) |
1da177e4 | 3317 | return PTR_ERR(from); |
1da177e4 | 3318 | |
dae6ad8f | 3319 | dentry = user_path_create(newdfd, newname, &path, 0); |
6902d925 DH |
3320 | error = PTR_ERR(dentry); |
3321 | if (IS_ERR(dentry)) | |
dae6ad8f | 3322 | goto out_putname; |
6902d925 | 3323 | |
dae6ad8f | 3324 | error = mnt_want_write(path.mnt); |
75c3f29d DH |
3325 | if (error) |
3326 | goto out_dput; | |
dae6ad8f | 3327 | error = security_path_symlink(&path, dentry, from); |
be6d3e56 KT |
3328 | if (error) |
3329 | goto out_drop_write; | |
dae6ad8f | 3330 | error = vfs_symlink(path.dentry->d_inode, dentry, from); |
be6d3e56 | 3331 | out_drop_write: |
dae6ad8f | 3332 | mnt_drop_write(path.mnt); |
75c3f29d | 3333 | out_dput: |
6902d925 | 3334 | dput(dentry); |
dae6ad8f AV |
3335 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
3336 | path_put(&path); | |
6902d925 | 3337 | out_putname: |
1da177e4 LT |
3338 | putname(from); |
3339 | return error; | |
3340 | } | |
3341 | ||
3480b257 | 3342 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3343 | { |
3344 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
3345 | } | |
3346 | ||
1da177e4 LT |
3347 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
3348 | { | |
3349 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 3350 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3351 | int error; |
3352 | ||
3353 | if (!inode) | |
3354 | return -ENOENT; | |
3355 | ||
a95164d9 | 3356 | error = may_create(dir, new_dentry); |
1da177e4 LT |
3357 | if (error) |
3358 | return error; | |
3359 | ||
3360 | if (dir->i_sb != inode->i_sb) | |
3361 | return -EXDEV; | |
3362 | ||
3363 | /* | |
3364 | * A link to an append-only or immutable file cannot be created. | |
3365 | */ | |
3366 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
3367 | return -EPERM; | |
acfa4380 | 3368 | if (!dir->i_op->link) |
1da177e4 | 3369 | return -EPERM; |
7e79eedb | 3370 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
3371 | return -EPERM; |
3372 | ||
3373 | error = security_inode_link(old_dentry, dir, new_dentry); | |
3374 | if (error) | |
3375 | return error; | |
3376 | ||
7e79eedb | 3377 | mutex_lock(&inode->i_mutex); |
aae8a97d AK |
3378 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
3379 | if (inode->i_nlink == 0) | |
3380 | error = -ENOENT; | |
8de52778 AV |
3381 | else if (max_links && inode->i_nlink >= max_links) |
3382 | error = -EMLINK; | |
aae8a97d AK |
3383 | else |
3384 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 3385 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 3386 | if (!error) |
7e79eedb | 3387 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
3388 | return error; |
3389 | } | |
3390 | ||
3391 | /* | |
3392 | * Hardlinks are often used in delicate situations. We avoid | |
3393 | * security-related surprises by not following symlinks on the | |
3394 | * newname. --KAB | |
3395 | * | |
3396 | * We don't follow them on the oldname either to be compatible | |
3397 | * with linux 2.0, and to avoid hard-linking to directories | |
3398 | * and other special files. --ADM | |
3399 | */ | |
2e4d0924 HC |
3400 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
3401 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
3402 | { |
3403 | struct dentry *new_dentry; | |
dae6ad8f | 3404 | struct path old_path, new_path; |
11a7b371 | 3405 | int how = 0; |
1da177e4 | 3406 | int error; |
1da177e4 | 3407 | |
11a7b371 | 3408 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 3409 | return -EINVAL; |
11a7b371 AK |
3410 | /* |
3411 | * To use null names we require CAP_DAC_READ_SEARCH | |
3412 | * This ensures that not everyone will be able to create | |
3413 | * handlink using the passed filedescriptor. | |
3414 | */ | |
3415 | if (flags & AT_EMPTY_PATH) { | |
3416 | if (!capable(CAP_DAC_READ_SEARCH)) | |
3417 | return -ENOENT; | |
3418 | how = LOOKUP_EMPTY; | |
3419 | } | |
3420 | ||
3421 | if (flags & AT_SYMLINK_FOLLOW) | |
3422 | how |= LOOKUP_FOLLOW; | |
c04030e1 | 3423 | |
11a7b371 | 3424 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 3425 | if (error) |
2ad94ae6 AV |
3426 | return error; |
3427 | ||
dae6ad8f | 3428 | new_dentry = user_path_create(newdfd, newname, &new_path, 0); |
1da177e4 | 3429 | error = PTR_ERR(new_dentry); |
6902d925 | 3430 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
3431 | goto out; |
3432 | ||
3433 | error = -EXDEV; | |
3434 | if (old_path.mnt != new_path.mnt) | |
3435 | goto out_dput; | |
3436 | error = mnt_want_write(new_path.mnt); | |
75c3f29d DH |
3437 | if (error) |
3438 | goto out_dput; | |
dae6ad8f | 3439 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 KT |
3440 | if (error) |
3441 | goto out_drop_write; | |
dae6ad8f | 3442 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry); |
be6d3e56 | 3443 | out_drop_write: |
dae6ad8f | 3444 | mnt_drop_write(new_path.mnt); |
75c3f29d | 3445 | out_dput: |
6902d925 | 3446 | dput(new_dentry); |
dae6ad8f AV |
3447 | mutex_unlock(&new_path.dentry->d_inode->i_mutex); |
3448 | path_put(&new_path); | |
1da177e4 | 3449 | out: |
2d8f3038 | 3450 | path_put(&old_path); |
1da177e4 LT |
3451 | |
3452 | return error; | |
3453 | } | |
3454 | ||
3480b257 | 3455 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 3456 | { |
c04030e1 | 3457 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
3458 | } |
3459 | ||
1da177e4 LT |
3460 | /* |
3461 | * The worst of all namespace operations - renaming directory. "Perverted" | |
3462 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
3463 | * Problems: | |
3464 | * a) we can get into loop creation. Check is done in is_subdir(). | |
3465 | * b) race potential - two innocent renames can create a loop together. | |
3466 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 3467 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
3468 | * story. |
3469 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 3470 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
3471 | * whether the target exists). Solution: try to be smart with locking |
3472 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 3473 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
3474 | * move will be locked. Thus we can rank directories by the tree |
3475 | * (ancestors first) and rank all non-directories after them. | |
3476 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 3477 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
3478 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
3479 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
3480 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 3481 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 3482 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 3483 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 3484 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
3485 | * locking]. |
3486 | */ | |
75c96f85 AB |
3487 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
3488 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3489 | { |
3490 | int error = 0; | |
9055cba7 | 3491 | struct inode *target = new_dentry->d_inode; |
8de52778 | 3492 | unsigned max_links = new_dir->i_sb->s_max_links; |
1da177e4 LT |
3493 | |
3494 | /* | |
3495 | * If we are going to change the parent - check write permissions, | |
3496 | * we'll need to flip '..'. | |
3497 | */ | |
3498 | if (new_dir != old_dir) { | |
f419a2e3 | 3499 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3500 | if (error) |
3501 | return error; | |
3502 | } | |
3503 | ||
3504 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3505 | if (error) | |
3506 | return error; | |
3507 | ||
1d2ef590 | 3508 | dget(new_dentry); |
d83c49f3 | 3509 | if (target) |
1b1dcc1b | 3510 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
3511 | |
3512 | error = -EBUSY; | |
3513 | if (d_mountpoint(old_dentry) || d_mountpoint(new_dentry)) | |
3514 | goto out; | |
3515 | ||
8de52778 AV |
3516 | error = -EMLINK; |
3517 | if (max_links && !target && new_dir != old_dir && | |
3518 | new_dir->i_nlink >= max_links) | |
3519 | goto out; | |
3520 | ||
3cebde24 SW |
3521 | if (target) |
3522 | shrink_dcache_parent(new_dentry); | |
9055cba7 SW |
3523 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
3524 | if (error) | |
3525 | goto out; | |
3526 | ||
1da177e4 | 3527 | if (target) { |
9055cba7 SW |
3528 | target->i_flags |= S_DEAD; |
3529 | dont_mount(new_dentry); | |
1da177e4 | 3530 | } |
9055cba7 SW |
3531 | out: |
3532 | if (target) | |
3533 | mutex_unlock(&target->i_mutex); | |
1d2ef590 | 3534 | dput(new_dentry); |
e31e14ec | 3535 | if (!error) |
349457cc MF |
3536 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3537 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3538 | return error; |
3539 | } | |
3540 | ||
75c96f85 AB |
3541 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3542 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 | 3543 | { |
51892bbb | 3544 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3545 | int error; |
3546 | ||
3547 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3548 | if (error) | |
3549 | return error; | |
3550 | ||
3551 | dget(new_dentry); | |
1da177e4 | 3552 | if (target) |
1b1dcc1b | 3553 | mutex_lock(&target->i_mutex); |
51892bbb SW |
3554 | |
3555 | error = -EBUSY; | |
1da177e4 | 3556 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
51892bbb SW |
3557 | goto out; |
3558 | ||
3559 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3560 | if (error) | |
3561 | goto out; | |
3562 | ||
3563 | if (target) | |
3564 | dont_mount(new_dentry); | |
3565 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) | |
3566 | d_move(old_dentry, new_dentry); | |
3567 | out: | |
1da177e4 | 3568 | if (target) |
1b1dcc1b | 3569 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3570 | dput(new_dentry); |
3571 | return error; | |
3572 | } | |
3573 | ||
3574 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3575 | struct inode *new_dir, struct dentry *new_dentry) | |
3576 | { | |
3577 | int error; | |
3578 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3579 | const unsigned char *old_name; |
1da177e4 LT |
3580 | |
3581 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3582 | return 0; | |
3583 | ||
3584 | error = may_delete(old_dir, old_dentry, is_dir); | |
3585 | if (error) | |
3586 | return error; | |
3587 | ||
3588 | if (!new_dentry->d_inode) | |
a95164d9 | 3589 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3590 | else |
3591 | error = may_delete(new_dir, new_dentry, is_dir); | |
3592 | if (error) | |
3593 | return error; | |
3594 | ||
acfa4380 | 3595 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3596 | return -EPERM; |
3597 | ||
0eeca283 RL |
3598 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3599 | ||
1da177e4 LT |
3600 | if (is_dir) |
3601 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3602 | else | |
3603 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3604 | if (!error) |
3605 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3606 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3607 | fsnotify_oldname_free(old_name); |
3608 | ||
1da177e4 LT |
3609 | return error; |
3610 | } | |
3611 | ||
2e4d0924 HC |
3612 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3613 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3614 | { |
2ad94ae6 AV |
3615 | struct dentry *old_dir, *new_dir; |
3616 | struct dentry *old_dentry, *new_dentry; | |
3617 | struct dentry *trap; | |
1da177e4 | 3618 | struct nameidata oldnd, newnd; |
2ad94ae6 AV |
3619 | char *from; |
3620 | char *to; | |
3621 | int error; | |
1da177e4 | 3622 | |
2ad94ae6 | 3623 | error = user_path_parent(olddfd, oldname, &oldnd, &from); |
1da177e4 LT |
3624 | if (error) |
3625 | goto exit; | |
3626 | ||
2ad94ae6 | 3627 | error = user_path_parent(newdfd, newname, &newnd, &to); |
1da177e4 LT |
3628 | if (error) |
3629 | goto exit1; | |
3630 | ||
3631 | error = -EXDEV; | |
4ac91378 | 3632 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3633 | goto exit2; |
3634 | ||
4ac91378 | 3635 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3636 | error = -EBUSY; |
3637 | if (oldnd.last_type != LAST_NORM) | |
3638 | goto exit2; | |
3639 | ||
4ac91378 | 3640 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3641 | if (newnd.last_type != LAST_NORM) |
3642 | goto exit2; | |
3643 | ||
0612d9fb OH |
3644 | oldnd.flags &= ~LOOKUP_PARENT; |
3645 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3646 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3647 | |
1da177e4 LT |
3648 | trap = lock_rename(new_dir, old_dir); |
3649 | ||
49705b77 | 3650 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3651 | error = PTR_ERR(old_dentry); |
3652 | if (IS_ERR(old_dentry)) | |
3653 | goto exit3; | |
3654 | /* source must exist */ | |
3655 | error = -ENOENT; | |
3656 | if (!old_dentry->d_inode) | |
3657 | goto exit4; | |
3658 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3659 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3660 | error = -ENOTDIR; | |
3661 | if (oldnd.last.name[oldnd.last.len]) | |
3662 | goto exit4; | |
3663 | if (newnd.last.name[newnd.last.len]) | |
3664 | goto exit4; | |
3665 | } | |
3666 | /* source should not be ancestor of target */ | |
3667 | error = -EINVAL; | |
3668 | if (old_dentry == trap) | |
3669 | goto exit4; | |
49705b77 | 3670 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3671 | error = PTR_ERR(new_dentry); |
3672 | if (IS_ERR(new_dentry)) | |
3673 | goto exit4; | |
3674 | /* target should not be an ancestor of source */ | |
3675 | error = -ENOTEMPTY; | |
3676 | if (new_dentry == trap) | |
3677 | goto exit5; | |
3678 | ||
9079b1eb DH |
3679 | error = mnt_want_write(oldnd.path.mnt); |
3680 | if (error) | |
3681 | goto exit5; | |
be6d3e56 KT |
3682 | error = security_path_rename(&oldnd.path, old_dentry, |
3683 | &newnd.path, new_dentry); | |
3684 | if (error) | |
3685 | goto exit6; | |
1da177e4 LT |
3686 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3687 | new_dir->d_inode, new_dentry); | |
be6d3e56 | 3688 | exit6: |
9079b1eb | 3689 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 LT |
3690 | exit5: |
3691 | dput(new_dentry); | |
3692 | exit4: | |
3693 | dput(old_dentry); | |
3694 | exit3: | |
3695 | unlock_rename(new_dir, old_dir); | |
3696 | exit2: | |
1d957f9b | 3697 | path_put(&newnd.path); |
2ad94ae6 | 3698 | putname(to); |
1da177e4 | 3699 | exit1: |
1d957f9b | 3700 | path_put(&oldnd.path); |
1da177e4 | 3701 | putname(from); |
2ad94ae6 | 3702 | exit: |
1da177e4 LT |
3703 | return error; |
3704 | } | |
3705 | ||
a26eab24 | 3706 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3707 | { |
3708 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3709 | } | |
3710 | ||
1da177e4 LT |
3711 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3712 | { | |
3713 | int len; | |
3714 | ||
3715 | len = PTR_ERR(link); | |
3716 | if (IS_ERR(link)) | |
3717 | goto out; | |
3718 | ||
3719 | len = strlen(link); | |
3720 | if (len > (unsigned) buflen) | |
3721 | len = buflen; | |
3722 | if (copy_to_user(buffer, link, len)) | |
3723 | len = -EFAULT; | |
3724 | out: | |
3725 | return len; | |
3726 | } | |
3727 | ||
3728 | /* | |
3729 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3730 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3731 | * using) it for any given inode is up to filesystem. | |
3732 | */ | |
3733 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3734 | { | |
3735 | struct nameidata nd; | |
cc314eef | 3736 | void *cookie; |
694a1764 | 3737 | int res; |
cc314eef | 3738 | |
1da177e4 | 3739 | nd.depth = 0; |
cc314eef | 3740 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3741 | if (IS_ERR(cookie)) |
3742 | return PTR_ERR(cookie); | |
3743 | ||
3744 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3745 | if (dentry->d_inode->i_op->put_link) | |
3746 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3747 | return res; | |
1da177e4 LT |
3748 | } |
3749 | ||
3750 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3751 | { | |
3752 | return __vfs_follow_link(nd, link); | |
3753 | } | |
3754 | ||
3755 | /* get the link contents into pagecache */ | |
3756 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3757 | { | |
ebd09abb DG |
3758 | char *kaddr; |
3759 | struct page *page; | |
1da177e4 | 3760 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3761 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3762 | if (IS_ERR(page)) |
6fe6900e | 3763 | return (char*)page; |
1da177e4 | 3764 | *ppage = page; |
ebd09abb DG |
3765 | kaddr = kmap(page); |
3766 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3767 | return kaddr; | |
1da177e4 LT |
3768 | } |
3769 | ||
3770 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3771 | { | |
3772 | struct page *page = NULL; | |
3773 | char *s = page_getlink(dentry, &page); | |
3774 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3775 | if (page) { | |
3776 | kunmap(page); | |
3777 | page_cache_release(page); | |
3778 | } | |
3779 | return res; | |
3780 | } | |
3781 | ||
cc314eef | 3782 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3783 | { |
cc314eef | 3784 | struct page *page = NULL; |
1da177e4 | 3785 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3786 | return page; |
1da177e4 LT |
3787 | } |
3788 | ||
cc314eef | 3789 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3790 | { |
cc314eef LT |
3791 | struct page *page = cookie; |
3792 | ||
3793 | if (page) { | |
1da177e4 LT |
3794 | kunmap(page); |
3795 | page_cache_release(page); | |
1da177e4 LT |
3796 | } |
3797 | } | |
3798 | ||
54566b2c NP |
3799 | /* |
3800 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3801 | */ | |
3802 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3803 | { |
3804 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3805 | struct page *page; |
afddba49 | 3806 | void *fsdata; |
beb497ab | 3807 | int err; |
1da177e4 | 3808 | char *kaddr; |
54566b2c NP |
3809 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3810 | if (nofs) | |
3811 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3812 | |
7e53cac4 | 3813 | retry: |
afddba49 | 3814 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3815 | flags, &page, &fsdata); |
1da177e4 | 3816 | if (err) |
afddba49 NP |
3817 | goto fail; |
3818 | ||
e8e3c3d6 | 3819 | kaddr = kmap_atomic(page); |
1da177e4 | 3820 | memcpy(kaddr, symname, len-1); |
e8e3c3d6 | 3821 | kunmap_atomic(kaddr); |
afddba49 NP |
3822 | |
3823 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3824 | page, fsdata); | |
1da177e4 LT |
3825 | if (err < 0) |
3826 | goto fail; | |
afddba49 NP |
3827 | if (err < len-1) |
3828 | goto retry; | |
3829 | ||
1da177e4 LT |
3830 | mark_inode_dirty(inode); |
3831 | return 0; | |
1da177e4 LT |
3832 | fail: |
3833 | return err; | |
3834 | } | |
3835 | ||
0adb25d2 KK |
3836 | int page_symlink(struct inode *inode, const char *symname, int len) |
3837 | { | |
3838 | return __page_symlink(inode, symname, len, | |
54566b2c | 3839 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3840 | } |
3841 | ||
92e1d5be | 3842 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3843 | .readlink = generic_readlink, |
3844 | .follow_link = page_follow_link_light, | |
3845 | .put_link = page_put_link, | |
3846 | }; | |
3847 | ||
2d8f3038 | 3848 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3849 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3850 | EXPORT_SYMBOL(follow_down); |
3851 | EXPORT_SYMBOL(follow_up); | |
3852 | EXPORT_SYMBOL(get_write_access); /* binfmt_aout */ | |
3853 | EXPORT_SYMBOL(getname); | |
3854 | EXPORT_SYMBOL(lock_rename); | |
1da177e4 LT |
3855 | EXPORT_SYMBOL(lookup_one_len); |
3856 | EXPORT_SYMBOL(page_follow_link_light); | |
3857 | EXPORT_SYMBOL(page_put_link); | |
3858 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3859 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3860 | EXPORT_SYMBOL(page_symlink); |
3861 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
d1811465 | 3862 | EXPORT_SYMBOL(kern_path); |
16f18200 | 3863 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 3864 | EXPORT_SYMBOL(inode_permission); |
1da177e4 LT |
3865 | EXPORT_SYMBOL(unlock_rename); |
3866 | EXPORT_SYMBOL(vfs_create); | |
3867 | EXPORT_SYMBOL(vfs_follow_link); | |
3868 | EXPORT_SYMBOL(vfs_link); | |
3869 | EXPORT_SYMBOL(vfs_mkdir); | |
3870 | EXPORT_SYMBOL(vfs_mknod); | |
3871 | EXPORT_SYMBOL(generic_permission); | |
3872 | EXPORT_SYMBOL(vfs_readlink); | |
3873 | EXPORT_SYMBOL(vfs_rename); | |
3874 | EXPORT_SYMBOL(vfs_rmdir); | |
3875 | EXPORT_SYMBOL(vfs_symlink); | |
3876 | EXPORT_SYMBOL(vfs_unlink); | |
3877 | EXPORT_SYMBOL(dentry_unhash); | |
3878 | EXPORT_SYMBOL(generic_readlink); |