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 | */ | |
91a27b2a | 120 | void final_putname(struct filename *name) |
1da177e4 | 121 | { |
91a27b2a JL |
122 | __putname(name->name); |
123 | kfree(name); | |
124 | } | |
125 | ||
126 | static struct filename * | |
127 | getname_flags(const char __user *filename, int flags, int *empty) | |
128 | { | |
129 | struct filename *result, *err; | |
130 | char *kname; | |
3f9f0aa6 | 131 | int len; |
4043cde8 | 132 | |
91a27b2a JL |
133 | /* FIXME: create dedicated slabcache? */ |
134 | result = kzalloc(sizeof(*result), GFP_KERNEL); | |
3f9f0aa6 | 135 | if (unlikely(!result)) |
4043cde8 EP |
136 | return ERR_PTR(-ENOMEM); |
137 | ||
91a27b2a JL |
138 | kname = __getname(); |
139 | if (unlikely(!kname)) { | |
140 | err = ERR_PTR(-ENOMEM); | |
141 | goto error_free_name; | |
142 | } | |
143 | ||
144 | result->name = kname; | |
145 | result->uptr = filename; | |
146 | len = strncpy_from_user(kname, filename, PATH_MAX); | |
147 | if (unlikely(len < 0)) { | |
148 | err = ERR_PTR(len); | |
3f9f0aa6 | 149 | goto error; |
91a27b2a | 150 | } |
3f9f0aa6 LT |
151 | |
152 | /* The empty path is special. */ | |
153 | if (unlikely(!len)) { | |
154 | if (empty) | |
4043cde8 | 155 | *empty = 1; |
3f9f0aa6 LT |
156 | err = ERR_PTR(-ENOENT); |
157 | if (!(flags & LOOKUP_EMPTY)) | |
158 | goto error; | |
1da177e4 | 159 | } |
3f9f0aa6 LT |
160 | |
161 | err = ERR_PTR(-ENAMETOOLONG); | |
162 | if (likely(len < PATH_MAX)) { | |
163 | audit_getname(result); | |
164 | return result; | |
165 | } | |
166 | ||
167 | error: | |
91a27b2a JL |
168 | __putname(kname); |
169 | error_free_name: | |
170 | kfree(result); | |
3f9f0aa6 | 171 | return err; |
1da177e4 LT |
172 | } |
173 | ||
91a27b2a JL |
174 | struct filename * |
175 | getname(const char __user * filename) | |
f52e0c11 | 176 | { |
f7493e5d | 177 | return getname_flags(filename, 0, NULL); |
f52e0c11 | 178 | } |
91a27b2a | 179 | EXPORT_SYMBOL(getname); |
f52e0c11 | 180 | |
1da177e4 | 181 | #ifdef CONFIG_AUDITSYSCALL |
91a27b2a | 182 | void putname(struct filename *name) |
1da177e4 | 183 | { |
5ac3a9c2 | 184 | if (unlikely(!audit_dummy_context())) |
91a27b2a JL |
185 | return audit_putname(name); |
186 | final_putname(name); | |
1da177e4 | 187 | } |
1da177e4 LT |
188 | #endif |
189 | ||
e77819e5 LT |
190 | static int check_acl(struct inode *inode, int mask) |
191 | { | |
84635d68 | 192 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
193 | struct posix_acl *acl; |
194 | ||
e77819e5 | 195 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
196 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
197 | if (!acl) | |
e77819e5 | 198 | return -EAGAIN; |
3567866b AV |
199 | /* no ->get_acl() calls in RCU mode... */ |
200 | if (acl == ACL_NOT_CACHED) | |
201 | return -ECHILD; | |
206b1d09 | 202 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
203 | } |
204 | ||
205 | acl = get_cached_acl(inode, ACL_TYPE_ACCESS); | |
206 | ||
207 | /* | |
4e34e719 CH |
208 | * A filesystem can force a ACL callback by just never filling the |
209 | * ACL cache. But normally you'd fill the cache either at inode | |
210 | * instantiation time, or on the first ->get_acl call. | |
e77819e5 | 211 | * |
4e34e719 CH |
212 | * If the filesystem doesn't have a get_acl() function at all, we'll |
213 | * just create the negative cache entry. | |
e77819e5 LT |
214 | */ |
215 | if (acl == ACL_NOT_CACHED) { | |
4e34e719 CH |
216 | if (inode->i_op->get_acl) { |
217 | acl = inode->i_op->get_acl(inode, ACL_TYPE_ACCESS); | |
218 | if (IS_ERR(acl)) | |
219 | return PTR_ERR(acl); | |
220 | } else { | |
221 | set_cached_acl(inode, ACL_TYPE_ACCESS, NULL); | |
222 | return -EAGAIN; | |
223 | } | |
e77819e5 LT |
224 | } |
225 | ||
226 | if (acl) { | |
227 | int error = posix_acl_permission(inode, acl, mask); | |
228 | posix_acl_release(acl); | |
229 | return error; | |
230 | } | |
84635d68 | 231 | #endif |
e77819e5 LT |
232 | |
233 | return -EAGAIN; | |
234 | } | |
235 | ||
5909ccaa | 236 | /* |
948409c7 | 237 | * This does the basic permission checking |
1da177e4 | 238 | */ |
7e40145e | 239 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 240 | { |
26cf46be | 241 | unsigned int mode = inode->i_mode; |
1da177e4 | 242 | |
8e96e3b7 | 243 | if (likely(uid_eq(current_fsuid(), inode->i_uid))) |
1da177e4 LT |
244 | mode >>= 6; |
245 | else { | |
e77819e5 | 246 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 247 | int error = check_acl(inode, mask); |
b74c79e9 NP |
248 | if (error != -EAGAIN) |
249 | return error; | |
1da177e4 LT |
250 | } |
251 | ||
252 | if (in_group_p(inode->i_gid)) | |
253 | mode >>= 3; | |
254 | } | |
255 | ||
256 | /* | |
257 | * If the DACs are ok we don't need any capability check. | |
258 | */ | |
9c2c7039 | 259 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 260 | return 0; |
5909ccaa LT |
261 | return -EACCES; |
262 | } | |
263 | ||
264 | /** | |
b74c79e9 | 265 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 266 | * @inode: inode to check access rights for |
8fd90c8d | 267 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
268 | * |
269 | * Used to check for read/write/execute permissions on a file. | |
270 | * We use "fsuid" for this, letting us set arbitrary permissions | |
271 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
272 | * are used for other things. |
273 | * | |
274 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
275 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
276 | * It would then be called again in ref-walk mode. | |
5909ccaa | 277 | */ |
2830ba7f | 278 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
279 | { |
280 | int ret; | |
281 | ||
282 | /* | |
948409c7 | 283 | * Do the basic permission checks. |
5909ccaa | 284 | */ |
7e40145e | 285 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
286 | if (ret != -EACCES) |
287 | return ret; | |
1da177e4 | 288 | |
d594e7ec AV |
289 | if (S_ISDIR(inode->i_mode)) { |
290 | /* DACs are overridable for directories */ | |
1a48e2ac | 291 | if (inode_capable(inode, CAP_DAC_OVERRIDE)) |
d594e7ec AV |
292 | return 0; |
293 | if (!(mask & MAY_WRITE)) | |
1a48e2ac | 294 | if (inode_capable(inode, CAP_DAC_READ_SEARCH)) |
d594e7ec AV |
295 | return 0; |
296 | return -EACCES; | |
297 | } | |
1da177e4 LT |
298 | /* |
299 | * Read/write DACs are always overridable. | |
d594e7ec AV |
300 | * Executable DACs are overridable when there is |
301 | * at least one exec bit set. | |
1da177e4 | 302 | */ |
d594e7ec | 303 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) |
1a48e2ac | 304 | if (inode_capable(inode, CAP_DAC_OVERRIDE)) |
1da177e4 LT |
305 | return 0; |
306 | ||
307 | /* | |
308 | * Searching includes executable on directories, else just read. | |
309 | */ | |
7ea66001 | 310 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 311 | if (mask == MAY_READ) |
1a48e2ac | 312 | if (inode_capable(inode, CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
313 | return 0; |
314 | ||
315 | return -EACCES; | |
316 | } | |
317 | ||
3ddcd056 LT |
318 | /* |
319 | * We _really_ want to just do "generic_permission()" without | |
320 | * even looking at the inode->i_op values. So we keep a cache | |
321 | * flag in inode->i_opflags, that says "this has not special | |
322 | * permission function, use the fast case". | |
323 | */ | |
324 | static inline int do_inode_permission(struct inode *inode, int mask) | |
325 | { | |
326 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
327 | if (likely(inode->i_op->permission)) | |
328 | return inode->i_op->permission(inode, mask); | |
329 | ||
330 | /* This gets set once for the inode lifetime */ | |
331 | spin_lock(&inode->i_lock); | |
332 | inode->i_opflags |= IOP_FASTPERM; | |
333 | spin_unlock(&inode->i_lock); | |
334 | } | |
335 | return generic_permission(inode, mask); | |
336 | } | |
337 | ||
cb23beb5 | 338 | /** |
0bdaea90 DH |
339 | * __inode_permission - Check for access rights to a given inode |
340 | * @inode: Inode to check permission on | |
341 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
cb23beb5 | 342 | * |
0bdaea90 | 343 | * Check for read/write/execute permissions on an inode. |
948409c7 AG |
344 | * |
345 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
0bdaea90 DH |
346 | * |
347 | * This does not check for a read-only file system. You probably want | |
348 | * inode_permission(). | |
cb23beb5 | 349 | */ |
0bdaea90 | 350 | int __inode_permission(struct inode *inode, int mask) |
1da177e4 | 351 | { |
e6305c43 | 352 | int retval; |
1da177e4 | 353 | |
3ddcd056 | 354 | if (unlikely(mask & MAY_WRITE)) { |
1da177e4 LT |
355 | /* |
356 | * Nobody gets write access to an immutable file. | |
357 | */ | |
358 | if (IS_IMMUTABLE(inode)) | |
359 | return -EACCES; | |
360 | } | |
361 | ||
3ddcd056 | 362 | retval = do_inode_permission(inode, mask); |
1da177e4 LT |
363 | if (retval) |
364 | return retval; | |
365 | ||
08ce5f16 SH |
366 | retval = devcgroup_inode_permission(inode, mask); |
367 | if (retval) | |
368 | return retval; | |
369 | ||
d09ca739 | 370 | return security_inode_permission(inode, mask); |
1da177e4 LT |
371 | } |
372 | ||
0bdaea90 DH |
373 | /** |
374 | * sb_permission - Check superblock-level permissions | |
375 | * @sb: Superblock of inode to check permission on | |
55852635 | 376 | * @inode: Inode to check permission on |
0bdaea90 DH |
377 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) |
378 | * | |
379 | * Separate out file-system wide checks from inode-specific permission checks. | |
380 | */ | |
381 | static int sb_permission(struct super_block *sb, struct inode *inode, int mask) | |
382 | { | |
383 | if (unlikely(mask & MAY_WRITE)) { | |
384 | umode_t mode = inode->i_mode; | |
385 | ||
386 | /* Nobody gets write access to a read-only fs. */ | |
387 | if ((sb->s_flags & MS_RDONLY) && | |
388 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
389 | return -EROFS; | |
390 | } | |
391 | return 0; | |
392 | } | |
393 | ||
394 | /** | |
395 | * inode_permission - Check for access rights to a given inode | |
396 | * @inode: Inode to check permission on | |
397 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
398 | * | |
399 | * Check for read/write/execute permissions on an inode. We use fs[ug]id for | |
400 | * this, letting us set arbitrary permissions for filesystem access without | |
401 | * changing the "normal" UIDs which are used for other things. | |
402 | * | |
403 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
404 | */ | |
405 | int inode_permission(struct inode *inode, int mask) | |
406 | { | |
407 | int retval; | |
408 | ||
409 | retval = sb_permission(inode->i_sb, inode, mask); | |
410 | if (retval) | |
411 | return retval; | |
412 | return __inode_permission(inode, mask); | |
413 | } | |
414 | ||
5dd784d0 JB |
415 | /** |
416 | * path_get - get a reference to a path | |
417 | * @path: path to get the reference to | |
418 | * | |
419 | * Given a path increment the reference count to the dentry and the vfsmount. | |
420 | */ | |
421 | void path_get(struct path *path) | |
422 | { | |
423 | mntget(path->mnt); | |
424 | dget(path->dentry); | |
425 | } | |
426 | EXPORT_SYMBOL(path_get); | |
427 | ||
1d957f9b JB |
428 | /** |
429 | * path_put - put a reference to a path | |
430 | * @path: path to put the reference to | |
431 | * | |
432 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
433 | */ | |
434 | void path_put(struct path *path) | |
1da177e4 | 435 | { |
1d957f9b JB |
436 | dput(path->dentry); |
437 | mntput(path->mnt); | |
1da177e4 | 438 | } |
1d957f9b | 439 | EXPORT_SYMBOL(path_put); |
1da177e4 | 440 | |
19660af7 | 441 | /* |
31e6b01f | 442 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
443 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
444 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
445 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
446 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
447 | * got stuck, so ref-walk may continue from there. If this is not successful | |
448 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
449 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 450 | */ |
31e6b01f | 451 | |
32a7991b AV |
452 | static inline void lock_rcu_walk(void) |
453 | { | |
454 | br_read_lock(&vfsmount_lock); | |
455 | rcu_read_lock(); | |
456 | } | |
457 | ||
458 | static inline void unlock_rcu_walk(void) | |
459 | { | |
460 | rcu_read_unlock(); | |
461 | br_read_unlock(&vfsmount_lock); | |
462 | } | |
463 | ||
31e6b01f | 464 | /** |
19660af7 AV |
465 | * unlazy_walk - try to switch to ref-walk mode. |
466 | * @nd: nameidata pathwalk data | |
467 | * @dentry: child of nd->path.dentry or NULL | |
39191628 | 468 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 469 | * |
19660af7 AV |
470 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
471 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
472 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 473 | */ |
19660af7 | 474 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry) |
31e6b01f NP |
475 | { |
476 | struct fs_struct *fs = current->fs; | |
477 | struct dentry *parent = nd->path.dentry; | |
5b6ca027 | 478 | int want_root = 0; |
31e6b01f NP |
479 | |
480 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
5b6ca027 AV |
481 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
482 | want_root = 1; | |
31e6b01f NP |
483 | spin_lock(&fs->lock); |
484 | if (nd->root.mnt != fs->root.mnt || | |
485 | nd->root.dentry != fs->root.dentry) | |
486 | goto err_root; | |
487 | } | |
488 | spin_lock(&parent->d_lock); | |
19660af7 AV |
489 | if (!dentry) { |
490 | if (!__d_rcu_to_refcount(parent, nd->seq)) | |
491 | goto err_parent; | |
492 | BUG_ON(nd->inode != parent->d_inode); | |
493 | } else { | |
94c0d4ec AV |
494 | if (dentry->d_parent != parent) |
495 | goto err_parent; | |
19660af7 AV |
496 | spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED); |
497 | if (!__d_rcu_to_refcount(dentry, nd->seq)) | |
498 | goto err_child; | |
499 | /* | |
500 | * If the sequence check on the child dentry passed, then | |
501 | * the child has not been removed from its parent. This | |
502 | * means the parent dentry must be valid and able to take | |
503 | * a reference at this point. | |
504 | */ | |
505 | BUG_ON(!IS_ROOT(dentry) && dentry->d_parent != parent); | |
506 | BUG_ON(!parent->d_count); | |
507 | parent->d_count++; | |
508 | spin_unlock(&dentry->d_lock); | |
509 | } | |
31e6b01f | 510 | spin_unlock(&parent->d_lock); |
5b6ca027 | 511 | if (want_root) { |
31e6b01f NP |
512 | path_get(&nd->root); |
513 | spin_unlock(&fs->lock); | |
514 | } | |
515 | mntget(nd->path.mnt); | |
516 | ||
32a7991b | 517 | unlock_rcu_walk(); |
31e6b01f NP |
518 | nd->flags &= ~LOOKUP_RCU; |
519 | return 0; | |
19660af7 AV |
520 | |
521 | err_child: | |
31e6b01f | 522 | spin_unlock(&dentry->d_lock); |
19660af7 | 523 | err_parent: |
31e6b01f NP |
524 | spin_unlock(&parent->d_lock); |
525 | err_root: | |
5b6ca027 | 526 | if (want_root) |
31e6b01f NP |
527 | spin_unlock(&fs->lock); |
528 | return -ECHILD; | |
529 | } | |
530 | ||
4ce16ef3 | 531 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 532 | { |
4ce16ef3 | 533 | return dentry->d_op->d_revalidate(dentry, flags); |
34286d66 NP |
534 | } |
535 | ||
9f1fafee AV |
536 | /** |
537 | * complete_walk - successful completion of path walk | |
538 | * @nd: pointer nameidata | |
39159de2 | 539 | * |
9f1fafee AV |
540 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
541 | * Revalidate the final result, unless we'd already done that during | |
542 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
543 | * success, -error on failure. In case of failure caller does not | |
544 | * need to drop nd->path. | |
39159de2 | 545 | */ |
9f1fafee | 546 | static int complete_walk(struct nameidata *nd) |
39159de2 | 547 | { |
16c2cd71 | 548 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 549 | int status; |
39159de2 | 550 | |
9f1fafee AV |
551 | if (nd->flags & LOOKUP_RCU) { |
552 | nd->flags &= ~LOOKUP_RCU; | |
553 | if (!(nd->flags & LOOKUP_ROOT)) | |
554 | nd->root.mnt = NULL; | |
555 | spin_lock(&dentry->d_lock); | |
556 | if (unlikely(!__d_rcu_to_refcount(dentry, nd->seq))) { | |
557 | spin_unlock(&dentry->d_lock); | |
32a7991b | 558 | unlock_rcu_walk(); |
9f1fafee AV |
559 | return -ECHILD; |
560 | } | |
561 | BUG_ON(nd->inode != dentry->d_inode); | |
562 | spin_unlock(&dentry->d_lock); | |
563 | mntget(nd->path.mnt); | |
32a7991b | 564 | unlock_rcu_walk(); |
9f1fafee AV |
565 | } |
566 | ||
16c2cd71 AV |
567 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
568 | return 0; | |
569 | ||
570 | if (likely(!(dentry->d_flags & DCACHE_OP_REVALIDATE))) | |
39159de2 JL |
571 | return 0; |
572 | ||
16c2cd71 AV |
573 | if (likely(!(dentry->d_sb->s_type->fs_flags & FS_REVAL_DOT))) |
574 | return 0; | |
575 | ||
576 | /* Note: we do not d_invalidate() */ | |
4ce16ef3 | 577 | status = d_revalidate(dentry, nd->flags); |
39159de2 JL |
578 | if (status > 0) |
579 | return 0; | |
580 | ||
16c2cd71 | 581 | if (!status) |
39159de2 | 582 | status = -ESTALE; |
16c2cd71 | 583 | |
9f1fafee | 584 | path_put(&nd->path); |
39159de2 JL |
585 | return status; |
586 | } | |
587 | ||
2a737871 AV |
588 | static __always_inline void set_root(struct nameidata *nd) |
589 | { | |
f7ad3c6b MS |
590 | if (!nd->root.mnt) |
591 | get_fs_root(current->fs, &nd->root); | |
2a737871 AV |
592 | } |
593 | ||
6de88d72 AV |
594 | static int link_path_walk(const char *, struct nameidata *); |
595 | ||
31e6b01f NP |
596 | static __always_inline void set_root_rcu(struct nameidata *nd) |
597 | { | |
598 | if (!nd->root.mnt) { | |
599 | struct fs_struct *fs = current->fs; | |
c28cc364 NP |
600 | unsigned seq; |
601 | ||
602 | do { | |
603 | seq = read_seqcount_begin(&fs->seq); | |
604 | nd->root = fs->root; | |
c1530019 | 605 | nd->seq = __read_seqcount_begin(&nd->root.dentry->d_seq); |
c28cc364 | 606 | } while (read_seqcount_retry(&fs->seq, seq)); |
31e6b01f NP |
607 | } |
608 | } | |
609 | ||
f1662356 | 610 | static __always_inline int __vfs_follow_link(struct nameidata *nd, const char *link) |
1da177e4 | 611 | { |
31e6b01f NP |
612 | int ret; |
613 | ||
1da177e4 LT |
614 | if (IS_ERR(link)) |
615 | goto fail; | |
616 | ||
617 | if (*link == '/') { | |
2a737871 | 618 | set_root(nd); |
1d957f9b | 619 | path_put(&nd->path); |
2a737871 AV |
620 | nd->path = nd->root; |
621 | path_get(&nd->root); | |
16c2cd71 | 622 | nd->flags |= LOOKUP_JUMPED; |
1da177e4 | 623 | } |
31e6b01f | 624 | nd->inode = nd->path.dentry->d_inode; |
b4091d5f | 625 | |
31e6b01f NP |
626 | ret = link_path_walk(link, nd); |
627 | return ret; | |
1da177e4 | 628 | fail: |
1d957f9b | 629 | path_put(&nd->path); |
1da177e4 LT |
630 | return PTR_ERR(link); |
631 | } | |
632 | ||
1d957f9b | 633 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
634 | { |
635 | dput(path->dentry); | |
4ac91378 | 636 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
637 | mntput(path->mnt); |
638 | } | |
639 | ||
7b9337aa NP |
640 | static inline void path_to_nameidata(const struct path *path, |
641 | struct nameidata *nd) | |
051d3812 | 642 | { |
31e6b01f NP |
643 | if (!(nd->flags & LOOKUP_RCU)) { |
644 | dput(nd->path.dentry); | |
645 | if (nd->path.mnt != path->mnt) | |
646 | mntput(nd->path.mnt); | |
9a229683 | 647 | } |
31e6b01f | 648 | nd->path.mnt = path->mnt; |
4ac91378 | 649 | nd->path.dentry = path->dentry; |
051d3812 IK |
650 | } |
651 | ||
b5fb63c1 CH |
652 | /* |
653 | * Helper to directly jump to a known parsed path from ->follow_link, | |
654 | * caller must have taken a reference to path beforehand. | |
655 | */ | |
656 | void nd_jump_link(struct nameidata *nd, struct path *path) | |
657 | { | |
658 | path_put(&nd->path); | |
659 | ||
660 | nd->path = *path; | |
661 | nd->inode = nd->path.dentry->d_inode; | |
662 | nd->flags |= LOOKUP_JUMPED; | |
663 | ||
664 | BUG_ON(nd->inode->i_op->follow_link); | |
665 | } | |
666 | ||
574197e0 AV |
667 | static inline void put_link(struct nameidata *nd, struct path *link, void *cookie) |
668 | { | |
669 | struct inode *inode = link->dentry->d_inode; | |
6d7b5aae | 670 | if (inode->i_op->put_link) |
574197e0 AV |
671 | inode->i_op->put_link(link->dentry, nd, cookie); |
672 | path_put(link); | |
673 | } | |
674 | ||
800179c9 KC |
675 | int sysctl_protected_symlinks __read_mostly = 1; |
676 | int sysctl_protected_hardlinks __read_mostly = 1; | |
677 | ||
678 | /** | |
679 | * may_follow_link - Check symlink following for unsafe situations | |
680 | * @link: The path of the symlink | |
55852635 | 681 | * @nd: nameidata pathwalk data |
800179c9 KC |
682 | * |
683 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
684 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
685 | * in a sticky world-writable directory. This is to protect privileged | |
686 | * processes from failing races against path names that may change out | |
687 | * from under them by way of other users creating malicious symlinks. | |
688 | * It will permit symlinks to be followed only when outside a sticky | |
689 | * world-writable directory, or when the uid of the symlink and follower | |
690 | * match, or when the directory owner matches the symlink's owner. | |
691 | * | |
692 | * Returns 0 if following the symlink is allowed, -ve on error. | |
693 | */ | |
694 | static inline int may_follow_link(struct path *link, struct nameidata *nd) | |
695 | { | |
696 | const struct inode *inode; | |
697 | const struct inode *parent; | |
698 | ||
699 | if (!sysctl_protected_symlinks) | |
700 | return 0; | |
701 | ||
702 | /* Allowed if owner and follower match. */ | |
703 | inode = link->dentry->d_inode; | |
81abe27b | 704 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
705 | return 0; |
706 | ||
707 | /* Allowed if parent directory not sticky and world-writable. */ | |
708 | parent = nd->path.dentry->d_inode; | |
709 | if ((parent->i_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) | |
710 | return 0; | |
711 | ||
712 | /* Allowed if parent directory and link owner match. */ | |
81abe27b | 713 | if (uid_eq(parent->i_uid, inode->i_uid)) |
800179c9 KC |
714 | return 0; |
715 | ||
ffd8d101 | 716 | audit_log_link_denied("follow_link", link); |
800179c9 KC |
717 | path_put_conditional(link, nd); |
718 | path_put(&nd->path); | |
719 | return -EACCES; | |
720 | } | |
721 | ||
722 | /** | |
723 | * safe_hardlink_source - Check for safe hardlink conditions | |
724 | * @inode: the source inode to hardlink from | |
725 | * | |
726 | * Return false if at least one of the following conditions: | |
727 | * - inode is not a regular file | |
728 | * - inode is setuid | |
729 | * - inode is setgid and group-exec | |
730 | * - access failure for read and write | |
731 | * | |
732 | * Otherwise returns true. | |
733 | */ | |
734 | static bool safe_hardlink_source(struct inode *inode) | |
735 | { | |
736 | umode_t mode = inode->i_mode; | |
737 | ||
738 | /* Special files should not get pinned to the filesystem. */ | |
739 | if (!S_ISREG(mode)) | |
740 | return false; | |
741 | ||
742 | /* Setuid files should not get pinned to the filesystem. */ | |
743 | if (mode & S_ISUID) | |
744 | return false; | |
745 | ||
746 | /* Executable setgid files should not get pinned to the filesystem. */ | |
747 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
748 | return false; | |
749 | ||
750 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
751 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
752 | return false; | |
753 | ||
754 | return true; | |
755 | } | |
756 | ||
757 | /** | |
758 | * may_linkat - Check permissions for creating a hardlink | |
759 | * @link: the source to hardlink from | |
760 | * | |
761 | * Block hardlink when all of: | |
762 | * - sysctl_protected_hardlinks enabled | |
763 | * - fsuid does not match inode | |
764 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
765 | * - not CAP_FOWNER | |
766 | * | |
767 | * Returns 0 if successful, -ve on error. | |
768 | */ | |
769 | static int may_linkat(struct path *link) | |
770 | { | |
771 | const struct cred *cred; | |
772 | struct inode *inode; | |
773 | ||
774 | if (!sysctl_protected_hardlinks) | |
775 | return 0; | |
776 | ||
777 | cred = current_cred(); | |
778 | inode = link->dentry->d_inode; | |
779 | ||
780 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, | |
781 | * otherwise, it must be a safe source. | |
782 | */ | |
81abe27b | 783 | if (uid_eq(cred->fsuid, inode->i_uid) || safe_hardlink_source(inode) || |
800179c9 KC |
784 | capable(CAP_FOWNER)) |
785 | return 0; | |
786 | ||
a51d9eaa | 787 | audit_log_link_denied("linkat", link); |
800179c9 KC |
788 | return -EPERM; |
789 | } | |
790 | ||
def4af30 | 791 | static __always_inline int |
574197e0 | 792 | follow_link(struct path *link, struct nameidata *nd, void **p) |
1da177e4 | 793 | { |
7b9337aa | 794 | struct dentry *dentry = link->dentry; |
6d7b5aae AV |
795 | int error; |
796 | char *s; | |
1da177e4 | 797 | |
844a3917 AV |
798 | BUG_ON(nd->flags & LOOKUP_RCU); |
799 | ||
0e794589 AV |
800 | if (link->mnt == nd->path.mnt) |
801 | mntget(link->mnt); | |
802 | ||
6d7b5aae AV |
803 | error = -ELOOP; |
804 | if (unlikely(current->total_link_count >= 40)) | |
805 | goto out_put_nd_path; | |
806 | ||
574197e0 AV |
807 | cond_resched(); |
808 | current->total_link_count++; | |
809 | ||
68ac1234 | 810 | touch_atime(link); |
1da177e4 | 811 | nd_set_link(nd, NULL); |
cd4e91d3 | 812 | |
36f3b4f6 | 813 | error = security_inode_follow_link(link->dentry, nd); |
6d7b5aae AV |
814 | if (error) |
815 | goto out_put_nd_path; | |
36f3b4f6 | 816 | |
86acdca1 | 817 | nd->last_type = LAST_BIND; |
def4af30 AV |
818 | *p = dentry->d_inode->i_op->follow_link(dentry, nd); |
819 | error = PTR_ERR(*p); | |
6d7b5aae | 820 | if (IS_ERR(*p)) |
408ef013 | 821 | goto out_put_nd_path; |
6d7b5aae AV |
822 | |
823 | error = 0; | |
824 | s = nd_get_link(nd); | |
825 | if (s) { | |
826 | error = __vfs_follow_link(nd, s); | |
b5fb63c1 CH |
827 | if (unlikely(error)) |
828 | put_link(nd, link, *p); | |
1da177e4 | 829 | } |
6d7b5aae AV |
830 | |
831 | return error; | |
832 | ||
833 | out_put_nd_path: | |
98f6ef64 | 834 | *p = NULL; |
6d7b5aae | 835 | path_put(&nd->path); |
6d7b5aae | 836 | path_put(link); |
1da177e4 LT |
837 | return error; |
838 | } | |
839 | ||
31e6b01f NP |
840 | static int follow_up_rcu(struct path *path) |
841 | { | |
0714a533 AV |
842 | struct mount *mnt = real_mount(path->mnt); |
843 | struct mount *parent; | |
31e6b01f NP |
844 | struct dentry *mountpoint; |
845 | ||
0714a533 AV |
846 | parent = mnt->mnt_parent; |
847 | if (&parent->mnt == path->mnt) | |
31e6b01f | 848 | return 0; |
a73324da | 849 | mountpoint = mnt->mnt_mountpoint; |
31e6b01f | 850 | path->dentry = mountpoint; |
0714a533 | 851 | path->mnt = &parent->mnt; |
31e6b01f NP |
852 | return 1; |
853 | } | |
854 | ||
f015f126 DH |
855 | /* |
856 | * follow_up - Find the mountpoint of path's vfsmount | |
857 | * | |
858 | * Given a path, find the mountpoint of its source file system. | |
859 | * Replace @path with the path of the mountpoint in the parent mount. | |
860 | * Up is towards /. | |
861 | * | |
862 | * Return 1 if we went up a level and 0 if we were already at the | |
863 | * root. | |
864 | */ | |
bab77ebf | 865 | int follow_up(struct path *path) |
1da177e4 | 866 | { |
0714a533 AV |
867 | struct mount *mnt = real_mount(path->mnt); |
868 | struct mount *parent; | |
1da177e4 | 869 | struct dentry *mountpoint; |
99b7db7b | 870 | |
962830df | 871 | br_read_lock(&vfsmount_lock); |
0714a533 | 872 | parent = mnt->mnt_parent; |
3c0a6163 | 873 | if (parent == mnt) { |
962830df | 874 | br_read_unlock(&vfsmount_lock); |
1da177e4 LT |
875 | return 0; |
876 | } | |
0714a533 | 877 | mntget(&parent->mnt); |
a73324da | 878 | mountpoint = dget(mnt->mnt_mountpoint); |
962830df | 879 | br_read_unlock(&vfsmount_lock); |
bab77ebf AV |
880 | dput(path->dentry); |
881 | path->dentry = mountpoint; | |
882 | mntput(path->mnt); | |
0714a533 | 883 | path->mnt = &parent->mnt; |
1da177e4 LT |
884 | return 1; |
885 | } | |
886 | ||
b5c84bf6 | 887 | /* |
9875cf80 DH |
888 | * Perform an automount |
889 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
890 | * were called with. | |
1da177e4 | 891 | */ |
9875cf80 DH |
892 | static int follow_automount(struct path *path, unsigned flags, |
893 | bool *need_mntput) | |
31e6b01f | 894 | { |
9875cf80 | 895 | struct vfsmount *mnt; |
ea5b778a | 896 | int err; |
9875cf80 DH |
897 | |
898 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
899 | return -EREMOTE; | |
900 | ||
0ec26fd0 MS |
901 | /* We don't want to mount if someone's just doing a stat - |
902 | * unless they're stat'ing a directory and appended a '/' to | |
903 | * the name. | |
904 | * | |
905 | * We do, however, want to mount if someone wants to open or | |
906 | * create a file of any type under the mountpoint, wants to | |
907 | * traverse through the mountpoint or wants to open the | |
908 | * mounted directory. Also, autofs may mark negative dentries | |
909 | * as being automount points. These will need the attentions | |
910 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 911 | */ |
0ec26fd0 | 912 | if (!(flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
d94c177b | 913 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
0ec26fd0 MS |
914 | path->dentry->d_inode) |
915 | return -EISDIR; | |
916 | ||
9875cf80 DH |
917 | current->total_link_count++; |
918 | if (current->total_link_count >= 40) | |
919 | return -ELOOP; | |
920 | ||
921 | mnt = path->dentry->d_op->d_automount(path); | |
922 | if (IS_ERR(mnt)) { | |
923 | /* | |
924 | * The filesystem is allowed to return -EISDIR here to indicate | |
925 | * it doesn't want to automount. For instance, autofs would do | |
926 | * this so that its userspace daemon can mount on this dentry. | |
927 | * | |
928 | * However, we can only permit this if it's a terminal point in | |
929 | * the path being looked up; if it wasn't then the remainder of | |
930 | * the path is inaccessible and we should say so. | |
931 | */ | |
49084c3b | 932 | if (PTR_ERR(mnt) == -EISDIR && (flags & LOOKUP_PARENT)) |
9875cf80 DH |
933 | return -EREMOTE; |
934 | return PTR_ERR(mnt); | |
31e6b01f | 935 | } |
ea5b778a | 936 | |
9875cf80 DH |
937 | if (!mnt) /* mount collision */ |
938 | return 0; | |
31e6b01f | 939 | |
8aef1884 AV |
940 | if (!*need_mntput) { |
941 | /* lock_mount() may release path->mnt on error */ | |
942 | mntget(path->mnt); | |
943 | *need_mntput = true; | |
944 | } | |
19a167af | 945 | err = finish_automount(mnt, path); |
9875cf80 | 946 | |
ea5b778a DH |
947 | switch (err) { |
948 | case -EBUSY: | |
949 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 950 | return 0; |
ea5b778a | 951 | case 0: |
8aef1884 | 952 | path_put(path); |
ea5b778a DH |
953 | path->mnt = mnt; |
954 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 955 | return 0; |
19a167af AV |
956 | default: |
957 | return err; | |
ea5b778a | 958 | } |
19a167af | 959 | |
463ffb2e AV |
960 | } |
961 | ||
9875cf80 DH |
962 | /* |
963 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 964 | * - Flagged for transit management (autofs) |
9875cf80 DH |
965 | * - Flagged as mountpoint |
966 | * - Flagged as automount point | |
967 | * | |
968 | * This may only be called in refwalk mode. | |
969 | * | |
970 | * Serialization is taken care of in namespace.c | |
971 | */ | |
972 | static int follow_managed(struct path *path, unsigned flags) | |
1da177e4 | 973 | { |
8aef1884 | 974 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
975 | unsigned managed; |
976 | bool need_mntput = false; | |
8aef1884 | 977 | int ret = 0; |
9875cf80 DH |
978 | |
979 | /* Given that we're not holding a lock here, we retain the value in a | |
980 | * local variable for each dentry as we look at it so that we don't see | |
981 | * the components of that value change under us */ | |
982 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
983 | managed &= DCACHE_MANAGED_DENTRY, | |
984 | unlikely(managed != 0)) { | |
cc53ce53 DH |
985 | /* Allow the filesystem to manage the transit without i_mutex |
986 | * being held. */ | |
987 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
988 | BUG_ON(!path->dentry->d_op); | |
989 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 990 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 991 | if (ret < 0) |
8aef1884 | 992 | break; |
cc53ce53 DH |
993 | } |
994 | ||
9875cf80 DH |
995 | /* Transit to a mounted filesystem. */ |
996 | if (managed & DCACHE_MOUNTED) { | |
997 | struct vfsmount *mounted = lookup_mnt(path); | |
998 | if (mounted) { | |
999 | dput(path->dentry); | |
1000 | if (need_mntput) | |
1001 | mntput(path->mnt); | |
1002 | path->mnt = mounted; | |
1003 | path->dentry = dget(mounted->mnt_root); | |
1004 | need_mntput = true; | |
1005 | continue; | |
1006 | } | |
1007 | ||
1008 | /* Something is mounted on this dentry in another | |
1009 | * namespace and/or whatever was mounted there in this | |
1010 | * namespace got unmounted before we managed to get the | |
1011 | * vfsmount_lock */ | |
1012 | } | |
1013 | ||
1014 | /* Handle an automount point */ | |
1015 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
1016 | ret = follow_automount(path, flags, &need_mntput); | |
1017 | if (ret < 0) | |
8aef1884 | 1018 | break; |
9875cf80 DH |
1019 | continue; |
1020 | } | |
1021 | ||
1022 | /* We didn't change the current path point */ | |
1023 | break; | |
1da177e4 | 1024 | } |
8aef1884 AV |
1025 | |
1026 | if (need_mntput && path->mnt == mnt) | |
1027 | mntput(path->mnt); | |
1028 | if (ret == -EISDIR) | |
1029 | ret = 0; | |
a3fbbde7 | 1030 | return ret < 0 ? ret : need_mntput; |
1da177e4 LT |
1031 | } |
1032 | ||
cc53ce53 | 1033 | int follow_down_one(struct path *path) |
1da177e4 LT |
1034 | { |
1035 | struct vfsmount *mounted; | |
1036 | ||
1c755af4 | 1037 | mounted = lookup_mnt(path); |
1da177e4 | 1038 | if (mounted) { |
9393bd07 AV |
1039 | dput(path->dentry); |
1040 | mntput(path->mnt); | |
1041 | path->mnt = mounted; | |
1042 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1043 | return 1; |
1044 | } | |
1045 | return 0; | |
1046 | } | |
1047 | ||
62a7375e IK |
1048 | static inline bool managed_dentry_might_block(struct dentry *dentry) |
1049 | { | |
1050 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT && | |
1051 | dentry->d_op->d_manage(dentry, true) < 0); | |
1052 | } | |
1053 | ||
9875cf80 | 1054 | /* |
287548e4 AV |
1055 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1056 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1057 | */ |
1058 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
287548e4 | 1059 | struct inode **inode) |
9875cf80 | 1060 | { |
62a7375e | 1061 | for (;;) { |
c7105365 | 1062 | struct mount *mounted; |
62a7375e IK |
1063 | /* |
1064 | * Don't forget we might have a non-mountpoint managed dentry | |
1065 | * that wants to block transit. | |
1066 | */ | |
287548e4 | 1067 | if (unlikely(managed_dentry_might_block(path->dentry))) |
ab90911f | 1068 | return false; |
62a7375e IK |
1069 | |
1070 | if (!d_mountpoint(path->dentry)) | |
1071 | break; | |
1072 | ||
9875cf80 DH |
1073 | mounted = __lookup_mnt(path->mnt, path->dentry, 1); |
1074 | if (!mounted) | |
1075 | break; | |
c7105365 AV |
1076 | path->mnt = &mounted->mnt; |
1077 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1078 | nd->flags |= LOOKUP_JUMPED; |
9875cf80 | 1079 | nd->seq = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1080 | /* |
1081 | * Update the inode too. We don't need to re-check the | |
1082 | * dentry sequence number here after this d_inode read, | |
1083 | * because a mount-point is always pinned. | |
1084 | */ | |
1085 | *inode = path->dentry->d_inode; | |
9875cf80 | 1086 | } |
9875cf80 DH |
1087 | return true; |
1088 | } | |
1089 | ||
dea39376 | 1090 | static void follow_mount_rcu(struct nameidata *nd) |
287548e4 | 1091 | { |
dea39376 | 1092 | while (d_mountpoint(nd->path.dentry)) { |
c7105365 | 1093 | struct mount *mounted; |
dea39376 | 1094 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry, 1); |
287548e4 AV |
1095 | if (!mounted) |
1096 | break; | |
c7105365 AV |
1097 | nd->path.mnt = &mounted->mnt; |
1098 | nd->path.dentry = mounted->mnt.mnt_root; | |
dea39376 | 1099 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
287548e4 AV |
1100 | } |
1101 | } | |
1102 | ||
31e6b01f NP |
1103 | static int follow_dotdot_rcu(struct nameidata *nd) |
1104 | { | |
31e6b01f NP |
1105 | set_root_rcu(nd); |
1106 | ||
9875cf80 | 1107 | while (1) { |
31e6b01f NP |
1108 | if (nd->path.dentry == nd->root.dentry && |
1109 | nd->path.mnt == nd->root.mnt) { | |
1110 | break; | |
1111 | } | |
1112 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
1113 | struct dentry *old = nd->path.dentry; | |
1114 | struct dentry *parent = old->d_parent; | |
1115 | unsigned seq; | |
1116 | ||
1117 | seq = read_seqcount_begin(&parent->d_seq); | |
1118 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 1119 | goto failed; |
31e6b01f NP |
1120 | nd->path.dentry = parent; |
1121 | nd->seq = seq; | |
1122 | break; | |
1123 | } | |
1124 | if (!follow_up_rcu(&nd->path)) | |
1125 | break; | |
1126 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
31e6b01f | 1127 | } |
dea39376 AV |
1128 | follow_mount_rcu(nd); |
1129 | nd->inode = nd->path.dentry->d_inode; | |
31e6b01f | 1130 | return 0; |
ef7562d5 AV |
1131 | |
1132 | failed: | |
1133 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1134 | if (!(nd->flags & LOOKUP_ROOT)) |
1135 | nd->root.mnt = NULL; | |
32a7991b | 1136 | unlock_rcu_walk(); |
ef7562d5 | 1137 | return -ECHILD; |
31e6b01f NP |
1138 | } |
1139 | ||
cc53ce53 DH |
1140 | /* |
1141 | * Follow down to the covering mount currently visible to userspace. At each | |
1142 | * point, the filesystem owning that dentry may be queried as to whether the | |
1143 | * caller is permitted to proceed or not. | |
cc53ce53 | 1144 | */ |
7cc90cc3 | 1145 | int follow_down(struct path *path) |
cc53ce53 DH |
1146 | { |
1147 | unsigned managed; | |
1148 | int ret; | |
1149 | ||
1150 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1151 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1152 | /* Allow the filesystem to manage the transit without i_mutex | |
1153 | * being held. | |
1154 | * | |
1155 | * We indicate to the filesystem if someone is trying to mount | |
1156 | * something here. This gives autofs the chance to deny anyone | |
1157 | * other than its daemon the right to mount on its | |
1158 | * superstructure. | |
1159 | * | |
1160 | * The filesystem may sleep at this point. | |
1161 | */ | |
1162 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1163 | BUG_ON(!path->dentry->d_op); | |
1164 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 1165 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 1166 | path->dentry, false); |
cc53ce53 DH |
1167 | if (ret < 0) |
1168 | return ret == -EISDIR ? 0 : ret; | |
1169 | } | |
1170 | ||
1171 | /* Transit to a mounted filesystem. */ | |
1172 | if (managed & DCACHE_MOUNTED) { | |
1173 | struct vfsmount *mounted = lookup_mnt(path); | |
1174 | if (!mounted) | |
1175 | break; | |
1176 | dput(path->dentry); | |
1177 | mntput(path->mnt); | |
1178 | path->mnt = mounted; | |
1179 | path->dentry = dget(mounted->mnt_root); | |
1180 | continue; | |
1181 | } | |
1182 | ||
1183 | /* Don't handle automount points here */ | |
1184 | break; | |
1185 | } | |
1186 | return 0; | |
1187 | } | |
1188 | ||
9875cf80 DH |
1189 | /* |
1190 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1191 | */ | |
1192 | static void follow_mount(struct path *path) | |
1193 | { | |
1194 | while (d_mountpoint(path->dentry)) { | |
1195 | struct vfsmount *mounted = lookup_mnt(path); | |
1196 | if (!mounted) | |
1197 | break; | |
1198 | dput(path->dentry); | |
1199 | mntput(path->mnt); | |
1200 | path->mnt = mounted; | |
1201 | path->dentry = dget(mounted->mnt_root); | |
1202 | } | |
1203 | } | |
1204 | ||
31e6b01f | 1205 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1206 | { |
2a737871 | 1207 | set_root(nd); |
e518ddb7 | 1208 | |
1da177e4 | 1209 | while(1) { |
4ac91378 | 1210 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1211 | |
2a737871 AV |
1212 | if (nd->path.dentry == nd->root.dentry && |
1213 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1214 | break; |
1215 | } | |
4ac91378 | 1216 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1217 | /* rare case of legitimate dget_parent()... */ |
1218 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1219 | dput(old); |
1220 | break; | |
1221 | } | |
3088dd70 | 1222 | if (!follow_up(&nd->path)) |
1da177e4 | 1223 | break; |
1da177e4 | 1224 | } |
79ed0226 | 1225 | follow_mount(&nd->path); |
31e6b01f | 1226 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1227 | } |
1228 | ||
baa03890 | 1229 | /* |
bad61189 MS |
1230 | * This looks up the name in dcache, possibly revalidates the old dentry and |
1231 | * allocates a new one if not found or not valid. In the need_lookup argument | |
1232 | * returns whether i_op->lookup is necessary. | |
1233 | * | |
1234 | * dir->d_inode->i_mutex must be held | |
baa03890 | 1235 | */ |
bad61189 | 1236 | static struct dentry *lookup_dcache(struct qstr *name, struct dentry *dir, |
201f956e | 1237 | unsigned int flags, bool *need_lookup) |
baa03890 | 1238 | { |
baa03890 | 1239 | struct dentry *dentry; |
bad61189 | 1240 | int error; |
baa03890 | 1241 | |
bad61189 MS |
1242 | *need_lookup = false; |
1243 | dentry = d_lookup(dir, name); | |
1244 | if (dentry) { | |
1245 | if (d_need_lookup(dentry)) { | |
1246 | *need_lookup = true; | |
1247 | } else if (dentry->d_flags & DCACHE_OP_REVALIDATE) { | |
201f956e | 1248 | error = d_revalidate(dentry, flags); |
bad61189 MS |
1249 | if (unlikely(error <= 0)) { |
1250 | if (error < 0) { | |
1251 | dput(dentry); | |
1252 | return ERR_PTR(error); | |
1253 | } else if (!d_invalidate(dentry)) { | |
1254 | dput(dentry); | |
1255 | dentry = NULL; | |
1256 | } | |
1257 | } | |
1258 | } | |
1259 | } | |
baa03890 | 1260 | |
bad61189 MS |
1261 | if (!dentry) { |
1262 | dentry = d_alloc(dir, name); | |
1263 | if (unlikely(!dentry)) | |
1264 | return ERR_PTR(-ENOMEM); | |
baa03890 | 1265 | |
bad61189 | 1266 | *need_lookup = true; |
baa03890 NP |
1267 | } |
1268 | return dentry; | |
1269 | } | |
1270 | ||
44396f4b | 1271 | /* |
bad61189 MS |
1272 | * Call i_op->lookup on the dentry. The dentry must be negative but may be |
1273 | * hashed if it was pouplated with DCACHE_NEED_LOOKUP. | |
1274 | * | |
1275 | * dir->d_inode->i_mutex must be held | |
44396f4b | 1276 | */ |
bad61189 | 1277 | static struct dentry *lookup_real(struct inode *dir, struct dentry *dentry, |
72bd866a | 1278 | unsigned int flags) |
44396f4b | 1279 | { |
44396f4b JB |
1280 | struct dentry *old; |
1281 | ||
1282 | /* Don't create child dentry for a dead directory. */ | |
bad61189 | 1283 | if (unlikely(IS_DEADDIR(dir))) { |
e188dc02 | 1284 | dput(dentry); |
44396f4b | 1285 | return ERR_PTR(-ENOENT); |
e188dc02 | 1286 | } |
44396f4b | 1287 | |
72bd866a | 1288 | old = dir->i_op->lookup(dir, dentry, flags); |
44396f4b JB |
1289 | if (unlikely(old)) { |
1290 | dput(dentry); | |
1291 | dentry = old; | |
1292 | } | |
1293 | return dentry; | |
1294 | } | |
1295 | ||
a3255546 | 1296 | static struct dentry *__lookup_hash(struct qstr *name, |
72bd866a | 1297 | struct dentry *base, unsigned int flags) |
a3255546 | 1298 | { |
bad61189 | 1299 | bool need_lookup; |
a3255546 AV |
1300 | struct dentry *dentry; |
1301 | ||
72bd866a | 1302 | dentry = lookup_dcache(name, base, flags, &need_lookup); |
bad61189 MS |
1303 | if (!need_lookup) |
1304 | return dentry; | |
a3255546 | 1305 | |
72bd866a | 1306 | return lookup_real(base->d_inode, dentry, flags); |
a3255546 AV |
1307 | } |
1308 | ||
1da177e4 LT |
1309 | /* |
1310 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1311 | * small and for now I'd prefer to have fast path as straight as possible. | |
1312 | * It _is_ time-critical. | |
1313 | */ | |
697f514d MS |
1314 | static int lookup_fast(struct nameidata *nd, struct qstr *name, |
1315 | struct path *path, struct inode **inode) | |
1da177e4 | 1316 | { |
4ac91378 | 1317 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1318 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1319 | int need_reval = 1; |
1320 | int status = 1; | |
9875cf80 DH |
1321 | int err; |
1322 | ||
b04f784e NP |
1323 | /* |
1324 | * Rename seqlock is not required here because in the off chance | |
1325 | * of a false negative due to a concurrent rename, we're going to | |
1326 | * do the non-racy lookup, below. | |
1327 | */ | |
31e6b01f NP |
1328 | if (nd->flags & LOOKUP_RCU) { |
1329 | unsigned seq; | |
12f8ad4b | 1330 | dentry = __d_lookup_rcu(parent, name, &seq, nd->inode); |
5a18fff2 AV |
1331 | if (!dentry) |
1332 | goto unlazy; | |
1333 | ||
12f8ad4b LT |
1334 | /* |
1335 | * This sequence count validates that the inode matches | |
1336 | * the dentry name information from lookup. | |
1337 | */ | |
1338 | *inode = dentry->d_inode; | |
1339 | if (read_seqcount_retry(&dentry->d_seq, seq)) | |
1340 | return -ECHILD; | |
1341 | ||
1342 | /* | |
1343 | * This sequence count validates that the parent had no | |
1344 | * changes while we did the lookup of the dentry above. | |
1345 | * | |
1346 | * The memory barrier in read_seqcount_begin of child is | |
1347 | * enough, we can use __read_seqcount_retry here. | |
1348 | */ | |
31e6b01f NP |
1349 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) |
1350 | return -ECHILD; | |
31e6b01f | 1351 | nd->seq = seq; |
5a18fff2 | 1352 | |
fa4ee159 MS |
1353 | if (unlikely(d_need_lookup(dentry))) |
1354 | goto unlazy; | |
24643087 | 1355 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
4ce16ef3 | 1356 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1357 | if (unlikely(status <= 0)) { |
1358 | if (status != -ECHILD) | |
1359 | need_reval = 0; | |
1360 | goto unlazy; | |
1361 | } | |
24643087 | 1362 | } |
31e6b01f NP |
1363 | path->mnt = mnt; |
1364 | path->dentry = dentry; | |
d6e9bd25 AV |
1365 | if (unlikely(!__follow_mount_rcu(nd, path, inode))) |
1366 | goto unlazy; | |
1367 | if (unlikely(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT)) | |
1368 | goto unlazy; | |
1369 | return 0; | |
5a18fff2 | 1370 | unlazy: |
19660af7 AV |
1371 | if (unlazy_walk(nd, dentry)) |
1372 | return -ECHILD; | |
5a18fff2 AV |
1373 | } else { |
1374 | dentry = __d_lookup(parent, name); | |
9875cf80 | 1375 | } |
5a18fff2 | 1376 | |
81e6f520 AV |
1377 | if (unlikely(!dentry)) |
1378 | goto need_lookup; | |
1379 | ||
1380 | if (unlikely(d_need_lookup(dentry))) { | |
44396f4b | 1381 | dput(dentry); |
81e6f520 | 1382 | goto need_lookup; |
5a18fff2 | 1383 | } |
81e6f520 | 1384 | |
5a18fff2 | 1385 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) |
4ce16ef3 | 1386 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1387 | if (unlikely(status <= 0)) { |
1388 | if (status < 0) { | |
1389 | dput(dentry); | |
1390 | return status; | |
1391 | } | |
1392 | if (!d_invalidate(dentry)) { | |
1393 | dput(dentry); | |
81e6f520 | 1394 | goto need_lookup; |
5a18fff2 | 1395 | } |
24643087 | 1396 | } |
697f514d | 1397 | |
9875cf80 DH |
1398 | path->mnt = mnt; |
1399 | path->dentry = dentry; | |
1400 | err = follow_managed(path, nd->flags); | |
89312214 IK |
1401 | if (unlikely(err < 0)) { |
1402 | path_put_conditional(path, nd); | |
9875cf80 | 1403 | return err; |
89312214 | 1404 | } |
a3fbbde7 AV |
1405 | if (err) |
1406 | nd->flags |= LOOKUP_JUMPED; | |
9875cf80 | 1407 | *inode = path->dentry->d_inode; |
1da177e4 | 1408 | return 0; |
81e6f520 AV |
1409 | |
1410 | need_lookup: | |
697f514d MS |
1411 | return 1; |
1412 | } | |
1413 | ||
1414 | /* Fast lookup failed, do it the slow way */ | |
1415 | static int lookup_slow(struct nameidata *nd, struct qstr *name, | |
1416 | struct path *path) | |
1417 | { | |
1418 | struct dentry *dentry, *parent; | |
1419 | int err; | |
1420 | ||
1421 | parent = nd->path.dentry; | |
81e6f520 AV |
1422 | BUG_ON(nd->inode != parent->d_inode); |
1423 | ||
1424 | mutex_lock(&parent->d_inode->i_mutex); | |
72bd866a | 1425 | dentry = __lookup_hash(name, parent, nd->flags); |
81e6f520 AV |
1426 | mutex_unlock(&parent->d_inode->i_mutex); |
1427 | if (IS_ERR(dentry)) | |
1428 | return PTR_ERR(dentry); | |
697f514d MS |
1429 | path->mnt = nd->path.mnt; |
1430 | path->dentry = dentry; | |
1431 | err = follow_managed(path, nd->flags); | |
1432 | if (unlikely(err < 0)) { | |
1433 | path_put_conditional(path, nd); | |
1434 | return err; | |
1435 | } | |
1436 | if (err) | |
1437 | nd->flags |= LOOKUP_JUMPED; | |
1438 | return 0; | |
1da177e4 LT |
1439 | } |
1440 | ||
52094c8a AV |
1441 | static inline int may_lookup(struct nameidata *nd) |
1442 | { | |
1443 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1444 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1445 | if (err != -ECHILD) |
1446 | return err; | |
19660af7 | 1447 | if (unlazy_walk(nd, NULL)) |
52094c8a AV |
1448 | return -ECHILD; |
1449 | } | |
4ad5abb3 | 1450 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1451 | } |
1452 | ||
9856fa1b AV |
1453 | static inline int handle_dots(struct nameidata *nd, int type) |
1454 | { | |
1455 | if (type == LAST_DOTDOT) { | |
1456 | if (nd->flags & LOOKUP_RCU) { | |
1457 | if (follow_dotdot_rcu(nd)) | |
1458 | return -ECHILD; | |
1459 | } else | |
1460 | follow_dotdot(nd); | |
1461 | } | |
1462 | return 0; | |
1463 | } | |
1464 | ||
951361f9 AV |
1465 | static void terminate_walk(struct nameidata *nd) |
1466 | { | |
1467 | if (!(nd->flags & LOOKUP_RCU)) { | |
1468 | path_put(&nd->path); | |
1469 | } else { | |
1470 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1471 | if (!(nd->flags & LOOKUP_ROOT)) |
1472 | nd->root.mnt = NULL; | |
32a7991b | 1473 | unlock_rcu_walk(); |
951361f9 AV |
1474 | } |
1475 | } | |
1476 | ||
3ddcd056 LT |
1477 | /* |
1478 | * Do we need to follow links? We _really_ want to be able | |
1479 | * to do this check without having to look at inode->i_op, | |
1480 | * so we keep a cache of "no, this doesn't need follow_link" | |
1481 | * for the common case. | |
1482 | */ | |
7813b94a | 1483 | static inline int should_follow_link(struct inode *inode, int follow) |
3ddcd056 LT |
1484 | { |
1485 | if (unlikely(!(inode->i_opflags & IOP_NOFOLLOW))) { | |
1486 | if (likely(inode->i_op->follow_link)) | |
1487 | return follow; | |
1488 | ||
1489 | /* This gets set once for the inode lifetime */ | |
1490 | spin_lock(&inode->i_lock); | |
1491 | inode->i_opflags |= IOP_NOFOLLOW; | |
1492 | spin_unlock(&inode->i_lock); | |
1493 | } | |
1494 | return 0; | |
1495 | } | |
1496 | ||
ce57dfc1 AV |
1497 | static inline int walk_component(struct nameidata *nd, struct path *path, |
1498 | struct qstr *name, int type, int follow) | |
1499 | { | |
1500 | struct inode *inode; | |
1501 | int err; | |
1502 | /* | |
1503 | * "." and ".." are special - ".." especially so because it has | |
1504 | * to be able to know about the current root directory and | |
1505 | * parent relationships. | |
1506 | */ | |
1507 | if (unlikely(type != LAST_NORM)) | |
1508 | return handle_dots(nd, type); | |
697f514d | 1509 | err = lookup_fast(nd, name, path, &inode); |
ce57dfc1 | 1510 | if (unlikely(err)) { |
697f514d MS |
1511 | if (err < 0) |
1512 | goto out_err; | |
1513 | ||
1514 | err = lookup_slow(nd, name, path); | |
1515 | if (err < 0) | |
1516 | goto out_err; | |
1517 | ||
1518 | inode = path->dentry->d_inode; | |
ce57dfc1 | 1519 | } |
697f514d MS |
1520 | err = -ENOENT; |
1521 | if (!inode) | |
1522 | goto out_path_put; | |
1523 | ||
7813b94a | 1524 | if (should_follow_link(inode, follow)) { |
19660af7 AV |
1525 | if (nd->flags & LOOKUP_RCU) { |
1526 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
697f514d MS |
1527 | err = -ECHILD; |
1528 | goto out_err; | |
19660af7 AV |
1529 | } |
1530 | } | |
ce57dfc1 AV |
1531 | BUG_ON(inode != path->dentry->d_inode); |
1532 | return 1; | |
1533 | } | |
1534 | path_to_nameidata(path, nd); | |
1535 | nd->inode = inode; | |
1536 | return 0; | |
697f514d MS |
1537 | |
1538 | out_path_put: | |
1539 | path_to_nameidata(path, nd); | |
1540 | out_err: | |
1541 | terminate_walk(nd); | |
1542 | return err; | |
ce57dfc1 AV |
1543 | } |
1544 | ||
b356379a AV |
1545 | /* |
1546 | * This limits recursive symlink follows to 8, while | |
1547 | * limiting consecutive symlinks to 40. | |
1548 | * | |
1549 | * Without that kind of total limit, nasty chains of consecutive | |
1550 | * symlinks can cause almost arbitrarily long lookups. | |
1551 | */ | |
1552 | static inline int nested_symlink(struct path *path, struct nameidata *nd) | |
1553 | { | |
1554 | int res; | |
1555 | ||
b356379a AV |
1556 | if (unlikely(current->link_count >= MAX_NESTED_LINKS)) { |
1557 | path_put_conditional(path, nd); | |
1558 | path_put(&nd->path); | |
1559 | return -ELOOP; | |
1560 | } | |
1a4022f8 | 1561 | BUG_ON(nd->depth >= MAX_NESTED_LINKS); |
b356379a AV |
1562 | |
1563 | nd->depth++; | |
1564 | current->link_count++; | |
1565 | ||
1566 | do { | |
1567 | struct path link = *path; | |
1568 | void *cookie; | |
574197e0 AV |
1569 | |
1570 | res = follow_link(&link, nd, &cookie); | |
6d7b5aae AV |
1571 | if (res) |
1572 | break; | |
1573 | res = walk_component(nd, path, &nd->last, | |
1574 | nd->last_type, LOOKUP_FOLLOW); | |
574197e0 | 1575 | put_link(nd, &link, cookie); |
b356379a AV |
1576 | } while (res > 0); |
1577 | ||
1578 | current->link_count--; | |
1579 | nd->depth--; | |
1580 | return res; | |
1581 | } | |
1582 | ||
3ddcd056 LT |
1583 | /* |
1584 | * We really don't want to look at inode->i_op->lookup | |
1585 | * when we don't have to. So we keep a cache bit in | |
1586 | * the inode ->i_opflags field that says "yes, we can | |
1587 | * do lookup on this inode". | |
1588 | */ | |
1589 | static inline int can_lookup(struct inode *inode) | |
1590 | { | |
1591 | if (likely(inode->i_opflags & IOP_LOOKUP)) | |
1592 | return 1; | |
1593 | if (likely(!inode->i_op->lookup)) | |
1594 | return 0; | |
1595 | ||
1596 | /* We do this once for the lifetime of the inode */ | |
1597 | spin_lock(&inode->i_lock); | |
1598 | inode->i_opflags |= IOP_LOOKUP; | |
1599 | spin_unlock(&inode->i_lock); | |
1600 | return 1; | |
1601 | } | |
1602 | ||
bfcfaa77 LT |
1603 | /* |
1604 | * We can do the critical dentry name comparison and hashing | |
1605 | * operations one word at a time, but we are limited to: | |
1606 | * | |
1607 | * - Architectures with fast unaligned word accesses. We could | |
1608 | * do a "get_unaligned()" if this helps and is sufficiently | |
1609 | * fast. | |
1610 | * | |
1611 | * - Little-endian machines (so that we can generate the mask | |
1612 | * of low bytes efficiently). Again, we *could* do a byte | |
1613 | * swapping load on big-endian architectures if that is not | |
1614 | * expensive enough to make the optimization worthless. | |
1615 | * | |
1616 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we | |
1617 | * do not trap on the (extremely unlikely) case of a page | |
1618 | * crossing operation. | |
1619 | * | |
1620 | * - Furthermore, we need an efficient 64-bit compile for the | |
1621 | * 64-bit case in order to generate the "number of bytes in | |
1622 | * the final mask". Again, that could be replaced with a | |
1623 | * efficient population count instruction or similar. | |
1624 | */ | |
1625 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1626 | ||
f68e556e | 1627 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1628 | |
f68e556e | 1629 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1630 | |
1631 | static inline unsigned int fold_hash(unsigned long hash) | |
1632 | { | |
1633 | hash += hash >> (8*sizeof(int)); | |
1634 | return hash; | |
1635 | } | |
1636 | ||
1637 | #else /* 32-bit case */ | |
1638 | ||
bfcfaa77 LT |
1639 | #define fold_hash(x) (x) |
1640 | ||
1641 | #endif | |
1642 | ||
1643 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1644 | { | |
1645 | unsigned long a, mask; | |
1646 | unsigned long hash = 0; | |
1647 | ||
1648 | for (;;) { | |
e419b4cc | 1649 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1650 | if (len < sizeof(unsigned long)) |
1651 | break; | |
1652 | hash += a; | |
f132c5be | 1653 | hash *= 9; |
bfcfaa77 LT |
1654 | name += sizeof(unsigned long); |
1655 | len -= sizeof(unsigned long); | |
1656 | if (!len) | |
1657 | goto done; | |
1658 | } | |
1659 | mask = ~(~0ul << len*8); | |
1660 | hash += mask & a; | |
1661 | done: | |
1662 | return fold_hash(hash); | |
1663 | } | |
1664 | EXPORT_SYMBOL(full_name_hash); | |
1665 | ||
bfcfaa77 LT |
1666 | /* |
1667 | * Calculate the length and hash of the path component, and | |
1668 | * return the length of the component; | |
1669 | */ | |
1670 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1671 | { | |
36126f8f LT |
1672 | unsigned long a, b, adata, bdata, mask, hash, len; |
1673 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; | |
bfcfaa77 LT |
1674 | |
1675 | hash = a = 0; | |
1676 | len = -sizeof(unsigned long); | |
1677 | do { | |
1678 | hash = (hash + a) * 9; | |
1679 | len += sizeof(unsigned long); | |
e419b4cc | 1680 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1681 | b = a ^ REPEAT_BYTE('/'); |
1682 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1683 | ||
1684 | adata = prep_zero_mask(a, adata, &constants); | |
1685 | bdata = prep_zero_mask(b, bdata, &constants); | |
1686 | ||
1687 | mask = create_zero_mask(adata | bdata); | |
1688 | ||
1689 | hash += a & zero_bytemask(mask); | |
bfcfaa77 LT |
1690 | *hashp = fold_hash(hash); |
1691 | ||
36126f8f | 1692 | return len + find_zero(mask); |
bfcfaa77 LT |
1693 | } |
1694 | ||
1695 | #else | |
1696 | ||
0145acc2 LT |
1697 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1698 | { | |
1699 | unsigned long hash = init_name_hash(); | |
1700 | while (len--) | |
1701 | hash = partial_name_hash(*name++, hash); | |
1702 | return end_name_hash(hash); | |
1703 | } | |
ae942ae7 | 1704 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1705 | |
200e9ef7 LT |
1706 | /* |
1707 | * We know there's a real path component here of at least | |
1708 | * one character. | |
1709 | */ | |
1710 | static inline unsigned long hash_name(const char *name, unsigned int *hashp) | |
1711 | { | |
1712 | unsigned long hash = init_name_hash(); | |
1713 | unsigned long len = 0, c; | |
1714 | ||
1715 | c = (unsigned char)*name; | |
1716 | do { | |
1717 | len++; | |
1718 | hash = partial_name_hash(c, hash); | |
1719 | c = (unsigned char)name[len]; | |
1720 | } while (c && c != '/'); | |
1721 | *hashp = end_name_hash(hash); | |
1722 | return len; | |
1723 | } | |
1724 | ||
bfcfaa77 LT |
1725 | #endif |
1726 | ||
1da177e4 LT |
1727 | /* |
1728 | * Name resolution. | |
ea3834d9 PM |
1729 | * This is the basic name resolution function, turning a pathname into |
1730 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1731 | * |
ea3834d9 PM |
1732 | * Returns 0 and nd will have valid dentry and mnt on success. |
1733 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1734 | */ |
6de88d72 | 1735 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 LT |
1736 | { |
1737 | struct path next; | |
1da177e4 | 1738 | int err; |
1da177e4 LT |
1739 | |
1740 | while (*name=='/') | |
1741 | name++; | |
1742 | if (!*name) | |
086e183a | 1743 | return 0; |
1da177e4 | 1744 | |
1da177e4 LT |
1745 | /* At this point we know we have a real path component. */ |
1746 | for(;;) { | |
1da177e4 | 1747 | struct qstr this; |
200e9ef7 | 1748 | long len; |
fe479a58 | 1749 | int type; |
1da177e4 | 1750 | |
52094c8a | 1751 | err = may_lookup(nd); |
1da177e4 LT |
1752 | if (err) |
1753 | break; | |
1754 | ||
200e9ef7 | 1755 | len = hash_name(name, &this.hash); |
1da177e4 | 1756 | this.name = name; |
200e9ef7 | 1757 | this.len = len; |
1da177e4 | 1758 | |
fe479a58 | 1759 | type = LAST_NORM; |
200e9ef7 | 1760 | if (name[0] == '.') switch (len) { |
fe479a58 | 1761 | case 2: |
200e9ef7 | 1762 | if (name[1] == '.') { |
fe479a58 | 1763 | type = LAST_DOTDOT; |
16c2cd71 AV |
1764 | nd->flags |= LOOKUP_JUMPED; |
1765 | } | |
fe479a58 AV |
1766 | break; |
1767 | case 1: | |
1768 | type = LAST_DOT; | |
1769 | } | |
5a202bcd AV |
1770 | if (likely(type == LAST_NORM)) { |
1771 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1772 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd AV |
1773 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
1774 | err = parent->d_op->d_hash(parent, nd->inode, | |
1775 | &this); | |
1776 | if (err < 0) | |
1777 | break; | |
1778 | } | |
1779 | } | |
fe479a58 | 1780 | |
200e9ef7 | 1781 | if (!name[len]) |
1da177e4 | 1782 | goto last_component; |
200e9ef7 LT |
1783 | /* |
1784 | * If it wasn't NUL, we know it was '/'. Skip that | |
1785 | * slash, and continue until no more slashes. | |
1786 | */ | |
1787 | do { | |
1788 | len++; | |
1789 | } while (unlikely(name[len] == '/')); | |
1790 | if (!name[len]) | |
b356379a | 1791 | goto last_component; |
200e9ef7 | 1792 | name += len; |
1da177e4 | 1793 | |
ce57dfc1 AV |
1794 | err = walk_component(nd, &next, &this, type, LOOKUP_FOLLOW); |
1795 | if (err < 0) | |
1796 | return err; | |
1da177e4 | 1797 | |
ce57dfc1 | 1798 | if (err) { |
b356379a | 1799 | err = nested_symlink(&next, nd); |
1da177e4 | 1800 | if (err) |
a7472bab | 1801 | return err; |
31e6b01f | 1802 | } |
3ddcd056 LT |
1803 | if (can_lookup(nd->inode)) |
1804 | continue; | |
1da177e4 | 1805 | err = -ENOTDIR; |
3ddcd056 | 1806 | break; |
1da177e4 LT |
1807 | /* here ends the main loop */ |
1808 | ||
1da177e4 | 1809 | last_component: |
b356379a AV |
1810 | nd->last = this; |
1811 | nd->last_type = type; | |
086e183a | 1812 | return 0; |
1da177e4 | 1813 | } |
951361f9 | 1814 | terminate_walk(nd); |
1da177e4 LT |
1815 | return err; |
1816 | } | |
1817 | ||
70e9b357 AV |
1818 | static int path_init(int dfd, const char *name, unsigned int flags, |
1819 | struct nameidata *nd, struct file **fp) | |
31e6b01f NP |
1820 | { |
1821 | int retval = 0; | |
31e6b01f NP |
1822 | |
1823 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
16c2cd71 | 1824 | nd->flags = flags | LOOKUP_JUMPED; |
31e6b01f | 1825 | nd->depth = 0; |
5b6ca027 AV |
1826 | if (flags & LOOKUP_ROOT) { |
1827 | struct inode *inode = nd->root.dentry->d_inode; | |
73d049a4 AV |
1828 | if (*name) { |
1829 | if (!inode->i_op->lookup) | |
1830 | return -ENOTDIR; | |
1831 | retval = inode_permission(inode, MAY_EXEC); | |
1832 | if (retval) | |
1833 | return retval; | |
1834 | } | |
5b6ca027 AV |
1835 | nd->path = nd->root; |
1836 | nd->inode = inode; | |
1837 | if (flags & LOOKUP_RCU) { | |
32a7991b | 1838 | lock_rcu_walk(); |
5b6ca027 AV |
1839 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
1840 | } else { | |
1841 | path_get(&nd->path); | |
1842 | } | |
1843 | return 0; | |
1844 | } | |
1845 | ||
31e6b01f | 1846 | nd->root.mnt = NULL; |
31e6b01f NP |
1847 | |
1848 | if (*name=='/') { | |
e41f7d4e | 1849 | if (flags & LOOKUP_RCU) { |
32a7991b | 1850 | lock_rcu_walk(); |
e41f7d4e AV |
1851 | set_root_rcu(nd); |
1852 | } else { | |
1853 | set_root(nd); | |
1854 | path_get(&nd->root); | |
1855 | } | |
1856 | nd->path = nd->root; | |
31e6b01f | 1857 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1858 | if (flags & LOOKUP_RCU) { |
1859 | struct fs_struct *fs = current->fs; | |
1860 | unsigned seq; | |
31e6b01f | 1861 | |
32a7991b | 1862 | lock_rcu_walk(); |
c28cc364 | 1863 | |
e41f7d4e AV |
1864 | do { |
1865 | seq = read_seqcount_begin(&fs->seq); | |
1866 | nd->path = fs->pwd; | |
1867 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1868 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1869 | } else { | |
1870 | get_fs_pwd(current->fs, &nd->path); | |
1871 | } | |
31e6b01f | 1872 | } else { |
2903ff01 | 1873 | struct fd f = fdget_raw(dfd); |
31e6b01f NP |
1874 | struct dentry *dentry; |
1875 | ||
2903ff01 AV |
1876 | if (!f.file) |
1877 | return -EBADF; | |
31e6b01f | 1878 | |
2903ff01 | 1879 | dentry = f.file->f_path.dentry; |
31e6b01f | 1880 | |
f52e0c11 | 1881 | if (*name) { |
2903ff01 AV |
1882 | if (!S_ISDIR(dentry->d_inode->i_mode)) { |
1883 | fdput(f); | |
1884 | return -ENOTDIR; | |
1885 | } | |
31e6b01f | 1886 | |
4ad5abb3 | 1887 | retval = inode_permission(dentry->d_inode, MAY_EXEC); |
2903ff01 AV |
1888 | if (retval) { |
1889 | fdput(f); | |
1890 | return retval; | |
1891 | } | |
f52e0c11 | 1892 | } |
31e6b01f | 1893 | |
2903ff01 | 1894 | nd->path = f.file->f_path; |
e41f7d4e | 1895 | if (flags & LOOKUP_RCU) { |
2903ff01 AV |
1896 | if (f.need_put) |
1897 | *fp = f.file; | |
e41f7d4e | 1898 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
32a7991b | 1899 | lock_rcu_walk(); |
e41f7d4e | 1900 | } else { |
2903ff01 AV |
1901 | path_get(&nd->path); |
1902 | fdput(f); | |
e41f7d4e | 1903 | } |
31e6b01f | 1904 | } |
31e6b01f | 1905 | |
31e6b01f | 1906 | nd->inode = nd->path.dentry->d_inode; |
9b4a9b14 | 1907 | return 0; |
9b4a9b14 AV |
1908 | } |
1909 | ||
bd92d7fe AV |
1910 | static inline int lookup_last(struct nameidata *nd, struct path *path) |
1911 | { | |
1912 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1913 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1914 | ||
1915 | nd->flags &= ~LOOKUP_PARENT; | |
1916 | return walk_component(nd, path, &nd->last, nd->last_type, | |
1917 | nd->flags & LOOKUP_FOLLOW); | |
1918 | } | |
1919 | ||
9b4a9b14 | 1920 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
ee0827cd | 1921 | static int path_lookupat(int dfd, const char *name, |
9b4a9b14 AV |
1922 | unsigned int flags, struct nameidata *nd) |
1923 | { | |
70e9b357 | 1924 | struct file *base = NULL; |
bd92d7fe AV |
1925 | struct path path; |
1926 | int err; | |
31e6b01f NP |
1927 | |
1928 | /* | |
1929 | * Path walking is largely split up into 2 different synchronisation | |
1930 | * schemes, rcu-walk and ref-walk (explained in | |
1931 | * Documentation/filesystems/path-lookup.txt). These share much of the | |
1932 | * path walk code, but some things particularly setup, cleanup, and | |
1933 | * following mounts are sufficiently divergent that functions are | |
1934 | * duplicated. Typically there is a function foo(), and its RCU | |
1935 | * analogue, foo_rcu(). | |
1936 | * | |
1937 | * -ECHILD is the error number of choice (just to avoid clashes) that | |
1938 | * is returned if some aspect of an rcu-walk fails. Such an error must | |
1939 | * be handled by restarting a traditional ref-walk (which will always | |
1940 | * be able to complete). | |
1941 | */ | |
bd92d7fe | 1942 | err = path_init(dfd, name, flags | LOOKUP_PARENT, nd, &base); |
ee0827cd | 1943 | |
bd92d7fe AV |
1944 | if (unlikely(err)) |
1945 | return err; | |
ee0827cd AV |
1946 | |
1947 | current->total_link_count = 0; | |
bd92d7fe AV |
1948 | err = link_path_walk(name, nd); |
1949 | ||
1950 | if (!err && !(flags & LOOKUP_PARENT)) { | |
bd92d7fe AV |
1951 | err = lookup_last(nd, &path); |
1952 | while (err > 0) { | |
1953 | void *cookie; | |
1954 | struct path link = path; | |
800179c9 KC |
1955 | err = may_follow_link(&link, nd); |
1956 | if (unlikely(err)) | |
1957 | break; | |
bd92d7fe | 1958 | nd->flags |= LOOKUP_PARENT; |
574197e0 | 1959 | err = follow_link(&link, nd, &cookie); |
6d7b5aae AV |
1960 | if (err) |
1961 | break; | |
1962 | err = lookup_last(nd, &path); | |
574197e0 | 1963 | put_link(nd, &link, cookie); |
bd92d7fe AV |
1964 | } |
1965 | } | |
ee0827cd | 1966 | |
9f1fafee AV |
1967 | if (!err) |
1968 | err = complete_walk(nd); | |
bd92d7fe AV |
1969 | |
1970 | if (!err && nd->flags & LOOKUP_DIRECTORY) { | |
1971 | if (!nd->inode->i_op->lookup) { | |
1972 | path_put(&nd->path); | |
bd23a539 | 1973 | err = -ENOTDIR; |
bd92d7fe AV |
1974 | } |
1975 | } | |
16c2cd71 | 1976 | |
70e9b357 AV |
1977 | if (base) |
1978 | fput(base); | |
ee0827cd | 1979 | |
5b6ca027 | 1980 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
2a737871 AV |
1981 | path_put(&nd->root); |
1982 | nd->root.mnt = NULL; | |
1983 | } | |
bd92d7fe | 1984 | return err; |
ee0827cd | 1985 | } |
31e6b01f | 1986 | |
ee0827cd AV |
1987 | static int do_path_lookup(int dfd, const char *name, |
1988 | unsigned int flags, struct nameidata *nd) | |
1989 | { | |
1990 | int retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); | |
1991 | if (unlikely(retval == -ECHILD)) | |
1992 | retval = path_lookupat(dfd, name, flags, nd); | |
1993 | if (unlikely(retval == -ESTALE)) | |
1994 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); | |
31e6b01f | 1995 | |
f78570dd | 1996 | if (likely(!retval)) |
bfcec708 | 1997 | audit_inode(name, nd->path.dentry, flags & LOOKUP_PARENT); |
170aa3d0 | 1998 | return retval; |
1da177e4 LT |
1999 | } |
2000 | ||
79714f72 AV |
2001 | /* does lookup, returns the object with parent locked */ |
2002 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 2003 | { |
79714f72 AV |
2004 | struct nameidata nd; |
2005 | struct dentry *d; | |
2006 | int err = do_path_lookup(AT_FDCWD, name, LOOKUP_PARENT, &nd); | |
2007 | if (err) | |
2008 | return ERR_PTR(err); | |
2009 | if (nd.last_type != LAST_NORM) { | |
2010 | path_put(&nd.path); | |
2011 | return ERR_PTR(-EINVAL); | |
2012 | } | |
2013 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); | |
1e0ea001 | 2014 | d = __lookup_hash(&nd.last, nd.path.dentry, 0); |
79714f72 AV |
2015 | if (IS_ERR(d)) { |
2016 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); | |
2017 | path_put(&nd.path); | |
2018 | return d; | |
2019 | } | |
2020 | *path = nd.path; | |
2021 | return d; | |
5590ff0d UD |
2022 | } |
2023 | ||
d1811465 AV |
2024 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2025 | { | |
2026 | struct nameidata nd; | |
2027 | int res = do_path_lookup(AT_FDCWD, name, flags, &nd); | |
2028 | if (!res) | |
2029 | *path = nd.path; | |
2030 | return res; | |
2031 | } | |
2032 | ||
16f18200 JJS |
2033 | /** |
2034 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2035 | * @dentry: pointer to dentry of the base directory | |
2036 | * @mnt: pointer to vfs mount of the base directory | |
2037 | * @name: pointer to file name | |
2038 | * @flags: lookup flags | |
e0a01249 | 2039 | * @path: pointer to struct path to fill |
16f18200 JJS |
2040 | */ |
2041 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2042 | const char *name, unsigned int flags, | |
e0a01249 | 2043 | struct path *path) |
16f18200 | 2044 | { |
e0a01249 AV |
2045 | struct nameidata nd; |
2046 | int err; | |
2047 | nd.root.dentry = dentry; | |
2048 | nd.root.mnt = mnt; | |
2049 | BUG_ON(flags & LOOKUP_PARENT); | |
5b6ca027 | 2050 | /* the first argument of do_path_lookup() is ignored with LOOKUP_ROOT */ |
e0a01249 AV |
2051 | err = do_path_lookup(AT_FDCWD, name, flags | LOOKUP_ROOT, &nd); |
2052 | if (!err) | |
2053 | *path = nd.path; | |
2054 | return err; | |
16f18200 JJS |
2055 | } |
2056 | ||
057f6c01 JM |
2057 | /* |
2058 | * Restricted form of lookup. Doesn't follow links, single-component only, | |
2059 | * needs parent already locked. Doesn't follow mounts. | |
2060 | * SMP-safe. | |
2061 | */ | |
eead1911 | 2062 | static struct dentry *lookup_hash(struct nameidata *nd) |
057f6c01 | 2063 | { |
72bd866a | 2064 | return __lookup_hash(&nd->last, nd->path.dentry, nd->flags); |
1da177e4 LT |
2065 | } |
2066 | ||
eead1911 | 2067 | /** |
a6b91919 | 2068 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
2069 | * @name: pathname component to lookup |
2070 | * @base: base directory to lookup from | |
2071 | * @len: maximum length @len should be interpreted to | |
2072 | * | |
a6b91919 RD |
2073 | * Note that this routine is purely a helper for filesystem usage and should |
2074 | * not be called by generic code. Also note that by using this function the | |
eead1911 CH |
2075 | * nameidata argument is passed to the filesystem methods and a filesystem |
2076 | * using this helper needs to be prepared for that. | |
2077 | */ | |
057f6c01 JM |
2078 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
2079 | { | |
057f6c01 | 2080 | struct qstr this; |
6a96ba54 | 2081 | unsigned int c; |
cda309de | 2082 | int err; |
057f6c01 | 2083 | |
2f9092e1 DW |
2084 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
2085 | ||
6a96ba54 AV |
2086 | this.name = name; |
2087 | this.len = len; | |
0145acc2 | 2088 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
2089 | if (!len) |
2090 | return ERR_PTR(-EACCES); | |
2091 | ||
6a96ba54 AV |
2092 | while (len--) { |
2093 | c = *(const unsigned char *)name++; | |
2094 | if (c == '/' || c == '\0') | |
2095 | return ERR_PTR(-EACCES); | |
6a96ba54 | 2096 | } |
5a202bcd AV |
2097 | /* |
2098 | * See if the low-level filesystem might want | |
2099 | * to use its own hash.. | |
2100 | */ | |
2101 | if (base->d_flags & DCACHE_OP_HASH) { | |
2102 | int err = base->d_op->d_hash(base, base->d_inode, &this); | |
2103 | if (err < 0) | |
2104 | return ERR_PTR(err); | |
2105 | } | |
eead1911 | 2106 | |
cda309de MS |
2107 | err = inode_permission(base->d_inode, MAY_EXEC); |
2108 | if (err) | |
2109 | return ERR_PTR(err); | |
2110 | ||
72bd866a | 2111 | return __lookup_hash(&this, base, 0); |
057f6c01 JM |
2112 | } |
2113 | ||
1fa1e7f6 AW |
2114 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2115 | struct path *path, int *empty) | |
1da177e4 | 2116 | { |
2d8f3038 | 2117 | struct nameidata nd; |
91a27b2a | 2118 | struct filename *tmp = getname_flags(name, flags, empty); |
1da177e4 | 2119 | int err = PTR_ERR(tmp); |
1da177e4 | 2120 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
2121 | |
2122 | BUG_ON(flags & LOOKUP_PARENT); | |
2123 | ||
91a27b2a | 2124 | err = do_path_lookup(dfd, tmp->name, flags, &nd); |
1da177e4 | 2125 | putname(tmp); |
2d8f3038 AV |
2126 | if (!err) |
2127 | *path = nd.path; | |
1da177e4 LT |
2128 | } |
2129 | return err; | |
2130 | } | |
2131 | ||
1fa1e7f6 AW |
2132 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
2133 | struct path *path) | |
2134 | { | |
f7493e5d | 2135 | return user_path_at_empty(dfd, name, flags, path, NULL); |
1fa1e7f6 AW |
2136 | } |
2137 | ||
91a27b2a JL |
2138 | static struct filename * |
2139 | user_path_parent(int dfd, const char __user *path, struct nameidata *nd) | |
2ad94ae6 | 2140 | { |
91a27b2a | 2141 | struct filename *s = getname(path); |
2ad94ae6 AV |
2142 | int error; |
2143 | ||
2144 | if (IS_ERR(s)) | |
91a27b2a | 2145 | return s; |
2ad94ae6 | 2146 | |
91a27b2a JL |
2147 | error = do_path_lookup(dfd, s->name, LOOKUP_PARENT, nd); |
2148 | if (error) { | |
2ad94ae6 | 2149 | putname(s); |
91a27b2a JL |
2150 | return ERR_PTR(error); |
2151 | } | |
2ad94ae6 | 2152 | |
91a27b2a | 2153 | return s; |
2ad94ae6 AV |
2154 | } |
2155 | ||
1da177e4 LT |
2156 | /* |
2157 | * It's inline, so penalty for filesystems that don't use sticky bit is | |
2158 | * minimal. | |
2159 | */ | |
2160 | static inline int check_sticky(struct inode *dir, struct inode *inode) | |
2161 | { | |
8e96e3b7 | 2162 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2163 | |
1da177e4 LT |
2164 | if (!(dir->i_mode & S_ISVTX)) |
2165 | return 0; | |
8e96e3b7 | 2166 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2167 | return 0; |
8e96e3b7 | 2168 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2169 | return 0; |
1a48e2ac | 2170 | return !inode_capable(inode, CAP_FOWNER); |
1da177e4 LT |
2171 | } |
2172 | ||
2173 | /* | |
2174 | * Check whether we can remove a link victim from directory dir, check | |
2175 | * whether the type of victim is right. | |
2176 | * 1. We can't do it if dir is read-only (done in permission()) | |
2177 | * 2. We should have write and exec permissions on dir | |
2178 | * 3. We can't remove anything from append-only dir | |
2179 | * 4. We can't do anything with immutable dir (done in permission()) | |
2180 | * 5. If the sticky bit on dir is set we should either | |
2181 | * a. be owner of dir, or | |
2182 | * b. be owner of victim, or | |
2183 | * c. have CAP_FOWNER capability | |
2184 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2185 | * links pointing to it. | |
2186 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2187 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2188 | * 9. We can't remove a root or mountpoint. | |
2189 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
2190 | * nfs_async_unlink(). | |
2191 | */ | |
858119e1 | 2192 | static int may_delete(struct inode *dir,struct dentry *victim,int isdir) |
1da177e4 LT |
2193 | { |
2194 | int error; | |
2195 | ||
2196 | if (!victim->d_inode) | |
2197 | return -ENOENT; | |
2198 | ||
2199 | BUG_ON(victim->d_parent->d_inode != dir); | |
4fa6b5ec | 2200 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
1da177e4 | 2201 | |
f419a2e3 | 2202 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2203 | if (error) |
2204 | return error; | |
2205 | if (IS_APPEND(dir)) | |
2206 | return -EPERM; | |
2207 | if (check_sticky(dir, victim->d_inode)||IS_APPEND(victim->d_inode)|| | |
f9454548 | 2208 | IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode)) |
1da177e4 LT |
2209 | return -EPERM; |
2210 | if (isdir) { | |
2211 | if (!S_ISDIR(victim->d_inode->i_mode)) | |
2212 | return -ENOTDIR; | |
2213 | if (IS_ROOT(victim)) | |
2214 | return -EBUSY; | |
2215 | } else if (S_ISDIR(victim->d_inode->i_mode)) | |
2216 | return -EISDIR; | |
2217 | if (IS_DEADDIR(dir)) | |
2218 | return -ENOENT; | |
2219 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2220 | return -EBUSY; | |
2221 | return 0; | |
2222 | } | |
2223 | ||
2224 | /* Check whether we can create an object with dentry child in directory | |
2225 | * dir. | |
2226 | * 1. We can't do it if child already exists (open has special treatment for | |
2227 | * this case, but since we are inlined it's OK) | |
2228 | * 2. We can't do it if dir is read-only (done in permission()) | |
2229 | * 3. We should have write and exec permissions on dir | |
2230 | * 4. We can't do it if dir is immutable (done in permission()) | |
2231 | */ | |
a95164d9 | 2232 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 LT |
2233 | { |
2234 | if (child->d_inode) | |
2235 | return -EEXIST; | |
2236 | if (IS_DEADDIR(dir)) | |
2237 | return -ENOENT; | |
f419a2e3 | 2238 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2239 | } |
2240 | ||
1da177e4 LT |
2241 | /* |
2242 | * p1 and p2 should be directories on the same fs. | |
2243 | */ | |
2244 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2245 | { | |
2246 | struct dentry *p; | |
2247 | ||
2248 | if (p1 == p2) { | |
f2eace23 | 2249 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2250 | return NULL; |
2251 | } | |
2252 | ||
a11f3a05 | 2253 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2254 | |
e2761a11 OH |
2255 | p = d_ancestor(p2, p1); |
2256 | if (p) { | |
2257 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2258 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2259 | return p; | |
1da177e4 LT |
2260 | } |
2261 | ||
e2761a11 OH |
2262 | p = d_ancestor(p1, p2); |
2263 | if (p) { | |
2264 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2265 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2266 | return p; | |
1da177e4 LT |
2267 | } |
2268 | ||
f2eace23 IM |
2269 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
2270 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
1da177e4 LT |
2271 | return NULL; |
2272 | } | |
2273 | ||
2274 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2275 | { | |
1b1dcc1b | 2276 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2277 | if (p1 != p2) { |
1b1dcc1b | 2278 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2279 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2280 | } |
2281 | } | |
2282 | ||
4acdaf27 | 2283 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2284 | bool want_excl) |
1da177e4 | 2285 | { |
a95164d9 | 2286 | int error = may_create(dir, dentry); |
1da177e4 LT |
2287 | if (error) |
2288 | return error; | |
2289 | ||
acfa4380 | 2290 | if (!dir->i_op->create) |
1da177e4 LT |
2291 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2292 | mode &= S_IALLUGO; | |
2293 | mode |= S_IFREG; | |
2294 | error = security_inode_create(dir, dentry, mode); | |
2295 | if (error) | |
2296 | return error; | |
312b63fb | 2297 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2298 | if (!error) |
f38aa942 | 2299 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2300 | return error; |
2301 | } | |
2302 | ||
73d049a4 | 2303 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2304 | { |
3fb64190 | 2305 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2306 | struct inode *inode = dentry->d_inode; |
2307 | int error; | |
2308 | ||
bcda7652 AV |
2309 | /* O_PATH? */ |
2310 | if (!acc_mode) | |
2311 | return 0; | |
2312 | ||
1da177e4 LT |
2313 | if (!inode) |
2314 | return -ENOENT; | |
2315 | ||
c8fe8f30 CH |
2316 | switch (inode->i_mode & S_IFMT) { |
2317 | case S_IFLNK: | |
1da177e4 | 2318 | return -ELOOP; |
c8fe8f30 CH |
2319 | case S_IFDIR: |
2320 | if (acc_mode & MAY_WRITE) | |
2321 | return -EISDIR; | |
2322 | break; | |
2323 | case S_IFBLK: | |
2324 | case S_IFCHR: | |
3fb64190 | 2325 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2326 | return -EACCES; |
c8fe8f30 CH |
2327 | /*FALLTHRU*/ |
2328 | case S_IFIFO: | |
2329 | case S_IFSOCK: | |
1da177e4 | 2330 | flag &= ~O_TRUNC; |
c8fe8f30 | 2331 | break; |
4a3fd211 | 2332 | } |
b41572e9 | 2333 | |
3fb64190 | 2334 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2335 | if (error) |
2336 | return error; | |
6146f0d5 | 2337 | |
1da177e4 LT |
2338 | /* |
2339 | * An append-only file must be opened in append mode for writing. | |
2340 | */ | |
2341 | if (IS_APPEND(inode)) { | |
8737c930 | 2342 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2343 | return -EPERM; |
1da177e4 | 2344 | if (flag & O_TRUNC) |
7715b521 | 2345 | return -EPERM; |
1da177e4 LT |
2346 | } |
2347 | ||
2348 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2349 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2350 | return -EPERM; |
1da177e4 | 2351 | |
f3c7691e | 2352 | return 0; |
7715b521 | 2353 | } |
1da177e4 | 2354 | |
e1181ee6 | 2355 | static int handle_truncate(struct file *filp) |
7715b521 | 2356 | { |
e1181ee6 | 2357 | struct path *path = &filp->f_path; |
7715b521 AV |
2358 | struct inode *inode = path->dentry->d_inode; |
2359 | int error = get_write_access(inode); | |
2360 | if (error) | |
2361 | return error; | |
2362 | /* | |
2363 | * Refuse to truncate files with mandatory locks held on them. | |
2364 | */ | |
2365 | error = locks_verify_locked(inode); | |
2366 | if (!error) | |
ea0d3ab2 | 2367 | error = security_path_truncate(path); |
7715b521 AV |
2368 | if (!error) { |
2369 | error = do_truncate(path->dentry, 0, | |
2370 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2371 | filp); |
7715b521 AV |
2372 | } |
2373 | put_write_access(inode); | |
acd0c935 | 2374 | return error; |
1da177e4 LT |
2375 | } |
2376 | ||
d57999e1 DH |
2377 | static inline int open_to_namei_flags(int flag) |
2378 | { | |
8a5e929d AV |
2379 | if ((flag & O_ACCMODE) == 3) |
2380 | flag--; | |
d57999e1 DH |
2381 | return flag; |
2382 | } | |
2383 | ||
d18e9008 MS |
2384 | static int may_o_create(struct path *dir, struct dentry *dentry, umode_t mode) |
2385 | { | |
2386 | int error = security_path_mknod(dir, dentry, mode, 0); | |
2387 | if (error) | |
2388 | return error; | |
2389 | ||
2390 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); | |
2391 | if (error) | |
2392 | return error; | |
2393 | ||
2394 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2395 | } | |
2396 | ||
1acf0af9 DH |
2397 | /* |
2398 | * Attempt to atomically look up, create and open a file from a negative | |
2399 | * dentry. | |
2400 | * | |
2401 | * Returns 0 if successful. The file will have been created and attached to | |
2402 | * @file by the filesystem calling finish_open(). | |
2403 | * | |
2404 | * Returns 1 if the file was looked up only or didn't need creating. The | |
2405 | * caller will need to perform the open themselves. @path will have been | |
2406 | * updated to point to the new dentry. This may be negative. | |
2407 | * | |
2408 | * Returns an error code otherwise. | |
2409 | */ | |
2675a4eb AV |
2410 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
2411 | struct path *path, struct file *file, | |
2412 | const struct open_flags *op, | |
64894cf8 | 2413 | bool got_write, bool need_lookup, |
2675a4eb | 2414 | int *opened) |
d18e9008 MS |
2415 | { |
2416 | struct inode *dir = nd->path.dentry->d_inode; | |
2417 | unsigned open_flag = open_to_namei_flags(op->open_flag); | |
2418 | umode_t mode; | |
2419 | int error; | |
2420 | int acc_mode; | |
d18e9008 MS |
2421 | int create_error = 0; |
2422 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; | |
2423 | ||
2424 | BUG_ON(dentry->d_inode); | |
2425 | ||
2426 | /* Don't create child dentry for a dead directory. */ | |
2427 | if (unlikely(IS_DEADDIR(dir))) { | |
2675a4eb | 2428 | error = -ENOENT; |
d18e9008 MS |
2429 | goto out; |
2430 | } | |
2431 | ||
62b259d8 | 2432 | mode = op->mode; |
d18e9008 MS |
2433 | if ((open_flag & O_CREAT) && !IS_POSIXACL(dir)) |
2434 | mode &= ~current_umask(); | |
2435 | ||
f8310c59 | 2436 | if ((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT)) { |
d18e9008 | 2437 | open_flag &= ~O_TRUNC; |
47237687 | 2438 | *opened |= FILE_CREATED; |
d18e9008 MS |
2439 | } |
2440 | ||
2441 | /* | |
2442 | * Checking write permission is tricky, bacuse we don't know if we are | |
2443 | * going to actually need it: O_CREAT opens should work as long as the | |
2444 | * file exists. But checking existence breaks atomicity. The trick is | |
2445 | * to check access and if not granted clear O_CREAT from the flags. | |
2446 | * | |
2447 | * Another problem is returing the "right" error value (e.g. for an | |
2448 | * O_EXCL open we want to return EEXIST not EROFS). | |
2449 | */ | |
64894cf8 AV |
2450 | if (((open_flag & (O_CREAT | O_TRUNC)) || |
2451 | (open_flag & O_ACCMODE) != O_RDONLY) && unlikely(!got_write)) { | |
2452 | if (!(open_flag & O_CREAT)) { | |
d18e9008 MS |
2453 | /* |
2454 | * No O_CREATE -> atomicity not a requirement -> fall | |
2455 | * back to lookup + open | |
2456 | */ | |
2457 | goto no_open; | |
2458 | } else if (open_flag & (O_EXCL | O_TRUNC)) { | |
2459 | /* Fall back and fail with the right error */ | |
64894cf8 | 2460 | create_error = -EROFS; |
d18e9008 MS |
2461 | goto no_open; |
2462 | } else { | |
2463 | /* No side effects, safe to clear O_CREAT */ | |
64894cf8 | 2464 | create_error = -EROFS; |
d18e9008 MS |
2465 | open_flag &= ~O_CREAT; |
2466 | } | |
2467 | } | |
2468 | ||
2469 | if (open_flag & O_CREAT) { | |
38227f78 | 2470 | error = may_o_create(&nd->path, dentry, mode); |
d18e9008 MS |
2471 | if (error) { |
2472 | create_error = error; | |
2473 | if (open_flag & O_EXCL) | |
2474 | goto no_open; | |
2475 | open_flag &= ~O_CREAT; | |
2476 | } | |
2477 | } | |
2478 | ||
2479 | if (nd->flags & LOOKUP_DIRECTORY) | |
2480 | open_flag |= O_DIRECTORY; | |
2481 | ||
30d90494 AV |
2482 | file->f_path.dentry = DENTRY_NOT_SET; |
2483 | file->f_path.mnt = nd->path.mnt; | |
2484 | error = dir->i_op->atomic_open(dir, dentry, file, open_flag, mode, | |
47237687 | 2485 | opened); |
d9585277 | 2486 | if (error < 0) { |
d9585277 AV |
2487 | if (create_error && error == -ENOENT) |
2488 | error = create_error; | |
d18e9008 MS |
2489 | goto out; |
2490 | } | |
2491 | ||
2492 | acc_mode = op->acc_mode; | |
47237687 | 2493 | if (*opened & FILE_CREATED) { |
d18e9008 MS |
2494 | fsnotify_create(dir, dentry); |
2495 | acc_mode = MAY_OPEN; | |
2496 | } | |
2497 | ||
d9585277 | 2498 | if (error) { /* returned 1, that is */ |
30d90494 | 2499 | if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 2500 | error = -EIO; |
d18e9008 MS |
2501 | goto out; |
2502 | } | |
30d90494 | 2503 | if (file->f_path.dentry) { |
d18e9008 | 2504 | dput(dentry); |
30d90494 | 2505 | dentry = file->f_path.dentry; |
d18e9008 | 2506 | } |
62b2ce96 SW |
2507 | if (create_error && dentry->d_inode == NULL) { |
2508 | error = create_error; | |
2509 | goto out; | |
2510 | } | |
d18e9008 MS |
2511 | goto looked_up; |
2512 | } | |
2513 | ||
2514 | /* | |
2515 | * We didn't have the inode before the open, so check open permission | |
2516 | * here. | |
2517 | */ | |
2675a4eb AV |
2518 | error = may_open(&file->f_path, acc_mode, open_flag); |
2519 | if (error) | |
2520 | fput(file); | |
d18e9008 MS |
2521 | |
2522 | out: | |
2523 | dput(dentry); | |
2675a4eb | 2524 | return error; |
d18e9008 | 2525 | |
d18e9008 MS |
2526 | no_open: |
2527 | if (need_lookup) { | |
72bd866a | 2528 | dentry = lookup_real(dir, dentry, nd->flags); |
d18e9008 | 2529 | if (IS_ERR(dentry)) |
2675a4eb | 2530 | return PTR_ERR(dentry); |
d18e9008 MS |
2531 | |
2532 | if (create_error) { | |
2533 | int open_flag = op->open_flag; | |
2534 | ||
2675a4eb | 2535 | error = create_error; |
d18e9008 MS |
2536 | if ((open_flag & O_EXCL)) { |
2537 | if (!dentry->d_inode) | |
2538 | goto out; | |
2539 | } else if (!dentry->d_inode) { | |
2540 | goto out; | |
2541 | } else if ((open_flag & O_TRUNC) && | |
2542 | S_ISREG(dentry->d_inode->i_mode)) { | |
2543 | goto out; | |
2544 | } | |
2545 | /* will fail later, go on to get the right error */ | |
2546 | } | |
2547 | } | |
2548 | looked_up: | |
2549 | path->dentry = dentry; | |
2550 | path->mnt = nd->path.mnt; | |
2675a4eb | 2551 | return 1; |
d18e9008 MS |
2552 | } |
2553 | ||
d58ffd35 | 2554 | /* |
1acf0af9 | 2555 | * Look up and maybe create and open the last component. |
d58ffd35 MS |
2556 | * |
2557 | * Must be called with i_mutex held on parent. | |
2558 | * | |
1acf0af9 DH |
2559 | * Returns 0 if the file was successfully atomically created (if necessary) and |
2560 | * opened. In this case the file will be returned attached to @file. | |
2561 | * | |
2562 | * Returns 1 if the file was not completely opened at this time, though lookups | |
2563 | * and creations will have been performed and the dentry returned in @path will | |
2564 | * be positive upon return if O_CREAT was specified. If O_CREAT wasn't | |
2565 | * specified then a negative dentry may be returned. | |
2566 | * | |
2567 | * An error code is returned otherwise. | |
2568 | * | |
2569 | * FILE_CREATE will be set in @*opened if the dentry was created and will be | |
2570 | * cleared otherwise prior to returning. | |
d58ffd35 | 2571 | */ |
2675a4eb AV |
2572 | static int lookup_open(struct nameidata *nd, struct path *path, |
2573 | struct file *file, | |
2574 | const struct open_flags *op, | |
64894cf8 | 2575 | bool got_write, int *opened) |
d58ffd35 MS |
2576 | { |
2577 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 2578 | struct inode *dir_inode = dir->d_inode; |
d58ffd35 MS |
2579 | struct dentry *dentry; |
2580 | int error; | |
54ef4872 | 2581 | bool need_lookup; |
d58ffd35 | 2582 | |
47237687 | 2583 | *opened &= ~FILE_CREATED; |
201f956e | 2584 | dentry = lookup_dcache(&nd->last, dir, nd->flags, &need_lookup); |
d58ffd35 | 2585 | if (IS_ERR(dentry)) |
2675a4eb | 2586 | return PTR_ERR(dentry); |
d58ffd35 | 2587 | |
d18e9008 MS |
2588 | /* Cached positive dentry: will open in f_op->open */ |
2589 | if (!need_lookup && dentry->d_inode) | |
2590 | goto out_no_open; | |
2591 | ||
2592 | if ((nd->flags & LOOKUP_OPEN) && dir_inode->i_op->atomic_open) { | |
64894cf8 | 2593 | return atomic_open(nd, dentry, path, file, op, got_write, |
47237687 | 2594 | need_lookup, opened); |
d18e9008 MS |
2595 | } |
2596 | ||
54ef4872 MS |
2597 | if (need_lookup) { |
2598 | BUG_ON(dentry->d_inode); | |
2599 | ||
72bd866a | 2600 | dentry = lookup_real(dir_inode, dentry, nd->flags); |
54ef4872 | 2601 | if (IS_ERR(dentry)) |
2675a4eb | 2602 | return PTR_ERR(dentry); |
54ef4872 MS |
2603 | } |
2604 | ||
d58ffd35 MS |
2605 | /* Negative dentry, just create the file */ |
2606 | if (!dentry->d_inode && (op->open_flag & O_CREAT)) { | |
2607 | umode_t mode = op->mode; | |
2608 | if (!IS_POSIXACL(dir->d_inode)) | |
2609 | mode &= ~current_umask(); | |
2610 | /* | |
2611 | * This write is needed to ensure that a | |
2612 | * rw->ro transition does not occur between | |
2613 | * the time when the file is created and when | |
2614 | * a permanent write count is taken through | |
015c3bbc | 2615 | * the 'struct file' in finish_open(). |
d58ffd35 | 2616 | */ |
64894cf8 AV |
2617 | if (!got_write) { |
2618 | error = -EROFS; | |
d58ffd35 | 2619 | goto out_dput; |
64894cf8 | 2620 | } |
47237687 | 2621 | *opened |= FILE_CREATED; |
d58ffd35 MS |
2622 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2623 | if (error) | |
2624 | goto out_dput; | |
312b63fb AV |
2625 | error = vfs_create(dir->d_inode, dentry, mode, |
2626 | nd->flags & LOOKUP_EXCL); | |
d58ffd35 MS |
2627 | if (error) |
2628 | goto out_dput; | |
2629 | } | |
d18e9008 | 2630 | out_no_open: |
d58ffd35 MS |
2631 | path->dentry = dentry; |
2632 | path->mnt = nd->path.mnt; | |
2675a4eb | 2633 | return 1; |
d58ffd35 MS |
2634 | |
2635 | out_dput: | |
2636 | dput(dentry); | |
2675a4eb | 2637 | return error; |
d58ffd35 MS |
2638 | } |
2639 | ||
31e6b01f | 2640 | /* |
fe2d35ff | 2641 | * Handle the last step of open() |
31e6b01f | 2642 | */ |
2675a4eb AV |
2643 | static int do_last(struct nameidata *nd, struct path *path, |
2644 | struct file *file, const struct open_flags *op, | |
2645 | int *opened, const char *pathname) | |
fb1cc555 | 2646 | { |
a1e28038 | 2647 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 2648 | int open_flag = op->open_flag; |
77d660a8 | 2649 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 2650 | bool got_write = false; |
bcda7652 | 2651 | int acc_mode = op->acc_mode; |
a1eb3315 | 2652 | struct inode *inode; |
77d660a8 | 2653 | bool symlink_ok = false; |
16b1c1cd MS |
2654 | struct path save_parent = { .dentry = NULL, .mnt = NULL }; |
2655 | bool retried = false; | |
16c2cd71 | 2656 | int error; |
1f36f774 | 2657 | |
c3e380b0 AV |
2658 | nd->flags &= ~LOOKUP_PARENT; |
2659 | nd->flags |= op->intent; | |
2660 | ||
1f36f774 AV |
2661 | switch (nd->last_type) { |
2662 | case LAST_DOTDOT: | |
176306f5 | 2663 | case LAST_DOT: |
fe2d35ff AV |
2664 | error = handle_dots(nd, nd->last_type); |
2665 | if (error) | |
2675a4eb | 2666 | return error; |
1f36f774 | 2667 | /* fallthrough */ |
1f36f774 | 2668 | case LAST_ROOT: |
9f1fafee | 2669 | error = complete_walk(nd); |
16c2cd71 | 2670 | if (error) |
2675a4eb | 2671 | return error; |
bfcec708 | 2672 | audit_inode(pathname, nd->path.dentry, 0); |
ca344a89 | 2673 | if (open_flag & O_CREAT) { |
fe2d35ff | 2674 | error = -EISDIR; |
2675a4eb | 2675 | goto out; |
fe2d35ff | 2676 | } |
e83db167 | 2677 | goto finish_open; |
1f36f774 | 2678 | case LAST_BIND: |
9f1fafee | 2679 | error = complete_walk(nd); |
16c2cd71 | 2680 | if (error) |
2675a4eb | 2681 | return error; |
bfcec708 | 2682 | audit_inode(pathname, dir, 0); |
e83db167 | 2683 | goto finish_open; |
1f36f774 | 2684 | } |
67ee3ad2 | 2685 | |
ca344a89 | 2686 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
2687 | if (nd->last.name[nd->last.len]) |
2688 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
bcda7652 | 2689 | if (open_flag & O_PATH && !(nd->flags & LOOKUP_FOLLOW)) |
77d660a8 | 2690 | symlink_ok = true; |
fe2d35ff | 2691 | /* we _can_ be in RCU mode here */ |
a1eb3315 | 2692 | error = lookup_fast(nd, &nd->last, path, &inode); |
71574865 MS |
2693 | if (likely(!error)) |
2694 | goto finish_lookup; | |
2695 | ||
2696 | if (error < 0) | |
2675a4eb | 2697 | goto out; |
71574865 MS |
2698 | |
2699 | BUG_ON(nd->inode != dir->d_inode); | |
b6183df7 MS |
2700 | } else { |
2701 | /* create side of things */ | |
2702 | /* | |
2703 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
2704 | * has been cleared when we got to the last component we are | |
2705 | * about to look up | |
2706 | */ | |
2707 | error = complete_walk(nd); | |
2708 | if (error) | |
2675a4eb | 2709 | return error; |
fe2d35ff | 2710 | |
bfcec708 | 2711 | audit_inode(pathname, dir, 0); |
b6183df7 MS |
2712 | error = -EISDIR; |
2713 | /* trailing slashes? */ | |
2714 | if (nd->last.name[nd->last.len]) | |
2675a4eb | 2715 | goto out; |
b6183df7 | 2716 | } |
a2c36b45 | 2717 | |
16b1c1cd | 2718 | retry_lookup: |
64894cf8 AV |
2719 | if (op->open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
2720 | error = mnt_want_write(nd->path.mnt); | |
2721 | if (!error) | |
2722 | got_write = true; | |
2723 | /* | |
2724 | * do _not_ fail yet - we might not need that or fail with | |
2725 | * a different error; let lookup_open() decide; we'll be | |
2726 | * dropping this one anyway. | |
2727 | */ | |
2728 | } | |
a1e28038 | 2729 | mutex_lock(&dir->d_inode->i_mutex); |
64894cf8 | 2730 | error = lookup_open(nd, path, file, op, got_write, opened); |
d58ffd35 | 2731 | mutex_unlock(&dir->d_inode->i_mutex); |
a1e28038 | 2732 | |
2675a4eb AV |
2733 | if (error <= 0) { |
2734 | if (error) | |
d18e9008 MS |
2735 | goto out; |
2736 | ||
47237687 | 2737 | if ((*opened & FILE_CREATED) || |
2675a4eb | 2738 | !S_ISREG(file->f_path.dentry->d_inode->i_mode)) |
77d660a8 | 2739 | will_truncate = false; |
d18e9008 | 2740 | |
bfcec708 | 2741 | audit_inode(pathname, file->f_path.dentry, 0); |
d18e9008 MS |
2742 | goto opened; |
2743 | } | |
fb1cc555 | 2744 | |
47237687 | 2745 | if (*opened & FILE_CREATED) { |
9b44f1b3 | 2746 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 2747 | open_flag &= ~O_TRUNC; |
77d660a8 | 2748 | will_truncate = false; |
bcda7652 | 2749 | acc_mode = MAY_OPEN; |
d58ffd35 | 2750 | path_to_nameidata(path, nd); |
e83db167 | 2751 | goto finish_open_created; |
fb1cc555 AV |
2752 | } |
2753 | ||
2754 | /* | |
3134f37e | 2755 | * create/update audit record if it already exists. |
fb1cc555 | 2756 | */ |
3134f37e | 2757 | if (path->dentry->d_inode) |
bfcec708 | 2758 | audit_inode(pathname, path->dentry, 0); |
fb1cc555 | 2759 | |
d18e9008 MS |
2760 | /* |
2761 | * If atomic_open() acquired write access it is dropped now due to | |
2762 | * possible mount and symlink following (this might be optimized away if | |
2763 | * necessary...) | |
2764 | */ | |
64894cf8 | 2765 | if (got_write) { |
d18e9008 | 2766 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 2767 | got_write = false; |
d18e9008 MS |
2768 | } |
2769 | ||
fb1cc555 | 2770 | error = -EEXIST; |
f8310c59 | 2771 | if ((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT)) |
fb1cc555 AV |
2772 | goto exit_dput; |
2773 | ||
9875cf80 DH |
2774 | error = follow_managed(path, nd->flags); |
2775 | if (error < 0) | |
2776 | goto exit_dput; | |
fb1cc555 | 2777 | |
a3fbbde7 AV |
2778 | if (error) |
2779 | nd->flags |= LOOKUP_JUMPED; | |
2780 | ||
decf3400 MS |
2781 | BUG_ON(nd->flags & LOOKUP_RCU); |
2782 | inode = path->dentry->d_inode; | |
5f5daac1 MS |
2783 | finish_lookup: |
2784 | /* we _can_ be in RCU mode here */ | |
fb1cc555 | 2785 | error = -ENOENT; |
54c33e7f MS |
2786 | if (!inode) { |
2787 | path_to_nameidata(path, nd); | |
2675a4eb | 2788 | goto out; |
54c33e7f | 2789 | } |
9e67f361 | 2790 | |
d45ea867 MS |
2791 | if (should_follow_link(inode, !symlink_ok)) { |
2792 | if (nd->flags & LOOKUP_RCU) { | |
2793 | if (unlikely(unlazy_walk(nd, path->dentry))) { | |
2794 | error = -ECHILD; | |
2675a4eb | 2795 | goto out; |
d45ea867 MS |
2796 | } |
2797 | } | |
2798 | BUG_ON(inode != path->dentry->d_inode); | |
2675a4eb | 2799 | return 1; |
d45ea867 | 2800 | } |
fb1cc555 | 2801 | |
16b1c1cd MS |
2802 | if ((nd->flags & LOOKUP_RCU) || nd->path.mnt != path->mnt) { |
2803 | path_to_nameidata(path, nd); | |
2804 | } else { | |
2805 | save_parent.dentry = nd->path.dentry; | |
2806 | save_parent.mnt = mntget(path->mnt); | |
2807 | nd->path.dentry = path->dentry; | |
2808 | ||
2809 | } | |
decf3400 | 2810 | nd->inode = inode; |
a3fbbde7 AV |
2811 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
2812 | error = complete_walk(nd); | |
16b1c1cd MS |
2813 | if (error) { |
2814 | path_put(&save_parent); | |
2675a4eb | 2815 | return error; |
16b1c1cd | 2816 | } |
fb1cc555 | 2817 | error = -EISDIR; |
050ac841 | 2818 | if ((open_flag & O_CREAT) && S_ISDIR(nd->inode->i_mode)) |
2675a4eb | 2819 | goto out; |
af2f5542 MS |
2820 | error = -ENOTDIR; |
2821 | if ((nd->flags & LOOKUP_DIRECTORY) && !nd->inode->i_op->lookup) | |
2675a4eb | 2822 | goto out; |
bfcec708 | 2823 | audit_inode(pathname, nd->path.dentry, 0); |
e83db167 | 2824 | finish_open: |
6c0d46c4 | 2825 | if (!S_ISREG(nd->inode->i_mode)) |
77d660a8 | 2826 | will_truncate = false; |
6c0d46c4 | 2827 | |
0f9d1a10 AV |
2828 | if (will_truncate) { |
2829 | error = mnt_want_write(nd->path.mnt); | |
2830 | if (error) | |
2675a4eb | 2831 | goto out; |
64894cf8 | 2832 | got_write = true; |
0f9d1a10 | 2833 | } |
e83db167 | 2834 | finish_open_created: |
bcda7652 | 2835 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 2836 | if (error) |
2675a4eb | 2837 | goto out; |
30d90494 AV |
2838 | file->f_path.mnt = nd->path.mnt; |
2839 | error = finish_open(file, nd->path.dentry, NULL, opened); | |
2840 | if (error) { | |
30d90494 | 2841 | if (error == -EOPENSTALE) |
f60dc3db | 2842 | goto stale_open; |
015c3bbc | 2843 | goto out; |
f60dc3db | 2844 | } |
a8277b9b | 2845 | opened: |
2675a4eb | 2846 | error = open_check_o_direct(file); |
015c3bbc MS |
2847 | if (error) |
2848 | goto exit_fput; | |
2675a4eb | 2849 | error = ima_file_check(file, op->acc_mode); |
aa4caadb MS |
2850 | if (error) |
2851 | goto exit_fput; | |
2852 | ||
2853 | if (will_truncate) { | |
2675a4eb | 2854 | error = handle_truncate(file); |
aa4caadb MS |
2855 | if (error) |
2856 | goto exit_fput; | |
0f9d1a10 | 2857 | } |
ca344a89 | 2858 | out: |
64894cf8 | 2859 | if (got_write) |
0f9d1a10 | 2860 | mnt_drop_write(nd->path.mnt); |
16b1c1cd | 2861 | path_put(&save_parent); |
e276ae67 | 2862 | terminate_walk(nd); |
2675a4eb | 2863 | return error; |
fb1cc555 | 2864 | |
fb1cc555 AV |
2865 | exit_dput: |
2866 | path_put_conditional(path, nd); | |
ca344a89 | 2867 | goto out; |
015c3bbc | 2868 | exit_fput: |
2675a4eb AV |
2869 | fput(file); |
2870 | goto out; | |
015c3bbc | 2871 | |
f60dc3db MS |
2872 | stale_open: |
2873 | /* If no saved parent or already retried then can't retry */ | |
2874 | if (!save_parent.dentry || retried) | |
2875 | goto out; | |
2876 | ||
2877 | BUG_ON(save_parent.dentry != dir); | |
2878 | path_put(&nd->path); | |
2879 | nd->path = save_parent; | |
2880 | nd->inode = dir->d_inode; | |
2881 | save_parent.mnt = NULL; | |
2882 | save_parent.dentry = NULL; | |
64894cf8 | 2883 | if (got_write) { |
f60dc3db | 2884 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 2885 | got_write = false; |
f60dc3db MS |
2886 | } |
2887 | retried = true; | |
2888 | goto retry_lookup; | |
fb1cc555 AV |
2889 | } |
2890 | ||
13aab428 | 2891 | static struct file *path_openat(int dfd, const char *pathname, |
73d049a4 | 2892 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 2893 | { |
fe2d35ff | 2894 | struct file *base = NULL; |
30d90494 | 2895 | struct file *file; |
9850c056 | 2896 | struct path path; |
47237687 | 2897 | int opened = 0; |
13aab428 | 2898 | int error; |
31e6b01f | 2899 | |
30d90494 AV |
2900 | file = get_empty_filp(); |
2901 | if (!file) | |
31e6b01f NP |
2902 | return ERR_PTR(-ENFILE); |
2903 | ||
30d90494 | 2904 | file->f_flags = op->open_flag; |
31e6b01f | 2905 | |
73d049a4 | 2906 | error = path_init(dfd, pathname, flags | LOOKUP_PARENT, nd, &base); |
31e6b01f | 2907 | if (unlikely(error)) |
2675a4eb | 2908 | goto out; |
31e6b01f | 2909 | |
fe2d35ff | 2910 | current->total_link_count = 0; |
73d049a4 | 2911 | error = link_path_walk(pathname, nd); |
31e6b01f | 2912 | if (unlikely(error)) |
2675a4eb | 2913 | goto out; |
1da177e4 | 2914 | |
2675a4eb AV |
2915 | error = do_last(nd, &path, file, op, &opened, pathname); |
2916 | while (unlikely(error > 0)) { /* trailing symlink */ | |
7b9337aa | 2917 | struct path link = path; |
def4af30 | 2918 | void *cookie; |
574197e0 | 2919 | if (!(nd->flags & LOOKUP_FOLLOW)) { |
73d049a4 AV |
2920 | path_put_conditional(&path, nd); |
2921 | path_put(&nd->path); | |
2675a4eb | 2922 | error = -ELOOP; |
40b39136 AV |
2923 | break; |
2924 | } | |
800179c9 KC |
2925 | error = may_follow_link(&link, nd); |
2926 | if (unlikely(error)) | |
2927 | break; | |
73d049a4 AV |
2928 | nd->flags |= LOOKUP_PARENT; |
2929 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
574197e0 | 2930 | error = follow_link(&link, nd, &cookie); |
c3e380b0 | 2931 | if (unlikely(error)) |
2675a4eb AV |
2932 | break; |
2933 | error = do_last(nd, &path, file, op, &opened, pathname); | |
574197e0 | 2934 | put_link(nd, &link, cookie); |
806b681c | 2935 | } |
10fa8e62 | 2936 | out: |
73d049a4 AV |
2937 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) |
2938 | path_put(&nd->root); | |
fe2d35ff AV |
2939 | if (base) |
2940 | fput(base); | |
2675a4eb AV |
2941 | if (!(opened & FILE_OPENED)) { |
2942 | BUG_ON(!error); | |
30d90494 | 2943 | put_filp(file); |
16b1c1cd | 2944 | } |
2675a4eb AV |
2945 | if (unlikely(error)) { |
2946 | if (error == -EOPENSTALE) { | |
2947 | if (flags & LOOKUP_RCU) | |
2948 | error = -ECHILD; | |
2949 | else | |
2950 | error = -ESTALE; | |
2951 | } | |
2952 | file = ERR_PTR(error); | |
2953 | } | |
2954 | return file; | |
1da177e4 LT |
2955 | } |
2956 | ||
13aab428 AV |
2957 | struct file *do_filp_open(int dfd, const char *pathname, |
2958 | const struct open_flags *op, int flags) | |
2959 | { | |
73d049a4 | 2960 | struct nameidata nd; |
13aab428 AV |
2961 | struct file *filp; |
2962 | ||
73d049a4 | 2963 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 2964 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 2965 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 2966 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 2967 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
13aab428 AV |
2968 | return filp; |
2969 | } | |
2970 | ||
73d049a4 AV |
2971 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
2972 | const char *name, const struct open_flags *op, int flags) | |
2973 | { | |
2974 | struct nameidata nd; | |
2975 | struct file *file; | |
2976 | ||
2977 | nd.root.mnt = mnt; | |
2978 | nd.root.dentry = dentry; | |
2979 | ||
2980 | flags |= LOOKUP_ROOT; | |
2981 | ||
bcda7652 | 2982 | if (dentry->d_inode->i_op->follow_link && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
2983 | return ERR_PTR(-ELOOP); |
2984 | ||
2985 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_RCU); | |
2986 | if (unlikely(file == ERR_PTR(-ECHILD))) | |
2987 | file = path_openat(-1, name, &nd, op, flags); | |
2988 | if (unlikely(file == ERR_PTR(-ESTALE))) | |
2989 | file = path_openat(-1, name, &nd, op, flags | LOOKUP_REVAL); | |
2990 | return file; | |
2991 | } | |
2992 | ||
ed75e95d | 2993 | struct dentry *kern_path_create(int dfd, const char *pathname, struct path *path, int is_dir) |
1da177e4 | 2994 | { |
c663e5d8 | 2995 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d | 2996 | struct nameidata nd; |
c30dabfe | 2997 | int err2; |
ed75e95d AV |
2998 | int error = do_path_lookup(dfd, pathname, LOOKUP_PARENT, &nd); |
2999 | if (error) | |
3000 | return ERR_PTR(error); | |
1da177e4 | 3001 | |
c663e5d8 CH |
3002 | /* |
3003 | * Yucky last component or no last component at all? | |
3004 | * (foo/., foo/.., /////) | |
3005 | */ | |
ed75e95d AV |
3006 | if (nd.last_type != LAST_NORM) |
3007 | goto out; | |
3008 | nd.flags &= ~LOOKUP_PARENT; | |
3009 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
c663e5d8 | 3010 | |
c30dabfe JK |
3011 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
3012 | err2 = mnt_want_write(nd.path.mnt); | |
c663e5d8 CH |
3013 | /* |
3014 | * Do the final lookup. | |
3015 | */ | |
ed75e95d AV |
3016 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
3017 | dentry = lookup_hash(&nd); | |
1da177e4 | 3018 | if (IS_ERR(dentry)) |
a8104a9f | 3019 | goto unlock; |
c663e5d8 | 3020 | |
a8104a9f | 3021 | error = -EEXIST; |
e9baf6e5 | 3022 | if (dentry->d_inode) |
a8104a9f | 3023 | goto fail; |
c663e5d8 CH |
3024 | /* |
3025 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3026 | * From the vfs_mknod() POV we just have a negative dentry - | |
3027 | * all is fine. Let's be bastards - you had / on the end, you've | |
3028 | * been asking for (non-existent) directory. -ENOENT for you. | |
3029 | */ | |
ed75e95d | 3030 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
a8104a9f | 3031 | error = -ENOENT; |
ed75e95d | 3032 | goto fail; |
e9baf6e5 | 3033 | } |
c30dabfe JK |
3034 | if (unlikely(err2)) { |
3035 | error = err2; | |
a8104a9f | 3036 | goto fail; |
c30dabfe | 3037 | } |
ed75e95d | 3038 | *path = nd.path; |
1da177e4 | 3039 | return dentry; |
1da177e4 | 3040 | fail: |
a8104a9f AV |
3041 | dput(dentry); |
3042 | dentry = ERR_PTR(error); | |
3043 | unlock: | |
ed75e95d | 3044 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
c30dabfe JK |
3045 | if (!err2) |
3046 | mnt_drop_write(nd.path.mnt); | |
ed75e95d AV |
3047 | out: |
3048 | path_put(&nd.path); | |
1da177e4 LT |
3049 | return dentry; |
3050 | } | |
dae6ad8f AV |
3051 | EXPORT_SYMBOL(kern_path_create); |
3052 | ||
921a1650 AV |
3053 | void done_path_create(struct path *path, struct dentry *dentry) |
3054 | { | |
3055 | dput(dentry); | |
3056 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
a8104a9f | 3057 | mnt_drop_write(path->mnt); |
921a1650 AV |
3058 | path_put(path); |
3059 | } | |
3060 | EXPORT_SYMBOL(done_path_create); | |
3061 | ||
dae6ad8f AV |
3062 | struct dentry *user_path_create(int dfd, const char __user *pathname, struct path *path, int is_dir) |
3063 | { | |
91a27b2a | 3064 | struct filename *tmp = getname(pathname); |
dae6ad8f AV |
3065 | struct dentry *res; |
3066 | if (IS_ERR(tmp)) | |
3067 | return ERR_CAST(tmp); | |
91a27b2a | 3068 | res = kern_path_create(dfd, tmp->name, path, is_dir); |
dae6ad8f AV |
3069 | putname(tmp); |
3070 | return res; | |
3071 | } | |
3072 | EXPORT_SYMBOL(user_path_create); | |
3073 | ||
1a67aafb | 3074 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3075 | { |
a95164d9 | 3076 | int error = may_create(dir, dentry); |
1da177e4 LT |
3077 | |
3078 | if (error) | |
3079 | return error; | |
3080 | ||
975d6b39 | 3081 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3082 | return -EPERM; |
3083 | ||
acfa4380 | 3084 | if (!dir->i_op->mknod) |
1da177e4 LT |
3085 | return -EPERM; |
3086 | ||
08ce5f16 SH |
3087 | error = devcgroup_inode_mknod(mode, dev); |
3088 | if (error) | |
3089 | return error; | |
3090 | ||
1da177e4 LT |
3091 | error = security_inode_mknod(dir, dentry, mode, dev); |
3092 | if (error) | |
3093 | return error; | |
3094 | ||
1da177e4 | 3095 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3096 | if (!error) |
f38aa942 | 3097 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3098 | return error; |
3099 | } | |
3100 | ||
f69aac00 | 3101 | static int may_mknod(umode_t mode) |
463c3197 DH |
3102 | { |
3103 | switch (mode & S_IFMT) { | |
3104 | case S_IFREG: | |
3105 | case S_IFCHR: | |
3106 | case S_IFBLK: | |
3107 | case S_IFIFO: | |
3108 | case S_IFSOCK: | |
3109 | case 0: /* zero mode translates to S_IFREG */ | |
3110 | return 0; | |
3111 | case S_IFDIR: | |
3112 | return -EPERM; | |
3113 | default: | |
3114 | return -EINVAL; | |
3115 | } | |
3116 | } | |
3117 | ||
8208a22b | 3118 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 3119 | unsigned, dev) |
1da177e4 | 3120 | { |
2ad94ae6 | 3121 | struct dentry *dentry; |
dae6ad8f AV |
3122 | struct path path; |
3123 | int error; | |
1da177e4 | 3124 | |
8e4bfca1 AV |
3125 | error = may_mknod(mode); |
3126 | if (error) | |
3127 | return error; | |
1da177e4 | 3128 | |
dae6ad8f AV |
3129 | dentry = user_path_create(dfd, filename, &path, 0); |
3130 | if (IS_ERR(dentry)) | |
3131 | return PTR_ERR(dentry); | |
2ad94ae6 | 3132 | |
dae6ad8f | 3133 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3134 | mode &= ~current_umask(); |
dae6ad8f | 3135 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3136 | if (error) |
a8104a9f | 3137 | goto out; |
463c3197 | 3138 | switch (mode & S_IFMT) { |
1da177e4 | 3139 | case 0: case S_IFREG: |
312b63fb | 3140 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
1da177e4 LT |
3141 | break; |
3142 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3143 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3144 | new_decode_dev(dev)); |
3145 | break; | |
3146 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3147 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3148 | break; |
1da177e4 | 3149 | } |
a8104a9f | 3150 | out: |
921a1650 | 3151 | done_path_create(&path, dentry); |
1da177e4 LT |
3152 | return error; |
3153 | } | |
3154 | ||
8208a22b | 3155 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
3156 | { |
3157 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
3158 | } | |
3159 | ||
18bb1db3 | 3160 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3161 | { |
a95164d9 | 3162 | int error = may_create(dir, dentry); |
8de52778 | 3163 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3164 | |
3165 | if (error) | |
3166 | return error; | |
3167 | ||
acfa4380 | 3168 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3169 | return -EPERM; |
3170 | ||
3171 | mode &= (S_IRWXUGO|S_ISVTX); | |
3172 | error = security_inode_mkdir(dir, dentry, mode); | |
3173 | if (error) | |
3174 | return error; | |
3175 | ||
8de52778 AV |
3176 | if (max_links && dir->i_nlink >= max_links) |
3177 | return -EMLINK; | |
3178 | ||
1da177e4 | 3179 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3180 | if (!error) |
f38aa942 | 3181 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3182 | return error; |
3183 | } | |
3184 | ||
a218d0fd | 3185 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 3186 | { |
6902d925 | 3187 | struct dentry *dentry; |
dae6ad8f AV |
3188 | struct path path; |
3189 | int error; | |
1da177e4 | 3190 | |
dae6ad8f | 3191 | dentry = user_path_create(dfd, pathname, &path, 1); |
6902d925 | 3192 | if (IS_ERR(dentry)) |
dae6ad8f | 3193 | return PTR_ERR(dentry); |
1da177e4 | 3194 | |
dae6ad8f | 3195 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3196 | mode &= ~current_umask(); |
dae6ad8f | 3197 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3198 | if (!error) |
3199 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3200 | done_path_create(&path, dentry); |
1da177e4 LT |
3201 | return error; |
3202 | } | |
3203 | ||
a218d0fd | 3204 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
3205 | { |
3206 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
3207 | } | |
3208 | ||
1da177e4 | 3209 | /* |
a71905f0 | 3210 | * The dentry_unhash() helper will try to drop the dentry early: we |
c0d02594 | 3211 | * should have a usage count of 1 if we're the only user of this |
a71905f0 SW |
3212 | * dentry, and if that is true (possibly after pruning the dcache), |
3213 | * then we drop the dentry now. | |
1da177e4 LT |
3214 | * |
3215 | * A low-level filesystem can, if it choses, legally | |
3216 | * do a | |
3217 | * | |
3218 | * if (!d_unhashed(dentry)) | |
3219 | * return -EBUSY; | |
3220 | * | |
3221 | * if it cannot handle the case of removing a directory | |
3222 | * that is still in use by something else.. | |
3223 | */ | |
3224 | void dentry_unhash(struct dentry *dentry) | |
3225 | { | |
dc168427 | 3226 | shrink_dcache_parent(dentry); |
1da177e4 | 3227 | spin_lock(&dentry->d_lock); |
64252c75 | 3228 | if (dentry->d_count == 1) |
1da177e4 LT |
3229 | __d_drop(dentry); |
3230 | spin_unlock(&dentry->d_lock); | |
1da177e4 LT |
3231 | } |
3232 | ||
3233 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
3234 | { | |
3235 | int error = may_delete(dir, dentry, 1); | |
3236 | ||
3237 | if (error) | |
3238 | return error; | |
3239 | ||
acfa4380 | 3240 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3241 | return -EPERM; |
3242 | ||
1d2ef590 | 3243 | dget(dentry); |
1b1dcc1b | 3244 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
3245 | |
3246 | error = -EBUSY; | |
1da177e4 | 3247 | if (d_mountpoint(dentry)) |
912dbc15 SW |
3248 | goto out; |
3249 | ||
3250 | error = security_inode_rmdir(dir, dentry); | |
3251 | if (error) | |
3252 | goto out; | |
3253 | ||
3cebde24 | 3254 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3255 | error = dir->i_op->rmdir(dir, dentry); |
3256 | if (error) | |
3257 | goto out; | |
3258 | ||
3259 | dentry->d_inode->i_flags |= S_DEAD; | |
3260 | dont_mount(dentry); | |
3261 | ||
3262 | out: | |
1b1dcc1b | 3263 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 3264 | dput(dentry); |
912dbc15 | 3265 | if (!error) |
1da177e4 | 3266 | d_delete(dentry); |
1da177e4 LT |
3267 | return error; |
3268 | } | |
3269 | ||
5590ff0d | 3270 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3271 | { |
3272 | int error = 0; | |
91a27b2a | 3273 | struct filename *name; |
1da177e4 LT |
3274 | struct dentry *dentry; |
3275 | struct nameidata nd; | |
3276 | ||
91a27b2a JL |
3277 | name = user_path_parent(dfd, pathname, &nd); |
3278 | if (IS_ERR(name)) | |
3279 | return PTR_ERR(name); | |
1da177e4 LT |
3280 | |
3281 | switch(nd.last_type) { | |
0612d9fb OH |
3282 | case LAST_DOTDOT: |
3283 | error = -ENOTEMPTY; | |
3284 | goto exit1; | |
3285 | case LAST_DOT: | |
3286 | error = -EINVAL; | |
3287 | goto exit1; | |
3288 | case LAST_ROOT: | |
3289 | error = -EBUSY; | |
3290 | goto exit1; | |
1da177e4 | 3291 | } |
0612d9fb OH |
3292 | |
3293 | nd.flags &= ~LOOKUP_PARENT; | |
c30dabfe JK |
3294 | error = mnt_want_write(nd.path.mnt); |
3295 | if (error) | |
3296 | goto exit1; | |
0612d9fb | 3297 | |
4ac91378 | 3298 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 3299 | dentry = lookup_hash(&nd); |
1da177e4 | 3300 | error = PTR_ERR(dentry); |
6902d925 DH |
3301 | if (IS_ERR(dentry)) |
3302 | goto exit2; | |
e6bc45d6 TT |
3303 | if (!dentry->d_inode) { |
3304 | error = -ENOENT; | |
3305 | goto exit3; | |
3306 | } | |
be6d3e56 KT |
3307 | error = security_path_rmdir(&nd.path, dentry); |
3308 | if (error) | |
c30dabfe | 3309 | goto exit3; |
4ac91378 | 3310 | error = vfs_rmdir(nd.path.dentry->d_inode, dentry); |
0622753b | 3311 | exit3: |
6902d925 DH |
3312 | dput(dentry); |
3313 | exit2: | |
4ac91378 | 3314 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
c30dabfe | 3315 | mnt_drop_write(nd.path.mnt); |
1da177e4 | 3316 | exit1: |
1d957f9b | 3317 | path_put(&nd.path); |
1da177e4 LT |
3318 | putname(name); |
3319 | return error; | |
3320 | } | |
3321 | ||
3cdad428 | 3322 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3323 | { |
3324 | return do_rmdir(AT_FDCWD, pathname); | |
3325 | } | |
3326 | ||
1da177e4 LT |
3327 | int vfs_unlink(struct inode *dir, struct dentry *dentry) |
3328 | { | |
3329 | int error = may_delete(dir, dentry, 0); | |
3330 | ||
3331 | if (error) | |
3332 | return error; | |
3333 | ||
acfa4380 | 3334 | if (!dir->i_op->unlink) |
1da177e4 LT |
3335 | return -EPERM; |
3336 | ||
1b1dcc1b | 3337 | mutex_lock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
3338 | if (d_mountpoint(dentry)) |
3339 | error = -EBUSY; | |
3340 | else { | |
3341 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3342 | if (!error) { |
1da177e4 | 3343 | error = dir->i_op->unlink(dir, dentry); |
bec1052e | 3344 | if (!error) |
d83c49f3 | 3345 | dont_mount(dentry); |
bec1052e | 3346 | } |
1da177e4 | 3347 | } |
1b1dcc1b | 3348 | mutex_unlock(&dentry->d_inode->i_mutex); |
1da177e4 LT |
3349 | |
3350 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3351 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
ece95912 | 3352 | fsnotify_link_count(dentry->d_inode); |
e234f35c | 3353 | d_delete(dentry); |
1da177e4 | 3354 | } |
0eeca283 | 3355 | |
1da177e4 LT |
3356 | return error; |
3357 | } | |
3358 | ||
3359 | /* | |
3360 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3361 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3362 | * writeout happening, and we don't want to prevent access to the directory |
3363 | * while waiting on the I/O. | |
3364 | */ | |
5590ff0d | 3365 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 3366 | { |
2ad94ae6 | 3367 | int error; |
91a27b2a | 3368 | struct filename *name; |
1da177e4 LT |
3369 | struct dentry *dentry; |
3370 | struct nameidata nd; | |
3371 | struct inode *inode = NULL; | |
3372 | ||
91a27b2a JL |
3373 | name = user_path_parent(dfd, pathname, &nd); |
3374 | if (IS_ERR(name)) | |
3375 | return PTR_ERR(name); | |
2ad94ae6 | 3376 | |
1da177e4 LT |
3377 | error = -EISDIR; |
3378 | if (nd.last_type != LAST_NORM) | |
3379 | goto exit1; | |
0612d9fb OH |
3380 | |
3381 | nd.flags &= ~LOOKUP_PARENT; | |
c30dabfe JK |
3382 | error = mnt_want_write(nd.path.mnt); |
3383 | if (error) | |
3384 | goto exit1; | |
0612d9fb | 3385 | |
4ac91378 | 3386 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
49705b77 | 3387 | dentry = lookup_hash(&nd); |
1da177e4 LT |
3388 | error = PTR_ERR(dentry); |
3389 | if (!IS_ERR(dentry)) { | |
3390 | /* Why not before? Because we want correct error value */ | |
50338b88 TE |
3391 | if (nd.last.name[nd.last.len]) |
3392 | goto slashes; | |
1da177e4 | 3393 | inode = dentry->d_inode; |
50338b88 | 3394 | if (!inode) |
e6bc45d6 TT |
3395 | goto slashes; |
3396 | ihold(inode); | |
be6d3e56 KT |
3397 | error = security_path_unlink(&nd.path, dentry); |
3398 | if (error) | |
c30dabfe | 3399 | goto exit2; |
4ac91378 | 3400 | error = vfs_unlink(nd.path.dentry->d_inode, dentry); |
c30dabfe | 3401 | exit2: |
1da177e4 LT |
3402 | dput(dentry); |
3403 | } | |
4ac91378 | 3404 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
1da177e4 LT |
3405 | if (inode) |
3406 | iput(inode); /* truncate the inode here */ | |
c30dabfe | 3407 | mnt_drop_write(nd.path.mnt); |
1da177e4 | 3408 | exit1: |
1d957f9b | 3409 | path_put(&nd.path); |
1da177e4 LT |
3410 | putname(name); |
3411 | return error; | |
3412 | ||
3413 | slashes: | |
3414 | error = !dentry->d_inode ? -ENOENT : | |
3415 | S_ISDIR(dentry->d_inode->i_mode) ? -EISDIR : -ENOTDIR; | |
3416 | goto exit2; | |
3417 | } | |
3418 | ||
2e4d0924 | 3419 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3420 | { |
3421 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3422 | return -EINVAL; | |
3423 | ||
3424 | if (flag & AT_REMOVEDIR) | |
3425 | return do_rmdir(dfd, pathname); | |
3426 | ||
3427 | return do_unlinkat(dfd, pathname); | |
3428 | } | |
3429 | ||
3480b257 | 3430 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
3431 | { |
3432 | return do_unlinkat(AT_FDCWD, pathname); | |
3433 | } | |
3434 | ||
db2e747b | 3435 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3436 | { |
a95164d9 | 3437 | int error = may_create(dir, dentry); |
1da177e4 LT |
3438 | |
3439 | if (error) | |
3440 | return error; | |
3441 | ||
acfa4380 | 3442 | if (!dir->i_op->symlink) |
1da177e4 LT |
3443 | return -EPERM; |
3444 | ||
3445 | error = security_inode_symlink(dir, dentry, oldname); | |
3446 | if (error) | |
3447 | return error; | |
3448 | ||
1da177e4 | 3449 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3450 | if (!error) |
f38aa942 | 3451 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3452 | return error; |
3453 | } | |
3454 | ||
2e4d0924 HC |
3455 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
3456 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3457 | { |
2ad94ae6 | 3458 | int error; |
91a27b2a | 3459 | struct filename *from; |
6902d925 | 3460 | struct dentry *dentry; |
dae6ad8f | 3461 | struct path path; |
1da177e4 LT |
3462 | |
3463 | from = getname(oldname); | |
2ad94ae6 | 3464 | if (IS_ERR(from)) |
1da177e4 | 3465 | return PTR_ERR(from); |
1da177e4 | 3466 | |
dae6ad8f | 3467 | dentry = user_path_create(newdfd, newname, &path, 0); |
6902d925 DH |
3468 | error = PTR_ERR(dentry); |
3469 | if (IS_ERR(dentry)) | |
dae6ad8f | 3470 | goto out_putname; |
6902d925 | 3471 | |
91a27b2a | 3472 | error = security_path_symlink(&path, dentry, from->name); |
a8104a9f | 3473 | if (!error) |
91a27b2a | 3474 | error = vfs_symlink(path.dentry->d_inode, dentry, from->name); |
921a1650 | 3475 | done_path_create(&path, dentry); |
6902d925 | 3476 | out_putname: |
1da177e4 LT |
3477 | putname(from); |
3478 | return error; | |
3479 | } | |
3480 | ||
3480b257 | 3481 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3482 | { |
3483 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
3484 | } | |
3485 | ||
1da177e4 LT |
3486 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry) |
3487 | { | |
3488 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 3489 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3490 | int error; |
3491 | ||
3492 | if (!inode) | |
3493 | return -ENOENT; | |
3494 | ||
a95164d9 | 3495 | error = may_create(dir, new_dentry); |
1da177e4 LT |
3496 | if (error) |
3497 | return error; | |
3498 | ||
3499 | if (dir->i_sb != inode->i_sb) | |
3500 | return -EXDEV; | |
3501 | ||
3502 | /* | |
3503 | * A link to an append-only or immutable file cannot be created. | |
3504 | */ | |
3505 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
3506 | return -EPERM; | |
acfa4380 | 3507 | if (!dir->i_op->link) |
1da177e4 | 3508 | return -EPERM; |
7e79eedb | 3509 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
3510 | return -EPERM; |
3511 | ||
3512 | error = security_inode_link(old_dentry, dir, new_dentry); | |
3513 | if (error) | |
3514 | return error; | |
3515 | ||
7e79eedb | 3516 | mutex_lock(&inode->i_mutex); |
aae8a97d AK |
3517 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
3518 | if (inode->i_nlink == 0) | |
3519 | error = -ENOENT; | |
8de52778 AV |
3520 | else if (max_links && inode->i_nlink >= max_links) |
3521 | error = -EMLINK; | |
aae8a97d AK |
3522 | else |
3523 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
7e79eedb | 3524 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 3525 | if (!error) |
7e79eedb | 3526 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
3527 | return error; |
3528 | } | |
3529 | ||
3530 | /* | |
3531 | * Hardlinks are often used in delicate situations. We avoid | |
3532 | * security-related surprises by not following symlinks on the | |
3533 | * newname. --KAB | |
3534 | * | |
3535 | * We don't follow them on the oldname either to be compatible | |
3536 | * with linux 2.0, and to avoid hard-linking to directories | |
3537 | * and other special files. --ADM | |
3538 | */ | |
2e4d0924 HC |
3539 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
3540 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
3541 | { |
3542 | struct dentry *new_dentry; | |
dae6ad8f | 3543 | struct path old_path, new_path; |
11a7b371 | 3544 | int how = 0; |
1da177e4 | 3545 | int error; |
1da177e4 | 3546 | |
11a7b371 | 3547 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 3548 | return -EINVAL; |
11a7b371 AK |
3549 | /* |
3550 | * To use null names we require CAP_DAC_READ_SEARCH | |
3551 | * This ensures that not everyone will be able to create | |
3552 | * handlink using the passed filedescriptor. | |
3553 | */ | |
3554 | if (flags & AT_EMPTY_PATH) { | |
3555 | if (!capable(CAP_DAC_READ_SEARCH)) | |
3556 | return -ENOENT; | |
3557 | how = LOOKUP_EMPTY; | |
3558 | } | |
3559 | ||
3560 | if (flags & AT_SYMLINK_FOLLOW) | |
3561 | how |= LOOKUP_FOLLOW; | |
c04030e1 | 3562 | |
11a7b371 | 3563 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 3564 | if (error) |
2ad94ae6 AV |
3565 | return error; |
3566 | ||
dae6ad8f | 3567 | new_dentry = user_path_create(newdfd, newname, &new_path, 0); |
1da177e4 | 3568 | error = PTR_ERR(new_dentry); |
6902d925 | 3569 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
3570 | goto out; |
3571 | ||
3572 | error = -EXDEV; | |
3573 | if (old_path.mnt != new_path.mnt) | |
3574 | goto out_dput; | |
800179c9 KC |
3575 | error = may_linkat(&old_path); |
3576 | if (unlikely(error)) | |
3577 | goto out_dput; | |
dae6ad8f | 3578 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 3579 | if (error) |
a8104a9f | 3580 | goto out_dput; |
dae6ad8f | 3581 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry); |
75c3f29d | 3582 | out_dput: |
921a1650 | 3583 | done_path_create(&new_path, new_dentry); |
1da177e4 | 3584 | out: |
2d8f3038 | 3585 | path_put(&old_path); |
1da177e4 LT |
3586 | |
3587 | return error; | |
3588 | } | |
3589 | ||
3480b257 | 3590 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 3591 | { |
c04030e1 | 3592 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
3593 | } |
3594 | ||
1da177e4 LT |
3595 | /* |
3596 | * The worst of all namespace operations - renaming directory. "Perverted" | |
3597 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
3598 | * Problems: | |
3599 | * a) we can get into loop creation. Check is done in is_subdir(). | |
3600 | * b) race potential - two innocent renames can create a loop together. | |
3601 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 3602 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 LT |
3603 | * story. |
3604 | * c) we have to lock _three_ objects - parents and victim (if it exists). | |
1b1dcc1b | 3605 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
3606 | * whether the target exists). Solution: try to be smart with locking |
3607 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 3608 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
3609 | * move will be locked. Thus we can rank directories by the tree |
3610 | * (ancestors first) and rank all non-directories after them. | |
3611 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 3612 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
3613 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
3614 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
3615 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 3616 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 3617 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 3618 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 3619 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
3620 | * locking]. |
3621 | */ | |
75c96f85 AB |
3622 | static int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, |
3623 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 LT |
3624 | { |
3625 | int error = 0; | |
9055cba7 | 3626 | struct inode *target = new_dentry->d_inode; |
8de52778 | 3627 | unsigned max_links = new_dir->i_sb->s_max_links; |
1da177e4 LT |
3628 | |
3629 | /* | |
3630 | * If we are going to change the parent - check write permissions, | |
3631 | * we'll need to flip '..'. | |
3632 | */ | |
3633 | if (new_dir != old_dir) { | |
f419a2e3 | 3634 | error = inode_permission(old_dentry->d_inode, MAY_WRITE); |
1da177e4 LT |
3635 | if (error) |
3636 | return error; | |
3637 | } | |
3638 | ||
3639 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3640 | if (error) | |
3641 | return error; | |
3642 | ||
1d2ef590 | 3643 | dget(new_dentry); |
d83c49f3 | 3644 | if (target) |
1b1dcc1b | 3645 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
3646 | |
3647 | error = -EBUSY; | |
3648 | if (d_mountpoint(old_dentry) || d_mountpoint(new_dentry)) | |
3649 | goto out; | |
3650 | ||
8de52778 AV |
3651 | error = -EMLINK; |
3652 | if (max_links && !target && new_dir != old_dir && | |
3653 | new_dir->i_nlink >= max_links) | |
3654 | goto out; | |
3655 | ||
3cebde24 SW |
3656 | if (target) |
3657 | shrink_dcache_parent(new_dentry); | |
9055cba7 SW |
3658 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); |
3659 | if (error) | |
3660 | goto out; | |
3661 | ||
1da177e4 | 3662 | if (target) { |
9055cba7 SW |
3663 | target->i_flags |= S_DEAD; |
3664 | dont_mount(new_dentry); | |
1da177e4 | 3665 | } |
9055cba7 SW |
3666 | out: |
3667 | if (target) | |
3668 | mutex_unlock(&target->i_mutex); | |
1d2ef590 | 3669 | dput(new_dentry); |
e31e14ec | 3670 | if (!error) |
349457cc MF |
3671 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) |
3672 | d_move(old_dentry,new_dentry); | |
1da177e4 LT |
3673 | return error; |
3674 | } | |
3675 | ||
75c96f85 AB |
3676 | static int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, |
3677 | struct inode *new_dir, struct dentry *new_dentry) | |
1da177e4 | 3678 | { |
51892bbb | 3679 | struct inode *target = new_dentry->d_inode; |
1da177e4 LT |
3680 | int error; |
3681 | ||
3682 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry); | |
3683 | if (error) | |
3684 | return error; | |
3685 | ||
3686 | dget(new_dentry); | |
1da177e4 | 3687 | if (target) |
1b1dcc1b | 3688 | mutex_lock(&target->i_mutex); |
51892bbb SW |
3689 | |
3690 | error = -EBUSY; | |
1da177e4 | 3691 | if (d_mountpoint(old_dentry)||d_mountpoint(new_dentry)) |
51892bbb SW |
3692 | goto out; |
3693 | ||
3694 | error = old_dir->i_op->rename(old_dir, old_dentry, new_dir, new_dentry); | |
3695 | if (error) | |
3696 | goto out; | |
3697 | ||
3698 | if (target) | |
3699 | dont_mount(new_dentry); | |
3700 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) | |
3701 | d_move(old_dentry, new_dentry); | |
3702 | out: | |
1da177e4 | 3703 | if (target) |
1b1dcc1b | 3704 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3705 | dput(new_dentry); |
3706 | return error; | |
3707 | } | |
3708 | ||
3709 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, | |
3710 | struct inode *new_dir, struct dentry *new_dentry) | |
3711 | { | |
3712 | int error; | |
3713 | int is_dir = S_ISDIR(old_dentry->d_inode->i_mode); | |
59b0df21 | 3714 | const unsigned char *old_name; |
1da177e4 LT |
3715 | |
3716 | if (old_dentry->d_inode == new_dentry->d_inode) | |
3717 | return 0; | |
3718 | ||
3719 | error = may_delete(old_dir, old_dentry, is_dir); | |
3720 | if (error) | |
3721 | return error; | |
3722 | ||
3723 | if (!new_dentry->d_inode) | |
a95164d9 | 3724 | error = may_create(new_dir, new_dentry); |
1da177e4 LT |
3725 | else |
3726 | error = may_delete(new_dir, new_dentry, is_dir); | |
3727 | if (error) | |
3728 | return error; | |
3729 | ||
acfa4380 | 3730 | if (!old_dir->i_op->rename) |
1da177e4 LT |
3731 | return -EPERM; |
3732 | ||
0eeca283 RL |
3733 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
3734 | ||
1da177e4 LT |
3735 | if (is_dir) |
3736 | error = vfs_rename_dir(old_dir,old_dentry,new_dir,new_dentry); | |
3737 | else | |
3738 | error = vfs_rename_other(old_dir,old_dentry,new_dir,new_dentry); | |
123df294 AV |
3739 | if (!error) |
3740 | fsnotify_move(old_dir, new_dir, old_name, is_dir, | |
5a190ae6 | 3741 | new_dentry->d_inode, old_dentry); |
0eeca283 RL |
3742 | fsnotify_oldname_free(old_name); |
3743 | ||
1da177e4 LT |
3744 | return error; |
3745 | } | |
3746 | ||
2e4d0924 HC |
3747 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
3748 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3749 | { |
2ad94ae6 AV |
3750 | struct dentry *old_dir, *new_dir; |
3751 | struct dentry *old_dentry, *new_dentry; | |
3752 | struct dentry *trap; | |
1da177e4 | 3753 | struct nameidata oldnd, newnd; |
91a27b2a JL |
3754 | struct filename *from; |
3755 | struct filename *to; | |
2ad94ae6 | 3756 | int error; |
1da177e4 | 3757 | |
91a27b2a JL |
3758 | from = user_path_parent(olddfd, oldname, &oldnd); |
3759 | if (IS_ERR(from)) { | |
3760 | error = PTR_ERR(from); | |
1da177e4 | 3761 | goto exit; |
91a27b2a | 3762 | } |
1da177e4 | 3763 | |
91a27b2a JL |
3764 | to = user_path_parent(newdfd, newname, &newnd); |
3765 | if (IS_ERR(to)) { | |
3766 | error = PTR_ERR(to); | |
1da177e4 | 3767 | goto exit1; |
91a27b2a | 3768 | } |
1da177e4 LT |
3769 | |
3770 | error = -EXDEV; | |
4ac91378 | 3771 | if (oldnd.path.mnt != newnd.path.mnt) |
1da177e4 LT |
3772 | goto exit2; |
3773 | ||
4ac91378 | 3774 | old_dir = oldnd.path.dentry; |
1da177e4 LT |
3775 | error = -EBUSY; |
3776 | if (oldnd.last_type != LAST_NORM) | |
3777 | goto exit2; | |
3778 | ||
4ac91378 | 3779 | new_dir = newnd.path.dentry; |
1da177e4 LT |
3780 | if (newnd.last_type != LAST_NORM) |
3781 | goto exit2; | |
3782 | ||
c30dabfe JK |
3783 | error = mnt_want_write(oldnd.path.mnt); |
3784 | if (error) | |
3785 | goto exit2; | |
3786 | ||
0612d9fb OH |
3787 | oldnd.flags &= ~LOOKUP_PARENT; |
3788 | newnd.flags &= ~LOOKUP_PARENT; | |
4e9ed2f8 | 3789 | newnd.flags |= LOOKUP_RENAME_TARGET; |
0612d9fb | 3790 | |
1da177e4 LT |
3791 | trap = lock_rename(new_dir, old_dir); |
3792 | ||
49705b77 | 3793 | old_dentry = lookup_hash(&oldnd); |
1da177e4 LT |
3794 | error = PTR_ERR(old_dentry); |
3795 | if (IS_ERR(old_dentry)) | |
3796 | goto exit3; | |
3797 | /* source must exist */ | |
3798 | error = -ENOENT; | |
3799 | if (!old_dentry->d_inode) | |
3800 | goto exit4; | |
3801 | /* unless the source is a directory trailing slashes give -ENOTDIR */ | |
3802 | if (!S_ISDIR(old_dentry->d_inode->i_mode)) { | |
3803 | error = -ENOTDIR; | |
3804 | if (oldnd.last.name[oldnd.last.len]) | |
3805 | goto exit4; | |
3806 | if (newnd.last.name[newnd.last.len]) | |
3807 | goto exit4; | |
3808 | } | |
3809 | /* source should not be ancestor of target */ | |
3810 | error = -EINVAL; | |
3811 | if (old_dentry == trap) | |
3812 | goto exit4; | |
49705b77 | 3813 | new_dentry = lookup_hash(&newnd); |
1da177e4 LT |
3814 | error = PTR_ERR(new_dentry); |
3815 | if (IS_ERR(new_dentry)) | |
3816 | goto exit4; | |
3817 | /* target should not be an ancestor of source */ | |
3818 | error = -ENOTEMPTY; | |
3819 | if (new_dentry == trap) | |
3820 | goto exit5; | |
3821 | ||
be6d3e56 KT |
3822 | error = security_path_rename(&oldnd.path, old_dentry, |
3823 | &newnd.path, new_dentry); | |
3824 | if (error) | |
c30dabfe | 3825 | goto exit5; |
1da177e4 LT |
3826 | error = vfs_rename(old_dir->d_inode, old_dentry, |
3827 | new_dir->d_inode, new_dentry); | |
3828 | exit5: | |
3829 | dput(new_dentry); | |
3830 | exit4: | |
3831 | dput(old_dentry); | |
3832 | exit3: | |
3833 | unlock_rename(new_dir, old_dir); | |
c30dabfe | 3834 | mnt_drop_write(oldnd.path.mnt); |
1da177e4 | 3835 | exit2: |
1d957f9b | 3836 | path_put(&newnd.path); |
2ad94ae6 | 3837 | putname(to); |
1da177e4 | 3838 | exit1: |
1d957f9b | 3839 | path_put(&oldnd.path); |
1da177e4 | 3840 | putname(from); |
2ad94ae6 | 3841 | exit: |
1da177e4 LT |
3842 | return error; |
3843 | } | |
3844 | ||
a26eab24 | 3845 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3846 | { |
3847 | return sys_renameat(AT_FDCWD, oldname, AT_FDCWD, newname); | |
3848 | } | |
3849 | ||
1da177e4 LT |
3850 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen, const char *link) |
3851 | { | |
3852 | int len; | |
3853 | ||
3854 | len = PTR_ERR(link); | |
3855 | if (IS_ERR(link)) | |
3856 | goto out; | |
3857 | ||
3858 | len = strlen(link); | |
3859 | if (len > (unsigned) buflen) | |
3860 | len = buflen; | |
3861 | if (copy_to_user(buffer, link, len)) | |
3862 | len = -EFAULT; | |
3863 | out: | |
3864 | return len; | |
3865 | } | |
3866 | ||
3867 | /* | |
3868 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
3869 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
3870 | * using) it for any given inode is up to filesystem. | |
3871 | */ | |
3872 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3873 | { | |
3874 | struct nameidata nd; | |
cc314eef | 3875 | void *cookie; |
694a1764 | 3876 | int res; |
cc314eef | 3877 | |
1da177e4 | 3878 | nd.depth = 0; |
cc314eef | 3879 | cookie = dentry->d_inode->i_op->follow_link(dentry, &nd); |
694a1764 MS |
3880 | if (IS_ERR(cookie)) |
3881 | return PTR_ERR(cookie); | |
3882 | ||
3883 | res = vfs_readlink(dentry, buffer, buflen, nd_get_link(&nd)); | |
3884 | if (dentry->d_inode->i_op->put_link) | |
3885 | dentry->d_inode->i_op->put_link(dentry, &nd, cookie); | |
3886 | return res; | |
1da177e4 LT |
3887 | } |
3888 | ||
3889 | int vfs_follow_link(struct nameidata *nd, const char *link) | |
3890 | { | |
3891 | return __vfs_follow_link(nd, link); | |
3892 | } | |
3893 | ||
3894 | /* get the link contents into pagecache */ | |
3895 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
3896 | { | |
ebd09abb DG |
3897 | char *kaddr; |
3898 | struct page *page; | |
1da177e4 | 3899 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 3900 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 3901 | if (IS_ERR(page)) |
6fe6900e | 3902 | return (char*)page; |
1da177e4 | 3903 | *ppage = page; |
ebd09abb DG |
3904 | kaddr = kmap(page); |
3905 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
3906 | return kaddr; | |
1da177e4 LT |
3907 | } |
3908 | ||
3909 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
3910 | { | |
3911 | struct page *page = NULL; | |
3912 | char *s = page_getlink(dentry, &page); | |
3913 | int res = vfs_readlink(dentry,buffer,buflen,s); | |
3914 | if (page) { | |
3915 | kunmap(page); | |
3916 | page_cache_release(page); | |
3917 | } | |
3918 | return res; | |
3919 | } | |
3920 | ||
cc314eef | 3921 | void *page_follow_link_light(struct dentry *dentry, struct nameidata *nd) |
1da177e4 | 3922 | { |
cc314eef | 3923 | struct page *page = NULL; |
1da177e4 | 3924 | nd_set_link(nd, page_getlink(dentry, &page)); |
cc314eef | 3925 | return page; |
1da177e4 LT |
3926 | } |
3927 | ||
cc314eef | 3928 | void page_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie) |
1da177e4 | 3929 | { |
cc314eef LT |
3930 | struct page *page = cookie; |
3931 | ||
3932 | if (page) { | |
1da177e4 LT |
3933 | kunmap(page); |
3934 | page_cache_release(page); | |
1da177e4 LT |
3935 | } |
3936 | } | |
3937 | ||
54566b2c NP |
3938 | /* |
3939 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
3940 | */ | |
3941 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
3942 | { |
3943 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 3944 | struct page *page; |
afddba49 | 3945 | void *fsdata; |
beb497ab | 3946 | int err; |
1da177e4 | 3947 | char *kaddr; |
54566b2c NP |
3948 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
3949 | if (nofs) | |
3950 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 3951 | |
7e53cac4 | 3952 | retry: |
afddba49 | 3953 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 3954 | flags, &page, &fsdata); |
1da177e4 | 3955 | if (err) |
afddba49 NP |
3956 | goto fail; |
3957 | ||
e8e3c3d6 | 3958 | kaddr = kmap_atomic(page); |
1da177e4 | 3959 | memcpy(kaddr, symname, len-1); |
e8e3c3d6 | 3960 | kunmap_atomic(kaddr); |
afddba49 NP |
3961 | |
3962 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
3963 | page, fsdata); | |
1da177e4 LT |
3964 | if (err < 0) |
3965 | goto fail; | |
afddba49 NP |
3966 | if (err < len-1) |
3967 | goto retry; | |
3968 | ||
1da177e4 LT |
3969 | mark_inode_dirty(inode); |
3970 | return 0; | |
1da177e4 LT |
3971 | fail: |
3972 | return err; | |
3973 | } | |
3974 | ||
0adb25d2 KK |
3975 | int page_symlink(struct inode *inode, const char *symname, int len) |
3976 | { | |
3977 | return __page_symlink(inode, symname, len, | |
54566b2c | 3978 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 KK |
3979 | } |
3980 | ||
92e1d5be | 3981 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
3982 | .readlink = generic_readlink, |
3983 | .follow_link = page_follow_link_light, | |
3984 | .put_link = page_put_link, | |
3985 | }; | |
3986 | ||
2d8f3038 | 3987 | EXPORT_SYMBOL(user_path_at); |
cc53ce53 | 3988 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 LT |
3989 | EXPORT_SYMBOL(follow_down); |
3990 | EXPORT_SYMBOL(follow_up); | |
f6d2ac5c | 3991 | EXPORT_SYMBOL(get_write_access); /* nfsd */ |
1da177e4 | 3992 | EXPORT_SYMBOL(lock_rename); |
1da177e4 LT |
3993 | EXPORT_SYMBOL(lookup_one_len); |
3994 | EXPORT_SYMBOL(page_follow_link_light); | |
3995 | EXPORT_SYMBOL(page_put_link); | |
3996 | EXPORT_SYMBOL(page_readlink); | |
0adb25d2 | 3997 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 LT |
3998 | EXPORT_SYMBOL(page_symlink); |
3999 | EXPORT_SYMBOL(page_symlink_inode_operations); | |
d1811465 | 4000 | EXPORT_SYMBOL(kern_path); |
16f18200 | 4001 | EXPORT_SYMBOL(vfs_path_lookup); |
f419a2e3 | 4002 | EXPORT_SYMBOL(inode_permission); |
1da177e4 LT |
4003 | EXPORT_SYMBOL(unlock_rename); |
4004 | EXPORT_SYMBOL(vfs_create); | |
4005 | EXPORT_SYMBOL(vfs_follow_link); | |
4006 | EXPORT_SYMBOL(vfs_link); | |
4007 | EXPORT_SYMBOL(vfs_mkdir); | |
4008 | EXPORT_SYMBOL(vfs_mknod); | |
4009 | EXPORT_SYMBOL(generic_permission); | |
4010 | EXPORT_SYMBOL(vfs_readlink); | |
4011 | EXPORT_SYMBOL(vfs_rename); | |
4012 | EXPORT_SYMBOL(vfs_rmdir); | |
4013 | EXPORT_SYMBOL(vfs_symlink); | |
4014 | EXPORT_SYMBOL(vfs_unlink); | |
4015 | EXPORT_SYMBOL(dentry_unhash); | |
4016 | EXPORT_SYMBOL(generic_readlink); |