]>
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 | 224 | } else if (len <= PATH_MAX) { |
30ce4d19 | 225 | const size_t size = offsetof(struct filename, iname[1]); |
08518549 PM |
226 | struct filename *tmp; |
227 | ||
30ce4d19 | 228 | tmp = kmalloc(size, GFP_KERNEL); |
08518549 PM |
229 | if (unlikely(!tmp)) { |
230 | __putname(result); | |
231 | return ERR_PTR(-ENOMEM); | |
232 | } | |
233 | tmp->name = (char *)result; | |
08518549 PM |
234 | result = tmp; |
235 | } else { | |
236 | __putname(result); | |
237 | return ERR_PTR(-ENAMETOOLONG); | |
238 | } | |
239 | memcpy((char *)result->name, filename, len); | |
c4ad8f98 LT |
240 | result->uptr = NULL; |
241 | result->aname = NULL; | |
55422d0b | 242 | result->refcnt = 1; |
fd3522fd | 243 | audit_getname(result); |
c4ad8f98 | 244 | |
c4ad8f98 LT |
245 | return result; |
246 | } | |
247 | ||
91a27b2a | 248 | void putname(struct filename *name) |
1da177e4 | 249 | { |
55422d0b PM |
250 | BUG_ON(name->refcnt <= 0); |
251 | ||
252 | if (--name->refcnt > 0) | |
253 | return; | |
254 | ||
fd2f7cb5 | 255 | if (name->name != name->iname) { |
55422d0b PM |
256 | __putname(name->name); |
257 | kfree(name); | |
258 | } else | |
259 | __putname(name); | |
1da177e4 | 260 | } |
1da177e4 | 261 | |
e77819e5 LT |
262 | static int check_acl(struct inode *inode, int mask) |
263 | { | |
84635d68 | 264 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
265 | struct posix_acl *acl; |
266 | ||
e77819e5 | 267 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
268 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
269 | if (!acl) | |
e77819e5 | 270 | return -EAGAIN; |
3567866b | 271 | /* no ->get_acl() calls in RCU mode... */ |
b8a7a3a6 | 272 | if (is_uncached_acl(acl)) |
3567866b | 273 | return -ECHILD; |
206b1d09 | 274 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
275 | } |
276 | ||
2982baa2 CH |
277 | acl = get_acl(inode, ACL_TYPE_ACCESS); |
278 | if (IS_ERR(acl)) | |
279 | return PTR_ERR(acl); | |
e77819e5 LT |
280 | if (acl) { |
281 | int error = posix_acl_permission(inode, acl, mask); | |
282 | posix_acl_release(acl); | |
283 | return error; | |
284 | } | |
84635d68 | 285 | #endif |
e77819e5 LT |
286 | |
287 | return -EAGAIN; | |
288 | } | |
289 | ||
5909ccaa | 290 | /* |
948409c7 | 291 | * This does the basic permission checking |
1da177e4 | 292 | */ |
7e40145e | 293 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 294 | { |
26cf46be | 295 | unsigned int mode = inode->i_mode; |
1da177e4 | 296 | |
8e96e3b7 | 297 | if (likely(uid_eq(current_fsuid(), inode->i_uid))) |
1da177e4 LT |
298 | mode >>= 6; |
299 | else { | |
e77819e5 | 300 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 301 | int error = check_acl(inode, mask); |
b74c79e9 NP |
302 | if (error != -EAGAIN) |
303 | return error; | |
1da177e4 LT |
304 | } |
305 | ||
306 | if (in_group_p(inode->i_gid)) | |
307 | mode >>= 3; | |
308 | } | |
309 | ||
310 | /* | |
311 | * If the DACs are ok we don't need any capability check. | |
312 | */ | |
9c2c7039 | 313 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 314 | return 0; |
5909ccaa LT |
315 | return -EACCES; |
316 | } | |
317 | ||
318 | /** | |
b74c79e9 | 319 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 320 | * @inode: inode to check access rights for |
8fd90c8d | 321 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
322 | * |
323 | * Used to check for read/write/execute permissions on a file. | |
324 | * We use "fsuid" for this, letting us set arbitrary permissions | |
325 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
326 | * are used for other things. |
327 | * | |
328 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
329 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
330 | * It would then be called again in ref-walk mode. | |
5909ccaa | 331 | */ |
2830ba7f | 332 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
333 | { |
334 | int ret; | |
335 | ||
336 | /* | |
948409c7 | 337 | * Do the basic permission checks. |
5909ccaa | 338 | */ |
7e40145e | 339 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
340 | if (ret != -EACCES) |
341 | return ret; | |
1da177e4 | 342 | |
d594e7ec AV |
343 | if (S_ISDIR(inode->i_mode)) { |
344 | /* DACs are overridable for directories */ | |
d594e7ec | 345 | if (!(mask & MAY_WRITE)) |
23adbe12 AL |
346 | if (capable_wrt_inode_uidgid(inode, |
347 | CAP_DAC_READ_SEARCH)) | |
d594e7ec | 348 | return 0; |
23adbe12 | 349 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_OVERRIDE)) |
1da177e4 | 350 | return 0; |
2a4c2242 SS |
351 | return -EACCES; |
352 | } | |
1da177e4 LT |
353 | |
354 | /* | |
355 | * Searching includes executable on directories, else just read. | |
356 | */ | |
7ea66001 | 357 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 358 | if (mask == MAY_READ) |
23adbe12 | 359 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_READ_SEARCH)) |
1da177e4 | 360 | return 0; |
2a4c2242 SS |
361 | /* |
362 | * Read/write DACs are always overridable. | |
363 | * Executable DACs are overridable when there is | |
364 | * at least one exec bit set. | |
365 | */ | |
366 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) | |
367 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_OVERRIDE)) | |
368 | return 0; | |
1da177e4 LT |
369 | |
370 | return -EACCES; | |
371 | } | |
4d359507 | 372 | EXPORT_SYMBOL(generic_permission); |
1da177e4 | 373 | |
3ddcd056 LT |
374 | /* |
375 | * We _really_ want to just do "generic_permission()" without | |
376 | * even looking at the inode->i_op values. So we keep a cache | |
377 | * flag in inode->i_opflags, that says "this has not special | |
378 | * permission function, use the fast case". | |
379 | */ | |
380 | static inline int do_inode_permission(struct inode *inode, int mask) | |
381 | { | |
382 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
383 | if (likely(inode->i_op->permission)) | |
384 | return inode->i_op->permission(inode, mask); | |
385 | ||
386 | /* This gets set once for the inode lifetime */ | |
387 | spin_lock(&inode->i_lock); | |
388 | inode->i_opflags |= IOP_FASTPERM; | |
389 | spin_unlock(&inode->i_lock); | |
390 | } | |
391 | return generic_permission(inode, mask); | |
392 | } | |
393 | ||
0bdaea90 DH |
394 | /** |
395 | * sb_permission - Check superblock-level permissions | |
396 | * @sb: Superblock of inode to check permission on | |
55852635 | 397 | * @inode: Inode to check permission on |
0bdaea90 DH |
398 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) |
399 | * | |
400 | * Separate out file-system wide checks from inode-specific permission checks. | |
401 | */ | |
402 | static int sb_permission(struct super_block *sb, struct inode *inode, int mask) | |
403 | { | |
404 | if (unlikely(mask & MAY_WRITE)) { | |
405 | umode_t mode = inode->i_mode; | |
406 | ||
407 | /* Nobody gets write access to a read-only fs. */ | |
bc98a42c | 408 | if (sb_rdonly(sb) && (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) |
0bdaea90 DH |
409 | return -EROFS; |
410 | } | |
411 | return 0; | |
412 | } | |
413 | ||
414 | /** | |
415 | * inode_permission - Check for access rights to a given inode | |
416 | * @inode: Inode to check permission on | |
417 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
418 | * | |
419 | * Check for read/write/execute permissions on an inode. We use fs[ug]id for | |
420 | * this, letting us set arbitrary permissions for filesystem access without | |
421 | * changing the "normal" UIDs which are used for other things. | |
422 | * | |
423 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
424 | */ | |
425 | int inode_permission(struct inode *inode, int mask) | |
426 | { | |
427 | int retval; | |
428 | ||
429 | retval = sb_permission(inode->i_sb, inode, mask); | |
430 | if (retval) | |
431 | return retval; | |
4bfd054a EB |
432 | |
433 | if (unlikely(mask & MAY_WRITE)) { | |
434 | /* | |
435 | * Nobody gets write access to an immutable file. | |
436 | */ | |
437 | if (IS_IMMUTABLE(inode)) | |
438 | return -EPERM; | |
439 | ||
440 | /* | |
441 | * Updating mtime will likely cause i_uid and i_gid to be | |
442 | * written back improperly if their true value is unknown | |
443 | * to the vfs. | |
444 | */ | |
445 | if (HAS_UNMAPPED_ID(inode)) | |
446 | return -EACCES; | |
447 | } | |
448 | ||
449 | retval = do_inode_permission(inode, mask); | |
450 | if (retval) | |
451 | return retval; | |
452 | ||
453 | retval = devcgroup_inode_permission(inode, mask); | |
454 | if (retval) | |
455 | return retval; | |
456 | ||
457 | return security_inode_permission(inode, mask); | |
0bdaea90 | 458 | } |
4d359507 | 459 | EXPORT_SYMBOL(inode_permission); |
0bdaea90 | 460 | |
5dd784d0 JB |
461 | /** |
462 | * path_get - get a reference to a path | |
463 | * @path: path to get the reference to | |
464 | * | |
465 | * Given a path increment the reference count to the dentry and the vfsmount. | |
466 | */ | |
dcf787f3 | 467 | void path_get(const struct path *path) |
5dd784d0 JB |
468 | { |
469 | mntget(path->mnt); | |
470 | dget(path->dentry); | |
471 | } | |
472 | EXPORT_SYMBOL(path_get); | |
473 | ||
1d957f9b JB |
474 | /** |
475 | * path_put - put a reference to a path | |
476 | * @path: path to put the reference to | |
477 | * | |
478 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
479 | */ | |
dcf787f3 | 480 | void path_put(const struct path *path) |
1da177e4 | 481 | { |
1d957f9b JB |
482 | dput(path->dentry); |
483 | mntput(path->mnt); | |
1da177e4 | 484 | } |
1d957f9b | 485 | EXPORT_SYMBOL(path_put); |
1da177e4 | 486 | |
894bc8c4 | 487 | #define EMBEDDED_LEVELS 2 |
1f55a6ec AV |
488 | struct nameidata { |
489 | struct path path; | |
1cf2665b | 490 | struct qstr last; |
1f55a6ec AV |
491 | struct path root; |
492 | struct inode *inode; /* path.dentry.d_inode */ | |
493 | unsigned int flags; | |
9883d185 | 494 | unsigned seq, m_seq; |
1f55a6ec AV |
495 | int last_type; |
496 | unsigned depth; | |
756daf26 | 497 | int total_link_count; |
697fc6ca AV |
498 | struct saved { |
499 | struct path link; | |
fceef393 | 500 | struct delayed_call done; |
697fc6ca | 501 | const char *name; |
0450b2d1 | 502 | unsigned seq; |
894bc8c4 | 503 | } *stack, internal[EMBEDDED_LEVELS]; |
9883d185 AV |
504 | struct filename *name; |
505 | struct nameidata *saved; | |
fceef393 | 506 | struct inode *link_inode; |
9883d185 AV |
507 | unsigned root_seq; |
508 | int dfd; | |
3859a271 | 509 | } __randomize_layout; |
1f55a6ec | 510 | |
9883d185 | 511 | static void set_nameidata(struct nameidata *p, int dfd, struct filename *name) |
894bc8c4 | 512 | { |
756daf26 N |
513 | struct nameidata *old = current->nameidata; |
514 | p->stack = p->internal; | |
c8a53ee5 AV |
515 | p->dfd = dfd; |
516 | p->name = name; | |
756daf26 | 517 | p->total_link_count = old ? old->total_link_count : 0; |
9883d185 | 518 | p->saved = old; |
756daf26 | 519 | current->nameidata = p; |
894bc8c4 AV |
520 | } |
521 | ||
9883d185 | 522 | static void restore_nameidata(void) |
894bc8c4 | 523 | { |
9883d185 | 524 | struct nameidata *now = current->nameidata, *old = now->saved; |
756daf26 N |
525 | |
526 | current->nameidata = old; | |
527 | if (old) | |
528 | old->total_link_count = now->total_link_count; | |
e1a63bbc | 529 | if (now->stack != now->internal) |
756daf26 | 530 | kfree(now->stack); |
894bc8c4 AV |
531 | } |
532 | ||
533 | static int __nd_alloc_stack(struct nameidata *nd) | |
534 | { | |
bc40aee0 AV |
535 | struct saved *p; |
536 | ||
537 | if (nd->flags & LOOKUP_RCU) { | |
6da2ec56 | 538 | p= kmalloc_array(MAXSYMLINKS, sizeof(struct saved), |
bc40aee0 AV |
539 | GFP_ATOMIC); |
540 | if (unlikely(!p)) | |
541 | return -ECHILD; | |
542 | } else { | |
6da2ec56 | 543 | p= kmalloc_array(MAXSYMLINKS, sizeof(struct saved), |
894bc8c4 | 544 | GFP_KERNEL); |
bc40aee0 AV |
545 | if (unlikely(!p)) |
546 | return -ENOMEM; | |
547 | } | |
894bc8c4 AV |
548 | memcpy(p, nd->internal, sizeof(nd->internal)); |
549 | nd->stack = p; | |
550 | return 0; | |
551 | } | |
552 | ||
397d425d EB |
553 | /** |
554 | * path_connected - Verify that a path->dentry is below path->mnt.mnt_root | |
555 | * @path: nameidate to verify | |
556 | * | |
557 | * Rename can sometimes move a file or directory outside of a bind | |
558 | * mount, path_connected allows those cases to be detected. | |
559 | */ | |
560 | static bool path_connected(const struct path *path) | |
561 | { | |
562 | struct vfsmount *mnt = path->mnt; | |
95dd7758 | 563 | struct super_block *sb = mnt->mnt_sb; |
397d425d | 564 | |
95dd7758 EB |
565 | /* Bind mounts and multi-root filesystems can have disconnected paths */ |
566 | if (!(sb->s_iflags & SB_I_MULTIROOT) && (mnt->mnt_root == sb->s_root)) | |
397d425d EB |
567 | return true; |
568 | ||
569 | return is_subdir(path->dentry, mnt->mnt_root); | |
570 | } | |
571 | ||
894bc8c4 AV |
572 | static inline int nd_alloc_stack(struct nameidata *nd) |
573 | { | |
da4e0be0 | 574 | if (likely(nd->depth != EMBEDDED_LEVELS)) |
894bc8c4 AV |
575 | return 0; |
576 | if (likely(nd->stack != nd->internal)) | |
577 | return 0; | |
578 | return __nd_alloc_stack(nd); | |
579 | } | |
580 | ||
7973387a AV |
581 | static void drop_links(struct nameidata *nd) |
582 | { | |
583 | int i = nd->depth; | |
584 | while (i--) { | |
585 | struct saved *last = nd->stack + i; | |
fceef393 AV |
586 | do_delayed_call(&last->done); |
587 | clear_delayed_call(&last->done); | |
7973387a AV |
588 | } |
589 | } | |
590 | ||
591 | static void terminate_walk(struct nameidata *nd) | |
592 | { | |
593 | drop_links(nd); | |
594 | if (!(nd->flags & LOOKUP_RCU)) { | |
595 | int i; | |
596 | path_put(&nd->path); | |
597 | for (i = 0; i < nd->depth; i++) | |
598 | path_put(&nd->stack[i].link); | |
84a2bd39 | 599 | if (nd->flags & LOOKUP_ROOT_GRABBED) { |
102b8af2 | 600 | path_put(&nd->root); |
84a2bd39 | 601 | nd->flags &= ~LOOKUP_ROOT_GRABBED; |
102b8af2 | 602 | } |
7973387a AV |
603 | } else { |
604 | nd->flags &= ~LOOKUP_RCU; | |
7973387a AV |
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 | ||
ee594bff AV |
642 | static bool legitimize_root(struct nameidata *nd) |
643 | { | |
644 | if (!nd->root.mnt || (nd->flags & LOOKUP_ROOT)) | |
645 | return true; | |
84a2bd39 | 646 | nd->flags |= LOOKUP_ROOT_GRABBED; |
ee594bff AV |
647 | return legitimize_path(nd, &nd->root, nd->root_seq); |
648 | } | |
649 | ||
19660af7 | 650 | /* |
31e6b01f | 651 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
652 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
653 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
57e3715c | 654 | * normal reference counts on dentries and vfsmounts to transition to ref-walk |
19660af7 AV |
655 | * mode. Refcounts are grabbed at the last known good point before rcu-walk |
656 | * got stuck, so ref-walk may continue from there. If this is not successful | |
657 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
658 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 659 | */ |
31e6b01f NP |
660 | |
661 | /** | |
19660af7 AV |
662 | * unlazy_walk - try to switch to ref-walk mode. |
663 | * @nd: nameidata pathwalk data | |
39191628 | 664 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 665 | * |
4675ac39 AV |
666 | * unlazy_walk attempts to legitimize the current nd->path and nd->root |
667 | * for ref-walk mode. | |
668 | * Must be called from rcu-walk context. | |
7973387a AV |
669 | * Nothing should touch nameidata between unlazy_walk() failure and |
670 | * terminate_walk(). | |
31e6b01f | 671 | */ |
4675ac39 | 672 | static int unlazy_walk(struct nameidata *nd) |
31e6b01f | 673 | { |
31e6b01f NP |
674 | struct dentry *parent = nd->path.dentry; |
675 | ||
676 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
e5c832d5 | 677 | |
4675ac39 AV |
678 | nd->flags &= ~LOOKUP_RCU; |
679 | if (unlikely(!legitimize_links(nd))) | |
4675ac39 | 680 | goto out1; |
84a2bd39 AV |
681 | if (unlikely(!legitimize_path(nd, &nd->path, nd->seq))) |
682 | goto out; | |
ee594bff AV |
683 | if (unlikely(!legitimize_root(nd))) |
684 | goto out; | |
4675ac39 AV |
685 | rcu_read_unlock(); |
686 | BUG_ON(nd->inode != parent->d_inode); | |
687 | return 0; | |
688 | ||
84a2bd39 | 689 | out1: |
4675ac39 AV |
690 | nd->path.mnt = NULL; |
691 | nd->path.dentry = NULL; | |
4675ac39 AV |
692 | out: |
693 | rcu_read_unlock(); | |
694 | return -ECHILD; | |
695 | } | |
696 | ||
697 | /** | |
698 | * unlazy_child - try to switch to ref-walk mode. | |
699 | * @nd: nameidata pathwalk data | |
700 | * @dentry: child of nd->path.dentry | |
701 | * @seq: seq number to check dentry against | |
702 | * Returns: 0 on success, -ECHILD on failure | |
703 | * | |
704 | * unlazy_child attempts to legitimize the current nd->path, nd->root and dentry | |
705 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
706 | * @nd. Must be called from rcu-walk context. | |
707 | * Nothing should touch nameidata between unlazy_child() failure and | |
708 | * terminate_walk(). | |
709 | */ | |
710 | static int unlazy_child(struct nameidata *nd, struct dentry *dentry, unsigned seq) | |
711 | { | |
712 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
713 | ||
e5c832d5 | 714 | nd->flags &= ~LOOKUP_RCU; |
7973387a AV |
715 | if (unlikely(!legitimize_links(nd))) |
716 | goto out2; | |
717 | if (unlikely(!legitimize_mnt(nd->path.mnt, nd->m_seq))) | |
718 | goto out2; | |
4675ac39 | 719 | if (unlikely(!lockref_get_not_dead(&nd->path.dentry->d_lockref))) |
7973387a | 720 | goto out1; |
48a066e7 | 721 | |
15570086 | 722 | /* |
4675ac39 AV |
723 | * We need to move both the parent and the dentry from the RCU domain |
724 | * to be properly refcounted. And the sequence number in the dentry | |
725 | * validates *both* dentry counters, since we checked the sequence | |
726 | * number of the parent after we got the child sequence number. So we | |
727 | * know the parent must still be valid if the child sequence number is | |
15570086 | 728 | */ |
4675ac39 AV |
729 | if (unlikely(!lockref_get_not_dead(&dentry->d_lockref))) |
730 | goto out; | |
84a2bd39 AV |
731 | if (unlikely(read_seqcount_retry(&dentry->d_seq, seq))) |
732 | goto out_dput; | |
e5c832d5 LT |
733 | /* |
734 | * Sequence counts matched. Now make sure that the root is | |
735 | * still valid and get it if required. | |
736 | */ | |
84a2bd39 AV |
737 | if (unlikely(!legitimize_root(nd))) |
738 | goto out_dput; | |
8b61e74f | 739 | rcu_read_unlock(); |
31e6b01f | 740 | return 0; |
19660af7 | 741 | |
7973387a AV |
742 | out2: |
743 | nd->path.mnt = NULL; | |
744 | out1: | |
745 | nd->path.dentry = NULL; | |
e5c832d5 | 746 | out: |
8b61e74f | 747 | rcu_read_unlock(); |
84a2bd39 AV |
748 | return -ECHILD; |
749 | out_dput: | |
750 | rcu_read_unlock(); | |
751 | dput(dentry); | |
31e6b01f NP |
752 | return -ECHILD; |
753 | } | |
754 | ||
4ce16ef3 | 755 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 756 | { |
a89f8337 AV |
757 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) |
758 | return dentry->d_op->d_revalidate(dentry, flags); | |
759 | else | |
760 | return 1; | |
34286d66 NP |
761 | } |
762 | ||
9f1fafee AV |
763 | /** |
764 | * complete_walk - successful completion of path walk | |
765 | * @nd: pointer nameidata | |
39159de2 | 766 | * |
9f1fafee AV |
767 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
768 | * Revalidate the final result, unless we'd already done that during | |
769 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
770 | * success, -error on failure. In case of failure caller does not | |
771 | * need to drop nd->path. | |
39159de2 | 772 | */ |
9f1fafee | 773 | static int complete_walk(struct nameidata *nd) |
39159de2 | 774 | { |
16c2cd71 | 775 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 776 | int status; |
39159de2 | 777 | |
9f1fafee | 778 | if (nd->flags & LOOKUP_RCU) { |
9f1fafee AV |
779 | if (!(nd->flags & LOOKUP_ROOT)) |
780 | nd->root.mnt = NULL; | |
4675ac39 | 781 | if (unlikely(unlazy_walk(nd))) |
9f1fafee | 782 | return -ECHILD; |
9f1fafee AV |
783 | } |
784 | ||
16c2cd71 AV |
785 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
786 | return 0; | |
787 | ||
ecf3d1f1 | 788 | if (likely(!(dentry->d_flags & DCACHE_OP_WEAK_REVALIDATE))) |
39159de2 JL |
789 | return 0; |
790 | ||
ecf3d1f1 | 791 | status = dentry->d_op->d_weak_revalidate(dentry, nd->flags); |
39159de2 JL |
792 | if (status > 0) |
793 | return 0; | |
794 | ||
16c2cd71 | 795 | if (!status) |
39159de2 | 796 | status = -ESTALE; |
16c2cd71 | 797 | |
39159de2 JL |
798 | return status; |
799 | } | |
800 | ||
18d8c860 | 801 | static void set_root(struct nameidata *nd) |
31e6b01f | 802 | { |
7bd88377 | 803 | struct fs_struct *fs = current->fs; |
c28cc364 | 804 | |
9e6697e2 AV |
805 | if (nd->flags & LOOKUP_RCU) { |
806 | unsigned seq; | |
807 | ||
808 | do { | |
809 | seq = read_seqcount_begin(&fs->seq); | |
810 | nd->root = fs->root; | |
811 | nd->root_seq = __read_seqcount_begin(&nd->root.dentry->d_seq); | |
812 | } while (read_seqcount_retry(&fs->seq, seq)); | |
813 | } else { | |
814 | get_fs_root(fs, &nd->root); | |
84a2bd39 | 815 | nd->flags |= LOOKUP_ROOT_GRABBED; |
9e6697e2 | 816 | } |
31e6b01f NP |
817 | } |
818 | ||
1d957f9b | 819 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
820 | { |
821 | dput(path->dentry); | |
4ac91378 | 822 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
823 | mntput(path->mnt); |
824 | } | |
825 | ||
7b9337aa NP |
826 | static inline void path_to_nameidata(const struct path *path, |
827 | struct nameidata *nd) | |
051d3812 | 828 | { |
31e6b01f NP |
829 | if (!(nd->flags & LOOKUP_RCU)) { |
830 | dput(nd->path.dentry); | |
831 | if (nd->path.mnt != path->mnt) | |
832 | mntput(nd->path.mnt); | |
9a229683 | 833 | } |
31e6b01f | 834 | nd->path.mnt = path->mnt; |
4ac91378 | 835 | nd->path.dentry = path->dentry; |
051d3812 IK |
836 | } |
837 | ||
248fb5b9 AV |
838 | static int nd_jump_root(struct nameidata *nd) |
839 | { | |
840 | if (nd->flags & LOOKUP_RCU) { | |
841 | struct dentry *d; | |
842 | nd->path = nd->root; | |
843 | d = nd->path.dentry; | |
844 | nd->inode = d->d_inode; | |
845 | nd->seq = nd->root_seq; | |
846 | if (unlikely(read_seqcount_retry(&d->d_seq, nd->seq))) | |
847 | return -ECHILD; | |
848 | } else { | |
849 | path_put(&nd->path); | |
850 | nd->path = nd->root; | |
851 | path_get(&nd->path); | |
852 | nd->inode = nd->path.dentry->d_inode; | |
853 | } | |
854 | nd->flags |= LOOKUP_JUMPED; | |
855 | return 0; | |
856 | } | |
857 | ||
b5fb63c1 | 858 | /* |
6b255391 | 859 | * Helper to directly jump to a known parsed path from ->get_link, |
b5fb63c1 CH |
860 | * caller must have taken a reference to path beforehand. |
861 | */ | |
6e77137b | 862 | void nd_jump_link(struct path *path) |
b5fb63c1 | 863 | { |
6e77137b | 864 | struct nameidata *nd = current->nameidata; |
b5fb63c1 CH |
865 | path_put(&nd->path); |
866 | ||
867 | nd->path = *path; | |
868 | nd->inode = nd->path.dentry->d_inode; | |
869 | nd->flags |= LOOKUP_JUMPED; | |
b5fb63c1 CH |
870 | } |
871 | ||
b9ff4429 | 872 | static inline void put_link(struct nameidata *nd) |
574197e0 | 873 | { |
21c3003d | 874 | struct saved *last = nd->stack + --nd->depth; |
fceef393 | 875 | do_delayed_call(&last->done); |
6548fae2 AV |
876 | if (!(nd->flags & LOOKUP_RCU)) |
877 | path_put(&last->link); | |
574197e0 AV |
878 | } |
879 | ||
561ec64a LT |
880 | int sysctl_protected_symlinks __read_mostly = 0; |
881 | int sysctl_protected_hardlinks __read_mostly = 0; | |
30aba665 SM |
882 | int sysctl_protected_fifos __read_mostly; |
883 | int sysctl_protected_regular __read_mostly; | |
800179c9 KC |
884 | |
885 | /** | |
886 | * may_follow_link - Check symlink following for unsafe situations | |
55852635 | 887 | * @nd: nameidata pathwalk data |
800179c9 KC |
888 | * |
889 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
890 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
891 | * in a sticky world-writable directory. This is to protect privileged | |
892 | * processes from failing races against path names that may change out | |
893 | * from under them by way of other users creating malicious symlinks. | |
894 | * It will permit symlinks to be followed only when outside a sticky | |
895 | * world-writable directory, or when the uid of the symlink and follower | |
896 | * match, or when the directory owner matches the symlink's owner. | |
897 | * | |
898 | * Returns 0 if following the symlink is allowed, -ve on error. | |
899 | */ | |
fec2fa24 | 900 | static inline int may_follow_link(struct nameidata *nd) |
800179c9 KC |
901 | { |
902 | const struct inode *inode; | |
903 | const struct inode *parent; | |
2d7f9e2a | 904 | kuid_t puid; |
800179c9 KC |
905 | |
906 | if (!sysctl_protected_symlinks) | |
907 | return 0; | |
908 | ||
909 | /* Allowed if owner and follower match. */ | |
fceef393 | 910 | inode = nd->link_inode; |
81abe27b | 911 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
912 | return 0; |
913 | ||
914 | /* Allowed if parent directory not sticky and world-writable. */ | |
aa65fa35 | 915 | parent = nd->inode; |
800179c9 KC |
916 | if ((parent->i_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) |
917 | return 0; | |
918 | ||
919 | /* Allowed if parent directory and link owner match. */ | |
2d7f9e2a SF |
920 | puid = parent->i_uid; |
921 | if (uid_valid(puid) && uid_eq(puid, inode->i_uid)) | |
800179c9 KC |
922 | return 0; |
923 | ||
31956502 AV |
924 | if (nd->flags & LOOKUP_RCU) |
925 | return -ECHILD; | |
926 | ||
ea841baf | 927 | audit_inode(nd->name, nd->stack[0].link.dentry, 0); |
94b9d9b7 | 928 | audit_log_link_denied("follow_link"); |
800179c9 KC |
929 | return -EACCES; |
930 | } | |
931 | ||
932 | /** | |
933 | * safe_hardlink_source - Check for safe hardlink conditions | |
934 | * @inode: the source inode to hardlink from | |
935 | * | |
936 | * Return false if at least one of the following conditions: | |
937 | * - inode is not a regular file | |
938 | * - inode is setuid | |
939 | * - inode is setgid and group-exec | |
940 | * - access failure for read and write | |
941 | * | |
942 | * Otherwise returns true. | |
943 | */ | |
944 | static bool safe_hardlink_source(struct inode *inode) | |
945 | { | |
946 | umode_t mode = inode->i_mode; | |
947 | ||
948 | /* Special files should not get pinned to the filesystem. */ | |
949 | if (!S_ISREG(mode)) | |
950 | return false; | |
951 | ||
952 | /* Setuid files should not get pinned to the filesystem. */ | |
953 | if (mode & S_ISUID) | |
954 | return false; | |
955 | ||
956 | /* Executable setgid files should not get pinned to the filesystem. */ | |
957 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
958 | return false; | |
959 | ||
960 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
961 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
962 | return false; | |
963 | ||
964 | return true; | |
965 | } | |
966 | ||
967 | /** | |
968 | * may_linkat - Check permissions for creating a hardlink | |
969 | * @link: the source to hardlink from | |
970 | * | |
971 | * Block hardlink when all of: | |
972 | * - sysctl_protected_hardlinks enabled | |
973 | * - fsuid does not match inode | |
974 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
f2ca3796 | 975 | * - not CAP_FOWNER in a namespace with the inode owner uid mapped |
800179c9 KC |
976 | * |
977 | * Returns 0 if successful, -ve on error. | |
978 | */ | |
979 | static int may_linkat(struct path *link) | |
980 | { | |
593d1ce8 EB |
981 | struct inode *inode = link->dentry->d_inode; |
982 | ||
983 | /* Inode writeback is not safe when the uid or gid are invalid. */ | |
984 | if (!uid_valid(inode->i_uid) || !gid_valid(inode->i_gid)) | |
985 | return -EOVERFLOW; | |
800179c9 KC |
986 | |
987 | if (!sysctl_protected_hardlinks) | |
988 | return 0; | |
989 | ||
800179c9 KC |
990 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, |
991 | * otherwise, it must be a safe source. | |
992 | */ | |
cc658db4 | 993 | if (safe_hardlink_source(inode) || inode_owner_or_capable(inode)) |
800179c9 KC |
994 | return 0; |
995 | ||
94b9d9b7 | 996 | audit_log_link_denied("linkat"); |
800179c9 KC |
997 | return -EPERM; |
998 | } | |
999 | ||
30aba665 SM |
1000 | /** |
1001 | * may_create_in_sticky - Check whether an O_CREAT open in a sticky directory | |
1002 | * should be allowed, or not, on files that already | |
1003 | * exist. | |
1004 | * @dir: the sticky parent directory | |
1005 | * @inode: the inode of the file to open | |
1006 | * | |
1007 | * Block an O_CREAT open of a FIFO (or a regular file) when: | |
1008 | * - sysctl_protected_fifos (or sysctl_protected_regular) is enabled | |
1009 | * - the file already exists | |
1010 | * - we are in a sticky directory | |
1011 | * - we don't own the file | |
1012 | * - the owner of the directory doesn't own the file | |
1013 | * - the directory is world writable | |
1014 | * If the sysctl_protected_fifos (or sysctl_protected_regular) is set to 2 | |
1015 | * the directory doesn't have to be world writable: being group writable will | |
1016 | * be enough. | |
1017 | * | |
1018 | * Returns 0 if the open is allowed, -ve on error. | |
1019 | */ | |
1020 | static int may_create_in_sticky(struct dentry * const dir, | |
1021 | struct inode * const inode) | |
1022 | { | |
1023 | if ((!sysctl_protected_fifos && S_ISFIFO(inode->i_mode)) || | |
1024 | (!sysctl_protected_regular && S_ISREG(inode->i_mode)) || | |
1025 | likely(!(dir->d_inode->i_mode & S_ISVTX)) || | |
1026 | uid_eq(inode->i_uid, dir->d_inode->i_uid) || | |
1027 | uid_eq(current_fsuid(), inode->i_uid)) | |
1028 | return 0; | |
1029 | ||
1030 | if (likely(dir->d_inode->i_mode & 0002) || | |
1031 | (dir->d_inode->i_mode & 0020 && | |
1032 | ((sysctl_protected_fifos >= 2 && S_ISFIFO(inode->i_mode)) || | |
1033 | (sysctl_protected_regular >= 2 && S_ISREG(inode->i_mode))))) { | |
1034 | return -EACCES; | |
1035 | } | |
1036 | return 0; | |
1037 | } | |
1038 | ||
3b2e7f75 AV |
1039 | static __always_inline |
1040 | const char *get_link(struct nameidata *nd) | |
1da177e4 | 1041 | { |
ab104923 | 1042 | struct saved *last = nd->stack + nd->depth - 1; |
1cf2665b | 1043 | struct dentry *dentry = last->link.dentry; |
fceef393 | 1044 | struct inode *inode = nd->link_inode; |
6d7b5aae | 1045 | int error; |
0a959df5 | 1046 | const char *res; |
1da177e4 | 1047 | |
8fa9dd24 N |
1048 | if (!(nd->flags & LOOKUP_RCU)) { |
1049 | touch_atime(&last->link); | |
1050 | cond_resched(); | |
c6718543 | 1051 | } else if (atime_needs_update(&last->link, inode)) { |
4675ac39 | 1052 | if (unlikely(unlazy_walk(nd))) |
bc40aee0 | 1053 | return ERR_PTR(-ECHILD); |
8fa9dd24 | 1054 | touch_atime(&last->link); |
bc40aee0 | 1055 | } |
574197e0 | 1056 | |
bda0be7a N |
1057 | error = security_inode_follow_link(dentry, inode, |
1058 | nd->flags & LOOKUP_RCU); | |
1059 | if (unlikely(error)) | |
6920a440 | 1060 | return ERR_PTR(error); |
36f3b4f6 | 1061 | |
86acdca1 | 1062 | nd->last_type = LAST_BIND; |
4c4f7c19 | 1063 | res = READ_ONCE(inode->i_link); |
d4dee48b | 1064 | if (!res) { |
fceef393 AV |
1065 | const char * (*get)(struct dentry *, struct inode *, |
1066 | struct delayed_call *); | |
1067 | get = inode->i_op->get_link; | |
8c1b4566 | 1068 | if (nd->flags & LOOKUP_RCU) { |
fceef393 | 1069 | res = get(NULL, inode, &last->done); |
6b255391 | 1070 | if (res == ERR_PTR(-ECHILD)) { |
4675ac39 | 1071 | if (unlikely(unlazy_walk(nd))) |
6b255391 | 1072 | return ERR_PTR(-ECHILD); |
fceef393 | 1073 | res = get(dentry, inode, &last->done); |
6b255391 AV |
1074 | } |
1075 | } else { | |
fceef393 | 1076 | res = get(dentry, inode, &last->done); |
8c1b4566 | 1077 | } |
fceef393 | 1078 | if (IS_ERR_OR_NULL(res)) |
fab51e8a | 1079 | return res; |
fab51e8a AV |
1080 | } |
1081 | if (*res == '/') { | |
9e6697e2 AV |
1082 | if (!nd->root.mnt) |
1083 | set_root(nd); | |
248fb5b9 AV |
1084 | if (unlikely(nd_jump_root(nd))) |
1085 | return ERR_PTR(-ECHILD); | |
fab51e8a AV |
1086 | while (unlikely(*++res == '/')) |
1087 | ; | |
1da177e4 | 1088 | } |
fab51e8a AV |
1089 | if (!*res) |
1090 | res = NULL; | |
0a959df5 AV |
1091 | return res; |
1092 | } | |
6d7b5aae | 1093 | |
f015f126 DH |
1094 | /* |
1095 | * follow_up - Find the mountpoint of path's vfsmount | |
1096 | * | |
1097 | * Given a path, find the mountpoint of its source file system. | |
1098 | * Replace @path with the path of the mountpoint in the parent mount. | |
1099 | * Up is towards /. | |
1100 | * | |
1101 | * Return 1 if we went up a level and 0 if we were already at the | |
1102 | * root. | |
1103 | */ | |
bab77ebf | 1104 | int follow_up(struct path *path) |
1da177e4 | 1105 | { |
0714a533 AV |
1106 | struct mount *mnt = real_mount(path->mnt); |
1107 | struct mount *parent; | |
1da177e4 | 1108 | struct dentry *mountpoint; |
99b7db7b | 1109 | |
48a066e7 | 1110 | read_seqlock_excl(&mount_lock); |
0714a533 | 1111 | parent = mnt->mnt_parent; |
3c0a6163 | 1112 | if (parent == mnt) { |
48a066e7 | 1113 | read_sequnlock_excl(&mount_lock); |
1da177e4 LT |
1114 | return 0; |
1115 | } | |
0714a533 | 1116 | mntget(&parent->mnt); |
a73324da | 1117 | mountpoint = dget(mnt->mnt_mountpoint); |
48a066e7 | 1118 | read_sequnlock_excl(&mount_lock); |
bab77ebf AV |
1119 | dput(path->dentry); |
1120 | path->dentry = mountpoint; | |
1121 | mntput(path->mnt); | |
0714a533 | 1122 | path->mnt = &parent->mnt; |
1da177e4 LT |
1123 | return 1; |
1124 | } | |
4d359507 | 1125 | EXPORT_SYMBOL(follow_up); |
1da177e4 | 1126 | |
b5c84bf6 | 1127 | /* |
9875cf80 DH |
1128 | * Perform an automount |
1129 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
1130 | * were called with. | |
1da177e4 | 1131 | */ |
756daf26 | 1132 | static int follow_automount(struct path *path, struct nameidata *nd, |
9875cf80 | 1133 | bool *need_mntput) |
31e6b01f | 1134 | { |
9875cf80 | 1135 | struct vfsmount *mnt; |
ea5b778a | 1136 | int err; |
9875cf80 DH |
1137 | |
1138 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
1139 | return -EREMOTE; | |
1140 | ||
0ec26fd0 MS |
1141 | /* We don't want to mount if someone's just doing a stat - |
1142 | * unless they're stat'ing a directory and appended a '/' to | |
1143 | * the name. | |
1144 | * | |
1145 | * We do, however, want to mount if someone wants to open or | |
1146 | * create a file of any type under the mountpoint, wants to | |
1147 | * traverse through the mountpoint or wants to open the | |
1148 | * mounted directory. Also, autofs may mark negative dentries | |
1149 | * as being automount points. These will need the attentions | |
1150 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 1151 | */ |
756daf26 | 1152 | if (!(nd->flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
5d38f049 IK |
1153 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
1154 | path->dentry->d_inode) | |
1155 | return -EISDIR; | |
0ec26fd0 | 1156 | |
756daf26 N |
1157 | nd->total_link_count++; |
1158 | if (nd->total_link_count >= 40) | |
9875cf80 DH |
1159 | return -ELOOP; |
1160 | ||
1161 | mnt = path->dentry->d_op->d_automount(path); | |
1162 | if (IS_ERR(mnt)) { | |
1163 | /* | |
1164 | * The filesystem is allowed to return -EISDIR here to indicate | |
1165 | * it doesn't want to automount. For instance, autofs would do | |
1166 | * this so that its userspace daemon can mount on this dentry. | |
1167 | * | |
1168 | * However, we can only permit this if it's a terminal point in | |
1169 | * the path being looked up; if it wasn't then the remainder of | |
1170 | * the path is inaccessible and we should say so. | |
1171 | */ | |
756daf26 | 1172 | if (PTR_ERR(mnt) == -EISDIR && (nd->flags & LOOKUP_PARENT)) |
9875cf80 DH |
1173 | return -EREMOTE; |
1174 | return PTR_ERR(mnt); | |
31e6b01f | 1175 | } |
ea5b778a | 1176 | |
9875cf80 DH |
1177 | if (!mnt) /* mount collision */ |
1178 | return 0; | |
31e6b01f | 1179 | |
8aef1884 AV |
1180 | if (!*need_mntput) { |
1181 | /* lock_mount() may release path->mnt on error */ | |
1182 | mntget(path->mnt); | |
1183 | *need_mntput = true; | |
1184 | } | |
19a167af | 1185 | err = finish_automount(mnt, path); |
9875cf80 | 1186 | |
ea5b778a DH |
1187 | switch (err) { |
1188 | case -EBUSY: | |
1189 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 1190 | return 0; |
ea5b778a | 1191 | case 0: |
8aef1884 | 1192 | path_put(path); |
ea5b778a DH |
1193 | path->mnt = mnt; |
1194 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 1195 | return 0; |
19a167af AV |
1196 | default: |
1197 | return err; | |
ea5b778a | 1198 | } |
19a167af | 1199 | |
463ffb2e AV |
1200 | } |
1201 | ||
9875cf80 DH |
1202 | /* |
1203 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 1204 | * - Flagged for transit management (autofs) |
9875cf80 DH |
1205 | * - Flagged as mountpoint |
1206 | * - Flagged as automount point | |
1207 | * | |
1208 | * This may only be called in refwalk mode. | |
1209 | * | |
1210 | * Serialization is taken care of in namespace.c | |
1211 | */ | |
756daf26 | 1212 | static int follow_managed(struct path *path, struct nameidata *nd) |
1da177e4 | 1213 | { |
8aef1884 | 1214 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
1215 | unsigned managed; |
1216 | bool need_mntput = false; | |
8aef1884 | 1217 | int ret = 0; |
9875cf80 DH |
1218 | |
1219 | /* Given that we're not holding a lock here, we retain the value in a | |
1220 | * local variable for each dentry as we look at it so that we don't see | |
1221 | * the components of that value change under us */ | |
6aa7de05 | 1222 | while (managed = READ_ONCE(path->dentry->d_flags), |
9875cf80 DH |
1223 | managed &= DCACHE_MANAGED_DENTRY, |
1224 | unlikely(managed != 0)) { | |
cc53ce53 DH |
1225 | /* Allow the filesystem to manage the transit without i_mutex |
1226 | * being held. */ | |
1227 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1228 | BUG_ON(!path->dentry->d_op); | |
1229 | BUG_ON(!path->dentry->d_op->d_manage); | |
fb5f51c7 | 1230 | ret = path->dentry->d_op->d_manage(path, false); |
cc53ce53 | 1231 | if (ret < 0) |
8aef1884 | 1232 | break; |
cc53ce53 DH |
1233 | } |
1234 | ||
9875cf80 DH |
1235 | /* Transit to a mounted filesystem. */ |
1236 | if (managed & DCACHE_MOUNTED) { | |
1237 | struct vfsmount *mounted = lookup_mnt(path); | |
1238 | if (mounted) { | |
1239 | dput(path->dentry); | |
1240 | if (need_mntput) | |
1241 | mntput(path->mnt); | |
1242 | path->mnt = mounted; | |
1243 | path->dentry = dget(mounted->mnt_root); | |
1244 | need_mntput = true; | |
1245 | continue; | |
1246 | } | |
1247 | ||
1248 | /* Something is mounted on this dentry in another | |
1249 | * namespace and/or whatever was mounted there in this | |
48a066e7 AV |
1250 | * namespace got unmounted before lookup_mnt() could |
1251 | * get it */ | |
9875cf80 DH |
1252 | } |
1253 | ||
1254 | /* Handle an automount point */ | |
1255 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
756daf26 | 1256 | ret = follow_automount(path, nd, &need_mntput); |
9875cf80 | 1257 | if (ret < 0) |
8aef1884 | 1258 | break; |
9875cf80 DH |
1259 | continue; |
1260 | } | |
1261 | ||
1262 | /* We didn't change the current path point */ | |
1263 | break; | |
1da177e4 | 1264 | } |
8aef1884 AV |
1265 | |
1266 | if (need_mntput && path->mnt == mnt) | |
1267 | mntput(path->mnt); | |
e9742b53 AV |
1268 | if (ret == -EISDIR || !ret) |
1269 | ret = 1; | |
8402752e AV |
1270 | if (need_mntput) |
1271 | nd->flags |= LOOKUP_JUMPED; | |
1272 | if (unlikely(ret < 0)) | |
1273 | path_put_conditional(path, nd); | |
1274 | return ret; | |
1da177e4 LT |
1275 | } |
1276 | ||
cc53ce53 | 1277 | int follow_down_one(struct path *path) |
1da177e4 LT |
1278 | { |
1279 | struct vfsmount *mounted; | |
1280 | ||
1c755af4 | 1281 | mounted = lookup_mnt(path); |
1da177e4 | 1282 | if (mounted) { |
9393bd07 AV |
1283 | dput(path->dentry); |
1284 | mntput(path->mnt); | |
1285 | path->mnt = mounted; | |
1286 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1287 | return 1; |
1288 | } | |
1289 | return 0; | |
1290 | } | |
4d359507 | 1291 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 | 1292 | |
fb5f51c7 | 1293 | static inline int managed_dentry_rcu(const struct path *path) |
62a7375e | 1294 | { |
fb5f51c7 IK |
1295 | return (path->dentry->d_flags & DCACHE_MANAGE_TRANSIT) ? |
1296 | path->dentry->d_op->d_manage(path, true) : 0; | |
62a7375e IK |
1297 | } |
1298 | ||
9875cf80 | 1299 | /* |
287548e4 AV |
1300 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1301 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1302 | */ |
1303 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
254cf582 | 1304 | struct inode **inode, unsigned *seqp) |
9875cf80 | 1305 | { |
62a7375e | 1306 | for (;;) { |
c7105365 | 1307 | struct mount *mounted; |
62a7375e IK |
1308 | /* |
1309 | * Don't forget we might have a non-mountpoint managed dentry | |
1310 | * that wants to block transit. | |
1311 | */ | |
fb5f51c7 | 1312 | switch (managed_dentry_rcu(path)) { |
b8faf035 N |
1313 | case -ECHILD: |
1314 | default: | |
ab90911f | 1315 | return false; |
b8faf035 N |
1316 | case -EISDIR: |
1317 | return true; | |
1318 | case 0: | |
1319 | break; | |
1320 | } | |
62a7375e IK |
1321 | |
1322 | if (!d_mountpoint(path->dentry)) | |
b8faf035 | 1323 | return !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
62a7375e | 1324 | |
474279dc | 1325 | mounted = __lookup_mnt(path->mnt, path->dentry); |
9875cf80 DH |
1326 | if (!mounted) |
1327 | break; | |
c7105365 AV |
1328 | path->mnt = &mounted->mnt; |
1329 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1330 | nd->flags |= LOOKUP_JUMPED; |
254cf582 | 1331 | *seqp = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1332 | /* |
1333 | * Update the inode too. We don't need to re-check the | |
1334 | * dentry sequence number here after this d_inode read, | |
1335 | * because a mount-point is always pinned. | |
1336 | */ | |
1337 | *inode = path->dentry->d_inode; | |
9875cf80 | 1338 | } |
f5be3e29 | 1339 | return !read_seqretry(&mount_lock, nd->m_seq) && |
b8faf035 | 1340 | !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
287548e4 AV |
1341 | } |
1342 | ||
31e6b01f NP |
1343 | static int follow_dotdot_rcu(struct nameidata *nd) |
1344 | { | |
4023bfc9 | 1345 | struct inode *inode = nd->inode; |
31e6b01f | 1346 | |
9875cf80 | 1347 | while (1) { |
aed434ad | 1348 | if (path_equal(&nd->path, &nd->root)) |
31e6b01f | 1349 | break; |
31e6b01f NP |
1350 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
1351 | struct dentry *old = nd->path.dentry; | |
1352 | struct dentry *parent = old->d_parent; | |
1353 | unsigned seq; | |
1354 | ||
4023bfc9 | 1355 | inode = parent->d_inode; |
31e6b01f | 1356 | seq = read_seqcount_begin(&parent->d_seq); |
aed434ad AV |
1357 | if (unlikely(read_seqcount_retry(&old->d_seq, nd->seq))) |
1358 | return -ECHILD; | |
31e6b01f NP |
1359 | nd->path.dentry = parent; |
1360 | nd->seq = seq; | |
397d425d EB |
1361 | if (unlikely(!path_connected(&nd->path))) |
1362 | return -ENOENT; | |
31e6b01f | 1363 | break; |
aed434ad AV |
1364 | } else { |
1365 | struct mount *mnt = real_mount(nd->path.mnt); | |
1366 | struct mount *mparent = mnt->mnt_parent; | |
1367 | struct dentry *mountpoint = mnt->mnt_mountpoint; | |
1368 | struct inode *inode2 = mountpoint->d_inode; | |
1369 | unsigned seq = read_seqcount_begin(&mountpoint->d_seq); | |
1370 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) | |
1371 | return -ECHILD; | |
1372 | if (&mparent->mnt == nd->path.mnt) | |
1373 | break; | |
1374 | /* we know that mountpoint was pinned */ | |
1375 | nd->path.dentry = mountpoint; | |
1376 | nd->path.mnt = &mparent->mnt; | |
1377 | inode = inode2; | |
1378 | nd->seq = seq; | |
31e6b01f | 1379 | } |
31e6b01f | 1380 | } |
aed434ad | 1381 | while (unlikely(d_mountpoint(nd->path.dentry))) { |
b37199e6 AV |
1382 | struct mount *mounted; |
1383 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry); | |
aed434ad AV |
1384 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) |
1385 | return -ECHILD; | |
b37199e6 AV |
1386 | if (!mounted) |
1387 | break; | |
1388 | nd->path.mnt = &mounted->mnt; | |
1389 | nd->path.dentry = mounted->mnt.mnt_root; | |
4023bfc9 | 1390 | inode = nd->path.dentry->d_inode; |
b37199e6 | 1391 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
b37199e6 | 1392 | } |
4023bfc9 | 1393 | nd->inode = inode; |
31e6b01f NP |
1394 | return 0; |
1395 | } | |
1396 | ||
cc53ce53 DH |
1397 | /* |
1398 | * Follow down to the covering mount currently visible to userspace. At each | |
1399 | * point, the filesystem owning that dentry may be queried as to whether the | |
1400 | * caller is permitted to proceed or not. | |
cc53ce53 | 1401 | */ |
7cc90cc3 | 1402 | int follow_down(struct path *path) |
cc53ce53 DH |
1403 | { |
1404 | unsigned managed; | |
1405 | int ret; | |
1406 | ||
6aa7de05 | 1407 | while (managed = READ_ONCE(path->dentry->d_flags), |
cc53ce53 DH |
1408 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { |
1409 | /* Allow the filesystem to manage the transit without i_mutex | |
1410 | * being held. | |
1411 | * | |
1412 | * We indicate to the filesystem if someone is trying to mount | |
1413 | * something here. This gives autofs the chance to deny anyone | |
1414 | * other than its daemon the right to mount on its | |
1415 | * superstructure. | |
1416 | * | |
1417 | * The filesystem may sleep at this point. | |
1418 | */ | |
1419 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1420 | BUG_ON(!path->dentry->d_op); | |
1421 | BUG_ON(!path->dentry->d_op->d_manage); | |
fb5f51c7 | 1422 | ret = path->dentry->d_op->d_manage(path, false); |
cc53ce53 DH |
1423 | if (ret < 0) |
1424 | return ret == -EISDIR ? 0 : ret; | |
1425 | } | |
1426 | ||
1427 | /* Transit to a mounted filesystem. */ | |
1428 | if (managed & DCACHE_MOUNTED) { | |
1429 | struct vfsmount *mounted = lookup_mnt(path); | |
1430 | if (!mounted) | |
1431 | break; | |
1432 | dput(path->dentry); | |
1433 | mntput(path->mnt); | |
1434 | path->mnt = mounted; | |
1435 | path->dentry = dget(mounted->mnt_root); | |
1436 | continue; | |
1437 | } | |
1438 | ||
1439 | /* Don't handle automount points here */ | |
1440 | break; | |
1441 | } | |
1442 | return 0; | |
1443 | } | |
4d359507 | 1444 | EXPORT_SYMBOL(follow_down); |
cc53ce53 | 1445 | |
9875cf80 DH |
1446 | /* |
1447 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1448 | */ | |
1449 | static void follow_mount(struct path *path) | |
1450 | { | |
1451 | while (d_mountpoint(path->dentry)) { | |
1452 | struct vfsmount *mounted = lookup_mnt(path); | |
1453 | if (!mounted) | |
1454 | break; | |
1455 | dput(path->dentry); | |
1456 | mntput(path->mnt); | |
1457 | path->mnt = mounted; | |
1458 | path->dentry = dget(mounted->mnt_root); | |
1459 | } | |
1460 | } | |
1461 | ||
eedf265a EB |
1462 | static int path_parent_directory(struct path *path) |
1463 | { | |
1464 | struct dentry *old = path->dentry; | |
1465 | /* rare case of legitimate dget_parent()... */ | |
1466 | path->dentry = dget_parent(path->dentry); | |
1467 | dput(old); | |
1468 | if (unlikely(!path_connected(path))) | |
1469 | return -ENOENT; | |
1470 | return 0; | |
1471 | } | |
1472 | ||
397d425d | 1473 | static int follow_dotdot(struct nameidata *nd) |
1da177e4 LT |
1474 | { |
1475 | while(1) { | |
030c7e0b | 1476 | if (path_equal(&nd->path, &nd->root)) |
1da177e4 | 1477 | break; |
4ac91378 | 1478 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
eedf265a EB |
1479 | int ret = path_parent_directory(&nd->path); |
1480 | if (ret) | |
1481 | return ret; | |
1da177e4 LT |
1482 | break; |
1483 | } | |
3088dd70 | 1484 | if (!follow_up(&nd->path)) |
1da177e4 | 1485 | break; |
1da177e4 | 1486 | } |
79ed0226 | 1487 | follow_mount(&nd->path); |
31e6b01f | 1488 | nd->inode = nd->path.dentry->d_inode; |
397d425d | 1489 | return 0; |
1da177e4 LT |
1490 | } |
1491 | ||
baa03890 | 1492 | /* |
f4fdace9 OD |
1493 | * This looks up the name in dcache and possibly revalidates the found dentry. |
1494 | * NULL is returned if the dentry does not exist in the cache. | |
baa03890 | 1495 | */ |
e3c13928 AV |
1496 | static struct dentry *lookup_dcache(const struct qstr *name, |
1497 | struct dentry *dir, | |
6c51e513 | 1498 | unsigned int flags) |
baa03890 | 1499 | { |
a89f8337 | 1500 | struct dentry *dentry = d_lookup(dir, name); |
bad61189 | 1501 | if (dentry) { |
a89f8337 AV |
1502 | int error = d_revalidate(dentry, flags); |
1503 | if (unlikely(error <= 0)) { | |
1504 | if (!error) | |
1505 | d_invalidate(dentry); | |
1506 | dput(dentry); | |
1507 | return ERR_PTR(error); | |
bad61189 MS |
1508 | } |
1509 | } | |
baa03890 NP |
1510 | return dentry; |
1511 | } | |
1512 | ||
44396f4b | 1513 | /* |
a03ece5f AV |
1514 | * Parent directory has inode locked exclusive. This is one |
1515 | * and only case when ->lookup() gets called on non in-lookup | |
1516 | * dentries - as the matter of fact, this only gets called | |
1517 | * when directory is guaranteed to have no in-lookup children | |
1518 | * at all. | |
44396f4b | 1519 | */ |
e3c13928 | 1520 | static struct dentry *__lookup_hash(const struct qstr *name, |
72bd866a | 1521 | struct dentry *base, unsigned int flags) |
a3255546 | 1522 | { |
6c51e513 | 1523 | struct dentry *dentry = lookup_dcache(name, base, flags); |
a03ece5f AV |
1524 | struct dentry *old; |
1525 | struct inode *dir = base->d_inode; | |
a3255546 | 1526 | |
6c51e513 | 1527 | if (dentry) |
bad61189 | 1528 | return dentry; |
a3255546 | 1529 | |
a03ece5f AV |
1530 | /* Don't create child dentry for a dead directory. */ |
1531 | if (unlikely(IS_DEADDIR(dir))) | |
1532 | return ERR_PTR(-ENOENT); | |
1533 | ||
6c51e513 AV |
1534 | dentry = d_alloc(base, name); |
1535 | if (unlikely(!dentry)) | |
1536 | return ERR_PTR(-ENOMEM); | |
1537 | ||
a03ece5f AV |
1538 | old = dir->i_op->lookup(dir, dentry, flags); |
1539 | if (unlikely(old)) { | |
1540 | dput(dentry); | |
1541 | dentry = old; | |
1542 | } | |
1543 | return dentry; | |
a3255546 AV |
1544 | } |
1545 | ||
e97cdc87 | 1546 | static int lookup_fast(struct nameidata *nd, |
254cf582 AV |
1547 | struct path *path, struct inode **inode, |
1548 | unsigned *seqp) | |
1da177e4 | 1549 | { |
4ac91378 | 1550 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1551 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 | 1552 | int status = 1; |
9875cf80 DH |
1553 | int err; |
1554 | ||
b04f784e NP |
1555 | /* |
1556 | * Rename seqlock is not required here because in the off chance | |
5d0f49c1 AV |
1557 | * of a false negative due to a concurrent rename, the caller is |
1558 | * going to fall back to non-racy lookup. | |
b04f784e | 1559 | */ |
31e6b01f NP |
1560 | if (nd->flags & LOOKUP_RCU) { |
1561 | unsigned seq; | |
766c4cbf | 1562 | bool negative; |
da53be12 | 1563 | dentry = __d_lookup_rcu(parent, &nd->last, &seq); |
5d0f49c1 | 1564 | if (unlikely(!dentry)) { |
4675ac39 | 1565 | if (unlazy_walk(nd)) |
5d0f49c1 | 1566 | return -ECHILD; |
e9742b53 | 1567 | return 0; |
5d0f49c1 | 1568 | } |
5a18fff2 | 1569 | |
12f8ad4b LT |
1570 | /* |
1571 | * This sequence count validates that the inode matches | |
1572 | * the dentry name information from lookup. | |
1573 | */ | |
63afdfc7 | 1574 | *inode = d_backing_inode(dentry); |
766c4cbf | 1575 | negative = d_is_negative(dentry); |
5d0f49c1 | 1576 | if (unlikely(read_seqcount_retry(&dentry->d_seq, seq))) |
12f8ad4b LT |
1577 | return -ECHILD; |
1578 | ||
1579 | /* | |
1580 | * This sequence count validates that the parent had no | |
1581 | * changes while we did the lookup of the dentry above. | |
1582 | * | |
1583 | * The memory barrier in read_seqcount_begin of child is | |
1584 | * enough, we can use __read_seqcount_retry here. | |
1585 | */ | |
5d0f49c1 | 1586 | if (unlikely(__read_seqcount_retry(&parent->d_seq, nd->seq))) |
31e6b01f | 1587 | return -ECHILD; |
5a18fff2 | 1588 | |
254cf582 | 1589 | *seqp = seq; |
a89f8337 | 1590 | status = d_revalidate(dentry, nd->flags); |
209a7fb2 | 1591 | if (likely(status > 0)) { |
5d0f49c1 AV |
1592 | /* |
1593 | * Note: do negative dentry check after revalidation in | |
1594 | * case that drops it. | |
1595 | */ | |
1596 | if (unlikely(negative)) | |
1597 | return -ENOENT; | |
1598 | path->mnt = mnt; | |
1599 | path->dentry = dentry; | |
1600 | if (likely(__follow_mount_rcu(nd, path, inode, seqp))) | |
e9742b53 | 1601 | return 1; |
24643087 | 1602 | } |
4675ac39 | 1603 | if (unlazy_child(nd, dentry, seq)) |
209a7fb2 AV |
1604 | return -ECHILD; |
1605 | if (unlikely(status == -ECHILD)) | |
1606 | /* we'd been told to redo it in non-rcu mode */ | |
1607 | status = d_revalidate(dentry, nd->flags); | |
5a18fff2 | 1608 | } else { |
e97cdc87 | 1609 | dentry = __d_lookup(parent, &nd->last); |
5d0f49c1 | 1610 | if (unlikely(!dentry)) |
e9742b53 | 1611 | return 0; |
a89f8337 | 1612 | status = d_revalidate(dentry, nd->flags); |
9875cf80 | 1613 | } |
5a18fff2 | 1614 | if (unlikely(status <= 0)) { |
e9742b53 | 1615 | if (!status) |
5d0f49c1 | 1616 | d_invalidate(dentry); |
5542aa2f | 1617 | dput(dentry); |
5d0f49c1 | 1618 | return status; |
24643087 | 1619 | } |
766c4cbf AV |
1620 | if (unlikely(d_is_negative(dentry))) { |
1621 | dput(dentry); | |
1622 | return -ENOENT; | |
1623 | } | |
5d0f49c1 | 1624 | |
9875cf80 DH |
1625 | path->mnt = mnt; |
1626 | path->dentry = dentry; | |
756daf26 | 1627 | err = follow_managed(path, nd); |
e9742b53 | 1628 | if (likely(err > 0)) |
63afdfc7 | 1629 | *inode = d_backing_inode(path->dentry); |
8402752e | 1630 | return err; |
697f514d MS |
1631 | } |
1632 | ||
1633 | /* Fast lookup failed, do it the slow way */ | |
88d8331a AV |
1634 | static struct dentry *__lookup_slow(const struct qstr *name, |
1635 | struct dentry *dir, | |
1636 | unsigned int flags) | |
697f514d | 1637 | { |
88d8331a | 1638 | struct dentry *dentry, *old; |
1936386e | 1639 | struct inode *inode = dir->d_inode; |
d9171b93 | 1640 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
1936386e | 1641 | |
1936386e | 1642 | /* Don't go there if it's already dead */ |
94bdd655 | 1643 | if (unlikely(IS_DEADDIR(inode))) |
88d8331a | 1644 | return ERR_PTR(-ENOENT); |
94bdd655 | 1645 | again: |
d9171b93 | 1646 | dentry = d_alloc_parallel(dir, name, &wq); |
94bdd655 | 1647 | if (IS_ERR(dentry)) |
88d8331a | 1648 | return dentry; |
94bdd655 | 1649 | if (unlikely(!d_in_lookup(dentry))) { |
a89f8337 | 1650 | if (!(flags & LOOKUP_NO_REVAL)) { |
949a852e AV |
1651 | int error = d_revalidate(dentry, flags); |
1652 | if (unlikely(error <= 0)) { | |
94bdd655 | 1653 | if (!error) { |
949a852e | 1654 | d_invalidate(dentry); |
94bdd655 AV |
1655 | dput(dentry); |
1656 | goto again; | |
1657 | } | |
949a852e AV |
1658 | dput(dentry); |
1659 | dentry = ERR_PTR(error); | |
1660 | } | |
1661 | } | |
94bdd655 AV |
1662 | } else { |
1663 | old = inode->i_op->lookup(inode, dentry, flags); | |
1664 | d_lookup_done(dentry); | |
1665 | if (unlikely(old)) { | |
1666 | dput(dentry); | |
1667 | dentry = old; | |
949a852e AV |
1668 | } |
1669 | } | |
e3c13928 | 1670 | return dentry; |
1da177e4 LT |
1671 | } |
1672 | ||
88d8331a AV |
1673 | static struct dentry *lookup_slow(const struct qstr *name, |
1674 | struct dentry *dir, | |
1675 | unsigned int flags) | |
1676 | { | |
1677 | struct inode *inode = dir->d_inode; | |
1678 | struct dentry *res; | |
1679 | inode_lock_shared(inode); | |
1680 | res = __lookup_slow(name, dir, flags); | |
1681 | inode_unlock_shared(inode); | |
1682 | return res; | |
1683 | } | |
1684 | ||
52094c8a AV |
1685 | static inline int may_lookup(struct nameidata *nd) |
1686 | { | |
1687 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1688 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1689 | if (err != -ECHILD) |
1690 | return err; | |
4675ac39 | 1691 | if (unlazy_walk(nd)) |
52094c8a AV |
1692 | return -ECHILD; |
1693 | } | |
4ad5abb3 | 1694 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1695 | } |
1696 | ||
9856fa1b AV |
1697 | static inline int handle_dots(struct nameidata *nd, int type) |
1698 | { | |
1699 | if (type == LAST_DOTDOT) { | |
9e6697e2 AV |
1700 | if (!nd->root.mnt) |
1701 | set_root(nd); | |
9856fa1b | 1702 | if (nd->flags & LOOKUP_RCU) { |
70291aec | 1703 | return follow_dotdot_rcu(nd); |
9856fa1b | 1704 | } else |
397d425d | 1705 | return follow_dotdot(nd); |
9856fa1b AV |
1706 | } |
1707 | return 0; | |
1708 | } | |
1709 | ||
181548c0 AV |
1710 | static int pick_link(struct nameidata *nd, struct path *link, |
1711 | struct inode *inode, unsigned seq) | |
d63ff28f | 1712 | { |
626de996 | 1713 | int error; |
1cf2665b | 1714 | struct saved *last; |
756daf26 | 1715 | if (unlikely(nd->total_link_count++ >= MAXSYMLINKS)) { |
626de996 AV |
1716 | path_to_nameidata(link, nd); |
1717 | return -ELOOP; | |
1718 | } | |
bc40aee0 | 1719 | if (!(nd->flags & LOOKUP_RCU)) { |
7973387a AV |
1720 | if (link->mnt == nd->path.mnt) |
1721 | mntget(link->mnt); | |
d63ff28f | 1722 | } |
626de996 AV |
1723 | error = nd_alloc_stack(nd); |
1724 | if (unlikely(error)) { | |
bc40aee0 | 1725 | if (error == -ECHILD) { |
ad1633a1 AV |
1726 | if (unlikely(!legitimize_path(nd, link, seq))) { |
1727 | drop_links(nd); | |
1728 | nd->depth = 0; | |
1729 | nd->flags &= ~LOOKUP_RCU; | |
1730 | nd->path.mnt = NULL; | |
1731 | nd->path.dentry = NULL; | |
ad1633a1 | 1732 | rcu_read_unlock(); |
4675ac39 | 1733 | } else if (likely(unlazy_walk(nd)) == 0) |
ad1633a1 | 1734 | error = nd_alloc_stack(nd); |
bc40aee0 AV |
1735 | } |
1736 | if (error) { | |
1737 | path_put(link); | |
1738 | return error; | |
1739 | } | |
626de996 AV |
1740 | } |
1741 | ||
ab104923 | 1742 | last = nd->stack + nd->depth++; |
1cf2665b | 1743 | last->link = *link; |
fceef393 AV |
1744 | clear_delayed_call(&last->done); |
1745 | nd->link_inode = inode; | |
0450b2d1 | 1746 | last->seq = seq; |
d63ff28f AV |
1747 | return 1; |
1748 | } | |
1749 | ||
8f64fb1c | 1750 | enum {WALK_FOLLOW = 1, WALK_MORE = 2}; |
31d66bcd | 1751 | |
3ddcd056 LT |
1752 | /* |
1753 | * Do we need to follow links? We _really_ want to be able | |
1754 | * to do this check without having to look at inode->i_op, | |
1755 | * so we keep a cache of "no, this doesn't need follow_link" | |
1756 | * for the common case. | |
1757 | */ | |
8f64fb1c AV |
1758 | static inline int step_into(struct nameidata *nd, struct path *path, |
1759 | int flags, struct inode *inode, unsigned seq) | |
3ddcd056 | 1760 | { |
31d66bcd AV |
1761 | if (!(flags & WALK_MORE) && nd->depth) |
1762 | put_link(nd); | |
8f64fb1c AV |
1763 | if (likely(!d_is_symlink(path->dentry)) || |
1764 | !(flags & WALK_FOLLOW || nd->flags & LOOKUP_FOLLOW)) { | |
1765 | /* not a symlink or should not follow */ | |
1766 | path_to_nameidata(path, nd); | |
1767 | nd->inode = inode; | |
1768 | nd->seq = seq; | |
d63ff28f | 1769 | return 0; |
8f64fb1c | 1770 | } |
a7f77542 AV |
1771 | /* make sure that d_is_symlink above matches inode */ |
1772 | if (nd->flags & LOOKUP_RCU) { | |
8f64fb1c | 1773 | if (read_seqcount_retry(&path->dentry->d_seq, seq)) |
a7f77542 AV |
1774 | return -ECHILD; |
1775 | } | |
8f64fb1c | 1776 | return pick_link(nd, path, inode, seq); |
3ddcd056 LT |
1777 | } |
1778 | ||
4693a547 | 1779 | static int walk_component(struct nameidata *nd, int flags) |
ce57dfc1 | 1780 | { |
caa85634 | 1781 | struct path path; |
ce57dfc1 | 1782 | struct inode *inode; |
254cf582 | 1783 | unsigned seq; |
ce57dfc1 AV |
1784 | int err; |
1785 | /* | |
1786 | * "." and ".." are special - ".." especially so because it has | |
1787 | * to be able to know about the current root directory and | |
1788 | * parent relationships. | |
1789 | */ | |
4693a547 AV |
1790 | if (unlikely(nd->last_type != LAST_NORM)) { |
1791 | err = handle_dots(nd, nd->last_type); | |
1c4ff1a8 | 1792 | if (!(flags & WALK_MORE) && nd->depth) |
4693a547 AV |
1793 | put_link(nd); |
1794 | return err; | |
1795 | } | |
254cf582 | 1796 | err = lookup_fast(nd, &path, &inode, &seq); |
e9742b53 | 1797 | if (unlikely(err <= 0)) { |
697f514d | 1798 | if (err < 0) |
f0a9ba70 | 1799 | return err; |
e3c13928 AV |
1800 | path.dentry = lookup_slow(&nd->last, nd->path.dentry, |
1801 | nd->flags); | |
1802 | if (IS_ERR(path.dentry)) | |
1803 | return PTR_ERR(path.dentry); | |
7500c38a | 1804 | |
e3c13928 AV |
1805 | path.mnt = nd->path.mnt; |
1806 | err = follow_managed(&path, nd); | |
1807 | if (unlikely(err < 0)) | |
f0a9ba70 | 1808 | return err; |
697f514d | 1809 | |
7500c38a AV |
1810 | if (unlikely(d_is_negative(path.dentry))) { |
1811 | path_to_nameidata(&path, nd); | |
1812 | return -ENOENT; | |
1813 | } | |
1814 | ||
254cf582 | 1815 | seq = 0; /* we are already out of RCU mode */ |
d4565649 | 1816 | inode = d_backing_inode(path.dentry); |
ce57dfc1 | 1817 | } |
697f514d | 1818 | |
8f64fb1c | 1819 | return step_into(nd, &path, flags, inode, seq); |
ce57dfc1 AV |
1820 | } |
1821 | ||
bfcfaa77 LT |
1822 | /* |
1823 | * We can do the critical dentry name comparison and hashing | |
1824 | * operations one word at a time, but we are limited to: | |
1825 | * | |
1826 | * - Architectures with fast unaligned word accesses. We could | |
1827 | * do a "get_unaligned()" if this helps and is sufficiently | |
1828 | * fast. | |
1829 | * | |
bfcfaa77 LT |
1830 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we |
1831 | * do not trap on the (extremely unlikely) case of a page | |
1832 | * crossing operation. | |
1833 | * | |
1834 | * - Furthermore, we need an efficient 64-bit compile for the | |
1835 | * 64-bit case in order to generate the "number of bytes in | |
1836 | * the final mask". Again, that could be replaced with a | |
1837 | * efficient population count instruction or similar. | |
1838 | */ | |
1839 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1840 | ||
f68e556e | 1841 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1842 | |
468a9428 | 1843 | #ifdef HASH_MIX |
bfcfaa77 | 1844 | |
468a9428 | 1845 | /* Architecture provides HASH_MIX and fold_hash() in <asm/hash.h> */ |
bfcfaa77 | 1846 | |
468a9428 | 1847 | #elif defined(CONFIG_64BIT) |
0fed3ac8 | 1848 | /* |
2a18da7a GS |
1849 | * Register pressure in the mixing function is an issue, particularly |
1850 | * on 32-bit x86, but almost any function requires one state value and | |
1851 | * one temporary. Instead, use a function designed for two state values | |
1852 | * and no temporaries. | |
1853 | * | |
1854 | * This function cannot create a collision in only two iterations, so | |
1855 | * we have two iterations to achieve avalanche. In those two iterations, | |
1856 | * we have six layers of mixing, which is enough to spread one bit's | |
1857 | * influence out to 2^6 = 64 state bits. | |
1858 | * | |
1859 | * Rotate constants are scored by considering either 64 one-bit input | |
1860 | * deltas or 64*63/2 = 2016 two-bit input deltas, and finding the | |
1861 | * probability of that delta causing a change to each of the 128 output | |
1862 | * bits, using a sample of random initial states. | |
1863 | * | |
1864 | * The Shannon entropy of the computed probabilities is then summed | |
1865 | * to produce a score. Ideally, any input change has a 50% chance of | |
1866 | * toggling any given output bit. | |
1867 | * | |
1868 | * Mixing scores (in bits) for (12,45): | |
1869 | * Input delta: 1-bit 2-bit | |
1870 | * 1 round: 713.3 42542.6 | |
1871 | * 2 rounds: 2753.7 140389.8 | |
1872 | * 3 rounds: 5954.1 233458.2 | |
1873 | * 4 rounds: 7862.6 256672.2 | |
1874 | * Perfect: 8192 258048 | |
1875 | * (64*128) (64*63/2 * 128) | |
0fed3ac8 | 1876 | */ |
2a18da7a GS |
1877 | #define HASH_MIX(x, y, a) \ |
1878 | ( x ^= (a), \ | |
1879 | y ^= x, x = rol64(x,12),\ | |
1880 | x += y, y = rol64(y,45),\ | |
1881 | y *= 9 ) | |
bfcfaa77 | 1882 | |
0fed3ac8 | 1883 | /* |
2a18da7a GS |
1884 | * Fold two longs into one 32-bit hash value. This must be fast, but |
1885 | * latency isn't quite as critical, as there is a fair bit of additional | |
1886 | * work done before the hash value is used. | |
0fed3ac8 | 1887 | */ |
2a18da7a | 1888 | static inline unsigned int fold_hash(unsigned long x, unsigned long y) |
0fed3ac8 | 1889 | { |
2a18da7a GS |
1890 | y ^= x * GOLDEN_RATIO_64; |
1891 | y *= GOLDEN_RATIO_64; | |
1892 | return y >> 32; | |
0fed3ac8 GS |
1893 | } |
1894 | ||
bfcfaa77 LT |
1895 | #else /* 32-bit case */ |
1896 | ||
2a18da7a GS |
1897 | /* |
1898 | * Mixing scores (in bits) for (7,20): | |
1899 | * Input delta: 1-bit 2-bit | |
1900 | * 1 round: 330.3 9201.6 | |
1901 | * 2 rounds: 1246.4 25475.4 | |
1902 | * 3 rounds: 1907.1 31295.1 | |
1903 | * 4 rounds: 2042.3 31718.6 | |
1904 | * Perfect: 2048 31744 | |
1905 | * (32*64) (32*31/2 * 64) | |
1906 | */ | |
1907 | #define HASH_MIX(x, y, a) \ | |
1908 | ( x ^= (a), \ | |
1909 | y ^= x, x = rol32(x, 7),\ | |
1910 | x += y, y = rol32(y,20),\ | |
1911 | y *= 9 ) | |
bfcfaa77 | 1912 | |
2a18da7a | 1913 | static inline unsigned int fold_hash(unsigned long x, unsigned long y) |
0fed3ac8 | 1914 | { |
2a18da7a GS |
1915 | /* Use arch-optimized multiply if one exists */ |
1916 | return __hash_32(y ^ __hash_32(x)); | |
0fed3ac8 GS |
1917 | } |
1918 | ||
bfcfaa77 LT |
1919 | #endif |
1920 | ||
2a18da7a GS |
1921 | /* |
1922 | * Return the hash of a string of known length. This is carfully | |
1923 | * designed to match hash_name(), which is the more critical function. | |
1924 | * In particular, we must end by hashing a final word containing 0..7 | |
1925 | * payload bytes, to match the way that hash_name() iterates until it | |
1926 | * finds the delimiter after the name. | |
1927 | */ | |
8387ff25 | 1928 | unsigned int full_name_hash(const void *salt, const char *name, unsigned int len) |
bfcfaa77 | 1929 | { |
8387ff25 | 1930 | unsigned long a, x = 0, y = (unsigned long)salt; |
bfcfaa77 LT |
1931 | |
1932 | for (;;) { | |
fcfd2fbf GS |
1933 | if (!len) |
1934 | goto done; | |
e419b4cc | 1935 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1936 | if (len < sizeof(unsigned long)) |
1937 | break; | |
2a18da7a | 1938 | HASH_MIX(x, y, a); |
bfcfaa77 LT |
1939 | name += sizeof(unsigned long); |
1940 | len -= sizeof(unsigned long); | |
bfcfaa77 | 1941 | } |
2a18da7a | 1942 | x ^= a & bytemask_from_count(len); |
bfcfaa77 | 1943 | done: |
2a18da7a | 1944 | return fold_hash(x, y); |
bfcfaa77 LT |
1945 | } |
1946 | EXPORT_SYMBOL(full_name_hash); | |
1947 | ||
fcfd2fbf | 1948 | /* Return the "hash_len" (hash and length) of a null-terminated string */ |
8387ff25 | 1949 | u64 hashlen_string(const void *salt, const char *name) |
fcfd2fbf | 1950 | { |
8387ff25 LT |
1951 | unsigned long a = 0, x = 0, y = (unsigned long)salt; |
1952 | unsigned long adata, mask, len; | |
fcfd2fbf GS |
1953 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; |
1954 | ||
8387ff25 LT |
1955 | len = 0; |
1956 | goto inside; | |
1957 | ||
fcfd2fbf | 1958 | do { |
2a18da7a | 1959 | HASH_MIX(x, y, a); |
fcfd2fbf | 1960 | len += sizeof(unsigned long); |
8387ff25 | 1961 | inside: |
fcfd2fbf GS |
1962 | a = load_unaligned_zeropad(name+len); |
1963 | } while (!has_zero(a, &adata, &constants)); | |
1964 | ||
1965 | adata = prep_zero_mask(a, adata, &constants); | |
1966 | mask = create_zero_mask(adata); | |
2a18da7a | 1967 | x ^= a & zero_bytemask(mask); |
fcfd2fbf | 1968 | |
2a18da7a | 1969 | return hashlen_create(fold_hash(x, y), len + find_zero(mask)); |
fcfd2fbf GS |
1970 | } |
1971 | EXPORT_SYMBOL(hashlen_string); | |
1972 | ||
bfcfaa77 LT |
1973 | /* |
1974 | * Calculate the length and hash of the path component, and | |
d6bb3e90 | 1975 | * return the "hash_len" as the result. |
bfcfaa77 | 1976 | */ |
8387ff25 | 1977 | static inline u64 hash_name(const void *salt, const char *name) |
bfcfaa77 | 1978 | { |
8387ff25 LT |
1979 | unsigned long a = 0, b, x = 0, y = (unsigned long)salt; |
1980 | unsigned long adata, bdata, mask, len; | |
36126f8f | 1981 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; |
bfcfaa77 | 1982 | |
8387ff25 LT |
1983 | len = 0; |
1984 | goto inside; | |
1985 | ||
bfcfaa77 | 1986 | do { |
2a18da7a | 1987 | HASH_MIX(x, y, a); |
bfcfaa77 | 1988 | len += sizeof(unsigned long); |
8387ff25 | 1989 | inside: |
e419b4cc | 1990 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1991 | b = a ^ REPEAT_BYTE('/'); |
1992 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1993 | ||
1994 | adata = prep_zero_mask(a, adata, &constants); | |
1995 | bdata = prep_zero_mask(b, bdata, &constants); | |
36126f8f | 1996 | mask = create_zero_mask(adata | bdata); |
2a18da7a | 1997 | x ^= a & zero_bytemask(mask); |
36126f8f | 1998 | |
2a18da7a | 1999 | return hashlen_create(fold_hash(x, y), len + find_zero(mask)); |
bfcfaa77 LT |
2000 | } |
2001 | ||
2a18da7a | 2002 | #else /* !CONFIG_DCACHE_WORD_ACCESS: Slow, byte-at-a-time version */ |
bfcfaa77 | 2003 | |
fcfd2fbf | 2004 | /* Return the hash of a string of known length */ |
8387ff25 | 2005 | unsigned int full_name_hash(const void *salt, const char *name, unsigned int len) |
0145acc2 | 2006 | { |
8387ff25 | 2007 | unsigned long hash = init_name_hash(salt); |
0145acc2 | 2008 | while (len--) |
fcfd2fbf | 2009 | hash = partial_name_hash((unsigned char)*name++, hash); |
0145acc2 LT |
2010 | return end_name_hash(hash); |
2011 | } | |
ae942ae7 | 2012 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 2013 | |
fcfd2fbf | 2014 | /* Return the "hash_len" (hash and length) of a null-terminated string */ |
8387ff25 | 2015 | u64 hashlen_string(const void *salt, const char *name) |
fcfd2fbf | 2016 | { |
8387ff25 | 2017 | unsigned long hash = init_name_hash(salt); |
fcfd2fbf GS |
2018 | unsigned long len = 0, c; |
2019 | ||
2020 | c = (unsigned char)*name; | |
e0ab7af9 | 2021 | while (c) { |
fcfd2fbf GS |
2022 | len++; |
2023 | hash = partial_name_hash(c, hash); | |
2024 | c = (unsigned char)name[len]; | |
e0ab7af9 | 2025 | } |
fcfd2fbf GS |
2026 | return hashlen_create(end_name_hash(hash), len); |
2027 | } | |
f2a031b6 | 2028 | EXPORT_SYMBOL(hashlen_string); |
fcfd2fbf | 2029 | |
200e9ef7 LT |
2030 | /* |
2031 | * We know there's a real path component here of at least | |
2032 | * one character. | |
2033 | */ | |
8387ff25 | 2034 | static inline u64 hash_name(const void *salt, const char *name) |
200e9ef7 | 2035 | { |
8387ff25 | 2036 | unsigned long hash = init_name_hash(salt); |
200e9ef7 LT |
2037 | unsigned long len = 0, c; |
2038 | ||
2039 | c = (unsigned char)*name; | |
2040 | do { | |
2041 | len++; | |
2042 | hash = partial_name_hash(c, hash); | |
2043 | c = (unsigned char)name[len]; | |
2044 | } while (c && c != '/'); | |
d6bb3e90 | 2045 | return hashlen_create(end_name_hash(hash), len); |
200e9ef7 LT |
2046 | } |
2047 | ||
bfcfaa77 LT |
2048 | #endif |
2049 | ||
1da177e4 LT |
2050 | /* |
2051 | * Name resolution. | |
ea3834d9 PM |
2052 | * This is the basic name resolution function, turning a pathname into |
2053 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 2054 | * |
ea3834d9 PM |
2055 | * Returns 0 and nd will have valid dentry and mnt on success. |
2056 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 2057 | */ |
6de88d72 | 2058 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 2059 | { |
1da177e4 | 2060 | int err; |
32cd7468 | 2061 | |
9b5858e9 AV |
2062 | if (IS_ERR(name)) |
2063 | return PTR_ERR(name); | |
1da177e4 LT |
2064 | while (*name=='/') |
2065 | name++; | |
2066 | if (!*name) | |
9e18f10a | 2067 | return 0; |
1da177e4 | 2068 | |
1da177e4 LT |
2069 | /* At this point we know we have a real path component. */ |
2070 | for(;;) { | |
d6bb3e90 | 2071 | u64 hash_len; |
fe479a58 | 2072 | int type; |
1da177e4 | 2073 | |
52094c8a | 2074 | err = may_lookup(nd); |
2a18da7a | 2075 | if (err) |
3595e234 | 2076 | return err; |
1da177e4 | 2077 | |
8387ff25 | 2078 | hash_len = hash_name(nd->path.dentry, name); |
1da177e4 | 2079 | |
fe479a58 | 2080 | type = LAST_NORM; |
d6bb3e90 | 2081 | if (name[0] == '.') switch (hashlen_len(hash_len)) { |
fe479a58 | 2082 | case 2: |
200e9ef7 | 2083 | if (name[1] == '.') { |
fe479a58 | 2084 | type = LAST_DOTDOT; |
16c2cd71 AV |
2085 | nd->flags |= LOOKUP_JUMPED; |
2086 | } | |
fe479a58 AV |
2087 | break; |
2088 | case 1: | |
2089 | type = LAST_DOT; | |
2090 | } | |
5a202bcd AV |
2091 | if (likely(type == LAST_NORM)) { |
2092 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 2093 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd | 2094 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
a060dc50 | 2095 | struct qstr this = { { .hash_len = hash_len }, .name = name }; |
da53be12 | 2096 | err = parent->d_op->d_hash(parent, &this); |
5a202bcd | 2097 | if (err < 0) |
3595e234 | 2098 | return err; |
d6bb3e90 LT |
2099 | hash_len = this.hash_len; |
2100 | name = this.name; | |
5a202bcd AV |
2101 | } |
2102 | } | |
fe479a58 | 2103 | |
d6bb3e90 LT |
2104 | nd->last.hash_len = hash_len; |
2105 | nd->last.name = name; | |
5f4a6a69 AV |
2106 | nd->last_type = type; |
2107 | ||
d6bb3e90 LT |
2108 | name += hashlen_len(hash_len); |
2109 | if (!*name) | |
bdf6cbf1 | 2110 | goto OK; |
200e9ef7 LT |
2111 | /* |
2112 | * If it wasn't NUL, we know it was '/'. Skip that | |
2113 | * slash, and continue until no more slashes. | |
2114 | */ | |
2115 | do { | |
d6bb3e90 LT |
2116 | name++; |
2117 | } while (unlikely(*name == '/')); | |
8620c238 AV |
2118 | if (unlikely(!*name)) { |
2119 | OK: | |
368ee9ba | 2120 | /* pathname body, done */ |
8620c238 AV |
2121 | if (!nd->depth) |
2122 | return 0; | |
2123 | name = nd->stack[nd->depth - 1].name; | |
368ee9ba | 2124 | /* trailing symlink, done */ |
8620c238 AV |
2125 | if (!name) |
2126 | return 0; | |
2127 | /* last component of nested symlink */ | |
8f64fb1c | 2128 | err = walk_component(nd, WALK_FOLLOW); |
1c4ff1a8 AV |
2129 | } else { |
2130 | /* not the last component */ | |
8f64fb1c | 2131 | err = walk_component(nd, WALK_FOLLOW | WALK_MORE); |
8620c238 | 2132 | } |
ce57dfc1 | 2133 | if (err < 0) |
3595e234 | 2134 | return err; |
1da177e4 | 2135 | |
ce57dfc1 | 2136 | if (err) { |
626de996 | 2137 | const char *s = get_link(nd); |
5a460275 | 2138 | |
a1c83681 | 2139 | if (IS_ERR(s)) |
3595e234 | 2140 | return PTR_ERR(s); |
d40bcc09 AV |
2141 | err = 0; |
2142 | if (unlikely(!s)) { | |
2143 | /* jumped */ | |
b9ff4429 | 2144 | put_link(nd); |
d40bcc09 | 2145 | } else { |
fab51e8a AV |
2146 | nd->stack[nd->depth - 1].name = name; |
2147 | name = s; | |
2148 | continue; | |
d40bcc09 | 2149 | } |
31e6b01f | 2150 | } |
97242f99 AV |
2151 | if (unlikely(!d_can_lookup(nd->path.dentry))) { |
2152 | if (nd->flags & LOOKUP_RCU) { | |
4675ac39 | 2153 | if (unlazy_walk(nd)) |
97242f99 AV |
2154 | return -ECHILD; |
2155 | } | |
3595e234 | 2156 | return -ENOTDIR; |
97242f99 | 2157 | } |
1da177e4 | 2158 | } |
1da177e4 LT |
2159 | } |
2160 | ||
edc2b1da | 2161 | /* must be paired with terminate_walk() */ |
c8a53ee5 | 2162 | static const char *path_init(struct nameidata *nd, unsigned flags) |
31e6b01f | 2163 | { |
c8a53ee5 | 2164 | const char *s = nd->name->name; |
31e6b01f | 2165 | |
c0eb027e LT |
2166 | if (!*s) |
2167 | flags &= ~LOOKUP_RCU; | |
edc2b1da AV |
2168 | if (flags & LOOKUP_RCU) |
2169 | rcu_read_lock(); | |
c0eb027e | 2170 | |
31e6b01f | 2171 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ |
980f3ea2 | 2172 | nd->flags = flags | LOOKUP_JUMPED | LOOKUP_PARENT; |
31e6b01f | 2173 | nd->depth = 0; |
5b6ca027 | 2174 | if (flags & LOOKUP_ROOT) { |
b18825a7 DH |
2175 | struct dentry *root = nd->root.dentry; |
2176 | struct inode *inode = root->d_inode; | |
93893862 AV |
2177 | if (*s && unlikely(!d_can_lookup(root))) |
2178 | return ERR_PTR(-ENOTDIR); | |
5b6ca027 AV |
2179 | nd->path = nd->root; |
2180 | nd->inode = inode; | |
2181 | if (flags & LOOKUP_RCU) { | |
5b6ca027 | 2182 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
8f47a016 | 2183 | nd->root_seq = nd->seq; |
48a066e7 | 2184 | nd->m_seq = read_seqbegin(&mount_lock); |
5b6ca027 AV |
2185 | } else { |
2186 | path_get(&nd->path); | |
2187 | } | |
368ee9ba | 2188 | return s; |
5b6ca027 AV |
2189 | } |
2190 | ||
31e6b01f | 2191 | nd->root.mnt = NULL; |
248fb5b9 AV |
2192 | nd->path.mnt = NULL; |
2193 | nd->path.dentry = NULL; | |
31e6b01f | 2194 | |
48a066e7 | 2195 | nd->m_seq = read_seqbegin(&mount_lock); |
fd2f7cb5 | 2196 | if (*s == '/') { |
9e6697e2 | 2197 | set_root(nd); |
248fb5b9 | 2198 | if (likely(!nd_jump_root(nd))) |
ef55d917 | 2199 | return s; |
ef55d917 | 2200 | return ERR_PTR(-ECHILD); |
c8a53ee5 | 2201 | } else if (nd->dfd == AT_FDCWD) { |
e41f7d4e AV |
2202 | if (flags & LOOKUP_RCU) { |
2203 | struct fs_struct *fs = current->fs; | |
2204 | unsigned seq; | |
31e6b01f | 2205 | |
e41f7d4e AV |
2206 | do { |
2207 | seq = read_seqcount_begin(&fs->seq); | |
2208 | nd->path = fs->pwd; | |
ef55d917 | 2209 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e AV |
2210 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
2211 | } while (read_seqcount_retry(&fs->seq, seq)); | |
2212 | } else { | |
2213 | get_fs_pwd(current->fs, &nd->path); | |
ef55d917 | 2214 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2215 | } |
ef55d917 | 2216 | return s; |
31e6b01f | 2217 | } else { |
582aa64a | 2218 | /* Caller must check execute permissions on the starting path component */ |
c8a53ee5 | 2219 | struct fd f = fdget_raw(nd->dfd); |
31e6b01f NP |
2220 | struct dentry *dentry; |
2221 | ||
2903ff01 | 2222 | if (!f.file) |
368ee9ba | 2223 | return ERR_PTR(-EBADF); |
31e6b01f | 2224 | |
2903ff01 | 2225 | dentry = f.file->f_path.dentry; |
31e6b01f | 2226 | |
edc2b1da AV |
2227 | if (*s && unlikely(!d_can_lookup(dentry))) { |
2228 | fdput(f); | |
2229 | return ERR_PTR(-ENOTDIR); | |
f52e0c11 | 2230 | } |
31e6b01f | 2231 | |
2903ff01 | 2232 | nd->path = f.file->f_path; |
e41f7d4e | 2233 | if (flags & LOOKUP_RCU) { |
34a26b99 AV |
2234 | nd->inode = nd->path.dentry->d_inode; |
2235 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
e41f7d4e | 2236 | } else { |
2903ff01 | 2237 | path_get(&nd->path); |
34a26b99 | 2238 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2239 | } |
34a26b99 | 2240 | fdput(f); |
368ee9ba | 2241 | return s; |
31e6b01f | 2242 | } |
9b4a9b14 AV |
2243 | } |
2244 | ||
3bdba28b | 2245 | static const char *trailing_symlink(struct nameidata *nd) |
95fa25d9 AV |
2246 | { |
2247 | const char *s; | |
fec2fa24 | 2248 | int error = may_follow_link(nd); |
deb106c6 | 2249 | if (unlikely(error)) |
3bdba28b | 2250 | return ERR_PTR(error); |
95fa25d9 | 2251 | nd->flags |= LOOKUP_PARENT; |
fab51e8a | 2252 | nd->stack[0].name = NULL; |
3b2e7f75 | 2253 | s = get_link(nd); |
deb106c6 | 2254 | return s ? s : ""; |
95fa25d9 AV |
2255 | } |
2256 | ||
caa85634 | 2257 | static inline int lookup_last(struct nameidata *nd) |
bd92d7fe AV |
2258 | { |
2259 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
2260 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
2261 | ||
2262 | nd->flags &= ~LOOKUP_PARENT; | |
1c4ff1a8 | 2263 | return walk_component(nd, 0); |
bd92d7fe AV |
2264 | } |
2265 | ||
4f757f3c AV |
2266 | static int handle_lookup_down(struct nameidata *nd) |
2267 | { | |
2268 | struct path path = nd->path; | |
2269 | struct inode *inode = nd->inode; | |
2270 | unsigned seq = nd->seq; | |
2271 | int err; | |
2272 | ||
2273 | if (nd->flags & LOOKUP_RCU) { | |
2274 | /* | |
2275 | * don't bother with unlazy_walk on failure - we are | |
2276 | * at the very beginning of walk, so we lose nothing | |
2277 | * if we simply redo everything in non-RCU mode | |
2278 | */ | |
2279 | if (unlikely(!__follow_mount_rcu(nd, &path, &inode, &seq))) | |
2280 | return -ECHILD; | |
2281 | } else { | |
2282 | dget(path.dentry); | |
2283 | err = follow_managed(&path, nd); | |
2284 | if (unlikely(err < 0)) | |
2285 | return err; | |
2286 | inode = d_backing_inode(path.dentry); | |
2287 | seq = 0; | |
2288 | } | |
2289 | path_to_nameidata(&path, nd); | |
2290 | nd->inode = inode; | |
2291 | nd->seq = seq; | |
2292 | return 0; | |
2293 | } | |
2294 | ||
9b4a9b14 | 2295 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2296 | static int path_lookupat(struct nameidata *nd, unsigned flags, struct path *path) |
9b4a9b14 | 2297 | { |
c8a53ee5 | 2298 | const char *s = path_init(nd, flags); |
bd92d7fe | 2299 | int err; |
31e6b01f | 2300 | |
9b5858e9 | 2301 | if (unlikely(flags & LOOKUP_DOWN) && !IS_ERR(s)) { |
4f757f3c | 2302 | err = handle_lookup_down(nd); |
5f336e72 AV |
2303 | if (unlikely(err < 0)) |
2304 | s = ERR_PTR(err); | |
4f757f3c AV |
2305 | } |
2306 | ||
3bdba28b AV |
2307 | while (!(err = link_path_walk(s, nd)) |
2308 | && ((err = lookup_last(nd)) > 0)) { | |
2309 | s = trailing_symlink(nd); | |
bd92d7fe | 2310 | } |
9f1fafee AV |
2311 | if (!err) |
2312 | err = complete_walk(nd); | |
bd92d7fe | 2313 | |
deb106c6 AV |
2314 | if (!err && nd->flags & LOOKUP_DIRECTORY) |
2315 | if (!d_can_lookup(nd->path.dentry)) | |
bd23a539 | 2316 | err = -ENOTDIR; |
625b6d10 AV |
2317 | if (!err) { |
2318 | *path = nd->path; | |
2319 | nd->path.mnt = NULL; | |
2320 | nd->path.dentry = NULL; | |
2321 | } | |
2322 | terminate_walk(nd); | |
bd92d7fe | 2323 | return err; |
ee0827cd | 2324 | } |
31e6b01f | 2325 | |
31d921c7 DH |
2326 | int filename_lookup(int dfd, struct filename *name, unsigned flags, |
2327 | struct path *path, struct path *root) | |
ee0827cd | 2328 | { |
894bc8c4 | 2329 | int retval; |
9883d185 | 2330 | struct nameidata nd; |
abc9f5be AV |
2331 | if (IS_ERR(name)) |
2332 | return PTR_ERR(name); | |
9ad1aaa6 AV |
2333 | if (unlikely(root)) { |
2334 | nd.root = *root; | |
2335 | flags |= LOOKUP_ROOT; | |
2336 | } | |
9883d185 | 2337 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2338 | retval = path_lookupat(&nd, flags | LOOKUP_RCU, path); |
ee0827cd | 2339 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2340 | retval = path_lookupat(&nd, flags, path); |
ee0827cd | 2341 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2342 | retval = path_lookupat(&nd, flags | LOOKUP_REVAL, path); |
31e6b01f | 2343 | |
f78570dd | 2344 | if (likely(!retval)) |
ff0ebee2 | 2345 | audit_inode(name, path->dentry, 0); |
9883d185 | 2346 | restore_nameidata(); |
e4bd1c1a | 2347 | putname(name); |
170aa3d0 | 2348 | return retval; |
1da177e4 LT |
2349 | } |
2350 | ||
8bcb77fa | 2351 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2352 | static int path_parentat(struct nameidata *nd, unsigned flags, |
391172c4 | 2353 | struct path *parent) |
8bcb77fa | 2354 | { |
c8a53ee5 | 2355 | const char *s = path_init(nd, flags); |
9b5858e9 | 2356 | int err = link_path_walk(s, nd); |
8bcb77fa AV |
2357 | if (!err) |
2358 | err = complete_walk(nd); | |
391172c4 AV |
2359 | if (!err) { |
2360 | *parent = nd->path; | |
2361 | nd->path.mnt = NULL; | |
2362 | nd->path.dentry = NULL; | |
2363 | } | |
2364 | terminate_walk(nd); | |
8bcb77fa AV |
2365 | return err; |
2366 | } | |
2367 | ||
5c31b6ce | 2368 | static struct filename *filename_parentat(int dfd, struct filename *name, |
391172c4 AV |
2369 | unsigned int flags, struct path *parent, |
2370 | struct qstr *last, int *type) | |
8bcb77fa AV |
2371 | { |
2372 | int retval; | |
9883d185 | 2373 | struct nameidata nd; |
8bcb77fa | 2374 | |
5c31b6ce AV |
2375 | if (IS_ERR(name)) |
2376 | return name; | |
9883d185 | 2377 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2378 | retval = path_parentat(&nd, flags | LOOKUP_RCU, parent); |
8bcb77fa | 2379 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2380 | retval = path_parentat(&nd, flags, parent); |
8bcb77fa | 2381 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2382 | retval = path_parentat(&nd, flags | LOOKUP_REVAL, parent); |
391172c4 AV |
2383 | if (likely(!retval)) { |
2384 | *last = nd.last; | |
2385 | *type = nd.last_type; | |
c9b07eab | 2386 | audit_inode(name, parent->dentry, AUDIT_INODE_PARENT); |
5c31b6ce AV |
2387 | } else { |
2388 | putname(name); | |
2389 | name = ERR_PTR(retval); | |
391172c4 | 2390 | } |
9883d185 | 2391 | restore_nameidata(); |
5c31b6ce | 2392 | return name; |
8bcb77fa AV |
2393 | } |
2394 | ||
79714f72 AV |
2395 | /* does lookup, returns the object with parent locked */ |
2396 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 2397 | { |
5c31b6ce AV |
2398 | struct filename *filename; |
2399 | struct dentry *d; | |
391172c4 AV |
2400 | struct qstr last; |
2401 | int type; | |
51689104 | 2402 | |
5c31b6ce AV |
2403 | filename = filename_parentat(AT_FDCWD, getname_kernel(name), 0, path, |
2404 | &last, &type); | |
51689104 PM |
2405 | if (IS_ERR(filename)) |
2406 | return ERR_CAST(filename); | |
5c31b6ce | 2407 | if (unlikely(type != LAST_NORM)) { |
391172c4 | 2408 | path_put(path); |
5c31b6ce AV |
2409 | putname(filename); |
2410 | return ERR_PTR(-EINVAL); | |
79714f72 | 2411 | } |
5955102c | 2412 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 2413 | d = __lookup_hash(&last, path->dentry, 0); |
79714f72 | 2414 | if (IS_ERR(d)) { |
5955102c | 2415 | inode_unlock(path->dentry->d_inode); |
391172c4 | 2416 | path_put(path); |
79714f72 | 2417 | } |
51689104 | 2418 | putname(filename); |
79714f72 | 2419 | return d; |
5590ff0d UD |
2420 | } |
2421 | ||
d1811465 AV |
2422 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2423 | { | |
abc9f5be AV |
2424 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2425 | flags, path, NULL); | |
d1811465 | 2426 | } |
4d359507 | 2427 | EXPORT_SYMBOL(kern_path); |
d1811465 | 2428 | |
16f18200 JJS |
2429 | /** |
2430 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2431 | * @dentry: pointer to dentry of the base directory | |
2432 | * @mnt: pointer to vfs mount of the base directory | |
2433 | * @name: pointer to file name | |
2434 | * @flags: lookup flags | |
e0a01249 | 2435 | * @path: pointer to struct path to fill |
16f18200 JJS |
2436 | */ |
2437 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2438 | const char *name, unsigned int flags, | |
e0a01249 | 2439 | struct path *path) |
16f18200 | 2440 | { |
9ad1aaa6 | 2441 | struct path root = {.mnt = mnt, .dentry = dentry}; |
9ad1aaa6 | 2442 | /* the first argument of filename_lookup() is ignored with root */ |
abc9f5be AV |
2443 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2444 | flags , path, &root); | |
16f18200 | 2445 | } |
4d359507 | 2446 | EXPORT_SYMBOL(vfs_path_lookup); |
16f18200 | 2447 | |
3c95f0dc AV |
2448 | static int lookup_one_len_common(const char *name, struct dentry *base, |
2449 | int len, struct qstr *this) | |
057f6c01 | 2450 | { |
3c95f0dc AV |
2451 | this->name = name; |
2452 | this->len = len; | |
2453 | this->hash = full_name_hash(base, name, len); | |
6a96ba54 | 2454 | if (!len) |
3c95f0dc | 2455 | return -EACCES; |
6a96ba54 | 2456 | |
21d8a15a AV |
2457 | if (unlikely(name[0] == '.')) { |
2458 | if (len < 2 || (len == 2 && name[1] == '.')) | |
3c95f0dc | 2459 | return -EACCES; |
21d8a15a AV |
2460 | } |
2461 | ||
6a96ba54 | 2462 | while (len--) { |
3c95f0dc | 2463 | unsigned int c = *(const unsigned char *)name++; |
6a96ba54 | 2464 | if (c == '/' || c == '\0') |
3c95f0dc | 2465 | return -EACCES; |
6a96ba54 | 2466 | } |
5a202bcd AV |
2467 | /* |
2468 | * See if the low-level filesystem might want | |
2469 | * to use its own hash.. | |
2470 | */ | |
2471 | if (base->d_flags & DCACHE_OP_HASH) { | |
3c95f0dc | 2472 | int err = base->d_op->d_hash(base, this); |
5a202bcd | 2473 | if (err < 0) |
3c95f0dc | 2474 | return err; |
5a202bcd | 2475 | } |
eead1911 | 2476 | |
3c95f0dc AV |
2477 | return inode_permission(base->d_inode, MAY_EXEC); |
2478 | } | |
2479 | ||
0da0b7fd DH |
2480 | /** |
2481 | * try_lookup_one_len - filesystem helper to lookup single pathname component | |
2482 | * @name: pathname component to lookup | |
2483 | * @base: base directory to lookup from | |
2484 | * @len: maximum length @len should be interpreted to | |
2485 | * | |
2486 | * Look up a dentry by name in the dcache, returning NULL if it does not | |
2487 | * currently exist. The function does not try to create a dentry. | |
2488 | * | |
2489 | * Note that this routine is purely a helper for filesystem usage and should | |
2490 | * not be called by generic code. | |
2491 | * | |
2492 | * The caller must hold base->i_mutex. | |
2493 | */ | |
2494 | struct dentry *try_lookup_one_len(const char *name, struct dentry *base, int len) | |
2495 | { | |
2496 | struct qstr this; | |
2497 | int err; | |
2498 | ||
2499 | WARN_ON_ONCE(!inode_is_locked(base->d_inode)); | |
2500 | ||
2501 | err = lookup_one_len_common(name, base, len, &this); | |
2502 | if (err) | |
2503 | return ERR_PTR(err); | |
2504 | ||
2505 | return lookup_dcache(&this, base, 0); | |
2506 | } | |
2507 | EXPORT_SYMBOL(try_lookup_one_len); | |
2508 | ||
3c95f0dc AV |
2509 | /** |
2510 | * lookup_one_len - filesystem helper to lookup single pathname component | |
2511 | * @name: pathname component to lookup | |
2512 | * @base: base directory to lookup from | |
2513 | * @len: maximum length @len should be interpreted to | |
2514 | * | |
2515 | * Note that this routine is purely a helper for filesystem usage and should | |
2516 | * not be called by generic code. | |
2517 | * | |
2518 | * The caller must hold base->i_mutex. | |
2519 | */ | |
2520 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) | |
2521 | { | |
8613a209 | 2522 | struct dentry *dentry; |
3c95f0dc AV |
2523 | struct qstr this; |
2524 | int err; | |
2525 | ||
2526 | WARN_ON_ONCE(!inode_is_locked(base->d_inode)); | |
2527 | ||
2528 | err = lookup_one_len_common(name, base, len, &this); | |
cda309de MS |
2529 | if (err) |
2530 | return ERR_PTR(err); | |
2531 | ||
8613a209 AV |
2532 | dentry = lookup_dcache(&this, base, 0); |
2533 | return dentry ? dentry : __lookup_slow(&this, base, 0); | |
057f6c01 | 2534 | } |
4d359507 | 2535 | EXPORT_SYMBOL(lookup_one_len); |
057f6c01 | 2536 | |
bbddca8e N |
2537 | /** |
2538 | * lookup_one_len_unlocked - filesystem helper to lookup single pathname component | |
2539 | * @name: pathname component to lookup | |
2540 | * @base: base directory to lookup from | |
2541 | * @len: maximum length @len should be interpreted to | |
2542 | * | |
2543 | * Note that this routine is purely a helper for filesystem usage and should | |
2544 | * not be called by generic code. | |
2545 | * | |
2546 | * Unlike lookup_one_len, it should be called without the parent | |
2547 | * i_mutex held, and will take the i_mutex itself if necessary. | |
2548 | */ | |
2549 | struct dentry *lookup_one_len_unlocked(const char *name, | |
2550 | struct dentry *base, int len) | |
2551 | { | |
2552 | struct qstr this; | |
bbddca8e | 2553 | int err; |
20d00ee8 | 2554 | struct dentry *ret; |
bbddca8e | 2555 | |
3c95f0dc | 2556 | err = lookup_one_len_common(name, base, len, &this); |
bbddca8e N |
2557 | if (err) |
2558 | return ERR_PTR(err); | |
2559 | ||
20d00ee8 LT |
2560 | ret = lookup_dcache(&this, base, 0); |
2561 | if (!ret) | |
2562 | ret = lookup_slow(&this, base, 0); | |
2563 | return ret; | |
bbddca8e N |
2564 | } |
2565 | EXPORT_SYMBOL(lookup_one_len_unlocked); | |
2566 | ||
eedf265a EB |
2567 | #ifdef CONFIG_UNIX98_PTYS |
2568 | int path_pts(struct path *path) | |
2569 | { | |
2570 | /* Find something mounted on "pts" in the same directory as | |
2571 | * the input path. | |
2572 | */ | |
2573 | struct dentry *child, *parent; | |
2574 | struct qstr this; | |
2575 | int ret; | |
2576 | ||
2577 | ret = path_parent_directory(path); | |
2578 | if (ret) | |
2579 | return ret; | |
2580 | ||
2581 | parent = path->dentry; | |
2582 | this.name = "pts"; | |
2583 | this.len = 3; | |
2584 | child = d_hash_and_lookup(parent, &this); | |
2585 | if (!child) | |
2586 | return -ENOENT; | |
2587 | ||
2588 | path->dentry = child; | |
2589 | dput(parent); | |
2590 | follow_mount(path); | |
2591 | return 0; | |
2592 | } | |
2593 | #endif | |
2594 | ||
1fa1e7f6 AW |
2595 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2596 | struct path *path, int *empty) | |
1da177e4 | 2597 | { |
abc9f5be AV |
2598 | return filename_lookup(dfd, getname_flags(name, flags, empty), |
2599 | flags, path, NULL); | |
1da177e4 | 2600 | } |
b853a161 | 2601 | EXPORT_SYMBOL(user_path_at_empty); |
1fa1e7f6 | 2602 | |
8033426e | 2603 | /** |
197df04c | 2604 | * mountpoint_last - look up last component for umount |
8033426e | 2605 | * @nd: pathwalk nameidata - currently pointing at parent directory of "last" |
8033426e JL |
2606 | * |
2607 | * This is a special lookup_last function just for umount. In this case, we | |
2608 | * need to resolve the path without doing any revalidation. | |
2609 | * | |
2610 | * The nameidata should be the result of doing a LOOKUP_PARENT pathwalk. Since | |
2611 | * mountpoints are always pinned in the dcache, their ancestors are too. Thus, | |
2612 | * in almost all cases, this lookup will be served out of the dcache. The only | |
2613 | * cases where it won't are if nd->last refers to a symlink or the path is | |
2614 | * bogus and it doesn't exist. | |
2615 | * | |
2616 | * Returns: | |
2617 | * -error: if there was an error during lookup. This includes -ENOENT if the | |
ba8f4613 | 2618 | * lookup found a negative dentry. |
8033426e | 2619 | * |
ba8f4613 AV |
2620 | * 0: if we successfully resolved nd->last and found it to not to be a |
2621 | * symlink that needs to be followed. | |
8033426e JL |
2622 | * |
2623 | * 1: if we successfully resolved nd->last and found it to be a symlink | |
ba8f4613 | 2624 | * that needs to be followed. |
8033426e JL |
2625 | */ |
2626 | static int | |
ba8f4613 | 2627 | mountpoint_last(struct nameidata *nd) |
8033426e JL |
2628 | { |
2629 | int error = 0; | |
8033426e | 2630 | struct dentry *dir = nd->path.dentry; |
ba8f4613 | 2631 | struct path path; |
8033426e | 2632 | |
35759521 AV |
2633 | /* If we're in rcuwalk, drop out of it to handle last component */ |
2634 | if (nd->flags & LOOKUP_RCU) { | |
4675ac39 | 2635 | if (unlazy_walk(nd)) |
deb106c6 | 2636 | return -ECHILD; |
8033426e JL |
2637 | } |
2638 | ||
2639 | nd->flags &= ~LOOKUP_PARENT; | |
2640 | ||
2641 | if (unlikely(nd->last_type != LAST_NORM)) { | |
2642 | error = handle_dots(nd, nd->last_type); | |
35759521 | 2643 | if (error) |
deb106c6 | 2644 | return error; |
ba8f4613 | 2645 | path.dentry = dget(nd->path.dentry); |
949a852e | 2646 | } else { |
ba8f4613 AV |
2647 | path.dentry = d_lookup(dir, &nd->last); |
2648 | if (!path.dentry) { | |
949a852e AV |
2649 | /* |
2650 | * No cached dentry. Mounted dentries are pinned in the | |
2651 | * cache, so that means that this dentry is probably | |
2652 | * a symlink or the path doesn't actually point | |
2653 | * to a mounted dentry. | |
2654 | */ | |
ba8f4613 | 2655 | path.dentry = lookup_slow(&nd->last, dir, |
949a852e | 2656 | nd->flags | LOOKUP_NO_REVAL); |
ba8f4613 AV |
2657 | if (IS_ERR(path.dentry)) |
2658 | return PTR_ERR(path.dentry); | |
bcceeeba | 2659 | } |
8033426e | 2660 | } |
ba8f4613 AV |
2661 | if (d_is_negative(path.dentry)) { |
2662 | dput(path.dentry); | |
deb106c6 | 2663 | return -ENOENT; |
8033426e | 2664 | } |
ba8f4613 | 2665 | path.mnt = nd->path.mnt; |
8f64fb1c | 2666 | return step_into(nd, &path, 0, d_backing_inode(path.dentry), 0); |
8033426e JL |
2667 | } |
2668 | ||
2669 | /** | |
197df04c | 2670 | * path_mountpoint - look up a path to be umounted |
2a78b857 | 2671 | * @nd: lookup context |
8033426e | 2672 | * @flags: lookup flags |
c8a53ee5 | 2673 | * @path: pointer to container for result |
8033426e JL |
2674 | * |
2675 | * Look up the given name, but don't attempt to revalidate the last component. | |
606d6fe3 | 2676 | * Returns 0 and "path" will be valid on success; Returns error otherwise. |
8033426e JL |
2677 | */ |
2678 | static int | |
c8a53ee5 | 2679 | path_mountpoint(struct nameidata *nd, unsigned flags, struct path *path) |
8033426e | 2680 | { |
c8a53ee5 | 2681 | const char *s = path_init(nd, flags); |
368ee9ba | 2682 | int err; |
9b5858e9 | 2683 | |
3bdba28b | 2684 | while (!(err = link_path_walk(s, nd)) && |
ba8f4613 | 2685 | (err = mountpoint_last(nd)) > 0) { |
3bdba28b | 2686 | s = trailing_symlink(nd); |
8033426e | 2687 | } |
ba8f4613 AV |
2688 | if (!err) { |
2689 | *path = nd->path; | |
2690 | nd->path.mnt = NULL; | |
2691 | nd->path.dentry = NULL; | |
2692 | follow_mount(path); | |
2693 | } | |
deb106c6 | 2694 | terminate_walk(nd); |
8033426e JL |
2695 | return err; |
2696 | } | |
2697 | ||
2d864651 | 2698 | static int |
668696dc | 2699 | filename_mountpoint(int dfd, struct filename *name, struct path *path, |
2d864651 AV |
2700 | unsigned int flags) |
2701 | { | |
9883d185 | 2702 | struct nameidata nd; |
cbaab2db | 2703 | int error; |
668696dc AV |
2704 | if (IS_ERR(name)) |
2705 | return PTR_ERR(name); | |
9883d185 | 2706 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2707 | error = path_mountpoint(&nd, flags | LOOKUP_RCU, path); |
2d864651 | 2708 | if (unlikely(error == -ECHILD)) |
c8a53ee5 | 2709 | error = path_mountpoint(&nd, flags, path); |
2d864651 | 2710 | if (unlikely(error == -ESTALE)) |
c8a53ee5 | 2711 | error = path_mountpoint(&nd, flags | LOOKUP_REVAL, path); |
2d864651 | 2712 | if (likely(!error)) |
c9b07eab | 2713 | audit_inode(name, path->dentry, AUDIT_INODE_NOEVAL); |
9883d185 | 2714 | restore_nameidata(); |
668696dc | 2715 | putname(name); |
2d864651 AV |
2716 | return error; |
2717 | } | |
2718 | ||
8033426e | 2719 | /** |
197df04c | 2720 | * user_path_mountpoint_at - lookup a path from userland in order to umount it |
8033426e JL |
2721 | * @dfd: directory file descriptor |
2722 | * @name: pathname from userland | |
2723 | * @flags: lookup flags | |
2724 | * @path: pointer to container to hold result | |
2725 | * | |
2726 | * A umount is a special case for path walking. We're not actually interested | |
2727 | * in the inode in this situation, and ESTALE errors can be a problem. We | |
2728 | * simply want track down the dentry and vfsmount attached at the mountpoint | |
2729 | * and avoid revalidating the last component. | |
2730 | * | |
2731 | * Returns 0 and populates "path" on success. | |
2732 | */ | |
2733 | int | |
197df04c | 2734 | user_path_mountpoint_at(int dfd, const char __user *name, unsigned int flags, |
8033426e JL |
2735 | struct path *path) |
2736 | { | |
cbaab2db | 2737 | return filename_mountpoint(dfd, getname(name), path, flags); |
8033426e JL |
2738 | } |
2739 | ||
2d864651 AV |
2740 | int |
2741 | kern_path_mountpoint(int dfd, const char *name, struct path *path, | |
2742 | unsigned int flags) | |
2743 | { | |
cbaab2db | 2744 | return filename_mountpoint(dfd, getname_kernel(name), path, flags); |
2d864651 AV |
2745 | } |
2746 | EXPORT_SYMBOL(kern_path_mountpoint); | |
2747 | ||
cbdf35bc | 2748 | int __check_sticky(struct inode *dir, struct inode *inode) |
1da177e4 | 2749 | { |
8e96e3b7 | 2750 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2751 | |
8e96e3b7 | 2752 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2753 | return 0; |
8e96e3b7 | 2754 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2755 | return 0; |
23adbe12 | 2756 | return !capable_wrt_inode_uidgid(inode, CAP_FOWNER); |
1da177e4 | 2757 | } |
cbdf35bc | 2758 | EXPORT_SYMBOL(__check_sticky); |
1da177e4 LT |
2759 | |
2760 | /* | |
2761 | * Check whether we can remove a link victim from directory dir, check | |
2762 | * whether the type of victim is right. | |
2763 | * 1. We can't do it if dir is read-only (done in permission()) | |
2764 | * 2. We should have write and exec permissions on dir | |
2765 | * 3. We can't remove anything from append-only dir | |
2766 | * 4. We can't do anything with immutable dir (done in permission()) | |
2767 | * 5. If the sticky bit on dir is set we should either | |
2768 | * a. be owner of dir, or | |
2769 | * b. be owner of victim, or | |
2770 | * c. have CAP_FOWNER capability | |
2771 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2772 | * links pointing to it. | |
0bd23d09 EB |
2773 | * 7. If the victim has an unknown uid or gid we can't change the inode. |
2774 | * 8. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2775 | * 9. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2776 | * 10. We can't remove a root or mountpoint. | |
2777 | * 11. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1da177e4 LT |
2778 | * nfs_async_unlink(). |
2779 | */ | |
b18825a7 | 2780 | static int may_delete(struct inode *dir, struct dentry *victim, bool isdir) |
1da177e4 | 2781 | { |
63afdfc7 | 2782 | struct inode *inode = d_backing_inode(victim); |
1da177e4 LT |
2783 | int error; |
2784 | ||
b18825a7 | 2785 | if (d_is_negative(victim)) |
1da177e4 | 2786 | return -ENOENT; |
b18825a7 | 2787 | BUG_ON(!inode); |
1da177e4 LT |
2788 | |
2789 | BUG_ON(victim->d_parent->d_inode != dir); | |
593d1ce8 EB |
2790 | |
2791 | /* Inode writeback is not safe when the uid or gid are invalid. */ | |
2792 | if (!uid_valid(inode->i_uid) || !gid_valid(inode->i_gid)) | |
2793 | return -EOVERFLOW; | |
2794 | ||
4fa6b5ec | 2795 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
1da177e4 | 2796 | |
f419a2e3 | 2797 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2798 | if (error) |
2799 | return error; | |
2800 | if (IS_APPEND(dir)) | |
2801 | return -EPERM; | |
b18825a7 DH |
2802 | |
2803 | if (check_sticky(dir, inode) || IS_APPEND(inode) || | |
0bd23d09 | 2804 | IS_IMMUTABLE(inode) || IS_SWAPFILE(inode) || HAS_UNMAPPED_ID(inode)) |
1da177e4 LT |
2805 | return -EPERM; |
2806 | if (isdir) { | |
44b1d530 | 2807 | if (!d_is_dir(victim)) |
1da177e4 LT |
2808 | return -ENOTDIR; |
2809 | if (IS_ROOT(victim)) | |
2810 | return -EBUSY; | |
44b1d530 | 2811 | } else if (d_is_dir(victim)) |
1da177e4 LT |
2812 | return -EISDIR; |
2813 | if (IS_DEADDIR(dir)) | |
2814 | return -ENOENT; | |
2815 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2816 | return -EBUSY; | |
2817 | return 0; | |
2818 | } | |
2819 | ||
2820 | /* Check whether we can create an object with dentry child in directory | |
2821 | * dir. | |
2822 | * 1. We can't do it if child already exists (open has special treatment for | |
2823 | * this case, but since we are inlined it's OK) | |
2824 | * 2. We can't do it if dir is read-only (done in permission()) | |
036d5236 EB |
2825 | * 3. We can't do it if the fs can't represent the fsuid or fsgid. |
2826 | * 4. We should have write and exec permissions on dir | |
2827 | * 5. We can't do it if dir is immutable (done in permission()) | |
1da177e4 | 2828 | */ |
a95164d9 | 2829 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 | 2830 | { |
036d5236 | 2831 | struct user_namespace *s_user_ns; |
14e972b4 | 2832 | audit_inode_child(dir, child, AUDIT_TYPE_CHILD_CREATE); |
1da177e4 LT |
2833 | if (child->d_inode) |
2834 | return -EEXIST; | |
2835 | if (IS_DEADDIR(dir)) | |
2836 | return -ENOENT; | |
036d5236 EB |
2837 | s_user_ns = dir->i_sb->s_user_ns; |
2838 | if (!kuid_has_mapping(s_user_ns, current_fsuid()) || | |
2839 | !kgid_has_mapping(s_user_ns, current_fsgid())) | |
2840 | return -EOVERFLOW; | |
f419a2e3 | 2841 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2842 | } |
2843 | ||
1da177e4 LT |
2844 | /* |
2845 | * p1 and p2 should be directories on the same fs. | |
2846 | */ | |
2847 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2848 | { | |
2849 | struct dentry *p; | |
2850 | ||
2851 | if (p1 == p2) { | |
5955102c | 2852 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
1da177e4 LT |
2853 | return NULL; |
2854 | } | |
2855 | ||
fc64005c | 2856 | mutex_lock(&p1->d_sb->s_vfs_rename_mutex); |
1da177e4 | 2857 | |
e2761a11 OH |
2858 | p = d_ancestor(p2, p1); |
2859 | if (p) { | |
5955102c AV |
2860 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT); |
2861 | inode_lock_nested(p1->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2862 | return p; |
1da177e4 LT |
2863 | } |
2864 | ||
e2761a11 OH |
2865 | p = d_ancestor(p1, p2); |
2866 | if (p) { | |
5955102c AV |
2867 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2868 | inode_lock_nested(p2->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2869 | return p; |
1da177e4 LT |
2870 | } |
2871 | ||
5955102c AV |
2872 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2873 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT2); | |
1da177e4 LT |
2874 | return NULL; |
2875 | } | |
4d359507 | 2876 | EXPORT_SYMBOL(lock_rename); |
1da177e4 LT |
2877 | |
2878 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2879 | { | |
5955102c | 2880 | inode_unlock(p1->d_inode); |
1da177e4 | 2881 | if (p1 != p2) { |
5955102c | 2882 | inode_unlock(p2->d_inode); |
fc64005c | 2883 | mutex_unlock(&p1->d_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2884 | } |
2885 | } | |
4d359507 | 2886 | EXPORT_SYMBOL(unlock_rename); |
1da177e4 | 2887 | |
4acdaf27 | 2888 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2889 | bool want_excl) |
1da177e4 | 2890 | { |
a95164d9 | 2891 | int error = may_create(dir, dentry); |
1da177e4 LT |
2892 | if (error) |
2893 | return error; | |
2894 | ||
acfa4380 | 2895 | if (!dir->i_op->create) |
1da177e4 LT |
2896 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2897 | mode &= S_IALLUGO; | |
2898 | mode |= S_IFREG; | |
2899 | error = security_inode_create(dir, dentry, mode); | |
2900 | if (error) | |
2901 | return error; | |
312b63fb | 2902 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2903 | if (!error) |
f38aa942 | 2904 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2905 | return error; |
2906 | } | |
4d359507 | 2907 | EXPORT_SYMBOL(vfs_create); |
1da177e4 | 2908 | |
8e6c848e AV |
2909 | int vfs_mkobj(struct dentry *dentry, umode_t mode, |
2910 | int (*f)(struct dentry *, umode_t, void *), | |
2911 | void *arg) | |
2912 | { | |
2913 | struct inode *dir = dentry->d_parent->d_inode; | |
2914 | int error = may_create(dir, dentry); | |
2915 | if (error) | |
2916 | return error; | |
2917 | ||
2918 | mode &= S_IALLUGO; | |
2919 | mode |= S_IFREG; | |
2920 | error = security_inode_create(dir, dentry, mode); | |
2921 | if (error) | |
2922 | return error; | |
2923 | error = f(dentry, mode, arg); | |
2924 | if (!error) | |
2925 | fsnotify_create(dir, dentry); | |
2926 | return error; | |
2927 | } | |
2928 | EXPORT_SYMBOL(vfs_mkobj); | |
2929 | ||
a2982cc9 EB |
2930 | bool may_open_dev(const struct path *path) |
2931 | { | |
2932 | return !(path->mnt->mnt_flags & MNT_NODEV) && | |
2933 | !(path->mnt->mnt_sb->s_iflags & SB_I_NODEV); | |
2934 | } | |
2935 | ||
f0bb5aaf | 2936 | static int may_open(const struct path *path, int acc_mode, int flag) |
1da177e4 | 2937 | { |
3fb64190 | 2938 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2939 | struct inode *inode = dentry->d_inode; |
2940 | int error; | |
2941 | ||
2942 | if (!inode) | |
2943 | return -ENOENT; | |
2944 | ||
c8fe8f30 CH |
2945 | switch (inode->i_mode & S_IFMT) { |
2946 | case S_IFLNK: | |
1da177e4 | 2947 | return -ELOOP; |
c8fe8f30 CH |
2948 | case S_IFDIR: |
2949 | if (acc_mode & MAY_WRITE) | |
2950 | return -EISDIR; | |
2951 | break; | |
2952 | case S_IFBLK: | |
2953 | case S_IFCHR: | |
a2982cc9 | 2954 | if (!may_open_dev(path)) |
1da177e4 | 2955 | return -EACCES; |
c8fe8f30 CH |
2956 | /*FALLTHRU*/ |
2957 | case S_IFIFO: | |
2958 | case S_IFSOCK: | |
1da177e4 | 2959 | flag &= ~O_TRUNC; |
c8fe8f30 | 2960 | break; |
4a3fd211 | 2961 | } |
b41572e9 | 2962 | |
62fb4a15 | 2963 | error = inode_permission(inode, MAY_OPEN | acc_mode); |
b41572e9 DH |
2964 | if (error) |
2965 | return error; | |
6146f0d5 | 2966 | |
1da177e4 LT |
2967 | /* |
2968 | * An append-only file must be opened in append mode for writing. | |
2969 | */ | |
2970 | if (IS_APPEND(inode)) { | |
8737c930 | 2971 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2972 | return -EPERM; |
1da177e4 | 2973 | if (flag & O_TRUNC) |
7715b521 | 2974 | return -EPERM; |
1da177e4 LT |
2975 | } |
2976 | ||
2977 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2978 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2979 | return -EPERM; |
1da177e4 | 2980 | |
f3c7691e | 2981 | return 0; |
7715b521 | 2982 | } |
1da177e4 | 2983 | |
e1181ee6 | 2984 | static int handle_truncate(struct file *filp) |
7715b521 | 2985 | { |
f0bb5aaf | 2986 | const struct path *path = &filp->f_path; |
7715b521 AV |
2987 | struct inode *inode = path->dentry->d_inode; |
2988 | int error = get_write_access(inode); | |
2989 | if (error) | |
2990 | return error; | |
2991 | /* | |
2992 | * Refuse to truncate files with mandatory locks held on them. | |
2993 | */ | |
d7a06983 | 2994 | error = locks_verify_locked(filp); |
7715b521 | 2995 | if (!error) |
ea0d3ab2 | 2996 | error = security_path_truncate(path); |
7715b521 AV |
2997 | if (!error) { |
2998 | error = do_truncate(path->dentry, 0, | |
2999 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 3000 | filp); |
7715b521 AV |
3001 | } |
3002 | put_write_access(inode); | |
acd0c935 | 3003 | return error; |
1da177e4 LT |
3004 | } |
3005 | ||
d57999e1 DH |
3006 | static inline int open_to_namei_flags(int flag) |
3007 | { | |
8a5e929d AV |
3008 | if ((flag & O_ACCMODE) == 3) |
3009 | flag--; | |
d57999e1 DH |
3010 | return flag; |
3011 | } | |
3012 | ||
d3607752 | 3013 | static int may_o_create(const struct path *dir, struct dentry *dentry, umode_t mode) |
d18e9008 | 3014 | { |
1328c727 | 3015 | struct user_namespace *s_user_ns; |
d18e9008 MS |
3016 | int error = security_path_mknod(dir, dentry, mode, 0); |
3017 | if (error) | |
3018 | return error; | |
3019 | ||
1328c727 SF |
3020 | s_user_ns = dir->dentry->d_sb->s_user_ns; |
3021 | if (!kuid_has_mapping(s_user_ns, current_fsuid()) || | |
3022 | !kgid_has_mapping(s_user_ns, current_fsgid())) | |
3023 | return -EOVERFLOW; | |
3024 | ||
d18e9008 MS |
3025 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); |
3026 | if (error) | |
3027 | return error; | |
3028 | ||
3029 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
3030 | } | |
3031 | ||
1acf0af9 DH |
3032 | /* |
3033 | * Attempt to atomically look up, create and open a file from a negative | |
3034 | * dentry. | |
3035 | * | |
3036 | * Returns 0 if successful. The file will have been created and attached to | |
3037 | * @file by the filesystem calling finish_open(). | |
3038 | * | |
00a07c15 AV |
3039 | * If the file was looked up only or didn't need creating, FMODE_OPENED won't |
3040 | * be set. The caller will need to perform the open themselves. @path will | |
3041 | * have been updated to point to the new dentry. This may be negative. | |
1acf0af9 DH |
3042 | * |
3043 | * Returns an error code otherwise. | |
3044 | */ | |
2675a4eb AV |
3045 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
3046 | struct path *path, struct file *file, | |
3047 | const struct open_flags *op, | |
3ec2eef1 | 3048 | int open_flag, umode_t mode) |
d18e9008 | 3049 | { |
384f26e2 | 3050 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; |
d18e9008 | 3051 | struct inode *dir = nd->path.dentry->d_inode; |
d18e9008 | 3052 | int error; |
d18e9008 | 3053 | |
384f26e2 | 3054 | if (!(~open_flag & (O_EXCL | O_CREAT))) /* both O_EXCL and O_CREAT */ |
d18e9008 | 3055 | open_flag &= ~O_TRUNC; |
d18e9008 | 3056 | |
d18e9008 MS |
3057 | if (nd->flags & LOOKUP_DIRECTORY) |
3058 | open_flag |= O_DIRECTORY; | |
3059 | ||
30d90494 AV |
3060 | file->f_path.dentry = DENTRY_NOT_SET; |
3061 | file->f_path.mnt = nd->path.mnt; | |
0fb1ea09 | 3062 | error = dir->i_op->atomic_open(dir, dentry, file, |
44907d79 | 3063 | open_to_namei_flags(open_flag), mode); |
6fbd0714 | 3064 | d_lookup_done(dentry); |
384f26e2 | 3065 | if (!error) { |
64e1ac4d AV |
3066 | if (file->f_mode & FMODE_OPENED) { |
3067 | /* | |
3068 | * We didn't have the inode before the open, so check open | |
3069 | * permission here. | |
3070 | */ | |
3071 | int acc_mode = op->acc_mode; | |
3072 | if (file->f_mode & FMODE_CREATED) { | |
3073 | WARN_ON(!(open_flag & O_CREAT)); | |
3074 | fsnotify_create(dir, dentry); | |
3075 | acc_mode = 0; | |
3076 | } | |
3077 | error = may_open(&file->f_path, acc_mode, open_flag); | |
3078 | if (WARN_ON(error > 0)) | |
3079 | error = -EINVAL; | |
3080 | } else if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { | |
2675a4eb | 3081 | error = -EIO; |
03da633a | 3082 | } else { |
384f26e2 AV |
3083 | if (file->f_path.dentry) { |
3084 | dput(dentry); | |
3085 | dentry = file->f_path.dentry; | |
03da633a | 3086 | } |
73a09dd9 | 3087 | if (file->f_mode & FMODE_CREATED) |
384f26e2 | 3088 | fsnotify_create(dir, dentry); |
a01e718f AV |
3089 | if (unlikely(d_is_negative(dentry))) { |
3090 | error = -ENOENT; | |
3091 | } else { | |
3092 | path->dentry = dentry; | |
3093 | path->mnt = nd->path.mnt; | |
00a07c15 | 3094 | return 0; |
a01e718f | 3095 | } |
62b2ce96 | 3096 | } |
d18e9008 | 3097 | } |
d18e9008 | 3098 | dput(dentry); |
2675a4eb | 3099 | return error; |
d18e9008 MS |
3100 | } |
3101 | ||
d58ffd35 | 3102 | /* |
1acf0af9 | 3103 | * Look up and maybe create and open the last component. |
d58ffd35 | 3104 | * |
00a07c15 | 3105 | * Must be called with parent locked (exclusive in O_CREAT case). |
1acf0af9 | 3106 | * |
00a07c15 AV |
3107 | * Returns 0 on success, that is, if |
3108 | * the file was successfully atomically created (if necessary) and opened, or | |
3109 | * the file was not completely opened at this time, though lookups and | |
3110 | * creations were performed. | |
3111 | * These case are distinguished by presence of FMODE_OPENED on file->f_mode. | |
3112 | * In the latter case dentry returned in @path might be negative if O_CREAT | |
3113 | * hadn't been specified. | |
1acf0af9 | 3114 | * |
00a07c15 | 3115 | * An error code is returned on failure. |
d58ffd35 | 3116 | */ |
2675a4eb AV |
3117 | static int lookup_open(struct nameidata *nd, struct path *path, |
3118 | struct file *file, | |
3119 | const struct open_flags *op, | |
3ec2eef1 | 3120 | bool got_write) |
d58ffd35 MS |
3121 | { |
3122 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 3123 | struct inode *dir_inode = dir->d_inode; |
1643b43f | 3124 | int open_flag = op->open_flag; |
d58ffd35 | 3125 | struct dentry *dentry; |
1643b43f | 3126 | int error, create_error = 0; |
1643b43f | 3127 | umode_t mode = op->mode; |
6fbd0714 | 3128 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
d58ffd35 | 3129 | |
ce8644fc AV |
3130 | if (unlikely(IS_DEADDIR(dir_inode))) |
3131 | return -ENOENT; | |
d58ffd35 | 3132 | |
73a09dd9 | 3133 | file->f_mode &= ~FMODE_CREATED; |
6fbd0714 AV |
3134 | dentry = d_lookup(dir, &nd->last); |
3135 | for (;;) { | |
3136 | if (!dentry) { | |
3137 | dentry = d_alloc_parallel(dir, &nd->last, &wq); | |
3138 | if (IS_ERR(dentry)) | |
3139 | return PTR_ERR(dentry); | |
3140 | } | |
3141 | if (d_in_lookup(dentry)) | |
3142 | break; | |
d58ffd35 | 3143 | |
6fbd0714 AV |
3144 | error = d_revalidate(dentry, nd->flags); |
3145 | if (likely(error > 0)) | |
3146 | break; | |
3147 | if (error) | |
3148 | goto out_dput; | |
3149 | d_invalidate(dentry); | |
3150 | dput(dentry); | |
3151 | dentry = NULL; | |
3152 | } | |
3153 | if (dentry->d_inode) { | |
6c51e513 | 3154 | /* Cached positive dentry: will open in f_op->open */ |
d18e9008 | 3155 | goto out_no_open; |
6c51e513 | 3156 | } |
d18e9008 | 3157 | |
1643b43f AV |
3158 | /* |
3159 | * Checking write permission is tricky, bacuse we don't know if we are | |
3160 | * going to actually need it: O_CREAT opens should work as long as the | |
3161 | * file exists. But checking existence breaks atomicity. The trick is | |
3162 | * to check access and if not granted clear O_CREAT from the flags. | |
3163 | * | |
3164 | * Another problem is returing the "right" error value (e.g. for an | |
3165 | * O_EXCL open we want to return EEXIST not EROFS). | |
3166 | */ | |
3167 | if (open_flag & O_CREAT) { | |
3168 | if (!IS_POSIXACL(dir->d_inode)) | |
3169 | mode &= ~current_umask(); | |
3170 | if (unlikely(!got_write)) { | |
3171 | create_error = -EROFS; | |
3172 | open_flag &= ~O_CREAT; | |
3173 | if (open_flag & (O_EXCL | O_TRUNC)) | |
3174 | goto no_open; | |
3175 | /* No side effects, safe to clear O_CREAT */ | |
3176 | } else { | |
3177 | create_error = may_o_create(&nd->path, dentry, mode); | |
3178 | if (create_error) { | |
3179 | open_flag &= ~O_CREAT; | |
3180 | if (open_flag & O_EXCL) | |
3181 | goto no_open; | |
3182 | } | |
3183 | } | |
3184 | } else if ((open_flag & (O_TRUNC|O_WRONLY|O_RDWR)) && | |
3185 | unlikely(!got_write)) { | |
3186 | /* | |
3187 | * No O_CREATE -> atomicity not a requirement -> fall | |
3188 | * back to lookup + open | |
3189 | */ | |
3190 | goto no_open; | |
d18e9008 MS |
3191 | } |
3192 | ||
6ac08709 | 3193 | if (dir_inode->i_op->atomic_open) { |
1643b43f | 3194 | error = atomic_open(nd, dentry, path, file, op, open_flag, |
3ec2eef1 | 3195 | mode); |
1643b43f AV |
3196 | if (unlikely(error == -ENOENT) && create_error) |
3197 | error = create_error; | |
3198 | return error; | |
d18e9008 | 3199 | } |
54ef4872 | 3200 | |
1643b43f | 3201 | no_open: |
6fbd0714 | 3202 | if (d_in_lookup(dentry)) { |
12fa5e24 AV |
3203 | struct dentry *res = dir_inode->i_op->lookup(dir_inode, dentry, |
3204 | nd->flags); | |
6fbd0714 | 3205 | d_lookup_done(dentry); |
12fa5e24 AV |
3206 | if (unlikely(res)) { |
3207 | if (IS_ERR(res)) { | |
3208 | error = PTR_ERR(res); | |
3209 | goto out_dput; | |
3210 | } | |
3211 | dput(dentry); | |
3212 | dentry = res; | |
3213 | } | |
54ef4872 MS |
3214 | } |
3215 | ||
d58ffd35 | 3216 | /* Negative dentry, just create the file */ |
1643b43f | 3217 | if (!dentry->d_inode && (open_flag & O_CREAT)) { |
73a09dd9 | 3218 | file->f_mode |= FMODE_CREATED; |
ce8644fc | 3219 | audit_inode_child(dir_inode, dentry, AUDIT_TYPE_CHILD_CREATE); |
ce8644fc AV |
3220 | if (!dir_inode->i_op->create) { |
3221 | error = -EACCES; | |
d58ffd35 | 3222 | goto out_dput; |
ce8644fc AV |
3223 | } |
3224 | error = dir_inode->i_op->create(dir_inode, dentry, mode, | |
1643b43f | 3225 | open_flag & O_EXCL); |
d58ffd35 MS |
3226 | if (error) |
3227 | goto out_dput; | |
ce8644fc | 3228 | fsnotify_create(dir_inode, dentry); |
d58ffd35 | 3229 | } |
1643b43f AV |
3230 | if (unlikely(create_error) && !dentry->d_inode) { |
3231 | error = create_error; | |
3232 | goto out_dput; | |
d58ffd35 | 3233 | } |
d18e9008 | 3234 | out_no_open: |
d58ffd35 MS |
3235 | path->dentry = dentry; |
3236 | path->mnt = nd->path.mnt; | |
00a07c15 | 3237 | return 0; |
d58ffd35 MS |
3238 | |
3239 | out_dput: | |
3240 | dput(dentry); | |
2675a4eb | 3241 | return error; |
d58ffd35 MS |
3242 | } |
3243 | ||
31e6b01f | 3244 | /* |
fe2d35ff | 3245 | * Handle the last step of open() |
31e6b01f | 3246 | */ |
896475d5 | 3247 | static int do_last(struct nameidata *nd, |
3ec2eef1 | 3248 | struct file *file, const struct open_flags *op) |
fb1cc555 | 3249 | { |
a1e28038 | 3250 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 3251 | int open_flag = op->open_flag; |
77d660a8 | 3252 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 3253 | bool got_write = false; |
bcda7652 | 3254 | int acc_mode = op->acc_mode; |
254cf582 | 3255 | unsigned seq; |
a1eb3315 | 3256 | struct inode *inode; |
896475d5 | 3257 | struct path path; |
16c2cd71 | 3258 | int error; |
1f36f774 | 3259 | |
c3e380b0 AV |
3260 | nd->flags &= ~LOOKUP_PARENT; |
3261 | nd->flags |= op->intent; | |
3262 | ||
bc77daa7 | 3263 | if (nd->last_type != LAST_NORM) { |
fe2d35ff | 3264 | error = handle_dots(nd, nd->last_type); |
deb106c6 | 3265 | if (unlikely(error)) |
2675a4eb | 3266 | return error; |
e83db167 | 3267 | goto finish_open; |
1f36f774 | 3268 | } |
67ee3ad2 | 3269 | |
ca344a89 | 3270 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
3271 | if (nd->last.name[nd->last.len]) |
3272 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
3273 | /* we _can_ be in RCU mode here */ | |
254cf582 | 3274 | error = lookup_fast(nd, &path, &inode, &seq); |
e9742b53 | 3275 | if (likely(error > 0)) |
71574865 MS |
3276 | goto finish_lookup; |
3277 | ||
3278 | if (error < 0) | |
deb106c6 | 3279 | return error; |
71574865 MS |
3280 | |
3281 | BUG_ON(nd->inode != dir->d_inode); | |
6583fe22 | 3282 | BUG_ON(nd->flags & LOOKUP_RCU); |
b6183df7 MS |
3283 | } else { |
3284 | /* create side of things */ | |
3285 | /* | |
3286 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
3287 | * has been cleared when we got to the last component we are | |
3288 | * about to look up | |
3289 | */ | |
3290 | error = complete_walk(nd); | |
e8bb73df | 3291 | if (error) |
2675a4eb | 3292 | return error; |
fe2d35ff | 3293 | |
c9b07eab | 3294 | audit_inode(nd->name, dir, AUDIT_INODE_PARENT); |
b6183df7 | 3295 | /* trailing slashes? */ |
deb106c6 AV |
3296 | if (unlikely(nd->last.name[nd->last.len])) |
3297 | return -EISDIR; | |
b6183df7 | 3298 | } |
a2c36b45 | 3299 | |
9cf843e3 | 3300 | if (open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
64894cf8 AV |
3301 | error = mnt_want_write(nd->path.mnt); |
3302 | if (!error) | |
3303 | got_write = true; | |
3304 | /* | |
3305 | * do _not_ fail yet - we might not need that or fail with | |
3306 | * a different error; let lookup_open() decide; we'll be | |
3307 | * dropping this one anyway. | |
3308 | */ | |
3309 | } | |
9cf843e3 AV |
3310 | if (open_flag & O_CREAT) |
3311 | inode_lock(dir->d_inode); | |
3312 | else | |
3313 | inode_lock_shared(dir->d_inode); | |
3ec2eef1 | 3314 | error = lookup_open(nd, &path, file, op, got_write); |
9cf843e3 AV |
3315 | if (open_flag & O_CREAT) |
3316 | inode_unlock(dir->d_inode); | |
3317 | else | |
3318 | inode_unlock_shared(dir->d_inode); | |
a1e28038 | 3319 | |
00a07c15 AV |
3320 | if (error) |
3321 | goto out; | |
d18e9008 | 3322 | |
00a07c15 | 3323 | if (file->f_mode & FMODE_OPENED) { |
73a09dd9 | 3324 | if ((file->f_mode & FMODE_CREATED) || |
496ad9aa | 3325 | !S_ISREG(file_inode(file)->i_mode)) |
77d660a8 | 3326 | will_truncate = false; |
d18e9008 | 3327 | |
76ae2a5a | 3328 | audit_inode(nd->name, file->f_path.dentry, 0); |
d18e9008 MS |
3329 | goto opened; |
3330 | } | |
fb1cc555 | 3331 | |
73a09dd9 | 3332 | if (file->f_mode & FMODE_CREATED) { |
9b44f1b3 | 3333 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 3334 | open_flag &= ~O_TRUNC; |
77d660a8 | 3335 | will_truncate = false; |
62fb4a15 | 3336 | acc_mode = 0; |
896475d5 | 3337 | path_to_nameidata(&path, nd); |
e83db167 | 3338 | goto finish_open_created; |
fb1cc555 AV |
3339 | } |
3340 | ||
d18e9008 MS |
3341 | /* |
3342 | * If atomic_open() acquired write access it is dropped now due to | |
3343 | * possible mount and symlink following (this might be optimized away if | |
3344 | * necessary...) | |
3345 | */ | |
64894cf8 | 3346 | if (got_write) { |
d18e9008 | 3347 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3348 | got_write = false; |
d18e9008 MS |
3349 | } |
3350 | ||
e6ec03a2 AV |
3351 | error = follow_managed(&path, nd); |
3352 | if (unlikely(error < 0)) | |
3353 | return error; | |
3354 | ||
6583fe22 AV |
3355 | if (unlikely(d_is_negative(path.dentry))) { |
3356 | path_to_nameidata(&path, nd); | |
3357 | return -ENOENT; | |
3358 | } | |
3359 | ||
3360 | /* | |
3361 | * create/update audit record if it already exists. | |
3362 | */ | |
3363 | audit_inode(nd->name, path.dentry, 0); | |
3364 | ||
deb106c6 AV |
3365 | if (unlikely((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT))) { |
3366 | path_to_nameidata(&path, nd); | |
3367 | return -EEXIST; | |
3368 | } | |
fb1cc555 | 3369 | |
254cf582 | 3370 | seq = 0; /* out of RCU mode, so the value doesn't matter */ |
d4565649 | 3371 | inode = d_backing_inode(path.dentry); |
766c4cbf | 3372 | finish_lookup: |
8f64fb1c | 3373 | error = step_into(nd, &path, 0, inode, seq); |
deb106c6 | 3374 | if (unlikely(error)) |
d63ff28f | 3375 | return error; |
bc77daa7 | 3376 | finish_open: |
8f64fb1c | 3377 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
a3fbbde7 | 3378 | error = complete_walk(nd); |
fac7d191 | 3379 | if (error) |
2675a4eb | 3380 | return error; |
76ae2a5a | 3381 | audit_inode(nd->name, nd->path.dentry, 0); |
30aba665 SM |
3382 | if (open_flag & O_CREAT) { |
3383 | error = -EISDIR; | |
3384 | if (d_is_dir(nd->path.dentry)) | |
3385 | goto out; | |
3386 | error = may_create_in_sticky(dir, | |
3387 | d_backing_inode(nd->path.dentry)); | |
3388 | if (unlikely(error)) | |
3389 | goto out; | |
3390 | } | |
af2f5542 | 3391 | error = -ENOTDIR; |
44b1d530 | 3392 | if ((nd->flags & LOOKUP_DIRECTORY) && !d_can_lookup(nd->path.dentry)) |
2675a4eb | 3393 | goto out; |
4bbcbd3b | 3394 | if (!d_is_reg(nd->path.dentry)) |
77d660a8 | 3395 | will_truncate = false; |
6c0d46c4 | 3396 | |
0f9d1a10 AV |
3397 | if (will_truncate) { |
3398 | error = mnt_want_write(nd->path.mnt); | |
3399 | if (error) | |
2675a4eb | 3400 | goto out; |
64894cf8 | 3401 | got_write = true; |
0f9d1a10 | 3402 | } |
e83db167 | 3403 | finish_open_created: |
6ac08709 AV |
3404 | error = may_open(&nd->path, acc_mode, open_flag); |
3405 | if (error) | |
3406 | goto out; | |
aad888f8 | 3407 | BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */ |
ae2bb293 | 3408 | error = vfs_open(&nd->path, file); |
fac7d191 | 3409 | if (error) |
015c3bbc | 3410 | goto out; |
a8277b9b | 3411 | opened: |
6035a27b | 3412 | error = ima_file_check(file, op->acc_mode); |
fe9ec829 | 3413 | if (!error && will_truncate) |
2675a4eb | 3414 | error = handle_truncate(file); |
ca344a89 | 3415 | out: |
c80567c8 AV |
3416 | if (unlikely(error > 0)) { |
3417 | WARN_ON(1); | |
3418 | error = -EINVAL; | |
3419 | } | |
64894cf8 | 3420 | if (got_write) |
0f9d1a10 | 3421 | mnt_drop_write(nd->path.mnt); |
2675a4eb | 3422 | return error; |
fb1cc555 AV |
3423 | } |
3424 | ||
af7bd4dc AG |
3425 | struct dentry *vfs_tmpfile(struct dentry *dentry, umode_t mode, int open_flag) |
3426 | { | |
af7bd4dc AG |
3427 | struct dentry *child = NULL; |
3428 | struct inode *dir = dentry->d_inode; | |
3429 | struct inode *inode; | |
3430 | int error; | |
3431 | ||
3432 | /* we want directory to be writable */ | |
3433 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
3434 | if (error) | |
3435 | goto out_err; | |
3436 | error = -EOPNOTSUPP; | |
3437 | if (!dir->i_op->tmpfile) | |
3438 | goto out_err; | |
3439 | error = -ENOMEM; | |
cdf01226 | 3440 | child = d_alloc(dentry, &slash_name); |
af7bd4dc AG |
3441 | if (unlikely(!child)) |
3442 | goto out_err; | |
3443 | error = dir->i_op->tmpfile(dir, child, mode); | |
3444 | if (error) | |
3445 | goto out_err; | |
3446 | error = -ENOENT; | |
3447 | inode = child->d_inode; | |
3448 | if (unlikely(!inode)) | |
3449 | goto out_err; | |
3450 | if (!(open_flag & O_EXCL)) { | |
3451 | spin_lock(&inode->i_lock); | |
3452 | inode->i_state |= I_LINKABLE; | |
3453 | spin_unlock(&inode->i_lock); | |
3454 | } | |
fdb2410f | 3455 | ima_post_create_tmpfile(inode); |
af7bd4dc AG |
3456 | return child; |
3457 | ||
3458 | out_err: | |
3459 | dput(child); | |
3460 | return ERR_PTR(error); | |
3461 | } | |
3462 | EXPORT_SYMBOL(vfs_tmpfile); | |
3463 | ||
c8a53ee5 | 3464 | static int do_tmpfile(struct nameidata *nd, unsigned flags, |
60545d0d | 3465 | const struct open_flags *op, |
3ec2eef1 | 3466 | struct file *file) |
60545d0d | 3467 | { |
625b6d10 | 3468 | struct dentry *child; |
625b6d10 | 3469 | struct path path; |
c8a53ee5 | 3470 | int error = path_lookupat(nd, flags | LOOKUP_DIRECTORY, &path); |
60545d0d AV |
3471 | if (unlikely(error)) |
3472 | return error; | |
625b6d10 | 3473 | error = mnt_want_write(path.mnt); |
60545d0d AV |
3474 | if (unlikely(error)) |
3475 | goto out; | |
af7bd4dc AG |
3476 | child = vfs_tmpfile(path.dentry, op->mode, op->open_flag); |
3477 | error = PTR_ERR(child); | |
684e73be | 3478 | if (IS_ERR(child)) |
60545d0d | 3479 | goto out2; |
625b6d10 AV |
3480 | dput(path.dentry); |
3481 | path.dentry = child; | |
c8a53ee5 | 3482 | audit_inode(nd->name, child, 0); |
69a91c23 | 3483 | /* Don't check for other permissions, the inode was just created */ |
62fb4a15 | 3484 | error = may_open(&path, 0, op->open_flag); |
60545d0d AV |
3485 | if (error) |
3486 | goto out2; | |
625b6d10 | 3487 | file->f_path.mnt = path.mnt; |
be12af3e | 3488 | error = finish_open(file, child, NULL); |
60545d0d | 3489 | out2: |
625b6d10 | 3490 | mnt_drop_write(path.mnt); |
60545d0d | 3491 | out: |
625b6d10 | 3492 | path_put(&path); |
60545d0d AV |
3493 | return error; |
3494 | } | |
3495 | ||
6ac08709 AV |
3496 | static int do_o_path(struct nameidata *nd, unsigned flags, struct file *file) |
3497 | { | |
3498 | struct path path; | |
3499 | int error = path_lookupat(nd, flags, &path); | |
3500 | if (!error) { | |
3501 | audit_inode(nd->name, path.dentry, 0); | |
ae2bb293 | 3502 | error = vfs_open(&path, file); |
6ac08709 AV |
3503 | path_put(&path); |
3504 | } | |
3505 | return error; | |
3506 | } | |
3507 | ||
c8a53ee5 AV |
3508 | static struct file *path_openat(struct nameidata *nd, |
3509 | const struct open_flags *op, unsigned flags) | |
1da177e4 | 3510 | { |
30d90494 | 3511 | struct file *file; |
13aab428 | 3512 | int error; |
31e6b01f | 3513 | |
ea73ea72 | 3514 | file = alloc_empty_file(op->open_flag, current_cred()); |
1afc99be AV |
3515 | if (IS_ERR(file)) |
3516 | return file; | |
31e6b01f | 3517 | |
bb458c64 | 3518 | if (unlikely(file->f_flags & __O_TMPFILE)) { |
3ec2eef1 | 3519 | error = do_tmpfile(nd, flags, op, file); |
5f336e72 | 3520 | } else if (unlikely(file->f_flags & O_PATH)) { |
6ac08709 | 3521 | error = do_o_path(nd, flags, file); |
5f336e72 AV |
3522 | } else { |
3523 | const char *s = path_init(nd, flags); | |
3524 | while (!(error = link_path_walk(s, nd)) && | |
3525 | (error = do_last(nd, file, op)) > 0) { | |
3526 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
3527 | s = trailing_symlink(nd); | |
3528 | } | |
3529 | terminate_walk(nd); | |
806b681c | 3530 | } |
7c1c01ec | 3531 | if (likely(!error)) { |
aad888f8 | 3532 | if (likely(file->f_mode & FMODE_OPENED)) |
7c1c01ec AV |
3533 | return file; |
3534 | WARN_ON(1); | |
3535 | error = -EINVAL; | |
16b1c1cd | 3536 | } |
7c1c01ec AV |
3537 | fput(file); |
3538 | if (error == -EOPENSTALE) { | |
3539 | if (flags & LOOKUP_RCU) | |
3540 | error = -ECHILD; | |
3541 | else | |
3542 | error = -ESTALE; | |
2675a4eb | 3543 | } |
7c1c01ec | 3544 | return ERR_PTR(error); |
1da177e4 LT |
3545 | } |
3546 | ||
669abf4e | 3547 | struct file *do_filp_open(int dfd, struct filename *pathname, |
f9652e10 | 3548 | const struct open_flags *op) |
13aab428 | 3549 | { |
9883d185 | 3550 | struct nameidata nd; |
f9652e10 | 3551 | int flags = op->lookup_flags; |
13aab428 AV |
3552 | struct file *filp; |
3553 | ||
9883d185 | 3554 | set_nameidata(&nd, dfd, pathname); |
c8a53ee5 | 3555 | filp = path_openat(&nd, op, flags | LOOKUP_RCU); |
13aab428 | 3556 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3557 | filp = path_openat(&nd, op, flags); |
13aab428 | 3558 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3559 | filp = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3560 | restore_nameidata(); |
13aab428 AV |
3561 | return filp; |
3562 | } | |
3563 | ||
73d049a4 | 3564 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
f9652e10 | 3565 | const char *name, const struct open_flags *op) |
73d049a4 | 3566 | { |
9883d185 | 3567 | struct nameidata nd; |
73d049a4 | 3568 | struct file *file; |
51689104 | 3569 | struct filename *filename; |
f9652e10 | 3570 | int flags = op->lookup_flags | LOOKUP_ROOT; |
73d049a4 AV |
3571 | |
3572 | nd.root.mnt = mnt; | |
3573 | nd.root.dentry = dentry; | |
3574 | ||
b18825a7 | 3575 | if (d_is_symlink(dentry) && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
3576 | return ERR_PTR(-ELOOP); |
3577 | ||
51689104 | 3578 | filename = getname_kernel(name); |
a1c83681 | 3579 | if (IS_ERR(filename)) |
51689104 PM |
3580 | return ERR_CAST(filename); |
3581 | ||
9883d185 | 3582 | set_nameidata(&nd, -1, filename); |
c8a53ee5 | 3583 | file = path_openat(&nd, op, flags | LOOKUP_RCU); |
73d049a4 | 3584 | if (unlikely(file == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3585 | file = path_openat(&nd, op, flags); |
73d049a4 | 3586 | if (unlikely(file == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3587 | file = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3588 | restore_nameidata(); |
51689104 | 3589 | putname(filename); |
73d049a4 AV |
3590 | return file; |
3591 | } | |
3592 | ||
fa14a0b8 | 3593 | static struct dentry *filename_create(int dfd, struct filename *name, |
1ac12b4b | 3594 | struct path *path, unsigned int lookup_flags) |
1da177e4 | 3595 | { |
c663e5d8 | 3596 | struct dentry *dentry = ERR_PTR(-EEXIST); |
391172c4 AV |
3597 | struct qstr last; |
3598 | int type; | |
c30dabfe | 3599 | int err2; |
1ac12b4b JL |
3600 | int error; |
3601 | bool is_dir = (lookup_flags & LOOKUP_DIRECTORY); | |
3602 | ||
3603 | /* | |
3604 | * Note that only LOOKUP_REVAL and LOOKUP_DIRECTORY matter here. Any | |
3605 | * other flags passed in are ignored! | |
3606 | */ | |
3607 | lookup_flags &= LOOKUP_REVAL; | |
3608 | ||
5c31b6ce AV |
3609 | name = filename_parentat(dfd, name, lookup_flags, path, &last, &type); |
3610 | if (IS_ERR(name)) | |
3611 | return ERR_CAST(name); | |
1da177e4 | 3612 | |
c663e5d8 CH |
3613 | /* |
3614 | * Yucky last component or no last component at all? | |
3615 | * (foo/., foo/.., /////) | |
3616 | */ | |
5c31b6ce | 3617 | if (unlikely(type != LAST_NORM)) |
ed75e95d | 3618 | goto out; |
c663e5d8 | 3619 | |
c30dabfe | 3620 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
391172c4 | 3621 | err2 = mnt_want_write(path->mnt); |
c663e5d8 CH |
3622 | /* |
3623 | * Do the final lookup. | |
3624 | */ | |
391172c4 | 3625 | lookup_flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
5955102c | 3626 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 3627 | dentry = __lookup_hash(&last, path->dentry, lookup_flags); |
1da177e4 | 3628 | if (IS_ERR(dentry)) |
a8104a9f | 3629 | goto unlock; |
c663e5d8 | 3630 | |
a8104a9f | 3631 | error = -EEXIST; |
b18825a7 | 3632 | if (d_is_positive(dentry)) |
a8104a9f | 3633 | goto fail; |
b18825a7 | 3634 | |
c663e5d8 CH |
3635 | /* |
3636 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3637 | * From the vfs_mknod() POV we just have a negative dentry - | |
3638 | * all is fine. Let's be bastards - you had / on the end, you've | |
3639 | * been asking for (non-existent) directory. -ENOENT for you. | |
3640 | */ | |
391172c4 | 3641 | if (unlikely(!is_dir && last.name[last.len])) { |
a8104a9f | 3642 | error = -ENOENT; |
ed75e95d | 3643 | goto fail; |
e9baf6e5 | 3644 | } |
c30dabfe JK |
3645 | if (unlikely(err2)) { |
3646 | error = err2; | |
a8104a9f | 3647 | goto fail; |
c30dabfe | 3648 | } |
181c37b6 | 3649 | putname(name); |
1da177e4 | 3650 | return dentry; |
1da177e4 | 3651 | fail: |
a8104a9f AV |
3652 | dput(dentry); |
3653 | dentry = ERR_PTR(error); | |
3654 | unlock: | |
5955102c | 3655 | inode_unlock(path->dentry->d_inode); |
c30dabfe | 3656 | if (!err2) |
391172c4 | 3657 | mnt_drop_write(path->mnt); |
ed75e95d | 3658 | out: |
391172c4 | 3659 | path_put(path); |
181c37b6 | 3660 | putname(name); |
1da177e4 LT |
3661 | return dentry; |
3662 | } | |
fa14a0b8 AV |
3663 | |
3664 | struct dentry *kern_path_create(int dfd, const char *pathname, | |
3665 | struct path *path, unsigned int lookup_flags) | |
3666 | { | |
181c37b6 AV |
3667 | return filename_create(dfd, getname_kernel(pathname), |
3668 | path, lookup_flags); | |
fa14a0b8 | 3669 | } |
dae6ad8f AV |
3670 | EXPORT_SYMBOL(kern_path_create); |
3671 | ||
921a1650 AV |
3672 | void done_path_create(struct path *path, struct dentry *dentry) |
3673 | { | |
3674 | dput(dentry); | |
5955102c | 3675 | inode_unlock(path->dentry->d_inode); |
a8104a9f | 3676 | mnt_drop_write(path->mnt); |
921a1650 AV |
3677 | path_put(path); |
3678 | } | |
3679 | EXPORT_SYMBOL(done_path_create); | |
3680 | ||
520ae687 | 3681 | inline struct dentry *user_path_create(int dfd, const char __user *pathname, |
1ac12b4b | 3682 | struct path *path, unsigned int lookup_flags) |
dae6ad8f | 3683 | { |
181c37b6 | 3684 | return filename_create(dfd, getname(pathname), path, lookup_flags); |
dae6ad8f AV |
3685 | } |
3686 | EXPORT_SYMBOL(user_path_create); | |
3687 | ||
1a67aafb | 3688 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3689 | { |
a95164d9 | 3690 | int error = may_create(dir, dentry); |
1da177e4 LT |
3691 | |
3692 | if (error) | |
3693 | return error; | |
3694 | ||
94f82008 | 3695 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3696 | return -EPERM; |
3697 | ||
acfa4380 | 3698 | if (!dir->i_op->mknod) |
1da177e4 LT |
3699 | return -EPERM; |
3700 | ||
08ce5f16 SH |
3701 | error = devcgroup_inode_mknod(mode, dev); |
3702 | if (error) | |
3703 | return error; | |
3704 | ||
1da177e4 LT |
3705 | error = security_inode_mknod(dir, dentry, mode, dev); |
3706 | if (error) | |
3707 | return error; | |
3708 | ||
1da177e4 | 3709 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3710 | if (!error) |
f38aa942 | 3711 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3712 | return error; |
3713 | } | |
4d359507 | 3714 | EXPORT_SYMBOL(vfs_mknod); |
1da177e4 | 3715 | |
f69aac00 | 3716 | static int may_mknod(umode_t mode) |
463c3197 DH |
3717 | { |
3718 | switch (mode & S_IFMT) { | |
3719 | case S_IFREG: | |
3720 | case S_IFCHR: | |
3721 | case S_IFBLK: | |
3722 | case S_IFIFO: | |
3723 | case S_IFSOCK: | |
3724 | case 0: /* zero mode translates to S_IFREG */ | |
3725 | return 0; | |
3726 | case S_IFDIR: | |
3727 | return -EPERM; | |
3728 | default: | |
3729 | return -EINVAL; | |
3730 | } | |
3731 | } | |
3732 | ||
87c4e192 DB |
3733 | long do_mknodat(int dfd, const char __user *filename, umode_t mode, |
3734 | unsigned int dev) | |
1da177e4 | 3735 | { |
2ad94ae6 | 3736 | struct dentry *dentry; |
dae6ad8f AV |
3737 | struct path path; |
3738 | int error; | |
972567f1 | 3739 | unsigned int lookup_flags = 0; |
1da177e4 | 3740 | |
8e4bfca1 AV |
3741 | error = may_mknod(mode); |
3742 | if (error) | |
3743 | return error; | |
972567f1 JL |
3744 | retry: |
3745 | dentry = user_path_create(dfd, filename, &path, lookup_flags); | |
dae6ad8f AV |
3746 | if (IS_ERR(dentry)) |
3747 | return PTR_ERR(dentry); | |
2ad94ae6 | 3748 | |
dae6ad8f | 3749 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3750 | mode &= ~current_umask(); |
dae6ad8f | 3751 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3752 | if (error) |
a8104a9f | 3753 | goto out; |
463c3197 | 3754 | switch (mode & S_IFMT) { |
1da177e4 | 3755 | case 0: case S_IFREG: |
312b63fb | 3756 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
05d1a717 MZ |
3757 | if (!error) |
3758 | ima_post_path_mknod(dentry); | |
1da177e4 LT |
3759 | break; |
3760 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3761 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3762 | new_decode_dev(dev)); |
3763 | break; | |
3764 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3765 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3766 | break; |
1da177e4 | 3767 | } |
a8104a9f | 3768 | out: |
921a1650 | 3769 | done_path_create(&path, dentry); |
972567f1 JL |
3770 | if (retry_estale(error, lookup_flags)) { |
3771 | lookup_flags |= LOOKUP_REVAL; | |
3772 | goto retry; | |
3773 | } | |
1da177e4 LT |
3774 | return error; |
3775 | } | |
3776 | ||
87c4e192 DB |
3777 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
3778 | unsigned int, dev) | |
3779 | { | |
3780 | return do_mknodat(dfd, filename, mode, dev); | |
3781 | } | |
3782 | ||
8208a22b | 3783 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d | 3784 | { |
87c4e192 | 3785 | return do_mknodat(AT_FDCWD, filename, mode, dev); |
5590ff0d UD |
3786 | } |
3787 | ||
18bb1db3 | 3788 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3789 | { |
a95164d9 | 3790 | int error = may_create(dir, dentry); |
8de52778 | 3791 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3792 | |
3793 | if (error) | |
3794 | return error; | |
3795 | ||
acfa4380 | 3796 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3797 | return -EPERM; |
3798 | ||
3799 | mode &= (S_IRWXUGO|S_ISVTX); | |
3800 | error = security_inode_mkdir(dir, dentry, mode); | |
3801 | if (error) | |
3802 | return error; | |
3803 | ||
8de52778 AV |
3804 | if (max_links && dir->i_nlink >= max_links) |
3805 | return -EMLINK; | |
3806 | ||
1da177e4 | 3807 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3808 | if (!error) |
f38aa942 | 3809 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3810 | return error; |
3811 | } | |
4d359507 | 3812 | EXPORT_SYMBOL(vfs_mkdir); |
1da177e4 | 3813 | |
0101db7a | 3814 | long do_mkdirat(int dfd, const char __user *pathname, umode_t mode) |
1da177e4 | 3815 | { |
6902d925 | 3816 | struct dentry *dentry; |
dae6ad8f AV |
3817 | struct path path; |
3818 | int error; | |
b76d8b82 | 3819 | unsigned int lookup_flags = LOOKUP_DIRECTORY; |
1da177e4 | 3820 | |
b76d8b82 JL |
3821 | retry: |
3822 | dentry = user_path_create(dfd, pathname, &path, lookup_flags); | |
6902d925 | 3823 | if (IS_ERR(dentry)) |
dae6ad8f | 3824 | return PTR_ERR(dentry); |
1da177e4 | 3825 | |
dae6ad8f | 3826 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3827 | mode &= ~current_umask(); |
dae6ad8f | 3828 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3829 | if (!error) |
3830 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3831 | done_path_create(&path, dentry); |
b76d8b82 JL |
3832 | if (retry_estale(error, lookup_flags)) { |
3833 | lookup_flags |= LOOKUP_REVAL; | |
3834 | goto retry; | |
3835 | } | |
1da177e4 LT |
3836 | return error; |
3837 | } | |
3838 | ||
0101db7a DB |
3839 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
3840 | { | |
3841 | return do_mkdirat(dfd, pathname, mode); | |
3842 | } | |
3843 | ||
a218d0fd | 3844 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d | 3845 | { |
0101db7a | 3846 | return do_mkdirat(AT_FDCWD, pathname, mode); |
5590ff0d UD |
3847 | } |
3848 | ||
1da177e4 LT |
3849 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) |
3850 | { | |
3851 | int error = may_delete(dir, dentry, 1); | |
3852 | ||
3853 | if (error) | |
3854 | return error; | |
3855 | ||
acfa4380 | 3856 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3857 | return -EPERM; |
3858 | ||
1d2ef590 | 3859 | dget(dentry); |
5955102c | 3860 | inode_lock(dentry->d_inode); |
912dbc15 SW |
3861 | |
3862 | error = -EBUSY; | |
7af1364f | 3863 | if (is_local_mountpoint(dentry)) |
912dbc15 SW |
3864 | goto out; |
3865 | ||
3866 | error = security_inode_rmdir(dir, dentry); | |
3867 | if (error) | |
3868 | goto out; | |
3869 | ||
3870 | error = dir->i_op->rmdir(dir, dentry); | |
3871 | if (error) | |
3872 | goto out; | |
3873 | ||
8767712f | 3874 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3875 | dentry->d_inode->i_flags |= S_DEAD; |
3876 | dont_mount(dentry); | |
8ed936b5 | 3877 | detach_mounts(dentry); |
116b9731 | 3878 | fsnotify_rmdir(dir, dentry); |
912dbc15 SW |
3879 | |
3880 | out: | |
5955102c | 3881 | inode_unlock(dentry->d_inode); |
1d2ef590 | 3882 | dput(dentry); |
912dbc15 | 3883 | if (!error) |
1da177e4 | 3884 | d_delete(dentry); |
1da177e4 LT |
3885 | return error; |
3886 | } | |
4d359507 | 3887 | EXPORT_SYMBOL(vfs_rmdir); |
1da177e4 | 3888 | |
f459dffa | 3889 | long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3890 | { |
3891 | int error = 0; | |
91a27b2a | 3892 | struct filename *name; |
1da177e4 | 3893 | struct dentry *dentry; |
f5beed75 AV |
3894 | struct path path; |
3895 | struct qstr last; | |
3896 | int type; | |
c6ee9206 JL |
3897 | unsigned int lookup_flags = 0; |
3898 | retry: | |
c1d4dd27 AV |
3899 | name = filename_parentat(dfd, getname(pathname), lookup_flags, |
3900 | &path, &last, &type); | |
91a27b2a JL |
3901 | if (IS_ERR(name)) |
3902 | return PTR_ERR(name); | |
1da177e4 | 3903 | |
f5beed75 | 3904 | switch (type) { |
0612d9fb OH |
3905 | case LAST_DOTDOT: |
3906 | error = -ENOTEMPTY; | |
3907 | goto exit1; | |
3908 | case LAST_DOT: | |
3909 | error = -EINVAL; | |
3910 | goto exit1; | |
3911 | case LAST_ROOT: | |
3912 | error = -EBUSY; | |
3913 | goto exit1; | |
1da177e4 | 3914 | } |
0612d9fb | 3915 | |
f5beed75 | 3916 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3917 | if (error) |
3918 | goto exit1; | |
0612d9fb | 3919 | |
5955102c | 3920 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 3921 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 | 3922 | error = PTR_ERR(dentry); |
6902d925 DH |
3923 | if (IS_ERR(dentry)) |
3924 | goto exit2; | |
e6bc45d6 TT |
3925 | if (!dentry->d_inode) { |
3926 | error = -ENOENT; | |
3927 | goto exit3; | |
3928 | } | |
f5beed75 | 3929 | error = security_path_rmdir(&path, dentry); |
be6d3e56 | 3930 | if (error) |
c30dabfe | 3931 | goto exit3; |
f5beed75 | 3932 | error = vfs_rmdir(path.dentry->d_inode, dentry); |
0622753b | 3933 | exit3: |
6902d925 DH |
3934 | dput(dentry); |
3935 | exit2: | |
5955102c | 3936 | inode_unlock(path.dentry->d_inode); |
f5beed75 | 3937 | mnt_drop_write(path.mnt); |
1da177e4 | 3938 | exit1: |
f5beed75 | 3939 | path_put(&path); |
1da177e4 | 3940 | putname(name); |
c6ee9206 JL |
3941 | if (retry_estale(error, lookup_flags)) { |
3942 | lookup_flags |= LOOKUP_REVAL; | |
3943 | goto retry; | |
3944 | } | |
1da177e4 LT |
3945 | return error; |
3946 | } | |
3947 | ||
3cdad428 | 3948 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3949 | { |
3950 | return do_rmdir(AT_FDCWD, pathname); | |
3951 | } | |
3952 | ||
b21996e3 BF |
3953 | /** |
3954 | * vfs_unlink - unlink a filesystem object | |
3955 | * @dir: parent directory | |
3956 | * @dentry: victim | |
3957 | * @delegated_inode: returns victim inode, if the inode is delegated. | |
3958 | * | |
3959 | * The caller must hold dir->i_mutex. | |
3960 | * | |
3961 | * If vfs_unlink discovers a delegation, it will return -EWOULDBLOCK and | |
3962 | * return a reference to the inode in delegated_inode. The caller | |
3963 | * should then break the delegation on that inode and retry. Because | |
3964 | * breaking a delegation may take a long time, the caller should drop | |
3965 | * dir->i_mutex before doing so. | |
3966 | * | |
3967 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3968 | * be appropriate for callers that expect the underlying filesystem not | |
3969 | * to be NFS exported. | |
3970 | */ | |
3971 | int vfs_unlink(struct inode *dir, struct dentry *dentry, struct inode **delegated_inode) | |
1da177e4 | 3972 | { |
9accbb97 | 3973 | struct inode *target = dentry->d_inode; |
1da177e4 LT |
3974 | int error = may_delete(dir, dentry, 0); |
3975 | ||
3976 | if (error) | |
3977 | return error; | |
3978 | ||
acfa4380 | 3979 | if (!dir->i_op->unlink) |
1da177e4 LT |
3980 | return -EPERM; |
3981 | ||
5955102c | 3982 | inode_lock(target); |
8ed936b5 | 3983 | if (is_local_mountpoint(dentry)) |
1da177e4 LT |
3984 | error = -EBUSY; |
3985 | else { | |
3986 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3987 | if (!error) { |
5a14696c BF |
3988 | error = try_break_deleg(target, delegated_inode); |
3989 | if (error) | |
b21996e3 | 3990 | goto out; |
1da177e4 | 3991 | error = dir->i_op->unlink(dir, dentry); |
8ed936b5 | 3992 | if (!error) { |
d83c49f3 | 3993 | dont_mount(dentry); |
8ed936b5 | 3994 | detach_mounts(dentry); |
116b9731 | 3995 | fsnotify_unlink(dir, dentry); |
8ed936b5 | 3996 | } |
bec1052e | 3997 | } |
1da177e4 | 3998 | } |
b21996e3 | 3999 | out: |
5955102c | 4000 | inode_unlock(target); |
1da177e4 LT |
4001 | |
4002 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
4003 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
9accbb97 | 4004 | fsnotify_link_count(target); |
e234f35c | 4005 | d_delete(dentry); |
1da177e4 | 4006 | } |
0eeca283 | 4007 | |
1da177e4 LT |
4008 | return error; |
4009 | } | |
4d359507 | 4010 | EXPORT_SYMBOL(vfs_unlink); |
1da177e4 LT |
4011 | |
4012 | /* | |
4013 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 4014 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
4015 | * writeout happening, and we don't want to prevent access to the directory |
4016 | * while waiting on the I/O. | |
4017 | */ | |
da2f1362 | 4018 | long do_unlinkat(int dfd, struct filename *name) |
1da177e4 | 4019 | { |
2ad94ae6 | 4020 | int error; |
1da177e4 | 4021 | struct dentry *dentry; |
f5beed75 AV |
4022 | struct path path; |
4023 | struct qstr last; | |
4024 | int type; | |
1da177e4 | 4025 | struct inode *inode = NULL; |
b21996e3 | 4026 | struct inode *delegated_inode = NULL; |
5d18f813 JL |
4027 | unsigned int lookup_flags = 0; |
4028 | retry: | |
da2f1362 | 4029 | name = filename_parentat(dfd, name, lookup_flags, &path, &last, &type); |
91a27b2a JL |
4030 | if (IS_ERR(name)) |
4031 | return PTR_ERR(name); | |
2ad94ae6 | 4032 | |
1da177e4 | 4033 | error = -EISDIR; |
f5beed75 | 4034 | if (type != LAST_NORM) |
1da177e4 | 4035 | goto exit1; |
0612d9fb | 4036 | |
f5beed75 | 4037 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
4038 | if (error) |
4039 | goto exit1; | |
b21996e3 | 4040 | retry_deleg: |
5955102c | 4041 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 4042 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 LT |
4043 | error = PTR_ERR(dentry); |
4044 | if (!IS_ERR(dentry)) { | |
4045 | /* Why not before? Because we want correct error value */ | |
f5beed75 | 4046 | if (last.name[last.len]) |
50338b88 | 4047 | goto slashes; |
1da177e4 | 4048 | inode = dentry->d_inode; |
b18825a7 | 4049 | if (d_is_negative(dentry)) |
e6bc45d6 TT |
4050 | goto slashes; |
4051 | ihold(inode); | |
f5beed75 | 4052 | error = security_path_unlink(&path, dentry); |
be6d3e56 | 4053 | if (error) |
c30dabfe | 4054 | goto exit2; |
f5beed75 | 4055 | error = vfs_unlink(path.dentry->d_inode, dentry, &delegated_inode); |
c30dabfe | 4056 | exit2: |
1da177e4 LT |
4057 | dput(dentry); |
4058 | } | |
5955102c | 4059 | inode_unlock(path.dentry->d_inode); |
1da177e4 LT |
4060 | if (inode) |
4061 | iput(inode); /* truncate the inode here */ | |
b21996e3 BF |
4062 | inode = NULL; |
4063 | if (delegated_inode) { | |
5a14696c | 4064 | error = break_deleg_wait(&delegated_inode); |
b21996e3 BF |
4065 | if (!error) |
4066 | goto retry_deleg; | |
4067 | } | |
f5beed75 | 4068 | mnt_drop_write(path.mnt); |
1da177e4 | 4069 | exit1: |
f5beed75 | 4070 | path_put(&path); |
5d18f813 JL |
4071 | if (retry_estale(error, lookup_flags)) { |
4072 | lookup_flags |= LOOKUP_REVAL; | |
4073 | inode = NULL; | |
4074 | goto retry; | |
4075 | } | |
da2f1362 | 4076 | putname(name); |
1da177e4 LT |
4077 | return error; |
4078 | ||
4079 | slashes: | |
b18825a7 DH |
4080 | if (d_is_negative(dentry)) |
4081 | error = -ENOENT; | |
44b1d530 | 4082 | else if (d_is_dir(dentry)) |
b18825a7 DH |
4083 | error = -EISDIR; |
4084 | else | |
4085 | error = -ENOTDIR; | |
1da177e4 LT |
4086 | goto exit2; |
4087 | } | |
4088 | ||
2e4d0924 | 4089 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
4090 | { |
4091 | if ((flag & ~AT_REMOVEDIR) != 0) | |
4092 | return -EINVAL; | |
4093 | ||
4094 | if (flag & AT_REMOVEDIR) | |
4095 | return do_rmdir(dfd, pathname); | |
4096 | ||
da2f1362 | 4097 | return do_unlinkat(dfd, getname(pathname)); |
5590ff0d UD |
4098 | } |
4099 | ||
3480b257 | 4100 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d | 4101 | { |
da2f1362 | 4102 | return do_unlinkat(AT_FDCWD, getname(pathname)); |
5590ff0d UD |
4103 | } |
4104 | ||
db2e747b | 4105 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 4106 | { |
a95164d9 | 4107 | int error = may_create(dir, dentry); |
1da177e4 LT |
4108 | |
4109 | if (error) | |
4110 | return error; | |
4111 | ||
acfa4380 | 4112 | if (!dir->i_op->symlink) |
1da177e4 LT |
4113 | return -EPERM; |
4114 | ||
4115 | error = security_inode_symlink(dir, dentry, oldname); | |
4116 | if (error) | |
4117 | return error; | |
4118 | ||
1da177e4 | 4119 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 4120 | if (!error) |
f38aa942 | 4121 | fsnotify_create(dir, dentry); |
1da177e4 LT |
4122 | return error; |
4123 | } | |
4d359507 | 4124 | EXPORT_SYMBOL(vfs_symlink); |
1da177e4 | 4125 | |
b724e846 DB |
4126 | long do_symlinkat(const char __user *oldname, int newdfd, |
4127 | const char __user *newname) | |
1da177e4 | 4128 | { |
2ad94ae6 | 4129 | int error; |
91a27b2a | 4130 | struct filename *from; |
6902d925 | 4131 | struct dentry *dentry; |
dae6ad8f | 4132 | struct path path; |
f46d3567 | 4133 | unsigned int lookup_flags = 0; |
1da177e4 LT |
4134 | |
4135 | from = getname(oldname); | |
2ad94ae6 | 4136 | if (IS_ERR(from)) |
1da177e4 | 4137 | return PTR_ERR(from); |
f46d3567 JL |
4138 | retry: |
4139 | dentry = user_path_create(newdfd, newname, &path, lookup_flags); | |
6902d925 DH |
4140 | error = PTR_ERR(dentry); |
4141 | if (IS_ERR(dentry)) | |
dae6ad8f | 4142 | goto out_putname; |
6902d925 | 4143 | |
91a27b2a | 4144 | error = security_path_symlink(&path, dentry, from->name); |
a8104a9f | 4145 | if (!error) |
91a27b2a | 4146 | error = vfs_symlink(path.dentry->d_inode, dentry, from->name); |
921a1650 | 4147 | done_path_create(&path, dentry); |
f46d3567 JL |
4148 | if (retry_estale(error, lookup_flags)) { |
4149 | lookup_flags |= LOOKUP_REVAL; | |
4150 | goto retry; | |
4151 | } | |
6902d925 | 4152 | out_putname: |
1da177e4 LT |
4153 | putname(from); |
4154 | return error; | |
4155 | } | |
4156 | ||
b724e846 DB |
4157 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
4158 | int, newdfd, const char __user *, newname) | |
4159 | { | |
4160 | return do_symlinkat(oldname, newdfd, newname); | |
4161 | } | |
4162 | ||
3480b257 | 4163 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4164 | { |
b724e846 | 4165 | return do_symlinkat(oldname, AT_FDCWD, newname); |
5590ff0d UD |
4166 | } |
4167 | ||
146a8595 BF |
4168 | /** |
4169 | * vfs_link - create a new link | |
4170 | * @old_dentry: object to be linked | |
4171 | * @dir: new parent | |
4172 | * @new_dentry: where to create the new link | |
4173 | * @delegated_inode: returns inode needing a delegation break | |
4174 | * | |
4175 | * The caller must hold dir->i_mutex | |
4176 | * | |
4177 | * If vfs_link discovers a delegation on the to-be-linked file in need | |
4178 | * of breaking, it will return -EWOULDBLOCK and return a reference to the | |
4179 | * inode in delegated_inode. The caller should then break the delegation | |
4180 | * and retry. Because breaking a delegation may take a long time, the | |
4181 | * caller should drop the i_mutex before doing so. | |
4182 | * | |
4183 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4184 | * be appropriate for callers that expect the underlying filesystem not | |
4185 | * to be NFS exported. | |
4186 | */ | |
4187 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, struct inode **delegated_inode) | |
1da177e4 LT |
4188 | { |
4189 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 4190 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
4191 | int error; |
4192 | ||
4193 | if (!inode) | |
4194 | return -ENOENT; | |
4195 | ||
a95164d9 | 4196 | error = may_create(dir, new_dentry); |
1da177e4 LT |
4197 | if (error) |
4198 | return error; | |
4199 | ||
4200 | if (dir->i_sb != inode->i_sb) | |
4201 | return -EXDEV; | |
4202 | ||
4203 | /* | |
4204 | * A link to an append-only or immutable file cannot be created. | |
4205 | */ | |
4206 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
4207 | return -EPERM; | |
0bd23d09 EB |
4208 | /* |
4209 | * Updating the link count will likely cause i_uid and i_gid to | |
4210 | * be writen back improperly if their true value is unknown to | |
4211 | * the vfs. | |
4212 | */ | |
4213 | if (HAS_UNMAPPED_ID(inode)) | |
4214 | return -EPERM; | |
acfa4380 | 4215 | if (!dir->i_op->link) |
1da177e4 | 4216 | return -EPERM; |
7e79eedb | 4217 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
4218 | return -EPERM; |
4219 | ||
4220 | error = security_inode_link(old_dentry, dir, new_dentry); | |
4221 | if (error) | |
4222 | return error; | |
4223 | ||
5955102c | 4224 | inode_lock(inode); |
aae8a97d | 4225 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
f4e0c30c | 4226 | if (inode->i_nlink == 0 && !(inode->i_state & I_LINKABLE)) |
aae8a97d | 4227 | error = -ENOENT; |
8de52778 AV |
4228 | else if (max_links && inode->i_nlink >= max_links) |
4229 | error = -EMLINK; | |
146a8595 BF |
4230 | else { |
4231 | error = try_break_deleg(inode, delegated_inode); | |
4232 | if (!error) | |
4233 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
4234 | } | |
f4e0c30c AV |
4235 | |
4236 | if (!error && (inode->i_state & I_LINKABLE)) { | |
4237 | spin_lock(&inode->i_lock); | |
4238 | inode->i_state &= ~I_LINKABLE; | |
4239 | spin_unlock(&inode->i_lock); | |
4240 | } | |
5955102c | 4241 | inode_unlock(inode); |
e31e14ec | 4242 | if (!error) |
7e79eedb | 4243 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
4244 | return error; |
4245 | } | |
4d359507 | 4246 | EXPORT_SYMBOL(vfs_link); |
1da177e4 LT |
4247 | |
4248 | /* | |
4249 | * Hardlinks are often used in delicate situations. We avoid | |
4250 | * security-related surprises by not following symlinks on the | |
4251 | * newname. --KAB | |
4252 | * | |
4253 | * We don't follow them on the oldname either to be compatible | |
4254 | * with linux 2.0, and to avoid hard-linking to directories | |
4255 | * and other special files. --ADM | |
4256 | */ | |
46ea89eb DB |
4257 | int do_linkat(int olddfd, const char __user *oldname, int newdfd, |
4258 | const char __user *newname, int flags) | |
1da177e4 LT |
4259 | { |
4260 | struct dentry *new_dentry; | |
dae6ad8f | 4261 | struct path old_path, new_path; |
146a8595 | 4262 | struct inode *delegated_inode = NULL; |
11a7b371 | 4263 | int how = 0; |
1da177e4 | 4264 | int error; |
1da177e4 | 4265 | |
11a7b371 | 4266 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 4267 | return -EINVAL; |
11a7b371 | 4268 | /* |
f0cc6ffb LT |
4269 | * To use null names we require CAP_DAC_READ_SEARCH |
4270 | * This ensures that not everyone will be able to create | |
4271 | * handlink using the passed filedescriptor. | |
11a7b371 | 4272 | */ |
f0cc6ffb LT |
4273 | if (flags & AT_EMPTY_PATH) { |
4274 | if (!capable(CAP_DAC_READ_SEARCH)) | |
4275 | return -ENOENT; | |
11a7b371 | 4276 | how = LOOKUP_EMPTY; |
f0cc6ffb | 4277 | } |
11a7b371 AK |
4278 | |
4279 | if (flags & AT_SYMLINK_FOLLOW) | |
4280 | how |= LOOKUP_FOLLOW; | |
442e31ca | 4281 | retry: |
11a7b371 | 4282 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 4283 | if (error) |
2ad94ae6 AV |
4284 | return error; |
4285 | ||
442e31ca JL |
4286 | new_dentry = user_path_create(newdfd, newname, &new_path, |
4287 | (how & LOOKUP_REVAL)); | |
1da177e4 | 4288 | error = PTR_ERR(new_dentry); |
6902d925 | 4289 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
4290 | goto out; |
4291 | ||
4292 | error = -EXDEV; | |
4293 | if (old_path.mnt != new_path.mnt) | |
4294 | goto out_dput; | |
800179c9 KC |
4295 | error = may_linkat(&old_path); |
4296 | if (unlikely(error)) | |
4297 | goto out_dput; | |
dae6ad8f | 4298 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 4299 | if (error) |
a8104a9f | 4300 | goto out_dput; |
146a8595 | 4301 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry, &delegated_inode); |
75c3f29d | 4302 | out_dput: |
921a1650 | 4303 | done_path_create(&new_path, new_dentry); |
146a8595 BF |
4304 | if (delegated_inode) { |
4305 | error = break_deleg_wait(&delegated_inode); | |
d22e6338 OD |
4306 | if (!error) { |
4307 | path_put(&old_path); | |
146a8595 | 4308 | goto retry; |
d22e6338 | 4309 | } |
146a8595 | 4310 | } |
442e31ca | 4311 | if (retry_estale(error, how)) { |
d22e6338 | 4312 | path_put(&old_path); |
442e31ca JL |
4313 | how |= LOOKUP_REVAL; |
4314 | goto retry; | |
4315 | } | |
1da177e4 | 4316 | out: |
2d8f3038 | 4317 | path_put(&old_path); |
1da177e4 LT |
4318 | |
4319 | return error; | |
4320 | } | |
4321 | ||
46ea89eb DB |
4322 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
4323 | int, newdfd, const char __user *, newname, int, flags) | |
4324 | { | |
4325 | return do_linkat(olddfd, oldname, newdfd, newname, flags); | |
4326 | } | |
4327 | ||
3480b257 | 4328 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4329 | { |
46ea89eb | 4330 | return do_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4331 | } |
4332 | ||
bc27027a MS |
4333 | /** |
4334 | * vfs_rename - rename a filesystem object | |
4335 | * @old_dir: parent of source | |
4336 | * @old_dentry: source | |
4337 | * @new_dir: parent of destination | |
4338 | * @new_dentry: destination | |
4339 | * @delegated_inode: returns an inode needing a delegation break | |
520c8b16 | 4340 | * @flags: rename flags |
bc27027a MS |
4341 | * |
4342 | * The caller must hold multiple mutexes--see lock_rename()). | |
4343 | * | |
4344 | * If vfs_rename discovers a delegation in need of breaking at either | |
4345 | * the source or destination, it will return -EWOULDBLOCK and return a | |
4346 | * reference to the inode in delegated_inode. The caller should then | |
4347 | * break the delegation and retry. Because breaking a delegation may | |
4348 | * take a long time, the caller should drop all locks before doing | |
4349 | * so. | |
4350 | * | |
4351 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4352 | * be appropriate for callers that expect the underlying filesystem not | |
4353 | * to be NFS exported. | |
4354 | * | |
1da177e4 LT |
4355 | * The worst of all namespace operations - renaming directory. "Perverted" |
4356 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
4357 | * Problems: | |
0117d427 | 4358 | * |
d03b29a2 | 4359 | * a) we can get into loop creation. |
1da177e4 LT |
4360 | * b) race potential - two innocent renames can create a loop together. |
4361 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 4362 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 | 4363 | * story. |
6cedba89 BF |
4364 | * c) we have to lock _four_ objects - parents and victim (if it exists), |
4365 | * and source (if it is not a directory). | |
1b1dcc1b | 4366 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
4367 | * whether the target exists). Solution: try to be smart with locking |
4368 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 4369 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
4370 | * move will be locked. Thus we can rank directories by the tree |
4371 | * (ancestors first) and rank all non-directories after them. | |
4372 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 4373 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
4374 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
4375 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
4376 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 4377 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 4378 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 4379 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 4380 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
4381 | * locking]. |
4382 | */ | |
bc27027a MS |
4383 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, |
4384 | struct inode *new_dir, struct dentry *new_dentry, | |
520c8b16 | 4385 | struct inode **delegated_inode, unsigned int flags) |
1da177e4 | 4386 | { |
bc27027a MS |
4387 | int error; |
4388 | bool is_dir = d_is_dir(old_dentry); | |
bc27027a | 4389 | struct inode *source = old_dentry->d_inode; |
9055cba7 | 4390 | struct inode *target = new_dentry->d_inode; |
da1ce067 MS |
4391 | bool new_is_dir = false; |
4392 | unsigned max_links = new_dir->i_sb->s_max_links; | |
49d31c2f | 4393 | struct name_snapshot old_name; |
bc27027a | 4394 | |
8d3e2936 | 4395 | if (source == target) |
bc27027a MS |
4396 | return 0; |
4397 | ||
4398 | error = may_delete(old_dir, old_dentry, is_dir); | |
4399 | if (error) | |
4400 | return error; | |
4401 | ||
da1ce067 | 4402 | if (!target) { |
bc27027a | 4403 | error = may_create(new_dir, new_dentry); |
da1ce067 MS |
4404 | } else { |
4405 | new_is_dir = d_is_dir(new_dentry); | |
4406 | ||
4407 | if (!(flags & RENAME_EXCHANGE)) | |
4408 | error = may_delete(new_dir, new_dentry, is_dir); | |
4409 | else | |
4410 | error = may_delete(new_dir, new_dentry, new_is_dir); | |
4411 | } | |
bc27027a MS |
4412 | if (error) |
4413 | return error; | |
4414 | ||
2773bf00 | 4415 | if (!old_dir->i_op->rename) |
bc27027a | 4416 | return -EPERM; |
1da177e4 LT |
4417 | |
4418 | /* | |
4419 | * If we are going to change the parent - check write permissions, | |
4420 | * we'll need to flip '..'. | |
4421 | */ | |
da1ce067 MS |
4422 | if (new_dir != old_dir) { |
4423 | if (is_dir) { | |
4424 | error = inode_permission(source, MAY_WRITE); | |
4425 | if (error) | |
4426 | return error; | |
4427 | } | |
4428 | if ((flags & RENAME_EXCHANGE) && new_is_dir) { | |
4429 | error = inode_permission(target, MAY_WRITE); | |
4430 | if (error) | |
4431 | return error; | |
4432 | } | |
1da177e4 LT |
4433 | } |
4434 | ||
0b3974eb MS |
4435 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry, |
4436 | flags); | |
1da177e4 LT |
4437 | if (error) |
4438 | return error; | |
4439 | ||
49d31c2f | 4440 | take_dentry_name_snapshot(&old_name, old_dentry); |
1d2ef590 | 4441 | dget(new_dentry); |
da1ce067 | 4442 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4443 | lock_two_nondirectories(source, target); |
4444 | else if (target) | |
5955102c | 4445 | inode_lock(target); |
9055cba7 SW |
4446 | |
4447 | error = -EBUSY; | |
7af1364f | 4448 | if (is_local_mountpoint(old_dentry) || is_local_mountpoint(new_dentry)) |
9055cba7 SW |
4449 | goto out; |
4450 | ||
da1ce067 | 4451 | if (max_links && new_dir != old_dir) { |
bc27027a | 4452 | error = -EMLINK; |
da1ce067 | 4453 | if (is_dir && !new_is_dir && new_dir->i_nlink >= max_links) |
bc27027a | 4454 | goto out; |
da1ce067 MS |
4455 | if ((flags & RENAME_EXCHANGE) && !is_dir && new_is_dir && |
4456 | old_dir->i_nlink >= max_links) | |
4457 | goto out; | |
4458 | } | |
da1ce067 | 4459 | if (!is_dir) { |
bc27027a | 4460 | error = try_break_deleg(source, delegated_inode); |
8e6d782c BF |
4461 | if (error) |
4462 | goto out; | |
da1ce067 MS |
4463 | } |
4464 | if (target && !new_is_dir) { | |
4465 | error = try_break_deleg(target, delegated_inode); | |
4466 | if (error) | |
4467 | goto out; | |
8e6d782c | 4468 | } |
2773bf00 | 4469 | error = old_dir->i_op->rename(old_dir, old_dentry, |
18fc84da | 4470 | new_dir, new_dentry, flags); |
51892bbb SW |
4471 | if (error) |
4472 | goto out; | |
4473 | ||
da1ce067 | 4474 | if (!(flags & RENAME_EXCHANGE) && target) { |
8767712f AV |
4475 | if (is_dir) { |
4476 | shrink_dcache_parent(new_dentry); | |
bc27027a | 4477 | target->i_flags |= S_DEAD; |
8767712f | 4478 | } |
51892bbb | 4479 | dont_mount(new_dentry); |
8ed936b5 | 4480 | detach_mounts(new_dentry); |
bc27027a | 4481 | } |
da1ce067 MS |
4482 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) { |
4483 | if (!(flags & RENAME_EXCHANGE)) | |
4484 | d_move(old_dentry, new_dentry); | |
4485 | else | |
4486 | d_exchange(old_dentry, new_dentry); | |
4487 | } | |
51892bbb | 4488 | out: |
da1ce067 | 4489 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4490 | unlock_two_nondirectories(source, target); |
4491 | else if (target) | |
5955102c | 4492 | inode_unlock(target); |
1da177e4 | 4493 | dput(new_dentry); |
da1ce067 | 4494 | if (!error) { |
f4ec3a3d | 4495 | fsnotify_move(old_dir, new_dir, &old_name.name, is_dir, |
da1ce067 MS |
4496 | !(flags & RENAME_EXCHANGE) ? target : NULL, old_dentry); |
4497 | if (flags & RENAME_EXCHANGE) { | |
f4ec3a3d | 4498 | fsnotify_move(new_dir, old_dir, &old_dentry->d_name, |
da1ce067 MS |
4499 | new_is_dir, NULL, new_dentry); |
4500 | } | |
4501 | } | |
49d31c2f | 4502 | release_dentry_name_snapshot(&old_name); |
0eeca283 | 4503 | |
1da177e4 LT |
4504 | return error; |
4505 | } | |
4d359507 | 4506 | EXPORT_SYMBOL(vfs_rename); |
1da177e4 | 4507 | |
ee81feb6 DB |
4508 | static int do_renameat2(int olddfd, const char __user *oldname, int newdfd, |
4509 | const char __user *newname, unsigned int flags) | |
1da177e4 | 4510 | { |
2ad94ae6 AV |
4511 | struct dentry *old_dentry, *new_dentry; |
4512 | struct dentry *trap; | |
f5beed75 AV |
4513 | struct path old_path, new_path; |
4514 | struct qstr old_last, new_last; | |
4515 | int old_type, new_type; | |
8e6d782c | 4516 | struct inode *delegated_inode = NULL; |
91a27b2a JL |
4517 | struct filename *from; |
4518 | struct filename *to; | |
f5beed75 | 4519 | unsigned int lookup_flags = 0, target_flags = LOOKUP_RENAME_TARGET; |
c6a94284 | 4520 | bool should_retry = false; |
2ad94ae6 | 4521 | int error; |
520c8b16 | 4522 | |
0d7a8555 | 4523 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
da1ce067 MS |
4524 | return -EINVAL; |
4525 | ||
0d7a8555 MS |
4526 | if ((flags & (RENAME_NOREPLACE | RENAME_WHITEOUT)) && |
4527 | (flags & RENAME_EXCHANGE)) | |
520c8b16 MS |
4528 | return -EINVAL; |
4529 | ||
0d7a8555 MS |
4530 | if ((flags & RENAME_WHITEOUT) && !capable(CAP_MKNOD)) |
4531 | return -EPERM; | |
4532 | ||
f5beed75 AV |
4533 | if (flags & RENAME_EXCHANGE) |
4534 | target_flags = 0; | |
4535 | ||
c6a94284 | 4536 | retry: |
c1d4dd27 AV |
4537 | from = filename_parentat(olddfd, getname(oldname), lookup_flags, |
4538 | &old_path, &old_last, &old_type); | |
91a27b2a JL |
4539 | if (IS_ERR(from)) { |
4540 | error = PTR_ERR(from); | |
1da177e4 | 4541 | goto exit; |
91a27b2a | 4542 | } |
1da177e4 | 4543 | |
c1d4dd27 AV |
4544 | to = filename_parentat(newdfd, getname(newname), lookup_flags, |
4545 | &new_path, &new_last, &new_type); | |
91a27b2a JL |
4546 | if (IS_ERR(to)) { |
4547 | error = PTR_ERR(to); | |
1da177e4 | 4548 | goto exit1; |
91a27b2a | 4549 | } |
1da177e4 LT |
4550 | |
4551 | error = -EXDEV; | |
f5beed75 | 4552 | if (old_path.mnt != new_path.mnt) |
1da177e4 LT |
4553 | goto exit2; |
4554 | ||
1da177e4 | 4555 | error = -EBUSY; |
f5beed75 | 4556 | if (old_type != LAST_NORM) |
1da177e4 LT |
4557 | goto exit2; |
4558 | ||
0a7c3937 MS |
4559 | if (flags & RENAME_NOREPLACE) |
4560 | error = -EEXIST; | |
f5beed75 | 4561 | if (new_type != LAST_NORM) |
1da177e4 LT |
4562 | goto exit2; |
4563 | ||
f5beed75 | 4564 | error = mnt_want_write(old_path.mnt); |
c30dabfe JK |
4565 | if (error) |
4566 | goto exit2; | |
4567 | ||
8e6d782c | 4568 | retry_deleg: |
f5beed75 | 4569 | trap = lock_rename(new_path.dentry, old_path.dentry); |
1da177e4 | 4570 | |
f5beed75 | 4571 | old_dentry = __lookup_hash(&old_last, old_path.dentry, lookup_flags); |
1da177e4 LT |
4572 | error = PTR_ERR(old_dentry); |
4573 | if (IS_ERR(old_dentry)) | |
4574 | goto exit3; | |
4575 | /* source must exist */ | |
4576 | error = -ENOENT; | |
b18825a7 | 4577 | if (d_is_negative(old_dentry)) |
1da177e4 | 4578 | goto exit4; |
f5beed75 | 4579 | new_dentry = __lookup_hash(&new_last, new_path.dentry, lookup_flags | target_flags); |
0a7c3937 MS |
4580 | error = PTR_ERR(new_dentry); |
4581 | if (IS_ERR(new_dentry)) | |
4582 | goto exit4; | |
4583 | error = -EEXIST; | |
4584 | if ((flags & RENAME_NOREPLACE) && d_is_positive(new_dentry)) | |
4585 | goto exit5; | |
da1ce067 MS |
4586 | if (flags & RENAME_EXCHANGE) { |
4587 | error = -ENOENT; | |
4588 | if (d_is_negative(new_dentry)) | |
4589 | goto exit5; | |
4590 | ||
4591 | if (!d_is_dir(new_dentry)) { | |
4592 | error = -ENOTDIR; | |
f5beed75 | 4593 | if (new_last.name[new_last.len]) |
da1ce067 MS |
4594 | goto exit5; |
4595 | } | |
4596 | } | |
1da177e4 | 4597 | /* unless the source is a directory trailing slashes give -ENOTDIR */ |
44b1d530 | 4598 | if (!d_is_dir(old_dentry)) { |
1da177e4 | 4599 | error = -ENOTDIR; |
f5beed75 | 4600 | if (old_last.name[old_last.len]) |
0a7c3937 | 4601 | goto exit5; |
f5beed75 | 4602 | if (!(flags & RENAME_EXCHANGE) && new_last.name[new_last.len]) |
0a7c3937 | 4603 | goto exit5; |
1da177e4 LT |
4604 | } |
4605 | /* source should not be ancestor of target */ | |
4606 | error = -EINVAL; | |
4607 | if (old_dentry == trap) | |
0a7c3937 | 4608 | goto exit5; |
1da177e4 | 4609 | /* target should not be an ancestor of source */ |
da1ce067 MS |
4610 | if (!(flags & RENAME_EXCHANGE)) |
4611 | error = -ENOTEMPTY; | |
1da177e4 LT |
4612 | if (new_dentry == trap) |
4613 | goto exit5; | |
4614 | ||
f5beed75 AV |
4615 | error = security_path_rename(&old_path, old_dentry, |
4616 | &new_path, new_dentry, flags); | |
be6d3e56 | 4617 | if (error) |
c30dabfe | 4618 | goto exit5; |
f5beed75 AV |
4619 | error = vfs_rename(old_path.dentry->d_inode, old_dentry, |
4620 | new_path.dentry->d_inode, new_dentry, | |
520c8b16 | 4621 | &delegated_inode, flags); |
1da177e4 LT |
4622 | exit5: |
4623 | dput(new_dentry); | |
4624 | exit4: | |
4625 | dput(old_dentry); | |
4626 | exit3: | |
f5beed75 | 4627 | unlock_rename(new_path.dentry, old_path.dentry); |
8e6d782c BF |
4628 | if (delegated_inode) { |
4629 | error = break_deleg_wait(&delegated_inode); | |
4630 | if (!error) | |
4631 | goto retry_deleg; | |
4632 | } | |
f5beed75 | 4633 | mnt_drop_write(old_path.mnt); |
1da177e4 | 4634 | exit2: |
c6a94284 JL |
4635 | if (retry_estale(error, lookup_flags)) |
4636 | should_retry = true; | |
f5beed75 | 4637 | path_put(&new_path); |
2ad94ae6 | 4638 | putname(to); |
1da177e4 | 4639 | exit1: |
f5beed75 | 4640 | path_put(&old_path); |
1da177e4 | 4641 | putname(from); |
c6a94284 JL |
4642 | if (should_retry) { |
4643 | should_retry = false; | |
4644 | lookup_flags |= LOOKUP_REVAL; | |
4645 | goto retry; | |
4646 | } | |
2ad94ae6 | 4647 | exit: |
1da177e4 LT |
4648 | return error; |
4649 | } | |
4650 | ||
ee81feb6 DB |
4651 | SYSCALL_DEFINE5(renameat2, int, olddfd, const char __user *, oldname, |
4652 | int, newdfd, const char __user *, newname, unsigned int, flags) | |
4653 | { | |
4654 | return do_renameat2(olddfd, oldname, newdfd, newname, flags); | |
4655 | } | |
4656 | ||
520c8b16 MS |
4657 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
4658 | int, newdfd, const char __user *, newname) | |
4659 | { | |
ee81feb6 | 4660 | return do_renameat2(olddfd, oldname, newdfd, newname, 0); |
520c8b16 MS |
4661 | } |
4662 | ||
a26eab24 | 4663 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4664 | { |
ee81feb6 | 4665 | return do_renameat2(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4666 | } |
4667 | ||
787fb6bc MS |
4668 | int vfs_whiteout(struct inode *dir, struct dentry *dentry) |
4669 | { | |
4670 | int error = may_create(dir, dentry); | |
4671 | if (error) | |
4672 | return error; | |
4673 | ||
4674 | if (!dir->i_op->mknod) | |
4675 | return -EPERM; | |
4676 | ||
4677 | return dir->i_op->mknod(dir, dentry, | |
4678 | S_IFCHR | WHITEOUT_MODE, WHITEOUT_DEV); | |
4679 | } | |
4680 | EXPORT_SYMBOL(vfs_whiteout); | |
4681 | ||
5d826c84 | 4682 | int readlink_copy(char __user *buffer, int buflen, const char *link) |
1da177e4 | 4683 | { |
5d826c84 | 4684 | int len = PTR_ERR(link); |
1da177e4 LT |
4685 | if (IS_ERR(link)) |
4686 | goto out; | |
4687 | ||
4688 | len = strlen(link); | |
4689 | if (len > (unsigned) buflen) | |
4690 | len = buflen; | |
4691 | if (copy_to_user(buffer, link, len)) | |
4692 | len = -EFAULT; | |
4693 | out: | |
4694 | return len; | |
4695 | } | |
4696 | ||
fd4a0edf MS |
4697 | /** |
4698 | * vfs_readlink - copy symlink body into userspace buffer | |
4699 | * @dentry: dentry on which to get symbolic link | |
4700 | * @buffer: user memory pointer | |
4701 | * @buflen: size of buffer | |
4702 | * | |
4703 | * Does not touch atime. That's up to the caller if necessary | |
4704 | * | |
4705 | * Does not call security hook. | |
4706 | */ | |
4707 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4708 | { | |
4709 | struct inode *inode = d_inode(dentry); | |
f2df5da6 AV |
4710 | DEFINE_DELAYED_CALL(done); |
4711 | const char *link; | |
4712 | int res; | |
fd4a0edf | 4713 | |
76fca90e MS |
4714 | if (unlikely(!(inode->i_opflags & IOP_DEFAULT_READLINK))) { |
4715 | if (unlikely(inode->i_op->readlink)) | |
4716 | return inode->i_op->readlink(dentry, buffer, buflen); | |
4717 | ||
4718 | if (!d_is_symlink(dentry)) | |
4719 | return -EINVAL; | |
4720 | ||
4721 | spin_lock(&inode->i_lock); | |
4722 | inode->i_opflags |= IOP_DEFAULT_READLINK; | |
4723 | spin_unlock(&inode->i_lock); | |
4724 | } | |
fd4a0edf | 4725 | |
4c4f7c19 | 4726 | link = READ_ONCE(inode->i_link); |
f2df5da6 AV |
4727 | if (!link) { |
4728 | link = inode->i_op->get_link(dentry, inode, &done); | |
4729 | if (IS_ERR(link)) | |
4730 | return PTR_ERR(link); | |
4731 | } | |
4732 | res = readlink_copy(buffer, buflen, link); | |
4733 | do_delayed_call(&done); | |
4734 | return res; | |
fd4a0edf MS |
4735 | } |
4736 | EXPORT_SYMBOL(vfs_readlink); | |
1da177e4 | 4737 | |
d60874cd MS |
4738 | /** |
4739 | * vfs_get_link - get symlink body | |
4740 | * @dentry: dentry on which to get symbolic link | |
4741 | * @done: caller needs to free returned data with this | |
4742 | * | |
4743 | * Calls security hook and i_op->get_link() on the supplied inode. | |
4744 | * | |
4745 | * It does not touch atime. That's up to the caller if necessary. | |
4746 | * | |
4747 | * Does not work on "special" symlinks like /proc/$$/fd/N | |
4748 | */ | |
4749 | const char *vfs_get_link(struct dentry *dentry, struct delayed_call *done) | |
4750 | { | |
4751 | const char *res = ERR_PTR(-EINVAL); | |
4752 | struct inode *inode = d_inode(dentry); | |
4753 | ||
4754 | if (d_is_symlink(dentry)) { | |
4755 | res = ERR_PTR(security_inode_readlink(dentry)); | |
4756 | if (!res) | |
4757 | res = inode->i_op->get_link(dentry, inode, done); | |
4758 | } | |
4759 | return res; | |
4760 | } | |
4761 | EXPORT_SYMBOL(vfs_get_link); | |
4762 | ||
1da177e4 | 4763 | /* get the link contents into pagecache */ |
6b255391 | 4764 | const char *page_get_link(struct dentry *dentry, struct inode *inode, |
fceef393 | 4765 | struct delayed_call *callback) |
1da177e4 | 4766 | { |
ebd09abb DG |
4767 | char *kaddr; |
4768 | struct page *page; | |
6b255391 AV |
4769 | struct address_space *mapping = inode->i_mapping; |
4770 | ||
d3883d4f AV |
4771 | if (!dentry) { |
4772 | page = find_get_page(mapping, 0); | |
4773 | if (!page) | |
4774 | return ERR_PTR(-ECHILD); | |
4775 | if (!PageUptodate(page)) { | |
4776 | put_page(page); | |
4777 | return ERR_PTR(-ECHILD); | |
4778 | } | |
4779 | } else { | |
4780 | page = read_mapping_page(mapping, 0, NULL); | |
4781 | if (IS_ERR(page)) | |
4782 | return (char*)page; | |
4783 | } | |
fceef393 | 4784 | set_delayed_call(callback, page_put_link, page); |
21fc61c7 AV |
4785 | BUG_ON(mapping_gfp_mask(mapping) & __GFP_HIGHMEM); |
4786 | kaddr = page_address(page); | |
6b255391 | 4787 | nd_terminate_link(kaddr, inode->i_size, PAGE_SIZE - 1); |
ebd09abb | 4788 | return kaddr; |
1da177e4 LT |
4789 | } |
4790 | ||
6b255391 | 4791 | EXPORT_SYMBOL(page_get_link); |
1da177e4 | 4792 | |
fceef393 | 4793 | void page_put_link(void *arg) |
1da177e4 | 4794 | { |
fceef393 | 4795 | put_page(arg); |
1da177e4 | 4796 | } |
4d359507 | 4797 | EXPORT_SYMBOL(page_put_link); |
1da177e4 | 4798 | |
aa80deab AV |
4799 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) |
4800 | { | |
fceef393 | 4801 | DEFINE_DELAYED_CALL(done); |
6b255391 AV |
4802 | int res = readlink_copy(buffer, buflen, |
4803 | page_get_link(dentry, d_inode(dentry), | |
fceef393 AV |
4804 | &done)); |
4805 | do_delayed_call(&done); | |
aa80deab AV |
4806 | return res; |
4807 | } | |
4808 | EXPORT_SYMBOL(page_readlink); | |
4809 | ||
54566b2c NP |
4810 | /* |
4811 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
4812 | */ | |
4813 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
4814 | { |
4815 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 4816 | struct page *page; |
afddba49 | 4817 | void *fsdata; |
beb497ab | 4818 | int err; |
c718a975 | 4819 | unsigned int flags = 0; |
54566b2c NP |
4820 | if (nofs) |
4821 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 4822 | |
7e53cac4 | 4823 | retry: |
afddba49 | 4824 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 4825 | flags, &page, &fsdata); |
1da177e4 | 4826 | if (err) |
afddba49 NP |
4827 | goto fail; |
4828 | ||
21fc61c7 | 4829 | memcpy(page_address(page), symname, len-1); |
afddba49 NP |
4830 | |
4831 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
4832 | page, fsdata); | |
1da177e4 LT |
4833 | if (err < 0) |
4834 | goto fail; | |
afddba49 NP |
4835 | if (err < len-1) |
4836 | goto retry; | |
4837 | ||
1da177e4 LT |
4838 | mark_inode_dirty(inode); |
4839 | return 0; | |
1da177e4 LT |
4840 | fail: |
4841 | return err; | |
4842 | } | |
4d359507 | 4843 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 | 4844 | |
0adb25d2 KK |
4845 | int page_symlink(struct inode *inode, const char *symname, int len) |
4846 | { | |
4847 | return __page_symlink(inode, symname, len, | |
c62d2555 | 4848 | !mapping_gfp_constraint(inode->i_mapping, __GFP_FS)); |
0adb25d2 | 4849 | } |
4d359507 | 4850 | EXPORT_SYMBOL(page_symlink); |
0adb25d2 | 4851 | |
92e1d5be | 4852 | const struct inode_operations page_symlink_inode_operations = { |
6b255391 | 4853 | .get_link = page_get_link, |
1da177e4 | 4854 | }; |
1da177e4 | 4855 | EXPORT_SYMBOL(page_symlink_inode_operations); |