Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/fs/namei.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
8 | /* | |
9 | * Some corrections by tytso. | |
10 | */ | |
11 | ||
12 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
13 | * lookup logic. | |
14 | */ | |
15 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
16 | */ | |
17 | ||
18 | #include <linux/init.h> | |
630d9c47 | 19 | #include <linux/export.h> |
44696908 | 20 | #include <linux/kernel.h> |
1da177e4 LT |
21 | #include <linux/slab.h> |
22 | #include <linux/fs.h> | |
23 | #include <linux/namei.h> | |
1da177e4 | 24 | #include <linux/pagemap.h> |
0eeca283 | 25 | #include <linux/fsnotify.h> |
1da177e4 LT |
26 | #include <linux/personality.h> |
27 | #include <linux/security.h> | |
6146f0d5 | 28 | #include <linux/ima.h> |
1da177e4 LT |
29 | #include <linux/syscalls.h> |
30 | #include <linux/mount.h> | |
31 | #include <linux/audit.h> | |
16f7e0fe | 32 | #include <linux/capability.h> |
834f2a4a | 33 | #include <linux/file.h> |
5590ff0d | 34 | #include <linux/fcntl.h> |
08ce5f16 | 35 | #include <linux/device_cgroup.h> |
5ad4e53b | 36 | #include <linux/fs_struct.h> |
e77819e5 | 37 | #include <linux/posix_acl.h> |
99d263d4 | 38 | #include <linux/hash.h> |
2a18da7a | 39 | #include <linux/bitops.h> |
aeaa4a79 | 40 | #include <linux/init_task.h> |
7c0f6ba6 | 41 | #include <linux/uaccess.h> |
1da177e4 | 42 | |
e81e3f4d | 43 | #include "internal.h" |
c7105365 | 44 | #include "mount.h" |
e81e3f4d | 45 | |
1da177e4 LT |
46 | /* [Feb-1997 T. Schoebel-Theuer] |
47 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
48 | * were necessary because of omirr. The reason is that omirr needs | |
49 | * to know the _real_ pathname, not the user-supplied one, in case | |
50 | * of symlinks (and also when transname replacements occur). | |
51 | * | |
52 | * The new code replaces the old recursive symlink resolution with | |
53 | * an iterative one (in case of non-nested symlink chains). It does | |
54 | * this with calls to <fs>_follow_link(). | |
55 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
56 | * replaced with a single function lookup_dentry() that can handle all | |
57 | * the special cases of the former code. | |
58 | * | |
59 | * With the new dcache, the pathname is stored at each inode, at least as | |
60 | * long as the refcount of the inode is positive. As a side effect, the | |
61 | * size of the dcache depends on the inode cache and thus is dynamic. | |
62 | * | |
63 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
64 | * resolution to correspond with current state of the code. | |
65 | * | |
66 | * Note that the symlink resolution is not *completely* iterative. | |
67 | * There is still a significant amount of tail- and mid- recursion in | |
68 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
69 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
70 | * may return different results than <fs>_follow_link(). Many virtual | |
71 | * filesystems (including /proc) exhibit this behavior. | |
72 | */ | |
73 | ||
74 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
75 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
76 | * and the name already exists in form of a symlink, try to create the new | |
77 | * name indicated by the symlink. The old code always complained that the | |
78 | * name already exists, due to not following the symlink even if its target | |
79 | * is nonexistent. The new semantics affects also mknod() and link() when | |
25985edc | 80 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
81 | * |
82 | * I don't know which semantics is the right one, since I have no access | |
83 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
84 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
85 | * "old" one. Personally, I think the new semantics is much more logical. | |
86 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
87 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
88 | * and in the old Linux semantics. | |
89 | */ | |
90 | ||
91 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
92 | * semantics. See the comments in "open_namei" and "do_link" below. | |
93 | * | |
94 | * [10-Sep-98 Alan Modra] Another symlink change. | |
95 | */ | |
96 | ||
97 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
98 | * inside the path - always follow. | |
99 | * in the last component in creation/removal/renaming - never follow. | |
100 | * if LOOKUP_FOLLOW passed - follow. | |
101 | * if the pathname has trailing slashes - follow. | |
102 | * otherwise - don't follow. | |
103 | * (applied in that order). | |
104 | * | |
105 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
106 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
107 | * During the 2.4 we need to fix the userland stuff depending on it - | |
108 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
109 | * XEmacs seems to be relying on it... | |
110 | */ | |
111 | /* | |
112 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 113 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
114 | * any extra contention... |
115 | */ | |
116 | ||
117 | /* In order to reduce some races, while at the same time doing additional | |
118 | * checking and hopefully speeding things up, we copy filenames to the | |
119 | * kernel data space before using them.. | |
120 | * | |
121 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
122 | * PATH_MAX includes the nul terminator --RR. | |
123 | */ | |
91a27b2a | 124 | |
fd2f7cb5 | 125 | #define EMBEDDED_NAME_MAX (PATH_MAX - offsetof(struct filename, iname)) |
7950e385 | 126 | |
51f39a1f | 127 | struct filename * |
91a27b2a JL |
128 | getname_flags(const char __user *filename, int flags, int *empty) |
129 | { | |
94b5d262 | 130 | struct filename *result; |
7950e385 | 131 | char *kname; |
94b5d262 | 132 | int len; |
4043cde8 | 133 | |
7ac86265 JL |
134 | result = audit_reusename(filename); |
135 | if (result) | |
136 | return result; | |
137 | ||
7950e385 | 138 | result = __getname(); |
3f9f0aa6 | 139 | if (unlikely(!result)) |
4043cde8 EP |
140 | return ERR_PTR(-ENOMEM); |
141 | ||
7950e385 JL |
142 | /* |
143 | * First, try to embed the struct filename inside the names_cache | |
144 | * allocation | |
145 | */ | |
fd2f7cb5 | 146 | kname = (char *)result->iname; |
91a27b2a | 147 | result->name = kname; |
7950e385 | 148 | |
94b5d262 | 149 | len = strncpy_from_user(kname, filename, EMBEDDED_NAME_MAX); |
91a27b2a | 150 | if (unlikely(len < 0)) { |
94b5d262 AV |
151 | __putname(result); |
152 | return ERR_PTR(len); | |
91a27b2a | 153 | } |
3f9f0aa6 | 154 | |
7950e385 JL |
155 | /* |
156 | * Uh-oh. We have a name that's approaching PATH_MAX. Allocate a | |
157 | * separate struct filename so we can dedicate the entire | |
158 | * names_cache allocation for the pathname, and re-do the copy from | |
159 | * userland. | |
160 | */ | |
94b5d262 | 161 | if (unlikely(len == EMBEDDED_NAME_MAX)) { |
fd2f7cb5 | 162 | const size_t size = offsetof(struct filename, iname[1]); |
7950e385 JL |
163 | kname = (char *)result; |
164 | ||
fd2f7cb5 AV |
165 | /* |
166 | * size is chosen that way we to guarantee that | |
167 | * result->iname[0] is within the same object and that | |
168 | * kname can't be equal to result->iname, no matter what. | |
169 | */ | |
170 | result = kzalloc(size, GFP_KERNEL); | |
94b5d262 AV |
171 | if (unlikely(!result)) { |
172 | __putname(kname); | |
173 | return ERR_PTR(-ENOMEM); | |
7950e385 JL |
174 | } |
175 | result->name = kname; | |
94b5d262 AV |
176 | len = strncpy_from_user(kname, filename, PATH_MAX); |
177 | if (unlikely(len < 0)) { | |
178 | __putname(kname); | |
179 | kfree(result); | |
180 | return ERR_PTR(len); | |
181 | } | |
182 | if (unlikely(len == PATH_MAX)) { | |
183 | __putname(kname); | |
184 | kfree(result); | |
185 | return ERR_PTR(-ENAMETOOLONG); | |
186 | } | |
7950e385 JL |
187 | } |
188 | ||
94b5d262 | 189 | result->refcnt = 1; |
3f9f0aa6 LT |
190 | /* The empty path is special. */ |
191 | if (unlikely(!len)) { | |
192 | if (empty) | |
4043cde8 | 193 | *empty = 1; |
94b5d262 AV |
194 | if (!(flags & LOOKUP_EMPTY)) { |
195 | putname(result); | |
196 | return ERR_PTR(-ENOENT); | |
197 | } | |
1da177e4 | 198 | } |
3f9f0aa6 | 199 | |
7950e385 | 200 | result->uptr = filename; |
c4ad8f98 | 201 | result->aname = NULL; |
7950e385 JL |
202 | audit_getname(result); |
203 | return result; | |
1da177e4 LT |
204 | } |
205 | ||
91a27b2a JL |
206 | struct filename * |
207 | getname(const char __user * filename) | |
f52e0c11 | 208 | { |
f7493e5d | 209 | return getname_flags(filename, 0, NULL); |
f52e0c11 AV |
210 | } |
211 | ||
c4ad8f98 LT |
212 | struct filename * |
213 | getname_kernel(const char * filename) | |
214 | { | |
215 | struct filename *result; | |
08518549 | 216 | int len = strlen(filename) + 1; |
c4ad8f98 LT |
217 | |
218 | result = __getname(); | |
219 | if (unlikely(!result)) | |
220 | return ERR_PTR(-ENOMEM); | |
221 | ||
08518549 | 222 | if (len <= EMBEDDED_NAME_MAX) { |
fd2f7cb5 | 223 | result->name = (char *)result->iname; |
08518549 PM |
224 | } else if (len <= PATH_MAX) { |
225 | struct filename *tmp; | |
226 | ||
227 | tmp = kmalloc(sizeof(*tmp), GFP_KERNEL); | |
228 | if (unlikely(!tmp)) { | |
229 | __putname(result); | |
230 | return ERR_PTR(-ENOMEM); | |
231 | } | |
232 | tmp->name = (char *)result; | |
08518549 PM |
233 | result = tmp; |
234 | } else { | |
235 | __putname(result); | |
236 | return ERR_PTR(-ENAMETOOLONG); | |
237 | } | |
238 | memcpy((char *)result->name, filename, len); | |
c4ad8f98 LT |
239 | result->uptr = NULL; |
240 | result->aname = NULL; | |
55422d0b | 241 | result->refcnt = 1; |
fd3522fd | 242 | audit_getname(result); |
c4ad8f98 | 243 | |
c4ad8f98 LT |
244 | return result; |
245 | } | |
246 | ||
91a27b2a | 247 | void putname(struct filename *name) |
1da177e4 | 248 | { |
55422d0b PM |
249 | BUG_ON(name->refcnt <= 0); |
250 | ||
251 | if (--name->refcnt > 0) | |
252 | return; | |
253 | ||
fd2f7cb5 | 254 | if (name->name != name->iname) { |
55422d0b PM |
255 | __putname(name->name); |
256 | kfree(name); | |
257 | } else | |
258 | __putname(name); | |
1da177e4 | 259 | } |
1da177e4 | 260 | |
e77819e5 LT |
261 | static int check_acl(struct inode *inode, int mask) |
262 | { | |
84635d68 | 263 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
264 | struct posix_acl *acl; |
265 | ||
e77819e5 | 266 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
267 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
268 | if (!acl) | |
e77819e5 | 269 | return -EAGAIN; |
3567866b | 270 | /* no ->get_acl() calls in RCU mode... */ |
b8a7a3a6 | 271 | if (is_uncached_acl(acl)) |
3567866b | 272 | return -ECHILD; |
206b1d09 | 273 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
274 | } |
275 | ||
2982baa2 CH |
276 | acl = get_acl(inode, ACL_TYPE_ACCESS); |
277 | if (IS_ERR(acl)) | |
278 | return PTR_ERR(acl); | |
e77819e5 LT |
279 | if (acl) { |
280 | int error = posix_acl_permission(inode, acl, mask); | |
281 | posix_acl_release(acl); | |
282 | return error; | |
283 | } | |
84635d68 | 284 | #endif |
e77819e5 LT |
285 | |
286 | return -EAGAIN; | |
287 | } | |
288 | ||
5909ccaa | 289 | /* |
948409c7 | 290 | * This does the basic permission checking |
1da177e4 | 291 | */ |
7e40145e | 292 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 293 | { |
26cf46be | 294 | unsigned int mode = inode->i_mode; |
1da177e4 | 295 | |
8e96e3b7 | 296 | if (likely(uid_eq(current_fsuid(), inode->i_uid))) |
1da177e4 LT |
297 | mode >>= 6; |
298 | else { | |
e77819e5 | 299 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 300 | int error = check_acl(inode, mask); |
b74c79e9 NP |
301 | if (error != -EAGAIN) |
302 | return error; | |
1da177e4 LT |
303 | } |
304 | ||
305 | if (in_group_p(inode->i_gid)) | |
306 | mode >>= 3; | |
307 | } | |
308 | ||
309 | /* | |
310 | * If the DACs are ok we don't need any capability check. | |
311 | */ | |
9c2c7039 | 312 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 313 | return 0; |
5909ccaa LT |
314 | return -EACCES; |
315 | } | |
316 | ||
317 | /** | |
b74c79e9 | 318 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 319 | * @inode: inode to check access rights for |
8fd90c8d | 320 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
321 | * |
322 | * Used to check for read/write/execute permissions on a file. | |
323 | * We use "fsuid" for this, letting us set arbitrary permissions | |
324 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
325 | * are used for other things. |
326 | * | |
327 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
328 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
329 | * It would then be called again in ref-walk mode. | |
5909ccaa | 330 | */ |
2830ba7f | 331 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
332 | { |
333 | int ret; | |
334 | ||
335 | /* | |
948409c7 | 336 | * Do the basic permission checks. |
5909ccaa | 337 | */ |
7e40145e | 338 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
339 | if (ret != -EACCES) |
340 | return ret; | |
1da177e4 | 341 | |
d594e7ec AV |
342 | if (S_ISDIR(inode->i_mode)) { |
343 | /* DACs are overridable for directories */ | |
d594e7ec | 344 | if (!(mask & MAY_WRITE)) |
23adbe12 AL |
345 | if (capable_wrt_inode_uidgid(inode, |
346 | CAP_DAC_READ_SEARCH)) | |
d594e7ec | 347 | return 0; |
23adbe12 | 348 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_OVERRIDE)) |
1da177e4 | 349 | return 0; |
2a4c2242 SS |
350 | return -EACCES; |
351 | } | |
1da177e4 LT |
352 | |
353 | /* | |
354 | * Searching includes executable on directories, else just read. | |
355 | */ | |
7ea66001 | 356 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 357 | if (mask == MAY_READ) |
23adbe12 | 358 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_READ_SEARCH)) |
1da177e4 | 359 | return 0; |
2a4c2242 SS |
360 | /* |
361 | * Read/write DACs are always overridable. | |
362 | * Executable DACs are overridable when there is | |
363 | * at least one exec bit set. | |
364 | */ | |
365 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) | |
366 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_OVERRIDE)) | |
367 | return 0; | |
1da177e4 LT |
368 | |
369 | return -EACCES; | |
370 | } | |
4d359507 | 371 | EXPORT_SYMBOL(generic_permission); |
1da177e4 | 372 | |
3ddcd056 LT |
373 | /* |
374 | * We _really_ want to just do "generic_permission()" without | |
375 | * even looking at the inode->i_op values. So we keep a cache | |
376 | * flag in inode->i_opflags, that says "this has not special | |
377 | * permission function, use the fast case". | |
378 | */ | |
379 | static inline int do_inode_permission(struct inode *inode, int mask) | |
380 | { | |
381 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
382 | if (likely(inode->i_op->permission)) | |
383 | return inode->i_op->permission(inode, mask); | |
384 | ||
385 | /* This gets set once for the inode lifetime */ | |
386 | spin_lock(&inode->i_lock); | |
387 | inode->i_opflags |= IOP_FASTPERM; | |
388 | spin_unlock(&inode->i_lock); | |
389 | } | |
390 | return generic_permission(inode, mask); | |
391 | } | |
392 | ||
0bdaea90 DH |
393 | /** |
394 | * sb_permission - Check superblock-level permissions | |
395 | * @sb: Superblock of inode to check permission on | |
55852635 | 396 | * @inode: Inode to check permission on |
0bdaea90 DH |
397 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) |
398 | * | |
399 | * Separate out file-system wide checks from inode-specific permission checks. | |
400 | */ | |
401 | static int sb_permission(struct super_block *sb, struct inode *inode, int mask) | |
402 | { | |
403 | if (unlikely(mask & MAY_WRITE)) { | |
404 | umode_t mode = inode->i_mode; | |
405 | ||
406 | /* Nobody gets write access to a read-only fs. */ | |
bc98a42c | 407 | if (sb_rdonly(sb) && (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) |
0bdaea90 DH |
408 | return -EROFS; |
409 | } | |
410 | return 0; | |
411 | } | |
412 | ||
413 | /** | |
414 | * inode_permission - Check for access rights to a given inode | |
415 | * @inode: Inode to check permission on | |
416 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
417 | * | |
418 | * Check for read/write/execute permissions on an inode. We use fs[ug]id for | |
419 | * this, letting us set arbitrary permissions for filesystem access without | |
420 | * changing the "normal" UIDs which are used for other things. | |
421 | * | |
422 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
423 | */ | |
424 | int inode_permission(struct inode *inode, int mask) | |
425 | { | |
426 | int retval; | |
427 | ||
428 | retval = sb_permission(inode->i_sb, inode, mask); | |
429 | if (retval) | |
430 | return retval; | |
4bfd054a EB |
431 | |
432 | if (unlikely(mask & MAY_WRITE)) { | |
433 | /* | |
434 | * Nobody gets write access to an immutable file. | |
435 | */ | |
436 | if (IS_IMMUTABLE(inode)) | |
437 | return -EPERM; | |
438 | ||
439 | /* | |
440 | * Updating mtime will likely cause i_uid and i_gid to be | |
441 | * written back improperly if their true value is unknown | |
442 | * to the vfs. | |
443 | */ | |
444 | if (HAS_UNMAPPED_ID(inode)) | |
445 | return -EACCES; | |
446 | } | |
447 | ||
448 | retval = do_inode_permission(inode, mask); | |
449 | if (retval) | |
450 | return retval; | |
451 | ||
452 | retval = devcgroup_inode_permission(inode, mask); | |
453 | if (retval) | |
454 | return retval; | |
455 | ||
456 | return security_inode_permission(inode, mask); | |
0bdaea90 | 457 | } |
4d359507 | 458 | EXPORT_SYMBOL(inode_permission); |
0bdaea90 | 459 | |
5dd784d0 JB |
460 | /** |
461 | * path_get - get a reference to a path | |
462 | * @path: path to get the reference to | |
463 | * | |
464 | * Given a path increment the reference count to the dentry and the vfsmount. | |
465 | */ | |
dcf787f3 | 466 | void path_get(const struct path *path) |
5dd784d0 JB |
467 | { |
468 | mntget(path->mnt); | |
469 | dget(path->dentry); | |
470 | } | |
471 | EXPORT_SYMBOL(path_get); | |
472 | ||
1d957f9b JB |
473 | /** |
474 | * path_put - put a reference to a path | |
475 | * @path: path to put the reference to | |
476 | * | |
477 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
478 | */ | |
dcf787f3 | 479 | void path_put(const struct path *path) |
1da177e4 | 480 | { |
1d957f9b JB |
481 | dput(path->dentry); |
482 | mntput(path->mnt); | |
1da177e4 | 483 | } |
1d957f9b | 484 | EXPORT_SYMBOL(path_put); |
1da177e4 | 485 | |
894bc8c4 | 486 | #define EMBEDDED_LEVELS 2 |
1f55a6ec AV |
487 | struct nameidata { |
488 | struct path path; | |
1cf2665b | 489 | struct qstr last; |
1f55a6ec AV |
490 | struct path root; |
491 | struct inode *inode; /* path.dentry.d_inode */ | |
492 | unsigned int flags; | |
9883d185 | 493 | unsigned seq, m_seq; |
1f55a6ec AV |
494 | int last_type; |
495 | unsigned depth; | |
756daf26 | 496 | int total_link_count; |
697fc6ca AV |
497 | struct saved { |
498 | struct path link; | |
fceef393 | 499 | struct delayed_call done; |
697fc6ca | 500 | const char *name; |
0450b2d1 | 501 | unsigned seq; |
894bc8c4 | 502 | } *stack, internal[EMBEDDED_LEVELS]; |
9883d185 AV |
503 | struct filename *name; |
504 | struct nameidata *saved; | |
fceef393 | 505 | struct inode *link_inode; |
9883d185 AV |
506 | unsigned root_seq; |
507 | int dfd; | |
3859a271 | 508 | } __randomize_layout; |
1f55a6ec | 509 | |
9883d185 | 510 | static void set_nameidata(struct nameidata *p, int dfd, struct filename *name) |
894bc8c4 | 511 | { |
756daf26 N |
512 | struct nameidata *old = current->nameidata; |
513 | p->stack = p->internal; | |
c8a53ee5 AV |
514 | p->dfd = dfd; |
515 | p->name = name; | |
756daf26 | 516 | p->total_link_count = old ? old->total_link_count : 0; |
9883d185 | 517 | p->saved = old; |
756daf26 | 518 | current->nameidata = p; |
894bc8c4 AV |
519 | } |
520 | ||
9883d185 | 521 | static void restore_nameidata(void) |
894bc8c4 | 522 | { |
9883d185 | 523 | struct nameidata *now = current->nameidata, *old = now->saved; |
756daf26 N |
524 | |
525 | current->nameidata = old; | |
526 | if (old) | |
527 | old->total_link_count = now->total_link_count; | |
e1a63bbc | 528 | if (now->stack != now->internal) |
756daf26 | 529 | kfree(now->stack); |
894bc8c4 AV |
530 | } |
531 | ||
532 | static int __nd_alloc_stack(struct nameidata *nd) | |
533 | { | |
bc40aee0 AV |
534 | struct saved *p; |
535 | ||
536 | if (nd->flags & LOOKUP_RCU) { | |
537 | p= kmalloc(MAXSYMLINKS * sizeof(struct saved), | |
538 | GFP_ATOMIC); | |
539 | if (unlikely(!p)) | |
540 | return -ECHILD; | |
541 | } else { | |
542 | p= kmalloc(MAXSYMLINKS * sizeof(struct saved), | |
894bc8c4 | 543 | GFP_KERNEL); |
bc40aee0 AV |
544 | if (unlikely(!p)) |
545 | return -ENOMEM; | |
546 | } | |
894bc8c4 AV |
547 | memcpy(p, nd->internal, sizeof(nd->internal)); |
548 | nd->stack = p; | |
549 | return 0; | |
550 | } | |
551 | ||
397d425d EB |
552 | /** |
553 | * path_connected - Verify that a path->dentry is below path->mnt.mnt_root | |
554 | * @path: nameidate to verify | |
555 | * | |
556 | * Rename can sometimes move a file or directory outside of a bind | |
557 | * mount, path_connected allows those cases to be detected. | |
558 | */ | |
559 | static bool path_connected(const struct path *path) | |
560 | { | |
561 | struct vfsmount *mnt = path->mnt; | |
562 | ||
563 | /* Only bind mounts can have disconnected paths */ | |
564 | if (mnt->mnt_root == mnt->mnt_sb->s_root) | |
565 | return true; | |
566 | ||
567 | return is_subdir(path->dentry, mnt->mnt_root); | |
568 | } | |
569 | ||
894bc8c4 AV |
570 | static inline int nd_alloc_stack(struct nameidata *nd) |
571 | { | |
da4e0be0 | 572 | if (likely(nd->depth != EMBEDDED_LEVELS)) |
894bc8c4 AV |
573 | return 0; |
574 | if (likely(nd->stack != nd->internal)) | |
575 | return 0; | |
576 | return __nd_alloc_stack(nd); | |
577 | } | |
578 | ||
7973387a AV |
579 | static void drop_links(struct nameidata *nd) |
580 | { | |
581 | int i = nd->depth; | |
582 | while (i--) { | |
583 | struct saved *last = nd->stack + i; | |
fceef393 AV |
584 | do_delayed_call(&last->done); |
585 | clear_delayed_call(&last->done); | |
7973387a AV |
586 | } |
587 | } | |
588 | ||
589 | static void terminate_walk(struct nameidata *nd) | |
590 | { | |
591 | drop_links(nd); | |
592 | if (!(nd->flags & LOOKUP_RCU)) { | |
593 | int i; | |
594 | path_put(&nd->path); | |
595 | for (i = 0; i < nd->depth; i++) | |
596 | path_put(&nd->stack[i].link); | |
102b8af2 AV |
597 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
598 | path_put(&nd->root); | |
599 | nd->root.mnt = NULL; | |
600 | } | |
7973387a AV |
601 | } else { |
602 | nd->flags &= ~LOOKUP_RCU; | |
603 | if (!(nd->flags & LOOKUP_ROOT)) | |
604 | nd->root.mnt = NULL; | |
605 | rcu_read_unlock(); | |
606 | } | |
607 | nd->depth = 0; | |
608 | } | |
609 | ||
610 | /* path_put is needed afterwards regardless of success or failure */ | |
611 | static bool legitimize_path(struct nameidata *nd, | |
612 | struct path *path, unsigned seq) | |
613 | { | |
614 | int res = __legitimize_mnt(path->mnt, nd->m_seq); | |
615 | if (unlikely(res)) { | |
616 | if (res > 0) | |
617 | path->mnt = NULL; | |
618 | path->dentry = NULL; | |
619 | return false; | |
620 | } | |
621 | if (unlikely(!lockref_get_not_dead(&path->dentry->d_lockref))) { | |
622 | path->dentry = NULL; | |
623 | return false; | |
624 | } | |
625 | return !read_seqcount_retry(&path->dentry->d_seq, seq); | |
626 | } | |
627 | ||
628 | static bool legitimize_links(struct nameidata *nd) | |
629 | { | |
630 | int i; | |
631 | for (i = 0; i < nd->depth; i++) { | |
632 | struct saved *last = nd->stack + i; | |
633 | if (unlikely(!legitimize_path(nd, &last->link, last->seq))) { | |
634 | drop_links(nd); | |
635 | nd->depth = i + 1; | |
636 | return false; | |
637 | } | |
638 | } | |
639 | return true; | |
640 | } | |
641 | ||
19660af7 | 642 | /* |
31e6b01f | 643 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
644 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
645 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
57e3715c | 646 | * normal reference counts on dentries and vfsmounts to transition to ref-walk |
19660af7 AV |
647 | * mode. Refcounts are grabbed at the last known good point before rcu-walk |
648 | * got stuck, so ref-walk may continue from there. If this is not successful | |
649 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
650 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 651 | */ |
31e6b01f NP |
652 | |
653 | /** | |
19660af7 AV |
654 | * unlazy_walk - try to switch to ref-walk mode. |
655 | * @nd: nameidata pathwalk data | |
39191628 | 656 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 657 | * |
4675ac39 AV |
658 | * unlazy_walk attempts to legitimize the current nd->path and nd->root |
659 | * for ref-walk mode. | |
660 | * Must be called from rcu-walk context. | |
7973387a AV |
661 | * Nothing should touch nameidata between unlazy_walk() failure and |
662 | * terminate_walk(). | |
31e6b01f | 663 | */ |
4675ac39 | 664 | static int unlazy_walk(struct nameidata *nd) |
31e6b01f | 665 | { |
31e6b01f NP |
666 | struct dentry *parent = nd->path.dentry; |
667 | ||
668 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
e5c832d5 | 669 | |
4675ac39 AV |
670 | nd->flags &= ~LOOKUP_RCU; |
671 | if (unlikely(!legitimize_links(nd))) | |
672 | goto out2; | |
673 | if (unlikely(!legitimize_path(nd, &nd->path, nd->seq))) | |
674 | goto out1; | |
675 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
676 | if (unlikely(!legitimize_path(nd, &nd->root, nd->root_seq))) | |
677 | goto out; | |
678 | } | |
679 | rcu_read_unlock(); | |
680 | BUG_ON(nd->inode != parent->d_inode); | |
681 | return 0; | |
682 | ||
683 | out2: | |
684 | nd->path.mnt = NULL; | |
685 | nd->path.dentry = NULL; | |
686 | out1: | |
687 | if (!(nd->flags & LOOKUP_ROOT)) | |
688 | nd->root.mnt = NULL; | |
689 | out: | |
690 | rcu_read_unlock(); | |
691 | return -ECHILD; | |
692 | } | |
693 | ||
694 | /** | |
695 | * unlazy_child - try to switch to ref-walk mode. | |
696 | * @nd: nameidata pathwalk data | |
697 | * @dentry: child of nd->path.dentry | |
698 | * @seq: seq number to check dentry against | |
699 | * Returns: 0 on success, -ECHILD on failure | |
700 | * | |
701 | * unlazy_child attempts to legitimize the current nd->path, nd->root and dentry | |
702 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
703 | * @nd. Must be called from rcu-walk context. | |
704 | * Nothing should touch nameidata between unlazy_child() failure and | |
705 | * terminate_walk(). | |
706 | */ | |
707 | static int unlazy_child(struct nameidata *nd, struct dentry *dentry, unsigned seq) | |
708 | { | |
709 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
710 | ||
e5c832d5 | 711 | nd->flags &= ~LOOKUP_RCU; |
7973387a AV |
712 | if (unlikely(!legitimize_links(nd))) |
713 | goto out2; | |
714 | if (unlikely(!legitimize_mnt(nd->path.mnt, nd->m_seq))) | |
715 | goto out2; | |
4675ac39 | 716 | if (unlikely(!lockref_get_not_dead(&nd->path.dentry->d_lockref))) |
7973387a | 717 | goto out1; |
48a066e7 | 718 | |
15570086 | 719 | /* |
4675ac39 AV |
720 | * We need to move both the parent and the dentry from the RCU domain |
721 | * to be properly refcounted. And the sequence number in the dentry | |
722 | * validates *both* dentry counters, since we checked the sequence | |
723 | * number of the parent after we got the child sequence number. So we | |
724 | * know the parent must still be valid if the child sequence number is | |
15570086 | 725 | */ |
4675ac39 AV |
726 | if (unlikely(!lockref_get_not_dead(&dentry->d_lockref))) |
727 | goto out; | |
728 | if (unlikely(read_seqcount_retry(&dentry->d_seq, seq))) { | |
729 | rcu_read_unlock(); | |
730 | dput(dentry); | |
731 | goto drop_root_mnt; | |
19660af7 | 732 | } |
e5c832d5 LT |
733 | /* |
734 | * Sequence counts matched. Now make sure that the root is | |
735 | * still valid and get it if required. | |
736 | */ | |
737 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
5a8d87e8 AV |
738 | if (unlikely(!legitimize_path(nd, &nd->root, nd->root_seq))) { |
739 | rcu_read_unlock(); | |
740 | dput(dentry); | |
741 | return -ECHILD; | |
7973387a | 742 | } |
31e6b01f | 743 | } |
31e6b01f | 744 | |
8b61e74f | 745 | rcu_read_unlock(); |
31e6b01f | 746 | return 0; |
19660af7 | 747 | |
7973387a AV |
748 | out2: |
749 | nd->path.mnt = NULL; | |
750 | out1: | |
751 | nd->path.dentry = NULL; | |
e5c832d5 | 752 | out: |
8b61e74f | 753 | rcu_read_unlock(); |
d0d27277 LT |
754 | drop_root_mnt: |
755 | if (!(nd->flags & LOOKUP_ROOT)) | |
756 | nd->root.mnt = NULL; | |
31e6b01f NP |
757 | return -ECHILD; |
758 | } | |
759 | ||
4ce16ef3 | 760 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 761 | { |
a89f8337 AV |
762 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) |
763 | return dentry->d_op->d_revalidate(dentry, flags); | |
764 | else | |
765 | return 1; | |
34286d66 NP |
766 | } |
767 | ||
9f1fafee AV |
768 | /** |
769 | * complete_walk - successful completion of path walk | |
770 | * @nd: pointer nameidata | |
39159de2 | 771 | * |
9f1fafee AV |
772 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
773 | * Revalidate the final result, unless we'd already done that during | |
774 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
775 | * success, -error on failure. In case of failure caller does not | |
776 | * need to drop nd->path. | |
39159de2 | 777 | */ |
9f1fafee | 778 | static int complete_walk(struct nameidata *nd) |
39159de2 | 779 | { |
16c2cd71 | 780 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 781 | int status; |
39159de2 | 782 | |
9f1fafee | 783 | if (nd->flags & LOOKUP_RCU) { |
9f1fafee AV |
784 | if (!(nd->flags & LOOKUP_ROOT)) |
785 | nd->root.mnt = NULL; | |
4675ac39 | 786 | if (unlikely(unlazy_walk(nd))) |
9f1fafee | 787 | return -ECHILD; |
9f1fafee AV |
788 | } |
789 | ||
16c2cd71 AV |
790 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
791 | return 0; | |
792 | ||
ecf3d1f1 | 793 | if (likely(!(dentry->d_flags & DCACHE_OP_WEAK_REVALIDATE))) |
39159de2 JL |
794 | return 0; |
795 | ||
ecf3d1f1 | 796 | status = dentry->d_op->d_weak_revalidate(dentry, nd->flags); |
39159de2 JL |
797 | if (status > 0) |
798 | return 0; | |
799 | ||
16c2cd71 | 800 | if (!status) |
39159de2 | 801 | status = -ESTALE; |
16c2cd71 | 802 | |
39159de2 JL |
803 | return status; |
804 | } | |
805 | ||
18d8c860 | 806 | static void set_root(struct nameidata *nd) |
31e6b01f | 807 | { |
7bd88377 | 808 | struct fs_struct *fs = current->fs; |
c28cc364 | 809 | |
9e6697e2 AV |
810 | if (nd->flags & LOOKUP_RCU) { |
811 | unsigned seq; | |
812 | ||
813 | do { | |
814 | seq = read_seqcount_begin(&fs->seq); | |
815 | nd->root = fs->root; | |
816 | nd->root_seq = __read_seqcount_begin(&nd->root.dentry->d_seq); | |
817 | } while (read_seqcount_retry(&fs->seq, seq)); | |
818 | } else { | |
819 | get_fs_root(fs, &nd->root); | |
820 | } | |
31e6b01f NP |
821 | } |
822 | ||
1d957f9b | 823 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
824 | { |
825 | dput(path->dentry); | |
4ac91378 | 826 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
827 | mntput(path->mnt); |
828 | } | |
829 | ||
7b9337aa NP |
830 | static inline void path_to_nameidata(const struct path *path, |
831 | struct nameidata *nd) | |
051d3812 | 832 | { |
31e6b01f NP |
833 | if (!(nd->flags & LOOKUP_RCU)) { |
834 | dput(nd->path.dentry); | |
835 | if (nd->path.mnt != path->mnt) | |
836 | mntput(nd->path.mnt); | |
9a229683 | 837 | } |
31e6b01f | 838 | nd->path.mnt = path->mnt; |
4ac91378 | 839 | nd->path.dentry = path->dentry; |
051d3812 IK |
840 | } |
841 | ||
248fb5b9 AV |
842 | static int nd_jump_root(struct nameidata *nd) |
843 | { | |
844 | if (nd->flags & LOOKUP_RCU) { | |
845 | struct dentry *d; | |
846 | nd->path = nd->root; | |
847 | d = nd->path.dentry; | |
848 | nd->inode = d->d_inode; | |
849 | nd->seq = nd->root_seq; | |
850 | if (unlikely(read_seqcount_retry(&d->d_seq, nd->seq))) | |
851 | return -ECHILD; | |
852 | } else { | |
853 | path_put(&nd->path); | |
854 | nd->path = nd->root; | |
855 | path_get(&nd->path); | |
856 | nd->inode = nd->path.dentry->d_inode; | |
857 | } | |
858 | nd->flags |= LOOKUP_JUMPED; | |
859 | return 0; | |
860 | } | |
861 | ||
b5fb63c1 | 862 | /* |
6b255391 | 863 | * Helper to directly jump to a known parsed path from ->get_link, |
b5fb63c1 CH |
864 | * caller must have taken a reference to path beforehand. |
865 | */ | |
6e77137b | 866 | void nd_jump_link(struct path *path) |
b5fb63c1 | 867 | { |
6e77137b | 868 | struct nameidata *nd = current->nameidata; |
b5fb63c1 CH |
869 | path_put(&nd->path); |
870 | ||
871 | nd->path = *path; | |
872 | nd->inode = nd->path.dentry->d_inode; | |
873 | nd->flags |= LOOKUP_JUMPED; | |
b5fb63c1 CH |
874 | } |
875 | ||
b9ff4429 | 876 | static inline void put_link(struct nameidata *nd) |
574197e0 | 877 | { |
21c3003d | 878 | struct saved *last = nd->stack + --nd->depth; |
fceef393 | 879 | do_delayed_call(&last->done); |
6548fae2 AV |
880 | if (!(nd->flags & LOOKUP_RCU)) |
881 | path_put(&last->link); | |
574197e0 AV |
882 | } |
883 | ||
561ec64a LT |
884 | int sysctl_protected_symlinks __read_mostly = 0; |
885 | int sysctl_protected_hardlinks __read_mostly = 0; | |
800179c9 KC |
886 | |
887 | /** | |
888 | * may_follow_link - Check symlink following for unsafe situations | |
55852635 | 889 | * @nd: nameidata pathwalk data |
800179c9 KC |
890 | * |
891 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
892 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
893 | * in a sticky world-writable directory. This is to protect privileged | |
894 | * processes from failing races against path names that may change out | |
895 | * from under them by way of other users creating malicious symlinks. | |
896 | * It will permit symlinks to be followed only when outside a sticky | |
897 | * world-writable directory, or when the uid of the symlink and follower | |
898 | * match, or when the directory owner matches the symlink's owner. | |
899 | * | |
900 | * Returns 0 if following the symlink is allowed, -ve on error. | |
901 | */ | |
fec2fa24 | 902 | static inline int may_follow_link(struct nameidata *nd) |
800179c9 KC |
903 | { |
904 | const struct inode *inode; | |
905 | const struct inode *parent; | |
2d7f9e2a | 906 | kuid_t puid; |
800179c9 KC |
907 | |
908 | if (!sysctl_protected_symlinks) | |
909 | return 0; | |
910 | ||
911 | /* Allowed if owner and follower match. */ | |
fceef393 | 912 | inode = nd->link_inode; |
81abe27b | 913 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
914 | return 0; |
915 | ||
916 | /* Allowed if parent directory not sticky and world-writable. */ | |
aa65fa35 | 917 | parent = nd->inode; |
800179c9 KC |
918 | if ((parent->i_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) |
919 | return 0; | |
920 | ||
921 | /* Allowed if parent directory and link owner match. */ | |
2d7f9e2a SF |
922 | puid = parent->i_uid; |
923 | if (uid_valid(puid) && uid_eq(puid, inode->i_uid)) | |
800179c9 KC |
924 | return 0; |
925 | ||
31956502 AV |
926 | if (nd->flags & LOOKUP_RCU) |
927 | return -ECHILD; | |
928 | ||
1cf2665b | 929 | audit_log_link_denied("follow_link", &nd->stack[0].link); |
800179c9 KC |
930 | return -EACCES; |
931 | } | |
932 | ||
933 | /** | |
934 | * safe_hardlink_source - Check for safe hardlink conditions | |
935 | * @inode: the source inode to hardlink from | |
936 | * | |
937 | * Return false if at least one of the following conditions: | |
938 | * - inode is not a regular file | |
939 | * - inode is setuid | |
940 | * - inode is setgid and group-exec | |
941 | * - access failure for read and write | |
942 | * | |
943 | * Otherwise returns true. | |
944 | */ | |
945 | static bool safe_hardlink_source(struct inode *inode) | |
946 | { | |
947 | umode_t mode = inode->i_mode; | |
948 | ||
949 | /* Special files should not get pinned to the filesystem. */ | |
950 | if (!S_ISREG(mode)) | |
951 | return false; | |
952 | ||
953 | /* Setuid files should not get pinned to the filesystem. */ | |
954 | if (mode & S_ISUID) | |
955 | return false; | |
956 | ||
957 | /* Executable setgid files should not get pinned to the filesystem. */ | |
958 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
959 | return false; | |
960 | ||
961 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
962 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
963 | return false; | |
964 | ||
965 | return true; | |
966 | } | |
967 | ||
968 | /** | |
969 | * may_linkat - Check permissions for creating a hardlink | |
970 | * @link: the source to hardlink from | |
971 | * | |
972 | * Block hardlink when all of: | |
973 | * - sysctl_protected_hardlinks enabled | |
974 | * - fsuid does not match inode | |
975 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
f2ca3796 | 976 | * - not CAP_FOWNER in a namespace with the inode owner uid mapped |
800179c9 KC |
977 | * |
978 | * Returns 0 if successful, -ve on error. | |
979 | */ | |
980 | static int may_linkat(struct path *link) | |
981 | { | |
800179c9 KC |
982 | struct inode *inode; |
983 | ||
984 | if (!sysctl_protected_hardlinks) | |
985 | return 0; | |
986 | ||
800179c9 KC |
987 | inode = link->dentry->d_inode; |
988 | ||
989 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, | |
990 | * otherwise, it must be a safe source. | |
991 | */ | |
cc658db4 | 992 | if (safe_hardlink_source(inode) || inode_owner_or_capable(inode)) |
800179c9 KC |
993 | return 0; |
994 | ||
a51d9eaa | 995 | audit_log_link_denied("linkat", link); |
800179c9 KC |
996 | return -EPERM; |
997 | } | |
998 | ||
3b2e7f75 AV |
999 | static __always_inline |
1000 | const char *get_link(struct nameidata *nd) | |
1da177e4 | 1001 | { |
ab104923 | 1002 | struct saved *last = nd->stack + nd->depth - 1; |
1cf2665b | 1003 | struct dentry *dentry = last->link.dentry; |
fceef393 | 1004 | struct inode *inode = nd->link_inode; |
6d7b5aae | 1005 | int error; |
0a959df5 | 1006 | const char *res; |
1da177e4 | 1007 | |
8fa9dd24 N |
1008 | if (!(nd->flags & LOOKUP_RCU)) { |
1009 | touch_atime(&last->link); | |
1010 | cond_resched(); | |
598e3c8f | 1011 | } else if (atime_needs_update_rcu(&last->link, inode)) { |
4675ac39 | 1012 | if (unlikely(unlazy_walk(nd))) |
bc40aee0 | 1013 | return ERR_PTR(-ECHILD); |
8fa9dd24 | 1014 | touch_atime(&last->link); |
bc40aee0 | 1015 | } |
574197e0 | 1016 | |
bda0be7a N |
1017 | error = security_inode_follow_link(dentry, inode, |
1018 | nd->flags & LOOKUP_RCU); | |
1019 | if (unlikely(error)) | |
6920a440 | 1020 | return ERR_PTR(error); |
36f3b4f6 | 1021 | |
86acdca1 | 1022 | nd->last_type = LAST_BIND; |
d4dee48b AV |
1023 | res = inode->i_link; |
1024 | if (!res) { | |
fceef393 AV |
1025 | const char * (*get)(struct dentry *, struct inode *, |
1026 | struct delayed_call *); | |
1027 | get = inode->i_op->get_link; | |
8c1b4566 | 1028 | if (nd->flags & LOOKUP_RCU) { |
fceef393 | 1029 | res = get(NULL, inode, &last->done); |
6b255391 | 1030 | if (res == ERR_PTR(-ECHILD)) { |
4675ac39 | 1031 | if (unlikely(unlazy_walk(nd))) |
6b255391 | 1032 | return ERR_PTR(-ECHILD); |
fceef393 | 1033 | res = get(dentry, inode, &last->done); |
6b255391 AV |
1034 | } |
1035 | } else { | |
fceef393 | 1036 | res = get(dentry, inode, &last->done); |
8c1b4566 | 1037 | } |
fceef393 | 1038 | if (IS_ERR_OR_NULL(res)) |
fab51e8a | 1039 | return res; |
fab51e8a AV |
1040 | } |
1041 | if (*res == '/') { | |
9e6697e2 AV |
1042 | if (!nd->root.mnt) |
1043 | set_root(nd); | |
248fb5b9 AV |
1044 | if (unlikely(nd_jump_root(nd))) |
1045 | return ERR_PTR(-ECHILD); | |
fab51e8a AV |
1046 | while (unlikely(*++res == '/')) |
1047 | ; | |
1da177e4 | 1048 | } |
fab51e8a AV |
1049 | if (!*res) |
1050 | res = NULL; | |
0a959df5 AV |
1051 | return res; |
1052 | } | |
6d7b5aae | 1053 | |
f015f126 DH |
1054 | /* |
1055 | * follow_up - Find the mountpoint of path's vfsmount | |
1056 | * | |
1057 | * Given a path, find the mountpoint of its source file system. | |
1058 | * Replace @path with the path of the mountpoint in the parent mount. | |
1059 | * Up is towards /. | |
1060 | * | |
1061 | * Return 1 if we went up a level and 0 if we were already at the | |
1062 | * root. | |
1063 | */ | |
bab77ebf | 1064 | int follow_up(struct path *path) |
1da177e4 | 1065 | { |
0714a533 AV |
1066 | struct mount *mnt = real_mount(path->mnt); |
1067 | struct mount *parent; | |
1da177e4 | 1068 | struct dentry *mountpoint; |
99b7db7b | 1069 | |
48a066e7 | 1070 | read_seqlock_excl(&mount_lock); |
0714a533 | 1071 | parent = mnt->mnt_parent; |
3c0a6163 | 1072 | if (parent == mnt) { |
48a066e7 | 1073 | read_sequnlock_excl(&mount_lock); |
1da177e4 LT |
1074 | return 0; |
1075 | } | |
0714a533 | 1076 | mntget(&parent->mnt); |
a73324da | 1077 | mountpoint = dget(mnt->mnt_mountpoint); |
48a066e7 | 1078 | read_sequnlock_excl(&mount_lock); |
bab77ebf AV |
1079 | dput(path->dentry); |
1080 | path->dentry = mountpoint; | |
1081 | mntput(path->mnt); | |
0714a533 | 1082 | path->mnt = &parent->mnt; |
1da177e4 LT |
1083 | return 1; |
1084 | } | |
4d359507 | 1085 | EXPORT_SYMBOL(follow_up); |
1da177e4 | 1086 | |
b5c84bf6 | 1087 | /* |
9875cf80 DH |
1088 | * Perform an automount |
1089 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
1090 | * were called with. | |
1da177e4 | 1091 | */ |
756daf26 | 1092 | static int follow_automount(struct path *path, struct nameidata *nd, |
9875cf80 | 1093 | bool *need_mntput) |
31e6b01f | 1094 | { |
9875cf80 | 1095 | struct vfsmount *mnt; |
ea5b778a | 1096 | int err; |
9875cf80 DH |
1097 | |
1098 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
1099 | return -EREMOTE; | |
1100 | ||
0ec26fd0 MS |
1101 | /* We don't want to mount if someone's just doing a stat - |
1102 | * unless they're stat'ing a directory and appended a '/' to | |
1103 | * the name. | |
1104 | * | |
1105 | * We do, however, want to mount if someone wants to open or | |
1106 | * create a file of any type under the mountpoint, wants to | |
1107 | * traverse through the mountpoint or wants to open the | |
1108 | * mounted directory. Also, autofs may mark negative dentries | |
1109 | * as being automount points. These will need the attentions | |
1110 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 1111 | */ |
756daf26 | 1112 | if (!(nd->flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
5d38f049 IK |
1113 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
1114 | path->dentry->d_inode) | |
1115 | return -EISDIR; | |
0ec26fd0 | 1116 | |
756daf26 N |
1117 | nd->total_link_count++; |
1118 | if (nd->total_link_count >= 40) | |
9875cf80 DH |
1119 | return -ELOOP; |
1120 | ||
1121 | mnt = path->dentry->d_op->d_automount(path); | |
1122 | if (IS_ERR(mnt)) { | |
1123 | /* | |
1124 | * The filesystem is allowed to return -EISDIR here to indicate | |
1125 | * it doesn't want to automount. For instance, autofs would do | |
1126 | * this so that its userspace daemon can mount on this dentry. | |
1127 | * | |
1128 | * However, we can only permit this if it's a terminal point in | |
1129 | * the path being looked up; if it wasn't then the remainder of | |
1130 | * the path is inaccessible and we should say so. | |
1131 | */ | |
756daf26 | 1132 | if (PTR_ERR(mnt) == -EISDIR && (nd->flags & LOOKUP_PARENT)) |
9875cf80 DH |
1133 | return -EREMOTE; |
1134 | return PTR_ERR(mnt); | |
31e6b01f | 1135 | } |
ea5b778a | 1136 | |
9875cf80 DH |
1137 | if (!mnt) /* mount collision */ |
1138 | return 0; | |
31e6b01f | 1139 | |
8aef1884 AV |
1140 | if (!*need_mntput) { |
1141 | /* lock_mount() may release path->mnt on error */ | |
1142 | mntget(path->mnt); | |
1143 | *need_mntput = true; | |
1144 | } | |
19a167af | 1145 | err = finish_automount(mnt, path); |
9875cf80 | 1146 | |
ea5b778a DH |
1147 | switch (err) { |
1148 | case -EBUSY: | |
1149 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 1150 | return 0; |
ea5b778a | 1151 | case 0: |
8aef1884 | 1152 | path_put(path); |
ea5b778a DH |
1153 | path->mnt = mnt; |
1154 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 1155 | return 0; |
19a167af AV |
1156 | default: |
1157 | return err; | |
ea5b778a | 1158 | } |
19a167af | 1159 | |
463ffb2e AV |
1160 | } |
1161 | ||
9875cf80 DH |
1162 | /* |
1163 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 1164 | * - Flagged for transit management (autofs) |
9875cf80 DH |
1165 | * - Flagged as mountpoint |
1166 | * - Flagged as automount point | |
1167 | * | |
1168 | * This may only be called in refwalk mode. | |
1169 | * | |
1170 | * Serialization is taken care of in namespace.c | |
1171 | */ | |
756daf26 | 1172 | static int follow_managed(struct path *path, struct nameidata *nd) |
1da177e4 | 1173 | { |
8aef1884 | 1174 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
1175 | unsigned managed; |
1176 | bool need_mntput = false; | |
8aef1884 | 1177 | int ret = 0; |
9875cf80 DH |
1178 | |
1179 | /* Given that we're not holding a lock here, we retain the value in a | |
1180 | * local variable for each dentry as we look at it so that we don't see | |
1181 | * the components of that value change under us */ | |
6aa7de05 | 1182 | while (managed = READ_ONCE(path->dentry->d_flags), |
9875cf80 DH |
1183 | managed &= DCACHE_MANAGED_DENTRY, |
1184 | unlikely(managed != 0)) { | |
cc53ce53 DH |
1185 | /* Allow the filesystem to manage the transit without i_mutex |
1186 | * being held. */ | |
1187 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1188 | BUG_ON(!path->dentry->d_op); | |
1189 | BUG_ON(!path->dentry->d_op->d_manage); | |
fb5f51c7 | 1190 | ret = path->dentry->d_op->d_manage(path, false); |
cc53ce53 | 1191 | if (ret < 0) |
8aef1884 | 1192 | break; |
cc53ce53 DH |
1193 | } |
1194 | ||
9875cf80 DH |
1195 | /* Transit to a mounted filesystem. */ |
1196 | if (managed & DCACHE_MOUNTED) { | |
1197 | struct vfsmount *mounted = lookup_mnt(path); | |
1198 | if (mounted) { | |
1199 | dput(path->dentry); | |
1200 | if (need_mntput) | |
1201 | mntput(path->mnt); | |
1202 | path->mnt = mounted; | |
1203 | path->dentry = dget(mounted->mnt_root); | |
1204 | need_mntput = true; | |
1205 | continue; | |
1206 | } | |
1207 | ||
1208 | /* Something is mounted on this dentry in another | |
1209 | * namespace and/or whatever was mounted there in this | |
48a066e7 AV |
1210 | * namespace got unmounted before lookup_mnt() could |
1211 | * get it */ | |
9875cf80 DH |
1212 | } |
1213 | ||
1214 | /* Handle an automount point */ | |
1215 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
756daf26 | 1216 | ret = follow_automount(path, nd, &need_mntput); |
9875cf80 | 1217 | if (ret < 0) |
8aef1884 | 1218 | break; |
9875cf80 DH |
1219 | continue; |
1220 | } | |
1221 | ||
1222 | /* We didn't change the current path point */ | |
1223 | break; | |
1da177e4 | 1224 | } |
8aef1884 AV |
1225 | |
1226 | if (need_mntput && path->mnt == mnt) | |
1227 | mntput(path->mnt); | |
e9742b53 AV |
1228 | if (ret == -EISDIR || !ret) |
1229 | ret = 1; | |
8402752e AV |
1230 | if (need_mntput) |
1231 | nd->flags |= LOOKUP_JUMPED; | |
1232 | if (unlikely(ret < 0)) | |
1233 | path_put_conditional(path, nd); | |
1234 | return ret; | |
1da177e4 LT |
1235 | } |
1236 | ||
cc53ce53 | 1237 | int follow_down_one(struct path *path) |
1da177e4 LT |
1238 | { |
1239 | struct vfsmount *mounted; | |
1240 | ||
1c755af4 | 1241 | mounted = lookup_mnt(path); |
1da177e4 | 1242 | if (mounted) { |
9393bd07 AV |
1243 | dput(path->dentry); |
1244 | mntput(path->mnt); | |
1245 | path->mnt = mounted; | |
1246 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1247 | return 1; |
1248 | } | |
1249 | return 0; | |
1250 | } | |
4d359507 | 1251 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 | 1252 | |
fb5f51c7 | 1253 | static inline int managed_dentry_rcu(const struct path *path) |
62a7375e | 1254 | { |
fb5f51c7 IK |
1255 | return (path->dentry->d_flags & DCACHE_MANAGE_TRANSIT) ? |
1256 | path->dentry->d_op->d_manage(path, true) : 0; | |
62a7375e IK |
1257 | } |
1258 | ||
9875cf80 | 1259 | /* |
287548e4 AV |
1260 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1261 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1262 | */ |
1263 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
254cf582 | 1264 | struct inode **inode, unsigned *seqp) |
9875cf80 | 1265 | { |
62a7375e | 1266 | for (;;) { |
c7105365 | 1267 | struct mount *mounted; |
62a7375e IK |
1268 | /* |
1269 | * Don't forget we might have a non-mountpoint managed dentry | |
1270 | * that wants to block transit. | |
1271 | */ | |
fb5f51c7 | 1272 | switch (managed_dentry_rcu(path)) { |
b8faf035 N |
1273 | case -ECHILD: |
1274 | default: | |
ab90911f | 1275 | return false; |
b8faf035 N |
1276 | case -EISDIR: |
1277 | return true; | |
1278 | case 0: | |
1279 | break; | |
1280 | } | |
62a7375e IK |
1281 | |
1282 | if (!d_mountpoint(path->dentry)) | |
b8faf035 | 1283 | return !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
62a7375e | 1284 | |
474279dc | 1285 | mounted = __lookup_mnt(path->mnt, path->dentry); |
9875cf80 DH |
1286 | if (!mounted) |
1287 | break; | |
c7105365 AV |
1288 | path->mnt = &mounted->mnt; |
1289 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1290 | nd->flags |= LOOKUP_JUMPED; |
254cf582 | 1291 | *seqp = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1292 | /* |
1293 | * Update the inode too. We don't need to re-check the | |
1294 | * dentry sequence number here after this d_inode read, | |
1295 | * because a mount-point is always pinned. | |
1296 | */ | |
1297 | *inode = path->dentry->d_inode; | |
9875cf80 | 1298 | } |
f5be3e29 | 1299 | return !read_seqretry(&mount_lock, nd->m_seq) && |
b8faf035 | 1300 | !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
287548e4 AV |
1301 | } |
1302 | ||
31e6b01f NP |
1303 | static int follow_dotdot_rcu(struct nameidata *nd) |
1304 | { | |
4023bfc9 | 1305 | struct inode *inode = nd->inode; |
31e6b01f | 1306 | |
9875cf80 | 1307 | while (1) { |
aed434ad | 1308 | if (path_equal(&nd->path, &nd->root)) |
31e6b01f | 1309 | break; |
31e6b01f NP |
1310 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
1311 | struct dentry *old = nd->path.dentry; | |
1312 | struct dentry *parent = old->d_parent; | |
1313 | unsigned seq; | |
1314 | ||
4023bfc9 | 1315 | inode = parent->d_inode; |
31e6b01f | 1316 | seq = read_seqcount_begin(&parent->d_seq); |
aed434ad AV |
1317 | if (unlikely(read_seqcount_retry(&old->d_seq, nd->seq))) |
1318 | return -ECHILD; | |
31e6b01f NP |
1319 | nd->path.dentry = parent; |
1320 | nd->seq = seq; | |
397d425d EB |
1321 | if (unlikely(!path_connected(&nd->path))) |
1322 | return -ENOENT; | |
31e6b01f | 1323 | break; |
aed434ad AV |
1324 | } else { |
1325 | struct mount *mnt = real_mount(nd->path.mnt); | |
1326 | struct mount *mparent = mnt->mnt_parent; | |
1327 | struct dentry *mountpoint = mnt->mnt_mountpoint; | |
1328 | struct inode *inode2 = mountpoint->d_inode; | |
1329 | unsigned seq = read_seqcount_begin(&mountpoint->d_seq); | |
1330 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) | |
1331 | return -ECHILD; | |
1332 | if (&mparent->mnt == nd->path.mnt) | |
1333 | break; | |
1334 | /* we know that mountpoint was pinned */ | |
1335 | nd->path.dentry = mountpoint; | |
1336 | nd->path.mnt = &mparent->mnt; | |
1337 | inode = inode2; | |
1338 | nd->seq = seq; | |
31e6b01f | 1339 | } |
31e6b01f | 1340 | } |
aed434ad | 1341 | while (unlikely(d_mountpoint(nd->path.dentry))) { |
b37199e6 AV |
1342 | struct mount *mounted; |
1343 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry); | |
aed434ad AV |
1344 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) |
1345 | return -ECHILD; | |
b37199e6 AV |
1346 | if (!mounted) |
1347 | break; | |
1348 | nd->path.mnt = &mounted->mnt; | |
1349 | nd->path.dentry = mounted->mnt.mnt_root; | |
4023bfc9 | 1350 | inode = nd->path.dentry->d_inode; |
b37199e6 | 1351 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
b37199e6 | 1352 | } |
4023bfc9 | 1353 | nd->inode = inode; |
31e6b01f NP |
1354 | return 0; |
1355 | } | |
1356 | ||
cc53ce53 DH |
1357 | /* |
1358 | * Follow down to the covering mount currently visible to userspace. At each | |
1359 | * point, the filesystem owning that dentry may be queried as to whether the | |
1360 | * caller is permitted to proceed or not. | |
cc53ce53 | 1361 | */ |
7cc90cc3 | 1362 | int follow_down(struct path *path) |
cc53ce53 DH |
1363 | { |
1364 | unsigned managed; | |
1365 | int ret; | |
1366 | ||
6aa7de05 | 1367 | while (managed = READ_ONCE(path->dentry->d_flags), |
cc53ce53 DH |
1368 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { |
1369 | /* Allow the filesystem to manage the transit without i_mutex | |
1370 | * being held. | |
1371 | * | |
1372 | * We indicate to the filesystem if someone is trying to mount | |
1373 | * something here. This gives autofs the chance to deny anyone | |
1374 | * other than its daemon the right to mount on its | |
1375 | * superstructure. | |
1376 | * | |
1377 | * The filesystem may sleep at this point. | |
1378 | */ | |
1379 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1380 | BUG_ON(!path->dentry->d_op); | |
1381 | BUG_ON(!path->dentry->d_op->d_manage); | |
fb5f51c7 | 1382 | ret = path->dentry->d_op->d_manage(path, false); |
cc53ce53 DH |
1383 | if (ret < 0) |
1384 | return ret == -EISDIR ? 0 : ret; | |
1385 | } | |
1386 | ||
1387 | /* Transit to a mounted filesystem. */ | |
1388 | if (managed & DCACHE_MOUNTED) { | |
1389 | struct vfsmount *mounted = lookup_mnt(path); | |
1390 | if (!mounted) | |
1391 | break; | |
1392 | dput(path->dentry); | |
1393 | mntput(path->mnt); | |
1394 | path->mnt = mounted; | |
1395 | path->dentry = dget(mounted->mnt_root); | |
1396 | continue; | |
1397 | } | |
1398 | ||
1399 | /* Don't handle automount points here */ | |
1400 | break; | |
1401 | } | |
1402 | return 0; | |
1403 | } | |
4d359507 | 1404 | EXPORT_SYMBOL(follow_down); |
cc53ce53 | 1405 | |
9875cf80 DH |
1406 | /* |
1407 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1408 | */ | |
1409 | static void follow_mount(struct path *path) | |
1410 | { | |
1411 | while (d_mountpoint(path->dentry)) { | |
1412 | struct vfsmount *mounted = lookup_mnt(path); | |
1413 | if (!mounted) | |
1414 | break; | |
1415 | dput(path->dentry); | |
1416 | mntput(path->mnt); | |
1417 | path->mnt = mounted; | |
1418 | path->dentry = dget(mounted->mnt_root); | |
1419 | } | |
1420 | } | |
1421 | ||
eedf265a EB |
1422 | static int path_parent_directory(struct path *path) |
1423 | { | |
1424 | struct dentry *old = path->dentry; | |
1425 | /* rare case of legitimate dget_parent()... */ | |
1426 | path->dentry = dget_parent(path->dentry); | |
1427 | dput(old); | |
1428 | if (unlikely(!path_connected(path))) | |
1429 | return -ENOENT; | |
1430 | return 0; | |
1431 | } | |
1432 | ||
397d425d | 1433 | static int follow_dotdot(struct nameidata *nd) |
1da177e4 LT |
1434 | { |
1435 | while(1) { | |
2a737871 AV |
1436 | if (nd->path.dentry == nd->root.dentry && |
1437 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1438 | break; |
1439 | } | |
4ac91378 | 1440 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
eedf265a EB |
1441 | int ret = path_parent_directory(&nd->path); |
1442 | if (ret) | |
1443 | return ret; | |
1da177e4 LT |
1444 | break; |
1445 | } | |
3088dd70 | 1446 | if (!follow_up(&nd->path)) |
1da177e4 | 1447 | break; |
1da177e4 | 1448 | } |
79ed0226 | 1449 | follow_mount(&nd->path); |
31e6b01f | 1450 | nd->inode = nd->path.dentry->d_inode; |
397d425d | 1451 | return 0; |
1da177e4 LT |
1452 | } |
1453 | ||
baa03890 | 1454 | /* |
f4fdace9 OD |
1455 | * This looks up the name in dcache and possibly revalidates the found dentry. |
1456 | * NULL is returned if the dentry does not exist in the cache. | |
baa03890 | 1457 | */ |
e3c13928 AV |
1458 | static struct dentry *lookup_dcache(const struct qstr *name, |
1459 | struct dentry *dir, | |
6c51e513 | 1460 | unsigned int flags) |
baa03890 | 1461 | { |
a89f8337 | 1462 | struct dentry *dentry = d_lookup(dir, name); |
bad61189 | 1463 | if (dentry) { |
a89f8337 AV |
1464 | int error = d_revalidate(dentry, flags); |
1465 | if (unlikely(error <= 0)) { | |
1466 | if (!error) | |
1467 | d_invalidate(dentry); | |
1468 | dput(dentry); | |
1469 | return ERR_PTR(error); | |
bad61189 MS |
1470 | } |
1471 | } | |
baa03890 NP |
1472 | return dentry; |
1473 | } | |
1474 | ||
44396f4b | 1475 | /* |
a03ece5f AV |
1476 | * Parent directory has inode locked exclusive. This is one |
1477 | * and only case when ->lookup() gets called on non in-lookup | |
1478 | * dentries - as the matter of fact, this only gets called | |
1479 | * when directory is guaranteed to have no in-lookup children | |
1480 | * at all. | |
44396f4b | 1481 | */ |
e3c13928 | 1482 | static struct dentry *__lookup_hash(const struct qstr *name, |
72bd866a | 1483 | struct dentry *base, unsigned int flags) |
a3255546 | 1484 | { |
6c51e513 | 1485 | struct dentry *dentry = lookup_dcache(name, base, flags); |
a03ece5f AV |
1486 | struct dentry *old; |
1487 | struct inode *dir = base->d_inode; | |
a3255546 | 1488 | |
6c51e513 | 1489 | if (dentry) |
bad61189 | 1490 | return dentry; |
a3255546 | 1491 | |
a03ece5f AV |
1492 | /* Don't create child dentry for a dead directory. */ |
1493 | if (unlikely(IS_DEADDIR(dir))) | |
1494 | return ERR_PTR(-ENOENT); | |
1495 | ||
6c51e513 AV |
1496 | dentry = d_alloc(base, name); |
1497 | if (unlikely(!dentry)) | |
1498 | return ERR_PTR(-ENOMEM); | |
1499 | ||
a03ece5f AV |
1500 | old = dir->i_op->lookup(dir, dentry, flags); |
1501 | if (unlikely(old)) { | |
1502 | dput(dentry); | |
1503 | dentry = old; | |
1504 | } | |
1505 | return dentry; | |
a3255546 AV |
1506 | } |
1507 | ||
e97cdc87 | 1508 | static int lookup_fast(struct nameidata *nd, |
254cf582 AV |
1509 | struct path *path, struct inode **inode, |
1510 | unsigned *seqp) | |
1da177e4 | 1511 | { |
4ac91378 | 1512 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1513 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 | 1514 | int status = 1; |
9875cf80 DH |
1515 | int err; |
1516 | ||
b04f784e NP |
1517 | /* |
1518 | * Rename seqlock is not required here because in the off chance | |
5d0f49c1 AV |
1519 | * of a false negative due to a concurrent rename, the caller is |
1520 | * going to fall back to non-racy lookup. | |
b04f784e | 1521 | */ |
31e6b01f NP |
1522 | if (nd->flags & LOOKUP_RCU) { |
1523 | unsigned seq; | |
766c4cbf | 1524 | bool negative; |
da53be12 | 1525 | dentry = __d_lookup_rcu(parent, &nd->last, &seq); |
5d0f49c1 | 1526 | if (unlikely(!dentry)) { |
4675ac39 | 1527 | if (unlazy_walk(nd)) |
5d0f49c1 | 1528 | return -ECHILD; |
e9742b53 | 1529 | return 0; |
5d0f49c1 | 1530 | } |
5a18fff2 | 1531 | |
12f8ad4b LT |
1532 | /* |
1533 | * This sequence count validates that the inode matches | |
1534 | * the dentry name information from lookup. | |
1535 | */ | |
63afdfc7 | 1536 | *inode = d_backing_inode(dentry); |
766c4cbf | 1537 | negative = d_is_negative(dentry); |
5d0f49c1 | 1538 | if (unlikely(read_seqcount_retry(&dentry->d_seq, seq))) |
12f8ad4b LT |
1539 | return -ECHILD; |
1540 | ||
1541 | /* | |
1542 | * This sequence count validates that the parent had no | |
1543 | * changes while we did the lookup of the dentry above. | |
1544 | * | |
1545 | * The memory barrier in read_seqcount_begin of child is | |
1546 | * enough, we can use __read_seqcount_retry here. | |
1547 | */ | |
5d0f49c1 | 1548 | if (unlikely(__read_seqcount_retry(&parent->d_seq, nd->seq))) |
31e6b01f | 1549 | return -ECHILD; |
5a18fff2 | 1550 | |
254cf582 | 1551 | *seqp = seq; |
a89f8337 | 1552 | status = d_revalidate(dentry, nd->flags); |
209a7fb2 | 1553 | if (likely(status > 0)) { |
5d0f49c1 AV |
1554 | /* |
1555 | * Note: do negative dentry check after revalidation in | |
1556 | * case that drops it. | |
1557 | */ | |
1558 | if (unlikely(negative)) | |
1559 | return -ENOENT; | |
1560 | path->mnt = mnt; | |
1561 | path->dentry = dentry; | |
1562 | if (likely(__follow_mount_rcu(nd, path, inode, seqp))) | |
e9742b53 | 1563 | return 1; |
24643087 | 1564 | } |
4675ac39 | 1565 | if (unlazy_child(nd, dentry, seq)) |
209a7fb2 AV |
1566 | return -ECHILD; |
1567 | if (unlikely(status == -ECHILD)) | |
1568 | /* we'd been told to redo it in non-rcu mode */ | |
1569 | status = d_revalidate(dentry, nd->flags); | |
5a18fff2 | 1570 | } else { |
e97cdc87 | 1571 | dentry = __d_lookup(parent, &nd->last); |
5d0f49c1 | 1572 | if (unlikely(!dentry)) |
e9742b53 | 1573 | return 0; |
a89f8337 | 1574 | status = d_revalidate(dentry, nd->flags); |
9875cf80 | 1575 | } |
5a18fff2 | 1576 | if (unlikely(status <= 0)) { |
e9742b53 | 1577 | if (!status) |
5d0f49c1 | 1578 | d_invalidate(dentry); |
5542aa2f | 1579 | dput(dentry); |
5d0f49c1 | 1580 | return status; |
24643087 | 1581 | } |
766c4cbf AV |
1582 | if (unlikely(d_is_negative(dentry))) { |
1583 | dput(dentry); | |
1584 | return -ENOENT; | |
1585 | } | |
5d0f49c1 | 1586 | |
9875cf80 DH |
1587 | path->mnt = mnt; |
1588 | path->dentry = dentry; | |
756daf26 | 1589 | err = follow_managed(path, nd); |
e9742b53 | 1590 | if (likely(err > 0)) |
63afdfc7 | 1591 | *inode = d_backing_inode(path->dentry); |
8402752e | 1592 | return err; |
697f514d MS |
1593 | } |
1594 | ||
1595 | /* Fast lookup failed, do it the slow way */ | |
88d8331a AV |
1596 | static struct dentry *__lookup_slow(const struct qstr *name, |
1597 | struct dentry *dir, | |
1598 | unsigned int flags) | |
697f514d | 1599 | { |
88d8331a | 1600 | struct dentry *dentry, *old; |
1936386e | 1601 | struct inode *inode = dir->d_inode; |
d9171b93 | 1602 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
1936386e | 1603 | |
1936386e | 1604 | /* Don't go there if it's already dead */ |
94bdd655 | 1605 | if (unlikely(IS_DEADDIR(inode))) |
88d8331a | 1606 | return ERR_PTR(-ENOENT); |
94bdd655 | 1607 | again: |
d9171b93 | 1608 | dentry = d_alloc_parallel(dir, name, &wq); |
94bdd655 | 1609 | if (IS_ERR(dentry)) |
88d8331a | 1610 | return dentry; |
94bdd655 | 1611 | if (unlikely(!d_in_lookup(dentry))) { |
a89f8337 | 1612 | if (!(flags & LOOKUP_NO_REVAL)) { |
949a852e AV |
1613 | int error = d_revalidate(dentry, flags); |
1614 | if (unlikely(error <= 0)) { | |
94bdd655 | 1615 | if (!error) { |
949a852e | 1616 | d_invalidate(dentry); |
94bdd655 AV |
1617 | dput(dentry); |
1618 | goto again; | |
1619 | } | |
949a852e AV |
1620 | dput(dentry); |
1621 | dentry = ERR_PTR(error); | |
1622 | } | |
1623 | } | |
94bdd655 AV |
1624 | } else { |
1625 | old = inode->i_op->lookup(inode, dentry, flags); | |
1626 | d_lookup_done(dentry); | |
1627 | if (unlikely(old)) { | |
1628 | dput(dentry); | |
1629 | dentry = old; | |
949a852e AV |
1630 | } |
1631 | } | |
e3c13928 | 1632 | return dentry; |
1da177e4 LT |
1633 | } |
1634 | ||
88d8331a AV |
1635 | static struct dentry *lookup_slow(const struct qstr *name, |
1636 | struct dentry *dir, | |
1637 | unsigned int flags) | |
1638 | { | |
1639 | struct inode *inode = dir->d_inode; | |
1640 | struct dentry *res; | |
1641 | inode_lock_shared(inode); | |
1642 | res = __lookup_slow(name, dir, flags); | |
1643 | inode_unlock_shared(inode); | |
1644 | return res; | |
1645 | } | |
1646 | ||
52094c8a AV |
1647 | static inline int may_lookup(struct nameidata *nd) |
1648 | { | |
1649 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1650 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1651 | if (err != -ECHILD) |
1652 | return err; | |
4675ac39 | 1653 | if (unlazy_walk(nd)) |
52094c8a AV |
1654 | return -ECHILD; |
1655 | } | |
4ad5abb3 | 1656 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1657 | } |
1658 | ||
9856fa1b AV |
1659 | static inline int handle_dots(struct nameidata *nd, int type) |
1660 | { | |
1661 | if (type == LAST_DOTDOT) { | |
9e6697e2 AV |
1662 | if (!nd->root.mnt) |
1663 | set_root(nd); | |
9856fa1b | 1664 | if (nd->flags & LOOKUP_RCU) { |
70291aec | 1665 | return follow_dotdot_rcu(nd); |
9856fa1b | 1666 | } else |
397d425d | 1667 | return follow_dotdot(nd); |
9856fa1b AV |
1668 | } |
1669 | return 0; | |
1670 | } | |
1671 | ||
181548c0 AV |
1672 | static int pick_link(struct nameidata *nd, struct path *link, |
1673 | struct inode *inode, unsigned seq) | |
d63ff28f | 1674 | { |
626de996 | 1675 | int error; |
1cf2665b | 1676 | struct saved *last; |
756daf26 | 1677 | if (unlikely(nd->total_link_count++ >= MAXSYMLINKS)) { |
626de996 AV |
1678 | path_to_nameidata(link, nd); |
1679 | return -ELOOP; | |
1680 | } | |
bc40aee0 | 1681 | if (!(nd->flags & LOOKUP_RCU)) { |
7973387a AV |
1682 | if (link->mnt == nd->path.mnt) |
1683 | mntget(link->mnt); | |
d63ff28f | 1684 | } |
626de996 AV |
1685 | error = nd_alloc_stack(nd); |
1686 | if (unlikely(error)) { | |
bc40aee0 | 1687 | if (error == -ECHILD) { |
ad1633a1 AV |
1688 | if (unlikely(!legitimize_path(nd, link, seq))) { |
1689 | drop_links(nd); | |
1690 | nd->depth = 0; | |
1691 | nd->flags &= ~LOOKUP_RCU; | |
1692 | nd->path.mnt = NULL; | |
1693 | nd->path.dentry = NULL; | |
1694 | if (!(nd->flags & LOOKUP_ROOT)) | |
1695 | nd->root.mnt = NULL; | |
1696 | rcu_read_unlock(); | |
4675ac39 | 1697 | } else if (likely(unlazy_walk(nd)) == 0) |
ad1633a1 | 1698 | error = nd_alloc_stack(nd); |
bc40aee0 AV |
1699 | } |
1700 | if (error) { | |
1701 | path_put(link); | |
1702 | return error; | |
1703 | } | |
626de996 AV |
1704 | } |
1705 | ||
ab104923 | 1706 | last = nd->stack + nd->depth++; |
1cf2665b | 1707 | last->link = *link; |
fceef393 AV |
1708 | clear_delayed_call(&last->done); |
1709 | nd->link_inode = inode; | |
0450b2d1 | 1710 | last->seq = seq; |
d63ff28f AV |
1711 | return 1; |
1712 | } | |
1713 | ||
8f64fb1c | 1714 | enum {WALK_FOLLOW = 1, WALK_MORE = 2}; |
31d66bcd | 1715 | |
3ddcd056 LT |
1716 | /* |
1717 | * Do we need to follow links? We _really_ want to be able | |
1718 | * to do this check without having to look at inode->i_op, | |
1719 | * so we keep a cache of "no, this doesn't need follow_link" | |
1720 | * for the common case. | |
1721 | */ | |
8f64fb1c AV |
1722 | static inline int step_into(struct nameidata *nd, struct path *path, |
1723 | int flags, struct inode *inode, unsigned seq) | |
3ddcd056 | 1724 | { |
31d66bcd AV |
1725 | if (!(flags & WALK_MORE) && nd->depth) |
1726 | put_link(nd); | |
8f64fb1c AV |
1727 | if (likely(!d_is_symlink(path->dentry)) || |
1728 | !(flags & WALK_FOLLOW || nd->flags & LOOKUP_FOLLOW)) { | |
1729 | /* not a symlink or should not follow */ | |
1730 | path_to_nameidata(path, nd); | |
1731 | nd->inode = inode; | |
1732 | nd->seq = seq; | |
d63ff28f | 1733 | return 0; |
8f64fb1c | 1734 | } |
a7f77542 AV |
1735 | /* make sure that d_is_symlink above matches inode */ |
1736 | if (nd->flags & LOOKUP_RCU) { | |
8f64fb1c | 1737 | if (read_seqcount_retry(&path->dentry->d_seq, seq)) |
a7f77542 AV |
1738 | return -ECHILD; |
1739 | } | |
8f64fb1c | 1740 | return pick_link(nd, path, inode, seq); |
3ddcd056 LT |
1741 | } |
1742 | ||
4693a547 | 1743 | static int walk_component(struct nameidata *nd, int flags) |
ce57dfc1 | 1744 | { |
caa85634 | 1745 | struct path path; |
ce57dfc1 | 1746 | struct inode *inode; |
254cf582 | 1747 | unsigned seq; |
ce57dfc1 AV |
1748 | int err; |
1749 | /* | |
1750 | * "." and ".." are special - ".." especially so because it has | |
1751 | * to be able to know about the current root directory and | |
1752 | * parent relationships. | |
1753 | */ | |
4693a547 AV |
1754 | if (unlikely(nd->last_type != LAST_NORM)) { |
1755 | err = handle_dots(nd, nd->last_type); | |
1c4ff1a8 | 1756 | if (!(flags & WALK_MORE) && nd->depth) |
4693a547 AV |
1757 | put_link(nd); |
1758 | return err; | |
1759 | } | |
254cf582 | 1760 | err = lookup_fast(nd, &path, &inode, &seq); |
e9742b53 | 1761 | if (unlikely(err <= 0)) { |
697f514d | 1762 | if (err < 0) |
f0a9ba70 | 1763 | return err; |
e3c13928 AV |
1764 | path.dentry = lookup_slow(&nd->last, nd->path.dentry, |
1765 | nd->flags); | |
1766 | if (IS_ERR(path.dentry)) | |
1767 | return PTR_ERR(path.dentry); | |
7500c38a | 1768 | |
e3c13928 AV |
1769 | path.mnt = nd->path.mnt; |
1770 | err = follow_managed(&path, nd); | |
1771 | if (unlikely(err < 0)) | |
f0a9ba70 | 1772 | return err; |
697f514d | 1773 | |
7500c38a AV |
1774 | if (unlikely(d_is_negative(path.dentry))) { |
1775 | path_to_nameidata(&path, nd); | |
1776 | return -ENOENT; | |
1777 | } | |
1778 | ||
254cf582 | 1779 | seq = 0; /* we are already out of RCU mode */ |
d4565649 | 1780 | inode = d_backing_inode(path.dentry); |
ce57dfc1 | 1781 | } |
697f514d | 1782 | |
8f64fb1c | 1783 | return step_into(nd, &path, flags, inode, seq); |
ce57dfc1 AV |
1784 | } |
1785 | ||
bfcfaa77 LT |
1786 | /* |
1787 | * We can do the critical dentry name comparison and hashing | |
1788 | * operations one word at a time, but we are limited to: | |
1789 | * | |
1790 | * - Architectures with fast unaligned word accesses. We could | |
1791 | * do a "get_unaligned()" if this helps and is sufficiently | |
1792 | * fast. | |
1793 | * | |
bfcfaa77 LT |
1794 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we |
1795 | * do not trap on the (extremely unlikely) case of a page | |
1796 | * crossing operation. | |
1797 | * | |
1798 | * - Furthermore, we need an efficient 64-bit compile for the | |
1799 | * 64-bit case in order to generate the "number of bytes in | |
1800 | * the final mask". Again, that could be replaced with a | |
1801 | * efficient population count instruction or similar. | |
1802 | */ | |
1803 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1804 | ||
f68e556e | 1805 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1806 | |
468a9428 | 1807 | #ifdef HASH_MIX |
bfcfaa77 | 1808 | |
468a9428 | 1809 | /* Architecture provides HASH_MIX and fold_hash() in <asm/hash.h> */ |
bfcfaa77 | 1810 | |
468a9428 | 1811 | #elif defined(CONFIG_64BIT) |
0fed3ac8 | 1812 | /* |
2a18da7a GS |
1813 | * Register pressure in the mixing function is an issue, particularly |
1814 | * on 32-bit x86, but almost any function requires one state value and | |
1815 | * one temporary. Instead, use a function designed for two state values | |
1816 | * and no temporaries. | |
1817 | * | |
1818 | * This function cannot create a collision in only two iterations, so | |
1819 | * we have two iterations to achieve avalanche. In those two iterations, | |
1820 | * we have six layers of mixing, which is enough to spread one bit's | |
1821 | * influence out to 2^6 = 64 state bits. | |
1822 | * | |
1823 | * Rotate constants are scored by considering either 64 one-bit input | |
1824 | * deltas or 64*63/2 = 2016 two-bit input deltas, and finding the | |
1825 | * probability of that delta causing a change to each of the 128 output | |
1826 | * bits, using a sample of random initial states. | |
1827 | * | |
1828 | * The Shannon entropy of the computed probabilities is then summed | |
1829 | * to produce a score. Ideally, any input change has a 50% chance of | |
1830 | * toggling any given output bit. | |
1831 | * | |
1832 | * Mixing scores (in bits) for (12,45): | |
1833 | * Input delta: 1-bit 2-bit | |
1834 | * 1 round: 713.3 42542.6 | |
1835 | * 2 rounds: 2753.7 140389.8 | |
1836 | * 3 rounds: 5954.1 233458.2 | |
1837 | * 4 rounds: 7862.6 256672.2 | |
1838 | * Perfect: 8192 258048 | |
1839 | * (64*128) (64*63/2 * 128) | |
0fed3ac8 | 1840 | */ |
2a18da7a GS |
1841 | #define HASH_MIX(x, y, a) \ |
1842 | ( x ^= (a), \ | |
1843 | y ^= x, x = rol64(x,12),\ | |
1844 | x += y, y = rol64(y,45),\ | |
1845 | y *= 9 ) | |
bfcfaa77 | 1846 | |
0fed3ac8 | 1847 | /* |
2a18da7a GS |
1848 | * Fold two longs into one 32-bit hash value. This must be fast, but |
1849 | * latency isn't quite as critical, as there is a fair bit of additional | |
1850 | * work done before the hash value is used. | |
0fed3ac8 | 1851 | */ |
2a18da7a | 1852 | static inline unsigned int fold_hash(unsigned long x, unsigned long y) |
0fed3ac8 | 1853 | { |
2a18da7a GS |
1854 | y ^= x * GOLDEN_RATIO_64; |
1855 | y *= GOLDEN_RATIO_64; | |
1856 | return y >> 32; | |
0fed3ac8 GS |
1857 | } |
1858 | ||
bfcfaa77 LT |
1859 | #else /* 32-bit case */ |
1860 | ||
2a18da7a GS |
1861 | /* |
1862 | * Mixing scores (in bits) for (7,20): | |
1863 | * Input delta: 1-bit 2-bit | |
1864 | * 1 round: 330.3 9201.6 | |
1865 | * 2 rounds: 1246.4 25475.4 | |
1866 | * 3 rounds: 1907.1 31295.1 | |
1867 | * 4 rounds: 2042.3 31718.6 | |
1868 | * Perfect: 2048 31744 | |
1869 | * (32*64) (32*31/2 * 64) | |
1870 | */ | |
1871 | #define HASH_MIX(x, y, a) \ | |
1872 | ( x ^= (a), \ | |
1873 | y ^= x, x = rol32(x, 7),\ | |
1874 | x += y, y = rol32(y,20),\ | |
1875 | y *= 9 ) | |
bfcfaa77 | 1876 | |
2a18da7a | 1877 | static inline unsigned int fold_hash(unsigned long x, unsigned long y) |
0fed3ac8 | 1878 | { |
2a18da7a GS |
1879 | /* Use arch-optimized multiply if one exists */ |
1880 | return __hash_32(y ^ __hash_32(x)); | |
0fed3ac8 GS |
1881 | } |
1882 | ||
bfcfaa77 LT |
1883 | #endif |
1884 | ||
2a18da7a GS |
1885 | /* |
1886 | * Return the hash of a string of known length. This is carfully | |
1887 | * designed to match hash_name(), which is the more critical function. | |
1888 | * In particular, we must end by hashing a final word containing 0..7 | |
1889 | * payload bytes, to match the way that hash_name() iterates until it | |
1890 | * finds the delimiter after the name. | |
1891 | */ | |
8387ff25 | 1892 | unsigned int full_name_hash(const void *salt, const char *name, unsigned int len) |
bfcfaa77 | 1893 | { |
8387ff25 | 1894 | unsigned long a, x = 0, y = (unsigned long)salt; |
bfcfaa77 LT |
1895 | |
1896 | for (;;) { | |
fcfd2fbf GS |
1897 | if (!len) |
1898 | goto done; | |
e419b4cc | 1899 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1900 | if (len < sizeof(unsigned long)) |
1901 | break; | |
2a18da7a | 1902 | HASH_MIX(x, y, a); |
bfcfaa77 LT |
1903 | name += sizeof(unsigned long); |
1904 | len -= sizeof(unsigned long); | |
bfcfaa77 | 1905 | } |
2a18da7a | 1906 | x ^= a & bytemask_from_count(len); |
bfcfaa77 | 1907 | done: |
2a18da7a | 1908 | return fold_hash(x, y); |
bfcfaa77 LT |
1909 | } |
1910 | EXPORT_SYMBOL(full_name_hash); | |
1911 | ||
fcfd2fbf | 1912 | /* Return the "hash_len" (hash and length) of a null-terminated string */ |
8387ff25 | 1913 | u64 hashlen_string(const void *salt, const char *name) |
fcfd2fbf | 1914 | { |
8387ff25 LT |
1915 | unsigned long a = 0, x = 0, y = (unsigned long)salt; |
1916 | unsigned long adata, mask, len; | |
fcfd2fbf GS |
1917 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; |
1918 | ||
8387ff25 LT |
1919 | len = 0; |
1920 | goto inside; | |
1921 | ||
fcfd2fbf | 1922 | do { |
2a18da7a | 1923 | HASH_MIX(x, y, a); |
fcfd2fbf | 1924 | len += sizeof(unsigned long); |
8387ff25 | 1925 | inside: |
fcfd2fbf GS |
1926 | a = load_unaligned_zeropad(name+len); |
1927 | } while (!has_zero(a, &adata, &constants)); | |
1928 | ||
1929 | adata = prep_zero_mask(a, adata, &constants); | |
1930 | mask = create_zero_mask(adata); | |
2a18da7a | 1931 | x ^= a & zero_bytemask(mask); |
fcfd2fbf | 1932 | |
2a18da7a | 1933 | return hashlen_create(fold_hash(x, y), len + find_zero(mask)); |
fcfd2fbf GS |
1934 | } |
1935 | EXPORT_SYMBOL(hashlen_string); | |
1936 | ||
bfcfaa77 LT |
1937 | /* |
1938 | * Calculate the length and hash of the path component, and | |
d6bb3e90 | 1939 | * return the "hash_len" as the result. |
bfcfaa77 | 1940 | */ |
8387ff25 | 1941 | static inline u64 hash_name(const void *salt, const char *name) |
bfcfaa77 | 1942 | { |
8387ff25 LT |
1943 | unsigned long a = 0, b, x = 0, y = (unsigned long)salt; |
1944 | unsigned long adata, bdata, mask, len; | |
36126f8f | 1945 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; |
bfcfaa77 | 1946 | |
8387ff25 LT |
1947 | len = 0; |
1948 | goto inside; | |
1949 | ||
bfcfaa77 | 1950 | do { |
2a18da7a | 1951 | HASH_MIX(x, y, a); |
bfcfaa77 | 1952 | len += sizeof(unsigned long); |
8387ff25 | 1953 | inside: |
e419b4cc | 1954 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1955 | b = a ^ REPEAT_BYTE('/'); |
1956 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1957 | ||
1958 | adata = prep_zero_mask(a, adata, &constants); | |
1959 | bdata = prep_zero_mask(b, bdata, &constants); | |
36126f8f | 1960 | mask = create_zero_mask(adata | bdata); |
2a18da7a | 1961 | x ^= a & zero_bytemask(mask); |
36126f8f | 1962 | |
2a18da7a | 1963 | return hashlen_create(fold_hash(x, y), len + find_zero(mask)); |
bfcfaa77 LT |
1964 | } |
1965 | ||
2a18da7a | 1966 | #else /* !CONFIG_DCACHE_WORD_ACCESS: Slow, byte-at-a-time version */ |
bfcfaa77 | 1967 | |
fcfd2fbf | 1968 | /* Return the hash of a string of known length */ |
8387ff25 | 1969 | unsigned int full_name_hash(const void *salt, const char *name, unsigned int len) |
0145acc2 | 1970 | { |
8387ff25 | 1971 | unsigned long hash = init_name_hash(salt); |
0145acc2 | 1972 | while (len--) |
fcfd2fbf | 1973 | hash = partial_name_hash((unsigned char)*name++, hash); |
0145acc2 LT |
1974 | return end_name_hash(hash); |
1975 | } | |
ae942ae7 | 1976 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1977 | |
fcfd2fbf | 1978 | /* Return the "hash_len" (hash and length) of a null-terminated string */ |
8387ff25 | 1979 | u64 hashlen_string(const void *salt, const char *name) |
fcfd2fbf | 1980 | { |
8387ff25 | 1981 | unsigned long hash = init_name_hash(salt); |
fcfd2fbf GS |
1982 | unsigned long len = 0, c; |
1983 | ||
1984 | c = (unsigned char)*name; | |
e0ab7af9 | 1985 | while (c) { |
fcfd2fbf GS |
1986 | len++; |
1987 | hash = partial_name_hash(c, hash); | |
1988 | c = (unsigned char)name[len]; | |
e0ab7af9 | 1989 | } |
fcfd2fbf GS |
1990 | return hashlen_create(end_name_hash(hash), len); |
1991 | } | |
f2a031b6 | 1992 | EXPORT_SYMBOL(hashlen_string); |
fcfd2fbf | 1993 | |
200e9ef7 LT |
1994 | /* |
1995 | * We know there's a real path component here of at least | |
1996 | * one character. | |
1997 | */ | |
8387ff25 | 1998 | static inline u64 hash_name(const void *salt, const char *name) |
200e9ef7 | 1999 | { |
8387ff25 | 2000 | unsigned long hash = init_name_hash(salt); |
200e9ef7 LT |
2001 | unsigned long len = 0, c; |
2002 | ||
2003 | c = (unsigned char)*name; | |
2004 | do { | |
2005 | len++; | |
2006 | hash = partial_name_hash(c, hash); | |
2007 | c = (unsigned char)name[len]; | |
2008 | } while (c && c != '/'); | |
d6bb3e90 | 2009 | return hashlen_create(end_name_hash(hash), len); |
200e9ef7 LT |
2010 | } |
2011 | ||
bfcfaa77 LT |
2012 | #endif |
2013 | ||
1da177e4 LT |
2014 | /* |
2015 | * Name resolution. | |
ea3834d9 PM |
2016 | * This is the basic name resolution function, turning a pathname into |
2017 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 2018 | * |
ea3834d9 PM |
2019 | * Returns 0 and nd will have valid dentry and mnt on success. |
2020 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 2021 | */ |
6de88d72 | 2022 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 2023 | { |
1da177e4 | 2024 | int err; |
32cd7468 | 2025 | |
1da177e4 LT |
2026 | while (*name=='/') |
2027 | name++; | |
2028 | if (!*name) | |
9e18f10a | 2029 | return 0; |
1da177e4 | 2030 | |
1da177e4 LT |
2031 | /* At this point we know we have a real path component. */ |
2032 | for(;;) { | |
d6bb3e90 | 2033 | u64 hash_len; |
fe479a58 | 2034 | int type; |
1da177e4 | 2035 | |
52094c8a | 2036 | err = may_lookup(nd); |
2a18da7a | 2037 | if (err) |
3595e234 | 2038 | return err; |
1da177e4 | 2039 | |
8387ff25 | 2040 | hash_len = hash_name(nd->path.dentry, name); |
1da177e4 | 2041 | |
fe479a58 | 2042 | type = LAST_NORM; |
d6bb3e90 | 2043 | if (name[0] == '.') switch (hashlen_len(hash_len)) { |
fe479a58 | 2044 | case 2: |
200e9ef7 | 2045 | if (name[1] == '.') { |
fe479a58 | 2046 | type = LAST_DOTDOT; |
16c2cd71 AV |
2047 | nd->flags |= LOOKUP_JUMPED; |
2048 | } | |
fe479a58 AV |
2049 | break; |
2050 | case 1: | |
2051 | type = LAST_DOT; | |
2052 | } | |
5a202bcd AV |
2053 | if (likely(type == LAST_NORM)) { |
2054 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 2055 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd | 2056 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
a060dc50 | 2057 | struct qstr this = { { .hash_len = hash_len }, .name = name }; |
da53be12 | 2058 | err = parent->d_op->d_hash(parent, &this); |
5a202bcd | 2059 | if (err < 0) |
3595e234 | 2060 | return err; |
d6bb3e90 LT |
2061 | hash_len = this.hash_len; |
2062 | name = this.name; | |
5a202bcd AV |
2063 | } |
2064 | } | |
fe479a58 | 2065 | |
d6bb3e90 LT |
2066 | nd->last.hash_len = hash_len; |
2067 | nd->last.name = name; | |
5f4a6a69 AV |
2068 | nd->last_type = type; |
2069 | ||
d6bb3e90 LT |
2070 | name += hashlen_len(hash_len); |
2071 | if (!*name) | |
bdf6cbf1 | 2072 | goto OK; |
200e9ef7 LT |
2073 | /* |
2074 | * If it wasn't NUL, we know it was '/'. Skip that | |
2075 | * slash, and continue until no more slashes. | |
2076 | */ | |
2077 | do { | |
d6bb3e90 LT |
2078 | name++; |
2079 | } while (unlikely(*name == '/')); | |
8620c238 AV |
2080 | if (unlikely(!*name)) { |
2081 | OK: | |
368ee9ba | 2082 | /* pathname body, done */ |
8620c238 AV |
2083 | if (!nd->depth) |
2084 | return 0; | |
2085 | name = nd->stack[nd->depth - 1].name; | |
368ee9ba | 2086 | /* trailing symlink, done */ |
8620c238 AV |
2087 | if (!name) |
2088 | return 0; | |
2089 | /* last component of nested symlink */ | |
8f64fb1c | 2090 | err = walk_component(nd, WALK_FOLLOW); |
1c4ff1a8 AV |
2091 | } else { |
2092 | /* not the last component */ | |
8f64fb1c | 2093 | err = walk_component(nd, WALK_FOLLOW | WALK_MORE); |
8620c238 | 2094 | } |
ce57dfc1 | 2095 | if (err < 0) |
3595e234 | 2096 | return err; |
1da177e4 | 2097 | |
ce57dfc1 | 2098 | if (err) { |
626de996 | 2099 | const char *s = get_link(nd); |
5a460275 | 2100 | |
a1c83681 | 2101 | if (IS_ERR(s)) |
3595e234 | 2102 | return PTR_ERR(s); |
d40bcc09 AV |
2103 | err = 0; |
2104 | if (unlikely(!s)) { | |
2105 | /* jumped */ | |
b9ff4429 | 2106 | put_link(nd); |
d40bcc09 | 2107 | } else { |
fab51e8a AV |
2108 | nd->stack[nd->depth - 1].name = name; |
2109 | name = s; | |
2110 | continue; | |
d40bcc09 | 2111 | } |
31e6b01f | 2112 | } |
97242f99 AV |
2113 | if (unlikely(!d_can_lookup(nd->path.dentry))) { |
2114 | if (nd->flags & LOOKUP_RCU) { | |
4675ac39 | 2115 | if (unlazy_walk(nd)) |
97242f99 AV |
2116 | return -ECHILD; |
2117 | } | |
3595e234 | 2118 | return -ENOTDIR; |
97242f99 | 2119 | } |
1da177e4 | 2120 | } |
1da177e4 LT |
2121 | } |
2122 | ||
c8a53ee5 | 2123 | static const char *path_init(struct nameidata *nd, unsigned flags) |
31e6b01f | 2124 | { |
c8a53ee5 | 2125 | const char *s = nd->name->name; |
31e6b01f | 2126 | |
c0eb027e LT |
2127 | if (!*s) |
2128 | flags &= ~LOOKUP_RCU; | |
2129 | ||
31e6b01f | 2130 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ |
980f3ea2 | 2131 | nd->flags = flags | LOOKUP_JUMPED | LOOKUP_PARENT; |
31e6b01f | 2132 | nd->depth = 0; |
5b6ca027 | 2133 | if (flags & LOOKUP_ROOT) { |
b18825a7 DH |
2134 | struct dentry *root = nd->root.dentry; |
2135 | struct inode *inode = root->d_inode; | |
93893862 AV |
2136 | if (*s && unlikely(!d_can_lookup(root))) |
2137 | return ERR_PTR(-ENOTDIR); | |
5b6ca027 AV |
2138 | nd->path = nd->root; |
2139 | nd->inode = inode; | |
2140 | if (flags & LOOKUP_RCU) { | |
8b61e74f | 2141 | rcu_read_lock(); |
5b6ca027 | 2142 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
8f47a016 | 2143 | nd->root_seq = nd->seq; |
48a066e7 | 2144 | nd->m_seq = read_seqbegin(&mount_lock); |
5b6ca027 AV |
2145 | } else { |
2146 | path_get(&nd->path); | |
2147 | } | |
368ee9ba | 2148 | return s; |
5b6ca027 AV |
2149 | } |
2150 | ||
31e6b01f | 2151 | nd->root.mnt = NULL; |
248fb5b9 AV |
2152 | nd->path.mnt = NULL; |
2153 | nd->path.dentry = NULL; | |
31e6b01f | 2154 | |
48a066e7 | 2155 | nd->m_seq = read_seqbegin(&mount_lock); |
fd2f7cb5 | 2156 | if (*s == '/') { |
9e6697e2 | 2157 | if (flags & LOOKUP_RCU) |
8b61e74f | 2158 | rcu_read_lock(); |
9e6697e2 | 2159 | set_root(nd); |
248fb5b9 | 2160 | if (likely(!nd_jump_root(nd))) |
ef55d917 | 2161 | return s; |
248fb5b9 | 2162 | nd->root.mnt = NULL; |
ef55d917 AV |
2163 | rcu_read_unlock(); |
2164 | return ERR_PTR(-ECHILD); | |
c8a53ee5 | 2165 | } else if (nd->dfd == AT_FDCWD) { |
e41f7d4e AV |
2166 | if (flags & LOOKUP_RCU) { |
2167 | struct fs_struct *fs = current->fs; | |
2168 | unsigned seq; | |
31e6b01f | 2169 | |
8b61e74f | 2170 | rcu_read_lock(); |
c28cc364 | 2171 | |
e41f7d4e AV |
2172 | do { |
2173 | seq = read_seqcount_begin(&fs->seq); | |
2174 | nd->path = fs->pwd; | |
ef55d917 | 2175 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e AV |
2176 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
2177 | } while (read_seqcount_retry(&fs->seq, seq)); | |
2178 | } else { | |
2179 | get_fs_pwd(current->fs, &nd->path); | |
ef55d917 | 2180 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2181 | } |
ef55d917 | 2182 | return s; |
31e6b01f | 2183 | } else { |
582aa64a | 2184 | /* Caller must check execute permissions on the starting path component */ |
c8a53ee5 | 2185 | struct fd f = fdget_raw(nd->dfd); |
31e6b01f NP |
2186 | struct dentry *dentry; |
2187 | ||
2903ff01 | 2188 | if (!f.file) |
368ee9ba | 2189 | return ERR_PTR(-EBADF); |
31e6b01f | 2190 | |
2903ff01 | 2191 | dentry = f.file->f_path.dentry; |
31e6b01f | 2192 | |
fd2f7cb5 | 2193 | if (*s) { |
44b1d530 | 2194 | if (!d_can_lookup(dentry)) { |
2903ff01 | 2195 | fdput(f); |
368ee9ba | 2196 | return ERR_PTR(-ENOTDIR); |
2903ff01 | 2197 | } |
f52e0c11 | 2198 | } |
31e6b01f | 2199 | |
2903ff01 | 2200 | nd->path = f.file->f_path; |
e41f7d4e | 2201 | if (flags & LOOKUP_RCU) { |
8b61e74f | 2202 | rcu_read_lock(); |
34a26b99 AV |
2203 | nd->inode = nd->path.dentry->d_inode; |
2204 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
e41f7d4e | 2205 | } else { |
2903ff01 | 2206 | path_get(&nd->path); |
34a26b99 | 2207 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2208 | } |
34a26b99 | 2209 | fdput(f); |
368ee9ba | 2210 | return s; |
31e6b01f | 2211 | } |
9b4a9b14 AV |
2212 | } |
2213 | ||
3bdba28b | 2214 | static const char *trailing_symlink(struct nameidata *nd) |
95fa25d9 AV |
2215 | { |
2216 | const char *s; | |
fec2fa24 | 2217 | int error = may_follow_link(nd); |
deb106c6 | 2218 | if (unlikely(error)) |
3bdba28b | 2219 | return ERR_PTR(error); |
95fa25d9 | 2220 | nd->flags |= LOOKUP_PARENT; |
fab51e8a | 2221 | nd->stack[0].name = NULL; |
3b2e7f75 | 2222 | s = get_link(nd); |
deb106c6 | 2223 | return s ? s : ""; |
95fa25d9 AV |
2224 | } |
2225 | ||
caa85634 | 2226 | static inline int lookup_last(struct nameidata *nd) |
bd92d7fe AV |
2227 | { |
2228 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
2229 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
2230 | ||
2231 | nd->flags &= ~LOOKUP_PARENT; | |
1c4ff1a8 | 2232 | return walk_component(nd, 0); |
bd92d7fe AV |
2233 | } |
2234 | ||
4f757f3c AV |
2235 | static int handle_lookup_down(struct nameidata *nd) |
2236 | { | |
2237 | struct path path = nd->path; | |
2238 | struct inode *inode = nd->inode; | |
2239 | unsigned seq = nd->seq; | |
2240 | int err; | |
2241 | ||
2242 | if (nd->flags & LOOKUP_RCU) { | |
2243 | /* | |
2244 | * don't bother with unlazy_walk on failure - we are | |
2245 | * at the very beginning of walk, so we lose nothing | |
2246 | * if we simply redo everything in non-RCU mode | |
2247 | */ | |
2248 | if (unlikely(!__follow_mount_rcu(nd, &path, &inode, &seq))) | |
2249 | return -ECHILD; | |
2250 | } else { | |
2251 | dget(path.dentry); | |
2252 | err = follow_managed(&path, nd); | |
2253 | if (unlikely(err < 0)) | |
2254 | return err; | |
2255 | inode = d_backing_inode(path.dentry); | |
2256 | seq = 0; | |
2257 | } | |
2258 | path_to_nameidata(&path, nd); | |
2259 | nd->inode = inode; | |
2260 | nd->seq = seq; | |
2261 | return 0; | |
2262 | } | |
2263 | ||
9b4a9b14 | 2264 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2265 | static int path_lookupat(struct nameidata *nd, unsigned flags, struct path *path) |
9b4a9b14 | 2266 | { |
c8a53ee5 | 2267 | const char *s = path_init(nd, flags); |
bd92d7fe | 2268 | int err; |
31e6b01f | 2269 | |
368ee9ba AV |
2270 | if (IS_ERR(s)) |
2271 | return PTR_ERR(s); | |
4f757f3c AV |
2272 | |
2273 | if (unlikely(flags & LOOKUP_DOWN)) { | |
2274 | err = handle_lookup_down(nd); | |
2275 | if (unlikely(err < 0)) { | |
2276 | terminate_walk(nd); | |
2277 | return err; | |
2278 | } | |
2279 | } | |
2280 | ||
3bdba28b AV |
2281 | while (!(err = link_path_walk(s, nd)) |
2282 | && ((err = lookup_last(nd)) > 0)) { | |
2283 | s = trailing_symlink(nd); | |
2284 | if (IS_ERR(s)) { | |
2285 | err = PTR_ERR(s); | |
2286 | break; | |
bd92d7fe AV |
2287 | } |
2288 | } | |
9f1fafee AV |
2289 | if (!err) |
2290 | err = complete_walk(nd); | |
bd92d7fe | 2291 | |
deb106c6 AV |
2292 | if (!err && nd->flags & LOOKUP_DIRECTORY) |
2293 | if (!d_can_lookup(nd->path.dentry)) | |
bd23a539 | 2294 | err = -ENOTDIR; |
625b6d10 AV |
2295 | if (!err) { |
2296 | *path = nd->path; | |
2297 | nd->path.mnt = NULL; | |
2298 | nd->path.dentry = NULL; | |
2299 | } | |
2300 | terminate_walk(nd); | |
bd92d7fe | 2301 | return err; |
ee0827cd | 2302 | } |
31e6b01f | 2303 | |
625b6d10 | 2304 | static int filename_lookup(int dfd, struct filename *name, unsigned flags, |
9ad1aaa6 | 2305 | struct path *path, struct path *root) |
ee0827cd | 2306 | { |
894bc8c4 | 2307 | int retval; |
9883d185 | 2308 | struct nameidata nd; |
abc9f5be AV |
2309 | if (IS_ERR(name)) |
2310 | return PTR_ERR(name); | |
9ad1aaa6 AV |
2311 | if (unlikely(root)) { |
2312 | nd.root = *root; | |
2313 | flags |= LOOKUP_ROOT; | |
2314 | } | |
9883d185 | 2315 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2316 | retval = path_lookupat(&nd, flags | LOOKUP_RCU, path); |
ee0827cd | 2317 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2318 | retval = path_lookupat(&nd, flags, path); |
ee0827cd | 2319 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2320 | retval = path_lookupat(&nd, flags | LOOKUP_REVAL, path); |
31e6b01f | 2321 | |
f78570dd | 2322 | if (likely(!retval)) |
625b6d10 | 2323 | audit_inode(name, path->dentry, flags & LOOKUP_PARENT); |
9883d185 | 2324 | restore_nameidata(); |
e4bd1c1a | 2325 | putname(name); |
170aa3d0 | 2326 | return retval; |
1da177e4 LT |
2327 | } |
2328 | ||
8bcb77fa | 2329 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2330 | static int path_parentat(struct nameidata *nd, unsigned flags, |
391172c4 | 2331 | struct path *parent) |
8bcb77fa | 2332 | { |
c8a53ee5 | 2333 | const char *s = path_init(nd, flags); |
368ee9ba AV |
2334 | int err; |
2335 | if (IS_ERR(s)) | |
2336 | return PTR_ERR(s); | |
2337 | err = link_path_walk(s, nd); | |
8bcb77fa AV |
2338 | if (!err) |
2339 | err = complete_walk(nd); | |
391172c4 AV |
2340 | if (!err) { |
2341 | *parent = nd->path; | |
2342 | nd->path.mnt = NULL; | |
2343 | nd->path.dentry = NULL; | |
2344 | } | |
2345 | terminate_walk(nd); | |
8bcb77fa AV |
2346 | return err; |
2347 | } | |
2348 | ||
5c31b6ce | 2349 | static struct filename *filename_parentat(int dfd, struct filename *name, |
391172c4 AV |
2350 | unsigned int flags, struct path *parent, |
2351 | struct qstr *last, int *type) | |
8bcb77fa AV |
2352 | { |
2353 | int retval; | |
9883d185 | 2354 | struct nameidata nd; |
8bcb77fa | 2355 | |
5c31b6ce AV |
2356 | if (IS_ERR(name)) |
2357 | return name; | |
9883d185 | 2358 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2359 | retval = path_parentat(&nd, flags | LOOKUP_RCU, parent); |
8bcb77fa | 2360 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2361 | retval = path_parentat(&nd, flags, parent); |
8bcb77fa | 2362 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2363 | retval = path_parentat(&nd, flags | LOOKUP_REVAL, parent); |
391172c4 AV |
2364 | if (likely(!retval)) { |
2365 | *last = nd.last; | |
2366 | *type = nd.last_type; | |
2367 | audit_inode(name, parent->dentry, LOOKUP_PARENT); | |
5c31b6ce AV |
2368 | } else { |
2369 | putname(name); | |
2370 | name = ERR_PTR(retval); | |
391172c4 | 2371 | } |
9883d185 | 2372 | restore_nameidata(); |
5c31b6ce | 2373 | return name; |
8bcb77fa AV |
2374 | } |
2375 | ||
79714f72 AV |
2376 | /* does lookup, returns the object with parent locked */ |
2377 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 2378 | { |
5c31b6ce AV |
2379 | struct filename *filename; |
2380 | struct dentry *d; | |
391172c4 AV |
2381 | struct qstr last; |
2382 | int type; | |
51689104 | 2383 | |
5c31b6ce AV |
2384 | filename = filename_parentat(AT_FDCWD, getname_kernel(name), 0, path, |
2385 | &last, &type); | |
51689104 PM |
2386 | if (IS_ERR(filename)) |
2387 | return ERR_CAST(filename); | |
5c31b6ce | 2388 | if (unlikely(type != LAST_NORM)) { |
391172c4 | 2389 | path_put(path); |
5c31b6ce AV |
2390 | putname(filename); |
2391 | return ERR_PTR(-EINVAL); | |
79714f72 | 2392 | } |
5955102c | 2393 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 2394 | d = __lookup_hash(&last, path->dentry, 0); |
79714f72 | 2395 | if (IS_ERR(d)) { |
5955102c | 2396 | inode_unlock(path->dentry->d_inode); |
391172c4 | 2397 | path_put(path); |
79714f72 | 2398 | } |
51689104 | 2399 | putname(filename); |
79714f72 | 2400 | return d; |
5590ff0d UD |
2401 | } |
2402 | ||
d1811465 AV |
2403 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2404 | { | |
abc9f5be AV |
2405 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2406 | flags, path, NULL); | |
d1811465 | 2407 | } |
4d359507 | 2408 | EXPORT_SYMBOL(kern_path); |
d1811465 | 2409 | |
16f18200 JJS |
2410 | /** |
2411 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2412 | * @dentry: pointer to dentry of the base directory | |
2413 | * @mnt: pointer to vfs mount of the base directory | |
2414 | * @name: pointer to file name | |
2415 | * @flags: lookup flags | |
e0a01249 | 2416 | * @path: pointer to struct path to fill |
16f18200 JJS |
2417 | */ |
2418 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2419 | const char *name, unsigned int flags, | |
e0a01249 | 2420 | struct path *path) |
16f18200 | 2421 | { |
9ad1aaa6 | 2422 | struct path root = {.mnt = mnt, .dentry = dentry}; |
9ad1aaa6 | 2423 | /* the first argument of filename_lookup() is ignored with root */ |
abc9f5be AV |
2424 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2425 | flags , path, &root); | |
16f18200 | 2426 | } |
4d359507 | 2427 | EXPORT_SYMBOL(vfs_path_lookup); |
16f18200 | 2428 | |
3c95f0dc AV |
2429 | static int lookup_one_len_common(const char *name, struct dentry *base, |
2430 | int len, struct qstr *this) | |
057f6c01 | 2431 | { |
3c95f0dc AV |
2432 | this->name = name; |
2433 | this->len = len; | |
2434 | this->hash = full_name_hash(base, name, len); | |
6a96ba54 | 2435 | if (!len) |
3c95f0dc | 2436 | return -EACCES; |
6a96ba54 | 2437 | |
21d8a15a AV |
2438 | if (unlikely(name[0] == '.')) { |
2439 | if (len < 2 || (len == 2 && name[1] == '.')) | |
3c95f0dc | 2440 | return -EACCES; |
21d8a15a AV |
2441 | } |
2442 | ||
6a96ba54 | 2443 | while (len--) { |
3c95f0dc | 2444 | unsigned int c = *(const unsigned char *)name++; |
6a96ba54 | 2445 | if (c == '/' || c == '\0') |
3c95f0dc | 2446 | return -EACCES; |
6a96ba54 | 2447 | } |
5a202bcd AV |
2448 | /* |
2449 | * See if the low-level filesystem might want | |
2450 | * to use its own hash.. | |
2451 | */ | |
2452 | if (base->d_flags & DCACHE_OP_HASH) { | |
3c95f0dc | 2453 | int err = base->d_op->d_hash(base, this); |
5a202bcd | 2454 | if (err < 0) |
3c95f0dc | 2455 | return err; |
5a202bcd | 2456 | } |
eead1911 | 2457 | |
3c95f0dc AV |
2458 | return inode_permission(base->d_inode, MAY_EXEC); |
2459 | } | |
2460 | ||
2461 | /** | |
2462 | * lookup_one_len - filesystem helper to lookup single pathname component | |
2463 | * @name: pathname component to lookup | |
2464 | * @base: base directory to lookup from | |
2465 | * @len: maximum length @len should be interpreted to | |
2466 | * | |
2467 | * Note that this routine is purely a helper for filesystem usage and should | |
2468 | * not be called by generic code. | |
2469 | * | |
2470 | * The caller must hold base->i_mutex. | |
2471 | */ | |
2472 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) | |
2473 | { | |
2474 | struct qstr this; | |
2475 | int err; | |
2476 | ||
2477 | WARN_ON_ONCE(!inode_is_locked(base->d_inode)); | |
2478 | ||
2479 | err = lookup_one_len_common(name, base, len, &this); | |
cda309de MS |
2480 | if (err) |
2481 | return ERR_PTR(err); | |
2482 | ||
72bd866a | 2483 | return __lookup_hash(&this, base, 0); |
057f6c01 | 2484 | } |
4d359507 | 2485 | EXPORT_SYMBOL(lookup_one_len); |
057f6c01 | 2486 | |
bbddca8e N |
2487 | /** |
2488 | * lookup_one_len_unlocked - filesystem helper to lookup single pathname component | |
2489 | * @name: pathname component to lookup | |
2490 | * @base: base directory to lookup from | |
2491 | * @len: maximum length @len should be interpreted to | |
2492 | * | |
2493 | * Note that this routine is purely a helper for filesystem usage and should | |
2494 | * not be called by generic code. | |
2495 | * | |
2496 | * Unlike lookup_one_len, it should be called without the parent | |
2497 | * i_mutex held, and will take the i_mutex itself if necessary. | |
2498 | */ | |
2499 | struct dentry *lookup_one_len_unlocked(const char *name, | |
2500 | struct dentry *base, int len) | |
2501 | { | |
2502 | struct qstr this; | |
bbddca8e | 2503 | int err; |
20d00ee8 | 2504 | struct dentry *ret; |
bbddca8e | 2505 | |
3c95f0dc | 2506 | err = lookup_one_len_common(name, base, len, &this); |
bbddca8e N |
2507 | if (err) |
2508 | return ERR_PTR(err); | |
2509 | ||
20d00ee8 LT |
2510 | ret = lookup_dcache(&this, base, 0); |
2511 | if (!ret) | |
2512 | ret = lookup_slow(&this, base, 0); | |
2513 | return ret; | |
bbddca8e N |
2514 | } |
2515 | EXPORT_SYMBOL(lookup_one_len_unlocked); | |
2516 | ||
eedf265a EB |
2517 | #ifdef CONFIG_UNIX98_PTYS |
2518 | int path_pts(struct path *path) | |
2519 | { | |
2520 | /* Find something mounted on "pts" in the same directory as | |
2521 | * the input path. | |
2522 | */ | |
2523 | struct dentry *child, *parent; | |
2524 | struct qstr this; | |
2525 | int ret; | |
2526 | ||
2527 | ret = path_parent_directory(path); | |
2528 | if (ret) | |
2529 | return ret; | |
2530 | ||
2531 | parent = path->dentry; | |
2532 | this.name = "pts"; | |
2533 | this.len = 3; | |
2534 | child = d_hash_and_lookup(parent, &this); | |
2535 | if (!child) | |
2536 | return -ENOENT; | |
2537 | ||
2538 | path->dentry = child; | |
2539 | dput(parent); | |
2540 | follow_mount(path); | |
2541 | return 0; | |
2542 | } | |
2543 | #endif | |
2544 | ||
1fa1e7f6 AW |
2545 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2546 | struct path *path, int *empty) | |
1da177e4 | 2547 | { |
abc9f5be AV |
2548 | return filename_lookup(dfd, getname_flags(name, flags, empty), |
2549 | flags, path, NULL); | |
1da177e4 | 2550 | } |
b853a161 | 2551 | EXPORT_SYMBOL(user_path_at_empty); |
1fa1e7f6 | 2552 | |
8033426e | 2553 | /** |
197df04c | 2554 | * mountpoint_last - look up last component for umount |
8033426e | 2555 | * @nd: pathwalk nameidata - currently pointing at parent directory of "last" |
8033426e JL |
2556 | * |
2557 | * This is a special lookup_last function just for umount. In this case, we | |
2558 | * need to resolve the path without doing any revalidation. | |
2559 | * | |
2560 | * The nameidata should be the result of doing a LOOKUP_PARENT pathwalk. Since | |
2561 | * mountpoints are always pinned in the dcache, their ancestors are too. Thus, | |
2562 | * in almost all cases, this lookup will be served out of the dcache. The only | |
2563 | * cases where it won't are if nd->last refers to a symlink or the path is | |
2564 | * bogus and it doesn't exist. | |
2565 | * | |
2566 | * Returns: | |
2567 | * -error: if there was an error during lookup. This includes -ENOENT if the | |
ba8f4613 | 2568 | * lookup found a negative dentry. |
8033426e | 2569 | * |
ba8f4613 AV |
2570 | * 0: if we successfully resolved nd->last and found it to not to be a |
2571 | * symlink that needs to be followed. | |
8033426e JL |
2572 | * |
2573 | * 1: if we successfully resolved nd->last and found it to be a symlink | |
ba8f4613 | 2574 | * that needs to be followed. |
8033426e JL |
2575 | */ |
2576 | static int | |
ba8f4613 | 2577 | mountpoint_last(struct nameidata *nd) |
8033426e JL |
2578 | { |
2579 | int error = 0; | |
8033426e | 2580 | struct dentry *dir = nd->path.dentry; |
ba8f4613 | 2581 | struct path path; |
8033426e | 2582 | |
35759521 AV |
2583 | /* If we're in rcuwalk, drop out of it to handle last component */ |
2584 | if (nd->flags & LOOKUP_RCU) { | |
4675ac39 | 2585 | if (unlazy_walk(nd)) |
deb106c6 | 2586 | return -ECHILD; |
8033426e JL |
2587 | } |
2588 | ||
2589 | nd->flags &= ~LOOKUP_PARENT; | |
2590 | ||
2591 | if (unlikely(nd->last_type != LAST_NORM)) { | |
2592 | error = handle_dots(nd, nd->last_type); | |
35759521 | 2593 | if (error) |
deb106c6 | 2594 | return error; |
ba8f4613 | 2595 | path.dentry = dget(nd->path.dentry); |
949a852e | 2596 | } else { |
ba8f4613 AV |
2597 | path.dentry = d_lookup(dir, &nd->last); |
2598 | if (!path.dentry) { | |
949a852e AV |
2599 | /* |
2600 | * No cached dentry. Mounted dentries are pinned in the | |
2601 | * cache, so that means that this dentry is probably | |
2602 | * a symlink or the path doesn't actually point | |
2603 | * to a mounted dentry. | |
2604 | */ | |
ba8f4613 | 2605 | path.dentry = lookup_slow(&nd->last, dir, |
949a852e | 2606 | nd->flags | LOOKUP_NO_REVAL); |
ba8f4613 AV |
2607 | if (IS_ERR(path.dentry)) |
2608 | return PTR_ERR(path.dentry); | |
bcceeeba | 2609 | } |
8033426e | 2610 | } |
ba8f4613 AV |
2611 | if (d_is_negative(path.dentry)) { |
2612 | dput(path.dentry); | |
deb106c6 | 2613 | return -ENOENT; |
8033426e | 2614 | } |
ba8f4613 | 2615 | path.mnt = nd->path.mnt; |
8f64fb1c | 2616 | return step_into(nd, &path, 0, d_backing_inode(path.dentry), 0); |
8033426e JL |
2617 | } |
2618 | ||
2619 | /** | |
197df04c | 2620 | * path_mountpoint - look up a path to be umounted |
2a78b857 | 2621 | * @nd: lookup context |
8033426e | 2622 | * @flags: lookup flags |
c8a53ee5 | 2623 | * @path: pointer to container for result |
8033426e JL |
2624 | * |
2625 | * Look up the given name, but don't attempt to revalidate the last component. | |
606d6fe3 | 2626 | * Returns 0 and "path" will be valid on success; Returns error otherwise. |
8033426e JL |
2627 | */ |
2628 | static int | |
c8a53ee5 | 2629 | path_mountpoint(struct nameidata *nd, unsigned flags, struct path *path) |
8033426e | 2630 | { |
c8a53ee5 | 2631 | const char *s = path_init(nd, flags); |
368ee9ba AV |
2632 | int err; |
2633 | if (IS_ERR(s)) | |
2634 | return PTR_ERR(s); | |
3bdba28b | 2635 | while (!(err = link_path_walk(s, nd)) && |
ba8f4613 | 2636 | (err = mountpoint_last(nd)) > 0) { |
3bdba28b AV |
2637 | s = trailing_symlink(nd); |
2638 | if (IS_ERR(s)) { | |
2639 | err = PTR_ERR(s); | |
8033426e | 2640 | break; |
3bdba28b | 2641 | } |
8033426e | 2642 | } |
ba8f4613 AV |
2643 | if (!err) { |
2644 | *path = nd->path; | |
2645 | nd->path.mnt = NULL; | |
2646 | nd->path.dentry = NULL; | |
2647 | follow_mount(path); | |
2648 | } | |
deb106c6 | 2649 | terminate_walk(nd); |
8033426e JL |
2650 | return err; |
2651 | } | |
2652 | ||
2d864651 | 2653 | static int |
668696dc | 2654 | filename_mountpoint(int dfd, struct filename *name, struct path *path, |
2d864651 AV |
2655 | unsigned int flags) |
2656 | { | |
9883d185 | 2657 | struct nameidata nd; |
cbaab2db | 2658 | int error; |
668696dc AV |
2659 | if (IS_ERR(name)) |
2660 | return PTR_ERR(name); | |
9883d185 | 2661 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2662 | error = path_mountpoint(&nd, flags | LOOKUP_RCU, path); |
2d864651 | 2663 | if (unlikely(error == -ECHILD)) |
c8a53ee5 | 2664 | error = path_mountpoint(&nd, flags, path); |
2d864651 | 2665 | if (unlikely(error == -ESTALE)) |
c8a53ee5 | 2666 | error = path_mountpoint(&nd, flags | LOOKUP_REVAL, path); |
2d864651 | 2667 | if (likely(!error)) |
668696dc | 2668 | audit_inode(name, path->dentry, 0); |
9883d185 | 2669 | restore_nameidata(); |
668696dc | 2670 | putname(name); |
2d864651 AV |
2671 | return error; |
2672 | } | |
2673 | ||
8033426e | 2674 | /** |
197df04c | 2675 | * user_path_mountpoint_at - lookup a path from userland in order to umount it |
8033426e JL |
2676 | * @dfd: directory file descriptor |
2677 | * @name: pathname from userland | |
2678 | * @flags: lookup flags | |
2679 | * @path: pointer to container to hold result | |
2680 | * | |
2681 | * A umount is a special case for path walking. We're not actually interested | |
2682 | * in the inode in this situation, and ESTALE errors can be a problem. We | |
2683 | * simply want track down the dentry and vfsmount attached at the mountpoint | |
2684 | * and avoid revalidating the last component. | |
2685 | * | |
2686 | * Returns 0 and populates "path" on success. | |
2687 | */ | |
2688 | int | |
197df04c | 2689 | user_path_mountpoint_at(int dfd, const char __user *name, unsigned int flags, |
8033426e JL |
2690 | struct path *path) |
2691 | { | |
cbaab2db | 2692 | return filename_mountpoint(dfd, getname(name), path, flags); |
8033426e JL |
2693 | } |
2694 | ||
2d864651 AV |
2695 | int |
2696 | kern_path_mountpoint(int dfd, const char *name, struct path *path, | |
2697 | unsigned int flags) | |
2698 | { | |
cbaab2db | 2699 | return filename_mountpoint(dfd, getname_kernel(name), path, flags); |
2d864651 AV |
2700 | } |
2701 | EXPORT_SYMBOL(kern_path_mountpoint); | |
2702 | ||
cbdf35bc | 2703 | int __check_sticky(struct inode *dir, struct inode *inode) |
1da177e4 | 2704 | { |
8e96e3b7 | 2705 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2706 | |
8e96e3b7 | 2707 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2708 | return 0; |
8e96e3b7 | 2709 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2710 | return 0; |
23adbe12 | 2711 | return !capable_wrt_inode_uidgid(inode, CAP_FOWNER); |
1da177e4 | 2712 | } |
cbdf35bc | 2713 | EXPORT_SYMBOL(__check_sticky); |
1da177e4 LT |
2714 | |
2715 | /* | |
2716 | * Check whether we can remove a link victim from directory dir, check | |
2717 | * whether the type of victim is right. | |
2718 | * 1. We can't do it if dir is read-only (done in permission()) | |
2719 | * 2. We should have write and exec permissions on dir | |
2720 | * 3. We can't remove anything from append-only dir | |
2721 | * 4. We can't do anything with immutable dir (done in permission()) | |
2722 | * 5. If the sticky bit on dir is set we should either | |
2723 | * a. be owner of dir, or | |
2724 | * b. be owner of victim, or | |
2725 | * c. have CAP_FOWNER capability | |
2726 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2727 | * links pointing to it. | |
0bd23d09 EB |
2728 | * 7. If the victim has an unknown uid or gid we can't change the inode. |
2729 | * 8. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2730 | * 9. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2731 | * 10. We can't remove a root or mountpoint. | |
2732 | * 11. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1da177e4 LT |
2733 | * nfs_async_unlink(). |
2734 | */ | |
b18825a7 | 2735 | static int may_delete(struct inode *dir, struct dentry *victim, bool isdir) |
1da177e4 | 2736 | { |
63afdfc7 | 2737 | struct inode *inode = d_backing_inode(victim); |
1da177e4 LT |
2738 | int error; |
2739 | ||
b18825a7 | 2740 | if (d_is_negative(victim)) |
1da177e4 | 2741 | return -ENOENT; |
b18825a7 | 2742 | BUG_ON(!inode); |
1da177e4 LT |
2743 | |
2744 | BUG_ON(victim->d_parent->d_inode != dir); | |
4fa6b5ec | 2745 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
1da177e4 | 2746 | |
f419a2e3 | 2747 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2748 | if (error) |
2749 | return error; | |
2750 | if (IS_APPEND(dir)) | |
2751 | return -EPERM; | |
b18825a7 DH |
2752 | |
2753 | if (check_sticky(dir, inode) || IS_APPEND(inode) || | |
0bd23d09 | 2754 | IS_IMMUTABLE(inode) || IS_SWAPFILE(inode) || HAS_UNMAPPED_ID(inode)) |
1da177e4 LT |
2755 | return -EPERM; |
2756 | if (isdir) { | |
44b1d530 | 2757 | if (!d_is_dir(victim)) |
1da177e4 LT |
2758 | return -ENOTDIR; |
2759 | if (IS_ROOT(victim)) | |
2760 | return -EBUSY; | |
44b1d530 | 2761 | } else if (d_is_dir(victim)) |
1da177e4 LT |
2762 | return -EISDIR; |
2763 | if (IS_DEADDIR(dir)) | |
2764 | return -ENOENT; | |
2765 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2766 | return -EBUSY; | |
2767 | return 0; | |
2768 | } | |
2769 | ||
2770 | /* Check whether we can create an object with dentry child in directory | |
2771 | * dir. | |
2772 | * 1. We can't do it if child already exists (open has special treatment for | |
2773 | * this case, but since we are inlined it's OK) | |
2774 | * 2. We can't do it if dir is read-only (done in permission()) | |
036d5236 EB |
2775 | * 3. We can't do it if the fs can't represent the fsuid or fsgid. |
2776 | * 4. We should have write and exec permissions on dir | |
2777 | * 5. We can't do it if dir is immutable (done in permission()) | |
1da177e4 | 2778 | */ |
a95164d9 | 2779 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 | 2780 | { |
036d5236 | 2781 | struct user_namespace *s_user_ns; |
14e972b4 | 2782 | audit_inode_child(dir, child, AUDIT_TYPE_CHILD_CREATE); |
1da177e4 LT |
2783 | if (child->d_inode) |
2784 | return -EEXIST; | |
2785 | if (IS_DEADDIR(dir)) | |
2786 | return -ENOENT; | |
036d5236 EB |
2787 | s_user_ns = dir->i_sb->s_user_ns; |
2788 | if (!kuid_has_mapping(s_user_ns, current_fsuid()) || | |
2789 | !kgid_has_mapping(s_user_ns, current_fsgid())) | |
2790 | return -EOVERFLOW; | |
f419a2e3 | 2791 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2792 | } |
2793 | ||
1da177e4 LT |
2794 | /* |
2795 | * p1 and p2 should be directories on the same fs. | |
2796 | */ | |
2797 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2798 | { | |
2799 | struct dentry *p; | |
2800 | ||
2801 | if (p1 == p2) { | |
5955102c | 2802 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
1da177e4 LT |
2803 | return NULL; |
2804 | } | |
2805 | ||
fc64005c | 2806 | mutex_lock(&p1->d_sb->s_vfs_rename_mutex); |
1da177e4 | 2807 | |
e2761a11 OH |
2808 | p = d_ancestor(p2, p1); |
2809 | if (p) { | |
5955102c AV |
2810 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT); |
2811 | inode_lock_nested(p1->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2812 | return p; |
1da177e4 LT |
2813 | } |
2814 | ||
e2761a11 OH |
2815 | p = d_ancestor(p1, p2); |
2816 | if (p) { | |
5955102c AV |
2817 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2818 | inode_lock_nested(p2->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2819 | return p; |
1da177e4 LT |
2820 | } |
2821 | ||
5955102c AV |
2822 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2823 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT2); | |
1da177e4 LT |
2824 | return NULL; |
2825 | } | |
4d359507 | 2826 | EXPORT_SYMBOL(lock_rename); |
1da177e4 LT |
2827 | |
2828 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2829 | { | |
5955102c | 2830 | inode_unlock(p1->d_inode); |
1da177e4 | 2831 | if (p1 != p2) { |
5955102c | 2832 | inode_unlock(p2->d_inode); |
fc64005c | 2833 | mutex_unlock(&p1->d_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2834 | } |
2835 | } | |
4d359507 | 2836 | EXPORT_SYMBOL(unlock_rename); |
1da177e4 | 2837 | |
4acdaf27 | 2838 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2839 | bool want_excl) |
1da177e4 | 2840 | { |
a95164d9 | 2841 | int error = may_create(dir, dentry); |
1da177e4 LT |
2842 | if (error) |
2843 | return error; | |
2844 | ||
acfa4380 | 2845 | if (!dir->i_op->create) |
1da177e4 LT |
2846 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2847 | mode &= S_IALLUGO; | |
2848 | mode |= S_IFREG; | |
2849 | error = security_inode_create(dir, dentry, mode); | |
2850 | if (error) | |
2851 | return error; | |
312b63fb | 2852 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2853 | if (!error) |
f38aa942 | 2854 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2855 | return error; |
2856 | } | |
4d359507 | 2857 | EXPORT_SYMBOL(vfs_create); |
1da177e4 | 2858 | |
8e6c848e AV |
2859 | int vfs_mkobj(struct dentry *dentry, umode_t mode, |
2860 | int (*f)(struct dentry *, umode_t, void *), | |
2861 | void *arg) | |
2862 | { | |
2863 | struct inode *dir = dentry->d_parent->d_inode; | |
2864 | int error = may_create(dir, dentry); | |
2865 | if (error) | |
2866 | return error; | |
2867 | ||
2868 | mode &= S_IALLUGO; | |
2869 | mode |= S_IFREG; | |
2870 | error = security_inode_create(dir, dentry, mode); | |
2871 | if (error) | |
2872 | return error; | |
2873 | error = f(dentry, mode, arg); | |
2874 | if (!error) | |
2875 | fsnotify_create(dir, dentry); | |
2876 | return error; | |
2877 | } | |
2878 | EXPORT_SYMBOL(vfs_mkobj); | |
2879 | ||
a2982cc9 EB |
2880 | bool may_open_dev(const struct path *path) |
2881 | { | |
2882 | return !(path->mnt->mnt_flags & MNT_NODEV) && | |
2883 | !(path->mnt->mnt_sb->s_iflags & SB_I_NODEV); | |
2884 | } | |
2885 | ||
f0bb5aaf | 2886 | static int may_open(const struct path *path, int acc_mode, int flag) |
1da177e4 | 2887 | { |
3fb64190 | 2888 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2889 | struct inode *inode = dentry->d_inode; |
2890 | int error; | |
2891 | ||
2892 | if (!inode) | |
2893 | return -ENOENT; | |
2894 | ||
c8fe8f30 CH |
2895 | switch (inode->i_mode & S_IFMT) { |
2896 | case S_IFLNK: | |
1da177e4 | 2897 | return -ELOOP; |
c8fe8f30 CH |
2898 | case S_IFDIR: |
2899 | if (acc_mode & MAY_WRITE) | |
2900 | return -EISDIR; | |
2901 | break; | |
2902 | case S_IFBLK: | |
2903 | case S_IFCHR: | |
a2982cc9 | 2904 | if (!may_open_dev(path)) |
1da177e4 | 2905 | return -EACCES; |
c8fe8f30 CH |
2906 | /*FALLTHRU*/ |
2907 | case S_IFIFO: | |
2908 | case S_IFSOCK: | |
1da177e4 | 2909 | flag &= ~O_TRUNC; |
c8fe8f30 | 2910 | break; |
4a3fd211 | 2911 | } |
b41572e9 | 2912 | |
62fb4a15 | 2913 | error = inode_permission(inode, MAY_OPEN | acc_mode); |
b41572e9 DH |
2914 | if (error) |
2915 | return error; | |
6146f0d5 | 2916 | |
1da177e4 LT |
2917 | /* |
2918 | * An append-only file must be opened in append mode for writing. | |
2919 | */ | |
2920 | if (IS_APPEND(inode)) { | |
8737c930 | 2921 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2922 | return -EPERM; |
1da177e4 | 2923 | if (flag & O_TRUNC) |
7715b521 | 2924 | return -EPERM; |
1da177e4 LT |
2925 | } |
2926 | ||
2927 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2928 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2929 | return -EPERM; |
1da177e4 | 2930 | |
f3c7691e | 2931 | return 0; |
7715b521 | 2932 | } |
1da177e4 | 2933 | |
e1181ee6 | 2934 | static int handle_truncate(struct file *filp) |
7715b521 | 2935 | { |
f0bb5aaf | 2936 | const struct path *path = &filp->f_path; |
7715b521 AV |
2937 | struct inode *inode = path->dentry->d_inode; |
2938 | int error = get_write_access(inode); | |
2939 | if (error) | |
2940 | return error; | |
2941 | /* | |
2942 | * Refuse to truncate files with mandatory locks held on them. | |
2943 | */ | |
d7a06983 | 2944 | error = locks_verify_locked(filp); |
7715b521 | 2945 | if (!error) |
ea0d3ab2 | 2946 | error = security_path_truncate(path); |
7715b521 AV |
2947 | if (!error) { |
2948 | error = do_truncate(path->dentry, 0, | |
2949 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2950 | filp); |
7715b521 AV |
2951 | } |
2952 | put_write_access(inode); | |
acd0c935 | 2953 | return error; |
1da177e4 LT |
2954 | } |
2955 | ||
d57999e1 DH |
2956 | static inline int open_to_namei_flags(int flag) |
2957 | { | |
8a5e929d AV |
2958 | if ((flag & O_ACCMODE) == 3) |
2959 | flag--; | |
d57999e1 DH |
2960 | return flag; |
2961 | } | |
2962 | ||
d3607752 | 2963 | static int may_o_create(const struct path *dir, struct dentry *dentry, umode_t mode) |
d18e9008 | 2964 | { |
1328c727 | 2965 | struct user_namespace *s_user_ns; |
d18e9008 MS |
2966 | int error = security_path_mknod(dir, dentry, mode, 0); |
2967 | if (error) | |
2968 | return error; | |
2969 | ||
1328c727 SF |
2970 | s_user_ns = dir->dentry->d_sb->s_user_ns; |
2971 | if (!kuid_has_mapping(s_user_ns, current_fsuid()) || | |
2972 | !kgid_has_mapping(s_user_ns, current_fsgid())) | |
2973 | return -EOVERFLOW; | |
2974 | ||
d18e9008 MS |
2975 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); |
2976 | if (error) | |
2977 | return error; | |
2978 | ||
2979 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2980 | } | |
2981 | ||
1acf0af9 DH |
2982 | /* |
2983 | * Attempt to atomically look up, create and open a file from a negative | |
2984 | * dentry. | |
2985 | * | |
2986 | * Returns 0 if successful. The file will have been created and attached to | |
2987 | * @file by the filesystem calling finish_open(). | |
2988 | * | |
2989 | * Returns 1 if the file was looked up only or didn't need creating. The | |
2990 | * caller will need to perform the open themselves. @path will have been | |
2991 | * updated to point to the new dentry. This may be negative. | |
2992 | * | |
2993 | * Returns an error code otherwise. | |
2994 | */ | |
2675a4eb AV |
2995 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
2996 | struct path *path, struct file *file, | |
2997 | const struct open_flags *op, | |
1643b43f | 2998 | int open_flag, umode_t mode, |
2675a4eb | 2999 | int *opened) |
d18e9008 | 3000 | { |
384f26e2 | 3001 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; |
d18e9008 | 3002 | struct inode *dir = nd->path.dentry->d_inode; |
d18e9008 | 3003 | int error; |
d18e9008 | 3004 | |
384f26e2 | 3005 | if (!(~open_flag & (O_EXCL | O_CREAT))) /* both O_EXCL and O_CREAT */ |
d18e9008 | 3006 | open_flag &= ~O_TRUNC; |
d18e9008 | 3007 | |
d18e9008 MS |
3008 | if (nd->flags & LOOKUP_DIRECTORY) |
3009 | open_flag |= O_DIRECTORY; | |
3010 | ||
30d90494 AV |
3011 | file->f_path.dentry = DENTRY_NOT_SET; |
3012 | file->f_path.mnt = nd->path.mnt; | |
0fb1ea09 AV |
3013 | error = dir->i_op->atomic_open(dir, dentry, file, |
3014 | open_to_namei_flags(open_flag), | |
3015 | mode, opened); | |
6fbd0714 | 3016 | d_lookup_done(dentry); |
384f26e2 AV |
3017 | if (!error) { |
3018 | /* | |
3019 | * We didn't have the inode before the open, so check open | |
3020 | * permission here. | |
3021 | */ | |
3022 | int acc_mode = op->acc_mode; | |
3023 | if (*opened & FILE_CREATED) { | |
3024 | WARN_ON(!(open_flag & O_CREAT)); | |
3025 | fsnotify_create(dir, dentry); | |
3026 | acc_mode = 0; | |
3027 | } | |
3028 | error = may_open(&file->f_path, acc_mode, open_flag); | |
3029 | if (WARN_ON(error > 0)) | |
3030 | error = -EINVAL; | |
3031 | } else if (error > 0) { | |
30d90494 | 3032 | if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 3033 | error = -EIO; |
03da633a | 3034 | } else { |
384f26e2 AV |
3035 | if (file->f_path.dentry) { |
3036 | dput(dentry); | |
3037 | dentry = file->f_path.dentry; | |
03da633a | 3038 | } |
384f26e2 AV |
3039 | if (*opened & FILE_CREATED) |
3040 | fsnotify_create(dir, dentry); | |
a01e718f AV |
3041 | if (unlikely(d_is_negative(dentry))) { |
3042 | error = -ENOENT; | |
3043 | } else { | |
3044 | path->dentry = dentry; | |
3045 | path->mnt = nd->path.mnt; | |
3046 | return 1; | |
3047 | } | |
62b2ce96 | 3048 | } |
d18e9008 | 3049 | } |
d18e9008 | 3050 | dput(dentry); |
2675a4eb | 3051 | return error; |
d18e9008 MS |
3052 | } |
3053 | ||
d58ffd35 | 3054 | /* |
1acf0af9 | 3055 | * Look up and maybe create and open the last component. |
d58ffd35 MS |
3056 | * |
3057 | * Must be called with i_mutex held on parent. | |
3058 | * | |
1acf0af9 DH |
3059 | * Returns 0 if the file was successfully atomically created (if necessary) and |
3060 | * opened. In this case the file will be returned attached to @file. | |
3061 | * | |
3062 | * Returns 1 if the file was not completely opened at this time, though lookups | |
3063 | * and creations will have been performed and the dentry returned in @path will | |
3064 | * be positive upon return if O_CREAT was specified. If O_CREAT wasn't | |
3065 | * specified then a negative dentry may be returned. | |
3066 | * | |
3067 | * An error code is returned otherwise. | |
3068 | * | |
3069 | * FILE_CREATE will be set in @*opened if the dentry was created and will be | |
3070 | * cleared otherwise prior to returning. | |
d58ffd35 | 3071 | */ |
2675a4eb AV |
3072 | static int lookup_open(struct nameidata *nd, struct path *path, |
3073 | struct file *file, | |
3074 | const struct open_flags *op, | |
64894cf8 | 3075 | bool got_write, int *opened) |
d58ffd35 MS |
3076 | { |
3077 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 3078 | struct inode *dir_inode = dir->d_inode; |
1643b43f | 3079 | int open_flag = op->open_flag; |
d58ffd35 | 3080 | struct dentry *dentry; |
1643b43f | 3081 | int error, create_error = 0; |
1643b43f | 3082 | umode_t mode = op->mode; |
6fbd0714 | 3083 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
d58ffd35 | 3084 | |
ce8644fc AV |
3085 | if (unlikely(IS_DEADDIR(dir_inode))) |
3086 | return -ENOENT; | |
d58ffd35 | 3087 | |
47237687 | 3088 | *opened &= ~FILE_CREATED; |
6fbd0714 AV |
3089 | dentry = d_lookup(dir, &nd->last); |
3090 | for (;;) { | |
3091 | if (!dentry) { | |
3092 | dentry = d_alloc_parallel(dir, &nd->last, &wq); | |
3093 | if (IS_ERR(dentry)) | |
3094 | return PTR_ERR(dentry); | |
3095 | } | |
3096 | if (d_in_lookup(dentry)) | |
3097 | break; | |
d58ffd35 | 3098 | |
6fbd0714 AV |
3099 | error = d_revalidate(dentry, nd->flags); |
3100 | if (likely(error > 0)) | |
3101 | break; | |
3102 | if (error) | |
3103 | goto out_dput; | |
3104 | d_invalidate(dentry); | |
3105 | dput(dentry); | |
3106 | dentry = NULL; | |
3107 | } | |
3108 | if (dentry->d_inode) { | |
6c51e513 | 3109 | /* Cached positive dentry: will open in f_op->open */ |
d18e9008 | 3110 | goto out_no_open; |
6c51e513 | 3111 | } |
d18e9008 | 3112 | |
1643b43f AV |
3113 | /* |
3114 | * Checking write permission is tricky, bacuse we don't know if we are | |
3115 | * going to actually need it: O_CREAT opens should work as long as the | |
3116 | * file exists. But checking existence breaks atomicity. The trick is | |
3117 | * to check access and if not granted clear O_CREAT from the flags. | |
3118 | * | |
3119 | * Another problem is returing the "right" error value (e.g. for an | |
3120 | * O_EXCL open we want to return EEXIST not EROFS). | |
3121 | */ | |
3122 | if (open_flag & O_CREAT) { | |
3123 | if (!IS_POSIXACL(dir->d_inode)) | |
3124 | mode &= ~current_umask(); | |
3125 | if (unlikely(!got_write)) { | |
3126 | create_error = -EROFS; | |
3127 | open_flag &= ~O_CREAT; | |
3128 | if (open_flag & (O_EXCL | O_TRUNC)) | |
3129 | goto no_open; | |
3130 | /* No side effects, safe to clear O_CREAT */ | |
3131 | } else { | |
3132 | create_error = may_o_create(&nd->path, dentry, mode); | |
3133 | if (create_error) { | |
3134 | open_flag &= ~O_CREAT; | |
3135 | if (open_flag & O_EXCL) | |
3136 | goto no_open; | |
3137 | } | |
3138 | } | |
3139 | } else if ((open_flag & (O_TRUNC|O_WRONLY|O_RDWR)) && | |
3140 | unlikely(!got_write)) { | |
3141 | /* | |
3142 | * No O_CREATE -> atomicity not a requirement -> fall | |
3143 | * back to lookup + open | |
3144 | */ | |
3145 | goto no_open; | |
d18e9008 MS |
3146 | } |
3147 | ||
6ac08709 | 3148 | if (dir_inode->i_op->atomic_open) { |
1643b43f AV |
3149 | error = atomic_open(nd, dentry, path, file, op, open_flag, |
3150 | mode, opened); | |
3151 | if (unlikely(error == -ENOENT) && create_error) | |
3152 | error = create_error; | |
3153 | return error; | |
d18e9008 | 3154 | } |
54ef4872 | 3155 | |
1643b43f | 3156 | no_open: |
6fbd0714 | 3157 | if (d_in_lookup(dentry)) { |
12fa5e24 AV |
3158 | struct dentry *res = dir_inode->i_op->lookup(dir_inode, dentry, |
3159 | nd->flags); | |
6fbd0714 | 3160 | d_lookup_done(dentry); |
12fa5e24 AV |
3161 | if (unlikely(res)) { |
3162 | if (IS_ERR(res)) { | |
3163 | error = PTR_ERR(res); | |
3164 | goto out_dput; | |
3165 | } | |
3166 | dput(dentry); | |
3167 | dentry = res; | |
3168 | } | |
54ef4872 MS |
3169 | } |
3170 | ||
d58ffd35 | 3171 | /* Negative dentry, just create the file */ |
1643b43f | 3172 | if (!dentry->d_inode && (open_flag & O_CREAT)) { |
47237687 | 3173 | *opened |= FILE_CREATED; |
ce8644fc | 3174 | audit_inode_child(dir_inode, dentry, AUDIT_TYPE_CHILD_CREATE); |
ce8644fc AV |
3175 | if (!dir_inode->i_op->create) { |
3176 | error = -EACCES; | |
d58ffd35 | 3177 | goto out_dput; |
ce8644fc AV |
3178 | } |
3179 | error = dir_inode->i_op->create(dir_inode, dentry, mode, | |
1643b43f | 3180 | open_flag & O_EXCL); |
d58ffd35 MS |
3181 | if (error) |
3182 | goto out_dput; | |
ce8644fc | 3183 | fsnotify_create(dir_inode, dentry); |
d58ffd35 | 3184 | } |
1643b43f AV |
3185 | if (unlikely(create_error) && !dentry->d_inode) { |
3186 | error = create_error; | |
3187 | goto out_dput; | |
d58ffd35 | 3188 | } |
d18e9008 | 3189 | out_no_open: |
d58ffd35 MS |
3190 | path->dentry = dentry; |
3191 | path->mnt = nd->path.mnt; | |
2675a4eb | 3192 | return 1; |
d58ffd35 MS |
3193 | |
3194 | out_dput: | |
3195 | dput(dentry); | |
2675a4eb | 3196 | return error; |
d58ffd35 MS |
3197 | } |
3198 | ||
31e6b01f | 3199 | /* |
fe2d35ff | 3200 | * Handle the last step of open() |
31e6b01f | 3201 | */ |
896475d5 | 3202 | static int do_last(struct nameidata *nd, |
2675a4eb | 3203 | struct file *file, const struct open_flags *op, |
76ae2a5a | 3204 | int *opened) |
fb1cc555 | 3205 | { |
a1e28038 | 3206 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 3207 | int open_flag = op->open_flag; |
77d660a8 | 3208 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 3209 | bool got_write = false; |
bcda7652 | 3210 | int acc_mode = op->acc_mode; |
254cf582 | 3211 | unsigned seq; |
a1eb3315 | 3212 | struct inode *inode; |
896475d5 | 3213 | struct path path; |
16c2cd71 | 3214 | int error; |
1f36f774 | 3215 | |
c3e380b0 AV |
3216 | nd->flags &= ~LOOKUP_PARENT; |
3217 | nd->flags |= op->intent; | |
3218 | ||
bc77daa7 | 3219 | if (nd->last_type != LAST_NORM) { |
fe2d35ff | 3220 | error = handle_dots(nd, nd->last_type); |
deb106c6 | 3221 | if (unlikely(error)) |
2675a4eb | 3222 | return error; |
e83db167 | 3223 | goto finish_open; |
1f36f774 | 3224 | } |
67ee3ad2 | 3225 | |
ca344a89 | 3226 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
3227 | if (nd->last.name[nd->last.len]) |
3228 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
3229 | /* we _can_ be in RCU mode here */ | |
254cf582 | 3230 | error = lookup_fast(nd, &path, &inode, &seq); |
e9742b53 | 3231 | if (likely(error > 0)) |
71574865 MS |
3232 | goto finish_lookup; |
3233 | ||
3234 | if (error < 0) | |
deb106c6 | 3235 | return error; |
71574865 MS |
3236 | |
3237 | BUG_ON(nd->inode != dir->d_inode); | |
6583fe22 | 3238 | BUG_ON(nd->flags & LOOKUP_RCU); |
b6183df7 MS |
3239 | } else { |
3240 | /* create side of things */ | |
3241 | /* | |
3242 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
3243 | * has been cleared when we got to the last component we are | |
3244 | * about to look up | |
3245 | */ | |
3246 | error = complete_walk(nd); | |
e8bb73df | 3247 | if (error) |
2675a4eb | 3248 | return error; |
fe2d35ff | 3249 | |
76ae2a5a | 3250 | audit_inode(nd->name, dir, LOOKUP_PARENT); |
b6183df7 | 3251 | /* trailing slashes? */ |
deb106c6 AV |
3252 | if (unlikely(nd->last.name[nd->last.len])) |
3253 | return -EISDIR; | |
b6183df7 | 3254 | } |
a2c36b45 | 3255 | |
9cf843e3 | 3256 | if (open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
64894cf8 AV |
3257 | error = mnt_want_write(nd->path.mnt); |
3258 | if (!error) | |
3259 | got_write = true; | |
3260 | /* | |
3261 | * do _not_ fail yet - we might not need that or fail with | |
3262 | * a different error; let lookup_open() decide; we'll be | |
3263 | * dropping this one anyway. | |
3264 | */ | |
3265 | } | |
9cf843e3 AV |
3266 | if (open_flag & O_CREAT) |
3267 | inode_lock(dir->d_inode); | |
3268 | else | |
3269 | inode_lock_shared(dir->d_inode); | |
896475d5 | 3270 | error = lookup_open(nd, &path, file, op, got_write, opened); |
9cf843e3 AV |
3271 | if (open_flag & O_CREAT) |
3272 | inode_unlock(dir->d_inode); | |
3273 | else | |
3274 | inode_unlock_shared(dir->d_inode); | |
a1e28038 | 3275 | |
2675a4eb AV |
3276 | if (error <= 0) { |
3277 | if (error) | |
d18e9008 MS |
3278 | goto out; |
3279 | ||
47237687 | 3280 | if ((*opened & FILE_CREATED) || |
496ad9aa | 3281 | !S_ISREG(file_inode(file)->i_mode)) |
77d660a8 | 3282 | will_truncate = false; |
d18e9008 | 3283 | |
76ae2a5a | 3284 | audit_inode(nd->name, file->f_path.dentry, 0); |
d18e9008 MS |
3285 | goto opened; |
3286 | } | |
fb1cc555 | 3287 | |
47237687 | 3288 | if (*opened & FILE_CREATED) { |
9b44f1b3 | 3289 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 3290 | open_flag &= ~O_TRUNC; |
77d660a8 | 3291 | will_truncate = false; |
62fb4a15 | 3292 | acc_mode = 0; |
896475d5 | 3293 | path_to_nameidata(&path, nd); |
e83db167 | 3294 | goto finish_open_created; |
fb1cc555 AV |
3295 | } |
3296 | ||
d18e9008 MS |
3297 | /* |
3298 | * If atomic_open() acquired write access it is dropped now due to | |
3299 | * possible mount and symlink following (this might be optimized away if | |
3300 | * necessary...) | |
3301 | */ | |
64894cf8 | 3302 | if (got_write) { |
d18e9008 | 3303 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3304 | got_write = false; |
d18e9008 MS |
3305 | } |
3306 | ||
e6ec03a2 AV |
3307 | error = follow_managed(&path, nd); |
3308 | if (unlikely(error < 0)) | |
3309 | return error; | |
3310 | ||
6583fe22 AV |
3311 | if (unlikely(d_is_negative(path.dentry))) { |
3312 | path_to_nameidata(&path, nd); | |
3313 | return -ENOENT; | |
3314 | } | |
3315 | ||
3316 | /* | |
3317 | * create/update audit record if it already exists. | |
3318 | */ | |
3319 | audit_inode(nd->name, path.dentry, 0); | |
3320 | ||
deb106c6 AV |
3321 | if (unlikely((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT))) { |
3322 | path_to_nameidata(&path, nd); | |
3323 | return -EEXIST; | |
3324 | } | |
fb1cc555 | 3325 | |
254cf582 | 3326 | seq = 0; /* out of RCU mode, so the value doesn't matter */ |
d4565649 | 3327 | inode = d_backing_inode(path.dentry); |
766c4cbf | 3328 | finish_lookup: |
8f64fb1c | 3329 | error = step_into(nd, &path, 0, inode, seq); |
deb106c6 | 3330 | if (unlikely(error)) |
d63ff28f | 3331 | return error; |
bc77daa7 | 3332 | finish_open: |
8f64fb1c | 3333 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
a3fbbde7 | 3334 | error = complete_walk(nd); |
fac7d191 | 3335 | if (error) |
2675a4eb | 3336 | return error; |
76ae2a5a | 3337 | audit_inode(nd->name, nd->path.dentry, 0); |
fb1cc555 | 3338 | error = -EISDIR; |
44b1d530 | 3339 | if ((open_flag & O_CREAT) && d_is_dir(nd->path.dentry)) |
2675a4eb | 3340 | goto out; |
af2f5542 | 3341 | error = -ENOTDIR; |
44b1d530 | 3342 | if ((nd->flags & LOOKUP_DIRECTORY) && !d_can_lookup(nd->path.dentry)) |
2675a4eb | 3343 | goto out; |
4bbcbd3b | 3344 | if (!d_is_reg(nd->path.dentry)) |
77d660a8 | 3345 | will_truncate = false; |
6c0d46c4 | 3346 | |
0f9d1a10 AV |
3347 | if (will_truncate) { |
3348 | error = mnt_want_write(nd->path.mnt); | |
3349 | if (error) | |
2675a4eb | 3350 | goto out; |
64894cf8 | 3351 | got_write = true; |
0f9d1a10 | 3352 | } |
e83db167 | 3353 | finish_open_created: |
6ac08709 AV |
3354 | error = may_open(&nd->path, acc_mode, open_flag); |
3355 | if (error) | |
3356 | goto out; | |
4aa7c634 MS |
3357 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
3358 | error = vfs_open(&nd->path, file, current_cred()); | |
fac7d191 | 3359 | if (error) |
015c3bbc | 3360 | goto out; |
fac7d191 | 3361 | *opened |= FILE_OPENED; |
a8277b9b | 3362 | opened: |
2675a4eb | 3363 | error = open_check_o_direct(file); |
fe9ec829 AV |
3364 | if (!error) |
3365 | error = ima_file_check(file, op->acc_mode, *opened); | |
3366 | if (!error && will_truncate) | |
2675a4eb | 3367 | error = handle_truncate(file); |
ca344a89 | 3368 | out: |
fe9ec829 AV |
3369 | if (unlikely(error) && (*opened & FILE_OPENED)) |
3370 | fput(file); | |
c80567c8 AV |
3371 | if (unlikely(error > 0)) { |
3372 | WARN_ON(1); | |
3373 | error = -EINVAL; | |
3374 | } | |
64894cf8 | 3375 | if (got_write) |
0f9d1a10 | 3376 | mnt_drop_write(nd->path.mnt); |
2675a4eb | 3377 | return error; |
fb1cc555 AV |
3378 | } |
3379 | ||
af7bd4dc AG |
3380 | struct dentry *vfs_tmpfile(struct dentry *dentry, umode_t mode, int open_flag) |
3381 | { | |
af7bd4dc AG |
3382 | struct dentry *child = NULL; |
3383 | struct inode *dir = dentry->d_inode; | |
3384 | struct inode *inode; | |
3385 | int error; | |
3386 | ||
3387 | /* we want directory to be writable */ | |
3388 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
3389 | if (error) | |
3390 | goto out_err; | |
3391 | error = -EOPNOTSUPP; | |
3392 | if (!dir->i_op->tmpfile) | |
3393 | goto out_err; | |
3394 | error = -ENOMEM; | |
cdf01226 | 3395 | child = d_alloc(dentry, &slash_name); |
af7bd4dc AG |
3396 | if (unlikely(!child)) |
3397 | goto out_err; | |
3398 | error = dir->i_op->tmpfile(dir, child, mode); | |
3399 | if (error) | |
3400 | goto out_err; | |
3401 | error = -ENOENT; | |
3402 | inode = child->d_inode; | |
3403 | if (unlikely(!inode)) | |
3404 | goto out_err; | |
3405 | if (!(open_flag & O_EXCL)) { | |
3406 | spin_lock(&inode->i_lock); | |
3407 | inode->i_state |= I_LINKABLE; | |
3408 | spin_unlock(&inode->i_lock); | |
3409 | } | |
3410 | return child; | |
3411 | ||
3412 | out_err: | |
3413 | dput(child); | |
3414 | return ERR_PTR(error); | |
3415 | } | |
3416 | EXPORT_SYMBOL(vfs_tmpfile); | |
3417 | ||
c8a53ee5 | 3418 | static int do_tmpfile(struct nameidata *nd, unsigned flags, |
60545d0d AV |
3419 | const struct open_flags *op, |
3420 | struct file *file, int *opened) | |
3421 | { | |
625b6d10 | 3422 | struct dentry *child; |
625b6d10 | 3423 | struct path path; |
c8a53ee5 | 3424 | int error = path_lookupat(nd, flags | LOOKUP_DIRECTORY, &path); |
60545d0d AV |
3425 | if (unlikely(error)) |
3426 | return error; | |
625b6d10 | 3427 | error = mnt_want_write(path.mnt); |
60545d0d AV |
3428 | if (unlikely(error)) |
3429 | goto out; | |
af7bd4dc AG |
3430 | child = vfs_tmpfile(path.dentry, op->mode, op->open_flag); |
3431 | error = PTR_ERR(child); | |
684e73be | 3432 | if (IS_ERR(child)) |
60545d0d | 3433 | goto out2; |
625b6d10 AV |
3434 | dput(path.dentry); |
3435 | path.dentry = child; | |
c8a53ee5 | 3436 | audit_inode(nd->name, child, 0); |
69a91c23 | 3437 | /* Don't check for other permissions, the inode was just created */ |
62fb4a15 | 3438 | error = may_open(&path, 0, op->open_flag); |
60545d0d AV |
3439 | if (error) |
3440 | goto out2; | |
625b6d10 AV |
3441 | file->f_path.mnt = path.mnt; |
3442 | error = finish_open(file, child, NULL, opened); | |
60545d0d AV |
3443 | if (error) |
3444 | goto out2; | |
3445 | error = open_check_o_direct(file); | |
af7bd4dc | 3446 | if (error) |
60545d0d AV |
3447 | fput(file); |
3448 | out2: | |
625b6d10 | 3449 | mnt_drop_write(path.mnt); |
60545d0d | 3450 | out: |
625b6d10 | 3451 | path_put(&path); |
60545d0d AV |
3452 | return error; |
3453 | } | |
3454 | ||
6ac08709 AV |
3455 | static int do_o_path(struct nameidata *nd, unsigned flags, struct file *file) |
3456 | { | |
3457 | struct path path; | |
3458 | int error = path_lookupat(nd, flags, &path); | |
3459 | if (!error) { | |
3460 | audit_inode(nd->name, path.dentry, 0); | |
3461 | error = vfs_open(&path, file, current_cred()); | |
3462 | path_put(&path); | |
3463 | } | |
3464 | return error; | |
3465 | } | |
3466 | ||
c8a53ee5 AV |
3467 | static struct file *path_openat(struct nameidata *nd, |
3468 | const struct open_flags *op, unsigned flags) | |
1da177e4 | 3469 | { |
368ee9ba | 3470 | const char *s; |
30d90494 | 3471 | struct file *file; |
47237687 | 3472 | int opened = 0; |
13aab428 | 3473 | int error; |
31e6b01f | 3474 | |
30d90494 | 3475 | file = get_empty_filp(); |
1afc99be AV |
3476 | if (IS_ERR(file)) |
3477 | return file; | |
31e6b01f | 3478 | |
30d90494 | 3479 | file->f_flags = op->open_flag; |
31e6b01f | 3480 | |
bb458c64 | 3481 | if (unlikely(file->f_flags & __O_TMPFILE)) { |
c8a53ee5 | 3482 | error = do_tmpfile(nd, flags, op, file, &opened); |
f15133df | 3483 | goto out2; |
60545d0d AV |
3484 | } |
3485 | ||
6ac08709 AV |
3486 | if (unlikely(file->f_flags & O_PATH)) { |
3487 | error = do_o_path(nd, flags, file); | |
3488 | if (!error) | |
3489 | opened |= FILE_OPENED; | |
3490 | goto out2; | |
3491 | } | |
3492 | ||
c8a53ee5 | 3493 | s = path_init(nd, flags); |
368ee9ba AV |
3494 | if (IS_ERR(s)) { |
3495 | put_filp(file); | |
3496 | return ERR_CAST(s); | |
3497 | } | |
3bdba28b | 3498 | while (!(error = link_path_walk(s, nd)) && |
76ae2a5a | 3499 | (error = do_last(nd, file, op, &opened)) > 0) { |
73d049a4 | 3500 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); |
3bdba28b AV |
3501 | s = trailing_symlink(nd); |
3502 | if (IS_ERR(s)) { | |
3503 | error = PTR_ERR(s); | |
2675a4eb | 3504 | break; |
3bdba28b | 3505 | } |
806b681c | 3506 | } |
deb106c6 | 3507 | terminate_walk(nd); |
f15133df | 3508 | out2: |
2675a4eb AV |
3509 | if (!(opened & FILE_OPENED)) { |
3510 | BUG_ON(!error); | |
30d90494 | 3511 | put_filp(file); |
16b1c1cd | 3512 | } |
2675a4eb AV |
3513 | if (unlikely(error)) { |
3514 | if (error == -EOPENSTALE) { | |
3515 | if (flags & LOOKUP_RCU) | |
3516 | error = -ECHILD; | |
3517 | else | |
3518 | error = -ESTALE; | |
3519 | } | |
3520 | file = ERR_PTR(error); | |
3521 | } | |
3522 | return file; | |
1da177e4 LT |
3523 | } |
3524 | ||
669abf4e | 3525 | struct file *do_filp_open(int dfd, struct filename *pathname, |
f9652e10 | 3526 | const struct open_flags *op) |
13aab428 | 3527 | { |
9883d185 | 3528 | struct nameidata nd; |
f9652e10 | 3529 | int flags = op->lookup_flags; |
13aab428 AV |
3530 | struct file *filp; |
3531 | ||
9883d185 | 3532 | set_nameidata(&nd, dfd, pathname); |
c8a53ee5 | 3533 | filp = path_openat(&nd, op, flags | LOOKUP_RCU); |
13aab428 | 3534 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3535 | filp = path_openat(&nd, op, flags); |
13aab428 | 3536 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3537 | filp = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3538 | restore_nameidata(); |
13aab428 AV |
3539 | return filp; |
3540 | } | |
3541 | ||
73d049a4 | 3542 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
f9652e10 | 3543 | const char *name, const struct open_flags *op) |
73d049a4 | 3544 | { |
9883d185 | 3545 | struct nameidata nd; |
73d049a4 | 3546 | struct file *file; |
51689104 | 3547 | struct filename *filename; |
f9652e10 | 3548 | int flags = op->lookup_flags | LOOKUP_ROOT; |
73d049a4 AV |
3549 | |
3550 | nd.root.mnt = mnt; | |
3551 | nd.root.dentry = dentry; | |
3552 | ||
b18825a7 | 3553 | if (d_is_symlink(dentry) && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
3554 | return ERR_PTR(-ELOOP); |
3555 | ||
51689104 | 3556 | filename = getname_kernel(name); |
a1c83681 | 3557 | if (IS_ERR(filename)) |
51689104 PM |
3558 | return ERR_CAST(filename); |
3559 | ||
9883d185 | 3560 | set_nameidata(&nd, -1, filename); |
c8a53ee5 | 3561 | file = path_openat(&nd, op, flags | LOOKUP_RCU); |
73d049a4 | 3562 | if (unlikely(file == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3563 | file = path_openat(&nd, op, flags); |
73d049a4 | 3564 | if (unlikely(file == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3565 | file = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3566 | restore_nameidata(); |
51689104 | 3567 | putname(filename); |
73d049a4 AV |
3568 | return file; |
3569 | } | |
3570 | ||
fa14a0b8 | 3571 | static struct dentry *filename_create(int dfd, struct filename *name, |
1ac12b4b | 3572 | struct path *path, unsigned int lookup_flags) |
1da177e4 | 3573 | { |
c663e5d8 | 3574 | struct dentry *dentry = ERR_PTR(-EEXIST); |
391172c4 AV |
3575 | struct qstr last; |
3576 | int type; | |
c30dabfe | 3577 | int err2; |
1ac12b4b JL |
3578 | int error; |
3579 | bool is_dir = (lookup_flags & LOOKUP_DIRECTORY); | |
3580 | ||
3581 | /* | |
3582 | * Note that only LOOKUP_REVAL and LOOKUP_DIRECTORY matter here. Any | |
3583 | * other flags passed in are ignored! | |
3584 | */ | |
3585 | lookup_flags &= LOOKUP_REVAL; | |
3586 | ||
5c31b6ce AV |
3587 | name = filename_parentat(dfd, name, lookup_flags, path, &last, &type); |
3588 | if (IS_ERR(name)) | |
3589 | return ERR_CAST(name); | |
1da177e4 | 3590 | |
c663e5d8 CH |
3591 | /* |
3592 | * Yucky last component or no last component at all? | |
3593 | * (foo/., foo/.., /////) | |
3594 | */ | |
5c31b6ce | 3595 | if (unlikely(type != LAST_NORM)) |
ed75e95d | 3596 | goto out; |
c663e5d8 | 3597 | |
c30dabfe | 3598 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
391172c4 | 3599 | err2 = mnt_want_write(path->mnt); |
c663e5d8 CH |
3600 | /* |
3601 | * Do the final lookup. | |
3602 | */ | |
391172c4 | 3603 | lookup_flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
5955102c | 3604 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 3605 | dentry = __lookup_hash(&last, path->dentry, lookup_flags); |
1da177e4 | 3606 | if (IS_ERR(dentry)) |
a8104a9f | 3607 | goto unlock; |
c663e5d8 | 3608 | |
a8104a9f | 3609 | error = -EEXIST; |
b18825a7 | 3610 | if (d_is_positive(dentry)) |
a8104a9f | 3611 | goto fail; |
b18825a7 | 3612 | |
c663e5d8 CH |
3613 | /* |
3614 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3615 | * From the vfs_mknod() POV we just have a negative dentry - | |
3616 | * all is fine. Let's be bastards - you had / on the end, you've | |
3617 | * been asking for (non-existent) directory. -ENOENT for you. | |
3618 | */ | |
391172c4 | 3619 | if (unlikely(!is_dir && last.name[last.len])) { |
a8104a9f | 3620 | error = -ENOENT; |
ed75e95d | 3621 | goto fail; |
e9baf6e5 | 3622 | } |
c30dabfe JK |
3623 | if (unlikely(err2)) { |
3624 | error = err2; | |
a8104a9f | 3625 | goto fail; |
c30dabfe | 3626 | } |
181c37b6 | 3627 | putname(name); |
1da177e4 | 3628 | return dentry; |
1da177e4 | 3629 | fail: |
a8104a9f AV |
3630 | dput(dentry); |
3631 | dentry = ERR_PTR(error); | |
3632 | unlock: | |
5955102c | 3633 | inode_unlock(path->dentry->d_inode); |
c30dabfe | 3634 | if (!err2) |
391172c4 | 3635 | mnt_drop_write(path->mnt); |
ed75e95d | 3636 | out: |
391172c4 | 3637 | path_put(path); |
181c37b6 | 3638 | putname(name); |
1da177e4 LT |
3639 | return dentry; |
3640 | } | |
fa14a0b8 AV |
3641 | |
3642 | struct dentry *kern_path_create(int dfd, const char *pathname, | |
3643 | struct path *path, unsigned int lookup_flags) | |
3644 | { | |
181c37b6 AV |
3645 | return filename_create(dfd, getname_kernel(pathname), |
3646 | path, lookup_flags); | |
fa14a0b8 | 3647 | } |
dae6ad8f AV |
3648 | EXPORT_SYMBOL(kern_path_create); |
3649 | ||
921a1650 AV |
3650 | void done_path_create(struct path *path, struct dentry *dentry) |
3651 | { | |
3652 | dput(dentry); | |
5955102c | 3653 | inode_unlock(path->dentry->d_inode); |
a8104a9f | 3654 | mnt_drop_write(path->mnt); |
921a1650 AV |
3655 | path_put(path); |
3656 | } | |
3657 | EXPORT_SYMBOL(done_path_create); | |
3658 | ||
520ae687 | 3659 | inline struct dentry *user_path_create(int dfd, const char __user *pathname, |
1ac12b4b | 3660 | struct path *path, unsigned int lookup_flags) |
dae6ad8f | 3661 | { |
181c37b6 | 3662 | return filename_create(dfd, getname(pathname), path, lookup_flags); |
dae6ad8f AV |
3663 | } |
3664 | EXPORT_SYMBOL(user_path_create); | |
3665 | ||
1a67aafb | 3666 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3667 | { |
a95164d9 | 3668 | int error = may_create(dir, dentry); |
1da177e4 LT |
3669 | |
3670 | if (error) | |
3671 | return error; | |
3672 | ||
975d6b39 | 3673 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3674 | return -EPERM; |
3675 | ||
acfa4380 | 3676 | if (!dir->i_op->mknod) |
1da177e4 LT |
3677 | return -EPERM; |
3678 | ||
08ce5f16 SH |
3679 | error = devcgroup_inode_mknod(mode, dev); |
3680 | if (error) | |
3681 | return error; | |
3682 | ||
1da177e4 LT |
3683 | error = security_inode_mknod(dir, dentry, mode, dev); |
3684 | if (error) | |
3685 | return error; | |
3686 | ||
1da177e4 | 3687 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3688 | if (!error) |
f38aa942 | 3689 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3690 | return error; |
3691 | } | |
4d359507 | 3692 | EXPORT_SYMBOL(vfs_mknod); |
1da177e4 | 3693 | |
f69aac00 | 3694 | static int may_mknod(umode_t mode) |
463c3197 DH |
3695 | { |
3696 | switch (mode & S_IFMT) { | |
3697 | case S_IFREG: | |
3698 | case S_IFCHR: | |
3699 | case S_IFBLK: | |
3700 | case S_IFIFO: | |
3701 | case S_IFSOCK: | |
3702 | case 0: /* zero mode translates to S_IFREG */ | |
3703 | return 0; | |
3704 | case S_IFDIR: | |
3705 | return -EPERM; | |
3706 | default: | |
3707 | return -EINVAL; | |
3708 | } | |
3709 | } | |
3710 | ||
8208a22b | 3711 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 3712 | unsigned, dev) |
1da177e4 | 3713 | { |
2ad94ae6 | 3714 | struct dentry *dentry; |
dae6ad8f AV |
3715 | struct path path; |
3716 | int error; | |
972567f1 | 3717 | unsigned int lookup_flags = 0; |
1da177e4 | 3718 | |
8e4bfca1 AV |
3719 | error = may_mknod(mode); |
3720 | if (error) | |
3721 | return error; | |
972567f1 JL |
3722 | retry: |
3723 | dentry = user_path_create(dfd, filename, &path, lookup_flags); | |
dae6ad8f AV |
3724 | if (IS_ERR(dentry)) |
3725 | return PTR_ERR(dentry); | |
2ad94ae6 | 3726 | |
dae6ad8f | 3727 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3728 | mode &= ~current_umask(); |
dae6ad8f | 3729 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3730 | if (error) |
a8104a9f | 3731 | goto out; |
463c3197 | 3732 | switch (mode & S_IFMT) { |
1da177e4 | 3733 | case 0: case S_IFREG: |
312b63fb | 3734 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
05d1a717 MZ |
3735 | if (!error) |
3736 | ima_post_path_mknod(dentry); | |
1da177e4 LT |
3737 | break; |
3738 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3739 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3740 | new_decode_dev(dev)); |
3741 | break; | |
3742 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3743 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3744 | break; |
1da177e4 | 3745 | } |
a8104a9f | 3746 | out: |
921a1650 | 3747 | done_path_create(&path, dentry); |
972567f1 JL |
3748 | if (retry_estale(error, lookup_flags)) { |
3749 | lookup_flags |= LOOKUP_REVAL; | |
3750 | goto retry; | |
3751 | } | |
1da177e4 LT |
3752 | return error; |
3753 | } | |
3754 | ||
8208a22b | 3755 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
3756 | { |
3757 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
3758 | } | |
3759 | ||
18bb1db3 | 3760 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3761 | { |
a95164d9 | 3762 | int error = may_create(dir, dentry); |
8de52778 | 3763 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3764 | |
3765 | if (error) | |
3766 | return error; | |
3767 | ||
acfa4380 | 3768 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3769 | return -EPERM; |
3770 | ||
3771 | mode &= (S_IRWXUGO|S_ISVTX); | |
3772 | error = security_inode_mkdir(dir, dentry, mode); | |
3773 | if (error) | |
3774 | return error; | |
3775 | ||
8de52778 AV |
3776 | if (max_links && dir->i_nlink >= max_links) |
3777 | return -EMLINK; | |
3778 | ||
1da177e4 | 3779 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3780 | if (!error) |
f38aa942 | 3781 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3782 | return error; |
3783 | } | |
4d359507 | 3784 | EXPORT_SYMBOL(vfs_mkdir); |
1da177e4 | 3785 | |
a218d0fd | 3786 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 3787 | { |
6902d925 | 3788 | struct dentry *dentry; |
dae6ad8f AV |
3789 | struct path path; |
3790 | int error; | |
b76d8b82 | 3791 | unsigned int lookup_flags = LOOKUP_DIRECTORY; |
1da177e4 | 3792 | |
b76d8b82 JL |
3793 | retry: |
3794 | dentry = user_path_create(dfd, pathname, &path, lookup_flags); | |
6902d925 | 3795 | if (IS_ERR(dentry)) |
dae6ad8f | 3796 | return PTR_ERR(dentry); |
1da177e4 | 3797 | |
dae6ad8f | 3798 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3799 | mode &= ~current_umask(); |
dae6ad8f | 3800 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3801 | if (!error) |
3802 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3803 | done_path_create(&path, dentry); |
b76d8b82 JL |
3804 | if (retry_estale(error, lookup_flags)) { |
3805 | lookup_flags |= LOOKUP_REVAL; | |
3806 | goto retry; | |
3807 | } | |
1da177e4 LT |
3808 | return error; |
3809 | } | |
3810 | ||
a218d0fd | 3811 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
3812 | { |
3813 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
3814 | } | |
3815 | ||
1da177e4 LT |
3816 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) |
3817 | { | |
3818 | int error = may_delete(dir, dentry, 1); | |
3819 | ||
3820 | if (error) | |
3821 | return error; | |
3822 | ||
acfa4380 | 3823 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3824 | return -EPERM; |
3825 | ||
1d2ef590 | 3826 | dget(dentry); |
5955102c | 3827 | inode_lock(dentry->d_inode); |
912dbc15 SW |
3828 | |
3829 | error = -EBUSY; | |
7af1364f | 3830 | if (is_local_mountpoint(dentry)) |
912dbc15 SW |
3831 | goto out; |
3832 | ||
3833 | error = security_inode_rmdir(dir, dentry); | |
3834 | if (error) | |
3835 | goto out; | |
3836 | ||
3cebde24 | 3837 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3838 | error = dir->i_op->rmdir(dir, dentry); |
3839 | if (error) | |
3840 | goto out; | |
3841 | ||
3842 | dentry->d_inode->i_flags |= S_DEAD; | |
3843 | dont_mount(dentry); | |
8ed936b5 | 3844 | detach_mounts(dentry); |
912dbc15 SW |
3845 | |
3846 | out: | |
5955102c | 3847 | inode_unlock(dentry->d_inode); |
1d2ef590 | 3848 | dput(dentry); |
912dbc15 | 3849 | if (!error) |
1da177e4 | 3850 | d_delete(dentry); |
1da177e4 LT |
3851 | return error; |
3852 | } | |
4d359507 | 3853 | EXPORT_SYMBOL(vfs_rmdir); |
1da177e4 | 3854 | |
5590ff0d | 3855 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3856 | { |
3857 | int error = 0; | |
91a27b2a | 3858 | struct filename *name; |
1da177e4 | 3859 | struct dentry *dentry; |
f5beed75 AV |
3860 | struct path path; |
3861 | struct qstr last; | |
3862 | int type; | |
c6ee9206 JL |
3863 | unsigned int lookup_flags = 0; |
3864 | retry: | |
c1d4dd27 AV |
3865 | name = filename_parentat(dfd, getname(pathname), lookup_flags, |
3866 | &path, &last, &type); | |
91a27b2a JL |
3867 | if (IS_ERR(name)) |
3868 | return PTR_ERR(name); | |
1da177e4 | 3869 | |
f5beed75 | 3870 | switch (type) { |
0612d9fb OH |
3871 | case LAST_DOTDOT: |
3872 | error = -ENOTEMPTY; | |
3873 | goto exit1; | |
3874 | case LAST_DOT: | |
3875 | error = -EINVAL; | |
3876 | goto exit1; | |
3877 | case LAST_ROOT: | |
3878 | error = -EBUSY; | |
3879 | goto exit1; | |
1da177e4 | 3880 | } |
0612d9fb | 3881 | |
f5beed75 | 3882 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3883 | if (error) |
3884 | goto exit1; | |
0612d9fb | 3885 | |
5955102c | 3886 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 3887 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 | 3888 | error = PTR_ERR(dentry); |
6902d925 DH |
3889 | if (IS_ERR(dentry)) |
3890 | goto exit2; | |
e6bc45d6 TT |
3891 | if (!dentry->d_inode) { |
3892 | error = -ENOENT; | |
3893 | goto exit3; | |
3894 | } | |
f5beed75 | 3895 | error = security_path_rmdir(&path, dentry); |
be6d3e56 | 3896 | if (error) |
c30dabfe | 3897 | goto exit3; |
f5beed75 | 3898 | error = vfs_rmdir(path.dentry->d_inode, dentry); |
0622753b | 3899 | exit3: |
6902d925 DH |
3900 | dput(dentry); |
3901 | exit2: | |
5955102c | 3902 | inode_unlock(path.dentry->d_inode); |
f5beed75 | 3903 | mnt_drop_write(path.mnt); |
1da177e4 | 3904 | exit1: |
f5beed75 | 3905 | path_put(&path); |
1da177e4 | 3906 | putname(name); |
c6ee9206 JL |
3907 | if (retry_estale(error, lookup_flags)) { |
3908 | lookup_flags |= LOOKUP_REVAL; | |
3909 | goto retry; | |
3910 | } | |
1da177e4 LT |
3911 | return error; |
3912 | } | |
3913 | ||
3cdad428 | 3914 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3915 | { |
3916 | return do_rmdir(AT_FDCWD, pathname); | |
3917 | } | |
3918 | ||
b21996e3 BF |
3919 | /** |
3920 | * vfs_unlink - unlink a filesystem object | |
3921 | * @dir: parent directory | |
3922 | * @dentry: victim | |
3923 | * @delegated_inode: returns victim inode, if the inode is delegated. | |
3924 | * | |
3925 | * The caller must hold dir->i_mutex. | |
3926 | * | |
3927 | * If vfs_unlink discovers a delegation, it will return -EWOULDBLOCK and | |
3928 | * return a reference to the inode in delegated_inode. The caller | |
3929 | * should then break the delegation on that inode and retry. Because | |
3930 | * breaking a delegation may take a long time, the caller should drop | |
3931 | * dir->i_mutex before doing so. | |
3932 | * | |
3933 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3934 | * be appropriate for callers that expect the underlying filesystem not | |
3935 | * to be NFS exported. | |
3936 | */ | |
3937 | int vfs_unlink(struct inode *dir, struct dentry *dentry, struct inode **delegated_inode) | |
1da177e4 | 3938 | { |
9accbb97 | 3939 | struct inode *target = dentry->d_inode; |
1da177e4 LT |
3940 | int error = may_delete(dir, dentry, 0); |
3941 | ||
3942 | if (error) | |
3943 | return error; | |
3944 | ||
acfa4380 | 3945 | if (!dir->i_op->unlink) |
1da177e4 LT |
3946 | return -EPERM; |
3947 | ||
5955102c | 3948 | inode_lock(target); |
8ed936b5 | 3949 | if (is_local_mountpoint(dentry)) |
1da177e4 LT |
3950 | error = -EBUSY; |
3951 | else { | |
3952 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3953 | if (!error) { |
5a14696c BF |
3954 | error = try_break_deleg(target, delegated_inode); |
3955 | if (error) | |
b21996e3 | 3956 | goto out; |
1da177e4 | 3957 | error = dir->i_op->unlink(dir, dentry); |
8ed936b5 | 3958 | if (!error) { |
d83c49f3 | 3959 | dont_mount(dentry); |
8ed936b5 EB |
3960 | detach_mounts(dentry); |
3961 | } | |
bec1052e | 3962 | } |
1da177e4 | 3963 | } |
b21996e3 | 3964 | out: |
5955102c | 3965 | inode_unlock(target); |
1da177e4 LT |
3966 | |
3967 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3968 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
9accbb97 | 3969 | fsnotify_link_count(target); |
e234f35c | 3970 | d_delete(dentry); |
1da177e4 | 3971 | } |
0eeca283 | 3972 | |
1da177e4 LT |
3973 | return error; |
3974 | } | |
4d359507 | 3975 | EXPORT_SYMBOL(vfs_unlink); |
1da177e4 LT |
3976 | |
3977 | /* | |
3978 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3979 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3980 | * writeout happening, and we don't want to prevent access to the directory |
3981 | * while waiting on the I/O. | |
3982 | */ | |
da2f1362 | 3983 | long do_unlinkat(int dfd, struct filename *name) |
1da177e4 | 3984 | { |
2ad94ae6 | 3985 | int error; |
1da177e4 | 3986 | struct dentry *dentry; |
f5beed75 AV |
3987 | struct path path; |
3988 | struct qstr last; | |
3989 | int type; | |
1da177e4 | 3990 | struct inode *inode = NULL; |
b21996e3 | 3991 | struct inode *delegated_inode = NULL; |
5d18f813 JL |
3992 | unsigned int lookup_flags = 0; |
3993 | retry: | |
da2f1362 | 3994 | name = filename_parentat(dfd, name, lookup_flags, &path, &last, &type); |
91a27b2a JL |
3995 | if (IS_ERR(name)) |
3996 | return PTR_ERR(name); | |
2ad94ae6 | 3997 | |
1da177e4 | 3998 | error = -EISDIR; |
f5beed75 | 3999 | if (type != LAST_NORM) |
1da177e4 | 4000 | goto exit1; |
0612d9fb | 4001 | |
f5beed75 | 4002 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
4003 | if (error) |
4004 | goto exit1; | |
b21996e3 | 4005 | retry_deleg: |
5955102c | 4006 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 4007 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 LT |
4008 | error = PTR_ERR(dentry); |
4009 | if (!IS_ERR(dentry)) { | |
4010 | /* Why not before? Because we want correct error value */ | |
f5beed75 | 4011 | if (last.name[last.len]) |
50338b88 | 4012 | goto slashes; |
1da177e4 | 4013 | inode = dentry->d_inode; |
b18825a7 | 4014 | if (d_is_negative(dentry)) |
e6bc45d6 TT |
4015 | goto slashes; |
4016 | ihold(inode); | |
f5beed75 | 4017 | error = security_path_unlink(&path, dentry); |
be6d3e56 | 4018 | if (error) |
c30dabfe | 4019 | goto exit2; |
f5beed75 | 4020 | error = vfs_unlink(path.dentry->d_inode, dentry, &delegated_inode); |
c30dabfe | 4021 | exit2: |
1da177e4 LT |
4022 | dput(dentry); |
4023 | } | |
5955102c | 4024 | inode_unlock(path.dentry->d_inode); |
1da177e4 LT |
4025 | if (inode) |
4026 | iput(inode); /* truncate the inode here */ | |
b21996e3 BF |
4027 | inode = NULL; |
4028 | if (delegated_inode) { | |
5a14696c | 4029 | error = break_deleg_wait(&delegated_inode); |
b21996e3 BF |
4030 | if (!error) |
4031 | goto retry_deleg; | |
4032 | } | |
f5beed75 | 4033 | mnt_drop_write(path.mnt); |
1da177e4 | 4034 | exit1: |
f5beed75 | 4035 | path_put(&path); |
5d18f813 JL |
4036 | if (retry_estale(error, lookup_flags)) { |
4037 | lookup_flags |= LOOKUP_REVAL; | |
4038 | inode = NULL; | |
4039 | goto retry; | |
4040 | } | |
da2f1362 | 4041 | putname(name); |
1da177e4 LT |
4042 | return error; |
4043 | ||
4044 | slashes: | |
b18825a7 DH |
4045 | if (d_is_negative(dentry)) |
4046 | error = -ENOENT; | |
44b1d530 | 4047 | else if (d_is_dir(dentry)) |
b18825a7 DH |
4048 | error = -EISDIR; |
4049 | else | |
4050 | error = -ENOTDIR; | |
1da177e4 LT |
4051 | goto exit2; |
4052 | } | |
4053 | ||
2e4d0924 | 4054 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
4055 | { |
4056 | if ((flag & ~AT_REMOVEDIR) != 0) | |
4057 | return -EINVAL; | |
4058 | ||
4059 | if (flag & AT_REMOVEDIR) | |
4060 | return do_rmdir(dfd, pathname); | |
4061 | ||
da2f1362 | 4062 | return do_unlinkat(dfd, getname(pathname)); |
5590ff0d UD |
4063 | } |
4064 | ||
3480b257 | 4065 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d | 4066 | { |
da2f1362 | 4067 | return do_unlinkat(AT_FDCWD, getname(pathname)); |
5590ff0d UD |
4068 | } |
4069 | ||
db2e747b | 4070 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 4071 | { |
a95164d9 | 4072 | int error = may_create(dir, dentry); |
1da177e4 LT |
4073 | |
4074 | if (error) | |
4075 | return error; | |
4076 | ||
acfa4380 | 4077 | if (!dir->i_op->symlink) |
1da177e4 LT |
4078 | return -EPERM; |
4079 | ||
4080 | error = security_inode_symlink(dir, dentry, oldname); | |
4081 | if (error) | |
4082 | return error; | |
4083 | ||
1da177e4 | 4084 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 4085 | if (!error) |
f38aa942 | 4086 | fsnotify_create(dir, dentry); |
1da177e4 LT |
4087 | return error; |
4088 | } | |
4d359507 | 4089 | EXPORT_SYMBOL(vfs_symlink); |
1da177e4 | 4090 | |
2e4d0924 HC |
4091 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
4092 | int, newdfd, const char __user *, newname) | |
1da177e4 | 4093 | { |
2ad94ae6 | 4094 | int error; |
91a27b2a | 4095 | struct filename *from; |
6902d925 | 4096 | struct dentry *dentry; |
dae6ad8f | 4097 | struct path path; |
f46d3567 | 4098 | unsigned int lookup_flags = 0; |
1da177e4 LT |
4099 | |
4100 | from = getname(oldname); | |
2ad94ae6 | 4101 | if (IS_ERR(from)) |
1da177e4 | 4102 | return PTR_ERR(from); |
f46d3567 JL |
4103 | retry: |
4104 | dentry = user_path_create(newdfd, newname, &path, lookup_flags); | |
6902d925 DH |
4105 | error = PTR_ERR(dentry); |
4106 | if (IS_ERR(dentry)) | |
dae6ad8f | 4107 | goto out_putname; |
6902d925 | 4108 | |
91a27b2a | 4109 | error = security_path_symlink(&path, dentry, from->name); |
a8104a9f | 4110 | if (!error) |
91a27b2a | 4111 | error = vfs_symlink(path.dentry->d_inode, dentry, from->name); |
921a1650 | 4112 | done_path_create(&path, dentry); |
f46d3567 JL |
4113 | if (retry_estale(error, lookup_flags)) { |
4114 | lookup_flags |= LOOKUP_REVAL; | |
4115 | goto retry; | |
4116 | } | |
6902d925 | 4117 | out_putname: |
1da177e4 LT |
4118 | putname(from); |
4119 | return error; | |
4120 | } | |
4121 | ||
3480b257 | 4122 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
4123 | { |
4124 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
4125 | } | |
4126 | ||
146a8595 BF |
4127 | /** |
4128 | * vfs_link - create a new link | |
4129 | * @old_dentry: object to be linked | |
4130 | * @dir: new parent | |
4131 | * @new_dentry: where to create the new link | |
4132 | * @delegated_inode: returns inode needing a delegation break | |
4133 | * | |
4134 | * The caller must hold dir->i_mutex | |
4135 | * | |
4136 | * If vfs_link discovers a delegation on the to-be-linked file in need | |
4137 | * of breaking, it will return -EWOULDBLOCK and return a reference to the | |
4138 | * inode in delegated_inode. The caller should then break the delegation | |
4139 | * and retry. Because breaking a delegation may take a long time, the | |
4140 | * caller should drop the i_mutex before doing so. | |
4141 | * | |
4142 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4143 | * be appropriate for callers that expect the underlying filesystem not | |
4144 | * to be NFS exported. | |
4145 | */ | |
4146 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, struct inode **delegated_inode) | |
1da177e4 LT |
4147 | { |
4148 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 4149 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
4150 | int error; |
4151 | ||
4152 | if (!inode) | |
4153 | return -ENOENT; | |
4154 | ||
a95164d9 | 4155 | error = may_create(dir, new_dentry); |
1da177e4 LT |
4156 | if (error) |
4157 | return error; | |
4158 | ||
4159 | if (dir->i_sb != inode->i_sb) | |
4160 | return -EXDEV; | |
4161 | ||
4162 | /* | |
4163 | * A link to an append-only or immutable file cannot be created. | |
4164 | */ | |
4165 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
4166 | return -EPERM; | |
0bd23d09 EB |
4167 | /* |
4168 | * Updating the link count will likely cause i_uid and i_gid to | |
4169 | * be writen back improperly if their true value is unknown to | |
4170 | * the vfs. | |
4171 | */ | |
4172 | if (HAS_UNMAPPED_ID(inode)) | |
4173 | return -EPERM; | |
acfa4380 | 4174 | if (!dir->i_op->link) |
1da177e4 | 4175 | return -EPERM; |
7e79eedb | 4176 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
4177 | return -EPERM; |
4178 | ||
4179 | error = security_inode_link(old_dentry, dir, new_dentry); | |
4180 | if (error) | |
4181 | return error; | |
4182 | ||
5955102c | 4183 | inode_lock(inode); |
aae8a97d | 4184 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
f4e0c30c | 4185 | if (inode->i_nlink == 0 && !(inode->i_state & I_LINKABLE)) |
aae8a97d | 4186 | error = -ENOENT; |
8de52778 AV |
4187 | else if (max_links && inode->i_nlink >= max_links) |
4188 | error = -EMLINK; | |
146a8595 BF |
4189 | else { |
4190 | error = try_break_deleg(inode, delegated_inode); | |
4191 | if (!error) | |
4192 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
4193 | } | |
f4e0c30c AV |
4194 | |
4195 | if (!error && (inode->i_state & I_LINKABLE)) { | |
4196 | spin_lock(&inode->i_lock); | |
4197 | inode->i_state &= ~I_LINKABLE; | |
4198 | spin_unlock(&inode->i_lock); | |
4199 | } | |
5955102c | 4200 | inode_unlock(inode); |
e31e14ec | 4201 | if (!error) |
7e79eedb | 4202 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
4203 | return error; |
4204 | } | |
4d359507 | 4205 | EXPORT_SYMBOL(vfs_link); |
1da177e4 LT |
4206 | |
4207 | /* | |
4208 | * Hardlinks are often used in delicate situations. We avoid | |
4209 | * security-related surprises by not following symlinks on the | |
4210 | * newname. --KAB | |
4211 | * | |
4212 | * We don't follow them on the oldname either to be compatible | |
4213 | * with linux 2.0, and to avoid hard-linking to directories | |
4214 | * and other special files. --ADM | |
4215 | */ | |
2e4d0924 HC |
4216 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
4217 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
4218 | { |
4219 | struct dentry *new_dentry; | |
dae6ad8f | 4220 | struct path old_path, new_path; |
146a8595 | 4221 | struct inode *delegated_inode = NULL; |
11a7b371 | 4222 | int how = 0; |
1da177e4 | 4223 | int error; |
1da177e4 | 4224 | |
11a7b371 | 4225 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 4226 | return -EINVAL; |
11a7b371 | 4227 | /* |
f0cc6ffb LT |
4228 | * To use null names we require CAP_DAC_READ_SEARCH |
4229 | * This ensures that not everyone will be able to create | |
4230 | * handlink using the passed filedescriptor. | |
11a7b371 | 4231 | */ |
f0cc6ffb LT |
4232 | if (flags & AT_EMPTY_PATH) { |
4233 | if (!capable(CAP_DAC_READ_SEARCH)) | |
4234 | return -ENOENT; | |
11a7b371 | 4235 | how = LOOKUP_EMPTY; |
f0cc6ffb | 4236 | } |
11a7b371 AK |
4237 | |
4238 | if (flags & AT_SYMLINK_FOLLOW) | |
4239 | how |= LOOKUP_FOLLOW; | |
442e31ca | 4240 | retry: |
11a7b371 | 4241 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 4242 | if (error) |
2ad94ae6 AV |
4243 | return error; |
4244 | ||
442e31ca JL |
4245 | new_dentry = user_path_create(newdfd, newname, &new_path, |
4246 | (how & LOOKUP_REVAL)); | |
1da177e4 | 4247 | error = PTR_ERR(new_dentry); |
6902d925 | 4248 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
4249 | goto out; |
4250 | ||
4251 | error = -EXDEV; | |
4252 | if (old_path.mnt != new_path.mnt) | |
4253 | goto out_dput; | |
800179c9 KC |
4254 | error = may_linkat(&old_path); |
4255 | if (unlikely(error)) | |
4256 | goto out_dput; | |
dae6ad8f | 4257 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 4258 | if (error) |
a8104a9f | 4259 | goto out_dput; |
146a8595 | 4260 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry, &delegated_inode); |
75c3f29d | 4261 | out_dput: |
921a1650 | 4262 | done_path_create(&new_path, new_dentry); |
146a8595 BF |
4263 | if (delegated_inode) { |
4264 | error = break_deleg_wait(&delegated_inode); | |
d22e6338 OD |
4265 | if (!error) { |
4266 | path_put(&old_path); | |
146a8595 | 4267 | goto retry; |
d22e6338 | 4268 | } |
146a8595 | 4269 | } |
442e31ca | 4270 | if (retry_estale(error, how)) { |
d22e6338 | 4271 | path_put(&old_path); |
442e31ca JL |
4272 | how |= LOOKUP_REVAL; |
4273 | goto retry; | |
4274 | } | |
1da177e4 | 4275 | out: |
2d8f3038 | 4276 | path_put(&old_path); |
1da177e4 LT |
4277 | |
4278 | return error; | |
4279 | } | |
4280 | ||
3480b257 | 4281 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4282 | { |
c04030e1 | 4283 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4284 | } |
4285 | ||
bc27027a MS |
4286 | /** |
4287 | * vfs_rename - rename a filesystem object | |
4288 | * @old_dir: parent of source | |
4289 | * @old_dentry: source | |
4290 | * @new_dir: parent of destination | |
4291 | * @new_dentry: destination | |
4292 | * @delegated_inode: returns an inode needing a delegation break | |
520c8b16 | 4293 | * @flags: rename flags |
bc27027a MS |
4294 | * |
4295 | * The caller must hold multiple mutexes--see lock_rename()). | |
4296 | * | |
4297 | * If vfs_rename discovers a delegation in need of breaking at either | |
4298 | * the source or destination, it will return -EWOULDBLOCK and return a | |
4299 | * reference to the inode in delegated_inode. The caller should then | |
4300 | * break the delegation and retry. Because breaking a delegation may | |
4301 | * take a long time, the caller should drop all locks before doing | |
4302 | * so. | |
4303 | * | |
4304 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4305 | * be appropriate for callers that expect the underlying filesystem not | |
4306 | * to be NFS exported. | |
4307 | * | |
1da177e4 LT |
4308 | * The worst of all namespace operations - renaming directory. "Perverted" |
4309 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
4310 | * Problems: | |
0117d427 | 4311 | * |
d03b29a2 | 4312 | * a) we can get into loop creation. |
1da177e4 LT |
4313 | * b) race potential - two innocent renames can create a loop together. |
4314 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 4315 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 | 4316 | * story. |
6cedba89 BF |
4317 | * c) we have to lock _four_ objects - parents and victim (if it exists), |
4318 | * and source (if it is not a directory). | |
1b1dcc1b | 4319 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
4320 | * whether the target exists). Solution: try to be smart with locking |
4321 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 4322 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
4323 | * move will be locked. Thus we can rank directories by the tree |
4324 | * (ancestors first) and rank all non-directories after them. | |
4325 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 4326 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
4327 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
4328 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
4329 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 4330 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 4331 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 4332 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 4333 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
4334 | * locking]. |
4335 | */ | |
bc27027a MS |
4336 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, |
4337 | struct inode *new_dir, struct dentry *new_dentry, | |
520c8b16 | 4338 | struct inode **delegated_inode, unsigned int flags) |
1da177e4 | 4339 | { |
bc27027a MS |
4340 | int error; |
4341 | bool is_dir = d_is_dir(old_dentry); | |
bc27027a | 4342 | struct inode *source = old_dentry->d_inode; |
9055cba7 | 4343 | struct inode *target = new_dentry->d_inode; |
da1ce067 MS |
4344 | bool new_is_dir = false; |
4345 | unsigned max_links = new_dir->i_sb->s_max_links; | |
49d31c2f | 4346 | struct name_snapshot old_name; |
bc27027a | 4347 | |
8d3e2936 | 4348 | if (source == target) |
bc27027a MS |
4349 | return 0; |
4350 | ||
4351 | error = may_delete(old_dir, old_dentry, is_dir); | |
4352 | if (error) | |
4353 | return error; | |
4354 | ||
da1ce067 | 4355 | if (!target) { |
bc27027a | 4356 | error = may_create(new_dir, new_dentry); |
da1ce067 MS |
4357 | } else { |
4358 | new_is_dir = d_is_dir(new_dentry); | |
4359 | ||
4360 | if (!(flags & RENAME_EXCHANGE)) | |
4361 | error = may_delete(new_dir, new_dentry, is_dir); | |
4362 | else | |
4363 | error = may_delete(new_dir, new_dentry, new_is_dir); | |
4364 | } | |
bc27027a MS |
4365 | if (error) |
4366 | return error; | |
4367 | ||
2773bf00 | 4368 | if (!old_dir->i_op->rename) |
bc27027a | 4369 | return -EPERM; |
1da177e4 LT |
4370 | |
4371 | /* | |
4372 | * If we are going to change the parent - check write permissions, | |
4373 | * we'll need to flip '..'. | |
4374 | */ | |
da1ce067 MS |
4375 | if (new_dir != old_dir) { |
4376 | if (is_dir) { | |
4377 | error = inode_permission(source, MAY_WRITE); | |
4378 | if (error) | |
4379 | return error; | |
4380 | } | |
4381 | if ((flags & RENAME_EXCHANGE) && new_is_dir) { | |
4382 | error = inode_permission(target, MAY_WRITE); | |
4383 | if (error) | |
4384 | return error; | |
4385 | } | |
1da177e4 LT |
4386 | } |
4387 | ||
0b3974eb MS |
4388 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry, |
4389 | flags); | |
1da177e4 LT |
4390 | if (error) |
4391 | return error; | |
4392 | ||
49d31c2f | 4393 | take_dentry_name_snapshot(&old_name, old_dentry); |
1d2ef590 | 4394 | dget(new_dentry); |
da1ce067 | 4395 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4396 | lock_two_nondirectories(source, target); |
4397 | else if (target) | |
5955102c | 4398 | inode_lock(target); |
9055cba7 SW |
4399 | |
4400 | error = -EBUSY; | |
7af1364f | 4401 | if (is_local_mountpoint(old_dentry) || is_local_mountpoint(new_dentry)) |
9055cba7 SW |
4402 | goto out; |
4403 | ||
da1ce067 | 4404 | if (max_links && new_dir != old_dir) { |
bc27027a | 4405 | error = -EMLINK; |
da1ce067 | 4406 | if (is_dir && !new_is_dir && new_dir->i_nlink >= max_links) |
bc27027a | 4407 | goto out; |
da1ce067 MS |
4408 | if ((flags & RENAME_EXCHANGE) && !is_dir && new_is_dir && |
4409 | old_dir->i_nlink >= max_links) | |
4410 | goto out; | |
4411 | } | |
4412 | if (is_dir && !(flags & RENAME_EXCHANGE) && target) | |
4413 | shrink_dcache_parent(new_dentry); | |
4414 | if (!is_dir) { | |
bc27027a | 4415 | error = try_break_deleg(source, delegated_inode); |
8e6d782c BF |
4416 | if (error) |
4417 | goto out; | |
da1ce067 MS |
4418 | } |
4419 | if (target && !new_is_dir) { | |
4420 | error = try_break_deleg(target, delegated_inode); | |
4421 | if (error) | |
4422 | goto out; | |
8e6d782c | 4423 | } |
2773bf00 | 4424 | error = old_dir->i_op->rename(old_dir, old_dentry, |
18fc84da | 4425 | new_dir, new_dentry, flags); |
51892bbb SW |
4426 | if (error) |
4427 | goto out; | |
4428 | ||
da1ce067 | 4429 | if (!(flags & RENAME_EXCHANGE) && target) { |
bc27027a MS |
4430 | if (is_dir) |
4431 | target->i_flags |= S_DEAD; | |
51892bbb | 4432 | dont_mount(new_dentry); |
8ed936b5 | 4433 | detach_mounts(new_dentry); |
bc27027a | 4434 | } |
da1ce067 MS |
4435 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) { |
4436 | if (!(flags & RENAME_EXCHANGE)) | |
4437 | d_move(old_dentry, new_dentry); | |
4438 | else | |
4439 | d_exchange(old_dentry, new_dentry); | |
4440 | } | |
51892bbb | 4441 | out: |
da1ce067 | 4442 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4443 | unlock_two_nondirectories(source, target); |
4444 | else if (target) | |
5955102c | 4445 | inode_unlock(target); |
1da177e4 | 4446 | dput(new_dentry); |
da1ce067 | 4447 | if (!error) { |
49d31c2f | 4448 | fsnotify_move(old_dir, new_dir, old_name.name, is_dir, |
da1ce067 MS |
4449 | !(flags & RENAME_EXCHANGE) ? target : NULL, old_dentry); |
4450 | if (flags & RENAME_EXCHANGE) { | |
4451 | fsnotify_move(new_dir, old_dir, old_dentry->d_name.name, | |
4452 | new_is_dir, NULL, new_dentry); | |
4453 | } | |
4454 | } | |
49d31c2f | 4455 | release_dentry_name_snapshot(&old_name); |
0eeca283 | 4456 | |
1da177e4 LT |
4457 | return error; |
4458 | } | |
4d359507 | 4459 | EXPORT_SYMBOL(vfs_rename); |
1da177e4 | 4460 | |
520c8b16 MS |
4461 | SYSCALL_DEFINE5(renameat2, int, olddfd, const char __user *, oldname, |
4462 | int, newdfd, const char __user *, newname, unsigned int, flags) | |
1da177e4 | 4463 | { |
2ad94ae6 AV |
4464 | struct dentry *old_dentry, *new_dentry; |
4465 | struct dentry *trap; | |
f5beed75 AV |
4466 | struct path old_path, new_path; |
4467 | struct qstr old_last, new_last; | |
4468 | int old_type, new_type; | |
8e6d782c | 4469 | struct inode *delegated_inode = NULL; |
91a27b2a JL |
4470 | struct filename *from; |
4471 | struct filename *to; | |
f5beed75 | 4472 | unsigned int lookup_flags = 0, target_flags = LOOKUP_RENAME_TARGET; |
c6a94284 | 4473 | bool should_retry = false; |
2ad94ae6 | 4474 | int error; |
520c8b16 | 4475 | |
0d7a8555 | 4476 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
da1ce067 MS |
4477 | return -EINVAL; |
4478 | ||
0d7a8555 MS |
4479 | if ((flags & (RENAME_NOREPLACE | RENAME_WHITEOUT)) && |
4480 | (flags & RENAME_EXCHANGE)) | |
520c8b16 MS |
4481 | return -EINVAL; |
4482 | ||
0d7a8555 MS |
4483 | if ((flags & RENAME_WHITEOUT) && !capable(CAP_MKNOD)) |
4484 | return -EPERM; | |
4485 | ||
f5beed75 AV |
4486 | if (flags & RENAME_EXCHANGE) |
4487 | target_flags = 0; | |
4488 | ||
c6a94284 | 4489 | retry: |
c1d4dd27 AV |
4490 | from = filename_parentat(olddfd, getname(oldname), lookup_flags, |
4491 | &old_path, &old_last, &old_type); | |
91a27b2a JL |
4492 | if (IS_ERR(from)) { |
4493 | error = PTR_ERR(from); | |
1da177e4 | 4494 | goto exit; |
91a27b2a | 4495 | } |
1da177e4 | 4496 | |
c1d4dd27 AV |
4497 | to = filename_parentat(newdfd, getname(newname), lookup_flags, |
4498 | &new_path, &new_last, &new_type); | |
91a27b2a JL |
4499 | if (IS_ERR(to)) { |
4500 | error = PTR_ERR(to); | |
1da177e4 | 4501 | goto exit1; |
91a27b2a | 4502 | } |
1da177e4 LT |
4503 | |
4504 | error = -EXDEV; | |
f5beed75 | 4505 | if (old_path.mnt != new_path.mnt) |
1da177e4 LT |
4506 | goto exit2; |
4507 | ||
1da177e4 | 4508 | error = -EBUSY; |
f5beed75 | 4509 | if (old_type != LAST_NORM) |
1da177e4 LT |
4510 | goto exit2; |
4511 | ||
0a7c3937 MS |
4512 | if (flags & RENAME_NOREPLACE) |
4513 | error = -EEXIST; | |
f5beed75 | 4514 | if (new_type != LAST_NORM) |
1da177e4 LT |
4515 | goto exit2; |
4516 | ||
f5beed75 | 4517 | error = mnt_want_write(old_path.mnt); |
c30dabfe JK |
4518 | if (error) |
4519 | goto exit2; | |
4520 | ||
8e6d782c | 4521 | retry_deleg: |
f5beed75 | 4522 | trap = lock_rename(new_path.dentry, old_path.dentry); |
1da177e4 | 4523 | |
f5beed75 | 4524 | old_dentry = __lookup_hash(&old_last, old_path.dentry, lookup_flags); |
1da177e4 LT |
4525 | error = PTR_ERR(old_dentry); |
4526 | if (IS_ERR(old_dentry)) | |
4527 | goto exit3; | |
4528 | /* source must exist */ | |
4529 | error = -ENOENT; | |
b18825a7 | 4530 | if (d_is_negative(old_dentry)) |
1da177e4 | 4531 | goto exit4; |
f5beed75 | 4532 | new_dentry = __lookup_hash(&new_last, new_path.dentry, lookup_flags | target_flags); |
0a7c3937 MS |
4533 | error = PTR_ERR(new_dentry); |
4534 | if (IS_ERR(new_dentry)) | |
4535 | goto exit4; | |
4536 | error = -EEXIST; | |
4537 | if ((flags & RENAME_NOREPLACE) && d_is_positive(new_dentry)) | |
4538 | goto exit5; | |
da1ce067 MS |
4539 | if (flags & RENAME_EXCHANGE) { |
4540 | error = -ENOENT; | |
4541 | if (d_is_negative(new_dentry)) | |
4542 | goto exit5; | |
4543 | ||
4544 | if (!d_is_dir(new_dentry)) { | |
4545 | error = -ENOTDIR; | |
f5beed75 | 4546 | if (new_last.name[new_last.len]) |
da1ce067 MS |
4547 | goto exit5; |
4548 | } | |
4549 | } | |
1da177e4 | 4550 | /* unless the source is a directory trailing slashes give -ENOTDIR */ |
44b1d530 | 4551 | if (!d_is_dir(old_dentry)) { |
1da177e4 | 4552 | error = -ENOTDIR; |
f5beed75 | 4553 | if (old_last.name[old_last.len]) |
0a7c3937 | 4554 | goto exit5; |
f5beed75 | 4555 | if (!(flags & RENAME_EXCHANGE) && new_last.name[new_last.len]) |
0a7c3937 | 4556 | goto exit5; |
1da177e4 LT |
4557 | } |
4558 | /* source should not be ancestor of target */ | |
4559 | error = -EINVAL; | |
4560 | if (old_dentry == trap) | |
0a7c3937 | 4561 | goto exit5; |
1da177e4 | 4562 | /* target should not be an ancestor of source */ |
da1ce067 MS |
4563 | if (!(flags & RENAME_EXCHANGE)) |
4564 | error = -ENOTEMPTY; | |
1da177e4 LT |
4565 | if (new_dentry == trap) |
4566 | goto exit5; | |
4567 | ||
f5beed75 AV |
4568 | error = security_path_rename(&old_path, old_dentry, |
4569 | &new_path, new_dentry, flags); | |
be6d3e56 | 4570 | if (error) |
c30dabfe | 4571 | goto exit5; |
f5beed75 AV |
4572 | error = vfs_rename(old_path.dentry->d_inode, old_dentry, |
4573 | new_path.dentry->d_inode, new_dentry, | |
520c8b16 | 4574 | &delegated_inode, flags); |
1da177e4 LT |
4575 | exit5: |
4576 | dput(new_dentry); | |
4577 | exit4: | |
4578 | dput(old_dentry); | |
4579 | exit3: | |
f5beed75 | 4580 | unlock_rename(new_path.dentry, old_path.dentry); |
8e6d782c BF |
4581 | if (delegated_inode) { |
4582 | error = break_deleg_wait(&delegated_inode); | |
4583 | if (!error) | |
4584 | goto retry_deleg; | |
4585 | } | |
f5beed75 | 4586 | mnt_drop_write(old_path.mnt); |
1da177e4 | 4587 | exit2: |
c6a94284 JL |
4588 | if (retry_estale(error, lookup_flags)) |
4589 | should_retry = true; | |
f5beed75 | 4590 | path_put(&new_path); |
2ad94ae6 | 4591 | putname(to); |
1da177e4 | 4592 | exit1: |
f5beed75 | 4593 | path_put(&old_path); |
1da177e4 | 4594 | putname(from); |
c6a94284 JL |
4595 | if (should_retry) { |
4596 | should_retry = false; | |
4597 | lookup_flags |= LOOKUP_REVAL; | |
4598 | goto retry; | |
4599 | } | |
2ad94ae6 | 4600 | exit: |
1da177e4 LT |
4601 | return error; |
4602 | } | |
4603 | ||
520c8b16 MS |
4604 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
4605 | int, newdfd, const char __user *, newname) | |
4606 | { | |
4607 | return sys_renameat2(olddfd, oldname, newdfd, newname, 0); | |
4608 | } | |
4609 | ||
a26eab24 | 4610 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4611 | { |
520c8b16 | 4612 | return sys_renameat2(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4613 | } |
4614 | ||
787fb6bc MS |
4615 | int vfs_whiteout(struct inode *dir, struct dentry *dentry) |
4616 | { | |
4617 | int error = may_create(dir, dentry); | |
4618 | if (error) | |
4619 | return error; | |
4620 | ||
4621 | if (!dir->i_op->mknod) | |
4622 | return -EPERM; | |
4623 | ||
4624 | return dir->i_op->mknod(dir, dentry, | |
4625 | S_IFCHR | WHITEOUT_MODE, WHITEOUT_DEV); | |
4626 | } | |
4627 | EXPORT_SYMBOL(vfs_whiteout); | |
4628 | ||
5d826c84 | 4629 | int readlink_copy(char __user *buffer, int buflen, const char *link) |
1da177e4 | 4630 | { |
5d826c84 | 4631 | int len = PTR_ERR(link); |
1da177e4 LT |
4632 | if (IS_ERR(link)) |
4633 | goto out; | |
4634 | ||
4635 | len = strlen(link); | |
4636 | if (len > (unsigned) buflen) | |
4637 | len = buflen; | |
4638 | if (copy_to_user(buffer, link, len)) | |
4639 | len = -EFAULT; | |
4640 | out: | |
4641 | return len; | |
4642 | } | |
4643 | ||
4644 | /* | |
4645 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
6b255391 AV |
4646 | * have ->get_link() not calling nd_jump_link(). Using (or not using) it |
4647 | * for any given inode is up to filesystem. | |
1da177e4 | 4648 | */ |
d16744ec MS |
4649 | static int generic_readlink(struct dentry *dentry, char __user *buffer, |
4650 | int buflen) | |
1da177e4 | 4651 | { |
fceef393 | 4652 | DEFINE_DELAYED_CALL(done); |
5f2c4179 AV |
4653 | struct inode *inode = d_inode(dentry); |
4654 | const char *link = inode->i_link; | |
694a1764 | 4655 | int res; |
cc314eef | 4656 | |
d4dee48b | 4657 | if (!link) { |
fceef393 | 4658 | link = inode->i_op->get_link(dentry, inode, &done); |
d4dee48b AV |
4659 | if (IS_ERR(link)) |
4660 | return PTR_ERR(link); | |
4661 | } | |
680baacb | 4662 | res = readlink_copy(buffer, buflen, link); |
fceef393 | 4663 | do_delayed_call(&done); |
694a1764 | 4664 | return res; |
1da177e4 LT |
4665 | } |
4666 | ||
fd4a0edf MS |
4667 | /** |
4668 | * vfs_readlink - copy symlink body into userspace buffer | |
4669 | * @dentry: dentry on which to get symbolic link | |
4670 | * @buffer: user memory pointer | |
4671 | * @buflen: size of buffer | |
4672 | * | |
4673 | * Does not touch atime. That's up to the caller if necessary | |
4674 | * | |
4675 | * Does not call security hook. | |
4676 | */ | |
4677 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4678 | { | |
4679 | struct inode *inode = d_inode(dentry); | |
4680 | ||
76fca90e MS |
4681 | if (unlikely(!(inode->i_opflags & IOP_DEFAULT_READLINK))) { |
4682 | if (unlikely(inode->i_op->readlink)) | |
4683 | return inode->i_op->readlink(dentry, buffer, buflen); | |
4684 | ||
4685 | if (!d_is_symlink(dentry)) | |
4686 | return -EINVAL; | |
4687 | ||
4688 | spin_lock(&inode->i_lock); | |
4689 | inode->i_opflags |= IOP_DEFAULT_READLINK; | |
4690 | spin_unlock(&inode->i_lock); | |
4691 | } | |
fd4a0edf | 4692 | |
76fca90e | 4693 | return generic_readlink(dentry, buffer, buflen); |
fd4a0edf MS |
4694 | } |
4695 | EXPORT_SYMBOL(vfs_readlink); | |
1da177e4 | 4696 | |
d60874cd MS |
4697 | /** |
4698 | * vfs_get_link - get symlink body | |
4699 | * @dentry: dentry on which to get symbolic link | |
4700 | * @done: caller needs to free returned data with this | |
4701 | * | |
4702 | * Calls security hook and i_op->get_link() on the supplied inode. | |
4703 | * | |
4704 | * It does not touch atime. That's up to the caller if necessary. | |
4705 | * | |
4706 | * Does not work on "special" symlinks like /proc/$$/fd/N | |
4707 | */ | |
4708 | const char *vfs_get_link(struct dentry *dentry, struct delayed_call *done) | |
4709 | { | |
4710 | const char *res = ERR_PTR(-EINVAL); | |
4711 | struct inode *inode = d_inode(dentry); | |
4712 | ||
4713 | if (d_is_symlink(dentry)) { | |
4714 | res = ERR_PTR(security_inode_readlink(dentry)); | |
4715 | if (!res) | |
4716 | res = inode->i_op->get_link(dentry, inode, done); | |
4717 | } | |
4718 | return res; | |
4719 | } | |
4720 | EXPORT_SYMBOL(vfs_get_link); | |
4721 | ||
1da177e4 | 4722 | /* get the link contents into pagecache */ |
6b255391 | 4723 | const char *page_get_link(struct dentry *dentry, struct inode *inode, |
fceef393 | 4724 | struct delayed_call *callback) |
1da177e4 | 4725 | { |
ebd09abb DG |
4726 | char *kaddr; |
4727 | struct page *page; | |
6b255391 AV |
4728 | struct address_space *mapping = inode->i_mapping; |
4729 | ||
d3883d4f AV |
4730 | if (!dentry) { |
4731 | page = find_get_page(mapping, 0); | |
4732 | if (!page) | |
4733 | return ERR_PTR(-ECHILD); | |
4734 | if (!PageUptodate(page)) { | |
4735 | put_page(page); | |
4736 | return ERR_PTR(-ECHILD); | |
4737 | } | |
4738 | } else { | |
4739 | page = read_mapping_page(mapping, 0, NULL); | |
4740 | if (IS_ERR(page)) | |
4741 | return (char*)page; | |
4742 | } | |
fceef393 | 4743 | set_delayed_call(callback, page_put_link, page); |
21fc61c7 AV |
4744 | BUG_ON(mapping_gfp_mask(mapping) & __GFP_HIGHMEM); |
4745 | kaddr = page_address(page); | |
6b255391 | 4746 | nd_terminate_link(kaddr, inode->i_size, PAGE_SIZE - 1); |
ebd09abb | 4747 | return kaddr; |
1da177e4 LT |
4748 | } |
4749 | ||
6b255391 | 4750 | EXPORT_SYMBOL(page_get_link); |
1da177e4 | 4751 | |
fceef393 | 4752 | void page_put_link(void *arg) |
1da177e4 | 4753 | { |
fceef393 | 4754 | put_page(arg); |
1da177e4 | 4755 | } |
4d359507 | 4756 | EXPORT_SYMBOL(page_put_link); |
1da177e4 | 4757 | |
aa80deab AV |
4758 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) |
4759 | { | |
fceef393 | 4760 | DEFINE_DELAYED_CALL(done); |
6b255391 AV |
4761 | int res = readlink_copy(buffer, buflen, |
4762 | page_get_link(dentry, d_inode(dentry), | |
fceef393 AV |
4763 | &done)); |
4764 | do_delayed_call(&done); | |
aa80deab AV |
4765 | return res; |
4766 | } | |
4767 | EXPORT_SYMBOL(page_readlink); | |
4768 | ||
54566b2c NP |
4769 | /* |
4770 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
4771 | */ | |
4772 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
4773 | { |
4774 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 4775 | struct page *page; |
afddba49 | 4776 | void *fsdata; |
beb497ab | 4777 | int err; |
c718a975 | 4778 | unsigned int flags = 0; |
54566b2c NP |
4779 | if (nofs) |
4780 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 4781 | |
7e53cac4 | 4782 | retry: |
afddba49 | 4783 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 4784 | flags, &page, &fsdata); |
1da177e4 | 4785 | if (err) |
afddba49 NP |
4786 | goto fail; |
4787 | ||
21fc61c7 | 4788 | memcpy(page_address(page), symname, len-1); |
afddba49 NP |
4789 | |
4790 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
4791 | page, fsdata); | |
1da177e4 LT |
4792 | if (err < 0) |
4793 | goto fail; | |
afddba49 NP |
4794 | if (err < len-1) |
4795 | goto retry; | |
4796 | ||
1da177e4 LT |
4797 | mark_inode_dirty(inode); |
4798 | return 0; | |
1da177e4 LT |
4799 | fail: |
4800 | return err; | |
4801 | } | |
4d359507 | 4802 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 | 4803 | |
0adb25d2 KK |
4804 | int page_symlink(struct inode *inode, const char *symname, int len) |
4805 | { | |
4806 | return __page_symlink(inode, symname, len, | |
c62d2555 | 4807 | !mapping_gfp_constraint(inode->i_mapping, __GFP_FS)); |
0adb25d2 | 4808 | } |
4d359507 | 4809 | EXPORT_SYMBOL(page_symlink); |
0adb25d2 | 4810 | |
92e1d5be | 4811 | const struct inode_operations page_symlink_inode_operations = { |
6b255391 | 4812 | .get_link = page_get_link, |
1da177e4 | 4813 | }; |
1da177e4 | 4814 | EXPORT_SYMBOL(page_symlink_inode_operations); |