]>
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; |
894bc8c4 | 511 | } *stack, internal[EMBEDDED_LEVELS]; |
1f55a6ec AV |
512 | }; |
513 | ||
756daf26 | 514 | static struct nameidata *set_nameidata(struct nameidata *p) |
894bc8c4 | 515 | { |
756daf26 N |
516 | struct nameidata *old = current->nameidata; |
517 | p->stack = p->internal; | |
518 | p->total_link_count = old ? old->total_link_count : 0; | |
519 | current->nameidata = p; | |
520 | return old; | |
894bc8c4 AV |
521 | } |
522 | ||
756daf26 | 523 | static void restore_nameidata(struct nameidata *old) |
894bc8c4 | 524 | { |
756daf26 N |
525 | struct nameidata *now = current->nameidata; |
526 | ||
527 | current->nameidata = old; | |
528 | if (old) | |
529 | old->total_link_count = now->total_link_count; | |
530 | if (now->stack != now->internal) { | |
531 | kfree(now->stack); | |
532 | now->stack = now->internal; | |
894bc8c4 AV |
533 | } |
534 | } | |
535 | ||
536 | static int __nd_alloc_stack(struct nameidata *nd) | |
537 | { | |
e269f2a7 | 538 | struct saved *p = kmalloc(MAXSYMLINKS * sizeof(struct saved), |
894bc8c4 AV |
539 | GFP_KERNEL); |
540 | if (unlikely(!p)) | |
541 | return -ENOMEM; | |
542 | memcpy(p, nd->internal, sizeof(nd->internal)); | |
543 | nd->stack = p; | |
544 | return 0; | |
545 | } | |
546 | ||
547 | static inline int nd_alloc_stack(struct nameidata *nd) | |
548 | { | |
da4e0be0 | 549 | if (likely(nd->depth != EMBEDDED_LEVELS)) |
894bc8c4 AV |
550 | return 0; |
551 | if (likely(nd->stack != nd->internal)) | |
552 | return 0; | |
553 | return __nd_alloc_stack(nd); | |
554 | } | |
555 | ||
19660af7 | 556 | /* |
31e6b01f | 557 | * Path walking has 2 modes, rcu-walk and ref-walk (see |
19660af7 AV |
558 | * Documentation/filesystems/path-lookup.txt). In situations when we can't |
559 | * continue in RCU mode, we attempt to drop out of rcu-walk mode and grab | |
560 | * normal reference counts on dentries and vfsmounts to transition to rcu-walk | |
561 | * mode. Refcounts are grabbed at the last known good point before rcu-walk | |
562 | * got stuck, so ref-walk may continue from there. If this is not successful | |
563 | * (eg. a seqcount has changed), then failure is returned and it's up to caller | |
564 | * to restart the path walk from the beginning in ref-walk mode. | |
31e6b01f | 565 | */ |
31e6b01f NP |
566 | |
567 | /** | |
19660af7 AV |
568 | * unlazy_walk - try to switch to ref-walk mode. |
569 | * @nd: nameidata pathwalk data | |
570 | * @dentry: child of nd->path.dentry or NULL | |
6e9918b7 | 571 | * @seq: seq number to check dentry against |
39191628 | 572 | * Returns: 0 on success, -ECHILD on failure |
31e6b01f | 573 | * |
19660af7 AV |
574 | * unlazy_walk attempts to legitimize the current nd->path, nd->root and dentry |
575 | * for ref-walk mode. @dentry must be a path found by a do_lookup call on | |
576 | * @nd or NULL. Must be called from rcu-walk context. | |
31e6b01f | 577 | */ |
6e9918b7 | 578 | static int unlazy_walk(struct nameidata *nd, struct dentry *dentry, unsigned seq) |
31e6b01f NP |
579 | { |
580 | struct fs_struct *fs = current->fs; | |
581 | struct dentry *parent = nd->path.dentry; | |
582 | ||
583 | BUG_ON(!(nd->flags & LOOKUP_RCU)); | |
e5c832d5 LT |
584 | |
585 | /* | |
48a066e7 AV |
586 | * After legitimizing the bastards, terminate_walk() |
587 | * will do the right thing for non-RCU mode, and all our | |
588 | * subsequent exit cases should rcu_read_unlock() | |
589 | * before returning. Do vfsmount first; if dentry | |
590 | * can't be legitimized, just set nd->path.dentry to NULL | |
591 | * and rely on dput(NULL) being a no-op. | |
e5c832d5 | 592 | */ |
48a066e7 | 593 | if (!legitimize_mnt(nd->path.mnt, nd->m_seq)) |
e5c832d5 | 594 | return -ECHILD; |
e5c832d5 | 595 | nd->flags &= ~LOOKUP_RCU; |
15570086 | 596 | |
48a066e7 AV |
597 | if (!lockref_get_not_dead(&parent->d_lockref)) { |
598 | nd->path.dentry = NULL; | |
d870b4a1 | 599 | goto out; |
48a066e7 AV |
600 | } |
601 | ||
15570086 LT |
602 | /* |
603 | * For a negative lookup, the lookup sequence point is the parents | |
604 | * sequence point, and it only needs to revalidate the parent dentry. | |
605 | * | |
606 | * For a positive lookup, we need to move both the parent and the | |
607 | * dentry from the RCU domain to be properly refcounted. And the | |
608 | * sequence number in the dentry validates *both* dentry counters, | |
609 | * since we checked the sequence number of the parent after we got | |
610 | * the child sequence number. So we know the parent must still | |
611 | * be valid if the child sequence number is still valid. | |
612 | */ | |
19660af7 | 613 | if (!dentry) { |
e5c832d5 LT |
614 | if (read_seqcount_retry(&parent->d_seq, nd->seq)) |
615 | goto out; | |
19660af7 AV |
616 | BUG_ON(nd->inode != parent->d_inode); |
617 | } else { | |
e5c832d5 LT |
618 | if (!lockref_get_not_dead(&dentry->d_lockref)) |
619 | goto out; | |
6e9918b7 | 620 | if (read_seqcount_retry(&dentry->d_seq, seq)) |
e5c832d5 | 621 | goto drop_dentry; |
19660af7 | 622 | } |
e5c832d5 LT |
623 | |
624 | /* | |
625 | * Sequence counts matched. Now make sure that the root is | |
626 | * still valid and get it if required. | |
627 | */ | |
628 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
629 | spin_lock(&fs->lock); | |
630 | if (nd->root.mnt != fs->root.mnt || nd->root.dentry != fs->root.dentry) | |
631 | goto unlock_and_drop_dentry; | |
31e6b01f NP |
632 | path_get(&nd->root); |
633 | spin_unlock(&fs->lock); | |
634 | } | |
31e6b01f | 635 | |
8b61e74f | 636 | rcu_read_unlock(); |
31e6b01f | 637 | return 0; |
19660af7 | 638 | |
e5c832d5 LT |
639 | unlock_and_drop_dentry: |
640 | spin_unlock(&fs->lock); | |
641 | drop_dentry: | |
8b61e74f | 642 | rcu_read_unlock(); |
15570086 | 643 | dput(dentry); |
d0d27277 | 644 | goto drop_root_mnt; |
e5c832d5 | 645 | out: |
8b61e74f | 646 | rcu_read_unlock(); |
d0d27277 LT |
647 | drop_root_mnt: |
648 | if (!(nd->flags & LOOKUP_ROOT)) | |
649 | nd->root.mnt = NULL; | |
31e6b01f NP |
650 | return -ECHILD; |
651 | } | |
652 | ||
4ce16ef3 | 653 | static inline int d_revalidate(struct dentry *dentry, unsigned int flags) |
34286d66 | 654 | { |
4ce16ef3 | 655 | return dentry->d_op->d_revalidate(dentry, flags); |
34286d66 NP |
656 | } |
657 | ||
9f1fafee AV |
658 | /** |
659 | * complete_walk - successful completion of path walk | |
660 | * @nd: pointer nameidata | |
39159de2 | 661 | * |
9f1fafee AV |
662 | * If we had been in RCU mode, drop out of it and legitimize nd->path. |
663 | * Revalidate the final result, unless we'd already done that during | |
664 | * the path walk or the filesystem doesn't ask for it. Return 0 on | |
665 | * success, -error on failure. In case of failure caller does not | |
666 | * need to drop nd->path. | |
39159de2 | 667 | */ |
9f1fafee | 668 | static int complete_walk(struct nameidata *nd) |
39159de2 | 669 | { |
16c2cd71 | 670 | struct dentry *dentry = nd->path.dentry; |
39159de2 | 671 | int status; |
39159de2 | 672 | |
9f1fafee | 673 | if (nd->flags & LOOKUP_RCU) { |
9f1fafee AV |
674 | if (!(nd->flags & LOOKUP_ROOT)) |
675 | nd->root.mnt = NULL; | |
6e9918b7 | 676 | if (unlikely(unlazy_walk(nd, NULL, 0))) |
9f1fafee | 677 | return -ECHILD; |
9f1fafee AV |
678 | } |
679 | ||
16c2cd71 AV |
680 | if (likely(!(nd->flags & LOOKUP_JUMPED))) |
681 | return 0; | |
682 | ||
ecf3d1f1 | 683 | if (likely(!(dentry->d_flags & DCACHE_OP_WEAK_REVALIDATE))) |
39159de2 JL |
684 | return 0; |
685 | ||
ecf3d1f1 | 686 | status = dentry->d_op->d_weak_revalidate(dentry, nd->flags); |
39159de2 JL |
687 | if (status > 0) |
688 | return 0; | |
689 | ||
16c2cd71 | 690 | if (!status) |
39159de2 | 691 | status = -ESTALE; |
16c2cd71 | 692 | |
39159de2 JL |
693 | return status; |
694 | } | |
695 | ||
2a737871 AV |
696 | static __always_inline void set_root(struct nameidata *nd) |
697 | { | |
7bd88377 | 698 | get_fs_root(current->fs, &nd->root); |
2a737871 AV |
699 | } |
700 | ||
7bd88377 | 701 | static __always_inline unsigned set_root_rcu(struct nameidata *nd) |
31e6b01f | 702 | { |
7bd88377 AV |
703 | struct fs_struct *fs = current->fs; |
704 | unsigned seq, res; | |
c28cc364 | 705 | |
7bd88377 AV |
706 | do { |
707 | seq = read_seqcount_begin(&fs->seq); | |
708 | nd->root = fs->root; | |
709 | res = __read_seqcount_begin(&nd->root.dentry->d_seq); | |
710 | } while (read_seqcount_retry(&fs->seq, seq)); | |
711 | return res; | |
31e6b01f NP |
712 | } |
713 | ||
1d957f9b | 714 | static void path_put_conditional(struct path *path, struct nameidata *nd) |
051d3812 IK |
715 | { |
716 | dput(path->dentry); | |
4ac91378 | 717 | if (path->mnt != nd->path.mnt) |
051d3812 IK |
718 | mntput(path->mnt); |
719 | } | |
720 | ||
7b9337aa NP |
721 | static inline void path_to_nameidata(const struct path *path, |
722 | struct nameidata *nd) | |
051d3812 | 723 | { |
31e6b01f NP |
724 | if (!(nd->flags & LOOKUP_RCU)) { |
725 | dput(nd->path.dentry); | |
726 | if (nd->path.mnt != path->mnt) | |
727 | mntput(nd->path.mnt); | |
9a229683 | 728 | } |
31e6b01f | 729 | nd->path.mnt = path->mnt; |
4ac91378 | 730 | nd->path.dentry = path->dentry; |
051d3812 IK |
731 | } |
732 | ||
b5fb63c1 CH |
733 | /* |
734 | * Helper to directly jump to a known parsed path from ->follow_link, | |
735 | * caller must have taken a reference to path beforehand. | |
736 | */ | |
6e77137b | 737 | void nd_jump_link(struct path *path) |
b5fb63c1 | 738 | { |
6e77137b | 739 | struct nameidata *nd = current->nameidata; |
b5fb63c1 CH |
740 | path_put(&nd->path); |
741 | ||
742 | nd->path = *path; | |
743 | nd->inode = nd->path.dentry->d_inode; | |
744 | nd->flags |= LOOKUP_JUMPED; | |
b5fb63c1 CH |
745 | } |
746 | ||
b9ff4429 | 747 | static inline void put_link(struct nameidata *nd) |
574197e0 | 748 | { |
21c3003d | 749 | struct saved *last = nd->stack + --nd->depth; |
237d8b32 | 750 | struct inode *inode = last->inode; |
b9ff4429 AV |
751 | if (last->cookie && inode->i_op->put_link) |
752 | inode->i_op->put_link(last->link.dentry, last->cookie); | |
753 | path_put(&last->link); | |
574197e0 AV |
754 | } |
755 | ||
561ec64a LT |
756 | int sysctl_protected_symlinks __read_mostly = 0; |
757 | int sysctl_protected_hardlinks __read_mostly = 0; | |
800179c9 KC |
758 | |
759 | /** | |
760 | * may_follow_link - Check symlink following for unsafe situations | |
55852635 | 761 | * @nd: nameidata pathwalk data |
800179c9 KC |
762 | * |
763 | * In the case of the sysctl_protected_symlinks sysctl being enabled, | |
764 | * CAP_DAC_OVERRIDE needs to be specifically ignored if the symlink is | |
765 | * in a sticky world-writable directory. This is to protect privileged | |
766 | * processes from failing races against path names that may change out | |
767 | * from under them by way of other users creating malicious symlinks. | |
768 | * It will permit symlinks to be followed only when outside a sticky | |
769 | * world-writable directory, or when the uid of the symlink and follower | |
770 | * match, or when the directory owner matches the symlink's owner. | |
771 | * | |
772 | * Returns 0 if following the symlink is allowed, -ve on error. | |
773 | */ | |
fec2fa24 | 774 | static inline int may_follow_link(struct nameidata *nd) |
800179c9 KC |
775 | { |
776 | const struct inode *inode; | |
777 | const struct inode *parent; | |
778 | ||
779 | if (!sysctl_protected_symlinks) | |
780 | return 0; | |
781 | ||
782 | /* Allowed if owner and follower match. */ | |
237d8b32 | 783 | inode = nd->stack[0].inode; |
81abe27b | 784 | if (uid_eq(current_cred()->fsuid, inode->i_uid)) |
800179c9 KC |
785 | return 0; |
786 | ||
787 | /* Allowed if parent directory not sticky and world-writable. */ | |
788 | parent = nd->path.dentry->d_inode; | |
789 | if ((parent->i_mode & (S_ISVTX|S_IWOTH)) != (S_ISVTX|S_IWOTH)) | |
790 | return 0; | |
791 | ||
792 | /* Allowed if parent directory and link owner match. */ | |
81abe27b | 793 | if (uid_eq(parent->i_uid, inode->i_uid)) |
800179c9 KC |
794 | return 0; |
795 | ||
1cf2665b | 796 | audit_log_link_denied("follow_link", &nd->stack[0].link); |
800179c9 KC |
797 | return -EACCES; |
798 | } | |
799 | ||
800 | /** | |
801 | * safe_hardlink_source - Check for safe hardlink conditions | |
802 | * @inode: the source inode to hardlink from | |
803 | * | |
804 | * Return false if at least one of the following conditions: | |
805 | * - inode is not a regular file | |
806 | * - inode is setuid | |
807 | * - inode is setgid and group-exec | |
808 | * - access failure for read and write | |
809 | * | |
810 | * Otherwise returns true. | |
811 | */ | |
812 | static bool safe_hardlink_source(struct inode *inode) | |
813 | { | |
814 | umode_t mode = inode->i_mode; | |
815 | ||
816 | /* Special files should not get pinned to the filesystem. */ | |
817 | if (!S_ISREG(mode)) | |
818 | return false; | |
819 | ||
820 | /* Setuid files should not get pinned to the filesystem. */ | |
821 | if (mode & S_ISUID) | |
822 | return false; | |
823 | ||
824 | /* Executable setgid files should not get pinned to the filesystem. */ | |
825 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) | |
826 | return false; | |
827 | ||
828 | /* Hardlinking to unreadable or unwritable sources is dangerous. */ | |
829 | if (inode_permission(inode, MAY_READ | MAY_WRITE)) | |
830 | return false; | |
831 | ||
832 | return true; | |
833 | } | |
834 | ||
835 | /** | |
836 | * may_linkat - Check permissions for creating a hardlink | |
837 | * @link: the source to hardlink from | |
838 | * | |
839 | * Block hardlink when all of: | |
840 | * - sysctl_protected_hardlinks enabled | |
841 | * - fsuid does not match inode | |
842 | * - hardlink source is unsafe (see safe_hardlink_source() above) | |
843 | * - not CAP_FOWNER | |
844 | * | |
845 | * Returns 0 if successful, -ve on error. | |
846 | */ | |
847 | static int may_linkat(struct path *link) | |
848 | { | |
849 | const struct cred *cred; | |
850 | struct inode *inode; | |
851 | ||
852 | if (!sysctl_protected_hardlinks) | |
853 | return 0; | |
854 | ||
855 | cred = current_cred(); | |
856 | inode = link->dentry->d_inode; | |
857 | ||
858 | /* Source inode owner (or CAP_FOWNER) can hardlink all they like, | |
859 | * otherwise, it must be a safe source. | |
860 | */ | |
81abe27b | 861 | if (uid_eq(cred->fsuid, inode->i_uid) || safe_hardlink_source(inode) || |
800179c9 KC |
862 | capable(CAP_FOWNER)) |
863 | return 0; | |
864 | ||
a51d9eaa | 865 | audit_log_link_denied("linkat", link); |
800179c9 KC |
866 | return -EPERM; |
867 | } | |
868 | ||
3b2e7f75 AV |
869 | static __always_inline |
870 | const char *get_link(struct nameidata *nd) | |
1da177e4 | 871 | { |
ab104923 | 872 | struct saved *last = nd->stack + nd->depth - 1; |
1cf2665b | 873 | struct dentry *dentry = last->link.dentry; |
237d8b32 | 874 | struct inode *inode = last->inode; |
6d7b5aae | 875 | int error; |
0a959df5 | 876 | const char *res; |
1da177e4 | 877 | |
844a3917 AV |
878 | BUG_ON(nd->flags & LOOKUP_RCU); |
879 | ||
574197e0 | 880 | cond_resched(); |
574197e0 | 881 | |
3b2e7f75 | 882 | touch_atime(&last->link); |
cd4e91d3 | 883 | |
37882db0 | 884 | error = security_inode_follow_link(dentry); |
6d7b5aae | 885 | if (error) |
6920a440 | 886 | return ERR_PTR(error); |
36f3b4f6 | 887 | |
86acdca1 | 888 | nd->last_type = LAST_BIND; |
d4dee48b AV |
889 | res = inode->i_link; |
890 | if (!res) { | |
6e77137b | 891 | res = inode->i_op->follow_link(dentry, &last->cookie); |
fab51e8a | 892 | if (IS_ERR_OR_NULL(res)) { |
6920a440 | 893 | last->cookie = NULL; |
fab51e8a AV |
894 | return res; |
895 | } | |
896 | } | |
897 | if (*res == '/') { | |
898 | if (!nd->root.mnt) | |
899 | set_root(nd); | |
900 | path_put(&nd->path); | |
901 | nd->path = nd->root; | |
902 | path_get(&nd->root); | |
903 | nd->inode = nd->path.dentry->d_inode; | |
904 | nd->flags |= LOOKUP_JUMPED; | |
905 | while (unlikely(*++res == '/')) | |
906 | ; | |
1da177e4 | 907 | } |
fab51e8a AV |
908 | if (!*res) |
909 | res = NULL; | |
0a959df5 AV |
910 | return res; |
911 | } | |
6d7b5aae | 912 | |
31e6b01f NP |
913 | static int follow_up_rcu(struct path *path) |
914 | { | |
0714a533 AV |
915 | struct mount *mnt = real_mount(path->mnt); |
916 | struct mount *parent; | |
31e6b01f NP |
917 | struct dentry *mountpoint; |
918 | ||
0714a533 AV |
919 | parent = mnt->mnt_parent; |
920 | if (&parent->mnt == path->mnt) | |
31e6b01f | 921 | return 0; |
a73324da | 922 | mountpoint = mnt->mnt_mountpoint; |
31e6b01f | 923 | path->dentry = mountpoint; |
0714a533 | 924 | path->mnt = &parent->mnt; |
31e6b01f NP |
925 | return 1; |
926 | } | |
927 | ||
f015f126 DH |
928 | /* |
929 | * follow_up - Find the mountpoint of path's vfsmount | |
930 | * | |
931 | * Given a path, find the mountpoint of its source file system. | |
932 | * Replace @path with the path of the mountpoint in the parent mount. | |
933 | * Up is towards /. | |
934 | * | |
935 | * Return 1 if we went up a level and 0 if we were already at the | |
936 | * root. | |
937 | */ | |
bab77ebf | 938 | int follow_up(struct path *path) |
1da177e4 | 939 | { |
0714a533 AV |
940 | struct mount *mnt = real_mount(path->mnt); |
941 | struct mount *parent; | |
1da177e4 | 942 | struct dentry *mountpoint; |
99b7db7b | 943 | |
48a066e7 | 944 | read_seqlock_excl(&mount_lock); |
0714a533 | 945 | parent = mnt->mnt_parent; |
3c0a6163 | 946 | if (parent == mnt) { |
48a066e7 | 947 | read_sequnlock_excl(&mount_lock); |
1da177e4 LT |
948 | return 0; |
949 | } | |
0714a533 | 950 | mntget(&parent->mnt); |
a73324da | 951 | mountpoint = dget(mnt->mnt_mountpoint); |
48a066e7 | 952 | read_sequnlock_excl(&mount_lock); |
bab77ebf AV |
953 | dput(path->dentry); |
954 | path->dentry = mountpoint; | |
955 | mntput(path->mnt); | |
0714a533 | 956 | path->mnt = &parent->mnt; |
1da177e4 LT |
957 | return 1; |
958 | } | |
4d359507 | 959 | EXPORT_SYMBOL(follow_up); |
1da177e4 | 960 | |
b5c84bf6 | 961 | /* |
9875cf80 DH |
962 | * Perform an automount |
963 | * - return -EISDIR to tell follow_managed() to stop and return the path we | |
964 | * were called with. | |
1da177e4 | 965 | */ |
756daf26 | 966 | static int follow_automount(struct path *path, struct nameidata *nd, |
9875cf80 | 967 | bool *need_mntput) |
31e6b01f | 968 | { |
9875cf80 | 969 | struct vfsmount *mnt; |
ea5b778a | 970 | int err; |
9875cf80 DH |
971 | |
972 | if (!path->dentry->d_op || !path->dentry->d_op->d_automount) | |
973 | return -EREMOTE; | |
974 | ||
0ec26fd0 MS |
975 | /* We don't want to mount if someone's just doing a stat - |
976 | * unless they're stat'ing a directory and appended a '/' to | |
977 | * the name. | |
978 | * | |
979 | * We do, however, want to mount if someone wants to open or | |
980 | * create a file of any type under the mountpoint, wants to | |
981 | * traverse through the mountpoint or wants to open the | |
982 | * mounted directory. Also, autofs may mark negative dentries | |
983 | * as being automount points. These will need the attentions | |
984 | * of the daemon to instantiate them before they can be used. | |
9875cf80 | 985 | */ |
756daf26 N |
986 | if (!(nd->flags & (LOOKUP_PARENT | LOOKUP_DIRECTORY | |
987 | LOOKUP_OPEN | LOOKUP_CREATE | LOOKUP_AUTOMOUNT)) && | |
0ec26fd0 MS |
988 | path->dentry->d_inode) |
989 | return -EISDIR; | |
990 | ||
756daf26 N |
991 | nd->total_link_count++; |
992 | if (nd->total_link_count >= 40) | |
9875cf80 DH |
993 | return -ELOOP; |
994 | ||
995 | mnt = path->dentry->d_op->d_automount(path); | |
996 | if (IS_ERR(mnt)) { | |
997 | /* | |
998 | * The filesystem is allowed to return -EISDIR here to indicate | |
999 | * it doesn't want to automount. For instance, autofs would do | |
1000 | * this so that its userspace daemon can mount on this dentry. | |
1001 | * | |
1002 | * However, we can only permit this if it's a terminal point in | |
1003 | * the path being looked up; if it wasn't then the remainder of | |
1004 | * the path is inaccessible and we should say so. | |
1005 | */ | |
756daf26 | 1006 | if (PTR_ERR(mnt) == -EISDIR && (nd->flags & LOOKUP_PARENT)) |
9875cf80 DH |
1007 | return -EREMOTE; |
1008 | return PTR_ERR(mnt); | |
31e6b01f | 1009 | } |
ea5b778a | 1010 | |
9875cf80 DH |
1011 | if (!mnt) /* mount collision */ |
1012 | return 0; | |
31e6b01f | 1013 | |
8aef1884 AV |
1014 | if (!*need_mntput) { |
1015 | /* lock_mount() may release path->mnt on error */ | |
1016 | mntget(path->mnt); | |
1017 | *need_mntput = true; | |
1018 | } | |
19a167af | 1019 | err = finish_automount(mnt, path); |
9875cf80 | 1020 | |
ea5b778a DH |
1021 | switch (err) { |
1022 | case -EBUSY: | |
1023 | /* Someone else made a mount here whilst we were busy */ | |
19a167af | 1024 | return 0; |
ea5b778a | 1025 | case 0: |
8aef1884 | 1026 | path_put(path); |
ea5b778a DH |
1027 | path->mnt = mnt; |
1028 | path->dentry = dget(mnt->mnt_root); | |
ea5b778a | 1029 | return 0; |
19a167af AV |
1030 | default: |
1031 | return err; | |
ea5b778a | 1032 | } |
19a167af | 1033 | |
463ffb2e AV |
1034 | } |
1035 | ||
9875cf80 DH |
1036 | /* |
1037 | * Handle a dentry that is managed in some way. | |
cc53ce53 | 1038 | * - Flagged for transit management (autofs) |
9875cf80 DH |
1039 | * - Flagged as mountpoint |
1040 | * - Flagged as automount point | |
1041 | * | |
1042 | * This may only be called in refwalk mode. | |
1043 | * | |
1044 | * Serialization is taken care of in namespace.c | |
1045 | */ | |
756daf26 | 1046 | static int follow_managed(struct path *path, struct nameidata *nd) |
1da177e4 | 1047 | { |
8aef1884 | 1048 | struct vfsmount *mnt = path->mnt; /* held by caller, must be left alone */ |
9875cf80 DH |
1049 | unsigned managed; |
1050 | bool need_mntput = false; | |
8aef1884 | 1051 | int ret = 0; |
9875cf80 DH |
1052 | |
1053 | /* Given that we're not holding a lock here, we retain the value in a | |
1054 | * local variable for each dentry as we look at it so that we don't see | |
1055 | * the components of that value change under us */ | |
1056 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1057 | managed &= DCACHE_MANAGED_DENTRY, | |
1058 | unlikely(managed != 0)) { | |
cc53ce53 DH |
1059 | /* Allow the filesystem to manage the transit without i_mutex |
1060 | * being held. */ | |
1061 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1062 | BUG_ON(!path->dentry->d_op); | |
1063 | BUG_ON(!path->dentry->d_op->d_manage); | |
1aed3e42 | 1064 | ret = path->dentry->d_op->d_manage(path->dentry, false); |
cc53ce53 | 1065 | if (ret < 0) |
8aef1884 | 1066 | break; |
cc53ce53 DH |
1067 | } |
1068 | ||
9875cf80 DH |
1069 | /* Transit to a mounted filesystem. */ |
1070 | if (managed & DCACHE_MOUNTED) { | |
1071 | struct vfsmount *mounted = lookup_mnt(path); | |
1072 | if (mounted) { | |
1073 | dput(path->dentry); | |
1074 | if (need_mntput) | |
1075 | mntput(path->mnt); | |
1076 | path->mnt = mounted; | |
1077 | path->dentry = dget(mounted->mnt_root); | |
1078 | need_mntput = true; | |
1079 | continue; | |
1080 | } | |
1081 | ||
1082 | /* Something is mounted on this dentry in another | |
1083 | * namespace and/or whatever was mounted there in this | |
48a066e7 AV |
1084 | * namespace got unmounted before lookup_mnt() could |
1085 | * get it */ | |
9875cf80 DH |
1086 | } |
1087 | ||
1088 | /* Handle an automount point */ | |
1089 | if (managed & DCACHE_NEED_AUTOMOUNT) { | |
756daf26 | 1090 | ret = follow_automount(path, nd, &need_mntput); |
9875cf80 | 1091 | if (ret < 0) |
8aef1884 | 1092 | break; |
9875cf80 DH |
1093 | continue; |
1094 | } | |
1095 | ||
1096 | /* We didn't change the current path point */ | |
1097 | break; | |
1da177e4 | 1098 | } |
8aef1884 AV |
1099 | |
1100 | if (need_mntput && path->mnt == mnt) | |
1101 | mntput(path->mnt); | |
1102 | if (ret == -EISDIR) | |
1103 | ret = 0; | |
8402752e AV |
1104 | if (need_mntput) |
1105 | nd->flags |= LOOKUP_JUMPED; | |
1106 | if (unlikely(ret < 0)) | |
1107 | path_put_conditional(path, nd); | |
1108 | return ret; | |
1da177e4 LT |
1109 | } |
1110 | ||
cc53ce53 | 1111 | int follow_down_one(struct path *path) |
1da177e4 LT |
1112 | { |
1113 | struct vfsmount *mounted; | |
1114 | ||
1c755af4 | 1115 | mounted = lookup_mnt(path); |
1da177e4 | 1116 | if (mounted) { |
9393bd07 AV |
1117 | dput(path->dentry); |
1118 | mntput(path->mnt); | |
1119 | path->mnt = mounted; | |
1120 | path->dentry = dget(mounted->mnt_root); | |
1da177e4 LT |
1121 | return 1; |
1122 | } | |
1123 | return 0; | |
1124 | } | |
4d359507 | 1125 | EXPORT_SYMBOL(follow_down_one); |
1da177e4 | 1126 | |
b8faf035 | 1127 | static inline int managed_dentry_rcu(struct dentry *dentry) |
62a7375e | 1128 | { |
b8faf035 N |
1129 | return (dentry->d_flags & DCACHE_MANAGE_TRANSIT) ? |
1130 | dentry->d_op->d_manage(dentry, true) : 0; | |
62a7375e IK |
1131 | } |
1132 | ||
9875cf80 | 1133 | /* |
287548e4 AV |
1134 | * Try to skip to top of mountpoint pile in rcuwalk mode. Fail if |
1135 | * we meet a managed dentry that would need blocking. | |
9875cf80 DH |
1136 | */ |
1137 | static bool __follow_mount_rcu(struct nameidata *nd, struct path *path, | |
254cf582 | 1138 | struct inode **inode, unsigned *seqp) |
9875cf80 | 1139 | { |
62a7375e | 1140 | for (;;) { |
c7105365 | 1141 | struct mount *mounted; |
62a7375e IK |
1142 | /* |
1143 | * Don't forget we might have a non-mountpoint managed dentry | |
1144 | * that wants to block transit. | |
1145 | */ | |
b8faf035 N |
1146 | switch (managed_dentry_rcu(path->dentry)) { |
1147 | case -ECHILD: | |
1148 | default: | |
ab90911f | 1149 | return false; |
b8faf035 N |
1150 | case -EISDIR: |
1151 | return true; | |
1152 | case 0: | |
1153 | break; | |
1154 | } | |
62a7375e IK |
1155 | |
1156 | if (!d_mountpoint(path->dentry)) | |
b8faf035 | 1157 | return !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
62a7375e | 1158 | |
474279dc | 1159 | mounted = __lookup_mnt(path->mnt, path->dentry); |
9875cf80 DH |
1160 | if (!mounted) |
1161 | break; | |
c7105365 AV |
1162 | path->mnt = &mounted->mnt; |
1163 | path->dentry = mounted->mnt.mnt_root; | |
a3fbbde7 | 1164 | nd->flags |= LOOKUP_JUMPED; |
254cf582 | 1165 | *seqp = read_seqcount_begin(&path->dentry->d_seq); |
59430262 LT |
1166 | /* |
1167 | * Update the inode too. We don't need to re-check the | |
1168 | * dentry sequence number here after this d_inode read, | |
1169 | * because a mount-point is always pinned. | |
1170 | */ | |
1171 | *inode = path->dentry->d_inode; | |
9875cf80 | 1172 | } |
f5be3e29 | 1173 | return !read_seqretry(&mount_lock, nd->m_seq) && |
b8faf035 | 1174 | !(path->dentry->d_flags & DCACHE_NEED_AUTOMOUNT); |
287548e4 AV |
1175 | } |
1176 | ||
31e6b01f NP |
1177 | static int follow_dotdot_rcu(struct nameidata *nd) |
1178 | { | |
4023bfc9 | 1179 | struct inode *inode = nd->inode; |
7bd88377 AV |
1180 | if (!nd->root.mnt) |
1181 | set_root_rcu(nd); | |
31e6b01f | 1182 | |
9875cf80 | 1183 | while (1) { |
31e6b01f NP |
1184 | if (nd->path.dentry == nd->root.dentry && |
1185 | nd->path.mnt == nd->root.mnt) { | |
1186 | break; | |
1187 | } | |
1188 | if (nd->path.dentry != nd->path.mnt->mnt_root) { | |
1189 | struct dentry *old = nd->path.dentry; | |
1190 | struct dentry *parent = old->d_parent; | |
1191 | unsigned seq; | |
1192 | ||
4023bfc9 | 1193 | inode = parent->d_inode; |
31e6b01f NP |
1194 | seq = read_seqcount_begin(&parent->d_seq); |
1195 | if (read_seqcount_retry(&old->d_seq, nd->seq)) | |
ef7562d5 | 1196 | goto failed; |
31e6b01f NP |
1197 | nd->path.dentry = parent; |
1198 | nd->seq = seq; | |
1199 | break; | |
1200 | } | |
1201 | if (!follow_up_rcu(&nd->path)) | |
1202 | break; | |
4023bfc9 | 1203 | inode = nd->path.dentry->d_inode; |
31e6b01f | 1204 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
31e6b01f | 1205 | } |
b37199e6 AV |
1206 | while (d_mountpoint(nd->path.dentry)) { |
1207 | struct mount *mounted; | |
1208 | mounted = __lookup_mnt(nd->path.mnt, nd->path.dentry); | |
1209 | if (!mounted) | |
1210 | break; | |
1211 | nd->path.mnt = &mounted->mnt; | |
1212 | nd->path.dentry = mounted->mnt.mnt_root; | |
4023bfc9 | 1213 | inode = nd->path.dentry->d_inode; |
b37199e6 | 1214 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); |
f5be3e29 | 1215 | if (read_seqretry(&mount_lock, nd->m_seq)) |
b37199e6 AV |
1216 | goto failed; |
1217 | } | |
4023bfc9 | 1218 | nd->inode = inode; |
31e6b01f | 1219 | return 0; |
ef7562d5 AV |
1220 | |
1221 | failed: | |
ef7562d5 | 1222 | return -ECHILD; |
31e6b01f NP |
1223 | } |
1224 | ||
cc53ce53 DH |
1225 | /* |
1226 | * Follow down to the covering mount currently visible to userspace. At each | |
1227 | * point, the filesystem owning that dentry may be queried as to whether the | |
1228 | * caller is permitted to proceed or not. | |
cc53ce53 | 1229 | */ |
7cc90cc3 | 1230 | int follow_down(struct path *path) |
cc53ce53 DH |
1231 | { |
1232 | unsigned managed; | |
1233 | int ret; | |
1234 | ||
1235 | while (managed = ACCESS_ONCE(path->dentry->d_flags), | |
1236 | unlikely(managed & DCACHE_MANAGED_DENTRY)) { | |
1237 | /* Allow the filesystem to manage the transit without i_mutex | |
1238 | * being held. | |
1239 | * | |
1240 | * We indicate to the filesystem if someone is trying to mount | |
1241 | * something here. This gives autofs the chance to deny anyone | |
1242 | * other than its daemon the right to mount on its | |
1243 | * superstructure. | |
1244 | * | |
1245 | * The filesystem may sleep at this point. | |
1246 | */ | |
1247 | if (managed & DCACHE_MANAGE_TRANSIT) { | |
1248 | BUG_ON(!path->dentry->d_op); | |
1249 | BUG_ON(!path->dentry->d_op->d_manage); | |
ab90911f | 1250 | ret = path->dentry->d_op->d_manage( |
1aed3e42 | 1251 | path->dentry, false); |
cc53ce53 DH |
1252 | if (ret < 0) |
1253 | return ret == -EISDIR ? 0 : ret; | |
1254 | } | |
1255 | ||
1256 | /* Transit to a mounted filesystem. */ | |
1257 | if (managed & DCACHE_MOUNTED) { | |
1258 | struct vfsmount *mounted = lookup_mnt(path); | |
1259 | if (!mounted) | |
1260 | break; | |
1261 | dput(path->dentry); | |
1262 | mntput(path->mnt); | |
1263 | path->mnt = mounted; | |
1264 | path->dentry = dget(mounted->mnt_root); | |
1265 | continue; | |
1266 | } | |
1267 | ||
1268 | /* Don't handle automount points here */ | |
1269 | break; | |
1270 | } | |
1271 | return 0; | |
1272 | } | |
4d359507 | 1273 | EXPORT_SYMBOL(follow_down); |
cc53ce53 | 1274 | |
9875cf80 DH |
1275 | /* |
1276 | * Skip to top of mountpoint pile in refwalk mode for follow_dotdot() | |
1277 | */ | |
1278 | static void follow_mount(struct path *path) | |
1279 | { | |
1280 | while (d_mountpoint(path->dentry)) { | |
1281 | struct vfsmount *mounted = lookup_mnt(path); | |
1282 | if (!mounted) | |
1283 | break; | |
1284 | dput(path->dentry); | |
1285 | mntput(path->mnt); | |
1286 | path->mnt = mounted; | |
1287 | path->dentry = dget(mounted->mnt_root); | |
1288 | } | |
1289 | } | |
1290 | ||
31e6b01f | 1291 | static void follow_dotdot(struct nameidata *nd) |
1da177e4 | 1292 | { |
7bd88377 AV |
1293 | if (!nd->root.mnt) |
1294 | set_root(nd); | |
e518ddb7 | 1295 | |
1da177e4 | 1296 | while(1) { |
4ac91378 | 1297 | struct dentry *old = nd->path.dentry; |
1da177e4 | 1298 | |
2a737871 AV |
1299 | if (nd->path.dentry == nd->root.dentry && |
1300 | nd->path.mnt == nd->root.mnt) { | |
1da177e4 LT |
1301 | break; |
1302 | } | |
4ac91378 | 1303 | if (nd->path.dentry != nd->path.mnt->mnt_root) { |
3088dd70 AV |
1304 | /* rare case of legitimate dget_parent()... */ |
1305 | nd->path.dentry = dget_parent(nd->path.dentry); | |
1da177e4 LT |
1306 | dput(old); |
1307 | break; | |
1308 | } | |
3088dd70 | 1309 | if (!follow_up(&nd->path)) |
1da177e4 | 1310 | break; |
1da177e4 | 1311 | } |
79ed0226 | 1312 | follow_mount(&nd->path); |
31e6b01f | 1313 | nd->inode = nd->path.dentry->d_inode; |
1da177e4 LT |
1314 | } |
1315 | ||
baa03890 | 1316 | /* |
bad61189 MS |
1317 | * This looks up the name in dcache, possibly revalidates the old dentry and |
1318 | * allocates a new one if not found or not valid. In the need_lookup argument | |
1319 | * returns whether i_op->lookup is necessary. | |
1320 | * | |
1321 | * dir->d_inode->i_mutex must be held | |
baa03890 | 1322 | */ |
bad61189 | 1323 | static struct dentry *lookup_dcache(struct qstr *name, struct dentry *dir, |
201f956e | 1324 | unsigned int flags, bool *need_lookup) |
baa03890 | 1325 | { |
baa03890 | 1326 | struct dentry *dentry; |
bad61189 | 1327 | int error; |
baa03890 | 1328 | |
bad61189 MS |
1329 | *need_lookup = false; |
1330 | dentry = d_lookup(dir, name); | |
1331 | if (dentry) { | |
39e3c955 | 1332 | if (dentry->d_flags & DCACHE_OP_REVALIDATE) { |
201f956e | 1333 | error = d_revalidate(dentry, flags); |
bad61189 MS |
1334 | if (unlikely(error <= 0)) { |
1335 | if (error < 0) { | |
1336 | dput(dentry); | |
1337 | return ERR_PTR(error); | |
5542aa2f EB |
1338 | } else { |
1339 | d_invalidate(dentry); | |
bad61189 MS |
1340 | dput(dentry); |
1341 | dentry = NULL; | |
1342 | } | |
1343 | } | |
1344 | } | |
1345 | } | |
baa03890 | 1346 | |
bad61189 MS |
1347 | if (!dentry) { |
1348 | dentry = d_alloc(dir, name); | |
1349 | if (unlikely(!dentry)) | |
1350 | return ERR_PTR(-ENOMEM); | |
baa03890 | 1351 | |
bad61189 | 1352 | *need_lookup = true; |
baa03890 NP |
1353 | } |
1354 | return dentry; | |
1355 | } | |
1356 | ||
44396f4b | 1357 | /* |
13a2c3be BF |
1358 | * Call i_op->lookup on the dentry. The dentry must be negative and |
1359 | * unhashed. | |
bad61189 MS |
1360 | * |
1361 | * dir->d_inode->i_mutex must be held | |
44396f4b | 1362 | */ |
bad61189 | 1363 | static struct dentry *lookup_real(struct inode *dir, struct dentry *dentry, |
72bd866a | 1364 | unsigned int flags) |
44396f4b | 1365 | { |
44396f4b JB |
1366 | struct dentry *old; |
1367 | ||
1368 | /* Don't create child dentry for a dead directory. */ | |
bad61189 | 1369 | if (unlikely(IS_DEADDIR(dir))) { |
e188dc02 | 1370 | dput(dentry); |
44396f4b | 1371 | return ERR_PTR(-ENOENT); |
e188dc02 | 1372 | } |
44396f4b | 1373 | |
72bd866a | 1374 | old = dir->i_op->lookup(dir, dentry, flags); |
44396f4b JB |
1375 | if (unlikely(old)) { |
1376 | dput(dentry); | |
1377 | dentry = old; | |
1378 | } | |
1379 | return dentry; | |
1380 | } | |
1381 | ||
a3255546 | 1382 | static struct dentry *__lookup_hash(struct qstr *name, |
72bd866a | 1383 | struct dentry *base, unsigned int flags) |
a3255546 | 1384 | { |
bad61189 | 1385 | bool need_lookup; |
a3255546 AV |
1386 | struct dentry *dentry; |
1387 | ||
72bd866a | 1388 | dentry = lookup_dcache(name, base, flags, &need_lookup); |
bad61189 MS |
1389 | if (!need_lookup) |
1390 | return dentry; | |
a3255546 | 1391 | |
72bd866a | 1392 | return lookup_real(base->d_inode, dentry, flags); |
a3255546 AV |
1393 | } |
1394 | ||
1da177e4 LT |
1395 | /* |
1396 | * It's more convoluted than I'd like it to be, but... it's still fairly | |
1397 | * small and for now I'd prefer to have fast path as straight as possible. | |
1398 | * It _is_ time-critical. | |
1399 | */ | |
e97cdc87 | 1400 | static int lookup_fast(struct nameidata *nd, |
254cf582 AV |
1401 | struct path *path, struct inode **inode, |
1402 | unsigned *seqp) | |
1da177e4 | 1403 | { |
4ac91378 | 1404 | struct vfsmount *mnt = nd->path.mnt; |
31e6b01f | 1405 | struct dentry *dentry, *parent = nd->path.dentry; |
5a18fff2 AV |
1406 | int need_reval = 1; |
1407 | int status = 1; | |
9875cf80 DH |
1408 | int err; |
1409 | ||
b04f784e NP |
1410 | /* |
1411 | * Rename seqlock is not required here because in the off chance | |
1412 | * of a false negative due to a concurrent rename, we're going to | |
1413 | * do the non-racy lookup, below. | |
1414 | */ | |
31e6b01f NP |
1415 | if (nd->flags & LOOKUP_RCU) { |
1416 | unsigned seq; | |
766c4cbf | 1417 | bool negative; |
da53be12 | 1418 | dentry = __d_lookup_rcu(parent, &nd->last, &seq); |
5a18fff2 AV |
1419 | if (!dentry) |
1420 | goto unlazy; | |
1421 | ||
12f8ad4b LT |
1422 | /* |
1423 | * This sequence count validates that the inode matches | |
1424 | * the dentry name information from lookup. | |
1425 | */ | |
63afdfc7 | 1426 | *inode = d_backing_inode(dentry); |
766c4cbf | 1427 | negative = d_is_negative(dentry); |
12f8ad4b LT |
1428 | if (read_seqcount_retry(&dentry->d_seq, seq)) |
1429 | return -ECHILD; | |
766c4cbf AV |
1430 | if (negative) |
1431 | return -ENOENT; | |
12f8ad4b LT |
1432 | |
1433 | /* | |
1434 | * This sequence count validates that the parent had no | |
1435 | * changes while we did the lookup of the dentry above. | |
1436 | * | |
1437 | * The memory barrier in read_seqcount_begin of child is | |
1438 | * enough, we can use __read_seqcount_retry here. | |
1439 | */ | |
31e6b01f NP |
1440 | if (__read_seqcount_retry(&parent->d_seq, nd->seq)) |
1441 | return -ECHILD; | |
5a18fff2 | 1442 | |
254cf582 | 1443 | *seqp = seq; |
24643087 | 1444 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) { |
4ce16ef3 | 1445 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1446 | if (unlikely(status <= 0)) { |
1447 | if (status != -ECHILD) | |
1448 | need_reval = 0; | |
1449 | goto unlazy; | |
1450 | } | |
24643087 | 1451 | } |
31e6b01f NP |
1452 | path->mnt = mnt; |
1453 | path->dentry = dentry; | |
254cf582 | 1454 | if (likely(__follow_mount_rcu(nd, path, inode, seqp))) |
b8faf035 | 1455 | return 0; |
5a18fff2 | 1456 | unlazy: |
254cf582 | 1457 | if (unlazy_walk(nd, dentry, seq)) |
19660af7 | 1458 | return -ECHILD; |
5a18fff2 | 1459 | } else { |
e97cdc87 | 1460 | dentry = __d_lookup(parent, &nd->last); |
9875cf80 | 1461 | } |
5a18fff2 | 1462 | |
81e6f520 AV |
1463 | if (unlikely(!dentry)) |
1464 | goto need_lookup; | |
1465 | ||
5a18fff2 | 1466 | if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE) && need_reval) |
4ce16ef3 | 1467 | status = d_revalidate(dentry, nd->flags); |
5a18fff2 AV |
1468 | if (unlikely(status <= 0)) { |
1469 | if (status < 0) { | |
1470 | dput(dentry); | |
1471 | return status; | |
1472 | } | |
5542aa2f EB |
1473 | d_invalidate(dentry); |
1474 | dput(dentry); | |
1475 | goto need_lookup; | |
24643087 | 1476 | } |
697f514d | 1477 | |
766c4cbf AV |
1478 | if (unlikely(d_is_negative(dentry))) { |
1479 | dput(dentry); | |
1480 | return -ENOENT; | |
1481 | } | |
9875cf80 DH |
1482 | path->mnt = mnt; |
1483 | path->dentry = dentry; | |
756daf26 | 1484 | err = follow_managed(path, nd); |
8402752e | 1485 | if (likely(!err)) |
63afdfc7 | 1486 | *inode = d_backing_inode(path->dentry); |
8402752e | 1487 | return err; |
81e6f520 AV |
1488 | |
1489 | need_lookup: | |
697f514d MS |
1490 | return 1; |
1491 | } | |
1492 | ||
1493 | /* Fast lookup failed, do it the slow way */ | |
cc2a5271 | 1494 | static int lookup_slow(struct nameidata *nd, struct path *path) |
697f514d MS |
1495 | { |
1496 | struct dentry *dentry, *parent; | |
697f514d MS |
1497 | |
1498 | parent = nd->path.dentry; | |
81e6f520 AV |
1499 | BUG_ON(nd->inode != parent->d_inode); |
1500 | ||
1501 | mutex_lock(&parent->d_inode->i_mutex); | |
cc2a5271 | 1502 | dentry = __lookup_hash(&nd->last, parent, nd->flags); |
81e6f520 AV |
1503 | mutex_unlock(&parent->d_inode->i_mutex); |
1504 | if (IS_ERR(dentry)) | |
1505 | return PTR_ERR(dentry); | |
697f514d MS |
1506 | path->mnt = nd->path.mnt; |
1507 | path->dentry = dentry; | |
8402752e | 1508 | return follow_managed(path, nd); |
1da177e4 LT |
1509 | } |
1510 | ||
52094c8a AV |
1511 | static inline int may_lookup(struct nameidata *nd) |
1512 | { | |
1513 | if (nd->flags & LOOKUP_RCU) { | |
4ad5abb3 | 1514 | int err = inode_permission(nd->inode, MAY_EXEC|MAY_NOT_BLOCK); |
52094c8a AV |
1515 | if (err != -ECHILD) |
1516 | return err; | |
6e9918b7 | 1517 | if (unlazy_walk(nd, NULL, 0)) |
52094c8a AV |
1518 | return -ECHILD; |
1519 | } | |
4ad5abb3 | 1520 | return inode_permission(nd->inode, MAY_EXEC); |
52094c8a AV |
1521 | } |
1522 | ||
9856fa1b AV |
1523 | static inline int handle_dots(struct nameidata *nd, int type) |
1524 | { | |
1525 | if (type == LAST_DOTDOT) { | |
1526 | if (nd->flags & LOOKUP_RCU) { | |
70291aec | 1527 | return follow_dotdot_rcu(nd); |
9856fa1b AV |
1528 | } else |
1529 | follow_dotdot(nd); | |
1530 | } | |
1531 | return 0; | |
1532 | } | |
1533 | ||
951361f9 AV |
1534 | static void terminate_walk(struct nameidata *nd) |
1535 | { | |
1536 | if (!(nd->flags & LOOKUP_RCU)) { | |
1537 | path_put(&nd->path); | |
1538 | } else { | |
1539 | nd->flags &= ~LOOKUP_RCU; | |
5b6ca027 AV |
1540 | if (!(nd->flags & LOOKUP_ROOT)) |
1541 | nd->root.mnt = NULL; | |
8b61e74f | 1542 | rcu_read_unlock(); |
951361f9 | 1543 | } |
15439726 AV |
1544 | while (unlikely(nd->depth)) |
1545 | put_link(nd); | |
951361f9 AV |
1546 | } |
1547 | ||
181548c0 AV |
1548 | static int pick_link(struct nameidata *nd, struct path *link, |
1549 | struct inode *inode, unsigned seq) | |
d63ff28f | 1550 | { |
626de996 | 1551 | int error; |
1cf2665b | 1552 | struct saved *last; |
756daf26 | 1553 | if (unlikely(nd->total_link_count++ >= MAXSYMLINKS)) { |
626de996 AV |
1554 | path_to_nameidata(link, nd); |
1555 | return -ELOOP; | |
1556 | } | |
d63ff28f AV |
1557 | if (nd->flags & LOOKUP_RCU) { |
1558 | if (unlikely(nd->path.mnt != link->mnt || | |
254cf582 | 1559 | unlazy_walk(nd, link->dentry, seq))) { |
d63ff28f AV |
1560 | return -ECHILD; |
1561 | } | |
1562 | } | |
cd179f44 AV |
1563 | if (link->mnt == nd->path.mnt) |
1564 | mntget(link->mnt); | |
626de996 AV |
1565 | error = nd_alloc_stack(nd); |
1566 | if (unlikely(error)) { | |
cd179f44 | 1567 | path_put(link); |
626de996 AV |
1568 | return error; |
1569 | } | |
1570 | ||
ab104923 | 1571 | last = nd->stack + nd->depth++; |
1cf2665b | 1572 | last->link = *link; |
ab104923 | 1573 | last->cookie = NULL; |
181548c0 | 1574 | last->inode = inode; |
d63ff28f AV |
1575 | return 1; |
1576 | } | |
1577 | ||
3ddcd056 LT |
1578 | /* |
1579 | * Do we need to follow links? We _really_ want to be able | |
1580 | * to do this check without having to look at inode->i_op, | |
1581 | * so we keep a cache of "no, this doesn't need follow_link" | |
1582 | * for the common case. | |
1583 | */ | |
254cf582 | 1584 | static inline int should_follow_link(struct nameidata *nd, struct path *link, |
181548c0 AV |
1585 | int follow, |
1586 | struct inode *inode, unsigned seq) | |
3ddcd056 | 1587 | { |
d63ff28f AV |
1588 | if (likely(!d_is_symlink(link->dentry))) |
1589 | return 0; | |
1590 | if (!follow) | |
1591 | return 0; | |
181548c0 | 1592 | return pick_link(nd, link, inode, seq); |
3ddcd056 LT |
1593 | } |
1594 | ||
4693a547 AV |
1595 | enum {WALK_GET = 1, WALK_PUT = 2}; |
1596 | ||
1597 | static int walk_component(struct nameidata *nd, int flags) | |
ce57dfc1 | 1598 | { |
caa85634 | 1599 | struct path path; |
ce57dfc1 | 1600 | struct inode *inode; |
254cf582 | 1601 | unsigned seq; |
ce57dfc1 AV |
1602 | int err; |
1603 | /* | |
1604 | * "." and ".." are special - ".." especially so because it has | |
1605 | * to be able to know about the current root directory and | |
1606 | * parent relationships. | |
1607 | */ | |
4693a547 AV |
1608 | if (unlikely(nd->last_type != LAST_NORM)) { |
1609 | err = handle_dots(nd, nd->last_type); | |
1610 | if (flags & WALK_PUT) | |
1611 | put_link(nd); | |
1612 | return err; | |
1613 | } | |
254cf582 | 1614 | err = lookup_fast(nd, &path, &inode, &seq); |
ce57dfc1 | 1615 | if (unlikely(err)) { |
697f514d | 1616 | if (err < 0) |
f0a9ba70 | 1617 | return err; |
697f514d | 1618 | |
caa85634 | 1619 | err = lookup_slow(nd, &path); |
697f514d | 1620 | if (err < 0) |
f0a9ba70 | 1621 | return err; |
697f514d | 1622 | |
63afdfc7 | 1623 | inode = d_backing_inode(path.dentry); |
254cf582 | 1624 | seq = 0; /* we are already out of RCU mode */ |
766c4cbf | 1625 | err = -ENOENT; |
caa85634 | 1626 | if (d_is_negative(path.dentry)) |
766c4cbf | 1627 | goto out_path_put; |
ce57dfc1 | 1628 | } |
697f514d | 1629 | |
4693a547 AV |
1630 | if (flags & WALK_PUT) |
1631 | put_link(nd); | |
181548c0 | 1632 | err = should_follow_link(nd, &path, flags & WALK_GET, inode, seq); |
d63ff28f AV |
1633 | if (unlikely(err)) |
1634 | return err; | |
caa85634 | 1635 | path_to_nameidata(&path, nd); |
ce57dfc1 | 1636 | nd->inode = inode; |
254cf582 | 1637 | nd->seq = seq; |
ce57dfc1 | 1638 | return 0; |
697f514d MS |
1639 | |
1640 | out_path_put: | |
caa85634 | 1641 | path_to_nameidata(&path, nd); |
697f514d | 1642 | return err; |
ce57dfc1 AV |
1643 | } |
1644 | ||
bfcfaa77 LT |
1645 | /* |
1646 | * We can do the critical dentry name comparison and hashing | |
1647 | * operations one word at a time, but we are limited to: | |
1648 | * | |
1649 | * - Architectures with fast unaligned word accesses. We could | |
1650 | * do a "get_unaligned()" if this helps and is sufficiently | |
1651 | * fast. | |
1652 | * | |
bfcfaa77 LT |
1653 | * - non-CONFIG_DEBUG_PAGEALLOC configurations (so that we |
1654 | * do not trap on the (extremely unlikely) case of a page | |
1655 | * crossing operation. | |
1656 | * | |
1657 | * - Furthermore, we need an efficient 64-bit compile for the | |
1658 | * 64-bit case in order to generate the "number of bytes in | |
1659 | * the final mask". Again, that could be replaced with a | |
1660 | * efficient population count instruction or similar. | |
1661 | */ | |
1662 | #ifdef CONFIG_DCACHE_WORD_ACCESS | |
1663 | ||
f68e556e | 1664 | #include <asm/word-at-a-time.h> |
bfcfaa77 | 1665 | |
f68e556e | 1666 | #ifdef CONFIG_64BIT |
bfcfaa77 LT |
1667 | |
1668 | static inline unsigned int fold_hash(unsigned long hash) | |
1669 | { | |
99d263d4 | 1670 | return hash_64(hash, 32); |
bfcfaa77 LT |
1671 | } |
1672 | ||
1673 | #else /* 32-bit case */ | |
1674 | ||
bfcfaa77 LT |
1675 | #define fold_hash(x) (x) |
1676 | ||
1677 | #endif | |
1678 | ||
1679 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) | |
1680 | { | |
1681 | unsigned long a, mask; | |
1682 | unsigned long hash = 0; | |
1683 | ||
1684 | for (;;) { | |
e419b4cc | 1685 | a = load_unaligned_zeropad(name); |
bfcfaa77 LT |
1686 | if (len < sizeof(unsigned long)) |
1687 | break; | |
1688 | hash += a; | |
f132c5be | 1689 | hash *= 9; |
bfcfaa77 LT |
1690 | name += sizeof(unsigned long); |
1691 | len -= sizeof(unsigned long); | |
1692 | if (!len) | |
1693 | goto done; | |
1694 | } | |
a5c21dce | 1695 | mask = bytemask_from_count(len); |
bfcfaa77 LT |
1696 | hash += mask & a; |
1697 | done: | |
1698 | return fold_hash(hash); | |
1699 | } | |
1700 | EXPORT_SYMBOL(full_name_hash); | |
1701 | ||
bfcfaa77 LT |
1702 | /* |
1703 | * Calculate the length and hash of the path component, and | |
d6bb3e90 | 1704 | * return the "hash_len" as the result. |
bfcfaa77 | 1705 | */ |
d6bb3e90 | 1706 | static inline u64 hash_name(const char *name) |
bfcfaa77 | 1707 | { |
36126f8f LT |
1708 | unsigned long a, b, adata, bdata, mask, hash, len; |
1709 | const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS; | |
bfcfaa77 LT |
1710 | |
1711 | hash = a = 0; | |
1712 | len = -sizeof(unsigned long); | |
1713 | do { | |
1714 | hash = (hash + a) * 9; | |
1715 | len += sizeof(unsigned long); | |
e419b4cc | 1716 | a = load_unaligned_zeropad(name+len); |
36126f8f LT |
1717 | b = a ^ REPEAT_BYTE('/'); |
1718 | } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants))); | |
1719 | ||
1720 | adata = prep_zero_mask(a, adata, &constants); | |
1721 | bdata = prep_zero_mask(b, bdata, &constants); | |
1722 | ||
1723 | mask = create_zero_mask(adata | bdata); | |
1724 | ||
1725 | hash += a & zero_bytemask(mask); | |
9226b5b4 | 1726 | len += find_zero(mask); |
d6bb3e90 | 1727 | return hashlen_create(fold_hash(hash), len); |
bfcfaa77 LT |
1728 | } |
1729 | ||
1730 | #else | |
1731 | ||
0145acc2 LT |
1732 | unsigned int full_name_hash(const unsigned char *name, unsigned int len) |
1733 | { | |
1734 | unsigned long hash = init_name_hash(); | |
1735 | while (len--) | |
1736 | hash = partial_name_hash(*name++, hash); | |
1737 | return end_name_hash(hash); | |
1738 | } | |
ae942ae7 | 1739 | EXPORT_SYMBOL(full_name_hash); |
0145acc2 | 1740 | |
200e9ef7 LT |
1741 | /* |
1742 | * We know there's a real path component here of at least | |
1743 | * one character. | |
1744 | */ | |
d6bb3e90 | 1745 | static inline u64 hash_name(const char *name) |
200e9ef7 LT |
1746 | { |
1747 | unsigned long hash = init_name_hash(); | |
1748 | unsigned long len = 0, c; | |
1749 | ||
1750 | c = (unsigned char)*name; | |
1751 | do { | |
1752 | len++; | |
1753 | hash = partial_name_hash(c, hash); | |
1754 | c = (unsigned char)name[len]; | |
1755 | } while (c && c != '/'); | |
d6bb3e90 | 1756 | return hashlen_create(end_name_hash(hash), len); |
200e9ef7 LT |
1757 | } |
1758 | ||
bfcfaa77 LT |
1759 | #endif |
1760 | ||
1da177e4 LT |
1761 | /* |
1762 | * Name resolution. | |
ea3834d9 PM |
1763 | * This is the basic name resolution function, turning a pathname into |
1764 | * the final dentry. We expect 'base' to be positive and a directory. | |
1da177e4 | 1765 | * |
ea3834d9 PM |
1766 | * Returns 0 and nd will have valid dentry and mnt on success. |
1767 | * Returns error and drops reference to input namei data on failure. | |
1da177e4 | 1768 | */ |
6de88d72 | 1769 | static int link_path_walk(const char *name, struct nameidata *nd) |
1da177e4 | 1770 | { |
1da177e4 | 1771 | int err; |
32cd7468 | 1772 | |
1da177e4 LT |
1773 | while (*name=='/') |
1774 | name++; | |
1775 | if (!*name) | |
9e18f10a | 1776 | return 0; |
1da177e4 | 1777 | |
1da177e4 LT |
1778 | /* At this point we know we have a real path component. */ |
1779 | for(;;) { | |
d6bb3e90 | 1780 | u64 hash_len; |
fe479a58 | 1781 | int type; |
1da177e4 | 1782 | |
52094c8a | 1783 | err = may_lookup(nd); |
1da177e4 | 1784 | if (err) |
3595e234 | 1785 | return err; |
1da177e4 | 1786 | |
d6bb3e90 | 1787 | hash_len = hash_name(name); |
1da177e4 | 1788 | |
fe479a58 | 1789 | type = LAST_NORM; |
d6bb3e90 | 1790 | if (name[0] == '.') switch (hashlen_len(hash_len)) { |
fe479a58 | 1791 | case 2: |
200e9ef7 | 1792 | if (name[1] == '.') { |
fe479a58 | 1793 | type = LAST_DOTDOT; |
16c2cd71 AV |
1794 | nd->flags |= LOOKUP_JUMPED; |
1795 | } | |
fe479a58 AV |
1796 | break; |
1797 | case 1: | |
1798 | type = LAST_DOT; | |
1799 | } | |
5a202bcd AV |
1800 | if (likely(type == LAST_NORM)) { |
1801 | struct dentry *parent = nd->path.dentry; | |
16c2cd71 | 1802 | nd->flags &= ~LOOKUP_JUMPED; |
5a202bcd | 1803 | if (unlikely(parent->d_flags & DCACHE_OP_HASH)) { |
a060dc50 | 1804 | struct qstr this = { { .hash_len = hash_len }, .name = name }; |
da53be12 | 1805 | err = parent->d_op->d_hash(parent, &this); |
5a202bcd | 1806 | if (err < 0) |
3595e234 | 1807 | return err; |
d6bb3e90 LT |
1808 | hash_len = this.hash_len; |
1809 | name = this.name; | |
5a202bcd AV |
1810 | } |
1811 | } | |
fe479a58 | 1812 | |
d6bb3e90 LT |
1813 | nd->last.hash_len = hash_len; |
1814 | nd->last.name = name; | |
5f4a6a69 AV |
1815 | nd->last_type = type; |
1816 | ||
d6bb3e90 LT |
1817 | name += hashlen_len(hash_len); |
1818 | if (!*name) | |
bdf6cbf1 | 1819 | goto OK; |
200e9ef7 LT |
1820 | /* |
1821 | * If it wasn't NUL, we know it was '/'. Skip that | |
1822 | * slash, and continue until no more slashes. | |
1823 | */ | |
1824 | do { | |
d6bb3e90 LT |
1825 | name++; |
1826 | } while (unlikely(*name == '/')); | |
8620c238 AV |
1827 | if (unlikely(!*name)) { |
1828 | OK: | |
368ee9ba | 1829 | /* pathname body, done */ |
8620c238 AV |
1830 | if (!nd->depth) |
1831 | return 0; | |
1832 | name = nd->stack[nd->depth - 1].name; | |
368ee9ba | 1833 | /* trailing symlink, done */ |
8620c238 AV |
1834 | if (!name) |
1835 | return 0; | |
1836 | /* last component of nested symlink */ | |
4693a547 | 1837 | err = walk_component(nd, WALK_GET | WALK_PUT); |
8620c238 | 1838 | } else { |
4693a547 | 1839 | err = walk_component(nd, WALK_GET); |
8620c238 | 1840 | } |
ce57dfc1 | 1841 | if (err < 0) |
3595e234 | 1842 | return err; |
1da177e4 | 1843 | |
ce57dfc1 | 1844 | if (err) { |
626de996 | 1845 | const char *s = get_link(nd); |
5a460275 | 1846 | |
3595e234 AV |
1847 | if (unlikely(IS_ERR(s))) |
1848 | return PTR_ERR(s); | |
d40bcc09 AV |
1849 | err = 0; |
1850 | if (unlikely(!s)) { | |
1851 | /* jumped */ | |
b9ff4429 | 1852 | put_link(nd); |
d40bcc09 | 1853 | } else { |
fab51e8a AV |
1854 | nd->stack[nd->depth - 1].name = name; |
1855 | name = s; | |
1856 | continue; | |
d40bcc09 | 1857 | } |
31e6b01f | 1858 | } |
3595e234 AV |
1859 | if (unlikely(!d_can_lookup(nd->path.dentry))) |
1860 | return -ENOTDIR; | |
1da177e4 | 1861 | } |
1da177e4 LT |
1862 | } |
1863 | ||
368ee9ba AV |
1864 | static const char *path_init(int dfd, const struct filename *name, |
1865 | unsigned int flags, struct nameidata *nd) | |
31e6b01f NP |
1866 | { |
1867 | int retval = 0; | |
fd2f7cb5 | 1868 | const char *s = name->name; |
31e6b01f NP |
1869 | |
1870 | nd->last_type = LAST_ROOT; /* if there are only slashes... */ | |
980f3ea2 | 1871 | nd->flags = flags | LOOKUP_JUMPED | LOOKUP_PARENT; |
31e6b01f | 1872 | nd->depth = 0; |
368ee9ba | 1873 | nd->total_link_count = 0; |
5b6ca027 | 1874 | if (flags & LOOKUP_ROOT) { |
b18825a7 DH |
1875 | struct dentry *root = nd->root.dentry; |
1876 | struct inode *inode = root->d_inode; | |
fd2f7cb5 | 1877 | if (*s) { |
44b1d530 | 1878 | if (!d_can_lookup(root)) |
368ee9ba | 1879 | return ERR_PTR(-ENOTDIR); |
73d049a4 AV |
1880 | retval = inode_permission(inode, MAY_EXEC); |
1881 | if (retval) | |
368ee9ba | 1882 | return ERR_PTR(retval); |
73d049a4 | 1883 | } |
5b6ca027 AV |
1884 | nd->path = nd->root; |
1885 | nd->inode = inode; | |
1886 | if (flags & LOOKUP_RCU) { | |
8b61e74f | 1887 | rcu_read_lock(); |
5b6ca027 | 1888 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); |
48a066e7 | 1889 | nd->m_seq = read_seqbegin(&mount_lock); |
5b6ca027 AV |
1890 | } else { |
1891 | path_get(&nd->path); | |
1892 | } | |
368ee9ba | 1893 | return s; |
5b6ca027 AV |
1894 | } |
1895 | ||
31e6b01f | 1896 | nd->root.mnt = NULL; |
31e6b01f | 1897 | |
48a066e7 | 1898 | nd->m_seq = read_seqbegin(&mount_lock); |
fd2f7cb5 | 1899 | if (*s == '/') { |
e41f7d4e | 1900 | if (flags & LOOKUP_RCU) { |
8b61e74f | 1901 | rcu_read_lock(); |
7bd88377 | 1902 | nd->seq = set_root_rcu(nd); |
e41f7d4e AV |
1903 | } else { |
1904 | set_root(nd); | |
1905 | path_get(&nd->root); | |
1906 | } | |
1907 | nd->path = nd->root; | |
31e6b01f | 1908 | } else if (dfd == AT_FDCWD) { |
e41f7d4e AV |
1909 | if (flags & LOOKUP_RCU) { |
1910 | struct fs_struct *fs = current->fs; | |
1911 | unsigned seq; | |
31e6b01f | 1912 | |
8b61e74f | 1913 | rcu_read_lock(); |
c28cc364 | 1914 | |
e41f7d4e AV |
1915 | do { |
1916 | seq = read_seqcount_begin(&fs->seq); | |
1917 | nd->path = fs->pwd; | |
1918 | nd->seq = __read_seqcount_begin(&nd->path.dentry->d_seq); | |
1919 | } while (read_seqcount_retry(&fs->seq, seq)); | |
1920 | } else { | |
1921 | get_fs_pwd(current->fs, &nd->path); | |
1922 | } | |
31e6b01f | 1923 | } else { |
582aa64a | 1924 | /* Caller must check execute permissions on the starting path component */ |
2903ff01 | 1925 | struct fd f = fdget_raw(dfd); |
31e6b01f NP |
1926 | struct dentry *dentry; |
1927 | ||
2903ff01 | 1928 | if (!f.file) |
368ee9ba | 1929 | return ERR_PTR(-EBADF); |
31e6b01f | 1930 | |
2903ff01 | 1931 | dentry = f.file->f_path.dentry; |
31e6b01f | 1932 | |
fd2f7cb5 | 1933 | if (*s) { |
44b1d530 | 1934 | if (!d_can_lookup(dentry)) { |
2903ff01 | 1935 | fdput(f); |
368ee9ba | 1936 | return ERR_PTR(-ENOTDIR); |
2903ff01 | 1937 | } |
f52e0c11 | 1938 | } |
31e6b01f | 1939 | |
2903ff01 | 1940 | nd->path = f.file->f_path; |
e41f7d4e | 1941 | if (flags & LOOKUP_RCU) { |
8b61e74f | 1942 | rcu_read_lock(); |
34a26b99 AV |
1943 | nd->inode = nd->path.dentry->d_inode; |
1944 | nd->seq = read_seqcount_begin(&nd->path.dentry->d_seq); | |
e41f7d4e | 1945 | } else { |
2903ff01 | 1946 | path_get(&nd->path); |
34a26b99 | 1947 | nd->inode = nd->path.dentry->d_inode; |
e41f7d4e | 1948 | } |
34a26b99 | 1949 | fdput(f); |
368ee9ba | 1950 | return s; |
31e6b01f | 1951 | } |
31e6b01f | 1952 | |
31e6b01f | 1953 | nd->inode = nd->path.dentry->d_inode; |
4023bfc9 | 1954 | if (!(flags & LOOKUP_RCU)) |
368ee9ba | 1955 | return s; |
4023bfc9 | 1956 | if (likely(!read_seqcount_retry(&nd->path.dentry->d_seq, nd->seq))) |
368ee9ba | 1957 | return s; |
4023bfc9 AV |
1958 | if (!(nd->flags & LOOKUP_ROOT)) |
1959 | nd->root.mnt = NULL; | |
1960 | rcu_read_unlock(); | |
368ee9ba | 1961 | return ERR_PTR(-ECHILD); |
9b4a9b14 AV |
1962 | } |
1963 | ||
893b7775 AV |
1964 | static void path_cleanup(struct nameidata *nd) |
1965 | { | |
1966 | if (nd->root.mnt && !(nd->flags & LOOKUP_ROOT)) { | |
1967 | path_put(&nd->root); | |
1968 | nd->root.mnt = NULL; | |
1969 | } | |
9b4a9b14 AV |
1970 | } |
1971 | ||
3bdba28b | 1972 | static const char *trailing_symlink(struct nameidata *nd) |
95fa25d9 AV |
1973 | { |
1974 | const char *s; | |
fec2fa24 | 1975 | int error = may_follow_link(nd); |
deb106c6 | 1976 | if (unlikely(error)) |
3bdba28b | 1977 | return ERR_PTR(error); |
95fa25d9 | 1978 | nd->flags |= LOOKUP_PARENT; |
fab51e8a | 1979 | nd->stack[0].name = NULL; |
3b2e7f75 | 1980 | s = get_link(nd); |
deb106c6 | 1981 | return s ? s : ""; |
95fa25d9 AV |
1982 | } |
1983 | ||
caa85634 | 1984 | static inline int lookup_last(struct nameidata *nd) |
bd92d7fe AV |
1985 | { |
1986 | if (nd->last_type == LAST_NORM && nd->last.name[nd->last.len]) | |
1987 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
1988 | ||
1989 | nd->flags &= ~LOOKUP_PARENT; | |
deb106c6 | 1990 | return walk_component(nd, |
4693a547 AV |
1991 | nd->flags & LOOKUP_FOLLOW |
1992 | ? nd->depth | |
1993 | ? WALK_PUT | WALK_GET | |
1994 | : WALK_GET | |
1995 | : 0); | |
bd92d7fe AV |
1996 | } |
1997 | ||
9b4a9b14 | 1998 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
5eb6b495 | 1999 | static int path_lookupat(int dfd, const struct filename *name, |
9b4a9b14 AV |
2000 | unsigned int flags, struct nameidata *nd) |
2001 | { | |
368ee9ba | 2002 | const char *s = path_init(dfd, name, flags, nd); |
bd92d7fe | 2003 | int err; |
31e6b01f | 2004 | |
368ee9ba AV |
2005 | if (IS_ERR(s)) |
2006 | return PTR_ERR(s); | |
3bdba28b AV |
2007 | while (!(err = link_path_walk(s, nd)) |
2008 | && ((err = lookup_last(nd)) > 0)) { | |
2009 | s = trailing_symlink(nd); | |
2010 | if (IS_ERR(s)) { | |
2011 | err = PTR_ERR(s); | |
2012 | break; | |
bd92d7fe AV |
2013 | } |
2014 | } | |
9f1fafee AV |
2015 | if (!err) |
2016 | err = complete_walk(nd); | |
bd92d7fe | 2017 | |
deb106c6 AV |
2018 | if (!err && nd->flags & LOOKUP_DIRECTORY) |
2019 | if (!d_can_lookup(nd->path.dentry)) | |
bd23a539 | 2020 | err = -ENOTDIR; |
deb106c6 AV |
2021 | if (err) |
2022 | terminate_walk(nd); | |
16c2cd71 | 2023 | |
893b7775 | 2024 | path_cleanup(nd); |
bd92d7fe | 2025 | return err; |
ee0827cd | 2026 | } |
31e6b01f | 2027 | |
873f1eed | 2028 | static int filename_lookup(int dfd, struct filename *name, |
ee0827cd AV |
2029 | unsigned int flags, struct nameidata *nd) |
2030 | { | |
894bc8c4 | 2031 | int retval; |
756daf26 | 2032 | struct nameidata *saved_nd = set_nameidata(nd); |
894bc8c4 | 2033 | |
894bc8c4 | 2034 | retval = path_lookupat(dfd, name, flags | LOOKUP_RCU, nd); |
ee0827cd | 2035 | if (unlikely(retval == -ECHILD)) |
5eb6b495 | 2036 | retval = path_lookupat(dfd, name, flags, nd); |
ee0827cd | 2037 | if (unlikely(retval == -ESTALE)) |
5eb6b495 | 2038 | retval = path_lookupat(dfd, name, flags | LOOKUP_REVAL, nd); |
31e6b01f | 2039 | |
f78570dd | 2040 | if (likely(!retval)) |
adb5c247 | 2041 | audit_inode(name, nd->path.dentry, flags & LOOKUP_PARENT); |
756daf26 | 2042 | restore_nameidata(saved_nd); |
170aa3d0 | 2043 | return retval; |
1da177e4 LT |
2044 | } |
2045 | ||
8bcb77fa AV |
2046 | /* Returns 0 and nd will be valid on success; Retuns error, otherwise. */ |
2047 | static int path_parentat(int dfd, const struct filename *name, | |
2048 | unsigned int flags, struct nameidata *nd) | |
2049 | { | |
368ee9ba AV |
2050 | const char *s = path_init(dfd, name, flags, nd); |
2051 | int err; | |
2052 | if (IS_ERR(s)) | |
2053 | return PTR_ERR(s); | |
2054 | err = link_path_walk(s, nd); | |
8bcb77fa AV |
2055 | if (!err) |
2056 | err = complete_walk(nd); | |
deb106c6 AV |
2057 | if (err) |
2058 | terminate_walk(nd); | |
8bcb77fa AV |
2059 | path_cleanup(nd); |
2060 | return err; | |
2061 | } | |
2062 | ||
2063 | static int filename_parentat(int dfd, struct filename *name, | |
2064 | unsigned int flags, struct nameidata *nd) | |
2065 | { | |
2066 | int retval; | |
2067 | struct nameidata *saved_nd = set_nameidata(nd); | |
2068 | ||
2069 | retval = path_parentat(dfd, name, flags | LOOKUP_RCU, nd); | |
2070 | if (unlikely(retval == -ECHILD)) | |
2071 | retval = path_parentat(dfd, name, flags, nd); | |
2072 | if (unlikely(retval == -ESTALE)) | |
2073 | retval = path_parentat(dfd, name, flags | LOOKUP_REVAL, nd); | |
2074 | ||
2075 | if (likely(!retval)) | |
2076 | audit_inode(name, nd->path.dentry, LOOKUP_PARENT); | |
2077 | restore_nameidata(saved_nd); | |
2078 | return retval; | |
2079 | } | |
2080 | ||
79714f72 AV |
2081 | /* does lookup, returns the object with parent locked */ |
2082 | struct dentry *kern_path_locked(const char *name, struct path *path) | |
5590ff0d | 2083 | { |
51689104 | 2084 | struct filename *filename = getname_kernel(name); |
79714f72 AV |
2085 | struct nameidata nd; |
2086 | struct dentry *d; | |
51689104 PM |
2087 | int err; |
2088 | ||
2089 | if (IS_ERR(filename)) | |
2090 | return ERR_CAST(filename); | |
2091 | ||
8bcb77fa | 2092 | err = filename_parentat(AT_FDCWD, filename, 0, &nd); |
51689104 PM |
2093 | if (err) { |
2094 | d = ERR_PTR(err); | |
2095 | goto out; | |
2096 | } | |
79714f72 AV |
2097 | if (nd.last_type != LAST_NORM) { |
2098 | path_put(&nd.path); | |
51689104 PM |
2099 | d = ERR_PTR(-EINVAL); |
2100 | goto out; | |
79714f72 AV |
2101 | } |
2102 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); | |
1e0ea001 | 2103 | d = __lookup_hash(&nd.last, nd.path.dentry, 0); |
79714f72 AV |
2104 | if (IS_ERR(d)) { |
2105 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); | |
2106 | path_put(&nd.path); | |
51689104 | 2107 | goto out; |
79714f72 AV |
2108 | } |
2109 | *path = nd.path; | |
51689104 PM |
2110 | out: |
2111 | putname(filename); | |
79714f72 | 2112 | return d; |
5590ff0d UD |
2113 | } |
2114 | ||
d1811465 AV |
2115 | int kern_path(const char *name, unsigned int flags, struct path *path) |
2116 | { | |
2117 | struct nameidata nd; | |
74eb8cc5 AV |
2118 | struct filename *filename = getname_kernel(name); |
2119 | int res = PTR_ERR(filename); | |
2120 | ||
2121 | if (!IS_ERR(filename)) { | |
2122 | res = filename_lookup(AT_FDCWD, filename, flags, &nd); | |
2123 | putname(filename); | |
2124 | if (!res) | |
2125 | *path = nd.path; | |
2126 | } | |
d1811465 AV |
2127 | return res; |
2128 | } | |
4d359507 | 2129 | EXPORT_SYMBOL(kern_path); |
d1811465 | 2130 | |
16f18200 JJS |
2131 | /** |
2132 | * vfs_path_lookup - lookup a file path relative to a dentry-vfsmount pair | |
2133 | * @dentry: pointer to dentry of the base directory | |
2134 | * @mnt: pointer to vfs mount of the base directory | |
2135 | * @name: pointer to file name | |
2136 | * @flags: lookup flags | |
e0a01249 | 2137 | * @path: pointer to struct path to fill |
16f18200 JJS |
2138 | */ |
2139 | int vfs_path_lookup(struct dentry *dentry, struct vfsmount *mnt, | |
2140 | const char *name, unsigned int flags, | |
e0a01249 | 2141 | struct path *path) |
16f18200 | 2142 | { |
74eb8cc5 AV |
2143 | struct filename *filename = getname_kernel(name); |
2144 | int err = PTR_ERR(filename); | |
2145 | ||
e0a01249 | 2146 | BUG_ON(flags & LOOKUP_PARENT); |
74eb8cc5 AV |
2147 | |
2148 | /* the first argument of filename_lookup() is ignored with LOOKUP_ROOT */ | |
2149 | if (!IS_ERR(filename)) { | |
2150 | struct nameidata nd; | |
2151 | nd.root.dentry = dentry; | |
2152 | nd.root.mnt = mnt; | |
2153 | err = filename_lookup(AT_FDCWD, filename, | |
2154 | flags | LOOKUP_ROOT, &nd); | |
2155 | if (!err) | |
2156 | *path = nd.path; | |
2157 | putname(filename); | |
2158 | } | |
e0a01249 | 2159 | return err; |
16f18200 | 2160 | } |
4d359507 | 2161 | EXPORT_SYMBOL(vfs_path_lookup); |
16f18200 | 2162 | |
eead1911 | 2163 | /** |
a6b91919 | 2164 | * lookup_one_len - filesystem helper to lookup single pathname component |
eead1911 CH |
2165 | * @name: pathname component to lookup |
2166 | * @base: base directory to lookup from | |
2167 | * @len: maximum length @len should be interpreted to | |
2168 | * | |
a6b91919 | 2169 | * Note that this routine is purely a helper for filesystem usage and should |
9e7543e9 | 2170 | * not be called by generic code. |
eead1911 | 2171 | */ |
057f6c01 JM |
2172 | struct dentry *lookup_one_len(const char *name, struct dentry *base, int len) |
2173 | { | |
057f6c01 | 2174 | struct qstr this; |
6a96ba54 | 2175 | unsigned int c; |
cda309de | 2176 | int err; |
057f6c01 | 2177 | |
2f9092e1 DW |
2178 | WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex)); |
2179 | ||
6a96ba54 AV |
2180 | this.name = name; |
2181 | this.len = len; | |
0145acc2 | 2182 | this.hash = full_name_hash(name, len); |
6a96ba54 AV |
2183 | if (!len) |
2184 | return ERR_PTR(-EACCES); | |
2185 | ||
21d8a15a AV |
2186 | if (unlikely(name[0] == '.')) { |
2187 | if (len < 2 || (len == 2 && name[1] == '.')) | |
2188 | return ERR_PTR(-EACCES); | |
2189 | } | |
2190 | ||
6a96ba54 AV |
2191 | while (len--) { |
2192 | c = *(const unsigned char *)name++; | |
2193 | if (c == '/' || c == '\0') | |
2194 | return ERR_PTR(-EACCES); | |
6a96ba54 | 2195 | } |
5a202bcd AV |
2196 | /* |
2197 | * See if the low-level filesystem might want | |
2198 | * to use its own hash.. | |
2199 | */ | |
2200 | if (base->d_flags & DCACHE_OP_HASH) { | |
da53be12 | 2201 | int err = base->d_op->d_hash(base, &this); |
5a202bcd AV |
2202 | if (err < 0) |
2203 | return ERR_PTR(err); | |
2204 | } | |
eead1911 | 2205 | |
cda309de MS |
2206 | err = inode_permission(base->d_inode, MAY_EXEC); |
2207 | if (err) | |
2208 | return ERR_PTR(err); | |
2209 | ||
72bd866a | 2210 | return __lookup_hash(&this, base, 0); |
057f6c01 | 2211 | } |
4d359507 | 2212 | EXPORT_SYMBOL(lookup_one_len); |
057f6c01 | 2213 | |
1fa1e7f6 AW |
2214 | int user_path_at_empty(int dfd, const char __user *name, unsigned flags, |
2215 | struct path *path, int *empty) | |
1da177e4 | 2216 | { |
2d8f3038 | 2217 | struct nameidata nd; |
91a27b2a | 2218 | struct filename *tmp = getname_flags(name, flags, empty); |
1da177e4 | 2219 | int err = PTR_ERR(tmp); |
1da177e4 | 2220 | if (!IS_ERR(tmp)) { |
2d8f3038 AV |
2221 | |
2222 | BUG_ON(flags & LOOKUP_PARENT); | |
2223 | ||
873f1eed | 2224 | err = filename_lookup(dfd, tmp, flags, &nd); |
1da177e4 | 2225 | putname(tmp); |
2d8f3038 AV |
2226 | if (!err) |
2227 | *path = nd.path; | |
1da177e4 LT |
2228 | } |
2229 | return err; | |
2230 | } | |
2231 | ||
1fa1e7f6 AW |
2232 | int user_path_at(int dfd, const char __user *name, unsigned flags, |
2233 | struct path *path) | |
2234 | { | |
f7493e5d | 2235 | return user_path_at_empty(dfd, name, flags, path, NULL); |
1fa1e7f6 | 2236 | } |
4d359507 | 2237 | EXPORT_SYMBOL(user_path_at); |
1fa1e7f6 | 2238 | |
873f1eed JL |
2239 | /* |
2240 | * NB: most callers don't do anything directly with the reference to the | |
2241 | * to struct filename, but the nd->last pointer points into the name string | |
2242 | * allocated by getname. So we must hold the reference to it until all | |
2243 | * path-walking is complete. | |
2244 | */ | |
91a27b2a | 2245 | static struct filename * |
f5beed75 AV |
2246 | user_path_parent(int dfd, const char __user *path, |
2247 | struct path *parent, | |
2248 | struct qstr *last, | |
2249 | int *type, | |
9e790bd6 | 2250 | unsigned int flags) |
2ad94ae6 | 2251 | { |
f5beed75 | 2252 | struct nameidata nd; |
91a27b2a | 2253 | struct filename *s = getname(path); |
2ad94ae6 AV |
2254 | int error; |
2255 | ||
9e790bd6 JL |
2256 | /* only LOOKUP_REVAL is allowed in extra flags */ |
2257 | flags &= LOOKUP_REVAL; | |
2258 | ||
2ad94ae6 | 2259 | if (IS_ERR(s)) |
91a27b2a | 2260 | return s; |
2ad94ae6 | 2261 | |
8bcb77fa | 2262 | error = filename_parentat(dfd, s, flags, &nd); |
91a27b2a | 2263 | if (error) { |
2ad94ae6 | 2264 | putname(s); |
91a27b2a JL |
2265 | return ERR_PTR(error); |
2266 | } | |
f5beed75 AV |
2267 | *parent = nd.path; |
2268 | *last = nd.last; | |
2269 | *type = nd.last_type; | |
2ad94ae6 | 2270 | |
91a27b2a | 2271 | return s; |
2ad94ae6 AV |
2272 | } |
2273 | ||
8033426e | 2274 | /** |
197df04c | 2275 | * mountpoint_last - look up last component for umount |
8033426e JL |
2276 | * @nd: pathwalk nameidata - currently pointing at parent directory of "last" |
2277 | * @path: pointer to container for result | |
2278 | * | |
2279 | * This is a special lookup_last function just for umount. In this case, we | |
2280 | * need to resolve the path without doing any revalidation. | |
2281 | * | |
2282 | * The nameidata should be the result of doing a LOOKUP_PARENT pathwalk. Since | |
2283 | * mountpoints are always pinned in the dcache, their ancestors are too. Thus, | |
2284 | * in almost all cases, this lookup will be served out of the dcache. The only | |
2285 | * cases where it won't are if nd->last refers to a symlink or the path is | |
2286 | * bogus and it doesn't exist. | |
2287 | * | |
2288 | * Returns: | |
2289 | * -error: if there was an error during lookup. This includes -ENOENT if the | |
2290 | * lookup found a negative dentry. The nd->path reference will also be | |
2291 | * put in this case. | |
2292 | * | |
2293 | * 0: if we successfully resolved nd->path and found it to not to be a | |
2294 | * symlink that needs to be followed. "path" will also be populated. | |
2295 | * The nd->path reference will also be put. | |
2296 | * | |
2297 | * 1: if we successfully resolved nd->last and found it to be a symlink | |
2298 | * that needs to be followed. "path" will be populated with the path | |
2299 | * to the link, and nd->path will *not* be put. | |
2300 | */ | |
2301 | static int | |
197df04c | 2302 | mountpoint_last(struct nameidata *nd, struct path *path) |
8033426e JL |
2303 | { |
2304 | int error = 0; | |
2305 | struct dentry *dentry; | |
2306 | struct dentry *dir = nd->path.dentry; | |
2307 | ||
35759521 AV |
2308 | /* If we're in rcuwalk, drop out of it to handle last component */ |
2309 | if (nd->flags & LOOKUP_RCU) { | |
6e9918b7 | 2310 | if (unlazy_walk(nd, NULL, 0)) |
deb106c6 | 2311 | return -ECHILD; |
8033426e JL |
2312 | } |
2313 | ||
2314 | nd->flags &= ~LOOKUP_PARENT; | |
2315 | ||
2316 | if (unlikely(nd->last_type != LAST_NORM)) { | |
2317 | error = handle_dots(nd, nd->last_type); | |
35759521 | 2318 | if (error) |
deb106c6 | 2319 | return error; |
35759521 AV |
2320 | dentry = dget(nd->path.dentry); |
2321 | goto done; | |
8033426e JL |
2322 | } |
2323 | ||
2324 | mutex_lock(&dir->d_inode->i_mutex); | |
2325 | dentry = d_lookup(dir, &nd->last); | |
2326 | if (!dentry) { | |
2327 | /* | |
2328 | * No cached dentry. Mounted dentries are pinned in the cache, | |
2329 | * so that means that this dentry is probably a symlink or the | |
2330 | * path doesn't actually point to a mounted dentry. | |
2331 | */ | |
2332 | dentry = d_alloc(dir, &nd->last); | |
2333 | if (!dentry) { | |
bcceeeba | 2334 | mutex_unlock(&dir->d_inode->i_mutex); |
deb106c6 | 2335 | return -ENOMEM; |
8033426e | 2336 | } |
35759521 | 2337 | dentry = lookup_real(dir->d_inode, dentry, nd->flags); |
bcceeeba DJ |
2338 | if (IS_ERR(dentry)) { |
2339 | mutex_unlock(&dir->d_inode->i_mutex); | |
deb106c6 | 2340 | return PTR_ERR(dentry); |
bcceeeba | 2341 | } |
8033426e JL |
2342 | } |
2343 | mutex_unlock(&dir->d_inode->i_mutex); | |
2344 | ||
35759521 | 2345 | done: |
698934df | 2346 | if (d_is_negative(dentry)) { |
35759521 | 2347 | dput(dentry); |
deb106c6 | 2348 | return -ENOENT; |
8033426e | 2349 | } |
191d7f73 AV |
2350 | if (nd->depth) |
2351 | put_link(nd); | |
35759521 | 2352 | path->dentry = dentry; |
295dc39d | 2353 | path->mnt = nd->path.mnt; |
181548c0 AV |
2354 | error = should_follow_link(nd, path, nd->flags & LOOKUP_FOLLOW, |
2355 | d_backing_inode(dentry), 0); | |
deb106c6 | 2356 | if (unlikely(error)) |
d63ff28f | 2357 | return error; |
295dc39d | 2358 | mntget(path->mnt); |
35759521 | 2359 | follow_mount(path); |
deb106c6 | 2360 | return 0; |
8033426e JL |
2361 | } |
2362 | ||
2363 | /** | |
197df04c | 2364 | * path_mountpoint - look up a path to be umounted |
8033426e JL |
2365 | * @dfd: directory file descriptor to start walk from |
2366 | * @name: full pathname to walk | |
606d6fe3 | 2367 | * @path: pointer to container for result |
8033426e | 2368 | * @flags: lookup flags |
8033426e JL |
2369 | * |
2370 | * Look up the given name, but don't attempt to revalidate the last component. | |
606d6fe3 | 2371 | * Returns 0 and "path" will be valid on success; Returns error otherwise. |
8033426e JL |
2372 | */ |
2373 | static int | |
668696dc | 2374 | path_mountpoint(int dfd, const struct filename *name, struct path *path, |
46afd6f6 | 2375 | struct nameidata *nd, unsigned int flags) |
8033426e | 2376 | { |
368ee9ba AV |
2377 | const char *s = path_init(dfd, name, flags, nd); |
2378 | int err; | |
2379 | if (IS_ERR(s)) | |
2380 | return PTR_ERR(s); | |
3bdba28b AV |
2381 | while (!(err = link_path_walk(s, nd)) && |
2382 | (err = mountpoint_last(nd, path)) > 0) { | |
2383 | s = trailing_symlink(nd); | |
2384 | if (IS_ERR(s)) { | |
2385 | err = PTR_ERR(s); | |
8033426e | 2386 | break; |
3bdba28b | 2387 | } |
8033426e | 2388 | } |
deb106c6 | 2389 | terminate_walk(nd); |
46afd6f6 | 2390 | path_cleanup(nd); |
8033426e JL |
2391 | return err; |
2392 | } | |
2393 | ||
2d864651 | 2394 | static int |
668696dc | 2395 | filename_mountpoint(int dfd, struct filename *name, struct path *path, |
2d864651 AV |
2396 | unsigned int flags) |
2397 | { | |
756daf26 | 2398 | struct nameidata nd, *saved; |
cbaab2db | 2399 | int error; |
668696dc AV |
2400 | if (IS_ERR(name)) |
2401 | return PTR_ERR(name); | |
756daf26 | 2402 | saved = set_nameidata(&nd); |
46afd6f6 | 2403 | error = path_mountpoint(dfd, name, path, &nd, flags | LOOKUP_RCU); |
2d864651 | 2404 | if (unlikely(error == -ECHILD)) |
46afd6f6 | 2405 | error = path_mountpoint(dfd, name, path, &nd, flags); |
2d864651 | 2406 | if (unlikely(error == -ESTALE)) |
46afd6f6 | 2407 | error = path_mountpoint(dfd, name, path, &nd, flags | LOOKUP_REVAL); |
2d864651 | 2408 | if (likely(!error)) |
668696dc | 2409 | audit_inode(name, path->dentry, 0); |
756daf26 | 2410 | restore_nameidata(saved); |
668696dc | 2411 | putname(name); |
2d864651 AV |
2412 | return error; |
2413 | } | |
2414 | ||
8033426e | 2415 | /** |
197df04c | 2416 | * user_path_mountpoint_at - lookup a path from userland in order to umount it |
8033426e JL |
2417 | * @dfd: directory file descriptor |
2418 | * @name: pathname from userland | |
2419 | * @flags: lookup flags | |
2420 | * @path: pointer to container to hold result | |
2421 | * | |
2422 | * A umount is a special case for path walking. We're not actually interested | |
2423 | * in the inode in this situation, and ESTALE errors can be a problem. We | |
2424 | * simply want track down the dentry and vfsmount attached at the mountpoint | |
2425 | * and avoid revalidating the last component. | |
2426 | * | |
2427 | * Returns 0 and populates "path" on success. | |
2428 | */ | |
2429 | int | |
197df04c | 2430 | user_path_mountpoint_at(int dfd, const char __user *name, unsigned int flags, |
8033426e JL |
2431 | struct path *path) |
2432 | { | |
cbaab2db | 2433 | return filename_mountpoint(dfd, getname(name), path, flags); |
8033426e JL |
2434 | } |
2435 | ||
2d864651 AV |
2436 | int |
2437 | kern_path_mountpoint(int dfd, const char *name, struct path *path, | |
2438 | unsigned int flags) | |
2439 | { | |
cbaab2db | 2440 | return filename_mountpoint(dfd, getname_kernel(name), path, flags); |
2d864651 AV |
2441 | } |
2442 | EXPORT_SYMBOL(kern_path_mountpoint); | |
2443 | ||
cbdf35bc | 2444 | int __check_sticky(struct inode *dir, struct inode *inode) |
1da177e4 | 2445 | { |
8e96e3b7 | 2446 | kuid_t fsuid = current_fsuid(); |
da9592ed | 2447 | |
8e96e3b7 | 2448 | if (uid_eq(inode->i_uid, fsuid)) |
1da177e4 | 2449 | return 0; |
8e96e3b7 | 2450 | if (uid_eq(dir->i_uid, fsuid)) |
1da177e4 | 2451 | return 0; |
23adbe12 | 2452 | return !capable_wrt_inode_uidgid(inode, CAP_FOWNER); |
1da177e4 | 2453 | } |
cbdf35bc | 2454 | EXPORT_SYMBOL(__check_sticky); |
1da177e4 LT |
2455 | |
2456 | /* | |
2457 | * Check whether we can remove a link victim from directory dir, check | |
2458 | * whether the type of victim is right. | |
2459 | * 1. We can't do it if dir is read-only (done in permission()) | |
2460 | * 2. We should have write and exec permissions on dir | |
2461 | * 3. We can't remove anything from append-only dir | |
2462 | * 4. We can't do anything with immutable dir (done in permission()) | |
2463 | * 5. If the sticky bit on dir is set we should either | |
2464 | * a. be owner of dir, or | |
2465 | * b. be owner of victim, or | |
2466 | * c. have CAP_FOWNER capability | |
2467 | * 6. If the victim is append-only or immutable we can't do antyhing with | |
2468 | * links pointing to it. | |
2469 | * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR. | |
2470 | * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR. | |
2471 | * 9. We can't remove a root or mountpoint. | |
2472 | * 10. We don't allow removal of NFS sillyrenamed files; it's handled by | |
2473 | * nfs_async_unlink(). | |
2474 | */ | |
b18825a7 | 2475 | static int may_delete(struct inode *dir, struct dentry *victim, bool isdir) |
1da177e4 | 2476 | { |
63afdfc7 | 2477 | struct inode *inode = d_backing_inode(victim); |
1da177e4 LT |
2478 | int error; |
2479 | ||
b18825a7 | 2480 | if (d_is_negative(victim)) |
1da177e4 | 2481 | return -ENOENT; |
b18825a7 | 2482 | BUG_ON(!inode); |
1da177e4 LT |
2483 | |
2484 | BUG_ON(victim->d_parent->d_inode != dir); | |
4fa6b5ec | 2485 | audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE); |
1da177e4 | 2486 | |
f419a2e3 | 2487 | error = inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2488 | if (error) |
2489 | return error; | |
2490 | if (IS_APPEND(dir)) | |
2491 | return -EPERM; | |
b18825a7 DH |
2492 | |
2493 | if (check_sticky(dir, inode) || IS_APPEND(inode) || | |
2494 | IS_IMMUTABLE(inode) || IS_SWAPFILE(inode)) | |
1da177e4 LT |
2495 | return -EPERM; |
2496 | if (isdir) { | |
44b1d530 | 2497 | if (!d_is_dir(victim)) |
1da177e4 LT |
2498 | return -ENOTDIR; |
2499 | if (IS_ROOT(victim)) | |
2500 | return -EBUSY; | |
44b1d530 | 2501 | } else if (d_is_dir(victim)) |
1da177e4 LT |
2502 | return -EISDIR; |
2503 | if (IS_DEADDIR(dir)) | |
2504 | return -ENOENT; | |
2505 | if (victim->d_flags & DCACHE_NFSFS_RENAMED) | |
2506 | return -EBUSY; | |
2507 | return 0; | |
2508 | } | |
2509 | ||
2510 | /* Check whether we can create an object with dentry child in directory | |
2511 | * dir. | |
2512 | * 1. We can't do it if child already exists (open has special treatment for | |
2513 | * this case, but since we are inlined it's OK) | |
2514 | * 2. We can't do it if dir is read-only (done in permission()) | |
2515 | * 3. We should have write and exec permissions on dir | |
2516 | * 4. We can't do it if dir is immutable (done in permission()) | |
2517 | */ | |
a95164d9 | 2518 | static inline int may_create(struct inode *dir, struct dentry *child) |
1da177e4 | 2519 | { |
14e972b4 | 2520 | audit_inode_child(dir, child, AUDIT_TYPE_CHILD_CREATE); |
1da177e4 LT |
2521 | if (child->d_inode) |
2522 | return -EEXIST; | |
2523 | if (IS_DEADDIR(dir)) | |
2524 | return -ENOENT; | |
f419a2e3 | 2525 | return inode_permission(dir, MAY_WRITE | MAY_EXEC); |
1da177e4 LT |
2526 | } |
2527 | ||
1da177e4 LT |
2528 | /* |
2529 | * p1 and p2 should be directories on the same fs. | |
2530 | */ | |
2531 | struct dentry *lock_rename(struct dentry *p1, struct dentry *p2) | |
2532 | { | |
2533 | struct dentry *p; | |
2534 | ||
2535 | if (p1 == p2) { | |
f2eace23 | 2536 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
1da177e4 LT |
2537 | return NULL; |
2538 | } | |
2539 | ||
a11f3a05 | 2540 | mutex_lock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 | 2541 | |
e2761a11 OH |
2542 | p = d_ancestor(p2, p1); |
2543 | if (p) { | |
2544 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT); | |
2545 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_CHILD); | |
2546 | return p; | |
1da177e4 LT |
2547 | } |
2548 | ||
e2761a11 OH |
2549 | p = d_ancestor(p1, p2); |
2550 | if (p) { | |
2551 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); | |
2552 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_CHILD); | |
2553 | return p; | |
1da177e4 LT |
2554 | } |
2555 | ||
f2eace23 | 2556 | mutex_lock_nested(&p1->d_inode->i_mutex, I_MUTEX_PARENT); |
d1b72cc6 | 2557 | mutex_lock_nested(&p2->d_inode->i_mutex, I_MUTEX_PARENT2); |
1da177e4 LT |
2558 | return NULL; |
2559 | } | |
4d359507 | 2560 | EXPORT_SYMBOL(lock_rename); |
1da177e4 LT |
2561 | |
2562 | void unlock_rename(struct dentry *p1, struct dentry *p2) | |
2563 | { | |
1b1dcc1b | 2564 | mutex_unlock(&p1->d_inode->i_mutex); |
1da177e4 | 2565 | if (p1 != p2) { |
1b1dcc1b | 2566 | mutex_unlock(&p2->d_inode->i_mutex); |
a11f3a05 | 2567 | mutex_unlock(&p1->d_inode->i_sb->s_vfs_rename_mutex); |
1da177e4 LT |
2568 | } |
2569 | } | |
4d359507 | 2570 | EXPORT_SYMBOL(unlock_rename); |
1da177e4 | 2571 | |
4acdaf27 | 2572 | int vfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
312b63fb | 2573 | bool want_excl) |
1da177e4 | 2574 | { |
a95164d9 | 2575 | int error = may_create(dir, dentry); |
1da177e4 LT |
2576 | if (error) |
2577 | return error; | |
2578 | ||
acfa4380 | 2579 | if (!dir->i_op->create) |
1da177e4 LT |
2580 | return -EACCES; /* shouldn't it be ENOSYS? */ |
2581 | mode &= S_IALLUGO; | |
2582 | mode |= S_IFREG; | |
2583 | error = security_inode_create(dir, dentry, mode); | |
2584 | if (error) | |
2585 | return error; | |
312b63fb | 2586 | error = dir->i_op->create(dir, dentry, mode, want_excl); |
a74574aa | 2587 | if (!error) |
f38aa942 | 2588 | fsnotify_create(dir, dentry); |
1da177e4 LT |
2589 | return error; |
2590 | } | |
4d359507 | 2591 | EXPORT_SYMBOL(vfs_create); |
1da177e4 | 2592 | |
73d049a4 | 2593 | static int may_open(struct path *path, int acc_mode, int flag) |
1da177e4 | 2594 | { |
3fb64190 | 2595 | struct dentry *dentry = path->dentry; |
1da177e4 LT |
2596 | struct inode *inode = dentry->d_inode; |
2597 | int error; | |
2598 | ||
bcda7652 AV |
2599 | /* O_PATH? */ |
2600 | if (!acc_mode) | |
2601 | return 0; | |
2602 | ||
1da177e4 LT |
2603 | if (!inode) |
2604 | return -ENOENT; | |
2605 | ||
c8fe8f30 CH |
2606 | switch (inode->i_mode & S_IFMT) { |
2607 | case S_IFLNK: | |
1da177e4 | 2608 | return -ELOOP; |
c8fe8f30 CH |
2609 | case S_IFDIR: |
2610 | if (acc_mode & MAY_WRITE) | |
2611 | return -EISDIR; | |
2612 | break; | |
2613 | case S_IFBLK: | |
2614 | case S_IFCHR: | |
3fb64190 | 2615 | if (path->mnt->mnt_flags & MNT_NODEV) |
1da177e4 | 2616 | return -EACCES; |
c8fe8f30 CH |
2617 | /*FALLTHRU*/ |
2618 | case S_IFIFO: | |
2619 | case S_IFSOCK: | |
1da177e4 | 2620 | flag &= ~O_TRUNC; |
c8fe8f30 | 2621 | break; |
4a3fd211 | 2622 | } |
b41572e9 | 2623 | |
3fb64190 | 2624 | error = inode_permission(inode, acc_mode); |
b41572e9 DH |
2625 | if (error) |
2626 | return error; | |
6146f0d5 | 2627 | |
1da177e4 LT |
2628 | /* |
2629 | * An append-only file must be opened in append mode for writing. | |
2630 | */ | |
2631 | if (IS_APPEND(inode)) { | |
8737c930 | 2632 | if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND)) |
7715b521 | 2633 | return -EPERM; |
1da177e4 | 2634 | if (flag & O_TRUNC) |
7715b521 | 2635 | return -EPERM; |
1da177e4 LT |
2636 | } |
2637 | ||
2638 | /* O_NOATIME can only be set by the owner or superuser */ | |
2e149670 | 2639 | if (flag & O_NOATIME && !inode_owner_or_capable(inode)) |
7715b521 | 2640 | return -EPERM; |
1da177e4 | 2641 | |
f3c7691e | 2642 | return 0; |
7715b521 | 2643 | } |
1da177e4 | 2644 | |
e1181ee6 | 2645 | static int handle_truncate(struct file *filp) |
7715b521 | 2646 | { |
e1181ee6 | 2647 | struct path *path = &filp->f_path; |
7715b521 AV |
2648 | struct inode *inode = path->dentry->d_inode; |
2649 | int error = get_write_access(inode); | |
2650 | if (error) | |
2651 | return error; | |
2652 | /* | |
2653 | * Refuse to truncate files with mandatory locks held on them. | |
2654 | */ | |
d7a06983 | 2655 | error = locks_verify_locked(filp); |
7715b521 | 2656 | if (!error) |
ea0d3ab2 | 2657 | error = security_path_truncate(path); |
7715b521 AV |
2658 | if (!error) { |
2659 | error = do_truncate(path->dentry, 0, | |
2660 | ATTR_MTIME|ATTR_CTIME|ATTR_OPEN, | |
e1181ee6 | 2661 | filp); |
7715b521 AV |
2662 | } |
2663 | put_write_access(inode); | |
acd0c935 | 2664 | return error; |
1da177e4 LT |
2665 | } |
2666 | ||
d57999e1 DH |
2667 | static inline int open_to_namei_flags(int flag) |
2668 | { | |
8a5e929d AV |
2669 | if ((flag & O_ACCMODE) == 3) |
2670 | flag--; | |
d57999e1 DH |
2671 | return flag; |
2672 | } | |
2673 | ||
d18e9008 MS |
2674 | static int may_o_create(struct path *dir, struct dentry *dentry, umode_t mode) |
2675 | { | |
2676 | int error = security_path_mknod(dir, dentry, mode, 0); | |
2677 | if (error) | |
2678 | return error; | |
2679 | ||
2680 | error = inode_permission(dir->dentry->d_inode, MAY_WRITE | MAY_EXEC); | |
2681 | if (error) | |
2682 | return error; | |
2683 | ||
2684 | return security_inode_create(dir->dentry->d_inode, dentry, mode); | |
2685 | } | |
2686 | ||
1acf0af9 DH |
2687 | /* |
2688 | * Attempt to atomically look up, create and open a file from a negative | |
2689 | * dentry. | |
2690 | * | |
2691 | * Returns 0 if successful. The file will have been created and attached to | |
2692 | * @file by the filesystem calling finish_open(). | |
2693 | * | |
2694 | * Returns 1 if the file was looked up only or didn't need creating. The | |
2695 | * caller will need to perform the open themselves. @path will have been | |
2696 | * updated to point to the new dentry. This may be negative. | |
2697 | * | |
2698 | * Returns an error code otherwise. | |
2699 | */ | |
2675a4eb AV |
2700 | static int atomic_open(struct nameidata *nd, struct dentry *dentry, |
2701 | struct path *path, struct file *file, | |
2702 | const struct open_flags *op, | |
64894cf8 | 2703 | bool got_write, bool need_lookup, |
2675a4eb | 2704 | int *opened) |
d18e9008 MS |
2705 | { |
2706 | struct inode *dir = nd->path.dentry->d_inode; | |
2707 | unsigned open_flag = open_to_namei_flags(op->open_flag); | |
2708 | umode_t mode; | |
2709 | int error; | |
2710 | int acc_mode; | |
d18e9008 MS |
2711 | int create_error = 0; |
2712 | struct dentry *const DENTRY_NOT_SET = (void *) -1UL; | |
116cc022 | 2713 | bool excl; |
d18e9008 MS |
2714 | |
2715 | BUG_ON(dentry->d_inode); | |
2716 | ||
2717 | /* Don't create child dentry for a dead directory. */ | |
2718 | if (unlikely(IS_DEADDIR(dir))) { | |
2675a4eb | 2719 | error = -ENOENT; |
d18e9008 MS |
2720 | goto out; |
2721 | } | |
2722 | ||
62b259d8 | 2723 | mode = op->mode; |
d18e9008 MS |
2724 | if ((open_flag & O_CREAT) && !IS_POSIXACL(dir)) |
2725 | mode &= ~current_umask(); | |
2726 | ||
116cc022 MS |
2727 | excl = (open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT); |
2728 | if (excl) | |
d18e9008 | 2729 | open_flag &= ~O_TRUNC; |
d18e9008 MS |
2730 | |
2731 | /* | |
2732 | * Checking write permission is tricky, bacuse we don't know if we are | |
2733 | * going to actually need it: O_CREAT opens should work as long as the | |
2734 | * file exists. But checking existence breaks atomicity. The trick is | |
2735 | * to check access and if not granted clear O_CREAT from the flags. | |
2736 | * | |
2737 | * Another problem is returing the "right" error value (e.g. for an | |
2738 | * O_EXCL open we want to return EEXIST not EROFS). | |
2739 | */ | |
64894cf8 AV |
2740 | if (((open_flag & (O_CREAT | O_TRUNC)) || |
2741 | (open_flag & O_ACCMODE) != O_RDONLY) && unlikely(!got_write)) { | |
2742 | if (!(open_flag & O_CREAT)) { | |
d18e9008 MS |
2743 | /* |
2744 | * No O_CREATE -> atomicity not a requirement -> fall | |
2745 | * back to lookup + open | |
2746 | */ | |
2747 | goto no_open; | |
2748 | } else if (open_flag & (O_EXCL | O_TRUNC)) { | |
2749 | /* Fall back and fail with the right error */ | |
64894cf8 | 2750 | create_error = -EROFS; |
d18e9008 MS |
2751 | goto no_open; |
2752 | } else { | |
2753 | /* No side effects, safe to clear O_CREAT */ | |
64894cf8 | 2754 | create_error = -EROFS; |
d18e9008 MS |
2755 | open_flag &= ~O_CREAT; |
2756 | } | |
2757 | } | |
2758 | ||
2759 | if (open_flag & O_CREAT) { | |
38227f78 | 2760 | error = may_o_create(&nd->path, dentry, mode); |
d18e9008 MS |
2761 | if (error) { |
2762 | create_error = error; | |
2763 | if (open_flag & O_EXCL) | |
2764 | goto no_open; | |
2765 | open_flag &= ~O_CREAT; | |
2766 | } | |
2767 | } | |
2768 | ||
2769 | if (nd->flags & LOOKUP_DIRECTORY) | |
2770 | open_flag |= O_DIRECTORY; | |
2771 | ||
30d90494 AV |
2772 | file->f_path.dentry = DENTRY_NOT_SET; |
2773 | file->f_path.mnt = nd->path.mnt; | |
2774 | error = dir->i_op->atomic_open(dir, dentry, file, open_flag, mode, | |
47237687 | 2775 | opened); |
d9585277 | 2776 | if (error < 0) { |
d9585277 AV |
2777 | if (create_error && error == -ENOENT) |
2778 | error = create_error; | |
d18e9008 MS |
2779 | goto out; |
2780 | } | |
2781 | ||
d9585277 | 2782 | if (error) { /* returned 1, that is */ |
30d90494 | 2783 | if (WARN_ON(file->f_path.dentry == DENTRY_NOT_SET)) { |
2675a4eb | 2784 | error = -EIO; |
d18e9008 MS |
2785 | goto out; |
2786 | } | |
30d90494 | 2787 | if (file->f_path.dentry) { |
d18e9008 | 2788 | dput(dentry); |
30d90494 | 2789 | dentry = file->f_path.dentry; |
d18e9008 | 2790 | } |
03da633a AV |
2791 | if (*opened & FILE_CREATED) |
2792 | fsnotify_create(dir, dentry); | |
2793 | if (!dentry->d_inode) { | |
2794 | WARN_ON(*opened & FILE_CREATED); | |
2795 | if (create_error) { | |
2796 | error = create_error; | |
2797 | goto out; | |
2798 | } | |
2799 | } else { | |
2800 | if (excl && !(*opened & FILE_CREATED)) { | |
2801 | error = -EEXIST; | |
2802 | goto out; | |
2803 | } | |
62b2ce96 | 2804 | } |
d18e9008 MS |
2805 | goto looked_up; |
2806 | } | |
2807 | ||
2808 | /* | |
2809 | * We didn't have the inode before the open, so check open permission | |
2810 | * here. | |
2811 | */ | |
03da633a AV |
2812 | acc_mode = op->acc_mode; |
2813 | if (*opened & FILE_CREATED) { | |
2814 | WARN_ON(!(open_flag & O_CREAT)); | |
2815 | fsnotify_create(dir, dentry); | |
2816 | acc_mode = MAY_OPEN; | |
2817 | } | |
2675a4eb AV |
2818 | error = may_open(&file->f_path, acc_mode, open_flag); |
2819 | if (error) | |
2820 | fput(file); | |
d18e9008 MS |
2821 | |
2822 | out: | |
2823 | dput(dentry); | |
2675a4eb | 2824 | return error; |
d18e9008 | 2825 | |
d18e9008 MS |
2826 | no_open: |
2827 | if (need_lookup) { | |
72bd866a | 2828 | dentry = lookup_real(dir, dentry, nd->flags); |
d18e9008 | 2829 | if (IS_ERR(dentry)) |
2675a4eb | 2830 | return PTR_ERR(dentry); |
d18e9008 MS |
2831 | |
2832 | if (create_error) { | |
2833 | int open_flag = op->open_flag; | |
2834 | ||
2675a4eb | 2835 | error = create_error; |
d18e9008 MS |
2836 | if ((open_flag & O_EXCL)) { |
2837 | if (!dentry->d_inode) | |
2838 | goto out; | |
2839 | } else if (!dentry->d_inode) { | |
2840 | goto out; | |
2841 | } else if ((open_flag & O_TRUNC) && | |
e36cb0b8 | 2842 | d_is_reg(dentry)) { |
d18e9008 MS |
2843 | goto out; |
2844 | } | |
2845 | /* will fail later, go on to get the right error */ | |
2846 | } | |
2847 | } | |
2848 | looked_up: | |
2849 | path->dentry = dentry; | |
2850 | path->mnt = nd->path.mnt; | |
2675a4eb | 2851 | return 1; |
d18e9008 MS |
2852 | } |
2853 | ||
d58ffd35 | 2854 | /* |
1acf0af9 | 2855 | * Look up and maybe create and open the last component. |
d58ffd35 MS |
2856 | * |
2857 | * Must be called with i_mutex held on parent. | |
2858 | * | |
1acf0af9 DH |
2859 | * Returns 0 if the file was successfully atomically created (if necessary) and |
2860 | * opened. In this case the file will be returned attached to @file. | |
2861 | * | |
2862 | * Returns 1 if the file was not completely opened at this time, though lookups | |
2863 | * and creations will have been performed and the dentry returned in @path will | |
2864 | * be positive upon return if O_CREAT was specified. If O_CREAT wasn't | |
2865 | * specified then a negative dentry may be returned. | |
2866 | * | |
2867 | * An error code is returned otherwise. | |
2868 | * | |
2869 | * FILE_CREATE will be set in @*opened if the dentry was created and will be | |
2870 | * cleared otherwise prior to returning. | |
d58ffd35 | 2871 | */ |
2675a4eb AV |
2872 | static int lookup_open(struct nameidata *nd, struct path *path, |
2873 | struct file *file, | |
2874 | const struct open_flags *op, | |
64894cf8 | 2875 | bool got_write, int *opened) |
d58ffd35 MS |
2876 | { |
2877 | struct dentry *dir = nd->path.dentry; | |
54ef4872 | 2878 | struct inode *dir_inode = dir->d_inode; |
d58ffd35 MS |
2879 | struct dentry *dentry; |
2880 | int error; | |
54ef4872 | 2881 | bool need_lookup; |
d58ffd35 | 2882 | |
47237687 | 2883 | *opened &= ~FILE_CREATED; |
201f956e | 2884 | dentry = lookup_dcache(&nd->last, dir, nd->flags, &need_lookup); |
d58ffd35 | 2885 | if (IS_ERR(dentry)) |
2675a4eb | 2886 | return PTR_ERR(dentry); |
d58ffd35 | 2887 | |
d18e9008 MS |
2888 | /* Cached positive dentry: will open in f_op->open */ |
2889 | if (!need_lookup && dentry->d_inode) | |
2890 | goto out_no_open; | |
2891 | ||
2892 | if ((nd->flags & LOOKUP_OPEN) && dir_inode->i_op->atomic_open) { | |
64894cf8 | 2893 | return atomic_open(nd, dentry, path, file, op, got_write, |
47237687 | 2894 | need_lookup, opened); |
d18e9008 MS |
2895 | } |
2896 | ||
54ef4872 MS |
2897 | if (need_lookup) { |
2898 | BUG_ON(dentry->d_inode); | |
2899 | ||
72bd866a | 2900 | dentry = lookup_real(dir_inode, dentry, nd->flags); |
54ef4872 | 2901 | if (IS_ERR(dentry)) |
2675a4eb | 2902 | return PTR_ERR(dentry); |
54ef4872 MS |
2903 | } |
2904 | ||
d58ffd35 MS |
2905 | /* Negative dentry, just create the file */ |
2906 | if (!dentry->d_inode && (op->open_flag & O_CREAT)) { | |
2907 | umode_t mode = op->mode; | |
2908 | if (!IS_POSIXACL(dir->d_inode)) | |
2909 | mode &= ~current_umask(); | |
2910 | /* | |
2911 | * This write is needed to ensure that a | |
2912 | * rw->ro transition does not occur between | |
2913 | * the time when the file is created and when | |
2914 | * a permanent write count is taken through | |
015c3bbc | 2915 | * the 'struct file' in finish_open(). |
d58ffd35 | 2916 | */ |
64894cf8 AV |
2917 | if (!got_write) { |
2918 | error = -EROFS; | |
d58ffd35 | 2919 | goto out_dput; |
64894cf8 | 2920 | } |
47237687 | 2921 | *opened |= FILE_CREATED; |
d58ffd35 MS |
2922 | error = security_path_mknod(&nd->path, dentry, mode, 0); |
2923 | if (error) | |
2924 | goto out_dput; | |
312b63fb AV |
2925 | error = vfs_create(dir->d_inode, dentry, mode, |
2926 | nd->flags & LOOKUP_EXCL); | |
d58ffd35 MS |
2927 | if (error) |
2928 | goto out_dput; | |
2929 | } | |
d18e9008 | 2930 | out_no_open: |
d58ffd35 MS |
2931 | path->dentry = dentry; |
2932 | path->mnt = nd->path.mnt; | |
2675a4eb | 2933 | return 1; |
d58ffd35 MS |
2934 | |
2935 | out_dput: | |
2936 | dput(dentry); | |
2675a4eb | 2937 | return error; |
d58ffd35 MS |
2938 | } |
2939 | ||
31e6b01f | 2940 | /* |
fe2d35ff | 2941 | * Handle the last step of open() |
31e6b01f | 2942 | */ |
896475d5 | 2943 | static int do_last(struct nameidata *nd, |
2675a4eb | 2944 | struct file *file, const struct open_flags *op, |
669abf4e | 2945 | int *opened, struct filename *name) |
fb1cc555 | 2946 | { |
a1e28038 | 2947 | struct dentry *dir = nd->path.dentry; |
ca344a89 | 2948 | int open_flag = op->open_flag; |
77d660a8 | 2949 | bool will_truncate = (open_flag & O_TRUNC) != 0; |
64894cf8 | 2950 | bool got_write = false; |
bcda7652 | 2951 | int acc_mode = op->acc_mode; |
254cf582 | 2952 | unsigned seq; |
a1eb3315 | 2953 | struct inode *inode; |
16b1c1cd | 2954 | struct path save_parent = { .dentry = NULL, .mnt = NULL }; |
896475d5 | 2955 | struct path path; |
16b1c1cd | 2956 | bool retried = false; |
16c2cd71 | 2957 | int error; |
1f36f774 | 2958 | |
c3e380b0 AV |
2959 | nd->flags &= ~LOOKUP_PARENT; |
2960 | nd->flags |= op->intent; | |
2961 | ||
bc77daa7 | 2962 | if (nd->last_type != LAST_NORM) { |
fe2d35ff | 2963 | error = handle_dots(nd, nd->last_type); |
deb106c6 | 2964 | if (unlikely(error)) |
2675a4eb | 2965 | return error; |
e83db167 | 2966 | goto finish_open; |
1f36f774 | 2967 | } |
67ee3ad2 | 2968 | |
ca344a89 | 2969 | if (!(open_flag & O_CREAT)) { |
fe2d35ff AV |
2970 | if (nd->last.name[nd->last.len]) |
2971 | nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY; | |
2972 | /* we _can_ be in RCU mode here */ | |
254cf582 | 2973 | error = lookup_fast(nd, &path, &inode, &seq); |
71574865 MS |
2974 | if (likely(!error)) |
2975 | goto finish_lookup; | |
2976 | ||
2977 | if (error < 0) | |
deb106c6 | 2978 | return error; |
71574865 MS |
2979 | |
2980 | BUG_ON(nd->inode != dir->d_inode); | |
b6183df7 MS |
2981 | } else { |
2982 | /* create side of things */ | |
2983 | /* | |
2984 | * This will *only* deal with leaving RCU mode - LOOKUP_JUMPED | |
2985 | * has been cleared when we got to the last component we are | |
2986 | * about to look up | |
2987 | */ | |
2988 | error = complete_walk(nd); | |
e8bb73df | 2989 | if (error) |
2675a4eb | 2990 | return error; |
fe2d35ff | 2991 | |
33e2208a | 2992 | audit_inode(name, dir, LOOKUP_PARENT); |
b6183df7 | 2993 | /* trailing slashes? */ |
deb106c6 AV |
2994 | if (unlikely(nd->last.name[nd->last.len])) |
2995 | return -EISDIR; | |
b6183df7 | 2996 | } |
a2c36b45 | 2997 | |
16b1c1cd | 2998 | retry_lookup: |
64894cf8 AV |
2999 | if (op->open_flag & (O_CREAT | O_TRUNC | O_WRONLY | O_RDWR)) { |
3000 | error = mnt_want_write(nd->path.mnt); | |
3001 | if (!error) | |
3002 | got_write = true; | |
3003 | /* | |
3004 | * do _not_ fail yet - we might not need that or fail with | |
3005 | * a different error; let lookup_open() decide; we'll be | |
3006 | * dropping this one anyway. | |
3007 | */ | |
3008 | } | |
a1e28038 | 3009 | mutex_lock(&dir->d_inode->i_mutex); |
896475d5 | 3010 | error = lookup_open(nd, &path, file, op, got_write, opened); |
d58ffd35 | 3011 | mutex_unlock(&dir->d_inode->i_mutex); |
a1e28038 | 3012 | |
2675a4eb AV |
3013 | if (error <= 0) { |
3014 | if (error) | |
d18e9008 MS |
3015 | goto out; |
3016 | ||
47237687 | 3017 | if ((*opened & FILE_CREATED) || |
496ad9aa | 3018 | !S_ISREG(file_inode(file)->i_mode)) |
77d660a8 | 3019 | will_truncate = false; |
d18e9008 | 3020 | |
adb5c247 | 3021 | audit_inode(name, file->f_path.dentry, 0); |
d18e9008 MS |
3022 | goto opened; |
3023 | } | |
fb1cc555 | 3024 | |
47237687 | 3025 | if (*opened & FILE_CREATED) { |
9b44f1b3 | 3026 | /* Don't check for write permission, don't truncate */ |
ca344a89 | 3027 | open_flag &= ~O_TRUNC; |
77d660a8 | 3028 | will_truncate = false; |
bcda7652 | 3029 | acc_mode = MAY_OPEN; |
896475d5 | 3030 | path_to_nameidata(&path, nd); |
e83db167 | 3031 | goto finish_open_created; |
fb1cc555 AV |
3032 | } |
3033 | ||
3034 | /* | |
3134f37e | 3035 | * create/update audit record if it already exists. |
fb1cc555 | 3036 | */ |
896475d5 AV |
3037 | if (d_is_positive(path.dentry)) |
3038 | audit_inode(name, path.dentry, 0); | |
fb1cc555 | 3039 | |
d18e9008 MS |
3040 | /* |
3041 | * If atomic_open() acquired write access it is dropped now due to | |
3042 | * possible mount and symlink following (this might be optimized away if | |
3043 | * necessary...) | |
3044 | */ | |
64894cf8 | 3045 | if (got_write) { |
d18e9008 | 3046 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3047 | got_write = false; |
d18e9008 MS |
3048 | } |
3049 | ||
deb106c6 AV |
3050 | if (unlikely((open_flag & (O_EXCL | O_CREAT)) == (O_EXCL | O_CREAT))) { |
3051 | path_to_nameidata(&path, nd); | |
3052 | return -EEXIST; | |
3053 | } | |
fb1cc555 | 3054 | |
756daf26 | 3055 | error = follow_managed(&path, nd); |
deb106c6 AV |
3056 | if (unlikely(error < 0)) |
3057 | return error; | |
a3fbbde7 | 3058 | |
decf3400 | 3059 | BUG_ON(nd->flags & LOOKUP_RCU); |
63afdfc7 | 3060 | inode = d_backing_inode(path.dentry); |
254cf582 | 3061 | seq = 0; /* out of RCU mode, so the value doesn't matter */ |
deb106c6 | 3062 | if (unlikely(d_is_negative(path.dentry))) { |
896475d5 | 3063 | path_to_nameidata(&path, nd); |
deb106c6 | 3064 | return -ENOENT; |
54c33e7f | 3065 | } |
766c4cbf | 3066 | finish_lookup: |
d63ff28f AV |
3067 | if (nd->depth) |
3068 | put_link(nd); | |
181548c0 AV |
3069 | error = should_follow_link(nd, &path, nd->flags & LOOKUP_FOLLOW, |
3070 | inode, seq); | |
deb106c6 | 3071 | if (unlikely(error)) |
d63ff28f | 3072 | return error; |
fb1cc555 | 3073 | |
896475d5 AV |
3074 | if (unlikely(d_is_symlink(path.dentry)) && !(open_flag & O_PATH)) { |
3075 | path_to_nameidata(&path, nd); | |
deb106c6 | 3076 | return -ELOOP; |
a5cfe2d5 AV |
3077 | } |
3078 | ||
896475d5 AV |
3079 | if ((nd->flags & LOOKUP_RCU) || nd->path.mnt != path.mnt) { |
3080 | path_to_nameidata(&path, nd); | |
16b1c1cd MS |
3081 | } else { |
3082 | save_parent.dentry = nd->path.dentry; | |
896475d5 AV |
3083 | save_parent.mnt = mntget(path.mnt); |
3084 | nd->path.dentry = path.dentry; | |
16b1c1cd MS |
3085 | |
3086 | } | |
decf3400 | 3087 | nd->inode = inode; |
254cf582 | 3088 | nd->seq = seq; |
a3fbbde7 | 3089 | /* Why this, you ask? _Now_ we might have grown LOOKUP_JUMPED... */ |
bc77daa7 | 3090 | finish_open: |
a3fbbde7 | 3091 | error = complete_walk(nd); |
16b1c1cd MS |
3092 | if (error) { |
3093 | path_put(&save_parent); | |
2675a4eb | 3094 | return error; |
16b1c1cd | 3095 | } |
bc77daa7 | 3096 | audit_inode(name, nd->path.dentry, 0); |
fb1cc555 | 3097 | error = -EISDIR; |
44b1d530 | 3098 | if ((open_flag & O_CREAT) && d_is_dir(nd->path.dentry)) |
2675a4eb | 3099 | goto out; |
af2f5542 | 3100 | error = -ENOTDIR; |
44b1d530 | 3101 | if ((nd->flags & LOOKUP_DIRECTORY) && !d_can_lookup(nd->path.dentry)) |
2675a4eb | 3102 | goto out; |
4bbcbd3b | 3103 | if (!d_is_reg(nd->path.dentry)) |
77d660a8 | 3104 | will_truncate = false; |
6c0d46c4 | 3105 | |
0f9d1a10 AV |
3106 | if (will_truncate) { |
3107 | error = mnt_want_write(nd->path.mnt); | |
3108 | if (error) | |
2675a4eb | 3109 | goto out; |
64894cf8 | 3110 | got_write = true; |
0f9d1a10 | 3111 | } |
e83db167 | 3112 | finish_open_created: |
bcda7652 | 3113 | error = may_open(&nd->path, acc_mode, open_flag); |
ca344a89 | 3114 | if (error) |
2675a4eb | 3115 | goto out; |
4aa7c634 MS |
3116 | |
3117 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ | |
3118 | error = vfs_open(&nd->path, file, current_cred()); | |
3119 | if (!error) { | |
3120 | *opened |= FILE_OPENED; | |
3121 | } else { | |
30d90494 | 3122 | if (error == -EOPENSTALE) |
f60dc3db | 3123 | goto stale_open; |
015c3bbc | 3124 | goto out; |
f60dc3db | 3125 | } |
a8277b9b | 3126 | opened: |
2675a4eb | 3127 | error = open_check_o_direct(file); |
015c3bbc MS |
3128 | if (error) |
3129 | goto exit_fput; | |
3034a146 | 3130 | error = ima_file_check(file, op->acc_mode, *opened); |
aa4caadb MS |
3131 | if (error) |
3132 | goto exit_fput; | |
3133 | ||
3134 | if (will_truncate) { | |
2675a4eb | 3135 | error = handle_truncate(file); |
aa4caadb MS |
3136 | if (error) |
3137 | goto exit_fput; | |
0f9d1a10 | 3138 | } |
ca344a89 | 3139 | out: |
64894cf8 | 3140 | if (got_write) |
0f9d1a10 | 3141 | mnt_drop_write(nd->path.mnt); |
16b1c1cd | 3142 | path_put(&save_parent); |
2675a4eb | 3143 | return error; |
fb1cc555 | 3144 | |
015c3bbc | 3145 | exit_fput: |
2675a4eb AV |
3146 | fput(file); |
3147 | goto out; | |
015c3bbc | 3148 | |
f60dc3db MS |
3149 | stale_open: |
3150 | /* If no saved parent or already retried then can't retry */ | |
3151 | if (!save_parent.dentry || retried) | |
3152 | goto out; | |
3153 | ||
3154 | BUG_ON(save_parent.dentry != dir); | |
3155 | path_put(&nd->path); | |
3156 | nd->path = save_parent; | |
3157 | nd->inode = dir->d_inode; | |
3158 | save_parent.mnt = NULL; | |
3159 | save_parent.dentry = NULL; | |
64894cf8 | 3160 | if (got_write) { |
f60dc3db | 3161 | mnt_drop_write(nd->path.mnt); |
64894cf8 | 3162 | got_write = false; |
f60dc3db MS |
3163 | } |
3164 | retried = true; | |
3165 | goto retry_lookup; | |
fb1cc555 AV |
3166 | } |
3167 | ||
60545d0d AV |
3168 | static int do_tmpfile(int dfd, struct filename *pathname, |
3169 | struct nameidata *nd, int flags, | |
3170 | const struct open_flags *op, | |
3171 | struct file *file, int *opened) | |
3172 | { | |
3173 | static const struct qstr name = QSTR_INIT("/", 1); | |
3174 | struct dentry *dentry, *child; | |
3175 | struct inode *dir; | |
5eb6b495 | 3176 | int error = path_lookupat(dfd, pathname, |
60545d0d AV |
3177 | flags | LOOKUP_DIRECTORY, nd); |
3178 | if (unlikely(error)) | |
3179 | return error; | |
3180 | error = mnt_want_write(nd->path.mnt); | |
3181 | if (unlikely(error)) | |
3182 | goto out; | |
3183 | /* we want directory to be writable */ | |
3184 | error = inode_permission(nd->inode, MAY_WRITE | MAY_EXEC); | |
3185 | if (error) | |
3186 | goto out2; | |
3187 | dentry = nd->path.dentry; | |
3188 | dir = dentry->d_inode; | |
3189 | if (!dir->i_op->tmpfile) { | |
3190 | error = -EOPNOTSUPP; | |
3191 | goto out2; | |
3192 | } | |
3193 | child = d_alloc(dentry, &name); | |
3194 | if (unlikely(!child)) { | |
3195 | error = -ENOMEM; | |
3196 | goto out2; | |
3197 | } | |
3198 | nd->flags &= ~LOOKUP_DIRECTORY; | |
3199 | nd->flags |= op->intent; | |
3200 | dput(nd->path.dentry); | |
3201 | nd->path.dentry = child; | |
3202 | error = dir->i_op->tmpfile(dir, nd->path.dentry, op->mode); | |
3203 | if (error) | |
3204 | goto out2; | |
3205 | audit_inode(pathname, nd->path.dentry, 0); | |
69a91c23 ER |
3206 | /* Don't check for other permissions, the inode was just created */ |
3207 | error = may_open(&nd->path, MAY_OPEN, op->open_flag); | |
60545d0d AV |
3208 | if (error) |
3209 | goto out2; | |
3210 | file->f_path.mnt = nd->path.mnt; | |
3211 | error = finish_open(file, nd->path.dentry, NULL, opened); | |
3212 | if (error) | |
3213 | goto out2; | |
3214 | error = open_check_o_direct(file); | |
f4e0c30c | 3215 | if (error) { |
60545d0d | 3216 | fput(file); |
f4e0c30c AV |
3217 | } else if (!(op->open_flag & O_EXCL)) { |
3218 | struct inode *inode = file_inode(file); | |
3219 | spin_lock(&inode->i_lock); | |
3220 | inode->i_state |= I_LINKABLE; | |
3221 | spin_unlock(&inode->i_lock); | |
3222 | } | |
60545d0d AV |
3223 | out2: |
3224 | mnt_drop_write(nd->path.mnt); | |
3225 | out: | |
3226 | path_put(&nd->path); | |
3227 | return error; | |
3228 | } | |
3229 | ||
669abf4e | 3230 | static struct file *path_openat(int dfd, struct filename *pathname, |
73d049a4 | 3231 | struct nameidata *nd, const struct open_flags *op, int flags) |
1da177e4 | 3232 | { |
368ee9ba | 3233 | const char *s; |
30d90494 | 3234 | struct file *file; |
47237687 | 3235 | int opened = 0; |
13aab428 | 3236 | int error; |
31e6b01f | 3237 | |
30d90494 | 3238 | file = get_empty_filp(); |
1afc99be AV |
3239 | if (IS_ERR(file)) |
3240 | return file; | |
31e6b01f | 3241 | |
30d90494 | 3242 | file->f_flags = op->open_flag; |
31e6b01f | 3243 | |
bb458c64 | 3244 | if (unlikely(file->f_flags & __O_TMPFILE)) { |
60545d0d | 3245 | error = do_tmpfile(dfd, pathname, nd, flags, op, file, &opened); |
f15133df | 3246 | goto out2; |
60545d0d AV |
3247 | } |
3248 | ||
368ee9ba AV |
3249 | s = path_init(dfd, pathname, flags, nd); |
3250 | if (IS_ERR(s)) { | |
3251 | put_filp(file); | |
3252 | return ERR_CAST(s); | |
3253 | } | |
3bdba28b AV |
3254 | while (!(error = link_path_walk(s, nd)) && |
3255 | (error = do_last(nd, file, op, &opened, pathname)) > 0) { | |
73d049a4 | 3256 | nd->flags &= ~(LOOKUP_OPEN|LOOKUP_CREATE|LOOKUP_EXCL); |
3bdba28b AV |
3257 | s = trailing_symlink(nd); |
3258 | if (IS_ERR(s)) { | |
3259 | error = PTR_ERR(s); | |
2675a4eb | 3260 | break; |
3bdba28b | 3261 | } |
806b681c | 3262 | } |
deb106c6 | 3263 | terminate_walk(nd); |
893b7775 | 3264 | path_cleanup(nd); |
f15133df | 3265 | out2: |
2675a4eb AV |
3266 | if (!(opened & FILE_OPENED)) { |
3267 | BUG_ON(!error); | |
30d90494 | 3268 | put_filp(file); |
16b1c1cd | 3269 | } |
2675a4eb AV |
3270 | if (unlikely(error)) { |
3271 | if (error == -EOPENSTALE) { | |
3272 | if (flags & LOOKUP_RCU) | |
3273 | error = -ECHILD; | |
3274 | else | |
3275 | error = -ESTALE; | |
3276 | } | |
3277 | file = ERR_PTR(error); | |
3278 | } | |
3279 | return file; | |
1da177e4 LT |
3280 | } |
3281 | ||
669abf4e | 3282 | struct file *do_filp_open(int dfd, struct filename *pathname, |
f9652e10 | 3283 | const struct open_flags *op) |
13aab428 | 3284 | { |
756daf26 | 3285 | struct nameidata nd, *saved_nd = set_nameidata(&nd); |
f9652e10 | 3286 | int flags = op->lookup_flags; |
13aab428 AV |
3287 | struct file *filp; |
3288 | ||
73d049a4 | 3289 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_RCU); |
13aab428 | 3290 | if (unlikely(filp == ERR_PTR(-ECHILD))) |
73d049a4 | 3291 | filp = path_openat(dfd, pathname, &nd, op, flags); |
13aab428 | 3292 | if (unlikely(filp == ERR_PTR(-ESTALE))) |
73d049a4 | 3293 | filp = path_openat(dfd, pathname, &nd, op, flags | LOOKUP_REVAL); |
756daf26 | 3294 | restore_nameidata(saved_nd); |
13aab428 AV |
3295 | return filp; |
3296 | } | |
3297 | ||
73d049a4 | 3298 | struct file *do_file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
f9652e10 | 3299 | const char *name, const struct open_flags *op) |
73d049a4 | 3300 | { |
756daf26 | 3301 | struct nameidata nd, *saved_nd; |
73d049a4 | 3302 | struct file *file; |
51689104 | 3303 | struct filename *filename; |
f9652e10 | 3304 | int flags = op->lookup_flags | LOOKUP_ROOT; |
73d049a4 AV |
3305 | |
3306 | nd.root.mnt = mnt; | |
3307 | nd.root.dentry = dentry; | |
3308 | ||
b18825a7 | 3309 | if (d_is_symlink(dentry) && op->intent & LOOKUP_OPEN) |
73d049a4 AV |
3310 | return ERR_PTR(-ELOOP); |
3311 | ||
51689104 PM |
3312 | filename = getname_kernel(name); |
3313 | if (unlikely(IS_ERR(filename))) | |
3314 | return ERR_CAST(filename); | |
3315 | ||
756daf26 | 3316 | saved_nd = set_nameidata(&nd); |
51689104 | 3317 | file = path_openat(-1, filename, &nd, op, flags | LOOKUP_RCU); |
73d049a4 | 3318 | if (unlikely(file == ERR_PTR(-ECHILD))) |
51689104 | 3319 | file = path_openat(-1, filename, &nd, op, flags); |
73d049a4 | 3320 | if (unlikely(file == ERR_PTR(-ESTALE))) |
51689104 | 3321 | file = path_openat(-1, filename, &nd, op, flags | LOOKUP_REVAL); |
756daf26 | 3322 | restore_nameidata(saved_nd); |
51689104 | 3323 | putname(filename); |
73d049a4 AV |
3324 | return file; |
3325 | } | |
3326 | ||
fa14a0b8 | 3327 | static struct dentry *filename_create(int dfd, struct filename *name, |
1ac12b4b | 3328 | struct path *path, unsigned int lookup_flags) |
1da177e4 | 3329 | { |
c663e5d8 | 3330 | struct dentry *dentry = ERR_PTR(-EEXIST); |
ed75e95d | 3331 | struct nameidata nd; |
c30dabfe | 3332 | int err2; |
1ac12b4b JL |
3333 | int error; |
3334 | bool is_dir = (lookup_flags & LOOKUP_DIRECTORY); | |
3335 | ||
3336 | /* | |
3337 | * Note that only LOOKUP_REVAL and LOOKUP_DIRECTORY matter here. Any | |
3338 | * other flags passed in are ignored! | |
3339 | */ | |
3340 | lookup_flags &= LOOKUP_REVAL; | |
3341 | ||
8bcb77fa | 3342 | error = filename_parentat(dfd, name, lookup_flags, &nd); |
ed75e95d AV |
3343 | if (error) |
3344 | return ERR_PTR(error); | |
1da177e4 | 3345 | |
c663e5d8 CH |
3346 | /* |
3347 | * Yucky last component or no last component at all? | |
3348 | * (foo/., foo/.., /////) | |
3349 | */ | |
ed75e95d AV |
3350 | if (nd.last_type != LAST_NORM) |
3351 | goto out; | |
3352 | nd.flags &= ~LOOKUP_PARENT; | |
3353 | nd.flags |= LOOKUP_CREATE | LOOKUP_EXCL; | |
c663e5d8 | 3354 | |
c30dabfe JK |
3355 | /* don't fail immediately if it's r/o, at least try to report other errors */ |
3356 | err2 = mnt_want_write(nd.path.mnt); | |
c663e5d8 CH |
3357 | /* |
3358 | * Do the final lookup. | |
3359 | */ | |
ed75e95d | 3360 | mutex_lock_nested(&nd.path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
6a9f40d6 | 3361 | dentry = __lookup_hash(&nd.last, nd.path.dentry, nd.flags); |
1da177e4 | 3362 | if (IS_ERR(dentry)) |
a8104a9f | 3363 | goto unlock; |
c663e5d8 | 3364 | |
a8104a9f | 3365 | error = -EEXIST; |
b18825a7 | 3366 | if (d_is_positive(dentry)) |
a8104a9f | 3367 | goto fail; |
b18825a7 | 3368 | |
c663e5d8 CH |
3369 | /* |
3370 | * Special case - lookup gave negative, but... we had foo/bar/ | |
3371 | * From the vfs_mknod() POV we just have a negative dentry - | |
3372 | * all is fine. Let's be bastards - you had / on the end, you've | |
3373 | * been asking for (non-existent) directory. -ENOENT for you. | |
3374 | */ | |
ed75e95d | 3375 | if (unlikely(!is_dir && nd.last.name[nd.last.len])) { |
a8104a9f | 3376 | error = -ENOENT; |
ed75e95d | 3377 | goto fail; |
e9baf6e5 | 3378 | } |
c30dabfe JK |
3379 | if (unlikely(err2)) { |
3380 | error = err2; | |
a8104a9f | 3381 | goto fail; |
c30dabfe | 3382 | } |
ed75e95d | 3383 | *path = nd.path; |
1da177e4 | 3384 | return dentry; |
1da177e4 | 3385 | fail: |
a8104a9f AV |
3386 | dput(dentry); |
3387 | dentry = ERR_PTR(error); | |
3388 | unlock: | |
ed75e95d | 3389 | mutex_unlock(&nd.path.dentry->d_inode->i_mutex); |
c30dabfe JK |
3390 | if (!err2) |
3391 | mnt_drop_write(nd.path.mnt); | |
ed75e95d AV |
3392 | out: |
3393 | path_put(&nd.path); | |
1da177e4 LT |
3394 | return dentry; |
3395 | } | |
fa14a0b8 AV |
3396 | |
3397 | struct dentry *kern_path_create(int dfd, const char *pathname, | |
3398 | struct path *path, unsigned int lookup_flags) | |
3399 | { | |
51689104 PM |
3400 | struct filename *filename = getname_kernel(pathname); |
3401 | struct dentry *res; | |
3402 | ||
3403 | if (IS_ERR(filename)) | |
3404 | return ERR_CAST(filename); | |
3405 | res = filename_create(dfd, filename, path, lookup_flags); | |
3406 | putname(filename); | |
3407 | return res; | |
fa14a0b8 | 3408 | } |
dae6ad8f AV |
3409 | EXPORT_SYMBOL(kern_path_create); |
3410 | ||
921a1650 AV |
3411 | void done_path_create(struct path *path, struct dentry *dentry) |
3412 | { | |
3413 | dput(dentry); | |
3414 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
a8104a9f | 3415 | mnt_drop_write(path->mnt); |
921a1650 AV |
3416 | path_put(path); |
3417 | } | |
3418 | EXPORT_SYMBOL(done_path_create); | |
3419 | ||
1ac12b4b JL |
3420 | struct dentry *user_path_create(int dfd, const char __user *pathname, |
3421 | struct path *path, unsigned int lookup_flags) | |
dae6ad8f | 3422 | { |
91a27b2a | 3423 | struct filename *tmp = getname(pathname); |
dae6ad8f AV |
3424 | struct dentry *res; |
3425 | if (IS_ERR(tmp)) | |
3426 | return ERR_CAST(tmp); | |
fa14a0b8 | 3427 | res = filename_create(dfd, tmp, path, lookup_flags); |
dae6ad8f AV |
3428 | putname(tmp); |
3429 | return res; | |
3430 | } | |
3431 | EXPORT_SYMBOL(user_path_create); | |
3432 | ||
1a67aafb | 3433 | int vfs_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 | 3434 | { |
a95164d9 | 3435 | int error = may_create(dir, dentry); |
1da177e4 LT |
3436 | |
3437 | if (error) | |
3438 | return error; | |
3439 | ||
975d6b39 | 3440 | if ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)) |
1da177e4 LT |
3441 | return -EPERM; |
3442 | ||
acfa4380 | 3443 | if (!dir->i_op->mknod) |
1da177e4 LT |
3444 | return -EPERM; |
3445 | ||
08ce5f16 SH |
3446 | error = devcgroup_inode_mknod(mode, dev); |
3447 | if (error) | |
3448 | return error; | |
3449 | ||
1da177e4 LT |
3450 | error = security_inode_mknod(dir, dentry, mode, dev); |
3451 | if (error) | |
3452 | return error; | |
3453 | ||
1da177e4 | 3454 | error = dir->i_op->mknod(dir, dentry, mode, dev); |
a74574aa | 3455 | if (!error) |
f38aa942 | 3456 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3457 | return error; |
3458 | } | |
4d359507 | 3459 | EXPORT_SYMBOL(vfs_mknod); |
1da177e4 | 3460 | |
f69aac00 | 3461 | static int may_mknod(umode_t mode) |
463c3197 DH |
3462 | { |
3463 | switch (mode & S_IFMT) { | |
3464 | case S_IFREG: | |
3465 | case S_IFCHR: | |
3466 | case S_IFBLK: | |
3467 | case S_IFIFO: | |
3468 | case S_IFSOCK: | |
3469 | case 0: /* zero mode translates to S_IFREG */ | |
3470 | return 0; | |
3471 | case S_IFDIR: | |
3472 | return -EPERM; | |
3473 | default: | |
3474 | return -EINVAL; | |
3475 | } | |
3476 | } | |
3477 | ||
8208a22b | 3478 | SYSCALL_DEFINE4(mknodat, int, dfd, const char __user *, filename, umode_t, mode, |
2e4d0924 | 3479 | unsigned, dev) |
1da177e4 | 3480 | { |
2ad94ae6 | 3481 | struct dentry *dentry; |
dae6ad8f AV |
3482 | struct path path; |
3483 | int error; | |
972567f1 | 3484 | unsigned int lookup_flags = 0; |
1da177e4 | 3485 | |
8e4bfca1 AV |
3486 | error = may_mknod(mode); |
3487 | if (error) | |
3488 | return error; | |
972567f1 JL |
3489 | retry: |
3490 | dentry = user_path_create(dfd, filename, &path, lookup_flags); | |
dae6ad8f AV |
3491 | if (IS_ERR(dentry)) |
3492 | return PTR_ERR(dentry); | |
2ad94ae6 | 3493 | |
dae6ad8f | 3494 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3495 | mode &= ~current_umask(); |
dae6ad8f | 3496 | error = security_path_mknod(&path, dentry, mode, dev); |
be6d3e56 | 3497 | if (error) |
a8104a9f | 3498 | goto out; |
463c3197 | 3499 | switch (mode & S_IFMT) { |
1da177e4 | 3500 | case 0: case S_IFREG: |
312b63fb | 3501 | error = vfs_create(path.dentry->d_inode,dentry,mode,true); |
1da177e4 LT |
3502 | break; |
3503 | case S_IFCHR: case S_IFBLK: | |
dae6ad8f | 3504 | error = vfs_mknod(path.dentry->d_inode,dentry,mode, |
1da177e4 LT |
3505 | new_decode_dev(dev)); |
3506 | break; | |
3507 | case S_IFIFO: case S_IFSOCK: | |
dae6ad8f | 3508 | error = vfs_mknod(path.dentry->d_inode,dentry,mode,0); |
1da177e4 | 3509 | break; |
1da177e4 | 3510 | } |
a8104a9f | 3511 | out: |
921a1650 | 3512 | done_path_create(&path, dentry); |
972567f1 JL |
3513 | if (retry_estale(error, lookup_flags)) { |
3514 | lookup_flags |= LOOKUP_REVAL; | |
3515 | goto retry; | |
3516 | } | |
1da177e4 LT |
3517 | return error; |
3518 | } | |
3519 | ||
8208a22b | 3520 | SYSCALL_DEFINE3(mknod, const char __user *, filename, umode_t, mode, unsigned, dev) |
5590ff0d UD |
3521 | { |
3522 | return sys_mknodat(AT_FDCWD, filename, mode, dev); | |
3523 | } | |
3524 | ||
18bb1db3 | 3525 | int vfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 | 3526 | { |
a95164d9 | 3527 | int error = may_create(dir, dentry); |
8de52778 | 3528 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3529 | |
3530 | if (error) | |
3531 | return error; | |
3532 | ||
acfa4380 | 3533 | if (!dir->i_op->mkdir) |
1da177e4 LT |
3534 | return -EPERM; |
3535 | ||
3536 | mode &= (S_IRWXUGO|S_ISVTX); | |
3537 | error = security_inode_mkdir(dir, dentry, mode); | |
3538 | if (error) | |
3539 | return error; | |
3540 | ||
8de52778 AV |
3541 | if (max_links && dir->i_nlink >= max_links) |
3542 | return -EMLINK; | |
3543 | ||
1da177e4 | 3544 | error = dir->i_op->mkdir(dir, dentry, mode); |
a74574aa | 3545 | if (!error) |
f38aa942 | 3546 | fsnotify_mkdir(dir, dentry); |
1da177e4 LT |
3547 | return error; |
3548 | } | |
4d359507 | 3549 | EXPORT_SYMBOL(vfs_mkdir); |
1da177e4 | 3550 | |
a218d0fd | 3551 | SYSCALL_DEFINE3(mkdirat, int, dfd, const char __user *, pathname, umode_t, mode) |
1da177e4 | 3552 | { |
6902d925 | 3553 | struct dentry *dentry; |
dae6ad8f AV |
3554 | struct path path; |
3555 | int error; | |
b76d8b82 | 3556 | unsigned int lookup_flags = LOOKUP_DIRECTORY; |
1da177e4 | 3557 | |
b76d8b82 JL |
3558 | retry: |
3559 | dentry = user_path_create(dfd, pathname, &path, lookup_flags); | |
6902d925 | 3560 | if (IS_ERR(dentry)) |
dae6ad8f | 3561 | return PTR_ERR(dentry); |
1da177e4 | 3562 | |
dae6ad8f | 3563 | if (!IS_POSIXACL(path.dentry->d_inode)) |
ce3b0f8d | 3564 | mode &= ~current_umask(); |
dae6ad8f | 3565 | error = security_path_mkdir(&path, dentry, mode); |
a8104a9f AV |
3566 | if (!error) |
3567 | error = vfs_mkdir(path.dentry->d_inode, dentry, mode); | |
921a1650 | 3568 | done_path_create(&path, dentry); |
b76d8b82 JL |
3569 | if (retry_estale(error, lookup_flags)) { |
3570 | lookup_flags |= LOOKUP_REVAL; | |
3571 | goto retry; | |
3572 | } | |
1da177e4 LT |
3573 | return error; |
3574 | } | |
3575 | ||
a218d0fd | 3576 | SYSCALL_DEFINE2(mkdir, const char __user *, pathname, umode_t, mode) |
5590ff0d UD |
3577 | { |
3578 | return sys_mkdirat(AT_FDCWD, pathname, mode); | |
3579 | } | |
3580 | ||
1da177e4 | 3581 | /* |
a71905f0 | 3582 | * The dentry_unhash() helper will try to drop the dentry early: we |
c0d02594 | 3583 | * should have a usage count of 1 if we're the only user of this |
a71905f0 SW |
3584 | * dentry, and if that is true (possibly after pruning the dcache), |
3585 | * then we drop the dentry now. | |
1da177e4 LT |
3586 | * |
3587 | * A low-level filesystem can, if it choses, legally | |
3588 | * do a | |
3589 | * | |
3590 | * if (!d_unhashed(dentry)) | |
3591 | * return -EBUSY; | |
3592 | * | |
3593 | * if it cannot handle the case of removing a directory | |
3594 | * that is still in use by something else.. | |
3595 | */ | |
3596 | void dentry_unhash(struct dentry *dentry) | |
3597 | { | |
dc168427 | 3598 | shrink_dcache_parent(dentry); |
1da177e4 | 3599 | spin_lock(&dentry->d_lock); |
98474236 | 3600 | if (dentry->d_lockref.count == 1) |
1da177e4 LT |
3601 | __d_drop(dentry); |
3602 | spin_unlock(&dentry->d_lock); | |
1da177e4 | 3603 | } |
4d359507 | 3604 | EXPORT_SYMBOL(dentry_unhash); |
1da177e4 LT |
3605 | |
3606 | int vfs_rmdir(struct inode *dir, struct dentry *dentry) | |
3607 | { | |
3608 | int error = may_delete(dir, dentry, 1); | |
3609 | ||
3610 | if (error) | |
3611 | return error; | |
3612 | ||
acfa4380 | 3613 | if (!dir->i_op->rmdir) |
1da177e4 LT |
3614 | return -EPERM; |
3615 | ||
1d2ef590 | 3616 | dget(dentry); |
1b1dcc1b | 3617 | mutex_lock(&dentry->d_inode->i_mutex); |
912dbc15 SW |
3618 | |
3619 | error = -EBUSY; | |
7af1364f | 3620 | if (is_local_mountpoint(dentry)) |
912dbc15 SW |
3621 | goto out; |
3622 | ||
3623 | error = security_inode_rmdir(dir, dentry); | |
3624 | if (error) | |
3625 | goto out; | |
3626 | ||
3cebde24 | 3627 | shrink_dcache_parent(dentry); |
912dbc15 SW |
3628 | error = dir->i_op->rmdir(dir, dentry); |
3629 | if (error) | |
3630 | goto out; | |
3631 | ||
3632 | dentry->d_inode->i_flags |= S_DEAD; | |
3633 | dont_mount(dentry); | |
8ed936b5 | 3634 | detach_mounts(dentry); |
912dbc15 SW |
3635 | |
3636 | out: | |
1b1dcc1b | 3637 | mutex_unlock(&dentry->d_inode->i_mutex); |
1d2ef590 | 3638 | dput(dentry); |
912dbc15 | 3639 | if (!error) |
1da177e4 | 3640 | d_delete(dentry); |
1da177e4 LT |
3641 | return error; |
3642 | } | |
4d359507 | 3643 | EXPORT_SYMBOL(vfs_rmdir); |
1da177e4 | 3644 | |
5590ff0d | 3645 | static long do_rmdir(int dfd, const char __user *pathname) |
1da177e4 LT |
3646 | { |
3647 | int error = 0; | |
91a27b2a | 3648 | struct filename *name; |
1da177e4 | 3649 | struct dentry *dentry; |
f5beed75 AV |
3650 | struct path path; |
3651 | struct qstr last; | |
3652 | int type; | |
c6ee9206 JL |
3653 | unsigned int lookup_flags = 0; |
3654 | retry: | |
f5beed75 AV |
3655 | name = user_path_parent(dfd, pathname, |
3656 | &path, &last, &type, lookup_flags); | |
91a27b2a JL |
3657 | if (IS_ERR(name)) |
3658 | return PTR_ERR(name); | |
1da177e4 | 3659 | |
f5beed75 | 3660 | switch (type) { |
0612d9fb OH |
3661 | case LAST_DOTDOT: |
3662 | error = -ENOTEMPTY; | |
3663 | goto exit1; | |
3664 | case LAST_DOT: | |
3665 | error = -EINVAL; | |
3666 | goto exit1; | |
3667 | case LAST_ROOT: | |
3668 | error = -EBUSY; | |
3669 | goto exit1; | |
1da177e4 | 3670 | } |
0612d9fb | 3671 | |
f5beed75 | 3672 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3673 | if (error) |
3674 | goto exit1; | |
0612d9fb | 3675 | |
f5beed75 AV |
3676 | mutex_lock_nested(&path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
3677 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); | |
1da177e4 | 3678 | error = PTR_ERR(dentry); |
6902d925 DH |
3679 | if (IS_ERR(dentry)) |
3680 | goto exit2; | |
e6bc45d6 TT |
3681 | if (!dentry->d_inode) { |
3682 | error = -ENOENT; | |
3683 | goto exit3; | |
3684 | } | |
f5beed75 | 3685 | error = security_path_rmdir(&path, dentry); |
be6d3e56 | 3686 | if (error) |
c30dabfe | 3687 | goto exit3; |
f5beed75 | 3688 | error = vfs_rmdir(path.dentry->d_inode, dentry); |
0622753b | 3689 | exit3: |
6902d925 DH |
3690 | dput(dentry); |
3691 | exit2: | |
f5beed75 AV |
3692 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
3693 | mnt_drop_write(path.mnt); | |
1da177e4 | 3694 | exit1: |
f5beed75 | 3695 | path_put(&path); |
1da177e4 | 3696 | putname(name); |
c6ee9206 JL |
3697 | if (retry_estale(error, lookup_flags)) { |
3698 | lookup_flags |= LOOKUP_REVAL; | |
3699 | goto retry; | |
3700 | } | |
1da177e4 LT |
3701 | return error; |
3702 | } | |
3703 | ||
3cdad428 | 3704 | SYSCALL_DEFINE1(rmdir, const char __user *, pathname) |
5590ff0d UD |
3705 | { |
3706 | return do_rmdir(AT_FDCWD, pathname); | |
3707 | } | |
3708 | ||
b21996e3 BF |
3709 | /** |
3710 | * vfs_unlink - unlink a filesystem object | |
3711 | * @dir: parent directory | |
3712 | * @dentry: victim | |
3713 | * @delegated_inode: returns victim inode, if the inode is delegated. | |
3714 | * | |
3715 | * The caller must hold dir->i_mutex. | |
3716 | * | |
3717 | * If vfs_unlink discovers a delegation, it will return -EWOULDBLOCK and | |
3718 | * return a reference to the inode in delegated_inode. The caller | |
3719 | * should then break the delegation on that inode and retry. Because | |
3720 | * breaking a delegation may take a long time, the caller should drop | |
3721 | * dir->i_mutex before doing so. | |
3722 | * | |
3723 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3724 | * be appropriate for callers that expect the underlying filesystem not | |
3725 | * to be NFS exported. | |
3726 | */ | |
3727 | int vfs_unlink(struct inode *dir, struct dentry *dentry, struct inode **delegated_inode) | |
1da177e4 | 3728 | { |
9accbb97 | 3729 | struct inode *target = dentry->d_inode; |
1da177e4 LT |
3730 | int error = may_delete(dir, dentry, 0); |
3731 | ||
3732 | if (error) | |
3733 | return error; | |
3734 | ||
acfa4380 | 3735 | if (!dir->i_op->unlink) |
1da177e4 LT |
3736 | return -EPERM; |
3737 | ||
9accbb97 | 3738 | mutex_lock(&target->i_mutex); |
8ed936b5 | 3739 | if (is_local_mountpoint(dentry)) |
1da177e4 LT |
3740 | error = -EBUSY; |
3741 | else { | |
3742 | error = security_inode_unlink(dir, dentry); | |
bec1052e | 3743 | if (!error) { |
5a14696c BF |
3744 | error = try_break_deleg(target, delegated_inode); |
3745 | if (error) | |
b21996e3 | 3746 | goto out; |
1da177e4 | 3747 | error = dir->i_op->unlink(dir, dentry); |
8ed936b5 | 3748 | if (!error) { |
d83c49f3 | 3749 | dont_mount(dentry); |
8ed936b5 EB |
3750 | detach_mounts(dentry); |
3751 | } | |
bec1052e | 3752 | } |
1da177e4 | 3753 | } |
b21996e3 | 3754 | out: |
9accbb97 | 3755 | mutex_unlock(&target->i_mutex); |
1da177e4 LT |
3756 | |
3757 | /* We don't d_delete() NFS sillyrenamed files--they still exist. */ | |
3758 | if (!error && !(dentry->d_flags & DCACHE_NFSFS_RENAMED)) { | |
9accbb97 | 3759 | fsnotify_link_count(target); |
e234f35c | 3760 | d_delete(dentry); |
1da177e4 | 3761 | } |
0eeca283 | 3762 | |
1da177e4 LT |
3763 | return error; |
3764 | } | |
4d359507 | 3765 | EXPORT_SYMBOL(vfs_unlink); |
1da177e4 LT |
3766 | |
3767 | /* | |
3768 | * Make sure that the actual truncation of the file will occur outside its | |
1b1dcc1b | 3769 | * directory's i_mutex. Truncate can take a long time if there is a lot of |
1da177e4 LT |
3770 | * writeout happening, and we don't want to prevent access to the directory |
3771 | * while waiting on the I/O. | |
3772 | */ | |
5590ff0d | 3773 | static long do_unlinkat(int dfd, const char __user *pathname) |
1da177e4 | 3774 | { |
2ad94ae6 | 3775 | int error; |
91a27b2a | 3776 | struct filename *name; |
1da177e4 | 3777 | struct dentry *dentry; |
f5beed75 AV |
3778 | struct path path; |
3779 | struct qstr last; | |
3780 | int type; | |
1da177e4 | 3781 | struct inode *inode = NULL; |
b21996e3 | 3782 | struct inode *delegated_inode = NULL; |
5d18f813 JL |
3783 | unsigned int lookup_flags = 0; |
3784 | retry: | |
f5beed75 AV |
3785 | name = user_path_parent(dfd, pathname, |
3786 | &path, &last, &type, lookup_flags); | |
91a27b2a JL |
3787 | if (IS_ERR(name)) |
3788 | return PTR_ERR(name); | |
2ad94ae6 | 3789 | |
1da177e4 | 3790 | error = -EISDIR; |
f5beed75 | 3791 | if (type != LAST_NORM) |
1da177e4 | 3792 | goto exit1; |
0612d9fb | 3793 | |
f5beed75 | 3794 | error = mnt_want_write(path.mnt); |
c30dabfe JK |
3795 | if (error) |
3796 | goto exit1; | |
b21996e3 | 3797 | retry_deleg: |
f5beed75 AV |
3798 | mutex_lock_nested(&path.dentry->d_inode->i_mutex, I_MUTEX_PARENT); |
3799 | dentry = __lookup_hash(&last, path.dentry, lookup_flags); | |
1da177e4 LT |
3800 | error = PTR_ERR(dentry); |
3801 | if (!IS_ERR(dentry)) { | |
3802 | /* Why not before? Because we want correct error value */ | |
f5beed75 | 3803 | if (last.name[last.len]) |
50338b88 | 3804 | goto slashes; |
1da177e4 | 3805 | inode = dentry->d_inode; |
b18825a7 | 3806 | if (d_is_negative(dentry)) |
e6bc45d6 TT |
3807 | goto slashes; |
3808 | ihold(inode); | |
f5beed75 | 3809 | error = security_path_unlink(&path, dentry); |
be6d3e56 | 3810 | if (error) |
c30dabfe | 3811 | goto exit2; |
f5beed75 | 3812 | error = vfs_unlink(path.dentry->d_inode, dentry, &delegated_inode); |
c30dabfe | 3813 | exit2: |
1da177e4 LT |
3814 | dput(dentry); |
3815 | } | |
f5beed75 | 3816 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
1da177e4 LT |
3817 | if (inode) |
3818 | iput(inode); /* truncate the inode here */ | |
b21996e3 BF |
3819 | inode = NULL; |
3820 | if (delegated_inode) { | |
5a14696c | 3821 | error = break_deleg_wait(&delegated_inode); |
b21996e3 BF |
3822 | if (!error) |
3823 | goto retry_deleg; | |
3824 | } | |
f5beed75 | 3825 | mnt_drop_write(path.mnt); |
1da177e4 | 3826 | exit1: |
f5beed75 | 3827 | path_put(&path); |
1da177e4 | 3828 | putname(name); |
5d18f813 JL |
3829 | if (retry_estale(error, lookup_flags)) { |
3830 | lookup_flags |= LOOKUP_REVAL; | |
3831 | inode = NULL; | |
3832 | goto retry; | |
3833 | } | |
1da177e4 LT |
3834 | return error; |
3835 | ||
3836 | slashes: | |
b18825a7 DH |
3837 | if (d_is_negative(dentry)) |
3838 | error = -ENOENT; | |
44b1d530 | 3839 | else if (d_is_dir(dentry)) |
b18825a7 DH |
3840 | error = -EISDIR; |
3841 | else | |
3842 | error = -ENOTDIR; | |
1da177e4 LT |
3843 | goto exit2; |
3844 | } | |
3845 | ||
2e4d0924 | 3846 | SYSCALL_DEFINE3(unlinkat, int, dfd, const char __user *, pathname, int, flag) |
5590ff0d UD |
3847 | { |
3848 | if ((flag & ~AT_REMOVEDIR) != 0) | |
3849 | return -EINVAL; | |
3850 | ||
3851 | if (flag & AT_REMOVEDIR) | |
3852 | return do_rmdir(dfd, pathname); | |
3853 | ||
3854 | return do_unlinkat(dfd, pathname); | |
3855 | } | |
3856 | ||
3480b257 | 3857 | SYSCALL_DEFINE1(unlink, const char __user *, pathname) |
5590ff0d UD |
3858 | { |
3859 | return do_unlinkat(AT_FDCWD, pathname); | |
3860 | } | |
3861 | ||
db2e747b | 3862 | int vfs_symlink(struct inode *dir, struct dentry *dentry, const char *oldname) |
1da177e4 | 3863 | { |
a95164d9 | 3864 | int error = may_create(dir, dentry); |
1da177e4 LT |
3865 | |
3866 | if (error) | |
3867 | return error; | |
3868 | ||
acfa4380 | 3869 | if (!dir->i_op->symlink) |
1da177e4 LT |
3870 | return -EPERM; |
3871 | ||
3872 | error = security_inode_symlink(dir, dentry, oldname); | |
3873 | if (error) | |
3874 | return error; | |
3875 | ||
1da177e4 | 3876 | error = dir->i_op->symlink(dir, dentry, oldname); |
a74574aa | 3877 | if (!error) |
f38aa942 | 3878 | fsnotify_create(dir, dentry); |
1da177e4 LT |
3879 | return error; |
3880 | } | |
4d359507 | 3881 | EXPORT_SYMBOL(vfs_symlink); |
1da177e4 | 3882 | |
2e4d0924 HC |
3883 | SYSCALL_DEFINE3(symlinkat, const char __user *, oldname, |
3884 | int, newdfd, const char __user *, newname) | |
1da177e4 | 3885 | { |
2ad94ae6 | 3886 | int error; |
91a27b2a | 3887 | struct filename *from; |
6902d925 | 3888 | struct dentry *dentry; |
dae6ad8f | 3889 | struct path path; |
f46d3567 | 3890 | unsigned int lookup_flags = 0; |
1da177e4 LT |
3891 | |
3892 | from = getname(oldname); | |
2ad94ae6 | 3893 | if (IS_ERR(from)) |
1da177e4 | 3894 | return PTR_ERR(from); |
f46d3567 JL |
3895 | retry: |
3896 | dentry = user_path_create(newdfd, newname, &path, lookup_flags); | |
6902d925 DH |
3897 | error = PTR_ERR(dentry); |
3898 | if (IS_ERR(dentry)) | |
dae6ad8f | 3899 | goto out_putname; |
6902d925 | 3900 | |
91a27b2a | 3901 | error = security_path_symlink(&path, dentry, from->name); |
a8104a9f | 3902 | if (!error) |
91a27b2a | 3903 | error = vfs_symlink(path.dentry->d_inode, dentry, from->name); |
921a1650 | 3904 | done_path_create(&path, dentry); |
f46d3567 JL |
3905 | if (retry_estale(error, lookup_flags)) { |
3906 | lookup_flags |= LOOKUP_REVAL; | |
3907 | goto retry; | |
3908 | } | |
6902d925 | 3909 | out_putname: |
1da177e4 LT |
3910 | putname(from); |
3911 | return error; | |
3912 | } | |
3913 | ||
3480b257 | 3914 | SYSCALL_DEFINE2(symlink, const char __user *, oldname, const char __user *, newname) |
5590ff0d UD |
3915 | { |
3916 | return sys_symlinkat(oldname, AT_FDCWD, newname); | |
3917 | } | |
3918 | ||
146a8595 BF |
3919 | /** |
3920 | * vfs_link - create a new link | |
3921 | * @old_dentry: object to be linked | |
3922 | * @dir: new parent | |
3923 | * @new_dentry: where to create the new link | |
3924 | * @delegated_inode: returns inode needing a delegation break | |
3925 | * | |
3926 | * The caller must hold dir->i_mutex | |
3927 | * | |
3928 | * If vfs_link discovers a delegation on the to-be-linked file in need | |
3929 | * of breaking, it will return -EWOULDBLOCK and return a reference to the | |
3930 | * inode in delegated_inode. The caller should then break the delegation | |
3931 | * and retry. Because breaking a delegation may take a long time, the | |
3932 | * caller should drop the i_mutex before doing so. | |
3933 | * | |
3934 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
3935 | * be appropriate for callers that expect the underlying filesystem not | |
3936 | * to be NFS exported. | |
3937 | */ | |
3938 | int vfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, struct inode **delegated_inode) | |
1da177e4 LT |
3939 | { |
3940 | struct inode *inode = old_dentry->d_inode; | |
8de52778 | 3941 | unsigned max_links = dir->i_sb->s_max_links; |
1da177e4 LT |
3942 | int error; |
3943 | ||
3944 | if (!inode) | |
3945 | return -ENOENT; | |
3946 | ||
a95164d9 | 3947 | error = may_create(dir, new_dentry); |
1da177e4 LT |
3948 | if (error) |
3949 | return error; | |
3950 | ||
3951 | if (dir->i_sb != inode->i_sb) | |
3952 | return -EXDEV; | |
3953 | ||
3954 | /* | |
3955 | * A link to an append-only or immutable file cannot be created. | |
3956 | */ | |
3957 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
3958 | return -EPERM; | |
acfa4380 | 3959 | if (!dir->i_op->link) |
1da177e4 | 3960 | return -EPERM; |
7e79eedb | 3961 | if (S_ISDIR(inode->i_mode)) |
1da177e4 LT |
3962 | return -EPERM; |
3963 | ||
3964 | error = security_inode_link(old_dentry, dir, new_dentry); | |
3965 | if (error) | |
3966 | return error; | |
3967 | ||
7e79eedb | 3968 | mutex_lock(&inode->i_mutex); |
aae8a97d | 3969 | /* Make sure we don't allow creating hardlink to an unlinked file */ |
f4e0c30c | 3970 | if (inode->i_nlink == 0 && !(inode->i_state & I_LINKABLE)) |
aae8a97d | 3971 | error = -ENOENT; |
8de52778 AV |
3972 | else if (max_links && inode->i_nlink >= max_links) |
3973 | error = -EMLINK; | |
146a8595 BF |
3974 | else { |
3975 | error = try_break_deleg(inode, delegated_inode); | |
3976 | if (!error) | |
3977 | error = dir->i_op->link(old_dentry, dir, new_dentry); | |
3978 | } | |
f4e0c30c AV |
3979 | |
3980 | if (!error && (inode->i_state & I_LINKABLE)) { | |
3981 | spin_lock(&inode->i_lock); | |
3982 | inode->i_state &= ~I_LINKABLE; | |
3983 | spin_unlock(&inode->i_lock); | |
3984 | } | |
7e79eedb | 3985 | mutex_unlock(&inode->i_mutex); |
e31e14ec | 3986 | if (!error) |
7e79eedb | 3987 | fsnotify_link(dir, inode, new_dentry); |
1da177e4 LT |
3988 | return error; |
3989 | } | |
4d359507 | 3990 | EXPORT_SYMBOL(vfs_link); |
1da177e4 LT |
3991 | |
3992 | /* | |
3993 | * Hardlinks are often used in delicate situations. We avoid | |
3994 | * security-related surprises by not following symlinks on the | |
3995 | * newname. --KAB | |
3996 | * | |
3997 | * We don't follow them on the oldname either to be compatible | |
3998 | * with linux 2.0, and to avoid hard-linking to directories | |
3999 | * and other special files. --ADM | |
4000 | */ | |
2e4d0924 HC |
4001 | SYSCALL_DEFINE5(linkat, int, olddfd, const char __user *, oldname, |
4002 | int, newdfd, const char __user *, newname, int, flags) | |
1da177e4 LT |
4003 | { |
4004 | struct dentry *new_dentry; | |
dae6ad8f | 4005 | struct path old_path, new_path; |
146a8595 | 4006 | struct inode *delegated_inode = NULL; |
11a7b371 | 4007 | int how = 0; |
1da177e4 | 4008 | int error; |
1da177e4 | 4009 | |
11a7b371 | 4010 | if ((flags & ~(AT_SYMLINK_FOLLOW | AT_EMPTY_PATH)) != 0) |
c04030e1 | 4011 | return -EINVAL; |
11a7b371 | 4012 | /* |
f0cc6ffb LT |
4013 | * To use null names we require CAP_DAC_READ_SEARCH |
4014 | * This ensures that not everyone will be able to create | |
4015 | * handlink using the passed filedescriptor. | |
11a7b371 | 4016 | */ |
f0cc6ffb LT |
4017 | if (flags & AT_EMPTY_PATH) { |
4018 | if (!capable(CAP_DAC_READ_SEARCH)) | |
4019 | return -ENOENT; | |
11a7b371 | 4020 | how = LOOKUP_EMPTY; |
f0cc6ffb | 4021 | } |
11a7b371 AK |
4022 | |
4023 | if (flags & AT_SYMLINK_FOLLOW) | |
4024 | how |= LOOKUP_FOLLOW; | |
442e31ca | 4025 | retry: |
11a7b371 | 4026 | error = user_path_at(olddfd, oldname, how, &old_path); |
1da177e4 | 4027 | if (error) |
2ad94ae6 AV |
4028 | return error; |
4029 | ||
442e31ca JL |
4030 | new_dentry = user_path_create(newdfd, newname, &new_path, |
4031 | (how & LOOKUP_REVAL)); | |
1da177e4 | 4032 | error = PTR_ERR(new_dentry); |
6902d925 | 4033 | if (IS_ERR(new_dentry)) |
dae6ad8f AV |
4034 | goto out; |
4035 | ||
4036 | error = -EXDEV; | |
4037 | if (old_path.mnt != new_path.mnt) | |
4038 | goto out_dput; | |
800179c9 KC |
4039 | error = may_linkat(&old_path); |
4040 | if (unlikely(error)) | |
4041 | goto out_dput; | |
dae6ad8f | 4042 | error = security_path_link(old_path.dentry, &new_path, new_dentry); |
be6d3e56 | 4043 | if (error) |
a8104a9f | 4044 | goto out_dput; |
146a8595 | 4045 | error = vfs_link(old_path.dentry, new_path.dentry->d_inode, new_dentry, &delegated_inode); |
75c3f29d | 4046 | out_dput: |
921a1650 | 4047 | done_path_create(&new_path, new_dentry); |
146a8595 BF |
4048 | if (delegated_inode) { |
4049 | error = break_deleg_wait(&delegated_inode); | |
d22e6338 OD |
4050 | if (!error) { |
4051 | path_put(&old_path); | |
146a8595 | 4052 | goto retry; |
d22e6338 | 4053 | } |
146a8595 | 4054 | } |
442e31ca | 4055 | if (retry_estale(error, how)) { |
d22e6338 | 4056 | path_put(&old_path); |
442e31ca JL |
4057 | how |= LOOKUP_REVAL; |
4058 | goto retry; | |
4059 | } | |
1da177e4 | 4060 | out: |
2d8f3038 | 4061 | path_put(&old_path); |
1da177e4 LT |
4062 | |
4063 | return error; | |
4064 | } | |
4065 | ||
3480b257 | 4066 | SYSCALL_DEFINE2(link, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4067 | { |
c04030e1 | 4068 | return sys_linkat(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4069 | } |
4070 | ||
bc27027a MS |
4071 | /** |
4072 | * vfs_rename - rename a filesystem object | |
4073 | * @old_dir: parent of source | |
4074 | * @old_dentry: source | |
4075 | * @new_dir: parent of destination | |
4076 | * @new_dentry: destination | |
4077 | * @delegated_inode: returns an inode needing a delegation break | |
520c8b16 | 4078 | * @flags: rename flags |
bc27027a MS |
4079 | * |
4080 | * The caller must hold multiple mutexes--see lock_rename()). | |
4081 | * | |
4082 | * If vfs_rename discovers a delegation in need of breaking at either | |
4083 | * the source or destination, it will return -EWOULDBLOCK and return a | |
4084 | * reference to the inode in delegated_inode. The caller should then | |
4085 | * break the delegation and retry. Because breaking a delegation may | |
4086 | * take a long time, the caller should drop all locks before doing | |
4087 | * so. | |
4088 | * | |
4089 | * Alternatively, a caller may pass NULL for delegated_inode. This may | |
4090 | * be appropriate for callers that expect the underlying filesystem not | |
4091 | * to be NFS exported. | |
4092 | * | |
1da177e4 LT |
4093 | * The worst of all namespace operations - renaming directory. "Perverted" |
4094 | * doesn't even start to describe it. Somebody in UCB had a heck of a trip... | |
4095 | * Problems: | |
d03b29a2 | 4096 | * a) we can get into loop creation. |
1da177e4 LT |
4097 | * b) race potential - two innocent renames can create a loop together. |
4098 | * That's where 4.4 screws up. Current fix: serialization on | |
a11f3a05 | 4099 | * sb->s_vfs_rename_mutex. We might be more accurate, but that's another |
1da177e4 | 4100 | * story. |
6cedba89 BF |
4101 | * c) we have to lock _four_ objects - parents and victim (if it exists), |
4102 | * and source (if it is not a directory). | |
1b1dcc1b | 4103 | * And that - after we got ->i_mutex on parents (until then we don't know |
1da177e4 LT |
4104 | * whether the target exists). Solution: try to be smart with locking |
4105 | * order for inodes. We rely on the fact that tree topology may change | |
a11f3a05 | 4106 | * only under ->s_vfs_rename_mutex _and_ that parent of the object we |
1da177e4 LT |
4107 | * move will be locked. Thus we can rank directories by the tree |
4108 | * (ancestors first) and rank all non-directories after them. | |
4109 | * That works since everybody except rename does "lock parent, lookup, | |
a11f3a05 | 4110 | * lock child" and rename is under ->s_vfs_rename_mutex. |
1da177e4 LT |
4111 | * HOWEVER, it relies on the assumption that any object with ->lookup() |
4112 | * has no more than 1 dentry. If "hybrid" objects will ever appear, | |
4113 | * we'd better make sure that there's no link(2) for them. | |
e4eaac06 | 4114 | * d) conversion from fhandle to dentry may come in the wrong moment - when |
1b1dcc1b | 4115 | * we are removing the target. Solution: we will have to grab ->i_mutex |
1da177e4 | 4116 | * in the fhandle_to_dentry code. [FIXME - current nfsfh.c relies on |
c41b20e7 | 4117 | * ->i_mutex on parents, which works but leads to some truly excessive |
1da177e4 LT |
4118 | * locking]. |
4119 | */ | |
bc27027a MS |
4120 | int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, |
4121 | struct inode *new_dir, struct dentry *new_dentry, | |
520c8b16 | 4122 | struct inode **delegated_inode, unsigned int flags) |
1da177e4 | 4123 | { |
bc27027a MS |
4124 | int error; |
4125 | bool is_dir = d_is_dir(old_dentry); | |
4126 | const unsigned char *old_name; | |
4127 | struct inode *source = old_dentry->d_inode; | |
9055cba7 | 4128 | struct inode *target = new_dentry->d_inode; |
da1ce067 MS |
4129 | bool new_is_dir = false; |
4130 | unsigned max_links = new_dir->i_sb->s_max_links; | |
bc27027a MS |
4131 | |
4132 | if (source == target) | |
4133 | return 0; | |
4134 | ||
4135 | error = may_delete(old_dir, old_dentry, is_dir); | |
4136 | if (error) | |
4137 | return error; | |
4138 | ||
da1ce067 | 4139 | if (!target) { |
bc27027a | 4140 | error = may_create(new_dir, new_dentry); |
da1ce067 MS |
4141 | } else { |
4142 | new_is_dir = d_is_dir(new_dentry); | |
4143 | ||
4144 | if (!(flags & RENAME_EXCHANGE)) | |
4145 | error = may_delete(new_dir, new_dentry, is_dir); | |
4146 | else | |
4147 | error = may_delete(new_dir, new_dentry, new_is_dir); | |
4148 | } | |
bc27027a MS |
4149 | if (error) |
4150 | return error; | |
4151 | ||
7177a9c4 | 4152 | if (!old_dir->i_op->rename && !old_dir->i_op->rename2) |
bc27027a | 4153 | return -EPERM; |
1da177e4 | 4154 | |
520c8b16 MS |
4155 | if (flags && !old_dir->i_op->rename2) |
4156 | return -EINVAL; | |
4157 | ||
1da177e4 LT |
4158 | /* |
4159 | * If we are going to change the parent - check write permissions, | |
4160 | * we'll need to flip '..'. | |
4161 | */ | |
da1ce067 MS |
4162 | if (new_dir != old_dir) { |
4163 | if (is_dir) { | |
4164 | error = inode_permission(source, MAY_WRITE); | |
4165 | if (error) | |
4166 | return error; | |
4167 | } | |
4168 | if ((flags & RENAME_EXCHANGE) && new_is_dir) { | |
4169 | error = inode_permission(target, MAY_WRITE); | |
4170 | if (error) | |
4171 | return error; | |
4172 | } | |
1da177e4 LT |
4173 | } |
4174 | ||
0b3974eb MS |
4175 | error = security_inode_rename(old_dir, old_dentry, new_dir, new_dentry, |
4176 | flags); | |
1da177e4 LT |
4177 | if (error) |
4178 | return error; | |
4179 | ||
bc27027a | 4180 | old_name = fsnotify_oldname_init(old_dentry->d_name.name); |
1d2ef590 | 4181 | dget(new_dentry); |
da1ce067 | 4182 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4183 | lock_two_nondirectories(source, target); |
4184 | else if (target) | |
1b1dcc1b | 4185 | mutex_lock(&target->i_mutex); |
9055cba7 SW |
4186 | |
4187 | error = -EBUSY; | |
7af1364f | 4188 | if (is_local_mountpoint(old_dentry) || is_local_mountpoint(new_dentry)) |
9055cba7 SW |
4189 | goto out; |
4190 | ||
da1ce067 | 4191 | if (max_links && new_dir != old_dir) { |
bc27027a | 4192 | error = -EMLINK; |
da1ce067 | 4193 | if (is_dir && !new_is_dir && new_dir->i_nlink >= max_links) |
bc27027a | 4194 | goto out; |
da1ce067 MS |
4195 | if ((flags & RENAME_EXCHANGE) && !is_dir && new_is_dir && |
4196 | old_dir->i_nlink >= max_links) | |
4197 | goto out; | |
4198 | } | |
4199 | if (is_dir && !(flags & RENAME_EXCHANGE) && target) | |
4200 | shrink_dcache_parent(new_dentry); | |
4201 | if (!is_dir) { | |
bc27027a | 4202 | error = try_break_deleg(source, delegated_inode); |
8e6d782c BF |
4203 | if (error) |
4204 | goto out; | |
da1ce067 MS |
4205 | } |
4206 | if (target && !new_is_dir) { | |
4207 | error = try_break_deleg(target, delegated_inode); | |
4208 | if (error) | |
4209 | goto out; | |
8e6d782c | 4210 | } |
7177a9c4 | 4211 | if (!old_dir->i_op->rename2) { |
520c8b16 MS |
4212 | error = old_dir->i_op->rename(old_dir, old_dentry, |
4213 | new_dir, new_dentry); | |
4214 | } else { | |
7177a9c4 | 4215 | WARN_ON(old_dir->i_op->rename != NULL); |
520c8b16 MS |
4216 | error = old_dir->i_op->rename2(old_dir, old_dentry, |
4217 | new_dir, new_dentry, flags); | |
4218 | } | |
51892bbb SW |
4219 | if (error) |
4220 | goto out; | |
4221 | ||
da1ce067 | 4222 | if (!(flags & RENAME_EXCHANGE) && target) { |
bc27027a MS |
4223 | if (is_dir) |
4224 | target->i_flags |= S_DEAD; | |
51892bbb | 4225 | dont_mount(new_dentry); |
8ed936b5 | 4226 | detach_mounts(new_dentry); |
bc27027a | 4227 | } |
da1ce067 MS |
4228 | if (!(old_dir->i_sb->s_type->fs_flags & FS_RENAME_DOES_D_MOVE)) { |
4229 | if (!(flags & RENAME_EXCHANGE)) | |
4230 | d_move(old_dentry, new_dentry); | |
4231 | else | |
4232 | d_exchange(old_dentry, new_dentry); | |
4233 | } | |
51892bbb | 4234 | out: |
da1ce067 | 4235 | if (!is_dir || (flags & RENAME_EXCHANGE)) |
bc27027a MS |
4236 | unlock_two_nondirectories(source, target); |
4237 | else if (target) | |
4238 | mutex_unlock(&target->i_mutex); | |
1da177e4 | 4239 | dput(new_dentry); |
da1ce067 | 4240 | if (!error) { |
123df294 | 4241 | fsnotify_move(old_dir, new_dir, old_name, is_dir, |
da1ce067 MS |
4242 | !(flags & RENAME_EXCHANGE) ? target : NULL, old_dentry); |
4243 | if (flags & RENAME_EXCHANGE) { | |
4244 | fsnotify_move(new_dir, old_dir, old_dentry->d_name.name, | |
4245 | new_is_dir, NULL, new_dentry); | |
4246 | } | |
4247 | } | |
0eeca283 RL |
4248 | fsnotify_oldname_free(old_name); |
4249 | ||
1da177e4 LT |
4250 | return error; |
4251 | } | |
4d359507 | 4252 | EXPORT_SYMBOL(vfs_rename); |
1da177e4 | 4253 | |
520c8b16 MS |
4254 | SYSCALL_DEFINE5(renameat2, int, olddfd, const char __user *, oldname, |
4255 | int, newdfd, const char __user *, newname, unsigned int, flags) | |
1da177e4 | 4256 | { |
2ad94ae6 AV |
4257 | struct dentry *old_dentry, *new_dentry; |
4258 | struct dentry *trap; | |
f5beed75 AV |
4259 | struct path old_path, new_path; |
4260 | struct qstr old_last, new_last; | |
4261 | int old_type, new_type; | |
8e6d782c | 4262 | struct inode *delegated_inode = NULL; |
91a27b2a JL |
4263 | struct filename *from; |
4264 | struct filename *to; | |
f5beed75 | 4265 | unsigned int lookup_flags = 0, target_flags = LOOKUP_RENAME_TARGET; |
c6a94284 | 4266 | bool should_retry = false; |
2ad94ae6 | 4267 | int error; |
520c8b16 | 4268 | |
0d7a8555 | 4269 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
da1ce067 MS |
4270 | return -EINVAL; |
4271 | ||
0d7a8555 MS |
4272 | if ((flags & (RENAME_NOREPLACE | RENAME_WHITEOUT)) && |
4273 | (flags & RENAME_EXCHANGE)) | |
520c8b16 MS |
4274 | return -EINVAL; |
4275 | ||
0d7a8555 MS |
4276 | if ((flags & RENAME_WHITEOUT) && !capable(CAP_MKNOD)) |
4277 | return -EPERM; | |
4278 | ||
f5beed75 AV |
4279 | if (flags & RENAME_EXCHANGE) |
4280 | target_flags = 0; | |
4281 | ||
c6a94284 | 4282 | retry: |
f5beed75 AV |
4283 | from = user_path_parent(olddfd, oldname, |
4284 | &old_path, &old_last, &old_type, lookup_flags); | |
91a27b2a JL |
4285 | if (IS_ERR(from)) { |
4286 | error = PTR_ERR(from); | |
1da177e4 | 4287 | goto exit; |
91a27b2a | 4288 | } |
1da177e4 | 4289 | |
f5beed75 AV |
4290 | to = user_path_parent(newdfd, newname, |
4291 | &new_path, &new_last, &new_type, lookup_flags); | |
91a27b2a JL |
4292 | if (IS_ERR(to)) { |
4293 | error = PTR_ERR(to); | |
1da177e4 | 4294 | goto exit1; |
91a27b2a | 4295 | } |
1da177e4 LT |
4296 | |
4297 | error = -EXDEV; | |
f5beed75 | 4298 | if (old_path.mnt != new_path.mnt) |
1da177e4 LT |
4299 | goto exit2; |
4300 | ||
1da177e4 | 4301 | error = -EBUSY; |
f5beed75 | 4302 | if (old_type != LAST_NORM) |
1da177e4 LT |
4303 | goto exit2; |
4304 | ||
0a7c3937 MS |
4305 | if (flags & RENAME_NOREPLACE) |
4306 | error = -EEXIST; | |
f5beed75 | 4307 | if (new_type != LAST_NORM) |
1da177e4 LT |
4308 | goto exit2; |
4309 | ||
f5beed75 | 4310 | error = mnt_want_write(old_path.mnt); |
c30dabfe JK |
4311 | if (error) |
4312 | goto exit2; | |
4313 | ||
8e6d782c | 4314 | retry_deleg: |
f5beed75 | 4315 | trap = lock_rename(new_path.dentry, old_path.dentry); |
1da177e4 | 4316 | |
f5beed75 | 4317 | old_dentry = __lookup_hash(&old_last, old_path.dentry, lookup_flags); |
1da177e4 LT |
4318 | error = PTR_ERR(old_dentry); |
4319 | if (IS_ERR(old_dentry)) | |
4320 | goto exit3; | |
4321 | /* source must exist */ | |
4322 | error = -ENOENT; | |
b18825a7 | 4323 | if (d_is_negative(old_dentry)) |
1da177e4 | 4324 | goto exit4; |
f5beed75 | 4325 | new_dentry = __lookup_hash(&new_last, new_path.dentry, lookup_flags | target_flags); |
0a7c3937 MS |
4326 | error = PTR_ERR(new_dentry); |
4327 | if (IS_ERR(new_dentry)) | |
4328 | goto exit4; | |
4329 | error = -EEXIST; | |
4330 | if ((flags & RENAME_NOREPLACE) && d_is_positive(new_dentry)) | |
4331 | goto exit5; | |
da1ce067 MS |
4332 | if (flags & RENAME_EXCHANGE) { |
4333 | error = -ENOENT; | |
4334 | if (d_is_negative(new_dentry)) | |
4335 | goto exit5; | |
4336 | ||
4337 | if (!d_is_dir(new_dentry)) { | |
4338 | error = -ENOTDIR; | |
f5beed75 | 4339 | if (new_last.name[new_last.len]) |
da1ce067 MS |
4340 | goto exit5; |
4341 | } | |
4342 | } | |
1da177e4 | 4343 | /* unless the source is a directory trailing slashes give -ENOTDIR */ |
44b1d530 | 4344 | if (!d_is_dir(old_dentry)) { |
1da177e4 | 4345 | error = -ENOTDIR; |
f5beed75 | 4346 | if (old_last.name[old_last.len]) |
0a7c3937 | 4347 | goto exit5; |
f5beed75 | 4348 | if (!(flags & RENAME_EXCHANGE) && new_last.name[new_last.len]) |
0a7c3937 | 4349 | goto exit5; |
1da177e4 LT |
4350 | } |
4351 | /* source should not be ancestor of target */ | |
4352 | error = -EINVAL; | |
4353 | if (old_dentry == trap) | |
0a7c3937 | 4354 | goto exit5; |
1da177e4 | 4355 | /* target should not be an ancestor of source */ |
da1ce067 MS |
4356 | if (!(flags & RENAME_EXCHANGE)) |
4357 | error = -ENOTEMPTY; | |
1da177e4 LT |
4358 | if (new_dentry == trap) |
4359 | goto exit5; | |
4360 | ||
f5beed75 AV |
4361 | error = security_path_rename(&old_path, old_dentry, |
4362 | &new_path, new_dentry, flags); | |
be6d3e56 | 4363 | if (error) |
c30dabfe | 4364 | goto exit5; |
f5beed75 AV |
4365 | error = vfs_rename(old_path.dentry->d_inode, old_dentry, |
4366 | new_path.dentry->d_inode, new_dentry, | |
520c8b16 | 4367 | &delegated_inode, flags); |
1da177e4 LT |
4368 | exit5: |
4369 | dput(new_dentry); | |
4370 | exit4: | |
4371 | dput(old_dentry); | |
4372 | exit3: | |
f5beed75 | 4373 | unlock_rename(new_path.dentry, old_path.dentry); |
8e6d782c BF |
4374 | if (delegated_inode) { |
4375 | error = break_deleg_wait(&delegated_inode); | |
4376 | if (!error) | |
4377 | goto retry_deleg; | |
4378 | } | |
f5beed75 | 4379 | mnt_drop_write(old_path.mnt); |
1da177e4 | 4380 | exit2: |
c6a94284 JL |
4381 | if (retry_estale(error, lookup_flags)) |
4382 | should_retry = true; | |
f5beed75 | 4383 | path_put(&new_path); |
2ad94ae6 | 4384 | putname(to); |
1da177e4 | 4385 | exit1: |
f5beed75 | 4386 | path_put(&old_path); |
1da177e4 | 4387 | putname(from); |
c6a94284 JL |
4388 | if (should_retry) { |
4389 | should_retry = false; | |
4390 | lookup_flags |= LOOKUP_REVAL; | |
4391 | goto retry; | |
4392 | } | |
2ad94ae6 | 4393 | exit: |
1da177e4 LT |
4394 | return error; |
4395 | } | |
4396 | ||
520c8b16 MS |
4397 | SYSCALL_DEFINE4(renameat, int, olddfd, const char __user *, oldname, |
4398 | int, newdfd, const char __user *, newname) | |
4399 | { | |
4400 | return sys_renameat2(olddfd, oldname, newdfd, newname, 0); | |
4401 | } | |
4402 | ||
a26eab24 | 4403 | SYSCALL_DEFINE2(rename, const char __user *, oldname, const char __user *, newname) |
5590ff0d | 4404 | { |
520c8b16 | 4405 | return sys_renameat2(AT_FDCWD, oldname, AT_FDCWD, newname, 0); |
5590ff0d UD |
4406 | } |
4407 | ||
787fb6bc MS |
4408 | int vfs_whiteout(struct inode *dir, struct dentry *dentry) |
4409 | { | |
4410 | int error = may_create(dir, dentry); | |
4411 | if (error) | |
4412 | return error; | |
4413 | ||
4414 | if (!dir->i_op->mknod) | |
4415 | return -EPERM; | |
4416 | ||
4417 | return dir->i_op->mknod(dir, dentry, | |
4418 | S_IFCHR | WHITEOUT_MODE, WHITEOUT_DEV); | |
4419 | } | |
4420 | EXPORT_SYMBOL(vfs_whiteout); | |
4421 | ||
5d826c84 | 4422 | int readlink_copy(char __user *buffer, int buflen, const char *link) |
1da177e4 | 4423 | { |
5d826c84 | 4424 | int len = PTR_ERR(link); |
1da177e4 LT |
4425 | if (IS_ERR(link)) |
4426 | goto out; | |
4427 | ||
4428 | len = strlen(link); | |
4429 | if (len > (unsigned) buflen) | |
4430 | len = buflen; | |
4431 | if (copy_to_user(buffer, link, len)) | |
4432 | len = -EFAULT; | |
4433 | out: | |
4434 | return len; | |
4435 | } | |
5d826c84 | 4436 | EXPORT_SYMBOL(readlink_copy); |
1da177e4 LT |
4437 | |
4438 | /* | |
4439 | * A helper for ->readlink(). This should be used *ONLY* for symlinks that | |
4440 | * have ->follow_link() touching nd only in nd_set_link(). Using (or not | |
4441 | * using) it for any given inode is up to filesystem. | |
4442 | */ | |
4443 | int generic_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4444 | { | |
cc314eef | 4445 | void *cookie; |
d4dee48b | 4446 | const char *link = dentry->d_inode->i_link; |
694a1764 | 4447 | int res; |
cc314eef | 4448 | |
d4dee48b | 4449 | if (!link) { |
6e77137b | 4450 | link = dentry->d_inode->i_op->follow_link(dentry, &cookie); |
d4dee48b AV |
4451 | if (IS_ERR(link)) |
4452 | return PTR_ERR(link); | |
4453 | } | |
680baacb | 4454 | res = readlink_copy(buffer, buflen, link); |
6e77137b | 4455 | if (dentry->d_inode->i_op->put_link) |
680baacb | 4456 | dentry->d_inode->i_op->put_link(dentry, cookie); |
694a1764 | 4457 | return res; |
1da177e4 | 4458 | } |
4d359507 | 4459 | EXPORT_SYMBOL(generic_readlink); |
1da177e4 | 4460 | |
1da177e4 LT |
4461 | /* get the link contents into pagecache */ |
4462 | static char *page_getlink(struct dentry * dentry, struct page **ppage) | |
4463 | { | |
ebd09abb DG |
4464 | char *kaddr; |
4465 | struct page *page; | |
1da177e4 | 4466 | struct address_space *mapping = dentry->d_inode->i_mapping; |
090d2b18 | 4467 | page = read_mapping_page(mapping, 0, NULL); |
1da177e4 | 4468 | if (IS_ERR(page)) |
6fe6900e | 4469 | return (char*)page; |
1da177e4 | 4470 | *ppage = page; |
ebd09abb DG |
4471 | kaddr = kmap(page); |
4472 | nd_terminate_link(kaddr, dentry->d_inode->i_size, PAGE_SIZE - 1); | |
4473 | return kaddr; | |
1da177e4 LT |
4474 | } |
4475 | ||
4476 | int page_readlink(struct dentry *dentry, char __user *buffer, int buflen) | |
4477 | { | |
4478 | struct page *page = NULL; | |
5d826c84 | 4479 | int res = readlink_copy(buffer, buflen, page_getlink(dentry, &page)); |
1da177e4 LT |
4480 | if (page) { |
4481 | kunmap(page); | |
4482 | page_cache_release(page); | |
4483 | } | |
4484 | return res; | |
4485 | } | |
4d359507 | 4486 | EXPORT_SYMBOL(page_readlink); |
1da177e4 | 4487 | |
6e77137b | 4488 | const char *page_follow_link_light(struct dentry *dentry, void **cookie) |
1da177e4 | 4489 | { |
cc314eef | 4490 | struct page *page = NULL; |
680baacb AV |
4491 | char *res = page_getlink(dentry, &page); |
4492 | if (!IS_ERR(res)) | |
4493 | *cookie = page; | |
4494 | return res; | |
1da177e4 | 4495 | } |
4d359507 | 4496 | EXPORT_SYMBOL(page_follow_link_light); |
1da177e4 | 4497 | |
680baacb | 4498 | void page_put_link(struct dentry *dentry, void *cookie) |
1da177e4 | 4499 | { |
cc314eef | 4500 | struct page *page = cookie; |
680baacb AV |
4501 | kunmap(page); |
4502 | page_cache_release(page); | |
1da177e4 | 4503 | } |
4d359507 | 4504 | EXPORT_SYMBOL(page_put_link); |
1da177e4 | 4505 | |
54566b2c NP |
4506 | /* |
4507 | * The nofs argument instructs pagecache_write_begin to pass AOP_FLAG_NOFS | |
4508 | */ | |
4509 | int __page_symlink(struct inode *inode, const char *symname, int len, int nofs) | |
1da177e4 LT |
4510 | { |
4511 | struct address_space *mapping = inode->i_mapping; | |
0adb25d2 | 4512 | struct page *page; |
afddba49 | 4513 | void *fsdata; |
beb497ab | 4514 | int err; |
1da177e4 | 4515 | char *kaddr; |
54566b2c NP |
4516 | unsigned int flags = AOP_FLAG_UNINTERRUPTIBLE; |
4517 | if (nofs) | |
4518 | flags |= AOP_FLAG_NOFS; | |
1da177e4 | 4519 | |
7e53cac4 | 4520 | retry: |
afddba49 | 4521 | err = pagecache_write_begin(NULL, mapping, 0, len-1, |
54566b2c | 4522 | flags, &page, &fsdata); |
1da177e4 | 4523 | if (err) |
afddba49 NP |
4524 | goto fail; |
4525 | ||
e8e3c3d6 | 4526 | kaddr = kmap_atomic(page); |
1da177e4 | 4527 | memcpy(kaddr, symname, len-1); |
e8e3c3d6 | 4528 | kunmap_atomic(kaddr); |
afddba49 NP |
4529 | |
4530 | err = pagecache_write_end(NULL, mapping, 0, len-1, len-1, | |
4531 | page, fsdata); | |
1da177e4 LT |
4532 | if (err < 0) |
4533 | goto fail; | |
afddba49 NP |
4534 | if (err < len-1) |
4535 | goto retry; | |
4536 | ||
1da177e4 LT |
4537 | mark_inode_dirty(inode); |
4538 | return 0; | |
1da177e4 LT |
4539 | fail: |
4540 | return err; | |
4541 | } | |
4d359507 | 4542 | EXPORT_SYMBOL(__page_symlink); |
1da177e4 | 4543 | |
0adb25d2 KK |
4544 | int page_symlink(struct inode *inode, const char *symname, int len) |
4545 | { | |
4546 | return __page_symlink(inode, symname, len, | |
54566b2c | 4547 | !(mapping_gfp_mask(inode->i_mapping) & __GFP_FS)); |
0adb25d2 | 4548 | } |
4d359507 | 4549 | EXPORT_SYMBOL(page_symlink); |
0adb25d2 | 4550 | |
92e1d5be | 4551 | const struct inode_operations page_symlink_inode_operations = { |
1da177e4 LT |
4552 | .readlink = generic_readlink, |
4553 | .follow_link = page_follow_link_light, | |
4554 | .put_link = page_put_link, | |
4555 | }; | |
1da177e4 | 4556 | EXPORT_SYMBOL(page_symlink_inode_operations); |