]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/namei.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * Some corrections by tytso. | |
9 | */ | |
10 | ||
11 | /* [Feb 1997 T. Schoebel-Theuer] Complete rewrite of the pathname | |
12 | * lookup logic. | |
13 | */ | |
14 | /* [Feb-Apr 2000, AV] Rewrite to the new namespace architecture. | |
15 | */ | |
16 | ||
17 | #include <linux/init.h> | |
630d9c47 | 18 | #include <linux/export.h> |
44696908 | 19 | #include <linux/kernel.h> |
1da177e4 LT |
20 | #include <linux/slab.h> |
21 | #include <linux/fs.h> | |
22 | #include <linux/namei.h> | |
1da177e4 | 23 | #include <linux/pagemap.h> |
0eeca283 | 24 | #include <linux/fsnotify.h> |
1da177e4 LT |
25 | #include <linux/personality.h> |
26 | #include <linux/security.h> | |
6146f0d5 | 27 | #include <linux/ima.h> |
1da177e4 LT |
28 | #include <linux/syscalls.h> |
29 | #include <linux/mount.h> | |
30 | #include <linux/audit.h> | |
16f7e0fe | 31 | #include <linux/capability.h> |
834f2a4a | 32 | #include <linux/file.h> |
5590ff0d | 33 | #include <linux/fcntl.h> |
08ce5f16 | 34 | #include <linux/device_cgroup.h> |
5ad4e53b | 35 | #include <linux/fs_struct.h> |
e77819e5 | 36 | #include <linux/posix_acl.h> |
99d263d4 | 37 | #include <linux/hash.h> |
1da177e4 LT |
38 | #include <asm/uaccess.h> |
39 | ||
e81e3f4d | 40 | #include "internal.h" |
c7105365 | 41 | #include "mount.h" |
e81e3f4d | 42 | |
1da177e4 LT |
43 | /* [Feb-1997 T. Schoebel-Theuer] |
44 | * Fundamental changes in the pathname lookup mechanisms (namei) | |
45 | * were necessary because of omirr. The reason is that omirr needs | |
46 | * to know the _real_ pathname, not the user-supplied one, in case | |
47 | * of symlinks (and also when transname replacements occur). | |
48 | * | |
49 | * The new code replaces the old recursive symlink resolution with | |
50 | * an iterative one (in case of non-nested symlink chains). It does | |
51 | * this with calls to <fs>_follow_link(). | |
52 | * As a side effect, dir_namei(), _namei() and follow_link() are now | |
53 | * replaced with a single function lookup_dentry() that can handle all | |
54 | * the special cases of the former code. | |
55 | * | |
56 | * With the new dcache, the pathname is stored at each inode, at least as | |
57 | * long as the refcount of the inode is positive. As a side effect, the | |
58 | * size of the dcache depends on the inode cache and thus is dynamic. | |
59 | * | |
60 | * [29-Apr-1998 C. Scott Ananian] Updated above description of symlink | |
61 | * resolution to correspond with current state of the code. | |
62 | * | |
63 | * Note that the symlink resolution is not *completely* iterative. | |
64 | * There is still a significant amount of tail- and mid- recursion in | |
65 | * the algorithm. Also, note that <fs>_readlink() is not used in | |
66 | * lookup_dentry(): lookup_dentry() on the result of <fs>_readlink() | |
67 | * may return different results than <fs>_follow_link(). Many virtual | |
68 | * filesystems (including /proc) exhibit this behavior. | |
69 | */ | |
70 | ||
71 | /* [24-Feb-97 T. Schoebel-Theuer] Side effects caused by new implementation: | |
72 | * New symlink semantics: when open() is called with flags O_CREAT | O_EXCL | |
73 | * and the name already exists in form of a symlink, try to create the new | |
74 | * name indicated by the symlink. The old code always complained that the | |
75 | * name already exists, due to not following the symlink even if its target | |
76 | * is nonexistent. The new semantics affects also mknod() and link() when | |
25985edc | 77 | * the name is a symlink pointing to a non-existent name. |
1da177e4 LT |
78 | * |
79 | * I don't know which semantics is the right one, since I have no access | |
80 | * to standards. But I found by trial that HP-UX 9.0 has the full "new" | |
81 | * semantics implemented, while SunOS 4.1.1 and Solaris (SunOS 5.4) have the | |
82 | * "old" one. Personally, I think the new semantics is much more logical. | |
83 | * Note that "ln old new" where "new" is a symlink pointing to a non-existing | |
84 | * file does succeed in both HP-UX and SunOs, but not in Solaris | |
85 | * and in the old Linux semantics. | |
86 | */ | |
87 | ||
88 | /* [16-Dec-97 Kevin Buhr] For security reasons, we change some symlink | |
89 | * semantics. See the comments in "open_namei" and "do_link" below. | |
90 | * | |
91 | * [10-Sep-98 Alan Modra] Another symlink change. | |
92 | */ | |
93 | ||
94 | /* [Feb-Apr 2000 AV] Complete rewrite. Rules for symlinks: | |
95 | * inside the path - always follow. | |
96 | * in the last component in creation/removal/renaming - never follow. | |
97 | * if LOOKUP_FOLLOW passed - follow. | |
98 | * if the pathname has trailing slashes - follow. | |
99 | * otherwise - don't follow. | |
100 | * (applied in that order). | |
101 | * | |
102 | * [Jun 2000 AV] Inconsistent behaviour of open() in case if flags==O_CREAT | |
103 | * restored for 2.4. This is the last surviving part of old 4.2BSD bug. | |
104 | * During the 2.4 we need to fix the userland stuff depending on it - | |
105 | * hopefully we will be able to get rid of that wart in 2.5. So far only | |
106 | * XEmacs seems to be relying on it... | |
107 | */ | |
108 | /* | |
109 | * [Sep 2001 AV] Single-semaphore locking scheme (kudos to David Holland) | |
a11f3a05 | 110 | * implemented. Let's see if raised priority of ->s_vfs_rename_mutex gives |
1da177e4 LT |
111 | * any extra contention... |
112 | */ | |
113 | ||
114 | /* In order to reduce some races, while at the same time doing additional | |
115 | * checking and hopefully speeding things up, we copy filenames to the | |
116 | * kernel data space before using them.. | |
117 | * | |
118 | * POSIX.1 2.4: an empty pathname is invalid (ENOENT). | |
119 | * PATH_MAX includes the nul terminator --RR. | |
120 | */ | |
91a27b2a | 121 | |
fd2f7cb5 | 122 | #define EMBEDDED_NAME_MAX (PATH_MAX - offsetof(struct filename, iname)) |
7950e385 | 123 | |
51f39a1f | 124 | struct filename * |
91a27b2a JL |
125 | getname_flags(const char __user *filename, int flags, int *empty) |
126 | { | |
94b5d262 | 127 | struct filename *result; |
7950e385 | 128 | char *kname; |
94b5d262 | 129 | int len; |
4043cde8 | 130 | |
7ac86265 JL |
131 | result = audit_reusename(filename); |
132 | if (result) | |
133 | return result; | |
134 | ||
7950e385 | 135 | result = __getname(); |
3f9f0aa6 | 136 | if (unlikely(!result)) |
4043cde8 EP |
137 | return ERR_PTR(-ENOMEM); |
138 | ||
7950e385 JL |
139 | /* |
140 | * First, try to embed the struct filename inside the names_cache | |
141 | * allocation | |
142 | */ | |
fd2f7cb5 | 143 | kname = (char *)result->iname; |
91a27b2a | 144 | result->name = kname; |
7950e385 | 145 | |
94b5d262 | 146 | len = strncpy_from_user(kname, filename, EMBEDDED_NAME_MAX); |
91a27b2a | 147 | if (unlikely(len < 0)) { |
94b5d262 AV |
148 | __putname(result); |
149 | return ERR_PTR(len); | |
91a27b2a | 150 | } |
3f9f0aa6 | 151 | |
7950e385 JL |
152 | /* |
153 | * Uh-oh. We have a name that's approaching PATH_MAX. Allocate a | |
154 | * separate struct filename so we can dedicate the entire | |
155 | * names_cache allocation for the pathname, and re-do the copy from | |
156 | * userland. | |
157 | */ | |
94b5d262 | 158 | if (unlikely(len == EMBEDDED_NAME_MAX)) { |
fd2f7cb5 | 159 | const size_t size = offsetof(struct filename, iname[1]); |
7950e385 JL |
160 | kname = (char *)result; |
161 | ||
fd2f7cb5 AV |
162 | /* |
163 | * size is chosen that way we to guarantee that | |
164 | * result->iname[0] is within the same object and that | |
165 | * kname can't be equal to result->iname, no matter what. | |
166 | */ | |
167 | result = kzalloc(size, GFP_KERNEL); | |
94b5d262 AV |
168 | if (unlikely(!result)) { |
169 | __putname(kname); | |
170 | return ERR_PTR(-ENOMEM); | |
7950e385 JL |
171 | } |
172 | result->name = kname; | |
94b5d262 AV |
173 | len = strncpy_from_user(kname, filename, PATH_MAX); |
174 | if (unlikely(len < 0)) { | |
175 | __putname(kname); | |
176 | kfree(result); | |
177 | return ERR_PTR(len); | |
178 | } | |
179 | if (unlikely(len == PATH_MAX)) { | |
180 | __putname(kname); | |
181 | kfree(result); | |
182 | return ERR_PTR(-ENAMETOOLONG); | |
183 | } | |
7950e385 JL |
184 | } |
185 | ||
94b5d262 | 186 | result->refcnt = 1; |
3f9f0aa6 LT |
187 | /* The empty path is special. */ |
188 | if (unlikely(!len)) { | |
189 | if (empty) | |
4043cde8 | 190 | *empty = 1; |
94b5d262 AV |
191 | if (!(flags & LOOKUP_EMPTY)) { |
192 | putname(result); | |
193 | return ERR_PTR(-ENOENT); | |
194 | } | |
1da177e4 | 195 | } |
3f9f0aa6 | 196 | |
7950e385 | 197 | result->uptr = filename; |
c4ad8f98 | 198 | result->aname = NULL; |
7950e385 JL |
199 | audit_getname(result); |
200 | return result; | |
1da177e4 LT |
201 | } |
202 | ||
91a27b2a JL |
203 | struct filename * |
204 | getname(const char __user * filename) | |
f52e0c11 | 205 | { |
f7493e5d | 206 | return getname_flags(filename, 0, NULL); |
f52e0c11 AV |
207 | } |
208 | ||
c4ad8f98 LT |
209 | struct filename * |
210 | getname_kernel(const char * filename) | |
211 | { | |
212 | struct filename *result; | |
08518549 | 213 | int len = strlen(filename) + 1; |
c4ad8f98 LT |
214 | |
215 | result = __getname(); | |
216 | if (unlikely(!result)) | |
217 | return ERR_PTR(-ENOMEM); | |
218 | ||
08518549 | 219 | if (len <= EMBEDDED_NAME_MAX) { |
fd2f7cb5 | 220 | result->name = (char *)result->iname; |
08518549 PM |
221 | } else if (len <= PATH_MAX) { |
222 | struct filename *tmp; | |
223 | ||
224 | tmp = kmalloc(sizeof(*tmp), GFP_KERNEL); | |
225 | if (unlikely(!tmp)) { | |
226 | __putname(result); | |
227 | return ERR_PTR(-ENOMEM); | |
228 | } | |
229 | tmp->name = (char *)result; | |
08518549 PM |
230 | result = tmp; |
231 | } else { | |
232 | __putname(result); | |
233 | return ERR_PTR(-ENAMETOOLONG); | |
234 | } | |
235 | memcpy((char *)result->name, filename, len); | |
c4ad8f98 LT |
236 | result->uptr = NULL; |
237 | result->aname = NULL; | |
55422d0b | 238 | result->refcnt = 1; |
fd3522fd | 239 | audit_getname(result); |
c4ad8f98 | 240 | |
c4ad8f98 LT |
241 | return result; |
242 | } | |
243 | ||
91a27b2a | 244 | void putname(struct filename *name) |
1da177e4 | 245 | { |
55422d0b PM |
246 | BUG_ON(name->refcnt <= 0); |
247 | ||
248 | if (--name->refcnt > 0) | |
249 | return; | |
250 | ||
fd2f7cb5 | 251 | if (name->name != name->iname) { |
55422d0b PM |
252 | __putname(name->name); |
253 | kfree(name); | |
254 | } else | |
255 | __putname(name); | |
1da177e4 | 256 | } |
1da177e4 | 257 | |
e77819e5 LT |
258 | static int check_acl(struct inode *inode, int mask) |
259 | { | |
84635d68 | 260 | #ifdef CONFIG_FS_POSIX_ACL |
e77819e5 LT |
261 | struct posix_acl *acl; |
262 | ||
e77819e5 | 263 | if (mask & MAY_NOT_BLOCK) { |
3567866b AV |
264 | acl = get_cached_acl_rcu(inode, ACL_TYPE_ACCESS); |
265 | if (!acl) | |
e77819e5 | 266 | return -EAGAIN; |
3567866b AV |
267 | /* no ->get_acl() calls in RCU mode... */ |
268 | if (acl == ACL_NOT_CACHED) | |
269 | return -ECHILD; | |
206b1d09 | 270 | return posix_acl_permission(inode, acl, mask & ~MAY_NOT_BLOCK); |
e77819e5 LT |
271 | } |
272 | ||
2982baa2 CH |
273 | acl = get_acl(inode, ACL_TYPE_ACCESS); |
274 | if (IS_ERR(acl)) | |
275 | return PTR_ERR(acl); | |
e77819e5 LT |
276 | if (acl) { |
277 | int error = posix_acl_permission(inode, acl, mask); | |
278 | posix_acl_release(acl); | |
279 | return error; | |
280 | } | |
84635d68 | 281 | #endif |
e77819e5 LT |
282 | |
283 | return -EAGAIN; | |
284 | } | |
285 | ||
5909ccaa | 286 | /* |
948409c7 | 287 | * This does the basic permission checking |
1da177e4 | 288 | */ |
7e40145e | 289 | static int acl_permission_check(struct inode *inode, int mask) |
1da177e4 | 290 | { |
26cf46be | 291 | unsigned int mode = inode->i_mode; |
1da177e4 | 292 | |
8e96e3b7 | 293 | if (likely(uid_eq(current_fsuid(), inode->i_uid))) |
1da177e4 LT |
294 | mode >>= 6; |
295 | else { | |
e77819e5 | 296 | if (IS_POSIXACL(inode) && (mode & S_IRWXG)) { |
7e40145e | 297 | int error = check_acl(inode, mask); |
b74c79e9 NP |
298 | if (error != -EAGAIN) |
299 | return error; | |
1da177e4 LT |
300 | } |
301 | ||
302 | if (in_group_p(inode->i_gid)) | |
303 | mode >>= 3; | |
304 | } | |
305 | ||
306 | /* | |
307 | * If the DACs are ok we don't need any capability check. | |
308 | */ | |
9c2c7039 | 309 | if ((mask & ~mode & (MAY_READ | MAY_WRITE | MAY_EXEC)) == 0) |
1da177e4 | 310 | return 0; |
5909ccaa LT |
311 | return -EACCES; |
312 | } | |
313 | ||
314 | /** | |
b74c79e9 | 315 | * generic_permission - check for access rights on a Posix-like filesystem |
5909ccaa | 316 | * @inode: inode to check access rights for |
8fd90c8d | 317 | * @mask: right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC, ...) |
5909ccaa LT |
318 | * |
319 | * Used to check for read/write/execute permissions on a file. | |
320 | * We use "fsuid" for this, letting us set arbitrary permissions | |
321 | * for filesystem access without changing the "normal" uids which | |
b74c79e9 NP |
322 | * are used for other things. |
323 | * | |
324 | * generic_permission is rcu-walk aware. It returns -ECHILD in case an rcu-walk | |
325 | * request cannot be satisfied (eg. requires blocking or too much complexity). | |
326 | * It would then be called again in ref-walk mode. | |
5909ccaa | 327 | */ |
2830ba7f | 328 | int generic_permission(struct inode *inode, int mask) |
5909ccaa LT |
329 | { |
330 | int ret; | |
331 | ||
332 | /* | |
948409c7 | 333 | * Do the basic permission checks. |
5909ccaa | 334 | */ |
7e40145e | 335 | ret = acl_permission_check(inode, mask); |
5909ccaa LT |
336 | if (ret != -EACCES) |
337 | return ret; | |
1da177e4 | 338 | |
d594e7ec AV |
339 | if (S_ISDIR(inode->i_mode)) { |
340 | /* DACs are overridable for directories */ | |
23adbe12 | 341 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_OVERRIDE)) |
d594e7ec AV |
342 | return 0; |
343 | if (!(mask & MAY_WRITE)) | |
23adbe12 AL |
344 | if (capable_wrt_inode_uidgid(inode, |
345 | CAP_DAC_READ_SEARCH)) | |
d594e7ec AV |
346 | return 0; |
347 | return -EACCES; | |
348 | } | |
1da177e4 LT |
349 | /* |
350 | * Read/write DACs are always overridable. | |
d594e7ec AV |
351 | * Executable DACs are overridable when there is |
352 | * at least one exec bit set. | |
1da177e4 | 353 | */ |
d594e7ec | 354 | if (!(mask & MAY_EXEC) || (inode->i_mode & S_IXUGO)) |
23adbe12 | 355 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_OVERRIDE)) |
1da177e4 LT |
356 | return 0; |
357 | ||
358 | /* | |
359 | * Searching includes executable on directories, else just read. | |
360 | */ | |
7ea66001 | 361 | mask &= MAY_READ | MAY_WRITE | MAY_EXEC; |
d594e7ec | 362 | if (mask == MAY_READ) |
23adbe12 | 363 | if (capable_wrt_inode_uidgid(inode, CAP_DAC_READ_SEARCH)) |
1da177e4 LT |
364 | return 0; |
365 | ||
366 | return -EACCES; | |
367 | } | |
4d359507 | 368 | EXPORT_SYMBOL(generic_permission); |
1da177e4 | 369 | |
3ddcd056 LT |
370 | /* |
371 | * We _really_ want to just do "generic_permission()" without | |
372 | * even looking at the inode->i_op values. So we keep a cache | |
373 | * flag in inode->i_opflags, that says "this has not special | |
374 | * permission function, use the fast case". | |
375 | */ | |
376 | static inline int do_inode_permission(struct inode *inode, int mask) | |
377 | { | |
378 | if (unlikely(!(inode->i_opflags & IOP_FASTPERM))) { | |
379 | if (likely(inode->i_op->permission)) | |
380 | return inode->i_op->permission(inode, mask); | |
381 | ||
382 | /* This gets set once for the inode lifetime */ | |
383 | spin_lock(&inode->i_lock); | |
384 | inode->i_opflags |= IOP_FASTPERM; | |
385 | spin_unlock(&inode->i_lock); | |
386 | } | |
387 | return generic_permission(inode, mask); | |
388 | } | |
389 | ||
cb23beb5 | 390 | /** |
0bdaea90 DH |
391 | * __inode_permission - Check for access rights to a given inode |
392 | * @inode: Inode to check permission on | |
393 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
cb23beb5 | 394 | * |
0bdaea90 | 395 | * Check for read/write/execute permissions on an inode. |
948409c7 AG |
396 | * |
397 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
0bdaea90 DH |
398 | * |
399 | * This does not check for a read-only file system. You probably want | |
400 | * inode_permission(). | |
cb23beb5 | 401 | */ |
0bdaea90 | 402 | int __inode_permission(struct inode *inode, int mask) |
1da177e4 | 403 | { |
e6305c43 | 404 | int retval; |
1da177e4 | 405 | |
3ddcd056 | 406 | if (unlikely(mask & MAY_WRITE)) { |
1da177e4 LT |
407 | /* |
408 | * Nobody gets write access to an immutable file. | |
409 | */ | |
410 | if (IS_IMMUTABLE(inode)) | |
411 | return -EACCES; | |
412 | } | |
413 | ||
3ddcd056 | 414 | retval = do_inode_permission(inode, mask); |
1da177e4 LT |
415 | if (retval) |
416 | return retval; | |
417 | ||
08ce5f16 SH |
418 | retval = devcgroup_inode_permission(inode, mask); |
419 | if (retval) | |
420 | return retval; | |
421 | ||
d09ca739 | 422 | return security_inode_permission(inode, mask); |
1da177e4 | 423 | } |
bd5d0856 | 424 | EXPORT_SYMBOL(__inode_permission); |
1da177e4 | 425 | |
0bdaea90 DH |
426 | /** |
427 | * sb_permission - Check superblock-level permissions | |
428 | * @sb: Superblock of inode to check permission on | |
55852635 | 429 | * @inode: Inode to check permission on |
0bdaea90 DH |
430 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) |
431 | * | |
432 | * Separate out file-system wide checks from inode-specific permission checks. | |
433 | */ | |
434 | static int sb_permission(struct super_block *sb, struct inode *inode, int mask) | |
435 | { | |
436 | if (unlikely(mask & MAY_WRITE)) { | |
437 | umode_t mode = inode->i_mode; | |
438 | ||
439 | /* Nobody gets write access to a read-only fs. */ | |
440 | if ((sb->s_flags & MS_RDONLY) && | |
441 | (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) | |
442 | return -EROFS; | |
443 | } | |
444 | return 0; | |
445 | } | |
446 | ||
447 | /** | |
448 | * inode_permission - Check for access rights to a given inode | |
449 | * @inode: Inode to check permission on | |
450 | * @mask: Right to check for (%MAY_READ, %MAY_WRITE, %MAY_EXEC) | |
451 | * | |
452 | * Check for read/write/execute permissions on an inode. We use fs[ug]id for | |
453 | * this, letting us set arbitrary permissions for filesystem access without | |
454 | * changing the "normal" UIDs which are used for other things. | |
455 | * | |
456 | * When checking for MAY_APPEND, MAY_WRITE must also be set in @mask. | |
457 | */ | |
458 | int inode_permission(struct inode *inode, int mask) | |
459 | { | |
460 | int retval; | |
461 | ||
462 | retval = sb_permission(inode->i_sb, inode, mask); | |
463 | if (retval) | |
464 | return retval; | |
465 | return __inode_permission(inode, mask); | |
466 | } | |
4d359507 | 467 | EXPORT_SYMBOL(inode_permission); |
0bdaea90 | 468 | |
5dd784d0 JB |
469 | /** |
470 | * path_get - get a reference to a path | |
471 | * @path: path to get the reference to | |
472 | * | |
473 | * Given a path increment the reference count to the dentry and the vfsmount. | |
474 | */ | |
dcf787f3 | 475 | void path_get(const struct path *path) |
5dd784d0 JB |
476 | { |
477 | mntget(path->mnt); | |
478 | dget(path->dentry); | |
479 | } | |
480 | EXPORT_SYMBOL(path_get); | |
481 | ||
1d957f9b JB |
482 | /** |
483 | * path_put - put a reference to a path | |
484 | * @path: path to put the reference to | |
485 | * | |
486 | * Given a path decrement the reference count to the dentry and the vfsmount. | |
487 | */ | |
dcf787f3 | 488 | void path_put(const struct path *path) |
1da177e4 | 489 | { |
1d957f9b JB |
490 | dput(path->dentry); |
491 | mntput(path->mnt); | |
1da177e4 | 492 | } |
1d957f9b | 493 | EXPORT_SYMBOL(path_put); |
1da177e4 | 494 | |
894bc8c4 | 495 | #define EMBEDDED_LEVELS 2 |
1f55a6ec AV |
496 | struct nameidata { |
497 | struct path path; | |
1cf2665b | 498 | struct qstr last; |
1f55a6ec AV |
499 | struct path root; |
500 | struct inode *inode; /* path.dentry.d_inode */ | |
501 | unsigned int flags; | |
502 | unsigned seq, m_seq; | |
503 | int last_type; | |
504 | unsigned depth; | |
756daf26 | 505 | int total_link_count; |
697fc6ca AV |
506 | struct saved { |
507 | struct path link; | |
508 | void *cookie; | |
509 | const char *name; | |
237d8b32 | 510 | struct inode *inode; |
0450b2d1 | 511 | unsigned seq; |
894bc8c4 | 512 | } *stack, internal[EMBEDDED_LEVELS]; |
1f55a6ec AV |
513 | }; |
514 | ||
756daf26 | 515 | static struct nameidata *set_nameidata(struct nameidata *p) |
894bc8c4 | 516 | { |
756daf26 N |
517 | struct nameidata *old = current->nameidata; |
518 | p->stack = p->internal; | |
519 | p->total_link_count = old ? old->total_link_count : 0; | |
520 | current->nameidata = p; | |
521 | return old; | |
894bc8c4 AV |
522 | } |
523 | ||
756daf26 | 524 | static void restore_nameidata(struct nameidata *old) |
894bc8c4 | 525 | { |
756daf26 N |
526 | struct nameidata *now = current->nameidata; |
527 | ||
528 | current->nameidata = old; | |
529 | if (old) | |
530 | old->total_link_count = now->total_link_count; | |
531 | if (now->stack != now->internal) { | |
532 | kfree(now->stack); | |
533 | now->stack = now->internal; | |
894bc8c4 AV |
534 | } |
535 | } | |
536 | ||
537 | static int __nd_alloc_stack(struct nameidata *nd) | |
538 | { | |
e269f2a7 | 539 | struct saved *p = kmalloc(MAXSYMLINKS * sizeof(struct saved), |
894bc8c4 AV |
540 | GFP_KERNEL); |
541 | if (unlikely(!p)) | |
542 | return -ENOMEM; | |
543 | memcpy(p, nd->internal, sizeof(nd->internal)); | |
544 | nd->stack = p; | |
545 | return 0; | |
546 | } | |
547 | ||
548 | static inline int nd_alloc_stack(struct nameidata *nd) | |
549 | { | |
da4e0be0 | 550 | if (likely(nd->depth != EMBEDDED_LEVELS)) |
894bc8c4 AV |
551 | return 0; |
552 | if (likely(nd->stack != nd->internal)) | |
553 | return 0; | |
554 | return __nd_alloc_stack(nd); | |
555 | } | |
556 | ||
7973387a AV |
557 | static void drop_links(struct nameidata *nd) |
558 | { | |
559 | int i = nd->depth; | |
560 | while (i--) { | |
561 | struct saved *last = nd->stack + i; | |
562 | struct inode *inode = last->inode; | |
563 | if (last->cookie && inode->i_op->put_link) { | |
564 | inode->i_op->put_link(inode, last->cookie); | |
565 | last->cookie = NULL; | |
566 | } | |
567 | } | |
568 | } | |
569 | ||
570 | static void terminate_walk(struct nameidata *nd) | |
571 | { | |
572 | drop_links(nd); | |
573 | if (!(nd->flags & LOOKUP_RCU)) { | |
574 | int i; | |
575 | path_put(&nd->path); | |
576 | for (i = 0; i < nd->depth; i++) | |
577 | path_put(&nd->stack[i].link); | |
578 | } else { | |
579 | nd->flags &= ~LOOKUP_RCU; | |
580 | if (!(nd->flags & LOOKUP_ROOT)) | |
581 | nd->root.mnt = NULL; | |
582 | rcu_read_unlock(); | |
583 | } | |
584 | nd->depth = 0; | |
585 | } | |
586 | ||
587 | /* path_put is needed afterwards regardless of success or failure */ | |
588 | static bool legitimize_path(struct nameidata *nd, | |
589 | struct path *path, unsigned seq) | |
590 | { | |
591 | int res = __legitimize_mnt(path->mnt, nd->m_seq); | |
592 | if (unlikely(res)) { | |
593 | if (res > 0) | |
594 | path->mnt = NULL; | |
595 | path->dentry = NULL; | |
596 | return false; | |
597 | } | |
598 | if (unlikely(!lockref_get_not_dead(&path->dentry->d_lockref))) { | |
599 | path->dentry = NULL; | |
600 | return false; | |
601 | } | |
602 | return !read_seqcount_retry(&path->dentry->d_seq, seq); | |
603 | } | |
604 | ||
605 | static bool legitimize_links(struct nameidata *nd) | |
606 | { | |
607 | int i; | |
608 | for (i = 0; i < nd->depth; i++) { | |
609 | struct saved *last = nd->stack + i; | |
610 | if (unlikely(!legitimize_path(nd, &last->link, last->seq))) { | |
611 | drop_links(nd); | |
612 | nd->depth = i + 1; | |
613 | return false; | |
614 | } | |
615 | } | |
616 | return true; | |
617 | } | |
618 | ||
19660af7 | 619 | /* |
31e6b01f | 620 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
621 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
622 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
623 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
624 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
625 | * got stuck, so ref-walk may continue from there. If this is not successful | |
626 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
627 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 628 | */ |
31e6b01f NP |
629 | |
630 | /** | |
19660af7 AV |
631 | * unlazy_walk - try to switch to ref-walk mode. |
632 | * @nd: nameidata pathwalk data | |
633 | * @dentry: child of nd->path.dentry or NULL | |
6e9918b7 | 634 | * @seq: seq number to check dentry against |
39191628 | 635 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 636 | * |
19660af7 AV |
637 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
638 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
639 | * @nd or NULL. Must be called from rcu-walk context. | |
7973387a AV |
640 | * Nothing should touch nameidata between unlazy_walk() failure and |
641 | * terminate_walk(). | |
31e6b01f | 642 | */ |
6e9918b7 | 643 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry, unsigned seq) |
31e6b01f NP |
644 | { |
645 | struct fs_struct *fs = current->fs; | |
646 | struct dentry *parent = nd->path.dentry; | |
647 | ||
648 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
e5c832d5 | 649 | |
e5c832d5 | 650 | nd->flags &= ~LOOKUP_RCU; |
7973387a AV |
651 | if (unlikely(!legitimize_links(nd))) |
652 | goto out2; | |
653 | if (unlikely(!legitimize_mnt(nd->path.mnt, nd->m_seq))) | |
654 | goto out2; | |
655 | if (unlikely(!lockref_get_not_dead(&parent->d_lockref))) | |
656 | goto out1; | |
48a066e7 | 657 | |
15570086 LT |
658 | /* |
659 | * For a negative lookup, the lookup sequence point is the parents | |
660 | * sequence point, and it only needs to revalidate the parent dentry. | |
661 | * | |
662 | * For a positive lookup, we need to move both the parent and the | |
663 | * dentry from the RCU domain to be properly refcounted. And the | |
664 | * sequence number in the dentry validates *both* dentry counters, | |
665 | * since we checked the sequence number of the parent after we got | |
666 | * the child sequence number. So we know the parent must still | |
667 | * be valid if the child sequence number is still valid. | |
668 | */ | |
19660af7 | 669 | if (!dentry) { |
e5c832d5 LT |
670 | if (read_seqcount_retry(&parent->d_seq, nd->seq)) |
671 | goto out; | |
19660af7 AV |
672 | BUG_ON(nd->inode != parent->d_inode); |
673 | } else { | |
e5c832d5 LT |
674 | if (!lockref_get_not_dead(&dentry->d_lockref)) |
675 | goto out; | |
6e9918b7 | 676 | if (read_seqcount_retry(&dentry->d_seq, seq)) |
e5c832d5 | 677 | goto drop_dentry; |
19660af7 | 678 | } |
e5c832d5 LT |
679 | |
680 | /* | |
681 | * Sequence counts matched. Now make sure that the root is | |
682 | * still valid and get it if required. | |
683 | */ | |
684 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
685 | spin_lock(&fs->lock); | |
7973387a AV |
686 | if (unlikely(!path_equal(&nd->root, &fs->root))) { |
687 | spin_unlock(&fs->lock); | |
688 | goto drop_dentry; | |
689 | } | |
31e6b01f NP |
690 | path_get(&nd->root); |
691 | spin_unlock(&fs->lock); | |
692 | } | |
31e6b01f | 693 | |
8b61e74f | 694 | rcu_read_unlock(); |
31e6b01f | 695 | return 0; |
19660af7 | 696 | |
e5c832d5 | 697 | drop_dentry: |
8b61e74f | 698 | rcu_read_unlock(); |
15570086 | 699 | dput(dentry); |
d0d27277 | 700 | goto drop_root_mnt; |
7973387a AV |
701 | out2: |
702 | nd->path.mnt = NULL; | |
703 | out1: | |
704 | nd->path.dentry = NULL; | |
e5c832d5 | 705 | out: |
8b61e74f | 706 | rcu_read_unlock(); |
d0d27277 LT |
707 | drop_root_mnt: |
708 | if (!(nd->flags & LOOKUP_ROOT)) | |
709 | nd->root.mnt = NULL; | |
31e6b01f NP |
710 | return -ECHILD; |
711 | } | |
712 | ||
7973387a AV |
713 | static int unlazy_link(struct nameidata *nd, struct path *link, unsigned seq) |
714 | { | |
715 | if (unlikely(!legitimize_path(nd, link, seq))) { | |
716 | drop_links(nd); | |
717 | nd->depth = 0; | |
718 | nd->flags &= ~LOOKUP_RCU; | |
719 | nd->path.mnt = NULL; | |
720 | nd->path.dentry = NULL; | |
721 | if (!(nd->flags & LOOKUP_ROOT)) | |
722 | nd->root.mnt = NULL; | |
723 | rcu_read_unlock(); | |
724 | } else if (likely(unlazy_walk(nd, NULL, 0)) == 0) { | |
725 | return 0; | |
726 | } | |
727 | path_put(link); | |
728 | return -ECHILD; | |
729 | } | |
730 | ||
4ce16ef3 | 731 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 732 | { |
4ce16ef3 | 733 | return dentry->d_op->d_revalidate(dentry, flags); |
34286d66 NP |
734 | } |
735 | ||
9f1fafee AV |
736 | /** |
737 | * complete_walk - successful completion of path walk | |
738 | * @nd: pointer nameidata | |
39159de2 | 739 | * |
9f1fafee AV |
740 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
741 | * Revalidate the final result, unless we'd already done that during | |
742 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
743 | * success, -error on failure. In case of failure caller does not | |
744 | * need to drop nd->path. | |
39159de2 | 745 | */ |
9f1fafee | 746 | static int complete_walk(struct nameidata *nd) |
39159de2 | 747 | { |
16c2cd71 | 748 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 749 | int status; |
39159de2 | 750 | |
9f1fafee | 751 | if (nd->flags & LOOKUP_RCU) { |
9f1fafee AV |
752 | if (!(nd->flags & LOOKUP_ROOT)) |
753 | nd->root.mnt = NULL; | |
6e9918b7 | 754 | if (unlikely(unlazy_walk(nd, NULL, 0))) |
9f1fafee | 755 | return -ECHILD; |
9f1fafee AV |
756 | } |
757 | ||
16c2cd71 AV |
758 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
759 | return 0; | |
760 | ||
ecf3d1f1 | 761 | if (likely(!(dentry->d_flags & DCACHE_OP_WEAK_REVALIDATE))) |
39159de2 JL |
762 | return 0; |
763 | ||
ecf3d1f1 | 764 | status = dentry->d_op->d_weak_revalidate(dentry, nd->flags); |
39159de2 JL |
765 | if (status > 0) |
766 | return 0; | |
767 | ||
16c2cd71 | 768 | if (!status) |
39159de2 | 769 | status = -ESTALE; |
16c2cd71 | 770 | |
39159de2 JL |
771 | return status; |
772 | } | |
773 | ||
2a737871 AV |
774 | static __always_inline void set_root(struct nameidata *nd) |
775 | { | |
7bd88377 | 776 | get_fs_root(current->fs, &nd->root); |
2a737871 AV |
777 | } |
778 | ||
7bd88377 | 779 | static __always_inline unsigned set_root_rcu(struct nameidata *nd) |
31e6b01f | 780 | { |
7bd88377 AV |
781 | struct fs_struct *fs = current->fs; |
782 | unsigned seq, res; | |
c28cc364 | 783 | |
7bd88377 AV |
784 | do { |
785 | seq = read_seqcount_begin(&fs->seq); | |
786 | nd->root = fs->root; | |
787 | res = __read_seqcount_begin(&nd->root.dentry->d_seq); | |
788 | } while (read_seqcount_retry(&fs->seq, seq)); | |
789 | return res; | |
31e6b01f NP |
790 | } |
791 | ||
1d957f9b | 792 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
793 | { |
794 | dput(path->dentry); | |
4ac91378 | 795 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
796 | mntput(path->mnt); |
797 | } | |
798 | ||
7b9337aa NP |
799 | static inline void path_to_nameidata(const struct path *path, |
800 | struct nameidata *nd) | |
051d3812 | 801 | { |
31e6b01f NP |
802 | if (!(nd->flags & LOOKUP_RCU)) { |
803 | dput(nd->path.dentry); | |
804 | if (nd->path.mnt != path->mnt) | |
805 | mntput(nd->path.mnt); | |
9a229683 | 806 | } |
31e6b01f | 807 | nd->path.mnt = path->mnt; |
4ac91378 | 808 | nd->path.dentry = path->dentry; |
051d3812 IK |
809 | } |
810 | ||
b5fb63c1 CH |
811 | /* |
812 | * Helper to directly jump to a known parsed path from ->follow_link, | |
813 | * caller must have taken a reference to path beforehand. | |
814 | */ | |
6e77137b | 815 | void nd_jump_link(struct path *path) |
b5fb63c1 | 816 | { |
6e77137b | 817 | struct nameidata *nd = current->nameidata; |
b5fb63c1 CH |
818 | path_put(&nd->path); |
819 | ||
820 | nd->path = *path; | |
821 | nd->inode = nd->path.dentry->d_inode; | |
822 | nd->flags |= LOOKUP_JUMPED; | |
b5fb63c1 CH |
823 | } |
824 | ||
b9ff4429 | 825 | static inline void put_link(struct nameidata *nd) |
574197e0 | 826 | { |
21c3003d | 827 | struct saved *last = nd->stack + --nd->depth; |
237d8b32 | 828 | struct inode *inode = last->inode; |
b9ff4429 | 829 | if (last->cookie && inode->i_op->put_link) |
5f2c4179 | 830 | inode->i_op->put_link(inode, last->cookie); |
6548fae2 AV |
831 | if (!(nd->flags & LOOKUP_RCU)) |
832 | path_put(&last->link); | |
574197e0 AV |
833 | } |
834 | ||
561ec64a LT |
835 | int sysctl_protected_symlinks __read_mostly = 0; |
836 | int sysctl_protected_hardlinks __read_mostly = 0; | |
800179c9 KC |
837 | |
838 | /** | |
839 | * may_follow_link - Check symlink following for unsafe situations | |
55852635 | 840 | * @nd: nameidata pathwalk data |
800179c9 KC |
841 | * |
842 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
843 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
844 | * in a sticky world-writable directory. This is to protect privileged | |
845 | * processes from failing races against path names that may change out | |
846 | * from under them by way of other users creating malicious symlinks. | |
847 | * It will permit symlinks to be followed only when outside a sticky | |
848 | * world-writable directory, or when the uid of the symlink and follower | |
849 | * match, or when the directory owner matches the symlink's owner. | |
850 | * | |
851 | * Returns 0 if following the symlink is allowed, -ve on error. | |
852 | */ | |
fec2fa24 | 853 | static inline int may_follow_link(struct nameidata *nd) |
800179c9 KC |
854 | { |
855 | const struct inode *inode; | |
856 | const struct inode *parent; | |
857 | ||
858 | if (!sysctl_protected_symlinks) | |
859 | return 0; | |
860 | ||
861 | /* Allowed if owner and follower match. */ | |
237d8b32 | 862 | inode = nd->stack[0].inode; |
81abe27b | 863 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
864 | return 0; |
865 | ||
866 | /* Allowed if parent directory not sticky and world-writable. */ | |
867 | parent = nd->path.dentry->d_inode; | |
868 | if ((parent->i_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) | |
869 | return 0; | |
870 | ||
871 | /* Allowed if parent directory and link owner match. */ | |
81abe27b | 872 | if (uid_eq(parent->i_uid, inode->i_uid)) |
800179c9 KC |
873 | return 0; |
874 | ||
31956502 AV |
875 | if (nd->flags & LOOKUP_RCU) |
876 | return -ECHILD; | |
877 | ||
1cf2665b | 878 | audit_log_link_denied("follow_link", &nd->stack[0].link); |
800179c9 KC |
879 | return -EACCES; |
880 | } | |
881 | ||
882 | /** | |
883 | * safe_hardlink_source - Check for safe hardlink conditions | |
884 | * @inode: the source inode to hardlink from | |
885 | * | |
886 | * Return false if at least one of the following conditions: | |
887 | * - inode is not a regular file | |
888 | * - inode is setuid | |
889 | * - inode is setgid and group-exec | |
890 | * - access failure for read and write | |
891 | * | |
892 | * Otherwise returns true. | |
893 | */ | |
894 | static bool safe_hardlink_source(struct inode *inode) | |
895 | { | |
896 | umode_t mode = inode->i_mode; | |
897 | ||
898 | /* Special files should not get pinned to the filesystem. */ | |
899 | if (!S_ISREG(mode)) | |
900 | return false; | |
901 | ||
902 | /* Setuid files should not get pinned to the filesystem. */ | |
903 | if (mode & S_ISUID) | |
904 | return false; | |
905 | ||
906 | /* Executable setgid files should not get pinned to the filesystem. */ | |
907 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
908 | return false; | |
909 | ||
910 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
911 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
912 | return false; | |
913 | ||
914 | return true; | |
915 | } | |
916 | ||
917 | /** | |
918 | * may_linkat - Check permissions for creating a hardlink | |
919 | * @link: the source to hardlink from | |
920 | * | |
921 | * Block hardlink when all of: | |
922 | * - sysctl_protected_hardlinks enabled | |
923 | * - fsuid does not match inode | |
924 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
925 | * - not CAP_FOWNER | |
926 | * | |
927 | * Returns 0 if successful, -ve on error. | |
928 | */ | |
929 | static int may_linkat(struct path *link) | |
930 | { | |
931 | const struct cred *cred; | |
932 | struct inode *inode; | |
933 | ||
934 | if (!sysctl_protected_hardlinks) | |
935 | return 0; | |
936 | ||
937 | cred = current_cred(); | |
938 | inode = link->dentry->d_inode; | |
939 | ||
940 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, | |
941 | * otherwise, it must be a safe source. | |
942 | */ | |
81abe27b | 943 | if (uid_eq(cred->fsuid, inode->i_uid) || safe_hardlink_source(inode) || |
800179c9 KC |
944 | capable(CAP_FOWNER)) |
945 | return 0; | |
946 | ||
a51d9eaa | 947 | audit_log_link_denied("linkat", link); |
800179c9 KC |
948 | return -EPERM; |
949 | } | |
950 | ||
3b2e7f75 AV |
951 | static __always_inline |
952 | const char *get_link(struct nameidata *nd) | |
1da177e4 | 953 | { |
ab104923 | 954 | struct saved *last = nd->stack + nd->depth - 1; |
1cf2665b | 955 | struct dentry *dentry = last->link.dentry; |
237d8b32 | 956 | struct inode *inode = last->inode; |
6d7b5aae | 957 | int error; |
0a959df5 | 958 | const char *res; |
1da177e4 | 959 | |
844a3917 AV |
960 | BUG_ON(nd->flags & LOOKUP_RCU); |
961 | ||
574197e0 | 962 | cond_resched(); |
574197e0 | 963 | |
3b2e7f75 | 964 | touch_atime(&last->link); |
cd4e91d3 | 965 | |
bda0be7a N |
966 | error = security_inode_follow_link(dentry, inode, |
967 | nd->flags & LOOKUP_RCU); | |
968 | if (unlikely(error)) | |
6920a440 | 969 | return ERR_PTR(error); |
36f3b4f6 | 970 | |
86acdca1 | 971 | nd->last_type = LAST_BIND; |
d4dee48b AV |
972 | res = inode->i_link; |
973 | if (!res) { | |
6e77137b | 974 | res = inode->i_op->follow_link(dentry, &last->cookie); |
fab51e8a | 975 | if (IS_ERR_OR_NULL(res)) { |
6920a440 | 976 | last->cookie = NULL; |
fab51e8a AV |
977 | return res; |
978 | } | |
979 | } | |
980 | if (*res == '/') { | |
981 | if (!nd->root.mnt) | |
982 | set_root(nd); | |
983 | path_put(&nd->path); | |
984 | nd->path = nd->root; | |
985 | path_get(&nd->root); | |
986 | nd->inode = nd->path.dentry->d_inode; | |
987 | nd->flags |= LOOKUP_JUMPED; | |
988 | while (unlikely(*++res == '/')) | |
989 | ; | |
1da177e4 | 990 | } |
fab51e8a AV |
991 | if (!*res) |
992 | res = NULL; | |
0a959df5 AV |
993 | return res; |
994 | } | |
6d7b5aae | 995 | |
31e6b01f NP |
996 | static int follow_up_rcu(struct path *path) |
997 | { | |
0714a533 AV |
998 | struct mount *mnt = real_mount(path->mnt); |
999 | struct mount *parent; | |
31e6b01f NP |
1000 | struct dentry *mountpoint; |
1001 | ||
0714a533 AV |
1002 | parent = mnt->mnt_parent; |
1003 | if (&parent->mnt == path->mnt) | |
31e6b01f | 1004 | return 0; |
a73324da | 1005 | mountpoint = mnt->mnt_mountpoint; |
31e6b01f | 1006 | path->dentry = mountpoint; |
0714a533 | 1007 | path->mnt = &parent->mnt; |
31e6b01f NP |
1008 | return 1; |
1009 | } | |
1010 | ||
f015f126 DH |
1011 | /* |
1012 | * follow_up - Find the mountpoint of path's vfsmount | |
1013 | * | |
1014 | * Given a path, find the mountpoint of its source file system. | |
1015 | * Replace @path with the path of the mountpoint in the parent mount. | |
1016 | * Up is towards /. | |
1017 | * | |
1018 | * Return 1 if we went up a level and 0 if we were already at the | |
1019 | * root. | |
1020 | */ | |
bab77ebf | 1021 | int follow_up(struct path *path) |
1da177e4 | 1022 | { |
0714a533 AV |
1023 | struct mount *mnt = real_mount(path->mnt); |
1024 | struct mount *parent; | |
1da177e4 | 1025 | struct dentry *mountpoint; |
99b7db7b | 1026 | |
48a066e7 | 1027 | read_seqlock_excl(&mount_lock); |
0714a533 | 1028 | parent = mnt->mnt_parent; |
3c0a6163 | 1029 | if (parent == mnt) { |
48a066e7 | 1030 | read_sequnlock_excl(&mount_lock); |
1da177e4 LT |
1031 | return 0; |
1032 | } | |
0714a533 | 1033 | mntget(&parent->mnt); |
a73324da | 1034 | mountpoint = dget(mnt->mnt_mountpoint); |
48a066e7 | 1035 | read_sequnlock_excl(&mount_lock); |
bab77ebf AV |
1036 | dput(path->dentry); |
1037 | path->dentry = mountpoint; | |
1038 | mntput(path->mnt); | |
0714a533 | 1039 | path->mnt = &parent->mnt; |
1da177e4 LT |
1040 | return 1; |
1041 | } | |
4d359507 | 1042 | EXPORT_SYMBOL(follow_up); |
1da177e4 | 1043 | |
b5c84bf6 | 1044 | /* |
9875cf80 DH |
1045 | * Perform an automount |
1046 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
1047 | * were called with. | |
1da177e4 | 1048 | */ |
756daf26 | 1049 | static int follow_automount(struct path *path, struct nameidata *nd, |
9875cf80 | 1050 | bool *need_mntput) |
31e6b01f | 1051 | { |
9875cf80 | 1052 | struct vfsmount *mnt; |
ea5b778a | 1053 | int err; |
9875cf80 DH |
1054 | |
1055 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
1056 | return -EREMOTE; | |
1057 | ||
0ec26fd0 MS |
1058 | /* We don't want to mount if someone's just doing a stat - |
1059 | * unless they're stat'ing a directory and appended a '/' to | |
1060 | * the name. | |
1061 | * | |
1062 | * We do, however, want to mount if someone wants to open or | |
1063 | * create a file of any type under the mountpoint, wants to | |
1064 | * traverse through the mountpoint or wants to open the | |
1065 | * mounted directory. Also, autofs may mark negative dentries | |
1066 | * as being automount points. These will need the attentions | |
1067 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 1068 | */ |
756daf26 N |
1069 | if (!(nd->flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
1070 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && | |
0ec26fd0 MS |
1071 | path->dentry->d_inode) |
1072 | return -EISDIR; | |
1073 | ||
756daf26 N |
1074 | nd->total_link_count++; |
1075 | if (nd->total_link_count >= 40) | |
9875cf80 DH |
1076 | return -ELOOP; |
1077 | ||
1078 | mnt = path->dentry->d_op->d_automount(path); | |
1079 | if (IS_ERR(mnt)) { | |
1080 | /* | |
1081 | * The filesystem is allowed to return -EISDIR here to indicate | |
1082 | * it doesn't want to automount. For instance, autofs would do | |
1083 | * this so that its userspace daemon can mount on this dentry. | |
1084 | * | |
1085 | * However, we can only permit this if it's a terminal point in | |
1086 | * the path being looked up; if it wasn't then the remainder of | |
1087 | * the path is inaccessible and we should say so. | |
1088 | */ | |
756daf26 | 1089 | if (PTR_ERR(mnt) == -EISDIR && (nd->flags & LOOKUP_PARENT)) |
9875cf80 DH |
1090 | return -EREMOTE; |
1091 | return PTR_ERR(mnt); | |
31e6b01f | 1092 | } |
ea5b778a | 1093 | |
9875cf80 DH |
1094 | if (!mnt) /* mount collision */ |
1095 | return 0; | |
31e6b01f | 1096 | |
8aef1884 AV |
1097 | if (!*need_mntput) { |
1098 | /* lock_mount() may release path->mnt on error */ | |
1099 | mntget(path->mnt); | |
1100 | *need_mntput = true; | |
1101 | } | |
19a167af | 1102 | err = finish_automount(mnt, path); |
9875cf80 | 1103 | |
ea5b778a DH |
1104 | switch (err) { |
1105 | case -EBUSY: | |
1106 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 1107 | return 0; |
ea5b778a | 1108 | case 0: |
8aef1884 | 1109 | path_put(path); |
ea5b778a DH |
1110 | path->mnt = mnt; |
1111 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 1112 | return 0; |
19a167af AV |
1113 | default: |
1114 | return err; | |
ea5b778a | 1115 | } |
19a167af | 1116 | |
463ffb2e AV |
1117 | } |
1118 | ||
9875cf80 DH |
1119 | /* |
1120 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 1121 | * - Flagged for transit management (autofs) |
9875cf80 DH |
1122 | * - Flagged as mountpoint |
1123 | * - Flagged as automount point | |
1124 | * | |
1125 | * This may only be called in refwalk mode. | |
1126 | * | |
1127 | * Serialization is taken care of in namespace.c | |
1128 | */ | |
756daf26 | 1129 | static int follow_managed(struct path *path, struct nameidata *nd) |
1da177e4 | 1130 | { |
8aef1884 | 1131 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
1132 | unsigned managed; |
1133 | bool need_mntput = false; | |
8aef1884 | 1134 | int ret = 0; |
9875cf80 DH |
1135 | |
1136 | /* Given that we're not holding a lock here, we retain the value in a | |
1137 | * local variable for each dentry as we look at it so that we don't see | |
1138 | * the components of that value change under us */ | |
1139 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1140 | managed &= DCACHE_MANAGED_DENTRY, | |
1141 | unlikely(managed != 0)) { | |
cc53ce53 DH |
1142 | /* Allow the filesystem to manage the transit without i_mutex |
1143 | * being held. */ | |
1144 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1145 | BUG_ON(!path->dentry->d_op); | |
1146 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 1147 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 1148 | if (ret < 0) |
8aef1884 | 1149 | break; |
cc53ce53 DH |
1150 | } |
1151 | ||
9875cf80 DH |
1152 | /* Transit to a mounted filesystem. */ |
1153 | if (managed & DCACHE_MOUNTED) { | |
1154 | struct vfsmount *mounted = lookup_mnt(path); | |
1155 | if (mounted) { | |
1156 | dput(path->dentry); | |
1157 | if (need_mntput) | |
1158 | mntput(path->mnt); | |
1159 | path->mnt = mounted; | |
1160 | path->dentry = dget(mounted->mnt_root); | |
1161 | need_mntput = true; | |
1162 | continue; | |
1163 | } | |
1164 | ||
1165 | /* Something is mounted on this dentry in another | |
1166 | * namespace and/or whatever was mounted there in this | |
48a066e7 AV |
1167 | * namespace got unmounted before lookup_mnt() could |
1168 | * get it */ | |
9875cf80 DH |
1169 | } |
1170 | ||
1171 | /* Handle an automount point */ | |
1172 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
756daf26 | 1173 | ret = follow_automount(path, nd, &need_mntput); |
9875cf80 | 1174 | if (ret < 0) |
8aef1884 | 1175 | break; |
9875cf80 DH |
1176 | continue; |
1177 | } | |
1178 | ||
1179 | /* We didn't change the current path point */ | |
1180 | break; | |
1da177e4 | 1181 | } |
8aef1884 AV |
1182 | |
1183 | if (need_mntput && path->mnt == mnt) | |
1184 | mntput(path->mnt); | |
1185 | if (ret == -EISDIR) | |
1186 | ret = 0; | |
8402752e AV |
1187 | if (need_mntput) |
1188 | nd->flags |= LOOKUP_JUMPED; | |
1189 | if (unlikely(ret < 0)) | |
1190 | path_put_conditional(path, nd); | |
1191 | return ret; | |
1da177e4 LT |
1192 | } |
1193 | ||
cc53ce53 | 1194 | int follow_down_one(struct path *path) |
1da177e4 LT |
1195 | { |
1196 | struct vfsmount *mounted; | |
1197 | ||
1c755af4 | 1198 | mounted = lookup_mnt(path); |
1da177e4 | 1199 | if (mounted) { |
9393bd07 AV |
1200 | dput(path->dentry); |
1201 | mntput(path->mnt); | |
1202 | path->mnt = mounted; | |
1203 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1204 | return 1; |
1205 | } | |
1206 | return 0; | |
1207 | } | |
4d359507 | 1208 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 | 1209 | |
b8faf035 | 1210 | static inline int managed_dentry_rcu(struct dentry *dentry) |
62a7375e | 1211 | { |
b8faf035 N |
1212 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT) ? |
1213 | dentry->d_op->d_manage(dentry, true) : 0; | |
62a7375e IK |
1214 | } |
1215 | ||
9875cf80 | 1216 | /* |
287548e4 AV |
1217 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1218 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1219 | */ |
1220 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
254cf582 | 1221 | struct inode **inode, unsigned *seqp) |
9875cf80 | 1222 | { |
62a7375e | 1223 | for (;;) { |
c7105365 | 1224 | struct mount *mounted; |
62a7375e IK |
1225 | /* |
1226 | * Don't forget we might have a non-mountpoint managed dentry | |
1227 | * that wants to block transit. | |
1228 | */ | |
b8faf035 N |
1229 | switch (managed_dentry_rcu(path->dentry)) { |
1230 | case -ECHILD: | |
1231 | default: | |
ab90911f | 1232 | return false; |
b8faf035 N |
1233 | case -EISDIR: |
1234 | return true; | |
1235 | case 0: | |
1236 | break; | |
1237 | } | |
62a7375e IK |
1238 | |
1239 | if (!d_mountpoint(path->dentry)) | |
b8faf035 | 1240 | return !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
62a7375e | 1241 | |
474279dc | 1242 | mounted = __lookup_mnt(path->mnt, path->dentry); |
9875cf80 DH |
1243 | if (!mounted) |
1244 | break; | |
c7105365 AV |
1245 | path->mnt = &mounted->mnt; |
1246 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1247 | nd->flags |= LOOKUP_JUMPED; |
254cf582 | 1248 | *seqp = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1249 | /* |
1250 | * Update the inode too. We don't need to re-check the | |
1251 | * dentry sequence number here after this d_inode read, | |
1252 | * because a mount-point is always pinned. | |
1253 | */ | |
1254 | *inode = path->dentry->d_inode; | |
9875cf80 | 1255 | } |
f5be3e29 | 1256 | return !read_seqretry(&mount_lock, nd->m_seq) && |
b8faf035 | 1257 | !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
287548e4 AV |
1258 | } |
1259 | ||
31e6b01f NP |
1260 | static int follow_dotdot_rcu(struct nameidata *nd) |
1261 | { | |
4023bfc9 | 1262 | struct inode *inode = nd->inode; |
7bd88377 AV |
1263 | if (!nd->root.mnt) |
1264 | set_root_rcu(nd); | |
31e6b01f | 1265 | |
9875cf80 | 1266 | while (1) { |
31e6b01f NP |
1267 | if (nd->path.dentry == nd->root.dentry && |
1268 | nd->path.mnt == nd->root.mnt) { | |
1269 | break; | |
1270 | } | |
1271 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
1272 | struct dentry *old = nd->path.dentry; | |
1273 | struct dentry *parent = old->d_parent; | |
1274 | unsigned seq; | |
1275 | ||
4023bfc9 | 1276 | inode = parent->d_inode; |
31e6b01f NP |
1277 | seq = read_seqcount_begin(&parent->d_seq); |
1278 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 1279 | goto failed; |
31e6b01f NP |
1280 | nd->path.dentry = parent; |
1281 | nd->seq = seq; | |
1282 | break; | |
1283 | } | |
1284 | if (!follow_up_rcu(&nd->path)) | |
1285 | break; | |
4023bfc9 | 1286 | inode = nd->path.dentry->d_inode; |
31e6b01f | 1287 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
31e6b01f | 1288 | } |
b37199e6 AV |
1289 | while (d_mountpoint(nd->path.dentry)) { |
1290 | struct mount *mounted; | |
1291 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry); | |
1292 | if (!mounted) | |
1293 | break; | |
1294 | nd->path.mnt = &mounted->mnt; | |
1295 | nd->path.dentry = mounted->mnt.mnt_root; | |
4023bfc9 | 1296 | inode = nd->path.dentry->d_inode; |
b37199e6 | 1297 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
f5be3e29 | 1298 | if (read_seqretry(&mount_lock, nd->m_seq)) |
b37199e6 AV |
1299 | goto failed; |
1300 | } | |
4023bfc9 | 1301 | nd->inode = inode; |
31e6b01f | 1302 | return 0; |
ef7562d5 AV |
1303 | |
1304 | failed: | |
ef7562d5 | 1305 | return -ECHILD; |
31e6b01f NP |
1306 | } |
1307 | ||
cc53ce53 DH |
1308 | /* |
1309 | * Follow down to the covering mount currently visible to userspace. At each | |
1310 | * point, the filesystem owning that dentry may be queried as to whether the | |
1311 | * caller is permitted to proceed or not. | |
cc53ce53 | 1312 | */ |
7cc90cc3 | 1313 | int follow_down(struct path *path) |
cc53ce53 DH |
1314 | { |
1315 | unsigned managed; | |
1316 | int ret; | |
1317 | ||
1318 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1319 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1320 | /* Allow the filesystem to manage the transit without i_mutex | |
1321 | * being held. | |
1322 | * | |
1323 | * We indicate to the filesystem if someone is trying to mount | |
1324 | * something here. This gives autofs the chance to deny anyone | |
1325 | * other than its daemon the right to mount on its | |
1326 | * superstructure. | |
1327 | * | |
1328 | * The filesystem may sleep at this point. | |
1329 | */ | |
1330 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1331 | BUG_ON(!path->dentry->d_op); | |
1332 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 1333 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 1334 | path->dentry, false); |
cc53ce53 DH |
1335 | if (ret < 0) |
1336 | return ret == -EISDIR ? 0 : ret; | |
1337 | } | |
1338 | ||
1339 | /* Transit to a mounted filesystem. */ | |
1340 | if (managed & DCACHE_MOUNTED) { | |
1341 | struct vfsmount *mounted = lookup_mnt(path); | |
1342 | if (!mounted) | |
1343 | break; | |
1344 | dput(path->dentry); | |
1345 | mntput(path->mnt); | |
1346 | path->mnt = mounted; | |
1347 | path->dentry = dget(mounted->mnt_root); | |
1348 | continue; | |
1349 | } | |
1350 | ||
1351 | /* Don't handle automount points here */ | |
1352 | break; | |
1353 | } | |
1354 | return 0; | |
1355 | } | |
4d359507 | 1356 | EXPORT_SYMBOL(follow_down); |
cc53ce53 | 1357 | |
9875cf80 DH |
1358 | /* |
1359 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1360 | */ | |
1361 | static void follow_mount(struct path *path) | |
1362 | { | |
1363 | while (d_mountpoint(path->dentry)) { | |
1364 | struct vfsmount *mounted = lookup_mnt(path); | |
1365 | if (!mounted) | |
1366 | break; | |
1367 | dput(path->dentry); | |
1368 | mntput(path->mnt); | |
1369 | path->mnt = mounted; | |
1370 | path->dentry = dget(mounted->mnt_root); | |
1371 | } | |
1372 | } | |
1373 | ||
31e6b01f | 1374 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1375 | { |
7bd88377 AV |
1376 | if (!nd->root.mnt) |
1377 | set_root(nd); | |
e518ddb7 | 1378 | |
1da177e4 | 1379 | while(1) { |
4ac91378 | 1380 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1381 | |
2a737871 AV |
1382 | if (nd->path.dentry == nd->root.dentry && |
1383 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1384 | break; |
1385 | } | |
4ac91378 | 1386 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1387 | /* rare case of legitimate dget_parent()... */ |
1388 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1389 | dput(old); |
1390 | break; | |
1391 | } | |
3088dd70 | 1392 | if (!follow_up(&nd->path)) |
1da177e4 | 1393 | break; |
1da177e4 | 1394 | } |
79ed0226 | 1395 | follow_mount(&nd->path); |
31e6b01f | 1396 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1397 | } |
1398 | ||
baa03890 | 1399 | /* |
bad61189 MS |
1400 | * This looks up the name in dcache, possibly revalidates the old dentry and |
1401 | * allocates a new one if not found or not valid. In the need_lookup argument | |
1402 | * returns whether i_op->lookup is necessary. | |
1403 | * | |
1404 | * dir->d_inode->i_mutex must be held | |
baa03890 | 1405 | */ |
bad61189 | 1406 | static struct dentry *lookup_dcache(struct qstr *name, struct dentry *dir, |
201f956e | 1407 | unsigned int flags, bool *need_lookup) |
baa03890 | 1408 | { |
baa03890 | 1409 | struct dentry *dentry; |
bad61189 | 1410 | int error; |
baa03890 | 1411 | |
bad61189 MS |
1412 | *need_lookup = false; |
1413 | dentry = d_lookup(dir, name); | |
1414 | if (dentry) { | |
39e3c955 | 1415 | if (dentry->d_flags & DCACHE_OP_REVALIDATE) { |
201f956e | 1416 | error = d_revalidate(dentry, flags); |
bad61189 MS |
1417 | if (unlikely(error <= 0)) { |
1418 | if (error < 0) { | |
1419 | dput(dentry); | |
1420 | return ERR_PTR(error); | |
5542aa2f EB |
1421 | } else { |
1422 | d_invalidate(dentry); | |
bad61189 MS |
1423 | dput(dentry); |
1424 | dentry = NULL; | |
1425 | } | |
1426 | } | |
1427 | } | |
1428 | } | |
baa03890 | 1429 | |
bad61189 MS |
1430 | if (!dentry) { |
1431 | dentry = d_alloc(dir, name); | |
1432 | if (unlikely(!dentry)) | |
1433 | return ERR_PTR(-ENOMEM); | |
baa03890 | 1434 | |
bad61189 | 1435 | *need_lookup = true; |
baa03890 NP |
1436 | } |
1437 | return dentry; | |
1438 | } | |
1439 | ||
44396f4b | 1440 | /* |
13a2c3be BF |
1441 | * Call i_op->lookup on the dentry. The dentry must be negative and |
1442 | * unhashed. | |
bad61189 MS |
1443 | * |
1444 | * dir->d_inode->i_mutex must be held | |
44396f4b | 1445 | */ |
bad61189 | 1446 | static struct dentry *lookup_real(struct inode *dir, struct dentry *dentry, |
72bd866a | 1447 | unsigned int flags) |
44396f4b | 1448 | { |
44396f4b JB |
1449 | struct dentry *old; |
1450 | ||
1451 | /* Don't create child dentry for a dead directory. */ | |
bad61189 | 1452 | if (unlikely(IS_DEADDIR(dir))) { |
e188dc02 | 1453 | dput(dentry); |
44396f4b | 1454 | return ERR_PTR(-ENOENT); |
e188dc02 | 1455 | } |
44396f4b | 1456 | |
72bd866a | 1457 | old = dir->i_op->lookup(dir, dentry, flags); |
44396f4b JB |
1458 | if (unlikely(old)) { |
1459 | dput(dentry); | |
1460 | dentry = old; | |
1461 | } | |
1462 | return dentry; | |
1463 | } | |
1464 | ||
a3255546 | 1465 | static struct dentry *__lookup_hash(struct qstr *name, |
72bd866a | 1466 | struct dentry *base, unsigned int flags) |
a3255546 | 1467 | { |
bad61189 | 1468 | bool need_lookup; |
a3255546 AV |
1469 | struct dentry *dentry; |
1470 | ||
72bd866a | 1471 | dentry = lookup_dcache(name, base, flags, &need_lookup); |
bad61189 MS |
1472 | if (!need_lookup) |
1473 | return dentry; | |
a3255546 | 1474 | |
72bd866a | 1475 | return lookup_real(base->d_inode, dentry, flags); |
a3255546 AV |
1476 | } |
1477 | ||
1da177e4 LT |
1478 | /* |
1479 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1480 | * small and for now I'd prefer to have fast path as straight as possible. | |
1481 | * It _is_ time-critical. | |
1482 | */ | |
e97cdc87 | 1483 | static int lookup_fast(struct nameidata *nd, |
254cf582 AV |
1484 | struct path *path, struct inode **inode, |
1485 | unsigned *seqp) | |
1da177e4 | 1486 | { |
4ac91378 | 1487 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1488 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1489 | int need_reval = 1; |
1490 | int status = 1; | |
9875cf80 DH |
1491 | int err; |
1492 | ||
b04f784e NP |
1493 | /* |
1494 | * Rename seqlock is not required here because in the off chance | |
1495 | * of a false negative due to a concurrent rename, we're going to | |
1496 | * do the non-racy lookup, below. | |
1497 | */ | |
31e6b01f NP |
1498 | if (nd->flags & LOOKUP_RCU) { |
1499 | unsigned seq; | |
766c4cbf | 1500 | bool negative; |
da53be12 | 1501 | dentry = __d_lookup_rcu(parent, &nd->last, &seq); |
5a18fff2 AV |
1502 | if (!dentry) |
1503 | goto unlazy; | |
1504 | ||
12f8ad4b LT |
1505 | /* |
1506 | * This sequence count validates that the inode matches | |
1507 | * the dentry name information from lookup. | |
1508 | */ | |
63afdfc7 | 1509 | *inode = d_backing_inode(dentry); |
766c4cbf | 1510 | negative = d_is_negative(dentry); |
12f8ad4b LT |
1511 | if (read_seqcount_retry(&dentry->d_seq, seq)) |
1512 | return -ECHILD; | |
766c4cbf AV |
1513 | if (negative) |
1514 | return -ENOENT; | |
12f8ad4b LT |
1515 | |
1516 | /* | |
1517 | * This sequence count validates that the parent had no | |
1518 | * changes while we did the lookup of the dentry above. | |
1519 | * | |
1520 | * The memory barrier in read_seqcount_begin of child is | |
1521 | * enough, we can use __read_seqcount_retry here. | |
1522 | */ | |
31e6b01f NP |
1523 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) |
1524 | return -ECHILD; | |
5a18fff2 | 1525 | |
254cf582 | 1526 | *seqp = seq; |
24643087 | 1527 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
4ce16ef3 | 1528 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1529 | if (unlikely(status <= 0)) { |
1530 | if (status != -ECHILD) | |
1531 | need_reval = 0; | |
1532 | goto unlazy; | |
1533 | } | |
24643087 | 1534 | } |
31e6b01f NP |
1535 | path->mnt = mnt; |
1536 | path->dentry = dentry; | |
254cf582 | 1537 | if (likely(__follow_mount_rcu(nd, path, inode, seqp))) |
b8faf035 | 1538 | return 0; |
5a18fff2 | 1539 | unlazy: |
254cf582 | 1540 | if (unlazy_walk(nd, dentry, seq)) |
19660af7 | 1541 | return -ECHILD; |
5a18fff2 | 1542 | } else { |
e97cdc87 | 1543 | dentry = __d_lookup(parent, &nd->last); |
9875cf80 | 1544 | } |
5a18fff2 | 1545 | |
81e6f520 AV |
1546 | if (unlikely(!dentry)) |
1547 | goto need_lookup; | |
1548 | ||
5a18fff2 | 1549 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) |
4ce16ef3 | 1550 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1551 | if (unlikely(status <= 0)) { |
1552 | if (status < 0) { | |
1553 | dput(dentry); | |
1554 | return status; | |
1555 | } | |
5542aa2f EB |
1556 | d_invalidate(dentry); |
1557 | dput(dentry); | |
1558 | goto need_lookup; | |
24643087 | 1559 | } |
697f514d | 1560 | |
766c4cbf AV |
1561 | if (unlikely(d_is_negative(dentry))) { |
1562 | dput(dentry); | |
1563 | return -ENOENT; | |
1564 | } | |
9875cf80 DH |
1565 | path->mnt = mnt; |
1566 | path->dentry = dentry; | |
756daf26 | 1567 | err = follow_managed(path, nd); |
8402752e | 1568 | if (likely(!err)) |
63afdfc7 | 1569 | *inode = d_backing_inode(path->dentry); |
8402752e | 1570 | return err; |
81e6f520 AV |
1571 | |
1572 | need_lookup: | |
697f514d MS |
1573 | return 1; |
1574 | } | |
1575 | ||
1576 | /* Fast lookup failed, do it the slow way */ | |
cc2a5271 | 1577 | static int lookup_slow(struct nameidata *nd, struct path *path) |
697f514d MS |
1578 | { |
1579 | struct dentry *dentry, *parent; | |
697f514d MS |
1580 | |
1581 | parent = nd->path.dentry; | |
81e6f520 AV |
1582 | BUG_ON(nd->inode != parent->d_inode); |
1583 | ||
1584 | mutex_lock(&parent->d_inode->i_mutex); | |
cc2a5271 | 1585 | dentry = __lookup_hash(&nd->last, parent, nd->flags); |
81e6f520 AV |
1586 | mutex_unlock(&parent->d_inode->i_mutex); |
1587 | if (IS_ERR(dentry)) | |
1588 | return PTR_ERR(dentry); | |
697f514d MS |
1589 | path->mnt = nd->path.mnt; |
1590 | path->dentry = dentry; | |
8402752e | 1591 | return follow_managed(path, nd); |
1da177e4 LT |
1592 | } |
1593 | ||
52094c8a AV |
1594 | static inline int may_lookup(struct nameidata *nd) |
1595 | { | |
1596 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1597 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1598 | if (err != -ECHILD) |
1599 | return err; | |
6e9918b7 | 1600 | if (unlazy_walk(nd, NULL, 0)) |
52094c8a AV |
1601 | return -ECHILD; |
1602 | } | |
4ad5abb3 | 1603 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1604 | } |
1605 | ||
9856fa1b AV |
1606 | static inline int handle_dots(struct nameidata *nd, int type) |
1607 | { | |
1608 | if (type == LAST_DOTDOT) { | |
1609 | if (nd->flags & LOOKUP_RCU) { | |
70291aec | 1610 | return follow_dotdot_rcu(nd); |
9856fa1b AV |
1611 | } else |
1612 | follow_dotdot(nd); | |
1613 | } | |
1614 | return 0; | |
1615 | } | |
1616 | ||
181548c0 AV |
1617 | static int pick_link(struct nameidata *nd, struct path *link, |
1618 | struct inode *inode, unsigned seq) | |
d63ff28f | 1619 | { |
626de996 | 1620 | int error; |
1cf2665b | 1621 | struct saved *last; |
756daf26 | 1622 | if (unlikely(nd->total_link_count++ >= MAXSYMLINKS)) { |
626de996 AV |
1623 | path_to_nameidata(link, nd); |
1624 | return -ELOOP; | |
1625 | } | |
d63ff28f | 1626 | if (nd->flags & LOOKUP_RCU) { |
7973387a | 1627 | if (unlikely(unlazy_link(nd, link, seq))) |
d63ff28f | 1628 | return -ECHILD; |
7973387a AV |
1629 | } else { |
1630 | if (link->mnt == nd->path.mnt) | |
1631 | mntget(link->mnt); | |
d63ff28f | 1632 | } |
626de996 AV |
1633 | error = nd_alloc_stack(nd); |
1634 | if (unlikely(error)) { | |
cd179f44 | 1635 | path_put(link); |
626de996 AV |
1636 | return error; |
1637 | } | |
1638 | ||
ab104923 | 1639 | last = nd->stack + nd->depth++; |
1cf2665b | 1640 | last->link = *link; |
ab104923 | 1641 | last->cookie = NULL; |
181548c0 | 1642 | last->inode = inode; |
0450b2d1 | 1643 | last->seq = seq; |
d63ff28f AV |
1644 | return 1; |
1645 | } | |
1646 | ||
3ddcd056 LT |
1647 | /* |
1648 | * Do we need to follow links? We _really_ want to be able | |
1649 | * to do this check without having to look at inode->i_op, | |
1650 | * so we keep a cache of "no, this doesn't need follow_link" | |
1651 | * for the common case. | |
1652 | */ | |
254cf582 | 1653 | static inline int should_follow_link(struct nameidata *nd, struct path *link, |
181548c0 AV |
1654 | int follow, |
1655 | struct inode *inode, unsigned seq) | |
3ddcd056 | 1656 | { |
d63ff28f AV |
1657 | if (likely(!d_is_symlink(link->dentry))) |
1658 | return 0; | |
1659 | if (!follow) | |
1660 | return 0; | |
181548c0 | 1661 | return pick_link(nd, link, inode, seq); |
3ddcd056 LT |
1662 | } |
1663 | ||
4693a547 AV |
1664 | enum {WALK_GET = 1, WALK_PUT = 2}; |
1665 | ||
1666 | static int walk_component(struct nameidata *nd, int flags) | |
ce57dfc1 | 1667 | { |
caa85634 | 1668 | struct path path; |
ce57dfc1 | 1669 | struct inode *inode; |
254cf582 | 1670 | unsigned seq; |
ce57dfc1 AV |
1671 | int err; |
1672 | /* | |
1673 | * "." and ".." are special - ".." especially so because it has | |
1674 | * to be able to know about the current root directory and | |
1675 | * parent relationships. | |
1676 | */ | |
4693a547 AV |
1677 | if (unlikely(nd->last_type != LAST_NORM)) { |
1678 | err = handle_dots(nd, nd->last_type); | |
1679 | if (flags & WALK_PUT) | |
1680 | put_link(nd); | |
1681 | return err; | |
1682 | } | |
254cf582 | 1683 | err = lookup_fast(nd, &path, &inode, &seq); |
ce57dfc1 | 1684 | if (unlikely(err)) { |
697f514d | 1685 | if (err < 0) |
f0a9ba70 | 1686 | return err; |
697f514d | 1687 | |
caa85634 | 1688 | err = lookup_slow(nd, &path); |
697f514d | 1689 | if (err < 0) |
f0a9ba70 | 1690 | return err; |
697f514d | 1691 | |
63afdfc7 | 1692 | inode = d_backing_inode(path.dentry); |
254cf582 | 1693 | seq = 0; /* we are already out of RCU mode */ |
766c4cbf | 1694 | err = -ENOENT; |
caa85634 | 1695 | if (d_is_negative(path.dentry)) |
766c4cbf | 1696 | goto out_path_put; |
ce57dfc1 | 1697 | } |
697f514d | 1698 | |
4693a547 AV |
1699 | if (flags & WALK_PUT) |
1700 | put_link(nd); | |
181548c0 | 1701 | err = should_follow_link(nd, &path, flags & WALK_GET, inode, seq); |
d63ff28f AV |
1702 | if (unlikely(err)) |
1703 | return err; | |
caa85634 | 1704 | path_to_nameidata(&path, nd); |
ce57dfc1 | 1705 | nd->inode = inode; |
254cf582 | 1706 | nd->seq = seq; |
ce57dfc1 | 1707 | return 0; |
697f514d MS |
1708 | |
1709 | out_path_put: | |
caa85634 | 1710 | path_to_nameidata(&path, nd); |
697f514d | 1711 | return err; |
ce57dfc1 AV |
1712 | } |
1713 | ||
bfcfaa77 LT |
1714 | /* |
1715 | * We can do the critical dentry name comparison and hashing | |
1716 | * operations one word at a time, but we are limited to: | |
1717 | * | |
1718 | * - Architectures with fast unaligned word accesses. We could | |
1719 | * do a "get_unaligned()" if this helps and is sufficiently | |
1720 | * fast. | |
1721 | * | |
bfcfaa77 LT |
1722 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we |
1723 | * do not trap on the (extremely unlikely) case of a page | |
1724 | * crossing operation. | |
1725 | * | |
1726 | * - Furthermore, we need an efficient 64-bit compile for the | |
1727 | * 64-bit case in order to generate the "number of bytes in | |
1728 | * the final mask". Again, that could be replaced with a | |
1729 | * efficient population count instruction or similar. | |
1730 | */ | |
1731 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1732 | ||
f68e556e | 1733 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1734 | |
f68e556e | 1735 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1736 | |
1737 | static inline unsigned int fold_hash(unsigned long hash) | |
1738 | { | |
99d263d4 | 1739 | return hash_64(hash, 32); |
bfcfaa77 LT |
1740 | } |
1741 | ||
1742 | #else /* 32-bit case */ | |
1743 | ||
bfcfaa77 LT |
1744 | #define fold_hash(x) (x) |
1745 | ||
1746 | #endif | |
1747 | ||
1748 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1749 | { | |
1750 | unsigned long a, mask; | |
1751 | unsigned long hash = 0; | |
1752 | ||
1753 | for (;;) { | |
e419b4cc | 1754 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1755 | if (len < sizeof(unsigned long)) |
1756 | break; | |
1757 | hash += a; | |
f132c5be | 1758 | hash *= 9; |
bfcfaa77 LT |
1759 | name += sizeof(unsigned long); |
1760 | len -= sizeof(unsigned long); | |
1761 | if (!len) | |
1762 | goto done; | |
1763 | } | |
a5c21dce | 1764 | mask = bytemask_from_count(len); |
bfcfaa77 LT |
1765 | hash += mask & a; |
1766 | done: | |
1767 | return fold_hash(hash); | |
1768 | } | |
1769 | EXPORT_SYMBOL(full_name_hash); | |
1770 | ||
bfcfaa77 LT |
1771 | /* |
1772 | * Calculate the length and hash of the path component, and | |
d6bb3e90 | 1773 | * return the "hash_len" as the result. |
bfcfaa77 | 1774 | */ |
d6bb3e90 | 1775 | static inline u64 hash_name(const char *name) |
bfcfaa77 | 1776 | { |
36126f8f LT |
1777 | unsigned long a, b, adata, bdata, mask, hash, len; |
1778 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; | |
bfcfaa77 LT |
1779 | |
1780 | hash = a = 0; | |
1781 | len = -sizeof(unsigned long); | |
1782 | do { | |
1783 | hash = (hash + a) * 9; | |
1784 | len += sizeof(unsigned long); | |
e419b4cc | 1785 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1786 | b = a ^ REPEAT_BYTE('/'); |
1787 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1788 | ||
1789 | adata = prep_zero_mask(a, adata, &constants); | |
1790 | bdata = prep_zero_mask(b, bdata, &constants); | |
1791 | ||
1792 | mask = create_zero_mask(adata | bdata); | |
1793 | ||
1794 | hash += a & zero_bytemask(mask); | |
9226b5b4 | 1795 | len += find_zero(mask); |
d6bb3e90 | 1796 | return hashlen_create(fold_hash(hash), len); |
bfcfaa77 LT |
1797 | } |
1798 | ||
1799 | #else | |
1800 | ||
0145acc2 LT |
1801 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1802 | { | |
1803 | unsigned long hash = init_name_hash(); | |
1804 | while (len--) | |
1805 | hash = partial_name_hash(*name++, hash); | |
1806 | return end_name_hash(hash); | |
1807 | } | |
ae942ae7 | 1808 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1809 | |
200e9ef7 LT |
1810 | /* |
1811 | * We know there's a real path component here of at least | |
1812 | * one character. | |
1813 | */ | |
d6bb3e90 | 1814 | static inline u64 hash_name(const char *name) |
200e9ef7 LT |
1815 | { |
1816 | unsigned long hash = init_name_hash(); | |
1817 | unsigned long len = 0, c; | |
1818 | ||
1819 | c = (unsigned char)*name; | |
1820 | do { | |
1821 | len++; | |
1822 | hash = partial_name_hash(c, hash); | |
1823 | c = (unsigned char)name[len]; | |
1824 | } while (c && c != '/'); | |
d6bb3e90 | 1825 | return hashlen_create(end_name_hash(hash), len); |
200e9ef7 LT |
1826 | } |
1827 | ||
bfcfaa77 LT |
1828 | #endif |
1829 | ||
1da177e4 LT |
1830 | /* |
1831 | * Name resolution. | |
ea3834d9 PM |
1832 | * This is the basic name resolution function, turning a pathname into |
1833 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1834 | * |
ea3834d9 PM |
1835 | * Returns 0 and nd will have valid dentry and mnt on success. |
1836 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1837 | */ |
6de88d72 | 1838 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 1839 | { |
1da177e4 | 1840 | int err; |
32cd7468 | 1841 | |
1da177e4 LT |
1842 | while (*name=='/') |
1843 | name++; | |
1844 | if (!*name) | |
9e18f10a | 1845 | return 0; |
1da177e4 | 1846 | |
1da177e4 LT |
1847 | /* At this point we know we have a real path component. */ |
1848 | for(;;) { | |
d6bb3e90 | 1849 | u64 hash_len; |
fe479a58 | 1850 | int type; |
1da177e4 | 1851 | |
52094c8a | 1852 | err = may_lookup(nd); |
1da177e4 | 1853 | if (err) |
3595e234 | 1854 | return err; |
1da177e4 | 1855 | |
d6bb3e90 | 1856 | hash_len = hash_name(name); |
1da177e4 | 1857 | |
fe479a58 | 1858 | type = LAST_NORM; |
d6bb3e90 | 1859 | if (name[0] == '.') switch (hashlen_len(hash_len)) { |
fe479a58 | 1860 | case 2: |
200e9ef7 | 1861 | if (name[1] == '.') { |
fe479a58 | 1862 | type = LAST_DOTDOT; |
16c2cd71 AV |
1863 | nd->flags |= LOOKUP_JUMPED; |
1864 | } | |
fe479a58 AV |
1865 | break; |
1866 | case 1: | |
1867 | type = LAST_DOT; | |
1868 | } | |
5a202bcd AV |
1869 | if (likely(type == LAST_NORM)) { |
1870 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1871 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd | 1872 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
a060dc50 | 1873 | struct qstr this = { { .hash_len = hash_len }, .name = name }; |
da53be12 | 1874 | err = parent->d_op->d_hash(parent, &this); |
5a202bcd | 1875 | if (err < 0) |
3595e234 | 1876 | return err; |
d6bb3e90 LT |
1877 | hash_len = this.hash_len; |
1878 | name = this.name; | |
5a202bcd AV |
1879 | } |
1880 | } | |
fe479a58 | 1881 | |
d6bb3e90 LT |
1882 | nd->last.hash_len = hash_len; |
1883 | nd->last.name = name; | |
5f4a6a69 AV |
1884 | nd->last_type = type; |
1885 | ||
d6bb3e90 LT |
1886 | name += hashlen_len(hash_len); |
1887 | if (!*name) | |
bdf6cbf1 | 1888 | goto OK; |
200e9ef7 LT |
1889 | /* |
1890 | * If it wasn't NUL, we know it was '/'. Skip that | |
1891 | * slash, and continue until no more slashes. | |
1892 | */ | |
1893 | do { | |
d6bb3e90 LT |
1894 | name++; |
1895 | } while (unlikely(*name == '/')); | |
8620c238 AV |
1896 | if (unlikely(!*name)) { |
1897 | OK: | |
368ee9ba | 1898 | /* pathname body, done */ |
8620c238 AV |
1899 | if (!nd->depth) |
1900 | return 0; | |
1901 | name = nd->stack[nd->depth - 1].name; | |
368ee9ba | 1902 | /* trailing symlink, done */ |
8620c238 AV |
1903 | if (!name) |
1904 | return 0; | |
1905 | /* last component of nested symlink */ | |
4693a547 | 1906 | err = walk_component(nd, WALK_GET | WALK_PUT); |
8620c238 | 1907 | } else { |
4693a547 | 1908 | err = walk_component(nd, WALK_GET); |
8620c238 | 1909 | } |
ce57dfc1 | 1910 | if (err < 0) |
3595e234 | 1911 | return err; |
1da177e4 | 1912 | |
ce57dfc1 | 1913 | if (err) { |
626de996 | 1914 | const char *s = get_link(nd); |
5a460275 | 1915 | |
3595e234 AV |
1916 | if (unlikely(IS_ERR(s))) |
1917 | return PTR_ERR(s); | |
d40bcc09 AV |
1918 | err = 0; |
1919 | if (unlikely(!s)) { | |
1920 | /* jumped */ | |
b9ff4429 | 1921 | put_link(nd); |
d40bcc09 | 1922 | } else { |
fab51e8a AV |
1923 | nd->stack[nd->depth - 1].name = name; |
1924 | name = s; | |
1925 | continue; | |
d40bcc09 | 1926 | } |
31e6b01f | 1927 | } |
3595e234 AV |
1928 | if (unlikely(!d_can_lookup(nd->path.dentry))) |
1929 | return -ENOTDIR; | |
1da177e4 | 1930 | } |
1da177e4 LT |
1931 | } |
1932 | ||
368ee9ba AV |
1933 | static const char *path_init(int dfd, const struct filename *name, |
1934 | unsigned int flags, struct nameidata *nd) | |
31e6b01f NP |
1935 | { |
1936 | int retval = 0; | |
fd2f7cb5 | 1937 | const char *s = name->name; |
31e6b01f NP |
1938 | |
1939 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
980f3ea2 | 1940 | nd->flags = flags | LOOKUP_JUMPED | LOOKUP_PARENT; |
31e6b01f | 1941 | nd->depth = 0; |
368ee9ba | 1942 | nd->total_link_count = 0; |
5b6ca027 | 1943 | if (flags & LOOKUP_ROOT) { |
b18825a7 DH |
1944 | struct dentry *root = nd->root.dentry; |
1945 | struct inode *inode = root->d_inode; | |
fd2f7cb5 | 1946 | if (*s) { |
44b1d530 | 1947 | if (!d_can_lookup(root)) |
368ee9ba | 1948 | return ERR_PTR(-ENOTDIR); |
73d049a4 AV |
1949 | retval = inode_permission(inode, MAY_EXEC); |
1950 | if (retval) | |
368ee9ba | 1951 | return ERR_PTR(retval); |
73d049a4 | 1952 | } |
5b6ca027 AV |
1953 | nd->path = nd->root; |
1954 | nd->inode = inode; | |
1955 | if (flags & LOOKUP_RCU) { | |
8b61e74f | 1956 | rcu_read_lock(); |
5b6ca027 | 1957 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
48a066e7 | 1958 | nd->m_seq = read_seqbegin(&mount_lock); |
5b6ca027 AV |
1959 | } else { |
1960 | path_get(&nd->path); | |
1961 | } | |
368ee9ba | 1962 | return s; |
5b6ca027 AV |
1963 | } |
1964 | ||
31e6b01f | 1965 | nd->root.mnt = NULL; |
31e6b01f | 1966 | |
48a066e7 | 1967 | nd->m_seq = read_seqbegin(&mount_lock); |
fd2f7cb5 | 1968 | if (*s == '/') { |
e41f7d4e | 1969 | if (flags & LOOKUP_RCU) { |
8b61e74f | 1970 | rcu_read_lock(); |
7bd88377 | 1971 | nd->seq = set_root_rcu(nd); |
e41f7d4e AV |
1972 | } else { |
1973 | set_root(nd); | |
1974 | path_get(&nd->root); | |
1975 | } | |
1976 | nd->path = nd->root; | |
31e6b01f | 1977 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1978 | if (flags & LOOKUP_RCU) { |
1979 | struct fs_struct *fs = current->fs; | |
1980 | unsigned seq; | |
31e6b01f | 1981 | |
8b61e74f | 1982 | rcu_read_lock(); |
c28cc364 | 1983 | |
e41f7d4e AV |
1984 | do { |
1985 | seq = read_seqcount_begin(&fs->seq); | |
1986 | nd->path = fs->pwd; | |
1987 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1988 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1989 | } else { | |
1990 | get_fs_pwd(current->fs, &nd->path); | |
1991 | } | |
31e6b01f | 1992 | } else { |
582aa64a | 1993 | /* Caller must check execute permissions on the starting path component */ |
2903ff01 | 1994 | struct fd f = fdget_raw(dfd); |
31e6b01f NP |
1995 | struct dentry *dentry; |
1996 | ||
2903ff01 | 1997 | if (!f.file) |
368ee9ba | 1998 | return ERR_PTR(-EBADF); |
31e6b01f | 1999 | |
2903ff01 | 2000 | dentry = f.file->f_path.dentry; |
31e6b01f | 2001 | |
fd2f7cb5 | 2002 | if (*s) { |
44b1d530 | 2003 | if (!d_can_lookup(dentry)) { |
2903ff01 | 2004 | fdput(f); |
368ee9ba | 2005 | return ERR_PTR(-ENOTDIR); |
2903ff01 | 2006 | } |
f52e0c11 | 2007 | } |
31e6b01f | 2008 | |
2903ff01 | 2009 | nd->path = f.file->f_path; |
e41f7d4e | 2010 | if (flags & LOOKUP_RCU) { |
8b61e74f | 2011 | rcu_read_lock(); |
34a26b99 AV |
2012 | nd->inode = nd->path.dentry->d_inode; |
2013 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
e41f7d4e | 2014 | } else { |
2903ff01 | 2015 | path_get(&nd->path); |
34a26b99 | 2016 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 2017 | } |
34a26b99 | 2018 | fdput(f); |
368ee9ba | 2019 | return s; |
31e6b01f | 2020 | } |
31e6b01f | 2021 | |
31e6b01f | 2022 | nd->inode = nd->path.dentry->d_inode; |
4023bfc9 | 2023 | if (!(flags & LOOKUP_RCU)) |
368ee9ba | 2024 | return s; |
4023bfc9 | 2025 | if (likely(!read_seqcount_retry(&nd->path.dentry->d_seq, nd->seq))) |
368ee9ba | 2026 | return s; |
4023bfc9 AV |
2027 | if (!(nd->flags & LOOKUP_ROOT)) |
2028 | nd->root.mnt = NULL; | |
2029 | rcu_read_unlock(); | |
368ee9ba | 2030 | return ERR_PTR(-ECHILD); |
9b4a9b14 AV |
2031 | } |
2032 | ||
893b7775 AV |
2033 | static void path_cleanup(struct nameidata *nd) |
2034 | { | |
2035 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
2036 | path_put(&nd->root); | |
2037 | nd->root.mnt = NULL; | |
2038 | } | |
9b4a9b14 AV |
2039 | } |
2040 | ||
3bdba28b | 2041 | static const char *trailing_symlink(struct nameidata *nd) |
95fa25d9 AV |
2042 | { |
2043 | const char *s; | |
fec2fa24 | 2044 | int error = may_follow_link(nd); |
deb106c6 | 2045 | if (unlikely(error)) |
3bdba28b | 2046 | return ERR_PTR(error); |
95fa25d9 | 2047 | nd->flags |= LOOKUP_PARENT; |
fab51e8a | 2048 | nd->stack[0].name = NULL; |
3b2e7f75 | 2049 | s = get_link(nd); |
deb106c6 | 2050 | return s ? s : ""; |
95fa25d9 AV |
2051 | } |
2052 | ||
caa85634 | 2053 | static inline int lookup_last(struct nameidata *nd) |
bd92d7fe AV |
2054 | { |
2055 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
2056 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
2057 | ||
2058 | nd->flags &= ~LOOKUP_PARENT; | |
deb106c6 | 2059 | return walk_component(nd, |
4693a547 AV |
2060 | nd->flags & LOOKUP_FOLLOW |
2061 | ? nd->depth | |
2062 | ? WALK_PUT | WALK_GET | |
2063 | : WALK_GET | |
2064 | : 0); | |
bd92d7fe AV |
2065 | } |
2066 | ||
9b4a9b14 | 2067 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
5eb6b495 | 2068 | static int path_lookupat(int dfd, const struct filename *name, |
9b4a9b14 AV |
2069 | unsigned int flags, struct nameidata *nd) |
2070 | { | |
368ee9ba | 2071 | const char *s = path_init(dfd, name, flags, nd); |
bd92d7fe | 2072 | int err; |
31e6b01f | 2073 | |
368ee9ba AV |
2074 | if (IS_ERR(s)) |
2075 | return PTR_ERR(s); | |
3bdba28b AV |
2076 | while (!(err = link_path_walk(s, nd)) |
2077 | && ((err = lookup_last(nd)) > 0)) { | |
2078 | s = trailing_symlink(nd); | |
2079 | if (IS_ERR(s)) { | |
2080 | err = PTR_ERR(s); | |
2081 | break; | |
bd92d7fe AV |
2082 | } |
2083 | } | |
9f1fafee AV |
2084 | if (!err) |
2085 | err = complete_walk(nd); | |
bd92d7fe | 2086 | |
deb106c6 AV |
2087 | if (!err && nd->flags & LOOKUP_DIRECTORY) |
2088 | if (!d_can_lookup(nd->path.dentry)) | |
bd23a539 | 2089 | err = -ENOTDIR; |
deb106c6 AV |
2090 | if (err) |
2091 | terminate_walk(nd); | |
16c2cd71 | 2092 | |
893b7775 | 2093 | path_cleanup(nd); |
bd92d7fe | 2094 | return err; |
ee0827cd | 2095 | } |
31e6b01f | 2096 | |
873f1eed | 2097 | static int filename_lookup(int dfd, struct filename *name, |
ee0827cd AV |
2098 | unsigned int flags, struct nameidata *nd) |
2099 | { | |
894bc8c4 | 2100 | int retval; |
756daf26 | 2101 | struct nameidata *saved_nd = set_nameidata(nd); |
894bc8c4 | 2102 | |
894bc8c4 | 2103 | retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); |
ee0827cd | 2104 | if (unlikely(retval == -ECHILD)) |
5eb6b495 | 2105 | retval = path_lookupat(dfd, name, flags, nd); |
ee0827cd | 2106 | if (unlikely(retval == -ESTALE)) |
5eb6b495 | 2107 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); |
31e6b01f | 2108 | |
f78570dd | 2109 | if (likely(!retval)) |
adb5c247 | 2110 | audit_inode(name, nd->path.dentry, flags & LOOKUP_PARENT); |
756daf26 | 2111 | restore_nameidata(saved_nd); |
170aa3d0 | 2112 | return retval; |
1da177e4 LT |
2113 | } |
2114 | ||
8bcb77fa AV |
2115 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
2116 | static int path_parentat(int dfd, const struct filename *name, | |
2117 | unsigned int flags, struct nameidata *nd) | |
2118 | { | |
368ee9ba AV |
2119 | const char *s = path_init(dfd, name, flags, nd); |
2120 | int err; | |
2121 | if (IS_ERR(s)) | |
2122 | return PTR_ERR(s); | |
2123 | err = link_path_walk(s, nd); | |
8bcb77fa AV |
2124 | if (!err) |
2125 | err = complete_walk(nd); | |
deb106c6 AV |
2126 | if (err) |
2127 | terminate_walk(nd); | |
8bcb77fa AV |
2128 | path_cleanup(nd); |
2129 | return err; | |
2130 | } | |
2131 | ||
2132 | static int filename_parentat(int dfd, struct filename *name, | |
2133 | unsigned int flags, struct nameidata *nd) | |
2134 | { | |
2135 | int retval; | |
2136 | struct nameidata *saved_nd = set_nameidata(nd); | |
2137 | ||
2138 | retval = path_parentat(dfd, name, flags | LOOKUP_RCU, nd); | |
2139 | if (unlikely(retval == -ECHILD)) | |
2140 | retval = path_parentat(dfd, name, flags, nd); | |
2141 | if (unlikely(retval == -ESTALE)) | |
2142 | retval = path_parentat(dfd, name, flags | LOOKUP_REVAL, nd); | |
2143 | ||
2144 | if (likely(!retval)) | |
2145 | audit_inode(name, nd->path.dentry, LOOKUP_PARENT); | |
2146 | restore_nameidata(saved_nd); | |
2147 | return retval; | |
2148 | } | |
2149 | ||
79714f72 AV |
2150 | /* does lookup, returns the object with parent locked */ |
2151 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 2152 | { |
51689104 | 2153 | struct filename *filename = getname_kernel(name); |
79714f72 AV |
2154 | struct nameidata nd; |
2155 | struct dentry *d; | |
51689104 PM |
2156 | int err; |
2157 | ||
2158 | if (IS_ERR(filename)) | |
2159 | return ERR_CAST(filename); | |
2160 | ||
8bcb77fa | 2161 | err = filename_parentat(AT_FDCWD, filename, 0, &nd); |
51689104 PM |
2162 | if (err) { |
2163 | d = ERR_PTR(err); | |
2164 | goto out; | |
2165 | } | |
79714f72 AV |
2166 | if (nd.last_type != LAST_NORM) { |
2167 | path_put(&nd.path); | |
51689104 PM |
2168 | d = ERR_PTR(-EINVAL); |
2169 | goto out; | |
79714f72 AV |
2170 | } |
2171 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); | |
1e0ea001 | 2172 | d = __lookup_hash(&nd.last, nd.path.dentry, 0); |
79714f72 AV |
2173 | if (IS_ERR(d)) { |
2174 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); | |
2175 | path_put(&nd.path); | |
51689104 | 2176 | goto out; |
79714f72 AV |
2177 | } |
2178 | *path = nd.path; | |
51689104 PM |
2179 | out: |
2180 | putname(filename); | |
79714f72 | 2181 | return d; |
5590ff0d UD |
2182 | } |
2183 | ||
d1811465 AV |
2184 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2185 | { | |
2186 | struct nameidata nd; | |
74eb8cc5 AV |
2187 | struct filename *filename = getname_kernel(name); |
2188 | int res = PTR_ERR(filename); | |
2189 | ||
2190 | if (!IS_ERR(filename)) { | |
2191 | res = filename_lookup(AT_FDCWD, filename, flags, &nd); | |
2192 | putname(filename); | |
2193 | if (!res) | |
2194 | *path = nd.path; | |
2195 | } | |
d1811465 AV |
2196 | return res; |
2197 | } | |
4d359507 | 2198 | EXPORT_SYMBOL(kern_path); |
d1811465 | 2199 | |
16f18200 JJS |
2200 | /** |
2201 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2202 | * @dentry: pointer to dentry of the base directory | |
2203 | * @mnt: pointer to vfs mount of the base directory | |
2204 | * @name: pointer to file name | |
2205 | * @flags: lookup flags | |
e0a01249 | 2206 | * @path: pointer to struct path to fill |
16f18200 JJS |
2207 | */ |
2208 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2209 | const char *name, unsigned int flags, | |
e0a01249 | 2210 | struct path *path) |
16f18200 | 2211 | { |
74eb8cc5 AV |
2212 | struct filename *filename = getname_kernel(name); |
2213 | int err = PTR_ERR(filename); | |
2214 | ||
e0a01249 | 2215 | BUG_ON(flags & LOOKUP_PARENT); |
74eb8cc5 AV |
2216 | |
2217 | /* the first argument of filename_lookup() is ignored with LOOKUP_ROOT */ | |
2218 | if (!IS_ERR(filename)) { | |
2219 | struct nameidata nd; | |
2220 | nd.root.dentry = dentry; | |
2221 | nd.root.mnt = mnt; | |
2222 | err = filename_lookup(AT_FDCWD, filename, | |
2223 | flags | LOOKUP_ROOT, &nd); | |
2224 | if (!err) | |
2225 | *path = nd.path; | |
2226 | putname(filename); | |
2227 | } | |
e0a01249 | 2228 | return err; |
16f18200 | 2229 | } |
4d359507 | 2230 | EXPORT_SYMBOL(vfs_path_lookup); |
16f18200 | 2231 | |
eead1911 | 2232 | /** |
a6b91919 | 2233 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
2234 | * @name: pathname component to lookup |
2235 | * @base: base directory to lookup from | |
2236 | * @len: maximum length @len should be interpreted to | |
2237 | * | |
a6b91919 | 2238 | * Note that this routine is purely a helper for filesystem usage and should |
9e7543e9 | 2239 | * not be called by generic code. |
eead1911 | 2240 | */ |
057f6c01 JM |
2241 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
2242 | { | |
057f6c01 | 2243 | struct qstr this; |
6a96ba54 | 2244 | unsigned int c; |
cda309de | 2245 | int err; |
057f6c01 | 2246 | |
2f9092e1 DW |
2247 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
2248 | ||
6a96ba54 AV |
2249 | this.name = name; |
2250 | this.len = len; | |
0145acc2 | 2251 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
2252 | if (!len) |
2253 | return ERR_PTR(-EACCES); | |
2254 | ||
21d8a15a AV |
2255 | if (unlikely(name[0] == '.')) { |
2256 | if (len < 2 || (len == 2 && name[1] == '.')) | |
2257 | return ERR_PTR(-EACCES); | |
2258 | } | |
2259 | ||
6a96ba54 AV |
2260 | while (len--) { |
2261 | c = *(const unsigned char *)name++; | |
2262 | if (c == '/' || c == '\0') | |
2263 | return ERR_PTR(-EACCES); | |
6a96ba54 | 2264 | } |
5a202bcd AV |
2265 | /* |
2266 | * See if the low-level filesystem might want | |
2267 | * to use its own hash.. | |
2268 | */ | |
2269 | if (base->d_flags & DCACHE_OP_HASH) { | |
da53be12 | 2270 | int err = base->d_op->d_hash(base, &this); |
5a202bcd AV |
2271 | if (err < 0) |
2272 | return ERR_PTR(err); | |
2273 | } | |
eead1911 | 2274 | |
cda309de MS |
2275 | err = inode_permission(base->d_inode, MAY_EXEC); |
2276 | if (err) | |
2277 | return ERR_PTR(err); | |
2278 | ||
72bd866a | 2279 | return __lookup_hash(&this, base, 0); |
057f6c01 | 2280 | } |
4d359507 | 2281 | EXPORT_SYMBOL(lookup_one_len); |
057f6c01 | 2282 | |
1fa1e7f6 AW |
2283 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2284 | struct path *path, int *empty) | |
1da177e4 | 2285 | { |
2d8f3038 | 2286 | struct nameidata nd; |
91a27b2a | 2287 | struct filename *tmp = getname_flags(name, flags, empty); |
1da177e4 | 2288 | int err = PTR_ERR(tmp); |
1da177e4 | 2289 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
2290 | |
2291 | BUG_ON(flags & LOOKUP_PARENT); | |
2292 | ||
873f1eed | 2293 | err = filename_lookup(dfd, tmp, flags, &nd); |
1da177e4 | 2294 | putname(tmp); |
2d8f3038 AV |
2295 | if (!err) |
2296 | *path = nd.path; | |
1da177e4 LT |
2297 | } |
2298 | return err; | |
2299 | } | |
2300 | ||
1fa1e7f6 AW |
2301 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
2302 | struct path *path) | |
2303 | { | |
f7493e5d | 2304 | return user_path_at_empty(dfd, name, flags, path, NULL); |
1fa1e7f6 | 2305 | } |
4d359507 | 2306 | EXPORT_SYMBOL(user_path_at); |
1fa1e7f6 | 2307 | |
873f1eed JL |
2308 | /* |
2309 | * NB: most callers don't do anything directly with the reference to the | |
2310 | * to struct filename, but the nd->last pointer points into the name string | |
2311 | * allocated by getname. So we must hold the reference to it until all | |
2312 | * path-walking is complete. | |
2313 | */ | |
91a27b2a | 2314 | static struct filename * |
f5beed75 AV |
2315 | user_path_parent(int dfd, const char __user *path, |
2316 | struct path *parent, | |
2317 | struct qstr *last, | |
2318 | int *type, | |
9e790bd6 | 2319 | unsigned int flags) |
2ad94ae6 | 2320 | { |
f5beed75 | 2321 | struct nameidata nd; |
91a27b2a | 2322 | struct filename *s = getname(path); |
2ad94ae6 AV |
2323 | int error; |
2324 | ||
9e790bd6 JL |
2325 | /* only LOOKUP_REVAL is allowed in extra flags */ |
2326 | flags &= LOOKUP_REVAL; | |
2327 | ||
2ad94ae6 | 2328 | if (IS_ERR(s)) |
91a27b2a | 2329 | return s; |
2ad94ae6 | 2330 | |
8bcb77fa | 2331 | error = filename_parentat(dfd, s, flags, &nd); |
91a27b2a | 2332 | if (error) { |
2ad94ae6 | 2333 | putname(s); |
91a27b2a JL |
2334 | return ERR_PTR(error); |
2335 | } | |
f5beed75 AV |
2336 | *parent = nd.path; |
2337 | *last = nd.last; | |
2338 | *type = nd.last_type; | |
2ad94ae6 | 2339 | |
91a27b2a | 2340 | return s; |
2ad94ae6 AV |
2341 | } |
2342 | ||
8033426e | 2343 | /** |
197df04c | 2344 | * mountpoint_last - look up last component for umount |
8033426e JL |
2345 | * @nd: pathwalk nameidata - currently pointing at parent directory of "last" |
2346 | * @path: pointer to container for result | |
2347 | * | |
2348 | * This is a special lookup_last function just for umount. In this case, we | |
2349 | * need to resolve the path without doing any revalidation. | |
2350 | * | |
2351 | * The nameidata should be the result of doing a LOOKUP_PARENT pathwalk. Since | |
2352 | * mountpoints are always pinned in the dcache, their ancestors are too. Thus, | |
2353 | * in almost all cases, this lookup will be served out of the dcache. The only | |
2354 | * cases where it won't are if nd->last refers to a symlink or the path is | |
2355 | * bogus and it doesn't exist. | |
2356 | * | |
2357 | * Returns: | |
2358 | * -error: if there was an error during lookup. This includes -ENOENT if the | |
2359 | * lookup found a negative dentry. The nd->path reference will also be | |
2360 | * put in this case. | |
2361 | * | |
2362 | * 0: if we successfully resolved nd->path and found it to not to be a | |
2363 | * symlink that needs to be followed. "path" will also be populated. | |
2364 | * The nd->path reference will also be put. | |
2365 | * | |
2366 | * 1: if we successfully resolved nd->last and found it to be a symlink | |
2367 | * that needs to be followed. "path" will be populated with the path | |
2368 | * to the link, and nd->path will *not* be put. | |
2369 | */ | |
2370 | static int | |
197df04c | 2371 | mountpoint_last(struct nameidata *nd, struct path *path) |
8033426e JL |
2372 | { |
2373 | int error = 0; | |
2374 | struct dentry *dentry; | |
2375 | struct dentry *dir = nd->path.dentry; | |
2376 | ||
35759521 AV |
2377 | /* If we're in rcuwalk, drop out of it to handle last component */ |
2378 | if (nd->flags & LOOKUP_RCU) { | |
6e9918b7 | 2379 | if (unlazy_walk(nd, NULL, 0)) |
deb106c6 | 2380 | return -ECHILD; |
8033426e JL |
2381 | } |
2382 | ||
2383 | nd->flags &= ~LOOKUP_PARENT; | |
2384 | ||
2385 | if (unlikely(nd->last_type != LAST_NORM)) { | |
2386 | error = handle_dots(nd, nd->last_type); | |
35759521 | 2387 | if (error) |
deb106c6 | 2388 | return error; |
35759521 AV |
2389 | dentry = dget(nd->path.dentry); |
2390 | goto done; | |
8033426e JL |
2391 | } |
2392 | ||
2393 | mutex_lock(&dir->d_inode->i_mutex); | |
2394 | dentry = d_lookup(dir, &nd->last); | |
2395 | if (!dentry) { | |
2396 | /* | |
2397 | * No cached dentry. Mounted dentries are pinned in the cache, | |
2398 | * so that means that this dentry is probably a symlink or the | |
2399 | * path doesn't actually point to a mounted dentry. | |
2400 | */ | |
2401 | dentry = d_alloc(dir, &nd->last); | |
2402 | if (!dentry) { | |
bcceeeba | 2403 | mutex_unlock(&dir->d_inode->i_mutex); |
deb106c6 | 2404 | return -ENOMEM; |
8033426e | 2405 | } |
35759521 | 2406 | dentry = lookup_real(dir->d_inode, dentry, nd->flags); |
bcceeeba DJ |
2407 | if (IS_ERR(dentry)) { |
2408 | mutex_unlock(&dir->d_inode->i_mutex); | |
deb106c6 | 2409 | return PTR_ERR(dentry); |
bcceeeba | 2410 | } |
8033426e JL |
2411 | } |
2412 | mutex_unlock(&dir->d_inode->i_mutex); | |
2413 | ||
35759521 | 2414 | done: |
698934df | 2415 | if (d_is_negative(dentry)) { |
35759521 | 2416 | dput(dentry); |
deb106c6 | 2417 | return -ENOENT; |
8033426e | 2418 | } |
191d7f73 AV |
2419 | if (nd->depth) |
2420 | put_link(nd); | |
35759521 | 2421 | path->dentry = dentry; |
295dc39d | 2422 | path->mnt = nd->path.mnt; |
181548c0 AV |
2423 | error = should_follow_link(nd, path, nd->flags & LOOKUP_FOLLOW, |
2424 | d_backing_inode(dentry), 0); | |
deb106c6 | 2425 | if (unlikely(error)) |
d63ff28f | 2426 | return error; |
295dc39d | 2427 | mntget(path->mnt); |
35759521 | 2428 | follow_mount(path); |
deb106c6 | 2429 | return 0; |
8033426e JL |
2430 | } |
2431 | ||
2432 | /** | |
197df04c | 2433 | * path_mountpoint - look up a path to be umounted |
8033426e JL |
2434 | * @dfd: directory file descriptor to start walk from |
2435 | * @name: full pathname to walk | |
606d6fe3 | 2436 | * @path: pointer to container for result |
8033426e | 2437 | * @flags: lookup flags |
8033426e JL |
2438 | * |
2439 | * Look up the given name, but don't attempt to revalidate the last component. | |
606d6fe3 | 2440 | * Returns 0 and "path" will be valid on success; Returns error otherwise. |
8033426e JL |
2441 | */ |
2442 | static int | |
668696dc | 2443 | path_mountpoint(int dfd, const struct filename *name, struct path *path, |
46afd6f6 | 2444 | struct nameidata *nd, unsigned int flags) |
8033426e | 2445 | { |
368ee9ba AV |
2446 | const char *s = path_init(dfd, name, flags, nd); |
2447 | int err; | |
2448 | if (IS_ERR(s)) | |
2449 | return PTR_ERR(s); | |
3bdba28b AV |
2450 | while (!(err = link_path_walk(s, nd)) && |
2451 | (err = mountpoint_last(nd, path)) > 0) { | |
2452 | s = trailing_symlink(nd); | |
2453 | if (IS_ERR(s)) { | |
2454 | err = PTR_ERR(s); | |
8033426e | 2455 | break; |
3bdba28b | 2456 | } |
8033426e | 2457 | } |
deb106c6 | 2458 | terminate_walk(nd); |
46afd6f6 | 2459 | path_cleanup(nd); |
8033426e JL |
2460 | return err; |
2461 | } | |
2462 | ||
2d864651 | 2463 | static int |
668696dc | 2464 | filename_mountpoint(int dfd, struct filename *name, struct path *path, |
2d864651 AV |
2465 | unsigned int flags) |
2466 | { | |
756daf26 | 2467 | struct nameidata nd, *saved; |
cbaab2db | 2468 | int error; |
668696dc AV |
2469 | if (IS_ERR(name)) |
2470 | return PTR_ERR(name); | |
756daf26 | 2471 | saved = set_nameidata(&nd); |
46afd6f6 | 2472 | error = path_mountpoint(dfd, name, path, &nd, flags | LOOKUP_RCU); |
2d864651 | 2473 | if (unlikely(error == -ECHILD)) |
46afd6f6 | 2474 | error = path_mountpoint(dfd, name, path, &nd, flags); |
2d864651 | 2475 | if (unlikely(error == -ESTALE)) |
46afd6f6 | 2476 | error = path_mountpoint(dfd, name, path, &nd, flags | LOOKUP_REVAL); |
2d864651 | 2477 | if (likely(!error)) |
668696dc | 2478 | audit_inode(name, path->dentry, 0); |
756daf26 | 2479 | restore_nameidata(saved); |
668696dc | 2480 | putname(name); |
2d864651 AV |
2481 | return error; |
2482 | } | |
2483 | ||
8033426e | 2484 | /** |
197df04c | 2485 | * user_path_mountpoint_at - lookup a path from userland in order to umount it |
8033426e JL |
2486 | * @dfd: directory file descriptor |
2487 | * @name: pathname from userland | |
2488 | * @flags: lookup flags | |
2489 | * @path: pointer to container to hold result | |
2490 | * | |
2491 | * A umount is a special case for path walking. We're not actually interested | |
2492 | * in the inode in this situation, and ESTALE errors can be a problem. We | |
2493 | * simply want track down the dentry and vfsmount attached at the mountpoint | |
2494 | * and avoid revalidating the last component. | |
2495 | * | |
2496 | * Returns 0 and populates "path" on success. | |
2497 | */ | |
2498 | int | |
197df04c | 2499 | user_path_mountpoint_at(int dfd, const char __user *name, unsigned int flags, |
8033426e JL |
2500 | struct path *path) |
2501 | { | |
cbaab2db | 2502 | return filename_mountpoint(dfd, getname(name), path, flags); |
8033426e JL |
2503 | } |
2504 | ||
2d864651 AV |
2505 | int |
2506 | kern_path_mountpoint(int dfd, const char *name, struct path *path, | |
2507 | unsigned int flags) | |
2508 | { | |
cbaab2db | 2509 | return filename_mountpoint(dfd, getname_kernel(name), path, flags); |
2d864651 AV |
2510 | } |
2511 | EXPORT_SYMBOL(kern_path_mountpoint); | |
2512 | ||
cbdf35bc | 2513 | int __check_sticky(struct inode *dir, struct inode *inode) |
1da177e4 | 2514 | { |
8e96e3b7 | 2515 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2516 | |
8e96e3b7 | 2517 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2518 | return 0; |
8e96e3b7 | 2519 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2520 | return 0; |
23adbe12 | 2521 | return !capable_wrt_inode_uidgid(inode, CAP_FOWNER); |
1da177e4 | 2522 | } |
cbdf35bc | 2523 | EXPORT_SYMBOL(__check_sticky); |
1da177e4 LT |
2524 | |
2525 | /* | |
2526 | * Check whether we can remove a link victim from directory dir, check | |
2527 | * whether the type of victim is right. | |
2528 | * 1. We can't do it if dir is read-only (done in permission()) | |
2529 | * 2. We should have write and exec permissions on dir | |
2530 | * 3. We can't remove anything from append-only dir | |
2531 | * 4. We can't do anything with immutable dir (done in permission()) | |
2532 | * 5. If the sticky bit on dir is set we should either | |
2533 | * a. be owner of dir, or | |
2534 | * b. be owner of victim, or | |
2535 | * c. have CAP_FOWNER capability | |
2536 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2537 | * links pointing to it. | |
2538 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2539 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2540 | * 9. We can't remove a root or mountpoint. | |
2541 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
2542 | * nfs_async_unlink(). | |
2543 | */ | |
b18825a7 | 2544 | static int may_delete(struct inode *dir, struct dentry *victim, bool isdir) |
1da177e4 | 2545 | { |
63afdfc7 | 2546 | struct inode *inode = d_backing_inode(victim); |
1da177e4 LT |
2547 | int error; |
2548 | ||
b18825a7 | 2549 | if (d_is_negative(victim)) |
1da177e4 | 2550 | return -ENOENT; |
b18825a7 | 2551 | BUG_ON(!inode); |
1da177e4 LT |
2552 | |
2553 | BUG_ON(victim->d_parent->d_inode != dir); | |
4fa6b5ec | 2554 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
1da177e4 | 2555 | |
f419a2e3 | 2556 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2557 | if (error) |
2558 | return error; | |
2559 | if (IS_APPEND(dir)) | |
2560 | return -EPERM; | |
b18825a7 DH |
2561 | |
2562 | if (check_sticky(dir, inode) || IS_APPEND(inode) || | |
2563 | IS_IMMUTABLE(inode) || IS_SWAPFILE(inode)) | |
1da177e4 LT |
2564 | return -EPERM; |
2565 | if (isdir) { | |
44b1d530 | 2566 | if (!d_is_dir(victim)) |
1da177e4 LT |
2567 | return -ENOTDIR; |
2568 | if (IS_ROOT(victim)) | |
2569 | return -EBUSY; | |
44b1d530 | 2570 | } else if (d_is_dir(victim)) |
1da177e4 LT |
2571 | return -EISDIR; |
2572 | if (IS_DEADDIR(dir)) | |
2573 | return -ENOENT; | |
2574 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2575 | return -EBUSY; | |
2576 | return 0; | |
2577 | } | |
2578 | ||
2579 | /* Check whether we can create an object with dentry child in directory | |
2580 | * dir. | |
2581 | * 1. We can't do it if child already exists (open has special treatment for | |
2582 | * this case, but since we are inlined it's OK) | |
2583 | * 2. We can't do it if dir is read-only (done in permission()) | |
2584 | * 3. We should have write and exec permissions on dir | |
2585 | * 4. We can't do it if dir is immutable (done in permission()) | |
2586 | */ | |
a95164d9 | 2587 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 | 2588 | { |
14e972b4 | 2589 | audit_inode_child(dir, child, AUDIT_TYPE_CHILD_CREATE); |
1da177e4 LT |
2590 | if (child->d_inode) |
2591 | return -EEXIST; | |
2592 | if (IS_DEADDIR(dir)) | |
2593 | return -ENOENT; | |
f419a2e3 | 2594 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2595 | } |
2596 | ||
1da177e4 LT |
2597 | /* |
2598 | * p1 and p2 should be directories on the same fs. | |
2599 | */ | |
2600 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2601 | { | |
2602 | struct dentry *p; | |
2603 | ||
2604 | if (p1 == p2) { | |
f2eace23 | 2605 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2606 | return NULL; |
2607 | } | |
2608 | ||
a11f3a05 | 2609 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2610 | |
e2761a11 OH |
2611 | p = d_ancestor(p2, p1); |
2612 | if (p) { | |
2613 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2614 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2615 | return p; | |
1da177e4 LT |
2616 | } |
2617 | ||
e2761a11 OH |
2618 | p = d_ancestor(p1, p2); |
2619 | if (p) { | |
2620 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2621 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2622 | return p; | |
1da177e4 LT |
2623 | } |
2624 | ||
f2eace23 | 2625 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
d1b72cc6 | 2626 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT2); |
1da177e4 LT |
2627 | return NULL; |
2628 | } | |
4d359507 | 2629 | EXPORT_SYMBOL(lock_rename); |
1da177e4 LT |
2630 | |
2631 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2632 | { | |
1b1dcc1b | 2633 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2634 | if (p1 != p2) { |
1b1dcc1b | 2635 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2636 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2637 | } |
2638 | } | |
4d359507 | 2639 | EXPORT_SYMBOL(unlock_rename); |
1da177e4 | 2640 | |
4acdaf27 | 2641 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2642 | bool want_excl) |
1da177e4 | 2643 | { |
a95164d9 | 2644 | int error = may_create(dir, dentry); |
1da177e4 LT |
2645 | if (error) |
2646 | return error; | |
2647 | ||
acfa4380 | 2648 | if (!dir->i_op->create) |
1da177e4 LT |
2649 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2650 | mode &= S_IALLUGO; | |
2651 | mode |= S_IFREG; | |
2652 | error = security_inode_create(dir, dentry, mode); | |
2653 | if (error) | |
2654 | return error; | |
312b63fb | 2655 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2656 | if (!error) |
f38aa942 | 2657 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2658 | return error; |
2659 | } | |
4d359507 | 2660 | EXPORT_SYMBOL(vfs_create); |
1da177e4 | 2661 | |
73d049a4 | 2662 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2663 | { |
3fb64190 | 2664 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2665 | struct inode *inode = dentry->d_inode; |
2666 | int error; | |
2667 | ||
bcda7652 AV |
2668 | /* O_PATH? */ |
2669 | if (!acc_mode) | |
2670 | return 0; | |
2671 | ||
1da177e4 LT |
2672 | if (!inode) |
2673 | return -ENOENT; | |
2674 | ||
c8fe8f30 CH |
2675 | switch (inode->i_mode & S_IFMT) { |
2676 | case S_IFLNK: | |
1da177e4 | 2677 | return -ELOOP; |
c8fe8f30 CH |
2678 | case S_IFDIR: |
2679 | if (acc_mode & MAY_WRITE) | |
2680 | return -EISDIR; | |
2681 | break; | |
2682 | case S_IFBLK: | |
2683 | case S_IFCHR: | |
3fb64190 | 2684 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2685 | return -EACCES; |
c8fe8f30 CH |
2686 | /*FALLTHRU*/ |
2687 | case S_IFIFO: | |
2688 | case S_IFSOCK: | |
1da177e4 | 2689 | flag &= ~O_TRUNC; |
c8fe8f30 | 2690 | break; |
4a3fd211 | 2691 | } |
b41572e9 | 2692 | |
3fb64190 | 2693 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2694 | if (error) |
2695 | return error; | |
6146f0d5 | 2696 | |
1da177e4 LT |
2697 | /* |
2698 | * An append-only file must be opened in append mode for writing. | |
2699 | */ | |
2700 | if (IS_APPEND(inode)) { | |
8737c930 | 2701 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2702 | return -EPERM; |
1da177e4 | 2703 | if (flag & O_TRUNC) |
7715b521 | 2704 | return -EPERM; |
1da177e4 LT |
2705 | } |
2706 | ||
2707 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2708 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2709 | return -EPERM; |
1da177e4 | 2710 | |
f3c7691e | 2711 | return 0; |
7715b521 | 2712 | } |
1da177e4 | 2713 | |
e1181ee6 | 2714 | static int handle_truncate(struct file *filp) |
7715b521 | 2715 | { |
e1181ee6 | 2716 | struct path *path = &filp->f_path; |
7715b521 AV |
2717 | struct inode *inode = path->dentry->d_inode; |
2718 | int error = get_write_access(inode); | |
2719 | if (error) | |
2720 | return error; | |
2721 | /* | |
2722 | * Refuse to truncate files with mandatory locks held on them. | |
2723 | */ | |
d7a06983 | 2724 | error = locks_verify_locked(filp); |
7715b521 | 2725 | if (!error) |
ea0d3ab2 | 2726 | error = security_path_truncate(path); |
7715b521 AV |
2727 | if (!error) { |
2728 | error = do_truncate(path->dentry, 0, | |
2729 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2730 | filp); |
7715b521 AV |
2731 | } |
2732 | put_write_access(inode); | |
acd0c935 | 2733 | return error; |
1da177e4 LT |
2734 | } |
2735 | ||
d57999e1 DH |
2736 | static inline int open_to_namei_flags(int flag) |
2737 | { | |
8a5e929d AV |
2738 | if ((flag & O_ACCMODE) == 3) |
2739 | flag--; | |
d57999e1 DH |
2740 | return flag; |
2741 | } | |
2742 | ||
d18e9008 MS |
2743 | static int may_o_create(struct path *dir, struct dentry *dentry, umode_t mode) |
2744 | { | |
2745 | int error = security_path_mknod(dir, dentry, mode, 0); | |
2746 | if (error) | |
2747 | return error; | |
2748 | ||
2749 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); | |
2750 | if (error) | |
2751 | return error; | |
2752 | ||
2753 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2754 | } | |
2755 | ||
1acf0af9 DH |
2756 | /* |
2757 | * Attempt to atomically look up, create and open a file from a negative | |
2758 | * dentry. | |
2759 | * | |
2760 | * Returns 0 if successful. The file will have been created and attached to | |
2761 | * @file by the filesystem calling finish_open(). | |
2762 | * | |
2763 | * Returns 1 if the file was looked up only or didn't need creating. The | |
2764 | * caller will need to perform the open themselves. @path will have been | |
2765 | * updated to point to the new dentry. This may be negative. | |
2766 | * | |
2767 | * Returns an error code otherwise. | |
2768 | */ | |
2675a4eb AV |
2769 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
2770 | struct path *path, struct file *file, | |
2771 | const struct open_flags *op, | |
64894cf8 | 2772 | bool got_write, bool need_lookup, |
2675a4eb | 2773 | int *opened) |
d18e9008 MS |
2774 | { |
2775 | struct inode *dir = nd->path.dentry->d_inode; | |
2776 | unsigned open_flag = open_to_namei_flags(op->open_flag); | |
2777 | umode_t mode; | |
2778 | int error; | |
2779 | int acc_mode; | |
d18e9008 MS |
2780 | int create_error = 0; |
2781 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; | |
116cc022 | 2782 | bool excl; |
d18e9008 MS |
2783 | |
2784 | BUG_ON(dentry->d_inode); | |
2785 | ||
2786 | /* Don't create child dentry for a dead directory. */ | |
2787 | if (unlikely(IS_DEADDIR(dir))) { | |
2675a4eb | 2788 | error = -ENOENT; |
d18e9008 MS |
2789 | goto out; |
2790 | } | |
2791 | ||
62b259d8 | 2792 | mode = op->mode; |
d18e9008 MS |
2793 | if ((open_flag & O_CREAT) && !IS_POSIXACL(dir)) |
2794 | mode &= ~current_umask(); | |
2795 | ||
116cc022 MS |
2796 | excl = (open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT); |
2797 | if (excl) | |
d18e9008 | 2798 | open_flag &= ~O_TRUNC; |
d18e9008 MS |
2799 | |
2800 | /* | |
2801 | * Checking write permission is tricky, bacuse we don't know if we are | |
2802 | * going to actually need it: O_CREAT opens should work as long as the | |
2803 | * file exists. But checking existence breaks atomicity. The trick is | |
2804 | * to check access and if not granted clear O_CREAT from the flags. | |
2805 | * | |
2806 | * Another problem is returing the "right" error value (e.g. for an | |
2807 | * O_EXCL open we want to return EEXIST not EROFS). | |
2808 | */ | |
64894cf8 AV |
2809 | if (((open_flag & (O_CREAT | O_TRUNC)) || |
2810 | (open_flag & O_ACCMODE) != O_RDONLY) && unlikely(!got_write)) { | |
2811 | if (!(open_flag & O_CREAT)) { | |
d18e9008 MS |
2812 | /* |
2813 | * No O_CREATE -> atomicity not a requirement -> fall | |
2814 | * back to lookup + open | |
2815 | */ | |
2816 | goto no_open; | |
2817 | } else if (open_flag & (O_EXCL | O_TRUNC)) { | |
2818 | /* Fall back and fail with the right error */ | |
64894cf8 | 2819 | create_error = -EROFS; |
d18e9008 MS |
2820 | goto no_open; |
2821 | } else { | |
2822 | /* No side effects, safe to clear O_CREAT */ | |
64894cf8 | 2823 | create_error = -EROFS; |
d18e9008 MS |
2824 | open_flag &= ~O_CREAT; |
2825 | } | |
2826 | } | |
2827 | ||
2828 | if (open_flag & O_CREAT) { | |
38227f78 | 2829 | error = may_o_create(&nd->path, dentry, mode); |
d18e9008 MS |
2830 | if (error) { |
2831 | create_error = error; | |
2832 | if (open_flag & O_EXCL) | |
2833 | goto no_open; | |
2834 | open_flag &= ~O_CREAT; | |
2835 | } | |
2836 | } | |
2837 | ||
2838 | if (nd->flags & LOOKUP_DIRECTORY) | |
2839 | open_flag |= O_DIRECTORY; | |
2840 | ||
30d90494 AV |
2841 | file->f_path.dentry = DENTRY_NOT_SET; |
2842 | file->f_path.mnt = nd->path.mnt; | |
2843 | error = dir->i_op->atomic_open(dir, dentry, file, open_flag, mode, | |
47237687 | 2844 | opened); |
d9585277 | 2845 | if (error < 0) { |
d9585277 AV |
2846 | if (create_error && error == -ENOENT) |
2847 | error = create_error; | |
d18e9008 MS |
2848 | goto out; |
2849 | } | |
2850 | ||
d9585277 | 2851 | if (error) { /* returned 1, that is */ |
30d90494 | 2852 | if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 2853 | error = -EIO; |
d18e9008 MS |
2854 | goto out; |
2855 | } | |
30d90494 | 2856 | if (file->f_path.dentry) { |
d18e9008 | 2857 | dput(dentry); |
30d90494 | 2858 | dentry = file->f_path.dentry; |
d18e9008 | 2859 | } |
03da633a AV |
2860 | if (*opened & FILE_CREATED) |
2861 | fsnotify_create(dir, dentry); | |
2862 | if (!dentry->d_inode) { | |
2863 | WARN_ON(*opened & FILE_CREATED); | |
2864 | if (create_error) { | |
2865 | error = create_error; | |
2866 | goto out; | |
2867 | } | |
2868 | } else { | |
2869 | if (excl && !(*opened & FILE_CREATED)) { | |
2870 | error = -EEXIST; | |
2871 | goto out; | |
2872 | } | |
62b2ce96 | 2873 | } |
d18e9008 MS |
2874 | goto looked_up; |
2875 | } | |
2876 | ||
2877 | /* | |
2878 | * We didn't have the inode before the open, so check open permission | |
2879 | * here. | |
2880 | */ | |
03da633a AV |
2881 | acc_mode = op->acc_mode; |
2882 | if (*opened & FILE_CREATED) { | |
2883 | WARN_ON(!(open_flag & O_CREAT)); | |
2884 | fsnotify_create(dir, dentry); | |
2885 | acc_mode = MAY_OPEN; | |
2886 | } | |
2675a4eb AV |
2887 | error = may_open(&file->f_path, acc_mode, open_flag); |
2888 | if (error) | |
2889 | fput(file); | |
d18e9008 MS |
2890 | |
2891 | out: | |
2892 | dput(dentry); | |
2675a4eb | 2893 | return error; |
d18e9008 | 2894 | |
d18e9008 MS |
2895 | no_open: |
2896 | if (need_lookup) { | |
72bd866a | 2897 | dentry = lookup_real(dir, dentry, nd->flags); |
d18e9008 | 2898 | if (IS_ERR(dentry)) |
2675a4eb | 2899 | return PTR_ERR(dentry); |
d18e9008 MS |
2900 | |
2901 | if (create_error) { | |
2902 | int open_flag = op->open_flag; | |
2903 | ||
2675a4eb | 2904 | error = create_error; |
d18e9008 MS |
2905 | if ((open_flag & O_EXCL)) { |
2906 | if (!dentry->d_inode) | |
2907 | goto out; | |
2908 | } else if (!dentry->d_inode) { | |
2909 | goto out; | |
2910 | } else if ((open_flag & O_TRUNC) && | |
e36cb0b8 | 2911 | d_is_reg(dentry)) { |
d18e9008 MS |
2912 | goto out; |
2913 | } | |
2914 | /* will fail later, go on to get the right error */ | |
2915 | } | |
2916 | } | |
2917 | looked_up: | |
2918 | path->dentry = dentry; | |
2919 | path->mnt = nd->path.mnt; | |
2675a4eb | 2920 | return 1; |
d18e9008 MS |
2921 | } |
2922 | ||
d58ffd35 | 2923 | /* |
1acf0af9 | 2924 | * Look up and maybe create and open the last component. |
d58ffd35 MS |
2925 | * |
2926 | * Must be called with i_mutex held on parent. | |
2927 | * | |
1acf0af9 DH |
2928 | * Returns 0 if the file was successfully atomically created (if necessary) and |
2929 | * opened. In this case the file will be returned attached to @file. | |
2930 | * | |
2931 | * Returns 1 if the file was not completely opened at this time, though lookups | |
2932 | * and creations will have been performed and the dentry returned in @path will | |
2933 | * be positive upon return if O_CREAT was specified. If O_CREAT wasn't | |
2934 | * specified then a negative dentry may be returned. | |
2935 | * | |
2936 | * An error code is returned otherwise. | |
2937 | * | |
2938 | * FILE_CREATE will be set in @*opened if the dentry was created and will be | |
2939 | * cleared otherwise prior to returning. | |
d58ffd35 | 2940 | */ |
2675a4eb AV |
2941 | static int lookup_open(struct nameidata *nd, struct path *path, |
2942 | struct file *file, | |
2943 | const struct open_flags *op, | |
64894cf8 | 2944 | bool got_write, int *opened) |
d58ffd35 MS |
2945 | { |
2946 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 2947 | struct inode *dir_inode = dir->d_inode; |
d58ffd35 MS |
2948 | struct dentry *dentry; |
2949 | int error; | |
54ef4872 | 2950 | bool need_lookup; |
d58ffd35 | 2951 | |
47237687 | 2952 | *opened &= ~FILE_CREATED; |
201f956e | 2953 | dentry = lookup_dcache(&nd->last, dir, nd->flags, &need_lookup); |
d58ffd35 | 2954 | if (IS_ERR(dentry)) |
2675a4eb | 2955 | return PTR_ERR(dentry); |
d58ffd35 | 2956 | |
d18e9008 MS |
2957 | /* Cached positive dentry: will open in f_op->open */ |
2958 | if (!need_lookup && dentry->d_inode) | |
2959 | goto out_no_open; | |
2960 | ||
2961 | if ((nd->flags & LOOKUP_OPEN) && dir_inode->i_op->atomic_open) { | |
64894cf8 | 2962 | return atomic_open(nd, dentry, path, file, op, got_write, |
47237687 | 2963 | need_lookup, opened); |
d18e9008 MS |
2964 | } |
2965 | ||
54ef4872 MS |
2966 | if (need_lookup) { |
2967 | BUG_ON(dentry->d_inode); | |
2968 | ||
72bd866a | 2969 | dentry = lookup_real(dir_inode, dentry, nd->flags); |
54ef4872 | 2970 | if (IS_ERR(dentry)) |
2675a4eb | 2971 | return PTR_ERR(dentry); |
54ef4872 MS |
2972 | } |
2973 | ||
d58ffd35 MS |
2974 | /* Negative dentry, just create the file */ |
2975 | if (!dentry->d_inode && (op->open_flag & O_CREAT)) { | |
2976 | umode_t mode = op->mode; | |
2977 | if (!IS_POSIXACL(dir->d_inode)) | |
2978 | mode &= ~current_umask(); | |
2979 | /* | |
2980 | * This write is needed to ensure that a | |
2981 | * rw->ro transition does not occur between | |
2982 | * the time when the file is created and when | |
2983 | * a permanent write count is taken through | |
015c3bbc | 2984 | * the 'struct file' in finish_open(). |
d58ffd35 | 2985 | */ |
64894cf8 AV |
2986 | if (!got_write) { |
2987 | error = -EROFS; | |
d58ffd35 | 2988 | goto out_dput; |
64894cf8 | 2989 | } |
47237687 | 2990 | *opened |= FILE_CREATED; |
d58ffd35 MS |
2991 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2992 | if (error) | |
2993 | goto out_dput; | |
312b63fb AV |
2994 | error = vfs_create(dir->d_inode, dentry, mode, |
2995 | nd->flags & LOOKUP_EXCL); | |
d58ffd35 MS |
2996 | if (error) |
2997 | goto out_dput; | |
2998 | } | |
d18e9008 | 2999 | out_no_open: |
d58ffd35 MS |
3000 | path->dentry = dentry; |
3001 | path->mnt = nd->path.mnt; | |
2675a4eb | 3002 | return 1; |
d58ffd35 MS |
3003 | |
3004 | out_dput: | |
3005 | dput(dentry); | |
2675a4eb | 3006 | return error; |
d58ffd35 MS |
3007 | } |
3008 | ||
31e6b01f | 3009 | /* |
fe2d35ff | 3010 | * Handle the last step of open() |
31e6b01f | 3011 | */ |
896475d5 | 3012 | static int do_last(struct nameidata *nd, |
2675a4eb | 3013 | struct file *file, const struct open_flags *op, |
669abf4e | 3014 | int *opened, struct filename *name) |
fb1cc555 | 3015 | { |
a1e28038 | 3016 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 3017 | int open_flag = op->open_flag; |
77d660a8 | 3018 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 3019 | bool got_write = false; |
bcda7652 | 3020 | int acc_mode = op->acc_mode; |
254cf582 | 3021 | unsigned seq; |
a1eb3315 | 3022 | struct inode *inode; |
16b1c1cd | 3023 | struct path save_parent = { .dentry = NULL, .mnt = NULL }; |
896475d5 | 3024 | struct path path; |
16b1c1cd | 3025 | bool retried = false; |
16c2cd71 | 3026 | int error; |
1f36f774 | 3027 | |
c3e380b0 AV |
3028 | nd->flags &= ~LOOKUP_PARENT; |
3029 | nd->flags |= op->intent; | |
3030 | ||
bc77daa7 | 3031 | if (nd->last_type != LAST_NORM) { |
fe2d35ff | 3032 | error = handle_dots(nd, nd->last_type); |
deb106c6 | 3033 | if (unlikely(error)) |
2675a4eb | 3034 | return error; |
e83db167 | 3035 | goto finish_open; |
1f36f774 | 3036 | } |
67ee3ad2 | 3037 | |
ca344a89 | 3038 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
3039 | if (nd->last.name[nd->last.len]) |
3040 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
3041 | /* we _can_ be in RCU mode here */ | |
254cf582 | 3042 | error = lookup_fast(nd, &path, &inode, &seq); |
71574865 MS |
3043 | if (likely(!error)) |
3044 | goto finish_lookup; | |
3045 | ||
3046 | if (error < 0) | |
deb106c6 | 3047 | return error; |
71574865 MS |
3048 | |
3049 | BUG_ON(nd->inode != dir->d_inode); | |
b6183df7 MS |
3050 | } else { |
3051 | /* create side of things */ | |
3052 | /* | |
3053 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
3054 | * has been cleared when we got to the last component we are | |
3055 | * about to look up | |
3056 | */ | |
3057 | error = complete_walk(nd); | |
e8bb73df | 3058 | if (error) |
2675a4eb | 3059 | return error; |
fe2d35ff | 3060 | |
33e2208a | 3061 | audit_inode(name, dir, LOOKUP_PARENT); |
b6183df7 | 3062 | /* trailing slashes? */ |
deb106c6 AV |
3063 | if (unlikely(nd->last.name[nd->last.len])) |
3064 | return -EISDIR; | |
b6183df7 | 3065 | } |
a2c36b45 | 3066 | |
16b1c1cd | 3067 | retry_lookup: |
64894cf8 AV |
3068 | if (op->open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
3069 | error = mnt_want_write(nd->path.mnt); | |
3070 | if (!error) | |
3071 | got_write = true; | |
3072 | /* | |
3073 | * do _not_ fail yet - we might not need that or fail with | |
3074 | * a different error; let lookup_open() decide; we'll be | |
3075 | * dropping this one anyway. | |
3076 | */ | |
3077 | } | |
a1e28038 | 3078 | mutex_lock(&dir->d_inode->i_mutex); |
896475d5 | 3079 | error = lookup_open(nd, &path, file, op, got_write, opened); |
d58ffd35 | 3080 | mutex_unlock(&dir->d_inode->i_mutex); |
a1e28038 | 3081 | |
2675a4eb AV |
3082 | if (error <= 0) { |
3083 | if (error) | |
d18e9008 MS |
3084 | goto out; |
3085 | ||
47237687 | 3086 | if ((*opened & FILE_CREATED) || |
496ad9aa | 3087 | !S_ISREG(file_inode(file)->i_mode)) |
77d660a8 | 3088 | will_truncate = false; |
d18e9008 | 3089 | |
adb5c247 | 3090 | audit_inode(name, file->f_path.dentry, 0); |
d18e9008 MS |
3091 | goto opened; |
3092 | } | |
fb1cc555 | 3093 | |
47237687 | 3094 | if (*opened & FILE_CREATED) { |
9b44f1b3 | 3095 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 3096 | open_flag &= ~O_TRUNC; |
77d660a8 | 3097 | will_truncate = false; |
bcda7652 | 3098 | acc_mode = MAY_OPEN; |
896475d5 | 3099 | path_to_nameidata(&path, nd); |
e83db167 | 3100 | goto finish_open_created; |
fb1cc555 AV |
3101 | } |
3102 | ||
3103 | /* | |
3134f37e | 3104 | * create/update audit record if it already exists. |
fb1cc555 | 3105 | */ |
896475d5 AV |
3106 | if (d_is_positive(path.dentry)) |
3107 | audit_inode(name, path.dentry, 0); | |
fb1cc555 | 3108 | |
d18e9008 MS |
3109 | /* |
3110 | * If atomic_open() acquired write access it is dropped now due to | |
3111 | * possible mount and symlink following (this might be optimized away if | |
3112 | * necessary...) | |
3113 | */ | |
64894cf8 | 3114 | if (got_write) { |
d18e9008 | 3115 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3116 | got_write = false; |
d18e9008 MS |
3117 | } |
3118 | ||
deb106c6 AV |
3119 | if (unlikely((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT))) { |
3120 | path_to_nameidata(&path, nd); | |
3121 | return -EEXIST; | |
3122 | } | |
fb1cc555 | 3123 | |
756daf26 | 3124 | error = follow_managed(&path, nd); |
deb106c6 AV |
3125 | if (unlikely(error < 0)) |
3126 | return error; | |
a3fbbde7 | 3127 | |
decf3400 | 3128 | BUG_ON(nd->flags & LOOKUP_RCU); |
63afdfc7 | 3129 | inode = d_backing_inode(path.dentry); |
254cf582 | 3130 | seq = 0; /* out of RCU mode, so the value doesn't matter */ |
deb106c6 | 3131 | if (unlikely(d_is_negative(path.dentry))) { |
896475d5 | 3132 | path_to_nameidata(&path, nd); |
deb106c6 | 3133 | return -ENOENT; |
54c33e7f | 3134 | } |
766c4cbf | 3135 | finish_lookup: |
d63ff28f AV |
3136 | if (nd->depth) |
3137 | put_link(nd); | |
181548c0 AV |
3138 | error = should_follow_link(nd, &path, nd->flags & LOOKUP_FOLLOW, |
3139 | inode, seq); | |
deb106c6 | 3140 | if (unlikely(error)) |
d63ff28f | 3141 | return error; |
fb1cc555 | 3142 | |
896475d5 AV |
3143 | if (unlikely(d_is_symlink(path.dentry)) && !(open_flag & O_PATH)) { |
3144 | path_to_nameidata(&path, nd); | |
deb106c6 | 3145 | return -ELOOP; |
a5cfe2d5 AV |
3146 | } |
3147 | ||
896475d5 AV |
3148 | if ((nd->flags & LOOKUP_RCU) || nd->path.mnt != path.mnt) { |
3149 | path_to_nameidata(&path, nd); | |
16b1c1cd MS |
3150 | } else { |
3151 | save_parent.dentry = nd->path.dentry; | |
896475d5 AV |
3152 | save_parent.mnt = mntget(path.mnt); |
3153 | nd->path.dentry = path.dentry; | |
16b1c1cd MS |
3154 | |
3155 | } | |
decf3400 | 3156 | nd->inode = inode; |
254cf582 | 3157 | nd->seq = seq; |
a3fbbde7 | 3158 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
bc77daa7 | 3159 | finish_open: |
a3fbbde7 | 3160 | error = complete_walk(nd); |
16b1c1cd MS |
3161 | if (error) { |
3162 | path_put(&save_parent); | |
2675a4eb | 3163 | return error; |
16b1c1cd | 3164 | } |
bc77daa7 | 3165 | audit_inode(name, nd->path.dentry, 0); |
fb1cc555 | 3166 | error = -EISDIR; |
44b1d530 | 3167 | if ((open_flag & O_CREAT) && d_is_dir(nd->path.dentry)) |
2675a4eb | 3168 | goto out; |
af2f5542 | 3169 | error = -ENOTDIR; |
44b1d530 | 3170 | if ((nd->flags & LOOKUP_DIRECTORY) && !d_can_lookup(nd->path.dentry)) |
2675a4eb | 3171 | goto out; |
4bbcbd3b | 3172 | if (!d_is_reg(nd->path.dentry)) |
77d660a8 | 3173 | will_truncate = false; |
6c0d46c4 | 3174 | |
0f9d1a10 AV |
3175 | if (will_truncate) { |
3176 | error = mnt_want_write(nd->path.mnt); | |
3177 | if (error) | |
2675a4eb | 3178 | goto out; |
64894cf8 | 3179 | got_write = true; |
0f9d1a10 | 3180 | } |
e83db167 | 3181 | finish_open_created: |
bcda7652 | 3182 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 3183 | if (error) |
2675a4eb | 3184 | goto out; |
4aa7c634 MS |
3185 | |
3186 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ | |
3187 | error = vfs_open(&nd->path, file, current_cred()); | |
3188 | if (!error) { | |
3189 | *opened |= FILE_OPENED; | |
3190 | } else { | |
30d90494 | 3191 | if (error == -EOPENSTALE) |
f60dc3db | 3192 | goto stale_open; |
015c3bbc | 3193 | goto out; |
f60dc3db | 3194 | } |
a8277b9b | 3195 | opened: |
2675a4eb | 3196 | error = open_check_o_direct(file); |
015c3bbc MS |
3197 | if (error) |
3198 | goto exit_fput; | |
3034a146 | 3199 | error = ima_file_check(file, op->acc_mode, *opened); |
aa4caadb MS |
3200 | if (error) |
3201 | goto exit_fput; | |
3202 | ||
3203 | if (will_truncate) { | |
2675a4eb | 3204 | error = handle_truncate(file); |
aa4caadb MS |
3205 | if (error) |
3206 | goto exit_fput; | |
0f9d1a10 | 3207 | } |
ca344a89 | 3208 | out: |
64894cf8 | 3209 | if (got_write) |
0f9d1a10 | 3210 | mnt_drop_write(nd->path.mnt); |
16b1c1cd | 3211 | path_put(&save_parent); |
2675a4eb | 3212 | return error; |
fb1cc555 | 3213 | |
015c3bbc | 3214 | exit_fput: |
2675a4eb AV |
3215 | fput(file); |
3216 | goto out; | |
015c3bbc | 3217 | |
f60dc3db MS |
3218 | stale_open: |
3219 | /* If no saved parent or already retried then can't retry */ | |
3220 | if (!save_parent.dentry || retried) | |
3221 | goto out; | |
3222 | ||
3223 | BUG_ON(save_parent.dentry != dir); | |
3224 | path_put(&nd->path); | |
3225 | nd->path = save_parent; | |
3226 | nd->inode = dir->d_inode; | |
3227 | save_parent.mnt = NULL; | |
3228 | save_parent.dentry = NULL; | |
64894cf8 | 3229 | if (got_write) { |
f60dc3db | 3230 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3231 | got_write = false; |
f60dc3db MS |
3232 | } |
3233 | retried = true; | |
3234 | goto retry_lookup; | |
fb1cc555 AV |
3235 | } |
3236 | ||
60545d0d AV |
3237 | static int do_tmpfile(int dfd, struct filename *pathname, |
3238 | struct nameidata *nd, int flags, | |
3239 | const struct open_flags *op, | |
3240 | struct file *file, int *opened) | |
3241 | { | |
3242 | static const struct qstr name = QSTR_INIT("/", 1); | |
3243 | struct dentry *dentry, *child; | |
3244 | struct inode *dir; | |
5eb6b495 | 3245 | int error = path_lookupat(dfd, pathname, |
60545d0d AV |
3246 | flags | LOOKUP_DIRECTORY, nd); |
3247 | if (unlikely(error)) | |
3248 | return error; | |
3249 | error = mnt_want_write(nd->path.mnt); | |
3250 | if (unlikely(error)) | |
3251 | goto out; | |
3252 | /* we want directory to be writable */ | |
3253 | error = inode_permission(nd->inode, MAY_WRITE | MAY_EXEC); | |
3254 | if (error) | |
3255 | goto out2; | |
3256 | dentry = nd->path.dentry; | |
3257 | dir = dentry->d_inode; | |
3258 | if (!dir->i_op->tmpfile) { | |
3259 | error = -EOPNOTSUPP; | |
3260 | goto out2; | |
3261 | } | |
3262 | child = d_alloc(dentry, &name); | |
3263 | if (unlikely(!child)) { | |
3264 | error = -ENOMEM; | |
3265 | goto out2; | |
3266 | } | |
3267 | nd->flags &= ~LOOKUP_DIRECTORY; | |
3268 | nd->flags |= op->intent; | |
3269 | dput(nd->path.dentry); | |
3270 | nd->path.dentry = child; | |
3271 | error = dir->i_op->tmpfile(dir, nd->path.dentry, op->mode); | |
3272 | if (error) | |
3273 | goto out2; | |
3274 | audit_inode(pathname, nd->path.dentry, 0); | |
69a91c23 ER |
3275 | /* Don't check for other permissions, the inode was just created */ |
3276 | error = may_open(&nd->path, MAY_OPEN, op->open_flag); | |
60545d0d AV |
3277 | if (error) |
3278 | goto out2; | |
3279 | file->f_path.mnt = nd->path.mnt; | |
3280 | error = finish_open(file, nd->path.dentry, NULL, opened); | |
3281 | if (error) | |
3282 | goto out2; | |
3283 | error = open_check_o_direct(file); | |
f4e0c30c | 3284 | if (error) { |
60545d0d | 3285 | fput(file); |
f4e0c30c AV |
3286 | } else if (!(op->open_flag & O_EXCL)) { |
3287 | struct inode *inode = file_inode(file); | |
3288 | spin_lock(&inode->i_lock); | |
3289 | inode->i_state |= I_LINKABLE; | |
3290 | spin_unlock(&inode->i_lock); | |
3291 | } | |
60545d0d AV |
3292 | out2: |
3293 | mnt_drop_write(nd->path.mnt); | |
3294 | out: | |
3295 | path_put(&nd->path); | |
3296 | return error; | |
3297 | } | |
3298 | ||
669abf4e | 3299 | static struct file *path_openat(int dfd, struct filename *pathname, |
73d049a4 | 3300 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 3301 | { |
368ee9ba | 3302 | const char *s; |
30d90494 | 3303 | struct file *file; |
47237687 | 3304 | int opened = 0; |
13aab428 | 3305 | int error; |
31e6b01f | 3306 | |
30d90494 | 3307 | file = get_empty_filp(); |
1afc99be AV |
3308 | if (IS_ERR(file)) |
3309 | return file; | |
31e6b01f | 3310 | |
30d90494 | 3311 | file->f_flags = op->open_flag; |
31e6b01f | 3312 | |
bb458c64 | 3313 | if (unlikely(file->f_flags & __O_TMPFILE)) { |
60545d0d | 3314 | error = do_tmpfile(dfd, pathname, nd, flags, op, file, &opened); |
f15133df | 3315 | goto out2; |
60545d0d AV |
3316 | } |
3317 | ||
368ee9ba AV |
3318 | s = path_init(dfd, pathname, flags, nd); |
3319 | if (IS_ERR(s)) { | |
3320 | put_filp(file); | |
3321 | return ERR_CAST(s); | |
3322 | } | |
3bdba28b AV |
3323 | while (!(error = link_path_walk(s, nd)) && |
3324 | (error = do_last(nd, file, op, &opened, pathname)) > 0) { | |
73d049a4 | 3325 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); |
3bdba28b AV |
3326 | s = trailing_symlink(nd); |
3327 | if (IS_ERR(s)) { | |
3328 | error = PTR_ERR(s); | |
2675a4eb | 3329 | break; |
3bdba28b | 3330 | } |
806b681c | 3331 | } |
deb106c6 | 3332 | terminate_walk(nd); |
893b7775 | 3333 | path_cleanup(nd); |
f15133df | 3334 | out2: |
2675a4eb AV |
3335 | if (!(opened & FILE_OPENED)) { |
3336 | BUG_ON(!error); | |
30d90494 | 3337 | put_filp(file); |
16b1c1cd | 3338 | } |
2675a4eb AV |
3339 | if (unlikely(error)) { |
3340 | if (error == -EOPENSTALE) { | |
3341 | if (flags & LOOKUP_RCU) | |
3342 | error = -ECHILD; | |
3343 | else | |
3344 | error = -ESTALE; | |
3345 | } | |
3346 | file = ERR_PTR(error); | |
3347 | } | |
3348 | return file; | |
1da177e4 LT |
3349 | } |
3350 | ||
669abf4e | 3351 | struct file *do_filp_open(int dfd, struct filename *pathname, |
f9652e10 | 3352 | const struct open_flags *op) |
13aab428 | 3353 | { |
756daf26 | 3354 | struct nameidata nd, *saved_nd = set_nameidata(&nd); |
f9652e10 | 3355 | int flags = op->lookup_flags; |
13aab428 AV |
3356 | struct file *filp; |
3357 | ||
73d049a4 | 3358 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 3359 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 3360 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 3361 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 3362 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
756daf26 | 3363 | restore_nameidata(saved_nd); |
13aab428 AV |
3364 | return filp; |
3365 | } | |
3366 | ||
73d049a4 | 3367 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
f9652e10 | 3368 | const char *name, const struct open_flags *op) |
73d049a4 | 3369 | { |
756daf26 | 3370 | struct nameidata nd, *saved_nd; |
73d049a4 | 3371 | struct file *file; |
51689104 | 3372 | struct filename *filename; |
f9652e10 | 3373 | int flags = op->lookup_flags | LOOKUP_ROOT; |
73d049a4 AV |
3374 | |
3375 | nd.root.mnt = mnt; | |
3376 | nd.root.dentry = dentry; | |
3377 | ||
b18825a7 | 3378 | if (d_is_symlink(dentry) && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
3379 | return ERR_PTR(-ELOOP); |
3380 | ||
51689104 PM |
3381 | filename = getname_kernel(name); |
3382 | if (unlikely(IS_ERR(filename))) | |
3383 | return ERR_CAST(filename); | |
3384 | ||
756daf26 | 3385 | saved_nd = set_nameidata(&nd); |
51689104 | 3386 | file = path_openat(-1, filename, &nd, op, flags | LOOKUP_RCU); |
73d049a4 | 3387 | if (unlikely(file == ERR_PTR(-ECHILD))) |
51689104 | 3388 | file = path_openat(-1, filename, &nd, op, flags); |
73d049a4 | 3389 | if (unlikely(file == ERR_PTR(-ESTALE))) |
51689104 | 3390 | file = path_openat(-1, filename, &nd, op, flags | LOOKUP_REVAL); |
756daf26 | 3391 | restore_nameidata(saved_nd); |
51689104 | 3392 | putname(filename); |
73d049a4 AV |
3393 | return file; |
3394 | } | |
3395 | ||
fa14a0b8 | 3396 | static struct dentry *filename_create(int dfd, struct filename *name, |
1ac12b4b | 3397 | struct path *path, unsigned int lookup_flags) |
1da177e4 | 3398 | { |
c663e5d8 | 3399 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d | 3400 | struct nameidata nd; |
c30dabfe | 3401 | int err2; |
1ac12b4b JL |
3402 | int error; |
3403 | bool is_dir = (lookup_flags & LOOKUP_DIRECTORY); | |
3404 | ||
3405 | /* | |
3406 | * Note that only LOOKUP_REVAL and LOOKUP_DIRECTORY matter here. Any | |
3407 | * other flags passed in are ignored! | |
3408 | */ | |
3409 | lookup_flags &= LOOKUP_REVAL; | |
3410 | ||
8bcb77fa | 3411 | error = filename_parentat(dfd, name, lookup_flags, &nd); |
ed75e95d AV |
3412 | if (error) |
3413 | return ERR_PTR(error); | |
1da177e4 | 3414 | |
c663e5d8 CH |
3415 | /* |
3416 | * Yucky last component or no last component at all? | |
3417 | * (foo/., foo/.., /////) | |
3418 | */ | |
ed75e95d AV |
3419 | if (nd.last_type != LAST_NORM) |
3420 | goto out; | |
3421 | nd.flags &= ~LOOKUP_PARENT; | |
3422 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
c663e5d8 | 3423 | |
c30dabfe JK |
3424 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
3425 | err2 = mnt_want_write(nd.path.mnt); | |
c663e5d8 CH |
3426 | /* |
3427 | * Do the final lookup. | |
3428 | */ | |
ed75e95d | 3429 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
6a9f40d6 | 3430 | dentry = __lookup_hash(&nd.last, nd.path.dentry, nd.flags); |
1da177e4 | 3431 | if (IS_ERR(dentry)) |
a8104a9f | 3432 | goto unlock; |
c663e5d8 | 3433 | |
a8104a9f | 3434 | error = -EEXIST; |
b18825a7 | 3435 | if (d_is_positive(dentry)) |
a8104a9f | 3436 | goto fail; |
b18825a7 | 3437 | |
c663e5d8 CH |
3438 | /* |
3439 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3440 | * From the vfs_mknod() POV we just have a negative dentry - | |
3441 | * all is fine. Let's be bastards - you had / on the end, you've | |
3442 | * been asking for (non-existent) directory. -ENOENT for you. | |
3443 | */ | |
ed75e95d | 3444 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
a8104a9f | 3445 | error = -ENOENT; |
ed75e95d | 3446 | goto fail; |
e9baf6e5 | 3447 | } |
c30dabfe JK |
3448 | if (unlikely(err2)) { |
3449 | error = err2; | |
a8104a9f | 3450 | goto fail; |
c30dabfe | 3451 | } |
ed75e95d | 3452 | *path = nd.path; |
1da177e4 | 3453 | return dentry; |
1da177e4 | 3454 | fail: |
a8104a9f AV |
3455 | dput(dentry); |
3456 | dentry = ERR_PTR(error); | |
3457 | unlock: | |
ed75e95d | 3458 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
c30dabfe JK |
3459 | if (!err2) |
3460 | mnt_drop_write(nd.path.mnt); | |
ed75e95d AV |
3461 | out: |
3462 | path_put(&nd.path); | |
1da177e4 LT |
3463 | return dentry; |
3464 | } | |
fa14a0b8 AV |
3465 | |
3466 | struct dentry *kern_path_create(int dfd, const char *pathname, | |
3467 | struct path *path, unsigned int lookup_flags) | |
3468 | { | |
51689104 PM |
3469 | struct filename *filename = getname_kernel(pathname); |
3470 | struct dentry *res; | |
3471 | ||
3472 | if (IS_ERR(filename)) | |
3473 | return ERR_CAST(filename); | |
3474 | res = filename_create(dfd, filename, path, lookup_flags); | |
3475 | putname(filename); | |
3476 | return res; | |
fa14a0b8 | 3477 | } |
dae6ad8f AV |
3478 | EXPORT_SYMBOL(kern_path_create); |
3479 | ||
921a1650 AV |
3480 | void done_path_create(struct path *path, struct dentry *dentry) |
3481 | { | |
3482 | dput(dentry); | |
3483 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
a8104a9f | 3484 | mnt_drop_write(path->mnt); |
921a1650 AV |
3485 | path_put(path); |
3486 | } | |
3487 | EXPORT_SYMBOL(done_path_create); | |
3488 | ||
1ac12b4b JL |
3489 | struct dentry *user_path_create(int dfd, const char __user *pathname, |
3490 | struct path *path, unsigned int lookup_flags) | |
dae6ad8f | 3491 | { |
91a27b2a | 3492 | struct filename *tmp = getname(pathname); |
dae6ad8f AV |
3493 | struct dentry *res; |
3494 | if (IS_ERR(tmp)) | |
3495 | return ERR_CAST(tmp); | |
fa14a0b8 | 3496 | res = filename_create(dfd, tmp, path, lookup_flags); |
dae6ad8f AV |
3497 | putname(tmp); |
3498 | return res; | |
3499 | } | |
3500 | EXPORT_SYMBOL(user_path_create); | |
3501 | ||
1a67aafb | 3502 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3503 | { |
a95164d9 | 3504 | int error = may_create(dir, dentry); |
1da177e4 LT |
3505 | |
3506 | if (error) | |
3507 | return error; | |
3508 | ||
975d6b39 | 3509 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3510 | return -EPERM; |
3511 | ||
acfa4380 | 3512 | if (!dir->i_op->mknod) |
1da177e4 LT |
3513 | return -EPERM; |
3514 | ||
08ce5f16 SH |
3515 | error = devcgroup_inode_mknod(mode, dev); |
3516 | if (error) | |
3517 | return error; | |
3518 | ||
1da177e4 LT |
3519 | error = security_inode_mknod(dir, dentry, mode, dev); |
3520 | if (error) | |
3521 | return error; | |
3522 | ||
1da177e4 | 3523 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3524 | if (!error) |
f38aa942 | 3525 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3526 | return error; |
3527 | } | |
4d359507 | 3528 | EXPORT_SYMBOL(vfs_mknod); |
1da177e4 | 3529 | |
f69aac00 | 3530 | static int may_mknod(umode_t mode) |
463c3197 DH |
3531 | { |
3532 | switch (mode & S_IFMT) { | |
3533 | case S_IFREG: | |
3534 | case S_IFCHR: | |
3535 | case S_IFBLK: | |
3536 | case S_IFIFO: | |
3537 | case S_IFSOCK: | |
3538 | case 0: /* zero mode translates to S_IFREG */ | |
3539 | return 0; | |
3540 | case S_IFDIR: | |
3541 | return -EPERM; | |
3542 | default: | |
3543 | return -EINVAL; | |
3544 | } | |
3545 | } | |
3546 | ||
8208a22b | 3547 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 3548 | unsigned, dev) |
1da177e4 | 3549 | { |
2ad94ae6 | 3550 | struct dentry *dentry; |
dae6ad8f AV |
3551 | struct path path; |
3552 | int error; | |
972567f1 | 3553 | unsigned int lookup_flags = 0; |
1da177e4 | 3554 | |
8e4bfca1 AV |
3555 | error = may_mknod(mode); |
3556 | if (error) | |
3557 | return error; | |
972567f1 JL |
3558 | retry: |
3559 | dentry = user_path_create(dfd, filename, &path, lookup_flags); | |
dae6ad8f AV |
3560 | if (IS_ERR(dentry)) |
3561 | return PTR_ERR(dentry); | |
2ad94ae6 | 3562 | |
dae6ad8f | 3563 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3564 | mode &= ~current_umask(); |
dae6ad8f | 3565 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3566 | if (error) |
a8104a9f | 3567 | goto out; |
463c3197 | 3568 | switch (mode & S_IFMT) { |
1da177e4 | 3569 | case 0: case S_IFREG: |
312b63fb | 3570 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
1da177e4 LT |
3571 | break; |
3572 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3573 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3574 | new_decode_dev(dev)); |
3575 | break; | |
3576 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3577 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3578 | break; |
1da177e4 | 3579 | } |
a8104a9f | 3580 | out: |
921a1650 | 3581 | done_path_create(&path, dentry); |
972567f1 JL |
3582 | if (retry_estale(error, lookup_flags)) { |
3583 | lookup_flags |= LOOKUP_REVAL; | |
3584 | goto retry; | |
3585 | } | |
1da177e4 LT |
3586 | return error; |
3587 | } | |
3588 | ||
8208a22b | 3589 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
3590 | { |
3591 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
3592 | } | |
3593 | ||
18bb1db3 | 3594 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3595 | { |
a95164d9 | 3596 | int error = may_create(dir, dentry); |
8de52778 | 3597 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3598 | |
3599 | if (error) | |
3600 | return error; | |
3601 | ||
acfa4380 | 3602 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3603 | return -EPERM; |
3604 | ||
3605 | mode &= (S_IRWXUGO|S_ISVTX); | |
3606 | error = security_inode_mkdir(dir, dentry, mode); | |
3607 | if (error) | |
3608 | return error; | |
3609 | ||
8de52778 AV |
3610 | if (max_links && dir->i_nlink >= max_links) |
3611 | return -EMLINK; | |
3612 | ||
1da177e4 | 3613 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3614 | if (!error) |
f38aa942 | 3615 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3616 | return error; |
3617 | } | |
4d359507 | 3618 | EXPORT_SYMBOL(vfs_mkdir); |
1da177e4 | 3619 | |
a218d0fd | 3620 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 3621 | { |
6902d925 | 3622 | struct dentry *dentry; |
dae6ad8f AV |
3623 | struct path path; |
3624 | int error; | |
b76d8b82 | 3625 | unsigned int lookup_flags = LOOKUP_DIRECTORY; |
1da177e4 | 3626 | |
b76d8b82 JL |
3627 | retry: |
3628 | dentry = user_path_create(dfd, pathname, &path, lookup_flags); | |
6902d925 | 3629 | if (IS_ERR(dentry)) |
dae6ad8f | 3630 | return PTR_ERR(dentry); |
1da177e4 | 3631 | |
dae6ad8f | 3632 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3633 | mode &= ~current_umask(); |
dae6ad8f | 3634 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3635 | if (!error) |
3636 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3637 | done_path_create(&path, dentry); |
b76d8b82 JL |
3638 | if (retry_estale(error, lookup_flags)) { |
3639 | lookup_flags |= LOOKUP_REVAL; | |
3640 | goto retry; | |
3641 | } | |
1da177e4 LT |
3642 | return error; |
3643 | } | |
3644 | ||
a218d0fd | 3645 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
3646 | { |
3647 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
3648 | } | |
3649 | ||
1da177e4 | 3650 | /* |
a71905f0 | 3651 | * The dentry_unhash() helper will try to drop the dentry early: we |
c0d02594 | 3652 | * should have a usage count of 1 if we're the only user of this |
a71905f0 SW |
3653 | * dentry, and if that is true (possibly after pruning the dcache), |
3654 | * then we drop the dentry now. | |
1da177e4 LT |
3655 | * |
3656 | * A low-level filesystem can, if it choses, legally | |
3657 | * do a | |
3658 | * | |
3659 | * if (!d_unhashed(dentry)) | |
3660 | * return -EBUSY; | |
3661 | * | |
3662 | * if it cannot handle the case of removing a directory | |
3663 | * that is still in use by something else.. | |
3664 | */ | |
3665 | void dentry_unhash(struct dentry *dentry) | |
3666 | { | |
dc168427 | 3667 | shrink_dcache_parent(dentry); |
1da177e4 | 3668 | spin_lock(&dentry->d_lock); |
98474236 | 3669 | if (dentry->d_lockref.count == 1) |
1da177e4 LT |
3670 | __d_drop(dentry); |
3671 | spin_unlock(&dentry->d_lock); | |
1da177e4 | 3672 | } |
4d359507 | 3673 | EXPORT_SYMBOL(dentry_unhash); |
1da177e4 LT |
3674 | |
3675 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
3676 | { | |
3677 | int error = may_delete(dir, dentry, 1); | |
3678 | ||
3679 | if (error) | |
3680 | return error; | |
3681 | ||
acfa4380 | 3682 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3683 | return -EPERM; |
3684 | ||
1d2ef590 | 3685 | dget(dentry); |
1b1dcc1b | 3686 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
3687 | |
3688 | error = -EBUSY; | |
7af1364f | 3689 | if (is_local_mountpoint(dentry)) |
912dbc15 SW |
3690 | goto out; |
3691 | ||
3692 | error = security_inode_rmdir(dir, dentry); | |
3693 | if (error) | |
3694 | goto out; | |
3695 | ||
3cebde24 | 3696 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3697 | error = dir->i_op->rmdir(dir, dentry); |
3698 | if (error) | |
3699 | goto out; | |
3700 | ||
3701 | dentry->d_inode->i_flags |= S_DEAD; | |
3702 | dont_mount(dentry); | |
8ed936b5 | 3703 | detach_mounts(dentry); |
912dbc15 SW |
3704 | |
3705 | out: | |
1b1dcc1b | 3706 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 3707 | dput(dentry); |
912dbc15 | 3708 | if (!error) |
1da177e4 | 3709 | d_delete(dentry); |
1da177e4 LT |
3710 | return error; |
3711 | } | |
4d359507 | 3712 | EXPORT_SYMBOL(vfs_rmdir); |
1da177e4 | 3713 | |
5590ff0d | 3714 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3715 | { |
3716 | int error = 0; | |
91a27b2a | 3717 | struct filename *name; |
1da177e4 | 3718 | struct dentry *dentry; |
f5beed75 AV |
3719 | struct path path; |
3720 | struct qstr last; | |
3721 | int type; | |
c6ee9206 JL |
3722 | unsigned int lookup_flags = 0; |
3723 | retry: | |
f5beed75 AV |
3724 | name = user_path_parent(dfd, pathname, |
3725 | &path, &last, &type, lookup_flags); | |
91a27b2a JL |
3726 | if (IS_ERR(name)) |
3727 | return PTR_ERR(name); | |
1da177e4 | 3728 | |
f5beed75 | 3729 | switch (type) { |
0612d9fb OH |
3730 | case LAST_DOTDOT: |
3731 | error = -ENOTEMPTY; | |
3732 | goto exit1; | |
3733 | case LAST_DOT: | |
3734 | error = -EINVAL; | |
3735 | goto exit1; | |
3736 | case LAST_ROOT: | |
3737 | error = -EBUSY; | |
3738 | goto exit1; | |
1da177e4 | 3739 | } |
0612d9fb | 3740 | |
f5beed75 | 3741 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3742 | if (error) |
3743 | goto exit1; | |
0612d9fb | 3744 | |
f5beed75 AV |
3745 | mutex_lock_nested(&path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
3746 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); | |
1da177e4 | 3747 | error = PTR_ERR(dentry); |
6902d925 DH |
3748 | if (IS_ERR(dentry)) |
3749 | goto exit2; | |
e6bc45d6 TT |
3750 | if (!dentry->d_inode) { |
3751 | error = -ENOENT; | |
3752 | goto exit3; | |
3753 | } | |
f5beed75 | 3754 | error = security_path_rmdir(&path, dentry); |
be6d3e56 | 3755 | if (error) |
c30dabfe | 3756 | goto exit3; |
f5beed75 | 3757 | error = vfs_rmdir(path.dentry->d_inode, dentry); |
0622753b | 3758 | exit3: |
6902d925 DH |
3759 | dput(dentry); |
3760 | exit2: | |
f5beed75 AV |
3761 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
3762 | mnt_drop_write(path.mnt); | |
1da177e4 | 3763 | exit1: |
f5beed75 | 3764 | path_put(&path); |
1da177e4 | 3765 | putname(name); |
c6ee9206 JL |
3766 | if (retry_estale(error, lookup_flags)) { |
3767 | lookup_flags |= LOOKUP_REVAL; | |
3768 | goto retry; | |
3769 | } | |
1da177e4 LT |
3770 | return error; |
3771 | } | |
3772 | ||
3cdad428 | 3773 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3774 | { |
3775 | return do_rmdir(AT_FDCWD, pathname); | |
3776 | } | |
3777 | ||
b21996e3 BF |
3778 | /** |
3779 | * vfs_unlink - unlink a filesystem object | |
3780 | * @dir: parent directory | |
3781 | * @dentry: victim | |
3782 | * @delegated_inode: returns victim inode, if the inode is delegated. | |
3783 | * | |
3784 | * The caller must hold dir->i_mutex. | |
3785 | * | |
3786 | * If vfs_unlink discovers a delegation, it will return -EWOULDBLOCK and | |
3787 | * return a reference to the inode in delegated_inode. The caller | |
3788 | * should then break the delegation on that inode and retry. Because | |
3789 | * breaking a delegation may take a long time, the caller should drop | |
3790 | * dir->i_mutex before doing so. | |
3791 | * | |
3792 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3793 | * be appropriate for callers that expect the underlying filesystem not | |
3794 | * to be NFS exported. | |
3795 | */ | |
3796 | int vfs_unlink(struct inode *dir, struct dentry *dentry, struct inode **delegated_inode) | |
1da177e4 | 3797 | { |
9accbb97 | 3798 | struct inode *target = dentry->d_inode; |
1da177e4 LT |
3799 | int error = may_delete(dir, dentry, 0); |
3800 | ||
3801 | if (error) | |
3802 | return error; | |
3803 | ||
acfa4380 | 3804 | if (!dir->i_op->unlink) |
1da177e4 LT |
3805 | return -EPERM; |
3806 | ||
9accbb97 | 3807 | mutex_lock(&target->i_mutex); |
8ed936b5 | 3808 | if (is_local_mountpoint(dentry)) |
1da177e4 LT |
3809 | error = -EBUSY; |
3810 | else { | |
3811 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3812 | if (!error) { |
5a14696c BF |
3813 | error = try_break_deleg(target, delegated_inode); |
3814 | if (error) | |
b21996e3 | 3815 | goto out; |
1da177e4 | 3816 | error = dir->i_op->unlink(dir, dentry); |
8ed936b5 | 3817 | if (!error) { |
d83c49f3 | 3818 | dont_mount(dentry); |
8ed936b5 EB |
3819 | detach_mounts(dentry); |
3820 | } | |
bec1052e | 3821 | } |
1da177e4 | 3822 | } |
b21996e3 | 3823 | out: |
9accbb97 | 3824 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3825 | |
3826 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3827 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
9accbb97 | 3828 | fsnotify_link_count(target); |
e234f35c | 3829 | d_delete(dentry); |
1da177e4 | 3830 | } |
0eeca283 | 3831 | |
1da177e4 LT |
3832 | return error; |
3833 | } | |
4d359507 | 3834 | EXPORT_SYMBOL(vfs_unlink); |
1da177e4 LT |
3835 | |
3836 | /* | |
3837 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3838 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3839 | * writeout happening, and we don't want to prevent access to the directory |
3840 | * while waiting on the I/O. | |
3841 | */ | |
5590ff0d | 3842 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 3843 | { |
2ad94ae6 | 3844 | int error; |
91a27b2a | 3845 | struct filename *name; |
1da177e4 | 3846 | struct dentry *dentry; |
f5beed75 AV |
3847 | struct path path; |
3848 | struct qstr last; | |
3849 | int type; | |
1da177e4 | 3850 | struct inode *inode = NULL; |
b21996e3 | 3851 | struct inode *delegated_inode = NULL; |
5d18f813 JL |
3852 | unsigned int lookup_flags = 0; |
3853 | retry: | |
f5beed75 AV |
3854 | name = user_path_parent(dfd, pathname, |
3855 | &path, &last, &type, lookup_flags); | |
91a27b2a JL |
3856 | if (IS_ERR(name)) |
3857 | return PTR_ERR(name); | |
2ad94ae6 | 3858 | |
1da177e4 | 3859 | error = -EISDIR; |
f5beed75 | 3860 | if (type != LAST_NORM) |
1da177e4 | 3861 | goto exit1; |
0612d9fb | 3862 | |
f5beed75 | 3863 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3864 | if (error) |
3865 | goto exit1; | |
b21996e3 | 3866 | retry_deleg: |
f5beed75 AV |
3867 | mutex_lock_nested(&path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
3868 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); | |
1da177e4 LT |
3869 | error = PTR_ERR(dentry); |
3870 | if (!IS_ERR(dentry)) { | |
3871 | /* Why not before? Because we want correct error value */ | |
f5beed75 | 3872 | if (last.name[last.len]) |
50338b88 | 3873 | goto slashes; |
1da177e4 | 3874 | inode = dentry->d_inode; |
b18825a7 | 3875 | if (d_is_negative(dentry)) |
e6bc45d6 TT |
3876 | goto slashes; |
3877 | ihold(inode); | |
f5beed75 | 3878 | error = security_path_unlink(&path, dentry); |
be6d3e56 | 3879 | if (error) |
c30dabfe | 3880 | goto exit2; |
f5beed75 | 3881 | error = vfs_unlink(path.dentry->d_inode, dentry, &delegated_inode); |
c30dabfe | 3882 | exit2: |
1da177e4 LT |
3883 | dput(dentry); |
3884 | } | |
f5beed75 | 3885 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
1da177e4 LT |
3886 | if (inode) |
3887 | iput(inode); /* truncate the inode here */ | |
b21996e3 BF |
3888 | inode = NULL; |
3889 | if (delegated_inode) { | |
5a14696c | 3890 | error = break_deleg_wait(&delegated_inode); |
b21996e3 BF |
3891 | if (!error) |
3892 | goto retry_deleg; | |
3893 | } | |
f5beed75 | 3894 | mnt_drop_write(path.mnt); |
1da177e4 | 3895 | exit1: |
f5beed75 | 3896 | path_put(&path); |
1da177e4 | 3897 | putname(name); |
5d18f813 JL |
3898 | if (retry_estale(error, lookup_flags)) { |
3899 | lookup_flags |= LOOKUP_REVAL; | |
3900 | inode = NULL; | |
3901 | goto retry; | |
3902 | } | |
1da177e4 LT |
3903 | return error; |
3904 | ||
3905 | slashes: | |
b18825a7 DH |
3906 | if (d_is_negative(dentry)) |
3907 | error = -ENOENT; | |
44b1d530 | 3908 | else if (d_is_dir(dentry)) |
b18825a7 DH |
3909 | error = -EISDIR; |
3910 | else | |
3911 | error = -ENOTDIR; | |
1da177e4 LT |
3912 | goto exit2; |
3913 | } | |
3914 | ||
2e4d0924 | 3915 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3916 | { |
3917 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3918 | return -EINVAL; | |
3919 | ||
3920 | if (flag & AT_REMOVEDIR) | |
3921 | return do_rmdir(dfd, pathname); | |
3922 | ||
3923 | return do_unlinkat(dfd, pathname); | |
3924 | } | |
3925 | ||
3480b257 | 3926 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
3927 | { |
3928 | return do_unlinkat(AT_FDCWD, pathname); | |
3929 | } | |
3930 | ||
db2e747b | 3931 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3932 | { |
a95164d9 | 3933 | int error = may_create(dir, dentry); |
1da177e4 LT |
3934 | |
3935 | if (error) | |
3936 | return error; | |
3937 | ||
acfa4380 | 3938 | if (!dir->i_op->symlink) |
1da177e4 LT |
3939 | return -EPERM; |
3940 | ||
3941 | error = security_inode_symlink(dir, dentry, oldname); | |
3942 | if (error) | |
3943 | return error; | |
3944 | ||
1da177e4 | 3945 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3946 | if (!error) |
f38aa942 | 3947 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3948 | return error; |
3949 | } | |
4d359507 | 3950 | EXPORT_SYMBOL(vfs_symlink); |
1da177e4 | 3951 | |
2e4d0924 HC |
3952 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
3953 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3954 | { |
2ad94ae6 | 3955 | int error; |
91a27b2a | 3956 | struct filename *from; |
6902d925 | 3957 | struct dentry *dentry; |
dae6ad8f | 3958 | struct path path; |
f46d3567 | 3959 | unsigned int lookup_flags = 0; |
1da177e4 LT |
3960 | |
3961 | from = getname(oldname); | |
2ad94ae6 | 3962 | if (IS_ERR(from)) |
1da177e4 | 3963 | return PTR_ERR(from); |
f46d3567 JL |
3964 | retry: |
3965 | dentry = user_path_create(newdfd, newname, &path, lookup_flags); | |
6902d925 DH |
3966 | error = PTR_ERR(dentry); |
3967 | if (IS_ERR(dentry)) | |
dae6ad8f | 3968 | goto out_putname; |
6902d925 | 3969 | |
91a27b2a | 3970 | error = security_path_symlink(&path, dentry, from->name); |
a8104a9f | 3971 | if (!error) |
91a27b2a | 3972 | error = vfs_symlink(path.dentry->d_inode, dentry, from->name); |
921a1650 | 3973 | done_path_create(&path, dentry); |
f46d3567 JL |
3974 | if (retry_estale(error, lookup_flags)) { |
3975 | lookup_flags |= LOOKUP_REVAL; | |
3976 | goto retry; | |
3977 | } | |
6902d925 | 3978 | out_putname: |
1da177e4 LT |
3979 | putname(from); |
3980 | return error; | |
3981 | } | |
3982 | ||
3480b257 | 3983 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3984 | { |
3985 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
3986 | } | |
3987 | ||
146a8595 BF |
3988 | /** |
3989 | * vfs_link - create a new link | |
3990 | * @old_dentry: object to be linked | |
3991 | * @dir: new parent | |
3992 | * @new_dentry: where to create the new link | |
3993 | * @delegated_inode: returns inode needing a delegation break | |
3994 | * | |
3995 | * The caller must hold dir->i_mutex | |
3996 | * | |
3997 | * If vfs_link discovers a delegation on the to-be-linked file in need | |
3998 | * of breaking, it will return -EWOULDBLOCK and return a reference to the | |
3999 | * inode in delegated_inode. The caller should then break the delegation | |
4000 | * and retry. Because breaking a delegation may take a long time, the | |
4001 | * caller should drop the i_mutex before doing so. | |
4002 | * | |
4003 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4004 | * be appropriate for callers that expect the underlying filesystem not | |
4005 | * to be NFS exported. | |
4006 | */ | |
4007 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, struct inode **delegated_inode) | |
1da177e4 LT |
4008 | { |
4009 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 4010 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
4011 | int error; |
4012 | ||
4013 | if (!inode) | |
4014 | return -ENOENT; | |
4015 | ||
a95164d9 | 4016 | error = may_create(dir, new_dentry); |
1da177e4 LT |
4017 | if (error) |
4018 | return error; | |
4019 | ||
4020 | if (dir->i_sb != inode->i_sb) | |
4021 | return -EXDEV; | |
4022 | ||
4023 | /* | |
4024 | * A link to an append-only or immutable file cannot be created. | |
4025 | */ | |
4026 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
4027 | return -EPERM; | |
acfa4380 | 4028 | if (!dir->i_op->link) |
1da177e4 | 4029 | return -EPERM; |
7e79eedb | 4030 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
4031 | return -EPERM; |
4032 | ||
4033 | error = security_inode_link(old_dentry, dir, new_dentry); | |
4034 | if (error) | |
4035 | return error; | |
4036 | ||
7e79eedb | 4037 | mutex_lock(&inode->i_mutex); |
aae8a97d | 4038 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
f4e0c30c | 4039 | if (inode->i_nlink == 0 && !(inode->i_state & I_LINKABLE)) |
aae8a97d | 4040 | error = -ENOENT; |
8de52778 AV |
4041 | else if (max_links && inode->i_nlink >= max_links) |
4042 | error = -EMLINK; | |
146a8595 BF |
4043 | else { |
4044 | error = try_break_deleg(inode, delegated_inode); | |
4045 | if (!error) | |
4046 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
4047 | } | |
f4e0c30c AV |
4048 | |
4049 | if (!error && (inode->i_state & I_LINKABLE)) { | |
4050 | spin_lock(&inode->i_lock); | |
4051 | inode->i_state &= ~I_LINKABLE; | |
4052 | spin_unlock(&inode->i_lock); | |
4053 | } | |
7e79eedb | 4054 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 4055 | if (!error) |
7e79eedb | 4056 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
4057 | return error; |
4058 | } | |
4d359507 | 4059 | EXPORT_SYMBOL(vfs_link); |
1da177e4 LT |
4060 | |
4061 | /* | |
4062 | * Hardlinks are often used in delicate situations. We avoid | |
4063 | * security-related surprises by not following symlinks on the | |
4064 | * newname. --KAB | |
4065 | * | |
4066 | * We don't follow them on the oldname either to be compatible | |
4067 | * with linux 2.0, and to avoid hard-linking to directories | |
4068 | * and other special files. --ADM | |
4069 | */ | |
2e4d0924 HC |
4070 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
4071 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
4072 | { |
4073 | struct dentry *new_dentry; | |
dae6ad8f | 4074 | struct path old_path, new_path; |
146a8595 | 4075 | struct inode *delegated_inode = NULL; |
11a7b371 | 4076 | int how = 0; |
1da177e4 | 4077 | int error; |
1da177e4 | 4078 | |
11a7b371 | 4079 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 4080 | return -EINVAL; |
11a7b371 | 4081 | /* |
f0cc6ffb LT |
4082 | * To use null names we require CAP_DAC_READ_SEARCH |
4083 | * This ensures that not everyone will be able to create | |
4084 | * handlink using the passed filedescriptor. | |
11a7b371 | 4085 | */ |
f0cc6ffb LT |
4086 | if (flags & AT_EMPTY_PATH) { |
4087 | if (!capable(CAP_DAC_READ_SEARCH)) | |
4088 | return -ENOENT; | |
11a7b371 | 4089 | how = LOOKUP_EMPTY; |
f0cc6ffb | 4090 | } |
11a7b371 AK |
4091 | |
4092 | if (flags & AT_SYMLINK_FOLLOW) | |
4093 | how |= LOOKUP_FOLLOW; | |
442e31ca | 4094 | retry: |
11a7b371 | 4095 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 4096 | if (error) |
2ad94ae6 AV |
4097 | return error; |
4098 | ||
442e31ca JL |
4099 | new_dentry = user_path_create(newdfd, newname, &new_path, |
4100 | (how & LOOKUP_REVAL)); | |
1da177e4 | 4101 | error = PTR_ERR(new_dentry); |
6902d925 | 4102 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
4103 | goto out; |
4104 | ||
4105 | error = -EXDEV; | |
4106 | if (old_path.mnt != new_path.mnt) | |
4107 | goto out_dput; | |
800179c9 KC |
4108 | error = may_linkat(&old_path); |
4109 | if (unlikely(error)) | |
4110 | goto out_dput; | |
dae6ad8f | 4111 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 4112 | if (error) |
a8104a9f | 4113 | goto out_dput; |
146a8595 | 4114 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry, &delegated_inode); |
75c3f29d | 4115 | out_dput: |
921a1650 | 4116 | done_path_create(&new_path, new_dentry); |
146a8595 BF |
4117 | if (delegated_inode) { |
4118 | error = break_deleg_wait(&delegated_inode); | |
d22e6338 OD |
4119 | if (!error) { |
4120 | path_put(&old_path); | |
146a8595 | 4121 | goto retry; |
d22e6338 | 4122 | } |
146a8595 | 4123 | } |
442e31ca | 4124 | if (retry_estale(error, how)) { |
d22e6338 | 4125 | path_put(&old_path); |
442e31ca JL |
4126 | how |= LOOKUP_REVAL; |
4127 | goto retry; | |
4128 | } | |
1da177e4 | 4129 | out: |
2d8f3038 | 4130 | path_put(&old_path); |
1da177e4 LT |
4131 | |
4132 | return error; | |
4133 | } | |
4134 | ||
3480b257 | 4135 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4136 | { |
c04030e1 | 4137 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4138 | } |
4139 | ||
bc27027a MS |
4140 | /** |
4141 | * vfs_rename - rename a filesystem object | |
4142 | * @old_dir: parent of source | |
4143 | * @old_dentry: source | |
4144 | * @new_dir: parent of destination | |
4145 | * @new_dentry: destination | |
4146 | * @delegated_inode: returns an inode needing a delegation break | |
520c8b16 | 4147 | * @flags: rename flags |
bc27027a MS |
4148 | * |
4149 | * The caller must hold multiple mutexes--see lock_rename()). | |
4150 | * | |
4151 | * If vfs_rename discovers a delegation in need of breaking at either | |
4152 | * the source or destination, it will return -EWOULDBLOCK and return a | |
4153 | * reference to the inode in delegated_inode. The caller should then | |
4154 | * break the delegation and retry. Because breaking a delegation may | |
4155 | * take a long time, the caller should drop all locks before doing | |
4156 | * so. | |
4157 | * | |
4158 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4159 | * be appropriate for callers that expect the underlying filesystem not | |
4160 | * to be NFS exported. | |
4161 | * | |
1da177e4 LT |
4162 | * The worst of all namespace operations - renaming directory. "Perverted" |
4163 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
4164 | * Problems: | |
d03b29a2 | 4165 | * a) we can get into loop creation. |
1da177e4 LT |
4166 | * b) race potential - two innocent renames can create a loop together. |
4167 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 4168 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 | 4169 | * story. |
6cedba89 BF |
4170 | * c) we have to lock _four_ objects - parents and victim (if it exists), |
4171 | * and source (if it is not a directory). | |
1b1dcc1b | 4172 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
4173 | * whether the target exists). Solution: try to be smart with locking |
4174 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 4175 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
4176 | * move will be locked. Thus we can rank directories by the tree |
4177 | * (ancestors first) and rank all non-directories after them. | |
4178 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 4179 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
4180 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
4181 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
4182 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 4183 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 4184 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 4185 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 4186 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
4187 | * locking]. |
4188 | */ | |
bc27027a MS |
4189 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, |
4190 | struct inode *new_dir, struct dentry *new_dentry, | |
520c8b16 | 4191 | struct inode **delegated_inode, unsigned int flags) |
1da177e4 | 4192 | { |
bc27027a MS |
4193 | int error; |
4194 | bool is_dir = d_is_dir(old_dentry); | |
4195 | const unsigned char *old_name; | |
4196 | struct inode *source = old_dentry->d_inode; | |
9055cba7 | 4197 | struct inode *target = new_dentry->d_inode; |
da1ce067 MS |
4198 | bool new_is_dir = false; |
4199 | unsigned max_links = new_dir->i_sb->s_max_links; | |
bc27027a MS |
4200 | |
4201 | if (source == target) | |
4202 | return 0; | |
4203 | ||
4204 | error = may_delete(old_dir, old_dentry, is_dir); | |
4205 | if (error) | |
4206 | return error; | |
4207 | ||
da1ce067 | 4208 | if (!target) { |
bc27027a | 4209 | error = may_create(new_dir, new_dentry); |
da1ce067 MS |
4210 | } else { |
4211 | new_is_dir = d_is_dir(new_dentry); | |
4212 | ||
4213 | if (!(flags & RENAME_EXCHANGE)) | |
4214 | error = may_delete(new_dir, new_dentry, is_dir); | |
4215 | else | |
4216 | error = may_delete(new_dir, new_dentry, new_is_dir); | |
4217 | } | |
bc27027a MS |
4218 | if (error) |
4219 | return error; | |
4220 | ||
7177a9c4 | 4221 | if (!old_dir->i_op->rename && !old_dir->i_op->rename2) |
bc27027a | 4222 | return -EPERM; |
1da177e4 | 4223 | |
520c8b16 MS |
4224 | if (flags && !old_dir->i_op->rename2) |
4225 | return -EINVAL; | |
4226 | ||
1da177e4 LT |
4227 | /* |
4228 | * If we are going to change the parent - check write permissions, | |
4229 | * we'll need to flip '..'. | |
4230 | */ | |
da1ce067 MS |
4231 | if (new_dir != old_dir) { |
4232 | if (is_dir) { | |
4233 | error = inode_permission(source, MAY_WRITE); | |
4234 | if (error) | |
4235 | return error; | |
4236 | } | |
4237 | if ((flags & RENAME_EXCHANGE) && new_is_dir) { | |
4238 | error = inode_permission(target, MAY_WRITE); | |
4239 | if (error) | |
4240 | return error; | |
4241 | } | |
1da177e4 LT |
4242 | } |
4243 | ||
0b3974eb MS |
4244 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry, |
4245 | flags); | |
1da177e4 LT |
4246 | if (error) |
4247 | return error; | |
4248 | ||
bc27027a | 4249 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
1d2ef590 | 4250 | dget(new_dentry); |
da1ce067 | 4251 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4252 | lock_two_nondirectories(source, target); |
4253 | else if (target) | |
1b1dcc1b | 4254 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
4255 | |
4256 | error = -EBUSY; | |
7af1364f | 4257 | if (is_local_mountpoint(old_dentry) || is_local_mountpoint(new_dentry)) |
9055cba7 SW |
4258 | goto out; |
4259 | ||
da1ce067 | 4260 | if (max_links && new_dir != old_dir) { |
bc27027a | 4261 | error = -EMLINK; |
da1ce067 | 4262 | if (is_dir && !new_is_dir && new_dir->i_nlink >= max_links) |
bc27027a | 4263 | goto out; |
da1ce067 MS |
4264 | if ((flags & RENAME_EXCHANGE) && !is_dir && new_is_dir && |
4265 | old_dir->i_nlink >= max_links) | |
4266 | goto out; | |
4267 | } | |
4268 | if (is_dir && !(flags & RENAME_EXCHANGE) && target) | |
4269 | shrink_dcache_parent(new_dentry); | |
4270 | if (!is_dir) { | |
bc27027a | 4271 | error = try_break_deleg(source, delegated_inode); |
8e6d782c BF |
4272 | if (error) |
4273 | goto out; | |
da1ce067 MS |
4274 | } |
4275 | if (target && !new_is_dir) { | |
4276 | error = try_break_deleg(target, delegated_inode); | |
4277 | if (error) | |
4278 | goto out; | |
8e6d782c | 4279 | } |
7177a9c4 | 4280 | if (!old_dir->i_op->rename2) { |
520c8b16 MS |
4281 | error = old_dir->i_op->rename(old_dir, old_dentry, |
4282 | new_dir, new_dentry); | |
4283 | } else { | |
7177a9c4 | 4284 | WARN_ON(old_dir->i_op->rename != NULL); |
520c8b16 MS |
4285 | error = old_dir->i_op->rename2(old_dir, old_dentry, |
4286 | new_dir, new_dentry, flags); | |
4287 | } | |
51892bbb SW |
4288 | if (error) |
4289 | goto out; | |
4290 | ||
da1ce067 | 4291 | if (!(flags & RENAME_EXCHANGE) && target) { |
bc27027a MS |
4292 | if (is_dir) |
4293 | target->i_flags |= S_DEAD; | |
51892bbb | 4294 | dont_mount(new_dentry); |
8ed936b5 | 4295 | detach_mounts(new_dentry); |
bc27027a | 4296 | } |
da1ce067 MS |
4297 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) { |
4298 | if (!(flags & RENAME_EXCHANGE)) | |
4299 | d_move(old_dentry, new_dentry); | |
4300 | else | |
4301 | d_exchange(old_dentry, new_dentry); | |
4302 | } | |
51892bbb | 4303 | out: |
da1ce067 | 4304 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4305 | unlock_two_nondirectories(source, target); |
4306 | else if (target) | |
4307 | mutex_unlock(&target->i_mutex); | |
1da177e4 | 4308 | dput(new_dentry); |
da1ce067 | 4309 | if (!error) { |
123df294 | 4310 | fsnotify_move(old_dir, new_dir, old_name, is_dir, |
da1ce067 MS |
4311 | !(flags & RENAME_EXCHANGE) ? target : NULL, old_dentry); |
4312 | if (flags & RENAME_EXCHANGE) { | |
4313 | fsnotify_move(new_dir, old_dir, old_dentry->d_name.name, | |
4314 | new_is_dir, NULL, new_dentry); | |
4315 | } | |
4316 | } | |
0eeca283 RL |
4317 | fsnotify_oldname_free(old_name); |
4318 | ||
1da177e4 LT |
4319 | return error; |
4320 | } | |
4d359507 | 4321 | EXPORT_SYMBOL(vfs_rename); |
1da177e4 | 4322 | |
520c8b16 MS |
4323 | SYSCALL_DEFINE5(renameat2, int, olddfd, const char __user *, oldname, |
4324 | int, newdfd, const char __user *, newname, unsigned int, flags) | |
1da177e4 | 4325 | { |
2ad94ae6 AV |
4326 | struct dentry *old_dentry, *new_dentry; |
4327 | struct dentry *trap; | |
f5beed75 AV |
4328 | struct path old_path, new_path; |
4329 | struct qstr old_last, new_last; | |
4330 | int old_type, new_type; | |
8e6d782c | 4331 | struct inode *delegated_inode = NULL; |
91a27b2a JL |
4332 | struct filename *from; |
4333 | struct filename *to; | |
f5beed75 | 4334 | unsigned int lookup_flags = 0, target_flags = LOOKUP_RENAME_TARGET; |
c6a94284 | 4335 | bool should_retry = false; |
2ad94ae6 | 4336 | int error; |
520c8b16 | 4337 | |
0d7a8555 | 4338 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
da1ce067 MS |
4339 | return -EINVAL; |
4340 | ||
0d7a8555 MS |
4341 | if ((flags & (RENAME_NOREPLACE | RENAME_WHITEOUT)) && |
4342 | (flags & RENAME_EXCHANGE)) | |
520c8b16 MS |
4343 | return -EINVAL; |
4344 | ||
0d7a8555 MS |
4345 | if ((flags & RENAME_WHITEOUT) && !capable(CAP_MKNOD)) |
4346 | return -EPERM; | |
4347 | ||
f5beed75 AV |
4348 | if (flags & RENAME_EXCHANGE) |
4349 | target_flags = 0; | |
4350 | ||
c6a94284 | 4351 | retry: |
f5beed75 AV |
4352 | from = user_path_parent(olddfd, oldname, |
4353 | &old_path, &old_last, &old_type, lookup_flags); | |
91a27b2a JL |
4354 | if (IS_ERR(from)) { |
4355 | error = PTR_ERR(from); | |
1da177e4 | 4356 | goto exit; |
91a27b2a | 4357 | } |
1da177e4 | 4358 | |
f5beed75 AV |
4359 | to = user_path_parent(newdfd, newname, |
4360 | &new_path, &new_last, &new_type, lookup_flags); | |
91a27b2a JL |
4361 | if (IS_ERR(to)) { |
4362 | error = PTR_ERR(to); | |
1da177e4 | 4363 | goto exit1; |
91a27b2a | 4364 | } |
1da177e4 LT |
4365 | |
4366 | error = -EXDEV; | |
f5beed75 | 4367 | if (old_path.mnt != new_path.mnt) |
1da177e4 LT |
4368 | goto exit2; |
4369 | ||
1da177e4 | 4370 | error = -EBUSY; |
f5beed75 | 4371 | if (old_type != LAST_NORM) |
1da177e4 LT |
4372 | goto exit2; |
4373 | ||
0a7c3937 MS |
4374 | if (flags & RENAME_NOREPLACE) |
4375 | error = -EEXIST; | |
f5beed75 | 4376 | if (new_type != LAST_NORM) |
1da177e4 LT |
4377 | goto exit2; |
4378 | ||
f5beed75 | 4379 | error = mnt_want_write(old_path.mnt); |
c30dabfe JK |
4380 | if (error) |
4381 | goto exit2; | |
4382 | ||
8e6d782c | 4383 | retry_deleg: |
f5beed75 | 4384 | trap = lock_rename(new_path.dentry, old_path.dentry); |
1da177e4 | 4385 | |
f5beed75 | 4386 | old_dentry = __lookup_hash(&old_last, old_path.dentry, lookup_flags); |
1da177e4 LT |
4387 | error = PTR_ERR(old_dentry); |
4388 | if (IS_ERR(old_dentry)) | |
4389 | goto exit3; | |
4390 | /* source must exist */ | |
4391 | error = -ENOENT; | |
b18825a7 | 4392 | if (d_is_negative(old_dentry)) |
1da177e4 | 4393 | goto exit4; |
f5beed75 | 4394 | new_dentry = __lookup_hash(&new_last, new_path.dentry, lookup_flags | target_flags); |
0a7c3937 MS |
4395 | error = PTR_ERR(new_dentry); |
4396 | if (IS_ERR(new_dentry)) | |
4397 | goto exit4; | |
4398 | error = -EEXIST; | |
4399 | if ((flags & RENAME_NOREPLACE) && d_is_positive(new_dentry)) | |
4400 | goto exit5; | |
da1ce067 MS |
4401 | if (flags & RENAME_EXCHANGE) { |
4402 | error = -ENOENT; | |
4403 | if (d_is_negative(new_dentry)) | |
4404 | goto exit5; | |
4405 | ||
4406 | if (!d_is_dir(new_dentry)) { | |
4407 | error = -ENOTDIR; | |
f5beed75 | 4408 | if (new_last.name[new_last.len]) |
da1ce067 MS |
4409 | goto exit5; |
4410 | } | |
4411 | } | |
1da177e4 | 4412 | /* unless the source is a directory trailing slashes give -ENOTDIR */ |
44b1d530 | 4413 | if (!d_is_dir(old_dentry)) { |
1da177e4 | 4414 | error = -ENOTDIR; |
f5beed75 | 4415 | if (old_last.name[old_last.len]) |
0a7c3937 | 4416 | goto exit5; |
f5beed75 | 4417 | if (!(flags & RENAME_EXCHANGE) && new_last.name[new_last.len]) |
0a7c3937 | 4418 | goto exit5; |
1da177e4 LT |
4419 | } |
4420 | /* source should not be ancestor of target */ | |
4421 | error = -EINVAL; | |
4422 | if (old_dentry == trap) | |
0a7c3937 | 4423 | goto exit5; |
1da177e4 | 4424 | /* target should not be an ancestor of source */ |
da1ce067 MS |
4425 | if (!(flags & RENAME_EXCHANGE)) |
4426 | error = -ENOTEMPTY; | |
1da177e4 LT |
4427 | if (new_dentry == trap) |
4428 | goto exit5; | |
4429 | ||
f5beed75 AV |
4430 | error = security_path_rename(&old_path, old_dentry, |
4431 | &new_path, new_dentry, flags); | |
be6d3e56 | 4432 | if (error) |
c30dabfe | 4433 | goto exit5; |
f5beed75 AV |
4434 | error = vfs_rename(old_path.dentry->d_inode, old_dentry, |
4435 | new_path.dentry->d_inode, new_dentry, | |
520c8b16 | 4436 | &delegated_inode, flags); |
1da177e4 LT |
4437 | exit5: |
4438 | dput(new_dentry); | |
4439 | exit4: | |
4440 | dput(old_dentry); | |
4441 | exit3: | |
f5beed75 | 4442 | unlock_rename(new_path.dentry, old_path.dentry); |
8e6d782c BF |
4443 | if (delegated_inode) { |
4444 | error = break_deleg_wait(&delegated_inode); | |
4445 | if (!error) | |
4446 | goto retry_deleg; | |
4447 | } | |
f5beed75 | 4448 | mnt_drop_write(old_path.mnt); |
1da177e4 | 4449 | exit2: |
c6a94284 JL |
4450 | if (retry_estale(error, lookup_flags)) |
4451 | should_retry = true; | |
f5beed75 | 4452 | path_put(&new_path); |
2ad94ae6 | 4453 | putname(to); |
1da177e4 | 4454 | exit1: |
f5beed75 | 4455 | path_put(&old_path); |
1da177e4 | 4456 | putname(from); |
c6a94284 JL |
4457 | if (should_retry) { |
4458 | should_retry = false; | |
4459 | lookup_flags |= LOOKUP_REVAL; | |
4460 | goto retry; | |
4461 | } | |
2ad94ae6 | 4462 | exit: |
1da177e4 LT |
4463 | return error; |
4464 | } | |
4465 | ||
520c8b16 MS |
4466 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
4467 | int, newdfd, const char __user *, newname) | |
4468 | { | |
4469 | return sys_renameat2(olddfd, oldname, newdfd, newname, 0); | |
4470 | } | |
4471 | ||
a26eab24 | 4472 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4473 | { |
520c8b16 | 4474 | return sys_renameat2(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4475 | } |
4476 | ||
787fb6bc MS |
4477 | int vfs_whiteout(struct inode *dir, struct dentry *dentry) |
4478 | { | |
4479 | int error = may_create(dir, dentry); | |
4480 | if (error) | |
4481 | return error; | |
4482 | ||
4483 | if (!dir->i_op->mknod) | |
4484 | return -EPERM; | |
4485 | ||
4486 | return dir->i_op->mknod(dir, dentry, | |
4487 | S_IFCHR | WHITEOUT_MODE, WHITEOUT_DEV); | |
4488 | } | |
4489 | EXPORT_SYMBOL(vfs_whiteout); | |
4490 | ||
5d826c84 | 4491 | int readlink_copy(char __user *buffer, int buflen, const char *link) |
1da177e4 | 4492 | { |
5d826c84 | 4493 | int len = PTR_ERR(link); |
1da177e4 LT |
4494 | if (IS_ERR(link)) |
4495 | goto out; | |
4496 | ||
4497 | len = strlen(link); | |
4498 | if (len > (unsigned) buflen) | |
4499 | len = buflen; | |
4500 | if (copy_to_user(buffer, link, len)) | |
4501 | len = -EFAULT; | |
4502 | out: | |
4503 | return len; | |
4504 | } | |
5d826c84 | 4505 | EXPORT_SYMBOL(readlink_copy); |
1da177e4 LT |
4506 | |
4507 | /* | |
4508 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
4509 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
4510 | * using) it for any given inode is up to filesystem. | |
4511 | */ | |
4512 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4513 | { | |
cc314eef | 4514 | void *cookie; |
5f2c4179 AV |
4515 | struct inode *inode = d_inode(dentry); |
4516 | const char *link = inode->i_link; | |
694a1764 | 4517 | int res; |
cc314eef | 4518 | |
d4dee48b | 4519 | if (!link) { |
5f2c4179 | 4520 | link = inode->i_op->follow_link(dentry, &cookie); |
d4dee48b AV |
4521 | if (IS_ERR(link)) |
4522 | return PTR_ERR(link); | |
4523 | } | |
680baacb | 4524 | res = readlink_copy(buffer, buflen, link); |
5f2c4179 AV |
4525 | if (inode->i_op->put_link) |
4526 | inode->i_op->put_link(inode, cookie); | |
694a1764 | 4527 | return res; |
1da177e4 | 4528 | } |
4d359507 | 4529 | EXPORT_SYMBOL(generic_readlink); |
1da177e4 | 4530 | |
1da177e4 LT |
4531 | /* get the link contents into pagecache */ |
4532 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
4533 | { | |
ebd09abb DG |
4534 | char *kaddr; |
4535 | struct page *page; | |
1da177e4 | 4536 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 4537 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 4538 | if (IS_ERR(page)) |
6fe6900e | 4539 | return (char*)page; |
1da177e4 | 4540 | *ppage = page; |
ebd09abb DG |
4541 | kaddr = kmap(page); |
4542 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
4543 | return kaddr; | |
1da177e4 LT |
4544 | } |
4545 | ||
4546 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4547 | { | |
4548 | struct page *page = NULL; | |
5d826c84 | 4549 | int res = readlink_copy(buffer, buflen, page_getlink(dentry, &page)); |
1da177e4 LT |
4550 | if (page) { |
4551 | kunmap(page); | |
4552 | page_cache_release(page); | |
4553 | } | |
4554 | return res; | |
4555 | } | |
4d359507 | 4556 | EXPORT_SYMBOL(page_readlink); |
1da177e4 | 4557 | |
6e77137b | 4558 | const char *page_follow_link_light(struct dentry *dentry, void **cookie) |
1da177e4 | 4559 | { |
cc314eef | 4560 | struct page *page = NULL; |
680baacb AV |
4561 | char *res = page_getlink(dentry, &page); |
4562 | if (!IS_ERR(res)) | |
4563 | *cookie = page; | |
4564 | return res; | |
1da177e4 | 4565 | } |
4d359507 | 4566 | EXPORT_SYMBOL(page_follow_link_light); |
1da177e4 | 4567 | |
5f2c4179 | 4568 | void page_put_link(struct inode *unused, void *cookie) |
1da177e4 | 4569 | { |
cc314eef | 4570 | struct page *page = cookie; |
680baacb AV |
4571 | kunmap(page); |
4572 | page_cache_release(page); | |
1da177e4 | 4573 | } |
4d359507 | 4574 | EXPORT_SYMBOL(page_put_link); |
1da177e4 | 4575 | |
54566b2c NP |
4576 | /* |
4577 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
4578 | */ | |
4579 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
4580 | { |
4581 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 4582 | struct page *page; |
afddba49 | 4583 | void *fsdata; |
beb497ab | 4584 | int err; |
1da177e4 | 4585 | char *kaddr; |
54566b2c NP |
4586 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
4587 | if (nofs) | |
4588 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 4589 | |
7e53cac4 | 4590 | retry: |
afddba49 | 4591 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 4592 | flags, &page, &fsdata); |
1da177e4 | 4593 | if (err) |
afddba49 NP |
4594 | goto fail; |
4595 | ||
e8e3c3d6 | 4596 | kaddr = kmap_atomic(page); |
1da177e4 | 4597 | memcpy(kaddr, symname, len-1); |
e8e3c3d6 | 4598 | kunmap_atomic(kaddr); |
afddba49 NP |
4599 | |
4600 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
4601 | page, fsdata); | |
1da177e4 LT |
4602 | if (err < 0) |
4603 | goto fail; | |
afddba49 NP |
4604 | if (err < len-1) |
4605 | goto retry; | |
4606 | ||
1da177e4 LT |
4607 | mark_inode_dirty(inode); |
4608 | return 0; | |
1da177e4 LT |
4609 | fail: |
4610 | return err; | |
4611 | } | |
4d359507 | 4612 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 | 4613 | |
0adb25d2 KK |
4614 | int page_symlink(struct inode *inode, const char *symname, int len) |
4615 | { | |
4616 | return __page_symlink(inode, symname, len, | |
54566b2c | 4617 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 | 4618 | } |
4d359507 | 4619 | EXPORT_SYMBOL(page_symlink); |
0adb25d2 | 4620 | |
92e1d5be | 4621 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
4622 | .readlink = generic_readlink, |
4623 | .follow_link = page_follow_link_light, | |
4624 | .put_link = page_put_link, | |
4625 | }; | |
1da177e4 | 4626 | EXPORT_SYMBOL(page_symlink_inode_operations); |