]>
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); | |
102b8af2 AV |
599 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { |
600 | path_put(&nd->root); | |
601 | nd->root.mnt = NULL; | |
602 | } | |
7973387a AV |
603 | } else { |
604 | nd->flags &= ~LOOKUP_RCU; | |
605 | if (!(nd->flags & LOOKUP_ROOT)) | |
606 | nd->root.mnt = NULL; | |
607 | rcu_read_unlock(); | |
608 | } | |
609 | nd->depth = 0; | |
610 | } | |
611 | ||
612 | /* path_put is needed afterwards regardless of success or failure */ | |
613 | static bool legitimize_path(struct nameidata *nd, | |
614 | struct path *path, unsigned seq) | |
615 | { | |
616 | int res = __legitimize_mnt(path->mnt, nd->m_seq); | |
617 | if (unlikely(res)) { | |
618 | if (res > 0) | |
619 | path->mnt = NULL; | |
620 | path->dentry = NULL; | |
621 | return false; | |
622 | } | |
623 | if (unlikely(!lockref_get_not_dead(&path->dentry->d_lockref))) { | |
624 | path->dentry = NULL; | |
625 | return false; | |
626 | } | |
627 | return !read_seqcount_retry(&path->dentry->d_seq, seq); | |
628 | } | |
629 | ||
630 | static bool legitimize_links(struct nameidata *nd) | |
631 | { | |
632 | int i; | |
633 | for (i = 0; i < nd->depth; i++) { | |
634 | struct saved *last = nd->stack + i; | |
635 | if (unlikely(!legitimize_path(nd, &last->link, last->seq))) { | |
636 | drop_links(nd); | |
637 | nd->depth = i + 1; | |
638 | return false; | |
639 | } | |
640 | } | |
641 | return true; | |
642 | } | |
643 | ||
19660af7 | 644 | /* |
31e6b01f | 645 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
646 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
647 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
57e3715c | 648 | * normal reference counts on dentries and vfsmounts to transition to ref-walk |
19660af7 AV |
649 | * mode. Refcounts are grabbed at the last known good point before rcu-walk |
650 | * got stuck, so ref-walk may continue from there. If this is not successful | |
651 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
652 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 653 | */ |
31e6b01f NP |
654 | |
655 | /** | |
19660af7 AV |
656 | * unlazy_walk - try to switch to ref-walk mode. |
657 | * @nd: nameidata pathwalk data | |
39191628 | 658 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 659 | * |
4675ac39 AV |
660 | * unlazy_walk attempts to legitimize the current nd->path and nd->root |
661 | * for ref-walk mode. | |
662 | * Must be called from rcu-walk context. | |
7973387a AV |
663 | * Nothing should touch nameidata between unlazy_walk() failure and |
664 | * terminate_walk(). | |
31e6b01f | 665 | */ |
4675ac39 | 666 | static int unlazy_walk(struct nameidata *nd) |
31e6b01f | 667 | { |
31e6b01f NP |
668 | struct dentry *parent = nd->path.dentry; |
669 | ||
670 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
e5c832d5 | 671 | |
4675ac39 AV |
672 | nd->flags &= ~LOOKUP_RCU; |
673 | if (unlikely(!legitimize_links(nd))) | |
674 | goto out2; | |
675 | if (unlikely(!legitimize_path(nd, &nd->path, nd->seq))) | |
676 | goto out1; | |
677 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
678 | if (unlikely(!legitimize_path(nd, &nd->root, nd->root_seq))) | |
679 | goto out; | |
680 | } | |
681 | rcu_read_unlock(); | |
682 | BUG_ON(nd->inode != parent->d_inode); | |
683 | return 0; | |
684 | ||
685 | out2: | |
686 | nd->path.mnt = NULL; | |
687 | nd->path.dentry = NULL; | |
688 | out1: | |
689 | if (!(nd->flags & LOOKUP_ROOT)) | |
690 | nd->root.mnt = NULL; | |
691 | out: | |
692 | rcu_read_unlock(); | |
693 | return -ECHILD; | |
694 | } | |
695 | ||
696 | /** | |
697 | * unlazy_child - try to switch to ref-walk mode. | |
698 | * @nd: nameidata pathwalk data | |
699 | * @dentry: child of nd->path.dentry | |
700 | * @seq: seq number to check dentry against | |
701 | * Returns: 0 on success, -ECHILD on failure | |
702 | * | |
703 | * unlazy_child attempts to legitimize the current nd->path, nd->root and dentry | |
704 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
705 | * @nd. Must be called from rcu-walk context. | |
706 | * Nothing should touch nameidata between unlazy_child() failure and | |
707 | * terminate_walk(). | |
708 | */ | |
709 | static int unlazy_child(struct nameidata *nd, struct dentry *dentry, unsigned seq) | |
710 | { | |
711 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
712 | ||
e5c832d5 | 713 | nd->flags &= ~LOOKUP_RCU; |
7973387a AV |
714 | if (unlikely(!legitimize_links(nd))) |
715 | goto out2; | |
716 | if (unlikely(!legitimize_mnt(nd->path.mnt, nd->m_seq))) | |
717 | goto out2; | |
4675ac39 | 718 | if (unlikely(!lockref_get_not_dead(&nd->path.dentry->d_lockref))) |
7973387a | 719 | goto out1; |
48a066e7 | 720 | |
15570086 | 721 | /* |
4675ac39 AV |
722 | * We need to move both the parent and the dentry from the RCU domain |
723 | * to be properly refcounted. And the sequence number in the dentry | |
724 | * validates *both* dentry counters, since we checked the sequence | |
725 | * number of the parent after we got the child sequence number. So we | |
726 | * know the parent must still be valid if the child sequence number is | |
15570086 | 727 | */ |
4675ac39 AV |
728 | if (unlikely(!lockref_get_not_dead(&dentry->d_lockref))) |
729 | goto out; | |
730 | if (unlikely(read_seqcount_retry(&dentry->d_seq, seq))) { | |
731 | rcu_read_unlock(); | |
732 | dput(dentry); | |
733 | goto drop_root_mnt; | |
19660af7 | 734 | } |
e5c832d5 LT |
735 | /* |
736 | * Sequence counts matched. Now make sure that the root is | |
737 | * still valid and get it if required. | |
738 | */ | |
739 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
5a8d87e8 AV |
740 | if (unlikely(!legitimize_path(nd, &nd->root, nd->root_seq))) { |
741 | rcu_read_unlock(); | |
742 | dput(dentry); | |
743 | return -ECHILD; | |
7973387a | 744 | } |
31e6b01f | 745 | } |
31e6b01f | 746 | |
8b61e74f | 747 | rcu_read_unlock(); |
31e6b01f | 748 | return 0; |
19660af7 | 749 | |
7973387a AV |
750 | out2: |
751 | nd->path.mnt = NULL; | |
752 | out1: | |
753 | nd->path.dentry = NULL; | |
e5c832d5 | 754 | out: |
8b61e74f | 755 | rcu_read_unlock(); |
d0d27277 LT |
756 | drop_root_mnt: |
757 | if (!(nd->flags & LOOKUP_ROOT)) | |
758 | nd->root.mnt = NULL; | |
31e6b01f NP |
759 | return -ECHILD; |
760 | } | |
761 | ||
4ce16ef3 | 762 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 763 | { |
a89f8337 AV |
764 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) |
765 | return dentry->d_op->d_revalidate(dentry, flags); | |
766 | else | |
767 | return 1; | |
34286d66 NP |
768 | } |
769 | ||
9f1fafee AV |
770 | /** |
771 | * complete_walk - successful completion of path walk | |
772 | * @nd: pointer nameidata | |
39159de2 | 773 | * |
9f1fafee AV |
774 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
775 | * Revalidate the final result, unless we'd already done that during | |
776 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
777 | * success, -error on failure. In case of failure caller does not | |
778 | * need to drop nd->path. | |
39159de2 | 779 | */ |
9f1fafee | 780 | static int complete_walk(struct nameidata *nd) |
39159de2 | 781 | { |
16c2cd71 | 782 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 783 | int status; |
39159de2 | 784 | |
9f1fafee | 785 | if (nd->flags & LOOKUP_RCU) { |
9f1fafee AV |
786 | if (!(nd->flags & LOOKUP_ROOT)) |
787 | nd->root.mnt = NULL; | |
4675ac39 | 788 | if (unlikely(unlazy_walk(nd))) |
9f1fafee | 789 | return -ECHILD; |
9f1fafee AV |
790 | } |
791 | ||
16c2cd71 AV |
792 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
793 | return 0; | |
794 | ||
ecf3d1f1 | 795 | if (likely(!(dentry->d_flags & DCACHE_OP_WEAK_REVALIDATE))) |
39159de2 JL |
796 | return 0; |
797 | ||
ecf3d1f1 | 798 | status = dentry->d_op->d_weak_revalidate(dentry, nd->flags); |
39159de2 JL |
799 | if (status > 0) |
800 | return 0; | |
801 | ||
16c2cd71 | 802 | if (!status) |
39159de2 | 803 | status = -ESTALE; |
16c2cd71 | 804 | |
39159de2 JL |
805 | return status; |
806 | } | |
807 | ||
18d8c860 | 808 | static void set_root(struct nameidata *nd) |
31e6b01f | 809 | { |
7bd88377 | 810 | struct fs_struct *fs = current->fs; |
c28cc364 | 811 | |
9e6697e2 AV |
812 | if (nd->flags & LOOKUP_RCU) { |
813 | unsigned seq; | |
814 | ||
815 | do { | |
816 | seq = read_seqcount_begin(&fs->seq); | |
817 | nd->root = fs->root; | |
818 | nd->root_seq = __read_seqcount_begin(&nd->root.dentry->d_seq); | |
819 | } while (read_seqcount_retry(&fs->seq, seq)); | |
820 | } else { | |
821 | get_fs_root(fs, &nd->root); | |
822 | } | |
31e6b01f NP |
823 | } |
824 | ||
1d957f9b | 825 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
826 | { |
827 | dput(path->dentry); | |
4ac91378 | 828 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
829 | mntput(path->mnt); |
830 | } | |
831 | ||
7b9337aa NP |
832 | static inline void path_to_nameidata(const struct path *path, |
833 | struct nameidata *nd) | |
051d3812 | 834 | { |
31e6b01f NP |
835 | if (!(nd->flags & LOOKUP_RCU)) { |
836 | dput(nd->path.dentry); | |
837 | if (nd->path.mnt != path->mnt) | |
838 | mntput(nd->path.mnt); | |
9a229683 | 839 | } |
31e6b01f | 840 | nd->path.mnt = path->mnt; |
4ac91378 | 841 | nd->path.dentry = path->dentry; |
051d3812 IK |
842 | } |
843 | ||
248fb5b9 AV |
844 | static int nd_jump_root(struct nameidata *nd) |
845 | { | |
846 | if (nd->flags & LOOKUP_RCU) { | |
847 | struct dentry *d; | |
848 | nd->path = nd->root; | |
849 | d = nd->path.dentry; | |
850 | nd->inode = d->d_inode; | |
851 | nd->seq = nd->root_seq; | |
852 | if (unlikely(read_seqcount_retry(&d->d_seq, nd->seq))) | |
853 | return -ECHILD; | |
854 | } else { | |
855 | path_put(&nd->path); | |
856 | nd->path = nd->root; | |
857 | path_get(&nd->path); | |
858 | nd->inode = nd->path.dentry->d_inode; | |
859 | } | |
860 | nd->flags |= LOOKUP_JUMPED; | |
861 | return 0; | |
862 | } | |
863 | ||
b5fb63c1 | 864 | /* |
6b255391 | 865 | * Helper to directly jump to a known parsed path from ->get_link, |
b5fb63c1 CH |
866 | * caller must have taken a reference to path beforehand. |
867 | */ | |
6e77137b | 868 | void nd_jump_link(struct path *path) |
b5fb63c1 | 869 | { |
6e77137b | 870 | struct nameidata *nd = current->nameidata; |
b5fb63c1 CH |
871 | path_put(&nd->path); |
872 | ||
873 | nd->path = *path; | |
874 | nd->inode = nd->path.dentry->d_inode; | |
875 | nd->flags |= LOOKUP_JUMPED; | |
b5fb63c1 CH |
876 | } |
877 | ||
b9ff4429 | 878 | static inline void put_link(struct nameidata *nd) |
574197e0 | 879 | { |
21c3003d | 880 | struct saved *last = nd->stack + --nd->depth; |
fceef393 | 881 | do_delayed_call(&last->done); |
6548fae2 AV |
882 | if (!(nd->flags & LOOKUP_RCU)) |
883 | path_put(&last->link); | |
574197e0 AV |
884 | } |
885 | ||
561ec64a LT |
886 | int sysctl_protected_symlinks __read_mostly = 0; |
887 | int sysctl_protected_hardlinks __read_mostly = 0; | |
30aba665 SM |
888 | int sysctl_protected_fifos __read_mostly; |
889 | int sysctl_protected_regular __read_mostly; | |
800179c9 KC |
890 | |
891 | /** | |
892 | * may_follow_link - Check symlink following for unsafe situations | |
55852635 | 893 | * @nd: nameidata pathwalk data |
800179c9 KC |
894 | * |
895 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
896 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
897 | * in a sticky world-writable directory. This is to protect privileged | |
898 | * processes from failing races against path names that may change out | |
899 | * from under them by way of other users creating malicious symlinks. | |
900 | * It will permit symlinks to be followed only when outside a sticky | |
901 | * world-writable directory, or when the uid of the symlink and follower | |
902 | * match, or when the directory owner matches the symlink's owner. | |
903 | * | |
904 | * Returns 0 if following the symlink is allowed, -ve on error. | |
905 | */ | |
fec2fa24 | 906 | static inline int may_follow_link(struct nameidata *nd) |
800179c9 KC |
907 | { |
908 | const struct inode *inode; | |
909 | const struct inode *parent; | |
2d7f9e2a | 910 | kuid_t puid; |
800179c9 KC |
911 | |
912 | if (!sysctl_protected_symlinks) | |
913 | return 0; | |
914 | ||
915 | /* Allowed if owner and follower match. */ | |
fceef393 | 916 | inode = nd->link_inode; |
81abe27b | 917 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
918 | return 0; |
919 | ||
920 | /* Allowed if parent directory not sticky and world-writable. */ | |
aa65fa35 | 921 | parent = nd->inode; |
800179c9 KC |
922 | if ((parent->i_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) |
923 | return 0; | |
924 | ||
925 | /* Allowed if parent directory and link owner match. */ | |
2d7f9e2a SF |
926 | puid = parent->i_uid; |
927 | if (uid_valid(puid) && uid_eq(puid, inode->i_uid)) | |
800179c9 KC |
928 | return 0; |
929 | ||
31956502 AV |
930 | if (nd->flags & LOOKUP_RCU) |
931 | return -ECHILD; | |
932 | ||
ea841baf | 933 | audit_inode(nd->name, nd->stack[0].link.dentry, 0); |
94b9d9b7 | 934 | audit_log_link_denied("follow_link"); |
800179c9 KC |
935 | return -EACCES; |
936 | } | |
937 | ||
938 | /** | |
939 | * safe_hardlink_source - Check for safe hardlink conditions | |
940 | * @inode: the source inode to hardlink from | |
941 | * | |
942 | * Return false if at least one of the following conditions: | |
943 | * - inode is not a regular file | |
944 | * - inode is setuid | |
945 | * - inode is setgid and group-exec | |
946 | * - access failure for read and write | |
947 | * | |
948 | * Otherwise returns true. | |
949 | */ | |
950 | static bool safe_hardlink_source(struct inode *inode) | |
951 | { | |
952 | umode_t mode = inode->i_mode; | |
953 | ||
954 | /* Special files should not get pinned to the filesystem. */ | |
955 | if (!S_ISREG(mode)) | |
956 | return false; | |
957 | ||
958 | /* Setuid files should not get pinned to the filesystem. */ | |
959 | if (mode & S_ISUID) | |
960 | return false; | |
961 | ||
962 | /* Executable setgid files should not get pinned to the filesystem. */ | |
963 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
964 | return false; | |
965 | ||
966 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
967 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
968 | return false; | |
969 | ||
970 | return true; | |
971 | } | |
972 | ||
973 | /** | |
974 | * may_linkat - Check permissions for creating a hardlink | |
975 | * @link: the source to hardlink from | |
976 | * | |
977 | * Block hardlink when all of: | |
978 | * - sysctl_protected_hardlinks enabled | |
979 | * - fsuid does not match inode | |
980 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
f2ca3796 | 981 | * - not CAP_FOWNER in a namespace with the inode owner uid mapped |
800179c9 KC |
982 | * |
983 | * Returns 0 if successful, -ve on error. | |
984 | */ | |
985 | static int may_linkat(struct path *link) | |
986 | { | |
593d1ce8 EB |
987 | struct inode *inode = link->dentry->d_inode; |
988 | ||
989 | /* Inode writeback is not safe when the uid or gid are invalid. */ | |
990 | if (!uid_valid(inode->i_uid) || !gid_valid(inode->i_gid)) | |
991 | return -EOVERFLOW; | |
800179c9 KC |
992 | |
993 | if (!sysctl_protected_hardlinks) | |
994 | return 0; | |
995 | ||
800179c9 KC |
996 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, |
997 | * otherwise, it must be a safe source. | |
998 | */ | |
cc658db4 | 999 | if (safe_hardlink_source(inode) || inode_owner_or_capable(inode)) |
800179c9 KC |
1000 | return 0; |
1001 | ||
94b9d9b7 | 1002 | audit_log_link_denied("linkat"); |
800179c9 KC |
1003 | return -EPERM; |
1004 | } | |
1005 | ||
30aba665 SM |
1006 | /** |
1007 | * may_create_in_sticky - Check whether an O_CREAT open in a sticky directory | |
1008 | * should be allowed, or not, on files that already | |
1009 | * exist. | |
1010 | * @dir: the sticky parent directory | |
1011 | * @inode: the inode of the file to open | |
1012 | * | |
1013 | * Block an O_CREAT open of a FIFO (or a regular file) when: | |
1014 | * - sysctl_protected_fifos (or sysctl_protected_regular) is enabled | |
1015 | * - the file already exists | |
1016 | * - we are in a sticky directory | |
1017 | * - we don't own the file | |
1018 | * - the owner of the directory doesn't own the file | |
1019 | * - the directory is world writable | |
1020 | * If the sysctl_protected_fifos (or sysctl_protected_regular) is set to 2 | |
1021 | * the directory doesn't have to be world writable: being group writable will | |
1022 | * be enough. | |
1023 | * | |
1024 | * Returns 0 if the open is allowed, -ve on error. | |
1025 | */ | |
1026 | static int may_create_in_sticky(struct dentry * const dir, | |
1027 | struct inode * const inode) | |
1028 | { | |
1029 | if ((!sysctl_protected_fifos && S_ISFIFO(inode->i_mode)) || | |
1030 | (!sysctl_protected_regular && S_ISREG(inode->i_mode)) || | |
1031 | likely(!(dir->d_inode->i_mode & S_ISVTX)) || | |
1032 | uid_eq(inode->i_uid, dir->d_inode->i_uid) || | |
1033 | uid_eq(current_fsuid(), inode->i_uid)) | |
1034 | return 0; | |
1035 | ||
1036 | if (likely(dir->d_inode->i_mode & 0002) || | |
1037 | (dir->d_inode->i_mode & 0020 && | |
1038 | ((sysctl_protected_fifos >= 2 && S_ISFIFO(inode->i_mode)) || | |
1039 | (sysctl_protected_regular >= 2 && S_ISREG(inode->i_mode))))) { | |
1040 | return -EACCES; | |
1041 | } | |
1042 | return 0; | |
1043 | } | |
1044 | ||
3b2e7f75 AV |
1045 | static __always_inline |
1046 | const char *get_link(struct nameidata *nd) | |
1da177e4 | 1047 | { |
ab104923 | 1048 | struct saved *last = nd->stack + nd->depth - 1; |
1cf2665b | 1049 | struct dentry *dentry = last->link.dentry; |
fceef393 | 1050 | struct inode *inode = nd->link_inode; |
6d7b5aae | 1051 | int error; |
0a959df5 | 1052 | const char *res; |
1da177e4 | 1053 | |
8fa9dd24 N |
1054 | if (!(nd->flags & LOOKUP_RCU)) { |
1055 | touch_atime(&last->link); | |
1056 | cond_resched(); | |
c6718543 | 1057 | } else if (atime_needs_update(&last->link, inode)) { |
4675ac39 | 1058 | if (unlikely(unlazy_walk(nd))) |
bc40aee0 | 1059 | return ERR_PTR(-ECHILD); |
8fa9dd24 | 1060 | touch_atime(&last->link); |
bc40aee0 | 1061 | } |
574197e0 | 1062 | |
bda0be7a N |
1063 | error = security_inode_follow_link(dentry, inode, |
1064 | nd->flags & LOOKUP_RCU); | |
1065 | if (unlikely(error)) | |
6920a440 | 1066 | return ERR_PTR(error); |
36f3b4f6 | 1067 | |
86acdca1 | 1068 | nd->last_type = LAST_BIND; |
d4dee48b AV |
1069 | res = inode->i_link; |
1070 | if (!res) { | |
fceef393 AV |
1071 | const char * (*get)(struct dentry *, struct inode *, |
1072 | struct delayed_call *); | |
1073 | get = inode->i_op->get_link; | |
8c1b4566 | 1074 | if (nd->flags & LOOKUP_RCU) { |
fceef393 | 1075 | res = get(NULL, inode, &last->done); |
6b255391 | 1076 | if (res == ERR_PTR(-ECHILD)) { |
4675ac39 | 1077 | if (unlikely(unlazy_walk(nd))) |
6b255391 | 1078 | return ERR_PTR(-ECHILD); |
fceef393 | 1079 | res = get(dentry, inode, &last->done); |
6b255391 AV |
1080 | } |
1081 | } else { | |
fceef393 | 1082 | res = get(dentry, inode, &last->done); |
8c1b4566 | 1083 | } |
fceef393 | 1084 | if (IS_ERR_OR_NULL(res)) |
fab51e8a | 1085 | return res; |
fab51e8a AV |
1086 | } |
1087 | if (*res == '/') { | |
9e6697e2 AV |
1088 | if (!nd->root.mnt) |
1089 | set_root(nd); | |
248fb5b9 AV |
1090 | if (unlikely(nd_jump_root(nd))) |
1091 | return ERR_PTR(-ECHILD); | |
fab51e8a AV |
1092 | while (unlikely(*++res == '/')) |
1093 | ; | |
1da177e4 | 1094 | } |
fab51e8a AV |
1095 | if (!*res) |
1096 | res = NULL; | |
0a959df5 AV |
1097 | return res; |
1098 | } | |
6d7b5aae | 1099 | |
f015f126 DH |
1100 | /* |
1101 | * follow_up - Find the mountpoint of path's vfsmount | |
1102 | * | |
1103 | * Given a path, find the mountpoint of its source file system. | |
1104 | * Replace @path with the path of the mountpoint in the parent mount. | |
1105 | * Up is towards /. | |
1106 | * | |
1107 | * Return 1 if we went up a level and 0 if we were already at the | |
1108 | * root. | |
1109 | */ | |
bab77ebf | 1110 | int follow_up(struct path *path) |
1da177e4 | 1111 | { |
0714a533 AV |
1112 | struct mount *mnt = real_mount(path->mnt); |
1113 | struct mount *parent; | |
1da177e4 | 1114 | struct dentry *mountpoint; |
99b7db7b | 1115 | |
48a066e7 | 1116 | read_seqlock_excl(&mount_lock); |
0714a533 | 1117 | parent = mnt->mnt_parent; |
3c0a6163 | 1118 | if (parent == mnt) { |
48a066e7 | 1119 | read_sequnlock_excl(&mount_lock); |
1da177e4 LT |
1120 | return 0; |
1121 | } | |
0714a533 | 1122 | mntget(&parent->mnt); |
a73324da | 1123 | mountpoint = dget(mnt->mnt_mountpoint); |
48a066e7 | 1124 | read_sequnlock_excl(&mount_lock); |
bab77ebf AV |
1125 | dput(path->dentry); |
1126 | path->dentry = mountpoint; | |
1127 | mntput(path->mnt); | |
0714a533 | 1128 | path->mnt = &parent->mnt; |
1da177e4 LT |
1129 | return 1; |
1130 | } | |
4d359507 | 1131 | EXPORT_SYMBOL(follow_up); |
1da177e4 | 1132 | |
b5c84bf6 | 1133 | /* |
9875cf80 DH |
1134 | * Perform an automount |
1135 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
1136 | * were called with. | |
1da177e4 | 1137 | */ |
756daf26 | 1138 | static int follow_automount(struct path *path, struct nameidata *nd, |
9875cf80 | 1139 | bool *need_mntput) |
31e6b01f | 1140 | { |
9875cf80 | 1141 | struct vfsmount *mnt; |
ea5b778a | 1142 | int err; |
9875cf80 DH |
1143 | |
1144 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
1145 | return -EREMOTE; | |
1146 | ||
0ec26fd0 MS |
1147 | /* We don't want to mount if someone's just doing a stat - |
1148 | * unless they're stat'ing a directory and appended a '/' to | |
1149 | * the name. | |
1150 | * | |
1151 | * We do, however, want to mount if someone wants to open or | |
1152 | * create a file of any type under the mountpoint, wants to | |
1153 | * traverse through the mountpoint or wants to open the | |
1154 | * mounted directory. Also, autofs may mark negative dentries | |
1155 | * as being automount points. These will need the attentions | |
1156 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 1157 | */ |
756daf26 | 1158 | if (!(nd->flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
5d38f049 IK |
1159 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && |
1160 | path->dentry->d_inode) | |
1161 | return -EISDIR; | |
0ec26fd0 | 1162 | |
756daf26 N |
1163 | nd->total_link_count++; |
1164 | if (nd->total_link_count >= 40) | |
9875cf80 DH |
1165 | return -ELOOP; |
1166 | ||
1167 | mnt = path->dentry->d_op->d_automount(path); | |
1168 | if (IS_ERR(mnt)) { | |
1169 | /* | |
1170 | * The filesystem is allowed to return -EISDIR here to indicate | |
1171 | * it doesn't want to automount. For instance, autofs would do | |
1172 | * this so that its userspace daemon can mount on this dentry. | |
1173 | * | |
1174 | * However, we can only permit this if it's a terminal point in | |
1175 | * the path being looked up; if it wasn't then the remainder of | |
1176 | * the path is inaccessible and we should say so. | |
1177 | */ | |
756daf26 | 1178 | if (PTR_ERR(mnt) == -EISDIR && (nd->flags & LOOKUP_PARENT)) |
9875cf80 DH |
1179 | return -EREMOTE; |
1180 | return PTR_ERR(mnt); | |
31e6b01f | 1181 | } |
ea5b778a | 1182 | |
9875cf80 DH |
1183 | if (!mnt) /* mount collision */ |
1184 | return 0; | |
31e6b01f | 1185 | |
8aef1884 AV |
1186 | if (!*need_mntput) { |
1187 | /* lock_mount() may release path->mnt on error */ | |
1188 | mntget(path->mnt); | |
1189 | *need_mntput = true; | |
1190 | } | |
19a167af | 1191 | err = finish_automount(mnt, path); |
9875cf80 | 1192 | |
ea5b778a DH |
1193 | switch (err) { |
1194 | case -EBUSY: | |
1195 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 1196 | return 0; |
ea5b778a | 1197 | case 0: |
8aef1884 | 1198 | path_put(path); |
ea5b778a DH |
1199 | path->mnt = mnt; |
1200 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 1201 | return 0; |
19a167af AV |
1202 | default: |
1203 | return err; | |
ea5b778a | 1204 | } |
19a167af | 1205 | |
463ffb2e AV |
1206 | } |
1207 | ||
9875cf80 DH |
1208 | /* |
1209 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 1210 | * - Flagged for transit management (autofs) |
9875cf80 DH |
1211 | * - Flagged as mountpoint |
1212 | * - Flagged as automount point | |
1213 | * | |
1214 | * This may only be called in refwalk mode. | |
1215 | * | |
1216 | * Serialization is taken care of in namespace.c | |
1217 | */ | |
756daf26 | 1218 | static int follow_managed(struct path *path, struct nameidata *nd) |
1da177e4 | 1219 | { |
8aef1884 | 1220 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
1221 | unsigned managed; |
1222 | bool need_mntput = false; | |
8aef1884 | 1223 | int ret = 0; |
9875cf80 DH |
1224 | |
1225 | /* Given that we're not holding a lock here, we retain the value in a | |
1226 | * local variable for each dentry as we look at it so that we don't see | |
1227 | * the components of that value change under us */ | |
6aa7de05 | 1228 | while (managed = READ_ONCE(path->dentry->d_flags), |
9875cf80 DH |
1229 | managed &= DCACHE_MANAGED_DENTRY, |
1230 | unlikely(managed != 0)) { | |
cc53ce53 DH |
1231 | /* Allow the filesystem to manage the transit without i_mutex |
1232 | * being held. */ | |
1233 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1234 | BUG_ON(!path->dentry->d_op); | |
1235 | BUG_ON(!path->dentry->d_op->d_manage); | |
fb5f51c7 | 1236 | ret = path->dentry->d_op->d_manage(path, false); |
cc53ce53 | 1237 | if (ret < 0) |
8aef1884 | 1238 | break; |
cc53ce53 DH |
1239 | } |
1240 | ||
9875cf80 DH |
1241 | /* Transit to a mounted filesystem. */ |
1242 | if (managed & DCACHE_MOUNTED) { | |
1243 | struct vfsmount *mounted = lookup_mnt(path); | |
1244 | if (mounted) { | |
1245 | dput(path->dentry); | |
1246 | if (need_mntput) | |
1247 | mntput(path->mnt); | |
1248 | path->mnt = mounted; | |
1249 | path->dentry = dget(mounted->mnt_root); | |
1250 | need_mntput = true; | |
1251 | continue; | |
1252 | } | |
1253 | ||
1254 | /* Something is mounted on this dentry in another | |
1255 | * namespace and/or whatever was mounted there in this | |
48a066e7 AV |
1256 | * namespace got unmounted before lookup_mnt() could |
1257 | * get it */ | |
9875cf80 DH |
1258 | } |
1259 | ||
1260 | /* Handle an automount point */ | |
1261 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
756daf26 | 1262 | ret = follow_automount(path, nd, &need_mntput); |
9875cf80 | 1263 | if (ret < 0) |
8aef1884 | 1264 | break; |
9875cf80 DH |
1265 | continue; |
1266 | } | |
1267 | ||
1268 | /* We didn't change the current path point */ | |
1269 | break; | |
1da177e4 | 1270 | } |
8aef1884 AV |
1271 | |
1272 | if (need_mntput && path->mnt == mnt) | |
1273 | mntput(path->mnt); | |
e9742b53 AV |
1274 | if (ret == -EISDIR || !ret) |
1275 | ret = 1; | |
8402752e AV |
1276 | if (need_mntput) |
1277 | nd->flags |= LOOKUP_JUMPED; | |
1278 | if (unlikely(ret < 0)) | |
1279 | path_put_conditional(path, nd); | |
1280 | return ret; | |
1da177e4 LT |
1281 | } |
1282 | ||
cc53ce53 | 1283 | int follow_down_one(struct path *path) |
1da177e4 LT |
1284 | { |
1285 | struct vfsmount *mounted; | |
1286 | ||
1c755af4 | 1287 | mounted = lookup_mnt(path); |
1da177e4 | 1288 | if (mounted) { |
9393bd07 AV |
1289 | dput(path->dentry); |
1290 | mntput(path->mnt); | |
1291 | path->mnt = mounted; | |
1292 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1293 | return 1; |
1294 | } | |
1295 | return 0; | |
1296 | } | |
4d359507 | 1297 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 | 1298 | |
fb5f51c7 | 1299 | static inline int managed_dentry_rcu(const struct path *path) |
62a7375e | 1300 | { |
fb5f51c7 IK |
1301 | return (path->dentry->d_flags & DCACHE_MANAGE_TRANSIT) ? |
1302 | path->dentry->d_op->d_manage(path, true) : 0; | |
62a7375e IK |
1303 | } |
1304 | ||
9875cf80 | 1305 | /* |
287548e4 AV |
1306 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1307 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1308 | */ |
1309 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
254cf582 | 1310 | struct inode **inode, unsigned *seqp) |
9875cf80 | 1311 | { |
62a7375e | 1312 | for (;;) { |
c7105365 | 1313 | struct mount *mounted; |
62a7375e IK |
1314 | /* |
1315 | * Don't forget we might have a non-mountpoint managed dentry | |
1316 | * that wants to block transit. | |
1317 | */ | |
fb5f51c7 | 1318 | switch (managed_dentry_rcu(path)) { |
b8faf035 N |
1319 | case -ECHILD: |
1320 | default: | |
ab90911f | 1321 | return false; |
b8faf035 N |
1322 | case -EISDIR: |
1323 | return true; | |
1324 | case 0: | |
1325 | break; | |
1326 | } | |
62a7375e IK |
1327 | |
1328 | if (!d_mountpoint(path->dentry)) | |
b8faf035 | 1329 | return !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
62a7375e | 1330 | |
474279dc | 1331 | mounted = __lookup_mnt(path->mnt, path->dentry); |
9875cf80 DH |
1332 | if (!mounted) |
1333 | break; | |
c7105365 AV |
1334 | path->mnt = &mounted->mnt; |
1335 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1336 | nd->flags |= LOOKUP_JUMPED; |
254cf582 | 1337 | *seqp = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1338 | /* |
1339 | * Update the inode too. We don't need to re-check the | |
1340 | * dentry sequence number here after this d_inode read, | |
1341 | * because a mount-point is always pinned. | |
1342 | */ | |
1343 | *inode = path->dentry->d_inode; | |
9875cf80 | 1344 | } |
f5be3e29 | 1345 | return !read_seqretry(&mount_lock, nd->m_seq) && |
b8faf035 | 1346 | !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
287548e4 AV |
1347 | } |
1348 | ||
31e6b01f NP |
1349 | static int follow_dotdot_rcu(struct nameidata *nd) |
1350 | { | |
4023bfc9 | 1351 | struct inode *inode = nd->inode; |
31e6b01f | 1352 | |
9875cf80 | 1353 | while (1) { |
aed434ad | 1354 | if (path_equal(&nd->path, &nd->root)) |
31e6b01f | 1355 | break; |
31e6b01f NP |
1356 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
1357 | struct dentry *old = nd->path.dentry; | |
1358 | struct dentry *parent = old->d_parent; | |
1359 | unsigned seq; | |
1360 | ||
4023bfc9 | 1361 | inode = parent->d_inode; |
31e6b01f | 1362 | seq = read_seqcount_begin(&parent->d_seq); |
aed434ad AV |
1363 | if (unlikely(read_seqcount_retry(&old->d_seq, nd->seq))) |
1364 | return -ECHILD; | |
31e6b01f NP |
1365 | nd->path.dentry = parent; |
1366 | nd->seq = seq; | |
397d425d EB |
1367 | if (unlikely(!path_connected(&nd->path))) |
1368 | return -ENOENT; | |
31e6b01f | 1369 | break; |
aed434ad AV |
1370 | } else { |
1371 | struct mount *mnt = real_mount(nd->path.mnt); | |
1372 | struct mount *mparent = mnt->mnt_parent; | |
1373 | struct dentry *mountpoint = mnt->mnt_mountpoint; | |
1374 | struct inode *inode2 = mountpoint->d_inode; | |
1375 | unsigned seq = read_seqcount_begin(&mountpoint->d_seq); | |
1376 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) | |
1377 | return -ECHILD; | |
1378 | if (&mparent->mnt == nd->path.mnt) | |
1379 | break; | |
1380 | /* we know that mountpoint was pinned */ | |
1381 | nd->path.dentry = mountpoint; | |
1382 | nd->path.mnt = &mparent->mnt; | |
1383 | inode = inode2; | |
1384 | nd->seq = seq; | |
31e6b01f | 1385 | } |
31e6b01f | 1386 | } |
aed434ad | 1387 | while (unlikely(d_mountpoint(nd->path.dentry))) { |
b37199e6 AV |
1388 | struct mount *mounted; |
1389 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry); | |
aed434ad AV |
1390 | if (unlikely(read_seqretry(&mount_lock, nd->m_seq))) |
1391 | return -ECHILD; | |
b37199e6 AV |
1392 | if (!mounted) |
1393 | break; | |
1394 | nd->path.mnt = &mounted->mnt; | |
1395 | nd->path.dentry = mounted->mnt.mnt_root; | |
4023bfc9 | 1396 | inode = nd->path.dentry->d_inode; |
b37199e6 | 1397 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
b37199e6 | 1398 | } |
4023bfc9 | 1399 | nd->inode = inode; |
31e6b01f NP |
1400 | return 0; |
1401 | } | |
1402 | ||
cc53ce53 DH |
1403 | /* |
1404 | * Follow down to the covering mount currently visible to userspace. At each | |
1405 | * point, the filesystem owning that dentry may be queried as to whether the | |
1406 | * caller is permitted to proceed or not. | |
cc53ce53 | 1407 | */ |
7cc90cc3 | 1408 | int follow_down(struct path *path) |
cc53ce53 DH |
1409 | { |
1410 | unsigned managed; | |
1411 | int ret; | |
1412 | ||
6aa7de05 | 1413 | while (managed = READ_ONCE(path->dentry->d_flags), |
cc53ce53 DH |
1414 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { |
1415 | /* Allow the filesystem to manage the transit without i_mutex | |
1416 | * being held. | |
1417 | * | |
1418 | * We indicate to the filesystem if someone is trying to mount | |
1419 | * something here. This gives autofs the chance to deny anyone | |
1420 | * other than its daemon the right to mount on its | |
1421 | * superstructure. | |
1422 | * | |
1423 | * The filesystem may sleep at this point. | |
1424 | */ | |
1425 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1426 | BUG_ON(!path->dentry->d_op); | |
1427 | BUG_ON(!path->dentry->d_op->d_manage); | |
fb5f51c7 | 1428 | ret = path->dentry->d_op->d_manage(path, false); |
cc53ce53 DH |
1429 | if (ret < 0) |
1430 | return ret == -EISDIR ? 0 : ret; | |
1431 | } | |
1432 | ||
1433 | /* Transit to a mounted filesystem. */ | |
1434 | if (managed & DCACHE_MOUNTED) { | |
1435 | struct vfsmount *mounted = lookup_mnt(path); | |
1436 | if (!mounted) | |
1437 | break; | |
1438 | dput(path->dentry); | |
1439 | mntput(path->mnt); | |
1440 | path->mnt = mounted; | |
1441 | path->dentry = dget(mounted->mnt_root); | |
1442 | continue; | |
1443 | } | |
1444 | ||
1445 | /* Don't handle automount points here */ | |
1446 | break; | |
1447 | } | |
1448 | return 0; | |
1449 | } | |
4d359507 | 1450 | EXPORT_SYMBOL(follow_down); |
cc53ce53 | 1451 | |
9875cf80 DH |
1452 | /* |
1453 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1454 | */ | |
1455 | static void follow_mount(struct path *path) | |
1456 | { | |
1457 | while (d_mountpoint(path->dentry)) { | |
1458 | struct vfsmount *mounted = lookup_mnt(path); | |
1459 | if (!mounted) | |
1460 | break; | |
1461 | dput(path->dentry); | |
1462 | mntput(path->mnt); | |
1463 | path->mnt = mounted; | |
1464 | path->dentry = dget(mounted->mnt_root); | |
1465 | } | |
1466 | } | |
1467 | ||
eedf265a EB |
1468 | static int path_parent_directory(struct path *path) |
1469 | { | |
1470 | struct dentry *old = path->dentry; | |
1471 | /* rare case of legitimate dget_parent()... */ | |
1472 | path->dentry = dget_parent(path->dentry); | |
1473 | dput(old); | |
1474 | if (unlikely(!path_connected(path))) | |
1475 | return -ENOENT; | |
1476 | return 0; | |
1477 | } | |
1478 | ||
397d425d | 1479 | static int follow_dotdot(struct nameidata *nd) |
1da177e4 LT |
1480 | { |
1481 | while(1) { | |
030c7e0b | 1482 | if (path_equal(&nd->path, &nd->root)) |
1da177e4 | 1483 | break; |
4ac91378 | 1484 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
eedf265a EB |
1485 | int ret = path_parent_directory(&nd->path); |
1486 | if (ret) | |
1487 | return ret; | |
1da177e4 LT |
1488 | break; |
1489 | } | |
3088dd70 | 1490 | if (!follow_up(&nd->path)) |
1da177e4 | 1491 | break; |
1da177e4 | 1492 | } |
79ed0226 | 1493 | follow_mount(&nd->path); |
31e6b01f | 1494 | nd->inode = nd->path.dentry->d_inode; |
397d425d | 1495 | return 0; |
1da177e4 LT |
1496 | } |
1497 | ||
baa03890 | 1498 | /* |
f4fdace9 OD |
1499 | * This looks up the name in dcache and possibly revalidates the found dentry. |
1500 | * NULL is returned if the dentry does not exist in the cache. | |
baa03890 | 1501 | */ |
e3c13928 AV |
1502 | static struct dentry *lookup_dcache(const struct qstr *name, |
1503 | struct dentry *dir, | |
6c51e513 | 1504 | unsigned int flags) |
baa03890 | 1505 | { |
a89f8337 | 1506 | struct dentry *dentry = d_lookup(dir, name); |
bad61189 | 1507 | if (dentry) { |
a89f8337 AV |
1508 | int error = d_revalidate(dentry, flags); |
1509 | if (unlikely(error <= 0)) { | |
1510 | if (!error) | |
1511 | d_invalidate(dentry); | |
1512 | dput(dentry); | |
1513 | return ERR_PTR(error); | |
bad61189 MS |
1514 | } |
1515 | } | |
baa03890 NP |
1516 | return dentry; |
1517 | } | |
1518 | ||
44396f4b | 1519 | /* |
a03ece5f AV |
1520 | * Parent directory has inode locked exclusive. This is one |
1521 | * and only case when ->lookup() gets called on non in-lookup | |
1522 | * dentries - as the matter of fact, this only gets called | |
1523 | * when directory is guaranteed to have no in-lookup children | |
1524 | * at all. | |
44396f4b | 1525 | */ |
e3c13928 | 1526 | static struct dentry *__lookup_hash(const struct qstr *name, |
72bd866a | 1527 | struct dentry *base, unsigned int flags) |
a3255546 | 1528 | { |
6c51e513 | 1529 | struct dentry *dentry = lookup_dcache(name, base, flags); |
a03ece5f AV |
1530 | struct dentry *old; |
1531 | struct inode *dir = base->d_inode; | |
a3255546 | 1532 | |
6c51e513 | 1533 | if (dentry) |
bad61189 | 1534 | return dentry; |
a3255546 | 1535 | |
a03ece5f AV |
1536 | /* Don't create child dentry for a dead directory. */ |
1537 | if (unlikely(IS_DEADDIR(dir))) | |
1538 | return ERR_PTR(-ENOENT); | |
1539 | ||
6c51e513 AV |
1540 | dentry = d_alloc(base, name); |
1541 | if (unlikely(!dentry)) | |
1542 | return ERR_PTR(-ENOMEM); | |
1543 | ||
a03ece5f AV |
1544 | old = dir->i_op->lookup(dir, dentry, flags); |
1545 | if (unlikely(old)) { | |
1546 | dput(dentry); | |
1547 | dentry = old; | |
1548 | } | |
1549 | return dentry; | |
a3255546 AV |
1550 | } |
1551 | ||
e97cdc87 | 1552 | static int lookup_fast(struct nameidata *nd, |
254cf582 AV |
1553 | struct path *path, struct inode **inode, |
1554 | unsigned *seqp) | |
1da177e4 | 1555 | { |
4ac91378 | 1556 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1557 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 | 1558 | int status = 1; |
9875cf80 DH |
1559 | int err; |
1560 | ||
b04f784e NP |
1561 | /* |
1562 | * Rename seqlock is not required here because in the off chance | |
5d0f49c1 AV |
1563 | * of a false negative due to a concurrent rename, the caller is |
1564 | * going to fall back to non-racy lookup. | |
b04f784e | 1565 | */ |
31e6b01f NP |
1566 | if (nd->flags & LOOKUP_RCU) { |
1567 | unsigned seq; | |
766c4cbf | 1568 | bool negative; |
da53be12 | 1569 | dentry = __d_lookup_rcu(parent, &nd->last, &seq); |
5d0f49c1 | 1570 | if (unlikely(!dentry)) { |
4675ac39 | 1571 | if (unlazy_walk(nd)) |
5d0f49c1 | 1572 | return -ECHILD; |
e9742b53 | 1573 | return 0; |
5d0f49c1 | 1574 | } |
5a18fff2 | 1575 | |
12f8ad4b LT |
1576 | /* |
1577 | * This sequence count validates that the inode matches | |
1578 | * the dentry name information from lookup. | |
1579 | */ | |
63afdfc7 | 1580 | *inode = d_backing_inode(dentry); |
766c4cbf | 1581 | negative = d_is_negative(dentry); |
5d0f49c1 | 1582 | if (unlikely(read_seqcount_retry(&dentry->d_seq, seq))) |
12f8ad4b LT |
1583 | return -ECHILD; |
1584 | ||
1585 | /* | |
1586 | * This sequence count validates that the parent had no | |
1587 | * changes while we did the lookup of the dentry above. | |
1588 | * | |
1589 | * The memory barrier in read_seqcount_begin of child is | |
1590 | * enough, we can use __read_seqcount_retry here. | |
1591 | */ | |
5d0f49c1 | 1592 | if (unlikely(__read_seqcount_retry(&parent->d_seq, nd->seq))) |
31e6b01f | 1593 | return -ECHILD; |
5a18fff2 | 1594 | |
254cf582 | 1595 | *seqp = seq; |
a89f8337 | 1596 | status = d_revalidate(dentry, nd->flags); |
209a7fb2 | 1597 | if (likely(status > 0)) { |
5d0f49c1 AV |
1598 | /* |
1599 | * Note: do negative dentry check after revalidation in | |
1600 | * case that drops it. | |
1601 | */ | |
1602 | if (unlikely(negative)) | |
1603 | return -ENOENT; | |
1604 | path->mnt = mnt; | |
1605 | path->dentry = dentry; | |
1606 | if (likely(__follow_mount_rcu(nd, path, inode, seqp))) | |
e9742b53 | 1607 | return 1; |
24643087 | 1608 | } |
4675ac39 | 1609 | if (unlazy_child(nd, dentry, seq)) |
209a7fb2 AV |
1610 | return -ECHILD; |
1611 | if (unlikely(status == -ECHILD)) | |
1612 | /* we'd been told to redo it in non-rcu mode */ | |
1613 | status = d_revalidate(dentry, nd->flags); | |
5a18fff2 | 1614 | } else { |
e97cdc87 | 1615 | dentry = __d_lookup(parent, &nd->last); |
5d0f49c1 | 1616 | if (unlikely(!dentry)) |
e9742b53 | 1617 | return 0; |
a89f8337 | 1618 | status = d_revalidate(dentry, nd->flags); |
9875cf80 | 1619 | } |
5a18fff2 | 1620 | if (unlikely(status <= 0)) { |
e9742b53 | 1621 | if (!status) |
5d0f49c1 | 1622 | d_invalidate(dentry); |
5542aa2f | 1623 | dput(dentry); |
5d0f49c1 | 1624 | return status; |
24643087 | 1625 | } |
766c4cbf AV |
1626 | if (unlikely(d_is_negative(dentry))) { |
1627 | dput(dentry); | |
1628 | return -ENOENT; | |
1629 | } | |
5d0f49c1 | 1630 | |
9875cf80 DH |
1631 | path->mnt = mnt; |
1632 | path->dentry = dentry; | |
756daf26 | 1633 | err = follow_managed(path, nd); |
e9742b53 | 1634 | if (likely(err > 0)) |
63afdfc7 | 1635 | *inode = d_backing_inode(path->dentry); |
8402752e | 1636 | return err; |
697f514d MS |
1637 | } |
1638 | ||
1639 | /* Fast lookup failed, do it the slow way */ | |
88d8331a AV |
1640 | static struct dentry *__lookup_slow(const struct qstr *name, |
1641 | struct dentry *dir, | |
1642 | unsigned int flags) | |
697f514d | 1643 | { |
88d8331a | 1644 | struct dentry *dentry, *old; |
1936386e | 1645 | struct inode *inode = dir->d_inode; |
d9171b93 | 1646 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
1936386e | 1647 | |
1936386e | 1648 | /* Don't go there if it's already dead */ |
94bdd655 | 1649 | if (unlikely(IS_DEADDIR(inode))) |
88d8331a | 1650 | return ERR_PTR(-ENOENT); |
94bdd655 | 1651 | again: |
d9171b93 | 1652 | dentry = d_alloc_parallel(dir, name, &wq); |
94bdd655 | 1653 | if (IS_ERR(dentry)) |
88d8331a | 1654 | return dentry; |
94bdd655 | 1655 | if (unlikely(!d_in_lookup(dentry))) { |
a89f8337 | 1656 | if (!(flags & LOOKUP_NO_REVAL)) { |
949a852e AV |
1657 | int error = d_revalidate(dentry, flags); |
1658 | if (unlikely(error <= 0)) { | |
94bdd655 | 1659 | if (!error) { |
949a852e | 1660 | d_invalidate(dentry); |
94bdd655 AV |
1661 | dput(dentry); |
1662 | goto again; | |
1663 | } | |
949a852e AV |
1664 | dput(dentry); |
1665 | dentry = ERR_PTR(error); | |
1666 | } | |
1667 | } | |
94bdd655 AV |
1668 | } else { |
1669 | old = inode->i_op->lookup(inode, dentry, flags); | |
1670 | d_lookup_done(dentry); | |
1671 | if (unlikely(old)) { | |
1672 | dput(dentry); | |
1673 | dentry = old; | |
949a852e AV |
1674 | } |
1675 | } | |
e3c13928 | 1676 | return dentry; |
1da177e4 LT |
1677 | } |
1678 | ||
88d8331a AV |
1679 | static struct dentry *lookup_slow(const struct qstr *name, |
1680 | struct dentry *dir, | |
1681 | unsigned int flags) | |
1682 | { | |
1683 | struct inode *inode = dir->d_inode; | |
1684 | struct dentry *res; | |
1685 | inode_lock_shared(inode); | |
1686 | res = __lookup_slow(name, dir, flags); | |
1687 | inode_unlock_shared(inode); | |
1688 | return res; | |
1689 | } | |
1690 | ||
52094c8a AV |
1691 | static inline int may_lookup(struct nameidata *nd) |
1692 | { | |
1693 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1694 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1695 | if (err != -ECHILD) |
1696 | return err; | |
4675ac39 | 1697 | if (unlazy_walk(nd)) |
52094c8a AV |
1698 | return -ECHILD; |
1699 | } | |
4ad5abb3 | 1700 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1701 | } |
1702 | ||
9856fa1b AV |
1703 | static inline int handle_dots(struct nameidata *nd, int type) |
1704 | { | |
1705 | if (type == LAST_DOTDOT) { | |
9e6697e2 AV |
1706 | if (!nd->root.mnt) |
1707 | set_root(nd); | |
9856fa1b | 1708 | if (nd->flags & LOOKUP_RCU) { |
70291aec | 1709 | return follow_dotdot_rcu(nd); |
9856fa1b | 1710 | } else |
397d425d | 1711 | return follow_dotdot(nd); |
9856fa1b AV |
1712 | } |
1713 | return 0; | |
1714 | } | |
1715 | ||
181548c0 AV |
1716 | static int pick_link(struct nameidata *nd, struct path *link, |
1717 | struct inode *inode, unsigned seq) | |
d63ff28f | 1718 | { |
626de996 | 1719 | int error; |
1cf2665b | 1720 | struct saved *last; |
756daf26 | 1721 | if (unlikely(nd->total_link_count++ >= MAXSYMLINKS)) { |
626de996 AV |
1722 | path_to_nameidata(link, nd); |
1723 | return -ELOOP; | |
1724 | } | |
bc40aee0 | 1725 | if (!(nd->flags & LOOKUP_RCU)) { |
7973387a AV |
1726 | if (link->mnt == nd->path.mnt) |
1727 | mntget(link->mnt); | |
d63ff28f | 1728 | } |
626de996 AV |
1729 | error = nd_alloc_stack(nd); |
1730 | if (unlikely(error)) { | |
bc40aee0 | 1731 | if (error == -ECHILD) { |
ad1633a1 AV |
1732 | if (unlikely(!legitimize_path(nd, link, seq))) { |
1733 | drop_links(nd); | |
1734 | nd->depth = 0; | |
1735 | nd->flags &= ~LOOKUP_RCU; | |
1736 | nd->path.mnt = NULL; | |
1737 | nd->path.dentry = NULL; | |
1738 | if (!(nd->flags & LOOKUP_ROOT)) | |
1739 | nd->root.mnt = NULL; | |
1740 | rcu_read_unlock(); | |
4675ac39 | 1741 | } else if (likely(unlazy_walk(nd)) == 0) |
ad1633a1 | 1742 | error = nd_alloc_stack(nd); |
bc40aee0 AV |
1743 | } |
1744 | if (error) { | |
1745 | path_put(link); | |
1746 | return error; | |
1747 | } | |
626de996 AV |
1748 | } |
1749 | ||
ab104923 | 1750 | last = nd->stack + nd->depth++; |
1cf2665b | 1751 | last->link = *link; |
fceef393 AV |
1752 | clear_delayed_call(&last->done); |
1753 | nd->link_inode = inode; | |
0450b2d1 | 1754 | last->seq = seq; |
d63ff28f AV |
1755 | return 1; |
1756 | } | |
1757 | ||
8f64fb1c | 1758 | enum {WALK_FOLLOW = 1, WALK_MORE = 2}; |
31d66bcd | 1759 | |
3ddcd056 LT |
1760 | /* |
1761 | * Do we need to follow links? We _really_ want to be able | |
1762 | * to do this check without having to look at inode->i_op, | |
1763 | * so we keep a cache of "no, this doesn't need follow_link" | |
1764 | * for the common case. | |
1765 | */ | |
8f64fb1c AV |
1766 | static inline int step_into(struct nameidata *nd, struct path *path, |
1767 | int flags, struct inode *inode, unsigned seq) | |
3ddcd056 | 1768 | { |
31d66bcd AV |
1769 | if (!(flags & WALK_MORE) && nd->depth) |
1770 | put_link(nd); | |
8f64fb1c AV |
1771 | if (likely(!d_is_symlink(path->dentry)) || |
1772 | !(flags & WALK_FOLLOW || nd->flags & LOOKUP_FOLLOW)) { | |
1773 | /* not a symlink or should not follow */ | |
1774 | path_to_nameidata(path, nd); | |
1775 | nd->inode = inode; | |
1776 | nd->seq = seq; | |
d63ff28f | 1777 | return 0; |
8f64fb1c | 1778 | } |
a7f77542 AV |
1779 | /* make sure that d_is_symlink above matches inode */ |
1780 | if (nd->flags & LOOKUP_RCU) { | |
8f64fb1c | 1781 | if (read_seqcount_retry(&path->dentry->d_seq, seq)) |
a7f77542 AV |
1782 | return -ECHILD; |
1783 | } | |
8f64fb1c | 1784 | return pick_link(nd, path, inode, seq); |
3ddcd056 LT |
1785 | } |
1786 | ||
4693a547 | 1787 | static int walk_component(struct nameidata *nd, int flags) |
ce57dfc1 | 1788 | { |
caa85634 | 1789 | struct path path; |
ce57dfc1 | 1790 | struct inode *inode; |
254cf582 | 1791 | unsigned seq; |
ce57dfc1 AV |
1792 | int err; |
1793 | /* | |
1794 | * "." and ".." are special - ".." especially so because it has | |
1795 | * to be able to know about the current root directory and | |
1796 | * parent relationships. | |
1797 | */ | |
4693a547 AV |
1798 | if (unlikely(nd->last_type != LAST_NORM)) { |
1799 | err = handle_dots(nd, nd->last_type); | |
1c4ff1a8 | 1800 | if (!(flags & WALK_MORE) && nd->depth) |
4693a547 AV |
1801 | put_link(nd); |
1802 | return err; | |
1803 | } | |
254cf582 | 1804 | err = lookup_fast(nd, &path, &inode, &seq); |
e9742b53 | 1805 | if (unlikely(err <= 0)) { |
697f514d | 1806 | if (err < 0) |
f0a9ba70 | 1807 | return err; |
e3c13928 AV |
1808 | path.dentry = lookup_slow(&nd->last, nd->path.dentry, |
1809 | nd->flags); | |
1810 | if (IS_ERR(path.dentry)) | |
1811 | return PTR_ERR(path.dentry); | |
7500c38a | 1812 | |
e3c13928 AV |
1813 | path.mnt = nd->path.mnt; |
1814 | err = follow_managed(&path, nd); | |
1815 | if (unlikely(err < 0)) | |
f0a9ba70 | 1816 | return err; |
697f514d | 1817 | |
7500c38a AV |
1818 | if (unlikely(d_is_negative(path.dentry))) { |
1819 | path_to_nameidata(&path, nd); | |
1820 | return -ENOENT; | |
1821 | } | |
1822 | ||
254cf582 | 1823 | seq = 0; /* we are already out of RCU mode */ |
d4565649 | 1824 | inode = d_backing_inode(path.dentry); |
ce57dfc1 | 1825 | } |
697f514d | 1826 | |
8f64fb1c | 1827 | return step_into(nd, &path, flags, inode, seq); |
ce57dfc1 AV |
1828 | } |
1829 | ||
bfcfaa77 LT |
1830 | /* |
1831 | * We can do the critical dentry name comparison and hashing | |
1832 | * operations one word at a time, but we are limited to: | |
1833 | * | |
1834 | * - Architectures with fast unaligned word accesses. We could | |
1835 | * do a "get_unaligned()" if this helps and is sufficiently | |
1836 | * fast. | |
1837 | * | |
bfcfaa77 LT |
1838 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we |
1839 | * do not trap on the (extremely unlikely) case of a page | |
1840 | * crossing operation. | |
1841 | * | |
1842 | * - Furthermore, we need an efficient 64-bit compile for the | |
1843 | * 64-bit case in order to generate the "number of bytes in | |
1844 | * the final mask". Again, that could be replaced with a | |
1845 | * efficient population count instruction or similar. | |
1846 | */ | |
1847 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1848 | ||
f68e556e | 1849 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1850 | |
468a9428 | 1851 | #ifdef HASH_MIX |
bfcfaa77 | 1852 | |
468a9428 | 1853 | /* Architecture provides HASH_MIX and fold_hash() in <asm/hash.h> */ |
bfcfaa77 | 1854 | |
468a9428 | 1855 | #elif defined(CONFIG_64BIT) |
0fed3ac8 | 1856 | /* |
2a18da7a GS |
1857 | * Register pressure in the mixing function is an issue, particularly |
1858 | * on 32-bit x86, but almost any function requires one state value and | |
1859 | * one temporary. Instead, use a function designed for two state values | |
1860 | * and no temporaries. | |
1861 | * | |
1862 | * This function cannot create a collision in only two iterations, so | |
1863 | * we have two iterations to achieve avalanche. In those two iterations, | |
1864 | * we have six layers of mixing, which is enough to spread one bit's | |
1865 | * influence out to 2^6 = 64 state bits. | |
1866 | * | |
1867 | * Rotate constants are scored by considering either 64 one-bit input | |
1868 | * deltas or 64*63/2 = 2016 two-bit input deltas, and finding the | |
1869 | * probability of that delta causing a change to each of the 128 output | |
1870 | * bits, using a sample of random initial states. | |
1871 | * | |
1872 | * The Shannon entropy of the computed probabilities is then summed | |
1873 | * to produce a score. Ideally, any input change has a 50% chance of | |
1874 | * toggling any given output bit. | |
1875 | * | |
1876 | * Mixing scores (in bits) for (12,45): | |
1877 | * Input delta: 1-bit 2-bit | |
1878 | * 1 round: 713.3 42542.6 | |
1879 | * 2 rounds: 2753.7 140389.8 | |
1880 | * 3 rounds: 5954.1 233458.2 | |
1881 | * 4 rounds: 7862.6 256672.2 | |
1882 | * Perfect: 8192 258048 | |
1883 | * (64*128) (64*63/2 * 128) | |
0fed3ac8 | 1884 | */ |
2a18da7a GS |
1885 | #define HASH_MIX(x, y, a) \ |
1886 | ( x ^= (a), \ | |
1887 | y ^= x, x = rol64(x,12),\ | |
1888 | x += y, y = rol64(y,45),\ | |
1889 | y *= 9 ) | |
bfcfaa77 | 1890 | |
0fed3ac8 | 1891 | /* |
2a18da7a GS |
1892 | * Fold two longs into one 32-bit hash value. This must be fast, but |
1893 | * latency isn't quite as critical, as there is a fair bit of additional | |
1894 | * work done before the hash value is used. | |
0fed3ac8 | 1895 | */ |
2a18da7a | 1896 | static inline unsigned int fold_hash(unsigned long x, unsigned long y) |
0fed3ac8 | 1897 | { |
2a18da7a GS |
1898 | y ^= x * GOLDEN_RATIO_64; |
1899 | y *= GOLDEN_RATIO_64; | |
1900 | return y >> 32; | |
0fed3ac8 GS |
1901 | } |
1902 | ||
bfcfaa77 LT |
1903 | #else /* 32-bit case */ |
1904 | ||
2a18da7a GS |
1905 | /* |
1906 | * Mixing scores (in bits) for (7,20): | |
1907 | * Input delta: 1-bit 2-bit | |
1908 | * 1 round: 330.3 9201.6 | |
1909 | * 2 rounds: 1246.4 25475.4 | |
1910 | * 3 rounds: 1907.1 31295.1 | |
1911 | * 4 rounds: 2042.3 31718.6 | |
1912 | * Perfect: 2048 31744 | |
1913 | * (32*64) (32*31/2 * 64) | |
1914 | */ | |
1915 | #define HASH_MIX(x, y, a) \ | |
1916 | ( x ^= (a), \ | |
1917 | y ^= x, x = rol32(x, 7),\ | |
1918 | x += y, y = rol32(y,20),\ | |
1919 | y *= 9 ) | |
bfcfaa77 | 1920 | |
2a18da7a | 1921 | static inline unsigned int fold_hash(unsigned long x, unsigned long y) |
0fed3ac8 | 1922 | { |
2a18da7a GS |
1923 | /* Use arch-optimized multiply if one exists */ |
1924 | return __hash_32(y ^ __hash_32(x)); | |
0fed3ac8 GS |
1925 | } |
1926 | ||
bfcfaa77 LT |
1927 | #endif |
1928 | ||
2a18da7a GS |
1929 | /* |
1930 | * Return the hash of a string of known length. This is carfully | |
1931 | * designed to match hash_name(), which is the more critical function. | |
1932 | * In particular, we must end by hashing a final word containing 0..7 | |
1933 | * payload bytes, to match the way that hash_name() iterates until it | |
1934 | * finds the delimiter after the name. | |
1935 | */ | |
8387ff25 | 1936 | unsigned int full_name_hash(const void *salt, const char *name, unsigned int len) |
bfcfaa77 | 1937 | { |
8387ff25 | 1938 | unsigned long a, x = 0, y = (unsigned long)salt; |
bfcfaa77 LT |
1939 | |
1940 | for (;;) { | |
fcfd2fbf GS |
1941 | if (!len) |
1942 | goto done; | |
e419b4cc | 1943 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1944 | if (len < sizeof(unsigned long)) |
1945 | break; | |
2a18da7a | 1946 | HASH_MIX(x, y, a); |
bfcfaa77 LT |
1947 | name += sizeof(unsigned long); |
1948 | len -= sizeof(unsigned long); | |
bfcfaa77 | 1949 | } |
2a18da7a | 1950 | x ^= a & bytemask_from_count(len); |
bfcfaa77 | 1951 | done: |
2a18da7a | 1952 | return fold_hash(x, y); |
bfcfaa77 LT |
1953 | } |
1954 | EXPORT_SYMBOL(full_name_hash); | |
1955 | ||
fcfd2fbf | 1956 | /* Return the "hash_len" (hash and length) of a null-terminated string */ |
8387ff25 | 1957 | u64 hashlen_string(const void *salt, const char *name) |
fcfd2fbf | 1958 | { |
8387ff25 LT |
1959 | unsigned long a = 0, x = 0, y = (unsigned long)salt; |
1960 | unsigned long adata, mask, len; | |
fcfd2fbf GS |
1961 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; |
1962 | ||
8387ff25 LT |
1963 | len = 0; |
1964 | goto inside; | |
1965 | ||
fcfd2fbf | 1966 | do { |
2a18da7a | 1967 | HASH_MIX(x, y, a); |
fcfd2fbf | 1968 | len += sizeof(unsigned long); |
8387ff25 | 1969 | inside: |
fcfd2fbf GS |
1970 | a = load_unaligned_zeropad(name+len); |
1971 | } while (!has_zero(a, &adata, &constants)); | |
1972 | ||
1973 | adata = prep_zero_mask(a, adata, &constants); | |
1974 | mask = create_zero_mask(adata); | |
2a18da7a | 1975 | x ^= a & zero_bytemask(mask); |
fcfd2fbf | 1976 | |
2a18da7a | 1977 | return hashlen_create(fold_hash(x, y), len + find_zero(mask)); |
fcfd2fbf GS |
1978 | } |
1979 | EXPORT_SYMBOL(hashlen_string); | |
1980 | ||
bfcfaa77 LT |
1981 | /* |
1982 | * Calculate the length and hash of the path component, and | |
d6bb3e90 | 1983 | * return the "hash_len" as the result. |
bfcfaa77 | 1984 | */ |
8387ff25 | 1985 | static inline u64 hash_name(const void *salt, const char *name) |
bfcfaa77 | 1986 | { |
8387ff25 LT |
1987 | unsigned long a = 0, b, x = 0, y = (unsigned long)salt; |
1988 | unsigned long adata, bdata, mask, len; | |
36126f8f | 1989 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; |
bfcfaa77 | 1990 | |
8387ff25 LT |
1991 | len = 0; |
1992 | goto inside; | |
1993 | ||
bfcfaa77 | 1994 | do { |
2a18da7a | 1995 | HASH_MIX(x, y, a); |
bfcfaa77 | 1996 | len += sizeof(unsigned long); |
8387ff25 | 1997 | inside: |
e419b4cc | 1998 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1999 | b = a ^ REPEAT_BYTE('/'); |
2000 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
2001 | ||
2002 | adata = prep_zero_mask(a, adata, &constants); | |
2003 | bdata = prep_zero_mask(b, bdata, &constants); | |
36126f8f | 2004 | mask = create_zero_mask(adata | bdata); |
2a18da7a | 2005 | x ^= a & zero_bytemask(mask); |
36126f8f | 2006 | |
2a18da7a | 2007 | return hashlen_create(fold_hash(x, y), len + find_zero(mask)); |
bfcfaa77 LT |
2008 | } |
2009 | ||
2a18da7a | 2010 | #else /* !CONFIG_DCACHE_WORD_ACCESS: Slow, byte-at-a-time version */ |
bfcfaa77 | 2011 | |
fcfd2fbf | 2012 | /* Return the hash of a string of known length */ |
8387ff25 | 2013 | unsigned int full_name_hash(const void *salt, const char *name, unsigned int len) |
0145acc2 | 2014 | { |
8387ff25 | 2015 | unsigned long hash = init_name_hash(salt); |
0145acc2 | 2016 | while (len--) |
fcfd2fbf | 2017 | hash = partial_name_hash((unsigned char)*name++, hash); |
0145acc2 LT |
2018 | return end_name_hash(hash); |
2019 | } | |
ae942ae7 | 2020 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 2021 | |
fcfd2fbf | 2022 | /* Return the "hash_len" (hash and length) of a null-terminated string */ |
8387ff25 | 2023 | u64 hashlen_string(const void *salt, const char *name) |
fcfd2fbf | 2024 | { |
8387ff25 | 2025 | unsigned long hash = init_name_hash(salt); |
fcfd2fbf GS |
2026 | unsigned long len = 0, c; |
2027 | ||
2028 | c = (unsigned char)*name; | |
e0ab7af9 | 2029 | while (c) { |
fcfd2fbf GS |
2030 | len++; |
2031 | hash = partial_name_hash(c, hash); | |
2032 | c = (unsigned char)name[len]; | |
e0ab7af9 | 2033 | } |
fcfd2fbf GS |
2034 | return hashlen_create(end_name_hash(hash), len); |
2035 | } | |
f2a031b6 | 2036 | EXPORT_SYMBOL(hashlen_string); |
fcfd2fbf | 2037 | |
200e9ef7 LT |
2038 | /* |
2039 | * We know there's a real path component here of at least | |
2040 | * one character. | |
2041 | */ | |
8387ff25 | 2042 | static inline u64 hash_name(const void *salt, const char *name) |
200e9ef7 | 2043 | { |
8387ff25 | 2044 | unsigned long hash = init_name_hash(salt); |
200e9ef7 LT |
2045 | unsigned long len = 0, c; |
2046 | ||
2047 | c = (unsigned char)*name; | |
2048 | do { | |
2049 | len++; | |
2050 | hash = partial_name_hash(c, hash); | |
2051 | c = (unsigned char)name[len]; | |
2052 | } while (c && c != '/'); | |
d6bb3e90 | 2053 | return hashlen_create(end_name_hash(hash), len); |
200e9ef7 LT |
2054 | } |
2055 | ||
bfcfaa77 LT |
2056 | #endif |
2057 | ||
1da177e4 LT |
2058 | /* |
2059 | * Name resolution. | |
ea3834d9 PM |
2060 | * This is the basic name resolution function, turning a pathname into |
2061 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 2062 | * |
ea3834d9 PM |
2063 | * Returns 0 and nd will have valid dentry and mnt on success. |
2064 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 2065 | */ |
6de88d72 | 2066 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 2067 | { |
1da177e4 | 2068 | int err; |
32cd7468 | 2069 | |
9b5858e9 AV |
2070 | if (IS_ERR(name)) |
2071 | return PTR_ERR(name); | |
1da177e4 LT |
2072 | while (*name=='/') |
2073 | name++; | |
2074 | if (!*name) | |
9e18f10a | 2075 | return 0; |
1da177e4 | 2076 | |
1da177e4 LT |
2077 | /* At this point we know we have a real path component. */ |
2078 | for(;;) { | |
d6bb3e90 | 2079 | u64 hash_len; |
fe479a58 | 2080 | int type; |
1da177e4 | 2081 | |
52094c8a | 2082 | err = may_lookup(nd); |
2a18da7a | 2083 | if (err) |
3595e234 | 2084 | return err; |
1da177e4 | 2085 | |
8387ff25 | 2086 | hash_len = hash_name(nd->path.dentry, name); |
1da177e4 | 2087 | |
fe479a58 | 2088 | type = LAST_NORM; |
d6bb3e90 | 2089 | if (name[0] == '.') switch (hashlen_len(hash_len)) { |
fe479a58 | 2090 | case 2: |
200e9ef7 | 2091 | if (name[1] == '.') { |
fe479a58 | 2092 | type = LAST_DOTDOT; |
16c2cd71 AV |
2093 | nd->flags |= LOOKUP_JUMPED; |
2094 | } | |
fe479a58 AV |
2095 | break; |
2096 | case 1: | |
2097 | type = LAST_DOT; | |
2098 | } | |
5a202bcd AV |
2099 | if (likely(type == LAST_NORM)) { |
2100 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 2101 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd | 2102 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
a060dc50 | 2103 | struct qstr this = { { .hash_len = hash_len }, .name = name }; |
da53be12 | 2104 | err = parent->d_op->d_hash(parent, &this); |
5a202bcd | 2105 | if (err < 0) |
3595e234 | 2106 | return err; |
d6bb3e90 LT |
2107 | hash_len = this.hash_len; |
2108 | name = this.name; | |
5a202bcd AV |
2109 | } |
2110 | } | |
fe479a58 | 2111 | |
d6bb3e90 LT |
2112 | nd->last.hash_len = hash_len; |
2113 | nd->last.name = name; | |
5f4a6a69 AV |
2114 | nd->last_type = type; |
2115 | ||
d6bb3e90 LT |
2116 | name += hashlen_len(hash_len); |
2117 | if (!*name) | |
bdf6cbf1 | 2118 | goto OK; |
200e9ef7 LT |
2119 | /* |
2120 | * If it wasn't NUL, we know it was '/'. Skip that | |
2121 | * slash, and continue until no more slashes. | |
2122 | */ | |
2123 | do { | |
d6bb3e90 LT |
2124 | name++; |
2125 | } while (unlikely(*name == '/')); | |
8620c238 AV |
2126 | if (unlikely(!*name)) { |
2127 | OK: | |
368ee9ba | 2128 | /* pathname body, done */ |
8620c238 AV |
2129 | if (!nd->depth) |
2130 | return 0; | |
2131 | name = nd->stack[nd->depth - 1].name; | |
368ee9ba | 2132 | /* trailing symlink, done */ |
8620c238 AV |
2133 | if (!name) |
2134 | return 0; | |
2135 | /* last component of nested symlink */ | |
8f64fb1c | 2136 | err = walk_component(nd, WALK_FOLLOW); |
1c4ff1a8 AV |
2137 | } else { |
2138 | /* not the last component */ | |
8f64fb1c | 2139 | err = walk_component(nd, WALK_FOLLOW | WALK_MORE); |
8620c238 | 2140 | } |
ce57dfc1 | 2141 | if (err < 0) |
3595e234 | 2142 | return err; |
1da177e4 | 2143 | |
ce57dfc1 | 2144 | if (err) { |
626de996 | 2145 | const char *s = get_link(nd); |
5a460275 | 2146 | |
a1c83681 | 2147 | if (IS_ERR(s)) |
3595e234 | 2148 | return PTR_ERR(s); |
d40bcc09 AV |
2149 | err = 0; |
2150 | if (unlikely(!s)) { | |
2151 | /* jumped */ | |
b9ff4429 | 2152 | put_link(nd); |
d40bcc09 | 2153 | } else { |
fab51e8a AV |
2154 | nd->stack[nd->depth - 1].name = name; |
2155 | name = s; | |
2156 | continue; | |
d40bcc09 | 2157 | } |
31e6b01f | 2158 | } |
97242f99 AV |
2159 | if (unlikely(!d_can_lookup(nd->path.dentry))) { |
2160 | if (nd->flags & LOOKUP_RCU) { | |
4675ac39 | 2161 | if (unlazy_walk(nd)) |
97242f99 AV |
2162 | return -ECHILD; |
2163 | } | |
3595e234 | 2164 | return -ENOTDIR; |
97242f99 | 2165 | } |
1da177e4 | 2166 | } |
1da177e4 LT |
2167 | } |
2168 | ||
edc2b1da | 2169 | /* must be paired with terminate_walk() */ |
c8a53ee5 | 2170 | static const char *path_init(struct nameidata *nd, unsigned flags) |
31e6b01f | 2171 | { |
c8a53ee5 | 2172 | const char *s = nd->name->name; |
31e6b01f | 2173 | |
c0eb027e LT |
2174 | if (!*s) |
2175 | flags &= ~LOOKUP_RCU; | |
edc2b1da AV |
2176 | if (flags & LOOKUP_RCU) |
2177 | rcu_read_lock(); | |
c0eb027e | 2178 | |
31e6b01f | 2179 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ |
980f3ea2 | 2180 | nd->flags = flags | LOOKUP_JUMPED | LOOKUP_PARENT; |
31e6b01f | 2181 | nd->depth = 0; |
5b6ca027 | 2182 | if (flags & LOOKUP_ROOT) { |
b18825a7 DH |
2183 | struct dentry *root = nd->root.dentry; |
2184 | struct inode *inode = root->d_inode; | |
93893862 AV |
2185 | if (*s && unlikely(!d_can_lookup(root))) |
2186 | return ERR_PTR(-ENOTDIR); | |
5b6ca027 AV |
2187 | nd->path = nd->root; |
2188 | nd->inode = inode; | |
2189 | if (flags & LOOKUP_RCU) { | |
5b6ca027 | 2190 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
8f47a016 | 2191 | nd->root_seq = nd->seq; |
48a066e7 | 2192 | nd->m_seq = read_seqbegin(&mount_lock); |
5b6ca027 AV |
2193 | } else { |
2194 | path_get(&nd->path); | |
2195 | } | |
368ee9ba | 2196 | return s; |
5b6ca027 AV |
2197 | } |
2198 | ||
31e6b01f | 2199 | nd->root.mnt = NULL; |
248fb5b9 AV |
2200 | nd->path.mnt = NULL; |
2201 | nd->path.dentry = NULL; | |
31e6b01f | 2202 | |
48a066e7 | 2203 | nd->m_seq = read_seqbegin(&mount_lock); |
fd2f7cb5 | 2204 | if (*s == '/') { |
9e6697e2 | 2205 | set_root(nd); |
248fb5b9 | 2206 | if (likely(!nd_jump_root(nd))) |
ef55d917 | 2207 | return s; |
ef55d917 | 2208 | return ERR_PTR(-ECHILD); |
c8a53ee5 | 2209 | } else if (nd->dfd == AT_FDCWD) { |
e41f7d4e AV |
2210 | if (flags & LOOKUP_RCU) { |
2211 | struct fs_struct *fs = current->fs; | |
2212 | unsigned seq; | |
31e6b01f | 2213 | |
e41f7d4e AV |
2214 | do { |
2215 | seq = read_seqcount_begin(&fs->seq); | |
2216 | nd->path = fs->pwd; | |
ef55d917 | 2217 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e AV |
2218 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
2219 | } while (read_seqcount_retry(&fs->seq, seq)); | |
2220 | } else { | |
2221 | get_fs_pwd(current->fs, &nd->path); | |
ef55d917 | 2222 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2223 | } |
ef55d917 | 2224 | return s; |
31e6b01f | 2225 | } else { |
582aa64a | 2226 | /* Caller must check execute permissions on the starting path component */ |
c8a53ee5 | 2227 | struct fd f = fdget_raw(nd->dfd); |
31e6b01f NP |
2228 | struct dentry *dentry; |
2229 | ||
2903ff01 | 2230 | if (!f.file) |
368ee9ba | 2231 | return ERR_PTR(-EBADF); |
31e6b01f | 2232 | |
2903ff01 | 2233 | dentry = f.file->f_path.dentry; |
31e6b01f | 2234 | |
edc2b1da AV |
2235 | if (*s && unlikely(!d_can_lookup(dentry))) { |
2236 | fdput(f); | |
2237 | return ERR_PTR(-ENOTDIR); | |
f52e0c11 | 2238 | } |
31e6b01f | 2239 | |
2903ff01 | 2240 | nd->path = f.file->f_path; |
e41f7d4e | 2241 | if (flags & LOOKUP_RCU) { |
34a26b99 AV |
2242 | nd->inode = nd->path.dentry->d_inode; |
2243 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
e41f7d4e | 2244 | } else { |
2903ff01 | 2245 | path_get(&nd->path); |
34a26b99 | 2246 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2247 | } |
34a26b99 | 2248 | fdput(f); |
368ee9ba | 2249 | return s; |
31e6b01f | 2250 | } |
9b4a9b14 AV |
2251 | } |
2252 | ||
3bdba28b | 2253 | static const char *trailing_symlink(struct nameidata *nd) |
95fa25d9 AV |
2254 | { |
2255 | const char *s; | |
fec2fa24 | 2256 | int error = may_follow_link(nd); |
deb106c6 | 2257 | if (unlikely(error)) |
3bdba28b | 2258 | return ERR_PTR(error); |
95fa25d9 | 2259 | nd->flags |= LOOKUP_PARENT; |
fab51e8a | 2260 | nd->stack[0].name = NULL; |
3b2e7f75 | 2261 | s = get_link(nd); |
deb106c6 | 2262 | return s ? s : ""; |
95fa25d9 AV |
2263 | } |
2264 | ||
caa85634 | 2265 | static inline int lookup_last(struct nameidata *nd) |
bd92d7fe AV |
2266 | { |
2267 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
2268 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
2269 | ||
2270 | nd->flags &= ~LOOKUP_PARENT; | |
1c4ff1a8 | 2271 | return walk_component(nd, 0); |
bd92d7fe AV |
2272 | } |
2273 | ||
4f757f3c AV |
2274 | static int handle_lookup_down(struct nameidata *nd) |
2275 | { | |
2276 | struct path path = nd->path; | |
2277 | struct inode *inode = nd->inode; | |
2278 | unsigned seq = nd->seq; | |
2279 | int err; | |
2280 | ||
2281 | if (nd->flags & LOOKUP_RCU) { | |
2282 | /* | |
2283 | * don't bother with unlazy_walk on failure - we are | |
2284 | * at the very beginning of walk, so we lose nothing | |
2285 | * if we simply redo everything in non-RCU mode | |
2286 | */ | |
2287 | if (unlikely(!__follow_mount_rcu(nd, &path, &inode, &seq))) | |
2288 | return -ECHILD; | |
2289 | } else { | |
2290 | dget(path.dentry); | |
2291 | err = follow_managed(&path, nd); | |
2292 | if (unlikely(err < 0)) | |
2293 | return err; | |
2294 | inode = d_backing_inode(path.dentry); | |
2295 | seq = 0; | |
2296 | } | |
2297 | path_to_nameidata(&path, nd); | |
2298 | nd->inode = inode; | |
2299 | nd->seq = seq; | |
2300 | return 0; | |
2301 | } | |
2302 | ||
9b4a9b14 | 2303 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2304 | static int path_lookupat(struct nameidata *nd, unsigned flags, struct path *path) |
9b4a9b14 | 2305 | { |
c8a53ee5 | 2306 | const char *s = path_init(nd, flags); |
bd92d7fe | 2307 | int err; |
31e6b01f | 2308 | |
9b5858e9 | 2309 | if (unlikely(flags & LOOKUP_DOWN) && !IS_ERR(s)) { |
4f757f3c | 2310 | err = handle_lookup_down(nd); |
5f336e72 AV |
2311 | if (unlikely(err < 0)) |
2312 | s = ERR_PTR(err); | |
4f757f3c AV |
2313 | } |
2314 | ||
3bdba28b AV |
2315 | while (!(err = link_path_walk(s, nd)) |
2316 | && ((err = lookup_last(nd)) > 0)) { | |
2317 | s = trailing_symlink(nd); | |
bd92d7fe | 2318 | } |
9f1fafee AV |
2319 | if (!err) |
2320 | err = complete_walk(nd); | |
bd92d7fe | 2321 | |
deb106c6 AV |
2322 | if (!err && nd->flags & LOOKUP_DIRECTORY) |
2323 | if (!d_can_lookup(nd->path.dentry)) | |
bd23a539 | 2324 | err = -ENOTDIR; |
625b6d10 AV |
2325 | if (!err) { |
2326 | *path = nd->path; | |
2327 | nd->path.mnt = NULL; | |
2328 | nd->path.dentry = NULL; | |
2329 | } | |
2330 | terminate_walk(nd); | |
bd92d7fe | 2331 | return err; |
ee0827cd | 2332 | } |
31e6b01f | 2333 | |
625b6d10 | 2334 | static int filename_lookup(int dfd, struct filename *name, unsigned flags, |
9ad1aaa6 | 2335 | struct path *path, struct path *root) |
ee0827cd | 2336 | { |
894bc8c4 | 2337 | int retval; |
9883d185 | 2338 | struct nameidata nd; |
abc9f5be AV |
2339 | if (IS_ERR(name)) |
2340 | return PTR_ERR(name); | |
9ad1aaa6 AV |
2341 | if (unlikely(root)) { |
2342 | nd.root = *root; | |
2343 | flags |= LOOKUP_ROOT; | |
2344 | } | |
9883d185 | 2345 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2346 | retval = path_lookupat(&nd, flags | LOOKUP_RCU, path); |
ee0827cd | 2347 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2348 | retval = path_lookupat(&nd, flags, path); |
ee0827cd | 2349 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2350 | retval = path_lookupat(&nd, flags | LOOKUP_REVAL, path); |
31e6b01f | 2351 | |
f78570dd | 2352 | if (likely(!retval)) |
625b6d10 | 2353 | audit_inode(name, path->dentry, flags & LOOKUP_PARENT); |
9883d185 | 2354 | restore_nameidata(); |
e4bd1c1a | 2355 | putname(name); |
170aa3d0 | 2356 | return retval; |
1da177e4 LT |
2357 | } |
2358 | ||
8bcb77fa | 2359 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
c8a53ee5 | 2360 | static int path_parentat(struct nameidata *nd, unsigned flags, |
391172c4 | 2361 | struct path *parent) |
8bcb77fa | 2362 | { |
c8a53ee5 | 2363 | const char *s = path_init(nd, flags); |
9b5858e9 | 2364 | int err = link_path_walk(s, nd); |
8bcb77fa AV |
2365 | if (!err) |
2366 | err = complete_walk(nd); | |
391172c4 AV |
2367 | if (!err) { |
2368 | *parent = nd->path; | |
2369 | nd->path.mnt = NULL; | |
2370 | nd->path.dentry = NULL; | |
2371 | } | |
2372 | terminate_walk(nd); | |
8bcb77fa AV |
2373 | return err; |
2374 | } | |
2375 | ||
5c31b6ce | 2376 | static struct filename *filename_parentat(int dfd, struct filename *name, |
391172c4 AV |
2377 | unsigned int flags, struct path *parent, |
2378 | struct qstr *last, int *type) | |
8bcb77fa AV |
2379 | { |
2380 | int retval; | |
9883d185 | 2381 | struct nameidata nd; |
8bcb77fa | 2382 | |
5c31b6ce AV |
2383 | if (IS_ERR(name)) |
2384 | return name; | |
9883d185 | 2385 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2386 | retval = path_parentat(&nd, flags | LOOKUP_RCU, parent); |
8bcb77fa | 2387 | if (unlikely(retval == -ECHILD)) |
c8a53ee5 | 2388 | retval = path_parentat(&nd, flags, parent); |
8bcb77fa | 2389 | if (unlikely(retval == -ESTALE)) |
c8a53ee5 | 2390 | retval = path_parentat(&nd, flags | LOOKUP_REVAL, parent); |
391172c4 AV |
2391 | if (likely(!retval)) { |
2392 | *last = nd.last; | |
2393 | *type = nd.last_type; | |
2394 | audit_inode(name, parent->dentry, LOOKUP_PARENT); | |
5c31b6ce AV |
2395 | } else { |
2396 | putname(name); | |
2397 | name = ERR_PTR(retval); | |
391172c4 | 2398 | } |
9883d185 | 2399 | restore_nameidata(); |
5c31b6ce | 2400 | return name; |
8bcb77fa AV |
2401 | } |
2402 | ||
79714f72 AV |
2403 | /* does lookup, returns the object with parent locked */ |
2404 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 2405 | { |
5c31b6ce AV |
2406 | struct filename *filename; |
2407 | struct dentry *d; | |
391172c4 AV |
2408 | struct qstr last; |
2409 | int type; | |
51689104 | 2410 | |
5c31b6ce AV |
2411 | filename = filename_parentat(AT_FDCWD, getname_kernel(name), 0, path, |
2412 | &last, &type); | |
51689104 PM |
2413 | if (IS_ERR(filename)) |
2414 | return ERR_CAST(filename); | |
5c31b6ce | 2415 | if (unlikely(type != LAST_NORM)) { |
391172c4 | 2416 | path_put(path); |
5c31b6ce AV |
2417 | putname(filename); |
2418 | return ERR_PTR(-EINVAL); | |
79714f72 | 2419 | } |
5955102c | 2420 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 2421 | d = __lookup_hash(&last, path->dentry, 0); |
79714f72 | 2422 | if (IS_ERR(d)) { |
5955102c | 2423 | inode_unlock(path->dentry->d_inode); |
391172c4 | 2424 | path_put(path); |
79714f72 | 2425 | } |
51689104 | 2426 | putname(filename); |
79714f72 | 2427 | return d; |
5590ff0d UD |
2428 | } |
2429 | ||
d1811465 AV |
2430 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2431 | { | |
abc9f5be AV |
2432 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2433 | flags, path, NULL); | |
d1811465 | 2434 | } |
4d359507 | 2435 | EXPORT_SYMBOL(kern_path); |
d1811465 | 2436 | |
16f18200 JJS |
2437 | /** |
2438 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2439 | * @dentry: pointer to dentry of the base directory | |
2440 | * @mnt: pointer to vfs mount of the base directory | |
2441 | * @name: pointer to file name | |
2442 | * @flags: lookup flags | |
e0a01249 | 2443 | * @path: pointer to struct path to fill |
16f18200 JJS |
2444 | */ |
2445 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2446 | const char *name, unsigned int flags, | |
e0a01249 | 2447 | struct path *path) |
16f18200 | 2448 | { |
9ad1aaa6 | 2449 | struct path root = {.mnt = mnt, .dentry = dentry}; |
9ad1aaa6 | 2450 | /* the first argument of filename_lookup() is ignored with root */ |
abc9f5be AV |
2451 | return filename_lookup(AT_FDCWD, getname_kernel(name), |
2452 | flags , path, &root); | |
16f18200 | 2453 | } |
4d359507 | 2454 | EXPORT_SYMBOL(vfs_path_lookup); |
16f18200 | 2455 | |
3c95f0dc AV |
2456 | static int lookup_one_len_common(const char *name, struct dentry *base, |
2457 | int len, struct qstr *this) | |
057f6c01 | 2458 | { |
3c95f0dc AV |
2459 | this->name = name; |
2460 | this->len = len; | |
2461 | this->hash = full_name_hash(base, name, len); | |
6a96ba54 | 2462 | if (!len) |
3c95f0dc | 2463 | return -EACCES; |
6a96ba54 | 2464 | |
21d8a15a AV |
2465 | if (unlikely(name[0] == '.')) { |
2466 | if (len < 2 || (len == 2 && name[1] == '.')) | |
3c95f0dc | 2467 | return -EACCES; |
21d8a15a AV |
2468 | } |
2469 | ||
6a96ba54 | 2470 | while (len--) { |
3c95f0dc | 2471 | unsigned int c = *(const unsigned char *)name++; |
6a96ba54 | 2472 | if (c == '/' || c == '\0') |
3c95f0dc | 2473 | return -EACCES; |
6a96ba54 | 2474 | } |
5a202bcd AV |
2475 | /* |
2476 | * See if the low-level filesystem might want | |
2477 | * to use its own hash.. | |
2478 | */ | |
2479 | if (base->d_flags & DCACHE_OP_HASH) { | |
3c95f0dc | 2480 | int err = base->d_op->d_hash(base, this); |
5a202bcd | 2481 | if (err < 0) |
3c95f0dc | 2482 | return err; |
5a202bcd | 2483 | } |
eead1911 | 2484 | |
3c95f0dc AV |
2485 | return inode_permission(base->d_inode, MAY_EXEC); |
2486 | } | |
2487 | ||
0da0b7fd DH |
2488 | /** |
2489 | * try_lookup_one_len - filesystem helper to lookup single pathname component | |
2490 | * @name: pathname component to lookup | |
2491 | * @base: base directory to lookup from | |
2492 | * @len: maximum length @len should be interpreted to | |
2493 | * | |
2494 | * Look up a dentry by name in the dcache, returning NULL if it does not | |
2495 | * currently exist. The function does not try to create a dentry. | |
2496 | * | |
2497 | * Note that this routine is purely a helper for filesystem usage and should | |
2498 | * not be called by generic code. | |
2499 | * | |
2500 | * The caller must hold base->i_mutex. | |
2501 | */ | |
2502 | struct dentry *try_lookup_one_len(const char *name, struct dentry *base, int len) | |
2503 | { | |
2504 | struct qstr this; | |
2505 | int err; | |
2506 | ||
2507 | WARN_ON_ONCE(!inode_is_locked(base->d_inode)); | |
2508 | ||
2509 | err = lookup_one_len_common(name, base, len, &this); | |
2510 | if (err) | |
2511 | return ERR_PTR(err); | |
2512 | ||
2513 | return lookup_dcache(&this, base, 0); | |
2514 | } | |
2515 | EXPORT_SYMBOL(try_lookup_one_len); | |
2516 | ||
3c95f0dc AV |
2517 | /** |
2518 | * lookup_one_len - filesystem helper to lookup single pathname component | |
2519 | * @name: pathname component to lookup | |
2520 | * @base: base directory to lookup from | |
2521 | * @len: maximum length @len should be interpreted to | |
2522 | * | |
2523 | * Note that this routine is purely a helper for filesystem usage and should | |
2524 | * not be called by generic code. | |
2525 | * | |
2526 | * The caller must hold base->i_mutex. | |
2527 | */ | |
2528 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) | |
2529 | { | |
8613a209 | 2530 | struct dentry *dentry; |
3c95f0dc AV |
2531 | struct qstr this; |
2532 | int err; | |
2533 | ||
2534 | WARN_ON_ONCE(!inode_is_locked(base->d_inode)); | |
2535 | ||
2536 | err = lookup_one_len_common(name, base, len, &this); | |
cda309de MS |
2537 | if (err) |
2538 | return ERR_PTR(err); | |
2539 | ||
8613a209 AV |
2540 | dentry = lookup_dcache(&this, base, 0); |
2541 | return dentry ? dentry : __lookup_slow(&this, base, 0); | |
057f6c01 | 2542 | } |
4d359507 | 2543 | EXPORT_SYMBOL(lookup_one_len); |
057f6c01 | 2544 | |
bbddca8e N |
2545 | /** |
2546 | * lookup_one_len_unlocked - filesystem helper to lookup single pathname component | |
2547 | * @name: pathname component to lookup | |
2548 | * @base: base directory to lookup from | |
2549 | * @len: maximum length @len should be interpreted to | |
2550 | * | |
2551 | * Note that this routine is purely a helper for filesystem usage and should | |
2552 | * not be called by generic code. | |
2553 | * | |
2554 | * Unlike lookup_one_len, it should be called without the parent | |
2555 | * i_mutex held, and will take the i_mutex itself if necessary. | |
2556 | */ | |
2557 | struct dentry *lookup_one_len_unlocked(const char *name, | |
2558 | struct dentry *base, int len) | |
2559 | { | |
2560 | struct qstr this; | |
bbddca8e | 2561 | int err; |
20d00ee8 | 2562 | struct dentry *ret; |
bbddca8e | 2563 | |
3c95f0dc | 2564 | err = lookup_one_len_common(name, base, len, &this); |
bbddca8e N |
2565 | if (err) |
2566 | return ERR_PTR(err); | |
2567 | ||
20d00ee8 LT |
2568 | ret = lookup_dcache(&this, base, 0); |
2569 | if (!ret) | |
2570 | ret = lookup_slow(&this, base, 0); | |
2571 | return ret; | |
bbddca8e N |
2572 | } |
2573 | EXPORT_SYMBOL(lookup_one_len_unlocked); | |
2574 | ||
eedf265a EB |
2575 | #ifdef CONFIG_UNIX98_PTYS |
2576 | int path_pts(struct path *path) | |
2577 | { | |
2578 | /* Find something mounted on "pts" in the same directory as | |
2579 | * the input path. | |
2580 | */ | |
2581 | struct dentry *child, *parent; | |
2582 | struct qstr this; | |
2583 | int ret; | |
2584 | ||
2585 | ret = path_parent_directory(path); | |
2586 | if (ret) | |
2587 | return ret; | |
2588 | ||
2589 | parent = path->dentry; | |
2590 | this.name = "pts"; | |
2591 | this.len = 3; | |
2592 | child = d_hash_and_lookup(parent, &this); | |
2593 | if (!child) | |
2594 | return -ENOENT; | |
2595 | ||
2596 | path->dentry = child; | |
2597 | dput(parent); | |
2598 | follow_mount(path); | |
2599 | return 0; | |
2600 | } | |
2601 | #endif | |
2602 | ||
1fa1e7f6 AW |
2603 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2604 | struct path *path, int *empty) | |
1da177e4 | 2605 | { |
abc9f5be AV |
2606 | return filename_lookup(dfd, getname_flags(name, flags, empty), |
2607 | flags, path, NULL); | |
1da177e4 | 2608 | } |
b853a161 | 2609 | EXPORT_SYMBOL(user_path_at_empty); |
1fa1e7f6 | 2610 | |
8033426e | 2611 | /** |
197df04c | 2612 | * mountpoint_last - look up last component for umount |
8033426e | 2613 | * @nd: pathwalk nameidata - currently pointing at parent directory of "last" |
8033426e JL |
2614 | * |
2615 | * This is a special lookup_last function just for umount. In this case, we | |
2616 | * need to resolve the path without doing any revalidation. | |
2617 | * | |
2618 | * The nameidata should be the result of doing a LOOKUP_PARENT pathwalk. Since | |
2619 | * mountpoints are always pinned in the dcache, their ancestors are too. Thus, | |
2620 | * in almost all cases, this lookup will be served out of the dcache. The only | |
2621 | * cases where it won't are if nd->last refers to a symlink or the path is | |
2622 | * bogus and it doesn't exist. | |
2623 | * | |
2624 | * Returns: | |
2625 | * -error: if there was an error during lookup. This includes -ENOENT if the | |
ba8f4613 | 2626 | * lookup found a negative dentry. |
8033426e | 2627 | * |
ba8f4613 AV |
2628 | * 0: if we successfully resolved nd->last and found it to not to be a |
2629 | * symlink that needs to be followed. | |
8033426e JL |
2630 | * |
2631 | * 1: if we successfully resolved nd->last and found it to be a symlink | |
ba8f4613 | 2632 | * that needs to be followed. |
8033426e JL |
2633 | */ |
2634 | static int | |
ba8f4613 | 2635 | mountpoint_last(struct nameidata *nd) |
8033426e JL |
2636 | { |
2637 | int error = 0; | |
8033426e | 2638 | struct dentry *dir = nd->path.dentry; |
ba8f4613 | 2639 | struct path path; |
8033426e | 2640 | |
35759521 AV |
2641 | /* If we're in rcuwalk, drop out of it to handle last component */ |
2642 | if (nd->flags & LOOKUP_RCU) { | |
4675ac39 | 2643 | if (unlazy_walk(nd)) |
deb106c6 | 2644 | return -ECHILD; |
8033426e JL |
2645 | } |
2646 | ||
2647 | nd->flags &= ~LOOKUP_PARENT; | |
2648 | ||
2649 | if (unlikely(nd->last_type != LAST_NORM)) { | |
2650 | error = handle_dots(nd, nd->last_type); | |
35759521 | 2651 | if (error) |
deb106c6 | 2652 | return error; |
ba8f4613 | 2653 | path.dentry = dget(nd->path.dentry); |
949a852e | 2654 | } else { |
ba8f4613 AV |
2655 | path.dentry = d_lookup(dir, &nd->last); |
2656 | if (!path.dentry) { | |
949a852e AV |
2657 | /* |
2658 | * No cached dentry. Mounted dentries are pinned in the | |
2659 | * cache, so that means that this dentry is probably | |
2660 | * a symlink or the path doesn't actually point | |
2661 | * to a mounted dentry. | |
2662 | */ | |
ba8f4613 | 2663 | path.dentry = lookup_slow(&nd->last, dir, |
949a852e | 2664 | nd->flags | LOOKUP_NO_REVAL); |
ba8f4613 AV |
2665 | if (IS_ERR(path.dentry)) |
2666 | return PTR_ERR(path.dentry); | |
bcceeeba | 2667 | } |
8033426e | 2668 | } |
ba8f4613 AV |
2669 | if (d_is_negative(path.dentry)) { |
2670 | dput(path.dentry); | |
deb106c6 | 2671 | return -ENOENT; |
8033426e | 2672 | } |
ba8f4613 | 2673 | path.mnt = nd->path.mnt; |
8f64fb1c | 2674 | return step_into(nd, &path, 0, d_backing_inode(path.dentry), 0); |
8033426e JL |
2675 | } |
2676 | ||
2677 | /** | |
197df04c | 2678 | * path_mountpoint - look up a path to be umounted |
2a78b857 | 2679 | * @nd: lookup context |
8033426e | 2680 | * @flags: lookup flags |
c8a53ee5 | 2681 | * @path: pointer to container for result |
8033426e JL |
2682 | * |
2683 | * Look up the given name, but don't attempt to revalidate the last component. | |
606d6fe3 | 2684 | * Returns 0 and "path" will be valid on success; Returns error otherwise. |
8033426e JL |
2685 | */ |
2686 | static int | |
c8a53ee5 | 2687 | path_mountpoint(struct nameidata *nd, unsigned flags, struct path *path) |
8033426e | 2688 | { |
c8a53ee5 | 2689 | const char *s = path_init(nd, flags); |
368ee9ba | 2690 | int err; |
9b5858e9 | 2691 | |
3bdba28b | 2692 | while (!(err = link_path_walk(s, nd)) && |
ba8f4613 | 2693 | (err = mountpoint_last(nd)) > 0) { |
3bdba28b | 2694 | s = trailing_symlink(nd); |
8033426e | 2695 | } |
ba8f4613 AV |
2696 | if (!err) { |
2697 | *path = nd->path; | |
2698 | nd->path.mnt = NULL; | |
2699 | nd->path.dentry = NULL; | |
2700 | follow_mount(path); | |
2701 | } | |
deb106c6 | 2702 | terminate_walk(nd); |
8033426e JL |
2703 | return err; |
2704 | } | |
2705 | ||
2d864651 | 2706 | static int |
668696dc | 2707 | filename_mountpoint(int dfd, struct filename *name, struct path *path, |
2d864651 AV |
2708 | unsigned int flags) |
2709 | { | |
9883d185 | 2710 | struct nameidata nd; |
cbaab2db | 2711 | int error; |
668696dc AV |
2712 | if (IS_ERR(name)) |
2713 | return PTR_ERR(name); | |
9883d185 | 2714 | set_nameidata(&nd, dfd, name); |
c8a53ee5 | 2715 | error = path_mountpoint(&nd, flags | LOOKUP_RCU, path); |
2d864651 | 2716 | if (unlikely(error == -ECHILD)) |
c8a53ee5 | 2717 | error = path_mountpoint(&nd, flags, path); |
2d864651 | 2718 | if (unlikely(error == -ESTALE)) |
c8a53ee5 | 2719 | error = path_mountpoint(&nd, flags | LOOKUP_REVAL, path); |
2d864651 | 2720 | if (likely(!error)) |
57d46577 | 2721 | audit_inode(name, path->dentry, flags & LOOKUP_NO_EVAL); |
9883d185 | 2722 | restore_nameidata(); |
668696dc | 2723 | putname(name); |
2d864651 AV |
2724 | return error; |
2725 | } | |
2726 | ||
8033426e | 2727 | /** |
197df04c | 2728 | * user_path_mountpoint_at - lookup a path from userland in order to umount it |
8033426e JL |
2729 | * @dfd: directory file descriptor |
2730 | * @name: pathname from userland | |
2731 | * @flags: lookup flags | |
2732 | * @path: pointer to container to hold result | |
2733 | * | |
2734 | * A umount is a special case for path walking. We're not actually interested | |
2735 | * in the inode in this situation, and ESTALE errors can be a problem. We | |
2736 | * simply want track down the dentry and vfsmount attached at the mountpoint | |
2737 | * and avoid revalidating the last component. | |
2738 | * | |
2739 | * Returns 0 and populates "path" on success. | |
2740 | */ | |
2741 | int | |
197df04c | 2742 | user_path_mountpoint_at(int dfd, const char __user *name, unsigned int flags, |
8033426e JL |
2743 | struct path *path) |
2744 | { | |
cbaab2db | 2745 | return filename_mountpoint(dfd, getname(name), path, flags); |
8033426e JL |
2746 | } |
2747 | ||
2d864651 AV |
2748 | int |
2749 | kern_path_mountpoint(int dfd, const char *name, struct path *path, | |
2750 | unsigned int flags) | |
2751 | { | |
cbaab2db | 2752 | return filename_mountpoint(dfd, getname_kernel(name), path, flags); |
2d864651 AV |
2753 | } |
2754 | EXPORT_SYMBOL(kern_path_mountpoint); | |
2755 | ||
cbdf35bc | 2756 | int __check_sticky(struct inode *dir, struct inode *inode) |
1da177e4 | 2757 | { |
8e96e3b7 | 2758 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2759 | |
8e96e3b7 | 2760 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2761 | return 0; |
8e96e3b7 | 2762 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2763 | return 0; |
23adbe12 | 2764 | return !capable_wrt_inode_uidgid(inode, CAP_FOWNER); |
1da177e4 | 2765 | } |
cbdf35bc | 2766 | EXPORT_SYMBOL(__check_sticky); |
1da177e4 LT |
2767 | |
2768 | /* | |
2769 | * Check whether we can remove a link victim from directory dir, check | |
2770 | * whether the type of victim is right. | |
2771 | * 1. We can't do it if dir is read-only (done in permission()) | |
2772 | * 2. We should have write and exec permissions on dir | |
2773 | * 3. We can't remove anything from append-only dir | |
2774 | * 4. We can't do anything with immutable dir (done in permission()) | |
2775 | * 5. If the sticky bit on dir is set we should either | |
2776 | * a. be owner of dir, or | |
2777 | * b. be owner of victim, or | |
2778 | * c. have CAP_FOWNER capability | |
2779 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2780 | * links pointing to it. | |
0bd23d09 EB |
2781 | * 7. If the victim has an unknown uid or gid we can't change the inode. |
2782 | * 8. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2783 | * 9. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2784 | * 10. We can't remove a root or mountpoint. | |
2785 | * 11. We don't allow removal of NFS sillyrenamed files; it's handled by | |
1da177e4 LT |
2786 | * nfs_async_unlink(). |
2787 | */ | |
b18825a7 | 2788 | static int may_delete(struct inode *dir, struct dentry *victim, bool isdir) |
1da177e4 | 2789 | { |
63afdfc7 | 2790 | struct inode *inode = d_backing_inode(victim); |
1da177e4 LT |
2791 | int error; |
2792 | ||
b18825a7 | 2793 | if (d_is_negative(victim)) |
1da177e4 | 2794 | return -ENOENT; |
b18825a7 | 2795 | BUG_ON(!inode); |
1da177e4 LT |
2796 | |
2797 | BUG_ON(victim->d_parent->d_inode != dir); | |
593d1ce8 EB |
2798 | |
2799 | /* Inode writeback is not safe when the uid or gid are invalid. */ | |
2800 | if (!uid_valid(inode->i_uid) || !gid_valid(inode->i_gid)) | |
2801 | return -EOVERFLOW; | |
2802 | ||
4fa6b5ec | 2803 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
1da177e4 | 2804 | |
f419a2e3 | 2805 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2806 | if (error) |
2807 | return error; | |
2808 | if (IS_APPEND(dir)) | |
2809 | return -EPERM; | |
b18825a7 DH |
2810 | |
2811 | if (check_sticky(dir, inode) || IS_APPEND(inode) || | |
0bd23d09 | 2812 | IS_IMMUTABLE(inode) || IS_SWAPFILE(inode) || HAS_UNMAPPED_ID(inode)) |
1da177e4 LT |
2813 | return -EPERM; |
2814 | if (isdir) { | |
44b1d530 | 2815 | if (!d_is_dir(victim)) |
1da177e4 LT |
2816 | return -ENOTDIR; |
2817 | if (IS_ROOT(victim)) | |
2818 | return -EBUSY; | |
44b1d530 | 2819 | } else if (d_is_dir(victim)) |
1da177e4 LT |
2820 | return -EISDIR; |
2821 | if (IS_DEADDIR(dir)) | |
2822 | return -ENOENT; | |
2823 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2824 | return -EBUSY; | |
2825 | return 0; | |
2826 | } | |
2827 | ||
2828 | /* Check whether we can create an object with dentry child in directory | |
2829 | * dir. | |
2830 | * 1. We can't do it if child already exists (open has special treatment for | |
2831 | * this case, but since we are inlined it's OK) | |
2832 | * 2. We can't do it if dir is read-only (done in permission()) | |
036d5236 EB |
2833 | * 3. We can't do it if the fs can't represent the fsuid or fsgid. |
2834 | * 4. We should have write and exec permissions on dir | |
2835 | * 5. We can't do it if dir is immutable (done in permission()) | |
1da177e4 | 2836 | */ |
a95164d9 | 2837 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 | 2838 | { |
036d5236 | 2839 | struct user_namespace *s_user_ns; |
14e972b4 | 2840 | audit_inode_child(dir, child, AUDIT_TYPE_CHILD_CREATE); |
1da177e4 LT |
2841 | if (child->d_inode) |
2842 | return -EEXIST; | |
2843 | if (IS_DEADDIR(dir)) | |
2844 | return -ENOENT; | |
036d5236 EB |
2845 | s_user_ns = dir->i_sb->s_user_ns; |
2846 | if (!kuid_has_mapping(s_user_ns, current_fsuid()) || | |
2847 | !kgid_has_mapping(s_user_ns, current_fsgid())) | |
2848 | return -EOVERFLOW; | |
f419a2e3 | 2849 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2850 | } |
2851 | ||
1da177e4 LT |
2852 | /* |
2853 | * p1 and p2 should be directories on the same fs. | |
2854 | */ | |
2855 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2856 | { | |
2857 | struct dentry *p; | |
2858 | ||
2859 | if (p1 == p2) { | |
5955102c | 2860 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
1da177e4 LT |
2861 | return NULL; |
2862 | } | |
2863 | ||
fc64005c | 2864 | mutex_lock(&p1->d_sb->s_vfs_rename_mutex); |
1da177e4 | 2865 | |
e2761a11 OH |
2866 | p = d_ancestor(p2, p1); |
2867 | if (p) { | |
5955102c AV |
2868 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT); |
2869 | inode_lock_nested(p1->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2870 | return p; |
1da177e4 LT |
2871 | } |
2872 | ||
e2761a11 OH |
2873 | p = d_ancestor(p1, p2); |
2874 | if (p) { | |
5955102c AV |
2875 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2876 | inode_lock_nested(p2->d_inode, I_MUTEX_CHILD); | |
e2761a11 | 2877 | return p; |
1da177e4 LT |
2878 | } |
2879 | ||
5955102c AV |
2880 | inode_lock_nested(p1->d_inode, I_MUTEX_PARENT); |
2881 | inode_lock_nested(p2->d_inode, I_MUTEX_PARENT2); | |
1da177e4 LT |
2882 | return NULL; |
2883 | } | |
4d359507 | 2884 | EXPORT_SYMBOL(lock_rename); |
1da177e4 LT |
2885 | |
2886 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2887 | { | |
5955102c | 2888 | inode_unlock(p1->d_inode); |
1da177e4 | 2889 | if (p1 != p2) { |
5955102c | 2890 | inode_unlock(p2->d_inode); |
fc64005c | 2891 | mutex_unlock(&p1->d_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2892 | } |
2893 | } | |
4d359507 | 2894 | EXPORT_SYMBOL(unlock_rename); |
1da177e4 | 2895 | |
4acdaf27 | 2896 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2897 | bool want_excl) |
1da177e4 | 2898 | { |
a95164d9 | 2899 | int error = may_create(dir, dentry); |
1da177e4 LT |
2900 | if (error) |
2901 | return error; | |
2902 | ||
acfa4380 | 2903 | if (!dir->i_op->create) |
1da177e4 LT |
2904 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2905 | mode &= S_IALLUGO; | |
2906 | mode |= S_IFREG; | |
2907 | error = security_inode_create(dir, dentry, mode); | |
2908 | if (error) | |
2909 | return error; | |
312b63fb | 2910 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2911 | if (!error) |
f38aa942 | 2912 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2913 | return error; |
2914 | } | |
4d359507 | 2915 | EXPORT_SYMBOL(vfs_create); |
1da177e4 | 2916 | |
8e6c848e AV |
2917 | int vfs_mkobj(struct dentry *dentry, umode_t mode, |
2918 | int (*f)(struct dentry *, umode_t, void *), | |
2919 | void *arg) | |
2920 | { | |
2921 | struct inode *dir = dentry->d_parent->d_inode; | |
2922 | int error = may_create(dir, dentry); | |
2923 | if (error) | |
2924 | return error; | |
2925 | ||
2926 | mode &= S_IALLUGO; | |
2927 | mode |= S_IFREG; | |
2928 | error = security_inode_create(dir, dentry, mode); | |
2929 | if (error) | |
2930 | return error; | |
2931 | error = f(dentry, mode, arg); | |
2932 | if (!error) | |
2933 | fsnotify_create(dir, dentry); | |
2934 | return error; | |
2935 | } | |
2936 | EXPORT_SYMBOL(vfs_mkobj); | |
2937 | ||
a2982cc9 EB |
2938 | bool may_open_dev(const struct path *path) |
2939 | { | |
2940 | return !(path->mnt->mnt_flags & MNT_NODEV) && | |
2941 | !(path->mnt->mnt_sb->s_iflags & SB_I_NODEV); | |
2942 | } | |
2943 | ||
f0bb5aaf | 2944 | static int may_open(const struct path *path, int acc_mode, int flag) |
1da177e4 | 2945 | { |
3fb64190 | 2946 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2947 | struct inode *inode = dentry->d_inode; |
2948 | int error; | |
2949 | ||
2950 | if (!inode) | |
2951 | return -ENOENT; | |
2952 | ||
c8fe8f30 CH |
2953 | switch (inode->i_mode & S_IFMT) { |
2954 | case S_IFLNK: | |
1da177e4 | 2955 | return -ELOOP; |
c8fe8f30 CH |
2956 | case S_IFDIR: |
2957 | if (acc_mode & MAY_WRITE) | |
2958 | return -EISDIR; | |
2959 | break; | |
2960 | case S_IFBLK: | |
2961 | case S_IFCHR: | |
a2982cc9 | 2962 | if (!may_open_dev(path)) |
1da177e4 | 2963 | return -EACCES; |
c8fe8f30 CH |
2964 | /*FALLTHRU*/ |
2965 | case S_IFIFO: | |
2966 | case S_IFSOCK: | |
1da177e4 | 2967 | flag &= ~O_TRUNC; |
c8fe8f30 | 2968 | break; |
4a3fd211 | 2969 | } |
b41572e9 | 2970 | |
62fb4a15 | 2971 | error = inode_permission(inode, MAY_OPEN | acc_mode); |
b41572e9 DH |
2972 | if (error) |
2973 | return error; | |
6146f0d5 | 2974 | |
1da177e4 LT |
2975 | /* |
2976 | * An append-only file must be opened in append mode for writing. | |
2977 | */ | |
2978 | if (IS_APPEND(inode)) { | |
8737c930 | 2979 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2980 | return -EPERM; |
1da177e4 | 2981 | if (flag & O_TRUNC) |
7715b521 | 2982 | return -EPERM; |
1da177e4 LT |
2983 | } |
2984 | ||
2985 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2986 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2987 | return -EPERM; |
1da177e4 | 2988 | |
f3c7691e | 2989 | return 0; |
7715b521 | 2990 | } |
1da177e4 | 2991 | |
e1181ee6 | 2992 | static int handle_truncate(struct file *filp) |
7715b521 | 2993 | { |
f0bb5aaf | 2994 | const struct path *path = &filp->f_path; |
7715b521 AV |
2995 | struct inode *inode = path->dentry->d_inode; |
2996 | int error = get_write_access(inode); | |
2997 | if (error) | |
2998 | return error; | |
2999 | /* | |
3000 | * Refuse to truncate files with mandatory locks held on them. | |
3001 | */ | |
d7a06983 | 3002 | error = locks_verify_locked(filp); |
7715b521 | 3003 | if (!error) |
ea0d3ab2 | 3004 | error = security_path_truncate(path); |
7715b521 AV |
3005 | if (!error) { |
3006 | error = do_truncate(path->dentry, 0, | |
3007 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 3008 | filp); |
7715b521 AV |
3009 | } |
3010 | put_write_access(inode); | |
acd0c935 | 3011 | return error; |
1da177e4 LT |
3012 | } |
3013 | ||
d57999e1 DH |
3014 | static inline int open_to_namei_flags(int flag) |
3015 | { | |
8a5e929d AV |
3016 | if ((flag & O_ACCMODE) == 3) |
3017 | flag--; | |
d57999e1 DH |
3018 | return flag; |
3019 | } | |
3020 | ||
d3607752 | 3021 | static int may_o_create(const struct path *dir, struct dentry *dentry, umode_t mode) |
d18e9008 | 3022 | { |
1328c727 | 3023 | struct user_namespace *s_user_ns; |
d18e9008 MS |
3024 | int error = security_path_mknod(dir, dentry, mode, 0); |
3025 | if (error) | |
3026 | return error; | |
3027 | ||
1328c727 SF |
3028 | s_user_ns = dir->dentry->d_sb->s_user_ns; |
3029 | if (!kuid_has_mapping(s_user_ns, current_fsuid()) || | |
3030 | !kgid_has_mapping(s_user_ns, current_fsgid())) | |
3031 | return -EOVERFLOW; | |
3032 | ||
d18e9008 MS |
3033 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); |
3034 | if (error) | |
3035 | return error; | |
3036 | ||
3037 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
3038 | } | |
3039 | ||
1acf0af9 DH |
3040 | /* |
3041 | * Attempt to atomically look up, create and open a file from a negative | |
3042 | * dentry. | |
3043 | * | |
3044 | * Returns 0 if successful. The file will have been created and attached to | |
3045 | * @file by the filesystem calling finish_open(). | |
3046 | * | |
00a07c15 AV |
3047 | * If the file was looked up only or didn't need creating, FMODE_OPENED won't |
3048 | * be set. The caller will need to perform the open themselves. @path will | |
3049 | * have been updated to point to the new dentry. This may be negative. | |
1acf0af9 DH |
3050 | * |
3051 | * Returns an error code otherwise. | |
3052 | */ | |
2675a4eb AV |
3053 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
3054 | struct path *path, struct file *file, | |
3055 | const struct open_flags *op, | |
3ec2eef1 | 3056 | int open_flag, umode_t mode) |
d18e9008 | 3057 | { |
384f26e2 | 3058 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; |
d18e9008 | 3059 | struct inode *dir = nd->path.dentry->d_inode; |
d18e9008 | 3060 | int error; |
d18e9008 | 3061 | |
384f26e2 | 3062 | if (!(~open_flag & (O_EXCL | O_CREAT))) /* both O_EXCL and O_CREAT */ |
d18e9008 | 3063 | open_flag &= ~O_TRUNC; |
d18e9008 | 3064 | |
d18e9008 MS |
3065 | if (nd->flags & LOOKUP_DIRECTORY) |
3066 | open_flag |= O_DIRECTORY; | |
3067 | ||
30d90494 AV |
3068 | file->f_path.dentry = DENTRY_NOT_SET; |
3069 | file->f_path.mnt = nd->path.mnt; | |
0fb1ea09 | 3070 | error = dir->i_op->atomic_open(dir, dentry, file, |
44907d79 | 3071 | open_to_namei_flags(open_flag), mode); |
6fbd0714 | 3072 | d_lookup_done(dentry); |
384f26e2 | 3073 | if (!error) { |
64e1ac4d AV |
3074 | if (file->f_mode & FMODE_OPENED) { |
3075 | /* | |
3076 | * We didn't have the inode before the open, so check open | |
3077 | * permission here. | |
3078 | */ | |
3079 | int acc_mode = op->acc_mode; | |
3080 | if (file->f_mode & FMODE_CREATED) { | |
3081 | WARN_ON(!(open_flag & O_CREAT)); | |
3082 | fsnotify_create(dir, dentry); | |
3083 | acc_mode = 0; | |
3084 | } | |
3085 | error = may_open(&file->f_path, acc_mode, open_flag); | |
3086 | if (WARN_ON(error > 0)) | |
3087 | error = -EINVAL; | |
3088 | } else if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { | |
2675a4eb | 3089 | error = -EIO; |
03da633a | 3090 | } else { |
384f26e2 AV |
3091 | if (file->f_path.dentry) { |
3092 | dput(dentry); | |
3093 | dentry = file->f_path.dentry; | |
03da633a | 3094 | } |
73a09dd9 | 3095 | if (file->f_mode & FMODE_CREATED) |
384f26e2 | 3096 | fsnotify_create(dir, dentry); |
a01e718f AV |
3097 | if (unlikely(d_is_negative(dentry))) { |
3098 | error = -ENOENT; | |
3099 | } else { | |
3100 | path->dentry = dentry; | |
3101 | path->mnt = nd->path.mnt; | |
00a07c15 | 3102 | return 0; |
a01e718f | 3103 | } |
62b2ce96 | 3104 | } |
d18e9008 | 3105 | } |
d18e9008 | 3106 | dput(dentry); |
2675a4eb | 3107 | return error; |
d18e9008 MS |
3108 | } |
3109 | ||
d58ffd35 | 3110 | /* |
1acf0af9 | 3111 | * Look up and maybe create and open the last component. |
d58ffd35 | 3112 | * |
00a07c15 | 3113 | * Must be called with parent locked (exclusive in O_CREAT case). |
1acf0af9 | 3114 | * |
00a07c15 AV |
3115 | * Returns 0 on success, that is, if |
3116 | * the file was successfully atomically created (if necessary) and opened, or | |
3117 | * the file was not completely opened at this time, though lookups and | |
3118 | * creations were performed. | |
3119 | * These case are distinguished by presence of FMODE_OPENED on file->f_mode. | |
3120 | * In the latter case dentry returned in @path might be negative if O_CREAT | |
3121 | * hadn't been specified. | |
1acf0af9 | 3122 | * |
00a07c15 | 3123 | * An error code is returned on failure. |
d58ffd35 | 3124 | */ |
2675a4eb AV |
3125 | static int lookup_open(struct nameidata *nd, struct path *path, |
3126 | struct file *file, | |
3127 | const struct open_flags *op, | |
3ec2eef1 | 3128 | bool got_write) |
d58ffd35 MS |
3129 | { |
3130 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 3131 | struct inode *dir_inode = dir->d_inode; |
1643b43f | 3132 | int open_flag = op->open_flag; |
d58ffd35 | 3133 | struct dentry *dentry; |
1643b43f | 3134 | int error, create_error = 0; |
1643b43f | 3135 | umode_t mode = op->mode; |
6fbd0714 | 3136 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
d58ffd35 | 3137 | |
ce8644fc AV |
3138 | if (unlikely(IS_DEADDIR(dir_inode))) |
3139 | return -ENOENT; | |
d58ffd35 | 3140 | |
73a09dd9 | 3141 | file->f_mode &= ~FMODE_CREATED; |
6fbd0714 AV |
3142 | dentry = d_lookup(dir, &nd->last); |
3143 | for (;;) { | |
3144 | if (!dentry) { | |
3145 | dentry = d_alloc_parallel(dir, &nd->last, &wq); | |
3146 | if (IS_ERR(dentry)) | |
3147 | return PTR_ERR(dentry); | |
3148 | } | |
3149 | if (d_in_lookup(dentry)) | |
3150 | break; | |
d58ffd35 | 3151 | |
6fbd0714 AV |
3152 | error = d_revalidate(dentry, nd->flags); |
3153 | if (likely(error > 0)) | |
3154 | break; | |
3155 | if (error) | |
3156 | goto out_dput; | |
3157 | d_invalidate(dentry); | |
3158 | dput(dentry); | |
3159 | dentry = NULL; | |
3160 | } | |
3161 | if (dentry->d_inode) { | |
6c51e513 | 3162 | /* Cached positive dentry: will open in f_op->open */ |
d18e9008 | 3163 | goto out_no_open; |
6c51e513 | 3164 | } |
d18e9008 | 3165 | |
1643b43f AV |
3166 | /* |
3167 | * Checking write permission is tricky, bacuse we don't know if we are | |
3168 | * going to actually need it: O_CREAT opens should work as long as the | |
3169 | * file exists. But checking existence breaks atomicity. The trick is | |
3170 | * to check access and if not granted clear O_CREAT from the flags. | |
3171 | * | |
3172 | * Another problem is returing the "right" error value (e.g. for an | |
3173 | * O_EXCL open we want to return EEXIST not EROFS). | |
3174 | */ | |
3175 | if (open_flag & O_CREAT) { | |
3176 | if (!IS_POSIXACL(dir->d_inode)) | |
3177 | mode &= ~current_umask(); | |
3178 | if (unlikely(!got_write)) { | |
3179 | create_error = -EROFS; | |
3180 | open_flag &= ~O_CREAT; | |
3181 | if (open_flag & (O_EXCL | O_TRUNC)) | |
3182 | goto no_open; | |
3183 | /* No side effects, safe to clear O_CREAT */ | |
3184 | } else { | |
3185 | create_error = may_o_create(&nd->path, dentry, mode); | |
3186 | if (create_error) { | |
3187 | open_flag &= ~O_CREAT; | |
3188 | if (open_flag & O_EXCL) | |
3189 | goto no_open; | |
3190 | } | |
3191 | } | |
3192 | } else if ((open_flag & (O_TRUNC|O_WRONLY|O_RDWR)) && | |
3193 | unlikely(!got_write)) { | |
3194 | /* | |
3195 | * No O_CREATE -> atomicity not a requirement -> fall | |
3196 | * back to lookup + open | |
3197 | */ | |
3198 | goto no_open; | |
d18e9008 MS |
3199 | } |
3200 | ||
6ac08709 | 3201 | if (dir_inode->i_op->atomic_open) { |
1643b43f | 3202 | error = atomic_open(nd, dentry, path, file, op, open_flag, |
3ec2eef1 | 3203 | mode); |
1643b43f AV |
3204 | if (unlikely(error == -ENOENT) && create_error) |
3205 | error = create_error; | |
3206 | return error; | |
d18e9008 | 3207 | } |
54ef4872 | 3208 | |
1643b43f | 3209 | no_open: |
6fbd0714 | 3210 | if (d_in_lookup(dentry)) { |
12fa5e24 AV |
3211 | struct dentry *res = dir_inode->i_op->lookup(dir_inode, dentry, |
3212 | nd->flags); | |
6fbd0714 | 3213 | d_lookup_done(dentry); |
12fa5e24 AV |
3214 | if (unlikely(res)) { |
3215 | if (IS_ERR(res)) { | |
3216 | error = PTR_ERR(res); | |
3217 | goto out_dput; | |
3218 | } | |
3219 | dput(dentry); | |
3220 | dentry = res; | |
3221 | } | |
54ef4872 MS |
3222 | } |
3223 | ||
d58ffd35 | 3224 | /* Negative dentry, just create the file */ |
1643b43f | 3225 | if (!dentry->d_inode && (open_flag & O_CREAT)) { |
73a09dd9 | 3226 | file->f_mode |= FMODE_CREATED; |
ce8644fc | 3227 | audit_inode_child(dir_inode, dentry, AUDIT_TYPE_CHILD_CREATE); |
ce8644fc AV |
3228 | if (!dir_inode->i_op->create) { |
3229 | error = -EACCES; | |
d58ffd35 | 3230 | goto out_dput; |
ce8644fc AV |
3231 | } |
3232 | error = dir_inode->i_op->create(dir_inode, dentry, mode, | |
1643b43f | 3233 | open_flag & O_EXCL); |
d58ffd35 MS |
3234 | if (error) |
3235 | goto out_dput; | |
ce8644fc | 3236 | fsnotify_create(dir_inode, dentry); |
d58ffd35 | 3237 | } |
1643b43f AV |
3238 | if (unlikely(create_error) && !dentry->d_inode) { |
3239 | error = create_error; | |
3240 | goto out_dput; | |
d58ffd35 | 3241 | } |
d18e9008 | 3242 | out_no_open: |
d58ffd35 MS |
3243 | path->dentry = dentry; |
3244 | path->mnt = nd->path.mnt; | |
00a07c15 | 3245 | return 0; |
d58ffd35 MS |
3246 | |
3247 | out_dput: | |
3248 | dput(dentry); | |
2675a4eb | 3249 | return error; |
d58ffd35 MS |
3250 | } |
3251 | ||
31e6b01f | 3252 | /* |
fe2d35ff | 3253 | * Handle the last step of open() |
31e6b01f | 3254 | */ |
896475d5 | 3255 | static int do_last(struct nameidata *nd, |
3ec2eef1 | 3256 | struct file *file, const struct open_flags *op) |
fb1cc555 | 3257 | { |
a1e28038 | 3258 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 3259 | int open_flag = op->open_flag; |
77d660a8 | 3260 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 3261 | bool got_write = false; |
bcda7652 | 3262 | int acc_mode = op->acc_mode; |
254cf582 | 3263 | unsigned seq; |
a1eb3315 | 3264 | struct inode *inode; |
896475d5 | 3265 | struct path path; |
16c2cd71 | 3266 | int error; |
1f36f774 | 3267 | |
c3e380b0 AV |
3268 | nd->flags &= ~LOOKUP_PARENT; |
3269 | nd->flags |= op->intent; | |
3270 | ||
bc77daa7 | 3271 | if (nd->last_type != LAST_NORM) { |
fe2d35ff | 3272 | error = handle_dots(nd, nd->last_type); |
deb106c6 | 3273 | if (unlikely(error)) |
2675a4eb | 3274 | return error; |
e83db167 | 3275 | goto finish_open; |
1f36f774 | 3276 | } |
67ee3ad2 | 3277 | |
ca344a89 | 3278 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
3279 | if (nd->last.name[nd->last.len]) |
3280 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
3281 | /* we _can_ be in RCU mode here */ | |
254cf582 | 3282 | error = lookup_fast(nd, &path, &inode, &seq); |
e9742b53 | 3283 | if (likely(error > 0)) |
71574865 MS |
3284 | goto finish_lookup; |
3285 | ||
3286 | if (error < 0) | |
deb106c6 | 3287 | return error; |
71574865 MS |
3288 | |
3289 | BUG_ON(nd->inode != dir->d_inode); | |
6583fe22 | 3290 | BUG_ON(nd->flags & LOOKUP_RCU); |
b6183df7 MS |
3291 | } else { |
3292 | /* create side of things */ | |
3293 | /* | |
3294 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
3295 | * has been cleared when we got to the last component we are | |
3296 | * about to look up | |
3297 | */ | |
3298 | error = complete_walk(nd); | |
e8bb73df | 3299 | if (error) |
2675a4eb | 3300 | return error; |
fe2d35ff | 3301 | |
76ae2a5a | 3302 | audit_inode(nd->name, dir, LOOKUP_PARENT); |
b6183df7 | 3303 | /* trailing slashes? */ |
deb106c6 AV |
3304 | if (unlikely(nd->last.name[nd->last.len])) |
3305 | return -EISDIR; | |
b6183df7 | 3306 | } |
a2c36b45 | 3307 | |
9cf843e3 | 3308 | if (open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
64894cf8 AV |
3309 | error = mnt_want_write(nd->path.mnt); |
3310 | if (!error) | |
3311 | got_write = true; | |
3312 | /* | |
3313 | * do _not_ fail yet - we might not need that or fail with | |
3314 | * a different error; let lookup_open() decide; we'll be | |
3315 | * dropping this one anyway. | |
3316 | */ | |
3317 | } | |
9cf843e3 AV |
3318 | if (open_flag & O_CREAT) |
3319 | inode_lock(dir->d_inode); | |
3320 | else | |
3321 | inode_lock_shared(dir->d_inode); | |
3ec2eef1 | 3322 | error = lookup_open(nd, &path, file, op, got_write); |
9cf843e3 AV |
3323 | if (open_flag & O_CREAT) |
3324 | inode_unlock(dir->d_inode); | |
3325 | else | |
3326 | inode_unlock_shared(dir->d_inode); | |
a1e28038 | 3327 | |
00a07c15 AV |
3328 | if (error) |
3329 | goto out; | |
d18e9008 | 3330 | |
00a07c15 | 3331 | if (file->f_mode & FMODE_OPENED) { |
73a09dd9 | 3332 | if ((file->f_mode & FMODE_CREATED) || |
496ad9aa | 3333 | !S_ISREG(file_inode(file)->i_mode)) |
77d660a8 | 3334 | will_truncate = false; |
d18e9008 | 3335 | |
76ae2a5a | 3336 | audit_inode(nd->name, file->f_path.dentry, 0); |
d18e9008 MS |
3337 | goto opened; |
3338 | } | |
fb1cc555 | 3339 | |
73a09dd9 | 3340 | if (file->f_mode & FMODE_CREATED) { |
9b44f1b3 | 3341 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 3342 | open_flag &= ~O_TRUNC; |
77d660a8 | 3343 | will_truncate = false; |
62fb4a15 | 3344 | acc_mode = 0; |
896475d5 | 3345 | path_to_nameidata(&path, nd); |
e83db167 | 3346 | goto finish_open_created; |
fb1cc555 AV |
3347 | } |
3348 | ||
d18e9008 MS |
3349 | /* |
3350 | * If atomic_open() acquired write access it is dropped now due to | |
3351 | * possible mount and symlink following (this might be optimized away if | |
3352 | * necessary...) | |
3353 | */ | |
64894cf8 | 3354 | if (got_write) { |
d18e9008 | 3355 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3356 | got_write = false; |
d18e9008 MS |
3357 | } |
3358 | ||
e6ec03a2 AV |
3359 | error = follow_managed(&path, nd); |
3360 | if (unlikely(error < 0)) | |
3361 | return error; | |
3362 | ||
6583fe22 AV |
3363 | if (unlikely(d_is_negative(path.dentry))) { |
3364 | path_to_nameidata(&path, nd); | |
3365 | return -ENOENT; | |
3366 | } | |
3367 | ||
3368 | /* | |
3369 | * create/update audit record if it already exists. | |
3370 | */ | |
3371 | audit_inode(nd->name, path.dentry, 0); | |
3372 | ||
deb106c6 AV |
3373 | if (unlikely((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT))) { |
3374 | path_to_nameidata(&path, nd); | |
3375 | return -EEXIST; | |
3376 | } | |
fb1cc555 | 3377 | |
254cf582 | 3378 | seq = 0; /* out of RCU mode, so the value doesn't matter */ |
d4565649 | 3379 | inode = d_backing_inode(path.dentry); |
766c4cbf | 3380 | finish_lookup: |
8f64fb1c | 3381 | error = step_into(nd, &path, 0, inode, seq); |
deb106c6 | 3382 | if (unlikely(error)) |
d63ff28f | 3383 | return error; |
bc77daa7 | 3384 | finish_open: |
8f64fb1c | 3385 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
a3fbbde7 | 3386 | error = complete_walk(nd); |
fac7d191 | 3387 | if (error) |
2675a4eb | 3388 | return error; |
76ae2a5a | 3389 | audit_inode(nd->name, nd->path.dentry, 0); |
30aba665 SM |
3390 | if (open_flag & O_CREAT) { |
3391 | error = -EISDIR; | |
3392 | if (d_is_dir(nd->path.dentry)) | |
3393 | goto out; | |
3394 | error = may_create_in_sticky(dir, | |
3395 | d_backing_inode(nd->path.dentry)); | |
3396 | if (unlikely(error)) | |
3397 | goto out; | |
3398 | } | |
af2f5542 | 3399 | error = -ENOTDIR; |
44b1d530 | 3400 | if ((nd->flags & LOOKUP_DIRECTORY) && !d_can_lookup(nd->path.dentry)) |
2675a4eb | 3401 | goto out; |
4bbcbd3b | 3402 | if (!d_is_reg(nd->path.dentry)) |
77d660a8 | 3403 | will_truncate = false; |
6c0d46c4 | 3404 | |
0f9d1a10 AV |
3405 | if (will_truncate) { |
3406 | error = mnt_want_write(nd->path.mnt); | |
3407 | if (error) | |
2675a4eb | 3408 | goto out; |
64894cf8 | 3409 | got_write = true; |
0f9d1a10 | 3410 | } |
e83db167 | 3411 | finish_open_created: |
6ac08709 AV |
3412 | error = may_open(&nd->path, acc_mode, open_flag); |
3413 | if (error) | |
3414 | goto out; | |
aad888f8 | 3415 | BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */ |
ae2bb293 | 3416 | error = vfs_open(&nd->path, file); |
fac7d191 | 3417 | if (error) |
015c3bbc | 3418 | goto out; |
a8277b9b | 3419 | opened: |
6035a27b | 3420 | error = ima_file_check(file, op->acc_mode); |
fe9ec829 | 3421 | if (!error && will_truncate) |
2675a4eb | 3422 | error = handle_truncate(file); |
ca344a89 | 3423 | out: |
c80567c8 AV |
3424 | if (unlikely(error > 0)) { |
3425 | WARN_ON(1); | |
3426 | error = -EINVAL; | |
3427 | } | |
64894cf8 | 3428 | if (got_write) |
0f9d1a10 | 3429 | mnt_drop_write(nd->path.mnt); |
2675a4eb | 3430 | return error; |
fb1cc555 AV |
3431 | } |
3432 | ||
af7bd4dc AG |
3433 | struct dentry *vfs_tmpfile(struct dentry *dentry, umode_t mode, int open_flag) |
3434 | { | |
af7bd4dc AG |
3435 | struct dentry *child = NULL; |
3436 | struct inode *dir = dentry->d_inode; | |
3437 | struct inode *inode; | |
3438 | int error; | |
3439 | ||
3440 | /* we want directory to be writable */ | |
3441 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); | |
3442 | if (error) | |
3443 | goto out_err; | |
3444 | error = -EOPNOTSUPP; | |
3445 | if (!dir->i_op->tmpfile) | |
3446 | goto out_err; | |
3447 | error = -ENOMEM; | |
cdf01226 | 3448 | child = d_alloc(dentry, &slash_name); |
af7bd4dc AG |
3449 | if (unlikely(!child)) |
3450 | goto out_err; | |
3451 | error = dir->i_op->tmpfile(dir, child, mode); | |
3452 | if (error) | |
3453 | goto out_err; | |
3454 | error = -ENOENT; | |
3455 | inode = child->d_inode; | |
3456 | if (unlikely(!inode)) | |
3457 | goto out_err; | |
3458 | if (!(open_flag & O_EXCL)) { | |
3459 | spin_lock(&inode->i_lock); | |
3460 | inode->i_state |= I_LINKABLE; | |
3461 | spin_unlock(&inode->i_lock); | |
3462 | } | |
3463 | return child; | |
3464 | ||
3465 | out_err: | |
3466 | dput(child); | |
3467 | return ERR_PTR(error); | |
3468 | } | |
3469 | EXPORT_SYMBOL(vfs_tmpfile); | |
3470 | ||
c8a53ee5 | 3471 | static int do_tmpfile(struct nameidata *nd, unsigned flags, |
60545d0d | 3472 | const struct open_flags *op, |
3ec2eef1 | 3473 | struct file *file) |
60545d0d | 3474 | { |
625b6d10 | 3475 | struct dentry *child; |
625b6d10 | 3476 | struct path path; |
c8a53ee5 | 3477 | int error = path_lookupat(nd, flags | LOOKUP_DIRECTORY, &path); |
60545d0d AV |
3478 | if (unlikely(error)) |
3479 | return error; | |
625b6d10 | 3480 | error = mnt_want_write(path.mnt); |
60545d0d AV |
3481 | if (unlikely(error)) |
3482 | goto out; | |
af7bd4dc AG |
3483 | child = vfs_tmpfile(path.dentry, op->mode, op->open_flag); |
3484 | error = PTR_ERR(child); | |
684e73be | 3485 | if (IS_ERR(child)) |
60545d0d | 3486 | goto out2; |
625b6d10 AV |
3487 | dput(path.dentry); |
3488 | path.dentry = child; | |
c8a53ee5 | 3489 | audit_inode(nd->name, child, 0); |
69a91c23 | 3490 | /* Don't check for other permissions, the inode was just created */ |
62fb4a15 | 3491 | error = may_open(&path, 0, op->open_flag); |
60545d0d AV |
3492 | if (error) |
3493 | goto out2; | |
625b6d10 | 3494 | file->f_path.mnt = path.mnt; |
be12af3e | 3495 | error = finish_open(file, child, NULL); |
60545d0d | 3496 | out2: |
625b6d10 | 3497 | mnt_drop_write(path.mnt); |
60545d0d | 3498 | out: |
625b6d10 | 3499 | path_put(&path); |
60545d0d AV |
3500 | return error; |
3501 | } | |
3502 | ||
6ac08709 AV |
3503 | static int do_o_path(struct nameidata *nd, unsigned flags, struct file *file) |
3504 | { | |
3505 | struct path path; | |
3506 | int error = path_lookupat(nd, flags, &path); | |
3507 | if (!error) { | |
3508 | audit_inode(nd->name, path.dentry, 0); | |
ae2bb293 | 3509 | error = vfs_open(&path, file); |
6ac08709 AV |
3510 | path_put(&path); |
3511 | } | |
3512 | return error; | |
3513 | } | |
3514 | ||
c8a53ee5 AV |
3515 | static struct file *path_openat(struct nameidata *nd, |
3516 | const struct open_flags *op, unsigned flags) | |
1da177e4 | 3517 | { |
30d90494 | 3518 | struct file *file; |
13aab428 | 3519 | int error; |
31e6b01f | 3520 | |
ea73ea72 | 3521 | file = alloc_empty_file(op->open_flag, current_cred()); |
1afc99be AV |
3522 | if (IS_ERR(file)) |
3523 | return file; | |
31e6b01f | 3524 | |
bb458c64 | 3525 | if (unlikely(file->f_flags & __O_TMPFILE)) { |
3ec2eef1 | 3526 | error = do_tmpfile(nd, flags, op, file); |
5f336e72 | 3527 | } else if (unlikely(file->f_flags & O_PATH)) { |
6ac08709 | 3528 | error = do_o_path(nd, flags, file); |
5f336e72 AV |
3529 | } else { |
3530 | const char *s = path_init(nd, flags); | |
3531 | while (!(error = link_path_walk(s, nd)) && | |
3532 | (error = do_last(nd, file, op)) > 0) { | |
3533 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); | |
3534 | s = trailing_symlink(nd); | |
3535 | } | |
3536 | terminate_walk(nd); | |
806b681c | 3537 | } |
7c1c01ec | 3538 | if (likely(!error)) { |
aad888f8 | 3539 | if (likely(file->f_mode & FMODE_OPENED)) |
7c1c01ec AV |
3540 | return file; |
3541 | WARN_ON(1); | |
3542 | error = -EINVAL; | |
16b1c1cd | 3543 | } |
7c1c01ec AV |
3544 | fput(file); |
3545 | if (error == -EOPENSTALE) { | |
3546 | if (flags & LOOKUP_RCU) | |
3547 | error = -ECHILD; | |
3548 | else | |
3549 | error = -ESTALE; | |
2675a4eb | 3550 | } |
7c1c01ec | 3551 | return ERR_PTR(error); |
1da177e4 LT |
3552 | } |
3553 | ||
669abf4e | 3554 | struct file *do_filp_open(int dfd, struct filename *pathname, |
f9652e10 | 3555 | const struct open_flags *op) |
13aab428 | 3556 | { |
9883d185 | 3557 | struct nameidata nd; |
f9652e10 | 3558 | int flags = op->lookup_flags; |
13aab428 AV |
3559 | struct file *filp; |
3560 | ||
9883d185 | 3561 | set_nameidata(&nd, dfd, pathname); |
c8a53ee5 | 3562 | filp = path_openat(&nd, op, flags | LOOKUP_RCU); |
13aab428 | 3563 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3564 | filp = path_openat(&nd, op, flags); |
13aab428 | 3565 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3566 | filp = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3567 | restore_nameidata(); |
13aab428 AV |
3568 | return filp; |
3569 | } | |
3570 | ||
73d049a4 | 3571 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
f9652e10 | 3572 | const char *name, const struct open_flags *op) |
73d049a4 | 3573 | { |
9883d185 | 3574 | struct nameidata nd; |
73d049a4 | 3575 | struct file *file; |
51689104 | 3576 | struct filename *filename; |
f9652e10 | 3577 | int flags = op->lookup_flags | LOOKUP_ROOT; |
73d049a4 AV |
3578 | |
3579 | nd.root.mnt = mnt; | |
3580 | nd.root.dentry = dentry; | |
3581 | ||
b18825a7 | 3582 | if (d_is_symlink(dentry) && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
3583 | return ERR_PTR(-ELOOP); |
3584 | ||
51689104 | 3585 | filename = getname_kernel(name); |
a1c83681 | 3586 | if (IS_ERR(filename)) |
51689104 PM |
3587 | return ERR_CAST(filename); |
3588 | ||
9883d185 | 3589 | set_nameidata(&nd, -1, filename); |
c8a53ee5 | 3590 | file = path_openat(&nd, op, flags | LOOKUP_RCU); |
73d049a4 | 3591 | if (unlikely(file == ERR_PTR(-ECHILD))) |
c8a53ee5 | 3592 | file = path_openat(&nd, op, flags); |
73d049a4 | 3593 | if (unlikely(file == ERR_PTR(-ESTALE))) |
c8a53ee5 | 3594 | file = path_openat(&nd, op, flags | LOOKUP_REVAL); |
9883d185 | 3595 | restore_nameidata(); |
51689104 | 3596 | putname(filename); |
73d049a4 AV |
3597 | return file; |
3598 | } | |
3599 | ||
fa14a0b8 | 3600 | static struct dentry *filename_create(int dfd, struct filename *name, |
1ac12b4b | 3601 | struct path *path, unsigned int lookup_flags) |
1da177e4 | 3602 | { |
c663e5d8 | 3603 | struct dentry *dentry = ERR_PTR(-EEXIST); |
391172c4 AV |
3604 | struct qstr last; |
3605 | int type; | |
c30dabfe | 3606 | int err2; |
1ac12b4b JL |
3607 | int error; |
3608 | bool is_dir = (lookup_flags & LOOKUP_DIRECTORY); | |
3609 | ||
3610 | /* | |
3611 | * Note that only LOOKUP_REVAL and LOOKUP_DIRECTORY matter here. Any | |
3612 | * other flags passed in are ignored! | |
3613 | */ | |
3614 | lookup_flags &= LOOKUP_REVAL; | |
3615 | ||
5c31b6ce AV |
3616 | name = filename_parentat(dfd, name, lookup_flags, path, &last, &type); |
3617 | if (IS_ERR(name)) | |
3618 | return ERR_CAST(name); | |
1da177e4 | 3619 | |
c663e5d8 CH |
3620 | /* |
3621 | * Yucky last component or no last component at all? | |
3622 | * (foo/., foo/.., /////) | |
3623 | */ | |
5c31b6ce | 3624 | if (unlikely(type != LAST_NORM)) |
ed75e95d | 3625 | goto out; |
c663e5d8 | 3626 | |
c30dabfe | 3627 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
391172c4 | 3628 | err2 = mnt_want_write(path->mnt); |
c663e5d8 CH |
3629 | /* |
3630 | * Do the final lookup. | |
3631 | */ | |
391172c4 | 3632 | lookup_flags |= LOOKUP_CREATE | LOOKUP_EXCL; |
5955102c | 3633 | inode_lock_nested(path->dentry->d_inode, I_MUTEX_PARENT); |
391172c4 | 3634 | dentry = __lookup_hash(&last, path->dentry, lookup_flags); |
1da177e4 | 3635 | if (IS_ERR(dentry)) |
a8104a9f | 3636 | goto unlock; |
c663e5d8 | 3637 | |
a8104a9f | 3638 | error = -EEXIST; |
b18825a7 | 3639 | if (d_is_positive(dentry)) |
a8104a9f | 3640 | goto fail; |
b18825a7 | 3641 | |
c663e5d8 CH |
3642 | /* |
3643 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3644 | * From the vfs_mknod() POV we just have a negative dentry - | |
3645 | * all is fine. Let's be bastards - you had / on the end, you've | |
3646 | * been asking for (non-existent) directory. -ENOENT for you. | |
3647 | */ | |
391172c4 | 3648 | if (unlikely(!is_dir && last.name[last.len])) { |
a8104a9f | 3649 | error = -ENOENT; |
ed75e95d | 3650 | goto fail; |
e9baf6e5 | 3651 | } |
c30dabfe JK |
3652 | if (unlikely(err2)) { |
3653 | error = err2; | |
a8104a9f | 3654 | goto fail; |
c30dabfe | 3655 | } |
181c37b6 | 3656 | putname(name); |
1da177e4 | 3657 | return dentry; |
1da177e4 | 3658 | fail: |
a8104a9f AV |
3659 | dput(dentry); |
3660 | dentry = ERR_PTR(error); | |
3661 | unlock: | |
5955102c | 3662 | inode_unlock(path->dentry->d_inode); |
c30dabfe | 3663 | if (!err2) |
391172c4 | 3664 | mnt_drop_write(path->mnt); |
ed75e95d | 3665 | out: |
391172c4 | 3666 | path_put(path); |
181c37b6 | 3667 | putname(name); |
1da177e4 LT |
3668 | return dentry; |
3669 | } | |
fa14a0b8 AV |
3670 | |
3671 | struct dentry *kern_path_create(int dfd, const char *pathname, | |
3672 | struct path *path, unsigned int lookup_flags) | |
3673 | { | |
181c37b6 AV |
3674 | return filename_create(dfd, getname_kernel(pathname), |
3675 | path, lookup_flags); | |
fa14a0b8 | 3676 | } |
dae6ad8f AV |
3677 | EXPORT_SYMBOL(kern_path_create); |
3678 | ||
921a1650 AV |
3679 | void done_path_create(struct path *path, struct dentry *dentry) |
3680 | { | |
3681 | dput(dentry); | |
5955102c | 3682 | inode_unlock(path->dentry->d_inode); |
a8104a9f | 3683 | mnt_drop_write(path->mnt); |
921a1650 AV |
3684 | path_put(path); |
3685 | } | |
3686 | EXPORT_SYMBOL(done_path_create); | |
3687 | ||
520ae687 | 3688 | inline struct dentry *user_path_create(int dfd, const char __user *pathname, |
1ac12b4b | 3689 | struct path *path, unsigned int lookup_flags) |
dae6ad8f | 3690 | { |
181c37b6 | 3691 | return filename_create(dfd, getname(pathname), path, lookup_flags); |
dae6ad8f AV |
3692 | } |
3693 | EXPORT_SYMBOL(user_path_create); | |
3694 | ||
1a67aafb | 3695 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3696 | { |
a95164d9 | 3697 | int error = may_create(dir, dentry); |
1da177e4 LT |
3698 | |
3699 | if (error) | |
3700 | return error; | |
3701 | ||
94f82008 | 3702 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3703 | return -EPERM; |
3704 | ||
acfa4380 | 3705 | if (!dir->i_op->mknod) |
1da177e4 LT |
3706 | return -EPERM; |
3707 | ||
08ce5f16 SH |
3708 | error = devcgroup_inode_mknod(mode, dev); |
3709 | if (error) | |
3710 | return error; | |
3711 | ||
1da177e4 LT |
3712 | error = security_inode_mknod(dir, dentry, mode, dev); |
3713 | if (error) | |
3714 | return error; | |
3715 | ||
1da177e4 | 3716 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3717 | if (!error) |
f38aa942 | 3718 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3719 | return error; |
3720 | } | |
4d359507 | 3721 | EXPORT_SYMBOL(vfs_mknod); |
1da177e4 | 3722 | |
f69aac00 | 3723 | static int may_mknod(umode_t mode) |
463c3197 DH |
3724 | { |
3725 | switch (mode & S_IFMT) { | |
3726 | case S_IFREG: | |
3727 | case S_IFCHR: | |
3728 | case S_IFBLK: | |
3729 | case S_IFIFO: | |
3730 | case S_IFSOCK: | |
3731 | case 0: /* zero mode translates to S_IFREG */ | |
3732 | return 0; | |
3733 | case S_IFDIR: | |
3734 | return -EPERM; | |
3735 | default: | |
3736 | return -EINVAL; | |
3737 | } | |
3738 | } | |
3739 | ||
87c4e192 DB |
3740 | long do_mknodat(int dfd, const char __user *filename, umode_t mode, |
3741 | unsigned int dev) | |
1da177e4 | 3742 | { |
2ad94ae6 | 3743 | struct dentry *dentry; |
dae6ad8f AV |
3744 | struct path path; |
3745 | int error; | |
972567f1 | 3746 | unsigned int lookup_flags = 0; |
1da177e4 | 3747 | |
8e4bfca1 AV |
3748 | error = may_mknod(mode); |
3749 | if (error) | |
3750 | return error; | |
972567f1 JL |
3751 | retry: |
3752 | dentry = user_path_create(dfd, filename, &path, lookup_flags); | |
dae6ad8f AV |
3753 | if (IS_ERR(dentry)) |
3754 | return PTR_ERR(dentry); | |
2ad94ae6 | 3755 | |
dae6ad8f | 3756 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3757 | mode &= ~current_umask(); |
dae6ad8f | 3758 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3759 | if (error) |
a8104a9f | 3760 | goto out; |
463c3197 | 3761 | switch (mode & S_IFMT) { |
1da177e4 | 3762 | case 0: case S_IFREG: |
312b63fb | 3763 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
05d1a717 MZ |
3764 | if (!error) |
3765 | ima_post_path_mknod(dentry); | |
1da177e4 LT |
3766 | break; |
3767 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3768 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3769 | new_decode_dev(dev)); |
3770 | break; | |
3771 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3772 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3773 | break; |
1da177e4 | 3774 | } |
a8104a9f | 3775 | out: |
921a1650 | 3776 | done_path_create(&path, dentry); |
972567f1 JL |
3777 | if (retry_estale(error, lookup_flags)) { |
3778 | lookup_flags |= LOOKUP_REVAL; | |
3779 | goto retry; | |
3780 | } | |
1da177e4 LT |
3781 | return error; |
3782 | } | |
3783 | ||
87c4e192 DB |
3784 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
3785 | unsigned int, dev) | |
3786 | { | |
3787 | return do_mknodat(dfd, filename, mode, dev); | |
3788 | } | |
3789 | ||
8208a22b | 3790 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d | 3791 | { |
87c4e192 | 3792 | return do_mknodat(AT_FDCWD, filename, mode, dev); |
5590ff0d UD |
3793 | } |
3794 | ||
18bb1db3 | 3795 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3796 | { |
a95164d9 | 3797 | int error = may_create(dir, dentry); |
8de52778 | 3798 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3799 | |
3800 | if (error) | |
3801 | return error; | |
3802 | ||
acfa4380 | 3803 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3804 | return -EPERM; |
3805 | ||
3806 | mode &= (S_IRWXUGO|S_ISVTX); | |
3807 | error = security_inode_mkdir(dir, dentry, mode); | |
3808 | if (error) | |
3809 | return error; | |
3810 | ||
8de52778 AV |
3811 | if (max_links && dir->i_nlink >= max_links) |
3812 | return -EMLINK; | |
3813 | ||
1da177e4 | 3814 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3815 | if (!error) |
f38aa942 | 3816 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3817 | return error; |
3818 | } | |
4d359507 | 3819 | EXPORT_SYMBOL(vfs_mkdir); |
1da177e4 | 3820 | |
0101db7a | 3821 | long do_mkdirat(int dfd, const char __user *pathname, umode_t mode) |
1da177e4 | 3822 | { |
6902d925 | 3823 | struct dentry *dentry; |
dae6ad8f AV |
3824 | struct path path; |
3825 | int error; | |
b76d8b82 | 3826 | unsigned int lookup_flags = LOOKUP_DIRECTORY; |
1da177e4 | 3827 | |
b76d8b82 JL |
3828 | retry: |
3829 | dentry = user_path_create(dfd, pathname, &path, lookup_flags); | |
6902d925 | 3830 | if (IS_ERR(dentry)) |
dae6ad8f | 3831 | return PTR_ERR(dentry); |
1da177e4 | 3832 | |
dae6ad8f | 3833 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3834 | mode &= ~current_umask(); |
dae6ad8f | 3835 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3836 | if (!error) |
3837 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3838 | done_path_create(&path, dentry); |
b76d8b82 JL |
3839 | if (retry_estale(error, lookup_flags)) { |
3840 | lookup_flags |= LOOKUP_REVAL; | |
3841 | goto retry; | |
3842 | } | |
1da177e4 LT |
3843 | return error; |
3844 | } | |
3845 | ||
0101db7a DB |
3846 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
3847 | { | |
3848 | return do_mkdirat(dfd, pathname, mode); | |
3849 | } | |
3850 | ||
a218d0fd | 3851 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d | 3852 | { |
0101db7a | 3853 | return do_mkdirat(AT_FDCWD, pathname, mode); |
5590ff0d UD |
3854 | } |
3855 | ||
1da177e4 LT |
3856 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) |
3857 | { | |
3858 | int error = may_delete(dir, dentry, 1); | |
3859 | ||
3860 | if (error) | |
3861 | return error; | |
3862 | ||
acfa4380 | 3863 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3864 | return -EPERM; |
3865 | ||
1d2ef590 | 3866 | dget(dentry); |
5955102c | 3867 | inode_lock(dentry->d_inode); |
912dbc15 SW |
3868 | |
3869 | error = -EBUSY; | |
7af1364f | 3870 | if (is_local_mountpoint(dentry)) |
912dbc15 SW |
3871 | goto out; |
3872 | ||
3873 | error = security_inode_rmdir(dir, dentry); | |
3874 | if (error) | |
3875 | goto out; | |
3876 | ||
3877 | error = dir->i_op->rmdir(dir, dentry); | |
3878 | if (error) | |
3879 | goto out; | |
3880 | ||
8767712f | 3881 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3882 | dentry->d_inode->i_flags |= S_DEAD; |
3883 | dont_mount(dentry); | |
8ed936b5 | 3884 | detach_mounts(dentry); |
912dbc15 SW |
3885 | |
3886 | out: | |
5955102c | 3887 | inode_unlock(dentry->d_inode); |
1d2ef590 | 3888 | dput(dentry); |
912dbc15 | 3889 | if (!error) |
1da177e4 | 3890 | d_delete(dentry); |
1da177e4 LT |
3891 | return error; |
3892 | } | |
4d359507 | 3893 | EXPORT_SYMBOL(vfs_rmdir); |
1da177e4 | 3894 | |
f459dffa | 3895 | long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3896 | { |
3897 | int error = 0; | |
91a27b2a | 3898 | struct filename *name; |
1da177e4 | 3899 | struct dentry *dentry; |
f5beed75 AV |
3900 | struct path path; |
3901 | struct qstr last; | |
3902 | int type; | |
c6ee9206 JL |
3903 | unsigned int lookup_flags = 0; |
3904 | retry: | |
c1d4dd27 AV |
3905 | name = filename_parentat(dfd, getname(pathname), lookup_flags, |
3906 | &path, &last, &type); | |
91a27b2a JL |
3907 | if (IS_ERR(name)) |
3908 | return PTR_ERR(name); | |
1da177e4 | 3909 | |
f5beed75 | 3910 | switch (type) { |
0612d9fb OH |
3911 | case LAST_DOTDOT: |
3912 | error = -ENOTEMPTY; | |
3913 | goto exit1; | |
3914 | case LAST_DOT: | |
3915 | error = -EINVAL; | |
3916 | goto exit1; | |
3917 | case LAST_ROOT: | |
3918 | error = -EBUSY; | |
3919 | goto exit1; | |
1da177e4 | 3920 | } |
0612d9fb | 3921 | |
f5beed75 | 3922 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3923 | if (error) |
3924 | goto exit1; | |
0612d9fb | 3925 | |
5955102c | 3926 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 3927 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 | 3928 | error = PTR_ERR(dentry); |
6902d925 DH |
3929 | if (IS_ERR(dentry)) |
3930 | goto exit2; | |
e6bc45d6 TT |
3931 | if (!dentry->d_inode) { |
3932 | error = -ENOENT; | |
3933 | goto exit3; | |
3934 | } | |
f5beed75 | 3935 | error = security_path_rmdir(&path, dentry); |
be6d3e56 | 3936 | if (error) |
c30dabfe | 3937 | goto exit3; |
f5beed75 | 3938 | error = vfs_rmdir(path.dentry->d_inode, dentry); |
0622753b | 3939 | exit3: |
6902d925 DH |
3940 | dput(dentry); |
3941 | exit2: | |
5955102c | 3942 | inode_unlock(path.dentry->d_inode); |
f5beed75 | 3943 | mnt_drop_write(path.mnt); |
1da177e4 | 3944 | exit1: |
f5beed75 | 3945 | path_put(&path); |
1da177e4 | 3946 | putname(name); |
c6ee9206 JL |
3947 | if (retry_estale(error, lookup_flags)) { |
3948 | lookup_flags |= LOOKUP_REVAL; | |
3949 | goto retry; | |
3950 | } | |
1da177e4 LT |
3951 | return error; |
3952 | } | |
3953 | ||
3cdad428 | 3954 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3955 | { |
3956 | return do_rmdir(AT_FDCWD, pathname); | |
3957 | } | |
3958 | ||
b21996e3 BF |
3959 | /** |
3960 | * vfs_unlink - unlink a filesystem object | |
3961 | * @dir: parent directory | |
3962 | * @dentry: victim | |
3963 | * @delegated_inode: returns victim inode, if the inode is delegated. | |
3964 | * | |
3965 | * The caller must hold dir->i_mutex. | |
3966 | * | |
3967 | * If vfs_unlink discovers a delegation, it will return -EWOULDBLOCK and | |
3968 | * return a reference to the inode in delegated_inode. The caller | |
3969 | * should then break the delegation on that inode and retry. Because | |
3970 | * breaking a delegation may take a long time, the caller should drop | |
3971 | * dir->i_mutex before doing so. | |
3972 | * | |
3973 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3974 | * be appropriate for callers that expect the underlying filesystem not | |
3975 | * to be NFS exported. | |
3976 | */ | |
3977 | int vfs_unlink(struct inode *dir, struct dentry *dentry, struct inode **delegated_inode) | |
1da177e4 | 3978 | { |
9accbb97 | 3979 | struct inode *target = dentry->d_inode; |
1da177e4 LT |
3980 | int error = may_delete(dir, dentry, 0); |
3981 | ||
3982 | if (error) | |
3983 | return error; | |
3984 | ||
acfa4380 | 3985 | if (!dir->i_op->unlink) |
1da177e4 LT |
3986 | return -EPERM; |
3987 | ||
5955102c | 3988 | inode_lock(target); |
8ed936b5 | 3989 | if (is_local_mountpoint(dentry)) |
1da177e4 LT |
3990 | error = -EBUSY; |
3991 | else { | |
3992 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3993 | if (!error) { |
5a14696c BF |
3994 | error = try_break_deleg(target, delegated_inode); |
3995 | if (error) | |
b21996e3 | 3996 | goto out; |
1da177e4 | 3997 | error = dir->i_op->unlink(dir, dentry); |
8ed936b5 | 3998 | if (!error) { |
d83c49f3 | 3999 | dont_mount(dentry); |
8ed936b5 EB |
4000 | detach_mounts(dentry); |
4001 | } | |
bec1052e | 4002 | } |
1da177e4 | 4003 | } |
b21996e3 | 4004 | out: |
5955102c | 4005 | inode_unlock(target); |
1da177e4 LT |
4006 | |
4007 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
4008 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
9accbb97 | 4009 | fsnotify_link_count(target); |
e234f35c | 4010 | d_delete(dentry); |
1da177e4 | 4011 | } |
0eeca283 | 4012 | |
1da177e4 LT |
4013 | return error; |
4014 | } | |
4d359507 | 4015 | EXPORT_SYMBOL(vfs_unlink); |
1da177e4 LT |
4016 | |
4017 | /* | |
4018 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 4019 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
4020 | * writeout happening, and we don't want to prevent access to the directory |
4021 | * while waiting on the I/O. | |
4022 | */ | |
da2f1362 | 4023 | long do_unlinkat(int dfd, struct filename *name) |
1da177e4 | 4024 | { |
2ad94ae6 | 4025 | int error; |
1da177e4 | 4026 | struct dentry *dentry; |
f5beed75 AV |
4027 | struct path path; |
4028 | struct qstr last; | |
4029 | int type; | |
1da177e4 | 4030 | struct inode *inode = NULL; |
b21996e3 | 4031 | struct inode *delegated_inode = NULL; |
5d18f813 JL |
4032 | unsigned int lookup_flags = 0; |
4033 | retry: | |
da2f1362 | 4034 | name = filename_parentat(dfd, name, lookup_flags, &path, &last, &type); |
91a27b2a JL |
4035 | if (IS_ERR(name)) |
4036 | return PTR_ERR(name); | |
2ad94ae6 | 4037 | |
1da177e4 | 4038 | error = -EISDIR; |
f5beed75 | 4039 | if (type != LAST_NORM) |
1da177e4 | 4040 | goto exit1; |
0612d9fb | 4041 | |
f5beed75 | 4042 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
4043 | if (error) |
4044 | goto exit1; | |
b21996e3 | 4045 | retry_deleg: |
5955102c | 4046 | inode_lock_nested(path.dentry->d_inode, I_MUTEX_PARENT); |
f5beed75 | 4047 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); |
1da177e4 LT |
4048 | error = PTR_ERR(dentry); |
4049 | if (!IS_ERR(dentry)) { | |
4050 | /* Why not before? Because we want correct error value */ | |
f5beed75 | 4051 | if (last.name[last.len]) |
50338b88 | 4052 | goto slashes; |
1da177e4 | 4053 | inode = dentry->d_inode; |
b18825a7 | 4054 | if (d_is_negative(dentry)) |
e6bc45d6 TT |
4055 | goto slashes; |
4056 | ihold(inode); | |
f5beed75 | 4057 | error = security_path_unlink(&path, dentry); |
be6d3e56 | 4058 | if (error) |
c30dabfe | 4059 | goto exit2; |
f5beed75 | 4060 | error = vfs_unlink(path.dentry->d_inode, dentry, &delegated_inode); |
c30dabfe | 4061 | exit2: |
1da177e4 LT |
4062 | dput(dentry); |
4063 | } | |
5955102c | 4064 | inode_unlock(path.dentry->d_inode); |
1da177e4 LT |
4065 | if (inode) |
4066 | iput(inode); /* truncate the inode here */ | |
b21996e3 BF |
4067 | inode = NULL; |
4068 | if (delegated_inode) { | |
5a14696c | 4069 | error = break_deleg_wait(&delegated_inode); |
b21996e3 BF |
4070 | if (!error) |
4071 | goto retry_deleg; | |
4072 | } | |
f5beed75 | 4073 | mnt_drop_write(path.mnt); |
1da177e4 | 4074 | exit1: |
f5beed75 | 4075 | path_put(&path); |
5d18f813 JL |
4076 | if (retry_estale(error, lookup_flags)) { |
4077 | lookup_flags |= LOOKUP_REVAL; | |
4078 | inode = NULL; | |
4079 | goto retry; | |
4080 | } | |
da2f1362 | 4081 | putname(name); |
1da177e4 LT |
4082 | return error; |
4083 | ||
4084 | slashes: | |
b18825a7 DH |
4085 | if (d_is_negative(dentry)) |
4086 | error = -ENOENT; | |
44b1d530 | 4087 | else if (d_is_dir(dentry)) |
b18825a7 DH |
4088 | error = -EISDIR; |
4089 | else | |
4090 | error = -ENOTDIR; | |
1da177e4 LT |
4091 | goto exit2; |
4092 | } | |
4093 | ||
2e4d0924 | 4094 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
4095 | { |
4096 | if ((flag & ~AT_REMOVEDIR) != 0) | |
4097 | return -EINVAL; | |
4098 | ||
4099 | if (flag & AT_REMOVEDIR) | |
4100 | return do_rmdir(dfd, pathname); | |
4101 | ||
da2f1362 | 4102 | return do_unlinkat(dfd, getname(pathname)); |
5590ff0d UD |
4103 | } |
4104 | ||
3480b257 | 4105 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d | 4106 | { |
da2f1362 | 4107 | return do_unlinkat(AT_FDCWD, getname(pathname)); |
5590ff0d UD |
4108 | } |
4109 | ||
db2e747b | 4110 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 4111 | { |
a95164d9 | 4112 | int error = may_create(dir, dentry); |
1da177e4 LT |
4113 | |
4114 | if (error) | |
4115 | return error; | |
4116 | ||
acfa4380 | 4117 | if (!dir->i_op->symlink) |
1da177e4 LT |
4118 | return -EPERM; |
4119 | ||
4120 | error = security_inode_symlink(dir, dentry, oldname); | |
4121 | if (error) | |
4122 | return error; | |
4123 | ||
1da177e4 | 4124 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 4125 | if (!error) |
f38aa942 | 4126 | fsnotify_create(dir, dentry); |
1da177e4 LT |
4127 | return error; |
4128 | } | |
4d359507 | 4129 | EXPORT_SYMBOL(vfs_symlink); |
1da177e4 | 4130 | |
b724e846 DB |
4131 | long do_symlinkat(const char __user *oldname, int newdfd, |
4132 | const char __user *newname) | |
1da177e4 | 4133 | { |
2ad94ae6 | 4134 | int error; |
91a27b2a | 4135 | struct filename *from; |
6902d925 | 4136 | struct dentry *dentry; |
dae6ad8f | 4137 | struct path path; |
f46d3567 | 4138 | unsigned int lookup_flags = 0; |
1da177e4 LT |
4139 | |
4140 | from = getname(oldname); | |
2ad94ae6 | 4141 | if (IS_ERR(from)) |
1da177e4 | 4142 | return PTR_ERR(from); |
f46d3567 JL |
4143 | retry: |
4144 | dentry = user_path_create(newdfd, newname, &path, lookup_flags); | |
6902d925 DH |
4145 | error = PTR_ERR(dentry); |
4146 | if (IS_ERR(dentry)) | |
dae6ad8f | 4147 | goto out_putname; |
6902d925 | 4148 | |
91a27b2a | 4149 | error = security_path_symlink(&path, dentry, from->name); |
a8104a9f | 4150 | if (!error) |
91a27b2a | 4151 | error = vfs_symlink(path.dentry->d_inode, dentry, from->name); |
921a1650 | 4152 | done_path_create(&path, dentry); |
f46d3567 JL |
4153 | if (retry_estale(error, lookup_flags)) { |
4154 | lookup_flags |= LOOKUP_REVAL; | |
4155 | goto retry; | |
4156 | } | |
6902d925 | 4157 | out_putname: |
1da177e4 LT |
4158 | putname(from); |
4159 | return error; | |
4160 | } | |
4161 | ||
b724e846 DB |
4162 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
4163 | int, newdfd, const char __user *, newname) | |
4164 | { | |
4165 | return do_symlinkat(oldname, newdfd, newname); | |
4166 | } | |
4167 | ||
3480b257 | 4168 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4169 | { |
b724e846 | 4170 | return do_symlinkat(oldname, AT_FDCWD, newname); |
5590ff0d UD |
4171 | } |
4172 | ||
146a8595 BF |
4173 | /** |
4174 | * vfs_link - create a new link | |
4175 | * @old_dentry: object to be linked | |
4176 | * @dir: new parent | |
4177 | * @new_dentry: where to create the new link | |
4178 | * @delegated_inode: returns inode needing a delegation break | |
4179 | * | |
4180 | * The caller must hold dir->i_mutex | |
4181 | * | |
4182 | * If vfs_link discovers a delegation on the to-be-linked file in need | |
4183 | * of breaking, it will return -EWOULDBLOCK and return a reference to the | |
4184 | * inode in delegated_inode. The caller should then break the delegation | |
4185 | * and retry. Because breaking a delegation may take a long time, the | |
4186 | * caller should drop the i_mutex before doing so. | |
4187 | * | |
4188 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4189 | * be appropriate for callers that expect the underlying filesystem not | |
4190 | * to be NFS exported. | |
4191 | */ | |
4192 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, struct inode **delegated_inode) | |
1da177e4 LT |
4193 | { |
4194 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 4195 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
4196 | int error; |
4197 | ||
4198 | if (!inode) | |
4199 | return -ENOENT; | |
4200 | ||
a95164d9 | 4201 | error = may_create(dir, new_dentry); |
1da177e4 LT |
4202 | if (error) |
4203 | return error; | |
4204 | ||
4205 | if (dir->i_sb != inode->i_sb) | |
4206 | return -EXDEV; | |
4207 | ||
4208 | /* | |
4209 | * A link to an append-only or immutable file cannot be created. | |
4210 | */ | |
4211 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
4212 | return -EPERM; | |
0bd23d09 EB |
4213 | /* |
4214 | * Updating the link count will likely cause i_uid and i_gid to | |
4215 | * be writen back improperly if their true value is unknown to | |
4216 | * the vfs. | |
4217 | */ | |
4218 | if (HAS_UNMAPPED_ID(inode)) | |
4219 | return -EPERM; | |
acfa4380 | 4220 | if (!dir->i_op->link) |
1da177e4 | 4221 | return -EPERM; |
7e79eedb | 4222 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
4223 | return -EPERM; |
4224 | ||
4225 | error = security_inode_link(old_dentry, dir, new_dentry); | |
4226 | if (error) | |
4227 | return error; | |
4228 | ||
5955102c | 4229 | inode_lock(inode); |
aae8a97d | 4230 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
f4e0c30c | 4231 | if (inode->i_nlink == 0 && !(inode->i_state & I_LINKABLE)) |
aae8a97d | 4232 | error = -ENOENT; |
8de52778 AV |
4233 | else if (max_links && inode->i_nlink >= max_links) |
4234 | error = -EMLINK; | |
146a8595 BF |
4235 | else { |
4236 | error = try_break_deleg(inode, delegated_inode); | |
4237 | if (!error) | |
4238 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
4239 | } | |
f4e0c30c AV |
4240 | |
4241 | if (!error && (inode->i_state & I_LINKABLE)) { | |
4242 | spin_lock(&inode->i_lock); | |
4243 | inode->i_state &= ~I_LINKABLE; | |
4244 | spin_unlock(&inode->i_lock); | |
4245 | } | |
5955102c | 4246 | inode_unlock(inode); |
e31e14ec | 4247 | if (!error) |
7e79eedb | 4248 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
4249 | return error; |
4250 | } | |
4d359507 | 4251 | EXPORT_SYMBOL(vfs_link); |
1da177e4 LT |
4252 | |
4253 | /* | |
4254 | * Hardlinks are often used in delicate situations. We avoid | |
4255 | * security-related surprises by not following symlinks on the | |
4256 | * newname. --KAB | |
4257 | * | |
4258 | * We don't follow them on the oldname either to be compatible | |
4259 | * with linux 2.0, and to avoid hard-linking to directories | |
4260 | * and other special files. --ADM | |
4261 | */ | |
46ea89eb DB |
4262 | int do_linkat(int olddfd, const char __user *oldname, int newdfd, |
4263 | const char __user *newname, int flags) | |
1da177e4 LT |
4264 | { |
4265 | struct dentry *new_dentry; | |
dae6ad8f | 4266 | struct path old_path, new_path; |
146a8595 | 4267 | struct inode *delegated_inode = NULL; |
11a7b371 | 4268 | int how = 0; |
1da177e4 | 4269 | int error; |
1da177e4 | 4270 | |
11a7b371 | 4271 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 4272 | return -EINVAL; |
11a7b371 | 4273 | /* |
f0cc6ffb LT |
4274 | * To use null names we require CAP_DAC_READ_SEARCH |
4275 | * This ensures that not everyone will be able to create | |
4276 | * handlink using the passed filedescriptor. | |
11a7b371 | 4277 | */ |
f0cc6ffb LT |
4278 | if (flags & AT_EMPTY_PATH) { |
4279 | if (!capable(CAP_DAC_READ_SEARCH)) | |
4280 | return -ENOENT; | |
11a7b371 | 4281 | how = LOOKUP_EMPTY; |
f0cc6ffb | 4282 | } |
11a7b371 AK |
4283 | |
4284 | if (flags & AT_SYMLINK_FOLLOW) | |
4285 | how |= LOOKUP_FOLLOW; | |
442e31ca | 4286 | retry: |
11a7b371 | 4287 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 4288 | if (error) |
2ad94ae6 AV |
4289 | return error; |
4290 | ||
442e31ca JL |
4291 | new_dentry = user_path_create(newdfd, newname, &new_path, |
4292 | (how & LOOKUP_REVAL)); | |
1da177e4 | 4293 | error = PTR_ERR(new_dentry); |
6902d925 | 4294 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
4295 | goto out; |
4296 | ||
4297 | error = -EXDEV; | |
4298 | if (old_path.mnt != new_path.mnt) | |
4299 | goto out_dput; | |
800179c9 KC |
4300 | error = may_linkat(&old_path); |
4301 | if (unlikely(error)) | |
4302 | goto out_dput; | |
dae6ad8f | 4303 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 4304 | if (error) |
a8104a9f | 4305 | goto out_dput; |
146a8595 | 4306 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry, &delegated_inode); |
75c3f29d | 4307 | out_dput: |
921a1650 | 4308 | done_path_create(&new_path, new_dentry); |
146a8595 BF |
4309 | if (delegated_inode) { |
4310 | error = break_deleg_wait(&delegated_inode); | |
d22e6338 OD |
4311 | if (!error) { |
4312 | path_put(&old_path); | |
146a8595 | 4313 | goto retry; |
d22e6338 | 4314 | } |
146a8595 | 4315 | } |
442e31ca | 4316 | if (retry_estale(error, how)) { |
d22e6338 | 4317 | path_put(&old_path); |
442e31ca JL |
4318 | how |= LOOKUP_REVAL; |
4319 | goto retry; | |
4320 | } | |
1da177e4 | 4321 | out: |
2d8f3038 | 4322 | path_put(&old_path); |
1da177e4 LT |
4323 | |
4324 | return error; | |
4325 | } | |
4326 | ||
46ea89eb DB |
4327 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
4328 | int, newdfd, const char __user *, newname, int, flags) | |
4329 | { | |
4330 | return do_linkat(olddfd, oldname, newdfd, newname, flags); | |
4331 | } | |
4332 | ||
3480b257 | 4333 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4334 | { |
46ea89eb | 4335 | return do_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4336 | } |
4337 | ||
bc27027a MS |
4338 | /** |
4339 | * vfs_rename - rename a filesystem object | |
4340 | * @old_dir: parent of source | |
4341 | * @old_dentry: source | |
4342 | * @new_dir: parent of destination | |
4343 | * @new_dentry: destination | |
4344 | * @delegated_inode: returns an inode needing a delegation break | |
520c8b16 | 4345 | * @flags: rename flags |
bc27027a MS |
4346 | * |
4347 | * The caller must hold multiple mutexes--see lock_rename()). | |
4348 | * | |
4349 | * If vfs_rename discovers a delegation in need of breaking at either | |
4350 | * the source or destination, it will return -EWOULDBLOCK and return a | |
4351 | * reference to the inode in delegated_inode. The caller should then | |
4352 | * break the delegation and retry. Because breaking a delegation may | |
4353 | * take a long time, the caller should drop all locks before doing | |
4354 | * so. | |
4355 | * | |
4356 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4357 | * be appropriate for callers that expect the underlying filesystem not | |
4358 | * to be NFS exported. | |
4359 | * | |
1da177e4 LT |
4360 | * The worst of all namespace operations - renaming directory. "Perverted" |
4361 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
4362 | * Problems: | |
0117d427 | 4363 | * |
d03b29a2 | 4364 | * a) we can get into loop creation. |
1da177e4 LT |
4365 | * b) race potential - two innocent renames can create a loop together. |
4366 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 4367 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 | 4368 | * story. |
6cedba89 BF |
4369 | * c) we have to lock _four_ objects - parents and victim (if it exists), |
4370 | * and source (if it is not a directory). | |
1b1dcc1b | 4371 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
4372 | * whether the target exists). Solution: try to be smart with locking |
4373 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 4374 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
4375 | * move will be locked. Thus we can rank directories by the tree |
4376 | * (ancestors first) and rank all non-directories after them. | |
4377 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 4378 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
4379 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
4380 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
4381 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 4382 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 4383 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 4384 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 4385 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
4386 | * locking]. |
4387 | */ | |
bc27027a MS |
4388 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, |
4389 | struct inode *new_dir, struct dentry *new_dentry, | |
520c8b16 | 4390 | struct inode **delegated_inode, unsigned int flags) |
1da177e4 | 4391 | { |
bc27027a MS |
4392 | int error; |
4393 | bool is_dir = d_is_dir(old_dentry); | |
bc27027a | 4394 | struct inode *source = old_dentry->d_inode; |
9055cba7 | 4395 | struct inode *target = new_dentry->d_inode; |
da1ce067 MS |
4396 | bool new_is_dir = false; |
4397 | unsigned max_links = new_dir->i_sb->s_max_links; | |
49d31c2f | 4398 | struct name_snapshot old_name; |
bc27027a | 4399 | |
8d3e2936 | 4400 | if (source == target) |
bc27027a MS |
4401 | return 0; |
4402 | ||
4403 | error = may_delete(old_dir, old_dentry, is_dir); | |
4404 | if (error) | |
4405 | return error; | |
4406 | ||
da1ce067 | 4407 | if (!target) { |
bc27027a | 4408 | error = may_create(new_dir, new_dentry); |
da1ce067 MS |
4409 | } else { |
4410 | new_is_dir = d_is_dir(new_dentry); | |
4411 | ||
4412 | if (!(flags & RENAME_EXCHANGE)) | |
4413 | error = may_delete(new_dir, new_dentry, is_dir); | |
4414 | else | |
4415 | error = may_delete(new_dir, new_dentry, new_is_dir); | |
4416 | } | |
bc27027a MS |
4417 | if (error) |
4418 | return error; | |
4419 | ||
2773bf00 | 4420 | if (!old_dir->i_op->rename) |
bc27027a | 4421 | return -EPERM; |
1da177e4 LT |
4422 | |
4423 | /* | |
4424 | * If we are going to change the parent - check write permissions, | |
4425 | * we'll need to flip '..'. | |
4426 | */ | |
da1ce067 MS |
4427 | if (new_dir != old_dir) { |
4428 | if (is_dir) { | |
4429 | error = inode_permission(source, MAY_WRITE); | |
4430 | if (error) | |
4431 | return error; | |
4432 | } | |
4433 | if ((flags & RENAME_EXCHANGE) && new_is_dir) { | |
4434 | error = inode_permission(target, MAY_WRITE); | |
4435 | if (error) | |
4436 | return error; | |
4437 | } | |
1da177e4 LT |
4438 | } |
4439 | ||
0b3974eb MS |
4440 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry, |
4441 | flags); | |
1da177e4 LT |
4442 | if (error) |
4443 | return error; | |
4444 | ||
49d31c2f | 4445 | take_dentry_name_snapshot(&old_name, old_dentry); |
1d2ef590 | 4446 | dget(new_dentry); |
da1ce067 | 4447 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4448 | lock_two_nondirectories(source, target); |
4449 | else if (target) | |
5955102c | 4450 | inode_lock(target); |
9055cba7 SW |
4451 | |
4452 | error = -EBUSY; | |
7af1364f | 4453 | if (is_local_mountpoint(old_dentry) || is_local_mountpoint(new_dentry)) |
9055cba7 SW |
4454 | goto out; |
4455 | ||
da1ce067 | 4456 | if (max_links && new_dir != old_dir) { |
bc27027a | 4457 | error = -EMLINK; |
da1ce067 | 4458 | if (is_dir && !new_is_dir && new_dir->i_nlink >= max_links) |
bc27027a | 4459 | goto out; |
da1ce067 MS |
4460 | if ((flags & RENAME_EXCHANGE) && !is_dir && new_is_dir && |
4461 | old_dir->i_nlink >= max_links) | |
4462 | goto out; | |
4463 | } | |
da1ce067 | 4464 | if (!is_dir) { |
bc27027a | 4465 | error = try_break_deleg(source, delegated_inode); |
8e6d782c BF |
4466 | if (error) |
4467 | goto out; | |
da1ce067 MS |
4468 | } |
4469 | if (target && !new_is_dir) { | |
4470 | error = try_break_deleg(target, delegated_inode); | |
4471 | if (error) | |
4472 | goto out; | |
8e6d782c | 4473 | } |
2773bf00 | 4474 | error = old_dir->i_op->rename(old_dir, old_dentry, |
18fc84da | 4475 | new_dir, new_dentry, flags); |
51892bbb SW |
4476 | if (error) |
4477 | goto out; | |
4478 | ||
da1ce067 | 4479 | if (!(flags & RENAME_EXCHANGE) && target) { |
8767712f AV |
4480 | if (is_dir) { |
4481 | shrink_dcache_parent(new_dentry); | |
bc27027a | 4482 | target->i_flags |= S_DEAD; |
8767712f | 4483 | } |
51892bbb | 4484 | dont_mount(new_dentry); |
8ed936b5 | 4485 | detach_mounts(new_dentry); |
bc27027a | 4486 | } |
da1ce067 MS |
4487 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) { |
4488 | if (!(flags & RENAME_EXCHANGE)) | |
4489 | d_move(old_dentry, new_dentry); | |
4490 | else | |
4491 | d_exchange(old_dentry, new_dentry); | |
4492 | } | |
51892bbb | 4493 | out: |
da1ce067 | 4494 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4495 | unlock_two_nondirectories(source, target); |
4496 | else if (target) | |
5955102c | 4497 | inode_unlock(target); |
1da177e4 | 4498 | dput(new_dentry); |
da1ce067 | 4499 | if (!error) { |
49d31c2f | 4500 | fsnotify_move(old_dir, new_dir, old_name.name, is_dir, |
da1ce067 MS |
4501 | !(flags & RENAME_EXCHANGE) ? target : NULL, old_dentry); |
4502 | if (flags & RENAME_EXCHANGE) { | |
4503 | fsnotify_move(new_dir, old_dir, old_dentry->d_name.name, | |
4504 | new_is_dir, NULL, new_dentry); | |
4505 | } | |
4506 | } | |
49d31c2f | 4507 | release_dentry_name_snapshot(&old_name); |
0eeca283 | 4508 | |
1da177e4 LT |
4509 | return error; |
4510 | } | |
4d359507 | 4511 | EXPORT_SYMBOL(vfs_rename); |
1da177e4 | 4512 | |
ee81feb6 DB |
4513 | static int do_renameat2(int olddfd, const char __user *oldname, int newdfd, |
4514 | const char __user *newname, unsigned int flags) | |
1da177e4 | 4515 | { |
2ad94ae6 AV |
4516 | struct dentry *old_dentry, *new_dentry; |
4517 | struct dentry *trap; | |
f5beed75 AV |
4518 | struct path old_path, new_path; |
4519 | struct qstr old_last, new_last; | |
4520 | int old_type, new_type; | |
8e6d782c | 4521 | struct inode *delegated_inode = NULL; |
91a27b2a JL |
4522 | struct filename *from; |
4523 | struct filename *to; | |
f5beed75 | 4524 | unsigned int lookup_flags = 0, target_flags = LOOKUP_RENAME_TARGET; |
c6a94284 | 4525 | bool should_retry = false; |
2ad94ae6 | 4526 | int error; |
520c8b16 | 4527 | |
0d7a8555 | 4528 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
da1ce067 MS |
4529 | return -EINVAL; |
4530 | ||
0d7a8555 MS |
4531 | if ((flags & (RENAME_NOREPLACE | RENAME_WHITEOUT)) && |
4532 | (flags & RENAME_EXCHANGE)) | |
520c8b16 MS |
4533 | return -EINVAL; |
4534 | ||
0d7a8555 MS |
4535 | if ((flags & RENAME_WHITEOUT) && !capable(CAP_MKNOD)) |
4536 | return -EPERM; | |
4537 | ||
f5beed75 AV |
4538 | if (flags & RENAME_EXCHANGE) |
4539 | target_flags = 0; | |
4540 | ||
c6a94284 | 4541 | retry: |
c1d4dd27 AV |
4542 | from = filename_parentat(olddfd, getname(oldname), lookup_flags, |
4543 | &old_path, &old_last, &old_type); | |
91a27b2a JL |
4544 | if (IS_ERR(from)) { |
4545 | error = PTR_ERR(from); | |
1da177e4 | 4546 | goto exit; |
91a27b2a | 4547 | } |
1da177e4 | 4548 | |
c1d4dd27 AV |
4549 | to = filename_parentat(newdfd, getname(newname), lookup_flags, |
4550 | &new_path, &new_last, &new_type); | |
91a27b2a JL |
4551 | if (IS_ERR(to)) { |
4552 | error = PTR_ERR(to); | |
1da177e4 | 4553 | goto exit1; |
91a27b2a | 4554 | } |
1da177e4 LT |
4555 | |
4556 | error = -EXDEV; | |
f5beed75 | 4557 | if (old_path.mnt != new_path.mnt) |
1da177e4 LT |
4558 | goto exit2; |
4559 | ||
1da177e4 | 4560 | error = -EBUSY; |
f5beed75 | 4561 | if (old_type != LAST_NORM) |
1da177e4 LT |
4562 | goto exit2; |
4563 | ||
0a7c3937 MS |
4564 | if (flags & RENAME_NOREPLACE) |
4565 | error = -EEXIST; | |
f5beed75 | 4566 | if (new_type != LAST_NORM) |
1da177e4 LT |
4567 | goto exit2; |
4568 | ||
f5beed75 | 4569 | error = mnt_want_write(old_path.mnt); |
c30dabfe JK |
4570 | if (error) |
4571 | goto exit2; | |
4572 | ||
8e6d782c | 4573 | retry_deleg: |
f5beed75 | 4574 | trap = lock_rename(new_path.dentry, old_path.dentry); |
1da177e4 | 4575 | |
f5beed75 | 4576 | old_dentry = __lookup_hash(&old_last, old_path.dentry, lookup_flags); |
1da177e4 LT |
4577 | error = PTR_ERR(old_dentry); |
4578 | if (IS_ERR(old_dentry)) | |
4579 | goto exit3; | |
4580 | /* source must exist */ | |
4581 | error = -ENOENT; | |
b18825a7 | 4582 | if (d_is_negative(old_dentry)) |
1da177e4 | 4583 | goto exit4; |
f5beed75 | 4584 | new_dentry = __lookup_hash(&new_last, new_path.dentry, lookup_flags | target_flags); |
0a7c3937 MS |
4585 | error = PTR_ERR(new_dentry); |
4586 | if (IS_ERR(new_dentry)) | |
4587 | goto exit4; | |
4588 | error = -EEXIST; | |
4589 | if ((flags & RENAME_NOREPLACE) && d_is_positive(new_dentry)) | |
4590 | goto exit5; | |
da1ce067 MS |
4591 | if (flags & RENAME_EXCHANGE) { |
4592 | error = -ENOENT; | |
4593 | if (d_is_negative(new_dentry)) | |
4594 | goto exit5; | |
4595 | ||
4596 | if (!d_is_dir(new_dentry)) { | |
4597 | error = -ENOTDIR; | |
f5beed75 | 4598 | if (new_last.name[new_last.len]) |
da1ce067 MS |
4599 | goto exit5; |
4600 | } | |
4601 | } | |
1da177e4 | 4602 | /* unless the source is a directory trailing slashes give -ENOTDIR */ |
44b1d530 | 4603 | if (!d_is_dir(old_dentry)) { |
1da177e4 | 4604 | error = -ENOTDIR; |
f5beed75 | 4605 | if (old_last.name[old_last.len]) |
0a7c3937 | 4606 | goto exit5; |
f5beed75 | 4607 | if (!(flags & RENAME_EXCHANGE) && new_last.name[new_last.len]) |
0a7c3937 | 4608 | goto exit5; |
1da177e4 LT |
4609 | } |
4610 | /* source should not be ancestor of target */ | |
4611 | error = -EINVAL; | |
4612 | if (old_dentry == trap) | |
0a7c3937 | 4613 | goto exit5; |
1da177e4 | 4614 | /* target should not be an ancestor of source */ |
da1ce067 MS |
4615 | if (!(flags & RENAME_EXCHANGE)) |
4616 | error = -ENOTEMPTY; | |
1da177e4 LT |
4617 | if (new_dentry == trap) |
4618 | goto exit5; | |
4619 | ||
f5beed75 AV |
4620 | error = security_path_rename(&old_path, old_dentry, |
4621 | &new_path, new_dentry, flags); | |
be6d3e56 | 4622 | if (error) |
c30dabfe | 4623 | goto exit5; |
f5beed75 AV |
4624 | error = vfs_rename(old_path.dentry->d_inode, old_dentry, |
4625 | new_path.dentry->d_inode, new_dentry, | |
520c8b16 | 4626 | &delegated_inode, flags); |
1da177e4 LT |
4627 | exit5: |
4628 | dput(new_dentry); | |
4629 | exit4: | |
4630 | dput(old_dentry); | |
4631 | exit3: | |
f5beed75 | 4632 | unlock_rename(new_path.dentry, old_path.dentry); |
8e6d782c BF |
4633 | if (delegated_inode) { |
4634 | error = break_deleg_wait(&delegated_inode); | |
4635 | if (!error) | |
4636 | goto retry_deleg; | |
4637 | } | |
f5beed75 | 4638 | mnt_drop_write(old_path.mnt); |
1da177e4 | 4639 | exit2: |
c6a94284 JL |
4640 | if (retry_estale(error, lookup_flags)) |
4641 | should_retry = true; | |
f5beed75 | 4642 | path_put(&new_path); |
2ad94ae6 | 4643 | putname(to); |
1da177e4 | 4644 | exit1: |
f5beed75 | 4645 | path_put(&old_path); |
1da177e4 | 4646 | putname(from); |
c6a94284 JL |
4647 | if (should_retry) { |
4648 | should_retry = false; | |
4649 | lookup_flags |= LOOKUP_REVAL; | |
4650 | goto retry; | |
4651 | } | |
2ad94ae6 | 4652 | exit: |
1da177e4 LT |
4653 | return error; |
4654 | } | |
4655 | ||
ee81feb6 DB |
4656 | SYSCALL_DEFINE5(renameat2, int, olddfd, const char __user *, oldname, |
4657 | int, newdfd, const char __user *, newname, unsigned int, flags) | |
4658 | { | |
4659 | return do_renameat2(olddfd, oldname, newdfd, newname, flags); | |
4660 | } | |
4661 | ||
520c8b16 MS |
4662 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
4663 | int, newdfd, const char __user *, newname) | |
4664 | { | |
ee81feb6 | 4665 | return do_renameat2(olddfd, oldname, newdfd, newname, 0); |
520c8b16 MS |
4666 | } |
4667 | ||
a26eab24 | 4668 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4669 | { |
ee81feb6 | 4670 | return do_renameat2(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4671 | } |
4672 | ||
787fb6bc MS |
4673 | int vfs_whiteout(struct inode *dir, struct dentry *dentry) |
4674 | { | |
4675 | int error = may_create(dir, dentry); | |
4676 | if (error) | |
4677 | return error; | |
4678 | ||
4679 | if (!dir->i_op->mknod) | |
4680 | return -EPERM; | |
4681 | ||
4682 | return dir->i_op->mknod(dir, dentry, | |
4683 | S_IFCHR | WHITEOUT_MODE, WHITEOUT_DEV); | |
4684 | } | |
4685 | EXPORT_SYMBOL(vfs_whiteout); | |
4686 | ||
5d826c84 | 4687 | int readlink_copy(char __user *buffer, int buflen, const char *link) |
1da177e4 | 4688 | { |
5d826c84 | 4689 | int len = PTR_ERR(link); |
1da177e4 LT |
4690 | if (IS_ERR(link)) |
4691 | goto out; | |
4692 | ||
4693 | len = strlen(link); | |
4694 | if (len > (unsigned) buflen) | |
4695 | len = buflen; | |
4696 | if (copy_to_user(buffer, link, len)) | |
4697 | len = -EFAULT; | |
4698 | out: | |
4699 | return len; | |
4700 | } | |
4701 | ||
fd4a0edf MS |
4702 | /** |
4703 | * vfs_readlink - copy symlink body into userspace buffer | |
4704 | * @dentry: dentry on which to get symbolic link | |
4705 | * @buffer: user memory pointer | |
4706 | * @buflen: size of buffer | |
4707 | * | |
4708 | * Does not touch atime. That's up to the caller if necessary | |
4709 | * | |
4710 | * Does not call security hook. | |
4711 | */ | |
4712 | int vfs_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4713 | { | |
4714 | struct inode *inode = d_inode(dentry); | |
f2df5da6 AV |
4715 | DEFINE_DELAYED_CALL(done); |
4716 | const char *link; | |
4717 | int res; | |
fd4a0edf | 4718 | |
76fca90e MS |
4719 | if (unlikely(!(inode->i_opflags & IOP_DEFAULT_READLINK))) { |
4720 | if (unlikely(inode->i_op->readlink)) | |
4721 | return inode->i_op->readlink(dentry, buffer, buflen); | |
4722 | ||
4723 | if (!d_is_symlink(dentry)) | |
4724 | return -EINVAL; | |
4725 | ||
4726 | spin_lock(&inode->i_lock); | |
4727 | inode->i_opflags |= IOP_DEFAULT_READLINK; | |
4728 | spin_unlock(&inode->i_lock); | |
4729 | } | |
fd4a0edf | 4730 | |
f2df5da6 AV |
4731 | link = inode->i_link; |
4732 | if (!link) { | |
4733 | link = inode->i_op->get_link(dentry, inode, &done); | |
4734 | if (IS_ERR(link)) | |
4735 | return PTR_ERR(link); | |
4736 | } | |
4737 | res = readlink_copy(buffer, buflen, link); | |
4738 | do_delayed_call(&done); | |
4739 | return res; | |
fd4a0edf MS |
4740 | } |
4741 | EXPORT_SYMBOL(vfs_readlink); | |
1da177e4 | 4742 | |
d60874cd MS |
4743 | /** |
4744 | * vfs_get_link - get symlink body | |
4745 | * @dentry: dentry on which to get symbolic link | |
4746 | * @done: caller needs to free returned data with this | |
4747 | * | |
4748 | * Calls security hook and i_op->get_link() on the supplied inode. | |
4749 | * | |
4750 | * It does not touch atime. That's up to the caller if necessary. | |
4751 | * | |
4752 | * Does not work on "special" symlinks like /proc/$$/fd/N | |
4753 | */ | |
4754 | const char *vfs_get_link(struct dentry *dentry, struct delayed_call *done) | |
4755 | { | |
4756 | const char *res = ERR_PTR(-EINVAL); | |
4757 | struct inode *inode = d_inode(dentry); | |
4758 | ||
4759 | if (d_is_symlink(dentry)) { | |
4760 | res = ERR_PTR(security_inode_readlink(dentry)); | |
4761 | if (!res) | |
4762 | res = inode->i_op->get_link(dentry, inode, done); | |
4763 | } | |
4764 | return res; | |
4765 | } | |
4766 | EXPORT_SYMBOL(vfs_get_link); | |
4767 | ||
1da177e4 | 4768 | /* get the link contents into pagecache */ |
6b255391 | 4769 | const char *page_get_link(struct dentry *dentry, struct inode *inode, |
fceef393 | 4770 | struct delayed_call *callback) |
1da177e4 | 4771 | { |
ebd09abb DG |
4772 | char *kaddr; |
4773 | struct page *page; | |
6b255391 AV |
4774 | struct address_space *mapping = inode->i_mapping; |
4775 | ||
d3883d4f AV |
4776 | if (!dentry) { |
4777 | page = find_get_page(mapping, 0); | |
4778 | if (!page) | |
4779 | return ERR_PTR(-ECHILD); | |
4780 | if (!PageUptodate(page)) { | |
4781 | put_page(page); | |
4782 | return ERR_PTR(-ECHILD); | |
4783 | } | |
4784 | } else { | |
4785 | page = read_mapping_page(mapping, 0, NULL); | |
4786 | if (IS_ERR(page)) | |
4787 | return (char*)page; | |
4788 | } | |
fceef393 | 4789 | set_delayed_call(callback, page_put_link, page); |
21fc61c7 AV |
4790 | BUG_ON(mapping_gfp_mask(mapping) & __GFP_HIGHMEM); |
4791 | kaddr = page_address(page); | |
6b255391 | 4792 | nd_terminate_link(kaddr, inode->i_size, PAGE_SIZE - 1); |
ebd09abb | 4793 | return kaddr; |
1da177e4 LT |
4794 | } |
4795 | ||
6b255391 | 4796 | EXPORT_SYMBOL(page_get_link); |
1da177e4 | 4797 | |
fceef393 | 4798 | void page_put_link(void *arg) |
1da177e4 | 4799 | { |
fceef393 | 4800 | put_page(arg); |
1da177e4 | 4801 | } |
4d359507 | 4802 | EXPORT_SYMBOL(page_put_link); |
1da177e4 | 4803 | |
aa80deab AV |
4804 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) |
4805 | { | |
fceef393 | 4806 | DEFINE_DELAYED_CALL(done); |
6b255391 AV |
4807 | int res = readlink_copy(buffer, buflen, |
4808 | page_get_link(dentry, d_inode(dentry), | |
fceef393 AV |
4809 | &done)); |
4810 | do_delayed_call(&done); | |
aa80deab AV |
4811 | return res; |
4812 | } | |
4813 | EXPORT_SYMBOL(page_readlink); | |
4814 | ||
54566b2c NP |
4815 | /* |
4816 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
4817 | */ | |
4818 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
4819 | { |
4820 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 4821 | struct page *page; |
afddba49 | 4822 | void *fsdata; |
beb497ab | 4823 | int err; |
c718a975 | 4824 | unsigned int flags = 0; |
54566b2c NP |
4825 | if (nofs) |
4826 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 4827 | |
7e53cac4 | 4828 | retry: |
afddba49 | 4829 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 4830 | flags, &page, &fsdata); |
1da177e4 | 4831 | if (err) |
afddba49 NP |
4832 | goto fail; |
4833 | ||
21fc61c7 | 4834 | memcpy(page_address(page), symname, len-1); |
afddba49 NP |
4835 | |
4836 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
4837 | page, fsdata); | |
1da177e4 LT |
4838 | if (err < 0) |
4839 | goto fail; | |
afddba49 NP |
4840 | if (err < len-1) |
4841 | goto retry; | |
4842 | ||
1da177e4 LT |
4843 | mark_inode_dirty(inode); |
4844 | return 0; | |
1da177e4 LT |
4845 | fail: |
4846 | return err; | |
4847 | } | |
4d359507 | 4848 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 | 4849 | |
0adb25d2 KK |
4850 | int page_symlink(struct inode *inode, const char *symname, int len) |
4851 | { | |
4852 | return __page_symlink(inode, symname, len, | |
c62d2555 | 4853 | !mapping_gfp_constraint(inode->i_mapping, __GFP_FS)); |
0adb25d2 | 4854 | } |
4d359507 | 4855 | EXPORT_SYMBOL(page_symlink); |
0adb25d2 | 4856 | |
92e1d5be | 4857 | const struct inode_operations page_symlink_inode_operations = { |
6b255391 | 4858 | .get_link = page_get_link, |
1da177e4 | 4859 | }; |
1da177e4 | 4860 | EXPORT_SYMBOL(page_symlink_inode_operations); |